Relatively little is known regarding the genetics of F. circinatum, with the bulk of knowledge at this level relating to its phylogeny and diagnostics,^6,7 as well as to its population biology.^8,9 Previous studies have, for example, shown that F. circinatum is a heterothallic fungus capable of both sexual and asexual reproduction. Unlike many other Ascomycete pathogens, sexual and asexual reproduction of F. circinatum have been shown in regions of the world where F. circinatum has been introduced relatively recently.^10,11,12 Furthermore, studies have also shown that the fungus probably originated in Mexico or Central America and that it has been accidently introduced into pine-growing regions around the world.¹³ In all cases, however, these previous DNA-based studies have utilised information from either housekeeping loci or microsatellite-rich regions, which in most cases represent small or limited portions of the pathogen’s genome.

Whole-genome analysis procedures such as genetic linkage mapping and genome sequence comparisons have increased our understanding of the genetic basis of various biological phenomena in fungi. Well-known examples include the development of spores in Pleurotus pulmonarius¹⁴ and the development of ectomycorrhizal symbiosis in Laccaria bicolor.¹⁵ Such whole-genome approaches have also shed light on the evolution of fungal pathogenicity,^16,17 which has also been particularly true for Fusarium species such as Fusarium oxysporum, Fusarium verticillioides and Fusarium graminearum.^18,19 The fact that the genomic data for these Fusarium species are in the public domain, and that a framework map is available for F. circinatum,²⁰ therefore presents ideal opportunities to understand the genetic basis for pathogenicityin the pitch canker fungus.

The aim of this study was to sequence, assemble and annotate the genome of F. circinatum. In addition, we present a preliminary analysis of putative gene clusters that are unique to F. circinatum and we compare three loci of this genome with the genomes of three close relatives: F. oxysporum, F. verticillioides and F. graminearum (Figure 2). From a South African perspective, this study will have significant impact – not only because the pitch canker pathogen is the first eukaryotic organism for which the entire genome has been sequenced in Africa, but also because the project strongly promotes human capacity development in the field of genome sequencing on the African continent. Furthermore, data emerging from this sequence will promote many studies concerning the pathogen and potentially lead to innovating approaches to reduce the losses that the pathogen is causing in South Africa and elsewhere in the world.

FIGURE 1: Symptoms of pitch canker caused by Fusarium circinatum on Pinus radiata near Cape Town, South Africa: (a) dying trees in a plantation, (b) a dead branch in the crown of an infected tree, (c) stem canker resulting from infection of a pruning wound and (d) resin (pitch) impregnated wood.

FIGURE 2: Evolutionary relationships amongst the Fusarium species for which whole-genome sequence information is available.

In order to confirm the integrity of the assembled F. circinatum genome, we interrogated the assembly for the presence and order of the open reading frames (ORFs) known to be encoded at the mating type (MAT) locus of this fungus. From previous research it is known that the mating type of F. circinatum isolate FSP34 is MAT 1.^20,26 Within the F. circinatum assembly we thus expected to find three MAT-1 ORFs (MAT 1.1.1, MAT 1.1.2 and MAT 1.1.3) and the entire region to be flanked by genes encoding a cytoskeleton assembly control protein (SLA1) and a DNA lyase (APN1).^25,27 Local Basic Local Alignment Search Tool (BLAST) analysis of the assembly indicated that a single contig (Contig00012) contained MAT-1 sequences. Examination of this 25 000 bp contig confirmed the presence of the genes, in both the same orientation and order as those found in other Fusarium species (Figure 3). This process of verification was repeated on two additional contigs (data not shown) containing genes that were of interest and also confirmed the accuracy of the assembly.

