Recruitment and remodeling of an ancient gene regulatory network during land plant evolution

Abstract

The evolution of multicellular organisms was made possible by the evolution of underlying gene regulatory networks. In animals, the core of gene regulatory networks consists of kernels, stable subnetworks of transcription factors that are highly conserved in distantly related species. However, in plants it is not clear when and how kernels evolved. We show here that RSL (ROOT HAIR DEFECTIVE SIX-LIKE) transcription factors form an ancient land plant kernel controlling caulonema differentiation in the moss Physcomitrella patens and root hair development in the flowering plant Arabidopsis thaliana. Phylogenetic analyses suggest that RSL proteins evolved in aquatic charophyte algae or in early land plants, and have been conserved throughout land plant radiation. Genetic and transcriptional analyses in loss of function A. thaliana and P. patens mutants suggest that the transcriptional interactions in the RSL kernel were remodeled and became more hierarchical during the evolution of vascular plants. We predict that other gene regulatory networks that control development in derived groups of plants may have originated in the earliest land plants or in their ancestors, the Charophycean algae.

The development of multicellular organisms is controlled by gene regulatory networks (GRNs) and the reorganization of GRN architecture is thought to be a major factor underlying morphological evolution (1–5). GRNs are hierarchic and modular structures where four major component classes can be identified (3): at the periphery of GRNs are differentiation gene batteries encoding proteins that execute cell type-specific functions (such as building a pigmented cell); upstream of these batteries are switches that allow or prevent subcircuits to function in specific developmental contexts, and “plug-ins,” small subcircuits that are flexibly and repeatedly used during development (such as signal transduction pathways); at the core of GRNs are kernels, small conserved subcircuits that execute specific developmental functions (such as defining spatial patterns in an embryo). Kernels are comprised of transcription factors that are highly conserved in distantly related species and are unusually stable components of GRNs.

Ancient kernels that regulate body plan and organ development are highly conserved among diverse groups of metazoans (animals) (6–10). In contrast, the core components of plant GRNs are difficult to identify because of the dynamic nature of plant genome evolution and the plastic character of plant development. Floral homeotic genes form a relatively recent kernel controlling flower development (11). Homologs of KNOX and LEAFY transcription factors control shoot development in vascular plants and sporophyte development in mosses (12–15). KNOX/BEL genes also control the development of the diploid phase in unicellular chlorophytes (16) and the haploid-to-diploid transition in mosses (15), suggesting that KNOX and LEAFY genes may be core members of ancient GRNs that control diploid development in plants. Auxin signaling (17, 18), ethylene perception (19), abscisic acid signaling (20), and several small RNAs are conserved between mosses and flowering plants (21, 22), suggesting that many switches and plug-ins of land plant GRNs have been conserved since before the evolution of vascular plants over 440 million y ago. However, the architecture and evolutionary history of these hypothetical ancient GRN kernels are mostly unknown.

In the angiosperm Arabidopsis thaliana, root hair development is controlled by the basic helix-loop-helix (bHLH) transcription factors AtRHD6 (A. thaliana ROOT HAIR DEFECTIVE 6) and AtRSL1 (A. thaliana RHD SIX-LIKE 1); their homologs in the moss Physcomitrella patens (Pp), PpRSL1 and PpRSL2, control the development of filamentous rooting structures: caulonema and rhizoids (23–25). This finding suggests that RSL genes belong to an ancient land plant GRN that controls the differentiation of cells with a rooting function. In A. thaliana, AtRHD6 was also found to form a transcriptional mechanism with two other bHLH transcription factors, AtRSL2 and AtRSL4 (26).

Here we test the hypothesis that the RSL mechanism is an ancient land plant kernel. We show that RSL genes form a transcriptional network that controls root hair development in A. thaliana and protonema development in P. patens. RSL genes form two ancient lineages that evolved in charophyte algae or in the first land plants and have been conserved during land plant evolution. Functional and expression analysis of the RSL genes in A. thaliana and in P. patens indicate that the two lineages form a transcriptional regulatory network in both species. Taken together, our results suggest that the RSL genes form a kernel that evolved over 450 million y ago and was recruited to control the development of root hairs during the evolution of vascular plants.

Results

RSL Network Controls Root Hair Development in A. thaliana.

