Unraveling the genetic expression of the highly variable immune receptors of a killer Vendelbosch, S.

Transcription

1 UvA-DARE (Digital Academic Repository) Unraveling the genetic expression of the highly variable immune receptors of a killer Vendelbosch, S. Link to publication Citation for published version (APA): Vendelbosch, S. (2015). Unraveling the genetic expression of the highly variable immune receptors of a killer General rights It is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), other than for strictly personal, individual use, unless the work is under an open content license (like Creative Commons). Disclaimer/Complaints regulations If you believe that digital publication of certain material infringes any of your rights or (privacy) interests, please let the Library know, stating your reasons. In case of a legitimate complaint, the Library will make the material inaccessible and/or remove it from the website. Please Ask the Library: or a letter to: Library of the University of Amsterdam, Secretariat, Singel 425, 1012 WP Amsterdam, The Netherlands. You will be contacted as soon as possible. UvA-DARE is a service provided by the library of the University of Amsterdam ( Download date: 27 Dec 2018

2

3 UNRAVELING THE GENETIC EXPRESSION OF THE HIGHLY VARIABLE IMMUNE RECEPTORS OF A KILLER Sanne Vendelbosch

4 Sanne Vendelbosch, The Netherlands, All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopy, recording or any information storage and retrieval system, without written permission of the author. Cover design: Taco Vendelbosch Printed by: Off Page ISBN: Printing of this thesis was financially supported by Sanquin Blood Supply Foundation, Amsterdam, The Netherlands.

5 UNRAVELING THE GENETIC EXPRESSION OF THE HIGHLY VARIABLE IMMUNE RECEPTORS OF A KILLER ACADEMISCH PROEFSCHIFT ter verkrijging van de graad van doctor aan de Universiteit van Amsterdam op gezag van de Rector Magnificus prof. dr. D.C. van den Boom ten overstaan van een door het College voor Promoties ingestelde commissie, in het openbaar te verdedigen in de Agnietenkapel op woensdag 23 september 2015, te 12:00 uur door Sanne Vendelbosch geboren te Vuren

6 PROMOTIECOMMISSIE Promotor: prof. dr. T.W. Kuijpers Universiteit van Amsterdam Copromotor: dr. T.K. van den Berg Universiteit van Amsterdam Overige leden: prof. dr. R.J.M. ten Berge Universiteit van Amsterdam prof. dr. F.H.J. Claas Universiteit Leiden prof. dr. M.G.J. Tilanus Universiteit van Maastricht prof. dr. H. Spits Universiteit van Amsterdam prof. dr. D.L.P. Baeten Universiteit van Amsterdam Faculteit der Geneeskunde

7 TABLE OF CONTENTS Chapter 1 General Introduction 7 Chapter 2 Extensive variation in gene copy number at the killer 21 immunoglobulin-like receptor locus in humans Chapter 3 Novel insights in the genomic organization and hot-spots 53 of recombination in the human KIR locus through analysis of intergenic regions Chapter 4 Interleukin (IL)-15 contributes to variegated expression 75 of Killer Immunoglobulin-like Receptors on human adult and neonatal natural killer cells Chapter 5 KIR3DL1 and KIR3DL2 gene copy number variation in axial 97 spondyloarthritis Chapter 6 Study on the protective effect of the KIR3DL1 gene 103 in Ankylosing Spondylitis Chapter 7 Summary and General Discussion 119 Addendum References 129 Nederlandse samenvatting 141 Portfolio 149 Dankwoord 155

8

9 GENERAL INTRODUCTION 1

10

11 GENERAL INTRODUCTION The immune system is a complex network that protects the human body against internal and external threats. It consists of an innate and an adaptive system. Natural killer (NK) cells constitute the first line of innate immunity against viral infections and tumor development, but also have the ability to mount adaptive immune responses. Cytotoxicity of NK cells is mainly regulated by immune receptors on the cell membrane of which the Killer Immunoglobulin-like Receptors (KIRs) are one of the most important families. The KIR family consists of multiple activating and inhibiting genes that are distributed unequally across the population due to gene presence and gene copy number variation (CNV). Each KIR has its own ligand specificity creating a specific reaction within an NK cell, as will be discussed in more detail below. Since every individual has its own KIR gene repertoire, the NK cells of every individual could potentially react in a unique manner against external threats. We know that individuals can indeed react differently to viral infections or tumor cell manifestations. For example, some people can cope better when infected with human immunodeficiency virus type 1 (HIV-1) or hepatitis B virus (HBV) than others. Also, almost every person will generate tumor cells at least once in a life time, but some immune systems kill these cells before a proper malignancy becomes manifest, while tumor development in another susceptible genetic background may generate a serious threat. Notably these are all examples of diseases to which NK cells form a critical part of the host defense, therefore the variable KIR family might play a major role in the ability of NK cells to cope with these situations. To test this hypothesis, we set out to study KIR gene variation, expression variation and their contribution to disease. In this introductory chapter I will first give some more background on NK cells and their functions, discuss what is known about the involvement of KIR and then provide a synopsis of the scope of the studies described in this thesis. 1 NATURAL KILLER CELLS Natural killer cell homeostasis Natural killer (NK) cells are granular lymphocytes critical for innate immunity, in particular with respect to the killing of tumor cells and elimination of virus-infected host cells. They comprise 10-15% of all lymphocytes in blood. NK cells primarily develop from a common lymphoid progenitor (CLP) in the bone marrow, although some reports suggest that these cells can also develop postnatally in lymph nodes and the liver (Andrews and Smyth 2009; Freud et al. 2005). They resemble to some extent the most important lymphocytes within the immune system, i.e. the T cells, with a major difference that T cells develop in the thymus and in contrast to the antigen-specific T-cell receptor (TCR) the immune receptors of NK cells are non-rearranging (Sun and Lanier 2011). As a result, the development of NK cells, unlike that of T cells, is independent of the recombinase activating genes (RAG). T cells are phenotypically marked by the expression of CD3, which is the signal-transducing component of the TCR. Mature NK cells are phenotypically distinguishable by the expression of CD56 and the lack of CD3 expression. Although the NK cell maturation process is still not completely understood and research defining the different developmental stages is still ongoing, certain events during NK cell 9

12 GENERAL INTRODUCTION 1 maturation are generally accepted. Maturation of NK cells can be monitored by the expression of certain NK cell markers and cytotoxicity receptors. NK cell cytotoxicity is primarily mediated by NK cell receptors (NKR) of which the three major families are the KIRs, the C-type lectins such as NKG2A and NKG2C, which form heterodimers with CD94, and NKG2D, and the natural cytotoxicity receptors (NCR) including NKp33, NKp44 and NKp46 (Farag and Caligiuri 2006; Pegram et al. 2011). Other activating receptors are CD244 (2B4) and CD16 (FcγRIIIa), while DNAM-1 and NKR-P1 are co-stimulatory receptors (Pegram et al. 2011). Another group of inhibitory receptors expressed on NK cells are the LILR (leukocyte immunoglobulin-like receptor), although the function of these receptors remains poorly understood (Brown et al. 2004; Pegram et al. 2011). The main NK cell population is marked by expression of a neural cell adhesion molecule, CD56, and constitutes of CD56bright and CD56dim cells. The CD56bright cells are CD16dim/neg, NKG2A+ and KIR- and are thought to be early NK precursors of the more abundant and mature CD56dim cells, which are CD16bright and express KIRs (Björkström et al. 2010; Farag and Caligiuri 2006). Also, it has been suggested that variation in expression levels of CD27 and CD57 mark the differentiation stages of NK cells (Björkström et al. 2010; Vossen et al. 2008). CD56bright cells differ in NKR expression from CD56dim cells, which probably explains the difference in cytotoxicity of the two populations. Where the CD56bright cells excrete high levels of immunoregulatory cytokines upon activation, they are less cytotoxic compared to CD56dim cells which in turn excrete negligible levels of cytokines when activated (Björkström et al. 2010; Farag and Caligiuri 2006; Vossen et al. 2008). Because of practical reasons NK cell development and function has been studied extensively in mice and the results of these studies have often been used to explain the human situation. Although the immune systems of mice and humans are highly similar, the NK cells of the two species show important differences. For instance, mice do not have the same immune receptors as humans do; where humans have KIRs, mice have LY49 receptors (Kim et al. 2002). Also, there are some important differences in NK cell homeostasis, as was shown by interleukin-15 (IL-15) blocking studies in both mice and humans (Lebrec et al. 2013). It is therefore critical that any research in mice is confirmed with human models. Natural killer cell education The discriminating characteristic of NK cells as compared to other immune cells is that NK cells can sense whether a host cell has become pathogenic or not. Clearly, they have to be tolerant towards healthy cells and become cytotoxic towards unhealthy cells. To distinguish the two, they sense whether antigen presentation on the target cell, specifically by major histocompatibility complex class I (MHC-I) molecules, is normal. An abnormal MHC-I expression pattern triggers the NK cell into killing of the target cell. During development and their entire life-time, NK cells are educated to recognize self through a ligand-instructed model by MHC-I (Jaeger and Vivier 2012; Jobim M and Jobim L F 2008; Orr and Lanier 2010) (Figure 1A). This instruction model has shown that in the mere absence of MHC-I molecules, NK cells are not well educated, show inefficient cytotoxicity and have problems to target MHC-I deficient cells (Fernandez et al. 2005; Kim et al. 2005). 10

13 GENERAL INTRODUCTION NK cell education in humans was shown to depend on both KIR receptors and C-type lectins, and their ligands play an important role in the process. In humans, MHC class I molecules comprise the classical (class Ia) human leukocyte antigens (HLA)-A, -B, and -C, and the nonclassical (class Ib) HLA-E, -F, -G and -H molecules. As will be explained in more detail below, HLA-A, -B, -C and -G are ligands for KIR (See Table 1), whereas HLA-E is a ligand for CD94/ NKG2A and CD94/NKG2C. HLA-E is the best-characterized MHC-Ib molecule, first described as a non-polymorphic ligand of the CD94/NKG2 receptors on NK cells. Both NKG2A and NKG2C pair with CD94 to form inhibitory and activating receptors specific for the HLA-E-canonical peptide complex. HLA-E preferentially binds to and presents leader sequence peptides derived from classical MHC class I molecules. The other important MHC-Ib molecule with a role in NK cell function is HLA-G, which expression is restricted to the fetal trophoblast cells that invade the maternal decidua during early pregnancy and acts as potential ligand for decidual NK cells expressing KIR2DL4 (Rajagopalan and Long 2012). 1 Table 1. Known HLA ligands for Killer Immunoglobulin-like Receptors KIR Ligand KIR2DL1 HLA-C2 KIR2DL2 HLA-C1, HLA-B*73, -B*46, some HLA-C2 KIR2DL3 HLA-C1, HLA-B*73, -B*46 KIR2DL4 HLA-G KIR2DL5A+B KIR2DS1 HLA-C2 (weak) KIR2DS2 HLA-C1 (weak) KIR2DS3 KIR2DS4 HLA-C*01 (weak), -*02, -*04, -*05, -*14, -*16, HLA-A*11 KIR2DS5 KIR3DL1 HLA-Bw4 (I80>T80) except HLA-B*13:01/02, HLA-A*23, -*24, -*25, -*32 KIR3DL2 HLA-A*3,-*11 (weak) KIR3DL3 KIR3DS1 HLA-Bw4 (weak) HLA subgroup HLA alleles HLA-C1 Alleles with asparagine at position 80 in the α -helix (HLA-C Asn80). Incl. HLA-C*01, *03, *07,* 08, *12, *13, *14, *16 and HLA-B*4601 and *7301. HLA-C2 Alleles with lysine at position 80 in the α -helix (HLA-C Lys80). Incl. HLA-C*02, *04, *05, *06, *15, *17, *18 HLA-Bw4 Alleles sharing a similar epitope at position in α1-helix. Incl. HLA-B*05, *5102, *5103, *13, *17, *27, *37, *38, *44, *47, *49, *51, *52, *53, *57, *58, *59, *63, *77 and HLA-A*09, *23, *24, *2403, *25, *32 Binding affinity of KIR3DL1 is stronger when there is an isoleucine at position 80 compared to a threonine. (Biassoni 2001; Foley et al. 2008; Graef et al. 2009; Pende et al. 1996; Rajagopalan and Long 1999; Stern et al. 2008b; Winter et al. 1998). 11

14 GENERAL INTRODUCTION 1 The expression of inhibitory receptors on NK cells was shown to increase upon interaction with self-mhc-i molecules, which leads to increased interaction of inhibitory receptors with MHC-I on host cells. This interaction between inhibitory receptor and self-mhc-i molecules induces MHC-I-dependent tolerance to self (Jaeger and Vivier 2012; Orr and Lanier 2010). This manner of education results in a higher sensitivity of the NK cell, which is now licensed to kill by self-mhc-i molecules, ensuring efficient target cell killing. Because of this, NK cells are activated when there is no interaction of inhibiting receptors and the cell is thus able to recognize missing self on tumor cells that downregulate MHC-I expression to escape detection by CD8-positive T cells. When a cell has not been educated by self-mhc-i, either because it is still a naïve NK cell or because there was no self-mhc-i ligand available, MHC-Iindependent tolerance to self can be induced, resulting in a higher activation threshold and thus a less efficient target cell killing (Jaeger and Vivier 2012; Orr and Lanier 2010). Several groups have suggested that external factors like viral infections can have a longlasting effect on the NK cell pool, suggesting additional mechanisms of NK cell education independent of MHC-I (Béziat et al. 2013a; Björkström et al. 2011; Gumá et al. 2004). However, this concept of alternative education has only been recently established and is as yet incompletely understood, as are many mechanistic aspects of the education process as well. Natural killer cell activation By being in a tightly balanced state between activation and inhibition NK cells contribute to a continuous immune surveillance. They circulate in the blood, ready to recognize and react to aberrant cells, such as virus-infected cells or tumor-transformed cells. Immune receptors on the cell membrane sense whether antigen presentation on the host cell is normal, resulting in activation of the NK cell when a host cell over-expresses activating receptor ligands on its membrane (Figure 1C) or when expression of inhibitory receptors (e.g. MHC-I) is low or absent (Figure 1B). In addition to direct interaction with the target cells NK cells can also react to indirect signals produced by other innate cells like macrophages or dendritic cells (O Connor et al. 2006). These signals can be pathogen-associated molecular patterns (PAMPs), which interact with Toll-like receptors (TLR) expressed on the cell surface of NK cells, or accessory cytokines like IL-2, IL-12, IL-15, IL-18 and interferon-α (IFN-α) cells (Adib-Conquy et al. 2014; O Connor et al. 2006). Their ability to mount a quick immune response without any need for antigen or sensitization or through antibody-dependent cellular cytotoxicity (ADCC) makes them key players in innate immunity. Once activated they can directly secrete cytotoxins such as perforin and granzyme B from preformed stored cytotoxic granules for destruction of the target cell, cytokines such as IFN-γ, tumor necrosis factor-α (TNF-α) and chemokines to impair tumor development as well as to initiate adaptive immune responses (Degli-Esposti and Smyth 2005). In addition to the innate immune responses of NK cells, recent studies have shown that NK cells are able to mount a more efficient cytotoxic response when a pathogen has been encountered before, suggesting some level of immunological memory, typical of adaptive immune cells (O Leary et al. 2006; Paust et al. 2010; Sun and Lanier 2011; Vivier et al. 2011). Sensitization with a hapten or virus generates antigen specific memory NK cells, which relocate 12

15 GENERAL INTRODUCTION A Host Tolerance KIR MHC-I 1 Education NK + Activation C Tumor Viral B Activation Figure 1. Natural killer cells stay in a constant balance between activation and inhibition. (A) KIRs on the cell membrane bind to MHC class I (MHC-I) molecules on healthy host cells. The cell receives as much activating as inhibiting signals, thereby leaving the cell in a resting state and thus causing tolerance towards the host cell. During development the cell has been educated by self-mhc-i molecules to recognize selfcells. (B) Tumor cells often downregulate the MHC-I on their membrane. The NK cell recognizes this state due to missing self antigens and is activated. (C) Virus infected cells present pathogenic peptides in their MHC-I molecules. KIRs bind these MHC-I molecules and transmit more activating signals than inhibiting signals to the cell, resulting in activation. to the liver where they can reside several months. After rechallenge, the antigen specific NK cells move towards the site of infection where they can mount a more efficient immune response (Paust and von Adrian 2011). However, these adaptive functions of NK cells have so far mainly been studied in mice, while studies on adaptive functions of human NK cells, which are shown to have different properties than murine NK cells, have only recently been commenced (Fauriat et al. 2009; Jaeger and Vivier 2012), leaving many questions about these memory NK cells. KILLER IMMUNOGLOBULIN-LIKE RECEPTORS Genes and haplotypes The KIR genes are clustered on the KIR locus located in the leukocyte receptor complex (LCR) on chromosome 19q13.4 (Trowsdale et al. 2001). In humans, seventeen highly homologous genes, have been identified to date; six activating (KIR2DS1-5, and KIR3DS1), nine inhibitory (KIR2DL1-4, KIR2DL5A, KIR2DL5B and KIR3DL1-3) and 2 pseudogenes (KIR2DP1 and KIR3DP1). For all of these genes a large number of allelic variants exists, which resemble each other up to 98% in sequence homology (Parham 2005; Uhrberg et al. 1997). Distribution of the genes depends on a limited set 13

16 GENERAL INTRODUCTION 1 of KIR haplotypes, all of which contain the same so-called framework genes (KIR3DL3, KIR3DP1, KIR2DL4 and KIR3DL2). The other KIR genes may or may not be part of one of these haplotypes, designated A and B haplotype. The KIR locus is composed of an A or B haplotype from the centromeric region (the part closest to the chromosome centromere) and an A or B haplotype from the telomeric region (the part closest to the chromosome telomere) (Figure 2). Also, because of the high homology between KIR genes, genome recombination events within the KIR locus due to homologous recombination are quite frequent, leading to novel hybrid genes and the duplication or deletion of multiple genes at the same time (Traherne et al. 2010). All these factors contribute to gene copy number variation (CNV) across the population, meaning that there can be 0, 1, 2 or even 3 copies of a specific KIR gene in a given individual. How these genes are organized and distributed has proven difficult to elucidate because of technical difficulties to discriminate the different genes due to the extremely high homology among the KIR genes combined with their high level of polymorphic variation. A major challenge in this field is to clarify the structural organization of the locus and to unravel and understand the genetics of the KIR locus in relation to the function of NK cells in the immune system in more detail. Proteins KIR molecules contain up to three immunoglobulin domains: D0, D1 and D2. KIRs designated with KIR3D contain all three domains, while KIR2D molecules contain only two domains. Type 1 KIR2D molecules miss D0 since exon 3 of the KIRs is spliced out, while type 2 KIR2D molecules do not have D1 since exon 4 of these genes is missing. KIRs can have either a long (L) or a short (S) cytoplasmic tail, which determines whether a receptor is inhibitory or activating. The MHC-I are the natural ligands for most KIRs as will be explained in more detail below. When the KIR receptor is engaged by one of its ligands, the receptor confers an activating or inhibiting signal to the cell, depending on the make-up of their cytoplasmic tail. A positively charged amino acid in the intracellular domain of the activating receptors can interact with the adaptor molecule DAP-12 which in turn contains immunoreceptor tyrosine-based activation motif (ITAM)- containing signaling molecules, while the long cytoplasmic tail of inhibiting receptors contains 1 or 2 immunoreceptor tyrosine-based inhibition motifs (ITIMs). Both motifs are phosphorylated Centromeric region Telomeric region A B 3DL3 2DL3 2DP1 2DL1 3DP1 2DS3 3DL3 2DS2 2DL2 2DL5B 2DS5 2DP1 2DL1 3DP1 2DL4 3DL1 2DS4 3DL2 2DS3 2DL4 3DS1 2DL5A 2DS5 2DS1 3DL2 Figure 2. The KIR locus is a highly variable region. A complete locus contains a centromeric part and a telomeric part, both of which can contain genes belonging to an A haplotype or a B haplotype. Each haplotype contains the same framework genes (in black). (Adapted from Vendelbosch S, et al; Extensive variation in gene copy number at the Killer Immunoglobulin-like Receptor locus in humans; PLoS ONE 2013; 8: e67619; Chapter 2 of this thesis.) 14

17 GENERAL INTRODUCTION by Src family kinases, upon which the ITIMs signal via recruited tyrosine phosphatases SHP-1 and SHP-2 and the ITAMs via Syk/Zap70 family protein tyrosine kinases (Purdy and Campbell 2009). The exception is KIR2DL4 which contains both an ITIM in its long cytoplasmic tail and an unknown motif that induces activating signals (Figure 3). As mentioned before, the net result of inhibiting and activating signals will determine to which extent an NK cell will become cytotoxic or not under certain conditions (Kulkarni et al. 2008; Purdy and Campbell 2009). In addition to its genetic variation, KIR protein expression is highly variable across the population owing to different factors that are still not all completely understood. First, KIR expression is clonally distributed, i.e. every NK cell expresses its own KIR repertoire depending on the DNA methylation status of each KIR in the cell (Chan et al. 2003; Uhrberg 2005). Methylation takes place at CpG sites located in and around the KIR genes and this prevents transcription factors from binding the promoter region and thus initiating transcription. Most de novo methylation takes place before birth, creating a relatively stable imprint on the NK cell population (Santourlidis et al. 2008; Turker 2002). Although some KIR positive NK cell populations seem to be able to expand more than others due to environmental factors, as is the case after cytomegalovirus infection which leaves a long-lasting imprint on the NK cell population (Béziat et al. 2013a), the general idea is that an NK cell clone is unable to change its expression pattern. 1 Type 1 Type 2 KIR3DS1 KIR2DS1 KIR2DS2 KIR2DS3 KIR2DS4 KIR2DS5 KIR2DL1 KIR2DL2 KIR2DL3 KIR2DL4 KIR2DL5A KIR2DL5B KIR3DL1 KIR3DL2 KIR3DL3 D0 D1 D2 D1 D2 D1 D2 D0 D2 D0 D2 D0 D1 D2 ITAM ITAM DAP12 ITAM ITAM DAP12 ITIM ITIM ITIM?? ITIM ITIM ITIM ITIM Figure 3. Composition of Killer Immunoglobulin-like Receptor molecules. Each KIR molecule has 2 or 3 Ig-like domains which are exposed to the exterior of the cell. Type I molecules contain only Domain 1 (D1) and D2 while type II molecules contain only D0 and D2. They are anchored to the cell membrane (grey dots) by the transmembrane domain and can have either a short or a long cytoplasmic tail. A short tail contains positively charged amino acid residues that can interact with DAP12 which in turn contains ITAMs, resulting in an activating signal upon engagement of the receptor. Long cytoplasmic tails contain 1 or 2 ITIMs which can induce an inhibiting signal to the cell. 15

18 GENERAL INTRODUCTION 1 The proportion of KIR-positive cells is different for each individual and is, at least for some KIR genes, directly related to the number of gene copies available in the genome, as we (see chapter 4) and others have shown (Béziat et al. 2013b; Vendelbosch et al. 2014). Apart from the methylation status of a gene, which determines whether a gene is available for transcription, the level of transcription may be dependent on the allele. Minor differences, like single nucleotide polymorphisms (SNPs), in the allele or promoter region can influence how much protein is presented on the cell membrane. This variation can be caused by differences in transcription factor binding sites in the promoter region, as was shown in a few cases (Li et al. 2008; McErlean et al. 2010), or because of folding or localization defects as was shown for some alleles of KIR3DL1 and KIR2DS4 (Middleton et al. 2007; Pando et al. 2003). Finally, NK cell activation can have a direct, and probably short-lived, effect on the level of KIR expression due to increased phosphorylation of transcription factors. This variation of KIRs on both the genetic and the protein level makes this family of receptors extremely difficult to study, resulting in few reports on the actual function of individual KIR receptors. MHC class I molecules MHC molecules are one of the most important immune modulators in the context of self and non-self recognition. In humans, they are also called HLA, especially when referring to the polymorphic alleles in the MHC system. The two classes of MHC molecules present endogenous peptides to cytotoxic T lymphocytes (CTL) and NK cells. Class I molecules are present on all nucleated cells, whereas class II molecules are expressed on specialized antigen presenting cells (APCs) (Penn and Ilmonen 2001). The MHC genes belong to one of the most diverse and complex gene families due to their extreme polymorphic nature. MHC class I molecules are the natural ligands for KIRs and other NKRs, but since the KIR and MHC families reside on different chromosomes, they segregate independently within the population. This means that the ligand for a certain KIR is not necessarily expressed by cells from that individual. Nevertheless, at the population level the natural selection of KIR genes and MHC genes through evolution seems to be influenced by each other (Parham et al. 2012). Although MHC ligands have been described for the majority of the (inhibitory) KIRs, not all ligands have as yet been found (Table 1). Also, it has been suggested that there are other non-mhc ligands for KIR genes, and more in particular for the activating KIRs, although none have been defined so far (Katz et al. 2004). As shown in Table 1, most KIRs can bind multiple HLA alleles, but they do not bind each molecule with the same affinity. Many KIR seem to have a binding affinity to HLA-C1 molecules, allotypes of HLA-C with an asparagine at position 80 in the α-helix, or HLA-C2 molecules, which have a lysine at this position33. However, also some HLA-A and HLA-B alleles were found to bind KIR and HLA-G, expressed mainly in placental cells in the uterus, binds KIR2DL4 specifically (Rajagopalan and Long 2012). The strength of the activating or inhibiting signal depends on the strength of binding between KIR and HLA molecule. The same holds true for other receptorligand interactions on NK cells, such as the binding of CD94/NKG2A and CD94/NKG2C to HLA-E. The net result of all NKR signals will determine whether an NK cell will be activated or inhibited (Kulkarni et al. 2008; Purdy and Campbell 2009). 16

