Enlightening the black and white: species delimitation and UNITE species hypothesis testing in the Russula albonigra species complex

Russula albonigra is considered a well-known species, morphologically delimited by the context of the basidiomata blackening without intermediate reddening, and the menthol-cooling taste of the lamellae. It is supposed to have a broad ecological range and a large distribution area. A thorough molecular analysis based on four nuclear markers (ITS, LSU, RPB2 and TEF1-α) shows this traditional concept of R. albonigra s. lat. represents a species complex consisting of at least five European, three North American, and one Chinese species. Morphological study shows traditional characters used to delimit R. albonigra are not always reliable. Therefore, a new delimitation of the R. albonigra complex is proposed and a key to the described European species of R. subgen. Compactae is presented. A lectotype and an epitype are designated for R. albonigra and three new European species are described: R. ambusta, R. nigrifacta, and R. ustulata. Different thresholds of UNITE species hypotheses were tested against the taxonomic data. The distance threshold of 0.5% gives a perfect match to the phylogenetically defined species within the R. albonigra complex. Publicly available sequence data can contribute to species delimitation and increase our knowledge on ecology and distribution, but the pitfalls are short and low quality sequences. Supplementary Information The online version contains supplementary material available at 10.1186/s43008-021-00064-0.


INTRODUCTION
Molecular identification of species gained importance over the last decades (Matute and Sepulveda 2019). As new techniques became available and more easily accessible, the number of publications using sequence data increased immensely (Hibbett et al. 2011;Kõljalg et al. 2013;Nilsson et al. 2018). The main problems with molecular identification are poor taxon coverage and misidentifications in many public sequence databases, as well as high infraspecific variability of DNA regions causing poor performance of the barcoding gap (Kõljalg et al. 2005;Badotti et al. 2017;Hofstetter et al. 2019). To overcome some of these problems, the UNITE database was created (https://unite.ut.ee/). UNITE targets the still most widely used, universal fungal barcode: the nuclear internal transcribed spacer (ITS) region to provide highquality reference records (Nilsson et al. 2018). UNITE groups individual ITS sequences into species hypotheses (SHs) at several distance thresholds (i.e. between 0 and 3%), each assigned a unique digital object identifier (DOI) which allows unambiguous reference across studies. These species hypotheses are either assigned automatically with a representative sequence or manually by a taxonomic expert with a reference sequence (Kõljalg et al. 2013;Nilsson et al. 2018).
Most Basidiomycota, the second largest phylum of fungi, are Agaricomycotina (Naranjo-Ortiz and Gabaldon 2019). A considerable part of this subphylum's diversity is concentrated in large genera of ectomycorrhizaforming agarics, e.g. Cortinarius and Russula, which exhibit very different evolutionary rates (Ryberg and Matheny 2012;Varga et al. 2019). Recent studies on lineages of closely related Russula members revealed that they often comprise closely related species diversified by ecological adaptation and isolation by distance or disjunction (Adamčík et al. 2016b;Caboň et al. 2019;Looney et al. 2020). This makes the recognising of these species relevant, despite their similarity in the ITS barcode. The efforts to barcode fungal species sometimes fail, due to lack of taxonomic studies addressing the species concept based on type collections (Kõljalg et al. 2020). Our current studies on Russula subgen. Compactae recognised Russula albonigra as one such example. Its morphological concept is historically established, and the species is traditionally recognised by the moderately distant, relatively narrow lamellae, the menthol-cooling taste of the lamellae, and the surface of the cap and the stipe, as well as the lamellae, that are strongly and rapidly blackening (hence, resulting in black-and-white contrast of bruised and untouched parts). Most publications report blackening without intermediate reddening of the context and surface (Romagnesi 1967;Adamčík and Buyck 2014). Microscopically, the species is defined by a low reticulate spore ornamentation and pileocystidia without apical knobs (Romagnesi 1967). However, our current search in two databases, which seek to adopt a nomenclatural concept for fungal operational units defined by the ITS barcode, recovered inconsistencies. Samples identified as R. albonigra in the BOLD database (https://www.boldsystems.org/) do not match any of the species hypothesis under this species name published in UNITE (https://unite.ut.ee/), and vice versa.
In this study, we use sequence data of four DNA markers and a detailed morphological revision to test the taxonomic status within the R. albonigra complex. To test different distance thresholds of UNITE species hypotheses, we used the strict genealogical concordance assessing the extent of genetic concordance across loci, the coalescent based species delimitation modelling the genealogical history of individuals back to a common ancestor and morphological differences.

