Bacterial profiles of the oral, vaginal, and rectal mucosa and colostrum of periparturient sows

The commensal microbiota influences the health, feeding efficiency, and reproductive performance of sows. The microbiota composition in the alimentary and genitourinary tracts and in colostrum/milk during pregnancy and lactation also impacts the developing microbiota and immune system of the piglets and subsequently their growth and health. Knowledge of the microbial compositions is important for evaluation of these effects and for discovering ways to improve the health and productivity of the sows. Oral, vaginal, and rectal mucosa were sampled from 32 sows of variable parity in late pregnancy on four commercial piglet-producing farms in Finland. Colostrum samples were taken within 6 hours of delivery of the first piglet. Microbial compositions were analyzed by 16S rRNA gene amplicon sequencing. Moderate differences in diversity and composition were observed between farms. The most abundant genera of the oral microbiota were Rothia, Moraxella, and Streptococcus. The rectal microbiota was dominated by Clostridium sensu stricto 1. Streptococcus was the most abundant genus in the vagina and colostrum. Differences in relative abundances of genera were detected between primiparous and multiparous sows. Some of these differences were detected across all the farms; in the multiparous sows, the relative abundances of Fusobacterium and Neisseria were lower in oral samples. Clostridium sensu stricto 1, Romboutsia, and Lachnospiraceae_UCG_007 were higher and Prevotella lower in rectal samples, and Streptococcus higher in colostrum samples. In vaginal samples, approximately half of the multiparous sows had significantly higher relative abundances of the genera Fusobacterium and Streptococcus than the primiparous sows. Among the differentially abundant taxa, F. necrophorum and F. nucleatum were identified in oral samples, Fusobacterium gastrosuis and Fusobacterium necrophorum in vaginal samples, and Streptococcus dysgalactiae in colostrum samples. Most of the differences were due to unidentified species within the respective genus. This study provides a comprehensive overview of the mucosal and colostrum microbiota of periparturient sows during normal production conditions on Finnish commercial farms, including potentially interesting differences in the relative abundances of several genera at different mucosal sites between sows of low and high parity.

This study provides a comprehensive overview of the mucosal and colostrum microbiota of periparturient sows during normal production conditions on Finnish commercial farms, including potentially interesting differences in the relative abundances of several genera at different mucosal sites between sows of low and high parity.

Background
The mucosal surfaces, skin, feces, and glandular excretions of the mammalian body are all inhabited by distinct communities of resident microbes.Once established early in life, these microbiotas are relatively stable [1][2][3].Commensal microbes provide health and wellbeing benefits to the host by improving maintenance of the epithelial barrier [4] and providing protection against pathogens [5].They also influence host immune development and immune functions [6][7][8][9] and feed utilization [10].Furthermore, commensal microbes have effects on neural development and function and may even impact host mood and behavior [11,12].
Due to the potential benefits of the microbiota on the health and performance of the host, there has been a growing interest in the composition of the commensal microbiota of production animals, including pigs.These studies have mostly focused on growing pigs [13,14], but the commensal microbiota of pregnant sows has also gained attention to an increasing extent [15][16][17][18][19][20][21][22][23][24].Physiological factors, including phase of the reproduction cycle [16][17][18][19] and parity [21][22][23], can modulate the microbial compositions of the adult animals.
Environmental factors, such as diet, stress, and exposure to antibiotics [13,[25][26][27], can also have impacts.The fecal, vaginal, and colostrum (or milk) microbiotas of the sow in turn influence the developing microbiota and immune system of the offspring during pregnancy and lactation [6,22,24,28,29].This has long-term effects on the health and growth of the piglets [6,30].Thus, knowledge on sow microbiota composition is valuable in evaluating its potential benefits and disadvantages to the offspring.
In this study, we explored the bacterial profiles of the oral, vaginal, and rectal mucosa and colostrum of 32 periparturient sows using 16S RNA gene amplicon sequencing.The study population included both primiparous and multiparous sows from four commercial Finnish piglet-producing farms.The bacterial profiles were compared between farms, parity groups, and individuals.

