Plants ranging from the small weed Arabidopsis to the giant

Transcription

1 Multiple feedback loops through cytokinin signaling control stem cell number within the Arabidopsis shoot meristem Sean P. Gordon, Vijay S. Chickarmane, Carolyn Ohno 1, and Elliot M. Meyerowitz 2 Division of Biology , California Institute of Technology, 1200 East California Boulevard, Pasadena, CA Contributed by Elliot M. Meyerowitz, July 27, 2009 (sent for review June 3, 2009) A central unanswered question in stem cell biology, both in plants and in animals, is how the spatial organization of stem cell niches are maintained as cells move through them. We address this question for the shoot apical meristem (SAM) which harbors pluripotent stem cells responsible for growth of above-ground tissues in flowering plants. We find that localized perception of the plant hormone cytokinin establishes a spatial domain in which cell fate is respecified through induction of the master regulator WUSCHEL as cells are displaced during growth. Cytokinin-induced WUSCHEL expression occurs through both CLAVATA-dependent and CLAVATA-independent pathways. Computational analysis shows that feedback between cytokinin response and genetic regulators predicts their relative patterning, which we confirm experimentally. Our results also may explain how increasing cytokinin concentration leads to the first steps in reestablishing the shoot stem cell niche in vitro. clavata wuschel computational modeling Plants ranging from the small weed Arabidopsis to the giant sequoia tree, maintain growth of stems, leaves, flowers, and branches through the action of stem cells. In the model plant Arabidopsis, as in other flowering plants, stem cells which give rise to above-ground tissues reside in a structure termed the shoot apical meristem (SAM) (1, 2). The Arabidopsis SAM is composed of three functionally distinct zones. The central zone (CZ) at the tip of the SAM harbors pluripotent stem cells which are necessary for the indeterminate growth and development of the plant. As the plant grows, CZ cells become either multipotent peripheral zone (PZ) cells on the sides of the meristem, capable of differentiating to leaf and flower primordia, or multipotent rib meristem (RM) cells beneath, which can differentiate to the cell types of the stem (3). Positions of zones within the meristem are maintained even as individual cells are displaced from the CZ through the PZ and RM into differentiating tissues. Molecular mechanisms by which meristematic zones are maintained as cells comprising these domains change remains a fundamental question in plant biology (1, 4). One mechanism involves the transmembrane receptor kinase CLAVATA1 (CLV1), expressed in cells of the RM (5). Its ligand, the extracellular peptide product of the CLAVATA3 (CLV3) gene, is produced in the CZ (6), and when it signals the RM cells, they reduce the activity of the WUSCHEL (WUS) gene, which codes for a homeodomain transcription factor also expressed in the RM (7, 8). WUS activity is nonautonomously necessary for the maintenance of the CZ cells as pluripotent stem cells, and therefore for persistence of the SAM (9). Loss of CLV3 activity causes enlargement of the CZ by conversion of PZ cells on the PZ-CZ border to CZ cells within hours, followed by enlargement of the SAM through increased cell division or reduced differentiation, or both, over days (10). Multiple lines of evidence show that the plant hormone cytokinin is involved in the CLV/WUS circuit, as well as SAM formation, maintenance and growth (1). Cytokinins stimulate the formation of new shoot apical meristems in culture (11). Cytokinin application rescues the SHOOTMERISTEMLESS (STM) mutant, which lacks the ability to maintain the SAM (12), and STM induces cytokinin biosynthetic genes (13, 14). Mutants for the LOG gene of rice, which encodes an enzyme that catalyzes the production of active cytokinins in the apical stem cell region of the SAM, have reduced shoot meristem size and prematurely terminate floral meristems (15, 16). Cytokinins act via receptors of the histidine kinase class (AHK2, 3, and 4), which when activated transfer phosphoryl groups to histidine phosphotransfer proteins (HPTs) and thence to two classes of Arabidopsis response regulators (ARRs) (17, 18). The Type-B ARRs activate transcription of cytokinin-induced target genes; Type-A ARRs negatively regulate cytokinin signaling (18 20). WUS has recently been shown to repress the genes for Type-A ARRs, thus likely increasing cytokinin signaling (21). Furthermore, overexpression of a Type-A ARR reduces WUS RNA levels, and can mimic the wus mutant phenotype (21) [SAM termination (22)]. Cytokinin treatment induces CLV loss-of-function phenotypes and causes increased WUS and decreased CLV1 expression (23, 24). In this study, we reveal multiple feedback loops between cytokinin response and WUS which influences gene expression and patterning within the Arabidopsis SAM. We use live imaging and an array of reporters to show that cytokinin perception and response is localized within the SAM where it regulates the pattern of WUS, a key positive genetic regulator of stem cell fate. We demonstrate that cytokinin signaling activates WUS expression through both CLV-dependent and CLV-independent pathways. We develop a computational model of cytokinin signaling which shows that feedback between cytokinin response and key genetic regulators determines the probability that a cell will express WUS. Given that WUS-expressing cells promote stem cell number and recent evidence that stem cells are a source of active cytokinins (16), our results may support a positive feedback loop between stem cells and underlying RM cells that maintains the organization of the SAM as stem cells are displaced during growth. Results CLV-Dependent and CLV-Independent Regulation of WUS by Cytokinin Signaling. Prior studies have shown that WUS expression within the SAM is partly restricted spatially through negative feedback from the CLV pathway shown in Fig. 1A. Recent studies have shown that treatment of plants with high levels of cytokinin leads Author contributions: S.P.G. and V.S.C. designed research; S.P.G. performed research; C.O. contributed new reagents/analytic tools; V.S.C. performed the computational analysis and model building; S.P.G., V.S.C., and E.M.M. analyzed data; and S.P.G., V.C., and E.M.M. wrote the paper. The authors declare no conflict of interest. See Commentary on page Present address: Developmental Biology Unit, EMBL, Meyerhofstrasse 1, Heidelberg, Germany. 2 To whom correspondence should be addressed. meyerow@its.caltech.edu. This article contains supporting information online at /DCSupplemental. PLANT BIOLOGY SEE COMMENTARY cgi doi pnas PNAS September 22, 2009 vol. 106 no

2 A P B WUS E G CLV1/3 stem cells WUS? T4 T4 CK C CLV1/3 P + cycloheximide - mock cycloheximide F H D T24 Fig. 1. CLV-independent regulation of WUS by cytokinin. (A) WUS/CLV interactions in a cross section of the SAM. WUS expression in RM cells promotes stem cell fate in overlying cells. Stem cells in turn secrete diffusible CLV3 ligand that binds to its receptor CLV1 in the RM leading to WUS downregulation. P labels organ primordia. (B) Hypothetical circuit in which cytokinin (CK) treatment leads to higher levels of WUS through suppression of CLV1.(Cand D) CLV1 (C), or ARR5 (D) transcript after 24 h of mock or cytokinin treatment. (E and F) relative WUS transcript in wild-type, clv1 11, and clv3 2 seedlings after mock or cytokinin treatment for (E)4h,or(F)24h(P 0.05). (G) Cytokinin induction of WUS for4h in absence ( ) or presence ( ) of 30 min cycloheximide (10 M) pretreatment. (H) Enhancement of carpel number in cytokinin treated clv1 and clv3 2 mutants compared to wild type (two-way ANOVA, F 81, P ). qrt-pcr error bars indicate SEM from three biological replicates. to CLV loss-of-function phenotypes and causes increased WUS and decreased CLV1 expression (23, 24). These data lead to a qualitative model in which cytokinin treatment increases WUS expression through suppression of CLV-mediated negative feedback on WUS levels (Fig. 1B). To test this starting hypothesis, we quantified the effect of cytokinin treatment on CLV1 and ARR5 transcription, as measured by quantitative reverse transcriptase PCR (qrt-pcr). As previously reported (20, 23), 24 h of cytokinin treatment reduced CLV1 RNA levels and increased RNA for the Type-A ARR, ARR5 (Fig. 1 C and D). To test whether repression of the CLV pathway is the only mechanism of WUS induction by cytokinin (23), we performed cytokinin treatments in a clv1 11 loss-of-function mutant. At 4 and 24 h after cytokinin treatment, WUS RNA increased in both wildtype and mutant lines compared to mock-treated samples (Fig. 1E and F); by 24 h WUS transcript was increased approximately 40-fold in both genotypes. Pretreatment of the plants with the protein synthesis inhibitor cycloheximide did not prevent this induction (Fig. 1G), suggesting a direct effect. Cytokinin treatment also induced WUS transcript accumulation in a clv3 2 loss-of-function mutant background, suggesting that induction in the clv1 11 mutant is not due to redundant function of related CLV3-dependent kinases active in the SAM, such as BAM1, 2 and 3 (25), CLV2 (26), or CORYNE (27). We observed greater phenotypic enhancement of floral organ number (an indicator of increased floral meristem size and stem cell activity) by cytokinin in clv1 and clv3 mutants compared with wild type, suggesting a synergistic interaction between cytokinin and CLV loss of function (Fig. 1H, two-way ANOVA, F 81, P ). In contrast to wild type, cytokinin (benzylaminopurine, BAP) treatment of clv mutants resulted in massive enlargement of the SAM and floral meristems (Fig. S1). Similar fold induction of WUS transcript in wild type and clv mutants after continuous cytokinin treatment reveals the existence of CLV-independent mechanisms of cytokinin-induced WUS expression (Fig. 1 E G). However, greater phenotypic enhancement in CLV loss-offunction background indicates that the CLV pathway limits the effect of transient perturbations in cytokinin signaling and therefore indicates that there are also CLV-dependent effects (Fig. 1H; see Computational Modeling in SI Appendix). Feedback Between Cytokinin Signaling and the WUS/CLV Circuit Influences Patterning of Gene Expression. After 24 h, CLV1 and ARR5 transcript levels were altered at low cytokinin concentrations. However, increase in WUS transcript occurred only at high concentration (Fig. 2A). Finer dilutions showed a steep rise in WUS transcript and corresponding decrease in ARR5 beginning at 400 M and peaking near 600 M (Fig. S1). We reasoned that the observed increase of WUS transcript after cytokinin perturbation could be indicative of a role for endogenous cytokinin response in influencing the pattern of WUS expression. Cytokinin response can be visualized at high resolution using a synthetic reporter, ptcs::gfp (28), which reports downstream activation of the cytokinin signaling pathway. Our data showing the sharper and higher threshold of cytokinin required for WUS induction as compared to ARR5 (Fig. 2A) indicates that WUS should closely overlap spatially with high levels of cytokinin response and drop off sharply in cells less responsive to cytokinin within the SAM. Consistent with this hypothesis, the ptcs::gfp reporter for cytokinin response was activated in a similar domain to WUS (Fig. 2 B D). ptcs::gfp expression mirrored temporal dynamics of WUS reporter expression during floral meristem development and SAM regeneration in culture (Fig. S2), consistent with a model where WUS is spatially regulated by cytokinin signaling during development (29). Previous studies have shown that WUS directly suppresses the transcription of a subset of Type-A ARRs involved in negative feedback on the cytokinin signaling pathway. Thus WUS likely increases cytokinin signaling (21). From the perspective of a gene regulatory network, regulation of WUS levels by cytokinin signaling, either by CLV-dependent or independent pathways, completes a positive feedback loop between cytokinin signaling and WUS. Cytokinin-induced increase in WUS levels leads to greater suppression of Type-A ARRs which leads higher cytokinin signaling and thus higher WUS levels. To better understand the nature of these potential positive feedback loops we used computational modeling (see Computational Modeling in SI Appendix for details) to plot predicted steady state values of WUS as a function of cytokinin signaling (Fig. 2 F J). We first considered the three hypothetical networks shown in Fig. 2E.In the circuit displayed in Fig. 2E (1), cytokinin signaling regulates WUS through suppression of CLV1 alone. Alternatively, in the cgi doi pnas Gordon et al.

