American Journal of Botany 90(5): 707±715. 2003.

DIVERSIFICATION OF ANDROMONOECY IN SECTION LASIOCARPA (): THE ROLES OF PHENOTYPIC PLASTICITY AND ARCHITECTURE1

JILL S. MILLER2 AND PAMELA K. DIGGLE

Department of Environmental, Population, and Organismic Biology, University of Colorado, Campus Box #334, Boulder, Colorado 80309 USA

Quantitative analyses of sexual expression show extensive interspeci®c variation in the strength of andromonoecy (proportions of hermaphroditic and staminate ¯owers) among Solanum species in the monophyletic section Lasiocarpa. The roles of phenotypic plasticity and inter- and intra-in¯orescence architecture in the diversi®cation of andromonoecy within this small clade were analyzed. Four species that represent a range of expression of andromonoecy were examined. Staminate ¯owers produced within in¯orescences ranged from 3% (S. candidum)to7%(S. ferox) in weakly andromonoecious species and from 39% (S. pseudolulo) to 60% (S. quitoense) in more strongly andromonoecious species. Manipulation of fruit set on clonal replicates of multiple genotypes demonstrated variation among species for phenotypic plasticity. The strongly andromonoecious species, S. pseudolulo and S. quitoense, were not plastic and produced a large proportion of staminate ¯owers regardless of fruiting treatment, whereas S. candidum and S. ferox were phenotypically plastic and produced signi®cantly more staminate ¯owers in the presence of developing fruit. Staminate ¯ower production of all four species varied both within and among in¯orescences. A greater proportion of staminate ¯owers were produced in distal (later produced) in¯orescences. Within in¯orescences, hermaphroditic ¯owers occurred in basal positions, whereas staminate ¯owers, when produced, occurred more distally. This pattern of staminate ¯ower production is qualitatively the same in all species investigated; however, quantitative variation in the transition from hermaphroditic to staminate ¯ower production within and among in¯orescences is associated with variation in the strength of andromonoecy. At least three factors have contributed to the diversi®cation of andromonoecy in section Lasiocarpa including the presence or absence of phenotypic plasticity in response to fruit set, quantitative variation in intra- and inter-in¯orescence architectural effects, and total ¯ower production.

Key words: andromonoecy; architecture; development; phenotypic plasticity; sexual systems; Solanaceae; Solanum.

