Journal of Heredity Advance Access originally published online on August 31, 2005
Journal of Heredity 2005 96(5):550-556; doi:10.1093/jhered/esi098
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The Genetic Basis of Naturally Occurring Pollen Color Dimorphisms in Nigella degenii (Ranunculaceae)
From the Department of Ecology, Section of Plant Ecology and Systematics, Sölvegatan 37, Lund University, SE-22362 Lund, Sweden
Address correspondence to Stefan Andersson at the address above, or e-mail: Stefan.Andersson{at}ekol.lu.se.
 |
Abstract |
|---|
 Top Abstract Materials and MethodsResultsDiscussionReferences |
|---|
Nigella degenii ssp. barbro and ssp. jenny differ from related taxa in being dimorphic for pollen color, with some plants having dark pollen and others light pollen. In this study we performed experimental crosses to determine whether the difference in pollen color is governed by few or many loci and whether the two subspecies utilize the same gene to control pollen color. Patterns of segregation in crosses between morphs show that dark pollen is dominant over light pollen and that a single major gene is responsible for most of the variation in pollen color. Consequently it should be relatively easy for pollen color dimorphisms to establish and spread in these subspecies. Aberrant segregation ratios were attributed to genetic factors that reduced the expression of the allele conferring dark pollen or processes that sorted between color morphs during seed development. Crosses between dark pollen plants from different subspecies showed signs of complementation in the F2 generation, but the frequency of the light morph was too low to support a model involving complementary action of recessive alleles at two separate loci. Based on this and other observations, we hypothesize that the pollen color difference is controlled by the same major locus in the two subspecies.
Flower color polymorphism—the co-occurrence of two or more distinct flower color phenotypes within the same population—is a conspicuous feature of many plants, particularly in species whose flowers typically contain blue, red, or purple anthocyanin pigments (Warren and Mackenzie 2001). Such polymorphisms may be selectively neutral or a consequence of balancing selection forces involving not only pollinators (Brown and Clegg 1984; Mogford 1974), but also pleiotropic relationships between the expression of floral anthocyanins and related compounds that influence photoprotection, stress tolerance, disease defense, and herbivory resistance (Clegg and Durbin 2000; Koes et al. 1994; Levin and Brack 1995; Schemske and Bierzychudek 2001; Warren and Mackenzie 2001). Selection on floral color polymorphisms is most efficient when differences in flower color are controlled by one, or a few, major genes, when novel color morphs are specified by dominant alleles, and when a species can achieve the same polymorphism by different genetic mechanisms (Haldane 1924; Lande 1983; Macnair 1976).
While much attention has focused on corolla color polymorphisms, only a few studies have considered color polymorphisms involving sexual organs such as pistils, stamens, and pollen (Rafinski 1979; Wolfe 2001). One exception is Strid (1970), who documented a conspicuous pollen color dimorphism in two geographically and morphologically distinct subspecies of Nigella degenii Vier., ssp. barbro Strid and ssp. jenny Strid. This dimorphism is expressed at the sporophytic level, with some individuals producing pale or dark yellow (henceforth "light") pollen and others violet ("dark") pollen. The proportion of plants with dark pollen shows extensive variation among populations, ranging from 0 to 0.93 in ssp. barbro and from 0.02 to 0.71 in ssp. jenny (Jorgensen TH and Andersson S, unpublished observations). Related species have light pollen (Strid 1970), indicating that the dimorphism, or more strictly the dark pollen type, is derived within N. degenii.
Pollen color differences in other plants have been found to involve a single-gene system with complete dominance for dark pollen (Gerats et al. 1985; Mehlenbacher and Smith 2002; Qiao et al. 1993; Wakelin et al. 2003), a pattern that also applies to many petal color dimorphisms [for references, see Levin and Brack (1995)]. However, it is still uncertain whether the single-gene model holds for N. degenii (Strid 1970) and whether the two subspecies of N. degenii utilize the same gene to control pollen color. In the present study we obtained data from extensive crossing experiments to address these questions. Following the detection of single-gene inheritance in most of the crosses, we also examined whether aberrant segregation ratios could be attributed to genetic background effects or processes that sorted between color morphs before flowering.
