Bulk Segregant Analysis of an Induced Floral Mutant Identifies a MIXTA-Like R2R3 MYB Controlling Nectar Guide Formation in Mimulus lewisii

Abstract

The genetic and developmental basis of many ecologically important floral traits (e.g., carotenoid pigmentation, corolla tube structure, nectar volume, pistil and stamen length) remains poorly understood. Here we analyze a chemically induced floral mutant of Mimulus lewisii through bulk segregant analysis and transgenic experiments and identify a MIXTA-like R2R3 MYB gene that controls nectar guide formation in M. lewisii flowers, which involves epidermal cell development and carotenoid pigmentation.

Mimulus, MIXTA-like MYB, nectar guides, trichome, carotenoids

THE rapid adaptive radiation of the >250,000 species of flowering plants has produced an astonishing diversity of flower morphology. Uncovering the genetic basis (i.e., genes and genetic pathways/networks) of floral trait variation is a fundamental step toward understanding the origin and evolution of these “endless forms” (Darwin 1859). Floral trait diversification is often thought to be driven principally by plant–pollinator interactions (Darwin 1862; Grant and Grant 1965; Stebbins 1970; Fenster et al. 2004; Harder and Johnson 2009). Therefore, an ideal experimental system to study the genetic basis of flower diversification should include diverse phenotypes that interact with different pollinators and be amenable to rigorous genetic and developmental analysis.

The foremost plant genetic model system, Arabidopsis thaliana, has been instrumental in unraveling the genes and pathways involved in making flowers from leaves (i.e., the origin of the first flower) (Coen and Meyerowitz 1991; Theissen 2001; Glover 2007). However, being a self-fertilizing species, it has little variation in floral traits that are important for pollinator interactions (e.g., color, shape, rewards, display). Another important plant model system, Antirrhinum majus, has been invaluable for our understanding of the genetic and developmental basis of floral organ identity, flower symmetry, and anthocyanin pigmentation (Coen and Meyerowitz 1991; Schwarz-Sommer et al. 2003; Glover 2007), largely thanks to its endogenous active transposable elements (TEs) that allow gene isolation by transposon tagging. However, the lack of standard genomic resources (e.g., genome assembly) and a routine stable transformation protocol has impeded exploitation of this system to study floral traits for which TE-induced mutants are not available (e.g., carotenoid pigmentation, corolla tube formation and elaboration, and stamen and pistil length).

Mimulus (monkeyflowers) represents an emerging model system that complements the aforementioned, well-established study systems, especially for exploring the diversification of flower morphology. The 160–200 species in the genus exhibit tremendous variation in floral traits and interact with a diverse array of pollinators (Wu et al. 2008). Of particular interest to us are Mimulus lewisii and M. cardinalis, sister species that are genetically very similar but display dramatically different flower phenotypes and are pollinated by bumblebees and hummingbirds, respectively (Hiesey et al. 1971; Bradshaw et al. 1995; Ramsey et al. 2003). These species have several features that greatly facilitate genetic analysis, including high fecundity (∼1000 seeds per fruit), short generation time (3 months), and relatively small genome size (∼500 Mb).

Recently, we have developed genomic resources for M. lewisii and M. cardinalis (Yuan et al. 2013), in conjunction with community resources developed for the other model species in the genus, M. guttatus (http://www.mimulusevolution.org/; http://www.phytozome.net/cgi-bin/gbrowse/mimulus/). More importantly, we have established an efficient in planta transformation system for M. lewisii, which allows transgenic experiments to be performed to characterize gene function and developmental processes rigorously. In the previous study (Yuan et al. 2013), we have demonstrated that these genomic resources and functional tools enable fine dissection of the genetic basis of flower color variation between M. lewisii and M. cardinalis. However, using this system to understand the genetics and development of flower diversification in other angiosperms—at the most fundamental level—is limited by the existing natural floral trait variation between M. lewisii and M. cardinalis. To overcome this limitation, we initiated a large-scale ethyl methanesulfonate (EMS) mutagenesis experiment using M. lewisii inbred line LF10, to generate novel flower phenotypes that have potential ecological relevance (Owen and Bradshaw 2011). Studying the developmental genetic basis of these mutant phenotypes presumably will generate useful knowledge for understanding the genetic basis of similar phenotypes found in natural species across the angiosperm phylogeny. Here we present an exemplar case, describing the discovery of a MIXTA-like R2R3 MYB gene that controls the formation of nectar guides in M. lewisii by analyzing an EMS mutant.

