Hybrid speciation through sorting of parental ...lithornis.nmsu.edu › ~phoude › Hermansen_et_al_2014...ences and structural rearrangements induced by recom-bination (Rieseberg

Hybrid speciation through sorting of parentalincompatibilities in Italian sparrows

JO S. HERMANSEN,* FREDRIK HAAS,* 1 CASSANDRA N. TRIER,*1 RICHARD I. BAILEY,*

ALEXANDER J . NEDERBRAGT,* ALFONSO MARZAL† and GLENN-PETER SÆ TRE*

*Department of Biosciences, Centre for Ecological and Evolutionary Synthesis, University of Oslo, PO Box 1066, Blindern,

N-0316 Oslo, Norway, †Department of Zoology, University of Extremadura, ES-06071 Badajoz, Spain

Abstract

Speciation by hybridization is emerging as a significant contributor to biological

diversification. Yet, little is known about the relative contributions of (i) evolutionary

novelty and (ii) sorting of pre-existing parental incompatibilities to the build-up of

reproductive isolation under this mode of speciation. Few studies have addressed

empirically whether hybrid animal taxa are intrinsically isolated from their parents,

and no study has so far investigated by which of the two aforementioned routes intrin-

sic barriers evolve. Here, we show that sorting of pre-existing parental incompatibili-

ties contributes to intrinsic isolation of a hybrid animal taxon. Using a genomic cline

framework, we demonstrate that the sex-linked and mitonuclear incompatibilities iso-

lating the homoploid hybrid Italian sparrow at its two geographically separated

hybrid–parent boundaries represent a subset of those contributing to reproductive iso-

lation between its parent species, house and Spanish sparrows. Should such a sorting

mechanism prove to be pervasive, the circumstances promoting homoploid hybrid spe-

ciation may be broader than currently thought, and indeed, there may be many cryptic

hybrid taxa separated from their parent species by sorted, inherited incompatibilities.

Keywords: birds, hybridization, introgression, Passer, reproductive barriers, speciation

Received 28 May 2014; revision received 29 August 2014; accepted 1 September 2014

Introduction

Natural hybridization is increasingly recognized as an

important contributor to biological diversification

(Abbott et al. 2013; Seehausen et al. 2014). Existing taxa

exchange adaptations through introgression (Heliconius

Genome Consortium 2012) and new species form—not

only in the face of gene flow (Nosil 2012), but also as a

direct consequence of interbreeding (Rieseberg 1997;

Mallet 2007; Rieseberg & Willis 2007; Mavarez & Lin-

ares 2008; Abbott & Rieseberg 2012; Abbott et al. 2013).

The latter process, known as hybrid speciation, encom-

passes the origin of genetically admixed taxa that are

reproductively isolated from their parents. The develop-

ment of reproductive barriers plays a central role in

hybrid speciation, yet the nature and development of

such barriers remain poorly understood (Rieseberg

1997; Mallet 2007; Abbott et al. 2013; Schumer et al.

2014). An important task in hybrid speciation research

is therefore to unveil the evolution of the barriers that

separate hybrid taxa from their parent species.

Both pre- and postzygotic barriers have been shown

to act between hybrid taxa and their parents (Rieseberg

1997; Mallet 2007; Mavarez & Linares 2008; Abbott et al.

2013). A recent survey of the topic suggested that eco-

logical divergence and geographic isolation are likely to

be of greater importance for the establishment of hybrid

taxa than are intrinsic postzygotic barriers, especially in

animals (Abbott & Rieseberg 2012). The evolution of

intrinsic reproductive isolation between hybrids and

parents is poorly studied and understood, however

(Buerkle et al. 2000; Coyne & Orr 2004; Jiggins et al.

2008; Schumer et al. 2014). The only direct empirical evi-

dence comes from the Helianthus sunflower system,

Correspondence: Glenn-Peter Sætre, Fax: +47 22854001;

E-mail: [email protected] authors contributed equally to this work.

© 2014 John Wiley & Sons Ltd

Molecular Ecology (2014) 23, 5831–5842 doi: 10.1111/mec.12910

where intrinsic isolation seems to result from a mixture

of sorting of pre-existing chromosomal structural differ-

ences and structural rearrangements induced by recom-

bination (Rieseberg et al. 1995, 2003; Lai et al. 2005). To

what extent intrinsic barriers separating hybrids from

their parents develop through sorting of pre-existing

parental incompatibilities or through de-novo epistatic

interactions in the hybrid genomes therefore remains an

open question. This question can, however–as sug-

gested by Rieseberg (1997)–be addressed by comparing

the locations of genomic regions contributing to intrin-

sic reproductive barriers between the parent species

with those isolating the hybrid species from their par-

ents. If the genomic regions contributing to reproduc-

tive isolation in a hybrid taxon are a subset of those

acting between its parents, then the sorting hypothesis

would be supported.

Here, we apply the approach suggested by Rieseberg

(1997) to investigate the evolution of intrinsic barriers

during homoploid hybrid speciation in Passer sparrows.

In this species complex, interbreeding between house

P. domesticus and Spanish sparrows P. hispaniolensis has

resulted in a taxon of hybrid origin, the Italian sparrow

P. italiae, which genetically and phenotypically is a mix-

ture of its parent species (Meise 1936; Elgvin et al. 2011;

Hermansen et al. 2011; Trier et al. 2014). The Italian spar-

row is a small seed-eating bird occupying a human-

commensal niche similar to that of the house sparrow

(Summers-Smith 1988). It is distributed over the entire

Italian Peninsula and meets the house sparrow in a nar-

row hybrid zone in the Alps. It lives sympatrically with

the Spanish sparrow in restricted areas of mainland Italy

(Summers-Smith 1988; Hermansen et al. 2011) and

exchanges migrants with Spanish sparrows from Sardi-

nia (Trier et al. 2014). The Italian sparrow’s parent spe-

cies are themselves broadly sympatric across large parts

of the Spanish sparrow range, maintaining phenotypic

distinctness in all but a few locations (Summers-Smith

1988). This demonstrates the existence of reproductive

barriers typically isolating house and Spanish sparrows.

