INTRODUCTIONBarley (Hordeum vulgare L.) is one of the majorcereal crops. It occupies fourth place among cerealsin the world and second place in Russia by productionquantity and cultivation area [1]. The importance ofbarley has been accepted since ancient times and usedin the food, feed and brewing industries due to itsversatility, excellent adaptation capabilities and superiorproperties [2].The increased interest in barley as a source of foodand fodder has resulted in a huge number of studies ofassociated microorganisms. It is known that barley canbe infected with a broad variety of plant-pathogenic andtoxigenic fungi, many of which may persist in grains.The Bipolaris, Pyrenophora, Phaeosphaeria, Alternaria,and Fusarium genera are considered to be prevailingfungi in barley grain worldwide [3, 4]. Species of the lasttwo genera are well known as mycotoxin producers, withFusarium spp. being the most dangerous food and feedcontaminants.Cultivation-based methods have traditionallyestablished modern vision on geographical andecological distribution and biodiversity of micromycetes.These methods cannot provide accurate data on taxoncomposition because some microorganisms do not havespecific characteristics to be identified, and some appearto be noncultivable. Thus, data on the mycobiome ofmany substrates, including barley grain, is likely to beincomplete.In recent decades, molecular methods have shiftedmicrobiological research to a new level, making itpossible to investigate hidden taxonomical biodiversity.Next-generation sequencing (NGS), implementedon various independent technical platforms, becamethe most promising method for conducting researchprojects aimed at revealing fungal or bacterialcomposition [5–7]. Several studies have focused on avariety of agricultural subjects [8–12]. Advances in thisfield led to consideration that NGS-based methods aresuitable as incipient techniques for seed testing [13].In Denmark, 454 pyrosequencing of theinternal transcribed spacer 1 (ITS1) of the nuclearribosomal DNA (rDNA) has been chosen to recoverthe composition of fungal communities associatedwith wheat grain [14]. NGS revealed a significantlyhigher level of biodiversity than it was observedin previous culturing studies. Another appropriate454 pyrosequencing of both ITS regions was doneto study the mycobiome of barley grain in westernCanada [3]. It demonstrated that geographic locationand agronomical practices were the determining factorsexplaining the observed differences in the fungalcommunities associated with barley. Such studies maycontribute to a better understanding of fungal speciescompositions in cereals. They may also lead to moreaccurate food-quality testing and the precise design ofcrop protection strategies that would reduce the level offungal contamination of agricultural products.The objective of this study was to revise thetaxonomical variety of fungi contaminated the surfaceand infected barley grains harvested in the northwesternregion of Russia. We hypothesized that grain mycobiomecould be significantly differ on the surface and insidethe grain, and the difference may depend on the cropyear. In our research we used the traditional agar platebasedmethod and the contemporary method based on454 pyrosequencing of ITS2.STUDY OBJECTS AND METHODSSampling. Grain samples of five spring barley cultivars(Suzdalets, Krinichnyy, Moskovskiy 86, Tatum,and Belgorodskiy 110) were received in 2014 and2015 from the State Experimental Station (Volosovo,Leningradskaya oblast, Russia, 59°31’N, 29°28’E).Small grain cereals on this station were cultivated withno fungicide treatments. The grain samples intendedfor fungi isolation and DNA extraction were storedseparately at 4°C and –20°C respectively.Extraction of DNA. Two representative subsamplesconsisting of 500 grains were picked from each sample.One subsample was placed in a 50 mL polypropylenetube and subjected to superficial sterilization. The grainswere consistently washed with 20 mL of 2% sodiumhypochlorite solution containing 0.1% of sodium dodecylsulphate (SDS). They were then washed once with 5%sodium hypochlorite, two times with deionized water(diH2O), and finally rinsed with 98% ethanol. Witheach step the mixtures were actively stirred up within2 min, and the flushing solution was then decantedwithout the grain. The ethanol was removed by burningat regular stirring during 10 s. After this, the grainswere homogenized in sterile disposable chambers on aTube Mill Control (IKA, Germany) grinder. The othersubsample was similarly homogenized, but the step ofsuperficial