Causes of Pneumonia Epizootics among Bighorn Sheep, Western United States, 2008–2010

Mycoplasma ovipneumoniae is a primary pathogen.

I n North America, epizootic pneumonia is a devastating, population-limiting disease of bighorn sheep (Ovis canadensis) (1)(2)(3)(4)(5). Anecdotal and experimental evidence suggests that in at least some instances, this disease may be introduced into bighorn sheep populations by contact with domestic sheep or goats (5,6). When the disease is fi rst introduced, outbreaks affect animals of all ages (1)(2)(3). During subsequent years or decades, sporadic cases of pneumonia in adult sheep and annual epizootics of pneumonia in lambs may continue (7)(8)(9)(10).
Considering the dramatic and severe character of epizootic bighorn sheep pneumonia, the etiology is surprisingly unclear. Findings of gross and histopathologic examinations of lung tissue strongly suggest bacterial etiology: anterior-ventral distribution, suppurative infl ammation, and abundant bacterial colonies. In domestic ruminants, bacterial pneumonia frequently occurs secondary to viral infections or other pulmonary insults, but extensive efforts to detect such underlying factors for bighorn sheep pneumonia have generally been nonproductive. For example, although evidence of infection or exposure to respiratory viruses, especially respiratory syncytial virus and parainfl uenza virus, is frequently found in healthy and pneumonia-affected populations, no consistent association between the disease and any virus has been found (11)(12)(13). As a result, most research attention has been directed toward bacterial respiratory pathogens that may act as primary infectious agents, particularly leukotoxinexpressing Mannheimia haemolytica, which is highly lethal to bighorn sheep after experimental challenge (5,14). Other Pasteurellaceae, particularly Bibersteinia trehalosi and Pasteurella multocida, have been more frequently isolated from pneumonia-affected animals during natural outbreaks than has M. haemolytica (11,12,15). Another candidate pathogen, Mycoplasma ovipneumoniae, has recently been isolated from pneumonia-affected bighorn sheep during 2 epizootics (11,16,17); antibodies against this agent were detected in bighorn sheep from 9 populations undergoing

Causes of Pneumonia Epizootics among Bighorn Sheep, Western
United States, 2008-2010 pneumonia epizootics but were absent in 9 nonaffected populations (17). In experiments, M. ovipneumoniae has been shown to predispose bighorn sheep to M. haemolytica pneumonia (18). When M. ovipneumoniae-free domestic sheep were commingled with bighorn sheep, the bighorn sheep survived at unprecedented rates (19). Development of effective methods for managing, preventing, or treating an infectious disease requires a clear understanding of its underlying etiology. However, clarifying the etiology can be diffi cult, particularly for primary infections (e.g., HIV) that are characteristically associated with multiple opportunistic infections that may be more lethal than the epidemic agent itself. During 2008-2010, epizootic pneumonia of bighorn sheep was detected in at least 5 western US states. These epizootics provided an opportunity to conduct a comparative study of the etiology of this disease (Table 1).
Conventional microbiological methods can fail to isolate agents because of their fastidious in vitro growth requirements or intermicrobial interactions; thus, for agent isolation, we used 2 culture-independent methods (agentspecifi c PCRs and 16S clone libraries) in addition to conventional bacterial cultures (17,(20)(21)(22). We expected that primary etiologic agents could be differentiated from opportunistic agents by 1) their detection at high prevalence in affected animals, 2) the presence of single (clonal) strain types within each outbreak, and 3) their uncommon or lack of detection in animals from healthy populations (11,(22)(23)(24). Therefore, to clarify the etiology of epizootic pneumonia, we applied these criteria to the bacterial respiratory pathogens detected in multiple bighorn sheep epizootics.

Bighorn Sheep Populations
The study sample consisted of 8 demographically independent bighorn sheep populations in 5 states (Montana, Nevada, Washington, Oregon, and South Dakota) that had been affected by epizootic pneumonia during 2008-2010 and for which lung tissue specimens from >4 affected animals were available (Table 1). In 6 of these populations, the disease affected bighorn sheep of all ages; in the other 2 populations, in which the disease had previously affected sheep of all ages, the disease was restricted to lambs. Convenience samples were selected among those available from each epizootic: the sample of pneumonia-affected animals consisted of the fi rst 4-6 sheep for which pneumonia had been confi rmed by gross or microscopic lesions. Sheep initially selected for analysis but later determined to have lacked gross or microscopic lesions characteristic of pneumonia were retained in the study but analyzed separately. Negative controls consisted of animals with no gross or histopathologic evidence of pneumonia that died or were culled from 2 closely observed healthy populations.

