Isolation, Characterization, and Epizootiology of Clostridioides cuniculi from Immunodeficient Mice with Enteric Disease

Abstract

Mouse strains deficient in adaptive and innate immune functions, such as NSG, NSG-SGM3, and NBSGW, are highly susceptible to opportunistic infections. Over a period of 7 mo, 1,193 mice from the above 3 strains in an SPF barrier were observed with mild loose stool (LS). Affected mice had minimal weight loss and mortality. Histopathology revealed erosion of the jejunal villi with neutrophilic inflammation and Gram-positive bacterial rods adhering to the cecal mucosa with varying degrees of mucosal hyperplasia, epithelial vacuolation, and apoptosis. Anaerobic culture revealed a clostridial species that could not be speciated using standard biochemical phenotyping. Further, Clostridioides difficle and Clostridioides perfringens ELISA on intestinal contents were negative for toxins. We performed a challenge study by exposing naïve NSG mice to dirty bedding from affected cages; metagenomics on pre- and postchallenge feces identified and associated the etiopathogenesis to Clostridioides cuniculi. Whole genome sequencing and phylogenetic analysis confirmed the identity of C. cuniculi. The isolate was sensitive to trimethoprim-sulfamethoxazole (TMS). TMS was effective in abrogating signs of LS and clearing infection in mice in studies. A probe-based real-time PCR specific for C. cuniculi was established. This assay was used to screen environmental and fomite contamination and potential use in rack-level screening. We traced the source of the outbreak to a NBSGW breeding colony. However, in our observation, spontaneous C. cuniculi-induced disease was only seen in the presence of an irradiated diet in the breeding NBSGW strain and not in the breeding colonies of NSG or NSG-SGM3 strains. Interestingly, we observed that exposure to infected feces from NBSGW-induced LS in both NSG and NSG-SGM3 mice. This investigation provides insights into the etiopathogenesis and probable source of sporadic clostridial infections in immunodeficient mice and lays the groundwork for its prevention and surveillance in immunodeficient mouse colonies.

Introduction

Immunodeficient mouse strains, including NOD.Cg-Prkdc^scid Il2rg^tm1Wjl/SzJ (NSG) and its variants like NOD.Cg-Prkdc^scidIl2rg^tm1Wjl Tg(CMV-IL3,CSF2,KITLG)1Eav/MloySzJ (NSG-SGM3), (NOD.Cg-Kit^W-41J Prkdc^scid Il2rg^tm1Wjl/WaskJ), and NOD.Cg-Kit^W-41J Tyr ⁺ Prkdc^scid Il2rg^tm1Wjl/ThomJ (NSGW41) are highly susceptible to opportunistic bacterial infections.^1–3 This includes spontaneous enteric disease manifested as colitis or typhlocolitis accompanied by enterotoxemia in mice naturally infected with Clostridioides difficile.^4,5 Other clostridial species like Clostridioides perfringens have also been associated with enterocolitis in NSG and related mouse strains.⁶ Interestingly, C. perfringens-induced enteropathy has also been described in immunocompetent inbred C57BL/6J and outbred Swiss-derived ICR mice in the second week of nursing large litters.^7,8 High-energy needs for heavy lactation and the associated increased feed consumption are speculated to favor clostridial proliferation and toxin production in a manner reminiscent of clostridial enterotoxemia in sheep and goats.^8,9 Another well-known antecedent in C. difficile enterocolitis is gut dysbiosis after antibiotic therapy.^4,10 Precipitating factors like antibiotic therapy and intensive breeding are ubiquitous in contemporary murine colonies and opportunistic clostridial species may be part of the murine gut microbiome or the macroenvironment, presenting challenges to biosecurity and experimental endpoints.

Clostridial species have been poorly characterized or underrecognized as pathogens in rodents. However, metagenomics has identified Clostridioides species in outbreaks of diarrhea in immunodeficient mice.¹¹ In that diarrheal outbreak, Clostridium celatum and Clostridiales bacterium V202-01 along with Candidatus Arthromitus and Bifidobacterium pseudolongum PV8-2 were enriched in the feces of the affected mice.¹¹ Ultimately, 75% of the colony was affected by high morbidity and low mortality, potentially indicating a single causative agent that could efficiently transmit through a barrier facility. Of the 4 enriched organisms, C. celatum is a possible candidate as it is not part of the normal gut microbiome of laboratory mice.¹² However, C. celatum is part of the normal gut microbiome of humans and is known to cause systemic and tissue infections in surgical patients.^13,14 As such, transmission from humans to immunodeficient mice could have been the source of infection. Another possibility, given the large and heterogenous nature of the genus Clostridioides, which includes more than 200 species is misidentification or misattribution due to the incomplete reference genome databases used for metagenomic analysis.¹⁴

In this article, we describe a similar outbreak of loose stool with high morbidity and low mortality in NSG, NSG-SGM3, and NBSGW mouse strains. Feces from affected mice were consistently culture positive for Clostridioides (Enterocloster) clostridioformis as identified by automated biochemical phenotyping but with atypical features. We conducted challenge studies that included metagenomic analysis of the feces of challenged mice collected before and after exposure to feces from affected mice. This analysis revealed in the postexposure, but not the preexposure feces, the presence of C. cuniculi, a novel organism recently reported to be associated with epizootic enteropathy of rabbits, a devastating disease with high mortality rates where animals succumb to diarrhea.¹⁵ Upon further analysis of the C. cuniculi genome, we observed that C. celatum and C. cuniculi share a close phylogenetic relationship. This relationship highlights an underappreciated challenge for microbial identification, where taxonomy is heavily dependent on comprehensive reference databases. As C. cuniculi was described in 2018 followed by genome sequencing and deposition in National Center for Biotechnology Information (NCBI) in 2019, misidentification of C. cuniculi as C. celatum or another phylogenetically related Clostridioides species likely occurred in reports in mice published before the isolation and characterization of C. cuniculi and its subsequent incorporation in reference databases.

