In this paper, we present new genetic evidence based on an analysis of the Mitochondrial (mtDNA) cytochrome c oxidase subunit I (COI) gene for several hundred individuals of Nearctic Euchloe. We compare the genetic distances between taxa and use various clustering methods to construct a molecular cladogram to match with those previously deduced solely by morphological study. We compare and discuss the new results and revise the North American species accordingly. We assign the Nearctic species to groups previously erected by Back et al. (2008) for Palaearctic Euchloe. They defined six species groups of Euchloe butterflies based on morphological and molecular data. Species group I consists of E. falloui (Allerd) a very peculiar species of Northern Africa and the Near East. Species group II consists of several taxa related to E. creusa, which occurs in several subspecies from the Altai as north as Wrangel Island of Russia. Species group III was defined by Southern Mediterranean E. belemia (Esper) and related species of the Canary Islands. E. daphalis (Moore) is the representative of group IV, and E. ausonia (Hübner, 1804) defined species group V. Species group VI was named after E. crameri Butler, but also includes E. insularis (Staudinger) from Corsica and Sardinia, as well as E. melanochloros Röber from Northern Africa.

Six species of Nearctic Euchloe are currently recognized (Pelham 2008). The most recent revision by Opler (1966–1974) included only four species (i.e. E. ausonides, E. hyantis, E. olympia, and E. creusa), but, subsequently, Euchloe guaymasensis (Opler 1986), and Euchloe ogilvia (Back 1990), were described as new; later Euchloe ausonia naina, a Palaearctic subspecies of Euchloe ausonia, was reported from Alaska and the Yukon Territory (Layberry et al. 1998). Most recently, this subspecies was found in the Mackenzie Mountains, Northwest Territories (Chris Schmidt pers. comm.). Several authors have considered that E. ogilvia constitutes a subjective junior synonym of Euchloe ausonides (Layberry et al. 1988).

Several studies have made use of molecular markers to clarify taxonomic relationships of species in question (e.g. Hajibabaei et al. 2006). “DNA-barcoding,” which uses a portion of the mitochondrial cytochrome oxidase gene (COI) to detect population divergences, is becoming accepted as a library of DNA sequence information of all organisms on earth ( www.barcoding.si.edu). This approach revealed, for example, a group of cryptic species in what had been thought a single species of skipper (Astraptes fulgerator (Walch)) (Hebert et al. 2004). In whites (Pieridae), molecular analyses have been used to present molecular phylogenies (Pollock et al. 1998; Braby et al. 2006; Chew & Watt 2006). In a previous study we analysed Palearctic Anthocharidini butterflies on a broad scale and confirmed species groups; In addition, we encountered some hitherto undetected species (Back et al. 2008). Our objective here is to integrate Nearctic Euchloe into an existing species group concept, and to complete molecular investigation of Holarctic Euchloe.

Material and Methods

Insect material. Taxa selection was focused on the completion of an existing dataset (Back et al. 2008). Nearctic specimens were selected from the collections of Werner Back and John F. Emmel, as well as those of several institutions—C.P. Gillette Museum of Arthropod Diversity, Colorado State University; Essig Museum of Entomology, University of California, Berkeley, and the Los Angeles County Museum of Natural History. Other critical specimens were donated or purchased. Samples were stored in absolute alcohol, dried, or pinned and dried. In some cases, specimens had been rehydrated in a relaxer prior to spreading. In most cases, three legs were used for isolation and analysis of nucleic acids.

