Introduction

Sorghum (Sorghum bicolor (L.)) was domesticated in northeast Africa 4–5000 years ago, although anthropological records indicate it has been consumed by humans as early as 8000 BC1,2. In 2016, the U.S. was the world’s leading producer of sorghum, followed by Nigeria, Sudan, Mexico, India, and China (https://www.worldsorghumproduction.com/previous-year.asp). In the U.S., sorghum is mostly grown as a grain crop, although it is also produced for silage and syrup3,4, and has recently garnered interest as a feedstock for cellulosic ethanol production5 and in culinary applications due to its gluten-free properties6. Sorghum is naturally adapted to harsh growing environments, particularly locations where heat, drought, and marginal soils limit the profitability of crops such as corn or wheat7.

Sorghum production, particularly in warm, humid growing areas, is hindered by a wide variety of fungal foliar diseases8. Noteworthy in the southeastern U.S. is target leaf spot, caused by Bipolaris cookei [=B. sorghicola = Drechslera sorghicola = Helminthosporium sorghicola 9]. For many decades after its initial description10, target leaf spot was considered to be a disease of minor concern in the U.S.11 and other regions of the world12,13. However, within two decades after its first reported observation on grain sorghum in Mississippi in 198614, target leaf spot has become widespread in the lower Mississippi river valley. Foliar lesions are distinctly rectangular (Fig. 1A and B), and leaf spots often contain scalariform bands of pigmentation that inspired the disease’s name. The pathogen is generally restrained by the major veins of leaves, but lesions can coalesce under heavy disease pressure, resulting in irregular patches of chlorosis and/or premature leaf death10. Most species of Bipolaris with defined pathogenicity lifestyles are categorized as necrotrophic pathogens15, although hemibiotrophic species have been noted16. In B. cookei, the growth habit underlying pathogenesis on sorghum has not yet been determined. However, the rapid onset of necrosis after infection, unimpeded intracellular growth during host colonization, and an apparent lack of specialized infectious structures associated with biotrophy (such as infectious hyphae or haustoria) collectively suggest a necrotrophic lifestyle12.

Figure 1
figure 1

Target leaf spot of sorghum. (A) Sorghum plants with symptoms of target leaf spot in the field. (B) Necrotic lesions caused by target leaf spot on a sorghum leaf.

The molecular basis of target leaf spot is poorly understood. Quantitative trait loci underlying resistance to target leaf spot have been explored to some extent17,18. A single recessive resistance gene (ds1) was positionally cloned and postulated to encode a leucine-rich receptor kinase family protein19,20. Comparative transcriptomics during infection of resistant and susceptible hybrids with B. cookei identified both plant and fungal genes 12 and 24 hr after infection21,22. However, only 160 transcripts were attributed to B. cookei, likely due to the low amount of fungal biomass and difficulty mapping transcripts due to the lack of a reference genome sequence for the pathogen.

To address the lack of genomic resources of B. cookei and provide new insights into the mechanistic basis of target leaf spot, a draft genome assembly of B. cookei was assembled and analyzed. Broad categories of genes involved in pathogenesis, such as carbohydrate-active enzymes and genes involved in secondary metabolism, were identified and assessed through comparative genomics. Additionally, novel genes were identified that could play important roles in pathogenicity.

Results

Genome sequencing, assembly, and annotation

The B. cookei genome was sequenced with Illumina and Pacific Biosciences (PacBio) sequencing technologies. Illumina sequencing produced 176,463,008 reads (2 × 100 bp paired-end sequences), and PacBio sequencing produced 251,565 reads with lengths of 35 to 21,246 bp (average length = 2,427 bp). A hybrid assembly pipeline (Supplementary Fig. S1) incorporated both types of reads to produce a draft genome assembly of 36,171,030 bp organized into 321 scaffolds (average length = 112 kb) (Table 1).

Table 1 Genome assembly statistics of B. cookei compared with other Bipolaris species.

Approximately 14% of the B. cookei draft genome sequence was comprised of transposable elements (TEs). Of these, DNA transposons (class 2 TEs) corresponded to 9% of the B. cookei genome, and retrotransposons (class 1 TEs), comprised 3% (Supplementary Fig. S2). The majority of the classified DNA transposons contained a DDE superfamily transposase motif23, which comprised approximately 3 Mb of the draft assembly.

To assist gene prediction, RNA-seq was performed from in vitro cultures of B. cookei. RNA was sequenced with Ion Torrent technology, which generated 6,273,768 reads (average length = 247 bp). After RNA-seq reads were mapped to the B. cookei genome assembly, 17,998 transcripts were reconstructed, corresponding to 11,451 distinct loci. With reconstructed transcripts and proteins from closely related species as evidence, 11,189 distinct protein-encoding genes were predicted (average length = 1707 bp) (Table 2). Assessment of predicted genes with BUSCO software showed 95.7% gene completeness, with 0.2% duplication, and 2.9% fragmentation. Functional characterization of the predicted B. cookei proteins revealed that 10,801 proteins (96%) had at least one homologous sequence in the NCBI nr database (e-value < 1e-5), 8,422 proteins (75%) had a conserved domain matching the InterPro database, and 9,034 proteins (80%) had a GO term attributed by Blast2GO.

Table 2 Gene prediction statistics of B. cookei and other Bipolaris species.

Homology-based analyses with mating-type loci from related Bipolaris species indicated that B. cookei strain LSLP18.3 contained a MAT-1 gene (Bc_02106). The predicted protein encoded by Bc_02106 shared 87% amino acid identity with MAT-1 from B. maydis, and was located within a genomic region conserved between B. maydis and B. cookei (Supplementary Fig. S3).

