Distinct host–pathogen and microbiome responses of aoudad (Ammotragus lervia) and bighorn sheep (Ovis canadensis) following exposure to Mycoplasma ovipneumoniae with or without co-exposure to leukotoxigenic Pasteurellaceae

Abstract

Background

Pathogens can shape their host communities over various timescales. The potential role of host–pathogen co-evolution in driving contemporary shifts in disease ecology is becoming increasingly important as host species emerge and persist outside their native ranges. In North America, Mycoplasma ovipneumoniae can cause fatal pneumonia epizootics in native bighorn sheep (Ovis canadensis), whereas introduced free-ranging sympatric aoudad (Ammotragus lervia) typically act as asymptomatic reservoirs. We describe the responses of two hosts with different biogeographical histories to shared pathogen exposure through a lens of potential host–pathogen co-evolution. Specifically, this work integrates findings on microbiome composition and host transcriptomic responses in aoudad and bighorn sheep before and after controlled exposure to M. ovipneumoniae, with or without leukotoxigenic Pasteurellaceae.

Results

Aoudad maintained significantly higher microbial richness (Shannon) in the lower respiratory tract, whereas bighorn sheep experienced post-exposure microbiome perturbations and enhanced growth of some opportunistic taxa. Both molecular detection of and humoral antibodies to M. ovipneumoniae reduced the relative abundance of key genera (e.g., Bibersteinia, Mannheimia, Pasteurella, Roseomonas) in the upper respiratory tract, but post-exposure bacterial community alteration was more pronounced in bighorn sheep. Transcriptome profiling revealed that, compared to aoudad, bighorn sheep upregulated pro-inflammatory and oxidative-stress pathways—including interleukin-1, interleukin-12, and NF-κB signaling—alongside reactive oxygen species generation. In contrast, aoudad exhibited comparatively muted inflammatory signatures, enhanced expression of molecular chaperones, antigen-processing machinery, and integrin-mediated regulatory genes (notably CD46, ILK, and NFKBIZ). Network analysis identified distinct hub genes likely underpinning effective pathogen clearance and mucosal resilience in aoudad versus immunopathology in bighorn sheep.

Conclusions

Our integrated microbiome and transcriptomic data underscore the importance of co-evolutionary history in driving host-specific responses to shared respiratory pathogens. Aoudad display microbiome stability and balanced immunoregulation, whereas bighorn sheep suffer dysbiosis and excessive inflammation, potentially increasing mortality risk. Explicit consideration of evolutionary and ecological context regarding host–pathogen co-evolution may increase overall understanding of observed pathobiological and epidemiological patterns commonly targeted for disease management interventions.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12917-026-05375-1.

Background

Pathogens shape ecological systems in many ways, and their effects are appreciated across multiple biologically significant timescales. Specifically, pathogens can influence fitness and life history traits, speciation processes and species distributions, and ecological community dynamics [1–5]. These aspects are often targets for conservationists and wildlife managers [6], so understanding the potential impacts of pathogens on these processes is crucial. Parasites exhibit impacts on populations that are relatively short at an ecological scale, but these impacts compound and generate spatiotemporally specific evolutionary consequences [7, 8]. Specifically, environmental variation at an ecological scale can drive variation in infection probability for a given host, as well as host responses to infection [7, 9]. Great variation in responses to infections also exist at the individual and species levels [9, 10]. As associations between host and pathogen continue, the nature of host–pathogen relationships can change through co-evolution (or co-accommodation) [6, 11, 12]. The continuation of host–pathogen interactions can be expected to decrease the virulence and/or presence of clinical disease, as well as increase the genetic diversity of the pathogen [12]. Thus, patterns of co-evolution between host and pathogen can shape how hosts respond to infection and how the pathogen is maintained across susceptible hosts. The anthropogenic movement of animals and their pathogens plus modification of habitat and landscape features that govern pathogen exposure all serve to uniquely alter host–pathogen relationships [5, 13]. These anthropogenic changes frequently occur in a shorter timeframe than the processes governing host–pathogen co-evolution in the absence of human disturbance [14]. Consequently, altering host–pathogen systems can cause acute and/or persistent issues for wildlife conservation and management.

The introduction of nonindigenous Artiodactyls and their pathogens to North America has significantly impacted native species. Examples include introduction of exotic lice from introduced fallow deer (Dama dama) causing hair-loss syndrome in mule deer (Odocoileus hemionus) in Pacific Northwestern United States [15], Brucella abortus spillover from domestic cattle (Bos taurus) to Rocky Mountain elk (Cervus canadensis) in Wyoming [16], and recent die-offs of Wyoming pronghorn (Antilocapra americana) due to Mycoplasma bovis maintained by domestic cattle [17]. In cases of novel/emerging infections resulting from the anthropogenic movement of animals, one potential driver of these conservation and management issues may ultimately lie in the differences in the co-evolutionary history between hosts and pathogens.

While the examples above demonstrate relatively recent disease issues stemming from nonindigenous hosts and pathogens, fatal diseases of desert bighorn sheep (Ovis canadensis; BHS) following contacts with domestic sheep (Ovis aries; DS) have long been known to negatively impact BHS populations. Evidence of BHS mortality associated with novel/introduced pathogen(s) from DS contact dates as early as 1942 [18]. However, the coincidence of large decreases in BHS populations with widespread expansion of DS across the western United States in the late 1890 s may suggest fatal pathogens spilled over earlier than 1942 [19, 20]. Up to 60% of DS harbor the bacterial agent Mycoplasma ovipneumoniae with little or no disease [21]. Moreover, this agent has been implicated in fatal pneumonia epizootics in BHS across the United States and Canada [21–23]. Physical separation of DS and BHS has successfully mitigated pneumonia epizootics in BHS, and is a primary tool used by wildlife managers to benefit BHS health [24]. Indeed, M. ovipneumoniae multi-locus strain typing efforts have repeatedly implicated DS as the pathogen source for BHS dying of respiratory disease [21, 25]. At the population level, a single DS flock harbors a high diversity of M. ovipneumoniae strains whereas a BHS population during a pneumonia epizootic typically harbors one distinct strain [21]. Recently, this pattern of high strain diversity in DS has been identified as a signal of historical host–pathogen co-evolution that is absent in BHS [26]. While DS are frequently implicated in respiratory disease epizootics in BHS, recent diagnostic cases demonstrate that these epizootics can also occur in BHS without apparent sharing of M. ovipneumoniae, Pasteurella multocida, or Bibersteinia trehalosi strains [27]. Clinically, exposure of BHS to M. ovipneumoniae results in a greater decrease in upper respiratory bacterial clearance than is seen in DS [28, 29]. Given the pathobiology of the agent, the defective clearance was likely secondary to ciliostasis and/or loss of cilia on tracheoepithelial cells [30, 31]. Another co-evolutionary difference between these host species is reflected in toxicity patterns of the Pasteurellaceae species each host typically harbors in the upper respiratory tract. Domestic sheep harbor leukotoxigenic Pasteurellaceae species in the upper respiratory tract more frequently than do BHS [32]. Further, BHS neutrophils are significantly more susceptible to Pasteurellaceae-derived leukotoxin-A than DS neutrophils [33]. Leukotoxigenic Pasteurellacea and M. ovipneumoniae have synergistic pathological effects on the BHS respiratory tract, making disease more severe than with M. ovipneumoniae alone [29]. Taken together, the clinical, pathobiological, and epidemiological patterns observed in the DS-BHS respiratory disease system reveal tractable patterns of host–pathogen co-evolution that result in significant wildlife management outcomes.

