vol. 159, supplement

the american naturalist

march 2002

Genetics of Floral Traits Influencing Reproductive Isolation between Aquilegia formosa and Aquilegia pubescens

Scott A. Hodges,* Justen B. Whittall, Michelle Fulton, and Ji Y. Yang

Department of Ecology, Evolution, and Marine Biology, University of California, Santa Barbara, California 93106, and the White Mountain Research Station, Bishop, California 93514

abstract: Reproductive isolation between Aquilegia formosa and Aquilegia pubescens is influenced by differences in their flowers through their effects on pollinator visitation and pollen transfer. Here, we investigate the genetic basis of floral characters differentiating these species. We found that in addition to the effects of flower orientation and the length of nectar spurs previously described, other characters such as flower color or odor affect hawkmoth visitation. Repeatability of measurements in an F2 population ranged from 0.53 to 0.83 among five floral traits, indicating that using the means of multiple measures per plant will substantially increase the power of a quantitative trait locus (QTL) analysis. Integration of floral traits was indicated by significant correlations among traits in an F2 population. In a separate F2 population we found that QTL for different floral traits were often closely associated, indicating that linkage or pleiotropy cause at least some of this integration. In addition, we found QTL for all floral traits examined. Because Aquilegia species are largely interfertile and vary extensively in both floral morphology and ecology, they offer the opportunity for QTL studies of a wide range of characters affecting reproductive isolation. Keywords: flower color, orientation, QTL, nectar spurs.

The ability to dissect the genetic basis of complex traits has generally been possible for only a few model organisms (Coyne 1995). Recently, however, the tools of molecular biology have begun to provide the large number of genetic markers needed for detailed genetic analyses of many natural species. These advances offer the opportunity to determine whether there are general patterns to the genetic architecture of speciation (e.g., Bradshaw et al. 1995, 1998; Rieseberg et al. 1996), and therefore, it is important to determine the genetic basis of reproductive isolation for a number of species. Conducting a quantitative trait locus * Corresponding author; e-mail: [email protected]. Am. Nat. 2002. Vol. 159, pp. S51–S60. 䉷 2002 by The University of Chicago. 0003-0147/2002/15903S-0005$15.00. All rights reserved.

(QTL) analysis, however, requires the production of large populations of known pedigree and the analysis of tens if not hundreds of genetic markers for each individual. Thus, it is important to initiate QTL experiments only after traits known to influence reproductive isolation have been identified. One potential area of focus is on species groups that have undergone adaptive radiations. Often these groups already have had traits identified that have likely contributed to extensive species diversification (e.g., Darwin’s finches [Lack 1943; Grant 1986], Caribbean Annolis lizards [Jackman et al. 1997], African cichlid fishes [Fryer and Iles 1972]). In addition, adaptive radiations offer a rich opportunity to link studies ranging from adaptation and ecology, to morphology, physiology, and behavior with genetic analyses (Schluter 2000). Among flowering plants, the evolution of floral characters has long been considered a major avenue for achieving reproductive isolation and generating species diversity (Grant 1949). Floral characters may effect reproductive isolation either through differential pollination due to pollinator behavior or through differential pollen transfer (Grant 1949). The importance of floral evolution in speciation may be widespread because a recent phylogenetic analysis of the flowering plants found a significant correlation between pollination mode and diversification (Dodd et al. 1999). Across multiple independent lineages, the evolution of floral nectar spurs is highly correlated with significant increases in species numbers (Hodges and Arnold 1995; Hodges 1997a) and therefore, the evolution of floral nectar spurs has been proposed as a key innovation for these groups (Hodges and Arnold 1995; Hodges 1997a). Nectar spurs are tubular outgrowths of floral parts with nectar produced at their tips. In the process of probing these tubes to collect nectar, animals pollinate the plants. Nectar spurs could directly promote reproductive isolation if simple changes in morphology and color reduce or enhance visitation or the transfer of pollen by different types of pollinators (Hodges and Arnold 1995; Hodges 1997a, 1997b). In particular, evidence suggests that the columbine genus, Aquilegia, has undergone a recent and extremely

