OF OXALIS ALPINA (OXALIDACEAE) IN THE SKY ISLANDS OF THE SONORAN DESERT1

THE EVOLUTION OF DISTYLY FROM TRISTYLY IN POPULATIONS

OF OXALIS ALPINA (OXALIDACEAE) IN THE SKY ISLANDS OF THE

SONORAN DESERT1

STEPHEN G. WELLER,2,6 CESAR A. DOMINGUEZ,3 FRANCISCO E. MOLINA-FREANER,4 JUAN FORNONI,3 AND

GRETCHEN LEBUHN5

2Department of Ecology and Evolutionary Biology, University of California, Irvine, California 92697 USA; 3Departamento de

Ecologıa Evolutiva, Instituto de Ecologıa, Universidad Nacional Autonoma de Mexico, Apartado Postal 70-275, C.P. 04510,

Mexico Distrito Federal, Mexico; 4Departamento de Ecologıa de la Biodiversidad, Estacion Regional del Noroeste, Instituto de

Ecologıa, Universidad Nacional Autonoma de Mexico, Apartado Postal 1354, C.P. 83000, Hermosillo, Sonora, Mexico; and5Department of Biology, San Francisco State University, 1600 Holloway Avenue, San Francisco, California 94132 USA

The evolution of distyly from tristyly was investigated in populations of Oxalis alpina at high elevations throughout the Sky

Islands of the Sonoran Desert. Incompatibility systems in tristylous populations, where self-incompatible short-, mid-, and long-

styled morphs occur in populations, vary from those typical of tristylous species in which each morph is equally capable of

fertilizing ovules of the other two morphs, to breeding systems in which incompatibility relationships are asymmetric. In these

populations, selection against the allele controlling expression of the mid-styled morph is likely. The degree of modification of

incompatibility in the short- and long-styled morphs in 10 populations was strongly associated with fewer mid-styled morphs,

supporting models predicting the effect of these modifications of incompatibility on frequency of the mid-styled morph. Self-

compatibility of the mid-styled morph may be important for maintaining the frequency of this morph, depending on the level of

self-pollination, self-fertilization, and the extent of inbreeding depression. Modifications of incompatibility in tristylous

populations and the distribution of distylous populations of O. alpina in the Sky Island region have similar geographic

components, indicating the potential importance of historical factors in the evolution of distyly from tristyly.

Key words: distyly; heterostyly; illegitimate crosses; legitimate crosses; Oxalis alpina; self-incompatibility; Sky Islands;

tristyly.

Most flowering plant species are hermaphroditic, and manyhave the capacity for self-fertilization. Charles Darwin (1900)first demonstrated the deleterious consequences of self-fertilization and suggested that many reproductive systems inflowering plants have evolved to promote outcrossing (matingswith other individuals in the same population). Understandingthe evolutionary forces that underlie modifications of plantbreeding systems, which themselves may influence geneticdiversity, gene flow, population structure, and the potential forspeciation, is a major challenge. Heterostylous breedingsystems have been useful model systems for addressingquestions about breeding system evolution. In heterostylousbreeding systems, two or three floral morphs occur in apopulation (Barrett, 1992). Because of strong self-incompati-bility, each floral morph is normally incapable of producingseed after self-fertilization. In tristyly, three floral morphs occurin populations (Fig. 1A). Flowers have three levels at whichorgans can occur; stamens occupy two levels, while stigmas

occur in the third. Morphs differ in the levels at which stigmasand anthers occur (Fig. 1A). Compatible matings are generallythose occurring between anthers and stigmas located at thesame level in the flower and were termed legitimate by Darwin(1877). Because there are three floral morphs, there are sixcategories of legitimate pollinations (L 3 l/S, L 3 l/M, etc.; seeFig. 1A for explanation of cross notation) that typicallyproduce seeds. The numerous categories of illegitimate cross-and self-pollinations (L 3 m/S, L 3 s/M, etc.) normally fail toproduce seeds.

The genetic system controlling tristyly, in those cases thathave been investigated, usually consists of two linked orunlinked loci, each with two alleles (Weller, 1976b; Lewis andJones, 1992). Theoretical studies have indicated that thissystem of genetic control leads to equal numbers of the threemorphs in populations (Charlesworth, 1979; Heuch, 1979). Indistylous species, only two morphs occur in populations (Fig.1B); as in the case of tristyly, fertilizations that are capable ofleading to seed production are normally those occurringbetween anthers and stigmas at the same level. Distyly isfound in approximately 25 families of flowering plants(Barrett, 1992). Tristyly is much less common and is foundin only six flowering plants families (Barrett, 1993; Thompsonet al., 1996). Distyly also occurs in four families with tristyly(Amaryllidaceae, Linaceae, Lythraceae, and Oxalidaceae) andis thought to be derived from tristyly in the Lythraceae andOxalidaceae (Weller, 1992).

Theoretical analyses indicate that the most obvious cause forthe loss of floral morphs in a tristylous population is fitnessdifferences among the morphs. Morph-specific fitness disad-vantages may occur when pollinators discriminate among floral

1 Manuscript received 31 May 2006; revision accepted 7 May 2007.

The authors thank W. Yang, R. Basile, and Y. Alliman for plant care; A.

Vuong, L. Duong, and A. Andres for help with the crossing program; A. Sakai

for reading the manuscript and for her help in the greenhouse; S. Barrett for

comments on the manuscript; T. van Devender, P. Jenkins, and R. Felger for

advice on field sites; L. Portillo and R. Jaime for showing them populations of

O. alpina in the Sierra Los Ajos; H. Aguayo, B. Brown, and C. Varela for

access to field sites; the U. S. Forest Service for permits; and R. Perez Ishiwara

for assistance in the field. This research was supported by grants from the

University of California Institute for Mexico and the United States (UC

MEXUS), CONACYT (47858-Q), the National Autonomous University of

Mexico (PAPIIT IN217 803), and the University of California at Irvine.6 Author for correspondence (e-mail: [email protected])

972

American Journal of Botany 94(6): 972–985. 2007.

morphs, when one of the morphs undergoes more self-fertilization and a reduction in fitness of offspring due toinbreeding depression (Charlesworth, 1979) or when theefficiency of pollen transfer is not symmetrical and one ortwo morphs suffer pollen limitation (e.g., Hodgins and Barrett,2006). Similarly, modifications of heterostylous incompatibil-ity allowing production of seeds following illegitimate crossescould lead to loss of a floral morph, assuming the occurrence ofillegitimate pollen flow and pollen limitation (Charlesworth,1979). A much less explored alternative is the loss of floralmorphs in tristylous populations independent of fitnessdifferences among the morphs. These losses may occur ifpatterns of pollen flow and compatibility favor the propagationof one of the morph-determining alleles or if one of thesealleles is more susceptible to drift (Eckert and Barrett, 1992).