Previous work has shown that the Italian sparrow

has developed reproductive barriers against both parent

species (Hermansen et al. 2011; Trier et al. 2014). Evi-

dence suggests that intrinsic barriers consist mainly of

sex-linked and mitonuclear incompatibilities (Trier et al.

2014). A set of seven candidate reproductive isolation

(RI) loci have been identified, three of which exhibit

steep clines at the Italian-house sparrow boundary and

four which exhibit steep clines at the Italian–Spanish

sparrow boundary. Among these candidate RI loci, sex

chromosome (Z)-linkage is overrepresented, and the

patterns of variation suggest a role for mitonuclear con-

flict in isolating Italian and Spanish sparrows (Trier

et al. 2014).

Here, we take the next step in the study of reproduc-

tive isolation in the Passer system and investigate

whether these intrinsic barriers have formed through

sorting of pre-existing incompatibilities or through

de-novo epistatic interactions in the hybrid genome. We

accomplish this by replicating the hybrid–parents study

of Trier et al. (2014) in an area of parental species symp-

atry and hybridization on the Iberian Peninsula.

Materials and methods

Focal population

Our focal population is situated on the Iberian Penin-

sula, close to Badajoz, Extremadura, Spain (38.64866N,

�7.215453E, Fig. 1). This is an area where house and

Spanish sparrows occur sympatrically, including breed-

ing in the same stork nests. As in our previous hybrid–

parents study (Trier et al. 2014), only male individuals,

which are diploid for the Z chromosome, were analysed

to avoid issues related to haplodiploidy of the Z chro-

mosome. One hundred and sixty-three males (pheno-

typic house sparrows: N = 76, phenotypic Spanish

sparrows: N = 87) were caught using mist nets, and a

blood sample (20–50 lL) for use in genetic analyses was

taken from each individual by puncturing a brachial

vein. DNA was extracted from blood samples stored in

standard buffer using Qiagen DNeasy 96 Blood and Tis-

sue Kits (Qiagen N.V., Venlo, the Netherlands) accord-

ing to the manufacturer’s instructions with the minor

adjustment of adding 100 lL of blood/buffer in the ini-

tial step. Authorization to catch birds and take blood

samples was obtained from the appropriate authorities,

and all birds were released upon finished sampling.

Reference populations

As reference house sparrows, we used previously pub-

lished data from two allopatric populations; Oslo, Nor-

way (N = 58) and Hradec Kr�alov�e, Czech Republic

(N = 27) (Fig. 1; Trier et al. 2014). Population genetic

analyses have shown that the house sparrow exhibits

very little geographic structuring genetically (Sætre

et al. 2012). This is likely a result of a recent, massive

range and population expansion associated with the

sparrows’ recent adaptation to human commensalism in

connection with the rise and spread of agriculture

(Sætre et al. 2012). House sparrows from Norway and

Czech Republic are therefore suitable allopatric refer-

ence populations for use in comparative analyses with

the sympatric area on the Iberian Peninsula.

As reference Spanish sparrows, we used a population

on the Gargano Peninsula in east-central Italy (N = 52)

(Fig. 1). This reference population was chosen based on


5832 J . S . HERMANSEN ET AL.

an assessment of hybrid indices from our available col-

lection of samples. This assessment indicated that these

birds, albeit locally sympatric with Italian sparrows, are

genetically pure Spanish sparrows and hence should

give good estimates of pure Spanish sparrow allele fre-

quencies. In contrast, allopatric Spanish sparrows from

Sardinia show evidence of some gene exchange with

mainland Italian sparrows, through episodes of migra-

tion (Trier et al. 2014; see also Fig. 2A).

Hybrid–parent comparison

To obtain a direct comparison with hybrid–parent

boundaries, we reanalysed the data set utilized by Trier

et al. (2014) using the exact same marker set and paren-

tal reference populations as used when analysing the

focal sympatric parental population. The hybrid–parents

data set consists of 57 populations of Italian sparrows,

Italian-house hybrids and parapatric house sparrows

from the Italian peninsula and the Alps (N = 385), as

well as parapatric Spanish sparrows from Sardinia

(N = 48).

Marker set

We utilized 77 species-informative SNP markers (Table

S1, Supporting information) from 75 genes developed

through 454 transcriptome sequencing of cDNA from

heart, liver and brain tissues (described in Trier et al.

2014). To obtain species-informative SNPs from 454

transcriptome data of six house and six Spanish spar-

rows lacking reference genomes, we mapped against

the most closely related reference genome available,

that of the zebra finch Taeniopygia guttata (Warren et al.

2010). First, we mapped all cDNA reads from the house

sparrows against the zebra finch transcriptome and

used the consensus called on the mapped reads as ref-

erence for mapping all cDNA reads from both house

and Spanish sparrows. Then, we singled out the bases

where all reads from one species exhibited a different

House sparrowPasser domesticus

Italian sparrowPasser italiae

Spanish sparrowPasser hispaniolensis

Oslo

Sardinia

Gargano

Badajoz (3)

Hradec Králove (2)

(1)

(4)

(5)

Fig. 1 Geographic distribution of the studied Passer taxa. Grey indicates the distribution of the house sparrow, orange indicates the

distribution of the Italian sparrow, and red indicates the distribution of the Spanish sparrow. Hatched areas indicate regions where

the house sparrow and Spanish sparrow (red-grey) or Italian sparrow and Spanish sparrow (red-orange) distribution overlap. Black

dots refer to populations presented in Fig. 2A. Bird drawings indicate species-specific male plumage characteristics of the house

sparrow, Italian sparrow and Spanish sparrow.


HYBRID SPECIATION BY SORTING OF INCOMPATIBILITIES 5833

base than that of the other species, that is putatively spe-

cies-diagnostic SNPs. These steps were repeated a

second time starting with the cDNA reads from the

Spanish sparrows (for detailed description of the

species-diagnostic SNP detection method, see

https://github.com/lexnederbragt/454_transcriptome_snps).

