Content uploaded by Axel Meyer
Author content
All content in this area was uploaded by Axel Meyer
Content may be subject to copyright.
q
2005 by the American Society of Ichthyologists and Herpetologists
Copeia, 2005(2), pp. 320–339
Historical Biogeography of the New-World Pupfish Genus Cyprinodon
(Teleostei: Cyprinodontidae)
A
NTHONY
A. E
CHELLE
,E
VAN
W. C
ARSON
,A
LICE
F. E
CHELLE
,R.A.V
AN
D
EN
B
USSCHE
,
T
HOMAS
E. D
OWLING
,
AND
A
XEL
M
EYER
Analysis of mtDNA sequence variation (2,548 bp from ND2, cytb, and part of the
control region) indicates that the genus Cyprinodon began diverging in the Late Mio-
cene from a common ancestor with Megupsilon, a monotypic genus on the Mesa del
Norte of Mexico. The geographic pattern of mtDNA variation, with estimates of
divergence time, suggests that by the end of the Miocene Cyprinodon occurred from
the Atlantic Coast and West Indies to near the western margin of North America via
ancestral Rio Grande and Colorado River systems. Phylogeographic structure within
the major mtDNA complexes supports a variety of hypotheses from geology and
previous phylogenetic analyses for Late Neogene connections among basins in south-
western North America now separated by formidable barriers to dispersal. Com-
parison of the mtDNA tree with previous phylogenetic inferences from allozymes
indicates that reticulate evolution involving divergent lineages probably was impor-
tant in the history of Cyprinodon.
El ana´lisis de variacio´ n de secuencias de ADNmt (ND2, cytb, y parte de la regio´n
reguladora; 2,548 pb) indica que el ge´nero Cyprinodon empezo´ a divergir en el mio-
ceno tardı´o de un antepasado comu´ n con Megupsilon,unge´nero monotı´pico de la
Mesa del Norte de Me´xico. El patro´n geogra´ fico de la variacio´n de ADNmt, con
estimadas de los tiempos divergencias, sugiere que al llegar al fin del mioceno Cypri-
nodon ocurrio´ desde la costa Atla´ntica y las Antillas hasta casi el margen oeste de
Norteame´rica por los sistemas pluviales antiguos del rı´o Bravo y del rı´o Colorado.
La estructura filogeogra´fica entre los grupos principales de ADNmt apoya a una
variedad de hipo´tesis de la geologı´a y de los ana´lisis filogene´ ticos anteriores para
conexiones del neogeno tardı´o entre cuencas del suroeste de Norteame´rica que
ahora esta´n separadas por barreras imponentes contra la dispersio´n. La compara-
cio´n del a´rbol de ADNmt con deducciones filogene´ticos anteriores de alozimas
indica que la evolucio´n reticulada que incluye lı´neas divergentes probablemente fue
importante en la historia del Cyprinodon.
F
ISH distributions and systematics have con-
tributed importantly to a large body of hy-
potheses for the paleohydrology and associated
biogeography of arid regions of North America,
most of which have emphasized relatively recent
geological/climatological events of Pleistocene
times (Hubbs and Miller, 1948; Miller, 1948; re-
views in Hocutt and Wiley, 1986). In contrast,
Minckley et al. (1986) suggested, primarily from
regional plate-tectonics, that the relevant time-
frame for understanding the biogeography of
extant western fishes might extend to Middle
Tertiary. Correspondingly, a review of molecular
phylogenetic studies of fishes in the Great Basin
region suggested that many lineages are much
older than generally appreciated (Smith et al.,
2002).
In this paper, we use mitochondrial DNA var-
iation to assess the historical biogeography of
the pupfish genus Cyprinodon, the most wide-
spread fish genus in a region comprising the
warm deserts and southern Great Plains of
southwestern North America (Miller, 1981). As
a group, cyprinodontids of this region inhabit
lowland waters where they are less likely than
many other fishes to cross basin divides and,
therefore, more likely to preserve ancient pat-
terns of geographic variation (but see Minckley
et al., 2002, for a possible exception in the Mex-
ican highlands).
Approximately 30 of the 50 or so species of
Cyprinodon occur in arid regions of southwest-
ern North America where they are primarily dis-
tributed allopatrically as endemics to relatively
small bodies of water, often single streams or
spring-systems in isolated valley floors (Miller,
1981). The remaining 20 or so species occur as
local endemics in the West Indies-Yucatan Pen-
insula region or as wide-ranging species with al-
lopatric distributions in coastal or coastal-plain
waters from Massachusetts to Venezuela (Smith
et al., 1990; Wildekamp, 1995).
321ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
Fig. 1. Distribution of samples and basins. DV
5
Death Valley System, GB
5
Guzma´n Basin, TB
5
Tu-
larosa Basin, CC
5
Cuatro Cie´negas Basin, RM
5
Rı´o
Mezquital, SPB
5
Sandia and Potosi Basins. Basins not
labeled are Devils River (7), Alamito Creek (8), Rı´o
Florido (9), and Laguna Santiaguillo (16). Localities
4 and 22 both comprise two sites in close proximity
(Material Examined).
The paleohydrology of southwestern North
America suggests many opportunities for sec-
ondary contact among previously allopatric fish
lineages (Hubbs and Miller, 1948; Minckley et
al., 1986; Smith and Miller, 1986). This and the
present, primarily allopatric occurrences of
pupfishes imply that, subsequent to most in-
stances of secondary contact, the number of lin-
eages in basins of concern was reduced to one
by hybridization and genetic introgression or by
extinctions leaving no genetic remnant.
Studies of reproductive compatibility among
extant pupfishes suggest that ancient contact
between divergent pupfishes generally would
have led to genetic introgression. Laboratory
experiments with three members of a small spe-
cies flock on the Yucatan Peninsula revealed in-
terfertility for all combinations and indicated
pre-mating reproductive isolation for only one
of the three (Strecker and Kodric-Brown, 2000).
Experiments with allopatric pupfishes show lit-
tle evidence of pre- or post-mating isolation be-
tween divergent species (Turner and Liu, 1977;
Cokendolpher, 1980; Villwock, 1982). In the
wild, hybridization is indicated in all three mor-
phologically well-studied instances of contact
between species (Minckley, 1969; Humphries
and Miller, 1981; Minckley and Minckley, 1986),
and genetic introgression exists in all known sit-
uations where endemic species were exposed to
anthropogenic introduction of a non-native
pupfish (Echelle and Echelle, 1994, 1996;
Childs et al., 1996), including one instance of
introgression over more than 400 river-kilome-
ters in less than five years (Echelle and Connor,
1989).
Hybridization confounds attempts to recover
phylogenetic relationships, but it can, at the
same time, inform reconstructions of biogeo-
graphic history because it might represent the
only evidence of past contact between lineages
(e.g., DeMarais et al., 1992). In this study, sev-
eral such instances are indicated by the present
geography of pupfishes and conflicting phylo-
genetic inferences from allozyme variation
(Echelle and Echelle, 1992, 1993a, 1998) and
the mtDNA results presented here.
M
ATERIALS AND
M
ETHODS
Sampling.—We obtained collections of 26 spe-
cies of Cyprinodon from 38 localities (Material
Examined, Fig. 1). We also used GenBank se-
quences for nine additional species not includ-
ed in our collections (see below). The complete
dataset includes 24 of the 29 historically extant
species of Cyprinodon in arid regions of south-
western North America and 11 of the approxi-
mately 20 species (some undescribed) known
from coastal Atlantic and Gulf of Mexico region
and the West Indies. One of the excluded spe-
cies (C. salvadori) is a recently described form
from the Rı´o Conchos basin in Chihuahua (Lo-
zano, 2002) that is similar to C. eximius and an-
other (C. latifasciatus) has been extinct since the
1950s (Miller, 1964). The three remaining, arid-
land species excluded from the analysis (C. ce-
ciliae,C. inmemoriam, and C. longidorsalis) repre-
sent a monophyletic group of four similar spe-
cies (Lozano and Contreras, 1993; Echelle et al.,
1995) from the Sandia Basin, Nuevo Leon, one
of which, C. veronicae, is included here. This
group was recently extirpated in the wild and is
extinct except for C. longidorsalis and C. veroni-
cae, which are being maintained in various
aquarium facilities (Contreras and Lozano,
1996; pers. comm.).
The remaining species not examined are con-
sidered close relatives of C. variegatus (Miller,
1962; Smith et al., 1990; Lozano and Contreras,
1999): C. bobmilleri from a Gulf of Mexico drain-
age in Nuevo Leon, C. hubbsi from inland Flori-
322 COPEIA, 2005, NO. 2
da, two undescribed species from Cuba (M.
Smith, pers. comm.), C. laciniatus and an unde-
scribed species flock (Holtmeier, 2001) from the
Bahamas, and C. higuey and C. nichollsi from His-
paniola. Surveys of mtDNA variation in pupfish
from the Bahamas, Hispaniola, other sites in the
West Indies, and the Atlantic and Gulf of Mexico
coasts (Bunt, 2001; L. Fuselier et al., unpubl.
data) indicate that our collections capture the
major clades of mtDNA variation in this region.
Except for C. dearborni from Venezuela, spe-
cies of Cyprinodon with large or fragmented geo-
graphic ranges were collected from multiple lo-
calities: six for the wide-ranging C. variegatus
(three each from the Atlantic and Gulf of Mex-
ico coasts), three encompassing the extremes of
the range of C. eximius, two for the coastal form
on the Yucatan Peninsula (C. artifrons), and one
for each of two isolated populations of C. rubro-
fluviatilis (Brazos and Red rivers), C. elegans
(Phantom Lake and San Solomon springs), and
C. nazas (Rı´o Nazas and Rı´o Guatimape). A
form Miller (1976) considered a third popula-
tion of C. nazas is treated as a separate species
(C. sp., Rı´o Aguanaval) on the basis of a phy-
logenetic analysis of allozymes (Echelle and
Echelle, 1998) and results presented herein. Fi-
nally, for the species from Cuatro Cie´negas, C.
atrorus and C. bifasciatus, we used three speci-
mens from the two most divergent mtDNA lin-
eages detected in a survey of this basin (E.
Carson, unpubl. data).
The phylogenetic analysis included three
monotypic genera: Jordanella, a species restrict-
ed to fresh and brackish waters in Florida, and
Cualac and Megupsilon, each of which is endemic
to a separate, springfed system on the Mexican
Plateau. Phylogenetic analyses of allozymes
(Echelle and Echelle, 1993b), mtDNA (Parker
and Kornfield, 1995), and morphology (Costa,
1997) indicate that Cyprinodon is closely related
to these genera, with allozymes and mtDNA in-
dicating a sister relationship to Megupsilon.
Data collection.—DNA was extracted from muscle
or liver tissue by the methods of either Long-
mire et al. (1997) or Hillis et al. (1990). We
used restriction fragment length polymorphism
(RFLP) to identify variant mtDNA haplotypes in
six collections of the wide-ranging C. variegatus
(n
5
12 each), each of two collections of C.
rubrofluviatilis (n
5
6 each), one of the two col-
lections of C. elegans (n
5
6), and the collections
of C. bovinus and C. pecosensis (n
5
6 each). Af-
ter digesting each sample with 14 six-base rec-
ognizing restriction enzymes as described by
Echelle and Dowling (1992), the fragments
were end-labeled with
a
32
P, electrophoresed in
0.8% agarose gels, and visualized by autoradi-
ography. This identified 19 composite haplo-
types: C. variegatus
5
9, C. elegans
5
2, C. rubro-
fluviatilis
5
4, C. pecosensis
5
2, and C. bovinus
5
2. We sequenced one sample of each com-
posite RFLP haplotype, and, for the remaining
collections, two specimens each except the fol-
lowing (n
5
1): one of the two collections of C.
artifrons, one of the three collections of C. exi-
mius, two pupfishes of the western clade (C. ma-
cularius and C. nevadensis), the two Hispaniolan
species, and the three genera, Megupsilon,Cua-
lac, and Jordanella.