The completeness of the F. circinatum genome sequence was determined by subjecting the sequence to the CEGMA (Core Eukaryotic Genes Mapping Approach) pipeline.²⁸ A defined set of conserved protein families known to occur in all eukaryotes was used for the analysis.²⁸ This procedure also allows for the production of an initial set of reliable gene annotations in a eukaryotic genome, even in a draft form. The analysis revealed that the F. circinatum genome sequence assembly included the large majority (95%) of the genes common to other eukaryotes. The assembled F. circinatum genome was thus at least 95% complete. Future studies will seek to verify whether the missing genes are indeed not encoded by the pitch canker fungus.

Fusarium circinatum is a haploid fungus and the isolate sequenced was established from a single spore. Therefore, as opposed to diploid or polyploid organisms, only a single allele would be found at any particular locus in the genome. This simplifies the genome assembly process for haploid species, which generally requires less sequence coverage to produce an accurate assembly. Based on the estimated size of the F. circinatum genome and the amount of sequence information generated, an 11X sequence coverage was obtained. We were, therefore, confident that the genome of F. circinatum had been sequenced close to completeness and that the accuracy and integrity of the assembly was as good as could reasonably be expected.

FIGURE 3: Diagrammatic comparison of the mating type (MAT) locus in Fusarium circinatum, F. verticillioides, F. oxysporum and F. graminearum, based on the available genome sequences.

Whilst computer annotation programs have become substantially more sophisticated, final annotations typically need to be done manually, which currently presents the most substantial obstacle for all genome projects.^33,34 In this study, we used the program Apollo³⁵ to manually annotate and curate the F. circinatum genome. Apollo can directly utilise the sequence output from MAKER and this program also has the advantage of being relatively user friendly for biologists not familiar with computer programming.

In addition to utilising manual curation for the F. circinatum annotation, we followed the novel strategy of engaging students as annotators in the process. This approach was adopted because the skills required for curating a simple eukaryotic genome require little more than a basic degree in the biological sciences with some molecular biology focus. By following this approach we were able to achieve our second aim of promoting human capacity in the field of genome annotation in South Africa. A team of 20 graduate student volunteers was identified for this study. The students were then exposed to a 2-day training course in which the theoretical background involved in gene and genome structure was reinforced and the basic concepts and requirements of the annotation process were learned.

All the annotators were supplied with a number of contigs to curate and a support programme was implemented to assist those annotators that encountered problems. In most cases the learning curve for members of the annotation team was considerable, but tackling the annotation process in this way clearly highlightedthe value of genome sequences to a biological sciences programme. The project made it possible to not only foster an appreciation of the methodologies and approaches associated with genome sequencing projects, but also provided a large number of graduate students with the opportunity to become experienced in the process of genome sequence annotation.

During the curation, each predicted ORF was compared with the predicted genes from the genomes of F. verticillioides, F. oxysporum and F. graminearum. What was immediately obvious was that about 70% of the F. circinatum ORFs were most similar to those of F. verticillioides, which is consistent with the fact that these two fungi are more closely related to one another than to the other two species (Figure 2). In many cases, when a F. circinatum ORF was not most similar to one in F. verticillioides, the dissimilarity was found to be as a result of differences in intron prediction between the two genomes. Although the ORFs in F. verticillioides have been annotated using FGENESH³⁶ that also utilises a hidden Markov model-based algorithm to find genes, the genome of this fungus has not been subject to much manual annotation. Also, the numbers of predicted ORFs in F. circinatum and F. verticillioides differed considerably. Compared to F. circinatum, which has about 15 000 ORFs, F. verticillioides contains only about 13 500 ORFs. Of the ORFs apparently missing in F. verticillioides, a significant proportion had, in fact, not been annotated, despite the availability of EST evidence in many cases. This absence suggests that the annotation of F. verticillioides as presented on the Broad Institute website²⁵ requires additional analyses which would probably increase the number of predicted ORFs in this genome by as much as 5%.