3 E ptcs F A B C D Cytokinin signaling (1) ARR5 WUS (3) CLV1 CLV3 (2) (1) H phosphorylated Type-B ARR I Cytokinin signaling (2) ARR5 G WUS WUS CLV1 (3*) CLV3 network displayed in Fig. 2E (2) WUS transcription is activated through a CLV-independent mechanism. In contrast, Fig. 2E (3) shows a network in which WUS is regulated by both CLVdependent and independent mechanisms, as suggested from our experiments. Plots in Fig. 2F show that increases in WUS in models considering CLV-dependent or CLV-independent regulation of WUS alone are limited. In contrast, when cytokinin signaling activates WUS through both mechanisms a massive increase in WUS occurs. In network Fig. 2E (1), WUS increase is bounded as the maximum level of WUS which can be achieved is equivalent to the clv1 mutant, a roughly 3 4-fold increase in WUS compared to wild type as shown in Fig. 1E (mock treated L-er versus mock treated clv1 11). In network (2) WUS increase is bounded by the presence of negative feedback from the CLV pathway. Higher WUS levels leads to higher levels of CLV3 that suppresses WUS transcription. In comparison, in network (3) (3) J Cytokinin signaling (3) ARR5 WUS CLV1 Type A ARR Fig. 2. Feedback between cytokinin signaling and the WUS/CLV circuit influences patterning of gene expression. (A) Relative CLV1, ARR5, and WUS RNA transcript at varying cytokinin concentrations. (B D) ptcs::gfp expression in the SAM (B), early flower bud (C), or cross section of SAM (D). (E) (1) Cytokinin activates WUS through suppression of CLV1 or (2) a CLV-independent pathway or (3) through both mechanisms. (F) Predicted steady state WUS levels at varying levels of cytokinin signaling for circuits (1, green line), (2, red line), and (3, blue line). (G) Steady state WUS levels for network (3) including CLV negative feedback compared to network 3 lacking the CLV pathway (3*). (H J) spatial distribution of phosphorylated B-type ARR (H), WUS (I), or Type-A ARR (J) for network 2E (3). Axis of the plots correspond to a section of the meristem using arbitrary units in which 0,0 marks the center of the meristem. Error bars indicate SEM from two biological replicates. (Scale bars, 20 mm.) CLV3 WUS can be induced an order of magnitude greater than in the first two cases, similar to the experimentally observed an approximate 40-fold increase. Suppression of CLV1 transcription by cytokinin signaling allows CLV-independent induction of WUS to occur with less suppression from the CLV pathway. The effect of functional CLV negative feedback is shown in Fig. 2G. Plots in Fig. 2G show that in the absence of CLV negative feedback (3*), WUS is induced at lower levels of signaling than in circuits in which the CLV pathway is present (3). Negative feedback on cytokinin signaling through Type-A ARRs also contributes to the high threshold required for WUS induction. We used our computational model (see Computational Modeling in SI Appendix) to predict the pattern of components within the circuit displayed in Fig. 2E (3), given our data showing a central peak of cytokinin signaling within the SAM (Fig. 2 B D). Plots of predicted steady state values of activated B-Type ARR, WUS, and the Type-A ARR, ARR5 (known to be suppressed by WUS) are show in Fig. 2 H J. These plots show that WUS is predicted to closely overlap with cytokinin signaling as we observe experimentally. In contrast, ARR5 is predicted to be suppressed where cytokinin signaling is highest and expressed strongly in a peripheral ring-shaped domain. Distribution of Cytokinin Receptor, Cytokinin Response, and WUS Correlate in Individual Cells Where ARR5 Is Suppressed. To test the predictions of our model, we experimentally determined the relative spatial expression of WUS and the Type-A ARR, ARR5. We observed that a transcriptional reporter for ARR5 was suppressed in the WUS domain but expressed strongly in adjacent cells forming a ring-like expression pattern (Fig. 3 A C), consistent with the predictions of our computational model. Activation of WUS expression by cytokinin perturbation and overlap of WUS expression with the reporter of downstream cytokinin response, ptcs::gfp, suggested that endogenous cytokinin response might act as a positional cue for patterning WUS transcription. Upstream of WUS function, cytokinin response is governed by cytokinin receptor availability and the local concentration of cytokinin. Therefore, localized cytokinin response within the center of the SAM could be indicative of either a higher local concentration of cytokinin or increased perception of cytokinin in these cells through localized receptor expression. To investigate the latter possibility we determined the distribution of cytokinin receptor expression within the SAM. Indeed, fluorescent reporters for the cytokinin receptor AHK4 (30), and WUS transcription were expressed in overlapping domains within the SAM and were correlated in individual cells (Fig. 3 D H). AHK4 and WUS reporters were similarly regulated during floral meristem development, expanded similarly in the clv3 2 mutant and were both altered in super-enlarged cytokinin-treated clv3 2 SAMs (Fig. 3 I L and Fig. S1). AHK4 and WUS reporters also overlapped during SAM regeneration in culture. AHK4 reporter was induced in cultured cells during pretreatment on auxin-rich medium known to promote regeneration (Fig. S2). Transfer to cytokinin-rich medium results in WUS induction (29) in cells marked by the AHK4 reporter in developing SAMs (Fig. S2). Cytokinin Regulates Domain of Cytokinin Signaling Output, WUS and CLV3 Expression. The above results suggested that cytokinin receptor distribution initiates a gradient of cytokinin signaling peaking within the center of the SAM and which patterns WUS expression in multiple contexts. This model predicts that treatment of plants with exogenous cytokinin would extend sufficient signaling to cells farther from the center of the SAM which have lower levels of receptor, causing WUS activation in an expanded domain. Indeed, live imaging before and after 12 h of cytokinin treatment showed expansion of the WUS reporter expression domain (Fig. 4 A D and Fig. S3). We observed respecification of PLANT BIOLOGY SEE COMMENTARY Gordon et al. PNAS September 22, 2009 vol. 106 no

