# paper.pdf ## Metadata - PDFFormatVersion=1.7 - Language=en-US - IsLinearized=true - IsAcroFormPresent=false - IsXFAPresent=false - IsCollectionPresent=false - IsSignaturesPresent=false - CreationDate=D:20260611192245+05'30' - Creator=Adobe InDesign 19.5 (Windows) - ModDate=D:20260611192246+05'30' - Producer=Adobe PDF Library 17.0 - Title=The role of dog keeping in the home microbiota and its impact on children's health - Trapped=False - Custom.WPS-ARTICLEDOI=10.1111/pai.70408 - Custom.WPS-JOURNALDOI=10.1111/(ISSN)1399-3038 - Custom.WPS-PROCLEVEL=3 - jav:journal_article_version=VoR - prism:doi=10.1111/pai.70408 - pdfx:wps-articledoi=10.1111/pai.70408 - pdfx:wps-journaldoi=10.1111/(ISSN)1399-3038 - pdfx:wps-proclevel=3 - dc:title=The role of dog keeping in the home microbiota and its impact on children's health ## Contents ### Page 1 Pediatr Allergy Immunol. 2026;37:e70408. | 1 of 11https://doi.org/10.1111/pai.70408wileyonlinelibrary.com/journal/paiReceived: 11 September 2025 | Revised: 1 April 2026 | Accepted: 2 June 2026DOI: 10.1111/pai.70408O R I G I N A L A R T I C L E The role of dog keeping in the home microbiota and its impact on children's health Jenni M. Mäki1,2 | Pirkka V. Kirjavainen1,3 | Martin Täubel1 | Pauli Tuoresmäki1 |Eija Piippo-Savolainen2,4 | Katri Backman2,4 | Juha Pekkanen1,5 | Anne M. Karvonen1This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.© 2026 The Author(s). Pediatric Allergy and Immunology published by European Academy of Allergy and Clinical Immunology and John Wiley & Sons Ltd.1 Department of Public Health, FinnishInstitute for Health and Welfare, Kuopio,Finland 2Department of Paediatrics, KuopioUniversity Hospital, Kuopio, Finland3 Department of Clinical Nutrition, Institute of Public Health and Clinical Nutrition, University of Eastern Finland,Kuopio, Finland4University of Eastern Finland, Kuopio,Finland 5 Department of Public Health, University of Helsinki, Helsinki, FinlandCorrespondenceJenni M. Mäki, Department of PublicHealth, Finnish Institute for Health andWelfare, Neulaniementie 4, 70701 Kuopio,Finland. Email: jmtolvan@student.uef.fiFunding informationAcademy of Finland, Grant/AwardNumber: 296814, 296817, 139021,287675, 308253 and 308254; JuhaVainio Säätiö, Grant/Award Number:201710468; The Foundation for PediatricResearch; Yrjö Jahnssonin Säätiö; SuomenKulttuurirahasto; Kuopion SeudunHengityssäätiö; Terveyden ja hyvinvoinninlaitos; Tampereen Tuberkuloosisäätiö;Päivikki ja Sakari Sohlbergin Säätiö;Maatalousyrittäjien eläkelaitos - MelaEditor: Jon GenuneitAbstract Background: Children with dog contacts in early childhood are healthier and use lessantibiotics during the first year of life. We have earlier identified dog-associated bac- terial and fungal signals in house-dust microbiota. The aim of this study was to deter-mine whether these signals in house dust mediate the dog keeping associated positive health effects in the first year of life.Methods: We studied 368 Finnish children from pregnancy and collected weekly diarydata on otitis, fever, usage of antibiotics, and dog keeping from 9th to 52nd postnatalweeks. The dog-associated signals in house dust microbiota included the abundanceof 12 bacterial and two fungal genera, which were determined alongside richnesswith Illumina MiSeq sequencing from floor dust samples collected at the child age of2 months. Bacterial cell concentrations in floor dust were measured with quantitativePCR (qPCR). Generalized estimating equations (GEE) were used for statistical analysesof the microbe-health associations.Results: Eight of the dog-associated genera individually explained 10%–23% of theprotective associations between dog keeping and at least one health outcome. The strongest effect was observed for the Pasteurella in relation to fever weeks. In addi- tion, three correlated bacterial genera together explained 25% of the dog-associatedreduction in antibiotic use. Bacterial or fungal richness, total cell concentrations, orproportion of human-sourced bacteria in house dust did not significantly