A total of 65 specimens were sequenced for
the entire NADH dehydrogenase subunit 2
(ND2; 1047 bp) and cytochrome b(cytb; 1140
bp) genes and a 361-bp section of the control
region adjacent to tRNA
Thr
. The primers ND2B-
L and ND2E-H (Broughton and Gold, 2000)
were used to amplify ND2 in 50
m
l reactions
under the following conditions: 93 C for 3 min;
30 cycles of 94 C for 1 min, 50 C for 1 min, and
72 C for 2 min; 72 C for 30 min. We used those
primers to sequence the amplified fragment
with a Perkin-Elmer ABI Prism 377 automated
sequencer. Haplotypes from the same collection
locality were compared and all discrepancies
were either verified by eye in reassessments of
the computer-generated sequence histograms
or by re-sequencing. In some instances, we also
used degenerate internal primers (forward, 5
9
–
CTYAARATTGGNCTYGCYCCCCT–3
9
; back-
ward, 5
9
–RATNGAKGTGAGKGCRGGGGC–3
9
)
to obtain sequences originating around ND2
base positions 330 (for ward sequence) and 700
(reverse sequence). Cytochrome bwas amplified
with primers LA and HA (Schmidt et al., 1998)
in 50
m
l reactions with the following thermal
profile (30 cycles): 94 C for 1 min, 48 C for 1
min, and 72 C for 2 min. The same primers
were used to sequence the amplified gene with
an ABI Prism 377 automated sequencer. Hap-
lotypes from the same sample were then com-
pared and all discrepancies were verified as not-
ed for ND2. For the control region, primers
L15926 (Kocher et al., 1989) and H16498 (Mey-
er et al., 1990) were used in both amplification
and sequencing. Eleven control-region sequenc-
es (seven C. variegatus, two C. rubrofluviatilis, and
two C. pecosensis) were obtained with methods
described by Meyer et al. (1994); the remaining
sequences were obtained with the amplification
and sequencing conditions described for ND2.
To allow re-assessment of relationships in a
clade of ten species referred to as the western
pupfishes (Echelle and Echelle, 1992), includ-
ing three species not sequenced in this study (C.
diabolis,C. eremus, and C. salinus), we used 14
323ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
ND2/control-region sequences from GenBank:
12 from Duvernell and Turner’s (1998) survey
of the four species of the Death Valley System
(AF028281–91, AF028293, AF028300–9), and
two from a survey (Echelle et al., 2000) of C.
eremus (AF199001, AF199002, AF198983, and
AF198984).
Finally, to more thoroughly represent the
pupfishes of the Yucatan Peninsula, we used 14
control-region sequences (GenBank accession
numbers AY006375–6, AY006378–86, AY007978–
9, L37113) representing all haplotypes detected
in 17 specimens of C. artifrons and 5–12 fish
from each of six species in a species flock in
Laguna Chichancanab (Strecker et al., 1996;
Strecker, 2002). The monophyly of the Yucatan
pupfishes has not been assessed in an analysis
that included pupfishes from other regions.
Phylogenetic analyses.—Sequence alignment was
performed with CLUSTAL W (Thompson et al.,
1994). Indels were treated as missing data. Un-
less otherwise mentioned, all analyses were per-
formed with PAUP* (vers. 4.0b8w, D. L. Swof-
ford, PAUP*: phylogenetic analysis using parsi-
mony [*and other methods], Sinauer, Sunder-
land, MA, 2002). The complete dataset,
including the GenBank sequences (control-re-
gion for the Yucatan pupfishes and ND2/con-
trol region for the western pupfishes), was used
in a heuristic, maximum-parsimony analysis of
relationships. For this analysis, we used TBR
branch swapping, 20 random addition repli-
cates, and equal weighting of characters. To
avoid excessive computer time resulting primar-
ily from low resolution for the Yucatan sequenc-
es, maximum number of trees for branch-swap-
ping was held to 100 and branches with mini-
mum lengths of zero were collapsed. Using the
same restrictions, nonparametric bootstrapping
(Felsenstein, 1985) with 500 replicates (ten ran-
dom addition sequences per replicate) was used
to assess support for clades. To compare the
fast-parsimony method used here with a more
thorough search, we subjected a dataset com-
prising only the ND2/cytb/control region se-
quences we detected to the same analyses ex-
cept that branches of length zero were not col-
lapsed and there was no limit on number of
trees held for branch swapping.
For a second approach to phylogenetic analy-
sis, we subjected the composite ND2, cytb, and
control-region haplotypes detected in this study
to the Bayesian method provided by MrBayes 2.0
(Hulsenbeck and Ronquist, 2001). ModelTest
3.04 (Posada and Crandall, 1998) indicated that
the base-substitution model best fitting the data
was the general-time-reversible model with some
sites assumed to be invariable and with variable
sites following a discrete gamma distribution
(GTR
1
I
1G
). With this model, we first per-
formed an analysis in which the nucleotide sub-
stitution parameters were treated as unknown
variables with uniform prior values and estimated
as part of the analysis. The analysis began with
random starting trees and was run for 1
3
10
6
generations, with the initial 1,000 cycles discard-
ed as burn-in. In a separate Bayesian analysis, we
partitioned the characters into seven classes
(control region and base positions 1, 2, and 3 for
both ND2 and cytb) and used a site-specific gam-
ma model (GTR
1
SS
G
). The analysis was run
for 1
3
10
6
generations, with the initial 3,600
cycles discarded as burn-in. To assure stationarity,
we graphically monitored the fluctuating value of
the likelihood in both Bayesian analyses and ran
the GTR
1
I
1G
analysis three times. Bayesian
analysis produces, for all tree nodes, true poste-
rior probabilities under the assumed substitution
model (Rannala and Yang, 1996). We considered
clades significantly supported when probabilities
were
.
0.95.
Estimating divergence times.—To assess the robust-
ness of our estimates of divergence time, we
used two different approaches, one based on a
uniform ND2 clock developed independently of
our data, and another that is based on our ND2
data and allows rate variation in different parts
of the tree. In the first approach, we used a mo-
lecular clock for the genus Aphanius, which is
sometimes mentioned as the Old World equiv-
alent of Cyprinodon (e.g., Villwock, 1982). Both
groups include extremely hardy, arid-land or
coastal cyprinodontids existing in a highly frag-
mented aquatic landscape. Two reasonably well-
understood vicariant events are postulated for
Aphanius: (1) the development, 5.5 million
years ago (Mya), of the Strait of Gibraltar, which
apparently initiated divergence of A. iberus and
A. baeticus (Perdices et al., 2001; Doadrio et al.,
2002) and (2) the transgression, 13 Mya, of the
Red Sea into the Wadi Sirhan of Jordan, result-
ing in divergence producing the sister species
A. sirhani and A. dispar (Hrbek and Meyer,
2003).
To calibrate an ND2 clock for Aphanius,we
first produced a minimum-evolution tree for
the 49 sequences detected in 16 species of the
genus (Hrbek and Meyer, 2003; GenBank ac-
cession numbers AF449287–335), with Valencia
letourneuxi as the outgroup (accession numbers
AF449336–7). The tree was based on gamma-
corrected, Tamura-Nei distances (maximum-
likelihood estimate of
a5
0.37). To adjust for
inconsistent substitution rates detected in rela-
324 COPEIA, 2005, NO. 2
Fig. 2. Strict consensus of maximum-parsimony
trees. Terminal nodes with locality numbers in paren-
theses represent haplotypes generated in the present
study. Taxa in boxes represent GenBank sequences.
Weakly divergent (uncorrected p
5
0.001–0.009) ter-
minal nodes are collapsed to one except C. nevadensis,
for which all are retained to show paraphyly with re-
spect to C. diabolis, and the Lake Chichancanab/
C. artifrons node, a polytomy of shared haplotypes.
Numbers above branches
5
bootstrap support. Aster-
isks
5
Bayesian probabilities
.
95% (model
5
GTR
1
SS
G
); nodes uptree of arrows include GenBank se-
quences not subjected to Bayesian analysis. The node
with a dashed line received Bayesian and bootstrap
support but was not retained in the strict consensus
of the shortest maximum-parsimony trees.
tive-rate tests (Tajima, 1993) with MEGA2 2.1
(Kumar et al., 2001), we imported the tree into
TreeEdit (vers. 1.0a8; A. Rambaut and M.
Charleston; Dept. of Zoology, Univ. of Oxford,
Oxford, U.K.) and used nonparametric-rate-
smoothing (NPRS; Sanderson, 1997) to adjust
for unequal rates of substitution on different
branches. NPRS smoothes the rapidity of rate
change among neighboring branches with a
function that penalizes rates changing too rap-
idly from branch to branch. The result is an
ultrametric tree with terminal nodes equidistant
from the root of the tree. The ND2 distances
for A. iberus vs. A. baeticus (17.1%) and A. sirhani
vs. A. dispar (32.2%) were read from the result-
ing tree, giving calibrations for pairwise diver-
gence of, respectively, 3.1% per Myr and 2.5%
per Myr (
5
1.6% and 1.2% for individual line-
ages). A similar procedure then was used to ob-
tain interclade divergences from our ND2 da-
taset, excluding all haplotypes detected by Du-
vernell and Turner (1998) for C. nevadensis ex-
cept two (H and I) that, with our sample,
represent the extremes for the species. Inter-
clade distances were taken directly from the
NPRS tree derived from the minimum-evolu-
tion tree for gamma-corrected Tamura-Nei dis-
tances (maximum likelihood estimate of
a5
0.29).
The second method of estimating divergence
times employed the multidivtime software (http:
//statgen.ncsu.edu/thorne/multidivtime.html;
Kishino et al., 2001; Weigmann et al., 2003).
This procedure allows substitution-rate variation
among branches, with the assumption that the
rate is more similar among nearby branches
than among more distant branches. Sequences
and an assumed topology are first input into
PAML 3.0c (Yang, 1997) to compute maximum
likelihood estimates of transition/transversion
ratios, rate heterogeneity among sites, and nu-
cleotide frequencies. Multidivtime uses this out-
put, together with the original topology and as-
sociated sequences, to derive Bayesian estimates
of divergence time and their 95% confidence
intervals. The method is most effective when
the timing of one or more interior nodes in the
topology is constrained based on the fossil re-
cord or, as in our analysis, assumed times of vi-
cariance (Weigmann et al., 2003).
For the multidivtime/PAML analysis, we used
the topology in Figure 2 and, to increase com-
putational tractability, we reduced the ND2 da-
taset used with the Aphanius clock to 38 se-
quences. This dataset included one ND2 hap-
lotype per species except for 2 sequences from
C. rubrofluviatilis (one each from the highly di-
vergent Red and Brazos river populations) and,
to preserve examples of mtDNA paraphyly,
three sequences from C. v. variegatus, four from
C. eximius, and two from C. nevadensis. To deter-
mine if the ND2 tree is consistent with the total-
information topology in Figure 2, we performed
the more thorough maximum-parsimony anal-
yses described above on the subset of 38 se-
quences used in estimates of time. The results
were completely consistent with Figure 2 except
that fewer nodes received bootstrap support.
The multidivitime procedure requires some
biologically plausible prior estimates for certain
parameters, but they can be vague (hence they
are given large error estimates) because the
Markov chain Monte Carlo procedure (MCMC)
employed in the analysis is relatively robust to
325ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
their variation (Weigmann et al., 2003). We
used 0.10 (SD
5
0.10) as the prior estimate of
the mean for rate of substitution per time unit
(1 unit
5
10 million years) at the ingroup root
and, following rationale of Weigmann et al.
(2003), 1.0 as a log-scale estimate of the vari-
ance in rate of change per unit time. As the
most plausible estimate of time from the pres-
ent to the root of the ingroup (Cualac,Megup-
silon, and Cyprinodon), we used 11 Mya (1.1 time
units; SD
5
1.1) based on results from the mo-
lecular clock developed with Aphanius. One
might view this as a violation of the intent to
use two independent approaches. However, our
experiments with values of 5 Mya, 11 Mya, and
20 Mya, indicated that the analysis was relatively
insensitive to this prior. Given the internal time
constraints described below, the posterior solu-
tions were, respectively, 11.7 Mya, 13.4 Mya, and
14.6 Mya, and the results elsewhere in the tree
had no effect on biogeographic conclusions.
For the remaining priors, we followed Weig-
mann et al. (2003) and used the default values
in multidivtime. The analysis included a burn-
in period of 100,000 cycles of the MCMC pro-
cedure followed by sampling of parameters ev-
ery 100 cycles until 10,000 samples were col-
lected. This design appeared adequate because
means of divergence time and confidence in-
tervals from different runs were similar to the
third significant figure.
