Regulatory modules controlling maize inflorescence architecture

Abstract

Genetic control of branching is a primary determinant of yield, regulating seed number and harvesting ability, yet little is known about the molecular networks that shape grain-bearing inflorescences of cereal crops. Here, we used the maize (Zea mays) inflorescence to investigate gene networks that modulate determinacy, specifically the decision to allow branch growth. We characterized developmental transitions by associating spatiotemporal expression profiles with morphological changes resulting from genetic perturbations that disrupt steps in a pathway controlling branching. Developmental dynamics of genes targeted in vivo by the transcription factor RAMOSA1, a key regulator of determinacy, revealed potential mechanisms for repressing branches in distinct stem cell populations, including interactions with KNOTTED1, a master regulator of stem cell maintenance. Our results uncover discrete developmental modules that function in determining grass-specific morphology and provide a basis for targeted crop improvement and translation to other cereal crops with comparable inflorescence architectures.

Figure 1.

Molecular signatures of auxin response are detected prior to changes in morphology. (A) Normal progression of axillary meristem initiation in wild-type ears occurs in a developmental gradient from tip to base. SPMs are formed at 1 mm (B), and SMs are formed at 2 mm (C,E). (D) Expression of the DR5-ER∷RFP reporter is strongly polarized to either side of developing SPMs in wild-type ears and these maxima indicate where new SM primordia will form. There is no DR5 signal detected between maxima in wild-type ears (gray arrow). (F) SPMs in ra1 mutants take on a fate similar to indeterminate BMs, reiterating the SPM developmental program, and (G,H,J) do not produce SMs by 2 mm. (I) In ra1 mutants, a weak DR5 signal is observed spanning the central domain of indeterminate SPMs joining the two maxima (white arrow), similar to that observed in tassel BMs. (K–M) In tassels, basal BMs are initiated first before the IM switches to produce determinate SPMs. (N) DR5 expression is observed across the central domain of indeterminate BMs (white arrow), connecting the maxima formed on opposite flanks. (Red asterisks) Determinate spikelet pair meristem (SPM); (green asterisks) indeterminate branch meristem (BM); scale bars, 250 μm in all panels except D (right), I (right), and N, where scale bars = 100 μm; DR5 expression views in D and I are taken from the section in white boxes in C and H, respectively.

Figure 2

Genetic perturbation of the RAMOSA pathway. (A) Differentially expressed (DE) genes and (B) TFs shared among ramosa (ra) mutant ears at 1- and 2-mm stages (corrected P < 0.05). (C) GO enrichment of biological processes for DE genes shared between different mutants at 1 and 2 mm. Venn diagrams above each set of columns are shaded to represent DE genes shared among mutants (from left to right): DE in ra1, ra2, and ra3; DE in ra1 and ra2 only; DE in ra1 and ra3 only; DE in ra1; DE in ra3. (D) Expression changes for DE TF genes in 1-mm ra mutants relative to wild-type siblings. TFs were grouped by family and the number of DE family members is indicated. Each column represents average expression differences across the TF family in a single mutant (ra1, ra2, or ra3); for TFs DE in more than one mutant, individual mutant profiles are shown, but grouped according to shaded area of Venn diagrams above. From left to right, TF expression profiles are shown if DE in ra1, ra2, and ra3; ra1 and ra2 only; ra1 and ra3 only; ra1; ra2; ra3. (Blue-to-red) Up- to down-regulation. (E) Expression profiles for individual members of two TF families: 13 TCP genes were significantly down-regulated in one or more ra mutants at 1 mm and expression changes across 1- to 2-mm stages are shown for all mutants; 12 MADS-box TF family members showed dynamic expression differences in 1-mm ra1, ra2, ra3 mutants, and 1- to 2-mm wild-type tassel primordia. TCP family members are represented by maize Gramene gene ID: GRMZMx₍₈₎; ¹AC199782.5_FG003; ²AC205574.3_FG006; MADS maize gene names or closest ortholog in rice are shown.

Figure 3

Developmental modules for SPM determinacy. (A) Expression signatures across wild-type libraries were used to cluster genes with dynamic expression during ear and/or tassel development. Twenty k-means clusters fell into four distinct clades of expression: enriched in SM/FM; tassel; 1- and 2-mm ear; IM/SPM (from top to bottom). Each cluster is assigned a number identifier (left) and the number of genes associated with each cluster is indicated (right). The heatmap represents cluster centers; (white-to-dark) low-to-high expression. Clusters 8 and 11 were highly enriched for DE genes in 1-mm ra mutants (blue arrows). (B) Genes in cluster 11 and (C) DE in all three mutants at 1 mm were either coordinately up- or down-regulated and (D) genes in cluster 8 and (E) DE in all mutants were almost entirely down-regulated. (F) DE genes in cluster 8 tended to be most strongly down-regulated in ra1 mutants consistent with a more severe phenotype. (G) Co-expressed genes in cluster 8 were also co-expressed across ra mutant backgrounds. Expression profiles are shown for examples of known genes implicated in determinacy and a gene of unknown function with grass-specific lineage. (H) Among these, an ortholog of ROXY (GRMZM2G442791) and (I) a CUC-like NAC TF (GRMZM2G393433) were temporally co-expressed in largely adjacent domains. (J) Of 31 cis-regulatory motifs significantly enriched within proximal promoters of genes co-expressed in cluster 8 and DE in 1-mm ra1 mutants, the 20 with the highest enrichment in this group of genes relative to genome-wide occurrences are shown.

