Synthesis and Characterization of Citrate and Polymer Stabilized Lanthanide Trifluoride Nanoparticles

Transcription

1 Synthesis and Characterization of Citrate and Polymer Stabilized Lanthanide Trifluoride Nanoparticles by Rohan Alvares A thesis submitted in conformity with the requirements for the degree of Master of Science Graduate Department of Chemistry University of Toronto Copyright by Rohan Alvares 2009

2 Synthesis and Characterization of Citrate and Polymer Stabilized Lanthanide Trifluoride Nanoparticles Rohan Alvares Master of Science Department of Chemistry University of Toronto 2009 Abstract Citrate-coated gadolinium trifluoride (Cit-GdF 3 ) and poly(acrylic acid)-coated nanoparticles (PAA-GdF 3 NPs) were synthesized, the former reproduced from literature (though using more refined conditions), the latter through a new, two-step, ligand exchange method. Diamagnetic nanoparticle analogs (Cit-YF 3 NPs) were prepared to investigate citrate interactions with the nanoparticle surface using NMR. Citrate was found to bind in numerous conformations, with a total of between % bound at 0 ºC. Exchange studies revealed short residence lifetimes of one and twelve seconds respectively for bound and free forms of citrate (0 ºC), perhaps explaining the colloidal instability of these nanoparticles. PAA-GdF 3 NPs were synthesized by first producing their Cit-GdF 3 counterparts, and then exchanging citrate for PAA. The impetus behind this latter synthesis was the relative enhancement in stability and relaxivity attainable by these nanoparticles. The displacement of citrate by PAA was verified using diffusion NMR studies. ii

3 Acknowledgements My sincere thanks goes to Prof. P. M. Macdonald, Prof. R. S. Prosser, Dr. Ferenc Evanics and Dr. Ronald Soong for teaching me the basics of operating the NMR spectrometers and helping execute experiments. I would further like to thank Prof. Macdonald and Prof. Prosser for providing me with sound guidance, giving me free reign (within limits) to design my own experiments and exercise thought, and also for shaping a wayward undergraduate into a more complete, thorough, organized and sceptical scientist. Although the nanoparticle ligand exchange protocol was developed independently, with guidance from Professors Macdonald and Prosser, equal credit goes to Evelyn Cheung for working seamlessly and in conjunction with me in developing it. She has since undertaken the development of much needed size and dispersity improvements to the syntheses while I focussed exclusively on finishing the NMR based characterization of the citrate-coated lanthanide trifluoride nanoparticles. I would also like to thank my colleagues in the Prosser and Macdonald labs for their company and for helping to create a safe and organized, yet enjoyable working environment. Among my colleagues, I would once again like to single out Evelyn Cheung who soldiering with me through highs and lows in the lab, and with whom I shared some insightful and stimulating discussions about the nanoparticle project, princesses and life in general. Lastly, and most importantly, I would like to extend my sincere gratitude to my family for their unwavering support and understanding. iii

4 Table of Contents Abstract...ii Acknowledgements...iii Table of Contents...iv List of Figures...vi List of Tables...vii Abbreviations...viii 1 Introduction Nanoparticle size - Tumour angiogenesis and clearance pathways Nanoparticle size Blood-brain barrier considerations Relaxivity MRI contrast agents Iron Oxide Nanoparticles Gadolinium chelates (Gd-chelates) Albumin-(gadolinium-DTPA) complexes and MS-325 (Gadofosveset) Viral MRI contrast agents Dextran-(Gd-DTPA) Liposomes Dendrimers Metallofullerenes Zeolites Gadolinium-based Nanoparticles Water Soluble Lanthanide Nanoparticle Synthesis Coprecipitation Polyol-Mediated Nanoparticle Synthesis Microemulsion methods Hydrothermal synthesis Thermodynamics of nanoparticle formation Thermodynamics La Mer model of nucleation and growth NMR experiments Pulsed Field Gradient Stimulated Echo (PFGSTE) experiment Correlation Spectroscopy (COSY) Exchange Spectroscopy (EXSY) Selective Inversion Recovery (SIR) Experimental Section Materials Methods Citrate coated lanthanide nanoparticle synthesis PAA coated lanthanide nanoparticle synthesis Electron Microscopy Nanoparticle Size Analysis Zeta potential measurements NMR iv

5 3 Results and Discussion Ligand Properties citrate and poly(acrylic acid) (PAA) Nanoparticle synthesis procedures Citrate-coated lanthanide trifluoride nanoparticle synthesis Poly(acrylic acid)-coated lanthanide trifluoride nanoparticle synthesis Synthesis Verification and Characterization of Size Polydispersity PAA 25 -LnF 3 NP - Direct Synthesis Images Control of Nanoparticles Size and Dispersity Rate of Lanthanide Addition Reduction in Total Lanthanide Feedstock Lump-sum lanthanide stock additions NMR Studies One-Dimensional Proton (1D 1H) Spectra of Citrate and Cit-YF 3 Nanoparticles Correlation Spectroscopy (COSY) Exchange and Population Calculations - Diffusion Studies Exchange spectroscopy (EXSY) Selective Inversion Recovery Verification of PAA ligand exchange Conclusion Supporting Data Supporting Data (for Figure 3.11) Supporting Data (for Figure 3.12) References v

6 List of Figures Figure 1.1 Size dependence of contrast agent diffusion in blood vessels...3 Figure 1.2 Tissue permeability size dependence...4 Figure 1.3 Pathways across the blood brain barrier...6 Figure 1.4 First and second coordination spheres of a gadolinium complex...7 Figure 1.5 Various contrast agents chelate, MS-325, virus, dextran...13 Figure 1.6 Various contrast agents liposome, dendrimer, fullerene, zeolite...16 Figure 1.7 Nanoparticle synthesis - reverse micelle...20 Figure 1.8 Thermodynamics of nanoparticle formation critical size...23 Figure 1.9 La Mer diagram colloidal nucleation and growth...24 Figure 1.10 Pulsed field gradient stimulated echo (PFGSTE) pulse sequence...25 Figure 1.11 Multiple quantum suppression CHIRP based z-filter...26 Figure 1.12 Correlation spectroscopy (COSY) NMR pulse sequence...27 Figure 1.13 COSY spectrum...28 Figure 1.14 Exchange spectroscopy (EXSY) NMR pulse sequence...29 Figure 1.15 Selective inversion recovery (SIR) pulse sequence...30 Figure 3.1 Citric acid and poly(acrylic acid) structures...36 Figure 3.2 PAA-GdF3 nanoparticle STEM images and size dispersions...41 Figure /10 Gd/EuF 3 energy dispersive X-ray linescan...42 Figure 3.4 STEM images variation in lanthanide rate of addition...45 Figure 3.5 STEM images reduction in total lanthanide feedstock...47 Figure 3.6 STEM images lump-sum addition of lanthanide feedstock...48 Figure 3.7 STEM images Cit-YF 3 and PAA-YF 3 nanoparticles...50 Figure 3.8 NMR spectra one dimensional proton results...52 Figure 3.9 NMR spectra one dimensional proton temperature variation...56 Figure 3.10 Citrate carboxylate coordination modes...57 Figure 3.11 NMR spectra Cit-YF 3 nanoparticle COSY spectrum...59 Figure 3.12 NMR spectra 1:1, Y:citrate COSY control spectrum...60 Figure 3.13 NMR spectra 1:10, Y:citrate COSY control spectrum...61 Figure 3.14 NMR spectra Diffusion Peak Set choices...64 Figure 3.15 NMR spectra PFGSTE diffusion decay of Cit-YF 3 nanoparticles...64 Figure 3.16 Logarithmic plot of PFGSTE citrate peak decays...65 Figure 3.17 Example of a simulated diffusion decay...65 Figure 3.18 NMR spectra EXSY spectra...67 Figure 3.19 NMR spectra SIR experiment...71 Figure 3.20 CIFIT fits of SIR results...71 Figure 3.21 NMR spectra - PFGSTE diffusion decay of PAA-YF 3 nanoparticles...73 Figure 3.22 Logarithmic plot of PFGSTE PAA peak decays...73 vi

7 List of Tables Table 3.1 Nanoparticle size distribution...44 Table 3.2 Comparison of citrate NMR parameters...54 Table 3.3 Comparison of citrate binding states...62 Table 3.4 PAA-YF 3 diffusion summary...74 Table SD1 Chemical shift assignments for Cit-YF 3 nanoparticles...77 Table SD2 Chemical shift assignments for 1:1, Y:citrate control...78 vii

8 Abbreviations Roman letters 1D1H one dimensional proton (spectrum) AB strong coupling AX weak coupling BBB blood-brain barrier BBTB blood-brain tumour barrier BCNU 1,3-bis(2-chloroethyl)-1-nitrosourea or carmustine CA contrast agent CCMV cowpea chlorotic mottle virus CT computed X-ray tomography Cit-GdF 3 NPs citrate-coated gadolinium trifluoride nanoparticle Cit-LnF 3 NPs citrate-coated lanthanide trifluoride nanoparticles COSY correlation spectroscopy D bound diffusion coefficient of the bound state D free diffusion coefficient of the free state D obs observed diffusion coefficient DTPA diethylene triamine pentaacetic acid E o standard reduction potential EDX energy dispersive X-ray EXSY exchange spectroscopy g amplitude of the gradient in the PFGSTE experiment g(n) free energy of the aggregate g B free energy of a bulk molecule g S free energy of an interfacial molecule G2 dendrimer generation, where the number be 2 or larger Gd-DPTA gadolinium-diethylene triamine pentaacetic acid chelate (Magnequist ) HD hydrodynamic diameter i a number (1 or 2) referring to either the spin lattice or spin-spin relaxation mechanisms k Boltzmann s constant k CM exchange constants when going from State C to State M k DM exchange constants when going from State D to State M k MC exchange constants when going from State M to State C k MD exchange constants when going from State M to State D LMCA low molecular weight contrast agent MMCA macromolecular contrast agent MPION micrometer-sized paramagnetic iron oxide nanoparticle MRI magnetic resonance imaging MS-325 a small molecule with an albumin binding diphenylcyclohexyl lipophilic group MW molecular weight n number of precursors n B number of bulk molecules n S number of interfacial molecules NMR nuclear magnetic resonance NOESY nuclear Overhauser effect spectroscopy viii

9 NP nanoparticle NSF nephrogenic systemic fibrosis O/W oil-in-water p bound proportion of bound population p free proportion of free population PAA poly(acrylic) acid PAA 25 poly(acrylic) acid with 25 acrylic acid monomers PAA-GdF 3 NPs poly(acrylic acid)-coated gadolinium trifluoride nanoparticles PAA-LnF 3 NPs poly(acrylic acid)-coated lanthanide trifluoride nanoparticles PAMAM polyamidoamine PDI polydispersity index PEG poly(ethylene glycol) PET positron emission tomography PFGSTE pulsed field gradient stimulated echo PGSE pulsed gradient spin echo PTFE poly(tetrafluoroethylene) r H distance between the metal center and the coordinated water molecules r i relaxivity R relaxation rate R c critical radius R i IS inner sphere contribution to relaxation R i o diamagnetic contribution to relaxation R i obs observed relaxation rate R i OS outer sphere contribution to relaxation RES reticuloendothelial system S electron spin quantum number SIR selective inversion recovery SPECT single photon emission computed tomography SPION super-paramagnetic iron oxide nanoparticle STE stimulated echo STEM scanning transmission electron microscope/microscopy q number of bound water molecules T temperature T 1 spin lattice relaxation time T 1M relaxation time of bound water molecules T 1e spin-lattice electron spin relaxation time T 2 spin-spin relaxation time T 2e spin-spin electron spin relaxation time T ie electron spin relaxation time USPION ultra-small super-paramagnetic iron oxide nanoparticle W/O water-in-oil x precursor concentration Y:Cit ratio of yttrium(iii) ions to citrate ix

10 Greek Letters mixing time in the PFGSTE experiment t time dependent zero field splitting γ of γ I proton magnetogyric ratio / also surface tension δ duration of gradient in the PFGSTE experiment µ chemical potential of a precursor µ B chemical potential of a bulk precursor µ B Bohr magneton τ i correlation time τ 1 spin-lattice correlation time / or delay in the PFGSTE experiment τ 2 spin-spin correlation time / or delay in the PFGSTE experiment τ M residence lifetime of bound water molecules or exchange correlation time τ R rotational correlation time τ V electron relaxation correlation time ω I proton Larmor frequency ω S electron Larmor frequency x