For internal time constraints we restricted two
nodes as follows: (1) the lower and upper limits
for time of divergence between C. macularius
and C. eremus were set at 1.1 Mya and 1.7 Mya,
representing the timing of the Pinacate volcanic
eruptions (Lynch, 1981) assumed to have initi-
ated divergence of those two species (Turner,
1983; Echelle et al., 2000), and (2) the upper
limit for divergence of C. rubrofluviatilis from
species in the Pecos River was set at 4.5 Mya.
The latter represents the 4–5 Mya cessation of
Ogallala deposition in the southern Great
Plains (Chapin and Cather, 1994). As described
later, this led to formation of the upper Pecos
River, which, in turn, helps explain the zooge-
ography of pupfish, other fishes (Echelle and
Echelle, 1978; Smith and Miller, 1986), and
snails (Hershler et al., 2002).
Reticulate evolution.—We searched for evidence of
potential reticulate evolution by comparing phy-
logenetic inferences from mtDNA with those
from allozymes. Conclusions are tentative be-
cause of low resolution in the allozyme results.
Pupfishes exhibit high levels of allele sharing for
allozymes, with few fixed differences among
groups (Turner, 1974; Echelle and Echelle, 1992,
1998). Consequently, the phylogenetic trees are
based on allele-frequency parsimony (Berlocher
and Swofford, 1997), and clades often are sup-
ported by frequencies of alleles shared with oth-
er clades. To reduce errors from poor resolution,
we consider only those conflicts supported by at
least one synapomorphic allozyme allele not
shared with other nodes.
R
ESULTS
We detected 58 mtDNA haplotypes among the
141 specimens surveyed (by RFLP or directly by
sequencing). All sequences are deposited in
GenBank (accession numbers AY901992–
AY902165). The control-region alignments had
eight indels; none occurred in ND2 or cytb. Se-
quences were identical in six instances where two
specimens were sequenced from the same collec-
tion: C. dearborni,C. artifrons from Cancun, C. tu-
larosa,C. macrolepis, Aguanaval pupfish (C. sp.),
and C. albivelis. With one exception (C. eximius
from Rı´o Florido; uncorrected p
5
0.057), level
of divergence among haplotypes from the same
collection was low, ranging from 0.001 to 0.013.
Maximum uncorrected p-values were 0.176 and
0.223 for, respectively, cytband ND2 (for both
genes, Jordanella vs. a species of Cyprinodon) and
0.194 for the control region (Cualac vs. three
samples of Cyprinodon). There were 777 parsi-
mony informative characters representing 26%
(96/361), 28% (322/1140), and 34% (359/
1047), respectively, of the control region, cytb,
and ND2 bases.
Phylogenetic analyses.—The faster approach to
maximum-parsimony analysis of the complete
dataset of 58 composite ND2/cytb/control-re-
gion haplotypes and 28 partial haplotypes ob-
tained from GenBank produced a topology
(Fig. 2) that, for the samples in common, was
identical to the topology obtained using a more
thorough parsimony approach with only the 58
haplotypes we detected, and differences in lev-
els of bootstrap support were negligible. Thus,
only results from the complete dataset are de-
scribed here. This analysis produced 446 trees
(length
5
2,715; CI
5
0.42; RI
5
0.80) differing
only in topology within terminal clades, primar-
ily the one consisting of 15 haplotypes from the
Yucatan Peninsula. All nodes in the strict con-
sensus tree received greater than 50% bootstrap
support except for two deep nodes within Cypi-
nodon (Fig. 2).
The results for the western pupfish group
(clade 5, Fig. 2) were, except for C. radiosus,
consistent with previous mtDNA-based analyses
(Echelle and Dowling, 1992; Duvernell and
326 COPEIA, 2005, NO. 2
Turner, 1998). Those studies supported, albeit
weakly, a sister relationship between C. radiosus
and the C. macularius-C. eremus clade, whereas
we found weak bootstrap support (56%) for C.
radiosus as the basal member of the western
pupfishes, a result consistent with lack of Bayes-
ian support for relationships among the four
basal nodes of the clade (Fig. 2).
The maximum parsimony bootstrap analysis
supported (98%) mtDNA monophyly for the
Yucatan Peninsula pupfishes, C. artifrons, and
the Lake Chichancanab species flock (Fig. 2).
The 14 haplotypes previously detected in this
region and an additional haplotype we found in
C. artifrons were weakly divergent (uncorrected
p
5
0.000–0.014), and relationships were un-
resolved excepting 65% support for a cluster of
two of the nine haplotypes known only in C.
artifrons.
The two models of substitution (GTR
1
I
1
G
and GTR
1
SS
G
) used in the Bayesian anal-
ysis of the 59 ND2/cytb/control-region haplo-
types resulted in identical topologies, and the
nodes with probabilities
.
0.95 were the same.
With one exception, the Bayesian and maxi-
mum parsimony cladograms were identical
when collapsed to nodes having, respectively,
.
0.95 Bayesian probabilities or
.
50% maximum
parsimony bootstrap support (Fig. 2). The ex-
ception is the maximum-parsimony placement
(67% bootstrap support) of clade 1 (Yucatan
pupfishes) as the basal lineage of Cyprinodon,an
inference not receiving Bayesian support.
None of the nodes connecting the five pri-
mary mtDNA clades in Cyprinodon received
Bayesian or maximum parsimony support ex-
cept the just mentioned conflict for clade 1.
Correspondingly, the optimal trees from the
two approaches showed no areas of congruence
for relationships among those clades. Thus, re-
lationships among the five basal clades within
Cyprinodon are unresolved.
Bayesian analysis and maximum-parsimony
resulted in nearly identical, well-supported to-
pologies within each of the seven major mtDNA
clades of Cyprinodon (Fig. 2), all of which, ex-
cept for a minor difference within clade 2A,
were consistent with the minimum-evolution
tree (Fig. 3) to which the molecular clock from
Aphanius was applied. The one disparity in-
volved the Red and Brazos river forms of C. ru-
brofluviatilis. The strict-consensus, maximum-
parsimony tree (Fig. 2) supported monophyly
of this species, but with no bootstrap support,
whereas the Bayesian tree indicated paraphyly
with respect to the C. pecosensis-C. bovinus-C. ele-
gans clade, also with no statistical support. Thus,
we consider the question of monophyly for C.
rubrofluviatilis to be unresolved.
Hereafter in this paper, we refer to the seven
major clades with names indicating region of
primary occurrence (Figs. 3 and 4A): clade 1
5
Yucatan (Laguna Chichancanab and coastal ar-
eas on the Yucatan Peninsula), 2A
5
maritime
(coastal Atlantic, Gulf of Mexico, and Hispan-
iola; Tularosa Basin of the northern Chihua-
huan Desert is the only occurrence in the in-
terior of North America), 2B
5
southern Great
Plains-northern Chihuahuan Desert (Red, Braz-
os, and Pecos river basins), 3
5
Sandia-Potosı´
(Sandia and Potosı´ basins, Nuevo Leon, Mexi-
co), 4A
5
Rı´o Conchos-middle Rio Grande (Rı´o
Conchos, Alamito Creek, and Devils River; all in
the middle Rio Grande Basin), 4B
5
Old Rı´o
Nazas (rı´os Nazas, Aguanaval and Mezquital, La-
guna Santiaguillo, and Cuatro Cie´negas Basin),
and 5
5
western (Guzma´n Basin, lower Colo-
rado River, and Death Valley System).
Divergence times.—The two methods of estimat-
ing divergence times gave similar results
throughout the tree (Table 1). The largest ab-
solute differences were at the base of the tree,
with differences of 2.5 Myr and 1.7 Myr in esti-
mated times of divergence between Cyprinodon
and, respectively, Cualac and Megupsilon. The
differences were effectively negligible within
Cyprinodon. The major discrepancy was the time
(1.9 Myr vs. 2.6 Myr) separating the Aguanaval
pupfish (C. sp.) from the Cuatro Cienegas pup-
fishes (C. atrorus and C. bifasciatus), and none
of the differences resulted in conflicting bio-
geographic hypotheses.
Mitochondrial DNA/allozyme conflicts.—The syn-
thesis of the various phylogenetic studies of
mtDNA and allozyme variation (Fig. 5) indi-
cates six instances of potential conflict, five sup-
ported by synapomorphic allozyme alleles. The
sixth conflict involves a polymorphism in the
Rı´o Florido population of C. eximius for two
markedly divergent mtDNAs (uncorrected di-
vergence
5
5.7%), one being only 1.3% diver-
gent from the mtDNA detected in C. macrolepis.
Thorough geographic surveys of mtDNA varia-
tion in C. bifasciatus (E. Carson, unpubl. data)
and C. tularosa (Stockwell et al., 1998) indicate
fixation for the mtDNA lineages involved in the
mtDNA/allozyme conflicts for those species.
Similar fixation in the other instances of con-
flict seems likely despite small sample sizes: C.
elegans was sampled from two localities in its
small range (n
5
2 and 6); C. pachycephalus (n
5
2) is restricted to an extremely small spring
system; and C. radiosus (n
5
2), which has un-
327ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
Fig. 3. Estimated timing of cladogenetic events. Ages indicated are as follows (20.3 Mya in Early Miocene)
and the start of each of the following (from Berggren et al., 1995): Late Miocene (11.2), Pliocene (5.2), and
Pleistocene (2.5 Mya). Based on the ultrametric, minimum-evolution tree, with timing based on the clock
(2.8%/Mya) developed for Aphanius. Timing closely resembles that from the multidivtime approach (Table
1), but that approach does not produce an ultrametric tree in which all terminal nodes are at time zero.
dergone severe population bottlenecking be-
cause of human activity (Miller and Pister,
1971), has appeared almost devoid of allozyme
variation (Turner, 1974; Echelle and Echelle,
1993a).
D
ISCUSSION
Estimating divergence time from molecular
divergence is fraught with difficulties (Doyle
and Donoghue, 1993; Bermingham et al.,
1997), and our estimates have wide confidence
intervals (Table 1). Nonetheless, the close
agreement reported here for divergence times
derived by two extremely divergent approaches
indicates that the means for the estimates are
sufficiently robust to provide reasonable hy-
potheses of event times. We also note that the
estimates for the western pupfish group (clade
5) are similar to those obtained from still an-
other approach, where sequence divergences
estimated from restriction-site data for whole
mtDNA (Echelle and Dowling, 1992) were used
with a molecular clock based primarily on the
goodeid fossil record and Webb’s (1998)
mtDNA sequence data for goodeids (Smith et
al., 2002).
Our results support the monotypic genus Me-
gupsilon as the sister-group of a monophyletic
Cyprinodon, in agreement with results from al-
lozymes for a diversity of pupfishes (Echelle and
Echelle, 1993b), as well as those from analyses
328 COPEIA, 2005, NO. 2
Fig. 4. Geography and vicariance history of
mtDNA clades in Cyprinodon: (A) Geographic distri-
bution of the seven primary mtDNA-clades. 1
5
Yu-
catan, 2A
5
maritime, 2B
5
southern Great Plains-
northern Chihuahuan Desert, 3
5
Sandia-Potosi, 4A
5
Rı´o Conchos-middle Rio Grande, 4B
5
Old Rı´o
Nazas, 5
5
western. Clade 2A is the maritime group,
with a basal C. variegatus ovinus-C. bondi clade, and a
clade with the following topology: (C. tularosa [C. dear-
borni {C. variegatus,C. sp. from Lake Enriquillo}]). (B)
Some hypothetical paleosystems (arrows) and esti-
mated times of vicariance for pupfish in each system.
Basin abbreviations as in Fig. 1. Lettering in boxes
5
age of associated vicariance (average of the estimates
in Table 1) and clades involved (labeled as in Fig. 2).
of mtDNA (Parker and Kornfield, 1995) and
morphology (Costa, 1997) that included, re-
spectively, two and three species of Cyprinodon.
Estimated divergence time suggests that the
genera began diverging in Late Miocene about
7–9 Mya. Branch lengths and unresolved rela-
tionships among the five most basal mtDNA
clades of Cyprinodon indicate origins rather
closely together in time
;
6–7 Mya (Table 1; Fig.
3). This is consistent with the estimated age of
the only fossil assigned to the genus Cyprinodon,
a specimen from Death Valley of Late Miocene
or Early Pliocene age (Miller, 1945, 1981).
