Divergence in flowering time is a key contributor to reproductive isolation between incipient species, as it enforces habitat specialization and causes assortative mating even in sympatry. Understanding the genetic basis of flowering time divergence illuminates the origins and maintenance of species barriers.
We investigated the genetics of divergence in critical photoperiod for flowering between yellow monkeyflowers Mimulus guttatus (outcrosser, summer flowering) and Mimulus nasutus (selfer, spring flowering). We used quantitative trait locus (QTL) mapping of F2 hybrids and fine-mapping in nearly isogenic lines to characterize the genomic regions underlying a > 2 h critical photoperiod difference between allopatric populations, and then tested whether the same QTLs control flowering time in sympatry.
We identified two major QTLs that almost completely explain M. nasutus's ability to flower in early spring; they are shared by allopatric and sympatric population pairs. The smaller QTL is coincident with one that differentiates ecotypes within M. guttatus, but the larger effect QTL appears unique to M. nasutus.
Unlike floral traits associated with mating system divergence, large interspecific differences in flowering phenology depend on only a few loci. Major critical photoperiod QTLs may be ‘speciation genes’ and also restrict interspecific gene flow in secondary sympatry.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
The timing of reproduction is a key determinant of both individual fitness and population success. In flowering plants, the optimal time for reproduction may depend on local abiotic factors, such as seasonal water availability, and biotic factors such as the availability of pollinators or density of seed predators (reviewed in Elzinga et al., 2007). Therefore, flowering phenology is an important axis for adaptive divergence among plant populations and species (Hall & Willis, 2006; Martin & Willis, 2007; Méndez-Vigo et al., 2011; Leinonen et al., 2012; Nakazato et al., 2013). Furthermore, because divergence in peak flowering time causes assortative mating, phenological shifts can directly cause pre-mating reproductive isolation. Thus, flowering time is a potential ‘magic trait’ in plants (Gavrilets & Vose, 2007) and understanding the genetic basis of divergence in flowering time and flowering cues is central to a full understanding of plant speciation (Rieseberg & Willis, 2007; Rieseberg & Blackman, 2010). Nonetheless, surprisingly little is known about the genetic architecture, much less the molecular genetic basis and evolutionary history, of interspecific differences in flowering phenology.
Flowering is a complex trait requiring an integrated developmental response to both internal and external conditions. In the molecular model plant Arabidopsis thaliana, the regulatory network controlling flowering is generally divided into multiple cue-dependent plus autonomous pathways funneling through a handful of key integrator genes (reviewed in Amasino, 2010), although these pathways are increasingly understood to be highly complex (Andrés & Coupland, 2012), and not independent from traits such as germination (Wilczek et al., 2010; Chiang et al., 2012). Photoperiod is among the most important cues for flowering in many plants, as it can be a reliable predictor of future seasonal environments (Amasino, 2010). Long-day plants, such as A. thaliana, require daylengths greater than a critical photoperiod (e.g. 14 h) to initiate floral development, whereas short-day species such as rice (Oryza spp.) require daylengths less than a critical photoperiod. Other taxa, such as the yellow monkeyflower Mimulus guttatus (Friedman & Willis, 2013), resemble long-day plants in requiring a daylength cue greater than some minimum threshold to flower, but vary in critical photoperiod across a broad range including both short (< 12 h) and long days.
Simple mutations in known flowering time pathway genes can underlie large differences in flowering behavior among natural populations. In A. thaliana, for example, allelic variation at a handful of loci explains much of the variation in flowering behavior across diverse A. thaliana accessions (Alonso-Blanco et al., 2009; Salome et al., 2011). Major quantitative trait loci (QTLs) have also been identified for inter-variety differences in flowering in rice (Yano et al., 2000) and for the evolution of early flowering in domesticated sunflower (Helianthus; Blackman et al., 2011) and peas (Pisum spp.) (Weller et al., 2012). By contrast, a massive quantitative genetic analysis in maize (Zea mays), a diverse outbred species, found a highly polygenic basis to standing variation for flowering phenology (Buckler et al., 2009). From these studies, it is clear that mutations with major effects on flowering time may be fixed by the strong selection of domestication; however, it is not yet clear whether they also contribute to the (often large) differences in flowering phenology that can reproductively isolate co-occurring species.
