Sequence data collection

Publicly available sequence data for described Phytophthora species were downloaded for nine loci (28S, 60SL10, Btub, EF1α, Enl, HSP90, TigA, ITS, and CoxI) from GenBank, drawing on previous phylogenetic works, species descriptions for new species, and the molecular key developed as part of IDPhy (S1 Table) [14]. All of these are nuclear loci with the exception of the CoxI mitochondrial locus.

Unfortunately, many Phytophthora isolates in GenBank have either been misidentified or renamed as more studies are done [16]. We used ITS and CoxI datasets from the well- researched IDPhy types or extype collections where available. For the other loci, we drew on well-supported sequence data from prior phylogenies of the genus. For the newer species not included in previous phylogenies or IDPhy, we used sequence data from species description papers. We also reviewed the isolates used for taxa that have known misidentifications like P. richardiae to ensure we used sequence data from the correct species. By comparing our cultures and notes, the culture collection and notes of Mannon Gallegly, the fact sheets provided on IDPhy, and recent species descriptions we were able to synthesize a thorough and accurate dataset of Phytophthora sequences.

Building on this dataset, we sequenced seven of these loci (28S, 60SL10, Btub, EF1α, Enl, HSP90, and TigA) for two additional species of Phytophthora which were recently described, Phytophthora acaciae and Phytophthora betacei (S1 Table) [45, 46]. Phytophthora acaciae was described as a species in Brazil in 2019 and infects black wattle (Acacia mearnsii) [45]. Phytophthora betacei was described in Colombia in 2018 and is a pathogen of tree tomato (Solanum betaceum), an important crop species in that area [46]. This species was previously classified as P. andina lineage EC-3 in earlier published work [47, 48] and is of interest to us in a related project on phylogenetic placement of this species. We also tested placement of P. cryptogea, P. ramorum and P. nicotianae in the tree using multilocus sequence datasets generated in our lab.

Isolates of P. acaciae were obtained from Dauri Tessmen, Universidade Estadual de Maringá, Parana, Brazil. Isolates of P. betacei were obtained from Sylvia Restrepo, University of Los Andes, Colombia. DNA extraction was performed using the CTAB method as described previously [49]. The primers used to amplify each of the seven loci were from previously published works (S2 Table) [14, 25, 27, 28, 32, 50]. PCR reactions were done in 50 μL volumes. Each 50 μL reaction contained 5 μL of 10X PCR buffer (Genesee, SanDiego, CA), 2.5 μL dNTPs (2 mM per nucleotide), 2 μL each 10 μM forward and reverse primer, 1.8 μL MgCl₂ (50mg/mL), 0.25 μL BSA (20mg/mL), 0.2 μL Taq (5U/μL) (Genesee, SanDiego, CA), with the remainder to 49 μL as dd H₂O. The final 1 μL consisted of sample DNA. Thermal cycling protocol consisted of 94°C for 5 minutes; then cycles of 94°C for 2 minutes, an annealing step specific to the primers, and 72°C for 2 minutes; followed by a final extension period at 72°C for 2 minutes (S2 Table). For the primers amplifying the TigA and HS90 loci, the annealing step was for 30 seconds with temperatures ranging from 64°C to 62°C, and 12 cycles completed at each temperature. For the other regions, the annealing step was for 30 seconds with temperatures ranging from 60°C to 53°C, and 3 cycles completed at each temperature plus an additional fifteen cycles with an annealing temperature of 53°C. Temperature reductions in the annealing step over time, known as touchdown PCR, allowed for amplification of multiple loci in one thermocycler protocol. Detailed descriptions of the primers, their optimum annealing temperatures, and the sources they were adapted from can be found in S2 Table. Amplicons from the PCR reactions that were expected to contain the locus of interest were sequenced using Sanger sequencing at the Genomic Sciences Laboratory at North Carolina State University.

In total, sequence data was collected for 194 Phytophthora species, 30 informally described Phytophthora taxa, and 3 outgroups from related oomycete genera (S1 Table). Compared to the most recent phylogeny using this larger set of loci, this represents the addition of 50 new taxa [14].

Phylogenetic inference and visualization

Phylogenetic trees were inferred with 1000 bootstrap replicates under the GTRGAMMA model using RAxML version 8 via the CIPRES REST API implemented in the DeCIFR toolkit (https://tools.decifr.hpc.ncsu.edu/denovo) [51, 52]. We partitioned the data by locus for our phylogeny. We also rebuilt the tree using IQTREE with partitioning by codon as well as separate partitions for the third codon positions.