19 GENERAL INTRODUCTION KIR IN DISEASE Viral infections As explained above, KIR function is difficult to study. Nevertheless, there have been multiple studies characterizing function of one or multiple KIRs. Within the immune system, the NK cell is the first line of defense against viral infections. For instance, KIRs seem to play a role in human immunodeficiency virus type 1 (HIV-1) infection. In particular, it has been shown that a higher number of genes for KIR3DS1, or a higher number of genes for KIR3DL1 in the presence of KIR3DS1 and the appropriate ligands, relates to a better individual resistance against human immunodeficiency virus type 1 (HIV-1). This resistance was related to an inhibition of replication of HIV-1 and was associated with the expansion of KIR3DS1 protein-expressing NK cells (Pelak et al. 2011). This study not only showed the contribution of KIR in the defense against HIV-1, it also showed the immunological relevance of KIR gene CNV in disease. The importance of KIR gene CNV in relation to disease was also supported by the correlation of increased clearance of hepatitis C virus (HCV) in individuals with two copies of KIR2DL3 compared to those with one or no copies (Khakoo et al. 2004). 1 Tumor development The second function for constant NK cell immune surveillance is the detection of transformed cells and prevention of tumor development. Tumor cells often have down-modulated MHC-I expression, leading to activation of the NK cell and subsequent destruction of the tumor cell (Purdy and Campbell 2009). During hematopoietic stem cell transplantation (HSCT) donor-kir / recipient-ligand mismatching has been shown to be beneficial for the killing of lymphomas and solid tumors by alloreactive NK cells and for prevention of graft-versus-host-disease (GVHD) (Leung 2011; Ruggeri et al. 2002). Also, the number of activating KIRs on NK cells of the donor may be related to a beneficial effect for patients with acute myeloid leukemia (AML) receiving HSCT (Cooley et al. 2008). Transplantation Besides the desired effects that NK cells have against viral infection and tumor growth, some undesired effects related to transplant rejection, autoimmunity and chronic inflammatory diseases are also caused by these cells. One can imagine that the cells that live in such a precarious balance between active and resting state, could attack healthy cells when overactivated. Or that the same cells which are supposed to distinguish aberrant host cells from healthy cells will recognize transplanted cells as non-self, causing rejection of the transplanted tissue. For this reason, studies have been performed to identify the optimal combination of KIR and HLA ligand of both donor and patient during various transplantation settings. HLA molecules have long been confirmed as critical factors during transplantation and HLA matching for organ transplantations is widely applied. Still, even HLA-matched transplants are prone to rejection, probably because additional factors are involved, such as KIR. The alloreactive characteristics of inhibitory KIR-ligand mismatched NK cells during transplantation may generate a welcome anti-tumor effect, but it can also lead to rejection of 17

20 GENERAL INTRODUCTION 1 organs after transplantation (van Bergen et al. 2011). However, the presence of activating KIR genes in the donor grafts has been suggested to improve survival after unrelated HLA-matched HSCT (Gallez-Hawkins et al. 2011; Venstrom et al. 2010). Also, it seems that the presence of inhibitory KIR genes in the recipient s genotype in the absence of the appropriate HLA ligands contributes to a successful transplantation, whereas inhibitory KIR genes with matched HLA ligands in the recipient in combination with a mismatched HLA ligand in the donor results in poor graft function in the long-term (van Bergen et al. 2011; Venstrom et al. 2009). The exact contribution of KIR genes to the success of transplantation has proven difficult to validate, as there are so many variables during transplantation, of which KIR genotype variation may only be one of the important factors. However, the data so far does suggest that a favorable combination of HLA and KIR genotyping, i.e. inhibitory KIR genes of the patient matched with their HLA ligands in the donor, do enhance the chance of success of transplantation (Kunert et al. 2007; Stern et al. 2010; van Bergen et al. 2011; Venstrom et al. 2009). Auto-immune diseases Although it is not always clear whether NK cells play a part in disease onset, pathogenesis or resistance, they have been implicated in a number of the most prevalent auto-immune disorders. Some well studied examples are rheumatoid arthritis (RA), systemic lupus erythematosus (SLE) and type 1 diabetes mellitus (T1DM), due to relatively lower NK cell numbers in the peripheral blood of patients (Kúsnierczyk 2013). Instead, NK cells seem to have accumulated in the affected tissues in these patients (Fogel et al. 2013). Whether NK cells are contributing to the disease in any way, or whether NK cell involvement is merely a response to the manifestations of the disease is still unclear. However, several studies concerning the association of KIR with certain auto-immune disorders suggest that NK cells play a role during disease pathogenesis. In general, it seems that inhibitory KIRs are more often associated with a protective effect against auto-immunity, while the activating KIRs seem to increase the risk for disease development (Fogel et al. 2013). This is also the case for gene association studies which have found that KIR2DL2, KIR2DL3 and KIR3DL1 to be negatively associated with RA, while KIR2DS2 and KIR3DS1 where positively associated (Prakash et al. 2014; Ramírez-De los Santos et al. 2012). In this respect, we have studied the link between KIR and Ankylosing Spondylitis (for more details see chapters 5 and 6). Furthermore, the occurrence of type 1 diabetes was also found to be associated with more activating KIRs, specifically KIR2DS2, in combination with its HLA-ligand; HLA-C1 (van der Slik et al. 2003). Finally, development of the skin disease psoriasis vulgaris, which has some autoimmune features, is associated with the gene KIR2DS1 (Kúsnierczyk 2013). In summary, whenever a host cell is behaving in a way it should not, NK cells seem to be around to monitor this behavior and to take action if necessary. Moreover, whenever such a situation occurs, KIRs seem to be guiding the actions, at least in part. At the same time however, the actual contribution of each of the individual KIRs as well as their collective activities within the human immune system, are still poorly understood. 18

21 GENERAL INTRODUCTION SCOPE OF THIS THESIS The central aim of the studies described in this thesis was to provide insight into the genetics and function of the human KIR family and its individual members. To understand the role of KIRs in disease and immunity, it is first necessary to understand the genetic distribution and how this relates to protein expression and ultimately NK cell function. The high homology between KIR genes has thus far hampered the development of proper genotyping tools. Therefore, we have developed a novel method to KIR genotype large cohorts of individuals and to determine CNV, as described in Chapter 2. In this chapter we also show that the KIR locus is subject to extreme gene copy number variation. How the KIR locus is organized, was further investigated in Chapter 3, where we used the intergenic sequences between the adjacent KIR genes to study the human KIR locus organization. We described some novel insights into the distribution of the KIR genes across the haplotypes and the discovery of novel hot-spots of recombination, causing an additional level of variation within the KIR locus. In addition to KIR variation at the genomic level, KIR protein expression is also highly variable across the population. In Chapter 4 we related KIR genotype, CNV and allelic variation with RNA and protein expression levels. This chapter also focuses on the ability of interleukin-15 to increase KIR protein expression on NK cells in vitro. Finally, we have studied the role of KIR genotype in disease. Ankylosing Spondylitis (AS) is an auto-immune disease affecting the spine and sacroiliac joints in particular and is highly associated with the allele HLA-B27. HLA-B27 is a natural ligand for KIR3DL1 and KIR3DL2, therefore we studied whether CNV of these genes was associated with the development of AS in Chapter 5, and this appeared not to be the case. However, in Chapter 6 we showed that in a large Caucasian cohort the KIR3DL1 gene is in fact associated with disease course or severity rather than development of disease or susceptibility. In Chapter 7 we summarize all results described in this thesis and discuss our findings in light of diagnostic significance and a future scientific perspective. 1 19

22

23 2 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE KILLER IMMUNOGLOBULIN-LIKE RECEPTOR LOCUS IN HUMANS Sanne Vendelbosch, Martin de Boer, Remko ATW Gouw, Cynthia KY Ho, Judy Geissler, Wendy TN Swelsen, Michael J Moorhouse, Neubury M Lardy, Dirk Roos, Timo K van den Berg and Taco W Kuijpers Published in: PLOS ONE 2013, 8 (6), e67619

24 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 ABSTRACT Killer immunoglobulin-like receptors (KIRs) are involved in the regulation of natural killer cell cytotoxicity. Within the human genome seventeen KIR genes are present, which all contain a large number of allelic variants. The high level of homology among KIR genes has hampered KIR genotyping in larger cohorts, and determination of gene copy number variation (CNV) has been difficult. We have designed a multiplex ligation-dependent probe amplification (MLPA) technique for genotyping and CNV determination in one single assay and validated the results by next-generation sequencing and with a KIR gene-specific short tandem repeat assay. In this way, we demonstrate in a cohort of 120 individuals a high level of CNV for all KIR genes except for the framework genes KIR3DL3 and KIR3DL2. Application of our MLPA assay in segregation analyses of families from the Centre d Etude du Polymorphisme Humaine, previously KIRgenotyped by classical techniques, confirmed an earlier reported duplication and resulted in the identification of a novel duplication event in one of these families. In summary, our KIR MLPA assay allows rapid and accurate KIR genotyping and CNV detection, thus rendering improved transplantation programs and oncology treatment feasible, and enables more detailed studies on the role of KIRs in human (auto)immunity and infectious disease. 22

25 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS INTRODUCTION The immune system is a complex network that protects the human body against internal and external threats. It consists of an innate and an adaptive system. Natural killer (NK) cells constitute the first line of innate immunity against viral infections and tumor development. Recent studies have shown that NK cells can also be educated during development and that memory NK cells are able to mount a more effective cytokine response upon reactivation, suggesting adaptive functions (Sun J.C. and Lanier L.L 2009; Vivier et al. 2011). The importance of human NK cells has become clear through studies concerning cancer, viral infections and donor/recipient panels in transplantation settings. The cytotoxicity of NK cells is regulated through large families of molecules with a homologous structure, belonging to the C-type lectin-like molecules and immunoglobulin-like receptors (Parham et al. 2010). The most prominent family that regulates NK cell cytotoxicity in humans comprises the killer immunoglobulin-like receptors (KIRs). Binding of a KIR to one of its ligands, Human Leukocyte Antigens (HLA) class-i molecules and possibly others, triggers either an activating or an inhibiting signal, thereby controlling the immune reactivity of NK cells. A widely studied application is the matching of KIRs in donors and recipients in transplantation settings (Kunert et al. 2007; Ruggeri et al. 2002; Stern et al. 2010; van Bergen et al. 2011). Alloreactive NK cells can lead to rejection of organs after transplantation (van Bergen et al. 2011). In contrast, donor KIR-ligand mismatching with the recipient during hematopoietic stem cell transplantation (HSCT) may be required for the beneficial effects of NK cell alloreactivity against tumor cells (Leung 2011; Ruggeri et al. 2002). Apart from inhibitory KIR-ligand mismatches, the presence of activating KIR genes in the donor grafts has been shown to improve survival after unrelated HLA-matched HSCT (Gallez-Hawkins et al. 2011; Venstrom et al. 2010). These studies and others have shown that a favorable combination of HLA and KIR genotyping of both donor and recipient can strongly enhance the chance of success of transplantation. The genes for the KIR family are present on human chromosome 19q13.4 (Trowsdale 2001). In humans, seventeen highly homologous genes, including two pseudogenes, have been identified to date. For all of these genes a large number of allelic variants exists, which resemble each other up to 98% in sequence homology. Distribution of the genes is thought to occur on two basic haplotypes, designated A and B, both of which contain the same so-called framework genes (KIR3DL3, KIR3DP1, KIR2DL4 and KIR3DL2). Other KIR genes may or may not be present on one of the haplotypes (Parham 2005; Uhrberg et al. 1997). In addition, individuals may carry more than two copies of a KIR gene, as has been shown so far only in case of KIR3DL1 and KIR3DS1 (Martin et al. 2003; Pelak et al. 2011; Traherne et al. 2010). The immunological relevance of copy number variation (CNV) in the KIR3DL1 and KIR3DS1 genes was shown by Pelak and coworkers, who described that a higher number of genes relates to a better individual resistance against human immunodeficiency virus type 1 (HIV-1). This resistance was related to a higher number of KIR3DS1 protein-expressing NK cells and an inhibition of replication of HIV-1 (Pelak et al. 2011). The importance of CNV in relation to disease was also supported by the correlation of increased clearance of hepatitis C virus (HCV) in individuals with of two copies of KIR2DL3 compared to those with one or no copies (Khakoo et al. 2004). 2 23

26 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 Tools to study the extent and functional meaning of this inter-individual KIR locus variation on a larger scale are currently lacking. From other immune receptors we know that extensive CNV exists and that the number of genes can relate to protein levels, leading to a difference in disease outcome. We have observed this for the genes encoding the human Fc-gamma receptors (FcγR) and complement factor 4 (Breunis et al. 2008; Breunis et al. 2009; Wouters et al. 2009). To study the level of CNV in the KIR gene cluster, we have used a convenient, sensitive and efficient MLPA-based method to detect all KIR genes in one assay. The introduction of a synthetically derived calibrator has allowed us to accurately quantify KIR gene CNV. Validation of the method by comparing the MLPA method with the standard polymerase chain reaction (PCR) with sequence-specific primers (SSP) in a large cohort of individuals indicated that the MLPA method provided more accurate genotyping. The MLPA method showed an unexpected range of CNV at the KIR locus. Segregation analyses in pedigrees previously genotyped by PCR-SSP confirmed the strength and accuracy of the MLPA method and helped to identify duplication events in the KIR gene cluster in one of these families. Taken together, CNV at the KIR locus is extensive. The KIR MLPA assay can accurately determine an individual s KIR genotype in a highly efficient manner, allowing routine KIR genotyping in transplantation programs and offering the opportunity to increase the success rates of transplantation or graft-versus-leukemia effects. Also, genotype-phenotype relations may be studied in greater detail to better understand the exact role of KIRs in health and disease. RESULTS KIR genotyping by MLPA methodology In designing the set of probes for the KIR MLPA we used the following principles. First, making use of publicly available databases ( we selected single nucleotide polymorphisms (SNPs) in each KIR gene that selectively distinguish one gene from all others. In some cases two of those SNPs were combined together in one probe set that consists of three probe parts, to make the probe segregating between the KIR genes. This has previously been applied in the MLPA assay for the complement genes in the HLA class III region (Wouters et al. 2009). The KIR gene-specific SNPs were chosen such that all allelic variants would be recognized by the corresponding probe. Secondly, some specific, additional KIR variant sequences were selected to make a further distinction among the variants of a particular KIR gene. In this way, a probe-based distinction was possible between wild-type KIR2DS4 and truncated variants of this gene (KIR2DS4*003-*010, *012, *013) (Crum et al. 2000; Hsu et al. 2002b; Middleton et al. 2007). Next, we designed our probes mainly inside exons of the selected gene sequences (Supplementary Tables 1 and 2). One allelic variant of KIR3DL1 (KIR3DL1*054) is not detected, because this allele contains a SNP at the specific probe-binding site. Where possible, two separate independent probes per KIR gene were included in the assay as an internal check on the presence or absence of any of the genes at the KIR gene cluster. The latter is particularly important for excluding putative false-negative results due to SNPs that have not been documented yet. To avoid, on the one hand, possible competition between probes for 24

27 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS the same KIR gene and on the other hand to accommodate all probe amplification products in the same assay, we divided the probes over three different probe mixes (Supplementary Table 3). Accuracy of the KIR MLPA assay The MLPA method was subsequently performed to genotype 120 Caucasian individuals (Table 1). The occurrence frequency of KIR genes in our study population of control individuals was similar to the frequency reported in the Allele Frequency Net Database ( allelefrequencies.net/kir6002a.asp; update November 2011 (Gonzalez-Galarza et al. 2011)) (Table 1), hence lending credibility to the MLPA-derived genotyping of KIR genes. PCR with sequence-specific primers (SSP) was performed routinely in parallel to evaluate the accuracy of the MLPA assay by comparing the results in the same cohort of 120 individuals (Table 1). Upon comparison, the datasets of both assays matched for 99.5%. Of the 17 KIR genes per donor typed by both methods, in these 120 individuals 12 individuals showed a mismatch in a single KIR gene. All KIR genes of 5 individuals were amplified by long-range PCR and sequenced by Nextgeneration sequencing (NGS) on a 454 FLX Roche Genome Sequencer to resolve these 2 Table 1. KIR genotype frequency as determined by MLPA, compared with data from the allele frequency database and determination by PCR-SSP. KIR Gene MLPA Allele frequency Database* MLPA vs PCR-SSP 2DL1 95,8% % 99,2% 2DL2 53,3% % 100% 2DL3 87,5% 84-97% 100% 2DL4 100,0% % 100% 2DL5 56,3% % 98,3% 2DS1 37,5% % 93,3% 2DS2 53,8% 40-63% 100% 2DS3 31,7% % 100% 2DS4all 96,6% % 100% 2DS4wt 39,2% % 100% 2DS4trunc 81,7% % 100%** 2DS5 33,3% % 100% 2DP1 95,8% % 99,2% 3DL1 96,7% % 100% 3DL2 100,0% 100,0% 100% 3DL3 100,0% 100,0% 100% 3DS1 41,7% 27-63% 100% 3DP1 100,0% % 100% n=120; * Range of Caucasoid populations ** SSP data available for 25 donors 25

28 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 discrepancies (Figure 1). In all cases, the methodological discrepancies appeared to be due to false-positive or false-negative PCR reactions in the PCR-SSP. Upon re-evaluation of the PCR-SSP data, some explanations for the mismatches were found. For one, KIR2DS1 is represented by a long 1800 base-pair PCR product, leading to some false-negative results for this KIR gene in the PCR-SSP, due to insufficient amplification. Also, the particular PCR product detected for KIR2DP1 appeared to be too long, leading to an aspecific product and a false-positive result for KIR2DP1 in the PCR-SSP. From these results we conclude that the KIR MLPA method performed with a higher degree of accuracy than the PCR-SSP method. Quantification of KIR CNV with a calibrator The proposed haplotypes (Figure 2A) contain a variable number of KIR genes within each individual s genome, leading to gene copy number variation (CNV). Within each genome particular genes may be absent or present once, twice and sometimes even more than twice (Martin et al. 2003; Pelak et al. 2011; Traherne et al. 2010; Trowsdale et al. 2001; Uhrberg et al. 1997). Our previous experience with the MLPA methodology had already indicated that this technique allows a reliable assessment of the gene copy number in the human genome (Breunis et al. 2008; Wouters et al. 2009). However, the MLPA also has some disadvantages, one of which is the use of a reference DNA sample that is employed to determine the gene copy number and to reduce inter-assay variation. Since the KIR genotype of this reference % of all unique reads (per donor) Donor 1 Donor 2 Donor 3 Donor 4 Donor 5 0 MLPA SSP-PCR KIR2DL1 KIR2DL5 KIR2DS1 KIR2DP1 KIR3DL3 Figure 1. Validation of the accuracy of genotyping with MLPA by next-generation sequencing (NGS). Five of the genomes with mismatches between MLPA and PCR-SSP were analyzed by NGS. The percentage of reads that aligned against a reference for a KIR gene gives an impression about the presence of the KIR gene in that specific donor. Circles indicate the mismatches found between MLPA and PCR-SSP. KIR3DL3 is a control gene as it is present in all donors. 26

29 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS sample is unknown, this does not allow an accurate determination of individual KIR gene CNV. To circumvent this problem we created a synthetic control as an internal calibrator for our KIR MLPA method. This internal standard consists of a DNA vector that includes the sequences of all of the probe-binding sites once. The gene-assembly method, as previously described by Stemmer and co-workers, was adapted to create the KIR calibrator (Stemmer et al. 1995). With this calibrator we analyzed the cohort of 120 individuals for CNV (Figure 2B, Supplementary Tables 4 and 6). According to our previous experience the number of 0, 1, 2, 3 and >3 genes can be discriminated in a highly reproducible manner. Several results should be noted. First, the framework genes KIR3DL2 and KIR3DL3 were represented by 2 copies in most individuals, which fits with the current A and B haplotype distribution system (Parham 2005). However, the other two framework genes KIR3DP1 and KIR2DL4, separating the two blocks of genes in each haplotype (Figure 2A), have been deleted or duplicated in a small percentage of individuals. KIR2DL4 occurs 2 times in 88% of the individuals, which suggests that translocation events have taken place in 12% of the individuals. The same seems true for KIR3DP1, which occurs 2 times in 84% of the individuals. Furthermore, we observed variation in gene copy numbers for KIR2DL1, KIR2DP1, KIR2DL3 and KIR3DL1. The KIR2DL3 and KIR3DL1 genes, which are considered to be specific for the so-called A haplotype (Uhrberg et al. 1997), occur in relatively high numbers, with >70% of individuals carrying 2 or 3 copies. Finally, several of the activating genes of the so-called B haplotype (Uhrberg et al. 1997), such as KIR2DS1, KIR2DS2 and KIR2DS5, are rarely seen in 2 or more copies within the genome of the population tested. Within the tested population, the average number of copies per inhibiting receptor was 1.2 (not including the framework genes), while the average number of copies for activating receptors was Quantification of CNV by MLPA validated by sequencing To validate whether our MLPA could determine CNV accurately, we needed an independent assay that could quantify the number of copies of the same gene or the number of the same gene fragments. For this purpose, we used the presence of short tandem repeats (STRs) in the KIR gene locus. STRs are short sequences that have been duplicated multiple times, resulting in stretches of conserved repeated sequences. In KIR genes, several STR regions exist, one of which is a repeat of 4 base pairs, AGAT, located in intron 4 of most KIR genes (Martin et al. 2000; Sambrook et al. 2005; Trowsdale et al. 2001). Closer examination of this region showed that each KIR gene contains an invariable number of repeats, with the exception of KIR2DL4 and KIR2DL5, which lack the typical intron 4 of the other KIR genes. Within this STR region single nucleotide polymorphisms (SNPs) are present, which allows for identification of the individual KIR genes even if the number of the repeats is equal. It is impossible to determine CNV of the KIR genes by these STRs by MLPA methodology. Apart from the many technical issues with respect to the design of such probes, these STRs may not only be present within the KIR gene cluster but might occur more widely throughout the human genome. However, it is possible to use the information about STRs within the KIR gene cluster for an independent validation of KIR CNV quantification obtained by the MLPA method. NGS enabled specific STR detection to determine the presence or absence of KIR genes (except KIR2DL4 and KIR2DL5). CNV can be quantified by determining the coverage of the amplified 27

30 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 A A B 3DL3 2DL3 2DP1 2DL1 3DP1 2DL4 3DL1 3DL3 2DS2 2DL2 2DL5b 2DS3 2DS5 2DP1 2DL1 3DP1 2DL4 3DS1 2DL5a 2DS3 2DS5 2DS1 3DL2 Framework gene A Haplotype A Haplotype A + B B Haplotype B 2DS4 3DL2 B DL DL DS DL DL DS DS DP DL DP DL DL DS DS DS4 WT % of test panel DS4 truncated DL # KIR gene copies Figure 2. KIR gene copy number determination of a large cohort by MLPA. (A) Overview of the Killer Immunoglobulin-like Receptor gene family. The genes are arranged in two haplotypes (A and B), which both always contain the framework genes as depicted in black. Both haplotypes may or may not contain the genes in white. The genes depicted in grey are supposedly found only on haplotype A or B. (B) KIR gene copy numbers were determined for 120 individuals by MLPA. Shown is a graphical representation of the percentages of donors with 0, 1, 2, 3 or more than 3 copies of each KIR gene in their genome. product, i.e. the relative occurrence of a certain STR length and specific sequence, as a signature of the corresponding KIR gene. Hence, to evaluate the accuracy of CNV determination in the KIR MLPA assay, primers were designed to amplify these repeat stretches in 12 samples. The PCR products were specific for each KIR gene because of the sequence and length (ranging from 390 to 450 bp; Supplementary Table 5). Subtle sequence differences allow for a distinction between KIR genes with the same number of repeats, except for KIR2DL1 and KIR2DS1, which are inseparable in this region. NGS yielded information about the copy number of the KIR-specific STRs of these 12 samples, which corresponded 100% with the MLPA test results (Figure 3). The STR assay thus confirmed the MLPA data for the KIR genes that can be detected with this assay. 28

