Distinct healthy and atopic canine gut microbiota is influenced by diet and antibiotics

The rising trend in non-communicable chronic inflammatory diseases coincides with changes in Western lifestyle. While changes in the human microbiota may play a central role in the development of chronic diseases, estimating the contribution of associated lifestyle factors remains challenging. We studied the influence of lifestyle—diet, antibiotic use, and residential environment with housing and family—on the gut microbiota of healthy and owner-reported atopic pet dogs, searching for associations between the lifestyle factors, atopy and microbiota. The results showed that atopic and healthy dogs had contrasting gut microbial composition. The gut microbiota also differed between two breeds, Labrador Retriever and Finnish Lapphund, selected for our study. Among all lifestyle factors studied, diet was most significantly associated with gut microbiota but only weakly with atopic symptoms. Thus, diet- and atopy-associated changes in the microbiota were not interrelated. Instead, the severity of symptoms was positively associated with the usage of antibiotics, which in turn was associated with the microbiota composition. Urban lifestyle was significantly associated with the increased prevalence of allergies but not with the gut microbiota. Our results from pet dogs supported previous evidence from humans, demonstrating that antibiotics, gut microbiota and atopic manifestation are interrelated. This congruence suggests that canine atopy might be a promising model for understanding the aetiology of human allergy.


Background
There is substantial evidence suggesting that the rapidly increasing incidence of inflammatory disorders results from the lack of sufficient immune education by diverse microbiota of environmental origin [1,2]. However, as indicated by both indirect evidence and recently discovered 2. Material and methods

Selection of participants, collection of samples and background information
The study included 155 privately owned pet dogs, of which 34 were mother dogs and the rest their adult offspring. Two dog breeds, Labrador Retriever (n = 59) and Finnish Lapphund (n = 96), were selected as described in Lehtimäki et al. [33]. Shortly, the selected breeds are common in Finland and occur equally in rural and urban areas. Labrador Retrievers are more prone to develop atopy than Finnish Lapphunds [34], facilitating the examination of breed-and atopy-associated changes in microbiota.
Faecal samples were collected by the dog owners at home during a limited period in November 2014 to avoid any seasonal effects. Following detailed instructions, owners sampled a spoonful using a tube with spatula (Sarstedt #80.623.022) from the central part of the faeces, a part that had not touched the ground and was not hard. Samples were placed in the household freezer (−18°C) immediately or within an hour after collection and were later transported as frozen to the laboratory freezer (−20°C). Owners also filled a comprehensive questionnaire, described and validated in Lehtimäki et al. [33], where information about atopic symptoms and environmental, lifestyle as well as dietary factors of their dogs were collected.

Sequencing of the bacterial 16S rRNA gene and subsequent bioinformatics
DNA was extracted from faecal samples with QIAamp DNA Stool Mini Kit (Qiagen GmbH, Hilden, Germany) and, as previously described [35], was used as a template in PCR to amplify the V1-V3 regions of the bacterial 16S rRNA gene that was subsequently sequenced as in Lehtimäki et al. [33].
The processing of the 16S rRNA gene sequences with the subsequent clustering and annotation of operational taxonomic units (OTUs) was performed as recently described in detail [33]. Briefly summarized, after removing the primers from the raw reads by cutadapt v. 1.9.1. [36], pair-end reads were merged using USEARCH v. 9.2.64 [37] and merged reads quality-trimmed. Sample-specific trimmed reads were dereplicated in VSEARCH v. 2.64 [38]. These sequences were sorted by size, and rare sequences (abundance less than 10) were removed. Chimeras and other artefacts were removed using the unoise2 command [39] and subsequently the reads were mapped to OTUs with 99% similarity threshold using VSEARCH command usearch_global. The representative sequences of OTUs were classified with SILVA rRNA gene sequence database (v. 123) [40,41] using Mothur [42] command classify.seqs with 80% similarity cut-off.
The cumulative-sum scaling [43,44] was used to normalize the processed 16S rRNA gene sequence counts. Potential contaminants were removed after the normalization before further analysis, as described in Lehtimäki et al. [33], with the exception that we do not have controls for the sampling performed by dog owners.

