Susceptibility of Aedes albopictus, Ae. aegypti and human populations to Ross River virus in Kuala Lumpur, Malaysia

Background Emerging arboviruses such as chikungunya and Zika viruses have unexpectedly caused widespread outbreaks in tropical and subtropical regions recently. Ross River virus (RRV) is endemic in Australia and has epidemic potential. In Malaysia, Aedes mosquitoes are abundant and drive dengue and chikungunya outbreaks. We assessed risk of an RRV outbreak in Kuala Lumpur, Malaysia by determining vector competence of local Aedes mosquitoes and local seroprevalence as a proxy of human population susceptibility. Methodology/Principal findings We assessed oral susceptibility of Malaysian Ae. aegypti and Ae. albopictus by real-time PCR to an Australian RRV strain SW2089. Replication kinetics in midgut, head and saliva were determined at 3 and 10 days post-infection (dpi). With a 3 log10 PFU/ml blood meal, infection rate was higher in Ae. albopictus (60%) than Ae. aegypti (15%; p<0.05). Despite similar infection rates at 5 and 7 log10 PFU/ml blood meals, Ae. albopictus had significantly higher viral loads and required a significantly lower median oral infectious dose (2.7 log10 PFU/ml) than Ae. aegypti (4.2 log10 PFU/ml). Ae. albopictus showed higher vector competence, with higher viral loads in heads and saliva, and higher transmission rate (RRV present in saliva) of 100% at 10 dpi, than Ae. aegypti (41%). Ae. aegypti demonstrated greater barriers at either midgut escape or salivary gland infection, and salivary gland escape. We then assessed seropositivity against RRV among 240 Kuala Lumpur inpatients using plaque reduction neutralization, and found a low rate of 0.8%. Conclusions/Significance Both Ae. aegypti and Ae. albopictus are susceptible to RRV, but Ae. albopictus displays greater vector competence. Extensive travel links with Australia, abundant Aedes vectors, and low population immunity places Kuala Lumpur, Malaysia at risk of an imported RRV outbreak. Surveillance and increased diagnostic awareness and capacity are imperative to prevent establishment of new arboviruses in Malaysia.

Introduction Recent re-emerging diseases caused by arboviruses such as chikungunya (CHIKV), dengue (DENV) and Zika (ZIKV) viruses are major public health issues [1]. These arboviruses have adapted to epidemic cycles in which virus transmission is sustained mainly by Aedes (Ae.) aegypti and Ae. albopictus mosquitoes. The recent and unexpected global spread of CHIKV and ZIKV from previously geographically restricted areas has increased awareness of future arboviral threats, driven by travel and expansion of Aedes distribution [2].
One arbovirus with epidemic potential is Ross River virus (RRV; Togaviridae family) which is endemic in Australia and the south Pacific region [3,4]. RRV is within the Semliki Forest serocomplex, which includes CHIKV, O'nyong-nyong virus and Mayaro virus [5]. RRV is usually non-fatal and causes rash, fever, fatigue, polyarthritis and myalgia, but debilitating arthritis may persist for up to 6 months [6]. Large outbreaks of RRV occurred in 1979 and 1980 in the Pacific islands of Fiji, the Cook Islands, Papua New Guinea, New Caledonia, and Vanuatu [7][8][9]. No major outbreaks have been reported outside of Australia since. The main vectors for RRV in Australia are Culex annulirostris, Ae. vigilax and Ae. camptorhynchus [10]. The main sylvatic hosts are considered to be marsupials such as kangaroos and wallabies, but nonmacropod hosts may be important in Fiji and French Polynesia [11,12]. There are currently no approved vaccines or antivirals.
Arboviruses meet various tissue barriers in a mosquito vector before being successfully transmitted to a new host via mosquito bite [13]. Following ingestion in a blood meal, the virus must overcome the midgut infection barrier (MIB) to infect midgut epithelial cells. Then, it must pass through the midgut escape barrier (MEB) to disseminate to other tissue and organs; in particular it must overcome the salivary gland infection barrier (SGIB) to productively infect salivary gland cells. Finally, the virus must pass through the salivary gland escape barrier (SGEB) to be shed in the saliva, ready to enter the bloodstream of the next host bitten by the mosquito [14,15]. These midgut and salivary gland barriers are critical for determining vector competence, and differ according to the arbovirus-mosquito combination [13].
The risk of local transmission of an arbovirus depends on the simultaneous presence of the virus, competent mosquitoes and susceptible human hosts. As RRV is able to infect multiple mosquito and vertebrate species, it has potential to establish viral transmissions, especially in areas with abundant mosquito vectors and naive populations [16]. While Ae. albopictus has been shown to be a competent vector for RRV and may act as a bridge vector for RRV with a potential risk for disease emergence [17], the role of Ae. aegypti is less certain [18]. Both Ae. aegypti and Ae. albopictus are widely distributed in tropical regions, especially in Southeast Asia, and are the main vectors of CHIKV, DENV and ZIKV in Malaysia [19][20][21]. As there is considerable traffic between Australia and Malaysia, there is a potential that RRV may be introduced by a viraemic traveller. Therefore, the aim of our study was to investigate the potential for an RRV outbreak in Malaysia by determining vector competence of local Ae. aegypti and Ae. albopictus mosquitoes for RRV. In addition, we assessed RRV serostatus among Malaysians in Kuala Lumpur as a proxy of population susceptibility. By detecting RRV in midguts, heads and saliva of the mosquito vectors, we can also make inferences about the main mosquito tissue barriers faced by RRV.

