Morphology and Pollination Biology of an Intersectional Hybrid of Costus (Costaceae)

KENNETH J. SYTSMA Department of Botany,University of Wisconsin, Madison, Wisconsin 53706

RICHARD W. PIPPEN Biology Department,Western Michigan University, Kalamazoo, Michigan 49008

ABSTRACT.An intersectionalhybridization event between hummingbird-pollinated Costus pulverulentus(sect. Ornithophilus)and large bee-pollinated C. guanaiensisvar. macrostrobilus(sect. Costus) in Belize, Central America, is documented by analyses of morphology and pollination biology. Morphologically the hybrid is intermediatebetween the parental species and retains floral char- acteristicsadaptive to both hummingbird and bee pollination. Nectar secretion rate and sugar concentrationbut not sucrose:hexose ratio in the hybrid are intermediatebetween those of the parental species. The hybridization event probably arose by breakdown of stricthummingbird pollination in Costuspulverulentus.

CostusL. (Costaceae) occurs in moist habitats mingbird-pollinatedspecies because it is visit- in the Neotropics,and is characterizedby a spi- ed and pollinated by both euglossine bees and rally arranged phyllotaxyand showy terminal hummingbirds. inflorescence.Costus is often found associated Many species of Costusare sympatricand are with species of Calathea,Canna, Heliconia,and visitedby the same pollinators(Schemske 1981; Renealmiaalong the banks of riversand streams pers. obs.). Hybridization has been described and more disturbed habitats like foresttrails, forspecies of Costuswithin each section (Maas secondary forests, lightgaps, roadsides, and 1972, 1977) but to date there are no records of plantations. hybridization between bee-pollinated and In his earlier monograph of the Neotropical hummingbird-pollinatedspecies (P. J.M. Maas Costoideae, Maas (1972) divided the genus Cos- pers. comm.). This study presents evidence for tus into two subgenera: Cadalvena,represented intersectional hybridization involving two by seven species restrictedto South America; Costusspecies, one adapted to bee-pollination and Costus,consisting of 40 species rangingfrom and the other to hummingbird-pollination. northernCentral America to South America. In In 1976, a new Costuswas discovered on the a more recent treatment(Maas 1977), two pre- road to San Jose, 2.5 kilometers north of the sumably natural sections based on labellum junction with the San Antonio Road, Toledo characteristicsare recognized in subg. Costus. District, Belize. A single population of nine Section Costusis characterizedby bee-pollinat- plants was found growing at the road's edge ed flowers;the labellum is white to yellow, and for 50 meters.This site in moist tropical forest has a short broad tube and an open exposed of the Mayan mountains at elevations of 200- limb or landing platformoften striped with red 250 meters is an area of active road construc- or purple. Section Ornithophilusis characterized tion and slash/burn agriculture. by hummingbird-pollinatedflowers; the label- The two suspected parental species occur lum is yellow, orange, or red, and has a narrow, within 100 metersof the putative hybrid pop- thick-walled,fleshy tube barely exceeding the ulation. Costus guanaiensis var. macrostrobilus corolla. Several species, most notably C. malor- (Schumann) Maas (sect. Costus),a robustspecies tieanusWendl. and closely related C. pictusD. attaining heights over 3.5 meters,is found in Don, appear to be intermediatein characterbe- more exposed and drier roadside habitats. Cos- tween the two sections.According to F. G. Stiles tus pulverulentusC. B. Presl (sect. Ornithophilus) (in Maas 1977), C. malortieanusrepresents a is found in shaded and usually wetterroadside transition between bee-pollinated and hum- habitats than either C. guanaiensisvar. macro- 353

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 354 SYSTEMATIC BOTANY [Volume 10