Here, we investigate QTLs underlying a striking difference in the critical photoperiod for flowering between the yellow monkeyflowers Mimulus guttatus and Mimulus nasutus. Mimulus guttatus (common yellow monkeyflower) is an extremely diverse, primarily outcrossing wildflower of wet soils across western North America (Wu et al., 2008). Mimulus nasutus, a highly selfing species that appears to be recently derived from within M. guttatus (Sweigart & Willis, 2003), co-occurs with annual ecotypes of M. guttatus in patches throughout their shared range. In addition to being divergent in mating system and associated floral traits (Fishman et al., 2002), M. nasutus tends to occupy drier (or more ephemerally wet) shallow-soiled microsites and to flower earlier than even annual M. guttatus. Hybridization does occur at some sympatric sites and there is also evidence of ongoing interspecific introgression (Sweigart & Willis, 2003), but divergence in flowering time is a key pre-mating barrier between these species (Kiang & Hamrick, 1978; Martin & Willis, 2007). In the glasshouse and field, M. nasutus accessions flower readily under daylengths as short as 9 h, whereas M. guttatus is highly variable but often requires at least 13–14 h of daylight to initiate flowering (Friedman & Willis, 2013). Thus, genetic differences in critical photoperiod are an important determinant of the realized differences in flowering phenology that contribute to reproductive isolation between M. nasutus and M. guttatus.
In addition to addressing the general question of whether a key plant speciation trait has diverged by major or minor steps, QTL mapping can begin to address fundamental questions about the repeatability of evolutionary transitions and the origins of interspecific variation. Recently, Friedman & Willis (2013) mapped major QTLs for both critical photoperiod and vernalization requirement differences between populations of M. guttatus. This close comparative context allows us to ask whether the genetic regions (and potentially genes) involved in the divergence of M. nasutus from an M. guttatus-like ancestor are novel or appear to be drawn from existing variation within M. guttatus. In addition, we can directly compare the genetic architecture of flowering time divergence to that of other species-diagnostic traits, such as highly polygenic floral traits associated with the evolution of selfing (Fishman et al., 2002). Finally, by identifying strong positional candidate genes with fine-mapping, we can assess whether genes or gene families implicated in flowering time variation in other taxa are also likely to be associated with critical photoperiod divergence between species of Mimulus.
We focus mapping efforts on a well-characterized, highly divergent pair of lines from allopatric Oregon populations: Iron Mountain M. guttatus (IM62) and Sherar's Falls (SF) M. nasutus. The IM62 line was derived from a high-elevation annual population that flowers in a narrow window between snowmelt and summer drought (June–July); it requires 13–14 h of daylight to flower (Friedman & Willis, 2013; L. Fishman pers. obs.) By contrast, all M. nasutus accessions will flower under 10-h days (Friedman & Willis, 2013), consistent with early spring flowering on substrates with very ephemeral water availability (including the spray-misted basalt cliffs from which SF was collected). We first used bulk segregant analysis and targeted genotyping of large SF × IM62 F2 populations (replicated in 2 yr) to map two major loci that together confer M. nasutus's ability to flower under early spring daylengths (< 12-h days). Secondly, we used nearly isogenic lines (NILs) to confirm and fine-map those QTLs to physically contiguous scaffolds of the M. guttatus draft genome, allowing us to identify strong functional and positional candidate genes for the interspecific divergence. We then extended our analysis to a pair of sympatric populations, and used targeted mapping to demonstrate that the two major loci identified in our allopatric cross are also the primary cause of flowering cue divergence in sympatry.
Materials and Methods
Study system and plant material
The Mimulus guttatus DC species complex (Phrymaceae) is a morphologically and ecologically diverse group of partially cross-fertile taxa. The outcrossing yellow monkeyflower M. guttatus is found from the Aleutian Islands to Baja California and east to the Rocky Mountains. Mimulus nasutus Greene, which generally self-pollinates before corolla opening, is the most widespread of several selfing taxa derived from M. guttatus, and has a broadly overlapping range. We used inbred lines derived from collections at three sites. The high-elevation Iron Mountain, OR (IM) M. guttatus population consists of bee-pollinated annuals that germinate in fall (overwintering under snow) or late spring, flower at mid-summer, and die by early August. The IM62 line (5+ generations inbred) was used to generate the M. guttatus reference genome (www.Phytozome.org). The Sherar's Falls, OR (SF) M. nasutus line was collected from basalt rocks flanking a waterfall on the Deschutes River, is naturally inbred, and has been maintained by inbreeding. Hybrids of SF and IM62 lines were previously used for linkage and QTL mapping (Fishman et al., 2001, 2002), and for genetic characterization of postmating prezygotic (Fishman et al., 2008) and postzygotic (Fishman & Willis, 2005; Sweigart et al., 2006; Case & Willis, 2008; Barr & Fishman, 2010) barriers. The Catherine Creek, WA (CAC) populations occupy seasonal seeps on the north side of the Columbia River Gorge (< 200 km from SF and IM). The two species co-occur in numerous patches throughout this area, but M. nasutus tends to occupy rockier microsites (A. Kenney & A. Sweigart, pers. obs.)