The mtDNA results provide little support for
the suggestion (Echelle and Echelle, 1992,
1998) that various inland lineages might have
arisen independently from coastal forms. This
hypothesis predicts that inland clades on the
mainland will be interleaved among coastal
clades (Lovejoy and Collette, 2001), and, with
one exception, there is no evidence for such a
pattern. The exception is C. tularosa of the Tu-
larosa Basin in southcentral New Mexico, whose
mtDNA arose rather recently (
;
1.6–1.9 Mya)
within the maritime clade. One or more of the
three major inland clades might have a similar
maritime origin but unresolved relationships
among these clades preclude testing this hy-
pothesis.
Atlantic and Gulf Coasts and West Indies.—The
pupfishes from this region represent two deeply
divergent mtDNA clades: the Yucatan clade
from the Yucatan Peninsula (clade 1, Fig. 2) and
the wide-ranging maritime clade (2A), which,
with one exception, C. tularosa from southcen-
tral New Mexico, occurs in coastal mainland
pupfishes of North and South America and is-
lands of the West Indies. The monophyly of the
Yucatan clade is consistent with the suggestion
from genetic similarities that the coastal species,
C. artifrons, was ancestral to the Laguna Chi-
chancanab species flock, a complex of six eco-
logically diverse taxa that, based on past lake
levels (Covich and Stuiver, 1974) and low levels
of genetic divergence, might be only 8,000 years
old (Humphries, 1984; Strecker et al., 1996).
Because of limited sampling, we cannot ex-
clude the possibility that the Yucatan mtDNA
clade occurs outside of the Yucatan Peninsula
or that C. artifrons, a wide-ranging pupfish on
the peninsula and nearby islands, does not car-
ry representatives of other mtDNA clades de-
tected in this study. However, surveys of pupfish-
es in the Bahamas (Bunt, 2001) and on Hispan-
iola, Grand Cayman, and the Florida Keys (L.
Fuselier et al., unpubl. data) so far reveal hap-
lotypes belonging only to the maritime clade.
Guzma´ n Basin, Lower Colorado River, and Death
Valley System.—ND2 divergence among basal lin-
eages of the western pupfish clade indicates
that, by
;
3.6–3.8 Mya, pupfishes occurred in
the regions now occupied by the Death Valley
System (Owens Valley and the Amargosa River-
Death Valley region), lower Colorado River, and
Guzma´n Basin of northwestern Chihuahua (Fig.
4B).
329ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
Fig. 5. A synthesis of phylogenetic studies of mtDNA and allozymes in pupfishes. Nodes for mtDNA had
.
50% bootstrap support and, in the present study,
.
95% Bayesian probability. For allozymes, asterisks denote
nodes supported by at least one synapomorphic allele; remaining allozyme nodes were in both the shortest
allozyme trees and the mtDNA tree. Thick vertical lines denote conflicts between mtDNA and the best-sup-
ported allozyme inferences. Labeled boxes denote major mtDNA clades (Figs. 3 and 4). Letters on trees
indicate sources other than this paper: A
5
Parker and Kornfield (1995) for placement of Cualac,Cyprinodon,
and Megupsilon with Jordanella;B
5
Echelle et al. (1995) for monophyly of a group of species from Sandia and
Potosı´ basins (C. veronicae and C. alvarezi here); C, D, E, and F
5
Echelle and Echelle (1992, 1993a, 1993b,
and 1998) for, respectively, monophyly of the C. variegatus complex and the Death Valley System pupfishes and
relationships among genera and among Mexican Plateau species and representatives of other lineages.
There are two competing hypotheses for the
lower Colorado River-Death Valley connection:
(1) a Miocene-Early Pliocene route via an Amar-
gosa-Colorado paleoriver that preceded devel-
opment of the Death Valley trough (Howard,
1996), or (2) a Plio-Pleistocene route via con-
nections among successively lower basins be-
tween the Death Valley System and the lower
Colorado River (Miller, 1948, 1981; Smith et al.,
2002). The older Amargosa-Colorado paleoriver
seems more compatible with the mtDNA esti-
mate of divergence time. However, as discussed
below, the possibility of a subsequent invasion
of the Death Valley System would help explain
other genetic aspects of the western pupfish
clade.
Branch lengths in the mtDNA tree (Fig. 3)
suggest a Pliocene connection between Guzma´n
Basin and the lower Colorado River at about the
time of the Colorado River-Death Valley con-
nection. At that time Guzma´n Basin was occu-
pied by Lake Cabeza de Vaca, an immense Mio-
cene to Late Pleistocene complex of basins ex-
tending from northwestern Chihuahua into
southern New Mexico (Strain, 1966; Seager et
al., 1984). Topography and geological data, in-
cluding a Miocene valley-basalt flow, indicate
potential connections between this system and
the Gila River (Miller, 1981), which flows west-
erly to the lower Colorado River. The Gila River
was integrated with the lower Colorado River by
5 Mya and probably earlier (Howard, 1996).
The period from 5 Mya through the Pleistocene
saw accelerated tectonic activity in the associat-
ed Rio Grande Rift (Seager et al., 1984), and
such activity might have severed westward con-
nections between the Lake Cabeza de Vaca com-
plex and the lower Colorado River.
A problematic portion of the western-pupfish
mtDNA clade is the indicated sister relationship
330 COPEIA, 2005, NO. 2
T
ABLE
1. C
OMPARISON OF MT
DNA D
IVERGENCE
T
IMES
B
ASED ON
R
ESULTS
O
BTAINED FROM THE
M
ULTIDIVTIME
A
NALYSIS AND THE
M
OLECULAR
C
LOCK FOR
Aphanius. Except where otherwise noted, the label for each node
consists of two taxa, one for each of the two descendant clades, some with locality numbers. Values given are
the mean and 95% confidence interval from multidivtime and, for the Aphanius clock, the estimates based on
the two calibrations for gamma-corrected Tamura-Nei divergence (2.4% and 3.1% per Mya) and the average
of the two (2.8%).
Node
Divergence time (Mya)
Multidivtime Aphanius clock
Jordanella/Cyprinodon — 20.3
(18.4–23.8)
Cualac/Cyprinodon 13.4
(7.3–22.6)
10.9
(9.8–12.7)
Megupsilon/Cyprinodon 8.5
(5.3–12.6)
6.8
(6.1–7.9)
Among clades 1–5 of Cyprinodon 5.5–7.2
(6.3; 3.5–10.2)
a
5.5–5.8
(4.7–7.9)
Clade 2
Clade 2A/2B 4.6
(3.0–6.0)
4.9
(4.5–5.8)
C. bondi vs C. ovinus 2.5
(1.3–3.9)
2.0
(1.8–2.4)
C. bondi/C. variegatus 2.9
(1.6–4.4)
2.4
(2.2–2.8)
C. tularosa/C. variegatus 1.9
(1.0–3.2)
1.6
(1.4–1.8)
C. rubrofluviatilis/C. pecosensis 3.7
b
(2.4–4.5)
3.6
(3.3–4.2)
C. rubrofluviatilis (1 & 2) 3.5
(2.2–4.3)
3.4
(3.0–3.9)
C. pecosensis/C. elegans 2.6
(1.5–3.8)
2.3
(2.1–2.7)
C. pecosensis/C. bovinus 0.6
(0.2–1.2)
0.5
(0.5–0.6)
Clade 3
C. veronicae/C. alvarezi 2.1
(1.0–3.6)
2.0
(1.8–2.3)
Clade 4
Clade 4A/4B 4.6
(2.8–6.7)
4.9
(4.5–5.8)
C. eximius (7)/C. eximius (8) 0.1
(0.05–0.4)
0.4
(0.4–0.5)
C. eximius (7)/C. eximius (9) 0.4
(0.1–1.0)
0.4
(0.4–0.5)
C. eximius (7)/C. macrolepis 2.9
(1.6–4.9)
3.4
(3.0–3.8)
C. macrolepis/C. pachycephalus 3.0
(1.6–4.9)
3.4
(3.0–3.7)
C. meeki/C. nazas 3.0
(1.6–4.7)
3.4
(3.1–4.0)
C. meeki/C. atrorus 4.1
(2.4–6.0)
4.8
(4.3–5.5)
C. atrorus/C. sp. (Aguanaval) 1.9
(0.9–3.3)
2.6
(2.4–3.1)
Clade 5
C. radiosus/C. macularius 3.8
(2.5–5.6)
3.6
(3.3–4.3)
C. macularius/C. eremus 1.3
c
(1.1–1.6)
1.6
(1.5–1.9)
C. macularius/C. diabolis 3.5
(2.3–5.0)
3.9
(3.5–4.5)
331ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
T
ABLE
1. Continued.
Node
Divergence time (Mya)
Multidivtime Aphanius clock
C. fontinalis/C. diabolis 2.7
(1.5–4.2)
3.1
(2.8–3.7)
C. nevadensis/C. salinus 1.0
(0.4–1.9)
1.0
(0.9–1.2)
C. nevadensis/C. diabolis 0.5–0.7
d
(0.1–1.4)
0.5–0.6
(0.5–0.8)
C. nevadensis/2–3 haplotypes
e
0.7
(0.2–1.4)
0.7
(0.6–0.8)
a
In parentheses, average divergence time among the five clades and the range from the lower 95% confidence limit for the smallest interclade
value (5.5) to the higher limit for the largest (7.2).
b
Timing constrained to have an upper limit of 4.5 Mya.
c
Timing constrained to be between 1.1 Mya and 1.7 Mya.
d
Times shown are between C. diabolis and the two haplotypes of C. nevadensis included in the analysis.
e
See Materials and Methods for haplotypes included.
between C. fontinalis of Guzma´n Basin and the
Amargosa River-Death Valley pupfishes (C. ne-
vadensis,C. diabolis, and C. salinus). This is enig-
matic because the suggested sister groups are
separated by the lower Colorado River region,
which is inhabited by C. macularius. Allozymes
support, albeit weakly, the geographically more
plausible hypothesis that the Amargosa River-
Death Valley pupfishes are part of a monophy-
letic complex in the Death Valley System that
includes C. radiosus in nearby Owens Valley. This
conflict is discussed in a later section.
Ancestral Rio Grande and Mesa del Norte.—The
three Guzma´n Basin members of the western
pupfish clade (C. albivelis,C. fontinalis, and C.
pisteri) and most other southwestern pupfishes
occur in remnant streams and endorheic basins
of Smith and Miller’s (1986) ancestral Rio
Grande System. In the Miocene and Pliocene,
the upper Rio Grande occupied a developing
north-south succession of fault basins in the Rio
Grande Rift from Colorado to southwestern
New Mexico. By 3.4–4.5 Mya, basins of south-
central New Mexico, including parts of the Lake
Cabeza de Vaca complex, probably formed the
terminus of an ancestral upper Rio Grande
(Seager et al., 1984; Repenning and May, 1986;
Mack et al., 1993). By
;
2 Mya, the river was
emptying into the Tularosa-Hueco complex of
basins at Fillmore Pass 40 km north of the pres-
ent city of El Paso (Mack et al., 1996). From
there, it probably flowed to the Gulf of Mexico
in roughly the present course (Hawley et al.,
1976).
The Tularosa Basin-lower Rio Grande con-
nection
;
2 Mya is supported by the mtDNA es-
timate for the age of C. tularosa (
;
1.6–1.9 Myr)
and that of a springsnail Juturnia tularosae (2–3
Myr; Hershler et al., 2002). The closest mtDNA
relatives of these endemics to the Tularosa Ba-
sin are, respectively, the coastally distributed C.
v. variegatus-C. dearborni clade and J. kosteri in the
Pecos River. Early Pleistocene uplifting and
westward tilting of the San Andres-Franklin
mountains diverted flow away from the Tularo-
sa-Hueco Basin and to the west of El Paso into
Mesilla Basin of the Lake Cabeza de Vaca com-
plex (Strain, 1966; Hawley et al., 1976). This
contributed to the isolation of Tularosa Basin
and temporarily severed the connection be-
tween the upper and lower Rio Grande.