Overview of sequencing data
16S rRNA gene amplicon sequencing of sow oral, vaginal, and rectal mucosa and colostrum samples (n = 126) resulted in 6 749 409 high-quality reads, with an average of 53 567 (+ SD 12 960) reads per sample.The reads were mapped to 5379 ASVs.Details of the sequencing data for each sample type are shown in Supplementary file 1, Supplementary table 1.

Sow oral microbiota
The median alpha diversity of farm D samples was higher than that of the others (Fig 2a

Sow vaginal microbiota
There were modest differences in the microbial diversities between the sows from the different farms, with farm A and C samples having higher alpha diversity than the others (Fig

Sow colostrum microbiota
Although colostrum samples from farm A showed higher alpha diversity than others, the difference from the other farms was not significant (Fig 5a).In PCoA, most of the samples

Parity effects on microbial compositions
The alpha diversities between primiparous and multiparous sows were not significantly different, and the groups overlapped in PCoA (Supplementary file 2,  In oral samples, the RAs of the genera Fusobacterium and Neisseria were significantly lower in multiparous compared with primiparous sows.The lower RA of the genus Fusobacterium in the oral samples of multiparous sows was due to ASVs likely representing several species.Among these, F. necrophorum and F. nucleatum were unequivocally identified based on BLAST sequence comparisons; the others were unknown.ASVs likely representing several species of Neisseria had lower RAs in multiparous sows.These were unknown or the identification was ambiguous.
In vaginal samples, there was a marked tendency towards higher RAs of the genera Fusobacterium and Streptococcus in multiparous compared with primiparous sows (Fig 6b Right panel).Around half of the samples in multiparous sows had higher RAs of these genera than primiparous sows.The RA of Fusobacteria was consistently higher in multiparous than in primiparous sows in the samples from all the four farms, although the difference was not statistically significant.The RAs of Streptococcus were higher in multiparous sows on three farms.The differences of the RAs of the other genera were either not significant between the groups or were not observed in all the farms.The differences in Fusobacteria were due to ASVs that were identified as Fusobacterium gastrosuis and Fusobacterium necrophorum.The RAs of several species of the genus Streptococcus differed between the groups.
In rectal samples, the RAs of the genera Clostridium sensu stricto 1, Romboutsia, and Lachnospiraceae_UCG_007 were significantly higher and Prevotella lower in multiparous compared with primiparous sows (Fig 6c Right panel).The difference was observed in samples from all the four farms.In contrast, the difference in the RA of Lactobacillus between the groups displayed in dbRDA was mainly due to the exceptionally high RA of Lactobacillus in only one primiparous sow.None of the ASVs with different levels in the two groups could be assigned to known species.
In colostrum, the RA of the genus Streptococcus was higher in multiparous than primiparous sows.This was due to colostrum samples from three multiparous sows having RAs of This was due to the high RA of an ASV matching the sequence of Streptococcus dysgalactiae in these samples.The average RA of this species in the other colostrum samples of multiparous sows was not higher than that of primiparous sows and remained < 0.2%.No clear relation in the abundances of S. dysgalactiae in colostrum versus vaginal samples for the affected individuals was detected.Sow D6-01 had higher than average RA (16%) of S. dysgalactiae also in the vagina, while the RAs for the other affected individuals were < 0.7%.ASVs matching to S. dysgalactiae were not detected in any of the rectal samples.

Discussion
The aim of this study was to obtain an overview of the mucosal and colostrum microbiota of sows on Finnish commercial farms under normal production conditions.The other studies in this field are from settings that were different with respect to geographical location, climate, animal breeds, and animal husbandry.We characterized the microbial communities present on oral, vaginal, and rectal mucosa of late pregnant sows and in colostrum from four pigletproducing farms.We detected differences between the farms both in microbial diversity and the RAs of genera.We also observed differences between primiparous and multiparous sows in the RAs of various genera in all the sample types.Some of these differences were shared between all the farms, while others were limited to only some of them.
Moderate differences between the farms were detected both in microbial diversity and RAs of the main genera in all the sample types.Although standard Finnish animal husbandry procedures were used on all the farms, many factors can contribute to differences in microbiota compositions, such as the genetic background of the sows [31,32], details in feeding, [25,26,33], antibiotic use and stress [27,34], and characteristics of the physical and microbial environment [35], especially during early life when the microbiota of the sows was first established ( [24,28,36,37].