In our investigation, we traced the likely source of infection to an SPF and stringently managed breeding colony of NBSGW mice that are more immunodeficient than NSG strain due to the functional impairment of hematopoietic stem cells and decreased circulating neutrophils.¹⁶ Multiple isolates of C. cuniculi were characterized and used to develop a specific probe-based real-time PCR for screening mouse colonies and the macroenvironment. Furthermore, we successfully used antibacterial treatment within the immunodeficient barrier facility to control and eliminate C. cuniculi that has resolved the occurrence of loose stool (LS). We also present epizootiological evidence for differential predisposition of NBSGW strain over NSG (and NSG-SGM3) to C. cuniculi when fed irradiated natural ingredient diet, raising concerns for clostridial endospores to survive irradiation to germinate and colonize the gut of susceptible murine hosts.

Materials and Methods

Case presentation.

The immunodeficient mice involved in this investigation were housed in a barrier facility within the centralized vivarium of St. Jude Children’s Research Hospital (SJCRH). The SJCRH animal care and use program is accredited by AAALAC International and operates in accordance with the National Research Council’s Guide for the Care and Use of Laboratory Animals. All the mice studied were on an IACUC protocol either for breeding and sentinel housing or for implanting patient-derived xenografts and therapeutic and imaging studies. Only animals from approved vendors were admitted to the immunodeficient colonies. Founders for NSG, NSG-SGM3, and NBSGW strains were imported from The Jackson Laboratory (Bar Harbor, ME). Upon arrival, breeders tested negative by PCR for mouse adenovirus type 1 and 2; mouse hepatitis virus; mouse norovirus; mouse parvovirus; minute virus of mice; mouse rotavirus; rodent chaphamaparvovirus 1; Theiler murine encephalomyelitis virus; β-streptococcus groups B, C, and G; Bordetella pseudohinzii; Corynebacterium bovis; Chlamydia muridarum; Helicobacter genus; Klebsiella oxytoca; Klebsiella pneumoniae; Proteus mirabilis; Rodentibacter heylii; Rodentibacter pneumotropicus; Staphylococcus aureus; Streptococcus pneumoniae; Enterobacter hormaechei; Demodex species; Entamoeba species; mite; pinworm; Pneumocystis; and Tritrichomonas genus. The mice were positive for murine astrovirus 1 (MuAstV1) and Staphylococcus xylosus. Further, all the mice in this study were negative by exhaust air duct (EAD; Charles River Laboratories, Wilmington, MA) and pooled dirty bedding samples (IDEXX, Westbrook, ME) for opportunistic bacterial agents Klebsiella spp., Rodentibacter spp., Proteus spp., and Staphylococcus aureus. Mice were housed as breeding pairs or in groups of 5 postweaning in autoclaved individually ventilated cages with 1/4-inch corncob (The Andersons, Maumee, OH) or 1/8-inch pelleted cellulose (Safe Complete Care Competence, Rosenberg, Germany) as bedding. Drinking water was provided through an automatic watering system (Avidity Science, Waterford, WI) after reverse osmosis, UV light irradiation, and chlorination (4 ppm). The diets were irradiated LabDiet 5058 (standard diet; LabDiet, Richmond, IN) or as part of treatment 5TK4 (modified Lab Diet 5058 containing 0.025% trimethoprim and 0.124% of sulfamethoxazole; TMS diet). NBSGW strains were transitioned from LabDiet 5058 to autoclavable LabDiet 5K52, the diet used by the vendor. Cages were changed once every 2 wk or spot changed as needed. MB-10 (Quip Laboratories, Wilmington, DE) or Peroxigard RTU (Virox Technologies, Oakville, Ontario, Canada) was used for decontamination in the animal holding areas and for cage change activities before transitioning to Defender (TrustMedical, Lakeville, MN), a chlorine-based sporicidal disinfectant. Cages and cage parts on racks were shrouded and sterilized in a pass-through bulk autoclave into the sterile side of the barrier. All cage changes and procedures except imaging and surgical manipulations were conducted inside a biologic safety cabinet. However, autoclaved and irradiated diets were dispensed into the cages directly from their original bags stored inside feed bins placed next to the biological safety cabinets. Irradiated diet bags were surface disinfected with Peroxiguard RTU or Defender before moving into the barrier.

In 2024, between February and August, we identified 1,193 cases of LS in this facility. Of these, 1,154 were cages with at least one mouse with LS with the remaining 39 recorded as individual mouse cases. The cases were restricted to mouse strains NSG, NSG-SGM3, NSGW41 and NBSGW. Additional testing for known viruses, aerobic bacteria, and parasites was negative, and there was no history of outbreaks or breaches to biosecurity anywhere in the program. All patient-derived xenograft samples were screened by multiplex PCR (VRL Animal HealthDiagnostics, Gaithersburg, MD) with no history of positive human or murine pathogens in these samples. However, on average 8.7% of the fecal samples from the clinical cases in nonbreeding study rooms were positive by anaerobic culture for E. clostridioformis but with atypical patterns by automated biochemical phenotyping. We later identified the organism as C. cuniculi (see below). The number of cases of LS peaked from 159 reports in February to 334 in April before tapering down to 9 cases in August coinciding with the implementation of sporicidal disinfectant, antibacterial treatments, segregation of infected colonies, and elimination of cross-infection of NSG-SGM3 breeding colony from the cohoused NBSGW breeding colony, which was the source of the infection. The breeding and study mice were housed on separate floors of the same facility and managed by nonoverlapping staff. The relationship of the various breeding colonies with the study mice is depicted in Figure 1A.

Figure 1.

Separate and dedicated breeding rooms for (A) NSG and NSG-SGM3 with 8- to 10-wk old mice transferred to study rooms for almost 5 years with no history of LS or enteric pathology until (B; red asterisk depicts infected mice) the introduction of NBSGW for breeding to the room housing breeding NSG-SGM3 colony. All breeding colonies were on the same irradiated rodent diet before and during the outbreak of LS and enteric disease.

Necropsy, histopathology, and clinical pathology.