4 B C D E F G H I WUS AHK4 WUS AHK4 WUS ARR5 WUS A AHK4 AHK4 WUS Frequency J K L i Fig. 3. AHK4 and WUS expression correlate in individual cells where ARR5 is suppressed. (A C) ARR5 (green) reporter down regulation within the WUS domain (red) and organ primordia (AHP6 domain, Fig. S3). Inset in (C) plots ARR5 and WUS intensity (yellow line indicates profile). (D F) Cytokinin receptor (AHK4, green) and WUS reporter (red) overlap within the SAM (center) or floral meristems (peripheral). Cross sections displayed below. (G) AHK4 and WUS overlap in single cells. (H) pixel intensity of AHK4 (x axis) and WUS (y axis) reporters in wild-type flowers (correlation coefficient R 0.79, upward trend indicates positive correlation). (I) WUS (red) and AHK4 (green) in cytokinin treated clv3 2 SAM and floral meristems (arrows) compared to untreated clv3 2 mutants (J L). [Scale bars, 20 m except for 10 m in (G) and 100 m in (I).] cells that previously did not express WUS, indicating that WUS domain expansion was not solely due to increased cell division in the RM (Fig. 4 A D). Therefore, similarly to recruitment of surrounding cells into CZ cells after loss of CLV3 activity (10), cytokinin increase leads to recruitment of surrounding cells into WUS expressing cells. After 24 h of cytokinin treatment we observed expanded ptcs::gfp and pclv3::gfp-er reporter expression within inflorescence and floral meristems which was not observed in mock treated samples (Fig. 4 E H). Cytokinin treatment was also sufficient to induce ectopic WUS expression, but only in cells which express high levels of cytokinin receptor (Fig. S3) (30). Cytokinin-Induced Increase of WUS Transcript and Related Meristem Phenotypes Requires a CLV-Independent Pathway Through an AHK2/ AHK4-Dependent Mechanism. Our results suggest that cytokinin receptor distribution controls the distribution of cytokinin response and thereby influences the pattern of WUS expression within the SAM. To determine whether the induction of WUS by cytokinin perturbation requires functional cytokinin receptors we quantified WUS levels in cytokinin receptor loss-of-function backgrounds (Fig. 5A). Of the three characterized cytokinin 兩 Fig. 4. Cytokinin regulates domain of cytokinin signaling output, WUS and CLV3 expression. (A D) live imaging of WUS reporter (green) before (A and B) and after 12 h of cytokinin treatment (1mM BAP) (C and D). Numbering in (A D) registers cells in (A and B) to the same cells in (C and D) after 12 h. Arrow marks floral meristem. (E) CLV3 reporter (green) in plants after 24 h of mock treatment (n 5) or (F) cytokinin treatment (n 5). (G) ptcs::gfp reporter (green) after 24 h of mock treatment (n 5) as compared to (H) 24 h of cytokinin treatment (n 5). Membranes are marked with FM4 64 dye (A D and G and H) or 29 1 membrane YFP marker (10) (E and F). (Scale bars, 50 m.) receptors only AHK2 and AHK4 mutants showed significantly lower relative WUS transcript levels after cytokinin treatment as compared to wild type (one-way Anova, F 42, P 0.05, Fig. Gordon et al.

5 A C B 5A). In contrast, relative WUS transcript was not significantly different between cytokinin treated AHK3 mutant and wild-type samples. Consistent with this observation, clv mutant-like phenotypes associated with WUS misregulation after cytokinin treatment were not observed in the ahk2 2 mutant but were observed in the ahk3 3 mutant similar to wild-type plants (Fig. 5C). In comparison to WUS, cytokinin-induced suppression of CLV1 transcript occurred at a similar magnitude in all backgrounds (Fig. 5B). These data indicate that the AHK2 and AHK4 receptors are required for cytokinin-induced up-regulation of WUS transcript levels and associated clv mutant-like phenotypes. In contrast, cytokinin-induced suppression of CLV1 transcript does not have specific requirements for individual receptors. The fact that CLV1 was suppressed in all backgrounds but WUS up-regulation and cytokinin-induced clv mutant-like phenotypes were blocked in the AHK2 mutant (ahk2 2), suggests that induction of clv mutant-like phenotypes by cytokinin treatment requires a CLV-independent pathway of WUS induction by cytokinin. D CKs stem cells WUS CLV1/3 Fig. 5. Cytokinin regulates WUS expression through an AHK2/AHK4 dependent mechanism while CLV1 suppression has no requirement for individual receptors. (A) Relative WUS or (B) CLV1 transcript levels in wild-type and individual cytokinin receptor mutants after 24 h of mock treatment or cytokinin treatment. (C) Cytokinin-induced clv mutant-like carpel number phenotypes in wild type (COL), ahk2 2, and ahk3 3 mutants. (D) Hypothetical positive feedback between apical stem cells and RM cells. Apical stem cells produce active cytokinins (CKs) perceived by RM cells expressing sufficient cytokinin receptor to activate WUS expression. WUS, in turn, promotes stem cell fate in apical cells. Negative feedback from the CLV pathway is also shown. Discussion We propose that within the shoot meristem a standing gradient of cytokinin response, dictated in part by cytokinin receptor distribution, acts as spatial reference to inform cells of their position. As cells move into the RM, high cytokinin signaling triggers cell respecification through induction of WUS. Given recent evidence for localized production of active cytokinins in shoot stem cells (15), our results support a feedback principle for maintenance of stem cell niche organization during growth (Fig. 5D). Stem cells specify RM cell fate through production of active cytokinins which are locally perceived by underlying cells leading to induction of the master regulator WUS. In turn, WUSexpressing cells in the RM promote stem cell fate in overlying cells. Such positive reinforcement between these two domains could maintain their juxtaposition as the apical stem cells are displaced during post embryonic growth. During in vitro reestablishment of the shoot stem cell niche in tissue culture, high cytokinin signaling triggers induction of ectopic WUS expression leading to stem cell fate in surrounding cells (29). Ectopic WUS expression is sufficient for induction of shoot tissues and WUS is functionally required for de novo formation of the SAM in vitro (29, 31, 32). Therefore induction of WUS through cytokinin treatment may be a key link in triggering the formation of shoot tissues in culture. Cytokinin is sufficient to induce WUS expression in the stele of root explants where AHK4 is expressed (Fig. S3). Auxin pretreatment of tissue explants is used to enhance the efficiency of regeneration in culture (29). Our results show that auxin treatment leads to callus formation associated with broad up-regulation of the AHK4 receptor (Fig. S2). Thus, it is possible that the ability of auxin pretreatment to enhance regeneration in culture is mediated through the up-regulation of cytokinin receptor expression. This enables a larger population of cells to be competent to respond to cytokinin and trigger high cytokinin signaling required for up-regulation of WUS when explants are subsequently induced with cytokinin. The ability of cytokinin signaling to alter cell fate through induction of WUS expression is a common thread that links in vitro regeneration of shoot tissues and normal shoot development. Cytokinin receptors appear to be redundant in many contexts (18). One known example of cytokinin receptor specificity is in the control of leaf senescence in Arabidopsis that specifically requires the AHK3 receptor (33). We demonstrate that cytokinin-induced up-regulation of WUS transcript is mediated primarily through AHK2 and AHK4 dependent pathways and does not require the AHK3 receptor. Furthermore, cytokinin-induced clv mutant-like phenotypes associated with WUS misregulation are not observed in the ahk2 2 mutant. Published microarray data suggests that expression of the cytokinin receptors is not involved in a positive feedback loop with WUS, as transient overexpression of WUS leads to reduction of AHK4 levels and does not significantly alter AHK2 levels (21). Previous computational models have addressed how WUS expression is confined to a small number of cells within the SAM (34), and recently how the CLV and WUS cell populations maintain each other through feedback between WUS and the CLV pathway (35). In the first study, several alternative hypotheses to maintain WUS spatial pattern were considered. A model that assumed a localized activation of WUS was best able to reproduce its experimentally observed pattern. In this study we propose that cytokinin is a potential candidate for an activator similar to the hypothetical activator described in the above study, which is locally perceived within the SAM and thus influences the WUS expression pattern. Our study represents an attempt to computationally model the cytokinin signaling pathway. We then integrate this model with components of the WUS/CLV feedback system. Hence, our model is similar to previous studies (34, 35) in how components of the WUS/CLV pathway interact to spatially maintain their regions within the SAM. However, unlike previous studies, the model reported here provides an understanding of hormonal feedback on gene expression, through a detailed study of the cytokinin perception network. This study (see Computational Modeling in SI Appendix) reveals the functionality of several nested feedback loops which suggest a threshold-dependent activation of WUS, as a function of cytokinin. The threshold for activation occurs because sufficient cytokinin must build up before: (i) negative feedback of the CLV pathway is suppressed; (ii) sufficient WUS is accumulated to repress Type-A ARRs and promote further increases in WUS transcription. The positive feedback inherent in this circuit has PLANT BIOLOGY SEE COMMENTARY Gordon et al. PNAS September 22, 2009 vol. 106 no

