Breeding system variation, genetics and evolution in the Turneraceae

Authors


Author for correspondence: Joel Shore Tel: +1 416 7362100 ext. 33492 Fax: +1 416 7365698 Email: shore@yorku.ca

Summary

We review the genetics and evolution of breeding systems in the Turneraceae. Distyly occurs in seven of 10 genera and 81% of species. The remaining species are homostylous. Polyploid evolution has been significant in Turnera. Approximately 60% of species are polyploid ranging from diploid through decaploid. No relationship between breeding system and polyploidy is evident. The genetics of distyly involves a one-locus two-allele system (S and s). Evidence from crosses with homostylous species and mutants is consistent with the possibility that a ‘Primula-type’ supergene underlies distyly but does not prove this to be the case. A polygalacturonase, and an α-dioxygenase specific to the transmitting tissue of short-styled plants both exhibit morph-limited expression in concert with predictions from an evolutionary model. The function of the proteins in distyly, if any, is unknown. We have begun constructing a fine-scale genetic map of Turnera. Two genetic markers lie within 0.2 cm of the distyly locus. This should provide a starting point for positional cloning of the distyly locus and reveal the genetic architecture and molecular basis of distyly.

Introduction

Distyly has provided an important model system in plant genetics and breeding system evolution since Darwin's investigations (Darwin, 1877; Barrett, 1992; Ornduff, 1992). Hypotheses on the genetic architecture, evolution and breakdown of distyly have emerged from theoretical studies and empirical work on various distylous taxa, and in particular, from Primula species. In this paper we describe breeding system variation in the Turneraceae and explore some of these hypotheses. Specifically, we will provide the first modern comprehensive review of breeding systems in the Turneraceae, describing breeding system variation, its inheritance and evolution. We re-examine whether there is a relationship between breeding system and polyploidy. We review the genetic basis of distyly in Turnera, and evidence for and against the hypothesis that a gene complex or ‘supergene’ determines distyly in Turnera. Finally, we review our progress in determining the molecular genetic basis of distyly.

Turneraceae and the evolution of distyly

The Turneraceae are in the order Malpighiales (APG, 2003). Among the 40 or so families in the Malpighiales (APG, 2003), four (Erythroxylaceae, Hypericaceae, Linaceae and Turneraceae) possess distylous species. Molecular phylogenetic analyses indicate that the Turneraceae are, however, sister to the Passifloraceae, and together, both are sister to the Malesherbiaceae (Davis & Chase, 2004). Neither the Passifloraceae nor Malesherbiaceae possess any heterostylous species. Based upon this recent molecular phylogenetic evidence, distyly must have arisen independently at least once within the Turneraceae.

The Turneraceae comprise 10 genera and 226 species and subspecific taxa (Table 1) most widely distributed in the Neotropics which possesses approximately 170 species in four genera. The bulk of the species occur in just two genera: Piriqueta with 45 and Turnera with c. 128 species. Adenoa Arbo is a monotypic genus endemic to Cuba, Mathurina penduliflora Balf. f. is endemic to Rodrigues island, Erblichia has one species in America and four in Madagascar, and the remaining genera/species occur in Africa (Urban, 1883; Arbo, 1979, 1995b, 1997, 2000, 2004, 2005).

