Cryptic diversity in Tranzscheliella spp. (Ustilaginales) is driven by host switches

Species of Tranzscheliella have been reported as pathogens of more than 30 genera of grasses (Poaceae). In this study, a combined morphological and molecular phylogenetic approach was used to examine 33 specimens provisionally identified as belonging to the T. hypodytes species complex. The phylogenetic analysis resolved several well-supported clades that corresponded to known and novel species of Tranzscheliella. Four new species are described and illustrated. In addition, a new combination in Tranzscheliella is proposed for Sorosporium reverdattoanum. Cophylogenetic analyses assessed by distance-based and event-cost based methods, indicated host switches are likely the prominent force driving speciation in Tranzscheliella.

probabilities. The inferred phylogenetic trees were consistent with each other, and only the PhyML tree is shown (Figs 1 and 2). Tranzscheliella spp. formed a well-supported monophyletic clade in the Ustilaginaceae (Fig. 1). Thirty-three specimens provisionally identified as belonging to the T. hypodytes species clustered in the seven well-supported clades (Fig. 2). Thirty-five haplotypes of specimens provisionally identified as belonging to the T. hypodytes species complex, one as T. minima and one as T. williamsii, were used for coalescent analyses. The single-threshold general mixed Yule coalescent (GMYC) supported ten putative species, but this species delimitation scenario was not well supported by the likelihood ratio (LR) test (single-threshold: LR = 5.670471, P = 0.0587047). The multiple-threshold GMYC model provided a better fit to the ultrametric tree than a null model of uniform coalescent branching across the entire tree (multiple-threshold: LR = 7.168903, P = 0.02775189), which supported the delimitation of the taxa into thirteen putative species. The species delimitation results from GMYC and PTP analyses are summarized in Fig. 2. There was a high congruence between the PTP and multi-loci phylogenetic analyses. Both PTP and multiple-threshold GMYC analyses recovered six clades. Two clades formed single PTP groups, but multiple-threshold analysis separated each of these clades into two to three subclades. Another clade was recovered as a single group by phylogenetic analyses, but multiple-threshold GMYC and PTP analyses split this clade into two and three subclades respectively (Fig. 2). Based on concordant results from GMYC, PTP models and phylogenetic analyses, nine strongly supported clades were resolved, which represented four new species, a new combination, T. minima, a reduced T. hypodytes s. lat., and an unidentified Tranzscheliella sp. from South America. The pairwise identity of ITS sequences derived from the type of each species is shown in Table 2.
Cophylogeny analysis. The co-evolutionary relationships of the host and fungi are shown in Fig. 3. The global ParaFit test indicated that congruence between the phylogenies of Tranzscheliella species and their hosts was not significant (P = 0.50505) ( Table 3). This indicated that co-speciation was not the major evolutionary force driving pathogen diversity and distribution on hosts. For the event based approach, all the reconstructions under different cost regimes were significantly better than those generated in the randomized test. Although different cost values were assigned to duplication, loss/sorting and failure to diverge, the event number inferred from analyses remained constant (0-1 duplication, 5-6 loss/sorting and 5 failure to diverge). The lowest costs were yielded by cost regime four and six, which penalized cospeciation. These two reconstructions comprised 0 cospeciation, 0 duplication, 6 host switches, 6 loss and 5 failures to diverge (Table 4). Taxonomy. Schlechtendal 20 first described Caeoma hypodytes, which was subsequently transferred to several genera, namely, Ustilago, Erysibe, Uredo, Cintractia and Tranzscheliella. Hirschhorn 21 considered that Ustilago hypodytes was a nomen dubium and proposed a neotype (referring to it as a lectotype) on Elymus arenarius (the type host) collected in 1884 by P. Sydow near Berlin, Germany, which had the advantage of being widely distributed in Rabenhorst's Fungi Europea Exsiccata, Ser. 2, no. 3201. This species was subsequently transferred to T. hypodytes 8 . The nomenclature and taxonomy of T. hypodytes is confused, with numerous synonyms as well as misidentified hosts reported in the scientific literature 1 . Tranzscheliella hypodytes has long been recognized as a species complex rather than a single species 9 .
