Despite the potential importance of divergent reproductive phenologies as a barrier to gene flow, we know less about the genetics of this factor than we do about any other isolating barrier. Here, we report on the genetic architecture of divergent flowering phenologies that result in substantial reproductive isolation between the naturally hybridizing plant species Iris fulva and I. brevicaulis. I. fulva initiates and terminates flowering significantly earlier than I. brevicaulis. We examined line crosses of reciprocal F₁ and backcross (BC₁) hybrids and determined that flowering time was polygenic in nature. We further defined quantitative trait loci (QTL) that affect the initiation of flowering in each of these species. QTL analyses were performed separately for two different growing seasons in the greenhouse, as well as in two field plots where experimental plants were placed into nature. For BCIF hybrids (BC₁ toward I. fulva), 14 of 17 detected QTL caused flowering to occur later in the season when I. brevicaulis alleles were present, while the remaining 3 caused flowering to occur earlier. In BCIB hybrids (BC₁ toward I. brevicaulis), 11 of 15 detected QTL caused flowering to occur earlier in the season when introgressed I. fulva alleles were present, while the remaining 4 caused flowering to occur later. These ratios are consistent with expectations of selection (as opposed to drift) promoting flowering divergence in the evolutionary history of these species. Furthermore, epistatic interactions among the QTL also reflected the same trends, with the majority of epistatic effects causing later flowering than expected in BCIF hybrids and earlier flowering in BCIB hybrids. Overlapping QTL that influenced flowering time across all four habitat/treatment types were not detected, indicating that increasing the sample size of genotyped plants would likely increase the number of significant QTL found in this study.

---

EARLY stages of speciation usually involve the evolution of numerous prezygotic and postzygotic isolating barriers that act in concert to reduce the chance for gene flow between genetically and phenotypically divergent populations. Recent, detailed experiments in which the relative contributions of numerous prezygotic and postzygotic barriers were quantified consistently revealed prezygotic barriers to be the strongest impediments to interspecific gene flow (RAMSEY et al. 2003; HUSBAND and SABARA 2004; HURT et al. 2005; WHITEMAN and SEMLITSCH 2005; MARTIN and WILLIS 2007). If divergence between organisms often occurs simultaneously with gene flow (see ARNOLD 2006 for a review), the strength of such prezygotic barriers may be fundamentally important for the establishment of new evolutionary lineages. Given (i) the sequential nature of reproductive barriers and (ii) that phenological isolation (isolation due to the timing of reproduction of the two species) occurs first in the sequence (e.g., MARTIN and WILLIS 2007), this factor has the potential to be the most important barrier preventing the formation of viable hybrids. However, despite this importance, COYNE and ORR (2004, p. 21) concluded "We know less about the genetics of temporal [phenological] isolation than about the genetics of any other isolating barrier."

Quantitative trait locus (QTL) mapping studies have been utilized extensively as a means for dissecting the genetic architecture of a number of postzygotic barriers (e.g., SLOTMAN et al. 2004; MARTIN et al. 2005; MOYLE and GRAHAM 2005; MOEHRING et al. 2006a), due largely to the fact that hybrid inviability and hybrid sterility are much more tractable under laboratory conditions than are most prezygotic barriers. Fewer studies of the genetic architecture of prezygotic isolation have been carried out, since prezygotic barriers must often be studied using more labor-intensive field conditions. Exceptions to this include studies of mating displays involved in sexual isolation in the laboratory between insect species (e.g., Drosophila: GLEASON et al. 2002; GLEASON and RITCHIE 2004; MOEHRING et al. 2004, 2006b), as well as laboratory studies examining divergent floral displays believed to be important in promoting pollinator isolation in flowering plants (e.g., Mimulus: BRADSHAW et al. 1995, 1998; FISHMAN et al. 2002; Aquilegia: HODGES et al. 2002). However, we have been unable to find any reports of studies examining the genetic architecture of phenological isolation.

In this study, we examined F₁ and reciprocal backcross populations of two closely related, and naturally hybridizing, species of Iris, Iris fulva and I. brevicaulis, to determine whether or not the phenotypic expression of flowering time was polygenic in nature and to test for QTL responsible for this flowering difference. These two species are broadly sympatric throughout the North American Mississippi River drainage system (VIOSCA 1935). In sympatric populations, divergence of flowering phenology acts as a major barrier for F₁ hybrid formation (CRUZAN and ARNOLD 1994). In this regard, I. fulva initiates flowering 1 month earlier than I. brevicaulis in southern Louisiana populations (CRUZAN and ARNOLD 1994). Consistent with this observation, few F₁ genotypes have been found in areas of sympatry. However, when F₁ crosses are made by artificial means, the resulting progeny are vigorous and completely fertile, suggesting that prezygotic barriers (including flowering phenology) are largely responsible for the paucity of F₁ hybrid production in natural populations. Interestingly, the rare formation of F₁ hybrids has been evolutionarily significant, as reflected by numerous hybrid zones containing a wide array of later-generation hybrids (ARNOLD 1993; CRUZAN and ARNOLD 1993; JOHNSTON et al. 2001).

