Abstract

The formation of hybrids among closely related species has been observed in numerous plant taxa. Selection by pollinators on floral traits can act as an early reproductive isolating barrier and may be especially important when there is overlap in distribution and flowering time. In this study, we use Quantitative Trait Locus (QTL) mapping based on 293 codominant SNP markers in an F2 population (n = 328) to assess the size, magnitude, and location of the genetic regions controlling floral traits known to be important for pollinator attraction in 2 species of Lousiana Irises, Iris fulva and Iris hexagona. We also evaluate correlations among F2 traits and identify transgression in the hybrid population. Overall, we observe that differences in most floral traits between I. fulva and I. hexagona are controlled by multiple QTLs and are distributed across several linkage groups. We also find evidence of transgression at several QTL, suggesting that hybridization can contribute to generating phenotypic variation, which may be adaptive in rapidly changing environments.

The formation of hybrids among closely related species has been observed in numerous taxa and is thought to be especially common in plants (Mallet 2005; Whitney et al. 2010). A number of studies investigating the evolutionary consequences of hybridization, including the potential for novel trait formation (Howarth and Baum 2005), the transfer of adaptive alleles (e.g., adaptive introgression; Anderson 1949; Parsons et al. 2011), and the formation of novel hybrid species (Donovan et al. 2010) have demonstrated the importance of hybridization as an evolutionary force. Both the synergistic effects of the pre- and postzygotic isolating barriers and the underlying genomic architecture of the species involved are important for determining the outcomes of hybridization (Fishman et al. 2002; Hodges et al. 2002). Identifying the relationship between the unique combinations of traits that result from hybridization and their underlying genetic architecture can provide a more comprehensive picture of how phenotypic traits evolve. In this study, we report on the results of a QTL mapping study designed to examine the genetic basis of floral traits important for prezygotic isolation in 2 naturally interfertile Louisiana Irises.

Reproductive isolation is often achieved through divergence in floral traits, which has been implicated as an avenue to speciation (Kay and Sargent 2009). In nascent plant species, where postzygotic isolating barriers are often incomplete, prezygotic isolating barriers are thought to play an important role in the maintenance of species boundaries (Widmer et al. 2009). For example, selection by pollinators on suites of traits, or pollination syndromes, has been linked to adaptive radiations in numerous genera (Hodges 1997; Stuurman et al. 2004; Kay and Sargent 2009; Puzey et al. 2012). The genetic dissection of floral traits using hybrid populations can identify which traits influence speciation. This first requires evaluating which traits are attractive to pollinators through observational studies, followed by measuring a range of phenotypes and associating them with a sufficiently large number of genetic markers. A QTL mapping approach can determine the number, location, and magnitude of genomic regions associated with phenotypic traits principal to pollinators.

In this study, we used a QTL mapping approach to examine quantitative trait variation in 2 closely related Louisiana Irises, Iris hexagona (Walter) and Iris fulva (Ker-Gawler). These 2 species are broadly sympatric, and interspecific hybrids have been observed across coastal Louisiana. It is easy to distinguish between the 2 species due to their dramatically different floral and vegetative phenotypes. Iris hexagona is tall (~2 m) and characterized by large purple flowers with distinctive yellow nectar guides, which are attractive to their primary pollinator, bumble bees (Viosca 1935; Emms and Arnold 2000). Iris fulva is relatively short (~1 m) with deep crimson colored flowers that lack nectar guides but produce larger volumes of nectar, as is typical of plants pollinated by hummingbirds (Viosca 1935; Wesselingh and Arnold 2000). These species are found in different habitats; I. hexagona in open, wet habitats and I. fulva in intermittently flooded, forested wetlands. However, both natural- and human-mediated disturbances are thought to have caused changes in the landscape that have facilitated hybridization between the 2 species (Riley 1938; Arnold et al. 1990). Iris hexagona and I. fulva have overlapping flowering phenologies with peak flowering occurring from March to April (Arnold 1993), which also increases the likelihood of gene flow in areas of geographic overlap.

Despite this large overlap in flowering phenologies, experimental arrays have shown a deficiency in F1 hybrid production when these species are in close proximity (Arnold 1993), suggesting that later-acting isolating barriers prevent hybridization between these taxa. Pollinator isolation also likely acts to reduce hybridization between these taxa because the species are visited by different pollinator groups. Iris fulva is primarily visited by hummingbirds, whereas I. hexagona is primarily visited by bees (Emms and Arnold 2000). Both long-range cues (e.g., color) and short-range cues (e.g., nectar) and also structural floral traits (e.g., anther extension) influenced pollinator visitation and success (Emms and Arnold 2000), suggesting that floral traits are an important component of prezygotic isolation in this system. Here, we use a QTL mapping approach to investigate the genetic basis of floral traits that are divergent in I. fulva and I. hexagona to examine the size, magnitude, and direction of QTL. Though we have previously defined the genetic architecture of reproductive and ecological adaptations for I. fulva (Bouck et al. 2007), this represents the first such analysis for I. hexagona.

Materials and Methods

F2 Mapping Population

Individuals of I. hexagona and I. fulva were collected from wild allopatric populations in 1989 and have been maintained through clonal propagation (Supplementary Material online). In 1990, a cross between a single I. hexagona (IhA16; seed parent) and a single I. fulva (If194; pollen parent) was made. A single F1 plant was selfed to produce a population of 444 F2 hybrid individuals with a range of floral phenotypes, which have been maintained under standard greenhouse conditions at the University of Georgia (Table 1).

