Morphological Divergence of Native and Recently Established Populations of White Sands ( tularosa) Author(s): Michael L. Collyer, James M. Novak, Craig A. Stockwell Source: Copeia, Vol. 2005, No. 1 (Feb. 24, 2005), pp. 1-11 Published by: American Society of Ichthyologists and Herpetologists Stable URL: http://www.jstor.org/stable/4098615 . Accessed: 13/01/2011 13:16

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=asih. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

American Society of Ichthyologists and Herpetologists is collaborating with JSTOR to digitize, preserve and extend access to Copeia.

http://www.jstor.org 2005, No. 1 COPEIA February24

Copeia,2005(1), pp. 1-11

Morphological Divergence of Native and Recently Established Populations of (Cyprinodontularosa)

MICHAELL. COLLYER,JAMES M. NOVAK,AND CRAIG A. STOCKWELL

We used landmark-based geometric morphometric methods to describe patterns of body shape variation and shape covariation with size among populations of the threatened White Sands Pupfish (Cyprinodon tularosa), a species that occurs in dissimilar aquatic habitats. White Sands Pupfish populations include two genetically distinct, native populations that have been historically isolated in Salt Creek, a saline river, and Malpais Spring, a brackish spring. In addition, two populations were es- tablished approximately 30 years before this study by translocation of fish from Salt Creek to Lost River (a saline river) and Mound Spring (a brackish spring). We found significant body shape variation among populations and between males and females. Body shapes were more slender for females than for males and more slender for saline river populations than brackish spring populations. Introductions of pupfish to new habitats resulted in significant departures in body shape and shape allometry from the native Salt Creek population. Shape divergence was more pronounced for the Mound Spring population, which is consistent with a greater change in abiotic conditions. Although Mound Spring pupfish, like Malpais Spring pupfish, were more deep-bodied than saline river pupfish, differences in body shape and the level of sexual dimorphism were significant between the two brackish spring populations, indicating that deep-bodied shapes may be achieved from different anatomical con- figurations. The significant shape divergence of introduced populations warrants consideration for the conservation of this rare species, as creation of refuge pop- ulations for native stocks is a current management strategy.

UPFISHES the methods for the of Rohlf P of Cyprinodonrepre- analysis shape (e.g., sent a group of fishes known for their abil- and Slice, 1990). These techniques, inclusively ity to tolerate variable environmental conditions named geometric morphometrics (Rohlf and in desert aquatic habitats of North America. Marcus 1993; Adams et al., 2004), describe the Morphological diversification of inland species spatial arrangement of anatomical features of is thought to be associated with isolation of pop- organisms, providing statistical (Rohlf, 1999) ulations in ecologically disparate, remnant and visual (e.g., Caldecutt and Adams, 1998; Ad- aquatic habitats following desiccation of large ams and Rohlf, 2000; Rfiber and Adams, 2001) Pleistocene lakes (Miller, 1981). Early accounts advantages to traditional approaches based on of species descriptions were based chiefly on dif- linear distance measures. These techniques ferences in morphometric and meristic data have been used increasingly over the past de- (e.g., Hubbs, 1932; Miller, 1943, 1948). These cade with morphological data from fishes for early papers represent exhaustive studies of spe- studies of phylogenetics and species descrip- cies descriptions, where detailed morphometric tions (e.g., Corti and Crosetti, 1996; Cavalcanti and meristic measurements were compared, et al., 1999; Douglas et al., 2001), ontogenetic trait-by-trait. However, intraspecific ecological allometry (e.g., Walker, 1993; Hood and Heins, morphology comparisons have not been consid- 2000; Gallo-Da-Silva et al., 2002), trophic mor- ered in detail for this group of fishes, despite phology (e.g., Caldecutt and Adams, 1999; Rfib- their renowned ability to survive in a variety of er and Adams, 2001), ecological morphology aquatic desert habitats. (e.g., Corti et al., 1996; Walker, 1996, 1997), and Recent advances in morphometric analytical osteology (e.g., Loy et al., 1999, 2001). Geo- techniques have provided statistically powerful metric morphometric methods allow fine-scale

