Diversity within Aspergillus niger Clade and Description of a New Species: Aspergillus vinaceus sp. nov.

Diversity of species within Aspergillus niger clade, currently represented by A. niger sensu stricto and A. welwitshiae, was investigated combining three-locus gene sequences, Random Amplified Polymorphic DNA, secondary metabolites profile and morphology. Firstly, approximately 700 accessions belonging to this clade were investigated using calmodulin gene sequences. Based on these sequences, eight haplotypes were clearly identified as A. niger (n = 247) and 17 as A. welwitschiae (n = 403). However, calmodulin sequences did not provide definitive species identities for six haplotypes. To elucidate the taxonomic position of these haplotypes, two other loci, part of the beta-tubulin gene and part of the RNA polymerase II gene, were sequenced and used to perform an analysis of Genealogical Concordance Phylogenetic Species Recognition. This analysis enabled the recognition of two new phylogenetic species. One of the new phylogenetic species showed morphological and chemical distinguishable features in comparison to the known species A. welwitschiae and A. niger. This species is illustrated and described as Aspergillus vinaceus sp. nov. In contrast to A. niger and A. welwitschiae, A. vinaceus strains produced asperazine, but none of them were found to produce ochratoxin A and/or fumonisins. Sclerotium production on laboratory media, which does not occur in strains of A. niger and A. welwitschiae, and strictly sclerotium-associated secondary metabolites (14-Epi-hydroxy-10,23-dihydro-24,25-dehydroaflavinine; 10,23-Dihydro-24,25-dehydroaflavinine; 10,23-Dihydro-24,25-dehydro-21-oxo-aflavinine) were found in A. vinaceus. The strain type of A. vinaceus sp. nov. is ITAL 47,456 (T) (=IBT 35556).


Introduction
The black aspergilli are among the most important fungal groups in the genus Aspergillus. These aspergilli include species causing food spoilage and mycotoxin production, often being reported as the third most common Aspergillus spp. associated with invasive human disease and aspergillomas [1,2]. Conversely, this fungal group includes several of the most industrially important species. For instance,

Diversity and Phylogenetic Relationships of CaM Haplotypes
All sequences (the Brazilian Dataset of CaM Sequences and the GenBank Dataset of CaM Sequences) were aligned using Multiple Sequence Comparison by Log-Expectation (MUSCLE) [18] in the MEGA 7 computational package [19].
The haplotype identification, haplotype diversity (Hd) and nucleotide diversity (π) were performed using the DNA sp5.10 computational package [20]. The phylogenetic relationships among the haplotypes were first examined using CaM locus sequences. The CaM sequences of up to 3 representatives (when available) from each of the haplotypes were aligned with homologous sequences from holotype/neotype strains of A. section Nigri species. The phylogenetic tree was constructed using a maximum likelihood method with the Kimura-2-parameter model [21] (previously calculated in MEGA7 and jModelTest2) with discrete Gamma distribution and invariant sites (G + I) with Subtree Pruning Regrafting (SPR level 5) and a very weak branch swap filter. Sequence identity was also confirmed using a GenBank Nucleotide Basic Local Alignment Search Tool (BLASTn) [22].

Sequencing Beta-Tubulin (BenA) and RNA Polymerase II (RPB2) Genes
To elucidate the taxonomic position of the haplotypes identified in the present study, two other loci, part of the beta-tubulin (BenA) and part of the RNA polymerase II (RPB2) genes, both considered alternative markers for A. section Nigri [5], were sequenced and used to compose the Genealogical Concordance Phylogenetic Species Recognition (GCPSR) analysis. Up to 10 strains (when available) were selected for each haplotype present in the Brazilian dataset. Unfortunately, the GCPSR step was not possible for haplotypes found exclusively in the GenBank dataset, as we did not have physical access to these strains, and there are no deposits on GenBank for all loci (CaM, BenA and RPB2) of representatives of these haplotypes (H19, H20, H22 and H23).
The primers used for the amplification of the BenA gene were those described by Glass and Donaldson [17], with amplification conditions shown in Table 1. The sequencing protocol was the same as that used by Taniwaki et al. [16].