Sampling
This study is based on collections from sampling trips in Belgium (2016Belgium ( , 2017Belgium ( and 2018, Italy (1997Italy ( , 2000Italy ( and 2016, Norway (2016), Slovakia (2003, 2006, 2008, 2009, 2011), and Sweden (2016. All collections are deposited in the Herbarium Universitatis Gandavensis (GENT) or the Slovak Academy of Sciences (SAV). Supplementary collections were requested from the Mycological Department of the National Museum in Prague, Czech Republic (PRM), and from the personal collections of Felix Hampe, Jesko Kleine, and Helga Marxmüller (the latter recently deposited in the State Museum of Natural History Karlsruhe, KR).
Samples that could belong in the Russula albonigra complex were selected based on morphology: (1) collections that were identified as R. albonigra in the field based on macro-morphology; and (2) fungarium collections labelled as R. albonigra based on both macro-and micro-morphological observations. The selected samples were molecularly checked using the ITS marker as a guideline and not included for further study when not placed within the R. albonigra complex.

Morphological analysis
The macroscopic description is based on observations from fresh material, with colour codes following Kornerup and Wanscher (1978), guaiac reactions referring to Chalange (2014), and spore print colour codes following the scale of Romagnesi (1967). The microscopic description and terminology follow Adamčík et al. (2019). Microscopic characters were studied from dried material, spores were observed in Melzer's reagent, elements of the hymenium and pileipellis were observed in Congo red in L4 after ca. 10 s in KOH 10%. Basidiospores were measured using a crosshair eyepiece on a Zeiss Axioskop 2 microscope. Line drawings of spores were made based on stacked photographs (Nikon Eclipse Ni-U microscope, stacking software: Extended Depth of Field, Nikon Nis Elements module) at an original magnification of 5000×. Measurements of other elements were made using an eyepiece micrometer and line drawings were prepared with the aid of a camera lucida (Olympus U-DA) on an Olympus CX21 or BX43 microscope, at original magnifications of 1500× or 2000×. Tissues were mounted in Cresyl Blue (Buyck 1989), sulfovanillin (Cabon et al. 2017) and treated with carbolfuchsin (Romagnesi 1967) to observe the presence and colour changes of incrustations and cystidium contents. All cited collections in the species descriptions have been sequenced, at least for ITS.
The key provided in the Taxonomy section is based on our observations of species within the R. albonigra complex, and following the traditional species concepts in literature for the other taxa. Russula clementinae is not included in the key as it is interpreted as a synonym of R. densifolia (Sarnari 1998).

Molecular analysis
DNA extraction and amplification was performed in either the molecular laboratory of Ghent University or that of the Slovak Academy of Sciences. Sequencing of PRM collections was conducted within the study of Leonhardt et al. (2019).
In Ghent University, DNA from fresh material was extracted using the CTAB extraction described in Nuytinck and Verbeken (2003). DNA from dried material was extracted using a modified CTAB protocol (Tel-Zur et al. 1999; modified by Meise Botanic Garden and Research Group Mycology of Ghent University). The original protocol was optimized for cacti species. These plant types tend to have larger cells compared to fungi. Hence, the ratio DNA content:biomass for fungi is much higher. Therefore, less biomass is needed as starting material and buffer volumes were adjusted consequently. Because mucilaginous polysaccharides have not been observed in previous Fungal DNA extractions the use of sorbitol in our extractions was omitted. Additionally, the use of CTAB as detergent to brake open fungal cells seemed to suffice to access the fungal DNA in an efficient way. Hence, the extra sarkosyl added in the lysis step was also omitted from our protocol. Protocols for PCR amplification follow Le et al. (2007). In the Slovak Academy of Sciences, total genomic DNA was extracted from dried material using the EZNA Fungal DNA Mini Kit (Omega Bio-Tek, Norcross, GA, USA) following the manufacturer's instruction. Amplification of DNA was performed in a PCR reaction mix consisting of approximately 2 ng/μl of template DNA, forward and reverse primers (10 pmol/μl), 5× HOT FIREPol® Blend Master Mix (Solis BioDyne, Tartu, Estonia) and molecular grade water added up to 20 μl. Four nuclear markers were amplified: (1) the internal transcribed spacer region of ribosomal DNA (ITS), comprising the ITS1 and ITS2 spacer regions and the ribosomal gene 5.8S, using primers ITS1-F and ITS4 (White et al. 1990;Gardes and Bruns 1993); (2) a part of the ribosomal large subunit 28S region (LSU), using primers LR0R and LR5 (Moncalvo et al. 2000); (3) the region between the conserved domains 6 and 7 of the second largest subunit of the RNA polymerase II (RPB2), using primers bRPB2-6F and fRPB2-7cR or bRPB2-7.1R (Liu et al. 1999;Matheny 2005); and (4) the translation elongation factor 1-alpha (TEF1α), using primer pairs EF1-1018F and EF1-1620R or tef1F and tef1R (Morehouse et al. 2003;Stielow et al. 2015). PCR products from Ghent University were sequenced using an automated ABI 3730 XL capillary sequencer at Macrogen. In the Slovak Academy of Sciences, the PCR products were purified using Qiaquick PCR Purification Kit (Qiagen, Hilden, Germany) and samples were sequenced by the Seqme company (Dobříš, Czech Republic).
