Influence of Water Delivery Method on the Gut Microbiome in Laboratory Mice (Mus musculus)

Abstract

Nuances related to the milieu of the gastrointestinal tract have led to investigations of environmental (or extrinsic) factors, like feed sources and fluid intake, and their influences on the gut microbiome in research animals. Water is typically provided to laboratory mice either by reusable water bottle (RWB), housing rack automatic water (RAW) delivery, or single-use disposable plastic pouch (DPP). In this study, the influence of differing water delivery methods on gut microbiome stability was evaluated in immunocompetent (n = 36 B6; 18 male [M]:18 female [F]) and immunocompromised (n = 36 NOG; 18 M:18 F) strains of mice. Mice were housed on a single IVC rack in sex-specific groups and provided with autoclaved caging and bedding, irradiated feed, and chlorinated, reverse-osmosis water provided by one of 3 delivery methods (8 cages per method). Access to the room was restricted to select personnel to conduct animal care and sample collection tasks. Fecal pellets (n = 2) were collected from each animal every other week, and water samples were collected weekly for analysis. Over the course of the study, bacteria were detected in 11% of the RWB samples (7 of 63) and 4% of the RAW samples (1 of 25). DPP samples were consistently free of bacterial contamination. Shotgun metagenomics and statistical analyses revealed overt shifts in gut microbiota in the majority of mice throughout the study (21 of 25 cages). Histologic examinations of organs from representative clinically normal study mice (n = 12) were unremarkable. With minimal exceptions, microbiome shifts were statistically significant across cage cohorts, despite attempts to control experimental variables. This study is the first to demonstrate that the water delivery method does not impart a significant influence on gut microbiota stability in research rodents and highlights the need to document water type, treatment, and delivery method as extrinsic factors in reporting animal studies.

Introduction

Water is one of many essential provisions for rodents used in research, and expectations regarding water quality and delivery are outlined in the Guide for the Care and Use of Laboratory Animals.¹ In brief, access to potable, uncontaminated drinking water is required, and periodic water quality assurance checks for pH levels, hardness, and microbial or chemical contamination may be necessary.¹ To minimize or eliminate potential contaminants, water may be purified or treated; however, common water treatment methods have the potential to influence animal physiology.^1,2 For example, water acidification and the mode of acidification are known to eliminate bacterial contaminants yet have been shown to impact animal behaviors, gut microbiomes, and fecal metabolomes of mice.³ It is prudent to mention that commercial animal vendors that breed and supply research rodents treat their drinking water in distinct ways; this results in a variety of acidification methods, pH levels, filtration mechanisms, and UV-radiation exposure for the water sources provided to transgenic strains and outbred mouse stocks, all of which may impact water consumption by these animals and ultimate model phenotypes.^4,5

Microbiomes are closely integrated with immunologic and metabolic functions in mammalian species, and this highly complex microbial-host ecosystem can have a crucial impact on host health.^6–8 With the increasing recognition that water directly impacts gut microbiota, there are surprisingly few evidence-based publications that have investigated the impact of water source on the mouse microbiome.^9,10 In a study⁹ of differing water sources (tap water (autoclaved and nonautoclaved), bottled mineral water, and disinfected water) provided to BALB/c female mice over a time course of 23 days, gut microbiota shifts were attributed to the varying water types, with further speculation that these gut perturbations could impact antibiotic resistance. Further, adequate water intake is critical for maintaining bacterial and immunologic homeostasis in the gut, thereby enhancing host defenses against enteric pathogens, as has been shown in humans and laboratory mice.¹¹ Water consumption, and subsequent levels of hydration, have been shown to accelerate the transfer of interleukins to mesenteric lymph in rats, which may contribute to the maintenance of innate immunity.¹² Overall, due to these impacts of fluid intake by research animals, water type, treatment, and delivery methods should be considered and documented as extrinsic factors, in addition to other factors like caging, bedding, enrichment, room cycle lighting, feed type, and baseline gut microbiome and gut microbiota assessments.^13–17

Despite regulatory guidance on water quality for research rodents, methods of water treatment are rarely reported in publications, which may compromise experimental reproducibility. A recent survey⁴ of microbiome publications (n = 76) found that papers underreport water type and delivery method, despite reporting numerous other extrinsic variables, like specific pathogen-free (SPF) facility status, chow type, and processing (autoclaved or irradiated), bedding, temperature control, housing (individually ventilated or static racks) and light/dark cycle. This survey⁴ of the literature estimated that approximately 62% (47 of 76) of papers did not report any information regarding the water source for animals and approximately 24% (18 of 76) only reported limited aspects of water treatment (for example, filtered, sterile, or acidified, notably without mention of pH or the acid type used), leaving only approximately 14% of papers that included sufficient detail to replicate watering aspects of the published study. Notably, this survey did not include whether common water delivery methods for rodents (rack autowater [RAW], autoclaved reusable water bottles [RWB], and disposable plastic pouches [DPP]) were described in publications, despite indications that water quality is crucial for research animal health.¹⁸