We investigated Oxalis alpina to test whether modificationsof tristylous incompatibility relationships have resulted in theevolution of distyly in a number of populations of this species(Weller, 1976a; Charlesworth, 1979). The importance of thesemodifications was suggested by observations of incompatibilityfrom populations of O. alpina in the Sky Islands (or MadreanArchipelago), isolated mountain ranges in the Sonoran Desertthat have coniferous forest at higher elevations. In typical

tristylous species, including tristylous species of Oxalis fromsouthern Mexico (Weller, 1980), only legitimate pollinations(pollen transfer between stamens and stigmas occurring at thesame level) lead to fertilization and seed set (Fig. 1A). Incontrast, in the two populations investigated in the Sky Islands,incompatibility reactions were modified from the typicalcondition of southern Mexican species (Weller, 1976a, 1980).In the short- and long-styled morphs there was no differentiationof the incompatibility reactions of the two anther whorls, andthe two morphs were completely intercompatible (for the long-styled morph, the L 3 m/S and L 3 l/S crosses had equivalentseed production; for the short-styled morph, the S 3 m/L and S3 s/L crosses had equivalent seed production; Fig. 2).

Modifications of the L 3 m/S and S 3 m/L incompatibilityreactions are likely to result in selection against the mid-allelefor two reasons (Weller, 1976a). From the perspective of malefitness, pollen from two stamen whorls of the short-styledmorph, but only a single stamen whorl of the mid-styled morphis capable of fertilizing ovules of long-styled morphs (Fig. 2).Similarly, pollen from two stamen whorls of the long-styledmorph, but only a single stamen whorl of the mid-styled morphis capable of fertilizing ovules of short-styled morph (Fig. 2).Unless the proportion of legitimate pollen flow (S 3 s/L and L3 l/S crosses) is high relative to illegitimate but compatiblepollen flow (S 3 m/L and L 3 m/S crosses), pollen from theshort- and long-styled morphs should outcompete pollen frommid-styled morph (Weller, 1976a). From the perspective offemale fitness, pollen from only two stamen whorls, the mid-stamen whorl of the short- and long-styled morphs, is capableof fertilizing mid-ovules, while pollen from three stamenwhorls is capable of fertilizing ovules of both short- and long-styled morphs (Fig. 2), an approximate 50% advantage for theshort- and long-styled morphs. Hence, the hypothesis predictsan alteration in the pattern of segregation within thosepopulations where incompatibility reactions of the short- andlong-styled morphs have been modified. In populations of O.alpina there is no evidence for reduced fecundity of the mid-styled morph (Weller, 1981a), presumably because pollinatorstransfer adequate pollen for full seed set. These results alsoindicate that it is unlikely that lower fecundity of the mid-styledmorph has resulted in reduced frequency of this morph andconsequent modifications of incompatibility in the short- andlong-styled morphs.

The consequences of loss of incompatibility differentiationfor the evolution of distyly were modeled by DeborahCharlesworth (1979). With relatively high levels of selfingand inbreeding depression, a modifier that resulted in loss ofincompatibility differentiation usually resulted in loss of themid-morph. Self-pollination of the mid-morph was consideredmore likely because stigmas are located between the two antherwhorls (Charlesworth, 1979). Pollen limitation increased thelikelihood of loss of the mid-morph. In contrast, whenlegitimate pollen flow was high (e.g., fewer L 3 mS and S 3mL pollinations), retention of the mid-morph was more likely.Charlesworth (1979) suggested that the ability of mid-levelpollen to function in the L 3 m/S and S 3 m/L crosses shouldbe accompanied by a corresponding failure of M 3 m/S and M3 m/L crosses.

In this paper, we describe variation in incompatibility among13 Sky Island populations of O. alpina and determine howthese differences are associated with style-morph variation. Inparticular, we test the hypothesis that modifications ofillegitimate crosses of the short- and long-styled morphs are

Fig. 1. Incompatibility relationships in (A) a typical tristylous speciesand in (B) a typical distylous species. Only crosses between anthers andstigmas at the same level, termed legitimate by Darwin (1877), are capableof leading to fertilization of ovules and seed production. In (A), L 3 l/Srefers to cross in which L is the long-styled morph used as a female parent,l is the long-anther whorl of the male parent, and S is the short-styledmorph used as a male parent; M 3 m/S refers to cross in which M is themid-styled morph used as a female parent, m is the mid-anther whorl ofthe male parent, and S is the short-styled morph used as the male parent,etc. In (B), cross terminology is the same as for tristylous populations,although in distylous populations, S 3 m/L and L 3 m/S crosses areconsidered legitimate because the mid-morph is absent. Diagrams fromWeller (1976a).

June 2007] WELLER ET AL.—EVOLUTION OF DISTYLY FROM TRISTYLY IN OXALIS 973

responsible for the reduction in frequency of the mid-styledmorph and evolution of distyly, as predicted from earlierstudies (Weller, 1976a, 1986; Charlesworth, 1979). Self-compatibility was investigated because of the possibility thatgreater selfing of the mid-styled morph relative to the short-and long-styled morphs, and expression of inbreedingdepression in the progeny of mid-styled individuals, couldselect against this morph (Charlesworth, 1979). Seed produc-tion of the M 3 m/S and M 3 m/L crosses was related to thedegree of incompatibility modification to determine whetherloss of incompatibility differentiation is associated withreductions of fecundity for legitimate crosses of the mid-styledmorph, as predicted by Charlesworth (1979). The overall goalof this study was to determine how the mid-styled morph inpopulations of O. alpina could be lost, despite apparentfrequency-dependent advantage resulting from the tristylousbreeding system.

MATERIALS AND METHODS

Study species—Oxalis alpina (Rose) Knuth (section Ionoxalis) is awidespread species ranging from Guatemala to the southwestern United States(Denton, 1973). This scapose species grows from a small bulb in response tosummer rains. Sessile bulblets are formed in the leaf axils in populations fromthe Sky Island region. Cytogeographic studies, which have shown that thehaploid chromosome number varies from 7 to 42, suggest that O. alpina is notmonophyletic (Weller and Denton, 1976). In contrast, populations in the SkyIsland region of Arizona, New Mexico, and Sonora are likely to bemonophyletic based on similar morphology, uniform tetraploidy (Weller andDenton, 1976), and production of viable hybrid seed (Weller, 1978).

Style-morph representation—We surveyed 18 tristylous populations andsix distylous populations for style-morph representation in 2001–2004.Populations were defined physically as an area where plants occurred in aspatially contiguous distribution, usually in moister canyons, separated fromother populations by intervening unsuitable habitat (usually dry ridges). In thefield, we defined flowering individuals by their physical separation fromneighboring plants; spatially distinct flowering individuals might representramets of the same genet, although neighboring plants often had different stylemorphs, indicating that clonal growth in O. alpina may be limited. In threetristylous populations, style-morph frequencies were estimated from nonflow-ering bulbs collected in the field and grown in the greenhouse. Severalpopulations sampled in 1977–1979 were also sampled in 2001, 2003, or 2004.We sampled a single population for style-morph frequencies in most mountain

ranges, although more intensive sampling of some mountain ranges was carriedout in earlier studies (Weller, 1979). G tests for heterogeneity of style-morphrepresentation were carried out for 1977–1979 to identify fluctuations in morphfrequencies likely to result from sampling, and for all sampling years (includingsurveys in 2001–2004) to determine the likelihood of changes in morphrepresentation that had occurred over the longer time interval. Equality ofmorph representation among years was investigated using G tests. Samplessizes averaged 387 (range 116–1723) for distylous populations and 395 (range55–1004) for tristylous populations. Sample sizes varied among populations inpart because of variation in total population size.

