Evolutionary dependence of host type and chasmothecial appendage morphology in obligate plant parasites belonging to Erysipheae (powdery mildew, Erysiphaceae)

ABSTRACT Evolutionary relationships between the morphological and ecological traits of fungi are poorly understood. The appendages of chasmothecia, which are sexual reproductive organs of Erysiphaceae, are considered to play a crucial role in the overwintering strategies of these fungi on host plants. Previous studies suggested that both the host type and appendage morphology evolved at the same nodes and transitioned from complex appendages on deciduous hosts to simple appendages on herb/evergreen hosts. However, the evolutionary dependence between host type and appendage morphology remains unproven owing to the limited species data used in analyses. To elucidate the evolutionary relationship between host type and appendage morphology, we used phylogenetic comparative methods (PCMs) to investigate the state transition, ancestral state, evolutionary dependence, and contingent evolution within Erysipheae, the largest and most diverse tribe in Erysiphaceae. Our PCMs, based on a comprehensive data set of Erysipheae, revealed that the most ancestral states were deciduous host types and complex appendages. From these ancestral states, convergent evolution toward the herb/evergreen host types and simple appendages occurred multiple times at the same nodes. For the first time in Erysiphaceae, we detected an evolutionary dependence between host type and appendage morphology. This is one of the few examples in which evolutionary dependence between host phenology and morphological traits in plant-parasitic fungi was demonstrated using PCMs. Appendage simplification on herb/evergreen hosts and complications on deciduous hosts can be reasonably explained by the functional advantages of each appendage type in different overwintering strategies. These expected appendage functions can explain approximately 90% of host type and appendage morphology combinations observed in the analyzed taxa. However, our results also highlighted the occurrence of evolutionary shifts that deviate from the expected advantages of each appendage morphology. These seemingly irrational shifts might be interpretable from the flexibility of overwintering strategies and quantification of appendage functions.


INTRODUCTION
The function of morphologies and their adaptive evolution constitutes a fundamental issue in evolutionary biology; however, existing research has primarily focused on visible plants and animals (Emlen 2008;Petersen and Kellogg 2022;Suzuki 2017).Fungi lack numerous distinctive morphological characteristics compared with plants and animals and thus have not received as much attention as macroorganisms regarding their morphological evolution (Nagy et al. 2011;Shirouzu et al. 2022).In particular, knowledge on the evolutionary relationships among the morphological and ecological traits in fungi is still limited (Aguilar-Trigueros et al. 2023;Janošík et al. 2023).
Obligate parasitic fungi, which are dependent on living host cells for nutrition (Gilbert and Parker 2023), are considered to be influenced by interactions with host organisms during their trait evolution.Thus, their morphological traits could provide evidence of adaptive evolution on a specific host.In the current study, we focused on a group of obligate plant parasites, the powdery mildews, as suitable models for exploring the evolutionary relationship between fungal morphology and host plants.Powdery mildews are ascomycetous fungi belonging to Erysiphaceae, which includes approximately 1000 species in 19 genera and represent globally noteworthy pathogens that cause diseases on approximately 10 000 species of angiosperms (Amano 1986;Braun and Cook 2012).Species of Erysiphaceae can reproduce both asexually through conidia and sexually via sexual reproductive organs known as chasmothecia.Chasmothecia have appendages with characteristic morphologies, such as mycelioid, circinate, or branched (FIG.1).These appendage morphologies are considered important taxonomic traits of the fungal group (Braun and Cook 2012).Moreover, chasmothecia are thought to function as overwintering (or oversummering) organs (Spencer 1978), and ascospores released from the overwintered chasmothecia establish infections on the leaves of their host in the following year.Morphological differences in chasmothecial appendages are thought to be related to the overwintering strategies of each fungus on different hosts.For example, chasmothecia with complex appendages, either circinate or branched, have been observed to overwinter by moving from the original substrata (e.g., leaves) to secondary substrata (e.g., branches and trunks; Gadoury and Pearson 1988;Grove and Boal 1991).In contrast, chasmothecia with simple appendages, such as mycelioid, are thought to overwinter while remaining on the original substrata (Gadoury et al. 2010).Takamatsu et al. (2000) and Takamatsu (2004Takamatsu ( , 2013) compared the phylogenetic relationships and phenotypic characteristics of Erysipheae and Cystotheceae, which are two tribes in Erysiphaceae containing species with diverse appendage morphologies, based on parsimonious considerations.Results showed that the evolution from branched or circinate appendages to mycelioid appendages occurred at the same nodes as the shift of hosts occurred from trees to herbaceous plants within these lineages.Based on these results, Takamatsu (2004) hypothesized that the appendage morphology of chasmothecia has a function related to the overwintering strategies adapted to a specific host; hence, appendages exhibit morphological variations across different host types.Shirouzu et al. (2022) conducted analyses using phylogenetic comparative methods (PCMs) for Cystotheceae and suggested that a shift in host type from deciduous trees to herbs or evergreen trees and a shift in appendage morphology from complex (circinate or branched) to simple (rudimentary or mycelioid) occurred two or three times at the same nodes.Based on these findings, Shirouzu et al. (2022) suggested that an evolutionary transition from complex to simple appendages might have occurred, allowing for more effective overwintering on herbs or evergreen trees.Moreover, they demonstrated that PCMs can be utilized to test statistical models of trait evolution in powdery mildews.However, this approach has only been applied to Cystotheceae, and the overall trend in evolutionary dependence of host type and appendage morphology has not been statistically supported.Cystotheceae is a medium-sized tribe (ca.120 species; Braun and Cook 2012), and only a limited subset of taxa (59 taxa) was used in the analysis (Shirouzu et al. 2022).This limitation may have affected the accumulation of sufficient information for evolutionary analysis.By examining Erysipheae (ca.380 species; Braun and Cook 2012), the largest and most diverse tribe of Erysiphaceae, we may be able to clarify the evolutionary dependence and general evolutionary patterns of these traits.The current study aimed to elucidate the evolutionary relationship between host type and appendage morphology; to this end, we analyzed the state transition, ancestral state, evolutionary dependence, and contingent evolution using PCMs based on a comprehensive data set of Erysipheae.