Nonindigenous Artiodactyl introductions to North America are not limited to domestic commercial hoofstock. The exotic hoofstock industry is comprised of 1.5–2 million animals across 125 different species in the United States [34]. Animal husbandry practices for exotic hoofstock vary, and many species exist as free-ranging feral populations [35]. One of these species is the aoudad (Ammotragus lervia; AD). Aoudad are gregarious Caprinae native to northern Africa but have been extensively introduced outside of their native range to Spain, Italy, Greece, and the United States including Oregon, New Mexico, Texas, Arizona, California and Utah [36]. The largest North American populations occur in Texas where at least 20,000 individuals occupy the landscape [37]. Recent work has experimentally demonstrated that AD can function as a reservoir, maintaining M. ovipneumoniae within its populations, as well as a transmission source of M. ovipneumoniae for other species (such as BHS) with varying clinical responses, raising concerns about the estimated 8,000 feral aoudad occupying habitat crucial to BHS in Texas [38].

The phylogenetics of AD are complex [39], raising questions about their degree of co-evolution with and subsequent pathological response to M. ovipneumoniae and leukotoxigenic Pasteurellaceae exposure. The native range and biogeographic history of AD and DS suggest AD had opportunities for historical exposure to and potential co-evolution with M. ovipneumoniae and leukotoxigenic Pasteurellaceae [36, 40]. While the genomic structure of M. ovipneumoniae in domestic Caprinae demonstrates evidence of host–pathogen-co-evolution [26], no similar molecular evidence of this pattern in AD has been published to the knowledge of these authors. The increasing number of host species outside of Caprinae with M. ovipneumoniae infections [41–43] further illustrates this pathogen’s significance for wild ungulate species.

The respiratory microbiome is increasingly recognized as a significant driver of pathobiological outcomes to pathogen exposure [44]. The complex, often multi-pathogen, etiology of BHS respiratory disease highlights the importance of native microbiota and subsequent microbiome shifts observed in hosts with varying co-evolution histories before and after exposure to pathogens. As part of a separate effort to evaluate transmission propensity of M. ovipneumoniae and leukotoxigenic Pasteurellaceae from AD to BHS, we initiated an opportunistic multi-omics analysis. Specifically, we described the bacterial microbiome and transcriptomic signals of AD and BHS exposed to M. ovipneumoniae in the presence and absence of leukotoxigenic Pasteurellaceae. Our descriptions of microbiome and transcriptomic responses to respiratory pathogens in nonindigenous AD and native BHS seek to provide crucial information to assess and predict individual and population level outcomes of interspecific contacts. Further, our efforts illustrate the importance of characterizing features consistent with differential host–pathogen co-evolution histories to facilitate wildlife conservation and management. We present molecular implications of host–pathogen co-evolution in an applied eco-pathological study system, which may extend to applications in other disease systems, especially those involving species introductions.

Materials and methods

Experimental design

This study, approved under Texas A&M AgriLife Animal Care and Use Committee permit 2020-038A, involved capturing 20 BHS (11 females, nine males, aged 0 to 4 years) from the Elephant Mountain Wildlife Management Area (30°02′03.2" N 103°33′57.6″ W) and relocating them to Mason Mountain Wildlife Management Area (30°49′18.8″N 99°13′20.0″W) on study day zero. Bighorn sheep were subjected to one of three treatments: 1) not allowed contacts with experimentally infected AD (control BHS, n = 6) or other BHS, 2) allowed contacts with AD experimentally infected with DS-origin M. ovipnuemoniae (MOP BHS, n = 6), or 3) allowed contacts with AD experimentally infected with DS-origin M. ovipnuemoniae and leukotoxigenic Pasteurellaceae (wash BHS; n = 7). Aoudad (eight females, four males, aged 0 to 2 years) were captured in a semi-free ranging setting in Mason County, Texas and brought to the study site on day 40 (after the 39-day BHS acclimatization period) and subjected to one of two treatments: 1) experimentally infected with DS-origin M. ovipnuemoniae (MOP AD, n = 6), or 2) experimentally infected with DS-origin M. ovipnuemoniae and leukotoxigenic Pasteurellaceae (wash AD, n = 6). Contacts between MOP AD and MOP BHS began by sharing a water source (days 40 through 72) followed by direct physical commingling within the same pen (days 95–111) whereas contacts between wash AD and wash BHS were limited to shared water use (days 40 through 72) due to the onset of clinical respiratory disease followed by necessary euthanasia for wash BHS. Pens between each treatment (MOP AD near MOP BHS; wash AD near wash BHS; Fig. S7) were adjacent with dividing fencelines modified to selectively prevent direct interactions and restrict pathogen transmission routes during the water sharing period (days 40 through 72). The control BHS pen was separated from those of all other experimental animals.