Inspection of the annotated output for the F. circinatum assembly revealed further discrepancies amongst the results of the different predictions programs employed by MAKER. For example, Genemark predicted 15 713 ORFs, whilst Augustus predicted 14 210 ORFs. By manually curating the annotation, it was thus possible to evaluate the various ORF prediction outputs of the pipeline in terms of intron–exon boundaries and EST evidence for F. circinatum and the other Fusarium species. After the manual curation, the F. circinatum assembly contained 15 049 predicted ORFs, with an accuracy of at least 90% for the combined gene prediction of these two programs.

From the curation it was also observed that most often the contigs terminated in intergenic regions. Although this could be ascribed to the reduced ability of the gene prediction programs to find ORFs in the absence of 3’ or 5’ gene signatures, the CGEMA output indicated that more than 95% of the core eukaryotic genes were present in the F. circinatum. A more likely explanation is that the assembly program Newbler was not able to assemble across DNA repeat regions, which are most often found in the intergenic regions.

The fumonisin gene cluster has been well characterised in F. verticillioides.³⁸ By making use of this information, we were able to compare the organisation and composition of this cluster in the genomes of F. verticillioides, F. oxysporum, F. circinatum and F. graminearum (Figure 4). Our results confirm previous reports that these genes are absent in both F. circinatum and F. graminearum, neither of which has ever been shown to produce fumonisins.^37,38,39,40 This gene cluster is also missingfrom the genome of the isolate of F. oxysporum for which the genome is available, but this locus is known to be present in another isolate (strain FRC O-1890) of the same species.⁴¹ It is interesting that one of the genes (ORF 20) flanking this locus is in a different orientation in F. oxysporum and F. circinatum and entirely missing from the F. graminearum genome. In addition, in F. circinatum ORF 20 is placed between the genes Znf1 and Zdb1, whilst these two genes are alongside each other in

F. verticillioides, F. oxysporum and F. graminearum. These rearrangements and deletions thus suggest that recombination, especially in the regions flanking the cluster, determines whether fumonisin will be produced. Such recombination events could potentially facilitate horizontal transfer of this cluster amongst unrelated strains, which has been suggested to explain the patchy distribution of fumonisin production amongst lineages of Fusarium.⁴⁰ A number of Fusarium species have been shown to produce the mycotoxin fusarin C, although its role in disease has not been established.^37,38 Comparison ofthe fusarin C gene cluster in the four Fusarium species revealed its presence in F. circinatum, as well as in F. verticillioides and F. graminearum (Figure 5). Within the cluster, the gene order in F. circinatum is different from that found in F. verticillioides and F. graminearum, where the gene order is similar. In all instances the regions flanking this cluster were also unique. The fact that the locus seems to occur in a different position in the genomes of F. verticillioides, F. graminearum and F. circinatum, thus suggests that this gene cluster has been translocated more than once during the evolution of this fungal genus. In the case of F. circinatum, this translocation has been accompanied by a change in the gene order.

FIGURE 4: Diagrammatic comparison of the FUM locus in Fusarium circinatum, F. verticillioides, F. oxysporum and F. graminearum, based on the available genome sequences.

FIGURE 5: Diagrammatic comparison of the fusarin locus in Fusarium circinatum, F. verticillioides, F. oxysporum and F. graminearum, based on the available genome sequences.

The genome of an isolate of the pitch canker pathogen, F. circinatum, was sequenced and manually annotated. A more detailed analysis of the genome will be published during the coming year. The sequence data is available at Genbank Bioproject: PRJNA41113; ID: 41113 Locus Tag Prefix: FCIRG and on the FABI website (http://www.fabinet.up.ac.za/genomes). The F. circinatum genome, whilst being in the expected size range for a Fusarium genome, has almost 1000 more protein coding genes than its closest relative, F. verticillioides. F. oxysporum has 3000 more protein coding genes than F. verticillioides and many of these have been proposedto originate from the acquisition of lineage-specific genomic regions.¹⁹ Understanding the origin of the additional 1000 genes in F. circinatum will add to this intriguing hypothesis.