11 1 Introduction A host of new techniques have surfaced in the field of biomedical imaging over the last century. Examples of some real-time, non-invasive methods include computed X-ray tomography (CT), magnetic resonance imaging (MRI), optical imaging, positron emission tomography (PET), single photon emission computed tomography (SPECT) and ultrasound. 1 PET and SPECT are often metabolic in nature. MRI and CT are often more anatomical with mm or worse resolution. Of these imaging platforms, MRI is one of the most important tools for biomedical research and diagnostic clinical medicine. 2 It is the technique of choice for imaging of the brain and central nervous system, detecting tumours and for assessing cardiac function. Often, however, contrast agents may be needed to enhance local signal of target tissue or improve resolution. Currently (2006), approximately 35 % of MRI procedures utilize such agents. 3 MRI contrast agents can assist in the early detection of cancer. Superparamagnetic iron oxide nanoparticles (SPIONs) have been used to detect lesions as small as 2 3 mm in liver tissue, 4 while ultra-small super-paramagnetic iron oxide nanoparticles (USPIONs) have detected lymph node metatheses in the 5 10 mm range. 5 However, these negative contrast agents suffer from drawbacks (discussed later) and construction of novel gadolinium based nanoparticles could alleviate these issues. Herein, the synthesis and verification of poly(acrylic acid)-coated, gadolinium trifluoride (PAA-GdF 3 ) nanoparticles is reported, along with the physical characterization of their citrate-coated GdF 3 nanoparticle precursors. Background information on the targeted tissue (malignant tumours), clearance pathways, physical barriers and relaxivity mechanisms is presented to provide structural considerations that influence in vivo nanoparticle function. Examples of different MRI contrast agents are reviewed, followed by discussion of different nanoparticle synthesis procedures and the thermodynamics of nanoparticle formation. Finally, NMR techniques, employed in the physical characterization of the nanoparticles, are described. 1

12 1.1 Nanoparticle size - Tumour angiogenesis and clearance pathways Size is an important parameter in the design of MRI contrast agents for cancer imaging. Ideal size, for enhanced blood retention time and nanoparticle to tumour localization, is dependent upon differences between normal and tumour vasculature permeability, as well as constraints imposed by the reticuloendothelial (RES) and excretory systems. 6 Tumour vascular permeability is induced by angiogenesis, the process through which new blood vessels are formed. When tumours exceed 1-2 mm in diameter, nutrient diffusion is no longer sufficient to nourish the outer cells of a tumour. Consequently, new blood vessel growth is needed to ensure tumour growth. 6 There is a fundamental structural difference between these and normal vessels. The latter are comprised of three layers that create a water-tight seal useful for carrying nutrients. Surrounding the inner endothelial layer, whose cells are held together by tight junctions, is a tightly adherent baseline membrane. This in turn, is surrounded by pericytes and smooth muscle cells. In contrast, vessel walls in tumours are incompletely formed and fragile. Large gaps exist between the endothelial cells and the baseline membrane, and the pericytes and smooth muscles are only loosely adherent. Consequently, tumour vessels are hyperpermeable. 7 Size plays a key role in the localization of contrast agents to tumour regions. Consequently, contrast agents are divided into two categories based on size: low molecular weight contrast agents (LMCAs) have molecular weights less than 1 kda, whereas macromolecular contrast agents (MMCAs) have molecular weights typically greater than 30 kda (size comparison: 20 kda ~ 10 nm). 6 LMCAs, such as Gd-DTPA, can localize in tumour cells due to their ability to diffuse faster through hyperpermeable tumour vessels than normal vessels. However, their non-selective permeable nature (i.e. larger first pass fractions in normal vasculature) results in a reduced blood circulation time. Although MMCAs diffuse more slowly through hyperpermeable vessels, they have longer intravascular retention in normal tissue that results in a greater localized concentration at tumour sites over time. The differing size-dependent ability of contrast agents to diffuse though vasculature is shown in Figure 1.1. In general, normal microvessels are less permeable to molecules whose diameters are greater than 5-10 nm 8 2

13 because the average effective pore size in normal intact endothelium is ~ 5 nm. 9 On the other hand, tumour vessels have been known to let through molecules as large as nm. 10 On a cautionary note, hyperpermeability is not restricted to tumour vessels but is also observed in severe inflammatory regions and where reparative tissues are present. 6 Figure 1.1 A representation of the differing ability of (A) small and (B) large macromolecular contrast agents to diffuse through blood vessels (Reprinted from Eur. J. Radiol. 2006, 60, Copyright (2006), with permission from Elsevier). The renal and hepatic systems impose further limits on size. Particles with a hydrodynamic diameter (HD) less than 6 nm are removed through glomerular filtration in the kidneys and undergo renal clearance. As such, they have a short blood circulation time. However, particles with an HD greater than 8 nm are not readily excreted in this fashion. Thus, approximately doubling particle size (from ~ 5 to ~ 11 nm) results in a 11, 12 thirty fold increase in blood-retention time. Nanoparticles greater than 100 nm in size are easily removed by the reticuloendothelial system (RES; liver, spleen and bone marrow). 1 Particles in the size range between these two extremes are not efficiently 9, 11 cleared by either mechanism and hence exhibit longer circulation times. In conclusion, numerous factors need to be taken into consideration when choosing the optimal size of a nanoparticle or MMCA. The hyperpermeability of tumour vasculature, size dependent permeability of normal vasculature, and size selective filtering by the renal and hepatic systems all play a role in determining nanoparticle size. The general consensus seems to be that particles between nm are good for tumour imaging. 1, 9 3

14 1.2 Nanoparticle size Blood-brain barrier considerations The blood-brain barrier (BBB) is noted for its ability to act as a natural, physiological wall that separates the central nervous system structures (brain and spinal cord) from direct contact with circulating blood. It is characterized by tight interendothelial junctions, few pinocytotic vesicles and no fenestrations. 13 Drugs typically cross the BBB through diffusion. Such a process results in an 8-log difference in permeability, relative to the liver, for an immunoglobulin (size ~ 57 Å in radius, Figure 1.2). 13 However, the ability of the blood-brain tumour barrier (BBTB) to act as a physiological wall can be partly compromised due to the type of vasculature present. Three kinds may exist: non-fenestrated capillaries (i.e. like normal brain), fenestrated capillaries (permeable to small but not large molecules, maximum channel width of 5.5 nm 14 ) and capillaries with interendothelial gaps as large as 1 µm. 13 Figure 1.2 Permeability of various tissues by molecules of different sizes. The data points at a molecular radius of ~ 57 Å correspond to an immunoglobulin (Reprinted by permission of Duke University Press: Neuro-oncology 2000, 2, 45-59, copyright (2000)). Size and blood retention time are critical parameters when assessing whether a particle can localize to brain tumours. Small molecules have proved to have reduced ability to accumulate in brain tumours due to low blood retention times. The behaviour 4

15 BBTB. 16 Two alternative methods to cross the BBB are also available. Disruption of the of 1,3-bis (2-chloroethyl)-1-nitrosourea (BCNU, carmustine) is a case in point. Despite its small size (<1 nm) and known ability to penetrate the BBTB, BCNU cannot accumulate in therapeutic quantities in malignant gliomas. 15 This is because, due to renal filtering, it has a short blood half-time. Consequently, it was theorized that particles with sizes large enough to evade renal filters ( > 5-6 nm), yet small enough to bypass hepatic filters ( < nm) might have better success due to longer blood-circulation times. Another particle size constraint is imposed by the pore size in malignant gliomas. In 2008, Sarin et al., using Gd-DTPA functionalized dendrimers, estimated the maximum contrast agent size that can penetrate fenestrations and gaps in the BBTB of RG-2 malignant gliomas to be nm in diameter. They concluded that nanoparticles between ~ nm may be able to be delivered across a minimally comprised tight junctions can allow particles to traverse the BBB. Osmotic opening can be induced by intracarotid administration of arabinose or mannitol solutions that result in the interendothelial tight junctions widening by about 40 nm in diameter. 17 Disruption can also be achieved by bradykinin analogues such as RMP Alternatively, receptormediated transcytosis can also allow particles to cross the BBB. Nanoparticles coated with polysorbate 80 adsorb apolipoproteins which can bind to brain capillary endothelial receptors, 19 while those conjugated to transferrin can bind to their corresponding transferrin brain endothelial receptors, 20 both of which allow receptor-mediated transcytosis to occur. 18 Figure 1.3 shows different pathways molecules use to cross the blood brain barrier, including receptor mediated transcytosis. 5

16 Figure 1.3 Five different pathways across the blood-brain barrier, (d) depicts receptor mediated transcytosis (Reprinted by permission from Macmillan Publishers Ltd: Nature Reviews Neuroscience, 2006, 7, 41-53, copyright (2006)). 1.3 Relaxivity The ability of a MRI contrast agent to induce contrast is captured by a property called relaxivity (r i ). It is defined by the equation, Ri ri = Equation 1.1 [ CA] where R is the relaxation rate, CA the contrast agent and i an index referring to either the spin-lattice (a value of 1) or spin-spin (a value of 2) relaxation mechanisms. The observed relaxation rate, R obs i, is a sum of three contributions: one diamagnetic and two paramagnetic ones (Equation 1.2). The diamagnetic contribution (R o i ) is due to all parameters that influence relaxation except paramagnetic ones. The first paramagnetic component, an inner sphere contribution (R IS i ), results from coordination and exchange of water molecules with the metal complex (Figure 1.4). The second, an outer sphere contribution (R OS i ), is due to water molecules influenced by the metal, but not in direct contact with it (usually nearby diffusing water molecules). Both the paramagnetic 6

17 contributions are caused by interactions of the electron dipole moment of the metal with the nuclear dipole moment of the proton (dipolar relaxation mechanism). 3 obs i o i IS i OS i R = R + R + R Equation 1.2 Occasionally in literature a second coordination sphere is mentioned, where relaxation is caused by water molecules which are usually hydrogen bonded to polar groups on the complex or nanoparticle. 21 Also, one can well imagine a separate contribution due to exchangeable protons present on the ligand or chelate. Figure 1.4. A representation of first and second coordination spheres for a gadolinium complex. Additional parameters, such as r H, a, τ R and τ M are indicated (European Journal of Inorganic Chemistry, 2000, Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission). Inner sphere paramagnetic contributions to relaxivity have been described by Solomon-Bloembergen-Morgan theory. A set of equations for such contributions to spinlattice relaxation is presented below. The R 1 IS is directly proportional to the number of bound water molecules (q) and the concentration of the paramagnetic contrast agent (CA), while inversely proportional to the mean residence lifetime of bound water molecules (τ m ) 3, 21 and their relaxation time (T 1m ). IS 1 [ CA] q ( T +τ ) R = Equation m m The latter (T 1m ) is dependent on the spectral density function (Equation 1.4). It is inversely proportional to the sixth power of the distance between the metal center and the 3, 21 coordinated water molecules (r H ) and dependent upon the correlation time (τ i ). ( S + 1) 3τ 7τ ( ) ( ) ω I τ 1 + ωs τ = 2 γ I g µ B S 6 2 1m 15 rh 1 T Equation 1.4 7

18 Other parameters in the equation include γ I, the proton magnetogyric ratio, g, the Lande factor for the free electron, µ B, the Bohr magneton, S, the electron spin quantum number, ω I and ω S, the respective proton and electron Larmor frequencies and τ 1 and τ 2 the spinlattice and spin-spin correlation times respectively. The last two parameters depend on three different correlation times (Equation 1.5) = + + Equation 1.5 τ i Tie τ m τ R T ie is the electron spin relaxation time, τ R the rotational correlation time and τ m the exchange correlation time (Figure 1.4). Proton spin-lattice relaxation is dependent on both T 1e and T 2e through the dispersive parts of Equation 1.4. These two parameters are also field-dependent (Equation 1.6 and 1.7) t τ v 4τ v T1 e = [ 4S( S + 1) 3] Equation τ v ωs 1+ 4τ v ωs 2 1 t 5τ v 2τ v T2e = [ 4S( S + 1) 3] 3τ v Equation τ v ωs 1+ 4τ v ωs t is the time-dependent zero field splitting and τ v is the electron relaxation correlation time. The outer sphere relaxation is inversely proportional to the distance of closest approach, a (Figure 1.4), and the diffusion coefficient, D. OS OS 1 R1 = C [ 7J ( ωs ) + 3J ( ω I )] Equation 1.8 ad C OS is a constant. For small Gd-chelates this contribution to relaxivity amounts to forty percent, 21 while ten percent can be attributed for macromolecular contrast agents. 23 It has long been recognized that larger contrast agents would have greater paramagnetic contributions to relaxivity due to slower tumbling times. At field strengths typically used for MRI (20 63 MHz) the rotational correlation time is usually so fast for small Gd-chelates that it is the dominant contributor to relaxivity. For a small molecule ( Da) a typical value of T ie is approximately 1 ns at 0.5 T (~ 20 MHz), while τ R is about ps. 3 Botta has mentioned studies where slowing down the rotation has resulted in improvements to relaxivity. 21 8