mediate thedog-associated health effects.Conclusions: Specific dog-associated bacterial and fungal genera in house dust, butnot richness, explained partly the associations between dog keeping and lower preva-lence of respiratory infections.K E Y W O R D Schildren, dog, house dust microbiota, indoor microbiome, respiratory infection ### Page 2 2 of 11 |MÄKI et al.1 | I NTRO D U C TI O NExposure to dogs and other pet animals may have positive healtheffects on human health such as higher frequency of physical activity;children interacting with a dog seem to have increased confidence and prevention of respiratory tract infections1–3 which are known to increase the risk for childhood asthma. 4 Some studies suggestthat exposure to a dog reduces the risk of allergic diseases, 5–7 while some claim that the dog increases the risk of allergy and asthma in children. 8 However, it is not known what is behind this phenomenon.One possibility is the microbial exposures in early childhood, 6,9–11which are known to affect development of the child's immune system and microbiota and therefore the risk of disease.12,13 For example,the microbiota in the home has been found to explain some of theprotective effects of cattle farms against asthma and allergies.10,14 Earlier studies have determined the impact of pets on microbi-ota of the home9,11,15–20 and the human gut. 21,22 Pets alter the mi-crobiota of the homes 9,17 and exposure to furry pets is seen in theinfant gut microbiota composition at 3–4 months of age.23 Whetherthese effects on house dust microbiota mediate the pet-associatedbeneficial health effects in children, such as the lower prevalence of respiratory infections, has not been studied.In the same Finnish birth cohort (LUKAS2), as in the presentstudy,1 we previously reported that children who had dog or cat con- tact(s) at home early in life were mostly healthy, had less fever andotitis, and tended to need fewer courses of antibiotics before theage of 12 months. One plausible mechanism of these pet-associatedbeneficial effects is through modifying the early-life indoor micro- bial exposures and its consequent beneficial effects on maturationof the immunological defense and human microbiome. We char- acterized the effects of dog keeping on house dust microbiota inLUKAS2 homes and determined the reproducibility of these effectsin a German birth cohort, LISA.9 The relative abundances of twelve bacterial and two fungal genera were significantly higher in dog thannon-dog households. In addition, bacterial richness (Chao1) and theconcentrations of Gram-negative bacteria were significantly higher,and the proportion of human-sourced bacteria as well as the con- centration of Gram-positive bacteria were lower in the householdswith dog compared to non-dog households. Fungal richness (Chao1)was higher in the dog homes in LUKAS2, but not in LISA, and the dif-ference was significant only when the ground was covered by snow.The aim of this study was to determine whether the above-mentioned bacterial and fungal signatures in house dust could ex-plain the association between dog ownership and the positive health effects in children in the first year of life.2 | M E TH O DSTwo birth cohorts (LUKAS1 and LUKAS2) consisting of 442 Finnishchildren were followed up from the third trimester of pregnancythrough to 1 year of age.24 All mothers gave birth betweenSeptember 2002 and May 2005. The ethical permission was grantedby the Research Ethics Committee, Hospital District of NorthernSavo, Kuopio. A written informed consent was obtained from theparents of the participating children.Parents filled weekly questionnaires consisting of questions re- lated to infectious symptoms during the past 7 days, children's an- imal contacts and breastfeeding (exclusive, partial, not breastfed).The diary questionnaires began from the 9th postnatal week andcontinued up to 1 year of age. Parents were first enquired whethertheir child had been “hale and hearty” during the last 7 days. If not,parents also answered questions