MATERIALS AND METHODS
DNA sequences and fungal traits.-Wedownloaded the nuc rDNA sequences of 28S D1-D2 regions (28S), internal transcribed spacer region ITS1-5.8S-ITS2(ITS), and 18S as well as the protein-coding region of MCM7 (minichromosome maintenance complex component 7) of Erysipheae (186 taxa) from the National Center for Biotechnology Information (NCBI) on 10 April 2023 (SUPPLEMENTARY TABLE 1).We selected the fungal sequence and trait data of 186 taxa based on the following two criteria: (i) reliable species identification through morphology and DNA sequences and (ii) confirmation of chasmothecium formation on each host type.Phenotypic characteristics of the Erysipheae species were determined based on the morphological descriptions in previous studies (SUPPLEMENTARY TABLE 1; SUPPLEMENTARY DATA).The host types were classified as deciduous broad-leaved trees (deciduous), evergreen broad-leaved trees (evergreen), and herbaceous plants (herb).For species with multiple host types, the major host type was selected based on the collection records of the specimens (SUPPLEMENTARY TABLE 1; Braun and Cook 2012).Appendage morphologies were classified as uncinate-circinate (circinate), branched, clavate with mucilaginous material (clavate), and mycelioid (FIG.1).
Circinate, branched, and clavate appendages were defined as complex, whereas mycelioid appendages were defined as simple.The appendages of Erysiphe uncinuloides and E. longiappendiculata are mentioned as "uncinate-circinate to helicoid at the apex" in the original description (Siahaan et al. 2018).However, photographs of their chasmothecia show the appendages as long undulating mycelia, rather than setiform, and they are estimated to be functionally almost identical to mycelioid.Therefore, the appendages of these two species were treated as mycelioid.The appendage of E. bremeri was described as "dichotomously branched" at the apex by Braun and Cook (2012).However, the overall characteristic of appendage was mentioned as "mycelioid" in the same description, with line drawing also depicting mycelioid features.Taken together, we considered the appendage of E. bremeri to be mycelioid.
Phylogenetic tree estimation.-Weconstructed data sets consisting of the 28S, ITS, 18S, and MCM7 sequences, followed by a multiple sequence alignment using MAFFT 7 (https://mafft.cbrc.jp/alignment/server;Katoh et al. 2019).The aligned data set was uploaded to TreeBASE with ID S30755 (http://purl.org/phylo/treebase/phylows/study/TB2:S30755).To estimate a Bayesian tree, we used MrBayes 3.2.7a in the CIPRES science gateway (Miller et al. 2015;Ronquist et al. 2012).To select optimal substitution models for each DNA region and codon, we used Kakusan4 (Tanabe 2011) based on BIC4 (gene proportional-codon proportional model).The SYM+G model was selected for 28S and ITS, K80+G was selected for 18S and 1st and 2nd codons of MCM7, and HKY85+G for 3rd codon of MCM7.We performed two runs with eight chains of Metropolis-coupled Markov chain Monte Carlo [(MC) 3 ] iterations, keeping one tree per 1000 generations.We assessed the convergence of the (MC) 3 procedure from the average standard deviation of split frequencies (<0.01) using the MrBayes software.After running 50 000 000 generations (100 002 samples were obtained), we discarded the first 80 002 samples as burn-in and used the remaining 20 000 samples to calculate a 50% majority-rule consensus tree and determined the posterior probabilities (PPs) for individual branches (SUPPLEMENTARY FIG. 1).
Random forest.-Toassess the relative importance of phylogenetic distance and traits among fungal species in predicting host type, we performed a random forest analysis.First, we calculated Moran's eigenvector maps (MEMs; Dray et al. 2006) for phylogenetic distances (pMEMs) to perform a distance-based analysis of the phylogeny among the fungal species according to the methods of Tedersoo et al. (2013).We first calculated the pairwise phylogenetic distances for each species based on the Bayesian tree.Next, we used the distance matrix to calculate the pMEM vectors with the ADESPATIAL package (0.3.16) of R 4.2.0 (R Core Team 2022).The pMEM vectors are orthogonal, represent the amount of phylogenetic variation proportional to their eigenvalues, and are listed according to the decreasing amount of explained variance (Tedersoo et al. 2013).We used 64 pMEM vectors (pMEM1-pMEM64) with positive Moran's I values.Random forest analysis was performed using the R package RANDOMFOREST (4.7-1.1;ntree = 500) with 64 phylogenetic vectors and 10 trait variables (chasmothecial diameter, appendage morphology, appendage number, appendage length, ascus number, ascus length, ascus width, ascospore number, ascospore length, and ascospore width; SUPPLEMENTARY DATA) as predictors.If the trait had a range value, such as diameter or length, the median was used as the value of the trait.If multiple host types were present, the major host type was used (SUPPLEMENTARY TABLE 1).