TABLE 1. Morphological charactersof Costusgua- were standardized against the mean of char- naiensis,C. pulverulentus,and the putativehybrid used acter measurementsfrom individuals of Costus in constructingAndersonian hybrid indices and Eu- guanaiensisvar. macrostrobilusand C. pulverulen- clideandistance coefficients. tus located more than five kilometersfrom the study site. The corresponding mean measure- 1. Sheathdiameter (mm). 2. Ligulelength (mm). 3. ments in the two referencepopulations were Petiolelength (mm). 4. Leaf length(cm). 5. Leaf scaled between zero and one with zero repre- width(cm). 6. Inflorescencelength (cm). 7. Inflo- rescencewidth (cm). 8. Plantheight (cm). 9. Bract senting the measurementof C. guanaiensisand color. 10. Bract length (mm). 11. Bractwidth one the measurement of C. pulverulentus.For (mm). 12. Bractappendage length (mm). 13. Bract each character,individuals of the putative hy- appendagewidth (mm). 14. Dilaceratingbract fi- brid and suspected parental populations were bers, presence or absence. 15. Bracteolelength assigned the value of "0" if the characterhad (mm). 16. Calyx length (mm). 17. Calyx width a value on the reference scale less than 0.25, (mm). 18.Calyx lobe length(mm). 19.Corolla col- the value "0.5" if greaterthan or equal to 0.25 or. 20. Corollatube length (mm). 21. Corollatube and less than or equal to 0.75, and the value diameter(mm). 22. Corollalobe length(mm). 23. "1.0" if greaterthan 0.75. The hybridindex for Corolla lobe width (mm). 24. Stamen length each individual was obtained by summingthese (mm). 25. Stamenwidth (mm). 26. Antherlength (mm). 27. Stylelength (mm). values over all 27 characters. The Euclidean distance diagram requires that two referencepoints be formedfrom character ranges withinthe populations hypothesized to strobilus(hereafter simply referred to as C. be hybridizing. The reference point assigned guanaiensis)or the putative hybrid and forms to each putative parental population is com- large populations throughoutthe area. All three posed of the set of 27 characters.The value of Costuspopulations flowerduring the rainy sea- each character for a reference point is either son from mid-June through late-November. the maximum or minimum value observed in Costusscaber Ruis & Pavon (sect. Ornithophilus) the population. The maximum is used if the is the only other Costus species in the sur- mean of the characteris greaterthan the cor- rounding region, but it was not found in the responding mean of the other parental popu- hybrid zone. Based on floral and vegetative lation. The minimum is used if the mean of characteristicsit was evident thatC. scabercould thatparticular character is less than the respec- not be a parental species of the putative hybrid. tive mean of the otherparental population. The charactersfor each individual in the putative METHODS hybridpopulation, the suspected parental pop- Morphology, pollen viability, chromosome ulations, as well as the two reference points, number, and pollination biology were exam- are then ranged between zero and one (see ined in the putative hybrid and suspected Wells 1980). The distance of each individual parental species to verify the occurrence of from the reference points is calculated using intersectional hybridization. Vegetative char- Euclidean distance equations (see Wells 1980) acteristicswere obtained fromdried specimens and plotted onto the distance diagram using a collected in the studyarea and surroundingre- compass and ruler. gion of Toledo District.Voucher specimens are Pollen viability was assayed by the double deposited in WMU. Floral characteristicswere stainingtechnique of Owczaryak (1952). Viable obtained from material preserved in FAA. pollen grains were easily identified by the Twenty-sevenmorphological characters(table combination of methyl green stained pollen 1) were examined from seven individuals of exine and phloxine stained cytoplasm.The vi- the putative hybrid, nine individuals of C. able pollen grains expand and stain with both guanaiensis,and 12 individuals of C. pulverulen- dyes whereas the aborted pollen grains remain tus. shrunken and take only the methyl green ex- These characterswere used to produce both ine stain. At least 500 pollen grains fromeach Andersonian hybrid indices (Anderson 1936) of at least four individuals per taxon were and hybridindices based on Euclidean distance scored. Mitoticchromosome counts were made coefficients(Wells 1980). The hybrid indices using young root tips pre-soaked in p-dichlo-

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 1985] SYTSMA & PIPPEN: COSTUS 355

TABLE 2. Comparison of select morphological charactersof Costusguanaiensis, the putative hybrid,and C. pulverulentus in Belize.

