Molecular Ecology (2012) 21, 2655–2670 doi: 10.1111/j.1365-294X.2012.05557.

Host-jump drives rapid and recent ecological speciation

of the emergent fungal pathogen Colletotrichum kahawae
D I O G O N . S I L V A , * †, P E D R O T A L H I N H A S , * L E I C A I , ‡ L U Z O L O M A N U E L , § E L I J A H K .
G I C H U R U , – A N D R E I A L O U R E I R O , * V Í T O R V Á R Z E A , * O C T Á V I O S . P A U L O † and D O R A
BATISTA*
*CIFC ⁄ IICT—Centro de Investigação das Ferrugens do Cafeeiro ⁄ Instituto de Investigação Cientı́fica Tropical, Quinta do
Marquês, 2784-505 Oeiras, Portugal, †Computational Biology and Population Genomics Group, Centro de Biologia Ambiental,
Faculdade de Ciências da Universidade de Lisboa, Campo Grande, 1749-016 Lisboa, Portugal, ‡State Key Laboratory of
Mycology, Institute of Microbiology, Chinese Academy of Sciences, No.10, North 4th Ring Road West, Beijing 100190, China,
§Instituto Nacional do Café de Angola, Luanda, Angola, –Coffee Research Foundation, P.O. Box 4, 00232 Ruiru, Kenya

Abstract
Ecological speciation through host-shift has been proposed as a major route for the
appearance of novel fungal pathogens. The growing awareness of their negative impact
on global economies and public health created an enormous interest in identifying the
factors that are most likely to promote their emergence in nature. In this work, a
combination of pathological, molecular and geographical data was used to investigate the
recent emergence of the fungus Colletotrichum kahawae. C. kahawae emerged as a
specialist pathogen causing coffee berry disease in Coffea arabica, owing to its
unparalleled adaptation of infecting green coffee berries. Contrary to current hypotheses,
our results suggest that a recent host-jump underlay the speciation of C. kahawae from a
generalist group of fungi seemingly harmless to coffee berries. We posit that immigrant
inviability and a predominantly asexual behaviour could have been instrumental in
driving speciation by creating pleiotropic interactions between local adaptation and
reproductive patterns. Moreover, we estimate that C. kahawae began its diversification
at <2200 BP leaving a very short time frame since the divergence from its sibling lineage
(c. 5600 BP), during which a severe drop in C. kahawae’s effective population size
occurred. This further supports a scenario of recent introduction and subsequent
adaptation to C. arabica. Phylogeographical data revealed low levels of genetic
polymorphism but provided the first geographically consistent population structure of
C. kahawae, inferring the Angolan population as the most ancestral and the East African
populations as the most recently derived. Altogether, these results highlight the
significant role of host specialization and asexuality in the emergence of fungal
pathogens through ecological speciation.
Keywords: adaptation, BEAST, Coffea arabica, coffee berry disease, divergence population genetics,
plant pathogen
Received 18 August 2011; revision received 7 February 2012; accepted 16 February 2012.

tion are fundamental (Rundle & Nosil 2005; Schluter

Introduction
Ecological speciation is being increasingly recognized as
an important mechanism driving the origin of species,
in which the contributions of ecology and natural selec-
Correspondence: Diogo N. Silva, Fax: 00351 214 544 689;
E-mail: o.diogo.silva@gmail.com
2009). By definition, ecological speciation occurs when
reproductive isolation between populations arises as a
consequence of ecologically based divergent selection
promoting the fixation of different advantageous alleles
in each of the contrasting environments (Schluter &
Conte 2009). Host specialization can be a powerful
driver of divergent selection, promoting ecological

 2012 Blackwell Publishing Ltd


2656 D . N . S I L V A E T A L .
speciation via host-shift or host-jump, depending on
whether the involved hosts are genetically similar or
distant, respectively. Evidence is accumulating from
disparate taxa, such as phytophagous insects (Winkler
et al. 2009), coral dwelling fish (Munday et al. 2004)
and pathogenic fungi (Giraud et al. 2010), showing that
specialization to different hosts can effectively lead to
population subdivision, adaptation and divergence.
Host-shifts may represent a way by which incipient
pathogen species can escape from direct resource com-
petition, thus experiencing relatively high fitness even if
they are initially poorly adapted to the new environ-
ment (Dieckmann et al. 2004). The adaptation to the
new habitat can then drive the evolution of partial
reproductive isolation within a few generations (Hen-
dry et al. 2007). During this period, adaptive divergence
itself is likely to be of key importance in restricting gene
flow, thereby allowing the completion of speciation
regardless of the geographical distribution of the
diverging populations (Dieckmann et al. 2004). Indeed,
verbal and mathematical models suggest that when host
specialization and local adaptation are coupled with
assortative mating, gene flow can be severely or totally
restricted even in the absence of extrinsic barriers (Nosil
et al. 2005; Schluter & Conte 2009; Gladieux et al. 2011).
Moreover, if one considers the effects of pleiotropy, the
genetic associations between disruptively selected traits
and the mechanism of reproductive isolation do not
depend exclusively upon the establishment of linkage,
rendering ecological speciation likely and swift (Via
2001; Servedio et al. 2011).
Elucidating the biotic factors and environmental cir-
cumstances that are most likely to promote ecological
speciation in nature can be challenging using contem-
poraneous data, since time often obscures the condi-
tions at the moment of speciation (Coyne & Orr 2004).
In this regard, addressing emergent fungal diseases can
be particularly insightful, as their recent origins provide
a unique opportunity to study the mechanisms under-
lying ecological speciation (Stukenbrock & McDonald
2008). Fungal diseases have been emerging at an
increasing rate on a wide range of host plants, posing
tremendous threats to global economy and food safety
(Jones et al. 2008). The growing awareness of their
worldwide impact has been followed by the need of
understanding the events leading to the emergence of
such species (Stukenbrock & McDonald 2008). In addi-
tion, they also represent interesting models of specia-
tion mechanisms for two other reasons. First, as human
activities led to dramatic changes in the ecosystems at a
global scale, with vast areas cultivated with single
crops ⁄ genotypes and with the global exchange of germ-
plasm (Kareiva et al. 2007), diverse ecological opportu-
nities have been created for the rapid emergence and
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
transmission of novel fungal diseases (Stukenbrock &
McDonald 2008). Second, fungi possess several exclu-
sive and remarkable features that facilitate rapid eco-
logical speciation, especially through host-shift (Giraud
et al. 2010). In some examples of local adaptation to
their hosts, fungi revealed capable of: (i) undergoing
frequent asexual reproduction with rare events of sex-
ual recombination; (ii) generating a large number of
spores, which increases adaptive variation input by
mutation; or (iii) mating only within the host, creating
pleiotropic interactions between host specialization and
assortative mating (Giraud et al. 2010 and references
therein).
Herein, we aimed to elucidate the speciation event of
an emergent fungal pathogen, Colletotrichum kahawae,
from within the generalist and cosmopolitan Colletotri-
chum gloeosporioides complex. C. kahawae emerged as a
specialist pathogen causing coffee berry disease (pro-
duces anthracnose symptoms, with sunken necroses
leading to fruit drop or mummification) on Arabica cof-
fee (Coffea arabica), owing to its unique adaptation of
infecting green berries, an ecological niche previously
unoccupied by other fungi (Firman & Waller 1977).
Although C. arabica is grown in South America, Africa
and Asia, C. kahawae is still restricted to Africa where it
was first reported in 1922 in Kenya (McDonald 1926).
However, and despite the African origin of C. arabica,
the cultivation of this crop is also more recent in this
continent, with the first attempts in the 18th century but
most dating from the second half of the 19th century
(Bigger 2006). The coffee berry disease pathogen cur-
rently occurs in nearly all African regions where C. arab-
ica is grown, particularly above 1400 m, ravaging
plantations and causing up to 80% yield losses annually.
Nonetheless, very little is known about its origin besides
the information provided by historical data and field
reports. According to these sources, C. kahawae would
have originated in Kenya from C. gloeosporioides sensu
lato populations inhabiting Coffea spp. hosts (frequently
isolated from ripe or damaged coffee berries) either by:
(i) mutation from a mildly parasitic form in C. arabica
(Nutman & Roberts 1960); (ii) hybridization between
C. gloeosporioides strains from other Coffea spp. hosts
(Robinson 1976); or (iii) shifting from another Coffea
spp., where it would be present as an harmless fungal
strain (Robinson 1974). Currently, these hypotheses
remain untested, and the relationship between C. kaha-
wae and C. gloeosporioides s.l. isolates has been poorly
investigated, so that it is unclear how the specialization
to the new niche was accomplished. Some studies have
attempted to provide insights by investigating the
genetic variability and population structure of C. kaha-
wae, albeit with little success. Concordant with the puta-
tive recent origin and predominantly asexual
 2012 Blackwell Publishing Ltd

