Wolves, dogs and humans in regular contact can mutually impact each other’s skin microbiota

In contrast to humans and dogs, the skin microbiota of wolves is yet to be described. Here, we investigated the skin microbiota of dogs and wolves kept in outdoor packs at the Wolf Science Center (WSC) via 16S rRNA gene amplicon sequencing. Skin swab samples were also collected from human care takers and their pet dogs. When comparing the three canine groups, representing different degrees of human contact to the care takers and each other, the pet dogs showed the highest level of diversity. Additionally, while human skin was dominated by a few abundant phylotypes, the skin microbiota of the care takers who had particularly close contact with the WSC animals was more similar to the microbiota of dogs and wolves compared to the humans who had less contact with these animals. Our results suggest that domestication may have an impact on the diversity of the skin microbiota, and that the canine skin microbiota can be shared with humans, depending on the level of interaction.

cohabitation with dogs leads to an exchange of gut as well as skin microbes between dogs and humans 2,38,40 . Sharing skin microbiota between dogs and humans has also been reported at the level of individual correlations specific to dyads living in the same household, and this process even affects the microbial exchange between cohabiting humans 2,38 . Interestingly, one study that included not only dogs but also cats as pets found weaker effects in this regard 41 , raising the possibility that dogs may have an especially strong impact on the microbiota of cohabiting humans. Given that pet owners tend to establish especially close relationships and engage in the most diverse activities with their dogs, this closeness as well as the long evolutionary history of dog-human cohabitation may contribute to the successful establishment of exchanged microbes between dogs and humans.
In the current study, we aimed to investigate the skin microbiota of wolves, dogs and humans that all had a varying amount of contact with each other and partly inhabited the same environment. To do so, we sampled dogs and wolves that had been raised by human care takers and were kept in a game park setting at the Wolf Science Center (WSC) for the purpose of behavioural and cognitive research. As such, both groups of animals had a similar but limited amount of contact to humans as compared to pets. Furthermore, the care takers of these animals and the pet dogs belonging to them were also sampled, in order to investigate whether various levels of contact with humans corresponds to different microbial compositions inhabiting the skin of composition in dogs. Indeed, the human participants and the pet dogs had varying amounts of contact to the wolves and dogs kept at the game park (WSC dogs and wolves), ranging from frequent direct contact (animal trainers) to medium or low contact (researchers studying animal behaviour). By comparing these groups of humans and canines we were able to show that the pet dogs had the highest level of diversity in their skin microbiomes. Additionally, the trainers that had particularly close contact with the animals had a more similar skin microbiome to the wolves and dogs, which was observed when comparing the distribution of various taxonomic levels. Lastly, the human skin was dominated by a few highly abundant phylotypes. Overall, our results suggest that exposure to various environments can have a large impact on the diversity of the skin microbiome and that the canine skin microbiome can be shared with humans to a degree that depends on the level of interaction.

Results and discussion
Humans have the least and pet dogs have the most diverse skin microbiota. Out of all four groups, species richness and diversity were lowest in human skin microbiota, whereas the pet dog group had the highest species richness and diversity. All four groups differed significantly to each other (Kruskal-Wallis; Chao1, chi-squared = 20.828, df = 3, p < 0.001; Shannon, chi-squared = 23.332, df = 3, p < 0.001). In the following pairwise comparisons all groups differed significantly from each other (p < 0.005), except for the WSC dog group vs. the WSC wolf group, that showed a very similar species richness and diversity (Dunn's-test for multiple comparisons; Chao1, p = 0.574; Shannon, p = 0.579; Fig. 1.). The low species richness and diversity in the human skin microbiota has also been shown in previous studies 42,43 and might be driven by physiological differences of the skin between humans and canines, such as pH and hair covering, as well as by skin hygiene practices and differences in regular environmental contact 44,45 . For humans, protection from invasion by microorganisms is controlled by microbial desiccation, competition with resident microbiota, and an acidic pH. The average reported cutaneous pH of humans is 4.8, while the average cutaneous pH of dogs is 7.4, suggesting that the higher pH might support a more complex skin microbiota composition, as compared to humans 45 . Moreover, Clemente et al. 33 showed that the skin microbiome of uncontacted humans living outdoors is significantly more diverse than that of westernized people, supporting the assumption that modern habits, including personal and environmental hygiene, lead to a decrease in skin microbiota species richness and diversity 33,46 . With respect to the high diversity of the pet dogs' microbiota, we suggest that the pet dog skin was regularly exposed to a diverse set of environmental microbiomes, both indoors and outdoors. The WSC animals (wolves and dogs), in contrast, are restricted to the game park environment.
