In animal systems, master regulatory transcription factors (TFs) mediate stem cell maintenance through a direct transcriptional repression of differentiation promoting TFs. Whether similar mechanisms operate in plants is not known. In plants, shoot apical meristems serve as reservoirs of stem cells that provide cells for all above ground organs. WUSCHEL, a homeodomain TF produced in cells of the niche, migrates into adjacent cells where it specifies stem cells. Through high‐resolution genomic analysis, we show that WUSCHEL represses a large number of genes that are expressed in differentiating cells including a group of differentiation promoting TFs involved in leaf development. We show that WUS directly binds to the regulatory regions of differentiation promoting TFs; KANADI1, KANADI2, ASYMMETRICLEAVES2 and YABBY3 to repress their expression. Predictions from a computational model, supported by live imaging, reveal that WUS‐mediated repression prevents premature differentiation of stem cell progenitors, being part of a minimal regulatory network for meristem maintenance. Our work shows that direct transcriptional repression of differentiation promoting TFs is an evolutionarily conserved logic for stem cell regulation.
The plant stem cell regulator WUSCHEL is shown to repress differentiation‐promoting transcription factors. This regulatory network is analyzed with a computational model of the three‐dimensional shoot stem cell niche and a combination of genetic perturbation and live imaging.
We find that the transcription factor (TF) WUSCHEL (WUS) directly binds to the promoters and represses a group of genes including key TFs involved in differentiation thus keeping them repressed in the stem cells of the plant shoot, a mechanistic logic that is similar to animal stem cell regulation.
We use a three‐dimensional computational model of the plant shoot stem cell niche to show that the WUS‐mediated repression of the differentiation program along with the previously reported activation of its own negative regulator leads to a robust stem cell homeostasis in a dynamic growth environment.
Live imaging of target genes upon transient manipulation of WUS levels is combined with model perturbations to validate the proposed network and to connect it with a large body of previous experimental work.
Genome‐wide expression analysis and chromatin immunoprecipitation (ChIP) assays in both mice and human embryonic stem (ES) cells have shown that a core group of transcription factors (TFs) represses several differentiation‐promoting TFs by directly binding to their promoters (Boyer et al, 2005; Loh et al, 2006). Unlike animal systems, stem cell homeostasis in plant systems does not involve cell behaviors such as physical asymmetric cell divisions, oriented cell divisions and cell migration (Reddy, 2008). For example, the shoot apical meristem (SAM) stem cell niche that resides at the tip of each shoot harbours a set of pluripotent stem cells within the central zone (CZ) and provides cells for the development of all aboveground organs (Figure 1A; Steeves and Sussex, 1989). The progeny of stem cells enters the flanking peripheral zone (PZ) where they differentiate into leaves or flowers (Figure 1A). The Rib‐meristem (RM)/organizing center (OC) located beneath the CZ functions as niche and provides cues for stem cell specification (Figure 1A; Mayer et al, 1998). It is believed that regulated patterns of cell division rates of stem cells and their progenitors are important in the timely transition of stem cell progenitors to differentiation pathways (Meyerowitz, 1997). In this context, it is essential to understand the molecular logic that underlies stem cell maintenance in the SAM stem cell niche.
The homeodomain TF WUSCHEL (WUS), which is expressed in the OC/niche, has been shown to be necessary and sufficient for stem cell specification in overlying cells of the CZ (Mayer et al, 1998; Schoof et al, 2000). Besides stem cell specification, WUS restricts its own levels by activating CLAVATA3 (CLV3) transcription in the CZ (Figure 1A; Fletcher et al, 1999; Brand et al, 2000). CLV3, a small secreted peptide, activates CLAVATA1 receptor kinase pathway to restrict WUS transcription to few cells, thus forming a feedback system (Clark et al, 1997; Ogawa et al, 2008). A recent study has shown that WUS protein synthesized in the niche migrates into adjacent cells and activates CLV3 transcription by binding to the promoter (Yadav et al, 2011). This study also reveals that the WUS protein gradient extends into the PZ where stem cell progenitors differentiate (Yadav et al, 2011). Understanding of the WUS function in cells of the PZ may reveal molecular logic of stem cell maintenance. Two lines of evidence have suggested that WUS represses differentiation. First, a live imaging study has shown that WUS determines organ primordia size by regulating patterns of differentiation in the PZ (Yadav et al, 2010). Second, genomic analysis has revealed that WUS downregulates genes expressed in leaves including genes that encode components of transcriptional repression complex (Busch et al, 2010). However, the repressed target genes presented do not provide immediate clues to their function in regulating stem cell homeostasis (Busch et al, 2010).
Here, we have developed a WUS‐responsive gene network at a higher spatial resolution, which reveals that WUS represses several key TFs implicated in various aspects of leaf differentiation. By employing biochemical methods, we further show that WUS directly binds to the promoter regions of KANADI1, KANADI2, YABBY3 and ASYMMETRIC LEAVES2, TF encoding genes that are expressed in the PZ and mediate early aspects of leaf differentiation. We develop a cell‐based computational model of the SAM by integrating WUS‐mediated direct transcriptional repression of differentiation with a CLAVATA/WUSCHEL regulatory network model. The differential equation model predicts that WUS function is required to prevent premature differentiation of stem cell progenitors. We validate model perturbation predictions through transient manipulation of WUS levels followed by live imaging, which shows that WUS‐mediated direct transcriptional repression is required to keep differentiation program at a distance from the meristem center/stem cell domain.