Character C. guanazenszs Putative hybrid C. pulverulentus

Bract length/width (mm) 42-55/22-42 41-42/26-32 28-39/19-26 Bractappendage Present Present Absent Bractcolor Green Yellow-green Orange-red Dilacerating bract fibers Absent Present Present Bracteole length (mm) 25-40 22-27 13-21 Calyx length (mm) 11-17 9.5-16 7.5-10 Floral structure Large, flaringlabellum Tubular labellum, Narrow tubular label- or "bee landing plat- short platform, lum, no platform, form",opening 15-20 opening 6-9 mm opening 2-3 mm mm wide wide wide Floral color White with pink-red Red, yellowish red Orange-red to red stripes,yellowish platform white platform Floral texture Thin, membranaceous Thick, waxy Thick, waxy Corolla tube length (mm) 76-78 62-65 40-45 Labellum length (mm) 85-89 63-64 41-45

robenzene for nine hours, hydrolyzed in 10% reflexed bract appendages typical of C. gua- HCl at 60?C for six minutes, and squashed in naiensisand many other species in sect. Costus lacto-propionicorcein. (figs. 1-2, 4-5). The floral structureand color Nectar secretionpatterns were quantifiedby of the putative hybrid,however, is more rem- repeatedly sampling bagged flowers with mi- iniscent of the red, narrow, and thick tubular crocapillarytubes at hourlyintervals from dawn flowersof C. pulverulentusand other members (0600 hr) until the flowerseither fell fromthe of hummingbird-pollinatedsect. Ornithophilus inflorescenceor stopped producing measurable (figs.2-5). The presence of a wide floralopen- nectar (1500 hr). Nectar volume and nectar ing and a small, yellowish "landing platform" concentration (% sucrose equivalents on a (labellum lobe) indicate that the putative hy- weight/totalweight basis) were recorded for brid has retained some of the typical bee-pol- each flower at each hour interval. Total sugar lination charactersof sect. Costus,and C. gua- content (mg sucrose equivalents) was calculat- naiensis in particular. The presence of ed from standard sucrose conversion values dilaceratingfibers on exposed bractmargins of (Weast 1974). Records of visitorsthat used both the putative hybrid,a trait found exclusively pollen and nectar resources were made in C. pulverulentus(Maas 1972), furtherdocu- throughout an 8-hour period at each popula- mentsthe parentalinfluence of this species and tion. Nectar composition (sugars, amino acids, excludes C. scaberof the same sectionfrom being and other organic substances) was kindly de- considered one of the parental species. terminedby I. Baker, Universityof California, Andersonian hybrid indices of individuals Berkeley,on fourindividuals of Costuspulveru- fromthe study area populations of Costusgua- lentusand three each of C. guanaiensisand the naiensis,C. pulverulentus,and the putativehybrid putative hybrid following methods described are depicted in figure6. Based on the 27 char- (Baker and Baker 1976, 1979). actersused to generatethe indices, the putative hybridindividuals are clearly intermediatebe- RESULTS tween individuals of the two putative parental Morphology. Individuals of the new Costus species. The sample mean of the indices forin- population are consistently intermediate be- dividuals of the putative hybrid population tween individuals of C. guanaiensisand C. pul- (10.1) is not significantlydifferent from the ex- verulentusin both vegetative and floralcharac- pected mean (13.5) nor the actual mean (13.9) teristics(table 2). The putative hybrid has the of the indices between the putative parental large, ovoid, and leafy inflorescenceand the species (two-tailedt-test, P > 0.05 in both cases).