H O S T - J U M P S P E C I A T I O N O F C . K A H A W A E 2657
reproduction, genetic variability of C. kahawae was
found to be very low, preventing inferences about its
population structure, origin and spread (Bridge et al.
2008; Manuel et al. 2010).
The present study intends to assess the current
hypotheses for the emergence and spread of C. kahawae
and to investigate the process of host specialization,
using information from a multilocus sequencing
approach. We used an extensive sampling of C. kaha-
wae’s populations, as well as C. gloeosporioides s.l. iso-
lates from Coffea spp. and several other hosts
worldwide. In this endeavour, we made use of novel
statistical techniques on the inference of phylogenetic
relationships among species, estimation of divergence
times and reconstruction of past demographic history.
These analyses were coupled with information from
geographical modelling and pathogenicity tests to pro-
vide a multidisciplinary view of such speciation event.
Methods
Fungal material
A total of 102 isolates from the Colletotrichum gloeospo-
rioides complex were obtained from culture collections
and field samples, comprising three main groups: (i)
29 C. gloeosporioides s.l. isolates collected from Coffea
spp. hosts throughout plantations in South America,
Africa and Asia; (ii) 14 C. gloeosporioides s.l. isolates
collected from eleven host species (other than Coffea
spp.) and five geographical locations, previously
selected based on preliminary phylogenetic analyses
and culture availability; and (iii) 59 Colletotrichum kah-
awae isolates obtained from infected green Coffea arab-
ica berries in ten African countries, covering most of
its current range (Table S1, Supporting information).
Ex-holotypes (cultures obtained from the taxonomical
type strain) from three recently described species
within the C. gloeosporioides complex from C. arabica
hosts in Thailand were included: Colletotrichum sia-
mense, Colletotrichum asianum and Colletotrichum fructi-
cola (Prihastuti et al. 2009). Isolates from Citrus limon
and Olea europaea hosts were also used, representing
the C. gloeosporioides sensu stricto epitype (Cannon
et al. 2008) based on BLAST results of the complete
rDNA internal transcribed spacer (ITS) (GenBank
accession number EU371022 [E value = 0, Max
ident. = 98%]) and b-tubulin 2 (b-tub2) (GenBank
accession number FJ907445 [E value = 0, Max
ident. = 99%]). Five isolates of Colletotrichum fragariae
were selected as outgroup taxa for the phylogenetic
analyses. Culturing and DNA extraction from fungal
isolates were performed as previously described (Silva
et al. 2012).
 2012 Blackwell Publishing Ltd
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Molecular data
Six nuclear markers previously described in detail by
Silva et al. (2012) were employed for this study: the ITS
region, b-tub2, Apn25L, MAT1-2-1, ApMAT and MAT5L.
Primers and PCR conditions for ApMAT, MAT5L,
Apn25L and MAT1-2-1 were as described in the study by
Silva et al. (2012). PCR amplification of the ITS region
was carried out with primers ITS1Ext ⁄ ITS4Ext (Brown
et al. 1996) and b-tub2 using primers T1 ⁄ T2 (O’Donnell
& Cigelnik 1997). PCR products yielding one clear band
were purified using SURECLEAN (BioLine). When products
presented a multiband profile, the band of the expected
size was excised and purified using the Silica Bead DNA
Gel Extraction kit (Fermentas). Sequencing reactions
were performed using the BIGDYE version 3.1 chemistry
(Applied Biosystems) on an ABI prism 310 automated
sequencer. Amplicons were sequenced in both directions,
and chromatograms were manually checked for errors in
SEQUENCHER v4.0.5 (Gene Codes Corporation).
Phylogenetic analyses
Multiple sequence alignments were constructed for each
nuclear sequence data set in MAFFT v6.717b (Katoh &
Toh 2009), using the L-INS-i method. Individual data
sets were concatenated into a combined matrix using
the CONCATENATOR v1.1.0 software (Pina-Martins & Paulo
2008). We used maximum likelihood (ML) and a Bayes-
ian framework with Markov chain Monte Carlo
(BMCMC) to reconstruct phylogenies from the separate
and combined data sets. For ML analyses, MODELTEST
v3.7 (Posada & Crandall 1998) was used to select the
best fit model of DNA sequence evolution, under the
Akaike information criterion (AIC). ML heuristic
searches were performed in PAUP* v4.0d99 (Swofford
2003) with 100 replicates, random sequence addition
and a tree-bisection reconnection (TBR) branch swap-
ping algorithm. Nonparametric bootstrap was con-
ducted using 1000 pseudoreplicates with 10 random
additions and TBR branch swapping. The BMCMC
analysis was run in MRBAYES v3.1.2 (Ronquist & Huelsen-
beck 2003) with the optimal model of sequence evolu-
tion selected under the AIC, as implemented in
MRMODELTEST v2.3 (Nylander 2004). The MAT1-2-1 data
set was partitioned into codon positions because two
different substitution models were estimated for the
first and second positions combined and third position.
Posterior probabilities were generated from 1 · 107 gen-
erations, sampling at every 1000th iteration, and the
analysis was run three times with one cold and three
incrementally heated Metropolis-coupled Monte Carlo
Markov chains, starting from random trees. The

2658 D . N . S I L V A E T A L .
achievement of the stationary phase was checked using
TRACER V1.4 (Rambaut & Drummond 2007), and 1 · 106
generations were discarded as burn-in. Trees from dif-
ferent runs were then combined and summarized in a
majority rule 50% consensus tree.
As species assignment for most C. gloeosporioides s.l.
isolates was initially uncertain, we diagnosed poten-
tially distinct species in a phylogenetic context using
the multilocus genealogical concordance approach (Tay-
lor et al. 2000). Phylogenetic species were required to
meet the following criteria: (i) monophyly, (ii) strong
phylogenetic support (e.g. bootstrap, posterior probabil-
ities) in the multilocus analysis and (iii) genealogical
concordance, as conflict between gene trees could be
interpreted as recombination among individuals within
a species (Taylor et al. 2000).
Divergence dating of the species tree and demographic
analysis
The Bayesian MCMC analysis implemented in *BEAST
(Heled & Drummond 2010), an extension of the BEAST
v.1.6.1 software package (Drummond & Rambaut 2007),
was used to joint estimate the species tree and diver-
gence times from our multilocus data set. Preliminary
simulations detected a substantial heterogeneity in the
substitution rates of the different sequenced regions,
rendering inappropriate the attribution of a single clock
model. Thus, we analysed eight partition schemes of
increasing complexity, ranging from a single partition
of substitution and clock models to the partitioning of
the entire data set in eight unlinked substitution models
and six unlinked clock models, according to gene ⁄ inter-
genic positions (Table S2, Supporting information). The
best fit model of sequence evolution for each partition
was estimated using MODELTEST v3.7 (Posada & Crandall
1998), under the AIC. We used a birth-and-death pro-
cess as species tree prior and a piecewise linear and
constant root model for population size. Preliminary
runs using an uncorrelated lognormal relaxed clock
revealed that the rr (‘CoefficientOfVariation’) parameter
value for the ITS partition was consistently <0.3 with a
frequency histogram abutting 0, suggesting that this
partition does not significantly deviate from a strict
clock. Therefore, where applicable in our partition
schemes, we used a strict clock for the ITS partition and
uncorrelated lognormal molecular clocks for all other
partitions.
Genetic distances were converted into geological time
units using an averaged substitution rate of 8.8 · 10)9
per bp per year corresponding to the median of esti-
mates for several nuclear genomic regions obtained by
Kasuga et al. (2002). We fixed the substitution rate of
ITS with this value, as it was the only genomic region
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
from our data set included in the calibration study and
estimated the rates for all other partitions using uni-
form priors [0, 1]. The final results from the *BEAST
MCMC analyses were obtained by combining two inde-
pendent runs with 1 · 108 generations each, sampling
parameters every 10 000 iterations, using LOGCOMBINER
v1.6.1 (Drummond & Rambaut 2007). TRACER v1.4 (Ram-
baut & Drummond 2007) was used to assess conver-
gence and mixing for all parameters by visually
inspecting the trace log and estimating the effective
sample size (ESS) for each parameter; all ESS values
were >200 indicating adequate mixing. A conservative
10% of each analysis was discarded as burn-in.
To objectively select the most appropriate partition,
we tested our partition schemes with a Bayes factors
(BF) analysis. Following a previous study (Brown &
Lemmon 2007), we considered values for the test statis-
tic 2ln(BF)12 between model 1 and model 2 above 10 to
be evidence of significant support for model 2, values
between 10 and )10 to indicate ambiguity and values
below 10 to indicate strong support for model 1. When
faced with ambiguity, we opted for the simpler model.
BF were determined by calculating the marginal likeli-
hood for each scheme using TRACER v1.4 (Rambaut &
Drummond 2007).
The demographic history of C. kahawae’s populations
was reconstructed using the Gaussian Markov random
fields (GMRF) Bayesian Skyride (Minin et al. 2008),
implemented in BEAST v1.6.1 (Drummond & Rambaut
2007). The GMRF Bayesian Skyride was only specified
for the most appropriate partition previously selected
using BF. The analysis was run twice for 1 · 108 genera-
tions, sampling at every 10 000 generations after an ini-
tial burn-in of 10%. As in the *BEAST analysis, the
performance of the MCMC procedure was evaluated in
TRACER v1.4 (Rambaut & Drummond 2007), which was
also used to create the Bayesian Skyride plot with the
median and corresponding credibility intervals of the
estimated demographic parameters.
Population genetics analyses in Colletotrichum
kahawae and UG1
According to the phylogenetic analyses described
above, a group of C. gloeosporioides s.l. isolates very
close to C. kahawae was detected (Fig. 1), here named
Undescribed group 1 (UG1). To better understand their
relationship, median-joining haplotype networks were
constructed for each locus using the program NETWORK
v4.6.0.0 (Bandelt et al. 1999). DNA polymorphism statis-
tics, measures of divergence using the net divergence
statistic (Da, Nei 1987) and analyses of recombination
based on the four-gamete test (Hudson & Kaplan 1985)
were computed in DNASP v5.0 (Rozas et al. 2003). To
 2012 Blackwell Publishing Ltd
 1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
H O S T - J U M P S P E C I A T I O N O F C . K A H A W A E 2659