Phenotypic and bulk segregant analyses (2009)
In Spring 2009, we conducted a bulk segregant analysis of an SF × IM62 F2 population grown under short days in our glasshouse at the University of Montana. Seeds were planted into 96-well flats containing Sunshine #1 soil less potting mix (Sungro Horticulture, Agawam, MA, USA) on 10 December 2008, chilled for 4 d at 4°C to promote germination, and then placed in a glasshouse with a 12 : 12 h day : night cycle. Daytime temperatures were c. 20–28°C and nighttime temperatures were c. 13–15°C. Supplemental light (600-W high-pressure sodium Philips Son-T Agro bulbs; Philips North America, Andover, MA, USA) was provided during the 12-h day, and plants were bottom-watered daily. After germination, seedlings were thinned to 1 per well. Parental controls (n = 24 each) were randomized in a single flat and the control flat position within the bench of F2 flats was rotated weekly. Plants were checked for floral buds weekly until 20 March 2009.
Tissue was collected from a subset of early-flowering plants (buds before 15 March; n = 96) and from nonflowering plants (no buds by 20 March; n = 356) for selective genotyping. DNA was extracted using a standard CTAB-chloroform protocol modified for 96-well format. For bulk segregant analysis, we constructed four flowering pools and four nonflowering pools (n = 4 individuals per pool). The eight pools were genotyped at 96 previously mapped MgSTS (MimulusguttatusSequence Tagged Site) markers, which are exon-primed and amplify intron-length polymorphisms (e-prefix in text and figures; for details on MgSTS markers, see www.mimulusevolution.org). Four to six MgSTS markers (forward primer 5′ fluorescent-labeled) from each of the 14 linkage groups were assembled into multiplexes and amplified using standard touchdown PCR protocols. The PCR products were run on an ABI 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA) with an in-lane size standard and informative polymorphisms were scored using Genemapper software (Applied Biosystems) and verified by eye. Markers were flagged as positives if two or more of the four flowering pools were homozygous for M. nasutus alleles, but nonflowering pools appeared heterozygous. Positive markers, plus additional linked markers on two linkage groups, were then genotyped in the F2 hybrids (n = 452) following the same protocols.
QTL mapping (2010)
In spring 2010, we grew a second F2 population (n = 768) for targeted QTL mapping. Growth conditions were similar (supplemental light and 12-h daylengths), but we included seven additional inbred lines from the IM M. guttatus population (n = 4 each) along with parental line controls (IM62, n = 12; SF, n = 8). Seeds were planted on 5 January and plants were scored as flowering if they initiated buds by 21 March, when natural daylengths reached 12 h. Supplemental light was then provided for 16 h per day and plants monitored for flowering for 100 d.
F2 hybrids were genotyped at MgSTS markers (e-prefix) as previously, and at new markers designed in the targeted QTL regions on LG7 and LG8 (m-prefix). Briefly, we used the draft genome build of M. guttatus (www.Phytozome.org, V2.0), plus whole-genome resequence data for the M. nasutus SF line (aligned to the IM62 reference with bwa; Li & Durbin, 2009) to design exon-primed markers spanning intron indel polymorphisms between SF and IM62. These markers (Supporting Information Table S1) were amplified and genotyped using the same PCR and scoring protocols as for existing MgSTS markers. For QTL mapping, flowering versus not flowering by 21 March was coded as a binary trait (0 versus 1). We conducted QTL mapping in Windows QTL Cartographer, using the composite interval mapping module (CIM; window size 10 cM; three cofactors). We also used the categorical multiple interval mapping function in QTL Cartographer (Li et al., 2006), but the results were nearly identical to those for CIM; we present the standard CIM analysis here. To test for interactions between the major QTLs, we used logistic regression in jmp 10 (SAS Institute, Cary, NC, USA), with the two markers closest to the QTL peaks and their interaction as fixed effects.
We previously generated two sets of NILs with M. guttatus (IM62) and M. nasutus (SF) introgressions in the opposite genetic background (Fishman & Willis, 2005; Sweigart et al., 2006). Each fourth-generation NIL (M. nasutus background, BN4; M. guttatus background, BG4) carries a unique complement of heterozygous introgressions embedded in a genome that is expected to be 93.75% homozygous for parental alleles. To catalogue introgressed regions, we genotyped all NILs at MgSTS markers distributed throughout the genome (L. Fishman et al., unpublished). NILs were then selfed to generate homozygous introgressions as necessary. We selected three NILs with introgressions spanning the LG7 and/or LG8 flowering time QTLs for further genetic analyses. For two of these NILs, M. nasutus is the recurrent parent: BN4.216 is heterozygous at the LG7 QTL and BN4.37 is homozygous for M. guttatus alleles at the LG8 QTL. The remaining NIL, BG4.327, has M. guttatus as the recurrent parent and is heterozygous at both the LG7 and LG8 QTLs. For each NIL, we grew selfed progeny under the same conditions as the 2010 F2 grow-out and scored plants for floral bud initiation. We genotyped all selfed progeny (or, for BN4.37, a small subset of the selfed progeny) for markers linked to the LG7 and LG8 QTLs.