Forward and reverse sequences were assembled into contigs and edited where needed with BioloMICS (BioAware SA NV). All sequences generated were deposited in GenBank (Table 1).

Phylogenetic analysis
Identifications of publicly available sequences of fungi often match contrasting taxonomic species concepts. To provide reliable sampling in accordance with traditional species concepts we selected three representative collections of each European species described within R. subgen. Compactae, identified using the most recent and reliable keys (Romagnesi 1967;Sarnari 1998). These collections, together with the collections of R. albonigra s. lat. were sequenced by us (Table 1). For non-European species, sequences used in Adamčík et al. (2019) or Buyck et al. (2020) were included if multiple of the markers used in this study were available for these samples, with an ITS sequence obligatory. Four species of Russula subgen. Archaeae were used as an outgroup, because the recent phylogenies of the genus place this subgenus as sister to R. subgen. Compactae.
Sequences were aligned using the online version of the multiple sequence alignment program MAFFT v7 (Katoh and Toh 2008), using the E-INS-i strategy. Trailing ends of the alignments were trimmed and the alignments were manually edited when necessary in MEGA7 (Kumar et al. 2016). The alignments can be obtained from the first author and TreeBASE (Submission ID 26815). The alignments were partitioned into following partitions: ITS-LSU-alignment: partial 18S, ITS1, 5.8S, ITS2, LSU; RPB2-alignment: the RPB2 intron and the first, second and third codon positions of the exon; TEF1α-alignment: the first and second intron and the first, second and third codon positions. PartitionFinder2 was used to find the appropriate partitioning scheme and substitution models using the Akaike information criterion (AICc) with a greedy search over all models (Guindon et al. 2010;Lanfear et al. 2012;Lanfear et al. 2017). Maximum likelihood (ML) analyses were conducted with IQ-Tree (Nguyen et al. 2014;Chernomor et al. 2016) using standard bootstrapping analysis (1000 replicates). Bayesian inference (BI) was executed with MrBayes v3.2.6 (Huelsenbeck and Ronquist 2001;Ronquist and Huelsenbeck 2003). Two independent parallel runs of one cold and three heated chains were run for ten million (single-locus datasets) or twenty million generations (multi-locus dataset) with a sample frequency of 100. Potential Scale Reduction Factor (PSRF) values approached 1.0. Convergence and Effective Sample Size (ESS) statistics of the runs were also examined with Tracer v1.7.1 (Rambaut et al. 2018). A burn-in sample of 20% was excluded before constructing the majority rule consensus tree. Analyses were first performed on each alignment separately and visually checked for incongruence. Significant incongruence was assumed if two different relationships (one monophyletic and the other non-monophyletic) for any set of taxa were supported with bootstrap values (BS) ≥ 70 or posterior probabilities (PP) ≥ 90. The resulting gene trees did not show any supported conflicts, therefore all alignments could be concatenated. The concatenated alignment was used for the multi-locus phylogenetic analyses ( Fig. 1).