Andromonoecy has historically been de®ned as a sexual novation of developmental mechanisms that allow for the pro- system in which produce both hermaphroditic and fe- duction of two distinct ¯oral types. Comparative developmen- male-sterile (hereafter, staminate) ¯owers. Andromonoecy oc- tal and morphological analyses of closely related species that curs in approximately 4000 species of ¯owering plants from vary in the magnitude of andromonoecy (i.e., in the relative at least 33 families (21 listed in Yampolsky and Yampolsky, production of hermaphroditic and staminate ¯owers) can iden- 1922; references for additional families available on request) tify those features that have been modi®ed in association with and has clearly evolved independently in numerous lin- the evolution and diversi®cation of this sexual system. Once eages. Considerable attention has focused on the adaptive sig- we understand how complex phenotypes such as andromon- ni®cance of andromonoecy (Bertin, 1982a; Whalen and Cos- oecy evolve, we can create and test more re®ned hypotheses tich, 1986; Anderson and Symon, 1989; Spalik, 1991; Diggle, for why they evolve. 1993, 1994; Emms, 1996; Emms et al., 1997). Andromonoecy In the plant family Solanaceae, andromonoecy is well doc- is widely thought to provide a mechanism to independently umented in Solanum subgenus Leptostemonum and has been adjust allocation to male and female function; individuals are reported for 13 of 22 described sections (Whalen and Costich, assumed to vary production of hermaphroditic and staminate 1986). Lasiocarpa is a small section of 12 species, within sub- ¯owers in response to changes in resource availability (re- genus Leptostemonum, that is strongly supported as monophy- viewed in Bertin, 1982a; but see Podolsky, 1993). Thus, an- letic based on morphological, allozyme, karyotype, and chlo- dromonoecy is often considered a form of phenotypic plastic- roplast restriction site data (Heiser, 1972, 1987; Whalen and ity. Regardless of the particular advantages of andromonoecy, Caruso, 1983; Bernardello et al., 1994; Bruneau et al., 1995). however, the evolution of this sexual system requires the in- Although all members of Lasiocarpa are andromonoecious, the numbers of hermaphroditic and staminate ¯owers produced 1 Manuscript received 6 August 2002; revision accepted 6 December 2002. among species in this section varies considerably (Whalen et The authors thank C. Heiser, L. Bohs, and the Botanical Garden of Nij- al., 1981). Some species are characterized as weakly andro- megen for seed accessions; T. Lemieux for providing valuable advice and monoecious (e.g., S. hirtum Vahl., Diggle, 1993; S. stramon- support regarding greenhouse space; H. Bechtold and K. Ryerson for assisting with data collection; R. Khorsand, T. Forbis, C. McGraw, and M. Allsop for ifolium Jacq., Whalen et al., 1981); that is, plants produce rel- assistance in the greenhouse; V. Ebbert for the use of a light meter; the De- atively few staminate ¯owers. In contrast, other species bear partment of E. P. O. Biology at the University of Colorado for greenhouse many staminate ¯owers and are described as strongly andro- space; and R. A. Levin and members of the Diggle and Friedman laboratory monoecious (e.g., S. quitoense Lam., Whalen and Costich, group for comments on earlier drafts of the manuscript. This work was sup- 1986; S. vestissimum Dunal., Whalen et al., 1981). The diver- ported by grants from the University of Colorado and the National Science sity of sexual expression among species in section Lasiocarpa Foundation DEB 9982489 to P.K.D. 2 Author for reprint requests, current address: Department of Biology, Am- makes this an ideal group for comparative analyses. herst College, Amherst, Massachusetts USA (Tel. 413-542-2168; FAX 413- Detailed investigation of Solanum hirtum (section Lasio- 542-7955; e-mail: [email protected]). carpa) has shown that sexual expression is phenotypically 707 708 AMERICAN JOURNAL OF BOTANY [Vol. 90 plastic; fruit-bearing plants produce a greater proportion of Lasiocarpa of Solanum subgenus Leptostemonum. All members of section staminate ¯owers than genetically identical plants that are pre- Lasiocarpa are sexually reproducing, self-compatible, and andromonoecious, vented from producing fruit (Diggle, 1993). Because fruit de- producing both hermaphroditic and staminate ¯owers (Whalen et al., 1981). velopment poses a signi®cant drain on resources available for The four species included were chosen to incorporate variation within section continued development (Lloyd, 1980), plasticity of sexual ex- Lasiocarpa. For example, vegetative size ranged from small, bushy species pression in S. hirtum is consistent with hypotheses for the evo- (S. pseudolulo, ϳ1.5 m tall, with considerable lateral branching) to larger, lution of andromonoecy that suggest that the production of treelike species (S. quitoense, Ͼ3 m tall, with minimal lateral branching). staminate ¯owers is a mechanism for adjusting allocation to Flower size and production also varied considerably. Flower production male and female function. As continued fruit production be- ranged from 5 or 6 ¯owers per in¯orescence for S. pseudolulo and S. ferox comes resource-limited, subsequent production of staminate (Whalen et al., 1981; J. S. Miller and P. K. Diggle, unpublished data) to Ͼ10 ¯owers allows reallocation of resources away from ``costly'' ¯owers per in¯orescence for S. candidum and S. quitoense. Flowers from S. quitoense were also 60±80% larger compared to the other species. gynoecia toward increased male function, vegetative growth, Seed for S. candidum was obtained from L. Bohs (University of Utah, Salt or storage (reviewed in Bertin, 1982a; Whalen and Costich, Lake City, Utah, USA) and was collected from La Cangreja Reserve in Costa 1986; Spalik, 1991). These analyses of S. hirtum indicate that Rica (Bohs no. 98±104). Solanum ferox var. lasiocarpum was collected from consideration of plasticity should be integral in comparative Indonesia (M. Ansyar 9605), and seed was obtained from C. Heiser (Univer- studies of the expression of andromonoecy in section Lasio- sity of Indiana, Bloomington, Indiana, USA). The Botanical Garden of Nij- carpa. megen, Netherlands, provided seed for S. pseudolulo (accession no. To compare andromonoecy and developmental plasticity 824750021). Seed for S. quitoense was acquired in Quito, Ecuador by C. among species, sexual expression must be characterized at Heiser. Voucher specimens for all species are housed at the University of multiple levels of morphological organization. In S. hirtum the Colorado herbarium (COLO). occurrence of staminate ¯owers varied both within and among in¯orescences. All in¯orescences on fruit-bearing plants pro- Experimental designÐPlants were grown from seed and clonally replicated duced both hermaphroditic and staminate ¯owers; however, by vegetative cuttings to produce genetically identical plants. For Solanum the proportion of staminate ¯owers increased in successively candidum, S. pseudolulo, and S. quitoense, four replicate clones were pro- developing in¯orescences (Diggle, 1994). We refer to this var- duced from either six (S. pseudolulo) or eight (S. candidum, S. quitoense) iation in sexual expression as an inter-in¯orescence architec- genotypes for a total of 88 plants. For S. ferox, six genotypes were used, and tural effect. In addition, within in¯orescences, ¯owers in basal eight replicate clones were produced per genotype (8 clones ϫ 6 genotypes positions were predictably hermaphroditic, that is, the pheno- ϭ 48 plants). All plants were grown in 11-L (3-gal) pots containing a 2 : 1 type of these ¯owers was ®xed regardless of treatment. Only mix of Fafard Growing Mix #2 (Conrad Fafard, Agawam, Massachusetts, the phenotype of ¯owers in distal positions within in¯ores- USA) to Persolite (Persolite Products, Florence, Colorado, USA) plus Os- cences was plastic and varied with fruiting treatment. These mocote 13-13-13 slow-release fertilizer (Scotts Company, Marysville, Ohio, ¯owers developed as hermaphroditic in the absence of fruit USA). Plants were watered daily with 150±200 ppm of Excel Magnitrate and as staminate on fruit-bearing plants (Diggle, 1994). We fertilizer (Scotts Company). Two (S. candidum, S. pseudolulo, S. quitoense) refer to variation in the developmental potential of individual or four (S. ferox) replicates for each genotype were randomly assigned posi- tions in each of two greenhouse rooms (110 and 111), and within each room ¯owers within in¯orescences as an intra-in¯orescence archi- these replicates were randomly assigned a fruiting treatment: ϩFruit or tectural effect (sensu Diggle, 1995). Both inter- and intra- ϪFruit. in¯orescence architectural effects varied among genotypes of An important determinant of plant resource status is fruit set; plants with S. hirtum and affected the magnitude of plasticity and the developing fruit have fewer resources available for growth and continued strength of andromonoecy. Therefore, we also ask how these reproduction than plants with no developing fruit (Lloyd, 1980). Because in- architectural effects are expressed in other species in section ¯orescences are produced continuously, fruits, ¯owers, and ¯ower buds occur Lasiocarpa and whether these effects contribute to the diverse simultaneously on each branch, and experimental manipulation of plant fruit- expression of andromonoecy among species. ing status is an effective, biologically relevant way to manipulate resource To investigate the potential roles of phenotypic plasticity status. For plants in the ϩFruit treatment, all open hermaphroditic ¯owers and architectural effects in the diversi®cation of andromon- were pollinated every other day using a mixture of pollen collected from oecy, we studied sexual expression for four additional species several (Ն3 genotypes) conspeci®c pollen donors. Hermaphroditic ¯owers in section Lasiocarpa: Lindl., S. ferox var. remained open for 2±3 d; therefore, most ¯owers were pollinated at least lasiocarpum (Dunal) Miq., S. pseudolulo Heiser, and S. qui- twice, thus maximizing the potential for successful fertilization. In contrast, toense Lam. The four species were chosen to encompass wide no ¯owers were pollinated on plants in the ϪFruit treatment and any auto- variation in the degree of andromonoecy. Speci®cally, we con- gamously produced fruits were removed shortly after their initiation. Autog- sider four questions for each of these species. (1) What is the amous fruit production was rare in S. ferox, S. pseudolulo, and S. quitoense, degree of andromonoecy and to what extent does andromon- but occurred occasionally in S. candidum. oecy vary among these closely related species? (2) Is sexual Plants of these species have a predictable architecture; branches are sym- expression phenotypically plastic with respect to fruiting status podial and produce in¯orescences sequentially and continuously. Two branch- (i.e., resource availability)? (3) How does inter-in¯orescence es on each plant were haphazardly selected and censused every other day. In¯orescences were numbered from the base of the branch such that basal, variation affect whole plant sexual expression? (4) What is the early-developing in¯orescences were numbered ®rst. Floral positions within distribution of hermaphroditic and staminate ¯owers within in¯orescences were numbered distally from the basal-most position. In¯ores- in¯orescences, and is there a predictable architectural com- cence position, ¯ower position, ¯oral sexual phenotype (i.e., hermaphroditic ponent of staminate ¯ower production within in¯orescences? or staminate), and fruit production (ϩFruit treatment only) were recorded for all ¯owers on 10 in¯orescences on each of two marked branches per individ- MATERIALS AND METHODS ual. Hermaphroditic ¯owers are easily distinguished from staminate ¯owers by their larger size and long-exserted styles that exceed the anthers (Whalen Study speciesÐSolanum candidum, S. ferox var. lasiocarpum, S. pseudo- et al., 1981; Whalen and Costich, 1986; Diggle, 1991; J. S. Miller and P. K. lulo, and S. quitoense are shrubby, lignescent perennials included in section Diggle, unpublished data). The experiment continued from July 2000 until May 2003] MILLER AND DIGGLEÐANDROMONOECY IN SOLANUM SECT. LASIOCARPA 709