S52 The American Naturalist rapid radiation of taxa at or very near the time that floral nectar spurs evolved (Hodges and Arnold 1995; Hodges 1997a). Although all species exhibit the same floral bauplan, the genus is remarkable for wide variation in floral morphology and color associated with different pollinators (Grant 1952). Species of Aquilegia also occur in diverse habitats, including warm-temperate forests, alpine zones, and desert springs (Munz 1946). Despite the striking diversity in floral morphology and ecology, species in the genus are largely interfertile (Prazmo 1965; Taylor 1967), and therefore, genetic analysis of the traits differentiating species is possible. Furthermore, the interfertility among species suggests that the evolution of prezygotic reproductive isolation may be a prominent feature of speciation in this genus. Differences in floral morphology among species of Aquilegia have been directly implicated in promoting reproductive isolation. Grant (1952), in a classic article, suggested that differences in floral morphology were the primary factors in maintaining Aquilegia formosa and Aquilegia pubescens as separate species. These two species are differentiated by the color of the nectar spurs and petal blades, the length of the spurs and blades, and the orientation of the flowers. Grant postulated that visitation by different pollinators (hummingbirds to A. formosa and hawkmoths to A. pubescens) was influenced by the floral differences, causing assortative mating and the maintenance of these two species. Though this view was challenged (Chase and Raven 1975), we have recently shown (Fulton and Hodges 1999) that the major visitors to these two species do differentially visit these species based on floral characters. We found that these visitors discriminated between A. formosa and A. pubescens both in natural stands and at artificial arrays of flowers (Fulton and Hodges 1999). Furthermore, by experimentally manipulating flowers, we have shown that both flower orientation and nectar spur length affect reproductive isolation between these two species. Here, we expand on our earlier work to test whether flower color may also affect pollinator visitation to A. formosa and A. pubescens. Because the floral traits differentiating these species have been directly implicated as affecting reproductive isolation, we have initiated an analysis of their genetic basis. Such an analysis through quantitative trait locus (QTL) mapping would allow us to determine how complicated, genetically, a shift from one floral form to another could be (e.g., Bradshaw et al. 1995, 1998). We first examine two of these traits, flower orientation and flower color, in detail and show how careful consideration of phenotypes and their measurement will improve our ability to accurately quantify QTL. We then present a phenotypic analysis of an F2 population and a separate QTL analysis for these same floral traits.

Methods Pollinator Choice Experiments To test whether hawkmoths discriminate between Aquilegia formosa and Aquilegia pubescens flowers solely on the basis of flower orientation, we conducted an array experiment with three treatments: A. pubescens (natural upright), A. formosa (natural pendent), and A. formosa (manipulated upright). These three treatments were alternated in a hexagonal array of five flowers per treatment. Four arrays were observed on each of seven nights between August 17 and 27, 1998, for 1–1.5 h each night between 1900 and 2030 hours, resulting in a total of 32 h of arrayobservations. Each array was placed near A. pubescens flowers and observed by a different person, who recorded the number of hawkmoth visits to each flower type. Genetics of Floral Characters We measured five floral characters that distinguish A. formosa and A. pubescens: the length and color of petal spurs and petal blades and the angle of flower orientation at anthesis (fig. 1; Grant 1952; Chase and Raven 1975; Hodges and Arnold 1994). Calipers were used to make length measurements (fig. 1). A protractor fitted with a plumb line was used to measure the angle of flowers relative to gravity (0⬚ for pendent flowers, 180⬚ for completely vertical flowers). Occasionally, pendent flowers were oriented toward the stem of the plant instead of the usual orientation of away from the stem (see fig. 1). In these instances, we assigned negative orientation values. Because flower characters in Aquilegia can change during the ontogeny of the flower (see below and Kappert 1944), we restricted our measurements to flowers just beginning anthesis.

Figure 1: Floral characters measured in this study. Left, line drawing of a flower of an F2 plant from a cross between Aquilegia formosa and Aquilegia pubescens. The orientation of the flower is measured as the angle of the flower axis relative to gravity. Right, line drawing of a single petal; the four measurements made are indicated.

Genetics of Floral Traits in Aquilegia S53 To determine the phenotypic distributions of the parental species, we made floral measurements on plants along the Bishop Creek drainage, Inyo County, California. Aquilegia formosa plants were measured at low elevations (1,950–3,000 m) and A. pubescens plants were measured at high elevations (3,340–3,900 m). Narrow hybrid zones are found at mid-elevations (3,100–3,300 m) along this drainage (Hodges and Arnold 1994). For each plant, the orientation of a single flower just beginning anthesis was measured in the field. The flower was collected and brought to the White Mountain Research Station in Bishop, California, where we made morphological measurements during the evening. On the following day, we made spectral measurements to quantify color using natural sunlight. To assess the genetic basis of the floral characters, we first produced F1 seeds between A. formosa and A. pubescens. We made these crosses in the field along the Bishop Creek drainage using low-elevation (2,350 m) A. formosa plants as the maternal plants and crossing them with pollen from high-elevation (3,900 m) A. pubescens plants. The resulting seeds were germinated and grown in the greenhouse at the University of California, Santa Barbara (UCSB). To avoid inbreeding, we crossed two F1 plants that had different parents to produce the F2 seed. These F2 seeds were germinated and grown at UCSB. To reduce potential intraplant variation, we restricted our measurements to the first flowers that opened on each plant. We measured all floral characters (fig. 1) on the first flowers that opened, and we attempted to measure at least five flowers per plant. On subsequent flowers, we only measured flower orientation. We used the mean of the measurements on each plant as its phenotypic value. For color measurements, we collected single petals and kept them on ice. Color was measured on the same day, usually within 1–2 h of collection. We used a standard microscope lamp as a light source. We quantified the color of spurs and blades first by calculating standardized reflectance spectra. We used a fiber optics spectrometer (Ocean Optics S2000, Dunedin, Fla.) to record the amount of light reflected (400–700 nm) from petal blades and spurs. We collected light through a fiber optics cable set at an angle of 45⬚ relative to the flower surface and 1 cm from the flower. For each flower, we also recorded the intensity of light striking the flowers by recording the reflectance from a white standard (Labsphere, North Sutton, N.H.). At each wavelength, we divided the amount of light reflected from the flower by the amount of incident light in order to calculate a standardized reflectance spectrum. We used these standardized reflectance spectra to calculate aspects of color using the segmental classification method of Endler (1990). Briefly, this classification uses the proportion of total brightness