about symptoms and diseases.Based on the earlier report on dog-associated health effects in thiscohort,1 the numbers of reported healthy weeks, weeks of antibioticuse, weeks of otitis and weeks of fever were studied (Table 1). Themaximum number of diary weeks was 44. We excluded children withless than 23 weeks of diary responses from analyses (n = 27) (moreinformation on Data S1 and in Bergroth et al.1).2.1 | Dust samplingLiving room floor dust samples were collected at 2 months of age usinga nylon sampling sock and by vacuuming an area of 1 m 2 from a rugfor 2 min or an area of 4 m2 from a smooth floor for 2 min. Dust sam- ples were homogenized by sieving through a sterile strainer, dried ina dessicator at 4°C and then stored at −20°C until DNA extraction.2.2 | Characterization of the microbesThe house dust microbiota characterization has been described indetail previously.9 In brief, DNA from dust samples was extractedusing a bead-beading method, followed by kit-based clean-up aconcentration. The bacterial/archaeal 16S ribosomal RNA gene V4region was amplified using 515F/806R primers 25 and the fungal in- ternal transcribed spacer (ITS) region by ITS1F/ITS2 primers.26 Thegenerated DNA amplicons were sequenced with Illumina MiSeqwith V3 chemistry. The bioinformatic processing followed the same pipeline as we have earlier described.9 In short, the sequences wereclustered into operational taxonomic units (OTUs) at 97% similarityusing the open-reference OTU-picking approach. Richness (Chao1)was calculated with QIIME (Quantitative Insights Into MicrobialEcology).27 For the analysis of individual fungal taxa, taxonomicKey messageDog-associated microbiota in the house dust affectpositively children's health during the first year of life.Dog-associated bacterial and fungal genera or theircombinations in house dust may explain up to 1/5 of theassociations between dog keeping and lower prevalence of respiratory infections in the first year of life. ### Page 3 | 3 of 11MÄKI et al.classification was done using FHiTINGS (Fungal High throughputTaxonomy Identification in NGS).28 After quality filtering, data onbacterial microbiota was available from the homes of 394 children and on fungal microbiota from the homes of 382 children.Bacterial concentrations in house dust were measured usingquantitative polymerase chain reaction (qPCR) duplex assays forparallel determination of Gram-positive and Gram-negative bacte- rial DNA.29 Microbial concentrations were analyzed as cell equiva- lents per milligram dust (CE/mg). In addition to these, three allergens(cat allergen (Fel d1), dust mite Dermatophagoides farina (Der f1), andstorage dust mite Dermatophagoides pteronyssinus (Der p1)) had beenpreviously determined in LUKAS1 from the 2mo dust sample.2.3 | Statistical analysesThe microbial signatures studied in here for the mediation effectwere selected based on our earlier findings in LUKAS2.9 The se- lection of microbial genera associated with dog-households wasbased on FDR-corrected p-values in ANCOM analyses.30 These in- cluded 12 bacterial genus-level taxa, two fungal genera and bacte- rial and fungal richness, the concentrations of Gram-negative andGram-positive bacteria, and proportion of human- sourced bacteria[human source proxy, HSP] in house dust (see more informationin the Data S1). Correlations between the different microbial sig-natures were calculated using Spearman's rank correlation, whichwere also used for network analyses for studying taxa interactions(ρ ≥ 0.5). For statistical analyses, the microbial signatures were di- vided into tertiles (see descriptives in Table S1), except HSP, forwhich the ln-logarithm variable was used, and the results were ex- pressed per interquartile range (IQR) change. If the percentage ofsamples under detection level was above 33.3%, then the valuesabove detection level were divided into two equal classes usingmedian as cut off. Generalized estimating equations (GEE) methodTA B L E 1 Characteristics of the study population.Study population Diary weeksN = 