Approximately 0.8 Mya (Mack et al., 1993),
the upper and lower Rio Grande basins were re-
integrated via a more southern connection
when the upper basin was diverted from the
Lake Cabeza de Vaca complex by basin overfill
and canyon cutting near El Paso (Hawley et al.,
1976; Seager et al., 1984). The history of con-
nections between the lower Rio Grande and its
major tributaries is not well understood, but a
minimum age for the lower Pecos River is
;
13
Myr based on dating of Gatun˜a formation sink-
hole-fill, near Orla, Reeves County, Texas, and
the connection between the Rı´o Conchos and
the Rio Grande might be as recent as early Pli-
ocene ( J. W. Hawley, pers. comm.). The Rio
Conchos crosses ancient mountain ranges sep-
arated by valley floors partially filled with thick
alluvium of Late Tertiary or Quaternary age, in-
dicating that drainage was not continually
through-flowing to the Rio Grande (Smith and
Miller, 1986).
The pupfishes on the Mesa del Norte of Mex-
ico east and south of Guzma´ n Basin carr y a well-
supported mtDNA complex (clade 4A–B) that,
early in the evolution of Cyprinodon (
;
5 Mya),
split into two major clades, one now in the Rı´o
Conchos-Middle Rio Grande and one support-
ing recognition of the Old Rı´o Nazas, a pa-
332 COPEIA, 2005, NO. 2
leosystem originally inferred from fish and her-
petofaunal distributions (Meek, 1904; Conant,
1963). The Old Rı´o Nazas would have extended
from the western highlands of Mexico, across
what is now the Chihuahuan Desert, to the Rio
Grande. The progenitor of the Old Rı´o Nazas
clade might have evolved in isolation on the
Mesa del Norte, starting in early Pliocene time.
The mtDNA tree corroborates the indication
from allozymes (Echelle and Echelle, 1998) of
an Old Rı´o Nazas clade (clade 4B) comprising
the pupfishes of Durango and Zacatecas (rı´os
Nazas, Aguanaval, and Mezquital, and Laguna
Santiaguillo) and one (allozymes) or both
(mtDNA) pupfishes in the Cuatro Cie´negas Ba-
sin. This argues for one of three hypotheses
(Arellano, 1951; Conant, 1963) for the connec-
tion of the Old Rı´o Nazas with the Rio Grande:
namely via Rı´o Salado (Fig. 4B), the Pleistocene
outlet for Cuatro Cie´negas Basin. The ND2 di-
vergence between the Rı´o Aguanaval pupfish
(C. sp.) and the Cuatro Cie´negas pupfishes in-
dicates severance of a connection across the
Chihuahuan Desert
;
1.9–2.6 Mya.
The only fishes known from the small, en-
dorheic Sandia and Potosı´ basins are the five
species of the C. alvarezi-C. veronicae clade and
the monotypic genus Megupsilon, suggesting that
these are relics of an ancient fauna isolated in
this region of the Mesa del Norte. Resolution of
C. alvarezi-C. veronicae as one of the basal
mtDNA clades in Cyprinodon and inclusion of
the Sandia and Potosı´ basins in Smith and
Miller’s (1986) Rio Grande System indicates
early isolation from the ancestral Rio Grande
(Fig. 3).
Southern Great Plains-Northern Chihuahuan Des-
ert.—Divergence among pupfishes carr ying the
southern Great Plains-northern Chihuahuan
Desert mtDNA clade might have involved frag-
mentation of a system (Gustavson and Finley,
1985; Conner and Suttkus, 1986) extending
eastward from the southern Rocky Mountains to
the Gulf of Mexico. This ancestral southern
Great Plains system (Fig. 4B) would have com-
prised a complex of low-gradient streams occa-
sionally interconnected by cross-grading as they
meandered over the relatively flat surface of the
Miocene-to-Pliocene Ogallala formation (Con-
ner and Suttkus, 1986). Vicariance and mtDNA
lineage sorting from a polymorphic ancestor in
this system might explain the short internodal
branches (Fig. 3) among the three markedly di-
vergent, basal clades of this complex, two rep-
resenting the Red and Brazos river forms of C.
rubrofluviatilis. The lack of support for a mono-
phyletic C. rubrofluviatilis (see Results) also oc-
curred in a phylogenetic analysis of allozymes
(Echelle and Echelle, 1992).
Range fragmentation leading to the present
geography of the southern Great Plains-north-
ern Chihuahuan Desert mtDNA clade might
have been triggered by cessation of Ogallala de-
position, an event that occurred 4–5 Mya (Cha-
pin and Cather, 1994) and was associated with
entrenchment of the Red, Brazos, and upper
Pecos rivers (Gustavson and Finley, 1985). For
the Pecos River, dissolution of Permian salt beds
and resultant landscape subsidence led to pond-
ing of portions of eastwardly flowing streams,
including the Brazos River, in a trough forming
the proto-Pecos River Valley in eastern New
Mexico. Subsequent integration of the upper
and lower Pecos occurred through headward
subsidence and erosion of the much older lower
Pecos River. Estimated divergence time (
;
3.6
Mya) for basal members of the southern Great
Plains-northern Chihuahuan Desert clade is
consistent with this model of river evolution.
Caveats and reticulate evolution.—The estimated
ages for some recently evolved populations/
taxa may be substantial overestimates because of
mtDNA polymorphism in ancestral species (see
Avise, 2000). The size of this error relative to
estimated age generally should be more pro-
nounced for younger taxa. For example, the es-
timated age of C. diabolis (
;
0.5 Myr) may be an
order of magnitude too great. Earlier workers
suggested an age of only 10,000 to 20,000 years
(Hubbs and Miller, 1948; Miller, 1981), which
seems reasonable from geology. Devils Hole, the
fault cavern supporting C. diabolis, apparently
did not open to the surface until about 50,000
years ago (Winograd et al., 1988), and this al-
most certainly preceded isolation of C. diabolis
from C. nevadensis.
If mtDNA diversity in ancestral C. nevadensis-
C. diabolis was near that of existing populations
of C. nevadensis, then C. diabolis would be much
younger than the 0.5 Myr minimum indicated
for its mtDNA. Both C. n. nevadensis and C. n.
mionectes are polymorphic for shared mtDNA
lineages that have been diverging from each
other and from the mtDNA of C. diabolis for
;
0.5–0.6 Myr. The lack of reciprocal monophyly
probably reflects incomplete lineage sorting
since isolation of C. diabolis. Such isolation
might have occurred anytime from the last few
thousand years to about 500,000 years ago, and
the former seems likely given the geological his-
tory of Devils Hole.
Various mtDNA/allozyme conflicts in phylo-
genetic inferences (Fig. 5) suggest that the bio-
geographic history of Cyprinodon is more com-
333ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
plex than indicated by mtDNA alone. Such con-
flicts might reflect random lineage sorting from
polymorphic ancestors, genetic introgression,
selection, or poor resolution with one or the
other dataset (Avise, 1997). Without further
data, selection is not a useful hypothesis to ex-
plain the observed conflicts, leaving lineage
sorting and introgression as potential evolution-
ary factors. As explained earlier, interpretations
are compromised by relatively low confidence
in phylogenetic resolution for allozyme varia-
tion in pupfish. Nonetheless, the major clades
indicated by allozymes generally are consistent
with the mtDNA tree, and most conflicts suggest
hypotheses of secondary contact and genetic in-
trogression that are geographically plausible.
The Rı´o Florido population of C. eximius
clearly illustrates reticulate evolution. Past hy-
bridization with C. macrolepis, a nearby, now-iso-
lated spring-dweller, is indicated by geographic
distributions of both allozymes (Echelle and
Echelle, 1998) and mtDNA lineages (Fig. 5).
With lineage sorting over time, the Rı´o Florido
population should become fixed for one or the
other of the divergent mtDNA lineages it now
carries, a process occurring more rapidly for
mtDNA than for nuclear genes because of the
effectively haploid nature and maternal trans-
mission of vertebrate mtDNA (Avise, 1994).
Such fixation following introgressive hybrid-
ization between divergent pupfishes might ex-
plain other mtDNA/allozyme conflicts indicat-
ed in Fig. 5. Three of the five remaining con-
flicts readily suggest reasonable hypotheses of
genetic introgression. Two involve locally co-oc-
curring species (C. bifasciatus and C. atrorus;C.
pachycephalus and C. eximius) in the only known
situations of natural contact and hybridization
between pupfish species in southwestern North
America (Minckley, 1969; Minckley and Minck-
ley, 1986). The third situation suggests contact
and introgressive hybridization between C. ele-
gans, which is endemic to isolated springs in the
relatively old lower Pecos River, and a C. pecosen-
sis-C. bovinus progenitor that was in the upper
Pecos River prior to the proposed integration
of the upper and lower basins.
The mtDNA/allozyme conflict in placement
of C. tularosa is more difficult to explain. It po-
tentially involved hybridization between a pup-
fish of the southern Great Plains-northern Chi-
huahuan Desert clade and a member of the
maritime clade that, by at least Early Pleisto-
cene, gained access to Tularosa Basin via the
Rio Grande (Fig. 5). Additional studies of nu-
clear DNA will be needed to evaluate this hy-
pothesis.
The previously mentioned mtDNA/allozyme
conflict for the western pupfishes (clade 5, Fig.
5) might be explained by mtDNA-lineage sort-
ing, either from an original ancestor in the Guz-
man Basin-lower Colorado River-Death Valley
region, or subsequent to secondary contact and
introgressive hybridization between divergent
pupfishes (Echelle and Echelle, 1993a). Regard-
ing the latter alternative, multiple instances of
secondary contact would explain the curious
lack of allozyme loci showing fixed differences
among eight of the nine species, the exception
being the peripherally distributed C. radiosus
(Echelle and Echelle, 1993a). Otherwise, the
only known examples of such a lack of fixed
allozyme differences in Cyprinodon are the sym-
patric species in Lake Chichancanab (Humph-
ries, 1984), the coastal species C. variegatus and
C. artifrons (Darling, 1976), and C. macrolepis
and C. eximius, which, as mentioned earlier,
were in contact relatively recently.
Comparisons with other organisms.—The allopatry
of the primary mtDNA clades of Cyprinodon
might reflect major vicariant events for the
aquatic biota of southwestern North America.
This predicts congruence in timing of cladoge-
netic events in pupfishes with those of other
aquatic organisms in the region.
Mitochondrial DNA surveys of two fishes,
Cyprinella lutrensis and Fundulus zebrinus, geo-
graphically overlap the southern Great Plains-
northern Chihuahuan Desert mtDNA-clade for
pupfishes. The cyprinid mirrored the marked
divergence (uncorrected
5
6.2%) between the
Red and Brazos river forms of C. rubrofluviatilis,
with 8.4% divergence (restriction-fragment-
length analysis) between a clade in the Brazos
River and another in more northern popula-
tions (Richardson and Gold, 1995). However,
the fundulid phylogeography seems more re-
cent, with a shallow clade (
;
2% divergence) of
populations from the Red, Brazos, and Pecos
rivers (Krieser, 2001; Krieser et al., 2001).
Relationships among the basal mtDNA clades
in Cyprinodon are unresolved, but the areas oc-
cupied by those in the ancestral Rio Grande Sys-
tem correspond well to those for four of May-
den’s (1989) morphologically recognized clades
of the Cyprinella formosa-lepida species complex:
Guzma´n Basin, Gulf Coastal Plain, Rı´o Conchos-
Middle Rio Grande, and Cuatro Cie´negas-Low-
er Rı´o Grande Basin. The results for pupfish
and the C. formosa-lepida group show little con-
gruence with phylogenetic inferences for other
fishes in the ancestral Rio Grande System, per-
haps because of differing tendencies to occupy
upland streams where dispersal or transport
across basin divides is more likely. Pupfishes and
334 COPEIA, 2005, NO. 2
the C. formosa-lepida group primarily occupy low-
lands and valley floors, whereas other groups
examined phylogenetically, including the cypri-
nid genus Dionda (Mayden et al., 1992) and
western catostomids (Smith, 1992a), are more
common in upland streams. The results for
Dionda do, however, parallel those for pupfish
by being consistent with recognition of an Old
Rı´o Nazas-lower Rio Grande system and by sup-
porting a sister relationship between an ances-
tral form in that system and species in the Rı´o
Conchos-Middle Rio Grande.
Hydrobiid snails often are restricted to isolat-
ed spring systems in southwestern North Amer-
ica and their phylogenetic pattern should be es-
pecially informative regarding patterns of vicar-
iance. Studies by Hershler et al. (1999a, 1999b,
2002) and Hurt (2004) indicate a more ancient
history than that of Cyprinodon, but both groups
support (1) Early Pliocene connections between
Guzma´n Basin, the Colorado River, and the
Death Valley System, (2) similar timing for con-
nections between eastward flowing southern
Great Plains streams and the Pecos River (Cypri-
nodon, 3.6 Mya; hydrobiids, 4–6 Mya), and (3)
the contact
;
2 Mya between the Tularosa Basin
and the Rio Grande. Further comparisons await
additional studies of hydrobiids, particularly
those on the Mexican Plateau.