Incompatibility relationships in populations of O. alpina—We collectedbulbs from 13 tristylous and one distylous population in Arizona, New Mexico,and Sonora for investigation of incompatibility relationships. The distylouspopulation was included in the crossing study to confirm incompatibility resultsobtained for other distylous populations throughout the range of O. alpina(Weller, 1976a). We attempted to analyze incompatibility relationships in atleast one population of O. alpina from each mountain range in the Sonoran SkyIslands. A few inaccessible populations or populations with inadequatepopulation sizes (Sierra Los Pinitos, Sierra El Tigre, Sierra La Madera) werenot included in the study. Bulbs were sampled throughout each population tominimize the chance that genets were represented more than once. Bulbs wereplanted in a soilless mix at the University of California Irvine greenhouse. Allcrosses were made during the summer months when O. alpina responds towater and grows actively. To assess the status of incompatibility in populations,we used approximately 10 individuals of each style morph as seed parents incrosses and an equal or larger number of plants as male parents. In eachtristylous population, there were six legitimate crosses (Fig. 1A), 12 illegitimatecrosses (crosses between individuals where anthers and stigmas occur atdifferent levels), and six illegitimate self-pollinations (both anther whorls ofeach morph used for self-pollinations). For each individual used as a maternalplant, two to five crosses per category were carried out, yielding a minimum of750 crosses per tristylous population. Control flowers (tagged but notpollinated) were used to monitor any background pollination. There wasvirtually no evidence that potential insect pollinators gained access to thegreenhouse, and data from control crosses are not presented. Capsules werecollected 13–15 d following pollination, before explosive dehiscence. Theaverage number of viable seeds per capsule was used as a measure ofincompatibility following a cross.

Statistical analysis of crossing results—Preplanned comparisons (SASInstitute, 1989) were used to test whether the S 3 m/L and L 3 m/S crosses hadthe same level of compatibility as legitimate crosses with short- and long-styledmorphs as female parents (S 3 s/L, S 3 s/M, L 3 l/S, L 3 l/M). Loss ofincompatibility differentiation between the mid- and long-anther whorls of theshort-styled morph and the short- and mid-anther whorls of the long-styledmorph was indicated when the preplanned contrast was not significant; asignificant contrast indicated either partial or no modification of incompatibilitydifferentiation. The contrast between these legitimate and illegitimate crosses iscritical because of previous work indicating that high compatibility of the S 3

m/L and L 3 m/S crosses is likely to favor loss of the mid-styled morph(Weller, 1976a; Charlesworth, 1979). Preplanned contrasts were also used todetermine whether the S 3 m/L and L 3 m/S crosses had greater seed set thanthe remaining illegitimate crosses (S 3 l/M, S 3 m/S, S 3 l/S, M 3 l/S, M 3 s/L,M 3 s/M, M 3 l/M, L 3 s/M, L 3 m/L, L 3 s/L). Lack of significance indicatedthat incompatibility of the S 3 m/L and L 3 m/S crosses was similar to the otherillegitimate cross categories; a significant value for the contrast indicated someloss of incompatibility differentiation for the S 3 m/L and L 3 m/S crosses.Results from the two preplanned contrasts for each population were used tocharacterize tristylous incompatibility populations as having (1) retainedancestral tristylous incompatibility (contrast of four legitimate and twoillegitimate crosses significant; contrast of S 3 m/L and L 3 m/S crosseswith remaining 10 illegitimate crosses insignificant), (2) partial loss ofincompatibility differentiation (contrast of four legitimate and two illegitimatecrosses significant; contrast of S 3 m/L and L 3 m/S crosses with remaining 10illegitimate crosses also significant), or (3) complete loss of incompatibilitydifferentiation (contrast of four legitimate and two illegitimate crossesinsignificant; contrast of S 3 m/L and L 3 m/S crosses with remaining 10illegitimate crosses significant).

All legitimate crosses were compared using Tukey’s post hoc tests (SASInstitute, 1989) to detect potential differences in viable seeds per capsule andespecially modifications of mid-morph fecundity. Seed production following

Fig. 2. Modified incompatibility relationships in a population ofOxalis alpina from Morse Canyon in the Chiricahua Mts. (based on datafrom Weller, 1976a). Two illegitimate crosses, S 3 m/L and L 3 m/S(solid lines), led to seed production as high as those of the legitimate S 3 s/L and L 3 l/S crosses (dotted lines).

974 AMERICAN JOURNAL OF BOTANY [Vol. 94

self-fertilization was also compared using post hoc analyses to determinewhether some morphs, particularly the mid-styled morph, have evolveddetectable self-compatibility. Values of seed production for all cross categorieswere tested for deviation from zero using t tests. Potential differences in midself-compatibility were compared for populations where incompatibilitydifferentiation was absent, partial, or complete with a one-way ANOVA onuntransformed seed production data. All statistical analyses were carried outusing SAS version 9.1 (SAS Institute, 2002–2003).

Loss of incompatibility differentiation and fecundity of the M 3 m/S andM 3 m/L crosses—We used regression to investigate the relationship betweenrelative seed production of the legitimate M 3 m/S and M 3 m/L crosses to lossof incompatibility differentiation. Relative seed production of the M 3 m/S andM 3 m/L crosses was calculated by dividing the value for each of these crossesby the average value for seeds per capsule for the legitimate crosses with short-and long-styled morphs as seed parents. This analysis was carried out for all 13tristylous populations.

Loss of incompatibility differentiation and frequency of the mid-morph—The potential effect of the degree of incompatibility differentiationin the short- and long-styled morphs on the frequency of the mid-styled morphin populations was investigated by dividing the mean number of seeds percapsule for the L 3 m/S cross by the mean number of seeds per capsule for theL 3 l/S cross. This ratio approaches 1 as the incompatibility between the Lstigma and style and the m/S pollen is lost. Similarly, S 3 m/L values weredivided by S 3 s/L values. A value indicating the overall extent ofincompatibility differentiation was obtained for each population by averagingvalues for the short- and long-styled morphs, with higher values indicatinggreater loss of incompatibility differentiation. Regression of mid-frequenciesagainst these values was used to determine whether loss of incompatibilitydifferentiation is associated with reduced mid-frequency. Only those popula-tions with field surveys of style-morph frequencies were used in this analysis.

RESULTS

Style-morph frequencies—In the Sky Island region, distylyis apparently restricted to populations of O. alpina occurring inneighboring mountain ranges in Arizona (Table 1, Fig. 3) andthe Chiricahua Mts., where distylous populations occur at thenorthern end of the range and tristylous populations at thesouthern end of the range (Weller, 1979). Distylous popula-tions had equal numbers of short- and long-styled style morphs(defined as isoplethic by Heuch, 1979) in four of the sixpopulations surveyed in 2001–2004 (Table 1). Amongdistylous populations, the greatest deviations from isoplethyoccurred in the Sierra Ancha Mts., the most northern Arizonamountain range where O. alpina is found. In this mountainrange style morphs were found in large patches, possiblybecause of clonal growth. While tristylous populations inSonora were more likely to have equal numbers of short-, mid-,and long-styled morphs, style-morph frequencies deviatedgreatly from isoplethy in the more northern tristylouspopulations (Table 1). Populations in the Chiricahua Mts.(Morse Canyon, Fly Peak, South Fork Canyon, and Green-house Trail), near the Black River, and in the Pinos Altos Mts.had reduced frequencies of mid-styled morphs. Excesses of thelong-styled morph were characteristic of anisoplethic popula-tions (Table 1).

