A combined microscopy and single-cell sequencing approach reveals the ecology, morphology, and phylogeny of uncultured lineages of zoosporic fungi

ABSTRACT Environmental DNA analyses of fungal communities typically reveal a much larger diversity than can be ascribed to known species. Much of this hidden diversity lies within undescribed fungal lineages, especially the early diverging fungi (EDF). Although these EDF often represent new lineages even at the phylum level, they have never been cultured, making their morphology and ecology uncertain. One of the methods to characterize these uncultured fungi is a single-cell DNA sequencing approach. In this study, we established a large data set of single-cell sequences of EDF by manually isolating and photographing parasitic fungi on various hosts such as algae, protists, and micro-invertebrates, combined with subsequent long-read sequencing of the ribosomal DNA locus (rDNA). We successfully obtained rDNA sequences of 127 parasitic fungal cells, which clustered into 71 phylogenetic lineages belonging to seven phylum-level clades of EDF: Blastocladiomycota, Chytridiomycota, Aphelidiomycota, Rozellomycota, and three unknown phylum-level clades. Most of our single cells yielded novel sequences distinguished from both described taxa and existing metabarcoding data, indicating an expansive and hidden diversity of parasitic taxa of EDF. We also revealed an unexpected diversity of endobiotic Olpidium-like chytrids and hyper-parasitic lineages. Overall, by combining photographs of parasitic fungi with phylogenetic analyses, we were able to better understand the ecological function and morphology of many of the branches on the fungal tree of life known only from DNA sequences. IMPORTANCE Much of the diversity of microbes from natural habitats, such as soil and freshwater, comprise species and lineages that have never been isolated into pure culture. In part, this stems from a bias of culturing in favor of saprotrophic microbes over the myriad symbiotic ones that include parasitic and mutualistic relationships with other taxa. In the present study, we aimed to shed light on the ecological function and morphology of the many undescribed lineages of aquatic fungi by individually isolating and sequencing molecular barcodes from 127 cells of host-associated fungi using single-cell sequencing. By adding these sequences and their photographs into the fungal tree, we were able to understand the morphology of reproductive and vegetative structures of these novel fungi and to provide a hypothesized ecological function for them. These individual host-fungal cells revealed themselves to be complex environments despite their small size; numerous samples were hyper-parasitized with other zoosporic fungal lineages such as Rozellomycota.

Chytrid-like-1, Fig. 9), which is sister to Monoblepharidomycota, but statistical support for this relationship was not robust. Two lineages were placed in the clade FRESHOL1 reported previously (6) (Fig. 10). PSC-L70 was the cell of Staurastrum sp. filled by including Spirogyra sp. and attaching chytrid-like sporangia.

Research Article mBio
Morellospora (63), which produce Microsporidia-like spores but are phylogenetically placed in Rozellomycota and much shorter branched than canonical Microsporidia. PSC-L61 was distinguished from these previously reported Microsporidia-like taxa. Five lineages (PSC-L64-68) were isolated as epibiotic chytrids on green algae ( Fig. 4S through V, X) and diatoms (Fig. 4W). However, they were positioned in the Rozella clade, which comprises endoparasites of chytrids and oomycetes, suggesting they were hyper-para sites of chytrids. PSC-L69 (Fig. 4Y), which showed the same morphology as PSC-L23 (Fig. 3D) in Rhizophydiales, was sister to all other Rozellomycota taxa. This lineage is also a putative Rozella-like hyper-parasite (see Discussion). PSC-L60 was a tube-shaped zoosporangium in Oedogonium sp. (Fig. 4O) and placed in the LKM11 clade (64). PSC-L62 in a tardigrade (Fig. 4Q) and PSC-L63 in a putative broken rotifer body (Fig. 4R) were related to Paramicrosporidium spp. These zoosporangium-like structures in PSC-L60, L62, and L63 may not correspond to rozellids, and the sequences could be derived from hyper-parasites of these zoosporangia or undetected contaminated cells.