In addition to the nuclear genome, the B. cookei mitochondrial genome was assembled as a circular sequence of 135,790 bp with a GC content of 30% (Fig. 2). A total of 75 mitochondrial genes were predicted, including 12 of 14 highly conserved genes among fungal mitochondria: four subunits of the respiratory chain complexes (cox1, cox2, cox3 and cob), seven NADH dehydrogenase subunits (nad1, nad2, nad3, nad4, nad4L, nad5 and nad6), and one ATP synthase (atp6)24. Additionally, 30 genes encoding tRNAs were predicted, which were able to recognize all 20 standard amino acids; 2 rDNAs (small and large subunits); a gene encoding ribosomal protein S3 (rps3); and 30 ORFs predicted to encode homing endonucleases (21 from the LAGLIDADG family and 9 from the GIY-YIG family) located within intronic regions. The B. cookei mitochondrial genome assembly contained 36 introns (average length = 2,284 bp), which comprised a total of 82,220 bp. Comparative genomic analyses with the mitochondrial gene cob from Cercospora beticola 25 indicated that the mutation G143A, which is associated with resistance to QoI fungicides, is not present in B. cookei strain LSLP18.3 (Supplementary Fig. S4).

Figure 2
figure 2

Circular representation of the B. cookei mitochondrial genome. Mitochondrial genes are represented as rectangles on the outermost circle. Genes on the forward and reverse strand are drawn outward and inward, respectively. Genes encoding subunits of the electron transport chain of complex I (nad1–6) are in green, complex III (cox1–3) in cyan, complex IV (cob) in orange, ATP synthase (atp6) in yellow, rDNAs (rns and rnl) in red, tRNAs in black and ribosomal protein S3 (rps3) in blue. ORFs encoding homing endonucleases are drawn in purple, and introns are represented as white regions. The innermost circle represents the GC content, calculated with a sliding window of 1 kb and step size of 100 bp. Regions with the GC content below and above the average (30%), are in blue and orange, respectively. A grey circle designates the 50% GC content.

Carbohydrate-active enzymes (CAZymes)

A total of 589 genes encoding CAZymes were identified in the B. cookei genome (Table 3; Supplementary Table S1). The three most populated CAZyme families were CE10 with 38 genes, AA7 with 32 genes, and CE1 with 26 genes. Both CE10 and CE1 families include a great variety of esterases, such as acetyl xylan esterases (EC: 3.1.1.72), carboxylesterases (EC: 3.1.1.1), and acetylcholinesterases (EC: 3.1.1.7)26, whereas family AA7 includes gluco- and chitooligosaccharide oxidases (EC: 1.1.3.-)27.

Table 3 Number of genes encoding carbohydrate-active enzymes across different Bipolaris species.

CAZymes categorized as glycoside hydrolases (GHs) receive particular attention for their ability to hydrolyse glycosidic bonds between carbohydrates or between a carbohydrate and a non-carbohydrate moiety26. In B. cookei, the most populated GH families were GH16 with 16 genes, GH43 with 14 genes, and GH3 with 13 genes (Fig. 3). Family GH16 is comprised of numerous endoglucanases, which catalyze hemicellulose degradation28. Family GH43 includes enzymes such as xylanases, α-L-arabinofuranosidases, and β-D-galactosidases, that debranch and degrade hemicellulose and pectin29. Most B. cookei CAZymes from family GH3 were functionally annotated as β-glucosidases (EC: 3.2.1.21). Beta-glucosidases play an important role in the degradation of cellulose by converting cellobiose into glucose30.

Figure 3
figure 3

Heatmap showing the number of genes encoding glycoside hydrolases (GHs) and other CAZyme families that include plant cell wall degrading enzymes. B. cookei and 16 other members of the Dothideomycetes were hierarchically clustered. The full name of each species is provided in Supplementary Table S1.

Many members of the AA class of CAZymes are carbohydrate oxidases that assist other enzymes from GH, PL and CE classes to gain access to carbohydrates present in plant cell walls31. Two-thirds of the AA CAZymes in the B. cookei genome belonged to families AA7 (32 genes), AA9 (23 genes), and AA11 (11 genes). While family AA7 includes gluco- and chitooligosaccharide oxidases, family AA9 is represented by lytic polysaccharide monooxygenases (LPMOs) that help cellulases and hemicellulases break down cellulose and hemicellulose32. CAZymes from family AA9 often act in conjunction with cellobiose dehydrogenases (CDHs), found in family AA3, to accelerate oxidative degradation of cellulosic materials, including foliar tissues of plants32,33,34.

B. cookei and 16 other plant pathogenic members of the Dothideomycetes were clustered based on the number of CAZymes belonging to families likely involved in plant cell wall degradation. In concordance with previous studies, species within the Capnodiales and Pleosporales were organized in two distinct branches that resemble their phylogenetic placement35,36,37 (Fig. 3). The number of enzymes from families AA9 (LPMOs), GH64 (β-1,3-glucanases), CE5 (cutinases), PL1, and PL3 (pectin lyases) were notably different between Capnodiales and Pleosporales members. The number of GH13 enzymes, which include α-amylases and related enzymes involved in starch degradation38, was also markedly different. On average, members of the Capnodiales had 12.6 proteins from family GH13, whereas members of the Pleosporales had 6.9. Interestingly, hierarchical clustering grouped all analyzed Bipolaris spp. together in a distinct branch with the exclusion of B. cookei. B. cookei instead grouped most closely with Setosphaeria turcica in a different branch that also included Pyrenophora tritici-repentisP. teres f. teres, and Stagonospora nodorum (Fig. 3).