Development of a high spatial resolution WUS‐responsive gene network
To explain the function of WUS in stem cell maintenance, we developed a high‐resolution spatial map of the WUS‐regulated transcriptome. An earlier study has employed FACS‐mediated purification of distinct cell types of the stem cell niche to develop a cell type‐specific high‐resolution map (Yadav et al, 2009). The method that is based on isolating fluorescently labeled cell populations from SAMs of apetala1‐1;cauliflower1‐1 (ap1‐1;cal1‐1) double mutants has been shown to accurately report gene expression patterns and also to be highly sensitive in detecting transcripts of ∼1000 genes that have not been detected, in any developmental stages, when whole tissues were used (Yadav et al, 2009). We introduced a dexamethasone (Dex)‐inducible form of WUS consisting of the WUS protein‐coding region fused to the ligand‐binding domain of the rat glucocorticoid receptor (GR), expressed under a ubiquitous promoter (35S:17444292:WUS‐GR) into apetala1‐1;cauliflower1‐1 (ap1‐1;cal1‐1) double mutant background. 35S:17444292:WUS‐GR;ap1‐1;cal1‐1 SAMs that were treated with 10 μM Dex solution for 12 h revealed an increase in stem cell domain as revealed by CLV3 (a WUS target gene) promoter activity when compared with mock‐treated SAMs (Figure 1B and C; Yadav et al, 2011). WUS‐mediated activation of CLV3 was observed as early as 2 hs after Dex application with CLV3 expression reaching maximum levels by 4 h and saturating by 6 h after Dex application (Figure 1D). Therefore, SAMs were treated with 10 μM Dex solution for 4 h and RNA samples were hybridized to Arabidopsis ATH1 gene Chip (Affymetrix). To identify putative genes that are directly regulated by WUS, SAMs were, in independent experiments, simultaneously treated with 10 μM Dex and 10 μM Cycloheximide (Cyc), protein synthesis inhibitor and Cyc alone for 4 h. A comparison of Dex‐treated samples with mock identified 641 genes as differentially expressed (DEGs) (⩾/⩽2‐fold; P<0.01), which consisted of 238 upregulated genes and 403 downregulated genes (Supplementary Tables 1 and 2). While a comparison of Dex+Cyc with Cyc identified 457 genes as DEGs (⩾/⩽2‐fold; P<0.01), which consisted of 154 upregulated genes and 303 downregulated genes (Supplementary Tables 1 and 3). Next, we compared the overlap between genes that respond to WUS upon Dex treatment alone with Dex+Cyc treatment by constructing a four‐way Venn diagram (Supplementary Figure 1). This analysis revealed a set of 49 upregulated genes and 140 downregulated genes that were common to both treatment conditions (Supplementary Figure 1).
WUS represses a group of differentiation promoting TFs
To determine the role of WUS‐responsive genes, we mapped their patterns of expression on the various sub‐domains of the SAM. An earlier study has shown that 2361 genes are differentially expressed (⩾2‐fold; P>0.01) across three cell types including the CZ, the RM and the PZ (Yadav et al, 2009). To visualize the spatial distribution of WUS‐responsive genes, we superimposed DEGs on to the cell type‐specific gene expression map. In all, 17% (40 genes out of 238) transcripts activated by WUS in Dex versus mock comparison (⩾2‐fold; P>0.01) were mapped to the CZ and the RM while only 5% transcripts (11 genes) were found in the PZ cells (Figure 2A; Supplementary Table 1). An opposite trend was observed for genes that were downregulated wherein 38% (152 genes out of 403) (⩽2‐fold difference; P>0.01) transcripts were either mapped to the PZ or their expression was detected in both the PZ and the RM. Only 10% (42 genes) were found in the CZ and the RM (Figure 2B; see Supplementary Table 1), showing that WUS represses a large group of genes expressed in differentiating cells. The genes that were not mapped to any of the three cell types may be broadly expressed in SAMs. Alternately, the resolution of the expression map that involves just three cell types may be limited to resolve genes enriched in individual cell layers of the SAM since CLV3 expression domain overlaps with that of WUS (Yadav et al, 2009). In Dex+Cyc versus Cyc comparison, 17% (26 genes out of 154) WUS‐activated genes mapped to the CZ and the RM while only 8% (12 genes) were part of the PZ cells (Figure 2C; Supplementary Table 1). Among 303 downregulated genes, only 11% (33 genes) were mapped to the CZ and the RM while 29% (87 genes) of them were either part of the PZ or their expression was detected in both the PZ and the RM (Figure 2D; Supplementary Table 1). Activation of CLV3 and repression of ARABIDOPSIS RESPONSE REGULATOR 7 and 15 were observed in the presence of Cyc (Supplementary Table 1), which is consistent with them being direct transcriptional targets of WUS (Leibfried et al, 2005; Zhao et al, 2010; Yadav et al, 2011). To exclude artifacts, if any, from ap1‐1;cal1‐1 SAMs, we re‐analyzed a select set of genes identified in the microarray experiments by performing qRT–PCR analysis on RNA samples extracted from finely dissected wild‐type SAMs expressing 35S:17444292:WUS:GR. This analysis showed that all six of the activated targets tested including CLV3 were activated within 4 h of WUS induction in both the absence and the presence of Cyc (Figure 2E and F). A similar agreement between microarray analysis and qRT–PCR analysis was observed for 14 of the repressed targets analyzed in both the absence and the presence of Cyc (Figure 2G and H). Taken together, these results reveal that a majority of WUS‐activated genes are broadly expressed within the central parts of SAMs and that WUS represses a large group of genes expressed in the PZ. A similar trend was observed in Cyc‐treated SAMs, suggesting that WUS may repress genes expressed in differentiating cells of the PZ through a direct transcriptional control.