sterilization was skipped.Further, 240 mg of the ground subsamples weretransferred to a 2 mL Eppendorf tube, where DNAwas extracted with an AxyPrep Multisource GenomicDNA Miniprep Kit (Axygen, USA), according tothe centrifugal protocol for plant tissues and fungalmycelium. The DNA concentrations were measuredwith a Qubit 2.0 (Thermo Fisher Scientific, USA)using a dsDNA HS Assay Kit. Extracted DNA wasused for library preparation and subsequent universaltailed amplicon sequencing, as described for the454 Sequencing System.Pyrosequencing and primary data analyses.For amplicon library preparation we chose thetaxonomically significant ITS2 region, whichis commonly used, as well as ITS1, in DNAmetabarcoding studies of fungal diversity. To alarge extent, ITS1 and ITS2 have similar resultswhen used as DNA metabarcodes for fungi [15–17].However, the ITS2 region lacks the insertions commonlyfound in ITS1 and thus reduces length variation [18].This is important, as length variation can biascommunity pyrosequencing toward shorter amplicons.Also, ITS2 is the best-represented fungal genomicelement in the public databases [19, 20]. Therefore, instudies similar to our project, use of ITS1 obtained withfungal specific primer (ITS1F) [21] can be helpful ineliminating plant ITS amplification, and may turn outto be the method of choice in cases of mixed plant andfungal genomic DNA. However, it is necessary to takeinto account that ITS1F (with constricted, specific rangetoward exclusion of all eukarya except fungal taxa), maynot be able to amplify several fungal taxa because it ishampered with a high degree of mismatches relative tothe target sequences [22, 23].The ITS2 region was amplified witheukaryote-specific ITS3 and ITS4 primers(ITS3: GCATCGATGAAGAACGCAGC; ITS4:TCCTCCGCTTATTGATATGC) [22, 24]. Multiplexidentifiers (MIDs) were attached to the primers’ ends tocarry out in consequence the simultaneous analysis of allsamples.The amplicon library pool was sequenced with454 pyrosequencing on the GS Junior sequencer(Roche, USA) according to the recommendations ofthe manufacturer [25]. The ITS2 locus reads wereprocessed by QIIME Version 1.6.0 (Quantitative Insightsinto Microbial Ecology) [26]. To reduce the amountof erroneous sequences and thus increase the accuracy of the whole pipeline, the denoising procedure wasemployed [27].Next steps included assigning multiplexed reads tosamples based on their specific MIDs (demultiplexing),removing the low-quality or ambiguous reads,truncating primers, and other accessorial sequences.Chimeric sequences were detected using the UCHIMEalgorithm with the Unite database [28–30]. All of thereads were clustered into operational taxonomic units(OTUs) at 97% sequence similarity using the UCLUSTmethod [31]. Representative sequences were chosenaccording to their abundance between similar reads.Low-abundance OTUs, which have less than four copies(singletons, doubletons and tripletons), were deletedover all of the analysis [32]. All 454 pyrosequencingdata of the present investigation are available throughthe Sequence Read Archive (SRA) under BioProjectPRJNA353503, with run accession numbers fromSRR5022991 to SRR5023010 [33].Phylogenetic and statistical analyses. Taxonomicalidentification of representative sequences was carriedout by the BLAST method using Genbank databases [34,35]. Query coverage ≥ 99% was recognized assignificant. Query identity of ≥ 99% was consideredidentification at the species level; identity of ≥ 98–95%was considered reliable identification at the genus level.The smaller similarity Ribosomal Database Projectclassifier, along with the Unite database (minimumconfidence at 0.9), were implemented to assign OTUs toa higher taxonomic rank [29, 30, 36, 37].Alignment of representative sequences wascarried out using MAFFT algorithm G-INS-1 [38].A phylogenetic tree was conducted with MEGA5 usingthe Maximum Likelihood method, based on the Tamura-Nei model with 1000 bootstrap replicates [39–41].Vegdist and hclust R functions were used forcomputing Bray-Curtis dissimilarity indices andUPGMA hierarchical clustering of OTUs, showing theircoexistence in the samples [42]. Heatmaps were generatedwith QIIME 1.8.0 with log-transformed abundance data.OTUs were sorted by phylogenetic or hierarchical trees.Figure 1 Maximum likelihood consensus tree and fungal taxa heat map of sterilized and non-sterilized grain samples,which were harvested in 2014 and 2015 years. Read counts of each OTU were weighted according to sum of readsin the sample and log-transformed. White corresponds to low and blue