Both breeding and study mice belonging to strains NSG, NSG-SGM3, NSGW41, and NBSGW were submitted for necropsy and histopathology. For each mouse, initially, a full set of tissues was routinely fixed in 10% neutral-buffered formalin for a minimum of 2 days and a maximum of 2 wk. Bone specimens were decalcified in 10% formic acid. Later, once the enteric nature of the disease was established, only intestines were collected for histopathology. Fixed tissues were embedded in paraffin and sectioned at 4-µm thickness. Sections were mounted on slides and stained with routine hematoxylin and eosin, and a subset was stained with Gram stain.

Microbiology.

At SJCRH, within 20 to 40 min of collection feces were submitted to the microbiology laboratory in 1.5 mL Eppendorf tubes. Saline (0.45%) was dispensed into the tubes to soften the fecal material and sat for 10 to 15 min before agitating in a vortex. A 1-mL calibrated loop was used to streak the samples onto a trypticase soy agar (TSA) with 5% sheep’s blood plates. The plates were placed inside anaerobic boxes containing gas packs and anaerobic indicators. The boxes were incubated for 48 h at 37°C with 5% CO₂. After incubation, the plates were observed for large, flat, spreading colonies with irregular edges. Individual colonies were subcultured for isolation using the same culture methods as for the primary isolation. Individual colonies from the subcultured plates were picked to make 3.0 McFarland in 0.45% saline. Biochemical phenotyping was performed on these using an ANC card on a Vitek 2 Compact System (bioMérieux, Inc., Durham, NC).

At Charles River Laboratories (Wilmington, MA), C. cuniculi isolates were maintained on TSA with 5% sheep blood. Cultures were grown in an anaerobic box (Mitsubishi Gas Chemical Company) using MGC AnaeroPack (Mitsubishi Gas Chemical Company, Tokyo, Japan). Plate cultures were grown for 48 to 72 h at 37°C. Subcultures were performed in atmospheric oxygen as subcultures performed in an anaerobic chamber Whitley A35 Anaerobic Workstation (Don Whitley Scientific, Frederick, MD) did not return viable colonies. For broth cultures, a 24-h culture of 500 μL Fastidious Bacteria (FB) broth (Hardy Diagnostics, Springboro, OH) was inoculated and maintained in an anaerobic box at 37°C. The starter culture was then transferred to 9.5 mL of prereduced FB broth in an anaerobic box for 24 to 36 hours. Cultures were assessed by Gram stain to evaluate spore-formation status as a marker for culture stress. Cultures were pelleted by centrifugation and resuspended in a DNA/RNA shield for sequencing. Glycerol stocks from broth cultures were overlayed with sterile mineral oil to minimize the effects of oxygen but not all isolates remained viable after freezing.

Antibiotic sensitivity testing.

A 0.5 McFarland standard of C. cuniculi was used to streak the culture plate containing TSA with 5% sheep’s blood agar. TMS, vancomycin, and gentamicin were placed on the streaked agar plate followed by incubation for 48 h in an anaerobic box. The zones of inhibition were measured using calipers. Vancomycin was used as the positive control and gentamicin as the negative control due to the presumed innate sensitivity or resistance respectively for these antibiotics toward Clostridioides sp. This test was replicated 3 times.

C perfringens, C difficile, and toxin analysis.

Fecal samples were sent to California Animal Health and Food Safety Laboratory, University of California, Davis, for ELISA testing of C. difficile and C. perfringens toxins. C. difficle, C. perfringens, and related toxin PCR testing were conducted through Charles River Laboratories (Wilmington, MA).

Metagenomics studies and statistical analysis.

Two fecal pellets per mouse were collected before and after challenging with dirty bedding from LS mice. The pellets were collected into microbiome sample collection tubes containing nucleic acid preservative and shipped to TransnetYX (Memphis, TN) for DNA extraction, library preparation, and whole genome shotgun sequencing. Sequenced data was uploaded to One Codex analysis software as part of the service through TransnetYX. The read and abundance data in One Codex were downloaded as CSV files for statistical analysis. Alpha diversity (Shannon-diversity index) was reported as mean and standard deviation. Beta diversity was measured using the Bray-Curtis dissimilarity and the UniFrac distances and examined using the principal coordinate analysis (PCoA). A heatmap was created to visualize the abundance pattern of core species, followed by a hierarchical clustering analysis based on the Jensen-Shannon divergence. The core species were defined as bacterial species with a mean relative abundance greater than 1%. Linear discriminant analysis effect size (LEfSE) was performed to identify core species that discriminated between groups. All statistical analyses were performed using R version 4.4.0.

Whole genome characterization and phylogenetic analysis.

C. cuniculi cultures were pelleted by centrifugation and resuspended in DNA/RNA shield (Zymo Research, Irvine, CA) before sequencing submission. Bacterial Genome Sequencing was performed by Plasmidsaurus using Oxford Nanopore Technology with custom analysis and annotation. Plasmidsaurus provided extraction and amplification-free long-read sequencing library using v14 library prep chemistry followed by sequencing on an R10.4.1 flow cell (Oxford Nanopore, Oxford, United Kingdom). Raw data were processed using a custom analysis involving removing the bottom 5% worst reads via Filtlong v0.2.1 using default parameters. Reads were downsampled to 250-Mb via Filtlong to create a rough sketch of the assembly with Miniasm v0.3. Based on the Miniasm assembly, reads were redownsampled to approximately 100× coverage (or unchanged if less than 100× coverage) with heavy weight applied to remove low-quality reads to support plasmid read retention. Next, reads were run using Flye v2.9.1 to generate an assembly with parameters selected for high-quality ONT reads. The Flye assembly was polished via Medaka v1.8.0 using the reads generated during the Miniasm downsampling. Several follow-up analyses were then used, including annotation (Bakta v1.6.1), contig analysis (Bandage v0.8.1), genome completeness and contamination (CheckM v1.2.2), and species/plasmid identification (Mash v2.3 against RefSeq genomes+plasmids and Sourmash v4.6.1 against GenBank).

Additional bioinformatic analysis.

