Sequential divergence and the multiplicative origin of
community diversity
Glen R. Hooda,1, Andrew A. Forbesb, Thomas H. Q. Powella,c, Scott P. Egana,d,e, Gabriela Hamerlinckb, James J. Smithf,
and Jeffrey L. Federa,d
a
Department of Biological Sciences, University of Notre Dame, Notre Dame, IN 46556; bDepartment of Biology, University of Iowa, Iowa City, IA 52242;
c
Department of Entomology and Nematology, University of Florida, Gainesville, FL 32611; dEnvironmental Change Initiative and Advanced Diagnostics and
Therapeutics, University of Notre Dame, Notre Dame, IN 46556; eDepartment of Biosciences, Anderson Biological Laboratories, Rice University, Houston,
TX 77005; and fDepartment of Entomology and Lyman Briggs College, Michigan State University, East Lansing, MI 48824
Edited by Douglas Futuyma, State University of New York, Stony Brook, NY, and approved September 18, 2015 (received for review December 24, 2014)
Phenotypic and genetic variation in one species can influence the
composition of interacting organisms within communities and across
ecosystems. As a result, the divergence of one species may not be an
isolated process, as the origin of one taxon could create new niche
opportunities for other species to exploit, leading to the genesis of
many new taxa in a process termed “sequential divergence.” Here,
we test for such a multiplicative effect of sequential divergence in a
community of host-specific parasitoid wasps, Diachasma alloeum,
Utetes canaliculatus, and Diachasmimorpha mellea (Hymenoptera:
Braconidae), that attack Rhagoletis pomonella fruit flies (Diptera:
Tephritidae). Flies in the R. pomonella species complex radiated by
sympatrically shifting and ecologically adapting to new host plants,
the most recent example being the apple-infesting host race of
R. pomonella formed via a host plant shift from hawthorn-infesting
flies within the last 160 y. Using population genetics, field-based
behavioral observations, host fruit odor discrimination assays, and
analyses of life history timing, we show that the same host-related
ecological selection pressures that differentially adapt and reproduc-
tively isolate Rhagoletis to their respective host plants (host-associated
differences in the timing of adult eclosion, host fruit odor preference
and avoidance behaviors, and mating site fidelity) cascade through the
ecosystem and induce host-associated genetic divergence for each of
the three members of the parasitoid community. Thus, divergent se-
lection at lower trophic levels can potentially multiplicatively and rap-
idly amplify biodiversity at higher levels on an ecological time scale,
which may sequentially contribute to the rich diversity of life.
host plant adaptation | parasitoid | Rhagoletis | tritrophic interactions |
ecological speciation
P opulation divergence is a fundamental evolutionary process
contributing to the diversity of life (1). Studies of how new life
forms originate typically focus on how barriers to gene flow evolve
in specific lineages, resulting in their divergence into descendent
daughter taxa. As a result, evolutionary biologists now have a good
understanding of how variation within a population is trans-
formed by selection into differences between taxa (1–3). What is
less well understood is whether the divergence of one population
has consequences that ripple through the trophic levels of an
ecosystem and affect entire communities of interacting organ-
isms. Studies in paleontology (4–6), community ecology (7, 8),
systematics (8, 9), and ecosystem genetics (10, 11) suggest that
evolutionary change in one lineage can influence entire com-
munities of organisms. For example, when the genotype/pheno-
type of a “foundation” species influences the relative fitness of
other species, evolutionary change(s) in this genotype/phenotype
may affect organisms in adjacent trophic levels (10, 11). If these
evolutionary changes are linked to ecological adaptation and re-
productive isolation (RI), associated organisms may diverge in par-
allel, potentially creating entire coevolved communities distinct from
one another (12–15). Therefore, population divergence may not
always be an isolated process, as the differentiation of one taxon
could beget the divergence of many others.
E5980–E5989 | PNAS | Published online October 23, 2015
Such “sequential” or “cascading” divergence events may be
particularly relevant to understanding why some groups of or-
ganisms, like plants, the insects that feed on them, and the para-
sitoids that attack the insects, are more diverse and species-rich than
other groups (8, 9, 12–15). Specifically, when phytophagous insects
diversify by adapting to new host plants, they create a new habitat
for their parasitoids to exploit (Fig. 1). If a parasitoid shifts to the
new habitat, it can encounter the same divergent ecological selec-
tion pressures as its insect host, which could result in the parallel
divergence of insect host and parasitoid (12–15) (Fig. 1A). More-
over, sequential divergence may have multiplicative effects in gen-
erating biodiversity, as the shift of an insect to a new plant may open
a new niche opportunity for not just one but the entire community
of parasitoids attacking the insect host (12, 13) (Fig. 1B). However,
few convincing examples of sequential divergence exist (12–15), and
in no study is there both genetic and ecological evidence for
sequential divergence multiplicatively amplifying biodiversity.
Here, we test for the multiplicative effects of sequential di-
vergence in the community of parasitoid wasps (Hymenoptera:
Braconidae) that attack fruit flies in the Rhagoletis pomonella
sibling species complex (Diptera: Tephritidae) (Fig. 2). Several
features of the biology and biogeography of the Rhagoletis−para-
sitoid system make it ideal for investigating the multiplicative
divergence hypothesis and allow us to directly test multiple criteria
supporting sympatric host race formation (16) and sequential di-
vergence (12, 13), summarized in Table 1. First, concerning the
spatial context of divergence (criterion 1), R. pomonella complex
Significance
Understanding how new life forms originate is a central question
in biology. Population divergence is usually studied with respect
to how single lineages diverge into daughter taxa. However,
populations may not always differentiate in isolation; divergence
of one taxon could create new niche opportunities in higher tro-
phic levels, leading to the sequential origin of many new taxa.
Here, we show that this may be occurring for three species of
parasitoid wasps attacking Rhagoletis fruit flies. As flies shift and
adapt to new host plants, wasps follow suit and diverge in kind,
resulting in a multiplicative increase of diversity as the effects of
ecologically based divergent selection cascade through the eco-
system. Biodiversity therefore may potentially beget increasing
levels of biodiversity.
Author contributions: G.R.H., A.A.F., T.H.Q.P., S.P.E., and J.L.F. designed research; G.R.H.,
A.A.F., T.H.Q.P., S.P.E., G.H., J.J.S., and J.L.F. performed research; G.R.H., A.A.F., T.H.Q.P., and
J.L.F. contributed new reagents/analytic tools; G.R.H., A.A.F., S.P.E., G.H., and J.L.F. analyzed
data; and G.R.H., A.A.F., T.H.Q.P., S.P.E., G.H., J.J.S., and J.L.F. wrote the paper.
The authors declare no conflict of interest.
This article is a PNAS Direct Submission.
Data deposition: The mtDNA sequences reported in this paper have been deposited in the
GenBank database (accession nos. KT761291–KT761497). Raw data are available from the
Dryad Digital Repository, dx.doi.org/10.5061/dryad.5n72m.
1
To whom correspondence should be addressed. Email: ghood@nd.edu.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
1073/pnas.1424717112/-/DCSupplemental.
www.pnas.org/cgi/doi/10.1073/pnas.1424717112

