Modular regulation of floral traits by a PRE1 homolog in Mimulus verbenaceus: implications for the role of pleiotropy in floral integration

Abstract Floral traits often show correlated variation within and among species. For species with fused petals, strong correlations among corolla tube, stamen, and pistil length are particularly prevalent, and these three traits are considered an intra-floral functional module. Pleiotropy has long been implicated in such modular integration of floral traits, but empirical evidence based on actual gene function is scarce. We tested the role of pleiotropy in the expression of intra-floral modularity in the monkeyflower species Mimulus verbenaceus by transgenic manipulation of a homolog of Arabidopsis PRE1. Downregulation of MvPRE1 by RNA interference resulted in simultaneous decreases in the lengths of corolla tube, petal lobe, stamen, and pistil, but little change in calyx and leaf lengths or organ width. Overexpression of MvPRE1 caused increased corolla tube and stamen lengths, with little effect on other floral traits. Our results suggest that genes like MvPRE1 can indeed regulate multiple floral traits in a functional module but meanwhile have little effect on other modules, and that pleiotropic effects of these genes may have played an important role in the evolution of floral integration and intra-floral modularity.


Introduction
A flower is essentially a functional unit composed of multiple floral organs that must be arranged in a proper configuration to ensure pollination success. As such, floral traits often show correlated variation within and among species [1][2][3][4][5][6], a pattern commonly referred to as floral integration [7][8][9]. For sympetalous species (i.e. species bearing flowers with their petals fused into a corolla tube), strong correlations among corolla tube, stamen, and pistil length are prevalent [10][11][12][13][14][15][16][17][18]. These three traits likely form a common intra-floral functional module, as their coordination is critical for successful pollen transfer [8]. This raises the question: what is the genetic architecture underlying such intra-floral modules?
Pleiotropy and linkage disequilibrium are two main genetic hypotheses for the expression of modularity [19][20][21][22][23][24]. The pleiotropy hypothesis assumes that a single gene affects multiple traits simultaneously within a module, whereas the linkage disequilibrium hypothesis assumes that each individual trait within the module is controlled by a different gene but some mechanism (e.g. recombination suppression associated with heterochromatic regions or chromosomal inversions) maintains a tight linkage among these genes. Of the two hypotheses, pleiotropy has been considered a major driver of floral integration and intra-floral modularity [2,9,10], although conclusive evidence supporting this hypothesis is scarce. The strongest evidence so far probably comes from studies of wild radish (Raphanus) and tobacco (Nicotiana) flowers, where genetic correlations among corolla tube, stamen, and pistil length do not break down after several generations of random mating [2,10]. However, even in these cases the actual gene that presumably affects all three traits has not been identified, and the possibility that these traits are controlled by multiple distinct genes in a chromosomal region with recombination suppression cannot be completely ruled out.
An alternative strategy to test the potential role of pleiotropy in intra-floral modularity is transgenic perturbation of candidate genes that might affect multiple floral traits within a module while affecting other traits or modules far less. In this study, we employ this transgenic approach to test modular regulation of the lengths of corolla tube, stamen, and pistil, in the hummingbird-pollinated monkeyflower species M. verbenaceus. M. verbenaceus is part of the Mimulus lewisii species complex, a powerful new model for developmental genetics and evo-devo studies of pollinator-associated floral traits [25]. We chose this species because it has a small stature but large, showy flowers (Fig. 1), a short generation time, and relatively small genome (∼500 MB) with high-quality genomic resources (http://mimuba se.org/), and most importantly, is readily amenable to Agrobacterium-mediated stable transformation [26].
Choosing the appropriate candidate genes, however, was not trivial. Although a large number of genes are known to control plant organ size in Arabidopsis [27][28][29], few genes are known to specifically control floral organ length with little effect on other parts of the plant, especially in sympetalous species. Given that plant organ elongation often involves coordinated actions of gibberellin (GA), brassinosteroid (BR), and auxin, we reasoned that an "integrator" of these hormone signaling pathways might be a good candidate. In particular, if this integrator gene has multiple functionally redundant paralogs in the genome, one of these gene family members could be preferentially expressed in floral organs, thereby regulating the elongation of multiple floral organs with minimal effect on vegetative tissues. The PACLOBUTRAZOL RESISTANCE (PRE) gene family fits these criteria. There are several closely related PRE paralogs in most plant genomes [30][31][32]. They encode small helix-loop-helix (HLH) proteins (90 ∼ 110 amino acids) that act downstream of and integrate GA, BR, and auxin signaling to promote cell elongation in Arabidopsis [30,[33][34][35], tomato (Solanum lycopersicum) [36], cotton (Gossypium spp.) [32], and rice (Oryza sativa) [31,35,37]. As such, we considered them promising candidates in the modular regulation of floral organ elongation.
Here we report the characterization of PRE homologs in M. verbenaceus and transgenic perturbation of MvPRE1, the only homolog preferentially expressed in floral organs. Our results demonstrate that MvPRE1 positively regulates corolla tube length, petal lobe length, stamen length, and pistil length, but not calyx or leaf length. We suggest that genes like MvPRE1 may occupy a specialized hub position [38] in a genetic network and regulate multiple traits in a functional module while having little effect on other modules, and thus provide the suitable genetic materials for the evolution of phenotypic integration and modularity through pleiotropy.