Using two reciprocal backcross 1 (BC₁) mapping populations, we performed genome scans to search for QTL that affected the timing of flowering in two separate years under greenhouse conditions, as well as in two experimental plots established within the naturally occurring ranges of both species. This allowed us to define the genetic architecture of the flowering phenology of both species. Specifically, we asked:

1. How many genomic regions were responsible for affecting flowering phenology?
   
2. Did the identified QTL interact epistatically?
   
3. Did the same QTL affect flowering phenology in two different greenhouse years and in the two different experimental plots?
   

Addressing these questions allowed us to define the genomic regions that affect a large proportion of the prezygotic isolation between these two naturally hybridizing plant species.

Construction of mapping populations:

One wild-collected individual each from I. fulva (If174, collected from Terrebonne Parish, Louisiana) and I. brevicaulis (Ib72, collected from St. Martinville Parish, Louisiana) was used to make reciprocal, interspecific BC₁ populations. The two field sites where the pure species were found were markedly different in terms of microhabitat, with the I. fulva and I. brevicaulis genotypes occurring along the margins of a bayou and an open hardwood forest, respectively (M. L. ARNOLD, unpublished data). To minimize the within-species genetic variation, clones of these parental individuals were used to produce the F₁ parents and were then utilized as the recurrent backcross parents. F₁ plants were produced using Ib72 as the maternal parent and If174 as the pollen parent. Two F₁ genotypes, designated as F₁(2) and F₁(3), were used as pollen parents to produce the BC₁ hybrids. F₁(2) was used to pollinate several clones of If174, and F₁(3) was used as the pollen donor for several clones of Ib72. Several hundred I. fulva backcross (BCIF) and I. brevicaulis backcross (BCIB) hybrid seeds were ultimately produced in 1999. Seeds were germinated and the resulting plants were maintained in the Department of Plant Biology greenhouses at the University of Georgia by transplanting into 6-in. azalea pots shortly after germination. During the fall of all subsequent years, a single rhizome was broken off of each BC₁ plant and transplanted into 8-in. azalea pots. All BC₁ plants were assigned to random positions in the greenhouse. Replicates that survived throughout these 5 years in the greenhouse were planted into the field as described below.

Construction of linkage maps:

Two independent linkage maps were constructed using the Iris retroelement (IRRE) transposon display marker system designed by KENTNER et al. (2003) and the linkage mapping program Mapmaker 3.0 (LANDER et al. 1987; LINCOLN et al. 1992). BOUCK et al. (2005) provide a detailed description of the initial map construction protocol. In short, BCIB genotypes were determined for 414 IRRE transposon display markers (N = 230 plants), and BCIF genotypes were determined for 309 IRRE markers (N = 120 plants). For the BCIF map, 63 additional markers were added to the marker data set originally utilized in the study by BOUCK et al. (2005). The two reciprocal linkage maps are independent because transposon display markers possess dominant inheritance. The presence or absence of dominant I. fulva markers in BCIB individuals was ascertained, and these markers were used to create the BCIB map. Likewise, the presence or absence of dominant I. brevicaulis markers in BCIF individuals was also recorded and utilized for creating the BCIF map. Linkage groups (LGs) identified in both maps were labeled in order of the largest to the smallest lengths (in centimorgans, LG1–LG22). Shared designations between LGs from the two maps do not imply homology. The BCIB map consists of 142 framework markers and 22 LGs, with an average marker spacing of 12 cM. The BCIF map consists of 108 framework markers and 22LGs, with an average marker spacing of 13 cM. The BCIF map presented here has been modified somewhat relative to the map presented in BOUCK et al. (2005) due to the addition of new, informative markers identified in this study.

Assaying flowering phenology in the greenhouse:

The two greenhouse experiments were performed in the Department of Plant Biology facilities in the spring of 2002 and of 2003 and were part of a larger study examining a number of floral-trait QTL (BOUCK 2004). Data were collected from clones of the same individuals used to create the BCIF and BCIB maps. The total number of BCIF individuals that were used to map flowering phenology QTL was 67 and 57 during 2002 and 2003, respectively. The total number of BCIB genotypes that were scored for flowering phenology was 176 and 145 during 2002 and 2003, respectively. For each individual, the date at which the first flower appeared was recorded and assigned a numerical score (1 for the first day that a flower was observed, 2 for the second day, etc.). Phenological plots for the greenhouse plants are not presented here, as F₁ and pure-species plants were not included in the study, and only the date of first flowering was recorded for the BCIF and BCIB individuals.

