Inclusion criteria consists of a household with an infant between 4 and 6 months of age that has not received any complementary foods, has one pet dog that is over the age of 6 months that has been spayed or neutered, with a breed analysis over 50% for one or more of the following breeds (Labrador Retriever, Golden Retriever, German Shepard, and Poodle) (27), and consists of at least two cohabiting adults (e.g., partners and couples). Exclusion criteria includes households with an infant that has already received complementary foods, has been exposed to antibiotics in the last month either directly (28) or indirectly via breastmilk consumption (29), that were small- (30) or large-for-gestational age (31) at birth, born at <37 weeks gestation, have growth, chromosomal, or genetic abnormalities (32), have developmental delays, or severe comorbidities that could impact growth and development. Households that have more than one pet dog; or a pet dog that received antibiotics in the last month (33), have known history of being factious, do not have proof of rabies vaccination (34), have been diagnosed with predetermined comorbidities (35, 36), or have not been examined by a veterinarian at least once in the past 12 months will also be excluded. Additionally, households where any member smokes inside the home will be excluded (37). Households that meet the criteria for both infants and pet dogs will be included in the pilot study. If either the infant or the pet dog do not meet inclusion criteria the household will be excluded.

Potential eligible participants will be recruited from email and pilot study promotional materials distributed to Midwestern University faculty, staff, and students residing in the greater Phoenix metropolitan area. Individuals that are interested will be asked to contact the research team to assess eligibility. Caretakers will be screened over the phone to determine study eligibility. If deemed eligible and willing to participate, caretakers will be asked to provide a proof of current rabies vaccination and lack of aggressive or fractious behavior of the animal toward their veterinarian or unknown people in the home. Research staff will schedule the first in-home study visit only after receipt of proof of vaccination, predominant dog breed verification, and review of behavioral status of the animal. Informed consent will take place at the first in-home study visit. If the caretaker decides they do not want to participate during this visit, the visit will end, and research staff will have no future contact with the family. Caretakers will consent to at least three additional in-home visits by research staff. The first study visit will occur when the infant is between 4 and 6 months of age, before receipt of any complementary foods, the second at ~9 months of age, and the last when the infant is approximately 12 months of age and consuming the majority of their nutrients from semi-solid and solid foods.

At all home visits, research staff will assess demographic, health history, anthropometrics, dietary history, and temperament for both the infant and pet dog to ensure continued eligibility for inclusion. Research staff will also assess the condition of the home environment. Caretakers will receive instructions from research staff on how to collect fecal samples from both the infant and pet dog prior to the scheduled in-home visit. During the in-home study visits, research staff will collect all remaining samples from the infant, pet dog, and home environment.

Infant and pet dog weight change is a primary outcome of interest. Infant length and weight will be measured with portable recumbent length boards (nearest mm) and electronic digital scales (nearest 0.1 kg), respectively. Infants will be nude at the time of weighing and will be measured in triplicate with the average weight taken as the final measurement. The weight of the pet dog will be collected using a Redmon Precision Digital Pet Scale and the body condition score (BCS) will be described using the Nestle Purina Body Condition System for dogs (38).

Infant temperament will be measured using the Infant Behavior Questionnaire (39) which is composed of six scales and asks the caretaker to rate the frequency of the infant’s behavior across a variety of situations. The Canine Behavioral Assessment & Research Questionnaire (40) will be used to assess the dog’s behavior across 13 behavioral factors to describe changes in temperament. The caretaker will complete both questionnaires during the in-home study visits and consult study staff if needed.

An age-appropriate version of the Infant Feeding and Health Questionnaire (41) will be used to collect infant dietary data supplement use, breastfeeding and formula feeding practices, receipt of commercial baby foods, and general health-related questions (41). For pet dogs, dietary information will be collected via a Canine Signalment and Health History Form which will be administered at each study visit. This will include pet owners’ feeding methods.

Study staff will mail sample collection kits the week before planned in-home visits with instructions on how to collect the infant and pet dog stool sample the day prior to scheduled in-home visits. Study staff will call and text message caretakers the day prior to the scheduled in-home visit to remind them to collect the infant and pet dog stool samples. Caretakers will swab the infant’s diaper and pet dog stool immediately after defecation using sterile swabs and tubes containing DNeasy buffer. Samples will be stored in a sealed container at −20°C until research staff collect the tubes at scheduled in-home visits. The stool sample will be transferred on dry ice to the laboratory and stored at −80°C until further processing.

