Pig farmers are exposed to a lot of potentially hazardous substances during their daily routine, including high concentrations of bacteria which are present inside pig confinement buildings (1–3
). Many studies have monitored the overall air quality inside pig farms (4–9
) and the possible transmission of zoonotic bacteria (10–14
). The majority of studies have focused on sampling during a very short time period or even only one point in time and therefore do not address possible seasonal variations in pig farms. In Switzerland, there is a mean temperature variation of 18°C throughout the year, with low temperatures in winter, high temperatures in summer, and intermediate temperatures in spring and fall (15
). These temperature shifts lead to different behaviors and farm management habits in the pig farming industry. In winter, openings such as doors and windows are closed, and a lower ventilation rate is taken up to minimalize heat loss. There are studies that have investigated seasonal variabilities of bacteria, archaea, and fungi in the air of pig buildings, and they have found a seasonal effect (16–19
). However, none of these studies have investigated the influence of this variability on the human nasal microbiota. Recent studies have found evidence that there is substantial exchange of microbial communities among humans and the animals with which they are in close contact (14
). In a previous study, we also showed the presence of a specific microbiota pattern for each investigated pig farm in winter by a cross-sectional approach (14
). The present study used a longitudinal approach to address the seasonal variability of the nasal microbiota of pig farmers as well as the nasal microbiota of pigs and the air inside the pig confinement building. We utilized Illumina MiSeq sequencing in combination with the recently developed DADA2 pipeline (22
). The present study was performed to address the following questions. (i) What changes occur in the human nasal microbiota throughout the seasons in individuals with and without close contact with farm animals? (ii) Are these potential changes driven by the airborne microbiota and are they farm specific?
The understanding of the formation and function of microbial communities in built environments and how this impacts human health is important. In this study, we investigated the influence of seasons on the composition of microbial communities from air samples taken inside the pig barn and from nasal samples from pigs, pig farmers, cow farmers, and individuals who were not exposed to farm animals or the farming environment. We detected changes between seasons in both alpha and beta diversity analyses. Furthermore, we observed that the microbial composition was influenced by sampling location for samples from pig farms and from cow farmers.
made up the majority of the microbiota in air samples from inside the pig building during all seasons. A high Firmicutes
abundance in swine confinement buildings was also found in a previous study (23
). In another study, the outdoor bioaerosols in urban and rural environments were investigated, and the taxonomic profiles were shown to be quite similar throughout the year. However, the phylum Firmicutes
was not as abundant as we have observed inside pig farms (24
). This might indicate that the composition of the microbiota of the indoor air of pig farms is not substantially influenced by the outdoor airborne microbial communities, though we did not have outdoor samples to further explore this hypothesis in our study.
, and Actinobacteria
were the most abundant phyla in the nasal samples from humans for all seasons in our study. This is consistent with other studies that investigated the microbiota of the nasal vestibule, the nasopharynx, and the trachea in a longitudinal manner (25–27
). However, there was a significantly lower alpha diversity in summer in nasal samples from pig and cow farmers than in winter in our study. This change in alpha diversity was not seen in nonexposed controls who probably spent most of their time in indoor environments (offices and homes). Importantly, air samples and nasal samples from pigs showed the same trend as the nasal swabs from pig farmers, with increased alpha diversity values in winter. Previous studies also found higher diversity in winter than in summer for both bacteria and fungi in the air in pig confinement buildings (16
We previously revealed strong clustering based on sample type (14
), and the longitudinal data presented here demonstrate that this applies regardless of season. Nevertheless, a significant effect of season on the microbiota was observed in the human-derived sample types from pig and cow farmers but not nonexposed individuals. In young children and adults, it has also been shown that season acts as a major factor in driving the dynamics of the nasal microbiota (25
). In contrast to our findings, De Boeck et al. reported no correlation between season and microbial composition in healthy humans (29
). However, they only collected one sample per study participant, and it is likely that the interindividual variation was too strong to detect seasonal correlations.
In winter, samples from the same farm share more of their microbiota than do samples from different farms, indicating the presence of farm-specific microbiomes. Depending on the sample type comparison, this strong specificity is partially or fully lost in the following seasons. These differences can also be explained by the aforementioned hypothesis. Due to the low temperatures in winter, windows and especially doors are opened less frequently, and the airborne bacteria from the indoor air may be less likely to be diluted with outdoor air. This would increase the impacts of humans, pigs, and possibly various other sources in the pig building on the microbial composition and would ultimately promote the emergence of a farm-specific microbiota. A higher frequency of opening doors in spring, summer, and fall would lead to a decrease of this farm-specific effect.