Table 1

Traits measured in parents, F1s and F2s, are presented

Trait I. fulva (n = 12) I. hexagona (n = 7) F1 (n = 10) F2 (n = 328) DSM ESD DSM/ESD 
Days to flower 117.7–158.3 140.5–149.3 147.6–157.4 96–81 3.25 80.73 0.04 
148.87±3.06 145.62±0.98 150.01±0.97 129.77±0.98 
Flower height 49.87–81.33 102.3–156.1 92.92–127.8 55.75±147.2 −70.13 84.85 −0.83 
67.67±2.38 137.8±6.78 114.86±3.22 95.25±0.81 
Leaf length 52.9–81.2 96.0–132.8 88.3–117.5 45.6–132.3a −51.32 64.64 −0.79 
64.33±2.36 115.65±5.67 95.16±2.92 92.76±0.83 
Floral guide area — 31.8–62.8 3.6–7.5 0–78.2b −42.31 80.10 −0.53 
0±0 42.31±4.75 5.48±0.36 12.27±0.89 
Blade length 40.5–44.4 71.3–74.7 49.2–53.1 32.8–7.0a −31.42 9.73 −3.23 
42.02±0.49 73.44±0.61 51.59±0.37 47.41±0.30 
Blade width 25.5–30.3 49.3–52.1 34.9–38.2 20.4–47.3a −21.84 5.74 −3.81 
28.05±0.46 49.89±0.48 36.72±0.37 33.18±0.23 
Sepal stalk length 8.2–11.5 39.2–41.9 18.4–20.9 10.81–32.4 −31.01 4.06 −7.64 
9.36±0.25 40.37±0.34 19.81±0.22 20.45±0.20 
Anther extension 1.7–8.8 −14.7 to −10.5 −7.07 to −3.1 −25.05 to 7.61 15.15 5.90 2.57 
2.72±0.56 −12.43±0.56 −4.67±0.34 −7.42±0.25 
L38.77–121.3 87.59–109.76 29.53–66.7 22.43–199.54a −23.75 23.12 −1.03 
74.13±7.15 97.88±2.25 48.12±18.59 85.98±1.63 
a142.46–172.99 150.5–156.69 134.91–152.27 131.23–176.42b 4.92 10.14 0.49 
157.64±2.74 152.72±0.75 143.59±8.68 150.19±0.43 
b134.29–165.87 81.32–91.74 115.42–123.99 16.64–199.39a,b 60.04 6.85 8.76 
148.63±2.72 88.59±1.15 119.71±4.29 116.87±0.88 
Trait I. fulva (n = 12) I. hexagona (n = 7) F1 (n = 10) F2 (n = 328) DSM ESD DSM/ESD 
Days to flower 117.7–158.3 140.5–149.3 147.6–157.4 96–81 3.25 80.73 0.04 
148.87±3.06 145.62±0.98 150.01±0.97 129.77±0.98 
Flower height 49.87–81.33 102.3–156.1 92.92–127.8 55.75±147.2 −70.13 84.85 −0.83 
67.67±2.38 137.8±6.78 114.86±3.22 95.25±0.81 
Leaf length 52.9–81.2 96.0–132.8 88.3–117.5 45.6–132.3a −51.32 64.64 −0.79 
64.33±2.36 115.65±5.67 95.16±2.92 92.76±0.83 
Floral guide area — 31.8–62.8 3.6–7.5 0–78.2b −42.31 80.10 −0.53 
0±0 42.31±4.75 5.48±0.36 12.27±0.89 
Blade length 40.5–44.4 71.3–74.7 49.2–53.1 32.8–7.0a −31.42 9.73 −3.23 
42.02±0.49 73.44±0.61 51.59±0.37 47.41±0.30 
Blade width 25.5–30.3 49.3–52.1 34.9–38.2 20.4–47.3a −21.84 5.74 −3.81 
28.05±0.46 49.89±0.48 36.72±0.37 33.18±0.23 
Sepal stalk length 8.2–11.5 39.2–41.9 18.4–20.9 10.81–32.4 −31.01 4.06 −7.64 
9.36±0.25 40.37±0.34 19.81±0.22 20.45±0.20 
Anther extension 1.7–8.8 −14.7 to −10.5 −7.07 to −3.1 −25.05 to 7.61 15.15 5.90 2.57 
2.72±0.56 −12.43±0.56 −4.67±0.34 −7.42±0.25 
L38.77–121.3 87.59–109.76 29.53–66.7 22.43–199.54a −23.75 23.12 −1.03 
74.13±7.15 97.88±2.25 48.12±18.59 85.98±1.63 
a142.46–172.99 150.5–156.69 134.91–152.27 131.23–176.42b 4.92 10.14 0.49 
157.64±2.74 152.72±0.75 143.59±8.68 150.19±0.43 
b134.29–165.87 81.32–91.74 115.42–123.99 16.64–199.39a,b 60.04 6.85 8.76 
148.63±2.72 88.59±1.15 119.71±4.29 116.87±0.88 

Mean trait values and standard errors are displayed. The DSM is the difference in the species mean trait values (IF − IH). The environmental standard deviation (ESD) ranges are given for each trait and the trait averages across the flowers on a single plant are shown with standard errors.

aDenotes negative transgression.

bIndicates positive transgression.

Measurement of Phenotypic Traits

Floral and vegetative traits were measured for the parents, F1s and F2s, to capture the range of phenotypic variation found in I. fulva, I. hexagona, and early generation hybrids. Phenotypic measurements (n = 328) were collected in 2009. The date of each flower opening was recorded along with the length of the tallest leaf. Floral measurements were collected on the day of anthesis of the second flower and included: flower height from the base of the rhizome, sepal blade length, sepal blade width, sepal stalk length, anther extension, and the area of the floral guide (i.e., roughly triangular area calculated using the formula 1/2 length × width) (Bouck et al. 2007). All trait values were averaged for each individual (range = 1 to 15 flowers per plant; median = 3), and Pearson’s correlations and pairwise comparisons were calculated in JMP 9.0 Pro (SAS Institue, Cary, NC) for each class (e.g., I. fulva, I. hexagona, F1, and F2) and trait combination, respectively.