In total, three sets of phylogenetic trees were generated. First, a phylogenetic tree based on the eight nuclear loci was constructed after excluding taxa with sequence data for fewer than three loci. An additional phylogenetic tree consisting of only the mitochondrial CoxI locus was inferred for species that had sequence data available at this locus. Lastly, independent trees for each locus were also inferred separately. Trees for all nine loci were compared using the Hypha package module of Mesquite v3.51 implemented in the DeCIFR toolkit (https://tools.decifr.hpc.ncsu.edu/trees2hypha) [53, 54]. The tree based on the eight nuclear loci was selected for further analysis.

The resulting Phytophthora genus tree was uploaded into the Tree-Based Alignment Selector Toolkit (T-BAS version 2.3) developed by Ignazio Carbone and colleagues [36]. The T-BAS system was chosen because it has a phylogeny-based placement feature that allows incorporation of new taxa into an existing phylogenetic tree. Metadata for the taxa in the tree were collected using a custom Python script for web scraping with the package BeautifulSoup as well as manual data collection [55]. A portion of the metadata for some species was retrieved from IDPhy, a published, open-access Phytophthora database developed and curated by Gloria Abad and colleagues [16, 33]. Metadata for some traits was summarized into a smaller number of categories to facilitate visualization. For example, host range was characterized as ‘specific’ or ‘broad’ following the categories presented in [8]. Each species was manually encoded to fit into one of these categories based on the number of host plants and their relatedness. Phytophthora species with few closely related hosts were placed in the specific category; hosts in three or more plant families were considered broad. For species where host information was lacking, a no host described category was used. A full list of specific hosts was also retained in the metadata. Similarly, pathogen lifestyle was categorized into soilborne, aquatic, and/or aerial [33]. Above ground disease symptoms were categorized as aerial, while below ground root infections were categorized as soilborne. Specific pathogen lifestyle information was also retained in the metadata. Phytophthora species with unknown lifestyles were left blank. Sexual characteristics were encoded as homothallic (ie, self-fertile) or heterothallic (ie, outcrossing), where available, or coded as sterile using the definition of Brasier et al. [15]. All distribution data were both retained in full and summarized into a list of continents for each species. The visualization capabilities of T-BAS were used to add metadata such as clade, sexual characteristics, or lifestyle traits to the phylogenetic tree inferred from the sequence data. Users have the option to download single and multilocus DNA sequence alignments, metadata, and Newick-formatted trees for focal clades of interest.

Phylogeny validation

To test real-time phylogenetic placement onto the tree, sequence data for two newly described Phytophthora species sequenced in this study (P. betacei and P acaciae) were added to the phylogeny. We used P. betacei because it is a close sister species to P. andina but has a different mitochondrial haplotype and is of interest to us in related sequencing project. Phytophthora andina has a 1c mitochondrial haplotype and P. betacei is 1a.The new taxa were inserted in the tree using the phylogeny-based evolutionary placement algorithm (EPA) implemented in T-BAS [36, 56, 57]. Additionally, independently identified species of Phytophthora from nursery crop plants including P. nicotianae, P. tropicalis, P. palmivora, P. drechsleri, P. cryptogea, P. pseudocryptogea, and P. kelmania collected by graduate student and extension specialist Inga Meadows and species that infect rhododendron were placed in the multilocus tree using the Phytophthora reference tree as a backbone constraint tree and 1,000 bootstrap replicates in RAxML. A tutorial and practice FASTA files for placement of two “unknown” Phytophthora species can be found on the project GitHub repository (https://github.com/allisoncoomber/phytophthora_tbas) and in S1 File.

P. infestans microsatellite lineage classifier

A dataset of approximately 2200 Phytophthora infestans isolates of known lineages and their corresponding microsatellite genotypes were obtained from over thirty years of collecting by Jean Beagle Ristaino [4, 43, 58]. Using R, an applet was developed to compare newly typed isolates of P. infestans to this dataset of lineages [59]. For genotype identification, unknown isolates were compared to a reference genotype using Bruvo’s genetic distance algorithm, implemented by the R package Poppr [60, 61]. Bruvo’s distance was selected for comparing unknown genotypes to the reference because it compares the relative genetic closeness of samples regardless of the ploidy [60]. Phytophthora infestans has been documented to have variations in ploidy from diploid to tetraploid, with triploid clonal lineages being common [62]. Variations in ploidy can influence results of other common genetic distance methods, so Bruvo’s distance is a logical choice for this species [60]. A newly typed isolate was considered to belong to a genotype if it was within a threshold genetic distance of other isolates in that genotype (described below).

A user interface was developed for the applet using the R Shiny package. The SSR genotype classifier is linked to the USABlight.org website under the “Identify an SSR genotype” page. A tutorial describing how to use the classifier and example files can be found in S2 File.

A threshold genetic distance to differentiate lineages was estimated using the cutoff-predictor function from Poppr [61]. A histogram of Bruvo’s values for all pairwise comparisons also provided visual confirmation of the threshold value (S1 Fig). Biologically, this threshold distance approximates the distinction between clonally and sexually related pairs, allowing the classifier to distinguish clonal genotypes.