C 2005 by the American Society of Ichthyologists and Herpetologists 2 COPEIA, 2005, NO. 1 assessment of shape differences and, therefore, among populations, shape covariation with pup- could be valuable for discerning patterns of in- fish size (shape allometry), and sexual dimor- traspecific morphological variation. phism in body shape, to gain an understanding In this study, we use landmark-based, geo- of pupfish ecological morphology. metric morphometric methods to describe pat- terns of body shape variation and covariation MATERIALSAND METHODS with fish size among populations of White Sands Pupfish (Cyprinodon tularosa), which occur in We examined 393 specimens of White Sands dissimilar environments in the Tularosa Basin Pupfish collected from the four populations (N in southern New Mexico. The White Sands Pup- = 49-50 females and N = 45-50 males from each fish is listed as threatened by New Mexico and population). Descriptions of collection sites are composed of two native and two recently intro- described in detail by Collyer (2003). Nonmet- duced populations (Pittenger and Springer, ric multidimensional scaling revealed that 1999). The native populations occur in brackish White Sands Pupfish habitats can be distinctly spring (Malpais Spring) and saline river (Salt described as brackish springs (BS = Malpais Creek) habitats and have been isolated presum- Spring and Mound Spring) or saline rivers (SR ably since the desiccation of the Pleistocene = Salt Creek and Lost River) based on abiotic Lake Otero (Miller and Echelle, 1975; Pittenger factors including temperature, salinity, and wa- and Springer, 1999). Miller and Echelle (1975) ter flow (Collyer, 2003). noted morphological and meristic differences between the two native populations. Salt Creek Fish collection and landmark acquisition.--Adult Pupfish had a more slender body and scales that pupfish were used in this study to reduce the were smaller and more numerous. In addition, amount of intrapopulation shape variation Salt Creek Pupfish had less scalation of the bel- based on ontogeny. Pupfish were collected from ly. These differences did not warrant subspecies 26 May to 8 June 2001, at the middle sections designation, however, according to the authors. of Lost River (26-27 May) and Salt Creek (30 Nevertheless, genetic distinction of the two na- May), the springhead and outflow marsh of Mal- tive populations has been reported (Stockwell pais Spring (4-5 June), and the upper pond of et al., 1998; Iyengar et al., 2004) based on fixed Mound Spring (8 June). Fish were live-captured or nearly fixed differences in allele frequencies with unbaited minnow traps placed at depths of allozymes and microsatellites. less than one meter at Lost River and Salt The two introduced populations also occur in Creek, with beach seines at Mound Spring, and a brackish spring (Mound Spring) and a saline with a combination of the two techniques at river (Lost River). These populations were es- Malpais Spring. Fish were transported live to a tablished circa 1970 (Pittenger and Springer, research laboratory at Holloman Air Force 1999) from translocation of Salt Creek fish Base, sacrificed in approximately 500 mg/L tri- (Stockwell et al., 1998). They are not genetically caine methanesulfonate (MS 222; Summerfelt differentiated fi-om the Salt Creek population and Smith, 1990), separated into female and based on genetic distances using allozyme and male groups, and placed into ice baths. Fish microsatellite data (Stockwell et al., 1998), but were patted dry, labeled, and photographed the introductions resulted in shifts in allele fre- within two hours of sacrifice. quency from the Salt Creek population and loss Digital photographs were captured with a of some alleles at various microsatellite markers Minolta RD-175(r) digital camera mounted on for both introduced populations (Stockwell and a photography table approximately 0.25 m di- Mulvey, 1998; Stockwell et al., 1998; Miller, rectly above the table surface. The left lateral 2001). surface of each specimen was photographed us- The creation of refuge populations for native ing flash lighting. Thirteen landmarks on each genetic stocks is a current management strategy specimen (Fig. 1) were digitized using TPSDIG for the White Sands Pupfish. The introduction software, version 1.31 (F. J. Rohlf, 2001, un- of the Lost River and Mound Spring popula- publ.). These landmarks represent anatomical tions approximately three decades before this homologs that could be easily identified for study has served as a source of ongoing research each specimen and provide quantitative infor- to evaluate the ecological and evolutionary im- mation for describing shape. plications associated with creating refilge pop- ulations. The purpose of this study was to con- Shape analysis.--We used landmark-based geo- sider morphological divergence of pupfish pop- metric morphometrics techniques to quantify ulations introduced to new environments. Our body shape (Rohlf and Marcus, 1993; Adams et objectives were to describe body shape variation al., 2004). Geometric morphometric methods COLLYFR ET AL.-PUPFISH MORPHOLOGICAL DIVERGENCE 3

?r ?k'~ ?i 2; I -?? ? ? ?r? u ii--?? d: ?~ ,1: 1L4~~1~~;y t ilf: -? ?? ,?~:: ? a? "; $' -r! r zr ,., :f ?- nY -r ?x?i ?( r~4 J; ?- r; 4~c~~?~ r* 1?Y? i,,i. i,~ait~T'?;C~3? ??'I i, D ;r ~,?.1? ?le I ~?'-~~~m?;,(;~ii~~ r.if? ?~ ~p~~"il~ --m ;, ?P .. `t ?~ 'R 3)'-cM??:?1 r; ~c,..l i -c T *r'-.?6.g c ? :*:; .5: :?; 5: .k re i?~ I - ? I ,- ~-Q~- +? xLI i ?~7..~-(~?

...

.,.?e

rmr ~rr~i I??r 'ci tri~~ ?1~.2 ?;.,e r c. Fth rB 'k r -.i YibT;.s~r~_; s .?r c; ?. i) m. ~iL~ii ".-C~ ?P~:-c-;.s i b ~ ;? ?t-.,, F. s ~f? ~??~ls ??11 ? u ~S~ ?r. ~: u. ?.. ,c'J;c~ 3?;? ?; ": ~bF~'?iF I:?? b ~r s