Statistical analyses
2.3.1. Variables representing residential environment, atopy, lifestyle, dietary factors and antibiotic use The large questionnaire information was first simplified as reported in Lehtimäki et al. [33]. To summarize it here, we formed continuous variables from the questionnaire data using dimensionality reducing unconstrained methods. The continuous variables were then used in the statistical analyses to determine their effect on the gut microbiota. Each type of multivariate dataset, i.e. land-use types of residential environment, lifestyle factors and owner-reported atopic symptoms, were reduced to one or two summarizing variables using either principal component analysis (PCA; environment) or principal coordinate analysis (PCoA; lifestyle and allergy, based on Gover distance) [33]. The variables are further described in electronic supplementary material, additional file 1.
The formed continuous variables for environment, lifestyle and atopy were further transformed into categorical variables, as previously described [33]. Dogs were classified as living in either a rural or urban environment, having either a rural or urban lifestyle, and being either healthy or atopic, i.e. having owner-reported CAD.
Due to the complexity of canine atopic dermatitis (canine atopy) [45] and the uncertainty of using immunoglobulin E (IgE) as a clinical outcome in dogs [46], several cutaneous symptoms were questioned from dog owners as previously reported [33]. This validated questionnaire [47] was developed at the University of Helsinki by specialists in canine clinical nutrition, disease and royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 10: 221104 dermatology. The answers of atopic symptoms were then simplified with PCoA to represent ownerreported atopic symptoms in general or to rule out them.
Unlike in Lehtimäki et al. [33], here the information on dietary factors (appendix S1, questions numbered from 7.1 to 7.4. in Lehtimäki et al. [33]) and oral exposure to dirt, i.e. if dogs were allowed to eat or taste plants, soil and faeces, drink muddy water etc. outdoors (appendix S1, question number 8.6 in Lehtimäki et al. [33]) were not included in the PCoA of lifestyle factors. The aspects related to oral exposure to dirt were summarized in a separate PCoA based on Gower distance (electronic supplementary material, additional file 1). Related to diet, the dog owners reported in detail how they fed their dogs, during a period of seven days. To control the accuracy of the diet diaries, the owners were also asked to select one of nine alternatives (reported in appendix S1, question number 7.1 in Lehtimäki et al. [33]) to describe their dog's diet. If these answers did not correspond well with the other given information or if there were other unclarities, these owners (n = 19) were approached with further questions by email.
Based on the given information, we divided the used diets into three categories: (i) A non-heatprocessed (NHeP) low carbohydrate (LC) diet with raw (R) products (NHeP-LC-R), (ii) a heat-processed (HeP) high carbohydrate (HC) dry (D) kibble diet (HeP-HC-D), and (iii) a heat-processed (HeP) high carbohydrate (HC) moist (M) food diet (HeP-HC-M). The third category, i.e. HeP-HC-M diet, included canned or home-cooked food, or different heat-processed 'sausage packed' products, produced for dogs. For each dog, we estimated the percentage of food items belonging to HeP-HC-D, NHeP-LC-R and HeP-HC-M diets by their volume (ml) or weight (g), depending on which measure the dog owner had used in their dogs' food-diaries. As the most significant dietary factor, NHeP-LC-R was selected to be used in further statistics (electronic supplementary material, additional file 1, figure S2). Because actual quantities of different food types were only roughly reported by owners, we also approximated the percentage of NHeP-LC-R food into categories that described whether NHeP-LC-R items were used in the diet mostly/solely, partly or never. Additionally, we determined whether the diet included nutraceuticals such as probiotic and fat ( predominantly omega-3 oil) products daily, sometimes or never.
The use of antibiotics was asked in our original questionnaire [33], reporting if skin symptoms of dogs has been treated with antibiotics and if symptoms eased. Eighty-six dog owners passed this question, but also other questions concerning different medication used for skin symptoms. They most likely passed the questions because they reported no (n = 53) or very mild skin symptoms (n = 33) such as little skin itchiness (n = 14) or skin infections (n = 2) and never redness or eczema (n = 67). Indeed, our analysis categorized only one of these dogs as atopic, as previously described [33]. Thus, we assumed that these 86 dogs had not used antibiotics. In the statistical analyses, we used a categorical variable whether dogs have been treated with antibiotics (once or more) or not.