Fig. 1. The proportion of staminate ¯owers produced by four species in Solanum section Lasiocarpa. Plotted values are back-transformed means shown with 95% con®dence intervals for genotypes in the ϪFruit treatment (open circles), the ϩFruit treatment (closed circles), and the grand means across both

treatments (shaded squares). The main effect of fruiting treatment was signi®cant for S. candidum (F1,31 ϭ 38.74, P ϭ 0.0001) and S. ferox (F1,71 ϭ 46.06, P

ϭ 0.0001), but not for S. pseudolulo (F1,9 ϭ 0.24, P ϭ 0.6343) or S. quitoense (F1,29 ϭ 0.61, P ϭ 0.4419).

May 2001 for S. candidum, S. pseudolulo, and S. quitoense and from Septem- roditic ¯owers into fruit, respectively. No unpollinated plants ber 2001 to March 2002 for S. ferox. of any species produced mature fruit.

Statistical analysesÐTo determine the effects of fruiting treatment and in- ter-in¯orescence ontogeny on sexual expression, we ®rst calculated the pro- Plasticity of sexual expression and variation among ge- portion of staminate ¯owers produced within each in¯orescence for all in¯o- notypesÐPlasticity of sexual expression differed among spe- rescences on marked branches. Proportions were arcsin-square-root trans- cies. Solanum candidum and S. ferox were phenotypically formed and analyzed as a repeated measures design using either PROC plastic; individuals of S. candidum and S. ferox with devel- MIXED (S. candidum, S. ferox, and S. quitoense; Littell et al., 1996) or PROC oping fruit produced signi®cantly more staminate ¯owers than GLM (S. pseudolulo; Littell et al., 1991) in SAS. Because in¯orescence pro- genetically identical individuals lacking fruit (Fig. 1). The duction occurred over time along branches, the repeated measure in these analyses was the proportion of staminate ¯owers produced within in¯ores- fruiting treatment by greenhouse room interaction was signif- cences, which were numbered from the earliest in¯orescence (ϭ1) to the last icant for S. candidum (F1,31 ϭ 5.54, P ϭ 0.025). This inter- in¯orescence (ϭ10) along each branch. In the PROC MIXED analyses, effects action re¯ects the difference between fruiting treatments in the included in the model were fruiting treatment (overall phenotypic plasticity), two greenhouse rooms: the ϩFruit plants produced 6% more genotype, greenhouse room (overall block effect), in¯orescence position (in- staminate ¯owers than ϪFruit plants in room 110 and 19% ter-in¯orescence effect), and all interaction terms; the interactions of impor- more in room 111. Nevertheless, when greenhouse rooms were tance to our interpretation were genotype by treatment (genotypic variation analyzed separately for S. candidum, staminate ¯ower produc- for plasticity) and in¯orescence by treatment (plasticity of inter-in¯orescence tion was signi®cantly higher for the ϩFruit treatment com- variation). Because the PROC MIXED model was not appropriate for analysis pared to the ϪFruit treatment in each room (room 110, F1,15 of S. pseudolulo (i.e., the maximum likelihood model did not ®t the observed ϭ 17.01, P ϭ 0.0009; room 111, F1,16 ϭ 23.91, P ϭ 0.0002). data), we used PROC GLM to analyze data for this species. This model In contrast to S. candidum and S. ferox, S. pseudolulo and S. included fruiting treatment, genotype, greenhouse room, and all interactions quitoense were not plastic; staminate ¯ower production did not as between-subject effects, as well as in¯orescence position as the within- subjects effect. All of the interaction terms with in¯orescence position were depend on fruiting status (Fig. 1). also included. Because PROC GLM excludes data from subjects with missing All genotypes of S. candidum and S. ferox in the ϩFruit data, we included data from only the ®rst eight in¯orescences of S. pseudo- treatment produced more staminate ¯owers than genetically lulo. identical plants in the ϪFruit treatment (Fig. 2A, B). For S. For each species, we determined the effect of intra-in¯orescence architec- candidum, the magnitude of the response differed among ge- tural variation on sexual expression by calculating the probability of produc- notypes (compare slopes in Fig. 2A); however, the difference ing a staminate ¯ower at each ¯oral position averaged over all in¯orescence was not signi®cant (i.e., there was no treatment by genotype positions. Hermaphroditic ¯owers were assigned the number 0 and staminate interaction; F7,31 ϭ 0.67, P ϭ 0.70). Hence, there was no ge- ¯owers were assigned the number 1, and these values were averaged for each netic variation for plasticity. The treatment by genotype inter- ¯oral position within in¯orescences. We used nonparametric Kendall rank action for S. ferox was signi®cant (F5,71 ϭ 3.28, P ϭ 0.01); correlations to determine if the probability of producing a staminate ¯ower however, this variation was contributed largely by a single, increased with increasing ¯oral position within in¯orescences. nonresponsive genotype (genotype 4; Fig. 2B). In general, the nonplastic species showed more variation RESULTS among genotypes in staminate ¯ower production than the plas- Fruit production of pollinated plantsÐThe percentage of tic species (Fig. 2C, D). There was signi®cant variation among hermaphroditic ¯owers that set fruit in the ϩFruit treatment genotypes in the grand mean of the proportion of staminate ranged from 23.4% Ϯ 1.5% (means Ϯ 1 SE, N ϭ 807) for ¯owers produced for the nonplastic species, S. pseudolulo (F5,9 Solanum pseudolulo to 61.2% Ϯ 1.2% (N ϭ 1708) for S. fer- ϭ 9.60, P ϭ 0.002) and S. quitoense (F7,29 ϭ 2.60, P ϭ 0.03), ox. and S. candidum matured 26.6% Ϯ but not for the plastic species S. candidum (F7,31 ϭ 1.92, P ϭ 2.0% (N ϭ 489) and 31.7% Ϯ 1.1% (N ϭ 1848) of hermaph- 0.10) or S. ferox (F5,71 ϭ 0.89, P ϭ 0.49). 710 AMERICAN JOURNAL OF BOTANY [Vol. 90

