See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/230013606 Assessment of rampant genitalic variation in the spider genus Homalonychus (Araneae, Homalonychidae) Article in Invertebrate Biology · March 2009 DOI: 10.1111/j.1744-7410.2008.00157.x CITATIONS 11 READS 54 1 author: Sarah C. Crews California Academy of Sciences 25 PUBLICATIONS 249 CITATIONS SEE PROFILE All content following this page was uploaded by Sarah C. Crews on 30 January 2014. The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document and are linked to publications on ResearchGate, letting you access and read them immediately.
20
Embed
Assessment of rampant genitalic variation in the spider genus … · 2017. 7. 12. · Assessment of rampant genitalic variation in the spider genus Homalonychus (Araneae, Homalonychidae)
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Assessment of rampant genitalic variation in the spider genus Homalonychus(Araneae, Homalonychidae)
Sarah C. Crewsa
Division of Organisms and Environment, University of California, Berkeley, California 94720-3114, USA
Abstract. Animal genitalia are often complex and thought to vary little within species butdiffer between closely related species making them useful as primary characters in speciesdiagnosis. Spiders are no exception, with nearly all of the 40,462 (at the time of this writing)described species differentiated by genitalic characteristics. However, in some cases, the gen-italia of putative species are not uniform, but rather vary within species. When intraspecificvariation overlaps interspecific variation, it can be difficult (if not impossible) to place a nameon a specimen. The quantification of shape variation in genitalia has not often been at-tempted, probably because until recently it was not a methodologically and computationallysimple process. In the two currently recognized species of the spider genus Homalonychus,genitalic variation is rampant in both male and female structures, with some parts of thegenitalia (e.g., the retrolateral tibial apophysis) differing in each specimen examined. In thisstudy, geometric morphometric analysis employing landmark data is used to quantify bothintra- and interspecific variation in this genus. The large amount of variation is condensedinto two or three groups depending on the structures examined, and these groups correspondto either the two species or to previously established mitochondrial DNA clades within one ofthe species. The results also show that analyses of female structures do not separate thegroups as readily as the analyses of the male structures. The large amount of variation presentin some structures is not correlated with geography or population genetic structure.
Genitalia are complex structures and are thoughtto evolve rapidly and divergently due to sexual selec-tion (Eberhard 1985; Hosken & Stockley 2004). Ifthis is true, genitalia will likely be the first, and some-times only, morphological characters to differ be-tween recently diverged species, and thus usinggenitalia to diagnose species is widespread in arthro-pod systematics (Borror et al. 1992; Foelix 1996).However, genitalia can often be variable in form, al-though typically variability within species does notexceed that between species (Eberhard 1985).
Spider genitalia are composed of complex struc-tures, and usually the precise function of these struc-tures is unclear (Sierwald 1990; Huber 1995). Ingeneral, it is thought that spider genitalia are quitestable in structure across species (Foelix 1996). This isonly a generalization, and intraspecific variation of
spider genitalia has been observed (e.g., Vlijm &Dijkstra 1966; Hippa & Oksala 1983; Roth 1984;Roberts 1987; Jocque 2002). The variation can besubtle or quite pronounced, so much so that it can bedifficult to confidently identify specimens to speciesonly using genitalia (Vlijm & Dijkstra 1966; Huber &Perez-Gonzalez 2001). It is not completely clear whysuch distinct variation would occur, though it couldbe due to population divergence, incipient speciation,or selection, and it is also a possibility that the vari-ation could be random, or due to sclerotization of thegenitalia after the final molt.
The genus Homalonychus MARX 1891 currentlycomprises two species, Homalonychus theologusCHAMBERLIN 1924 and Homalonychus selenopoidesMARX 1891. These are non-web-building spiders in-habiting the arid regions of the North Americansouthwest (Fig. 1). They are found under rocks orother debris, on hillsides or in washes, and possessspecialized setae that allow them to attach fine soil tothemselves, presumably to aid in camouflage (Dom-ınguez & Jimenez 2005; Duncan et al. 2007). Follow-
Invertebrate Biology 128(2): 107–125.
r 2009, The Authors
Journal compilation r 2009, The American Microscopical Society, Inc.
