Figure 1

Figure 1. Characteristics of an RRV outbreak among Australian Defence Force (ADF) personnel during and after training in Shoalwater Bay Training Area (SWBTA), northeastern Australia, 2016. A) RRV attack rates among 3 ADF...

The first EPA outbreak was reported after a 12-day exercise (February 29–March 11, 2016). Forty-four personnel from a combat unit of 128 sought care at the Regimental Aid Post (RAP) with rash, headache, nausea, fatigue, lethargy, and joint and muscle pain at the conclusion of the exercise. RRV IgM was detected in acute- or convalescent-phase serum samples collected from 28 of these 44 persons 14–50 days after symptom onset (Table); attack rates ranged from 22% (confirmed cases only, 28/128 [22%]) to 34% (all suspected cases, 44/128 [34%]) (Figure 1, panel A).

Troops from the affected unit (referred to hereafter as the combat unit) had spent 3 days in heavy rain near Camp Growl in the training area (Figure 1, panel B). During the following 3 days, the unit moved over varied terrain, including dense bushland, in hot and humid conditions to an urban operation training facility (UOTF). Once at the UOTF, they conducted urban assault maneuvers against a 36-member opposition force for an additional 5 days. Upon conclusion of the UOTF component of the exercise, the combat unit marched back to Camp Growl, where they spent the night before returning to Brisbane, ≈630 km south of SWBTA. An administration unit of 30 persons was stationed in SWBTA during the same period but remained at Camp Growl throughout the exercise. In addition to the 44 members of the combat unit affected by EPA, 1 member from the administration unit reported symptoms of EPA, and RRV IgM was detected in this patient’s serum sample upon return to Brisbane (attack rate 3.3%) (Figure 1, panel A). None of the 36 members of the opposition force reported symptoms of EPA (Figure 1, panel A). This localized EPA outbreak had 2 distinct epidemic curves, with gaps of 7 days and ≈15 days between the peaks (Figure 1, panel C).

In 2017, a total of 43 members from 3 different units had symptoms of EPA during an exercise conducted from late April to mid-May in the same area as the combat unit affected in the 2016 outbreak. Thirty-one of the 43 patients provided a serum sample. Of these, 25 samples contained RRV IgM (Table). Neither the advancing routes of these 3 units through SWBTA nor the epidemiologic timeline of the 2017 EPA outbreak could be obtained because of ADF operational restrictions. Additional EPA cases might have occurred during both outbreaks, given that anecdotal evidence suggests that members who were unwell at that time chose not to seek care at the RAP. ADF members were not screened for RRV seroconversions over the course of the exercise, so asymptomatic RRV infections could not be detected. Because most RRV infections are asymptomatic, the number of infections during these outbreaks was probably higher than the number of cases reported.

The most common symptoms experienced in the 2016 and 2017 outbreaks were, respectively, polyarthritis and muscle pain (71% and 88%), arthralgia (67% and 88%), fatigue (71% and 88%), loss of appetite (52% and 75%), stiff neck (38% and 71%), fever (33% and 47%), rash (9.5% and 65%), headache (28% and 41%), and sore throat (14% and 24%) (Table). Most personnel recovered within 4–6 weeks of symptom onset, except for 3 from the 2016 outbreak who were unfit for deployment 3 months after illness onset because of ongoing signs and symptoms.

All study participants were aware of the ADF policy of dipping uniforms and bed nets in permethrin, but none of the participants, nor their unit commanders, requested support from the Preventive Medicine Company to undertake this procedure before or during either exercise, citing time constraints (Table). More than half of the personnel failed to comply with the sleeves-down policy because of the hot and humid weather conditions. All participants stated that they used commercial repellents, which contain relatively low concentrations of N,N-Diethyl-meta-toluamide (DEET) compared with the repellents issued by the ADF. About 57% (2016) and 82% (2017) of personnel reported sleeping under bed nets at night but noted that nets at times were abandoned because of their incompatibility with the tactical situation. Five soldiers from the 2017 outbreak reported constantly being assailed with mosquito bites (concentrated around their hands and legs) during the night even when inside their nets. Gloves and socks were worn when sleeping to address this issue. All participants reported being bitten by mosquitoes on a regular basis.

Figure 2

Figure 2. Maximum clade credibility tree based on analysis of 58 complete Ross River virus E2 sequences (1,266 nt) from outbreaks among Australian Defence Force personnel during and after training in Shoalwater Bay...