MvPRE1 is the major PRE-like gene expressed during flower development in M. verbenaceus
We identified seven PRE-like genes, MvPRE1-MvPRE7, from the MvBL genome by TBLASTN searches. They encode proteins that are 91 ∼ 105 a.a. long and share sequence similarity with AtPRE1 across the entire length ( Fig. 2a; Supplementary File S1). To determine which MvPRE genes may function in flower development, we performed RT-PCR across four stages of floral bud development (i.e. 5-mm, 15-mm, 20-mm, and 35-mm), and found that MvPRE1 is the only PRE-like gene with readily detectable expression across multiple floral developmental stages, with peak expression level at the 20-mm stage (Fig. 2b) when the corolla tube is rapidly growing through cell elongation [39]. Phylogenetic analysis suggests that MvPRE1 is closely related to tomato SlPRE1/Style2. 1 and SlPRE2 ( Fig. 2a; Supplementary Fig. S1). We further examined the spatial expression pattern of MvPRE1 in vegetative tissues and various floral organs dissected from 20-mm floral buds, and observed that MvPRE1 is highly expressed in the corolla tube, petal lobe, filament, and style, with intermediate expression levels in the leaf and stem, weak expression in the calyx and ovary, and no detectable expression in the anther and root (Fig. 2c).

MvPRE1 knockdown causes coordinated decreases in the lengths of corolla tube, style, and stamen
The spatial pattern of MvPRE1 expression suggests that it may play an important role in the elongation of the corolla, stamen filament, and style. To test this hypothesis, we knocked down the expression of MvPRE1 by transforming MvBL with an RNAi plasmid. We obtained five stable transgenic RNAi lines, three of which showed strong phenotypes and the other two showed intermediate phenotypes. Compared to the wild type, the three strong RNAi lines (RNAi-1, RNAi-3, RNAi-4) exhibited a moderately shorter stature (Fig. 3a) and slightly longer leaves ( Fig. 3b; Supplementary Fig. S2a), but no obvious change in other vegetative characters. Intriguingly, the three floral traits that often show strong genetic correlations in sympetalous species (i.e. corolla tube length, stamen length, and pistil length) exhibited coordinated changes in these RNAi lines: the corolla tube, stamen, and pistil are reduced in length by ∼20%, 28%, and 21%, respectively, compared to the wild type ( Fig. 3c-e). Petal lobe length was also reduced to a similar degree (∼20%), but calyx length showed little change (Fig. 3e). These results are consistent with the tissue-specific expression pattern of MvPRE1 in the wild-type flower, in that MvPRE1 has relatively high expression levels in the corolla tube, petal lobe, stamen filament, and style, but low expression levels in the leaf and calyx (Fig. 2c). To verify that these phenotypes are indeed caused by the knockdown of MvPRE1, we performed RT-qPCR on whole floral buds at the 20-mm stage and found that the expression level of MvPRE1 decreased nearly 10-fold in the strong RNAi lines (Fig. 3f). In addition, the different phenotypic strengths displayed by the intermediate and strong RNAi lines furnished an opportunity to test phenotypic correlation quantitatively ( Supplementary Fig. S3). As expected, the corolla tube length, stamen length, and pistil length were tightly correlated with each other ( Supplementary  Fig. S3f). The internode length was also correlated  with the three floral traits, but less strongly than that among the three floral traits. By contrast, leaf length showed only weak correlation with internode length, and no correlation with any of the three floral traits ( Supplementary Fig. S3f). Taken together, these results suggest that MvPRE1 coordinately regulates the length of multiple pollinator-associated floral traits (i.e. corolla tube, petal lobe, stamen, and style), but plays little role in calyx or leaf elongation in M. verbenaceus.