Multiplicative
Sequential Divergence Sequential Divergence
A B C
Allopatry
Time
Secondary contact
flies have formed in the absence of geographic isolation. In par-
ticular, the recent sympatric shift of R. pomonella from its ancestral
host plant hawthorn (Crataegus spp.) to introduced domesticated
apple (Malus domestica) in the last 160 y is an example of host race
formation in action (1–3, 17–19), the hypothesized initial stage of
ecological speciation (16, 17). The short time frame and sympatric
spatial context of R. pomonella’s shift to apple exclude passive
codivergence or speciation (Fig. 1C) as an explanation for differ-
entiation. Specifically, fly and wasp populations could not have
diverged in concert, because they became jointly geographically
separated in the past (Fig. 1C). Rather, if flies and wasps display
concordant adaptations, it is likely due to the direct effects of di-
vergent ecological selection resulting from host shifts cascading
from host plants to flies to parasitoids (Fig. 1 A and B).
Second, the apple and hawthorn host races of R. pomonella
belong to a closely related group of sibling species, including
Rhagoletis mendax (host: blueberry, Vaccinium spp.), Rhagoletis
zephyria (host: snowberry, Symphoricarpos spp.), and the unde-
scribed flowering dogwood fly (host: Cornus florida). All of these
taxa purportedly radiated via sympatric host shifts (17–23). In
addition, other species in the genus, such as the eastern cherry
fly, Rhagoletis cingulata (host: black cherry, Prunus serotina) are
sympatric with R. pomonella group flies (23). Thus, the potential
for sequential divergence in the Rhagoletis parasitoid community
extends beyond the host races, with multiple cooccurring fly re-
sources existing for wasps to attack, satisfying criterion 1.
Third, Rhagoletis in the eastern United States are attacked by a
community of host-specific endoparasitoid wasps that include
the species Diachasma alloeum, Utetes canaliculatus, and Diac-
hasmimorpha mellea (24, 25) (Fig. 2). All three species have a
free-living, sexually reproducing adult life stage. This life cycle
eliminates vertical transmission as a factor facilitating codi-
vergence. U. canaliculatus oviposits into Rhagoletis eggs laid be-
neath the skin of ripe fruit, while D. alloeum and D. mellea
oviposit into late instar larvae feeding within fruit (24). A degree
of niche partitioning for oviposition sites therefore exists among
species (25), potentially facilitating coexistence on the same fly
host. As a result, multiple host associations of wasps exist in close
geographic proximity, fulfilling the requirements of criterion 1
for sympatric race formation and sequential divergence for the
wasp community as well.
In addition, a previous study documented that one parasitoid
attacking Rhagoletis, D. alloeum, is undergoing sequential di-
Hood et al.
PNAS PLUS
Co-speciation
in Allopatry
Fig. 1. Three scenarios of codivergence in a host−
parasitoid system. (A) A single sequential divergence
event, (B) sequential divergence with multiplicative
amplification of biodiversity, and (C) cospeciation in
allopatry. In A, codivergence is driven by the cas-
cade of divergent ecological selection pressures
across trophic levels in sympatry. Here, a degree of
divergent ecological adaptation must accompany
the host shift such that parasitoids are not merely
moving between geographically separated hosts. In
B, the multiplicative effects of sequential divergence
can be seen as several members of the parasitoid
community diverge in parallel with their host. In C,
codivergence (cospeciation) occurs after host plant,
fly, and parasitoid populations become jointly geo-
graphically isolated (black bar), resulting in parallel
allopatric speciation. Here, little differentiation need
accompany the initial host shift of fly or parasitoid.
Cospeciation is not necessarily driven by the creation
and adaptation to new niches but by the concordant
geographic and reproductive separation of hosts and
EVOLUTION
parasitoids.
vergence (14). Population genetic surveys, field observations,
behavioral assays of host choice, and studies of life history timing
support the existence of an ecologically derived population of
D. alloeum attacking the recently formed apple-infesting host
race of R. pomonella, meeting criteria 2, 3, 4, 6, and 7. We hy-
pothesize that if U. canaliculatus and D. mellea are undergoing
sequential divergence, they will show similar patterns of host-
associated ecological and genetic divergence.
Two dimensions of divergent ecological selection generate RI
among host-associated populations of Rhagoletis and D. alloeum:
host-specific mating (habitat isolation) and differences in eclo-
sion phenology (temporal isolation). With respect to habitat
isolation, Rhagoletis (26) and D. alloeum (14) court and mate on
or near the fruit of their respective host plants. The most im-
portant long- to intermediate-range cues that flies use to find and
discriminate among plants are the volatile compounds emitted
from the surface of ripening fruit (27–30). Flies display genetically
based behavioral preference for natal fruit surface volatiles and
avoid the volatiles of alternative fruit (29). Similarly, D. alloeum
prefer natal and avoid nonnatal host fruit volatiles in behavioral
assays (14), supporting criterion 3. Consequently, differences in
host choice translate directly to mate choice, generating prezygotic
habitat-related RI for both flies and wasps, fulfilling criterion 4.
Additionally, host odor discrimination may also act as a postzygotic
barrier to gene flow in R. pomonella, as they suffer behavioral host
choice sterility mediated by a reduced chemosensory ability to find
suitable host fruit for mating and oviposition (30). Whether or not
hybrid D. alloeum display a similar behavior is unknown. Lastly,
criteria 5 is partially met for Rhagoletis, as the host fruit envi-
ronment has no effect on host odor discrimination behaviors for
hawthorn-origin R. pomonella reared in apple fruit, indicating
that host selection and fidelity are under (partial) genetic control
A B C D E
Fig. 2. The community of host-specific parasitoids that attack members
of the (A) Rhagoletis pomonella sibling species complex: (B) D. alloeum,
(C) D. mellea, and (D) U. canaliculatus that (E) emerge from the fly pupal
case as adults following overwintering. (Scale bar, 1 mm.)
PNAS | Published online October 23, 2015 | E5981
Table 1. Summary of conditions (criteria) conducive to and supporting hypotheses of sympatric host race formation (modified from
ref. 16) and sequential divergence (modified from refs. 12 and 13)
Criteria supporting hypotheses of sympatric host race
formation and sequential divergence Rp Da Dm Uc

Criterion 1. Shift to new host resource and multiple host-associations occur in yes yes yes yes
sympatry or close geographic proximity
Criterion 2. Host-associated populations form distinct genetic clusters (spatially yes yes yes yes
replicable), but experience gene flow at appreciable rates
Criterion 3. Females, but also potentially males, display host preferences and yes yes yes yes
discriminate among alternate hosts
Criterion 4. Host choice is linked to mate choice facilitating assortative mating yes yes yes yes
and resulting in prezygotic habitat isolation
Criterion 5. Host selection and fidelity are under some degree of genetic yes yes, tested not not
control and not due solely to maternal, learning, or environmental effects in one direction tested tested
Criterion 6. Differences in insect phenologies tracks differences in the host yes yes yes yes
phenologies resulting in temporal (allochronic) isolation
Criterion 7. Insect phenology under some degree of genetic control and not yes yes yes yes
due solely to maternal or environmental effects
Criterion 8. Fitness tradeoffs exist between host-associated populations yes, sometimes not tested not tested not tested
resulting in migrants and hybrids having reduced fitness

Also shown is whether these criteria have been empirically tested and confirmed in R. pomonella (Rp) species complex flies and three members of the
parasitoid wasp community attacking the flies: D. alloeum (Da), D. mellea (Dm), and U. canaliculatus (Uc). Data for Da are from Forbes et al. (14) and the
current study (criterion 5).

(27, 30). Similar experiments have not yet been conducted in
D. alloeum (but see Reciprocal Rearing of Diachasma).
With respect to temporal isolation (criterion 6), the timing of
overwinter diapause is an important host-related ecological ad-
aptation for Rhagoletis. The host plants of Rhagoletis fruit at
different times of the year (19, 31, 32). For example, apple va-
rieties favored by R. pomonella ripen 3–4 wk before native
hawthorns in sympatry. Thus, flies must eclose to coincide with
the availability of ripe fruit to find mates and oviposition sites.
Rhagoletis are univoltine, and their lifespan is short (1 mo).
Differences in eclosion timing between races therefore results in
partial allochronic mating isolation (19, 29–32). The differences
in eclosion timing also confer a degree of postzygotic isolation
because hybrids will possess eclosion patterns asynchronous with
fruit ripening (29, 31, 32). Rhagoletis attacking blueberries and
flowering dogwoods display similar differences in eclosion time
related to variation in host fruiting phenology (31).
The life cycle of D. alloeum mirrors that of Rhagoletis, gen-
erating the same divergent ecological selection pressures. As a
result, populations of D. alloeum attacking different Rhagoletis
eclose to match the phenology of fly larvae feeding within host
fruit (14). In addition, longevity of D. alloeum (∼2 wk) is half that
of Rhagoletis, generating even more pronounced allochronic
mating isolation compared to the fly (14), supporting criterion 6.
Significant allele frequency differences between sympatric pop-
ulations of D. alloeum attacking different fly hosts (criterion 2)
were associated with differences in eclosion time (14), confirm-
ing criterion 7. The same has also been found for Rhagoletis
(19, 31, 32), connecting host-related life history adaptation and
RI to patterns of genetic differentiation among flies and wasps.
Here, we test for the multiplicative hypothesis of sequential
divergence in the Rhagoletis−parasitoid system using the criteria
in Table 1 to frame our experimental approach. For populations
of U. canaliculatus and D. mellea attacking different Rhagoletis,
we first assessed mtDNA sequence variation and conducted
population genetics surveys using nuclear-encoded microsatellites
to test for host-related genetic differentiation (criterion 2). Second,
to test for host plant-related assortative mating caused by habitat
isolation, we coupled field observations of mating behavior of
U. canalicualtus and D. mellea with tests of host odor discrimination,
including preference for and avoidance of fruit surface volatile
compounds (criteria 3 and 4). To determine if genetic effects
contributed to host odor discrimination (criterion 5), we com-
pared the behavioral responses of D. alloeum reared through
nonnatal fly and host plant environments to parental and nonnatal
host fruit volatiles. Third, to assess the degree of allochronic iso-
lation due to variation in host phenology, we compared the timing
of adult eclosion of U. canaliculatus and D. mellea attacking
different fly populations in sympatry (criterion 6). Lastly, to link
host-associated genetic differentiation to divergence in life history
timing (criterion 7), we tested for associations between micro-
satellite genotypes and the timing of eclosion for U. canalicualtus
and D. mellea. Although host-associated fitness tradeoffs (criterion
8) have been inferred for several species of Rhagoletis feeding in
natal versus nonnatal fruit (33, 34), difficulty in reciprocally
transplanting wasps precludes these experiments at this time but
remain an area for future study.
Results and Discussion
mtDNA Divergence. To make a case for the sequential divergence
of U. canaliculatus and D. mellea, it must be shown that pop-
ulations attacking the derived apple fly do not represent un-
recognized sibling species associated with different Rhagoletis. In
this case, differences could be erroneously attributed to ecolog-
ical adaptation following a recent sympatric host shift when, in
fact, they represent older speciation events that potentially occurred
in allopatry. To test for the existence of cryptic sibling species, we
sequenced a 540-bp region of the mtDNA COI (cytochrome
oxidase subunit I) gene for specimens morphologically classified
as U. canaliculatus and D. mellea attacking apple, hawthorn,
blueberry, flowering dogwood, snowberry, and black cherry flies.
The presence of sibling species would be indicated by diverged
mtDNA haplotypes that, upon analysis using nuclear-encoded
microsatellites, would identify as genetically distinct populations,
implying an absence of gene flow.
Utetes. A maximum parsimony gene tree of Utetes populations
attacking apple, hawthorn, snowberry, and flowering dogwood
flies indicated three distinct mtDNA haplotype groups distin-
guished by ≥19 nucleotide substitutions (Fig. 3A). Haplotype A
was restricted to wasps attacking the flowering dogwood fly, and
haplotype B was found only in wasps attacking flowering dog-
wood and a single population of snowberry flies from State
College, PA (site 9). Haplotype C was most common, found in
U. canaliculatus attacking flowering dogwood, apple, hawthorn,
and a single population of snowberry flies from East Lansing, MI
(site 2). Thus, the possibility exists for three different cryptic
Utetes sibling species, with haplotype C being the most informative