Humans and pet dogs are less similar compared to WSC dogs and wolves, which are more similar to each other in their skin microbiota composition. The analysis of independent variable influence, i.e. sex (male vs. female), age class (sub adult vs. adult), last antibiotic treatment (early vs. late) and human contact did not reveal significant effects on the canine skin microbiota composition (multivariate PER-MANOVA in Adonis; diet, p = 0. 387; age, p = 0.136; sex, p = 0.114; last antibiotic treatment, p = 0.728, human contact, p = 0.539), but the groups differed significantly (multivariate PERMANOVA in Adonis; group, R 2 = 0.074, F model = 1.622, p = 0.0228). Also in beta diversity significant differences were detected between groups (PER-MANOVA; Bray-Curtis dissimilarity; pseudo R 2 = 0.180; F model = 3.741; p = 0.001). In the pairwise comparison, all groups differed significantly from each other (R 2 = 0.076-0.209, F model = 1.468-6.598, p = 0.006). In the visual inspection of tSNE plots which were based on Jensen-Shannon divergence and Bray-Curtis dissimilarity ( Fig. 2.), humans and pet dogs appear less similar to each other and the other groups, while WSC dogs and wolves appear more similar to each other in their skin microbiota composition. Not surprisingly, among the four groups humans appeared most distinct to all canines that grouped more closely together.
Physiological differences between canine species affecting their microbiota may be relatively small. DeCandia et al. (2019) 47 found out that coyotes, red and grey foxes living in the wild in North-America responded with a similar microbial community shift to a Sarcoptes scabiei mite infection. However, the environment and living conditions of individuals do lead to a difference in microbial composition within species 48 suggesting that environmental effects on microbiota are significant. These observations are also consistent with the pairwise alpha diversity comparisons above, where the only non-significant difference between groups was the comparison of WSC dogs to wolves.
While physiological differences of the skin are likely to contribute to the differences observed between humans and canines, in this study mutual contact and living in the same environment have likely reduced difference www.nature.com/scientificreports/ between the canine groups 44,45 . This is probably an important factor that has made the WSC dog and wolf samples most similar to each other ( Fig. 2.). This is consistent with the species richness and alpha diversity, and suggests that environmental exposure has a similar or stronger impact on shaping the skin microbiota than the evolutionary segregation of wolves and dogs due to domestication (see also 21,22,35 ). It has further been suggested that diet shapes the skin microbiome 49 . Here, diet does not explain the difference between the pet dog group and the WSC animals (PERMANOVA; diet, R 2 = 0.046, F model = 1.007, p = 0.386). Both dog groups, WSC dogs and pet dogs were fed a similar diet. However, the WSC dog and wolf groups microbial community structure was similar to each other despite being fed different diets (Supplementary Dataset 1).