Fig. 1 50% majority rule Bayesian tree with the concatenated six-marker data set illustrating the evolutionary relationships between
Colletotrichum kahawae and Colletotrichum gloeosporioides s.l. from coffee and other hosts. Isolates from Coffea spp. hosts are followed
by a black bar, while isolates from other hosts are followed by a grey bar and a code, for which a key is provided in the right. The
geographical origin of the isolates is also provided underlined. The tree was rooted with C. fragariae. C. kahawae is represented by a
single clade for clarity with the detailed phylogenetic reconstruction provided in Fig. 5.

assess the degree of genetic differentiation, F-statistics
were calculated for all molecular markers, as imple-
mented in ARLEQUIN (Excoffier et al. 2005), and the Snn
statistic (Hudson 2000) was estimated after 1000 permu-
tations.
Isolation-with-migration analyses
Before investigating the history of isolation between
these two groups, the standard neutral model was tested
for each marker using Tajima’s D and Fu and Li’s D* and
F*, as implemented in DNASP v5.0. We used the program
isolation-with-migration (IMa) (Hey & Nielsen 2004) to
assess whether an isolation with migration model would
fit the data significantly better than a strict isolation
model and to estimate population mutation parameters
(hCk, hUG1 and ancestral hA), migration rates (mCk>UG1
and mUG1>Ck) and the time since the divergence of the
ancestral population of C. kahawae and UG1. Because
IMa supports multilocus data sets with different molecu-
lar rates, the concatenated data set of six nuclear markers
was used to carry out the estimation of parameters.
Given the low sequence polymorphism and absence of
sites with multiple substitutions, we applied the infinite-
site model of sequence evolution for all loci. Multiple
preliminary runs were performed to determine the
appropriate prior upper bounds for the parameters and
to assess the most efficient heating scheme for the
Metropolis-coupled MCMC run. After these pilot runs,
we performed four independent runs with 100 Metropo-
lis-coupled chains with geometric heating (increment
parameters: g1 = 0.95; g2 = 0.5) for 5 · 105 steps, sam-
pling at every 10th iteration, after a burn-in period of
1 · 106 generations. Overall, 2 · 105 generations were
saved and combined for the final estimation of parame-
ters by executing the program in L-mode. The L-mode of
IMa was also used to estimate joint parameter distribu-
tions and to examine nested models with a reduced num-
ber of parameters, representing alternative hypothesis of
gene flow patterns. Hypothesis testing was performed by
comparing the ln-likelihood ratio (LLR) statistic between
two models to a v2 distribution with k degrees of free-
dom, where k is the difference in the number of parame-
ters between models (Hey & Nielsen 2007). To convert
parameter estimates into biologically meaningful units,
we used the per locus substitution rate of the ITS parti-
tion (1.369 · 10)6), obtained from the same averaged sub-
stitution rate per site per year as in the *BEAST analyses.

 2012 Blackwell Publishing Ltd

 1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2660 D . N . S I L V A E T A L .

conidia suspension ⁄ berry from each fungal isolate. For

Phylogeography of Colletotrichum kahawae
The relationships and structure of C. kahawae’s popula-
tions were assessed by ML and BMCMC methods as
described above. For each of the polymorphic sites
within C. kahawae, the ancestral state was estimated
using SNAP Workbench (Aylor et al. 2006). Relevant
alternative topologies were tested to assess the robust-
ness of the unconstrained inference of population’s rela-
tionships, using the topological test (SH) of Shimodaira
& Hasegawa (1999). One thousand replicates were per-
formed by resampling the partial likelihoods for each
site (RELL model). Using the IDRISI 15 ANDES software
(Eastman 2006), a topographical map of a partial region
of the African continent was modelled to highlight
regions above 1400 m of altitude and to assess their cor-
relation with C. kahawae’s haplotype distribution.
Pathogenicity assays
To assess the pathogenicity of Colletotrichum isolates
from other hosts on Arabica coffee fruits, inoculations
of fungal strains were carried out on green and ripe
detached berries. Based on the phylogenetic analyses
conducted above, all isolates from UG1 and Unde-
scribed group 2 (UG2) phylogenetic groups (Fig. 1)
were included, plus four C. kahawae isolates as positive
controls (Mal2, Cam1, Ang29 and Que2). For each iso-
late, 106 spore ⁄ mL spore suspensions were obtained
from 7-day-old sporulating cultures in Malt Extract
Agar medium (Cultimed, Panreac Quimica). Berries
were collected from coffee plants maintained at CIFC
greenhouses, washed three times with distilled water
and blotted dry with a sterilized paper. Two trials were
then undertaken: (i) 10 Arabica green coffee berries of
genotype CIFC-7963 and (ii) 10 Arabica ripe coffee ber-
ries of genotype CIFC-H420 ⁄ 2 susceptible to coffee
berry disease were inoculated with 10 lL droplets of
both trials, an additional 10-berry set was mock-inocu-
lated with 10 lL sterile water ⁄ berry as a negative con-
trol. After inoculation, berries were incubated for 24 h
in the dark in a moisture chamber at 25 C and then
maintained under these conditions but with a 12-h pho-
toperiod. Symptoms were scored at the 7th and 15th
days after inoculation, and berries were classified either
as showing coffee berry disease symptoms (1) or as
asymptomatic (0).
Results
Evolutionary relationships of Colletotrichum spp. from
coffee and other hosts
The evolutionary relationships between Colletotrichum
kahawae and Colletotrichum gloeosporioides s.l. isolates
were established by constructing a multilocus concate-
nated phylogeny of 102 samples of Colletotrichum from
coffee and other hosts collected worldwide. Using a six-
marker (ITS, b-tub2, ApMAT, Apn25L, MAT1-2-1 and
MAT5L) sequencing approach, we achieved a substan-
tial resolution of species relationships. A summary of
the phylogenetic information for each individual and
combined molecular markers is provided in Table 1.
Parsimony informative content within individual
nuclear sequences ranged from low (ITS: 3%; b-tub2:
13%) to moderate (Apn25L: 20%; MAT5L: 18%; MAT1-
2-1: 17%) and high (ApMAT: 37%). The combined data
set consisted of 3783 bp of sequence data, with 764 par-
simony informative sites (20%).
Phylogenetic reconstructions of the concatenated data
set using both ML and BMCMC methods yielded the
same topology with minor discrepancies in the support
for some nodes (Fig. 1). The phylogeny was mostly
resolved with high statistical support and revealed a
pattern of divergent evolution between three main lin-

Table 1 Summary of the phylogenetic information for the individual and combined nuclear regions used in this study

Nuclear region FL Indels NC PI %PI V Model* Model†

ITS 489 2 487 14 3 1 TrNef+I SYM+I

b-tub2 575 34 541 76 13 5 SYM+G SYM+G
Apn25L 883 36 847 172 20 32 GTR+G GTR+G
ApMAT 782 84 698 287 37 29 HKY+G HKY+G
MAT5L 213 6 194 38 18 11 TrN+G HKY+G
MAT1-2-1‡ 843 1 842 141 17 18 K81+I ⁄ GTR ⁄ SYM+G GTR+I ⁄ GTR ⁄ SYM
Combined 3783 130 3606 764 20 72 – –

ITS, internal transcribed spacer; FL, Fragment length; NC, nucleotide characters (bp); PI, Parsimony informative characters (bp); %PI,
% of Parsimony informative characters; V, variable uninformative characters (bp).
*Best fit model of DNA evolution, under the AIC, implemented in MODELTEST.
†
Best fit model of DNA evolution, under the AIC, implemented in MRMODELTEST.
‡
Models of DNA evolution correspond to 1st and 2nd codon position ⁄ 3rd codon position ⁄ intron.

