|Home | About | Journals | Submit | Contact Us | Français|
Author Information The National Center for Biotechnology Information Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/) accession numbers for the array data discussed in this manuscript are GSE15876 and GSE21338. Reprints and permissions information is available at www.nature.com/reprints. The authors declare no competing financial interests. Readers are welcome to comment on the online version of this article at www.nature.com/nature. Correspondence and requests for materials should be addressed to P.N.B. (email@example.com).
The development of multicellular organisms relies on the coordinated control of cell divisions leading to proper patterning and growth1–3. The molecular mechanisms underlying pattern formation, particularly the regulation of formative cell divisions, remain poorly understood. In Arabidopsis, formative divisions generating the root ground tissue are controlled by SHORTROOT (SHR) and SCARECROW (SCR)4–6. Here we show, using cell-type-specific transcriptional effects of SHR and SCR combined with data from chromatin immunoprecipitation-based microarray experiments, that SHR regulates the spatiotemporal activation of specific genes involved in cell division. Coincident with the onset of a specific formative division, SHR and SCR directly activate a D-type cyclin; furthermore, altering the expression of this cyclin resulted in formative division defects. Our results indicate that proper pattern formation is achieved through transcriptional regulation of specific cell-cycle genes in a cell-type- and developmental-stage-specific context. Taken together, we provide evidence for a direct link between developmental regulators, specific components of the cell-cycle machinery and organ patterning.
Growth and patterning are key processes that govern the development of multicellular organisms. In some cases, like early Drosophila embryogenesis7, these are independent. However, in many animals and plants, proper development frequently relies on tight coordination of growth and patterning. Disruption of this coordination can lead to unchecked cell growth, resulting in tumorigenesis or misshapen organs8. Although the molecular mechanisms involved in pattern formation9–11 and in cell-cycle control12–15 have been well characterized independently, few connections have been made between the two16,17. Here, we identify a new transcriptional regulatory relationship between two key regulators of organ patterning and specific components of the cell-cycle machinery.
SHR and SCR transcription factors control ground tissue patterning by regulating the formative asymmetric division in the immediate progeny of the ground tissue stem cells, known as cortex/endodermis initial (CEI) cells18,19. SHR acts non-cell autonomously20–22 to activate the expression of SCR in the quiescent centre, CEI and endodermis5, 6, 23,24,25. To gain insight into the role of the SHR/SCR network in controlling formative cell divisions, we expressed an inducible version of either SHR or SCR in its respective mutant background and characterized the timing of formative divisions after induction.
Before induction, SHR- and SCR-inducible plants had a single mutant ground tissue layer4,5 (Supplementary Fig. 1). After SHR induction, SCR expression was observed within 3 h, indicating that SHR rapidly activates its targets (Supplementary Fig. 1). The first periclinal (parallel to the direction of growth) division in the mutant ground tissue layer occurred 6 h after SHR induction (Fig. 1a, Supplementary Fig. 1 and Supplementary Movie 1) and earlier after SCR induction (Fig. 1b). Two layers of ground tissue with SCR expression in the quiescent centre and endodermis (Supplementary Fig. 1), along with a nearly complete Casparian band5, were detected 24 h after SHR induction (Supplementary Fig. 2). This underlines the combinatorial role of SHR and SCR in regulating formative cell divisions and also indicates that the two inducible systems have slightly different kinetics.
To understand the dynamics of the SHR/SCR regulatory network, we sorted ground tissue cells at several time points after SHR and SCR induction and performed microarray analysis on RNA from the sorted cells26. We found 2,478 and 1,903 differentially expressed genes throughout the SHR and SCR time courses, respectively (Fig. 1d, Supplementary Tables 1 and 2). Seven of the eight previously identified SHR direct target genes24,25 were differentially regulated in the SHR time course (Supplementary Fig. 3) and most of them were also identified in the SCR time course (Supplementary Fig. 3). For both inducible systems, the results for a subset of regulated genes were independently confirmed by quantitative PCR with reverse transcription (RT–qPCR) (Supplementary Fig. 3).