Ethics statement
The use of human serum samples was approved by the institution's Medical Ethics Committee (no. 2017116-5794). The committee does not require informed consent for retrospective studies of archived and anonymised samples.
Ae. aegypti and Ae. albopictus mosquitoes were collected from the field in Taman Seputeh, Kuala Lumpur. Mosquitoes were maintained under standard insectary conditions of 28 ± 1˚C and 80 ± 10% relative humidity with a 12 h:12 h photoperiod and fed with 10% sucrose supplemented with 0.1% vitamin B. The F3-F4 generation of the mosquitoes was used for virus infection. The mosquitoes were starved for 24 h prior to infection. Mosquito infection work was carried out in an arthropod containment level 2 laboratory.

Viruses
Ross River virus (RRV) strain SW2089 genotype 3/Eastern (GenBank accession number MN038260), kindly provided by Cheryl Johansen and David Smith from the University of Western Australia, was originally isolated from Ae. camptorhynchus mosquitoes in Mandurah, southwest Australia, in 1988. The virus has been passaged eight times in C6/36, Vero cells and BHK cells. The virus stock was stored at -80˚C and used for blood meals.
Chikungunya virus (CHIKV) strain MY/08/065 from the East/Central/South African (ECSA) genotype (GenBank accession number FN295485), was isolated from a patient from Johor, Malaysia in 2008 [19]. CHIKV was passaged three times in Vero cells and was used for the neutralization test.

Mosquito infection
To determine the optimal infectious dose, Ae. aegypti and Ae. albopictus mosquitoes were fed blood meals containing RRV at 3 log 10 , 5 log 10 and 7 log 10 PFU/ml. RRV was mixed 1:10 with fresh blood donated by one of the co-authors who has no history of RRV infection. The virusblood mixture was maintained at 37˚C and fed to seven-to eight-day old female Ae. aegypti and Ae. albopictus in the dark for 1 h using an artificial collagen membrane attached to a Hemotek meal reservoir (Discovery Workshops, UK). After feeding, the mosquitoes were snap frozen for 30 sec and ten engorged mosquitoes were transferred into individual cups and placed in a secure environmental chamber with the previously mentioned conditions. The experiments were conducted according to standard procedures in an arthropod containment level 2 laboratory. We selected 20 mosquitoes at 7 days post-infection (dpi), killed them by freezing, and placed each mosquito in a 1.5 ml zirconium beads tube (Benchmark Scientific, USA) prefilled with 500 μL of serum free DMEM. The mosquitoes were homogenized at 4000 rpm for 15 sec using a microtube homogenizer. The homogenates were stored at -80˚C and used for downstream viral RNA quantification in whole mosquitoes.