Fig. 2. Reaction norms for the proportion of staminate ¯owers produced within in¯orescences for (A) Solanum candidum, (B) S. ferox var. lasiocarpum, (C) S. pseudolulo, and (D) S. quitoense. Plotted values are back-transformed means for each genotype in the two treatments. Numbers at right indicate the genotype identity. Note differences in the scale of the y-axes.

Inter-in¯orescence effectsÐIn¯orescence position had a production was signi®cantly positively correlated with ¯oral strong and signi®cant effect on staminate ¯ower production position for S. candidum and S. ferox in both treatments (S. for all four species. In S. candidum and S. ferox, the in¯ores- candidum,TauϾ 0.84, Z Ͼ 3.9, P Ͻ 0.0001; S. ferox,Tauϭ cence effect depended on the fruiting treatment, and there was 1.0, Z Ͼ 3.7, P ϭ 0.0002), and this effect was more pro- a signi®cant in¯orescence by treatment interaction for these nounced in the ϩFruit treatment (Fig. 4A, B). There were species (S. candidum, F9,197 ϭ 8.10, P ϭ 0.0001; S. ferox, F9,535 similar architectural effects on the probability of producing ϭ 3.99, P ϭ 0.0001). Staminate ¯ower production in the staminate ¯owers for S. pseudolulo (Tau ϭ 1.0, Z ϭ 3.46, P ϪFruit treatment changed little (Ͻ15%) in successive in¯o- ϭ 0.0005; Fig. 4C) and S. quitoense (Tau ϭ 0.509, Z ϭ 2.85, rescences, whereas in the ϩFruit treatment staminate ¯ower P ϭ 0.0043; Fig. 4D). production increased dramatically with in¯orescence position (73% in S. candidum and 44% in S. ferox; Fig. 3A, B). For Block effectÐThere was a signi®cant main effect of green- S. pseudolulo and S. quitoense, in¯orescences in more distal house room on sexual expression for three of the four species positions produced signi®cantly more staminate ¯owers (S. (S. candidum, F1,31 ϭ 27.26, P ϭ 0.0001; S. ferox, F1,71 ϭ pseudolulo, F7,63 ϭ 10.84, P ϭ 0.0001; S. quitoense, F9,158 ϭ 49.24, P ϭ 0.0001; S. pseudolulo, F1,9 ϭ 11.84, P ϭ 0.007). 6.69, P ϭ 0.0001; Fig. 3C, D); however, there were no sig- Genotypes of S. candidum, S. ferox, and S. pseudolulo grown ni®cant interactions between in¯orescence position and fruit- in greenhouse room 111 produced, on average, 10% more sta- ing treatment for either species. minate ¯owers compared to identical genotypes housed in greenhouse room 110. Though not statistically signi®cant for

Intra-in¯orescence effectsÐThe probability of producing S. quitoense (F1,29 ϭ 0.90, P ϭ 0.3499), genotypes in this staminate ¯owers increased in distal ¯oral positions within in- species showed the same trend, producing 7% more staminate ¯orescences for all four species. Speci®cally, staminate ¯ower ¯owers in greenhouse room 111 compared to room 110. May 2003] MILLER AND DIGGLEÐANDROMONOECY IN SOLANUM SECT. LASIOCARPA 711

Fig. 3. Staminate ¯ower production in successive in¯orescence positions for (A) Solanum candidum, (B) S. ferox var. lasiocarpum, (C) S. pseudolulo, and (D) S. quitoense. For the plastic species, S. candidum and S. ferox var. lasiocarpum, open symbols represent the ϪFruit treatment mean and closed symbols represent the ϩFruit treatment mean. For the nonplastic species, S. pseudolulo and S. quitoense, the grand means are presented. Plotted values are back- transformed means shown with 95% con®dence intervals. Note differences in the scale of the y-axes; dotted lines indicate 50% staminate ¯ower production within in¯orescences.