DNA extraction, amplification and sequencing. The extraction of sample DNA, amplification and sequencing was performed by kmbs ( www.kmbioservices.com). Tissue from adult individuals was used for the proteinase K procedure with DNeasy tissue kit (QIAGEN), following the manufacturer's protocol for animal tissues with slight modifications. Digestion was performed for 12 hours. Then, DNA was eluted with buffer AE. Prior to PCR amplification, the DNA concentration was measured using a Fluorometer (BioRad, VersaFluorTM ) and adjusted to 50 ng/µl. A 326 bp fragment of the mitochondrial (mtDNA) cytochrome c oxidase subunit I (COI) gene was amplified by PCR (Biometra T professional) using protocols and primers as in Back et al. (2008). Direct sequencing of dye-labelled templates (BigDye V 1.1, Applied Biosystems) was carried out using an ABI 377 XL automated sequencer (Applied Biosystems). Single strand sequences were used in both directions to generate a consensus sequence for quality control of electropherograms using Sequence Navigator V 1.0.1 (Perkin Elmer). The sequences were aligned to the reference sequence of Pieris napi (Linnaeus, 1758) and P. brassicae (Linnaeus, 1758). The new sequences are deposited at EBI/GenBank. The accession numbers are shown in Table 1.

Sequence data analysis. The pairwise genetic distances were calculated using the NJ-algorithm of PAUP software (Phylogenetic Analysis Using Parsimony 4.0 ß 10 version, Swofford 2003). These values were compiled to calculate the ranges of genetic distance between the subgroups and the outgroup, respectively, using the software PAST v. 1.35 (Hammer et al. 2005). For each sample, the 25–75 percent quartiles are drawn using a box. The median is shown with a horizontal line inside the box. The minimal and maximal values are shown with short horizontal lines.

Phylogenetic analyses were conducted using the Maximum Parsimony (MP) algorithm of PAUP, the settings of a heuristic search (branch swapping, tree bisection and reconnection (TBR) were 10.000 starting trees, random stepwise addition sequences. Within each heuristic search 10 replicates were performed. For ML it is essential to use the most fitting evolution model for a given dataset. To calculate this, we used Modeltest V. 3.7 (Posada & Crandall 1998). Calculations using the received evolution model were done by the software PhyML (Guindon & Gascuel 2003). Bayesian analysis (BI) was performed using MrBayes V. 3.1.2 (Ronquist & Huelsenbeck 2003) with default settings: two times four chains, MCMC, chain temperature = 0.2, number of generations = 5.000.000.

Table 1.

Sample material. Asterisks in column “Sample-ID” refer to individuals investigated earlier (Back et al. 2008).

Continued.

Continued.

Fig. 1.

Neighbor joining tree of Euchloe taxa. DNA sequences of mtCOI gene contained 326 bp each. Outgroup rooting was set to Papilio rutulus. Representatives of yet established species groups I, III, IV, and VI were only single specimens. CI was 0.494, RI was 0.817. Bar indicates 0.005 changes. Values show support of nodes with posterior probability values of Bayesian analysis/results of Maximum Likelihood: 1 — branch was found; 0 — branch was not found/Maximum Parsimony majority rule consensus. See text for settings of tree search algorithms.

Molecular data. We accepted sequences divergences of 2 % or more as evidence of species-level status. This was in part based on both our previous studies of Anthocharidini (Back et al. 2006a, b, 2008) and those of others for other groups (Sperling 2000). Intraspecific variability, i.e. oscillation of sequence divergence within a single population is below 1 %, while we arbitrarily assume subspecies level from 1 to 2%.

Results

All of the sample individuals yielded DNA suitable for PCR. Amplified fragments yielded sequences with the expected 326 base pairs (bp) in length each, and base composition of the resulting nexus file including outgroup specimens showed the bias of A and T as typical for insects (A: 0.28%; C: 0.18%; G: 0.13%; T: 0.41%). Of 326 bp 215 bps were constant, 84 of 111 variable positions were parsimony-informative which means a portion of 25.77% positions with phylogenetic impact.

Haplotype diversity. Haplotype identity occurred regularly and was detected even between individuals several hundred miles apart. Within clade II, we found one haplotype in E. olympia (n=3, HTeo1, ref. specimen WB217), one haplotype in E. guaymasensis (n=2, HTeg1, ref. specimen WB223), two haplotypes in E. creusa (n=9, HTec1, ref. specimen WB220, HTec2, ref. specimen WB247), and 11 haplotypes in E. hyantis (n=23, HTeh1, ref. specimen WB177, HTeh2, ref. specimen WB227, HTeh3, ref. specimen WB228, HTeh4, ref. specimen WB204, HTeh5, ref. specimen WB321, HTeh6, ref. specimen WB201, HTeh7, ref. specimen WB202, HTeh8, ref. specimen WB203, HTeh9, ref. specimen WB233, HTeh10, ref. specimen WB321, HTeh11, ref. specimen WB257.