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 356 SYSTEMATICBOTANY [Volume 10

L 4~~

FIGS. 1-5. Flowersand inflorescencesof Costuspopulations in and aroundthe putativehybrid zone in Belize. 1. C. guanaiensis.2. C. pulverulentus.3. The putativehybrid. 4. Comparisonof inflorescences.a. C. guanaiensis.b. The putative hybrid. c. C. scaber.d. C. pulverulentus.5. Comparison of perianth. a. C. guanaiensis.b. The putative hybrid.c. C. pulverulentus.d. C. scaber.

Morphologicalintermediacy of the putative examined[Np] = 2017).Pollen viability was 85% hybridis shown in a Euclideandistance dia- for C. guanaiensis(NI = 6, NP = 2310) and 86% gram(fig. 7). The centralposition of points rep- forthe putative hybrid (NI = 4, NP = 4127). resentingindividuals from the putative hybrid Chromosomenumber. Mitotic chromosome populationwithout spreading between the pu- numbersfor Costus guanaiensis, C. pulverulentus, tativeparental positions suggest that the new and theirputative hybrid are the same (2n = Costuspopulation is a simpleF1 hybrid. 18).These counts are thesame as earliermitotic A probleminherent in any morphological counts of the two putativeparental species analysisof a presumedhybrid and itsparental (Maas 1972). species is thatmorphological intermediacy is Pollinationbiology. The threeCostus popula- nota necessaryoutcome in thehybrid depend- tionsare visitedby an arrayof faunaspecies ing on theaddictive effects of polygenic inher- (table3). Costusguanaiensis is pollinatedexclu- itance.A formalanalysis of the morphological sivelyby large,pollen and nectarcollecting fe- variationencountered in the putativehybrid male euglossinebees (Eulaemasp.). Pollen col- and its parentalspecies would requiresegre- lectingwas observedonly in theearly morning gationtests and backcrosses. (0700-0900hr) with nectar collecting the dom- Pollenviability. Pollen viabilityis 97% for inantactivity when the available pollen was Costuspulverulentus (number of individuals depleted.Visits by this species forthe nectar sampled[NJ = 5, totalnumber of pollen grains rewardwere consistently observed at about20-

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 1985] SYTSMA & PIPPEN: COSTUS 357

170

50- (n

3039

~20-

10 L

2 4 6 S 10o 12 14 16 13 20 22 2'4 26

HybridIndex

FIG. 6. Histogram of Andersonian hybridindices for Costusguanaiensis (black, N = 9), the putative hybrid (striped,N = 7), and C. vulverulentus(white, N = 12) based on 27 morphological characters. minute intervals.When entering the throatof verticalposition. The long-tonguedEulaema sp. the flower,the bees contact the single anther apparently can reach the nectar at the base of attached near the apex of the petaloid stamen the corolla because it can maintainthis position and the stigma positioned just above the an- forseveral seconds. Stamen contactoccurs dur- ther. Pollen deposition on the bee occurs on ing these visits but pollen deposition is con- the dorsal surface of the abdomen and thorax. fined to the head region. Reciprocal pollen Individuals of small, brilliantgreen euglossine transferby Eulaemasp. between C. pulverulentus bees (Euglossaspp.) were also observed visiting and C. guanaiensisis possible but probably very C. guanaiensis.These euglossine bees are inef- rare in mixed populations. fectivepollinators of C. guanaiensisbecause they The putative hybrid is visited by both the do not contacteither the antheror stigmawhen Eulaemasp. that regularlyvisit Costusguanaien- enteringthe throatof the flower. sis and the hermithummingbirds observed on Costuspulverulentus is visited by several pol- C. pulverulentus.The floraltube of the putative len vectorsbut most effectivelyby both hermit hybridflower is intermediatein size allowing and non-hermithummingbirds. Pollen depo- foranther and stigma contactby both types of sition occurs on either the upper portions of visitors. the bill or on the facial feathers.Regular visits Nectar secretionpatterns in Costusguanaiensis to these narrow-tubedflowers by the Eulaema and C. pulverulentusare somewhat similar with sp. that visits C. guanaiensiswere also observed. a mid-morningdecline in nectarvolume and a In one mixed patch of C. pulverulentusand C. subsequent peak just prior to noon (fig.8). Cos- guanaiensisoutside the hybridzone, individual tusguanaiensis, however, produces larger hour- bees visited both species on the same foraging ly volumes (x = 16.6 Al vs. x = 5.5 ,l) of more bout. Non-hermithummingbirds on the plants concentrated nectar (x = 26.5% vs. x = 16.9%) of C. pulverulentusroutinely would chase away than C. pulverulentus.The putative hybridis in- the intruding bees. The bees are too large to termediate in nectar secretion averaging 12.4 enter the narrow-tubedflowers of C. pulveru- Al hr-1 of 19.7% nectar. Nectar concentration lentusbut would cling to the flowerapex in the generally decreases during the day in all three