Fusarium species are well known for their production of mycotoxins. Whilst the study of mycotoxins is of particular importance for species that contaminate food and feed stocks, these toxins have also been shown to be important in plant pathogenesis.³⁸ A comparative analysis of the F. circinatum genome with the otherFusarium genomes currently available has enabled us to determine that this fungus does not contain the fumonisin gene cluster, although the genes for the synthesisof fusarin C are present. Whether or not the pitch canker pathogen is capable of producing this compound remains to be determined. Also, the role of fusarin C has not been established in human or animal disease, nor has its role in plant pathogenesis been determined. It is perhaps most noteworthy that in the case of both these gene clusters, significant deletions or insertions and rearrangements have occurred. There is also increasing evidence of horizontal gene transfer in fungi¹⁹ and it is likely that there would be particularly strong selective pressure in terms of acquiring the ability to produce mycotoxins. Further analysis of the F. circinatum genome will focus on this possibility.

Analysis of the unique gene clusters identified in F. circinatum, whilst interesting, did not contain any significant surprises. The genome of this fungus contains a large number of proteases or peptidases, lyases and transferases that deserve further investigation. For example, these proteins could be involved in the synthesis of secondary metabolites, which in turn have the potential to be involved in pathogenesis. Amongst the four Fusarium genomes compared, F. circinatum is unique inbeing a gymnosperm and tree pathogen. The other three Fusarium species are pathogens typically of angiosperm monocotyledonous crops and we thus expect that F. circinatum could have a unique set of genes involved in pathogenesis.

A number of viral proteins were also found in the gene clusters, suggesting the presence of viral genomes within the F. circinatum genome. We did not filter for transposable elements and some of the genes identified as viral proteins could in fact represent elements associated with transposons. There is also precedence for the presence of both retroviruses and transposons in fungal genomes.⁴²

The availability of the F. circinatum genome will strongly promote various projects currently being undertaken on the pathogen. For example, variation in pathogenicity was observed in the interspecies cross between F. circinatum and Fusarium subglutinans, which was used by De Vos et al.²⁰ to produce a genetic linkage map. The availability of both the linkage map and the genome sequence will be used to study some genome regions that are potentially associated with pathogenicity. These studies will include further comparative analyses of the genome for mycotoxin biosynthetic gene clusters and other secondary metabolic gene clusters. In addition, quantitative trait loci have been identified and linked to the growth of F. circinatum (De Vos, unpublished data). With the availability of the F. circinatum genome sequence, the genetic components of these loci can now be understood and this will add to knowledge regarding the mycelial growth in fungi.

Genome sequencing is rapidly growing in importance and there is little question that this field will impact increasingly on most aspects of biology. This growth is a logical continuation of the situation that existed little more than two decades ago when the ability to sequence relatively small numbers of genes began to influence the field. The number of complete genome sequences has increased from one to close to 1000 in just 15 years.⁴³ Apart from next generation sequencing that has substantially influenced this growth, new technological developments will ensure that this process continues, ultimately resulting in broad applications relating to genome sequences. This investigation, promoting the sequencing and annotation of a relatively small, but very significant genome, will undoubtedly come to represent a milestone in genome sequencing in South Africa.

An important element of genome sequencing is that the bulk of the work lies in annotating and interpreting the data. The entry point to this process is the laboratory component, the physical sequencing process. In South African terms, this is still relatively expensive, even for genomes as small as that of F. circinatum. However, the costs are dropping rapidly and suites of genomes, or the genomes of numerous strains of single species, are now being sequenced in consortium projects interested in comparative genomics. Whilst the wet laboratory work might be somewhat beyond the budgets of many South African laboratories, there are substantial opportunities for local scientists and students with knowledge of genome annotation to utilise genome data that is already freely available. In the future, there will be even larger volumes of genome data available for study. Some of the scientists involved in the investigation presented here will be well positioned to capitalise on this and related opportunities.