Among genes downregulated in the presence of Cyc, 37 were annotated as TFs (Supplementary Table 1). The repressed TFs included key TFs implicated in leaf polarity establishment and differentiation such as KANADI 1 and 2 (KAN1 and KAN2) (Kerstetter et al, 2001), ASYMMETRIC LEAVES 2 (AS2) (Lin et al, 2003) and YABBY 3 (YAB3) (Kumaran et al, 2002; Figure 2G and H). Genes encoding homeodomain TFs, KNAT1/BREVIPEDICELLUS (BP) (Figure 2G and H) and BLH5 (BELL1‐LIKE HOMEODOMAIN 5), that have been shown to be expressed in the PZ and the RM of SAMs, and implicated in cell fate specification and inflorescence stem growth were also repressed by WUS (Lincoln et al, 1994; Douglas et al, 2002; Venglat et al, 2002; Bhatt et al, 2004; Rutjens et al, 2009). This apart, WUS represses BHLH093, a FAMA class of basic helix‐loop‐helix TF required to promote differentiation of stomatal guard cells of leaves (Ohashi‐Ito and Bergmann, 2006) ANAC083/VND‐INTERACTING 2 (Yamaguchi et al, 2010), C2C2 domain‐DOF Zinc Finger TF‐DOF2.4 and AT4G24060 implicated in vascular development in leaves (Gardiner et al, 2010). The repressed genes also included SCARECROW (SCR) that encodes GRAS‐domain TF and GROWTH‐REGULATING FACTOR 6 (GRF6), both are involved in leaf growth (Kim et al, 2003; Dhondt et al, 2010). We analyzed the RNA localization patterns of a select set of differentiation promoting TFs repressed by WUS (Supplementary Figure 2). Expression of KAN1, KAN2 and AS2 was restricted to few cells of the PZ of both SAMs (Supplementary Figure 2A–C) and floral meristems (Supplementary Figure 2D–F). Taken together, our results show that WUS represses a group of differentiation promoting TFs thus excluding them from stem cells of the CZ.
WUS represses differentiation promoting TFs by binding to their regulatory regions
Next, we tested whether WUS directly binds to the regulatory regions of repressed TFs by performing ChIP coupled to qPCR analysis (ChIP‐qPCR) by using anti‐WUS (peptide)‐antibodies described in an earlier study (Yadav et al, 2011). We obtained immunoprecipitated DNA that was enriched in specific genomic regions of KAN1, KAN2, AS2 and YAB3 genes (Figure 3A–D). We used electrophoretic mobility shift assays (EMSAs) for fine mapping of WUS binding sites within these genomic regions by testing a series of short oligonucleotides for their ability to bind to WUS (Figure 3E). DNA sequences of KAN1, KAN2, AS2 and YAB3 that bound WUS revealed the presence of conserved TAAT core sequences (Figure 3F). Previous studies have shown a similar WUS binding specificity to TAAT containing sequence elements in EMSAs (Lohmann et al, 2001; Leibfried et al, 2005; Busch et al, 2010; Yadav et al, 2011). In some cases, for example the sequence of KAN2 650, AS2−480, AS2 750 and YAB3 500 oligonucleotides contained more than one TAAT motif (Figure 4H). Therefore, we explored the WUS binding specificity of KAN1, KAN2, AS2 and YAB3 promoter‐specific oligonucleotides by introducing mutations into all TAAT motifs present in them (Figure 4H). Mutations within the TAAT core sequences abolished WUS binding, suggesting that WUS binding to these sequences is specific (Figure 4A–G). In cases where more than one TAAT core was detected, a selective lack of binding was observed with mutations in certain TAAT elements and not others (Figure 4H). This analysis shows that TAAT sequences are important but not sufficient to determine WUS binding specificity. In addition, a ‘super shift’ was observed when KAN1 promoter oligonucleotides (Figure 3G) were incubated with anti‐WUS antibody confirming the presence of WUS protein in WUS‐DNA complex. We next tested whether the WUS‐binding element found in the KAN1 promoter can modulate transcription in a heterologous transient expression system by examining the LUCIFERASE (LUC) reporter expression levels in protoplasts isolated from mesophyll cells of Arabidopsis leaves. These results show WUS‐dependent repression of LUC when KAN1 promoter context containing the WUS‐binding element was used (Figure 4I). In contrast, the WUS‐dependent repression of LUC was abolished when KAN1 promoter context carrying a mutated WUS‐binding element was used (Figure 4I). Taken together, these results demonstrate that WUS protein binds to the regulatory regions of differentiation promoting TFs to repress their expression.
A computational model reveals the importance of direct transcriptional repression of differentiation in stem cell homeostasis
Thus far, our analysis suggests that stem cell maintenance may partly be a result of collective repression of differentiation promoting TFs. To understand the importance of the direct transcriptional repression of differentiation genes by WUS, we implemented these interactions in a differential equations model of a three‐dimensional SAM tissue, and integrated it with a model of the core CLV3‐WUS negative feedback stem cell regulatory network (Figure 5A; Supplementary information; Yadav et al, 2011). The model includes the direct activation of CLV3 transcription from the diffusing WUS protein (Yadav et al, 2011), as well as activation from a signal originating from the epidermal cell layer (Jönsson et al, 2003; Yadav et al, 2011). Although a signal originating in the L1 is yet to be identified in molecular experiments, we propose such a signal based on the behavior of a meristem model in specific mutants in which CLV3 expression is limited to the outermost cells even when the SAM undergoes dramatic morphogenetic changes (e.g., Brand et al, 2002; Supplementary Figure 3I; Supplementary information). The expression of WUS is negatively regulated from a diffusive CLV3 signal (Fletcher et al, 1999; Brand et al, 2000), and is activated by a static cytokinin/AHK4 signal located at the center of the OC (Supplementary Figure 4; Jönsson et al, 2005; Gordon et al, 2009; Yadav et al, 2011). In this model, we introduced interactions