The second flower from each plant was placed in a light box with a standardized white background and photographed (Nikon Coolpix 5000, macro setting, point auto focus, no flash). The camera was anchored 4″ above the light box. A Gretag McBeth Color checker chart (X-Rite, Inc.) was also photographed under the same conditions, and the Adobe Photoshop inCamera 4.0.1 filter plug was used to generate an ICC profile for each image to control for shadows or other artifacts associated with the 3D dimensionality of the flowers. Color values (3 per plant) were analyzed in Adobe Photoshop by selecting a circular area of 120×120 (14 400 total) pixels on the distal tip of each petal. Average L*a*b* values and a composite score were calculated for each plant. L*a*b* values were used because they are device independent and cover a larger color spectrum than RGB or CMYK values (Yam and Papadakis 2004). The L* value represents a lightness score, the a* component measures from green to red, and the b* component covers the blue to yellow spectrum (Yam and Papadakis 2004; Lehnert et al. 2011).

Transgressive Traits

Transgressive traits were calculated following the study by Rieseberg et al. (2003). In brief, transgressive traits were defined as traits in which the range of the F2 values exceeds or falls below the parental values by at least 2 SDs (Table 1).

Codominant Genetic Map Construction QTL mapping

Marker design and genotyping and linkage map construction are described in detail in the study by Ballerini et al. (2012). Briefly, 293 codominant SNP markers were genotyped using the GoldenGate Genotyping Assay for the VeraCode platform (Illumina, Inc., San Diego, CA), and a linkage map composed of 22 linkage groups (LGs) was assembled using MapMaker 3.0 (Ballerini et al., in review). Genome wide scans for quantitative trait loci (QTL) were performed using composite interval mapping (CIM) following the method of Zeng (1993, 1994) using QTL Cartographer (WinQTLcart 2.5), and map figures were produced using MapChart (2.2, Vorrips 2002). Each trait was analyzed separately using a forward and backward stepwise regression using 1000 permutations with a significance level of 0.05 to determine the threshold cutoff for each QTL.

To determine whether pairs of established QTL better explain the data in combination than individually, we performed a 2-way scan for epistasis. This scan was carried out using the scantwo function in Rqtl (Broman et al. 2003). Three odds ratios were calculated: 1) a full model that takes in account the additive effect at each of the 2 loci and the effect of their interaction, 2) a 2 locus additive-only model that only takes into account each locus’ independent additivity, and 3) an interaction likelihood that is the difference between the 2 locus full model (including the interaction) and the 2 locus strictly additive model (with no interaction effects). The full and additive model LOD scores were derived from a comparison to a null (no QTL at either locus nor an interaction.) Significances of interactions among marker pairs were assessed with interaction variances from a 2-way ANOVA on the raw marker data.

Results

Phenotypic Effects of Hybridization

The I. fulva and I. hexagona parents differed for all traits measured in this study with the exception of flowering time, which is consistent with observations of the 2 species in nature (Table 1). The F1, used as both a pollen and a seed parent, was intermediate for all of the traits included in this study (Table 1). For each trait, the difference in the species mean (DSM) and the environmental standard deviation (ESD) were calculated (Fishman et al. 2002; Table 1). These values were used to estimate the degree to which I. fulva and I. hexagona differ while controlling for environmental noise (i.e., DSM/ESD) (Fishman et al. 2002, Bouck et al. 2007). Trait values in I. hexagona were generally larger than those in I. fulva with the exception of anther extension, which was more exerted in I. fulva to facilitate the distribution of pollen by hummingbirds (Emms and Arnold 2000). The b* component of color differed the most; I. hexagona was more blue and I. fulva was closer to the yellow end of the spectrum. In the F2 population, traits spanned the range of phenotypes of the parents, with 7 traits exhibiting transgression (Table 1). Negative transgression was observed for leaf length, blade length, blade width, and 2 color components, L* (lightness scale) and b* (blue to yellow scale). Positive transgression was observed for floral guide area, a* (green to red scale) and b*. For example, some F2 floral guides were larger than the largest floral guide of I. hexagona by at least 2 SDs.

Correlations Among Traits

Trait correlations were uncommon in the parental and F1 generations (likely due to small sample sizes) but abundant in the F2 generation (Table 2). The DSM was calculated by subtracting the mean trait value of I. hexagona from the mean trait value of I. fulva (Table 1). Flower height, leaf length, floral guide area, blade length, blade width, and sepal stalk length were significantly different (P < 0.0001) in I. hexagona and I. fulva. The L* and b* values also differed between the 2 species (Figure 1; P < 0.05 and P < 0.001, respectively).

Table 2

Trait correlations for the F2 hybrid data

F2 hybrids Days to flower Flower height Leaf length Floral guide area Blade length Blade width Sepal stalk length Anther extension 
Days to flower — — — — — — — — 
Flower height 0.1138* — — — — — — — 
Leaf length 0.1950* 0.4162* — — — — — — 
Floral guide area 0.0423 0.0484 −0.0271 — — — — — 
Blade length −0.0545 0.2545* 0.2280* 0.0086 — — — — 
Blade width 0.0096 0.1681* 0.0643 0.0065 0.7154* — — — 
Sepal stalk length −0.2394* 0.1976* 0.1096* 0.2463* 0.3666* 0.2519* — — 
Anther extension −0.0379 −0.0929 −0.0629 0.0274 −0.0863 −0.1846* −0.1711 — 
F2 hybrids Days to flower Flower height Leaf length Floral guide area Blade length Blade width Sepal stalk length Anther extension 
Days to flower — — — — — — — — 
Flower height 0.1138* — — — — — — — 
Leaf length 0.1950* 0.4162* — — — — — — 
Floral guide area 0.0423 0.0484 −0.0271 — — — — — 
Blade length −0.0545 0.2545* 0.2280* 0.0086 — — — — 
Blade width 0.0096 0.1681* 0.0643 0.0065 0.7154* — — — 
Sepal stalk length −0.2394* 0.1976* 0.1096* 0.2463* 0.3666* 0.2519* — — 
Anther extension −0.0379 −0.0929 −0.0629 0.0274 −0.0863 −0.1846* −0.1711 — 

*Denotes significance at P ≤ 0.05.

Figure 1.

QTL map of I. fulva and I. hexagona. Solid bars indicate a “positive” trait value (increases the trait value) and empty bars indicate a negative trait value in reference to I. fulva alleles.

Figure 1.