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 358 SYSTEMATIC BOTANY [Volume 10

//~ 0

A B

0 .25 .50 75 1.0 SCALE FIG. 7. Euclidean distance diagram of Costusguanaiensis (Cg, N= 9), the putative hybrid (Cg x Cp, N= 7), and C. pulverulentus(Cp, N = 12) based on 27 morphological characters.A and B are referencepoints for C. guanaiensisand C. pulverulentusgenerated by the algorithm. populations. This decrease in nectarconcentra- glossine bee-visited C. guanaiensis (3.9:1). Sur- tion observed is dramatic in C. guanaiensisand prisingly,the putative hybrid has the highest the putativehybrid because theirflowers begin ratio (4.8:1). Proportionalto this increase in the to abscise earlier in the afternoon. disaccharide:monosaccharide ratio is an in- Because environmentalconditions inside and outside the flower can influence the trade-off between nectar concentration and volume TABLE 3. Observed visitors to flowers of Costus guanaiensis,the putative hybrid,and C. pulverulentus. (Corbet et al. 1979), the hourly nectar energy P + indicates effectivepollinator; P - indicates visitor sucrose equivalents (Bolten expendituresin mg ineffectivein pollination. et al. 1979) were calculated forthe three Costus populations (fig. 9). Costusguanaiensis, the pu- C. guanaiensis tativehybrid, and C. pulverulentusexpended 4.6, x C. C. gua- pulveru- C. pulveru- 2.5, and 0.9 mg sucrose equivalents per hour, Visitor naiensis lentus lentus respectively.Despite the wide fluctuationsin Amaziliatzacatl concentration,individuals nectar volume and (rufous-tailedhumming- of all three Costuspopulations exhibit signifi- bird) P+ P+ cantly linear cumulative energy expenditures A. candida in producing floral nectar, with the putative (white bellied emerald) P+ hybrid intermediatebetween the two putative Phaethornissuperciliosus parental species. (long-tailed hermit) P+ P+ Sugars detected in the floral nectar include P. longuemareus glucose and fructose (monosaccharides), su- (little hermit) P+ crose (disaccharide), and melezitose (trisac- Eulaemasp. (large euglossine bee) P+ P+ P-(?) charide) (table 4). Nectar fromeach population Euglossaspp. The hummingbird- was sucrose-dominant. (small, metallic euglos- visited Costus pulverulentushas a lower disac- sine bees) P- charide:monosaccharide ratio (2.2:1) than eu-

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 1985] SYTSMA& PIPPEN: COSTUS 359

40-40 ] < ~~Cg 30-30

20, 1 220

10 I10,

40 40 Cg X Cp

_ 30 30 I-

2 . * 20

10 10 z z * ~~~~~~~~~~~~~~~~~~~~0

404 40 Cp

30 630

20 20

10 10

6 10~~~~i' 1'2 1'4 Hour (a.m.)