Table 1.  List of all 226 species/subspecific taxa of Turneraceae and their breeding system, chromosome numbers and ploidy levels (where known)
SpeciesBreeding systema,eChromosome numbersPloidy levelse
Turnera (135 taxa)
Series Turnera (29 taxa)
Subseries Turnera (21 taxa)
T. aureliiHom1SC2n = 402Segmental allooctaploid
T. campanifloraHom1SC2n = 303Allohexaploid
T. candidaHom1SC2n = 104Diploid
T. coerulea var. coeruleaDis1SI2n = 10, 204Diploid and autotetraploid
T. coerulea var. surinamensisDis1SI2n = 104Diploid
T. concinnaDis1SI2n = 102Diploid
T. cuneiformisHom1SC2n = 405Segmental allooctaploid
T. fernandeziiDis1SI2n = 401,6Autooctaploid
T. grandidentataDis1SI2n = 202Segmental allotetraploid
T. grandifloraDis1SI2n = 106Diploid
T. krapovickasiiDis1SI2n = 10, 206Diploid and autotetraploid
T. lucidaHom1   
T. occidentalisHom1SC2n = 301Hexaploid
T. oculataHom1   
T. orientalisHom1SC2n = 302Segmental allohexaploid
T. scabraDis1SI2n = 10, 206Diploid and autotetraploid
T. subulataDis1SI2n = 10, 207Diploid and autotetraploid
T. thomasiiHom1   
T. ulmifolia var. acutaHom1SC2n = 308Hexaploid
T. ulmifolia var. ulmifoliaHom1SC2n = 309Segmental allohexaploid
T. velutinaHom1SC2n = 3010Segmental allohexaploid
Subseries Umbilicatae (8 taxa)
T. arcuataDis1 2n = 208Autotetraploid
T. coriaceaDis1 2n = 4011Octoploid
T. hermannioidesDis1SI2n = 106Diploid
T. joeliiDis1SI2n = 101,6Diploid
T. leptospermaDis1   
T. purpurascensDis1   
T. simulansDis1   
T. stenophyllaDis1   
Series Annulares (4 taxa)
T. annularisDis12   
T. aromaticaDis12   
T. brevifloraDis12   
T. odorataDis12   
Series Anomalae (14 taxa)
T. amazonianaHom?1   
T. bahiensisDis1   
T. blanchetianaDis1   
T. cearensisDis1   
T. chrysocephalaDis1   
T. discorsDis1   
T. gardnerianaHom?1   
T. involucrataDist1   
T. kuhlmannianaDist1   
T. laciniataDis1   
T. reginaeDis1   
T. sanctaDis1   
T. stipularisDis1   
T. tapajoensisHom?1   
Series Capitatae (10 taxa)
T. albicansDis12   
T. capitataDis12   
T. dasystylaHom12   
T. hatschbachiiDis12   
T. maracasanaDis12   
T. marmorataDis12   
T. pernambucensisDis?12   
T. princepsDis12   
T. schomburgkianaDis12   
T. waltherioidesDis12   
Series Leiocarpae (50 taxa)
T. acaulisHom13   
T. argenteaDis14   
T. callosaDis13   
T. cipoensisDis15 2n = 1411Diploid
T. crulsiiHom16   
T. curassavicaDis13   
T. dasytricha?   
T. dichotomaDis13   
T. discolorDis17   
T. dolichostigmaDis18   
T. ellipticaDis13   
T. foliosaDis16   
T. genistoidesDis13   
T. goyazensisDis16   
T. guianensisDis13   
T. harleyiDis19   
T. hasslerianaDis18SI2n = 14, 286Diploid and autotetraploid
T. hilaireanaDis13   
T. huberiDis14   
T. incanaDis13   
T. lamiifoliaDis13 2n = 286Tetraploid
T. lanceolataDis13   
T. lineataDis?   
T. longifloraDis13   
T. luetzelburgiiDis20   
T. melanorhizaDis16   
T. melochia?   
T. melochioidesDis13SC2n = 146Diploid
T. nanaDis13   
T. nervosaDis18SI2n = 146Diploid
T. oblongifoliaDis13   
T. opiferaDis13SI2n = 706Decaploid
T. paruanaDis14   
T. pinifoliaHom13   
T. pohlianaDis13   
T. pranceiDis15   
T. pumilea var. pumileaHom18SC2n = 146Diploid
T. pumilea var. piauhyensisDis21,42SC2n = 1442Diploid
T. revolutaDis17   
T. riedelianaDis13   
T. sidoides ssp. sidoidesDis and Hom22SI2n = 286,23Autotetraploid
T. sidoides ssp. carneaDis and Hom22SI2n = 14, 28, 426,23Diploid and autopolyploid
T. sidoides ssp. holosericeaDis and Hom22SI2n = 28, 426,23Autotetraploid and autohexaploid
T. sidoides ssp. integrifoliaDis22SI2n = 14, 28, 42,566,23Diploid and autopolyploid
T. sidoides ssp. pinnatifidaDis22SI2n = 14, 28, 426,23Diploid and autopolyploid
T. stachydifoliaDis13   
T. subnudaDis16   
T. tenuicaulisDis16   
T. trigonaDis13 2n = 148Diploid
T. uleanaDis20   
Series Microphyllae (5 taxa)
T. asymmetricaDis43   
T. calyptrocarpaDis12   
T. collotrichaDis19   
T. diffusaDis12   
T. hebepetalaDis12   
Series Papilliferae (2 taxa)
T. caatinganaDis12   
T. chamaedrifoliaDis and Hom12SC2n = 266Diploid
Series Salicifoliae (12 taxa)
T. amapaensis?   
T. brasiliensisDis24   
T. claussenianaDis24   
T. glazioviiDis24   
T. hindsianaDis24   
T. ignotaDis43   
T. panamensisDis24   
T. rupestrisDis24   
T. serrataDis24   
T. steyermarkiiDis24   
T. venosaDis24   
T. weddellianaDis24SI2n = 146Diploid
Series Stenodictyae (9 taxa)
T. acutaDis24   
T. annectensDis24   
T. aurantiacaDis24   
T. benthamianaDis24   
T. castilloiDis24   
T. cicatricosaDis24   
T. longipesDis24   
T. macrophyllaDis and Hom24 2n = 146Diploid
T. urbaniiHom24   
Series uncertain (1 species)
T. rubrobracteataDis15   
Piriqueta (48 taxa)
P. abairanaDis25   
P. araguaianaDis26   
P. asperifoliaDis26   
P. assuruensisDis and Hom26   
P. aureaDis26   
P. breviseminataDis26   
P. caiapoensisDis26   
P. capensisDis and Hom27,28   
P. carneaDis26   
P. cistoides ssp. cistoidesHom26SC2n = 146Diploid
P. cistoides ssp. carolinianaDis26SI2n = 1411,29Diploid
P. constellataDis26   
P. corumbensisDis26   
P. cristobaliaeDis26   
P. dentataDis26   
P. densifloraDis26   
P. douradinhaDis26   
P. duarteanaDis26 2n = 146Diploid
P. emasensisDis30   
P. flammeaDis26   
P. grandifoliaDis26   
P. guianensisDis26   
P. hapalaHom26   
P. lourteigiaeDis26   
P. mesoamericanaDis26   
P. mexicanaHom26   
P. morongiiDis and Hom26SC2n = 146,31Diploid
P. mortoniiHom26   
P. nanuzaeDis26   
P. nitidaDis26   
P. ochroleucaDis26 2n = 1431Diploid
P. plicataDis26   
P. racemosaDis and Hom26SC2n = 146Diploid
P. revolutaDis?25   
P. roseaDis26 2n = 14 and 286Diploid and autotetraploid
P. saraeDis26   
P. scabridaDis26   
P. sidifolia var. sidifoliaDis26   
P. sidifolia var. multifloraDis26 2n = 1431Diploid
P. suborbicularisDis26SI2n = 286Tetraploid
P. subsessilisDis26   
P. sulfureaDis26   
P. tamberlikii ssp. rotundifoliaDis26 2n = 2826Tetraploid
P. tamberlikii ssp. tamberlikiiDis26   
P. taubatensisDis26SI2n = 426,31Segmental allohexaploid
P. undulataDis26   
P. venezuelanaHom?26   
P. viscosaHom26   
Adenoa (1 species)
Adenoa cubensisHom32   
Erblichia (5 species)
E. antsingyaeHom33   
E. bernerianaHom33   
E. integrifoliaHom33   
E. madagascariensisHom33   
E. odorataHom33   
Hyalocalyx (1 species)
H. setiferusDisb,27 and Hom34   
Loewia (3 species)
L. glutinosaDis35   
L. microphyllaHomc,36   
L. tanaënsisDis?b,37   
Mathurina (1 species)
M. pendulifloraHom38   
Stapfiella (7 taxa)
S. claoxyloides?   
S. lucida var. lucidaHom39   
S. lucida var. pubescens?   
S. muricata?   
S. ulugurica?   
S. usambaricaDis?37   
S. zambesiensisHom27   
Streptopetalum (6 taxa)
S. arenariumDisc,35   
S. graminifoliumDis?c,40   
S. hildebrandtiiDisc,13   
S. luteoglandulosumDis27   
S. serratumHomb,c,d,13   
S. witteiDisc,27,39   
Tricliceras (19 taxa)
T. auriculatum?b,27   
T. bivinianumDisb,37   
T. brevicaule var. brevicauleDisb,27   
T. brevicaule var. rosulatum?b,27   
T. elatumDisb,27   
T. glanduliferumDis and Homb,27,28   
T. hirsutumDisb,27   
T. lacerataDisb,c,28   
T. lanceolatumDisb,27   
T. lobatumHomb,13,34   
T. longepedunculatumDisb,c,13,37   
T. longepedunculatum var. eratense?   
T. mossambicenseDisb,c,27,28   
T. pilosumHomb,13,34   
T. prittwitziiDis?c,41   
T. schinzii var. schinzii?   
T. schinzii var. juttaeDisb,d,27,41   
T. tanacetifoliumDisb,27,28   
T. xylorhizum?   