Assaying flowering phenology in the field:

The field experiment was performed in a cypress–mixed hardwood forest in which natural populations of I. brevicaulis and I. fulva are typically found (VIOSCA 1935; CRUZAN and ARNOLD 1993; JOHNSTON et al. 2001). Two experimental plots were established 1 km away from one another near the Choupique Bayou located in the U. S. Army Corps of Engineers Atchafalaya Basin Floodway in Louisiana. The elevation at both of these sites is 5.0 m above mean sea level, and the elevational change within each plot differed by <0.5 m. It should be noted that these plots are not the same as those described previously by MARTIN et al. (2006), as extensive and extended flooding largely destroyed the latter. Throughout this study, we will differentiate the two plots as either the "dry" or the "wet" site. We performed no soil moisture analysis, but we observed that during periods of heavy rainfall, much of the wet plot became inundated with 0.5 m of water. Furthermore, following the cessation of rainfall, several days passed before the water receded completely. In contrast, within the dry site, no standing water was ever observed for more than a single day during the field season.

In October 2005, we collected three to four rhizomes from each BCIF and BCIB plant (172 and 243 individuals, respectively), along with 43 I. fulva rhizomes, 62 I. brevicaulis rhizomes, and 47 F₁ rhizomes. We transplanted one to two of these rhizomes into each of the two plots. Within each plot, we randomly assigned positions to the various genotypes, spacing them 0.5 m apart. We noted the date at which each flower opened, and the date at which each flower wilted to the point where it was unattractive to pollinators (2 days post opening). For the majority of flowers, we were present to record the date of opening and wilting. However, for a few genotypes, we were not present when a flower opened, and in those instances, we estimated the date based upon floral morphology. Flowers are normally open for 2 days, with dramatic morphological differences apparent between 1-day-old and 2-day-old flowers. For example, stigmas do not become exposed to pollinators until day 2. Flowering phenology for the two field plots are given in Figure 1.

View larger version (35K): In this window In a new window Download PPT slide   FIGURE 1.— Flowering phenology for experimental I. fulva (red), BCIF (pink), F1 (yellow), BCIB (purple), and I. brevicaulis (blue) assessed on a daily basis at two separate sites. The census number (y-axis) is expressed here as a percentage of the maximum number of flowers open throughout the census period. (A) Dry site. A total of 146 I. fulva, 2005 BCIF, 470 F1, 2115 BCIB, and 51 I. brevicaulis flowers were censused throughout the entire flowering period. (B) Wet site. A total of 233 I. fulva, 2351 BCIF, 489 F1, 2469 BCIB, and 151 I. brevicaulis flowers were censused throughout the entire flowering period.

 
We utilized a two-way analysis of variance (ANOVA) to test whether there were significant differences among the mean flower initiation dates for the five different cross types. A "cross type" main effect (I. fulva, BCIF, F₁, BCIB, and I. brevicaulis), a "habitat" main effect (wet or dry), and a "cross type x habitat" interaction effect were included in the model (Table 1). We also used posthoc Tukey honestly significant differences (HSD) tests (corrected for multiple comparisons) to test whether there were significant differences among the cross types.

View this table: In this window In a new window   TABLE 1 Results of a two-way analysis of variance on "date of first flower" conducted in two separate field sites

 
Under a purely additive model of genetic inheritance, the expected mean flowering date of F₁ hybrids would be the average of the mean flowering dates of I. fulva and I. brevicaulis. Planned linear contrasts were used to test whether the mean of F₁ hybrids differed significantly from this additive model's expectation by comparing the F₁ hybrid class to that of the two parental species. Further, under a purely additive model of genetic inheritance, the expected mean flowering dates of the backcross hybrids would be 0.75 x the mean flowering dates of the recurrent parent + 0.25 x the mean flowering date of the other parent. Thus, we used two planned linear contrasts to test whether the mean of both BCIF and BCIB hybrids differed significantly from this additive model's expectation.

QTL analysis of flowering phenology:

We mapped flower initiation dates (day 1 for the date at which the first flower was observed, day 2 for the second date, etc.) in both BCIB and BCIF mapping populations planted in the greenhouse and in nature. In the greenhouse analyses, only a single copy of each genotype was present each year. For the field data, up to two copies of each BC₁ genotype were planted into each plot, and in instances where both copies flowered at a site, we used the mean of the two initiation dates of the two replicates to perform QTL mapping. No transformations were performed on the data. Separate genome scans for QTL affecting flowering phenology were performed for each of the two BC₁ maps, first using composite interval mapping (CIM) (ZENG 1993, 1994) followed by refinement with multiple interval mapping (MIM) methods (KAO et al. 1999; ZENG et al. 1999). Analyses were performed separately for the 2002 and 2003 greenhouse data sets and separately for the two plots used in the single 2006 field season. CIM tests were performed at 1-cM steps, with a 10-cM window size to exclude closely linked cofactors. The CIM program through forward and backward regression chose three cofactors for each QTL analysis. Genomewide threshold values were determined at the 5% level by running 1000 permutations for each map (CHURCHILL and DOERGE 1994; DOERGE and CHURCHILL 1996). MIM tends to have more power and precision for detecting QTL (KAO et al. 1999), so we used QTL detected in CIM analyses as an initial model for MIM mapping. If no QTL were found with CIM, we used forward and backward regression selection on all markers with partial r² probability set at 0.01 as the initial model for MIM. MIM was then used to (i) refine QTL positions, (ii) search for additional QTL, (iii) search for epistatic effects among detected QTL, and (iv) estimate individual QTL effects and the proportion of phenotypic variance explained by each of the QTL and significant interactions. Tests in MIM were performed at 1-cM steps, with Bayesian Information Criterion model selection and the penalty function set to c(n) = ln(n). Two-LOD support limits for each QTL were calculated around the most likely QTL position, as illustrated on the maps in Figures 3 and 4. The positions, effect size and directions, and the proportion of phenotypic variance explained by each QTL are presented in Tables 2 and 3. To test whether selection might have been important in causing flowering divergence, we utilized sign tests in each mapping population to determine whether the directionality (+ or –) of nonoverlapping QTL effects significantly differed from the null expectation of equal proportions (i.e., consistent with a model of neutrality).