QTL map of I. fulva and I. hexagona. Solid bars indicate a “positive” trait value (increases the trait value) and empty bars indicate a negative trait value in reference to I. fulva alleles.

QTL Analyses

In this study, QTL (n = 28) were identified for traits on 14 LGs (Figure 1; Table 3), and several regions showing overlapping QTL in traits important for pollinator attraction were revealed. Specifically, overlapping LOD intervals were observed on LG 2 for blade width, sepal stalk length, and anther extension; on LG 6 for sepal stalk length and color lightness (L*); on LG 12a for blade width and sepal stalk length; on LG 17 for blade width and color lightness (L*) (Figure 1). Sepal stalk length was correlated with blade width and blade length (r = 0.2262, P <0.0001; r = 0.3387, P <0.0001, respectively), which is consistent with the presence of colocalized QTL for these 3 traits.

Table 3

Location of each QTL is presented as the LG followed by position on the LG (in Kosambi cM) and the 2-LOD confidence intervals

Trait LG Position 2-LOD interval LR Additive effect Dominance effect R2 QTL effect sizes 
Leaf length (cm) 13 2.7 1.1–18.7 20.24 3.78 3.46 0.14 15 
Floral guide area (mm) 84.3 83.1–90.5 157.98 −8.39 −11.95 0.48 40 
Blade length (mm) 34.7 21.3–39.5 46.82 3.08 −0.43 0.30 20 
1a 8.2 0.0–13.5 19.18 −2.33 −1.22 0.31 15 
Blade width (mm) 0.0 0.0–22.2 29.80 −2.07 0.55 0.23 19 
12a 10.4 5.4–19.2 15.24 −2.02 1.41 0.26 18 
15 31.3 28.4–44.7 15.30 −1.24 −0.06 0.27 11 
16 20.5 12.8–25.5 15.52 −1.85 0.68 0.25 17 
17 5.3 0.0–5.4 28.20 −2.12 0.21 0.23 19 
Sepal stalk length (cm) 3.0 0.0–19.5 22.04 −1.23 0.12 0.33 
51.2 50.5–66.9 61.03 −2.23 −0.26 0.34 14 
67.7 66.9–73.8 58.16 −2.06 −0.69 0.34 13 
8/18 12.8 9.3–27.8 20.64 −1.63 0.63 0.34 
63.7 54.3–65.7 20.05 −1.64 0.94 0.37 11 
12a 5.2 1.8–5.3 24.86 −1.54 0.037 0.33 10 
12a 12.9 9.2–16.9 27.88 −1.73 0.44 0.34 11 
Anther extension (mm) 0.2 0.0–24.0 28.25 1.91 −0.49 0.27 25 
0–12.7 17.88 1.26 0.24 0.27 17 
43 35.3–47.2 20.77 1.19 0.51 0.27 16 
0–26 20.13 1.59 −0.14 0.27 21 
8/18 36.1 35.3–47.2 23.01 1.07 0.92 0.27 14 
L56.4 51.8–59.9 45.19 −17.59 0.48 0.33 146 
11 27.6 27.1–31.6 35.73 16.5 −9.27 0.32 139 
17 1.6 0.0–5.3 21.96 −5.75 −7.24 0.33 48 
a9.8 1.4–18.8 20.40 2.77 0.52 0.19 112 
65.5 58.2–75 22.86 3.41 −2.21 0.19 138 
b20.8 18.8–29.5 39.14 10.10 −3.55 0.27 34 
12.3 6.4–19.9 17.51 3.94 1.63 0.31 13 
8/18 8.9 0.4–12.6 21.97 6.26 −0.09 0.24 21 
Trait LG Position 2-LOD interval LR Additive effect Dominance effect R2 QTL effect sizes 
Leaf length (cm) 13 2.7 1.1–18.7 20.24 3.78 3.46 0.14 15 
Floral guide area (mm) 84.3 83.1–90.5 157.98 −8.39 −11.95 0.48 40 
Blade length (mm) 34.7 21.3–39.5 46.82 3.08 −0.43 0.30 20 
1a 8.2 0.0–13.5 19.18 −2.33 −1.22 0.31 15 
Blade width (mm) 0.0 0.0–22.2 29.80 −2.07 0.55 0.23 19 
12a 10.4 5.4–19.2 15.24 −2.02 1.41 0.26 18 
15 31.3 28.4–44.7 15.30 −1.24 −0.06 0.27 11 
16 20.5 12.8–25.5 15.52 −1.85 0.68 0.25 17 
17 5.3 0.0–5.4 28.20 −2.12 0.21 0.23 19 
Sepal stalk length (cm) 3.0 0.0–19.5 22.04 −1.23 0.12 0.33 
51.2 50.5–66.9 61.03 −2.23 −0.26 0.34 14 
67.7 66.9–73.8 58.16 −2.06 −0.69 0.34 13 
8/18 12.8 9.3–27.8 20.64 −1.63 0.63 0.34 
63.7 54.3–65.7 20.05 −1.64 0.94 0.37 11 
12a 5.2 1.8–5.3 24.86 −1.54 0.037 0.33 10 
12a 12.9 9.2–16.9 27.88 −1.73 0.44 0.34 11 
Anther extension (mm) 0.2 0.0–24.0 28.25 1.91 −0.49 0.27 25 
0–12.7 17.88 1.26 0.24 0.27 17 
43 35.3–47.2 20.77 1.19 0.51 0.27 16 
0–26 20.13 1.59 −0.14 0.27 21 
8/18 36.1 35.3–47.2 23.01 1.07 0.92 0.27 14 
L56.4 51.8–59.9 45.19 −17.59 0.48 0.33 146 
11 27.6 27.1–31.6 35.73 16.5 −9.27 0.32 139 
17 1.6 0.0–5.3 21.96 −5.75 −7.24 0.33 48 
a9.8 1.4–18.8 20.40 2.77 0.52 0.19 112 
65.5 58.2–75 22.86 3.41 −2.21 0.19 138 
b20.8 18.8–29.5 39.14 10.10 −3.55 0.27 34 
12.3 6.4–19.9 17.51 3.94 1.63 0.31 13 
8/18 8.9 0.4–12.6 21.97 6.26 −0.09 0.24 21 

LOD = Log Odds distance, LG = Linkage Group.

