Third-generation sequencing-based mapping and visualization of single nucleotide polymorphism, meiotic recombination, illegitimate mutation and repeat-induced point mutation

Abstract

Generation of new genetic diversity by crossover (CO) and non-crossover (NCO) is a fundamental process in eukaryotes. Fungi have played critical roles in studying this process because they permit tetrad analysis, which has been used by geneticists for several decades to determine meiotic recombination products. New genetic variations can also be generated in zygotes via illegitimate mutation (IM) and repeat-induced point mutation (RIP). RIP is a genome defense mechanism for preventing harmful expansion of transposable elements or duplicated sequences in filamentous fungi. Although the exact mechanism of RIP is unknown, the C:G to T:A mutations might result from DNA cytosine methylation. A comprehensive approach for understanding the molecular mechanisms underlying these important processes is to perform high-throughput mapping of CO, NCO, RIP and IM in zygotes bearing large numbers of heterozygous variant markers. To this aim, we developed ‘TSETA’, a versatile and user-friendly pipeline that utilizes high-quality and chromosome-level genome sequences involved in a single meiotic event of the industrial workhorse fungus Trichoderma reesei. TSETA not only can be applied to most sexual eukaryotes for genome-wide tetrad analysis, it also outcompetes most currently used methods for calling out single nucleotide polymorphisms between two or more intraspecies strains or isolates.

INTRODUCTION

Meiosis is a specialized cell cycle of sexual eukaryotes by which they undergo two rounds of division (Meiosis I and Meiosis II) to produce haploid gametes from diploid progenitor cells. A hallmark of meiosis is interhomolog recombination, i.e. exchange of genetic materials between homologous chromosomes, generating new genetic variations in the gametes. Tetrad analysis has been used in classical genetic experiments for several decades to determine meiotic recombination products such as crossovers (COs) and non-crossovers (NCOs). COs involve the exchange of flanking markers and possible gene conversion.

Over recent decades, the remarkable progress in next-generation sequencing (NGS) technologies has enabled high-throughput identification of meiotic recombination products bearing large numbers of heterozygous variant markers following hybrid meiosis, including single nucleotide variant (SNV) and insertion/deletion mutation (InDel). For convenience, Indels can be converted into SNVs by filling in hyphen symbols (-) or null (n) spaces. Accordingly, SNVs and Indels are collectively referred to as single nucleotide polymorphisms (SNPs). Accurate calling of SNPs and genotypes from NGS data is an essential prerequisite for genetics studies. However, there are two pitfalls to applying NGS-based platforms for SNP calling and genome-wide mapping of meiotic recombination products. First, due to the short lengths of NGS reads, it is difficult to accurately assemble the nucleotide sequences in chromosome regions hosting repetitive and/or high AT-biased sequences. As a result, the NGS reference genomes used for SNP calling are not only far from complete, but they also possess many unidentified nucleotides and assembly errors (1). Second, most bioinformatic pipelines used for SNP calling [e.g. Mummer4 (2) and or MapCaller (3)] only report SNPs present in the matching sequences but not in chromosomal regions with low or no sequence homology as well as those with repetitive or AT-biased sequences.

Third-generation sequencing (TGS) and mapping technologies have created a renaissance in high-quality genome sequencing. For example, PacBio's single-molecule real-time (SMRT) technology offers two principal advantages over NGS platforms for assessing high-quality and chromosome-level genome sequences: (i) it provides very long sequencing reads, with lengths 100 times longer than those of NGS; and (ii) it can easily cover DNA regions with high GC bias and repetitive sequences, facilitating better variant detection. However, PacBio reads typically have a high error rate (∼15% compared with ∼0.1% for Illumina reads) and, since the single-pass errors are distributed randomly, high sequence coverage can very rapidly wash out the single-pass errors when building consensus sequences (4).

Thanks to TGS technology, the availability of high-quality, near-complete genome sequences provides a new opportunity to improve genome-wide identification of sequence variations with single-nucleotide precision. In this study, we used meiosis of the filamentous workhorse fungus Trichoderma reesei (Teleomorph Hypocrea jecorina) as a model to demonstrate this application of TGS. Zygotes with highly diverse genome sequences are a prerequisite for genome-wide identification of meiotic recombination products. Therefore, we chose two T. reesei wild-isolate strains, QM6a and CBS999.97(MAT1-1), for sexual crossing (1,5-6). QM6a is the ancestor of all currently used cellulase-producing industrial strains (7–10) and it possesses a MAT1–2 mating-type locus (11). CBS999.97(MAT1-1) and CBS999.97(MAT1–2) were derived from two ascospores in a fruiting body of the CBS999.97 wild isolate strain (11). QM6a was originally isolated from one of the Solomon Islands, whereas CBS999.97 was sampled from French Guiana [see review in (9)]. Due to long-term geographic isolation, QM6a and CBS999.97 harbor high levels of sequence variation (6) (Figure 1). Sexual crossing of CBS999.97(MAT1-1) with CBS999.97(MAT1–2) or QM6a readily induces sexual development, resulting in fruiting bodies that contain linear asci each with 16 linearly arranged ascospores (i.e. the sexual spores) (11). The 16 ascospores in each ascus are generated from a meiosis event, followed by two post-meiotic rounds of mitosis (12). Accordingly, the 16 ascospores in each ascus can be classified into four genetically identical groups, with each group containing four ascospores.

Figure 1.

PacBio genome sequencing and assembly of QM6a and CBS999.97(MAT1-1). Collinearity of the QM6a and CBS999.97(MAT1-1) genomes is depicted in the inner circles of the diagrams. The Circos plot shows the synteny of QM6a and CBS999.97(MAT1-1) on the basis of the BLASTN results of chromosome-level nucleotide sequences, with the expect (e) value of 1e-20, >10 kb in length with >95% sequence identity (in gray). The two inversion alleles on ChI and ChV are indicated in magenta, whereas the mating-type loci MAT1-1 and MAT1–2 are in green. The GC contents (window size 5000 bp) of the seven chromosomes are shown in the first outer circle. The second outer circle indicates the seven chromosomes (I–VII) of QM6a (in blue) and CBS999.97(MAT1-1) (in red). The densities of SNP, SNV and InDel (window size 3000 bp) are depicted in the third, fourth and fifth circles, respectively. The D segments in QM6a (586 915 bp) and CBS999.97(MAT1-1) (529 007 bp) are indicated in orange. The L segments in QM6a (41 102 bp) and CBS999.97(MAT1-1) (26 657 bp) are indicated in black. Local inversions are shown in magenta. A reciprocal translocation between the D and L segments had been demonstrated previously by microarray-based comparative genomic hybridization (1) and electrophoretic karyotyping and Southern hybridization (6). D and L represent segmental duplication and loss (12), respectively, because most 16-spore asci (>90%) generated by sexual crosses between QM6a and CBS999.97(MAT1-1) contain four or eight inviable ascospores and an equal number of segmental aneuploid (SAN) ascospores. The viable SAN ascospores contain two D segments but no L segment (12).