Based on the species-diagnostic SNPs from the 12 sam-

ples, multiplex sets of PCR primers were designed and

all genotyped at each of 77 SNP loci (with the exception

of five loci where genotyping failed for 44 of the Span-

ish sparrow reference individuals, see Table S1, Sup-

porting information) using the Sequenom MassARRAY

platform at CIGENE, Norwegian University of Life Sci-

ences, �As, Norway. Genes were annotated by blasting

against the zebra finch and chicken Gallus gallus (Inter-

national Chicken Genome Sequencing Consortium 2004)

genomes. Two exceptions were SNPs within CHD1Z

and ND2 genes, which were genotyped using existing

primers (see Elgvin et al. 2011). The CHD1Z SNP in this

set is located within an intron.

We note that the six Spanish sparrows used for

transcriptome sequencing and SNP detection were

sampled from the same sympatric locality on the

Iberian Peninsula as the focal sparrows in this study,

whereas the six house sparrows used for transcriptome

sequencing and SNP detection were sampled in Oslo,

Norway. Thus, all else being equal, one might expect to

observe more introgression into house sparrows than

Spanish sparrows in the area of parental sympatry, as

the house sparrows in the parental sympatric popula-

tion were not among the populations used for identifi-

cation of species-informative markers. This potential

problem should not, however, adversely affect the com-

parison of locus-specific introgression between hybrid–

parents and parent–parent systems, that is the compara-

tive genomic clines analyses (see below).

Comparative investigation of introgression patterns

We used Bayesian admixture analysis as implemented

in STRUCTURE (Pritchard et al. 2000; Falush et al. 2003)

to compare the genetic makeup of individuals in the

focal sympatric area in Spain (house: N = 76, Spanish:

N = 87) to that of other populations of house and Span-

ish sparrows and hence to assess patterns of introgres-

sion. As representative allopatric house sparrow

(A)

(B) (C)

Fig. 2 Genetic composition of house and Spanish sparrows in parental sympatry. (A) Comparison of house and Spanish sparrow

genetic makeup using STRUCTURE. Each bar represents one individual, and the coloration of a bar represents the relative house

(white) and Spanish (black) genomic contribution, that is its admixture proportion. Numbers in lower left corners indicate sampling

localities shown in Fig. 1 except 6, which indicates F1 house X Spanish sparrow hybrids from aviary crosses. (B) BGC generated

Bayesian estimates of hybrid index (HI) for each individual in the focal sympatric population near Badajoz, Spain. HI estimates are

presented in ascending order, and 0 denotes pure house sparrow, whereas 1 denotes pure Spanish sparrow. The point estimates are

based on the median from the posterior distribution, and black lines indicate 95% credibility intervals. Colour of dots indicates the

species origin of the mitochondrial DNA of each individual; blue denoting house sparrow and red denoting Spanish sparrow origin.

(C) Frequency distribution of HI in parental sympatry.



populations, birds from Oslo, Norway (N = 58) and

Hradec Kr�alov�e, Czech Republic (N = 27) were used.

As representative Spanish sparrow populations, birds

from the Gargano Peninsula, Italy (N = 52) and Sardi-

nia, Italy (N = 48) were used. Known F1 aviary hybrids

between house (Oslo, Norway) and Spanish (Badajoz,

Spain) sparrows (N = 6) were also included to verify

that STRUCTURE assigned admixed individuals cor-

rectly. In STRUCTURE, the admixture model was used

and we assumed two groups (K = 2) with correlated

allele frequencies and ran 1 000 000 iterations after a

burn-in of 500 000 iterations.

Detection of SNPs associated with reproductiveisolation

To test for loci potentially associated with reproductive

isolation, we applied a genomic cline approach (Payseur

2010; Gompert & Buerkle 2011; Fitzpatrick 2013). This

approach takes advantage of the fact that introgression

can vary across the genome due to recombination in

admixed individuals. Genomic cline analysis quantifies

locus-specific introgression relative to a genome-wide

average as estimated by hybrid index or average gen-

ome-wide ancestry. We used genomic cline analysis to

identify loci exhibiting restricted introgression relative

to the genome-wide average, as such loci are candidates

for being associated with reproductive isolation. The

basic assumption is that alleles at loci responsible for

barriers to gene flow and neutral loci linked to such

selected loci will exhibit reduced introgression into the

foreign genomic background, due to selection against

unfit admixed individuals. This will result in steep and

possibly shifted genomic clines. By contrast, gene flow

at unlinked, neutral loci is not expected to be influenced

by such barriers. Neutral loci therefore exhibit shal-

lower clines that do not change more sharply with

changing hybrid index than a neutral expectation.

We used genomic cline analysis as implemented in

the BGC software (Gompert & Buerkle 2012). This

approach allows for the use of markers that are not

fixed in each parent species. BGC uses Markov Chain

Monte Carlo (MCMC) methods to obtain Bayesian esti-

mates of two parameters that describe the bias and rate

of locus-specific introgression into foreign genomic

background, based on ancestry (Gompert & Buerkle

2011, 2012). For a given locus, cline parameter a speci-

fies the probability of ancestry from one of the parental

species where an increase in probability produces a

positive parameter value and a decrease, a negative

parameter value. Cline parameter b, on the other hand,

specifies the rate of transition from low to high

probability of ancestry from the same parental species

as a function of hybrid index, with an increased rate

yielding a positive parameter value and a decreased

rate, a negative parameter value (Gompert & Buerkle

2012). a is therefore analogous to cline centre in geo-

graphic cline analysis, whereas b is analogous to cline

steepness. Locus-specific introgression differs from the

genome-wide average when credibility intervals for the

posterior probability distributions of a and b do not

overlap with zero (Gompert & Buerkle 2011, 2012).