Oral microbiota
Proteobacteria was the most abundant phylum in the sow oral samples in contrast to the other sample types, which were dominated by Firmicutes.At the genus level, the oral microbiota was dominated by Rothia, Moraxella, Streptococcus, and Actinobacillus, followed by Porphyromonas, Neisseria, and Corynebacterium.Species belonging to these genera are commonly detected in the oral cavity and upper respiratory tract of various mammalian species, such as humans, dogs, and cattle [38][39][40].Relatively few studies have addressed the composition of oral microbiota in pigs, mostly in piglets and growing pigs at various oral sites, such as tonsilla [41,42], gingiva, buccal mucosa, and floor of the mouth [43].The four top genera of our study were also identified as the most abundant in the saliva of sows and piglets in the study by Murase et al., where a similar procedure of sampling was used [44].
Overall, around half of the top 20 genera were shared between the two studies.The composition in our study also overlapped with two other studies on sow oral microbiota where cotton swabs [24] or ropes [45] were used in sample collection, with eight of our 20 top genera shared with those of each of the studies.Acinetobacter was the dominant genus in both of these studies, Streptococcus was among the most abundant genera, but Rothia, Moraxella, and Actinobacillus were either less abundant or not detected.Only the genera Streptococcus, Lactobacillus and Clostridium sensu stricto 1 were among the 20 most abundant in all the four studies, indicating variability of the RAs and prevalences of the oral genera.The differences in the reported compositions may be due to many factors, such as feed and the environmental microbial exposure of the sows.Detailed information on the effect of the reproductive cycle on the sow oral microbiota composition is not yet available.Li et al. reported a transient increase of Actinobacillus during parturition [24].Pregnancy is known to induce changes in the human oral microbiota, some of which may be related to adverse pregnancy outcomes [46].We detected significantly higher abundances of the genera Fusobacterium and Neisseria in multiparous versus primiparous sows from the four farms.To the best of our knowledge, the effects of parity on the oral microbiota of sows have not been reported previously.The difference in Fusobacterium was attributed to F. necrophorum.F. necrophorum is present in the mouth, upper respiratory, and gastrointestinal tract and is an opportunistic pathogen in humans and other species ( [47,48].In pigs, F. necrophorum can cause necrotic stomatitis and facial necrosis [49,50].The species of Neisseria involved could not be determined without ambiguity.Several species of Neisseria are associated with periodontal health in humans [38] and dogs [39,51].

Vaginal microbiota
Actinobacillus, Clostridium sensu stricto 1, Parvimonas, and Streptococcus were among the most abundant genera in the vaginal microbiota.These genera have also been reported to belong to the most abundant in pregnant sows by others [27,52,53].For the other top genera, there was more variation between the studies.Actinobacillus and Clostridium were also among the most abundant genera in weaning sows [54].The RAs of Lactobacilli in our study were comparatively low but still within the range of RAs reported in the other studies.
While lactobacilli are dominant in the human vaginal microbiota and have an important role in the maintenance of vaginal pH, the RAs of lactobacilli in other mammalian species, including pigs, are much lower [55,56]).
A marked proportion of the multiparous sows had higher RAs of the genera Fusobacterium and Streptococcus than primiparous sows.Although not statistically significant at the group level, this difference was seen in all farms and in around half of the multiparous sows.The species corresponding to the differentially abundant ASVs were F. gastrosuis and F. necrophorum.F. gastrosuis is resident in the pig gastrointestinal tract and is implicated in gastric histopathology [57].F. necrophorum is associated with metritis in dairy cows [58].
Parity-related differences in the sow vaginal microbiota have not been reported previously.