Vector competence analysis
Mosquitoes were orally infected using the optimum infectious doses previously determined and 20 mosquitoes were sampled at 3 and 10 dpi. The mosquitoes were immobilized by snap freezing and had their legs and wings removed to prevent escape. Each mosquito's proboscis was inserted into a pipette tip containing 10 μL of FBS with 10% sucrose and the mosquito allowed to salivate for 1 h. The saliva was then transferred into a PCR tube filled with 40 μL of serum free DMEM. The mosquito's midguts and heads were then dissected and placed individually into a 1.5 ml zirconium beads tube prefilled with 500 μL of serum free DMEM and homogenized. Dissection needles were soaked in 70% alcohol and thoroughly wiped between each mosquito. Vector competence was evaluated by calculating infection rate (number of PCR-positive midguts/number of tested midguts), dissemination rate (number of PCR-positive heads/number of PCR-positive midguts), transmission rate (number of PCR-positive saliva/number of PCR-positive heads), and transmission efficiency (number of PCR-positive saliva/number of mosquitoes fed with infectious blood meal).

Quantification of RRV viral load
To generate standard curves for RRV quantification, viral RNA was extracted from viral stock using QIAamp Viral RNA Mini Kit (Qiagen, Germany). The cDNA was synthesised with random primers using the Superscript III Reverse Transcriptase System (Thermo Fisher Scientific). PCR was performed using Q5 high fidelity DNA polymerase (NEB, UK), and 10 μM of E1 forward primer (5'-CGATGACGTGGGTACAGAGGA-3') and E1 reverse primer (5'-GTTACCAAGACCAGCACAACCA-3') to generate a 79 base pair fragment [22]. PCR cycling conditions were: 98˚C for 30 sec, then 25 cycles of 98˚C for 10 sec, 69˚C for 30 sec, 72˚C for 30 sec, and 72˚C for 2 min. The PCR fragment was cloned into pJET1.2/blunt cloning vector using CloneJET PCR Cloning Kit (Thermo Fisher Scientific), in vitro transcribed using Invitrogen MEGAShortScript Kit, and purified using MEGAClear kit (Thermo Fisher Scientific).
The RRV standard curves were generated using StepOnePlus Real-Time PCR System (Applied Biosystems, USA). The RNA standard was 10-fold serially diluted and added to the 10 μL real-time PCR reaction mixtures containing 4× Taqman Fast Virus 1-Step Master Mix (Thermo Fisher Scientific), 400 nM of E1 forward primer and E1 reverse primer, 250 nM of probe (5'-/56-FAM/TAGAGGGCC/ZEN/AGCCCACCTAACCCACTG/3IABkFQ/-3') [22], and RNAse-free water. PCR cycling conditions were: 1 cycle at 55˚C for 5 min, 95˚C for 20 sec, and 40 cycles at 95˚C for 3 sec and 60˚C for 30 sec. To determine the RRV viral load in the mosquitoes, viral RNA was extracted from mosquito homogenates and 1 μL of RNA was added to the 10 μL real-time PCR reaction mixture as previously mentioned. A blank control and uninfected blood-fed mosquitoes were used as negative controls. Triplicates were performed for each sample. Viral RNA in mosquito samples was quantified by comparing sample cycle threshold (Ct) values to standard curve Ct values. The limit of detection (LoD) of the assay is 1 copy/reaction, equivalent to 2.2 log 10 copies/organ and 1.7 log 10 copies/saliva. The limit of quantitation (LoQ) of the assay is 100 copies/reaction, equivalent to 4.2 log 10 copies/ organ and 3.7 log 10 copies/saliva.

Serum samples
We tested seroprevalence of RRV antibodies in residual serum samples collected from inpatients and sent to the diagnostic microbiology laboratory at the University Malaya Medical Centre (UMMC) in Kuala Lumpur. Samples were collected in 2018 and 2019, and we excluded samples sent for dengue (DENV) and CHIKV testing, as both are endemic in Malaysia and have similar clinical presentation to RRV. An older study showed a range of RRV seroprevalence of 0-4% in non-endemic countries in Southeast Asia [23]. Using an estimated prevalence of 3.5% and precision of 2.5%, we calculated a minimum sample size of 208. A total of 240 serum samples were obtained, with 30 samples per age groups <10, 10-19, 20-29, 30-39, 40-49, 50-59, 60-69 and �70 years.