As parental reference populations in the BGC analy-

ses, and hence to indicate parental allele frequencies,

we used house sparrows from Oslo, Norway (N = 58)

and Hradec Kr�alov�e, Czech Republic (N = 27), and

Spanish sparrows from the Gargano Peninsula, Italy

(N = 52). For the BGC analyses in parental sympatry,

all individuals from the sympatric parental population

were pooled into one admixed test population. Simi-

larly, for the reanalysis of the hybrid–parents data set,

all individuals from the Alps, the Italian peninsula

(excluding Spanish sparrows from the south-east Italian

sympatric zone), Sicily and Sardinia were pooled into

one admixed test population.

Genotypes at the mitochondrial locus ND2 were

coded as diploid homozygotes as BGC failed to run

with a haploid marker included. Ten independent runs

of 100 000 iterations each were run with the first 25 000

iterations discarded as burn-in, MCMC samples thinned

by recording every fifth value, while the rest of the

BGC settings were as default. We verified that the runs

gave qualitatively similar results, and for each system,

we present the results from the run with the best fit,

that is the run with the lowest mean negative log-likeli-

hood. SNPs were identified as significantly deviating

from genome-wide expectations when the 95% credibil-

ity intervals of the cline parameters a and b did not

cross zero. We also estimated the quantiles of the cline

parameters in the genome-wide distribution of intro-

gression, namely qa and qb. These quantiles represent

the position of each SNP on the posterior probability

distribution of the genome-wide average for each of aand b, respectively.The genomic cline model implemented in BGC

assumes no drift and that all loci have the same effec-

tive population size. As Z-linked loci have a lower

effective population size than autosomal markers (3/4),

this could lead to an overrepresentation of Z linkage of

steep clines simply as an effect of stronger drift at these

loci. To test for this, we also analysed the Z-linked and

autosomal loci in parental sympatry separately using

the same procedure as when analysing the full data set.

To compare the patterns of introgression between the

parent species in sympatry on the Iberian peninsula

with the introgression patterns between the Italian spar-

row and its parent species in Italy, we investigated

which markers exhibited significant excess ancestry



(significant a parameter) and/or steep clines (signifi-

cant, positive b parameter) in either or both systems.

We also estimated Pearson’s product–moment correla-

tion coefficients among systems for both cline parame-

ters (a and b) for the autosomal markers, for the

Z-linked markers, for the candidate intraspecific incom-

patibilities in the Italian sparrow and for the putative

reproductive isolation markers (see below). The correla-

tions were performed on point estimates based on the

median of the posterior distribution for a and b. Finally,we plotted the quantiles of the cline parameters in the

genome-wide distribution of introgression, qa and qb,to visualize where in the genome-wide distribution of

locus-specific introgression, the markers were situated.

Results

Patterns of introgression in parental sympatry

Our Bayesian admixture analysis revealed a pattern of

asymmetric introgression in the area of parental sympa-

try. The majority of individuals with house sparrow

phenotype showed signs of introgression when com-

pared to allopatric house sparrow populations (Fig. 2A).

Spanish sparrows, on the other hand, did not exhibit

signs of extensive introgression. They showed less intro-

gression than the insular Spanish sparrow population

on the Mediterranean island of Sardinia previously

shown to undergo gene exchange with Italian sparrow

migrants from mainland Italy (Fig. 2A; Trier et al. 2014).

With the aforementioned SNP ascertainment bias caveat

in mind, however, this apparent asymmetry in intro-

gression must be interpreted cautiously as it may to

some degree reflect an artefact of the SNP detection

strategy used in this study. Nonetheless, the relative

difference in introgression between allopatric and sym-

patric house sparrow populations as evident in Fig. 2A

is a signal of significant ongoing introgression in symp-

atry. Moreover, the distribution of hybrid indices esti-

mated using the Bayesian algorithm implemented in

BGC provided an informative basis for investigating

locus-specific introgression using genomic clines analy-

sis (Fig. 2B, C).

Genomic clines analysis in parental sympatry

Ten SNPs exhibited significant excess house sparrow

ancestry (negative a), whereas eight SNPs exhibited

excess Spanish sparrow ancestry (positive a) in parental

sympatry (Fig. S1, Supporting information). This indi-

cates that locus-specific introgression does not show the

same overall bias towards introgression into house

sparrows as previously described for admixture propor-

tion results from STRUCTURE. Furthermore, 19 SNPs

exhibited significantly steep clines (positive b), indicat-ing reduced introgression and hence a potential associa-

tion with reproductive isolation, whereas 10 SNPs

exhibited significantly shallow clines (negative b) (Fig.

S1, Supporting information). Of the 19 loci exhibiting

steeper clines than expected given the marker set aver-

age, 15 were Z-linked: a significant overrepresentation

of sex-linkage compared to the genome-wide expecta-

tion from other passerine birds (two-tailed binomial

test: null probability based on flycatcher genome (Elle-

gren et al. 2012) = 0.066, successes = 15, trials = 19,

P = 5.90 9 10�15). This remains a significant overrepre-

sentation of sex-linkage also when compared to the

proportion of sex-linked markers in our marker set

(Two-tailed binomial test: null probability based on pro-

portion of sex-linked marker in marker set = 0.273, suc-

cesses = 15, trials = 19, P = 4.10 9 10�6). One could

argue that this overrepresentation of Z-linked loci may

result from stronger drift due to lower effective popula-

tion size of Z-linked loci compared to autosomal loci,

and we cannot exclude the possibility that differences

in effective population size between Z and autosomes

may contribute to the observed pattern. However, our

independent analyses of Z-linked and autosomal loci

gave qualitatively the same results for the b parameter

as when all markers were analysed together (Z-linked

loci: N = 21, r = 0.964, P = 2.15 9 10�12, autosomal loci:

N = 55, r = 0954, P = 2.20.0 9 10�16, Fig. S2, Supporting

information). This indicates that there is a real overrep-

resentation of Z linkage among the candidate RI loci.