We identified 12 clusters of co-expressed genes downstream of either SHR or SCR (Supplementary Fig. 4). In both SHR and SCR time courses, known SHR targets were found in what we named the ‘SCR cluster’, as it included SCR (Supplementary Fig. 5). Notably, in the SCR-inducible system the expression level of these genes peaked at 3 h as compared to 6 h in the SHR-inducible system. The observed timing of the first formative division followed a similar progression. We therefore propose that genes in the SCR cluster may have a role in regulating formative cell divisions.
In each of the 12 clusters, we identified significantly enriched gene ontology categories (Supplementary Fig. 6, Supplementary Tables 3 and 4). In clusters activated 1 h after SHR (cluster I) and SCR induction (clusters IV and V) (Supplementary Fig. 4), transcription factors were significantly overrepresented. At 6 h after induction and coincident with the onset of formative divisions, gene ontology categories associated with cell-cycle progression and cyclin-dependent protein kinase (CDK) activity were overrepresented in both inducible systems (Supplementary Fig. 6).
To examine the role of the SHR/SCR network in ground tissue formative divisions, we identified genes that were differentially expressed in both inducible systems. In total, 860 genes were regulated by both SHR and SCR (Fig. 1d and Supplementary Table 5). We identified six clusters of co-expressed genes (Supplementary Fig. 7), and gene ontology category enrichment analysis showed predominant molecular functions of these clusters (Fig. 1e and Supplementary Table 6). Interestingly, one cluster (V) contained genes related to cell division and CDK activity. This indicates that SHR and SCR jointly regulate formative cell divisions by regulating cell-cycle machinery components.
To identify direct and indirect target genes of SHR, we performed chromatin immunoprecipitation-based microarray (ChIP-chip) experiments. We identified 266 genes as direct targets of SHR (Supplementary Table 7). Of these, 65 were differentially expressed in the SHR-inducible system (Fig. 2a and Supplementary Table 8). These genes were positively regulated by SHR and were found primarily in two major expression profiles, with activation at either 1 h or 6 h after SHR induction (Supplementary Fig. 8). The proportion of transcription factor genes was significantly higher than expected by chance (Fig. 2b and Supplementary Table 9), suggesting that SHR directly activates a regulatory cascade involved in several biological processes including cell-cycle progression. Coincident with the onset of periclinal divisions, we identified a D-type cyclin, CYCD6;1, that was regulated and whose promoter was directly bound by SHR (Fig. 2c). This gene was one of only six SHR direct target genes in the SCR cluster (Fig. 2d and Supplementary Fig. 9). To confirm direct binding of SHR to the CYCD6;1 promoter and determine whether SCR also binds this target, we performed ChIP–qPCR. We found significant enrichment for both SHR and SCR binding around 1 kilo-base (kb) upstream of the CYCD6;1 promoter (Supplementary Fig. 10). The temporal pattern of activation of CYCD6;1 in the SHR- and SCR-inducible systems, the binding of SHR and SCR to its promoter and the known protein–protein interaction of SHR and SCR25 support regulation of CYCD6;1 by both SHR and SCR.
To determine if CYCD6;1 expression is consistent with a role in the formative division of the immediate progeny of the CEI, the CEI-daughter cell, we generated a transcriptional reporter for CYCD6;1 (pCYCD6;1::GFP). We observed its expression specifically in CEI/CEI-daughter cells (Fig. 3a); however, we rarely detected expression in all eight CEI/CEI-daughter cells simultaneously, indicating that these formative divisions are not tightly synchronized. In cycd6;1 mutant seedlings (Supplementary Fig. 11), CEI-daughter cells showed significantly fewer formative divisions (Fig. 3b). Nevertheless, cortical cell number and overall meristem length were comparable to wild type (Supplementary Fig. 12). These results indicate that CYCD6;1 is not required for proliferative cell divisions but is specifically involved in formative divisions needed for proper ground tissue patterning. Because CYCD6;1 is also expressed in embryonic CEI-daughter cells (Fig. 3a), we examined ground tissue formative divisions during embryogenesis. cycd6;1 CEI-daughter cells showed significantly fewer periclinal divisions than wild type at both early heart and torpedo embryo stages, whereas all other embryonic root divisions appeared normal (Fig. 3c, d). Interestingly, by the mature embryo stage the number of periclinal divisions in cycd6;1 and wild-type CEI-daughter cells was more similar (Fig. 3e). We also observed no difference in the seed germination rate between wild type and cycd6;1 (Supplementary Fig. 13 and Supplementary Movie 2). These results indicate that the cycd6;1 phenotype is specific to a small number of formative divisions and developmental stages.