between WUS and the PZ, where KAN1 was used as a PZ representative. KAN1 expression as revealed by pKAN1:17444292:KAN1:GFP (KAN1 promoter driving the expression of KAN1:GFP translational fusion) shows that it is expressed in the outer edges of the PZ (Figure 5E). We introduced a direct transcriptional repression of KAN1 by the WUS protein as revealed by this study. To generate a sharp KAN1 expression close to the epidermis, we also introduced an activating signal originating from the L1 layer (Figure 5A). Note that this addition only keeps the KAN1 expression close to the epidermis, all results reported in this study would be the same without it, except that the PZ gene would be expressed also in internal layers of the PZ. An earlier study has shown that ubiquitous misexpression of KAN1 results in SAM termination which was argued to be a consequence of abaxialization of leaves (Kerstetter et al, 2001). We tested whether ubiquitous KAN1 misexpression leads to downregulation of WUS levels by using Dex‐inducible KAN1 (35S:17444292:KAN1:GR). An RT–PCR analysis of 35S:17444292:KAN1:GR seedlings exposed to 10 μM Dex for 5 days revealed a downregulation of WUS transcript levels (Supplementary Figure 5). An earlier study has also shown that YABBY3 and FILAMENTOUS FLOWER (FIL) which are expressed in differentiating cells of organ primordia are required to restrict WUS and CLV3 expression domains through a non‐cell autonomous mechanism (Goldshmidt et al, 2008). We identified YAB3 as one of the genes directly repressed by WUS (Figures 2C, 3D and E). From these data, we use KAN1 as a representative PZ signal modeled as a non‐diffusing protein repressing WUS expression. Since this interaction has not been verified at the molecular level, we used the model to investigate how different interactions from the PZ on the CLV3/WUS network behaved when ubiquitous expression of KAN1 was introduced (Supplementary Figure 6; Supplementary information). The only way in which long‐term downregulation of WUS could be achieved was if KAN1, possibly indirectly, represses WUS or alternatively, the WUS activation network (cytokinin or AHK4 in our model, Supplementary information). Transcription is modeled with Hill functions and proteins are linearly created based on RNA concentrations (Materials and methods; Supplementary information). RNA and proteins undergo mass action degradation, and CLV3 and WUS diffuse between neighboring cells. An optimization strategy was implemented to obtain sets of parameter values able to fit the expression domains of the three‐dimensional tissue template (Supplementary information). The domains were defined using confocal microscopy data (Figure 5C–E). Using this strategy, 229 parameter sets were extracted in which the model successfully describes the wild‐type expression domains (Figure 5B–E) and the WUS gradient (Figure 5F), showing that the regulatory network with a core where WUS acts as a hub with negative feedback with the CZ and positive (doubly negative) feedback with the PZ is sufficient to organize SAMs.
We also analyzed models where individual regulatory interactions were removed. None of the reduced models were able to match experiments, suggesting that the model represents a minimal set of interactions for this network (Supplementary information).
Transient gene manipulations and live imaging support repression model
We next tested the effects of transient manipulation of WUS levels on the stem cell niche in live imaging experiments. To transiently increase WUS level, we introduced Dex‐inducible 35S:17444292:WUS:GR into plants carrying both pKAN1:17444292:KAN1:GFP (nuclear localized) and the stem cell marker‐pCLV3:17444292:mGFP5‐ER (endoplasmic reticulum localized GFP). Within 24 h of Dex application, a dramatic expansion of the pCLV3 expression was observed that was accompanied by the loss of pKAN1 expression (n=4) (Figure 6A and B). The corresponding perturbation was applied to the computational model (Supplementary information). Consistent with the experimental observations, the model predicts a large expansion of the CLV3 expression domain along with a reduction in KAN1 expression (Figure 6E and F), which was a robust result for all 229 parameter sets optimized for wild‐type behavior (Table I; Supplementary Figure 7E). Conversely, the downregulation of WUS levels, achieved through the Dex‐inducible overexpression of CLV3 (35S:17444292:GR:LhG4;6XOP:17444292:CLV3) (Supplementary Figure 8), resulted in a gradual misexpression of KAN1 in stem cell progenitors that are located closer to the stem cell domain revealing premature differentiation of stem cell progenitors (n=6) (Figure 6C and D; Supplementary Figure 9). Introducing ubiquitous CLV3 expression in the model (Figure 6G; Supplementary information) resulted in a dramatic reduction in WUS expression (Supplementary Figures 7F and 3E), and in a gradual shift of KAN1 expression toward the CZ (Figure 6H; Supplementary Movie 1). The KAN1 expansion was, as opposed to the experimental result, continued throughout the CZ (Figure 6D and H), suggesting additional KAN1 repressors in the CZ, although a similar result could also be achieved in the model with a relatively weaker CLV3 overexpression (weaker 35S promoter activity) (Supplementary Figure 10). Earlier studies have shown that ubiquitous misexpression of KAN1 (Kerstetter et al, 2001) and AS2 (Lin et al, 2003) leads to differentiation and termination of SAM growth. We observed SAM and floral meristem termination upon misexpression of KAN1 under the influence of the WUS promoter (Supplementary Figure 11), which when applied in the model led to disappearance of WUS and CLV3 expression (Supplementary Figures 12B and 3H). Taken together, these observations show that WUS‐mediated transcriptional repression of differentiation promoting factors prevents premature differentiation of stem cell progenitors.