The differentiation of root hairs in A. thaliana is controlled by a regulatory mechanism that comprises the bHLH transcription factors AtRHD6, AtRSL1, AtRSL2, and AtRSL4: no root hairs differentiate in Atrhd6 Atrsl1 or in Atrsl2 Atrsl4 double-mutants (23, 26), and the transcription of AtRSL2 and AtRSL4 is positively regulated by AtRHD6 and AtRSL1 (26). These four genes belong to a phylogenetic group that also includes AtRSL3 and AtRSL5 (Fig. 1B). To determine if AtRSL3 and AtRSL5 also control root hair development, we characterized the phenotypes of Atrsl3, Atrsl5 and Atrsl2 Atrsl3 mutants. Root hairs of Atrsl3 and Atrsl5 single-mutants were indistinguishable from wild-type, but the root hairs of Atrsl2 Atrsl3 were shorter than in Atrsl2 single-mutants (Fig. 1A). Furthermore, the constitutive expression of AtRSL3 or AtRSL5 in the hairless Atrsl2 Atrsl4 double-mutant background could partially restore root hair development (Fig. S1). Taken together, these data indicate that each of the six A. thaliana RSL genes positively regulate root hair development. AtRHD6 and AtRSL1 genes were expressed early in the development of trichoblasts (the epidermal cells that give rise to root hair cells), but the expression disappeared before root hairs initiated (23) (Fig. 1C). In contrast, AtRSL2 and AtRSL4 were expressed later, specifically during root hair growth (26) (Fig. 1C). Because AtRSL3 and AtRSL5 also control root hair growth, we hypothesized that they would be expressed while root hairs elongate. Accordingly, GFP:AtRSL3 and GFP:AtRSL5 protein fusions expressed under the control of their respective native promoters accumulated in the nuclei of trichoblasts during root hair growth (Fig. 1C). Taken together, this finding indicates that AtRHD6 and AtRSL1 act earlier in root hair development than AtRSL2, AtRSL3, AtRSL4, and AtRSL5. Together, these results suggest that all six RSL genes are components of a transcriptional network that controls root hair development in A. thaliana.

Fig. 1.

RSL class I and RSL class II proteins control root hair development in A. thaliana. (A) Root hair phenotype of single and double mutants of RSL genes in A. thaliana. (Scale bar, 200 μm.) (B) Maximum-likelihood cladogram showing that the A. thaliana RSL genes fall into two classes. The tree was rooted with AtbHLH040 (27). (C) Promoter-GFP-protein constructs showing that RSL class I and class II proteins accumulate in the nuclei of root hair cells before and during root hair growth, respectively. GFP-AtRSL5 can only be detected after application of exogenous auxin. (Scale bars, 50 μm.)

RSL Class I and Class II Genes Were Present in Early Land Plants.

We hypothesized that RSL genes form a GRN kernel that is present in other land plants. To trace the evolutionary history of the RSL regulatory network and define the diversity of RSL genes in plants, we identified and retrieved RSL sequences from 12 different plant genomes. RSL proteins are characterized by a conserved C-terminal region, which includes a bHLH domain that extends into a conserved stretch of 14 amino acids, the RSL domain (Fig. 2A). Maximum-likelihood phylogenetic analyses using RSL sequences from different species show that RSL proteins form two distinct and ancient phylogenetic clades, which we named RSL class I and RSL class II (Fig. 2B and Fig. S2) [subfamilies VIIIc(1) and VIIIc(2) in ref. 27]. We found both RSL classes in all species of land plants for which a complete genomic sequence is available, including mosses, lycophytes, eudicots, and monocots (Fig. 2C), but we did not find RSL sequences in chlorophyte algae. This result indicates that, like most other plant bHLH subfamilies (27), RSL class I and class II proteins evolved sometime after the divergence of the chlorophyte and streptophyte lineages 700–1,000 million y ago (28, 29), but before the evolution of vascular plants over 443 million y ago (30). This finding means that RSL proteins evolved in multicellular streptophytes either before or shortly after their colonization of terrestrial environments.

Fig. 2.