to high number of reads Kri – Krinichnyy,Suz – Suzdalets, B110 – Belgorodskiy 110, Tat – Tatum, and M86 – Moskovskiy 86Beta diversity between samples wascalculated by beta_diversity.py script in QIIME withunweighted UniFrac metric [26, 43]. To check therobustness of estimated beta diversity, jackknifedanalysis, with 96 reads per sample depth and100 replicates, was performed. The results werevisualized with Principal coordinate analysis (PCoA) ina 2D scale plot.Figure 2 Boxplots depicting relative abundance (%) of the taxonomic ranks in sterilized and non-sterilized grain samplesharvested in 2014 and 2015. All taxonomical ranks are marked by different colors. The boxplots consist of square boundariesindicating the 25th and 75th percentiles. Whiskers indicate the 10th and 90th percentiles, and the line inside the box representsthe median. Outliers are not displayedBray-Curtis, weighted and unweighted UniFracdissimilarity indices were used for measuring thestrength and significance of sample groupings withPermutational Multivariate Analysis of Variance(PERMANOVA) and Analysis of Similarity (ANOSIM)with script compare_categories.py [26].Agar plate-based method of isolation andidentification of fungi. Representative subsamples (200grains) of each cultivar were surface-sterilized by beingshaken for 2 min in 5% sodium hypochlorite. Thenthey were rinsed twice in sterilized water. The surfacesterilizedgrains were placed on 90 mm Petri dishes (10grains per dish) with potato-sucrose agar (PSA) andincubated at 24°C for 10–14 days. The isolated fungalcolonies from every grain were identified by visual andmicroscopic observations according to Ellis, Gerlachand Nirenberg, Lawrence, Rotondo, and Gannibal,and Samson et al. [44–47]. To present data comparableto those obtained with NGS, relative abundance wascalculated as the number of all isolates of a certain taxondivided by the total number of fungal isolates (%). Amore conventional index of seed expertise, infectionfrequency, was calculated as the number of grainsinfected by the fungus (%).RESULTS AND DISCUSSIONNGS-based identification of fungi. Afterquality filtering and removal of nonfungal andchimeric sequences, in total, 8484 fungal readswere obtained and clustered into 43 operationaltaxonomic units (OTUs). The number of observedOTUs in the grain samples was ranged from 10to 27. The estimated OTU richness was higher onthe s urface ( Chao1 = 2 4.3 ± 4 .4, A CE = 2 6.2 ± 4 .9(2014); C hao1 = 3 0.7 ± 4 .2, A CE = 2 9.7 ± 1 .8 ( 2015)),than inside (Chao1 = 15.1 ± 1.5, ACE = 16.8 ± 1.9 (2014);Chao1 = 19.3 ± 2.0, ACE = 21.7 ± 1.7 (2015)) the barleygrains. The rarefaction curves pointed that the diversityof some samples might be underestimated, although allrarefaction curves were beyond the linear ranges.All of the OTUs were assigned to Basidiomycota andAscomycota phyla (Fig. 1).Primary Ascomycota prevail over Basidiomycota,but in non-sterilized grain, the ascomycetous readprevalence (percentage of reads) was significantlylower (Fig. 2). The 26 OTUs from Ascomycotabelonged to the families Nectriaceae, Dothioraceae,Microdochiaceae, Cladosporiaceae, Pleosporaceae,Sclerotiniaceae, Didymellaceae, Phaeosphaeriaceae,and unidentified Dothideomycetes, as well asundefined groups within Helotiales and Hypocreales.Seventeen basidiomycete OTUs belonged to yeastlikefungi from Entylomataceae, Cystofilobasidiaceae,and other undefined groups within Tremellales,Cystofilobasidiales, and Sporidiobolales.One of the most abundant OTU, otu166, assignedas Dothideomycetes, failed to be identified to a moreprecise taxonomical level. It has similar characteristics(BLASTn 100% query coverage and identity) withseveral GenBank sequences (e.g., EU552134, AJ279448,HG935454) which can be joined together only at therank of class. More likely, it coincides with Epicoccumnigrum, the one abundant Dothideomycete identifiedduring mycological analysis.Eleven minor OTUs (Alternaria related otu44, otu53,otu61, otu180, and otu190; Cryptococcus related otu29,otu60, otu218, and otu264; Davidiella related otu123;and Fusarium related otu89) had no close relation to anyknown species but appeared in the same samples wherea major OTU of a certain genus was abundant. However,potentially such satellite OTUs represent rare and/orpoorly studied species, but most likely they are technicalerrors or sequence variances, which can occur despite allfiltering and trimming procedures.From seven clustered OTUs that were assigned asAlternaria, the most abundant OTUs, otu18 and otu304,can refer to Alternaria and Pseudoalternaria