Later in root development, CYCD6;1 expression is restricted to a subset of endodermal cells (Fig. 3f) that undergo a second formative cell division to form middle cortex27. In wild type, middle cortex divisions were observed in 52% of roots, as compared to 12% in cycd6;1 (data not shown). In older plants, expression of pCYCD6;1::GFP was detected in lateral root primordia (Fig. 3g) and pericycle and phloem cells (data not shown). Each of these tissues undergoes formative divisions and all are within the SHR and/or SCR functional domains17,24.
Because the cycd6;1 and shr phenotypes are not identical, we proposed that there is functional redundancy in the regulation of this formative division. To test this, we analysed mutations in CYCD2;1 and CYCD5;1, other D-type cyclins downstream of SHR (Supplementary Table 1). We did not detect abnormal CEI-daughter divisions in either mutant. The double mutants, cycd6;1cycd2;1 and cycd6;1cycd5;1 (Supplementary Fig. 14), were not significantly different from cycd6;1, suggesting a higher level of functional redundancy, which may include other CYCD genes. Taken together, we provide evidence that the SHR/SCR transcriptional network exerts finely tuned control of the formative divisions within the CEI through the restricted spatiotemporal activation of a specific cell-cycle gene.
To identify further genes that participate in this formative division, we profiled transcripts expressed in CEI/CEI-daughter cells by sorting cells from plants containing pCYCD6;1::GFP. Of genes highly expressed in CEI/CEI-daughter cells (Supplementary Table 10), 234 were also differentially regulated in the SHR- and SCR-inducible systems. Notably, 32 showed G2/M-phase-specific expression during cell-cycle progression28 (Supplementary Fig. 15). On the basis of root-cell-type microarray expression data29, these mitotic genes are expressed in the same domain as CYCD6;1 (Fig. 4a), indicating that they are involved in tissue-specific cell divisions. Whether they are involved specifically in formative divisions or more generally in proliferative divisions remains an open question.
Recently, genes involved in formative cell divisions during lateral root formation have been identified17, 12 of which are regulated by SHR and SCR and expressed in CEI cells (Fig. 4b and Supplementary Table 11). Among these are two CDKs, CDKB2;1 and CDKB2;2, indicating that SHR/SCR regulation of these genes may have a role in formative divisions of ground tissue. To test this hypothesis, we ectopically expressed CDKB2;1 and CDKB2;2, as well as CYCD6;1, specifically in the ground tissue. Although root length in these lines was comparable to wild type, we observed extra formative divisions in endodermal cells (Fig. 4c, Supplementary Figs 16 and 17). We also ectopically expressed CYCD5;1, CYCD2;1 and CYCD1;1 and found no alterations in patterning, indicating that not every D-type cyclin is able to stimulate this formative division (Supplementary Fig. 18). Interestingly, CDKB2;1 and CDKB2;2 were not identified as direct targets of SHR, suggesting that they may be regulated by transcription factors activated 1 h after SHR and SCR induction. Taken together, our results indicate that CDKB2;1, CDKB2;2 and CYCD6;1 are involved in specific formative cell divisions downstream of the SHR/SCR network.
To assess whether CYCD6;1 and CDKB2;1 could function independently of SHR in promoting formative divisions, we ectopically expressed them in the shr mutant ground tissue. Both CYCD6;1 and CDKB2;1 were able to partially complement the shr formative division phenotype (27% and 25% of plants showed divisions, respectively) (Fig. 4d), whereas root length was comparable to shr-2-J0571 (Fig. 4e). This indicates that these genes have important roles in generating formative divisions downstream of SHR but that there are further genes involved in this process.
We identified a component of the cell-cycle machinery, CYCD6;1, as a key SHR/SCR downstream target, providing the first evidence of a direct molecular link between these key developmental regulators and a cell-cycle gene. We further showed that CYCD6;1 is expressed specifically at the time of the formative cell divisions regulated by SHR/SCR and that altering its expression results in formative division changes in both loss- and gain-of-function plants. Thus, tight spatiotemporal regulation of specific cell-cycle genes is required for proper root pattern formation.