6 the consequence of turning on WUS robustly as sufficient signaling is achieved. Additional exploration of the model suggests this positive feedback biases WUS to be expressed in a bistable switch-like manner (see Computational Modeling in SI Appendix). Experimentally, WUS is expressed in a switch-like mode, expressed strongly in some cells while absent from directly adjacent cells. Switch behavior is important as it gives robust output even if input, such as cytokinin signaling, fluctuates. In this study, relatively high levels of cytokinin were required to perturb the cytokinin signaling pathway and extend both the domain of cytokinin response and WUS expression. The requirement for high levels of cytokinin to disrupt the spatial distribution of cytokinin response is not surprising given that a multitude factors (Type-A ARRs, AHP6) act in a developmental context to suppress unrestrained signaling and maintain a stereotyped developmental pattern of response. Even at the relatively high levels of cytokinin used in this study, the spatial domain of cytokinin response changed only slightly when compared to its unperturbed pattern, as monitored by the ptcs::gfp reporter. The relatively sharp expression profile of the cytokinin receptors may also explain why strong cytokinin perturbations were required to extend the domain of cytokinin response and WUS expression. Cytokinin response within a given cell is the output of receptor concentration and cytokinin. Therefore cells with only low levels of receptor, such as cells farther away from the center of the SAM, require high levels of cytokinin to have significant cytokinin response. This study shows that cytokinin response regulates WUS expression through both CLV-dependent and CLV-independent mechanisms. Given that WUS promotes cytokinin response (21), our experimental and computational modeling data suggests that WUS and cytokinin signaling interact through multiple positive-feedback loops which ultimately control stem cell number in the SAM. Future studies will show whether cytokinin acts as signaling cue to relay information between cells in different domains of the SAM as cells comprising different zones change during growth of the plant. Materials and Methods Plant Materials and Reporter Constructs. clv1 11, clv3 2, and clv2 1 alleles in L-er background have been previously described (36, 37). The pwol::gfp line in Columbia (Col-0) background has been previously described (38), and it recapitulates expression patterns observed in the shoot and root via in situ hybridization (30). The parr5::gfp line in WS ecotype has been previously described (13). The pwus::gfp-er and pclv3::gfp-er lines have also been previously described (10, 34) (for details of other lines see SI Materials and Methods). Plant Growth and Cytokinin Treatment Conditions. Plants were grown as previously described (29). Cytokinin treatments with N6-benzylaminopurine (BAP; Sigma Aldrich Co.) were performed as described (23) except that shoots were sprayed with the respective solutions (for details see SI Materials and Methods). Quantitative Real-Time PCR (qrt-pcr). Quantitative real-time PCR (qrt-pcr) was performed with Roche Universal Probe Library hydrolysis probes. Each sample represents tissue harvested from 50 two-week-old seedlings just transitioned to flowering (for details see SI Materials and Methods). ACKNOWLEDGMENTS. We thank P. Benfey for the pwoodenleg::gfp line referred to here as the AHK4 reporter, pahk4::gfp. The parr5::gfp line was provided by J. Kieber and Y. Helariutta provided the pahp6::ahp6-gfp line. Cytokinin receptor mutants were provided by T. Kakimoto. We thank the Computable Plant group ( including H. Jönsson, M. Heisler, A. Roeder, and P. Tarr for comments on the manuscript; K. Sugimoto for advice with RT-PCR analysis of treated root explants; A. Martin for sample collection of cytokinin serial dilutions; and A. Garda for technical support. This study was supported by National Science Foundation grants IOS and IOS to E.M.M. 1. Sablowski R (2007) The dynamic plant stem cell niches. Curr Opin Plant Biol Dec;10: Fletcher JC, Meyerowitz EM (2000) Cell signaling within the shoot meristem. Curr Opin Plant Biol 3: Clark SE (2001) Cell signaling at the shoot meristem. Nat Rev Mol Cell Biol 2: Haecker A, Laux T (2001) Cell-cell signaling in the shoot meristem. Curr Opin Plant Biol 4: Clark SE, Williams RW, Meyerowitz EM (1997) The CLAVATA1 gene encodes a putative receptor kinase that controls shoot and floral meristem size in Arabidopsis. Cell 89: Fletcher JC, Brand U, Running MP, Simon R, Meyerowitz EM (1999) Signaling of cell fate decisions by CLAVATA3 in Arabidopsis shoot meristems. Science 283: Brand U, Fletcher JC, Hobe M, Meyerowitz EM, Simon R (2000) Dependence of stem cell fate in Arabidopsis on a feedback loop regulated by CLV3 activity. Science 289: Mayer KF, Schoof H, Haecker A, Lenhard M, Jurgens G, Laux T (1998) Role of WUSCHEL in regulating stem cell fate in the Arabidopsis shoot meristem. Cell 95: Schoof H, Lenhard M, Haecker A, Mayer KF, Jurgens G, Laux T (2000) The stem cell population of Arabidopsis shoot meristems in maintained by a regulatory loop between the CLAVATA and WUSCHEL genes. Cell 100: Reddy GV, Meyerowitz EM (2005) Stem-cell homeostasis and growth dynamics can be uncoupled in the Arabidopsis shoot apex. Science 310: Skoog F, Miller CO (1957) Chemical regulation of growth and organ formation in plant tissues cultured in vitro. Symp Soc Exp Biol 54: Long JA, Moan EI, Medford JI, Barton MK (1996) A member of the KNOTTED class of homeodomain proteins encoded by the STM gene of Arabidopsis. Nature 379: Yanai O, et al. (2005) Arabidopsis KNOXI proteins activate cytokinin biosynthesis. Curr Biol 15: Jasinski S, et al. (2005) KNOX action in Arabidopsis is mediated by coordinate regulation of cytokinin and gibberellin activities. Curr Biol 15: Tucker MR, Laux T (2007) Connecting the paths in plant stem cell regulation. Trends Cell Biol 17: Kurakawa T, et al. (2007) Direct control of shoot meristem activity by a cytokininactivating enzyme. Nature 445: Kakimoto T (2003) Perception and signal transduction of cytokinins. Annu Rev Plant Biol 54: Muller B, Sheen J (2007) Advances in cytokinin signaling. Science 318: To JP, et al. (2004) Type-A Arabidopsis response regulators are partially redundant negative regulators of cytokinin signaling. Plant Cell 16: To JP, Kieber JJ (2008) Cytokinin signaling: Two-components and more. Trends Plants Sci 13: Leibfried A, et al. (2005) WUSCHEL controls meristem function by direct regulation of cytokinin-inducible response regulators. Nature 438: Laux T, Mayer KF, Berger J, Jurgens G (1996) The WUSCHEL gene is required for shoot and floral meristem integrity in Arabidopsis. Development 122: Lindsay DL, K. Vipen V, Bonham-Smith PC (2006) Cytokinin-induced changes in CLAVATA1 and WUSCHEL expression temporally coincide with altered floral development in Arabidopsis. Plant Science 170: Clark SE, Running MP, Meyerowitz EM (1993) CLAVATA1, a regulator of meristem and flower development in Arabidopsis. Development 119: DeYoung BJ, Bickle KL, Schrage KJ, Muskett P, Patel K, Clark SE (2006) The CLAVATA1- related BAM1, BAM2 and BAM3 receptor kinase-like proteins are required for meristem function in Arabidopsis. Plant J 45: Kayes JM, Clark SE (1998) CLAVATA2, a regulator of meristem and organ development in Arabidopsis. Development 125: Muller R, Bleckmann A, Simon R (2008) The receptor kinase CORYNE of Arabidopsis transmits the stem cell-limiting signal CLAVATA3 independently of CLAVATA1. Plant Cell 20: Muller B, Sheen J (2008) Cytokinin and auxin interaction in root stem-cell specification during early embryogenesis. Nature 453: Gordon SP, Heisler MG, Reddy GV, Ohno C, Das P, Meyerowitz EM (2007) Pattern formation during de novo assembly of the Arabidopsis shoot meristem. Development 134: Mahonen AP, et al. (2006) Cytokinins regulate a bidirectional phosphorelay network in Arabidopsis. Curr Biol 16: Gallois JL, Woodward C, Reddy GV, Sablowski R (2002) Combined SHOOT MERISTEM- LESS and WUSCHEL trigger ectopic organogenesis in Arabidopsis. Development 129: Gallois JL, Nora FR, Mizukami Y, Sablowski R (2004) WUSCHEL induces shoot stem cell activity and developmental plasticity in the root meristem. Genes Dev 18: Kim HJ, et al. (2006) Cytokinin-mediated control of leaf longevity by AHK3 through phosphorylation of ARR2 in Arabidopsis. Proc Natl Acad Sci USA 103: Jonsson H, et al. (2005) Modeling the organization of the WUSCHEL expression domain in the shoot apical meristem. Bioinformatics 21:i Geier F, Lohmann JU, Gerstung M, Maier AT, Timmer J, Fleck C (2008) A quantitative and dynamic model for plant stem cell regulation. PLoS ONE 3:e Dievart A, et al. (2003) CLAVATA1 dominant-negative alleles reveal functional overlap between multiple receptor kinases that regulate meristem and organ development. Plant Cell 15: Clark SE, Running MP, Meyerowitz EM (1995) CLAVATA3 is a specific regulator of shoot and floral meristem development affecting the same processes as CLAVATA1. Development 121: Birnbaum K, et al. (2003) A gene expression map of the Arabidopsis root. Science 302: cgi doi pnas Gordon et al.