We previously established high-quality, near-complete genome sequences of QM6a (1), CBS999.97(MAT1-1) and CBS999.97(MAT1–2) (6) using the PacBio SMRT platform. A hallmark of these three T. reesei genomes (∼34 Mb in size) is the presence of ∼2250 AT-rich blocks each with length ≥500 base pairs (bp) (1,6). These AT-rich blocks might be evolutionary remnants of repeat-induced point mutation (RIP), a genome defense mechanism against transposable elements (TE) that is conserved in many Pezizomycotina fungi. RIP was first discovered in Neurospora crassa. It occurs premeiotically in haploid parental nuclei in preparation for karyogamy and meiosis. RIP detects duplications of chromosomal DNA above a certain length threshold (∼0.4 kbp), leading to C:G to T:A mutations and concomitant DNA cytosine methylation in a pairwise fashion on both strands of each DNA duplex [reviewed in (13–15)]. It has been proposed that these AT-rich blocks might have important functions in fungal-plant interactions, symbiosis, genome evolution (e.g. gene loss, duplication and neofunctionalization), genome organization and transcription (1,16–17).

AT- and repeat-rich blocks in the genomes of some plant pathogenic fungi and oomycetes are gene-sparse compartments that contain fast-evolving and/or virulence-associated genes. In contrast, AT- or repeat-poor compartments are gene-rich and evolutionarily conserved. The mosaic features of the genome architecture in these fungi led to the ‘two-speed genome’ hypothesis (18). Interestingly, recent resequencing efforts using TGS technology have demonstrated that not only do some rapidly evolving phytopathogens lack some key characteristics described by the ‘two-speed genome’ model (19,20), so too do several economically important Trichoderma spp. (our unpublished results). For example, the putative genes encoding virulence-associated proteins or secreted effectors are rarely located in their AT-rich or gene-poor compartments. Thus, further modes of genome organization may exist, likely representing different speeds during genome evolution [see review of (19)].

Nevertheless, these TGS-based resequencing results also reveal the need to develop better bioinformatic tools for accurate and comprehensive analyses of highly mosaic and compartmentalized genomes. Also noteworthy is that high content of AT-rich blocks impedes accurate SNP calling and genome-wide mapping of meiotic recombination products by NGS platforms. To overcome these problems, we isolated the 16 ascospores generated from an ascus produced by sexual crossing of QM6a and CBS999.97(MAT1-1). These ascospores were then germinated and vegetatively propagated into 16 F1 progeny mycelial colonies. After genotyping by genomic polymerase chain reaction (PCR), we isolated high-quality, high-molecular-weight genomic DNA from four representative F1 progeny (i.e. F_A, F_B, F_C and F_D) and then applied PacBio's SMRT technology to determine high-quality and chromosome-level genome sequences (6). Here, we further present a highly organized and modular computational framework named TSETA (Third-generation Sequencing to Enable Tetrad Analysis) through which we processed all six TGS-based and near-complete haploid genome sequences. Our results indicate that TSETA is a convenient method for further exploring the basic mechanisms of meiotic recombination, not only in filamentous fungi but also in other sexual eukaryotes.

MATERIALS AND METHODS

Fungi

Fungal growth, culture media, sexual crossing, single ascospore isolation, preparation of high-quality genomic DNA, PacBio SMRT genomic DNA sequencing and assembly were performed as described previously (1,12,21). The Circos program [(http://circos.ca/documentation/tutorials/quick_start/selection_and_scale/; (22)] was used to show synteny and conservation between the genome sequences of QM6a and CBS999.97(MAT1-1).

To isolate the four representative F1 progeny (F_A, F_B, F_C and F_D) used in this study, QM6a was sexually crossed with CBS999.97(MAT1-1) to produce fruiting bodies containing asci with 16 linearly arranged ascospores. Next, we applied the yeast tetrad dissection technique to sequentially separate the 16 ascospores from an ascus. Each ascospore was individually germinated and cultured. Genomic PCR genotyping of all 16 ascospores from one hexadecad further confirmed that each hexadecad was classified into four linearly arranged groups and each group contained four genetically indistinguishable ascospores (12). Finally, F_A, F_B, F_C and F_D were selected from each of the four linearly arranged groups.

Hardware, operating system and network

The TSETA software package was designed for a desktop or computing server running with x86–64 Linux operating system. Multithreaded processors are preferred to speed up the process since many steps can be configured to use multiple threads in parallel. TSETA was written in JavaScript or ECMAScript. TSETA has two different functional modes: (1) The “SNP" model is a powerful analytic tool for single-nucleotide-resolution comparison of different intraspecies genomes. (2) The “Tetrad" model of TSETA aims for genome-wide identification of genetic variations before and after a single meiotic events, including meiotic recombination products, illegitimate mutations and repeat-induced point mutation. A stable internet connection is required for installation and configuration of TSETA, as well as for retrieving the test data. For a fungal genome of length ∼34 Mb, we recommend having at least 200 Gb and 20 Gb of RAM for TGS-based genome assembly and TSETA analysis, respectively, as well as 10 Gb of free disk space. Definitions of the different genetic markers in the four representative F1 progeny are presented in Table 1. For clarity, IMs were further classified into three different types, i.e. IM1, IM2 and IM3, respectively. A flow chart explaining the steps of the “Tetrad" model of TSETA program is depicted in Figure 2.

Table 1.

Definitions of different genetic markers

Paternal and maternal genomes			Four representative F1 genomes
QM6a	CBS999.97(MAT1-1)	SNP	FA	FB	FC	FD	Genetic markers
C	A	SNVa	A	C	A	C	SNV 2:2
C	A	SNV	A	C	C	A	SNV 2:2
C	A	SNV	C	C	A	A	SNV 2:2
C	A	SNV	A	A	A	C	SNV 3:1/1:3 (NCO)
C	A	SNV	A	A	A	A	SNV 4:0 (2NCO)
C	A	SNV	A	T	A	T	SNV RIP (Q)d
A	C	SNV	A	T	A	T	SNV RIP (C)d
C	C	SNV	T	T	T	T	SNV RIP (Q and C)d
C		InDelb		T		T	InDel RIP (Q)d
	C	InDel		T		T	InDel RIP (C)d
C		InDel		C		C	InDel 2:2
C		InDel		C	C		InDel 2:2
C		InDel			C	C	InDel 2:2
C		InDel	C				InDel 1:3 (deletion)
C		InDel					InDel 0:4 (deletion)
C		InDel	C	C		C	InDel 3:1 (NCO)
C		InDel	C	C	C	C	InDel 4:0 (2NCO)
C	A	SNV	A		A	C	1n:3c
C	A	SNV	A		A		2n:2c
C	A	SNV				C	3n:1c
C	A	SNV					4n:0c
C	A	SNV	A	C	G	C	IM-1e
C	A	SNV	A	C	G	T	IM-1e
C		InDel		C		G	IM-1e
C		InDel		T		G	IM-1e
C		InDel	C	C		G	IM-2f
C		InDel	C	T	C	G	IM-2f
			A				IM-3g
				A	G		IM-3g
			A	A	A		IM-3g
			A	A	T	G	IM-3g

^aSNVs are the positions where neither QM6a nor CBS999.97(MAT1-1) is null.