DNA could not be extracted from an isoneotype (HUV 3784) of T. hypodytes. Further, we were unable to obtain a more recent European specimen of Tranzscheliella on Elymus arenarius. Morphologically, T. hypodytes has spore walls that are smooth under light microscopy and densely, minutely, uniformly verruculose under SEM (p. Sori in the culms and surrounding the upper internodes and axes of abortive inflorescences, initially covered by the leaf sheath, finally exposed, peridium absent, upper internodes and leaves reduced in size. Spore mass semi-agglutinated to powdery. Spores globose, ovoid, ellipsoidal to slightly irregular, 4.5-5.5 × (3.5− ) 4-4.5 (− 5) μ m, light olive-brown; wall c. 0.5 μ m, surface smooth, in SEM moderately, unevenly verruculose, punctuate between warts.   Table 3. Results of the cophylogenetic analyses with the distance-based approach ParaFit.    Note -The Chinese specimens of Tranzscheliella on Elymus dahuricus and Leymus secalinus (subfamily Pooideae, tribe Triticeae) formed an unresolved polytomy in the phylogenetic analysis (Fig. 2). There is a likelihood that this clade will contain T. hypodytes s. str., as the neotype was collected on Leymus arenarius from Germany in 1884 21 . Of note is that two specimens on Elymus dahuricus, which is native to Siberia, Mongolia and northern China, formed a strongly supported subclade that may represent a novel species. Taxonomic resolution of this polytomy needs to wait until the neotype of T. hypodytes has been sequenced and further specimens of Tranzscheliella on other triticoid grasses have been examined. Further, the spore morphology of the Chinese collections of T. hypodytes on species of Elymus and Leymus was similar to the type specimen of T. hypodytes on L. arenarius as compared with SEM images in Vánky 1 (page 1007). Vánky 1 listed several species of Elymus and Leymus as hosts of T. hypodytes s. lat., which is the classification that we assign to this clade. Sori in the culms and surrounding the upper internodes and axes of abortive inflorescences, initially covered by the leaf sheath, finally exposed, peridium absent, upper internodes and leaves reduced in size. Spore mass semi-agglutinated to powdery. Spores globose, ovoid, ellipsoidal to slightly irregular, (4. Note -Tranzscheliella lavrovii occurs on Cleistogenes hackelii (subfamily Chloridoideae, tribe Cynodonteae), which has synonyms in Diplachne and Kengia that were considered as hosts for existing names in Tranzscheliella. Vánky 1 lists Diplachne spp. as a host for four species of smut fungi, T. amplexa, T. hypodytes s. lat., T. serena and U. ornata. Of these, only T. amplexa and T. hypodytes s. lat., have small spores similar in size to T. lavrovii. However T. lavrovii has more densely verruculose spores in SEM than T. amplexa. In the phylogenetic analysis, T. lavrovii was distinct from other species studied, having ITS similarity ranging from 90-95% identity ( Table 2). Tranzscheliella lavrovii has slightly larger spores than the isolates of Tranzscheliella sp. on Stipa papposa Sori in the culms and surrounding the upper internodes and axes of abortive inflorescences, initially covered by the leaf sheath, finally exposed, peridium absent, upper internodes and leaves reduced in size. Spore mass semi-agglutinated to powdery. Spores globose, ovoid, ellipsoidal to slightly irregular, 3.5-4 (− 4.5) × 3-4 μ m, light olive-brown; wall c. 0.5 μ m, surface smooth, in SEM spore surface densely verruculose with irregular warts that fuse to create an irregular pattern on the spore surface. Note -Tranzscheliella linguoae is one of four species of Tranzscheliella that infects species of Achnatherum (subfamily Pooideae, tribe Stipeae) 1 , which is another large polyphyletic grass genus 22 . The other species are T. jacksonii, T. minima and T. williamsii 1 . Tranzscheliella linguoae has smaller spores than T. jacksonii (8-13.5 × 8-12 μ m) and T. williamsii (7-10 × 6-8 μ m) 1 . The sori of T. linguoae lack a peridium and differ from T. minima, which has sori with a silvery to whitish fungal peridium 1 . In the phylogenetic analysis, specimens of T. linguoae were resolved in a well-supported monophyletic clade (Fig. 2).