Results and Discussion

The ventral petal of the pink-flowered M. lewisii has two yellow hairy ridges as nectar guides for bumblebees (Figure 1A). This contrasting color pattern is typical of bee-pollinated flowers (Daumer 1958), including A. majus, although in Antirrhinum the yellow color is due to aurones (Jorgensen and Geissmann 1955), a type of flavonoid pigment, whereas in M. lewisii it is due to carotenoid pigments (Supporting Information, Figure S1). The ecological function of the nectar guides in attracting and properly orienting bumblebees into the flower during pollination has been demonstrated in M. lewisii by using an EMS mutant, guideless (Owen and Bradshaw 2011). This mutant displays a novel phenotype, lacking the yellow color and the brushy hairs (trichomes) in the nectar guides (Figure 1B), but without pleiotropic effects outside the flower. guideless was observed to segregate as a Mendelian recessive trait (Owen and Bradshaw 2011), but the gene identity remained unknown.

Figure 1

Phenotypic characterization of wild-type M. lewisii LF10 and the guideless mutant. Wild-type LF10 has two yellow ridges with brushy hairs (trichomes) on the ventral petal (A), conical cells on the inner epidermis of all petal lobes (C), and long (1–3 mm) single-celled trichomes in the nectar guides (E). In guideless mutants, there are neither yellow pigment nor brushy hairs on the ventral petal (B); the conical cells on the inner epidermis of petal lobes are much less elaborated (D), and the vestigial trichomes in the nectar guides are short (<50 μm) and stumpy (F). Bars on the SEM micrographs, 50 μm.

To identify the GUIDELESS gene, we carried out a bulk segregant analysis coupled with deep sequencing (Lister et al. 2009). We first crossed guideless (in the LF10 genetic background) with another M. lewisii inbred line, SL9, and pooled DNA samples from 100 F₂ segregants with the mutant phenotype (i.e., homozygous for the LF10 guideless allele). We then sequenced the pooled DNA sample to an average coverage of 55-fold (277 million 100-bp Illumina paired-end reads), and mapped the short reads to the SL9 genome using CLC Genomics Workbench. The GUIDELESS gene and tightly linked regions are expected to be homozygous for the LF10 genotype among all individuals displaying the mutant phenotype (Figure S2), which means that these regions are highly enriched in homozygous single nucleotide polymorphisms (SNPs) in the “F₂ reads–SL9 genome” alignment.

To generate the reference SL9 genome, we sequenced SL9 to an average coverage of 12-fold (82 million 75-bp Illumina paired-end reads), and de novo assembled the short reads into 86,563 contigs with an N50 of 2.3 kb, using CLC Genomics Workbench. We then aligned these contigs against the 14 chromosome-level superscaffolds of the M. guttatus genome using the software package MUMmer 3.0 (Kurtz et al. 2004), assuming gene collinearity between M. lewisii and M. guttatus. The M. lewisii and M. guttatus genomes are sufficiently diverged at nucleotide level that only the coding regions are readily alignable; therefore, only the genic regions of SL9 were captured in this genome alignment, with essentially all of the intergenic noncoding sequences being left out. This resulted in 14 “pseudoscaffolds” of SL9, which together contain ∼70 Mb of genic sequences.

We scanned all 14 pseudoscaffolds in 20-kb intervals for enrichment of homozygous SNPs and found one sharp peak at the beginning of pseudoscaffold 5 (Figure 2A). This peak corresponds to a 50-kb region on M. guttatus scaffold 66, which contains only nine genes (Figure 2B). A manual inspection of the pooled mutant sample sequences that match each of the nine genes revealed neither nonsense nor nonsynonymous mutations nor mutations that potentially affect intron splicing. Instead, we found a 2-bp frameshift insertion in the beginning of the third exon of mgv1a023545 (Figure 2B), which is a MIXTA-like R2R3 MYB gene and is the most promising candidate for GUIDELESS (Figure S3).

Figure 2

Identification of the GUIDELESS gene by bulk segregant analysis and deep sequencing. (A) Genome scan for regions that are enriched in homozygous SNPs. Each pseudoscaffold of the M. lewisii SL9 genome was binned into 20-kb intervals, and the number of homozygous SNPs in each 20-kb interval was plotted in a bar graph. (B) The 50-kb candidate interval contains nine genes, among which a MIXTA-like R2R3 MYB gene, mgv1a023545m, has a 2-bp insertion in the beginning of the third exon that disrupts the reading frame. (C) RT–PCR of GUIDELESS and the reference gene MlUBC (Yuan et al. 2013) in various tissue types and six stages of corolla development. GUIDELESS expression is restricted to the floral tissue and increases during early corolla development, with the highest level at the 9- to 10-mm stage and then decreases as the corolla matures.