31 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS DL1/2DS DL2 8 2DL3 10 2DP DS2 10 2DS3 10 2DS4 15 2DS % of reads per donor 15 3DL1 25 3DL3 15 3DS1 15 3DP # of copies according to MLPA Figure 3. Validation of the KIR MLPA method with an independent short tandem repeat (STR) assay. A primer set was used to amplify a stretch of repeats that occurs in intron 4 of most KIR genes. Nextgeneration sequencing (NGS) was used to sequence the amplification products based on both specific sequence and number of repeats. The graphs show the percentage of reads in NGS, aligned against the reference sequence of that KIR, in donors with a certain number of copies of that gene as determined by MLPA. The number of reads of framework gene KIR3DL2 was used as a reference to adjust for the total number of genes per donor. KIR2DL1 and KIR2DS1 have the same number of STR and no SNPs to distinguish one from the other and have therefore been combined. Several SNPs within the repeat region of KIR2DL3 and KIR3DL3 confirm the copy numbers as found with the MLPA method. Copy number variation validated within CEPH pedigrees The human genome diversity cell line panel (HGDP) and Centre d Etude du Polymorphisme Humaine (CEPH) bank contain a large number of cultured lymphoblastoid cell lines (LCLs) that are available for family segregation analyses and genome research ( hgdp/diversity.php). We typed the KIR genome of two generations of 3 CEPH families and a family from Israel. Family segregation analysis functions as a third form of validation, because the CEPH families have been extensively typed by other groups with PCR-SSP (Martin et al. 2008; Traherne et al. 2010). These families were typed in detail by KIR MLPA because of CNV and the indication of duplication events. Using the KIR gene calibrator as our reference sample, we determined the CNV for all the KIR genes of both parents and some or all children in these families and mapped possible inheritance patterns (Figure 4, Supplementary Figures 1, 2 and 3). For all families tested, the genes of all genotyped children were traced back to the two alleles of both parents. Interestingly, in one family (Figure 4) we witnessed a cross-over of part of allele B to allele A. The father has 2 copies each of KIR2DL5 and KIR2DS3, while the mother has none. These genes are present twice in the children that inherited the A allele of the father. As the siblings that 29

32 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 received the B allele only have one copy of each gene, this excludes the possibility that the genes were already duplicated in the father. This duplication event was confirmed with the STR assay described above for KIR2DS3 and with a qpcr on DNA for KIR2DL5 and KIR2DS3 (Supplementary Figure 4). The mother of family 1416 (Supplementary Figure 1) contained a duplication of KIR3DP1 and KIR2DL4, which was subsequently detected in three of the children. These data confirm the results of Martin and co-workers (Martin et al. 2003) for the same family, who established a crossing-over event of part of a haplotype between KIR2DL5 and KIR3DP1. The translocation event described by Traherne and co-workers (Traherne et al. 2010) in family 1413 (Supplementary Figure 2) can also be detected by the MLPA method. In addition, when testing another series of untyped control pedigrees we picked up a family with a lack of the framework genes KIR2DL4 and KIR3DP1 in all three children (Figure S3), a phenomenon that was previously reported to occur only rarely (Nowak et al. 2011). These observations lead us to conclude that the KIR gene cluster is subject to more translocation events than is generally assumed. Father A B 11 3DL3 1 2DL3 1 2DS2 1 2DL DP DL1 11 3DP1 11 2DL4 3DL DS DL DS3 2DS DS1 2DS4WT 2DS4trunc 11 3DL2 Mother C D 11 3DL3 1 2DL3 1 2DS2 1 2DL2 1 2DP1 1 2DL1 11 3DP1 11 2DL DL1 3DS1 2DL5 2DS3 2DS5 2DS DS4WT 2DS4trunc 11 3DL2 Child Child Child Child B D B C A C A D 11 3DL3 11 3DL3 11 3DL3 11 3DL DL3 1 2DL3 2DL3 1 2DL3 2DS2 1 2DS DS2 1 2DS2 2DL2 1 2DL DL2 1 2DL DP1 1 2DP1 1 2DP DP DL1 1 2DL1 1 2DL DL1 11 3DP1 11 3DP1 11 3DP1 11 3DP1 11 2DL4 11 2DL4 11 2DL4 11 2DL4 1 3DL1 1 3DL1 1 3DL1 1 3DL1 1 3DS1 1 3DS1 1 3DS1 1 3DS1 1 2DL5 1 2DL DL DL5 1 2DS3 1 2DS DS DS3 2DS5 2DS5 2DS5 2DS5 1 2DS1 1 2DS1 1 2DS1 1 2DS1 1 2DS4WT 1 2DS4WT 1 2DS4WT 1 2DS4WT 2DS4trunc 2DS4trunc 2DS4trunc 2DS4trunc 11 3DL2 11 3DL2 11 3DL2 11 3DL2 Figure 4. KIR gene pedigree analysis of a Centre d Etude du Polymorphisme Humaine family by KIR-MLPA. Children 04, 06, 09 and 16 of family 1347 show a translocation event involving KIR2DL5 and KIR2DS3. 30

33 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS MATERIALS AND METHODS Ethics Statement The study was approved by the Institutional Medical Ethics Committee of the Academic Medical Center in Amsterdam and was performed in accordance with the Declaration of Helsinki. Participants provided their written informed consent to participate in this study. 2 DNA isolation DNA was purified from 120 healthy donors with a QIAgen Blood Kit (Qiagen, Venlo, The Netherlands) according to the manufacturer s protocol. Multiplex ligation-dependent probe amplification The KIR MLPA is based on the method described by Breunis et al. (Breunis et al. 2008). In short, 100 ng of DNA was denatured at 98 C for 5 min in a thermocycler with heated lid and cooled to 25 C. Three µl of probe mixture, containing KIR-specific probes (Supplementary Table 1), control probes (Supplementary Table 2) and competitor probes, was added and denatured at 95 C for 1 min, then hybridized at 60 C for at least 16 hours. The mixture was cooled to 54 C before addition of 32 µl of Ligase 65 mixture (MRC Holland, Amsterdam, The Netherlands). Ligation took place at 54 C for 15 min, deactivation at 98 C for 5 min, and the mixture was cooled to 4 C. Bound probes were amplified by PCR: 10 µl of ligation mixture was added to 5 µl of Accuprime Taq Buffer, 0.4 µl of Accuprime Taq enzymes (AccuPrime Taq DNA Polymerase High Fidelity kit, Life Technologies, Carlsbad, CA, USA) 1.5 µl of 50 mm MgSO4 and 1.5 µl of a single primer pair of which one primer was labeled with a FAM group. PCR conditions included 36 cycles of 30 sec denaturation at 95 C, 30 sec annealing at 60 C and 1 min elongation at 68 C, followed by 20 min elongation at 68 C. Fluorescently labeled products were thus created, allowing for fragment analysis by capillary electrophoresis. 1.5 µl of the PCR reaction was mixed with 9 µl of deionized formamide and 1 µl of Promega Internal Lane Standard (Promega, Madison, WI, USA) and incubated at 90 C for 10 min before product separation on an ABI-3130XL (Applied Biosystems by Life Technologies). Fragment analysis was done with the program GeneMarker (version 1.90, SoftGenetics, State College, PA, USA). Next-generation sequencing For whole KIR genome sequencing, a long-range PCR was performed of each donor. Several primers were designed that amplified the whole KIR genome; FW universal (5 -gccaaataacatcctgtgcgctgctcagct-3 ), FW 3DL3 (5 -ctcacaacatcctgtgtgctgctaactga-3 ), FW 2DL4 (5 -cacatcgtctgcaccggtcagtcgagccga-3 ), REV universal (5 -ttggagaggtgggcaggggtcaagtg-3 ) and REV 3DP1 (5 -ctccatctgagggtcccctgaatgtg-3 ). The PCR reaction was performed with 500 µm dntps, 1x PCR Buffer 2 (Expand Long Template PCR system, Roche Diagnostics, 454 Life Sciences, Branford, CT, USA), 300 µm of both primers and 3.75 U polymerase enzyme mix (Expand Long Template PCR system, Roche Diagnostics). The PCR program consisted of an initial denaturation 31

34 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 of 94 C followed by 10 cycles of 10 sec at 94 C, 30 sec at 60 C and 12 min at 68 C, then another 20 cycles in which the denaturation at 94 C was extended to 15 sec, followed by final elongation of 7 min at 68 C. The resulting PCR products were up to 17 kb in length. The fragments were sequenced on a 454 FLX Roche Genome Sequencer (Roche Diagnostics, 454 Life Sciences, Branford, CT, USA), with the Shotgun Library method. ReferenceMapper software (Roche Diagnostics) was used to align the reads to a reference sequence. Synthesis of the KIR-MLPA calibrator The KIR-MLPA calibrator was created with an assembly method adjusted from Stemmer et al. (Stemmer et al. 1995). The intended synthetic product was divided in six separate sequences. Each sequence was cut up into small oligodeoxyribonucleotides of 60 base pairs that covered either the sense or the anti-sense strand with a 20-base pair overlap of each strand. These oligo s were synthesized by Life Technologies (Carlsbad, CA, USA). Multiple 60-mers were fused together in a PCR reaction: 2 µm oligo s, 0.2 µl of Accuprime Taq polymerase, 0.4 µl of Accuprime Pfx polymerase, 2 µl of Buffer I (Life Technologies) and de-ionized water were combined to a volume of 20 µl. This reaction contained a denaturation step at 94 C for 15 sec, then 55 cycles of 30 sec at 94 C, 30 sec at 52 C, 60 sec at 68 C. Subsequent amplification of 1 µl of this product with 5 µl of the two outer oligo s of each synthesized sequence, 0.5 µl of Accuprime Taq polymerase, 1 µl of Accuprime Pfx polymerase, 5 µl of Buffer I and de-ionized water up to 50 µl. This second PCR program included 24 cycles of 30 sec at 94 C, 30 sec at 52 C, 90 sec at 68 C. With this method, products of around 680 base pairs were synthesized. For larger sequences, 1 µl of two 680-mers were pooled together (with an overlap of 180 bp) and subjected to another round of PCR. Genes larger than 1180 bp were synthesized in multiple steps of these assembly rounds, until the right size was achieved. These products were sequenced by Sanger sequencing, to check for mutations. The mutations were mutated back to their original sequence by means of the Quickchange protocol (Stratagene, La Jolla, CA, USA) as described by the manufacturer. Transformation in E. coli and subsequent sequencing of the plasmids was used to check the final product. Copy number determination with the KIR-MLPA calibrator The calibrator was added in a comparable number of molecules as present in an average human genome of 100 ng. Because the calibrator has a length of approximately 11 kb, there is relatively little DNA present in each sample. To prevent the DNA from sticking to the plastic storage tube, we used 400 ng/ml herring sperm DNA as a decoy. Also, LoBind tubes (Eppendorf, Nijmegen, The Netherlands) were used to prevent the DNA from sticking to the plastic. The calibrator served as a reference and was run in parallel to the test samples in every MLPA run as an extra sample. To minimize variation within one experiment, one calibrator was present for every 7 samples. Analysis was performed with GeneMarker software (Softgenetics). The amounts of amplification product that were measured for all calibrator samples within one run were averaged and used as one reference sample. The software calculated a ratio for the peak height or area of each test sample 32

35 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS against the corresponding (averaged) calibrator peak. The peak pattern of the calibrator represented 2 copies of each KIR gene, after normalization with the software. Therefore, a ratio of 1 means that 2 copies of this gene were present in the test sample. The control probes provided a quality control for each test, as each gene targeted by these probes is present twice in each individual. Not all probes in our MLPA gave a good approximation of the copy number. The probes that were used for copy number determination are listed in Supplementary Table 4. When multiple probe sets were used for copy number determination, an average was calculated of all ratios. We apply as a rule of thumb that an (average) ratio of represents 1 copy, represents 2 copies, represents 3 copies and >1.75 represents 4 or more copies. For border cases multiple runs were taken into account. 2 PCR-SSP The polymerase chain reaction with sequence-specific primers (PCR-SSP) was performed as previously described (Kunert et al. 2007; Uhrberg et al. 1997). Sanger sequencing Sequence analysis was performed according to the manufacturer s protocol of the BigDyeTerminator cycle sequencing kit on an ABI-3130XL (Applied Biosystems by Life Technologies). Short tandem repeat assay Primers were designed around the short tandem repeat region in intron 4, present in all intron 4 containing KIR genes (sense 5 -ccaaagagaactagagagaccgagaggc-3, 5 ccaaatagaac tagagagactgagaggc-3, 5 -ccaaagagagctagagagaccgagaggc-3, 5 -ccaaaagggaactagagagactga gaggc-3 and anti-sense 5 -tgtgtccttgtgtcctgttcataactttctgc-3, 5 -tgtgtccttgtgtcctgtccataac tttctgc-3, 5 -tgtgtccttctgtcttgctcataactttctgc-3, 5 -tgtgtccttctgtcttgttcataactttctgc-3, 5 -tgt gtccttgtgtcccgttcataactttctgc-3 ). One PCR reaction amplified all short tandem repeat regions (Supplementary Table 5): 5 µl of 20 ng/µl DNA was mixed with 12.6 µl of de-ionized water, 3 µl of SalsaBuffer (MRC Holland), 1.2 µl of 10 mm dntp, 1 µl of each primer mix with a 10 pmol/primer/ µl concentration. 0.4 µl of Taq Start Antibody (Clontech, Saint-Germain-en-Laye, France) and 0.8 µl of Salsa DNA polymerase (MRC Holland) were combined on ice and added to the DNA mix. The PCR program consisted of 2 min denaturation at 94 C, followed by 36 cycles of 15 sec denaturation at 95 C, 1 min annealing at 58 C and 5 min elongation at 68 C. The final step was an elongation at 68 C of 20 min. The amplified fragments were processed on a 454 FLX Roche Genome Sequencer, by the Shotgun Library method, followed by analysis with ReferenceMapper. The percentage of reads per donor per KIR gene was normalized against the framework gene KIR3DL2 to compensate for the difference in total gene copy numbers between donors. The percentage of reads that were found per KIR gene represented a relative number of copies present of that gene. A reference sequence was used to sort the reads, based upon the sequences within the STR region of the KIRs, available from the IPD KIR database: (Release 2.4.0, 15 April 2011). 33

36 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 Quantitative DNA PCR A quantitative PCR on DNA was designed to validate gene copy number determination by MLPA. Primers were designed for KIR2DL5 (sense 5 -gctggctccacatcctcgtt-3 and anti-sense 3 -cccaagacgagagcgactca-3 ) and for KIR2DS3 (sense 5 -ccaagatcagcaagtgtgggttt-3 and antisense 5 -cttggcaggaggtatgaactcaa-3 ). The qpcr was performed as described earlier (van Mirre et al. 2006). The relative ratio is a result of the absolute numbers from the KIR genes compared to the absolute numbers of a PCR on exon 8 of the Cytochrome b558 beta chain, gene (CYBB) (Sense: 5 -atgtcaaatatttaagcaagcctac-3 and anti-sense 5 -acttgtccatgatatagttagacac-3 ). The CYBB numbers are corrected for sex, as this gene is present on the X-chromosome. DISCUSSION In this study we have demonstrated the presence of extensive CNV at the KIR gene cluster. To determine KIR genotype and copy number, we designed a novel MLPA assay, using a synthetic calibrator as an internal reference. We found more gene copies for inhibiting than for activating KIRs. Similar gene copy variation was unexpectedly found in a number of the framework genes. This suggests either a more complex role for these supposed framework genes in the haplotype system than assumed until now, and/or crossing-over events, deletions and duplications in the KIR gene cluster occur at a far higher extent than previously thought. Because more variation exists in the middle framework gene block containing KIR2DL4 and KIR3DP1 than in the outer framework genes, KIR3DL2 and KIR3DL3, frequent duplications within the KIR gene cluster seem likely. Although it has been shown before that duplications, crossing-over events and deletions in the KIR gene cluster can occur (Martin et al. 2003; Traherne et al. 2010), the frequency of these events has never been evaluated in detail. To confirm our findings, next-generation sequencing was applied to validate our assay for KIR genotyping. No false-positive or false-negative genotyping was apparent when checked by 454 sequencing. An STR assay was used as an independent technique to measure CNV in the KIR gene cluster. This assay confirmed the copy numbers as found by the MLPA method for those KIRs that could be identified with the STR assay. The analysis of hereditary patterns of KIRs in several families supports the importance of copy number determination. Using the MLPA assay, we detected several independent duplication/deletion events in previously typed CEPH families and in one additional family. One of these duplications represents a previously unreported event. To our knowledge, there are no analytical techniques available to easily detect these events at the KIR gene cluster, explaining the fact that one of these events went undetected in these standard pedigrees until now. The MLPA assay is designed to detect the presence of entire genes, based on the current knowledge. Novel genes and hybrid genes are not specifically targeted when using this technique. From our 454-sequencing data and the probe analysis of our MLPA method, we know that the hybrid gene KIR2DL5A/3DP1 (Martin et al. 2003) (3DP1*004) is detected in our MLPA method as KIR3DP1. The hybrid gene KIR2DL3/2DP1 (Traherne et al. 2010) is not detected at all and the hybrid gene KIR2DL1/2DS1 (Traherne et al. 2010) is detected by both the KIR2DL1 and KIR2DS1 probes. 34

37 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS As the first two genes are most likely pseudogenes, only the latter causes concern for proper detection with the current MLPA assay. The hybrid gene KIR2DL1/2DS1 likely resembles KIR2DS1 the most, as fusion of the two sequences occurs before the first transcribed domain, which would result in a false-positive KIR2DL1 scoring by the KIR MLPA technique. Nevertheless, the frequency of this hybrid gene is expected to be very low, as was reported by Traherne et al. (Traherne et al. 2010). In addition to the 2 families in which this gene was originally found, the hybrid gene was detected in only 3 (non-caucasian) individuals of 1214 samples from various ethnic origins. To the best of our knowledge, no protein expression has been confirmed to date. Also, specific allelic variants were not separately targeted. There are several KIR genes with allelic variants that are not expressed. Specific targeting of these allelic variants could give an even more accurate KIR genotype. Although the MLPA assay allows further expansion of probe sets when required, we have to consider the additional relative value of such probes to detect rare allelic variants because of the low frequency in cohort studies. From an evolutionary perspective, one would expect an immunological effect of CNV at the KIR gene cluster, as shown for other immune receptors (Breunis et al. 2008). In fact, studies have been published that indicate a relationship between CNV and transcription levels for several KIRs (McErlean et al. 2010), and also the relevance of CNV in the KIR gene cluster in relation to certain diseases has been suggested for KIR3DL1 and KIR3DS1 (Alter et al. 2011; Espeli et al. 2010; Khakoo et al. 2004; Pelak et al. 2011). An NK cell resides in a delicate balance between activated and inhibited state. Expression seems to be regulated in a clonal fashion, in which each cell expresses its own selection of receptors. From this notion, we can hypothesize that it is also a selection of several receptors that decides cell fate during disease. Indeed, the presence of some activating KIR-HLA pairs in combination with a lack of inhibitory receptor pairs has been associated with an increased risk of developing certain autoimmune diseases like Diabetes type 1 and Ankylosing Spondylitis (Tajik et al. 2011; van der Slik et al. 2003). Also, the total number of activating KIR genes might influence NK cell effector function, as has been suggested for CMV reactivation during HSCT (Zaia et al. 2009), even though we still do not know the nature of the ligands for most of these activating KIRs. The use of our KIR MLPA creates interesting extra dimensions to the analysis of such data panels, adding information on the presence or absence of CNV. KIRs are mostly expressed by NK cells, which have a role in cancer surveillance and host defense against viral infections. A pivotal role for NK cells in human disease was shown for the first time in a rare severe immunodeficiency disease (SCID) patient who was able to clear CMV infection in the absence of T cells (Kuijpers et al. 2008). A similar situation can be encountered early after myeloablative HSCT, since donor-derived NK cells reconstitute earlier than T cells (Foley et al. 2012; Leung 2011; Ruggeri et al. 2002). In experimental models and in clinical trials, NK cells have received much attention in transplantation settings during cancer treatment. NK cells have been demonstrated not to cause Graft-versus-host-disease (GVHD), even when alloreactive across major histocompatibility barriers (Ruggeri et al. 2002; Ruggeri et al. 2007). Unlike T cells that carry the risk of causing lethal GVHD, NK cells are safe as a form of cellular immunotherapy in the allogeneic HSCT setting. Indeed, alloreactive NK cells are exploited as 2 35

38 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 mediators of graft-versus-leukemia (GVL) effects in T-cell-depleted haploidentical HSCT for acute myeloid leukemia following HSCT, or as potential cellular treatment in high-risk leukemia patients instead of HLA-identical HSCT (Leung 2011; Ruggeri et al. 2002). In summary, there is a clear need for simple but highly accurate tools to study KIR genotype-phenotype relations. We have therefore developed our current KIR MLPA assay for this purpose. Irrespective of the relative contribution of inhibitory or activating KIRs or lectins such as NKG2C or NKG2D in viral clearance and/or transplantation outcome (Foley et al. 2012; Gallez-Hawkins et al. 2011), realizing the potential role of CNV on KIR expression and functional activity as indicated in the present study, our current KIR MLPA assay will also enable us to relate disease severity to CNV in the KIR gene cluster in more detail on a larger scale. In addition, our MLPA assay for KIRs may be helpful in providing a rapid and efficient method for complete genotyping and at the same time provide data on CNV as a selection basis for donors in transplantation programs to improve the prognosis of graft survival and outcome of patients. 36

39 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS SUPPLEMENT A B 11 3DL DL3 2DS2 2DL DP DL1 11 3DP1 11 2DL4 1 3DL1 3DL1*024N 1 3DS1 3DS1*049N 1 2DL5 2DS3 1 2DS5 1 2DS1 2DS4wt 1 2DS4trunc 11 3DL2 C D 11 3DL3 1 2DL3 1 2DS2 1 2DL DP DL1 12 3DP1 12 2DL DL1 3DL1*024N 1 3DS1 3DS1*049N 1 2DL5 1 2DS3 2DS5 2DS1 2DS4wt 1 1 2DS4trunc 11 3DL B C B D A D A D 11 3DL3 11 3DL3 11 3DL3 11 3DL DL3 1 2DL3 1 2DL3 1 2DL3 2DS2 1 2DS2 1 2DS2 1 2DS2 2DL2 1 2DL2 1 2DL2 1 2DL DP DP DP DP DL DL DL DL1 11 3DP1 12 3DP1 12 3DP1 12 3DP1 11 2DL4 12 2DL4 12 2DL4 12 2DL DL DL1 1 3DL1 1 3DL1 3DL1*024N 3DL1*024N 3DL1*024N 3DL1*024N 3DS1 1 3DS DS DS1 3DS1*049N 3DS1*049N 3DS1*049N 3DS1*049N 2DL5 1 2DL DL DL5 2DS3 1 2DS3 1 2DS3 1 2DS3 2DS5 2DS5 1 2DS5 1 2DS5 2DS1 2DS1 1 2DS1 1 2DS1 2DS4wt 2DS4wt 2DS4wt 2DS4wt 1 1 2DS4trunc 1 1 2DS4trunc 1 2DS4trunc 1 2DS4trunc 11 3DL2 11 3DL2 11 3DL2 11 3DL2 Supplementary Figure 1. KIR gene pedigree analysis of a Centre d Etude du Polymorphisme Humaine family by KIR MLPA. The mother of family 1416 has a duplication of two KIR genes on allele D, which was transferred to three of her genotyped children. 37

40 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS A B 11 3DL3 2DL3 1 2DS2 1 2DL2 2DP1 1 2DL1 1 3DP1 1 2DL4 1 3DL1 3DL1*024N 3DS1 3DS1*049N 2DL5 2DS3 2DS5 1 2DS1 2DS4wt 1 2DS4trunc 11 3DL2 C D 11 3DL3 1 2DL3 1 2DS2 1 2DL DP DL1 11 3DP1 11 2DL4 1 3DL1 3DL1*024N 1 3DS1 3DS1*049N 2 2DL5 2 2DS3 2DS5 1 2DS1 2DS4wt 1 2DS4trunc 11 3DL A C B C A D B D 11 3DL3 11 3DL3 11 3DL3 11 3DL3 1 2DL3 1 2DL3 2DL3 2DL3 2DS2 1 2DS2 1 2DS DS2 2DL2 1 2DL2 1 2DL DL2 1 2DP1 1 2DP1 1 2DP1 1 2DP DL1 1 2DL DL1 1 2DL1 1 3DP1 11 3DP1 1 3DP1 11 3DP1 1 2DL4 11 2DL4 1 2DL4 11 2DL4 1 3DL DL1 3DL1 1 3DL1 3DL1*024N 3DL1*024N 3DL1*024N 3DL1*024N 3DS1 3DS1 1 3DS1 1 3DS1 3DS1*049N 3DS1*049N 3DS1*049N 3DS1*049N 2DL5 2DL5 2 2DL5 2 2DL5 2DS3 2DS3 2 2DS3 2 2DS3 2DS5 2DS5 2DS5 2DS5 1 2DS1 2DS DS1 1 2DS1 2DS4wt 2DS4wt 2DS4wt 2DS4wt 1 2DS4trunc 1 1 2DS4trunc 2DS4trunc 1 2DS4trunc 11 3DL2 11 3DL2 11 3DL2 11 3DL2 Supplementary Figure 2. KIR gene pedigree analysis of a Centre d Etude du Polymorphisme Humaine family by KIR MLPA. The father of family 1413 has only one copy of both KIR2DL4 and KIR3DP1, which can be traced back to some of his children. 38