Table 2.

Species groups of genus Euchloe. Hitherto known members of respective group (Back et al. 2008) and new results.

Clade V bears four different haplotypes within E. ausonides (n=13): HTea1, ref. specimen WB051, HTea2, ref. specimen WB314, HTea3, ref. specimen WB106, and HTea4, ref. specimen WB215. Of all newly investigated nearctic specimens (n=50), 19 different haplotypes in 5 nominal taxa were detected.

Distance analysis. Genetic p-distances were from 0.0% between specimens of one population to 17.9% between least related taxa. Mean genetic distances of outgroup specimens of other or the respective family was about 15%, distances to specimens of genera of the family Pieridae, tribe Anthocharidini was approximately 10% (Fig. 1). The least genetic distance was found between Euchloe and Iberochloe, the closest relative to genus Euchloe.

Fig. 2.

Box plots of genetic distances. Box 1: genetic distance of outgroup taxon Papilio rutulus (AF044888, Papilionidae) to Euchloe taxa. Box 2: genetic distance of Pieris napi (DQ148917, Pieridae) to Euchloe taxa. Box 3: genetic distance of Anthocharis cardamines to Euchloe taxa. Box 4: genetic distance of Elphinstonia charlonia to Euchloe taxa. Box 5: genetic distance of Iberochloe species to Euchloe taxa. Outliers are shown with circles or asterisks when they exceed the 95% confidence interval.

Fig. 3.

Box plots of genetic distances within Euchloe species groups. I vs. II, I vs. III, I vs. IV, I vs. V, I vs. VI, II vs. III, II vs. IV, II vs. V, II vs. VI, III vs. IV, III vs V, III vs VI, IV vs. V, IV vs VI, V vs. VI.

Based on our sequence divergence results, we accept five species of Nearctic Euchloe as clearly resolved and genetically defined. These species are E. ausonides, E. creusa, E. hyantis, E. olympia, and E. guaymasensis. Genetic distance analysis clusters these species clearly into the species groups defined by Back et al. (2008) (Table 2).

Genetic distances of outgroup taxa vs. Euchloe specimens gradually decreased with phylogenetic relatedness. Values seem to reach a maximum at approximately 15% sequence divergence, as there is no significant difference between the Papilio and the Pieris individuals when comparing p-values (Fig. 2). At the genus level, values vary around 10% sequence divergence, when comparing genera of the same tribe (Anthocharidini). Noteworthy is the increasing fluctuation in combination with increasing relation to the ingroup.

Within the selected taxa of Nearctic Euchloe we compared genetic distances of the respective groups. Distances ranged from 3.68% to 9.02% between groups. Mean values varied from 4.72% for the closest related group V and VI to 9.43% for the most distantly related group I and II (Fig 3, table 3).

Genetic variation within groups was calculated solely for groups II, and V, while further splitting of group II was performed to compare values of species or species complex. Based on the tree we assigned specimens WB217, WB218, WB240, and DQ148941 to E. olympia, WB223 and WB224 to E. guaymasensis, WB094, WB095, WB179, WB180, WB220, WB245, WB246, WB232, and WB247 to E. creusa, WB177, WB227, WB234, WB228, WB178, WB200, WB204, WB205, WB216, WB226, WB229, WB230, WB232, WB242, WB243, WB231, WB201, WB202, WB308, WB203, WB233, WB321, and WB257 to E. hyantis. Within group V, all nearctic Euchloe specimens (WB051, WB107, Wb213, WB214, WB236, WB238, WB239, WB254, WB255, WB314, WB322, WB106, and WB215) were assigned to E. ausonides.

Table 3.

Comparison of genetic distances between species groups.