RSL proteins are conserved across land plants. (A) Alignment of conserved regions of the A. thaliana and P. patens RSL proteins. The position of the bHLH RSL domains is indicated by colored boxes; identical amino acids are represented in black. The sequence logos represent the multiple alignment of RSL class I and class II amino acid sequences from 13 plant species (Table S1); heights are proportional to sequence conservation in each position. (B) Maximum-likelihood tree of A. thaliana (red) and P. patens (green) RSL proteins. The tree was based on the bHLH and RSL domains of the alignment shown in A, together with the bHLH sequence of the outgroups AtbHLH040 and PpbHLH069 (27); approximate likelihood ratio test support values are indicated in the nodes. See also Fig. S2. (C) Number of RSL class I and RSL class II genes in different plant species (Table S1). (D) Six-day-old seedling roots of the A. thaliana wild-type Col-0, Atrsl2 Atrsl4 double-mutant, and Atrsl2 Atrsl4 expressing the P. patens genes PpRSL3, PpRSL4, PpRSL5, and PpRSL6 under the control of the constitutive CaMV 35S promoter. (Scale bar, 200 μm in the main figure and 50 μm in the close ups.)

The similarity of the amino acid sequences in the C-terminal region of RSL proteins in different land plants suggests that their molecular function may be conserved. To test this hypothesis, we transformed the hairless A. thaliana Atrsl2 Atrsl4 (RSL class II) double-mutant with P. patens RSL class II genes under the control of the constitutive cauliflower mosaic virus (CaMV) 35S promoter. Each of the moss genes could partially rescue the development of root hairs in Atrsl2 Atrsl4 double-mutant plants (Fig. 2D). The partial rescue ranged from the formation of small bulges in Atrsl2 Atrsl4 plants expressing PpRSL3 or PpRSL4 to the development of short tip-growing root hairs in plants expressing PpRSL5 or PpRSL6. Similarly, the expression of the A. thaliana RSL class II genes AtRSL3 and AtRSL5 in Atrsl2 Atrsl4 plants could partially rescue the development of root hairs (Fig. S1). Previously, Menand et al. (23) showed that a P. patens RSL class I protein could rescue the development of root hairs in the hairless A. thaliana RSL class I Atrhd6 Atrsl1 mutant. Taken together, these data indicate that the molecular function of RSL class I and RSL class II proteins is conserved between mosses and angiosperms.

RSL Network Controls Protonema Development in P. patens.

The conservation of RSL protein function in land plants supports the hypothesis that RSL genes are components of an ancient GRN that was present in early land plants. To further test this hypothesis, we determined if an RSL gene regulatory network exists in the moss P. patens; if the RSL network is present in both P. patens and A. thaliana then it would also have been present in their common ancestor, an early land plant. P. patens RSL class I genes control the chloronema-to-caulonema transition in the moss protonema (23). The protonema is a filamentous structure composed of chloronema and caulonema cells: chloronema are small, slow-growing cells with numerous chloroplasts that fulfill a predominantly assimilatory function; caulonema are larger, fast-growing cells that play an important role in substrate colonization (31). Young protonemata predominantly comprise cells with chloronema characteristics, but in older protonema several chloronema cells undergo a transition to caulonema. However, the differentiation of caulonema cells is totally abolished in the RSL class I Pprsl1 Pprsl2 double-mutant (23). If RSL class II genes function in the same pathway as RSL class I genes, we would expect that the chloronema-to-caulonema transition would also be defective in RSL class II mutants. To test this hypothesis, we generated P. patens plants that lack the function of single or paralogous pairs of RSL class II genes. Pprsl3 Pprsl4 double-mutants developed small and very dense protonemata composed predominantly of chloronema cells (Fig. 3), indicating that the chloronema-to-caulonema transition is defective. Unlike the Pprsl1 Pprsl2 RSL class I double-mutant, however, Pprsl3 Pprsl4 plants developed a few normal caulonema filaments. The cells in these filaments were morphologically identical to the cells of wild-type caulonema (Fig. S3), confirming that the phenotype of Pprsl3 Pprsl4 double-mutants was caused by a reduction in the proportion of cells that undergo the chloronema-to-caulonema transition and not by a general protonema growth defect. Pprsl3, Pprsl4, Pprsl5, and Pprsl6 single-mutants and Pprsl5 Pprsl6 double-mutants were phenotypically similar to wild-type protonemata (Fig. 3). We then constitutively expressed PpRSL3 and PpRSL4 under the control of the CaMV 35S promoter. The 35S:PpRSL3 plants were slightly smaller than wild-type and there was a strong reduction of caulonema development in 35S:PpRSL4 plants, which caused the development of very dense chloronema-rich protonemata (Fig. 3). These results indicate that ectopic expression or loss of function of PpRSL3 and PpRSL4 disrupts the chloronema-to-caulonema transition. Taken together, these data indicate that both RSL class I and class II genes control the chloronema-to-caulonema transition in the protonemata of P. patens, supporting the hypothesis that there is a functional RSL network in mosses.

