Genotyping of European Toxoplasma gondii strains by a new high-resolution next-generation sequencing-based method

In total, 170 T. gondii specimens, including 123 cell-culture isolates and 47 clinical samples, were analyzed according to the workflow depicted in Fig. 1 (also see Supplementary Figure 1). The sample set comprised specimens (Supplementary Table 1) from 19 different countries on the European continent (Fig. 2) and seven non-European countries or locations. Clinical samples originated from 12 different matrices and 15 animal species, including domestic, wild-living and zoo animals. Isolates were cell-cultured as described [17, 19, 20]. The isolates cultivated at Friedrich-Loeffler-Institut (FLI) included the type I reference strains RH_FLI and GT1_FLI, the type II references ME49_FLI and NTE_FLI and the type III reference NED_FLI. DNA was extracted by standard methods from cellular pellets or clinical material (Supplementary Table 1). A real-time PCR targeting TgREP-529 [21] was used to characterize DNAs quantitatively [22] (Supplementary Note 1).

All specimens were genotyped using 15 MS markers [10]. For the markers N60, M102 and AA, the fluorophore (FL) Atto550_Fl was used instead of NED_Fl for primer labelling. The reported fragment sizes of these three markers were numerically adjusted based on published guidelines [11]. Furthermore, all isolates were genotyped using nine PCR-RFLP markers [12, 23]. In both methods, the reference strains RH_FLI, ME49_FLI and NED_FLI were used as positive controls and water as a negative control. The PCR-RFLP and MS genotypes of the reference strains PRU and CZ-H3 described in the literature (Supplementary Table 1) were included in the data analysis.

Whole genome sequencing of 59 T. gondii isolates (Supplementary Table 1) was conducted with the Illumina NovaSeq 6000 system in 150 bp paired-end mode (Biodiversa s.r.l., Treviso, Italy). Raw read data of these genomes was processed bioinformatically with 21 publicly available and three reference genomes, ME49, PRU and CZ-H3 (accession numbers in Supplementary Table 1) as described in detail in Supplementary Note 2. Finally, genetic variants were detected and converted into genomic variant call format (gVCF) for further use.

HPRs (n = 55) identified in the gVCF files of the WGS analysis were assessed by Sanger sequencing (details in Supplementary Note 3 and Supplementary Table 2). Sanger HPR sequences obtained in this study for three representative isolates and ME49_FLI were aligned to a publicly available ME49 sequence (ToxoDB release 47), using the software Geneious Prime (version 2021.0.1), and SNPs detected by Sanger sequencing were compared to those identified by WGS analysis.

Library preparation was performed using the Ion AmpliSeq™ Library Kit Plus and IonCode™ Barcode Adaptors (Thermo Fisher Scientific, Waltham, MA, USA), following the manufacturer’s instructions.

For the initial multiplex PCR, an Ion AmpliSeq™ custom panel was designed (Ion AmpliSeq Designer, version 7.49, Thermo Fisher Scientific) by using a ME49 genome sequence (ToxoDB release 47), a BED file containing information about genetic variants of a subset (n = 43) of all available T. gondii genomes (Supplementary Note 2) and the locations of 24 Sanger sequencing confirmed target regions. Since six regions were excluded by the Ion AmpliSeq Designer, the panel consisted of 68 primers divided into two pools covering 18 regions (Supplementary Table 3). The final target regions were larger compared to the target regions initially identified by WGS analysis or covered these only partially (T16, T32, T51), as illustrated in Supplementary Figure 2a–b. PCR cycling conditions were 99 °C for 2 min, followed by 24, 26 or 28 cycles at 99 °C for 15 s and 60 °C for 8 min.

Adapter ligation was followed by size selection as described [24]. Library quality was checked using the High Sensitivity DNA Kit on an Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA) or by using the 4200 TapeStation (Agilent Technologies). Libraries were quantified with the QIAseq™ Library Quant Assay (Qiagen Sciences, Germantown, MD, USA), pooled including the Ion S5 Calibration Standard and the pools sequenced on an Ion 530 chip with an Ion S5 XL System (Thermo Fisher Scientific) in 400 bp-mode according to the manufacturer’s instructions.

For data analyses, a reference sequence set (accessible at https://​zenodo.​org/​; DOI: 10.5281/zenodo.8377016), in the following referred to as AmpliSeq-ME49-Reference, was created using the sequence data of each target region in the genome of ME49 (ToxoDB release 53) with additional 10 bp added to the 5′- and 3′-ends of the regions. The final target region corresponded in 17 of 18 targets to the amplicons generated by the primer panel as described in Supplementary Figure 2a–b and Supplementary Table 4. In the case of T26, the final target region was shortened for data analysis, because runs of consecutive thymine nucleotides (poly[T]) had led to ambiguous sequencing results. Sequence reads of each library were analyzed by reference mapping with the Torrent Mapping Alignment Program (TMAP-ion, version 3.4.0) to generate bam files for further use. Mapping quality (MQ) of the reads against the AmpliSeq-ME49-Reference was analyzed by Qualimap bamqc (v2.3) [25]. Furthermore, several BCFtools (v1.15.1 [using htslib 1.16]) [26] were employed, including the “mpileup” command to call variants in each Ion AmpliSeq record with mapped reads to generate library-specific variant call format (VCF) files. All Ion AmpliSeq VCF files were merged into a multiVCF file using the “bcftools merge” command and all variants were filtered by VCFtools (v0.1.16-20) using the hard filter criteria of MQ > 30 and read depth (DP) > 10. Moreover, the combination of Samtools faidx and the “bcftools consensus” command [26] was used to convert the VCF data into the FASTA format. The program Snp-sites (2.5.1) [27] was applied to extract the variable sites from the FASTA sequences. If parts of the AmpliSeq-ME49-Reference were not covered by the reads of specific genotypes, the corresponding nucleotides were indicated as “N” in the FASTA file. The aligned FASTA file containing respective library-specific SNPs was then converted into the NEXUS format and incorporated into SplitsTree4 software (version 4.18.1) [28] to generate unrooted phylogenetic networks using a neighbour-net method and 1000 bootstrap replicates. Sample IDs were replaced by numbers (Supplementary Table 1).

