Paired versus Single Housing of Ferrets (Mustela putorius furo) Does Not Influence Experimental Outcomes of Influenza A Virus Pathogenesis Studies

Abstract

Social housing of animals is well recognized as a fundamental element in providing for the well-being of social animals during laboratory research activities but is contraindicated in some experimental designs. To ascertain if social housing represented a potential confounder in studies determining influenza A virus (IAV) pathogenicity conducted in the ferret model, we analyzed data from ferrets inoculated with 56 distinct IAV strains, where animals were housed either singly or in pairs postinoculation. Parameters examined included frequency of lethal outcomes, timing and magnitude of clinical signs (weight loss and fever), and magnitude and kinetics of viral shedding in the upper respiratory tract. Statistical differences between IAV-inoculated ferrets housed in either setting were not consistently detected among any parameters examined, supporting that data from ferrets inoculated with a diverse range of avian- and mammalian-origin IAV housed singly or in social pairs may be combined and analyzed without confounding based on this variable. This study supports the utility of performing retrospective analyses to ensure results obtained from in vivo experimentation are interpreted to the highest standard possible. Further, it aligns with ethical obligations of researchers who perform in vivo experimentation to ensure these studies are designed to meet both animal welfare considerations and research goals.

Introduction

The housing environment of laboratory animals can greatly influence physiologic and behavioral parameters in numerous species.¹ Housing social animals in stable pairs or groups of compatible individuals is strongly recommended by The Guide for the Care and Use of Laboratory Animals² and is a well-established form of enrichment shown to elicit many positive outcomes and improvements in general welfare in rodents (eg, mice and rats),^3–5 rabbits,^6,7 nonhuman primates,^8–10 and other species.^11,12 Social housing promotes the expression of a wide array of species-specific normative behaviors and provides a means of buffering against potential stressors in the captive research environment such as exposure to novel environments and personnel, restraint, and common husbandry practices. Companionship can help reduce the perception of and reaction to aversive stimuli and blunt the physiologic response to stressors.^13–21 Conversely, individual housing of social species, defined as one animal housed alone in a cage with auditory and/or visual contact with conspecifics (not to be confused with isolation housing), can have many deleterious effects on behavior and physiology,^22,23 such as increased frequency and types of abnormal and self-directed behavior, increased fear response, decreased locomotion, increased visceral fat mass, and difficulty with thermoregulation.^7,18,24–26

Ferrets (Mustela putorius furo) are increasingly used in infectious disease laboratory settings, spanning a diversity of experimental designs and BSL environments.^27–29 In the case of influenza A viruses (IAV), ferrets represent the gold-standard small mammalian model to study viral pathogenicity, transmissibility, and tropism; data from these experiments inform critical pandemic risk assessment rubrics.³⁰ Risk assessment studies of this nature typically encapsulate experiments performed with animals housed individually (eg, airborne transmission studies) and in pairs (eg, transmission studies performed in the presence of direct contact, or within-host pathogenesis studies where all animals in the cage are similarly inoculated with virus).³¹ Laboratory-housed ferrets exhibit both solitary and social play behavior, with a preference for social play,^27,28 supporting the preference for socially housing animals in laboratory facilities. However, despite the numerous benefits of social housing and potential for undesirable behavioral and physiologic outcomes resulting from single housing, some experimental designs (notably those that include transmissible pathogens, like IAV) may necessarily preclude social housing to achieve research goals.²⁷ However, no evaluations have been performed to date assessing if single compared with pair housing of IAV-inoculated ferrets represents a confounder in quantifiable experimental outcomes.

We leveraged a publicly available online dataset of ferrets experimentally inoculated with a diverse range of IAV³² to examine if social housing influenced primary metrics of disease severity. This dataset aggregates records from previously conducted assessments of IAV pathogenicity and transmissibility in the ferret model; this study performed novel analyses on these previously collected and reported data. From records pulled from this dataset, a total of 332 ferrets were inoculated with one of 56 distinct IAV and returned individually to a cage. Twenty-four hours later, a susceptible recipient ferret was housed in one of 2 conditions: pair housed with an inoculated ferret in a shared enclosure for 14 days or housed individually in an adjacent cage separated with a perforated sidewall permitting auditory and visual communication with the inoculated ferret while maintaining physical separation. In agreement with our hypothesis that social housing would not represent a confounding variable, we found that both clinical and virologic metrics were generally indistinguishable between IAV-inoculated ferrets housed in either housing setting, supporting that comparable data could be generated from virus-inoculated animals under the conditions evaluated here.