Relationships between the gut microbiota, atopic symptoms, environmental and lifestyle factors
In examining the variation in gut microbial composition, i.e. betadiversity (defined by Bray-Curtis dissimilarity), distance-based redundancy analysis (db-RDA) in the vegan package [48] was used to assess the role of different explanatory factors (electronic supplementary material, additional file 1, figure S1). In the first db-RDA, illustrating overall patterns, dog health (dogs defined as atopic or healthy) and breed were used as constraining (categorical) variables. In the second db-RDA, demonstrating the effect of diet on the gut microbiota, the percentage of NHeP-LC-R items in dogs' daily intake was used as a constraining variable, after adjusting for the effects of nutraceutical use (probiotics and fat products) and HeP-HC-M diet. The remaining residual variation was then constrained by the percentage of NHeP-LC-R products. HeP-HC-D was not considered as it correlated strongly with NHeP-LC-R (Pearson correlation −0.80, p < 2.2 × 10 −16 ). In the third db-RDA, the severity of atopic symptoms was used as the explanatory variable. Overlapping effects of NHeP-LC-R and especially antibiotics were visualized along with the ordination. Finally, we tested whether urban and rural land-use and lifestyle patterns and oral exposure to dirt affected the microbial dissimilarity. The significance and effects of the studied variable was estimated by multiple regression on distance matrices (MRM), using 9999 permutations in the vegan package [48].
To support the compositional changes observed in db-RDA analyses, we specified those OTUs that significantly varied between the different groups, e.g. atopic and healthy dogs, or due to dietary and lifestyle constraints. For that purpose, we used a differential abundance analysis, applying a zeroinflated lognormal model in the package metagenomeSeq [43,44]. For relative abundances of bacterial genera, a generalized linear model with quasi-binomial distribution was used to model difference between the above-mentioned groups. Both models used the false discovery rate (fdr) to adjust p-values.

Atopy-and breed-associated patterns in the gut microbiota
We investigated the role of dietary, lifestyle and environmental factors in the gut microbiota of pet dogs (table 1, electronic supplementary material, additional file 1, figure S1), considering potential association with atopy (i.e. owner-reported CAD). First, we questioned if dogs' gut microbiota composition and atopy were associated. In both breeds, Finnish Lapphunds and Labrador Retrievers, atopy was significantly associated with gut microbial composition that differed between healthy and atopic dogs (p allergy = 0.019) (figure 1a). The association between microbiota and atopy could be found for OTUs, but no significantly different genera were observed. In healthy dogs (n = 125), an OTU of the genus Prevotella_9 was more abundant, while in atopic dogs (n = 30), OTUs representing the genera Escherichia-Shigella were enriched (figure 1b). Otu66 of Prevotella_9 was one of the major OTUs among all (n = 6382, of which 1473 OTUs were prevalent in at least 5% of samples and were included in the statistics) with a mean global relative abundance of 0.87% (0.94% in healthy, 0.61% in atopic). Only 17 OTUs were more abundant (a mean global relative abundance between 0.9 and 8.8%) than Otu66 of Prevotella_9. OTUs of Escherichia-Shigella were moderately abundant, together gathering a mean relative abundance of 0.24% (0.05% in healthy, 1.03% in atopic). Several OTUs annotated as Escherichia-Shigella were also significantly more abundant in those dogs that had been given NHeP-LC-R = non-heat-processed low-carbohydrate raw, HeP-HC-D = heat-processed high-carbohydrate dry, HeP-HC-M = heatprocessed high-carbohydrate moist items in diet. Rural and urban lifestyle categories were formed by dividing the first principal coordinate (PCoA-1, electronic supplementary material, figure S1), summarizing different land-use types and lifestyle factors, respectively, to positive scores (greater than 0, rural) and negative scores (less than 0, urban).
royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 10: 221104 antibiotics (n = 19, electronic supplementary material, additional file 1, figure S3). As 50% of atopic dogs, but only 3% of healthy dogs, had received antibiotics during their lifetime, the association between increasing abundance of Escherichia-Shigella and antibiotics could not be excluded. Secondly, to understand the role of breed in the gut microbial compositions, we characterized the association between breed and the gut microbiota. The gut microbiota composition clearly differed between breeds (p breed = 0.0012) (figure 1a). In line with that, breed was the main source of the interindividual variation in microbiota (15.7%), which was significantly higher among atopic Labrador Retrievers than atopic Finnish Lapphunds (electronic supplementary material, additional file 1, figure S4a). The severity of atopic symptoms seemed to increase the inter-individual variation in the gut microbiota, explaining the high variation among Labrador Retrievers, which were generally more atopic and more medicated by antibiotics than Lapphunds (electronic supplementary material, additional file 1, figure S4).
In general, many of the studied dietary, lifestyle and environmental factors seemed to be associated with the dog breeds and their microbiota, but in different ways. For instance, diet was associated differently with the gut microbiota of Finnish Lapphunds and Labrador Retrievers (table 2). As the gut microbiota and associated factors distinguished between the breeds, we investigated atopy-associated and coinciding dietary, lifestyle and environmental patterns in the microbiota-separately for both breeds.  NHeP-LC-R in diet (electronic supplementary material, additional file 1, figure S5). Diet and allergic symptoms were not statistically associated in Finnish Lapphunds. In Labrador Retrievers, the severity of atopic symptoms (figure 3, p = 0.032) and the use of antibiotics (table 2, electronic supplementary material, additional file 1, figure S5) were associated with the gut microbiota clearly stronger than diet (table 2). The severity of atopic symptoms and antibiotic use were tightly coupled, whereas diet and symptoms were weakly associated. Labrador Retrievers with the most severe symptoms had received antibiotics (figure 3), and had been, in some cases, on the NHeP-LC-R diet (figure 3, electronic supplementary material, additional file 1, figure S1b). These associations prevent complete distinction of compositional changes in the microbiota attributed to diet, medication or allergy. However, Escherichia-Shigella was highly enriched in atopic dogs (figure 1b) or dogs treated with antibiotics (electronic supplementary material, additional file 1, figure S3) but only slightly enriched by diet ( figure 2c).
Alpha diversity, measured as Shannon diversity index, was associated with the amount of NHeP-LC-R in the diet ( p = 0.0003). The dogs having NHeP-LC-R items as the major components in their diet (greater than 33.3%) had lower diversity of amplicon sequencing variants (ASVs) in their gut microbiota than the dogs having less raw items in their diet (electronic supplementary material, additional file 1, figure S6). Alpha diversity was not significantly associated to any other studied factors.

