Long-term molecular surveillance provides clues on a cattle origin for *Mycobacterium bovis* in Portugal

Animal tuberculosis caused by Mycobacterium bovis persists in multi-host systems worldwide and poses challenges to disease control due to transmission across livestock and wildlife. In Portugal, despite a long-running eradication program in cattle, regional heterogeneity in TB prevalence remains, particularly in central and south-eastern districts where cattle co-occur with abundant wild ungulates (red deer and wild boar) managed for hunting. The study addresses the research questions: how are M. bovis populations structured in space and time across key Portuguese hotspots, what are the likely transmission links among hosts, and can the origin and spread of epidemics be inferred? The purpose is to refine understanding of M. bovis population structure using molecular typing integrated with spatial statistics to inform targeted control at the livestock–wildlife interface. The work is important for improving eradication strategies, mitigating wildlife reservoirs’ impact, and addressing zoonotic risks.

Prior studies identify multiple wildlife reservoirs for M. bovis (e.g., badger in the UK/Ireland, African buffalo in South Africa, possum in New Zealand, white-tailed deer in the USA, red deer and wild boar in Iberia). In Portugal and Spain, spatial clusters of TB and high prevalence in wild ungulates are reported, with expansion linked to high host densities, artificial management, and super-shedders. Portugal’s eradication program since 1987 reduced national cattle herd prevalence, but regional disparities persist, with Alentejo showing higher rates. An epidemiological risk area for big game was established in 2011 mandating carcass inspection and laboratory confirmation. Molecular studies (spoligotyping and MIRU-VNTR) in Iberia demonstrated inter-species transmission and regional strain sharing; Eu2 clonal complex predominates in the Iberian Peninsula. However, quantification of each host’s role and fine-scale spatiotemporal dynamics remained limited, motivating the present long-term, multi-host, multi-region molecular epidemiology analysis.

Study area: Three Portuguese TB hotspots at the livestock–wildlife interface—districts of Castelo Branco (CB), Portalegre (PG), and Beja (BJ). Isolate collection: 948 M. bovis isolates (2002–2016; cattle n=384, red deer n=303, wild boar n=261). Time stratification: Period 1 (2002–2008), Period 2 (2009–2012), Period 3 (2013–2016). Sampling: Cattle tissues from sanitary slaughter/reactors or abattoir lesions; wildlife tissues from hunted animals with TB-compatible lesions within the official risk area. Culture on Stonebrink and Löwenstein-Jensen pyruvate media and liquid medium (BACTEC 9000 MB), 37°C up to 12 weeks. Species ID by gyrB PCR-REA or GenoType Mycobacterium; controls included M. tuberculosis H37Rv, M. bovis BCG, and no-template. Spatiotemporal clustering: SaTScan multinomial space-time scan (v9.6) on 723 geocoded cases (cattle n=249 across 135 herds; wildlife n=474 across 91 hunting areas). Cylindrical scanning windows (max 50% of time and case data), no overlapping clusters; significance via 999 Monte Carlo replications (p<0.05). Mapping in QGIS v3.10. Molecular typing: Spoligotyping for all isolates (n=948) using in-house membranes per standard protocol; profiles assigned via mbovis.org (SB codes). Trend analysis of top six SB profiles via linear regression. MIRU-VNTR: Representative subset (n=524) chosen to cover all SB groups and balance host/region/year; nine orphan SBs included. Panel of 8 loci (VNTR3232, ETR-A, ETR-B, ETR-C, QUB11a, QUB11b, MIRU4, MIRU26) by PCR; allele calling by amplicon size; controls included M. bovis BCG and water. Complete single-allele profiles (n=447) grouped into MIRU types; double-allele profiles (n=37) flagged as polyclonal; for double-allele cases, four extra loci (MIRU16, MIRU40, ETR-E, QUB26) were amplified. Profiles with non-amplifiable loci were excluded from MIRU-type classification. Genotypic profile definition: Combined SB profile + MIRU type. Discriminatory power (Simpson’s D) and allelic diversity (h) computed for typing methods (only complete single-allele MIRU profiles considered for D). Network analysis: Gephi v0.9.2 networks linking host–region nodes by counts of shared SB or combined genotypic profiles (absolute counts). Phylogenetics: Minimum Spanning Trees (MST) in BioNumerics v6.6 using combined SB+8-loci MIRU data (n=487, excluding double-allele profiles), categorical similarity, single-locus variant maximum distance. Bayesian population structure: STRUCTURE v2.3.4 on MIRU-VNTR profiles (n=487), admixture model, burn-in 10,000 and 100,000 MCMC iterations, K=2–15 with 10 replicates per K; delta K via Evanno method (STRUCTURE HARVESTER). Assignment threshold Q≥0.50. MSTs relabeled by inferred populations for congruence checks. Allelic richness: HP-RARE v1.0 rarefaction by region and STRUCTURE populations to standardize sample sizes. Statistics: Chi-square tests (α=0.05) with Monte Carlo exact p-values (10,000 resamples, 99% CI) for associations of SB or STRUCTURE populations with region and host (IBM SPSS v24).