Distylous populations sampled across years had considerableheterogeneity in style-morph representation, despite theoccurrence of isoplethy in most years. In two tristylouspopulations from the Huachuca Mts., the mid-styled morphremained at approximately the same frequency over a 23-yrperiod, although in one of these populations (Miller Canyon),the numbers of short- and long-styled morphs fluctuatedsubstantially. In two of the four populations in the Chiricahua

Mts., the frequency of the mid-styled morph dropped duringthis time interval (Table 1), while in the remaining twopopulations there was no significant heterogeneity across years.The frequency of the mid-morph appeared to decline in apopulation from the White Mts. sampled in 1977 and 2003.

Incompatibility relationships—Seed production followingcrosses within a distylous population from the Santa CatalinaMts. (voucher 706) was high for legitimate crosses (L 3 l/S, L3 m/S, S 3 s/L, and S 3 m/L in a distylous species; Fig. 1A)and very low for illegitimate crosses and selfs (Fig. 4A). Inpreplanned comparisons, there were no differences in seedproduction between the L 3 l/S and S 3 s/L crosses vs. the L 3m/S and S 3 m/L crosses (F ¼ 0, df ¼ 1, P ¼ 0.99). Post hoccomparisons indicated that short- and long-styled morphsproduced similar numbers of seeds (F¼ 1.76, df¼ 3, 37, P ¼0.17; Fig. 4A). Based on preplanned comparisons, seedproduction was much higher for the S 3 m/L and L 3 m/Scrosses than for illegitimate crosses (S 3 m/S, S 3 l/S, L 3 m/Land L 3 s/L in a distylous species), in which seed productiondid not deviate significantly from zero (F¼ 16.63, df¼ 1, P ,0.0001; Fig. 4A). Seed production following self-pollinationsdid not deviate significantly from zero.

Tristylous populations varied greatly in incompatibilityrelationships (Fig. 4B–N). Illegitimate crosses of the short-and long-styled morphs most likely to produce seeds were theS 3 m/L and L 3 m/S crosses. Illegitimate crosses of the mid-styled morph were more likely to have seed productionsignificantly greater than zero (39% of all crosses acrosspopulations) than illegitimate crosses involving either theshort- or long-styled morphs (10% of all crosses). Based onvariation in seed production of the illegitimate S 3 m/L and L3 m/S crosses, populations can be divided into three groups.The first group includes those populations, all in northernSonora (Fig. 4H, Sierra San Luis, Sonora, 958; Fig. 4J, SierraLa Mariquita, Sonora, 960; and Fig. 4K, Sierra Azul, Sonora,966) in which preplanned comparisons indicated that the fourlegitimate crosses using short-styled and long-styled morphshad significantly higher seed production than the twoillegitimate S 3 m/L and L 3 m/S crosses (Table 2a). In thesepopulations the S 3 m/L and L 3 m/S crosses also had the samelow level of seed production as the remaining illegitimatecrosses (preplanned comparisons, Table 2b).

In a second group that includes populations in Arizona andNew Mexico and one population from northern Sonora, loss ofincompatibility differentiation in the short- and long-styledmorphs appears to be complete; the illegitimate S 3 m/L and L3 m/S crosses have the same seed production as legitimatecrosses with short- and long-styled morphs as seeds parents(Table 2a, Fig. 4C, Pinos Altos Mts., New Mexico, 971; Fig.4D, Morse Canyon, Chiricahua Mts., Arizona, 727; Fig. 4E,Animas Mts., New Mexico, 973; and Fig. 4L, Sierra Los Ajos,Sonora, 967) and significantly greater seed production than theremaining illegitimate crosses (Table 2b). Incompatibility ofthe short- and long-styled morphs in these populationsresembles incompatibility in distylous populations.

Partial loss of incompatibility differentiation occurs in a thirdgroup of populations scattered throughout the Sky Islands (Fig.4B, White Mts., Arizona, 713; Fig. 4F, Miller Canyon,Huachuca Mts., Arizona, 702; Fig. 4G, Sierra San Jose,Sonora, 959; Fig. 4I, Sierra La Elenita, Sonora 956; Fig. 4M,Sierra Buenos Aires, Sonora, 961; Fig. 4N, Sierra La Purica,Sonora, 968). In these populations, the S 3 m/L and L 3 m/S


TABLE 1. Style-morph frequencies in distylous and tristylous populations of Oxalis alpina in Arizona and Mexico. For populations sampled acrossmultiple years, probabilities for significant heterogeneity are presented. Tests for equality of style-morph representation (isoplethy) are also presented.Populations are ordered geographically starting with populations in northwestern Arizona (AZ) (all distylous) to populations located in eastern andsoutheastern Arizona, New Mexico (NM), and Sonora (SON), Mexico (see Fig. 3).

Locality Voucher no. Year of survey

Style form frequency (%)

N Probability of heterogeneity Test for isoplethy G (P value)Short Mid Long

Distylous populations:Sierra Ancha Mts., AZ 781 1977 9 — 91 116 ,0.0001 92.67 (,0.0001)

2003 61 — 39 284 14.00 (0.0002)Pinal Mts., AZ 972 2003 60 — 40 85 3.42 (0.0642)Nuttall Ridge, Pinaleno Mts., AZ 954 1977 41 — 59 258 0.0151 8.25 (0.0041)

2001 53 47 196 0.51 (0.4751)Santa Catalina Mts., AZ 706 1977 57 — 43 137 0.2059 2.64 (0.1042)

2001 50 — 50 212 0.00 (1.000)Santa Rita Mts., AZ 699 2001 53 — 47 362 1.11 (0.2921)Barfoot Park, Chiricahua Mts., AZ 785 1978 48 — 52 400 0.0112 0.49 (0.4839)

2001 58 — 42 277 7.34 (0.0067)Crest Trail, Chiricahua Mts., AZ 776 1977 51 — 49 200 0.7117 0.08 (0.7773)

1978 48 — 52 486 0.2815 (including 2001) 1.19 (0.2753)1979 49 — 51 1723 1.51 (0.21912001 50 — 50 385 0.02 (0.8875)

Tristylous populations:White Mts., AZ 713 1977 15 36 49 204 0.0086 40.0 (,0.0001)

2003 25 26 49 210 20.8 (,0.0001)Black River, AZ 771 1977 40 11 49 204 0.2967 58.2 (,0.0001)

771 2003 35 15 50 407 82.2 (,0.0001)Pinos Altos Mts., NM 971 2003 38 21 41 309 24.4 (,0.0001)Morse Canyon, Chiricahua Mts., AZ 727 1977 23 30 47 200 0.0789 17.8 (0.0001)