The hologenome theory of evolution supports the idea that specific groups or species evolved together with their microbiomes, and that this symbiosis greatly affects their health status 50,51 . A loss of microbiome diversity can be caused by several factors and may impact health. For example, a decrease in species richness and diversity can be caused by specific living conditions over several generations, as seen in both humans and canines living in or close to urban environments, as compared to populations living under natural conditions 33 . Outside of these long-term shifts in the skin microbiomes of certain groups, similar changes in diversity that do not necessarily impact health, can also be caused within shorter time spans, as represented in our canine groups. The pet dogs within the current study were similar to each other in terms of beta diversity, but distinct from the WSC animals ( Fig. 2). Of the pet dogs, four out of twelve were originally born at the WSC and later on adopted by the human caretakers, while remaining on the same diet; their origin however apparently did not leave remaining track in their microbiome as they were not more similar to the WSC dogs than the other pet dogs. Lastly, three of the four humans, that were a priory classified as having close physical interactions with the WSC dogs and wolves were more similar to the WSC animals than the other human participants with less animal contact (PERMANOVA; R 2 = 0.167, F = 1.967, p = 0.077; Fig. 2). This was also found when conducting a PCoA ( Supplementary Fig. 1). Based on this finding, a new group named "human close" was used to label downstream analyses and included the three human samples that clustered closer to the animals.
The skin microbiota of humans with close contact to WSC animals is more similar to the microbiota of the WSC animals. LEfSe analysis revealed an enrichment of Acidobacteria, Verrucomicrobia and Proteobacteria (in particular Alphaproteobacteria and Gammaproteobacteria) in wolves, and Bacteroidetes and www.nature.com/scientificreports/ Proteobacteria (in particular Betaproteobacteria) in WSC dogs, whereas Firmicutes were enriched in human samples (Fig. 3A). Interestingly, not only the three canine groups, but also the humans with close contact to the WSC animals, had a microbiome that showed higher proportions of Proteobacteria, Acidobacteria, Verrucomicrobia, and Bacteroidetes (Fig. 4). The proportions of Firmicutes and Actinobacteria were lower in these groups compared to the humans without close contact to the WSC animals. This suggests that contact with the animals increased the ratio of gram negative to gram positive microorganisms on the skin, and the phylum level diversity. The Proteobacteria appeared to be particularly shared between humans with close animal contact and pet dogs (Fig. 4). The patterns observed at phylum level could also be seen on lower taxonomic levels in the LEfSe analysis, with a significant enrichment of f.e. Sphingomonadaceae in the wolf group (Fig. 3B), and f.e. Pseudomonadaceae in the WSC dog group (Fig. 3C). Consistent with this observation, humans with close contact to the WSC animals had microbial community shifts mainly caused by an increase in Pseudomonadaceae, Sphingomonadaceae and Flavobacteriaceae abundance and a decrease of Staphylococcacceae and Corynebacteriaceae abundance ( Supplementary Fig. 2). Lehtimaki et al. 39 found that intensive contact to forest and arable land (which was the case for our canine groups compared to the human individuals) correlated with a higher diversity of soil-based Proteobacteria and other soil-related taxa on skin, while in urban areas, skin-based Actinobacteria were more abundant. This was also the case in our human samples, with the exception of those from humans with very close contact to WSC animals. This supports the importance of human-animal interactions in providing exposure to environmental microbes and affecting the skin microbiome composition 2,23 . These findings also indicate that not only the pet www.nature.com/scientificreports/ dog skin microbiota seems to be greatly affected by their surroundings, but that the human skin microbiota can also show strong shifts when being exposed to canine animals and their environment. Coelho et al. 53 recently indicated that the similarities in the gut microbiome of dogs and humans might not be solely explained by direct transmission, but rather a function of similar physiology and lifestyle 53 . Lehtimaki et al. 54 also examined that a shared living environment as well as lifestyle patterns correlate with microbial similarities in dog-owner pairs and that they influence the structure of skin microbiota in both species. In our study, direct transmission might have also contributed to the similarities observed, as the skin was not cleaned before sampling thereby washing off allochthonous microorganisms, as applied in skin microbiota sampling of amphibians 43,55 . However, different to former studies on pets 2,38 , we found no evidence that owners and pet dogs living in the same household would have more similar skin microbiota to each other than to other pet dogs or other owners, respectively. A rural environment or countryside lifestyle are known to boost protection against allergic diseases 39 . This protection has suggested to be mediated by acquiring a more diverse microbiome or by exposure to environmental microbes that support immune tolerance 39 . In this study, the three humans with close contact to the WSC animals had a microbiota structure with altered composition (PERMANOVA; R 2 = 0.167, F = 1.967, p = 0.077) but without higher species richness and/or diversity. Whether these changes might benefit their hosts in terms of immune responses or allergies, as suggested previously 39,46,54 , needs further investigation.