T. williamsii
Tranzscheliella reverdattoana (Lavrov)  Note -Sorosporium reverdattoanum was described from a specimen of Lasiagrostis splendens (= Achnatherum splendens) (subfamily Pooideae, tribe Stipeae) collected in Kazakhstan 23 . Vánky 24 observed that the spores of this specimen had passed through the alimentary tracts of insects, becoming agglutinated and hence the generic placement in Sorosporium. The host, A. splendens, is especially interesting as it was shown to form a highly supported monophyletic clade that was distinct from other Old World Stipeae 22 . Further, Hamasha et al. 22 suggested that a new genus based on A. splendens was warranted, but only after clarification of the highly polyphyletic Achnatherum.
In making this new combination, we do not accept that S. reverdattoanum is a synonym of T. minima (type on Achnatherum hymenoides, USA) as considered by Vánky 1,24 . Tranzscheliella reverdattoana and T. minima both have very small spores (4-6 × 3.5-5 μ m for T. minima) that are densely verruculose in SEM 1 . However T. reverdattoana has spore surfaces with punctate warts between the verrucose warts in SEM, which are not seen in T. minima 1,24 . There was sequence data on GenBank for a specimen identified as T. minima (DQ191251) on Stipa occidentalis (subfamily Pooideae, tribe Stipeae) from the USA, which was found to be sister to T. reverdattoana in our phylogenetic analysis (Fig. 2). Despite not having DNA sequence data from the type specimen of S. reverdattoanum, we have chosen to transfer this species to Tranzscheliella on the basis of the (i) similar morphology between the isotype of S. reverdattoanum and the Chinese specimens, (ii) relative proximity of the collections in neighboring countries, i.e. China and Kazakhstan, (iii) unique phylogenetic placement of A. splendens and (iv) molecular diversity between the North American isolate of T. minima (represented by DQ191251) and the Chinese isolates studied here. Sori in the culms and surrounding the upper internodes and axes of abortive inflorescences, initially covered by the leaf sheath, finally exposed, peridium absent, upper internodes and leaves reduced in size. Spore mass semi-agglutinated to powdery. Spores globose, ovoid, ellipsoidal to slightly irregular, Note -Tranzscheliella schlechtendalii is one of six species of smut fungi in the Ustilaginaceae that infect Calamagrostis (subfamily Pooideae, tribe Poeae), which is a large polyphyletic grass genus 25 . The other species include four Ustilago stripe smuts (U. calamagrostidis, U. corcontica, U. scrobiculata and U. striiformis) 26 and T. hypodytes s. lat. 1 . The Ustilago stripe smuts all have larger spores that T. schlechtendalii. Vánky 1 listed "? Calamagrostis epigeios" as a host of T. hypodytes s. lat., although a specimen was not found in Herbarium Ustilaginales Vánky. In the phylogenetic analysis, T. schlechtendalii was resolved on a long branch in a well-supported monophyletic clade that was sister to all other Tranzscheliella species except T. williamsii (Fig. 2).
Tranzscheliella sp. Figure 4J-L. Sori in the culms and surrounding the upper internodes and axes of abortive inflorescences, initially covered by the leaf sheath, finally exposed, peridium absent, upper internodes and leaves reduced in size. Spore mass semi-agglutinated to powdery. Spores globose, ovoid, ellipsoidal to slightly irregular, ( Note -Tranzscheliella sp. occurs on two closely related grass species, Jarava plumosa and Nassella mucronata, (subfamily Pooideae, tribe Stipeae) 27 in South America. Vánky 1 listed three South American species, Ustilago nummularia 28 , U. stipicola 28 and U. spegazzinii 29 , as synonyms of T. hypodytes s. lat., which may represent this species. In the phylogenetic analysis, Tranzscheliella sp. was resolved in a well-supported clade (Fig. 2). Further work is needed to determine the identity of this South American species. Note -Tranzscheliella yupeitaniae occurs on two closely related grass species, Leymus chinensis and Psathyrostachys juncea (subfamily Pooideae, tribe Triticeae) 30,31 . Leymus contains about 50 species found in temperate regions of China and North America, and Psathyrostachys about 10 species from Russia, Turkey and China 31 . Several species of Leymus were listed as hosts of T. hypodytes s. lat. by Vánky 1 . Tranzscheliella yupeitaniae has spores that are densely, unevenly verruculose in SEM, which differ from the densely, minutely, uniformly verruculose spores of T. hypodytes s. str. 1 (p. 1007). In the phylogenetic analysis, T. yupeitaniae was resolved in a strongly supported clade (Fig. 2).