Each animal (experimental units, 20 BHS and 12 AD) was randomly allocated to their respective species-specific treatment group upon initial capture. Neither age nor sex were considered when allocating animals to treatment groups. Randomization was achieved by having an unbiased Texas Parks and Wildlife Department employee obtain and distribute animals from the trailer to treatment-specific facility pens. However, because all animals were fit with ear tags upon capture, and were retained throughout the experiment, research personnel were privy to animal allocation immediately. Authors were aware of individuals’ treatment group memberships prior to sample processing and data analyses. Prior to analysis, one MOP BHS was excluded from the study due to mortality (day 39; traumatic injury during chemical immobilization) occurring before experimental contacts with AD (day 40). All serum samples for both BHS (days 0 and 39) and AD (day 40 prior) tested “not detected” when submitted to the Washington Animal Disease Diagnostic Laboratory (WADDL; Pulman, WA, USA) for assessment of serological exposure to M. ovipneumoniae.

Sample sizes were selected based on a balance of facility capacity, spatial requirements for animal welfare, and consideration of the native population from which the BHS were captured. Additionally, the present study was post-hoc to investigate experimental transmission of M. ovipneumoniae from AD to BHS. Thus, our power analysis was conducted based upon detecting differences in time-to-clinical score onset between MOP BHS and wash BHS. Specifically, we utilized a fixed-design two arm sample size calculation for time-to-event data using nSurvival in the gsDesign package in program R [45].

Daily health assessments determined that euthanasia via intravenous sodium pentobarbital (390 mg/mL) administration (86 mg/kg) of animals in stage III plane II anesthesia was necessary for scores exceeding five for five consecutive days [46]. Specifically, we observed animals at least once daily to observe the presence/absence of: coughing, nasal discharge, ear paresis, lethargy, anorexia, head shaking, and/or nose licking. Animals had continuous access to fescue and alfalfa hay, natural forage, mineral licks, commercial sheep and goat pellets, and water. By day 155, all 20 BHS were removed via mortality or euthanasia. Seven AD were euthanized on day 155, one on day 168, and two on day 211. Tonsil swabs were collected from BHS on days 0 (n = 19) and 39 (n = 19) and from AD on days 40 (n = 12) and 72 (n = 11). Pooled respiratory tissue samples were collected from BHS (n = 19) and all AD (n = 12) upon their respective euthanasia and/or mortality dates (Supplementary Dataset 6).

Animal restraint and handling

Bighorn sheep were initially captured (day zero) using a helicopter net gun, restrained with leg hobbles and blindfolds, and administered intramuscular azaperone (15 mg) immediately and sustained-release haloperidol (0.2–0.7 mg/kg; ZooPharm, Laramie, Wyoming 82,070, USA) 30–45 min after capture to minimize stress [47]. Additionally, 50 mg of eprinomectin (LongRange, Boehringer Ingelheim Animal Health, Duluth, GA 30096, USA) was administered subcutaneously at initial capture to mitigate gastrointestinal helminth loads. The same sedation and eprinomectin protocol was followed for AD, but initial capture involved physical restraint with blindfolds and leg hobbles only on day 40. Thereafter (days 39, 40, 72, euthanasia) animals were chemically immobilized via remote dart projector for animal handling and sampling with nalbuphine, azaperone, and medetomidine compounded at 40, 10, and 10 mg/mL, respectively [48].

Inoculum preparation and administration

We utilized pooled nasal washings from six of eight M. ovipneumoniae-positive DS, as determined by quantitative polymerase chain reaction (qPCR) results from previously collected nasal swabs submitted to WADDL, to create our inoculum [49]. Nasal washings consisted of sterile phosphate-buffered saline washed through the nares of DS until at least 50 mL were obtained. The inoculum for wash AD was unaltered whereas MOP AD inoculum was exposed to 123.5 µg/mL of ceftiofur sodium (1 g Naxcel powder; Zoetis Parsippany, NJ 07054, USA) for two hours at 37 ℃ immediately prior to inoculation to eliminate non-target bacterial overgrowth. Aoudad were inoculated on day 40 via instillation syringes for the corresponding AD group (MOP AD; wash AD); 5 mL administered into each of the nares, 10 mL administered into the pharynx, and 2 mL administered into each conjunctival sac (Fig. S8). Like previous findings suggesting leukotoxigenic Pasteurellacea are common in DS, six of the eight nasal washings from DS tested PCR positive for the leukotoxin-A gene at WADDL [32].

Sample collection and handling

Sterile sheathed rayon-flocked nasal and palatine tonsil swabs were collected in 2 mL of tryptic soy broth and frozen immediately at −80 ℃. Retropharyngeal lymph nodes, tonsils, tracheobronchial lymph nodes, cranioventral and dorsocaudal lung sections, and trachea were aseptically collected, pooled, and immediately frozen at −80 ℃ in tryptic soy broth with bovine serum albumin (ThermoFisher Scientific, 339 Waltham, MA 02451, USA). Lung tissue samples within the respiratory tissue pools were biased towards margins of grossly apparent pathological change. All pooled respiratory tissue samples and nasal swab samples were submitted to WADDL for the detection of M. ovipneumoniae by qPCR. Palatine tonsillar swabs targeting the detection of Pasteurellaceae were subjected to aerobic and anaerobic culture at the Texas Veterinary Medical Diagnostic Laboratory (College Station, TX, USA). Pooled respiratory tissue samples, however, were submitted to WADDL for aerobic and anaerobic culture for the detection of Pasteurellaceae. Leukotoxigenicity of tissue samples was determined by PCR at WADDL. Additional details regarding chronological order and experimental timing of sample collection events and their use in the present analysis can be found in Supplementary Dataset 6. Materials, methods, and approaches for the microbiome and transcriptomic analyses can be found in the methodological appendix within the Supplemental Information file.

Results

Microbiome

After DNA extraction, we observed a mean DNA concentration of 6.18 ng/mL (SD: 8.42 ng/mL) and 33.16 ng/mL (SD: 12.37 ng/mL) in tonsillar swabs and LRT tissue samples, respectively. After full-length 16S rRNA gene PCR, we observed a mean DNA concentration of 8.76 ng/mL (SD: 11.53 ng/mL) and 17.83 ng/mL (SD: 18.58 ng/mL) in tonsillar swabs and LRT tissue samples, respectively. Fifty-seven tonsillar swabs and 30 LRT tissue samples were successfully sequenced. Seven hundred and eight bacterial genera were detected in both types of samples. Library size ranged from 9,412 to 275,547 with a median of 137,575 reads in tonsillar swab samples, and from 4,402 to 370,997 with a median of 127,049 in the LRT tissue samples. Rarefaction curves, which estimate the number of bacterial genera per sample when they plateau, and the distribution of the 20 most abundant bacterial genera per sample type and host species are shown in Figures S1-S4.