The human skin microbiota. A total of 15,181 ASVs were detected in all skin samples from the four groups. In the human group, 4,676 ASVs were found, of which 581 (12%) were shared with the pet dogs. Firmicutes, in particular Staphylococcaceae were identified as significantly enriched in humans in the LefSe analysis (Fig. 3A,D). The six most abundant ASVs in the human group were classified as Staphylococcus with relative abundances from 7.22 to 2.39%. The dominance of Staphylococcaceae is in line with previous literature 56 . Staphylococcus species are known to especially inhabit the human skin as commensals and opportunistic pathogens 57 . Other highly abundant phylotypes detected in the current dataset have also been described as normal inhabitants of the human skin, such as Corynebacteria, that show a higher sensitivity to environmental factors compared to staphylococci 58 . Cutibacterium has also been found on human skin samples. Cutibacterium acne is www.nature.com/scientificreports/ seen as an important species within this genus, which is linked to acne and balanced by specific Staphylococcus strains 59 . Several ASVs among the 50 most abundant, which were classified as Staphylococcus, Corynebacterium and Cutibacterium were significantly higher abundant in human samples in the pairwise comparison with canine skin samples ( Table 1).
The canine skin microbiota. This is the first study to describe the bacterial composition on the skin of wolves. Overall, the wolf skin was inhabited by similar abundant phylotypes compared to the dog skin. The most abundant ASVs in the three canine groups were classified as Pseudomonas, Rhodococcus, Staphylococcus, Micrococcus and Sphingomonas with relative abundances between 1.26 to 0.60% (Table 1, Supplementary Dataset 2). Several phylotypes were enriched in the wolf group in the LEfse analysis, among them Alpha-and Gammaproteobacteria, Sphingomonadaceae and Pseudomonadales. In the WSC dog group, Bacteroidetes, Betaproteobacteria and Flavobacteriaceae were enriched, while in the pet dog group Actinobacteria, Pseudomonadaceae and Sphingobacteria showed significant enrichment (Fig. 3). Only one of the 50 most abundant ASVs that differed significantly between the WSC wolf and WSC dog group (Sphingomonas ASV 48, enriched in WSC wolf skin samples compared to WSC dog skin samples), while seven ASVs differed significantly between the WSC wolf and pet dog groups (three Staphylococcus-ASVs, Streptococcus, Pedobacter, unclassified Actinomycetales and Sphingosinicella, Table 1). Except of Sphingosinicella all of these ASVs were enriched in the pet dog skin samples, compared to the WSC wolf skin samples. All highly abundant phylotypes have been found on the skin of healthy dogs before 2 .