Secondary metabolism genes

Plant pathogenic fungi possess a diverse array of genes involved in the biosynthesis of secondary metabolites (SMs). SMs are often produced by genes organized into clusters that contain one or more genes referred to as backbone genes, which are primarily classified as polyketide synthases (PKSs), non-ribosomal peptide synthetases (NRPSs), hybrid PKS-NRPSs, dimethylallyl tryptophan synthases (DMATs), or terpene synthases (TSs)39. B. cookei had 47 backbone genes involved in secondary metabolism, categorized as 20 PKSs, 17 NRPSs, 3 DMATs, and 7 TSs, which is consistent with other Bipolaris species analyzed so far (Supplementary Table S2). One of the PKSs was a type III PKS, and the remaining 19 were type I PKSs, which were further classified into 10 highly reducing (HR-PKS), 2 partially reducing (NR-PKS), and 7 non-reducing (NR-PKS) (Supplementary Table S3). Orthologs of one NR-PKS from B. cookei (Bc_10041) could not be identified in the genomes of other Bipolaris species via homology searches (Supplementary Table S4), and the most similar protein identified among all fungi corresponded to a hypothetical gene from another sorghum pathogen, Colletotrichum sublineolum (55% amino acid identity). Gene Bc_10041 was located in a genomic region enriched with repetitive DNA, in close proximity to two other backbone genes encoding NRPSs, Bc_10039 and Bc_10040 (Fig. 4A). Interestingly, putative orthologs of Bc_10039 were not detected in the genomes of other Bipolaris species, and Bc_10040 was orthologous to NPS9 (78% identity), a NRPS-encoding gene from B. maydis that was previously thought to be unique to this species15 (Supplementary Table S4).

Figure 4
figure 4

Secondary metabolism genes. (A) Genomic region around B. cookei gene Bc_10041, a PKS-encoding gene possibly acquired from Colletotrichum sublineolum via horizontal transfer. Gene Bc_10038 encodes a putative alpha-xylosidase, Bc_10039 and Bc_10040 both encode NRPSs, Bc_10042 and Bc_10043 are hypothetical genes. (B) Ophiobolin gene cluster in Aspergillus clavatus and the corresponding homologous cluster in B. cookei. A. clavatus genes that are part of the cluster are labelled oblA-oblD, and other genes are labelled with the JGI protein ID. oblA encodes a terpene synthase, oblB encodes a cytochrome P450, oblC encodes a flavin-dependent oxidase, and oblD encodes a major facilitator transport protein. Genes Bc_07939, Bc_07940, and Bc_07942 encode hypothetical proteins. Bc_07939 has a bZIP domain, and Bc_07942 has a NAD(P)-binding domain. Repetitive DNA is represented as small rectangles.

Phytotoxic sesquiterpenoids of the ophiobolin family have reportedly been produced by B. cookei and other Bipolaris species40,41,42. Recently, the gene cluster responsible for ophiobolin F biosynthesis was characterized in Aspergillus clavatus 43. Based on this report, a similar gene cluster was identified in B. cookei that contained homologs of the terpene synthase oblA (Bc_07938, 62% identity), the cytochrome P450 oblB (Bc_07937, 68% identity), the FAD dependent oxidoreductase oblC (Bc_07941, 52% identity), and the ABC transporter oblD (Bc_07943, 76% identity) (Fig. 4B). Two additional ORFs present in the A. clavatus ophiobolin cluster were also conserved in B. cookei: a homolog of Aspc_1602 (Bc_07939, 38% identity; a hypothetical protein with no conserved domains), and a homolog of Aspcl_1604 (Bc_07940, 43% identity; a putative bZIP transcription factor). The additional ORF in the B. cookei ophiobolin cluster that is not present in the A. clavatus cluster (Bc_07942) encoded a hypothetical protein containing a NAD(P)-binding domain.

The PKS-encoding gene Bc_07091 of B. cookei was highly homologous to PKS18 of B. maydis (98% amino acid identity), which is involved in melanin biosynthesis44. The draft genome sequence of B. cookei contained additional genes involved in melanin biosynthesis, including the 1,3,8-trihydroxynaphthalene reductase BRN1 (Bc_07086, 92% identity), the transcription factor CMR1 (Bc_07088, 97% identity), the 1,3,6,8-tetrahydroxynaphthalene reductase BRN2 (Bc_07555, 99% identity), and the scytalone dehydratase SCD1 (Bc_10431, 98% identity) (Supplementary Fig. S5).

Secretome and candidate effectors

A total of 1035 genes were identified in the B. cookei nuclear genome that encoded proteins containing a signal peptide for secretion. Among these genes, 170 were predicted to encode proteins containing transmembrane domains or a glycosylphosphatidylinositol (GPI) anchor (Supplementary Table S5), and thus were predicted to be cell surface proteins45. The remaining 865 proteins were determined to comprise the B. cookei secretome. Nearly half (42%) of the predicted secretome consisted of proteins implicated in aspects of primary metabolism. Specifically, the putative secretome contained 253 CAZymes, with AA families being the most abundant: 22 proteins from family AA9, 18 proteins from family AA7, and 15 proteins from family AA3. Additionally, 74 proteases, 43 peroxidases, and 107 lipases were identified in the B. cookei secretome (Supplementary Table S5).

