South Korea COVID-19 outbreak and response
When the Wuhan government first reported 27 cases of pneumonia on December 31, 2019, the South Korean government immediately issued the lowest threat alert (blue) and started to screen air travelers from Wuhan using infrared thermometers and a pan-corona virus PCR kit. Upon discovery of the first confirmed case on January 20, the government promptly raised the threat level to yellow and subsequently to orange on January 27 following the identification of a fourth case (Fig. 1). By February 17, 30 cases were identified, and 10 had already recovered. Additionally, 11 cases were confirmed to be caused by close contact, and 19 were confirmed to be imported (Table 1).
The first instance of a massive outbreak in the church was identified on February 18, 2020. This case was a member of the church without any international travel records. Ten close contacts of this individual, mainly including church members, were confirmed on the following day, and positive cases from the church subsequently increased rapidly (Fig. 1). On February 20, another outbreak was identified at a hospital complex, and subsequently, the first COVID-19-related death was reported. By February 22, the total number of cases rose to 346, among which 169 (48.8%) and 108 (31.2%) cases were related to the church outbreak and the regional hospital outbreak, respectively.
The government started screening all church members in Daegu and all individuals at the hospital complex. When the cases rose to 600 on February 23, the government raised the alert level to red, officially declaring a national emergency (Fig. 1). On February 25, the Korean Government started massive contact tracing and real-time RT-PCR testing on the church members in South Korea, completing around 240,000 tests in just 16 days by March 12 (Fig. 1). The resulting exponential increase in cases, however, led to a shortage of negative pressure isolation rooms and overall burden on the South Korean healthcare system. Following this peak, there were several minor outbreaks related to the known cases, but all were successfully contained owing to rapid screening efforts and preemptive response measures for vulnerable groups. Additionally, instances continued to be imported, particularly from Europe and North America, and travel-related cases became greater than 10% of the total daily confirmed cases on March 31. In response, the government enforced a mandatory 14-day self-quarantine for all travelers on April 1 (Fig. 1). On May 8, one local case was identified, whereas 11 were deemed imported.
Demographics of patients with COVID-19
As of May 8, approximately 60% of total confirmed cases were female; however, the mortality rate was higher in males than in females (3.02% versus 1.91%, respectively; Table 1). Mortality rates also increase with age (0% and 1% for individuals below 30 years old and below 60 years old, respectively; 25% for individuals 80 years old and older). Additionally, 80.5% of total cases were part of local cluster outbreaks, and 10.8% were imported. Notably, 5,212 confirmed cases related to the church outbreak were localized around Daegu (dark blue in Fig. 2), accounting for 48.1% of total confirmed cases in South Korea.
Genomic investigation cohort
We performed SARS-CoV-2 whole genome sequencing on 66 cases to investigate associations between the church outbreak and the initial cases within South Korea (Table 2). Short-read high-throughput sequencing was performed for whole genome coverage. The total number of filtered reads, number of reads mapped to the reference genome (Wuhan-Hu-1, NCBI Reference Sequence: NC_045512, GenBank accession number: MN908947), base coverage, and average sequencing depth are presented in Supplementary Table S1. As detailed in Table 2, we obtained viral sequences from i) 15 initial cases (out of 30) prior to the church outbreak, ii) 37 cases (out of 311) within the first 5 days of the church outbreak, and iii) 14 other cases presumed to be related to the church outbreak by contact tracing records.
Our genomic investigation estimated five transmission clusters among the 15 cases prior to the church outbreak (Fig. 3). The first cluster (C1) contained four cases, i.e., cases 5174, 7156, 1534, and 7834, where all but case 5174 showed sequences identical to that of one of the earliest cases (Wuhan-Hu-1 strain: NCBI Reference Sequence: NC_045512, GenBank accession number: MN908947), and case 5174 harbored one nonsynonymous mutation (G26144T).6 This aligned with the reported contact tracing results (Fig. 3a), as follows: i) case 7156 had interacted with a Wuhan resident and was a family member of case 1534; ii) case 7834 had met a confirmed case from Wuhan, and iii) case 5174 was from Wuhan. Taken together, all four individuals in C1 were either closely linked to Wuhan or had close contact of Wuhan cases, revealing a strong correlation between our genomic investigation and the epidemiological tracing.