^bInDels are the positions where only QM6a or CBS999.97(MAT1-1) is null.

^cFor the 1n:3, 2n:2, 3n:1 and 4n:0 markers, neither QM6a nor CBS999.97(MAT1-1) is null. Moreover, 1n, 2n, 3n and 4n represents the numbers of F1 progeny with a null at the corresponding position, respectively.

^dRIP involves C:G-to-T:A transitional mutation in duplicated DNA. RIP (Q) and RIP (C) are QM6a-specific and CBS999.97(MAT1-1)-specific RIP, respectively.

^eIM-1 is the position where ≥1 F1 progeny harbors an authentic nucleotide that is different from the two corresponding nucleotides in QM6a and CBS999.97(MAT1-1), respectively. RIP is excluded.

^fIM-2 are positions where either QM6a and CBS999.97(MAT1-1) is null, but the corresponding positions contain an authentic nucleotide in ≥3 F1 progeny.

^gIM-3 are positions where both QM6a and CBS999.97(MAT1-1) are nulls, but the corresponding positions contain an authentic nucleotide in ≥1 F1 progeny.

Figure 2.

Flow chart depict the BLASTN-guided and sectional MAFFT algorithm in the “Tetrad" mode of TSETA.

Genome-wide methylation profiling

QM6a and CBS999.97(MAT1-1) were inoculated on 90 mm petri-dishes with malt extract agar (MEA) medium at 25°C until full asexual sporulation was observed (∼5 days). Conidial spores were then harvested and inoculated in 5 ml potato dextrose medium (PDB) in a rolling incubator at 30°C for 16 h. The germinated hyphae were harvested for isolation of genomic DNA (1). Bisulfite conversion of the genomic DNA was carried out using the EZ DNA Methylation-Gold™ Kit (Zymo Research, Orange, CA, USA). Unmethylated cl857 Sam7 Lambda DNA (catalog # D1521; Promega, USA) was spiked-in (0.5% w/w) to estimate the bisulfite conversion efficiency. The genomic DNA was then subjected to library preparation using the Accel-NGS Methyl-Seq DNA library kit (catalog # 30024; Swift Bioscience, USA). The libraries were sequenced using a NextSeq 500 sequencing system (Illumina, USA) as described previously (23). BatMeth2, an open source software program (https://github.com/GuoliangLi-HZAU/BatMeth2/) (24), was used for genome-wide DNA methylation calling.

RESULTS

TSETA is a BLASTN-guided and sectional MAFFT program

There are two reasons why it is difficult to directly apply the MAFFT (for Multiple Alignment using Fast Fourier Transform) program (25–27) or other publicly available multiple genome sequence alignment (MGSA) software tools [e.g. Kalign (28)] to achieve accurate and fast MGSA of different T. reesei strains. Firstly, the length (34 Mbp) of the T. reesei genome is relatively large, so it requires quite a lot of computational time and space to complete the task. For example, it took our computing server (with 20 GB of RAM) 4–7 days to apply MAFFT alone for aligning the six homologous chromosomes of QM6a, CBS999.97(MAT1-1), F_A, F_B, F_C and F_D, respectively. Secondly, T. reesei contains ∼2250 AT-rich blocks, each of ≥ 500 bp (1,6). These AT-rich blocks are highly homologous to each other, resulting in many false-positive or non-allelic alignment results.

These two technical problems were overcome using TSETA and within ∼20 h. TSETA first employs a GC content (or guanine-cytosine content) calculation algorithm using a window of 500 bp (1). The longest AT-rich blocks in each chromosome are the centromeres, whereas the most 5′ AT-rich block and the most 3′ AT-rich block are likely to be (but not necessarily) the 5′ and 3′ telomeric or subtelomeric sequences (Figure 2A). We reported previously that, except for the left terminus of chromosome VI, all other chromosomal ends in QM6a, CBS999.97(MAT1-1) and CBS999.97(MAT1–2) contain 8–13 telomeric repeats (i.e. TTAGGG at 3′-termini and the reverse complement CCCTAA at 5′-termini) (1,6).

Next, TSETA utilizes a newly invented algorithm to carry out fast and low-cost MGSA by combining the standard nucleotide Basic Local Alignment Search Tool (BLASTN) (29,30) and MAFFT (26,27). Initially, users upload genomic sequences in FASTA format to initiate TSETA. TSETA then applies BLASTN to search and scan local homologous regions in the genomes of QM6a, CBS999.97(MAT1-1), F_A, F_B, F_C and F_D, respectively. In brief, the first 10 000 bp of nucleotide sequence at the 5′ terminus of a selected QM6a chromosome was assigned as the seed query to search for ‘best-matched’, ‘allelic’ and ‘non-complementary’ sequences at the 5′ terminus (1–30 000 bp) of the corresponding chromosomes (Figure 2B). If the first BLASTN search fails (i.e. the best-matched sequence is <1000 bp, nonallelic or complementary), the seed QM6a query can be further extended 10 000 bp in the 3′ direction. BLASTN searches are repeated until the best matched allelic sequences can be found in CBS999.97(MAT1-1), F_A, F_B, F_C and F_D, respectively (Figure 2B). It is important to note that the seed query often (but not always) contains telomeric and/or subtelomeric sequences. In other words, neither telomeric nor subtelomeric sequence is a prerequisite for this newly invented algorithm for MGSA. This is important because not all chromosomal termini can be completely sequenced and assembled even using TGS technology. Also noteworthy is that sequences with high AT-contents are difficult to align properly using MAFFT alone. Accordingly, TSETA employs a longer subject (i.e. 30 000 bp) to initiate the BLASTN searches.

Next, the seed QM6a query and the five best-matched allelic sequences (along with their 5′ flanking sequences) are compared by MAFFT to identify the smallest Search End (SSE). SSE is the very last identical nucleotide in the seed QM6a query that can find a corresponding allelic nucleotide in all five subjects, i.e. CBS999.97(MAT1-1), F_A, F_B, F_C and F_D, respectively. This nucleotide in QM6a is referred to as the ‘Consensus Query End’ or ‘CQE’ (Figure 2B). For the next and all subsequent rounds of BLASTN searching, new QM6a query sequences are sequentially selected from ‘Search Start (SS)’ to ‘Search End (SE)’. SS is equal to ‘CQE + 1 bp’, whereas SE is ‘CQE + 10 000 bp’ or ‘CQE + the remaining unsearched nucleotide sequences’ located at the 3′ terminus of the selected ‘QM6a chromosome’ (Figure 2B).