Genomic clines reanalysis of hybrid–parent data set

In our reanalysis of the hybrid–parents data set using a

different, larger, parental Spanish reference set than

previously, 26 SNPs exhibited significant excess house

sparrow ancestry, whereas 29 SNPs exhibited excess

Spanish sparrow ancestry. Furthermore, 13 SNPs exhib-

ited significantly steep clines, whereas 17 SNPs exhib-

ited significantly shallow clines (Fig. S3, Supporting

information). All Z-linked loci exhibiting steep clines at

the hybrid–parent range boundaries in Trier et al. (2014)

also exhibited significantly steep and shifted genomic

clines when reanalysing the hybrid–parent data set

(Figs 3 and 4; Fig. S3, Supporting information). That is,

the Z-linked candidate reproductive isolation loci

remained unchanged: CETN3 and CHD1Z at the Ital-

ian/house boundary and HSDL2, MCCC2 and GTF2H2

at the Italian/Spanish boundary. Autosomal candidate

marker RPS4 did not exhibit a significantly steep cline

in the reanalysis, however, but was the marker exhibit-

ing the strongest excess Spanish sparrow ancestry

(Figs 3 and 4; Fig. S3, Table S2, Supporting informa-

tion). In our previous study of hybrid–parent barriers,



(A) (B)

(C) (D)

Fig. 3 Comparison of locus-specific patterns of introgression between parental sympatry and hybrid–parent range boundaries. Dis-

played are cline estimates for the 77 markers used in our BGC analyses. Each symbol denotes the point estimate of the given parame-

ter based on the median of the posterior distribution. White symbols denote autosomal loci, black symbols denote Z-linked loci, grey

symbols denote putative internal incompatibilities in the Italian sparrow, red symbols denote the putative nuclear reproductive isola-

tion genes situated at hybrid–parent range boundaries, and the yellow symbol denotes mitochondrial marker ND2. The shape of the

symbol denotes the significance level in each comparison based on 95% credibility intervals of the posterior distribution. Squares

denote comparisons where the parameter estimates are significant in both parental sympatry and in the hybrid–parents analysis, tri-

angles denote comparisons where the parameter estimates are significant only in parental sympatry, diamonds denote comparisons

where estimates are significant only in the hybrid–parents analysis, circles denote comparisons where the parameter estimates do not

deviate from neutral expectations in any of the two systems, and finally, crosses denote where the estimates are significant in both

systems but in opposite direction. Numbers refer to the different putative RI markers as specified in panel A. In all panels, parame-

ters from Italy (hybrid–parents) are plotted against parameters from Spain (parental sympatry) (A) a (bias) parameter: positive and

negative values indicate bias in favour of Spanish sparrow and house sparrow alleles, respectively. Note different scale on axes. (B)

b (rate) parameter: positive values indicate restricted introgression, and negative values indicate elevated introgression. Note differ-

ent scale on axes. (C) qa parameter: values closer to 0 and 1 indicate increasing evidence for excess house or Spanish sparrows ances-

try, respectively. (D) qb parameter: values closer to 0 and 1 indicate increasing evidence for shallow and steep clines, respectively.



FST outlier analysis revealed RPS4 to be the strongest

candidate for being under directional selection in the

Italian-house hybrid zone in the Alps (Trier et al. 2014).

Moreover, analysis of spatial genetic variation using

GENELAND (Guillot et al. 2005a,b) showed that RPS4

exhibited a sharp shift from house to Spanish sparrow

genotypes in the Alps hybrid zone (Trier et al. 2014).

We thus retain RPS4 as a candidate RI locus between

Italian and Spanish sparrows when plotting the results

from the comparative genomic cline analysis as well as

in our correlation comparison of candidate hybrid–

parent RI loci. Furthermore, in our reanalysis, six loci

exhibited steep clines not situated at either hybrid–par-

ent boundary, compared with 15 such loci reported by

Trier et al. (2014). These six markers were, however, a

subset of the 15 loci that Trier et al. (2014) interpreted

as candidates for being unpurged intraspecific incom-

patibilities within the Italian sparrow.

Hybrid–parents intrinsic barriers represent a subset ofthe barriers isolating the parent species

All Z-linked candidate hybrid–parent RI loci (CETN3,

CHD1Z, HSDL2, MCCC2 and GTF2H2) were among the

19 loci exhibiting significantly steep clines in parental

sympatry (Figs 3 and 4; Figs S1 and S3, Supporting

information). Indeed, there was strong concordance

between the two systems in patterns of locus-specific

introgression for the candidate nuclear RI loci (Figs 3

and 4; Table 1, Supporting information, correlation

among systems for the a parameter for the nuclear

putative RI markers: N = 6, r = 0.986, P = 2.77 9 10�4;

correlation among systems for the b parameter for the

nuclear putative RI markers b: N = 6, r = 0.942,

P = 4.94 9 10�3).

Of the six candidate intraspecific incompatibility loci

in the Italian sparrow, three exhibited restricted intro-

gression also in parental sympatry (Fig. 3; Figs S1 and S3,

Table S2, Supporting information). Furthermore, more

Fig. 4 Comparison of genomic clines for loci associated with

hybrid–parent reproductive isolation in Italian sparrows. Geno-

mic clines for the hybrid–parents analysis in Italy (left panel)

and parental sympatry in Spain. Markers PIK3C3 (Z-linked)

and DYRK4 (autosomal) are shown as ‘control loci’ that do not

deviate from neutral expectations between hybrid and parent.

Significant excess house sparrow ancestry is indicated by H,

significant excess Spanish sparrow ancestry is indicated by S,

significantly steep clines are indicated by +, and significantly

narrow clines are indicated by �. For estimates not differing

from neutral expectations, this is indicated by ns. Red clines

denote loci with significant excess Spanish sparrow ancestry,

and blue clines denote loci with significant excess house spar-

row ancestry. Black clines are not significantly shifted in either

direction.



loci appeared to be involved in reproductive isolation in

parental sympatry than between hybrid and either parent

as ten loci exhibited steep clines only in parental sympa-

try (Fig. 3; Figs S1 and S3, Supporting information).