Coalescent species delimitation approaches
For species delimitation under the multispecies coalescent model a part of the alignment used in the multilocus phylogenetic analyses mentioned above, comprising members of the Russula albonigra complex, was used. A total of five potential species units (as proposed by the ML and BI trees) were evaluated as the full model. Two coalescent species delimitation methods were performed to test these species hypotheses. The first method was implemented in Bayesian Phylogenetics and Phylogeography, BP&P v4.3.8 (Yang 2015). We performed analysis A11 (Yang and Rannala 2014) for unguided species delimitation using rjMCMC algorithm 0 (Yang and Rannala 2010). Analyses were run with several values for the fine-tune parameter (ε = 2, 5, 10 and 20) and we assigned equal probabilities to the rooted species trees as a species model prior. Because the prior distributions on the ancestral population size and root age can affect the posterior probabilities of the model, we considered three different combinations of priors (based on the idea of Leache and Fujita (2010)): (1) θĨ G(3,0.002) and τ~IG(3,0.002), (2) θ~IG(3,0.02) and τ~IG(3,0.02), and (3) θ~IG(3,0.02) and τ~IG(3,0.002) (with α = 3 for a diffuse prior as proposed in the manual). For each combination of settings the analysis was run twice with a different seed (to confirm consistency between runs) for 200,000 generations (sampling interval of two) and a burn-in of 50,000. As a second species delimitation method we used the STACEY v1.2.5 (Jones 2017) package implemented in BEAST2 (Bouckaert et al. 2019). The xml-file for the BEAST2 runs were prepared in BEAUTi v2.6.3 (Drummond et al. 2012). We used following partitions: for the nrDNA (1) 5.8S, (2) ITS1 + ITS2 and (3) LSU; for the protein coding loci the introns and the first, second and third codon positions of the exons. PartitionFinder2 was used to find the appropriate substitution models. The substitution rate of each partition was estimated independently of the others. Clock and tree model parameters were estimated independently for the nrDNA and each protein coding locus. We used a lognormal, relaxed clock model and a Yule tree model. The Collapse Height parameter ε was set to 10 − 5 . The Collapse Weight parameter ω was estimated and given a uniform prior on [0,1] so that every number of species between five and one is regarded as equally likely a priori. We ran five parallel MCMC runs for one  billion generations sampling every 1000th tree. Convergence and Effective Sample Size (ESS) statistics of the runs were examined with Tracer v1.7.1. Twenty percent of each run was discarded as burn-in and the remaining posterior samples were combined using LogCombiner v2.6.3 (Drummond and Rambaut 2007) and used to calculate the most likely number of clusters (i.e., putative species), using SpeciesDelimitationAnalyzer (Jones et al. 2014).

Species hypothesis and threshold testing
ITS sequences generated by authors of this study and used in the multi-locus phylogeny were combined with all ITS sequences, either labelled as R. albonigra or showing high similarity (97%) to our sequences of R. albonigra s.lat., available on UNITE (https://unite.ut.ee/), GenBank (www.ncbi.nlm.nih.gov) and BOLD (https://www. boldsystems.org/) databases. Accession numbers are given in Fig. 2. The clade containing R. nigricans and R. dissimulans is in a basal position within R. sect. Nigricantinae (see Fig. 1) and is chosen as the outgroup. Short sequences (containing only ITS1 or ITS2) and sample UDB065518 (containing many ambiguities and differences in conserved domains compared to other sequences of the group) were excluded from the analysis. The sequences were aligned following the same principles as mentioned above. The alignment was partitioned into following partitions: ITS1, 5.8S and ITS2. A ML analysis was conducted with IQ-Tree (Nguyen et al. 2014;Chernomor et al. 2016) using the option to first test for the best substitution model (Kalyaanamoorthy et al. 2017) and standard bootstrapping analysis (1000 replicates). The excluded samples were later plotted on the tree based on similarities in distinguishing nucleotide positions (Additional file 1). We identified the single nucleotide positions distinguishing the species of the R. albonigra complex and compared the excluded samples to the good quality sequences based on these positions to estimate their placement in the tree. SH inclusiveness across sequence distance threshold values, as it is shown on UNITE, is plotted against the tree.

Multi-locus phylogenetic analyses
All 24 sequences generated by the authors of specimens putatively identified as R. albonigra, are placed within R. subgen. Compactae and group together with other members of R. sect. Nigricantinae (Fig. 1). All but two of these sequences are placed in one strongly supported clade, here further referred to as the R. albonigra complex. The two sequences outside the R. albonigra complex are placed in either the R. atramentosa clade or the R. anthracina clade (these sequences were not used for the final analysis and are not shown in the trees). Our molecular analysis shows the presence of five distinct European clusters within the R. albonigra complex ( Fig.  1). They form four well supported clades and one singleton collection on a long branch. The overall topology of the ML and BI tree was congruent. The name R. albonigra is assigned to a clade with two collections from the Czech Republic, PRM 924409 and PRM 934322, that originate from the type collecting area. Three species, R. ambusta, R. nigrifacta and R. ustulata, are described here as new. One species is represented by a singleton position in the tree, placed as sister to R. albonigra and is labelled as Russula sp. 1. These two species form a sister clade to a larger clade containing R. ambusta, R. nigrifacta, and R. ustulata. The relations within the latter clade are not well supported in the BI tree.