DOI: 10.1111/j.1744-7410.2008.00157.x
aCurrent address: Western Australia Museum, Locked
Fig. 1. Map showing distribution of Homalonychus species and collections used in this study. Numbers correspond to
those shown in Appendix A. Circles represent the western species H. theologus and diamonds represent the ‘‘eastern’’
speciesH. selenopoides. Arrows indicate areas Roth (1984) thought to harbor most of the genitalic variation present in the
group. AZ, Arizona; BC, Baja California; BCS, Baja California Sur; CA, California; NV, Nevada; SON, Sonora.
108 Crews
Invertebrate Biologyvol. 128, no. 2, spring 2009
ing Roth (1984), members ofH. theologus range fromCabo San Lucas, Baja California Sur, north to InyoCounty California, and east near Jean, Nevada (Fig.1). Members of H. selenopoides range from SonoraMexico, north through Arizona, southern Nevada(north of Jean to Mercury), and west into InyoCounty, CA (Fig. 1). In order to differentiate thetwo species of Homalonychus, Roth (1984) used gen-italic characters. These include the median and laterallobes of the epigynum, and the sperm ducts of theinternal genitalia; and on the palpi of the male, theretrolateral tibial apophysis (RTA), cymbial process,conductor, and embolus are used. Roth (1984) men-tioned and illustrated much variation in the epigynain H. selenopoides and suggested most of the varia-tion occurs in two geographic areas where the twospecies may come into contact (eastern ImperialCounty, CA and western Yuma County, AZ; NyeCounty, Nevada NV plus Inyo County, CA and SanBernardino County, CA and Clark County, NV)(Fig. 1), and suggested there was no variation in thepalpus of males from the same area. This observationis in stark contrast to what might be expected under astrict ‘‘lock-and-key’’ (Dufour 1844; Mayr 1963)mating system, because in areas where the two spe-cies overlap there should be less variability in order toprevent mating between different species. While astrict lock-and-key mating system has been demon-strated empirically in one specific case (Sota &Kubota 1998), several studies have shown thatarthropod genitalia generally do not conform to thelock-and-key hypothesis (Mayr 1963; Eberhard 1985and references therein; Opell & Ware 1989; Porter &Shapiro 1990; Arnqvist 1997). Also, it is unlikely thetwo species of Homalonychus come into contact inthe southern part of the range, as the Colorado Rivercurrently acts as a geographic barrier to gene flow,although they must in the north (Crews & Hedin2006; unpubl. data). There is no mention in Roth(1984), or elsewhere, of variation in H. theologus.After examining 4300 specimens, I have observedconsiderable variation among the epigyna in H. the-ologus, as well as in the RTAs and cymbial processesof males of both species (Fig. 2). Not only does vari-ation within the two species occur, but there is also aseemingly unusual degree of variation among speci-mens from the same locality.
In this study, geometric morphometric analyses(statistical analyses of shape) were used to quantifygenitalic variation in the spider genusHomalonychus.First, I ask if one sex is more variable than the other.Second, I ask if particular structures of the genitaliaof both sexes vary more than others. This is impor-tant because, as mentioned above, the exact role of
certain genitalic structures of spiders is unknown.Relative variation of male and female genitalic struc-tures may provide some clues to the importance ofthese structures. Finally, I ask if there is any corre-spondence between genitalic variation and popula-tion genetic structure, which has been shown to belinked to geography (Crews & Hedin 2006).
There are several advantages of using geometricmorphometrics as opposed to traditional morpho-metrics or linear measurements. A key reason is thatgeometric morphometric analyses present complexquantitative shape differences in an easily visualizedmanner (Zelditch et al. 2004). Also, when measuringthe distances between two points, there is redundancyassociated with these measurements that can weakenthe power of statistical tests (Adams & Funk 1997). Athird reason is that linear measurements are onlycapturing the affine or uniform components of shape,that is, the shape changes that leave parallel linesparallel and do not involve bending. Geometric mor-phometrics allows the non-affine components ofshape change to be examined, and these componentshave been demonstrated to contain more taxonomicinformation in certain organisms (Rohlf et al. 1996).Finally, geometric morphometrics can control forsize, scale, and position in ways linear measurementscannot.