occurring in four equally spaced (by l) segments (400–475 [B], 475–550 [G], 550–625 [Y], and 625–700 [R] nm). The segmental values are used to calculate chroma as C p 冑[(R ⫺ G)2 ⫹ (Y ⫺ B)2] and hue as arcsin ([Y ⫺ B]/C). Total brightness is calculated as the proportion of total incident light that is reflected. We measured temporal changes in flower orientation by following buds early in development through anthesis until the petals abscised from the flower. We measured two to three buds on each of 10 F2 plants and measured their orientation approximately every 12 h. We noted the time when anthesis began for each flower. We then compared the temporal changes in flower orientation among plants by standardizing all measurements relative to the time of anthesis. QTL Mapping We used a second F2 population derived in the same manner as described above to search for QTL for the five floral traits differentiating A. formosa and A. pubescens. In order to produce linkage maps, we used amplified fragment length polymorphisms (AFLP; Vos et al. 1995) to generate markers. Extraction of DNA was accomplished using Qiagen DNA extraction kits. Extracted DNA (250 ng) was digested with EcoR I and Mse I, ligated with adaptors matching the EcoR I site (E) and the Mse I site (M), and preamplified with primers matching the adaptor sequences plus a single base-pair extension (E-A and M-C). We then performed specific amplifications using primers with three base-pair extensions. Specific primers that matched the Eadaptors were end-labeled with either IRD-700 or IRD800 and analyzed on a LiCor 4200 DNA sequencer. Because of the dominant nature of most AFLP bands, we searched for markers that were in coupling phase for one of the parental species, that is, present in both grandparents of one species, absent in both grandparents of the alternate species, and present in both F1 parents. Bands with these patterns would then segregate in the F2 population and mark A. formosa or A. pubescens chromosomal segments. We used Mapmaker 3.0b to construct the linkage maps. Because of the dominant nature of most AFLP markers, we divided markers into two coupling phase groups, depending on which species possessed the marker. Thus, we constructed linkage maps for both A. formosa markers and A. pubescens markers. We used Map Manager QTX (version b11; Manly and Olson 1999) to identify potential QTL. We first used interval mapping to detect potential QTL for each trait. We used permutation tests (1,000 replicates; Churchill and Doerge 1994) to determine a genome-wide significance levels for each trait and then tested each linkage group separately for evidence of a QTL. We

S54 The American Naturalist Floral Characters Aquilegia formosa spurs had low reflectance at short wavelengths and high reflectance at longer wavelengths (shifting at about 550–625 nm; fig. 3A). In contrast, A. pubescens spurs had relatively high reflectance at shorter wavelengths and a more gradual transition to even higher reflectance at longer wavelengths (fig. 3A). Similar patterns were found for petal blades, except that the shift to high reflectance in A. formosa occurs between 500–525 nm (fig. 3B). We found significant differences between A. formosa (N p 253) and A. pubescens (N p 198) for measurements of chroma, brightness, and hue for both petal spurs (t p 45.3, t p ⫺30.3, t p ⫺10.3, respectively, all P ! .0001) and petal blades (t p 72.3, t p ⫺40.3, t p ⫺17.0, respectively, all P ! .0001). Measurements of chroma were nonoverlapping for petal blades and nearly so for petal spurs (table 1). Measurements of brightness overlapped Figure 2: Number of visits by hawkmoths to flowers with different characters in artificial arrays. Flowers of Aquilegia formosa were either unmanipulated (pendent) or tethered to be upright. Flowers of Aquilegia pubescens were unmanipulated (upright).

first classified linkage groups as being significant (a p .05) and highly significant (a p .001) for a QTL (Lander and Kruglyak 1995). After all linkage groups for one species had been analyzed for a character, we used markers linked to the QTL with the most significant effect as a cofactor to retest the linkage group with the next most significant association. Again, we performed permutation tests (1,000 replicates) to determine significance levels. If the second putative QTL was determined to be significant, then the third most significant linkage group was tested using markers linked to the first two QTL as cofactors. This strategy was repeated for each additional linkage group until no significant association was detected.