Methods

Ferret source and housing.

Castrated and descented 5- to 12-month-old male Fitch ferrets (Triple F Farms, Gillett, PA) were used for all experiments; animals were vaccinated for rabies and distemper before shipment and were confirmed to be serologically negative to currently circulating IAV before use. Ferrets were permitted a minimum of 72 hours of acclimatization following transfer to the procedure room before any experimental manipulations took place. Before inoculation, a subcutaneous temperature transponder (IPTT-300; BMDS; Avidity Sciences, Waterford, WI) was inserted into the dorsal space between the scapula of each ferret. Ad libitum access to water and pelleted feed (Lab Diet 5L14; PMI Nutritional, Richmond, IN) was provided. Housing rooms were on a 12:12-hour light:dark cycle. Cages included manipulanda enrichment (eg, dangling stainless steel toys); nutritional enrichment was not provided.

The stainless steel ferret cages housing ferrets during viral challenge studies were 56 × 42 × 42 cm. Pair-housed ferrets (n = 172) were housed in solid-walled cages; singly housed ferrets (n = 179) were in cages with a perforated side-wall (each opening <5 mm), approximately 1 cm apart from another cage with a perforated side-wall containing another singly housed ferret. Ferrets were randomly assigned to cages before the start of the experiment. Cages were placed within a Duo-Flow BioClean mobile environmental enclosure (Lab Products, Aberdeen, MD) for all studies with virus-inoculated animals. Experiments were performed under BSL2 or BSL3 conditions, including enhancements required by the USDA and the Federal Select Agent Program.³³ All animal manipulations were conducted under the guidance of the Centers for Disease Control and Prevention IACUC in an AAALAC International-accredited animal facility.

IAV infection in ferrets.

Ferrets were anesthetized for all specimen collections, including inoculation, with a ketamine cocktail (25 mg/kg ketamine, 2 mg/kg xylazine, with or without 0.05 mg/kg xtropine) by intramuscular injection in the hamstring. All ferrets were inoculated intranasally (1-mL volume) with a high dose (10⁵-10⁷ infectious units) of IAV (typically n = 3 ferrets per housing condition, per virus, with few exceptions).³² Following inoculation, all ferrets were singly housed in a cage for 24 hours, after which an additional serologically naive ferret was cohoused in the same cage (pair-housed animals) or in an adjacent cage (singly housed animals), with respective housing environments maintained through day 14 postinoculation. Temperatures and body weights were collected daily. Animals that met experimental endpoints warranting humane euthanasia (primarily due to weight loss greater than or equal to 25% of preinoculation levels or onset of neurologic symptoms) were humanely euthanized by intracardiac administration of 1 mL/kg, 390 mg pentobarbital sodium, and 50 mg phenytoin sodium per 100 mL (Euthanasia Solution; Med-Pharmex, Pomona, CA).

Ferret nasal wash (NW) specimens were collected on alternate days postinoculation and titered for the presence of infectious virus in embryonated hens’ eggs or MDCK cells to determine a 50% egg infectious dose (EID₅₀; limit of detection 10^1.5 EID₅₀) or pfu titer (limit of detection 10¹ pfu) as previously described.^32,34

Data source.

Experiments from which these data were aggregated, and the inclusion criteria governing the studies included for aggregation, have been published previously,³² and data are available online at data.cdc.gov (https://data.cdc.gov/National-Center-for-Immunization-and-Respiratory-D/An-aggregated-dataset-of-serially-collected-influe/cr56-k9wj/about_data). Fifty-six unique IAV were evaluated for pathogenicity in ferrets in both pair-housed and singly housed ferrets, with results from the primary characterization of each virus evaluated in ferrets included for analysis. IAV testing spanned multiple hemagglutinin subtypes (H1, H2, H3, H5, H7, and H9) that were of both avian origin and mammalian origin. Investigators who performed the source experiments pooled results from pair-housed and singly housed ferrets together, and as this analysis was not conceived until all previously conducted experiments were made publicly available, these investigators were fully blinded from how these group allocations would be examined in this study.