The Arabidopsis thaliana ecotypes Columbia (Col-0), Wassilewskija (WS) and C24 were used. Origin and ecotypes are as follows: shr-2 and pSHR::SHR:GFP shr-2 (Col-0); pSHR::SHR:GR shr-2 pSCR::GFP (Col-0); scr-4 (WS), J0571 enhancer trap (C24). The T-DNA insertion line for (At4g03270) cycd6;1 is from the GABI-Kat line (GK-368E07).
To characterize the SHR- and SCR-inducible lines, time-lapse laser scanning confocal microscopy (Zeiss LSM 510) was used. Confocal microscopy was used for observation of embryos and roots of 5- to 15-day-old plants stained with 10 μM propidium iodide. For the cycd6;1 phenotypic analyses, we counted the number of CEI cells in wild-type and cycd6;1 mutants and transverse confocal sections were taken just above the quiescent centre cells to determine if the ground tissue formative division had occurred within the CEI daughter.
To understand the dynamics of the SHR/SCR regulatory network, we performed time-course microarrays at 0, 1, 3, 6 and 12 h after SHR and SCR induction. Moreover, to examine the transcriptional effects specific to the ground tissue, we sorted the GFP-positive cells from the J0571 enhancer trap30 present in both inducible lines. For the SHR/SCR expression time-course microarrays and CEI-specific expression experiments, root tips were dissected at approximately 0.5 cm from the root tip. The protoplasts obtained from these dissected J0571 or pCYCD6;1::GFP lines were used for fluorescence-activated cell sorting26. Two biological replicates were performed for each experiment. Correlations between biological replicates were higher then 0.92 R2.
To determine whether genes regulated by SHR are direct or indirect targets, we performed ChIP-chip. We used the SHR time-course microarray data to determine the detection settings that would allow us to identify SHR direct targets with maximized specificity to sensitivity.
We thank D. Orlando and R. Twigg for generating the list of probes for the ChIP-array; J. Nieuwland, S. Maughan and C. Collins for construction of the CYCD6;1 reporter line and assistance in isolation of the cycd6;1 mutant; L. Sanz, F. Patell and S. Scofield for isolation of the cycd2;1 and cycd5;1 mutants; J. Dinneny, M. Noor and members of the Benfey laboratory for their comments on the manuscript. Funding to M.A.M.-R. is provided by the Ministerio de Ciencia y Innovacion (Spain). W.D. and J.A.H.M. were funded by a Biotechnology and Biological Sciences Research Council grant (BB/E022383) and the European Research Area in Plant Genomics network on Plant Stem Cells (BB/E024858). This work was funded by grants to P.N.B. from the NIH (RO1-GM043778 and P50-GM081883) and from the NSF (AT2010 0618304).
The Arabidopsis thaliana ecotypes Columbia (Col-0), Wassilewskija (WS) and C24 were used. Origin and ecotypes are as follows: shr-2 and pSHR::SHR:GFP shr-2 (Col-0)20; pSHR::SHR:GR shr-2 pSCR::GFP (Col-0)24; scr-4 (WS)31; J0571 enhancer trap (C24) (http://www.plantsci.cam.ac.uk/Haseloff/Home.html). Transfer DNA insertion lines: cycd6;1 (At4g03270) is from the GABI-Kat line described in NASC (GK-368E07); cycd2;1 (At2g22490) and cycd5;1 (At4g37630) are from NASC (SALK 049449 and FLAG 148D11, respectively). Primers listed in Supplementary Table 13a were used to verify that cycd6;1, cycd5;1 and cycd2;1 were homozygous lines. To check CYCD6;1 expression in cycd6;1, the primers listed in Supplementary Table 13b were used.
After surface sterilization using 50% bleach and 0.1% Tween for 5 min and then rinsing three times with sterile water, seeds were stratified at 4 °C for 2 days before being planted on standard media. Seeds were sown and allowed to germinate on vertically positioned Murashige and Skoog agar plates in a Percival incubator with 16 h daily illumination at 22 °C. For the plants grown for the microarray experiments, a layer of nylon mesh (Nitex Cat 03-100/44, Sefar) was added on top of the agar to facilitate transfer onto treatment media and/or root dissection.