Average nucleotide identity was determined using pyANI version 0.2.12 using the aniM under default parameters.¹⁷ NCBI deposited genomes for C. cuniculi (taxid: 2548455), C. celatum (taxid: 36834), C. saudiense (taxid: 1414720), and C. perfringens (taxid: 1502) were downloaded for inclusion (access date January 2025). Additional genomes for C. cuniculi and C. difficile sequenced from mouse isolates were included.

Genomes were further analyzed using the KBase predictive biology platform.¹⁸ Each isolate’s genome was uploaded to KBASE and annotated using RASTtk version 1.073 using default parameters. These genomes were added to a genome set containing additional reference genomes for C. cuniculi, C. saudiense, C. perfringens, C. celatum, and C. difficile as included in the pyANI analysis above. Relatedness of these genomes and an additional 20 closest genetic neighbors was determined using the KBase module Insert Set of Genomes Into SpeciesTree, version 2.2.0, which analyzes 49 clusters of orthologous group domains to generate a multiple sequence alignment followed by concatenation of the multiple sequence alignment and generation of a phylogenetic tree. The module uses FastTree2 with the -fastest setting.

C cuniculi PCR assay and targeted sequencing.

A proprietary real-time fluorogenic 5′-nuclease PCR assay specifically targeting an approximately 100-bp region within the gene encoding a AAA family ATPase specific for C cuniculi with homology across isolates sequenced for this study and the sequence data available in GenBank. The assay was used to determine the presence of genomic DNA in samples based on published methodology.¹⁹ Samples that amplified during initial testing were subsequently retested using DNA isolated from a retained lysate sample to confirm the original finding. A positive result was reported when the retested sample was confirmed positive. To monitor for successful DNA recovery after extraction and to assess whether PCR inhibitors were present, an exogenous nucleic acid recovery control assay was added to each sample after the lysis step and before magnetic nucleic acid isolation. The concentration of eluted nucleic acid in mock extracted samples (no sample material) was calibrated to approximately 40 copies of exogenous DNA/µL and compared with a 100-copy system suitability control. A second real-time fluorogenic 5′-nuclease PCR assay was used to target the exogenous template to serve as a sample suitability control and was performed simultaneously with the C. cuniculi assay. Nucleic acid recovery control assays for samples that had greater than a log₁₀ loss of template copies compared with control wells were diluted 1:4 and retested, reextracted, or both before accepting results as valid. To demonstrate the master mix and PCR amplification equipment function, a 100-copy/reaction positive control plasmid template containing the C. cuniculi target template was co-PCR amplified with the test sample. Additional primers amplifying an approximately 800-bp region encompassing the real-time PCR target region were used to confirm the specificity of the target region by sequencing. PCR products for select samples were purified and sequenced using the Sanger method (Genomics Core, Tufts University Medical School). Sequence results were further processed by trimming the sequencing primers and any undetermined nucleotide bases (Geneious Prime Software; version 2024.0.4; https://www.geneious.com). The clean consensus was analyzed by comparing it to isolates sequenced for this study and C cuniculi sequence data available in GenBank.

Results

Microbiological and histopathologic findings in index cases.

There were 1,193 cases of LS in NSG, NSG-SGM3, NBSGW, and NSGW41 strains of immunodeficient mice across 6 study rooms and 2 breeding rooms observed between February and August of 2024. Most if not all cages had greater than 3 mice and as such, LS once noticed in the cage was always present even though there were anecdotes of intermitted LS in cages with less than 3 mice. We tested 823 fecal samples using aerobic and anaerobic cultures. On average, 8.7% of these samples were positive by anaerobic culture and Vitek2 ANC card-based biochemical phenotyping for E. clostridioformis albeit with 90% to 94% probability. This lower than excellent (score of greater than 96%) probability was due to the contraindicating biopatterns of fermentations of α-L-arabinofuranoside, 5-bromo-4-chloro-3-indoxyl-α-mannoside, and 5-bromo-4chloro-3-indoxyl-β-N-acetyl-glucosamide and enzyme activities of L-pyrrolidonyl arylamidase and urease, indicating the possibility of misidentification of the anaerobe as E. clostridioformis and the lack of representation of C. cuniculi in the Vitek2 reference database. Henceforth, Clostridioides isolates from this study will be referred to as C. cuniculi. Necropsy on 13 of these index cases representing the first reported cases of LS involving NSG-SGM3 (n = 6), NSGW41 (n = 4), NSG (n = 2), and NBSGW (n = 1) strains did not reveal any gross abnormalities. On histopathology, pertinent lesions were restricted to the intestines. These mice often had erosion of the jejunal villi with neutrophilic inflammation, and all had Gram-positive bacterial rods adhering to the cecal mucosa with varying degrees of mucosal hyperplasia, epithelial vacuolation, and apoptosis (Figure 2A and B). Anaerobic bacterial isolates from the intestine of these mice were identical to the fecal isolates and stained as Gram-positive rods with central spores (Figure 2C and D).

Figure 2.

Histopathology of intestinal tract from index cases and challenge study. (A) Jejunum from an index case depicting erosion of the apical mucosa (arrows) with associated clusters of degenerate neutrophils (*) in the lumen. (B) Cecum from the same mouse shows mucosal hyperplasia and apoptosis (circles) with rodlike bacteria adhering to the epithelial surface (inset: Gram stain, arrows). (C) Gram strain of bacteria cultured from an index case, revealing they are Gram-positive, rod-shaped bacteria arranged in chains. (D) Romanowsky stain of the same bacteria, in which the formation of central spores (arrows) is present. Challenge study showing no pathology of (E) the cecum from an unexposed NSG mouse and (F) the cecum from an exposed mouse with loose stool exhibiting mucosal hyperplasia (note increased mucosal thickness) and apoptosis of the epithelium (circled).

Challenge study establishes C. cuniculi as the etiological agent.