Data analysis.

All analyses were performed in R v4.4.0. Figures were generated with ggplot2 v3.5.2,³⁵ wesanderson v0.3.7,³⁶ ggpubr v0.6.0,³⁷ survminer v0.5.0,³⁸ and patchwork v1.3.0.³⁹ AUC was calculated as previously described.⁴⁰ A Fisher exact test and a Shapiro-Wilk test for normality were conducted with base R stats. Levene’s test for homogeneity of variance was conducted with car v3.1-3.⁴¹ Paired t tests were calculated using ggpubr. P values less than 0.05 were considered statistically significant.

Results

Heterogeneity of IAV challenge strains used to inoculate ferrets.

IAV pathogenicity in mammals is a multifactorial trait that is influenced by a diversity of viral and host factors⁴²; accordingly, clinical symptoms can vary greatly in ferrets inoculated with IAV, ranging from asymptomatic to severe and fatal illness, depending on the strain used to infect. Since ferrets are used in public health-focused research to assess the pathogenicity of heterogeneous IAV strains, it is critical to include data from ferrets inoculated with diverse IAV strains when examining potential confounders that may modulate clinical signs detected during acute infection.

The 56 IAV strains in this study from which ferret data were sourced³² included 6 hemagglutinin subtypes (H1, H2, H3, H5, H7, and H9), representing the most abundant subtypes isolated from humans (H1, H2, and H3), swine (H1 and H3), and avian (H5, H7, and H9) reservoirs. Twenty-five viruses were of avian origin (isolated from an avian host or originated from an avian species during zoonotic spillover to humans), and 31 were of mammalian origin (inclusive of human origin, swine origin, or human infection with swine-origin virus). Ten of the 25 avian-origin viruses were classified as highly pathogenic avian influenza strains.⁴³ Viruses spanned a range of predicted receptor binding specificity: 24 viruses predicted to bind human-like α2-6-linked sialic acids, which are present at highest abundance in the ferret and human upper respiratory tract, 16 viruses predicted to bind avian-like α2-3-linked sialic acids that are present at highest abundance in the ferret and human lower respiratory tract, and 16 viruses predicted to bind both types of cellular receptors.^31,40 Collectively, the IAV strains used in this analysis provide a comprehensive and balanced diversity of viruses that encapsulates the full range of strains studied in ferrets in laboratory settings.

Social housing does not affect frequency or timing of lethal outcomes in IAV-inoculated ferrets.

We first assessed if lethal outcomes in virus-inoculated ferrets varied when animals were housed in either pair- or singly housed environments. In the dataset, 15 of 172 pair-housed, virus-inoculated ferrets (hemagglutinin [HA] subtype H1 [n = 1], H2 [n = 1], H5 [n = 6], and H7 [n = 7]) reached humane endpoints during the 14-day observation period (8.7%, inoculated with 7 different IAV strains), whereas 19 of 160 singly housed ferrets (HA subtype H1 [n = 5], H5 [n = 7], and H7 [n = 7]) reached humane endpoints (10.6%, inoculated with 11 different IAV strains); a comprehensive presentation of clinical and virological metrics associated with lethal outcomes in these ferrets is included in the dataset analyzed.³² Among the 12 unique IAV strains associated with lethal outcomes in any ferret, 6 IAV were associated with some proportion of lethal outcomes in both housing settings; all other strains were associated with one lethal outcome only per group in either housing setting. The difference in frequency of lethality outcomes was not statistically significant via a Fisher exact test (P = 0.59; odds ratio: 1.24 [0.597-2.73]). These data support that no meaningful difference in the frequency of lethal outcomes was observed between ferrets inoculated with the same IAV but housed in different social settings. Among ferrets with lethal outcomes, the timing of these lethal outcomes was similar (day range 3-10 postinoculation for pair-housed ferrets and 3-13 postinoculation for singly housed ferrets, with median day of lethal outcome day 7 and 8 postinoculation for pair-housed and singly housed ferrets, respectively), and no statistical difference in the survival probability was observed between ferrets housed in either setting (Figure 1). Collectively, these data indicate that the housing environment does not substantially modulate the frequency or timing of lethal events recorded in ferrets inoculated with IAV capable of causing severe and fatal disease in these animals.