|
Materials and Methods |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Study System
Nigella degenii (Ranunculaceae) is an annual, diploid (2n = 12) species that occupies disturbed habitats (e.g., abandoned fields, seashores, open patches in phrygana vegetation) in the Cyclades (Greece), where four geographically restricted subspecies have been recognized, including ssp. barbro (northwest Cyclades) and ssp. jenny (island of Syros). Flowers of Nigella have a conspicuous perianth, differentiated into an outer whorl of five white, petaloid sepals and an inner whorl of eight, stalked nectaries. The androecium comprises a variable number of stamens, which shed their pollen as the filaments curve outward during the male phase. The gynoecium consists of up to five completely united follicles (each with a separate style) and develops into a capsule with numerous seeds. The self-compatible, insect-pollinated flowers are protandrous and herkogamous, reducing within-flower selfing (Strid 1969, 1970). N. degenii ssp. barbro and ssp. jenny differ in nectary color, sepal shape, fruit morphology, and various features related to the pollen color dimorphisms characterizing these subspecies (Strid 1970).
Plant Material
This study involves plants from one population of N. degenii ssp. barbro (Mykonos, about 2.5 km north-northwest of the town) and one population of N. degenii ssp. jenny (Syros, about 300 m south of Kini), both sampled in 1993 and maintained in an insect-free greenhouse for several generations by random outcrossing within populations. Both populations were dimorphic for pollen color, the proportion of plants with dark pollen being approximately 14% in the initial population samples (Andersson S, unpublished observations).
Four generations of enforced self-pollination were used to establish five (partly) inbred lines for each color morph and population. Each line was derived from a single maternal plant in a base population and had reached (apparent) fixation for the intended color morph in the last generation (based on an examination of more than 20 progeny in each line). Lines were denoted as ML, MD, SL, or SD to indicate population origin (M for Mykonos, S for Syros) and pollen color (L for light pollen, D for dark pollen), and given a unique identification number within each category (e.g., ML1, SD5).
Segregation Analyses
In 2001–2003, we established a series of F1, F2, and backcross (BC) progenies from a number of reciprocal intermorph crosses, each pair representing a distinct combination of inbred lines from the same or different base populations. These "inbred-line crosses" represented four parent line combinations in the Mykonos population, four parent line combinations in the Syros population, and six interpopulation combinations. One cross (ML3 x MD1) was performed twice, using different parent plants from the same inbred lines. Flowers used as pollen recipients were emasculated during the male phase to avoid contamination with self-pollen and attempts were made to perform each cross reciprocally. Most crosses yielded progeny with dark pollen (indicating dominance for dark pollen), but a few families were dimorphic. In these cases we restricted the subsequent self- or cross-pollinations to F1 plants with dark pollen to maximize the probability that the resulting F2 or BC progenies segregated for pollen color. In the autumn of 2003 we planted F2 or BC seeds in separate pots and scored the resulting progeny for pollen color. Crosses that yielded too few plants (due to low seed set, germination ability, or survival rate) were excluded, which reduced the number of progenies for which both reciprocal families were available. Seventy-two percent of the F2 and BC seeds (4428 of 6075 sown seeds) resulted in plants that could be scored for pollen color.
As a second step, we extended the genetic analyses to intermorph crosses involving plants with no prior inbreeding history. These "outbred-plant crosses" were based on pairs of plants sampled from the same base population before the inbreeding phase (see above). These crosses represented 11 parent combinations in the Mykonos population and 9 parent combinations in the Syros population. Each cross was performed reciprocally and replicated up to six times (using different pairs of flowers), and the resulting seeds were sown into plug trays to provide data on pollen color. Given the low frequency (approximately 14%) of the dark morph in the base populations, most dark pollen parents were expected to be heterozygous at loci controlling pollen color, that is, to produce segregating progenies in crosses with the light morph. Seed set, quantified as the proportion of ovules that developed into seeds, was determined for each replicate cross (fruit) to assess the potential for differential selection on color morphs during seed maturation (see below). Seventy-eight percent of the sown seeds (2574 of 3315 seeds) resulted in plants that could be scored for pollen color.
