Shaping of root architecture is a quintessential developmental response that involves the concerted action of many different cell types, is highly dynamic and underpins root plasticity. To determine to what extent the environmental regulation of lateral root development is a product of cell type preferential activities, we tracked transcriptomic responses to two different treatments that both change root development in Arabidopsis thaliana, at an unprecedented level of temporal detail. We found that individual transcripts are expressed with a very high degree of temporal and spatial specificity, yet biological processes are commonly regulated, in a mechanism we term response non-redundancy. Using causative gene network inference to compare the genes regulated in different cell types and during responses to nitrogen and a biotic interaction we found that common transcriptional modules often regulate the same gene families, but control different individual members of these families, specific to response and cell type. This reinforces that the activity of a gene cannot be defined simply as molecular function; rather, it is a consequence of spatial location, expression timing and environmental responsiveness.
Supplemental Dataset 1
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961.Supplemental Dataset 1. RMA-normalised Nimblegen microarray data for all transcripts measured. The table lists the TAIR10 Arabidopsis Genome Initiative (AGI) gene IDs represented on the array (for design see GEO record GPL18735 for Nimblegen probe design) and their expression values in all 6 time series, averaged (“Mean”) for each replicate set; see GEO GSE91379 for complete raw and normalised individual replicate values. Gene symbols and gene descriptions are listed according to the TAIR10 annotation. If significantly differentially expressed within a time series (BATS), the cluster number is listed. If significantly differentially expressed between treated and untreated time series (CN GP2S etc), the cluster number is listed. Cluster numbers described in the manuscript text always refer to the within-CU/PU or N/Rhizobia vs. U clusters (orange columns). Empty cells indicate no evidence of DE within/between time series. Transcripts associated with genes that have previously been found to be affected by the protoplast generation treatment or FACS (as [18]) are marked (Proto-flagged) but not removed from the analysis.
Supplemental Dataset 2
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961.Supplemental Dataset 2. Cluster designations for differentially expressed transcripts. The table lists the TAIR10 AGI IDs and descriptions (as Supplemental Dataset 1), whether these are classed as transcriptional regulators for network analysis (‘TF category’) and the cluster numbers for DE transcripts. In addition, transcripts that are differentially expressed in both CU and PU (‘Core CU/PU’), are ‘N-response’ marker genes, clock genes or are CTSEs in cortex or pericycle are indicated (as referred to in the text).
Supplemental Dataset 3
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961.
Supplemental Dataset 3. Phenotypic analysis. (A) Average (ave) and standard error (SE) root trait values measured from 9 day old Col0 seedlings and for the same seedlings after 4 days of N, rhizobial or control/mock treatment with T-test values comparing phenotype values (bold=P<0.01). (B-C) Average (ave) and standard error (SE) root trait values measured from 12 day old wrky15 (B) and nlp8 (C) mutant plants on replete and deplete N with T-test values comparing phenotype values in mutant and silbling WT (bold=P<0.01); LR = lateral root, PR = primary root. (D) Average and SE (n=3 biological replicates) of qPCR measurement, qPCR primer sequences and qPCR product lengths.
Supplemental Dataset 4
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961. Supplemental Dataset 4. GO term overrepresentation analysis. GO terms (biological process/molecular function/cellular component) found to be overrepresented within CN/PN/CR/PR clusters (according to GOrilla [79] with hypergeometric test correction) or within core N-response groups (Figure 6C) (according to BioMaps implemented in VirtualPlant [80] using a Fisher exact test with FDR correction) are listed, together with the corrected P value
and gene membership; x=number of genes with GO term in test cluster/group; X=total number of genes in test cluster/group; n=number of genes with GO term in background; N=total number of genes in background.
Supplemental Dataset 5
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961. Supplemental Dataset 5. Network edge designations with validation information. For each causal regulatory network (A) CU, (B) CN, (C) CR, (D) PU, (E) PN, (F) PR, edges between regulator and target are listed with their posterior probability (pp), regulation type (induce or repress), and whether they are validated according to the O'Malley et al 2016 ampDAP-Seq data or DAP-seq data [42] (1=validated, 0=not validated, NT=not tested).
Supplemental Dataset 6
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961. Supplemental Dataset 6. Network statistics and module annotation and characteristics .(A) For each causal regulatory network 8 statistics are shown: number of edges, number of nodes, the node/edge ratio, number of connected TFs, number of targets of those TFs (including TFs that are targets), average targets per TF, the network clustering coefficient and the characteristic path length as given in Cytoscape. (B) Comparison of TF module size and timing response category for CN and PN networks. (C) Network statistics for major modules in each experiment are shown with a cut off of >=10 targets in at least one experiment. The status of each TF in the network in each experiment is shown (DE (0 not DE, 1 DE), indegree, outdegree). The combined outdegree is the sum of the shown out degrees across the six experiments, the number of experiments where outdegree is >7 is also shown. Maximum module size was in part dictated by the number of genes in the network, while the mean number of targets per TF was larger in CU and CN. This indicates that the pericycle and cortex networks have statistically different structures and this was confirmed by analysis of the network connectivity using size independent statistics; specifically there were differing levels of clustering, CN was the least clustered whereas PN had the longest path length while still being highly clustered. This indicates that the cortex is dominated by a few large modules, while the pericycle is more distributed amongst a larger number of modules, in fact extending to very large sizes. (D) Corroborated interactions based on the presence of cis-acting TF family binding sites for networks. Data derived from Franco-Zorrilla et al 2014 [83]. Source contains TFs that interact with a putative regulated gene (Target) and the type of interaction (activation/inhibition) is shown.
Supplemental Dataset 7
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961. Supplemental Dataset 7. Glutathione S-transferase gene family expression is highly variable. All members of the GST family the time series within which differential expression was found are indicated (or not DE), together with the cluster number for that time series. Genes shown on Figure 6 are highlighted in yellow and the time series that has been shown is indicated.
Supplemental Dataset 8
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961. Supplemental Dataset 8. Gene family regulation across cell types and treatments. (A) Summary of gene family regulation: number DE/non-DE, in which timeseries they are regulated and if transcripts are DE in one or both cell types. (B) Lists of the genes in each family together with cluster numbers if differentially expressed in a timeseries.
Supplemental Dataset 9
Supplemental Data. Walker et al. (2017). Plant Cell 10.1105/tpc.16.00961.Supplemental Dataset 9. Gene model analysis. 1547 differentially expressed transcripts that measure the expression of 711 genes, with each gene represented between 2-6 times. Cluster numbers, the transcript count per gene and a summary of whether expression of transcripts are “identical”, ”similar” or “different” based on the location, response and timing of differential expression for each gene are given (see text for explanation of analysis).
Supplemental File 1
CU full network session
Supplemental File 2
CN full network session
Supplemental File 3
CR full network session
Supplemental File 4
PU full network session
Supplemental File 5
PN full network session
Supplemental File 6
PR full network session
Supplemental File 7
CU backbone network session
Supplemental File 8
CN backbone network session
Supplemental File 9
CR backbone network session
Supplemental File 10
PU backbone network session
Supplemental File 11
PN backbone network session
Supplemental File 12
PR backbone network session
Supplemental File 13
CN backbone network session with shared TFs in middle
Supplemental File 14
PN backbone network session with shared TFs in middle
