Reconciling actual and inferred population histories in the house finch (Carpodacus mexicanus) by AFLP analysis.

Abstract

The house finch (Carpodacus mexicanus) is a native songbird of western North America that was introduced to the eastern United States and Hawaiian Islands in historic times. As such, it provides an unusually good opportunity to test the ability of molecular markers to recover recent details of a known population history. To investigate this prospect, genetic variation in 172 individuals from 16 populations in the western and eastern United States, southeastern Canada, Hawaiian Islands, and Mexico, as well as genetic variation in the closely related purple finch (Carpodacus purpureus) and Cassin's finch (Carpodacus cassinii) was studied by a semi-automated fluorescence-labeled amplified fragment length polymorphism (AFLP) marker system. A total of 363 markers were generated, of which 258 (71.2%) were polymorphic among species, 166 (61.4%) polymorphic among house finch subspecies, and 157 (60.2%) polymorphic among populations within the frontalis subspecies complex. Heterozygosities and interpopulation divergences revealed by the analysis appeared relatively low at all taxonomic levels, but there are few similar studies in avian populations with which to compare results. Whereas the known population history predicts that both eastern and Hawaiian finches should have been derived from within western populations, tree analysis using both populations and individuals as units suggests weak monophyly of eastern populations and indicates that Hawaiian populations are not clearly derived from California populations. However, the genetic distinctiveness of native and recently founded populations was disclosed by analyses of molecular variance as well as by a model-based assignment approach in which 98%, 94%, and 99% individuals from western, Hawaiian, and eastern regions, respectively, were assigned correctly to their populations without using prior information on population of origin, suggesting that these recent introductions have resulted in detectable differentiation without substantial loss of AFLP diversity. Our results indicate that AFLPs are a useful tool for population genetic and evolutionary studies of birds, particularly as a prelude to finding molecular markers linked to traits subjected to recent adaptive evolution.