We employed log-likelihood G tests for goodness-of-fit (Sokal and Rohlf 1995) to compare observed morph ratios with segregation ratios expected under the single-gene model. Individual G tests were carried out at a critical probability of 
' = /k, where is the experiment-wise error rate (0.05, 0.01, or 0.001) and k is the number of replicate progenies (tests) for the type of cross considered (Bonferroni adjustment). A replicated goodness-of-fit test was carried out for each type of progeny in the inbred-line crosses (for categories, see Table 1) and for each population in the outbred-plant crosses to evaluate the overall fit of the single-gene model (quantified as Gpooled) and to assess the heterogeneity in the proportion of morphs among replicate progenies (quantified as GH).
|
Complementation Tests
In addition to the segregation analyses, we performed a complementation analysis by crossing dark pollen plants from different populations (derived from the same inbred lines as were used in the intermorph crosses) and then searching for plants with light pollen in the F2 generation. Under the assumption of Mendelian segregation at two unlinked loci, 1/16 of the F2 progeny were expected to be homozygous recessive at both loci, that is, to express the light phenotype. Such "complementation" would indicate that the dark pollen type is conditioned by different dominant genes in the two populations. Comparisons of observed and predicted frequencies were carried out with a G test.
Selection Analyses
There is great potential for selective processes to alter progeny segregation ratios when only a few ovules develop into germinable seeds or when many seedlings die before flowering. To evaluate the role of selection during seed maturation, we assessed the relationship between the seed set of each flower in the outbred-plant experiment and the corresponding segregation ratio in the progeny generation (expressed as a fraction of the dark morph). Progeny segregation ratios were subjected to analysis of covariance (ANCOVA) involving: (1) "parent combination," a group variable that provided a control for genetic background effects, (2) seed set (covariate), and (3) the interaction between parent combination and seed set. A similar approach was used to examine the potential for selection between germination and flowering, as determined by the proportion of seeds that developed into flowering plants. In this case it was also possible to analyze data from the inbred-line crosses (using different BC or F2 categories as groups; see Table 1).
The residuals from the ANCOVAs were approximately normally distributed, hence no transformation was necessary. Initial analyses of data from outbred-plant crosses revealed a nonsignificant difference between the two source populations (F = 0.09, P = .773), so this factor was excluded from the final analysis.
|
Results |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Patterns of Segregation in Inbred-Line Crosses
The 15 intermorph crosses produced F1 progenies in one or both directions (as indicated by the presence of reciprocal or unidirectional F2 families in Table 1). Based on data pooled across families, 95% of the F1 progeny (106 of 111 plants) produced dark pollen. The five light pollen plants represented two dimorphic F1 families from crosses within the Syros population: SD10 (male) x SL1 (female) and SD9 (male) x SL2 (female). Backcrosses to the dark pollen morph always yielded progeny with dark pollen (Table 1), confirming the dominant nature of this phenotype.
Morph frequencies in segregating F2 or BC progenies generally conformed to the 3:1 or 1:1 ratio predicted under the single-gene model, regardless of whether the two parent plants represented the same or different populations (Table 1). There was a close fit to single-gene predictions after pooling data across progenies within the same F2 or BC category (Gpooled < 1, P > .05), the only exception being the F2 Syros group (Gpooled = 29.3, P < .01). The latter result was almost entirely due to the extreme excess of light pollen plants in the family SD10 (male) x SL1 (female) (1 dark versus 74 lights), as shown by the drastic reduction in the G value after excluding this F2 family from the analysis (Gpooled = 0.2, P > .05). This procedure also reduced the heterogeneity among replicate families within the F2 Syros group (GH = 3.3, P > .05 versus GH = 169.4, P < .001 for the full dataset). Remaining backcross and F2 categories showed moderate among-family variation (F2 Mykonos: GH = 16.1, P > .05; F2 between populations: GH = 39.0, P < .001; backcrosses: GH = 16.7, P < .05), with three significant deviations from single-gene predictions, each involving a slight excess of the light morph (Table 1).
Incidents of deviant segregation were unilateral: a significant deviation from the predicted Mendelian ratio in one family was always associated with normal Mendelian segregation in the reciprocal family (when available, Table 1). The replicated cross in the Mykonos population (ML3 x MD1) showed a weak, though significant, unilateral bias in one replicate but not in the other.