Plaque reduction neutralization test
Serum samples were tested using plaque reduction neutralization test (PRNT) for anti-RRV neutralizing antibodies [24]. Briefly, samples were heat-inactivated at 56˚C for 30 min, then diluted in 1× DPBS from 1:20 to 1:1280 and mixed with 70-80 PFU of RRV diluted in 2% FBS DMEM to a final volume of 200 μL. The virus-antibody mixture was incubated at 37˚C for 1.5 h before inoculation into 10 5.4 Vero cells per well in a 24-well plate. The plate was incubated for 1 h at 37˚C before replacing with plaque medium, 2 parts 3% carboxymethylcellulose and 3 parts 3.5% FBS-DMEM. After 3 days of incubation, cells were fixed with 3.7% of formaldehyde and stained with crystal violet. PRNT 50 was calculated as the dilution that reduced plaque formation by 50% relative to virus control. As both CHIKV (an alphavirus like RRV) and DENV (flavivirus) are endemic in Malaysia and cause similar symptoms to RRV, we checked for cross-reactivity between RRV, DENV and CHIKV antibodies. We tested 10 DENV-seropositive serum samples, confirmed by Panbio Dengue IgM/IgG ELISA (Abbott, USA), and 6 CHIKV-seropositive serum samples, confirmed by IgG ELISA [25], for ability to neutralize RRV.
Initial screening for neutralizing antibodies against RRV was carried out with samples at 1:10 dilution. Samples that reduced plaques by >75% relative to virus control were retested with 2-fold serial dilutions from 1:20 to 1:320 to determine PRNT 50 . Serum samples with detected anti-RRV antibodies were also tested for specificity against CHIKV by PRNT 50 and the plates were incubated for 2 days. Virus controls were spiked with serum samples from healthy donors with no serological evidence of past infection of CHIKV and RRV at a dilution of 1:10.

Statistical analysis
Fisher's exact test (two-tailed) was used to compare infection, dissemination, transmission and transmission efficiency rates for different mosquito species at the same time point or for the same mosquito species at different timepoints. The Mann-Whitney U test was used to compare distributions or medians of viral loads (with non-normal distributions) between the two mosquito species at each time point. Viral loads with normal distributions were compared with the unpaired t-test. These analyses were performed using GraphPad Prism 9.0 (GraphPad Software, USA). Probit regression was used to calculate the oral infectious doses needed to infect 50%, 75% and 90% of fed mosquitoes (OID 50 , OID 75 and OID 90 , respectively) [26], using IBM SPSS Statistics v25 (IBM, USA). Differences between OID values of the two mosquito species were considered statistically significant if the 95% confidence intervals (CI) of the OID ratios did not overlap [27]. Using outcomes of dissemination (detectable RRV RNA in heads) and transmission (detectable RRV RNA in saliva), the effects of days post-infection, midgut and head viral loads were determined using multiple regression analysis. Viral loads found to be significant predictors were then evaluated with receiver operating characteristic (ROC) curves for ability to predict dissemination or transmission [28]. Youden's J statistic (where J = sensitivity + specificity -1) was used to determine the viral load with the highest combined sensitivity and specificity. Two-tailed p-values of <0.05 were considered significant.
Due to the theoretical cross-reaction between alphaviruses, a sample was considered probable RRV-seropositive only if RRV PRNT 50 titres were �20 and additionally CHIKV PRNT 50 titre was <10 (undetectable). A sample was defined as possible RRV-seropositive with PRNT 50 titres of �20 and detectable anti-CHIKV antibodies, but if CHIKV PRNT 50 titre was �4-fold higher than RRV PRNT 50 , then that sample was considered probable CHIKV-seropositive. Seroprevalence rate was determined by dividing the number of probable cases over the total number of samples tested. Rates are reported with Wilson 95% CI (https://epitools.ausvet. com.au/ciproportion).