Alpha diversity

Alpha diversity, which captures the estimated number (richness) and relative abundance (evenness) of the microbial population in a sample, was different in tonsillar swabs and LRT tissue samples. Tonsillar swab richness (estimated number of species present), as measured by Chao1 index, was significantly positively associated with M. ovipneumoniae serological and qPCR status (Supplementary Table 1). Shannon index (capturing diversity in the microbiome composition of each sample) was significantly higher in wash than in MOP treatment groups in tonsillar swabs (Supplementary Table 2, Fig. 1). Shannon index was significantly positively associated with M. haemolytica detection in tonsillar swabs (Supplementary Table 1; Fig. 1D). In LRT tissue samples, the Chao1 index was not significantly associated with any of the available metadata variables (Supplementary Table 2, Figs. 2A-D). Shannon index in LRT tissue samples was significantly higher in AD, in older animals, and when T. pyogenes was not present (Supplementary Table 2, Figs. 2A-D).

Fig. 1.

Influence of host species, treatment group, Mycoplasma ovipneumoniae detection status (via PCR), and Mannheimia haemoytica detection status on alpha diversity (Chao1 and Shannon diversity indices; A-D, beta diversity given by Bray–Curtis (E–H), by Jaccard (I-L), and by Aitchison (M-P) in tonsillar swabs collected from bighorn sheep (Ovis canadensis; sheep) and aoudad (Ammotragus lervia).Con, control bighorn sheep; t1, MOP treatment group; t2, wash treatment group. Ellipses represent 95% confidence ellipses around the corresponding centroids. Missing values (“NA”) are shown in gray

Fig. 2.

Influence of host species, treatment group, Mycoplasma ovipneumoniae detection status (via PCR), and Mannheimia haemolytica detection status on alpha diversity (Chao1 and Shannon diversity indices; A-D, beta diversity given by Bray–Curtis (E–H), by Jaccard (I-L), and by Aitchison (M-P) in lower respiratory tract samples (lung parenchyma and respiratory-associated lymph nodes) collected from bighorn sheep (Ovis canadensis; sheep) and aoudad (Ammotragus lervia). t1, MOP treatment group; t2, wash treatment group. Ellipses represent 95% confidence ellipses around the corresponding centroids. Missing values (“NA”) are shown in gray

Beta diversity

Beta diversity, which measures how similar the microbial populations of two samples are, showed varying associations to the evaluated variables. Species, age, M. ovipneumoniae serological status and qPCR status, and B. trehalosi detection were associated with the differences in microbiome composition in tonsillar samples in at least two of the dissimilarity measures evaluated (Supplementary Table 3, Figs. 1E,G,I, M, and O). These variables explained a low to moderate percentage of the variance in the dissimilarity distances utilized in this study (from 1.4–18.2%). However, there was evidence of significant differences in the dispersion for species (Fig. 1M), and M. ovipneumoniae qPCR status (Fig. 1G and K; P < 0.05) in at least one of the dissimilarity distances.

Species, sex, age, treatment group, M. ovipneumoniae qPCR status, P. multocida and M. haemolytica prevalence were associated with the differences in microbiome composition in LRT tissue samples in at least two of the dissimilarity measures evaluated (Supplementary Table 4, Figs. 2F-M and P). These variables explained a low to moderate percentage of the variance in the dissimilarity distances utilized in this study (from 5.5 to 14.6%). However, there was evidence of significant differences in the dispersion of the dissimilarity distances in species, sex, and M. haemolytica prevalence (Figs. 2H, L and P; P < 0.05). It is important to note that the dispersion was not evaluated in the case of age, as the test of multivariate homogeneity of groups dispersions does not allow for a continuous predictor.

Bacterial differential abundance

In tonsillar swabs, several bacterial genera were differentially abundant by host species (42 genera), age (70 genera), M. ovipneumoniae seroprevalence (292 genera), M. ovipneumoniae qPCR status (287 genera), and B. trehalosi prevalence (1 genus). All differentially abundant bacterial genera are listed in Supplementary Dataset 1 (except for B. trehalosi prevalence). Roseomonas, a clinically significant genus for BHS respiratory disease, was more relatively abundant in BHS than in AD (Fig. 3A; BH-corrected P < 0.05) while Moraxella exhibited the inverse pattern (BH-corrected P < 0.01, Fig. 3A). Mycoplasma relative abundance increased with age without reaching statistical significance (Fig. 3B; BH-corrected P = 0.8450475), whereas the relative abundance of Roseomonas and Providencia decreased as age increased (BH-corrected P < 0.05; Fig. 3B).

Fig. 3.

Scatter plot of the median difference in abundance by transformed q value (volcano plot). Dots represent bacterial genera. The red dot represents a statistically significantly differentially abundant bacterial genus in tonsillar swabs by species (A), age (B), Mycoplasma ovipneumoniae seroprevalence (C), and PCR status (D). The horizontal dashed line represents Benjamini–Hochberg adjusted P value (q) < 0.05. The bacterial genera labeled in the volcano plot have been previously associated with respiratory disease in ruminants

Animals seropositive for anti-M. ovipneumoniae antibodies showed significantly lower relative abundance of Bibersteinia, Providencia, Mannheimia, Pasteurella, and Moraxella (BH-corrected P < 0.05; 3 C) compared to seronegative animals. Similar findings were observed for M. ovipneumoniae qPCR status (Fig. 3D). Selenomonas relative abundance was increased in B. trehalosi negative animals (BH-corrected P < 0.05).

No differential abundant genera associated with sex, treatments, M. haemolytica, nor T. pyogenes prevalence were observed in tonsillar swabs. Additionally, no bacterial genera were detected in LRT tissue samples to be differentially abundant depending on any of the available variables (species, sex, age, treatment groups, and the presence of M. ovipneumoniae, B. trehalosi, M. haemolytica, and T. pyogenes) in LRT tissue samples, with the exception of Clostridium, which was more relatively abundant in LRT tissue samples from BHS than in those from AD (BH-corrected P < 0.05).