While attempts are made to control variables that might impact microbiome research, a focus on standardizing water delivery and sources as key extrinsic factors is recommended. In the spirit of experimental reproducibility, water type, and delivery should be clarified when procuring research animals, with consideration of specific husbandry approaches at originating rodent vendors, through to delivery of animals to destination sites and final housing and husbandry arrangements. Due to the recognized gap in the scientific literature about the influence of water delivery and source on the microbiome, this study was designed to test the hypothesis that the stability of the gut microbiomes in immunocompetent mice (B6 strain) compared with immunodeficient mice (NOG strain) would be improved by providing water through a dedicated delivery type while controlling for cohoused intracage effects that have been observed in prior publications.^15,19 The overall intent was to identify relevant methods to limit mouse microbiome variability over time and disseminate this information throughout the research animal community.

Materials and Methods

Animals and originating water source.

Five-week-old male and female Black 6 (B6NTac; n = 36) and NOG (CIEA NOG; n = 36) mice were studied to assess immunocompetent and immunocompromised strains, respectively. The B6NTac females (n = 18) came from a vendor location (site A) distinct from the other mice. Per the vendor, at site A, animal drinking water is derived from wells on the property treated to human potable standards. The water is delivered into large holding tanks on site and then softened and distributed to each location. Within most barrier breeding facilities at site A, water is hyperchlorinated and then passed through a series of 0.2-μm filters. Filters are routinely assessed and changed when the pressure drop exceeds prescribed limits (12 PSI across the first 0.2-μm filter or 15 PSI across the filter bank). Animal drinking water is hyperchlorinated to 5 to 10 ppm at the animal consumption point. Water is provided ad libitum to barrier animals via an automatic watering system or individual water bottles. Conversely, all CIEA NOG males and females, as well as B6NTac males originated from site B (from differing rooms within the site), where water is acquired from municipal utilities. As with site A, water at site B is softened and chlorinated to 5 to 10 ppm at the animal consumption point. Water is passed through a series of filters culminating in 0.2-μm filters. Animal drinking water is supplied by automatic watering racks or water bottles to vendor mice.

Husbandry.

Mice were permitted to acclimate for 2 wk after arrival from vendor sites and the first fecal samples were collected on day 7, as a “baseline” sample, during this time frame. All procedures were performed within a facility accredited by AAALAC International with the approval of the University of Pennsylvania IACUC following the Guide for the Care and Use of Laboratory Animals.¹ Animals were separated by strain and sex and placed in cohorts (n = 3/cage; 12 cages/strain) with 8 cages/water type or 24 cages total in the study. Individual mice were identified by ear punch, as per IACUC guidelines. Animals were housed in IVC racks (NexGen MAX; Allentown LLC, Allentown, NJ). Cage components, including bedding (1/4-inch corncob bedding from Animal Specialties and Provisions, Quakertown, PA), cotton square nesting material (Ancare, Inc. Bellmore, NY), and water bottles/sipper tubes (Allentown) were autoclaved and changed every other week throughout the 8-wk study. Animals were housed in 3 columns across the rack, with NOG mice cages placed into rows located above B6 mice cages in each column and NOGs handled before B6 mice at cage changes. Mice were housed in a 12:12 light-dark cycle and provided with irradiated chow (5053 PicoLab Rodent Diet 20; LabDiet supplied through Animal Specialties and Provisions).

Caging equipment was changed within a Class II, Type A2 Biologic Safety Cabinet (NuAire, Plymouth, MN). The interior surfaces of the hood and external surfaces of IVC cages were sanitized (Super Sani-Cloth; PDI Healthcare, Inc., Woodcliff Lake, NJ) before the cages were opened. Animals were handled by dedicated personnel wearing disposable gloves sanitized with disinfectant (1:18:1 Clidox-S; Pharmacal Research Laboratories, Inc., Waterbury, CT; minimum of 8-min contact time) between cages to prevent cross-contamination.

Sentinel testing was not incorporated in this study, given the limited time of the experimental investigation. Staff access was highly restricted to limit variability related to personnel: only one animal caretaker (DA), 2 facility supervisors (JB; TBM), 1 graduate student (AK), and 2 veterinarians (FCH; JOM) were permitted to enter the room for daily health checks, supply delivery, cage changes, and sample collections. No other animals were housed in the study room during the experiment.

Experimental water sources.

