Potential risk of re-emergence of urban transmission of Yellow Fever virus in Brazil facilitated by competent Aedes populations

Yellow fever virus (YFV) causing a deadly viral disease is transmitted by the bite of infected mosquitoes. In Brazil, YFV is restricted to a forest cycle maintained between non-human primates and forest-canopy mosquitoes, where humans can be tangentially infected. Since late 2016, a growing number of human cases have been reported in Southeastern Brazil at the gates of the most populated areas of South America, the Atlantic coast, with Rio de Janeiro state hosting nearly 16 million people. We showed that the anthropophilic mosquitoes Aedes aegypti and Aedes albopictus as well as the YFV-enzootic mosquitoes Haemagogus leucocelaenus and Sabethes albiprivus from the YFV-free region of the Atlantic coast were highly susceptible to American and African YFV strains. Therefore, the risk of reemergence of urban YFV epidemics in South America is major with a virus introduced either from a forest cycle or by a traveler returning from the YFV-endemic region of Africa.


Results
Aedes mosquitoes were efficient to disseminate and transmit from day 14 post-infection. To define the days after infection for examining viral dissemination and transmission in Ae. aegypti and Ae. albopictus, we infected Ae. aegypti AE-GOI and Ae. albopictus AL-GOI collected in Goiânia in the state of Goiás, Central Brazil, in the YFV-epidemic/epizootic region with three YFV strains (two from Brazil (74018-1D and 4408-1E) and one from Senegal (S-79)). When examining AE-GOI, mosquitoes were infected and disseminate YFV from 7 days post-infection (dpi), and viral particles were only detected in mosquito saliva at 14 dpi (Fig. 2). For AL-GOI, infection and dissemination started from 7 dpi and detection of virus in saliva from 14 dpi (Fig. 2).
Interestingly, out of 447 female mosquitoes studied at 3 dpi, no mosquitoes showed infection. Therefore, specimens examined at 3 dpi were not considered in further analysis. Moreover, considering all Ae. aegypti populations on the one hand, and all Ae. albopictus populations on the other hand, logistic regression models showed similar rates of infection at 14 and 21 dpi (p = 0.10 and p = 0.73, respectively). Therefore, all further analyses were conducted considering 14 and 21 dpi together. Brazilian Aedes aegypti were susceptible to American as well as African YFV genotypes. To determine if Ae. aegypti populations from Brazil were competent vectors of YFV, three populations (AE-MAN, AE-GOI, AE-RIO collected in enzootic-, epizootic/epidemic-and free-YFV areas, respectively; Fig. 3) were experimentally infected with three YFV strains (74018-1D, 4408-1E and S-79). When analyzing viral infection, AE-MAN, AE-GOI and AE-RIO presented significant differences of IR for the three YFV strains (p < 0.05, Fig. 3) with values ranging from 30% (AE-MAN infected with S-79) to 85% (AE-RIO infected with 4408-1E). After controlling for population, virus and dpi, level of infection was higher with the strain 74018-1D in AE-GOI, was higher with the strain 4408-1E and 74018-1D in AE-MAN, while no difference was observed in AE-RIO (Table 1).
After midgut infection, the virus must propagate inside the mosquito hemocele to allow detecting a positive viral dissemination. Dissemination as determined by the presence of virus in mosquito heads, were similar regardless of the YFV strain except for AE-MAN which presented higher viral dissemination of YFV 4408-1E (90.47%) (Fig. 3, Table 2).

Figure 2.
Infection, Dissemination and Transmission of YFV by Aedes aegypti AA-GOI and Aedes albopictus AL-GOI from the epizootic/epidemic region of Goiânia. Mosquitoes were exposed to blood meals at a titer of 10 6 PFU/mL. Engorged females were maintained in laboratory conditions until examination at 3, 7, 14, 21 days post-infection. Mosquito thorax and abdomen were processed individually to determine the infection rate (IR, proportion of mosquitoes with infected body among the engorged mosquitoes). The mosquito head was used to define the disseminated infection rate (DIR, proportion of mosquitoes with infected head among infected mosquitoes) and the saliva collected from individual females to determine the transmission rate (TR, proportion of mosquitoes with infectious saliva among mosquitoes with disseminated infection). Asterisks refer to a significant difference (*p < 0.05, **p < 10 −2 , ***p < 10 −3 ). In brackets, the number of mosquitoes tested.