Standard molecular biology techniques and the Gateway system (Invitrogen) were used for the cloning procedures. The pSHR::SHR:GR shr-2-J0571 line was obtained by crossing shr-2 pSHR::SHR:GR T1 A1/124 to the J0571 enhancer trap line30. The pSHR::SHR:GR shr-2 pSCR::GFP line was obtained by crossing shr-2 pSHR::SHR:GR T1 A1/1 to the SCR reporter (pSCR2.0)24.To generate pSCR::SCR:GR, pSCR2.0 was used as a promoter region driving the SCR coding sequence and glucocorticoid receptor (GR). The pSCR::SCR:GR scr-4-J0571 line was obtained by transforming pSCR::SCR:GR into scr-4-J0571 (F4 homozygous lines). Genomic DNA from Arabidopsis ecotype Columbia was used as the template for amplification of the upstream regulatory promoter sequence for pCYCD6;1 (2,738 base pairs (bp)) using the cpromCYCD6;1 primers (Supplementary Table 13c) and the PCR product was then cloned into pENTR/D-TOPO. The promoter–reporter pCYCD6;1::GFP:GUS was cloned into the pKGWFS7-Dx destination vector. The coding sequences of CDKB2;1 (At1g76540), CDKB2;2 (At1g20930), SYP111/KNOLLE (At1g08560), CYCD1;1 (At1g70210), CYCD2;1 (At2g22490) and CYCD5;1 (At4g37630), obtained from The Arabidopsis Information Resource (TAIR)—U12632, U83912, C105142, U60123, U82523 and U66230, respectively—and CYCD6;1, were cloned in a pENTR/D-TOPO using the gene coding sequence (cds) amplified with primers as in Supplementary Table 13c. These constructs were then recombined in a two-component system where GAL4 is expressed in the ground tissue (line J0571)30, resulting in expression of the gene of interest, as shown by YFP expression, specifically in the ground tissue. The plasmids were then transformed in the wild-type Col-0 or C24 J0571 enhancer trap by the floral dip method32.
The inducible version of either SHR or SCR, expressed in its respective mutant background, was characterized using time-lapse laser scanning confocal microscopy. We morphologically characterized pSHR::SHR:GR shr-2-pSCR::GFP, the reporter of which is expressed in the quiescent centre, CEIs and endodermis24, pSHR::SHR:GR shr-2-J0571 and the pSCR::SCR:GR scr-4-J0571 enhancer trap, which specifically marks the ground tissue. Z stacks were taken to locate formative divisions in medial longitudinal sections of SHR- and SCR-induced plants.
Laser scanning confocal microscopy (Zeiss LSM 510) was used to examine embryos and roots of 5- to 15-day-old plants stained with 10 μM propidium iodide.
Fixation, staining and analysis of mature embryos were conducted as described33.
For the cycd6;1 phenotypic analyses, we counted the number of CEI cells in wild-type Col-0 (n = 24 and n = 31 of 4- and 5-day-old roots, respectively) and cycd6;1 mutants (n = 24 and n = 33 of 4- and 5-day-old roots, respectively). Transverse confocal sections were taken just above the quiescent centre cells to determine if formative divisions had occurred in any of the eight CEI-daughter cells. Images of longitudinal confocal sections were analysed to measure the root meristem length and count the number of cortex cells between the quiescent centre and the first elongated cell as described34.
To analyse middle-cortex formation in wild-type Col-0 and cycd6;1 roots, Z stacks were taken to locate formative divisions in endodermal cells. The root growth length was assayed by marking the position of the root tips daily on the back of the plate, beginning when the plants were 4 days old. After 4 consecutive days of marking each root (except for roots with the shr mutation, for which only 2 days were measured), the plates were scanned and the distance between each mark was assayed using ImageJ software.
Multiple plants (n > 50) of individual T3 lines for J0571-UAS::CYCD6;1:YFP (n = 4), J0571-UAS::CDKB2;1:YFP (n = 2) and J0571-UAS::CDKB2;2:YFP (n = 1) were examined and all of these lines showed ectopic formative divisions in the endodermis.