Methods
Adult female (241) and adult male (92) specimenswere borrowed from the American Museum of Nat-ural History (AMNH), the EssigMuseum at the Uni-versity of California (UC) Berkeley, UC Riverside,and the California Academy of Sciences (CAS), orare new collections (now deposited in the NationalMuseum of Natural History in Washington, DC).Digital images were acquired using a Nikon Coolpix990 digital camera (Nikon Inc., Melville, NY, USA)attached to an Olympus SZX12 dissecting micro-scope (Olympus America Inc., Center Valley, PA,USA). One image was captured of the epigynum, andone image of the internal genitalia. Only new collec-tions and the majority of specimens from the CASwere dissected, so only 132 images of the internalgenitalia were captured. Genitalia were cleared inclove oil or lactic acid, and remaining tissues weredissected away using forceps and an insect pin. Oneimage was captured of the left male palpus in dorsalview, one in the retrolateral view and one in lateralview. If the left palpus was unavailable, an image ofthe right palpus was captured and with the aid ofAdobe Photoshop (Adobe, San Jose, CA, USA) thisimage was horizontally inverted. A number (SCC
Analysis of genitalic variation 109
Invertebrate Biologyvol. 128, no. 2, spring 2009
Fig. 2. A sample of genitalic structures illustrating rampant variation. A. Median lobes of epigyna. B. Lateral lobes of
epigyna. C. Cymbial processes of left palpus. D. Retrolateral tibial apophyses of left palpus. H. t.5Homalonychus
theologus, the western species; H. s.5Homalonychus selenopoides, the eastern species.
110 Crews
Invertebrate Biologyvol. 128, no. 2, spring 2009
001–SCC 296) was placed in each vial for voucherpurposes. Fine sand in a dish with alcohol was usedto stabilize the genitalia and care was taken to alignthe genitalia in the same plane.
Morphometric and statistical analyses
Analyses consisted of scoring landmarks (Bookstein1991), a Procrustes analysis (Rohlf & Slice 1990), anaffine, partial warp analysis based on the thin-platespline (Rohlf et al. 1996), and a non-affine relativewarp analysis (RWA) (Rohlf 1993). These methodsare reviewed in detail in Adams et al. (2004), Zelditchet al. (2004), and references therein. First, to placelandmarks on the images, tpsDIG32 version 1.31(Rohlf 2002b) was used. On structures that are highlyvariable, locating homologous landmarks can berather difficult, but care was taken to ensure that thechosen landmarks were homologous in the sense thateach landmark is unambiguous. Bookstein (1991) de-fined landmarks as Type I–III, where Type I land-marks, or landmarks placed in areas where threestructures meet, are the only truly homologous land-marks. Type II landmarks include the end of a processor points of maximum and minimum curvature. TypeIII landmarks are defined by their distance from otherpoints. All three types were used in the current study.
In this study, I examined the median and laterallobes of the genitalia of the female and internally, aportion of the sperm ducts. This portion consists ofthe external openings of the sperm ducts to wherethey first disappear behind the spermathecae in ven-tral view (Fig. 3). The entire duct was not used be-cause in order to see it in its entirety, thespermathecae would need to be removed, and thusthe ducts would be damaged. The female genitalia orepigynum of entelgyne spiders consists of the primarygenital opening and infoldings that constitute thesperm ducts and spermathecae (Foelix 1996). Thesperm ducts are tubes through which the spermpass on their way to the spermathecae (Foelix 1996).
On the palpi, or male genitalia, I examined theembolus, median apophysis, cymbial process, andRTA. The precise functions of the many parts ofthe male genitalia of entelegyne spiders are poorlyknown and may not be homologous across taxa(Coddington 1990; Sierwald 1990; Griswold et al.1998, 2005; Agnarsson et al. 2007). Homology ofboth the median apophysis and cymbial process ishighly in doubt and their functions are unknown.The embolus is the structure from which the spermduct opens and transfers the sperm to the female. TheRTA is thought to guide and stabilize the male gen-italia during copulation (Sierwald 1990; Huber 1995).
Landmarks on each of the structures used in thisstudy are shown in Figs. 3A–4C. To ensure less am-biguity as resolution may be lost in the figures, a de-scription of the chosen landmarks is presented inTable 1. To ensure the female genitalia were alignedin the same plane, I conducted three separate ana-lyses on the landmarks of the median lobe: one usinglandmarks 1–5 on the entire lobe, another using land-marks 1–3 for the left half of the lobe, and a thirdusing landmarks 3–5 for the right half. Because allthree analyses produced the same results, I concludedthe genitalia were aligned in the same plane and theresults presented are from the analyses using land-marks 1–5. Similarly, different combinations of land-marks were used on the RTA, embolus, and medianapophysis, and always produced corresponding re-sults to those presented in this study.