We found that TSETA was able to accurately align the three inversion alleles on ChI (22,855 bp), ChIII (the MAT1–2 locus 36 490 bp and the MAT1-1 locus 37 189 bp) and ChV (59 633 bp), respectively (Figure 1). The MAT1–2 locus in QM6a contains one gene, mat1–2-1, whereas the MAT1-1 locus in CBS999.97(MAT1-1) has three genes, mat1–1-1, mat1–1-2 and mat1–1-3 (11) (Figure 3). In this case, the corresponding SE had been extended at least three times (i.e. from ‘CQE + 10,000 bp’ to ‘CQE + 40,000 bp’) before a new CQE was revealed by BLASTN.

Figure 3.

Pairwise comparison of the nucleotide sequences within and around the mating-type loci in QM6a and CBS999.97(MAT1-1). The tracks between two strains are color-coded to indicate nucleotide sequence identity. The mating-type genes (mat1–1-1, mat1–1-2, mat1–1-3 and mat1–2-1) are dissimilar in sequence but are found at the same loci on the third chromosomes and flanked by two evolutionarily conserved genes, the DNA lyase apn2 and the complex I intermediate-associated protein 30 gene cia30, respectively. Interestingly, the three mating-type genes (matA-1, matA-2, matA-3) in the MAT-A locus of Neurospora crassa are flanked by apn2 and the cytoskeleton assembly gene sla2.

TSETA enables comprehensive sequence alignment and comparative analyses

Next, MAFFT was applied sequentially to align all allelic sequences resulting from each round of BLASTN searches. We found that nearly all the best-matched allelic sequences had an expect (e) value of ‘0’ throughout the entire BLASTN-guided and sectional MAFFT process, except for centromeres and the large rDNA loci (see below). TSETA then compiles the MGSA results telomere-to-telomere and displays the assembled results on a computer screen (Figures 4 and 5).

Figure 4.

The TSETA Viewer allows easy and intuitive visualization of both meiotic recombination and RIP mutation events on a computer screen at scales ranging from the full-length chromosome landscape (A), 100-fold magnification (B), 5000-fold magnification (C), 50 000-fold magnification to single-nucleotide levels (D). The first two horizontal rows of sequence data represent the full-length sequences of the first chromosomes (ChI) of QM6a (in cyan) and CBS999.97(MAT1-1) (in magenta). The next four horizontal rows of sequence data represent full-length ChI of the four representative F1 progeny, respectively. Nucleotide sequences identical to those of parental QM6a and CBS999.97(MAT1-1) are also indicated in cyan and magenta (A–C) or in pale turquoise and pink (D), respectively. Some nucleotides in the four representative F1 progeny are indicated in gray, because they are identical to the corresponding nucleotides in both QM6a and CBS999.97(MAT1–1), with difficult to assign nucleotides in cyan or magenta, respectively. The gapped (deletion) regions are colored ‘white’. COs are located where 2:2 markers undergo a reciprocal genotype change. The GC tracts associated with a CO event are labeled with ‘red’ bars. The NCO products show either 3:1 or 4:0 segregation. Different genetic markers in F1 (Table 1) are marked by bars in a variety of colors. At the single-nucleotide level, all non-2:2 markers or the 2:2 markers located next to the boundaries of COs are highlighted in ‘cyan’ and ‘magenta’, respectively (D).

Figure 5.

Intuitive visualization of the large rDNA loci. (A) The ChVI chromosomes of QM6a, CBS999.97(MAT1-1) and the representative F1 progeny are shown as described in Figure 4. The large internal deletion (57 946 bp) within the ChVI centromere of the second F1 representative progeny is marked by a vertical black arrow. (B) The large rDNA loci are shown as an array of red arrows, with each red arrow representing an 18S-5.8S-26S rRNA gene cluster.

To make data interpretation easier, all gapped positions in the six chromosomes are automatically filled with a hyphen symbol or null (-). If there is a null (-) in QM6a or CBS999.97(MAT1-1) but not both, the gapped position will be considered as an InDel by TSETA. Some gapped regions have authentic nucleotides in both QM6a and CBS999.97(MAT1-1), but the corresponding positions in the four F1 progeny show 1n:3, 2n:2, 3n:1 or 4n:0 segregation, where 1n, 2n, 3n and 4n represent the numbers of F1 progeny with a null (-) at a selected nucleotide position (Table 1). These four markers (i.e. 1n:3, 2n:2, 3n:1 and 4n:0) are likely due to sequencing or assembly errors rather than non-mendelian segregation. TSETA also identifies and categorizes three different types of IMs (Table 1).

It is also important to note that the original coordination of all nucleotide sequences in the six chromosomes remain labeled by TSETA. Together, these ‘filling-in’ and ‘memory’ functions of TSETA allow users to comprehensively describe the MGSA results during subsequent structural and functional analyses.

Alignment of centromeres and large rDNA loci is problematic

There is an inevitable limitation of the BLASTN-guided and sectional MAFFT algorithm we have proposed here. This approach was unable to accurately align the nucleotide sequences in centromeres and those of large ribosomal DNA (rDNA) because AT- or repeat-rich sequences make up the major part of these two chromosomal regions. As a result, the best-matched sequences resulting from the corresponding BLASTN searches were often non-allelic. Centromeres are the most prominent or longest AT-rich blocks in each chromosome of QM6a and CBS999.97(MAT1-1), ranging from 150 kb up to 290 kb. Due to long-term geographic isolation, the two corresponding centromeres in each chromosome of QM6a and CBS999.97(MAT1-1) also show high levels of intraspecific sequence heterogeneity (6). Moreover, we reported previously that the seven QM6a centromeres collectively harbored 24 conserved sequences (≥90% identity; maximum length 8625 bp and minimum length 4847 bp) with a copy number per chromosome ranging from one to five (1). Interestingly, some of these centromeric sequences share significant sequence homology (∼65% identity and expect values = e⁻²³ to e⁻³⁵) to Foret1 (for Fusarium oxysporumretroid element), a repeated DNA sequence in the filamentous fungus Fusarium oxysporum. The overall structure of Foret1 is homologous to the ‘gypsy’ class of LTR-retrotransposons (31). Similar sequences have been found not only in all CBS999.97(MAT1-1) centromeres, but also in several other non-centromeric AT-rich blocks in the genomes of QM6a and CBS999.97(MAT1-1) (Li W.C., Liu H.C. and Wang T.F., unpublished results). The large rDNA loci of QM6a and CBS999.97(MAT1-1) reside in the right arm of chromosome ChVI, harboring 9 and 11 tandem ‘head-to-tail’ repeats, respectively. Each repeat (∼7.8 kb) contains an 18S-5.8S-26S rRNA gene cluster and a non-transcribed intergenic spacer (IGS) (1). Collectively, these characteristics explain why we chose a query length of 10 000 bp for each BLASTN search. In general, the shorter the query length, the faster TSETA can complete the task, but with more non-allelic or even ectopic alignment results.