Phylogenetic comparative methods.-Statistical
estimations of the state transition rates and ancestral state reconstructions were conducted using the Mk model based on the Bayesian framework in BayesTraits 4.0 (Pagel et al. 2004;Suzuki et al. 2022).Parameter fitting was conducted using reversible-jump Markov chain Monte Carlo, which assigns a zero to arbitrary parameter and calculates likelihood in each of Markov chains, providing z-scores for low-frequency transitions.To determine an appropriate statistical model for trait evolution, we conducted three analytical steps as follows: (i) First, the full model, which includes all of parameters, was calculated by estimating the z-score.Several constrained models were designed using the z-score obtained by parameter fitting of the full model as a threshold, and transitions with a z-score higher than the threshold values (>70%, >50%, and >30%) were omitted.(ii) Second, the constrained models were calculated with Bayes factors.(iii) Third, a comparison of the Bayes factors of the full and constrained models was performed to select the best-fit models.Using the best-fit models, we conducted the following analyses (i.e., analyses of evolutionary state transitions and ancestral state reconstruction).To test for the (in)dependence of a pair of traits during trait evolution, we used the Discrete program implemented in BayesTraits (Pagel and Meade 2006) and compared two statistical models, an independent model vs. a dependent model, using Bayes factor (>2: positive).If chasmothecium formation was observed on multiple host types, we specified multiple traits for ancestral state reconstruction and the major host type was used to test for (in)dependence of the traits of state changes (SUPPLEMENTARY TABLE 1).We ran 50 000 000 generations after 50 000 iterations were removed as a burn-in, sampling every 1000 generations to yield a sample of 50 000 iterations.The phylogenetic trees used in the analysis were generated by sampling 4000 trees from the Bayesian-computed phylogenetic trees, with all branches having lengths greater than zero.In the ancestral state reconstruction, posterior probabilities ≥0.90 were used as the criterion to determine the ancestral state at a node.Contingent evolution of host type and appendage morphology was analyzed according to Suzuki et al. (2014).

