Mycoplasma hyopneumoniae is the primary agent involved in porcine enzootic pneumonia and the Porcine Respiratory Disease Complex. Infections with this bacterium are highly prevalent in almost all swine-producing areas, causing significant economic losses to the pig industry worldwide (1). Information on the genetic diversity of M. hyopneumoniae can have implications in the design, identification, and application of pig health intervention strategies, such as emergence of new and drug-resistant strains, development of precision diagnostics, therapeutics and vaccines, as well as epidemiological studies.

To identify genetic variability in M. hyopneumoniae, different typing schemes have been reported, such as Random Amplification of Polymorphic DNA [RAPD-PCR (2)], Amplified Fragment Length Polymorphism [AFLP (2, 3)], Arbitrarily Primed-PCR [AP-PCR (4)] and Pulsed Field Gel Electrophoresis [PFGE (5)]. Although some of the above mentioned techniques have demonstrated high discriminatory power, currently, they are not widely used. In addition, Multiple-Locus Sequence Typing (MLST) and Multiple-Locus Variable-Number Tandem Repeat Analysis (MLVA) have been or are used for genetic typing of this microorganism worldwide (6–9, 43, 45, 46). While MLST can be laborious, costly, and has mainly been employed for research purposes (10–14), MLVA can be applied with ease, is less expensive, and has been increasingly used for M. hyopneumoniae typing (7, 9, 15, 16), even in field based investigations (17, 43). From a proposed repertoire of twenty-two Variable Number Tandem Repeats (VNTR) regions, four have proven to be highly polymorphic (6), namely p97R1, p146R3, H4 and H5. Although analyzing the four loci of the MLVA scheme has shown a high discriminatory power (6, 18), the detection of different M. hyopneumoniae genetic types has been evidenced using three (15, 19), two (7) or even one VNTR (20, 21).

Numerous epidemiological studies have been carried out using different MLVA schemes in various countries, at farm, regional and/or country level (7, 15–17). However, a global perspective on the implications of the genetic diversity of this etiological agent is lacking. To this extent, M. hyopneumoniae has been pointed out as a microorganism with limited clonality (13). Notwithstanding, information is needed on the overall allele frequency, distribution of MLVA types, and genetic structure of this pathogen. Therefore, the objectives of this study were to determine and to analyze the genetic structure of M. hyopneumoniae based on the p97R1-p146R3 MLVA typing scheme and to characterize, analyze and compare MLVA types among countries where the information was publicly available.

Locus p97R1 revealed 20 alleles, while locus p146R3 revealed 38 alleles, with tandem repeats (VNTRs) ranging from 10 to 48 (Supplementary Table S1; Figures 1, 2). For p97R1, the VNTR types with 9, 10 and 11 tandem repeats were identified in all countries, while unique alleles were reported in the United States (VNTRs 4, 16, 18 and 19), Belgium (VNTR 1), and Mexico (VNTR 20). For p146R3, the only type shared in all the countries was VNTR 17, while unique types were observed in the United States (VNTRs 10 and 11), Spain (VNTRs 30, 36, 46 and 47), and Brazil (VNTRs 33 and 48) (Supplementary Table S1). For p97R1, the country with the highest number of alleles (Na), effective alleles (Ne) and most diversity (h) and unbiased diversity (uh) by population was the United States (Table 2). For p146R3, the country with the highest number of alleles (Na) was Brazil. The highest number of effective alleles (Ne) and the most diversity (h) and unbiased (uh) diversity by population was Belgium (Table 2).