E5982 | www.pnas.org/cgi/doi/10.1073/pnas.1424717112 Hood et al.


for testing the sequential divergence hypothesis, attacking pop-
ulations of all four Rhagoletis hosts parasitized by U. canaliculatus.
Diachasmimorpha. In contrast to U. canaliculatus, there was no evi-
dence for a distinct mtDNA matrilineage among D. mellea attacking
apple, hawthorn, blueberry, or black cherry flies (Fig. 3B). Similar to
previous results from D. alloeum (14), several haplotypes were
found in D. mellea displaying low levels of sequence divergence
(≤1%). Haplotypes did not differentiate wasps in a host-specific
manner and were generally shared among all populations. The
pattern implies that host-associated populations of D. mellea are
relatively recently derived, including those attacking apple flies,
and rules out the presence of long-established cryptic species.
Microsatellite Differentiation.
Utetes. The pattern of microsatellite differentiation for 20 loci
suggested the existence of three cryptic species of Utetes likely
having diverged before the introduction of apples to North
America ∼400 y ago and the shift of R. pomonella to apples within
the last 160 y. The three genetic clusters observed in the micro-
satellite genetic distance network (Fig. S2) corresponded exactly
to the three major mtDNA haplotypes (Fig. 3A). Populations
defined by mtDNA haplotypes A, B, and C each possessed alleles
displaying high-frequency differences or private variants at many
loci. For example, allele 173 at locus UC57 ranged in frequency
from 0.45 to 0.76 among haplotype C wasp populations, was absent
PNAS PLUS
Within haplotype C, significant microsatellite differentiation
was observed among populations of U. canaliculatus attacking
snowberry, flowering dogwood, hawthorn, and apple flies. No
fixed or high-frequency private allele was found for any of the 20
microsatellite loci that distinguished wasps attacking different fly
hosts. Instead, populations shared major alleles in common that
differed in frequency (see data in the Dryad database), implying
that differentiation among populations is likely of relatively re-
cent origin with ongoing (or recently ceased) gene flow. Signif-
icant allele frequency differences were observed in each of the
six sympatric pairwise comparisons between wasps attacking
different fly hosts (Table S2). Seven loci (UC8, UC10, UC12,
UC16, UC22, UC47, UC54) displayed pronounced and consis-
tent differences (of similar magnitude and in the same direction)
across at least three pairwise sympatric comparisons (Table S2).
Similarly, generalized linear model (GLM) analyses identified
seven loci (UC8, UC12, UC14, UC32, UC47, UC52, UC53)
showing a significant host effect across the five sympatric paired
sites of wasps attacking apple and hawthorn flies (Table S3).
There was little evidence for a major effect of geography in the
GLM analysis, as only two loci (UC14, UC61) had a significant
latitude and/or host × latitude effect (Table S3). This result
contrasts with previous findings for R. pomonella (18, 32) and
D. alloeum (14) where latitudinal variation was pronounced. Thus,
the direction and magnitude of host-related allele frequency

in haplotype B, and was present at low frequency (∼0.06) in
haplotype A (see data in the Dryad database). Thus, all sub-
sequent analyses of U. canaliculatus focused on the widespread
mtDNA haplotype C populations to test for recent sequential
divergence.
EVOLUTION
differences were more consistent across sites for U. canal-
iculatus than for R. pomonella or D. alloeum. As a result, a
neighbor-joining (NJ) genetic distance network based on all 20
microsatellites distinguished populations attacking apple, haw-
thorn, flowering dogwood, and snowberry flies (Fig. 4A). Apple

A Utetes canaliculatus B Diachasmimorpha mellea

3(2), 6
1 substitution 1 1 substitution
1(3), 4(3), 6
1(2), 2(2), 5(2) 6
4(2)
2, 5 3
6
1(2), 2(2) 2
6
1 6
5(93) 1
6 2(2), 3(4), 6(3)
3
C
2, 3(2), 6
1, 2, 3(2), 6(2)
4(2), 6(3) 3(3)
3(4), 6, 8(3)
2(7) 6(2)
16(100)
2 6
5
2 3
1, 3, 5(2), 6, 7(3), 8
1(2)
1
7(2) 2(3)
6(2)
2(2)
9
4(3), 6(17) 6(2) Fig. 3. Maximum parsimony mtDNA COI gene trees
9 for (A) U. canaliculatus and (B) D. mellea attacking
4 6
hawthorn (red), apple (green), blueberry (blue), flow-
6 B 3 ering dogwood (yellow), snowberry (white), and black
4
9 3 cherry (black) flies. The three major haplotypes
9 1(5), 2(3), 3(3), 6(6) identified for U. canaliculatus are designated A, B,
14(100) 9(3) and C. Each tree is rooted with the outgroup taxon,
9 1(3), 2(3), 3(2), 6(6)
D. alloeum. Nonparenthetic numbers next to each
6(2)
6(3) circle indicate collecting site (see Table S1 and Fig. S1
4 for site designations). Numbers in parentheses in-
25(100) 6 4(3), 6(14)
3(2) dicate the number of identical haplotypes sequenced
6
6 A 3(3)
at each collecting site. The number of nucleotide sub-
stitutions (evolutionary steps) and bootstrap support
6(3) based on 10,000 replicates (in parentheses) are given
49(100) Diachasma alloeum 83(100)
Diachasma alloeum above longer branches.