Fig. i. Landmarkdefinition for morphometricanalyses of WhiteSands Pupfish. A male pupfishfrom MoundSpring is shown. generate shape variables from x, y Cartesian co- which the significance of observed values could ordinates of landmarks, with the effects of spec- be inferred as the probability of finding a great- imen size, orientation, and position held con- er distance by chance. Shape variation was stant (Rohlf and Slice, 1990). All nonshape var- graphically examined with the first three axes iation was held constant, mathematically, with a of a principal component analysis (PCA). generalized Procrustes analysis (GPA) of origi- Although Dp provides an unscaled metric of nal landmark data (Rohlf and Slice, 1990), shape difference between two points in shape which involves (1) a generalized least-squares space, the non-Euclidian nature of this space superimposition of original landmark data can be prohibitive for certain statistical analyses, scaled to the same size for all landmark config- such as the covariation of shape and other var- urations, and (2) centering subsequent size-free iables. The thin-plate spline (TPS) is a method landmark configurations at the origin of a com- that projects data from shape space into a tan- mon coordinate system. The superimposed gent space that is Euclidian (Bookstein, 1989, (aligned) landmark configurations represent 1991), thereby allowing statistical analyses of points in a multidimensional shape space where shape variation and covariation with other var- relationships of specimens are defined by the iables using linear models (Marcus, 1993; Mar- metric, Procrustes distance (the square root of cus et al. 1996). We used TPS to generate shape the sum of squared distances between homolo- variables (partial warp scores plus two uniforms gous landmarks). scores) from aligned landmark configurations Differences in Procrustes distance (Dr) produced by GPA (Bookstein, 1996; Rohlf and among population by sex groups (eight total) Bookstein, 2003). GPA and TPS were per- were assessed with permutation tests. Because formed using the TPSRELW software, version we observed strong evidence for sexual dimor- 1.29 (F. J. Rohlf, 2003, unpubl.). phism (see Results) individuals were randomly assigned to any of the four populations, within Analyses incorporatingshape covariationwith size.- female and male blocks, for 4999 iterations. In To consider the covariation in shape and size each iteration, we calculated the test statistic, (shape allometry), we used multivariate analysis ID,, - Dj, as a direct comparison of two shape of covariance (MANCOVA) for shape variables differences, i and j. Direct comparisons were (Klingenberg, 1996). Independent variables in- performed for females and males separately to cluded population, sex, size (covariate), and all compare shape divergence of Lost River and interactions. The metric of size used in this Mound Spring populations from Salt Creek to analysis was centroid size (CS), which is calcu- the shape differences between Salt Creek and lated as the square root of the sum of squared Malpais Spring, the two native populations. The distances of individual landmarks from the cen- permutation tests produced distributions of troid of the landmark configuration (Bookstein, 5000 statistics (including observed values) from 1991). In this study, CS was highly correlated 4 COPEIA, 2005, NO. 1 with two other size measures, standard length produced from means of aligned coordinates, (SL) and mass (Pearson r = 0.99 and 0.97, re- using the software, TPSSPLIN, version 1.20 (F. spectively). The metric of shape difference be- J. Rohlf, 2004, unpubl.), and for the regression tween groups (accounting for shape allometry) of shape on size using the software, TPSREGR, was calculated as the magnitude of the shape version 1.28 (E J. Rohlf, 2003, unpubl.) vector between least-squares means. This value was converted to a generalized Mahalanobis RESLtITrs (1936) distance, D,,, using the pooled within- group variance/covariance matrix of the multi- Body shape variation.-For the shape differences variate model, to assess statistical significance. we considered, Procrustes distances were largest Differences in shape allometry among popu- between female and male means, within popu- lation by sex groups were considered by com- lations, and smallest between Salt Creek and paring vectors of regression coefficients (b), Lost River means, suggesting that much shape which describe the covariation of shape and size variation could be attributed to sexual dimor- (see Appendix 1). The association of allometry phism (Table 1; Fig. 2). The only measure of vectors was considered by the angle, 0, between sexual dimorphism that differed from others them, calculated from the equation: 0 = was that of Mound Spring, which was signifi- for vectors and normalized to less < 0.0008). Among- cos' (b', b1) i j cantly dimorphic (P,,,ad same unit length, where T represents a vector population patterns of shape variation were sim- transpose. The vector correlation, r~, is related ilar for both females and males, with the great- to 0 by the following equation: r,1= cos0 (see est distances occurring for comparisons of e.g., B6gin and Roff, 2003). Thus, highly cor- Mound Spring to SR populations. The diver- related vectors (i.e., that express similar covari- gence of Mound Spring shapes from Salt Creek ation of shape and size) should have a small was significantly greater than the divergence of angle. Using a permutation procedure (de- Lost River shapes (P,,,, = 0.0002 for both fe- scribed in Appendix 1), we generated distribu- males and males) and was greater than the tions of 4999 random angles for every vector shape differences described for native popula- comparison. Vectors were considered signifi- tions (P,,,,,, = 0.0002 for both females and cantly different if the angle between them was males). Notably, the shape differences between greater than or equal to the empirical angle as- Mound Spring and Malpais Spring were nearly sociated with a one-tailed probability of = as large as any other population comparisons, 0.05 from the random distribution of 5000P,.,,,nd an- despite both populations occurring in springs. gles (including the observed value). Females had more slender body shapes than We also calculated 0 between shape vectors males and SR pupfish had more slender body between least-squares means to compare direc- shapes than BS pupfish (Fig. 2). Although BS tional differences in shape, holding the effects pupfish were deep-bodied in both populations, of allometry constant. Shape vector compari- the large difference in shape between them re- sons included vectors between means of females sulted from deep-bodied shapes produced by and males (i.e., measures of sexual dimor- different landmark configurations. Deforma- phism), compared among populations, and vec- tion grids revealed that the snout (landmark 1) tors between SR and BS environments. Distri- was more ventrally positioned, and the curva- butions of 5000 random angles were generated ture of the dorsal crest (landmarks 2, 3, and 4) for each vector comparison with the same pro- was more pronounced for Mound Spring fe- cedure for allometry vectors to determine the males and males. In addition, the Mound significance of vector directional differences. Spring shapes exhibited narrower anal fins Permutation procedures were performed us- (landmarks 8 and 9) and shallower caudal re- ing the software, Poptools, version 2.6.2 (G. M. gions (landmarks 5, 6, 7, and 8) for both fe- Hood, 2004, unpubl.). We verified parameter males and males. Thus, although Salt Creek fish estimates and multivariate test statistics for our were introduced into an environment at Mound multivariate model with the regression module Spring that was more similar to Malpais Spring, of NTYSYS-pc, version 2.1 (F. J. Rohlf, 2002, Ex- body shape did not converge on the Malpais eter Software, Setauket, New York). We also con- Spring form, despite becoming more deep-bod- ducted a two-way ANOVA on CS for the inde- ied. pendent variables, population and sex, to deter- mine whether size differences existed among Shape covariation with size.-MANCOVA indicat- population by sex combinations, using the soft- ed that the multivariate model for covariation ware, JMP, version 5.0.1 (SAS Institute, Inc., of shape and size among population by sex Cary, NC, 2002). Shape deformation grids were groups was significant: (Wilks' A = 0.00054; Es- COLLYER ET AL.-PUPFISH MORPHOLOGICAL DIVERGENCE 5

TABLE1. MATRICESOF BODY SHAPE DISTANCES FOR POPULATIONS OF WHITE SANDS PUPFISH. Procrustes distances (Dp) are represented in elements above matrix diagonals and generalized Mahalanobis distances (DM) are represented below. Distances corresponding to measures of sexual dimorphism within populations are also * shown. All DM-valuesare significant (P < 0.0001). denotes significantly smaller value than Dp between SC and MO 0.0002). ** denotes smaller than other measures of sexual (0.0002 (Prand = significantly Dp dimorphism < Pr,,nd < 0.0008). Abbreviations correspond to populations: Lost River (IR), Malpais Spring (MA), Mound Spring (MO), and Salt Creek (SC).