Distinct water delivery methods were investigated, including RAW, RWB (9-oz/266-mL volume with neoprene stoppers and stainless-steel sipper tubes; Ancare, Inc., Bellmore, NY), and DPP. Source water to the university is provided by the city of Philadelphia and is drawn from the 2 adjacent rivers, the Delaware and the Schuylkill Rivers (https://water.phila.gov/drinking-water/). Source water was treated by reverse osmosis and manually chlorinated at the facility source holding tank to achieve a range from 2 to 4 ppm (ExStik CL200 Waterproof Chlorine Meter; ExTech Instruments, Nashua, NH); the addition of chlorine to the water holding tank is designed to achieve a ratio of 50 mL of chlorine:5 gallons of H₂O. The RWB were filled with chlorinated reverse-osmosis water via a bottle filling station (TBJ, Inc., Chambersburg, PA) in the animal facility. Once filled, bottles were autoclaved with sipper tubes and stoppers in place. The RWB were then brought to the animal housing room the morning of the cage change. A dedicated supply of DPP (250-mL volume; Hydropac; Avidity Science LLC, Waterford, WI) was filled on the same day before the start of the study; this approach provided an abundance of pouches for the entire study. DPP were filled from a proprietary filling machine linked to the source water for the animal facility. A sterile disposable valve was attached to each DPP, extending from the pouch through the IVC wire bar lid for drinking access by the mice.

Autoclaved reusable bottles and disposable pouches were replaced during weekly cage changes. At the time of cleaning, water bottles and stoppers were sanitized by placement in a basket strainer and processed through the tunnel washer with exposure to acid (supplied by Pharmacal Research Laboratories, Waterbury, CT) delivered at a concentration of 1/2 ounce to every gallon of water in the machine. Bottles and stoppers were then autoclaved (90-min cycle). Drinking valves on the IVC rack were maintained in place throughout the study since specific cages were kept in exact locations during the 8 wk of housing. To avoid any potential disruption to the microbiome related to disinfectant exposure, the valves were not manipulated or wiped with chemical disinfectant during the experiment.

Mice were handled in the sequential order in which they were located on the rack (within dedicated columns based on water source type: NOG RWB animals, then B6 RWB cages, NOG RAW cages, B6 RAW cages, then NOG DPP cages, and finally B6 DPP cages. Water samples (50 mL) were collected weekly from the IVC rack, pouches, and water bottles using single-use sterile syringes. Sampling from DPP and RWB was performed within the room biosafety cabinet. Water from pouches was collected by inserting a hypodermic needle through a silicone patch affixed to the pouch surface that was sealed upon removal of the needle once samples were collected. RWB, with stopper removed, were sampled from the top with a sterile syringe and needle without contacting the bottle itself. Rack ports at the bottom of the rack manifold were opened and flushed for approximately 30 s before collection into two conical tubes. Samples included one 50-mL sample from each of the 8 RWBs, one 50-mL sample from each of the 8 DPPs, and two 50-mL samples from the rack port. One 50-mL sample from a RWB and one 50-mL sample from a DPP that were unused by animals were also collected for comparison. In addition, two 50-mL samples of central source water were collected from the storage tank located in the interstitial space of the facility at 3 time points throughout the experiment, day 0 (start), day 27 (midpoint), and day 55 (end) to ensure that the source water was within specifications. Water samples were sent weekly via local courier to VRL for bacterial counts and genus/strain identification. Once the water samples were received, initial cultures were performed by sample placement on nutrient agar; any subcultures were performed using blood agar. The samples were incubated at 37 °C for 48 h, per VRL procedures.

Fecal pellet collection.

Fresh fecal samples were collected (n = 2 pellets per mouse) from each animal for microbiome analysis beginning after one week of acclimation (day 7; baseline) and again at dedicated time points (days 14, 28, and 56) during the study. For collection, mice were placed singly into empty, autoclaved cages. Two excreted pellets were collected immediately with a sterile toothpick and submerged into a sample collection tube containing DNA stabilization buffer to ensure reproducibility, stability, and traceability (Transnetyx). Sample collection tubes were barcoded and labeled for each animal.

Fecal microbiome metadata for each mouse was documented according to mouse ID, sex, water source, genetic strain, and cage number. Reproducible DNA extraction was automated by Transnetyx for inhibitor-free, high-molecular-weight genomic DNA that captures the full microbial diversity of fecal samples. The extracted genomic DNA was converted into sequencing libraries using a method optimized for minimal bias. In brief, per Transnetyx (Cordova, TN) processes, the Qiagen DNeasy 96 PowerSoil Pro kit was used for extraction, the KAPA HyperPlus kit was used for library preparation, and the Illumina 2 × 150-bp sequencing kits were used on a NextSeq 2000. Unique dual-indexed adapters were used to ensure that reads and/or organisms were assigned correctly. After quality control review, the libraries were sequenced using shotgun sequencing (a depth of 2 million 2 × 150-bp read pairs), which enables species and strain level taxonomic resolution.

Sequencing data were uploaded automatically into analysis software and analyzed against the software database consisting of >127 K whole microbial reference genomes (One Codex; One Codex Inc., San Francisco, CA). Classification results were filtered through statistical postprocessing steps designed to eliminate false positive results caused by contamination or sequencing artifacts.

Statistical analyses of fecal samples.