A main strength of this study is its longitudinal nature, as this made it possible to account for the impact of the individual, such as from lifestyle choices or host genetics, on the nasal microbiota. We were able to correct for clustering on the level of the individual to accurately investigate the impact of season and location for pig farmers. By collecting different sample types from the same farm throughout the year, it was possible to show that the farm-specific effect is strongest in winter and partially or even completely lost in the subsequent seasons. A further strength of the study is that the recruitment of cow farmers and individuals without contact with farm animals as control groups made it evident that the nasal microbiota of people with a farming lifestyle are more heavily affected by seasonal changes than those of individuals who spend the majority of their lives in an indoor environment.
Finally, there are also limitations. A typical disadvantage in longitudinal studies is the difficulty of collecting follow-up samples (30
); some individuals were excluded, as they either dropped out of the study after one season of sampling or were found to be unsuitable in the downstream analysis. We were also unable to assess an effect of season for identical pigs, since they were difficult or impossible to follow-up after they were dispatched to other farms or slaughtered.
In conclusion, we showed that season has a significant effect on the human nasal microbiota and that this effect is much more distinct in individuals who work on farms with livestock. Future research should be conducted to investigate which factors are responsible for this strong impact of season.
MATERIALS AND METHODS
Ethical clearance for this study was obtained from the Human Research Ethics Committee of the Canton Vaud (243/14 and P_2017-00265) and the Veterinary Ethics Committee of the Canton Vaud (VD2903). The sampling locations were visited four times during the sample period: in winter (October to March), spring (April to May), summer (June to August), and fall (September). The periods of the seasons were chosen on the basis of previously observed temperature patterns in Switzerland based on meteorological data from the last ten years (15
). Winter and summer were characterized with cold (mean, 8°C) and warm (mean, 25°C) temperatures, respectively, while spring and fall were represented by moderate temperatures (mean, 18°C). Among the overall 30 visited farms, 10 were naturally and 20 were mechanically ventilated. With the exception of very few, windows were always closed in winter and also during the other seasons. However, all the doors were systematically opened during the hot days (always in summer and sometimes also in spring and fall).
Sampling of nasal swabs and air.
All pig farms (n
= 30) were visited, and on each farm, nasal samples from suckling or weaning pigs and pig farmers as well as air samples from inside the pig confinement building were collected as previously described (14
). As controls, two groups were sampled from the same geographic area and the same socioeconomic level. First, samples were collected from cow farmers who have a farming lifestyle and contact with cows but no contact with pigs. For the second control group, samples were taken from nonexposed individuals who work in offices with no farming lifestyle and no contact with any type of farm animal. All human participants provided a nasal sample taken from the anterior cavity; the participants took the samples themselves using a sterile cotton swab under the supervision of the study personnel.
Questionnaires were administered individually to each participant to determine health status. The same human individuals but different pigs were sampled for each season. Several pig farmer and pig samples originated from the same location, and several cow farmers also shared a location. Each nonexposed sample originated from a unique location, as all of these nonexposed participants lived and worked at independent locations. All samples were immediately transferred to 4°C after collection and then were stored at −20°C within 4 h.
Particle sampling of the air and 16S rRNA quantitative PCR.
A direct-reading portable optical counter, model Grimm 1.109 (Grimm Aerosol Technik GmbH & Co., Ainring, Germany), was used to measure particles inside pig confinement buildings. We measured mass fraction of inhalable particles (PM10). Units were expressed in microgram per cubic meter, and the log-transformed values were visualized. Sampling was performed in the central pen at a height of approximately 1 m for 10 min (one read every 10 s). The average PM10 concentrations were calculated for each farm, and seasonal differences were determined with a linear model (lmer function). Simultaneously and at the same location, airborne bacteria were sampled with a Coriolis μ air sampler (Bertin Technologies, Montigny-le-Bretonneux, France) at a flow rate of 300 liters/min for 10 min. Airborne particles were collected into a sterile cone containing 15 ml 0.005% Triton X-100 solution. Repeated sampling was not performed on the same day or during the same week to assess variations. However, in the international ISO norm (ISO16000-18) concerning the sampling of cultivable airborne fungi, it is mentioned that either 4 or 5 replicates representing a total volume of 300 liters are sufficient to assess daily exposure and variability, which is 10 times less than in our study (3,000 liters in total).