Females Males LR MA MO SC LR MA MO SC LR - 0.033 0.063 0.016*,a - 0.044 0.067 0.025*,a MA 4.219 - 0.059 0.024* 5.144 - 0.055 0.033* MO 5.419 5.360 - 0.062 5.397 5.396 - 0.058 SC 2.639 3.334 4.861 - 2.983 4.465 4.582 -

Sexual Dimorphism D)p D LR 0.075 6.490 MA 0.066 5.442 MO 0.048** 4.545 SC 0.071 5.230

t These values were also significantly smaller than Dp between SC and MA.

timated F330,4475 = 11.27; P < 0.0001). All terms than males. Females and males were not differ- in the model were significant, including the ent in size in the other three populations. How- Population X Sex X CS interaction (Wilks' A = ever, Mound Spring fish were significantly small- = 0.763; Estimated F66,1064 1.53; P = 0.005), in- er than Malpais Spring fish. dicating significant variation in shape allometry When accounting for the covariation of shape among population by sex groups. Further, CS and size, all pairwise shape distances (DM) con- significantly differed among population by sex sidered were significant (P 0.0001 for all dis- = groups (F3,385 5.82; P = 0.0007; Figs. 3, 4). tances; pairwise o- = 0.0031). Further, there was Pairwise comparisons (using Tukey's honest sig- much concordance with DMand Dpmeasures for nificant difference tests) revealed that SR fish all shape differences, with the exception that DM were significantly larger than BS fish and that between Mound Spring and Salt Creek was not females from Lost River were significantly larger larger than between Mound Spring and Malpais

PCII

= MA INAO N sc LR

II SPPC

PC I

Fig. 2. First three principal components (PC) of shape variation for White Sands Pupfish. The three PCs shown represent approximately 69% of the total shape variation. Values in the plot correspond to group means for females (circles) and males (squares). Populations are indicated with the following abbreviations: Lost River (LR), Malpais Spring (MA), Mound Spring (MO), and Salt Creek (SC). Deformation grids (with a scale factor of 2X) are shown to facilitate a visual understanding of the shape corresponding to group means. 6 COPEIA, 2005, NO. 1

-IiI--1 sc

LR

t-F--S BS MA

I I I I Mo 1 25 30 35 40 45 50 55 60 CentroidSize Fig. 3. Shape allometry of male White Sands Pupfish. Shape allometries are shown for each population (abbreviations the same as in Fig. 1) with saline river (SR) populations above and brackish spring (BS) pop- ulations below. Box plots show the range, 25th percentile, median, and 75th percentiles of centroid size data. Deformation grids are shown for the extents of the centroid size range to demonstrate the regression of shape on size. Deformation grids at large sizes are scaled 3x to facilitate a visual understanding of shape change associated with size.

F so t---] I LR - SR BS MA MA H

2530 35 40 45 50 55 60 CentroidSize

Fig. 4. Shape allometry of female White Sands Pupfish. Descriptions are the same as in Figure 3. COLLYER ET AL.-PUPFISH MORPHOLOGICAL DIVERGENCE 7

TABLE2. MATRICESOF TESTSTATISTICS FOR THE COVARIATION OF BODYSHAPE AND SIZE FOR POPULATIONS OF WHITESANDS PUPFISH. Vector correlation coefficients, rv,are represented in elements above matrix diagonals and angles, 0, (in degrees) are represented below. Test statisticsfor sexual dimorphismare also shown. Sig- nificantlydifferent vectors (rv < 0.350, 0 > 690) are italicized.Abbreviations are the same as in Table 1.

Females Males LR MA MO SC LR MA MO SC LR - -0.060 0.191 -0.043 - 0.759 0.870 0.788 MA 93.42 - 0.376 0.350 40.61 - 0.688 0.616 MO 78.97 67.92 - 0.422 29.51 46.51 - 0.796 SC 92.44 69.49 65.01 - 38.04 51.96 37.21 - SexualDimorphism r, 0 LR 0.454 63.015 MA 0.600 53.159 MO 0.373 68.113 SC 0.482 61.183

