Multi-locus phylogeny and taxonomy of an unresolved, heterogeneous species complex within the genus Golovinomyces (Ascomycota, Erysiphales), including G. ambrosiae, G. circumfusus and G. spadiceus

Background Previous phylogenetic analyses of species within the genus Golovinomyces (Ascomycota, Erysiphales), based on ITS and 28S rDNA sequence data, revealed a co-evolutionary relationship between powdery mildew species and hosts of certain tribes of the plant family Asteraceae. Golovinomyces growing on host plants belonging to the Heliantheae formed a single lineage, comprised of a morphologically differentiated complex of species, which included G. ambrosiae, G. circumfusus, and G. spadiceus. However, the lineage also encompassed sequences retrieved from Golovinomyces specimens on other Asteraceae tribes as well as other plant families, suggesting the involvement of a plurivorous species. A multilocus phylogenetic examination of this complex, using ITS, 28S, IGS (intergenic spacer), TUB2 (beta-tubulin), and CHS1 (chitin synthase I) sequence data was carried out to clarify the discrepancies between ITS and 28S rDNA sequence data and morphological differences. Furthermore, the circumscription of species and their host ranges were emended. Results The phylogenetic and morphological analyses conducted in this study revealed three distinct species named, viz., (1) G. ambrosiae emend. (including G. spadiceus), a plurivorous species that occurs on a multitude of hosts including, Ambrosia spp., multiple species of the Heliantheae and plant species of other tribes of Asteraceae including the Asian species of Eupatorium; (2) G. latisporus comb. nov. (≡ Oidium latisporum), the closely related, but morphologically distinct species confined to hosts of the Heliantheae genera Helianthus, Zinnia, and most likely Rudbeckia; and (3) G. circumfusus confined to Eupatorium cannabinum in Europe. Conclusions The present results provide strong evidence that the combination of multi-locus phylogeny and morphological analysis is an effective way to identify species in the genus Golovinomyces.


Background
Powdery mildews are obligate biotrophic ascomycetes that occur on a wide range of dicotyledonous and monocotyledonous host plants. The family Erysiphaceae has a nearly worldwide distribution, with the exception of the Antarctic region, and currently comprises around 900 species in 18 genera [1][2][3]. Golovinomyces was originally introduced by Braun [4] as a section of the genus Erysiphe (s. lat.) and was later raised to genus rank by Heluta [5]. Braun [6] and Braun and Takamatsu [7] accepted Golovinomyces as a distinct genus and established the new tribe Golovinomyceteae. Golovinomyces is characterized by having chasmothecia with mycelioid appendages, several, mostly 2spored asci, an asexual morph with catenescent conidia that lack fibrosin bodies, and mostly nipple-shaped appressoria. Golovinomyces currently encompasses 57 species and 5 varieties [1,[8][9][10][11][12][13]. Erysiphe cichoracearum [14] included nearly all of the species that are now assigned to Golovinomyces. Blumer [15,16] split E. cichoracearum sensu Salmon [14] into several species but continued to maintain the species E. cichoracearum in a very broad sense (covering collections on Asteraceae and on hosts of multiple other plant families). Braun [17] confined E. cichoracearum to powdery mildews on hosts of Asteraceae and assigned specimens on hosts belonging to other plant families to Erysiphe orontii. Phylogenetic analyses of Golovinomyces, based on ITS and 28S rDNA sequence data [18], suggested the co-evolution between Golovinomyces species and certain tribes of Asteraceae. Based on these results, Braun and Cook [1] introduced a much narrower species concept for this genus, which included two morphologically differentiated species on hosts belonging to the Heliantheae, viz., G. ambrosiae and G. spadiceus. However, in more detailed phylogenetic analyses of ITS and 28S rDNA sequences, including Golovinomyces species on Asteraceae hosts, Takamatsu et al., [19] found that powdery mildews on hosts of the Heliantheae (previously referred to as G. ambrosiae and G. spadiceus), on hosts of an Asian species of Eupatorium (G. circumfusus s. lat.) and on a multitude of other hosts, including those on other plant families, formed a single large, unresolved clade (lineage III in Takamatsu et al., [19]). The taxonomic interpretation of these results posed a serious problem since G. ambrosiae and G. spadiceus, as circumscribed in Braun and Cook [1], are two morphologically differentiated species. Hence, the resolution within phylogenetic trees based only on ITS sequences was in this case insufficient to discriminate closely allied species. Therefore, most subsequent authors followed the taxonomic treatment in Braun and Cook [1] and recognized G. ambrosiae and G. spadiceus as separate species within lineage III, based on morphological differences [20][21][22][23][24][25][26][27]. The morphological differences used to differentiate the species include above all, much broader conidia and dimorphic germ tubes belonging to the longitubus pattern within the Euoidium type of conidial germination in G. ambrosiae than in G. spadiceus [1]. Additional research has found G. spadiceus to be extremely plurivorous, occurring on hosts of the Heliantheae and other tribes of Asteraceae, e.g., Aster and Chrysanthemum [19], Chrysogonum [28], as well hosts of various other plant families, including Abelmoschus (Malvaceae) [29], Crotalaria (Fabaceae) [13], Persicaria (Polygonaceae) [11,13,30], Solanum (Solanaceae) [13], and Verbena (Verbenaceae) [13]. The taxonomic interpretation of the inclusion of a sequence obtained from a Japanese collection of powdery mildew on Eupatorium chinense in lineage III [19] caused an additional problem and raised the question whether the name G. circumfusus, originally described from Europe on Eupatorium cannabinum, is included in this species complex.
