Origin, distribution, and potential risk factors associated with influenza A virus in swine in two production systems in Guatemala

Abstract Background Guatemala is the country with the largest swine production in Central America; however, evidence of influenza A virus (IAV) in pigs has not been clearly delineated. Objectives In this study, we analyzed the presence and spatial distribution of IAV in commercial and backyard swine populations. Methods Samples from two nationwide surveys conducted in 2010 and 2011 were tested using virological (rRT‐PCR and virus isolation) and serological (ELISA and hemagglutination inhibition) assays to detect IAV. Results Influenza A virus was detected in 15.7% of the sampled pigs (30.6% of herds) in 2010 and in 11.7% (24.2% of herds) in 2011. The percentage of seropositive pigs was 10.6% (16.1% of herds) and 1.4% (3.1% of herds) for each year, respectively. Three pandemic H1N1 and one seasonal human‐like H3N2 viruses were isolated. Antibodies against viruses from different genetic clusters were detected. No reassortant strains with swine viruses were detected. The H3N2 virus was closely related to human viruses that circulated in Central America in 2010, distinct to the most recent human seasonal vaccine lineages. Spatial clusters of rRT‐PCR positive herds were detected each year by scan statistics. Conclusions Our results demonstrate circulation of IAV throughout Guatemala and identify commercial farms, animal health status, and age as potential risk factors associated with IAV infection and exposure. Detection of human‐origin viruses in pigs suggests a role for humans in the molecular epidemiology of IAV in swine in Guatemala and evidences gaps in local animal and human surveillance.


| INTRODUCTION
Interspecies transmission events of IAVs between humans and pigs play a significant role in the generation of novel reassortant strains that may spread among humans and/or swine populations. 1,2 The role of humans in the epidemiology of swine IAV has been increasingly recognized due to the accumulating evidence of reverse zoonotic transmission events observed over the past century. 3 Information on the prevalence and distribution of swine IAVs remains limited in many regions of the world, particularly in Latin America. 4 Recent studies suggest that introduction of human viruses into pigs may be a major driver in the evolution of IAV lineages that are exclusive to Latin America. 5 In general, animal husbandry practices in many of these countries resemble those from other regions that are believed to be associated with an increased risk of exposure to zoonotic influenza viruses. 6 In Central America, Guatemala is the country with the largest swine industry (estimated population size >2.7 million). 7 Although large-scale commercial farms with enclosed housing exist, most swine production is peridomestic-in household backyards or open smallholdings-without specialized equipment. In these systems, pigs are often free ranged or kept in contact with other domestic animals. 8 Serological evidence of H1N1 and H3N2 in pigs was documented previously 9 ; however, only a limited number of strains were used which may have resulted in limited detection of antigenic diversity. In Guatemala, vaccination against swine IAVs is generally not practiced, and it is neither recommended nor regulated by animal health authorities. To our knowledge, virus isolation from animal samples has not been attempted in Guatemala; consequently, the genetic diversity and the distribution of circulating virus strains in the country remain unknown. In this study, two nationwide surveys were performed in pigs in Guatemala to detect IAV. Viral infection and serological exposure were investigated using molecular and serological testing. The viruses' origin in the sampled population was identified through phylogenetic analysis. Information available on the type of production system, geographic location, and animal characteristics was used to identify potential risk factors and to analyze the spatial distribution of IAV-positive herds.

| Sample collection
Two nationwide surveys were conducted in Guatemala, one in 2010 (October) and one in 2011 (June-August). To demonstrate the presence of IAV in the pig population in Guatemala, 500 samples per year were collected throughout the country, sufficient to detect 1% circulation of virus with 99% confidence. 10 The samples were distributed proportionally to the swine population, by department (administrative subdivisions in Guatemala) and type of pig production system: smallto medium-scale commercial farms, or backyards. Pig production units  were collected and placed in 3 mL of viral transport medium with antibiotics and antimycotics. 11 Additionally, 2 mL of blood was collected from the orbital sinuous vein for antibody detection. Information on potential risk factors at the animal level (animal health status, age, and sex-only done in 2011) and at the herd level (herd size, type of PPU, geographic location) was collected at the time of sampling. Serum samples were tested for antibodies against IAV with the commercially available kit, IDEXX ELISA Influenza A Ab (IDEXX, Westbrook, ME). The cutoff value was validated and adjusted. 16 ELISA-positive samples were tested by hemagglutination inhibition (HI) assay with standard protocols, 11 against selected swine and human H1 and H3 viruses from different genetic clusters. Samples were considered positive to the antigen with the highest inhibition titer, and exposure to multiple viruses was considered positive when the inhibition titer was the same for more than one reference antigen and when positive to multiple subtypes.