The primer-pair used for the amplification of the RPB2 gene was designed in the present study. Primer design involved the recovery of RPB2 gene sequences from all species of the A. niger aggregate available in GenBank, the alignment of the sequences using MUSCLE in MEGA 7 and a visual search for regions conserved within and between species. The melting temperature, the formation of secondary structures and other parameters were analysed in silico using the Oligoanalyzer 3.1 (https://www.idtdna.com/calc/Analyzer/Home/Instructions). The primer sequences and the segment of the RPB2 gene ( 720 bp) are shown in Supplementary Figure S1. PCR amplification was performed according to the conditions shown in Table 1. All PCR products were purified and sequenced, in forward and reverse directions, on an ABI 3500 XL genetic analyser.

Genealogical Concordance Phylogenetic Species Recognition-GCPSR
Genealogical Concordance Phylogenetic Species Recognition (GCPSR) is a practical approach to Phylogenetic Species Recognition (PSR). The GCPSR protocol adopted in this study was that found in Dettman et al. [23]. Sequence data of three loci (CaM, BenA and RPB2) were analysed individually and in combination.
Initially, we recognised the independent evolutionary lineages (IELs) based on the criteria of genealogical concordance and genealogical non-discordance [23]. According to the criterion of genealogical concordance, a clade must be present in most of the phylogenies analysed, that is, when analysing the phylogeny of each gene tree separately, the evolutionary lineage must be consistent in most of them, regardless of support levels (Bootstrap Percentage, BP, or Posterior Probabilities, PP).
For the application of the genealogical concordance criterion, maximum parsimony (MP) and Bayesian Inference (BI) trees were constructed for each locus, individually. The alignments were performed using MUSCLE in MEGA 7. The maximum parsimony trees based on these sequence alignments were created using MEGA 7. The heuristic search option with 1000 bootstraps and Tree Bisection Reconnection (TBR) branch swapping were used to infer the most parsimonious trees, and a Majority-Rule Consensus (MRC 50%) was generated [24].
A Bayesian Inference (BI) analysis was conducted in MrBayes 3.2.3 [25]. The most suitable model of sequence evolution for each dataset was selected based on the lowest Akaike Information Criterion (AIC) value in jModeltest2 [26]. For BI, the Markov Chain Monte Carlo (MCMC) algorithm was run for 5 × 10 6 generations with a sample frequency of 100 and with 25% of trees removed for burn-in. Convergence diagnostics were monitored based on standard deviations of frequencies below 0.01. The consensus tree topology (MRC 50%) and the PP for the nodes were visualised using FigTree version 1.4.3 (http://tree.bio.ed.ac.uk/software/figtree/).
According to the criterion of genealogical non-discordance, a clade can be considered to be an IEL if it is statistically supported (BP and PP) in at least one phylogeny and is not contradicted at the same level of support by another locus. The identification of these clades is done through analysis of a semi-strict consensus tree combining the phylogenies of all analysed loci. Dettman et al. [23] stipulated an arbitrary value of statistical support as 70% for BP values and 0.95 for PP.
For the application of the criterion of genealogical non-discordance, semi-strict consensus trees (MP and BI) were constructed. The semi-strict consensus trees were constructed in the PAUP 4.0 [27] based on the phylogenies of the three loci analysed.
The framing on one of the two criteria (genealogical concordance/non-discordance) satisfies the condition for the recognition of an IEL. To determine an IEL as a phylogenetic species, we used the criteria of genetic differentiation and exhaustive subdivision [23].
For analysis of genetic differentiation and exhaustive subdivision, maximum parsimony and Bayesian trees were constructed with the combined data, as established in the Dettman protocol. Additionally, we also built a tree of maximum likelihood with the same combined dataset. The combined dataset was partitioned into three blocks (CaM, BenA and RPB2) in the Partitionfinder2 on XSEDE platform (v2.1.1) [28]. The best model for the CaM block (430 bp) was K80 + G (BI) and GTR + G + I (ML), that for the BenA block (453 bp) was K80 + G (BI) and K80 (ML) and that for the RPB2 block (625 bp) was GTR (BI) and K80 (ML). The programs used to construct the trees were MrBayes 3.2.3, RaxML v.8 on XSEDE [29] and PAUP 4.0.