Turnera is divided into nine series (Urban, 1883). The genus has a centre of high species diversity in the Brazilian states of Bahia, Goiás, and Minas Gerais, and a secondary centre in Paraguay (Arbo, 1987, 2004). Turnera species extend north into the southern USA and south into central Argentina. There are two Turnera species native to Africa (Turnera oculata Story and Turnera thomasii (Urb.) Story) and two species (Turnera ulmifolia L. and Turnera subulata Sm.) have been introduced into and have become weeds in various tropical regions of the world (Arbo, 2005).

Breeding systems

The taxonomic distribution of breeding systems in the Turneraceae has not been reviewed comprehensively since Urban's (1883) monograph, yet numerous species have been described over the past century. We review the distribution of breeding systems categorizing them into those that are distylous vs. homostylous. Distylous species show reciprocal herkogamy, and Turnera species that have been investigated also show pollen size and production dimorphisms, and are commonly strongly self-incompatible (Barrett, 1978; Barrett & Shore, 1985). There is somewhat of an asymmetry in the lengths to which incompatible pollen tubes grow (Tamari et al., 2001) as in many distylous taxa (Wedderburn & Richards, 1990; Dulberger, 1992; Barrett & Cruzan, 1994; Wong et al., 1994b). For the purpose of consistency with our previous work, we refer to species/populations exhibiting floral monomorphism as possessing a homostylous breeding system. Homostylous species commonly have anthers and stigmas in close proximity within a flower and are self-compatible, although some (e.g. Turnera campaniflora and Turnera velutina) show approach herkogamy (Barrett & Shore, 1987).