Fig. 4. Staminate ¯ower production within in¯orescences, averaged across all in¯orescence positions, for (A) Solanum candidum, (B) S. ferox var. lasio- carpum, (C) S. pseudolulo, and (D) S. quitoense. For the plastic species, S. candidum and S. ferox var. lasiocarpum, open symbols represent the ϪFruit treatment mean and closed symbols represent the ϩFruit treatment mean. For the nonplastic species, S. pseudolulo and S. quitoense, the grand means are presented. Error bars are Ϯ1 SE. 712 AMERICAN JOURNAL OF BOTANY [Vol. 90

DISCUSSION although plasticity is a mechanism for staminate ¯ower pro- duction by some taxa within Lasiocarpa, differences in plas- The 12 species of Solanum section Lasiocarpa are all an- ticity do not underlie the observed variation in andromonoecy. dromonoecious and presumably inherited this sexual system Our results demonstrate that analyses of phenotypic plastic- from a common ancestor. Nevertheless, quantitative analyses ity are always context speci®c. Although the four species dif- of sexual expression for ®ve species show extensive interspe- fered in their capacity to respond to the fruiting treatment, they ci®c variation in the degree of andromonoecy. We argue below all responded plastically to the different environmental con- that evolutionary modi®cation of developmental plasticity and ditions of the two greenhouse rooms. All species produced inter- and intra-in¯orescence architecture interact to produce more staminate ¯owers in greenhouse room 111 compared to the diverse expressions of andromonoecy observed among room 110, signi®cantly so for S. candidum, S. ferox, and S. species of section Lasiocarpa. pseudolulo. The two rooms are physically adjacent to each other and both maintained a mean temperature of 21ЊC. How- Andromonoecy in Solanum section LasiocarpaÐThe de- ever, room 111 was signi®cantly more shaded than room 110 gree of andromonoecy differs widely among species in Sola- (159.0 Ϯ 13.0 ␮mol vs. 264.3 Ϯ 33.4 ␮mol, respectively; F1,98 num section Lasiocarpa. Choice of the appropriate metric for ϭ 8.6, P ϭ 0.004). Studies of other andromonoecious species quantitative comparison among the species, however, is prob- have shown that decreased light is associated with staminate lematic. Although the ϪFruit treatment may not be realistic in ¯ower production (Aesculus pavia, Bertin, 1982b; Solanum natural populations, we argue that comparisons among species carolinense, Solomon, 1985). In addition, population sex ra- should be made using the ϪFruit treatment because patterns tios in some diphasic orchids are female biased in open hab- of sexual expression using data from plants without fruit mea- itats and male biased in closed canopies (Gregg, 1975, 1978; sures their inherent tendency to produce staminate ¯owers. In Zimmerman, 1991). The intuitive (but untested) explanation is addition, the ϪFruit treatment is equivalent across species, that plants growing under increased light have greater photo- whereas the ϩFruit treatment may differ among species due synthate availability and thus an enhanced ability to mature to differential fruit production or fruit size. It should be noted, fruit. It would be advantageous for plants under such condi- however, that the rank order of the strength of andromonoecy tions to produce hermaphroditic ¯owers. Whatever the proxi- is the same regardless of the measure used for comparison mate cue, staminate ¯ower production differed signi®cantly (Fig. 1). among rooms and represents a distinct form of plasticity in In the absence of fruit production, S. candidum and S. ferox sexual expression. (this study) and S. hirtum (Diggle, 1993) produce few stami- nate ¯owers (3%, 7%, and Ͻ1%, respectively; Fig. 1), and therefore should be considered weakly andromonoecious. In Architectural effects on sexual expression in section La- contrast, even in the absence of fruit production, S. quitoense siocarpaÐAll species of Lasiocarpa investigated bear stami- is strongly andromonoecious and produces nearly 60% sta- nate ¯owers in a predictable pattern that encompasses both minate ¯owers per in¯orescence. Solanum pseudolulo is in- within- and among-in¯orescence variation. At the level of termediate and produces 39% staminate ¯owers (Fig. 1). Al- branches, a greater proportion of staminate ¯owers are borne though all of these species are typologically andromonoecious in later-produced in¯orescences. For S. candidum and S. ferox, and are capable of producing both hermaphroditic and stami- much of this pattern is the result of plastic changes in sexual nate ¯owers, they differ dramatically in the magnitude of ex- expression in response to fruiting treatment. Staminate ¯ower pression of this sexual system, despite sharing a common or- production in basal in¯orescences of plants in the ϪFruit treat- igin of andromonoecy. ment was near zero and remained less than 15% even in the distal-most in¯orescences. Among in¯orescences of plants in Plasticity in Solanum section LasiocarpaÐPlasticity of al- the ϩFruit treatment, however, staminate ¯ower production location to male and female function generally is thought to increased by over 40% from basal to distal in¯orescences (Fig. be an important component of the evolution of andromonoecy 3A, B). A similar pattern of inter-in¯orescence variation in (reviewed in Bertin, 1982a) and is clearly the mechanism by association with plasticity occurred in S. hirtum (Diggle, which staminate ¯ower production occurs in S. hirtum (Diggle, 1993). This pattern of increased staminate ¯ower production 1993). Yet, analyses of four additional species of Lasiocarpa in successive in¯orescences is consistent with a plastic re- reveal that staminate ¯ower production is a phenotypically sponse by developing ¯owers to the presence of maturing fruit plastic response to fruit production in S. candidum and S. fer- at earlier (more basal) in¯orescence positions. ox, but is a ®xed aspect of the phenotype in S. quitoense and Staminate ¯ower production in the nonplastic species, S. S. pseudolulo (Fig. 1). Thus, andromonoecy and the produc- pseudolulo and S. quitoense, also increases in distal in¯ores- tion of staminate ¯owers in Lasiocarpa is not inevitably a cences (Fig. 3C, D); however, this pattern is not dependent on plastic response to the presence of developing fruit. fruiting treatment, rather, it is a ®xed aspect of the architecture The presence or absence of plasticity per se appears to be of each branch. The pattern of inter-in¯orescence increase in associated with differences in the strength of andromonoecy: the proportion of staminate ¯owers in the nonplastic species plastic species are weakly andromonoecious, whereas stami- is