The integrated computational model provides a systems description of gene expression regulation in a growing and proliferating SAM tissue
To further connect the new regulatory (repression) interactions from this study with previous work, a total of 17 perturbations including loss‐of‐function and ectopic expression of CLV3, WUS and KAN1, as well as transport malfunction of CLV‐WUS network components were applied to the computational model (Table I; Supplementary information). Simulations with the 229 different parameter sets show some variance in the quantitative behavior (Supplementary Figures 7 and 12), but interestingly the model interaction network is robust in the qualitative (increase versus decrease) response to almost all perturbations (Table I). All perturbations correspond qualitatively with experimentally measured changes in expression domains, where available, and other simulations can be seen as predictions of changes. Loss‐of‐function mutants have been repeatedly studied in the literature and these mutants were implemented in the model at the protein level making it possible to investigate the changes in mRNA levels of the perturbed gene due to the network interactions. The clv3 loss‐of‐function shows an increase in WUS transcripts and an increase in CLV3 promoter activity and subsequent expansion, and these simulations are consistent with experimental observations made upon transiently downregulating CLV3 expression (Reddy and Meyerowitz, 2005), although our static template cannot address the increase in SAM size. The PZ region is predicted to move outwards in these simulations (Supplementary Figure 3C). For the wus loss‐of‐function, CLV3 expression decreases (Brand et al, 2002), and in the model this is at least a four‐fold decrease (Supplementary Figure 7A). KAN1 expression is predicted to move inwards, comparable to the ubiquitous CLV3 expression simulations (Figure 6H), but since it is not a null‐mutant implementation (the amount of WUS is strongly decreased, but not set to zero), KAN1 expression does not cover all of the CZ (Supplementary Figure 3B). Ubiquitous CLV3 and WUS expression simulations are described previously (Figure 6), and simulations with ubiquitous KAN1 expression lead to extremely low CLV3 and WUS expression (Supplementary Figures 7G and 3F), in agreement with the decrease in WUS levels (Supplementary Figure 5), and the terminating SAM phenotype (Kerstetter et al, 2001). The same is true for the pWUS:17444292:KAN1 simulations (Supplementary Figure 3H, confer Supplementary Figure 11), and the pCLV3:17444292:WUS simulations show the non‐trivial change of both CLV3 and WUS expressions into high levels in the outermost three layer of cells (Supplementary Figure 3I), similar as reported in a previous model (Yadav et al, 2011), and as seen in experiments (Brand et al, 2002). KAN1 expression is predicted to be lost in the meristem in these simulations. Detailed descriptions of all other mutant simulations are given in Supplementary information.
The model is robust to local parameter perturbations (Supplementary Figure 13), and from the perturbations it is possible to follow some mechanisms in the model. The perturbation of parameters for KAN1 levels do not influence CLV3 or WUS at all; KAN1 does not influence the core network results including when perturbed, except for when KAN1 is expressed ectopically. More important is the dependence of CLV3/WUS on each other, where the direct interactions are seen, for example, when gene expression activities are altered (VW, VC), while the feedback mechanism leading to homeostasis is also indicated, for example, an increased CLV3 (WUS) production (PC, PW) leads to decreased CLV3 (WUS) expression. When the model was challenged in simulations including growth and proliferation (Supplementary Movie 2), the three studied expression domains showed little variation of size and expression intensity as cells moved through the different meristem regions. Taken together, the model shows that the WUS‐mediated regulatory network links the CZ and the PZ, and provides sufficient mechanistic framework to explain stem cell homeostasis in a proliferating SAM tissue.
Other WUS‐responsive gene functions and pathways
Besides activation of CLV3 and repression of differentiation promoting TFs, performing enrichment analyses of gene ontology (GO) terms (Supplementary Table 4) assessed other molecular pathways enriched in the WUS‐regulated transcriptome in the presence of cycloheximide (Supplementary Table 3). The WUS upregulated transcriptome contained genes involved in developmental processes (GO:0032502; P<0.05). This included ANAC018 and CUP SHAPED COTYLEDON1 (Takada et al, 2001) and EXCESS MICROSPOROCYTES1 (EMS1), a receptor‐like kinase that promotes cytokinesis during anther development (Zhao et al, 2002). CUC1 that encodes an NAC‐domain containing TF has been shown to express in presumptive cells that give rise to SAMs in early embryonic development (Takada et al, 2001). Therefore, WUS activation of CUC1 expression may be relevant in establishment or maintenance of embryonic SAMs. ARR4, a type‐A response regulator implicated in negatively regulating cytokinin signaling (To and Kieber, 2008) and seven genes related to response to plant hormone abscisic acid (GO:0009737; P<0.01) were also activated by WUS. WUS also activates GLABRA1 implicated in epidermal cell fate specification (GO:0001708; P<0.01). Taken together, WUS activates genes implicated in diverse plant processes including those that are critical for stem cell maintenance. The WUS downregulated transcriptome contained significantly enriched GO categories that represent indole derivatives and biosynthesis processes (GO:0042435, P<8.19 × 10−5), regulation of gene expression (GO:0010468, P<0.002) and cell‐to‐cell communication and the catalyzation of post‐translation modification (GO:0043687, P<0.01) (Supplementary Table 4). Auxin signaling has been shown to promote columella distal stem cell differentiation in Arabidopsis root (Ding and Friml, 2010) and promote leaf differentiation in SAMs (Veronoux et al, 2010). WUS represses ANTHRANILATE SYNTHASE ALPHA SUBUNIT1 (ASA1) that encodes a rate‐limiting step in tryptophan biosynthesis, TRYPTOPHAN AMINOTRANSFERASE RELATED2 (TAR2) that encodes a protein similar to TRYPTOPHAN AMINOTRANSFERASE OF ARABIDOPSIS1 involved in IAA biosynthesis and PHYTOALEXIN DEFICIENT3 (PAD3) that encodes a cytochrome P450 enzyme that catalyzes the conversion of dihydrocamalexic acid to indole derivative camalexin (Mano and Nemoto, 2012). The finding that WUS inhibits auxin biosynthetic genes may add another layer of regulation in preventing differentiation of stem cells in addition to the repression of differentiation promoting TFs. However, a high‐resolution functional analysis of these genes is necessary to integrate them into the current model of stem cell maintenance.