MvPRE1 overexpression leads to increases in corolla tube and stamen lengths, but not pistil length
Given that knockdown of MvPRE1 in MvBL causes decreases in the lengths of the corolla tube, stamen, pistil, and petal lobe, we expected that overexpression of MvPRE1 would increase the lengths of these floral organs, and maybe even increase the length of calyx, where the endogenous expression level of MvPRE1 is low. To test these predictions, we transformed MvBL with a 35S:MvPRE1-YFP plasmid. We obtained 40 stable transgenic lines, 11 of which showed obvious phenotypes and three (OE-28, OE-6, and OE-10) were selected for further analysis. As expected, corolla tube and stamen lengths increased significantly in the overexpression lines compared to the wild type ( Fig. 4c-e). However, pistil length did not change (Fig. 4e). As a result, the anthers are clearly positioned above the stigma in the overexpression lines, whereas in the wild type the anthers are at similar height to the stigma (Fig. 4d). We did not observe any obvious change in petal lobe or calyx length either (Fig. 4e). The overexpression lines also showed changes in vegetative tissues: the whole plant appears less robust, and leaves are shorter and narrower than the wild type (Fig. 4a,b; Supplementary Fig. S2b). RT-qPCR experiments verified that the expression levels of MvPRE1 increased 3 ∼ 4-fold in the three selected overexpression lines (Fig. 4f). The 35S:MvPRE1-YFP transgene also allowed us to examine subcellular localization of the MvPRE1 proteins, which are predominantly localized in the nucleus, with weaker YFP signals also detectable in the cytosol (Supplementary Fig. S4). This pattern of subcellular localization is very similar to AtPRE1 and AtPRE3/ATBS1 in Arabidopsis [30,34]. Taken together, these results further demonstrate that MvPRE1 positively regulates elongation of some floral organs, but its function is tissue dependent, as it appears to have no effect on the calyx and have negative effect on leaf size.

MvPRE1 promotes cell elongation in M. verbenaceus flowers
To determine whether the floral organ length changes in the MvPRE1 RNAi and overexpression lines are due to change in cell number or cell length, we profiled cell number and size along the dorsal side of the corolla tube of open flowers from the wild type (MvBL), RNAi line 4 (RNAi-4), and overexpression line 10 (OE-10). Consistent with a previous study [39], we found that in the wild-type MvBL, cell length increases gradually from the base of the corolla tube to the stamen insertion site, then plateaus for ∼50 cells with maximal length before gradually decreasing towards the distal end (Fig. 5a,d). Accordingly, cell shape changes from more rounded at the base to elongated rectangular in the plateau zone to puzzle-shaped at the distal end (Fig. 5a). RNAi-4 and OE-10 showed the same pattern of cell length variation along the corolla tube and had similar cell numbers to the wild type ( Fig. 5b-d). The main difference is that RNAi-4 has shorter cells than the wild type, especially in the zone near the stamen insertion site (from the 50th to the 150th cell), and OE-10 has longer cells ( Fig. 5d; Supplementary  Table S1). These results suggest that MvPRE1 regulates floral organ length by promoting cell elongation instead of cell division, consistent with the known function of its homologs in Arabidopsis, rice, and cotton [32,33,35].