Results Pollinator Discrimination Making Aquilegia formosa flowers upright increased hawkmoth visitation, compared to their natural pendent state (x 2 p 11.3, df p 1, P ! .001; fig. 2), but upright Aquilegia pubescens flowers received far more visits than upright A. formosa flowers (x 2 p 57.4, df p 1, P K .001; fig. 2). Because nectar spur length does not affect hawkmoth visitation (Fulton and Hodges 1999), these data indicate that other floral characters, such as color or odor, must influence hawkmoth discrimination between upright A. formosa and A. pubescens (fig. 2).

Figure 3: Reflectance spectra for flowers of Aquilegia formosa and Aquilegia pubescens. Reflectance spectra were calculated as described in the text for petal spurs (A) and petal blades (B). The means (heavy lines) and SDs (light lines) for 10 flowers are shown.

Genetics of Floral Traits in Aquilegia S55 Table 1: Summary of components of color distinguishing Aquilegia formosa and Aquilegia pubescens A. formosa mean (SD, range)

A. pubescens mean (SD, range)

Blade: Chroma .43 (.04, .35–.63) .17 (.03, .08–.30) Brightness .25 (.05, .12–.39) .48 (.07, .28–.69) Hue 56.4 (8.6, 21.8–84.2) 68.5 (5.7, 53.6–84.4) Spur: Chroma .52 (.08, .27–.74) .19 (.07, .08–.48) Brightness .16 (.04, .04–.30) .41 (.12, .12–.63) Hue 22.2 (7.8, 8.2–65.2) 36.2 (19.6, 1.0–84.6)

DSD 7.4 3.8 1.7 4.4 3.1 1.0

Note: Standardized reflectance spectra were used to calculate chroma, brightness, and hue for petal blades and petal spurs. The difference between mean values of the species in units of standard deviation (DSD) are given.

somewhat, while measurements of hue overlapped to a much greater extent (table 1). In terms of phenotypic standard deviations (DSD), mean values between the two species differed greatest for values of chroma, then brightness, then hue (table 1). Thus, measurements of chroma distinguished the two species to the greatest extent, and for simplicity, we restrict our analysis of color in the remainder of this article to values of chroma. The orientation of flowers changed substantially during development for all plants (e.g., fig. 4). Early in development, buds had an upright orientation and then became progressively more pendent. Subsequently, the flowers became progressively more upright. After flowering, fruit development begins and the fruit is fully upright. For five plants, measurements were made at early enough stages to quantify the minimal flower orientation by fitting second-order polynomial regressions (e.g., fig. 4). Using these regressions, the minimal angles for these five plants were estimated (range 24⬚–66⬚). The minimal angles were also estimated to occur between 36 and 83 h prior to anthesis. Thus, differences among F2 plants in flower orientation at anthesis are due to both the time that anthesis occurs during the developmental sequence and the degree that buds become pendent before turning upright (e.g., fig. 4). Because flower orientation influences reproductive isolation at the time of anthesis, we restrict our measurements to this time point in the remainder of our analyses. We measured one flower on each of 253 A. formosa plants and 198 A. pubescens plants for the five floral characters. All floral characters differed significantly between the two species (t p ⫺49.0, ⫺67.3, ⫺60.6, 72.3, and 45.3 for spur length, blade length, orientation, blade chroma, and spur chroma, respectively, all P ! .0001). Phenotypic measurements for blade length and blade chroma did not overlap, and the mean values differed by 6.8 and 7.4 SD, respectively (table 2; fig. 5). Spur length and flower orientation measurements were nearly nonoverlapping, and

the means differed by 5.7 and 4.7 SD, respectively (table 2; fig. 5). Identifying the presence and estimating the effect of a QTL depends on detecting phenotypic differences between classes of individuals distinguished by their genotypes at specific loci. Thus, reducing phenotypic variation due to nongenetic factors increases the accuracy of a QTL analysis. We therefore sought to reduce within-plant variation (due to environmental or measurement error) by using the mean of multiple measures on each plant. We then used ANOVA to estimate the repeatability of each trait (among-plant variance/total variance). This value is the maximal value for broad-sense heritability among F2 plants if only a single measure per plant is made. These estimates are maximal estimates of broad-sense heritability because our analysis cannot separate variance due to the special environment of each individual from variance due to genotype. We made a cross between two F1 plants that produced 273 seeds, of which 194 germinated and 153 flowered. We restricted our analysis to plants with a minimum of three