WUS has been shown to activate transcription of its own negative regulator CLV3 by directly binding to the promoter regions (Yadav et al, 2011). Here, we have shown that WUS represses transcription of differentiation promoting TFs by directly binding to the promoter regions. Thus, the WUS‐mediated repression of differentiation program observed in this study links the PZ with the well‐established feedback loop between cells of the CZ and the RM through a direct transcriptional control. The dual function of WUS in regulating its own level by activating a negative regulator and repressing differentiation would lead to a robust homeostatic mechanism by balancing stem cell numbers and differentiation rates of stem cell progenitors in a dynamic cellular environment. Our computational model, integrating the WUS‐PZ interactions extracted in this work with WUS‐CZ interactions, highlights this by robustly generating the spatial patterns for CLV3 and KAN1 expression, and the use of the multiple (17) model perturbations connects our findings with more than a decade of previous experimental work. Since no interaction in the model can be removed without losing ability to explain some experimental work, it also suggests that the current model provides a minimal model for regulating stem cell maintenance in the SAM. WUS has been shown to form a concentration gradient with higher levels in the OC/RM and lower levels in adjacent cells (Yadav et al, 2011). It is intriguing how a single WUS gradient results in the transcriptional activation of some genes, and the repression of other genes in placing a negative regulator at a distance from its own domain of expression while keeping the differentiation program repressed in stem cell progenitors, and the computational model resolves this by having different WUS thresholds for CZ and PZ regulation. WUS has been shown to act both as an activator and as a repressor of transcription (Ikeda et al, 2009). WUS utilizes DNA binding cis‐elements containing similar core sequences (TAAT) both to activate CLV3 transcription (Yadav et al, 2011) and to repress expression of differentiation promoting TFs (Figure 3F). Therefore, WUS may utilize spatially localized co‐activators and co‐repressors to modulate transcriptional output. The computational model requires an L1‐derived signal to position CLV3 expression domain within the CZ (Figure 5A; Jönsson et al, 2003; Yadav et al, 2011), which is consistent with the requirement of additional factors that may function along with WUS. One such spatial signal could be the local synthesis of active cytokinins in superficial cell layers as suggested by the expression patterns of LONELYGUY class of genes (Yadav et al, 2009; Chickarmane et al, 2012). Genome‐wide expression profiles that have been described will form a rich resource in identifying regional factors that may modulate WUS function (Yadav et al, 2009). WUS binds to some TAAT containing sequences but not all (Figure 4A–H), suggesting that sequence context may determine binding specificity. However, a comparison of WUS binding sequences of several WUS‐regulated genes did not reveal any consensus sequences around the TAAT core (Figures 3F and 4H). Shape analysis of cis‐element sequences that bind Drosophila homeodomain TF‐HOX has revealed that the width of the minor grove determines the binding specificities (Rohs et al, 2009; Slattery et al, 2011). Future studies unraveling genome‐wide WUS binding patterns along with sequence and shape analysis of WUS binding sites may provide insights into WUS binding specificity and transcriptional modulation.
The use of multiple parameter value sets for the model, and the large consistency for these to respond to perturbations shows the robustness of the modeled SAM gene regulatory network architecture, also during growth. Interestingly, the model predicts that the WUS and CLV3 domains can respond differently to perturbations, indicating the possibility for variability within plant populations. Our methodology, using populations of models (parameter values), is a step toward integrating statistical analysis of experimental population behavior within a modeling framework, something that might be necessary for elucidating the exact regulatory interactions present in the SAM.
Our work shows that WUS functions to prevent premature differentiation of stem cell progenitors. The computational model shows that the experimentally extracted interactions where WUS binds directly and represses the expression of TFs involved in PZ differentiation are sufficient to explain the dynamic changes caused by transient perturbations where expression domains grow or shrink at the boundaries. However, transient depletion of WUS fails to induce differentiation of long‐term stem cells within the CZ as suggested by lack of misexpression of auxin‐sensitive differentiation marker shown in an earlier study (Yadav et al, 2010) and lack of misexpression of KAN1 (Figure 6D; Supplementary Figure 9J). Leaf formation from the CZ was seldom observed in wus mutants instead they continue to produce leaves albeit aberrantly from a defective SAM in a ‘stop and go’ fashion (Laux et al, 1996). This suggests that long‐term stem cells may be active in these mutants and argues for function of overlapping mechanisms in protecting long‐term stem cells from differentiation signals, as also suggested by the computational model showing a stronger expression phenotype compared with the experiment. The WUSCHEL‐LIKE HOMEOBOX5 (WOX5) TF of the WUS family is expressed in the quiescent center of the root meristem and has been shown to repress differentiation of neighboring initials/stem cells in a non‐cell autonomous manner (Sarkar et al, 2007). This suggests that WOX‐mediated suppression of differentiation of stem cell progenitors is a conserved theme in both shoot and root stem cell niches. Future work should reveal whether WOX5 shares mechanistic similarities with WUS with regard to the nature of non‐cell autonomous signaling and molecular mechanisms of repression of differentiation.
An increase in WUS levels leads to de‐differentiation of stem cell progenitors into stem cells (Reddy and Meyerowitz, 2005; Yadav et al, 2010). Our work provides a molecular basis for de‐differentiation wherein WUS‐mediated direct repression of differentiation program plays an important role. Deciphering how stem cell progenitors are maintained in WUS‐responsive state requires an understanding of the molecular control of the cellular memory system. Studies on murine ES cells have identified bi‐valent stretches of both H3 lysine 27 methylation, a gene repressive chromatin mark and H3 lysine 4 methylation, a gene activating chromatin mark within genes which encode differentiation promoting TFs that are repressed by master regulatory stem cell promoting TFs (Bernstein et al, 2006; Mikkelsen et al, 2007). This suggests that the bi‐valent domains are responsible for keeping differentiation genes silent in ES cells, while keeping them primed for activation. Profiling of chromatin modifications in individual cell types of SAM stem cell niche, by using methods described earlier (Yadav et al, 2009), should reveal whether collaboration between repression of differentiation and epigenetic pathways is a conserved theme in maintaining stem cell progenitors in a flexible state.