Coalescent-based species delimitation
The full set of proposed species (i.e. five species) was recovered as the highest supported species model in the BP&P analysis under each combination of settings, with posterior probabilities ranging from 0.91 to 0.99. Also the STACEY analysis resulted in the highest probability (posterior probability of 0.99) for five minimal clusters (species). Both coalescent delimitation methods confirmed the species hypothesis for all five clusters in the multilocus phylogenetic analysis.

ITS analysis and optimal SH distance
When searching for UNITE species hypotheses labelled as R. albonigra, at every threshold, two species hypotheses are found that are not placed within the R. albonigra complex defined by our multi-locus analysis. Both are represented by a singleton sequence. The first, UDB024525 represents a collection from Lao People's Democratic Republic and seems more closely related to R. atramentosa. The second, JF908707 represents a collection from Italy with an isolated position in the phylogeny (Fig. 2).
The general topology of the ITS tree is congruent with the multi-locus tree and all sequences of the R. albonigra complex generated by the authors of this study are again placed within this monophyletic group (Fig. 2). The ITS analyses revealed the presence of three additional North American clusters and one Chinese collection of singleton position, within the R. albonigra complex. Two of the North American clusters are supported and probably represent well defined species, while the status of the three sequences from the USA (KF306041, JF834355 and KF306040) is uncertain and requires more sequence data to resolve. The singleton Chinese sample (KX441086) probably represents an undescribed species sister to R. albonigra. Furthermore, Fig. 2 shows an overview of the different UNITE species hypotheses within the R. albonigra complex at different thresholds. At a threshold of 1% or higher two species hypotheses are recognised (for SH numbers see red and pink boxes in the Fig. 2). The first one labelled as R. albonigra with the representative UNITE sequence UDB016040 covers R. ambusta, R. nigrifacta, R. ustulata and the North American species. The second one with the representative sequence JF519228 is labelled as Russula sp. and covers R. albonigra and the Chinese species. At a threshold of 0.5%, all European species (except for R. sp. 1 which is not represented by any public sequence) and two North American species are supported. Threshold < 0.5% gives additional units within the phylogenetic species that were not supported by our multi-locus analysis.

TAXONOMY
The species within the Russula albonigra complex are characterised by the moderately distant, relatively narrow lamellae, the context that is rapidly and strongly blackening, generally without intermediate reddening. In some cases though, some slight reddening is observed. The taste of the lamellae and flesh is never acrid, but can be menthol-cooling. Microscopically, the species are defined by spores with low and dense warts forming subreticulate to reticulate ornamentation, long pileocystidia (if present) and a cystidial content which is not reacting in sulfovanillin.
Description: Pileus large, 56-112 mm diam., planoconvex, at the centre with shallow but wide depression; margin deflexed, long involuted, not striated, smooth; pileus surface velvety and smooth near margin, towards the centre radially wrinkled or rugulose, centre smooth, dry, matt, almost not peeling (max. to 1/3 of the radius); young completely white, later becoming yellowish white (4A2), cream (4A3) to orange-grey (5B2) at the centre, more greying and blackening when old. Lamellae segmentiform to subventricose, to 6 mm deep, adnate to subdecurrent; snow white, later yellowish white (4A2), blackening with age or when bruised; with numerous lamellulae of different lengths, frequently forked near the stipe but also near the pileus margin, often anastomosed; moderately crowded to moderately distant, L = 190-260, l = 1 (one between each pair of long lamellae); edges even, concolorous, blackening with age.
