A draft genome for the African crocodilian trypanosome Trypanosoma grayi

The availability of genome sequence data has greatly enhanced our understanding of the adaptations of trypanosomatid parasites to their respective host environments. However, these studies remain somewhat restricted by modest taxon sampling, generally due to focus on the most important pathogens of humans. To address this problem, at least in part, we are releasing a draft genome sequence for the African crocodilian trypanosome, Trypanosoma grayi ANR4. This dataset comprises genomic DNA sequences assembled de novo into contigs, encompassing over 10,000 annotated putative open reading frames and predicted protein products. Using phylogenomic approaches we demonstrate that T. grayi is more closely related to Trypanosoma cruzi than it is to the African trypanosomes T. brucei, T. congolense and T. vivax, despite the fact T. grayi and the African trypanosomes are each transmitted by tsetse flies. The data are deposited in publicly accessible repositories where we hope they will prove useful to the community in evolutionary studies of the trypanosomatids.

The availability of genome sequence data has greatly enhanced our understanding of the adaptations of trypanosomatid parasites to their respective host environments. However, these studies remain somewhat restricted by modest taxon sampling, generally due to focus on the most important pathogens of humans. To address this problem, at least in part, we are releasing a draft genome sequence for the African crocodilian trypanosome, Trypanosoma grayi ANR4. This dataset comprises genomic DNA sequences assembled de novo into contigs, encompassing over 10,000 annotated putative open reading frames and predicted protein products. Using phylogenomic approaches we demonstrate that T. grayi is more closely related to Trypanosoma cruzi than it is to the African trypanosomes T. brucei, T. congolense and T. vivax, despite the fact T. grayi and the African trypanosomes are each transmitted by tsetse flies. The data are deposited in publicly accessible repositories where we hope they will prove useful to the community in evolutionary studies of the trypanosomatids.

Background & summary
Most trypanosomatid parasites of humans, livestock and plants are transmitted between hosts by invertebrate vectors. They are widespread, and are collectively responsible for substantial economic and health losses in many of the world's poorest regions. Within this group are the Leishmania species, the causative agents of Leishmaniasis, as well as the monophyletic Trypanosoma genus, which includes Trypanosoma cruzi and Trypanosoma brucei, causative organisms of Chagas disease and African trypanosomiasis respectively. Despite a probable single origin of vertebrate parasitism within this monophyletic group 1,2 , the challenge of escaping clearance by host immune responses has given rise to a variety of disparate parasitic lifestyles including intracellular parasitism (Leishmania spp. and T. cruzi) and antigenic variation (T. brucei) [3][4][5] .
Since the publication of the T. cruzi, T. brucei and Leishmania major genomes in 2005 [6][7][8] , several other trypanosomatid genome sequences have been released, including further Leishmania species and other African tsetse transmitted trypanosomes related to T. brucei [9][10][11][12] . The availability of sequence data for other T. brucei clade trypanosomes has increased our understanding of the evolution of the primary immune evasion strategy of this parasite as well as the evolution of cell surface molecules that represent the host-parasite interface 12,13 . Similarly, sequencing of disparate T. cruzi isolates has provided major insights into population structure and dynamics 14,15 . Though more species now have published genome sequences, sampling across the trypanosomatid phylogeny is limited and thus there are limited resources for comparative genomic investigations.
To address this key knowledge gap, here we provide a draft genome sequence of the African trypanosomatid parasite of crocodiles, Trypanosoma grayi (Data Citation 1 and Data Citation 2). T. grayi is an extracellular parasite of the bloodstream of crocodiles, and though it is transmitted by tsetse flies it is closely related to other trypanosome parasites of crocodiles in South America 16 . The trypanosome is taken up by tsetse flies in a bloodmeal and resides solely within the mid-and hindgut. Unlike salivarian trypanosomes, transmission between crocodile hosts occurs via oral contamination with infective metacyclics in tsetse faeces 17,18 . This faecal transmission strategy is employed by many other trypanosomes, including T. cruzi.
BLAST and OrthoMCL analysis of the genome sequence and predicted gene models respectively suggests that T. grayi possesses neither the T. brucei type VSG surface antigens nor the T. cruzi type mucin coat. Thus T. grayi may have evolved an alternative family of primary surface antigen genes, or possess a novel immune evasion strategy geared to survival in the reptilian bloodstream 19 . Both phylogenomic reconstruction and best-BLASTp analysis demonstrate that T. grayi is more closely related to T. cruzi than to T. brucei ( Figure 1 and Table 1). This result refines the phylogenetic position of T. grayi, that in previous studies using 18S ribosomal RNA and glycosomal glyceraldehyde dehydrogense (gGAPDH) genes was placed in a separate clade from both T. cruzi and T. brucei, often with other reptile or bird trypanosomes [20][21][22] . Additional taxon sampling in this region of the phylogenetic tree will be important for resolving these relationships further. We anticipate that these data will provide a useful comparator for evolutionary studies of the adaptations of trypanosomes to different vertebrate hosts, as well as increasing the available sequence data resources for this globally important group of parasites.
To generate the draft genome, DNA from T. grayi strain ANR4, isolated from the midgut of the tsetse fly Glossina palpalis gambiensis in The Gambia 20 was sequenced by 91 bp paired-end Illumina sequencing and assembled de novo into contigs (Data Citation 2). We inferred the phylogenetic position of T. grayi strain ANR4 through construction of a concatenated protein sequence phylogeny using 959 single copy nuclear encoded genes. We also confirmed that both the 18S ribosomal RNA sequence and gGAPDH sequence for our T. grayi strain ANR4 were 100% identical to those provided in GenBank (AJ005278 and AJ620257 respectively) for T. grayi. Furthermore, we have identified and annotated over 10,000 putative open reading frames and have submitted this information to public databases alongside the draft genome sequence.