Seven of the 10 genera within the Turneraceae possess distylous species (Tables 1 and 2). The three genera lacking distylous species are either monotypic (Adenoa and Mathurina) or possess few species (Erblichia; Table 1). Hyalocalyx is a monotypic distylous genus. In Tricliceras and Streptopetalum most species are distylous, while in Loewia, two of three species are distylous (see later). Stapfiella has six species with very small flowers (2–4 mm long). The reproductive system has not been described and the available material is very scarce. Stapfiella lucida var. lucida may be homostylous and Stapfiella usambarica is likely to be distylous. We have categorized the breeding systems of Stapfiella with a degree of uncertainty (Table 1). Interestingly, the flowers of species of four African genera (Hyalocalyx, Loewia, Streptopetalum & Tricliceras) possess stamens of different lengths, mostly two shorter and three longer within a flower, for both homostylous and distylous species (Table 1; Urban, 1883; Arbo, 2006).

Table 2.  Summary of numbers of species/subspecific taxa with distyly, homostyly or both breeding systemsa in various genera of Turneraceae and in the series of the genus Turnera
Genus or SeriesDistylyHomostylybDistyly and homostyly
  • a

    We ignored the uncertainty in categorization of breeding systems in Table 1.

  • b

    Note that some species/populations listed as homostylous, show varying degrees of approach herkogamy.

Turnera
Series Turnera17120
Series Annulares400
Series Anomalae1130
Series Capitatae910
Series Leiocarpe4343
Series Microphyllae500
Series Papilliferae201
Series Salicifoliae1100
Series Stenodictyae711
Turnera (totals)109215
Piriqueta3864
Adenoa010
Erblichia040
Hyalocalyx100
Loewia210
Mathurina010
Stapfiella120
Streptopetalum530
Tricliceras1231
Totals1684010

Distyly is widespread in Turnera and Piriqueta. In Turnera, distyly is the predominant breeding system (possessed by c. 80% of species) but there is variation in the proportion of distylous species across the taxonomic series (Table 2). Series Turnera possesses the lowest percentage of distylous species (59%) while four series possess only distylous species (Table 2). Interestingly, five species of Turnera possess both distylous and homostylous populations.

In the genus Piriqueta with 48 taxa (species, subspecies and taxonomic varieties), 38 (79%) are distylous, six are homostylous and four species are composed of populations, some of which are distylous and the others homostylous (Tables 1 and 2). Piriqueta assuruensis Urb., Piriqueta racemosa (Jacq.) Sw. and Piriqueta capensis (Harvey) Urb. (the last being the only African species of the genus) all possess both distylous and homostylous populations (Arbo, 1995b). Piriqueta viscosa Griseb. is a small-flowered diploid homostylous species sister to Piriqueta morongii Rolfe, the latter of which possesses both distylous and homostylous populations (Arbo, 1995b; Truyens et al., 2005). Plants in distylous populations of P. morongii are self-fertile (J. S. Shore, pers. obs.). No crosses between or within these species have been undertaken so the nature and inheritance of homostyly is unknown. Piriqueta cistoides (L.) Griseb. ssp. cistoides is a diploid homostylous subspecies, while Piriqueta cistoides ssp. caroliniana is the distylous subspecies that has likely given rise to the homostyle (Ornduff, 1970). Interestingly, there is considerable morphological and molecular variation among distylous populations of Piriqueta cistoides ssp. caroliniana sensu lato, in Florida. Cruzan (2005) has explored patterns of hybridization among them.

The breeding systems of the other genera of Turneraceae have not been studied to any great extent, and we have classified their breeding systems based largely upon surveys of herbarium material and/or literature surveys. Further work will be required to characterize their breeding systems.

The distribution of breeding systems among species should provide insights into the origins of distyly and homostyly in the family. This might allow the recognition of species that are primitive or ‘primary homostyles’ (should they occur) as opposed to those that are derived or ‘secondarily’ homostylous (sensu Ernst, 1955). Recent molecular phylogenetic analyses (Mast et al., 2006) indicate that there are apparently no primary homostyles within Primula. A comprehensive phylogenetic analysis will be required to assess breeding system evolution in the Turneraceae. In an initial phylogenetic analysis of c. 40 species, Truyens et al. (2005) indicated that homostyly appears to have evolved independently at least three times in Turnera.

Polyploidy and breeding systems

In a study of the T. ulmifolia complex, Barrett & Shore (1987) reported that diploids and autotetraploids were distylous and self-incompatible while allohexaploids were homostylous and self-compatible. Barrett & Shore (1987) speculated that reduced inbreeding depression resulting from allohexaploidy might have been responsible for allowing the spread of selfing homostyles through hexaploid populations.

There is a tendency towards the evolution of homostyly in polyploid Primula species (Kelso, 1992; Richards, 2003; Guggisberg et al., 2006). Richards (2003) indicated that chromosome numbers of 29 homostyle species of Primula are known and that 16 of these species are polyploid. The polyploids range from tetraploid through tetrakaidecaploid. A majority of the 13 diploid homostyles were at one time thought to be primitively monomorphic (Richards, 2003), but it now seems that distyly is the ancestral breeding system in Primula and so homostyly must be derived in these diploids (Mast et al., 2006). By comparison, only five distylous species are polyploid (4% of species), outside of two sections of the genus (Parryi and Auricula) where all species are either tetraploid or hexaploid (Richards, 2003).