For viral transmission to occur, the virus in the hemocele must reach mosquito salivary glands and be excreted with saliva expectorated by the mosquito. TR as determined by the presence of virus in saliva, were similar for all three Brazilian Ae. aegypti populations regardless of the YFV strain (p > 0.05; Fig. 3, Table 3). These results suggest that the salivary glands behave as a more efficient barrier for the release of viruses than the midgut for dissemination. Moreover, all mosquito populations present similar ability to deliver particles of the two YFV strains in saliva: 74018-1D responsible for epizootics from 1998 to 2001 4 and the new viral lineage, 4408-1E, causing increasing deaths of monkeys and human cases 15 . Aedes albopictus in Rio de Janeiro were very efficient to deliver particles of YFV from their saliva. To determine if Ae. albopictus populations from Brazil were as competent vectors as Ae. aegypti for YFV, Ae. albopictus from Manaus (AL-MAN), Goiânia (AL-GOI) and Rio (AL-RIO) were experimentally infected with three YFV strains (two from Brazil and one from Senegal). When comparing viral infection, AL-MAN and AL-RIO presented low (ranging from 10% for AL-RIO infected with S-79 to 21.42% for AL-MAN infected with S-79; p > 0.05; Fig. 4), and comparable IR values whatever the viral strain (p = 0.71 and p = 0.19, respectively; Table 1). On the other hand, AL-GOI showed significant differences (p < 0.05) with a higher IR value for S-79 (60.97%) ( Table 1). It must also be noted that the rate of infection was significantly lower in Ae. albopictus than in Ae. aegypti for all viral strains, except in females from Goiânia when infected with S79.
When analyzing viral dissemination represented by DIR, dissemination was lower in AL-GOI, then higher in AL-RIO and then even higher in AL-MAN (Fig. 4, Table 2). All three Ae. albopictus populations showed similar values (p > 0.05) for the three YFV, although for AL-MAN, dissemination tended to be higher with viral strain 1E ( Table 2). The highest DIR values were obtained for AL-MAN and AL-RIO populations. Interestingly, dissemination was similar in AL and in AE from Manaus and from Rio; while in female mosquitoes from GOI, the dissemination was significantly lower in AL than in AE (Table 2).
When considering viral transmission described by TR, AL-MAN population showed high TR values but based on low sample sizes (66.67% (6) with 74018-1D and 50% (4) with 4408-1E) and surprisingly, was not able to transmit the S-79 YFV from Africa. The population AL-GOI collected in the epizootic/epidemic region were able to excrete similarly all three YFV (p > 0.05) but was less efficient than AL-RIO (Fig. 4). These results suggest that AL-RIO and AL-MAN shared the same pattern of infection, dissemination and transmission with low IR, high DIR and high TR values suggesting the role of the midgut as the main barrier in the trajectory of the virus . Mosquitoes were exposed to blood meals at a titer of 10 6 PFU/mL. Engorged females were maintained in laboratory conditions until days 14-21 post-infection. Mosquitoes were processed as previously described to determine the infection rate (IR), the disseminated infection rate (DIR) and the transmission rate (TR). Asterisks refer to a significant difference (*p < 0.05, **p < 10 −2 ). In brackets, the number of mosquitoes tested. The map was created using software the CorelDraw X5 software (http://www. coreldraw.com/br/). to the mosquito salivary glands. Interestingly, AL-GOI was less susceptible to YFV than the other two mosquito populations. Like Ae. aegypti, Ae. albopictus were similarly susceptible to the former YFV 74018-1D lineage and the new viral lineage 4408-1E (Table 3).
Aedes mosquitoes from an African YFV-endemic country were similarly susceptible to American as well as African YFV genotypes. To test whether Ae. aegypti and Ae. albopictus from Congo were competent for both Brazilian and West African YFV strains, mosquitoes were infected with the three YFV strains. The population AE-CON and AL-CON showed IR ranging between 25% (when infected with 74018-1D) and 38.6% (when infected with 4408-1E). Whatever the viral strain, the infection rate was not different between AE-CON and AL-CON (p > 0.05; Table 1) Regarding dissemination, AE-CON and AL-CON also showed much higher viral dissemination rates than the other female mosquitoes (Fig. 5). Interestingly, American viral strains led to significantly higher dissemination than the African one both for AE-CON and for AL-CON (p < 0.05), but we did not observe any difference between AE-CON and AL-CON (p = 0.72). Viral transmission measured by TR was slightly lower except when infected with the YFV 4408-1E strain (73.33%). On the other hand, AL-CON presented roughly similar patterns than AE-CON with IR lower than 31.37% (S-79) and DIR higher than 68.75% (S-79). TR did not differ significantly (p > 0.05), varying from 36.36% (S-79) to 50% (74018-1D) strains. Thus Ae. aegypti and Ae. albopictus from Congo, city of Brazzaville presented similar vector competence indices when infected with YFV strains belonging to both the South America I and West Africa genotypes.
