Inferring the ecological niche of bat viruses closely related to SARS-CoV-2 using phylogeographic analyses of Rhinolophus species

Genetic analyses of Rhinolophus species identified as reservoirs of viruses closely related to SARS-CoV-2

Until now, SCoV2rCs have been found in four bat species of the genus Rhinolophus: R. acuminatus, R. affinis, R. malayanus, and R. shameli. The haplotype networks constructed using CO1 sequences of these four species are shown in Fig. 3. A star-like genetic pattern, characterized by one dominant haplotype and several satellite haplotypes was found for the two bat species endemic to Southeast Asia, i.e. R. acuminatus and R. shameli.

In the network of R. acuminatus, the most common haplotype (named Rac1 in online supplementary Table S1) was found in northern Cambodia, southern Laos, eastern Thailand and southern Vietnam, indicating recent gene flow among these populations. Since a virus related to SARS-CoV-2 (91.8% of genome identity), named RacCS203, was detected in five R. acuminatus bats caught in eastern Thailand in June 2020⁶, the genetic pattern obtained for this species suggests that viruses closely related to RacCS203 may have circulated in most southern regions of mainland Southeast Asia. In contrast, R. acuminatus bats collected in Borneo (M5) showed a divergent haplotype (separated by 12 mutations; haplogroup II), suggesting that the South China Sea between mainland Southeast Asia and Borneo constitutes a barrier to gene flow. Isolated populations of R. acuminatus described in northern Myanmar, Indonesia (Java and Sumatra) and the Philippines¹⁴ should be further studied.

The network of R. shameli shows a typical star-like pattern, the most common haplotype (named Rsh1 in online supplementary Table S1) being detected in northern Cambodia and Laos. Since a virus related to SARS-CoV-2 (93.1% of genome identity), named RshSTT200, was recently discovered in two R. shameli bats collected in northern Cambodia in December 2010⁷, the genetic pattern obtained for this species suggests that viruses closely related to RshSTT200 may have circulated, at least in the zone between northern Cambodia and central Laos. The bats sampled south to the Tonle Sap lake (n = 4; southern Cambodia and Vietnamese island of Phu Quoc) were found to be genetically isolated from northern populations (four mutations). However, further sampling in the south is required to confirm this result, as it may reveal CO1 sequences identical to the haplotypes detected in the north.

For the two species distributed in both China and Southeast Asia, i.e. R. affinis and R. malayanus, the genetic patterns are more complex with different haplogroups showing more than 1% of nucleotide divergence. In the network of R. affinis, there are three major haplogroups (named I, II and III in Fig. 3) separated by a minimum of seven mutations. The results are therefore in agreement with those previously published using CO1 and D-loop mitochondrial sequences¹⁵. The CO1 haplotypes detected in the localities sampled in southern China (ch1, ch4, ch5) are distantly related to the single haplotype available for central China (ch6), but they are also found in Laos, northern and central Vietnam, northern Thailand and northeastern Myanmar. This result suggests recent gene flow between populations from southern Yunnan and those from northern mainland Southeast Asia. Since a virus related to SARS-CoV-2 (96.2% of genome identity), named RaTG13, was detected in one R. affinis bat captured in southern Yunnan in 2013¹, the genetic pattern obtained for this species suggests that viruses closely related to RaTG13 may have circulated in the zone comprising southern Yunnan and northern mainland Southeast Asia.

In the network of R. malayanus, there are four major haplogroups (named I, II, III and IV in Fig. 3) separated by a minimum of seven mutations. The CO1 haplotypes detected in the localities sampled in southern China (ch2 and ch3) were also found in northern Laos (L1 and L3), suggesting recent gene flow between populations from these two countries. Since a virus related to SARS-CoV-2 (93.7% of genome identity), named RmYN02, was recently isolated from one R. malayanus bat collected in southern Yunnan in June 2019⁵, the genetic pattern obtained for this species suggests that viruses closely related to RmYN02 may have circulated, at least between southern Yunnan and northern Laos. In contrast, the bats sampled in Myanmar were found to be genetically isolated from other geographic populations (haplogroup II in Fig. 3).