Ae. albopictus is more susceptible than Ae. aegypti to oral infection with Ross River virus
Ae. aegypti and Ae. albopictus were fed infectious blood meals containing 3, 5 and 7 log 10 PFU/ ml of RRV, and infection rates and viral loads in the whole mosquitoes were determined at 7 dpi (S1 Data). Infection rate was significantly higher in Ae. albopictus compared to Ae. aegypti (Fisher's exact test: 60% vs 15%; p<0.05) at 3 log 10 PFU/ml, but comparable at 5 log 10 PFU/ml (75-95%) and 7 log 10 PFU/ml (100%) (Fig 1A). The OID 50 and OID 75 for Ae. aegypti were significantly higher than that of Ae. albopictus, although the OID 90 values were similar (Table 1). Increasing the infectious dose from 3 to 7 log 10 PFU/ml led to increasing viral loads in both mosquitoes, with Ae. albopictus demonstrating significantly higher viral loads than Ae. aegypti (Fig 1B). Most of the RRV RNA detected in Ae. aegypti at infectious doses of 3 to 5 log 10 PFU/ ml fell below the LoQ of the assay. A further observation was the bimodal distribution of viral loads in Ae. aegypti at each blood meal titre, with mosquitoes either showing an undetectable/ low viral load (at 3 and 5 log 10 PFU/ml blood meals), or a low/high viral load (at 7 log 10 PFU/ ml). The distributions for Ae. albopictus were more tightly clustered. As medians should not be directly compared for binomial distributions, we used the Mann-Whitney U test to compare ranks and distributions of the values, rather than compare medians.
By removing mosquitoes with undetectable RRV RNA, found in the 3 and 5 log 10 PFU/ml blood meal experiments, the bimodal distributions were no longer present in Ae. aegypti and median viral loads could be compared with Ae. albopictus ( Table 2). The bimodally distributed viral loads in Ae. aegypti fed the 7 log 10 PFU/ml blood meal were also analysed separately as a low (2.8-4.5 log 10 copies/mosquito) and high (8.7-9.9 log 10 copies/mosquito) viral load groups. There were no significant differences between the two mosquito species at 3 log 10 PFU/ml, although only 3 Ae. aegypti mosquitoes were infected at this dose. Despite similarly

PLOS NEGLECTED TROPICAL DISEASES
high infection rates, Ae. albopictus still had significantly higher median viral loads at both 5 and 7 log 10 PFU/ml blood meals, even when compared to the Ae. aegypti high viral load group.
These findings suggest that Ae. albopictus is more susceptible to oral infection with RRV than Ae. aegypti. For the optimum oral infectious doses of RRV for further infection, dissemination and transmission experiments, we selected 5 log 10 PFU/ml for Ae. albopictus and 7 log 10 PFU/ml for Ae. aegypti, as these doses exceeded the OID 90 values for each. The range for both these doses is similar to reported levels of viraemia in experimentally infected nonhuman vertebrates [6].

Malaysian Ae. aegypti and Ae. albopictus are potential RRV vectors
We determined vector competence (infection, dissemination, transmission rates and transmission efficiency) of Ae. aegypti and Ae. albopictus mosquitoes (S1 Data) using the previously determined optimum oral infectious doses of 7 log 10 PFU/ml and 5 log 10 PFU/ml, respectively. Midgut infection rates in Ae. aegypti at 3 and 10 dpi ranged between 85-100% (Fig 2A), with similar viral loads (Fig 2D). Virus dissemination increased from 70% at 3 dpi to 100% of the mosquito heads at 10 dpi (Fig 2B), with comparable viral loads (Fig 2E). Transmission was detected in 50% of the saliva at 3 dpi and remained low at 10 dpi at 41% (Fig 2C), with very low viral loads below the LoQ (Fig 2F).
When Ae. albopictus was fed with 5 log 10 PFU/ml blood meal, all midguts (100%) were infected at 3 dpi (Fig 3A). No significant difference was observed between the midgut viral loads at 3 and 10 dpi (Fig 3D), indicating high susceptibility of Ae. albopictus to RRV infection and low MIB. The dissemination rates were high at 95-100% at 3 and 10 dpi, with similar viral loads (Fig 3B and 3E). Transmission rates increased significantly over time from 47% (3 dpi) to 100% (10 dpi) (Fisher's exact test; **, p<0.005), though viral loads remained similar (Fig 3C  and 3F). Overall, the results indicate that Malaysian strains of both Aedes species are capable of transmitting RRV. Viral loads are measured in log 10 RNA copies/mosquito.
Ae. aegypti fed a blood meal of 7 log 10 PFU/mL showed a bimodal distribution of low and high viral loads (Fig 2), which are shown separately here. Median viral loads of the two mosquito species are compared at each time point, and significant differences are shown (Mann-Whitney U; *, p<0.01, **, p<0.0001). https://doi.org/10.1371/journal.pntd.0011423.t002