The One Codex microbiome analysis platform was used to organize microbiome clusters and extract the fecal microbiome relative abundance data, similar to previously published analyses.¹⁴ Microbiome analyses were evaluated by mouse strain, water source, and sex, and grouped by time point. Each cage of mice served as a unit to compare cage at day 7 compared with days 14, 28, and 56; however, mice from the original cage 8B were analyzed together as a cohort with cage 8B-2. In addition, cages were compared over time for both strains, with males assessed independently from females.

For Shannon α diversity, with multiple time points in the comparison, a Kruskal-Wallis test was applied to determine if there are any significant differences (a significance threshold of ≤0.05 was used for all comparisons) in familial taxa (of bacteria) between cage cohorts. If significance was found, a post hoc pairwise comparison between the various time points, using a Dunn test, was completed (see Tables 1 and 2). For β-diversity, comparing the different time points for each combination of strain/water source/sex) and a post hoc permutational multivariate ANOVA test was performed between each pair of time points, followed by a Benjamini-Hochberg false discovery control (that is adjusting for multiple testing).

Table 1.

Shannon index of α diversity across strain NOG cage cohorts (post hoc).

NOG female rack autowater (RAW)-cage 1A
Day	7	14	28	56
7	1	0.935	0.585	0.02a
14		1	0.585	0.02a
28			1	0.003a
56				1
NOG female rack autowater (RAW)-cage 2A
Day	7	14	28	56
7	1	0.653	0.078	0.001a
14		1	0.128	0.003a
28			1	0.128
56				1
NOG male rack autowater (RAW)-cage 3A
Day	7	14	28	56
7	1	0.011a	0.178	0.0001a
14		1	0.178	0.178
28			1	0.011a
56				1
NOG male rack autowater (RAW)-cage 4A
Day	7	14	28	56
7	1	0.935	0.006a	0.017a
14		1	0.006a	0.017a
28			1	0.680
56				1
NOG female reuseable bottle (RWB)-cage 2B
Day	7	14	28	56
7	1	0.025a	0.008a	0.623
14		1	0.623	0.008a
28			1	0.003a
56				1
NOG male reuseable bottle (RWB)-cage 4B
Day	7	14	28	56
7	1	0.357	0.0005a	0.074
14		1	0.007a	0.292
28			1	0.094
56				1
NOG female disposable pouch (DPP)-cage 1H
Day	7	14	28	56
7	1	0.470	0.008a	0.008a
14		1	0.046a	0.044a
28			1	0.902
56				1
NOG female disposable pouch (DPP)-cage 2H
Day	7	14	28	56
7	1	0.103	0.425	0.024a
14		1	0.024a	0.0001a
28			1	0.103
56				1
NOG male disposable pouch (DPP)-cage 3H
Day	7	14	28	56
7	1	0.653	0.046a	0.012a
14		1	0.018a	0.006a
28			1	0.585
56				1
NOG male disposable pouch (DPP)-cage 4H
Day	7	14	28	56
7	1	0.482	0.008a	0.014a
14		1	0.046a	0.009a
28			1	0.500
56				1

^a P < 0.05, threshold of significance.

Table 2.

Shannon index of α diversity across strain B6 cage cohorts (post hoc).

B6 female rack autowater (RAW)-cage 5A
Day	7	14	28	56
7	1	0.013a	0.061	0.061
14		1	0.526	0.526
28			1	0.903
56				1
B6 female rack autowater (RAW)-cage 6A
Day	7	14	28	56
7	1	0.0013a	0.061	0.032a
14		1	0.188	0.302
28			1	0.698
56				1
B6 male rack autowater (RAW)-cage 7A
Day	7	14	28	56
7	1	0.110	0.110	0.003a
14		1	0.967	0.157
28			1	0.157
56				1
B6 male rack autowater (RAW)-cage 8A
Day	7	14	28	56
7	1	0.668	0.002a	0.095a
14		1	0.004a	0.145
28			1	0.145
56				1
B6 female reuseable bottle (RWB)-cage 5B
Day	7	14	28	56
7	1	0.870	0.005a	0.309
14		1	0.005a	0.324
28			1	0.082
56				1
B6 female reuseable bottle (RWB)-cage 6B
Day	7	14	28	56
7	1	0.286	0.144	0.303
14		1	0.010a	0.870
28			1	0.010a
56				1
B6 male reuseable bottle (RWB)-cage 7B
Day	7	14	28	56
7	1	0.003a	0.01a	0.106
14		1	0.845	0.106
28			1	0.091
56				1
B6 male reuseable bottle (RWB)-cage 8B
Day	7	14	28	56
7	1	0.770	0.393	0.022a
14		1	0.330	0.049a
28			1	0.001a
56				1
B6 male disposable pouch (DPP)-cage 7H
Day	7	14	28	56
7	1	0.432	0.005a	0.744
14		1	0.044a	0.555
28			1	0.008a
56				1
B6 male disposable pouch (DPP)-cage 8H
Day	7	14	28	56
7	1	0.838	0.010a	0.313
14		1	0.010a	0.276
28			1	0.144
56				1

^a P < 0.05, threshold of significance.

