Studies in Gyromitra II: cryptic speciation in the Gyromitra gigas species complex; rediscovery of G. ussuriensis and G. americanigigas sp. nov.

Taxa in the Gyromitra gigas species complex were previously studied and their taxonomy resolved. During ongoing studies in this group, cryptic speciation was discovered in G. gigas. Sequences of the ITS and LSU regions from 75 specimens were included in maximum likelihood and Bayesian phylogenetic analyses to establish species boundaries and resolve species relationships. Sequence similarity comparisons were also conducted between the two ribosomal markers and between the ITS1 and ITS2 regions. Gyromitra gigas sensu stricto and two additional species were discovered within the G. gigas clade. Gyromitra ussuriensis was rediscovered as a distinct taxon and removed from synonymy under G. gigas. It occurs in central and eastern Asia, whereas G. gigas occurs mostly in Europe but also extends into central Asia. A neotype is designated for G. ussuriensis. A new species, Gyromitra americanigigas, is described and illustrated from eastern North America. Although morphology and the LSU exhibited little variation among the three species, the ITS1 and ITS2 regions displayed similar interspecific sequence variability around 0.5–1%, which is sufficient for species identification at the molecular level.

Gyromitra gigas is a widespread taxon that is infrequently collected throughout Europe (Carbone et al. 2018), but has also been reported from China, Japan, North America, and Russia (MyCoPortal 2022). An epitype (MBT 383600) specimen has been designated from the Czech Republic (Carbone et al. 2018). It is reported to grow near or on rotten logs and old stumps in woods of Abies, Betula, Carpinus, Picea, Populus tremula, Quercus, and Tilia from mid-March to early May (Carbone et al. 2018). ITS sequence data has recently shown its distribution to be limited to Europe, with single reports from China and Russia (Miller et al. 2020). Although the European name G. gigas has been frequently used for North American material, this species does not occur in North America. Rather, G. korfii occurs throughout eastern North America, whereas G. montana occurs primarily in western North America and Canada (Miller et al. 2020).
Gyromitra ussuriensis was described in 1950 from the Ussurisky Nature Reserve (formerly known as the Suputinsky Nature Reserve) in Far East Russia (Vassiljeva 1950). It is infrequently collected and reported to grow on Section Editor: Roland Kirschner * Andrew N. Miller amiller7@illinois.edu 1 rotten logs and stumps of Pinus koraiensis and on dead trunks and stumps of Betula costata from late May to early June. It was compared to G. gigas because of its similar ascospores, but distinguished by the smaller hymenophore with a free edge, longer stem, and its growth on wood (Vassiljeva 1950). Since he found no morphological differences, Raitviir treated G. ussuriensis as a synonym of G. gigas (Raitviir 1970) where it remains today (Carbone et al. 2018, Index Fungorum). During our ongoing taxonomic and systematic studies of Gyromitra Fr., two well-supported clades occurring near G. gigas were discovered that represent cryptic species, one previously described representing G. ussuriensis and the other an undescribed species. The goals of this study were to sample and sequence multiple representatives for these two taxa as well as G. gigas and to establish species boundaries, reconcile species relationships and biogeographical distributions, and assess the potential of ITS and LSU for resolving these cryptic species.

Specimens examined
Entire dried ascomata or small portions of the hymenophore of ascomata in 1.5 mL centrifuge tubes were sent to the first author either as loans or as gifts. Sequences generated during this study were obtained from DNA extracted directly from these dried ascomata, which were deposited at ILLS or are available at their home institution (BPI, CUP, F, ILLS, LE, MICH, MIN, NY, O, TAAM, TNS, and YSU). Fungarium acronyms follow Index Herbariorum (Thiers 2013, continuously updated).
Morphological descriptions are based on notes taken from fresh collections and associated photographs, dried fungarium specimens, and other sequenced material. Micromorphological examination followed Miller et al. (2020). The following were calculated for ascospore lengths and widths for each specimen: the range in minimum and maximum values, the mean length (L m ), mean width (W m ), length-width ratio (Q), and mean length-width ratio (Q m ). The lengths of the ascospore apiculi are included in the measurements of ascospore length. Small fragments of the hymenial layer were rehydrated in 5% KOH, washed in distilled water, and sections were prepared at 25 μm thickness using a Physitemp BSF-3 freezing stage mounted on a Leica SM2000 sliding microtome. Images of micromorphological features were captured with an Olympus DP22 digital camera mounted on a BX52 compound microscope using Olympus Imaging Software Cell^D and processed using Adobe Photoshop 2021 (Adobe Systems Inc., Mountain View, California).