A total of 185 M. hyopneumoniae typologies were identified among the 797 VNTR types analyzed. The most frequent VNTR types (p97R1-p146R3, respectively) were: 9–15 (8.4%), 11–21 (7.4%), 14–14 (6.5%), 13–17 (3.4%), 9–20 (3.3%), 16–24 (2.9%) and 8–16 (2.4%), which were present in all countries included in the study, except Belgium (Supplementary material S2). The remaining VNTR types were identified in less than 2.1% of the analyzed data. The 44.3% (82/185) of the identified types were deemed unique and were distributed in all countries (Supplementary material S2). In Argentina there were 13 VNTR types, being the most frequent 8–16 (27.7%) and 11–26 (15.4%). In Belgium, 20 types were recognized, with 12–17 (19.4%) and 7–34 (9.7%) being the most prevalent. In Brazil 52 typologies were identified, the most frequent: 12–14 (24.2%) and 11–21 (13.6%). In Mexico 12 types were observed, with 10–18, 10–19 and 10–17 in 30.8, 15.4 and 11.5%, respectively. In Spain 32 types were identified, including 9–20 (34.8%) and 11–46 (14.5%). In the United States a total of 101 types were observed, the most prominent being 9–15 (16.2%) and 11–21 (7.6%) (Supplementary material S2). All MLVA types identified in this investigation were not present in all countries, with no type present in more than three countries.

Similarity coefficients of Nei’s unbiased genetic distance between M. hyopneumoniae populations ranged from 0.471 to 1.943 (Table 3). The highest distance was observed between Mexico and Argentina (1.943%), while the highest Nei unbiased genetic identity was observed between Brazil and Belgium (0.6%; Table 4).

Results of the AMOVA indicated that, for both loci, most of the molecular variation in M. hyopneumoniae populations (65%), existed within populations (countries), with less variation between regions (28%) and among populations (7%). For the p97R1 locus, most of the variation existed within populations (90%) and 10% among populations, with no difference between regions. For the p146R3 locus, most of the variation existed within populations (61%), with less variation among regions (32%) and populations (7).

Analyses revealed three genetic clusters, with log posterior probabilities of −3611.72 ± 2.83 for the 797 types (Figure 3). Different countries in the study were marked in Figure 3 for ease of visualization, although a geographical differentiation was not set as part of the priors of the analysis. The frequency of each of the three clusters varied considerably among countries (Table 5). VNTR types whose genome belonged to a single cluster were deemed “pure” and typologies with admixed ancestry were identified in all clusters. At a general level, there were more VNTR types with admixed ancestry (43.3%) than pure types (Table 5). The largest proportion of VNTR types belonging to cluster I were reported in Brazil (41.9%), and in Spain for clusters II and III (34.8 and 20.3%, respectively). Mexico showed the largest proportion of admixed ancestry types (80.8%).

Bacterial population structures range from panmictic or non-clonal, to strictly clonal, but intermediate structures have also been recognized, such as epidemic (32). The different population genetic structures may be due to events like genetic recombination, geographical or ecological isolation, genetic drift, horizontal gene transfer, among others (32–35). While Salmonella enterica is a model of a clonal population, Neisseria gonorrhoeae and Helicobacter pylori are representative of non-clonal models (36). In the former, the same or very similar types of individuals can cause disease and/or outbreaks, and in the latter genetically unrelated individuals of a given type lead to malaise.

In the present study, the fact that more than 40% of the analyzed VNTR types belonged to admixed ancestry, three defined genetic clusters were identified, differences in diversity among countries was minimal, and no p97R1-p146R3 MLVA type was reported in all countries suggests that the genetic structure of M. hyopneumoniae is predominantly, but not exclusively, panmictic or non-clonal. On the other hand, the abundance of VNTR types, in addition to unique alleles identified in different countries strongly suggests that M. hyopneumoniae is a highly diverse microorganism.

Our findings indicate that three M. hyopneumoniae MLVA genetic clusters, which possibly evolved separately, are now seen combined globally, or at least in the analyzed countries. A similar pattern has been described for M. bovis, where three genetic clusters have been reported, with a preponderance of two (37). In the above mentioned study, the authors used MLST typing for 137 M. bovis isolates from 12 countries and observed the first cluster including isolates from multiple geographical origins (United Kingdom, Lithuania, Hungary, Germany, Italy, Israel, and Australia) and European origin only (Spain, Germany, France, and Lithuania). The second cluster including also isolates from several origins and all the strains from China and most of those of Israeli and Australian origin, and the third, including only two Israeli isolates (37). Moreover, similar patterns of geographical distribution of M. bovis isolates has been recognized using the MLVA typing method (38).