Patterns of Segregation in Outbred-Plant Crosses
Three intermorph crosses, all involving Mykonos plants, produced progeny with only dark pollen, presumably because the dark parent was homozygous dominant at the inferred major effect locus. Morph frequencies in dimorphic progenies showed a close agreement with the 1:1 ratio predicted under the single-gene model (Table 2), especially when data where pooled over families (Mykonos: Gpooled = 0.68, P > .05; Syros: Gpooled = 2.16, P > .05). Nevertheless, there was significant heterogeneity in the proportion of morphs among replicate families (Mykonos: GH = 61.88, P < .001; Syros: GH = 39.18, P < .001), with three families showing (unilateral) distortion of morph frequency, one involving a significant excess of the dark type and two involving a significant excess of the light type (Table 2).
|
Complementation Tests
Crosses between dark pollen plants from different populations yielded two plants with light pollen in the F2 generation: 1 of 70 plants in the cross MD5 x SD5 and 1 of 120 plants in the cross MD2 x SD9. A third cross produced plants with only dark pollen (MD4 x SD5, 90 plants). The overall frequency of the light morph (2 of 280 plants) was significantly lower than the frequency expected under a model involving complementary action of recessive alleles at two separate (unlinked) loci (17.5 plants; Gpooled = 24.67, P < .001).
Selection Analyses
Pollinated flowers varied greatly in the proportion of ovules that matured into seeds (range 47–100% for outbred-plant crosses) and there was extensive among-family variation in the proportion of sown seeds that resulted in flowering plants (range 40–92% for inbred-line crosses and 21–100% for outbred-plant crosses). The segregation ratio in the progeny generation showed no relationship with seed set (F = 0.49, P > .05; ANCOVA), but was affected by parent combination (F = 2.19, P < .05) and by the parent combination–seed set interaction (F = 2.43, P < .01). As for the significant interaction term, our data indicate large changes in the ranking of parent combinations between fruits with low and high seed set (crossing regression lines) and a tendency for the most extreme segregation ratios (within a given parent combination) to represent the low seed set category (Figure 1). The progeny segregation ratio was independent of survival rate and showed no significant interaction between survival rate and progeny type (inbred-line crosses) or parent combination (outbred-plant crosses) (F < 1.70, P > .19 in all cases; ANCOVA).
|
|
Discussion |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
Knowledge of the genetic basis of phenotypic variation—for example, the number of genes involved and the dominance relationships among the alleles at these loci—is fundamental when inferring the potential for selection to shape evolutionary change in floral characters, including color polymorphisms. To date, only a few investigations have explored the genetic control of pollen color, and these are restricted to plant species whose pollen coats normally contain dark pigments and where the light morph results from mutations that prevent the synthesis of dark pigments (Gerats et al. 1985; Mehlenbacher and Smith 2002; Qiao et al. 1993; Wakelin et al. 2003). In this study we examined the genetic basis of naturally occurring pollen color dimorphisms in N. degenii, a species in which the dark pollen morph represents the derived rather than the ancestral condition.
Genetic data from N. degenii indicate a Mendelian basis for most of the variation in pollen color: patterns of segregation in crosses between the different color morphs generally conformed to segregation of two alleles at a single major locus, regardless of whether the parents represented inbred or outbred genotypes from the same or from different populations (subspecies). Both populations showed dominance for the dark pollen type; consequently there should be great potential for local selection pressures to act on a newly risen mutant with dark pollen. Given the existence of a major locus controlling pollen color and the observed dominance relationship, it should be relatively easy for pollen color dimorphisms to establish and spread in N. degenii. The evolutionary lability of pollen color is also manifested by the simple genetic control of pollen color in Petunia hybrida (Gerats et al. 1985), Helianthus annuus (Qiao et al. 1993), Corylus avellana (Mehlenbacher and Smith 2002), and Eschscholzia californica (Wakelin et al. 2003), even though no persistent pollen color dimorphisms have evolved in these species.
In a previous study of N. degenii, we used field data, common garden experiments, and pollen competition experiments to explore the potential role of pollinators and postpollination processes in exerting selection on pollen color dimorphisms (Jorgensen TH and Andersson S, unpublished data). Our results show that pollinators can discriminate between plants with different pollen color and that the dark and light pollen morphs sometimes differ in fertilizing or siring success; however, the relative morph fitness is strongly determined by the type of pollinator (bumblebees versus honeybees), pollination treatment (one- versus two-donor pollinations), and the particular combination of plants used as pollen and seed parents in the pollen competition experiments. Pollen color has also been shown to influence pollination success in Campanula americana: halictid bees were more likely to discriminate against flowers without pollen when they foraged in arrays of flowers with tan-colored pollen than in arrays of flowers with purple pollen (Lau and Galloway 2004). Although these results suggest a role for selection in shaping evolutionary change in pollen color, there are no obvious between-species differences in habitat specificity or pollination biology that could explain why N. degenii possess pollen color dimorphisms while other Nigella species do not.