FIG. 8. Average hourly nectar production in Costusguanaiensis (Cg, N= 4), the putative hybrid (Cg x Cp, N = 4), and C. pulverulentus(Cp, N = 6). Bars represent1 standard error.

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 360 SYSTEMATIC BOTANY [Volume 10

* Cg r2= 0.98 stance where stronghybridization barriers have 60 * Cg X Cp r2= 0.92 been shown to be importantwithin a section of Costusis between C. alleniiMaas and C. laevis OCP r2= 0.95 Ruiz & Pavon (both of sect. Costus)(Schemske 40- 1981). These two unrelated species apparently have converged in floralcharacteristics to share the same pollinator, with selection operating li 20- in the formationor strengtheningof internal barriers to prevent hybridization (Schemske 1981). Costus guanaiensisof the same section, 6 8 10 12 14 16 however, has high seed set when pollinated by Hour (a.m.) either of these species. Despite the occurrence FIG. 9. Cumulative production of hourly sucrose of intrasectionalhybrids, hybridization in Cos- equivalents in floralnectar of Costusguanaiensis (Cg, tusmust be considered rare when the frequen- N = 4), the putative hybrid(Cg x Cp, N = 4), and C. cy of associations and transferof pollen be- pulverulentus(Cp, N = 6). tween sympatricspecies is taken into account. Even less is known about reproductivebar- riersbetween the two sectionsin Costus.To date crease in the presence of the trisaccharide me- no intersectionalhybrids have been reported. lezitose. This comparativestudy of the morphologyand Amino acid composition of floral nectar was pollination biology of the new Costuspopula- similar in Costus guanaiensis, C. pulverulentus,and tion in Belize with two sympatric species the putative hybrid. Trace amounts of alanine, stronglysupports the intersectionalhybrid or- glutamine, glycine, isoleucine, leucine, pro- igin of the new Costuspopulation. The inter- line, and valine were present in the nectars. sectional origin of the hybrid is best seen in Serine was the most abundant amino acid pres- features of the inflorescenceand flower. The ent with values of 3-4 on a scale of 1 to 5 with hybridhas acquired the squat, appendaged in- 5 representing the largest value. Proteins, florescenceof C. guanaiensisbut the yellowish phenols, alkaloids, and lipids were all absent white and shortly appendaged bracts ap- from the nectar. proach those of C. pulverulentus.The flower of the hybrid has the red tubular corolla of DISCUSSION hummingbird-pollinatedC. pulverulentusbut the Interspecific competition for pollinators and yellowish "landing platform"and large diam- the subsequent selection for reducing this com- eter opening of bee-pollinated C. guanaiensis. petition have often been cited as evolutionary The average hourly volume (16.6 ,l) and av- forces leading to the diversification of flower- erage concentration(26.5%) of Costusguanaien- ing times (Gentry 1974; Stiles 1977), flowering sis nectar is within the range of nectarproduc- morphologies (Schemske 1976; Stiles 1975), and tion for flowers pollinated by long-tongued flower color (Jones 1978) among congeneric and bees (Baker and Baker 1975); the smaller hourly sympatric species in the Neotropics. The diver- volume (5.5 ,l) of less concentrated (16.9%) sity of floral morphologies and colors in both nectar in C. pulverulentusis within the range of the bee- (sect. Costus) and hummingbird-polli- nectar production for flowers pollinated by nated (sect. Ornithophilus)species indicates that hummingbirds(Baker 1975; Bolten and Fein- floral evolution has been important in the di- singer 1978). It is not surprisingthat both hum- versification of Costus (Maas 1972, 1977; mingbirdsand large bees visit the hybridflow- Schemske 1981). er because it is structurally intermediate Whether this diversification within Costus has between the classical bee and hummingbird been accompanied by the formation of internal flowersand exhibitsan intermediacyin nectar genetic reproductive barriers is poorly known. production. The more than occasional occurrence of natural The sucrose-dominantnectar of the hybrid intrasectional hybrids (Maas 1972, 1977) might and parental Costuspopulations is concordant suggest that post-pollination barriers within with results fromprevious surveys (Baker and each section are weak. The only known in- Baker 1982, 1983) and pollinator preference

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 1985] SYTSMA & PIPPEN: COSTUS 361

TABLE4. Sugar composition in floralnectar of Costusguanaiensis, the putative hybrid,and C. pulverulentus.