Two different ecological niches for bat viruses related to either SARS-CoV or SARS-CoV-2

In the wild, sarbecoviruses were generally detected after examining fecal samples collected on dozens of bats. For instance, two sarbecoviruses were found⁷ among the total 59 bats collected at the same cave entrance in northern Cambodia in 2010 (unpublished data). However, this does not mean necessarily that sarbecoviruses were absent in negative samples, as degradation of RNA molecules and very low viral concentrations may prevent the detection of RNA viruses. Despite these difficulties, full genomes of Sarbecovirus have been sequenced from a wide diversity of horseshoe bat species collected in Asia, Africa and Europe^5,6,7,8,9,10. Therefore, there is no doubt that Rhinolophus species constitute the natural reservoir host of all sarbecoviruses^3,8. The genus Rhinolophus currently includes between 92¹¹ and 109¹⁶ insectivorous species that inhabit temperate and tropical regions of the Old World, with a higher biodiversity in Asia (63–68 out of the 92–109 described species) than in Africa (34–38 species), Europe (5 species) and Oceania (5 species). Although some Rhinolophus species are solitary, most of them are gregarious and live in large colonies or small groups generally in caves and hollow trees, but also in burrows, tunnels, abandonned mines, and old buildings^11,16. However, they prefer large caves with total darkness, where temperatures are stable and less affected by diurnal and seasonal climatic variations. Importantly, all Rhinolophus species in which sarbecoviruses were detected in previous studies^{1,5,6,7,8,9,17} are cave species that form small groups or colonies (up to several hundreds)^11,18,19.

In China, many SCoVrCs were previously detected in several horseshoe bat species, including Rhinolophus sinicus, Rhinolophus ferrumequinum (currently R. nippon)¹⁶, Rhinolophus macrotis (currently R. episcopus)¹⁶, Rhinolophus pearsoni, and Rhinolophus pusillus, and it has been shown that they circulate not only among conspecific bats from the same colony, but also between bat species inhabiting the same caves^17,20,21. The ecological niche predicted for bat SCoVrCs using a data set of 19 points (see online supplementary Table S2) is shown in Fig. 4. The AUC was 0.81. The value was > 95% CI null-model’s AUCs (0.68), indicating that the model performs significantly better than a random model (see online supplementary Fig. S1). The highest probabilities of occurrence (highlighted in green in Fig. 4) were found in Nepal, Bhutan, Bangladesh, northeastern India, northern Myanmar, northern Vietnam, most regions of China south of the Yellow River, Taiwan, North and South Korea, and southern Japan.

In Southeast Asia and southern China, SCoV2rCs have currently been found in four Rhinolophus species (R. acuminatus, R. affinis, R. malayanus and R. shameli)^1,6,7,8, but the greatest diversity of horseshoe bat species in mainland Southeast Asia (between 28 and 36 species)^11,16 suggests that many sarbecoviruses will be discovered soon. Despite the limited data currently available on SCoV2rCs, several arguments support that bat intraspecific and interspecific transmissions also occur with SCoV2rCs. Firslty, recent genomic studies have revealed that SCoV2rCs circulate and evolve among horseshoe bats of the same colony, as five very similar genomes (nucleotide distances between 0.03% and 0.10%) were sequenced from five R. acuminatus bats collected from the same colony in eastern Thailand⁶, and as two genomes differing at only three nucleotide positions (distance = 0.01%) were sequenced from two R. shameli bats collected at the same cave entrance on the same night⁷. Secondly, the discovery of four viruses closely related to SARS-CoV-2 (between 96.2 and 91.8% of genome identity) in four different species of Rhinolophus is a strong evidence that interspecific transmission occurred several times in the past. As detailed in online supplementary Table S1, these species were collected together in several localities of Cambodia (three species in C1, C2, and C5; two species in C8), Laos (four species in L10; three species in L9; two species in L1, L5, L8, L11), and Vietnam (two speciess in V10, V9, V17, V18). These data corroborate previous studies suggesting that sarbecoviruses can be transmitted, at least occasionally, between Rhinolophus species sharing the same caves.