| Statistical analysis
The percentages of positive and seropositive pigs detected by RT-PCR and ELISA were computed by year, type of PPU, and other collected variables. At the animal level, potential risk factors were tested by generalized estimating equations (GEE), to account for clustered observations from the same herd. An exchangeable correlation structure was assumed using robust variance estimates. 17 A bivariate model was considered using each risk factor (animal health status, age, or sex) as independent variable and virus detection (rRT-PCR) or exposure (ELISA) as the dependent variable. Due to demographical differences between the sampled populations, these associations were

| Spatial analysis
For spatial analysis, it was assumed that sampled PPUs are a representative spatial sample of the distribution of swineherds in Guatemala. The  20 Areas with high positivity rates were scanned, using a Bernoulli distribution as the probability model. An elliptical window shape was used with a maximum spatial cluster size of 50% of the population at risk and 999 Monte Carlo randomizations. 21 A robust standard error was used to account for the corrections made in the geographic coordinates, and clusters were considered significant when P<.1.
Sensitivity analysis was performed using a circular window shape and different maximum scanning window sizes to test for robustness of the clusters found. The clusters were mapped in Manifold 8 ® system.

| Virus sequence characterization and phylogenetic analysis
The  Table S1). Phylogenetic analyses were performed in MEGA 6.0. 23 Sequences were manually trimmed and final coding sequences were aligned with MUSCLE for codons. After alignment, datasets were subsampled to remove identical sequences and reduce sampling bias. Final phylogenetic trees were constructed using maximum-likelihood (ML) inference with the best-fit model of nucleotide substitution determined by the BIC criterion and Hasegawa-Kishino-Yano (HKY) with gamma distribution. Robustness of tree topologies was assessed with 100 neighbor-joining bootstrap replicates.

| Sample collection
Samples were collected from 188 herds in 2010 (500 pigs) and 199 herds in 2011 (499 pigs) ( Table 1). Summary statistics of the herd sizes and type of sampled PPUs are shown in Table S2. In 2010, sampled PPUs included commercial farms (45%, n=85, 329 pigs) and backyards (54%, n=101, 169 pigs). In addition, two samples were submitted from one agricultural school and one PPU not identified as farm or backyard.
In 2011, samples were submitted from 53 commercial farms (27%, 230 pigs), 141 backyards (71%, 257 pigs), one abattoir (4 pigs), an agricultural school (2 pigs), and two unidentified PPUs (5 and 1 pigs, respectively). Samples that were not from farms or backyards were tested in the laboratory, but they were excluded from all statistical analyses. Information on animal sex and age was collected only in 2011 (Table 1).

| Virus detection and spatial analysis
Influenza A virus in pigs was detected in all departments in both years, with the exception of two departments each year (Tables S3 and S4). Herd spatial clusters were located and tested by scan statistics.
No space-time clusters were observed; therefore, purely spatial analysis was performed for each year. In 2010, one cluster (P=.057) was observed located in the western part of the country (Table 3 and In 2011, a major cluster was found (P=.0075), located in an area that included at least 13 departments. This cluster had a circular shape with a diameter of 89.98 km and comprised 43 herds, from which 27 were positive for IAV (Table 3 and Figure 1, panel B).

| Serologic detection and HI assay
With respect to IAV seroprevalence, 10 Serological exposure to different subtypes varied between types of PPU: Exposure to H1 viruses of swine origin (including the α and ɣ clusters) was detected in commercial farms, whereas in backyards only exposure to pandemic H1 was found. Exposure to H3 clusters III and IV was detected in commercial farms in 2010 as single or multiple exposures. In backyards, exposure to H3 viruses was only found in samples with co-exposure to pandemic H1 (Table 4).

| Virus isolation, sequence characterization, and phylogenetic analysis
Unknown Weanling Juvenile   (Table S5), including two non-synonymous mutations in the HA (V251L and R222K), one in the PA (E688G), and two in the NA (G298A and a mixed base E462D). All of these mutations correspond to residues that are more frequently found in H1 viruses of swine    isolate was obtained from the department "El Progreso" where, according to data from the ministry of health, the reported incidence of acute respiratory infection in humans was higher than the national average in 2010. 37 With the current amount of IAV isolates and epidemiologic information available from pigs and humans, it is hard to analyze the extent of cross-species transmission between these hosts.
Future studies should focus at the swine-human interface to address these questions.  With respect to the location of the clusters, the cluster observed in 2010 is located in the highlands (1000 to 3000 MASL) where annual average temperatures range between 7 and 22°C. 39 The lower tem-