Spring for both females and males (Table 1). Directional differences in sexual dimorphism This finding does not alter the pattern of be- among populations were small but significant tween-population differentiation in shape; al- (0.74 < rv < 0.93; 21.60 < 0 < 41.2; 0.0002 < though Mound Spring is slightly less divergent Prand < 0.031). Differences were greatest for from Salt Creek when accounting for shape var- Mound Spring (300 < 0 < 41.2'). Directional iation related to size variation (or shape and differences for shape vectors between SR and size covariation), either shape distance demon- BS populations were larger but not necessarily strates that shape differences between Salt significant. The shape divergence of Mound Creek and Mound Spring are greater than Spring from Salt Creek was significantly differ- differences between Salt Creek and Mal- ent in direction = 0.002 for both females shape (Prand pais Spring. and males) from Malpais Spring (rv = 0.48 and The significant heterogeneity of shape allom- 0.28; 0 = 61.60 and 73.90; females and males, etries was caused by differences among popu- respectively). However, the same pattern was lations within females (Figs. 3, 4; Table 2). De- not observed when Lost River was the SR pop- spite differences in shape among populations, ulation considered (rv = 0.44 and 0.60; 0 = allometries were not differ- 63.90 and 53.50; = 0.102 and 0.333; females shape significantly Pran, ent among populations for males. Each popu- and males, respectively). Thus, although the dif- lation showed a positive association with body ference in directions for shape vectors between depth, curvature of the dorsal crest, and size SR and BS environments appear large com- (Fig. 3). For Females, however, significantly dif- pared to directional differences among sexual ferent allometries were observed for all SR-BS dimorphism vectors, these differences are not comparisons except one (Salt Creek compared greater than expected by chance when compar- to Mound Spring; Fig. 4; Table 2). For Salt ing with Lost River. This result is not surprising Creek and Lost River females, negative associa- given the larger shape distances observed for tions of body depth and size were observed, sexual dimorphism than interpopulation com- with Lost River fish having greater tapering of parisons within sex. the posterior body. At larger sizes, Malpais Spring females showed a flattening of the dorsal DISCUSSION crest and a more posterior position of the anal fin. Mound Spring females had exaggerated Introduction of White Sands Pupfish from dorsal crests at larger sizes-a trend that shows Salt Creek to new habitats lead to significant some similarity to male shape allometries. Thus, body shape divergence for new populations. Di- the significantly smaller level of sexual dimor- vergence in body shape was relatively modest phism in shape for the Mound Spring popula- for pupfish introduced to Lost River, a SR hab- tion might be because of shape allometry that itat like Salt Creek. is more malelike for BS females compared to Both SR populations were characterized by SR females (sexual dimorphism was also less for body shapes that were more slender than BS Malpais Spring than SR populations, but not sig- pupfish. However, the divergence of body shape nificantly so; Table 1). for pupfish in Mound Spring exceeded the level 8 COPEIA, 2005, NO. I of shape difference between Salt Creek and with adaptive morphology related to water flow Malpais Spring, the two native populations that and salinity. The body shapes associated with have been isolated since the desiccation of the Salt Creek and Lost River environments may re- Pleistocene Lake Otero (5000-10,000 years flect a combination of functional adaptations ago). The divergence of the Mound Spring such as (1) an advantage to living in high water body shape did not, however, converge on the flow (Vogel, 1994), or (2) reduced circumfer- Malpais Spring form, although both shapes ence (compared to fish length) as a result of were deep-bodied. This result indicates that high osmotic potential in high salinity. The ad- deep-bodied shapes may be achieved by alter- vantage of streamlined body shapes for fishes native landmark configurations and gives cre- that live in flowing water, as an adaptation to- dence to the morphological distinctiveness of ward reduced energetic cost via drag resistance, native populations. is well documented (e.g., Blake, 1983; Weihs Indeed, the Mound Spring body shape ap- and Webb, 1983; Videler, 1993). Low-speed ma- pears to have retained some features of a shape neuverability and neutral buoyancy can be aug- that evolved in a fluvial habitat. The shape of mented by highly compressed, laterally flat- the head (more ventrally positioned snout) and tened body shapes that facilitate pivoting mo- shallow caudal region were consistent with the tions (Webb, 1997), which may be less energet- Salt Creek shape. Thus, it is possible that the ically costly in the spring environment. significant shape difference between Mound Therefore, shape variation between SR and BS Spring and Malpais Spring shapes is associated environments may reflect greater streamlining with genetic constraint (Oster and Albrecht, of body shape in the SR environments. Howev- 1982; B6gin and Roff, 2003). This hypothesis is er, a tendency for pupfish to have slender body supported by the significant directional differ- shapes in other, perhaps less fluvial, saline en- ences of SR-BS shape vectors (i.e., the shape vironments has been noted (e.g., Miller, 1948). vectors between Salt Creek and Malpais Spring Experimental research on White Sands Pupfish and the shape vectors between Salt Creek and native populations has indicated that body Mound Spring are uncorrelated). This result shape is, in part, phenotypically plastic with a possibly indicates that the genetic covariances of negative association of body depth and salinity Salt Creek and Malpais Spring populations dif- (Collyer, 2003). Therefore, difference in body fer; thus, shape response to similar selection dif- shape between SR and BS pupfish may indicate fered between the two (see e.g., Klingenberg a phenotypically plastic response to variation in and Leamy, 2001; Begin and Roff, 2003). salinity levels among habitats. Such a hypothesis Overall, differences in shape allometry were is supported by the observed Lost River body subtle or not significant. Although the Mound shapes. Lost River Pupfish (which were intro- Spring habitat may have altered the shape of duced from the Salt Creek population) were pupfish introduced from Salt Creek, shape al- slightly more slender than Salt Creek shapes lometry did not significantly diverge. However, and the Lost River habitat is slightly more saline when compared to Lost River, the difference in (Stockwell and Mulvey, 1998). shape allometry between the two introduced Second, body shape variation may indicate populations