1978 31 22 47 1004 0.1317 (including 2001) 89.6 (,0.0001)1979 30 25 45 1003 63.2 (,0.0001)2001 27 25 48 508 44.7 (,0.0001)

Fly Peak, Chiricahau Mts., AZ 777 1977 23 30 47 325 0.0027 45.1 (,0.0001)1978 43 11 46 396 ,0.0001 (including 2001) 112.1 (,0.0001)1979 46 11 43 843 231.4 (,0.0001)2001 50 ,1 49 850 653.6 (,0.0001)

South Fork Canyon, Chiricahua Mts., AZ 779 1977 42 2 56 201 0.3745 34.9 (,0.0001)1978 44 4 52 194 0.1937 (including 2001) 102.8 (,0.0001)1979 51 2 47 55 36.1 (,0.0001)2001 51 3 46 119 67.3 (,0.0001)

Greenhouse Trail, Chiricahua Mts., AZ 789 1978 44 4 52 508 0.0863 284.9 (,0.0001)1979 38 6 56 500 ,0.0001 (including 2001) 245.0 (,0.0001)2001 45 ,1 54 445 341.1 (,0.0001)

Animas Mts., NM 973 2004 27 29 44 107 2.71 (0.2579)Miller Canyon, Huachuca Mts., AZ 702 1978 26 24 50 796 0.0004 90.2 (,0.0001)

1979 18 25 57 672 ,0.0001 (including 2001) 169.8 (,0.0001)2001 29 28 43 259 10.2 (0.0061)

Carr Peak, Huachuca Mts., AZ 787 1978 15 31 54 474 0.0453 109.7 (,0.0001)2001 13 39 48 389 90.0 (,0.0001)

Sierra San Jose, SON 959 2004 22 24 53 92 15.4 (0.0005)Sierra San Luis, SON 958 2001 46 21 34 111 10.43 (0.0055)Sierra La Elenita, SON 956 2001 39 32 29 109 1.43 (0.4892)Sierra La Mariquita, SON 960 2003 34 36 30 308 2.26 (0.3230)Sierra Azul, SON 966 2003 33 34 33 123 0.05 (0.9753)Sierra Los Ajos, SON 967 2003 34 33 33 259 0.05 (0.9753)Sierra Buenos Aires, SON 961 2004 33 33 34 300 0.06 (0.9704)Sierra La Purica, SON 968 2003 22 30 48 939 10.3 (0.0058)

Notes: Style-morph frequencies were sampled from plants flowering in the field with three exceptions (Animas Mts., Sierra La Elenita, and Sierra SanLuis), where style-morph frequencies were based on plants grown in the greenhouse from bulbs collected in the field either before or after flowering. Testsfor heterogeneity were first carried out for surveys in 1977–1979 (when multiple years were sampled) and then also including data from 2001–2003. Alltests for isoplethy were significant after a sequential Bonferroni correction using morph ratios from the most recent population surveys.

!Fig. 3. Distribution of tristylous and distylous populations of Oxalis alpina in the Sky Island region of the Sonoran Desert. Frequencies of short-, mid-,

and long-styled morphs are based on the most recent surveys for each population. For several populations, limited field observations or herbarium recordsindicate the presence of tristylous (T) or distylous (D) populations. The distribution of oak woodland and coniferous forest follows Marshall (1957) andMcLaughlin (1995), with modifications based on our field observations.



crosses produced significantly less seed than the legitimate

crossing categories (Table 2a; contrasts of four legitimate vs.

two illegitimate crosses were significant) but produced

significantly more seed than the other categories of illegitimate

crosses (Table 2b; contrasts of seed production for the S 3 m/L

and L 3 m/S crosses vs. the remaining illegitimate cross

categories were significant). The extent of modification varied

widely among these populations with partially modified

incompatibility differentiation. Populations with partially

modified incompatibility located in the White Mts. and Sierra

San Jose (Fig. 4B, 4G) closely resembled nearby populations

where differentiation of incompatibility was complete (Pinos

Fig. 4. Seed production per capsule following legitimate and illegitimate cross- and self-pollinations in 14 populations (A–N) of Oxalis alpina. Foreach cross, mean values are based on averages by maternal parent. Bars represent SE. Asterisks indicate that values of seed production differ significantlyfrom zero. Bars with identical lower case letters indicate either legitimate pollinations or self-pollinations that do not differ significantly in seed production.Collection numbers are vouchers for populations located at UC and US. (A) Distylous population from the Santa Catalina Mts., AZ (706). (B) Tristylouspopulation from the White Mts., AZ (713). (C) Tristylous population from the Pinos Altos Mts., NM (971). (D) Tristylous population from Morse Canyonin the Chiricahua Mts., AZ (727). (E) Tristylous population from the Animas Mts., NM (973). (F) Tristylous population from Miller Canyon in theHuachuca Mts. (702). (G) Tristylous population from the Sierra San Jose, Sonora (959). (H) Tristylous population from the Sierra San Luis, Sonora (958).(I) Tristylous population from the Sierra La Elenita, Sonora (956). (J) Tristylous population from the Sierra La Mariquita, Sonora (960). (K) Tristylouspopulation from the Sierra Azul, Sonora (966), (L) Tristylous population from the Sierra Los Ajos, Sonora (967). (M) Tristylous population from the SierraBuenos Aires, Sonora (961). (N) Tristylous population from the Sierra La Purica, Sonora (968).


Fig. 4. Continued.


Fig. 4. Continued.


Fig. 4. Continued.


Altos Mts. and Chiricahua Mts; Fig. 4C, 4D, respectively),while in several Sonoran populations (Sierras La Elenita,Buenos Aires, and La Purica; Fig. 4I, 4M, and 4N,respectively), modifications of incompatibility were slight,and incompatibility relationships were similar to those Sonoranpopulations with unmodified tristylous incompatibility rela-tionships. Overall, loss of incompatibility differentiation wasmost apparent in northern populations of O. alpina located inArizona and New Mexico (Fig. 3).

Variation in self-incompatibility in populations of O.alpina—Self-compatibility varied substantially among popu-lations of O. alpina, with a tendency for mid-styled morphs tohave greater self-compatibility than the short- or long-styledmorphs, especially in those populations with partial orcomplete loss of incompatibility differentiation (Table 2d;Fig. 4A–N). Significant post hoc contrasts indicated higherself-compatibility of the mid-styled vs. short- and long-styledmorphs for Morse Canyon, Chiricahua Mts. (727), Sierra SanJose (959), and the Pinos Altos Mts. (971), populations wheremodification of incompatibility differentiation is either com-plete (727, 971) or nearly complete (959). Mid-styled morphsfrom all four populations with complete loss of incompatibilitydifferentiation had significant seed production followingselfing (Fig. 4C–E and 4L). Three of six populations withpartial loss of incompatibility differentiation had significantself-compatibility in the mid-styled morph (Fig. 4B, 4G, and

4M), and one of three populations possessing full incompat-ibility differentiation had significant seed production followingselfing of mid-styled morphs (Fig. 4K). When self-compati-bility of the mid-styled morph, measured as the average of seedproduction for the M 3 l/M and M 3 s/M self-fertilizations,was compared for populations with unmodified, partiallymodified, and fully modified incompatibility, there were nosignificant differences (F¼0.30; df¼2, 10, P¼0.7464). Thereappeared to be little correlation between seed production of theM 3 l/M and M 3 s/M self-pollinations within mid-styledindividuals (Fig. 4).