 2012 Blackwell Publishing Ltd


H O S T - J U M P S P E C I A T I O N O F C . K A H A W A E 2661
eages derived from two ancestral splits. Since the split
from C. fragariae, the first lineage to diverge included
the C. kahawae clade, along with two closely related
groups of isolates (Da = 0.53%) from several hosts other
than Coffea spp. (UG1 and UG2). These two groups
include fungi comprising a very broad range of hosts
and geographical origins, as they were isolated from
fruits, twigs and leaf lesions or as leaf endophytes from
eleven host species belonging to different taxonomic
divisions and orders fairly distant from the taxonomic
placement of Coffea spp., in three continents (Fig. 1;
Table S1, Supporting information). The phylogenetic
proximity was particularly high between C. kahawae
and isolates from the UG1 group. According to the
genealogical concordance criteria alone, both groups
should be regarded as a single species, considering that
UG1 could not be recovered as a distinct monophyletic
group with sufficient statistical support neither in single
nor in multilocus phylogenetic reconstructions and that
C. kahawae could not be completely distinguished from
some UG1 isolates in four of six single-locus analyses
(Fig. S1, Supporting information). However, as UG1
and C. kahawae clearly represent distinct and mostly
differentiated biological entities (see Results, Pathoge-
nicity tests; Genetic divergence and differentiation
between C. kahawae and UG1), they will be considered
hereafter as two separate groups. Moreover, using the
combined data set, all C. kahawae isolates were mono-
phyletic and clearly distinguishable with high bootstrap
and posterior probability values (100 ⁄ 1). The remaining
C. gloeosporioides s.l. isolates clustered in a large mono-
phyletic group comprising the two remaining lineages.
The most anciently diverged lineage encompassed rep-
resentative isolates of C. gloeosporioides s.s. from lemon
and olive. The third main lineage was exclusively com-
posed by isolates from coffee hosts belonging to at least
four different species, which revealed a fairly high
divergence from C. kahawae (Da = 6.14%). In this line-
age, most of the unassigned C. gloeosporioides s.l. strains
grouped with the C. siamense ex-holotype in a large and
highly supported monophyletic clade, within which a
 2012 Blackwell Publishing Ltd
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
high degree of genealogical discordance was observed
(Fig. S1, Supporting information). We also found two
clonal samples (Ang100 and Ang101) that revealed to
be a distinct and well-supported new monophyletic
lineage (Undescribed group 3; UG3). Even though a for-
mal recognition of species boundaries within C. gloeo-
sporioides s.l. on coffee hosts is out of the scope of this
study, all of the diagnosed groups in Fig. 1 are in
accordance with the genealogical concordance phyloge-
netic species recognition guidelines. Only the ITS and
MAT5L partitions were unable to recover clear species
boundaries, most likely due to low nucleotide variabil-
ity and small sequence size, respectively.
Divergence time estimates of the species tree
and demographic analysis
From the eight partitioning schemes performed in
*BEAST, the BF analysis favoured the second most com-
plex model in which all six nuclear sequences presented
distinct substitution and clock models, with the MAT1-
2-1 data set further partitioned into intron and exon
regions with two codon partitioning, over similar mod-
els with three codon (2lnBF = 7) or without codon parti-
tioning (2lnBF = 37) (Table S3, Supporting information).
The generated maximum clade credibility tree (Fig. 2)
presented a similar topology to that produced by the
ML and BMCMC analyses of concatenated data for the
two first main lineages. However, species relationships
within the third main lineage were largely incongruent
between methods with respect to the relative phyloge-
netic position of taxa and statistical support. Unlike the
well-supported phylogeny obtained from the concate-
nated analyses, the *BEAST species tree revealed a much
weaker support for the relationships of these species
with C. fructicola, C. siamense and UG3 sharing a termi-
nal and nearly polytomic relationship, while C. asianum
was the most anciently diverged species.
The Bayesian MCMC procedure implemented in
*BEAST also allowed co-estimating the posterior distribu-
tion of the time to the most recent common ancestor
Fig. 2 Time-calibrated Bayesian species
tree resulting from the *BEAST analysis
based on the six-marker data set and on
the best partitioning scheme selected
from the Bayes factors analysis. Poster-
ior probabilities are given above the
respective node. Scale numbers corre-
spond to years before present.

2662 D . N . S I L V A E T A L .
(TMRCA), including 95% credibility intervals [highest
posterior density (HDP)], of any group of taxa. Thus, to
construct a time line of events, we estimated the
TMRCA of all C. kahawae isolates as well as the diver-
gence time between C. kahawae and three taxa sets of
particular interest: (i) all C. gloeosporioides s.l. isolates
from Coffea spp. hosts; (ii) all isolates from UG1; and
(iii) the closest phylogenetic representative of UG1, iso-
late Cg432 (Table 2). The TMRCA of C. kahawae was
estimated to be between 320 and 4784 BP [95% high
posterior density (HPD)] with a median of 2219 BP. This
species most likely diverged from the sampled C. gloeo-
sporioides s.l. species from coffee hosts around
403 610 BP (95% HPD, 164 000–686 000), contrasting
with the relatively shallow divergence of 11 547 BP
(95% HPD, 3574–21 735) from UG1. However, given
the lack of monophyly for all isolates in this group and
some heterogeneity in their phylogenetic distance to
C. kahawae (Da = 0.19–0.69%), we also estimated the
TMRCA between C. kahawae and the UG1 isolate
Cg432. This allowed a better approximation of the
divergence time between C. kahawae and the closest
non-C. kahawae relative, which was estimated to be
7840 BP (95% HPD, 2020–15 245).
The demographic history of C. kahawae’s populations
is depicted in Fig. 3 using a Bayesian Skyride plot
reconstruction. Setting the GMRF Bayesian Skyride as a
tree prior in the BEAST analysis resulted in older estima-
tions for the TRMCA of C. kahawae (mean: 10 598 BP;
95% HPD, 3207–19 889) and its divergence time from
the closest non-C. kahawae relative (mean: 26 578 BP;
95% HPD, 13 735–45 698). The time lag relative to the
*BEAST estimates was expected owing to methodological
differences between concatenation and species tree
methods (Heled & Drummond 2010), and thus, these
time estimates should be interpreted in relative terms.
The historical demography of C. kahawae revealed a
very recent and sharp drop in population size, without
significant evidence of recovery until the present day.
Table 2 Bayesian estimates of the time (in years) to the most
recent common ancestor (TMRCA) of four taxa sets of interest.
Values in parenthesis are 95% highest posterior densities inter-
vals. The divergence estimates were calculated using *BEAST
v1.6.1 with the best partitioning scheme previously selected
with a Bayes factors analysis
Taxa set TMRCA SD
Ck 2219 (320; 4784) 64
Ck & UG1 11 547 (3574; 21 735) 133
Ck & Cg432 7840 (2020; 15 245) 142
Ck & Cg from 403k (149k; 647k) 2.6k
Coffea spp.
SD, standard deviation of mean; ESS, effective sample size.
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
1.E8
1.E7
Population size
1.E6
1.E5
1.E4
0 25 000 50 000 75 000 100 000
Time (yrs)
Fig. 3 GMRF Bayesian Skyride plot depicting fluctuations in
the population size of Colletotrichum kahawae through time. The
x-axis is in units of years before present, and the y-axis is equal
to Nes. The thick black line is the median estimate, and the
grey dashed lines delimit the 95% highest posterior density.
The vertical dashed line represents the median of the time to
the most recent common ancestor between C. kahawae and the
closest non-C. kahawae relative.
Notably, the onset of the downfall in C. kahawae’s pop-
ulation size (c. 25 000 BP) approximately matches the
estimated time of divergence from the closest non-
C. kahawae relative, pointing to the occurrence of a bot-
tleneck only after their separation (Fig. 3). To assess the
potentially biasing effect of C. kahawae’s population
structure in the reconstruction of its demographic his-
tory, we also performed the same GMRF Bayesian sky-
ride analysis for each haplotypic group separately
(Angola, Cameroon and East Africa). The results were
similar across all data sets, suggesting that our demo-
graphic estimates were robust to the violation of the
panmitic population assumption owing to the genetic
structure of C. kahawae’s populations.
Genetic divergence and differentiation between
Colletotrichum kahawae and UG1
Haplotype networks constructed for each locus revealed
ESS a close relationship between C. kahawae and UG1
(Fig. 4). Full haplotypes were shared between some
428 UG1 isolates and C. kahawae in several individual mark-
1470 ers, although both groups could be completely distin-
743
guished in the Apn25L and MAT1-2-1 data sets.
2811
Table 3 summarizes several interspecific divergence
and differentiation statistics. Consistent with a very
recent separation, the net divergence was very low
 2012 Blackwell Publishing Ltd
 1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
H O S T - J U M P S P E C I A T I O N O F C . K A H A W A E 2663