Stipe 42-60 × 14-29 mm, cylindrical, firm and fleshy, longitudinally striated, velvety near the lamellae; white, later becoming greyish orange (5B4) near the base; interior solid, cortex ca. 2.5 mm thick. Context ca. 7 mm thick at mid-radius, hard, white, turns rapidly grey and then black on cut section, at surface also turns red before grey and black; turning orange with FeSO 4 , immediately dark blue with guaiac (strong reaction, +++); taste mild, slightly like mint (refreshing) in lamellae, odour weak of apples. Spore print white (Ia). Basidiospores    Notes: Russula albonigra was first described by Krombholz (1845) as Agaricus alboniger with only a brief description and no holotype designated. Later, Fries classified it in the genus Russula under its current name (Fries 1874) that became well-known and widely used in Europe. The illustration (Krombholz 1845: pl. 70, Figs. 16 and 17; Biodiversity Heritage Library), reproduced in Fig. 7 and made by Krombholz was suggested to serve as lectotype by Sarnari, although it was never formally designated (Sarnari 1998). The illustration is the only available original material and is hereby formally designated as the lectotype. The brief description and the plate itself are not sufficient to determine which species of the R. albonigra complex corresponds to R. albonigra. This makes the collecting area of Krombholz the most relevant criterion and in Krombholz (1845) it is mentioned that R. albonigra is found in Prague. Within the dataset, collection PRM 924409 was found only 31 km from the Prague city centre. Therefore, the clade in which this collection is placed, is chosen to represent R. albonigra. Specimen SAV F-20197 is here designated as epitype.
MycoBank: MB839080. Etymology: Refers to the appearance of the basidiomata, which look like they were burnt.
Diagnosis: Differs from the other species of the Russula albonigra complex by the intermediate reticulation and density of spore ornamentation and the presence of appendages, but lack of bifurcations on the pileocystidia.
Ecology Notes: The range of the spore size is large within this species. This results from the spore size difference between the collections. The holotype (SAV F-3558) has smaller spores and lower Q-value than collection FH 2008 ST01. The variability in the shape of the hymenial cystidia is also due to the difference between the collections with the holotype having cystidia of type 1 and collection FH 2008 ST01 having cystidia of type 2 (types referring to (1) and (2) in the description). Therefore, based on morphology, we could hypothesize that these collections represent different species. Nevertheless, we treat them here as the same species because phylogenetically there is no support for the hypothesis of two different species. All markers used in this study place these two collections together as the same species. Measurements are given for each collection separately in Supplementary Material 1. Of course, more collections and more micromorphological study are needed in order to understand this intraspecific variation. The UNITE species hypothesis at 0.5% corresponding to our concept of R. ambusta is based on a locked sequence and after this publication we will propose to change the reference sequence to the holotype of the species. De Lange & Adamčík,11,12 and 13).
Diagnosis: Differs from the other species of the Russula albonigra complex by the lack of both appendages and bifurcations on the pileocystidia.
Distribution: Known from Estonia, Italy, and Slovakia.  Notes: Besides the diagnostic features mentioned in the diagnosis, R. nigrifacta is characterised by its higher spore ornamentation (incomplete reticulum) and the thin pileipellis. These features are shared with R. albonigra, which is easily differentiated from the other species in the complex by the oily guttulate content of the cystidia. Furthermore, our data suggests that R. nigrifacta is associated with Quercus spp. (and possibly also with Carpinus betulus) in thermophilous habitats (thermophilous oak forests and Mediterranean oak forests). De Lange & Verbeken,sp. nov. (Figs. 3h,14 and 15).
Etymology: Refers to the appearance of the basidiomata, which look like they were burnt.
Diagnosis: Differs from the other species of the Russula albonigra complex by the absence (or rareness) of pileocystidia.
Ecology Notes: Our data suggests that Russula ustulata has a specific ecology, different from the other species in the complex. It is associated with coniferous trees in boreal forests or mountain habitats.
Notes: Although there is molecular and morphological support that this collection represents a species different from the other species in the Russula albonigra complex, the authors chose to not formally describe this Key to the European species of Russula subgen. Compactae species here as the description is based on a single collection only.

DISCUSSION
Our study recognises five species within the traditional concept of Russula albonigra and this name was the only available at the species rank. Before new names could be assigned to undescribed species, identification of the old name proofed to be challenging, especially because it was already used in the early Friesian period and its original description (Krombholz 1845) does not meet current morphological standards and does not provide sufficient detail. The species clade identified in this study as R. albonigra has the best match with the probable geographic and ecological origin of the species described by Krombholz.