In parental sympatry, mitochondrial marker ND2

exhibited the same steep cline and excess house spar-

row ancestry as in Italian sparrows (Figs 3 and 4; Figs

S1 and S3, Supporting information). Similarly, Z-linked

nuclear-encoded mitochondrial markers HSDL2 and

MCCC2 also exhibited the same steep clines and excess

house sparrow ancestry in parental sympatry as at the

Italian-Spanish boundary (Figs 3 and 4; Figs S1 and S3;

Table S2, Supporting information).

Discussion

In this study, we investigated the evolution of intrinsic

reproductive isolation in a hybrid species system. We

compared loci exhibiting restricted introgression at the

range boundaries between the hybrid Italian sparrow

and its parent species, house and Spanish sparrows, to

loci exhibiting restricted introgression in an area where

the parent species co-occur and hybridize. The rationale

behind our approach is that if the hybrid–parent barri-

ers have arisen through sorting of pre-existing parental

incompatibilities, the same loci should also act as barri-

ers between the parents when they hybridize in sympa-

try. If the hybrid–parent barriers have arisen through

de-novo interactions in the hybrid genome, no such con-

cordance is expected (Rieseberg 1997).

Hybrid–parent intrinsic barriers represent a subset ofthose isolating the parent species

Consistent with the main prediction from the sorting

hypothesis, all five Z-linked candidate RI loci between

the Italian sparrow and either parent species were

among the 19 loci exhibiting restricted introgression in

parental sympatry. Moreover, these five loci all exhib-

ited similar patterns of excess ancestry in parental

sympatry as in the hybrid taxon, indicating that the

sorting of these incompatibilities in the Italian sparrow

was driven by a deterministic process. Autosomal mar-

ker RPS4 did not exhibit a significantly steep cline in

parental sympatry, however, suggesting that this

locus may be involved in a novel barrier that has devel-

oped between the hybrid Italian sparrow and the house

sparrow.

Our comparative analysis further revealed a set of

ten autosomal loci to exhibit restricted introgression

only in parental sympatry. This supports the prediction

from the sorting hypothesis that hybrid–parent RI genes

should represent a subset of parent–parent RI genes. It

further suggests that the association of these loci with

reproductive isolation is caused by epistasis through

linkage disequilibrium, which has been broken down

by recombination in the hybrid taxon. In sum, our

results support the hypothesis that the previously

reported intrinsic barriers between the Italian sparrow

and its parents have arisen through the sorting of pre-

existing parental incompatibilities. Allelic mosaicism of

hybrid species genomes has previously been demon-

strated in Helianthus sunflowers (Rieseberg et al. 1995,

2003) as well as in Lycaeides (Gompert et al. 2006; Nice

et al. 2013) and Papilio butterflies (Kunte et al. 2011) and

cichlids (Keller et al. 2013), but this is the first time

mosaicism of parental intrinsic incompatibility alleles is

demonstrated.

Not all narrow clines in the hybrid Italian sparrow

are located at hybrid–parent range boundaries, indicat-

ing a possible presence of unpurged intraspecific

incompatibilities (Trier et al. 2014). Such incompatibili-

ties may become coupled with extrinsic barriers and

this can set the stage for further diversification of the

hybrid taxon (Bierne et al. 2011). Findings consistent

with isolation-by-adaptation have previously been

reported from the Passer system, so such a coupling

mechanism may be at work (Eroukhmanoff et al. 2013).

In our comparative genomic cline analysis, three of the

six candidate intraspecific incompatibilities exhibited

restricted introgression also in parental sympatry. This

suggests that a subset of the candidate intraspecific

incompatibilities in the Italian sparrow may represent

unpurged parental incompatibilities, albeit further care-

ful investigation is needed to unveil their potential role

in further diversification.

The set of markers associated with restricted intro-

gression in parental sympatry exhibited a strong over-

representation of Z chromosome linkage. This finding

must, however, be interpreted with caution. We cannot

fully control for the lower effective population size of

Z-linked markers in our analyses, and hence this

Table 1 Pearson’s product–moment correlations between cline

parameters in different marker groups. The correlations were

performed on point estimates based on the median of the pos-

terior distribution for a and b

Marker group N

a-Parameter b-Parameter

r P r P

RI (nuclear) 6 0.986 2.77 9 10�4 0.942 4.94 9 10�3

II 6 0.617 0.192 0.030 0.956

Z-linked 21 0.558 8.56 9 10�3 0.240 0.295

Autosomal 55 0.426 1.17 9 10�3 0.485 1.77 9 10�4

RI refers to candidate reproductive isolation loci and II refers

to candidate unpurged intraspecific incompatibilities within

the Italian sparrow.



finding may to some degree be influenced by elevated

drift at Z-linked loci. However, the finding is consistent

with a growing number of studies demonstrating that

sex chromosome linkage plays a disproportionate role

in reproductive barriers between species with chromo-

somal sex determination systems (Qvarnstr€om & Bailey

2009), including in the Passer sparrows (Elgvin et al.

2011; Trier et al. 2014). The disproportionate role for sex

chromosomes in reproductive isolation is thought to

result from the higher mutation rate in the male germ

line, combined with exposure of recessive alleles to

selection in the heterogametic sex, an overrepresenta-

tion of genes with sexual function on the sex chromo-

somes, high average linkage between the sex-linked

genes involved in reproductive isolation, and stronger

drift due to a lower effective population size of sex-

linked markers (Charlesworth et al. 1987; Coyne & Orr

2004; Qvarnstr€om & Bailey 2009; Mank et al. 2010; Sætre

& Sæther 2010; Ellegren et al. 2012).

Mother’s curse influences isolation both between theparents and between hybrid and parent

‘Mother’s curse’ is the phenomenon by which selection

in males has no direct effect on mitochondrial fitness

due to the maternal inheritance of mitochondria,

favouring female (irrespective of male) interests, conse-

quently creating a selective sieve allowing for the build-

up of male-detrimental mutations (Frank & Hurst 1996;

Gemmel et al. 2004). This in turn selects for suppressor

alleles that restore male fitness. In female-heterogametic

taxa (ZZ/ZW), Z-linked genes spend two-thirds of their

time in the male lineage, and suppressor alleles are

therefore expected to be disproportionately situated on

the Z chromosome (Rice 1984; Trier et al. 2014).