Rectal microbiota
The major rectal genera in the studied sows were Clostridium sensu stricto 1, Terrisporobacter, Lactobacilllus, and Romboutsia.Christensenellaceae_R−7_group, Prevotella, Turicibacter, and Rikenellaceae_RC9_gut_group were also present in high relative abundances.These genera were also among the most abundant in other recent studies on sows from several geographically distinct locations and varying study designs [15-17, 19-21, 23, 27, 61, 62].There is variation across studies in the genera reported as the most abundant in sow fecal or rectal samples, with Clostridium as the most abundant in some studies [19,23,62], including ours, and Prevotella, Treponema, or both in others [17,20,21].However, all the genera defined as the minimal common fecal core in growing pigs by Holman et al. [63] (Prevotella, Clostridium, Alloprevotella, Ruminococcus, and the RC9 gut group) were among the most abundant both in our study and in the other sow studies.In addition to these, Christensenellaceae_R−7_group was among the most abundant in most of the sow studies.
Parity was associated with the RAs of several genera in rectal samples.Consistent differences included a significantly higher RA of Clostridium sensu stricto 1, Romboutsia, and Lachnospiraceae_UCG_007 and lower RA of Prevotella in multiparous compared with primiparous sows.Higher RA of Clostridium sensu stricto 1 in third parity sows compared with first parity sows has also been reported earlier [23].Clostridium species are commensal bacteria known as butyrate producers in the gut [64].Lachnospiraceae are also producers of short-chain fatty acids [65].Prevotella contains species with the ability to degrade plant glycans [66].In growing pigs, Prevotella is a dominant genus that has multiple interactions with other microbial taxa; Prevotella species have effects on feeding efficiency [67].Higher RAs of Prevotellaceae and Bacteroidota in high vs low parity have been observed previously [21].Bacterial species belonging to the genus Romboutsia are gastrointestinal or environmental anaerobes identified from animal and human ileal samples [68,69].Little is known of their role in the gastrointestinal tract, but higher RAs of Romboutsia along with other commensal intestinal genera are associated with lower prevalence of type 2 diabetes and lower RAs with hypertension in humans [70,71].
The differences in the abundance of the genus Lactobacillus detected in our study were due to only a few individuals and could not be considered as consistent over the data.Parityrelated differences in the abundances of Lactobacillus amylovorus and L. reuteri have however been observed in pregnant sows by Berry et.al [22].They also observed increased abundance of Treponema bryantii with higher parity.The genus Treponema was detected in our study, albeit with lower RA than in some other studies, and its RA did not differ between the parity groups.Several species of Treponema were present, but no ASVs corresponding to T. bryantii were detected.

Colostrum microbiota
The most abundant genera in colostrum were Clostridium sensu stricto 1, Streptococcus, Staphylococcus, Corynebacterium, Terrisporobacter, Romboutsia, and Lactobacillus.Clostridium, Streptococcus, Staphylococcus, and Lactobacillus were identified as the most abundant genera in other studies [28,72].Pseudomonas was reported as the core genus by Li et al., with variation between the study breeds for the abundances of other genera [24].
The main reason for the variation in colostrum samples between the parity groups was the high RA of the genus Streptococcus in five (of 23) multiparous animals from three different farms.This was due to the exceptionally high RA of the species S. dysgalactiae.This species is a known pathogen that causes mastitis in cows [73] and has been found in increased abundance in sows with purulent vaginal discharge [74].When transmitted to newborn piglets it may cause arthritis and encephalitis [75].The source of S. dysgalactiae in the colostrum of the study sows remains unknown.It was present in vaginal samples of both affected and nonaffected individuals but was not detected in rectal samples.Environmental transmission is also possible [73].

Limitations of the study
This was an exploratory study on the compositions of mucosal and colostrum microbiota on sows reared under regular commercial production conditions.It was not specifically designed to study differences between sows of different parity.Therefore, the results presented must be considered preliminary.However, it is notable that our analysis uncovered differences between the rectal microbiotas of primiparous and multiparous sows that are similar to those previously reported from study settings specifically designed for that purpose.

