Multigene phylogeny, taxonomy and reclassification of Hyaloperonospora on Cardamine

Based on sequence data from cox1, cox2, ITS and LSU rDNA, it is shown that at least six species of Hyaloperonospora occur on the genus Cardamine, most of which were commonly classified under Peronospora dentariae. Based on sequences from their type hosts, Peronospora dentariae, Peronospora cardamines-laciniatae, Peronospora dentariae-macrophyllae, Peronospora malyi and Peronospora nasturtii-aquatici are combined into Hyaloperonospora, and their circumscription is clarified. Hyaloperonospora cardamines-enneaphyllos is described as a new species from Cardamine enneaphyllos. The host range of Hyaloperonospora nasturtii-aquatici, described from Nasturtium officinale, is shown to extend to various Cardamine species. Host range of species is shown to be highly diagnostic, with no overlap in their host range, but species commonly cannot be distinguished by morphology alone. Both cox1 and cox2 are confirmed to be good markers for phylogenetic species delimitation of closely related Hyaloperonospora species on Cardamine.

Although in general the narrow species concept of Gäumann (1918Gäumann ( , 1923 has been shown to be more appropriate, there are numerous problems in detail. Gäumann (1918Gäumann ( , 1923 described numerous Peronospora species from Brassicaceae based on subtle morphological differences and on evidence of high host specificity obtained by crossinoculation studies; however, as cross-inoculation studies could be performed only on a very limited number of hosts, most of his species were mainly based on host range in combination with often subtle differences in conidial size and shape. Therefore, he commonly classified accessions of various related host species under the same species if they were morphologically similar. As a result, he unfortunately did not select types for the species he described, but only a list of material examined, which especially becomes a problem if accessions from several hosts were classified under the same species. Recently, the extensive molecular phylogenetic study of Göker et al. (2009a) showed that several of his species were highly polyphyletic, as accessions from related hosts classified under the same species name by Gäumann (1918) are often not closely related, raising the problem of the correct naming of species. For this, nomenclatural and taxonomic decisions like lectotypification of heterogeneous entities are necessary. In addition, sequences have to be available for the species originating from the type host before appropriate reclassification can be achieved.
Five of these problematic taxa for which no nomenclatural decisions could yet be achieved due to reasons discussed above include Peronospora dentariae, P. cardamines-laciniatae, P. dentariae-macrophyllae, P. malyi and P. nasturtii-aquatici. Peronospora dentariae has already been described by Rabenhorst (1859) from an Italian collection from Cardamine (Dentaria) heptaphylla, a rather rare montane species confined to the Jura, south-western Alps and Apennines, but Gäumann (1918Gäumann ( , 1923 also placed collections from various other Cardamine species in that species. Peronospora cardamineslaciniatae, P. dentariae-macrophyllae and P. nasturtii-aquatici have been described by Gäumann (1918), based on differences in conidial sizes. Peronospora cardamines-laciniatae was described for accessions from the North American Cardamine laciniata, P. dentariae-macrophyllae for accessions from the East Asian Cardamine leucantha (syn. C. macrophylla var. dasyloba), and P. nasturtii-aquatici for accessions from Nasturtium officinale. Based on his conidial measurements, Gäumann (1918) assumed that accessions from the European Cardamine bulbifera could also belong to P. cardamineslaciniatae, but later expressed strong doubts that they are conspecific (Gäumann 1923). However, Gustavsson (1959) found no significant differences in conidial size between accessions from various Cardamine species either classified as P. dentariae or P. cardamines-lacinatae and he considered them to be synonymous. Finally, Lindtner (1957) described a fifth species, Peronospora malyi, from Cardamine graeca, based on larger conidia and oospores.
The extensive investigation of Göker et al. (2009a) showed that Hyaloperonospora accessions from various Cardamine species were placed in five distinct subgroups, two of which were part of their clade 1, and three of their clade 6. However, as neither material from the type host of Peronospora dentariae, C. heptaphylla, nor from the type host of P. cardamines-laciniatae, C. laciniata, were available for study at that time, it remained unclear to which clades the names P. dentariae and P. cardamines-laciniatae should be applied. Therefore, the various Hyaloperonospora clades on Cardamine could not be properly classified. Due to these uncertainties, P. nasturtii-aquatici could also not be properly classified, as accessions from Nasturtium officinale were contained in a clade comprising accessions from various Cardamine species, e.g., C. amara, C. hirsuta and C. pratensis, which were all placed in the older P. dentariae by Gäumann (1918). No material from Cardamine graeca, the type host of P. malyi, was included in the investigation of Göker et al. (2009a), so the status of that species also remained unresolved.
To clarify these nomenclatural uncertainties, recent collections were obtained for the type hosts of all four species described from Cardamine. Four genes (ITS, LSU, cox1, cox2) were sequenced and analyzed for numerous accessions from Cardamine to evaluate the host ranges and species boundaries.