Host transcriptomics

Co-expression module identification

Filtered genes (n = 15,024) were utilized for module placement via WGCNA one-step network construction. Network construction resulted in 23 color-coded modules containing 14,924 genes; 100 genes were not conserved into any co-expression module and were subsequently placed into the “grey” module (Fig. 4; Supplementary Dataset 2). Across the 23 assigned modules, the turquoise module possessed the largest number of co-expressed genes (n = 6,562) and the darkturquoise module possessed the smallest number of co-expressed genes (n = 33); the average size of each module was 659 ± 1,296 genes. The complete list of genes and module assignments is found in Supplementary Dataset 2.

Fig. 4.

Cluster dendrogram of 15,024 filtered genes generated through dissimilarity metrics (1- Topological Overlap Matrix) and hierarchical clustering. The x-axis represents the retained genes and their associated module placement (Module colors) and the y-axis represents the co-expressional distance (Height) within each module by their correlation value

Module-trait correlations

Independent Kendall’s Tau correlation coefficients were generated within WGCNA for identifying and visualizing module-trait relationships (Fig. 5; Supplementary Dataset 3). Regarding host species (AD, coded “0”; BHS, coded “1”), three modules possessed significant correlations: midnightblue (τ = −0.72, P = 0.001), brown (τ = 0.68, P = 0.003), and lightyellow (τ = 0.59, P = 0.01). Regarding treatment group (MOP or wash), one module possessed a significant correlation: purple (τ = 0.48, P = 0.05). For M. ovipneumoniae PCR values (positive or negative), one module possessed a significant correlation: purple (τ = 0.50, P = 0.04). Regarding bacterial isolation results, several modules possessed significant correlations. Concerning Pasteurella multocida (Pm), four modules possessed significant negative correlations: purple (τ = −0.57, P = 0.02), brown (τ = −0.52, P = 0.03), black (τ = −0.57, P = 0.02), and green (τ = −0.59, P = 0.01). Concerning Bibersteinia trehalosi (Btreh), one module was found to have positive correlation: grey60 (τ = 0.48, P = 0.05). Regarding Mannheimia haemolytica (Mh), one module was identified with significant positive correlation: brown (τ = 0.62, P = 0.008). No significant module-trait correlations were identified for tissue type nor Trueperella pyogenes isolation.

Fig. 5.

Kendall’s Tau module-trait correlation heatmap. Correlation coefficients between each expressional module and metadata trait of interest were calculated between the first principal component of each module and each specified trait (Tissue (lung = 0, lymph node = 1), Species (AD = 0, BHS = 1), M. ovipneumoniae PCR (Mycoplasma ovipneumoniae PCR results; negative = 0, positive = 1), Pm (Pasteurella multocida culture results; negative = 0, positive = 1), Btreh (P Bibersteinia trehalosi culture results; negative = 0, positive = 1), Mh (Mannheimia haemolytica culture results; negative = 0, positive = 1), Tpyo (Trueperella pyogenes culture results; negative = 0, positive = 1)). Scaling of yellow and purple colors indicate Kendall’s Tau correlation coefficients, which can range from 1 to −1, respectively, reflecting both the strength and direction of association between module expression and each trait

Regarding host species (AD, coded “0”; BHS, coded “1”), three modules possessed significant correlations: midnightblue (τ = −0.72, P = 0.001), brown (τ = 0.68, P = 0.003), and lightyellow (τ = 0.59, P = 0.01). Regarding treatment group (MOP or wash), one module possessed a significant correlation: purple (τ = 0.48, P = 0.05). For MOP PCR values (positive or negative), one module possessed a significant correlation: purple (τ = 0.50, P = 0.04). Regarding bacterial isolation results, several modules possessed significant correlations. Concerning Pasteurella multocida (Pm), four modules possessed significant negative correlations: purple (τ = −0.57, P = 0.02), brown (τ = −0.52, P = 0.03), black (τ = −0.57, P = 0.02), and green (τ = −0.59, P = 0.01). Concerning Bibersteinia trehalosi (Btreh), one module was found to have positive correlation: grey60 (τ = 0.48, P = 0.05). Regarding Mannheimia haemolytica (Mh), one module was identified with significant positive correlation: brown (τ = 0.62, P = 0.008). No significant module-trait correlations were identified for tissue type nor Trueperella pyogenes isolation.

Functional enrichment analyses

The significantly enriched pathways for the seven aforementioned modules (midnightblue, brown, lightyellow, purple, black, green, and grey60), are found in Supplementary Dataset 2. The black module enriched for two pathways, both related to metabolism. The brown module enriched for 191 pathways, primarily related to ribosomal processing of RNA, cellular metabolism, antigen processing and cross-presentation, neutrophil degranulation, cell response to stress and external stimuli, interleukin-1 and interleukin-12 signaling, B-cell NF-κB activation and downstream signaling, and response to infectious disease. The green module enriched for 38 pathways, primarily related to cellular chaperoning, RNA processing and metabolism, CLEC7A signaling, and NF-κB activation. The grey60 module enriched for two pathways related to viral infection and translation events. The lightyellow module enriched for 19 pathways, primarily related to oxidative phosphorylation, ATP synthesis, and cellular activity involving neurons. The midnightblue module enriched for eight pathways, primarily related to antigen processing and presentation, heat shock protein response, and attenuation of heat shock transcriptional response. The purple module enriched for 127 pathways, primarily related to tumor necrosis factor α, mitogen-activated protein kinase, and NF-κB signaling, multiple toll-like receptor cascades, interleukin-1, 4, 10, 17, and 18 signaling, integrin-mediated cellular adhesion, T-cell receptor activity and signaling, inflammatory response, and transforming growth factor β receptor signaling.