Molecular data
Methods for the extraction, PCR amplification, and sequencing of the internal transcribed spacer (ITS) region and the first 600 bp of the 5' end of 28S nuclear ribosomal large subunit (LSU) followed Miller et al. (2020). Sequences were produced at the Roy J. Carver Biotechnology Center at the University of Illinois Urbana-Champaign, and consensus sequences were assembled with Sequencher 5.4 (Gene Codes Corp., Ann Arbor, Michigan, USA).

Phylogenetic analyses
ITS and LSU datasets were individually aligned using MUSCLE® as implemented in Sequencher 5.4. Since most taxa had missing data in the two datasets, portions of the 5' and 3' ends were excluded from all analyses. Both the ITS and LSU datasets possessed little sequence variation in their final alignments so removal of ambiguously aligned regions was unnecessary. The ITS and LSU alignments are included as FASTA-formatted files (Supplementary). Both ITS and LSU alignments were rooted with G. montana based on previous analyses (Miller et al. 2020). The Hasegawa-Kishino-Yano (HKY) model (Hasegawa et al. 1985) was determined to be the best-fit model for both datasets by jModeltest (Darriba et al. 2012;Guindon and Gascuel 2003) based on the Akaike information criterion (AIC) (Posada and Buckley 2004). A maximum likelihood (ML) analysis with the HKY model and all parameters optimized was performed with 1000 bootstrap replicates using PhyML as implemented in Seaview 4.7 (Gouy et al. 2010). An additional ML analysis with a GTRCAT approximation and 1000 bootstrap replicates employing GAMMA model of rate heterogeneity and the rapid bootstrapping option (Stamatakis et al. 2008) was also performed using RAxML-HPC2 v.8.2.12 (Stamatakis 2014) in the CIPRES Science Gateway v.3.3 portal (Miller et al. 2010). Clades with bootstrap values (BV) ≥ 70% were considered significant and supported (Hillis and Bull 1993). Bayesian analyses were performed under the above model using MrBayes v 3.2.7 Ronquist 2001, 2005) on the CIPRES 3.3 portal. The Bayesian analyses were run for 1,000,000 generations which was when the average standard deviation of split frequencies was below 0.01 with trees saved every 1000 generations and burn-in set at 25%. Bayesian posterior probabilities (BPP) were determined from   a consensus tree using PAUP* 4.0b10 (Swofford 2002), and clades with BPP ≥ 95% were considered significant and strongly supported (Alfaro et al. 2003;Larget and Simon 1999).

Sequence similarity comparisons
Comparisons between ITS and LSU sequences and between the ITS1 and ITS2 regions were made in PAUP v.4.0a (build 166) (Swofford 2002) with distance set to uncorrected "p." Mean and range were calculated for infraspecific and interspecific variation. The ITS1 and ITS2 regions were delimited using the ITSx program in PlutoF (Abarenkov et al. 2010).