Evolutionary transition and ancestral state of host
type.-Based on the Bayesian trees of Erysipheae, we employed PCMs to estimate the evolutionary pattern of host types.Two statistical models (a full model and a constrained model) were postulated to elucidate the evolutionary transitions across host-type states (SUPPLEMENTARY FIG. 2).The full model allowed all transitions across states, and the constrained model rejected transitions from evergreen to deciduous or herb.Constrained model 1 was constructed by omitting transitions having a z-score higher than the threshold value (>50%) derived from the full model (SUPPLEMENTARY TABLE 2).Since constrained model 1 was a statistically unreasonable model and the initial starting set of parameters could not be found, we considered the full model preferable for subsequent analyses.Our estimation based on the full model revealed that all transitions across host types were possible (FIG.2a).Bidirectional transitions between deciduous and herb and transitions from deciduous to evergreen and from herb to evergreen were more likely to occur than from evergreen to deciduous or herb (z-score ≥20%; FIG.2a).The ancestral state reconstruction using the full model showed that the most ancestral state was deciduous (FIG. 3,SUPPLEMENTARY FIG. 4,SUPPLEMENTARY TABLE 3).PCM analyses clearly indicated that most extant species of deciduous types were inherited from a common ancestor of the host types.Convergent evolution from deciduous to herb, deciduous to evergreen, and herb to evergreen occurred repeatedly six, three, and two times, respectively.Furthermore, the deciduous type reverted five times from the herb type.

Evolutionary transition and ancestral state of appendage morphology.
-The evolutionary pattern of appendage morphology was estimated using PCMs.To explain the evolutionary transitions among the appendage morphology states, we considered four statistical models (SUPPLEMENTARY FIG. 3).One was a full model that allowed all transitions across states, whereas the remaining three were constrained models 1-3, which imposed specific restrictions.Constrained model 1 limited the transitions from branched to circinate or clavate, whereas constrained model 2, in addition to the constraints of model 1, restricted the transitions from mycelioid to circinate and circinate to branched.Constrained model 3 allowed transitions between mycelioid and branched, as well as from circinate to mycelioid or clavate.These constraints were derived by omitting transitions with a z-score higher than the threshold values (>70%, >50%, and >30%) estimated using the full model (SUPPLEMENTARY TABLE 4).The initial starting set of parameters could not be found for constrained model 3, since it was a statistically unreasonable model.Since statistical tests using the Bayes factor significantly supported constrained model 2 (SUPPLEMENTARY TABLE 5), it was considered preferable for subsequent analyses.Based on constrained model 2, transitions from circinate to mycelioid and bidirectional transitions between mycelioid and branched were more likely to occur than other transitions (z-score ≥20%; FIG.2b).The result suggested the existence of evolutionary constraints from branched to clavate and between circinate and branched.Although both are transitions between simple and complex appendages, there appears to be an evolutionary constraint on the transition from mycelioid to circinate; however, no such constraint exists between mycelioid and branched appendages.Mycelioid and branched morphologies may have a common basis, making it easy for them to shift their states.Ancestral state reconstruction using constrained model 2 indicated the most ancestral state to be circinate (FIG.3, SUPPLEMENTARY FIG. 4, SUPPLEMENTARY TABLE 6).PCM analyses clearly showed that most extant species of the circinate type are due to inheritance from the common ancestral state.Convergent evolution toward the mycelioid type occurred eight times repetitively, with three transitions from the circinate type and five from the branched type.The branched type probably evolved once from the circinate type (PP = 0.85; FIG. 3, SUPPLEMENTARY FIG. 4, SUPPLEMENTARY TABLE 6), and after the transition the latter never occurred.Although the transition from circinate to branched type had a low frequency in the phylogenetic tree as a whole (SUPPLEMENTARY TABLE 4), the ancestral state reconstruction indicated that the transition occurred at least once (FIG.3).The branched types in the present species are primarily due to the inheritance of host types from a common ancestor.In addition, the branched types reverted twice from the mycelioid types.
Evolutionary dependence and combination of host type and appendage morphology.-Weanalyzed the correlation between host type and appendage morphology with phylogenetic correction using two phylogenetic comparative analytical methods.One was based on independence vs. dependence analyses, showing the overall tendencies of the pairwise state transitions.The other was based on ancestral state reconstruction, which revealed the local tendencies along specific lineages.For the former analyses, PCM analyses using Discrete in BayesTraits indicated significant evolutionary dependence (Bayes factor = 11.33;TABLE 1) between the host type (deciduous vs. herb/evergreen) and appendage morphology (complex vs. simple).This was one of the few examples demonstrating evolutionary dependence between host phenology and morphological traits in plant-parasitic fungi using PCMs.The evolutionary dependence between host type and appendage morphology was also detected as state transitions at the same nodes in the ancestral state reconstruction.The ancestral state indicated that the mycelioid (simple) appendage evolved from the complex appendage, and the transitions of host type from deciduous to herb/evergreen occurred five times at the same nodes (FIG.3).Additionally, our analyses indicated that the branched (complex) appendage reverted from the mycelioid (simple) appendage, and the transition of host type from herb to deciduous occurred twice at the same nodes.These findings suggested that the evolution of host type and appendage morphology can occur bidirectionally between complex types on deciduous hosts and simple types on herb/evergreen hosts.These convergent evolutions and transitions at the same nodes further suggested the evolutionary dependence between host type and appendage morphology.In the character states observed within the analyzed species, all appendage types were found on deciduous hosts, whereas branched and mycelioid types were found on evergreen and herbaceous hosts (FIG.4).Regarding the ancestral states (PP ≥0.90), circinate, clavate, and branched types were identified on deciduous hosts, only mycelioid type exclusively on evergreen hosts, and branched and mycelioid types on herb hosts.Our random forest analysis demonstrated a corrected classification rate of 86.56% using phylogenetic vectors and traits, with an out-of-bag (OOB) error rate of 13.44%.Notably, among the predictors, appendage morphology and one phylogenetic vector (pMEM10) displayed a high mean decrease in Gini values (>4) when predicting the host type (SUPPLEMENTARY FIG. 5).pMEM10 primarily captured the distinctions between the clade containing E. frickii to E. mori (mycelioid on herbs and circinate on deciduous trees) and the clade containing E. longifilamentosa to E. trinae (mycelioid on evergreen trees; FIG. 3, SUPPLEMENTARY DATA).