View larger version (18K): In this window In a new window Download PPT slide   FIGURE 3.— Linkage map of dominant I. brevicaulis IRRE retrotransposon display markers segregating in the F1 hybrid used to produce BCIF hybrids. Significant QTL for flowering phenology are denoted (with 2-LOD confidence intervals) to the right of the linkage groups. Red bars represent regions where introgressed I. brevicaulis alleles caused flowering to initiate earlier, while blue bars represent regions where introgressed I. brevicaulis alleles caused flowering to initiate later. QTL analyses were performed during the two greenhouse years 2002 and 2003 as well as in two field plots (dry and wet) in 2006.

 

View larger version (19K): In this window In a new window Download PPT slide   FIGURE 4.— Linkage map of dominant I. fulva IRRE retrotransposon display markers segregating in the F1 hybrid used to produce BCIB hybrids. Significant QTL for flowering phenology are denoted (with 2-LOD confidence intervals) to the right of the linkage groups. Red bars represent regions where introgressed I. fulva alleles caused flowering to initiate earlier, while blue bars represent regions where introgressed I. fulva alleles caused flowering to initiate later. QTL analyses were performed during the two greenhouse years 2002 and 2003 as well as in two field plots (dry and wet) in 2006.

 

View this table: In this window In a new window   TABLE 2 QTL for flowering initiation in two field sites (dry and wet) and in two greenhouse seasons (2002 and 2003) detected in a BC1 (BCIF) population

 

View this table: In this window In a new window   TABLE 3 QTL for flowering initiation in two field sites (dry and wet) and in two greenhouse seasons (2002 and 2003) detected in a BC1 (BCIB) population

 

The mean flowering-initiation dates (±1 SE) at the two different sites were as follows: dry site: I. fulva—March 30 (±1.4 days); BCIF—April 10 (±0.4 days); F₁—April 14 (±0.7 days); BCIB—April 23 (±0.4 days); I. brevicaulis—May 2 (±0.4 days); wet site: I. fulva—March 30 (±1.4 days); BCIF—April 9 (±0.4 days); F₁—April 11 (±0.7 days); BCIB—April 22 (±0.4 days); I. brevicaulis—May 1 (0.7 days). Results of two-way ANOVA indicated that there was a main effect for "cross type," no main effect for "site," and no "site x cross type" interaction (Table 1). Examining the significant main effect (cross type) further, posthoc Tukey HSD tests accounting for multiple comparisons revealed that all cross types were statistically different from one another with respect to flower initiation (corrected P-values <0.05 for all 10 comparisons).

We tested whether the flowering-date phenotypes of F₁ and reciprocal backcross hybrids deviated significantly from an additive model of genetic inheritance. Under a purely additive model, the expected mean flowering date of F₁ hybrids would be the mean flowering date of both pure-species parents, and F₁ hybrids did deviate from this additive model, but only with marginal significance (linear contrasts, F = 2.91, 1 d.f., P = 0.088, Figure 2). For BCIF hybrids, given strict additivity, the expected mean flowering date of BCIF hybrids would be 0.75 x the mean flowering date of I. fulva + 0.25 x the mean flowering date of I. brevicaulis. The BCIF hybrids did in fact deviate significantly from this null additive model, with BCIF hybrids flowering only 3 days earlier than F₁ hybrids, on average, but 11.5 days later, on average, than pure I. fulva species (linear contrast, F = 4.44, 1 d.f., P = 0.035, Figure 2). The expected mean flowering date of BCIB hybrids, given a purely additive genetic model of inheritance, would be 0.75 x the mean flowering date of I. brevicaulis + 0.25 x the mean flowering date of I. fulva, and the BCIB hybrids did not significantly deviate from this model (linear contrast, F = 0.063, 1 d.f., P = 0.802, Figure 1).

View larger version (5K): In this window In a new window Download PPT slide   FIGURE 2.— Observed vs. expected mean flower initiation dates (±2 SE) for I. fulva, I. brevicaulis, F1, BCIF, and BCIB hybrids. The diagonal line denotes the expected mean flowering date under an additive model of gene action.