PLOS NEGLECTED TROPICAL DISEASES
Aedes mosquito and human susceptibility to Ross River virus, Malaysia

Ae. albopictus is a more competent vector for RRV than Ae. aegypti
We used different RRV blood meal concentrations for the two mosquitoes to ensure infection rates were >90%, which allows comparison of vector competence parameters with a similar starting infection rate (S1 Fig). Both Ae. aegypti and Ae. albopictus demonstrated high midgut infection rates (S1A Fig), with similar viral loads (S1D Fig). This suggests no significant MIB in both Aedes vectors. Dissemination to heads was also similar in Ae. aegypti and Ae. albopictus at 3 dpi (70% vs 95%) and 10 dpi (100% vs 100%) (S1B Fig), but there were significantly higher viral loads in Ae. albopictus heads (S1E Fig). We did not specifically dissect out and include salivary glands with the head, so some of the detectable RRV in heads would be from non-salivary tissue. Nevertheless, this finding suggests either a greater MEB and/or head tissue barrier (which may include the SGIB) in Ae. aegypti. Subsequently, for Ae. aegypti, only 41-50% of the saliva were positive for RRV at 3 and 10 dpi (S1C Fig), with very low viral loads below the LoQ (S1F Fig). This suggests Ae. aegypti presents a greater SGEB to RRV than Ae. albopictus, which showed 100% (p<0.005) dissemination at 10 dpi. As seen in the oral infectious dose experiments, the viral loads in some Ae. aegypti organs had a bimodal distribution. This was less evident for Ae. albopictus, apart from head viral load at 10 dpi. As transmission is a key endpoint, we removed saliva samples with undetectable RNA in order to compare the mean (normally distributed) viral loads in only the positive samples (S1 Table). When RRV was detectable in saliva, Ae. albopictus had significantly higher mean viral loads than Ae. aegypti at 3 dpi (3.9 vs 2.6 log 10 copies/sample) and 10 dpi (4.7 vs 2.6 log 10 copies/sample).
Overall, despite being infected with a viral dose which was 100-fold lower than Ae. aegypti, Ae. albopictus showed greater competence for RRV, with higher transmission rates and efficiency and viral loads in the saliva.

PLOS NEGLECTED TROPICAL DISEASES
Aedes mosquito and human susceptibility to Ross River virus, Malaysia

Viral loads as predictors for dissemination and transmission
There were no instances of positive heads (dissemination) with undetectable midgut viral load, and no instances of positive saliva (transmission) with undetectable virus in midguts or heads. This is consistent with infection of midguts and heads being required for RRV to be potentially transmissible in saliva. We carried out logistic regression to determine the predictive value of  midgut and head viral loads on dissemination and transmission rates (Table 3). In Ae. aegypti, midgut viral load was a good predictor (AUC of 0.84) of disseminated infection, but a poor predictor (AUC of 0.69) of transmission. In Ae. albopictus, viral load in the head was a fair predictor of detectable RRV in saliva (AUC of 0.77). The cut-off values giving the best combined sensitivity and specificity rates were 5.3 log 10 RNA copies/midgut for Ae. aegypti and 6.6 log 10 RNA copies/head for Ae. albopictus, but sensitivity rates were low, ranging from 58.6-85.7%.

