Here, we combine retrospective molecular clock analysis in a coalescent framework with a forward compartmental epidemiological model to estimate the timing of the SARS-CoV-2 index case in Hubei province. The inferred dynamics during these unobserved early days of SARS-CoV-2 highlight challenges in detecting and preventing nascent pandemics.
The high extinction rates we inferred suggest that
spillover of SARS-CoV-2-like viruses may be frequent, even if pandemics are rare (
28). Further, the same dynamics that characterized the establishment of SARS-CoV-2 in Hubei province may have played out all over the world as the virus was repeatedly introduced, but only occasionally took hold (
29,
30). The reports of cases in December 2019 and January 2020 in France and California that did not establish sustained transmission fit this pattern (
31?
33).
However, our results suggest that PCR evidence of SARS-CoV-2 in wastewater outside of China before November 2019 is unlikely to be valid (34) and the suggestion of international spread in mid-November or early-December 2019 should be viewed with skepticism (
35?
37), given that our results suggests fewer than 20 people infected with SARS-CoV-2 at this time (table S4 and fig. S11).
Our results also refute claims (38) of large numbers of patients requiring hospitalization due to COVID-19 in Hubei province prior to December 2019 (figs. S13 and S14). Nevertheless, SARS-CoV-2 may be detectable should archived wastewater samples or other biomaterials from Hubei province exist from early-to-mid November 2019, and incorporating these types of data in our model could further refine our timing estimates. Moreover, wastewater detection may present the best chance of early detection of future pandemics during the early phase of spread where we estimate very low numbers of infections.
Even though all of the earliest documented cases of COVID-19 were found in Hubei province, we cannot discount the possibility that the index case initially acquired the virus elsewhere. Nonetheless, our dating inference is insensitive to geography.
Further, our results suggest that if the virus first emerged in a rural community, it would have needed to migrate to an urban setting to avoid extinction.
The lack of reports of COVID-19 elsewhere in China in November and early-December suggest Hubei province is the location where human-to-human transmission chains were first established.
The circumstances surrounding the emergence of SARS-CoV-2 in Hubei province remain shrouded. Although SARS-CoV-2 is repeatedly adapting to spread among humans (
40,
41), our findings do not reveal whether the virus that first emerged was less fit than the virus that spread throughout China. Nevertheless, the inferred timing of the index case is generally similar in both of these scenarios, because less-fit viruses in our simulations that went extinct tended to do so very quickly.
Importantly, it is yet unknown whether the virus emerged directly from its animal reservoir, presumably horseshoe bats (42, 43), or first circulated in and possibly adapted to an intermediate host. Our estimates for the timing of the Hubei index case further distance this individual from the outbreak at the Huanan Seafood Wholesale Market. Finding the animal reservoir, or hypothetical intermediate host, will help to further narrow down the date, location, and circumstances of the original SARS-CoV-2 infection in humans. However, even in the absence of that information, coalescent-based approaches permit us to look back beyond the tMRCA and toward the earliest days of the COVID-19 pandemic. Although there was a pre-tMRCA fuse to the COVID-19 pandemic, it was almost certainly very short. This brief period of time suggests that future pandemics with similar characteristics to the COVID-19 pandemic permit only a narrow window for preemptive intervention.
https://science.sciencemag.org/content/early/2021/03/17/science.abf8003.full