 

Quantitative trait locus analysis:

Using CIM followed by refinement with MIM, we identified 17 QTL in the BCIF mapping population that affected flowering time in one or more field habitats or greenhouse years (Table 2, Figure 3). In this backcross population, LG1 possessed four QTL: three of the QTL had negative effects, meaning that the introgressed I. brevicaulis alleles caused flowering time to occur earlier. The confidence intervals of these three QTL were overlapping, suggesting that the genes causing earlier flowering times could be the same across the three separate habitats (greenhouse season 2002, wet and dry field sites; Table 2, Figure 3). A fourth QTL was detected at the topmost portion of the linkage group, and this QTL had a positive effect, meaning that the introgressed I. brevicaulis alleles caused flowering time to occur later (in the dry field site; Table 2, Figure 3). LG12 was the only other linkage group in the BCIF mapping population that revealed a QTL with negative effect (i.e., introgressed I. brevicaulis alleles caused flowering to initiate earlier in the dry field site). In contrast, a QTL that caused a later flowering time in the wet field site was also detected on LG12. These two QTL, given their opposite effects, are unlikely to be due to the same genes. The remaining 11 QTL detected in the BCIF mapping population caused flowering to initiate later. These loci included single QTL (i.e., affecting flowering time in only one of the four greenhouse or field habitats) found on LG2, LG6, LG8, LG9, and LG13 and two overlapping QTL (affecting flowering time in two of the four greenhouse or field habitats) found on LG5, LG7, and LG11. Four epistatic interactions were detected between QTL using MIM methodologies: between QTL 2 and 3 in the 2002 greenhouse study; between QTL 2 and 3 and 3 and 4 in the dry site; and between QTL 1 and 6 in the wet site (Table 2). In all instances, epistatic interactions between two alleles resulted in flowering time occurring later than expected, given a purely additive model.

In the BCIB mapping population, we utilized CIM followed by MIM and found a total of 15 QTL that affected flowering time (Table 3, Figure 4). Only 4 of the 15 detected QTL had positive effects (i.e., caused later flowering when I. fulva alleles were present). Two of these QTL were detected in the 2002 greenhouse study and were located on LG6 and LG15. The other two positive QTL were detected in the wet field site and were located on LG9 and LG13. All other QTL in the BCIB mapping population were negative; i.e., they caused earlier flowering when introgressed I. fulva alleles were present. Two overlapping, negative QTL were located on LG2 and were detected in the 2002 greenhouse sample and the dry field site. Another two overlapping QTL were located on LG7 and were detected in the 2002 and 2003 greenhouse studies. The remaining negative QTL were found singly (i.e., affected flowering time in only one of the four greenhouse or field habitats) on LG4, LG5, LG10, LG16, LG17, LG20, and LG21. Three epistatic interactions were detected in the BCIB mapping population: between QTL 2 and 3 and between QTL 4 and 5 in the 2002 greenhouse study and between QTL 1 and 4 in the wet field site (Table 3). The epistatic interactions detected in the greenhouse study caused an earlier flowering time when both I. fulva alleles at the different QTL were present. In contrast, the epistatic interaction detected in the wet site caused flowering time to occur later than expected in comparison to a purely additive model.

Patterns of flowering phenology:

As with many other closely related taxa that have the potential to hybridize, a number of reproductive isolating mechanisms interact to prevent F₁ hybrid formation in I. brevicaulis/I. fulva hybrid zones (see ARNOLD 1997 and 2006 for reviews). Divergent flowering phenologies between I. fulva and I. brevicaulis, with I. fulva initiating flowering earlier in the season, have been shown to be an important prezygotic barrier to hybridization (CRUZAN and ARNOLD 1994). In fact, this divergence of flowering time has the potential to be the most important barrier to gene flow between sympatric populations, since the two species must obviously overlap in their reproductive times for later-acting barriers to come into play. In this study, we showed that in two field sites planted in natural conditions, I. fulva and I. brevicaulis genotypes were nonoverlapping in their flowering times, with I. brevicaulis initiating flowering 9–10 days after the last I. fulva flowers were observed (Figure 1). In the absence of other results, the findings from this study would suggest that flowering phenology alone is sufficient to prevent gene flow between these two species. However, previous studies have shown some amount of overlap in natural, sympatric populations of I. fulva and I. brevicaulis (CRUZAN and ARNOLD 1994). Given this, along with the detection of numerous natural hybrid zones between the two species (e.g., ARNOLD 1993; CRUZAN and ARNOLD 1993; JOHNSTON et al. 2001), we can assume that flowering phenology is not an impermeable reproductive barrier. Assuming that there is intraspecific variability for the timing of flowering, an experiment involving larger sample sizes of the two species might reveal an overlap in flowering times [as, for example, was found in the much larger, natural population studied by CRUZAN and ARNOLD (1994)]. Regardless, flowering time is apparently an extremely strong (in an absolute sense) and important prezygotic barrier between I. fulva and I. brevicaulis. Furthermore, the observation of the nonoverlapping of flowering phenologies of the I. fulva and I. brevicaulis suggests that the formation of the natural hybrid populations is likely limited by the episodic formation of F₁ hybrids between these two Iris species.