In genetic differentiation, minor tip clades recognised as IELs must be relatively distinct and well-differentiated to avoid minor tip clades being recognised as phylogenetic species. In the exhaustive subdivision, all individuals must be assigned to a phylogenetic species. Therefore, if an individual was not included within an IEL, we collapsed this clade to the adjacent node, and this was done hierarchically until all the individuals were allocated within an IEL and the rankings possibilities are exhausted. The final groups were recognised as phylogenetic species.

Random Amplified Length Polymorphism (RAPD)
Random Amplified Polymorphic DNA (RAPD) analysis was applied to four phylogenetic species (7 strains representing each phylogenetic species). DNA amplifications were carried out using six arbitrary primers (OPAM02, OPAM07, OPAM10, OPAM11, OPAM13 and OPX11) in a Veriti ® Thermal Cycler (Applied Biosystems, Waltham, MA-USA). The conditions of the reactions were the same as those described in Fungaro et al. [30]. Amplified fragments were scored 1 for the presence and 0 for the absence in each strain. The program Ntsys-PC was used to obtain a similarity matrix using a Dice coefficient and a dendrogram with the Unweighted Pair Group Method with Arithmetic Mean (UPGMA). The allele frequencies of the sets of strains were estimated and submitted to pairwise ϕst analysis using Arlequin 3.5.2.2 [31].
Bayesian clustering was also used to assess genetic structuring of fungal groups using the program Structure 2.3.4 [32]. The number of clusters (K) was estimated using the admixture ancestral model, with K ranging from 1 to 8. Twenty independent runs of 10 5 Markov Chain Monte Carlo (MCMC) generations with 2 × 10 4 generations of "burn-in" were used for each value of K. To interpret the number of population clusters, ∆K [33], statistics built into the Structure harvester [34] virtual platform were used.

Morphological Analysis
The morphological analyses were conducted in accordance with the guidelines of Samson et al. [5]. For macromorphological observations, representatives from each phylogenetic species identified by GCPSR (PS1, PS2, PS3 and PS4) were inoculated in three points onto Czapek Yeast Autolysate agar (CYA), Malt Extract Agar (MEA), Creatine Sucrose agar (CREA) and Yeast Extract Sucrose Agar (YESA) and incubated for 7 days at 25 • C in the dark. In addition, for growth pattern examination, CYA plates were incubated at 37 and 42 • C. The experiments were performed in triplicate.
For scanning electron microscopy (SEM), 0.5 × 0.5 cm plugs (MEA, 7 days at 25 • C) were fixed in 2% glutaraldehyde solution at 4 • C, for 24 h, and later washed in 0.1 M sodium phosphate buffer, (three times, for 15 min). Subsequently, the plugs were post-fixed with 1% osmium tetroxide (aqueous solution) at room temperature for 2 h at 25 • C in the dark. After washing in 0.1 M sodium phosphate buffer, the fragments were dehydrated with ethanol at 70%, 80%, 90% and 100% for 10 min at each step. The samples were transferred to a critical-point dryer (Bal-Tec, CSDC 030) and gold-coated on a vaporiser (Bal-Tec, SDC050). The images were obtained with a Quanta 200 ® (Field Electron and Ion Company-FEI, Hillsboro, OR, USA) Scanning Microscope.

Secondary Metabolites Characterisation of PS3 Strains
Ten PS3 strains were analysed for secondary metabolites by extracting 3 agar plugs from 7 days incubated cultures on CYA and YESA at an incubation temperature of 25 • C in darkness. The extracts were analysed by ultra-high performance liquid chromatography with diode array detection (UHPLC-DAD) and an acetonitrile/water gradient (both eluents containing 0.05% trifluoroacetic acid) solution as mobile phase and compared to authentic small-molecule extrolite standards as previously described by Frisvad and Thrane [35] and Nielsen et al. [36].