In a recent molecular phylogenetic analysis of species in Primula sect. Aleuritia subsect. Aleuritia using cpDNA, Guggisberg et al. (2006) demonstrated multiple origins of polyploidy coupled with the evolution of homostyly. In this subsection of Primula, only a single polyploid (tetraploid) is distylous, all others are homostylous. Most of the polyploids in subsection Aleuritia are believed to be allopolyploids. In Damnacanthus (Rubiaceae), there is a complete association between diploidy and distyly vs. tetraploidy and monomorphism. The monomorphic species are phenotypically long-styled but their pollen is like that of short-styled plants (Naiki & Nagamasu, 2004). By contrast, molecular phylogenetic analyses of Amsinckia (Boraginaceae) do not reveal any strong associations between polyploidy and breeding system (Schoen et al., 1997).

Here we re-examine the pattern reported in Barrett & Shore (1987) reviewing a broader range of species for which cytological and breeding system data are now available. Polyploid evolution has been very significant in Turnera as c. 60% of the species/populations studied have chromosome numbers in the tetraploid through decaploid range. Evidence indicates that c. 35% of the polyploids are autopolyploids (Solís Neffa & Fernández, 2002). The mechanism of polyploidization is likely to be through unreduced gamete formation (Fernández & Arbo, 1990).

Polyploid distylous species occur in both Turnera and Piriqueta ranging from tetraploid through decaploid, or tetraploid through hexaploid, respectively (Table 1). It is possible that the distylous polyploids are restricted to those having had autopolyploid origins. This certainly appears to be the case for species in series Turnera (x = 5) including Turnera subulata, Turnera scabra Millsp., Turnera krapovickasii Arbo, and Turnera coerulea DC. var. coerulea, all of which occur at both diploid and autotetraploid levels (Fernández, 1987; Shore, 1991a,b; Truyens et al., 2005) while distylous Turnera fernandezii is an auto-octaploid (Fernández, 1987; Arbo, 2005; Table 1). Five subspecies occur within the T. sidoides L. complex (series Leiocarpae, x = 7). Two of the subspecies possess only distylous populations, while the other three subspecies possess both distylous and homostylous populations (Table 1). Ploidy levels range from diploid through octaploid, and all the polyploids appear to be autopolyploids (Arbo, 1985; Fernández, 1987; Solís Neffa & Fernández, 2000, 2002).

Exceptions to this pattern of autopolyploid distylous species occur for tetraploid Turnera grandidentata (Urb.) Arbo, which is a segmental allotetraploid (Fernández, 1987; Fernández & Arbo, 1990, 1993b), and Turnera chamaedrifolia Cambess. which is a ‘diploid’ having a base chromosome number of x = 13, as opposed to x = 5 or x = 7 for all other Turnera species (Fernández, 1987). Cytogenetic and phylogenetic evidence suggests that the increased base chromosome number of T. chamaedrifolia is the result of polyploid evolution and subsequent aneuploid reduction (Solís Neffa & Fernández, 2000; Truyens et al., 2005). Homostylous collections of this latter species have also been made. Interestingly, the distylous specimens cultivated in Corrientes Argentina were somewhat self-compatible and set a few autogamous seeds although the ovaries possess c. 90 ovules (Arbo, 2000).

Overall, there appears to be no impediment to the occurrence of distyly in polyploids. Is there, however, a tendency for homostyly to evolve preferentially in polyploids? Within series Turnera there are 29 species and/or subspecific taxa (Arbo, 2005). Eight of these are homostylous polyploid species (six hexaploids and two octaploids). The polyploids all appear to be allopolyploids based upon examination of meiosis in these species, although Fernández & Arbo (2000b) have suggested that octaploids Turnera aurelii and Turnera cuneiformis, and hexaploids Turnera orientalis, T. ulmifolia var. ulmifolia and T. velutina may be segmental allopolyploids based upon studies of meiosis in hybrids (Fernández & Arbo, 1993a, 2000a,b; Fernández, 1997). Turnera candida Arbo is the only known self-compatible diploid homostylous species in this series (Fernández & Arbo, 1996). The nature of homostyly is unknown in this species since compatibility relationships (Fig. 1) and inheritance studies with related distylous species have not been carried out. A single long-styled diploid hybrid (pollen fertility 19.7%) was obtained in a cross with Turnera grandiflora (Urb.) Arbo. Crosses of T. candida with short-styled hybrids of T. grandiflora × Turnera coerulea var. coerulea were unsuccessful (Fernández & Arbo, 1996).

Figure 1.

Compatibility relationships among long-styled, short-styled and homostyled plants.

While data from series Turnera appear to support a propensity for the evolution of homostyly in allopolyploids/segmental allopolyploids, in fact, it is unclear whether the homostylous polyploids have had independent origins. Molecular phylogenetic analysis of internal transcribed spacer (ITS) sequence data have not provided sufficient resolution to determine whether homostyly in these eight polyploids was the result of a single or multiple origins (Truyens et al., 2005). The two homostylous octaploids T. aurelii and T. cuneiformis certainly appear to share a hexaploid homostylous progenitor, T. orientalis, based upon studies of meiosis in hybrids and molecular phylogenetic analysis (Fernández & Arbo, 2000a; Truyens et al., 2005). Thus, homostyly did not arise independently in these octaploids.