Files containing the images with the chosen land-marks were imported into tpsRelw version 1.26(Rohlf 2002a). First, a tangent configuration of allthe images from each analysis was computed using ageneralized Procrustes analysis based on generalizedleast squares (GLS) following Gower (1971), butmodified as in Rohlf & Slice (1990). The construc-tion of the tangent configuration is equivalent to thesuperimposition of each corresponding landmark ofeach image on top of the other, and the distance be-tween the centroids, or gravitational centers, of eachcorresponding landmark is minimized using GLS.Once this tangent configuration of landmarks is com-puted, specimens were then compared with thisaverage, or consensus, configuration to analyzevariation in shape. The Procrustes superimpositionmethod has drawbacks, such as distributing differ-ences localized at a few landmarks over all of thelandmarks, but this is circumvented if shape change isnot localized at any particular landmark(s) (Rohlf &Slice 1990).
Next, the partial warp and RWAs were conducted.Because the program utilizes the thin-plate spline(Bookstein 1991), both uniform and non-uniformcomponents of shape change may be visualized. Uni-form or affine shape change leaves straight linesstraight and parallel lines parallel. This type ofchange is global and can be thought of as all land-marks being displaced at the same ‘‘rate’’ relative toothers. Non-uniform or non-affine changes are localchanges involving bending. This can be thought of ashow fast the shear changes relative to all local indi-vidual landmarks. The RWA is a principal compo-nents analysis (PCA) of shape space used to describevariation within the samples. Bookstein (1991) intro-duced a, a scaling parameter that can be appliedwhen calculating the relative warp scores. An a of 0
gives equal weight to all partial-warps, so large- andsmall-scale differences of shape change are equiva-lent. If one wishes to place more emphasis on large-scale or small-scale changes, a can be set above orbelow 0, respectively (Bookstein 1991; Rohlf 1993).In this study, RWAs were preformed with a set to 0,making it equivalent to a PCA. Hotelling’s T2 andcanonical variants analysis (CVA) were used to assesswhether differences in shape were significant andcluster analyses were also performed to examine theexistence of any correlations between geography andthe shapes of the genitalia. Hotelling’s T2, CVAs andthe cluster analyses of the male genitalic structureswere completed using the program PAST v1.06(Hammer et al. 2001). SPSS v. 10.0 (1999) was usedfor the cluster analyses of the female genitalia, asPAST v1.06 is not able to perform these analyseswith 4210 samples.
Results
Differences in the shape of the genitalic structuresexamined were so great that plots of relative warp 1against relative w distinguished two clusters (Fig. 5)corresponding to the western species and the ‘‘west-ern’’ species (Homalonychus selenopoides and Ho-
malonychus theologus, respectively). No furtheranalyses on the sperm ducts were conducted becausethe clusters present from the RWAof the sperm ductsand median lobe were redundant, as it seems that theshape of the examined portion of sperm ducts is de-pendent upon the shape of the median lobe or viceversa. There were fewer samples (132/241) for theanalysis of the sperm ducts, but the group separationwas consistent with the analysis of the median lobes.The least amount of separation of morphologicalclusters was found in the lateral lobes of the epigyna(Fig. 5B), while the greatest amount of separationwas found in the median lobes of the epigyna (Fig.5A) and the emboli (Fig. 5C). The amount of varia-
11
22
3344
5566
77
A
11
22
33
4455
66
77
88
B
12
34
C
11
2233
44
55
66
77
99
88
D
Fig. 4. Landmarks on male genitalia. Numbers refer to
landmarks listed in Table 1. A. Embolus. B. Median
apophysis. C. Cymbial process. D. Retrolateral tibial
apophysis.
55
44
33
22
11
A
11
2233
44
55 66
B
11
22
33
55
6644
C
Fig. 3. Landmarks on female genitalia. Numbers refer to
landmarks listed in Table 1. A. Median lobe. B. Lateral
tion explained by the first two and three axes of thePCA for each structure is shown in Table 2.