Contingent evolution of host type and appendage morphology.
-We further examined the evolutionary order dependence of the host type and the appendage morphology.Contingent analysis was performed for trait evolution to check whether the evolution of one of the paired traits was dependent on that of the other.Results showed that the evolutionary transition from simple to complex appendages occurred more likely on deciduous hosts than on herb/evergreen hosts (z-score ≥70%; FIG. 5, SUPPLEMENTARY TABLE 7), strongly suggesting that complex appendages on herb/evergreen hosts are supplied from deciduous hosts.This result in turn strongly suggested that the evolutionary potential of morphological transitions is tightly coupled with the host type.In contrast, the results showed that morphological simplification occurred in all host types (FIG. 5, SUPPLEMENTARY TABLE 7).Transitions of host types were suggested to occur in any state of appendage morphology.

DISCUSSION
Evolutionary relationship of host type and appendage morphology.-Weidentified, for the first time, the evolutionary dependence between host type and appendage morphology in Erysiphaceae (TABLE 1).This is one of the few examples demonstrating the   evolutionary dependence between host phenology and morphological traits in plant-parasitic fungi using PCMs (Janošík et al. 2023).The result reinforced the significance of considering the ecological features of host plants to unravel the trait evolution in plant-parasitic fungi.To further analyze the evolutionary relationships between these traits, we estimated contingent evolution and compared it with ancestral state reconstruction.Our results suggested that the states of extant species primarily arose through appendage simplifications and complications on deciduous hosts, appendage simplifications on herb/evergreen hosts, and host shifts occurring independent of appendage morphology (FIGS. 3,4,and FIG. 5).
In these evolutionary patterns, appendage simplifications on herb/evergreen hosts and appendage complications on deciduous hosts could be explained by the assumed advantages of each appendage morphology during overwintering.On herb or evergreen hosts, overwintering by remaining on the original substrata (e.g., leaves) with simple-appendage chasmothecia is considered advantageous (Gadoury et al. 2010;Shirouzu et al. 2022;Takamatsu 2004).On these host types, overwintering through migration from the original to secondary substrata via complex-appendage chasmothecia might pose a higher risk than overwintering by remaining on the original substrata with simple appendages (Gadoury et al. 2010;Shirouzu et al. 2022).Consequently, appendage complications on herb/evergreen hosts may be disadvantageous, making appendage simplification more likely on this host type.In addition, transitions to herb/evergreen hosts may alleviate the selection pressure favoring the development of higher-cost complex appendages, resulting in a convergence toward lower-cost simple appendages (Shirouzu et al. 2022).Conversely, on deciduous hosts, migrations of the complex-appendage chasmothecia from the original substrata to the secondary substrata (e.g., branches and trunks) are considered more favorable for overwintering (Gadoury and Pearson 1988;Grove and Boal 1991;Shirouzu et al. 2022;Takamatsu 2004).Our analysis, demonstrating that appendage complications are more probable on deciduous hosts, supported the results of a previous study, which considered that selection pressure toward complex appendages might be higher on this host type (Shirouzu et al. 2022).On deciduous hosts, the risks associated with remaining and overwintering on original substrata, which are deciduous during winter, likely outweigh the risks associated with the formation of complex appendages and migration to secondary substrata (Gadoury and Pearson 1988;Grove and Boal 1991;Shirouzu et al. 2022).This might allow the evolution of complex appendages to occur more likely on deciduous hosts than herb/evergreen hosts.