Bacterial hub genes and predicted protein–protein interactions

The WGCNA-based hub gene analyses resulted in the identification of 1,373 total hub genes as follows: MMblack-Pm: 234, MMbrown-Mh: 231, MMbrown-Pm: 332, MMgreen-Pm: 354, MMgrey60-Btreh: 21, MMpurple-Pm: 136, MMpurple-Movi: 65 (Supplementary Dataset 4). As the correlations with Pm and associated co-expression modules were all independently in the same direction (negative), these hub genes were combined for protein–protein interaction networking (n = 1,065). From the Pm-associated hub genes, 186 gene products were identified to have predicted physical interactions (Fig. 6). The physical interaction network demonstrated high interconnectivity between the eight distinct clusters, with a median cluster size of 16 gene products. From the Mh-associated hub genes from the brown module, 64 gene products were identified to have predicted physical interactions (Fig. 7). The physical interaction network demonstrated high interconnectivity between the six distinct clusters, with a median cluster size of 9 gene products. Of particular interest, the hub genes identified within MMpurple-Pm and MMpurple-Movi analyses shared complete overlap (n = 65). Here, 13 gene products were identified to have predicted physical interactions (Fig. 8). The physical interaction network demonstrated low interconnectivity between the three distinct clusters, with only one gene product, ACTN4 (Cluster 3), demonstrating unique identity compared to the other 12 gene products. All gene products and their interconnectivity scores within each network analysis are found in Supplementary Dataset 5.

Fig. 6.

Protein–protein interaction network of interconnected hub genes associated with Pasteurella multocida culture results from the lungs and respiratory lymph nodes. Hub genes identified from the black, brown, green, and purple modules (n^total = 1,065) were combined to generate an overlapping protein–protein interaction network. Edges (lines) between each node (gene product) represent a predicted interaction between two products (minimum interaction score of 0.900). Colors within each node represent the predicted k-means cluster based on interaction scores

Fig. 7.

Protein–protein interaction network of interconnected Mannheimia haemolytica-associated hub genes found in the brown module (n = 64). Edges (lines) between each node (gene product) represent a predicted interaction between two products (minimum interaction score of 0.400). Colors within each node represent the predicted k-means cluster based on interaction scores

Fig. 8.

Protein–protein interaction network of interconnected Mycoplasma ovipneumoniae and Pasteurella multocida-associated hub genes found in the purple module (n = 13). Edges (lines) between each node (gene product) represent a predicted interaction between two products (minimum interaction score of 0.400). Colors within each node represent the predicted k-means cluster based on interaction scores

Discussion

Microbiome

Associations with diversity and differential abundance in the upper and lower respiratory tracts

Bacterial diversity of the upper respiratory tract (tonsillar swabs) was significantly impacted by the treatment group (MOP vs wash groups) and detection of M. ovipneumoniae and associated antibodies, as well as M. haemolytica detection. The increasing microbiome diversity associated with M. haemolytica detection may underly the treatment group effect given that the inoculum preparation technique was designed to either retain or eliminate (leukotoxigenic) Pasteurellaceae detected in DS nasal washings [50]. Considering the documented synergistic role of M. haemolytica and M. ovipneumoniae in clinical disease progression [29, 51], this pattern is not surprising. The importance of this synergism is highlighted by similar bacterial diversity values in control and MOP BHS. Notably, bacterial diversity and composition were both impacted by M. ovipneumoniae detection, whose mucociliary disruption may have allowed normally excluded bacterial species to persist in the upper respiratory tract [30, 31]. Other studies have found that the effects of B. trehalosi on the microbiome depends on the presence [52] or absence [53] of M. haemolytica. Thus, it is reasonable to consider that the bacterial compositional changes driven by B. trehalosi were, at least in part, related to the interactions between M. ovipneumoniae and M. haemolytica stemming from the two inocula used in this study. Importantly, our sampling design did not allow us to determine the Pasteurellaceae source species of leukotoxin-A detected in DS nasal washes, though M. haemolytica is thought to possess a comparatively higher leukotoxin-A gene carriage rate than B trehalosi [54].

The relative abundance of approximately 40% of the bacterial genera shifted in M. ovipneumoniae qPCR positive tonsillar swabs, including clinically significant respiratory bacterial genera (Bibersteinia, Providencia, Mannheimia, and Pasteurella). This observation coupled with the finding of higher evenness (that is, less bacterial genera dominating the microbiome composition) in those tonsillar swabs, suggests profound changes in the upper respiratory microbiome when M. ovipneumoniae is present. Interestingly, the relative abundance of Mannheimia significantly decreased when M. ovipneumoniae was detected in opposition to the expected synergism. This finding warrants further research to elucidate the nuanced interactions between these two bacteria. Furthermore, our study was unable to obtain sufficient sequencing reads to detect significantly differentially abundant bacterial genera in LRT samples on the basis of M. ovipneumoniae despite its role in LRT colonization and subsequent microbiome shifts [55]. Recent work further illustrates the increasingly recognized importance of integrated microbiome shifts across organ systems, and their role in pathogenesis, specifically in DS exposed to M. ovipneumoniae [56].

Host species is a major determinant of the establishment and maintenance of a particular microbiome [57]. Therefore, it is expected that microbiome shifts, and the potential role of clinically significant respiratory bacteria vary between different host species. Previous efforts to describe microbiome shifts important to BHS disease progression identified Providencia, Rosemonas, and Moraxella as important determinants of disease outcome in BHS [58]. Bighorn sheep in the present study demonstrated higher abundances of Roseomonas compared to AD in the tonsillar swabs, like the previous work with free-ranging individuals [58].

In the LRT, microbiome diversity and composition were strongly impacted by host species, with AD exhibiting higher diversity in the LRT compared to BHS. Additionally, detection of clinically significant respiratory bacteria (M. ovipneumoniae, P. multocida, and M. haemolytica) was associated with differences in bacterial composition. Just as in the upper respiratory tract, the interactions between M. ovipneumoniae and M. haemolytica seem to drive microbiome responses in the LRT. The relative abundance of opportunistic pathogens typically acquired from the environment (e.g. Clostridium) was also significantly higher in BHS than AD. Taken together, there is evidence to suggest that AD and BHS naturally or experimentally exposed to M. ovipneumoniae and other clinically significant respiratory bacteria have distinct changes in the microbial populations in the respiratory environment, possibly stemming from distinct pathobiological responses associated with host–pathogen co-evolution. Microbiomes of BHS appeared more susceptible to microbiome destabilization and pathogen-associated shifts, consistent with prior evidence of impaired mucociliary clearance and heightened susceptibility to leukotoxigenic bacterial toxins [33]. Conversely, AD exhibited microbiome responses suggestive of resilience or tolerance, potentially indicative of greater immunological or ecological compatibility with M. ovipneumoniae and perhaps some leukotoxigenic species/variants of Pasteurellaceae.