Animal disposition.

During the experiments, one animal (ID: C, cage 7B; B6NTac male on RWB) was found deceased 6 days after the first fecal collection with no apparent signs of illness or injury. Twelve (6 per strain, 2 per water source, one male [M]:one female [F]) of the remaining mice (n = 71) were randomly selected and euthanized at the end of the study by CO₂ inhalation for necropsy and histologic assessment of tissues including brain, spleen, liver, kidney, gastrointestinal tract, reproductive tract, and lungs. The remainder of the study mice were transferred to approved institutional research and training protocols at the conclusion of experiments.

Results

Clinical health.

One cage with 2 mice injured from fighting was found during the first cage change/fecal collection (cage 8B; B6NTac males on RWB). The animals presented to veterinary staff with scabbed lesions but were otherwise bright, alert, and responsive with no changes to their mobility or activity. No topical treatment was provided, and the affected mice were removed from the aggressor for monitoring, creating an additional cage (8B-2) for the duration of the study. The 2 mice with fight wounds healed well with no further clinical issues. Water and fecal samples were collected from both cages for the duration of the study after separation for fighting. The cagemates of the mouse found deceased within the first experimental week remained healthy, and this cage remained coded as cage 7B; the 2 remaining male mice had fecal samples collected at days 14, 28, and 56.

Water analysis.

A total of 166 water samples were analyzed for bacterial content, discounting samples damaged or lost in transit. Throughout the study (Table 3), we identified positive water samples (total n = 9), which included Brevundimonas diminuta (n = 1), Pseudomonas stutzeri (n = 1), Pseudomonas fluorescens (n = 2), Bacillus spp. (n = 2), Sphingomonas paucimobilis (n = 2), and Escherichia coli (n = 1).

Table 3.

Overview of bacterial results (number of positives/total sample number) from water sources.

Study date	Rack autowater (RAW)	Reusable water bottle (RWB)	Disposable plastic pouch (DPP)	Originating tank
Day 0a	0/2	0/2	0/2	0/2
Day 6	1/2	3/9	0/9
Day 13b	0/2	1/10	0/9
Day 20	Samples lost in shipping
Day 27	0/3c	1/10	0/9	0/2
Day 34	0/4	0/10	0/9
Day 41	0/4	1/10	0/9
Day 48	0/4	2/10	0/8c
Day 55	0/4	0/10	0/9	0/2

^a Day 0 of study: before animal arrival and subsequent placement onto IVC rack.

^b Day 13: adjusted rack collection to allow water to run for 30 s before collection; unused water bottle was (+); an additional RWB bottle created from split of fighting B6NTac male mice cage (8B); decision to double samples from IVC rack port for remainder of study.

^c Days 27 and 48: loss of water samples that opened during shipment.

Nearly all positive samples (8 of 9; 88.9%) were collected from RWBs; one was collected from an unused RWB (P. stutzeri), and 7 were collected from used RWBs (3 from cages of B6NTac and 4 from cages of NOG). From the unused RWB sample, the culture of >100 P. stutzeri was attributed to contamination postautoclaving and was a singular occurrence. This particular source of bacteria was also never detected in any other sample throughout the rest of the experiment. Of the other 7/9 positive bacterial samples from RWB, 3 samples were found from cages carrying B6NTac mice (2 F cages; one M cage) and 4 samples were found from cages carrying NOG mice (2 F cages; 2 M cages). Interestingly, the 3 positive RWB bottles collected on the same date (day 6) contained 3 different types of bacteria, including E. coli (52 colonies/plate from cage 1B [NOG F]), S. paucimobilis (31 colonies/plate from cage 2B [NOG F]), and Bacillus (12 colonies/plate from cage 5B [B6 F]). While S. paucimobilis, Bacillus, and P. fluorescens bacteria all appeared on 2 dates of water sample collections, no single type of bacteria ever appeared twice from the same animal cage or source.

The remaining positive water sample was collected from the rack manifold port (RAW at the first collection time point [P. fluorescens]) and all subsequent collections were adjusted to allow for approximately 30 s of flushing of rack lines before sample capture. For the samples collected from DPPs and those obtained directly from the facility water source, no samples were found to contain bacteria at any time point.

The 6 types of bacteria that we identified likely can be attributed to different factors. S. paucimobilis appeared twice throughout the experiment: once on day 6 in the RWB from cage 2B (NOG F) with 31 colonies/plate and then again on day 27 in the RWB from cage 5B (B6 F) with colonies too numerous to count. This was the largest count of bacteria found in animal-used bottles. Bacillus also appeared twice throughout the experiment: once on Day 6 in the RWB from cage 5B (B6 F) with 12 colonies/plate, and then again on day 34 in the RWB from cage 3B (NOG M) with 37 colonies/plate. P. fluorescens was the last bacteria to appear twice: once on day 6 in the RAW rack source with 17 colonies/plate, and again on day 41 in the RWB from cage 3B (NOG M) with 5 colonies/plate. E. coli was found once on Day 6 in the RWB from cage 1B (NOG F) with 52 colonies/plate. B. diminuta was also only found once in the RWB from cage 8B (B6 M) with 21 colonies/plate.