41 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 Father A B 11 3DL3 2DL DS DL2 2DP1 2DL1 1 3DP1 1 2DL4 1 3DL1 3DS1 1 2DL5 2DS3 1 2DS5 1 2DS1 1 2DS4WT 2DS4trunc 11 3DL2 Mother C D 11 3DL3 1 2DL3 1 2DS2 1 2DL2 1 2DP1 1 2DL1 1 3DP1 1 2DL4 3DL1 1 3DS DL5 2DS DS DS1 2DS4WT 2DS4trunc 11 3DL2 Child 1 Child 2 Child 3 B D 11 3DL3 2DL DS DL2 2DP1 2DL1 3DP1 2DL4 3DL1 3DS DL5 2DS DS DS1 2DS4WT 2DS4trunc 11 3DL2 B D 11 3DL3 2DL DS DL2 2DP1 2DL1 3DP1 2DL4 3DL1 3DS DL5 2DS DS DS1 2DS4WT 2DS4trunc 11 3DL2 B D 11 3DL3 2DL DS DL2 2DP1 2DL1 3DP1 2DL4 3DL1 3DS DL5 2DS DS DS1 2DS4WT 2DS4trunc 11 3DL2 Supplementary Figure 3. KIR gene pedigree analysis of an Israeli family by KIR MLPA. Both parents carry only one copy of the framework genes KIR3DP1 and KIR2DL4, which results in a complete absence of those genes in all three children. 39

42 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 qpcr KIR2DL5 Relative ratio (KIR2DL5/CYBB) Control samples 1347 father 1347 child child 04 MLPA copy number qpcr KIR2DS3 Relative ratio (KIR2DS3/CYBB) Control family members 1347 father 1347 child child 04 MLPA copy number Supplementary Figure 4. Graphical representation of the quantitative PCR on DNA from Centre d Etude du Polymorphisme Humaine family 1347 and some control donors. (A) The relative product of KIR2DL5 compared to the occurrence of CYBB (corrected for sex). (B) The relative product of KIR2DS3 compared to the occurrence of CYBB (corrected for sex). 40

43 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS Supplementary Table 1. Overview probe sequences of the KIR MLPA technique. Probe Sequence (5-3 ) 2DL1 - #1 left GGGTTCCCTAAGGGTTGGACTACCCCATCGCTCTTCATGCTGGATCATTCACTCTGCATCCCAATGACAATG 2DL1 - #1 middle AGAAGAAAGTCTGGACACTCTCACCTATGATCACGATGTCCAGAGGGTCACTGGGAGCTGACAC 2DL1 - #1 right CTGATAGGGGGAGTGAGTAACAGAACCGTAGTCTAGATTGGATCTTGCTGGCAC 2DL1 - #2 left GGGTTCCCTAAGGGTTGGACATCCTGTGCGCTGCTGAGCTGAGCTCG 2DL1 - #2 right GTCGCGGCTGCCTGTCTGCTCCGGCAGTCTAGATTGGATCTTGCTGGCAC 2DL2 - #1 left GGGTTCCCTAAGGGTTGGACTGACCTTGGGCCCTGCAGAGAACCTACA 2DL2 - #1 right TTCATGGGCCTCCCCCTCCCTGGATGTCTAGATTGGATCTTGCTGGCAC 2DL2 - #2 left GGGTTCCCTAAGGGTTGGACATGTCCTATGATCCTAGAGCCTTAGCTGAGGAGCTTCCTGCTGATGATGGAGAT 2DL2 - #2 right AAGCATGGACAGATGCAGAGAGAAGACGAAGCTTGGGTGTGAGGGAGGTCTAGATTGGATCTTGCTGGCAC 2DL3 - #1 left GGGTTCCCTAAGGGTTGGACYRCACAGTTGRATCACTGCGTTTTCACACAGAGAAAAATCACTCRCCCTT 2DL3 - #1 middle CTCAGAGGCCCAAGACACCCCCAACAGATATCATCGTGTACACGGAACTTCCAAATGCTGAGCCCT 2DL3 - #1 right GATCCAAAGTTGTCTCCTGCCCATGAGCACCACAGTCAGGCCTTGTCTAGATTGGATCTTGCTGGCAC 2DL3 - #2 left GGGTTCCCTAAGGGTTGGACTGCTGCCTTGGGCCAGGGACCATCCTGTCTGTGAGGAACACACACCTGAGTGCTC 2DL3 - #2 middle CCATCCTGCTTCCCCACATGGCCCTGAGCTCTCTGGCCTCTGCTTCGTGAGACTTACTTTTTTTGTTGC 2DL3 - #2 right AGCACCAGCGATGAAGGAGAAAGAAGAGGAGGAGGATGAAGAGGATGTCTAGATTGGATCTTGCTGGCAC 2DL4 - #1 left GGGTTCCCTAAGGGTTGGACCTGCTTCAGAACATGGCTCTCTGCTGGGGAGACACCCAA 2DL4 - #1 middle TCTGCAGGCCCATAGTGTAACCCTGGTGCTCCTTCCCTTCCAGGACTCACCAAG 2DL4 - #1 right ACATGCCAGGATGATGACCGTGGGTGACATGGAGTCTAGATTGGATCTTGCTGGCAC 2DL4 - #2 left GGGTTCCCTAAGGGTTGGACCCGCTTAGAAAGAAGAAATGGGGAGAATCTTCTGAGCACAGGGAGGGAGGGGC 2DL4 - #2 middle AGCTCAACATACTCCTCTCTGAGGCGGCATCTCCTTCTCCCCAAGGTGGTCAGGACAAGCCCTTCTGC 2DL4 - #2 right TCTGCCTGGCCCAGCGCTGTGGTGCCTCAAGGAGGACACGTGACTCTTCGGTGGTCTAGATTGGATCTTGCTGGCAC 2DL5 - #1 left GGGTTCCCTAAGGGTTGGACTCAGGTGTGAGGGGAGCTGTGACAAGGAAGAACCTCC 2DL5 - #1 middle CTGAGGAAACTGCCTCTTCTTCCAGGTCTATT 2 41

44 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 Supplementary Table 1. Continued Probe Sequence (5-3 ) 2DL5 - #1 right TGGGAAACCTTCACTCTCAGCCCAGCCGGGTCTAGATTGGATCTTGCTGGCGC 2DL5 - #2 left GGGTTCCCTAAGGGTTGGACCGACCTACACATGCTTYRGCTCTCTCCATGACTC 2DL5 - #2 right ACCCTATGAGTGGTCAGACCCGAGTGACCCGTCTAGATTGGATCTTGCTGGCAC 2DL5b left GGGTTCCCTAAGGGTTGGACATGTTAGCACAGATTTTAGGCATCTCGTGTTCGGATAAAAATACATGAAAAGTCTTTCAC 2DL5b middle GTTAGCACAGATTTTAGGCATCTTGTGTTCGGGAGGTTGGATCTGAGACGTGTTGTGAGTTGGTCATAGTGAAGGACGT 2DL5b right GAGGTGCCAATTCTAGTGAGAACAATTTCCAGGAAGCCGTGTTCCGGTCTAGATTGGATCTTGCTGGCAC 2DP1 - #1 left GGGTTCCCTAAGGGTTGGACCCAAGGTGGTCAGGACAAGCCCTTGCTGTCTGCCTGGCCCAGCTC 2DP1 - #1 right TGTGGTGCCTCCAGGACATGTGATTCTTCGGTGTCTAGATTGGATCTTGCTGGCGC 2DP1 - #2 left GGGTTCCCTAAGGGTTGGACCAGGGACCTACAGATGCTACGGTTCTGTTACTCACTCCCC 2DP1 - #2 right CATCAGTTGTCAGCTCCCAGTGACCCTCTGGACATCGTCATGTCTAGATTGGATCTTGCTGGCAC 2DS1 - #1 left GGGTTCCCTAAGGGTTGGACACAGGGCCCATGAAAAGGCTGTTCCAGAATATTATGTTGTAGAGCTCAGGGACAGGCA 2DS1 - #1 middle CCCCATCTTCCTTTTACAGACTGAAGTTGTTAAACCCAAGATAAGAATGACACTGAAGAATCACATA 2DS1 - #1 right TCCTGGAGGCACCACAGGGCTTGGCCAGTCTAGATTGGATCTTGCTGGCAC 2DS1 - #2 left GGGTTCCCTAAGGGTTGGACTAGGAGACCGTGGAAAAGGCAATTCCCGA 2DS1 - #2 middle CCCACTGGTGAAATGTGGTGCTGATTTT 2DS1 - #2 right GACACTAAGTGGATGAAGCAGATGGATATAAGCGTCTAGATTGGATCTTGCTGGCAC 2DS2 left GGGTTCCCTAAGGGTTGGACCGGCCGAGCACCCCAGGGTCCTCTCTTCCC 2DS2 right AGTTTATGAGAGACTCCCTGACAGGACGTCTAGATTGGATCTTGCTGGCAC 2DS3 - #1 left GGGTTCCCTACGGGTTGGACCATCACGATGTCCAGAGGGTCACTGGGAGCTGAA 2DS3 - #1 right AACTGATAGGGGGAGTGAGGAACAGAGACCGTCTAGATTGGATCTTGCTGGCAC 2DS3 - #2 left GGGTTCCCTAAGGGTTGGACCGAAAGAGCCGAAGCATCTGTAGGTTCCTCCT 2DS3 - #2 right TGGGTGGCAGGGCCCAGAGGAAAGTCGGCCTGGAATGTCTAGATTGGATCTTGCTGGCAC 2DS4 - All left GGGTTCCCTAAGGGTTGGACTCCCCTCTCTGTGCAGAAGGAAGTGCTCAAACATGACATCC 42

45 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2DS4 - All right GACCAACATTGCAGGATGACTGTCTCTTCTGATTTCACCAGGTGACCTGGGAGTCTAGATTGGATCTTGCTGGCAC 2DS4 - WT/trunc left GGGTTCCCTAAGGGTTGGACCTTGGGCCCAGAGGAAAGTCRGCCTGGAATGTTCCGTKGAT 2DS4 - trunc middle GCTGCGCACTGCAGGGAGCCTACGTTCATGGGCCTCCCCYTCCCTGGATAG 2DS4 - trunc right ATGGAGCTGCAGGACAAGGTCACAGTCTAGATTGGATCTTGCTGGCAC 2DS4 - WT middle GCTGCGCACTGCAGGGAGCCTACGTTCATGGGCCTCCCCYTCCCTGGATAGATGGTAC 2DS4 - WT right CATGTCATAGGAGCTCCGGGAGCTGCAGGACAAGGWCACATTCTCTCTCTAGATTGGATCTTGCTGGCAC 2DS5 left GGGTTCCCTACGGGTTGGACCGAGTAAACCGGAAAATTTTCATCTGCACAGAGAGGGGACGTTTAAC 2DS5 middle CACACTTTGCCCCTCATTGGAGAGCACATTGATGGGGTCTCCAAGGG 2DS5 right CAACTTCTCCCTCGGTCGCATGACACAAGACCTGGCATAGCGAATACGTCTAGATTGGATCTTGCTGGCAC 3DL1 - #1 left GGGTTCCCTAAGGGTTGGACCCCTCAHGCCTCGYTGGACA 3DL1 - #1 middle GATCCATGATGGGGTCTCCAAGGCCAATTTCTCCATCGGTCCCATGATGCT 3DL1 - #2 left GGGTTCCCTAAGGGTTGGACTCAGCTCAGGTATGAGGGGAGCTATGACAAGGAAGAACCT 3DL1 - #2 middle CCCTGAGGAAACTGCCTCTTCTCCTTCCAGGTCC 3DL1 - #2 right ATATGAGAAACCTTCTCTCTCAGCCCAGCCGGGTCTAGATTGGATCTTGCTGGCAC 3DL1/S1 right TGCCCTTGCAGGGACCTACAGATGCTACGGTTCTGGTCTAGATTGGATCTTGCTGGCAC 3DL2 - #1 left GGGTTCCCTAAGGGTTGGAATCCACCCTAAGGTTTGGGGAKGGACTCACCCATGA 3DL2 - #1 right GTGGCCAGGCCCCCTGCAGCAAGAAGAACCCTGTCTAGATTGGATCTTGCTGGCGC 3DL2 - #2 left GGGTTCCCTAAGGGTTGGACCATGAAGCTCCTCAGCTATGGCTCTAGGATCATAAGACATGGGACAGACA 3DL2 - #2 middle CGGGTTTTCCTCACCTGTGACAGAAACAAGCAGTGGGTCACTTGAGTTTGACCACACGCA 3DL2 - #2 right GGGCAGGGCACGGAAAGAGCCGAAGCATCTGTAGTTCCCTCGTCTAGATTGGATCTTGCTGGCAC 3DL3 - #1 left GGGTTCCCTACGGGTTGGATAGATGCTTCGGCTCTTTCCGTGCCCTGCCC 3DL3 - #1 right CAYGCGTGGTCAGACCCGAGTGACCCGTCTAGATTGGATCTTGCTGGCAC 3DL3 - #2 left GGGTTCCCTAAGGGTTGGACAAGGGTGAGAGGCAGGTCTGTATTCTCTCACCTA 3DL3 - #2 middle CGACCACGATGTCCAGAGGGTCACTGGGAGCYGACAACTCATAGGGTA 3DL3 - #2 right AGTGAGTGACAGAACCAAAGCATCTGTAGTCTAGATTGGATCTTGCTGGCAC 2 43

46 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 Supplementary Table 1. Continued Probe Sequence (5-3 ) 3DP1 - #1 left GGGTTCCCTAAGGGTTGGACACCATGATCACCAGGGGGTTGCTGGGTGCTGACCACCCAGTGAGGA 3DP1 - #1 right AGTGTGGGTGTGAACCCCGACATCTGTAGGTCCCTGTCTAGATTGGATCTTGCTGGCGC 3DP1 - #2 left GGGTTCCCTAAGGGTTGGACGCTGAGGCCTGGAAAGGAATAGAGGGAGGGAGTGCCACATC 3DP1 - #2 right CTCCTCTCTAAGGTGGCGCCTCCTTCTCCCCCAGGTGTCTAGATTGGATCTTGCTGGCAC 3DS1 - #1 left GGGTTCCCTAAGGGTTGGACTTGTTCATCAGAATCCTGGAGAGAGGGAAATGCTGAGTGAGGGAGGGTGCTCACATTTTTC 3DS1 - #1 middle AGGACTCTTTGGGAATAACACTAGCCACGAGGCTGGGCCGAGGAGCACCTACCTCGCTGTTCAC 3DS1 - #1 right TTCTGTTCCCTGCAGGCTCTTGGTCCATTACAGCAGCATCTGTAGGAGACGTCTAGATTGGATCTTGCTGGCGC 3DS1 - #2 left GGGTTCCCTAAGGGTTGGACACTTCTTTCTGCACAAAGAGTGGATCTCTAAGG 3DS1 - #2 middle ACCCCTCACGCCTCGTTGGACAGATCCATGATGGGGTCTCCAAGGCCAATTTCTCCATCGGTTCCATGATGCG 44

47 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS Supplementary Table 2. Overview control probe sequences of the KIR MLPA technique. KIR Control Probe mix Probe Sequence (5-3 ) Control 1 - left GGGTTCCCTAAGGGTTGGACCAGCCAGCAGCAGCCCCAAGCTGATAAGATTAATCTAAAGAGCAAATTATGGT Control 1 - middle GTAATTTCCTATGCTGAAACTTTGTAGTTAATTTTTTAAAAAGGTTTCATTTTCCTATTGGTCTGATTT Control 1 - right CACAGGAACATTTTACCTGTTTGTGAGGCATTTTTTCTCCTGGTCTAGATTGGATCTTGCTGGCAC Control 2 - left GGGTTCCCTAAGGGTTGGATAGAGGGCTCCAGGTTATCCGGGCTC Control 2 - right TGCTTCTGCCGCCGCCGTCGGGGGTCTAGATTGGATCTTGCTGGCAC Control 3 - left GGGTTCCCTAAGGGTTGGATTCAGGAGTGATACTTCACAGATCCTGGAGGAA Control 3 - right AACATCCCAGTCCTTAAGGCCAAACTGAGTCTAGATTGGATCTTGCTGGCAC Control 4 - left GGGTTCCCTAAGGGTTGGACTGCGTGAGACCAAGCGCCGTCATGAGACCCGACTGGTG Control 4 - right GAGATTGACAATGGGAAGCAGCGTGAGTTTGAGAGTCTAGATTGGATCTTGCTGGCAC Control 8 - left GGGTTCCCTAAGGGTTGGACCTGCTCCCAGTAGGGTCAGCATCTGGACCCCAGGCTGA Control 8 - middle GAGTCAGGCTCTGATTCCAGATCTAGCCTCCATCATGAAGAAGCTCTTGACCAAGTATG Control 8 - right ACAACCTCTTTGAGACGTCCTTTCCCTACTCCATGTCTAGATTGGATCTTGCTGGCAC Control 9 - left GGGTTCCCTAAGGGTTGGACGCAGGACAGAAGGAGCAAGCTGTGGAATGGTATAAGAAAG Control 9 - middle GTATTGAAGAACTGGAAGAAGGAATAGCTGTTATAGTTACAGGACAAGGTAAGATTGTAT Control 9 - right TTGTTTATAGCCATCCCAAATTATGATATATTCACACTCTAGATTGGATCTTGCTGGCAC Control 10 - left GGGTTCCCTAAGGGTTGGACCTTCCCCATTGGTTTGTTATTGCAGATGAAGTGGAAAGGGAAGGACCTCTTTGATTTG Control 10 - middle GTGTGCCGGACTCTGGGGCTCCGAGAAACCTGGTTCTTTGGACTGCAGTACACAATCAAGGACACAGTG Control 10 - right GCCTGGCTCAAAATGGACAAGAAGGTTGGGCTAGAACTCGATGAAACTGGTGCTCTAGATTGGATCTTGCTGGCAC 2 45

48 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 Supplementary Table 3. KIR MLPA technique probe distribution over 3 mixes. Mix 1 Mix 2 Mix 3 KIR controls Probe Product Length Probe Product Length Probe Product Length Probe Target Product Length 2DL1 - #1 left 190 2DL1 - #2 left 97 2DL1 - #2 left 97 Control 1 - left IL DL1 - #1 middle 2DL1 - #2 right 2DL1 - #2 right Control 1 - middle 2DL1 - #1 right 2DL2 - #2 left 145 2DL3 - #2 left 214 Control 1 - right 2DL2 - #1 left 96 2DL2 - #2 right 2DL3 - #2 middle Control 2 - left FGF3 92 2DL2 - #1 right 2DL3 - #1 left 204 2DL3 - #2 right Control 2 - right 2DL3 - #1 left 204 2DL3 - #1 middle 2DL4 - #1 left 170 Control 3 - left BCAS DL3 - #1 middle 2DL3 - #1 right 2DL4 - #1 middle Control 3 - right 2DL3 - #1 right 2DL3 - #2 left 214 2DL4 - #1 right Control 4 - left LMNA 116 2DL4 - #1 left 170 2DL3 - #2 middle 2DL5b left 229 Control 4 - right 2DL4 - #1 middle 2DL3 - #2 right 2DL5b middle Control 8 - left GALT 175 2DL4 - #1 right 2DL4 - #2 left 218 2DL5b right Control 8 - middle 2DL5 - #1 left 142 2DL4 - #2 middle 2DP1 - #1 left 121 Control 8 - right 2DL5 - #1 middle 2DL4 - #2 right 2DP1 - #1 right Control 9 - left SPG DL5 - #1 right 2DL5 - #2 left 108 2DS1 - #2 left 134 Control 9 - middle 2DP1 - #1 left 121 2DL5 - #2 right 2DS1 - #2 middle Control 9 - right 2DP1 - #1 right 2DL5b left 229 2DS1 - #2 right Control 10 - left NF DS1 - #1 left 195 2DL5b middle 2DS3 - #1 left 108 Control 10 - middle 2DS1 - #1 middle 2DL5b right 2DS3 - #1 right Control 10 - right 2DS1 - #1 right 2DP1 - #2 left 125 2DS5 left 185 2DS3 - #1 left 108 2DP1 - #2 right 2DS5 middle 2DS3 - #1 right 2DS1 - #2 left 134 2DS5 right 2DS4 - All left 137 2DS1 - #2 middle 3DL1 - #1 left

49 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 2DS4 - All right 2DS1 - #2 right 3DL1 - #1 middle 2DS4 - WT/trunc left 160 2DS2 left 101 3DL1/S1 right 2DS4 - trunc middle 2DS2 right 3DL3 - #1 left 100 2DS4 - trunc right 2DS3 - #2 left 112 3DL3 - #1 right 2DS5 left 185 2DS3 - #2 right 3DP1 - #2 left 121 2DS5 middle 2DS4 - WT/trunc left 190 3DP1 - #2 right 2DS5 right 2DS4 - WT middle 3DL1 - #1 left 150 2DS4 - WT right 3DL1 - #1 middle 3DL1 - #2 left 150 3DL1/S1 right 3DL1 - #2 middle 3DL2 - #1 left 111 3DL1 - #2 right 3DL2 - #1 right 3DL2 - #2 left 195 3DL3 - #1 left 100 3DL2 - #2 middle 3DL3 - #1 right 3DL2 - #2 right 3DP1 - #1 left 125 3DL3 - #2 left 154 3DP1 - #1 right 3DL3 - #2 middle 3DS1 - #1 left 219 3DL3 - #2 right 3DS1 - #1 middle 3DP1 - #2 left 121 3DS1 - #1 right 3DP1 - #2 right 3DS1 - #2 left 185 3DS1 - #2 middle 3DL1/S1 right 47

50 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 Supplementary Table 4. Copy number determination by KIR MLPA technique. Gene detected Determining CNV 2DL1 Average of all probes 2DL2 Average of all probes 2DL3 Average of #2 in Mix 2 and 3 2DL4 Average of all probes 2DL5 * 2DL5 #2 in Mix 2 2DS1 Average of #2 multiplied by 2 in Mix 2 and 3 2DS2 2DS2 in Mix 2 2DS3 Average 2DS3 #2 in Mix 2 and 2DS3 #1 in Mix 3 2DS4 All 2DS4 All and independently copies of WT + trunc 2DS4 WT 2DS4 WT 2DS4 truncated 2DS4 trunc multiplied by 2 2DS5 Average of all probes 2DP1 Average of all probes 3DL1 Average of 3DL1 #1 in Mix 1 and 3 3DL2 Average of all probes 3DL3 Average of all probes 3DS1 Average of all probes 3DP1 3DP1 #1 in Mix1 * Detects both 2DL5A and B 48