Figure 4: The angle of flower orientation over time for two different F2 plants. Time is relative to the beginning of anthesis with values less than 0 during bud development and values greater than 0 during flowering. The measurements for each plant were fitted to a second-order polynomial function (solid circles: y p 80.5 ⫹ 0.42x ⫹ 0.003x 2, r p 0.85, F p 64.5, df p 2, 50, P ! .0001; open squares: y p 39.4 ⫹ 0.37x ⫹ 0.005x 2, r p 0.81, F p 40.5, df p 2, 44, P ! .0001). The two plants differ both in the minimal angle of orientation and the time, relative to anthesis, when the minimum is reached.

S56 The American Naturalist Table 2: Summary of floral characters differentiating Aquilegia formosa and Aquilegia pubescens and their F2 hybrids Character

A. formosa mean (SD)

A. pubescens mean (SD)

DSD

F2 mean (SD)

Spur length (mm) Blade length (mm) Flower orientation Spur chroma Blade chroma

19.6 (2.4) 3.8 (.7) 7.1 (11.5) .52 (.08) .43 (.04)

35.1 (4.2) 12.6 (1.9) 110.1 (22.7) .19 (.07) .17 (.03)

5.7 6.8 4.7 4.4 7.4

27.2 (3.0) 7.9 (1.2) 61.4 (32.6) .45 (.11) .34 (.08)

Repeatability, F .72, .68, .53, .69, .83,

F134, 550 p 14.4 F134, 550 p 12.8 F134, 1,149 p 11.8 F134, 548 p 12.9 F134, 550 p 26.7

Note: The species were measured in the field and F2 plants were measured at the University of California, Santa Barbara. The mean and standard deviation (SD) are given for each character in each population. The difference between mean values of the species in units of standard deviation (DSD) are given. Repeatability (SA2 /[SA2 ⫹ S 2 ]) for each trait was calculated in the F2 population using ANOVA; all P ! .0001 . Values for spur and blade chroma for the two species are repeated from table 1.

flowers measured for each trait (N p 135). On average, we measured 5.1 (range 3–9) flowers on each plant for spur length, blade length, spur chroma, and blade chroma and 9.5 (range 3–13) flowers per plant for flower orientation. For all characters, there was significant variation among plants, and estimates of repeatability ranged from 0.53 to 0.83 (table 2). Though measured in different environments, the mean phenotypic values for F2 plants were between the mean values measured for A. formosa and A. pubescens (table 2; fig. 5). The means of trait values for spur length, blade length, and orientation were nearly exactly intermediate between the means of the parental species (fig. 5). In contrast, the distributions of spur and blade chroma were skewed toward A. formosa–like values (fig. 5). The floral characters were not inherited completely independently, as evidenced by significant correlation coefficients among all trait values. The absolute value of the product-moment correlation coefficients average 0.47 (range 0.3–0.63; table 3). QTL Mapping We scored 36 AFLP markers for 248 individuals in a separate F2 population. Of the 36 AFLP markers we scored, 26 were dominant for A. pubescens and eight were dominant for A. formosa. Two markers (28-218 and 15-144) were found to be codominant with alleles for the two species, differing by two base pairs in length in each case. We named our markers as follows: the first number refers to the E-primer used (specific bases AAC [1], AAG [2], ACA [3], ACC [4], ACG [5], ACT [6], AGC [7], or AGG [8]); the second number refers to the M-primer used (specific bases CAA [1], CAC [2], CAG [3], CAT [4], CTA [5], CTC [6], CTG [7], or CTT [8]); these numbers are followed by the approximate number of base pairs of the amplification product. Thus, marker 41-466 refers to the 466 bp product amplified with E-ACC and M-CAA. Of