7 Supplement:Computational Modeling Sean Gordon, Vijay Chickarmane, Carolyn Ohno, and Elliot Meyerowitz INTRODUCTION In what follows we develop a mathematical model to describe the signaling and transcriptional network dynamics by which negative and positive feedback loops regulate cytokinin signaling and WUS expression. The main focus of the model is to explore several hypotheses, as to how WUS induction occurs as a function of cytokinin signaling and to understand how individual feedback loops contribute to the dynamics of the cytokinin perception pathway. As described in the main text, cytokinin perception in cells of the Arabidopsis thaliana meristem is mediated through a signaling system in which sensor histidine kinase receptors are activated by cytokinin and in response phosphorylate Arabidopsis histidine phosphotransfer proteins (AHPTs). Phosphorylated AHPTs transfer phosphorylation to downstream Arabidopsis Response Regulators (ARRs). (ARRs)[1]. Type-B ARRs which are phosphorylated become transcriptionally active. One direct target being the Type-A ARRs. The Type-A ARRs negatively regulate signaling [1, 2]which leads to a negative feedback signaling loop: Signaling Type-B ARR TypeA ARR Signaling. The transcription factor WUS, is known to suppress Type-A ARR transcription [3]. Hence these basic interactions summarize the signaling activity mediated through the ARR signaling pathway. WUS also participates in the well known CLV negative feedback loop regulating stem cell number. WUS activates CLV3 which when complexed with its receptor,clv1, negatively regulates WUS expression. Hence WUS regulates itself negatively through the CLV pathway. Recent experiments by Lindsay et al [4], show that cytokinin treatment of wild type plants leads to suppression of CLV1, and induction of WUS. The authors therefore suggest that cytokinin breaks the CLV-mediated negative feedback on WUS, and that this may explain WUS induction by cytokinin treatment. However, in the experiments described in the main text, WUS induction in clv mutant backgrounds suggests that there must be other mechanisms which operate to induce WUS. Hence we make, in addition to the ability of cytokinin to suppress CLV1, the hypothesis that WUS is induced by a CLV independent mechanism through the Type B ARR s as shown in Fig. 1. Network Model We make the following simplified assumptions, in building the mathematical model. Cytokinin signaling leads to down regulation of CLV1 [4]. 1

8 Cyt AHPT AHPT-P B A Bp HYP 1 C1 HYP 2 Wus C3 Figure 1: The cytokinin perception circuit. represent signaling as well transcriptional activation, represents negative feedback or repression (see text for type of interaction). Hyp 1: Cytokinin signaling represses CLV1, Hyp 2: Cytokinin signaling activates Wus(C1=CLV1, C3=CLV3. B P,AHPT P correspond to the phosphorylated forms of B,AHPT). Cyt corresponds to (cytokinin receptor). Cytokinin treated clv mutants show WUS induction, which suggests CLV-independent activation of WUS. Type-A ARRs mediate negative feedback on cytokinin signaling, either through competition with Type-B ARRs for phosphorylation from the AHPTs or through an independent mechanism involving protein-protein interactions [1, 2]. Type-A ARR negative regulation of signaling is modeled in a simple way by assuming that Type-A ARR s reduce signaling by a factor 1 (1) 1+α[A] (see Eq. 2, d[bp], where the strength of the negative feedback depends on the parameter dt α. Transcription factor dynamics is modeled based upon a thermodynamic model for transcription [5, 6, 7, 8], We also model transcription and translation as one combined process. The above can be translated into the following set of differential equations for the concentrations of the various species. d[ahp T P ] dt = ( [Cyt][AHP T ] K h1 +[AHP T ] ) v h1[ahp T P ] K h2 +[AHP T P ], 2

9 d[b P ] dt d[a] dt d[wus] dt d[clv 1] dt d[clv 3] dt d[clv ] dt = ( [AHP T P ][B] 1 )( K 1 +[B] 1+α[A] ) v 1[B P ] K 2 +[B P ], a 0 + a 1 [B P ] = 1+a 0 + a 1 [B P ]+a 2 [Wus] γ 1[A], n b 0 + b 1 [B P ] = 1+b 0 + b 1 [B P ]+b 2 [CLV ] γ 4[Wus] = = c 0 1+c 0 + c 1 [B P ] k 1[CLV 1][CLV 3] + k 2 [CLV ] γ 2 [CLV 1] d 1 [Wus] 1+d 1 [Wus] k 1[CLV 1][CLV 3] + k 2 [CLV ] γ 3 [CLV 3] = k 1 [CLV 1][CLV 3] k 2 [CLV ] (2) In the above equations we assume that cytokinin signaling input corresponding to (cytokinin receptor) phosphorylates AHP T. The AHP T P then causes phosphorylation of Type-B ARR s. We describe these by simple Michaelis-Menten equations. The parameter values used for the above equations are displayed in Table 1. We do not assume any type of cooperativity, except in a later section where we assume that WUS binds co-operatively to Type-A ARR as a repressor, where we assume n = 2. In the equations for AHP T, B P,wehavenotincluded production and degradation terms. Cytokinin does not alter the levels of AHPT s and Type B RR s, and in fact adding cyclohexamide does not prevent signaling(see Fig 1G main text). Hence we interpret the role of cytokinin as triggering the conversion of un-phosphorylated forms of AHPT and Type-B ARR to phosphorylated forms of the proteins. We have however tested the model by adding a low rate of production and degradation, to see if the results change significantly. We find that we get almost identical results. The reason is that by allowing a low rate of production+degradation, the total amount of phosphorylated+unphosphorylated proteins is conserved, hence this situation is almost identical to having no production/degradation of type B and AHPT s. We begin by analyzing the sub-system which consists of AHP T P, Type-A ARR, and Type-B ARR. We exclude all other interactions, since we are interested here in the dynamics of the core negative feedback of Type-A ARR on signaling. This amounts to assuming that the levels of cytokinin are below the threshold above which WUS is strongly induced. Hence, AHPTs Type-B ARR (phosphorylation), Type-A ARR signaling, and Type-B ARR Type-A ARR (transcription activation). In Fig 2 we plot Type-A ARR and Type-B ARR levels, for a perturbation in cytokinin concentration,(cytokinin concentration which starts at a high value(10), degrades in time) 1. This is plotted for the wild type, and for higher negative feedback, i.e α =1, 5, respectively. Cytokinin treatment first leads to a rise in activated (phosphorylated) Type-B ARR, which immediately increases transcription of Type-A ARR. The Type-A ARR, negatively feeds 1 We assume a degradation constant for cytokinin of 1min 1, which is higher than all the other degradation constants in the network. This implies that the system is first perturbed, by a pulse(cyt t=0 = 10), followed by the slow decay to steady state. 3

10 back on signaling, which in turn, restricts Type-B ARR activation. Hence as the feedback strength is increased (increasing α), we see from the plots that both Type-A ARR as well as Type-B ARR levels rise and fall over shorter time scales. The negative feedback therefore leads to a quick response of the system to perturbations. However, if the concentration of cytokinin is increased substantially, then WUS is significantly induced, leading to different dynamics. Hence the Type-A ARRs limit cytokinin signaling, filtering out small fluctuations which could initiate induction of a subset of downstream genes thereby leading to a different developmental program. From the figure, the relevant time scales are approximately within 1h Type A ARR WT WT stronger feedback Relative Levels Type B ARR Time Figure 2: Dynamics of A & B Type ARR s as a function of time (mins). Steady State Dynamics of WUS with respect to cytokinin concentration. We now discuss the steady state behavior of WUS as a function of cytokinin signaling for the various hypothesis. By this we mean that the input level of cytokinin is fixed, and the steady state values of the network components are computed. WUS induction in clv mutants point to a CLV-independent mechanism by which cytokinin activates WUS. We propose that WUS is induced through a CLV-independent mechanism, see Fig 1, lines marked as HYP2, which shows activation of WUS transcription through the core cytokinin pathway. Fig. 3 panel A displays the steady state values of WUS, Type-A ARR, Type-B ARR as a function of cytokinin signaling for the clv mutant, for HYP2(in the model we simply set b 2 =0,to represent the inability of the CLV complex to repress WUS), and in panel B, for the wild type (HYP1+HYP2). WUS induction is seen in both cases, with the difference being that the curve for WUS induction in (B) requires a higher threshold of cytokinin signaling to be fully induced. Referring to panel A, when induced by cytokinin signaling, Type-B ARRs transcribe Type-A ARRs, which in turn suppresses signaling through the core cytokinin pathway, and hence, this negative feedback keeps WUS induction at a low level. However, on further increase of cytokinin, a threshold of cytokinin signaling is reached such that 4

11 when sufficient WUS is produced, it suppresses Type-A ARRs (ARRs 5, 6, 7, 15), thereby relieving cytokinin signaling. This results in enhanced signaling and hence more WUS. Thus the two negative feedback regulatory interactions lead to positive feedback, which leads to an induction of WUS. In the wild type (panel B), we assume HYP1( [4]) and HYP2, the curve shows an even greater threshold in cytokinin required, before full WUS induction. Here, WUS effectively negatively regulates itself, since WUS activates CLV3, which when complexed with CLV1, leads to downregulation of WUS. Hence, sufficient cytokinin levels have to be build up to downregulate CLV1, so as to break the WUS self-negative feedback. Hence there are two interlooped mechanisms which induce WUS: the first is that increasing signaling leads to repression of Type-A ARR by WUS, and the second by repressing CLV1. The first of these occurs through WUS itself, which leads to positive feedback. The second mechanism allows for WUS levels to increase as cytokinin signaling increases. The curve (A) ARR5 B p WUS Concentration (B) Cytokinin Figure 3: The network models displayed as cartoons, corresponding to model(3), and CLV mutant. The curves for WUS, Type-A ARR, Type-B ARR as a function of cytokinin for clv mutant (panel A) and wild types (panel B). Panel A: b2 = 0,b1 = 0.25, Panel B: b2 =1,b1=0.25, all other parameters are as in Table1. for Type-A ARR predicts a remarkable behavior (panels B) for the wild type. Here, for low values of cytokinin before WUS is fully induced, Type-A ARR levels are higher. This is because WUS induction has to come through the signaling which is negatively regulated by the Type-A ARR s. As cytokinin signaling increases, WUS ultimately gets induced, and this leads to down-regulation of Type-A ARRs. Hence the curve for Type-A ARR shows a peak. Comparing the spatial profile(see later section), of Type-A, B ARR s, WUS, with dose response curves (Fig. 2(a) main text) confocal image data of Type-A ARR reporters (Fig. 3(a-c) main text), provides support for HYP2+HYP1 to be the mechanism by which WUS is induced by cytokinin. Throughout the rest of the text we will often refer to this as the wild type. A better understanding of the functional relevance of the various hypothesis to the dynamics of the network can be achieved by considering two cases:wus is assumed to be induced only by activation from the cytokinin signaling pathway alone (HYP2 in the absence of HYP1), or solely by suppression of CLV1 (HYP1 in the absence of HYP2). Fig. 4 5