Full text

Page 1

2852
? 2003 The Society for the Study of Evolution. All rights reserved.
Evolution, 57(12), 2003, pp. 2852–2864
RECONCILING ACTUAL AND INFERRED POPULATION HISTORIES IN THE HOUSE
FINCH (CARPODACUS MEXICANUS) BY AFLP ANALYSIS
ZHENSHAN WANG,1,2ALLAN J. BAKER,3,4GEOFFREY E. HILL,5,6AND SCOTT V. EDWARDS1,7
1Department of Biology, University of Washington, Seattle, Washington 98195
2E-mail: zhenshan@u.washington.edu
3Center for Biodiversity and Conservation Biology, Royal Ontario Museum, Toronto, Canada, M5S 2C6
4E-mail: allanb@rom.on.ca
5Department of Biological Sciences, Auburn University, Auburn, Alabama 36849-5414
6E-mail: ghill@acesag.auburn.edu
7E-mail: sedwards@u.washington.edu
Abstract.
to the eastern United States and Hawaiian Islands in historic times. As such, it provides an unusually good opportunity
to test the ability of molecular markers to recover recent details of a known population history. To investigate this
prospect, genetic variation in 172 individuals from 16 populations in the western and eastern United States, southeastern
Canada, Hawaiian Islands, and Mexico, as well as genetic variation in the closely related purple finch (Carpodacus
purpureus) and Cassin’s finch (Carpodacus cassinii) was studied by a semi-automated fluorescence-labeled amplified
fragment length polymorphism (AFLP) marker system. A total of 363 markers were generated, of which 258 (71.2%)
were polymorphic among species, 166 (61.4%) polymorphic among house finch subspecies, and 157 (60.2%) poly-
morphic among populations within the frontalis subspecies complex. Heterozygosities and interpopulation divergences
revealed by the analysis appeared relatively low at all taxonomic levels, but there are few similar studies in avian
populations with which to compare results. Whereas the known population history predicts that both eastern and
Hawaiian finches should have been derived from within western populations, tree analysis using both populations and
individuals as units suggests weak monophyly of eastern populations and indicates that Hawaiian populations are not
clearly derived from California populations. However, the genetic distinctiveness of native and recently founded
populations was disclosed by analyses of molecular variance as well as by a model-based assignment approach in
which 98%, 94%, and 99% individuals from western, Hawaiian, and eastern regions, respectively, were assigned
correctly to their populations without using prior information on population of origin, suggesting that these recent
introductions have resulted in detectable differentiation without substantial loss of AFLP diversity. Our results indicate
that AFLPs are a useful tool for population genetic and evolutionary studies of birds, particularly as a prelude to
finding molecular markers linked to traits subjected to recent adaptive evolution.
The house finch (Carpodacus mexicanus) is a native songbird of western North America that was introduced
Key words.
phylogeography.
Amplified fragment length polymorphism, analysis of molecular variance, gene diversity, house finch,
Received March 10, 2003.Accepted July 7, 2003.
Studies of population genetic structure provide windows
into the roles that selection, mutation, gene flow, and drift
play in processes such as local adaptation and speciation
(Barton and Clark 1990; Avise 1994; Slatkin 1994; Foster et
al. 1998). Because the actual histories of most species are
unknown, many population genetic analyses focusing on geo-
graphic patterns are necessarily based on the assumption that
the analyzed populations are in equilibrium with respect to
these forces. In recent years, however, the possibility of non-
equilibrium situations in many species has become clearer,
and a number of new methods promise to deal better with
nonequilibrium situations (Kuhner et al. 1998; Polanski et
al. 1998; Templeton 1998; Weiss and von Haesler 1998; Em-
erson et al. 2001). In the face of complex, nonequilibrium
situations, we can expect an enhanced ability to study the
process of speciation when we employ molecular genetic
tools to species with recent demographies whose details have
been recorded historically (Clegg et al. 2002).
The house finch (Carpodacus mexicanus) is a cardueline
finch with a wide distribution in North America. Before the
landscapes were modified by Europeans house finches were
found from Oaxaca, Mexico, to central Oregon and east to
Colorado and Texas (American Ornithologists’ Union 1983;
Hill 2002), with several morphologically distinct subspecies
in Mexico and on islands in the Pacific and Gulf of California.
Moore (1939) recognized 18 subspecies of house finches, but
Hill (1996) recently reduced this to 14 subspecies. All of the
house finches in the United States and Canada, except for
birds in the extreme lower Rio Grande Valley, belong to one
supspecies, C. m. frontalis (Hill 1996).
House finches probably originating from coastal California
and from the subspecies C. m. frontalis were introduced to
the Hawaiian Islands sometime before 1870 (Grinnell 1911)
and to Long Island, New York, in 1940 (Elliot and Arbib
1953). There is no record of the number of house finches that
were released in Hawaii, but because the birds were trans-
ported slowly in wooden ships, the number has been assumed
to be small. Wild house finches were first noticed in eastern
North America in New York City in 1940 (Elliot and Arbib
1953), but the number of birds that founded the eastern pop-
ulation is unknown. Hill (2002) speculated that, based on the
large number of house finches that were imported from Cal-
ifornia to East Coast cities in the early 20th century, 50 or
more individuals may have founded the eastern population
of house finches.
Since they were introduced to the New York City area 60
years ago, the western boundary of the eastern population’s
range has reached the easternmost extension of some western
populations, although populations in the Great Plains are
sparse and gene flow across this region is likely low. Despite