Research staff will collect infant and pet dog saliva samples, infant and pet dog skin samples, and environmental samples during the in-home study visits. Infant and pet dog saliva samples will be collected using SalivaBio collection kits according to the manufacturer protocol (42). Immediately after collection, the swab will be placed in a syringe to express the saliva into a cryovial for storage. The process will be repeated until a 200 ul saliva sample is collected. Research staff will use sterile swabs and tubes to collect samples from the infant’s hands and feet, and pet dog’s paws. One swab each will be used for hands, feet, front paws, and back paws. Research staff will use sterile swabs and tubes to collect samples from environmental surfaces within the home including the floor of the infant and pet dog feeding area, the high-chair surface and seat, the pet dog food and water bowl, and the main indoor play area. One swab will be used for each surface area. All collection swabs will be sealed in a conical tube, transferred on dry ice to the laboratory, and stored at −80°C until further processing.

Microbial genomic DNA will be extracted from fecal, saliva, and environmental samples, as described by the manufacturer (Qiagen Sciences LLC, Germantown, Maryland), using DNeasy PowerSoil Pro DNA isolation kits and a vortex-based beadbeater, the recommended method of the National Institute of Health Human Microbiome Project. Amplification of the 16S rRNA gene sequence will be completed in triplicate PCRs using 96-well plates. The V4 region of the 16S rRNA gene will be amplified by utilizing forward 515F primers and 806R reverse primers containing Illumina adapter sequences. PCR amplification, amplicon cleaning, and quantification will be performed following previously outlined protocols (43). Sequencing will be performed by using the Illumina MiSeq (SN M02149, with the MiSeq Control Software v 2.5.0.5). Pooled sequences will be de-multiplexed and quality filtered using the QIIME2 software package. Sequences will be assigned to operational taxonomic units (OTU) with a 97% similarity threshold using QIIME’s uclust-based open-reference OTU picking protocol against most recent SILVA or GreenGenes reference databases. Sequences that do not match the reference database will be clustered de novo including all sequences in the analysis. Core diversity analyses will be performed OTU tables, including alpha and beta diversity and taxonomic summaries as provided by QIIME software. Collaboration with scientists experienced in performing the above-mentioned bioinformatics is necessary for this pilot study.

For this pilot study, we will control for gender, age, and diet of the pet dog and human infant in all analyses. Furthermore, we will control for birth mode and breastfeeding status (44) of the human infant in all our analysis as research has demonstrated key differences in the GM of infants based on these factors. In addition to the above-mentioned covariates, for aims 2 and 3, we will also include the following human infant covariates: maternal pre-pregnancy body mass index (45), gestational weight gain (45), and maternal education (46) as previous research have shown their impact on adiposity or rapid infant weight gain among infants. In aims 2 and 3, we will also include covariates related to the most common comorbid conditions in dogs which lead to obesity, hypothyroidism and hyperadrenocorticism (35, 36). There are a number of environmental factors that may impact the health of the infant and pet dog. Several key household factors have been associated with greater risk of obesity among pet dogs including pet owner age, income, and frequency of exercise (25). The number of people residing in the household may impact microbial transfer between infant and pet dog (12). These variables will be assessed during pilot study participant evaluations and controlled for as needed in aims 2 and 3.

Additional factors related to details pertaining to the types of daily interactions that infants have with their pets such as licking/petting/other potential mechanisms of microbiota exchanges outside of feeding times will also be collected during this pilot study. Similarly, information related to animal husbandry practices and habits such as feeding table scraps or allowing a pet to clean the floor under the highchair will be collected through interviews during home visits with the primary adult caregiver. Due to sample size limitations in this pilot study these variables will not include as covariates. The availability and analysis of the resulting data will be utilized as an important tool to inform future research questions and hypothesis for larger follow-on studies.

In microbiome studies, power analyses and sample size estimates can be difficult because the true effect size is unknown. Because of this and that our proposed study is a pilot study, we based our sample size estimate on a recently published protocol paper for an observational study among infants followed through the first 2 years of life which estimated statistical power using simulation yielding a total sample size of 82 (47). This aligns with a sample size calculation using G*Power for ANCOVA with two groups with an a priori effect size of 0.47, 80% power, and 8 covariates, yielded a total sample size of 83. Furthermore, Kers and Saccenti recently published statistical power and sample size estimates from simulation of data from the Human Microbiome Project and found to achieve observe differences in microbial features of 25% with 80%, the sample size would range from approximately 40–90 for alpha diversity metrics, and approximately 15–50 for beta diversity metrics (48). Based on these publications, we feel that a total sample size of 82 is sufficient for a pilot study.