Phylogenetic relationship between single cells and environmental sequences
The concatenated data set analysis showed that many of our single cells represented novel lineages distinguished from described taxa (Fig. 6 to 10). This result complements many environmental DNA studies that have reported unknown fungal lineages (23,48).
To examine overlap between our single-cell lineages and sequences only known from environmental DNA, we conducted a phylogenetic analysis on a comprehensive 18S rDNA data set including described taxa, environmental sequences available from NCBI database, with a focus on the phylum Chytridiomycota and phylum incertae sedis clades (see Fig. S1 in the supplemental material). In this analysis, new sequences of PacBio metabarcoding analyses primarily from Michigan, USA (44), were also used. Many of these PacBio sequences are derived from the same locations as the single cells isolated in this study, which gives a good opportunity to compare the two methods, metabarcoding and single-cell analysis, for exploring novel fungal diversity. The ML tree (see Fig. S1) showed that the vast majority of PacBio sequences represent entirely new lineages. Only a few of the single-cell lineages were closely related to PacBio environmental sequences. The PSC-L8 clade, in Rhizophydiales, included sequence UMPB-228 from Ford Lake, the same place where the two diatom parasites were isolated. In Chytridiales, PSC-L39 included UMPB-32 detected from multiple freshwater environments including Lake Erie (see Data S1 in the supplemental material) where some of the single cells were isolated. This lineage was also found in lakes in Japan (15) and France (48). Sequence UMPB-232 was the most abundant in Bryant Bog (see Data S1) and was related to PSC-L33 isolated from the same location. Although overlap of lineages in the PacBio and single-cell data sets was low in terms of species, multiple lineages of single cells had as their most closely related sequence an OTU from the PacBio data set, e.g., PSC-L13 and 21 in Rhizophydiales; PSC-L33, 35, 38, 43, and 44 in Chytridiales; PSC-L45 and 46 in Lobulomycetales; PSC-L49 in Cladochytriales; and PSC-L71 in the FRESHOL1 clade (see Fig. S1). On the other hand, both Mesochytriales and Polyphagales were represented by multiple OTUs in the PacBio data set but were absent in the single-cell data. Despite these exceptions, the general pattern was one of significant overlap of taxonomic genera and families in these two culture-independent approaches.

DISCUSSION
By utilizing single-cell techniques, 71 EDF lineages were sampled, many of which were newly recognized branches in the phylogeny of EDF, even at the phylum-level. The approach in the present study focused on targeting and sequencing individual EDF cells one at a time. Photographs of the isolated cells have implications for the ecology, morphology, and life cycle of these newly discovered EDF lineages. Using these data, we discuss the (i) ecological role of these uncultivated lineages, (ii) ecology and morphology of novel phylum-level clades, (iii) phylogenetic diversity of an enigmatic chytrid genus Olpidium, and (iv) unexpected recovery of hyper-parasitic lineages. Finally, technical advances and challenges of single-cell approaches used in this study are also discussed.