In other series of the genus Turnera investigated, homostyly occurs in diploid T. pumilea L. The nature of homostyly is unknown in this species (Truyens et al., 2005) which also possesses a distylous variety, T. pumilea var. piauhyensis Urb.

All the polyploids known in Piriqueta are distylous (Table 1). Piriqueta suborbicularis and Piriqueta tamberlikii var. rotundifolia are tetraploids, Piriqueta rosea has diploid and autotetraploid populations, and finally Piriqueta taubatensis appears to be a segmental allohexaploid (Fernández, 1987; Lavia & Fernández, 1993; Arbo, 1995b).

In contrast to the observations of Barrett & Shore (1987), at present, it appears to be premature to conclude that there is a causal relationship between polyploidy and the evolution of homostyly in the Turneraceae. The investigation of further species in a phylogenetic context will be required to address this question rigorously. It will be necessary to determine the progenitors of the allopolyploid species to understand the contributions of reticulate evolution to breeding systems of allopolyploids.

Inheritance of distyly: genetic architecture in Turnera

Distyly is inherited by what appears to be a single Mendelian locus with two alleles (Lewis & Jones, 1992). Short-styled (thrum) plants are commonly heterozygous, Ss, while long-styled (pin) plants are recessive homozygotes, ss. In autotetraploid Primula obconica, Dowrick (1956) showed that the same dominance relationships hold, however, tetrasomic inheritance occurs at the locus. We have studied the inheritance of distyly in T. scabra and T. subulata (formerly T. ulmifolia var. intermedia and T. ulmifolia var. elegans, respectively) at both the diploid and autotetraploid levels (Shore & Barrett, 1985). We found a common pattern of inheritance. Short-styled plants are Ss and long-styled plants are ss in diploids, and we have demonstrated tetrasomic inheritance in the autotetraploids.

Detailed studies, particularly by Ernst (Ernst, 1955), indicate that distyly in Primula is determined by a series of tightly linked loci comprising a supergene. Ernst's data suggest that three loci may be separable by recombination and others have postulated the existence of, and provided evidence for additional linked genes (Dowrick, 1956; Kurian & Richards, 1997). Population genetic models indicate that the distylous polymorphism is able to establish if the genes for various traits are linked forming a supergene (Charlesworth & Charlesworth, 1979). In contrast, the phenotypic models of Lloyd & Webb, 1992) allow a broader range of genetic bases including the possibility of major linked genes determining androecial and gynoecial characters, as well as modifiers that are not necessarily linked. The expression of modifiers may be morph-limited and coordinated with the activities of the supergene.

Because of its historical precedence, we begin by reviewing the evidence for a supergene in Turnera and subsequently consider the model of Lloyd & Webb (1992). Under perhaps the simplest supergene model, three tightly linked loci are responsible for distyly (Lewis & Jones, 1992). The G/g locus determines style length and its incompatibility, the P/p locus determines pollen size and its incompatibility, while the A/a locus determines anther height. Under this model, short-styled plants are heterozygous at all three loci and are genotypically GPA/gpa while long-styled plants are homozygous, gpa/gpa.

There are three sources of experimental data bearing upon the question of whether there is a supergene in Turnera. The first approach is to cross (reciprocally) distylous species with homostylous species to explore compatibility relationships between long-styled, short-styled and homostyled plants. We have carried out these experiments for a number of homostylous species in series Turnera examining seed set and pollen tube growth (Barrett & Shore, 1987; Tamari et al., 2001). If the homostylous species of Turnera have origins analogous to the recombinant long-homostyles of Primula, then we can predict a particular pattern of compatibility among pollen from various sources (Fig. 1). Our data show such a pattern for six homostylous species investigated. Interestingly, when homostyles reject pollen from long-styled plants, the incompatibility response appears similar to that observed when long-styled plants are selfed. That is, pollen tubes germinate, penetrate the stigma and grow into the style where they are inhibited. Pollen plugs are visible in the pollen tubes (Tamari et al., 2001). When short-styled plants reject pollen from homostyles, pollen germinates, but is inhibited in the stigma and no callose plugs are apparent. The incompatibility reaction of short-styled plants to pollen from homostyles resembles their rejection response to self pollen (Tamari et al., 2001).

If the homostyles in Turnera have arisen via recombination within a supergene controlling distyly (following the Primula model), there is a predicted pattern of dominance relationships that should occur in the F1 progeny from crosses between distylous and homostylous species. That is, homostyles should appear to be determined by an allele, SH, of the distyly locus and the dominance hierarchy, S > SH > s, should occur. This is, in fact, the pattern of progeny observed for three homostylous species investigated including, T. ulmifolia, T. velutina and T. orientalis in crosses with distylous T. subulata or T. scabra (Shore & Barrett, 1985). Unfortunately, we have had little opportunity to explore the other homostylous species of Turnera in this way. The occurrence of a single long-styled plant from a cross between homostylous T. candida and distylous T. grandiflora (Fernández & Arbo, 1996), suggests that the origins of T. candida may not be the result of recombination within a supergene (should it exist), but this warrants further investigation.