A set of 233 proteins were classified as candidate effectors, or small secreted proteins (SSPs). Among these, 133 could not be assigned a putative function based on homology, and 149 had no conserved domains. One of the B. cookei SSPs (Bc_04981) was homologous to Ecp6 from the tomato pathogen Cladosporium fulvum 46. Ecp6 is an effector that contains lysin motif (LysM) domains, and sequesters chitin oligosaccharides released from the fungal cell wall to avoid host chitin-triggered immune responses46. The architecture of the three LysM domains found in Ecp6 was conserved in the B. cookei homolog Bc_04981 (Supplementary Fig. S6)

Expression-based analyses to identify candidate genes involved in pathogenesis

Fungal genes preferentially expressed during host infection are potentially involved in pathogenicity. To qualitatively identify B. cookei genes highly or exclusively expressed in planta, an RNA-seq data set was obtained from B. cookei grown on various defined culture media in vitro, representing the basal transcriptome, which was then compared to in planta expression data from sorghum leaves infected by B. cookei 21 (accessions: DRR006371 and DRR006373). Of the 66 million reads obtained by Yazawa and co-authors21, 537,582 (0.8%) mapped to the B. cookei draft genome assembly. Fungal RNA-seq reads identified in planta accounted for 54,295,782 bp, as compared to 1,422,960,112 bp (5,531,920 reads) obtained in vitro. A total of 37 genes were considered to be induced in planta, 16 of which had no evidence of expression in vitro (Table 4). Several of the identified genes induced in planta (e.g. oxidoreductase activity enzymes, major facilitator superfamily transporters, and SM backbone genes) were associated with fungal secondary metabolite biosynthesis or detoxification of compounds. Additionally, four genes encoding candidate effectors were also considered to be induced in planta.

Table 4 B. cookei genes considered induced during infection.

Interestingly, some of the genes induced in planta were clustered in two regions of the B. cookei genome (Fig. 5A and B). Both regions were rich in repetitive DNA, and contained signs of repeat-induced point (RIP) mutations. One region (Fig. 5B) contained four genes exclusively expressed in planta. These four genes corresponded to a CAZyme from family AA7, annotated as 6-hydroxy-d-nicotine oxidase (Bc_11188), a putative cytochrome P450 (Bc_10329), a putative dehydrogenase reductase (Bc_10330), and a protein containing a nuclear transport factor 2 superfamily domain, annotated as SnoaL-like polyketide cyclase family 2 (Bc_10331). Homology searches revealed that these four genes were not conserved in other Bipolaris species or members of the Pleosporaceae family (Supplementary Table S4). Additionally, the closest homolog of Bc_10331 was a hypothetical protein from the sorghum pathogen C. sublineolum (67% amino acid identity).

Figure 5
figure 5

Regions of the B. cookei genome containing genes induced during sorghum infection. B. cookei genes designated as induced during early stages of sorghum infection are indicated with “*”. Putative functions of the genes are (A) aldehyde dehydrogenase (Bc_04816), aromatic ring-opening dioxygenase (Bc_04818), X-Pro dipeptidyl-peptidase S15 (Bc_04819), and galactonate dehydratase (Bc_04820). (B) 6-hydroxy-d-nicotine oxidase (Bc_11188), cytochrome P450 (Bc_10329), short-chain dehydrogenase reductase (Bc_10330), snoal-like polyketide cyclase (Bc_10331). (C) Candidate effector (Bc_06468). Plots above the genes represent the RNA-seq coverage in vitro and in planta (logarithmic scale), and the RIP index (TpA/ApT) calculated with a sliding window of 300 bp and step size of 50 bp. The minimum RIP index considered as evidence of RIP (0.89)104 is marked with a horizontal grey line. Repetitive DNA is represented as small grey rectangles.

Additional genes with exclusive in planta expression evidence included a serine protease (Bc_04819), an additional AA7 CAZyme (Bc_06321), and two candidate effectors (Bc_06468 and Bc_04701) (Table 4). Bc_06468 was located within a genomic region rich in repetitive DNA that showed signs of RIP mutation (Fig. 5C), and was classified as a putative effector due to its small size (98 amino acids), predicted signal peptide for secretion, and lack of conserved domains. Interestingly, homologs of Bc_06468 were found only in B. maydis, B. zeicola, B. victoriae, B. sorokiniana, B. oryzae, S. turcica, P. tritici-repentis, Fusarium oxysporum and F. oxysporum f. sp. cubense (Supplementary Table S4).

The PKS-encoding gene exclusively expressed in planta (Bc_04579) was homologous to alt5 (82% identity), a PKS gene from Alternaria solani postulated to be involved in alternapyrone biosynthesis47. Homologs of other genes also predicted to be involved in alternapyrone biosynthesis were clustered with Bc_04579, including Bc_11190 (82% identity with the cytochrome P450 alt2), Bc_11189 (87% identity with the cytochrome P450 alt3), and Bc_04580 (76% identity with the FAD-dependent oxygenase/oxidase alt4) (Supplementary Fig. S7). Interestingly, these four genes were located on a 30 kb region of the B. cookei genome with little or no evidence of expression in vitro (Supplementary Fig. S8). This genomic region also contained four other genes predicted to encode an NAD-dependent epimerase (Bc_04582), a major facilitator superfamily transporter (Bc_04583), an acetylcholinesterase (Bc_04584), and a hypothetical protein (Bc_04585).