19 1.4 MRI contrast agents Certain deficiencies of MRI, e.g. the lack of contrast between tissues and lesions, have cemented the importance of MRI contrast agents. Current commercial agents include super and ultra-small paramagnetic iron oxide nanoparticles, used for T 2 - weighted imaging, and Gd-chelates, used for T 1 -weighted imaging. In addition, a variety of MRI contrast agents have appeared in literature and new ones are still being published. The need to generate these new contrast agents is driven by the shortcomings of the current ones. Iron oxide, T 2 contrast agents suffer because of their negative contrast effect and high magnetic susceptibility, causing image aberrations. Disadvantages of the Gd-chelates include relatively low relaxivities, a lack of targeting specificity and small size, the latter of which results in non-specific and rapid equilibrium between intravascular and interstitial compartments. 1 In addition, Gd-chelates have been found to increase the risk of developing nephrogenic systemic fibrosis (NSF), a serious medical disorder, in patients with acute or chronic severe renal insufficiency and patients with acute renal insufficiency. 24 The dearth of commercially available large positive contrast agents (MW > 30000) has, in part, led to a focus on the synthesis of large gadolinium based contrast agents. Designed to carry high gadolinium(iii) payloads, dendrimers, micelles and liposomes, virus vectors, proteins, polysaccharides, polyamino acids, zeolites and metallofullerenes have all been reported over the last 20 years. 6 nanoparticles have also been synthesized. 1 In addition, several gadolinium containing A brief description of these contrast agents, as well as the commercially available ones, is provided below, with a moderate focus on their relationship to cancer imaging Iron Oxide Nanoparticles Iron oxide nanoparticles, which dramatically shorten the T 2 relaxation time, are negative contrast agents. Because of the relationship between size, and biodistribution and blood half-life, they are divided into three classes according to diameter. These are micrometer-sized (MPION, a few micrometers), super (SPION, nm) and ultra- 1, 25 small (USPION, <50 nm) super-paramagnetic iron oxide nanoparticles. Commercial iron oxide MRI contrast agents include Feridex, Resovist and Combidex, all of which are 9

20 either SPIONs or USPIONs. These particles have an iron oxide core (Fe 3 O 4, Fe 2 O 3 ) with either a dextran or carboxydextran coating and induce high T 2 relaxivities ranging from mm -1 s -1 (1.5 T). 1 SPIONs have been used to diagnose liver diseases because they are selectively phagocytosed by Kupffer cells in the liver, spleen and bone marrow (i.e. RES). As diseased liver tissue, such as a liver tumour, has a deficiency of Kupffer cells, few SPIONs are taken up. The localized SPIONs result in a strong contrast between normal and diseased tissue, enabling detection of the disease. 1 In contrast, the smaller USPIONs have a longer blood circulation time and are used for lymph node imaging. 26 Two disadvantages have been noted for iron oxide nanoparticles. Firstly, contrast is achieved by inducing a decrease in signal in T 2 -weighted imaging. Secondly, their high magnetic susceptibility alters the magnetic field in neighbouring normal tissue which can make images unclear and reduce the background around lesions. 1, 25 applications use Gd-chelate contrast agents. Consequently, most Gadolinium chelates (Gd-chelates) As of 2006, gadolinium chelates (Gd-chelates) comprised all of the T 1 contrast agents used for MR imaging. These consisted of four clinically approved agents: the two anionic contrast agents, Gd(DTPA) 2- (Magnevist, logk Gd = 22.5) and Gd(DOTA) - (Dotarem, logk Gd = 24.7), and the two neutral versions, Gd(DTPA-BPA) (Omniscan, logk Gd = 16.85) and Gd(HPDO3A) (Prohance, logk Gd = 23.8) (Figure 1.5. A). 27 relaxivity enhancement induced by these agents is due to inner sphere and outer sphere contributions (refer to Relaxivity section). The small size of these agents makes them non-specific and they distribute principally in the intravascular and interstitial space. 28 They have been known to collect in the kidneys due to glomerular filtration and are 29, 30 excreted in this manner Albumin-(gadolinium-DTPA) complexes and MS-325 (Gadofosveset) Albumin-(gadolinium-DTPA) complexes were originally prepared by Ovid et al. in the late eighties. 31 The They typically contained covalently attached Gd-DTPA chelates, giving the macromolecule a molecular weight of roughly 92 kda (~ 6 nm). 10

21 Relaxivities of 14.9 mm -1 s -1, relative to gadolinium concentration, were reported (0.25T 6, 31 ~ 10 Hz, 37ºC). Numerous studies were conducted to demonstrate the MRI capabilities of albumin-(gd-dtpa) molecules. 6 Among them, enhanced signals were observed in liver, lung, spleen, kidney and brain tissue due to a T 1 -effect. 32 Furthermore, albumin-(gd- DTPA) assisted in characterization of microvessels in breast, sarcoma and prostate tumors However, a few disadvantages led to this contrast agent never reaching human trials. Prolonged retention of Gd (several weeks) was noted with incomplete and slow elimination of it. This is due to only 5% of albumin seeping out from the blood every hour, consequently leading to intravascular retention. 7, 36 observed to accumulate in liver and bone. Lastly, limiting its ability to act as an in vivo contrast agent. 37 Furthermore, it was albumin is possibly immunogenic, An MS-325 albumin conjugate was developed to overcome some of the limitations associated with the albumin-(gd-dtpa) macromolecule. MS-325, a small molecule of molecular weight (MW) 957 Da, consists of a diphenylcyclohexyl lipophilic group attached to a gadolinium chelate though a phosphodiester linkage (Figure 1.5. B). The lipophilic group binds strongly to albumin in a reversible, non-covalent manner. The molecule is injected in its free form and binds to albumin in vivo. Up to 30 MS-325 molecules can attach to a single albumin, leading to the formation of a macromolecule of about 68 kda. 6 The more rapid clearance time of MS-325, especially compared to albumin-(gd- DTPA), led to it becoming the first gadolinium based macromolecular agent to undergo human trials. The elimination half-lives were observed to be 2-3 hrs in primates and rabbits, and 25 minutes in rats. 38 Subsequent Phase (I, II and III) trials conducted in humans demonstrated good vasculature and arterial enhancement (obtained through MRI angiography) and no adverse effects European Union (EU). 30 Consequently, it was approved for use in the However, it demonstrated limited potential as a tumour imaging agent because no signification correlation was detected between it and either microvascular density (MVD) or tumour grade

22 1.4.4 Viral MRI contrast agents Recently, the cowpea chlorotic mottle virus (CCMV) protein cage (capsid) has been used as a frame for a viral MMCA (Figure 1.5. C). 43 The approximately 180 metal binding sites on the cage surface, usually for Ca 2+, were used to bind Gd 3+ ions instead. The T 1 and T 2 relaxivities of water protons were 202 and 376 mm -1 s -1 respectively (~1.5 T, 62 MHz, 23 ºC), which the authors claim were the highest known at that time (2005). However, the in vivo stability of the Gd-CCMV interaction needs to be better characterized and improved as the Gd-CCMV interaction displays a high dissociation constant compared to chelated gadolinium ions. In addition, high local concentrations of Ca 2+ could potentially compete with the Gd 3+ ions. As a comparison, Tb 3+ displayed a 100-fold greater affinity for the protein binding sites than Ca , 43 immunogenicity of this contrast agent needs to be collected. Also, more data on the Dextran-(Gd-DTPA) Dextran-(Gd-DTPA) is composed of a linear glucose polymer, dextran, to which is conjugated many Gd-DTPA complexes through hydrolysable bonds. 6 Dextrans of various different sizes can be prepared. One dextran contrast agent contained 15 Gd- DTPA complexes and possessed a weight of 75 kda. 45 It was found to remain intravascular for 1 hr, degraded more rapidly than albumin and had a short biological half-life (43 min). A larger dextran-(gd-dtpa) contrast agent was produced with approximately 187 Gd-complexes per dextran, a molecular weight of 165 kda and a diameter of 17.6 nm (Figure 1.5. D). 46 Its comparatively larger size resulted in it remaining intravascular for 58 hrs. When injected in rabbits with thigh tumours, it demonstrated less tumour contrast than a LMCA after 1 hr, more after 24 hrs and was still visible after 72 hrs. Less contrast was obtained after a short time (1 hr) due to the greater initial ability of LMCAs to localize at the tumour due to higher vascular permeability. Of particular note, the distribution and elimination of dextrans is found to be size and charge dependent. 47 Dextran use in contrast agents is advantageous because of its inexpensive price and recognized safety record (used for 50 years as a synthetic plasma expander). However, its intrinsic polydispersity makes permeability difficult to predict 6 and an increased incidence of anaphylactic reactions have been reported for larger dextrans

23 Figure 1.5. (A) Structures of (i) Gd(DOTA) -, (ii) Gd(HPDO3A), (iii) Gd(DTPA) 2- and (iv) Gd(DTPA-BPA). (B) The chemical structure of MS-325. (C) Reconstruction of the (i) cowpea chlorotic mottle virus (CCMV) protein cage 48 and (ii) CCMV Ca 2+ /Gd 3+ binding protein. 49 (D) The structure of dextran-(gd-dtpa), where ND is the number of monomers bound to GD-DTPA. [(A) Chem. Soc. Rev. 2006, 35, Reproduced by permission of The Royal Society of Chemistry. (B) Reprinted from Eur. J. Radiol. 2006, 60, Copyright (2006), with permission from Elsevier. (C) Magnetic Resonance in Medicine 2005, 54, Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission. (D) Reprinted from Acad. Radiol., 2004, 11, , Copyright (2004), with permission from Elsevier.] Liposomes Liposomes are spherical vessels that range in size from nm in diameter. They are composed of one of more bilayer phospholipid membranes (lamella) and contain a hydrophilic interior. If paramagnetic material is inserted into the membrane or aqueous interior, they can act as MRI contrast agents (Figure 1.6. A). 6 furthermore serve a dual role as drug targeting delivery vehicles. Liposomes can 13

24 Two liposome-(gd-dtpa) vesicles, 70 and 400 nm in size, were tested on rats with hepatic metathesis. The experiments demonstrated significant contrast enhancement between liver and tumour. 50 Further studies, which attached poly(ethylene glycol) (PEG) to the liposome, found that this molecule reduced uptake by the RES. 51 the liposome-peg vesicles were christened stealth liposomes. 52 Consequently, More recent work has focused on liposomes targeted with agents such as antibodies, 53 peptides, 54 folates, aptamers and polysaccharides. 55 The principal disadvantage of liposomes is their polydispersity. As this makes the synthesis difficult to reproduce, the future of targeted liposomes has been stated to be uncertain Dendrimers Dendrimers comprise a class of repeatedly branched, synthetically produced, spherical polymers that can be consistently and reproducibly created. Two examples are polyamidoamine (PAMAM) and diaminobutane core polypropylimine (DAB or PPI), the former of which is shown in Figure 1.6. (B). Both can be functionalized with a large number of Gd-chelates giving them potential to act as MRI contrast agents. The size and molecular weight of dendrimers increase with each subsequent generation, giving them different pharmacokinetic and pharmocodynamic abilities. Consequently, they can be used for different imaging applications. 6 For example, different generations of PAMAM, ranging from two to ten (denoted G2 to G10), were postulated to have potential to image renal, hepatic, tumour and vascular regions. Molecules of sizes 3-6 nm (G2-G4) are excreted via the kidney and can be used for renal imaging. 56 Sizes of 5-8 nm (G4, G5) can be used for tumour screening, as they can selectively permeate tumour vasculature. 6 Finally, sizes of greater than 8 nm (G6 and above) demonstrated good vascular enhancement, 57 while sizes of ~ 15 nm (G10) have potential for hepatic imaging Metallofullerenes Fullerenes are closed-caged molecules of carbon in the sp 2 hybridized state. They are the third allotrope of carbon, after diamond and graphite. 58 Soon after their discovery, it was found that metals could be trapped in fullerenes giving rise to the then novel metallofullerene complexes. 59 Water soluble, polyhydroxyl, gadolinium endoheral, metallofullerenes [(Gd@C 82 (OH) n, Gd-fullerenols] were synthesized and displayed r 1 and 14

25 r 2 relaxivities of 81 and 108 mm -1 s -1 respectively. 60 In vivo studies indicated signal enhancement in lung, liver, spleen and kidney. Entrapment by the RES, was postulated to be caused by either particle aggregation or association with plasma components such 61, 62 as albumin. One advantage of the Gd-fullerenols is that the carbon cage protects the gadolinium ion from leaching into the surrounding area, while another is that it protects the metal ion from chemical attack. 30 However, previous toxicity studies with Ho- C 82 (OH) n fullerenes indicated that as much as 10% accumulate in the bone tissue. 63 If the same holds true for the Gd-fullerenols, surface modification will be required to demonstrate any viability for commercial use. More recently, a gadofulleride, with carboxylate and hydroxyl functional groups, was prepared and conjugated to antibodies. 64 It was found to form aggregates of ~30 nm in diameter, providing insight as to why these particles localize in the RES. Aside from the Gd@C 82, two other basic gadolinium metallofullerene skeletons have been produced: Gd@C 60 and Sc x Gd 80. Usually the latter produces the highest relaxivities because three gadoliniums are enclosed within its cage. Recently, a surface modified Gd 3 N@C 80 with hydroxyl and carboxylate functionalities (Figure 1.6. C) was synthesized that displayed r 1 and r 2 relaxivities of 207 and 282 mm -1 s -1 respectively, fifty times larger than those of Omniscan and Magnequist. 30 (2.4T) Zeolites Zeolites are aminosilicates that contain a well-defined pore structure and channel system of molecular length. They are also chemically and thermally resistant. One type of zeolite, zeolite Y, is composed of eight sodalite cages connected by oxygen bridges (Figure 1.6. D). At the center is a cavity, called a supercage, that has an internal diameter of 11.8 Å. The diameter accessible to small molecules is 7.4 Å. Gd(III) ions can be nestled within the supercage and are held firm by the strong electrostatic interaction between them and the negative aminosilicate. 65 Negligible leaching of gadolinium was observed from the supercage. 66 The size of the zeolite was found to lie between nm in diameter. Relaxivities (r 1 ) ranging from 11.4 to 37.7 mm -1 s -1 have been reported (~1.5T, 37ºC)