Shedding light on the ecological role of dark matter fungi
The cells isolated in this study were from diverse hosts, ranging from amoebae to invertebrates and especially algae. Some lineages are readily recognized as obligate algal parasites belonging to known parasite-specific groups such as Zygophlyctidales and Aphelidiomycota. Also found were many lineages of alga-associated cells in well studied orders such as Rhizophydiales, whose diversity has long been investigated based on numerous strains of saprotrophic taxa (65)(66)(67). Recently though several families of obligate algal parasites were described (37)(38)(39)68). Our data revealed further hidden diversity of putative parasitic lineages, representing new families or genera in the order.
Although the isolated cells in this study were initially identified as "parasitic fungi, " some lineages we sampled in Chytridiales and Cladochytriales are putatively saprotrophic. When parasitic chytrids infect colonial algae, only dead cells are infected while living cells are uninfected (Fig. 2F, U, V and 3I AA, ). In contrast, in some colonies, all algal cells are uncolored, or their chloroplasts are exhausted ( Fig. 3O, P, R and Y), indicating the attaching chytrids grow on dead or moribund algae. Moreover, chytrids corresponding to PSC-L33 (Fig. 3M) in Chytridiales and PSC-L48 (Fig. 4A) in Cladochytriales were success fully isolated as pure cultures (data not shown). These data also inform hypothetical ecological functions of lineages that were only known from metabarcoding approaches. For example, algal parasitic lineages PSC-L21 and 22 in Rhizophydiales formed independent clades along with some environmental sequences from aquatic environments, implicating a role for this clade as parasites of algae. Similarly, PSC-L45-47 in Lobulomycetales formed a novel clade including some environmental sequences from aquatic and soil environments, indicating that these lineages are parasites of algae and protists. Although Zygophlyctidales was previously thought to be composed only of diatom parasites (21), a lineage of a green algal parasite (PSC-L31) sister to an environmental sequence from a lake was found. This result indicates that other environmental sequence lineages in the order could exist as parasites of algae other than diatoms.
Generally, however, most of the single-cell sequences were poor matches at the species level to sequences from cultures or environmental DNA. This speaks to just how poorly we understand the true species level diversity of EDF, and how much work remains to be done in describing these fungi. In some cases, sequences from clades that were readily recovered with metabarcoding were not detected. The most striking case is Gromochytriales and Mesochytriales, together containing a total of three described species, all of which are obligate parasites (20,69,70). Despite limited described species, Mesochytriales is represented by numerous environmental DNA sequences (20). In this study, chytrids belonging to these orders were not identified with a single-cell approach. Instead, additional diversity was revealed based on phylogenetic analysis using PacBio metabarcoding data (see Fig. S1). Many of the sequences from lakes or ponds and some OTUs from Lake Erie, were related to Mesochytrium penetrans. This species is a parasite of a small green alga, Chlorococcum minutum (71), yet the collection strategy adopted for Lake Erie samples biased for larger colonial and filamentous forms. Further, most effort on Lake Erie was aligned with a winter science initiative (72), a season where diatoms are the dominant taxa associated with ice-cover in the lake (73). Thus, chytrids parasitic on smaller single cells or on taxa more prevalent during the summer may have been overlooked. More single-cell analyses on parasitic chytrids on various algae are necessary to reveal hidden taxa in the order.
While the pictures of the isolated cells can be informative in inferring their ecological role, sometimes they may be misleading. Specifically, each "cell" is actually a number of cells that include host, parasite, associated bacteria, and hyper-parasites. The latter were particularly common with some cells, such as PSC023, giving both an obvious chytrid pathogen as well as a likely Rozella hyper-parasite (74). Indeed, the majority of the Rozellomycota detected in this study were found as "by-catch" present in samples appearing as normal chytrids infecting algae (Fig. 4S through W). This result provides a slight cautionary tale that some of the sequences emerging from this approach may be most appropriately assigned not as parasites of the primary host but as hyper-parasites, an observation consistent with earlier results (15).