The homozygous line for either J0571-UAS::CYCD6;1:YFP or J0571-UAS::CDKB2;1:YFP was crossed to shr-2-J0571 homozygous plants. Over-expression of CYCD6;1 or CDKB2;1 was able to partially complement the shr formative division phenotype (n = 42 and 47, respectively).
Many plants (n > 20) of individual T2 lines for J0571-UAS::CYCD1;1:YFP (n = 2), J0571-UAS::CYCD2;1:YFP (n = 1) and J0571-UAS::CYCD5;1:YFP (n = 5) were also examined and none of these lines showed further divisions in the endodermis. Standard error was used and a parametric Student's t-test (Figs 3b, ,4e)4e) and a non-parametric Wilcoxon test (Fig. 3c–e) were applied for the statistical analysis.
The Casparian band staining was performed on roots from 5- and 7-day-old plants as previously described24, using 1% agarose sections.
ChIP was conducted as previously described25 on roots of 5-day-old seedlings expressing pSHR::SHR:GFP in shr-2 and pSCR::SCR:GFP in scr-4. Immunoprecipitation was performed using a rabbit polyclonal antibody to GFP (ab290, Abcam). For the ChIP–qPCR, enrichment of putative target promoter-region DNA was determined using RT–qPCR and a qPCR efficiency of twofold amplifications per cycle was assumed. Immuno-precipitation from pSHR::SHR:GFP and pSCR::SCR:GFP was compared to mock DNA (bovine serum albumin was used instead of anti-GFP). To assay the in vivo binding of SHR and SCR to the promoter region of the CYCD6;1 gene using ChIP–qPCR, primers were designed to amplify 100–500 bp regions that overlap to span the 4 kb promoter length of the gene. The 500 bp resolution that we obtained in our ChIP–qPCR experiments was owing to technical constraints of primers and materials. The POLYUBIQUITIN 10 gene (At4g05320) was used to normalize the results between samples (Supplementary Table 13b). Tiling along the CYCD6;1 promoter region was done using the primers listed in Supplementary Table 13d and was designed to be in ascending order upstream from the CYCD6;1 ATG. The smaller peak that is visible only in the ChIP–qPCR data could have arisen because of many technical reasons that include variations of material and fixation conditions.
For the genome-wide identification of direct targets of SHR, a ChIP-chip method was developed. First, ChIP was performed on the roots of 5-day-old seedlings expressing the pSHR::SHR:GFP in shr-225, except that DNA-free protein A agarose beads (Invitrogen) were used. DNA from the ChIP and mock experiments was individually amplified using a random-primer-based genome amplification method (http://cat.ucsf.edu/pdfs/22_Round_A_B_C_protocol.pdf) with minor modifications. After labelling with Cy3 and Cy5, respectively, following an amino-allyl-dye coupling protocol (http://camd.bio.indiana.edu/files/amino-allyl-protocol.pdf), the DNA was cleaned up using the PCR purification kit (Qiagen) and 3 μg each from the ChIP and mock samples were taken and mixed for hybridization to a custom long oligonucleotide (~60 bases) Arabidopsis promoter microarray. The promoter micorarray was fabricated using the Agilent SurePrint technology, and has a capacity of 244,000 features. For each gene in the Arabidopsis genome, probes were designed for the intergenic region, the first intron, the 5′ untranslated region and the 3′ untranslated region sequence. The probes were evenly distributed with a density of 125 bp per probe for genes encoding transcription factors and 250 bp per probe for other genes. To identify microRNA gene targets, probes were separately designed for the promoter sequences of about 110 known miRNA genes (http://asrp.cgrb.oregonstate.edu/). Hybridization was performed according to the Agilent ChIP-chip protocol, and images were obtained using an Agilent microarray scanner (model G2565BA) at a resolution of 5 μm. Signal extraction and initial data processing were done using the Agilent feature extraction software. Two biological replicates were conducted.