Genetic diversity within species groups: group II. Species group II is comprised of E. olympia, E. guaymasensis, E. creusa, and the E. hyantis complex. Genetic diversity between these taxa varies from the closest between E. guaymasensis and E. hyantis (p= 0.03374233) to the most distant between E. olympia and E. hyantis (p= 0.06724877 ). Mean genetic distances between species is p= 0.04656292, which supports well differentiated species at the COI level (Fig. 4, Table 4).

Genetic diversity within species groups: group V. Species group V is comprised of E.ausonia, E. simplonia, E. ausonides, and E. persica. Genetic diversity was calculated between E. ausonides and their cospecies. P-values vary from p= 0.02760736 in closest neighbors (E. ausonia -E. ausonides, and E. ausonia -E. persica) to p= 0.0398773 in most distant ones (E. simplonia -E. ausonia) (Fig. 5, Table 5). Mean genetic distances between all species of this group is at p= 0.03441089, which supports the respective species distinctiveness.

Fig. 4.

Box plots of genetic distances of species within Euchloe species group II.

Phylogenetic analysis. Phylogenetic signal is found at several nodes of the NJ tree. The majority rule consensus tree of the 10 trees retained from MP search supported the monophyly of the genus Euchloe. All of the six groups were retained with highest support, with the exception of group II, which included group III and IV as sister group to a clade containing E. olympia and E. guaymasensis, respectively.

ML: The best fitting model of evolution as defined by Modeltest was the GTR+I+G model with a likelihood-value of -lnL = 1934.2101. Monophyly of Euchloe failed in ML analysis, when E. falloui clustered in a clade containing Elphinstonia, Anthocharis and Iberochloe. Besides this, all the other groups were retained and the topology was identical to the NJ tree.

Bayes: Bayesian inference confirmed monophyly of Euchloe with E. falloui as sister to all other Euchloe. Topology resembles most the NJ tree with the exception of an unresolved basal trichotomy, containing group IV, a clade with group V and VI, and a clade with group II and III.

Fig. 5.

Box plots of genetic distances of species within Euchloe species group V.

Table 4.

Comparison of genetic distances between species of group II.

Discussion

We realize that use of the COI barcode method is most suitable for the discrimination of genetic divergence of genera, species and subspecies, and one should not make firm conclusions with only a mitochondrial gene about phylogenetic relationships and lineages. Nevertheless we feel our analysis is novel and somewhat informative and may stand until more inclusive studies are performed. The limitation of this type of analysis was shown in our earlier paper (Back et al. 2006b), as the magnitude of sequence divergence was lower between E. bazae Fabiano and E. penia (Freyer) and not, as had been expected, between E. bazae and E. charlonia (Donzel). On the other hand, separation of putative species was considerably higher than that for other butterflies, and in general agreement with the magnitude of divergence found in earlier studies (e.g. Pollock et al. 1998).

Taxonomy. The Nearctic Euchloe are included below in species groups II and V of the extant cladogram for Palaearctic Euchloe (Back et al. 2008). In the following paragraphs, we will discuss how these placements agree or disagree with the species groups presented by Opler (1971) and others which were based on both larval and adult phenotypic characters, biology, and ecological occurrence.

Table 5.

Comparison of genetic distances between species of group V.

Species group II. This group is represented by the Holarctic E. creusa. When investigating individuals from Siberia it was noticed that there was no genetic diversity at all between populations at the investigated COI fragment, even though the dispersion of populations exceeded ranges of several 1000km and were as distant as Wrangel Island (Russia). In the present project we found that individuals from both British Columbia and Alaska fit perfectly into this previously described cluster, suggesting relatively recent trans-Beringian genetic exchange between these populations. An exception was an Alaskan individual (WB247) that showed some genetic divergence, but we considered it within the range of intraspecific sequence divergence.

The placement of E. creusa in this species group differs from the results of Opler (1971) who placed E. creusa with E. ausonides and E. olympia in the Ausonides species group based on adult characters: white scales in forewing discal black bar, valval shape, and juxta configuration. Moreover, both E. olympia and E. ausonides have nearly identical larval color patterns (Opler 1974).