The applet was also tested using a minimum representative set of distinct reference genotypes representing global populations of P. infestans to investigate which genotypes were similar and likely to be confused for each other. In this case, the reference set was developed by selecting the most common microsatellite profile for each genotype in the dataset. The remainder of isolates were then used as the tester dataset. These parameters produced a similar cutoff threshold to that estimated by Poppr.

To test the sensitivity and specificity of the classifier, the reference dataset was split into five randomly selected, non-overlapping samples. Each sample was classified using the remainder of the reference dataset (i.e., those not included in the sample). For some lineages which had only one or a few representative isolates, sensitivity and specificity could not be estimated.

Live phylogeny of Phytophthora

A total of 194 formally described Phytophthora species and 33 affiliated taxa were included in the genus phylogeny (Fig 2 and S1 Table). The maximum likelihood phylogeny was inferred with RAxML and includes 8 concatenated nuclear loci (28S, 60SL10, Btub, EF1α, ENL, HS90, ITS, and TigA). The tree was rooted with three outgroups from related oomycete genera (Halophytophthora fluviatilis, Phytopythium vexans, and Elongisporangium undulatum) (Fig 2). A full list of included species and accession numbers for each locus can be found in S1 Table. There was strong support for the outgroups as separate and distinct from the rest of the Phytophthora genus, with the Phytophthora genus appearing as a monophyletic clade (Fig 2). Note no downy mildew pathogen sequences were used in our phylogeny as our focus was on Phytophthora, but the genus Phytophthora is known to be paraphyletic with downy mildews residing within provisional clades as outlined by Bourret et al (2018) [15, 63].

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Phylogeny of the Phytophthora genus.

Radial phylogeny of the genus Phytophthora inferred with maximum likelihood and 1,000 bootstrap replicates for an alignment of 8 concatenated nuclear genes. Coloring on the inner ring indicates clade. Colors on the middle ring indicate substrate. Colors on the outer ring indicate host range (broad, specific, or no host reported). The branches in the tree are drawn to scale and the scale bar represents 0.06 substitutions per site.

https://doi.org/10.1371/journal.pone.0283540.g002

The results of this phylogeny were largely in agreement with what has been previously presented [14, 25–27, 29–31]. There was strong evidence for twelve subclades within the genus, and the relationships between clades were mostly in agreement with previous studies (Fig 3). Clade 11 contained one species, P. lilii, and clade 12 contains P. castenatorum, P quercina, P. tubulina and P. versiformis. Fifty species or closely affiliated taxa were placed in our tree that were not reported in the previous phylogeny by Yang et al. (2017) (Table 1 and S2 Fig). Placement of these taxa was overall in agreement with their previous descriptions (references for each in Table 1).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Phylogeny of major Phytophthora clades.

Collapsed phylogenetic tree of the genus Phytophthora showing the relationships between the clades, outgroups, and species that do not fit into the conventional clade system. Clade 12 is highlighted in pink. Phylogeny is inferred using maximum likelihood and 1,000 bootstrap replicates for an alignment of 8 concatenated nuclear loci. Bootstrap values (in percent) are shown above each branch. The branches in the tree are drawn to scale and the scale bar represents 0.06 substitutions per site.

https://doi.org/10.1371/journal.pone.0283540.g003

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Recently described Phytophthora taxa in TBAS phylogeny along with a summary of their associated metadata and species description paper.

https://doi.org/10.1371/journal.pone.0283540.t001

Several taxa that were previously given only a general clade in their species description papers were resolved to the subclade level by our study, similar to Chen et al.’s (2021) recent phylogeny (Table 1). For example, P. prodigiosa, P. pseudopolonica, P. rhizophorae, P. estuarina, and P. cacuminis were all previously described as belonging to clade 9 in their species description papers. Here, we resolve them to the subclade or subclade cluster level as 9a3, 9a3, 9a1, 9a1, and 9b respectively (S3 Fig), matching the results of Chen et al. 2021. Additional subclade distinctions and their previous clade classifications are listed in Table 1. Notably, the addition of several new species closely related to P. quercina including P. versiformis, P. tubulina, and P. castanetorum, thus allowing for a more detailed resolution of clade 12 (Fig 3). This clade was previously referred to as 3b [16, 33].

The topology of the CoxI mitochondrial tree differed from the nuclear derived trees, (S3 and S4 Figs). Because of this discordance, the CoxI locus was included as a single locus tree, but not with the other nuclear loci in the multilocus tree.

Phylogeny-based placement of P. acaciae grouped it in clade 2d with sister species P. bishii (S5 Fig). Phytophthora betacei was placed in clade 1c near both P. infestans and P. andina (Fig 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Phylogeny of clades 1, 3 4, 5, and 12.