Fecal analysis.

Microbiome analysis was completed for a total of 285 fecal samples, identifying the top 10 taxa (relative abundance by family, see Figure 1) within the sample. For Figures 2–4, each column represents a single mouse (for example, a, b, c) and the adjacent 3-column cluster represents the cage cohort. Across these figures, all images are arranged similarly: the 12-column cluster represents animals in the cage at 4 sampling times (days 7, 14, 28, and 56). Each figure represents a different water source; however, in each arrangement, cages 1 and 2 are NOG females, cages 3 and 4 are NOG males, cages 5 and 6 are B6NTac females, and cages 7 and 8 are B6NTac males.

Figure 1.

Color-coded legend of taxa (families) that were most commonly observed in microbiome analyses. The number in parentheses represents the National Center for Biotechnology Information taxonomic identifier (https://www.ncbi.nlm.nih.gov/) in case the strain nomenclature changes at a future date.

Figure 2.

Relative abundance levels for animals receiving rack autowater (RAW) during the study. Each cluster represents a cage of 3 animals that were profiled on days (d)7, d14, d28, and d56. All cages had significant shifts in α-diversity across time points throughout the study; the red dot is placed below the only cage (6A) in the study that did not have significant differences in β-diversity. NOG females are in cages 1 and 2, NOG males are in cages 3 and 4, B6NTac females are in cages 5 and 6, and B6NTac males are in cages 7 and 8.

Figure 4.

Relative abundance levels for animals receiving water from reusable water bottles (RWB) during the study. Each cluster represents a cage of 3 animals that were profiled at d7, d14, d28, and d56. Black dots are placed below the 2 cages of NOG animals (1B: females; 3B: males) that did not have significant differences in α diversity across time points. All other cages had significant shifts in microbiome throughout the study. NOG females are in cages 1 and 2, NOG males are in cages 3 and 4, B6NTac females are in cages 5 and 6, and B6NTac males are in cages 7 and 8 and 8B-2.

In Figure 2, representing rack autowater as the source, the male and female NOGs display a microbiome containing Tannerellaceae, Lactobacillaceae, and Lachnospiraceae at day 7. However, all NOG cages exhibit a shift in taxa by day 56: specifically, male mice show an increase in the presence of Muribaculaceae, while Prevotellaceae and Rikenellaceae appear to overtake Tannerellaceae in the females (Table 1). The B6NTac animals on RAW source have microbiomes generally consistent and dominated by Muribaculaceae, Prevotellaceae, and Rikenellaceae. One of the 2 B6NTac cages of females appeared to have only a minimal shift in relative abundance (cage 6A) across experimental sampling dates; this cage had significant α diversity across time points but did not exhibit β diversity that was significant (Table 2; Figure 2).

Mice of the NOG strain receiving water from plastic pouches (DPP) displayed a microbiome dominated by Tannerellaceae and Lachnospiraceae at day 7 that shifted to a microbiome comprised of Muribaculaceae, Prevotellaceae, and Rikenellaceae by da7 56 for 3 of the cages (2H, 3H, 4H; Figure 3; Table 1). The microbiome profile of B6NTac DPP females was the most stable over time (5H and 6H), while the profiles for B6NTac DPP males exhibited significant variability. The 2 male B6NTac DPP cages first showed a microbiome dominated by Oscillospiraceae and Lachnospiraceae but then shifted to a significantly different microbiome profile dominated by Muribaculaceae by day 28 that then returned to baseline taxa, not statistically different from day 7, by day 56 (Table 2).

Figure 3.

Relative abundance levels for animals receiving water from disposable plastic pouches (DPP) during the study. Each cluster represents a cage of 3 animals that were profiled at d7, d14, d28 and d56. Black dots are placed below the 2 cages of B6 Females (5H and 6H) that did not have significant differences in α diversity across time points. All other cages had significant shifts in microbiome throughout the study. NOG females are in cages 1 and 2, NOG males are in cages 3 and 4, B6NTac females are in cages 5 and 6, and B6NTac males are in cages 7 and 8.

The microbiome profiles for NOG mice receiving reusable bottles (RWB) remained fairly stable across all time points, and 2 of these cages did not have any significant shifts between sampling dates (1B and 3B; Figure 4; Table 1); gut microbiota for NOG mice was dominated by Tannerellaceae, Oscillospiraceae, Lactobacillaceae, and Lachnospiraceae. Microbiome profiles in B6NTac RWB (7B and 8B + 8B-2) mice were primarily comprised of Muribaculaceae and Rikenellaceae at days 7 and 56, with a midpoint shift to a profile containing a mixture of Tannerellaceae, Oscillospiraceae, Lactobacillaceae, and Lachnospiraceae at day 28 (Table 2).