Numerous sequence-based typing approaches for M. hyopneumoniae have been developed, being MLST and MLVA the most commonly used currently. However, the use of one or the other techniques should not be arbitrary. While the MLST scheme, which like Multiple Locus Enzyme Electrophoresis (MLEE) detects DNA changes in conserved genome areas, MLVA was developed for high discrimination between closely related strains (39). This information, combined with the intermediate genetic structure of the agent might explain some incongruities reported to date. For instance, while some studies using MLST associated enzootic pneumonia (EP) with identical or very similar M. hyopneumoniae genotypes (12, 13, 40), others using MLVA typing reported a higher prevalence and severity of EP-like lung lesions with higher diversity of the pathogen (41). Particularly for M. hyopneumoniae, the use of the MLVA assay has been recommended for differentiation of strains with identical MLST sequence types (10). Considering that the MLST scheme is useful for phylogenetic analysis, but often not discriminative enough for outbreak detection (39), and that MLVA outcomes can be highly polymorphic, the use of both typing approaches in combination for EP outbreak studies could yield insightful information.

In spite of the fact that VNTR types with mixed ancestry were identified in the six analyzed countries, in Mexico, it reached almost 81%. The source of this type of VNTR could be due, for instance, to genetic drive or horizontal gene transfer. Future studies are needed for a better understanding of this phenomenon. In addition to this, Mexico is the only country among those analyzed in which a live attenuated M. hyopneumoniae vaccine has recently been licensed (42). Further longitudinal studies to evaluate the stability of VNTR types from M. hyopneumoniae hosted in pigs vaccinated with live vaccines should be conducted. Interestingly, the VNTR type of the J strain (9–18), the most used strain for inactivated vaccines against EP worldwide (18), did not match exactly any of the analyzed VNTR types. Similar types, namely 9–17, 9–19 and 8–17, were also identified in Argentina, Belgium, Spain and the United States. Contrary to what was observed in Mexico, in Brazil more VNTR types belonging to cluster I than with admixed ancestry were identified, suggesting that gene flow may be restricted between the Brazilian M. hyopneumoniae populations and those of other countries. In addition, if the Brazilian populations originated from a single source of infection and if not many types with a different genetic composition have entered the country, it could be expected for the M. hyopneumoniae populations to have a predominance of a single cluster with very little presence of the other genetic combinations, which has been observed in this study.

Results from this investigation suggest that most variation in M. hyopneumoniae types (61–90%) was observed within populations (countries). Although it has been suggested that a certain M. hyopneumoniae MLVA type predominates and persists within each herd (17, 19), circulating M. hyopneumoniae genotypes are abundant, at herd and even at pig level (15, 43, 44). The molecular variation among populations (7–10%) was probably due to the existence of unique alleles in some countries and to that a small number of p97R1 and p146R3 alleles were present in all countries. Results also indicated that for both loci in combination, and for the p146R3 locus alone, approximately 30% of variation was observed between regions (America vs. Europe), which was probably due to the high polymorphism of the p146R3 locus (6). A practical implication of this finding would be to use p146R3 together with other polymorphic loci when making comparisons between the two continents.

In conclusion, this study reinforces the concept that M. hyopneumoniae is a genetically diverse pathogen with limited clonality. When typed by MLVA, three well-defined populations or clusters and their combinations were identified. The major cause of genetic variation of M. hyopneumoniae might be within populations, recognizing the bias in the analyzed data, as not all countries were analyzed using the same number of farms, nor randomly. This study constitutes a starting point for future investigations to provide greater detail of the molecular epidemiological behavior of this pathogen. The difficulty in culturing and isolating M. hyopneumoniae strains and the scarcity of genomes available, continues to be a limitation for the application of whole genome sequencing (WGS). Analysis of WGS of this pathogen could offer greater insight into genetic variability than that of the comparison of a small number of loci.