Figure 1.

Survival Probability of IAV-Inoculated Ferrets with Lethal Outcomes Housed in Pairs or Singly. Kaplan-Meier curve of ferrets inoculated with IAV housed in pairs (n = 15) or singly (n = 19) that met humane endpoints between days 1 and 14 postinoculation. Solid line represents day of lethal outcome; shading represents 95% CI.

Social housing does not modulate magnitude or timing of clinical symptoms in IAV-inoculated ferrets.

IAV-inoculated ferrets frequently exhibit fever and weight loss during the acute phase of infection, though the magnitude of both varies by the strain used to infect. These clinical signs are typically transient with ferrets returning to or approaching preinoculation baseline levels by the end of a 14-day observation period, though unresolved weight loss greater than 25% of preinoculation weight is possible. Among all IAV-inoculated ferrets in the study, peak fevers were detected on days 1-2 postinoculation (median day 2 postinoculation for both housing groups), and maximum weight loss was detected on days 7-8 postinoculation (median day 7 for both housing groups), supporting that the timing of these clinical signs did not vary based on the housing environment (Figure 2).

Figure 2.

Timing of Peak Maximum Temperature Rise and Weight Loss in IAV-Inoculated Ferrets Housed in Pairs or Singly. Day of (A) peak maximum rise over preinoculation baseline temperature or (B) peak maximum weight loss between days 1 and 14 postinoculation with all tested viruses in pair-housed or singly housed ferrets. Number of individual ferrets with peak values recorded on the day shown is captured in the legend. No color indicates no peak values were recorded on that day.

While the timing of these events was generally comparable, we next assessed if the magnitude of the clinical symptoms varied between ferrets inoculated with the same virus but housed in different social settings. Mean peak temperature readings recorded between days 1 and 5 postinoculation (during which the majority of fevers were detected⁴⁴) were calculated for all ferrets inoculated with the same IAV strain but housed in different social settings; comparable peak readings were found in both groups (1.47 ± 0.71 and 1.43 ± 0.75 °C in pair-housed and singly housed groups, respectively). While mean peak temperature values could vary between ferrets inoculated with the same IAV but housed in different social settings, differences in the aggregate were not statistically significant (in either the entire set of all IAV tested, or when stratified by host origin of the IAV used to inoculate) (Figure 3A and B).

Figure 3.

Mean Maximum Temperature Rise and Weight Loss in IAV-Inoculated Ferrets Housed in Pairs or Singly. (A and B) Mean maximum rise over preinoculation baseline temperature days 1-5 postinoculation (in °C) or (C and D) mean maximum weight loss days 1-14 postinoculation (percentage from baseline weight) in pair-housed or singly housed ferrets inoculated with (A and C) all tested viruses or (B and D) separated by avian or mammalian host origin. Each dot represents the mean of n ≥ 3 ferrets inoculated with the same IAV. Lines connecting dots indicate mean values obtained for ferrets inoculated with the same virus. Boxes depict median with first and third quartile; whiskers extend from each hinge to the largest/smallest value no more than 1.5 of the IQR. P value denotes paired t test.

Similar to mean peak temperature rise, the mean maximum percent weight loss recorded between days 1 and 14 postinoculation calculated for all ferrets inoculated with the same IAV strain but housed in different social settings was comparable (9.63 ± 6.09 and 10.57% ± 6.77%, respectively). Differences in the aggregate were P = 0.039 when using the entire dataset (Figure 3C) but did not reach statistical significance when stratified by host origin (Figure 3D). The modest difference in mean maximum percent weight loss observed (less than 1% between ferrets housed in different social settings when using the entire dataset) supports that biologically meaningful differences were not consistently observed. Furthermore, we assessed variance differences between groups using the Levene test. F statistics were low, and P values exceeded 0.05, indicating no significant heterogeneity of variance in these data. Collectively, these data support that variation in mean values between groups may be detected with both temperature and weight loss metrics, but these differences are not consistently skewed when animals are housed in one housing setting compared with another.