51 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS Supplementary Table 5. A sequence alignment of the short tandem repeat region in intron 4 of most KIRs. 3DS1_450 -CCAAATAGAACTAGAGAGACTGAGAGGCAGAGAAAGACAAGGAGACGGAGAGAGAGAGATGATAGATGGATAGATA GACGTAGATAGATGATAAATA---GG 3DL1_422 -CCAAATAGAACTAGAGAGACTGAGAGGCAGAGAAAGACAAGGAGACGGAGAGAGAGAGATGATAGATGGATAGATA GACGTAGATAGATGATAAATA---GG 3DL2_447 -CCAAATAGAACTAGAGAGACTGAGAGGCAGAGAAAGACAAGGAGATGGAGAGAGACAGATGATAGATGGATAGAYA GATATAGATAGATGATAAATA---GG 2DL1_405 -CCAAAGAGAACTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGATGGATAGATA TAGATAGATGATAAATA---GG 2DS1_405 -CCAAAGAGAACTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGATGGATAGATA TAGATAGATGATAAATA---GG 2DS5_401 -CCAAAGAGAACTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGATGGATAGATA TAGATAGATGATAAATA---GG 2DS4_424 -CCAAAGAGAACTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGATGGATAGATACAGATAGATGATGGATAGATATAGATAGATGATATATA---GG 3DP1_401 -CCAAAGAGAACTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGATGGATAGATA TAGATAGATGATAAATA---GG 2DL2_391 -CCAAAGAGAGCTAGAGAGACCGAGAGGCAGAGCAATACA GATGATAGATGGATAGATA TAGATAGATGATAAATA---GG 2DL3_391 -CCAAAGAGARCTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGATGGATAGATA TAGATAGATGATAAATA---GG 2DS2_395 -CCAAAGAGAACTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGATGGATAGATA TAGATAGATGATAAATA---GG 2DS3_392 -CCAAAGAGAACTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGATGGATAGATA TAGATAGATGATAAATA---GG 2DP1_395 -CCAAAGAGAACTAGAGAGACCGAGAGGCAGAGCAAGACA GATGATAGAWGGWTAGATA TAGATAGATGATAAATA---GG 3DL3_416 CCAAAAGGGAACTAGAGAGACTGAGAGGCAGAGAAAGACAAGGAGATGGAGAGAGACAGATGATAGATGGATAGATA GATATAGATAGATGAAAGATAAAAGG * *** ** ********** *********** ** *** ********* ** **** * *********** * *** ** 3DS1_450 TAGATGATAGATAATGGA----TTGGTTATAGATACATAGATGATGACTGAT AGATGATACATAGAGATGATGATGATGACGATGATGATGATAGACACATAGA 3DL1_422 TAGATGATAGATAATGGA----TTGGTTATAGATACATAGATGATGACTGAT AGATGATACATAGAGATGA CGATGATGATGATAGACACATAGA 3DL2_447 TAGATGATAGATAATGGA----TAGGTTATAGATACATAGATGATGATTGAT AGATGATACATAGAGATGATGATGATGATGATGATGAAGATAGATAGATAGA 2DL1_405 TAGATGATAGATAATAG GTTATAGATACATAGATGATGATTGATTGATTCATT----AATAGATGAGACATAGAGATGA TGATGATGAAGACASATAGATAGA 2DS1_405 TAGATGATAGATAATAG GTTATAGATACATAGATGATGATTGATTGATTCATT----AATAGATGAGACATAGAGATGA TGATGATGAAGACAGATAGATAGA 2DS5_401 TAGATGATAGATAATAG GTTATAGATACATAGATGATGATTGATTGATTCATT----AATAGATGAGACATAGAGATGA TGATGATGAAGACAGATAGA---- 2DS4_424 TAGATGATAGATAATAG GTTATAGATACATAGATGATGATTGATTGATTCATT----AATAGATGATACATAGAGATGA TGATGATGAAGATAGATGGATAGA 3DP1_401 TAGATGATAGATAATAG GTTATAGATACATAGATGATGATCGATTCATTCATTGATTAATCGATGATACATAGAGATGA TGAAGATGAAGATA----GATAGA 2DL2_391 TAGATGATAGATAATAG GTTAAAGATACATAGATGATGATTGATTGATTCATT----AATAGATAATACATAGAGATGA TGATGATGAAGACA----GATA-- 2DL3_391 TAGATGATAGATAATAG GTTAAAGATACATAGATGATGATTGATTGATTCATT----AATAGATAATACATAGAGATGA TGATGATGAAGACA----GATA-- 2DS2_395 TAGATGATAGATAATAG GTTAAAGATACATAGATGATGATTGATTGATTCATT----AATAGATAATACATAGAGATGA TGATGATGAAGACA----GATAGA 2DS3_392 TAGATGATAGATACTAG GTTATAGATACATAGATGATGATTGATTGATTCATT----AATAGATGAGACGTAGAGATGA TGATGA---AGACA----GATAGA 2DP1_395 TAGATGATAGATAATAG GTTAAAGATACATAGATGATGATTGATTGATTCATT----AATAGATGAGACATAGAGATGA TGATGATGAAGACA----GATAGA 3DL3_416 TAGATGATAGATAATAGAGAGACAGGTGATAGACAAATAGATGATGAATGACTGAT AGATGATATAGATAGACAA GTAGAAAGACAGACAGAT-GATATA ************* * * ** * *** * *********** ** *** * * * *** * * ** * * 3DS1_ TATATACATAGAT-----GATACATAAAT AGAGACAGAGAGGCAGACA----GAGAGGTAATAGAGAGAGAGATAGATG-ATACATA--TATAGATAATAGATGATTGAT 3DL1_ TATATACATAGAT-----GATACATAAAT AGAGACAGAGAGGCAGACA----GAGAGGTAATAGAGAGAGAGATAGATG-ATACATA--TATAGATAATAGATGATTGAT 3DL2_447 AGACACATATATAAATATATAGATACATAGATGATACATAGAGACTGACAGGCAGACA----GAGAGGTAATAGAGAGAGAGAGAGATG-ATACATAGATACAGATAATACATAGATGAT 2DL1_ T-----AATACATA GAGATACAGAGGCAGACATAGAG--AAATCATAGAGAGAGAGA--GATG-ATACATAGATATAGATAATAGATGATTGAT 2DS1_ T-----AATACATA GAGATACAGAGGCAGACATAGAG--AAATCATAGAGAGAGAGA--GATG-ATACATAGATATAGATAATAGATGATTGAT 2DS5_ T-----AATACATA GAGATAGAGAGGCAGACATAGAG--AAATCATAGAGAGAGAGA--GATG-ATACACAGATATAGATAATAGATGATTGAT 2DS4_ T-----AATACATA GAGATAGAGAGGAAGACAAAGAGAGAAATAATAGAGAGAGAGA--GATG-ATACATATATATAGATAATAGATGATTGAC 3DP1_ T-----AATACATA GAGATAGAGAGGCAGACAAAGAG--AAATCATAGAGAGAGAGA--GATG-ATACATAGATATAGATAATAGATGATTTTT 2DL2_ ATACGTA CAGATAGAGAGGCAGACA----G--AAATCATAGAGAGAGAG----ATG-ATACATACATATAAATAACAGATGATTGAT 2DL3_ ATACGTA CAGATAGAGAGGCAGACA----G--AAATCATAGAGAGAGAG----ATG-ATACATACATATAAATAACAGATGATTGAT 2DS2_ T-----AATACGTA CAGATAGAGAGGCAGACA----G--AAATCATAGAGAGAGAG----ATG-ATACATACATATAAATAACAGATGATTGAT 2DS3_ T-----AATACATA GAGATAGAGAGGCAGACA----G--AAGTCATAGAGAGAGAG----ATG-ATACATAGATATAGATAACAGATGATTGAT 2DP1_ T-----AATACATA GAGATAGAGAGGCAGACA----G--AAGTCATAGAGAGAGAG----ATG-ATACATAGATATAGATAACAGATGATTGAT 3DL3_ TA----AATAGATATAG AGAGATAGAAAGACAGATAAAC-ACATGATGATAGATGGATAG----ATGCATACATACATACAT-TGATTGATAGATGAT *** ** *** * ** *** * * ***** ** ** *** ***** * ** * * * ** * 2 49

52 EXTENSIVE VARIATION IN GENE COPY NUMBER AT THE HUMAN KIR LOCUS 2 Supplementary Table 5. Continued 3DS1_450 GGATAGATAGACAGATAGACAATTGATAGAGAGATAGATAAGTGATACATAAATATAGATGATAGATA----ATTTGTAGATAGACACAAAATAGATA AAT 3DL1_422 GGATAGATAGACAGAYAGACAATTGATAGAGAGATAGATAAGTGATACATAAATATAGATGATAGATA----ATTTGTAGATAGACACAAAATAGATA AAT 3DL2_447 TGATGGATAGACAGATAGACAATTGATAGA TAAATGATACATAGATATAGATGACAGATA----ATTTGTAGATAGACACAAAATAGATAGATAGA TAAT 2DL1_405 GGAT AGATAGACAATTGATGGA----TAAATAGATGATATATAGATATAGATGACAGGTAGAGAATTTGTAGATAGGCACCGAATAGATAAATAGATAGATCGATAGATAAT 2DS1_405 GGAT AGATAGACAATTGATGGA----TAAATAGATGATATATAGATATAGATGACAGGTAGAGAATTTGTAGATAGGCACCGAATAGATAAATAGATAGATCGATAGATAAT 2DS5_401 GGAT AGATAGAAAATTGATAGA----TAAATAGATGATATATAGATATAGATGACAGGTAGAGAATTTGTAGATAGGCACGGAATAGATAAATAGATAGATCGATAGATAAT 2DS4_424 GGAT AG----ACAATTGATAGA----TAAATAGATGATATATAGATATAGATGACAGGTAGAGAATTTGTAGATAGGCACCGAATAGATAAATAGATGGATTGATAGATAAT 3DP1_401 GGAT AG----ACAATTGATAGA----TAAATAGATTATATATAGATATAGATGACAGGTAGAGAATTTGTAGATAGGCACCAGATAGATAAATAGATATATCGATAGATAAT 2DL2_391 GGAT AGATAGACAACTGATAGA----TACATAGATGATATATAGATATAGATGACAGGTAGAGAATTTGTAGATAGGCACCGAATAGATAAATAGATAGATCGACAGATAAT 2DL3_391 GGAT AGATAGACAAGTGATAGA----TACATAGATGATATATAGATATAGATGACAGGTAGAGAATTTGTAGATAGRCACCGAATAGATAAATAGATAGATCGACAGATAAT 2DS2_395 GGAT AGATAGACAACTGATAGA----TACATAGATGATATATAGATATAGATGACAGGTAGAGAATTTGTAGATAGGCACCGAATAGATAAATAGATAGATCGACAGATAAT 2DS3_392 GGAT AGATAGACAAGTGATAGA----TACATAGATGATATATAGATATAGATGACAAGTAGAGAATTTGTAGATAGGCACCGAATAGATAAATAGATAGATCAACAGATAAT 2DP1_395 GGAT AGATAGACAAGTGATAGA----TACATAGATGATATATAGAYATAGATGACAGGTAGAGAATTTGTAGATAGGCACCGAATAGATAAATAGATAGATCGAYAGATAAT 3DL3_416 AGATA-----ACAGAGAGATAGGTCATAGA----TACACAGATGATG ATAGATGATAGATAC---ATACATAGATAAATGATAGATCGATCAATAGA TAGT *** ** * * * ** ** * * ** ******** * ** ** ****** ** *** * * 3DS1_450 AGATAGATCGATAGATAATAGATAGAAATGTGCAGAAAGTTATGAACAAGACAGAA 450 3DL1_422 AGATAGA AATGTGCAGAAAGTTATGAACAAGACAGAA 422 3DL2_447 AGATAGA AATATGCAGAAAGTTATGAACAAGACAGAA 447 2DL1_405 AGATAGA AATATGCAGAAAGTTATGRACAGGACACAA 405 2DS1_405 AGATAGA AATATGCAGAAAGTTATGAACAGGACACAA 405 2DS5_401 AGATAGA AATATGCAGAAAGTTATGAACAGGACACAA 401 2DS4_424 AGATAGA AATATGCAGAAAGTTATGAACGGGACACAA 424 3DP1_401 AGATAGA AATATGCAGAAAGTTATGAACAGGACACAA 401 2DL2_391 AGATAGA AATATGCAGAAAGTTATGAACAGGACACAA 391 2DL3_391 AGATAGA AATATGCAGAAAGTTATGAACAGGACACAA 391 2DS2_395 AGATAGA AATATGCAGAAAGTTATGAACAGGACACAA 395 2DS3_392 AGATAGA AATATGCAGAAAGTTATGAACAGGACACAA 392 2DP1_395 AGATAGA AATATGCAGAAAGTTATGAACAGGACACAA 395 3DL3_416 AGATAGA AATATGCAGAAAGTTATGAGCAAGACAGAA 416 ******* *** ************** * **** ** 50

53 Supplementary Table 6. KIR gene copy number distribution in a cohort of healthy individuals. Copies KIR Gene >3 2DL1 4,2% 21,7% 62,5% 11,7% 0,0% 2DL2 46,7% 45,8% 7,5% 0,0% 0,0% 2DL3 12,5% 51,7% 35,0% 0,8% 0,0% 2DL4 0,0% 3,3% 88,3% 8,3% 0,0% 2DL5 43,7% 39,5% 16,0% 0,8% 0,0% 2DS1 62,5% 31,7% 5,8% 0,0% 0,0% 2DS2 46,2% 49,6% 4,2% 0,0% 0,0% 2DS3 68,3% 26,7% 5,0% 0,0% 0,0% 2DS4all 3,4% 26,9% 68,1% 1,7% 0,0% 2DS4wt 60,8% 32,5% 6,7% 0,0% 0,0% 2DS4trunc 18,3% 50,0% 30,8% 0,8% 0,0% 2DS5 66,7% 32,5% 0,8% 0,0% 0,0% 2DP1 4,2% 24,2% 58,3% 13,3% 0,0% 3DL1 3,3% 34,2% 59,2% 3,3% 0,0% 3DL2 0,0% 0,0% 99,2% 0,8% 0,0% 3DL3 0,0% 0,0% 99,2% 0,8% 0,0% 3DS1 58,3% 17,5% 20,0% 4,2% 0,0% 3DP1 0,0% 15,0% 84,2% 0,8% 0,0% n=120

54

55 3 NOVEL INSIGHTS IN THE GENOMIC ORGANIZATION AND HOT-SPOTS OF RECOMBINATION IN THE HUMAN KIR LOCUS THROUGH ANALYSIS OF INTERGENIC REGIONS S Vendelbosch, M de Boer, K van Leeuwen, F Pourfarzad, J Geissler, TK van den Berg, TW Kuijpers Published in: Genes and Immunity 2015,

56 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 ABSTRACT The Killer Immunoglobulin-like Receptor (KIR) proteins constitute a family of highly homologous surface receptors involved in the regulation of the innate cytotoxicity of Natural Killer (NK) cells. Within the human genome seventeen KIR genes are present, many of which show large variation across the population due to the high number of allelic variants and copy number variation (CNV). KIR genotyping and CNV determination was used to map the KIR locus in a large cohort of more than 400 Caucasian individuals. Gene order and structure was determined by sequencespecific PCR of the intergenic regions. In this way, we could show that KIR3DL1 and KIR2DS4 gene variants are linked and that contrary to current views the gene KIR2DS5 is only present in the telomeric half of the KIR locus. Our study revealed novel insights in the highly organized distribution of KIR genes. Novel recombination hot-spots were identified that contribute to the diversity of KIR gene distribution in the Caucasian population. Next-generation sequencing (NGS) of the KIR intergenic regions allowed for a detailed single nucleotide polymorphism (SNP) analysis, which demonstrated several gene-specific as well as haplotype-specific nucleotides for a more accurate genotyping of this notoriously complex gene cluster. 54

57 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS INTRODUCTION Natural killer (NK) cells represent a lymphocyte subset critical for innate immunity, in particular with respect to the killing of tumor cells and elimination of virus-infected host cells. Although still debated, some studies in mice have suggested that NK cells can even elicit a form of memory function (O Leary et al. 2006; Paust et al. 2010). In humans, it is known that viral infection can have a long-lasting effect on the NK cell pool (Béziat et al. 2013a). Undesired effects of NK cell capabilities became clear in studies that showed their involvement in clinical manifestations of certain autoimmune diseases (Kunert et al. 2007; Ruggeri et al. 2002; van Bergen et al. 2011). Moreover, NK cells may play an important role in transplantation success rate (Kunert et al. 2007; Ruggeri et al. 2002; van Bergen et al. 2011; Venstrom et al. 2009). Cellular activation and cytotoxicity of human NK cells are regulated by cell surface-expressed molecules from various families of immune receptors, some of which are NK cell specific. One of these receptor families is the Killer Immunoglobulin-like Receptor (KIR) family. There are seventeen KIR genes described to date: six activating genes, nine inhibiting genes and two pseudogenes, all located together within the leukocyte receptor cluster on chromosome 19q13.4 (Trowsdale 2001). Although still a matter of debate, the genes KIR2DL2 and KIR2DL3 are often referred to as alleles of the same gene, as is also suggested for KIR3DL1/KIR3DS1 and KIR2DL1/KIR2DS1 (Hsu et al. 2002a; Uhrberg et al. 1997). As these genes (or alleles) are located at different positions on the KIR locus, we have chosen to consider them as individual genes in this manuscript. The activating receptors are marked by a short cytoplasmic tail that, upon engagement of the receptor, interacts with DAP12 and subsequently the immunoreceptor tyrosine-based activation motifs (ITAMs) elicit activating signals to the NK cell. The inhibiting receptors are equipped with longer cytoplasmic tails containing immunoreceptor tyrosine-based inhibition motifs (ITIMs). When engaged these receptors inhibit the NK cell from eliciting a cytotoxic response. Because of the high sequence homology between KIR genes in the human genome, a number of unequal cross-over and gene conversion events have taken place within the gene cluster, which has resulted in an unequal KIR gene distribution across the population (Jiang et al. 2012). Therefore large inter-individual diversity of allelic variation as well as extensive copy number variation (CNV) exist in human populations (Vendelbosch et al. 2013). The KIR locus genes have been divided into centromeric and telomeric segments. Each segment is separated by the genes KIR3DP1 and KIR2DL4. KIR3DL3 is always the first gene of the centromeric segment whereas the telomeric segment always has the KIR3DL2 gene at its end. These four KIR genes together are called framework genes as within this frame both segments of the KIR locus are formed. The presence of other KIR genes at this locus varies. Several studies have reported combinations of KIR genes that occur more often than others, suggesting a certain distribution and organization of these genes at the KIR locus resulting in genes specific for the A haplotype and other genes specific for the B haplotype (Hsu et al. 2002a; Jiang et al. 2012; Pyo et al. 2010; Uhrberg et al. 1997). Because of the sequence homology of these genes, genotyping is notoriously difficult and although the gene content of each locus can be determined, the order in which the genes reside on each locus remains difficult to determine. Although knowledge on the organizational structure of the KIR locus has taken major leaps in recent years due to novel (next generation) 3 55

58 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 sequencing techniques, controversies in literature still exist as to the exact organization of the KIR locus. For instance, the telomeric KIR genes KIR2DS1 and KIR2DS4 have sometimes been reported to be present on both the telomeric A and B haplotypes (Hsu et al. 2002a; Shilling et al. 2002; Vierra-Green et al. 2012), while other studies make a clear distinction with KIR2DS4 being a telomeric A-specific gene and KIR2DS1 being a telomeric B-specific gene (Pyo et al. 2010). For genetic association and functional expression studies regarding these KIR genes, it is crucial to know how this gene cluster is organized and how it can be accurately genotyped. In this study, using genotyping and sequencing data of intergenic regions, we have investigated the KIR gene locus in more detail and mapped the KIR loci of more than 400 Caucasian individuals. This effort improved our understanding of the KIR locus haplotype distribution. Moreover, detailed single nucleotide polymorphism (SNP) analysis of the intergenic regions provided additional sequence signatures useful for KIR genotyping techniques. Hence, for accurate typing of this complex gene cluster we propose an optimized KIR genotyping protocol. RESULTS Haplotype mapping of the KIR locus We determined the KIR genotype and CNV of all KIR genes in a cohort of 422 healthy Caucasian individuals. The arrangement of genes in the KIR locus was studied by allele specific PCR of the sequences in between the KIR genes. These intergenic regions, approximately 2 kb in length, were amplified with gene-specific primers (Supplementary Table 1), enabling us to target the exact location of each gene at the KIR locus. In this way a list of centromeric and telomeric haplotypes was obtained, of which the most common haplotypes are shown in Figure 1. The majority i.e. 90.8% of these unrelated, healthy and genotyped individuals carried these common KIR centromeric and telomeric haplotypes. Since each KIR locus was composed of a centromeric and a telomeric segment linked to the framework genes KIR3DP1 and KIR2DL4, only nine different common KIR haplotypes were found in these Caucasian individuals. The other individuals carried re-arrangements of these common haplotypes. By calculating the frequency of each haplotype and each possible combination of centromeric and telomeric segment, we determined the frequency that a certain haplotype occurred in the Caucasian population (Tables 1 and 2). In about 75% of the tested individuals, the frequency of the haplotypes could be generated without any assumptions, since these individuals had the same haplotypes on both alleles in either the centromeric or the telomeric segment. The frequency of the haplotype combinations are shown in the first part of Table 2. For the remaining individuals, the haplotype combinations were determined based on the results found in the first part of Table 2. For instance, one of the key assumptions was the lack of the combination of Cent-B1 with Tel-B1. Similar assumptions were made for the frequency of other haplotype segment combinations, resulting in a good indication for the frequency of haplotype combinations found in the complete cohort (lower part of Table 2). From this Table, we noticed that the frequency in which the B haplotypes of one segment combines with the B haplotype of the other segment was very low compared to the frequency of a B segment combination with 56

59 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS Table 1. Frequency of KIR haplotypes in a Caucasian cohort of 422 individuals. Percentages are calculated as the amount of haplotypes (N) counted in a total of 844 alleles. Haplotypes Frequency % (N) Centromeric part Haplotype A 71.1% (600) Haplotype B1 10.5% (89) Haplotype B2 16.0% (135) Non-standard arrangement 2.4% (20) Telomeric part Haplotype A 76.4% (645) Haplotype B1 4.5% (38) Haplotype B2 16.5% (138) Non-standard arrangement 2.6% (23) 3 an A segment. This suggests that these combinations of the KIR genes with centromeric and telomeric B segments have arisen recent in evolution. Syntenic segregation of KIR genes within the KIR locus The allele-specific PCR products of the intergenic regions of the KIR genes gave us the opportunity to deduce the order in which the KIR genes are present in the different haplotypes. This analysis showed which genes were always found in synteny and which genes had variable neighboring genes. For instance, the two adjacent genes KIR2DP1 and KIR2DL1 were linked in the centromeric segment and their presence in a haplotype was not random. In every individual that was tested positive for KIR2DL3, the KIR2DP1-KIR2DL1 gene combination was found in all cases, forming the centromeric haplotype-a segment (Cent-A). The centromeric haplotype-b segment is consistently marked by the presence of the KIR2DS2-KIR2DL2 gene combination, with either the KIR genes KIR2DL5B, KIR2DS3, KIR2DP1 and KIR2DL1 all present (forming Cent-B1) or all absent (forming Cent-B2) (Figure 1). Moreover with current typing methods the Cent-B1 haplotype is believed to contain either KIR2DS3 or KIR2DS5 (Jiang et al. 2012; Martin et al. 2008; Pyo et al. 2010). Instead, our data showed that KIR2DL5B was consistently followed only by KIR2DS3. This is explained in more detail below. The telomeric haplotype A (Tel-A) segment was as invariable as the Cent-A segment containing only the KIR3DL1 and KIR2DS4 genes. The telomeric haplotype B segment of the KIR locus has a predicted gene organization in which KIR3DS1 was always found to be linked to KIR2DL5A. This region was identified by the presence of the KIR2DS3 gene (forming Tel-B1) or the KIR2DS5 gene (forming Tel-B2), preceded by the KIR2DL5A gene. Both of these telomeric B haplotypes contained the KIR2DS1 gene (Figure 1). Our data show strong linkage between some KIR genes that have not been characterized before, which prompted us to investigate the KIR haplotypes in more detail by sequencing the intergenic regions. 57

60 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 Table 2. Frequency of KIR haplotype combinations in a Caucasian cohort. The top panel confers the combinations from individuals that had identical centromeric or telomeric segments, allowing exact calling of the total number of haplotype combinations (327 individuals; 654 alleles / combinations). The lower panel includes all the haplotype combinations from individuals with 4 different haplotype segments (both centromeric and telomeric alleles are different) that could be deduced based on the calculated frequencies. The results of 8 individuals remained inconclusive. Haplotype combinations Frequency Calculated n=327 Cent-A--- Tel-A 68,0% Cent-A--- Tel-B1 1,8% Cent-A--- Tel-B2 9,8% Cent-B1 --- Tel-A 5,5% Cent-B1 --- Tel-B1 0,2% Cent-B1 --- Tel-B2 0,5% Cent-B2 --- Tel-A 10,7% Cent-B2 --- Tel-B1 0,0% Cent-B2 --- Tel-B2 0,8% Non-standard arrangement 2,8% Deduced from above n=414 Cent-A--- Tel-A 54,2% Cent-A--- Tel-B1 3,4% Cent-A--- Tel-B2 12,6% Cent-B1 --- Tel-A 8,0% Cent-B1 --- Tel-B1 1,2% Cent-B1 --- Tel-B2 1,1% Cent-B2 --- Tel-A 13,5% Cent-B2 --- Tel-B1 0,0% Cent-B2 --- Tel-B2 2,4% Non-standard arrangement 3,6% KIR3DL1 and KIR2DS4 genes are coupled together To study the synteny between certain KIR genes in more detail, we sequenced the intergenic regions of the KIR locus using the gene specific primers (Supplementary Table 1). Initially, we focused on the most common haplotype A, since some consensus in its telomeric segment was found. The same number of copies was found for both KIR3DL1 and KIR2DS4 in each individual, suggesting that these genes were linked in the Caucasian population. Further analysis of this region showed that the KIR3DL1 gene was always found in combination with either a KIR2DS4 wildtype gene or one of the truncated variants of this gene (KIR2DS4*003-*010, *012, *013) (Figure 2A). After genotyping of the non-expressed KIR3DL1*004 allele (Pando et al. 2003), we found this allele in 27.8% of the individuals (15% of the analyzed haplotypes), which is in 58