The purpose of the present study was to clarify and resolve the taxonomy of this Golovinomyces complex using a multilocus approach, based on ITS, 28S, IGS, TUB2 and CHS1 DNA sequences. Multi-gene analyses are currently the method of choice to analyze phylogenetically and taxonomically difficult complexes of plant pathogenic fungi, including Colletotrichum spp. [31,32]. However, there is minimal multilocus data for the powdery mildews currently available. Most of the research involves the intraspecific genetic diversity in species such as Blumeria graminis [33,34], Erysiphe japonica [35], E. necator [36,37], Podosphaera xanthii [38] and Golovinomyces orontii [39]. Recently, the geographic and temporal distributions of four genotypes found in E. gracilis var. gracilis were studied based on a combination of data from the ITS, 28S rDNA and IGS regions [40]. Comprehensive applications of multilocus approaches to solve complex taxonomic-phylogenetic problems connected with the species level classification of the powdery mildews are still lacking. The present study is the first to use a multilocus approach to solve species distinction issues within the Erysiphales. An additional issue regarding the taxonomic conclusions drawn from phylogenetic results is also addressed in this study. Older taxonomic names are often available, but the application and allocation of such names are usually problematic. Because species names are based on their type collections, epitypifications, with appropriate new material, and exepitype sequences tend to be the main method to overcome these obstacles and to determine the application of older names. During the current study, this issue was addressed using international collaboration.

Sampling
A total of 69 specimens belonging to Golovinomyces ambrosiae, G. circumfusus, and G. spadiceus were examined, including 39 samples collected in China in recent years and 30 additional specimens from Germany, Japan, Russia, Switzerland, and the USA. Furthermore, eight specimens, consisting of three samples of G. magnicellulatus, three samples of Neoërysiphe galeopsidis, a sample of Arthrocladiella mougeotii and a sample of Erysiphe kenjiana, were used for phylogenetic analyses in this study. All of the plant materials used in this study were collected in the public gardens with Latin names or some are common ornamental plants which were identified by ourselves. Among the 69 specimens, ISC-F-0076752, ISC-F-0076753, and ISC-F-0076754 were deposited in the Herbarium of Iowa State University Fungi of Iowa, and the rest voucher specimens were deposited in the Herbarium of Mycology of Jilin Agricultural University. Names of the host plants, fungal species, locations and years of collection, voucher numbers and newly sequenced multi-gene accession numbers for the nucleotide sequence database (GenBank) in this study are given in Table 1.

Morphological examinations
For microscopic examinations, fresh samples were mounted in sterile water, and dried specimens, scraped from the leaf surface with a clean scalpel, were mounted in a drop of lactic acid on a microscope slide. Slides were examined using light microscopy with the total magnification at 200 and 400 (Zeiss Axio Scope A1, Germany). Fresh conidia were examined for the presence or absence of fibrosin bodies. A minimum of 30 measurements were made of asexual and sexual fungal structures. Germination of conidia was examined following the method of Hirata [41].