Secondary Metabolites Characterisation of Sclerotia in PS3 Strains
For secondary metabolites characterisation of sclerotia, the strain ITAL 47.456 was grown on MEA for 10 days at an incubation temperature of 25 • C in darkness. Sclerotia were recovered using the methodology described in Carvajal-Campos et al. [37]. Secondary metabolites were screened by liquid chromatography coupled to high-resolution mass spectrometry (HPLC-HRMS). The chromatographic system consisted of a RSLC3000 UHPLC (Thermo Scientific, Les Ulis, France) operating with a 5 µm Luna ® C18 column (125 × 2.0 mm Luna C18 5 µm column) (Thermo Scientific, Les Ulis, France). A gradient program of (A) water acidified with 0.05% formic acid and (B) acetonitrile was used at 30 • C and a flow rate of 0.2 mL min −1 as follows: 0 min 20% B, 30 min 50% B, 35 min 90% B, from 35 to 45 min 90% B, 50 min 20% B, from 50 to 60 min 20% B. A volume of 10 µL of each sample diluted twice with mobile phase A was injected. The mass spectrometer corresponded to an LTQ Orbitrap XL (Thermo Scientific, Les Ulis, France) fitted with an Electrospray Ionisation Source (ESI). ESI parameters for the negative mode were set as follows: spray voltage: 3.7 kV, sheath gas flow rate (N 2 ): 30 arbitrary units (a.u.), auxiliary gas flow rate (N 2 ): 10 a.u., capillary temperature: 350 • C, capillary voltage: −34 V and tube lens offset: −180 V. ESI parameters for the positive mode were set as follows: spray voltage: 5.5 kV, sheath gas flow rate (N 2 ): 30 arbitrary units (a.u.), auxiliary gas flow rate (N 2 ): 10 a.u., capillary temperature: 350 • C, capillary voltage: 1 V and tube lens offset: 45 V. High-resolution mass spectra were acquired between m/z 80 and 800 at a resolution of 7500. The calibration of the mass spectrometer was achieved using the calibration solution of Thermo Scientific in agreement with their protocol. MS/MS spectra were obtained with the collision-induced dissociation (CID) mode of the ion trap analyser at low resolution and normalised collision energy of 35%.

Diversity and Phylogenetic Relationships of CaM Haplotypes
The 418-character matrix of the aligned CaM sequences generated from the Brazilian Dataset of CaM Sequences (n = 403) plus the GenBank Dataset of CaM Sequences (n = 292) enabled the detection of 47 variable sites with 18 parsimony informative sites, yielding 31 CaM haplotypes (Supplementary  Table S1). Of the total number of haplotypes detected, 93.5% (29/31) were generated from Single Nucleotide Polymorphisms (SNP's) and only 6.5% (2/31) were generated from Insertion or Deletion variants (INDEL's).
By taking the 695 sequences as a single dataset, the overall haplotype diversity (Hd) and nucleotide diversity (π) were 0.654 ± 0.013 and 0.00990 ± 0.00100, respectively. When the Brazilian dataset (n = 403) and GenBank dataset (n = 292) sequences were analysed separately, higher diversity was noted in the GenBank dataset sequences (Hd = 0.686 ± 0.022; π = 0.00974 ± 0.00026) compared to the Brazilian dataset (Hd = 0.524 ± 0.025; π = 0.00762 ± 0.00044). The greater genetic diversity within GenBank dataset sequences is not surprising because GenBank receives sequences from strains isolated from different countries throughout the world and from a higher range of substrates than analysed here. The genetic relationships among haplotypes identified in the present study were inferred by phylogenetic analysis of the CaM nucleotide sequences. Most of the haplotypes are on well-supported branches with the neotype strain of A. welwitschiae (NRRL 4948) (98% bootstrap value) or with the type strain of A. niger (NRRL 326) (90% bootstrap value) and were confidently identified as A. welwitschiae (n = 403) or A. niger (n = 248), respectively ( Figure 1). The exceptions to this identification were six haplotypes (H19, H20, H21, H22, H23, H30), representing 45 strains, which were not resolved on well-supported branches in the A. niger or A welwitschiae clades ( Figure 1). Thus, the phylogenetic analyses using only CaM Among the six haplotypes unresolved, only two of them (H30 and H21) occurred in the Brazilian dataset, and it was possible to investigate them more deeply to determine whether they might represent new phylogenetic species.