Page 2

2853
AFLP ANALYSIS OF THE HOUSE FINCH
the very recent divergence of eastern and western populations
of this species, differences in morphological characters, be-
havior, and plumage colors have been detected in east-west
comparisons and at a variety of taxonomic levels both within
and between subspecies (Aldrich and Weske 1978; Power
1979; Aldrich 1982; Sprenkle and Blem 1984; Stangel 1985;
Wootton 1986). Many of these differences, including hatch-
ing order and sex allocation, are now known to be adaptive
(Badyaev et al. 2002).
Two unpublished master’s theses that examined genetic
diversity among native California populations and introduced
populations of house finches with mitochondrial DNA re-
striction fragment length polymorphisms and allozyme an-
alyis showed that house finches in the eastern United States
have retained most of the genetic diversity found in house
finches sampled in California. Similarly, house finch popu-
lations in the Hawaiian Islands have levels of genetic diver-
sity similar to those within California populations, although
in some comparisons there was loss of rare alleles (Benner
1991; Vasquez-Phillips 1992). These studies suggest that the
founder effect in this species—the shift in allele frequencies
accompanying the founding of novel populations—has not
been dramatic. The molecular tools used for these studies,
namely allozyme electrophoresis and restriction fragment
analysis of mitochondrial DNA, are less sensitive than some
recently developed methods. In recent years, a variety of
DNA-based techniques have been employed to study varia-
tion within and among species, for example, restriction frag-
ment length polymorphisms (RFLPs), random amplified
polymorphic DNAs (RAPDs), amplified fragment length
polymorphisms (AFLPs), microsatellite, and single nucleo-
tide polymorphism (SNPs; Sunnucks 2000; Brumfield et al.
2003). Advantages of these methods over allozyme include
the increased likelihood of neutral variation and their in-
creased saturation and better coverage of the genome (Storfar
1996). Recent methods also detect higher levels of variation
and, in the case of mitochondrial DNA, have increased sen-
sitivity to bottlenecks. AFLPs have emerged as a recent mo-
lecular method that detects significant levels of DNA poly-
morphism at a very large number of loci (Vos et al. 1995).
AFLP analysis is not technically difficult, yet it generates
large numbers of markers spanning the entire genome without
requiring any prior sequence knowledge, libraryconstruction,
or the characterization of DNA probes. In addition, newly
developed semi-automated fluorescence-labeled detection of
AFLPs and software has improved both fragment scoring and
data handling. The dominant nature of the AFLP markers
may lead to the underestimation of recessive allele frequen-
cies (Szmidt et al. 1996), and it is still unclear whether AFLPs
can or should be analyzed in a genealogical frameworkwithin
species. These potential problems, however, may be over-
come by examining a large number of loci.
Although the AFLP technique is widely used in plant map-
ping and population genetic studies (Travis et al. 1996), its
application to animals, especially for population genetic and
evolutionary studies, is still relatively new (Otsen et al. 1996;
Ajmone-Marsan et al. 1997; Questian et al. 1999; Busch et
al. 2000). Several studies have recently used AFLPs to re-
construct interspecfic relationships (Albertson et al. 1999; De
Knijff et al. 2001; Giannasi et al. 2001; Parsons and Shaw
2001), but few studies have applied the technique to questions
below the species level.
In choosing the AFLP technique for our study, we were
motivated at least as much by its ability to generate large
numbers of markers across the genome as by its potential
utility in phylogeography. In fact, were our motives purely
phylogeographic, we would have chosen single nucleotide
polymorphisms over AFLPs as our preferred marker (Brum-
field et al. 2003). However, eastern populations of the house
finch have recently been invaded by a bacterium, Mycoplasma
gallisepticum, that escaped from chickens and caused a large
decline in house finch numbers (Hochachka et al. 2000). Our
long-term goal is to identify genetic loci that have contributed
to resistance to this parasite and may have undergone selec-
tion as a result of the epizootic. The AFLP method is well
suited for generating a large number of markers that, in prin-
ciple, could ultimately aid in the discovery of loci involved
in the evolution of resistance. In addition, to our knowledge,
this is the first large-scale geographic survey using the AFLP
approach of an avian species below the species level, al-
though some smaller avian studies have been published
(Busch et al. 2000).
MATERIALS AND METHODS
Genetic Resources
We examined 163 individuals sampled from 16 house finch
populations from North America and the Hawaiian Islands,
as well as samples of two closely related sister species, purple
finches (C. purpureus; n ? 3) and Cassin’s finches (C. cas-
sinii; n ? 6) in this study (Fig. 1). Detailed information on
location, sample sizes per population, and collection dates
are presented in Table 1. Importantly, all samples were col-
lected prior to the onset of the Mycoplasma epizootic in any
given locality, a sampling scheme that ensures that AFLP
frequencies are uninfluenced by this disease. Blood samples
were stored in Queen’s lysis buffer (Seutin et al. 1991) after
collection in the field. Tissue samples were immediately fro-
zen in liquid nitrogen after collection, stored long term at
?80?C, and transported between laboratories in 100% ethanol
at room temperature.
DNA Isolation
DNA extraction was based on the protocol of Kempenaers
et al. (1999) with minor modifications. About 0.1 g tissue,
or 150 ?l blood, were diluted in Queen’s lysis buffer (Seutin
et al. 1991) and added to 600 ?l digestion buffer (100 mM
Tris-HCl, pH 8.0, 10 mM EDTA, 100 mM NaCl, and 1%
sodium dodecyl sulfate) with proteinase K (final concentra-
tion 50 ?g/ml). Tissue digestion took place in a water bath
at 55?C for 3–4 h. Before extracting DNA, 5 ?l RNase A
(10 mg/ml) was added to each sample and incubated at 55?C
for another 30 min. DNA extraction was carried out twice
with an equal volume of phenol:chloroform:isoamyl alcohol
(25:24:1) and once with an equal volume of chloroform:isoa-
myl alcohol (24:1). The DNA was precipitated in ethanol by
using 1/10 volume of 3 M sodium acetate, spun down, and
dried and diluted with TE buffer (10 mM Tris-HCl, pH 8.0,