Collapsed phylogeny of the genus Phytophthora showing clades 1 (red), 3 (yellow), 4 (green), 5 (beige), and 12 (pink) in detail. Subclade values are shown as variations in color hue. Phylogeny is inferred using maximum likelihood and 1,000 bootstrap replicates for an alignment of 8 concatenated nuclear loci. Bootstrap values (in percent) are shown above each branch. The branches in the tree are drawn to scale and the scale bar represents 0.06 substitutions per site (0.003 substitutions per site in the inset).

https://doi.org/10.1371/journal.pone.0283540.g004

Collaborator Inga Meadows tested the phylogenetic placement tool using 80 Phytophthora isolates spanning seven species (P. nicotianae, P. tropicalis, P. palmivora, P. drechsleri, P. cryptogea, P. pseudocryptogea, and P. kelmania) from other sequencing work. Phylogenetic placements matched previous identification, with the exception of eight isolates of P. cryptogea that were reclassified as P. pseudocryptogea using T BAS. In addition, species of Phytophthora reported on rhododendron were also correctly placed in the tree.

Metadata

Metadata, such as sexual reproductive strategies (heterothallic, homothallic, or sterile) were collated across the genus (S6 Fig). Since 2005 the number of homothallic species reported has exceeded the number of heterothallic species (S7 Fig). In general, the frequency of Phytophthora associated with different lifestyles (soilborne, aquatic, or aerial) has increased with additional species reported in each category over time (Fig 5A). Over the past 10 years, new Phytophthora species have been described more often from soilborne (includes root infecting species) than aerial lifestyles (Fig 5A). However, clusters of multiple aquatic species are present in Clades 6, 7, and 9 (Fig 6). In recent years, there has been a significant expansion in the number of surveys of water for Phytophthora species, likely explaining these results.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5.

(A) Histogram of Phytophthora species in each lifestyle category over time. Histogram showing the number of Phytophthora species in each lifestyle category in major phylogenies published since 2000. (B) Histogram of Phytophthora species sorted by host ranges over time. Histogram showing the number of Phytophthora species in each host range category in major phylogenies published since 2000. Host specificity was characterized as “broad” or “specific”. Species with reported hosts in three or more plant families were characterized as "broad”, while species with reported hosts in two or fewer plant families were described as "specific”.

https://doi.org/10.1371/journal.pone.0283540.g005

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Phylogeny of Phytophthora clades 6, 7, and 9.

Collapsed phylogeny of the genus Phytophthora showing 6 (green), 7 (teal), and 9 (blue) in detail. Subclade values are shown as variations in color hue. Phylogeny is inferred using maximum likelihood and 1,000 bootstrap replicates for an alignment of 8 concatenated nuclear loci. Bootstrap values (in percent) are shown for each branch. The branches in the tree are drawn to scale and the scale bar represents 0.06 substitutions per site.

https://doi.org/10.1371/journal.pone.0283540.g006

The level of host specificity also varies widely across the genus. Host specificity was characterized as “broad” or “specific” based on the number of reported hosts the pathogen has as well as the relatedness of the hosts. If a Phytophthora taxon was reported on three or more plant families it was characterized as “broad,” otherwise it was considered “specific.” It is important to note that reported hosts is not necessarily equivalent to host range as future pathogenicity trials could reveal additional hosts. This dichotomy in reported hosts of Phytophthora species was also acknowledged by Chen et al. (2021) in their recent phylogeny. Species with a broad list of reported hosts were clustered in Clades 1, 2, and 7 (Fig 2). However, the ability to cause disease on a wide variety of hosts is also generally widespread, being found at least once in all clades (Fig 2). In general, more recently described species fall in the “specific” reported host category than the “broad” reported host category, although this may change as these species are more thoroughly studied for host range (Fig 5B).

Phytophthora infestans SSR classifier

In order to differentiate clonal lineages within Phytophthora infestans we estimated a threshold genetic distance to separate lineages. This threshold genetic distance was estimated as 0.099 by Poppr’s cutoff prediction function. A histogram of all pairwise comparisons of Bruvo’s genetic distance for the included lineages was bimodal, indicating a population with mixed forms of sexual and asexual reproduction, as is characteristic of P. infestans (S1 Fig). Pairs with a Bruvo’s genetic distance below the threshold occurred in the first peak of the distribution and indicated clonal relationships. In other words, the pair consisted of clones of the same lineage or genotype. Above the threshold was another peak in the distribution of Bruvo’s distances. Pairs falling in this range of the distribution represent lineages, which are distinctly different, likely as a result of sexual recombination. In other words, these are lineages that should not be considered the same genotype. This result visually confirmed the threshold genetic distance as estimated by Poppr’s cutoff prediction function. The SSR classifier was parameterized to use a cutoff threshold of a Bruvo’s distance of 0.099 in order to consider an unknown P. infestans isolate a match to a described lineage.