Therefore, the hypotheses of selective neutrality and no

Fig. 4 Median-joining haplotype networks depicting the rela-
tionship between Colletotrichum kahawae and Undescribed
group 1 (UG1). Haplotypes are presented as pie charts with
size proportional to the number of individuals.
across all individual loci (Da = 0.03–0.35). Intriguingly,
of the 38 polymorphic sites among C. kahawae and
UG1, there was an absence of shared polymorphisms
and three polymorphic sites were fixed in the Apn25L
and MAT1-2-1 data sets. This suggests that while both
groups have separated quite recently, current levels of
gene flow should be low. Differentiation indexes, such
as FST and Snn estimates, reflect the same pattern as
they consistently show high and significant levels of
differentiation across all loci. Most of the polymorphic
sites were exclusive of the UG1 group (32), while only a
small fraction (3) was exclusive of C. kahawae, which
further supports the effective population decline of the
latter.
Isolation-with-migration analyses
For each marker data set, both Fu and Li’s D* and F*
and Tajima’s D tests were not significantly deviated
from 0 (P > 0.05; Table S4, Supporting information),
and no recombination events were observed within loci.
recombination within loci, both assumptions of the IMa
models, could not be rejected.
The approximate posterior density curves for the IMa
parameters that resulted from the IMa analyses are
shown in Fig. 5, and their ML estimates with respective
95% HPD intervals are provided in Table 4. Even
though there was no sufficient information in the data
to estimate the time since the split of the ancestral pop-
ulation of C. kahawae and UG1 parameter, all other
parameters had stronger signals in the data and
revealed a single major peak in their posterior distribu-
tions. The limiting information in the data as well as
the lack of shared polymorphisms between populations
may have also led to nonzero tails in the posterior dis-
tribution of hUG1, hA and mUG1>Ck that prevented a reli-
able estimation of the credibility intervals for these
parameters. Despite these limitations, the four replicate
runs gave similar ML estimates and posterior distribu-
tions for these parameters, suggesting that the program
was sampling from the stationary phase of the MCMC
run and that parameter values are robust and represen-
tative given their priors. To test the potential impact of
C. kahawae’s population structure on the estimation of
IMa parameters, we have performed additional runs for
each haplotypic group separately. As in the BEAST GMRF
Skyride analysis, the results were similar across data
sets suggesting a limited impact of C. kahawae’s popula-
tion structure on the estimation of parameters.
Estimates of current and ancestral effective popula-
tion sizes revealed a decline of circa 40-fold in the pop-
ulation size of C. kahawae compared to the ancestral
population size, suggesting that this species underwent
a population bottleneck since its separation from UG1.
Estimates of gene flow were highly asymmetrical. No
gene flow was detected from UG1 to C. kahawae, but
moderate and significant gene exchange was detected

Table 3 Divergence and differentiation between Undescribed group 1 (UG1) and Colletotrichum kahawae

Locus Da (%) V VF VS PCk* PUG1* FST† Snn‡ Rm‡

ApMAT 0.09 7 0 0 0 7 0.536*** 0.88596*** 0

Apn25L 0.12 13 1 0 0 12 0.868*** 0.93525*** 0
ITS 0.03 4 0 0 0 4 0.634** 0.88596*** 0
MAT1-2-1 0.31 7 2 0 1 4 0.845*** 1*** 0
MAT5L 0.35 2 0 0 0 2 0.913*** 0.96610*** 0
b-tub2 0.13 5 0 0 2 3 0.425** 0.84394*** 0
Total 0.14 38 3 0 3 32 0.757*** 1***

ITS, internal transcribed spacer; Da, Net divergence (Nei 1987); V, total number of polymorphic sites; Vf, number of fixed
polymorphisms; VS, number of shared polymorphisms.
*Number of exclusive polymorphisms for C. kahawae and UG1.
†
Differentiation statistics measured by FST (Excoffier et al. 2005) and Snn statistics (Hudson 2000).
‡
Minimum number of recombination events.
Statistical significance: **P < 0.05; ***P < 0.01.

 2012 Blackwell Publishing Ltd

 1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
2664 D . N . S I L V A E T A L .

Fig. 5 The posterior probabilities of

parameter estimates of the isolation-

0.11
6

0.15
with-migration model.
5

0.09
4

0.10
3

0.07
0.05
2
1

0.05
0
0

0.0 0.1 0.2 0.3 0.4 0.5 0 5 10 15 0 5 10 15

1.2

1.6
0.8

0.4
0.4

0.2
0
0

0 1 2 3 4 0 1 2 3 4

Table 4 Maximum likelihood estimates (MLE) and respective
95% HPD intervals for parameters of the isolation and migra-
tion model for the clade composed by Undescribed group 1
(UG1) and Colletotrichum kahawae (Ck)
hA h(Ck)
MLE 6.38 0.15
Lower 95% HPD 0 0.05
Higher 95% HPD 13.97 0.29
HPD, high posterior density.
from C. kahawae to UG1. However, it is important to
note that the full model did not fit the data significantly
better than any of the nested models compared, includ-
ing the strict isolation model (Table S5, Supporting
information). Thus, even though the full model reveals
an intriguing pattern of gene flow between C. kahawae
and UG1, our data could not reject the simpler hypoth-
esis that both groups have been isolated since their
divergence.
Phylogeography of Colletotrichum kahawae
Our results confirmed the extremely low genetic vari-
ability within the C. kahawae clade (p = 0.00076; S = 3).
However, three divergent haplotypes could be distin-
guished, named after their geographical location:
Angola, Cameroon and East Africa (Fig. 6). For the
three polymorphic sites within C. kahawae, ancestral
states were inferred from the C. gloeosporioides sampling
(Fig. 6a). The Angola haplotype presented a nucleotide
sequence identical to the inferred ancestral state, while
the Cameroon haplotype diverged by one nonsynony-
mous mutation at the MAT1-2-1 gene, which replaces a
serine for a proline residue. The East Africa haplotype
shared the same nonsynonymous mutation and
hUG1 mUG1>Ck mCk>UG1
diverged from the other haplotypes by two additional
5.45 0.88 1.45 intronic mutations at the b-tub2 marker. To assess the
1.09 0 0 robustness of this inference, we tested an alternative
12.29 2.73 2.59 topology, in which we constrained the East African
population as the most ancestral lineage and the Angola
population as the most derived. The likelihood score of
the resulting tree was worse than the unconstrained
topology, with a marginally significant P-value (SH
Test, P = 0.056). Moreover, we could not detect the
presence of migrant haplotypes within our C. kahawae
sampling. We further extended our phylogeographical
analysis by modelling a topographical map of a partial
region of Africa to highlight all regions above 1400 m
of altitude, and the result was embedded on Fig. 6b, as
rough grey areas. East African regions revealed a larger
extension of highlands, which coincides with the Great
Rift Valley area. Outside this area, highland regions are
sparsely distributed, mainly through South Africa,
Namibia, Angola and Cameroon, and isolated by long
distances of lowland regions.
Pathogenicity assays
In the pathogenicity test with green berries, only C. kah-
awae isolates produced necrotic and sunken lesions
characteristic of coffee berry disease symptoms on all
10 inoculated berries (10 ⁄ 10 berries scored ‘1’). Green
berries inoculated with isolates from UG1 and UG2

 2012 Blackwell Publishing Ltd

 1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
H O S T - J U M P S P E C I A T I O N O F C . K A H A W A E 2665

(a) (b)
Fig. 6 Phylogeographical patterns of Colletotrichum kahawae. (a) Detailed phylogenetic reconstruction of populations relationships
within the C. kahawae clade, using the combined six-marker data set. Bootstrap and posterior probability values are provided above
each branch. The three nucleotide combination provided below each branch represents the mutational events that occurred along the
evolution of the three populations. The combination of the Angola population was inferred as the ancestral state. Mutations are high-
lighted with an asterisk, and the source gene is provided. KA, nonsynonymous mutation; Ki, intronic mutations. (b) Geographical
distribution of the three divergent C. kahawae populations. Countries were highlighted with different colours corresponding to the
existing haplotypes. Rough grey areas (orange in color version) on the map represent regions above 1400 m of altitude.

revealed no necroses or any other symptoms (10 ⁄ 10 ber-
ries scored ‘0’). Conversely, neither C. kahawae nor iso-
lates from UG1 and UG2 groups were able to
successfully colonize ripe berries.
Discussion
To our knowledge, this study describes the first com-
prehensive analysis of the evolutionary relationships of
Colletotrichum kahawae with Colletotrichum gloeosporioides
s.l. isolates from diverse hosts worldwide, aiming at
enlightening the underlying speciation event. One of
the most striking findings was that none of the initial
hypotheses asserting that C. kahawae would have
emerged from a C. gloeosporioides s.l. gene pool from
Coffea spp. hosts could be supported by our results.
According to our phylogenetic analyses, C. kahawae
appears to be diverging from all C. gloeosporioides s.l.
isolates from coffee hosts since c. 403 000 BP, which is
inconsistent with a very recent origin by mutation and
adaptation (Nutman & Roberts 1960) or hybridization
(Robinson 1976) from these populations. The most nota-
ble inconsistency, however, derives from the existence
of a taxonomically unclassified phylogenetic group sam-
pled from several hosts other than Coffea spp. (UG1)
that is genetically similar to C. kahawae. Although unex-
pected, the close relationship of UG1 with C. kahawae as
well as its association with genetically distant hosts
clearly suggests an alternative scenario in which a host-
jump preceded the outbreak of coffee berry disease epi-
demics, although the diversity of hosts that this group
covers hamper any confident inference on the host
sources of this event. In Colletotrichum, host-shifts have
not been previously documented as a speciation driver,
but there is recent evidence that local adaptation and
host specialization have structured populations of some
species, such as C. cereale (Crouch et al. 2009). Notwith-
standing, a large amount of evidence has accumulated
showing that host-shift speciation, a particular case of
ecological speciation, is one of the main routes for the
emergence of fungal pathogens (Couch et al. 2005;
Stukenbrock & McDonald 2008; Zaffarano et al. 2008;
Giraud et al. 2010).
‘Rapid ecological speciation through host jump’
hypothesis
Considering the onset of C. kahawae’s diversification
and its divergence time from the closest non-C. kahawae
relative, C. kahawae and UG1 have only been separated
for an average of 5600 BP. This presents a remarkably
short period of time for speciation to occur and comes

 2012 Blackwell Publishing Ltd


2666 D . N . S I L V A E T A L .
in agreement with the observation that crop pathogens
can emerge rather swiftly in recent timescales (Stuken-
brock & McDonald 2008). However, the estimated time
period for the origin of C. kahawae is still older than
expected from historical data (Firman & Waller 1977)
and from the onset of the cultivation of Coffea arabica in
Africa (Bigger 2006). While C. kahawae could have
begun adapting to C. arabica in wild populations or on
an intermediate host, such as another Coffea spp., these
scenarios are not supported by our data. A more plausi-
ble explanation could be a downward bias in our molec-
ular rate calibration resultant from the time dependency
of mutation rates (Ho et al. 2005). It has been recently
demonstrated that mutation rates in recent lineages
accelerate exponentially towards the present, biasing
divergence time estimates to higher values when old
calibrations are employed (Ho et al. 2011). This would
leave room for fitting the more recent historical time
estimates in this molecular-derived scenario.
The close phylogenetic relationship between C. kaha-
wae and UG1 also raises the question of whether they
represent good species or simply structured populations
from a single species. Even though we lacked a solid
basis to consider both groups as separate species solely
from a phylogenetic standpoint, our results from the
pathogenicity tests showed that, unlike C. kahawae, iso-
lates from UG1 were unable to cause coffee berry dis-
ease on detached green berries and neither group was
able to infect ripe berries. Green Arabica coffee berries
thus seem to present an ecological source of divergent
selection to which C. kahawae successfully adapted,
while isolates from UG1 appear to suffer a sharp fitness
decrease that greatly compromises their survival. At the
molecular level, C. kahawae and UG1 also seem to rep-
resent well-differentiated groups according to a combi-
nation of significant and elevated differentiation
indexes across all studied loci and a complete segrega-
tion of polymorphic sites. Under the isolation with
migration model, migration estimates revealed an
absence of gene flow to C. kahawae but, unexpectedly, a
moderate amount of gene flow into UG1 was detected.
The transfer of genetic material from C. kahawae to UG1
after their divergence is intriguing because it implies
the occurrence of a sexual stage of C. kahawae in nature
as well as the existence of a suitable alternate host or
substrate where mating could occur, both of which are
yet to be reported (Firman & Waller 1977; Bridge et al.
2008). Nevertheless, the migration signal in our data
was not strong enough to allow the rejection of a strict
isolation model, although future work with a larger
sampling of the UG1 group and more variable loci is
still required. Overall, it is reasonable to conclude that
C. kahawae and UG1 represent ecologically distinct and
isolated groups that evolved significantly different path-
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ogenic abilities in a short period of time, from which
only C. kahawae should trigger major disease control
and biosecurity measures on C. arabica crops.
During the brief period that C. kahawae and UG1 were
separated, historical demographic reconstructions and
patterns of DNA polymorphism consistently showed a
sudden and severe drop in the effective population size
of C. kahawae, as might be expected following a host-
jump. For such a host-jump to occur, isolates from the
UG1 group must have been within the cruising range of
C. arabica during the initial stages of C. kahawae specia-
tion, even if the geographical distribution of both groups
became disjunct in latter stages. In such scenario, the
pattern of low genetic diversity in C. kahawae could
have been produced by the exertion of strong disruptive
selection during the first stages of adaptation to C. arab-
ica, coupled with the ability of the fungus to undergo
repeated cycles of asexual propagation. Asexual repro-
duction could greatly amplify new advantageous muta-
tions to very high frequencies along with the entire
genome by hitchhiking (Adolfatto 2001). This would
eliminate polymorphisms and maintain only the intact
genome of those individuals in the population having
the favoured mutations, evidencing the strong genetic
bottleneck and clonal pattern observed in C. kahawae as
well as the lack of shared polymorphisms with UG1.
Adaptation to a new host is generally most efficient
when gene flow of ancestral genes into the adapting
population is severely reduced or absent (Giraud et al.
2010), which may be difficult to achieve when the source
population lies within cruising range. Nonetheless, two
significant intrinsic barriers to gene flow may have
indeed evolved between C. kahawae and UG1 that can
explain their high differentiation and low levels of gene
flow. First, the process of host specialization itself can
give rise to an isolation mechanism from the earliest
stages of divergence in fungal populations (Gladieux
et al. 2010). Because most pathogenic Ascomycetes (Col-
letotrichum included; Cisar & TeBeest 1999) can only
mate on their host after mycelial development, any
genetic variation favouring an adaptation to C. arabica
can pleiotropically cause assortative mating and restrict
gene flow, as any unfit immigrant will be unable to grow
and reproduce in such environment (Giraud 2006;
Giraud et al. 2010). This means that the differential sur-
vivorship of ill-adapted UG1 and adapted C. kahawae
isolates on C. arabica constitutes a legitimate and signifi-
cant prezygotic reproductive barrier, recently named
immigrant inviability (Nosil et al. 2005). The unique
adaptation of C. kahawae could be viewed as a ‘magic
trait’ scenario (Gavrilets 2004), where assortative mating
arises as a by-product of host specialization. Under
strong selection, this barrier has already been shown to
quickly and significantly prevent gene flow in other fun-
 2012 Blackwell Publishing Ltd

H O S T - J U M P S P E C I A T I O N O F C . K A H A W A E 2667
gal pathogens undergoing adaptation to their hosts, such
as Venturia inaequalis’s sympatric populations from
apple varieties with and without the Vf resistance gene
(Gladieux et al. 2011). Second, the transition of C. kaha-
wae to a predominantly asexual reproductive strategy
could have been a major reproductive barrier, allowing
for multiple generations of selection for local adaptation
without the pernicious effect of recombination (Zeigler
1998; Giraud et al. 2010). Altogether, it is reasonable to
expect that the combination of immigrant inviability and
the predominantly asexual behaviour of C. kahawae
would have been rather effective in keeping populations
separated during the early stages of divergence. How-
ever, additional work with a strategic and focused sam-
pling of the UG1 group will be required to better
understand not only the geography of the latter stages in
C. kahawae speciation but also the mechanisms underly-
ing the adaptation of this fungal pathogen to C. arabica.
Phylogeography of Colletotrichum kahawae
The phylogeographical knowledge of C. kahawae can
provide important information that may help pinpoint-
ing potential regions for the occurrence of its source
population. In fact, our genetic analysis revealed the
first consistent and unambiguous population structure
of C. kahawae. Based on single nucleotide polymor-
phisms, three slightly divergent haplotypes highly cor-
related with their geographical distribution were found
(Angola, Cameroon and East Africa), confirming previ-
ous indications of an incipient geographical structuring
(Bridge et al. 2008). Strikingly, the ancestral state infer-
ence suggests that the Angola haplotype is the most
ancestral among the studied isolates, while those from
Kenya and the remaining East African countries clus-
tered together as the most derived lineage. This pro-
vides an alternative view for the geographical origin of
C. kahawae, as compared to the current understanding,
which follows the premise that C. kahawae co-evolved
with coffee hosts and is based on disease reports and
field observations. Arabica coffee occupies only a small
fraction of the plantations in Angola, mainly in the cen-
tral plateau, while 98% of coffee production derives
from C. canephora varieties. The first introductions of
C. arabica in this country occurred in the 18th century
(A. Mendes Gaspar, personal communication), and the
first reports of coffee berry disease date back to 1930
(Beynon et al. 1995), just 8 years after its discovery in
Kenya. However, we stress that information obtained
from historical data can be flawed because it is biased
towards regions where disease incidence is higher, and
where substantial scientific effort has been focused on
monitoring plant diseases. Moreover, as we suggested
above, the ancestral population of C. kahawae most
 2012 Blackwell Publishing Ltd
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
likely emerged from hosts other than Coffea spp., which
may circumvent the difficulties in reconciling our
results with historical data, as speciation via host-shift
can be rather swift. Adding the fact that the transport
of infected plant material has reached an unprecedented
global scale (Stukenbrock & McDonald 2008), this
greatly increases the likelihood for the hypothesis of a
host-jump of the ancestral population of C. kahawae into
the newly arrived C. arabica plants in Angola.
The population structure of C. kahawae also reveals no
evidence of migration between the geographical locations
of each haplotype. This may be explained by seldom
sequential introductions with subsequent geographical
isolation. Arabica coffee growing areas in Angola, Camer-
oon and East Africa are separated by extensive lowland
areas, which are not suitable for the pathogen or the host
(Firman & Waller 1977), thus representing a potentially
effective barrier to migration. In such scenario, bottleneck
events during these rare introductions can generate drift
pulses in each of the introduced populations, which
become genetically differentiated from each other while
retaining their source–introduction relationship (Estoup
& Guillemaud 2010). Accordingly, our results suggest
that after a hypothetical origin of the Angola population,
an introduction in the Cameroon followed and from there
to the East Africa countries, while each of the established
populations remained isolated in their respective high-
land areas. However, in invasion biology, evolutionary
scenarios are often characterized by small divergence
times, which may decrease the likelihood of identifying
the true source–sink relationship between populations
because of the stochasticity of the process (Estoup & Gu-
illemaud 2010). Moreover, given the vastness of coffee
plantations in the East African countries, particularly in
Ethiopia where C. arabica also occurs naturally, we can-
not exclude the presence of unsampled ancestral haplo-
types in these regions that could alter our inferences.
However, in such scenario, the exceeding number of
additional steps required to explain the current phylogeo-
graphical pattern renders this hypothesis unlikely. Thus,
even though our data set suggest an Angolan origin for
C. kahawae, sampling more isolates and polymorphic loci
will certainly provide a much more reliable and robust
insight on the evolution of C. kahawae’s populations.
Conclusions
Altogether, our work represents an important step in
the understanding of the evolutionary and speciation
history of Colletotrichum kahawae that will certainly have
implications and launch new directions on its future
research. Having found very little support for the
current understanding of these events, we postulate an
alternative hypothesis of rapid ecological speciation by

2668 D . N . S I L V A E T A L .
a very recent host-jump and subsequent host specializa-
tion to explain the emergence of C. kahawae on Arabica
coffee. Two far-reaching and intriguing implications of
this hypothesis are that new and severe fungal patho-
gens are indeed able to successfully adapt and speciate
in a remarkably short time span, particularly when dri-
ven by divergent natural selection, and that intrinsic
and common biological traits of fungal populations
may greatly facilitate their emergence. Our results also
highlight the power and value of molecular data when
inferring dissemination patterns of emerging pathogens.
Unlike the limited information retrieved from historical
data, the population structure of C. kahawae revealed an
alternative and more objective centre of origin and that
the topography of the African continent may have had
a pivotal role in shaping and limiting its dispersal.
Acknowledgements
At FCUL, we thank our colleagues, Ana Vieira and Tiago Jesus,
for discussions and constructive criticisms during the elabora-
tion of this manuscript. At CIFC ⁄ IICT, we appreciate the techni-
cal support provided by Sandra Sousa Emı́dio. For supplying
additional and valuable isolates for this work, we are also in
debt to Ana Paula Ramos, at ISA ⁄ UTL, Portugal, and Peter
Johnston and Bevan Weir, at Landcare Research, New Zealand.
Lei Cai acknowledges grants CAS-KSCX2-EW-J-6 ⁄ NSFC
31070020. This work was financially supported by the portu-
guese Foundation for Science and Technology (FCT) and IICT,
Ministério da Ciência, Tecnologia e Ensino Superior, Portugal.
References
Adolfatto P (2001) Adaptive hitchhiking effects on genome
variability. Current Opinion in Genetics & Development, 11,
635–641.
Aylor DL, Price EW, Carbone I (2006) SNAP: combine and
map modules for multilocus population genetic analysis.
Bioinformatics, 22, 1399–1401.
Bandelt HJ, Forster P, Röhl A (1999) Median-joining networks
for inferring intraspecific phylogenies. Molecular Biology and
Evolution, 16, 37–48.
Beynon S, Coddington A, Lewis BG, Várzea V (1995) Genetic
variation in the Coffee Berry Disease pathogen,
Colletotrichum kahawae. Physiological and Molecular Plant
Pathology, 46, 457–470.
Bigger M (2006) The dissemination of coffee cultivation
throughout the world. Tropical Agriculture Association
Newsletter, 26, 15–19.
Bridge PD, Waller JM, Davies D, Buddie AG (2008) Variability of
Colletotrichum kahawae in relation to other Colletotrichum
species from tropical perennial crops and the development of
diagnostic techniques. Journal of Phytopathology, 156, 274–280.
Brown JM, Lemmon AR (2007) The importance of data
partitioning and the utility of Bayes factors in Bayesian
phylogenetics. Systematic Biology, 56, 643–655.
Brown A, Sreenivasaprasad S, Timmer L (1996) Molecular
characterization of slow-growing orange and key lime
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
anthracnose strains of Colletotrichum from citrus as
C. acutatum. Phytopathology, 86, 523–527.
Cannon PF, Buddie AG, Bridge PD (2008) The typification of
Colletotrichum gloeosporioides. Mycotaxon, 104, 189–204.
Cisar C, TeBeest D (1999) Mating system of the filamentous
ascomycete, Glomerella cingulata. Current Genetics, 35, 127–
133.
Couch BC, Fudal I, Lebrun M-H et al. (2005) Origins of host-
specific populations of the blast pathogen Magnaporthe oryzae
in crop domestication with subsequent expansion of
pandemic clones on rice and weeds of rice. Genetics, 170,
613–630.
Coyne JA, Orr HA (2004) Speciation. Sinauer Associates, Inc.,
Sunderland, Massachusetts.
Crouch JA, Tredway L, Clarke B, Hillman B (2009)
Phylogenetic and population genetic divergence correspond
with habitat for the pathogen Colletotrichum cereale and allied
taxa across diverse grass communities. Molecular Ecology, 18,
123–135.
Dieckmann U, Doebeli M, Metz JAJ, Tautz D (2004) Adaptive
Speciation. Cambridge University Press, Cambridge,
Massachusetts.
Drummond AJ, Rambaut A (2007) BEAST: Bayesian
evolutionary analysis by sampling trees. BMC Evolutionary
Biology, 7, 214.
Eastman JR (2006) IDRISI Andes. Clark University, Worcester,
Massachusetts.
Estoup A, Guillemaud T (2010) Reconstructing routes of
invasion using genetic data: why, how and so what?
Molecular Ecology, 19, 4113–4130.
Excoffier L, Laval G, Schneider S (2005) Arlequin ver. 3.0: an
integrated software package for population genetics data
analysis. Evolutionary Bioinformatics Online, 1, 47–50.
Firman I, Waller J (1977) Coffee Berry Disease and other
Colletotrichum diseases of coffee. Phytopathological Papers, 20,
1–53.
Gavrilets S (2004) Fitness Landscapes and the Origin of Species.
Princeton University Press, Princeton, New Jersey.
Giraud T (2006) Selection against migrant pathogens: the
immigrant inviability barrier in pathogens. Heredity, 97, 316–
318.
Giraud T, Gladieux P, Gavrilets S (2010) Linking the
emergence of fungal plant diseases with ecological
speciation. Trends in Ecology and Evolution, 25, 387–395.
Gladieux P, Caffier V, Devaux M, Le Cam B (2010) Host-
specific differentiation among populations of Venturia
inaequalis causing scab on apple, pyracantha and loquat.
Fungal Genetics and Biology, 6, 511–521.
Gladieux P, Guérin F, Giraud T et al. (2011) Emergence of
novel fungal pathogens by ecological speciation: importance
of the reduced viability of immigrants. Molecular Ecology, 20,
4521–4532.
Heled J, Drummond AJ (2010) Bayesian inference of species
trees from multilocus data. Molecular Biology and Evolution,
27, 570–580.
Hendry AP, Nosil P, Rieseberg LH (2007) The speed of
ecological speciation. Functional Ecology, 21, 455–464.
Hey J, Nielsen R (2004) Multilocus methods for estimating
population sizes, migration rates and divergence time, with
application to the divergence of Drosophila pseudoobscura and
D. persimilis. Genetics, 167, 747–760.
 2012 Blackwell Publishing Ltd

H O S T - J U M P S P E C I A T I O N O F C . K A H A W A E 2669
Hey J, Nielsen R (2007) Integration within the Felsenstrein
equation for improved Markov chain Monte Carlo methods
in population genetics. Proceedings of the National Academy of
Sciences, USA, 104, 2785–2790.
Ho SYW, Phillips MJ, Cooper A, Drummond AJ (2005) Time
dependency of molecular rate estimates and systematic
overestimation of recent divergence times. Molecular Biology
and Evolution, 22, 1561–1568.
Ho SYW, Lanfear R, Bromham L et al. (2011) Time-dependent
rates of molecular evolution. Molecular Ecology, 20, 3087–
3101.
Hudson RR (2000) A new statistic for detecting genetic
differentiation. Genetics, 155, 2011–2014.
Hudson RR, Kaplan N (1985) Statistical properties of the
number of recombination events in the history of a sample
of DNA sequences. Genetics, 111, 147–164.
Jones KE, Patel NG, Levy MA et al. (2008) Global trends in
emerging infectious diseases. Nature, 451, 990–993.
Kareiva P, Watts S, McDonald R, Boucher T (2007)
Domesticated nature: shaping landscapes and ecosystems for
human welfare. Science, 103, 1866–1869.
Kasuga T, White TJ, Taylor JW (2002) Estimation of nucleotide
substitution rates in Eurotiomycete fungi. Molecular Biology,
19, 2318–2324.
Katoh K, Toh H (2009) Recent developments in the MAFFT
multiple sequence alignment program. Briefings in
Bioinformatics, 9, 286–298.
Manuel L, Talhinhas P, Várzea V, Neves-Martins J (2010)
Characterization of Colletotrichum kahawae isolates causing
coffee berry disease in Angola. Journal of Phytopathology, 158,
310–313.
McDonald J (1926) A preliminary account of a disease of green
coffee berries in Kenya. Transactions of the British Mycological
Society, 11, 145–154.
Minin VN, Bloomquist EW, Suchard MA (2008) Smooth
skyride through a rough skyline: Bayesian coalescent-based
inference of population dynamics. Molecular Biology and
Evolution, 25, 1459–1471.
Munday PL, Herwerden LV, Dudgeon CL (2004) Evidence for
sympatric speciation by host shift in the sea. Current Biology,
14, 1498–1504.
Nei M (1987) Molecular Evolutionary Genetics. Columbia
University Press, New York.
Nosil P, Vines TH, Funk DJ (2005) Reproductive isolation
caused by natural selection against immigrants from
divergent habitats. Evolution, 59, 715–719.
Nutman F, Roberts F (1960) Investigations on a disease of
Coffea arabica caused by a form of Colletotrichum coffeanum
Noack. I. Some factors affecting infection by the pathogen.
Transactions of the British Mycological Society, 43, 489–505.
Nylander JAA (2004) MrModeltest v2. Program distributed by
the author. Evolutionary Biology Center, Uppsala University.
O’Donnell K, Cigelnik E (1997) Two divergent intragenomic
rDNA ITS2 types within a monophyletic lineage of the
fungus Fusarium are nonorthologous. Molecular Phylogenetics
and Evolution, 7, 103–116.
Pina-Martins F, Paulo OS (2008) Concatenator: sequence data
matrices handling made easy. Molecular Ecology Resources, 8,
1254–1255.
Posada D, Crandall KA (1998) Modeltest: testing the model of
DNA substitution. Bioinformatics, 14, 817–818.
 2012 Blackwell Publishing Ltd
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Prihastuti H, Cai L, Chen H, McKenzie E, Hyde KD (2009)
Characterization of Colletotrichum species associated with
coffee berries in northern Thailand. Fungal Diversity, 39, 89–
109.
Rambaut A, Drummond AJ (2007) Tracer v1.4, Available from
http://beast.bio.ed.ac.uk/Tracer.
Robinson RA (1974) Terminal report of the FAO coffee
pathologist to the government of Ethiopia. FAO, Rome
AGO ⁄ 74 ⁄ 443, 16pp.
Robinson RA (1976) Plant Pathosystems. Springer-Verlag, Berlin.
Ronquist F, Huelsenbeck JP (2003) MrBayes 3: Bayesian
phylogenetic inference under mixed models. Bioinformatics,
19, 1572–1574.
Rozas J, Sánchez-DelBarrio JC, Messeguer X, Rozas AR (2003)
DnaSP: DNA polymorphism analyses by the coalescent and
other methods. Bioinformatics, 19, 2496–2497.
Rundle HD, Nosil P (2005) Ecological speciation. Ecology
Letters, 3, 336–352.
Schluter D (2009) Evidence for ecological speciation and its
alternative. Science, 323, 737–741.
Schluter D, Conte GL (2009) Genetics and ecological speciation.
Proceedings of the National Academy of Sciences, USA, 106,
9955–9962.
Servedio MR, Doorn GSV, Kopp M, Frame AM, Nosil P (2011)
Magic traits in speciation: ‘‘magic’’ but not rare? Trends in
Ecology and Evolution, 26, 389–397.
Shimodaira H, Hasegawa M (1999) Multiple comparisons of
log likelihoods with applications to phylogenetic inference.
Molecular Biology and Evolution, 16, 1114–1116.
Silva DN, Talhinhas P, Lei C, Várzea V, Paulo OS, Batista D
(2012) Application of the Apn2 ⁄ MAT locus to improve the
systematics of the Colletotrichum gloeosporioides complex: an
example from coffee (Coffea spp.) hosts. Mycologia,
doi:10.3852/11-145.
Stukenbrock E, McDonald B (2008) The origins of plant
pathogens in agro-ecosystems. Annual Review of
Phytopathology, 46, 75–100.
Swofford DL (2003) PAUP*: Phylogenetic Analysis Using
Parsimony (* and Other Methods), Version 4. Sinauer
Associates, Sunderland, Massachusetts.
Taylor JW, Jacobson DJ, Kroken S et al. (2000) Phylogenetic
species recognition and species concepts in fungi. Fungal
Genetics and Biology, 31, 21–32.
Via S (2001) Sympatric speciation in animals: the ugly duckling
grows up. Trends in Ecology and Evolution, 16, 381–390.
Winkler IS, Mitter C, Scheffer SJ (2009) Repeated climate-linked
host shifts have promoted diversification in a temperate
clade of leaf-mining flies. Proceedings of the National Academy
of Sciences, USA, 43, 18103–18108.
Zaffarano PL, McDonal BA, Linde CC (2008) Rapid speciation
following recent host shifts in the plant pathogenic fungus
Rhynchosporium. Evolution, 62, 1418–1436.
Zeigler RS (1998) Recombination in Magnaporthe grisea. Annual
Review of Phytopathology, 36, 249–275.
This work was part of the Msc thesis of D.N.S at CIFC ⁄ IICT in
collaboration with FCUL. D.N.S’s main research interests are
speciation, phylogenomics and population genomics of fungal
species with the application and development of bioinformatic
tools. P.T.’s research focuses on plant-microbe interactions

2670 D . N . S I L V A E T A L .
involving Hemileia vastatrix and Colletotrichum spp., including
molecular mechanisms of pathogenicity and resistance, epide-
miology and genetic diversity. L.C.’s main research interests
are in biodiversity, systematics and phylogenetics of micro-
fungi. L.M.’s research focuses on coffee management and
pathology. E.K.G. is a plant pathologist with bias towards deci-
phering the mechanisms of plant-pathogen interactions, their
evolution to generate diversity and markers for resistance ⁄ -
virulence. A.L.’s research interests focuses on molecular and
cytological mechanisms involved in Coffea – Colletotrichum kaha-
wae and Coffea - Hemileia vastatrix interactions along with
genetic diversity of these pathogens. V.V.’s research focuses on
the pathology of coffee main diseases, dealing with pathogen
physiologic specialization (pathotypes) to screen for resistant
coffee genotypes. O.S.P. is currently working on phylogeogra-
phy and evolutionary ecological genomics. D.B’s research is
focused on the population structure dynamics, virulence adap-
tation and evolution of Coffee pathogens using molecular mar-
kers, population genetics and genomics.
Data accessibility
List of isolates, natural hosts, countries and geographical
regions: Table S1 (Supporting information).
DNA sequences were deposited in the EMBL database with
the accession numbers presented in Table S1 (Supporting
information).
Multiple sequence alignments for each nuclear marker and
*BEAST, GMRF Skyride and IMa input files were archived in the
dryad repositories: doi:10.5061/dryad.c0f2n3mj.
1365294x, 2012, 11, Downloaded from https://onlinelibrary.wiley.com/doi/10.1111/j.1365-294X.2012.05557.x by Universidad De Cadiz, Wiley Online Library on [06/12/2023]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
Supporting information
Additional supporting information may be found in the online
version of this article.
Table S1 List of the isolates with information regarding their
species, natural host, geographic origin and GenBank accession
numbers.
Table S2 Partitioning schemes for the substitution and clock
models of *BEAST. Combined partitions are represented inside
‘()’. For the MAT1-2-1 partition, codon partitioning was also
performed for the first (CP1), second (CP2) and third positions
(CP3).
Table S3 2ln Bayes Factors results of comparisons of all parti-
tion strategies, calculated from the marginal likelihood of each
scheme using TRACER V1.4. A positive value indicates evidence
against alternative hypotheses.
Table S4 Summary of the neutrality test statistics for the
UG1—Colletotrichum kahawae group.
Table S5 Log Likelihood Ratio Test (LLRT) of nested models
with different patterns of gene flow compared with the Full
model, as estimated from IMa analyses.
Fig. S1 50% majority rule Bayesian tree for each molecular
marker used in this study.
Please note: Wiley-Blackwell is not responsible for the content
or functionality of any supporting information supplied by the
authors. Any queries (other than missing material) should be
directed to the corresponding author for the article.
 2012 Blackwell Publishing Ltd