Materials and methods
Plant growth and live imaging
Plasmid constructs, selection of transgenic lines and rescue analysis
The details of the 35S:17444292:WUS:GR (Yadav et al, 2010) and pCLV3:17444292:mGFP5‐ER (Reddy and Meyerowitz, 2005) transgenic lines have been described. The pKAN1:17444292:KAN1:GFP is a translational fusion that includes 4.95 kb of upstream sequence plus the entire coding sequence with introns fused as a translational fusion to eGFP. 35S:17444292:KAN1:GR and 35S:17444292:GR‐LhG4;6XOP:17444292:CLV3 plants have been described in Wu et al (2008) and Yadav et al (2010) respectively.
WUS induction and microarray experiments
Four‐week‐old 35S:17444292:WUS‐GR;ap1‐1;cal1‐1 plants were either treated with 10 μM solution of Dex or in combination with 10 μM cycloheximide (Cyc) for 4 h. Mock‐treated or Cyc‐treated plants served as controls. RNA was isolated from SAMs of 30–35 plants for each treatment by using RNeasy kit (Qiagen). Probe synthesis and array hybridizations were carried out as described earlier (Yadav et al, 2009). Microarray data analysis was performed in R using BioConductor packages. The probe set to gene locus mapping for ATH1 gene chip, removal of redundant probe sets and present call analysis was carried out as described earlier (Yadav et al, 2009). To assess differentially expressed genes, raw data CEL files were normalized using GCRMA package in R (Supplementary Table 5). DEG was determined with LIMMA package using the normalized expression values (Supplementary Table 1). The Benjamini and Hochberg method was selected to adjust P‐values (P) for multiple testing and to determine FDR. As confidence threshold an adjusted P⩽0.01 was used (Yadav et al, 2009). A gene was determined as enriched in a cell type when it was two‐fold upregulated. A similar method was used to determine WUS‐response genes. A gene was considered as a DEG if it was two‐fold upregulated or downregulated (P<0.01) in pairwise comparisons of Dex versus mock or Dex+Cyc versus Cyc. The MAS5 normalized expression values for three cell types were used to generate present (P), marginal (M) and absent (A) calls (Supplementary Table 6). The present call information (PMA values) from the Wilcoxon signed rank test of the MAS5 algorithm is provided for three cell types in Supplementary Table 6. The probe set showing present calls in all three replicates of a sample was considered as positive. GO analysis was performed as described in an earlier study (Yadav et al, 2009).
Four‐week‐old 35S:17444292:WUS‐GR wild type SAMs were either treated with 10 μM solution of Dex or in combination with 10 μM cycloheximide (Cyc) for 4 h. Mock‐treated or Cyc‐treated plants served as controls. 35S:17444292:KAN1:GR plants were exposed to Dex treatment for 5 days before RNA isolation. RNA was isolated using RNeasy kit (Qiagen). cDNA was reverse transcribed using ThermoScript RT (Invitrogen). qRT–PCRs were performed either using sensiMix SYBER kit (Bioline) or SYBR green (BIO‐RAD) on a BIO‐RAD iQ5 Cycler. Three replicates were used for each reaction and UBIQUTIN1 was used to normalize the mRNA levels.
ChIP and qPCR analysis
ChIP and qPCR were performed on 4‐week‐old 35S:17444292:WUS:GR;ap1‐1;cal1‐1 inflorescences, upon Dex or Mock treatment for 24 h, as described earlier (Yadav et al, 2011). Previously published anti‐WUS (peptide) antibodies were used (Yadav et al, 2011). qPCR analysis by using primers overlapping the gene body and regulatory regions of KAN1 KAN2, YAB3 and AS2 (Supplementary Table 7 for list of primers). Three independent sets of biological samples were analyzed.
Electrophoretic mobility shift assay
His‐WUS fusion protein was expressed in BL21 Escherichia coli after induction with 1 mM of IPTG for 5 h at 30°C and purified as described earlier (Yadav et al, 2011). Single‐stranded complementary oligonucleotide fragments corresponding to regulatory regions of KAN1 KAN2, YAB3 and AS2 (Supplementary Table 7) were radiolabeled as described earlier (Yadav et al, 2011). DNA‐protein binding reaction and electrophoresis was performed as described earlier (Yadav et al, 2011).
Protoplast isolation and transactivation assay
Mesophyll protoplast isolation and transactivation was performed as described earlier (Yadav et al, 2011) by using 200 bp sequence containing WUS binding element in KAN1 −1050 region.
Computational modeling and algorithms
The model is simulated on a meristem tissue template comprising 1366 (spherical) cells defined by spatial coordinates and size. Cell neighborhood is defined from the overlap between the spherical cells. The expression domains of WUS, CLV3 and KAN1 are manually defined with confocal microscopy data used as a template. The differential equation model describes gene expression regulation using Hill functions. Production of peptides and proteins, along with degradation are modeled with mass action kinetics, and transport between cells is passive (diffusion like). Two positional cues are included (WUS activator and epidermal cell layer; Supplementary Figure 4) producing three diffusing gene activators. The dynamics of the molecular concentrations are described by a system of differential equations given by where [X] represents concentration of molecule X, upper (lower) case represents mRNA (protein) concentrations. The network is present in each individual cell (cell indices have been omitted) and C, W and K are CLV3, WUS and KAN1 concentrations. A1 (A2) are the static positional cues and a1c, a1k and a2 their signaling molecules. Parameters Vx represents maximal expression rates (for gene x), kx are the Hill constants, and n the Hill coefficients (always equal to 2). Px are protein production rates, gx degradation rates and Dx passive transport rates (Δ represents the passive diffusion‐like transport).
The model and its parameters are described in more detail in Supplementary information. Simulations are performed using a fifth order Runge‐Kutta solver with adaptive steps.
In all, 229 model parameter value sets where simulations mirror the biological behavior at equilibrium were obtained with a stochastic gradient descent optimization approach (Supplementary information). The parameter values used for simulations in Figures 5 and 6 are provided in Supplementary Table 8.