Discovery of novel clades of early diverging fungi
Our approach was successful in revealing novel diversity at many taxonomic levels: species, genera, families, and even phyla. The novelty at the phylum level is consistent with recent phylogenetic analyses that have revealed that some parasitic fungi represent novel lineages worthy of phylum-level distinction (17,18). We found and characterized three phylum incertae sedis clades (Fig. 5), two of which are newly reported in the present study. The NC_OlpL-1 clade includes two single-cell lineages of Olpidium-like chytrids on rotifers and O. vermicola parasitic on nematode eggs (75), indicating that this clade represents animal-associated endobiotic chytrids. NC_OlpL-1 was sister to a previously reported undescribed phylum-level clade represented by a single-cell isolate of a Rhizosiphon-like chytrid on the cyanobacterium Dolichospermum from a lake in Germany (11). In the tree by Van den Wyngaert et al. (11), Dolichospermum parasites were sister to the Chytridiomycota + Monoblepharidomycota + Neocallimastigomycota clade without strong statistical support. Although these putative novel phylum clades are related to Kickellomycotina, Zoopagomycotina, and Entomophthoromycotina in our tree (Fig. 6), the exact phylogenetic position is uncertain. Phylogenomic analysis would clarify the evolutionary history and taxonomy of these enigmatic lineages.
The NC_ChyL-1 clade, which included isolates of epibiotic chytrids on Desmodesmus (PSC-L52) and Rhizophydium scenedesmi strain EPG01 on Graesiella sp (45), was sister to Monoblepharidomycota without strong statistical support. Previously, R. scenedesmi was shown to be sister to the genus Zygophlyctis in Chytridiomycota (45). However, another analysis showed that R. scenedesmi along with some environmental sequences were placed sister to Monoblepharidomycota (21) as with the present study. This clade could correspond to the clade GS13 defined by Tedersoo et al. (8) because one of their environmental sequences (GL02368_027_G2750 from Australian soil) was positioned within NC_ChyL-1.
The clade FRESHOL1 was originally defined by Arroyo et al. (6) in their metabarcod ing analysis of the Paraná River in Argentina. This clade was sister to all other fungi including Aphelidiomycota and Rozellomycota as with our analysis. PSC-L70 was a cell of Staurastrum sp. filled with a putative endoparasite (Fig. 4Z). The isolate correspond ing to PSC-L71 included chytrid zoosporangium-like cells on Spirogyra sp. (Fig. 4AAA). Although information on the life cycles of these two lineages is currently limited, there is the possibility that they are endoparasites of algae or chytrid-like organisms in these samples. Two deep-branching groups of fungi, Aphelidiomycota and Rozellomycota, are known as endoparasites of other organisms (22). The previously defined phylum-level clade NCLC1 is sister to Rozellomycota in our tree (Fig. 10) and is also comprised of putative endoparasites of marine diatoms (76). Given the phylogenetic position and host of the FRESHOL1 lineage, our findings strengthen the recently suggested hypothesis that the ancestor of Fungi sensu lato (including aphelids, rozellids, microsporidians, and canonical fungi) had a symbiotic relationship with cellulose-based cell-walled taxa (77). Further observations and phylogenetic analyses of the FRESHOL1 clade are pivotal to elucidate the early evolution of Holomycota lineages.

Phylogenetic diversity of Olpidium-like chytrids
We found Olpidium-like chytrids parasitic on various hosts such as adult rotifers ( Fig. 2A, C, and D), rotifer eggs (Fig. 2B), desmid algae (Fig. 2K through O and 3C through G), and pine pollen (Fig. 2AA). The genus Olpidium is characterized by a holocarpic thallus, namely a simple thallus composed of only a zoosporangium without rhizoids (78). All species are endobiotic parasites of algae, plants, fungi, protists, and micro-invertebrates (78). Early molecular phylogenetic analyses (79,80) revealed that plant root parasitic species are separated from core chytrid clades (e.g., Chytridiomycota and Blastocladio mycota) and related to zygomycetous fungi. A recent phylogenomic analysis showed that O. bornovanus parasitic on cucumber roots is sister to all terrestrial fungi (Dikarya + Mucoromycota + Zoopagomycota) (18). Olpidium-like chytrids obtained in the present study were not related to plant parasitic species (Olpidiomycota in Fig. 5) and were instead distributed among three other phylum-level clades.
Four lineages of rotifer parasites were placed in the NC_OlpL-1 clade (PSC-L1 and 2) and Blastocladiomycota (PSC-L3 and 4). The NC_OlpL-1 clade also included O. vermicola parasitic on nematode eggs. Apart from the plant parasitic lineage sister to Dikarya + zygomycetes (18), the NC_OlpL-1 clade is an additional putative independent phylum of Olpidium-like fungi. In Blastocladiomycota, PSC-L3 and 4 were related to taxa of the polyphyletic family Catenariaceae (81), which is characterized by polycentric thalli, consisting of catenulated zoosporangia connected by isthmuses (82). In both PSC-L3 and 4, multiple zoosporangia were seen in a single rotifer body but connections between zoosporangia were not visible. Some Olpidium species are known as rotifer parasites and often produce multiple zoosporangia in a single host, but early developmental stages have not been fully described (83)(84)(85). Some of these species could be related to Catenariaceae as with our rotifer parasites.
Our phylogenetic analysis clearly showed that the genus Olpidium is polyphyletic, and that host generally tracks phylogeny. Tedersoo et al. (31) suggested accommodating Olpidium in the phylum Olpidiomycota based on the phylogenetic position of plant parasitic species of Olpidium. However, this taxonomic treatment should be examined by investigating more taxa, especially the type species, O. endogenum, which is known as a parasite on green algae of the genus Closterium.