The lack of specific identity from automated biochemical phenotyping and the mild but unique intestinal pathology raised the concern for dysbiosis or infection with an unknown bacterial pathogen. To investigate this we exposed naive, anaerobic bacterial culture-negative NSG mice to infected bedding from LS cases from index animals. To control age-dependent changes to gut microbiome, we used 2 groups of NSG mice: unexposed (control; n = 4; 2 mice per cage) and exposed to infected bedding (treatment; n = 4; 2 mice per cage). Fecal samples were taken for culture (all mice in both groups) and metagenomics (one mice from each cage) during week 1 (preexposure) and week 3 (postexposure). Among the 4 exposed mice, 2 mice (that is one from each exposed cage) exhibited LS between days 8 and 11 and 3 were culture positive for the above-described anaerobic isolate. The unexposed mice remained culture negative with normal stool consistency during this period and had normal cecal morphology on histopathology (Figure 2E). In contrast, the cecal tissue from all 4 mice from the exposed group exhibited hyperplasia (arrows) and apoptosis (circles) of the epithelium (Figure 2F). There were no significant changes in the body weights between unexposed (mean: 24.5 g; SEM: 1.32) and exposed (mean: 23.75 g; SEM: 1.03) groups (P = 0.67). Metagenomic analysis for α diversity using the Shannon index in the unexposed group were 4.60 ± 0.04 and 4.39 ± 0.06 on weeks 1 and 3, respectively (P = 0.034) while the exposed group had values of 4.55 ± 0.14 and 6.02 ± 0.29 on weeks 1 and 3, respectively (P = 0.13). Beta diversity using UniFrac distance indicated a significant difference between control (that is all preexposure and week 3 unexposed samples) and week 3 exposed samples (Figure 3A, green ellipse represents 95% CI). These findings along with the microbiology and histopathology data prompted us to perform relative abundance analysis on the Clostridiaceae family. C. cuniculi was the only known pathogenic species in the sample set and was only identified in the week 3 exposed group (Figure 3B, arrows). We further evaluated the entire data using heatmap visualization of relative abundances of core bacterial species with a mean relative abundance greater than 1%. Hierarchical clustering analysis revealed 2 distinct clusters: one comprising the week 3 samples from the exposed group, and the other comprising the remaining 6 samples from control and baseline week 1 samples from the exposed group (Figure 3C). The 2 clusters were subjected to LEfSE. C. cuniculi, the only pathogenic species identified in this analysis was restricted to cluster 2 (that is the exposed group at week 3, Figure 3D). Further, ELISA on intestinal contents from this study was negative for C. perfringens and C. difficile toxins indicating toxin products from these species were not present and did not represent the suspected pathogenic Clostridioides species.

Figure 3.

Shotgun metagenomics from the challenge study, in which naïve, culture-negative NSG mice were exposed to infected bedding from LS cases. (A) PCoA of beta diversity using UniFrac distances. Green ellipses represent 95% CI for samples and exclude exposed mice from week 3. (B) Relative abundance for Clostridiaceae family. C. cuniculi is only present in the exposed group at week 3 (dotted outline and purple). (C) Heatmap visualizing the relative abundances of core bacterial species, defined as those with a mean relative abundance greater than 1%. Hierarchical clustering analysis revealed 2 distinct clusters: one comprising the week 3 samples from the exposed group, and the other comprising the remaining 6 samples. (D) LEfSE on Cluster 1 or Cluster 2 from C. C. cuniculi, the only pathogenic species identified, was restricted to Cluster 2 (i.e. exposed group at week 3).

Whole genome characterization and phylogenetic analysis.

Comparison of the genomes of 3 C. cuniculi isolates (W41, 819933, and 823059; Table 1) from mice in this study demonstrated significant similarity with GenBank isolates from chicken and rabbit, and further, agreement with the curated OneCodex database assignment as C. cuniculi. Average nucleotide identity (ANI) analysis reveals a clustering of the C. cuniculi isolates alongside a single C. saudiense (strain An435), and distinctly clustered from other isolates of C. celatum and C. saudiense (Figure 4A). Isolates appearing as red cells in the pairwise alignment share a percent identity >0.95 which is consistent with identification as the same species. Data support C. cuniculi as its own species, distinct from C. saudiense and C. celatum, and suggests that C. saudiense (strain An435) may be misnamed. In conjunction with percent identity, sufficient alignment coverage was attained to compare these genomes (Figure 4B), indicating sufficient data and appropriate comparison to draw such conclusions. Further, the identity of our isolates as C. cuniculi is validated by the pairwise comparison of genomes for similarity errors (Figure 4C) that measure the number of bases/positions that do not match exactly (including nonidentities + insertions + deletions) and analysis of alignment lengths (Figure 4D), defined as the count of bases contributed by each genome to the pairwise alignment between those genomes.

Table 1.

Genome features of murine C. cuniculi isolates

Item/genome features	Description
Item
Genome assembly data	W41-1	823059	819933
Assembly method	Filtlong + Miniasm + Flye + Medaka
Sequencing technology	Nanopore minION
No. of contigs	20	187	24
Assembled genome coverage	82×	11×	40×
Genome features
Size (Mb)	4.5 Mb	4.3 Mb	4.4 Mb
Maximum contiguous length (bp)	2,184,169	4,333,138	1,544,445
GC content (%)	28%	28.1%	28%
No. of coding sequences (total)	4,180	4,125	4,040
No. of genes	4,327	4,289	4,184
rRNA (23S, 16S, 5S)	30	25	31
tRNA	103	126	102
ncRNA	14	13	11
Regulatory	64	63	63

Figure 4.

Comparative genomics of C. cuniculi isolates relative to other Clostridioides species using average nucleotide to perform comparisons identifying homologous regions. Pairwise comparisons are used to identify (A) the proportion of identical aligned regions, (B) the amount of each genome that is aligned, (C) the number of unaligned or nonidentical bases, and (D) the total number of aligned bases in each genome.

In addition to ANI analysis, clusters of orthologous genes (COG) analysis of 49 marker genes demonstrates the phylogenetic relatedness of C. cuniculi genomes (Figure 5). Concatenation and alignment of the 49 marker genes demonstrate that C. cuniculi are closely related to previously published C. cuniculi isolates, the suspected misnamed C. saudiense (strain An435) and distinctly clustered apart from C. celatum isolates. Take together, the 2 independent methods of ANI and COG analysis support identification of C. cuniculi as the speciation of the Clostridioides species isolated from LS mice in this study.