The role of non-dietary lifestyle factors in canine atopy
Some dogs tend to eat or taste 'dirt' outside and thus are exposed to environmental or faecal microbes. Oral exposure to dirt, i.e. if dogs were allowed to eat or taste soil and faeces, drink muddy water etc., tended to be associated with gut microbial composition (table 2). Still, atopy was not significantly related to oral exposure to dirt in our data. The lifestyle variable, summarizing gut microbiota of Labrador Retrievers associated with atopy dbRDA-1, 3.1%  royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 10: 221104 several other aspects of urban and rural lifestyle, and living environment (land-use) was not associated with the gut microbial composition.

Discussion
Using a canine model, we investigated the associations of lifestyle and environmental factors-including diet and antibiotics-on the gut microbiota of atopic and healthy pet dogs of two different breeds. Our results suggest that the gut microbial composition is contrastingly different between atopic and healthy dogs as well as dogs on different diets. Recent investigations suggest that changes in the human gut microbiota induced by a Western lifestyle, e.g. dietary shift to 'high-fat, high-protein and high-sugar diet' [20], are associated with allergies due to altered immune function, e.g. via less short-chain fatty acids (SCFAs) produced from diet by gut microbes [4,8]. Some allergy-associated gut microbial signatures have been identified in humans, either preceding the disease [49,50] or after the onset of the disease [51][52][53]. Information on how and which gut microbes are associated with allergy or atopy (AD) in dogs is scarce and from small samples [25,26].
In general, diet is a major determinant of the mammalian gut microbial composition [27,28,54]. As expected, we also detected a clear difference in the gut microbiota of dogs that were on different diets. Among all dietary factors considered, NHeP-LC-R diet was most significantly related to the composition of the gut microbiota, particularly in Finnish Lapphunds. The consumption of NHeP-LC-R food, containing raw items with more animal proteins and fat and less carbohydrates, associated with the enrichment of Bacteroides and Lachnospiraceae in the gut. The consumption of HeP-HC-D food, rich in carbohydrates [23], is in turn associated with the increased abundances of Prevotella and Faecalibacterium. These compositional changes are well in line with previous reports, illustrating that Bacteroides is related to the consumption of animal proteins and fats, while Prevotella and Faecalibacterium indicate carbohydrate-based diet and fibres, both in dogs [28,55,56] and humans [57,58].
Do these observed diet-associated differences in the microbiota contribute to or protect from atopy, or on broader perspective, allergies? In human societies, Western diet is rich in processed food, which has repeatedly been associated with elevated allergy risk [4,8]. Processed dog food has also become very popular, paralleling the increase in canine allergies [24,30]. Considering these coincidental phenomena, it is relevant to assume that dietary changes may be involved also in dogs' allergic diseases. However, contrary to our expectations, dogs that consumed HeP-HC-D food were mostly healthy. The genera Prevotella and Faecalibacterium that were enriched in their gut, are known to produce SCFAs from dietary fibres and carbohydrates [59,60].
SCFAs, although not measured in our study, modulate immune tolerance of the host [61][62][63]. Lower concentration of SCFAs have been detected in humans suffering from atopic dermatitis (AD), as compared with subjects without AD [60]. However, dogs do not rely on the microbial fermentation of carbohydrates as much as humans [23]. In addition, the knowledge on the role of SCFAs in canine immune function is scarce, and relationships between SCFAs, diets as well as microbiota rarely studied and contradictory [56,64]. Thus, whether the aforementioned microbes have allergy/atopy-protecting capacity, e.g. via SCFAs, remains an open question that should be studied in controlled conditions. We found that the genus Prevotella_9 characterized the microbiota of healthy dogs while the genera Escherichia-Shigella was common in atopic dogs. In humans, the abundance of Prevotella has been negatively associated with some food allergies [52], while promoting several autoimmune diseases [65]. Escherichia-Shigella, in turn, has been associated with allergies in humans [50,53]. Thus, it seems, that Escherichia-Shigella indicates diseased conditions while Prevotella_9 may or may not be protective. Escherichia-Shigella increased also in the gut of dogs treated with antibiotics. As antibiotic use and atopic symptoms were associated, the exclusive designation of atopy-associated gut microbiota was not possible to report. As we received incomplete information on whether the antibiotics were given before or after the pet started to become atopic, we cannot either conclude that antibiotic treatment would play a predisposing role in CAD, even though this is likely, based on the existing literature. Antibiotics facilitate dysbiosis of the gut microbiota, which in turn can predispose to inflammatory disorders [5,66]. In terms of atopy, an overall risk increment of 41% for the development of AD has been reported for human subjects receiving even one course of antibiotics in early life [67]. Here, antibiotics were a source of inter-individual variation in microbiota, indicating perturbation to the gut microbiota.