An estimate of the additive allelic effect and an estimate of the dominance effect and the percentage of variance explained (R2). The QTL effect sizes are shown relative to the differences in the species mean (Fishman et al. 2002, Bouck et al. 2007).

The additive effect, dominance effect, and direction of QTL for traits are presented in Table 3. In Figure 1, solid bars indicate that I. fulva alleles increase the trait value, whereas open bars indicate that the I. hexagona alleles increase the trait value. For some traits, the effect direction of alleles at QTL was as predicted by the species averages. For example, I. fulva alleles at all of the QTL for anther extension had positive additive effects (i.e., alleles from I. fulva increased anther extension). Other traits were influenced by multiple QTLs with mixed effects. Sepal stalk length, blade width, and the L* trait for petal color lightness each had multiple QTL with some having positive additive effects (i.e., alleles from I. fulva increase the trait value) and others with negative additive effects (i.e., alleles from I. hexagona increase the trait value). More specifically, for 4 QTL (LGs 2, 6, 9, and 8/18), I. fulva alleles decreased sepal stalk length, and for 2 QTL (both on LG 12a), I. fulva alleles increased sepal stalk length. Similarly, 4 QTL (LGs 2, 15, 16, and 17) were identified where alleles from I. fulva decreased blade width and 1 QTL where alleles from I. fulva that increased blade width (LG 12a). Three QTL controlling lightness were identified, 2 in which I. fulva alleles increased lightness relative to I. hexagona alleles (LGs 11 and 17) and 1 in which I. fulva alleles decreased lightness with respect to I. hexagona alleles (LG 6). A QTL for leaf length (LG 13) was identified at which I. fulva alleles increased leaf length, which was counter to what might be expected because I. fulva has shorter leaves than I. hexagona. Of the 8 traits analyzed, 4 (leaf length, floral guide area, blade width, and sepal stalk length) showed significant signs of epistasis among QTL (Table 4). This does not mean that all other QTL act independently. It only means that the strength of the interaction is not strong enough to make it past the multiple comparisons tests.

Table 4

Estimation of interactions among QTL

 Mark1 Mark2 LG LG LOD.full LOD.add LOD.int F P 
Leaf length 342 78 10.88 7.89 2.986 4.260 0.002 
342 259 9.16 7.46 1.703 2.527 0.041 
Floral guide area 376 312 36.91 33.84 3.071 5.352 0.000 
376 92 11 32.09 31.25 0.839 3.622 0.007 
Blade width 148 175 8/18 10.19 5.91 4.281 5.272 0.000 
Sepal stalk length 243 10.04 6.29 3.751 4.328 0.002 
56 136 8/18 14 7.89 5.66 2.225 3.928 0.004 
247 369 14 1a 7.3 5.2 2.105 3.717 0.006 
109 53 13 18.62 15.69 2.934 3.303 0.011 
148 80 1a 7.94 5.57 2.371 2.707 0.031 
315 311 8/18 9.46 7.4 2.068 2.445 0.047 
 Mark1 Mark2 LG LG LOD.full LOD.add LOD.int F P 
Leaf length 342 78 10.88 7.89 2.986 4.260 0.002 
342 259 9.16 7.46 1.703 2.527 0.041 
Floral guide area 376 312 36.91 33.84 3.071 5.352 0.000 
376 92 11 32.09 31.25 0.839 3.622 0.007 
Blade width 148 175 8/18 10.19 5.91 4.281 5.272 0.000 
Sepal stalk length 243 10.04 6.29 3.751 4.328 0.002 
56 136 8/18 14 7.89 5.66 2.225 3.928 0.004 
247 369 14 1a 7.3 5.2 2.105 3.717 0.006 
109 53 13 18.62 15.69 2.934 3.303 0.011 
148 80 1a 7.94 5.57 2.371 2.707 0.031 
315 311 8/18 9.46 7.4 2.068 2.445 0.047 

Presented are the marker names nearest the QTL being tested and the chromosome number with which the markers are associated. LOD.full is the log-likelihood ratio of the full model (Q1 + Q2 + Q1 × Q2), LOD.add is the log-likelihood ratio of the additive model (Q1 + Q2), and LOD.int is the log-likelihood ratio of the difference between the 2 models (full model − additive model). F ratios and P values are from the interaction term of a 2-way Anova with phenotype as the response variable and genotypic configuration at the presented markers as factors.

QTL Effect Sizes

QTL effect sizes, calculated relative to the DSM, ranged in magnitude from 8% to 139% of the species difference (Bouck et al. 2007; Table 3). Color QTL (i.e., L*, a*, b*) exhibited the highest values, potentially due to the transgressive nature of these traits relative to the parents. Color QTL were distributed across 7 LGs demonstrating the complexity and polygenic nature of floral color. A majority of the color QTL (7/8) did not colocalize with QTL for other traits, the one exception being on LG17, where the QTL LOD intervals for L* overlapped with the QTL LOD intervals for blade width. QTL for all other traits (excluding color QTL) had effect sizes between 8% and 40%, with the average effect equaling 16% of the phenotypic species difference. Floral guide area and 2 L* QTL exhibited large dominance effects (Table 3). The only QTL identified for leaf length was highest in plants homozygous for I. fulva-like alleles (Figure 1).

