To improve species-level resolution within the “E. aquilegiae complex” and identify potential cryptic species, both single-locus and multilocus phylogenetic analyses were conducted (Fig. 1; Supplementary Fig. 1–7). While earlier analyses based on ITS+28S failed to support species-level distinctions, multilocus approaches revealed two well-supported clades corresponding to E. aquilegiae s. str. and E. hortensiae. Support for these species was also evident in single-locus trees. E. hortensiae was supported in the RPB2 (posterior probability = 0.99; Supplementary Fig. 6), and IGS (1.0; Supplementary Fig. 4). In contrast, E. aquilegiae s. str. received minimal support across single-locus trees, except for weak support in the CAM tree (0.7; Supplementary Fig. 1). There was high support for both species in the concatenated tree in both likelihood (≥70) and posterior probabilities (>0.9).
In addition to these species, several host-specialized lineages belonging to Erysiphe on Ranunculaceae formed distinct, well-supported clades in both ITS+28S and multilocus trees in a basal position to E. aquilegiae s. str and E. hortensiae. These include the recognition of the previously named E. nitida on Aconitum spp. and Delphinium spp. (tribe Delphinieae), E. clematidicola sp. nov., E. clematidis-montanae sp. nov., E. aquilegiae-coeruleae sp. nov., E. parnassiae, and E. thalictri sp. nov. Each of these taxa was supported in all trees for which sequence data were available.

Application of subspecific taxonomic ranks within the “E. aquilegiae complex”
Based on results of the multilocus analyses from specimens from throughout the world, Bradshaw et al. (2025a,c) identified many strongly supported subclades within the E. vaccinii complex, which reflect morphologically indistinguishable biological races, for which the taxonomic rank ‘forma’ was introduced. Feng et al. (2025) continued the application of this convention when deviding Sawadaea bicornis into two morphologically indistinguishable formae represented by well-supported subclades within the larger species clade. The conditions for the application of the rank forma in powdery mildews has been discussed, and the advantages and justifications for using a formal taxonomic rank, ruled by the Code (ICNafp, Turland et al., 2018) have been outlined (Bradshaw et al., 2025a). In contrast to the previously used informal designation “forma specialis” for cases of biological races detected in inoculation experiments, the category “forma” is subjected to clear rules (ICNafp), including the need for a formal description and the designation of a holotype. Furthermore, genetically characterized and established biological races described as “formae” allow objective identifications based on sequence analyses, which is not given in cross inoculation tests where reproducibility cannot be assured.
The conditions for the use of such “formae” must be critically outlined, above all in cases of genetic multilocus analyses, where local populations can be reflected in highly supported clades in phylogenetic trees due to the large amounts of genomic data. Moreover, applications of formae used in the case of E. vaccinii (Bradshaw et al., 2025a) cannot be fully applied to the intricate E. aquilegiae complex, in which genetically well-supported subclades that correspond to host data have not yet been sufficiently demonstrated. The two broadly distributed species, E. aquilegiae s. str. and E. hortensiae, are taxonomically complex, comprising isolates from multiple host species and exhibiting minimal evidence of host-associated divergence. We hypothesize that these lineages are in the early stages of speciation, and it seems host specificity has not developed as strictly in this system as in other powdery mildew systems. Considering that the splitting of these taxa into cryptic species remains unsupported, even in phylogenetic multilocus analyses with many genes employed, it is best to treat these complexes as species with wider circumscriptions. In multilocus analyses, and in some cases even in single loci analyses, the presence of well-supported subclades suggests the occurrence of some host-specific specialized races within these broadly distributed species, however clear morphological differences are usually not observable. These races may be recognized as formae—genetically fixed races showing biological or ecological specializations and/or host preferences—but caution is warranted when introducing new formae. Naming local races or individual populations should be avoided, as excessive subdivision would lead to an impractical and uninformative proliferation of formae. Determining whether observed genetic groupings represent local populations or races, rather than distinct formae, can be challenging; therefore, the following criteria should be taken into consideration: (1) A higher number of sequenced specimens, above that is common for species descriptions, should be involved. (2) The specimens should be from a wide geographical range and, when applicable, a wide sampling of hosts, particularly when the hosts are known to be infected in different regions of the world. Exceptions can be made in cases of endemic host plants with restricted distribution areas or agriculturally significant crops. (3) Multilocus/full genome analyses should always take precedence over ITS analyses as basis for decisions.
We prefer to accept only multilocus/full genome analyses as the basis for the introduction of formae, and our currently available ITS and multilocus analyses suggest possible occurrences of such formae, reflected in subclades, but the available data are not yet sufficient for final conclusions. To stabilize the taxonomy and allow future formae to be described as additional data, we have pointed out where potential (hypothetical) autonyms, E. aquilegiae “f. aquilegiae” and E. hortensiae “f. hortensiae” will fall once future formae are introduced (Fig. 1). These subclades within the E. aquilegiae and E. hortensiae clades are supported by multiple specimens in the concatenated and GS tree (Fig. 1 and Supplementary Fig. 3). E. hortensiae on its type host is also supported in the IGS tree (1.0).