Unexpected findings of hyper-parasites
In the present study, we found putative hyper-parasites within Chytridiomycota and Rozellomycota. The two lineages in Chytridiomycota were clearly recognizable as a chytrid zoosporangium on top of another parasite. PSC-L11 in Rhizophydiales was parasitic on an elongated zoosporangium inside Spirogyra sp. (Fig. 2R). The host of this chytrid could be an endoparasitic oomycete in algae. Regarding similar described species, Rhizophydium carpophilum is known as a parasite of oogonia and oospores of Saprolegnia and Achlya (89) and also reported as a hyper-parasite of endoparasitic Olpidiopsis infecting Achlya (90). Another hyper-parasitic chytrid isolated in the present study is PSC-L39 in Chytridiales, a spherical zoosporangium on an epibiotic chytrid parasite on S. binderanus (Fig. 3T). Its host could be the chytrid of PSC-L28 (Fig. 3I) in Zygophlyctidales because the shape of the zoosporangium is similar, and they were found in the same sample collected at Lake Erie. Currently, some 15 species are known as epibiotic chytrid parasites of other chytrids (78, 91). One of them, Septosperma anomalum, was reported as a hyper-parasite infecting diatom parasite such as Chytriomy ces tabellariae on Tabellaria flocculosa (92) and Zygophlyctis asterionellae on Asterionella formosa (93). Our isolates are distinguished from S. anomalum based on the shape of zoosporangium. Also, S. anomalum produces a unique resting spore with septation, which was not observed in our sample. Unfortunately, DNA sequence data are currently not available for any chytrid species parasitic on other chytrids, preventing comparison with our isolates. PSC-L39 corresponds to the clade CH_D including single-cell isolates from Lake Inba in Japan (15). These isolates were recorded as a chytrid parasite on Aulacoseira spp. but there is a possibility that its hyper-parasitic nature was overlooked.
The other single-cell lineages of hyper-parasites were found in Rozellomycota. These were isolated as epibiotic or endobiotic chytrid parasites of green algae or diatoms, but they were phylogenetically related to Rozella spp. (PSC-L64-69). The genus Rozella is well known as an endoparasite of chytrids or oomycetes (94). Rozella invades the host as an unwalled cell, consumes host cytoplasm by phagocytosis, and ultimately fills the entire host cell. Due to this endoparasitic nature, chytrid zoosporangia infected by Rozella might be difficult to detect, although some species cause hypertrophy or abnormal septation of the host cell (36,95). Therefore, infections by Rozella were likely overlooked in our isolates. The putative hosts of our Rozella isolates were speculated: PSC-L64 (Fig. 4S) on Chytridiales chytrid PSC-L38 on Desmidium sp. (Fig. 3S); PSC-L66 (Fig.  4V) on Lobulomycetales chytrid PSC-L45 on Spirogyra sp. (Fig. 3Z); PSC-L67 (Fig. 4W) on Rhizophydiales chytrid PSC-L8 on Ulnaria sp. (Fig. 2J); and PSC-L69 (Fig. 4Y) on Olpi dium-like chytrid PSC-L23 on Micrasterias truncata (Fig. 3D). Indeed, single-cell genomic analysis on the amplified genome of isolate PSC023 (lineage PSC-L23) revealed that the genome included both the host as well as the putative hyper-parasite corresponding to PSC-L69 (74). However, in our ONT sequencing, only a chytrid rDNA sequence was obtained in PSC023. We assume that biased PCR amplification occurred in this sample. Phylogenomic analysis showed that the Rozellomycota genome in PSC023 was sister to Rozella spp. although PSC-L69 separated from the Rozella clade in the present study. Our finding of hyper-parasitic Rozella indicates cryptic diversity of endoparasites infecting chytrids. These findings need to be taken into consideration when using single-cell approaches to infer nutritional mode from the recovered genomes.