Modification of fecundity of the M 3 m/S and M 3 m/Lcrosses—Loss of incompatibility differentiation was associatedwith modification of seed production of the legitimate M 3 m/Scross, as indicated by a significant regression of relativefecundity of the M 3 m/S cross (seed production of M 3 m/Sdivided by the average fecundity of legitimate crosses of theshort- and long-styled morphs) against a measure of loss ofincompatibility differentiation (F¼ 8.26; df¼ 1, 11; P¼ 0.015;Fig. 5). Populations with the most pronounced loss ofincompatibility differentiation of the short- and long-styledmorphs had the lowest relative seed production for the M 3 m/S cross (Table 2c). This trend was most apparent in populationsfrom the White, Pinos Altos, Chiricahua (Morse Canyon), andAnimas Mts. The relationship of M 3 m/L relative fecundity to

TABLE 2. Summary of results from preplanned comparisons and post hoc contrasts following crosses within tristylous populations of Oxalis alpina fromArizona, New Mexico, and Sonora (localities listed in Table 1). Analyses include (A) preplanned contrasts of the four legitimate crosses (S 3 s/L, S 3s/M, L 3 l/M, and L 3 l/S) vs. the illegitimate S 3 m/L and L 3 m/S crosses, (B) preplanned contrasts of illegitimate S 3 m/L and L 3 m/S crosseswith remaining illegitimate crosses, (C) post hoc comparisons among legitimate crosses using Tukey’s contrasts, and (D) post hoc comparisons ofseed production following self-pollination using Tukey’s contrasts.

Population, state, voucher no.

A) Preplanned contrast(F, P) of 4 legitimate crosses

vs. 2 illegitimate crosses

B) Preplanned contrast(F, P) of S 3 m/L and

L 3 m/S crosseswith remaining 10illegitimate crosses

C) ANOVA for alllegitimate crosses (F, P)and significant post hoc

contrasts (P , 0.05)

D) ANOVA for allself-pollinations (F, P)

and significant post hoccontrasts (P , 0.05)

E) Summary of extent ofmodification of incompatibility

differentiation

Morse, Chiricahua Mts., AZ 727 0.33, 0.5660 27.56, ,0.0001 2.81, 0.0225 8.61, ,0.0001 Complete modificationmids low M 3 s/M high

San Luis, SON 958 60.47, ,0.0001 0.81, 0.3714 0.73, 0.6033 0.60, 0.6966 No modificationLa Elenita, SON 956 19.93, ,0.0001 6.86, 0.0100 1.44, 0.2180 1.31, 0.2713 Partial modificationMiller, Huachuca Mts., AZ 702 6.41, 0.0143 19.74, ,0.0001 0.69, 0.6314 0.60, 0.6966 Partial modificationSan Jose, SON 959 17.79, ,0.0001 72.27, ,0.0001 11.41, ,0.0001 10.77, ,0.0001 Partial modification

M 3 m/S low M 3 s/M, M 3 l/M highLa Purica, SON 968 40.13, ,0.0001 37.81, ,0.0001 1.17, 0.3301 2.97, 0.0184 Partial modification

L 3 mL highLos Ajos, SON 967 2.62, 0.1099 47.57, ,0.0001 3.07, 0.0145 2.05, 0.0850 Complete modification

longs highAzul, SON 966 54.57, ,0.0001 0.22, 0.6419 0.02, 0.9998 1.63, 0.1653 No modificationBuenos Aires, SON 961 6.73, ,0.0001 6.47, 0.0126 0.30, 0.9118 2.54, 0.0398 Partial modificationLa Mariquita, SON 960 16.61, ,0.0001 0.06, 0.8098 0.81, 0.5455 0.49, 0.7792 No modificationAnimas, NM 973 1.82, 0.1202 6.16, ,0.0001 4.49, 0.0014 1.42, 0.2397 Complete modification

M 3 m/S lowWhite, AZ 713 9.24, 0.0035 110.64, ,0.0001 0.73, 0.6033 2.21, 0.0662 Partial modification

Pinos Altos, NM 971 1.62, 0.1655 19.35, ,0.0001 7.80, ,0.0001 9.03, ,0.0001 Complete modificationmids low M 3 l/M high

Notes: Retention of typical tristylous incompatibility relationships is indicated when the S 3 m/L and L 3 m/S crosses have low seed production relativeto the legitimate crosses with short- and long-styled morphs as seed parents (significant F values for preplanned comparisons in column A), and when the S3 m/L and L 3 m/S crosses have similar levels of seed production compared to the remaining illegitimate crosses (nonsignificant F values for preplannedcomparisons in column B). Evolution of incompatibility relationships resembling distyly is indicated by insignificant F values in column A and significantF values for column B. Partial loss of incompatibility differentiation is indicated when F values are significant for (A) and (B). Although a significant Fvalue was obtained for the Buenos Aires population when seed production following self-pollination was compared, the post hoc comparison was notsignificant.


incompatibility modification was not significant (F¼1.96, df¼1, 11; P ¼ 0.189).

Relationship between incompatibility and style morph-frequencies—Complete loss of incompatibility differentiationin the short- and long-styled morphs in the Chiricahua (MorseCanyon), Los Ajos, Animas, and Pinos Altos populations wasassociated with reduced frequencies of the mid-styled morph(average mid-frequency¼ 20.5%; N¼ 4, using a mean value ofmid-styled morph frequency in the Chiricahua Mts. of 7.5%averaged over all tristylous populations) relative to thosepopulations with partial loss of incompatibility differentiation(average mid-styled morph frequency ¼ 28.2%; N ¼ 6), orunmodified incompatibility differentiation (average mid-styledmorph frequency¼ 35%; N¼ 3). There was a strong negativerelationship between loss of incompatibility differentiation ofthe short- and long-styled morphs and the frequency of themid-styled morph (Fig. 6; F ¼ 12.7; df¼ 1, 8; P ¼ 0.0074).

DISCUSSION

Variation in incompatibility relationships in O. alpina andselection against the mid-allele—Incompatibility relationshipsin O. alpina show remarkable diversity over short geographicdistances, and differences among populations in the extent ofincompatibility differentiation between the stamen whorls ofthe short- and long-styled morphs were strongly related to thefrequency of mid-styled morphs. Loss of incompatibilitydifferentiation is expected to result in selection against themid allele (Fig. 2) (Weller, 1976a, 1986; Charlesworth, 1979),and in populations of O. alpina from the Sky Islands, the mid-

styled morph occurs in significantly lower frequencies inpopulations with greater seed production for S 3 m/L and L 3m/S crosses (Fig. 4). Reduction in frequency of the mid-styledmorph most likely results from modifications of incompatibil-ity in the short- and long-styled morphs, rather than fromreduced fitness of the mid-styled morph because in previousstudies (Weller, 1981a), no differences in seed production weredetected among morphs in the field. In most populations withmodified incompatibility, the S 3 m/L and L 3 m/S crosses hadsimilar seed production, although in several cases (e.g., SierrasLos Ajos, Buenos Aires, and La Purica; Fig. 4D) the fecunditywas substantially greater for the L 3 m/S cross than for the S 3m/L cross. The regression of mid-morph frequency againstincompatibility modification assumes that each populationrepresents an independent data point, although we do not knowthe phylogeographic history of these populations.