was significant for females. Mound sexual difference in behavior. Previous studies Spring females had a positive association of cur- (e.g., Cowles, 1934; Raney et al., 1953; Echelle, vature of the dorsal crest and size, much like 1973) have documented breeding territory de- males in all populations, but unlike females in fense of males in Cyprinodon.Barlow (1961) doc- Lost River. Thus, the level of sexual dimorphism umented lateral threat displays with Cyprinodon was possibly significantly smaller for the Mound macularinusthat were also observed with Cypri- Spring population because mean shapes are, in nodon pecosensisby Kodric-Brown (1977, 1978), part, a result of shape allometry. As a result, the who also noted that visual cues were important female and male means of Mound Spring con- in territorial defense by male Cyprinodonagainst verge slightly in shape space. The greater sexual conspecific males. Breeding territory defense by dimorphism of SR populations is best explained males has also been documented for the White by the positive association of body depth and Sands Pupfish (Suminski, 1977). Male size is an size for males but not for females. Nevertheless, important determinant in Cyprinodonfor the the more slender body shapes of SR fish were ability of males to obtain and defend territories distinct, regardless of the positive association of (Kodric-Brown, 1978). Thus, deep-bodied body depth and size for males. We outline two shapes may be associated with increased fitness possible ecological explanations for such a re- because of a functional or display advantage of lationship below. deep-bodied males to defend territories. Fur- First, body shape variation may be associated ther, this hypothesis is supported by the de- COLLYER ET AL.-PUPFISH MORPHOLOGICAL DIVERGENCE 9 creased association of body depth and size for morph), written and maintained by F. J. Rohlf. females, who do not defend breeding territo- The Poptools software is available at http:// ries. www.cse.csiro.au/poptools, written and main- The results of this study warrant further in- tained by G. M. Hood. vestigation into the adaptive significance of body shape in White Sands Pupfish and the con- LITERATURECITED comitant management of native populations. Based on the genetic distinctiveness of native populations (Stockwell et al., 1998) and their ADAMS,D. C., ANDFJ. ROHLF.2000. Ecological char- acter in Plethodon:biomechanical dif- geographic isolation in dissimilar environments, displacement the native of White Sands ferences found from a geometric morphometric populations Pupfish Proc. Natl. Acad. Sci. USA:4106-4111. are as study. managed separate evolutionarily signifi- AND D. E. SLICE.2004. Geometric cant units - (ESU; Ryder, 1986; Waples, 1991; morphometrics: ten years of progress following the Stockwell et al., 1998). Consistent with manage- "revolution." It. J. Zool. 71:5-16. ment strategies of short-lived desert fish species BARLOW,G. W. 1961. Social behavior of the Desert (Hendrickson and Brooks, 1991), the establish- Pupfish, Cyprinodonmacularius, in shore pools of the ment of refuge populations, as a hedge against Salton Sea, California. Ecology 39:580-587. extinction of native populations, is a goal of the BEGIN,M., AND D. A. ROFF.2003. The constancy of current management plan. Because Lost River the G matrix through species divergence and the effects of constraints on and Mound Spring populations could poten- quantitative genetics phe- serve as of the native notypic evolution: a case study in crickets. Evolution tially refuge populations 55:1107-1120. Salt Creek it is to evalu- population, important BLAKE,R. W. 1983. Fish locomotion. Univ. ate how novel environments could affect the Cambridge Press, Cambridge. of the Salt Creek ESU in ref- evolutionary legacy BOOKSTEIN,F. L. 1989. Principal warps: thin-plate uge environments. Our results indicate that splines and the decomposition of deformations. contemporary morphological evolution of ref- IEEETrans. Pattern. Anal. Mach. Intell. 11:567-85. uge populations is a potential concern. -. 1991. Morphometric tools for landmark data: geometry and biology. Cambridge Univ. Press, New York. ACKNOWLEDGMENTS -. 1996. Combining the tools of geometric mor- 131-151. In: Advances in We thank D. Rogowski, D. Harstad, D. Moen, phometrics, p. morpho- metrics. L. F. M. A. P. and H. Resier for assistance with field collec- Marcus, Corti, Loy, G. J. Nay- lor, and D. E. Slice (eds.). NATO ASI Series A, Life tions. We are to R. Ge- grateful Myers (Range Sciences. Vol. 284. Plenum Press, New York. Environmental ologist, Stewardship Division) CALDECUTT,W. C., ANDD. C. ADAMs.1998. Morpho- and T. A. Ladd (Director, Environment and metrics of trophic osteology in the Threespine Safety Directorate, U.S. Army, White Sands Mis- Stickleback, Gasterosteusaculeatus. Copeia 1998:827- sile Range) for arrangement of range visitation. 838. Public release of this article was approved by the CAVALCANTI, M. J., L. R. MONTEIRO, AND P. R. D. White Sands Missile Range with distribution un- LOPES.1999. Landmark-based morphometric anal- limited, by Operations Security (OPSEC) review ysis in selected species of serranid fishes (Percifor- mes: Zool. Stud. 38:287-294. on 18 June 2003. Pupfish were collected under Teleostei). New Mexico State 2887 to COLLYER,M. L. 2003. Ecological morphology of the collecting permit Sands CAS. Fish were sacrificed North Da- White Pupfish (Cyprinodontularosa). Unpubl. following Ph.D. diss., North Dakota State Univ., kota State Institutional Care Fargo. University CORTI,M., ANDD. CROSETTI.1996. varia- and Use Geographic protocol A0117. Laboratory support tion in the grey mullet: a geometric morphometric was provided by M. H. Reiser and Holloman Air analysis using partial warp scores. J. Fish Biol. 48: Force Base. This work benefited from discussion 255-269. with F. J. Rohlf, D. Slice, and D. Adams. Re- --, A. LoY, AND S. CATAUDELIA.1996. Form search was supported by a DOD legacy grant to changes in the sea bass, Dicentrarchus labrax (Mo- CAS, administered by M. H. Resier, 49 CES/ ronidae: Teleostei), after acclimation to freshwater: CEV, Environmental Flight, Holloman Air Force an analysis using shape coordinates. Environ. Biol. Fish. 47:165-175. Base, and an EPA Star (U-91597601- Fellowship R. B. 1934. Notes on the and 1) to MLC. Some of the data were con- COWLES, ecology breed- analyses habits of the Desert Minnow, macu- ducted at the Savannah River Labora- ing Cyprinodon Ecology Baird and Girard. 1934:40-42. with from the of Geor- larius Copeia tory support University DOUGLAS,M. E., M. R. DOUGLAS,J. M. LYNCH,AND D. to MLC. All software used is gia morphometrics M. McELROY.2001. Use of geometric morphomet- available at the SUNY at Stony Brook morpho- rics to differentiate Gila () within the metrics website (http://life.bio.sunysb.edu/ Upper Colorado River Basin. Copeia 2001:389-400. 10 COPEIA, 2005, NO. 1