Morphological analysis
Conidiophores and conidia were removed from the underneath of infected leaves, transferred to anhydrous lactic acid on a slide, carefully torn apart using forceps and needles, shortly heated using an alcohol burner and covered with a cover slip. Slides were examined and photographed using a Zeiss Axio Imager.A1 (Zeiss, Jena, Germany) microscope equipped with a Zeiss AxioCam ICc3 digital camera. Measurements are reported as maxima and minima in parentheses and the mean plus and minus the standard deviation of a number of measurements given in parentheses.

Sample sources
Information on the samples used for morphological analysis, sequencing and phylogenetic analyses is given Table 1. DNA extraction, PCR and sequencing For DNA extraction, infected dry host tissue was placed in 2-ml reaction tubes together with six sterile 2-mm glass beads and ground in a Retsch 200 mixer mill for 10 min at a frequency of 30 Hz. DNA was extracted using the modified CTAB protocol described in Riethmüller et al. (2002).

Phylogenetic analysis
To reveal the phylogenetic position of the Hyaloperonospora clades on Cardamine, an ITS-LSU data matrix was used. For this, the new sequences generated during the current study were combined with those of accessions from Cardamine and Nasturtium included in Göker et al. (2009a); in addition, sequences of representative species from throughout the tree of Göker et al. (2009a) were added (Table 1), with Perofascia lepidii as outgroup to root the trees. All alignments were produced with Muscle v.3.6 (Edgar 2004). For evaluation of species status, a multi-gene analysis of all four genes (ITS, LSU, cox1, cox2) was performed; due to lack of data available for other species, only accessions of all six clades from Cardamine were included, with Peronospora rumicis as outgroup to root the tree. Prior to phylogenetic analyses, the approach of Wiens (1998) was applied to test for significant levels of localized incongruence among the two gene partitions, using the level of bootstrap support (Sung et al. 2007). For this, the 70 % maximum parsimony (MP) bootstrap trees of the individual gene regions (ITS-LSU, cox1, cox2) were compared, which were calculated using the same parameters as for the combined analysis given below. No topological conflicts were observed between these bootstrap trees of genes, indicating the absence of significant incongruence and combinability of the matrices (Wiens 1998).
Maximum parsimony (MP) analysis was performed with PAUP* v.4.0 b10 (Swofford 2002), using 1,000 replicates of heuristic search with random addition of sequences and subsequent TBR branch swapping (MULTREES option in effect, COLLAPSE=MINBRLEN, steepest descent option not in effect), each replicate limited to 1 million rearrangements. All molecular characters were unordered and given equal weight; analyses were performed with gaps treated as missing data. Bootstrap analysis with 1,000 replicates was performed in the same way, but using 5 rounds of random sequence addition and subsequent branch swapping during each bootstrap replicate.
For maximum likelihood (ML) and Bayesian analyses, the well-known general time reversible model (GTR) was selected by Modeltest 3.6 (Posada and Crandall 1998) using the Akaike information criterion for all three genes; with a gamma distribution for the ITS-LSU (GTR+G) and with invariant sites and gamma distribution for the remaining sites for cox1 and cox2 (GTR+I+G). In the combined analyses of all gene regions, substitution parameters were estimated separately for each region. For ML analyses, 500 rounds of random addition of sequences as well as 500 fast bootstrap replicates were computed with RAxML (Stamatakis 2006) as implemented in raxmlGUI 0.95 (Silvestro and Michalak 2012) using the GTRGAMMA and GTRCAT substitution models, respectively. For Bayesian analyses using MrBayes v.3.1.2 (Huelsenbeck and Ronquist 2001), three parallel runs of four incrementally heated simultaneous Markov chains were performed over 1 million generations from which every 100th tree was sampled in each run. The first 500 trees were discarded, and a 90 % majority rule consensus of the remaining trees was computed to obtain posterior probabilities. The final matrix was deposited in TreeBASE (http://www.treebase.org) and is available under http://purl.org/phylo/treebase/phylows/study/TB2:S14017.