Mismatches between mitochondrial alleles and nuclear

suppressor alleles lead to detrimental effects in hybrids

whose parents differ in mitonuclear system (Frank &

Hurst 1996; Gemmel et al. 2004).

In parental sympatry, we found both mitochondrial

marker ND2 and Z-linked nuclear-encoded mitochon-

drial markers HSDL2 and MCCC2 to exhibit the same

steep clines and shifts in favour of the house sparrow

allele as in Italian sparrows at the Italian–Spanish

boundary (Trier et al. 2014). The finding of strong house

sparrow excess ancestry for mitochondrial DNA also in

parental sympatry suggests that the fixation of house

rather than Spanish sparrow mitochondrial DNA in the

Italian sparrow was driven by deterministic rather than

stochastic processes. The mechanism favouring house

sparrow mitochondria remains unknown, but the candi-

dacy of HSDL2 and MCCC2 as Z-linked mother’s curse

suppressors is strengthened as these genes exhibit

among the steepest clines both at the Italian-Spanish

sparrow boundary and in parental sympatry, as well as

exhibiting significant shifts in the same direction as

mtDNA in both systems.

Conclusions

Our findings suggest that intrinsic barriers isolating the

Italian sparrow from its parent species have mainly

developed through the sorting of pre-existing sex-linked

parental incompatibilities and that isolation in both

cases is driven in part by mitonuclear conflict involving

the Z chromosome. This represents the first evidence

that the sorting process contributes to the persistence of

a hybrid animal taxon. It remains to be further investi-

gated whether and to what extent novel evolution in

the hybrid Italian sparrow also contributes to reproduc-

tive isolation against its parent species, but autosomal

marker RPS4 suggests that de-novo epistatic interactions

may also play a role in this system. Should a sorting

mechanism similar to the one described here prove to

be pervasive, the circumstances promoting homoploid

hybrid speciation may be broader than currently sus-

pected, and indeed, there may be many cryptic hybrid

taxa separated at two boundaries by sorted, inherited

incompatibilities, as in the Italian sparrow.

Acknowledgements

We thank M. H. Tu, M. Moan, P. Munclinger, M. F. G. Rojas,

S. A. Sæther as well as numerous field assistants for help with

acquiring the data, B. Dogan for help with laboratory work

and N. H. Barton, A. Runemark and anonymous referees for

helpful comments on a previous draft of the manuscript. This

work was supported by The Research Council of Norway,

Molecular Life Science (MLS), University of Oslo, and Centre

for Ecological and Evolutionary Synthesis (CEES), University

of Oslo.

References

Abbott RJ, Rieseberg LH (2012) Hybrid speciation. In: eLS. John

Wiley & Sons, Ltd, Chichester. DOI: 10.1002/9780470015902.

a0001753.pub2.

Abbott R, Albach D, Ansell S et al. (2013) Hybridization and

speciation. Journal of Evolutionary Biology, 26, 229–246.Bierne N, Welch J, Loire E, Bonhomme F, David P (2011)

The coupling hypothesis: why genome scans may fail to

map local adaptation genes. Molecular Ecology, 20, 2044–

2072.

Buerkle CA, Morris RJ, Asmussen MA, Rieseberg LH (2000)

The likelihood of homoploid hybrid speciation. Heredity, 84,

441–451.Charlesworth B, Coyne JA, Barton NH (1987) The relative rates

of evolution of sex chromosomes and autosomes. The Ameri-

can Naturalist, 130, 113–146.

Coyne JA, Orr HA (2004) Speciation, 1st edn. Sinauer Associ-

ates, Sunderland.



Elgvin TO, Hermansen JS, Fijarczyk A et al. (2011) Hybrid spe-

ciation in sparrows II: a role for sex chromosomes? Molecular

Ecology, 20, 3823–3837.

Ellegren H, Smeds L, Burri R et al. (2012) The genomic land-

scape of species divergence in Ficedula flycatchers. Nature,

491, 756–760.Eroukhmanoff F, Hermansen JS, Bailey RI, Sæther SA, Sætre

G-P (2013) Local adaptation within a hybrid species. Hered-

ity, 111, 286–292.

Falush D, Stephens M, Pritchard JK (2003) Inference of

population structure using multilocus genotype data: linked

loci and correlated allele frequencies. Genetics, 164, 1567–1587.Fitzpatrick BM (2013) Alternative forms for genomic clines.

Ecology and Evolution, 3, 1951–1966.Frank SA, Hurst LD (1996) Mitochondria and male disease.

Nature, 383, 224.

Gemmel NJ, Metcalf VJ, Allendorf FW (2004) Mother’s curse:

the effect of mtDNA on individual fitness and population

viability. Trends in Ecology and Evolution, 19, 238–244.

Gompert Z, Buerkle CA (2011) Bayesian estimation of genomic

clines. Molecular Ecology, 20, 2111–2127.

Gompert Z, Buerkle CA (2012) bgc: software for Bayesian esti-

mation of genomic clines. Molecular Ecology Resources, 12,

1168–1176.Gompert Z, Fordyce JA, Forister ML, Shapiro AM, Nice CC

(2006) Homoploid hybrid speciation in an extreme habitat.

Science, 314, 1923–1925.

Guillot G, Estoup A, Mortier F, Cosson JF (2005a) A spatial sta-

tistical model for landscape genetics. Genetics, 170, 1261–1280.

Guillot G, Mortier F, Estoup A (2005b) GENELAND: a com-

puter package for landscape genetics. Molecular Ecology

Notes, 5, 712–715.Heliconius Genome Consortium (2012) Butterfly genome reveals

promiscuous exchange of mimicry adaptations among spe-

cies. Nature, 487, 94–98.