An examination of the flowering times of the three hybrid classes planted into the wild revealed the F₁ hybrids to be intermediate between the two species. Furthermore, as predicted for quantitative traits, BCIF were intermediate between I. fulva and F₁ hybrids, while BCIB hybrids were intermediate between I. brevicaulis and F₁ hybrids. These data are consistent with the detection of numerous natural hybrid zones between these species. Thus, although the divergent flowering times of I. fulva and I. brevicaulis greatly limit F₁ hybrid formation, these rare events produce effective intermediaries for gene flow between the two species. Furthermore, we would predict that the later-generation hybrids could also be conduits for gene flow, but primarily with the species with which they were backcrossed.

The genetic architecture of flowering phenology:

By examining the phenotypic expression of flowering times of the three hybrid classes used in this study, and comparing those flowering times to the parental genotypes, we can gain an initial estimate of the underlying genetic architecture responsible for the divergent phenologies of I. fulva and I. brevicaulis. If the divergent flowering times were due solely to a single gene that acted in a completely dominant/recessive fashion, then the F₁ flowering phenology would mirror that of one of the parents. Here, F₁ hybrids are intermediate between the two parents, suggesting that this simple model cannot explain the divergent flowering phenologies. In addition, by examining the flowering times of both reciprocal backcross hybrids (BCIF and BCIB), we see that the mean flowering times of those hybrids are intermediate between the F₁ and the recurrent parental species. Furthermore, the flowering phenotypes of many backcross individuals are indistinguishable from the recurrent parents. This suggests that the divergent flowering phenologies of I. fulva and I. brevicaulis are due to multiple genes. Furthermore, not all of the genes act in a purely additive fashion. Thus, the mean flowering phenology of BCIF hybrids is significantly later than would be expected, given a purely additive model of gene action (Figure 2).

We further examined the polygenic nature of this complex quantitative trait by performing QTL mapping of flower initiation in each of the BCIF and BCIB mapping populations in both greenhouse and field conditions. We identified 16 nonoverlapping QTL that affected flowering phenology in both BCIF and BCIB hybrids in at least one of the four greenhouse/field habitats. Our results indicate that these phenological QTL, depending on the habitats being examined, have small-to-large effect sizes. We would add a cautionary note that QTL mapping tends to overestimate the effect magnitude of QTL, especially with smaller sample sizes such as those found in our BCIF mapping population. Most importantly, the power to detect QTL of smaller effect is also dependent upon the sample size. The number of genotyped individuals that flowered in this study was relatively small. Thus, the number of QTL detected should be considered a minimum. Furthermore, the limited sample sizes may have contributed to the observation of no overlapping QTL detected across all four of the habitats (greenhouse and field) examined. Athough it is not unusual to detect QTL that are unique across habitat types, it is striking that none of the QTL detected in this study affected flowering phenology in all habitats examined. This is especially striking since "habitat" apparently did not significantly affect flowering time of plants in the field (Table 1). We therefore attribute this observation largely to our relatively low sample sizes, as it is difficult to argue that the different habitats are themselves responsible for the lack of detection of overlapping QTL. Readers should thus interpret these results with the following caveat: lack of detection of a QTL in one particular environment does not rule out the possibility that a QTL could be affecting the phenotype in that environment. However, if a QTL is detected in any particular environment, the QTL reflects a significant phenotype/genotype association. These data clearly show that multiple QTL, distributed throughout the genome, are responsible for the divergent flowering times between I. fulva and I. brevicaulis and should be viewed as "preliminary" only in the sense that an increased sample size will likely reveal even more overlapping QTL, as the power to detect them increases.

A consideration of the directionality of the detected QTL suggests that divergence in flowering time likely has an adaptive explanation. In the BCIF population, only 4 of the 17 QTL caused flowering time to occur earlier when I. brevicaulis (i.e., the later-flowering species) alleles were present. A sign test (assuming all of the QTL are independent) indicates that this significantly differs from the expected proportion if no selection was acting on flowering time (i.e., a neutral model: + QTL 13, – QTL 4, P = 0.049). Assuming that overlapping QTL (i.e., those that affected flowering time in multiple habitats, in the same direction, and whose confidence intervals overlapped) are the same also results in a significant sign test result (+ QTL 10, – QTL 2, P = 0.039). The opposite trend was detected in the BCIB backcross hybrids, with only 4 of the 15 detected QTL increasing flowering time when I. fulva (i.e., the earlier-flowering species) alleles were present. However, a sign test (assuming all QTL are independent) suggests only marginal significance (P = 0.12). Furthermore, assuming overlapping QTL are the same, no significant difference is detected (P = 0.27). Yet, combining both the BCIB and BCIF data sets reveals that the majority of QTL are in the expected direction (24/32), with results from a sign test on this proportion (P = 0.007) being consistent with the contribution of directional selection in promoting divergence in reproductive phenology. Furthermore, epistatic interactions among QTL predominantly affected phenology in the predicted direction. Thus, all four interactions between I. brevicaulis alleles in the BCIF plants caused later flowering times. Similarly, two of three epistatic interactions among I. fulva alleles in the BCIB population were associated with earlier flowering. Thus, not only does flowering phenology cause large-scale reproductive isolation between these two species, but also the genetic architecture fits a model of adaptive divergence.