The reference dataset for classifying genotypes consisted of 2,176 isolates of P. infestans and their corresponding microsatellite genotypes obtained from over thirty years of collecting by Jean Beagle Ristaino. The overall accuracy of the classifier was approximately 0.98, with an unweighted Kappa statistic of 0.977. Various summary statistics were also calculated for each individual class (lineage, n = 36) that isolates could be placed into (S3 Table). Some of the isolates included in the reference dataset had one or only a few representative genotypes. For these isolates, accuracy, precision, and other summary statistics were not calculated because there were no closely matching isolates. Many of these isolates are no longer in circulation (most of the US lineages except US-23, US-8, and US-11) or were transient. Importantly, the SSR classifier was able to identify the US genotypes from the recent past that are currently circulating (S3 Table). As collection of isolates continues, the classifier will continue to grow more robust with the increase in reference data. A list of all representative genotypes used to develop the classifier and their SSR profiles can be found in S4 Table. This includes several genotypes for which there were too few isolates to perform statistical testing.

To evaluate the performance of the P. infestans classifier with incomplete SSR genotype data, a testing dataset was developed that included genotyped lineages which had at least 7 but no more than 11 SSR loci genotyped. Many of these isolates were from historic samples of the FAM-1 lineage. For incomplete data, overall accuracy was 0.968 and the Kappa statistic was 0.954.

Out of all the genotypes in the dataset, the SSR classifier was unable to distinguish between some genotypes at the decided threshold. Most of these multi-genotype groups contain one genotype that is still in circulation, as well as some genotypes that are no longer found. In this case, the genotype that is still in circulation is returned. If the multi-genotype group contains multiple genotypes still in circulation, all those genotypes are returned after querying the database. For example, genotypes US-6 and US-7 could not be distinguished from US-11 by our classifier. However, US-6 and US-7 were transient lineages that sexually recombined to produce US-11 [96, 97]. Of these three lineages, only US-11 has persisted and so this is the lineage name that is returned by the classifier.

Genus tree placements

The TBAS Phytophthora tree presented here is designed to serve as a starting point for a curated, living phylogeny of the genus that may include species that are the direct ancestors of other contemporary taxa. Including at least three nuclear loci for well-described Phytophthora taxa and additional, newly described taxa has allowed for the resolution of conflicts within the genus phylogeny. Setting a minimum of three nuclear loci allowed for the inclusion of Phytophthora taxa that have been described that have few sequenced loci, while still maintaining the robustness of the tree. The loci included vary from species to species depending on what sequence data is available for each species, but all include some of the eight nuclear loci listed above. The nine loci used in this study all have utility in differentiating Phytophthora species, but the utility of each locus varies across the Phytophthora clades. [32]. Here, we found that including at least three nuclear loci (we recommend Btub, TigA, and ITS) allowed for the differentiation of all the species included in our tree. For species identification, we recommend using both a nuclear placement comprising at least three loci and placement into the mitochondrial CoxI tree. Hybrid Phytophthora taxa can be challenging to distinguish from their parental taxa, as is seen with the P. x alni species complex in our phylogenies [15, 48, 101].

This phylogeny is current at the time of submission of this publication. As new species are described researchers can query the tree and request placement of the species into the phylogeny by sending an email to us and following the submission guideline in the tutorial (S1 File).

Our work has resolved some issues of clade placement of some species and expanded the currently published phylogenies. For example, the placement of P. quercina has varied since its inclusion in the first molecular phylogeny of the Phytophthora genus [25]. Cooke et al. placed P. quercina into clade 3 with P. ilicis [25]. In their 2008 phylogeny, Blair and colleagues placed P. quercina in clade 4 with P. arecae, P. megakarya, P. palmivora, and P. sp. “quercetorum” [27]. In 2014, Martin and colleagues found P. quercina to be in its own clade with P. sp. ohioensis and not part of the canonical clade system [27]. In 2017 Yang and colleagues grouped P. quercina along with P. sp. ohioensis back into clade 4 [14]. On IDPhy, P. quercina is categorized as part of clade 3b in the ITS phylogeny [33]. Recently several new relatives of P. quercina, P. versiformis, P. tubulina, and P. castanetorum, have been described. By incorporating these new sister species and using a multilocus approach, it has been confirmed that these four species comprise a new clade 12 [13, 16, 63]. The phylogeny presented here supports this conclusion, with P. quercina, P. versiformis, P. tubulina, and P. castanetorum forming a unique monophyletic clade referred to as clade 12. This clade is separate and distinct from the rest of clade 3, forming a sister group to clades 1 and 4 (Fig 3).