Enrichment of repetitive DNA in proximity to pathogenicity-related genes

In some plant pathogenic fungi, genes important for pathogenicity are in close proximity to repetitive genomic regions35. Genes co-localized with repetitive elements are postulated to provide pathogens an evolutionary advantage due to allelic diversification induced by RIP activity. Consistent with other plant pathogenic fungi35, candidate effectors and backbone genes for secondary metabolite production in B. cookei were localized in closer proximity to repetitive elements than the overall predicted genes and genes highly conserved among members of the Ascomycota (Fig. 6). More specifically, nearly all B. cookei backbone genes (40 out of 47; 85%) and most candidate effectors (153 out of 233; 65%) had repetitive DNA within 5 kb up- or downstream of their ORFs, while about half of all predicted genes (5,478 out of 11,189; 49%) had an analogous arrangement. We also observed similar enrichment of repeats near SM backbone genes and SSPs in the close relatives B. maydis C5 and B. sorokiniana, albeit not as pronounced as in B. cookei (Fig. 6). Interestingly, only 26% and 24% of the B. maydis C5 and B. sorokiniana predicted genes, respectively, had a repeat within a neighbor region of 5 kb, which is much less than the ratio in B. cookei (49%). These results indicate that B. cookei has a substantial enrichment of repetitive elements near genes potentially important for pathogenicity, and that repeats are more dispersed in the B. cookei genome as compared to closely related species.

Figure 6
figure 6

Proximity of classes of genes to repetitive elements. The charts show the fraction of genes from B. cookei, B. maydis C5, and B. sorokiniana that have a repetitive element within an adjacent region up- or downstream of their ORFs of size shown on the horizontal axis. SM backbones: secondary metabolite backbone genes (PKSs, NRPSs, TPs, hybrid PKS-NRPSs, DMATs, and TPs); SSPs: small secreted proteins (candidate effectors); CAZymes: carbohydrate-active enzymes; USCOs: universal single-copy orthologs among members of the Acomycota88.

Discussion

The draft genome sequence of B. cookei strain LSLP18.3 expanded comparative genomics analyses among members of the order Pleosporales within the Dothidiomycetes class of fungi. To date, the genomes of at least 60 members of the Pleosporales have been sequenced48. Although the genome size of B. cookei is very similar to other members of the Pleosporales, the set of predicted genes in B. cookei is notably smaller than many close relatives (Supplementary Fig. S9; Supplementary Table S6). Particularly in the context of other Bipolaris species, B. cookei had contracted sets of CAZymes, secreted proteins, and candidate effectors. Conversely, B. cookei had a substantially larger mitochondrial genome than most other species of Dothideomycetes for which a complete mitochondrial genome is available. The proliferation of introns and ORFs encoding putative homing endonucleases were primarily responsible for the expanded size. Fungal mitochondria usually contain a set of 14 conserved protein-encoding genes required for electron transfer and oxidative phosphorylation24,49. However, exceptions have been previously noted. More precisely, the absence of the ATP synthase genes atp8 and atp9 was also observed in the B. cookei close relatives Stagonospora nodorum 50 and Shiraia bambusicola 51, the only Pleosporales members with mitochondrial genomes published to date. This result supports the hypothesis that the absence of atp8 and atp9 is a common genomic feature among members of the Pleosporales. Interestingly, B. cookei strain LSLP18.3 does not contain a G143A mutation within the mitochondrial cytochrome b (cob) gene, which is associated with resistance to QoI fungicides25,52. At the time strain LSLP18.3 was isolated (2009), fungicides were not commonly applied to sorghum grown in Arkansas, although applications of foliar fungicides on grain sorghum in Arkansas have become increasingly common in recent years. Resistance to QoI fungicides recently emerged in related Dothidiomycetes pathogens of soybean (e.g., Cercospora sojina and Cercospora c. f. flagellaris) and spread rapidly throughout the southeastern U.S.53,54. The finished mitochondrial genome of B. cookei will help facilitate future population-level studies to explore the potential emergence of fungicide resistance in this pathosystem.

Pathogens must adapt rapidly to changing environments in order to win ongoing evolutionary arms races with their hosts. The genome sequence of B. cookei provides evidence for diverse mechanisms promoting genetic variability and genome evolution. Regarding sexual reproduction, B. cookei strain LSLP18.3 possesses an ortholog of MAT-1 that lacks obvious evidence of inactivation associated with RIP mutations, transposon insertion, or other disruptive mutations. The presence of a presumably functional MAT-1 mating-type locus indicates the potential for heterothallic sexual reproduction in B. cookei, which would be consistent with other closely related Bipolaris species55,56,57. Sexual recombination is a central driver of adaptation among many plant pathogenic fungi by creating new combinations of genes and alleles involved in pathogenesis58. Byproducts of sexual recombination can also modify genome architecture by inducing genome modifications and rearrangements, such as duplications, inversions, and deletions59,60. In plant pathogenic fungi, the ability to induce changes in gene content and genomic architecture is a key component of rapid adaptive evolution60. Mobile genetic elements such as transposons are associated with genomic rearrangements and the creation of novel genes and alleles in many plant pathogenic fungi, and specific selective advantages have been documented to result from such genomic modifications61. Similar to the genomes of many other plant pathogenic fungi, the genome of B. cookei contains a large number of taxonomically diverse transposable elements. The substantial enrichment of SM backbone genes and SSPs near repetitive elements, compared to a negative association of housekeeping genes and other predicted elements of the core genome, is consistent with the presence of ‘repetitive islands’, in which pathogenicity-associated genes are enriched and isolated from components of the core genome61. The dynamic nature of such repetitive islands is associated, in part, through genome defense mechanisms such as RIP mutations62. Although RIP presumably evolved as a mechanism for fungi to inactivate duplicated sequences in their genomes, the sequences of single-copy genes in close proximity to repetitive elements can be altered via RIP slippage61,63, thus leading to the generation of novel alleles. Considering the widespread distribution of repetitive elements in the B. cookei genome, and their notably proximity to secondary metabolite genes and candidate effectors, repetitive elements are a plausible source of genomic variability and evolution in B. cookei. Yet another mechanism of fungal genome diversification is horizontal gene transfer (HGT), in which genetic material is exchanged between reproductively isolated organisms64. Circumstantial evidence for horizontal gene transfer in B. cookei is provided by the NR-PKS gene Bc_10041 and the hypothetical gene Bc_10331. Homologs of Bc_10041 and Bc_10331 were present in Colletotrichum sublineolum, yet absent from other sequenced Bipolaris species, which is inconsistent with a pattern of vertical inheritance from a common ancestor. Because B. cookei and C. sublineolum are both pathogens of the same host, share overlapping geographical distributions, and have been noted to co-infect individual sorghum leaves, populations of both pathogens have presumably been in more than adequate physical proximity for HGT events to have occurred.