Discussion

Studies of both plant and animal species have demonstrated that hybridization has the potential to generate phenotypic novelty in wild populations (Rieseberg et al. 1996; Arnold 2006; Parsons et al. 2011; Twyford and Ennos 2012). Given the environmental stochasticity of the southern regions of Louisiana (i.e., frequent flooding, drought, hurricanes, and salinization), the genetic variation introduced through interspecific hybridization may be advantageous. In this study, we evaluated how the underlying genetic architecture of floral and vegetative traits might influence quantitative trait variation and phenotypic expression in an F2 hybrid population of I. hexagona and I. fulva. These 2 species are naturally interfertile in southern Louisiana where phenotypically detectable hybrid zones are observed. Previous studies from several species have demonstrated that pollinators prefer floral traits associated with their expected pollination syndrome and that changes in allele frequency or reciprocal introgression of interspecific QTL can alter pollinator visitation. For example, shifts in pollinator visitation in response to color cues and nectar content in Mimulus have been shown to be controlled by a single mutation of large effect (Bradshaw and Schemske 2003), changes in floral morphology in Aquilegia are controlled by few QTL that affect pollinator visitation (Hodges et al. 2002), and pollinator shifts occur in response to scent controlled by few QTL in Petunia (Klahre et al. 2011). In Irises, bees prefer I. hexagona and hummingbirds prefer I. fulva due to several factors that contribute to pollinator success, including floral morphology (Emms and Arnold 2000). In previous studies, the smaller sepals of I. fulva proved challenging to bees that typically crawl into a flower, whereas the larger flowers of I. hexagona prohibited hummingbirds from acquiring nectar (Emms and Arnold 2000). Knowledge of the genetic basis of phenotypic differences is essential for understanding the processes that both facilitate hybridization and also maintain species integrity in wild populations. In this report, we demonstrated that there are many QTL underlying the traits comprising the pollination syndrome in Iris, and that while there is some overlap, there is also independence among QTL allowing traits to assort individually in hybrid populations. Furthermore, identifying QTL for traits known to be important for pollinator attraction and the successful transfer of gametes suggests that introgression of even single or a few QTL regions across species boundaries could alter pollinator visitation.

Transgression

Recombination of QTL in which alleles from the same parent have opposing effects can produce hybrid genotypes that are phenotypically extreme or novel in relation to the parental phenotypes (Rieseberg et al. 2003). Studies in I. fulva and Iris brevicaulis found evidence for transgression in 6 out of the 9 traits measured (Bouck et al. 2007); similarly, here we found evidence for transgression in 7 out of the 11 traits examined. QTL of mixed effect, such as those we observed, suggest that either genetic drift or antagonistic selection may be at work in these 2 species (Rieseberg et al. 2003). Due to the highly variable conditions in the coastal habitats (i.e., flooding, drought, and hurricanes), it is not surprising that selection pressures would fluctuate. Although we cannot be certain of the causes of the transgression observed in this study, it is clear that hybridization could be a potential source of novel phenotypic variation in wild populations. Indeed, a stabilized homoploid hybrid species derived from hybridization between I. fulva, I. hexagona, and a third species, I. brevicaulis, has phenotypic means outside of the parental species means for a number of traits (Randolph 1966).

QTL Important for Pollination

The 28 QTL identified in this study associated with floral traits known to be important for pollinator choice are useful for understanding how the genetic architecture influences quantitative trait variation in Louisiana Irises. Two traits, days to flower and flower height, were not associated with a QTL. Because the flowering phenology is similar between the 2 species, it may be difficult to detect an association between a particular locus and flowering time in this population; however, numerous QTL have been identified for flowering time between I. fulva and I. brevicaulis (another closely related species), which have disparate floral phenology (Ballerini et al. 2012). Previous studies in Irises have found that flowering time is a complex trait associated with many QTL regions that vary with environmental conditions (e.g., wet vs. dry) and developmental stages (e.g., age), which may also reduce the ability to detect QTL for this trait (Ballerini et al. 2012). We did not detect a QTL for flower height, which is surprising given that flower height is quite different in I. fulva and I. hexagona (Table 1), and flower height serves as an important long-range cue to pollinators (Emms and Arnold 2000). However, it is possible that epistatic interactions among loci obscured QTL associated with flower height.

Traits such as floral guide area, which has a single QTL that explains ~50% variation among individuals, and blade length, which has 2 essentially additive QTL that explain upwards of 60% of the variation among individuals, suggest that these traits are controlled by QTL of large effect. With regards to blade length, alternating a single allele from either of the QTL changes the trait value by 15–20%, leaving the opportunity for a sweep to fixation to produce a 30–40% change with 1 loci or approximately 70% with both loci. With the floral guide area QTL, the likelihood of an introgressive sweep is direction dependent as this QTL is swamped with dominance (Figure 2). The transition from homozygous HH to FH at this QTL causes a reduction of floral guide area by an average of 23mm2, but because of dominance, the transition between a homozygote (FF) and heterozygote (FH) produces effectively no change in trait value, making sweeps improbable in this direction. Another complication is the complexity of a trait’s genetic architecture. In this study, sepal stalk length is controlled by a network of at least 12 loci. With this trait, the combination of alleles needed to produce trait change may hinder an interspecies sweep. Co-occurring QTL on LG 2, LG 6, LG 8/18, LG 12a, and LG 17 show that there may be regions of gene clustering for traits important for pollination, suggesting that despite the phenotypic diversity observed in hybrids, there is selection for traits important for fitness outcomes to act in concert (Arnold 2006).

Figure 2.

Examples of dominance effects for 4 traits. (a) Trait values for floral guide area (LG 6) are strongest when I. hexagona alleles are present at MK 376. (b) Blade width (LG 15) is affected only slightly by dominance using MK 245. (c) The QTL for leaf length (LG 13) with opposite effects to what was expected based on parental phenotypes at MK 348. (d) The QTL for anther extension (LG 2) is associated with least recession of the anther in the FF homozygote at MK 094.

Figure 2.

Examples of dominance effects for 4 traits. (a) Trait values for floral guide area (LG 6) are strongest when I. hexagona alleles are present at MK 376. (b) Blade width (LG 15) is affected only slightly by dominance using MK 245. (c) The QTL for leaf length (LG 13) with opposite effects to what was expected based on parental phenotypes at MK 348. (d) The QTL for anther extension (LG 2) is associated with least recession of the anther in the FF homozygote at MK 094.