Discussion
In this study, we have shown that knocking down the expression of a PRE-like gene in the hummingbirdpollinated monkeyflower species M. verbenaceus caused simultaneous decreases in the lengths of corolla tube, petal lobe, stamen, and pistil, but little change in calyx and leaf lengths or organ width (Fig. 3b-e; Supplementary  Fig. S2a). For sympetalous species, the lengths of corolla tube, stamen, and pistil often show strong phenotypic and genetic correlation independent of other floral or vegetative traits, and these organs are considered to be a functional module [10][11][12][13][14][15][16][17][18]. Because the corolla tube can direct or limit pollinator entrance, integration of tube, stamen, and pistil lengths maintains the anthers and stigma in the appropriate position for contacting the pollinator [40,41]. Moreover, coordination of the relative positions of the anthers and stigma (herkogamy) mediates the efficacy of pollen transfer [17,42,43]. Pleiotropy has long been implicated in such patterns of modular integration in sympetalous species [9,44], but empirical evidence based on actual gene function is lacking. Our demonstration of the modular regulation of these traits by a single gene supports the role of pleiotropy in generating intra-floral integration.
The overexpression experiment suggests that the role of MvPRE1 in organ elongation depends on the developmental context. While downregulation of MvPRE1 affected all organs in the putative functional module, overexpression of MvPRE1 in M. verbenaceus led to increased corolla tube and stamen length, as expected, but did not change pistil or calyx length (Fig. 4c-e). In the leaf, overexpression of MvPRE1 even had the opposite effect: both leaf length and width were reduced ( Fig. 4b;  Supplementary Fig. S2b). Notably, AtPRE1 overexpression in Arabidopsis also resulted in narrower leaves [30], indicating that the leaf phenotypes we observed are not an oddity unique to M. verbenaceus.
In Arabidopsis, AtPRE1 is part of a triantagonistic HLH/bHLH system that controls cell elongation. In this     Regardless of the exact underlying mechanisms, the fact that transgenic manipulations of MvPRE1 trigger correlated responses of the three floral organs (corolla, stamen, and pistil) that are distinct from the leaf or calyx is evolutionarily and functionally significant. An important criterion of modularity is that the characters within a module are tightly correlated with each other, but less strongly correlated with those outside of the module [1,8,22]. MvPRE1 provides an example of how this can be realized by the pleiotropic effect of a single gene. Although manipulation of MvPRE1 did not exclusively affect the organs of the putative functional module, the effects on the serially homologous calyx and vegetative leaves were far less ( Fig. 3; Supplementary Fig. S3).
Pleiotropy and modularity may have played a critical role in the evolutionary diversification of organismal forms in general [21,22,24] and flower form and function in particular [9,44], because selection on one trait will result in correlated changes in functionally related traits. Indeed, floral traits associated with pollinator shifts have been subjected to quantitative trait locus (QTL) analysis in a wide range of sympetalous taxa, including Mimulus [11,46], Leptosiphon [47], Ipomopsis [48], Penstemon [18], Petunia [49], and Ipomoea [50]. These studies showed that corolla tube, stamen, and style lengths are not only genetically correlated but also frequently co-localized to the same QTL, consistent with a role of pleiotropic QTLs underlying intra-floral  It should be noted that our results on MvPRE1 only show that pleiotropy can produce intra-floral integration, it does not show pleiotropy actually did cause intrafloral integration during Mimulus evolution. To demonstrate the latter, one would need to first identify the presumptive gene that causes intra-or inter-specific covariation of traits that constitute a functional module, then demonstrate the phenotypic consequence and pleiotropic effect of the causal gene by generating high-resolution introgression lines (i.e. allelic substitution at single-gene level between the two forms) or direct transgenic manipulations. Unfortunately, although efforts in identifying causal genes responsible for intraor inter-specific flower color variation in Mimulus were successful [51][52][53][54][55][56], progress in genetic mapping of floral organ length variation has been slow, partly because organ length tends to be more polygenic than pigmentation traits [11,46,57]. However, with chromosomelevel genome assemblies for multiple Mimulus species becoming available recently (http://mimubase.org/), identification of casual genes underlying floral organ length variation should be soon within reach.
This study also raises the question of how many other genes may behave like MvPRE1 in modular regulation of floral traits. The majority of genes reported to control floral organ size in Arabidopsis also tend to have strong pleiotropic effects in vegetative tissues [27][28][29]. We still know very little about which genes specifically control floral organ size, shape, number, and spatial arrangement without affecting other plant parts, especially in non-Arabidopsis species. However, it is not difficult to envision the general properties of such genes: (i) They are likely "hub" genes in a gene regulatory network to control the development of multiple traits [23]; (ii) They should have closely related paralogs in a genome to allow parcellation (e.g. through sub-functionalization or evolution of tissue specific expression patterns) [22]. Given the impressive pace of technological improvements (e.g. long-read sequencing, genome editing) in the past few years, both classical model systems for floral organ elaboration like snapdragon [58] or petunia [59] and more recently developed plant systems such as Aquilegia [60], Sinningia [61], Torenia [62], Mimulus [25], and Nigella [63] are becoming more and more accessible to detailed genetic and developmental analyses. We are optimistic that studies in these systems will reveal the genetic networks that can regulate multiple floral traits in a functional module but meanwhile have little effect on other modules, and that investigations on the role of pleiotropy in the evolution of floral integration and intra-floral modularity will begin to focus on specific genes instead of presumptive "loci" without any knowledge of their molecular identities.