Regarding the overall assessment of β diversity, there were significant differences (P = 0.001) between time points for all comparisons except B6NTac females receiving RAW (Figure 2); this group (P = 0.11) was therefore excluded from the post hoc analysis that showed significance for all other cages (data not shown).

Necropsy and histopathology.

No clinical or health differences were identified throughout the study between the NOG (immunocompromised) and B6NTac (immunocompetent) strains of mice. There was no gross or histologic evidence of degeneration, systemic inflammation, or finding of infectious agents within any of the tissues examined in either strain. In addition, despite the detection of P. fluorescens from the unflushed RAW port early in the study, there was no evidence of ulcers, edema to the head, or disease of the gastrointestinal tract in the immunocompromised mice examined.²⁰

Discussion

In this project, the goal was to mitigate variation and enhance stability in mouse gut microbiome over time by controlling for housing location (dedicated rack/room), caging type (autoclaved IVCs), feed source (irradiated chow), bedding material (autoclaved corncob), animal personnel, and operational practices. Taking this one step further, the working hypothesis was that providing drinking water through dedicated delivery types would augment the stability of the gut microbiomes of mice, regardless of immune status and sex, over the experimental time course. This study is the first to assess water delivery methods commonly used in rodent facilities, specifically providing the same sourced/treated water by either IVC rack piping (autowater), autoclaved reusable water bottles changed biweekly, or via prefilled disposable pouches.

As a measure of individual microbiomes and cage effects of cohoused animals, fecal samples were collected and submitted for analysis on day 7 (1 wk after arrival as a baseline), on day 14 (approximately 2 wk after arrival to allow for acclimation to institutional housing conditions), midway through the study at day 28, and at the study endpoint on day 56. The fecal sample analyses permitted a specific view of the unique microbial composition (at the bacterial family level) of each study animal. Alpha diversity reflects the species richness of a sample, such that a higher α diversity is usually indicative of a richer microbiome, whereas low α diversity is often associated with dysbiosis. Regarding α-diversity, the majority of cages (21 of 25) demonstrated significant shifts between day 7 and at least one of the other sampling dates within that cohort of cagemates. Apart from the female B6NTac animals provided with disposable plastic pouches (cages 5H and 6H), only 2 of the NOG cages provided with water bottles (cages 1B [females] and 3B [males]) had minimal shifts in relative abundance that did not statistically differ over the 8-wk study. Beta diversity looks across microbial communities to assess the similarities or differences between samples; β-diversity was significant across all cages except one housing females on autowater (6A).

Overall, for this study, α diversity and β diversity did not exhibit specific clustering by water delivery method, but primarily by time. No definitive water source appeared to provide consistency to the mouse microbiome over time across strains and sexes. It is unusual to observe microbial families disappearing entirely without any known external changes, only to be replaced by new families for several weeks, and then see the original microbial families reappear later. Such a dramatic change posits the fact that there must be changes occurring in the animal facility with a significant impact on the microbiome despite extensive efforts to eliminate experimental variability. The changes seen in microbiome composition over time again highlights the importance of testing the gut microbiome throughout the study (and perhaps longer than 8 wk after arrival to the facility), as phenotypes are sensitive to microbial composition and inherent alterations to model phenotypes may challenge study replicability and reproducibility.^10,21

As a reminder, the B6NTac females were sourced from a vendor location (site A) distinct from the NOG mice and B6NTac males. Across the water sources, these B6NTac females (in cages 5A, 6A, 5H, 6H, 5B, and 6B) had comparable relative abundance levels of Muribaculaceae, Rikenellaceae, Prevotellaceae, Lactobacillaceae, and Lachnospiraceae at baseline assessments sampled on day 7 after arrival to our facility. For B6NTac females kept on disposable pouches of water, these animals had the most stable gut microbiota over the entire study, compared with any other cages, regardless of sex or sampling date. As mentioned previously, B6NTac females on autowater from the rack (6A) were the only animals that did not exhibit β diversity over sampling time points.

Throughout the study and across water sources, only 9 of 166 samples (5.4%) were found to contain bacteria. The majority of these were pulled from reusable water bottles. Knowing that oral flora from mice can contaminate water bottles, these findings were not entirely unexpected. Of note, rubber stoppers and sipper tubes at our institution are washed as a unit, placed in metal baskets, and processed by tunnel washer, before autoclaving. There is potential that sipper tubes may retain undetected biofilm after washing and autoclaving, which could have contributed to the positive samples collected from RWB; however, assessment of potential postprocessing residue within sipper tubes was not conducted as part of these studies.