Percentage (based on weight) Ratio of sucrose: glucose + Taxon N Glucose Fructose Sucrose Melezitose fructose C. guanaiensis 3 0.081 0.121 0.794 0.004 3.9:1 C. guanaiensis x C. pulverulentus 3 0.095 0.075 0.823 0.007 4.8:1 C. pulverulentus 4 0.155 0.156 0.688 0.001 2.2:1

(Stiles 1976) of hummingbirdand long-tongued DNA (rDNA) and chloroplastDNA of the three bee-visited flowers. It is interesting that the populations. For the few restrictionenzymes nectaravailable to the large Eulaemasp. bees is examined, rDNA from the hybrid appears to more sucrose-dominant(3.9:1, sucrose: hexose) be similar to that of C. guanaiensisand not to than thatavailable to hummingbirds(2.2:1). The that of C. pulverulentus(Sytsma unpubl. data). hybridflower nectar has a higher sucrose: hex- The occurrenceof an intersectionalhybrid in ose ratio (4.8:1) than eitherparental species and Costus between two species with divergent suggests that the inheritanceof nectar compo- modes of pollination raises several questions sition is more complex than usually believed. about reproductivebarriers in Costus.Has little A lack of qualitative variation in amino acid genetic incompatibilityarisen within each sec- complements of floral nectar between the pa- tion of Costusor even between sections during rental species prevented verificationof the hy- diversificationof the genus? Is species integrity brid origin of the newly found population of in Costusmaintained by mechanical barriersto Costusby comparison of amino acids (see Baker interspecificpollen movementor by pollinator and Baker 1976). specificity?Has hybridizationand subsequent The breakdown of strictlyhummingbird pol- gene flowbetween species played an important lination in Costuspulverulentus, as seen in mixed role within and between the two sections of Costus patches, is most likely responsible for Costus? In Heliconia (Heliconiaceae), another the formationof the hybrid.It is unclear if the Neotropical monocot genus exhibiting exten- hybridindividuals representthe F1products of sive floral differentiation,progamic (post-pol- a single hybridizationevent or if subsequent lination but pre-fertilization)barriers to hy- introgressionwith either parental species has bridization are well developed, although occurred.Transfer of pollen fromthe hybridto hybrids,as in Costus,are known but relatively either parental species is highly likely due to rare (Kress 1983). Answers to these questions the intermediatepollination syndrome of the concerning the biosystematics of Costus are hybrid. Unfortunately,attempts with water- being sought by a survey of post-pollination soluble dyes to observe pollen movement by barriers to hybridization and a DNA-based both hummingbirdand bee vectorsamong the phylogeneticanalysis of the subgenus Costus. hybrid and parental individuals have proven ACKNOWLEDGMENTS. Gratefulacknowledgment is unsuccessful. The unusually high pollen via- given to James Aldrich, Cory Berish, and Eric An- bility of the than that of Costus hybrid,greater derson forhelp in the field;to RobertCruden, Helen guanaiensis,would allow forbackcrossing if no Kennedy,and JohnKress forhelpful suggestions; and post-pollination compatibility barriers exist. to JackieSytsma for the illustrations. Because plants obtained fromthe three populations in the hybrid zone have not yet flow- LITERATURE CITED ered in synchrony,crosses among the three ANDERSON,E. 1936. A morphological populations have not been made. Furthersup- comparison of triploid and tetraploid interspecifichybrids in port for the of the new hybrid origin Costus Tradescantia.Genetics 21:61-65. population and clarificationof its statusas either BAKER,H. G. 1975. Sugar concentrationsin nectars an F1 or an introgressed hybrid population fromhummingbird flowers. Biotropica 7:37-42. might emerge from analysis of endonuclease and I. BAKER. 1975. Studies of nectar consti- restrictionsite variation in nuclear ribosomal tutionand pollinator-plantcoevolution. Pp. 100-