Targeted mapping in sympatric M. guttatus and M. nasutus hybrids
We grew reciprocal F2 hybrids of M. guttatus (CAC6G, three generations inbred) and M. nasutus (CAC9N, naturally inbred) lines from the sympatric Catherine Creek populations. We sowed seeds directly into 96-well flats containing Fafard® 3B soilless potting mix (Sungro Horticulture) on 9 February 2012, chilled them for 7 d at 4°C to promote germination, and then placed them in the University of Georgia Plant Biology glasshouse with a 12 : 12 h day : night cycle. Daytime temperatures were c. 17–24°C and nighttime temperatures were c. 13–17°C. We provided supplemental light (400-W high-pressure sodium bulbs) during the 12-h day and bottom-watered plants daily. After germination, we thinned seedlings to one per well. Reciprocal F2s were grown in separate sets of flats (n = 368 and 460 F2s with the M. nasutus and M. guttatus cytoplasmic backgrounds, respectively). Parental controls (one each per flat) were randomly dispersed within each flat and in additional F2 flats not used for mapping (total n = 21 M. nasutus and 26 M. guttatus).
We scored plants every other day for floral bud initiation, beginning on 17 March when natural daylength was c. 12 h. We considered plants that had initiated floral buds by 14 April (when natural daylength was c. 13 h) as flowering under short days and scored all other plants as not flowering. At Catherine Creek, M. nasutus begins flowering in late March/early April when daylength is c. 12.5 h (A. Kenney, pers. obs.). Therefore, the conditions in this experiment are relevant to variation found within this population. To test for an effect of the LG7 and LG8 QTLs on flowering under short days, we genotyped the F2s at markers flanking each QTL. We extracted DNA and genotyped plants at a combination of MgSTS and custom-designed markers, except that PCR products were run on an ABI 3730 Genetic Analyzer at the University of Georgia, and polymorphisms scored using Genemarker software (Soft Genetics, State College, PA, USA). For analyses, we exclusively used F2 individuals without recombination between the flanking markers. We performed χ2 tests in jmp pro 10 (SAS Institute) on the F2 hybrids to test for associations between QTL genotype and short-day flowering.
Phenotypic differentiation and bulk segregant analysis (2009)
In both the 2009 and 2010 grow-outs, all SF M. nasutus controls flowered under <12-h days, whereas no IM62 M. guttatus or other IM lines flowered under spring conditions. In 2009, two markers near the center of LG7 (MgSTS markers e34 and e251; both with three of four flowering pools homozygous for the SF allele) and two markers near the bottom of LG8 (e829 and e621; with three and two flowering pools homozygous for the SF allele, respectively) were flagged as positives, as was one marker (e779; two SF homozygous flowering pools) on LG12. The LG12 marker, however, was not associated with flowering in the larger genotyped data set (Pearson χ2; P =0.49; n = 427), so we focused on LG7 and LG8 markers for further genotyping. Single marker analyses of 10 markers across LG7 and four on the bottom of LG8 indicated major QTLs in each region, with e34 on LG7 (likelihood ratio; LR = 64.3) and e294 (LR = 71.8) most strongly associated with 12-h flowering. These two markers together explained 27% of the variance in the selected subset of F2s, with 70.0% of NN double homozygotes and 0.0% of GG double homozygotes flowering.
QTL mapping (2010)
In 2010, the LG7 (LR = 177.0; r2 = 0.23) and LG8 (LR = 46.8; r2 = 0.065) QTLs were again highly significant (Fig. 1). At the LG7 QTL, the M. nasutus allele was partially recessive (a = −0.28; d = −0.18; both P <0.01), whereas the LG8 QTL appeared nearly additive (a = −0.23; d = −0.05). Together, the two QTLs explained > 95% of the species difference in the response to flowering (Fig. 2), but there was no evidence of epistasis (genotypic interaction P >0.4 in logistic regression). That is, the ability to flower under short days depended on the additive action of alleles across the two loci, with c. 50% of NG (LG7, M. nasutus; LG8, M. guttatus) individuals flowering under 12-h days, but 96% of NN F2s flowering. Conversely, c. 25% of plants homozygous for M. nasutus alleles at the LG8 QTL (but homozygous for M. guttatus alleles at LG7) flowered, indicating that its effects are at least partially independent from the genotype at the more major LG7 QTL (Fig. 2).