Phylogenetic analyses
PCR amplification and Sanger sequencing of ITS and LSU were largely successful from DNA extracted from most specimens, even those over 60 years old (Table 1). The final ITS alignment of 75 sequences consisted of 738 nucleotides after the removal of nucleotides on the 5' and 3' ends due to missing characters in most taxa. No ambiguously aligned characters occurred in the ITS alignment. The ITS contained 49 parsimony-informative characters with gaps treated as missing characters: 34 in the ITS1 region and 15 in the ITS2 region. The final LSU alignment of 40 sequences consisted of 864 nucleotides after the removal of nucleotides on the 5' and 3' ends due to missing characters in most taxa. No ambiguous regions were present in the LSU dataset. The LSU contained only 3 parsimony-informative characters and lacked sufficient phylogenetic signal to differentiate among these putative taxa (data not shown) so phylogenetic relationships are based only on the ITS dataset.
Analyses of the ITS dataset generated identical most-likely trees in both the PhyML and RAxML analyses, except BV were higher in the RAxML tree (Fig. 1). Three distinct, wellsupported monophyletic clades were formed that corresponded to three closely related, but separate species. Gyromitra gigas formed a highly supported clade with 90% BV and significant BPP. The clade containing members of G. ussuriensis was supported with 70% BV and significant BPP. The new species, G. americanigigas, is well-supported with 96% BV, but without significant BPP. Gyromitra gigas and G. ussuriensis occurred as sister taxa in a clade supported by significant BPP.

Distribution
Each of these three species inhabits a specific geographic region with some overlap in central Russia between G. gigas and G. ussuriensis (Fig. 2). All known species in the G. gigas species complex are shown on the map for clarity. Gyromitra americanigigas occurs throughout northeastern USA and southeastern Canada. Its range overlaps with G. korfii in Michigan and New York and with G. montana in Michigan and Canada. Gyromitra gigas occupies a large range extending from western Europe to central Russia and overlaps with G. ticiniana in France, Italy, and Turkey and with G. ussuriensis in central Russia. Gyromitra ussuriensis occurs mostly in eastern China, Japan, eastern Russia, and South Korea. Gyromitra khanspurensis is known only from its type locality in Pakistan, whereas G. pseudogigas has only been collected in Sichuan Province of China. Etymology: Named for the combination of "americana" and "gigas" to describe this new species from North America that is closely related to the European G. gigas.
Ecology and distribution: Solitary to scattered on soil or decayed wood of Betula, Picea, Pinus korajensis, and Populus tremula in temperate coniferous and deciduous-coniferous forests in May and June. Known from eastern China, Japan, South Korea, and central and eastern Russia.

Sequence similarity comparisons
The ITS region was compared to LSU to investigate the infraspecific and interspecific variability of these two common molecular markers. Infraspecific ITS sequence variation on average was zero in G. americanigigas and G. gigas and 0.1% in G. ussuriensis (Table 2). Infraspecific LSU sequence variation on average was zero in all three species. Interspecific ITS sequence variation on average ranged from 0.7% between G. americanigigas and G. gigas to 1.2% between G. americanigigas and G. ussuriensis. This is compared to 5.6-6.5% sequence variation between these three species and G. montana. Interspecific LSU sequence variation on average ranged from zero between G. americanigigas and G. ussuriensis to 0.1% between the other two combinations. Sequence variation was equally low ranging from 0.2 to 0.4% between the three target species and G. montana. Whereas the ITS displayed a significant barcode gap and can be used to recognize these three species, the LSU had no barcode gap with infra-and interspecific variation nearly identical.
The ITS1 and ITS2 regions were compared to determine whether either one of these two regions could be used as a barcode marker for molecular identification of these taxa as is the case for most environmental sampling studies (Table 3). Infraspecific sequence variation in the ITS1 ranged between 0 and 0.3% and averaged 0-0.1%. Infraspecific sequence variation in the ITS2 ranged between 0 and 0.7% and averaged 0-0.1%. The ITS1 region contained more than twice the number of parsimony-informative characters compared to the ITS2 Table 2 Infraspecific and interspecific sequence variation of the ITS and LSU for G. americanigigas, G. gigas, G. ussuriensis, and the outgroup, G. montana. Mean and range (in parentheses) of percent differences based on uncorrected "p" sequence differences are shown for ITS along the upper diagonal and for LSU along the lower diagonal region (34 vs. 15), and thus, it was expected that, in general, the ITS1 would vary twice as much as the ITS2. Interspecific variability was three times higher in the ITS1 versus the ITS2 region in comparisons of G. montana to the three target species, a result consistent with our previous findings (Miller et al. 2020). Interspecific ITS1 sequence variation on average ranged from 0.9% between G. americanigigas and G. gigas (and 0.9% between G. americanigigas and G. ussuriensis) to 1.2% between G. gigas and G. ussuriensis. Interspecific ITS2 sequence variation on average ranged from 0.6% between G. americanigigas and G. gigas to 1.3% between G. americanigigas and G. ussuriensis. Whereas infraspecific variation averaged less than 0.1%, interspecific variation averaged 0.6-1.3% for all three species comparisons. The ITS1 and ITS2 regions both displayed similar sequence variability among these three species, justifying the use of either the ITS1 or ITS2 region or both as a molecular barcode marker for species identification.