Population susceptibility to RRV in Kuala Lumpur
We assessed seroprevalence of residual diagnostic serum from Kuala Lumpur hospital patients as a proxy of population susceptibility. No cross-neutralizing activity against RRV was demonstrated with 6 CHIKV-seropositive and 10 DENV-seropositive serum samples. Of the 240 patient samples tested, 4 were possible RRV-seropositive with PRNT 50 titres �20 (range, 20-80). The two samples with RRV PRNT 50 of 20 had CHIKV PRNT 50 titres which were 4-8-fold higher, and were considered probable CHIKV-seropositive. The other two samples (RRV PRNT 50 of 40 and 80) were considered probable RRV-seropositive as CHIKV PRNT 50 titres were <10. Overall, there was very low probable RRV-seropositivity (2/240, 0.8% [0.2%-3.0%]) in this sample of hospital inpatients from Kuala Lumpur.

Discussion
There is extensive travel from Australia to Malaysia (1.1 million airline passengers in the year ending June 2019) [29], and an abundance of Aedes vectors in Malaysia [30]. We therefore studied vector competence and seroprevalence data as a proxy for population susceptibility to assess risk for RRV transmission in Malaysia. RRV is thought endemic only in Australia and Papua New Guinea, maintained in nature by native macropods [10]. However, a wide range of mammals and birds have been infected experimentally, and RRV seropositivity has been found in common animals such as cows, goats and horses [31]. Significant RRV outbreaks and enzootic transmission have occurred in regions without marsupials, such as the Pacific Islands and metropolitan Australia [32,33]. High viraemia in humans and low seroprevalence in nonhuman vertebrates during these outbreaks strongly suggest human-mosquito-human transmission [31]. Thus, epidemic RRV transmission is possible without preferred animal hosts, if competent mosquito vectors are abundant, climatic conditions are ideal, and human herd immunity is low [34,35].
We showed that Malaysian field Ae. aegypti and Ae. albopictus are susceptible to RRV in a dose-dependent manner. Both species potentially transmit RRV, with virus RNA detectable in saliva from 3 dpi. Dose-dependent oral infection is important, as higher dose blood meals increase transmission of CHIKV by Ae. albopictus [36]. The OID 90 of Ae. aegypti (5.7 log 10 PFU/ml) and Ae. albopictus (4.3 log 10 PFU/ml) were comfortably below reported viraemia rates of up to 7.6 log 10 PFU/ml in experimentally infected mammals [6] and 8 log 10 PFU/ml in naturally infected horses [37]. Thus, Aedes mosquitoes may facilitate human-mosquito-human transmission, as reported in the Cook Islands RRV outbreak involving Ae. polynesiensis [8].
The short extrinsic incubation period (EIP) of 3 days observed in both Aedes species here is comparable with the established RRV vector Ae. vigilax in Australia [38]. A similarly short EIP in Ae. albopictus facilitated the large CHIKV outbreak in Malaysia in 2008 [19]. Ae. albopictus is a potential RRV vector in Australia, Asia and the USA [17,39,40], and has comparable vector competence with Ae. vigilax [41]. The mean transmission efficiency for enzootic RRV by Ae. albopictus may exceed epidemic viruses such as DENV, yellow fever virus and ZIKV; with its opportunistic feeding behaviour and wide host-feeding range, Ae. albopictus may serve as a bridge vector to transmit emerging viruses between animals and humans [42].
We further found that Ae. albopictus is a more competent RRV vector than Ae. aegypti, with lower threshold for oral infection and higher transmission efficiency (100% vs 35% at 10 dpi) despite a blood meal containing 100× lower RRV load. Although little is known about RRV and tissue barriers in Aedes vectors, there are several studies of other arboviruses [13,43]. We found no significant MIB to RRV in both Aedes species. However, Ae. aegypti showed either a greater MEB or SGIB than Ae. albopictus, resulting in significantly lower virus titres in heads. Our data indicates that a threshold of viral infection is important for midgut escape. A blood meal triggers transient disruption of midgut basal lamina, and an arbovirus must replicate to achieve that threshold quickly enough to pass through and disseminate [15]. Mosquito innate immune responses like the RNA interference system [44] also modulate viral escape from the midgut, as shown for Sindbis virus [45].