At the current time, whether B. cookei utilizes a hemibiotrophic or necrotrophic infection strategy is not fully resolved. Most plant pathogenic members of Bipolaris are generally considered to be necrotrophs15,35, although a few hemibiotrophic species are postulated to exist within the genus16. When considered in the context of the limited histopathological analyses of B. cookei infection12, the genome sequencing and analysis presented in this study lends support to the hypothesis that B. cookei is a necrotrophic pathogen. In particular, available information about gene content, such as the reduced set of candidate effectors, and gene expression are mostly consistent with a necrotrophic growth habit. The in planta expression set examined in this study was derived from an early stage of the infection process (24 hr after inoculation), at which time hemibiotrophs are typically initiating latent (asymptomatic) infections. As such, hemibiotrophic pathogens are thought to rely more heavily on suites of effectors that manipulate host defense reactions at early stages of infection, rather than phytotoxic secondary metabolites. However, genes for secondary metabolite production or detoxification in some hemibiotrophic fungi have been previously reported upregulated at early stages of infection65,66. Furthermore, it is important to note that the distinction between a necrotroph and a hemibiotroph is currently predicated on the existence of a consistent, definable latent phase before the visible onset of necrosis35,67, rather than a specific set of genetic or biochemical criteria. Thus, if the period between inoculation and symptom expression is compressed, as has been reported for B. cookei 12 the potential existence of an unusually short-lived hemibiotrophic phase cannot fully be discounted. Future histopathological experiments utilizing cell biology approaches, in conjunction with extensive transcriptional profiling, will be required to more conclusively define the infection strategy utilized by B. cookei.

The known host range of B. cookei is limited to certain members of the grass family (Poaceae), specifically species within the genera Sorghum and Zea 9. The current study provides a degree of insight into potential mechanisms underlying host specificity. In other fungal pathosystems, host specific toxins (HSTs) are important components of host-specific necrotrophy68,69. The B. cookei LSLP18.3 draft genome contains a substantial number of secondary metabolism genes and clusters, some of which have not yet been described in other fungal genome sequencing projects and thus are potentially unique to B. cookei. Interestingly, the SM backbone gene Bc_10041 is taxonomically restricted to B. cookei among Bipolaris genomes sequenced to date, yet an ortholog is present in the genome of the sorghum pathogen C. sublineolum. An intriguing possibility is that one or more metabolites requiring Bc_10041 for their biosynthesis could conceivably function as HSTs on sorghum. The gene expression profiling data analyzed in this study also suggest the potential existence of HSTs in this pathosystem. Among the 37 genes highly expressed in planta as opposed to in vitro, the majority encoded genes implicated in secondary metabolism, as well as some that appear to be involved in modification or detoxification of host phytoalexins. At this time, it is not clear whether B. cookei might utilize a suite of SM-derived HSTs, or if it may instead utilize a limited number of HSTs in conjunction with an arsenal of broad-spectrum necrosis-inducing toxins. Additionally, given that HSTs can be proteins rather than SMs, such as ToxA70, the potential contribution of diverse HSTs throughout various stages of the infection process is worthy of further investigation. When considering the potential roles of HSTs in host-specific necrotrophy, much less clear is the potential role of effectors, e.g., pathogen-derived proteins that modulate host defense. It is conceivable that necrotrophs could utilize effectors to manipulate host defense responses in such a way to accelerate cell death, which would be advantageous for pathogen growth. Gene Bc_06468 is particularly interesting as a candidate gene involved in necrotrophic pathogenesis. Although it may have a relatively conventional function such as suppressing host defense responses, it could potentially have a biochemical function related to the induction or maintenance of necrotrophy. However, its function is difficult to infer based on its structure, which contains no conserved domains, or its taxonomic distribution; putative orthologs were found only in related Bipolaris spp., S. turcica, P. tritici-repentis, and species of the F. oxysporum complex. Thus, functional characterization of Bc_06468 and other genes underlying pathogenesis will be required to clarify the molecular basis of host-specific necrotrophy in B. cookei.

In summary, this work presented a comprehensive analysis of genomic features in B. cookei and expression analysis of B. cookei genes induced at early stages of infection. The results provided important and novel insights into the molecular basis of target leaf spot on sorghum, and highlighted several candidate pathogenicity-related genes that were taxonomically restricted among Bipolaris species. Lastly, the draft genome assembly of B. cookei will serve as a valuable genomic resource for functional genomics studies and population genetics in this organism.

Materials and Methods

Fungal strain and culture conditions

Wild-type strain LSLP18 of B. cookei was isolated from a diseased sorghum plant collected at the University of Arkansas Lon Mann Cotton Research Station in Lee County, Arkansas, during the 2009 growing season. The isolate was selected for genome sequencing due to a high level of virulence and stable growth in routine laboratory culture. To confirm Koch’s postulates, the sorghum line BTx623 was grown in a growth chamber maintained at 25 °C under a 12 h photoperiod. A conidial suspension (5 × 104 conidia ml−1) in 0.01% Tween-20 was atomized onto the adaxial surface of sorghum leaves three weeks after planting until just before runoff. Inoculated plants were placed in transparent plastic bags and incubated in a growth chamber at 25 °C and near 100% relative humidity under a 12 h photoperiod. After 48 h, plants were removed from the bags and maintained in a growth chamber at 25 °C under a 12 h photoperiod until symptoms developed. Working cultures were maintained on V8 agar medium71. The fungus was stored as colonized agar in 30% (v/v) glycerol at −80 °C. For DNA isolation, a single-spore isolate (LSLP18.3) was derived from LSLP18 and grown in yeast extract peptone dextrose liquid medium72 for four days at 25 °C with shaking at 150 RPM.