To determine whether shape changes were due touniform or non-uniform components of shapechange, these components were analyzed separatelyfor each structure. Morphological clusters corre-sponding to eastern versus western species can bedistinguished in the PCA plots of the uniform andnon-uniform components (Fig. 6). Separation ofclusters is present in the plots of the uniform compo-nents of all structures although less so in the cymbialprocess (Fig. 6E), and a lot of overlap is present inthe plot of the lateral lobes (Fig. 6B). For the non-uniform component, separation of morphologicalclusters is present in all structures except the laterallobes (Fig. 6B), the median apophysis (Fig. 6D), andthe RTA (Fig. 6F).
Hotelling’s T2 was used to assess whether signifi-cant differences in shape exist between the two spe-cies. This test is conducted on theW or weight matrixfrom the partial warp analysis. Hotelling’s T2 re-quires a priori groups be defined. These two groupswere chosen based on collection locality and eithercorresponded to the ‘‘eastern’’ species (H. selenopo-ides) or the western species (H. theologus). Becausetwo distinct groups were already visible in the RWAof most structures, it is not surprising that the differ-ences in shape for these two groups were highly sig-nificant (Table 3).
The overall non-uniform shape deformations fromrelative warp axis 1 from the consensus configura-tions for both the median lobe and the embolus usingthe thin-plate spline are shown in Fig. 7. These twostructures display the greatest between-group differ-ences. The landmarks of both the embolus and me-dian lobe of the epigynum are more concentrated atthe negative extreme and more widespread at thepositive extreme. This results in a shorter, wider me-dian lobe and a more smoothly curving embolus atthe positive extreme. At the negative extreme, the
more concentrated landmarks result in a longer, nar-rower median lobe and a more abruptly curvingembolus.
Analyses of mitochondrial DNA (mtDNA) ofHo-malonychus (Crews & Hedin 2006) reveal three deep-ly divergent mtDNA clades. One corresponds to aneastern clade, or H. selenopoides, and the other twoare comprised of a H. theologus ‘‘north’’ clade (allsamples north of the mid-peninsular region of Baja)and a H. theologus ‘‘south’’ clade (all samples southof the mid-peninsular region of Baja). CVAs wereperformed to examine if three groups correspondingto the three mtDNA clades could be separated. Next,the eastern specimens were removed and PCAs,CVAs, and Hotelling’s T2 were performed on thespecimens comprising only the northern and south-ern groups.
A similar pattern of separation is present in theCVA plots for all of the structures (Fig. 8). There isvisible separation between the eastern and westerngroups, while near total overlap is present betweenthe western–northern and western–southern groups,and to a lesser extent, overlap between the western–southern and eastern groups. CVAs were alsoconducted on the northern and southern groups,excluding the eastern group for the uniform andnon-uniform components together as well as separately.
The results of the Hotelling’s T2 on the northernand southern groups are shown in Table 4. Becausethe increase in the number of comparisons can in-crease the likelihood of rejecting the null hypothesis,a Bonferroni correction procedure was applied tothese results, rendering all of them non-significant(p40.005) (results not shown), although the valuefor the median lobe approaches significance(p5 0.00594), and is significant for the non-uniformcomponent (p5 0.0015).
Cluster analyses were also performed on each ofthe structures to examine the presence of groups andwhether these groups corresponded to geography.Several different methods of clustering were per-formed as similar results from different types of an-alyses are thought to be superior to any one method.These clustering types included paired group Euclid-ean, average linkage between groups correlation,paired group Pearson correlation, paired group rhocorrelation, Ward’s method, single linkage Pearsoncorrelation, and single linkage rho correlation, andall methods produced equivalent results which aresummarized in Table 5.
Overall, structures that are visibly highly variable,like the RTA andmedian lobe (Fig. 2), are condensedinto two to three clusters when subjected to statisticalanalyses, such as RWAs and Hotelling’s T2. The lat-
Table 2. Amount of variation explained by PCA axes.
PCA, principal components analysis; RTA, retrolateral
Fig. 6. Plots of PC axis 2 against PC axis 1 for separate analyses of the uniform and non-uniform components of shape
change. A. Median lobes. B. Lateral lobes. C. Emboli. D. Median apophyses. E. Cymbial processes. F. RTAs. The first
plot represents the uniform component and the second plot represents the non-uniform component. The red crosses
correspond toHomalonychus theologus and the blue squares correspond toHomalonychus selenopoides. RTA, retrolateral
tibial apophysis.