To obtain sufficient DNA for sequencing, the DNA regions of TUB2 and CHS1 were amplified by two rounds of PCR with the same primer set. All PCR reactions were conducted in 25 μL volumes. The reaction components were 2.5 μL 10 × PCR Buffer (Mg 2+ plus) (TaKaRa, Japan), 2 μL dNTP Mixture (10 mM total, 2.5 mM each), 1 μL each primer (20 ng/μL), 2 μL of total genomic DNA, 0.1 μL Taq polymerase (TaKaRa, Japan) (5 U/μL) and sterile ddH 2 O up to a final volume of 25 μL. The PCR reactions were conducted under the following thermal cycling conditions: an initial denaturation step of 5 min at 95°C, 35 cycles of 1 min at 94°C, followed by 30 s at 52°C for annealing, and 2 min at 72°C for extension, and a final extension for 8 min at 72°C. A negative control that lacked template DNA was included in each set of reactions. PCR products were subjected to electrophoresis in a 1.2% agarose gel in 0.5× TBE buffer. The amplified DNA products were purified using Mag-MK PCR Products Purification Kit following the protocol of the manufacturer. Amplicons were sequenced in both directions with the same PCR primers using direct sequencing in a 3730xl DNA Analyzer (Applied Biosystems) by Sangon Biotech (Shanghai, China). The sequence reactions were conducted using the BigDye™ Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems) following instructions of the manufacturer.
The reaction components for the PCR conducted at the University of Washington were 5 uL AllTaq PCR Buffer (Qiagen, Germany), 0.5 uL dNTP mixture, 0.25 μL of each primer (100 uM), 2 μL of total genomic DNA, 0.5 μL, Taq Polymerase (Qiagen, Germany) and sterile ddH 2 O up to a final volume of 25 μL. DNA was purified by isopropanol precipitation. These sequences [(The 28S rDNA sequence from ISC-F-0076754 and IGS sequences from ISC-F-0076752 and ISC-F-0076753] were manually trimmed using Geneious version 11.0.2 (https://www.geneious.com) and deposited in GenBank.
All other new sequences obtained in the present study were edited by DNAMAN version 6.0 and BioEdit Sequence Alignment Editor version 7.0, and then deposited in GenBank ( Table 1). The ITS, 28S, IGS, TUB2 and CHS1 sequences were respectively aligned by ClastalW. Furthermore, a multilocus sequences alignment datasets file (ITS+28S + IGS + TUB2 + CHS1) including 40 strains from Table 1 was also used for phylogenetic analyses. The six alignments were further refined manually in MEGA 7.0 [49] and deposited in TreeBASE (http:// www.treebase.org/) under the Accession No. of S24404 (http://purl.org/phylo/treebase/phylows/study/TB2:S244 04). Phylogenetic trees were obtained from the sequence   "-" means failed to get sequence data using maximum parsimony (MP) in PAUP 4.0b [50]. The MP analyses were performed with heuristic search option using the tree bisection reconnection (TBR) algorithm with 100 random sequence additions to find the global optimum tree. All sites were treated as unordered and unweighted, with gaps treated as missing data. The strength of the internal branches of the resulting trees were tested with bootstrap (BS) analysis using 1000 replications. Tree scores, including tree length, consistency index (CI), retention index (RI), and rescaled consistency index (RC), were also calculated. Five phylogenetic trees were generated based on the ITS, 28S, IGS, TUB2 and CHS1 nucleotide sequences.

Phylogenetic analyses
Parsimoniuous trees were separately constructed based on sequences of five gene regions and their combination and the numerical data including the number of taxa and characters are shown in Table 3. The information of outgroup taxon for each phylogenetic tree was also included in Table 1. The phylogenetic trees based on the ITS and 28S rDNA sequences were topologically congruent and indicated that G. ambrosiae complex on many Asteraceae plants, including Eupatorium spp. from Japan, formed a single clade with 100 and 99% bootstrap support, respectively (see Additional files 1, 2: Figure S1, S2).
Golovinomyces circumfusus on E. cannabinum from Germany did not form a monophyletic group with G. ambrosiae complex in all phylogenies (see Additional file 1-5: Figure S1-S5 and Fig. 1). The phylogenetic tree of IGS was similar to ITS tree, with the G. ambrosiae complex formed a single clade with 100% bootstrap support based on the individual genes (see Additional file 3: Figure S3). However, the isolates from Helianthus spp. and some Zinnia spp. differed by one base from isolates on other host genera, and forming a subclade with 64% bootstrap support (see Additional file 3: Figure S3 pink clade). The G. ambrosiae complex included two groups, one identified as G. ambrosiae emend. (see Additional file 3: Figure S3 green clade) and the other as G. latisporus comb. nov. (see Additional file 3: Figure S3 pink clade), based on the phylogenetic analysis of the IGS. The G. ambrosiae complex in TUB2 and CHS1 trees was divided into two subgroups, viz. G. ambrosiae emend., including G. spadiceus with 91 and 85% bootstrap support respectively (see Additional files 4, 5: Figure S4, S5 green clade), and G. latisporus comb. nov. with 70 and 78% bootstrap support respectively (see Additional files 4, 5: Figure S4, S5 pink clade). In the G. ambrosiae emend. Clade the sequences of CHS1 from isolates on Ambrosiae artemisiifolia and A. trifida differed by one base from isolates on other hosts. Golovinomyces ambrosiae emend. is a plurivorous species that occurs on a multitude of hosts including, Ambrosia spp., multiple species from the Heliantheae and plant species   Fig. 1) were higher than in other trees that were constructed based on separate genes. Golovinomyces circumfusus on E. cannabinum from Europe, forming a single clade, represented a separate species based on the combined data analysis (see Fig. 1).