61 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS Centromeric region Cent-A 3DL3 2DL3 2DP1 2DL1 3DP1 Cent-B1 3DL3 2DS2 2DL2 2DL5B 2DS3 2DP1 2DL1 3DP1 Cent-B2 3DL3 2DS2 2DL2 3DP1 Recombinant haplotypes 3 Telomeric region Tel-A 2DL4 3DL1 2DS4 3DL2 Tel-B1 2DL4 3DS1 2DL5A 2DS3 2DS1 3DL2 Tel-B2 2DL4 3DS1 2DL5A 2DS5 2DS1 3DL2 Recombinant haplotypes Figure 1. Schematic overview of the basic haplotypes in a Caucasian cohort of 422 individuals. Each complete KIR haplotype consists of a centromeric segment and a telomeric segment, creating 9 possible haplotypes in total. concordance with previous studies (Gonzalez-Galarza et al. 2011). Remarkably, KIR3DL1*004 was identified only in combination with a truncated variant of KIR2DS4 and never with the KIR2DS4 wildtype allele (Figures 2A and 2B). On the other hand, the truncated KIR2DS4 allele variants could be present with either an expressed or non-expressed allele of KIR3DL1. PCR amplification of the intergenic region between KIR3DL1 and KIR2DS4 was successful in all individuals when these genes were present in their genome, confirming the linkage between these genes. Subsequent sequence analysis of this intergenic region revealed highly conserved sequences, with several SNPs being indicative of the presence of a functional KIR3DL1 gene, a full-length KIR2DS4 wildtype allele, or a truncated KIR2DS4 gene (Figure 2C). These data demonstrate that these two genes have co-evolved in the Caucasian population and show absolute linkage at the telomeric A haplotype of the KIR locus. KIR2DS5 is not present in the centromeric segment of the KIR locus Using the same approach of intergenic region sequence analysis we next focused on the KIR2DL5 gene, having two variants A and B, suspected to be derived from the same parental gene (Gomez- Lozano et al. 2002), either in the centromeric A (KIR2DL5B) or telomeric B (KIR2DL5A) segment. As mentioned before, these KIR2DL5 genes have been suggested to be linked to both KIR2DS3 and KIR2DS5, no distinction between KIR2DL5A and KIR2DL5B was made. Initial genotyping for the KIR2DL5A or KIR2DL5B genes at this locus by our multiplex ligation-dependent probe 59

62 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS A KIR alleles Frequency B KIR3DL1 all 77.1% KIR3DL1*expr 62.1% KIR3DL1* % 3 KIR2DS4 all 77.1% KIR2DS4 WT 20.1% KIR2DS4 trunc 57.0% Allele combinations KIR3DL1*expr - KIR2DS4 WT 20.4% KIR3DL1*004 - KIR2DS4 WT 0.0% KIR3DL1*expr - KIR2DS4 trunc 41.8% KIR3DL1*004 - KIR2DS4 trunc 15.0% C DL1*expr G T A T C A C G A A G A A C 2DS4 WT 3DL1*expr 3DL1*004 G C A T T T T A G G G R R Y A C G A T T T A G G A C G C 2DS4 trunc 2DS4 trunc Figure 2. The genes for KIR3DL1 and KIR2DS4 show strong linkage in the Caucasian population. (A) Frequency of the genes and the various combinations found in 422 individuals (counted per 844 alleles); making a distinction between the non-expressed allele (KIR3DL1*004) and all other alleles (KIR3DL1*expr) of KIR3DL1 and the full length (KIR2DS4 WT) and truncated (KIR2DS4 trunc) alleles of KIR2DS4. (B) Number of individuals carrying 1 or 2 KIR2DS4 WT alleles, 1 or 2 KIR2DS4 trunc alleles or both alleles in combination with a certain KIR3DL1 allele. (C) Detailed SNP analysis of the conserved region between KIR3DL1 and KIR2DS4. Varying nucleotides are indicated, numbered from the last base-pair position of the preceding gene. amplification (MLPA) method, resulted in a considerable number of individuals that were negative for KIR2DS3 or KIR2DS5, which enabled us to estimate to which extent these KIR2DL5 genes were linked to either KIR2DS3 or KIR2DS5. Our results indicated that KIR2DL5B was found together with KIR2DS5 in only 4 out of the 94 KIR2DS5-positive individuals, while the combination with KIR2DS3 was found in all of the remaining 90 individuals carrying a KIR2DL5B gene (Figures 3A and 3B). In fact, further investigation of these haplotypes carrying the KIR2DL5B-KIR2DS5 gene combination revealed, that these four individuals all showed an unequal cross-over between the Cent-B1 segment and the Tel-B2 segment. This recombination event had caused the absence of the KIR2DS3, KIR2DP1, KIR2DL1, KIR3DP1, KIR2DL4, KIR3DS1 and KIR2DL5A genes and created a very short hybrid B haplotype (Figure 3C). These data indicate that the KIR2DS5 gene is only located in the telomeric B segment and is not found in the centromeric B segment. After the identification of the genetic organization of the telomeric B haplotypes, further analysis showed that KIR2DL5A could be observed in combination with either KIR2DS3 or KIR2DS5 (Figure 3A and 3B). The combination of KIR2DL5A with KIR2DS5 was observed three times more than the combination with KIR2DS3. Subsequent SNP analysis of the intergenic regions showed no difference between KIR2DL5A-KIR2DS3 and KIR2DL5A-KIR2DS5 in the telomeric B haplotypes. In contrast, several differences were found when these intergenic regions were compared with the KIR2DL5B-KIR2DS3 60

63 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS region of the centromeric B segment (Figure 3D). The intergenic region of the rare KIR2DL5B-KIR2DS5 gene combination resembled the region between KIR2DL5A and KIR2DS5, suggesting that the location of the cross-over breakpoint is located somewhere between the start of the KIR2DL5B gene and the first specific SNP at 532 bp upstream from KIR2DS5 in the intergenic region (data not shown). Similarities and polymorphisms in the KIR intergenic regions Sequence analysis by next generation sequencing (NGS) of all the intergenic regions within the KIR locus of 24 Caucasian individuals generated data to create a detailed map of this region including the promoter regions of the KIR genes (Supplementary Figure 1). All intergenic regions were of the same length and were highly homologous, with the exception of the regions located directly 3 of the framework gene KIR3DL3 and 5 of the framework gene KIR3DL2. The region between KIR3DP1 and KIR2DL4 was not included in this study since we did not succeed in amplifying this 14 kb sequence by long-range PCR. Detailed sequence analysis of the intergenic regions on the KIR locus enables us to perform simple and fast haplotyping. Intergenic nucleotides in the centromeric segment of the locus between KIR2DL3-KIR2DP1, KIR2DS3-KIR2DP1 and KIR2DL2-KIR3DP1 will give the correct information of the 3 A KIR alleles Frequency B KIR2DL5 A 22.4% KIR2DL5 B 11.1% KIR2DS3 15.3% KIR2DS5 18.1% Allele combinations KIR2DL5 A - KIR2DS3 4.6% KIR2DL5 A - KIR2DS5 17.7% KIR2DL5 B - KIR2DS3 10.7% KIR2DL5 B - KIR2DS5 0.5% C 3DL3 2DS2 2DL2 2DL5B 2DS5 2DS1 3DL2 D DL5A 2DL5B C AT G C C G C G C T A C C T A T T G T GC C A Y A A T A C G T T G G C C A 2DS3/2DS5 2DS3 Figure 3. Presence of KIR2DS5 in the centromeric or telomeric half of the KIR locus. (A) Frequency of the genes KIR2DL5A, KIR2DL5B, KIR2DS3 and KIR2DS5 and the various combinations found in 422 individuals (counted per 844 alleles). (B) Number of individuals carrying 1 or 2 KIR2DL5A genes, 1 or 2 KIR2DL5B genes or both genes in combination with a either KIR2DS3 or KIR2DS5. (C) KIR genome of the individuals that carry a combination of KIR2DL5B and KIR2DS5. The contracted haplotype is the result of an unequal cross-over between the KIR2DL5B and KIR2DL5A gene, because of which KIR2DS3, KIR2DP1, KIR2DL1, KIR3DP1, KIR2DL4 and KIR3DS1 are deleted. (D) Detailed SNP analysis of the region between KIR2DL5A/B, KIR2DS3 and KIR2DS5. Varying nucleotides are indicated, numbered from the last base-pair position of the preceding gene. 61

64 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 complete centromeric KIR haplotype. By targeting just a few nucleotides in the intergenic region between KIR2DL4 and KIR3DL1 or KIR3DS1, haplotyping of the telomeric region can be performed to distinguish Tel-A from Tel-B. Specific nucleotides targeting the region between KIR3DL1-KIR2DS4 and KIR2DS1-KIR3DL2 will define all genes and gene variants in the telomeric segment of the KIR locus. The phylogenic tree of the KIR gene intergenic sequences visualizes which regions most resemble each other (Figure 4). As indicated above, the intergenic region following KIR3DL3, between KIR2DS2 and KIR2DL2 and the region just before KIR3DL2 formed a different branch on the phylogenic tree (in blue). In contrast to other intergenic KIR regions, these regions showed multiple poly-t stretches and contained sequence sections that were very different from other intergenic regions of the KIR locus (Supplementary Figure 1). A high degree of homology was observed between the intergenic regions between KIR3DL1-KIR2DS4 (Tel-A), KIR2DS3-KIR2DP1 (Cent-B1), KIR2DS3-KIR2DS1 (Tel-B1), KIR2DS5- KIR2DS1 (Tel-B2), KIR2DL5AB-KIR2DS3 (Cent-B1; Tel-B1) and KIR2DL5A-KIR2DS5 (Tel-B2) (Figure 4 in red). Although several SNPs exist that can distinguish the different regions from each other, the sequences shared a high degree of homology to each other, despite the fact that the genes in which these intergenic regions are located, do not resemble each other to a similar degree and come from different haplotypes at the KIR locus. In green (Figure 4) a number of intergenic sequences are listed that resemble each other. The similarity between the intergenic regions before KIR2DL5A, KIR2DL5B and KIR3DP1 was Figure 4. Detailed sequence analysis of the intergenic regions within the KIR locus of 24 individuals. The phylogenetic tree of the sequences between each KIR gene shows which regions show the most resemblance to each other. Highly similar sequences are highlighted in blue, red and green. The numbers at each node display the bootstrap score scaled to 1. 62

65 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS not unexpected since the sequences of these KIR genes themselves resemble each other to a high degree. In contrast, there was a lack of sequence similarity between the regions KIR2DS3- KIR2DP1 and KIR2DL3-KIR2DP1 (30%), suggesting that the A and B haplotypes have diverged quite early in evolution. The regions between KIR2DL3-KIR2DP1 and KIR2DP1-KIR2DL1 also showed a lack of similarity with any other intergenic region, confirming the conserved nature of these regions and supporting the notion that these genes show strong linkage as was proposed in Figure 1. These data also confirm why we did not observe any recombination events for this region as described below. 3 Differences between promoter regions of the KIR genes The detailed sequence information of the intergenic regions gave us the opportunity to study hypothetical transcription factor binding motifs in the promoter regions of the KIR genes. We checked the last 300 base pairs of each intergenic region before the transcription start site using GimmeMotifs (van Heeringen and Veenstra 2010) and identified 47 potential transcription factor binding sites (Supplementary Table 2). Some of these transcription factor binding sites are found only in a few KIR promoter regions, while others are present in every KIR promoter region multiple times. Most of the transcription factors found are probably not expressed in NK cells or may not be involved in KIR gene regulation. Therefore, a selection of transcription factors was made that are known to bind to KIR gene promoters. So far, the literature on KIR promoter regions has mainly focused on well-described and/or well-detectable KIR genes like KIR2DL5, KIR3DL1 and KIR2DS4, omitting the analysis of other KIR promoters (Davies et al. 2007; Gómez- Lozano et al. 2007; Li et al. 2008; Presnell et al. 2006; Trompeter et al. 2005; van Bergen et al. 2005). We have created an overview of every KIR promoter region in which we have mapped the binding sites of all known transcription factors involved in KIR gene regulation described in the literature to date (Figure 5) (Davies et al. 2007; Gómez-Lozano et al. 2007; Li et al. 2008; Presnell et al. 2006; Trompeter et al. 2005; van Bergen et al. 2005). From this overview, we can conclude that the transcription factors STAT (signal transducer and activator of transcription), Ets (E26 transformation-specific), CREB (camp response element-binding protein) and AP4 (activating enhancer binding protein 4) can bind to each KIR promoter and that there is at least 1 potential YY1 (Yin Yang 1) binding site present in every KIR promoter. On the other hand, Sp1 (specificity protein 1) and AML (acute myeloid leukemia 1 protein) can bind only to the promoters of a small number of the KIR gene. The AML (or RUNX (runt-related transcription factor 1)) binding site in the promoter regions of KIR2DL5B and KIR3DP1 contains a SNP, which is thought to cause silencing of these genes (Gómez-Lozano et al. 2007). Together these data suggest that it is likely that KIR gene promoter sequence variation has an impact on transcription factor recruitment and thus explains the difference in the expression level of a particular KIR gene Recombination hot-spots in the KIR locus Although the KIR locus seemed to be organized in a relatively constant and systematic way in the major part of the Caucasian population, variation within the locus is caused by several recombination 63

66 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 STAT/Ets YY1 CREB YY1 AML YY1 Ets AP4 Sp Consensus A A G T--GGGGCGCG 2DL4-3DL1 T--GGGGCGCR 2DL4-3DS1 T--GGGGCGCR 2DL5AB-2DS3 G TAGGGGTGCG 2DL5A-2DS5 G TAGGGGTGCG 2DS3-2DP1 G 2DS3S5-2DS1 G 2DP1-2DL1_A G T--CGGTCGCG 2DP1-2DL1_B2 G T--CGGTCGCG 2DL3-2DP1 G T--GGGGCACG 2DL1-3DP1_A A 2DL1-3DP1_B2 A 2DL2-3DP1 A 2DS4WT-3DL2 2DS4trunc-3DL2 2DS1-3DL2_B3 2DS1-3DL2_B5 2DS2-2DL2 3DL3-2DS2 3DL3-2DL3 3DS1-2DL5A G G 2DL2-2DL5B G A 3DL1-2DS4trunc 3DL1*004-2DS4trunc 3DL1-2DS4WT Figure 5. Potential transcription factor binding sites within the KIR promoter regions. A detailed comparison was performed of each KIR promoter sequence and the binding sites of known transcription factors involved in KIR transcription regulation (Davies et al. 2007; Li et al. 2008; Presnell et al. 2006; Trompeter et al. 2005; van Bergen et al. 2005). Highlighted in grey are the promoters that contain a binding site for the indicated transcription factor. For those promoters that cannot facilitate the indicated transcription factor, the difference in sequence compared to the consensus sequence is shown. events. One of the most common places to facilitate such events is the recombination hot-spot between the framework genes KIR3DP1 and KIR2DL4, causing the centromeric and the telomeric segments to be exchanged as is shown in Table 2 (Pyo et al. 2010). Recombination events can either take place in the region between the genes, as is the case for the recombination between KIR3DP1 and KIR2DL4, or within KIR genes hence creating so-called hybrid genes. A frequently observed phenomenon of a hybrid gene is the recombination between KIR3DP1 and KIR2DL5A that results in the KIR3DP1*004 allele (Gómez-Lozano et al. 2005; Jiang et al. 2012; Martin et al. 2003; Pyo et al. 2010), which is present in 1.2% of our Caucasian cohort. We observed signs of a recombination event in the centromeric or telomeric segment of the KIR locus in 9.2% of the Caucasian individuals. Most of the recombination events were found in a few specific regions, suggesting additional recombination hot-spots that have not yet been described previously. The recombination between KIR2DL5B in the centromeric region and KIR2DS5 in the telomeric region has already been mentioned above. Besides the four individuals with KIR haplotypes that lost an entire set of genes due to this unequal crossover event (Figure 3C), we found its counterpart in two individuals in our cohort, in which the KIR locus had gained a set of genes at this locus (Figure 6A). These individuals gained KIR2DS3, KIR2DP1, KIR2DL1, KIR3DP1, KIR2DL4 and KIR3DS1 through an unequal cross-over event between KIR2DL5A in Tel-B and KIR2DL5B in Cent-B1. In contrast to common belief, the KIR2DS2 gene was not always found to be linked to the KIR2DL2 gene. In 2% of the Caucasian individuals, KIR2DS2 was present in the absence of its 64

67 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS inhibiting counterpart. In these individuals, KIR2DS2 was directly linked to either KIR2DL5B, KIR2DP1 or KIR3DP1 in the centromeric segment of the B haplotype (Figure 6B), indicating that the genes normally present between the KIR2DS2 gene and the gene linked to it, were deleted. Finally, there were a few individuals (1%) that did not fit into any of the 9 common haplotypes because all the haplotype-specific KIR genes were absent (Figure 6C). In these cases, KIR3DL3 was followed by KIR2DP1, missing all functional KIR genes of the centromeric haplotype B since the protein KIR3DL3 is not expressed on the cell surface (Trundley et al. 2006). In another recombination event involving KIR3DL3, all KIR genes between KIR3DL3 and KIR2DS1 were absent, resulting in a fusion of the centromeric framework gene KIR3DL3 with a small part of the telomeric haplotype B segment. Although more recombination events have undoubtedly occurred within the KIR locus, the hot-spots described here contribute to the majority of the variation found in our cohort. 3 MATERIALS AND METHODS Ethics Statement The study was approved by the Institutional Medical Ethics Committee of the Academic Medical Center in Amsterdam and was performed in accordance with the Declaration of Helsinki. Participants provided their written informed consent to participate in this study. A 3DL3 2DS2 2DL2 2DL5B 2DL5A 2DS5 2DS1 3DL2 2DL4 3DS1 2DL5A 2DL5B 2DS3 2DP1 2DL1 3DP1 2DL4 3DS1 2DL5A3DP1 2DL4 3DL1 2DS4 3DL2 B 3DL3 2DS2 2DL5B 2DS3 2DP1 2DL1 3DP1 3DL3 2DS2 2DP1 2DL1 3DP1 3DL3 2DS2 3DP1 C 3DL3 2DP1 2DL1 3DP1 3DL3 2DS1 3DL2 Figure 6. Recombination hot-spots observed in a cohort of 422 individuals. (A) An often observed unequal cross-over event is seen within the genes for KIR2DL5A, KIR2DL5B and KIR3DP1, resulting in a deletion or an insertion of several genes, depending on which genes recombine with each other. (B) KIR2DS2 can be present without KIR2DL2. In these cases, the gene is followed by KIR2DP1, KIR3DP1 or KIR2DL5B, deleting KIR2DL2 and all genes in between, dependent on the gene it recombines with. (C) A small proportion of Caucasian individuals carry (part of) a centromeric KIR locus that does not contain haplotype specific KIR genes, caused by deletion of all KIR genes between KIR3DL3 and KIR2DP1 or KIR3DL3 and KIR2DS1. 65

68 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 KIR genotyping DNA was extracted from 422 healthy individuals, using the QIAgen Blood Kit (Qiagen, Frederick, MD, USA) according to the manufacturers protocol. KIR genotyping and copy number determination was performed with multiplex ligation-dependent probe amplification as described before (Vendelbosch et al. 2013). Shortly, oligodeoxyribonucleotide probes, designed to target a single gene or allele group, are hybridized to the genome. After binding and subsequent ligation, fluorescently labeled universal primers amplify the ligated probes. As every probe set has a different length, these amplification products can be separated and quantified by capillary electrophoresis. The product quantity is relative to the number of genes present in the genome, allowing copy number determination using GeneMarker software (version 1.90, SoftGenetics, State College, PA, USA). Inconclusive copy number results were verified by quantitative PCR as described before (Vendelbosch et al. 2013). Allele subtyping by PCR Since the KIR MLPA did not discriminate KIR2DL5A and KIR2DL5B genes, gene-specific PCRs were designed to differentiate the presence of these two genes. Gene-specific primer sets used were FW- KIR2DL5A (5 -gcagcaccatgtcgctcatggaca-3 ), FW-KIR2DL5B (5 -gcagcaccatgtcgctcatggacg-3 ) and REV-KIR2DL5A and B (5 -cacacttgggtgcccatggsttcg-3 ). Each 25 μl PCR reaction consisted of 200 µm dntps, 2.5 μl 10x Salsa PCR buffer (MRC Holland, Amsterdam, Netherlands), 10 pmol of each primer, 2.5 units Salsa polymerase (MRC Holland) and 275 ng TaqStart Antibody (Clontech, Saint-Germainen-Laye, France). Initial denaturation took place at 95 C for 2 minutes, followed by 35 cycles of 20 sec at 95 C, 30 sec at 60 C and 1 min at 72 C, and a final elongation step of 2 min at 72 C. The non-expressed KIR3DL1*004 and surface-expressed KIR3DL1 alleles (all except the allelic variant *004) were determined in a multiplex PCR, using primers based on those reported by Gardiner et al. (Gardiner et al. 2001); FW KIR3DL1*004 (5 -cagacacctgcatgttctc-3 ), FW KIR3DL1*expr (5 -tatcctcagcacgttccaagg-3 ) and REV KIR3DL1 (5 -gtacaagatggtatctgtag-3 ). The same PCR program was used as described above with the exception of the amount of FW primers, which was 5 pmol of each primer, and the elongation step, which was 1.5 min each cycle. Amplification and sequencing of the intergenic regions The order of KIR genes on the KIR locus was determined by sequence-specific PCR. The regions between KIR genes were amplified with primers designed specifically for each region, except of the region between KIR3DP1 and KIR2DL4 (Supplementary Table 1). All PCR assays were performed as described above with an elongation step of 2 min for each cycle to yield 2 kb products. To establish a reference database for the KIR intergenic regions, Sanger sequencing of the purified PCR products was performed according to the manufacturer s protocol of the BigDyeTerminator v1.1 cycle sequencing kit on an ABI-3130XL (Applied Biosystems, Life Technologies, Carlsbad, CA, USA). To expand our sequence database for the KIR intergenic regions and to study ambiguities between regions, we amplified these intergenic regions with universal primers, targeting all 66

69 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS KIR genes; FW intergenic (5 -cacttgacccctgcccacctctcc-3 ) and REV intergenic (5 -agctcagca gcgcacaggatgttatttggc-3 ). Next generation sequencing (NGS) was performed on a 454 FLX+ Roche Genome Sequencer (Roche Diagnostics, 454 Life Sciences, Branford, CT, USA), with the Shotgun Rapid Library method. The amplicon size varied between 40 and 920 bp with an average of 299 bp. A minimum coverage depth of 10 reads over the entire reference sequence was maintained for accurate analysis. The reads of the intergenic regions produced by 454 sequencing were aligned to the previous obtained reference sequence database by Sanger sequencing for the intergenic regions with DNASTAR SeqMan Ngen Software (DNASTAR, Inc., Madison, WI, USA). A detailed variant report for each intergenic region was created with SeqMan Pro (DNASTAR, Inc). An overview of the intergenic region sequences was created with BioEdit Sequence alignment Editor v (Hall 1999). Predictive analysis of the promoter regions of all intergenic regions was performed with GimmeMotifs Software (van Heeringen and Veenstra 2010). A phylogeny tree of the regions was created with Clustal Omega (EMBL-EBI 2014) using neighbor joining analysis. Visualization was performed using FigTree v1.4.2 (Institute of Evolutionary Biology, University of Edinburgh, 3 DISCUSSION In this study we mapped the location of the KIR genes within the KIR locus, in more than 400 Caucasian individuals, in detail by combining the results of our MLPA method for genotyping and CNV determination and by determining gene order by sequence-specific PCR. We have also provided the first detailed sequence map of the regions in between KIR genes by NGS of the KIR intergenic regions of 24 individuals. Our approach has resulted in a refinement and improved understanding of the organization of the human KIR locus. Despite the identification of some novel gene combinations, only a few basic haplotypes make up the KIR locus in more than 90% of the Caucasian population, revealing a much less variable haplotype structure of the KIR gene cluster than previously assumed (Figure 1). On the other hand, the allelic variation of the individual KIR genes creates a large portion of the variation within the basic haplotype structure, which makes this gene family the most variable among the non-rearranging receptors within the human genome. The genomic organization of these haplotypes was best revealed by syntenic coupling of some KIR genes. For instance, the genes KIR3DL1 and KIR2DS4 are tightly linked with a highly conserved 2 kb intergenic region (Figure 2). We have never found the gene KIR2DS1 to be coupled to KIR3DL1, as has previously been suggested in the literature (Hsu et al. 2002a; Norman et al. 2004; Shilling et al. 2002; Vierra-Green et al. 2012). Even though recombination events have commonly occurred at this locus (according to recombination hot-spots as described here and reported by previous family segregation analyses by us and others (Martin et al. 2003; Vendelbosch et al. 2013)), the KIR3DL1-KIR2DS4 linkage has always remained intact. Although others have observed haplotypes in which this syntenic bond is not observed, these difference may be attributed to the use of different and less detailed sequence information (Hsu et al. 2002a; Norman et al. 2004; Shilling et al. 2002). Within these conserved segments 67