Methods
Sequencing and assembly T. grayi strain ANR4 was grown in vitro in Cunningham's medium and genomic DNA was extracted from agarose plugs using standard phenol/chloroform methods. DNA was sequenced by 91 bp pairedend Illumina sequencing at the Beijing Genomics institute (www.genomics.cn/en/). Raw reads were subject to quality filtering using trimmomatic 23 . This was done to remove low quality bases and readpairs as well as contaminating adaptor sequences prior to assembly. Sequences were searched for all common Illumina adaptors (the default option) and the settings used for read processing by trimmomatic were 'LEADING:10 TRAILING:10 SLIDINGWINDOW:5:15 MINLEN:50'. The quality filtered pairedend reads were then subject to read error correction using the ALLPATHS-LG 24 ErrorCorreactReads.pl program using the default program settings. The corrected reads were then assembled using SGA 25 using default settings and setting the mimimum overlap length to 80. The assembled contigs were scaffolded by mapping the trimmed and filtered paired-end reads (described above) to the assembled contigs using BWA-MEM and scaffolding the contigs using the SGA 25 scaffolding algorithm using default program settings. The resultant scaffolds were then subject to fourteen rounds of assembly error correction and gap filling using Pilon (http://www.broadinstitute.org/software/pilon/) using the '-fix all' option and setting the expected ploidy to diploid. Following scaffolding and assembly error correction all filtered paired-end reads were mapped to the contig set using BWA-MEM 26 , paired-end reads that did not map to the assembly were isolated and the above assembly, scaffolding and correction process was repeated until all no-further reads could be assembled. The final draft assembly contained 2,963 sequences greater than 100 bp in length with an N50 of 16.7 kb and a total assembly length of 20.9 Mb and average coverage per assembled contig of~105X (Figure 2a,b).

ORF finding and annotation
The assembled draft genome of T. grayi was subject to gene model prediction using Augustus 27 . In brief, an initial set of gene models were predicted using gene prediction parameters inferred by training Augustus using the set of genes currently annotated in the T. cruzi genome. These gene model parameters were used to predict a training set of genes in the draft assembly of T. grayi. The training set of genes were then used for multiple iterations of prediction and training until prediction converged on a final set of gene models and no further genes could be detected. The identity of the T. grayi DNA used for sequencing was confirmed against database sequences for 18S ribosomal RNA and glycosomal GAPDH genes (AJ005278 and AJ620257 respectively).

Gene family analysis
The protein sequence files for a subset of available trypanosomatid genomes were downloaded from TriTrypDB. These were combined with the newly predicted protein sequences from T. grayi and subject to orthologue group clustering using OrthoMCL 28 . The presence of gene families in each species was analyzed and the overlap in gene family content between each species and that of the newly assembled

Phylogenetics for strain verification
Orthologous sequence groups that contained only single copy genes in each of the species that were subject to clustering were selected (n = 959). These single copy gene families were aligned using MergeAlign 29 and concatenated to form a super-alignment containing 119,006 aligned amino acid positions across all species (1,547,078 amino acids). This concatenated alignment was subject to  phylogenetic inference using bootstrapped maximum likelihood, Bayesian inference and bootstrapped neighbor joining methods. Maximum likelihood trees were inferred using FastTree 30 , utilizing the JTT model of amino acid substitution and CAT rates. A Bayesian inference tree was inferred using MrBayes v3.1.2 31 using the WAG model of amino acid substitution and gamma distributed rates approximated by four discrete gamma categories. Two runs each of four chains were initiated and allowed to run for 200,000 generations sampling every 500 generations. Convergence was assessed through visual inspection of log-likelihood traces and through analysis of the standard deviation of split frequencies. The analysis had reached stationary phase after 15,000 generations and these first 15,000 generations were discarded as burnin prior to inferring the consensus tree. The neighbor joining tree was inferred using QuickTree 32 using the default parameters. The final topology is shown in Fig. 1 and received 100% support at each bipartition from all methods.

Technical Validation
The contig statistics of the assembly are reported in Figure 2, and an example region of an assembly against several related trypanosomatid genomes is shown in Figure 3. Phylogenetic strain validation as described above confirmed the placement of T. grayi ANR4 with other species of genus Trypanosoma (Fig. 2) and identity of the sequenced genome here with the previously reported 18S and glycosomal GAPDH genes (AJ005278 and AJ620257 respectively). The phylogenomic position of T. grayi closer to T. cruzi than T. brucei is also supported by BLASTp analysis of all predicted open reading frames (Table 1).