Wild YFV vectors Hemagogus and Sabethes from Rio de Janeiro were highly competent to transmit Brazilian and African YFV strains. To examine if enzootic mosquitoes from Brazil, Hg. leucocelaenus and Sa. albiprivus were as susceptible as Ae. aegypti and Ae. albopictus, mosquitoes were infected with the three YFV strains. Both enzootic mosquito species showed pattern of infection similar to Ae. aegypti while Ae. albopictus was significantly less often infected (as previously shown in this study). Overall, dissemination occurred in 64.3% of mosquitoes, this rate of dissemination was not different between the four species (p = 0.34), between the three viruses (p = 0.14). Overall, transmission was observed in 36.6% and was not different between the four species (p = 0.85) nor between the three viruses (p = 0.95). IRs were higher than 48%, DIRs higher than 44% and TRs higher than 40% suggesting a limited role of both barriers, midgut and salivary glands, in the virus migratory route inside the mosquito. When mosquitoes with infectious saliva were considered according to the initial number of females tested, transmission efficiencies (TE) varied from 10.81% to 20% for Hg. leucocelaenus  Figure). Thus enzootic YFV vectors from Brazil were highly competent to transmit YFV from Brazil as well as from West Africa.

Discussion
Our results showed that Ae. aegypti and Ae. albopictus from YFV-free regions in Brazil were as susceptible as their counterparts in the endemic Amazon region to transmit YFV. Mosquitoes from the epizootic/epidemic region were broadly less competent suggesting that Ae. albopictus and to a lesser extent, Ae. aegypti, were unlikely to play the role of active bridge vector in the emergence zone even if it cannot definitively be excluded.
Since the main YF-vector Ae. aegypti has been eradicated from Brazil in 1957 9 , YFV had been maintained within a forest cycle in endemic/enzootic regions where waves of epizootics coincided with the renewal of non-human primate populations, taking an interval of 6-10 years 22 25 . We showed that Hg. leucocelaneus as well as Sa. albiprivus were highly susceptible to all three YFV strains corroborating their role as YFV-enzootic vectors 26 . They were able to experimentally deliver infectious viral particles in saliva after infection with the two YFV strains, the former 1D and the new viral lineage 1E, as well as the West African YFV strain isolated from Senegal in 1979. Their pattern of YFV infection,   27 . YFV is at the gates of the most densely populated zone in South America, the Brazilian Atlantic coast, where are located large cities like Rio de Janeiro (WHO, "Yellow fever -Brazil", 2017; http://www.who.int/csr/don/27-january-2017-yellow-fever-brazil/en/). Ae. albopictus colonizes regions surrounding the urban environment where Ae. aegypti remains the main vector of arboviruses pathogen to humans (dengue, chikungunya, Zika). As expected, we showed that both mosquito species were able to excrete YFV from day 14 28 . Populations of Ae. aegypti and Ae. albopictus from -endemic, -epizootic/epidemic, and YFV-free regions were compared for their performance to be infected, disseminate and transmit YFV. We brought out that mosquitoes from the YFV-endemic (i.e. Manaus) and YFV-free regions (i.e. Rio de Janeiro) were more capable after oral infections to expectorate the two American (74018-1D and 4408-1E) and one African (S-79) YFV strains than mosquitoes from the epizootic/epidemic area (i.e. Goiânia). These two geographically distant regions are linked by an emergence region where mosquitoes are susceptible to YFV infection. Ae. albopictus as well as Ae. aegypti from Goiânia are able to deliver the three YFV strains in their saliva after experimental infections but with much less efficiency. However, a poorly competent vector may play an important role in transmission if other conditions are met, e.g. high vector densities, high human-biting rate and high daily survival rates 29 . Moreover, it has been shown that human population has extensively contributed to YFV dispersal 2 . Some environmental/ecological barriers by inhibiting movements of vectors and hosts can prevent YFV spillovers from a forest cycle. However, these barriers became more and more anecdotal in Brazil as with population growth, cities are increasingly close to YFV-enzootic forests. Besides, because of their original habitat degradation and their high propensity to explore new environments, NHPs such as capuchins and marmosets have densely invaded parks and urban areas in Rio de Janeiro and other cities in the Atlantic coast. These new urban invaders colonize a large area around cities including patches of forests 30,31 where sylvatic YFV vectors can be abundant 32 . Considering all these conditions, it is difficult to understand why YF is not already in the urban areas of Brazil. However, it seems that this is changing with the recent detection of human cases at less than that 100 km apart from city of Rio de Janeiro  . Mosquitoes were exposed to blood meals at a titer of 10 6 PFU/mL. Engorged females were maintained in laboratory conditions until days 14-21 post-infection. Mosquitoes were processed as previously described to calculate the infection rate (IR), the disseminated infection rate (DIR) and the transmission rate (TR). Asterisks refer to a significant difference (*p < 0.05, **p < 10 −2 ). In brackets, the number of mosquitoes tested. The map was created using software the CorelDraw X5 software (http://www. coreldraw.com/br/). that transmission implicates urban vectors, mainly Ae. aegypti. Imported cases from Angola were later confirmed in China 33 stressing the risk of spread outside Africa in Aedes-infested countries through non-immunized travelers. All four mosquito species, Ae. aegypti, Ae. albopictus, Hg. leucocelaenus and Sa. albiprivus from Rio de Janeiro were highly susceptible to Brazilian as well as West African YFV lineages. The city of Rio de Janeiro is an important touristic and trade center hosting nearly 6.5 million inhabitants, and where annually converge 26 millions of Brazilians/tourists in airports and 19 million via roads (http://cidades.ibge.gov.br/xtras/perfil.php?cod-mun=330455). If the virus is introduced from an infected traveler into Rio de Janeiro, opportunities to initiate a vectorial transmission of YFV are multiple: (i) by anthropophilic mosquitoes such as Ae. aegypti and Ae. albopictus which are highly susceptible to YFV and (ii) by YFV-enzootic mosquitoes Hg. leucocelaenus and Sa. albiprivus colonizing the forest near Rio de Janeiro. Therefore, vaccination of travelers visiting Rio de Janeiro should be highly advised to limit the risk of introducing the virus from YFV-endemic areas.