For RNA sequencing, cultures of B. cookei were grown on eight different conditions on solidified media plates overlaid with cellophane (V8 juice agar at constant darkness and constant light; 0.2× strength potato dextrose agar; complete medium agar; minimal medium agar; minimal medium pH 3; minimal medium pH 8; minimal medium with ammonium as nitrogen source), liquid minimal medium, and yeast extract peptone dextrose medium72. The tissue from each culture was harvested five days after inoculation, flash frozen immediately, and ground in liquid nitrogen. The ground tissue was either used immediately for RNA extraction or stored at −80 °C.

DNA and RNA extraction and sequencing

For DNA isolation, strain LSLP18.3 was grown as described above. Tissue was collected by filtration, frozen in liquid nitrogen, and ground in a pre-chilled mortar and pestle containing sterile glass beads (0.5 mm) (Research Products International, Mt. Prospect, IL, USA). Genomic DNA was isolated with a modified cetyltrimethylammonium bromide (CTAB) method71. DNA was further purified with a Qiagen Genomic-tip 500/G column (Qiagen, Germantown, MD, USA) following the manufacturer’s recommendations. DNA quality and quantity was determined by agarose gel electrophoresis and a NanoDrop ND-1000 Spectrophotometer (Thermo Fisher Scientific, Waltham, MA, USA). Genome sequencing was performed with a hybrid sequencing approach. Library preparation (fragment size = 500 bp) and sequencing (one lane, paired-end, 100 bp read length) was performed by BGI Americas (Cambridge, MA, USA) with an Illumina HiSeq. 2000 Sequencing System (Illumina Inc., San Diego, CA, USA). Additionally, library preparation (library size = 3–10 kb) and sequencing (two SMRT cells) was performed by the Yale Center for Genome Analysis (Orange, CT, USA) with a PacBio RS II Sequencing System (Pacific Biosciences, Menlo Park, CA, USA).

For RNA isolation, strain LSLP18.3 was grown as described above. Tissue was collected, frozen in liquid nitrogen, and ground in a pre-chilled mortar and pestle containing sterile glass beads (0.5 mm) (Research Products International). Total RNA was isolated with a Direct-zol RNA MiniPrep Kit (Zymo Research, Irvine, CA, USA) following the manufacturer’s recommendations. DNA quality and quantity was determined by agarose gel electrophoresis and a NanoDrop ND-1000 Spectrophotometer (Thermo Fisher Scientific). Total RNA from each condition was mixed in equal amounts. Pooled total RNA (10 μg) was subjected to mRNA enrichment with a MagJET mRNA Enrichment Kit (Thermo Fisher Scientific). Ribosomal RNA-free mRNA was fragmented with RNase III (New England Biolabs, Ipswich, MA, USA) at 37 °C for 5 minutes. First strand cDNA was synthesized from the fragmented mRNA with random hexamers (Integrated DNA Technologies, Inc., Coralville, IA, USA) and M-MLV Reverse Transcriptase (Promega, Madison, WI, USA), and second strand synthesis was performed with a NEBNext mRNA Second Strand Synthesis Module (New England Biolabs). End repair was performed on the second strand DNA with a NEBNext End Repair Module (New England Biolabs). Following end repair, sequencing adapters were ligated to the double stranded cDNA with a NEBNext Fast DNA Fragmentation & Library Prep Set for Ion Torrent (New England Biolabs). Size selection for 480 bp was performed with an Agencourt AMPure XP kit (Beckman Coulter, Brea, CA, USA). A final PCR enrichment step was performed following the recommendations provided with the NEBNext Fast DNA Fragmentation & Library Prep Set for Ion Torrent to amplify fragments containing ligated adapters at both ends. The size distribution and quality of the library was determined with an Agilent Tapestation 2200 D1K (Agilent Technologies, Santa Clara, CA, USA). The library was sequenced on an Ion Torrent Personal Genome Machine System with an Ion 318 Chip Kit v2 (Thermo Fisher Scientific).

Genome assembly and repetitive DNA identification

The draft genome of B. cookei was assembled based on a custom hybrid strategy (Supplementary Fig. S1). The Illumina reads were corrected with Bayes Hammer within SPAdes v3.1 assembler73. Corrected Illumina reads were used to correct the PacBio reads with LoRDEC v0.274. Illumina and PacBio reads were analyzed with SPAdes v3.1, which produced a raw draft assembly. Contigs from the draft assembly were merged into scaffolds with successive iterations of SSPACE-Standard v3.075 and AHA from smrtanalysis suite 2.3 (http://www.pacb.com/support/software-downloads/). Genomic regions containing gaps were filled with GapFiller v1.1176 and PBJelly v14.7.1477.

A homology search with genes from the Stagonospora nodorum mitochondrial genome50 identified one assembled scaffold (scaffold75) corresponding to the B. cookei mitochondrial genome. Illumina and PacBio reads were mapped to the B. cookei assembly with Bowtie v2.2.978 and BLASR v1.3.179, respectively, and the reads that mapped to scaffold75 were given to SPAdes to produce a new assembly, which was improved with SSPACE-Standard v3.0. This produced a single contig with overlapping ends.