Analysis of genitalic variation 115
Invertebrate Biologyvol. 128, no. 2, spring 2009
eral lobes, however, while also somewhat observablyvariable, do not separate into discrete groups and arealmost completely overlapping between species. TheCVAs show a general pattern to geographic structur-ing only at a large scale.
Discussion
Examination and statistical analyses of the genita-lia of Homalonychus indicate there is more interspe-cific differentiation in more structures of the malegenitalia than the female genitalia. This conclusion isbased on the fact that in general, statistical analysesindicate there is more overlap between the two spe-cies in the structures of the female genitalia than thestructures of the male genitalia, yet even these over-lapping clusters have significantly different means.
Shape differences between the eastern and westerngroups can be visualized in both the emboli and themedian lobes. However, even though significant differ-ences are present in the other structures, the differencesin shape between the two groups are not clear. This issomewhat surprising as the differences are highly sig-nificant for both the uniform and non-uniform compo-nents of shape change. Although there is overlap in thetwo groups shown in the plots of some of the structures,the differences between the means are clear.
In a discriminant analysis of the median lobes (un-publ. data), some specimens from localities not previ-ously reported as having members of Homalonychusselenopoides present are included in theH. selenopoidesgroup (144, SCC_129, Mexico, BCS, Isla Gaviota,Bahia de La Paz; 75, SCC_032, CA: San BernardinoCO, Pisgah Crater), while the same is true for H. the-ologus (44, SCC_020, AZ: Yuma; 37, SCC_023, AZ:Yuma CO, N Gila Valley; 172, SCC_041, Mexico,Sonora, 25mi W Sonoita; 15, SCC_252, AZ: N AZExperimental Station). In the males, however, themorphological clusters are entirely consistent withRoth’s species boundaries. It should be observedthat males were not available from all areas where
there are ‘‘misplaced’’ females, but only Pisgah and theYuma area (see Appendix A).
There are multiple explanations for the apparentmisplacement of female spiders into the ‘‘wrong’’groups based on the discriminant analysis. It couldbe that these spiders are representatives of the otherspecies, though this is unlikely. The spiders are sed-entary in general, and adult females have never beencollected wandering around, while males seem towander occasionally (Crews & Hedin 2006; unpubl.data), implying that one would expect to find malesthat fall in the ‘‘wrong’’ group as well. As for the areain Eastern Imperial County where overlap is sup-posed to occur (Roth 1984), many recent collectionshave been made here and members of H. selenopo-ides, as defined by genitalia, and nuclear andmtDNA(Crews & Hedin 2006), have not been collected. An-other explanation for the discordance in the genitaliaof females and their geographic locations could sim-ply be that the specimens were mislabeled. If the re-sults from the discriminant analysis are due tospecimens being mislabeled, then the methods pre-sented here are good for examining this aspect if thereare suspect specimens.
There are several explanations for why genitaliaare so variable in this group. It could be due to in-cipient speciation, divergence, or selection. It is alsopossible that genitalic variation could be random
Fig. 7. Plots of landmarks showing positive and negative
shape deviation from the means. A. Median lobes of the
epigyna, and B. emboli of the palpi. The landmarks are
more concentrated at the negative extreme and more
widespread at the positive extreme, reflecting a shorter,
wider median lobe and a more smoothly curving embolus
at the positive extreme; at the negative extreme the more
concentrated landmarks reflect a longer, narrower median
lobe and a more abruptly curving embolus.
Table 3. p-values from Hotelling’s T2 for eastern versus
western groups. RTA, retrolateral tibial apophysis.
Structure Uniform and
non-uniform
Uniform Non-
uniform
Median lobe 5.036E�37 4.333E�37 2.387E�17Lateral lobe 2.008E�17 1.51E�20 0.2257
with respect to geography, or due to the way epigynasclerotize after ecdysis. While the general mating be-havior of Homalonychus has been observed (Dom-ınguez & Jimenez 2005), it has not been observed indetail. An important next step would be to figure out
how the parts of the genitalia fit together during cop-ulation. This would be possible using liquid nitrogenfixation and careful dissection and/or thin sectioningfor electron microscopy, or simpler histological sec-tioning and light microscopy (e.g., Huber 1993).
Fig. 8. CVA plots. A.Median lobes. B. Lateral lobes. C. Emboli. D.Median apophyses. E. Cymbial processes. F. RTAs.