Technical advances and challenges using single-cell technique
The approach outlined presents both advances over traditional methods of single-cell genomics that involve fluorescence-activated cell sorting (96) as well as challenges. The primary advantages are that the method allows images of the target species to be obtained and that the success rate of going from cell to sequence is higher. Among the 259 cells processed with multiple displacement amplification (MDA), DNA sequence data were successfully generated for 139 cells (54%) in total. Excluding putative contaminants (e.g., cercozoans) and fungus-like organisms (e.g., oomycetes and hyphochytrids), 129 cells (50%) were categorized as fungal sequences (Table S1). This rate is higher than previous single-cell sequencing studies (single-cell sorting + whole genome amplification + PCR and sequencing) on planktonic prokaryotes and protists, which had a 5%-38% success rate (97)(98)(99)(100)(101). Moreover, photos accompanied these cells. While these cells are no longer available for morphological analysis, their amplified DNA with high concentration (147-1,600 ng/µL, Table S1) is present, which is facilitating ongoing genome sequencing. We believe that the high success rate of amplification and sequencing of the target cells is likely due to the fact that these fungal cells are actually comprised of multiple nuclei, in many cases representing the near mature reproductive stages of the chytrid zoosporangia that may contain 5-50 or more nuclei. A final technical advance is the combination of single-cell approaches with long rDNA PCR. Amplification of the majority of the coding bases of the rRNA operon in addition to the highly variable internal transcribed spacer region allows for robust phylogenetic placement as well as discrimination at the species level (102). Amplification of both 18S and 28S regions allows the data to be compared to multiple data sets, given that there are disparate uses of the two regions in both environmental DNA and systematics studies (48,103).
There are also some disadvantages of the method. First, it is hard to scale up to a large number of cells because this is a manual approach in which each cell requires as much as an hour to find, clean, and pipette into a sample tube. Second, this approach requires considerable taxonomy, microscopy, and microbial natural history skills. These skills are lacking in most microbiology and mycology training. Third, there are biases in the targeting of hosts. Most of the isolated cells in this study were parasites of algae, although a few protists and micro-invertebrates were isolated. More diverse taxa could host parasitic EDF, but they may have been undersampled due to our limited ability to find them and diagnose them as infected. These biased isolations potentially hinder clarifying the diversity of EDF; such a limitation is less applicable to metabarcoding and metagenomic approaches. Finally, our samples are far from single cells, and often contain host cells, bacterial cells, and in several cases, hyper-parasites. This is both an advantage and disadvantage because it identifies interesting symbioses, but it also makes ascribing ecological function more complicated. Presence of host and bacterial DNA could limit the ability to sequence fungal genomes from these samples. In some cases, we were able to amplify host DNA in order to confirm species identity (data not shown), but in other cases, host DNA could not be recovered as presumably the parasite had already consumed it. Despite these disadvantages, the target single-cell isolation is a powerful method to investigate uncultured parasitic fungi, and its use will expand our understand ing of the ecology and phylogeny of EDF.

Sample collection and single-cell isolation
We collected 50-250 mL of water samples with detritus and/or plant material from ponds or lakes in Michigan in 2019-2021 (see Table S2 in the supplemental material). For Lake Erie, seston was collected by boat with a plankton net (≥20 µm) deployed 1-3 m from the surface, after which the collected material was transferred to a 50 mL conical centrifuge tube maintained at in situ water temperature in the dark. The samples were transferred to University of Michigan and incubated for ~1 month, at 20°C, under LED lighting. Water samples were observed using a Nikon TMS Inverted microscope (Nikon, Tokyo, Japan) to detect fungi associated with algae, micro-invertebrates, and protists. Detected fungal cells were photographed using Moticam X Camera (Motic, Hong Kong, China) or Dino-Eye Edge S Eyepiece Camera (AnMo Electronic Corporation, Taipei, Taiwan) digital cameras. Representative images were edited and assembled into plates using Adobe Photoshop. The cells were isolated manually using a manually prepared drawn-out glass capillary pipette. The isolated cells were washed by serial transfer in small drops (more than five) of UV-sterilized water, transferred into 200 µL PCR tubes with 1-2 µL of water, and kept at −80°C until DNA extraction.