26 Figure 1.6. (A) A stealth liposome capable of acting as an MRI contrast agent. (B) A representation of a PAMAM dendrimer, where n is the generation number. Gd-chelates can be attached to the amino ends. (C) An image of a Gd 3 N@C 80 metallofullerene. (D) A zeolite Y structure composed of eight sodalite cages (truncated octahedrons) connected by oxygen bridges. Within the supercage is nestled a Gd(III) atom. Water is shown to permeate through an accessible hole. [(A) and (B) Reprinted from Eur. J. Radiol. 2006, 60, Copyright (2006), with permission from Elsevier. (C) Reprinted in part with permission from Bioconjug. Chem. 2009, 20, Copyright 2009 American Chemical Society. (D) Chemistry-A European Journal 2005, 11, Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission] Gadolinium-based Nanoparticles Some recent effort has been directed towards the synthesis of gadolinium nanoparticles in hopes of producing better T 1 contrast agents. Gadolinium oxide (Gd 2 O 3 ), fluoride (GdF 3 ) 71 and phosphate (GdPO 4 ) 72 nanoparticles have all been synthesized. The Gd 2 O 3 nanoparticles had dextran, PEG and polysiloxane PEG coatings with sizes ranging from 3 26 nm in hydrodynamic diameter (HD). Two types of GdF 3 nanoparticles were produced. One was coated with citrate and had an average size of 129 nm, while the other was coated with 2-aminoethyl phosphate (AEP), doped with LaF 3 16

27 (20%), and had an average size of 51 nm in HD. The GdPO 4 nanoparticles, synthesized by a hydrothermal process, were coated with dextran and had an average HD of 23 nm. Nanoparticle relaxivity has been reported in numerous ways: relaxivity based on the number of particles, that based on surface area, and that based on the concentration of whole atoms have all been reported. Using the latter measure, relaxivities comparable to (GdF 3 core, r 1 = mm s, 14.2 T) 71 and a little over three times (GdPO 4 core, r 1 = 1 1 mm s, 7 T) 72 that of gadolinium chelates (Gd-DTPA, r 1 = 4.1 have been reported. 1 1 mm s, 7 T) 68 Poly(acrylic acid)-coated, gadolinium trifluoride nanoparticles (PAA-GdF 3 NPs) were herein synthesized based on the above GdF 3 nanoparticle protocol. Firstly, citratecoated, GdF 3 nanoparticles were synthesized and then the citrate ligand was exchanged for a PAA ligand. Citrate-coated yttrium trifluoride nanoparticles were produced in a similar fashion. 1.5 Water Soluble Lanthanide Nanoparticle Synthesis Several different nanoparticle syntheses have been reported that have produced water-soluble lanthanide nanoparticles. These include coprecipitation, 71 synthesis 73 and a microemulsion method. 74 a polyol Coprecipitation In coprecipitation syntheses, nucleation, growth, Ostwald ripening and/or agglomeration can occur simultaneously. Metal coprecipitation has commonly been achieved through reduction by use of agents such as borohydride and hydrazine (strong agents). As the free energy change must be negative, precipitation of any metal with a standard reduction potential (Eº) more positive than or V (the Eº for borohydride and hydrazine respectively) should be possible. Many first-, second- and third- row transition metals, as well as post-transition and some non-metals are ideal candidates for a reduction synthesis. 75 One such example involves the production of gold nanoparticles. In a well-known synthesis, nanoparticles were formed from auric acid and sodium citrate, the latter of which served the dual role of capping and reducing agent. 76 However, as the gadolinium (Gd 3+ ) ion in acidic and 17

28 media. 75 Nanoparticles produced though coprecipitation can also be formed though basic solution has a Eº of and V respectively, coprecipitation of this lanthanide through reduction is not possible using these two strong reducing agents. Furthermore, the reduction of metal ions with a potential more negative than V is either very difficult or not possible due to the instability of the cations in aqueous insoluble metal precursors, such as metal oxides. During such a synthesis, a capping ligand is often needed to prevent agglomeration. 75 Production of lanthanide nanoparticles through precipitation of insoluble lanthanide oxides, fluorides and phosphates are possible. Recently, water soluble, citrate-coated, gadolinium trifluoride nanoparticles were synthesized. The protocol specified the addition of gadolinium nitrate (Gd(NO 3 ) 3 to a neutralized solution of citrate and sodium fluoride, yielding roughly spherical, polydisperse nanoparticles with an average size of 129 nm. 71 The chief disadvantage of coprecipitation is that nucleation and growth occur at once, yielding nanoparticles of different sizes Polyol-Mediated Nanoparticle Synthesis In the polyol method, nanoparticle precursors are heated to a high temperature ( ºC) in a high boiling point polyol solvent (e.g. glycerol, diethylene glycol or glycerine). 77 The polyol serves the role of both solvent and capping agent. 73 This synthesis has been noted for its versatility and efficiency in the preparation of nanoscale materials. A host of different metals such as gold, platinum, and copper have yielded micron to submicron particles. 78 In addition, metal oxide, 79 phosphate, 80 sulphide 81 and halenogenide 73 forms have also been produced. Monodisperse and non-agglomerated nanoscale materials with sizes ranging from the mesoscale (>100 nm) to the nanoscale (~5 nm) have been achieved. 77 Several advantages of the polyol synthesis have been presented. Firstly, the polyol has sufficient polarity to solubilize inorganic salts. Secondly, more crystalline nanoparticles can be produced as nucleation and growth of the nanoscale materials can occur at high temperatures, as the polyols have high boiling points. Next, particle agglomeration is prevented and growth of the nanomaterial limited because of the 18

29 relative affinity of the polyol for the metal precursors. Finally, the synthesis is easy and suitable for production of large quantities of meso/nano-material. 77 Recently, lanthanum doped nanoparticles have been produced by polyol-mediated syntheses. The precursors were lanthanide chlorides and ammonium fluoride, whereas the reaction was conducted in one of three polyols: glycol, diethylene glycol and glycerol. Ammonium fluoride was used as the fluoride source as opposed to sodium fluoride because the latter is not as soluble in the polyol. 73 The lanthanide nanoparticles produced were spherical and had an average size of 5-7 nm regardless of the type of polyol employed. The nanoparticles exhibited good dispersity in both water and ethanol Microemulsion methods Micro-emulsions can result when combinations of oil, water, surfactant and cosurfactant are mixed. In these systems, the solution is optically isotropic however the molecules are not randomly orientated as they should be in a solution. Instead, two general types of drops can form. In oil-in-water (O/W) drops, the oil in located within the boundary of the surfactants coating the drop, while the reverse is true for water-in-oil (W/O, i.e. reverse micelle) drops. The latter is depicted in Figure 1.7. (A). In an equal volume by volume mixture of oil and water, O/W drops will form if the surfactant is more soluble in water that in oil. W/O drops will form if the reverse is true. 75 Microemulsions have been commonly used as micro- and nano-reactors in nanoparticle syntheses. 75 Reverse micelles in particular have been used to mix two components which react to form nanoparticles. Due to their small size, reverse micelles are subject to Brownian motion. Inevitably, the collisions between two reverse micelles with different internal compositions (Figure 1.7. (B) i and ii) sometimes lead to the formation of a short-lived dimer (Figure 1.7. (B) iii) whose lifetime is ~100 ns. The contents of the two reverse micelles mix before decoalescing. 82 results in equilibrium being achieved across the entire mixture. 75 Over time this process Advantages to using microemulsions to synthesize nanoparticles include control of nanoparticle size (though control of the size of the reverse micelle), and the inherent ability to produce more monodisperse nanoparticles. Reverse micelle size can be controlled by varying the amount of water to surfactant ratio. 75 However, microemulsions can also be relatively complex. Reaction rates and equilibria have been 19

30 found to be different from those in bulk solution and solvated ions can effect phase equilibria as well as reverse micelle stability, size and shape. 75 Figure 1.7. A) Model of a reverse micelle. B) The coalescence (iii) of two drops (i and ii) containing different chemicals, the result of which can produce nanoparticles. [(A) Reprinted by permission from Macmillan Publishers Ltd: Nature, 1943, 152, , copyright (1943). (B) Reprinted in part with permission from Chem. Rev. 2004, 104, Copyright 2004 American Chemical Society] Microemulsions have been used to synthesize nickel nanoparticles through metal reduction with hydrazine. 83 Similarly, core shell metal nanoparticles have been produced 84, 85 with reactive metal cores (e.g. Fe) being coated with inert metals (e.g. Au, Si). Of greater interest, YF 3 nanoparticles with relatively monodisperse sizes have recently been synthesized. Both amorphous and crystalline nanoparticles were produced. The diameter of amorphous spheres was found to be controllable between 6 and 50 nm. Slight variation in reaction conditions led to the production of monocrystalline, monodisperse, hexagonal and triangular crystals with diameters tuneable between 25 and 350 nm. Greater crystallinity was suggested to be achieved by a slower growth process caused by direct addition of ammonium hydrogen difluoride (NH 4 HF 2 ) to YCl 3 containing microemulsions in the latter synthesis

31 1.5.4 Hydrothermal synthesis High temperature, high pressure conditions may be achieved by conducting the reaction in a solvent well above its boiling point in a sealed vessel. Such reaction conditions are referred to as solvothermal processing, or in the case of water, hydrothermal processing. If the temperature is increased sufficiently, a supercritical liquid can be created displaying unique properties such as high viscosity, no surfacetension, and a high ability to dissolve compounds that ordinarily display low solubilities at ambient temperature. However, a synthesis need not be performed under supercritical conditions to be considered solvothermal or hydrothermal. 75 Advantages include high crystallinity of the nanoparticles, and increased solubility and reactivity of the metal salts and complexes at elevated temperatures. 75 Although no known water soluble lanthanide nanoparticles have been produced via this method, syntheses have been conducted 86 with similar conditions to those that produced water soluble, citrate coated, lanthanide trifluoride nanaparticles. 71 Consequently, particle insolubility in the hydrothermal synthesis might have been due to their exceedingly large size (on the supra-micrometer scale 86 ). 1.6 Thermodynamics of nanoparticle formation A coprecipitation synthesis method is characterized by nucleation, growth, coarsening and/or agglomeration occurring simultaneously. Often, the incidence or time lengths of some of these processes are undesirable because they result in unwanted nanoparticle attributes, such as size polydispersity. Hence, an understanding of nucleation and growth processes is crucial because it allows one to predict and control various aspects of the synthesis. Equations describing nanoparticle nucleation from a thermodynamic perspective are presented herein, along with the La Mer nucleation and growth model of colloids Thermodynamics The preparation of hydrosols in solution involves the presence of two phases: the free precursor and the bulk aggregate. These two phases have three main chemical potentials associated with them: that of the free precursor, the bulk hydrosol and the 21

32 surface of the hydrosol. Hydrosol formation will be spontaneous if there is a net reduction in the chemical potential of the system and vice versa. A more elaborate description of nucleation can be described with a series of equations. 88 The change in free energy that occurs as a result of aggregation is equal to the difference between the free energy of the aggregate, g(n), and that of the precursors in solution, nµ, G = g ( n) nµ Equation 1.9 where n is the number of molecules (or precursors) present and µ the chemical potential of each molecule. As µ depends on precursor concentration, x, and the temperature, T, Equation 1.9 can be modified to give, Θ G = g ( n) n( µ + kt ln x) Equation 1.10 where k is Boltzmann s constant. The aggregate term, g(n), can be further broken down into bulk and surface energy terms, g + where n B and n S are the number of bulk and interfacial molecules respectively, and g B B B S S ( n) = n g n g Equation 1.11 and g S the free energy associated with each bulk and interfacial molecule respectively. The bulk free energy term can be related to its chemical potential, µ B, through, B B B n g = nµ Equation 1.12 while the surface or interfacial energy term can be given as a function of its surface tension, γ, and a geometric factor, b gf. n S g S 2 / 3 = γ bgf n Equation 1.13 Equation 1.12 and 1.13 can be inserted into the general equation (Equation 1.10) to give, B Θ 2 / 3 ( µ + kt ln x) + γ b n G = n µ Equation 1.14 Since for a saturated concentration, the bulk term equals, B sat µ = µ Θ + kt ln x Equation 1.15 the free energy (Equation 1.14) can be rearranged to give, x 2 / 3 G = nkt ln + bgf n sat x γ Equation 1.16 For a spherical particle this equation can be revised to give an alternative form that includes a radial term, R. x G = nkt ln + 4πR 2 γ Equation 1.17 sat x gf 22