qualitatively similar to that of fruit-bearing plants of the nate ¯ower production is ®xed in strongly andromonoecious plastic species (S. candidum and S. ferox; Fig. 3), though sta- species. Considering only the plastic species, we cannot de- minate ¯ower production commences at earlier (more basal) termine whether variation in the magnitude of plasticity (the in¯orescences in the nonplastic species. For example, the ®rst difference in staminate ¯ower production between the two in¯orescence bears 20% (S. pseudolulo) to 40% (S. quitoense) treatments) is associated with variation in the degree of an- staminate ¯owers in nonplastic species compared to near zero dromonoecy; plasticity and the degree of andromonoecy are in the plastic species (compare Fig. 3A, B to C, D). Strong similar among S. ferox, S. candidum, and S. hirtum. Thus, andromonoecy exhibited by S. pseudolulo and S. quitoense is May 2003] MILLER AND DIGGLEÐANDROMONOECY IN SOLANUM SECT. LASIOCARPA 713 due, in part, to the onset of staminate ¯ower production at early in¯orescence positions. Within in¯orescences of each of the four species, staminate ¯owers can be produced at all ¯oral positions; however, the probability is very low at basal positions and increases distally (Fig. 4). Thus, the general pattern is one of hermaphroditic ¯owers in basal positions, and staminate ¯owers, if produced, in terminal portions of each in¯orescence. Although the pat- tern is qualitatively similar, there are quantitative differences in the average intra-in¯orescence pattern of staminate ¯ower production among the four species and between treatments of the plastic species. Two indices can be used for comparison: the ¯ower position at which the probability of producing a staminate ¯ower exceeds 50% (dotted line in Fig. 4) and the probability of staminate ¯ower production in the distal-most positions. In in¯orescences of the strongly andromonoecious species, S. quitoense and S. pseudolulo, staminate ¯ower pro- duction occurs early (Ͼ50% probability by ¯ower position 4± 5) and, in the distal-most positions, the probability of stami- nate ¯owers exceeds 80% (Fig. 4C, D). In contrast, in weakly andromonoecious S. ferox, staminate ¯ower production does not exceed 50% until ¯ower position 7 in fruit-bearing plants and never exceeds 50% in nonfruiting plants (Fig. 4B). In S. candidum, despite a distal increase in staminate ¯ower pro- duction, the average probability never exceeds 50% at any position in either treatment (Fig. 4A). Thus, within in¯ores- cences, strong andromonoecy is associated with an earlier tran- sition from hermaphroditic to staminate ¯ower production and a greater probability of staminate ¯owers in distal-most posi- tions. The preceding analysis considers the intra-in¯orescence pat- tern averaged across all 10 in¯orescences per branch (Fig. 4). While this summary allows for explicit quantitative compari- sons, it obscures the signi®cant variation in staminate ¯ower production among in¯orescences. Figure 5 depicts the com- bined effects of both sources of variation in sexual expression: intra- and inter-in¯orescence architectural effects. This repre- sentation con®rms that staminate ¯owers are produced distally within in¯orescences and that the transition from hermaphro- ditic to staminate ¯ower development occurs at more basal ¯oral positions within later, more distal in¯orescences. More- over, Fig. 5 emphasizes that strong andromonoecy is the result of an earlier transition from hermaphroditic to staminate ¯ower production both within and among in¯orescences. For exam- Fig. 5. Intra- and inter-in¯orescence effects on staminate ¯ower produc- ple, the strongly andromonoecious S. quitoense bears stami- tion for (A) Solanum candidum, (B) S. ferox var. lasiocarpum, (C) S. pseu- nate ¯owers on the ®rst in¯orescence, and within that in¯o- dolulo, and (D) S. quitoense. In¯orescence position along the branch is shown rescence, the transition to staminate ¯owers occurs at ¯ower schematically along the horizontal axis (for the ®rst 10 in¯orescences). Floral position six (Fig. 5D). In contrast, the weakly andromonoe- position within in¯orescences increases vertically. Light shading indicates a cious species produce staminate ¯owers at later in¯orescences Ͻ50% probability of producing a staminate ¯ower, whereas dark shading (2 or 5) and later ¯ower positions (8 or 11; S. ferox and S. indicates a Ն50% probability of producing a staminate ¯ower. For the plastic species, S. candidum and S. ferox var. lasiocarpum, the top panel is the ϪFruit candidum, respectively; Fig. 5A, B). These comparisons in- treatment and the lower panel is the ϩFruit treatment. For the nonplastic dicate that diversi®cation of andromonoecy has occurred by species, S. pseudolulo and S. quitoense, the grand means are presented. quantitative variation in both inter- and intra-in¯orescence po- sitional effects on ¯oral development. The number of ¯owers borne on each in¯orescence also can Comparisons among more closely related species shows this contribute to differences in the strength of andromonoecy. to be incorrect. Clearly, the addition of ¯owers, which are Hermaphroditic ¯ower number is comparable for S. pseudol- predominantly staminate, increases the degree of andromon- ulo and S. quitoense (Fig. 5C, D); however, the more strongly oecy for S. quitoense (Fig. 5D). andromonoecious S. quitoense produces considerably more (typically staminate) ¯owers per in¯orescence. Based on an Implications for andromonoecy in SolanumÐQuantitative informal survey of the genus Solanum, Whalen and Costich variation within an otherwise uniform architectural pattern ap- (1986) predicted that strong andromonoecy would not be as- pears to underlie differences in sexual expression within So- sociated with the production of excess staminate ¯owers. lanum section Lasiocarpa. This model may be extended to 714 AMERICAN JOURNAL OF BOTANY [Vol. 90 other members of the genus. General branching architecture oecy is typically assumed to be a form of plasticity, traditional (sensu Halle et al., 1978) is quite uniform throughout the ge- hypotheses would predict that andromonoecy arose in the an- nus Solanum: branches grow sympodially with each sympodial cestor of Lasiocarpa as a plastic response and has become a unit bearing a characteristic number of leaves and terminating ®xed developmental pattern in some species. in an in¯orescence (Child, 1979; Child and Lester, 1991). We also found an association between weak andromonoecy Thus, each leafy branch produces an indeterminate succession and plasticity and between strong andromonoecy and ®xed of in¯orescences and typically bears fruits, ¯owers, and ¯oral