Bayesian phylogenetic analyses of 58 RRV complete E2 sequences (1,266 nt) were performed by using TN93+Γ, HKY+Γ, and the GTR+Γ+I substitution models with both strict and relaxed clock models. All phylogenies placed the ADF RRV isolates collected in 2016 and 2017 (MIDI13.2016, MIDI4.2017, MIDI9.2017, and MIDI32.2017) into 2 distinct sublineages of lineage III (IIIE and IIIF). The TN93+Γ substitution model was the most highly ranked of those used, having the lowest Bayesian information criterion score (32–34) (Appendix Table 1), and coupled with the strict clock to produce the E2 phylogenetic tree (Figure 2). The remaining trees are provided in Appendix Figures 1–5.

Figure 3

Figure 3. Maximum clade credibility tree based on analysis of 32 complete Ross River virus nsP3 sequences (1,650 nt) from outbreaks among Australian Defence Force personnel during and after training in Shoalwater Bay...

Although the confidence level for the separation of the 2 sublineages differs depending on the model used (posterior probabilities range from 0.63 for HKY+Γ relaxed clock to 0.89 for GTR+Γ+I relaxed clock), the bifurcation of the 2 sublineages occurs in all of the trees produced (Figure 2; Appendix Figures 1–5). Furthermore, an analysis of 32 nsP3 sequences (1,650 nt) similarly demonstrates the bifurcation of the 2 proposed sublineages but with high posterior probability (≈1) for each of the HKY and GTR+Γ+I substitutional models with both strict and relaxed clocks. The tree for nsP3 based on the HKY substitution model with a strict clock (Figure 3) is also presented with alternative substitution models and clocks (Appendix Figures 6–8).

Collectively, these results suggest that different lineages of RRV were responsible for the 2016 and 2017 outbreaks. The 2016 isolate (MIDI13.2016) resembled strains of RRV recovered from patients in Brisbane (≈600 km south of SWBTA) in 2014 and 2015 and belongs to sublineage IIIE. Lineage III is thought to have evolved from 2 other lineages of RRV that are believed to be extinct. The 3 isolates recovered in 2017 (MIDI4.2017, MIDI9.2017, and MIDI32.2017) shared 99.8% nucleotide identity in pairwise comparisons and have been assigned to sublineage IIIF, which contains strains of RRV recovered from the east and west coasts of Australia in 2009 and in an isolate identified in Brisbane in 2015 (BNE 2885.2015).

We observed 6 amino acid differences between the 2016 (sublineage IIIE) and 2017 (sublineage IIIF) isolates from SWBTA (N132D, Y296H, T369A, I376M, A384T, and A389T) of which only 1 (I376M) was a conservative change. All substitutions are located in the A and C domains of the E2 protein, in areas involved in the interaction with other proteins (E1, capsid, and 6k), as well as in the process of budding of alphavirus envelope proteins from host cell membranes (35,36). The 2 glycosylation sites on the E2 protein were conserved in all 4 of the SWBTA isolates.

The state of Queensland, where SWBTA is situated, has the highest rates of EPA in Australia, consistently recording >1,000 cases each year (1). RRV-caused EPA cases are reported routinely in the areas surrounding Rockhampton (37), the nearest city to SWBTA. In 1997, nineteen RRV-caused EPA cases were reported among US Navy personnel during a joint Australia–US military exercise located at SWBTA (27). In 2004, RRV RNA was detected in multiple mosquito species collected in this training area (27,38). The most recent common ancestor of RRV (strain T48) originated from northeastern Australia (2,18), and this finding is in agreement with our phylogenetic analysis (Figure 2).

In light of these data, we reasonably believe that RRV has established its natural endemic cycle in SWBTA. The recurrent outbreaks of RRV infection within SWBTA and the intermittent presence of humans in this region suggests that the 2016 and 2017 infections were spillover events in which the virus was incidentally acquired by ADF soldiers from the natural endemic animal-mosquito-animal transmission cycle made possible by large populations of native animals (e.g., kangaroos and wallabies) and the wide variety of RRV vectors (21). However, phylogenetic analyses of nucleotide sequences of the E2 and nsP3 genes of RRV recovered from the ADF personnel in our study and from patients and mosquitoes from other parts of Australia do not support the notion of RRV strains being sequestered in endemic pockets across Australia. Instead, SWBTA isolates bore a remarkable similarity to those circulating in the wider community (Figures 2, 3; Appendix Figures 1–8). A more likely explanation is that strains of RRV are not regionally locked but spread around Australia by the movement of humans, as most likely occurred in the RRV outbreak in the Pacific during 1979–1980 (4–8). More surveillance and exhaustive genetic analysis might help to resolve the precise origin and nature of these outbreaks.