Death ECHEIE, A. A. 1973. Behavior of the pupfish, Cypri- . 1948. The cyprinodont fishes of the nodon rubrofluviatilis.Copeia 1973:68-76. Valley System of eastern California and southwest- GALLO-DA-SILVA,V., M.J. CAVAI[(A-NTI,AND A. TINToR- ern Nevada. Occ. Pap. Mus. Zool. Univ. Mich. 529: NI. 2002. Using geometric morphometrics to study 1-155. ontogenetic shape changes in Paralepidotusornatus . 1981. Coevolution of deserts and (Agassiz 1833-43) (: Semionotidae). (genus Cyprinodon)in the American Southwest, p. Arq. Mus. Nac. 60:117-120. 39-94. In: Fishes in North American deserts. R. J. HENRICKSON, D. A., AND J. E. BROOKS.1991. Trans- Naiman and D. L. Soltz (eds.). John Wiley and planting short-lived fishes in North American de- Sons, New York. serts: and ANDA. A. 1975. review, assessment, recommendations, p. , , ECHELLIE. Cyprinodontularosa, 283-298. In: Battle against extinction. W. L. Minck- a new cyprinodontid fish from the Tularosa Basin, ley and J. E. Deacon (eds.). Univ. of Arizona Press, New Mexico. Southwest. Nat. 19:365-377. Tucson. OSTER, G. AND P. ALBRECHT. 1982. Evolution and bi- Hoon, C. S., AND D. C. HEINS.2000. Ontogeny and fiurcation of developmental programs. Evolution allometry of body shape in the Blacktail Shiner, Cy- 36:444-459. venusta. 2000:270-275. 1999. Native prinella Copeia PITrENGER, J. S., AND C. L. SPRINGER. HI'BBS,C. L. 1932. Studies of fishes of the order Cv- range and conservation of the White Sands Pupfish XII. A new related to prindontes, genus Empetri- (Cyprinodontularosa). Southwest. Nat. 44:157-165. chthyes.Occ. Pap. Mus. Zool. Univ. Mich. 198:1-5. RANEY,E. C., R. H. BACKUS,R. W. CRAWFORD,AND C. C. A. D. ANDP. A. IYENGAR,A., STOcKWETI.I,,LArFIEID, R. ROBINS. 1953. Reproductive behavior in Cyprin- MORIN. of 2004. Characterisation microsatellite don variegatusLacepde, in Florida. Zoologica 38:97- markers in White Sands Pupfish (Cvprinodon tula- 105. Mol. Ecol. Notes 4:191-193. rosa). ROHLF,F. J. 1999. On the use of shape spaces to com- C. P. 1996. Multivariate KLINGENBERG, allometry, p. pare morphometric methods. Hystrix 11:1-17. 23-49. In: Advances in L. F. morphometrics Marcus, - , AND F. L. BOOKSTEIN.2003. Computing the P. and D. E. Slice M. Corti, A. Loy G. J. Naylor, uniform component of shape variation. Syst. Biol. (eds.). NATO ASI Series A, Life Sciences. Vol. 284. 52:66-69. Plenum Press, New York. ANDL. F. MARCUS.1993. A revolution in mor- , ANDL. LEAMY.2001. ,, J. Quantitative genetics phometrics. Trends Ecol. Evol. 8:129-132. of geometric shape in the mouse mandible. Evo- ANDD. E. SLICE.1990. Extensions of the Pro- lution 55:2342-2352. crustes, method for the superimposition of A. 1977. success and optimal KODRIC-BROWN, Reproductive landmarks. Zool. 39:40-59. the evolution of territories in Syst. breeding pupfish (Cy- AND D. C. ADAMS. 2001. con- Evolution 31:750-766. L., Evolutionary prinodon). R['BER, of and in . 1978. Establishment and defence of vergence body shape trophic morphology breeding cichlids from Lake Evol. Biol. 14: territories in a Tanganyika. J. pupfish (Cyprinodontidae: Cyprino- 325-332. Anim. Behav. 26:818-834. don). A. 1986. conservation and the di- A., P. BRONZI,AND S. MOLTENI.1999. RYDER,O. Species Lov, Morpho- lemma of the Trends Ecol. Evol 1:9-10. metrics in the characterisation of the cranial subspecies. ANDM. MULVEY.1998. of STOCKWTFI.,C. A., Phosphoglu- growth pattern Acipensernaccarii. J. Appl. Ichthy- and ol. 15:50-53. conate dehydrogenase polymorphism salinity in the White Sands Evolution. 52:1856- -- -, M. BERTEILLETTI,C. COSTA, L. FERIIN,AND S. Pupfish. 1860. CATAUDEIHA.2001. and tra- Shape changes growth AND A. G. 1998. Genetic evi- jectories in the early stages of three species of the , JONES. dence for two units of genus Diplodus (Perciformes, Sparidae). J. Morph. evolutionarily significant 250:24-33. White Sands Pupfish. Anim. Conserv. 1:213-226. R. R. 1977. Life of the White Sands MAHAIANOBIS,P. C. 1936. On the generalized distance St'MINSKI, history in New Mex- in statistics. Proc. Natl. Inst. Science, India 2:49-55. Pupfish and distribution of Cyprinodon ico. master's New Mexico State MARCUS,L. F. 1993. Some aspects of multivariate sta- Unpubl. thesis, Cruces. tistics for morphometrics, p. 95-130. In: Contribu- Univ., Ias R. ANDL. S. SMITH.1990. tions to morphometrics. Vol. 8. L. F. Marcus, E. Bel- SUMMERFELT,C., Anesthesia, lo, and A. Garcia-Valdecasas (eds.). Museo Nacional surgery, and related techniques, p. 213-271. In: C. B. Schreck and P. B. de Ciencias Naturales. Madrid, Spain. Methods for fish biology. American Fisheries Bethesda, , M. CORTI,A. LoY, G. J. NAYIOR,AND D. E. Moyle (eds.). Society SLICE. 1996. Advances in morphometrics. NATO MD. ASI Series A, Life Sciences. Vol. 284. Plenum Press, VII)EIR, J. J. 1993. Fish swimming. Chapman and New York. Hall, Inc., New York. MILLER,K. M. 2001. Pgdh Polymorphism in the White VocEI., S. 1994. Life in moving fluids: the physical Sands Pupfish: an investigation of a possible case of biology of flow. Princeton Univ. Press, Princeton, natural selection in the wild. Unpubl. master's the- NJ. sis, North Dakota State Univ, Fargo. WAILKER, J. A. 1993. Ontogenetic allometry of Three- MILLER, R. R. 1943. Cyprinodonsalinus, a new species Spine Stickleback body form using landmark-based of fish from Death Valley, California. Copeia 1943: morphometrics, p. 193-214. In: Contributions to 69-78. morphometrics. Vol. 8. L. F. Marcus, E. Bello, and COLLYER ET AL.-PUPFISH MORPHOLOGICAL DIVERGENCE 11