368 N = 15,686Characteristic n % n % Sex Male 184 50.0 – –Female 184 50.0 – – Birth weight<3480 120 32.6 – – 3480–3810 126 32.2 – –>3810 122 33.2 – –Dog ownership No dog – – 10,667 68.0Owning a dog(s) – – 5019 32.0BreastfeedingNo – – 6925 44.2 Partly – – 6524 41.6 Solely – – 2237 14.3 Number of older siblings 0 128 34.8 – –1 124 33.7 – – ≥ 2 116 31.5 – –Living environment Farm 112 30.4 – – Rural, non-farm 176 47.8 – –Suburban 80 21.7 – –Season of birthWinter 101 27.5 – –Spring 111 30.2 – –Summer 63 17.1 – – Autumn 93 25.3 – – Maternal smoking^No 344 93.5 – –Yes 24 6.5 – –Parental atopy No 168 45.7 – –Yes 200 54.4 – –CohortLUKAS1 192 52.2 – – LUKAS2 176 47.8 – –Healthy weeks No – – 5010 31.9 Yes – – 10,676 68.1Antibiotic usage (weeks)No – – 15,105 96.3 Yes – – 581 3.7(Continues)Study population Diary weeksN = 368 N = 15,686Otitis (weeks)No – – 14,869 97.4 Yes – – 399 2.6Missing 418 Fever weeks No – – 14,631 95.8 Yes – – 637 4.2Missing 418Note: N/n the number of observations (in study population) or weeks(in diary weeks) in total and within category; % the percentage of thecategory. Maternal smoking^ when children were at 2 months old. –indicates missing formation due to the information was collected eitherin questionnaires (during pregnancy or follow-up questionnaire at theage of 2 months) or in weekly diary.TA B L E 1 (Continued) ### Page 4 4 of 11 |MÄKI et al.with working correlation matrix AR (1) was used for statisticalanalyses. In the mediation analyses, the percentage change in the dog's estimate was calculated from estimates given by two mod- els: (1) an original model in which the confounding factors were ad- justed and (2) the original model in which was additionally adjustedfor a microbial signature. All models were adjusted similarly as inBergroth et al.1 in order to study the mediation effect of the dog's microbiota on the previously reported inverse association betweenthe dog and respiratory infections. A priori selected adjustmentswere related to general risk factors for allergic diseases gender,number of older siblings (none, 1, ≥2), maternal smoking (childrenage 2 months), parental atopy, weekly data on breastfeeding (exclu- sive, partial or not breastfed), birth weight (in tertiles), and seasonof birth (winter, spring, summer, autumn) and study characteristics(living environment (farm, rural non-farm, suburban), the order ofdiary month, and cohort (LUKAS1, LUKAS2)).3 | R E S U LT SA total of 15,686 diary weeks were recorded from the 368 childrenwho had information on any of the four outcomes and on dog keep- ing and either bacterial or fungal sequencing data and had no missingdata on a priori selected confounders (Table 1). The study samplecomprised an equal number of boys and girls. Twenty-four mothers(6.5%) smoked, and two of them smoked indoors. Of the children,112 (30.4%) lived on a farm, 176 (47.8%) in a rural area (non-farm),and 80 (21.7%) in a suburban area. Of the 14 dog-associated genera, Clostridium and Fusobacteriumhad the highest median abundance in dust samples (Table S1). Allof the 12 bacterial genera correlated positively with each other(mostly ρ = 0.2–0.4); correlations with the two fungal genera wereweaker, though still largely positive (Figure S1). We observed thestrongest correlations (ρ >0.7) between Pasteurella, Lampropedia, and Conchiformibius genera.3.1 | Associations between 14 genera and fourhealth outcomesThe higher levels of Collinsella were positively associated withhealthy weeks (Figure 1A) and inversely with the use of antibiotics(Figure 1B) and weeks with otitis (Figure 1C) when dog keeping wastaken account (Trend test p < .05, Table S2), and a similar tendency(p = .11) was seen with fever weeks (Figure 1D). The higher levelsof Ruminococcus candidate group and Udeniomyces were inversely associated with fever weeks. There was a tendency that higher lev-els of Conchiformibius and of Megamonas were associated with moreantibiotic treatment weeks and more fever weeks (p = .09 and 0.07),respectively. In models including all 14 genera