Hermansen JS, Sæther SA, Elgvin TO et al. (2011) Hybrid

speciation in sparrows I: phenotypic intermediacy, genetic

admixture and barriers to gene flow. Molecular Ecology, 20,

3812–3822.

International Chicken Genome Sequencing Consortium (2004)

Sequence and comparative analysis of the chicken genome

provide unique perspectives on vertebrate evolution. Nature,

432, 695–716.

Jiggins CD, Salazar C, Linares M, Mavarez J (2008) Hybrid trait

speciation and Heliconius butterflies. Philosophical Transactions

of the Royal Society B: Biological Sciences, 363, 3047–3054.Keller I, Wagner CE, Greuter L et al. (2013) Population genomic

signatures of divergent adaptation, gene flow and hybrid

speciation in the rapid radiation of Lake Victoria cichlid

fishes. Molecular Ecology, 22, 2848–2863.Kunte K, Shea C, Aardema ML et al. (2011) Sex chromosome

mosaicism and hybrid speciation among Tiger Swallowtail

butterflies. PLoS Genetics, 7, e1002274.

Lai Z, Nakazato T, Salmaso M et al. (2005) Extensive chromoso-

mal repatterning and the evolution of sterility barriers in

hybrid sunflower species. Genetics, 171, 291–303.Mallet J (2007) Hybrid speciation. Nature, 446, 279–283.

Mank J, Nam K, Ellegren H (2010) Faster Z-evolution is pre-

dominantly due to genetic drift. Molecular Biology and Evolu-

tion, 27, 661–670.

Mavarez J, Linares M (2008) Homoploid hybrid speciation in

animals. Molecular Ecology, 17, 4181–4185.Meise W (1936) Zur Systematik und Verbreitungsgeschichte

der Haus- und Weidensperlinge, Passer domesticus (L.) und

hispaniolensis (T.). Journal f€ur Ornithologie, 84, 631–672.

Nice CC, Gompert Z, Fordyce CA et al. (2013) Hybrid specia-

tion and independent evolution in lineages of alpine butter-

flies. Evolution, 67, 1055–1068.Nosil P (2012) Ecological Speciation, 1st edn. Oxford University

Press, Oxford.

Payseur BA (2010) Using differential introgression in hybrid

zones to identify genomic regions involved in speciation.

Molecular Ecology Resources, 10, 806–820.

Pritchard JK, Stephens M, Donnelly P (2000) Inference of popu-

lation structure using multilocus genotype data. Genetics,

155, 945–959.Qvarnstr€om A, Bailey RI (2009) Speciation through evolution

of sex-linked genes. Heredity, 102, 4–15.Rice WR (1984) Sex chromosomes and the evolution of sexual

dimorphism. Evolution, 38, 735–742.Rieseberg LH (1997) Hybrid origins of plant species.

Annual Review of Ecology and Systematics, 28, 359–389.Rieseberg LH, Willis JH (2007) Plant speciation. Science, 317,

910–914.Rieseberg LH, Van Fossen C, Desrochers AM (1995) Hybrid

speciation accompanied by genomic reorganization in wild

sunflowers. Nature, 375, 313–316.

Rieseberg LH, Raymond O, Rosenthal DM et al. (2003) Major

ecological transitions in wild sunflowers facilitated by

hybridization. Science, 301, 1211–1216.

Sætre G-P, Sæther SA (2010) Ecology and genetics of speciation

in Ficedula flycatchers. Molecular Ecology, 19, 1091–1106.

Sætre G-P, Riyahi S, Aliabadian M et al. (2012) Single origin of

human commensalism in the house sparrow. Journal of Evolu-

tionary Biology, 25, 788–796.Schumer M, Rosenthal GG, Andolfatto P (2014) How com-

mon is homoploid hybrid speciation? Evolution, 68, 1553–1560.

Seehausen O, Butlin RK, Keller I et al. (2014) Genomics and the

origin of species. Nature Reviews Genetics, 15, 176–192.

Summers-Smith JD (1988) The Sparrows: A Study of the Genus

Passer, 1st edn. T & AD Poyser, Calton.

Trier CN, Hermansen JS, Sætre G-P, Bailey RI (2014) Evidence

for mito-nuclear and sex-linked reproductive barriers

between the hybrid Italian sparrow and its parent species.

PLoS Genetics, 10, e1004075.

Warren WC, Clayton DF, Ellegren H et al. (2010) The genome

of a songbird. Nature, 464, 757–762.

The study was conceived and designed by J.S.H. and

G.P.S. Fieldwork was carried out by J.S.H., F.H., C.N.T.,

A.M. and G.P.S. Bioinformatics work was carried out

by A.J.N. The data were analysed by J.S.H. with contri-

butions from C.N.T., R.I.B. and F.H. The study was

written by J.S.H. with contributions from R.I.B., G.P.S.

and the other authors.



Data accessibility

Genotype data for all individuals used in the presented

analyses, estimated cline parameters for all markers,

hybrid indices, BGC input files and code used for run-

ning BGC are available through Dryad (doi:10.5061/

dryad.v6f4d).

Supporting information

Additional supporting information may be found in the online ver-

sion of this article.

Fig. S1 Individual Bayesian genomic clines for all 77 markers

in parental sympatry analysis.

Fig. S2 Comparison of estimates of genomic cline steepness in

parental sympatry for Z-linked and autosomal loci analysed

separately and together with all markers.

Fig. S3 Individual Bayesian genomic clines for all 77 markers

in hybrid–parents analysis.

Table S1 Details of SNPs used in analyses.

Table S2 Estimates of a-parameter (‘cline shift’) and b-parame-

ter (‘cline steepness’) in hybrid–parents and parental sympatry

analyses for candidate RI markers and putative unpurged

intraspecific incompatibilities in the Italian sparrow.



Hybrid speciation through sorting of parental ...lithornis.nmsu.edu › ~phoude › Hermansen_et_al_2014...ences and structural rearrangements induced by recom-bination (Rieseberg

Documents