33 Since the bulk free energy term rises faster with radius than the interfacial free energy term, there exists a critical radius (R c ) above which nucleation is spontaneous because G decreases (Figure 1.8). Figure 1.8. A depiction of the spontaneity of nuclei formation. Initially, there is an increase in G when the concentration, x, is below the saturation concentration, x sat. Nucleation is not spontaneous during this period. Eventually a critical concentration is reached, n c (corresponding to R c ), above which stable nuclei are formed. (Polyelectrolytes and Nanoparticles, 2007, pg 48, Chapter 3, Koetz, J., Kosmella, S., Fig. 3.1, reprinted with kind permission of Springer Science+Business Media). If the formed nanoparticles or colloids are polydisperse in size, an additional process, called Ostwald ripening, may occur that affects particle size. Since small particles have a greater concentration of surrounding solute molecules than larger ones, diffusion of solute molecules from small to larger particles will take place. This results in large particles growing and small particles shrinking La Mer model of nucleation and growth About sixty years ago, a model was proposed to describe nucleation and growth process that occur during colloidal formation (Figure 1.9). 87 Initially, concentration of the solute is increased over time until nucleation occurs at the critical limiting supersaturation. This is followed by a nucleation period during which new nuclei are formed and growth of existing nuclei take place. The consumption of solute, which is incorporated into new and growing nuclei, results in a decrease in concentration to a 23

34 point where no new nuclei are formed but where growth of existing nuclei can still occur. This concentration is called the nucleation concentration. Finally, growth of existing nuclei ceases to occur once the concentration of solute drops to its solubility level. A limitation of the model is that it does not take into account capping agents that are often used to stabilize and arrest the growth of the nanoparticles. Figure 1.9. A colloidal nucleation and growth model as proposed by La Mer and Dinegar. (Polyelectrolytes and Nanoparticles, 2007, pg 49, Chapter 3, Koetz, J., Kosmella, S., Fig. 3.2, reprinted with kind permission of Springer Science+Business Media). 1.7 NMR experiments The ligand serves many purposes in gadolinium trifluoride nanoparticles. It acts as a capping agent, provides steric and electrostatic stabilization, affects relaxivity directly though outer sphere and indirectly though inner sphere effects, and can assist in evasion of the cells filtering systems (e.g. polyethylene glycol, PEG). Hence determination of the exchange rate and stability constant is important towards nanoparticle function and toxicity. A number of different NMR experiments were used to probe the interaction of the ligands with the YF 3 nanoparticles. One dimensional proton (1D1H) and correlation spectroscopy (COSY) experiments were used to characterize the coupled citrate peaks. 24

35 Diffusion coefficients, which provided ligand exchange information, were extracted via pulsed field gradient stimulated echo (PFGSTE) sequences. Finally, the exchange network was mapped by using exchange spectroscopy (EXSY) and rate constants determined using selective inversion recovery (SIR) experiments. Sequences and explanations for PFGSTE, COSY, EXSY and SIR experiments are provided below Pulsed Field Gradient Stimulated Echo (PFGSTE) experiment Diffusion coefficients can be obtained using pulsed gradient spin echo (PGSE) or pulsed field gradient stimulated echo (PFGSTE) NMR experiments. The latter utilizes the pulse sequence shown in Figure 1.10, where τ 1 and τ 2 are delays between the pulses, δ and g the duration and magnitude of the gradient respectively, and the diffusion time ( = τ 2 + τ 1 ). Figure The pulse sequence of the PGSTE experiment, originally formulated by Tanner et al (Reprinted from Biochimica Et Biophysica Acta-Biomembranes, 2007, 1768, Copyright (2007), with permission from Elsevier). The PFGSTE can be understood by looking at the stimulated echo (STE) and gradient parts separately. In the STE, the chemical shifts are refocused but not the J- couplings. The gradients add an important feature to this sequence. The first one causes a dephasing of the magnetization proportional to the amplitude and duration of the gradient, the position of the nuclei along the z-axis, and the magnetogyric ratio, γ. In the absence of diffusion, the second gradient refocuses the magnetization resulting in no attenuation of signal due to diffusion processes. However, diffusion of molecules causes incomplete rephasing and consequently loss of signal. This relationship is captured by the following expression, 25

36 echoes. 90 The presence of zero quantum coherences, which severely distort the citrate ( γ g δ [ / 3] ) 2τ 2 τ 1 I = I exp exp o exp D δ Equation 1.18 T2 T1 where I is the observed intensity and I o the initial intensity. The advantage of this sequence over the spin echo (SE) is that the spins are aligned along the z-axis during interval τ 1 and are consequently dependent on T 1 as opposed to T 2 relaxation. For systems where the T 1 > T 2, this proves beneficial because signal reduction due to relaxation is reduced leaving more magnetization to be manipulated by diffusion parameters. 89 Of note, up to five echoes can result for a system in thermal equilibrium which experiences three radiofrequency pulses. The stimulated echo can be selected by setting the first and third delays between the pulses (of duration τ 2 ) equal to each other. A sixteen phase pulse sequence has been advanced to eliminate the other unwanted spectrum in a PFGSTE experiment, necessitated the incorporation of a zero quantum filter between the second and third 90º pulses (Figure 1.11). The filter consisted of a swept-frequency, adiabatic, 180º CHIRP pulse of duration τ f applied at the same time as a gradient, G f. The latter dephased the magnetization along the z-axis, resulting in it experiencing the swept-frequency pulse at different times, with subsequent echo formation occurring at 2ατ f (where α ranged from 0 to 1 along the length of the NMR tube). Thus, if an echo occurred at one position along the z-axis, the other positions would have dephased magnetizations. If the range of frequencies caused by the gradient is wide enough, the unwanted zero quantum magnetization cancels out. 91 Figure The CHIRP based z-filter. The two black bars represent the second and third 90º pulses in the PFGSTE sequence. G HS, a homospoil gradient, is incorporated along with the zero quantum filter to remove zero and non-zero quantum coherences that lead to antiphase peaks (Angewandte Chemie-International Edition 2003, 42, Copyright Wiley-VCH Verlag GmbH & Co. KGaA. Reproduced with permission). 26

37 1.7.2 Correlation Spectroscopy (COSY) The COSY experiment (Figure 1.12) can be used to elucidate J-coupling connections. It is a homonuclear, two-dimensional experiment typically used when observing 1 H nuclei. 92 Cross peaks in the spectrum indicate coupling, while diagonal peaks provide no significantly different information than that which can be obtained through a one-dimensional spectrum. Figure The pulse sequence used in a COSY experiment where both pulses are 90º x in phase. The white triangle represents the acquisition period. The essence of this experiment is explained through a product operator approach for a two spin (I and S) coupled system. For the I spin, only two out of the four signals present before the acquisition period are observable. These are inphase, ( J IS t ) sin ( Ω 1 I t1 ) I X cos π Equation 1.19 and antiphase, ( J ISt1 ) sin( Ω I t ) 2I Z SY sin π 1 Equation 1.20 components. J IS is the J-coupling constant between spin I and S, t 1 the time period between the two 90º pulses, and Ω I the frequency difference between the off resonance peak (of spin I) and the resonance frequency. The inphase term gives a signal in the direct dimension (ω 2 ) at Ω I and is modulated in the indirect dimension (ω 1 ) by the term sin( Ω t ) 1 1. The antiphase component gives a signal at Ω S in the direct dimension but is again modulated by ( ) sin Ω I t 1 in the indirect dimension. Consequently, the inphase term results in a diagonal peak at Ω I while the antiphase term results in a cross peak at Ω S (direct dimension) and Ω I (indirect dimension). After Fourier transformation, the inphase term (Equation 1.19) yields four double absorption mode lineshapes, 27

38 1 + AI ( Ω I + πj IS ) AS ( Ω S + πj IS ) AI ( Ω I + πj IS ) AS ( Ω S πj IS ) 4 Equation AI ( Ω I πj IS ) AS ( Ω S + πj IS ) AI ( Ω I πj IS ) AS ( Ω S πj IS ) 4 Similarly, after Fourier transformation four double absorption mode lineshapes are also obtained for the antiphase term (Equation 1.20), 1 + AI ( Ω I + πj IS ) AS ( Ω S + πj IS ) 4 1 AI ( Ω I + πj IS ) AS ( Ω S πj IS ) 4 Equation AI ( Ω I πj IS ) AS ( Ω S + πj IS ) AI ( Ω I πj IS ) AS ( Ω S πj IS ) 4 These signals give rise to diagonal and cross peaks obtained after analysis of the I spin. Complementary results are obtained after analysis of the S spin, giving the characteristic COSY spectrum (Figure 1.13) for a two spin system. 92 Figure An example of coupling in a COSY experiment between spins at frequencies of Ω I and Ω S. The black peaks are positive peaks, while the grey peaks are negative peaks. 28

39 1.7.3 Exchange Spectroscopy (EXSY) Exchange spectroscopy can be conducted to elucidate exchange networks. The pulse sequence is the same as that used in a two-dimensional nuclear Overhauser effect spectroscopy (NOESY) experiment (Figure 1.14). The latter spectrum is similar to the COSY and is employed to identify spins that are in close proximity ( < 5 Å) to each other. The difference between the two experiments is that COSY cross peaks result for coherence transfer though coupling, whereas NOESY cross peaks are due to cross relaxation. A brief description of the NOESY experiment is provided using a product operator approach. During the evolution period, the spins are frequency labelled according to their characteristic resonance frequencies. After the second 90º pulse, four product operator terms are obtained of which only the I z term, ( cos ( π J IS t ) cos ( Ω 1 I t1 ) I Z ) Equation 1.23 is relevant for this process. Cross relaxation can occur during the mixing period, τ, giving rise to the signals, ( Ω I t )( 1 RZ ) I Z + ( Ω St ) σ τ I Z + ( RZ + σ ) τ I Z cos 1 τ cos 1 Equation 1.24 where R z is the self-relaxation and σ the cross relaxation rate constants. The other terms have the same definitions as in the above presented COSY. The first term is modulated by Ω I in ω 1 resulting in a diagonal peak, while the second is modulated by Ω S in ω 1 giving a cross peak. The third term, an axial term, is not modulated in ω 1, consequently giving rise to a signal at a frequency of zero in the indirect dimension. These peaks can be eliminated by incorporating a two phase (x, -x) cycle. The third 90º pulse is employed to detect magnetization. In much the same manner, the frequency modulated z- magnetization created after the second 90º pulse is affected by exchange processes that also give rise to cross peaks. Figure A schematic of a two-dimensional NOESY pulse sequence. 29

40 1.7.4 Selective Inversion Recovery (SIR) The type of NMR experiment chosen to calculate exchange rate constants depends on the exchange regime under investigation. Line shape analysis is used in the intermediate exchange regime, where the exchange rate is comparable to the difference between chemical shifts (usually ~ Hz). Offset saturation methods can be used for faster exchange processes (usually > 10 3 Hz), while selective inversion recovery experiments can be employed for slower exchanges (usually Hz). 93 The SIR is a T 1 based experiment developed by Forsen and Hoffman. 94 based on perturbing the system away from equilibrium, by applying a selective 180º pulse, and then watching the effect on all peaks in the spectrum over time (Figure 1.15). The upper limit of its capabilities is dependent on the chemical shift difference between peaks (i.e. must have a distinct peak to invert), while the lower boundary is related to the relative T 1 relaxation rate (for the inverted peak) to exchange rate ratio. The latter must be fast enough to cause a change in the magnetizations of the non-inverted exchangeable peaks. The T 1 relaxation rate limits the experiment length (for one transient) because of signal loss due to relaxation. It is Figure A schematic of the selective inversion recovery pulse sequence. 30