Red crosses correspond to northern populations of Homalonychus theologus, black circles correspond to southern
populations of H. theologus, and blue squares correspond to Homalonychus selenopoides. CVA, canonical variants
analysis; RTAs, retrolateral tibial apophyses.
Analysis of genitalic variation 117
Invertebrate Biologyvol. 128, no. 2, spring 2009
Many arthropod groups can be difficult to diag-nose based on somatic characters (Adams & Funk1997; Bond & Sierwald 2002; Crews & Hedin 2006;Skevington et al. 2007), or genitalic characters due tolarge amounts of variation (Jocque 2002; Paquin &Hedin 2004; Crews & Hedin 2006) or the opposite inwhich the genitalia may seemingly be invariable(Bond & Sierwald 2002). Many of these studies, in-cluding the current one, indicate that genitalic shapeis an important character as it is likely to evolve rap-idly (i.e., Huber et al. 2005) and offers yet anothertool that can be used to assign specimens to partic-ular taxonomic groups, such as species.
Geometric morphometric analyses of both themale and female genitalia in Homalonychus revealedstatistical differences in shape mostly consistent withthe species boundaries of Roth (1984), but no signifi-cant differences were revealed corresponding tomtDNA clades of ‘‘northern’’ and ‘‘southern’’ H.theologus (Crews & Hedin 2006), and there were nodetectable population-level differences. On a largescale, the shape of the median lobe seems to bemostly coupled with geography, while the male em-
bolus shape is perfectly correlated with geography.Thus, the rate of mtDNA evolution and of the malegenitalia appear to be correlated.
It is possible that biogeographic barriers have keptthe two species apart, and the genitalia now differ be-cause of isolation and subsequent evolution. But,again, there is no known barrier which separates thetwo species in the northern part of the range, althoughit is possible that populations of H. theologus becameisolated on the Baja Peninsula and have moved north-ward; thus, contact between the two species is pri-mary. While there were more refugia in the SonoranDesert than in the western deserts (Mojave and Col-orado) during the Pleistocene, differences on such afine scale are not seen with the sampling and analysesconducted here. Nevertheless, apparent random gen-italic variation is no longer problematic in this groupwhen attempting to delimit species. These methodscould become even more useful when more is knownabout homology and functions of the different parts ofspider genitalia. For instance, data from Huber (1995)suggest that the RTA is used to guide and stabilize thepedipalp in the epigynum. However, observationshows that nearly each specimen has a differentRTA, although some of the components (e.g., the firstand last ‘‘spike’’; Fig. 2) seem relatively static. Thissuggests that different parts of the genitalia may beunder different selection pressures. Without rigorousexperimentation it is unknown whether stability ofparticular structures of the male genitalia is due to se-lection or other factors, but the results presented hereare indicative of such a trend.
Acknowledgments. I would like to acknowledgeNorman Platnick and Lou Sorkin at the AMNH,Charles Griswold and Darrell Ubick at the CAS, CherylBarr at the Essig Museum of Entomology at UC Berkeley,and Rick Vetter at UC Riverside for providing specimensand allowing their dissection. I am especially grateful toWilliam Eberhard and Bernhard Huber for manyinsightful and thoughtful suggestions which drasticallyimproved this manuscript. I am also grateful to JoeSpagna, Pierre Paquin, Lindsey Leighton, and MarshalHedin for comments on various drafts of this manuscript.I also wish to thank the American Arachnological SocietyVince Roth Fund for systematic research for providingfunds.
References
Adams DC & Funk DJ 1997. Morphometric inferences on
sibling species and sexual dimorphism in Neochlamisus
bebbianae leaf beetles: multivariate applications of the
thin-plate spline. Syst. Biol. 46: 180–194.
Table 4. Non-Bonferroni corrected p-values from Hotell-
ing’s T2 for western–northern versus western–southern
groups. RTA, retrolateral tibial apophysis.
Structure Uniform and
non-uniform
Uniform Non-
uniform
Median lobe 0.000396 0.9118 9.77E�5Lateral lobe 0.4757 — —
Embolus 0.8617 — —
Median apophysis 0.002374 0.3366 0.00598
Cymbial process 0.01549 0.666 0.0294
RTA 0.005837 0.2902 0.02984
Table 5. Summary of results of cluster analyses. mtDNA,