A. Garcia-Valdecasas(eds.). Museo Nacional de means for one population, or between population Ciencias Naturales. Madrid, Spain. means within females or males). The magnitude of . 1996. of Principal components body shape shape difference between two groups (adjusting for variation within an endemic radiation of threespine effects in the model) is the Euclidian distance of the stickleback, p. 321-334. In: Advances in morpho- shape vector: Di = IlAyi| = (Ay7' Ayi)1'2 for compari- metrics L. F. Marcus, M. Corti, A. Loy, G. J. P. Nay- son, lor, and D. E. Slice (eds.). NATO ASI Series A, Life i. the directions of different vectors Sciences. Vol. 284. Plenum Press, New York. We compared by the between them: . 1997. Ecological morphology of lacustrine calculating angle Threespine Stickleback Gasterosteus aculeatus L. I (Gasterosteidae) body shape. Biol. J. Linn. Soc. 61: 0 = cos- i 3-50. Di- Dj ', R. S. 1991. Pacific Salmon, WAPLES Oncorhynchusspp. for shape vectors i and j (Begin and Roff, 2003). This and the definition of "species" under the endan- is the same procedure for calculating the angle be- gered species act. Mar. Fish. Rev. 53:11-22. tween allometry vectors, b, which are vectors of re- WEBB,P. W. 1997. Swimming, p. 3-24. In: The physi- coefficients from the matrix B, to ology of fishes. D. H. Evans (ed.). CRC Press, Boca gression pertaining Raton, FL. size covariates described in X. WEIHS,D., ANDP. W. WEBB.1983. Optimization of lo- To test the significance of 0 for either shape vector comotion. p. 339-371. In: Fish biomechanics. D. or allometry vector comparisons, a permutation pro- Weihs and P. W. Blake (eds.). Praeger Publsihers, cedure was performed. Because significant popula- New York. tion and sex effects were observed in MANCOVA, we used a procedure that preserved these effects. First, DEPARTMENTOF (MLC, CAS) BIOLOGICALSCI- we determined the parameter estimates from the NORTH DAKOTA STE- ENCES, STATE UNIVERSITY, model Y = XIB, + e, where X, contained effects cod- NORTH DAKOTA VENS HALL, FARGO, 58102; ed for population and sex. For each permutation pro- AND (JMN) SAVANNAH RIVER ECOLOGY LABO- cedure iteration, the residuals were randomly per- UNIVERSITYOF DRAWER RATORY, GEORGIA, E, muted twice, to form two matrices of residuals, E*,~, AIKEN, SOUTH CAROLINA29802. PRESENTAD- and E*1,2, where the superscript, *, represents a ran- DRESSES: (MLC) DEPARTMENT OF ECOLOGY, domized form of the matrix. These matrices were AND ORGANISMAL DE- EVOLUTION, BIOLOGY, then used to calculate and Y*,, holding BI con- PARTMENTOF IOWA STATE UNIVER- Y*I STATISTICS, stant. The randomly generated response matrices 315F SNEDECOR IOWA SITY, HALL, AMES, 50011; were used in the model Y*i = X2B2 + E2, where X2 AND DEPARTMENTOF UNIVER- (JMN) BIOLOGY, contained effects and size covariates coded for pop- SITY OF SOUTH 189 CHURCHILL DAKOTA, ulation by sex groups. Random shape vectors and al- HAINES, SOUTH DAKOTA 57069. VERMILION, lometry vectors were calculated from the two sets of E-mail: (MLC) [email protected]; (JMN) data, as well as 0 between corresponding vectors (i.e., and [email protected]; (CAS) craig.stockwell@ 0 was calculated as the angle between two random Send to ndsu.nodak.edu. reprint requests versions of the same vector). MLC. Submitted: 18 Dec. 2003. 30 Accepted: We performed 4999 permutation iterations to gen- Oct. 2004. Section editor: M. E. Douglas. erate distributions of random angles, holding popu- lation and sex effects constant. This allowed us to test APPENDIX I the null hypothesis (no difference in direction be- tween observed vectors) by calculating the probability STATISTICALDETAILS that distributions of randomly generated angles con- The multivariate linear model used to describe tained observed angles. For every vector comparison, > to infer the shape vector relationships was of the form Y = XB + we calculated P,,and=P(Orandom observed) E, where Y is the matrix of partial warp scores and significance of directional differences in vectors. Two uniform scores, X is the design matrix coded for the assessments were performed for every comparison: effects of interests (e.g., sex, population, and size), B Oobse~vedwas calculated for two vectors; and 0randomwas is a matrix of regression coefficients, and E is a matrix measured for one vector (i.e., observedwas compared to of residuals. From the multivariate model, vectors of two distributions of Orandom). Any observed angle was (population by sex) least-squares means, y, were cal- not considered significant if we failed to reject the < culated. A shape vector, Ay, describes a difference in null hypothesis (at Pra,,d 0.05) for either random two group vectors (e.g. between female and male distribution.