Genealogical Concordance Phylogenetic Species Recognition (GCPSR)
Genealogical Concordance Phylogenetic Species Recognition (GCPSR) is a term originally suggested by Taylor et al. [39] which reflects the concept of genealogical concordance established in Avise and Ball [40], later operationalised by Dettman et al. [23]. Briefly, it is an analysis of multiple loci realised by the observation of the phylogenies and the statistical support of the clades in two scenarios: individual and consensus genealogies.
For the GCPSR implementation, nucleotide sequences of 3 loci (CaM, BenA and RPB2) were used in our study. A total of 56 strains isolated from Brazilian foodstuffs were selected according to the analysis of haplotypes based on the CaM locus (secondary barcode for phylogeny of A. section Nigri) and additionally sequenced for parts of the RPB2 and BenA genes. For GCPSR analysis, the alignment of the CaM locus resulted in a fragment of 430 bp (Hd = 0.863 ± 0.015; π = 0.01076 ± 0.00043), for the BenA locus, the final alignment resulted in 453 bp (Hd = 0.807 ± 0.024; π = 0.00395 ± 0.00024), and for the RPB2 locus, the final alignment resulted in 625 bp (Hd = 0.303 ± 0.067; π = 0.00340 ± 0.00075). As can be observed (Supplementary Figure S2), the CaM locus has more resolution power for phylogeny of the A. niger clade in relation to the other loci analysed. The combined data (CaM + BenA + RPB2) totalled 1508 bp (Hd = 0.880 ± 0.017; π = 0.00563 ± 0.00034).
Based on the criteria of genealogical concordance (Supplementary Figure S2) and genealogical non-discordance (Supplementary Figure S3), 8 independent evolutionary lineages (IELs) were recognised. IEL 1 and 2 were identified by the genealogical concordance criterion, and the remaining IELs were identified by the genealogical non-discordance criterion. As mentioned in the "Material and Methods Section" to recognise an IEL as a phylogenetic species (PS), the criteria of genetic differentiation and exhaustive subdivision [23] were used. The minor tip clades IEL 6 and 7 were excluded from the ranking process according to the genetic differentiation criterion.
The limits of the phylogenetic species were then established by means of the exhaustive subdivision criterion. Based on this criterion, IELs 3 and 8 were collapsed, and all individuals were harboured under IEL 1; therefore, this node was recognised as the limit of phylogenetic species 1 (PS1). A group of seven strains representing the IEL 4 formed the phylogenetic species 2 (PS2). This phylogenetic species was well-supported statistically in the combined three-locus datasets ( Figure 2) and CaM locus analyses (Supplementary Figure S2).
The IEL 2 harboured 10 strains, and following the ranking process, it was identified as phylogenetic species 3 (PS3). This phylogenetic species was well-supported statistically in all analyses in the individual and/or combined genealogies (Figure 2 and Supplementary Figures S2 and S3).
Under the exhaustive subdivision criterion, the node of IEL 5, which was well-supported in the combined analysis (three-locus dataset), was recognised as a phylogenetic species boundary and designated as phylogenetic species 4 (PS4). Fifteen strains examined in this study were allocated to this phylogenetic species (Figure 2).
The strains belonging to the PS1 and PS4 were those identified as A. welwitschiae (H3, H8, H25, H27) and A. niger (H14 and H15), respectively. On the basis of our data, A. welwitschiae neotypes strains (NRRL 4948 and ITEM 4509) (A. awamori sensu Perrone) [41] are assigned to H8, whereas the exotype strain of A. niger (NRRL 326) is assigned to H14. It is worth mentioning that the type strains of A. lacticoffeatus (CBS 101883) and A. foetidus (NRRL 341) are also contained in Haplotype H14 (Supplementary Table S1), these species were previously synonymised to A. niger stricto sensu.
The PS2 (H21) and PS3 (H30) were not collapsed to A. niger and A. welwitschiae, thereby becoming candidates for new species. The IEL 2 harboured 10 strains, and following the ranking process, it was identified as phylogenetic species 3 (PS3). This phylogenetic species was well-supported statistically in all analyses in the individual and/or combined genealogies (Figure 2 and Supplementary Figures S2 and  S3).