The traditional morphological concept of R. albonigra (Romagnesi 1967;Sarnari 1998;Kibby 2001) more or less agrees with our observations of the species complex. We think that in the field, for the preliminary identification of species to the R. albonigra complex, the following characters are most helpful: a strong and fast blackening of the surface resulting in a distinct contrast of wounded compared to untouched areas, the relatively sturdy and thick-fleshed basidiomata, a mild or refreshing taste of the context, moderately distant and strongly blackening lamellae and a dry, usually dull and not viscid pileus cuticle. However, macroscopically the species within the R. albonigra species complex are very alike and cannot be distinguished unambiguously. Nonetheless, we observed that the reaction to FeSO 4 could be interesting to differentiate some of the species. While R. albonigra has an orange reaction to FeSO 4 , both R. nigrifacta and R. ustulata have a greenish reaction. As this data is missing for R. ambusta and R. sp. 1, it is important to pay attention to this character in future collections. There are some interesting observations questioning the traditional characteristics used to define the R. albonigra complex. First of all, it seems that the absolute lack of any reddening is not a reliable feature, because it can be weak and easily overseen or vanishing quickly due to the strong and quick blackening. At least three of the species within the R. albonigra complex (R. albonigra, R. nigrifacta, and R. ustulata) comprise a collection where some weak reddening is observed at the surface or even of the context. Some variability about the reddening reaction was also noted by Romagnesi, who recognised Russula albonigra f. pseudonigricans with an intense reddening context (Romagnesi 1962;Romagnesi 1967). Attempts to get a sequence from the type material failed. The holotype of this form is in a bad condition which does not allow good microscopic observations. Furthermore, a microscopic study of a paratype suggested that holotype and paratype did not represent the same species. We suggest for now, until molecular data becomes available, not to draw any conclusions about the identity of R. albonigra f. pseudonigricans or its classification within the R. albonigra complex. Moreover, names described in our study have priority at species rank over any future combination of the f. pseudonigricans epithet (Art. 11.2 of the ICNafp). Although we believe that the reddening reaction of the context can be used as a diagnostic between species on opposite sides of the spectrum (i.e. species with a strong reddening reaction versus species without a clear reaction) some caution is needed.
Another traditional morphological character to define R. albonigra was the characteristic menthol taste in the lamellae. This does not seem to be a stable character because it was not observed in all collections and it can also depend on the subjective opinion of an individual person. Furthermore, this menthol-refreshing taste is also noted to possibly be present in R. atramentosa by Sarnari (1998).
We found the lack of a reaction of the cystidial content to sulfovanillin to be a good synapomorphic character to define the R. albonigra complex. Russula nigricans is the only species outside this complex also showing no clear reaction of the cystidial content to sulfovanillin. But the latter species can easily be distinguished from the R. albonigra complex by its thick and very distant lamellae, the strong reddening of the context and its pileocystidia that are much shorter (never exceeding 90 μm). Our conclusions about the delimitation of the species complex are only based on observations of European taxa of R. subgen. Compactae.
The second challenge of this study was to define morphological differences among the species of the R. albonigra complex defined by phylogenetic analyses. Due to the low morphological variability we could consider the species within the R. albonigra complex pseudocryptic species (i.e. species with a morphological resemblance that seems indistinguishable at first, but can be distinguished when using the appropriate characters; Delgat et al. 2019). This is a phenomenon that is widely distributed within the Russulaceae, especially in the genus Lactifluus , Van de Putte et al. 2010, Van de Putte 2012, De Crop et al. 2014, Van de Putte et al. 2016, Delgat et al. 2017, De Lange et al. 2018, Delgat et al. 2019), but also within the genus Russula (Adamčík et al. 2016a;Adamčík et al. 2016b;Caboň et al. 2019). The most striking microscopical differences between the species in the R. albonigra complex are the higher spore ornamentation of R. nigrifacta and R. albonigra compared to the other species within the complex. Russula ustulata and R. sp. 1 have an almost complete and denser reticulum than the incomplete reticulum of R. nigrifacta and R. albonigra. The ornamentation of the spores in R. ambusta is intermediate in reticulation and density. R. albonigra is distinguishable by the unique oily guttulate content of all cystidia. R. nigrifacta typically lacks appendages and bifurcations on the pileocystidia whereas these are present in R. albonigra and R. sp. 1. Russula ambusta lacks bifurcations but appendages are present. The most striking feature of R. ustulata is the absence (or rareness) of pileocystidia, whereas R. sp. 1 has very long and numerous pileocystidia. The thickness of the pileipellis is also an interesting character. R. albonigra has the thinnest pileipellis followed by R. nigrifacta, the other species in the complex have a much thicker pileipellis. The presence of inflated subterminal cells in the hyphal terminations of the pileipellis centre is typical for R. sp. 1 and not observed in the other species of the complex.