phenotypes. The association between plasticity and the degree primordia simultaneously. Within the genus, andromonoecy of andromonoecy may be related to developmental constraints has evolved and diversi®ed within the context of this stereo- imposed by the architecture of Solanum. Given that in¯ores- typical morphology. In all described andromonoecious Sola- cences invariably contain both hermaphroditic and staminate num (belonging to different sections or even to different sub- ¯owers, an effective way to increase staminate ¯ower produc- genera and likely representing more than one origin of andro- tion is to produce staminate ¯owers at earlier (more basal) monoecy), staminate ¯owers occur predominantly in distal po- positions both within and among in¯orescences. Moreover, for sitions within in¯orescences (S. torvum, Hossain, 1973; S. the plastic species, if initiation of fruit development is the sodomeum and S. campanulatum, Symon, 1979; S. margina- proximate cue for staminate ¯ower development, then the pro- tum, Dulberger et al., 1981; section Lasiocarpa, Whalen et al., duction of staminate ¯owers will not be possible until later in 1981; S. paliacanthum, Coleman and Coleman, 1982; S. car- branch ontogeny after signi®cant numbers of fruit producing olinense, Solomon, 1985; references to additional species in hermaphroditic ¯owers have developed. Hence, plastic plants Whalen and Costich, 1986 and Anderson and Symon, 1989). may necessarily produce more hermaphroditic ¯owers relative Other patterns of staminate ¯ower production, for example, to staminate ¯owers and will be classi®ed as weakly andro- basal staminate and distal hermaphroditic ¯owers within in¯o- monoecious. rescences or the production of exclusively hermaphroditic and exclusively staminate in¯orescences within the same branch, ConclusionsÐThat ¯owering plants as a group possess an have never been described for Solanum. Based on these ob- enormous diversity of sexual systems is well known (Darwin, servations, it is likely that what variation in sexual expression 1877). Surprisingly, the expression of a single sexual system exists among andromonoecious species of the genus as a can vary widely, even among members of a small clade. An- whole depends on the types of quantitative variation in archi- dromonoecy has been de®ned typologically as a single phe- tectural effects described here for section Lasiocarpa. notype, that is, the presence of both hermaphroditic and sta- Architectural effects may also have played a critical role in minate ¯owers on individual plants. Analysis of Solanum sec- the evolutionary transition from hermaphroditism to andro- tion Lasiocarpa shows that andromonoecy clearly varies monoecy in Solanum. In many hermaphroditic species (re- among species. The common ancestor of these species was gardless of taxon), fruit set is reduced or absent at distal po- likely andromonoecious, thus the variation documented here sitions within in¯orescences (reviewed in Diggle, 1995, 2002); represents evolutionary diversi®cation from a single ``andro- thus, many hermaphroditic taxa are functionally andromon- monoecious phenotype.'' At least three factors contribute to oecious. Intra-in¯orescence variation in female function has the diversi®cation of sexual expression in this group, including not been investigated explicitly in hermaphroditic species of the presence or absence of phenotypic plasticity in response Solanum; however, fruit size declines dramatically within in- to fruit set, intra- and inter-in¯orescence architectural effects, ¯orescences of cultivated tomato (Bertin, 1995), and distal and total ¯ower production. Consideration of the stereotypical hermaphroditic ¯owers in species of section Lasiocarpa rarely branching architecture of the genus Solanum and the recurring set fruit (P. K. Diggle, personal observation). Within Solanum, intra- and inter-in¯orescence pattern of staminate ¯ower pro- intra-in¯orescence declines in female function may be an evo- duction among all andromonoecious species indicates that lutionary antecedent of andromonoecy. If so, then the transi- these factors may also have been important in the evolutionary tion from functional to morphological andromonoecy would dynamics of andromonoecy in the genus as a whole. require only the origin of a developmental mechanism for pre- anthesis termination of ovary development in ¯owers that are LITERATURE CITED already functionally female-sterile. ANDERSON,G.J.,AND D. E. SYMON. 1989. Functional dioecy and andro- Evolutionary dynamics of plasticityÐAlthough variation in monoecy in Solanum. Evolution 43: 204±219. BERNARDELLO, L. M., C. B. HEISER, AND M. PIAZZANO. 1994. Karyotypic the magnitude of plasticity may not underlie species-level var- studies in Solanum section Lasiocarpa (Solanaceae). American Journal iation in andromonoecy, we found variation in plasticity per of Botany 81: 95±103. se. Within the small section Lasiocarpa, staminate ¯ower pro- BERTIN, R. I. 1982a. The evolution and maintenance of andromonoecy. Evo- duction is a phenotypically plastic response to fruit production lutionary Theory 6: 25±32. in some species and a ®xed aspect of the phenotype in others. BERTIN, R. I. 1982b. The ecology of sex expression in red buckeye. Ecology Interestingly, the architectural pattern of staminate ¯ower pro- 63: 445±456. BERTIN, R. I. 1995. Competition for assimilates and fruit position affect fruit duction is qualitatively similar in nonplastic species and fruit- set in indeterminate greenhouse tomato. Annals of Botany 75: 55±65. bearing plants of the plastic species. There is a transition from BRUNEAU, A., E. E. DICKSON, AND S. KNAPP. 1995. Congruence of chloro- hermaphroditic to staminate ¯ower production within each in- plast DNA restriction site characters with morphological and isozyme ¯orescence, and this transition occurs earlier (at more basal data in Solanum sect. Lasiocarpa. Canadian Journal of Botany 73: 1151± positions) on successive in¯orescences. Thus, it is qualitatively 1167. the same phenotype that is ®xed in some species and produced CHILD, A. 1979. A review of branching patterns in the Solanaceae. In J. G. Hawkes, R. N. Lester, and A. D. Skelding [eds.], The biology and tax- as a plastic response to fruit set in others. The pressing issue onomy of the Solanaceae, 345±356. Academic Press, London, UK. for future research is to understand the sequence of evolution- CHILD, A., AND R. N. LESTER. 1991. Life form and branching within the ary transitions between these phenotypes. Because andromon- Solanaceae. In J. G. Hawkes, R. N. Lester, M. Nee, and N. Estrada [eds.], May 2003] MILLER AND DIGGLEÐANDROMONOECY IN SOLANUM SECT. LASIOCARPA 715