As in previous iterations of the SF × IM62 F2 population grown under long days (Fishman et al., 2002, 2008), we observed transmission ratio distortion (TRD) toward excess M. nasutus alleles at LG7 markers and Mendelian segregation in the focal LG8 region. In this cross, multiple gametic mechanisms of TRD, including centromere-associated meiotic drive (LG11; Fishman & Willis, 2005) and competition among haploid pollen genotypes (c. 10 loci; Fishman et al., 2008), generally cause M. guttatus over-transmission. One explanation for the local TRD favoring M. nasutus would be differences in seed germination attributable to F2 genotype at the LG7 flowering QTL, as the environmentally dependent genetic pathways for flowering and germination partially overlap, at least in the Brassicaceae (Chiang et al., 2009). In that case, we would expect TRD to be strongest at the QTL peak (near e34) and to exhibit a pattern consistent with zygotic selection (e.g. reduced germination of M. guttatus homozygotes) rather than gametic selection. Instead, the frequency of the M. nasutus allele was highest at e741 (distal; 0.583; n = 644), lowest at e376 (proximal; 0.482; n = 443) and intermediate at e34 (0.565; n = 663). In addition, genotypic frequencies (N : H : G) at e34 were virtually identical to the expected values given the skewed allele frequencies (P > 0.5). Together, this suggests that M. nasutus-biased transmission on LG7 is not a pleiotropic effect of the major critical photoperiod QTL, but instead reflects gametic distortion at a more distal locus. Notably, the distal end of LG7, beyond e741, contains the dominant nuclear restorer (R f ) of cryptic mitochondrial male sterility in IM M. guttatus (Barr & Fishman, 2010).
Fine-mapping of QTLs in NILs (2011)
To refine the positions of the LG7 and LG8 QTLs and characterize their individual phenotypic effects, we grew the selfed progeny of three independent NILs under similar short-day glasshouse conditions. One of these lines, BG4.327, is doubly heterozygous for introgressions spanning the LG7 and LG8 QTLs in an otherwise M. guttatus genetic background. Only one of 23 selfed progeny from this line flowered under short days. Genotyping revealed that this individual alone carried homozygous M. nasutus introgressions for both flowering time QTLs (Fig. 3), underscoring their necessary roles in critical photoperiod divergence. Moreover, the recombination breakpoints on LG7 coincided precisely with the QTL identified in the F2 hybrids: the proximal breakpoint occurred between m9.01 and m114, whereas the distal breakpoint occurred between e369 and e259. For LG8, the proximal breakpoint in BG4.327 was located at the edge of the QTL, so did not provide any further refinement of this QTL.
The BN4.216 line carries a heterozygous introgression spanning the LG7 QTL in a nearly isogenic (including the LG8 QTL) M. nasutus genetic background. As expected, segregation of the LG7 introgression among the selfed progeny additively affected the propensity to flower under short days: 97.1% of M. nasutus homozygotes flowered, compared with only 56.9% of heterozygotes and 10.8% of M. guttatus homozygotes (n = 941). Because the M. nasutus allele is nearly completely penetrant against this isogenic background, recombinational mapping allowed us to place a distal boundary for the LG7 QTL at m110 (Fig. 3). The proximal boundary was at m9.01, consistent with mapping results above (Figs 1, 3). Despite a large mapping population (n =941), there were too few recombinants to further refine the LG7 QTL. Indeed, we observed a severe reduction in recombination in this region among the BN4.216-selfed progeny compared with our 2010 F2 population (3.5 cM versus 30 cM between e741 and e251), consistent with previous findings for heterospecific Mimulus NILs (Sweigart et al., 2006).
None of 84 selfed progeny of BN4.37 – a line that carries a homozygous M. guttatus introgression spanning the LG8 QTL in an M. nasutus (including the LG7 QTL) genetic background – flowered under short days. To define the introgression breakpoints in this line, and thereby refine the QTL it contains, we genotyped eight of these selfed progeny for additional markers designed from the genome sequence assembly. This approach localized the recombination breakpoint to between m12545 and m13970, placing a distal boundary on the LG8 QTL (Fig. 3).
Candidate gene identification
The LG7 QTL spans 2.5–3 cM (1–2 LOD drop, respectively) centered near marker e34, and the NIL data also restrict it to this region. The 5.35 cM region between flanking m114 and e369 markers corresponds to c. 530 kb of the V. 2.0 M. guttatus genome, indicating that 1 cM equals c. 100 kb. Thus, the QTL encompasses c. 250–300 kb containing 36–42 annotated genes (Fig. 3). Forward searches for Mimulus homologs of A. thaliana genes with functional annotations related to flowering identified only two loci on LG7 as possible candidates; both are near the QTL region. The only Mimulus match to Cryptochome 2 (CRY2; Migut.G00510) is adjacent to the gene containing e251, making it a positional candidate as well as a strong functional candidate for shifts in critical photoperiod (El-Assal et al., 2001). However, flanking markers m9.01 and m9.03 place this gene well outside of the QTL confidence interval (Fig. 1) and the NIL breakpoints exclude it (Fig. 3). The other candidate (Migut.G00570, a homolog of FLOWERING LOCUS T) is predicted to be physically coincident with the peak of the LG7 QTL (Fig. 3); this location was genetically confirmed by mapping an SF-IM62 intron-length polymorphism in Migut.G00570 as a marker (m26170) in the 2010 F2 mapping population (Fig. 1).