Figure 5.

Concatenated sequences from 49 clusters of orthologous genes for each genome were aligned to determine relatedness similarity. C. cuniculi genomes are bold and include information about host species from which the isolate was obtained.

Antibiotic sensitivity screening and treatment.

Due to the lack of standards for antibiotic sensitivity testing against C. cuniculi in the Vitek 2 system, we pursued agar disc diffusion testing to evaluate antibacterial sensitivity. Vancomycin and gentamicin were used to demonstrate their presumed intrinsic efficacy and resistance, respectively, against Clostridioides species, while TMS was used due to its widespread use in laboratory mice and availability as a feed premix. The ranges for the zones of inhibitions were between 6 to 9 mm for gentamicin, 19 mm for vancomycin, and 21 to 32 mm for TMS (Figure 6). TMS was administered orally through diet to nonbreeding mice of NSG and NSG-SGM3 strains in 81 cages from rooms with a history of LS and positive fecal culture for C. cuniculi. There were no new or ongoing cases of LS within one week of transition to the TMS diet and remained free of LS until the end of the studies ranging from 2 wk to over 6 mo. No effort was made to transition such cages back to a standard irradiated diet. Seventy-nine of 81 cages were culture negative for C. cuniculi when tested 5 to 6 wk after the transition to the TMS diet. Two naïve culture negative NSG mice on a regular diet were exposed to dirty bedding from cages on a TMS diet. Their fecal samples remained culture negative for C. cuniculi 21 days postexposure to dirty bedding. A newly established breeding colony of NBSGW founded with C. cuniculi culture negative breeders developed LS (40% of the cages) over a period of 3 mo postestablishment of the colony. Upon testing, 75% of the cages were culture positive. Subsequently, all cages were transitioned to the TMS diet, were cleared of LS within one week, and were culture negative for C. cuniculi after 5 wk. However, unlike the study cages of NSG and NSG-SGM3 strains, 16% of these NBSGW breeding strains developed LS and were culture positive for C. cuniculi over a period of 3 mo. During this period, 50% of the offspring cages weaned from the breeding cages were placed on a regular diet resulting in culturable C. cuniculi and the development of LS. We were unable to pursue a PCR-based assessment of the effectiveness of TMS treatment during this period as this resource was still in development at this time.

Figure 6.

Agar disc diffusion for assessing antibiotic effectiveness against C. cuniculi.

PCR-based environmental monitoring.

RT PCR was used to assess the effectiveness of fomite transmission in 2 study rooms housing NSG mice with LS that were culture positive for C. cuniculi. We detected C. cuniculi on the surface of several pieces of equipment, including horizontal exhaust plenums of a ventilated rack, a lixit, and a biosafety cabinet. In the ventilated rack tested, 2 samples pooled from 3 horizontal plenums corresponding to positive cages had copy numbers ranging from 10 to 100 while the rest of the plenums were negative and corresponded to empty slots or cages that were negative by RT PCR. LS animals showed higher estimated copy numbers per reaction for C. cuniculi compared with asymptomatic animals (Figure 7).

Figure 7.

Real-time PCR values from environmental testing (black) in comparison to fecal samples from LS mice (red) or asymptomatic animals (blue).

Identification of the source of infection and epizootiology.

The source of infection was traced to the NBSGW breeding colony on an irradiated diet based on the following epizootiological evidence.

A retrospective fecal microbiome analysis from 94 in-house bred mice (56 NSG, 12 NSG-SGM3, 7 NBSGW, 17 C57BL/6J, and 2 CD1) over a span of 4 y, which included the 3 y preceding the incidence of LS in immunodeficient mice, revealed that the presence of C. cuniculi was restricted to fecal samples submitted subsequent to the outbreak of LS and was confined to NBSGW, NSG-SGM3, and NSG mice (read abundance 0.71% to 4.8%). This eliminated immunocompetent mice as the potential source of C. cuniculi in immunodeficient mice.

In the NSG breeding colony housed in a dedicated room with an average cage count of 500 (average 4 mice per cage), C. cuniculi was undetectable by culture and PCR. Only NSG mice cohoused with NSG-SGM3 mice in study rooms were positive for C. cuniculi. In turn, the NSG-SGM3 mice originated from a long-standing breeding colony with over 5 y of healthy history until cohoused with NBSGW breeding mice. The founding NBSGW strains were introduced into this room directly from the vendor 6 mo before the first report of LS in study rooms cohousing NSG and NSG-SGM3 mice. Concurrent with the observation of LS in study rooms, we observed LS in breeding NBSGW and NSG-SGM3 strains with several cages testing positive for C. cuniculi by culture (Figure 1B).

NBSGW and NSG-SGM3 breeding colonies were reestablished in another dedicated room with vendor-derived mice. Both strains were screened for C. cuniculi and found to be negative and fed an irradiated diet for 2 wk before transitioning the NBSGW strain to the autoclavable diet used by the vendor. However, within 3 mo, the NBSGW strain developed LS as described earlier under the results on the effectiveness of the TMS diet. All NSG-SGM3 (39 cages; average 4 mice/cage) in the same room were culture and PCR negative and free of LS during this period. Similarly, a study room dedicated to the NBSGW strain housed vendor-sourced mice on the irradiated diet. Within 3 mo, close to 40% of the cages presented with LS and were identified by culture to be positive for C. cuniculi. This room was decontaminated with a sporicidal disinfectant and repopulated with vendor-sourced NBSGW strain but on an autoclaved diet and remained free of LS and C. cuniculi for 4 mo, when the study concluded. These findings are summarized in Table 2.

Table 2.