70 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 of genes, the KIR3DL1*004 allele was present in about 28% of the individuals and only occurred in combination with a truncated KIR2DS4 allele, suggesting a co-evolutionary event within this haplotype. Individuals with the non-expressed KIR3DL1*004 allele consequently also lacked the KIR2DS4 gene expression and therefore a telomeric haplotype A without any functional KIR gene. Such a strong linkage is important for the interpretation of gene association studies with disease conditions focusing on the KIR locus. For example, a genetic study in patients with leukemia suggested that the wild type allele for KIR2DS4 has a protective effect against CML (Giebel et al. 2008). The authors state that this association is only found when a distinction is made between the wildtype KIR2DS4 allele and the truncated KIR2DS4 allele. However, the KIR3DL1 gene is not taken into account in this study and a possible association through linkage could be easily overlooked. Therefore an optimal genetic association study in this context requires at least an integral analysis of the KIR haplotype. In contrast to current understanding, our haplotyping methods demonstrate that the centromeric KIR locus never contains the KIR2DS5 gene, which explains why this gene is found less frequent in the human population than one would assume when present in both the centromeric and telomeric segment (Gonzalez-Galarza et al. 2011; Vendelbosch et al. 2013). Although the KIR2DS3 gene was observed in both centromeric and telomeric segments, the predominant absence of the gene in the telomeric segment may again explain the low frequency of this specific KIR gene in the general Caucasian population. For the first time, a detailed sequence analysis of the intergenic regions between the KIR genes was obtained. These sequences provide a detailed overview of every promoter region at the KIR locus which can help to provide answers regarding their transcription regulation. A predictive transcription factor binding site tool showed that there is a high degree of variation in hypothetical transcription factor binding motifs in the promoter regions of KIR genes. More specifically, some of the transcription factors known to be involved in KIR gene regulation seem to selectively target only some KIR promoters. Although, this variation in KIR promoter regions might contribute to differences in transcription levels, the promoter activity is not the only factor controlling KIR gene expression regulation. Various epigenetic factors determine which specific KIR genes are transcribed in each cell. The methylation status of CpG islands within the KIR locus play a major role and also histone modifications have been reported to influence KIR transcription regulation (Cichocki et al. 2013; Uhrberg 2005). These factors should be taken into consideration when studying the transcriptional regulation of the individual KIR genes. The KIR intergenic sequences that we have obtained did provide an important reference data set to characterize each individual KIR gene arrangement in the KIR locus. They can help to identify different haplotypes and have provided a better understanding of the KIR locus organization in general. To complete a map of the entire KIR locus, allele-specific analysis needs to be performed to address whether some of the SNPs in the intergenic regions that we have identified are linked to certain allelic KIR gene variants. Such mapping efforts are likely to be very informative, as demonstrated in this study for KIR3DL1 and KIR2DS4, but also for KIR2DP1 in which Cent-A carried a SNP at nucleotide position 28 as a genetic signature for the *002, *005 or *008 alleles (Supplementary Figure 1 and the EMBL-EBI Immuno Polymorphism 68

71 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS Database). While these alleles were present only in Cent-A, other KIR2DP1 alleles were present only in the Cent-B1 segment. Sequence analysis and allele typing of the other genes at the KIR locus may provide more of these detailed genetic signatures, allowing for more accurate and complete genotyping of the KIR gene content in any given individual. In addition to the previously described recombination hot-spots at the KIR locus, we have identified some common recombination sites that can lead to the gain or loss of multiple genes at the same time. The recombination event between KIR2DL5B and KIR2DS5 that we describe in this manuscript, was observed previously in a Caucasian family (Vendelbosch et al. 2013) and in the Ga-Adangbe population in Ghana (Norman et al. 2013), suggesting that this haplotype is more commonly observed. Overall, it appears that recombination events at the KIR locus have taken place in 9.2% of the Caucasian population. Some of these recombination events must have taken place with breakpoints within the genes as exemplified by the recombination between KIR3DP1 and KIR2DL5A and -B, whereas other recombinations have their breakpoints within the intergenic regions. The formation of hybrid genes within the KIR locus has been previously ascribed to the presence of Alu repeats within the introns of the KIR genes (Traherne et al. 2010). As every intergenic region also has ~80% similarity with Alu sequences (data not shown), it seems very likely that all these elements have contributed to the recombination events at the KIR locus (Batzer and Deininger 2002). Past recombinations within the KIR locus have contributed to the high diversity of KIR haplotypes throughout the population, as we and others have previously observed (Jiang et al. 2012; Martin et al. 2003; Traherne et al. 2010; Vendelbosch et al. 2013). The recombination hot-spots that we have observed explain some of the obscurities that exist in literature regarding KIR gene organization, such as the possibility that KIR2DL5B can exist on the same haplotype segment as KIR2DS5 (Du et al. 2008; Hsu et al. 2002a; Vierra-Green et al. 2012). Here we have shown that this combination of genes can only occur if a recombination event has taken place, deleting several genes including KIR2DS3 and KIR2DL5A. Moreover, we have also shown that through recombination the KIR3DL1 and KIR2DS4 genes can be present on the same KIR haplotype as KIR3DS1, while in a classical haplotype this is never the case. A reason for ambiguities in the literature concerning KIR gene order may be the inability of the common genotyping methods to distinguish between the two KIR locus strands. Deletions and duplications are difficult to detect when typing only for the presence or combinations of certain genes. Such ambiguities can be overcome by using more advanced genotyping techniques that consider gene copy number variation. Overall, the KIR locus was found to be organized within certain limits. In this study 90.8% of all tested individuals had a KIR locus that comprised one or two out of nine possibilities of haplotype segment combinations identified thus far. As demonstrated in this study, the intergenic SNPs aided to identify entire haplotypes, simplifying KIR genotyping and allowing for the development of cheaper and easier tools for diagnostic purposes. SNP analysis has already been used by others to study KIR allele and gene linkage (Norman et al. 2004). With a more simple genotyping method it would be possible to haplotype at least 90% of Caucasians, having targeted only nine specific SNPs, which would be informative for genotyping arrays or other large-scale high-throughput assays. By using a quantitative SNP analysis method and by targeting three extra SNPs between KIR3DL3-KIR2DS2, KIR2DS2-KIR2DL2 and KIR2DL2-KIR2DL5B 3 69

72 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 one will be able to detect the most common recombination hot-spots, completing the genotyping of the entire KIR locus in the Caucasian population. Of course the synteny between KIR genes described here might differ substantially among different ethnic backgrounds, as is already suggested for an African population (Norman et al. 2013). However, a study on a Japanese population shows similar patterns of organization as we describe in this manuscript (Yawata et al. 2006), even though the haplotype frequencies differ from the Caucasian populations. Future studies along similar lines as the ones presented here are likely to shed light on the KIR gene organization in other ethnic populations. Using the sequence information of this study we can further bridge the gap between this extremely complex gene cluster and a more comprehensive interpretation of the KIR locus for further research on functional expression of KIR genes, genetic association studies as well as patient diagnostics in the near future. 70

73 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS SUPPLEMENT Supplementary Figure 1. Multiple sequence alignment of all the intergenic KIR regions. The Supplementary Figure can be found at: 3 Supplementary Table 1. Primers used to sequence all intergenic KIR regions. FW primers on end of gene Sequence 5 > 3 KIR2DL1/KIR2DL2 GCACCACAGTCAGGCCTTGACGC KIR2DL2 CACAAATCTGAATGTGCCTCACA KIR2DL3 CCAGCAGCTGGAATCTGAACGC KIR2DL4 GCGTGAGTCTCCATCTTTGA KIR2DL5/KIR3DL1 CTTCAGGCCCATAACTCCACCACT KIR2DL5 GACCCCTGCCCACCTCTCCAACTG KIR2DP1 GCAATCACACTGAGGAACTCAGAG KIR2DS1/KIR2DS2/KIR2DS4 CTGGCTTACTTCCTAGTCTTCC KIR2DS3 GTCTAAGGTCCCCACTGCCTGCAGC KIR2DS3/KIR2DS5 GCTGTTCCACCTTCTCTCATGCA KIR2DS4 GTGCTGTTCCACCTCCGTT KIR2DS4WT GGCACCAGAATCTATTTCTCT KIR3DL1 TTGTGATTTCAATGTAGCTG KIR3DL3 GGAACTCACAATTCCAAATCAAAT KIR3DS1 TTCCACACATACAAGAGGCTGCG REV primers on start of gene Sequence 5 > 3 KIR2DL1 CCGGAGCAGACAGGCAGCCGGGA KIR2DL2/KIR2DL3/KIR2DS2 CCTTGCGTCCTTCACTACGACGAG KIR2DL3 AGGAGGAGGATGTGGAACTGCGCT KIR2DL5A CTTCAGGCCCATAACTCCACCACT KIR2DL5B / KIR3DP1 ATTTGGCGCCCTGCCCATGCTGC KIR2DP1 CTCCCCTCCAGGTTCCTATCGG KIR2DS1 CAACTCCACCTCCAGGCCTATTTA KIR2DS3 CAGCGCACAGGATGTTATTTGCCT KIR2DS3 / KIR2DS5 GTGCAGACAGGCGGCCGCACCGCT KIR2DS4 TGGGAGGGTGACGTACGCAGGGTT KIR3DL1*002, * , *062 TGGATGGGCCTGGAGGGGAGAT KIR3DL1*001, *004, *005, *059, *063 CGCTCATGGTCGTCAGCATGGC KIR3DL2 ATTCCCTTCCAGGACTCACCATCG KIR3DS1 TGGGTGCTCGCTCAAGAGCGCAG 71

74 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS 3 Supplementary Table 2. Overview of the number of predicted transcription factor binding sites in the last 300 bp of each intergenic KIR sequence. 2DL1-3DP1_A 2DL1-3DP1_B1 2DL2-2DL5B 2DL2-3DP1 2DL3-2DP1 2DL4-3DL1 2DL4-3DS1 2DL5A-2DS5 2DL5AB-2DS3 2DP1-2DL1_A 2DP1-2DL1_B1 2DS1-3DL2_B1 2DS1-3DL2_B2 2DS2-2DL2 2DS3-2DP1 2DS3S5-2DS1 2DS4N-3DL2 2DS4WT-3DL2 3DL1*004-2DS4trunc 3DL1-2DS4trunc 3DL1-2DS4WT 3DL3-2DL3 3DL3-2DS2 3DS1-2DL5A Predicted TF binding site AP1_bZIP ARID3A_AT-rich_interactive ARNT_bHLH BARHL_Homeobox BRCA1_Other CEBPA_BZIP_C/EBP E2F1_E2F_DP EBF1_EBF EN1_Homeobox ETS1_ETS FOXC1_Forkhead FOX_Forkhead GATA2_GATA GFI/HDX_Homeobox GMEB1/2_C2H HBP1_HMG_box HIC_C2H HLTF_RING HOXA5_Homeobox HOXD9_Homeobox

75 INTERGENIC REGIONS OF THE HUMAN KIR LOCUS IRX_Homeobox_TALE/IRO MAFB_bZIP-MAF MEIS_Homeobox_TALE/MEIS MYB_MYB-HTH MZF1_C2H NFATC2_RHD NFE2L1/MAFG_BZIP NFI_CTF/NF-I NFI_CTF/NF-I PAX2_Paired PROC1Homeobox Rhox11_Homeobox RHOXF1_Homeobox SIX_Homeobox SMAD3_MAD SOX10_HMG_SOX SOX1_HMG_SOX SOX_HMG_SOX SPIB_ETS SP/KLF_C2H STAT_STAT TBX_T-box YY1_YY Zfp105_Zinc_finger Zfp128_Zinc_finger ZFX_C2H ZNF354C_C2H

76

77 4 INTERLEUKIN (IL)-15 CONTRIBUTES TO VARIEGATED EXPRESSION OF KILLER IMMUNOGLOBULIN-LIKE RECEPTORS ON HUMAN ADULT AND NEONATAL NATURAL KILLER CELLS Sanne Vendelbosch, Judy Geissler, Farzin Pourfarzad, Martin de Boer, Timo K van den Berg, and Taco W Kuijpers Manuscript in preparation.

78 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS 4 ABSTRACT Cytotoxicity of Natural Killer (NK) cells is regulated by various receptor families, including the family of Killer Immunoglobulin-like Receptors (KIRs). The KIR gene locus is composed of highly homologous and polymorphic KIR genes with varying haplotype composition and gene copy number variation (CNV) across the human population. In addition, the level of surface expression on NK cells differs per individual, not only because of KIR genotype but also due to clonal expansion of individual NK cells and variegated KIR expression levels. To date, the exact regulation of KIR expression is poorly understood. We therefore studied the role of interleukin-15 (IL-15) and IL-2 on the surface expression of KIRs on circulating NK cells in fully KIR-genotyped individuals. Culture of NK cells in the presence or absence of these cytokines did not influence the proportion of KIR+CD56+ NK cells. Remarkably, the presence of IL-15, but not IL-2, did result in an increased KIR surface expression within the KIR+CD56+ NK cell fraction. This effect was observed in both adult and neonatal NK cells. KIR CNV influenced the proportion of circulating KIR+ NK cells, although the KIR expression level was unaffected for most KIRs apart from KIR3DL1. STAT5 phosphorylation in NK cells was observed upon IL-15 but not IL-2 stimulation, consistent with a role for STAT5 phosphorylation in IL15-induced KIR expression regulation. Our data indicate that IL-15 is a potent regulator of KIR expression on NK cells, thereby contributing to the regulation and activation of these during immune activation. 76

79 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS INTRODUCTION Natural Killer (NK) cells are granular lymphocytes that contribute both to innate and adaptive processes of the immune system. Their cytotoxic properties make them key players in the early response against tumor cells and viral infections, and there is growing evidence for the existence of some sort of memory within the NK cell population (Paust and von Adrian 2011). NK cells practice a constant immune surveillance by distinguishing pathogenic and aberrant cells via multiple families of immune receptors, the most prominent of which are the C-type lectins and the Killer Immunoglobulin-like Receptors (KIR). The genes for C-type lectins are always present and therefore have the potential to be expressed on the cell surface of every individual in the human population while the KIRs show extreme variation, both in genome content and gene copy number variation (CNV). The KIR genes are located in the leukocyte receptor complex locus (LRC) on chromosome 19q13.4 (Trowsdale et al. 2001). There are seventeen KIR genes; six activating (KIR2DS1-5, and KIR3DS1), nine inhibiting (KIR2DL1-4, KIR2DL5A, KIR2DL5B and KIR3DL1-3) and 2 pseudogenes (KIR2DP1 and KIR3DP1). When the KIR receptor is engaged by one of its ligands, such as a human leukocyte antigen (HLA) class I molecule (Biassoni 2001) the receptor confers an activating or inhibiting signal to the cell, depending on the make-up of their cytoplasmic tail. The tail of the activating receptors can interact with DAP12 which in turn contains immunoreceptor tyrosinebased activation motif (ITAM)-containing signaling molecules, whereas the long cytoplasmic tail of inhibiting receptors contains 1 or 2 immunoreceptor tyrosine-based inhibition motifs (ITIMs). Both motifs are phosphorylated by Src family kinases, upon which the ITIMs signal via recruited tyrosine phosphatases SHP-1 and SHP-2 and the ITAMs via Syk/Zap70 family protein tyrosine kinases (Purdy and Campbell 2009). The genetic complexity of this gene family, which is primarily characterized by an extensive level of haplotypic and allelic variation, gives rise to an unprecedented degree of diversity, for a non-rearranging innate immune receptor family, across the human population (Jiang et al. 2012; Vendelbosch et al. 2013). In addition to its genetic variation, KIR expression is highly variable across the population owing to different factors that are still not well understood. KIR expression on NK cells has been found to be regulated at different levels. First, expression is clonally distributed, i.e. every NK cell expresses its own KIR repertoire depending on the methylation status of each KIR promoter region in the cell (Uhrberg 2005). It is also thought that the expressed KIR repertoire within a given individual is influenced by education or licensing of the NK cells via a ligand instructed model by MHC class I molecules (Schönberg et al. 2010). Secondly, the expression may depend on allelic variation of the KIR gene, as demonstrated for the KIR3DL1 gene having alleles that have been related to no, low and high expression (Li et al. 2008; Pando et al. 2003; Trundley et al. 2006). Finally, environmental or infectious triggers have a direct effect on KIR expression by NK cells, as has been reported for prior cytomegalovirus (CMV) infection (Béziat et al. 2013a; Gumá et al. 2004). Finally, an important contribution to variegated KIR expression is the observed CNV at the KIR gene locus, which contributes to the proportion of KIR-expressing cells (Béziat et al. 2013b). This variation in KIR receptor repertoire across the population forms an intriguing factor that could influence disease susceptibility. Indeed, the KIRs have an important role in NK cell tumor killing during hematopoietic 4 77

80 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS 4 stem cell transplantation (HSCT) (Khakoo et al. 2004; Leung 2011; Venstrom et al. 2010) and the outcome of viral infections like human immunodeficiency virus type 1 (HIV-1) (Pelak et al. 2011) and hepatitis C virus (HCV) (Khakoo et al. 2004) is associated with certain KIR receptors. In addition, several gene association studies have related the KIR gene locus with susceptibility to autoimmune diseases such as psoriasis and rheumatoid arthritis (Kúsnierczyk 2013). NK cell development, survival, proliferation and activation is thought to be controlled by cytokines, mainly interleukin-2 (IL-2) and IL-15 (Becknell and Caligiuri 2005; Carson et al. 1994). Both cytokines have affinity for the IL-2Rβ subunit (CD122) in complex with the common gamma chain γc (CD132), while each receptor has its own unique α-subunit (IL-2Rα and IL-15Rα) for further ligand binding affinity regulation and signaling specificity (Becknell and Caligiuri 2005; Pillet et al. 2009). As a consequence, IL-2 and IL-15 have many overlapping effects on the regulation of lymphocytes, although they seem to have different effects on NK cells due to sequential expression of the receptor subunits (Pillet et al. 2011). The role of these interleukins in NK cell homeostasis and activation has been investigated thoroughly in mouse models and human in vitro studies. Although NK cells share many similar characteristics in both species, they diverge on several important aspects of NK cell biology. For instance, IL-15 is suggested to have a critical role on NK cell survival (Brilot et al. 2007; Huntington et al. 2007), and indeed, in mice and non-human primates, NK cell numbers decrease significantly upon blocking of IL-15 (Lebrec et al. 2013). However, Lebrec and colleagues showed that blocking of trans-presented IL-15 by anti-il-15 Ab AMG 714 in humans did not result in the same reduction in NK cell numbers (Lebrec et al. 2013). Another major difference between the two species is the fact that mice have a lectin-like family of closely related Ly49 molecules instead of the human KIR gene family. These Ly49 lectin members can similarly inhibit or activate murine NK cells, but cannot be structurally compared to KIRs (Berry et al. 2014). To fully understand the effect of cytokine activation of NK cells, the regulation of KIR expression has to be assessed in a human model. In this study, we investigated KIR expression level regulation of human NK cells. Complete genotyping of the healthy test subjects allowed for accurate analysis of KIR surface staining by flow cytometry, both in relation to presence of each gene and gene CNV. In this model, we studied the effects of IL-2 and IL-15, being the most relevant NK cell-responsive cytokines currently in use in human (pre)clinical studies. RESULTS Screening set-up to determine KIR surface staining KIR genotype and gene copy numbers were determined in 12 healthy donors and 12 neonates using a validated MLPA assay previously described (Vendelbosch et al. 2013) (Supplementary Table 1). Commercially available KIR antibodies were tested for their specificity by cell surface staining followed by flow cytometry in the presence of their corresponding KIR gene, resulting in a reliable set of antibodies for the accurate determination of KIR surface expression on NK cells. The validated antibodies were included in our flow cytometry panel for NK cell KIR cell-surface expression (Table 1). Peripheral blood mononuclear cells (PBMCs) were incubated with monoclonal antibodies against CD56 and CD3 to gate for NK cells (CD3 CD56+) and counterstained with validated KIR- 78

81 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS specific antibodies. Since expression of CD56 decreased upon cell culture, the CD56+ NK cell population became inseparable from the CD56 cell population. Therefore, KIR expression was measured on all SSClow FSClow CD3 cells (Supplementary Figure 1). This population constitutes of 40-60% NK cells and 30-60% B lymphocytes. These ratios remained constant throughout the culture. B lymphocytes do not express KIRs, as was confirmed by counterstaining for CD19 (data not shown), therefore all SSClow FSClow CD3 cells gated for KIR expression in these PBMC cultures represented human NK cells. In this study, only data from individuals genotyped positive for any expressed KIR gene with KIR+ populations >1% were included (thus excluding KIR3DL1*004 and the truncated alleles of KIR2DS4). Although all available KIR specific antibodies are listed in Table 1, including those targeting the same KIR protein, the results of only some antibodies are shown in the figures to avoid unnecessary repetition of data. 4 Table 1. Monoclonal antibodies tested for KIR and NK marker expression screening. Antibody Clone Fluorochrome * KIR - Reliable KIR2DL1/2DS1 EB6.B PE 1 KIR2DL2/2DL3 DX27 PE 4 KIR2DL2/2DL3/2DS2 GL183 PE 1 KIR3DL1/3DS1 Z PE 1 KIR3DL1/3DS1 Z APC 1 KIR2DS APC 2 KIR2DS4 FES172 PE 1 KIR2DL PE 2 KIR2DL unconjugated 2 KIR3DL1 DX9 PE 3 KIR3DL1/3DL2 5,133 PE 4 NK receptors NKG2A Z199 PE 1 NKG2C PE 2 NKG2D PE 2 CD94 HP-3B1 PE 1 CD3 UCHT1 Alexa fluor CD56 B159 APC 3 KIR - Unreliable KIR2DL5 UP-R1 PE 6 KIR2DL PE 2 KIR2DL3 190IIC311 unconjugated 7 * Source: 1. Beckman Coulter, 2. R&D Systems, 3, BD Biosciences, 4. Miltenyi Biotec, 5. Life Technologies, 6. LifeSpan Bioscience 7. American Research Products Inc. 79

82 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS 4 Increase in KIR expression on NK cells upon stimulation with IL-15 KIR protein levels were measured on freshly isolated PBMCs or after a three day culture in the absence or presence of IL-2 or IL-15. The percentage of KIR+ NK cells within the freshly isolated CD3 cell population varied between donors from 0-40% for all KIRs except for KIR2DL4. KIR2DL4 did not show any surface staining as it is located in NK cell endosomes (Rajagopalan 2010) (Figure 1A). C-type lectin expression was similarly variable among donors (Figure 1B). The surface expression level (measured as mean fluorescence intensity (MFI)) in the KIR+ and C-type-lectin+ cell population was also variable (Figures 1C and 1D). After a three day culture period, both IL-2 and IL-15 induce NK cell proliferation, as described earlier (Kuijpers et al. 2011; Kuijpers et al. 2013). The KIR+ NK-cell fraction did not change significantly, either with or without stimulation (Figure 2A). The KIR cell-surface expression levels of unstimulated and IL-2-stimulated cells remained equal or decreased slightly after a three day culture, whereas KIR cell-surface expression on NK cells showed a significant increase upon incubation with IL-15 (Figure 2B). This increase in KIR expression level was observed for all KIRs in any given individual (depending on the individual s KIR genotype). After a 6 day culture period, the same observations regarding KIR expression in response to cytokines were made (data not shown). Moreover, the observed changes in short-term NK cell activation in PBMC cultures were repeated with highly purified NK cells, confirming our findings with PBMCs (data not shown). A B C MFI positive cells % positive cells D Figure 1. Variegated KIR expression levels in healthy individuals. (A) The percentage of KIR positive CD3 PBMCs from 12 healthy Caucausian KIR-genotyped donors, measured by flow cytometry. (B) The percentage of C-type lectin positive cells in the same donors as in A. (C) The KIR expression level and (D) the C-type lectin expression level as expressed in mean fluorescent intensity (MFI). 80