The bacterial load inside the pig confinement buildings was ascertained by 16S quantitative PCR (qPCR). In brief, DNA was extracted from the Coriolis air samples by using a Qiagen DNA Minikit (Qiagen, Hilden, Germany) and was subsequently used to quantify the copy numbers of total bacteria by using universal primers for the V9 region of the 16S rRNA gene (forward, 3′-CGGTGAATACGTTCYCGG-5′; reverse, 3′-GGWTACCTTGTTACGACTT-5′) (31
). The 25-μl qPCR mixture contained 12.5 μl of 2× SYBR green PCR master mix (Applied Biosystems, Foster City, CA, USA), 2.5 μl each of the 10 μM forward and reverse primers, 5 μl DNA template, and 2.5 μl of sterile DNA-free water. The experiments were carried out on the QuantStudio 7 Flex real-time PCR system using 0.1 ml MicroAmp Fast 96-well reaction plates (Applied Biosystems, Foster City, CA, USA). In each run, a standard curve was included, using serial 10-fold dilutions of Escherichia coli
DNA to determine the gene copy numbers. A negative control was included in every run. The following cycling conditions were used: 95°C for 10 min, followed by 40 cycles of 95°C for 15 s and 60°C for 1 min. The results were collected using the QuantStudio 6 and 7 Flex software. Each sample was measured in triplicates. Samples with high threshold cycle (CT
) values (>35) were excluded from analysis (17/105 [16.2%]). The 16S rRNA gene copy numbers were logarithmically transformed for analysis, and seasonal differences were determined using a linear model (lmer).
Sample processing and 16S rRNA amplification for microbiota analyses.
Sample processing for 16S rRNA microbiota analyses from nasal swabs and air samples was performed during the first 8 weeks of storage at −20°C. DNA extraction, amplification, and Illumina MiSeq sequencing were conducted as previously described (14
). Briefly, DNA was extracted using the Qiagen DNA Minikit (Qiagen, Hilden, Germany), and the V4 region of the 16S rRNA gene was amplified using forward (5′-GTGCCAGCMGCCGCGGTAA-3′) and reverse (5′-GGACTACHVGGGTWTCTAAT-3′) primers previously described (32
). Amplicons were submitted to the Next Generation Sequencing Platform at the University of Bern for indexing and pair-end sequencing (2 × 250 bp) on the Illumina MiSeq platform (San Diego, CA, USA). Samples with the following characteristics were excluded from the study: samples taken from individuals with antibiotics intake during sampling, samples from pig farmers working with pigs for less than 6 months, samples with a DNA concentration below 1 ng/µl after PCR and purification, and samples with less than 3,300 high-quality sequence reads after sequencing. In addition, samples from 11 individuals (2 cow farmers and 9 nonexposed) were excluded, because only one of the four samples remained after sample processing.
Several quality-control samples were included during our sampling and processing procedures. A clean nasal swab tip which was exposed for several seconds during sampling was used to check for contamination. In addition, an extraction control (200 µl phosphate-buffered saline [PBS]) was included for every batch of 60 samples, and a PCR control (10 µl sterile water) was included for each amplification batch to confirm that there was no reagent-based contamination.
Reads were analyzed using the DADA2 package version 1.5.0 and workflow (22
) in R version 3.1.2 (http://www.R-project.org
) as previously described (14
). The sequences were not rarefied for downstream analyses, since the DADA2 algorithm drastically reduces the issues of having different sequencing depths for the samples being compared, which is the main reason for rarefying (14
Diversity and source tracker analyses.
All calculations were performed in R utilizing the vegan package, and all graphs were created using the ggplot2 package.
Significant differences in phyla and families were assessed via ANOVA-like differential expression (ALDEx) analysis in R using the aldex2 package. Instances of the centered log-ratio transformation values were generated (aldex.clr function), and significant differences were assessed. Overall significant differences were investigated via an omnibus test (generalized linear model and Kruskal Wallis tests for one-way ANOVA with Benjamini-Hochberg [BH] correction [33
]; aldex.glm function), and significant differences between seasons were assessed using Wilcoxon rank tests with BH correction (33
) (aldex.ttest). Alpha diversity was assessed with the functions estimate (richness) and diversity (Shannon diversity indices [SDIs]). Significant differences between seasons were assessed with linear regression models with random effects to correct for clustering on location level, as several samples originated from the same location, and to correct for clustering on the individual identification (ID) level, as there were multiple samples from the same human individual (one sample per season). The regression model was calculated with the lmer function from the lmeTest package.
Beta diversity was assessed by unweighted Jaccard index (presence-absence based) and weighted Ružička index (abundance-based) of dissimilarity, and the distances were calculated using the vegdist function. Significant clustering between seasons was determined by permutational multivariate analysis of variance using 1,000 Monte Carlo permutation tests (PERMANOVA; adonis function) followed by Benjamini-Hochberg (BH) correction for multiple testing (33
). The post hoc
tests were performed with stratification for clustering on season level for pig and air samples, as the pig samples came from a different animal each season and the air sample was not taken at the exact same spot each time. Therefore, it was not possible to calculate the influence of season on variance for these sample types.
Significant differences between the seasons in the within and between farms dissimilarity plot and differences between sample types in the between seasons plot were assessed by Kruskal-Wallis rank sum tests (kruskal.test function) and post hoc
Mann-Whitney test (wilcox.test function) with Benjamini-Hochberg (BH) correction for multiple comparisons (33
). The SourceTracker software package was used to determine the potential sources of bacteria in pigs and air and their importance in the pig farmers’ noses sampled (34
The BioProject study accession number is PRJEB26637