The one and two LOD confidence intervals of the LG8 QTL spanned c. 6.5 and 17 cM, respectively (Fig. 1), but we were able to truncate the distal edge of the QTL with the BN4.37 NIL, a nonflowering introgression line that is M. nasutus homozygous for LG8 markers beyond m12545. This limited the LG8 QTL to the 300–500 kb interval between m41 or m12 (1 or 2 LOD drop) on the proximal side and the NIL breakpoint between m12545 and m13970 on the distal side (Fig. 3). This region contains c. 60–70 genes, including a homolog of GIBBERELLIC ACID INSENSITIVE (GAI; Migut.H02266) in the larger interval and, directly under the QTL peak, a cluster of three genes with homology to the MADS-box gene SHORT VEGETATIVE PHASE/AGAMOUS-LIKE 24 (SVP/AGL24; Migut.H02293, Migut.H02296, and Migut.H02298).
Assessment of LG7 and LG8 QTLs in hybrids of sympatric M. guttatus and M. nasutus
Catherine Creek M. nasutus and M. guttatus parental lines strongly differed in their ability to flower under short days (90.4% versus 19.2% flowered, respectively), although less so than the allopatric inbred parental lines. Some early flowering by the M. guttatus parent, in particular, may reflect the later assay for flowering in this experiment relative to the SF × IM62 experiments. Consistent with the 2010 SF × IM62 F2 results, 39% (324/828) of sympatric F2s initiated flower buds under short days. The reciprocal F2 groups flowered at the same overall rate (40% versus 38% for M. nasutus and M. guttatus cytoplasmic backgrounds, respectively; χ2 = 0.185; P =0.67) and showed the same expression of our focal QTL effects. In the full data set, both the LG7 and LG8 QTLs significantly affected floral bud initiation under short days (Fig. 4; LG7, χ2 = 53.5; P <0.0001; LG8, χ2 = 19.2; P <0.0001). Together, the two QTLs almost recapitulate the Catherine Creek parental phenotypes − 18% of the F2s homozygous for the M. guttatus allele at both QTLs flowered under short days and 73% of the individuals homozygous for M. nasutus at both QTLs flowered (Fig. 4); the pattern of inheritance is also very similar to that of the allopatric line cross (Fig. 2). These results demonstrate that the same two large-effect QTLs are important for flowering time divergence between both allopatric and sympatric populations of M. guttatus and M. nasutus.
We identified two major QTLs that together explain divergence between outcrosser M. guttatus and selfer M. nasutus in the ability to flower under short days, a key adaptive trait that also contributes to reproductive isolation. We fine-mapped the most important critical photoperiod QTL, on LG7, to a < 300-kb region containing a single candidate gene. This genomic region was not associated with critical photoperiod variation within M. guttatus in a previous study (Friedman & Willis, 2013), suggesting that the genetic basis of early spring flowering in M. nasutus may be novel. The smaller QTL, on the bottom of LG8, was fine-mapped to a region of c. 500 kb containing multiple candidates. In contrast to the LG7 QTL, this region also contains QTLs for critical photoperiod and vernalization requirement (Friedman & Willis, 2013), as well as flowering time under long days and other life-history traits within M. guttatus (Hall et al., 2006, 2010; Lowry & Willis, 2010). Analysis of M. guttatus-background NILs suggests that M. nasutus alleles at both loci are necessary and sufficient for the switch to obligate short-day flowering, and thus may act as major speciation genes. Nonetheless, the additive nature of both loci indicates that intermediate genotypes could still have been visible to natural selection for early flowering during the divergence of M. nasutus. Finally, the same QTLs identified in our cross of allopatric populations also fully account for differences in photoperiodic responses between M. nasutus and M. guttatus from a sympatric site, indicating that these QTLs may be important for adaptation species-wide and act as species barriers in areas of contact.
Genetic architecture of interspecific differences in critical photoperiod
QTL mapping can both characterize broad patterns of genetic architecture of complex traits and identify genomic regions containing candidate molecular variants for further evolutionary analysis. Our bulk segregant and targeted mapping approach primarily focused on the former goal, as it can only detect major QTLs; however, our results do rule out a polygenic architecture for critical photoperiod divergence. This contrasts with the genetic architecture of floral traits in the same SF M. nasutus and IM62 M. guttatus cross (Fishman et al., 2002). In that study, only three of 24 QTLs underlying floral traits accounted for more that 20% of the species difference in a given trait and F2 hybrids did not recapitulate parental values, suggesting that M. nasutus's transition to obligate self-fertilization was gradual, characterized by a ‘moving optimum’, and/or drawn from the abundant standing variation for mating system traits present within M. guttatus (Bodbyl Roels & Kelly, 2011). Our finding of a relatively small set of major mutations underlying short-day flowering suggests that M. nasutus's ability to occupy habitats with transient water availability may have been a relatively abrupt transition.