Discussion
Many of the specimens examined were herbarium specimens more than 5 years old that had not been housed in environmentally controlled conditions. The extraction and amplification of DNA from these specimens was challenging, most likely because of DNA degradation. In term of genealogical information, the ITS and LSU (linked rDNA loci) equate to a single locus. GMYC and PTP are methods primarily intended for delimiting species in single-locus molecular phylogenies 32,33 , and the species boundaries proposed by these methods are consistent with the phylogenetic species concept 34,35 . The GMYC and PTP analyses used in this study meet the basic requirements of these two methods. The GMYC method has a tendency to over-split and generate biologically unrealistic putative entities 36 . In this study, T. reverdattoana, T. schlechtendalii and Tranzscheliella sp., formed single PTP groups, although multiple-threshold analysis separated each of these species into two subclades (Fig. 2). These subclades were not well supported by phylogeny, morphological characters and host affiliations. Tranzscheliella schlechtendalii was sister to all other Tranzscheliella spp., with a large molecular distance (ITS sequence identity 82-89%), indicating missing data or undiscovered species.
Traditional species recognition criteria for smut fungi have been based on morphological and ecological characters, with emphasis on sori, spores, sterile cells and columellae, as well as pathogenicity on specific hosts 1,13 . A high degree of host specificity in most smut fungi, as postulated by earlier mycologists, has been largely confirmed by phylogenetic studies [11][12][13][14]19,37,38 . In this study, phylogenetic analyses of specimens of Tranzscheliella recognized eight distinct species as well as a clade that we retain as representing T. hypodytes s. lat. These seven species, T. lavrovii, T. linguoae, T. minima, T. reverdattoana, T. schlechtendalii, T. yupeitaniae and Tranzscheliella sp., appear restricted to specific grass species or closely related grass species. The unidentified Tranzscheliella sp. was found on two closely related grass species, Jarava plumosa and Nassella mucronata, from South America. Most of the remaining specimens were collected from China (Gansu, Inner Mongolia, Ningxia, Xinjiang and Qinghai) and neighboring countries.
It is highly likely that more species of Tranzscheliella await discovery as only 13 grass host species were included in our study. Our data showed that specimens from the same host species in different geographical regions were genetically closer than the specimens from the same geographical region on different hosts. This indicates the importance of host-adaption in the process of speciation. Cophylogenetic analyses showed that host switch was the best explanation for speciation in Tranzscheliella.

Materials and Methods
Specimens were borrowed from Queensland Plant Pathology Herbarium (BRIP) and Herbarium Mycologicum Academiae Sinicae (HMAS) ( Table 1). Spores were mounted in lactic acid (100% v/v) and examined under the light microscope. Means and standard deviations (SD) were calculated from at least 20 measurements. Ranges were expressed as (min.−) mean − SD-mean + SD (− max.) with values rounded to 0.5 μ m if below 20 μ m and 1.0 μ m if above 20 μ m. Images were captured by using a Nikon Eclipse 80i camera attached to a Nikon DS-Fi1 compound microscope with Nomarski differential interference contrast. For scanning electron microscopy (SEM), dried spores were dusted onto double-sided adhesive tape, fixed on specimen stubs, sputter coated with gold, ca. 20 nm thick, and examined with a FEI Quanta 200 electron microscope. Nomenclatural novelties and descriptions were registered in MycoBank (www.MycoBank.org).
DNA extraction, PCR amplification and sequencing. Fungal spores were removed from herbarium specimens with a fine needle and placed in cell lysis solution. For host tissue, dissected leaf samples were frozen in liquid nitrogen and ground with a mortar and pestle. Genomic DNA was extracted with the Gentra Puregene ® DNA Extraction Kit (Qiagen, Valencia, USA) according to the manufacturer's instructions.