Concluding comments.—The mtDNA tree for Cypri-
nodon seems to preserve signals of vicariance in
the Late Neogene biogeographic history of the
aquatic fauna of southwestern North America. It
appears, however, that mtDNA alone portrays an
incomplete picture of the historical biogeogra-
phy of Cyprinodon. The conflicting phylogenetic
inferences from mtDNA and allozymes indicate
that reticulate evolution involving divergent lin-
eages were important in the genetic history of
the group, and this probably is not unusual, at
least for teleost fishes. Hybridization is common
among and within extant genera of fishes
(Hubbs, 1955; Schwartz, 1972), and both the fos-
sil record (Smith, 1992b) and biochemical anal-
yses (Dowling and Hoeh, 1991; DeMarais et al.,
1992) indicate that introgression was not uncom-
mon in the past. The mtDNA/allozyme conflicts
in phylogenetic inferences suggest explicit hy-
potheses of vicariance and secondary contact
and introgression that can be tested with addi-
tional nuclear markers for Cyprinodon, phyloge-
netic analyses of other organisms in the region,
and geological studies of past surface-water con-
nections among basins.
M
ATERIAL
E
XAMINED
Collection localities for sequences generated
in this study. Parentheses show locality numbers
from Fig. 1, followed by number of specimens
surveyed for variation (RFLP or sequencing)
and number sequenced. Brackets
5
catalog
numbers for vouchers at Oklahoma State Uni-
versity Collection of Vertebrates (OSUS): Cypri-
nodon rubrofluviatilis (1:6/2) Prairie Dog Town
Fork of Red River, 20 km N Childress, Childress
Co., Texas, (2:6/2) Salt Fork of Brazos River at
Highway 380 bridge E Jayton, Stonewall Co.,
Texas; C. pecosensis (3:6/2) Salt Creek, at High-
way 285 bridge, Reeves Co., Texas; C. elegans
(4a:6/2) Giffin Spring canal at Toyahvale,
Reeves Co., Texas [18246], (4b:2/2) captive
stock at Uvalde National Fish Hatchery derived
from Phantom Lake Spring, Jeff Davis Co., Tex-
as; C. bovinus (5:6/2) Dexter National Fish
Hatchery stock derived from Diamond Y Draw,
Pecos Co., Texas; C. tularosa (6:2/2) Lost River
at Holloman Air Force Base, Otero Co., New
Mexico; C. eximius (7:2/2) Devils River at Paf-
fords Crossing, Val Verde Co., Texas, (8:1/1)
Alamito Creek, 5 km SE Presidio, Presidio Co.,
Texas, (9:2/2) Rı´o Florido, 1 km SE Villa Lopez,
Chihuahua [18240]; C. macrolepis (10:2/2) Ojo
de Hacienda Dolores, 11.2 km S Jimenez, Chi-
huahua [18228]; C. pachycephalus (11:2/2) Ojo
de San Diego, 57 km SE Chihuahua, Chihuahua
[18230]; C. fontinalis (12:2/2) Ojo de Carbo-
nera, 35 km W Villa Ahumada, Chihuahua
[18238]; C. pisteri (13:2/2) tributary of Rı´o Ca-
sas Grandes, 14.4 km W Janos, Chihuahua
[18237]; C. albivelis (14:2/2) springs ENE Ojo
de Arrey, 4.8 km SSE Galeana, Chihuahua
[18242]; C. nazas (15:2/2) outflow of Ojo de la
Concha, 9 km W Pen˜on Blanco, Durango
[18235], (16:2/2) tributary of Laguna Santia-
guillo, Rı´o Guatimape at Guatimape, Durango
[18244]; C. meeki (17:2/2) tributary of Rı´o del
Tunal, 40 km NE Durango, Durango [18224];
C. sp. Aguanaval (18:2/2) Rı´o Aguanaval at Ran-
cho Grande, Zacatecas [18232]; C. alvarezi (19:
2/2) irrigation canal from Ojo del Potosı´ at Eji-
do Catarino Rodriguez, Nuevo Leon [18243]; C.
veronicae (20:2/2) Charco Azul, 1 km S San Juan
de Avile´s, Nuevo Leon [18223]; C. atrorus (21:
1/1) Laguna Grande at inlet from Rı´o Churin-
ce, 15.5 Km SE Cuatro Cie´negas, Coahuila
[18231]; C. bifasciatus (22a:1/1) Pozo Churince
[18227] and (22b:1/1) Pozo Escobeda, respec-
tively, 16 km SE, 13 km S Cuatro Cie´negas, Coa-
huila; C. variegatus variegatus (23:12/2) Edin-
burg Water Supply Canal, 1 km N Edinburg, Hi-
dalgo Co., Texas [18317], (24:12/1) isolated
pool near Highway 90 bridge over Biloxi River,
335ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
Biloxi, Harrison County, Mississippi, (25:12/1)
roadside pool near Yankeetown, Levy Co., Flor-
ida, (26:12/2) Vero Beach at Riverside Park, In-
dian River Co., Florida, (27:12/1) Flagler Beach
at Highway 201, Flagler Co., Florida; C. variega-
tus ovinus, (28:12/1) S end of Chincoteaque Is-
land, Accomac Co., Virginia; C. bondi (29:1/1)
Lago del Fondo, Dominican Republic; C. sp.
‘‘Enriquillo pupfish’’ (30:1/1) Lake Enriquillo,
Dominican Republic; C. dearborni (31:2/2) Car-
lisle Salt Works at Canas, Island of Bonaire; C.
artifrons (32:1/1) tidepool at Rı´o Lagartos, Yu-
catan, (33:2/2) tidepool at Cancun, Quintana
Roo; C. radiosus (34:1/1) Warm Spring Refugi-
um, S of Bishop, Inyo Co., California [18309];
C. nevadensis (35:1/1) tailwaters of Shoshone
Spring in Amargosa River at State Highway 178
SE of Shoshone, Inyo Co., California; C. macu-
larius (36:1/1) captive population at Boyce
Thompson Arboretum, Pinal Co., Arizona, orig-
inally from Santa Clara Slough, Sonora; Megup-
silon aporus (37:1/1) captive stock from outdoor
tanks at Universidad de Autonoma de Nuevo
Leon, Monterrey, Nuevo Leon; Cualac tesselatus
(38:1/1) La Media Luna, 11 km S Rı´o Verde,
San Luis Potosı´ [18225]; Jordanella floridae (39:
1/1) 40 km NW West Palm Beach, Palm Beach
Co., Florida [18314].
A
CKNOWLEDGMENTS
We dedicate this paper to the memories of R.
Miller and W. Minckley. We thank S. Contreras
and L. Lozano for help with fieldwork and
many fruitful discussions; S. Cather, D. Love, R.
Myers, and especially J. Hawley, for consultation
on geology; R. Hershler for comments on drafts
of the manuscript; T. Bunt, R. Edwards, L.
Echelle, T. Echelle, H. Fitch, L. Fuselier, M.
Smith, B. Turner, and A. Valdez for help in ob-
taining specimens; M. Childs, S. Hoofer, L.
Knowles, T. Malloy, M. Hainey, G. Wilson, and
R. Pfau for help in the laboratory; T. Hrbek for
providing his dataset on Aphanius, S. Norris for
unpublished manuscripts, S. Fox for help with
the Spanish abstract, D. Swofford and M. Vences
for analytical advice, and the Government of
Mexico (SEDUE and Departamento de Pesca)
for permission to collect in 1989. The research
protocol satisfied requirements of the
Oklahoma State University Animal Care and
Use Committee. Collecting in Cuatro Ciene´gas
was done with permit DAH19990626-1-5 to D.
Hendrickson from the government of Mexico.
Funding provided by the National Science
Foundation (AAE and AFE, RVDB, TED, and
AM), the U.S. Fish and Wildlife Service (AAE
and AFE), Fond der Chemischen Industrie and
Deutsche Forschunggemeinschaft (to AM), and
the NASA Astrobiology Program (to TED).
L
ITERATURE
C
ITED
A
RELLANO
, A. R. V. 1951. Research on the continental
Neogene of Mexico. Am. J. Sci. 249:604–616.
A
VISE
, J. C. 1994. Molecular Markers, Natural History
and Evolution. Chapman and Hall, New York.
———. 1997. Identification and interpretation of mi-
tochondrial DNA stocks in marine species, p. 105–
136. In: Proceedings of the Stock Identification
Workshop. H. Kumph and E. L. Nakamura (eds.).
National Oceanographic and Atmospheric Admin-
istration, Panama City, Florida.
———. 2000. Phylogeography: The Histor y and For-
mation of Species. Harvard Univ. Press, Cambridge,
Masachusetts.
B
ERGGREN
, W. A., D. V. K
ENT
,C.C.S
WISHER
III,
AND
M.-P. A
UBRY
. 1995 A revised Cenozoic geochronol-
ogy and chronostratigraphy, p. 129–212. In: Geo-
chronology, Time Scales and Global Stratigraphic
Correlation. W. A. Berggren, D. V. Kent, M.-P.
Aubry, and J. Hardenbol (eds.). Soc. for Economic
Paleontology and Mineralogy, Tulsa, Oklahoma.
B
ERLOCHER
,S.H.,
AND
D. L. S
WOFFORD
. 1997. Search-
ing for phylogenetic trees under the frequency par-
simony criterion: an approximation using general-
ized parsimony. Syst. Biol. 46:211–215.
B
ERMINGHAM
, E., S. S. M
C
C
AFFERTY
,
AND
A. P. M
ARTIN
.
1997. Fish biogeography and molecular clocks: Per-
spectives from the Panamanian isthmus, p. 113–
128. In: Molecular Systematics of Fishes. T. D. Ko-
cher and C. A. Stepien (eds.). Academic Press, New
York.
B
ROUGHTON
,R.E.,
AND
J. R. G
OLD
. 2000. Phylogenetic
relationships in the North American cyprinid genus
Cyprinella (Actinopterygii: Cyprinidae) based on se-
quences of the mitochondrial ND2 and ND4L
genes. Copeia 2000:1–10.
B
UNT
, T. M. 2001. Reproductive isolation and diver-
gence in a young ‘‘species flock’’ of pupfish (Cypri-
nodon) from San Salvador Island, Bahamas. Un-
publ. M.S. thesis, Virginia Polytechnic Institute and
State University, Blacksburg, Virginia.
C
HAPIN
,C.E.,
AND
S. M. C
ATHER
. 1994. Tectonic set-
ting of the axial basins of the northern and central
Rio Grande rift, p. 5–25. In: Basins of the Rio
Grande Rift: Structure, Stratigraphy, and Tectonic
Setting. G. R. Keller, and S. M. Cather (eds.). Geol.
Soc. Amer. Spec. Pap. 291.
C
HILDS
, M. R., A. A. E
CHELLE
,
AND
T. E. D
OWLING
.
1996. Development of the hybrid swarm between
Pecos pupfish (Cyprinodontidae: Cyprinodon peco-
sensis) and sheepshead minnow (C. variegatus): a
perspective from allozymes and mtDNA. Evolution
50:2014–2022.
C
OKENDOLPHER
, J. C. 1980. Hybridization experiments
with the genus Cyprinodon (Teleostei: Cyprinodon-
tidae). Copeia 1980:173–176.
C
ONANT
, R. 1963. Semiaquatic snakes of the genus
Thamnophis from the isolated drainage system of
336 COPEIA, 2005, NO. 2
the Rio Nazas and adjacent areas in Mexico. Ibid.
1963:473–499.
C
ONNER
,J.V.,
AND
R. D. S
UTTKUS
. 1986. Zoogeogra-
phy of freshwater fishes of the western Gulf slope,
p. 413–456. In: The Zoogeography of North Amer-
ican Freshwater Fishes. C. H. Hocutt and E. O. Wi-
ley (eds.). John Wiley and Sons, New York.
C
ONTRERAS
-B.,
AND
M
A
.
DE
L
OZANO
. 1996. Extinction
of most Sandı´a and Potosı´ valleys (Nuevo Leo´n,
Mexico) endemic pupfishes, crayfishes and snails.