The spatiotemporal characteristics of the 2016 outbreak suggested that the virus was transmitted among soldiers by local mosquitoes after the initial infection. The first wave of soldiers who sought care at the RAP at Gallipoli Barracks, Brisbane, did so within a 2-week period of leaving the SWBTA, which is consistent with the 2–15-day incubation period proposed for RRV infection in humans (18) (Figure 1, panel C). Although this first wave of soldiers might have been infected with RRV by spillover from the suspected natural, endemic animal–mosquito–animal transmission cycle in SWBTA, few (if any) intermediate host-macropods exist in Gallipoli Barracks from which soldiers would become infected upon their return. Soldiers who sought care for EPA signs and symptoms 19–37 days after returning to the barracks are indicative of a potential secondary human-mosquito-human infection cycle in Gallipoli Barracks, possibly involving asymptomatic but viremic persons. This conclusion is further supported by the high level of homology shared by 3 RRV isolates from 2017.

No EPA cases were reported (except 1 case from the administration unit attached to the combat team at SWBTA) from other units at Gallipoli Barracks (a base containing >5,600 personnel who operate across 30 different units) upon the return of the infected personnel. This finding might simply reflect the probability of an extremely small number of infected mosquitoes finding a susceptible and unprotected human host among other units headquartered at discrete localities spread over an area of >200 hectares. However, without concurrent isolation of the virus from local mosquitoes in the vicinity of Gallipoli Barracks, this 2-wave theory, however plausible, remains speculative. Delayed or failed reporting of health concerns or symptoms of infected personnel can be a confounding issue when surveying soldiers attached to combat units. This difficulty is caused in part by a cultural resistance to seek medical attention for ailments of low to moderate severity; soldiers often opt to press on in the face of adversity.

A previous study based on phylogenies derived from short E2 gene sequences identified 2 lineages of RRV (3), whereas a more recent study (39) using longer sequences identified 3 (1 from eastern Australia, 1 from western Australia, and 1 from northeastern Australia). The phylogenetic analyses in our study, which used the complete nucleotide sequences of RRV E2 and nsP3 genes, confirm the presence of 3 lineages but suggest that lineage III should be divided into 6 sublineages. Lineage I viruses were classified as the northeastern Australia lineage; examples of this lineage have not been isolated since 1977. Lineage II was classified as the western Australia lineage; examples of this lineage have not been isolated since 1991 (Figures 2, 3; Appendix Table 1, Figures 1–8). The virus strain of RRV that was responsible for the outbreaks in Fiji, the Cook Islands, and Samoa during 1979–1980 has been classified as sublineage IIIA. The results of our study suggest that the lineage III viruses have been responsible for outbreaks of RRV infection in Australia since 1991. The temporal structure of the maximum clade credibility trees suggest that RRV is continuing to diversify under the pressure of natural selection and potential for further diversification exists given changes in the virus’ ecologic niche, cycles of transmission, or both. The importance of the 5 nonconservative changes in the amino acid sequences of the E2 proteins that occurred from 2016 to 2017 warrants further investigation to determine what role they might play in viral fitness. The implementation of continuous and routine viral sequencing, analysis, and monitoring will increase confidence in the proposed delineation of the sublineages within lineage III and further clarify the evolutionary process.

With the ADF’s commitment to maintaining the natural environment in its training areas, an ongoing risk for infection with RRV persists for personnel who do not have natural immunity through prior exposure to RRV and who are required to spend prolonged periods in this RRV-endemic area. A further risk is the potential for RRV to be exported to other countries through infected, asymptomatic personnel participating in multinational exercises, which occur on a regular basis in SWBTA. This risk is of particular concern for countries with mosquitoes known to be RRV vectors (40). The largest outbreak of RRV infection ever recorded occurred in the Pacific during 1979–1980 (4) and is a testament to the epidemic potential of the virus. The recent experience with Zika and chikungunya viruses underscores the serious threat posed to global health by the potential for previously obscure arboviruses to move from their historical cycles of transmission (41,42).

In conclusion, this investigation confirmed through viral isolation and sequence analysis that the outbreaks of EPA in the ADF in 2016 and 2017 were caused by 2 distinct sublineages of lineage III strains of RRV. Further, these outbreaks are most likely attributable to human-mosquito-human transmission.