We thank S. Wessler, R. Mauricio, and E. Kentner for contributions to the linkage mapping projects that form the basis for these analyses. We thank Y. Sapir, D. Wills, S. Taylor, and N. Sherman for helpful comments on previous versions of the manuscript. R. H. Martin and Y. Z. Martin provided invaluable assistance in the field. B. Weckerley provided invaluable statistical advice. We thank the Army Corps of Engineers Atchafalaya Basin Floodway System for field sites and field assistance. This study was supported by two National Science Foundation grants to M.L.A. (DEB-0074159 and DEB-0345123) and by a gift from M. Caillet.

ARNOLD, M. L., 1993 Iris nelsonii: origin and genetic composition of a homoploid hybrid species. Am. J. Bot. 80: 577–583.[CrossRef]

ARNOLD, M. L., 1997 Natural Hybridization and Evolution. Oxford University Press, New York.

ARNOLD, M. L., 2006 Evolution Through Genetic Exchange. Oxford University Press, New York.

BOUCK, A. C., 2004 The Genetic Architecture of Reproductive Isolation in Louisiana Irises. Ph.D. Dissertation, University of Georgia, Athens, GA.

BOUCK, A. C., R. PEELER, M. L. ARNOLD and S. R. WESSLER, 2005 Genetic mapping of species boundaries in Louisiana irises using IRRE retrotransposon display markers. Genetics 171: 1289–1303.[Abstract/Free Full Text]

BRADSHAW, H. D., JR., S. M. WILBERT, K. G. OTTO and D. W. SCHEMSKE, 1995 Genetic mapping of floral traits associated with reproductive isolation in monkeyflowers (Mimulus). Nature 376: 762–765.[CrossRef]

BRADSHAW, H. D., JR., K. G. OTTO, B. E. FREWEN, J. K. MCKAY and D. W. SCHEMSKE, 1998 Quantitative trait loci affecting differences in floral morphology between two species of monkeyflower (Mimulus). Genetics 149: 367–382.[Abstract/Free Full Text]

CHURCHILL, G. A., and R. W. DOERGE, 1994 Empirical threshold values for quantitative trait mapping. Genetics 138: 963–971.[Abstract]

COYNE, J. A., and H. A. ORR, 2004 Speciation. Sinauer Associates, Sunderland, MA.

CRUZAN, M. B., and M. L. ARNOLD, 1993 Ecological and genetic associations in an Iris hybrid zone. Evolution 47: 1432–1445.[CrossRef]

CRUZAN, M. B., and M. L. ARNOLD, 1994 Assortative mating and natural selection in an Iris hybrid zone. Evolution 48: 1946–1958.[CrossRef]

DOERGE, R. W., and G. A. CHURCHILL, 1996 Permutation tests for multiple loci affecting a quantitative character. Genetics 142: 285–294.[Abstract]

GLEASON, J. M., and M. G. RITCHIE, 2004 Do quantitative trait loci (QTL) for a courtship song difference between Drosophila simulans and D. sechellia coincide with candidate genes and intraspecific QTL? Genetics 166: 1303–1311.[Abstract/Free Full Text]

GLEASON, J. M., S. V. NUZHDIN and M. G. RITCHIE, 2002 Quantitative trait loci affecting a courtship signal in Drosophila melanogaster. Heredity 89: 1–6.[CrossRef][Medline]

FISHMAN, L., A. J. KELLY and J. H. WILLIS, 2002 Minor quantitative trait loci underlie floral traits associated with mating system divergence in Mimulus. Evolution 56: 2138–2155.[CrossRef][Medline]

HODGES, S. A., J. B. WHITALL, M. FULTON and J. Y. YANG, 2002 Genetics of floral traits influencing reproductive isolation between Aquilegia formosa and A. pubescens. Am. Nat. 159: S51–S60.[CrossRef][Medline]

HURT, C. R., M. FARZIN and P. W. HEDRICK, 2005 Premating, not postmating, barriers drive genetic dynamics in experimental hybrid populations of the endangered Sonoran topminnow. Genetics 171: 655–662.[Abstract/Free Full Text]

HUSBAND, B. C., and H. A. SABARA, 2004 Reproductive isolation between autotetraploids and their diploid progenitors in fireweed, Chamerion angustifolium (Onagraceae). New Phytol. 161: 703–713.[CrossRef]