Ichthyol. Explor. Freshwaters 7:33–40.
C
OSTA
, W. J. E. M. 1997. Phylogeny and classification
of the Cyprinodontidae revisited (Teleostei: Cypri-
nodontiformes): Are Andean and Anatolian killi-
fishes sister taxa? J. Comp. Biol. 2:1–17.
C
OVICH
, A.,
AND
M. S
TUIVER
. 1974. Changes in oxygen
18 as a measure of long-term fluctuations in tropi-
cal lake levels and molluscan populations. Limnol.
Oceanogr. 19:682–691.
D
ARLING
, J. 1976. Electrophoretic variation in Cypri-
nodon variegatus and systematics of some fishes of
the subfamily Cyprinodontinae. Unpubl. Ph.D.
diss., Yale Univ., New Haven, Connecticut.
D
E
M
ARAIS
, B. D., T. E. D
OWLING
,M.E.D
OUGLAS
,W.
L. M
INCKLEY
,
AND
P. C. M
ARSH
. 1992. Origin of Gila
seminuda (Teleostei: Cyprinidae) through introgres-
sive hybridization: Implications for evolution and
conservation. Proc. Nat. Acad. Sci. USA 89:2747–
2751.
D
OADRIO
, I., J. A. C
ARMONA
,
AND
C. F
ERNANDEZ
-D
EL
-
GADO
. 2002. Morphometric study of the Iberian Ap-
hanius (Actinopterygii, Cyprinodontiformes), with
description of a new species. Folia Zool. 51:67–79.
D
OWLING
,T.E.,
AND
W. R. H
OEH
. 1991. The extent of
introgression outside the hybrid zone between No-
tropis cornutus and Notropis chr ysocephalus (Teleostei:
Cyprinidae). Evolution 45:944–956.
D
OYLE
,J.A.,
AND
M. J. D
ONOGHUE
. 1993. Phylogenies
and angiosperm diversification. Paleobiol. 19:141–
167.
D
UVERNELL
, D. D.,
AND
B. J. T
URNER
. 1998. Evolution-
ary genetics of Death Valley pupfish populations:
mitochondrial DNA sequence variation and popu-
lation structure. Mol. Ecol. 7:279–288.
E
CHELLE
,A.A.,
AND
P. J. C
ONNOR
. 1989. Rapid, geo-
graphically extensive genetic introgression after
secondary contact between two pupfish species
(Cyprinodon, Cyprinodontidae). Evolution 43:717–
727.
———,
AND
T. E. D
OWLING
. 1992. Mitochondrial DNA
variation and evolution of the Death Valley pupfish-
es (Cyprinodon, Cyprinodontidae). Ibid. 46:193–206.
———,
AND
A. F. E
CHELLE
. 1978. The Pecos River
pupfish, Cyprinodon pecosensis n. sp. (Cyprinodonti-
dae), with comments on its evolutionary origin. Co-
peia 1978:569–582.
———,
AND
———. 1992. Mode and pattern of spe-
ciation in the evolution of inland pupfishes of the
Cyprinodon variegatus complex (Teleostei: Cyprino-
dontidae): an ancestor-descendant hypothesis, p.
691–709. In: Systematics, Historical Ecology and
North American Freshwater Fishes. R. L. Mayden
(ed.). Stanford Univ. Press, Stanford, California.
———,
AND
———. 1993a. Allozyme perspective on
mitochondrial DNA variation and evolution of the
Death Valley pupfishes (Cyprinodontidae: Cyprino-
don). Copeia 1993:275–287.
———,
AND
———. 1993b. Allozyme variation and
systematics of the New World cyprinodontines (Te-
leostei: Cyprinodontidae). Biochem. Syst. Ecol. 21:
583–590.
———,
AND
———. 1996. Genetic introgression of
endemic taxa by non-natives: a case study with Leon
Springs pupfish and sheepshead minnow. Conserv.
Biol. 11:153–161.
———,
AND
———. 1998. Evolutionar y relationships
of pupfishes in the Cyprinodon eximius complex (At-
herinomorpha: Cyprinodontiformes). Copeia 1998:
852–865.
———, ———, S. C
ONTRERAS
-B.,
AND
L. L
OZANO
-V.
1995. Genetic variation in the endangered fish fau-
na (Atheriniformes: Cyprinodontidae) associated
with Pluvial Lake Sandia, Nuevo Leon, Mexico.
Southwest. Nat. 40:11–17.
———, R. A. V
AN
D
EN
B
USSCHE
,T.P.M
ALLOY
,J
R
., M.
L. H
AYNIE
,
AND
C. O. M
INCKLEY
. 2000. Mitochon-
drial DNA variation in pupfishes assigned to the
species Cyprinodon macularius (Atherinomorpha:
Cyprinodontidae): taxonomic implications and
conservation genetics. Copeia 2000:353–364.
E
CHELLE
,A.F.,
AND
A. A. E
CHELLE
. 1994. Assessment
of genetic introgression between two pupfish spe-
cies, Cyprinodon elegans and C. variegatus (Cyprino-
dontidae), after more than 20 years of secondary
contact. Ibid. 1994:590–597.
F
ELSENSTEIN
, J. 1985. Confidence limits on phyloge-
nies: an approach using the bootstrap. Evolution
39:783–791.
G
USTAVSON
, T. C.,
AND
R. J. F
INLEY
. 1985. Late Ceno-
zoic geomorphic evolution of the Texas Panhandle
and northeastern New Mexico: case studies of
structural controls on regional drainage develop-
ment. Univ. Texas Bur. Econ. Geol. Rep. Invest.
148:1–42.
H
AWLEY
, J. W., G. O. B
ACHMAN
,
AND
K. M
ANLEY
. 1976.
Quaternary stratigraphy in the Basin and Range
and Great Plains provinces, New Mexico and west-
ern Texas, p. 235–274. In: Quaternary Stratigraphy
of North America. W. C. Mahaney (ed.). Dowden,
Hutchinson, and Ross, Stroudsburg, Pennsylvania.
H
RBEK
,T.,
AND
A. M
EYER
. 2003. Closing of the Tethys
Sea and the phylogeny of Eurasian killifishes
(Cyprinodontiformes: Cyprinodontidae). J. Evol.
Biol. 16:17–36.
H
ERSHLER
, R., H.-P. L
IU
,
AND
M. M
ULVEY
. 1999a. Phy-
logenetic relationships within the aquatic snail ge-
nus Tryonia: Implications for biogeography of the
North American southwest. Mol. Phyl. Evol. 13:377–
391.
———, ———,
AND
C. A. S
TOCKWELL
. 2002. A new
genus and species of aquatic gastropods (Rissooi-
dea: Hydrobiidae) from the North American South-
west: phylogenetic relationships and biogeography.
Proc. Biol. Soc. Washington 115:171–188.
———, M. M
ULVEY
,
AND
H.-P. L
IU
. 1999b. Biogeog-
raphy in the Death Valley region: evidence from
springsnails (Hydrobiidae: Tryonia). Zool. J. Linn.
Soc. 126:335–354.
337ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
H
ILLIS
, D. M., A. L
ARSON
,S.K.D
AVIS
,
AND
E. A. Z
IM
-
MER
. 1990. Nucleic acids III: sequencing, p. 318–
372. In: Molecular Systematics. D. M. Hillis and C.
Moritz (eds.). Sinauer, Sunderland, Massachusetts.
H
OCUTT
, C. H.,
AND
E. O. W
ILEY
. 1986. The Zooge-
ography of North American Freshwater Fishes.
John Wiley and Sons, New York.
H
OLTMEIER
, C. L. 2001. Heterochrony, maternal ef-
fects, and phenotypic variation among sympatric
pupfishes. Evolution 55:330–338.
H
OWARD
, J. L. 1996. Paleocene to Holocene paleo-
deltas of ancestral Colorado River offset by the San
Andreas fault system, southern California. Geology
24:783–786.
H
UBBS
, C. L. 1955. Hybridization between fish species
in nature. Syst. Zool. 4:1–20.
———,
AND
R. R. M
ILLER
. 1948. The Great Basin, with
emphasis on glacial and postglacial times. II. The
zoological evidence. Bull. Univ. Utah 38 Biol. Ser.
10:17–166.
H
ULSENBECK
,J.P.,
AND
F. R
ONQUIST
. 2001. MRBAYES:
Bayesian inference of phylogeny. Bioinformatics 17:
754–755.
H
UMPHRIES
, J. M. 1984. Genetics of speciation in pup-
fishes from Laguna Chichancanab, Mexico, p. 129–
140. In: Evolution of Fish Species Flocks. A. A.
Echelle and I. Kornfield (eds.). Univ. of Maine
Press, Orono, Maine.
———,
AND
R. R. M
ILLER
. 1981. A remarkable species
flock of Cyprinodon from Lake Chichancanab, Yu-
catan, Mexico. Copeia 1981:52–64.
H
URT
, C. R. 2004. Genetic divergence, population
structure and historical demography of rare
springsnails (Pyrgulopsis) in the lower Colorado Riv-
er basin. Mol. Ecol. 13:1173–1187.
K
ISHINO
, H., J. L. T
HORNE
,
AND
W. J. B
RUNO
. 2001.
Performance of a divergence time estimation meth-
od under a probabilistic model of rate evolution.
Mol. Biol. Evol. 18:352–361.
K
OCHER
, T. D., W. K. T
HOMAS
,A.M
EYER
,S.V.E
D
-
WARDS
,S.P
A
¨A
¨BO
,F.X.V
ILLABLANCA
,
AND
A. C. W
IL
-
SON
. 1989. Dynamics of mitochondrial DNA evolu-
tion in animals: amplification and sequencing with
conserved primers. Proc. Natl. Acad. Sci. USA 86:
6196–6200.
K
REISER
, B. R. 2001. Mitochondrial cytochrome b se-
quences support recogition of two cryptic species
of plains killifish, Fundulus zebrinus and Fundulus
kansae. Am. Midl. Nat. 146:199–209.
———, J. B. M
ITTON
,
AND
J. D. W
OODLING
. 2001.
Phylogeography of the plains killifish, Fundulus ze-
brinus. Evolution 2001:339–350.
K
UMAR
, S., K. T
AMURA
,I.B.J
AKOBSEN
,
AND
M. N
EI
.
2001. MEGA2: Molecular Evolutionary Genetics
Analysis software. Bioinformatics 17:1244–1245.
L
ONGMIRE
, J. L., M. M
ALTBIE
,
AND
R. J. B
AKER
. 1997.
Use of ‘‘Lysis Buffer’’ in DNA isolation and its im-
plication for museum collections. Occ. Pap. Mus.
Tex. Tech Univ. 163:1–3.
L
OVEJOY
, N. R.,
AND
B. B. C
OLLETTE
. 2001. Phyloge-
netic relationships of New World needlefishes (Te-
leostei: Belonidae) and the biogeography of tran-
sitions between marine and freshwater habitats. Co-
peia 2001:324–338.
L
OZANO
-V., M. L. 2002. Cyprinodon salvadori, new spe-
cies from the upper Rio Conchos, Chihuahua, Mex-
ico, with a revised key to the C. eximius complex
(Pisces, Teleostei: Cyprinodontidae, p. 15–22. In:
Libro Jubilar en Honor al Dr. Salvador Contreras
Balderas. M. L. Lozano-V. (ed.). Univ. Autonoma
de Nuevo Leon, Monterrey, Mexico.
———,
AND
S. C
ONTRERAS
-B. 1993. Four new species
of Cyprinodon from southern Nuevo Leon, Mexico,
with a key to the C. eximius complex (Teleostei:
Cyprinodontidae). Ichthyol. Explor. Freshwaters 4:
295–308.
———,
AND
———. 1999. Cyprinodon bobmilleri: New
species of pupfish from Nuevo Leo´n, Me´xico (Pi-
sces: Cyprinodontidae). Copeia 1999:382–387.
L
YNCH
, D. J. 1981. A model for volcanism in the Pi-
nacate volcanic field of northwestern Sonora, Mex-
ico. Abstr. Geol. Soc. Am. 13:93.
M
ACK
, G. H., S. L. S
ALYARDS
,
AND
W. C. J
AMES
. 1993.
Magnetostratigraphy of the Plio-Pleistocene Camp
Rice and Palomas formations in the Rio Grande rift
of southern New Mexico: Amer. J. Sci. 293:49–77.