Bacteriologic Cultures
Surfaces of affected lung tissue specimens were seared, and swab samples of deeper tissues were obtained and streaked onto Columbia blood agar plates (Hardy Diagnostics, Santa Maria, CA, USA). Pasteurellaceae were isolated and identifi ed by using routine methods (25) and then stored at −80°C in 30% buffered glycerol in brainheart infusion agar (Hardy Diagnostics).

DNA Template Preparation
DNA was extracted from 1.0-1.5 mL of fl uid collected from 1-2 g of fresh-frozen lung tissue macerated in 1 mL of phosphate buffered saline for 5 min by using a stomacher (Seward Stomacher 80 Laboratory Blender, Bohemia, NY, USA). DNA was extracted by using a QIAamp mini kit (QIAGEN, Valencia, CA, USA) according to the manufacturer's protocol.

PCR Detection of Respiratory Pathogens and lktA
To detect M. haemolytica, P. multocida, B. trehalosi, lktA, and M. ovipneumoniae, we used previously published

16S Analyses
To detect predominant microbial populations in the pneumonic lung tissues, we used a culture-independent method (17). In brief, we aseptically collected two 1-g samples of lung tissue from sites at least 10 cm apart in grossly abnormal (consolidated) tissue from 16 pneumoniaaffected animals, including 2 from each outbreak. Tissues were stomached and DNA was extracted (DNeasy Blood and Tissue Kit; QIAGEN) from 100-μg aliquots of each homogenate. Segments of 16S rDNA were PCR amplifi ed and cloned. Insert DNA was sequenced (vector primers T3 and M13, BigDye version 3.1, ABI PRISM Genetic Analyzer; Applied Biosystems, Foster City, CA, USA) from 16 clones derived from each homogenate, resulting in 32 sequences from each animal. DNA sequences were assigned to species (>99% identity) or genus (>97% identity) according to BLASTN GenBank search results (29). Clone sequences may be accessed in GenBank under accession nos. JN857366-857894.

Pulsed-Field Gel Electrophoresis for Pasteurellaceae
When available, Pasteurellaceae isolated from the study animals were obtained from the Washington Animal Disease Diagnostic Laboratory (Pullman, WA, USA). If such isolates were unavailable, we substituted banked isolates from other bighorn sheep involved in the same outbreaks. Isolates were subjected to pulsed-fi eld gel electrophoresis (PFGE) performed on a CHEF-DRII PFGE apparatus (Bio-Rad, Hercules, CA, USA) in 1% agarose gel (Seakem Gold Agarose; FMC Bio Products, Rockland, MD, USA) in 0.5× Tris borate EDTA buffer at 14°C for 20 h at 6 V/cm and a linear ramp of 1.0-30.0 s for ApaI or 0.5-40.0 s for SmaI. Salmonella serovar Braenderup H9812, digested with XbaI for 3 h at 37°C, was used as a size standard on each gel. Gels were stained with ethidium bromide and photographed under UV transillumination. PFGE data were analyzed by using BioNumerics version 4.6 (www.applied-maths.com/bionumerics/bionumerics. htm) with Dice coeffi cients and the unpaired pair group method with arithmetic means for clustering; tolerance and optimization parameters were set at 1%.

Intergenic Spacer Region Sequence typing for M. ovipneumoniae
In a preliminary study performed in our laboratory, ribosomal operon intergenic spacer (IGS) regions of M. ovipneumoniae from isolates from 6 bighorn sheep populations were amplifi ed by using the method described by Kong et al. (30) and sequenced (Amplicon Express, Pullman, WA, USA). Sequences were aligned and clustered by using Lasergene software (DNASTAR, Inc., Madison WI, USA). Each isolate exhibited a different IGS sequence, demonstrating the utility of IGS sequences for identifying strain diversity (data not shown). We used Primer3 software (http://frodo.wi.mit.edu/primer3/) to develop primers fl anking the variable IGS region, conserved among M. ovipneumoniae isolates, and divergent from IGS regions of the second most common sheep upper respiratory mycoplasma, M. arginini (Table 2). IGS PCR products were sequenced, and sequences were aligned and

Statistical Analyses
To evaluate the agreement between results of bacteriologic cultures and PCR tests for detection of P. multocida, M. haemolytica, and B. trehalosi, we used Cohen κ coeffi cients (31). To assess overall differences in prevalence of specifi c bacterial respiratory pathogens, we used χ 2 tests; for pairwise comparisons, we used the Marascuilo procedure (32) to control for multiple comparison problems, which might affect error rates. To assess associations between prevalence of different respiratory bacteria and mortality rates among different bighorn sheep populations, we used the Pearson correlation coeffi cient.