Hood et al. PNAS | Published online October 23, 2015 | E5983

and hawthorn fly-attacking wasps were positioned adjacently in
the network, consistent with the apple fly-attacking populations
having originated via a recent shift from hawthorn flies. Al-
though a shift of U. canaliculatus from the snowberry to apple fly
cannot be completely discounted, this scenario appears unlikely
because (i) the genetic distance from apple to snowberry fly-
attacking wasps is greater than that from apple to hawthorn fly-
attacking wasps and (ii) populations of the snowberry fly are
relatively rare and present at low densities in the eastern
United States compared with R. pomonella (35).
Diachasmimorpha. The pattern of microsatellite differentiation for
21 loci for D. mellea also supported the existence of a recently
derived population attacking apple flies. Similar to U. canal-
iculatus, D. alloeum (14), and R. pomonella (18, 22, 32), no fixed
or high-frequency private alleles were found that distinguish
D. mellea populations attacking different fly hosts (see data in the
Dryad database). However, significant and consistent allele fre-
quency differences were observed for eight loci (DM19, DM32,
DM42, DM48, DM51, DM61, DM62, DM65) in at least three
sympatric pairwise comparisons between D. mellea attacking
different fly hosts (Table S2). GLM analyses revealed four loci
(DM14, DM15, DM25, DM61) displaying a significant host ef-
fect across the four sympatric pairs of apple and hawthorn fly-
attacking wasp populations (Table S3). Unlike U. canaliculatus
but similar to Rhagoletis (18, 22, 32) and D. alloeum (14), a high
proportion (28.6%) of loci (DM1, DM6, DM14, DM15, DM19,
DM28) showed a significant host × latitude effect for D. mellea
(Table S3). An NJ network based on all 21 loci clustered all
apple fly-attacking D. mellea separately from populations para-
sitizing hawthorn, blueberry, and black cherry flies (Fig. 4B).
Similar to previous findings for D. alloeum (14), the network
suggested that D. mellea attacking blueberry flies could be the
source of the population parasitizing apple flies. However, this
conclusion is preliminary because (i) only a single population of
D. mellea attacking blueberry flies was genotyped in the study,
A Utetes canaliculatus B Diachasmimorpha mellea
N 6
8 observed over 45 h. Five of the 19 mating pairs (26%) were
100 1 remaining 14 pairs (74%) were observed on branches or leaves
3
7 = 3–45 cm). The mean time from first observation to initiation of
4 mating was 209 ± 35.1 s (range = 12–516 s). No mating pair or
2
6
6 Host Plant Odor Discrimination.
2 5 3
4 1
6 ination in U. canaliculatus similar to Rhagoletis (27, 30) and
3
1 2
1 C U. canaliculatus reared from snowberry, flowering dogwood,
2
3
6 of the Y tube containing their respective natal fruit volatiles
100
2 3
0.01 0.01
Fig. 4. NJ genetic distance networks of Nei’s D based on all microsatellite loci
for (A) U. canaliculatus and (B) D. mellea attacking hawthorn (red), apple
(green), blueberry (blue), flowering dogwood (yellow), snowberry (white), and
black cherry (black) flies. Bootstrap support values ≥ 50% based on 10,000
replicates across loci are given along branches. Numbers next to or inside cir-
cles indicate collecting sites (see Table S1 and Fig. S1 for site designations). The
node of U. canaliculatus labeled “N” refers to a northern population of wasps
attacking flowering dogwood that was pooled across multiple collecting sites.
E5984 | www.pnas.org/cgi/doi/10.1073/pnas.1424717112
(ii) at the sympatric site in Fennville, MI, more loci displayed
significant differences between wasps attacking blueberry and
apple flies (7 loci) than between blueberry and hawthorn flies (2
loci), and (iii) D. mellea is a relatively rare parasitoid of blue-
berry flies (24, 25). Thus, a hawthorn or black cherry fly ancestry
cannot be discounted.
The origins of the recently formed apple fly-attacking parasitoid
community still need to be established. Genetic results imply that
apple-infesting populations of D. alloeum are genetically more
closely related to wasps parasitizing the blueberry fly, R. mendax,
than the hawthorn-infesting race of R. pomonella (14). Although
inferences can be made regarding the origins of the apple-
infesting populations of U. canaliculatus and D. mellea, the genetic
results are more ambiguous than for D. alloeum. Regardless,
Rhagoletis and its parasitoids may not always cospeciate in a strict
1:1 follow-the-leader fashion. Wasps attacking different flies in
the community may take advantage of the new niche opportunity
provided by Rhagoletis host shifts, not necessarily just the para-
sitoid infesting ancestral fly hosts. Thus, adaptive starbursts of
sequential divergence may be the result of biodiversity radiating
from several different origins within the community. Regardless,
the genetic results support criterion 2 and multiplicative sequential
divergence by establishing the existence of host-associated genetic
differentiation among both U. canaliculatus and D. mellea.
Site of Mating Assembly.
Utetes. Field observations of mtDNA haplotype C U. canaliculatus
in stands of apple and hawthorn trees confirmed the use of host
fruit as the site of mating assembly. A total of 23 mating pairs were
detected over 45 h. Eleven of the 23 mating pairs (48%) were de-
tected directly on apple or hawthorn fruit, whereas the remaining
12 pairs (52%) were observed on branches or leaves an average of
8.52 ± 2.11 cm from the nearest host fruit (range = 5–30 cm). The
mean time from the first observation of a male and female to the
initiation of mating was 143 ± 19 s (range = 4–411 s). No mating
pair or individual U. canaliculatus was observed on any non-Rha-
goletis plant, underscoring the host-specific nature of mating.
Diachasmimorpha. Field observations of D. mellea in stands of
apple and hawthorn trees also confirmed the use of host fruit as
the site of mating assembly similar to U. canaliculatus, D. alloeum
(14), and R. pomonella (26). A total of 19 mating pairs were
detected directly on apple or hawthorn fruit, whereas the
an average of 12.37 ± 3.25 cm from the nearest host fruit (range
individual D. mellea was observed on any non-Rhagoletis plant.
Utetes. Results from Y-tube behavioral assays supported the ex-
istence of divergent host choice based on fruit odor discrim-
D. alloeum (14). Wasps displayed no tendency to orient to the
right or left arm of the Y tube when no fruit odor was present
(X2 = 1.2; P = 0.27). However, odor-naïve, mtDNA haplotype
apple, and hawthorn flies all preferentially oriented to the arm
compared with the odorless control arm (Fig. 5 and Table S4).
Wasps also avoided the odors of alternative hosts, orienting less
often to nonnatal fruit volatiles compared with the odorless
control (Fig. 5 and Table S4).
Diachasmimorpha. Results from behavioral assays also supported
the existence of divergent host choice in D. mellea. Wasps showed
no tendency to orient to either arm of the Y tube when no fruit
odor was present (X2 = 0.10; P = 0.76). However, odor-naïve
D. mellea attacking apple, hawthorn, blueberry, and black cherry
flies all preferred their respective natal host fruit volatiles and
avoided nonnatal fruit odors compared with the odorless control
arm (Fig. 5 and Table S4).
Hood et al.
 PNAS PLUS
Fig. 5. Host fruit odor discrimination of (A) D. alloeum,
(B) U. canaliculatus, and (C) D. mellea. Values represent
the percent increase (preference) or decrease (avoid-
ance) in the orientation of wasps to the arm of the
Y tube containing different host fruit volatiles relative
to the blank (odorless) control arm. Each host associ-
ated population of wasps displayed avoidance
behavior to all nonnatal fruit odors. We therefore
averaged avoidance values ± SEMs (gray bars) across
all nonnatal fruit assays (see Table S4 for individual
values and statistical significance). Data for D. alloeum
are taken from Forbes et al. (14). *P = 0.05; **P < 0.01;
***P < 0.001; ****P < 0.0001.

In many models of ecological speciation (1, 12, 13, 16, 17), in- contend they are unlikely to be major contributors to host choice

