The source of SARS-CoV, how it was introduced into humans, and where it may reemerge are critical questions related to disease control (16
). Understanding the mode of transmission of SARS-CoV into humans is essential in designing appropriate prevention and control measures for future SARS epidemics. We provide the first direct evidence that SARS-CoV was transmitted from palm civets to humans, and that a restaurant serving palm civets positive for this virus was the source of infection for 2 of 4 confirmed SARS patients during the resurgence of SARS in the winter of 2003–2004 (6,17
All 6 palm civets from the restaurant were positive for SARS-CoV. Partial S gene sequences were identical in both patients from this study and to 4 of 5 S gene sequences from palm civets from the restaurant, but different from more than 100 S gene sequences from SARS patients worldwide (6,10,12,13
). That the restaurant was an infection source was further supported by serologic investigation of restaurant employees. Specific IgG was detected in 2 of 39 employees, 1 with a history of close contact with these palm civets. However, we lack evidence that eating civet could transmit the virus because the employees had not eaten palm civet before SARS developed. The patients most likely were infected by close exposure to animals carrying SARS-CoV in the restaurant. This situation may be similar to those earliest index cases linked to markets or restaurants that occurred in winter of 2002–2003 (18
). Results of PCR tests conducted by the WHO were positive for SARS-CoV in specimens from the bottom of animal cages and the kitchen of the restaurant (19
Genome sequence analysis data strongly suggest that sporadic cases of SARS in Guangzhou in 2003–2004 were caused by SARS-CoV of animal origin. The 29-nt deletion was not observed in palm civets from the restaurant, but was present in almost all human isolates, and may have resulted from the adaptation and evolution of SARS-CoV in humans. SNVs in S gene sequences have been reported in several studies of the molecular evolution of SARS-CoV (6,7,12,13
). The characteristic SNV pattern of S genes has 21 nt. SARS-CoV isolated from palm civets at the restaurant had 0, 1, or 2 SNVs. However, viruses from several provinces of China and other countries had all 21 SNVs (7,10,14,15
). Viruses isolated in the early phase of the 2003 epidemic had 16–19 SNVs (6,7,13
). The SZ3 and SZ16 isolates from palm civets in 2003 had 11 SNVs (6
), while the S gene from the first case of SARS encountered in 2004 had only 3 SNVs () (13
). When the complete genomes of SARS-CoV from palm civets at the restaurant were compared with sequences of human isolates, 62 SNVs were identified. However, when the complete genome was compared with sequences of virus isolated from palm civets from animal markets in the 2003 epidemic, only 37 SNVs were identified.
Phylogenetic analysis of the S gene of SARS-CoV also showed that viruses from palm civets at the restaurant were more closely related to previously described viruses of animal origin, and these were more closely related to viruses isolated from patients during the early epidemic phase. Moreover, all SARS-CoV strains, including isolates from animal markets, had evolved from isolates in palm civets at the restaurant (). Clearly, SARS cases contracted at the restaurant were the result of recent interspecies transfer from a putative palm civet virus reservoir, rather than the result of circulation of SARS-CoV in the human population.
SNV and phylogenetic analysis also suggest that the virus responsible for SARS infections in 2004 was not yet able to cause severe disease in humans. Minor clinical symptoms and no subsequent transmission have been recognized as features of the recent SARS infections. These findings support our observations that SARS-like illness did not develop in any of the 257 contacts of the 4 patients, or in any of the health care workers attending them. However, epidemiologic data can only provide clues to the biologic characteristics of the virus. Therefore, experimental infection using animal models is necessary to measure the relative pathogenic potential of various strains of SARS-CoV isolated from human and animals.