Euchloe hyantis is another distinct species of species group II. Its association with species group II of Euchloe is not entirely unanticipated, because the butterflies show phenotypic similiarities to E. creusa, leading to an occasionally proposed conspecificity of these two species (Talbot 1932–1935). Because of phenotypic similiarities to Iberochloe tagis (e.g. the adult appearance, univoltinism, larvae green with violet and white lateral stripes) it would not have been surprising, if E. hyantis, as well as E. creusa, were assigned to the newly erected genus Iberochloe Back, Knebelsberger & Miller, 2008. The genetic data strictly rejects this with high support, as the genetic distance is close to 10%. However, it appears that larval color pattern is relatively unstable and should not be used for phylogenetic considerations. A case in point is shown by the larvae of closely related species Anthocharis belia (Linnaeus) and A. euphenoides (Staudinger), which show a similar range of variability (Back 1977).

We investigated the Euchloe hyantis group or complex most intensively, because apparent local variation, phenotypic differences, larval host use, and adult behavior all suggested that more than one species might be involved (Opler 1971, unpublished; J. Emmel personal communication). Many individuals throughout most of the range were selected for genetic analysis. However, we found only slight genetic differences between the populations that we examined with our fragment of COI. For example, we tried to discover whether the separation E. hyantis and E. lotta is justified (Beutenmüller 1898; Opler 1999). We found that close to nomenotypical individuals of E. hyantis from Napa Co., California, (type locality Mendocino, Mendocino Co., California) are of an identical haplotype as one individual E. hyantis lotta from near Glenwood Springs, Garfeld Co., Colorado (WB243). We refer this Colorado individual as closest to topotypical E. lotta (Kanab, Kane Co., Utah) (Opler, 2008).

Interestingly, a series of 20 individuals from a population at Eureka, Juab Co., Utah, had the forewing black discal patch especially broad in both sexes and a noticeable clear white base color, differing slightly from Arizona, Colorado, and Nevada populations, which possess a more creamy white dorsal ground color. These individuals closely resemble the lectotype of E. h. lotta from Kanab. The genetic differences are less than 1% (0.92%) and within our standard for subspecific range of sequence diversity. We consider this population to be within the range of topotypical lotta (Kanab, Utah) because of both its geographic and apparent genetic proximity. It is possible that some other populations that occur east of the Cascade — Sierra Nevada crest may eventually be described as subspecies as there is notable variation in some regions.

Likewise, the population from Jacumba, San Diego Co., California, close to the Mexican border (Peninsular Ranges Segregate) showed small sequence divergence from other E. hyanthis/E. lotta. Since some populations are found along the western edge of the Colorado Desert, one might consider these to be E. h. lotta, however, extending upward from the desert, this population blankets the east side of the Laguna Mountains to its summit. These have been considered closest to nomenotypic E. hyantis by Opler (1971, 1999) because of their narrow black discal patches on the forewings and the lack of pearly sheen in the white areas on the ventral hindwing.

In contrast, the two specimens of E. hyantis andrewsi were identical to the most common haplotype of hyantis/lotta-specimens. Somewhat more distinct are the populations found in southeastern Arizona. These belong to the E. h. lotta phenotype, but their forewings have much narrower black discal bars (Opler 1971). There seems to be a cline extending northward as the average bar width gradually increases as one approaches southern Utah. The Arizona specimens are 0.6–1.8% different from all the remaining E. hyantis populations.