Summary of several months of observation and testing that identified a strong association with NBSGW strain and exposure to irradiated diet

Breeding status/strain	Irradiated diet	Autoclaved diet
Breeding
NBSGW	Positive for C. cuniculi (incidence >40%) with LS on 3 separate occasions of starting a breeding colony within a 3-y period	Not done
NSG	Negative for C. cuniculi based on culture and PCR, and absence of LS during the last 5 y in a breeding colony of over 2,500 mice at steady state	Not done
Nonbreeding study rooms
NBSGW	Positive for C. cuniculi by clinical signs, culture, or PCR within 12 to 16 wk postweaning or arrival from vendor on 2 separate studies	Negative for C. cuniculi by culture and PCR and free of LS on 3 separate 16-wk studies
NSG	Negative for C. cuniculi based on culture and PCR and free of LS on several studies where NSG strain was not cohoused with NSG-SGM3 exposed to NBSGW	No data available

Discussion

In this study, we have identified and characterized C. cuniculi infection and enteric disease in the highly immunodeficient NSG, NSG-SGM3, and NBSGW mouse strains, which lack mature T cells, B cells, and natural killer cells. We also found epizootiological evidence that unlike NSG and NSG-SGM3, the NBSGW strain is uniquely predisposed to C. cuniculi infection through the feeding of an irradiated natural ingredient rodent diet. However, NSG and NSG-SGM3 strains get infected when exposed to infected NBSGW strains or indirectly through contaminated fomites.

After eliminating common viral and bacterial agents as potential pathogens, we suspected Clostridioides species as the potential etiological agents given the reports of C. difficile and C. perfringens in mice.^4,6 ELISA and PCR testing on intestinal contents or feces from the affected mice excluded this possibility. Further, we did not notice mortality in the affected mice nor pathologic changes in their intestines such as transmural hemorrhage, gaseous distention, or severe ulceration and submucosal edema that are typically associated with C. perfringens and C. difficile infections.^5,6 However, we noticed subtle changes to the intestinal tissue with frequent changes to the cecal epithelium in the form of cellular vacuolation and apoptosis. As such, we undertook comprehensive fecal microbiologic testing including anaerobic culturing. All identified aerobic organisms were commensals of laboratory mice while the anaerobic culturing and rapid biochemical phenotyping consistently isolated and identified E. clostridioformis, a member of the class Clostridia, but with atypical biochemical patterns that reduced the probability of match to 90 to 94%. This isolate was only present in immunodeficient mice with LS and was considered a strong suspect in the etiopathogenesis. To establish the identity of the organism and prove cause and effect, we took a 2-pronged approach of fecal metagenomics and whole genome sequencing of the isolates. For metagenomics, we challenged NSG mice from our colony with dirty bedding from LS mice that were culture positive for the anaerobic isolates. To enhance statistical power and account for the time-dependent variation in the metagenome, we compared pre- and postchallenge feces alongside samples from unchallenged mice. This analysis unambiguously identified C. cuniculi in 75% of the postexposure samples from the challenged group. Fifty percent of the challenged mice also exhibited clinical signs consistent with LS with a prepatent period of 8 to 11 days. Interestingly, although the α diversity in the exposed group after the 3-wk period was higher than the baseline in comparison to the unexposed control group, the difference was not statistically significant probably due to the higher variance in microbial composition after C. cuniculi infection. Nevertheless, like the C. cuniculi infection in rabbits,¹⁵ the α-diversity points to C. cuniculi as the primary driver of LS over dysbiosis. These findings prompted us to perform a retrospective analysis of fecal metagenomics data from 94 samples comprising immunocompetent and severely immunodeficient mouse strains. We were surprised that C. cuniculi was absent in all the fecal samples collected since 2021 except for the immunodeficient samples collected during the peak period of this outbreak. This clearly identified C. cuniculi as a recently introduced opportunistic agent to our immunodeficient mouse colony and cleared immunocompetent mice as the probable source of the outbreak. We later confirmed this conclusion using C. cuniculi-specific PCR testing.

C. cuniculi was originally described in 2018 when identified in association with epizootic rabbit enteropathy.¹⁵ The genomic sequence from that study was deposited in NCBI in 2019, followed by additional genomic sequences deposited in 2020 (one human) and in 2023 (2 chickens). However, the human isolate has an inconclusive taxonomy status and may not be C. cuniculi. Nevertheless, the nomenclature for C. cuniculi has been effectively published but not validated under the rules of the International Code of Nomenclature of Prokaryotes (Bacteriological Code). Whole genome sequencing of the 3 isolates in mice from our study showed significant similarity with the published rabbit and chicken isolates by forming a tight cluster in the ANI analysis and is phylogenetically related to C. celatum and C. saudiense. Interestingly, a previous study¹¹ investigating the cause of a diarrheal outbreak in immunocompromised mice using a metagenomic approach revealed the enrichment of 4 bacterial taxa that included C. celatum. However, the possibility of misidentification of C. cuniculi as C. celatum was likely if the database used for the analysis was devoid of the C. cuniculi genome due to the contemporaneous dates for that study and characterization of C. cuniculi in rabbits. Such a revelation points to the fact that C. cuniculi infection may be an important cause of idiopathic and sporadic diarrheal outbreaks in highly immunodeficient mice.