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions 362 SYSTEMATIC BOTANY [Volume 10

140 in Animaland plantcoevolution, eds. L. E. Gil- KRESS,W. J. 1983. Crossabilitybarriers in neotrop- bert and P. H. Raven. Austin: Univ. of Texas ical Heliconia.Ann. Bot. 51:131-148. Press. MAAS, P. J. M. 1972. Costoideae (Zingiberaceae). and . 1979. Sugar ratios in nectars. Flora Neotropica,Monograph 8. New York: Haf- Phytochem.Bull. 12:43-45. ner. and . 1982. Chemical constituentsof 1977. Renealmia(Zingiberaceae-Zingibero- nectar in relation to pollinator mechanisms and ideae) and Costoideae (Zingiberaceae). Flora phylogeny. Pp. 131-171 in Biochemicalaspects of Neotropica,Monograph 18. New York: Hafner. evolutionarybiology, ed. M. H. Nitecki. Chicago: OWCZARYAK,A. 1952. A rapid method formounting Univ. of Chicago Press. pollen grains, with special regard to sterility and . 1983. A brief historicalreview studies. Stain Technol. 27:249-251. of the chemistryof floralnectar. Pp. 126-152 in SCHEMSKE, D. W. 1976. Pollinator specificityin Lan- The biologyof nectaries,eds. B. Bentley and T. S. tana camaraand L. trifolia(Verbenaceae). Biotro- Elias. New York: Columbia Univ. Press. pica 8:260-264. BAKER,I. and H. G. BAKER. 1976. Analyses of amino . 1981. Floral convergence and pollinator acids in flowernectars of hybridsand their par- sharing in two bee-pollinated tropical herbs. ents, with phylogeneticimplications. New Phy- Ecology 62:946-954. tol. 76:87-98. STILES, F. G. 1975. Ecology, flowering phenology, BOLTEN, A. B. and P. FEINSINGER. 1978. Why do and hummingbirdpollination of some Costa Ri- hummingbirdflowers secrete dilute nectar?Bio- can Heliconiaspecies. Ecology 56:285-301. tropica 10:307-309. . 1976. Taste preferences,color preferences, - , , H. G. BAKER, and I. BAKER. 1979. On and flower choice in hummingbirds.The Con- the calculation of sugar concentrationin flower dor 78:10-26. nectar. Oecologia 41:301-304. . 1977. Coadapted competitors:The flower- CORBET, S. A., P. S. WILLMER,J. W. L. BEAMENT,D. M. ing seasons of hummingbird-pollinatedplants UNWIN,and 0. E. PRYS-JONES.1979. Post-secre- in a tropical forest.Science 198:1177-1178. torydeterminants of sugar concentrationin nec- WEAST, R. C., ed. 1974. Handbookof chemistryand tar. Plant Cell Envir. 2:293-308. physics,55th ed. Cleveland, Ohio: CRC Press. GENTRY,A. 1974. Coevolutionary patternsin Cen- WELLS, H. 1980. A distance coefficientas a hybrid- tral American Bignoniaceae. Ann. Missouri Bot. ization index: An example using Mimuluslongi- Gard. 61:728-759. florusand M. flemingii(Scrophulariaceae) from JONES, C. E. 1978. Pollinator constancyas a pre-pol- Santa Cruz Island, California. Taxon 29:53-65. lination isolating mechanismbetween sympatric species of Cercidium.Evolution 32:189-198.

This content downloaded from 128.104.46.206 on Mon, 16 Mar 2015 13:25:51 UTC All use subject to JSTOR Terms and Conditions