MIXTA-like genes are known to positively regulate trichome development and epidermal cell differentiation in Antirrhinum and other plants (Glover et al. 1998; Perez-Rodriguez et al. 2005; Baumann et al. 2007), which is consistent with the aborted trichome and epidermal conical cell development in the guideless mutant (Figure 1). However, to our knowledge, MIXTA-like genes have never before been associated with the regulation of carotenoid pigmentation, as the absence of yellow carotenoids in the ventral petal of guideless mutants would indicate. The expression of this MYB gene is restricted to floral tissue in LF10, and peaked at the 9–10 mm stage of corolla development (Figure 2C). This is consistent with the observation that the guideless mutant has no phenotypic effect outside the flower.

MIXTA-like R2R3 MYBs can be conveniently identified by a conserved signature motif, “HMAQWESARLEAEARLx-RxS” (Stracke et al. 2001; Brockington et al. 2013) (Figure S3). A TBLASTN search against the M. guttatus genome assembly (http://www.phytozome.net/cgi-bin/gbrowse/mimulus/) using this motif as query with an E-value cutoff of 1 retrieved the same set of MIXTA-like MYB genes identified in a previous study (Scoville et al. 2011) (Figure S4). Using the same search strategy for the M. lewisii genome assembly, we retrieved 10 putative MIXTA-like sequences, 2 of which contain multiple nonsense and frameshift mutations and are most likely pseudogenes. The other 8, including the GUIDELESS candidate (GenBank: KC139356) and MlMYBML1–MlMYBML7 (KC692454–KC692460), were annotated as bona fide MIXTA-like R2R3 MYB genes (Figure S4).

To confirm that the candidate MIXTA-like MYB is GUIDELESS, we wanted to rescue the guideless phenotype by transforming a genomic copy of the wild-type LF10 allele into the mutant background. However, a transposable element of unknown size, located 190 bp upstream of the ATG translation initiation codon, rendered our attempts to clone the promoter region of the wild-type allele unsuccessful. Therefore, we took an alternative approach––transforming the wild-type LF10 with an RNAi construct. Knocking down the expression of this gene in LF10 was expected to recapitulate the guideless phenotype.

An RNAi plasmid was constructed with a 339-bp fragment from the third exon of the candidate MYB gene (File S1). This fragment was BLASTed against the LF10 genome assembly to ensure target specificity. We obtained four independent RNAi lines that closely resemble the guideless mutant not only in gross morphology (i.e., much reduced trichome development and carotenoid pigmentation), but also in the fine structure of petal lobe and nectar guide epidermal cells (Figure 3 A–F). Presence of the transgene in these RNAi lines was verified by PCR using transgene-specific primers (Figure S5 and Table S1). Quantitative reverse-transcription PCR (qRT–PCR) showed a 70–80% knockdown of the candidate MYB gene in these transgenic lines (Figure 3G). We have also further verified that no other MIXTA-like genes were inadvertently knocked down in these RNAi lines (Figure S5), which was expected as the 339-bp region used in the RNAi construct is so divergent among the M. lewisii MIXTA-like paralogs that no obvious sequence similarity exists at the nucleotide level.

Figure 3

Characterization of the GUIDELESS RNAi transgenic lines. (A–D) Flower images of RNAi line 1, 4, 5, and 13, showing much reduced trichomes and carotenoid pigmentation in the nectar guides. (E and F) Aborted conical cell development on the petal lobes and aborted trichome elongation in the nectar guides of RNAi-4. Bars on the SEM micrographs, 50 μm. (G) qRT–PCR of GUIDELESS at the 10-mm corolla stage. GUIDELESS expression was knocked down 70–80% in the four RNAi transgenic lines. Bars, 1 SE from three biological replicates.

Taken together, all three lines of evidence led to the conclusion that GUIDELESS is a MIXTA-like R2R3 MYB gene necessary for the development of nectar guides in M. lewisii: (i) in the nine-gene interval mapped by the bulk segregant analysis, only this MIXTA-like gene contained a mutation that could severely interfere with protein function; (ii) the specific knockdown of this gene in the wild-type genetic background recapitulated the mutant phenotype; and (iii) elements of the guideless phenotype are consistent with known functions of previously characterized MIXTA-like genes in other plants.

It is worth noting that the guideless mutant does not produce completely flat cells on the petal inner epidermis (Figure 1D), nor does it completely lack trichomes in the nectar guides (Figure 1F). Instead, it produces less elaborated conical cells and very short, stumpy hairs. This indicates that the primary function of GUIDELESS is to promote cell elaboration (i.e., unidirectional cell expansion) once the outgrowth of a cell has been initiated, rather than determine cell fate in the first place. In this sense, GUIDELESS is functionally more similar to AmMYBML2/PhMYB1/AtMYB16 (Baumann et al. 2007) than MIXTA (Glover et al. 1998) or AmMYBML1 (Perez-Rodriguez et al. 2005), although phylogenetically GUIDELESS seems more closely related to the latter genes (Figure S4).