In all, 17 perturbations of the system, representing experimental mutants, are studied and modeled by modifying parameters or by adding reactions to the system (Supplementary information). All perturbations are simulated on all the 229 wild‐type equilibrium scenarios.
The robustness of the model is assessed using a local parameter sensitivity analysis and in a simulation including cell growth and division (Supplementary information).
Finally, a host of sub‐models missing individual regulatory interactions are optimized showing that the model is the minimal model able to reproduce the wild type and perturbations observed in the biological system. Details of the differential equations model, optimization algorithm and model perturbations are given in Supplementary information.
All microarray data used in this study are deposited at http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=fpixfyiwisgumby&acc=GSE29364 (accession number GSE29364).
All simulations were performed using in‐house developed software (http://dev.thep.lu.se/organism) and a zip archive containing source code, model files, tissue templates, and all sets of optimized parameter values are available as Supplementary Material or from the web page http://www.thep.lu.se/∼henrik/MSB2013/.
Live imaging work was done at microscopy core facility of Center for Plant Cell Biology (CEPCEB) and genomics work was done at Institute of Integrative Genome Biology (IIGB), UCR. We thank Yuval Eshed for 6XoP:17444292:KAN1, Thomas Laux for pWUS:17444292:LhG4 transgenic lines and Patricia Springer for 35S:17444292:KAN1‐GR transgenic lines and also for comments on manuscript. This work was supported by NSF grant (IOS‐1147250) to GVR, and by the Swedish Research Council and the Gatsby Charitable Foundation (HJ). MH acknowledges the Australian Research Council for present funding.
Author contributions: RKY, MP and GVR conceived research and designed experiments. RKY and MP performed experiments. JG and HJ developed computational models and performed simulations. RKY, MP, JG, TG, HJ and GVR analyzed the data. CO and MH developed KAN1:GFP reporter line used in live imaging experiments. TG helped with microarray analysis. RKY, MP, JG, HJ and GVR wrote the paper. All authors edited and approved the final version of the manuscript.
Conflict of Interest
The authors declare that they have no conflict of interest.
Supplementary text describing computational methods together with Supplementary Figures
Software for Computational Methods
This archive contains the simulation tool and scripts for running perturbations etc. for the computational work. When unzipping the archive a directory MSB_WCK is created. Instructions on how to use the software can be found in the README file in the directory
Supplementary Table 1
DEG Analysis of WUS‐GR responses mapped to three cell types. In this table columns E and F represent DEX / MOCK set, columns G and H represent DEXCYC / CYC set for 35S:17444292:WUS‐GR ap1‐1;cal1‐1 DEG analysis. Columns I‐P display 3 SAM cell samples (CLV3p, FILp and WUSp) including the comparison of protoplasted ap1‐1;cal1‐1 SAMs (IND) with non‐protoplasted SAMs (UNIND). All samples shows log2 fold change values and the adjusted p‐value (FDR) of the LIMMA package (e.g. DEX Mock 4hrs fold change and Dex Mock 4hrs adj p‐value). The GCRMA normalized expression values from Supplementary Table 5 were used for the DEG analysis. The column D provide 2 fold up regulated DEGs relative to 2 other samples for cell types. An overlap of WUS‐GR data set with cell type data set is shown in Figure 2A‐D. In addition, the table contains in column Q the detected genes for CLV3p, FILp and WUSp cell samples, respectively. For this MAS5 algorithm was used to make PMA calls, a genes scored as P in all the replicate was considered present and counted once (Supplementary Table 6).
Supplementary Table 2
WUS‐GR up and down regulated transcripts in the presence of Dex. Down regulated genes were sorted from Supplementary Table 1 after applying (≤ 2 fold; p > 0.01)
Supplementary Table 3
WUS‐GR up and down regulated transcripts in Dex + cycloheximide. Up regulated genes were sorted from Supplementary Table 1 after applying (= 2 fold; p > 0.01)
Supplementary Table 4
GO Term Enrichment for WUS regulated transcriptome. The table provides the GO term enrichment data for WUS regulated transcripts from Supplementary Table 3. The columns provide WUS up and down regulated set identifiers (DEG Sets), their sample sizes (Sample SZ), the GO identifier, the number of genes in the genome associated with a GO term (Node Size), the number of genes in the test sample associated with a GO term (Sample Match), the p‐value of the hypergeometric distribution test (P‐value), the Bonferroni corrected version of this p‐value (P‐value adjusted), the GO term, the Ontology type, and the Arabidopsis genes associated with a given GO term in the test sample (Sample AGIs)
Supplementary Table 5
GCRMA Normalized Expression Values. The table provides GCRMA normalized expression values of CYC (cycloheximide), DEX (dexamethasone), DEXCYC (dexamethasone plus cycloheximide) and MOCK treated 35S:17444292:WUS‐GR ap1‐1;cal1‐1 in columns (D‐O). Three cell types (CLV3p, WUSp and FILp) GCRMA normalized expression values presented in columns (P‐V). In addition, GCRMA normalized values for protoplasting induced and un induced replicates also provided in columns (X‐AA)
Supplementary Table 6
MAS5 Normalized Expression Values with Present Call Information. This table provides mean of MAS5 normalized expression values for three cell types with PMA calls. The present call information (PMA values) from the Wilcoxon signed rank test of the MAS5 algorithm is provided for the cell types in column (e.g. CLV3p PMA calls). The probe set showing present calls in all the replicate of a sample was considered as positive
Supplementary Table 7
List of primers used in this study
Supplementary Movie 1
Supplementary Movie 2
This is an open‐access article distributed under the terms of the Creative Commons Attribution License, which permits distribution, and reproduction in any medium, provided the original author and source are credited. This license does not permit commercial exploitation without specific permission.
- Copyright © 2013 EMBO and Macmillan Publishers Limited