EVOLUTION
cluding R. pomonella (1, 12, 20, 23, 29), host choice and mate choice for these wasps. In this regard, U. canaliculatus and D. mellea are
(the site where courtship and mating occurs) are linked, generating endoparasitoids that oviposit into fly eggs and larvae, respec-
habitat-related prezygotic RI among diverging populations. When tively, in fruit. Wasp larvae do not feed on their Rhagoletis hosts
we couple the results from host fruit odor discrimination assays and until larvae have left fruit and pupariated in the soil. Thus, wasps
field observations, the same is true for the community of wasps do not come in direct contact with the surface volatiles of rip-
attacking different Rhagoletis hosts, supporting criteria 3 and 4 and ening host fruit before returning to their respective host fruit to
multiplicative sequential divergence. We estimate that fruit odor mate and oviposit as adults the following year. In addition, wasps
discrimination potentially translates into ≥84% prezygotic RI eclose as adults from fly pupal cases buried in the soil for >300 d.
among U. canaliculatus attacking different Rhagoletis hosts at sym- Following eclosion, flies must move quickly to the soil surface or
patric sites (Table 2 and SI Text). Similarly, estimates for D. mellea be trapped underground, therefore making it difficult for wasps
ranged from 68% to 88% across populations (Table 2). to antennate fly pupal cases for host cues. Moreover, any volatile
We caution, however, that observations of site of mating may compound from the surface of ripe fruit, even if originally pre-
be skewed unintentionally by observers’ bias by focusing on lo- sent (which is unlikely, as fly larvae do not consume the skin of
cations in the field where insects are more likely to be present fruit but the flesh, which is chemically different), would have
(on host fruit) than not (on other nonhost plant tissue). Thus, it long since dissipated from the pupal case and surrounding soil.
is possible that we missed mating pairs behaving in a non-host- Furthermore, it is unlikely wasps would also learn to avoid odors
specific manner, thereby overestimating the importance of host emitted from nonnatal fruits from exposure to the natal host fly
choice and RI. However, we spent more time surveying nonhost pupal case or a similar host-related stimuli only. Thus, the bi-
plants (30 h) than host trees (15 h) for parasitoids (SI Text). ologies of U. canaliculatus and D. mellea suggest that fruit odor
Moreover, all of the wasps are conspicuous in color (bright red or
reddish black) and large compared with other parasitoid species
(Fig. 2 B−E). Consequently, wasps or mating pairs residing on a Table 2. Pairwise estimates of habitat isolation due to
nonhost plant would have likely been observed if present. differences in fruit odor discrimination behavior calculated
Reciprocal rearing of Diachasma. In the current study, we reared 22 between wasps attacking different host populations of
hawthorn fly-origin and five blueberry fly-origin D. alloeum through Rhagoletis
a single generation in R. pomonella infesting apples. Both hawthorn-
and blueberry-origin D. alloeum retained preferences for their re- Habitat isolation, %
spective natal host fruit odors, while avoiding apple volatiles (Fig. 6 Host comparison Da Uc Dm
and Table S4). Thus, rearing studies support a genetic basis for
behavioral differences in host fruit odor discrimination for both A vs. H 69 85 77
Rhagoletis (27) and D. alloeum (14), supporting, partially, criteria 5. A vs. B 68 68
At present, we cannot rear U. canaliculatus or D. mellea in A vs. S 79 84
alternate fly hosts to test for environmental or maternal effects A vs. C 85
on fruit odor discrimination, a shortcoming of our results and an A vs. D 89
important avenue for future research. However, examples of H vs. B 74 77
larval conditioning affecting adult preference are rare for holo- H vs. S 56 85
metabolous insects (36) and, when found, usually involve cases H vs. C 88
where juvenile life stages come in direct contact with plant tissues H vs. D 85
during development (37). Nevertheless, instances do exist in which
B vs. S 66
parasitoids learn cues from their insect host (38). For example, the
B vs. C 79
braconid wasp Aphidius colemani learns natal host preference by
S vs. D 85
antennating the mummy of its aphid host following adult emer-
gence. However, if A. colemani are dissected from their hosts Da, D. alloeum; Uc, U. canaliculatus; Dm, D. mellea; A, apple flies; H,
before adult emergence and allowed to complete development, hawthorn flies; S, snowberry flies; B, blueberry flies; C, black cherry flies.
natal host preference is reduced. Although we cannot completely Estimates for Da are calculated from host odor discrimination values from
discount such effects for D. mellea and U. canaliculatus, we Forbes et al. (14).

Hood et al. PNAS | Published online October 23, 2015 | E5985

discrimination may be (partially) genetically based, as is the case
for Rhagoletis (27, 30) and D. alloeum (14).
Eclosion Timing.
Utetes. Eclosion curves differed significantly among mtDNA
haplotype C U. canaliculatus attacking different fly hosts at
sympatric sites, tracking the eclosion times of their Rhagoletis
hosts and the fruiting times of their respective host plants (Fig.
7, Fig. S3, and Table S5). The one exception was at Dowagiac,
MI, where eclosion times of U. canaliculatus infesting apple
(102.83 ± 1.05 d; n = 35) and hawthorn (102.71 ± 1.18 d; n = 38)
flies did not differ (Fig. S3 and Table S5). Interestingly, similar
results were previously found for both R. pomonella and
D. alloeum at this site (Table S5) (14). Mean adult longevity for
U. canaliculatus was 9.54 ± 0.18 d (n = 167) in the laboratory.
Assuming Utetes have the same life span in nature (likely an
overestimate), we estimate that the observed differences in eclosion
time translate into allochronic prezygotic RI between host-associated
populations of U. canaliculatus at all sites (excluding Dowagiac, MI)
(11–96%) (Table 3).
Diachasmimorpha. Eclosion curves for D. mellea also differed sig-
nificantly among populations attacking different fly hosts at
sympatric sites (Fig. 7, Fig. S3, and Table S5). Mean adult lon-
gevity for D. mellea was 12.21 ± 0.30 d (n = 67) in the laboratory.
We estimate that the differences in eclosion times translate into
prezygotic RI between host-associated populations of D. mellea
at all sites (12–55%) (Table 3). We note that wasps may not
reach sexual maturity until several days after adult eclosion, po-
tentially increasing temporal isolation among host-associated pop-
ulations of all three species of parasitoids.
Genetic Correlations with Eclosion Time.
Utetes. Eight loci in females (UC10, UC12, UC14, UC18, UC47,
UC52, UC59, UC60) and eight in males (UC14, UC32, UC48,
UC52, UC53, UC54, UC59, UC60) displayed significant geno-
type or genotype × host effects with eclosion time (Table S6). Of
these loci, six showed a significant host effect between wasps
attacking apple and hawthorn flies in GLM analyses (Table S3),
linking the pattern of genetic differentiation with an important
host-related adaptation causing prezygotic RI. Overall, a signif-
icant stepwise multiple regression was found between seven loci
for females (r = 0.49; P < 0.0001) and six for males (r = 0.40; P =
0.03) with eclosion time. Similar relationships between multiple
microsatellites and eclosion time were found for Rhagoletis (18,
22) and D. alloeum (14).
Diachasmimorpha. Three microsatellite loci for females (DM18,
DM25, DM61) and five loci for males (DM18, DJ19, DM28,
DM33, DM36) displayed significant genotype or genotype × host
effects with eclosion time (Table S6). Of these loci, two displayed
a significant host effect between apple and hawthorn fly-attack-
ing wasps in GLM analyses (Table S3). Consequently, evidence
supporting criteria 6 and 7 was found for both D. mellea and
U. canaliculatus. However, unlike U. canaliculatus, D. alloeum
(14), and R. pomonella (18, 22), stepwise multiple regressions
revealed no significant relationship between microsatellites and
eclosion time for female or male D. mellea.
Conclusions
The concept that “biodiversity begets biodiversity” is deeply
engrained in biology (4–9, 12–15). Sequential divergence has
been proposed to help explain a number of diverse patterns,
including radiations following mass extinctions (4–6), species
diversity in the tropics (7), and macroevolutionary patterns of
diversification in herbivorous insects (8, 9). However, the process
has rarely been explicitly studied. Here, we document for the first
time, to our knowledge, the genetic signatures of and ecological
mechanisms promoting rapid, sequential divergence for three dis-
tinct species, as the same host-related ecological adaptations asso-
ciated with host choice and life history timing in Rhagoletis appear
to have evolved concordantly and multiplicatively in the parasitoid
community.
E5986 | www.pnas.org/cgi/doi/10.1073/pnas.1424717112
Odor Sources
Preference (+)
Hawthorn
Response to odor (%)
200
Apple
Blueberry
100
0
Avoidance (-)
-100
Haw Blue
Parental host
Fig. 6. Host fruit odor discrimination of hawthorn fly-origin and blueberry
fly-origin D. alloeum reared for a single generation on apple-infesting
R. pomonella in apple fruits. Values represent the percent increase (prefer-
ence) or decrease (avoidance) in the orientation of hawthorn-origin or
blueberry-origin D. alloeum to the arm of the Y-tube olfactometer con-
taining the indicated host fruit odor relative to blank odorless control
treatment. **P < 0.01; ***P < 0.001.
We caution, however, that our results represent a single da-
tum. Additional studies are needed to assess how common and
taxonomically widespread sequential divergence, evaluate its
importance for generating patterns of biodiversity, and assess its
role in assembling novel guilds or communities de novo. Condi-
tions promoting sequential divergence (Table 1 and Fig. 1) may
not be common for many species and may be difficult to empir-
ically test. Regardless, even if sequential divergence is rare, it could
still, through a multiplicative effect, make a nontrivial contribu-
tion to the origin of new taxa. Thus, for organisms such as insects
and their parasites that experience and partition resources on a
fine scale (39), the effects of new niche construction may cascade
through ecosystems and have an important effect on biodiversity.
Given that an estimated 10–30 million plant-feeding insect spe-
cies exist (40), that each phytophagous insect is likely attacked by
≥1 parasitoid (41), and that host-specific parasitoids account for
an estimated 25% of insects (40), sequential divergence may be
an underappreciated process (12–15).
Additional concerns regarding sequential divergence require
clarification. In our study, we associated the process with sym-
patric divergence with gene flow (criterion 1). This need not
always be the case. The critical element is that divergent eco-
logical selection pressures transcend trophic levels to generate
new biodiversity. Thus, sequential divergence may occur in any
geographic context. For example, sequential divergence is also
possible in allopatry if the evolution of RI is driven by adaptation
to a new host following migration of a parasite into a different
geographic area (Fig. 1C). However, verifying that population
divergence was due to the cascading effects of selection from
plant, to insect, to parasitoid, as opposed to genomic conflict and
RI arising as an inadvertent by-product of mutation order pro-
cesses (42, 43), is difficult in the absence of gene flow. Conse-
quently, we focused on an example of sequential divergence involving
sympatric host shifts, allowing us to directly document the mul-
tiplicative potential of the process in action. Thus, we excluded
mtDNA haplotypes A and B in Utetes from further analysis here,
but these taxa may represent cases of sequential divergence when
considered from a broader geographic perspective.
Hood et al.