Conclusions
Our study presents the microbial compositions of the mucosal sites and colostrum of commercially reared pregnant sows from Finnish piglet-producing farms.Diverse microbiotas with some variations between farms were discovered, with Rothia, Moraxella, and Streptococcus as the major genera in oral, Streptococcus in vaginal, Clostridium sensu stricto 1 in rectal, and Streptococcus in colostrum samples.Several potentially interesting differences between primiparous and multiparous sows were detected, involving Fusobacterium and Neisseria in oral, Fusobacterium and Streptococcus in vaginal, Clostridium sensu stricto 1, Romboutsia, Lachnospiraceae_UCG_007, Prevotella in rectal, and Streptococcus in colostrum samples.Further targeted studies are needed to define the differences related to parity and the impact of the sow microbiota on reproductive health.Otherwise, the farm staff followed their normal management practices during the study.The researchers inspected the study sows visually while supervising their farrowings.None of the sows were considered to have health concerns.

Farms, animals, and management
On each farm, the sows in the farrowing units were housed in farrowing pens (4.6-4.8 m 2 ) with crates and partly slatted floors.No bedding material was used for the sows and piglets in the farrowing pens.Within the farrowing pen, the piglets had a separate nest with a heat lamp and solid floor.The sows were fed a standard liquid feed three times a day in farms A-C and in farm D three to four times before farrowing, twice right after farrowing, and four times daily during the rest of the lactation.Further details of the study farms, animals, and sow management are shown in Table 1.
Table 1.Descriptive information about the four study farms, animals, and management.

Library preparation and 16S rRNA gene amplicon sequencing
The hypervariable regions V3-V4 of the 16S ribosomal RNA (rRNA) genes were sequenced using the Illumina MiSeq platform in the DNA core facility of the University of Helsinki, essentially as described previously [77,78] The final extension step was at 72°C for 5 min.A T100™ Thermal Cycler (Bio-Rad Laboratories) was used.The amplicons were treated with exonuclease I and shrimp alkaline phosphatase for 30 min at 37°C to remove excess free primers.The second-round PCR amplifications were performed using an Illumina forward and reverse primer set, Phusion Hot-Start II polymerase, High Fidelity buffer, and 2.5% DMSO.The following thermal cycling conditions were applied with an Arktik thermal cycler (Finnzymes/Thermo Scientific): initial denaturation at 98°C for 30 s, 18 cycles at 98°C for 10 s, 65°C for 30 s, 72°C for 10s, and a final extension at 72°C for 5 min.Sample libraries were then pooled, and the pool was purified with a bead wash (MagSi-NGS Plus 0.9x).The final 16S rRNA gene amplicons were sequenced on an Illumina MiSeq sequencer using the v2 600 cycle kit paired-end (325 bp + 285 bp).

Bioinformatics and statistics
The raw sequence data were processed as described in detail previously [78].Briefly, the read quality was inspected using FastQC and MultiQC [79,80].Primers and spacers were trimmed using Cutadapt version 1.10 [81].The mapping file was validated using Keemei [82].QIIME2 version 2022.2 and the DADA2 plugin were used to de-noise and filter the reads, call amplicon sequence variants (ASVs), and generate a feature table [83,84].
Taxonomy was assigned using the SILVA v138 99% database [85,86].Singleton sequences and sequences derived from chloroplasts or mitochondria were removed.One vaginal microbiota sample was excluded from further analyses as it was mostly composed of Delftia, a known reagent contaminant.Species-level identifications were confirmed by BLASTN 2.14.1+ search of the ASV sequences.
Statistical significances of alpha diversity differences were evaluated using the nonparametric Kruskal-Wallis rank-sum test and pairwise Wilcoxon rank-sum exact test with Bonferroni correction.Principal coordinates analyses (PCoA) and distance-based redundancy analysis (dbRDA) were performed using vegan 2.6-4 [91].The core microbiota heatmaps were generated using the microbiome package.Scater 1.18.6 package [92] was alpha diversity indices of the four sample types are shown in Fig 1a.Each sample type clustered separately in PCoA (Fig 1b).In PCoA, the oral samples were completely separated from the other sample types on the PCo1 axis.Vaginal samples were separated from the others on the PCo2 axis, with some overlap with the rectal cluster (Fig 1b).The colostrum samples clustered closest to but separate from the rectal samples and separate from the oral and vaginal samples (Fig 1b).

Proteobacteria ( 7 .
0%).The overall relative abundances (RAs) of the main phyla in each sample type are shown in Fig 1c.