JOHNSTON, J. A., R. A. WESSELINGH, A. C. BOUCK, L. A. DONOVAN and M. L. ARNOLD, 2001 Intimately linked or hardly speaking? The relationship between genotype and environmental gradients in a Louisiana Iris hybrid population. Mol. Ecol. 10: 673–681.[CrossRef][Medline]

KAO, C. H., Z-B. ZENG and R. D. TEASDALE, 1999 Multiple interval mapping for quantitative trait loci. Genetics 152: 1203–1216.[Abstract/Free Full Text]

KENTNER, E. K., M. L. ARNOLD and S. R. WESSLER, 2003 Characterization of high-copy-number retrotransposons from the large genomes of the Louisiana Iris species and their use as molecular markers. Genetics 164: 685–697.[Abstract/Free Full Text]

LANDER, E. S., P. GREEN, J. ABRAHAMSON, A. BARLOW, M. J. DALY et al., 1987 MAPMAKER: an interactive computer package for constructing primary genetic linkage maps of experimental and natural populations. Genomics 1: 174–181.[CrossRef][Medline]

LINCOLN, S., M. DALY and E. S. LANDER, 1992 Constructing genetic maps with MAPMAKER/EXP 3.0. Whitehead Institute Technical Report, Ed. 2. Whitehead Institute, Cambridge, MA.

MARTIN, N. H., and J. H. WILLIS, 2007 Ecological divergence associated with mating system causes nearly complete reproductive isolation between sympatric Mimulus species. Evolution 61: 68–82.[CrossRef][Medline]

MARTIN, N. H., A. C. BOUCK and M. L. ARNOLD, 2005 Loci affecting long-term hybrid survivorship in Louisiana Irises: implications for reproductive isolation and introgression. Evolution 59: 2116–2124.[CrossRef][Medline]

MARTIN, N. H., A. C. BOUCK and M. L. ARNOLD, 2006 Detecting adaptive trait introgression between Iris fulva and I. brevicaulis in highly selective field conditions. Genetics 172: 2481–2489.[Abstract/Free Full Text]

MOEHRING, A. J., J. LI, M. D. SCHUG, S. G. SMITH, M. DEANGELIS et al., 2004 Quantitative trait loci for sexual isolation between Drosophila simulans and D. mauritiana. Genetics 167: 1265–1274.[Abstract/Free Full Text]

MOEHRING, A. J., A. LLOPART, S. ELWYN, J. A. COYNE and T. F. C. MACKAY, 2006a The genetic basis of postzygotic reproductive isolation between Drosophila santomea and D. yakuba due to hybrid male sterility. Genetics 173: 225–233.[Abstract/Free Full Text]

MOEHRING, A. J., A. LLOPART, S. ELWYN, J. A. COYNE and T. F. C. MACKAY, 2006b The genetic basis of prezygotic reproductive isolation between Drosophila santomea and D. yakuba due to mating preference. Genetics 173: 215–223.[Abstract/Free Full Text]

MOYLE, L. C., and E. B. GRAHAM, 2005 Genetics of hybrid incompatibility between Lycopersicon esculentum and L. hirsutum. Genetics 169: 355–373.[Abstract/Free Full Text]

RAMSEY, J., H. D. BRADSHAW and D. W. SCHEMSKE, 2003 Components of reproductive isolation between the monkeyflowers Mimulus lewisii and M. cardinalis (Phrymaceae). Evolution 57: 1520–1534.[CrossRef][Medline]

SLOTMAN, M., A. D. TORRE and J. R. POWELL, 2004 The genetics of inviability and male sterility in hybrids between Anopheles gambiae and An. arabiensis. Genetics 167: 275–287.[Abstract/Free Full Text]

VIOSCA, P. J., 1935 The Irises of southeastern Louisiana: a taxonomic and ecological interpretation. Bull. Am. Iris Soc. 57: 3–56.

WHITEMAN, H. H., and R. D. SEMLITSCH, 2005 Asymmetric reproductive isolation among polymorphic salamanders. Biol. J. Linn. Soc. 86: 265–281.[CrossRef]

ZENG, Z-B., 1993 Theoretical basis of separation of multiple linked gene effects on mapping quantitative trait loci. Proc. Natl. Acad. Sci. USA 90: 10972–10976.[Abstract/Free Full Text]

ZENG, Z-B., 1994 Precision mapping of quantitative trait loci. Genetics 136: 1457–1468.[Abstract]

ZENG, Z-B., C-H. KAO and C. J. BASTEN, 1999 Estimating the genetic architecture of quantitative traits. Genet. Res. 4: 279–289.

Communicating editor: S. W. SCHAEFFER

International Access Link

Online ISS 1091-6490
Copyright 2008 by the Genetics Society of America
phone: 412-268-1812   fax: 412-268-1813   email: genetics-gsa{at}andrew.cmu.edu