The ecological niche of bat SCoV2rCs was firstly predicted using the four localities where bat viruses were previously detected^1,6,7,8 (Fig. 5a). The highest probabilities of occurrence (highlighted in green in Fig. 5a) were found in Southeast Asia rather than in China. However, the AUC was only 0.58, and the value was < 95% CI null-model’s AUCs (0.74), indicating that the model was not statically supported at a significance level of 0.05 (see on line supplementary Fig. S2). As expected, these preliminary results confirmed that more than four records are needed to increase the accuracy of the distribution model²². For that reason, we used a genetic approach to determine geographic localities where bat SCoV2rCs are more likely to be found. The CO1 sequences of the nine bats in which a SCoV2rC was detected are shown by red arrows in Fig. 3. For R. affinis, the CO1 haplotype sequenced for the bat found positive for a SCoV2rC in southern Yunnan (site named Ch1 in Figs. 2, 3) was not found in other sampled localities. For the three other bat species found positive for SCoV2rCs, identical CO1 sequences were detected in bats from 17 other geographic localities (see online supplementary Table S1). For R. acuminatus, the four CO1 haplotypes sequenced for the bats found positive for SCoV2rCs in eastern Thailand (site named T4 in Figs. 2, 3) were also found in four localities in Cambodia (C2, C4, C5, and C6), two localities in southern Laos (L10 and L11), and three localities in southern Vietnam (V13, V14, and V17). The results indicate high connectivity among R. acuminatus populations from eastern Thailand, Cambodia, southern Laos and southern Vietnam. For R. malayanus, the CO1 haplotype sequenced for the bat found positive for a SCoV2rC in southern Yunnan (site named Ch3 in Figs. 2, 3) was also found in another locality in southern Yunnan (Ch2) and two localities in northern Laos (L1 and L3). The results indicate high connectivity among R. malayanus populations from southern China and northern Laos. For R. shameli, the single CO1 haplotype sequenced for the two bats found positive for SCoV2rCs in northern Cambodia (site named C3 in Figs. 2, 3) was also found in three other localities in Cambodia (C1, C2, and C5) and five localities in Laos (L4, L6, L8, L9, and L10). The results indicate high connectivity among R. shameli populations from Cambodia and Laos. Based on these genetic data, the ecological niche of bat SCoV2rCs was predicted using 21 records corresponding to the four localities where bat viruses were previously detected^1,6,7,8 and the 17 localities where bats showed the same CO1 haplotype than virus-positive bats (data set B: 21 points; see online supplemntary Table S1 for details). The AUC was 0.96. The value was > 95% CI null-model’s AUCs (0.81), indicating that the model performs significantly better than a random model (see online supplementary Fig. S3). The areas showing the highest probabilities of occurrence (highlighted in green in Fig. 5b) include four main geographic areas: (i) southern Yunnan, northern Laos and bordering regions in northern Thailand and northwestern Vietnam; (ii) southern Laos, southwestern Vietnam, and northeastern Cambodia; (iii) the Cardamom Mountains in southwestern Cambodia and the East region of Thailand; and (iv) the Dawna Range in central Thailand and southeastern Myanmar.

Our results show that bat SCoVrCs and SCoV2rCs have different ecological niches: that of SCoVrCs covers mainly China and several adjacent countries and extends to latitudes between 18° and 43°N, whereas that of SCoV2rCs covers northern mainland Southeast Asia and extends to latitudes between 10° and 24°N. Most Rhinolophus species involved in the ecological niche of SCoVrCs have to hibernate in winter when insect populations become significantly less abundant. This may be different for most Rhinolophus species involved in the ecological niche of SCoVrC2s. Since this ecological difference may be crucial for the dynamics of viral transmission among bat populations, it needs to be further studied through comparative field surveys in different regions of China and Southeast Asia. The ecological niches of SCoVrCs and SCoV2rCs slightly overlap in the zone including southern Yunnan, northern Laos, and northern Vietnam (Figs. 4, 5b). This zone corresponds to the northern edge of tropical monsoon climate²³. Highly divergent sarbecoviruses of the two main lineages SCoVrCs and SCoV2rCs are expected to be found in sympatry in this area. This is confirmed by the discovery of both SCoVrCs and SCoV2rCs in horseshoe bats collected in southern Yunnan^1,6,21. Collectively, these data suggest that genomic recombination between viruses of the two divergent lineages are more likely to occur in bats roosting, at least seasonally, in the caves of these regions. Since highly recombinant viruses can threaten the benefit of vaccination campaigns, southern Yunnan, northern Laos, and northern Vietnam should be the targets of closer surveillance.