Plant materials and growth conditions
The M. verbenaceus inbred line MvBL (wild type) was previously described in [26]. All plants were grown in the University of Connecticut research greenhouses under natural light supplemented with sodium vapor lamps, ensuring a 16-hr day length with a light intensity of 110-160 μmol·m −2 ·s −1 . Plants were fertilized 2-3 timesper week.

Sequence retrieval and phylogenetic analysis
The amino acid (a.a.) sequence of Arabidopsis AtPRE1 (AT5G39860, 92 a.a.) was used as query to retrieve the M. verbenaceus homologs from the MvBL genome (version 2.0; http://mimubase.org/). TBLASTN searches, as implemented in the program TARGeT (http://newcizin.cyverse.org/) [64], resulted in seven genes encoding proteins that are 91-105 a.a. long and share sequence similarity with AtPRE1 across the entire length. We named these genes MvPRE1-MvPRE7. Multiple sequence alignment (Supplementary File S1) of all PRE-like proteins from Arabidopsis, tomato, rice, and M. verbenaceus were performed in AliView [65]. Phylogenetic analyses of the amino acid sequence alignment were performed using both the parsimony criterion and the maximum likelihood method. Parsimony analysis was performed using PAUP * version 4.0a168 [66], with 200 random stepwise addition replicates and TBR branch swapping with MULTREES option effective. Clade support was determined by bootstrap analyses of 200 replicates. Maximum likelihood analysis was performed using MEGA X [67] with the JTT amino acid substitution model and 1000 bootstrap replicates.