Results
After the exclusion of excessive leading and trailing gap regions and large insertions present in some single sequences, 2,172 characters were included in the ITS-LSU analyses, of which 428 were parsimony informative. MP analyses revealed 2,106 MP trees of 1,213 steps which were identical except for topologies within species, a polytomy containing H. lobulariae, H. brassicae and H. lunariae, a polytomy containing H. hesperidis and the H. cheiranthi-H. sisymbrii-sophiae clade, and a polytomy containing H. cardaminopsidis and the H. arabidopsidis-H. praecox clade. Backbone topology of the deeper unsupported nodes of the ML tree obtained with RAxML differed from the MP strict consensus tree (data not shown), but topologies of supported nodes were the same as in the MP analysis. One of the 2,106 MP trees is shown as phylogram in Fig. 1, with MP and ML bootstrap support above 60 % and posterior probabilities above 90 % given at first, second and third positions above/below the branches, respectively.
The combined matrix contained 3,412 characters (2,150 from ITS-LSU, 681 from cox1, 581 from cox2), from which 374 were parsimony informative. MP analyses revealed 54 MP trees of 885 steps which were identical except for topologies within species. Tree topology of the ML tree obtained with RAxML was fully compatible with the MP strict consensus tree (data not shown). One of the 54 MP trees is shown as phylogram in Fig. 2, with MP and ML bootstrap support above 60 % and posterior probabilities above 90 % given at first, second, and third positions above/below the branches, respectively.
In the ITS-LSU analyses (Fig. 1), the accessions from Cardamine were contained in three distinct clades of Göker et al. (2009a), one consisting of H. dentariae within clade 1,  Fig. 2 Phylogram showing 1 of 54 MP trees inferred from the combined ITS-LSU-cox1-cox2 sequence data matrix. MP and ML bootstrap support above 60 % and posterior probabilities above 90 % are given at first, second, and third positions, respectively, above/below the branches harboured only a single Hyaloperonospora species. The accessions from the previously not sampled C. enneaphyllos formed a distinct clade representing a new species described as H. cardamines-enneaphyllos below. Hyaloperonospora cardamines-enneaphyllos, H. malyi and H. dentariaemacrophyllae each contained only accessions from a single host species, whereas within H. cardamines-laciniatae, H. dentariae and H. nasturtii-aquatici accessions from several host species were placed. Within H. dentariae and H. nasturtii-aquatici, some substructure was observed in the combined analyses according to the host species, indicating the presence of host-specific lineages within species. Within the H. dentariae clade, the accessions from C. heptaphylla and C. pentaphyllos formed a highly supported monophylum, the latter again forming a highly supported subclade; however, accessions from C. bulbifera and C. impatiens were not contained in monophyletic lineages. Within H. nasturtiiaquatici, the accessions from C. hirsuta and from C. flexuosa each formed weakly supported subclades, the latter being included in a moderately supported clade together with the accessions from C. pratensis.
Habitat: On living leaves and stems of Cardamine enneaphyllos (Brassicaceae).