Figure 1 .
Figure 1.Overview of the oral, vaginal, rectal, and colostrum microbiota in 32 pregnant sows from four Finnish commercial farms, A-D.a) Alpha diversity, b) PCoA of Bray Curtis dissimilarities, c) relative abundances of major phyla.
), but the difference was significant only between farms D and C. In PCoA, the samples from different farms overlapped, but a partial separation of farm D samples from the others on PCo1 and of farm B on PCo2 could be observed (Fig 2b).In total, 113 genera were identified in the oral samples.The microbiota composition at the genus level is presented in Fig 2c.Twenty-one genera had 100% prevalence and together constituted 79% of the total abundance calculated from the RA data.The most abundant of these genera were Rothia, Moraxella, Streptococcus, Actinobacillus, an unidentified genus of the Neisseriaceae family, Porphyromonas, Neisseria and Corynebacterium, all present with RA >0.1% (Fig 2c).There were some apparent differences in the genus-level compositions between the farms with farm A and D sows having higher RA of Moraxella and farm D lower RA of Rothia than sows from the other farms (Fig 2d).Further, sows from farms A and D had lower RAs of Lactobacillus than those of farms B and C (Fig 2d).

Figure 2 .
Figure 2. Composition of the oral microbiota of pregnant sows from four Finnish commercial farms.a) Alpha diversity indices Shannon and Inverse Simpson, b) principal coordinates analysis based on Bray-Curtis dissimilarities, c) Core heatmap, d) Average relative abundances of the main genera in farms A-D.
3a).Only the difference between farm C and D was significant in Shannon.No differences were significant between the farms in Inverse Simpson's.The farms were not separated in PCoA, but three primiparous sows from farm A clustered separately on PCoA 1 (Fig 3b).In vaginal samples, 129 genera were identified.Altogether 13 of these were present in all samples constituting 42% of the total microbiota.Streptococcus, Clostridium sensu stricto 1, Peptostreptococcus, Parvimonas, Terrisporobacter , Prevotella, Delftia, Romboutsia, Lactobacillus, and Turicibacter were the most abundant of these genera (Fig 3c).Actinobacillus and Fusobacterium had also high relative abundances but remained undetected in one sample.Farm A samples had lower RAs of Streptococcus, Fusobacterium, and Actinobacillus than the others, but higher RAs of Porphyromonas, Ezakiella and Peptoniphilus.Farms B and D had higher RAs of Streptococcus, Fusobacterium, and Actinobacillus than farms A and C, and higher RA of Lactobacillus compared to other farms (Fig 3d).

Figure 3 .
Figure 3. Composition of the vaginal microbiota of pregnant sows from four Finnish commercial farms.a) Alpha diversity indices Shannon and Inverse Simpson, b) Principal coordinates analysis based on Bray-Curtis dissimilarities, c) Core heatmap, d) Average relative abundances of the main genera in farms A-D.

Figure 4 .
Figure 4. Composition of the rectal microbiota of pregnant sows from four Finnish commercial farms.a) Alpha diversity indices Shannon and Inverse Simpson, b) Principal coordinates analysis based on Bray-Curtis dissimilarities, c) Core heatmap, d) Average relative abundances of the main genera in farms A-D.
formed a cluster on PCo1, with three outliers belonging to two different farms clustering separately in the upper right corner (Fig 5b).Among the 159 genera identified, 13 were present in all samples and accounted for 64% of the total abundance.Clostridium sensu stricto 1, Terrisporobacter, Corynebacterium, Romboutsia, and Staphylococcus were the most abundant genera (Fig 5c).Samples from farm A had lower abundance of Streptococcus and higher abundance of Staphylococcus and Terrisporobacter than other farms.Samples from farm B had higher RA of Corynebacterium and Anaerococcus, and lower RA of Clostridium than the others (Fig 5d).Farm C had more Rothia than other farms, and farm D more Aerococcus (Fig 5d).

Figure 5 .
Figure 5. Composition of the colostrum microbiota of pregnant sows from four Finnish commercial farms.a) Alpha diversity indices Shannon and Inverse Simpson, b) Principal coordinates analysis based on Bray-Curtis dissimilarities, c) Core heatmap, d) Average relative abundances of the main genera in farms A-D.

