A total of 160 samples (81 stools and 79 sputum samples) from 62 patients with positive results by RT-PCR were included this study. Of these, 45 samples (36 sputum samples and 9 stools) from 29 patients (17 men and 12 women, with a median age of 32 years) yielded amplicons for the 5 targeted loci (Table 2). The patients came from 2 SARS-designated hospitals in Beijing, with disease onset ranging from March to May, 2003. Four patients had serious conditions during hospitalization, including pulmonary aggravation requiring oxygen ventilation or transfer to an intensive care unit. No patient died.

The sequences of the 45 positive specimens were compared with SARS-CoV genome sequences available from the public database (GenBank). The sequence variants in 5 loci (17564, 21721, 22222, 23823, and 27827) defined 3 kinds of motifs: GGCTC, TGTTT, and GATTC (Table 2). In addition, 4 new SNVs were identified at nucleotides 17620, 22077, 22589, and 27749 in >1 patient. These variations appeared independently in several isolates, which indicates that they are not RT-PCR artifacts. None of them had been previously reported, with 3 nucleotide substitutions leading to amino acid changes (Table 3).

Twelve patients in this study belonged to a cluster. They derived from an SSE indirectly linked with the earliest SARS patients in Beijing. The first 2 patients of this cluster, who became ill on March 10 and 21, respectively, harbored the GGCTC motif. The remaining patients, who became ill from March 31 to May 4, showed the TGTTT motif. Among patients outside of the cluster, 5 of 6 patients with onset date before April had the GGCTC motif, while the TGTTT motif became predominant later (9 of 11 patients until May 12). A new motif, GATTC, was found in 2 patients outside the cluster. In addition, no intrapatient variation was observed in the 5 amplicons from specimens collected at different times or from different sources (sputum or stools).

The possible role of genetic mutations in patients' prognosis was also investigated. The presence of nucleotide substitution was compared between 2 groups of patients: 1 with good prognosis (absence of pulmonary aggravation; n = 25) and 1 with adverse outcome (pulmonary aggravation 8–12 days after onset of symptoms requiring oxygen ventilation or transfer to ICU; n = 4). No mutation was found associated with disease severity (Table 2).

During the 2003 SARS epidemic, conventional epidemiologic investigation, aided by viral sequencing analysis, identified viral genetic signatures that are linked to geographic and temporal clusters of infection (4,10–12,15–18). Findings of these studies are summarized in the Figure, connecting the worldwide epidemic to a transmission event in hotel M in Hong Kong in late February 2003.

Beijing had experienced the SARS epidemic from March to June; however, only a few Beijing strains from the early epidemic have been analyzed in previous studies. Our study is the first to provide phylogenetic information on Beijing strains from the early and middle epidemic, as well as the late epidemic, by using the 5-locus motif of previous studies. The series of mutations in the 5-locus motif observed in Beijing followed the same path as isolates in Guangdong Province and the worldwide epidemic, i.e., the early introduction of GACTC motif was followed by transition to a GGCTC motif, before switching to a stable TGTTT motif. The observation of the same series of mutations occurring in 2 separate locations at different times suggests a dominant process of viral adaptation to the host. Moreover, this finding can expand our understanding of SARS-CoV response to selection pressures in humans, since early Beijing isolates (BJ01, BJ02, and BJ 03), which are traceable to Guangdong, underwent an independent selection process and would not be subject to the same sampling bias caused by superspreading events in Hong Kong isolates. The GGCTC→TGTTT switch was observed among patients belonging to the same cluster in this study, which rules out the possibility of the coincidental superposition of 2 epidemics (GGCTC and TGTTT) coexisting in Beijing.

The mutations involved in the GGCTC→TGTTT switch are responsible for amino acid changes in a nonstructural protein (17564, region Orf1b) in S protein (21721 and 22222) and in a noncoding region (27827, X3). We were not able to identify a correlation between these changes and the clinical status of patients. We did not find sequence variations in specimens obtained from the same patients either collected at different times or among different specimen types, which suggests that within-individual variations are rare in the partial genome of this study, although the phenomenon was described in a previous study (15). A new motif, GATTC, which represents a new transitional motif between GACTC and TGTTT, was described on 2 occasions in patients who were not part of the cluster. Similarly, 4 new SNVs were identified at nucleotides 17620, 22077, 22589, 27749.

In summary, this study confirms the evolution of SARS-CoV strains towards a TGTTT motif in positions 17564, 21721, 22222, 23823, and 27827 in Beijing, as was observed in Guangdong province before the hotel M outbreak in Hong Kong. Whether this motif is associated with higher transmission or virulence remains to be elucidated.