Overall N. caninum DNA findings in aborted foetuses
One hundred and ninety-eight aborted bovine fetuses from 165 farms located in the Po valley of Lombardy, i.e. the provinces of Cremona, Lodi, Mantova, and Milano, were included in the study. Out of 198 aborted fetuses 55 were positive for N. caninum by RT-qPCR, yielding a prevalence of 27.8% (presence of the parasite in pooled tissue homogenates of brain, lung, and liver); 43 farms recorded at least one positive fetus (26.1%). Overall, 55 samples from 43 herds of fetuses collected from 2015 to early 2019 were subjected to microsatellite typing.
Relationship between real time PCR results and success in genotyping
The majority of the 55 DNA samples subjected to multilocus microsatellite typing were typed at all 10 (n = 20), or at 9 (n = 15) loci. Nine DNA samples could be typed at 8 (n = 2), 7 (n = 2), or 6 (n = 6) loci. The remaining 10 DNAs could only be typed at 5 (n = 3), 4 (n = 3), 3 (n = 2), 2 (n = 1), or 1 (n = 1) of the loci.
The numbers of genotyped loci were related to the amount of parasitic DNA in the sample as reflected by the Ct values determined by real time PCR (Table 1).
Microsatellite typing on individual farms at different time points revealed differences in the MLMG results
In seven farms, more than one fetus had been sampled and detected N. caninum DNA typed in the study period from 2015–2018. Most of the differences in typing comprised of only one or two repeat units in MLMG (Table 2). However, in five farms, more prominent differences per locus (i.e. differences of 3 or even more repeat units per locus) were observed. In herd 6, i.e. the farm with the largest number of fetuses analysed, 48.1% of the comparisons of individual loci revealed differences (Table 2, Table 3). These more prominent differences affected the loci MS7 and MS21 (Chromosome 7a), MS10 (MS10.2, Chromosome 8), and MS6A (Chromosome 10) (Table 3).
Relationship between the spatial distance of sampling sites and the genetic distance of N. caninum isolates
For n = 20 of the N. caninum positive samples of bovine foetuses from Lombardy, which had been typed at the complete set of microsatellites loci (MS1B, 2, 3, 5, 6A, 6B, 7, 10, 12 and 21, this study), pairwise Bruvo’s genetic distances were calculated. Bruvo’s genetic distance and the geographical distance between sampling sites of the individual N. caninum-positive foetuses (i.e. based on the geographical coordinates of 17 farms) were tested for a correlation. Linear regression revealed that Bruvo’s genetic distance correlated statistically significantly with the geographical distance between the sampling sites (Fig. 1a; Table 4). The model (Model 1, Table 4) had an adjusted R2 of 19.5%. Including the number of days between the sampling of foetuses into the model did not improve the model significantly (Model 2, Table 4). Bruvo’s genetic distance with non-parametric bootstrapping allowed the identification of at least four groups, separated by bootstrap values > 50 (Fig. 1b). While the farms, from which the largest group of samples originated from (Fig. 1b, red), were located with two exceptions central-north in the Cremona district (Fig. 1c, red), the farms, where the remaining samples had been derived from, were located in the far north (Fig. 1c, green), in the south of the Cremona district (Fig. 1c, blue), or in the neighbouring district of Mantova (Fig. 1c, yellow).
An eBURST and PCoA analysis reveal different groups or clusters of N. caninum MLMGs in northern Italy
In addition to the microsatellite data established in this study (Lombardy; n = 25), further data on n = 50 N. caninum isolates from northern Italy covering mainly other regions than Lombardy (i.e., Piedmont n = 17, Veneto-Trento n = 6, including also one goat isolate), Lombardy (n = 1), and a further North-Italian bovine N. caninum isolate (n = 1) were analysed. These additional data were available from a recent study [31]. Since the latter study had employed a set of microsatellite markers that overlapped only partially with the one we used, our analysis was restricted to MS6A, MS6B, MS10, MS12 and MS21, for which data were available from both groups. Linkage disequilibrium (LD) was assessed for the entire population excluding the goat sample; results (IAS=0.0411, VD=1.8249, L = 15508, P = 0.0411) indicated LD because VD > L [39].
Network analysis using eBURST with double locus variation (DLV) among 6 loci and 3 MS10 motifs or sub-loci (MS10.1-3) revealed that many of N. caninum MLMGs from Lombardy (n = 11/25; this study) clustered separately (eBURST G4) from those obtained from Piedmont (n = 10/17, eBURST G1, G2, G3) or Veneto-Trento (n = 3/6, eBURST G1, G2) (Fig. 2a). The grouping of isolates from Piedmont and Veneto-Trento had already been assessed by others [31]. This previous grouping of particular N. caninum isolates into the eBURST groups G1, G2 and G3 matched perfectly (n = 10/10) with our grouping (Table S1). Genotyping of an Italian isolate, for which the province of origin was not known (violet dot) revealed results close to most isolates from Lombardy (group G4). The full representation of the MLMST network using the MST option in eBURST, including the complete data set (n = 49 N. caninum samples), shows the G4 samples in the centre with most of the Veneto-Trento samples on the right side (G1) and the Piedmont samples both left (G2, G3) and right (G1) in the network (Fig. 2b).
Investigating N. caninum positive samples (n = 48, excluding a sample, for which the province of origin was unknown, and the goat sample) by PCoA revealed a clear axis #1 separation of the samples from Piedmont (Fig. 3) and Lombardy (Fig. 3). While n = 17/17 samples from Piedmont were located on the left side, all from Lombardy (n = 17/25) were on the right side (Fig. 3). The majority (n = 3/5) of samples from Veneto-Trento (Fig. 3) were left or right in the PCoA graph.
When the eBURST grouping (DLV option) was compared to the PCoA result, eBURST G1 and eBURST G3 was located separately in the upper left quarter of the PCoA graph, while eBURST G2 was located in the lower left part (Fig. 3). Neospora caninum samples that had been separated by eBURST into G4 (exclusively Lombardy samples) clustered in both the upper and lower corners on the right side of the PCoA graph (Fig. 3).
Differences between the N. caninum populations in northern Italy were assessed FST analysis values (F) and Nei’s unbiased genetic distance (D) (Table 5). Both, results for F and D suggest that there is a statistically significant genetic difference between N. caninum from Piedmont and Lombardy as well as between Lombardy and Veneto-Trento (Table 5).
The graphical representation of the heterogeneity in numbers of repeat units, stratified for various loci, suggested differences between bovine N. caninum sampled in the North Italian regions of Lombardy, Piedmont, and Veneto-Trento (Fig. 4). The range between the lowest and highest number of repeats was largest for MS10.2, followed by MS6A, MS5 and MS7. However, the pairwise Wilcoxon test (BH adjusted P value) revealed statistically significant differences between regions only for MS5, MS6B, MS7, MS10.1, and MS10.2 (Fig. 4).