the A. pubescens markers, 22 were grouped into four linkage groups (A–D; fig. 6). Of the A. formosa markers, six were grouped into three linkage groups (E–G; fig. 6). The two codominant markers were used in the construction of both maps. Marker 28-218 mapped to both linkage groups D and E, while marker 15-144 mapped to linkage group F in the A. formosa map and was unlinked to any A. pubescens markers. We removed 10 A. pubescens markers from further analysis because they were closely linked (!5 cM) to other loci. We made floral measurements on 210 of the 248 individuals that we scored for AFLP markers. On average, we measured 5.2 flowers (range 1–12) per plant for spur length, blade length, spur chroma, and blade chroma and 9.0 flowers per plant (range 1–15) for flower orientation at anthesis. We emphasize that because of the limited number of markers used in this analysis, only a small portion of the genome is likely to be covered. This is illustrated by the fact that all species of Aquilegia have seven chromosome pairs (Prazmo 1965; Taylor 1967), but we only found three and four linkage groups for the two sets of markers. Furthermore, we do not know if the linkage groups we did find correspond to different chromosomes. On the A. formosa linkage groups, we found significant QTL for all floral traits tested (fig. 6). On the A. pubescens linkage groups, we found significant QTL for flower orientation, blade length, and spur length (fig. 6). Discussion Reproductive isolation between Aquilegia formosa and Aquilegia pubescens is clearly affected by floral characters through their effects on pollinator behavior and pollen transfer (Fulton and Hodges 1999; this article). Pollinators (hummingbirds, hawkmoths, and bees) differentially visit the flowers of these species both in natural populations and at arrays of flowers (Fulton and Hodges 1999). Hawkmoths discriminate against pendent but not short-spurred

Genetics of Floral Traits in Aquilegia S57

Figure 5: Frequency distributions of spur length (A), blade length (B), flower orientation (C), spur chroma (D), and blade chroma (E) for Aquilegia formosa (open bars) and Aquilegia pubescens (hatched bars) measured under field conditions and an F2 population (solid bars) measured at University of California, Santa Barbara.

flowers (Fulton and Hodges 1999). Shorter spur length, however, significantly reduces pollen removal by hawkmoths and, therefore, also influences reproductive isolation between these species (Fulton and Hodges 1999). Floral characters in addition to flower orientation must influence hawkmoth discrimination because hawkmoths preferentially visited naturally upright A. pubescens rather than upright A. formosa flowers (fig. 2). Because hawkmoths do not discriminate based on spur length (Fulton and Hodges 1999), characters such as flower color or odor must account for this differential visitation. We have not performed specific experiments to determine which floral characters hummingbirds or bees use to discriminate between A. formosa and A. pubescens. However, other studies suggest that spur shape likely plays an important role for hummingbird visitation. Grant and Temeles (1992) found that hummingbirds (Selasphorus rufus) feed with optimal efficiency from floral tubes with lengths

similar to the mean spur length for A. formosa (19.6 mm; fig. 5) and that floral tubes beyond 25 mm in length caused an exponential increase in the time hummingbirds took to extract nectar. Therefore, hummingbirds may discriminate against A. pubescens flowers due to their long spurs (Grant and Temeles 1992). The longer petal blades of A. pubescens may also obstruct hummingbird access to nectar and reduce their visitation. Because floral traits contribute to reproductive isolation between A. formosa and A. pubescens, we investigated the genetic architecture of these traits through QTL analysis. Our phenotypic analyses suggest that there is a strong genetic component to each of the key floral traits. It is important to note that the parental species were measured in the field and the F2 plants were measured at UCSB. However, parental species grown under the same conditions of the F2 population over 3 yr do not differ markedly from plants in the field (S. A. Hodges, personal obser-

S58 The American Naturalist Table 3: Product-moment correlation coefficients between mean floral measurements of F2 individuals derived from a cross between Aquilegia formosa and Aquilegia pubescens Spur length Blade length Flower orientation Blade chroma Spur chroma

.47 .59 ⫺.56 ⫺.30

(.32, .59) (.47, .69) (⫺.43, ⫺.66) (⫺.14, ⫺.45)

Blade length

Flower orientation

Blade chroma

.49 (.35, .61) ⫺.46 (⫺.31, ⫺.58) ⫺.33 (⫺.17, ⫺.48)

⫺.54 (⫺.41, ⫺.65) ⫺.36 (⫺.20, ⫺.50)

.63 (.52, .72)

Note: All correlations are significant at P ! .001 . Each correlation is followed by upper and lower 95% confidence intervals.

vation). As expected, for a large genetic component to phenotype, the F2 distributions for spur length, blade length, and flower orientation are largely intermediate between the parental distributions (fig. 5). The F2 distributions for spur and blade chroma extensively overlapped with the values for A. formosa, suggesting that there may be dominance for the A. formosa phenotype. Most important, estimates of repeatability were moderate to relatively high, suggesting a genetic component for each of these traits. A common question for evolutionary studies using QTL mapping is whether traits are controlled by a few genes of large effect or many genes of small effect, and therefore, it is critical to understand whether a particular analysis has the statistical power to distinguish between these possibilities. Though the size of the experimental population is critical for the power of a QTL analysis (Beavis 1994; Lynch and Walsh 1998), the heritability of a trait among individuals in the mapping population also influences the ability to detect QTL. This is due to the fact that the power to detect a particular QTL is based on its affect on the phenotypic rather than the genetic variation of the mapping population. Clearly, to maximize the chances of detecting QTL, it is important to reduce variation due to nongenetic factors. One way to accomplish this is to use the mean of multiple independent measurements as the phenotypic value for each individual. Multiple measures will improve the heritability of a trait by a factor of N/(1 ⫹ r [N ⫺ 1]), where N is the number of measurements and r is the repeatability (Haley and Andersson 1997). For example, in our analysis, flower orientation had a repeatability of only 0.53, and using the mean of nine measurements rather than a single measurement per individual will increase the heritability by a factor of 1.72. Put another way, our QTL analysis of 210 individuals has the same power to detect a QTL for flower orientation as a mapping population of 361 individuals where a single measure per individual is made. Using this strategy, we were able to detect a QTL for flower orientation (fig. 6). For many organisms it may be difficult or impractical to produce large segregating populations and relatively easy to make multiple measurements. In these cases, the power to detect QTL can be increased