Results
We detected 4 previously reported bacterial respiratory pathogens of bighorn sheep. We detected M. haemolytica, B. trehalosi, and P. multocida by using aerobic culture and species-specifi c PCR and M. ovipneumoniae by using PCR alone (20) (Table 3; online Appendix Table, wwwnc. cdc.gov/EID/article/18/1/11-1554-TA1.htm). Agreement between detection by culture and PCR varied by agent, ranging from no agreement (M. haemolytica, κ −0.02), to fair agreement (B. trehalosi, κ 0.22), to good agreement (P. multocida, κ 0.76). For the purposes of the following analyses, animals for which any agent was detected by either method were considered positive for that agent. Among the targeted agents, 3 (B. trehalosi, M. haemolytica, and M. ovipneumoniae) were detected in >1 animals from all 8 outbreak-affected populations and 1 (P. multocida) was detected in animals from 5 outbreak-affected populations ( Table 3).
Frequency of detecting B. trehalosi and P. multocida differed signifi cantly among outbreaks (p = 0.002 and 0.001, respectively). Similarly, PCR-based detection of lktA differed among outbreaks (p = 0.003). Although such differences could potentially contribute to the signifi cant differences in disease severity and mortality rates among the epizootics in this study (χ 2 184.7, 7 df, p<0.0001), the prevalence of B. trehalosi, P. multocida, or lktA did not correlate with estimated mortality rates in the 8 outbreaks included in this study (Tables 1, 3).
Strain typing to evaluate the genetic similarity of bacterial pathogens within and among outbreaks (23) detected only single IGS types of M. ovipneumoniae within each outbreak, whereas distinctly different IGS types were found for each epizootic with the exception of 2 populations in Montana (Figure). In contrast, the PFGE strain types of Pasteurellaceae isolated from within single outbreaks ranged from 0 to 7, including 0-7 B. trehalosi strains and 0-2 P. multocida strains (Table 4; online Appendix Table). Assessment of strain type diversity of M. haemolytica within outbreaks was not possible because this species was isolated only 1 time.
Among the agents and genes tested, M. ovipneumoniae was the only agent or gene that was detected at different frequencies for animals from epizootic-affected compared with non-epizootic-affected populations (Table 3; p<0.001). The frequency of M. ovipneumoniae and P. multocida detection in non-pneumonia-affected animals culled from epizootic populations was intermediate, signifi cantly lower than that in pneumonia-affected animals (p<0.01).