41 2 Experimental Section 2.1 Materials Sodium fluoride (99%), Y(NO 3 ) 3 6H 2 O (99.9%), Gd(NO 3 ) 3 6H 2 O (99.9%) and poly(acrylic acid) (PAA, MW = 1800 g/mol) were purchased from Sigma Aldrich (Mississauga, ON, Canada). Citric acid (99.5%), aqueous ammonium hydroxide (28-30%) and ph indicator strips (ph range 5-10) were acquired from EMD Chemicals Inc. (Gibbstown, NJ, USA). All chemicals were used as received. 2.2 Methods Citrate coated lanthanide nanoparticle synthesis Citrate-coated lanthanide trifluoride (Cit-LnF 3 ) nanoparticles (NPs) were produced through a slight modification of the van Veggel protocol. 71 A solution containing citric acid (0.41 g, 2.13 mmol) and sodium fluoride (0.13 g, 3.00 mmol) in distilled water (25 ml) was neutralized using concentrated ammonium hydroxide. The solution was then heated in a round bottom flask, on an IKAMAG RET basic safety magnetic stirrer (Wilmington, NC, USA), to a temperature of 75 ºC using an oil bath. Concurrently, the magnetic stirrer was set to 375 rpm and a one inch magnetic vane was employed. An aqueous lanthanide solution (2.13 mmol of Ln(NO 3 ) 3, 2 ml of distilled water) was titrated into the round bottom flask at an average rate of 2 ml/hr, using a 200 µl micropipette, in 10 µl increments. Initially, a slow rate was used (each drop added was allowed to re-dissolve) but as the reaction turned cloudy (~ 20 min) the additions were made at regular intervals of fifteen seconds. After ~ 30 min, the additions were made at shorter, but still regular intervals of ten seconds. The resulting mixture was left to react for a further two hours. Ethanol was then added to the mixture to precipitate the nanoparticles, which were subsequently isolated though centrifugation (7500 rpm, 5 min., supernatant decanted). This was followed by resuspension in water, precipitation with ethanol and centrifugation once again (7500 rpm, 5 min., supernatant decanted). The nanoparticles were then dried in a dessicator under house vacuum. 31

42 2.2.2 PAA coated lanthanide nanoparticle synthesis Citrate coated LnF 3 nanoparticles were first synthesized according to the above protocol. PAA was substituted for citrate by manual addition of 2 ml of a 128 mm PAA solution at a rate of 1 drop/second using a 2 ml syringe with a 26G1/2 needle. The exchange was allowed to proceed for 12 hours at 75ºC. Identically to the above Cit-LnF 3 NP procedure, nanoparticles were isolated though two ethanol precipitation steps Electron Microscopy Nanoparticles were verified though use of a Hitachi H-7000 scanning transmission electron microscope (STEM) operating at an accelerating voltage of 200 kv. The grids were prepared by coating 200 mesh copper grids (Sigma Aldrich, Mississauga, ON, Canada) with a solution of 0.5% formvar in 1,2-ethylene dichloride. A drop of lanthanide nanoparticle solution (~10 mg/ml) was placed on the grid for a minute, after which it was drawn off by capillary action using a Kimwipe and air-dried. Dark field images were then acquired Nanoparticle Size Analysis Nanoparticle sizes were measured using ImageJ. Upper and lower threshold limits (Image>Adjust>Threshold) and the scale (Analyze>SetScale..) were set using options in the program. Particles sizes were typically determined (Analyze>Analyze Particles..) with the following processing conditions: size between 100/314 infinity nm 2, and a circularity between Using the generated data, size distributions were plotted and average size and a polydispersity index value was computed for each synthesized nanoparticle batch Zeta potential measurements Zeta, ζ, potential measurements were conducted using a Zetasizer 3000HS (Malvern Instruments). The instrument was flushed with 15 ml of deionized water, before 5 ml of diluted nanoparticle solution was injected into it. Five replicates were taken for each nanoparticle sample, the average of which was reported as the ζ potential. 32

43 2.2.6 NMR NMR Calibration Standards A 90:10, D 2 O:H 2 O sample was prepared to calibrate the gradient strength of the 500 and 600 MHz spectrometers. In addition, an ethylene glycol standard (Varian Inc., Palo Alto, CA) was used to calibrate the temperature between 0 and 45ºC NMR Sample Preparation A citrate control (35 mm citric acid, 20 mm NaOH, ph 9.0), Cit-YF 3 NP (10 mg/ml NP, 10 mm NaOH, 90:10 D 2 O:H 2 O, ph 8.5), 1:1 Y:Cit control (23.9 mm citric acid, 25.3 mm Y(NO 3 ) 3, 106 mm NaOH, 90:10 D 2 O:H 2 O, ph 8.5), 1:10 Y:Cit control (21.7 mm citric acid, 2.1 mm Y(NO 3 ) 3, 31.9 mm NaOH, 90:10 D 2 O:H 2 O, ph 8.5) and PAA-YF 3 NP (19.3 mg/ml NP, 10 mm NaOH, 95:5 D 2 O:H 2 O, ph 8.5) samples were prepared in 5 mm Economy NMR tubes (Wilmad-Labglass, Vineland, NJ). The ph was tested using ph indicator strips (EMD, Gibbstown, NJ, USA) NMR Experiments One-dimensional proton NMR, COSY, PFGSTE, EXSY and SIR (CRYO) experiments were conducted on a Varian Unity 600 spectrometer using either an HFCN quad probe (Varian Inc., Palo Alto, CA) or a HCN cold-probe (Varian Inc, Palo Alto, CA). The pulse sequences for the various experiments are provided in the Introduction section. 1 H NMR experiments were performed at MHz. Typically, 90º pulse lengths of 10.5 µs and 8 µs were used for the HFCN and cold-probe respectively, while an exponential multiplication equivalent to 1 Hz line broadening was employed before Fourier transformation. A sufficient recycle delay (usually 5 x T 1 ) was selected to allow for essentially full relaxation in the 1D1H, PFGSTE, EXSY and SIR experiments. Parameters specific to each of the different NMR experiments are provided below Correlation Spectroscopy (COSY) A standard 1 H magnitude only COSY was acquired with a sweep width of 5500 Hz. The number of increments taken during the evolution period was 512, resulting in a 10.7 Hz resolution in the indirect dimension. Eight transients were taken per increment. 33

44 Pulsed Field Gradient Stimulated Echo (PFGSTE) Diffusion Experiments A pulsed field gradient stimulated echo (PFGSTE) experiment 95, 96 was modified to incorporate a CHIRP based z-filter, 91 between the first and second 90º pulses, to reduce the effects of undesired coherences. The duration of the CHIRP pulse was typically set to 25 ms, while the simultaneously applied gradient had the same duration and an amplitude of 0.76 T/m. This gradient was calibrated as per the recommendations of Thrippleton and Keeler. 91 employed to filter out unwanted echoes. 90 In addition, a sixteen step phase cycle scheme was General parameters of the experiment are as follows. The PFGSTE gradients were applied along the z direction and turned on for durations of 3 ms. The spin echo delay was typically set to between 5 10 ms and the T1 delay between ms, the spectral width was 4 khz and the data size 4-K. Finally, the gradient amplitude was arrayed between 15 values, chosen such that a ninety percent decay in peak intensity was achieved Exchange Spectroscopy (EXSY) A standard Varian EXSY pulse sequence, which included a pre-programmed CHIRP and homospoil gradient, was employed. The parameters of the CHIRP were once again calibrated according to the methods of Thrippleton and Keeler. 91 The duration of the CHIRP and the homospoil gradient was 50 and 2.5 ms respectively. The strength of the CHIRP gradient was 0.67 T/m while that of the homospoil was 0.89 T/m. Other experimental parameters include a spectral width of 5.5 khz, 1.5-K data points, eight transients and 256 increments, the latter of which yielded 21 Hz resolution in the indirect dimension. Four 2-D EXSY spectra were collected with mixing times of 100, 200, 400 and 800 ms Selective inversion recovery (SIR) Selective inversion recover experiments were conducted using a soft 180º E- BURP-1 pulse, 97 of length 25.3 ms, centered on the resonance to be inverted. The spectral width was 10 khz and the data size 20-K. Twenty exchange time delay values ranging from 1 ms to 5 s were employed. Each arrayed delay value was signal averaged over 512 transients. Standard non-selective inversion recovery experiments were used to estimate spin-lattice relaxation rates for the various observed states. 34

45 3 Results and Discussion Macromolecular, lanthanide contrast agents, with potential MRI-based cancer detection capabilities, have been investigated. Citrate-coated gadolinium trifluoride nanoparticles (Cit-GdF 3 NPs) were synthesized and their citrate surface binding and exchange properties characterized. Although these nanoparticles have been produced before, 71 more refined reaction conditions are presented that allow for greater control of important factors such as the size and dispersity of the nanoparticles. Furthermore, a new two-step synthesis protocol is advanced, and verified, whereby poly(acrylic acid)-coated gadolinium trifluoride nanoparticles (PAA-GdF 3 NPs) are synthesized from Cit-GdF 3 NP precursors (A similar ligand exchange and verification has been achieved with quantum dots 98 ). The tremendous versatility of this procedure is that any ligand can bind to the surface of the nanoparticle so long as it can adhere more strongly than citrate. This circumvents the need to venture into potentially difficult and tricky direct synthesis methods, while yielding the benefits afforded by new ligands. Note: All references to polydispersity, dispersity or dispersion, henceforth mentioned, refer to nanoparticle size distributions unless qualified otherwise. 3.1 Ligand Properties citrate and poly(acrylic acid) (PAA) Two different ligands were used to coat the lanthanide trifluoride nanoparticles: citrate and poly(acrylic acid) (Figure 3.1). The motivation behind the use of a polymer was driven by the deficiencies of citrate: primarily its relatively low binding affinity for the surface of the nanoparticle, which leads to higher in vivo toxicity and lower stability, and the comparatively low relaxivity of this nanoparticle-ligand conjugate. Binding affinity, absence of cross-linking ability between nanoparticles, and ease of functionalization were considerations taken into account when selecting the polymer type and length. The chosen PAA, a 25-mer (PAA 25 ), has a total of 25 carboxylic acid groups, as each acrylate monomer possesses one such group. Since the number of carboxylic acids represents the number of theoretical binding sites, PAA 25 may be expected to bind with greater affinity to the surface of the nanoparticle than citrate (3 carboxylates, 1 hydroxyl). In addition, the acidic nature of the polymer (pka ~ 4.58) facilitates a 35

46 stronger electrostatic interaction, at neutral conditions, with the positively charged nanoparticle surface. A short polymer size was chosen to prevent cross-linking of nanoparticles, where an acceptable PAA upper size length was restricted to a value less than the nanoparticle diameter. This constraint required a quantitative estimate of polymer size. Several different measures of length, such as the mean-squared end-to-end length (Flory radius), mean-squared radius of gyration and hydrodynamic radius, can be employed. However, as a short polymer in the fully deprotonated state is likely to exist in its fully extended conformation, the contour length, which is the physically possible maximum length, might be the most appropriate measurement. For PAA 25 this is calculated to be ~ 7 nm (assuming a carbon covalent radius of Å 99 ). As the synthesized nanoparticles are on average 50 nm or greater in diameter, PAA 25 is, by and large, unlikely to cause crosslinking. The presence of carboxylate groups is advantageous because they facilitate the conjugation of targeting molecules to the surface of the nanoparticle. For example, folic acid, whose receptor is overexpressed in forty percent of human cancers, 100 can be attached to PAA though a water soluble synthesis using N-(3-Dimethylaminopropyl)-N - ethylcarbodiimide hydrochloride (EDAC), N-hydroxyl succinimide (NHS) and an ethylene diamine linker. 101 Figure 3.1. The structures of (a) citric acid and (b) poly(acrylic acid). The Greek letters α and β are used to help discriminate the α-hydroxyl, α-carboxylate and the two β- carboxylates in citric acid. The letter n represents the number of monomers comprising the polymer. 36

47 3.2 Nanoparticle synthesis procedures Citrate-coated lanthanide trifluoride nanoparticle synthesis The citrate-coated lanthanide trifluoride nanoparticle (Cit-LnF 3 NP) synthesis involved the dropwise addition of a Gd(NO 3 ) 3 stock into a neutralized solution containing NaF and citrate (T = 75 ºC, stirring ~ 375 rpm on a IKAMAG RET basic safety magnetic stirrer). Upon addition of the first drop (~ 10 µl), a white precipitate (GdF 3 aggregate) initially formed before dissolving over a period of a few seconds. Each subsequent drop was added only after dissolution of the previous one. After addition of about 0.15 ml, the solution displayed a tinge of white, likely indicating the formation of nuclei large enough and in sufficient numbers to scatter light. The mixture turned cloudy and opaque after ~ 0.6 ml of lanthanide stock was added and the disappearance of the next drop could no longer be observed. At the end of the reaction the mixture remained an opaque white (If left to stand for an extended time (a few hours) in the round-bottom flask, a white precipitate was often observed, perhaps indicating that the nanoparticles were not colloidally stable under natural reaction conditions. Although an alternative explanation could be that large, insoluble non-nanoparticle material also formed, this was not observed in sufficient frequency (evaluated using STEM) to completely attribute the precipitate to this reason). Finally, addition of a non-solvent (e.g. acetone or ethanol) resulted in immediate precipitation of the nanoparticles from solution Poly(acrylic acid)-coated lanthanide trifluoride nanoparticle synthesis Following previous work, 102 a direct approach was initially taken when attempting to synthesize poly(acrylic acid)-coated gadolinium trifluoride nanoparticles (PAA 25 -GdF 3 NPs). Into a neutralized solution of PAA 25 and NaF, a stock solution of Gd(NO 3 ) 3 was added dropwise (T = 75 ºC, with stirring). Although one variation of the synthesis yielded remarkably monodisperse particles (Figure 3.2), synthesis reproducibility proved elusive and an indirect, two-step approach was henceforth taken to produce PAA 25 -GdF 3 nanoparticles. This firstly involved the production of citrate-coated gadolinium trifluoride nanoparticles (Cit-GdF 3 NPs), 71 followed by exchange of citrate for PAA 25 (Refer to experimental for specific details). Observations during and at the end of the reaction were similar to those encountered for Cit-LnF 3 NPs. 37