Collections used in this study often do not have precise ecological details to define ecological niches and host tree preferences. However, habitat type and geographical data suggest biological relevance to recognise closely related species (Ryberg 2015). Russula ambusta, R. nigrifacta, and R. ustulata are closely related but seem to inhabit ecologically different niches. Russula nigrifacta occurs both with Mediterranean oaks (Quercus ilex and Quercus suber) and Quercus robur and Carpinus betulus in thermophilous oak forests. Possibly, Russula sp. 1 has a similar ecology, it is only known from a single collection associated with Mediterranean oak (Quercus suber). Russula ustulata is up to now only known from boreal or mountain habitats, associated with coniferous trees (Picea sp., Pinus sp.). Russula ambusta and R. albonigra seem to have a similar ecology and are associated with a variety of trees in temperate to montane forest types. Russula ambusta was collected with Quercus robur, Pinus sylvestris, and Betula pendula; Russula albonigra with Fagus sylvatica, Abies alba, Picea abies, and Carpinus betulus.
Our study suggests that species within the Nigricantinae clade have a limited area of distribution, unlike what is often believed. All North American collections retrieved from GenBank and placed in the R. albonigra complex are not clustered with the European ones and probably represent different species (Fig. 2). The retrieved ITS data did not confirm that the distribution of R. albonigra is not transcontinental (Singer 1958;Hesler 1961;Shaffer 1962;Kibby and Fatto 1990;Thiers 1994), but rather it supports the hypothesis that none of the European taxa within R. subgen. Compactae are present in the United States (Adamčík and Buyck 2014). The North American collections represent at least two different species with a high macromorphological resemblance to R. albonigra and they may represent R. sordida and R. subsordida, both having a weak or negative reaction of the pileocystidia to sulfovanillin (Adamčík and Buyck 2014).
This study shows that the R. albonigra complex is also represented in Asia by a still undescribed Chinese species.
A multi-locus phylogeny resulting in a strong support of the European species within the R. albonigra complex, stimulated the detailed search for morphological differences between the species. The species described in this study are defined by integrated taxonomy combining multi-locus molecular data with detailed morphological and ecological data (i.e. distribution, climate, and host data).
Our study demonstrated that all species within the R. albonigra complex, supported by the strict genealogic concordance and coalescent-based species delimitation, are strictly distinguished at the threshold of 99.5%, that corresponds to a distance of 0.5% when performing a UNITE search. Even a distance of 1% results in only two UNITE species hypotheses both covering multiple phylogenetic species within the complex. This shows that there is a low genetic diversity of the ITS region between the species within this complex. A possible explanation for this low genetic diversity is that the species within the R. albonigra complex are only relatively recently diverged from each other, which could explain the relatively short branch lengths (Fig. 2) and the low morphological variability. The occupation of new habitats and the adaptation to new hosts could have caused the radiation seen.
This study shows that ITS sequence similarity thresholds of 97% commonly used in metabarcoding studies (Pauvert et al. 2019) are not sufficient to differentiate the phylogenetically defined species of the R. albonigra complex. This observation is also made for other species complexes within the genus Russula (Adamčík et al. 2016b). The species thresholds retrieved in this study agree with the conclusions of testing global fungal databases as training datasets, that predicted the optimal identity thresholds to discriminate filamentous fungal species as 99.6% or 99.3% for ITS (Vu et al. 2019;Vu et al. 2020). Badotti et al. (2017) rank Russula as the genus with only 38% probability of correct identification, but the study also ranks it into the group for which ITS is a good marker while another Russulaceae genus, Lactarius, is placed in a group for which ITS is a poor marker. We do not promote the use of a universal threshold value, but rather emphasise the importance of searching for the best threshold value, according to the fungal group of interest.
Our UNITE species hypothesis threshold testing meets both major problems highlighted by Kõljalg et al. (2013): (1) the lack of an inclusive, reliable public reference data set; and (2) the lack of means to refer to fungal species, for which no scientific name is available, in a standardized stable way. First, our singleton collection of R. sp. 1 is not represented in UNITE, the other four European species are represented by 2-20 sequences. Second, the