sectionsrespectively (Fig. 3) [46, 48].From ITS2 sequences combined into six OTUs anddesignated as Fusarium, several OTUs can be readilyassigned as two synapomorphic (Fig. 4) clades, similarto that described by Watanabe et al. [49]. Two OTUs(F. poae – otu224 and Fusarium sp. – otu259) wereabundant, but four OTUs appeared as solitary sequences.Distribution of Fusarium-related OTUs amongclusters corresponded to the prevalent toxic secondarymetabolite production. The first cluster consistedof Fusarium fungi that are able to produce thetrichothecene group metabolites. The subcluster 1aincluded F. sporotrichioides and F. langsethiae, whichare the main producers of type A trichothecenes(like T-2 and HT-2 toxins); the subcluster 1b includedF. culmorum and F. graminearum the producers oftype B trichothecenes (like DON, or NIV). SpeciesF. poae (subcluster 1c) produces trichothecenes of typesFigure 3 Maximum likelihood tree of OTUs related to theAlternaria genus. Bootstrap values based on 1000 replicates.Bootstrap values less than 50% are not presentedFigure 4 Maximum likelihood tree of OTUs related to theFusarium genus. Bootstrap values based on 1000 replicates.Bootstrap values less than 50% are not presentedA and B, and enniatins (ENNs). Fungus F. equiseti(subcluster 1d) is able to produce ENNs, but accordingto some authors, it can also produce a small amountof type A trichothecenes [50, 51]. The subcluster2 brought together Fusarium fungi that are able tosynthesize ENNs: F. tricinctum, F. avenaceum, andF. lateritium [52].Fourteen OTUs that were defined to the specieslevel belonged to Bipolaris sorokiniana, Fusariumpoae, Neoascochyta exitialis, Sarocladium strictum,Cystofilobasidium macerans, Udeniomyces pannonicus,Cryptococcus victoriae, Cryptococcustephrensis, Cryptococcus wieringae, Sporobolomycesruberrimus, Sporobolomyces roseus, Dioszegiahungarica, Aureobasidium pullulans, and Tilletiopsiswashingtonensis.In general, the mycobiome of nonsterilized barleygrains was characterized by a greater abundance ofBasidiomycetes in comparison with surface-sterilizedgrains. The most abundant fungi in nonsterilized grainswere Davidiella (Cladosporium) spp. and Cryptoccocusspp. After surface sterilization, the average abundanceof Fusarium, Alternaria, Pyrenophora, andPhaeosphaeria, as well as fungi from Dothideomycetes,increased, but the ratio of those taxa dependedon the year.Comparison of taxonomical structure and relativeabundance between groups of samples combined bycrop year (2014/2015) and type of treatment (sterilized/non-sterilized) reflected significant distinctions inboth cases (Fig. 5). Nevertheless, distinctions betweensterilized and non-sterilized grain mycobiota (ANOSIMR = 0.64, 0.76, 0.69; P = 0.001, 0.001, 0.001;PERMANOVA pseudo F = 9.89, 17.7, 10.93;P = 0.001, 0.001, 0.001; data shown successively forBray-Curtis, Weighted and Unweighted UniFraccommunity dissimilarity matrices) occurred to be morestrong, than those determined in successive crop years(ANOSIM R = 0.53, 0.21, 0.37; P = 0.001, 0.02, 0.006;PERMANOVA pseudo F = 8.03, 4.97, 5.02, P = 0.001,0.019, 0.003).The fungal species composition of non-sterilizedgrains differed from the mycobiome of surfacesterilizedgrains primary due to a higher abundanceof Basidiomycetes (Cryptococcus spp. and otherTremellales, and Cystofilobasidium macerans and otherCystofilobasidiaceae) and Davidiella ( Cladosporiumspp.) in the non-sterilized grains. All Basidiomycetesdisappeared or became sparse after surface sterilization.The most abundant of them, Cryptococcus tephrensis(otu204) and C. victoriae (otu124), were also revealedinside grains but in fewer samples and in lesser amounts.Several OTUs, e.g., Cryptococcus wieringae (otu183),Mrakiella sp. (otu254), and Dioszegia sp. (otu303),tended to present on seed surfaces during only oneyear. Mycobiomes observed in two different growingseasons differed by abundance of Pyrenophora sp.in 2014 and Fusarium spp. and Phaeosphaeria sp. in2015. The Alternaria, Bipolaris, and Epicoccum generawere relatively abundant in both sample sets. Moredetailed results of fungal coexistence in the samples areintroduced in Fig. 6.Agar plate-based method of identification offungi. From 87 to 117 fungal isolates per 100 grainswere obtained from each sample. As a result of twogrowing seasons, a total of 18 taxa of seed-borne fungiwere identified (Table 1). The taxonomic position ofsome fungi was vague due to the lack of sporulation(Mycelia sterilia). In both years, the appearance ofAlternaria, Bipolaris, Epicoccum, and Fusarium species Figure 6 Fungal