We overcame this difficulty by developing new alignment strategies. In short, the nucleotide sequences of centromeres and the large rDNA locus were automatically extracted via a user-friendly and intuitive input interface. Pairwise BLASTN searching only detected parental alleles, indicating that only two of the four representative F1 progeny inherited the same nucleotide sequences from either the paternal allele or the maternal allele (Figure 4). In other words, their nucleotide sequences are highly similar or even identical to either those in QM6a or those in CBS999.97(MAT1-1). Next, we used a JavaScript program to perform best-matched multiple sequence alignment, gap filling and visualization of the corresponding sequences in either the paternal QM6a genome or the maternal CBS999.97(MAT1-1) genome as well as their two closest F1 successors (Figure 5A). These results are consistent with previous reports that centromeres (32,33) and the large rDNA loci (34) are devoid of reciprocal recombination (crossing-over) between homologous chromosomes. Since the BLASTN-guided and sectional MAFFT algorithm failed to align the nucleotide sequences in QM6a centromeres to those of the corresponding CBS999.97(MAT1-1) centromeres, it is likely that nucleotide sequence heterogeneity between the two corresponding centromeres in QM6a and CBS999.97(MAT1-1) is too high to allow ‘allelic’ homology-directed recombinational repair. TSETA ends up categorizing and displaying centromeres and several longer AT-rich blocks as ‘strain-specific’ or solitary sequences (Figure 5A). As a result, we found no ‘allelic’ NCO or 2NCO within all seven centromeres of the four F1 representative progeny we examined here (Figures 4 and 5). Interestingly, it was reported previously that there are widespread genetic exchanges in maize in the form of gene conversion between CRM2-based centromeric markers (32). CRM2 is a highly enriched retrotransposon in maize centromeres and exhibits very low transposition rates. Since those CRM2-dependent gene conversions were detected by using a PCR-based method called transposon display (32), they might arise from both allelic and non-allelic or even ectopic homology-directed recombinational repair. Thus, we acknowledge that TSETA might lack the capability to detect non-allelic or ectopic gene conversion.

Notably, we observed a large internal deletion (>54 kb) in the ChVI centromere of the second of the representative F1 progeny. It is not clear whether this large internal deletion occurred during meiosis or the two rounds of postmeiotic mitosis. This large internal deletion might also result from incorrect chromosome assembly due to the high AT content and/or the repetitive nature of these sequences (Figure 5).

The copy numbers of the 18S-5.8S-26S rRNA gene cluster in F_A, F_B, F_C or F_D were often not completely identical to that of parental QM6a or CBS999.97(MAT1-1), probably due to intrachromosomal or unequal sister chromatid recombination [see review of (35)], nonhomologous end joining (34), as well as IMs, TGS sequencing or genome assembly errors. Accordingly, we incorporated a new function into TSETA to display the large rDNA loci as an array of red arrows. Each red arrow represents a single 18S-5.8S-26S rRNA gene cluster (Figure 5B). There might be some gene conversion events between the large rDNA loci of F_A, F_B, F_C and F_D (Figure 5B), but almost all these gene conversions are associated with the InDel genetic markers, including InDel 0:4, InDel 1:3, InDel 3:1 or InDel 4:0, respectively.

Accuracy assessment of TSETA

Three F1 genetic makers (i.e. 1n:3, 2n:2, IM-1; Table 1) are not caused by gene conversion, but instead they originate from TGS sequencing errors or from DNA replication errors during premeiotic S phase, post-meiotic mitosis or the post-germination growth phase. Therefore, the sum of these three markers can be used to estimate the overall quality of the six near-complete TGS genome sequences we determined in this study. Since there are differences in the 18S-5.8S-26S rRNA gene cluster copy number in QM6a, CBS999.97(MAT1-1), F_A, F_B, F_C and F_D, respectively, we estimate the total error rate of TSETA to be <0.0344% [(10 552 + 650 + 521)/34 139 199] or <0.0110 [(2681 + 650 + 402)/34 139 199] when the large rDNA loci are included or excluded, respectively. The overall genome size of QM6a is 34 139 199 bp, whereas the sums of these three markers with or without the large rDNA loci are 11 723 and 3733 bp, respectively. We reported previously that the PacBio SMRT sequencing error rate of the near-complete QM6a genome sequence was <0.0024% because the mitochondrial genome sequence of QM6a is completely identical to that previously reported (1).

In contrast, the overall number of four different markers (i.e. 3n:1, 4n:0, IM-2 and IM-3; Table 1) allowed us to determine the accuracy of the TSETA alignment algorithm (Table 2). Therefore, the sum of these three markers can be used to estimate the overall quality of the six near-complete TGS genome sequences we determined in this study. We estimate the total error assembly rate between these six near-complete genome sequences to be ∼0.0725% [(15 + 14 + 91 + 24 606)/34 139 199] or ∼0.0510% [(15 + 14 + 21 + 17 327)/34 139 199] when the rDNA loci are included or excluded, respectively. Collectively, we conclude that TSETA is a highly accurate tool for MGSA.

Table 2.

Summary of SNPs in the two parental genomes and genetic markers in the four F1 representative genomes detected by TSETA

Genomes	Chromosome	I	II	III	IV	V	VI (w/o large rDNA locusa)	VII	Total (w/o large rDNA locusa)
QM6a and CBS999.97(MAT1-1)	SNV	212365	13506	154861	103825	139638	131366 (130541)	131866	1011427 (1010602)
	InDel	895019	1339626	679705	1211117	589213	927569 (909572)	722107	6364383 (6346359))
	SNP	1107384	1477132	834566	1314942	728851	1058962 (1040113)	853973	7375810 (7356961)
Four representative F1 progeny FA, FB, FC and FD	SNV 2:2	212044	137460	154784	103782	139578	130480 (129734)	131608	1009736 (1008990)
	simple CO	2	1	1	1	1	1 (1)	1	8 (8)
	CO + NCO	3	2	2	2	3	2 (2)	2	16 (16)
	SNV 1:3 or 3:1 (NCO)	17	12	42	7	16	19 (15)	24	137 (133)
	SNV 4:0 (2NCO)	2	6	4	5	5	35 (27)	0	57 (49)
	SNV RIP (Q)	269	5	2	2	4	87 (87)	229	598 (598)
	SNV RIP (C)	3	0	2	4	1	6 (6)	3	19 (19)
	SNV RIP (both Q and C)	0	0	0	0	0	0	0	0
	InDel RIP (Q)	65	56	17	865	67	1391 (1391)	14	2475 (2475)
	InDel RIP (C)	9	15	10	13	4	8 (8)	14	73 (73)
	InDel 2:2	894062	1338780	678945	1209649	584665	832222 (823482)	721580	6259903 (6251163)
	InDel 1:3 (deletion)	577	408	330	382	4318	72 799 (64379)	260	79 074 (70654)
	InDel 0:4 (deletion)	159	286	203	113	88	20 213 (20160)	110	21 172 (21119)
	InDel 3:1 (NCO)	57	25	92	36	28	50 (45)	49	337 (332)
	InDel 4:0 (2NCO)	52	35	48	22	28	684 (28)	26	895 (239)
	1n:3	409	359	395	316	358	8473 (602)	242	10552 (2681)
	2n:2	6	2	8	1	1	631 (631)	1	650 (650)
	3n:1	4	1	3	0	5	2 (2)	0	15 (15)
	4n:0	0	0	3	1	9	0 (0)	1	14 (14)
	IM-1	51	23	66	36	35	256 (137)	54	521 (402)
	IM-2	0	5	3	3	0	79 (9)	1	91 (21)
	IM-3	446	645	340	357	324	22278(14999)	216	24606 (17327)

^aThe genetic markers of the large rDNA loci are included or excluded (w/o) due to the copy number differences of the 18S-5.8S-26S rRNA gene cluster in QM6a, CBS999.97(MAT1-1), F_A, F_B, F_C and F_D, respectively.