ITS was amplified with the primers M-ITS 1 11 and ITS4 11,39 . LSU was amplified with the primers LR0R/LR5 40 . For the host plant, plasmid DNA regions rbcL, ITS and trnH-psbA were amplified with the primers rbcLa-F/ rbcLa-R 41,42 , 17SE/26SE 43 and psbAF/trnHR 44 , respectively. The PCR protocols were conducted as described by Zhang et al. 45 , with annealing temperature 62 °C for ITS of smuts, 60 °C for LSU, and 56 °C for ITS of host plants, rbcL and trnH-psbA. PCR products were sent to Biomed (Beijing, China) for sequencing with the same primer pairs used for amplification. Contigs were assembled in Mega 5 46 .
Phylogenetic analyses. The DNA sequences included in this study (Table 1) were aligned online with MAFFT (mafft.cbrc.jp/alignment/server/index.html) (Katoh and Toh 2008) using the L-INS-i method. ML was implemented as a search criterion in RAxML 47 and PhyML 3.0 48 . GTRGAMMA was specified as the model of evolution in both programs. The RAxML analyses were run with a rapid Bootstrap analysis (command -f a) using a random starting tree and 1,000 ML bootstrap replicates. The PhyML analyses were implemented using the ATGC bioinformatics platform (available at: http://www.atgcmontpellier.fr/phyml/), with six substitution type and SPR tree improvement, and support obtained from an approximate likelihood ratio test 49 .
MrBayes was used to conduct a Markov Chain Monte Carlo (MCMC) search in a Bayesian analysis. Four runs, each consisting of four chains, were implemented until the standard deviation of split frequencies were 0.02. The cold chain was heated at a temperature of 0.25. Substitution model parameters were sampled every 1,000 generations and trees were saved every 1,000 generations. Convergence of the Bayesian analysis was confirmed using AWTY 50 (available at: ceb.csit.fsu.edu/awty/).

Coalescent-based species delimitation. GMYC analysis.
The combined ITS and LSU sequences were analysed under the single threshold model and the multi-threshold model. The alignments were stripped of non-unique haplotypes using Arlequin 3.1 51 . Haplotype alignments were used to generate gene trees using Beast 1.7.5 with an uncorrelated lognormal relaxed clock model 52 and nucleotide substitution model using the same parameters as in the Bayesian analysis. Four independent MCMC chains were run for 400,000,000 generations, with sampling every 10,000 generations, using the 'auto optimize' operators option, and a Yule tree prior. The effective sample size (ESS) of each run was determined using Tracer v1.5 and only trees with an ESS of at least 200 were kept 53 . Four separate tree files were combined by LogCombiner 54 (burnin = 40,000) with a reduced resample frequency of 200,000. The reduced tree samples were used to reconstruct the maximum clade credibility tree by TreeAnnotator 54 . The selected topologies were used to optimize the single-threshold and multi-threshold GMYC models online (http://species.h-its.org/gmyc/). PTP analysis. The RAxML gene trees were constructed using the same markers selected by GMYC analysis. The PTP analysis was conducted online (http://species.h-its.org/ptp/) with the following settings: 10,000 MCMC generations; thinning interval of 100 and burn-in of 0.2 34 .
Cospeciation analyses. TREEMAP 3b 55 was used to generate a tanglegram from the ML tree of Tranzscheliella spp. and their host plants. To assess cospeciation between the host and parasites, both distance-based and event-based methods were utilized for the cophylogenetic analyses. For each of these two analyses, the parasite topology was obtained by using PhyML analysis based on ITS and LSU alignment, including just one representative per putative species. The host topology was obtained by PhyML analysis based on rbcL, ITS and trnH-psbA alignment of the representative specimens. For the distance-based analyses of cophylogeny, COPYCAT 2.02 56 was used, which incorporated a wrapper for ParaFit 57 . The congruence between the host and parasites phylogenies were computed and statistical significance tests were assessed by comparing randomizing parasites and host association with 999 permutations 58,59 . Event-based analyses were run in Jane 4 60 . Jane 4 considers five types of co-evolutionary event, namely cospeciation, duplication, host switch, sorting and failure to diverge. As it is difficult to estimate the relative cost of events, a default event cost scheme (cospeciation = 0, duplication = 1, duplication and host switch = 2, sorting = 1, failure to diverge = 1) as well as 9 cost regimes derived from default one were tested. In all the analyses, the vertex-base cost model method has been implemented, with the number of generation has been set to 100, and population size to 300. And the statistical significance of reconstructions was evaluated with 1,000 random tip mapping permutations.