The few incidents of deviant segregation were distributed uniformly over the two study populations and generally involved an excess of the light pollen morph, the most extreme being one of the F2 progenies from the Syros crosses (1 dark versus 74 light). This and other, less extreme segregation distortions were always unilateral: the detection of a non-Mendelian morph ratio was always associated with a normal Mendelian ratio in the reciprocal family. In view of these findings, and the close agreement with single-gene predictions in a majority of the crosses, we hypothesize that some segregants failed to produce dark pollen pigments despite being heterozygous or homozygous for the allele conferring dark pollen, and that the magnitude of this penetrance-reducing effect was influenced by some nonnuclear (cytoplasmic) factor carried by some of the plants in the parent generation. However, it is difficult to explain why the replicated cross (ML3 x MD1) showed a unilateral bias in only one of the two sets of reciprocal crosses; since each inbred line was derived from a single female, one would expect all plants within a given line to have the same cytoplasm and to express the same progeny morph ratio bias in replicate crosses with another line. Although more detailed genetic data are needed to confirm the role of cytonuclear interactions, our observations clearly differ from the more strictly dominant expression of pollen color loci in other plants (Mehlenbacher and Smith 2002; Qiao et al. 1993; Wakelin et al. 2003; but see Gerats et al. 1985). The detection of such heterogeneity accentuates the advantage of performing multiple crosses—involving several unrelated parents—before any broad generalizations are made regarding the heritable basis of pollen color polymorphisms.
Some of the aberrant segregation was probably mediated by processes that sorted between morphs during the embryonic stages: progenies from outbred-plant crosses showed greater deviations from Mendelian ratios when there was an enhanced potential for postzygotic selection (low seed set) than when little or no selection was possible (high seed set). Interestingly, low seed set was not associated with a consistent bias in morph frequency: progenies from some parent combinations had an excess of the dark morph, whereas others had an excess of the light morph (Figure 1). These patterns can either be attributed to selection on the pollen color loci per se—mediated through pleiotropic associations between pollen color and early zygote vigor—or selection on alleles at linked loci. In this context we also note that previous common garden experiments have demonstrated morph-specific differences in vegetative survivorship, at least for plants of N. degenii ssp. barbro grown under stressful conditions (Jorgensen TH and Andersson S, unpublished data). Indirect selection during embryonic or vegetative stages extends the range of ecological factors that might influence the spread and maintenance of pollen color dimorphisms, a hypothesis that also applies to flower color polymorphisms in general (Clegg and Durbin 2000; Koes et al. 1994; Levin and Brack 1995; Schemske and Bierzychudek 2001; Warren and Mackenzie 2001).
Populations of N. degenii ssp. barbro and ssp. jenny are geographically isolated and have diverged for a wide variety of vegetative and reproductive characters (Andersson 1997; Strid 1970). For example, the pollen color dimorphism in ssp. jenny is expressed by both the pollen and anthers, whereas plants of ssp. barbro have dark anthers irrespective of pollen color. These observations not only suggest that N. degenii ssp. barbro and ssp. jenny have been genetically isolated for substantial periods of time, but also raise the possibility that dimorphic populations have evolved independently in the two subspecies. One testable prediction of this "multiple-origin scenario" is that the production of dark pollen pigments is controlled by different dominant genes in the two subspecies.