Repetitive DNA sequences in the B. cookei genome were identified with RepeatScout v1.0.580 using an l-mer value of 13, and filtered with the accessory script filter-stage-1.prl. Repetitive elements that appeared less than 10 times in the genome were filtered out with the accessory script filter-stage-2.prl and RepeatMasker open-4.0.5 (http://www.repeatmasker.org). Annotation of transposable elements was performed with TransposonPSI (http://transposonpsi.sourceforge.net). The Repeat-Induced Point (RIP) mutation index (TpA/ApT) was calculated with the dinucleotide Frequency function within the R package Biostrings v2.44.281 based on a sliding window approach with window size of 300 bp and step size of 50 bp. The distance of genes to repetitive elements was calculated with the function closest within BEDtools v2.2682.

Gene prediction and functional annotation

For gene prediction, RNA sequencing reads were first mapped to the draft genome with GSNAP v2014-10-0983 with the splicing option (-N) enabled. Then, mapped RNA-seq reads were analyzed with Cufflinks v2.2.184 to reconstruct transcripts with default parameters. Gene models were predicted with Maker pipeline v2.32.685 with the reconstructed transcripts from B. cookei as EST evidence, and protein sequences from B. maydis C5 and B. zeicola (http://genome.jgi.doe.gov/programs/fungi/index.jsf) as protein homology evidence. Gene models were selected with the accessory script Maker2zff within the Maker software package with default parameters to train the ab initio predictors SNAP v2006-07-2886 and Augustus v3.0.287. A new set of gene models was predicted with Maker adjusted to use SNAP and Augustus simultaneously with the B. cookei reconstructed transcripts, and the protein sequences from B. maydis C5 and B. zeicola. Gene completeness was assessed with BUSCO software v288 using protein assessment mode (-m protein) and the Ascomycota lineage (-l).

To functionally characterize genes, predicted protein sequences were queried against the NCBI nr database (update 03/2016) with BLAST v2.2.3189 and an e-value < 1e-5. BLAST results were analyzed with Blast2GO v3.090 to attribute functional descriptions and GO terms for each protein. Mitochondrial genes were identified and annotated with MFannot (http://megasun.bch.umontreal.ca/cgi-bin/mfannot/mfannotInterface.pl) and MITOS291. Uncharacterized ORFs in the mitochondrial genome were annotated with Blast2GO as described previously.

The mating type of B. cookei was determined with BLAST homology searches using as queries the mating genes MAT-1 (GenBank accession: CAA48464) and MAT-2 (GenBank accession: CAA48465) from B. maydis 92.

Carbohydrate-active enzymes were identified with a local installation of dbCAN v5.093 and HMM v3.1b1 (http://hmmer.org). Polyketide synthases (PKSs), nonribosomal peptide synthatases (NRPSs), and dimethylallyl transferases (DMTs) were identified with SMURF39, and terpene synthases (TSs) were identified with antiSMASH v3.0.594. PKSs were queried against the NCBI conserved domain database (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi) for further categorization. PKSs containing a starter unit:ACP transacylase (SAT) domain were classified as non-reducing PKSs, those containing an enoyl reductase (ER) domain were classified as highly reducing PKSs, and those that lacked both SAT and ER domains were classified as partially-reducing PKSs.

To identify putative secreted proteins, all predicted proteins were evaluated for the presence of signal peptides with SignalP v4.195 and TargetP v1.196. Proteins containing a signal peptide as well as two or more transmembrane domains, as determined with TMHMM v2.097, or a glycosylphosphatidylinositol (GPI) anchor, as determined with PredGPI98, were categorized as plasma membrane-associated rather than secreted. The remaining proteins (comprising the putative secretome) were further categorized into secreted proteases, lipases and peroxidases by performing BLAST searches against the MEROPS database v1199, Fungal Peroxidase Database100, and Lipase Engineering Database101, respectively, with a maximum e-value of 1e-5. Secreted proteins classified as effectors according to EffectorP v1.0102, or containing less than 300 amino acids with at least 2% of predicted residues corresponding to cysteine, calculated with EMBOSS package v6.6103, were considered to be candidate effectors.

Gene expression and comparative genomic analyses

RNA-seq reads from sorghum leaves infected with B. cookei 12 h and 24 h post-inoculation21 were combined and mapped to the B. cookei genome assembly with GSNAP v2014-10-09, and then compared with RNA-seq data obtained from B. cookei grown on various defined culture media in vitro. Reads mapped to each predicted ORF were counted with the function coverage within BEDtools v2.26. Genes with more than three in planta RNA-seq reads and zero in vitro RNA-seq reads were considered exclusively expressed in planta. In addition, genes with a ratio of at least 4:1 regarding in planta and in vitro RNA-seq reads, respectively, were also considered induced during sorghum infection.

Comparative genomic analyses with other Bipolaris species were performed with the genomes of B. maydis (=Cochliobolus heterostrophus) C5, B. sorokiniana (= C. sativus) ND90Pr, B. zeicola (=C. carbonum) 26-R-13, B. oryzae (= C. miyabeanus) ATCC 44560, and B. victoriae (= C. victoriae) FI315,35, obtained from the JGI website (http://genome.jgi.doe.gov/programs/fungi/index.jsf).

Data availability

This Whole Genome Shotgun project has been deposited at DDBJ/ENA/GenBank under the accession NRSV00000000. The version described in this paper is version NRSV01000000. The complete mitochondrial genome has been deposited at GenBank under accession MF784482. RNA-seq data obtained from culture media conditions have been deposited at NCBI sequence read archive (SRA) under accession SRR5957114.