The second transmission cluster (C2) consisted of cases 9273, 3542, 5431, 4536, 2314, a and 4023 (Fig. 3b). Case 9273 had been working in Wuhan when case 3542 had visited, and the sequence of the virus from case 9273 showed two mutations compared with case 3542 and case 5431, but shared two common mutations from the C1 founder sequence. Case 3542 had met case 5431, and the genomic sequences of the virus for these cases were identical. Case 4023 was closely related to case 5431, with the sequences showing a difference of only one mutation. Case 4536 had worked in Wuhan prior to diagnosis in South Korea and interacted with case 2314; the genomic sequence of the virus from case 4536 showed one mutation compared with that of case 3542, whereas cases 4536 and 2314 had two nucleotide substitutions. Therefore, all confirmed close-contact cases had either identical sequences or sequence pairs with one nucleotide difference, whereas potential exposure pairs showed only two mutations.
The three other sequence clusters were also consistent with contact tracing data (Fig. 3). The third transmission cluster (C3) has only one member, case 24, a Wuhan resident who did not have close contact with any of the other cases in South Korea. The fourth (C4) and fifth (C5) clusters each consisted of two members, who shared six and eight common mutation markers from the first cluster, respectively. Thus, our maximum likelihood tree constructed from mutation signatures was consistent with epidemiological contact tracing records, demonstrating the effectiveness of genomic sequencing in discerning close-contact transmission clusters.
Church outbreak analysis
Fig. 4 shows the maximum likelihood tree of 51 cases after the church outbreak alongside the 15 initial cases that were sequenced. Most post-outbreak cases were phylogenetically distinct from the prior cases and formed two closely related, yet independent clusters: C6 (Fig. 4a, dark blue brackets), which mainly consisted of church members in Daegu; and C7 (Fig. 4a, light blue brackets), which consisted of all and only cases from the hospital.
Additionally, both clusters displayed star-like phylogeny (Fig. 4b and 4c), a signature of random neutral evolution.7 Both clusters had multiple identical sequences at the tree center (13 out of 37 for cluster 6 and six out of 11 for cluster 7), an evolutionary pattern frequently observed in acute-stage human immunodeficiency virus sequences within an infected individual.8,9 This group of identical sequences is a putative founder of each transmission cluster and only one mutation away (G5572T nonsynonymous mutation for cluster 6 and C26681T synonymous mutation for cluster 7) from their most recent common ancestor sequence (Fig. 4a, red circle). The independent and random evolution patterns observed with the star-like phylogeny suggested allopatric speciation or diversification under geographic isolation. This was plausible given that most cases from the hospital occurred within the long-term care unit, which was isolated from the rest of the complex. Thus, the initial transmission of the two clusters may have occurred simultaneously at the hospital, based on the following observations: i) the founder strains of the two clusters were one base substitution only from their most probable common ancestor; ii) the common ancestor was not identified, despite the abundance of the two founders; and iii) the hospital had an isolated setting.
Additionally, the church cluster (C6) and hospital cluster (C7) were unique, sharing common markers from the Wuhan-Hu-1 strain (two nonsynonymous mutations, G11083T and G26144T, and one synonymous mutation, C14805T). As of May 8, 2020, no sequences within the two clusters matched any of the those registered in GISAID. Moreover, no GISAID sequences matched that of the most probable ancestor. However, we were able to identify a pair of identical strains that shared two common mutation markers (Fig. 4a, blue circle) with this ancestor. These strains (GISAID access numbers: EPI_ISL_410546 and EPI_ISL_412974) were isolated from individuals who traveled from Wuhan to Italy, suggesting that Wuhan may have been the origin of the church and hospital clusters.
The church cluster (C6) was a likely transmission origin of the other outbreaks that occurred outside Daegu, displaying genomic linkage with a transmission cluster associated with case 3407, that associated with case 2397, that with case 2058, and that with case 7239. Similarly, our genomic investigation suggested that multiple sporadic cases, such as cases 4237 and 2697, could also be attributed to the church cluster (cluster 6).