In their 2017 Phytophthora phylogeny, Yang and colleagues subdivided clade 9 into clades 9a and 9b, where 9b is monophyletic and 9a consists of three monophyletic subclades, 9a1, 9a2, 9a3 [14]. The addition of new species in our work supports this subdivision of clade 9 (Fig 6). Several species newly included in this phylogeny were previously characterized as only “clade 9” but here, we find support for a specific subclade (Fig 6 and Table 1). Phytophthora cacuminis was strongly supported as part of Clade 9b [74]. Phytophthora pseudopolonica, which was previously classified as clade 9, grouped closely with P. polonica in clade 9a3 [90]. The new species P. prodigiosa also clustered in clade 9a3, despite being classified as clade 9b by previous ITS-based phylogenies [33, 85]. Phytophthora estuarina, P. sp. lagoariana, P. sp. cuyabensis, and P. rhizophorae were all supported as members of clade 9a1 [79]. In general, the discovery of new species in clade 9 supports the current grouping of clade 9 into the subclades proposed by Yang et al. 2017 and by Chen et al. 2021 (Fig 6).

Clade placement was also confirmed for two species that we sequenced for this study including P. acaciae and P. betacei. Phytophthora acaciae was placed in clade 2d with close relatives P. bishii and P. frigida, which is in agreement with our previous work on this species (S5 Fig) [45]. Phytophthora betacei was confirmed as a member of clade 1c as previously reported (Fig 4) [46].

The mitochondrial CoxI locus was included in part because it has been a popular gene to use in barcoding in previous Phytophthora phylogenies [29, 30]. The topology of the CoxI mitochondrial tree differed from the nuclear derived trees, potentially because of uniparental mitochondrial inheritance, horizontal gene transfer, incomplete lineage sorting, and phylogenetic artifacts (S3 and S4 Figs). The mitochondrial tree we produced uses only the Cox1 locus, which provides less opportunity to differentiate haplotypes than a more detailed multilocus analysis. However, the Cox1 locus evolves much faster than some of the nuclear loci we used, so when used in combination with the nuclear genes, this locus may be useful to differentiate new haplotypes or new species. The CoxI locus was included as a single locus tree in theT BAS tool and is also available for use. It is not concatenated with the other nuclear loci in the multilocus tree.

The CoxI inferred tree is available for further analysis (S3 Fig and S1 Table). The presence of unexpectedly close relationships between CoxI loci for some species warrants further investigation into known and potential hybrids. For example, in the tree based on eight nuclear loci P. x serendipita is most closely related to P. aleatoria and P. alpina with very strong support (bootstrap value of 100) (Fig 4). However, in the tree inferred from the CoxI mitochondrial locus, P. x serendipita is a sister group to P. hedraiandra with very strong support (bootstrap value of 95) (S3 Fig). This indicates that P. x serendipita’s mitochondrial genome is more closely related to that of P. hedraiandra than the nuclear genome [95]. P. x serendipita is thought to be a hybrid of P. hedraiandra and P. cactorum [95]. Inheritance of the mitochondria from the P. hedraiandra parent would explain why it is closer to this species in the mitochondrial-based tree [95]. This is the case for the isolate included in our trees, but other isolates of this species may have P. cactorum as their mitochondrial parent and therefore place with P. cactorum in the mitochondrial phylogeny. More data, especially sequence data from additional mitochondrial loci, need to be collected for some species to further investigate potential hybridizations. Full mitochondrial genome genealogies would also be useful. The discordance of the CoxI tree with the other single locus nuclear trees, as well as previously published trees, led us to exclude the CoxI locus from our final tree (S4 Fig). Conflict between multilocus analysis and CoxI placement was also observed by Yang and Hong in their 2018 evaluation of Phytophthora markers [32].

Recently, it has also been proposed that the rps10 mitochondrial locus be used to barcode Phytophthora species [98]. Although both CoxI and rps10 can differentiate many Phytophthora species, it is important to include nuclear barcodes as well [63]. The frequency of hybridization events in the genus Phytophthora merits careful examination of both nuclear and mitochondrial genealogies in order to understand phylogenetic relationships [24, 94, 99, 101, 102, 103]. The addition of more mitochondrial loci into the CoxI tree such as rps10 might strengthen support for these relationships, which could further clarify potential hybridization events. The phylogenomic computational tools developed by Van Pouke et al, will also be useful for understanding species hybrids in the genus [24]. We recommend separate placements of genes of interest on both the nuclear and mitochondrial trees and examination of relationships on each.