Discussion
The results of this study support the hypothesis that M. ovipneumoniae is a primary agent in the etiology of epizootic bighorn sheep pneumonia in populations across the western United States and that it acts to induce secondary infection with opportunistic pathogens. M. ovipneumoniae was detected in the pneumonic lungs of >95% of study animals involved in the 8 discrete pneumonia epizootics, signifi cantly more frequently than any of the other respiratory agents sought except the bighorn sheep commensal bacterium B. trehalosi (34,35). We identifi ed identical ribosomal IGS strains of M. ovipneumoniae within the affected animals in each outbreak, consistent with epizootic spread (24); M. ovipneumoniae was not detected in the healthy populations sampled. Of note, the 2 populations in which identical IGS strains of M. ovipneumoniae were detected were separated by only ≈20 miles, suggesting the possibility that this strain was transmitted among these populations by movement of >1 M. ovipneumoniae-infected bighorn sheep.
The normal host range of M. ovipneumoniae (members of Old World Caprinae, including domestic sheep and moufl on, a closely related Eurasian sheep species) is consistent with many observations that epizootic bighorn sheep pneumonia frequently follows contact with these hosts (5,19). Previous experiments in which bighorn sheep were commingled with domestic sheep or moufl on each resulted in epizootic bighorn sheep pneumonia and, cumulatively, the death of 88 (98%) of 90 bighorn sheep; similar commingling experiments with other ungulates (deer, elk, llamas, horses, cattle, goats, mountain goats) resulted at most in sporadic deaths from bighorn sheep pneumonia (4 [7%] of 56) (19). In a recent study in which bighorn sheep were commingled with M. ovipneumoniaefree domestic sheep, the lack of epizootic bighorn sheep pneumonia was unprecedented (19). Together, these data support the hypothesis that bighorn sheep epizootic pneumonia results from cross-species transmission of M. ovipneumoniae from its normal host(s) to a naive, highly susceptible host: bighorn sheep.
Each of the other specifi c potential respiratory pathogens targeted failed to fulfi ll >1 expectations for a primary etiologic agent. B. trehalosi was detected in most animals regardless of their health status; exhibited diverse strain types within epizootics; and in most instances was detected in the absence of lktA, consistent with the nontoxigenic strains widely distributed in healthy and pneumonic bighorn sheep (36). M. haemolytica was similarly detected in about half of the animals regardless of their health status and in the absence of lktA. P. multocida was not detected at all in animals involved in 3 of the epizootics, but in those outbreaks in which it was present, it was detected at high prevalence and somewhat more frequently in pneumonia-affected than in healthy bighorn sheep. Furthermore, multiple isolates from those epizootics in which it was detected shared a high degree of genetic similarity, consistent with epizootic transmission (24). The frequencies with which B. trehalosi, P. multocida, and lktA were detected from animals in the different epizootics differed signifi cantly, although this fi nding did not correlate with mortality rates (Table 1). This conclusion is limited, however, by the possible confounding effect of the extensive culling conducted in several areas of the epizootics examined in this study. More research into factors that affect the severity of bighorn sheep pneumonia epizootics is clearly needed.
The analysis of prevalence of bacterial respiratory pathogens in the lung tissues of healthy animals from unaffected populations was comparatively limited by the small number of control specimens available. To more clearly defi ne the prevalence, infectivity, and virulence of M. ovipneumoniae, sampling of additional healthy bighorn sheep populations is needed. Although M. ovipneumoniae was not detected in the negative control animals examined in this study and although serologic evidence of exposure to M. ovipneumoniae is uncommon or rare in healthy bighorn sheep populations (17), several apparently healthy bighorn sheep populations with serologic and/or PCR evidence of exposure to M. ovipneumoniae have been identifi ed (data not shown). This fi nding demonstrates that not all exposures to this agent result in epizootic bronchopneumonia or, perhaps, that unrecognized previous epizootics had occurred. To clarify these fi ndings, more research, specifi cally including longitudinal observational studies and investigation of strain differences in virulence of M. ovipneumoniae (37,38), is needed.
Our universal eubacterial 16S rDNA approach used analysis of small clone libraries from each animal to detect those agents representing >10% of the 16S operons in lung tissue with high (>95%) confi dence. The 3 most frequently detected aerobic bacterial agents detected by using this method were P. multocida, M. ovipneumoniae,   Table). Furthermore, the lung tissues from animals affected by the 5 epizootics in Washington or Montana were obtained from bighorn sheep that were coughing and culled in an attempt to prevent further transmission of the disease; therefore, these specimens could represent animals at earlier stages of the disease when more consistent presence of causal agents would be expected. Consistent with previous reports of bighorn sheep in Hells Canyon (17), the predominance of obligate anaerobes (Fusobacterium, Prevotella, Clostridium, and Bacteroides spp.) among the lung fl ora was consistent with decreased clearance of inhaled oral fl ora from the lower respiratory tract. Impaired clearance of inhaled fl ora is expected subsequent to infection by virulent M. ovipneumoniae (38) or by leukotoxin-expressing Pasteurellaceae (39), albeit by different mechanisms.
To our knowledge, only 1 other study of epizootic bighorn sheep pneumonia has reported comparative microbiological fi ndings from pneumonia-affected animals involved in multiple discrete epizootics. Aune et al. (12) reported that Pasteurellaceae cultured from pneumoniaaffected animals differed somewhat among 4 bighorn sheep pneumonia epizootics in Montana during 1991-1996. P. multocida was isolated from pneumonic lung tissues of >1 animals during all 4 epizootics, although prevalence exceeded 50% during only 1 epizootic. Pasteurellaceae biotypes corresponding to B. trehalosi were isolated from animals involved in 3 of the 4 outbreaks, and Pasteurellaceae biotypes corresponding to M. haemolytica were isolated from animals in only 1 outbreak. The microbiology of epizootic pneumonia in Hells Canyon also has been described (11,15,40); results were broadly comparable to the conventional microbiology results reported here for Pasteurellaceae. All these studies differed from the study reported here in that the conventional microbiological methods used failed to recognize M. ovipneumoniae in affected lung tissues.
In summary, of the bacterial respiratory pathogens evaluated, M. ovipneumoniae was the only agent for which the data consistently met the criteria for a primary etiologic agent across all outbreaks. In contrast, the data were inconsistent with regard to a primary etiologic role for any Pasteurellaceae species. The likelihood of M. ovipneumoniae having a primary role in bighorn sheep pneumonia is consistent with the association between some epizootics of this disease and contact with domestic sheep because the latter carry this agent at high prevalence. Identifi cation of M. ovipneumoniae as the epizootic agent of bighorn sheep pneumonia may provide a useful focus for the development of specifi c preventative or therapeutic interventions.