The high morbidity with the very low mortality rate we observed in this outbreak prompted us to explore the potential of oral antibacterial therapy to control clinical signs and prevent shedding and spread of disease. Standard antibiotics like vancomycin and clindamycin with proven efficacy against obligate anaerobes like Clostridioides species were eliminated given its narrow spectrum and potential to significantly perturb the gut microbiome.²⁰ Unlike these narrow spectrum antibiotics, TMS is shown to minimally perturb gut microbiome and has potential efficacy against C. difficile when provided as a feed premix.^4,21 However, we were unsure whether the protection against C. difficile was direct or indirect through changes to the composition of certain microbial communities as such changes can be protective against pathogen colonization.^22,23 We pursued agar disk diffusion assay to compare the sensitivity of TMS toward C. difficile while comparing it against vancomycin, a well-established antibiotic with efficacy against Clostridioides species, and gentamicin, an aminoglycoside that is intrinsically ineffective against obligate anaerobes. This study, although qualitative and lacking information on the clinical breakpoint for susceptibility, clearly established the direct action of TMS against C. cuniculi. The availability of TMS as a feed premix, and widespread acceptance within the research community of using this feed in immunodeficient mice allowed us to transition to this diet in C. cuniculi-positive study rooms. Cages were free of LS in one week and remained so for the entire span of studies, which in several cases was over 8 wk. About 6 wk into TMS treatment, close to 97% of the study cages of NSG mice were culture negative. Concern for falsely negative culture results due to TMS in the feces prompted us to expose naïve NSG mice to pooled dirty bedding from culture-negative cages; these mice remained culture negative. Over the last 6 mo, feedback from the research community indicates no impact on study endpoints involving C. cuniculi infected and TMS-treated NSG and NSG-SGM3 strains used for patient-derived leukemia, brain tumor, and chimeric antigen receptor T-cell studies. In the C. cuniculi -positive NBSGW breeding colony, transition to TMS diet resulted in a LS-free period of 5 wk, at which time a subset became culture positive. Offspring from these mice also became culture positive with LS after weaning and transitioning to non-TMS diet. This highlights the strain-dependent efficacy of TMS against C. cuniculi and the importance of innate immunity through circulating neutrophils that are impaired in NBSGW for clearing the infection.¹⁶

Whole genome analysis of the 3 isolates also facilitated the design of a discriminating real-time PCR probe for C. cuniculi. However, we were not able to employ this assay during the early through the peak periods of the outbreak. By the time we started using PCR, the infection was already managed through a combination of quarantine, TMS diet, and depopulation of infected breeding colonies of NBSGW. However, during this period, we used the PCR to screen some of the last positive cages of mice and the environment to understand the spread of infection through fomites like doorknobs, biosafety cabinets, and imaging equipment. We also demonstrated the potential for plenum-level testing of individually ventilated racks and its specificity to differentiate rows containing positive cages from rows of negative cages. Due to the scarcity of infected cages, we were not able to further investigate the sensitivity (that is the minimum number of infected cages or mice per row) of this approach. However, such a methodology was successful in eliminating Corynebacterium bovis in a mouse colony and should be investigated.²⁴ Further, we used this PCR assay to confirm the SPF status of our breeding NSG colony that remained consistently negative by culture and free of LS during this outbreak. Finally, we used the PCR to confirm the C. cuniculi-free status of several C57BL/6 mouse colonies throughout our facility. This helped us focus our investigation on the source of this outbreak in the immunodeficient colonies.

Early during the outbreak, we suspected the source to be the NSG colony. This was because 70% to 80% of the immunodeficient mice in studies belonged to the NSG strain. However, concurrent findings of LS in a room dedicated to the breeding of NBSGW and NSG-SGM3 strains, transfer of NSG-SGM3 mice to several study rooms housing NSG mice, a history of no LS in the NSG-SGM3 breeding room until the introduction of the NBSGW strain, and the lack of LS in the room dedicated to the breeding of NSG strain indicated NBSGW strain as the most likely source of this outbreak. Further, we thrice established breeding and study colonies of NBSGW with vendor-sourced mice on an irradiated natural ingredient diet. On all 3 occasions, we noticed the onset of LS requiring depopulation despite stringent SPF conditions. In our third attempt, vendor mice were screened by culture for C. cuniculi and found to be negative before the commencement of breeding operations. When we noticed LS in this colony after 3 mo, we focused our attention on the irradiated diet. We have successfully housed NBSGW mice in a dedicated room on an autoclaved natural ingredient diet for over 4 mo without any evidence of LS or detection of C. cuniculi by culture and PCR. However, we continue to house NSG and NSG-SGM3 strains on the original irradiated diet without any evidence of LS or detection of C. cuniculi by culture and PCR. Taken together, we present strong circumstantial evidence for a permissive host environment in NBSGW for colonization with C. cuniculi spores that survives irradiation of a natural ingredient diet. Such a situation is not unlikely given published evidence of numerous Clostridioides species in nonsterile, natural ingredient laboratory animal diets.²⁵ Further, it is known that high levels of murine parvovirus in a natural ingredient diet can survive standard irradiation protocols.²⁶ As such, it is possible that spore-forming organisms like C. cuniculi may survive a dose of 25 kGy, a minimum dose typically used to irradiate natural ingredient rodent diets.²⁷ Our attempts at culturing and PCR identification of C. cuniculi in the irradiated diet were unsuccessful but not surprising given the ability of irradiation to reduce microbial count thus making sampling and direct identification extremely challenging. It is possible that a permissive host like NBSGW, when exposed in adequate numbers and length of time to an irradiated diet, may provide conditions conducive to the germination, proliferation, and colonization of C. cuniculi. Clearly, our observations show that once infected the NBSGW strain can efficiently infect NSG and NSG-SGM3 strains.

There is a pressing need to identify the source of C. cuniculi spores, notably, to definitively demonstrate the presence of C. cuniculi in irradiated natural ingredient diet. A starting point for this analysis could be the nonsterile diet and the feed ingredients commonly used in these diets. For example, fish meal may be a concern given reports of high prevalence (greater than 54%) of C. perfringens in fish for human consumption.²⁸ We also plan to investigate the host factors that make certain strains of mice like NBSGW highly susceptible to amplifying hosts so that measures can be taken to change operational practices in the care and use of such strains. A secondary hypothesis for why C. cuniculi infection may be variable extends from the observation that the rabbit C. cuniculi genome encodes an enterotoxin. Given the potential variation in microbial genetics between strains, including the capacity for horizontal gene transfer, one could speculate that toxin status could correlate with the severity of disease. Further investigation is ongoing to evaluate the genetics and relevance of the encoded enterotoxin.

In summary, our discovery of C. cuniculi as an opportunistic agent of severely immunodeficient mice suggests that the sequence of events we describe in this study may be currently unfolding in other institutions under different conditions, severity, and incidence resulting in sporadic diarrheal outbreaks, necessitating further investigation to understand its prevalence and impact on biosecurity and study outcomes.

Conflict of Interest

The authors have no conflicts of interest to declare.

Funding

The work was internally funded.