83 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS A 4 B Figure 2. IL-15 mediated KIR expression on adult cells. (A) The percentage of KIR positive CD3 PBMCs of 12 donors after 3-day culture. The ratio is calculated by comparing the values after culture with or without cytokines with the values of freshly isolated cells from the same donor (dotted line). (B) KIR expression level, expressed as MFI, on cells cultured without stimulus, with IL-2 or with IL-15 normalized to fresh cells (dotted line). Statistically significant differences in KIR staining compared to unstimulated cells: (***) p<0.001, (**) p<0.01 (*) p<0.05. IL-15 stimulation increases KIR expression level on neonatal NK cells Apart from changes in adult cells, neonatal PBMCs isolated from cord blood showed a similar response in KIR and C-type lectin protein expression level to IL-15 (Figures 3 and 4). The overall percentages of KIR+ NK cell populations seemed slightly lower, while the cell-surface expression levels of some KIRs were slightly higher than in adults. This was not significantly 81

84 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS 4 A C MFI positive cells % positive cells B D Figure 3. Neonate NK cell KIR expression levels. KIR expression levels on freshly isolated CD3- PBMCs from the cord blood of 12 KIR-genotyped neonates measured by flow cytometry. (A) The percentage of KIR+ cells and (B) C-type lectin positive cells. (C) The expression levels of KIR+ cell populations and (D) the C-type lectin expression levels of the positive cell populations expressed as mean fluorescent intensity (MFI). different apart for KIR3DL1/DS1 and KIR3DL1/L2 when staining with MoAbs (Figures 3A and 3C). The clone that stains KIR3DL1 specifically, DX9, does not show higher MFI, suggesting significantly higher surface expression levels of KIR3DS1 and KIR3DL2 (KIR3DL1/S1 p= and KIR3DL1/L2 p<0.0001) on neonatal NK cells compared to adult NK cells. As previously reported, the percentages of the C-type lectins NKG2A and CD94 were increased on neonatal NK cells compared to adult NK cells, whereas the level of C-type lectin surface expression was similar (Figures 3B and 3D) (Lee and Lin 2014; Schönberg et al. 2011; Wang et al. 2007). When stimulated with IL-15, the percentage of KIR+ cells did not change, whereas all KIRs showed increased surface expression levels compared to unstimulated cells (Figures 4A and 4B). Other NK cell markers affected by IL-15 IL-15 strongly influences lymphocyte cell survival, proliferation and cytotoxicity (Dunne et al. 2001; Huntington 2014). To investigate NK receptor expression and changes herein upon IL-15 stimulation we studied the surface expression levels of some characteristic NK cell markers. Unstimulated 3-day culture of NK cells resulted in a decreased CD56 surface expression level, while surface expression levels were upregulated upon stimulation with IL-15, but not when stimulated with IL-2 (Supplementary Figure 2A). The MFI of NKp46 and CD16 did not change upon culturing of the NK cells, with or without cytokines (data not shown). Neonatal NK cells 82

85 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS A 4 B Figure 4. IL-15 mediated KIR expression on neonate cells. Cord blood derived PBMCs from 12 neonates were stained for KIR expression before and after 3-day culture. The ratio is calculated by comparing the values after culture with or without cytokines with the values of freshly isolated cells from the same donor (dotted line). (A) Percentage of KIR positive cells. (B) KIR expression level, measured in MFI. Statistically significant differences in KIR staining were found with p<0.01 (**) and p<0.05 (*). showed a similar pattern in CD56 surface expression level after 3-day culture with or without IL-15 compared to adult NK cells (Supplementary Figure 2B). The surface expression levels of the different members of the C-type lectin family on adult cells were not equally affected by the cytokines. The percentages of all C-type-lectin+ NK cells remained equal except for CD94+ cells, which showed and increased positive population upon IL-2 or IL-15 stimulation compared to unstimulated cells (Supplementary Figure 3A). Surface 83

86 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS 4 expression levels of CD94 and the inhibitory NKG2A increased significantly upon stimulation with IL-15 (but not IL-2) compared to unstimulated cells, while surface expression of the activating NKG2C did not change. Similarly, the surface expression of the activating NKG2D was upregulated upon IL-15 stimulation (Supplementary Figure 3B). Since KIRs have been reported to be expressed on T cells (Anfossi et al. 2004; Battistini et al. 1997; van Bergen et al. 2009), we determined various KIRs on CD3+/CD8+ T cells. The percentage of KIR+CD8+ T cells was found to be very low (<1%) (data not shown) and therefore does not substantially affect our findings in NK cells. KIR protein expression level is related to KIR gene CNV The proportion of KIR+ NK cells, but not KIR surface expression level per se seems to be influenced by KIR gene copy numbers, as Beziat and colleagues have suggested before (Béziat et al. 2013b). Indeed, in our experimental set-up the percentage of KIR+ NK cells correlated with the respective KIR gene CNV for most of the KIRs when analyzed on fresh cells (Figure 5A). Any differences in results between the studies of Beziat and colleagues and our study (like the gene-dose effect of KIR2DL1) are likely the result of the study cohort size. Stimulation with IL-15 in a 3 day culture did not affect the percentage of KIR+ NK cells relative to the relation between gene CNV and KIR+ NK cell fractions; the gene-dose effect remains after IL-15 stimulation. The cell surface expression level of KIRs was not dependent on their gene copy number, either being measured on fresh cells or on IL-15-stimulated cells (Figure 5B). However, we did find an exception to the general lack of correlation between CNV gene dose effect and cell A B Figure 5. Effect of IL-15 and gene copy number variation (CNV) on KIR expression level. KIR expression on CD3 PBMCs of 12 KIR-genotyped donors measured by flow cytometry. (A) IL-15 induced percentage of KIR positive cells and (B) MFI of KIR positive cells compared to gene copy number. Different monoclonal antibodies were used to stain for KIR3DL1 (clone DX9) or both KIR3DL1/L2 (clone 5.133). Statistical significance: (**) p<0.01 and (*) p<

87 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS surface KIR expression level for KIR3DL1. The protein expression levels of KIR3DL1 and KIR3DL2 in fresh cells and IL-15-stimulated cells, measured in MFI with clone 5.133, seemed positively related to the number of KIR3DL1 gene copies. This correlation between gene CNV and KIR3DL1 cell surface expression was weak when stained with MoAb clone DX9, either before or after IL-15 stimulation. In conclusion, IL-15 did not change any relative correlation between KIR expression level and KIR gene copy number. Allelic variation contributes to KIR expression level variation The effect of KIR3DL1 gene copy number on its cell surface expression may also be affected by allelic variation. Although the phenomenon of KIR gene allelic variation has best been described to date for the KIR3DL1 gene, which contains alleles with high, low or no expression (Pando et al. 2003), it may be true for more KIR genes. The KIR2DS4 gene has an allelic variant that is not expressed at the cell surface but is thought to be translated into a secretory soluble protein (Middleton et al. 2007). To investigate the relation between allele, transcription and translation, we focused on these two genes. Freshly isolated NK cells from two individuals with the same genotype; a low- and a high-expression allele of KIR3DL1 and both the full-length (KIR2DS4) and truncated (KIR2DS4trunc) form of KIR2DS4, were sorted in NK cells with no, low or high KIR3DL1 cell surface expression (Figure 6A). While KIR2DS4 transcript quantity (normalized to NKp46 transcripts) was variable in each KIR3DL1+ NK cell subpopulation (Figure 6B), the relative KIR3DL1 transcript level was proportional to the protein surface expression in each population (Figure 6B). The truncated form of KIR2DS4trunc can be linked to either an expressed or a nonexpressed allele of KIR3DL1, KIR3DL1expr and KIR3DL1*004 respectively. The KIR3DL1*004 allele was never found in combination with a full-length KIR2DS4 allele (Vendelbosch et al. 2014). Measuring transcript levels of these non-protein expressed alleles, we observed that KIR2DS4 is hardly transcribed when linked to the expressed alleles of KIR3DL1 (Figure 6C), whereas transcription is much higher when linked to the non-expressed KIR3DL1*004. The difference in linking partner seems to affect the transcriptional activity of each allele. Since KIR transcription levels seemed to be directly proportional to KIR protein levels, the variable transcriptional activity may depend on differential transcription factor recruitment. The STAT (signal transducer and activator of transcription) transcription factor binding motif is present in the promoter of every single KIR gene and KIR-allele. As STAT members are reported to be regulated by IL-2 and IL-15, this is of interest for our cultures with these cytokines. 4 IL-15 induces STAT5 phosphorylation in NK cells Both IL-2 and IL-15 have been implicated in NK cell development through STAT5 phosphorylation (Mishra et al. 2014; Pillet et al. 2011). Moreover, KIR3DL1 reverse promoter activity in transfected YT NK cells by IL-2 and IL-15 was shown to require STAT5 activation by Presnell et al (Presnell et al. 2013). We therefore investigated whether this transcription factor was also responsible for the IL-15 induced KIR expression of circulating human NK cells. We incubated purified NK 85

88 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS A neg low high 4 B KIR2DS4 KIR3DL1 C Figure 6. Variegated KIR expression levels and allelic variation. NK cells from donors with different alleles for KIR3DL1 and KIR2DS4 were studied for KIR transcript levels. (A) NK cell populations of 2 donors with low and high expressing KIR3DL1 alleles and functional and truncated KIR2DS4 alleles were sorted into KIR3DL1 negative (neg), low expressing and high expressing cells. (B) KIR3DL1 and KIR2DS4 transcript of each of these populations, as in A (P2+P3 = neg, P4+P5= low and P6+P7= high), normalized to NKp46. (C) Relative transcript levels of NK cells from donors with at least one allele carrying a truncated KIR2DS4 allele coupled to a non-functional KIR3DL1*004 allele, compared to the transcript levels from donors with at least one allele carrying a truncated KIR2DS4 allele coupled to an expressing KIR3DL1 allele (KIR3DL1expr). (**) p<0.01. cells and NK cell-depleted PBMCs with either IL-2 or IL-15 and measured STAT5 phosphorylation levels. Stimulation with IL-2 or IL-15 both increase STAT5 phosphorylation in NK cell-depleted PBMCs. However, only IL-15 induces STAT5 phosphorylation in NK cells (Figure 7). The closely related STAT family member, STAT3, is not phosphorylated by either IL-2 or IL-15 in NK cells, while IL-15-induced STAT3 phosphorylation was increased in NK cell-depleted PBMCs. Although proliferation was induced by both of these growth factors (Dunne et al. 2001; Huntington 2014), these data demonstrate an NK cell differential STAT5 phosphorylation and KIR expression response to IL-2 and IL

89 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS pstat3 NK cells PBMCs ( NK cells) unstim IL-2 IL-15 unstim IL-2 IL-15 Tubulin pstat5 Tubulin 4 Figure 7. Differential phosphorylation of STAT5 by IL-2 and IL-15 in human lymphocytes. STAT3 and STAT5 phosphorylation of IL-2 and IL-15 stimulated NK cells and NK cell-depleted PBMCs from the same donors. Red: STAT3 or STAT5. Green: tubulin as loading control. MATERIALS AND METHODS Ethics statement Participants provided their written informed consent to participate in this study. For the cord blood samples, written informed consent from the mothers was obtained for all samples. The studies on both sample types were approved by the Institutional Medical Ethics Committee of the Academic Medical Center in Amsterdam and were performed in accordance with the Declaration of Helsinki. Genotyping DNA was extracted from the blood of healthy individuals or cord blood from neonates, using the QIAgen Blood Kit (Qiagen, Frederick, MD, USA) according to the manufacturers protocol. KIR genotyping and copy number determination was performed using the KIR multiplex ligation-dependent probe amplification assay as described (Vendelbosch et al. 2013). The (non-expressed) KIR3DL1*004 allele and all other (expressed) KIR3DL1 alleles were genotyped separately as described earlier (Vendelbosch et al. 2014). Primary cell isolation and culture Peripheral blood mononuclear cells (PBMCs) were isolated from the blood of healthy volunteers by Percoll density gradient centrifugation. Neonatal PBMCs were isolated from cord blood by Percoll density gradient centrifugation. NK cells were isolated from the PBMC fraction by negative selection to >95% purity using MACS human NK cell isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany). All experiments were performed with freshly isolated cells or cells cultured directly after isolation. Measurements of cultured cells are always compared with the measurements of freshly isolated cells of the same donor and the same blood withdrawal. 87

90 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS Primary cells were cultured with a density of 1x106 cells/ml in complete IMDM (PAA, Pasching Austria) supplemented with 10% FCS (Bodinco, Alkmaar, The Netherlands), 100 Units/ ml Penicillin, 100 μg/ml streptomycin and 2mM L-glutamine (PAA) at 37 C, 4%CO2. Primary cells were incubated with 20 Units/ml IL-2 or 1 ng/ml IL-15 where indicated (Sanquin Reagents). Sorting of specific KIR positive NK cell populations was performed on a BD FACSAria II Cell Sorter (BD Biosciences). 4 Flow cytometry Monoclonal antibodies against KIR and cell biomarkers obtained from companies as indicated (see also Table 1); α-cd8 clone 4H8 (Sanquin Reagents, Amsterdam, The Netherlands), α-cd19 clone HIB19 from Biolegend (San Diego, CA, USA), α-cd16 clone 3G8 from BD Biosciences (Franklin Lakes, NJ, USA), Beckman Coulter (Woerden, the Netherlands), R&D Systems (Minneapolis, MN, USA), Miltenyi Biotec, Life Technologies (Carlsbad, CA, USA), LifeSpan Bioscience (Seattle, WA, USA) or American Research Products Inc. (Waltham, MA, USA). Cells were measured on a FACS CANTO II (BD Biosciences) and data analyzed with BD FACSDiva Software (v7.0, BD Biosciences). Western-blotting Primary cells were isolated and left overnight in IMDM medium at 37 C, 4%CO2 and subsequently incubated with 20 Units/ml IL-2 or 1 ng/ml IL-15 for 15 min at 37 C, 4%CO2. Aliquots of 1x106 cells were transferred to ice and kept cold while centrifuged; pellets were taken up in lysis buffer (PBS containing 10 mm EDTA, 1x HALT (Thermo Fisher Scientific, Loughborough, UK)). Laemmli sample buffer was added (containing 50 mm Tris-HCl, ph6.8, 10% (v/v) glycerol, 5 mm DTT (DL-dithiothreitol, Sigma Aldrich, Saint Louis, MO, USA), 1% β-mercapthoethanol, 1% (w/v) SDS (sodiumdodecylsulphate, Sigma) and 100 μg/ml bromephenol blue) and samples were kept at 95 C for 30 min. Samples were loaded on 10% SDS page gels and blotted on nitrocellulose membranes. Nitrocellulose membranes were blocked in BSA 5% BSA (Sigma Aldrich) and stained using rabbit α-pstat3 clone 79D7 (Cell Signaling Technology, Danvers MA, USA), rabbit α-pstat5 clone 6HCLC (Thermo Fisher Scientific) and mouse α-tubulin clone DM1A (Sigma Aldrich). Antibody binding was visualized with IRDye 680LP and IRDye 800CW (Li-Cor, Lincoln, NE, USA) secondary antibodies on an Odyssey CLx (Li-Cor). Real time quantitative PCR Total RNA was isolated using the QiaAmp RNA Blood kit mini (Qiagen) according to manufacturers instructions. Subsequently, reverse transcription was performed with Superscript III first strand synthesis system for RT-PCR (Life Technologies). Detection and quantification was performed according to the method described before (van Mirre et al. 2006). Primers for KIR RNA quantification were adjusted from Chen et al. (Chen et al. 2008) KIR3DL1-FW (5 -GTGGTCGGCACCCAGCAA-3 ), KIR3DL1-REV (5 - AGCATC TGTAGGTCCCTGCAAGGGGAA-3 ), KIR2DS4WT-FW (5 -TCATCCTGCAATGTTGGACG-3 ), KIR2 DS4WT-REV (5 -TACATGTCATAGGAGCTCCG -3 ), KIR2DS4trunc-FW (5 -ATGGCGTGTGTTGGG 88

91 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS TTCTTC-3 ), KIR2DS4trunc-REV (5 -CCCCTCCCTGGATAGATCGA-3 ) and primers to detect NKp46 were described by McErlean et al. (McErlean et al. 2010): NKp46-FW (5 -CCGCCC AGGCTCAACACC-3 ), NKp46-REV (5 -ACGGGACTCCAGAAAGACCAT-3 ). Quantification was performed with the Lightcycler 2.0 Instrument (Roche Diagnostics, Branford, CT, USA). Statistical analysis Statistical differences were determined with one-way Anova followed by Bonferonni for multiple comparisons. Significance levels found were indicated with *** = p<0.001, **= p<0.01 and *= p<0.05. DISCUSSION In this study we have investigated KIR expression levels on NK cells in relation to the individual genotype. KIR expression level not only depends on clonal NK cell expansion and genotypic variation, but also on CNV for certain KIR genes or allelic variation for some of these genes. The latter was exemplified for KIR3DL1 and KIR2DS4. Upon activation with IL-15 or IL-2 with respect to the regulation of NK cells it is now well established that the two cytokines although binding to similar cytokine receptors have very different effects on pre-activated NK cells (Becknell and Caligiuri 2005; Pillet et al. 2011). Here, we show that stimulation with IL-15, but not IL-2, significantly increased KIR surface expression level on resting NK cells, although both resulted in a similar proliferative and survival response. Similarly, IL-15, but not IL-2, response on KIR expression on human NK cells is correlated with STAT5 phosphorylation. Apart from KIRs, CD56 was also upregulated by IL-15, but not by IL-2, in a similar fashion to the KIR receptors. In contrast, expression of NKp46 and CD16 were not significantly affected. Of the C-type lectins, expression of NKG2D and CD94 were significantly increased upon IL-15 stimulation compared to fresh cells. NKG2A was upregulated upon IL-15 activation, whereas NKG2C responded differently. NKG2D is an activating receptor that in contrast to either NKG2A or NKG2C does not dimerize with CD94 and therefore acts independently. CNV has been suggested to influence clonal distribution, or education, of KIRs on NK cells (Béziat et al. 2013b). We were able to confirm these findings in freshly isolated circulating NK cells and demonstrate a similar correlation between KIR gene CNV and the proportion of the KIR+ cell population before and after IL-15 stimulation. This suggests that all NK cells, irrespective of clonal expansion, may show an equal proliferative response to IL-15. In addition, our data suggest that CNV and gene-dose are correlated to the level of KIR3DL1 protein expression on NK cells, which was not observed for other KIRs. In addition to the CNV of KIR3DL1, allelic variation is another major factor that impacts the level of cell surface expression for this KIR. We found that the level of KIR3DL1 transcripts is proportional to the level of protein expression in sorted NK cell fractions, suggesting that lower expression of this protein was due to lower transcription rather than less efficient protein folding or membrane anchoring. Apart from the allelic variation of the KIR3DL1 gene, we also observed allelic variation for the truncated form of KIR2DS4. Although no known function for the truncated protein has been described to date, it has been suggested that the protein is expressed and excreted from the cell (Middleton et al. 2007). 4 89

92 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS 4 How genomic variations of KIR alleles are involved in their transcriptional regulation at the cellular level remains poorly understood. We have observed that the KIR transcript level was directly proportional to KIR protein expression. One contributing factor to KIR protein expression variation could be the presence of polymorphisms in the promoter region contributing to differential binding of transcription factors to alleles of the same KIR gene. Another factor may be the different clonal distribution and expansion of NK cells through education by self-hla ligands processes that may occur at different locations in the lymphoid system at a variable rate throughout life. Yet, in neonatal NK cell cultures we observed the same clonal diversity and variegated KIR protein expression levels as in adult NK cell cultures. Also, the effects of IL-2 and IL-15 on KIR expression levels were identical to those of adult cells. Neonatal NK cells have been shown to be less cytoxic due to an increased NKG2A+/CD94+ population compared to adult NK cells (Wang et al. 2007). The cytotoxicity of neonatal NK cells can be increased upon IL-15 stimulation similar to adult NK cells (Choi et al. 2004), while we have shown here that IL-15 also increases KIR expression levels. Although inhibitory interactions between KIR and their cognate HLA class I ligands abrogate effector responses of NK cells, they are also, somewhat paradoxically, required for the functional education of NK cells in a process referred to as NK cell licensing. However, there is no bias for expression of self-specific KIRs in neonates or healthy donors (Béziat et al. 2013b; Schönberg et al. 2011). Even though functional education was reported to promote NKG2C-driven proliferation after infection (Béziat et al. 2013b), the circulating fraction of NKG2C+ NK cells in adults is generally small after viral infection. In this respect, the KIR expression repertoire would be of great interest to assess on tissuederived NK cells, for instance from liver biopsies, to compare with the circulating NK cells. Together, these data suggest that although education and environmental factors may influence clonal distribution of KIR expression during childhood, the variegated KIR protein expression on NK cells is already largely embedded in the human genome. Certain environmental factors such CMV infection (Béziat et al. 2013a; Gumá et al. 2004), have been found to contribute to this variation although largely limited to the NKG2C+ NK cell population, which in our donors remained <10% of the CD3- PBMC fraction on average. Following CMV, randomly formed KIR repertoires are believed to become imprinted by the virus infection (Béziat et al. 2013a). The biased expression of self-specific inhibitory KIRs has been suggested not only for CMV but also in case of hantavirus, hepatitis viruses or HIV (Béziat et al. 2013a; Björkström et al. 2011). Although selective NKG2C and/or KIR+ NK cell expansion has also been induced in vitro by CMV-infected cells (Béziat et al. 2013a; Gumá et al. 2004). Upon comparing KIR expression levels between CMV-seropositive and CMV-negative individuals, we did not observe a difference in KIR or NKG2C protein expression levels, either before or after IL-15 stimulation (data not shown). Moreover, all (CMV-negative) neonatal blood samples had similar variegated KIR expression. Hence, the association between KIR expression levels and CMV-seropositivity may not be as absolute as suggested previously. Finally, our study demonstrated a clear difference in IL-15 versus IL-2 in activating NK cells. Some earlier studies have suggested that IL-15 and IL-2 have a similar effect on NK cell proliferation and KIR expression. A study by Romagnani and co-workers resulted in an increase 90

93 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS of the percentage of KIR+ NK cells derived from peripheral blood and lymph nodes after stimulation with IL-2 or IL-15 (Romagnani et al. 2007). Also, work by de Rham et al showed an increase and de novo upregulation of KIR expression after IL-2 stimulation (de Rham et al. 2007). The actual reasons for these apparent differences are not known at present. Previous studies have also suggested that IL-2 can phosphorylate STAT5 in a similar fashion to IL-15. Although future studies will have to be conducted to show direct interaction between STAT5 and the KIR promoter, we observed a difference in the ability of IL-15 and IL-2 to induce STAT5 phosphorylation in NK cells. The work from Presnell et all (Presnell et al. 2013) in particular, suggested that initiation of transcription of the KIR3DL1 reverse promoter by IL-2 or IL-15 required STAT5 phosphorylation. In our study IL-15 induced STAT5 phosphorylation in circulating primary NK cells and NK cell depleted PBMCs. IL-2 induced STAT5 phosphorylation in NK cell depleted PBMCs, showing that IL-2 was able to reach the cell surface receptors of the cells. However, IL-2 did not induce STAT5 phosphorylation in NK cells. The main difference in experimental set-up is the origin of the cell type that was used for these experiments. Presnell et al used reporter plasmids in transfected YT NK cells. Others who suggested that IL-2 can induce STAT5 phophorylation have used T-cells or IL-15 pre-activated NK cells (Johnston et al. 1995; Mahmud et al. 2013; Pillet et al. 2009). The differential effects of IL-2 and IL-15 on NK cells described here may be related to the sequential expression of IL-15Rα and IL-2Rα (Pillet et al. 2009). As IL-2Rα is not readily expressed on circulating NK cells, there is no STAT5 phosphorylation in IL-2 stimulated NK cells. Since STAT5 phosphorylation was induced in IL-2 stimulated NK cell-depleted PBMCs, the mechanism of sequential cytokine receptor expression is specific for NK cells. Upon immune activation various cells release cytokines to mount a rapid immune response (Huntington 2014). NK cell cytotoxicity is enhanced through binding and activation by IL-15 which increases the synthesis and release of perforin and granzymes (Mishra et al. 2014). It also induces increased anti-viral responses in NK cells as was shown for HIV (Gosselin et al. 1999; Lum et al. 2004). An NK cell may respond to reduced levels of HLA-class I expression or because of recognition of as yet unidentified ligands or stress-induced molecules with their activating KIR (Jobim M and Jobim L F 2008; Katz et al. 2004). Our data suggest that IL-15 contributes to regulation and activation of NK cells through an increased surveillance of other cells and therefore the NK cell is better able to sense potential threats and able to mount a faster and more efficient cytotoxic response through its activated state when needed. 4 91

94 IL-15 ENHANCES KIR EXPRESSION ON NK CELLS SUPPLEMENT 4 A CD56 NK culture Fresh Day 1 Day 2 Day 3 B PBMCs SSC FSC NK cells KIR positive cells CD56 Count CD3 KIR Supplementary Figure 1. CD56 expression of cultured NK cells decreases over time. (A) CD56 expression measured on fresh cells and cells in culture (without stimuli) for 1, 2 or 3 days (B) Gating strategy KIR positive CD3- cells. 92