Notes: Braun and Cook [1] assigned Golovinomyces on host species belonging to Eupatorium s. lat. From the northern hemisphere, including Europe, North America and northern regions of Asia, to G. circumfusus. This species seems to be confined to its type host, E. cannabinum, as collections from Asian species of Eupatorium pertain to G. ambrosiae. The affinity and identity of North American collections on Eupatorium perfoliatum, Eutrochium maculatum (≡ Eupatorium maculatum), and Eutrochium purpureum (≡ Eupatorium purpureum) remain unclear since sequence data and results of detailed morphological examinations of the asexual morphs on these hosts are not yet available. Golovinomyces on these hosts is common in North America, including several collections distributed in exsiccatae    fig. 1b).
Notes: Golovinomyces latisporus occurs on various Helianthus species almost worldwide. Zinnia angustifolia and Z. elegans are additional hosts proven by means of molecular methods. Golovinomyces collections found on various Rudbeckia spp. are assigned to G. latisporus with respect to the characters of the anamorph, although multilocus sequence analyses are still lacking. Taxonomy of a recently published record of "G. spadiceus" on Helianthus annuus in the United States [54] is unclear and urgently requires multilocus analyses for species identification. The identity of Golovinomyces on Iva spp. (axillaris, frutescens, xanthifolia) has not been sufficiently studied.

Discussion
The taxonomic history of the powdery mildews allied to Erysiphe cichoracearum dates back to de Candolle, in Lamarck and de Candolle [68]. He described E. cichoracearum on Scorzonera hispanica and Tragopogon porrifolius. Salmon [14] widened the concept of E. cichoracearum considerably by assigning powdery mildew on numerous hosts of various plant families to this species, including Helianthus spp. In previous circumscriptions, E. cichoracearum was characterized by having ascomata with mycelium-like appendages, several usually 2-spored asci, and conidia formed in chains without fibrosin bodies [14][15][16][17]. Braun [62] described the asexual morph of powdery mildew found on Helianthus × laetiflorus in Germany as Oidium latisporum based on the differences in conidial characters (most notably broader conidia) from collections of E. cichoracearum on various other hosts. Later, Braun [63] introduced the name E. cichoracearum var. latispora based on holomorphic North American type material, and cited E. ambrosiae as a possible synonym. Heluta [69] reallocated E. cichoracearum to Golovinomyces. Braun and Cook [1] split G. cichoracearum into several species based on molecular analyses of this complex which suggested a coevolutionary relationship between Golovinomyces species and tribes of Asteraceae [18].
Golovinomyces on hosts of the Heliantheae was divided into two species, G. ambrosiae and G. spadiceus, distinguished by clear morphological differences in their asexual morphs [1]. Type material of E. ambrosiae was examined, and this name was used for powdery mildew on Ambrosia, Helianthus, Iva, and Rudbeckia spp. E. ambrosiae was characterized by having broad ellipsoidovoid, doliiform to somewhat limoniform conidia, 25-45 × 15-27 μm (when fresh) with a length/width ratio < 2 (1.3-1.9, mostly 1.4-1.6), and dimorphic germ tubes that were long and filiform (longitubus pattern with the Euoidium conidial germination type) and consisted of a varying percentage of shorter germ tubes that were often swollen at the tip (ordinary Euoidium germ tubes) [1]. Whereas, the conidial shape and size, as well as the conidial germination pattern of G. spadiceus agrees with the common Euoidium type. These morphological differences were not reflected in a comprehensive phylogenetic analyses based on ITS and 28S rDNA powdery mildews previously referred to as G. ambrosiae and G. spadiceus. In the phylogenetic analyses, G. ambrosiae and G. spadiceus formed a single undifferentiated clade (lineage III in Takamatsu et al., [19]). Furthermore, this clade also encompassed sequences obtained from Golovinomyces on Eupatorium chinense in Japan [referred to as G. circumfusus based on the circumscription of this species in Braun and Cook [1] and the assumption that all Golovinomyces collections on various Eupatorium species in Asia, Europe and North America pertain to a single species] as well as sequences from Golovinomyces on numerous Asteraceae hosts from several tribes and even other families. The extensive host range exhibited by clade 3 suggests the involvement of a plurivorous species.