Host Transcriptomics

Metabolic pathway expression associated with Pasteurella multocida in both host species

Transcripts related to metabolic pathways (black module; Supplementary Dataset 3) were negatively correlated with the presence of Pasteurella multocida in the lungs of both AD and BHS. While the enriched pathways were relatively non-specific, studies evaluating the lung cells of both mice and goat lung cells have demonstrated that Pasteurella multocida induces a regulatory response related to cellular metabolic processes and may disrupt key signaling pathways that would promote host cell apoptosis during infection [59–61]. Our study corroborates these findings, as these two modules (black and brown; Supplementary Dataset 3; Supplementary Dataset 4) describe a similar relationship with cellular metabolism and NF-κB activation with the likelihood of isolating Pasteurella multocida from AD and BHS.

Possible in vivo bacterial inhibition of Mannheimia haemolytica and Pasteurella multocida mediated by cellular metabolic patterns of bighorn sheep

Interestingly, the relationship of expressional profiles from the brown module suggests an inverse relationship in Mannheimia haemolytica and Pasteurella multocida isolation specifically within BHS. Boukahil and Czuprynski (2018) identified a similar pattern from in vitro biofilm production on bovine bronchial epithelial cells, specifically where these two bacteria could inhibit each other’s biofilm development when in close proximity to each other on the epithelial surface [62]. Notably, Boukahil and Czuprynski (2018) did not establish a definitive antagonistic relationship between Mannheimia haemolytica and Pasteurella multocida, nor a mechanism by which they inhibit each other. Here, our study describes 139 genes within the brown module which are hub genes of both Mannheimia haemolytica and Pasteurella multocida (Supplementary Dataset 2; Supplementary Dataset 5), many of which are related to carbohydrate metabolism, mitochondrial translation activation, and G-protein chaperoning, thus serving as a possible mechanistic explanation to this phenomenon (Supplementary Dataset 2; Supplementary Dataset 4). Due to the clinical significance of these bacterial species in BHS pneumonia development, understanding host-specific responses to these agents upon exposure or co-exposure may highlight pathological differences related to ultimate disease outcomes.

Overlapping hub genes differentially associated with Pasteurella multocida and Mycoplasma ovipneumoniae

With respect to hub gene analyses, the further assessment of overlapping genes between modules, and the results of predictive protein–protein interactions, particular interest was given to those genes found within the purple module as it relates to Pasteurella multocida (negatively associated) and Mycoplasma ovipneumoniae (positively associated) detection. Of particular interest, ABL1, ACTN4, ILK, CD46, COL4A1, COL4A2, and NFKBIZ gene products have been shown to be directly involved in integrin regulation and signaling [63–68]. Further, CD46 has been shown to be a key receptor in downregulating complement activation, interacts with several β1 integrins, and may serve as a key binding receptor for bacterial pathogens [68–71]. Regarding infectious respiratory disease, integrins may serve as mediators of immune cell proliferation and molecular targets exploited by bacteria for cellular adhesion and immunological evasion [72]. Mycoplasma hyopneumoniae has been shown to interact with integrin β1-fibronectin to enter host cells and evade the immune system [73]. Similarly, Pasteurella multocida can produce a fibronectin-binding protein which interacts with immobilized fibronectin and type I collagen of host cells to promote colonization and invasion of host tissues [74]. Studies have demonstrated the paradoxical nature of integrin promotion and expression in pulmonary disease, as integrin-dependent mechanisms facilitate neutrophil migration and enhance macrophage-mediated clearance of bacteria, but persistent or dysregulated integrin activation may exacerbate tissue damage through sustained inflammation and bacterial infiltration or evasion [75–80].

Host species differences in immunological and neuroimmune-axis expressional profiles

Our work revealed three distinct expressional modules associated with species-level differences between AD and BHS, offering insights into potential mechanisms underlying susceptibility to infectious bacterial diseases. Three modules – midnightblue, brown, and lightyellow – exhibited significant correlations with species identity, highlighting divergent immunometabolic profiles that may influence host–pathogen interactions. Expressional profiles for pathways related to antigen processing and presentation, heat shock protein (HSP) responses, and attenuation of HSP transcriptional activity were depressed in BHS compared to AD. HSPs are critical in modulating immune responses during pathogen-induced stress and inflammation [81, 82]. Importantly, HSP70 and HSP90 proteins aid the host by protecting secreting intracellular proteins from proteolysis, support antigen presentation and immunoglobulin biosynthesis, and aid in cellular stabilization during infectious processes and/or heat stress [83–86]. In contrast, BHS had enriched immune pathways including interleukin-1 and interleukin-12 signaling, neutrophil degranulation, and B-cell NF-κB activation; these pathways are critical for initiating and sustaining inflammatory responses and adaptive immunity [87–90]. Bighorn sheep, which appear more susceptible to Mycoplasma-associated pneumonia and Pasteurella-related pleuropneumonia than AD, may mount a heightened but potentially dysregulated inflammatory response that exacerbates tissue damage during infection [38, 91, 92]. Similar immune and inflammatory profiles have been observed in domestic livestock species during the clinical onset of respiratory disease [93–96].

Bighorn sheep expression patterns were enriched for oxidative phosphorylation, ATP synthesis, and neuronally associated cellular activity. These findings suggest increased metabolic activity and possibly heightened neuroimmune regulation in BHS tissues. Enhanced mitochondrial activity has been associated with both immune activation and pathogenic oxidative stress [97–101]. In the context of infectious respiratory disease in livestock, excessive reactive oxygen species (ROS) production, a byproduct of elevated oxidative phosphorylation, can damage pulmonary tissue [102–104]. Moreover, the bidirectional relationship between peripheral neuronal activity and immune cell activation, termed the neuroimmune axis, has been implemented in airway diseases and cellular stress [105–108]. Here, it may be that BHS exhibit distinct tissue-level responses involving stress or inflammation-induced neurogenic pathways compared to AD.