Fig. 3.

RSL class II proteins control caulonema development in P. patens. (A) Protonemata of Pprsl3, Pprsl4, Pprsl5, Pprsl6, Pprsl3 Pprsl4, and Pprsl5 Pprsl6 loss-of-function mutants and constitutively expressed PpRSL3 and PpRSL4 genes in P. patens. Spores were germinated on minimal media for 3 wk. (Scale bars, 1 mm.) (B) Diameter of plants, shown as a stacked graph of the relative sizes of the inner chloronema-rich and the peripheral caulonema filament-rich regions (mean ± SD, n = 30). (C) Number of protruding caulonema filaments per plant (mean ± SD, n = 30 plants). *Significantly different from wild-type (P < 0.01 with Bonferroni multiple comparison correction); gray asterisks refer to the caulonema-rich region alone.

RSL Networks Have Different Topologies in A. thaliana and P. patens.

In A. thaliana, a primary characteristic of the RSL network is the positive regulation of the RSL class II genes AtRSL2 and AtRSL4 by RSL class I proteins (26). To determine if the regulatory interactions between the two RSL classes are conserved between P. patens and A. thaliana, we measured the relative changes in the steady-state levels of RSL mRNA in different rsl mutant backgrounds using quantitative RT-PCR (qRT-PCR). The levels of A. thaliana RSL class II mRNA were lower in Atrhd6 Atrsl1 double-mutants than in wild-type plants (Fig. 4A), indicating that RSL class I proteins positively regulate the transcription of all RSL class II genes. This conclusion is supported by the expression patterns of RSL genes in A. thaliana (Fig. 1C), which show that during differentiation of root hair cells, RSL class I genes are expressed before RSL class II genes. Levels of AtRSL3 and AtRSL5 mRNA were also lower in the Atrsl2 Atrsl4 double-mutant (Fig. 4A), implying the existence of at least three levels of regulation in the A. thaliana RSL network: RSL class I genes positively regulate AtRSL2 and AtRSL4, which in turn positively regulate the transcription of AtRSL3 and AtRSL5. In contrast, in the P. patens Pprsl1 Pprsl2 double-mutant only PpRSL6 mRNA levels are altered compared with the wild-type (Fig. 4B). The levels of each RSL mRNA are also similar in wild-type Pprsl3 Pprsl4 and Pprsl5 Pprsl6 double-mutants (Fig. 4B). This finding indicates that the expression of the different P. patens RSL genes is largely independent of the activity of other RSL proteins. Taken together, these results show that there are multilevel regulatory interactions between different RSL genes in A. thaliana, but fewer regulatory interactions between the P. patens RSL genes. We cannot discard the possibility that there are localized spatial or temporal interactions between P. patens RSL genes, which we would not be able to detect with a qRT-PCR using whole protonemata. In contrast with A. thaliana, where AtRSL genes are expressed specifically in trichoblast cells (RSL class I) and growing root hair cells (RSL class II) (26), PpRSL class I genes are loosely expressed throughout the protonema (25). The extent of the overlap between RSL class I and RSL class II gene expression during protonema development has yet to be investigated.

Fig. 4.

RSL genes and auxin form regulatory networks in P. patens and A. thaliana (A–F) qRT-PCRs showing the relative expression level of RSL genes in different A. thaliana (A and D) and P. patens (B, C, E, and F) mutant backgrounds and after NAA (1-naphthaleneacetic acid, a synthetic auxin) treatments. The expression levels are relative to wild-type (A–C) or untreated plants (D–F). The putative PpRSL7 transcript was not detected. Asterisk means absent or not determined. Bars represent the SD of three independent replicates. (G and H) Schematic representation of the regulatory interactions between the different RSL class I genes (red), RSL class II genes (blue), and auxin (green) in A. thaliana (G) and P. patens (H).