Social housing does not modulate upper respiratory tract viral loads in IAV-inoculated ferrets.

Following high-titer intranasal inoculation, IAV typically replicate in the ferret upper respiratory tract for at least 5 days before clearance; peak titers in NW specimens can nonetheless vary greatly depending on molecular determinants present in the inoculating strain.⁴⁵ To determine if viral titer loads from IAV-inoculated ferrets varied between housing environments, we examined 2 summary measures of viral replication in the ferret upper respiratory tract. Mean peak NW viral titer days 1-5 postinoculation (Figure 4A and B), or total AUC of NW titers collected days 1-8 postinoculation (Figure 4C and D), was determined for each virus tested in both housing environments and compared in a paired t test. Similar to clinical parameters, comparable mean values were observed between pair-housed and singly housed ferrets (using either the entire dataset or when viruses used to inoculate animals were stratified by host origin), with no statistically significant differences between groups when either peak NW or AUC summary metrics were assessed. These findings indicate that comparable kinetics and magnitude of viral replication were observed during acute infection, independent of the housing environment.

Figure 4.

Peak Nasal Wash (NW) Titer and AUC in IAV-Inoculated Ferrets Housed in Pairs or Singly. (A and B) Mean peak NW viral titer days 1-5 postinoculation (p.i.) or (C and D) total AUC of NW titers collected days 1-8 postinoculation in pair-housed or singly housed ferrets inoculated with (A and C) all tested viruses or (B and D) separated by avian or mammalian host origin. Each dot represents mean of n ≥ 3 ferrets inoculated with the same IAV. Lines connecting dots indicate mean values obtained for ferrets inoculated with the same virus. Boxes depict median with first and third quartile; whiskers extend from each hinge to the largest/smallest value no more than 1.5 of the IAR. P value denotes paired t test.

Discussion

The 3Rs (Reduction, Refinement, Replacement) compel research investigators to routinely assess ways in which experimental goals can be balanced with animal welfare needs.⁴⁶ While pair housing laboratory animals is a common form of enrichment, studies assessing different modes of respiratory pathogen transmissibility represent an example of scientifically justified single housing of animals in a laboratory setting. That said, even in the event of a strong scientific justification, the potential impact of different social housing environments should not be dismissed without examination, to both assess the relative contribution of potential confounders that may be present between groups, and to facilitate compilation of data obtained from animals in different housing settings in subsequent summary metrics. This analysis represents a first-in-field investigation of a common experimental variable present in IAV risk assessment studies performed in ferrets worldwide. Our finding of minimal to absent confounding of key experimental outcomes between IAV-inoculated ferrets housed in different settings supports combining data from inoculated animals housed in different social settings when analyzing and reporting results. Furthermore, this work supports best ethical practices for research investigators to look critically at well-established procedures and protocols to ensure studies performed in vivo continuously balance welfare considerations and research needs.

We provide here an example of using previously published data to retrospectively investigate if experimental outcomes were confounded by a well-controlled variable. This study represents a template for other research groups to perform similar analyses to assess potential confounding of experimental variables (including but not limited to housing environment), without the need for additional animal experimentation. In vivo-generated data pose several challenges for use in data science applications⁴⁷ but can enable a diversity of analyses to identify trends in data not present in individual studies. As evidenced here, knowledge gleaned from analyzing aggregated data may extend far beyond the objectives of the experiments as initially designed. Increased attention to leveraging aggregated datasets to investigate not only associations between different biologic, physiologic, and behavioral parameters in the context of in vivo experimentation,^40,48–51 but also ways in which variation in provided enrichment (including but not limited to social housing) may modulate these parameters, is warranted.