There is a lack of strong support in the both the nuclear and CoxI tree for the relationships between some species in the 1c clade. In the nuclear tree, there is strong support (bootstrap value of 90) for P. infestans, P. betacei, and P. andina forming a clade. However, the relationship between the next most closely related species, P. mirabilis, and these species, has weak support (bootstrap value of 70). Although the CoxI tree can also be used to inform these relationships, the relationships are not resolved with strong support in this tree either, indicating the need for more mitochondrial genome sequences. Given the recent addition of two new species in clade 1c, P. urerae and P. betacei, more whole genome sequence data and whole mitogenome data [98] will be needed to confidently understand the evolutionary relationships among 1c clade species. This clade is economically relevant and important in evolution studies and studies on the origins of the P. infestans. The inability to separate P. betacei from P. andina with strong support in our tree indicates that these species may be conspecific. A more comprehensive, genomic phylogeny of species in this clade should be conducted to strengthen confidence in the phylogeny. A finer resolution of the evolutionary relationships of P. andina to P. betacei is in progress in a related genome sequencing project in our lab.

Recent evidence has also shown that many genera of downy mildews place phylogenetically within the Phytophthora genus [63]. The downy mildew species are not included in the resource presented here because they are outside the scope of this project. We also did not expand to the provision 16 clade system. Although the genus Phytophthora is paraphyletic in that it includes the downy mildews in several nested clades within the genus tree [63], it has been proposed that the genus name Phytophthora be retained because it is both biologically and structurally cohesive [15]. We agree with this work. Evidence for the cohesion of the Phytophthora genus and the lack of distinguishing morphological or lifestyle traits to separate out the individual species within Phytophthora clades is furthered by our work. Many distinguishing traits, such as reproductive mode and ecological “life style”, appear to be spread across clades in the genus without a strong pattern. This indicates a shared evolutionary relationship between all of the species in the Phytophthora genus, regardless of later convergent evolution into multiple groups of downy mildews.

Metadata

Incorporating the metadata into the phylogenetic visualization allows the user to see relationships in biological characteristics across the phylogeny. For example, the recent proliferation in the description of species that were isolated from aquatic habitats is shown to have occurred in multiple clades (Fig 2). The primary lifestyle of Phytophthora species (broadly aerial, soilborne, or aquatic) are widely distributed, with small clusters of taxa that have different lifestyles appearing in multiple clades (Fig 2). Several recently described species from aquatic systems are clustered together in clade 6a and clade 9 (Fig 6). It appears that the ability to have success in different substrates has evolved multiple times in the genus.

Other than clade 5 and clade 3, which have entirely homothallic species, every clade has representatives of both homothallic and heterothallic species (S6 Fig). Both homothallism (self-fertile) and heterothallism (outcrossing) have been hypothesized to be the basal state of the genus [25, 28]. Blair et al. 2008 agreed with Kroon and colleague’s assessment that homothallism was the basal state based on the lack of heterothallism in both Clades 9 and 10, the basal clades [27]. Here we show that heterothallic species are also present in these clades including P. intercalaris in clade 10 and P. irrigata and others in clade 9 (S6 Fig). The ancestral reproductive mode in Phytophthora is therefore unclear as more than 12 transitions from one state to another are shown in our most supported tree (S6 Fig). Thus, reproductive strategy may be more transient than previously assumed. Others have shown that environmental stimuli including stress, aging and even fungicides can shift species such as P. infestans from heterothallism to homothallism [100,101].

Reported host range from broad to specific is also a variable trait in Phytophthora and has been shown here and in other work [15] (Fig 2). Again, this indicates relative ease in the evolutionary switch from wide to narrow host range strategy. Additionally, Phytophthora species have been clustered into two invasiveness groups based on host range as either generalists or specialists [8]. The categorization of each taxon into broad or specific host range could provide useful information to determine the potential invasiveness of a given species and be of use to regulatory agencies attempting to conduct risk analysis [8, 10].

Comparison to other databases

Several efforts have been made over the years to collect disparate information on Phytophthora species across the genus to facilitate identification, phylogenetics and evolutionary research. Notable examples include IDPhy, Phytophthora-ID for species and lineage identification, PhytophthoraDB and a lucid, dichotomous key for species identification [19, 23, 33–35].

IDPhy presents a wealth of morphological information and images and other metadata for species identification and is a valuable resource [16, 33]. This website contains fact sheets for Phytophthora species, morphological traits and molecular identification protocols to facilitate species identification. Sequence data for two genetic loci (ITS and CoxI) are also provided for many Phytophthora species as barcodes for species identification and can be blasted against via Genbank. In order to maintain compatibility with IDPhy, we have used the same sequence data for the ITS and CoxI regions for the 162 species identified as types or ex types in the IDPhy molecular key. We have also made an effort to synchronize metadata, such as lifestyle (soilborne, aerial, or aquatic) and distribution, with that provided by IDPhy.