TSETA is a versatile tool for genome-wide variant calling

One of our goals in developing TSETA was to accurately and comprehensively identify SNVs and Indels among the near-complete genome sequences of any two or more intraspecific strains of T. reesei. We considered SNVs as selected positions exhibiting different nucleotides (A, T, G or C) in QM6a and CBS999.97. Indels were defined as nucleotide sequences specific to QM6a or CBS999.97(MAT1-1), with the corresponding nucleotides in the four F1 progeny often (but not always) showing 2n:2 segregation.

The “Tetrad" mode of TSETA detected 1 011 427 SNVs and 6 364 383 Indels between QM6a and CBS999.97(MAT1-1) (Table 2). In contrast, the NGS-based SNP calling method [i.e. MUMmer (2)] only revealed 604 578 SNVs and 331 491 Indels between QM6a and CBS999.97(MAT1-1) (6) as strain-specific and repeated sequences were neglected. Therefore, TSETA is about ∼7.88 times more powerful than MUMmer [7 375 810 SNPs (1 011 427 SNVs and 6 364 383 Indels) versus 936 069 SNPs (604 578 SNVs + 331 491 Indels), respectively] in revealing intraspecific sequence diversity between genetically distinct strains or isolates.

Genome-wide identification of interhomolog recombination products, RIP and IMs

TSETA is versatile, possessing two powerful programs (‘Recombine Analyzer’ and ‘RIP Analyzer’) for genome-wide detection of CO, NCO, RIP and IM in each tetrad. The strategy of ‘Recombine Analyzer’ is similar to those employed by ‘Recombine (v2.1)’, a suite of programs originally designed solely for the detection and analysis of meiotic recombination in the budding yeast Saccharomyces cerevisiae (36). Initially, all markers in a tetrad are categorized according to whether or not they exhibit 2:2 (but not 2n:2) segregation. Next, the 2:2 markers are searched for locations where markers undergo a reciprocal genotype change, representing the locations of COs. The program then determines if any of 3:1 and/or 4:0 markers (collectively referred to as ‘non-2:2 markers’) fall within the boundaries of the COs. If they do, they are considered gene conversions associated with a CO (termed ‘CO+NCO’) but, if they do not, they are referred to as ‘simple CO’. Then, non-2:2 markers not associated with COs are deemed to be NCOs (only 3:1, but not 1n:3 or 3n:1) or 2NCOs (only 4:0, but not 4n:0) (Tables 2–4).

Table 4.

Summary of all NCO products (≥2 3:1 and/or 4:0 markers) detected by TSETA

Chromosome number	Chromosome Length (bp)	Position (bp)	GC tract length (bp)	GC tract marker	snp_start_out	snp_start_in	snp_end_in	snp_end_out
1	6835803	2841837	168	6	2841732	2841773	2841800	2842041
1	6835803	4349992	163	5	4349909	4349911	4349918	4350228
2	6234656	1379384	166	3	1379241	1379360	1379405	1379528
2	6234656	4817690	68	2	4817606	4817707	4817710	4817738
2	6234656	4955357	19	5	4955339	4955355	4955360	4955372
3	5311445	1308219	339	27	1308011	1308088	1308375	1308402
3	5311445	1581203	116	2	1581077	1581213	1581214	1581307
3	5311445	2314738	11	4	2314730	2314734	2314742	2314744
3	5311445	3935255	1041	19	3934647	3934821	3935580	3935970
3	5311445	4101515	385	9	4101219	4101426	4101619	4101796
4	4556834	1363135	51	3	1363073	1363147	1363155	1363166
4	4556834	1790509	3	2	1790506	1790509	1790510	1790511
4	4556834	1790513	3	2	1790511	1790513	1790514	1790515
4	4556834	3558827	9	3	3558820	3558825	3558831	3558832
4	4556834	3558857	9	3	3558851	3558854	3558857	3558865
4	4556834	3914233	5	2	3914230	3914232	3914235	3914236
5	4159965	7	11	4	1	1	7	17
5	4159965	783456	170	2	783355	783388	783475	783607
5	4159965	2906230	142	3	2906100	2906219	2906235	2906367
6	4000387	16382	33	3	16360	16371	16379	16417
6	4000387	2042371	759	2	2041331	2042652	2042737	2042764
6	4000387	2345006	171	2	2344776	2345065	2345066	2345116
6	4000387	2377073	99	2	2377012	2377035	2377040	2377205
6	4000387	2383990	146	2	2383772	2384061	2384062	2384063
6	4000387	3900903	85	2	3900794	3900927	3900944	3900946
6	4000387	3901276	633	3	3900959	3900960	3900968	3902216

Table 3.

Summary of all CO products detected by TSETA

Chromosome number	Chromosome Length (bp)	Position (bp)	GC tract length (bp)	GC tract marker	snp_start_out	snp_start_in	snp_end_in	snp_end_out	Type
1	6835803	290545	313	5	290360	290417	290660	290742	CO(NCO)
1	6835803	1376407	278	5	1376220	1376316	1376462	1376630	CO(NCO)
1	6835803	2846616	192	0	2846520	2846520	2846712	2846712	CO
1	6835803	4850122	164	0	4850040	4850040	4850204	4850204	CO
1	6835803	5942633	45	1	5942600	5942621	5942621	5942690	CO(NCO)
2	6234656	1078315	265	5	1078135	1078229	1078313	1078581	CO(NCO)
2	6234656	3067878	287	6	3067677	3067793	3067982	3068061	CO(NCO)
2	6234656	4581277	273	0	4581140	4581140	4581413	4581413	CO
3	5311445	389221	78	1	389129	389234	389234	389285	CO(NCO)
3	5311445	1639304	149	1	1639199	1639259	1639259	1639497	CO(NCO)
3	5311445	2230879	438	0	2230660	2230660	2231098	2231098	CO
4	4556834	823739	75	0	823701	823701	823776	823776	CO
4	4556834	1922011	290	6	1921803	1921929	1922088	1922223	CO(NCO)
4	4556834	3937471	185	1	3937229	3937528	3937528	3937598	CO(NCO)
5	4159965	906317	84	1	906203	906347	906347	906371	CO(NCO)
5	4159965	1651362	137	2	1651260	1651328	1651378	1651483	CO(NCO)
5	4159965	2640861	176	1	2640737	2640809	2640809	2641088	CO(NCO)
5	4159965	3965025	349	0	3964850	3964850	3965199	3965199	CO
6	4000387	640821	312	6	640643	640688	640864	641090	CO(NCO)
6	4000387	2245856	476	0	2245618	2245618	2246094	2246094	CO
6	4000387	3266023	167	1	3265860	3266018	3266018	3266194	CO(NCO)
7	3823438	1180772	373	0	1180585	1180585	1180958	1180958	CO
7	3823438	2891266	661	6	2890900	2890971	2891551	2891642	CO(NCO)
7	3823438	3411537	251	15	3411387	3411435	3411632	3411692	CO(NCO)