- Dataset and demography: 948 isolates across 3 hosts and 3 regions over 15 years; wildlife isolates became regular after 2011. Two significant space-time clusters (p=0.001): (1) Beja/Barrancos, 2004–2010, radius 173.17 km, cases=108 (cattle 105, red deer 2, wild boar 1), Relative Risk (RR): cattle 4.16, red deer 0.047, wild boar 0.025; (2) Castelo Branco/Rosmaninhal, 2012–2016, radius 31.49 km, cases=270 (cattle 13, red deer 131, wild boar 126), RR: wild boar 2.08, red deer 1.90, cattle 0.093. - Spoligotyping: 64 SB profiles (D=0.92), including 3 newly recorded (SB2529, SB2530, SB2531). Top six: SB1174 (n=158; 16.7%), SB0122 (n=123; 13.0%), SB0121 (n=111; 11.7%), SB1264 (n=92; 9.7%), SB0265 (n=75; 7.9%), SB0119 (n=57; 6.0%); together 65% of isolates. SB0119 and SB0121 decreased over time; SB0122 and SB1264 increased. - MIRU-VNTR: On n=524, obtained 447 complete single-allele profiles forming 157 MIRU types; 37 double-allele (polyclonal) profiles; 40 with one unamplified locus. Majority of MIRU types were singletons (60.1%). Most frequent MIRU types: M34, M205, M219, M225, M251, M261 (35% of subtyped isolates). Overall MIRU D=0.97; locus diversity h ranged 0.28–0.68 (highest ETR-A h=0.68; lowest MIRU26 h=0.28). All loci polymorphic (3–10 alleles per locus). - Combined genotypes: 208 SB+MIRU genotypic profiles (n=487). Profiles shared across hosts within regions; common across adjacent regions but not all three regions simultaneously. - Polyclonal infection: 37 cases with double alleles at 1–4 loci; 83.8% with a single affected locus; double alleles across all loci except QUB11a. Six wildlife cases (CB) had 2–4 loci with double alleles. - Regional diversity: SB profiles—CB (48 profiles; D≈0.90) > BJ (28; D≈0.90) > PG (23; D≈0.78). D decreased over time in all regions. Region-specific SB associations (p<0.01): Beja (SB0119, SB0120, SB0140, SB0265, SB0295, SB1172, SB1190); Castelo Branco (SB0122, SB1195, SB1264, SB1266); Portalegre (SB1167, SB1174, SB1483). - Host diversity: Red deer (39 SB profiles) > cattle (38) > wild boar (30); 17 SB profiles shared by all three hosts. Most frequent by host: cattle SB0121 (17.7%), red deer SB0122 (21.8%), wild boar SB1174 (21.8%). Host-specific SB associations (p<0.01): cattle (SB0119, SB0121, SB0140, SB0295, SB1090, SB1172); red deer (SB0122, SB1195, SB1264); wild boar (SB1174, SB1257, SB1264, SB1265). - Networks: Strong within-region connections, especially in CB; cattle showed many shared SB links across regions, but fewer high-strength links when considering combined genotypes. - Outbreak tracing: 48 chronic foci (≥3 isolates in different years; 19 herds, 29 hunting areas) and 14 small outbreaks (single year; mostly herds). Hunting areas yielded higher isolate counts and greater genotype diversity; repeated genotypic profiles within foci indicated persistence/reactivation, while multiple profiles suggested re-introduction/translocations. - Phylogenetics (MST): 14 clonal complexes (CC) covering 87.9% (n=428) of isolates; largest CC held 45% (n=220). CC4, CC5, CC8 exclusive to cattle; CC3 exclusive to red deer. Geographic segregation visible (e.g., CC3/6/12 only in CB; CC5 only in BJ). Star-like expansion patterns in CC1 (two branches), CC2, CC10. - Bayesian STRUCTURE: Best K=5 ancestral M. bovis populations (AM1–AM5). Geographic contrasts: AM1/2/4/5 predominant in CB; AM3 more evenly between CB and PG; AM2 underrepresented in PG; AM3 underrepresented in BJ. Host association: AM4 significantly associated with cattle; AM2 negatively associated with cattle; AM4 negatively with red deer. Allelic richness highest in AM4, lowest in AM5. - Clonal complex context: Many profiles showed spacer deletions consistent with European 2 (Eu2) predominance in Iberia; some evidence of Eu1/Eu3/Af2 signatures, though SNP confirmation not performed. - Overall inference: Phylogeographic segregation with Beja indicated as the most ancient population and Castelo Branco/Portalegre showing expanding, locally adapted, multi-host M. bovis populations.