Our PCMs further revealed state shifts that could not be fully explained by the distinctive overwintering function inferred from appendage morphology.Specifically, we detected appendage simplifications on deciduous hosts, shifts to deciduous hosts with simple appendages, and shifts to herb/evergreen hosts with complex appendages (FIGS. 3 and 5).Appendage simplification on deciduous hosts and shifts to deciduous hosts with simple appendages may occur when the significance of complex appendages in overwintering decreases.For example, this could happen when chasmothecia formed on buds, branches, or fruits remain as overwintering organs (e.g., E. diervillae and E. monascogera; Braun and Cook 2012).This expansion into non-leaf microhabitats may have allowed these fungi to complete their life cycle with simple-appendage chasmothecia, even on deciduous hosts.In certain species, appendages exhibit intermediate or ambiguous morphologies, which makes it difficult to deduce their functions during overwintering.For example, E. astragali, E. bremeri, E. hyperici, E. intermedia, E. longiappendiculata, E. russellii, E. trifoliorum, and E. uncinuloides exhibit intermediate states between complex and simple appendages (Braun and Cook 2012;Siahaan et al. 2018).These species may represent transitional stages in which appendage simplifications on herb/evergreen hosts are in progress.Erysiphe fimbriata, found on deciduous hosts, possesses mycelioid (simple) appendages that are exceptionally long (up to 5000 µm) and extend vertically from the upper surface of chasmothecia (Braun and Cook 2012).This appendage morphology may play a role during migration to secondary substrata.These species exhibit ambiguities in the classification and functional estimation of appendage morphology, which in turn limits the interpretations based on the current available data.To further understand the evolution of appendage morphology, additional research is warranted to investigate the flexibility of overwintering strategies and to quantify the functions of appendage morphologies.
Evolutionary constraint of host type and appendage morphology.-PCManalysis provides an opportunity to introduce conceptual quantitative evaluations and opens new doors for phenotypic trait evolution (Suzuki 2022).In addition to the results presented in previous studies on Cystotheceae (Shirouzu et al. 2022), our present study showed that trait evolution did not occur randomly in powdery mildew fungi.Instead, there seemed to be some constraints governing the transitions between specific states (FIG.2).For instance, shifts from evergreen to herb or deciduous type were estimated to less likely occur in Erysipheae (FIG.2), whereas analogous shifts between evergreen and herb types are known to be less probable in Cystotheceae (Shirouzu et al. 2022).Although examples of host jumping have been reported in obligate plant-parasitic fungi (Ploch et al. 2022), the underlying factors contributing to evolutionary constraints on the transitions between different hosts remain inadequately understood.In terms of the transitions in appendage morphology, Erysipheae exhibited a reduced likelihood of shifts between complex types, such as circinate and branched (FIG.2).Conversely, in Cystotheceae, transitions between simple types, such as rudimentary and mycelioid, were identified to less likely occur (Shirouzu et al. 2022).Our ancestral state estimation indicated that the shift from circinate to branched type might have occurred only once in Erysipheae, but this infrequent state transition subsequently became a widely conserved morphology within the extant species (FIG.3).These findings suggested that evolutionary constraints are lineage dependent and may play a crucial role in the adaptive evolution and speciation of powdery mildew.In Cystotheceae, shifts from simple to complex appendages were demonstrated to be unlikely (Shirouzu et al. 2022); however, in Erysipheae, transitions between mycelioid (simple) and branched (complex) appendages occur without constraints, indicating that these morphologies can be easily altered by simple genetic changes.The evolutionary shifts in appendage morphologies may be related to internal factors, such as fungal genetic mutations and morphogenetic processes.Although Erysiphaceae are important plant-pathogenic fungi with remarkable features, knowledge regarding their interactions with host plants and the developmental processes of their morphology still remains limited (Gadoury et al. 2010;Gadoury and Pearson 1988;Takamatsu et al. 1982).To enhance our understanding of evolutionary constraints in powdery mildew, further investigation is imperative, encompassing data on fungus-host interactions and comprehensive morphogenetic studies.