Perhaps the best evidence for a supergene would be the direct observation of recombinants at the ‘locus’. This would require extremely large sample sizes and flanking genetic markers to confirm that phenotypically novel progeny are the result of recombination and not mutation. While we have not detected directly any such recombinants, we have studied the inheritance of two homostylous mutants. The first mutant clearly did not arise as a result of meiotic recombination, but rather arose as a somatic mutant branch on an otherwise short-styled plant. The homostyle mutant is inherited as if determined by an allele of the distyly locus, and exhibits the dominance hierarchy above (Tamari et al., 2005). While not the result of meiotic recombination, the occurrence and inheritance of this mutant is consistent with the supergene hypothesis. The mutant also exhibits compatibility relationships and pollen size expected of a recombinant long-homostyle (Fig. 1).

We observed and studied a second homostyle ‘mutant’ in autotetraploid T. scabra. The long-homostyled plant was first observed as a seedling growing out of the pot of an adult plant. Studies of its inheritance also revealed that it is determined by an allele of the distyly locus showing the dominance hierarchy above. In this instance, tetrasomic inheritance occurs at the locus, and oddly, the original ‘mutant’ has the genotype SHSHss (Tamari et al., 2005). This homostyle also exhibits compatibility relationships and pollen size expected of a recombinant long homostyle (Fig. 1).

If a supergene occurs in Turnera, recombination frequencies within it might be extremely low. As another means of exploring whether a supergene might be present we have initiated a mutagenesis experiment. We generated short-styled plants of T. subulata that were genotypically SS. The S-allele is marked by alleles of two isozyme loci that lie on either side of the distyly locus (Athanasiou & Shore, 1997) so that we can confirm and follow the fate of mutated chromosomes. In a pilot study, we irradiated pollen of the SS short-styled plants and pollinated a long-styled plant homozygous for alternative alleles of the two linked isozyme loci. Out of approximately 1000 progeny, the vast majority of which were short-styled, we recovered two mutants. One is a long-homostyle and the other a long-styled plant. Unfortunately, both plants are female sterile, and we have been unable to transmit the putative mutant allele of distyly to progeny through the mutant's pollen. At present, we do not know the genetic basis of the mutant phenotypes. The occurrence of the mutants is certainly consistent with the hypothesis that a supergene determines distyly as we appear to have been able to independently mutate (or possibly delete the gene(s) or a portion of it, as X-rays commonly cause deletions) the putative G allele (which determines style length and its incompatibility) for the homostyle, and the putative ‘G’, ‘P’ and ‘A’ alleles for the long-styled mutant. An alternative explanation for these mutant phenotypes is that they are determined by new mutant alleles of a single gene that resides at the S-locus. At present we cannot distinguish between these possibilities.

In Primula, it has been suggested that it is difficult to obtain SS genotypes because a recessive lethal embedded in the distyly supergene may cause nonviability of progeny carrying this genotype (Kurian & Richards, 1997; Richards, 1997). Richards (2003) model for the evolution of distyly postulates the occurrence of a supergene which has embedded within it such a linked recessive lethal gene. In diploid Turnera spp. there is no evidence for such a linked recessive lethal. Data from selfing (or bud-selfing) short-styled plants has shown no departure from the expected three short-styled to one long-styled ratio, for plants of T. scabra and T. subulata (Shore & Barrett, 1985; Athanasiou & Shore, 1997). Furthermore, we have produced short-styled plants that are homozygous, SS, by exploiting a self-compatible short-styled plant of T. subulata. We have shown in test-crosses of these plants to long-styled plants, that only short-styled progeny are produced, confirming the homozygosity of the short-styled parents (J.S. Shore, unpublished).

The evolutionary model of Lloyd & Webb (1992) postulates the possible involvement of both linked genes (i.e. a supergene) as well as unlinked modifiers with morph-limited expression, in the genetic makeup of distyly. We reviewed (above) evidence that is consistent with, but certainly does not prove the existence of, a supergene in Turnera. Our work to discover morph-specific proteins, and their corresponding genes, has, however, provided the first evidence for genes with morph-limited expression (see later). We have demonstrated that both a polygalacturonase and an α-dioxygenase are expressed in the transmitting tissue of only short-styled plants (Athanasiou et al., 2003; Khosravi et al., 2004). Recently, McCubbin et al. (2006) provided evidence for morph-limited and/or differential expression of genes in Primula vulgaris.

Molecular genetic basis of distyly

One approach to discovering both the genetic architecture and molecular basis of distyly is to search directly for proteins specific to one or the other morph. This approach had been taken by Golynskaya et al. (1976), and Shivanna et al. (1981) for Primula, as well as for other distylous species, including Averrhoa carambola (Wong et al., 1994a) and Fagopyrum esculentum (Miljuš-Ðukićet al. 2004). McCubbin et al. (2006) have used subtractive hybridization to discover genes differentially expressed between the morphs of Primula vulgaris. They identified a number of different classes of genes involved potentially as downstream components of floral heteromorphism. None of the genes, however, appear to be at the S-locus (McCubbin et al., 2006).