The integrated molecular and spatial analyses elucidate the spatiotemporal dynamics of M. bovis within Portuguese hotspots and the roles of cattle and wildlife in transmission. Shared genotypes across hosts within the same place and time indicate both intra- and interspecific transmission at the livestock–wildlife interface. SaTScan clustering reveals two distinct epidemiological phases: an earlier cattle-driven cluster centered in Beja (2004–2010) and a later wildlife-driven cluster (especially wild boar) in Castelo Branco (2012–2016), aligning with increased wildlife surveillance from 2011 onward. MST and STRUCTURE demonstrate a largely clonal expansion pattern with geographic stratification, supporting local evolution and limited inter-regional flow. The significant association of the AM4 ancestral population with cattle, and the identification of Beja as harboring the most ancient population, suggest a primary role for cattle as an original reservoir with subsequent spillover and spread, particularly to red deer, while wild boar also contributes importantly to maintenance in certain areas. Reduction in genotypic diversity over time in some regions is consistent with selective sweeps and control pressure in cattle populations, whereas higher diversity and persistence in hunting areas reflect complex wildlife dynamics and management practices. Despite homoplasy concerns with spoligotyping/MIRU-VNTR, combining methods, network analyses, and STRUCTURE yielded congruent results, strengthening the inference of phylogeographic segregation and host associations relevant for targeted TB control.

Long-term molecular surveillance using spoligotyping and MIRU-VNTR across three Portuguese TB hotspots revealed five ancestral M. bovis populations with marked geographic structuring and multi-host circulation. Evidence points to Beja as the most ancient population and a strong association of one ancestral population (AM4) with cattle, supporting a cattle-origin clue for local epidemics, with red deer likely receiving direct transmission and wild boar acting as an important reservoir in specific areas. The work refines the understanding of population structure, documents inter- and intra-species transmission, and identifies spatial clusters and outbreak patterns at herd and hunting-area scales. These insights support adaptive, region-specific control strategies that integrate cattle testing and movement controls with wildlife management and biosecurity. Future research should incorporate whole genome sequencing to resolve transmission at finer scales, quantify inter- versus intra-species transmission, and evaluate environmental reservoirs, alongside longitudinal ecological data on host densities, movements, and management.

- Surveillance and sampling biases: Wildlife sampling was sporadic and uneven before 2011 and dependent on hunting activity and local practices thereafter; hunting bags varied by year and region. Early period (2002–2006) had few isolates (n=85; 9% of dataset). Only 76.4% of cases were geocoded for spatiotemporal analysis. - Methodological constraints: Spoligotyping and MIRU-VNTR are subject to homoplasy and have limited genomic resolution; MIRU analysis primarily used an 8-loci panel; profiles with double alleles excluded from some analyses. No SNP-based confirmation of clonal complexes (e.g., Eu2 via guaA) was performed. - Spatial modeling assumptions: Fixed location centroids assumed stable over time; non-overlapping clusters enforced; potential underdetection of smaller or overlapping clusters. - Generalizability: Study focused on three hotspot districts; findings may not directly extrapolate to other regions without similar host community structures and management practices.