Some populations of Euchloe hyantis in southeastern Arizona, in fact, show lowered proportion of green marbling on the ventral hindwings, reminiscent of Euchloe guaymasensis. This is noticeable in the Mule Mountains of Cochise County and the Santa Rita Mountains of Santa Cruz County, and may represent introgression from populations of E. guaymasensis in Sonora, Mexico. In fact, one population located west of Cuitapa, Sonora, Mexico, appears to be a hybrid swarm between E. hyantis lotta and E. guaymasensis. Most individuals in this population appear widely variable between the two parental phenotypes. We included a few individuals hyantis lotta x guaymasensis from this population in our analysis. In this area the ranges of E. hyantis lotta and E. guaymasensis are now parapatric. We believe this represents secondary contact of the species intermediated by the spread of weedy Descurainea pinnata (Walt.) Britt. (Brassicaceae), which thrives in historically overgrazed, disturbed habitats. Further south in Sonora, E. guaymasensis uses only Dryopetalum runcinatum A. Gray var. laxiflorum Rollins, a native crucifer, even though Descurainea pinnata is present at the same sites where we found no evidence of its use (Brock & Opler unpublished). According to the results of maternally inherited COI the hybrids cluster within the hyantis-clade from SW-Arizona and from the Sonoran Desert. An investigation of nuclear Histone3-gene (H3) revealed, that, besides the slower mutation rate of this gene, the genotype of the hybrids is identical to those of E. hyantis and E. guaymasensis, respectively. This result points towards the actual hybrid identity of these two species.

Because of the sensitivity of mtDNA sequences with respect to recognition of distinct species, the consequences of the results are the following. E. hyantis is a genetically robust species, and gene flow between populations is still possible and visible. However, because of the phenetic, biological, and spatial occurrence we feel that recognition as separate subspecies is justified, i.e. hyantis, lotta, andrewsi being quite reasonable. It is possible that when additional genes are sequenced, these taxa may in fact foster their genetic distinctiveness. Almost all populations of E. hyantis from the Pacific Divide and westward feed solely on Streptanthus species as their larval hosts; Streptanthus tortuosus Kellogg (Brassicaceae) is used in most instances (Opler 1974). To the east of the Pacific Divide, populations assignable to E. h. lotta are catholic in their host use and several unrelated genera of Brassicaceae are selected and consumed. Euchloe h. andrewsi, first decribed by Martin (1936) as a subspecies of Euchloe ausonides, is endemic to a small area of the San Bernardino where its sole host is the endemic Streptanthus bernardinus (Greene) Parish (Brassicaceae) (C. Henne personal communication). The large size and phenotype of E. h. andrewsi is distinctive and shows no evidence of blending with adjacent populations of E. h. lotta or E. h. hyantis, which occur nearby in the same mountain range. For that, some authors, including Opler, argue that this population looks and behaves like a species separate from other hyantis populations (Chris Schmidt pers. comm.) and therefore should be treated as distinct species.

The fact that E. guaymasensis also belongs to species group II had been expected due to the phenotype (butterfly, caterpillar, chrysalid [Brock in Interactive Listing of Mexican Butterflies]), but the distinct gap of 3.7% from the closest species (E. olympia) shows that its separation from the other Euchloe is of long-standing. However, these butterflies, despite their present-day long distance disjunct occurrence of roughly 1400 km, bear a slight similarity to E. olympia. In particular, both share the evanescent rosy pink shade along the costal margin of the ventral hindwing, have a sparse distribution of green marbling on the ventral hindwings, and a convex forewing outer margin. Moreover, no other Euchloe species share the rosy-pink iridescence due to an unstable pteridine pigment that changes after death.

Euchloe olympia also belongs to species group II, according to our results. This was surprising, as was the relatively high degree of similarity to E. guaymasensis, especially since the shape and color pattern of the caterpillar and chrysalid pointed towards inclusion with species group V. In fact, Opler (1971) had E. olympia as part of the ausonides species group. Given the short length of the COI fragment used in our study, such association may not be justified. A comparison with a more conservative, nuclear DNA (H3) shows, in contrast, that E. olympia is in fact more similar to E. ausonides (data not shown). Apart from that, the three samples are concordant despite their different localities. Likewise, the GenBank COI sequence DQ148941 from Larimer Co., CO (Chew & Watt 2006), shows only minor sequence divergence. Moreover, in a study of Euchloe olympia by Opler & Clench (1983), detailed phenetic analysis argued against the recognition of any subspecies throughout its range.