Under the exhaustive subdivision criterion, the node of IEL 5, which was well-supported in the combined analysis (three-locus dataset), was recognised as a phylogenetic species boundary and designated as phylogenetic species 4 (PS4). Fifteen strains examined in this study were allocated to this phylogenetic species (Figure 2).
The strains belonging to the PS1 and PS4 were those identified as A. welwitschiae (H3, H8, H25, H27) and A. niger (H14 and H15), respectively. On the basis of our data, A. welwitschiae neotypes Phylogenetic species 2 (PS2) was recognised by the sequence data collected from 7 Brazilian strains isolated from coffee beans, grape berries and onions bulbs. Based on CaM sequence, these seven strains (2.7, 2.6, 2.8, 47.514, 148.727, 102.2706, 13.09), as well as two clinical isolates (F8478 and F12140) characterised by Howard et al. [2] in their study of cryptic species in the Aspergillus niger complex, represent the haplotype H21 (Supplementary Table S1). There are five SNPs fixed for PS2 in relation to A. niger and four in relation to A. welwitschiae. Interestingly, among several A. niger and A. welwitschiae strains tested by Howard et al. [2], only the strain F12140 was highly cross-resistant to all azole drugs. Combining calmodulin and beta-tubulin sequencing, these authors observed that the F12140 isolate is most closely related to A. awamori (currently A. welwitschiae); however, by the cyp51A sequence, the isolate was most similar to A. niger, although bootstrap values were low (<70) for all analyses. Using cyp51A sequences, the strain F12140 appears to be even more divergent from A. awamori and A. niger [2]. It is also important to note that the CaM sequence of the strain denoted as ITEM 1148 (see Susca et al. [38]) (sequence not available in GenBank), recovered from Brazilian cashew nuts, is identical to those from our seven Brazilian strains recognised as PS2 and that the identity of this strain was considered ambiguous by Susca et al. [38].
PS3 (=Aspergillus vinaceus sp. nov.), represented by haplotype H30 (Supplementary Table S1), was exclusive to the Brazilian dataset on 12 July 2018. All strains were isolated from grape samples grown in the São Francisco Valley, Northeastern Brazil (Supplementary Table S1). However, on 22 December 2018, one identical CaM sequence (strain IHEM 18080, accession number MH644969) was deposited in GenBank by D'hooge et al. [42]. According to these authors, the position of this strain relative to A. welwitschiae and A. niger is not resolved, and they decided to provisionally name them A. aff. welwitschiae. Based on a concatenated tree (ITS, BenA and CaM), this strain was positioned next to A. welwitschiae with low support.
The concept of genealogical concordance has been widely used to recognise species boundaries, and has supported the description of new species [43][44][45][46][47][48][49][50]. As an example, recently, Frisvad et al. [51] revised the taxonomy of the Aspergillus section Flavi and with the help of GCPSR described eight new species. The GCPSR allows us to determine the limit of a species through a protocol based on the concordance/conflict of the different genealogies, thus reducing the subjectivity in recognising these limits. However, as pointed in Rintoul et al. [52], the limits of the species must be interpreted as explicit taxonomic hypotheses and must be tested with additional data, and/or based on other concepts.

RAPD Analysis
A total of 92 reproducible RAPD bands were obtained from six random primers. The dendrogram generated by UPGMA cluster analysis using the Dice similarity coefficient was congruent to the results obtained using sequence data. Four well-defined groups were found, with one of them being composed of A. niger strains (GI), another being composed of A. welwitschiae strains (GII) and the two other groups formed by strains here recognised as belonging to PS2 (GIV) and PS3 (GIII) (Figure 3).
Several studies have used "F" statistic and its analogs as indicative of the degree of genetic structuring/differentiation between several groups of fungi [53][54][55][56][57][58][59][60]. In our study, the pairwise ϕst analysis corroborated the GCPSR findings. The values of ϕst for the comparisons among the four phylogenetic species ranged from 0.63 to 0.89 (Figure 3), values that can be interpreted as a very large differentiation [61]. Massi et al. [62] used the value of Fst as support to discriminate Aspergillus nomius from Aspergillus pseudonomius, two closely related species. According to the authors, values of Fst above 0.73 are uncommon for strains of the same species.