U. canaliculatus
Rhagoletis
Host association
Fig. 7. Mean eclosion times averaged across collecting sites for Rhagoletis
(circles) infesting blueberry (blue), apple (green), snowberry (white), black
cherry (black), hawthorn (red), and flowering dogwood (yellow) host plants,
and D. alloeum (squares), U. canaliculatus (triangles), and D. mellea (dia-
monds) attacking each fly host. Data for D. alloeum are from Forbes et al.
(14). See Table S5 for individual site values ± SEMs.
We also defined the process with respect to “population di-
vergence” rather than using the previously defined term “se-
quential speciation” (14) to convey a focus on recent sympatric
host shifts for empirical testing and to remove any presumption
of the evolutionary outcome of our study. Although our results
support the sequential divergence hypothesis, it is not possible to
discern the long-term evolutionary fate of the species we studied.
We are unable to determine, with certainty, if these taxa will
diverge further along the “speciation continuum” (44), represent
“frozen” multistate polymorphisms in evolutionary equilibrium
between selection and gene flow, or have already attained species
status. Thus, case studies of speciation must rely on inferences
drawn from comparisons involving a series of related population
pairs at varying stages along the speciation continuum (44). We
therefore contend that host-specific populations of D. alloeum,
D. mellea, and U. canaliculatus provide an opportunity to make
inferences concerning the critical ecological attributes and adapta-
tions initiating sequential divergence that may result in speciation.
Hood et al.
PNAS PLUS
Recent theory suggests that, during ecological divergence, pop-
ulations may diverge incrementally through time before a threshold
level of differentiation is reached and the genomes of taxa rapidly
“congeal” into new species (45). Thus, populations that appear to
be in evolutionary limbo may be slowly accumulating new mutations
of small effect and progressively inching their way toward species
status. In this regard, the two cryptic Utetes taxa we discovered (Fig.
2A) imply that the wasps have the potential to diverge further.
However, it remains to be determined whether sequential di-
vergence alone (cascading ecological selection pressures), or in
conjunction with other mechanisms, contributes to the transition
from races or biotypes to species.
With respect to the current taxonomic status of wasp pop-
ulations, we cast our experimental design in terms of eight cri-
teria supporting sympatric host race formation and sequential
divergence (Table 1). Our results support, partially or in whole,
criteria 1–7 for Rhagoletis and its parasitoid community. Due to
difficulties in wasp husbandry, we could not test for host fly-
related fitness tradeoffs among populations (criterion 8). Re-
ciprocal fruit transplant studies of apple- and hawthorn-infesting
R. pomonella have failed to detect host-related differences in
larval feeding performance with respect to survivorship to pupar-
iation (46). However, tradeoffs in larval performance have been
documented for R. pomonella sibling species (33, 34) that may
distinguish species from the apple and hawthorn host races. If the
EVOLUTION
same is true for wasps, experiments investigating performance
tradeoffs may help establish the taxonomic status of D. alloeum,
D. mellea, and U. canaliculatus. Nevertheless, estimates of habitat
isolation via host odor discrimination and temporal isolation
via differences in eclosion timing substantially reproductively
isolate host-associated populations of D. alloeum, D. mellea, and
U. canaliculatus (Tables 2 and 3). Taking into account the relative
contribution of each measure of RI (SI Text), temporal isolation
and habitat isolation can act in concert to produce prezygotic RI
among pairs of D. alloeum (70–94%), U. canaliculatus (85–99%),
and D. mellea (79–90%). Moreover, other cues in addition to fruit
surface volatiles also likely increase host plant fidelity and habitat
isolation for the wasps (e.g., differences in fruit color, size, and
taste), as in Rhagoletis (47). Host fruit odor discrimination could
also contribute to postzygotic RI for the wasps due to hybrid
behavioral sterility, as it may for Rhagoletis (30).
Finally, competition may limit the potential for multiplicative
sequential divergence (25), placing limits on the overall number
of parasitoids that can simultaneously share the same host.
However, we discovered two cryptic, highly diverged haplotypes
of egg-attacking Utetes sharing snowberry and flowering dog-
wood hosts, suggesting that wasps can subdivide the fly resource
and avoid extinction. Regardless, competition does occur among
Rhagoletis parasitoids, as only one adult wasp ecloses when a
single fly is attacked by multiple parasitoids (25). Furthermore,
larval-attacking D. mellea and D. alloeum overlap in host use.
Thus, several species cooccur, and it is possible that interspecific
interactions may foster diversity by selecting for newly shifting
wasps to differentiate to avoid competition on novel hosts, re-
ducing temporal overlap and increasing their RI from ancestral
population (27). Indeed, eclosion time differences exist not only
within each species attacking different fly hosts but also between
parasitoid species sharing the same host (Table 2 and Fig. 7).
Therefore, a degree of temporal niche partitioning for oviposi-
tion sites exists among species, a common phenomenon among
parasitoids sharing the same host (48, 49). Although speculative,
this may permit coexistence of multiple species on shared fly
hosts. Moreover, escape from parasitism itself may also favor
host shifting in Rhagoletis (50). Thus, host plant, fly, and wasp
may play a tritrophic coevolutionary game of hide and seek that
spins off several new life forms.
Materials and Methods
Specimen Collection. Host fruits infested with parasitized flies were collected
from nine sites in the eastern United States from 2006 to 2012 (Fig. S1, Table
S1, and SI Text). Six of these sites for U. canaliculatus and four for D. mellea
PNAS | Published online October 23, 2015 | E5987
 Table 3. Pairwise estimates of both intraspecific (within) and interspecific (between) temporal isolation for wasps
attacking different and the same Rhagoletis host(s), respectively, at sympatric sites
Within-species
comparisons, %
Site no. Location Host comparison Da Uc
1 Grant, MI A vs. H 17 11
2 E. Lansing, MI A vs. H 34 30
A vs. S 5 52
H vs. S 53 47
3 Fennville, MI A vs. H 30 15
A vs. B 29
H vs. B 75
4 Dowagiac, MI A vs. H 5 3
5 Three Oaks, MI A vs. H 28
6 S. Bend, IN A vs. H 96
A vs. C
H vs. C
7 Urbana, IL A vs. H
See Table S1 and Fig. S1 for site designations. Da, D. alloeum; Uc, U. canaliculatus; Dm, D. mellea; A, apple flies; H, hawthorn flies;
S, snowberry flies; B, blueberry flies; C, black cherry flies. Data for Da from Fennville, MI, are modified from Forbes et al. (14).
represent locations where two or more Rhagoletis taxa were sympatric
(<1.5 km apart) or in close proximity (<10 km apart).
Genetic Methods. Twenty and 21 pairs of unique microsatellite primers were
developed for U. canaliculatus and D. mellea, respectively (see data in the
Dryad database). Compared with other types of markers, microsatellites
have high mutation rates and levels of polymorphism, characteristics useful
for revealing subtleties in population genetic structure associated with rapid
and recent divergence. We also PCR-amplified and sequenced a 540-bp region
of the mtDNA Cytochrome Oxidase subunit I (COI). For further details on ge-
netic methodology, see SI Text.
Analysis of Genetic Data. Microsatellite loci were tested for conformity to
Hardy−Weinberg equilibrium (HWE) in females (diploid) using Markov chain
randomization methods with default parameters in Micro-Checker (51). Of
the 41 microsatellites scored, no locus displayed significant deficiencies of
heterozygosity in more than two populations, implying markers represented
single-copy sequences that did not possess high frequencies of null alleles.
Furthermore, every haploid male possessed only a single allele for each
microsatellite, and we observed no obvious null allele at each locus. We
estimated linkage disequilibrium (LD) between pairs of microsatellites sep-
arately for each host-associated population using Burrow’s composite Δ (52),
which does not assume HWE or require phased data but instead provides a
joint metric of intralocus and interlocus disequilibria based solely on geno-
type frequencies. Thus, Δ is equivalent to the LD parameter D under HWE (52).
We combined multilocus haplotypes of individual males together at random,
based on their frequencies in a population, to construct diploid genotypes in
HWE to combine with females genotypes to estimate a single Δ value for each
population. These values were then transformed to a standardized correlation
coefficient and combined across populations using the method of Fisher (53) to
derive overall measures that were tested for significance by χ2 tests, as de-
scribed by Weir (52). In general, LD was low between loci of both species, an
indication that most markers were evolving independently. However, certain
microsatellites did display significant LD, and these loci were (i) arranged
stepwise with one another to form arrays interconnecting the majority of
markers genotyped for each species (Fig. S4) and (ii) used in association with
eclosion time relationships to pool alleles at loci for GLM analysis (Table S3).
We tested for a relationship between eclosion timing and microsatellite
genotypes separately for males and females by a two-way ANOVA with
genotype and host as main effects. Our analyses were performed on the
subset of individuals genotyped (Table S1) that were also phenotyped for
eclosion timing (see data in the Dryad database). We concentrated on the
sympatric pairs of wasps attacking hawthorn and apple flies, transforming
the eclosion time for each individual by its SD from the mean eclosion time for
each sex at the site from which it was collected. Normalizing eclosion times
E5988 | www.pnas.org/cgi/doi/10.1073/pnas.1424717112
Between-species comparisons, %
Dm Host Da vs. Uc Da vs. Dm Uc vs. Dm
55 A 11 9 13
H 30 8 22
12 A 31 51 17
H 74 74 4
S 85
54 A 21 18 12
33 H 40 37 21
38 B 19
A 53
H 31
17 A 67
21 H 21
16
H 85
allowed us to combine data across hosts and sites to test for genetic associa-
tions. We report the results for the main genotype effect and the genotype ×
host interaction because the host effect was significant in the ANOVA for all
loci. We also performed stepwise multiple regressions for all microsatellites
against normalized eclosion times separately for females and males.
We tested for significant allele frequency differences between host-
associated populations at sympatric sites using a nonparametric, Monte Carlo
approach. Microsatellite genotypes were randomly resampled with re-
placement from the combined data set for a locus at a site to determine the
probability from 100,000 replicates of generating a Nei’s genetic distance D
(54) greater than the observed value between populations by chance. We
also performed GLM analyses at sympatric sites to test for significant host
and latitude effects on microsatellite allele frequencies for each species. For
the GLM analyses, allele frequency was modeled as a quasi-binomial vari-
able, with host as a discrete factor and latitude as a continuous factor. The
GLM analysis required that we collapse the microsatellites to a diallelic
dataset. To accomplish this, we pooled different combinations of alleles at
each locus into two variant classes, using the diallelic combination that (i)
generated the highest level of LD with other loci (Fig. S4) and (ii) explained
the largest amount of variation in adult eclosion time (Table S6), similar to
Michel et al. (18) and Powell et al. (22) (see data in the Dryad database).
We assessed the genetic relationships among populations for COI mtDNA
by constructing maximum parsimony gene trees for U. canaliculatus and
D. mellea using Phylogeny Inference Package (PHYLIP) 3.69 (55). We exam-
ined relationships for the nuclear-encoded microsatellites by constructing
unrooted NJ networks for U. canaliculatus and D. mellea based on Nei’s
pairwise genetic distances calculated between populations using PHYLIP
3.69 (55), with bootstrap support determined by 10,000 replicates across loci.
Eclosion Study. The timing of adult eclosion was determined by monitoring fly
pupae once daily, following overwinter, for emerging flies and wasps. Adult
wasp longevity was calculated by measuring the interval, in days, between
adult emergence and death. We estimated the degree of allochronic isolation
between wasp populations due to differences in eclosion time and adult
according to the formula in SI Text.
Field Observations and Behavioral Assays. To assess the site of assembly for
mating, we conducted 45 h of observations of behavior for both U. canal-
iculatus and D. mellea in stands of apple and hawthorn trees, and nearby
non-Rhagoletis host plants, at sympatric study sites in Fennville and Grant,
MI (14) (SI Text). The responses of U. canaliculatus, D. mellea, and D. alloeum
to natal and nonnatal fruit volatiles were determined using a Y-tube olfac-
tometer based on the protocol previously developed for D. alloeum (14) (SI
Text). We estimated the degree of habitat isolation between wasp populations
due to host fruit odor discrimination according to the formula in SI Text.
Hood et al.