Fig 1 )
. However, we detected differences in the RAs of individual genera in each of the sample types.RAs of the main genera in all samples sorted by parity group are shown in Fig 6 (Left panel, a-d).Variation in the RAs of some of the major genera between individuals was extensive, especially of Lactobacillus in oral and rectal samples, Fusobacterium in vagina, and Streptococcus in colostrum.Distance-based redundancy analysis (dbRDA) shows the genera with largest contributions to the difference between the groups in each sample type (Fig 6, Middle panel).The genera with a tendency to higher RAs in multiparous sows point left and vice versa.The genera highlighted in dbRDA were tested for statistical significance and consistency of the difference across the farms.Fig 6 (Right panel) shows boxplots for the genera that were significantly and consistently different in each of the sample types.

Four
commercial piglet-producing farms in western and southwestern Finland participated voluntarily in the study during 2018-2019.The study farms are hereafter referred to as farms A-D.Detailed information about expected farrowing days and sow parities of one farrowing group in each farm were obtained from the farmers.The first farm visit was planned to enable the researchers to supervise the maximum number of farrowings in three days on each farm.

Sampling
Swab samples from sow oral, vaginal, and rectal mucosa were collected from 32 pregnant sows on average 1.5 (SD 1.2) days prior to farrowing.Factory clean protective gloves were used during sampling.Sterile flocked swabs (FLOQSwabs®, Copan Diagnostics Inc, CA, USA) were used for oral and vaginal samples.Oral samples were taken by wiping the buccal mucosa, and vaginal samples by opening the vulva with one hand and wiping the mucosa just inside the vagina with the swab.From one sow the oral sample was not obtained.Rectal samples were taken with sterile cotton swabs (Applimed SA, Châtel-St-Denis).The swabs were placed into cryotubes that were placed in a cool box, moved to a -18°C freezer within an hour and further to a -80°C freezer on arrival at the laboratory.Colostrum samples were milked within 6 hours from birth of the first piglet into sterile 10-ml plastic tubes after disinfection of the udder as described in detail in[76].From one sow the colostrum sample was not obtained.Colostrum samples were stored similarly as the other samples.DNA extractionDNA from all sample types was extracted using a ZymoBIOMICSTM DNA Miniprep Kit (Zymo Research, Irvine, CA) with minor modifications to the kit instructions as described below.Different sample types were processed at different times to avoid cross-contamination between samples.All samples were thawed on ice on extraction day.The sampling swabs were transferred with sterile instruments into lysis tubes (ZR BashingBead Lysis Tubes, Zymo Research, Irvine, CA), and 750 µl of ZymoBiomics Lysis Solution was added to each tube.Bacterial cells were lysed three times with a Fastprep® 24 (5.5 m/s, 60 s); the tubes were then centrifuged 16000 x g for 5 min at 4°C.The supernatant was transferred into ultraclean Eppendorf tubes and 200 µl of lysis solution was added to the tubes.The fast-prep and centrifugation steps were repeated.The supernatants were combined and centrifuged at 8000 x g for 1 min at room temperature.Plain sterile flocked swabs (FLOQSwabs®, Copan Diagnostics Inc, CA, USA) were used as negative controls for vaginal swabs during three of four analysis rounds.For oral and rectal swabs, negative controls were not included in the extraction round.For colostrum samples, 2 ml of thawed sample were transferred into ultraclean microcentrifuge tubes and centrifuged at 16000 x g for 10 min at +4 o C. Excess supernatant was removed until 200 µl of the supernatant remained.750 µl of lysis solution (ZymoBiomics Lysis Solution) and 19 µl 20 mg/ml proteinase K solution in storage buffer (Zymo Research, Irvine, CA, USA) were added into each tube and mixed.The tubes were incubated at 55°C for 30 min.The remaining steps of the extraction from fast-prepping onward were performed as for the other sample types.Ultrapure water (Zymo Research, Irvine, CA, USA), stored in ultraclean microcentrifuge tubes in a similar way to the colostrum samples, was used as a negative control for colostrum samples.After extraction, the DNA concentration of all sample types was measured with Qubit 3.0 Fluorometer (Thermo Fisher Scientific Inc., Waltham, MA).