In the Bayesian clustering analysis, it was observed that the ideal number of clusters to allocate all strains was 4 (the best K value according to the model described by Evanno et al. [33]). As expected, the strains of the putative new species (PS2 and PS3) were designated as a group different from the groups of A. niger and A. welwitschiae, which are represented by a colour-coded bar graph shown in Figure 3. These findings corroborated with UPGMA analysis and with the GCPSR results.
The genotypic data (GCPSR and RAPD) obtained in the present study allows the validation of PS2 and PS3 as new phylogenetic species pertaining to the A. niger clade. Molecular approaches provide the largest number of variable characters for fungal taxonomy; currently, many phylogenetic species are being recognised. However, according to the International Code of Nomenclature for algae, fungi and plants, for the description of new species, differential phenotypic characters must be presented. Therefore, we decided to investigate the phenotypic traits of phylogenetic species 2 and 3. Several studies have used "F" statistic and its analogs as indicative of the degree of genetic structuring/differentiation between several groups of fungi [53][54][55][56][57][58][59][60]. In our study, the pairwise φst analysis corroborated the GCPSR findings. The values of φst for the comparisons among the four phylogenetic species ranged from 0.63 to 0.89 (Figure 3), values that can be interpreted as a very large differentiation [61]. Massi et al. [62] used the value of Fst as support to discriminate Aspergillus nomius from Aspergillus pseudonomius, two closely related species. According to the authors, values of Fst above 0.73 are uncommon for strains of the same species.
In the Bayesian clustering analysis, it was observed that the ideal number of clusters to allocate all strains was 4 (the best K value according to the model described by Evanno et al. [33]). As expected, the strains of the putative new species (PS2 and PS3) were designated as a group different from the groups of A. niger and A. welwitschiae, which are represented by a colour-coded bar graph shown in Figure 3. These findings corroborated with UPGMA analysis and with the GCPSR results.
The genotypic data (GCPSR and RAPD) obtained in the present study allows the validation of PS2 and PS3 as new phylogenetic species pertaining to the A. niger clade. Molecular approaches provide the largest number of variable characters for fungal taxonomy; currently, many phylogenetic species are being recognised. However, according to the International Code of Nomenclature for algae, fungi and plants, for the description of new species, differential phenotypic characters must be presented. Therefore, we decided to investigate the phenotypic traits of phylogenetic species 2 and 3.

Taxonomy
A morphological analysis was carried out to establish a comparison between phylogenetic species found in GCPSR. No morphological differences were found among strains representing PS2, PS4 (A. niger) and PS1 (A. welwitschiae), however, the PS3 strains differ in their morphology when compared to any of the other phylogenetic species analysed here. The PS3 strains show reduced growth on MEA medium in relation to PS2, A. niger and A. welwitschiae strains ( Table 2). The colonies of PS3

Taxonomy
A morphological analysis was carried out to establish a comparison between phylogenetic species found in GCPSR. No morphological differences were found among strains representing PS2, PS4 (A. niger) and PS1 (A. welwitschiae), however, the PS3 strains differ in their morphology when compared to any of the other phylogenetic species analysed here. The PS3 strains show reduced growth on MEA medium in relation to PS2, A. niger and A. welwitschiae strains ( Table 2). The colonies of PS3 strains grown on CYA (25 • C) show white mycelial area staining with very sparse sporulation. PS3 strains have conidiophores (169 ± 23.1 µm), stipes width (18.3 ± 2.1 µm) and vesicles (68.2 ± 8.8 µm) of notably larger size in relation to the other species of the A. niger clade ( Table 2). The conidia (4.6 ± 0.6 µm) varied from finely roughened to echinulate (Figure 4), while A. niger and A. welwitschiae conidia varied from smooth to rough. PS3 strains produce sclerotia, with ovoid shape, colour varying from white to cream, measuring on average 1200 ± 300 µm ( Figure 4). Sclerotial development is an important prerequisite for sexual development in Aspergillus subgenus Circumdati, but only some species of Aspergillus section Nigri have been reported to produce sclerotia on well-known growth media. As shown by Frisvad et al. [63], no strains of A. niger sensu stricto produced sclerotia on conventional media. In certain strains of this species, sclerotium production may occur, but only when whole-raising or other plant parts are added to CYA medium (CYAR). A. welwitschiae did not produce sclerotia under any known conditions. from white to cream, measuring on average 1200 ± 300 µm ( Figure 4). Sclerotial development is an important prerequisite for sexual development in Aspergillus subgenus Circumdati, but only some species of Aspergillus section Nigri have been reported to produce sclerotia on well-known growth media. As shown by Frisvad et al. [63], no strains of A. niger sensu stricto produced sclerotia on conventional media. In certain strains of this species, sclerotium production may occur, but only when whole-raising or other plant parts are added to CYA medium (CYAR). A. welwitschiae did not produce sclerotia under any known conditions.  To conclude, the morphology data did not support the separation of PS2 from A. niger and A. welwitschiae; however, it is well-known that genetic changes can be observed before the fixation of the morphological and behavioural changes. In contrast, the PS3 can be diagnosed based on morphological characters and this phylogenetic species can also be framed in the morphological species concept.