Species concepts
Species within the G. gigas monophyletic group are characterized by stipitate ascomata, hymenophores that are saddle-shaped to irregularly lobed or cerebriform and wrinkled, and yellow-brown to brown to reddish brown, ribbed to sulcate, white to yellow-brown stipe, ellipsoid to fusiform ascospores that are roughened to finely reticulate and uniguttulate or triguttulate with inconspicuous to distinctive single apiculi that are up to 4 μm long. Although we attempted to apply a polyphasic approach to species delimitation, our morphological examination revealed no significant characters to distinguish G. americanigigas, G. gigas, and G. ussuriensis. However, there appears to be some geographical signal to the speciation that has occurred in this group (Fig. 2). Attempts to culture species in the G. gigas complex failed, and we are unaware of anyone else successfully doing so (Healy et al. 2022), preventing the use of intercompatibility tests to facilitate the delimitation of species boundaries under the biological species concept. The LSU cannot be used as a barcode marker since the infra-and interspecific variation among sequences of these three species was nearly identical. However, the ITS does contain an adequate barcode gap and can be used for species identification of these three species using a phylogenetic species concept.

Discussion
Although species concepts in the Gyromitra gigas species complex were believed to be resolved (Miller et al. 2020), some secrets still remained due to cryptic speciation. Three well-supported clades representing G. gigas, a new species (G. americanigigas), and a rediscovered species (G. ussuriensis) were revealed in our phylogenetic analyses (Fig. 1). Unfortunately, these three species do not possess any significant differences by which to delimit them morphologically. Although Vassiljeva (1950) originally described the ascospores of G. ussuriensis as being longer than in G. gigas, we could not confirm her results. She also distinguished it from G. gigas as having a smaller hymenophore with a free margin, long stipe, and growth on wood, but these characters are also  shared by G. gigas (Carbone et al. 2018). However, these three species do occupy somewhat distinct geographical areas with G. americanigigas confined to Canada and USA, G. gigas occurring mostly in Europe, and G. ussuriensis found in Asia (Fig. 2). Some overlap between G. gigas and G. ussuriensis does occur in central Russia, and G. gigas and G. ticiniana are both found in France, Italy, and Turkey. There is also overlap in the distributions of G. americanigigas, G. korfii, and G. montana, with all three species found in Michigan, USA. Sequencing of all or part of the ITS region is highly recommended for species identification in this group. The ITS1 and ITS2 regions both display a small, but adequate barcode gap for identifying these three species either through nBLAST similarity searches or phylogenetic analyses. The infraspecific variation averaged less than 0.1%, whereas the interspecific variation averaged 0.6-1.3% for all three species comparisons. Ideally, species hypotheses should be based on ITS interspecific variation higher than 0.5-1.5%, but as more fungal species complexes are studied and more cryptic speciation is discovered, these lower percentages for species differentiation using the ITS fungal barcode may be the norm (Lücking et al. 2020). Future studies of freshly collected specimens should sequence one or more protein-coding genes, or ideally whole genomes, and use genetic discontinuity models to test our species hypotheses (Matute and Sepúlveda 2019), which would most likely corroborate our ITS data derived from mostly older fungarium specimens.