Geographic patterns to breeding system modification in O.alpina—Oxalis alpina was probably continuously distributedthroughout the Sky Island region during the Pleistocene pluvialperiod because packrat–midden studies confirm that coniferousforests occurred at lower elevations during this period (VanDevender, 1977; Van Devender and Spaulding, 1979).Isolation of O. alpina to separate mountain ranges probablyoccurred following the Pleistocene, when drying caused theretreat of forests and species like O. alpina to higher elevations.The current distribution of distylous and tristylous populationscould have been established before isolation of O. alpinapopulations or after populations became disjunct on isolatedmountain ranges at the end of the pluvial period. The restrictionof distylous populations of O. alpina to mountain ranges in thenorthwestern part of the Sky Islands (Fig. 3) may indicate that asingle transition to distyly occurred in this region, prior toisolation of populations by post-Pleistocene drying. Popula-tions throughout the Sky Islands have probably been

Fig. 6. Regression of mid-styled morph frequency against the extent ofloss of incompatibility differentiation in 10 populations of Oxalis alpinasampled for style-morph frequency in the field. Higher values on the x axisindicate tristylous incompatibility more closely resembling incompatibilitytypical of distylous populations, when the S 3 s/L and S 3 m/L crosses aswell as the L 3 l/S and L 3 m/S crosses have equivalent seed production.The regression equation is y ¼�12.3x þ 35.4 (F ¼ 12.7; df ¼ 1, 8; P ¼0.0074).

Fig. 5. Regression of relative seed production of the M 3 m/S crossagainst the extent of loss of incompatibility differentiation in 13populations of Oxalis alpina. In each population, relative seed productionof the M 3 m/S cross was calculated by dividing seed production for the M3 m/S cross by the average value of seed production of the remaininglegitimate crosses. Higher values on the x axis indicate tristylousincompatibility more closely resembling incompatibility typical ofdistylous populations, when the S 3 s/L and S 3 m/L crosses as well asthe L 3 l/S and L 3 m/S crosses have equivalent seed production. Theregression equation is y ¼�0.803x þ 1.03 (F ¼ 8.26; df ¼ 1, 11; P ¼0.015).


genetically isolated from one another because drying occurredat the end of the Pleistocene pluvial period. Several studies ofother organisms in this region (e.g., Aquilegia, Strand et al.,1996; jumping spiders, Masta, 2000) have used phylogeo-graphic approaches and concluded that gene flow is limitedbetween these isolated mountain ranges. Phylogeographicinformation would be useful for determining how manytransitions to distyly have occurred in populations of O. alpinain the Sky Islands region.

The concentration of distylous populations in the northwest-ern part of the Sky Islands suggests that environmentaldifferences may be important in explaining modifications ofheterostylous reproductive systems in O. alpina. Despite thegeographic pattern of breeding system variation, however, noobvious environmental gradients or differences in pollinatorsare associated with the transition to distyly (Weller, 1981b; F.Baena [Departamento de Ecologıa Evolutiva, Instituto deEcologıa, Universidad Nacional Autonoma de Mexico], C.Domınguez, J. Fornoni, F. Molina-Freaner, S. Weller,unpublished observations). At present, we do not know whymodifications of incompatibility are more common in thenorthern Sky Island populations of O. alpina.

Changes in style-morph frequencies over time—Compar-isons of style-morph frequencies over a 24–26 yr periodindicate apparent stability in the frequency of the mid-morph,with the exception of several populations from the ChiricahuaMts., as well as a population from the White Mts., where thefrequency of the mid-morph declined significantly. Full orpartial loss of incompatibility differentiation is characteristic ofthese populations, indicating that modified incompatibilityrelationships may have resulted in decline of the mid-morph.Distylous populations fluctuated less in style-morph frequen-cies over the sampling interval, with the exception of the SierraAncha population, where clonal growth appears to predominateand leads to substantial differences in style-morph representa-tion, depending on the area sampled (S. Weller, C. Domınguez,and F. Molina-Freaner, unpublished observations).

Fitness of the mid-morph—Charlesworth (1979) suggestedthat in populations with modified incompatibility of the short-and long-styled morphs, the ability of mid-level pollen fromthese morphs to function on long and short stigmas would beinversely associated with function on mid stigmas. Reductionin fertility of the M 3 m/S cross in populations of O. alpinawith loss of incompatibility differentiation is consistent withprevious observations (Weller, 1976a), although there is noapparent loss of function of mid-level pollen from the long-styled morph on mid stigmas. High levels of natural pollination(Weller, 1981a), as well as excesses of the long-styled morphserving as pollen donor to the mid-styled morph, may explainthe absence of differences in seed production between themorphs of O. alpina despite low seed production following M3 m/S crosses.

Differences in seed production of the M 3 m/S and M 3 m/Lcrosses also explain the results of direct tests for the effects ofloss of incompatibility modification on mid-styled morphfrequency. Progeny raised from naturally pollinated plants inChiricahua populations were expected to show deficits of themid-styled morph because of pollen competition that favoredthe short- and long-styled morphs (Weller, 1986). Contrary tothis prediction, mid-styled morph excesses were found,particularly in the progeny of mid-styled morphs, which

produced almost no short-styled morphs in their offspring(Weller, 1986). This unexpected result can now be understoodin light of reduced seed production following the M 3 m/Scross in the Morse Canyon population in the Chiricahua Mts.(Fig. 4). Because the S allele is epistatic to the M allele (Weller,1976b), if the M 3 m/S cross produces few seeds, most mid-styled offspring will result from M 3 m/L crosses, and the Mallele is more likely to be expressed in these progeny. Of thefour populations with complete loss of incompatibilitydifferentiation (Chiricahua, Los Ajos, Animas, and PinosAltos), the M 3 m/S cross had low seed production in all butthe Los Ajos population. Because rejection of pollen fromshort-styled morph parents would increase the frequency ofmid-styled morphs in the progeny of mid-styled morphs, thismechanism may have been favored in mid-styled morphs thatare under negative selection as a result of incompatibilitymodifications in short- and long-styled morphs.

Significant self-fertility was observed for a number ofpopulations, and half of these cases involved self-fertility of themid-styled morph. Self-fertility of the mid-styled morph mightresult in selection against this morph, depending on the selfingrate and the extent of inbreeding depression (Charlesworth,1979). Alternatively, self-fertility of the mid-styled morphmight be advantageous in the face of modifications ofincompatibility in short-styled and long-styled morphs thatplace mid-styled morphs at a disadvantage during outcrossing.Assuming that inbreeding depression was minimal, selfingwould favor the mid-styled morph because of the occurrence ofonly mid- and long-styled morphs among progeny derivedfrom selfing. Preliminary data suggest that self-pollination ofmid-styled morphs is substantial (F. Baena, C. Domınguez, J.Fornoni, F. Molina-Freaner, S. Weller, unpublished observa-tions), although the degree of natural self-fertilization of mid-styled morphs is unknown. We are investigating selfing ratesand the extent of inbreeding depression in populations of O.alpina.