ACKNOWLEDGMENTS. We thank E. Archie, S. Berlocher, M. Brueseke,
T. Brown, M. Doellman, C. Finn, M. Glover, K. Gomez, T. Kubler,
J. McLachlan, P. Meyers, M. Pfrender, G. Ragland, D. Sanford, H. Schuler,
C. Tait, and especially P. Morton for helpful comments and support.
1. Coyne JA, Orr HA (2004) Speciation (Sinauer, Sunderland, MA).
2. Feder JL, Flaxman SM, Egan SP, Comeault AA, Nosil P (2013) Geographic mode of
speciation and genomic divergence. Annu Rev Ecol Syst 44:73–97.
3. Pinho C, Hey J (2010) Divergence with gene flow: Models and data. Annu Rev Ecol Syst
41:215–230.
4. Erwin DH (2008) Macroevolution of ecosystem engineering, niche construction and
diversity. Trends Ecol Evol 23(6):304–310.
5. Butterfield NJ (2001) The Ecology of the Cambrian Region (Columbia Univ Press, New
York).
6. Bush AM, Bambach RK, Erwin DH (2011) Ecosphere utilization during the Ediacaran
Radiation and the Cambrian Eco-explosion. Quantifying the Evolution of Early Life:
Numerical Approaches to the Evalucation of Fossils and Ancient Ecosystems, eds
Laflamme M, Schiffbauer J, Dornbos S (Springer, Doredrecht, The Netherlands), pp
111–133.
7. Emerson BC, Kolm N (2005) Species diversity can drive speciation. Nature 434(7036):
1015–1017.
8. Strong DR, Lawton JA, Southwood TRE (1984) Insects on Plants: Community Patterns
and Mechanisms (Blackwell Sci, Oxford).
9. Mitter C, Farrell B, Wiegmann B (1988) The phylogenetic study of adaptive zones: Has
phytophagy promoted insect diversification? Am Nat 132(1):107–128.
10. Whitham TG, et al. (2006) A framework for community and ecosystem genetics: From
genes to ecosystems. Nat Rev Genet 7(7):510–523.
11. Whitham TG, et al. (2008) Extending genomics to natural communities and ecosys-
tems. Science 320(5875):492–495.
12. Abrahamson WG, Blair CP (2008) Sequential radiation through host-race formation:
Herbivore diversity leads to diversity in natural enemies. Specialization, Speciation,
and Radiation: The Evolutionary Biology of Herbivorous Insects, ed Tilmon KJ (Univ
California Press, Oakland, CA), pp 188–202.
13. Feder JL, Forbes AA (2010) Sequential divergence and the diversity of parasitic insects.
Ecol Entomol 35(Suppl 1):67–76.
14. Forbes AA, Powell THQ, Stelinski LL, Smith JJ, Feder JL (2009) Sequential sympatric
speciation across trophic levels. Science 323(5915):776–779.
15. Stireman JO, Nason JD, Heard SB, Seehawer JM (2006) Cascading host-associated
genetic differentiation in parasitoids of phytophagous insects. Proc Biol Sci 273(1586):
523–530.
16. Drès M, Mallet J (2002) Host races in plant-feeding insects and their importance in
sympatric speciation. Philos Trans R Soc Lond B Biol Sci 357(1420):471–492.
17. Berlocher SH, Feder JL (2002) Sympatric speciation in phytophagous insects: Moving
beyond controversy? Annu Rev Entomol 47:773–815.
18. Michel AP, et al. (2010) Widespread genomic divergence during sympatric speciation.
Proc Natl Acad Sci USA 107(21):9724–9729.
19. Egan SP, et al. (2015) Experimental evidence of genome-wide impact of ecological
selection during early stages of speciation-with-gene-flow. Ecol Lett 18(8):817–825.
20. Berlocher SH (2000) Radiation and divergence in the Rhagoletis pomonella species
group: Inferences from allozymes. Evolution 54(2):543–557.
21. Xie X, et al. (2008) Radiation and divergence in the Rhagoletis pomonella species
complex: Inferences from DNA sequence data. J Evol Biol 21(3):900–913.
22. Powell THQ, et al. (2013) Genetic divergence along the speciation continuum: The
transition from host race to species in rhagoletis (Diptera: tephritidae). Evolution
67(9):2561–2576.
23. Bush GL (1969) Sympatric host race formation and speciation in frugivorous flies of
the genus Rhagoletis (Diptera: Tephritidae). Evolution 23(2):237–251.
24. Forbes AA, Hood GR, Feder JL (2010) Geographic ranges and host breadths of para-
sitoid wasps associated with the Rhagoletis pomonella (Diptera: Tephritidae) species
complex. Ann Entomol Soc Am 103(6):908–915.
25. Hood GR, Egan SP, Feder JL (2012) Interspecific competition and speciation in en-
doparasitoids. Evol Biol 39(2):219–230.
26. Prokopy RJ, Bennett EW, Bush GL (1971) Mating behavior in Rhagoletis pomonella
(Diptera: Tephritidae). I. Site of assembly. Can Entomol 103(10):1405–1409.
27. Linn C, Jr, et al. (2003) Fruit odor discrimination and sympatric host race formation in
Rhagoletis. Proc Natl Acad Sci USA 100(20):11490–11493.
28. Forbes AA, Fisher J, Feder JL (2005) Habitat avoidance: Overlooking an important
aspect of host-specific mating and sympatric speciation? Evolution 59(7):1552–1559.
29. Feder JL, et al. (1994) Host fidelity is an effective premating barrier between sym-
patric races of the apple maggot fly. Proc Natl Acad Sci USA 91(17):7990–7994.
30. Linn CE, Jr, et al. (2004) Postzygotic isolating factor in sympatric speciation in Rha-
goletis flies: Reduced response of hybrids to parental host-fruit odors. Proc Natl Acad
Sci USA 101(51):17753–17758.
31. Dambroski HR, Feder JL (2007) Host plant and latitude-related diapause variation in
Rhagoletis pomonella: A test for multifaceted life history adaptation on different
stages of diapause development. J Evol Biol 20(6):2101–2112.
32. Filchak KE, Roethele JB, Feder JL (2000) Natural selection and sympatric divergence in
the apple maggot Rhagoletis pomonella. Nature 407(6805):739–742.
33. Bierbaum TJ, Bush GL (1990) Genetic differentiation in the viability of sibling species
of Rhagoletis fruit-flies on host plants, and the influence of reduced hybrid viability
on reproductive isolation. Entomol Exp Appl 55(2):105–118.
Hood et al.
PNAS PLUS