Our complementation crosses showed some evidence for segregation in the F2 generation, but the frequency of the light morph was too low (less than 1%) to support a model involving complementary action of recessive alleles at two independent loci. Based on this result and the excess of the light pollen morph in some within-population crosses, we cannot exclude incomplete penetrance of dominant alleles—at the same locus—as the principal cause of the few light pollen plants that appeared in the complementation crosses. Consequently our results provide little support for rejecting the hypothesis that the two subspecies of N. degenii utilize the same major locus to control differences in pollen color. Further evaluation of the multiple-origin hypothesis will benefit from the identification of pigments responsible for light and dark pollen color in the two subspecies.
|
Acknowledgments |
|---|
This work was supported by a fellowship from the Danish Natural Science Research Council (to T.H.J.) and grants from the Swedish Natural Science Research Council (to S.A.), the Julie von Müllens Foundation (to T.H.J.), the Svend G. Fiedlers Foundation (to T.H.J.), and the Frimodt-Heineke Foundation (to T.H.J.). Thanks to Olympia Tassopoulou and colleagues at the Department of Environment, Prefecture of Syros, for encouraging the project and to D. Richardson for comments on the manuscript.
|
Footnotes |
|---|
Corresponding Editor: James Hamrick
Received December 3, 2004
Accepted May 26, 2005
|
References |
|---|
|
TopAbstractMaterials and MethodsResultsDiscussionReferences |
|---|
-
Andersson S, 1997. Genetic constraints on phenotypic evolution in Nigella (Ranunculaceae). Biol J Linn Soc 62:519–532.[CrossRef]
Brown BA and Clegg MT, 1984. Influence of flower color polymorphism on genetic transmission in a natural population of the common morning glory, Ipomoea purpurea. Evolution 38:796–803.[CrossRef]
Clegg MT and Durbin ML, 2000. Flower color variation: a model for the experimental study of evolution. Proc Natl Acad Sci USA 97:7016–7023.
Gerats AGM, Vrijlandt E, Wallroth M, and Schram AW, 1985. The influence of the genes An1, An2, and An4 on the activity of the enzyme UPD-glucose: flavonoid 3-O-glucosyltransferase in flowers of Petunia hybrida. Biochem Genet 23:591–598.[Medline]
Haldane JBS, 1924. A mathematical theory of natural and artificial selection. Trans Camb Philos Soc 23:19–41.
Koes RE, Quattrocchio F, and Mol JNM, 1994. The flavonoid biosynthetic pathway in plants: function and evolution. BioEssays 16:123–131.
Lande R, 1983. The response to selection on major and minor mutations affecting a metrical trait. Heredity 50:47–65.
Lau JA and Galloway LF, 2004. Effects of low-efficiency pollinators on plant fitness and floral trait evolution in Campanula americana (Campanulaceae). Oecologia 141:577–583.[CrossRef][Web of Science][Medline]
Levin D and Brack ET, 1995. Natural selection against white petals in Phlox. Evolution 49:1017–1022.
Macnair MR, 1976. The genetics of copper tolerance in Mimulus guttatus (Scrophulariaceae) (PhD dissertation). Liverpool: University of Liverpool.
Mehlenbacher SA and Smith DC, 2002. Inheritance of pollen color in hazelnut. Euphytica 127:303–307.[CrossRef]
Mogford DJ, 1974. Flower color polymorphism in Cirsium palustre. 2. Pollination. Heredity 33:257–263.
Qiao CG, Wang QY, Wang YL, and Sun CZ, 1993. Inheritance of white pollen in sunflower (Helianthus annuus). Indian J Agric Sci 63:518–519.
Rafinski JN, 1979. Geographic variability of flower color in Crocus scepusiensis (Iridaceae). Plant Syst Evol 131:107–125.[CrossRef]
Schemske DW and Bierzychudek P, 2001. Evolution of flower color in the desert annual Linanthus parryae: Wright revisited. Evolution 55:1269–1282.[CrossRef][Web of Science][Medline]
Sokal RR and Rohlf FJ, 1995. Biometry. New York: Freeman.
Strid A, 1969. Evolutionary trends in the breeding system of Nigella (Ranunculaceae). Bot Notiser 122:380–396.
Strid A, 1970. Studies in the Aegean flora. XVI. Biosystematics of the Nigella arvensis complex. Opera Bot 28:1–169.
Wakelin AM, Lister CE, and Conner AJ, 2003. Inheritance and biochemistry of pollen pigmentation in California poppy (Eschscholzia californica Cham.). Int J Plant Sci 164:867–875.[CrossRef]
Warren J and Mackenzie S, 2001. Why are all color combinations not equally represented as flower-color polymorphisms? New Phytol 151:237–241.[CrossRef]
Wolfe LM, 2001. Associations among multiple floral polymorphisms in Linum pubescens (Linaceae), a heterostylous plant. Int J Plant Sci 162:335–342.[CrossRef]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||