We also investigated the regulation of the RSL network by auxin, a key positive regulator of root hair development in angiosperms and the chloronema-to-caulonema transition in mosses (18, 32–34). Auxin positively regulates the expression of RSL class I genes in P. patens (24) (Fig. 4E) but had no effect on the expression of the A. thaliana RSL class I genes (Fig. 4D). However, the expression of RSL class II genes is highly responsive to auxin in both species. Exogenous auxin treatment increased the levels of AtRSL4 and AtRSL5 and reduced the levels of AtRSL2 and AtRSL3 mRNA in A. thaliana compared with untreated plants (Fig. 4D) (26). At low concentrations, exogenous auxin moderately increased the levels of all PpRSL class II mRNAs, whereas at higher concentrations it further increased the expression of PpRSL6 and decreased the levels of PpRSL3 (Fig. 4E). This finding means that auxin dynamically and strongly regulates the expression of RSL class II genes in both A. thaliana and P. patens. To confirm that endogenous auxin signaling modulates RSL gene expression, we compared RSL mRNA levels between wild-type P. patens and aux/iaa mutants. Aux/IAA proteins are transcriptional repressors that mediate auxin signaling in land plants; a set of aux/iaa P. patens mutants are auxin-resistant and display a delay or arrest in the chloronema-to-caulonema transition similar to the phenotypes observed in Pprsl mutants (18, 35). In two of these aux/iaa mutants (Ppiaa1A-113 and Ppiaa2-183), the mRNA levels of RSL class II genes were lower than in wild-type plants (Fig. 4C). Interestingly, a third aux/iaa mutant (Ppiaa1B-112) showed an inverse change in the levels of PpRSL3 and PpRSL4, confirming our observation that auxin has a dynamic effect on the expression of RSL class II genes. Taken together, these data demonstrate that auxin modulates the expression of RSL genes in P. patens.

We observed that the auxin-induced changes in the mRNA levels of AtRSL4, AtRSL5, PpRSL3, PpRSL5, and PpRSL6 were much larger in the Atrhd6 Atrsl1 and Pprsl1 Pprsl2 double-mutants than in wild-type (Fig. 4 D and F). This result indicates that RSL class I proteins negatively regulate the transcriptional responses of these RSL class II genes to auxin. Conversely, although levels of AtRSL2 and AtRSL3 mRNAs decreased upon auxin treatment in wild-type plants, they increased upon auxin treatment in Atrhd6 Atrsl1 double-mutants (Fig. 4 A and D). This finding indicates that RSL class I proteins are required for the auxin-induced repression of AtRSL2 and AtRSL3 in A. thaliana. The increase in steady-state mRNA levels of AtRSL2 and AtRSL3 in auxin-treated Atrhd6 Atrsl1 double-mutants is caused by the activity of AtRSL4; these genes were not induced when the Atrhd6 Atrsl1 Atrsl4 triple-mutant was treated with auxin (Fig. 4D). Taken together, these data indicate that RSL class I proteins modulate the transcriptional responses of RSL class II genes to auxin.

These positive and negative interactions between RSL genes can be incorporated into a regulatory network with numerous feed-forward loops, where auxin and RSL class I genes assume a central role in regulating the expression of RSL class II genes (Fig. 4 G and H). The many differences in the A. thaliana and P. patens RSL network architecture demonstrate that extensive rearrangements of these regulatory networks have occurred since mosses and angiosperms last shared a common ancestor. A fundamental difference in the networks between the two species is that although the expression of individual RSL genes in P. patens is largely independent of the activity of other RSL proteins, the A. thaliana network is more hierarchical with both RSL class I and RSL class II proteins regulating the expression of RSL class II genes.