Athanasiou & Shore (1997) identified protein expression differences between the morphs of Turnera subulata and Athanasiou et al. (2003) subsequently identified the proteins and their corresponding genes. One protein is a polygalacturonase specific to the transmitting tissue of the short-styled morph of species in series Turnera (Khosravi et al. 2003; Tamari & Shore, 2004). While linked to the distyly locus, the polygalacturonase gene is 4.6 cm distal to it (Athanasiou et al., 2003). Athanasiou et al. (2003) also showed that the pollen protein discovered by Athanasiou & Shore (1997) was also a polygalacturonase belonging to a clade (clade C) of pollen expressed plant polygalacturonases. In a survey of a additional species of series Turnera, Tamari & Shore (2004) were unable to confirm the short morph specificity of the pollen polygalacturonase.

Khosravi et al. (2004) discovered a 68 kDa protein specific to short styles. They showed, based upon sequence similarity, immunocytochemistry, and phylogenetic analysis, that the protein is an α-dioxygenase specific to the transmitting tissue of short styles of species in series Turnera. The gene encoding the α-dioxygenase is not closely linked to the distyly locus. α-Dioxygenases have only been recently discovered and they appear to play a role in signalling in response to plant pathogens (Sanz et al., 1998). The roles of both the polygalacturonase and α-dioxygenase in distyly, if any, remain to be determined (Tamari & Shore, 2006). We have continued to search for proteins distinguishing the morphs using two-dimensional gel electrophoresis and tandem mass spectrometry to identify the proteins. Two additional candidate proteins have been tentatively identified and others await identification and verification of morph-specificity (Khosravi et al., 2006).

Studies by Athanasiou et al. (2003) and Khosravi et al. (2004) clearly demonstrate that the S-allele of the distyly locus has the capacity to regulate the expression of two genes that do not reside at this locus. The proteins encoded by the style polygalacturonase gene and the α-dioxygenase gene both show morph-limited expression and are expressed only in the transmitting tissue of short-styled plants. These studies are in concert with predictions of the model of Lloyd & Webb (1992) which postulates that some of the genes involved in distyly might exhibit morph-limited expression. Unfortunately we do not yet know what role these two proteins play in distyly. This remains an important avenue for future research.

Genetic localization

Genetic localization is currently being used as a means to positionally clone the locus or loci determining distyly. Manfield et al. (2005) identified a random amplified polymorphic DNA (RAPD) marker linked to the S-locus (thrum allele) in P. vulgaris. They subsequently sequenced an 8.8 kb region corresponding to this marker, and used homostyles to further define the region encompassing the S-locus. A similar approach has been initiated in F. esculentum (Aii et al., 1998; Nagano et al., 2001a,b) where linked molecular markers and candidate BAC clones containing the S-locus have apparently been obtained.

We have recently initiated a mapped-based approach in Turnera (J. Labonne and J.S. Shore, unpublished). We did so by exploiting a mutant homostyle discovered by Tamari et al. (2005). The mutant allele was backcrossed into T. subulata and a backcross mapping population of approximately 700 progeny was generated. We used a range of molecular markers including isozymes, RAPD, intersimple sequence repeat (ISSR) and randomly amplified microsatellite polymorphism (RAMP) markers. We have now identified a number of closely linked markers, one of which is within approximately 0.2 cm of the distyly locus. We will continue mapping and construct a BAC library in an effort to clone the gene(s) at the distyly locus. This work should ultimately yield a clear understanding of the molecular genetic basis of distyly and a test of whether a supergene is involved.

Future prospects

While breeding systems in the Turneraceae are perhaps among the best studied for families possessing distylous species, considerable efforts will be required to resolve important outstanding questions. The origins of distyly in the family, including the number of origins, and the recognition of primitively homostylous or monomorphic species (should they occur) will necessitate that a broad phylogenetic analysis is undertaken. Coupled with this phylogeny will be the more basic need to characterize the breeding systems of a number of species, particularly in the African genera.

Polyploidy is rampant in Turnera. Continued chromosome number surveys and cytogenetic analyses within the Turneraceae will be required to explore the relationship between breeding system and polyploid evolution. The process of reticulate evolution will require that the progenitors of allopolyploid species are identified to fully understand breeding system evolution. More recent methods such as genomic in situ hybridization (GISH) are currently being used to explore the origins of some of the polyploids.

The inheritance of distyly and homostyly has been explored in a restricted number of species in series Turnera. Further study of species possessing both distylous and homostylous populations (e.g. Piriqueta morongii), and of sister species that vary in breeding system (e.g. homostylous Turnera candida and distylous T. grandiflora), will be important to understand the modes of origin of homostyly and/or provide further insight into whether a supergene underlies distyly in the Turneraceae. These species are also important candidates for ecological genetic investigations to explore the selective forces acting on breeding systems.

Finally, we require further investigations to elucidate the molecular genetic basis of distyly. Ideally, the development of transgenic or RNA interference methods in Turnera, would allow us to knock out genes for morph-specific proteins previously identified (e.g. style polygalacturonase) to explore their function. Similarly, the function of candidate genes identified through positional cloning can be explored in this manner. Once the locus or loci are identified, a clear picture of the genetic architecture of distyly should emerge.

Acknowledgements

We thank Jonathan Labonne for allowing us to refer to his unpublished data on genetic mapping in Turnera.

Ancillary