significantly, particularly for traits with low repeatability, by making multiple measurements per individual. The floral traits differentiating A. formosa and A. pubescens are integrated to some extent as shown by the significant correlations among traits in the F2 population. Linkage, pleiotropy, or common environmental effects must cause these correlations. In our QTL analysis, we found that either pleiotropy or linkage must explain some of this integration. All but one QTL mapped close to a QTL for another floral trait (fig. 6). Similarly, QTL affecting multiple floral traits has recently been found in Arabidopsis thaliana (Juenger et al. 2000). Because we cannot determine the exact position of QTL, either linkage or pleiotropy can explain these results. Integration of characters may help explain the limited introgression of floral characters beyond the narrow hybrid zones between these species (Hodges and Arnold 1994). Because selection likely acts against the introgression of each character we examined, tight linkage would delay introgression and pleiotropy would limit its extent compared to predictions based on each character in isolation. The prospects for future QTL mapping of traits influencing reproductive isolation between A. formosa and A. pubescens are quite good. In addition to floral traits, differences in habitat also likely influence reproductive isolation between these species (Chase and Raven 1975; Hodges and Arnold 1994). We have shown here how understanding the biology of specific traits and how accurate measures of them will improve our ability to detect QTL. In the future, Aquilegia species are very promising not only for identifying QTL but also for understanding their molecular basis and how natural selection acts on them. Because Aquilegia is self-compatible, recombinant inbred lines can be constructed for more detailed genetic analysis. Because most species are relatively easy to cross (Prazmo 1965; Taylor 1967), it should be possible to create nearisogenic lines in order to determine how specific QTL affect phenotype and plant fitness. Furthermore, natural hybrid zones exist between species (Grant 1952; Chase and Raven 1975; Hodges and Arnold 1994), and therefore, it may be possible to infer, using analyses of cline shapes, how natural selection acts on specific QTL. Using studies such as these, it should be possible to dissect the genetic

Genetics of Floral Traits in Aquilegia S59

Figure 6: Linkage maps and most likely positions of QTL for spur length (S), blade length (B), flower orientation (O), spur chroma (CS), and blade chroma (CB). AFLP markers were used to construct linkage maps that are grouped depending on which species was dominant for the coupling phase marker Aquilegia pubescens (A–D) or Aquilegia formosa (E–G). Marker names are indicated on the right in A. pubescens linkage groups and on the left for A. formosa linkage groups. A single codominant marker, 28-218, occurs in both groups and is indicated by a dashed line. The most likely location for each significant QTL is indicated. ∗ p P ! .05; ∗∗ p P ! .001. Scale bar p 10 cM.

basis and evolutionary consequences of traits that have influenced the adaptive radiation within Aquilegia. In the future, similar studies, conducted on a variety of species, will allow us to determine if there are general patterns to the evolution and genetic architecture of reproductive isolation and speciation.

Acknowledgments We thank S. Collie and B. Counterman for their collection of AFLP data and D. Bush for many helpful discussions. We thank S. Via and two anonymous reviewers for comments that greatly improved the manuscript. This study was funded by National Science Foundation grant DEB9726272 and a White Mountain Research Station Faculty Research grant to S.A.H.

Literature Cited Beavis, W. D. 1994. The power and deceit of QTL experiments: lessons from comparative QTL studies. Pages 250–266 in Proceedings of the Annual Corn and Sorghum Research Conference. American Seed Trade Association, Washington, D.C. Bradshaw, H. D., S. M. Wilbert, K. G. Otto, and D. W. Schemske. 1995. Genetic mapping of floral traits associated with reproductive isolation in monkey flowers (Mimulus). Nature 376:762–765. Bradshaw, H. D., 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. Chase, V. C., and P. H. Raven. 1975. Evolutionary and