taxa heat map of sterilized and non-sterilized grain samples harvested in 2014 and 2015. OTUs are groupedaccording to UPGMA clusterization of Bray-Curtis dissimilarity matrix representing coexistence of OTUs within samples.Read counts of each OTU were weighted according to sum of reads in the sample, and then the proportion of the OTU dominancebetween samples was calculated and log-transformed. White corresponds to OTU absence in sample and red to OTU with highrelative abundance across the samples Kri –Krinichnyy, Suz – Suzdalets, B110 – Belgorodskiy 110, Tat – Tatum,and M86 – Moskovskiy 86was common. Species of the genus Pyrenophora werefound only in the 2014 growing season. Some potentiallytoxigenic fungi, such as Penicillium, Aspergillus,Cladosporium and unidentified Zygomycota, were foundonly in a few samples. No Basidiomycetes were isolatedand identified by agar plate-based assay.In both years, fungi of the genus Alternariapredominated in barley grain samples. The membersof two sections, Alternaria and Infectoriae, weredetermined. More precise identification was notperformed, since species concept is debatable forAlternaria and Infectoriae [53–56] sections.Contamination of the barley grains by Fusariumspp. varied significantly in 2014 and 2015. In 2014, theFusarium infection frequency was low (0–4%) andrepresented by five species, of which F. avenaceumwas the most frequent (infection frequency up to 2.5%,relative abundance of isolates up to 2.2%). In 2015,the infection of barley grains with Fusarium spp. wasconsiderably higher (infection frequency 14–19%,isolate abundance 12–17%). Eight Fusarium species wereidentified; four of them were common for both years.Comparison of methods. In total, 43 OTUs assignedas Ascomycota (26) and Basidiomycota (17) wererevealed by DNA metabarcoding. Only 14 OTUs wereassigned to species level. From those species, only twowere reoccurred in traditional mycological analysis. Theother 12 species either were not detected among isolatesgrown up from grains on agar medium or were Myceliasterilia. At the same time, the conventional mycologicalseed test revealed 17 Ascomycetes, including 11 species,apart from some Zygomycetes and sterile Ascomycetes.Basidiomycetes were not recovered by conventionalassay. Two species (Fusarium poae and Bipolarissorokiniana), one section (Alternaria sect. Alternaria),and three genera (Davidiella [Cladosporium], Fusarium,and Pyrenophora) were formally common for bothassays. In general, the list of undoubtedly identifieddominant taxa coincides with the results of the NGSmycobiome study of Canadian barley grains [3].The predominant OTUs from Alternaria wereidentified as Alternaria and Pseudoalternaria sectionswhen Alternaria and Infectoriae sections were fixedduring mycological analysis. In both cases, taxawere identified to the section level. Such precisionis sufficient for the majority of practical purposes,e.g., for tests of seed, food, or feed-grain quality. Thebig section Infectoriae and lately described sectionPseudoalternaria are morphologically similar andphylogenetically close groups [46]. This obviouslycan be the cause of errors, if identification is based onmorphological features.Both methods similarly reflected a very lowabundance of Fusarium spp. in 2014 and a higherquantity in 2015. Traditional mycological analysisrevealed nine Fusarium species. DNA metabarcodingresults were more limited; only one OTU was identifiedas a certain species, F. poae, but the others wereassigned to a clade level. Phylogenetic resolutionderived from ITS2 is not useful in defining Fusariumspecies. Recently, Fusarium-specific primers targetingtranslation elongation factor 1 (TEF1) were evaluatedand successfully applied to analyze Fusariumcommunities in soil and plant material [57].The taxonomy of Fusarium fungi is confusing andvarious classification systems have been proposed [58].For Fusarium, chemotaxonomy is considered asupplement to traditional morphology-based taxonomy.Several fungal genes involved in trichothecene andenniatins biosynthesis have been defined and used fordevelopment of molecular assays aimed at identification.In spite of the ITS sequences used in our analysis, theresults strongly suggested the division of fungi based ontheir ability to produce metabolites. In the future, thiswill provide an opportunity to predict the severity ofgrain contamination by some mycotoxins according tothe number of certain identified OTUs.Fusarium avenaceum, F. poae, F. tricinctum, andF. sporotrichioides were the most abundantrepresentatives of the genus. They