Discussion

Our functional and phylogenetic analyses suggest that RSL class I and class II transcription factors form a GRN kernel that evolved in the common ancestors of mosses and vascular plants (Fig. 5). The RSL kernel likely evolved as a mechanism controlling cell-type transitions in aquatic charophyte algae or in the earliest land plants. After the divergence of mosses from other land plants, the RSL kernel controlled the differentiation of multicellular filamentous structures in mosses and was recruited to regulate the development of cellular projections from root epidermal cells (root hairs) in vascular plants. The architecture of the RSL kernel has changed since this divergence and became more hierarchical in the lineage that gave rise to A. thaliana than in the lineage that gave rise to P. patens. An alternative possibility is that the RSL kernel was already hierarchical in the first land plants and became reduced in the lineage that gave rise to modern mosses or, less likely, the RSL network evolved entirely independently in the two lineages.

Fig. 5.

Evolutionary history of the RSL network. RSL genes evolved in charophyte algae or in the first land plants, 500–1,000 million y ago. An ancestral RSL kernel was present in early land plants and was conserved during land plant evolution. Later, during vascular plant evolution, the RSL network was recruited to control the development of cellular projections from root epidermal cells (root hairs).

Our expression analyses suggest that a close interaction of the RSL kernel with an auxin signaling plug-in is a key structural feature of both the P. patens and A. thaliana GRNs (Fig. 4). Auxin is an important regulator of both protonema and root hair development and operates by regulating the expression of genes in the RSL network in both species. It is possible that other components of the higher-level GRN to which the RSL kernel belongs are partially conserved between mosses and angiosperms. We predict that there may be an overlap between the downstream differentiation gene batteries in both species. In A. thaliana, the RSL GRN regulates the transcription of a suite of genes that encode proteins involved in cell wall synthesis and modification, such as expansins, extensins, and peroxidases (26, 36). Therefore, we hypothesize that the RSL GRN controls the expansion of the root hair cell by regulating the expression of these growth-effector proteins in root hairs. Given the conservation of elements of the RSL GRN among mosses and seed plants, it is conceivable that homologs of some of these effector proteins are required for the chloronema-to-caulonema transition in mosses. That is, the RSL GRN in P. patens may control the expression of growth effectors as in the root hair of A. thaliana. If this is the case, it raises interesting questions about how these different GRN components were recruited and reassembled from the gametophyte to the sporophyte generation during the evolution of vascular plants.

The antiquity of the RSL kernel suggests that other plant regulatory networks may be derived from conserved GRN kernels that existed in the first land plants, almost 500 million y ago. Signaling pathways that can act as plug-ins of GRNs are conserved across land plants (18–20). Transcription factors, such as KNOX/BEL and LEAFY proteins, are conserved across plants (12–16, 37), but the networks in which these genes participate have not yet been described in early diverging groups of land plants. We predict that KNOX/BEL and LEAFY are also components of other plant kernels that control sporophyte development. If our hypotheses are correct, then the recruitment and modification of Pre-Cambrian and Early Paleozoic GRNs were important evolutionary mechanisms in multicellular plants. This process could have driven the generation of novel cell types and increased morphological diversity that occurred during the radiation of plants on land.

Materials and Methods

The Columbia-0 (Col-0) wild-type ecotype of A. thaliana and the Gransden wild-type strain of P. patens (Hedw.) Bruch and Schimp were used in this study. The lines Atrhd6-3 Atrsl1-1, Atrsl2-1, Atrsl4-1, Atrsl2-1 Atrsl4-1, Atrhd6-3 Atrsl1-1 Atrsl4-1, Pprsl1-1 Pprsl2-1, Ppiaa1A-113, Ppiaa1B-112, and Ppiaa2-183 were described previously (18, 23, 26). The T-DNA insertion Atrsl3-1: (GABI_422C06) was obtained from the Nottingham Arabidopsis Stock Centre. Atrsl5 was generated by introducing an artificial microRNA specifically targeting AtRSL5 into the wild-type. Two lines with significant repression for RSL5 transcript were picked out for further phenotypic analysis. The generation of Pprsl3, Pprsl4, Pprsl5, Pprsl6, Pprsl3 Pprsl4, and Pprsl5 Pprsl6 mutants and of Atrsl2 Atrsl4 plants transformed with 35S:AtRSL3, 35S:AtRSL5, 35S:PpRSL3, 35S:PpRSL4, 35S:PpRSL5, and 35S:PpRSL6 constructs is described in SI Materials and Methods and Fig. S4.