A somewhat similar phenotype with loss of yellow aurones and the mass of trichomes in the corolla throat has been described in the divaricata mutant of A. majus (Perez-Rodriguez et al. 2005). DIVARICATA is responsible for determining ventral petal identity (Galego and Almeida 2002) and is likely to directly regulate AmMYBL1, thereby restricting the expression of AmMYBL1 to the ventral petal (Perez-Rodriguez et al. 2005). The loss of trichomes in the corolla throat in divaricata is probably mediated by the down-regulation of AmMYBL1. The guideless mutant differs from divaricata in that the ventral petal identity is not affected in guideless, as the pair of ventral petal specific ridges is still present (Figure 1B). Furthermore, the DIVARICATA ortholog in M. lewisii is unlikely to directly regulate GUIDELESS, because GUIDELESS not only affects the formation of nectar guides in the ventral petal, but also regulates conical cell development in the dorsal and lateral petals (Figure 1).

Finally, an unexpected but intriguing observation of the guideless mutant is the loss of carotenoid pigmentation. This implies that GUIDELESS might directly regulate carotenoid production or deposition in the nectar guides, independent of unidirectional cell expansion during trichome and conical cell development. Alternatively, the loss of carotenoids could be the consequence of a defect in cell elaboration. Distinguishing these possibilities will require identifying the downstream target genes of GUIDELESS.

The GUIDELESS example highlights the potential of our collection of chemically induced M. lewisii mutants for contributing new knowledge of floral morphogenesis and diversification. The developmental genetics of many ecologically important floral traits (e.g., carotenoid pigmentation, corolla tube structure, touch-sensitive stigma, nectar concentration and volume, and various petal lobe ornaments) remains poorly understood, simply because these traits do not exist in the conventional plant genetic model system, A. thaliana. The induced mutants in M. lewisii furnish the raw materials to study these traits. Here we outline a general strategy to use these induced floral mutants for further rapid progress in understanding the genetic and developmental basis of floral trait variation. First, one can rapidly identify the genes underlying a particular M. lewisii mutant phenotype by bulk segregant analysis and manipulate the genes to study their function by stable transformation. Starting from these genes, one can discover other genes in the same genetic pathway/network by three complementary approaches: (i) characterizing nonallelic mutants with similar phenotypes; (ii) yeast two-hybrid screening to detect genes whose protein products physically interact with the newly discovered protein; and (iii) comparing transcriptomes of wild-type and mutants to identify downstream target genes. Once the genetic network underlying a particular floral trait (e.g., corolla tube formation and elaboration) is understood in Mimulus, one can then apply the Mimulus “gene toolbox” to dissect the developmental genetic basis of similar floral trait variation in nonmodel systems across the flowering plant phylogeny.

The defining characteristic of classical genetic model systems is the ability to go from phenotype to gene (or the reverse) with a high standard of experimental evidence. The advent of massively parallel DNA sequencing now makes it possible to develop—quickly and inexpensively—a sophisticated genetics/genomics toolkit for “emerging” model systems. Induced mutants have proven indispensable for unraveling genetic pathways and networks and must be part of the toolkit. Finally, stable transgenesis is required for rigorous testing of genetic hypotheses and precise characterization of developmental mechanisms. The rapid identification of the GUIDELESS gene through analyzing a chemically induced mutant, together with our recent work on fine dissection of the genetic basis of natural flower color variation between M. lewisii and M. cardinalis (Yuan et al. 2013), suggest that Mimulus is becoming such a “classical” genetic model system that is particularly suitable for studying flower diversification.

Acknowledgments

We are grateful to Brian Watson, James Vela, Doug Ewing, Jeanette Milne, Paul Beeman, and Erin Forbush for plant care. We also thank two anonymous reviewers for valuable comments on the manuscript. Piotr Mieczkowski at the University of North Carolina High Throughput Sequencing Facility supervised the Illumina sequencing. Wai Pang Chan (University of Washington Biology Imaging Center) helped with the scanning electron microscopy. The Mimulus guttatus genome sequencing consortium and Department of Energy Joint Genome Institute provided the chromosome-level assembly of the M. guttatus genome. This work was supported by National Science Foundation Frontiers in Integrative Biological Research grant 0328636 and National Institutes of Health grant 5R01GM088805.

Footnotes

Communicating editor: S. Poethig

Author notes

Supporting information is available online at http://www.genetics.org/lookup/suppl/doi:10.1534/genetics.113.151225/-/DC1.

Short read data have been deposited in the NCBI Short Read Archive (SRA056512); annotated GUIDELESS and other MIXTA-like gene sequences have been deposited in GenBank (KC139356 and KC692454–KC692460).