Funding was provided to J.L.F., A.A.F., and S.P.E. by the National Sci-
ence Foundation (NSF) (DEB-1145573) and to G.R.H. by the Indiana Academy
of Science, the Entomological Society of America, Sigma Xi, and NSF (DEB-
1310850).
34. Ragland GJ, et al. (2015) Differences in performance and transcriptome-wide gene
expression associated with Rhagoletis (Diptera: Tephritidae) larvae feeding in alter-
nate host fruit environments. Mol Ecol 24(11):2759–2776.
35. Gavrilovic V, et al. (2007) Rhagoletis zephyria (Diptera: Tephritidae) in the Great Lakes
Basin: A native insect on native hosts? Ann Entomol Soc Am 100(4):474–482.
36. Barron AB (2001) The life and death of Hopkins’ host-selection principle. J Insect
Behav 14(6):725–737.
37. Turlings TCJ, et al. (1993) Learning of host-finding cues by hymenopterous parasit-
oids. Insect Learning: Ecological and Evolutionary Perspectives, eds Papaj DR, Lewis A
(Chapman & Hall, New York), pp 51–78.
38. Storeck K, Poppy GM, van Emden HF, Powell W (2000) The role of plant chemical cues
in determining host preference in the generalist aphid parasitoid Aphidius colemani.
Entomol Exp Appl 97(1):41–46.
39. Bush GL (1993) A reaffirmation of Santa Rosalia, or why there are so many kinds of
small animals? Evolutionary Patterns and Processes, eds Edwards D, Lees DR (Aca-
demic, New York), pp 229–249.
40. Price PW (1980) Evolutionary Biology of Parasites (Princeton Univ Press, Princeton).
41. Hawkins BA, Lawton JH (1987) Species richness for parasitoids of British phytopha-
gous insects. Nature 326(6115):788–790.
42. Schluter D (2009) Evidence for ecological speciation and its alternative. Science
323(5915):737–741.
43. Nosil P, Flaxman SM (2011) Conditions for mutation-order speciation. Proc Biol Sci
278(1704):399–407.
44. Feder JL, Egan SP, Nosil P (2012) The genomics of speciation-with-gene-flow. Trends
EVOLUTION
Genet 28(7):342–350.
45. Flaxman SM, Wacholder AC, Feder JL, Nosil P (2014) Theoretical models of the in-
fluence of genomic architecture on the dynamics of speciation. Mol Ecol 23(16):
4074–4088.
46. Reissig WH, Smith DC (1978) Bionomics of Rhagoletis pomonella in Crataegus. Ann
Entomol Soc Am 71(2):155–159.
47. Prokopy RJ, Roitberg BD (1984) Foraging behavior of true fruit flies. Am Sci 72(1):
41–49.
48. Weis AR, Abrahamson WG (1985) Potential selective pressures by parasitoids on a
plant-herbivore interaction. Ecology 66(4):1261–1269.
49. Hackett-Jones E, Cobbold C, White A (2009) Coexistence of multiple parasitoids on a
single host due to differences in parasitoid phenology. Theor Ecol 2(1):19–31.
50. Feder JL (1995) The effects of parasitoids on sympatric host races of the apple maggot
fly, Rhagoletis pomonella (Diptera: Tephritidae). Ecology 76(3):801–813.
51. Van Oosterhout C, Hutchinson WF, Wills DPM, Shirley P (2004) Micro-Checker: Soft-
ware for identifying and correcting genotyping errors in microsatellite data. Mol Ecol
Notes 4(3):535–538.
52. Weir BS (1979) Inferences about linkage disequilibrium. Biometrics 35(1):235–254.
53. Sokal RR, Rohlf FJ (2012) Biometry: the Principles and Practice of Statistics in Biological
Research (Freeman, New York).
54. Nei M (1972) Genetic distance between populations. Am Nat 106(949):283–292.
55. Felsenstein J (2005) PHYLIP (Phylogeny Inference Package) Version 3.6 (Dep Genome
Sci, Univ Washington, Seattle, WA).
56. Lathrop FH, Newton RC (1933) The biology of Opius melleus Gahan, a parasite of the
blueberry maggot. J Agric Res 46(2):143–160.
57. Wharton WA, Marsh PM (1978) New world Opiinae (Hymenoptera: Braconidae)
parasitic on Tephritidae (Diptera). J Wash Acad Sci 68(4):147–167.
58. Wharton RA (1997) Generic relationships of opine Braconidae (Hymenoptera) para-
sitic on fruit infesting Tephritidae (Diptera). Contrib Am Entomol Inst 30:1–53.
59. Wharton RA, Yoder MJ (2015) Parasitoids of fruit-infesting tephritidae. Available at
paroffit.org/public/site/paroffit/home. Accessed July 2, 2015.
60. Simon C, et al. (1994) Evolution, weighting and phylogenetic utility of mitochondrial
gene sequences and a compilation of conserved polymerase chain reaction primers.
Ann Entomol Soc Am 87(6):651–701.
61. Nojima S, Linn C, Jr, Morris B, Zhang A, Roelofs W (2003) Identification of host fruit
volatiles from hawthorn (Crataegus spp.) attractive to hawthorn-origin Rhagoletis
pomonella flies. J Chem Ecol 29(2):321–336.
62. Nojima S, Linn C, Jr, Roelofs W (2003) Identification of host fruit volatiles from
flowering dogwood (Cornus florida) attractive to dogwood-origin Rhagoletis po-
monella flies. J Chem Ecol 29(10):2347–2357.
63. Powell THQ, Cha DH, Linn CE, Jr, Feder JL (2012) On the scent of standing variation for
speciation: Behavioral evidence for native sympatric host races of Rhagoletis pomo-
nella (Diptera: Tephritidae) in the southern United States. Evolution 66(9):2739–2756.
64. Powell THQ, Forbes AA, Hood GR, Feder JL (2014) Ecological adaptation and re-
productive isolation in sympatry: Genetic and phenotypic evidence for native host
races of Rhagoletis pomonella. Mol Ecol 23(3):688–704.
65. Ramsey J, Bradshaw HD, Jr, Schemske DW (2003) Components of reproductive iso-
lation between the monkeyflowers Mimulus lewisii and M. cardinalis (Phrymaceae).
Evolution 57(7):1520–1534.
66. Stelinski LL, Oakleaf R, Rodriguez-Saona C (2007) Oviposition-deterring pheromone
deposited on blueberry fruit by the parasitic wasp, Diachasma alloeum. Behaviour
144(4):429–445.
67. Jones TH, Godfray HCJ, Hassell MP (1996) Relative movement of a tephritid fly and its
parasitoid wasps. Oecologia 106(3):317–324.
PNAS | Published online October 23, 2015 | E5989