CONCLUSION
Our PCMs detected evolutionary dependence between host type and appendage morphology for the first time in Erysiphaceae, reaffirming the importance of the interplay between these traits in comprehending their evolutionary dynamics.This is one of the few examples demonstrating the evolutionary dependence between host phenology and morphological traits in plant-parasitic fungi using PCMs (Janošík et al. 2023).PCMs on powdery mildew have provided an opportunity to study the evolution of fungal morphology from the perspective of interactions between fungal parasites and host plants.Within the functionally diverse kingdom of fungi, powdery mildews are noteworthy plant parasites that reveal trait evolution in the context of their infection strategies.
Appendage simplifications on herb/evergreen hosts and complications on deciduous hosts can be reasonably explained by the functional advantages of each appendage in overwintering strategies (FIGS. 4 and 5;Takamatsu 2004;Shirouzu et al. 2022).The expected functions of the appendages can elucidate approximately 90% of the host-appendage combinations of the analyzed taxa (SUPPLEMENTARY TABLE 1).However, our results further showed the occurrence of evolutionary shifts that deviated from the functional advantages inferred from appendage morphology (FIGS. 3 and 5).These exceptions might be interpretable from the flexibility of the overwintering strategies and the quantification of appendage functions.

DISCLOSURE STATEMENT
No potential conflict of interest was reported by the author(s).

FUNDING
This work was supported by the Japan Society for the Promotion of Science (19K06124 and 23K05302).Takamatsu S, Ishizaki H, Kunoh H. 1982

Figure 2 .
Figure 2. State transitions of (a) host types and (b) appendage morphologies.Transitions with z-scores ≥20% are indicated by dashed arrows, and those with z-scores <20% are indicated by thick arrows.Asterisks indicate the most ancestral state.

Figure 3 .
Figure 3. Ancestral state reconstruction of host types and appendage morphologies.Asterisks indicate posterior probabilities (PPs) of supported states at each node (≥0.90).The deciduous type is the most ancestral, whereas the herb (orange arrows) and evergreen (blue arrows) types evolved secondarily.The deciduous types (yellow-green arrows) reverted from herb types.The circinate type is the most ancestral, and the mycelioid (orange arrows), clavate, and branched (blue arrows) types evolved secondarily.The branched types (blue arrows) reverted from mycelioid types.Transitions of host type and appendage morphology at the same nodes are indicated (yellow stars).

Figure 4 .
Figure 4. Number of combinations of host type and appendage morphology.The numbers of the present state are shown on the left, and the numbers of the ancestral state are shown on the right (PP ≥0.90).

Figure 5 .
Figure 5. Contingent evolution of host type and appendage morphology.The transition from (1, 0) to (1, 1) is not indicated owing to the z-score being ≥70%.Asterisk indicates the most ancestral state.

Table 1 .
Evolutionary correlations of host type and appendage morphology.