Transcriptomic implications for species-level disease susceptibility

These interspecies differences in co-expressed immune and metabolic modules may suggest that BHS rely on a more acute, inflammatory-based immune activation strategy, potentially leading to immunopathologies, while AD may benefit from a more stress-tolerant antigen processing system. This could partially explain observed epidemiological patterns where BHS populations suffer high mortality from bacterial pneumonia, especially following Mycoplasma spp. and Pasteurella spp. co-infections, whereas AD appear less impacted under similar pathogen exposure scenarios [38, 91, 92]. The more tempered transcriptomic responses in AD may be due to host–pathogen co-accommodation, wherein host responses are expected to become less severe when infections occur in species with pathogen exposures for a given host species repeated through evolutionary time [6, 8, 11]. Such patterns reinforce the role of differential host–pathogen co-evolution in governing disease susceptibility and highlight the potential consequences of pathogen spillovers from nonindigenous hosts to immunologically and evolutionarily naïve wildlife species. Further validation in tissue-specific expression contexts and functional assays is warranted, but these transcriptional signatures provide a valuable framework for understanding disease susceptibility differences between native and nonindigenous Caprinae and inform conservation and disease management strategies.

Limitations and considerations

As with many studies focused on free-ranging wildlife in captive settings, our study was met with challenges stemming from low sample sizes and lack of knowledge on pre-study pathogen exposure patterns in our study animals. These limitations introduce the possibility that observed patterns may have been impacted not only by species, sex, and treatment group, but also by inherent interindividual variations, differences in pathogen exposure histories, and differences in the duration of pathogen exposure between host species. Due to animal facility and animal procurement considerations, our design did not control confounders, particularly acclimatization period and pathogen exposure period differences between groups (BHS, 40 days; AD, 0 days). Additionally, the pathogenesis of M. ovipneumoniae-induced respiratory disease and subsequent transmission dynamics in both BHS and AD can be influenced by a variety of factors, including strain composition [25, 38]. This further limited our ability to ascertain the precise pathological stage of each individual relative to our microbiome and host transcriptomic findings. Finally, failure to collect nasal and tonsillar swab samples from each species and treatment group may further limit the generalizability of our results. A true multi-omics approach requires paired swab and tissue samples across the study period, which was not feasible given the animal handling requirements for these species in captivity. Our use of pooled lower respiratory tract samples for microbiome and transcriptomic analyses prevented us from formally assessing the full set of possible differences in host responses to infection across individual tissue types. Overall, we were limited by commonplace issues associated with wildlife research that warrant interpretive caution.

Conclusions

Natural and anthropogenic changes in the communities of host species, as well as the introduction and movement of pathogens, represent significant global biotic changes [5]. Here, we use the introduction of AD to the southwestern United States as a case study to elucidate potential outcomes of exposure to shared pathogens, and the potential reliance of these outcomes on differential host–pathogen co-evolutionary histories. With numerous mammalian introductions globally altering host–pathogen dynamics, understanding co-evolutionary relationships become critical for predicting disease emergence and transmission. Future work should prioritize integrating microbiome and expressional profiling from free-ranging individuals with ongoing epidemiological surveillance to better assess and manage potentially emerging disease threats at both a population and host community level.

Significant clinical differences exist between DS and BHS when exposed to M. ovipneumoniae and leukotoxigenic Pasteurellaceae [109, 110]. These host species genetically and biogeographically diverged from one another between 2.52 and 5.63 million years ago [111, 112], allowing much opportunity for different host–pathogen co-evolution patterns to occur. Indeed, multiple signatures consistent with host–pathogen co-evolution are present in DS and absent in BHS [26, 33]. While the co-evolutionary history of AD with M. ovipnuemoniae and leukotoxigenic Pasteurellacea is not known, assessing both respiratory microbiome and host transcriptomic profiles from these two host species (AD and BHS) allowed us to assess whether and how they differ when exposed to pathogens lethal to BHS. When combined with clinical data from known exposures of either the naïve or co-evolved host species to an agent of interest, these predictions could be improved. The applied nature of this host–pathogen system has implications for BHS health and epidemiological dynamics for M. ovipenumoniae in Texas, and elsewhere BHS and AD co-occur.

Our focus is to extend findings from this host–pathogen system to novel, emerging, and existing host–pathogen systems perturbed by anthropogenic activities. Characterizing functionally distinct processes for the microbiomes and transcriptomes for host species with (co-evolved) and without (naïve) co-evolution with a given agent presents a unique opportunity to describe how other altered host–pathogen systems may behave. Together, our microbiome and transcriptomic findings highlight complex host–pathogen-microbiome interactions shaped by evolutionary history and species-specific immunometabolic and microbiome responses. Such patterns reinforce the role of differential host–pathogen co-evolution in governing disease susceptibility and highlight the potential consequences of pathogen spillover from nonindigenous hosts to immunologically naïve wildlife species.

Authors’ contributions

WC was the senior project lead, procured funding, managed and conceptualized the project, and directed the scope of the study and prepared the manuscript; CP created the bioinformatic pipeline used to assess the microbiome data, and performed statistical analyses related to microbiome efforts; MS conceptualized and led the transcriptome efforts, performed all transcriptome statistical analyses, and prepared the manuscript; LT collected samples, conducted animal husbandry and handling, designed and conceptualized the study, procured funding, provided input on statistical analyses with coauthor assistance, guided the overall scope of the manuscript in its preparation, and managed tasks among coauthors; RV-C conceptualized and led the microbiome efforts, finalized all microbiome statistical analyses, and prepared the manuscript.

Funding

This study was funded by the Foundation for North American Wild Sheep, Grant-In-Aid Project Number 1920–71 (Walter E. Cook.) and the Texas Parks and Wildlife Department’s Wildlife Research Program through Federal Aid in Wildlife Restoration Act (Pittman-Robertson) Grant Program (“Evaluating the risk of Mycoplasma transmission from Aoudad to Bighorn Sheep”; TX W-207-R-1 F20AF10359-00). The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data availability

The datasets generated and/or analyzed during the current study are available as supplementary data files and/or data repository. All raw transcriptomic sequencing data and curated metadata produced by this study are available at the National Center for Biotechnology Information Gene Expression Omnibus (NCBI-GEO) under the accession number GSE295667 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE295667). All raw microbiome sequencing data and curated metadata produced by this study are available at the National Center for Biotechnology Information Sequence Read Archive (SRA) under the accession number PRJNA1306317 (https://www.ncbi.nlm.nih.gov/sra/?term=PRJNA1306317).

Declarations

Ethics approval and consent to participate

All work was authorized under the A&M AgriLife Animal Care and Use Committee permit 2020-038A. Reporting of methods, results, and study limitations complies with the ARRIVE 2.0 Essential 10 guidelines.

Consent for publication

Not applicable.

Competing interests

The authors declare that they have no competing interests.