Expression analysis using RT-(q)PCR
RNA extraction and cDNA synthesis were as described in [53]. Briefly, total RNA was extracted using the Spectrum Plant Total RNA Kit (Sigma-Aldrich). cDNA was synthesized from 1 μg of the DNaseI (Invitrogen) treated RNA using the SuperScript III First-Strand Synthesis System (Invitrogen) or GoScript™ Reverse Transcription Mix (Promega), then diluted 10-fold before RT-(q)PCR. RT-PCR was performed to determine the expression patterns of all seven MvPRE genes in different floral developmental stages of MvBL and the expression pattern of MvPRE1 in various tissue types. The M. verbenaceus ortholog of Arabidopsis ubiquitin-conjugating enzyme gene (At5g25760), MvUBC, was used as a reference gene following [53]. Whole floral buds at 20-mm stage were used for assaying relative transcript levels of MvPRE1 in wild-type MvBL and transgenic lines by RT-qPCR, which was performed using iQ SYBR Green Supermix (Bio-Rad) and a CFX96 Touch Real-Time PCR Detection System (Bio-Rad). For RT-qPCR experiments, three biological replicates were included for each genotype. Each replicate was a distinct 20-mm floral bud. The cDNA samples were amplified for 40 cycles of 95 • C for 15 s and 60 • C for 20 s. Amplification efficiencies for each primer pair were calculated using critical threshold values obtained from a dilution series  Figure S5) that designing paralog-specific primers was straightforward. To further verify the specificity of primers, each primer was BLASTed against the MvBL genome in the TARGeT platform (http://new-cizin.cyverse.org/) [64] with an E-value cut-off of 0.1, ensuring that each primer has only one match in the genome. The primers used for RT-(q) PCR were listed in Supplementary Table S2.

Plasmid construction and plant transformation
An RNA interference (RNAi) plasmid was constructed to knock down the expression of MvPRE1 in MvBL. Since the MvPRE1 gene is quite small, and BLASTing the fulllength coding sequence (CDS) (282 bp, without the stop codon) against the MvBL genome found no other genomic regions that perfectly match this CDS for a contiguous block longer than 16 bp (i.e. the MvPRE1 sequence is highly specific at the nucleotide level), we used the entire CDS for the RNAi plasmid. To this end, we first amplified the CDS from MvBL cDNA with the Phusion High-Fidelity DNA Polymerase (New England Biolabs), cloned it into the pDONOR207 vector (Invitrogen) by BP cloning reaction, and then cloned it into the pB7GWIWG2(I) vector [68] in both sense and antisense directions by LR recombination.
To build the overexpression plasmid, the full-length MvPRE1 CDS (without the stop codon) was first cloned into the pDONOR207 vector (Invitrogen) by BP reaction, and then cloned into the destination vector pEarleyGate 101 [69] by LR recombination. The pEarleyGate 101 vector drives transgene expression by the cauliflower mosaic virus (CaMV) 35S promoter and contains the YFP CDS fused in frame with the 3 end of the gene of interest (MvPRE1).
After verifying the plasmid sequence by Sanger sequencing, each final plasmid was transformed into Agrobacterum tumefaciens strain GV3101 and then transformed into the plants following the "floral spray/ vacuum infiltration" protocol described in [53]. The primers used for plasmid construction were listed in Supplementary Table S2.

Phenotypic analysis
To quantify the phenotypic differences among wild type, MvPRE1 RNAi, and MvPRE1 overexpression lines, we measured several floral traits using digital calipers, including lateral corolla tube (LCT) length, dorsal corolla tube (DCT) length, ventral petal lobe (VPL) length and width (Fig. 1b,c), pistil length (from the receptacle to the stigma), stamen length (from the receptacle to the anther), and calyx length. Ten fully open flowers were measured for each genotype. We also measured the length and width of the largest leaves (usually the fourth pair on the main stem) of mature plants. To determine whether the phenotypic differences between the wild type and the transgenic lines are due to cell number or cell size, we profiled the cell number and cell length along the dorsal side of the corolla tube (Fig. 1c) of mature flowers using a compound light microscope, following [39].