Distribution of breeding systems in Oxalis sectionIonoxalis—The distribution of tristylous and distylous popu-lations in the Sky Island mirrors the overall distribution ofheterostyly in section Ionoxalis, with tristylous species foundin southern Mexico and distylous species found morecommonly in northern Mexico and the United States. Withinsection Ionoxalis, most tristylous species are diploid and haverestricted ranges, while distylous species have broaderdistributions and higher ploidy levels, an overall patternsuggesting that distyly is derived from tristyly (Weller andDenton, 1976). Within O. alpina, which occurs fromGuatemala to New Mexico, ploidy levels are diverse, but theyare not correlated with the distribution of distylous andtristylous breeding systems. In the Sky Islands region, allpopulations have the same ploidy level (n ¼ 14). The closestpopulations outside of Sonora and Chihuahua with the samechromosome number as the Sky Islands populations occur indistylous populations in central Mexico. North of the SkyIslands region in New Mexico, populations of O. alpina arehexaploid (n¼ 21) and distylous. At present, the evolutionaryrelationships among populations of O. alpina throughout itsrange are poorly understood, although breeding-system labilitywithin O. alpina is largely confined to the Sky Islands region.Additional studies, particularly those that model the effects ofall modifications of self-incompatibility (J. Fornoni, C.Domınguez, F. Molina-Freaner, and S. Weller, unpublished


data) and investigate the extent of self-pollination, self-fertilization, and inbreeding depression in progeny of mid-styled morphs, will be particularly useful for understanding thedynamics of style-morph representation in populations of O.alpina.

LITERATURE CITED

BARRETT, S. C. H. 1992. Evolution and function of heterostyly. Springer-Verlag, Berlin, Germany.

BARRETT, S. C. H. 1993. The evolutionary biology of tristyly. In D.Futuyma and J. Antonovics [eds.], Oxford surveys in evolutionarybiology, vol. 9, 283–326. Oxford University Press, Oxford, UK.

CHARLESWORTH, D. 1979. The evolution and breakdown of tristyly.Evolution 33: 486–498.

DARWIN, C. 1877. The different forms of flowers on plants of the samespecies. Murray, London, England.

DARWIN, C. 1900. The effects of cross and self fertilisation in the vegetablekingdom, 2nd ed. Murray, London, UK.

DENTON, M. F. 1973. A monograph of Oxalis, section Ionoxalis(Oxalidaceae) in North America. Publications of the Museum.Michigan State University Biological Series 4: 455–615.

ECKERT, C. G., AND S. C. H. BARRETT. 1992. Stochastic loss of style morphsfrom populations of tristylous Lythrum salicaria and Decodonverticillatus (Lythraceae). Evolution 46: 1014–1029.

HEUCH, I. 1979. Equilibrium populations of heterostylous plants.Theoretical Population Biology 15: 43–57.

HODGINS, K. A., AND S. C. H. BARRETT. 2006. Female reproductive successand the evolution of mating-type frequencies in tristylous populations.New Phytologist: 171: 569–580.

LEWIS, D., AND D. A. JONES. 1992. The genetics of heterostyly. In S. C. H.Barrett [ed.], Evolution and function of heterostyly, 129–150.Springer-Verlag, Berlin, Germany.

MARSHALL, J. T. JR. 1957. Birds of pine–oak woodland in southernArizona and adjacent Mexico. Pacific Coast Avifauna Number 32.Cooper Ornithological Society, Berkeley, California, USA.

MASTA, S. E. 2000. Phylogeography of the jumping spider Habronattuspugillis (Aranae: Salticidae): recent vicariance of Sky Islandpopulations? Evolution 54: 1699–1711.

MCLAUGHLIN, S. 1995. An overview of the flora of the Sky Islands,southeastern Arizona: diversity, affinities, and insularity. In L. F.DeBano, G. J. Gottfied, R. H. Hamre, C. B. Edminster, P. F. Folliot,and A. Ortega-Rubio [eds.], Biodiversity and management of theMadrean Archipelago: the Sky Islands of southwestern United States

and northwestern Mexico, 60–70. USDA Forest Service, GeneralTechnical Report RM-GTR-264, Fort Collins, Colorado, USA.

SAS INSTITUTE. 1989. SAS/STAT user’s guide, version 6, 4th ed., vol. 2.SAS Institute, Cary, North Carolina, USA.

SAS INSTITUTE. 2002–2003. SAS, version 9.1. SAS Institute, Cary, NorthCarolina, USA.

STRAND, A. E., B. G. MILLIGAN, AND C. M. PRUITT. 1996. Are populationsislands? Analysis of chloroplast DNA variation in Aquilegia.Evolution 50: 1822–1829.

THOMPSON, J. D., T. PAILLIER, D. STRASBERG, AND D. MANIACACCI. 1996.Tristyly in the endangered Mascarene Island endemic Hugoniaserrata (Linaceae). American Journal of Botany 83: 1160–1167.

VAN DEVENDER, T. R. 1977. Holocene woodlands in the southwesterndeserts. Science 198: 189–192.

VAN DEVENDER, T. R., AND W. G. SPAULDING. 1979. Development ofvegetation and climate in the southwestern United States. Science204: 701–710.

WELLER, S. G. 1976a. Breeding system polymorphism in a heterostylousspecies. Evolution 30: 442–454.

WELLER, S. G. 1976b. The inheritance of tristyly in Oxalis sectionIonoxalis. Heredity 37: 387–393.

WELLER, S. G. 1978. Dispersal patterns and the evolution of distyly inOxalis alpina. Systematic Botany 3: 115–126.

WELLER, S. G. 1979. Variation in heterostylous reproductive systemsamong populations of Oxalis alpina in southeastern Arizona.Systematic Botany 4: 57–71.

WELLER, S. G. 1980. The incompatibility relationships of tristylous speciesof Oxalis section Ionoxalis of southern Mexico. Canadian Journal ofBotany 58: 1908–1911.

WELLER, S. G. 1981a. Fecundity in populations of Oxalis alpina insoutheastern Arizona. Evolution 35: 197–200.

WELLER, S. G. 1981b. Pollination biology of heteromorphic populations ofOxalis alpina in southeastern Arizona. Botanical Journal of theLinnean Society 83: 189–198.

WELLER, S. G. 1986. Factors influencing frequency of the mid-styledmorph in tristylous populations of Oxalis alpina. Evolution 40: 279–289.

WELLER, S. G. 1992. Evolutionary modifications of tristylous breedingsystems. In S. C. H. Barrett [ed.], Evolution and function ofheterostyly, 247–272. Springer-Verlag, Berlin, Germany.

WELLER, S. G., AND M. F. DENTON. 1976. Cytogeographic evidence for theevolution of distyly from tristyly in the North American species ofOxalis section Ionoxalis. American Journal of Botany 63: 120–125.


OF OXALIS ALPINA (OXALIDACEAE) IN THE SKY ISLANDS OF THE SONORAN DESERT1

Documents