simultaneously, Collinsellawas, independently of the other genera, positively associated withhealthy weeks (Figure 1A) and inversely with weeks with otitis(Figure 1C); Pasteurella was inversely associated with the antibiotictreatment weeks (Figure 1B) and weeks with fever (Figure 1D), andRuminococcus candidate group was inversely associated with feverweeks (p < .05, Table S3).3.2 | How much 14 different genera explain theassociation between a dog and a health effect?Out of the 14 genera, Conchiformibius and Lampropedia did not ex-plain the dog effect, whereas Collinsella, Fusobacterium, Mycoplasma, and Leucosporidiella explained only a minor part (<10%), and therest of the eight genera each explained 10%–23% of the associa-tion between dog keeping and at least one of the studied outcomes(Figure 2), which were in healthy weeks: Clostridium, Eubacteriumcandidate group, Helicobacter, Megamonas, and Pasteurella; in antibi-otic treatment weeks: Pasteurella and Sutterella; and in fever weeks:Pasteurella, Ruminococcus candidate group, and Udeniomyces. Noneof the genera explained the association between dog and otitis morethan 6%. In addition, the inverse associations between dog owner-ship and the use of antibiotics were no longer significant after ad-justment with Pasteurella or Sutterella, nor were fever weeks after adjustment with Pasteurella (p > .1, Figure 2).The simultaneous adjustment for the 14 genera did not, per se,explain the dog-associated associations with the health outcomes,but rather fortified these associations, except for healthy weeks, forwhich the effect remained unchanged (Table S3). Due to the co-occurrence of the genera, we checked whetherdog-associated genera together could explain more of the healthbenefits of dog keeping than the single genera, and thus, we usedtwo methods: (1) correlation-based network analyses and (2) sum-ming up the selected genera. The first method distinguished twogroups of three bacterial genera (Figure S2). When the first group(A), which included Collinsella, Ruminococcus candidate group, andSutterella, was adjusted in the model simultaneously with their in- teraction terms, they explained 2%–25% of the inverse associationbetween the dog and respiratory infections (Table 2). However, thesecond group (B) (Lampropedia, Pasteurella, and Conchiformibius) didnot explain the association but rather strengthened the protectiveeffect of the dog. In the second method, the relative abundancesof 2–7 genera, which were significantly (Trend test p < .1) associatedwith a health outcome when dog keeping was not taken into account,were summed up (Table S4). The sums of selected genera were for‘healthy weeks’ Clostridium, Collinsella, Eubacterium candidate group,Helicobacter, Megamonas, Pasteurella, and Ruminococcus candidate group; for the ‘antibiotic treatment weeks’ Collinsella and Pasteurella; and for the ‘weeks of fever’ Collinsella, Pasteurella, Ruminococcuscandidate group, and Udeniomyces. Only Collinsella genus was sig-nificantly associated with otitis and thus, the sum variable was not created. The sum variables were significantly associated with givenoutcomes (Table S4) and similar tendency (p ≤ .10) were seen when ### Page 5 | 5 of 11MÄKI et al.dog keeping was taken account (Table S5). The sum variables did notexplain more the effect between dog keeping and health benefitsthan the single genera did: the sum of selected genera for healthy,for antibiotic treatment, and for fever weeks explained 21%, 17%,and 15%, respectively.3.3 | Other microbial signatures and four healthoutcomesAmong the bacterial loads and richness in house dust, only the con- centration of Gram-negative bacteria (measured via qPCR) wasOne−by−one Mutually adjusted0.0 0.5 1.0 1.5 2.0 2.5 0.0 0.5 1.0 1.5 2.0 2.5Udeniomyces (fungi)Leucosporidiella (fungi)Sutterella Ruminococcus candidate groupPasteurella Mycoplasma Megamonas Lampropedia Helicobacter Fusobacterium Eubacterium candidate groupConchiformibius Collinsella ClostridiumaOR (95% CI) middle vs. lowest highest vs. lowest low vs.