Page 3

2854
ZHENSHAN WANG ET AL.
FIG. 1.
Numbers correspond to populations in Table 1.
Geographic locations of house finch populations, as well as outgroups sampled in this study. Circles represent populations.
TABLE 1.Population sample information used in the present investigation.
Popula-
tion
Population
code
Species/
subspecies
Country/
StateLocality
Tissue
for DNA
Sampling
period
Sample
size
1
2
3
4
5
6
7
8
9
CA-Goleta
CA-Los Alamos
TX
AR
CO
WA
HI
Canada
MI
ME
NY
OH
MD
PA
AL
Mexico
purple finch
Cassin’s finch
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
frontalis
griscomi
purpureus
cassinii
California
California
Texas
Arizona
Colorado
Washington
Hawaii
Canada
Michigan
Maine
New York
Ohio
Maryland
Pennsylvania
Alabama
Mexico
Washington
Washington
Goleta
Los Alamos
San Angelo
Marana, Tucson
Boulder
Kittitas and King Co.
Oahu, Mauna Kea, Maui
St. Catharines
Ann Arbor
south of Gorham
Mattituck Long Island
Cleveland
Laurel
Quarryville
Auburn
Guerrero
Clark, Kittitas, Pacific Co.
Asotin, Yakima
tissue
tissue
tissue
tissue
tissue
tissue
tissue
tissue
blood
tissue
tissue
tissue
tissue
tissue
blood
tissue
tissue
tissue
March 1984
April 1989
September 1991
April 1989
April 1989
1996–2000
July 1991
January 1984
February 1991
April 1991
May 1990
March 1990
May 1990
May 1990
1995
1990
1995–1998
1994–1995
10
10
10
10
10
6
17
10
10
10
10
10
10
10
10
10
3
6
10
11
12
13
14
15
16
17
18
1 mM EDTA). The diluted DNA samples were run on elec-
trophorestic gels to determine DNA quality and quantity.
Amplified Fragment Length Polymorphism Procedure and
Primer Screening
Our AFLP procedure followed the AFLP plant mapping
kit protocol (Perkin Elmer, Foster City, CA) with modifi-
cations. Restriction digestion and adaptor ligation were per-
formed simultaneously on 200 ng of genomic DNA using 7.5
units EcoRI (New England Biolabs, Beverly, MA), 1.5 units
MseI (New England Biolabs), and 67 Weiss units of T4 DNA
ligase (New England Biolabs). The restriction ligation re-
action was performed in 11-?l volume at room temperature
overnight. The restriction ligation products were subsequent-
ly diluted to 100 ?l using TE0.1(20 mM Tris-HCl, 0.1 mM
EDTA, pH 8.0). Polymerase chain reaction (PCR) amplifi-
cations were carried out in two steps as recommended by
Vos et al. (1995). Preselective amplifications were performed
by using 2 ?l of the diluted, restricted, and ligated DNA, 0.5
?l EcoRI and MseI preselective primers, and 7.5 ?l of AFLP
core mix supplied in the kit in a final volume of 10 ?l. The
preselective primers consisted of the adaptor primer sequence
with a single nucleotide extension at the 3? end. This pre-
selective PCR was performed with the following temperature
profile: 2 min at 72?C followed by 20 cycles of 20 sec at
94?C, 30 sec at 56?C, and 2 min at 72?C, then a holding step
at 60?C for 30 min. A 5-?l aliquot of each preselective am-
plification product was checked for quality on a 1.5% agarose
gel. For selective amplification, 3 ?l of the preselective am-
plified diluted DNA were added to 15 ?l of AFLP core mix,