Phytophthora-ID enables phylogenetic placement of unknown strains based on only 2 genes—CoxI and ITS sequences, and is not as robust as our tool [34]. However, Phytophthora-ID does not contain the full global set of known SSR lineages for P. infestans and many SSR lineages from the EuroBlight database are missing. The synthesis of genetic information with morphological characteristics and other descriptions is also not addressed by Phytophthora-ID.

Recently, the genus Phytophthora was included in the fourth edition of the “Genera of phytopathogenic fungi” series (GOPHY4) [31]. This included a whole genus tree of 192 species. However, this tree used a limited number of loci (up to four) from both the nuclear and mitochondrial genomes. Phytophthora species are known to undergo interspecies hybridizations, and including loci with discordant inheritance patterns can result in misleading phylogenies [102–104]. By including both nuclear and mitochondrial loci, but in separate phylogenies, we can mitigate these issues.

With a rapidly expanding genus, the need to update phylogenies and metadata descriptions to include newly described species and provisional species is critical. This tool will be useful to post provisional new species and give researchers access to expand phylogenetically new or interesting understudied clades of Phytophthora as a community. The contribution of the database presented here is to synthesize the wealth of sequence data and information that has been collected on Phytophthora, and to provide a more robust phylogenetic framework for the genus that can be updated at the pace of discovery by researchers in the community. In contrast to other resources which use a single gene e-barcoding strategy for quick identification, we have developed a more detailed phylogenetic tool.

Future directions

By incorporating the Phytophthora genus phylogeny into the T-BAS system, researchers will be able to add metadata, sequence data, and taxa as new species are described and our understanding of the genus continues to expand. Metadata that is not presented here, such as taxonomic characteristics or phenotype data, could easily be added to the tree if of interest to the larger community. Adding a new species will require submission of the metadata and sequence data through the T-BAS website so the data can be curated before uploading. Maintaining a robust and rapidly updating phylogeny will allow the streamlining of scientific information on the genus to keep pace with the discoveries of new species and characteristics. Incorporating the metadata directly into the phylogeny may also provide new insights for Phytophthora researchers of where new outbreaks are occurring. We plan to expand the mapping capabilities within the T-Bas tool. Thus, a community of research experts involved in Phytophthora research will need to be designated to become validators for the genus level phylogeny and the P. infestans classifier.

Although the Phytophthora genus tree enables rapid phylogenetic placement of newly sequenced isolates, phylogenetic placement alone cannot be used for species level identification. In addition to sequencing the loci used in this study for phylogenetic placement in T-BAS, other methods such as morphology should be used to verify the species an isolate belongs to. Moving forward, many newly described species need to be more thoroughly studied. Several of the newly described species in this phylogeny have only a few sequenced loci, which weakens phylogenetic placement. For example, clade 9 remains difficult to fully resolve in part because only 3–4 of the loci are sequenced for several new species (P. psuedopolonica, P. cacuminis, P. estuarina, and P. rhizophorae). Furthermore, key metadata characteristics for many species, such as sexual strategy and hosts are missing. Going forward, we recommend at least three of the nuclear loci utilized here be sequenced for new species descriptions in addition to morphological requirements. Placement in both the nuclear and mitochondrial trees is also recommended for species identification. In choosing which loci to include at a minimum, we suggest consulting Yang and Hong (2018) to select what loci might be most useful for the clade(s) of interest. Significant differences at these loci and in morphological traits is needed to warrant the description of a new species. These requirements will clarify the existing phylogeny and enable more robust phylogenetic placements of new species. Adding information into the phylogeny is facilitated by the T-BAS system, so new information about Phytophthora species can be readily available to the community, without the need to publish an updated description.

A review of the phylogeny also shows that newer species are often being described from natural ecosystems, such as riparian buffers, aquatic and forested areas, as opposed to agricultural systems. Further exploration of these systems is warranted, as it will likely reveal more species of Phytophthora that are yet to be discovered [8].

The utility of this tool for the plant-pathogenic Phytophthora genus serves as a proof of concept for other pathogen phylogenies. A separate live phylogeny for downy mildew fungi would be a great additional tool for the oomycete research community. Enhancing pathogen phylogenies with metadata and live taxon placement will facilitate research into a diversity of pathogen species. This tool also emphasizes the sharing and standardization of data–including the phylogeny, multiple sequence alignments and sequence data, biological sample data, specimen vouchers and other associated metadata. The comprehensive nature of this tool has been enabled by the wealth of published Phytophthora research that has been collected and shared by our research group and many other groups. We intend to contribute to and facilitate this trend by synthesizing these data streams. The efficacy of the live phylogeny presented here depends on open data sharing of not only the database tool but of cultures of newly described species between researchers and deposit in respected culture collections by the global Phytophthora research community. Many but not all of the species described herein are available to researchers from the Ristaino Phytophthora Collection at NC State.