‘Recombine Analyzer’ ignores all markers harboring hyphen symbols (1n:3, 2n:2, 3n:1 and 4n:0), IM1, IM2 and IM3 to avoid misidentification of false-positive CO and NCO products. Remember, the hyphen symbols were deliberately incorporated by TSETA to create user-friendly and comprehensive graphical representations of all six chromosomes.

‘RIP Analyzer’ detects the products of RIP. A hallmark of RIP is that it occurs prior to meiosis, thereby only affecting the two sister chromatids but not the two non-sister homologous chromosomes in the same zygote [reviewed in (13–15)]. One of the most difficult challenges in genome analysis is to faithfully distinguish a RIP event from a 2NCO event (with either a SNV or InDel 4:0 marker) in which all the nucleotides of the four F1 progeny are either all ‘T’ or ‘A’ but the two parental nucleotides are ‘T’ and ‘C’ or ‘A’ and ‘G’, respectively. Since RIP occurs prior to the formation of meiotic interhomolog recombination products, the 2NCO events with C:G to T:A mutations will not be considered as RIP if they locate between two adjacent CO and/or NCO events. Accordingly, we managed to integrate ‘RIP Analyzer’ with ‘Recombine Analyzer’ in TSETA, so that the nucleotides in the two corresponding sister chromatids that have undergone RIP can be accurately recognized according to the segregation pattern of their neighboring nucleotides in all four F1 progeny (Figure 2).

Notably, more RIP events were detected in the genome of QM6a [3.073 nt (598 SNV RIPs + 2475 InDel RIPs)] than in that of CBS999.97(MAT1-1) [92 nt (19 SNV RIPs and 73 InDel RIPs)] (Table 2 and Figures 3–5). This result is not surprising for two reasons. First, the CBS999.97(MAT1-1) strain we used here was previously derived from an ascospore from a fruiting body generated by sexual crossing of CBS999.97(MAT1-1) with CBS999.97(MAT1-1) (11), indicating that its genome had undergone at least one round of RIP. Secondly, QM6a is female sterile and has been propagated vegetatively for more than 70 years. Accordingly, the QM6a genome has accumulated more duplicated and RIP-prone sequences than that of CBS999.97(MAT1-1) (6). For example, a long RIP-prone sequence in ChVI (length: 47 420 bp; location: 1 155 679−1 203 098 bp; indicated by a square box in Figure 6A) is highly similar in nucleotide sequence (E-values = 0 and identities ≥77%) with three different long DNA segments in QM6a, including ChIV (4 288 615−4 344 503 bp), ChV (399 786−447 005 bp) and ChVI (993 647−10 457 763 bp), respectively. These three DNA segments in QM6a have underwent RIP during sexual crossing with CBS999.97(MAT1-1). Interestingly, the genome of CBS999.97(MAT1-1) does not possess similar or homologous sequence. It is not clear how QM6a acquired these long repetitive sequences during evolution.

Figure 6.

Intuitive visualization of genome-wide 5-methylcytosines in vegetative mycelia of QM6a and CBS999.97(MAT1-1). The ChVI chromosomes of QM6a, CBS999.97(MAT1-1) and the four representative F1 progeny are shown as described in Figure 4. Genomic DNA was isolated from vegetative mycelia of QM6a and CBS999.97(MAT1-1), respectively, and then subjected to whole genome bisulfite sequencing (see ‘Materials and Methods’ section). Genome-wide levels of 5-methylcytosines were calculated using BatMeth2 (24). The two bottom panels depict the 5-methylcytosine levels along the ChVI chromosomes of vegetative mycelia of QM6a and CBS999.97(MAT1-1), respectively. A long RIP-prone region in chromosome VI of QM6a is indicated by a square box.

Our results also reveal that the overall frequencies of 2NCO (i.e. 57 SNV 4:0 markers and 895 InDel 4:0 markers) are lower and higher than those of RIP in QM6a and CBS999.97(MAT1-1), respectively (Figures 4 and 5 and Table 2). Locally, few RIP-induced C:G-to-T:A mutations were detected in the centromeres of the four F1 progeny (Figures 4 and 5). No RIP event was detected within the large rDNA loci of the four F1 progeny (Table 2 and Figure 5A).

TSETA is a powerful tool for global and local visualization of sequence variants

We designed a high-performance viewing interface for TSETA to provide an intuitive user experience for all levels of genome resolution. It has a zoom function (≥10⁵-fold magnification) that can be manually controlled by turning a mouse wheel. The TSETA viewer allows users to conduct instantaneous and continuous visualization of all six near-complete chromosomes from the scale of the full-length chromosome landscape to individual nucleotides (Figure 4).

For global visualization, regions with sequences identical to QM6a and CBS999.97(MAT1-1) are colored ‘magenta’ or ‘cyan’, respectively. Gapped (deletion) regions are colored ‘white’. The red bars and blue bars mark CO and NCO, respectively, whereas black bars mark RIP. All the genetic markers depicted in Table 1 are also displayed in the TSETA interface (Figures 4 and 5).

For local nucleotide sequence visualization, QM6a-like and CBS999.97(MAT1-1)-like sequences are colored ‘pale turquoise’ and ‘pink’, respectively. All non-2:2 markers and the 2:2 markers located next to the boundaries of COs are also highlighted in local nucleotide sequences (in ‘cyan’ and ‘magenta’, respectively) (Figure 4–6).

TSETA is convenient for integration and visualization of 5-methylcytosines revealed by NGS-based bisulfite sequencing

RIP was thought to be mediated by 5-methylcytosines [reviewed in (13,14)]. Here, we also combined a NGS-based bisulfite sequencing method and TSETA to map and visualize 5-methylcytosines in the genomic DNA isolated from vegetative mycelia of QM6a and CBS999.97(MAT1-1), respectively (Figure 6). Using a control unmethylated phage DNA (see ‘Materials and Methods’ section), the bisulfite conversion efficiencies of QM6a and CBS999.97(MAT1-1) were determined to be 97.5%. Next, genome-wide DNA methylation calling was carried out using the publicly available software BatMeth2 (24). The genome-wide ratios of 5-methylcytosines versus the sum of 5-methylcytosines and cytosines were then visualized in TSETA. We found that the 5-methylcytosine components in both genomes consisted of relics that are highly AT-rich or RIP-prone (Figure 6). Most of these AT-rich relics are categorized and displayed as ‘strain-specific’ sequences in TSETA, so their 5-methylcytosine profiles can be visualized in isolation (Figure 6). This additional function of TSETA provides a powerful new means for comparative analyses of the impacts of 5-methylcytosine and strain-specific sequences during sexual development, such as RIP and meiotic silencing by unpaired DNA (MSUD). MSUD is a RNA interference (RNAi) process whereby DNA unpaired during meiosis causes silencing of all DNA homologous to it, including genes that are themselves paired (37,38).