Page 4

2855
AFLP ANALYSIS OF THE HOUSE FINCH
1 ?l (5 pmol) of selective MseI primer, and 1 ?l (1 pmol)
of the EcoRI selective primer labeled with fluorescent dye
(FAM, blue; or NED, yellow). Both the EcoRI and MseI
primers used in the selective amplification have three extra
nucleotides at the 3? end to reduce the number of the amplified
fragments. The cycle profile for the selective amplification
began with a 2-min denaturation at 94?C, then followed by
10 cycles with 20-sec denaturation at 94?C, 30-sec annealing
with the temperature decreased each cycle by 1?C from 66?C
to 57?C, and elongation 2 min at 72?C. The PCR was con-
tinued for 20 cycles with 20 sec at 94?C, 30 sec at 56?C, and
2 min at 72?C, followed by a holding step at 60?C for 30
min.
After checking the PCR products (10 ?l) on 1.5% agarose
gels, 0.4 ?l of blue-labeled and/or 0.75 ?l of yellow-labeled
amplification products were added to 1.05 ?l of loading buff-
er containing 0.67 ?l of deionized formamide, 0.13 ?l of
blue dextran, and 0.25 ?l of GeneScan-500 (Perkin Elmer)
ROX (red) labeled size standard. The samples were denatured
for 3 min at 90?C, then run on an ABI Prism 377 DNA
sequencer (Perkin Elmer) in a 5% Long Ranger (Combrex
Bioscience, Rockland, ME) gel for 6 h with a well-to-read
distance of 48 cm. The digital gel data was collected by ABI
Prism GeneScan analysis software (ver. 3.1.2). Each lane file
was analyzed for the presence and absence of AFLP products
at approximately 1-bp intervals using Genographer software
(Benham et al. 1999). With this labeling system, small frag-
ments typically have a stronger fluorescence signal than larg-
er fragments, and fluorescence signal decreases with increas-
ing fragment size. We thus analyzed different parts of the
gel with different intensity indices to visualize the fragments
with maximum clarity. Only unambiguously detectable frag-
ments were scored.
To determine the number of fragments generated by dif-
ferent primer pairs, 16 primer combinations (E-AAC, AAG,
ACC, ACT/M-CAC, CAG, CTA, CTG, where E is EcoRI and
M is MseI) were used in a preliminary screen of two indi-
viduals. Ten individuals from different populations were em-
ployed to confirm the preliminary screening results. Based
on these results, three primer combinations (E-AAC/M-CTG,
E-ACC/M-CAG, E-ACT/M-CTA) were chosen for all indi-
viduals. These primer combinations were selected because
they produced a manageable number of appropriately sized,
polymorphic, and well-separated markers.
Data Analysis
We considered each fragment position as a dominant locus
with two states: presence or absence. Amplification products
were scored as discrete, binary state (present/absent) for each
individual and labeled by primer combination and estimated
band size. A data matrix (individual x marker) containing the
band scoring information was transformed to allele frequen-
cies under the assumption that each amplified band corre-
sponds to a different AFLP locus.
Because AFLP markers must be analyzed as dominant loci,
the Hardy-Weinberg equilibrium assumption was made to
estimate population genetic diversity and genetic structure
parameters. Nei’s (1973) gene diversity was obtained by the
POPGENE version 1.31 program (Yeh et al. 1997). Gene
diversity was also obtained by a Bayesian method (Holsinger