12 shows simulated dose response curves for the two cases as well as compared to the combined case (HYP1+HYP2 wild type). The curves for WUS show that HYP1 alone or HYP2 alone lead to low levels of WUS induction. HYP1:Supression of CLV1 is insufficient to give major WUS Levels for HYP1, HYP2, HYP1+YP2 5 HYP2 HYP1 4.5 WT 4 Concentration Cytokinin Figure 4: The network models corresponding to models (1,2 & 3) displayed as cartoons. WUS as a function of cytokinin for HYP1, HYP2 and wild type. HYP1: b2 =1,b1= 0,c1 = 5, HYP2:b2 = 1,b1 = 0.25,c1 = 0, HYP1+HYP2:b2 = 1,b1 = 0.25,c1 = 5, all other parameters are as in Table1. increases in WUS levels, unless WUS is induced separately, HYP2:WUS activation, either through Type-A ARR regulated or an independent mechanism does not lead to massive induction of WUS, since the CLV-mediated negative feedback restricts WUS levels. This compared to the curve for the combined case (wild type), in which both mechanisms operate, shows that for robust WUS induction, both hypothesis must be considered. In summary, CLV negative feedback must be reduced, in addition to CLV independent activation to obtain the observed 38 fold induction of WUS(Fig 1 main text). Bistable Response of the Circuit Multiple core binding sites for WUS are present in an upstream region of the Type-A ARR, ARR7, to which WUS is known to bind, suggesting that WUS may bind to Type-A regulatory sequences in a cooperative manner [3]. The combined model including CLV1 suppression and simultaneous activation of WUS transcription through the core cytokinin pathway has 6

13 intrinsic positive feedback which would be enhanced through cooperative binding of WUS. Fig.5 shows WUS expression behaves in a switch-like manner, turning ON to high levels once cytokinin signaling levels cross a certain threshold. This network behavior which displays hysteresis, and hence memory, has two interesting consequences. First, once WUS expression turns ON, it is relatively well buffered by fluctuations in cytokinin. Second, it predicts a sharper spatial profile of WUS, as cells can only stably have either high or low levels of WUS. In situ hybridizations show that WUS is indeed expressed in a switch-like manner, strongly expressed in one cell while not present in a directly adjacent cell. Switch-like behavior of WUS activation is also consistent with its steep response curve to cytokinin (Fig. 2a main text) An interesting question is whether bistability in the network dynamics can result just 25 I+II I+II Coop 20 Relative WUS SN 5 SN Cytokinin signaling Figure 5: Bistability of the circuit due to co-operative binding of WUS on the Type-A ARR promoter (n = 2), all other parameters are as described in Table1). SN denotes saddle-node bifurcation, and the dotted lines represent unstable states. The system exhibits multiple steady states for a single value of cytokinin signaling. EX: at Cyt =1,WUS =22, 1, depending on the initial concentrations. due to the fact that there is already inherent positive feedback in the network(due to the two sequential negative feedbacks, namely, WUS A, A signaling, in addition to signaling WUS, both through Type B ARR, as well as by suppressing CLV1). Is cooperativity of WUS binding to Type-A ARR, necessary for bistability? In general it is difficult to show this, for the complex network described by Eq. 2. However, we will now demonstrate that bistability cannot arise simply from positive feedback, for a simplified network, in which we retain the important aspects of the network described by Eq. 2. Namely we consider only the dynamics between Type A ARR, Type B ARR and WUS. We therefore assume that activation of WUS which occurs both through suppression of CLV1 and activation through Type B ARR s can be replaced simply by activation by Type B ARR. We also ignore dynamics of the HPT s since it is upstream of the signaling. Hence, from Eq.2, we obtain the following simplified set of equations, d[b P ] dt 1 = [AHP T P ][B])( 1+α[A] ) v 1[B P ], 7

14 d[a] dt d[wus] dt = = a 0 + a 1 [B P ] 1+a 0 + a 1 [B P ]+a 2 [Wus] γ 1[A], n e 0 + e 1 [B P ] 1+e 0 + e 1 [B P ] γ 4[Wus] (3) where we have assumed that the Michaelis-Menten constants K 1,K 2 >> B, B P. Setting γ 1,γ 4 =1,a 0,e 0 = 0, and assuming non-coperativity of WUS binding i.e n =1,weobtain for the steady state values of [B P ], [A], [Wus], η(t [B P ]) = [B P ](1 + α[a]), ([A]) = a 1 [B P ] 1+a 1 [B P ]+a 2 [Wus], ([Wus]) = e 1 [B P ] 1+e 1 [B P ] (4) where η = [AHP T P ] v 1,andB + B P = T is the total protein number for B. By first replacing [Wus] as a function of B P, from the third equation into [A](second equation), we obtain an equation for B P, [ (η +1)a 1 e 1 αa 1 e 1 )]B P 3 +[ (η +1)(a 1 + e 1 + a 2 e 1 )+ηta 1 e 1 αa 1 ]B P 2 +[ηt(a 1 + e 1 + a 2 e 1 ) (η +1)]B P 1 +[ηt]b P 0 =0 (5) The roots of this equation determine the possible steady state values of B P. For bistability to occur there must be three positive roots. A qualitative analysis of the solutions can be done using we use the Descartes Sign Rule 2 We count the number of sign changes (N) that occur as we go through each of the terms of the polynomial for B P, including the last term, which is multiplied by B 0 P. There are at most N, N 2, N 4... positive solutions(the possibility for complex solutions reduce the number of solutions by 2). The coefficient of B 3 P is negative and the last term is positive. Hence, the only way to get three positive solutions is if the sign of second term and sign of the third term are positive and negative respectively, i.e ( )B 3 P (+)B 2 P ( )B 1 P (+)B 0 P. Hence we must have A>0,B < 0, where A = (η +1)(a 1 + e 1 + a 2 e 1 )+ηta 1 e 1 αa 1, B = ηt(a 1 + e 1 + a 2 e 1 ) (η + 1). Since v 1 =1, and 0 <AHPT P < 10, we have 0 <η<10. Let S =(a 1 + e 1 + a 2 e 1 ), then we have, B =10Sη (η +1). (6) B = 1, at η = 0, hence to guarantee the condition that B<0, for 0 <η<10, a limiting condition is that B =0,atη = 10. This results in S<0.11, from the above equation. This implies that, (a 1 + e 1 + a 2 e 1 ) < (7) 2 Anderson, B., Jackson, J. & Sitharam, M. Descartes rule of signs revisited. Amer. Math. Monthly 105, (1998). 8

15 We now use this to set limits for A. WecanwriteA as, A =( Ta 1 e 1 )η 0.11 αa 1. (8) For A to be positive, we must therefore satisfy from the first term in the above equation, 10a 1 e 1 > 0.11,wherewehaveusedT = 10. Since S =0.11, e 1,a 1,a 2 < 0.11, this implies that in this case, the above condition is not satisfied, since the product of a 1 e 1 <.01. Hence A<0, indicating that 3 sign changes are not possible in this case. However, with cooperativity of binding of Wus to Type A ARR s it is possible to obtain switch-like behavior. In Fig. 6, we plot Wus,A,B P as a function of HPT P for n = 4, which shows the switch-like response A B P Wus Concentration SN 2 SN HPT P Figure 6: Bistability of the circuit due to co-operative binding of WUS on the Type-A ARR promoter (n = 4). In addition we have, a 1 =5,a 2 =2,γ 1,γ 4 =0.025,α =10,v 1 =1,e 1 = 0.1. SN denotes saddle-node bifurcation, and the dotted lines represent unstable states. The system exhibits multiple steady states for a single value of cytokinin signaling. Spatial Models of Type A, B ARR s, WUS As shown in Fig. 2 B-D (Main Text), the cytokinin response gradient peaks in the center of the meristem, which follows from the cytokinin receptor concentration which also peaks in exactly the same location (see Fig 3 main text). Our goal was to model the 2 dimensional spatial distribution of Wuschel (corresponding to a longitudinal slice of the meristem (see Fig 2 B,C main text)). We therefore assume a central peak of cytokinin input(cyt in the equation for dahp T P Eq. 2) to the network model. The output for Type-A, phosphorylated Type-B dt ARR s and WUS, is computed from Fig. 3B, by reading as input the spatial distribution of cytokinin input, and the final distributions are plotted in Fig. 2 H-J (Main text). We therefore assume that a central cytokinin signaling gradient (cytokinin receptor) exists, which patterns the network components. Implicit in this assumption are the following: we treat cells as a continuum; cytokinin is assumed to diffuse rapidly into cells and hence 9