All statistical analyses will be completed using Statistical Package for the Social Sciences 25 (SPSS, Inc., an IBM Company, Chicago, Illinois) and QIIME 2020.2 statistical bioinformatics software packages. Data will be expressed as mean ± SD or median (interquartile range) for demographic, anthropometric, temperament scores, and microbiota frequencies based on the normality of the data. Phylogenetic diversity measures will be performed using QIIME 2 longitudinal plug-in to determine alpha (within-sample) diversity via Observed OTUs, Faith’s PD and Shannon’s Index to assess richness and evenness. Median alpha-diversity values will be visually reported using distribution comparison plots (box-and-whisker plots). Beta-diversity (between-sample) metrics to explore community composition will be assessed by principal coordinates analysis (PCoA) plots and visualized using EMPeror (via QIIME 2). Four ß-diversity metrics will be evaluated including Jaccard, Bray-Curtis, Unweighted Unifrac, and Weighted Unifrac. Outliers will be identified using the volatility visualizer to assess the volatility of the dependent variables over the independent variable. Visual identification of outliers that are disproportionately impacting the variance within individuals and groups will be removed. Scatterplots with the metric predictions of each sample on the x-axis and the residual for each sample on the y-axis will be examined to evaluate patterns or non-random distributions. Residuals should be roughly zero-centered and normal across the range of measured values. Non-normal data will be centered log-ratio transformed prior to analysis. To evaluate Aim 1, Pearson correlation will be used to examine associations between the GM of the human infant and pet dog with microbial composition and diversity of the skin, saliva, and home environment at each time point with alpha- and beta-diversity metrics for the different microbiome samples analyzed separately. Paired and difference distance comparisons will be evaluated to assess the changes in the gut microbiome between the human infant and pet dog at each time point. Group differences (human infants and pet dogs) will be evaluated by performing linear mixed effects (LME) models at each timepoint (before, during, and after CF) while controlling for diet (human infant and pet dog), gender (human infant and pet dog), age (human infant and pet dog), infant breastfeeding status, and infant birth mode as fixed effects and subject ID as random effect. Alpha- and beta-diversity metrics will be separately analyzed as dependent variables with group assignment as the independent variable. LME models will be fit with all main effect interactions as fixed effects, with insignificant terms removed from the final model for that outcome of interest. All analyses will be performed using the linear-mixed-effects action in the q2-longitudinal plugin for the QIIME 2 microbiome analysis platform. Findings will be considered significant based on a p < 0.05 level of each variables Z-score estimate.

To evaluate Aim 2, to examine longitudinal changes in microbiomes, LME models will be used to compare changes in pairwise first differences (i.e., magnitude of change) for alpha diversity metrics and pairwise first distances (i.e., rate of change) for beta-diversity metrics while controlling diet (human infant and pet dog), gender (human infant and pet dog), age (human infant and pet dog), infant breastfeeding status, infant birth mode, pre-pregnancy maternal BMI, gestational weight gain, maternal education, canine hypothyroidism and canine hyperadrenocorticism, household income and size, and age and exercise level of main pet caretaker as fixed effects and subject ID as random effect. First difference and distance calculations will be performed at fixed time intervals with the “before CF” timepoint representing the baseline measurement for the human infant and pet dog separately. The pairwise first difference and distance between subjects (human infant and pet dog) will be calculated for a given timepoint and included in the LME models as the independent variable. Subjects that are missing samples for this baseline timepoint will not be included in the analysis. LME models will be used to examine the relationship between the independent variables of interest and the longitudinal response that will include the intercept and slope term as both fixed and random effects. The fixed intercept and slope represent the regression line for the average subject, whereas the random intercept and slope reflect individual departures from the average line for each subject (49). LME models will be fit with all main effect interactions as fixed effects, with insignificant terms removed from the final model for that outcome of interest. All analyses will be performed using the linear-mixed-effects action in the q2-longitudinal plugin for the QIIME 2 microbiome analysis platform. Findings will be considered significant based on a p < 0.05 level of each variables Z-score estimate.

To evaluate Aim 3, assess whether the changes observed in Aim 2 are correlated with subsequent changes in infant and pet dog adiposity and temperament, we will use the q2-longitudinal’s feature-volatility action. First, we will visually analyze the data by evaluating volatility charts to identify specific timepoints where infant and pet dog temperament or adiposity appears to have a significant impact on microbial diversity. Volatility charts combine the attributes of control charts and spaghetti plots to produce interactive HTML-based visualization where we can select which alpha- or beta-diversity metric is plotted on the y axis, choose the categorical sample metadata column used to group the data. Next, we will use the feature-volatility action to produce an interactive visualization of the longitudinal abundance, importance, and descriptive statistics for all important features. The longitudinal abundance of each feature is plotted using volatility plots (as described above). Important features identified change over time with their abundance predictive of the specific time point when the sample was collected. Finally, we will use LME models (as described in Aim 2) to test whether alpha- and beta-diversity were impacted by changes in adiposity and temperament of the human infant and pet dog including the same covariates as described in Aim 2. For these models, the random effects in the model will include subject ID as the random intercept and timepoint as the slope.