To verify the results of the automated analysis described above, the sequence reads of each library were also mapped to a ME49 genome (ToxoDB release 53) using the Geneious Prime mapper (version 2021.0.1) and default settings. Coverage of the target regions was analyzed, and positions of potential SNPs were visually inspected.

The analytical sensitivity of the Ion AmpliSeq method was assessed with a set of serially diluted DNA of ME49_FLI previously used in a ring trial to harmonize MS typing [11]. The three dilutions were characterized by real-time PCR with Ct values of 23.79, 27.43 and 30.26, corresponding to DNA concentrations of 1 ng/μl, 0.1 ng/μl and 0.01 ng/μl. Each dilution was amplified with 24, 26 and 28 cycles in the Ion AmpliSeq multiplex PCR. The experiment was repeated three times.

Library preparation of 170 specimens was performed as described above. Based on the results of the sensitivity assessment, 24 cycles were defined as the standard protocol, but five isolates with Ct values > 24.5 and all clinical samples with Ct values > 22.0 were amplified with 28 cycles (Supplementary Table 1).

By MS typing, most specimens (n = 129) were genotyped as type II (Fig. 3, Supplementary Table 1). In addition, twelve specimens were categorized as type II variants as they showed a deviation on one MS lineage typing marker. Eight were W35, two TgM-A variants and one specimen each was a XI.1 or a B18 variant. Four specimens belonged to type I, 16 to type III and five specimens were categorized as type II × III recombinants. Furthermore, the six non-archetypal strains were classified as Africa 1, Caribbean 1, Caribbean 2, Caribbean 3 and Atypical.

Nine of the 86 type II isolates were PCR-RFLP genotyped as ToxoDB#1, while the remaining 77 type II and all type II variant isolates belonged to ToxoDB#3 (Fig. 3, Supplementary Table 1). MS type I corresponded to ToxoDB#10 and 11/13 type III isolates belonged to ToxoDB#2. Two type III isolates from Argentina were classified as ToxoDB#123 by PCR-RFLP. One of the type II × III recombinant isolates belonged to ToxoDB#3 and two to ToxoDB#2. The two remaining recombinants and the isolates MS typed as atypical, Caribbean 2 and 3 could not be assigned to any known PCR-RFLP ToxoDB number. MS type Africa 1 corresponded to ToxoDB#6 and Caribbean 1 to ToxoDB#13.

WGS data of 43 T. gondii type II genomes (Supplementary Table 1, Supplementary Figure 3) were used for the identification of HPRs (Supplementary Table 5). The mean number of reads per library was 62.1 M (range, 29.9–620.6 M) and the median depth of coverage after mapping to the ME49 genome, in the following referred to as ME49 reference, was 1077× in the case of ME49 (SRR6793863) and 15×–338× for the remaining 42 isolates. An average of 98.1% ± 1.4% standard deviation of each genome was mapped with over 10× coverage.

When mapping to ME49 reference, the SNPs found in the analyzed 43 T. gondii genomes sum up to a total of 65,006. The SNPs were used to identify target regions for the Ion AmpliSeq method (Table 1). SNPs were counted in non-overlapping windows of 333 bp, and four prioritization categories of target regions were defined. Nineteen target regions (Fig. 4) were categorized as first priority (20–35 SNPs), 37 as second priority (15–19 SNPs), 136 as third priority (10–14 SNPs) and 673 as fourth priority (5–9 SNPs). The first priority targets were located on seven chromosomes, mainly in subtelomeric regions. The second priority targets were located on ten chromosomes, while third and fourth priority targets were distributed all throughout the chromosomes (Fig. 4, Supplementary Table 5).

Sanger sequencing confirmed the WGS data of 24/55 tested SNP dense regions (Supplementary Table 6). Of the sequenced targets, 15.4% (2/13) of first priority, 50.0% (8/16) of second priority, 52.1% (12/23) of third priority and 66.7% (2/3) of the fourth priority targets were confirmed. SNP analysis was not possible for 13/55 regions, due to overlapping peaks in the Sanger sequences resulting in low Phred quality scores. Furthermore, the sequences of 4/55 regions were too short to cover the whole region after mapping to ME49, and in the case of 10/55 regions, no SNPs were detected by Sanger sequencing or observed SNPs were not in accordance with WGS data. The Ion AmpliSeq primer panel was designed using all 24 confirmed regions, containing 336 different SNPs. Six of the confirmed 24 regions were excluded from primer design as detailed in Methods.