Phylogenetic analysis
To clarify phylogenetic positions of single cell isolates, a phylogenetic analysis of a concatenated data set of 18S, 5.8S, and 28S rDNA sequences was performed (see Table  S3 in the supplemental material). The 5.8S rDNA sequences were extracted from the data of ITS1-5.8S-ITS2 using ITSx (113). Sequences were aligned using MAFFT v7.487 (114) with the "L-INS-I" method, and the alignment was trimmed using trimAl (115) with the "gappyout" method. The maximum likelihood (ML) tree was inferred with IQ-TREE 2 (116). The best model of each alignment was examined using ModelFinder (117) implemented in the IQ-TREE 2. According to the corrected Akaike information criterion (AICc), GTR + F + R9, JC + R4, and TIM3 + F + R9 models were selected for 18S, 5.8S, and 28S, respectively. An ML analysis was run with a partitioned model (118). The branch supports were assessed with standard non-parametric bootstrap analysis (100 replicates). The tree was visualized with FigTree (https://github.com/rambaut/figtree) and edited with Adobe Illustrator.

PacBio metabarcoding analysis
Single-cell sequence data were compared to a large collection of environmental DNA sequences utilizing a recently developed PacBio metabarcoding data set of 18S amplicons (44). These data are complementary because many of the sampling localities are shared (i.e., are from aquatic habitats in Michigan). For this analysis, we extracted reference sequences of 339 PacBio OTUs that were putatively identified as Chytridiomy cota using taxonomic assignment in Qiime v1.9.1 (119) with BLAST (120) and a curated version of the SILVA database (121), which was amended to include more EDF including Aphelidiomycota and Rozellomycota (available here as "Updated_Silva_Cryptos_Aphe lids.txt": https://github.com/Michigan-Mycology/Lab-Code-and-Hacks/tree/master/Cryp tomycota_ecology/Data_files/). We performed further manual curation of the sequences identified as Chytridiomycota for putative chimeras by phylogenetic analyses. We prepared a reference data set of 18S rDNA of cultured Chytridiomycota species, singlecell sequences from our and previous studies, and outgroup taxa (see Table S4 in the supplemental material). The sequences were aligned and trimmed as described above.
All OTU sequences were divided into two parts, the former and latter ~650 bp nucleoti des. The original and divided PacBio sequences were added into the reference alignment using MAFFT with the "--add" and "--keeplength" options. Maximum likelihood trees were inferred using FastTree (122) with the "-gtr" option. The trees were visualized using FigTree. The OTUs of the following results were excluded in the subsequent analyses: (i) the phylogenetic positions of divided sequences were clearly different and (ii) the sequence was extremely long branched. After manual curation, 123 OTUs were excluded leaving 216 OTUs in the final data set (see Table S5 in the supplemental material). To examine the phylogenetic position of these OTUs, we performed an ML analysis. We prepared the 18S data set of almost the entire Chytridiomycota by adding environ mental sequences available in GenBank (see Table S6 in the supplemental material) to the reference data set used above (see Table S4) and sequences were aligned and trimmed as above. Subsequently, the curated PacBio OTU sequences were added to this alignment as above. The ML tree was inferred using IQ-TREE 2 with the GTR + F + R6 model selected by ModelFinder. A standard non-parametric bootstrap analysis of 100 replicates was performed. The trees were visualized and edited as above.

DIRECT CONTRIBUTION
This article is a direct contribution from Timothy James, a Fellow of the American Academy of Microbiology, who arranged for and secured reviews by Luis Galindo, Oxford University, and Jeremy Wideman, Arizona State University.

DATA AVAILABILITY
The sequence data obtained in this study were deposited in GenBank under the accession numbers OQ687116-OQ687331 and OQ702805-OQ702950.

ADDITIONAL FILES
The following material is available online.