For the SHR/SCR expression time-course microarrays (at 0, 1, 3, 6 and 12 h after SHR and SCR induction) and CEI-specific expression experiments, root tips were dissected at approximately 0.5 cm from the root tip. The protoplasts obtained from these dissected J0571 or pCYCD6;1::GFP lines were used for fluorescence-activated cell sorting26.The GFP-positive sorted cell populations were subsequently frozen before RNA was extracted and used both for RT–qPCR analysis and for microarray analysis using a two-cycle Affymetrix amplification protocol. Total RNA isolation and probes for hybridization were prepared as described29. Two biological replicates were performed for each experiment when the correlations between biological replicates were higher than R2 = 0.92. R2 values were as follows: T0_SHR: 1vs.2 (0.950312295); T1_SHR: 1vs.2 (0.972531048); T2_SHR 1vs.2 (0.955786); T3_SHR: 1vs.2 (0.96980011); T4_SHR: 1vs.2 (0.957667748); T0_SCR: 1vs.2 (0.972756742); T1_SCR:1vs.2 (0.955672139); T2_SCR: 1vs.2 (0.972798562); T3_SCR: 1vs.2 (0.922900261); T4_SCR:1vs.2 (0.976237166); and CEI 1vs2 (0.977861208). A global normalization step was applied to the SHR- and SCR-induction time course using gcRMA. The gcRMA implementation in the Bioconductor and R software packages were used for background correction, quantile normalization and expression-estimate computation (http://www.bioconductor.org)35. False-discovery rates were calculated by the significance analysis of microarrays (SAM) algorithm36. Differentially expressed genes were identified based on fold-change in expression and a significant P-value threshold (twofold change and P value < 0.05). Figure of merit (FOM) led us to identify twelve clusters in both SHR and SCR time courses and six clusters for the common differentially regulated genes. K-means clustering, using a Pearson correlation as the distance metric, was used to group the co-expressed genes downstream of SHR and SCR. For the CEI sorted cell microarrays, mixed model software was used to globally normalize all arrays and to identify differentially expressed probesets24. Criteria used to determine enrichment of gene expression in a cell type are as described29. SCR5 and J0571 GFP marker lines were excluded from our enrichment analysis because their expression overlapped with the pCYCD6;1::GFP expression domain in the CEIs and CEI-daughter cells. Gene ontology enrichment categories were found using ChipEnrich software37 (http://www.arexdb.org/software.jsp). Heatmaps were obtained using TMeV freeware (http://www.tm4.org/).
Primers for RT–qPCR on the known targets of SHR were designed to be specific to their coding sequences and are listed in Supplementary Table 13b. As an endogenous control, we used a gene that did not show differential expression in either SHR or SCR microarrays, At3g52240 (Supplementary Table 13b).
Genes induced by dexamethasone alone were not included in the list of SHR and SCR differentially regulated genes (Supplementary Table 12). The Venn diagrams shown were obtained using VENNY, an interactive tool for comparing lists with Venn Diagrams (http://bioinfogp.cnb.csic.es/tools/venny/index.html). The gene identities were obtained using VirtualPlant 1.0 (http://virtualplant.bio.nyu.edu/cgi-bin/vpweb2/virtualplant.cgi). The expression microarrays and the ChIP-chip data have been deposited in the public functional genomics data repository Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo/).
For ChIP-chip analysis38, to assess genome-wide binding the processed signal ratios of each experiment were randomly sampled 10,000 times and an empirical P value was estimated for each ratio. The empirical P values for each probe of the independent ChIP-chip experiments were multiplied. Enrichment was scored by: (1) the length of DNA regions that were covered by probes below a defined P-value threshold; (2) a local P-value minimum (seed); and (3) the number of nucleotides allowed as gaps within called regions. For each combination of parameters, the detected regions were registered. A gene was assigned to an enriched region if that region was present within 4,000 bp upstream or 300 bp downstream of the transcription start site in an intron, or 300 bp downstream of the gene model. Each parameter combination produced a list of called regions and thus of assigned genes. The proportion of genes that were classified as regulated by SHR (Supplementary Table 1) was recorded for each list. The optimal enrichment detection criteria were defined by those parameters that yielded the highest proportion of SHR-regulated genes and a major fraction of the already described SHR direct genes. These settings were: probe P-value seed, P < 0.0002; probe P value, P < 0.002; minimum length of hybridization below P value, 210 nucleotides; maximum gap, 195 nucleotides. One-million-fold random sampling of lists containing the same number of genes yielded an empirical P value < 10−6. The code for the used Python scripts is available on request.