Overall, these findings indicate that the wolf skin microbiota is similar to the dog skin microbiota, but several phylotypes are more dominant in the WSC animals. This might be largely affected by living environment 21 , but whether and to what extent the similar skin microbiota is due to the wolf and dog subjects of the current study sharing the same environment is difficult to tell. A former study on red wolves has shown that living in a captive environment significantly affects the gut microbiota of the animals, and this effect is apparent even if the animals are fed with a diet reflective of their natural environment 60 . Given that our wolves also live in captivity and do share some of the facilities with the dogs, it is not surprising that the highly abundant ASVs on the WSC wolf skin were also present on the WSC dog skin. However, the fact that they were detected also on the skin of the pet dogs in this study suggests fundamental similarities in the skin microbiota of wolves and dogs, although the abundances of these phylotypes differed among groups. In this study, the shared environment of the wolves and WSC dogs seem to have impacted at least the profile of highly abundant phylotypes of the animals' skin microbiota (   2,46 ± 2,434 a 0,03 ± 0,051 b 0,00 ± 0,000 b 0,00 ± 0,000 b Staphylococcus 8 2,39 ± 2,409 a 0,01 ± 0,034 b 0,00 ± 0,000 b 0,00 ± 0,000 b Cutibacterium 9 2,21 ± 2,576 a 0,00 ± 0,000 b 0,02 ± 0,067 b 0,00 ± 0,000 b www.nature.com/scientificreports/ their differential diet and their divergent evolution, as also indicated by the lower similarity between the wolf and pet dog group (Dunn's test; Chao1, p = 0.025; Shannon, p = 0.041) compared to the wolf and WSC dog group (Dunn's test; Chao1, p = 0.574; Shannon, p = 0.579). The number of shared ASVs within the groups is shown in Supplementary Fig. 3. A huge amount of low abundant organisms was uniquely found in each canine group. The number of unique ASV was more than double in canine groups compared to the humans. The importance of a high skin microbiota diversity in canines, the function of unique low abundant phylotypes, as well as the importance of specific microbial enrichments in groups has not been investigated until now and has to be examined in future research.

Conclusion
Overall, the current results suggest that exposure to various environments and cohabitation affect the skin microbiota more than ecological and physiological changes that took place during the course of dog domestication and can lead to a significant shift of the skin microbiome both in dogs and in humans. Close contact with dogs and wolves shape the human skin microbiome by causing large shifts in the microbiota composition at the phylum and family level, leading to an increase in the ratio of gram negative to positive bacteria.

Material and methods
Ethics statement. Read processing. For all analyses, the forward reads from the dataset were processed into amplicon sequence variants (ASV) using DADA2 67 (version 1.9.1) with a forward cutoff of 30 and a length cutoff of 290 within the QIIME2 68 (version 2019.1) environment. All ASVs present in the negative control above 3% relative abundance were removed, leading to the removal of 10 ASVs. Before all beta-diversity and taxonomic comparisons were made, sequences that were classified at the class level as 'Chloroplast' were removed. A full overview on the absolute abundances per sample is shown in Supplementary Dataset 3. This data was used for all alpha and beta-diversity metrics, in addition to phylum level comparisons. Phylum level taxonomic assignment was conducted using the RDP package within R (version 1.20.0). The python and R code used in the analysis is available at [https:// github. com/ camer onstr achan/ WolfS kinCo mmuni ty2020]. www.nature.com/scientificreports/ Statistical analysis. Only samples with a total of more than 10,000 reads after quality control were kept.
Relative abundances were then calculated and used in downstream statistical analysis. Permutational multivariate analysis of variance (PERMANOVA) was used to examine the association between the microbial communities and independent variables including sex, diet, age, last antibiotic treatment and human contact. Species richness were calculated using the Chao1 index and Shannon index. The normality distribution of alpha diversity and ASVs was checked in semi-parametric factorial designs with corresponding histograms, Q-Q plots, residual versus fitted plots, the Shapiro-Wilk Test and skewness and kurtosis test for normality as well as for different transformations techniques such as log transformation. Additionally, the homogeneity of variances across samples was tested with the Levense's test. Due to the non-normal distribution of diversity data and ASV abundances, the Kruskal-Wallis test was applied followed by a Dunn's test for multiple comparisons. Further, means, standard deviations, and standard errors were calculated. The results were considered statistically significant at p value < 0.05 and the p value was adjusted for multiple comparison with the p-adjusted method Benjamini & Hochberg. The statistical analysis was done using the R statistical computing environment (Version 4.0.5 R Foundation for Statistical Computing, Vienna, Austria) using the package normtest, and stats and vegan.
For beta diversity analysis, T-distributed stochastic neighbor embedding (tSNE) plots, based on Jensen-Shannon divergence and Bray-Curtis dissimilarity, were calculated with the package tsne. To find biomarkers that explain the differences between the groups, LEfSe 69 was applied by using default values, except of the threshold on the logarithmic LDA score for discriminative features, which was set to 4.