et al. 2002), which does not assume that genotypes within
populations are in Hardy-Weinberg proportions. The per-
centage of polymorphic loci was obtained by Tools for Pop-
ulation Genetic Analysis (TFPGA) software version 1.3
(Miller 1997a). The proportion of shared bands (F) was cal-
culated using the RAPDPLOT program (Black 1996) based
on Dice’s (1945) similarity coefficient: Sxy? 2Nxy/(Nx? Ny),
where Nxyis the number of fragments in common between x
and y individuals, and Nx and Ny are the total number of
fragments in x and y individuals, respectively (see also Nei
and Li 1979). Nucleotide diversity (?) based on F, in prin-
ciple directly comparable to sequence data, was estimated by
the method of Innan et al. (1999). This method of estimating
nucleotide diversity is based on the fact that each AFLP
product represents a 16-bp sequence assay when using the
EcoRI and MseI restriction enzymes, which have 6-bp and 4-
bp recognition sequences, respectively, and three selective
nucleotides on each of the two AFLP selective amplification
primers. Therefore, each shared AFLP product indicates zero
nucleotide differences over 16 bp, whereas polymorphisms
reflect at least one nucleotide difference over 16 bp. The
actual number of differences that contribute to each poly-
morphism is a function of F, which can be used to determine
the overall number of nucleotide substitutions per site. The
standard deviation of ? was calculated by the jackknife meth-
od (Efron 1982) following Nei and Miller’s (1990) approach.
Hierarchical structuring of genetic variation and pairwise
?STdistances (analogous to FST-statistics at the molecular
level; Excoffier et al. 1992) among populations were mea-
sured using analysis of molecular variance (AMOVA) using
WINAMOVA version 1.55 (Excoffier et al. 1992; Stewart
and Excoffier 1996). Significance levels of the variance com-
ponents were based on 1000 permutations. A pairwise Eu-
clidean distance matrix and all input files needed for the
AMOVA analysis were produced using the AMOVA-PREP
program version 1.1 (Miller 1997b), which specifically pre-
pares dominant marker data for Excoffier et al.’s WINA-
MOVA. The parameter ?B, which is analogous to FST, was
also calculated by the Bayesian approach (Holsinger et al.
2002), which incorporates uncertainty about the magnitude
of within-population inbreeding.
Gene flow between pairs of populations based on the equa-
tion Nm ? 0.25(1/FST? 1) was calculated from ?ST-values
under the assumption of an infinite-island model of popu-
lation structure (Wright 1951). A matrix Mantel correspon-
dence test (Mantel 1967) between genetic distances and geo-
graphical distances among populations was carried out using
the TFPGA program. The geographical distances among pop-
ulations were obtained by using Distance Finder (available
via http://www.indo.com/distance).
A neighbor-joining tree for individuals were constructed
from the mean character distance with PAUP* version 4.0b8
(Swofford 2001). Bootstrap support was evaluated with 1000
replicates. To obtain a tree in which populations are taxo-
nomic units, gene frequency datasets was obtained by POP-
GENE version 1.31 program (Yeh et al. 1997). Then a boot-
strapped neighbor-joining tree was constructed by various
modules in the PHYLIP package (Felsenstein 1985) and vi-
sualized using TreeView version 1.6.1 (Page 1996).