As mentioned previously, although certain bacterial species that we detected in bottle samples have been linked to diseases, no animals from these cages displayed clinical signs of ill health that could be attributed to systemic bacterial infection.²² NOG animals demonstrated the most stable microbiome composition across the study when receiving RWB sources, despite the bottles having the majority of the positive bacterial samples. The DPP cages consistently remained free of bacterial contamination, and these pouches were all filled and remained present for use within the housing room from the start of the project. Their sterility also indicates that the design of the pouch drinking valve does not permit the refluxing of the mouse oral flora into the pouch, unlike the stopper and sipper tube mechanism on water bottles. The autowater source from rack valves was the only water provision for which all cages, across strains and sexes, had some significant shift in the microbiome. While the central water supply is treated biweekly with chlorine per facility standard operating procedures, this treatment practice may have been linked to unrecognized variables introduced to the water. Overall, the vast majority (approximately 95%) of the experimental water samples remained sterile, indicating that the water source and delivery do not appear to have profoundly impacted the stability of the microbiome; instead, variations likely could be attributed to myriad factors, including those observed and some likely yet to be determined.

Despite concerted efforts made throughout this study to control for husbandry factors and limit personnel influences, several circumstances arose during the study that could have impacted the microbiome profiles. For example, while the study mice were all procured from the same vendor on the same date, it was not until the results of day 14 fecal assessments that the authors intuited that the animals had a differing microbiome from each other. After contacting the vendor, it was determined that the B6NTac males and females did not come from the same vendor site; further, the NOG mice were from the same facility as the male B6NTac mice but were in a different housing room. Because vendor facilities have different methods for water, feed, and cage treatment, and inevitably expose the animals to different personnel, this likely impacted the divergent microbiomes across strains and sexes noted early in the study, as has been shown previously.^19,23,24 Importantly, the goal of this study was not to ensure all animals arrived or acclimated to the same baseline microbiome but rather to identify which water delivery method supports the most stable microbiome, considering the diversity of the initial compositions.

Retrospectively, additional extenuating circumstances inherent to animal husbandry and care operations may have impacted microbiome profiles during our study period, as has been described previously.¹⁴ Our animals received irradiated feed from the same bag until it was replaced with a fresh bag midstudy. While the lot and processing date should have been identical to the first bag, there did appear to be a notable change in the microbiome across the study mice at day 28. Since the animals did not receive any medical interventions nor exposure to disinfectants from the rack valves, it is presumed that the substantial change in the microbiome detected on day 28 was likely linked to the replacement bag of rodent chow. Additional challenges during the 2 months of the study included a latex spill into the Delaware River that might have influenced city water parameters and, within the housing facility floor, a cage washer breakdown that necessitated the caging equipment (water bottles and cage components) be shipped to another campus facility for processing, which introduced a variation to the attempts to control for husbandry practices. Finally, within the final days of the experiment, a building-wide fire alarm test was enacted with flashing lights and audible warnings distributed throughout the housing areas, which may have potentially been stressful for the animals and impacted the microbiome.

While many extraneous variables were meticulously accounted for in the design of the experiment, it was impossible to completely control all variables involved in the project. For a microbiome study specifically, any unusual factor can lead to a potential stress response in animals and a change in their microbiome as a result. In the future, it could be valuable to ensure that all participating animals have comparable baseline microbiomes from the animal vendor source at the study starting point, perhaps by only receiving mice that had all been raised in the same room and building at the vendor site. Another more stringent measure of microbiome control would be to inoculate animals with a fecal slurry of known microbial content after animals arrive at the destination facility to normalize the microbiome as a baseline.²⁵

Despite the variability in mouse microbiome profiles noted across the cages and water delivery methods, animals from both immunocompetent and immunocompromised strains remained healthy and did not exhibit clinical signs indicative of negative impacts of the gut microbiota changes. Overall, the data suggest that water delivery method does not substantially contribute to gut microbiota modulation and does not necessarily offer greater stability to one cage cohort over another. This study serves as a reminder that the microbiome is just one of many variables that are particularly challenging to control for, yet even with noted deviations, it does not mean that the overall experiment has been unsuccessful or uninformative. As other studies have emphasized, our results confirm that details about water provision over the lifetime of the animals (for example, from vendor to facility housing to experimental endpoints) should be documented as another extrinsic factor of importance. The goal, in line with the ARRIVE guidelines, is that disclosure of water source and delivery methods will help to promote transparency of methodologies so that experimental data can be generalized and contextually interpreted alongside similar scientific studies.²⁶

Conflict of Interest

FCH has served as a member of the Transnetyx Scientific Publication Committee since 2019, and CAB is employed as a consultant to Transnetyx. Pouches and pouch filling station were upgraded from earlier model maintained at the institution at no cost by Avidity Science.

Funding

The study was funded in part by Avidity Science (mouse costs, per diems, water collection supplies, and water analysis) and Transnetyx (fecal collection supplies, fecal sample analysis, and statistical assessments).