QTL Effects

Similar to other studies examining the underlying genetic architecture of floral traits, we found that many QTL of varying effect sizes contribute to the morphological characteristics known to influence pollinator visitation. However, the average effect sizes associated with QTL in this study were smaller relative to those reported in earlier studies of Irises (Bouck et al. 2007). This may be due in part to our relatively larger sample size of F2s, as small sample sizes can exaggerate the effects of detected QTL (Beavis 1994).

Extreme reshuffling of genetic material during hybridization promotes variation and has been proposed as a mechanism for avoiding genetic constraints (Parsons et al. 2011). We found that not only were we able to recover the full range of phenotypic variation expressed in the parental species but we also observed transgression for several traits measured in this study, consistent with this hypothesis.

The effect of additivity can be interpreted as the average change in the trait value when an alternative allele is substituted, which in this study can be interpreted as the acquisition of an I. fulva allele. The floral guide area exhibits both strong additive and dominance effects; the floral guide area in the presence of an I. fulva allele is dramatically reduced (Table 3; Figure 2). It is thus possible that the floral guide area phenotype is controlled by a QTL of major effect. Similarly, the additive effects were especially strong in the color QTL, where 2 out of 3 L* QTL reduce the lightness of the petals and all of the a* and b* QTL increase. Overall, this is consistant with I. fulva alleles increasing both the green to red values (a*) and the blue to yellow (b*) values. However, the lack of overlapping QTL for the color traits (L*, a*, and b*) is surprising given that color traits are in other species have been shown to be under the control of relatively few genes (Bradshaw and Schemske 2003, Zufall and Rausher 2003, Rausher 2008).

Potential for Introgression

Given the wide range of phenotypic divergence and the distribution of QTL observed in this F2 population, it is likely that I. fulva and I. hexagona have genomic regions that are potentially permeable to introgression. Iris nelsonii, which is thought to be a composite species of I. fulva, I. brevicaulis, and I. hexagona (Arnold 1993) and may have evolved to inhabit intermediate ecological niche space (Taylor et al. 2011), also supports hybridization and introgression, as a potential cause of species diversification.

Conclusions

The transfer of adaptive alleles across species boundaries can act as an evolutionary catalyst. The underlying genetic architecture of flowering traits in I. hexagona and I. fulva as outlined in this study demonstrates that hybridization can lead to transgression, thereby increasing phenotypic variation, though in this study these effects do not appear to be the result of overdominance. Furthermore, hybridization among species in a fluctuating environment may facilitate adaptation by increasing the genetic variation on which selection can act. For example, hybridization may introduce novel genotypes with a range of phenotypes—including extreme phenotypes relative to the parental species, which may be attractive to a wider range of pollinators, or increase the geographic range in which the plant can survive. The number and distribution of QTL for floral traits identified in this study are useful for understanding how floral traits assort both within and across species boundaries. For example, correlations among traits that are found in the same genome region suggest that traits important for pollinator attraction are genetically linked, whereas others assort independently. Future studies to evaluate whether hybrids between I. hexagona and I. fulva have demonstrable fitness advantages under a range of ecological conditions will further aid in understanding the role of hybridization in this species complex.

Supplementary Material

Supplementary material can be found at http://www.jhered.oxfordjournals.org/.

Funding

This work was supported by National Science Foundation grants DEB-0949479/0949424 (collaborative grant between M.L.A. and NH Martin of Texas State University) and DEB-1049757 (M.L.A.) and by funds from the Office of the Vice President for Research at the University of Georgia.

Acknowledgments

The authors thank M. Boyd and K. Tarner for assistance with plant maintenance, J. Aderinwale and J. Foley their contributions to lab and greenhouse work, and W. Bunch for assistance with floral phenotyping.

References

Anderson
E
.
1949
.
Introgressive hybridization
 .
New York
:
Wiley
.
Arnold
ML
.
1993
.
Iris Nelsonii (Iridaceae)—origin and genetic composition of a homoploid hybrid species
.
Am J Bot.
 
80
:
577
583
.
Arnold
ML
.
2006
.
Evolution through genetic exchange
 .
Oxford
:
Oxford University Press
.
Arnold
ML
Bennett
BD
Zimmer
EA
.
1990
.
Natural hybridization between Iris fulva and Iris hexagona—patterns of ribosomal DNA variation
.
Evolution
 .
44
:
1512
1521
.
Ballerini
ES
Brothers
AN
Tang
S
Knapp
SJ
Bouck
A
Taylor
SJ
Arnold
ML
Martin
NH
.
2012
.
QTL mapping reveals the genetic architecture of loci affecting pre- and post-zygotic isolating barriers in Louisiana Iris
.
BMC Plant Biol
 .
12
:
91
.
Ballerini
ES
Mockaitis
K
Barg
JG
Brothers
AN
Knapp
SJ
Martin
NH
Arnold
ML
.
Forthcoming
. Comparative mapping reveals different patterns of transmission ratio distortion in crosses of Iris fulva to Iris brevicaulis versus Iris hexagona. In Review.
Beavis
WD
.
1994
.
The power and deceit of QTL experiments: Lessons from comparative QTL studies. In: DB Wilkinson, editor. 49th Ann Corn Sorghum Res Conf
.
Chicago: Am Seed Trade Assoc.
  pp.
250
266
.
Bouck
A
Wessler
SR
Arnold
ML
.
2007
.
QTL analysis of floral traits in Louisiana iris hybrids
.
Evolution
 .
61
:
2308
2319
.
Bradshaw
HD
Schemske
DW
.
2003
.
Allele substitution at a flower colour locus produces a pollinator shift in monkeyflowers
.
Nature
 .
426
:
176
178
.
Broman
KW
Wu
H
Sen
S
Churchill
GA
.
2003
.
R/qtl: QTL mapping in experimental crosses
.
Bioinformatics
 .
19
:
889
890
.
Donovan
LA
Rosenthal
DR
Sanchez-Velenosi
M
Rieseberg
LH
Ludwig
F
.
2010
.
Are hybrid species more fit than ancestral parent species in the current hybrid species habitats?
J Evol Biol
 .
23
:
805
816
.
Emms
SK
Arnold
ML
.
2000
.
Site-to-site differences in pollinator visitation patterns in a Louisiana iris hybrid zone
.
Oikos
 .
91
:
568
578
.
Fishman
L
Kelly
AJ
Willis
JH
.
2002
.
Minor quantitative trait loci underlie floral traits associated with mating system divergence in Mimulus
.
Evolution
 .
56
:
2138
2155
.
Hodges
SA
.
1997
.
A rapid adaptive radiation via a key innovation in Aquilegia
. In:
Givinish
T
Sytsma
K
, editors.
Molecular evolution and adaptive radiations
 .
Cambridge (UK)
:
Cambridge University Press
. p.
391
405
.
Hodges
SA
Whittall
JB
Fulton
M
Yang
JY
.
2002
.
Genetics of floral traits influencing reproductive isolation between Aquilegia formosa and Aquilegia pubescens
.
Am Nat
 .
159
(
Suppl 3
):
S51
S60
.
Howarth
DG
Baum
DA
.
2005
.
Genealogical evidence of homoploid hybrid speciation in an adaptive radiation of Scaevola (goodeniaceae) in the Hawaiian Islands
.
Evolution
 .
59
:
948
961
.
Kay
KM
Sargent
RD
.
2009
.
The role of animal pollination in plant speciation: integrating ecology, geography, and genetics
.
Ann Rev Ecol Evol Syst.
 