Page 5

2856
ZHENSHAN WANG ET AL.
TABLE 2.Amplified fragment length polymorphism primer pairs used and their amplification results.
Primer
pair1
Size of
fragments
scored (bp)
Among Carpodacus
Total
bands
Polymorphic
bands
Polymorphic
bands (%)
Within house finch
Total
bands
Polymorphic
bands
Polymorphic
bands (%)
Within frontalis subspecies
Total
bands
Polymorphic
bands
Polymorphic
bands (%)
E-AAC/M-CTG
E-ACC/M-CAG
E-ACT/M-CTA
80–499
79–489
80–485
130
131
102
84
100
74
64.6
76.3
72.6
92
99
78
57
64
45
62.0
64.7
57.7
90
94
76
56
58
43
62.2
61.7
56.6
1E, EcoRI; M, MseI.
Treating either populations or individuals at taxonomic
units with AFLP data obsures details of gene history through
pedigrees at several hierarchical levels, because trees at these
levels are summaries of many independent gene genealogies.
Therefore, a model-based clustering method for using mul-
tilocus genotype data to infer population structure and assign
individuals to populations was implemented by the STRUC-
TURE program (Pritchard et al. 2000). This approach pro-
vides a coherent Bayesian framework for incorporating the
inherent uncertainty of parameter estimates into the inference
procedure and for evaluating the strength of evidence for the
inferred clustering. We chose to use a burn-in period of
30,000 iterations and collect data for 106iterations under a
no-admixture model without using prior population origin
information for running the program. For each dataset, we
ran three independent simulations of this length. Highly con-
sistent results were produced between independent runs.
RESULTS
Amplified Fragment Length Polymorphism Patterns
and Polymorphism
The three primer combinations generate a total of 363
bands that range in size from 79 to 499 bp, of which 258
(71.2%) are polymorphic across all 172 individuals. A total
of 166 (61.3%) bands are polymorphic across the 163 house
finch individuals, and 157 (60.2%) bands are polymorphic
across the 153 individuals representing the 15 populations
from the eastern and western United States and Hawaiian
Islands (C. m. frontalis subspecies; Table 2). Within house
finches, the primer pair E-AAC/M-CTG generates the largest
number of polymorphic bands and primer pair E-ACT/M-
CTA generates the smallest number of polymorphic bands
(Table 2). An example of an AFLP pattern reproduced by
Genographer software with primer pair E-ACC/M-CAG is
shown in Figure 2; visual inspection of this reproduced gel
clearly shows the distinct AFLP patterns found between the
three species analyzed and between Mexican and U.S. house
finch populations. We find 12, 11, and 41 bands specific to
(albeit not fixed within) C. cassini, C. purpureus and C. mex-
icanus, respectively, and three bands specific to but not fixed
in the Mexican populations of house finches. However, no
population-specific bands are detected among populations
within the C. m. frontalis subspecies. The complete data ma-
trix is available on the Web at http://depts.washington.edu/
scotte/AFLPdata.
Population Genetic Diversity
Population genetic diversity descriptive parameters are
summarized in Table 3. Within the frontalis subspecies, av-
erage heterozygosity ranges from 0.08 to 0.11, with an av-
erage of 0.10. Similar but slightly higher values are obtained
with the Bayesian approach. The percentage of polymorphic
loci ranges from 22.3 to 32.5, with an average of 29.9. Over-
all, average heterozygosity is relatively low and comparable
to the levels for allozyme study (Vasquez-Phillips 1992).
Estimated nucleotide diversity ranges from 0.0057 to 0.0085,
with an average of 0.0075, implying that a random pair of
house finches differs at approximately seven nucleotides per
1000 in the nuclear genome, a level that is relatively high
compared to direct estimates in humans and other species
from single nucleotide polymorphisms (Brumfield et al.
2003).
There is very little difference in levels of diversity between
the western and eastern regions and the Hawaiian Islands.
For all the descriptive parameters, the population from Can-
ada shows the highest average heterozygosity and percent
polymorphic loci, and the Washington population exhibits
the lowest values for all of the descriptive statistics, a result
that may be a consequence of the small sample size for this
population.
Population Structure
We partitioned the molecular variances into species, sub-
species, regions, population, and individual levels. AMOVA
results recover deep divisions between the griscomi and fron-
talis subspecies within the house finch (Table 4). Of the total
molecular variance within the house finch, 21% is attributable
to divergence between subspecies and 71% is found among
individuals within populations. However, when only the fron-
talis subspecies is considered, the vast majority (87.3%) of
the total variance is found among individuals within popu-
lations. Similar results are obtained when analyses are carried
out within the continental United States or within the eastern
or western regions (data not shown).
Phylogenetic Analysis
Neighbor-joining trees are constructed with populations as
units (Fig. 3). The phenograms reveal deep divisions between
C. mexicanus and its congeners, and between C. griscomi in
Mexico and C. m. frontalis subspecies within house finches.
One cluster in the tree consisting solely of easternpopulations
is most closely related to the Goleta, California, population.
Contrary to expectation from the known history, Hawaiian
populations are not clearly derived from western populations,
but instead fall outside all other frontalis populations. A sim-
ilar pattern is found in trees constructed with PAUP programs
with individuals as units, although bootstrap support is very
low (not shown). Though individuals from each population