Analyses in this study were restricted to frequently captured clinical parameters and measures of viral load in the ferret upper respiratory tract but did not assess the wide diversity of additional parameters that may be examined when performing standard risk assessments of IAV pathogenicity in ferrets.³¹ Beyond these parameters, there are numerous variables, including, but not limited to, ferret activity levels, frequency of other clinical symptoms such as nasal discharge or sneezing, lymphohematopoietic parameters, and extrapulmonary virus spread, that are often assessed in the context of IAV risk assessment studies.^52,53 While our analyses support that no meaningful differences were present in the clinical or virological parameters examined, other variables could plausibly be modulated by behavioral and/or physiologic alterations attributed to housing in different social settings, and may warrant further investigation. As shown in Figures 3 and 4, outbred ferrets may display high variability in clinical and virological parameters even when inoculated with the same virus. Performing similar analyses assessing potential confounding by social housing environment in inbred animal models such as mice, where per-animal variability would likely be reduced, would be of interest.

Pair/group housing of ferrets is common in laboratory settings as it represents a strong and cost-effective opportunity to provide species-specific physical activity and cognitive stimulation needs of this species.^27,54 Ferrets are social animals (both during play behavior when awake and sleeping in groups),^55,56 highlighting the critical role of social housing in laboratory settings to reflect species-specific behavioral needs. Despite these benefits, there are experimental research designs that preclude social housing. Infectious disease studies evaluating the transmission potential of a contagious pathogen represent one of the most frequently encountered experimental designs in laboratory settings, which may necessitate separating animals in individual cages. In these studies, to assess the capacity of a virus to transmit between mammals by infectious aerosols, animals must be physically separated and housed in separate, adjacent cages that prohibit all direct and/or indirect contact (eg, contact with shared food/water sources) between inoculated donor animals and susceptible recipient animals.³¹ Other study designs where single housing may be scientifically justified include studies designed to evaluate infectious pathogens in animals inoculated with different viral strains and/or receiving different treatments, along with studies assessing within-host perturbations (such as disturbances to the microbiome).⁵⁷

While the housing environments of inoculated ferrets were well controlled in this analysis, there are nonetheless limitations inherent in the study design. The IAV used in this study exhibits diverse transmissibility profiles in either pair-housing or single-housing environments.⁵⁰ As such, during the 14-day observation period, inoculated ferrets were pair housed with or housed singly but adjacent to naive ferrets. Consequently, naive ferrets could have become productively infected with IAV and exhibited variable clinical symptoms, potentially influencing the behavior of both animals within the housing environment. Singly housed ferrets had no direct or indirect contact with other conspecifics; however, the perforated side-wall of their cage permitted auditory and visual communication with a ferret in an adjacent cage while remaining fully physically separated. With few exceptions, both singly housed and pair-housed ferrets inoculated with the same IAV were housed within the same larger environmental enclosure (which holds up to 9 individual ferret cages), such that auditory communication was also possible between all animals within an experiment independent of housing occupancy; isolation housing (eg, a singly housed ferret with no auditory or visual communication to other animals) did not take place. Caging and housing environments of IAV-inoculated ferrets used in the experimental designs from which data were aggregated can vary substantially in the field,⁵¹ such that the overall breadth and quality of environmental enrichment provided to ferrets in both single housing and pair housing scenarios may not be directly translatable to other institutions performing similar work.

Aggregation of previously conducted experimental in vivo data is beneficial not only for improving animal welfare by reducing animal numbers. It also supports a variety of scientific purposes, including overcoming statistical limitations of small sample size from individually performed experiments, informing mathematical and computational models, and discovering data trends that are not easily identifiable in discrete studies, among other considerations.⁵⁸ Here, we leveraged preexisting data to provide supporting evidence that standard clinical and virological parameters obtained in IAV-inoculated ferrets are comparable independent of the housing option used (paired compared with single housed) during the experimental period. These analyses were possible without any new animal manipulation and support that social housing does not represent an experimental variable contraindicating combining results when conducting standard pathogenicity assessments of IAV infection in this model; both of these achievements contribute to Reduction and Refinement aims of the 3Rs. Continued efforts of researchers to leverage past experimental studies to gain new insight toward improving the rigor of existing laboratory in vivo models are warranted to continue supporting these critical goals.

Conflict of Interest

The authors have no conflicts of interest to declare.

Funding

This work was supported by the Centers for Disease Control and Prevention as part of the normal job duties of the authors.