16 establishes a steady state profile. The cytokinin signaling gradient is assumed to be of the form, cyt(x, y) =4exp (x 2 + y 2 )/2, (9) which corresponds to Cytokinin signaling as a peak at the center, disappearing at the boundary. The spatial plots are plotted in x, y axis ranging from 5 <x,y<5 arbitrary units. Type-A, phosphorylated Type-B ARR s, WUS read out their values from the network model (Fig. 3B) (also see Fig 2 main text). The previous spatial profiles computed assume no co-operative binding of WUS takes place on Type A ARR. However, if we now include cooperativity, then we can use the results for steady state values of Type-A,phosphorylated Type-B ARR and WUS, shown in Fig. 7 (using the same parameter values to obtain the bistable curve in Fig. 5), to read out values of Type-A ARR etc from knowledge of the input cytokinin signaling gradient(eq. 3). The spatial profiles are plotted for this case in Fig. 8, centered at (0 0). Phosphorylated Type A Type B WUS Relative Levels SN 5 SN Cytokinin Signaling Figure 7: Type-A, B ARR, levels for the bistable response curve. Type-A, Type-B ARR, and WUS are expressed in a switch-like manner, since the bistable curve of Fig. 7 clearly shows the all/nothing nature of the curve. Type-A ARR in spatially expressed as a ring, with almost zero levels at the center, where WUS is high. Interrogating Network Dynamics through variation of individual feedback loops. We discuss the effect on network dynamics, in particular the steady state levels of WUS as a function of cytokinin, for several different interactions, for the wild type(hyp1+hyp2). In each case the strength of an individual feedback loop is varied, and the WUS vs cytokinin 10

17 Figure 8: Spatial patterns of Type-A, B ARR, WUS for the switch-like case. curve is plotted. This way one can infer the contribution of that particular interaction to the circuit dynamics, by observing its effect on the steady state values of WUS vs cytokinin. In each case we vary a particular parameter p i over the three values 2p i,p i, p i, while keeping 2 all the other parameters unchanged( where the wild type parameters are given in Table1). 0.1 WUS Type A ARR Increasing repression of Type-A ARR by WUS (higher values of a 2, Fig 9), relieves negative feedback of the signaling (by Type-A ARR), and hence WUS levels increase. Therefore this interaction leads to higher positive feedback. Notice also that the threshold of signaling required to induce high values of WUS reduces, as lower levels of WUS are sufficient to equivalently suppress Type-A ARRs. 0.2 Type A ARR signaling The strength by which Type-A ARRs inhibit signaling is another important parameter regulating the dynamics of this model. Higher values of α correspond to stronger negative feedback mediated by Type-A ARRs on signaling, and hence results in lower levels of WUS, see Fig

18 25 20 wildtype a 2 + a 2 Relative WUS Cytokinin signaling Figure 9: WUS vs cytokinin as a function of a wildtype α + α Relative WUS Cytokinin signaling 0.3 cytokinin Type A ARR Figure 10: WUS vs cytokinin as a function of α Parameter a 1 controls how much Type-A is induced by phosphorylated Type-B ARR. Thus, increasing a 1, Fig 11, leads to higher induction of Type-A ARR per unit of phosphorylated Type-B ARR, and hence, higher negative feedback of cytokinin signaling. This leads to reduced values of WUS for the same amount of cytokinin signaling. 0.4 cytokinin CLV1 The parameter c 1 controls suppression of CLV1 by cytokinin signaling which relieves negative feedback of WUS on itself. Hence it is easier to induce WUS, asc 1 increases, Fig

19 25 20 wildtype a 1 + a 1 Relative WUS Cytokinin signaling Figure 11: WUS vs cytokinin as a function of a wildtype c 1 + c 1 Relative WUS Cytokinin signaling 0.5 cytokinin WUS Figure 12: WUS vs cytokinin as a function of c 1 The parameter b 1 controls the strength by which cytokinin signaling (phosphorylated Type- B ARR, for example) activates WUS. Therefore increasing b 1 leads to higher induction of WUS for an equivalent value of cytokinin signaling, Fig 13. Phenotypic effects of self-negative feedback of WUS due to the CLV pathway: Time Series plots of WUS for wild type and clv mutant Negative feedback in a system leads to a quicker response in damping out fluctuations. Consider a simple example of a gene which regulates its own expression negatively. This 13

20 30 25 wildtype b 1 + b 1 Relative WUS Cytokinin signaling Figure 13: WUS vs cytokinin as a function of b 1 could be thought to occur if the protein binds to its own promoter and prevents RNA polymerase from transcribing it. This system can be described by the following equations, dx dt = A γx, (10) 1+αxn where x represents the protein concentration, A, α, n&γ correspond to the basal rate of transcription, negative feedback strength, cooperativity coefficient and the degradation rate. Setting the right hand side to zero, this equation can be solved to give x s, the steady state value of the protein concentration. Consider a small perturbation δa(t), around the basal rate A, then the corresponding perturbation around x s,isδx(t), with dδx = δa dt A γx s ( αnγ2 n+1 x s + γ)δx. (11) A Notice that the negative feedback imposes a greater apparent degradation rate (the first term inside brackets), which can increase with stronger feedback. This explains why WUS levels remain at a higher level in the mutant, since in this case the CLV-mediated negative feedback is not active. In the main text a perturbation was applied, namely, treatment with cytokinin, and then different phenotypes were observed in terms of carpel number for clv mutants(fig. 1H, main text). We construct time-series plots Fig. 14, so as to mimic these experiments in which we assume that in each case, the network is initialized for [cytokinin]= 0, and then we simulate the system with [cytokinin]= 10, at t = 0, assuming that cytokinin degrades at a rate γ =1min 1. Hence the final steady state is the same as the initial state, and therefore the time series plots show the effects of a perturbation on the network. The normalized WUS(t) WUS(t=0) time series are plotted as, for 3 different cases. The curve for the mutant shows that relative WUS levels remain at slightly higher levels than when strong negative feedback is present. The mutant lacks the ability for CLV to suppress WUS, and hence there is no negative feedback, thereby leading to higher levels of WUS, and hence its greater phenotypic effect. Increasing negative feedback through stronger binding of the CLV complex to WUS, leads to a quicker fall in WUS levels. 14

21 4 3.5 WT MT Strong Feedback Relative WUS Time Figure 14: WUS vs Time for wild type (including increasing strength of feedback from CLV pathway, i.e how strongly the CLV complex suppresses WUS. black, blue and red correspond to increasing levels of feedback. b 2 =0, 1, 100. Simulations were carried out using MATLAB software (The Mathworks) and the Systems Biology Workbench (SBW/BioSPICE) tools [13]: JDesigner, and Jarnac. The bifurcation diagrams which were used to generate steady state curves for various species were generated using Oscill8 [14] 15

22 K h1 v h1 K h2 K 1 α v 1 K 2 a 0 a 1 a 2 b 0 b 1 b c 0 c 1 k 1 k 2 d Table 1: The concentration of the network components are in dimensionless units, the rate constants (transcription and degradation) are in units of min 1, and the Michaelis-Menten constants are dimensionless. We assume no form of cooperativity unless specifically mentioned, hence n = 1(in the section on bistable response of the circuit, we use n =2,since multiple binding sites for WUS were found on the Type A ARR promoter [3]). Although the results obtained from the model are robust to a wide range of parameters, one guideline we have adhered to is to set the parameters for the AHPT, Type A ARR, and Type B ARR, such that certain relevant time scales are obtained. Namely, the phosphorylation/dephosphorylation time scales of the AHPT s are 10min, and the time scales of the Type A & Type B ARR s, are < 1hr [9, 10, 11] (see Fig 2). To obtain these results we assume a degradation rate for the Type A ARR to be 0.025min 1. We also assume that all degradation rates γ i =0.025min 1 (i=1:5), the transcription/translation rates are chosen to be scaled to a maximum of unity(for parameter values used in typical models of bacterial signaling/regulatory networks see [12]). The total amount of AHPT, Type B ARR in their two different forms obey, AHP T + AHP T P =10,B+ B P = 10. References [1] Muller, B. and Sheen, J., Advances in cytokinin signaling. Science 318 (5847), 68 (2007), see also Hwang, I. and J. Sheen, Two-component circuitry in Arabidopsis cytokinin signal transduction. Nature, (6854): p [2] To, J.P.C. et al., Cytokinin Regulates Type-A Arabidopsis Response Regulator Activity and Protein Stability via Two-Component Phosphorelay. Plant Cell (December 7) (2007). [3] Leibfried, A., et al., WUSCHEL controls meristem function by direct regulation of cytokinin-inducible response regulators. Nature, (7071): p [4] Lindsay, D.L., V. K.Vipen, and P.C. Bonham-Smith, Cytokinin-induced changes in CLAVATA1 and WUSCHEL expression temporally coincide with altered floral development in Arabidopsis. Plant Science, : p [5] Shea MA, Ackers GK (1985) The OR control system of bacteriophage λ. Aphysical chemical model for gene regulation. J Mol Biol 181: [6] Buchler NE, Gerland U, Hwa T (2003) On schemes of combinatorial transcription logic. Proc Natl Acad Sci USA 100:

23 [7] Bintu L, et al. (2005) Transcriptional regulation by the numbers: models. Curr Opin Genet Dev 15: [8] Hasty J, Isaacs F, Dolnik M, McMillen D, Collins JJ (2001) Designer gene networks: Towards fundamental cellular control. Chaos 1: [9] A. P. Mahonen et al.(2006),cytokinins regulate a bidirectional phosphorelay network in Arabidopsis. Current Biology 16, [10] I. B. D Agostino, J. Derure, J. J. Kieber,(2000),Characterization of the response of the Arabidopsis response regulator gene family to cytokinin. Plant Physiology 124, [11] J. P. C. To et al.,(2007),cytokinin regulates type-a Arabidopsis Response Regulator activity and protein stability via two-component phosphorelay. Plant Cell, 19(12): [12] Mitrophanov AY, Groisman EA. Positive feedback in cellular control systems. Bioessays. (2008) 30(6): [13] Sauro HM, Hucka M, Finney A, Wellock C, Bolouri H, Doyle J, Kitano H (2003) Next generation simulation tools: the systems biology workbench and biospice integration. OMICS 7: [14] 17