48 3.3 Synthesis Verification and Characterization of Size Polydispersity Nanoparticle formation, and consequently synthesis success, was primarily verified using annular darkfield scanning transmission electron microscopy (STEM). In addition to images, energy dispersive X-ray (EDX) linescans were taken to verify nanoparticle composition. Figure 3.3 depicts one obtained for nanoparticles composed of europium and gadolinium. EDX analysis indicates a nanoparticle core composed of the elements gadolinium, europium, fluorine and likely sodium. Weaker signals (oxygen and carbon) corresponding to the citrate ligand are also observed. Unfortunately, the same analysis could not be obtained for subsequent Cit-GdF 3 and PAA-GdF 3 nanoparticles because the EDX feature is currently out of operation (and has been for the last few months). Sizes were extracted from dispersed nanoparticle images, which had high uniform contrast, using ImageJ, a computer program. As size was often polydisperse, the count was biased toward large sizes which appeared brighter in the images than much smaller ones. The number of nanoparticles counted varied between , which is much lower than the recommended number (>1000) for statistically significant analysis 103 (this low number was due to the prohibitive cost of imaging). However, the dispersion appeared to mesh satisfactorily with visual conclusions, and as only rough trends in nanoparticle synthesis were sought, these counts were deemed sufficient to ascertain them. Polydispersity indices (PDIs) were chosen to evaluate the breadth of the distribution over alternative methods such as standard deviation and variance. As PDIs were not commonly encountered when discussing nanoparticle size distribution in literature, a brief attempt will be made to justify their use. They were primarily employed because the calculated standard deviations are large numbers (in the ten thousands for NPs) which are cumbersome to deal with, whereas the PDI presents a neater way of presenting the same measure. In fact, the standard deviation can be quantitatively related to the PDI as will be shown. The polydispersity index is most commonly encountered when describing polymer molar mass dispersity and is stated to be, 38

49 M W PDI = Equation 3.1 M N where M W and M N are the weight average molar mass and the number average molar mass respectively. 104 Similar equations can also be used to represent nanoparticle size polydispersity. The PDI can be represented by the ratio, D W PDI = Equation 3.2 DN where D N and D W are defined as, N xdx x D N = Equation 3.3 N D W x x N xdx x = N D x x x 2 Equation 3.4 and where N x is the number of molecules of length x and D x the diameter of molecules of length x. The relationship between the PDI and the standard deviation (σ) can be shown to equal, 2 σ DW = 1 Equation 3.5 DN DN For a perfectly monodisperse sample, the PDI equals polydispersity leads to larger values. nanoparticles ranged from 1.02 to Increasing sample Typical PDI values for the synthesized 39

50 3.4 PAA 25 -LnF 3 NP - Direct Synthesis Images One synthesis yielded relatively monodisperse particles (Figure 3.2) that displayed T 1 and T 2 relaxivities of 38 and 80 Hz (mg/ml) -1 (1.5 T, 20 ºC), the former of which is about six times larger than that observed for commercial Gd chelates (e.g. Gd- DTPA, 6 Hz (mg/ml) -1, 1.5 T, 37 ºC). 105 The size of the nanoparticles ranged from nm, with a mean size of 112 nm and a polydispersity index (PDI) of Although this synthesis proved irreproducible, the high relaxivities obtained served as the impetus behind the creation of a new, two-step PAA 25 -LnF 3 nanoparticle synthesis protocol. On average, all nanoparticles (PAA-LnF 3 and Cit-LnF 3 NPs) were roughly spherical, with an irregular surface, regardless of lanthanide composition or ligand coating (Figure 3.2). This irregular surface may be due to either fast growth of the lattice or a hierarchal structure. In the former, nanoparticles may mature with defects in the lattice resulting in irregular expansion, whereas in the latter, building blocks may first form smaller components, which then nucleate to give the final nano-aggregate. It is possible that high resolution transmission electron microscopy (HR-TEM) might provide sufficient resolution to discriminate between these growth mechanisms. Regardless, the larger nanoparticle surface to volume ratio, due to surface irregularity, is beneficial as ninety percent of relaxivity is attributed to inner sphere effects

51 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) Figure 3.2. STEM images of relatively monodisperse PAA25-GdF3 NPs with an associated size dispersion profile. Evidence of finer structure, in a PAA25-YF3 nanoparticle, is visible in the central image. 41

52 Figure 3.3. An energy dispersive X-ray linescan of 90/10 Gd/EuF 3 nanoparticles. In the bar graphs, the x-axis gives the distance along the drawn line in the STEM image, while the y-axis is a measure of the X-ray signal observed. Significant signals from elements comprising the core of the nanoparticle (europium, gadolinium, fluorine and sodium) are present. Weaker signals from carbon and oxygen, likely from citrate, are also observed. A large background carbon signal is due to the formvar coating. Titanium and barium were selected as controls, as they were not present in the nanoparticle, and their very weak signals were as expected. 42

53 3.5 Control of Nanoparticles Size and Dispersity Nanoparticle biodistribution is closely linked to particle size. As a case in point, whereas small gold nanorods (~ 10 nm) were found to collect in liver, spleen, kidney, testis, thymus, heart, lung and brain, larger particles (50 and 250 nm) were only found to localize in the liver and spleen. 106 is crucial to desired function. Hence, control of nanoparticle size and polydispersity Cursory experiments were attempted to ascertain rough trends in the synthesis of Cit-GdF 3 nanoparticles. The rate of lanthanide addition and amount of lanthanide added were varied to examine effects on size and dispersity. Lump-sum additions of lanthanide were experimented with to observe the same. Unless otherwise stated, all reaction conditions were kept the same as in the protocol stated in the Experimental section Rate of Lanthanide Addition Based on the nucleation and growth model (Figure 1.9), it was expected that a fast rate of addition would yield smaller, more polydisperse nanoparticles if it results in a longer nucleation period. In such a period, the concentration of the free lanthanide trifluoride (LnF 3 ) building block will not drop below the nucleation concentration between additions of the lanthanide nitrate. An increase in the duration of the nucleation period will cause two effects. Firstly, it will lead to a greater number of nuclei forming. Secondly, it may induce an increase in polydispersity because nucleation and growth occur simultaneously during this period. Consequently, for a fast addition, the smaller size is due to the same number of building blocks being spread over a greater number of nuclei. Gadolinium addition was varied between slow (2 ml in 1 hr), intermediate (2 ml in 40 min) and fast (2 ml in 7 min) rates of additions (Figure 3.4). The slow addition yielded a larger average size, while the intermediate and fast additions produced nanoparticles of similar size. The size range of the slow addition was roughly twice as large as that of the intermediate and fast additions, while the PDI showed so discernable trend across the series (Table 3.1, Figure 3.4). In the slow addition, the large distribution likely indicated that nucleation occurred for a longer period than in the intermediate and fast additions perhaps due to fluctuation of the building block concentration about the 43

54 nucleation concentration during gadolinium addition. This might indicate that gadolinium addition, even in the case of a slow rate, is still occurring fast enough to cause multiple nucleation events. The relatively similar results for the intermediate and fast additions might demonstrate that there is a certain limit to the rate of addition above which increasing the rate does not yield significant differences in the size range. It is important to note, that slow and fast are relatively terms, whose relativity depend on the rate at which the building block is consumed (which is unknown). What was termed intermediate might, in fact, be fast, and what was thought to be slow was clearly fast enough to cause continuous nucleation. Table 3.1. Nanoparticle size distribution characteristics Size (nm) Parameter Varied Parameter value Minimum Maximum Range Average PDI Gadolinium rate of addition Slow (1 hr) Intermediate (40 min) Fast (7 min) Total gadolinium mmol* Lump-sum 1 ml addition 1.5 ml Yttrium Cit-YF nanoparticles PAA-YF * The total lanthanide added was half what is normally added (i.e mmol) 44

55 80 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) 80 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) 80 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) Figure 3.4. STEM images of syntheses with (a) slow (2 ml in 1 hr), (b) intermediate (2 ml in 40 min) and (c) fast (2 ml in 7 min) gadolinium nitrate rates of addition. Their corresponding size dispersion profiles are shown. 45

56 3.5.2 Reduction in Total Lanthanide Feedstock The amount of gadolinium feedstock and the rate of addition were both halved (1 ml in 1 hr), compared to the previously mentioned slow addition, in the expectation of affecting nanoparticle size and increasing monodispersity. A remarkable increase in the latter property was observed with roughly seventy percent of nanoparticles falling within a 50 nm range (Figure 3.5). The more monodisperse results obtained here mesh with the slow rate of addition (2 ml in 1 hr) results obtained in the previous section, where it was postulated, that decreasing the rate of addition might result in reduction or elimination of multiple nucleation events. The slower rate of addition appears to have reduced these events leading to more monodisperse nanoparticles of fairly large size. It is possible that if the same reaction conditions were used, except with an even slower rate of addition, a further improvement in monodispersity may be triggered. This is similar to what is predicted from the La Mer diagram, and what some published nanoparticle syntheses hope to or have achieved, 107 where effective separation of the nucleation and growth processes lead to a burst of nucleation, followed by diffusion controlled growth. The change in the lanthanide to fluoride ratio poses a potentially troubling problem as it might reduce the affinity of a ligand for the nanoparticle surface. In the original synthesis, this ratio was 1:2.3 resulting in a fluoride deficiency towards the end of the reaction. A distribution of LnF 3, LnF 2 +, LnF 2+ and Ln 3+ species may thus form, perhaps coating the nanoparticle with a more positive charge than otherwise (Zeta potential measurements presented later). However, as in this experiment the ratio was 1:4.6, lanthanide trifluorides are still forming towards the end of the addition possibly reducing the effective positive charge on the lanthanide surface. In the future, it would be interesting to verify the presence or absence of this affinity effect though use of NMR experiments (pulsed field gradient stimulated echo, selective inversion recovery) which can probe ligand exchange. complementary data. Additionally, zeta potential measurements can provide 46

57 80 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) Figure 3.5. STEM images of the same synthesis, at different magnifications (scale on the lower right of the images), where half the usual amount of lanthanide feedstock was added. Although the nanoparticle may appear to be aggregated in some of the images, this is actually due to improper focusing. The corresponding size distribution is also presented Lump-sum lanthanide stock additions In attempts to reduce the average nanoparticle size, it was hypothesized that a large lump-sum addition of the lanthanide nitrate stock would produce a larger number of nuclei. As stated previously, a greater number of nucleation centers for the same given amount of building block reduces average nanoparticle size. Also, if the resulting nucleation period was still short enough then no effective change in polydispersity would occur as nuclei would all form at approximately the same time. Two separate reactions were conducted where 1.00 and 1.50 ml of the 2.00 ml lanthanide stock was added in lump-sum additions. This was followed by slow addition of the remaining stock. The 1.00 ml lump-sum addition resulted in nanoparticles of a smaller average size and range (Figure 3.6, Table 3.1), while the 1.5 ml addition 47

58 produced larger, more polydisperse nanoparticles. The observed increase in polydispersity of the latter was expected due to the longer nucleation period, however the increase in average size compared to the 1.00 ml lump-sum addition was not. 80 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) 80 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) Figure 3.6. STEM images of (a) 1.00 and (b) 1.5 ml lump-sum additions of lanthanide feedstock. Their associated size dispersion profiles are provided. In conclusion, three methods have been covered that have shown potential to control nanoparticle size: lump-sum addition, rate of addition, and the amount of lanthanide added. Controlling the rate of addition and the amount of lanthanide appear key to producing monodisperse nanoparticles via a coprecipitation method. Other mechanical ways, such as stir-rate variation can also be used to control size. From the synthesis conducted, the general impression is that control of average size is easy but that of mondodispersity much harder. Also, as numerous ways: seeding growth, reflux ripening, size-selective precipitation and electrophoresis, have been used in the synthesis of monodisperse gold nanoparticles, 108 it is possible that some of these methods can also achieve the same for lanthanide nanoparticles, without undue effort. 48

59 3.6 NMR Studies Due to the importance of ligand interactions with the nanoparticle, a major part of this work involved the characterization of citrate binding to the nanoparticle surface. As the paramagnetic gadolinium induces severe broadening and consequent loss of peaks, diamagnetic lanthanide substitutes were used to form the core of the nanoparticle. Lanthanum (La), lutetium (Lu) and yttrium (Y) with ionic radii of 1.22, 1.03 and 1.08 Å respectively, all are able candidates. Although the latter is not strictly a lanthanide, it nevertheless displays properties very similar to these elements and is often grouped with them. Since the chemical properties of the lanthanides are usually dictated by ionic radius, yttrium was chosen because it has the one most similar to gadolinium (1.11 Å). 109 Cit-YF 3 and PAA-YF 3 nanoparticles were synthesized in the same manner as their gadolinium counterparts, except for the lanthanide employed. Nanoparticle presence was verified for all samples used for NMR studies. The second step (titration of the polymer into the Cit-YF 3 NP solution) of the two-step synthesis procedure for PAA-YF 3 NPs was found to result in imperceptible changes in nanoparticle size and is consequently expected to only result in direct substitution of PAA for citrate. Images and size distributions of yttrium nanoparticles are given in Figure 3.7 while average sizes, ranges and PDI values are provided in Table 3.1. The averaged (four values) zeta (ζ) potential was found to be -3.6 mv for Cit-YF 3 NPs indicating a negative coating on the nanoparticle surface. However, this was still considerably lower than that of their paramagnetic analogs (Cit-GdF 3 NPs [-73 mv] and PAA-LnF 3 NPs [-37 mv]). 110 To ensure the absence of excess citrate, which could influence later calculations of the population of free or bound (to the nanoparticle surface) ligand, multiple precipitations were conducted with ethanol and the amount of citrate in the nanoparticle precipitate quantified after each step using an acetonitrile standard. This particular standard was chosen because it was miscible with water, displayed peaks shifted away from citrate, was unlikely to bind to the nanoparticle, and had a strong single resonance in the proton spectrum. After each precipitation step, a measured weight of nanoparticle was collected and acidified (ph < 1) in an NMR tube to deprotonate citrate (pka 1 = 3.220, pka 2 = 4.837, pka 3 = at 0 ºC 111 ) thereby reducing its affinity for the nanoparticle surface. Consequently, peaks due to bound fractions disappeared yielding only the four 49

60 free citrate resonances which could be more accurately quantified. Acidification, for quantification purposes, is not a necessary step but can lead to more accurate estimates of the total citrate population. One ethanol precipitation step was found to lead to 28 % citrate by weight, while two resulted in 24 %. This likely indicates that very little excess citrate is present even after just one precipitation step. 80 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) 80 Nanoparticle Abundance (%) Nanoparticle Size Range (nm) Figure 3.7 (a) Cit-YF 3 and (b) PAA-YF 3 nanoparticles with their respective size distributions One-Dimensional Proton (1D 1H) Spectra of Citrate and Cit-YF 3 Nanoparticles One-dimensional proton (1D 1H) spectra were acquired for a citrate control and Cit-YF 3 nanoparticles for preliminary visualization and comparison of the systems. In addition, all resonances in the Cit-YF 3 spectrum were verified to unambiguously belong to citrate. Finally, a variable temperature series of 1D 1H spectra were used to evaluate the presence of exchange and the effect of temperature hysteresis. 50

61 Similar to literature, 112 the control displayed characteristic AB peaks arising from the strong, two-bond coupling of methylene protons (Figure 3.8. a) which result because of the chiral α-carbon and slow rotation about the α-carbon, β-carbon bond. For such a system, several equations can be used to equate frequency differences to parameters like the J-coupling constant for protons A and B (J AB ), and the chemical shift difference (δ) between the two proton resonances (ν A and ν B ). 113 The former is described by the equation, J AB = ν 1 ν 2 = ν 3 ν 4 Equation 3.6 where ν 1 and ν 2 are the outer and inner frequencies of the downfield doublet respectively, and ν 3 and ν 4 the inner and outer frequencies of the upfield doublet respectively. The latter is given by the relationship, [( ν ν )( ν )] 1/ 2 δ = ν 3 Equation 3.7 Finally, the ratio between the inner and outer frequencies (doublet ratio) can also be equated to the observed frequencies, Intensity of inner resonances ν 1 ν 4 doublet ratio = = Equation 3.8 Intensity of outer resonances ν 2 ν 3 For free citrate (citrate control), J AB was 15.0 Hz, δ 48.7 Hz, and the doublet ratio 1.61 (600 MHz, ph = 8.5, 0 ºC). The doublet ratio is important because it can provide a quantitative indication of the type of coupling (strong vs weak) and, as such, was used to derive some of the inferences stated here. For a strongly coupled system (AB), such as citrate, this ratio is greater than one, while for its weakly coupled counterpart (AX) it is equal to one (i.e. the resonances in each doublet are of the same intensity). 51

62 Figure 3.8. One dimensional spectra of (a) citrate (ph ~ 8.5, 600 MHz), with assigned proton resonances, (b) Cit-YF 3 NP (ph ~ 8.5, 600 MHz) and (c) acidified Cit-YF 3 NP (ph < 1, 500 MHz). All samples were taken at 0 ºC. The chemical shift and J-coupling difference between the citrate peaks in (a) and (c) is due to the ph difference between the samples. The Cit-YF 3 NPs displayed a multitude of peaks possibly corresponding to free and bound citrate states (Figure 3.8. b). Of these peaks, the four largest in the spectrum (2.22, 2.19, 2.11, 2.08 ppm) were inferred to be citrate in the free state because their doublet ratios (1.38 and 1.21) were closest to that of the citrate control (1.61). Although, 52

63 they seem to be quite different, it is argued, and proved later, that this is due to underlying bound states that skew the ratios. The rest of the peaks were taken to correspond to bound states. The best separated of these were two doublets, present at 2.68 and 2.59 ppm, that appeared to move from AB to AX systems. Their doublet ratios at 1.11 and 1.09 respectively, provided clear evidence of such a movement. The change in the ratios may be driven by a change in chemical shift induced by binding (carboxylates or hydroxyl) adjacent to one proton of the methylene, but not the other. The large upfield shift of the free citrate peaks in the nanoparticle sample (~ 0.3 ppm) was unexpected as it should have been identical to that observed in the citrate control. Similarly, the upfield shift of the bound citrates states were likewise unexpected as binding to the positive lanthanide surface should have induced, though an electron withdrawing effect, a downfield shift in the methylene resonances. It is possible that the nanoparticles (10 mg/ml) may have affected the chemical shift of the water reference peak. The one dimensional proton spectrum of an acidified nanoparticle sample was acquired, as a control, to confirm that all peaks in the spectrum were due to citrate. In some nanoparticle syntheses, citrate is often used as a metal reducing agent where the end result produces metal nanoparticles and citrate thermal decomposition products. The one dimensional proton spectrum of these products bears a resemblance to that of Cit-YF 3 NPs. 114 Consequently, an experiment was conducted to eliminate this possibility. It was hypothesized that in an acidified (ph < 1) nanoparticle sample, if the peaks were due to free and bound states, protonation would cause disappearance of bound peaks. Conversely, if due to thermal decomposition then these peaks would still be observed. Spectral evidence (Figure 3.8. c) indicates that the former hypothesis is true, and consequently renders the above conclusions and conjectures still meaningful. Spectral analysis is complicated by the ph and temperature dependence of citrate chemical shifts and J-couplings. These parameters vary over a ph range related to changing protonated, carboxylate fractions. Chemical shifts, for example, were found to vary by ~ 0.4 ppm between a ph range of A similar dependence on temperature is possible, due to slight changes (~ 0.1 ph units) in the pka values of citrate carboxylic acids over 0 50 ºC. 111 The ph dependence of citrate NMR parameters in 53

64 acidified and basic samples was found to be comparable to literature values (Table 3.2, note that the temperature is different between experimental and literature data). Further complications are posed by the ph-dependent chemical shift nature of water, to which the citrate peaks are referenced. Thus, NMR results can only be compared across samples under identical acidic/basic and thermal conditions. It should be noted that basicity was tested, in all samples, with ph paper with an error of ~ 0.25 units. Table 3.2 Comparison of different parameters in acidified Cit-YF 3 NP and citrate control samples with literature values ph < 1 ph ~ 8.5 Acidified NP sample (500 MHz, 0 ºC) Literature (400 MHz, 24 ºC) Citrate Control (600 MHz, 0 ºC) Literature (400 MHz, 24 ºC) Doublet ratio 1.41 Unknown 1.61 Unknown J AB (Hz) δ (Hz) 48.7 Unknown 88 Unknown (Hz) δ cit (ppm) Doublet ratio, J AB and δ are as defined in the discussion above is the frequency difference between the midpoints of the two doublets δ cit is the chemical shift of the midpoint of the citrate resonances Literature values from Moore et al. 112 Variable temperature (VT) one dimensional spectra were obtained to offer evidence of exchange between different citrate states in the nanoparticle sample (Figure 3.9). It was expected that an increase in temperature would cause an increase in exchange between states, which would result in exchange broadening at higher temperatures. The most telling example is provided by the peaks circled in red (region i), where at 0 ºC two doublets are present, at 25 ºC loss of double structure occurs and at 45 ºC only one broad peak is observed. Although it appears that exchange is occurring between the two doublets in region i, this is not true as the chemical shift difference between these doublets is constant (~ ppm) as temperature increases. Instead, exchange seems to occur with the major peaks (possibly the free citrate state) since the chemical shift difference between them decreases by about 0.03 ppm. The blue and green circled areas, regions ii and iii respectively, also display loss of peaks. It should be noted that in the absence of exchange a sharpening of peaks should be observed for free 54

65 citrate (four major peaks) as temperature is increased. This results because, for a small molecule (i.e. fast motion regime), increasing temperature causes a decrease in the rotational correlation time, increasing the T 2 relaxation time, which causes a narrowing of peaks. Consequently, loss of fine structure in region ii is possibly due to exchange with bound states. Temperature hysteresis was briefly examined in the same variable temperature series as above (Figure 3.9). Although, the same general features were present in the spectra, changes were observed especially in region i at 0 ºC, where one of the two doublets was either absent or present in much lower intensity. Thus, either the system experiences slight hysteresis effects or the sample was not given sufficient time (20 min.) to equilibrate. 55

66 Figure 3.9. Temperature dependence, between 0 45 ºC, of the citrate peaks in citratecoated YF 3 NPs. Other spectra, 25 ºC and 0 ºC, show the effect due to temperature hysteresis. The peak present at 3.2 ppm in the 0 ºC spectrum is due to residual ethanol from the precipitation step. 56

67 3.6.2 Correlation Spectroscopy (COSY) Several COSY spectra were acquired to complement the information provided by the one dimensional spectra. A COSY spectrum of Cit-YF 3 NPs was used to elucidate the observed complexity in the proton 1D spectrum. Control spectra with low (1:10) and intermediate (1:1) yttrium:citrate (Y:Cit) ratios were acquired to eliminate the possibility of citrate binding to free lanthanides. Unfortunately, high Y:Cit ratios of 4:1 or even 2:1 could not be collected due to lanthanide hydroxide precipitation at basic ph (~ 8.5). The COSY spectrum of the Cit-YF 3 NPs provided a wealth of information about the relationships of peaks in the one dimensional spectrum. At least ten different methylene states were observed with even the slightest peaks in the one dimensional spectrum accounted for (Figure 3.11, chemical shift assignments in the Supporting Data Table SD1). These peaks likely indicate numerous citrate binding geometries to the nanoparticle surface, and may resemble some of the binding modes present in lanthanide coordination polymers (Figure 3.10). The citrate has seven sites capable of coordinating to the metal: one from the α-hydroxyl, two from the α-carboxylate and four from the two β-carboxylates (Figure 3.1). In the lanthanide coordination polymers, the greatest number of citrate functional groups that coordinated to the lanthanide was three. These were the α-hydroxyl, α-carboxylate and one β-carboxylate, where each donated one oxygen to a particular metal site. 115 Figure Different carboxylate coordination modes to positively charged metal centers. (Reprinted from J. Mol. Struct. 2008, 877, Copyright (2008), with permission from Elsevier). The COSY spectrum of the intermediate Y:Cit (1:1) (Figure 3.12) and the low Y:Cit (1:10) (Figure 3.13) ratios displayed similarity in peaks with that of the nanoparticle sample. However, the presence and intensity of the peaks differed between 57

68 the 1:1 control and nanoparticle spectra, indicating that not all the bound states present in the nanoparticle are present in this sample, and if present, they can be in different amounts (Table 3.3). On the other hand, the citrate states present in the 1:10 control bore a remarkable similarity to those of the nanoparticle, although the intensity of the peaks are different as observed in the one dimensional spectra. In summary, the COSY control results do not conclusively confirm that the citrate peaks in the Cit-YF 3 NPs are due solely to binding to the nanoparticle, although the widely different one dimensional spectra may indicate that some states are more favoured due to constraints imposed by nanoparticle surface structure. 58

69 Figure One dimensional and COSY spectra of the Cit-YF 3 nanoparticles (0 ºC, ph ~ 8.5). The different states are labelled in capital letters, where each letter designates a particular state across all the COSY spectra. 59

70 Figure One dimensional and COSY spectra of the 1:1, Y:Cit control (0 ºC, ph ~ 8.5). The different states are labelled in capital letters, where each letter designates a particular state across all the COSY spectra. 60