Mainland Southeast Asia is the cradle of diversification of bat SCoV2rCs

Chinese researchers have actively sought sarbecoviruses in all Chinese provinces after the 2002–2004 SARS outbreak. They found many bat SCoVrCs^16,20,21 but only two SCoV2rCs^1,5 and both of them were discovered in southern Yunnan, the Chinese province bordering Southeast Asia. The ecological niches predicted herein for bat sarbecoviruses suggest that SCoVrCs are dominant in China (Fig. 4) while SCoV2rCs are present mostly in Southeast Asia (Fig. 5). This means that viruses similar to SARS-CoV-2 have been circulating for several decades throughout Southeast Asia, and that different species of bats have exchanged these viruses in the caves they inhabit. The data available on human cases and deaths caused by the COVID-19 pandemic² indirectly support the hypothesis that the cradle of diversification of bat SCoV2rCs is mainland Southeast Asia, and in particular the areas highlighted in green in Fig. 5b. Indeed, human populations in Cambodia, Laos, Thailand, and Vietnam appear to be much less affected by the COVID-19 pandemic than other countries of the region, such as Indonesia, Malaysia, Myanmar, and the Philippines (Fig. 6). This suggests that some human populations of Cambodia, Laos, Thailand, and Vietnam, in particular rural populations living in contact with wild animals for several generations, have a better immunity against SCoV2rCs because they have been regularly contaminated by bats and/or infected secondary hosts such as pangolins.

Pangolins contaminated by bats in Southeast Asia

Apart from bats, the Sunda pangolin (Manis javanica) and Chinese pangolin (Manis pentadactyla) are the only wild animals in which viruses related to SARS-CoV-2 have been found so far. However, these discoveries were made in a rather special context, that of pangolin trafficking. Several sick pangolins were seized by Chinese customs in Yunnan province in 2017 (unpublished data), in Guangxi province in 2017–2018²⁴ and in Guangdong province in 2019²⁵. Even if the viruses sequenced in pangolins are not that close to SARS-CoV-2 (one was 85% identical and the other 90%), they indicate that at least two sarbecoviruses could have been imported into China well before the emergence of COVID-19 epidemic. Indeed, it has been shown that Sunda pangolins collected from different Southeast Asian regions have contaminated each other while in captivity on Chinese territory³. It has been estimated that 43% of seized pangolins were infected by at least one SARS-CoV-2-like virus³. Such a high level of viral prevalence and the symptoms of acute interstitial pneumonia detected in most dead pangolins²⁴ indicate that captive pangolins are highly permissive to infection by SARS-CoV-2-like viruses. The question remained on how the Sunda pangolins became infected initially. Could it have been in their natural Southeast Asian environment, before being captured? The discovery of two new viruses close to SARS-CoV-2 in bats from Cambodia and Thailand^7,8 supports this hypothesis, as Rhinolophus bats and pangolins can meet, at least occasionally, in forests of Southeast Asia, possibly in caves, tree hollows or burrows. Further substantiating this hypothesis, the geographic distribution of Manis javanica²⁶ overlaps the ecological niche here predicted for bat SCoV2rCs (Fig. 5), and SARS-CoV-2 neutralizing antibodies have been recently detected in one of the ten pangolin sera sampled from February to July 2020 from three wildlife checkpoint stations in Thailand⁶. Collectively, these data strengthen the hypothesis that pangolin trafficking is responsible for multiple exports of viruses related to SARS-CoV-2 to China³.

Source: Ecology - nature.com