Figure 4.

Genome-wide binding profiles for RAMOSA1. (A) YFP-tagged RA1 is expressed in an adaxial domain subtending SPMs in developing inflorescences, and localized to the nucleus. (B) Distribution of RA1 binding relative to maize gene models showed strong enrichment −1.5 and +1.5 kb from the TSS. (C) Distribution of high-confidence peak summits across genomic features (numbers are based on percent of total). Within a genic region, (up) upstream; (body) gene body; (down) downstream. (D) Bound and modulated targets of RA1 grouped by functional class. (E) RA1 bound genes with known inflorescence phenotypes; zag1, ts2, ct2, and lg1. Examples of overlapping peaks in ear and tassel (zag1 and ts2) and HA- and YFP-tagged libraries (ct2 and lg1) are shown. (F) lg1 was up-regulated in ra1 mutant ears and in wild-type tassel primordia compared with wild-type ears. (G) Immunolocalization of the LG1 protein indicates its absence in wild-type ears (inset image shows determinate SPMs). (H) In ra1 mutant ears, LG1 is localized to the adaxial side of developing branch meristems and (I) in wild-type tassels is localized to the base of long branches. (J) Bound and modulated targets of RA1 were more strongly regulated at 2 mm compared with 1 mm. (K,L) Expression profiles represent cluster centers from Figure 3A: Repressed targets were largely co-expressed in clusters 2, 13, 17, and 18; activated targets were associated with clusters 8, 11, 12, and 19. (M) Analysis of high-confidence RA1 binding sites within gene promoters showed enrichment of de novo motifs: a GAGA-repeat element, a motif similar to the indeterminate1 (id1) binding site (P = 5.6 × 10⁻⁹), novel CAG-box and TG repeat motifs. The latter two were most strongly enriched in promoters of activated genes. (N) Motifs were enriched at specific positions relative to the center of RA1 binding sites.

Figure 5.

Integration of RA1- and KN1-dependent networks. (A) RA1 and KN1 bound 481 shared target genes (189 at the same genomic position), which was greater than expected by chance based on Fisher's test. (B) RA1 bound kn1 in its third regulatory intron. (C) Sixty-five targets were cobound by RA1 and KN1 at the same genomic position and differentially expressed in ra1 and/or kn1 loss-of-function mutants; these genes tended to have stronger dependence on RA1 than KN1 for their normal expression; (green to red) up- to down-regulation; (ln) natural log. (D) Expression profiles for 40 TFs cobound by RA1 and KN1 at overlapping genomic regions showed signatures of spatiotemporal regulation. TFs are listed by their family or protein domain name and, where provided, Arabidopsis ortholog name in brackets. (E) Three co-expressed HD-Zip Class I genes (indicated by an asterisk in D) were modulated targets of RA1 and/or KN1. All were significantly down-regulated ([*] P < 0.05) in kn1 tassels; GRMZM2G132367 was significantly down-regulated in ra1, ra2, and ra3 mutant ears by 2 mm and showed significant change ([**] P < 0.001) between 1 and 2 mm in ra1 and ra3, but its expression remained unchanged in wild-type ears from 1 to 2 mm. RA1 and KN1 also cobound putative intergenic regulatory regions ∼15 kb upstream of these HD-Zip genes. Shown are orthologs of ATHB6 (GRMZM2G132367) and ATHB21 (GRMZM2G104204). (F) IDD genes bound by both RA1 and KN1 were positively modulated by RA1. ZmIDD-p1 (GRMZM2G179677) was repressed by KN1 while expression of the LOOSE PLANT ARCHITECTURE 1 ortholog (GRMZM2G074032) was not significantly altered in kn1 mutants.

Figure 6.

Models for RA1-mediated regulation and integration with KN1-based meristem maintenance pathways. (A) RA1 and KN1 interact via gibberellic acid (GA) biosynthesis and signaling. RA1 may modulate GA levels in a spatiotemporal manner by activating genes for its biosynthesis and catabolism and negatively regulating a repressor of GA signaling, SPY. (B) RA1 interfaces with various developmental and regulatory networks, and interacts with KN1-based meristem maintenance via common targets and pathways. RA1 directly represses genes involved in chromatin and RNAi and positively regulates a suite of co-expressed determinacy factors. Promoters of the latter were enriched for binding sites of LFY, bZIP, and MADS-box TFs, and therefore activation of determinacy factors by RA1 could work in part through coregulation by these TFs. RA1 positively regulates a set of IDD TFs, including one that is co-bound and repressed by KN1, and negatively regulates lg1, which may play a role in BM identity, possibly by establishing a boundary. RA1 and KN1 also co-target genes related to floral transition, auxin biology, and the integration of environmental and developmental cues.