DISCUSSION

Single-nucleotide-resolution MGSA (rather than SNP calling) is the main strategic difference that distinguishes TSETA from other NGS-based software packages previously developed for genome-wide detection of meiotic recombination products. As long as a hybrid meiosis with large numbers of heterozygous variant markers is established, there are only two prerequisites to completing the tasks assigned to TSETA. First, high-quality and chromosome-level genome sequences involved in a meiosis event are necessary. Given the rapid development of TGS platforms (e.g. PacBio SMRT and Nanopore), obtaining such sequences is now feasible and affordable. Second, a fast, accurate and low-cost computational approach for MGSA of all complete haploid genomes involved in a meiotic event is required. We overcame inherent challenges with genomic analyses by combining in TSETA the power of the BLASTN-guided and sectional MAFFT algorithms.

Our results demonstrate that TSETA will prove to be a faster and timely tool for the filamentous fungal community for the following reasons. Firstly, TSETA outcompetes nearly all NGS-based sequencing technology and alignment software tools because it takes into account all SNPs in a hybrid zygote. Secondly, TSETA sequentially identifies all ‘allelic’ homologous sequences with a size greater than 1 kb in the six corresponding homologous chromosomes before and after a single meiotic event, including those in QM6a, CBS999.97(MAT1-1), F_A, F_B, F_C and F_D, respectively. These allelic DNA sequences are then used as anchors to carry out multiple genome sequence alignment (MGSA) and comparison of all six high-quality genomes. Subsequently, TSETA can identify and visualize DNA nucleotide sequence changes or even modifications during sexual development. Thirdly, TSETA is a powerful system to identify and visualize all highly variable, non-allelic or repetitive sequences (e.g. centromeres, AT-rich blocks and rDNA) before and after meiosis, as well as the solitary sequences not found in maternal or paternal genomes but that are present in F_A, F_B, F_C or F_D, respectively. Finally, our computing server (with 20 GB of RAM) only took ∼20 h for TSETA to complete all the tasks described here.

In this study, we have demonstrated that TSETA is also an appropriate tool for visualizing the 5-methylcytosines revealed by NGS-based bisulfite conversion. Interestingly, both PacBio SMRT and Nanopore sequencers allow direct detection of modified nucleotides in the DNA templates, including N6-methyladenine, 5-methylcytosine and 5-hydroxymethylcytosine (39,40), all of which can be further analyzed and visualized in TSETA. Therefore, this new MGSA algorithm is applicable to genome-wide identification, visualization and comparative analyses of many other DNA variants involved in various biological processes. For example, our results reveal that 5-methylcytosines are highly enriched in the RIP-prone regions in the genomes of QM6a and CBS999.97(MAT1-1) during vegetative growth. Presumably, these 5-methylcytosines will not have been converted to thymidines because RIP only occurs premeiotically (Figure 6). It is noteworthy that the GC content of a long RIP-prone region in chromosome VI of QM6a (indicated by a square box in Figure 6A) is higher (>25%) than those of AT-rich and gene-poor compartments, but lower (<10%) than the average GC content of the entire QM6a genome. Further investigations will reveal what structural features (i.e. duplicated or repetitive sequences and/or their GC contents) and genetic determinants are critical for the formation of 5-methylcytosines in the RIP-prone regions during vegetative growth.

In conclusion, TSETA is a fast and accurate software program that allows all researchers, no matter their level of bioinformatics experience, to perform genome-wide SNP calling among high-quality genome sequences of any two or more intraspecific strains. TSETA is also a powerful tool for genome-wide identification of interhomolog recombination product, RIP and IM before and after meiosis. Using T. reesei QM6a/CBS999.97(MAT1-1) hybrid meiosis as a model, we demonstrate that TSETA can complete analytical tasks regardless of the high contents of long and AT-rich sequences in T. reesei genomes. In principle, TSETA could be applied to many other sexual eukaryotes, including other Pezizomycotina fungi (e.g. Neurospora crassa) and those with high numbers of repetitive sequences and transposable elements (e.g. Basidiomycota fungi, higher plants and animals). For organisms with unique reproductive strategies, some modifications might be needed. For example, in many Basidiomycota, the dikaryotic state is maintained by clamp connections that help reestablish the pairs of compatible nuclei after synchronous mitotic divisions. During sexual development, the two nuclei fuse in a basidium and form a diploid zygote. Haploid spores (basidiospores) are then produced in the basidium. Thanks to rapid improvements of TGS and sequence alignment technology, unzipping haplotypes in diploid or even polyploid genomes is a growing trend of recent genome studies (41). Alternatively, sexually competent monokaryonic mycelia of many Basidiomycete species can be obtained via germination of either basidiospores or arthrospores. Arthrospores are conidia or asexual spores that are produced simply by the last cell on a dikaryonic hypha breaking off and dispersing as a propagule. Accordingly, the genome sequences of monokaryonic mycelia can be readily determined by TGS technology. In both scenarios, TSETA can be a versatile tool for genome-wide analysis of organisms with diploid or even polyploid zygotes.

DATA AVAILABILITY

The reference genomes used in this study were previously determined using the PacBio SMRT sequencing and assembly platforms: T. reesei QM6a (GenBank accession number: CP016232-CP016238) (1), CBS999.97(MAT1-1) (GenBank accession numbers: CP017983-CP017984 and CP020875-CP020879), and their four representative F1 progeny (GenBank accession numbers: CP021290-CP021317) (6). All of these complete genome sequences have been submitted to the National Center for Biotechnology Information (NCBI) under three BioProjects (PRJNA325840, PRJNA352653 and PRJNA386077). The NGS-based bisulfite sequencing results have also been submitted to NCBI under a BioProject (PRJNA635549).

SOFTWARE AVAILABILITY AND REQUIREMENTS

Project name: TSETA

Project home page: http://github.com/labASIMBTFWang/TSETA

Operating system: Linux.

Programming languages: ECMAScript.

Other requirements: BLASTN, MAFFT and Node.js.

License: GPL-3.0

Any restrictions to use by non-academics: License required.

FUNDING

Academia Sinica, Taiwan, Republic of China [AS-105-TP-B07, AS108-TP-B07 to T.F.W.]; Ministry of Science and Technology, Taiwan, Republic of China [MOST163-2311-B-001-016-MY3 to T.F.W.].

Conflict of interest statement. None declared.