Solanaceae III: taxonomy, chemistry, evolution. Royal Botanic Gardens HEISER, C. B. 1972. The relationship of the naranjilla, Solanum quitoense. Kew and Linnean Society, London, UK. Biotropica 4: 77±84. COLEMAN,J.R.,AND M. A. COLEMAN. 1982. Reproductive biology of an HEISER, C. B. 1987. Origins of Solanum lasiocarpum and S. repandum. Amer- andromonoecious Solanum (S. paliacanthum Dun.). Biotropica 14: 69± ican Journal of Botany 74: 1045±1048. 75. HOSSAIN, M. 1973. Observations on stylar heteromorphism in Solanum tor- DARWIN, C. 1877. The different forms of ¯owers on plants of the same vum Sw. (Solanaceae). Botanical Journal of the Linnean Society 66: 291± species. Murray, London, UK. 301. ᭧ DIGGLE, P. K. 1991. Labile sex expression in andromonoecious Solanum hir- LITTELL, R. C., R. J. FREUND, AND P. C. SPECTOR. 1991. SAS system for tum: pattern of variation in ¯oral structure. Canadian Journal of Botany linear models, 3rd edition. SAS Institute, Cary, North Carolina, USA. 69: 2033±2043. LITTELL, R. C., G. A. MILLIKEN,W.W.STROUP, AND R. D. WOLFINGER. ᭧ DIGGLE, P. K. 1993. Developmental plasticity, genetic variation, and the evo- 1996. SAS system for mixed models, 3rd ed. SAS Institute, Cary, North lution of andromonoecy in Solanum hirtum (Solanaceae). American Jour- Carolina, USA. nal of Botany 80: 967±973. LLOYD, D. G. 1980. Sexual strategies in plants I. An hypothesis of serial DIGGLE, P. K. 1994. The expression of andromonoecy in Solanum hirtum adjustment of maternal investment during one reproductive session. New (Solanaceae): phenotypic plasticity and ontogenetic contingency. Amer- Phytologist 86: 69±79. ican Journal of Botany 81: 1354±1365. PODOLSKY, R. D. 1993. Evolution of a ¯ower dimorphism: how effective is DIGGLE, P. K. 1995. Architectural effects and the interpretation of patterns pollen dispersal by ``male'' ¯owers? Ecology 74: 2255±2260. of fruit and seed development. Annual Review of Ecology and System- SOLOMON, B. P. 1985. Environmentally in¯uenced changes in sex expression atics 26: 531±552. in an andromonoecious plant. Ecology 66: 1321±1332. SPALIK, K. 1991. On the evolution of andromonoecy and `overproduction' DIGGLE, P. K. 2002. Architectural effects on ¯oral form and function: a re- of ¯owers: a resource allocation model. Biological Journal of the Lin- view. In T. Stuessy, E. HoÈrandl, and V. Mayer [eds.], Deep morphology: nean Society 42: 325±336. toward a renaissance of morphology in plant systematics. Koeltz, KoÈn- SYMON, D. E. 1979. Sex forms in Solanum (Solanaceae) and the role of igstein, Germany. pollen collecting insects. In J. G. Hawkes, R. N. Lester, and A. D. Skeld- DULBERGER, R., A. LEVY, AND D. PALEVITCH. 1981. Andromonoecy in So- ing [eds.], The biology and taxonomy of the Solanaceae, 385±397. Ac- lanum marginatum. Botanical Gazette 142: 259±261. ademic Press, London, UK. EMMS, S. K. 1996. Temporal patterns of seed set and decelerating ®tness WHALEN,M.D.,AND E. E. CARUSO. 1983. Phylogeny in Solanum sect. returns on female allocation in Zigadenus paniculatus (Liliaceae), an Lasiocarpa (Solanaceae): congruence of morphological and molecular andromonoecious lily. American Journal of Botany 83: 304±315. data. Systematic Botany 8: 369±380. MMS TRATTON AND NOW E , S. K., D. A. S , A. A. S . 1997. The effect of in¯o- WHALEN,M.D.,AND D. E. COSTICH. 1986. Andromonoecy in Solanum. In rescence size on male ®tness: experimental tests in the andromonoecious W. G. D'Arcy [ed.], Solanaceae: biology and systematics, 284±302. Co- lily, Zigadenus paniculatus. Evolution 51: 1481±1489. lumbia University Press, New York, New York, USA. GREGG, K. B. 1975. The effect of light intensity on sex expression in species WHALEN, M. D., D. E. COSTICH, AND C. B. HEISER. 1981. Taxonomy of of Cyanoches and Catasetum (Orchidaceae). Selbyana 1: 101±113. Solanum section Lasiocarpa. Gentes Herbarium 12: 41±129. GREGG, K. B. 1978. The interaction of light intensity, plant size, and nutrition YAMPOLSKY, C., AND H. YAMPOLSKY. 1922. Distribution of sex forms in the on sex expression in Cyanoches (Orchidaceae). Selbyana 2: 212±223. phanerogamic ¯ora. Bibliotheca Genetica 3: 1±62. HALLEÂ , F., R. A. A. OLDEMAN, AND P. B. TOMLINSON. 1978. Tropical trees ZIMMERMAN, J. K. 1991. Ecological correlates of labile sex expression in the and forests. An architectural analysis. Springer-Verlag, Berlin, Germany. orchid Catasetum viridi¯avum. Ecology 72: 597±608.