Discussion
Phylogenetic analyses of the ITS-LSU data are largely congruent with the results of Göker et al. (2009a) concerning the Hosts in bold denote type hosts. Note the variability of sporangial sizes commonly observed within collections from the same host, within the same Hyaloperonospora species, and the overlap in size range between most Hyaloperonospora species a range measurements listed in Gäumann (1918: 528) erroneous; approximate range was estimated from spore size line graphs (Gäumann 1918: 466) placement of Hyaloperonospora accessions from Cardamine ( Fig. 1) within their clades 1 and 6. However, there are differences in the deeper unsupported nodes of the tree which may be due to the more extensive taxon and accession sampling of Göker et al. (2009a). This is to be expected, as the deeper nodes of trees inferred from ITS-LSU data mostly lack support. Lack of support for deeper nodes in ITS as well as LSU phylogenies has also been reported from other downy mildew genera like Peronospora (e.g., Voglmayr 2003;Choi et al. 2007;García Blázquez et al. 2008;Göker et al. 2009b) or Plasmopara (e.g., Voglmayr et al. 2004;Voglmayr and Constantinescu 2008;Voglmayr and Thines 2007).
In the combined analyses, resolution as well as support within the main species clades is improved, showing that cox1 and cox2 add substantial resolution to the tree. cox1, chosen as barcoding locus for higher animals and considered to be the primary barcoding marker for organisms unless shown to be unsuitable (http://www.barcodeoflife.org), has also been shown to be an appropriate barcoding locus for oomycetes (Robideau et al. 2011), which is confirmed in the current study. cox2 shows similarly good resolution and may serve as an accessory barcoding marker; it also has some advantages over cox1, as it usually amplifies better especially in cases of low DNA quantity or older degraded samples (as also shown in Telle and Thines 2008), and thus cox2 sequences are available for many more species.
The results of the current study clearly show that the classification of Hyaloperonospora accessions from Cardamine proposed by Gäumann (1918Gäumann ( , 1923 has to be substantially revised. In his classification, accessions from Cardamine heptaphylla, C. impatiens, C. amara, C. flexuosa, C. hirsuta and C. pratensis were placed in Peronospora dentariae. However, these are actually contained within two distinct clades in the molecular phylogenies, the first two being included in clade 1 as H. dentariae, whereas the latter four are part of clade 6 of Göker et al. (2009a), belonging to H. nasturtiiaquatici (Figs. 1, 2). Likewise, his P. cardamines-laciniatae contained accessions from C. bulbifera and C. laciniata which are phylogenetically also contained within clade 1 and 6, respectively, the former being placed in H. dentariae. This is understandable, as the conidial sizes and shapes, which were primary criteria for species classification, are similar between most of these phylogenetically distinct entities (Table 2), and other features like conidiophore morphology are highly variable and also unsuitable for morphological distinction (Gäumann 1918, own observations). In addition, conidial sizes show some variability between different collections of the same species (Table 2) and may be dependent on host, environmental conditions, and ontogenetic state of the collection, which has been recently confirmed for Pseudoperonospora cubensis (Runge et al. 2012). The inability to distinguish them morphologically led Gustavsson (1959) to synonymize P. cardamines-laciniatae with P. dentariae. On the other hand, Lindtner (1957) accepted the classification of Gäumann, and distinguished the various species by their deviating mean lengths and widths. Based on conidial sizes, Săvulescu and Rayss (1930) classified accessions from C. enneaphyllos under P. dentariae-macrophyllae, which was subsequently followed by Lindtner (1957) and Kochman and Majewski (1970).
The data on conidial sizes (Table 2) show that most phylogenetic species occurring on Cardamine cannot be distinguished by morphology alone, at least not by characters usually used for species identification; they should be considered cryptic species which is quite common amongst downy mildews (Voglmayr 2008). However, the hosts are diagnostic for the different Hyaloperonospora species investigated in the present study. Within Hyaloperonospora, usually only a single species is observed on a given host species. In exceptional cases where more than one Hyaloperonospora species has been observed to occur on a single host species, they were morphologically quite distinct (Voglmayr and Göker 2011). No such case of overlapping host range has yet been observed in Hyaloperonospora on Cardamine despite extensive sampling. However, reliable species identification is nowadays only possible by sequence data, especially in groups which are incompletely sampled.
Remarkably, H. nasturtii-aquatici contains accessions from two distinct genera, Cardamine and Nasturtium. Long considered to be closely related to Rorippa and sometimes even classified within that genus, it has been shown that Nasturtium actually is the closest relative of Cardamine (Al-Shehbaz and Price 1998; Franzke et al. 1998). The conidial size recorded for Hyaloperonospora accessions from Nasturtium officinale (range 15-27×13-21 μm, av. 20 × 17 μm; see Table 2) fits well the measurements recorded for accessions from Cardamine species placed within H. nasturtii-aquatici (range 14-25×12.5-22 μm, av. lengths 17.3-20.5 μm, av. widths 15-17.5 μm; see Table 2). Some phylogenetic substructure was observed within both H. dentariae and H. nasturtii-aquatici, where subclades were formed according to the host species (Fig. 2), which indicates some host specificity also within species. This is evidence for active evolutionary radiation, which may result in speciation events, and some of these subclades may actually represent taxonomically separable entities on the subspecific level. Therefore, these species may be good candidates for population genetic investigations on a larger scale to investigate host specificity, gene flow, and speciation processes in detail. Remarkably, similar phylogenetic substructures were also observed in Albugo on various Cardamine species, which either formed phylogenetically distinct species or showed distinct substructures within a species according to their hosts (Ploch et al. 2010). This may imply that a common evolutionary pattern may be present in both Albugo and Hyaloperonospora on Cardamine, and a comparison of their divergence patterns could give some more general insights in the evolution of obligatory parasitic oomycetes in future studies.