In addition to morphological and DNA sequence features, it is strongly recommended to use secondary metabolites in species descriptions due to their high species specificity [64]. Secondary metabolite profiling has been used quite extensively for taxonomic purposes in Aspergillus section Nigri. Following the taxonomic schemes proposed by Samson et al. [5], we examined some strains (n = 10) of PS3 for secondary metabolite production on CYA and YESA by ultra-high-performance liquid chromatography using diode array detection (UHPLC-DAD). The strains produced acetyl-leucomelone, asperazine, funalenone, nigragillin, pyranonigrin A, tensidol B, malformins and very large amounts + --* Grown on CYA and YESA media. ** Sclerotia produced on MEA medium.
Notes: Aspergillus vinaceus is molecularly related to A. welwitschiae and A. niger, but clearly morphologically distinguishable. In CYA, colonies of A. vinaceus have a white prominent mycelial area. In MEA, the growth of A. vinaceus is significantly lower than the other species of the clade A. niger. In addition, it produces very large (900-1500 µm) white to cream colour sclerotia on MEA, CYA and YESA media, which do not occur in strains of A. niger and A. welwitschiae. Microscopically, the conidiophore, stipe width and vesicle are substantially larger than in the other closely related species. The conidia are echinulate rather than with black bars, as in A. niger and A. welwitschiae. A. vinaceus produces a spectrum of secondary metabolites which is different from A. niger and A. welwitschiae (Table 3).
Supplementary Materials: The following are available online at http://www.mdpi.com/2309-608X/6/4/371/s1: Table S1: Listing of the strains and distribution of haplotypes found in this study. Table S2: Secondary metabolites data of Aspergillus vinaceus sp. nov. sclerotia, produced in MEA medium, detected by liquid chromatography coupled to high-resolution mass spectrometry (HPLC-HRMS). Figure S1: Alignment of partial RPB2 gene sequences of A. niger aggregate. The boxes marked by arrows indicate the annealing site of the ANRP primer pair. * ANRP1-F or ANRP2-R designed in this study. Figure S2: Maximum parsimony tree (MR 50%) of A. niger clade species using the loci CaM, BenA and RPB2. Genealogical Concordance criterion: The lineages that occur in most of the single loci are highlighted in coloured boxes, and these lineages were identified as independent evolutionary lineages. Bootstrap values (BP) and/or posterior probabilities values (PP) higher than 70% and 0.70 respectively, are shown. A. carbonarius is the outgroup. Figure S3: Maximum parsimony tree (semi-strict consensus) of A. niger clade species using three-locus (CaM, BenA and RPB2) phylogenies. Genealogical non-discordance criterion: The lineages that were well-supported by at least one locus but not contradicted by any other locus are identified as independent evolutionary lineages (coloured boxes). Bootstrap values (BP and/or posterior probabilities values (pp) higher than 70% and 0.70 respectively, are shown. A. carbonarius is the outgroup. Figure S4: UV spectrum of metabolite "HUTI" (RT: retention time, RI: retention index, log A: logarithm of the area of the peak in the chromatogram). Figure S5: UV spectrum of metabolite "SURI" (RT: retention time, RI: retention index, log A: logarithm of the area of the peak in the chromatogram).