In contrast to this apparent relationship are other characters shared between E. olympia and E. ausonides including identical larval color pattern and genitalia features. The existence of individuals showing intermediate phenotypes between E. olympia and E. ausonides from Jefferson Co., Colorado and Natrona Co., Wyoming suggests periodic matings between these species (C.P. Gillette Museum, Colorado State University). These seem to be rare individuals and no evidence of retrogression is seen in either parent population.

Species group V. Our results strongly support the inclusion of E. ausonides in species group V. Scott (1986) included E. ausonides as a subspecific taxon under E. ausonia, apparently due to its superficial similarity, but our analysis shows it separated by at least 2.76% of genetic divergence in our fragment of COI.

The molecular analysis of numerous E. ausonides throughout its distribution resulted in a unitary outcome, similar to that for E. hyantis. The maximum value of sequence divergence was 1.2%. The main grouping, consisting of specimens from several localities from Yukon (ogilvia), Sierra Co., California (transmontana), Manitoba (mayi), British Columbia and Alberta, respectively, showed the same haplotype. A small magnitude of divergence from the above cluster was encountered solely in specimens from Mariposa County, California and Sausalito, Marin County, CA (0.3%), as well as individuals from Jefferson County, Colorado and Whitman County, Washington (0.9–1.2%).

We interpret these differences to be infrasubspecific genetic divergences and point mutations, leading to the assumption that E. ausonides populations have not been isolated since their occurrence and only slight genetic and phenotypic divergence have occurred to date. However, discrimination at the subspecific level because of distinct phenotypic characters of the four infraspecific taxa E. a. ausonides, E. a. mayi, E. a. coloradensis and E. a. ogilvia (e. g. different shape of chrysalid in E. a. ogilvia: Back 1990) seems reasonable to us. The validity of the subspecies E. a. insulana Guppy & Shepard and E. a. paleoreis K. Johnson was not tested as no individuals from these described taxa were included in our study.

Mating barriers. Behavioral mating strategies in butterflies have been detailed by Scott (1974, 1986) as perching or patrolling, and it is clear that all Euchloe, in common with all temperate Pierinae, have males that patrol in search of mates. Overlain on these mate location modes is the topographic character on which mate location takes place. In western North America, a clear division between use of peaks and ridges versus drainage courses may be made (Scott 1974). Shields (1967) detailed hill-topping behavior of many butterfly species at one site in California. In Anthocharidini, mate-seeking takes place almost entirely before noon, and the individuals wander more widely afterwards. Newly emerged females of hilltopping species fly to ridges and peaks for mating (Opler personal obs.), while females of drainage followers may be more random in orientation.

Euchloe populations and species may also be characterized as either hilltopping or drainage following. Euchloe olympia throughout its range is a ridge and hilltop follower, while Euchloe ausonides is a drainage follower in almost all cases. These two species are broadly sympatric and synchronic along the foothills of the Rocky Mountain front in Colorado, Wyoming, and Montana. Since these two species may hybridize (as shown by rare intermediates), we feel that the prime importance of the disparate mate-location venues is a mostly effective mating barrier. After mating, females scatter and seek nectar sources and, most importantly, their respective hosts for oviposition. In the case of the above species pair, oviposition choices are primarily Arabis species, although other Brassicaceae may also be selected (Opler 1974; Opler & Clench 1983).

For other North American Euchloe, hilltopping is the primary mate location strategy we have observed (unpublished), especially for Euchloe hyantis and northernmost Euchloe guaymasensis populations. Exceptions are some populations of Euchloe hyantis in the western Sierra Nevada foothills where males patrol the stands of their Streptanthus (Brassicaeae) hosts along steep cliffs (Shapiro pers. comm.), and more southern populations of Euchloe guaymasensis whose males patrol shaded hillsides (Brock & Opler pers. obs.). In northern Sonora where both Euchloe hyantis lotta and Euchloe guaymasensis use hilltopping behavior, the result may be open mating and hybrid swarms. We have no information on the mate-location strategy of Euchloe creusa.

Conclusions: