Population divergence in the wheat leaf rust fungus Puccinia triticina is correlated with wheat evolution

Abstract

Co-evolution of fungal pathogens with their host species during the domestication of modern crop varieties has likely affected the current genetic divergence of pathogen populations. The objective of this study was to determine if the evolutionary history of the obligate rust pathogen on wheat, Puccinia triticina, is correlated with adaptation to hosts with different ploidy levels. Sequence data from 15 loci with different levels of polymorphism were generated. Phylogenetic analyses (parsimony, Bayesian, maximum likelihood) showed the clear initial divergence of P. triticina isolates collected from Aegilops speltoides (the likely B genome donor of modern wheat) in Israel from the other isolates that were collected from tetraploid (AB genomes) durum wheat and hexaploid (ABD genomes) common wheat. Coalescence-based genealogy samplers also indicated that P. triticina on A. speltoides, diverged initially, followed by P. triticina isolates from durum wheat in Ethiopia and then by isolates from common wheat. Isolates of P. triticina found worldwide on cultivated durum wheat were the most recently coalesced and formed a clade nested within the isolates from common wheat. By a relative time scale, the divergence of P. triticinia as delimited by host specificity appears very recent. Significant reciprocal gene flow between isolates from common wheat and isolates from durum wheat that are found worldwide was detected, in addition to gene flow from isolates on common wheat to isolates on durum wheat in Ethiopia.

Introduction

The beginning of agriculture 12 000 years ago in the Fertile Crescent made possible the rise of human culture and civilization, and changed the genetic makeup of cereal populations through extensive domestication (Salamini et al., 2002). Hybridization of different genomes leading to ploidy changes is a landmark of wheat evolution. Archeological, genetic and molecular evidence revealed that modern cultivated hexaploid common wheat, Triticum aestivum L. with A, B, and D genomes is derived from the hybridization of tetraploid emmer wheat, T. turgidum ssp. dicoccum L. that has A and B genomes, and the wild diploid D genome species Triticum tauschii (Feldman, 2001) ca. 8000 years ago. In emmer wheat, the A genome donor is derived from T. urartu and the B genome donor is likely from Aegilops speltoides. Coupled with the intriguing history of wheat evolution is the corresponding adaptation of fungal pathogens from wild grass hosts to the founding genotypes of present day modern wheat cultivars. We are interested in determining if the historical divergence of an obligate parasite with different forms on diploid, tetraploid and hexaploid wheat hosts is congruent with the history of wheat evolution.

Documentation of the destructive nature of the rust diseases on wheat can be traced back to classical Greek and Roman eras and also to references in the Old Testament (Chester, 1946). Distinct species of the causal pathogens were recognized only in the 19th century when the individual species were identified (Bolton et al., 2008). Among the three rusts of wheat, leaf rust caused by Puccinia triticina Erikss. is the most prevalent (Chester, 1946), causing regular economic losses due to reduced number and size of kernels and decreased quality of flour. With a macrocyclic-heteroecious life cycle, P. triticina produces pycniospores and aeciospores on Thalictrum speciosissimum L., known as common meadow rue, and urediniospores, teliospores and basidiospores on wheat and wheat-related hosts (Mains and Jackson, 1921). The uredinial/telial hosts include hexaploid common wheat worldwide, tetraploid durum wheat (T. turgidum ssp. durum L., A and B genomes) in the Middle East, Ethiopia, Europe, Mexico and South America (Ordoñez and Kolmer, 2007); tetraploid emmer wheat in the Middle East (Manisterski et al., 2000) and Ethiopia (Andenow et al., 1997) and diploid A. speltoides in Israel (Yehuda et al., 2004). Puccinia triticina also occurs on A. cylindrica in the USA, but no occurrence has been reported in natural stands of other wild grasses such as A. sharonensis, A. tauschii and T. timopheevi, or the ancient diploid wheat T. monococcum (Bolton et al., 2008) although infections on these hosts can be obtained in artificial greenhouse inoculations (Anikster et al., 2005b). Isolates of P. triticina collected from modern hexaploid wheat are also virulent to wild emmer wheat, T. turgidum ssp. dicoccoides (Dinoor et al., 1991; Manisterski et al., 2000; Anikster et al., 2005a) in Israel but are generally avirulent to cultivated durum wheat that was derived from emmer wheat. Isolates virulent to A. speltoides are avirulent to hexaploid and tetraploid wheats. Sexual reproduction of P. triticina on T. speciosissimum and other Thalictrum spp. was sporadically reported in Portugal (D'Oliveira and Samborski, 1966), Italy (Casulii and Siniscalco, 1987), Spain (Young and D'Oliveira, 1982) and the United States (Levine and Hildreth, 1957). However, molecular marker data in population genetic studies indicate that in worldwide populations the clonal reproduction of urediniospores is highly predominant, with little if any evidence of sexual recombination (Goyeau et al., 2007; Kolmer and Ordoñez, 2007; Ordoñez and Kolmer, 2009; Ordoñez et al., 2010; Kolmer et al., 2011).

Molecular markers have been used to assess genetic variation within and between populations of P. triticina. Genotypes of P. triticina generated with random amplified polymorphism of DNA markers demonstrated a general correlation with the virulence phenotypes (Kolmer and Liu, 2000). Genotypes of P. triticina based on amplified fragment length polymorphism identified the possible introduction of new races in North America (Kolmer, 2001). Simple sequence repeat (SSR) markers (Duan et al., 2003; Szabo and Kolmer, 2007) have been used to characterize P. triticina populations from different continents (Kolmer and Ordoñez, 2007; Ordoñez and Kolmer, 2007; Ordoñez and Kolmer 2009; Ordoñez et al., 2010; Kolmer et al., 2011). SSR genotypes of P. triticina collected from A. speltoides were highly divergent from P. triticina genotypes from wheat. Comparison of P. triticina isolates indicated that isolates from tetraploid durum wheat are less variable than isolates from hexaploid common wheat (Ordoñez and Kolmer, 2007; Goyeau et al., 2011) for both virulence and SSR genotypes. The goal of the present study was to reconstruct the evolutionary history of P. triticina divergence as related to its adaptation to hosts with different ploidy levels. Our initial hypotheses are that P. triticina on A. speltoides represents the earliest form of P. triticina, and that P. triticina isolates with virulence to tetraploid durum wheat are derived from genotypes with virulence to hexaploid common wheat.

Materials and methods

Puccinia triticina isolates

Seventy P. triticina isolates were selected based on a neighbor-joining tree constructed from SSR genotypes from a worldwide collection (663 total isolates) in order to include isolates representing all major genotype groups and geographic regions. Among those, 48 isolates were from common wheat that originated from Central Asia, Europe, the Middle East, North America, South America, South Africa and New Zealand and 20 isolates from durum wheat that originated from the Middle East, Ethiopia, Europe, North America and South America (Table 1). Two samples collected from A. speltoides were from Israel. Genomic DNA of the isolates used in previous studies was used for obtaining single nucleotide polymorphism (SNP) data. In addition, we used DNA from two isolates of P. persistens, (Markova and Urban, 1998) a sister species to P. triticina, that were collected in the Czech Republic and Sweden from Elymus repens, as the outgroup in phylogenetic and coalescence analyses. Isolate numbers, countries of origin and SSR genotype group designations are listed in Table 1.

Table 1. Origins and hosts of Puccinia triticina isolates used for phylogenetic and coalescence study.

Species—isolatea	Haplotypes	Country	Population—SSR group	Reference
Puccinia triticina
On Aegilops speltoidesa
ISR850	H8, H9	Israel	—	Yehuda et al., 2004
ISR1400	H7, H10	Israel	—	Yehuda et al., 2004
On Triticum turgidum
ARG9_3	H12, H12	Argentina	—	Ordoñez and Kolmer, 2007
CA1_2	H12, H14	USA	—	Ordoñez and Kolmer, 2007
CH14_21	H5, H12	Chile	—	Ordoñez and Kolmer, 2007
M11_1	H12, H16	Mexico	—	Ordoñez and Kolmer, 2007
M14_3	H12, H15	Mexico	—	Ordoñez and Kolmer, 2007
ES1_1	H11, H18	Spain	EU_3	Kolmer et al., 2012
ES27	H11, H14	Spain	—	Ordoñez and Kolmer, 2007
F2_2	H12, H13	France	—	Ordoñez and Kolmer, 2007
F5_3		France		Ordoñez and Kolmer, 2007
ITA1_1	H11, H18	Italy	EU_3	Kolmer et al., 2012
ETH26_2	H12, H14	Ethiopia	ME_4	Kolmer et al., 2011
ETH4114_7		Ethiopia	ME_3	Kolmer et al., 2011
ETH4_1		Ethiopia	—	Ordoñez and Kolmer, 2007
E1_1		Ethiopia	—	Ordoñez and Kolmer, 2007
E6_1		Ethiopia	—	Ordoñez and Kolmer, 2007
E10_4		Ethiopia	—	Ordoñez and Kolmer, 2007
E11_1		Ethiopia	—	Ordoñez and Kolmer, 2007
E16_1		Ethiopia	—	Ordoñez and Kolmer, 2007
E16_4		Ethiopia	—	Ordoñez and Kolmer, 2007
E4090_3		Ethiopia	ME_3	Kolmer et al., 2011
On Triticum aestivum
AZB 1_2	H17, H19	Azerbaijian	CA_1	Kolmer and Ordoñez, 2007
AZB6_1		Azerbaijian	CA_1	Kolmer and Ordoñez, 2007
NKAZ51_2		Kazakhstan	CA_2	Kolmer and Ordoñez, 2007
S KAZ 44_3		Kazakhstan	CA_2	Kolmer and Ordoñez, 2007
AZB5_1		Azerbaijian	CA_3	Kolmer and Ordoñez, 2007
S KAZ 16B	H17, H30	Kazakhstan	CA_3	Kolmer and Ordoñez, 2007
TJK7_1	H6, H25	Tajikistan	CA_4	Kolmer and Ordoñez, 2007
UZB5_2		Uzbekistan	CA_4	Kolmer and Ordoñez, 2007
CZ20_09		Czech_Slovakia	EU_1	Kolmer et al., 2012
GBR1_1_09	H17, H23	UK	EU_1	Kolmer et al., 2012
GBR15_3_09		UK	EU_1	Kolmer et al., 2012
CZ4_09	H29, H29	Czech_Slovakia	EU_2	Kolmer et al., 2012
TUR29_1_09	H21, H22	Turkey	EU_2	Kolmer et al., 2012
CZ_SK95_9_2		Czech_Slovakia	EU_4	Kolmer et al., 2012
CZ18_09		Czech_Slovakia	EU_4	Kolmer et al., 2012
CZ10_9	H3, H4	Czech_Slovakia	EU_5	Kolmer et al., 2012
HG95_10_1		Hungary	EU_6	Kolmer et al., 2012
HG95_3_2	H17, H30	Hungary	EU_6	Kolmer et al., 2012
ES9_1		Spain	EU_7	Kolmer et al., 2012
ITA14_1	H18, H34	Italy	EU_7	Kolmer et al., 2012
CZ11_09		Czech_Slovakia	EU_8	Kolmer et al., 2012
TUR1_1_09	H17, H31	Turkey	EU_8	Kolmer et al., 2012
EGY_154_A	H26, H32	Egypt	ME_1	Kolmer et al., 2011
EGY_195_A		Egypt	ME_2	Kolmer et al., 2011
EGY_195_B		Egypt	ME_2	Kolmer et al., 2011
ISR_L_2		Israel	ME_2	Kolmer et al., 2011
99 NC	H24, H27	United States	NA_1	Ordoñez and Kolmer, 2009
Race_1	H17, H17	Canada	NA_1	Ordoñez and Kolmer, 2009
01_AL_21	H1, H17	United States	NA_2	Ordoñez and Kolmer, 2009
MI 91_2		United States	NA_2	Ordoñez and Kolmer, 2009
04 TX 67	H17, H31	United States	NA_3	Ordoñez and Kolmer, 2009
04_TX_32	H17, H35	United States	NA_3	Ordoñez and Kolmer, 2009
CAN 71_2_96		Canada	NA_4	Ordoñez and Kolmer, 2009
Race_9		Canada	NA_4	Ordoñez and Kolmer, 2009
04 MN 485		United States	NA_5	Ordoñez and Kolmer, 2009
99_NC	H17, H26	United States	NA_5	Ordoñez and Kolmer, 2009
NZ97_13_3		New Zealand	—	Kolmer, unpublished
NZ97_8_1	H17, H20	New Zealand		Kolmer, unpublished
PER95_82		Peru	SA_1	Ordoñez et al., 2010
URG6_2	H33, H33	Uruguay	SA_1	Ordoñez et al., 2010
BRA19_3	H2, H28	Brazil	SA_2	Ordoñez et al., 2010
URG4_1	H17, H25	Uruguay	SA_2	Ordoñez et al., 2010
ARG512_1		Argentina	SA_3	Ordoñez et al., 2010
URG17_3	H17, H31	Uruguay	SA_3	Ordoñez et al., 2010
CHL08_3_3	H29, H29	Chile	SA_4	Ordoñez et al., 2010
PER87_4107_2		Peru	SA_4	Ordoñez et al., 2010
SF96_4_1		South Africa	SAF	Kolmer, unpublished
SF98_1_2		South Africa	SAF	Kolmer, unpublished
P. persistens
HSZ0689		Sweden	—	Liu et al., 2013
HSZ0695		Czech republic	—	Liu et al., 2013

^aIsolates in bold font are included for compatible data set for GENETREE analysis, and converted into haplotypes listed in next column.

Primer design

To screen for hypervariable DNA loci, the intron regions of house-keeping genes and the flanking regions of SSR markers were examined. In addition, an in silico approach was pursued to select SNPs in hypervariable anonymous regions based on the comparison of Illumina reads of four P. triticina genomes, that is, isolates Race 1, Race 161, Race 9 and PRTUS 35 through the Integrative Genomics Viewer (IGV) (Robinson et al., 2011).

DNA sequences of targeted genes including RNA polymerase II largest subunit (RPB1), the 2nd largest unit (RPB2) and elongation factor 1 alpha (EF1-α) were extracted from the complete genome of Puccinia graminis f. sp. tritici published by the Broad Institute (http://www.broadinstitute.org/annotation/genome/puccinia_group/Regions.html). The extracted sequences were submitted into BLAST searches against the non-annotated genome sequence of Puccinia triticina 1-1 Race 1 in order to obtain the homolog sequences from P. triticina. The DNA sequence segments including multiple intron regions of individual homologous genes from P. triticina were input separately into Primerquest on the Integrated DNA Technologies (IDT) website (http://www.idtdna.com/Scitools/Applications/Primerquest/) as template to design appropriate PCR primers.

To design primers for SSR flanking regions, 16 pairs of SSR primers (Duan et al., 2003; Szabo and Kolmer, 2007) were used to BLAST search the P. triticina genome to locate loci containing targeted SSRs. DNA sequence 1 kb up and down stream from targeted SSR loci were downloaded and input as templates into Primerquest to design appropriate primers. All the output primers were compared with the P. triticina whole genome by BLAST to screen for the candidates that did not match with unspecified regions in the genome. The alignment of Illumina reads of the four P. triticina genomes were examined through the Integrative Genomics Viewer (Robinson et al., 2011). Thirty-six sites of 1.5 kb with 1–92 putative SNPs were selected randomly. Primer design was attempted for these sites based on uniqueness of the site in the genome, specificity of the primers and sequencing cost efficiency. BLAST search was conducted to determine the uniqueness of the site in the genome and specificity of the primers. To be cost efficient, the length of amplified DNA fragment was adjusted to 500–900 bp, including at least 1 SNP.

Polymerase chain reaction and sequencing

PCR was performed in 10 μl reactions containing 0.1 mM dNTPs, 2.0 pmoles of each primer, 1 μl 10 × Titanium Taq buffer, 0.1 μl 50 × Titanium Taq DNA Polymerase (BD Biosciences, Mississauga, ON, USA). Touch-down thermal cycling conditions were used initially (except for locus Ctg84-1, for which the annealing temperature was set as 56 °C): 95 °C for 3 min, followed by 10 cycles of 95 °C for 30 s, 63 °C (decreasing 0.5 °C per cycle) for 20 s, and 72 °C for 2 min, then 26–30 cycles of 95 °C for 30 s, 58 °C for 20 s, 72 °C for 2 min, followed by a final extension of 72 °C for 8 min. In the case when unspecified fragments were amplified, the annealing temperature was kept at 63 or 65 °C to minimize the unspecified amplification (Supplementary Table 1). E-Gel CloneWell SYBR Safe system (Invitrogen, Carlsbad, CA, USA) was used to purify the specified band when unspecified bands could not be eliminated by increasing annealing temperatures. Otherwise, PCR products were cleaned by ExoSAP-IT (USB Products Affymetrix, Inc., Cleveland, OH, USA) before being sent to the BioMedical Genomics Center (University of Minnesota, St Paul, MN, USA) for sequencing. PCR primers were used for sequencing most of the loci, except for SSR flanking loci where internal primers were designed to enhance the sequence quality.

DNA sequence examination and heterozygous site recognition

DNA fragments sequenced by forward and reverse primers were assembled using Sequencher 4.10 (Gene Codes Corporation, MI, USA). Heterozygous sites are common in the dikaryotic P. triticina genome. To recognize the heterozygous sites, the chromatograms of the two strands (sequenced by forward and reverse primers) were examined by eye and only the sites that met at least one of the following three criteria were accepted: (1) sites showed two equal height peaks in at least one high-quality strand (confidence of surrounding bases >40, max. 60, Supplementary Figures 1A and 1B); (2) sites showed alternate bases of high quality (Supplementary Figure 1C) from two strands; (3) sites showed the secondary peaks higher than 50% of the primary peaks (Supplementary Figure 1D) in both strands.

Haplotype inference, gene diversity, neutrality test and phylogentic network

DNA sequences from 15 selected loci were analyzed for polymorphisms and neutrality using DnaSP 5.10 (Rozas et al., 2003) locus–by-locus for 70 isolates. Due to the significant genetic difference between the two isolates from A. speltoides with the other isolates from common and durum wheat, we conducted separate polymorphism analyses for P. triticina isolates on common and durum wheat only, and for all P. triticina isolates, including the isolates from A. speltoides. Diploid sequence data (dikaryotic) were converted to haplotype data by PHASE v2.1 (Stephens et al., 2001) through a coalescence-based Bayesian method. Neutrality was evaluated using parameters Fu and Li's D*, Fu and Li's F* (Fu and Li, 1993) and Tajima's D (Tajima, 1989).

A phylogenetic network was generated by SplitsTree 4.13 (Huson and Bryant) for 70 P. triticina isolates with a total of 140 alleles. The NeighborNet was performed with variance as ordinary least squares (Bryant and Moulton, 2004). To accommodate incompatible and ambiguous signals in the data, the split network implies evolutionary history by parallel edges and branches. When the network is not significantly non-tree like, tree-based phylogenies can be performed. The presence of recombination within and among 15 loci was evaluated with the PHI_w test (Bruen et al. 2006) incorporated in SplitsTree 4.13 (Huson and Bryant 2006). This test is very general in that a single population is not assumed, and it determines only if a recombination signal is present in the sequence data. This approach is not sensitive to mutation rate correlation due to mutation hot spots that would falsely imply recombination.

Phylogenetic analyses

Phylogenetic analyses were conducted through the three different algorithms of parsimony, Bayesian and maximum likelihood. Parsimony analyses for each of the 15 selected loci and the concatenated DNA sequence matrix were conducted using PAUP*4.0b10 (Swofford, 1998). Heuristic searches with random stepwise addition of 100 replicates and tree bisection-reconnnection branch swapping were performed. A limit of 1 000 000 rearrangements was set for each replicate. Bootstrapping analyses were set with 200 replicates with full heuristic search of random step-wise addition of 20 replicates and limit of 10 000 rearrangements per replicate. A Bayesian phylogenetic analysis was conducted using MrBayes 3.1, with a GTR+G+I model of evolution. (Huelsenbeck and Ronquist, 2001). DNA substitution rates were set as a flat Dirichlet (1.0, 1.0, 1.0, 1.0), which allows the program to estimate the parameters while mixing among Markov chain Monte Carlo (MCMC) chains. Four chains of MCMC were performed and finished at 7 500 000 generations when the average deviation was lower than 0.01; sampling frequency was every 100 generations. The equilibrium of posterior probability was checked by plotting posterior probability value with generations in Excel. The first 1250 trees (1.67%) before equilibrium were discarded (burn-in). Maximum likelihood analyses were performed using GARLI 0.96 (Zwickl, 2006) with unspecified models. Two independent runs each with 10 000 000 generations were performed. Bootstrap replicates were 1000 with default settings. The two Puccinia persistens isolates were used as an outgroup.

Coalescence analysis

The coalescence-based program GENETREE 9.0 (Griffiths, http://www.maths. monash.edu.au/∼mbahlo/mpg/gtree.html) estimates most recent common ancestors for subdivided populations using a Markov chain simulation under an infinite-many-sites model. With the assumption that there is no recombination within and among loci, the DNA segregating sites must be screened for compatibility. Therefore, the isolates with gaps in sequence were eliminated and the Clade and Matrix (http://snap.cifr.ncsu) incorporated in SNAP workbench (Price and Carbone, 2005) was used to eliminate incompatible sites. An analysis of haplotypes of 35 selected isolates from common wheat, durum wheat and the two isolates from A. speltoides with a total of 88 compatible segregating sites was conducted. The matrix was submitted to GENETREE module in SNAP workbench. Two independent simulations were performed each with 100 000 steps; θ was obtained by maximum likelihood estimates. Initial backward migration matrix was inferred by the MIGRATE v. 3.0 module in SNAP workbench. Parameters were set as: short-chain 20, short-increase 200, short-sample 500 000; long-chain 10, long-increase 200, long-sample 500 000 and burn-in 10 000 000 (10%). Seven independent runs were performed in order to reach stabilized estimation.

The coalescence-based genealogy sampler BEAST reconstructs rooted and time-measured phylogenies based on strict or relaxed molecular clock models through Bayesian MCMC analysis for molecular sequences of multiple loci (Drummond et al., 2012). For the comparison with the results from GENETREE and general parsimony analyses, we analyzed two data sets: the 35 isolates (haplotypes) with 88 compatible segregating sites used for GENETREE, and 72 isolates (genotypes) for 10 395 nucleotides from 15 loci. BEAUti v 1.7.1 was used to generate XML input file for BEAST v 1.7.1. Since no time points are available for references, the strict clock model was adopted. HKY and GTR substitution models were attempted in alternate analyses for comparison; 10 000 000 generations and sample frequency 1000 were set for MCMC. The output was analyzed with TreeAnnotater v1.7.1 with burn-in as 1000 (10%). The final tree with highest probability was visualized with FigTree v1.3.1 (http://tree.bio.ed.ac.uk/software/figtree/).

Isolation and migration

Low levels of genetic divergence among populations can be a result of recent divergence or due to recurrent gene flow among populations. An isolation-with-migration model (Nielsen and Wakeley, 2001) implemented in the IMa2 program developed by Hey and Nielsen (2007) can discriminate between these two different models among multiple populations through an MCMC approach. The migration with isolation model was tested for four P. triticina populations based on host and geographic origins, which were recognized based on the phylogenetic analyses mentioned above. Three independent runs were performed for simulation with IMa2; parameters were set as: metropolis coupling implemented using 80 chains, heating scheme by geometric increment model (–hfg –hn80 –ha0.99 –hb0.5) and burn-in 1 000 000 steps; migration prior follows exponential distribution with mean set as 10 (-j 7, -m 10), population size scalar 5 (-q 5), splitting time (-t 10), all the priors were set based on preliminary analyses; running time as determined by checking the output files for stationarity stage.

Results

Screening for DNA polymorphic loci

Initially, sixteen P. triticina isolates were used to test primers designed for three house-keeping genes (EF1-α, RPB1 and RPB2), 16 SSR flanking loci and 17 anonymous hyper variable regions. Among a total of 83 pairs of primers, 15 pairs were selected for the present study as a result of the uniqueness of the site in the genome, specificity of the primers, success rate of PCR and sequencing cost efficiency. The selected loci included: one house-keeping gene (RPB2), four SSR flanking regions (two up and down streams amplified by two pair of primers) and ten anonymous hypervariable regions located in supercontigs 1, 5, 9, 10, 12, 34, 47, 67 and 84 (Supplementary Table 1). Although amplification was attempted with 70 isolates with all 15 loci, for certain isolates the fragments could not be amplified or sequenced with high quality. For each locus, from 50 to 69 isolates were amplified (Table 2). The non-amplified fragments were treated as missing data in the phylogenetic analyses. DNA sequences for 15 loci were submitted into GeneBank; accession numbers are JN592765–592776 and JX543322–543364 for RPB2; JN592805–592816 and JX543275–543321 for FPtssr91U; JN592817–592829 and JX543227–543274 for FPtssr91D; JN592830–592844 and JX542944–542985 for FPtssr161U; JN592845–592859 and JX543187–543226 for FPtssr161D; JN592876–592890 and JX542836–542885 for Ctg1-3; JN592891–592906 and JX542617–542671 for Ctg5-1; JN592907–592920 and JX543146–543186 for Ctg9-2; JN592921–592936 and JX542727–542783 for Ctg10-2; JN592937–592950 and JX542886–542943 for Ctg10-3; JN592951–592963 and JX542986–543040 for Ctg12-2; JN592964–592978 and JX543041–543095 for Ctg34-1; JN592979–592992 and JX543096–543145 for Ctg47-1; JN592993–593007 and JX542672–542726 for Ctg67-1; JX542784–542835 for Ctg84-1.

Table 2. Polymorphic sites, DNA variation and neutrality tests of 15 loci for Puccinia triticina isolates.

	Polymorphic sites									Neutrality testsa
	Number of samples	Number of sequences analyzed	Total sites	Sites excluding gaps and missing data	Variable sitesb	Singleton variable sites	Parsimony-informative sites	Variable sites in regions with gap	Number of polymorphic sites per kb	Fu and Li's D*	Fu and Li's F*	Tajima's D	PHIw recombination test (P value)c
RPB2	51	102	677	486	6	2	2	2	9	−1.47720	−1.01350	0.43399	NS (0.5)
FPtssr91U	57	114	842	617	5	1	1	3	6	−1.1–168	−0.69385	0.53891	NA
FPtssr91D	60	120	848	705	5	0	2	3	6	−0.62657	−0.25902	0.064935	NS (1.0)
FPtssr161U	54	108	577	410	8	1	3	4	14	−0.28165	−0.62160	−1.04443	NS (0.5)
FPtssr161D	53	106	612	456	18	8	5	5	29	3.30563**	3.25343**	−1.71255	NS (0.3871)
Ctg1-3	64	128	758	735	26	4	19	3	34	0.11648	0.58632	1.07481	NS (0.0512)
Ctg5-1	67	134	588	521	10	2	6	2	17	−0.44298	−0.11003	0.56497	NS (1.0)
Ctg9-2	50	100	788	648	4	1	3	0	5	−0.26472	0.00422	0.56915	NS (1.0)
Ctg10-2	67	134	799	721	4	0	4	0	5	0.91610	0.90854	0.46869	NS (1.0)
Ctg10-3	64	128	751	672	6	1	5	0	8	0.11315	0.36717	0.69720	NS (1.0)
Ctg12-2	63	126	425	288	19	1	11	7	45	0.80631	1.44164	1.92003	NS (1.0)
Ctg34-1	64	128	516	414	12	1	11	0	23	0.80394	1.22323	1.41667	NS (1.0)
Ctg47-1	57	114	606	403	28	2	16	10	42	0.69961	0.71309	0.42508	NS (0.7687)
Ctg67-1	64	128	650	379	32	3	8	21	49	−0.66012	−0.59934	−0.22203	NS (0.7148)
Ctg84-1	64	128	664	545	8	1	7	0	12	0.40629	0.26341	−0.12577	NS (1.0)
P. triticina on wheat and Aegilops speltoides
RPB2	53	106	677	486	13	3	7	3	19	−0.76132	−0.92160	−0.80240	NS (0.2095)
FPtssr161U	55	110	577	406	11	1	5	5	19	0.14364	−0.43817	−1.39400	NS (0.5)
Ctg5-1	68	136	588	521	21	11	8	2	36	−3.69668**	−3.27151**	−1.17507	NS (1.0)
Ctg9-2	52	104	788	648	7	1	6	0	9	0.31181	0.16301	−0.20288	NS (1.0)
Ctg10-2	69	138	799	712	8	3	5	0	10	−1.29576	−1.32406	−0.77360	NS (1.0)
Ctg10-3	66	132	751	672	15	1	14	1	20	1.01080	0.42118	−0.75956	NS (1.0)
Ctg12-2	65	130	425	288	28	3	16	9	66	0.25324	0.39635	0.45730	NS (1.0)
Ctg34-1	66	132	516	414	12	1	11	0	23	0.79930	1.22017	1.41568	NS (1.0)
Ctg67-1	66	132	650	379	32	3	8	21	49	−0.67164	−0.54633	−0.07807	NS (0.7024)

^aFor the tests labeled with **, P<0.02, indicating the null hypothesis of neutrality can be rejected at 2% level; *, 0.05<P<0.1, indicating the null hypothesis of neutrality can be rejected at 10% level, but not at 5%. For all other tests, P>0.1, indicating the null hypothesis of neutrality cannot be rejected at 10% level.

^bAnalyzed sites exclude gaps and missing data.

^cPHI_w recombination test was performed in SplitsTree 4.13.1, NA stands for not available due to too few informative characters.

DNA polymorphism of 15 loci in P. triticina isolates

SNPs from 15 loci examined in P. triticina isolates on common and durum wheat varied from 5–49 SNPs per kilobase. Two anonymous regions on contig 9 (Ctg9-2) and contig 10 (Ctg10-2) had the overall lowest polymorphism rates, whereas anonymous contig 67 (Ctg67-1) had the highest rate (Table 2). Neutrality tests resulted in P values larger than 0.1 for most of the loci, indicating neutral evolution, except for microsatellite PtSSR161 down stream region (FPtssr161D), where Fu and Li's D* and Fu and Li's F* P was lower than 0.02, suggesting the mutations in this locus deviated from selective neutrality. However, Tajima's D tests suggested that all loci did not deviate significantly from neutral evolution.

Nine loci were examined for the DNA variation among isolates on wheat and A. speltoides; Ctg12-2 had the highest polymorphism rate of 66 per kilobase. In neutrality tests, Fu and Li's D* and Fu and Li's F* P was lower than 0.02 for locus ctg5-1, indicating non-neutral selection pressure. The haplotype inference by PHASE was confirmed for one locus (Ctg12-2) through sequencing different alleles by using allele specific primers (data not shown).

Test for recombination between and within loci were negative. The PHI test for the concatenated sequence data of the 15 loci in SpiltsTree 4.13 for recombination between loci was non-significant (P=0.0949) and also did not detect any statistically significant evidence for recombination within each locus (Table 2). The split network generated by SplitsTree 4 was generally tree-like (Supplementary Figure 2). The grouping was congruent with that from the other phylogenetic approaches (see below), in which isolates from Ethiopian durum wheat split next to the isolates from Aegilops, followed by isolates from common wheat, and then isolates from cultivated durum wheat imbedded in the isolates from common wheat. This indicated that the sampling error was not significant and that tree-based phylogenetic approaches could be applied.

Phylogenetic analyses

Phylogenetic analyses based on individual loci determined that P. triticina isolates from A. speltoides were genetically distinct from P. triticina on wheat, but the relationships between the subpopulations on wheat were variable from locus to locus. The low inference is possibly because of homoplasy and limited variation for some loci (Supplementary Figure 3, Supplementary Table 2). With relatively little phylogenetic information available, stochastic or sampling error would result in the different tree topologies. As a result, a holistic approach was used to maximize the phylogenetic information (Smith, 2000). The phylogenetic trees that resulted from parsimony majority rule consensus, Bayesian and maximum likelihood analyses based on concatenated DNA sequences of 15 loci separated the P. triticina isolates with various degrees of support into four groups that correlated with host specificities and geographic locations (Figure 1): P. triticina on A. speltoides (indicated as Aegilops form in the following paragraphs), P. triticina on worldwide cultivated durum wheat (cultivated durum form), P. triticina on durum wheat from Ethiopia (Ethiopian durum form) and P. triticina on worldwide common wheat (common wheat form). Rooted with P. persistens, the phylogenies indicated that the Aegilops form diverged initially and appeared as a strongly supported clade with high support values 100/100/100/95 (majority consensus frequency/parsimony bootstrap/Bayesian posterior probability/maximum likelihood bootstrap), followed by the separation between the common wheat form (100/66/100/60) and the Ethiopian durum form. The Ethiopian durum form appeared as a weakly supported clade in the parsimony majority rule consensus tree and bootstrap support values of both parsimony and maximum likelihood analyses were lower than 50% (78/<50/66/<50). The cultivated durum form grouped as a clade nested within the common wheat form with varied support (63/<50/99/71). The matrix of concatenated DNA sequences from 15 loci and phylogenetic tree were submitted into TreeBASE. Study accession no. S13628. Reviewer access URL: http://purl.org/phylo/treebase/phylows/study/TB2: S13628?x-access-code=f8d52f18514ed95babae6dedf2d6f924 &format=html.

Figure 1.

Majority consensus of 20 854 most parsimonious trees based on DNA sequences of 15 loci. L=392, CI=0.786, RI=0.924, RC=0.726, HI=0.214, G-fit=−202.450. Values on branches are majority consensus frequency/Parsimony bootstrap/Bayesian posterior probability/maximum likelihood bootstrap. Hedged bar indicates P. triticina on durum wheat, gray bar indicates P. triticina from Ethiopian durum wheat, black bar indicates P. triticina on Aegilops speltoides, non-labeled isolates are P. triticina from common wheat.

Coalescence analysis

The coalescence tree by GENETREE with highest root probability for 35 isolates (35 haplotypes, Table 1) in three populations (the Ethiopia durum form was not included due to poor resolution that caused the program to crash) determined that the cultivated durum form coalesced the most recently and represented the most recently derived population, whereas the Aegilops form coalesced the earliest and represented the most ancestral population (Figure 2). The common wheat population was intermediate in coalescence between the Aegilops and cultivated durum forms.

Figure 2.

Coalescence tree with highest root probability. The vertical axis on the right stands for relative time scale, from the bottom to the top reflects from the present to the past. At the bottom are the haplotypes. Dots on the branches are the number of mutations. Haplotypes of cultivated durum form coalesced at time t1, common wheat form at t2 and Aegilops form at t3.

Time-scaled coalescence trees that resulted from BEAST analyses for 35 isolates (haplotypes) with 88 compatible segregating sites and 72 isolates (genotypes) for 10 395 nucleotides based on HKY and GTR substitution models revealed the same order of divergence for the P. triticina host-specified populations (only the tree of 72 isolates with HKY model is shown in Figure 3). The form on A. speltoides diverged first, followed by the Ethiopian durum form and then the forms on common wheat and cultivated durum wheat worldwide. The cultivated durum wheat form was recently derived than the common wheat form. Compared with the outgroup P. persistens, the most closely related species to P. triticina, the divergence of P. triticina populations are very recent (Figure 3). A distinction between P. triticina and P. persistens was not evident based on partial ITS and EF1-α sequence data, indicating that they are recently diverged species (M. Liu et al. 2013). By the relative scale, the divergence among P. triticina populations started 0.004 timescale ago, which is much later than the divergence between P. triticina and P. persistens, 0.015 timescale past.

Figure 3.

Rooted time-scaled phylogeny of 72 isolates by BEAST 1.7.1 based on 15 loci and HKY model. Time scale at the bottom from right (0.0) to left (0.015) indicates present to the past. The hedged bar shows P. triticina on cultivated durum wheat, the gray bar shows P. triticina on Ethiopian durum wheat, the black bar shows P. triticina on Aegilops speltoides, non-labeled isolates are P. triticina on common wheat.

Isolation and migration

The levels of gene flow among four populations detected through three runs of Ima2 analyses suggested current significant (P<0.001) reciprocal gene flow between the common wheat form and the cultivated durum wheat form (Table 3). There was also detectable gene flow from the common wheat form to the Ethiopian durum form. One of the three runs detected gene flow from the Ethiopian durum form to the cultivated durum form, but other two longer runs indicated the gene flow was not significant. There was no detectable gene flow to and from the Aegilops form.

Table 3. Gene flow analyses using IMa2, runtime information and migration rates for four populations of P. triticina.

Populations		Runs	Steps (after burn-in)	Migration rate 2NM (LLRtest)a
0	1			0>1b	1>0
CW	DW	Run 1	1 000 000	0.349 (26.389***)	0.156 (307.336***)
		Run 2	3 034 032	0.502 (29.254***)	0.168 (105.039***)
		Run 3	4 324 321	0.476 (30.177***)	0.179 (105.804***)
CW	ED	Run 1	1 000 000	0.073 (1.399 NS)	0.189 (8.139***)
		Run 2	3 034 032	0.050 (0.649 NS)	0.117 (8.068**)
		Run 3	4 324 321	0.096 (0.619 NS)	0.154 (8.805**)
CW	AG	Run 1	1 000 000	0.000 (−0.000 NS)	0.081 (0.972 NS)
		Run 2	3 034 032	0.000 (−0.000 NS)	0.030 (0.101 NS)
		Run 3	4 324 321	0.000 (−0.000 NS)	0.052 (0.305 NS)
DW	ED	Run 1	1 000 000	0.119 (7.423**)	0.026 (−0.202 NS)
		Run 2	3 034 032	0.008 (0.299 NS)	0.099 (1.397 NS)
		Run 3	4 324 321	0.007 (1.103 NS)	0.025 (−0.014 NS)
DW	AG	Run 1	1 000 000	0.000 (−0.000 NS)	0.113 (0.079 NS)
		Run 2	3 034 032	0.000 (−0.000 NS)	0.119 (0.690 NS)
		Run 3	4 324 321	0.000 (−0.000 NS)	0.041 (0.338 NS)
ED	AG	Run 1	1 000 000	0.000 (−0.000 NS)	0.009 (0.040 NS)
		Run 2	3 034 032	0.005 (0.121 NS)	0.000 (−0.000 NS)
		Run 3	4 324 321	0.040 (0.039 NS)	0.002 (−0.005 NS)

^a*P<0.05, **P<0.01, ***P<0.001.

^b0>1 stands for migration from population 0 to population 1 in the coalescence sense (backwards in time); as time moves forward, the migration is from population 1 to population 0.

Discussion

Divergence of P. triticina populations

In asexual populations, accumulated mutation, genetic drift and selection are recognized as the main forces that affect divergence of lineages (Kohn, 2005). Host specificity and adaptation is an additional key factor driving the evolution and divergence of clonal obligate plant pathogens. Among the diploid ancestors of wheat, P. triticina is found only on A. speltoides in natural stands. The isolates were found in two locations at Ahihud and Qazir in northern and central Israel. Morphological examination and DNA content study revealed the two collections to be P. triticina (Yehuda et al., 2004). Inoculation experiments showed that the collections from A. speltoides were avirulent to common and durum wheat. However, the isolates from A. speltoides and from cultivated durum and common wheat were sexually compatible, indicating that these could be considered as a biological single species. Yehuda et al. (2004) recognized the isolates from A. speltoides as a new forma specialis of P. triticina. Our study based on phylogenetic analysis, BEAST and GENETREE coalescence indicated that the P. triticina on A. speltoides was the earliest diverged lineage, which is in accordance with A. speltoides as a wheat progenitor.

Eight of the P. triticina isolates from durum wheat in Ethiopia were grouped separately from the durum wheat isolates from Mexico, Europe and South America and also from the isolates from common wheat. The wheat population in Ethiopia is unique which may account for the early divergence of this set of isolates. Emmer wheat was introduced to Ethiopia via the Nile from Egypt 5000 years ago (Zaharieva et al., 2010), and Ethiopia is one of the few countries where it is still grown. Emmer wheat is grown on marginal land throughout Ethiopia, and is grown by farmers in mixtures with durum landrace cultivars and common wheat in the Bale, Tigray and Welo regions. Ethiopia is considered as a center of diversity for tetraploid wheats (Payne et al., 2001). The P. triticina population has likely been selected on emmer and tetraploid durum landraces for thousands of years. Since emmer and durum wheat predate common hexaploid wheat, the P. triticina population on tetraploid wheat in Ethiopia may be a remnant of the P. triticina population before the widespread cultivation of common hexaploid wheat. The Bayesian phylogeny and the BEAST coalescence analysis strongly indicated that the Ethiopian durum isolates are ancestral to the isolates from common wheat and to the other isolates from durum wheat. The consensus phylogeny tree weakly supported the separation of these isolates in a single clade. Isolates from durum wheat in Ethiopia were also previously noted to have SSR genotypes distinct from isolates from cultivated durum wheat and common wheat in other parts of the world (Ordoñez and Kolmer, 2007; Kolmer et al., 2011).

P. triticina on common hexaploid wheat is found worldwide, nearly wherever wheat is grown. On a worldwide scale P. triticina is highly diverse for virulence to leaf rust resistance genes in wheat (Kolmer and Liu, 2000) and also for distinct groups of SSR genotypes (Ordoñez and Kolmer, 2009; Kolmer et al., 2012). Although P. triticina isolates from common wheat are generally virulent to wild and cultivated emmer wheat (Anikster et al., 2005a), these isolates are usually avirulent to the modern durum wheat cultivars that are grown in North America, Europe, the Middle East and North Africa. The BEAST and GENETREE analyses all indicated that 12 of the isolates with virulence to durum wheat from Europe, Mexico, South America, California and Ethiopia were derived from P. triticina with virulence to common wheat. Isolates adapted to durum wheat were closest in SNP variation to isolates from common wheat from Central Europe, Central Asia, Israel, New Zealand, South Africa and Spain, based on phylogenetic and BEAST analyses. The cultivated durum form is recently diverged from the common wheat form and adaptation of P. triticina to modern durum wheat cultivars is relatively recent. Isolates of P. triticina from cultivated durum wheat in France (Goyeau et al., 2011) and worldwide (Ordoñez and Kolmer, 2007) had very limited SSR genotype variation and virulence variation to leaf rust resistance genes compared to isolates from common wheat, supporting the hypothesis of recent adaptation to cultivated durum wheat. An isolate of P. triticina with virulence to durum wheat that was collected in Arizona, USA in 2009 had an SSR genotype and virulence very similar to isolates collected from common wheat in western Europe (J. Kolmer, unpublished data). One of the isolates from Ethiopia (E11_1) collected from durum wheat was grouped with the isolates from common wheat in the BEAST and phylogenetic analyses. Shortly after the expansion of domesticated emmer wheat ca. 10 000 years ago, durum wheat appeared in the Near East and replaced emmer as the leading tetraploid wheat by 2000 BC (Zohary and Hopf, 2000). Durum wheat may have originated from a small population of emmer wheat in southeastern Europe (Oliveira et al., 2012). Since emmer wheat is generally susceptible to P. triticina, leaf rust resistance in durum wheat may have only been recently acquired, perhaps during selection for grain quality and agronomic traits.

Phylogenetic and coalescence approaches and interpretation

The strength of the groupings of the P. triticina isolates differed among the analyses employed. In general, the three phylogenetic approaches strongly separated the isolates from Aegilops, but the parsimony and maximum likelihood methods provided weak support for separation of the Ethiopian durum isolates, the common wheat isolates and the isolates from cultivated durum wheat worldwide. This is likely due to low levels of sequence variation overall, resulting in low levels of signal for each separate region. Such stochastic or sampling error would explain the different tree topologies that were observed for the individual regions. As there were only 319 variable sites in total, any phylogenetic inference or groupings will be less robust. Also, since the groups are relatively recently derived, any phylogenetic separation would be expected to be weak in comparison to separation of P. persistens and the isolates from Aegilops, which were strongly supported. In contrast the Bayesian phylogeny and the coalescence approaches used in GENETREE and BEAST strongly supported the separation of the different isolate types. The GENETREE analysis was limited to only compatible sites while BEAST utilized complete sequence data, with compatible and incompatible sites. Both methods strongly placed the isolates from cultivated durum wheat that were found worldwide in a single group that was very recently derived from the isolates from common wheat, while BEAST also strongly grouped the isolates from durum wheat in Ethiopia in an earlier derived group. The differing results from the phylogenetic and coalescence approaches may be due to underlying differences in evolutionary models. BEAST in particular can be used for rigorous statistical inference and hypothesis testing of evolutionary models, in addition to generating phylogenies. BEAST is especially useful for dealing with heterogeneous data and incongruence among data sets (Drummond et al. 2012). The Bayesian inference employed with GENETREE and BEAST also performs better when dealing with missing data (Yang and Rannala, 2012) as occurred in each of the 15 loci examined.

Gene flow among populations

Reciprocal gene flow between the common wheat form and the cultivated durum wheat form and uni-directional gene flow from the common wheat form to the Ethiopian durum form were consistently detected through multiple analyses. These results seem logical when compared with the phylogenetic trees and coalescence-based genealogies: between the two most closely related lineages, the common wheat form and the cultivated durum wheat form, there were high levels of gene flow. There was no detectable gene flow between the Aegilops form and the other forms since it diverged the earliest. These results are consistent with the results from infection experiments that almost all A. speltoides accessions are highly resistant to the P. triticina isolates from common and durum wheat (Anikster et al., 2005b). However, the forms on Aegilops and wheat are sexually compatible, so potential gene flow between the Aegilops form and isolates from wheat cannot be excluded.

In conclusion, our study demonstrated the ability of the obligate plant pathogen P. triticina to expand its host range from the wheat progenitor host species Aegilops speltoides to the modern-day tetraploid and hexaploid wheat varieties. The P. triticina isolates with virulence to durum cultivars that have been found in Europe, South America, Mexico and the Middle East very recently arose from isolates from hexaploid common wheat, indicating a rapid adaptation to a genetically distinct host population. Finally, coalescence analyses indicated that isolates from tetraploid wheats in Ethiopia may be the oldest derived form of P. triticina found on wheat, predating the adaptation of P. triticina to the hexaploid wheats that are grown worldwide.

Data archiving

Sequence data of 15 P. triticina loci were deposited in GenBank with individual accession numbers listed in results section.

The authors declare no conflict of interest.