are the typicalpathogens of barley in northwestern Russia [59, 60].Most likely, multi-copy otu259 discovered by DNAmetabarcoding is associated with F. avenaceum, whichoccurred frequently on the barley grain.Both methods revealed pathogenic fungi fromPleosporaceae: Bipolaris and Pyrenophora. Thosefungi have different patterns of appearance through thecropping seasons. DNA metabarcoding demonstratedhigher sensitivity. Pyrenophora sp. colonies were notrecovered in 2015 at all, but several respective readswere obtained for 7 out of 10 samples.Davidiella (Cladosporium) associated reads wereabundant in DNA metabarcoding assay in non-sterilizedsamples but only single colonies were detected inthe agar plate-based test. Underestimation of relativeabundance of the fungus in the latter case can be resultof two reasons: rapidly spreading colonies suppress ormask slowly growing fungi, and infected individualgrains contain not uniform quantity of fungal biomassthat appear as insufficient correlation between the number of infected grains and the amount of fungalDNA in the whole sample.Four fungal genera revealed by only DNA metabarcodingcontained agents of cereal diseases(Neoascochyta, Botrytis, Microdochium, andPhaeosphaeria). The first three taxa were represented bysolitary reads. Phaeosphaeria (otu106 and otu215) werefound in 14 of 20 samples. In 2015, in surface-sterilizedsamples, the relative abundance of Phaeosphaeriareads varied between 11 and 36%. Sequences of otu106had the closest similarity (99%) with representativesequences of Parastagonospora avenae ( Septoriaavenae or Stagonospora avenae), widespread funguscausing leaf blotch of barley and some other cereals, andParastagonospora poagena, a recently described fungusfrom Poa sp. [61, 62]. Less abundant OTU, otu215, hada similarity of 98%, with several Phaeosphaeria speciesand with some unidentified endophytes.CONCLUSIONDNA metabarcoding, based on high-throughputsequencing, is a sensitive and powerful method of grainmycobiome analysis that provides large amounts of data.However, at this time, not all fungi can be identifiedto species level by molecular markers, especially byrDNA sequences. In spite of universality, rDNA hasa limitation as a taxonomic marker. The resolutionof the ITS sequence-based method is not enough todifferentiate many fungal species. For instance, manyFusarium species have nonorthologous copies of ITS2.Many other important plant pathogenic and toxigenicfungi also can be identified up to genus level, but that isnot always informative in the framework of mycologicalseed expertise. Erroneous and chimerical sequences, aswell as the lack of reference sequences of many species,still limits wide application of NGS-based technologiesin biodiversity studies.The most complete and credible results can beobtained when several approaches are implementedsimultaneously. Combining the results of DNAmetabarcoding and traditional culture-platingassay allowed us to revise the diversity of fungicolonizing on the surface of and inside barley grains inLeningradskaya oblast (northwest Russia).Fungal species diversity of barley grain revealedby DNA metabarcoding formally exceeded thetraditional microbiological culture-based agar plating:43 operational taxonomic units (OTUs) vs. 17 taxaof genus or species level. DNA metabarcoding assayallowed seven ascomycete taxa to be added to the totallist. Of those additional taxa, only Phaeosphaeria wasabundant internal fungus. Seventeen OTUs belongingmainly to surface-seed-borne, yeastlike Basidiomyceteswere completely outside the scope of traditional analysis.Meanwhile, routine mycological analysis, in contrast toDNA metabarcoding, resulted in precise identificationof practically important Fusarium species. On the otherside, due to DNA analysis, one Alternaria taxon wasreidentified as Alternaria section Pseudoalternariainstead of section Infectoriae.CONTRIBUTIONAuthors are equally related to the writing of themanuscript and are equally responsible for plagiarism.CONFLICT OF INTERESTThe authors declare that there is no conflict ofinterest regarding the publication of this article.ACKNOWLEDGEMENTSThe authors are grateful to A. Vagin (Volosovo StateSeed-Trial Ground, Russia) for providing seed samplesand key information on them, and to M. Gomzhina(All-Russian Institute of Plant Protection) for technicalassistance. Pyrosequencing was conducted using theequipment of the Core Centrum “Genomic Technologies,Proteomics and Cell Biology” at All-Russia ResearchInstitute for Agricultural Microbiology (ARRIAM,St. Petersburg, Russia).