S60 The American Naturalist ecological relationships between Aquilegia formosa and A. pubescens (Ranunculaceae): two perennial plants. Evolution 29:474–486. Churchill, S. E., and R. W. Doerge. 1994. Empirical threshold values for quantitative trait mapping. Genetics 138: 963–971. Coyne, J. A. 1995. Speciation in monkeyflowers. Nature 376:726–727. Dodd, M. E., J. Silvertown, and M. W. Chase. 1999. Phylogenetic analysis of trait evolution and species diversity variation among angiosperm families. Evolution 53: 732–744. Endler, J. A. 1990. On the measurement and classification of color in studies of animal color patterns. Biological Journal of the Linnean Society 41:315–352. Fryer, G., and T. D. Iles. 1972. The cichlid fishes of the great lakes of Africa. Oliver & Boyd, Edinburgh. Fulton, M., and S. A. Hodges. 1999. Floral isolation between Aquilegia formosa and A. pubescens. Proceedings of the Royal Society of London B, Biological Sciences 266:2247–2252. Grant, P. R. 1986. Ecology and evolution of Darwin’s finches. Princeton University Press, Princeton, N.J. Grant, V. 1949. Pollination systems as isolating mechanisms in angiosperms. Evolution 3:82–97. ———. 1952. Isolation and hybridization between Aquilegia formosa and A. pubescens. Aliso 2:341–360. Grant, V., and E. J. Temeles. 1992. Foraging ability of rufous hummingbirds on hummingbird flowers and hawkmoth flowers. Proceedings of the National Academy of Sciences of the USA 89:9400–9404. Haley, C. S., and L. Andersson. 1997. Linkage mapping of quantitative trait loci in plants and animals. Pages 49–71 in P. H. Dear, ed. Genome mapping: a practical approach. Oxford University Press, Oxford. Hodges, S. A. 1997a. Floral nectar spurs and diversification. International Journal of Plant Sciences 158(suppl.): S81–S88. ———. 1997b. Rapid radiation due to a key innovation in columbines (Ranunculaceae: Aquilegia). Pages 391–405 in T. J. Givnish and K. J. Sytsma, eds. Molecular evolution and adaptive radiation. Cambridge University Press, Cambridge. Hodges, S. A., and M. L. Arnold. 1994. Floral and ecological isolation between Aquilegia formosa and Aquilegia pubescens. Proceedings of the National Academy of Sciences of the USA 91:2493–2496.

———. 1995. Spurring plant diversification: are floral nectar spurs a key innovation? Proceedings of the Royal Society of London B, Biological Sciences 262:343–348. Jackman, T., J. B. Losos, A. Larson, and K. deQeiroz. 1997. Phylogenetic studies of convergent adaptive radiations in Caribbean Anolis lizards. Pages 535–557 in T. J. Givnish and K. J. Sytsma, eds. Molecular evolution and adaptive radiation. Cambridge University Press, Cambridge. Juenger, T., M. Purugganan, and T. F. C. Mackay. 2000. Quantitative trait loci for floral morphology in Arabidopsis thaliana. Genetics 156:1379–1392. Kappert, H. 1944. Untersuchungen uber Plasmonwinkungen bei Aquilegia (Gynodioecie-Herterosis-Gestalt des Sporus). Flora Allgemeine Botanische Zeitung. Neue Folge Bd. 37, Der Ganzen Reiche 137 Bd. Lack, D. 1943. Darwin’s finches. Cambridge University Press, Cambridge. Lander, E., and L. Kruglyak. 1995. Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nature Genetics 11:241–247. Lynch, M., and B. Walsh. 1998. Genetics and analysis of quantitative traits. Sinauer, Sunderland, Mass. Manly, K. F., and J. M. Olson. 1999. Overview of QTL mapping software and introduction to Map Manager QT. Mammalian Genome 10:327–334. Munz, P. A. 1946. Aquilegia: the cultivated and wild columbines. Gentes Herbarum 7:1–150. Prazmo, W. 1965. Cytogenetic studies on the genus Aquilegia. IV. Fertility relationships among the Aquilegia species. Acta Societatis Botanicorum Poloniae 34: 667–685. Rieseberg, L. H., B. Sinervo, C. R. Linder, M. C. Ungerer, and D. M. Arias. 1996. Role of gene interactions in hybrid speciation: evidence from ancient and experimental hybrids. Science (Washington, D.C.) 272: 741–745. Schluter, D. 2000. The ecology of adaptive radiation. Oxford Series in Ecology and Evolution. Oxford University Press, Oxford. Taylor, R. J. 1967. Interspecific hybridization and its evolutionary significance in the genus Aquilegia. Brittonia 19:374–390. Vos, P., R. Hogers, M. Bleeker, M. Reijans, T. V. D. Lee, M. Hornes, et al. 1995. AFLP: a new technique for DNA fingerprinting. Nucleic Acids Research 23:4407–4414.