Sequences from the five gene regions could not be obtained for all samples used in this study. The phylogenetic affinity of G. circumfusus could be clarified by the inclusion of sequences obtained from powdery mildew on E. cannabinum (type host) in Germany (type region). G. circumfusus on its type host does not cluster within the former "Heliantheae Clade" and is not closely allied with G. ambrosiae complex. It represents a well-supported species of its own, confined to E. cannabinum in Europe. Blumer ([16], p. 188) summarized results of previous inoculation tests carried out by himself and other authors and classified Erysiphe cichoracearum s. lat. on E. cannabinum as a biologically specialized form (f. sp. eupatorii), confined to this host. In order to stabilize the application of the old name Erysiphe circumfusa, described in the nineteenth century, an epitype has been designated. Powdery mildew on Asian Eupatorium spp. is not conspecific with G. circumfusus and pertains to a clade previously referred to as G. spadiceus [13]. This clade represents a plurivorous species on a wide range of hosts belonging to the Heliantheae and other tribes of Asteraceae as well as hosts of other plant families. However, the naming of this clade had to be corrected.
Sequences from Golovinomyces on Ambrosia spp. in Asia and North America do not cluster together with sequences obtained from Golovinomyces on Helianthus spp., but they pertain to the former plurivorous G. spadiceus. The morphological characters of the powdery mildew on Ambrosia also agree with that of G. spadiceus (the type material of Erysiphe ambrosiae contains chasmothecia, but the features of the asexual morph could not be properly examined). Hence, Braun [63] cited E. ambrosiae as a potential synonym of E. cichoracearum var. latispora. The application of the name E. ambrosiae in Braun and Cook [1], based on this questionable synonymy, must be classified as a misinterpretation. These results have nomenclatural and taxonomic consequences, viz., the older name Erysiphe ambrosiae, which has priority over G. spadiceus, is now the correct name for this plurivorous species, and G. spadiceus and its synonyms must be reduced to synonymy with G. ambrosiae. Finally, Golovinomyces on Helianthus spp., morphologically distinguished from the former G. spadiceus, turned out be genetically different as well (although undoubtedly closely allied to the latter species).
Since G. ambrosiae now represents an older name for the species previously referred to as G. spadiceus, it is necessary to rename the species on Helianthus. Hence, Oidium latisporum (= Erysiphe cichoracearum var. latispora), the oldest valid name for this taxon at the species level, is used as the basionym for the combination G. latisporus. This species is common with a near global distribution, and also occurs on Zinnia [sequences retrieved from Z. angustifolia (HAL 2338 F) refer to a German collection from a botanical garden in which the Zinnia grew close to Helianthus plants infected by G. latisporus]. Sequences retrieved from Z. elegans (HMJAU-PM91850) refer to a collection from the Sichuan province of China where no Helianthus plants grew. The powdery mildew on Rudbeckia coincides morphologically with G. latisporus. However, currently only ITS and 28S sequences are available [19]. Future examinations based on IGS, TUB2 and CHS1 are necessary to confirm the identity. In any case, the example of Zinnia shows that host plants of other genera, such as Helianthus or Iva, might also be infested by the two closely allied species, G. ambrosiae and G. latisporus. In order to answer this question, a combination of morphological examinations and phylogenetic analyses based on a multilocus approach are required in the future.

Conclusions
The phylogenetic analyses of multilocus sequence data, including ITS and 28S rDNA, IGS, TUB2, CHS1, and consideration of morphological characters enabled to resolve species delimitation in a heterogeneous complex within the genus Golovinomyces. The old names involved in this complex have been epitypified, providing ex-epitype sequence data, and three species were distinguished in the complex named G. ambrosiae emend. (including G. spadiceus), G. latisporus comb. nov. (≡ Oidium latisporum), and G. circumfusus confined to Eupatorium cannabinum in Europe. This research illustrated that such approaches are suitable and promising in cases of phylogenetically closely allied assemblages of powdery mildew species in which ITS analyses do not yield sufficient resolution.