40
:
637
656
.
Klahre
U
Gurba
A
Hermann
K
Saxenhofer
M
Bossolini
E
Guerin
PM
Kuhlemeier
C
.
2011
.
Pollinator choice in Petunia depends on two major genetic Loci for floral scent production
.
Curr Biol
 .
21
:
730
739
.
Lehnert
MS
Balaban
MO
Emmel
TC
.
2011
.
A new method for quantifying the color of insects
.
Fl Entom.
 
94
:
201
207
.
Mallet
J
.
2005
.
Hybridization as an invasion of the genome
.
Trends Ecol Evol
 .
20
:
229
237
.
Parsons
KJ
Son
YH
Albertson
RC
.
2011
.
Hybridization promotes evolvability in African Cichlids: connections between transgressive segregation and phenotypic integration
.
Evol Biol.
 
38
:
306
315
.
Puzey
JR
Gerbode
SJ
Hodges
SA
Kramer
EM
Mahadevan
L
.
2012
.
Evolution of spur-length diversity in Aquilegia petals is achieved solely through cell-shape anisotropy
.
Proc Biol Sci
 .
279
:
1640
1645
.
Randolph
LF
.
1996
.
Iris nelsonii, a new species of Louisiana Iris of hybrid origen
.
Baileya.
 
14
:
143
163.
Rausher
MD
.
2008
.
Evolutionary transitions in flower color
.
Int J Plant Sci.
 
169
:
7
21.
Rieseberg
LH
Arias
DM
Ungerer
MC
Linder
CR
Sinervo
B
.
1996
.
The effects of mating design on introgression between chromosomally divergent sunflower species
.
Theor App Genet.
 
93
:
633
644
.
Rieseberg
LH
Raymond
O
Rosenthal
DM
Lai
Z
Livingstone
K
Nakazato
T
Durphy
JL
Schwarzbach
AE
Donovan
LA
Lexer
C
.
2003
.
Major ecological transitions in wild sunflowers facilitated by hybridization
.
Science
 .
301
:
1211
1216
.
Riley
HP
.
1938
.
A character analysis of colonies of Iris fulva, iris hexagona var. giganticaerulea and natural hybrids
.
Am J Bot.
 
25
:
727
738
.
Stuurman
J
Hoballah
ME
Broger
L
Moore
J
Basten
C
Kuhlemeier
C
.
2004
.
Dissection of floral pollination syndromes in Petunia
.
Genetics
 .
168
:
1585
1599
.
Taylor
SJ
Willard
RW
Shaw
JP
Dobson
MC
Martin
NH
.
2011
.
Differential response of the homoploid hybrid species Iris nelsonii (Iridaceae) and its progenitors to abiotic habitat conditions
.
Am J Bot.
 
98
:
1309
1316
.
Twyford
AD
Ennos
RA
.
2012
.
Next-generation hybridization and introgression
.
Heredity (Edinb)
 .
108
:
179
189
.
Viosca
PJ
.
1935
.
The irses of southeastern Louisiana—a taxonomic and ecological interpretation
.
Bull Am Iris Soc.
 
57
:
3
56
.
Vorrips
RE
.
2002
.
MapChart: software for the graphical presentation of linkage maps and QTLs
.
J Hered.
 
93
:
77
78
.
Wesselingh
RA
Arnold
ML
.
2000
.
Pollinator behaviour and the evolution of Louisiana iris hybrid zones
.
J Evol Biol.
 
13
:
171
180
.
Whitney
KD
Ahern
JR
Campbell
LG
Albert
LP
King
MS
.
2010
.
Patterns of hybridization in plants
.
Pers Plant Ecol Evol Syst.
 
12
:
175
182
.
Widmer
A
Lexer
C
Cozzolino
S
.
2009
.
Evolution of reproductive isolation in plants
.
Heredity (Edinb)
 .
102
:
31
38
.
Yam
KL
Papadakis
SE
.
2004
.
A simple digital imaging method for measuring and analyzing color of food surfaces
.
J Food Eng.
 
61
:
137
142
.
Zeng
ZB
.
1993
.
Theoretical basis for separation of multiple linked gene effects in mapping quantitative trait loci
.
Proc Natl Acad Sci USA
 .
90
:
10972
10976
.
Zeng
ZB
.
1994
.
Precision mapping of quantitative trait loci
.
Genetics
 .
136
:
1457
1468
.
Zufall
RA
Rausher
MD
.
2004
.
Genetic changes associate with floral adaptation restrict future evolutionary potential
.
Nature
 .
428
:
847
850
.

Author notes

Corresponding Editor: John Stommel
Data deposited at Dryad: http://dx.doi.org/10.5061/dryad.fq48q