Despite being measured as a discrete trait and only involving two major QTLs, short-day flowering appeared surprisingly quantitative. That is, although the M. nasutus allele at the larger LG7 QTL is partially recessive, the total dose of M. nasutus alleles cumulatively affected the probability that an individual crossed the threshold and initiated flowering (Figs 2, 4). This inheritance pattern is very similar to that seen for a pair of critical photoperiod QTLs (LG8a and LG8b) distinguishing annual and perennial M. guttatus populations from northern California (Friedman & Willis, 2013). The additivity and lack of epistasis we see in F2 hybrids suggest that alleles promoting short-day flowering could have been readily recruited by natural selection. However, genetic background effects may also have been important during initial spread, as M. nasutus homozygosity at both loci was necessary for short-day flowering in the M. guttatus background BG4.327 progeny. In addition, in both SF × IM experiments, all M. nasutus controls flowered before even the fastest flowering hybrids, indicating that loci beyond our critical photoperiod QTLs may modulate the speed of flowering under short-day conditions.
Candidate genes in the context of environmental cues
The network of genes associated with photoperiod-dependent flowering in A. thaliana often also affect flowering phenology in other taxa, providing a clear pool of initial functional candidates (Hayama & Coupland, 2004). However, the traditional divisions of the flowering time gene network into discrete, cue-specific pathways (e.g. photoperiod and vernalization) are beginning to blur. For example, the flowering behavior of mutants (Wilczek et al., 2010) and mapping populations (Anderson et al., 2012a) assayed in growth chambers is not always replicable in the field. Our experiments were conducted in a glasshouse under naturally fluctuating day–night temperature conditions similar to those experienced by plants growing under short days in the field (i.e. cool nights and warm days). Thus, although we cannot pin interspecific differences in flowering time on photoperiod sensitivity independent of temperature, our results may better capture natural flowering time variation. Most notably, Friedman & Willis (2013) reported little or no difference in critical photoperiod between M. guttatus and M. nasutus lines from Catherine Creek grown in growth chambers at constant (21°C) temperature. However, these populations flower at distinct times in the field (A. Kenney & A. Sweigart, pers. obs.) and our experiment, and hybrids segregated for major critical photoperiod QTLs (Fig. 4). Because Catherine Creek populations are potentially hybridizing, introgression may contribute to genetic differences between the lines used in the two studies. However, the constant temperatures used in the earlier study may also have masked some ecologically relevant divergence in critical photoperiod that our experiment revealed.
For each QTL, we identified strong positional and functional candidates that will be accessible to targeted molecular characterization. The 35–45 genes under the major LG7 QTL included only one flowering time network candidate (Migut.G00570), which is annotated as the first two (of four expected) exons of a phosphatidylethanolamine-binding protein (PEBP) in the FLOWERING LOCUS T/TERMINAL FLOWER 1 (FT/TFL1) family. In A. thaliana, FT integrates signals to promote flowering, whereas TFL1 represses flowering (Pin & Nilsson, 2012); however, the expression patterns and interactions of FT/TFL1 family members can transcend simple repressor/promoter dichotomies (Ballerini & Kramer, 2011). Members of the FT/TFL1 family have been implicated in natural variation in flowering phenology, including photoperiod- and temperature-dependent responses, in diverse taxa (Blackman et al., 2010; Pin et al., 2010; Anderson et al., 2012b; Koskela et al., 2012; Grillo et al., 2013; Nitcher et al., 2013). The M. guttatus reference genome contains eight full FT/TFL1 family sequences (all four exons), as well as multiple, unlinked, partial sequences. Migut.G00570 may be only a partial PEBP sequence because it is misassembled, but appears to be expressed as annotated in M. guttatus (V2.0 annotation; www.Phytozome.org). Even if it does not encode a complete FT/TFL1 sequence, however, Migut.G00570 may actively regulate flowering. For example, an FT paralog with a frameshift mutation was recently shown to underlie a major photoperiod QTL in sunflower, indicating that apparently ‘nonfunctional’ FT-like genes can affect flowering phenology by interfering with other FT-like loci (Blackman et al., 2010). Molecular population genetic analyses of the FT/TFL1 family in Mimulus, as well as detailed characterization of polymorphism and expression patterns of the LG7 QTL candidate, hold great promise for understanding flowering time evolution in yellow monkeyflowers and beyond.
The LG8 QTL contained two functional candidates, a homolog of GAI under the left edge of the 2-LOD drop confidence interval and a cluster of three SVP/AGL24 family genes under the peak. In A. thaliana, GAI and related proteins block gibberellic acid signaling and repress flowering under short days (Wilson et al., 1992). SVP is a floral signal integrator that interacts with genes across flowering time pathways and directly regulates FT. Recently, a single amino acid substitution causing loss of SVP functionality was pinpointed as the source of a dramatic shift to early flowering under short-day conditions in some Asian accessions of A. thaliana (Méndez-Vigo et al., 2013). Thus, the cluster of three SVP-like genes under our LG8 QTL, which also overlaps with vernalization and photoperiod QTLs mapped within M. guttatus (Friedman & Willis, 2013), is an attractive target for further investigation of within- and among-species flowering time variation in the M. guttatus complex.