As the gut microbiota is also controlled by genetic factors [68,69], one might expect that the genetic determinants of atopy could affect microbial communities. Here, the compositional differences and interindividual variation in microbiota, as well as different atopy prevalence between breeds, indicate some royalsocietypublishing.org/journal/rsos R. Soc. Open Sci. 10: 221104 genetic contribution. This is supported by studies showing different genetic risk for atopy in dog breeds [34,70,71]. Whether the gut microbiota of atopic individuals represent a proxy of their lifestyle, genetic factors, or atopic conditions per se would be important to distinguish, in order to treat atopy by manipulating the gut microbiota [72]. Nevertheless, the source for atopic-associated microbial signatures could not be traced here. Neither could it be assessed how socio-economic status of dog owners in terms of household income, education and other related aspects may have influence on the resources and interest in their pet's health, and thus generalizability of our results.
In microbiome research, animal models are useful for testing causalities and gaining more mechanistic understanding, as human microbial communities tend to represent high inter-individual variation (e.g. due to the large variation in human lifestyles). For studying relationships between lifestyle-associated microbiota and human atopy, pet dogs have several advantages. For example, canine AD shares several characteristics in common with human AD [46,73]. Importantly, dogs share the residential environment and lifestyle, sometimes even diet, with humans, contrary to laboratory rodent models. Lifestyle factors, predisposing dogs and humans to atopy, can be the same, and atopy in dogs and their owners can occur concurrently [16], which was also recently demonstrated [74] for the dogs of this study. An atopic dog more likely has an atopic than healthy owner. Lately, Coelho et al. [28] also showed that the dietary responses of the canine gut microbiota are comparable to those observed in humans. These previous observations are in line with ours. We also showed that the pet dogs' gut microbiota reflects the use of antibiotics, as found in the human gut microbiota, and diverged between atopic and healthy individuals. Our observations with the previous findings suggest that pet dogs could be a promising model for understanding associations of atopy with gut microbiota and related external factors.

Conclusion
Our study, utilizing privately owned pet dogs, demonstrates an interplay between dietary factors and antibiotic use, the prevalence of atopy and gut microbiota. The gut microbiota tends to be strongly affected by the type of diet. In addition, the microbiota diverged between atopic and healthy individuals. This is probably due to the differences in their lifestyle, such as the frequent usage of antibiotics in atopic dogs. Escherichia-Shigella, enriched by antibiotic use, seems to be a potential atopy-contributing candidate, which should be further tested in experimental set-ups. Importantly, our results are consistent with the observations from human studies showing that the gut microbiota reflects dietary and medical factors and the allergy prevalence of humans. Therefore, our results are partly applicable to humans and support the use of the canine model in exploring aetiology of atopic and allergic diseases. An interesting future direction would be to include genome-wide genetic variants to explore associations between diet and other lifestyle, genetic risk, allergy or atopy prevalence and microbiota. Data accessibility. The 16S rRNA sequences supporting the results of this article are available in the NCBI SRA under the accession SRP152712. The normalized OTU counts and related taxonomy, metadata and the variables used in the statistical analyses are accessed as electronic supplementary material, additional file 2. R scripts for statistical analyses are included as additional files 3, 4 and 5. These files were created using the R package knitr [75].
The data are provided in the electronic supplementary material [76].