An effective YFV vaccine has been available since the 1930s. Unfortunately, incomplete coverage in regions at risk of infection is responsible for several thousands of deaths every year. In the Americas, to prevent enzootic spillovers with introduction of YFV into the urban cycle, people in contact with the jungle should be rapidly vaccinated in priority to prevent a potential urbanization of YFV. Mosquitoes. Ten American and African mosquito populations originated from three contrasting regions (enzootic, epidemic/epizootic and YFV-free areas) were challenged with YFV: four Ae. albopictus, four Ae. aegypti, one Haemagogus leucocelaenus (Dyar & Shannon) and one Sabethes albiprivus Theobald (Fig. 6, Table 4). We tested paired Ae. albopictus and Ae. aegypti populations simultaneously sampled in the same area (i.e. Brazil and Congo). Similarly, four species (Ae. albopictus, Ae. aegypti, Hg. leucocelaenus and Sa. albiprivus) collected in the Rio de Janeiro area were tested (Table 1). Populations were derived from eggs collected with ovitraps settled on the ground for sampling Ae. albopictus and Ae. aegypti in the urban and suburban sites, or suspended at the forest canopy at 5-16 m high for collecting Hg. leucocelaenus. When possible, the first generation in laboratory after collection (F1) of Ae. aegypti and Ae. albopictus was used for experimental infections. For Hg. leucocelaenus which cannot be maintained in laboratory conditions 34,35 , the F0 generation derived from eggs collected fortnightly in 2015 was directly used for infections. In the case of Sa. albiprivus, mosquitoes from a colony established in the laboratory since 2013 were used.

Methods
Ovitraps were provided with 1 to 3 wooden paddles. Eggs were hatched by submerging paddles in dechlorinated tap water for two consecutive days. Larvae were reared in pans (25 × 25 × 10 cm) containing one liter of   Mosquitoes were processed as follows: abdomen and thorax (herein after referred to as body) were tested for determining infection, head for dissemination and saliva for transmission. To determine viral infection and dissemination rates, each mosquito body and head were respectively ground in 500 μL and 300 μL of Leibovitz L15 medium (Invitrogen) supplemented with 2% fetal bovine serum (FBS), centrifuged at 10,000 × g for 5 min at +4 °C and inoculated onto monolayers of Ae. albopictus C6/36 cell culture in 96-well plates. After 1 h incubation at 28 °C, 150 μL of 2.4% CMC (carboxymethyl cellulose) in Leibovitz L15 medium supplemented with 10% FBS was added per well. After 5 days incubation at 28 °C, cells were fixed with 10% formaldehyde, washed, and revealed using hyperimmune ascetic fluid specific to YFV as the primary antibody and Alexa Fluor 488 goat anti-mouse IgG as the second antibody (Life Technologies) 28 . To estimate viral transmission, mosquito saliva was collected in individual pipette tips containing 5 μL FBS for 30 min as previously described 40 . FBS containing mosquito saliva was expelled into 45 μL of Leibovitz L15 medium, inoculated on Ae. albopictus C6/36 cell culture in 96-well plates and stained as described above.
Statistical analysis. Rates (infection, disseminated infection, and transmission) were described using median and inter-quartile range (IQR). The effect of species, population, YFV strain and duration on the rates was investigated using logistic linear regression models. Two-by-two interaction between species, population, YFV strain and duration was systematically investigated. Statistical analyses were conducted using the Stata software (StataCorp LP, Texas, and USA). P-values < 0.05 were considered significant.