The MIB and SGIB require the presence of appropriate receptors for virus binding and entry, which vary within organs. For example, SINV was detected in lateral but not medial salivary lobes of Ae. aegypti [46]. This may affect the SGEB as virus is released into apical cavities where saliva is stored before entering salivary ducts. In our study, a SGEB to RRV was evident early (3 dpi) in both Aedes species, which by 10 dpi had been overcome in Ae. albopictus, but not Ae. aegypti. SGEB is influenced by genetic differences within Ae. aegypti populations and viral lineages, as shown for CHIKV [47,48]. Two Western equine encephalitis virus strains showed different saliva shedding in Culex tarsalis, which was partially overcome by a chimeric recombinant containing structural genes from the more transmissible strain [49]. Thus, Aedes species in Malaysia will likely demonstrate varying competence for RRV strains other than SW2089 used in this study.
We found bimodal distributions of viral loads-particularly in Ae. aegypti-which may represent distinct mosquito or virus subpopulations. The association between higher Ae. aegypti midgut RRV loads with dissemination suggests that a viral load threshold is required to cross the barrier. The basis of intraspecies variation in vector competence is unknown, but involves complex interplay of numerous genes with environmental factors [50]. Virus quasispecies with greater vector adaptation may arise during infection; key envelope protein E1 mutations in CHIKV emerged in Ae. albopictus mosquitoes that could transmit CHIKV, but were absent in those that could not [51]. We could further explore this by whole genome sequencing of RRV from successfully infected Aedes mosquitoes demonstrating large differences in transmission.
We then assessed RRV seroprevalence as a proxy for population immunity in the event of imported RRV. Anti-RRV antibodies likely persist and protect for life after infection [52]. Our low rate of anti-RRV neutralizing antibodies of 0.8% (0.2%-3.0%) in Kuala Lumpur inpatients indicates susceptibility and low likelihood of unsuspected endemicity. This supports a 1975 study reporting zero seroprevalence in Southeast Asian populations, including Malaysia [23]. In contrast, seropositivity rates of >30% are reported from RRV-endemic Papua New Guinea [23] and Pacific islands where endemicity is suspected [3]. Furthermore, the lack of cross-neutralization by anti-CHIKV antibodies against RRV indicates that CHIKV endemicity in Malaysia will not protect human populations against RRV [53].
There are limitations to our work. Laboratory demonstration of vector competence does not replicate natural conditions, and surveillance of field mosquitoes should be done. To date, Ae. albopictus and Ae. aegypti have not been incriminated as natural RRV vectors [54], so it would be important to test vector competence of other Aedes and Culex species found in Malaysia. Single strains of RRV and each mosquito species were tested, and clearly more should be evaluated. There are at least 4 RRV genotypes (G1-G4) with G4 the most recently predominant, although it is most closely related to G3 [55], to which the study RRV strain SW2089 belongs. Ae. aegypti from Fiji, Townsville and Jakarta showed good vector competence to RRV [40,54], thus vector competence differences exist among mosquitoes of the same species from different regions. In addition, since PCR cannot distinguish between live and non-viable virus, future work should determine infectious virus titres to confirm dissemination and transmission; nevertheless, the presence of high viral loads of 4-5 log 10 copies in saliva (higher in organs) in Ae. albopictus strongly suggests viral replication occurred and that transmission is possible. Our seroprevalence findings are limited by bias from studying a single centre. However, convenience sampling of residual hospital serum gave similar estimates of influenza seroprevalence to cohort studies [56]. Future sampling should be more representative and include rural areas where Ae. albopictus predominates [57,58].
In summary, RRV is an arbovirus with epidemic potential to spread beyond its current geographical distribution. We show that Kuala Lumpur fulfils two key factors determining risk for RRV outbreaks: a naive host population and abundant competent mosquito vectors. Virus introduction into suitable conditions by a viraemic traveller, reported in neighbouring Singapore [59], could potentially trigger an outbreak, as seen in the South Pacific in 1979-1980 [8]. The greater vector competence of Ae. albopictus over Ae. aegypti has implications for disease distribution; the preference of epidemic CHIKV for Ae. albopictus led to disproportionate disease incidence in rural areas in Sarawak, Malaysia [60]. The possible roles of animals as reservoirs should be explored, although enzootic transmission is unlikely to be important if these two Aedes mosquito species are involved as vectors, as they predominantly feed on humans [61]. Nevertheless, early detection in countries at risk, particularly with travellers from RRVendemic areas, requires surveillance for RRV, increased diagnostic awareness and laboratory preparedness.