Evolutionary history and consequences for gene flow
To understand the role of our flowering time QTLs in speciation, we must consider both their evolutionary history within the M. guttatus species complex and their consequences for gene flow on secondary contact. In the first case, it is particularly important to understand the directionality of genetic and phenotypic divergence. That is, where did the M. nasutus LG7 and LG8 QTL alleles come from, and what (if any) role did they play in the initial divergence and reproductive isolation of M. nasutus? Phylogeographic data suggest that M. nasutus was derived once from within highly diverse and paraphyletic M. guttatus (Sweigart & Willis, 2003; Modliszewski & Willis, 2012). Its ancestor is likely to have been an annual, as M. nasutus is collinear with IM62 and other widespread annual M. guttatus ecotypes rather than perennials (Fishman et al., 2013). However, the geographical and elevational origin of M. nasutus is not yet clear. One plausible scenario is that M. nasutus evolved from a spring-flowering annual M. guttatus found at lower latitudes and altitudes than IM M. guttatus. In that scenario, the short critical photoperiod of M. nasutus could pre-date the origin of the species, and the causal QTLs would be shared with early-flowering M. guttatus. For the major LG7 QTL, however, this scenario is unlikely: Friedman & Willis (2013) found no evidence of LG7 involvement in critical photoperiod differences between LMC (a northern California annual M. guttatus with the same 10-h critical photoperiod as SF M. nasutus) and SWB (a northern California perennial with a > 13-h critical photoperiod). Therefore, the M. nasutus early-flowering LG7 QTL allele does not appear to be the common source of M. guttatus short-day flowering, and thus may be a unique adaptation to the extremely ephemeral habitats occupied by M. nasutus. Identification of the underlying molecular polymorphism, and characterization of its evolutionary history across the species complex, will be necessary to test this hypothesis. By contrast, the LG8 QTL is shared with within-M. guttatus QTLs for both vernalization requirement and critical photoperiod (Friedman & Willis, 2013). This co-localization may be coincidental, as multiple candidate genes lie under this broad QTL and the causal polymorphisms have not yet been identified. However, it is possible that the LMC-SWB LG8b photoperiod QTL of Friedman & Willis (2013) and our LG8 QTL share a common genetic basis. In that case, the more powerful M. nasutus-specific LG7 allele may have initially arisen in a relatively LMC-like M. guttatus background (e.g. one fixed for the early-flowering allele at the shared LG8 QTL), and thus could have spread rapidly when and where earlier spring flowering was favored.
Regardless of their origins, our QTLs probably contribute to current reproductive isolation between M. nasutus and M. guttatus.This may occur most importantly, but cryptically, by dictating habitat suitability on both regional and local scales (Sobel et al., 2010). By initiating flowering under < 12-h days, M. nasutus can reproduce in shallow-soil sites that dry out by late spring. By contrast, our M. guttatus accessions delay flowering and are thus restricted to less ephemeral sites that remain moist through midsummer. Thus, a short critical photoperiod is a key component of M. nasutus's drought avoidance strategy (Wu et al., 2010; Ivey & Carr, 2012). By facilitating habitat specialization, as well as by promoting asynchronous flowering where the two species do co-occur, M. nasutus's early flowering may be as important an isolating mechanism as autonomous self-fertilization. Ongoing investigations of phenotypic variation and genome-wide introgression patterns at Catherine Creek have the potential to reveal how the QTLs found in this study contribute to phenological behavior in nature, and whether (and under what environmental conditions) they promote assortative mating.
This work on the genetic basis for early spring flowering in selfer M. nasutus relative to congener M. guttatus provides insight into mechanisms of plant adaptation and speciation, particularly in the context of work on other traits in the same taxa (Fishman et al., 2002) and parallel transitions in the same trait (Friedman & Willis, 2013). Mimulus nasutus appears to have taken a novel genetic path to adapt its flowering time to extremely early-drying habitats, and it will be interesting to explore whether this pattern is repeated in other parallel transitions. For example, we recently described a population of M. guttatus from extreme thermal soils in Yellowstone National Park that appears to have independently evolved short-day flowering, as well as annuality and self-pollination, from the local perennial ecotype (Lekberg et al., 2012). Further study of the genetic and molecular basis of flowering variation across the M. guttatus complex holds great promise for understanding the causes and consequences of change in this evolutionarily labile and ecologically important, but genetically simple, trait.
We thank T. Amsler, C. M. Barr, N. Batora, D. Crowser, A. Garner, H. Garner, R. Hughes and J. Mason and the University of Georgia glasshouse staff for assistance with plant care, phenotyping, and genotyping. Funding was provided by NSF grants DEB-0846089 and DEB-0918902 to L.F. and by a grant from the University of Georgia Research Foundation to A.S. S.C. was supported by the Montana Integrative Learning Experience for Students (MILES) program, through a grant to The University of Montana from the Howard Hughes Medical Institute—Undergraduate Science Education Program; grant no. 52005905.