———, W. C. M
C
I
NTOSH
,M.R.L
EEDER
,
AND
H. C.
M
ONGER
. 1996. Plio-Pleistocene pumice floods in
the ancestral Rio Grande, southern Rio Grande rift,
USA. Sediment. Geol. 103:1–8.
M
AYDEN
, R. L. 1989. Phylogenetic studies of North
American minnows, with emphasis on the genus
Cyprinella (Teleostei: Cypriniformes). Misc. Publ.
Mus. Nat. Hist., Univ. Kansas 80:1–189.
———, R. H. M
ATSON
,
AND
D. M. H
ILLIS
. 1992. Spe-
ciation in the North American genus Dionda (Te-
leostei: Cypriniformes), p. 710–746. In: Systematics,
Historical Ecology and North American Freshwater
Fishes. R. L. Mayden (ed.). Stanford Univ. Press,
Stanford, California.
M
EEK
, S. E. 1904. The fresh-water fishes of Mexico
north of the Isthmus of Tehuantepec. Field Colum-
bian Mus. Zool. 93(Zool. Ser. 5):1–252.
M
EYER
, A., J. M. M
ORISSEY
,
AND
M. S
CHARTL
. 1994. Re-
current origin of a sexually selected trait in Xiphop-
horus fishes inferred from a molecular phylogeny.
Nature 368:539–542.
———, T. D. K
OCHER
,P.B
ASASIBWAKI
,
AND
A. C. W
IL
-
SON
. 1990. Monophyletic origin of Lake Victoria
fishes suggested by mitochondrial DNA sequences.
Ibid. 347:550–553.
M
ILLER
, R. R. 1945. Four new species of fossil cyprin-
odont fishes from eastern California. J. Wash. Acad.
Sci. 35:315–321.
———. 1948. The cyprinodont fishes of the Death
Valley system of eastern California and southwest-
ern Nevada. Misc. Publ. Mus. Zool., Univ. of Mich-
igan 68:1–155.
———. 1962. Taxonomic status of Cyprinodon baconi,
a killifish from Andros Island, Bahamas. Copeia
1962:836–837.
———. 1964. Redescription and illustration of Cypri-
nodon latifasciatus, an extinct cyprinodontid fish
from Coahuila, Mexico. Southwest. Nat. 9:62–67.
———. 1976. Four new fishes of the genus Cyprinodon
from Mexico, with a key to the Cyprinodon eximius
complex. Bull. So. Calif. Acad. Sci. 75:68–75.
———. 1981. Coevolution of deserts and pupfishes
338 COPEIA, 2005, NO. 2
(genus Cyprinodon) in the American southwest, p.
39–94. In: Fishes in North American deserts. R. J.
Naiman and D. L. Soltz (eds.). John Wiley and
Sons, New York.
———,
AND
E. P. P
ISTER
. 1971. Management of the
Owens pupfish, Cyprinodon radiosus, in Mono Co.,
California. Trans. Amer. Fish. Soc. 100:502–509.
M
INCKLEY
, W. L. 1969. Environments of the Bolso´n of
Cuatro Cie´ negas, Coahuila, Me´ xico, with special
reference to the aquatic biota. Sci. Ser., Univ. Texas
El Paso 2:1–65.
———,
AND
C. O. M
INCKLEY
. 1986. Cyprinodon pachy-
cephalus, a new species of pupfish (Cyprinodonti-
dae) from the Chihuahuan Desert of northern
Mexico. Copeia 1986:184–192.
———, D. A. H
ENDRICKSON
,
AND
C. E. B
OND
. 1986.
Geography of western North American freshwater
fishes: description and relationships to intraconti-
nental tectonism, p. 519–614. In: The Zoogeogra-
phy of North American Freshwater Fishes. C. H.
Hocutt and E. O. Wiley (eds.). John Wiley and
Sons, New York.
———, R. R. M
ILLER
,
AND
S. M. N
ORRIS
. 2002. Three
new pupfish species, Cyprinodon (Teleostei, Cypri-
nodontidae), from Chihuahua, Mexico, and Arizo-
na, USA. Copeia 2002:687–705.
P
ARKER
, A.,
AND
I. K
ORNFIELD
. 1995. Molecular per-
spective on evolution and zoogeography of cypri-
nodontid killifishes (Teleostei: Atherinomorpha).
Ibid. 1995:8–21.
P
ERDICES
, A., J. A. C
ARMONA
,C.F
ERNANDEZ
-D
ELGADO
,
AND
I. D
OADRIO
. 2001. Nuclear and mitochondrial
data reveal high genetic divergence among Atlantic
and Mediterranean populations of the Iberian kil-
lifish Aphanius iberus (Teleostei: Cyprinodontidae).
Heredity 87:314–324.
P
OSADA
, D.,
AND
K. A. C
RANDALL
. 1998. MODELTEST:
Testing the model of DNA substitution. Bioinfor-
matics 14:817–818.
R
ANNALA
, B.,
AND
Z. Y
ANG
. 1996. Probability distribu-
tion of molecular evolutionar y trees: A new method
of phylogenetic inference. J. Mol. Evol. 43:304–311.
R
EPENNING
, C. A.,
AND
S. R. M
AY
. 1986. New evidence
for the age of lower part of the Palomas Formation,
Truth or Consequences, New Mexico. N. M. Geol.
Soc., Guideb. 37:257–263.
R
ICHARDSON
, L. R.,
AND
J. R. G
OLD
. 1995. Evolution
of the Cyprinella lutrensis species group. III. Geo-
graphic variation in the mitochondrial DNA of
Cyprinella lutrensis–the influence of Pleistocene gla-
ciation on population dispersal and divergence.
Mol. Ecol. 4:163–171.
S
ANDERSON
, M. 1997. A nonparametric approach to
estimate divergence times in the absence of rate
constancy. Mol. Biol. Evol. 14:1218–1231.
S
CHMIDT
, T. R., J. P. B
IELAWSKI
,
AND
J. R. G
OLD
. 1998.
Molecular phylogenetics and evolution of the cy-
tochrome bgene in the cyprinid genus Lythrurus
(Actinopterygii: Cypriniformes). Copeia 1998:14–
22.
S
CHWARTZ
, F. J. 1972. World literature to fish hybrids,
with an analysis by family, species, and hybrid. Publ.
Gulf Coast Res. Lab. Mus. 3:1–328.
S
EAGER
, W. R., M. S
HAFIQUILLAH
,J.W.H
AWLEY
,
AND
R. F. M
ARVIN
. 1984. New K-Ar dates from basalts
and the evolution of the southern Rio Grande rift.
Geol. Soc. Am. Bull. 95:87–99.
S
MITH
, G. R. 1992a. Phylogeny and biogeography of
the Catostomidae, freshwater fishes of North Amer-
ica and Asia, p. 778–826. In: Systematics, Historical
Ecology, and North American Freshwater Fishes. R.
L. Mayden (ed.). Stanford Univ. Press, Stanford,
California.
———. 1992b. Introgression in fishes: significance for
paleontology, cladistics, and evolutionary rates. Syst.
Biol. 41:41–57.
———, T. E. D
OWLING
,K.G
OBALET
,T.L
UGASKI
,D.
S
HIOZAWA
,
AND
P. E
VANS
. 2002. Biogeography and
rates of evolution of Great Basin fishes, p. 175–234.
In: The Great Basin: Cenozoic Geology and Bioge-
ography. R. Hershler and D. Curry (eds.). Smith-
sonian Institution, Washington, DC.
S
MITH
,M.L.,
AND
R. R. M
ILLER
. 1986. The evolution
of the Rio Grande Basin as inferred from its fish
fauna, p. 457–485. In: The Zoogeography of North
American Freshwater Fishes. C. H. Hocutt and E.
O. Wiley (eds.). John Wiley and Sons, New York.
———, C. M. R
ODRIGUEZ
,
AND
C. L
YDEARD
. 1990. Sys-
tematics of Cyprinodon higuey n. sp. and Cyprinodon
jamaicensis Fowler from the Greater Antilles (Te-
leostei: Cyprinodontiformes). Am. Mus. Nov. 2990:
1–10.
S
TOCKWELL
, C. A., M. M
ULVEY
,
AND
A. G. J
ONES
. 1998.
Genetic evidence for two evolutionarily significant
units of White Sands pupfish. Animal Conserv. 1:
213–225.
S
TRAIN
, W. S. 1966. Blancan mammalian fauna and
Pleistocene formations, Hudspeth County, Texas.
Bull. Tex. Mem. Mus. 10:1–55.
S
TRECKER
, U. 2002. Cyprinodon esconditus, a new pup-
fish from Laguna Chichancanab, Yucatan, Mexico
(Cyprinodontidae). Cybium 26:301–307.
———,
AND
A. K
ODRIC
-B
ROWN
. 2000. Mating prefer-
ences in a species flock of Mexican pupfish. Biol. J.
Linn. Soc. 71:677–687.
———, C. G. M
EYER
,C.S
TURMBAUER
,
AND
H. W
IL
-
KENS
. 1996. Genetic divergence and speciation in
an extremely young species flock in Mexico formed
by the genus Cyprinodon (Cyprinodontidae, Teleos-
tei). Mol. Phyl. Evol. 6:143–149.
T
AJIMA
, F. 1993. Simple methods for testing molecular
clock hypothesis. Genetics 135:599–607.
T
HOMPSON
, J. D., D. G. H
IGGINS
,
AND
T. J. G
IBSON
.
1994. CLUSTAL W: improving the sensitivity of pro-
gressive multiple sequence alignment through se-
quence weighting, position-specific gap penalties
and weight matrix choice. Nuc. Acids Res. 22:4673–
4680.
T
URNER
, B. J. 1974. Genetic divergence of Death Val-
ley pupfish species: biochemical versus morpholog-
ical evidence. Evolution 28:281–294.
———. 1983. Genic variation and differentiation of
remnant natural populations of the desert pupfish,
C. macularius.Ibid. 37:690–700.
———,
AND
R. K. L
IU
. 1977. Extensive interspecific
genetic compatibility in the New World killifish ge-
nus Cyprinodon. Copeia 1977:259–269.
V
ILLWOCK
, W. 1982. Aphanius (Nardo, 1827) and
339ECHELLE ET AL.—PUPFISH BIOGEOGRAPHY
Cyprinodon (Lac. 1803) (Pisces: Cyprinodontidae),
an attempt for a genetic interpretation of specia-
tion. Z. Zool. Syst. Evol. 20:187–197.
W
EBB
, S. A. 1998. Phylogenetic studies of the family
Goodeidae (Teleostei, Cyprinodontiformes). Un-
publ. Ph.D. Diss., Univ. Michigan, Ann Arbor, Mich-
igan.
W
EIGMAN
, B. M., D. K. Y
EATES
,J.L.T
HORNE
,
AND
H.
K
ISHINO
. 2003. Time flies, a new molecular time-
scale for brachyceran fly evolution without a clock.
Syst. Biol. 52:745–756.
W
INOGRAD
, I. J., B. J. S
ZABO
,T.B.C
OPLEN
,
AND
A. C.
R
IGGS
. 1988. A 250,000-year climatic record from
Great Basin vein calcite: Implications for Milanko-
vitch theory. Science 242:1275–1280.
W
ILDEKAMP
, R. H. 1995. A world of killies: Atlas of the
oviparous cyprinodontiform fishes of the world. Vol
II. American Killifish Association, Mishawaka, In-
diana.
Y
ANG
, Z. 1997. PAML: a program package for phylo-
genetic analysis by maximum likelihood. Comput.
Apl. Biosci. 13:555–556.
(AAE, AFE, RAVDB) Z
OOLOGY
D
EPARTMENT
,
O
KLAHOMA
S
TATE
U
NIVERSITY
,S
TILLWATER
,
O
KLAHOMA
74078; (EWC, TED) D
EPARTMENT
OF
B
IOLOGY
,A
RIZONA
S
TATE
U
NIVERSITY
,T
EM
-
PE
,A
RIZONA
85287; (AM) D
EPARTMENT OF
B
I
-
OLOGY
,U
NIVERSITY OF
K
ONSTANZ
, D-78457
K
ONSTANZ
,G
ERMANY
. E-mail: (AAE) echelle@
okstate.edu. Send reprint requests to AAE.
Submitted: 21 April 2003. Accepted: 31 Dec.
2004. Section editor: R. M. Wood.