24 Supporting Information Gordon et al /pnas SI Materials and Methods Plant Materials and Reporter Constructs. clv1 11, clv3 2, and clv2 1 alleles in L-er background have been previously described (36, 37). The pwol::gfp line in Columbia (Col-0) background has been previously described (38) and it recapitulates expression patterns observed in the shoot and root via in situ hybridization (30). The parr5::gfp line in WS ecotype has been previously described (13). The pwus::gfp-er and pclv3::gfp-er lines have also been previously described (10, 34) (for details of other lines see below). The pwus::dsred-n7 construct in the T-DNA vector pmlbart (39) conferring Basta resistance in plants is composed of 3.33 kb of upstream regulatory sequence from the WUS gene fused to dsred followed by the N7 nuclear localization sequence (40) with 1.31 kb of WUS 3 -untranslated sequence. For double transgenic plants with various reporters, pwus::dsred-n7 was transformed into respective backgrounds as previously described (29). Reporter lines were subsequently crossed into mutant backgrounds. Cytokinin receptor mutants in Col-0 background were genotyped as previously reported (41). Quantitative Real-Time PCR (qrt-pcr). Quantitative real-time PCR (qrt-pcr) was performed with Roche Universal Probe Library hydrolysis probes. Each sample represents tissue harvested from 50 two-week-old seedlings just transitioned to flowering. Meristem tissue from 50 plants was harvested and pooled from seedlings after removing leaves, cotyledons and root tissue followed by liquid nitrogen flash freezing and homogenization. Relative expression by qrt-pcr was normalized to NM which has been shown to be a superior reference gene for qrt-pcr analysis, constant against various treatments, including cytokinin treatment (42). Transcript abundance of this gene is 1/10 of UBQ10 transcript levels and therefore more similar to WUS transcript levels. Similar trends were also observed using UBIQUITIN 10 (42). All samples were run in at least triplicate. Total RNA was isolated using the RNeasy mini kit (Qiagen). RNA concentration and quality was assayed using nanodrop spectrometer (Agilent). First-strand cdna synthesis was performed with 2 g total RNA using SuperScript II RNase H- reverse transcriptase (Invitrogen) and 20-mer oligo dt primers according to the manufacturer s instructions. Real-time PCR amplifications were performed in triplicate in 96-well plates in a 20- L reaction volume on a Roche LightCycler 480 system. Unlabeled gene-specific primers in combination with a gene-specific hydrolysis probe from the Roche Universal Probe Library Set were used to detect gene-specific amplification products. For WUS quantification, primers spanned two intron sequences which eliminated products from potential genomic DNA contamination under our PCR settings. Gene specific primers 5 -ggattttcagctactcttcaagcta-3 and 5 - ctgccttgactaagttgacacg-3 with UPL probe 157 were used to amplify and detect NM , the primers 5 -tcagagaacatcttgcctcgt-3 and 5 atttcacaggcttcaataagaaatc-3 with UPL probe 17 were used to amplify ARR5, 5 -ggatacatcgccccagagt-3 and 5 -tccaaattcaccaacaggttt-3 with UPL probe 33 were used for CLV1, the primers 5 -aaccaagaccatcatctctatcatc-3 and 5 - ccatcctccacctacgttgt-3 with probe 33 were used to amplify WUS, and the primers 5 -gaagttcaatgtttcgtttcatgt-3 and 5 -ggattatacaaggccccaaaa-3 with UPL probe 119 were used to amplify and detect UBQ10. Error bars of real-time qrt-pcr experiments in Fig. 1 are derived from three independent biological experiments each run in triplicate, the cytokinin serial dilution curve in Fig. 2A which is derived from two independent biological experiments. We show the mean and SEM between respective biological replicates. Analysis and plots in Fig. 5 are derived from pooled tissue run in triplicate. Plant Growth and Cytokinin Treatment Conditions. Plants were grown as previously described (29). Cytokinin treatments with N6-benzylaminopurine (BAP; Sigma Aldrich Co.) were performed as described (23) except that shoots were sprayed with the respective solutions. For analysis of carpel number, plants were treated three times at 1-week intervals. clv3 2 mutants and wild-type plants in Fig. 3F and Fig. S1F, I, and K were treated once every second day. Phenotypic analysis was performed on soil. Flowers at positions 3 20 of at least 10 plants were counted for carpel numbers of cytokinin and mock treated samples. At least two independent biological experiments were performed for each genotype. Imaging was performed as previously described (29). Membranes were stained with FM4 64 dye unless otherwise noted (29). Computational modeling is described in the Computational Modeling in SI Appendix. Cycloheximide Treatments. Plants were pretreated with 10 M cycloheximide (Sigma Aldrich Co.) for 30 min. then treated with respective BAP or mock solutions also containing 10 M cycloheximide and harvested 4haftertreatment. 39. Gleave AP (2002) A versatile binary vector system with a T-DNA organizational structure conducive to efficient integration of cloned DNA into the plant genome. Plant Mol Biol 20: Cutler SR, Ehrhardt DW, Griffitts JS, Somerville CR (2000) Random GFP::cDNA fusions enable visualization of subcellular structures in cells of Arabidopsis at a high frequency. Proc Natl Acad Sci USA 97: Higuchi M, et al. (2004) In planta functions of the Arabidopsis cytokinin receptor family. Proc Natl Acad Sci USA 101: Czechowski T, Stitt M, Altmann T, Udvardi MK, Scheible W (2005) Genome-wide identification and testing of superior reference genes for transcript normalization in Arabidopsis. Plant Physiol 139:5 17. Gordon et al. 1of5

25 Fig. S1. (A) Finer serial dilutions reveal a steep rise in WUS transcript starting around 400 M and peaking near 600 M, overlapping with a dip of ARR5 transcript. (B) Mean carpel numbers from either mock or cytokinin treated clv2 1 flowers. Error bars, SEM. (C and D) Phenotypes of mock (C) or treated (D) clv1 11 carpels. Cytokinin treatment enhanced the club shaped carpel phenotype. (E and F) clv3 2 mock (E) or cytokinin-treated (F) flowers. Treated flowers were observed with 14 carpels with a central mass of undifferentiated tissue (arrowhead) which was never observed in mock treated flowers. (G and H) Mock (G)or cytokinin-treated (H) clv3 2 meristems showing massive enlargement of the clv3 2 SAM (bounded by arrows) after cytokinin treatment. (I and J) Mock (I) or treated (J) clv1 11 meristems. Treated meristems were slightly wider (50 vs. 100 m) and qualitatively taller than mock treated meristems. (K) undifferentiated tissue within the center of the gynoecium (arrowheads, Fig. S1F) formed after repeated treatment of clv3 2 flowers was marked by WUS (red) and AHK4 (green) reporter expression. (L N) AHK4 (green) and WUS (red) reporter expression in the untreated linear fasciated clv3 2 SAM (arrows) and floral meristems or (O Q) in the grossly enlarged cytokinin treated clv3 2 SAM (arrows) and floral meristems. (Scale bars, 100 m.) Gordon et al. 2of5

26 Fig. S2. (A and B) AHK4 reporter expression in the untreated root (A) and proliferating cells after culture on auxin-rich medium (B). Receptor expression is both stronger and broader in auxin treated samples. (C and D) AHK4 and WUS reporter overlap in the developing rib zone of new shoot meristems forming from callus. (E and F) ptcs::gfp report is also active in the developing RM of regenerating SAMs in culture and peripheral callus cells. Error bars, 100 m in(a and B) 200 m in(e), and 50 m in(c, D and F). Arrowheads mark regenerating SAMs. Gordon et al. 3of5

27 Fig. S3. (A D) are representative results of live imaging experiments of WUS reporter (green) before (A and B)(n 10) and after 24 h of cytokinin treatment (C and D)(n 10). (E H) are representative results of live imaging experiments of WUS reporter (green) before (E and F)(n 10) and after 24 h of mock treatment (G and H)(n 10). WUS expression is induced in the adaxial sides of cytokinin treated leaves (J) but not mock treated samples (I). (K) Cytokinin treatment induces WUS expression in root explants. Semiquantitative RT-PCR on root explants after 0, 5, 10, and 15 days of culture in the absence ( ) or presence ( ) of the cytokinin (50 g/l kinetin). WUS is not expressed in root plants at 0 days of culture (lane 1) but becomes expressed after prolonged culture on MS media (lanes 2 10) and this expression is enhanced by the presence of cytokinin (lanes 3, 4, 6, 7, 9, and 10). (L and M) WUS reporter expression (green) in the stele of cytokinin treated root explants (M) but not untreated roots (L). (N) AHP6 reporter, pahp6::ahp6-gfp (green), a component known to repress cytokinin signaling 28 is expressed in organ primordia where Type-A ARRs are down-regulated. (O) AHK4/CRE1 is expressed in the root stele, where ectopic WUS is induced after cytokinin treatment. (Scale bars, 50 m.) Other Supporting Information Files SI Appendix Gordon et al. 4of5