Being in the epicenter of the COVID-19 pandemic, NewYork-Presbyterian Queens Hospital has received 185,870 specimens for SARS-CoV-2 RNA testing by diagnostic multiplex real-time PCR since the COVID-19 pandemic started in March 2020. In addition to those specimens, we analyzed specimens collected by mobile service vans from residential communities and workplaces. We used demographics, including residential and business addresses associated with collection sites, for the epidemiology analysis. We considered households as persons at the same residential address, identical business addresses as workplace, and family members as family. We determined traveler status by the home address and traveling inquiry performed during sampling. We advised all patients testing positive or exposed patients to follow the Centers for Disease Control and Prevention (CDC) quarantine guidelines.

To investigate SARS-CoV-2 mutations and variant epidemiology, we performed next-generation sequencing (NGS) on the positive specimens with real-time PCR cycle threshold (Ct) value <33 cycles and analyzed virus mutations among the specimens from our laboratory and LabQ Diagnostics (New York, New York, USA). We performed NGS by using the Illumina COVID-Seq test kit (https://www.illumina.com). We extracted viral RNA from a viral transport medium containing a nasopharyngeal swab specimen, then performed cDNA synthesis through reverse transcription using random hexamer primers. We amplified the cDNA of the viral genome by 2 separate PCR reactions and pooled the products together. The fragments underwent bead-based tagmentation to the adaptor sequences. Subsequently, the adaptor-tagged fragments underwent another round of PCR amplification. Using the purification beads, we pooled and cleaned the indexed tagged libraries. We clustered pooled libraries onto a flow cell and then sequenced on the NovaSeq 6000 sequencing system (Illumina). We used VarSeq version 2.2.2 (Golden Helix, https://www.goldenhelix.com) for sequence analysis; we used consensus sequences of these viruses as input to Nextclade version 1.10.1 (19) for quality control, mutation calling, and Nextstrain clade assignment. Viruses <29,000 nt in length or with Nextclade-assessed QC.overallStatus below good were considered low quality and removed.

To investigate the genetic relationship between Omicron viruses in NYC, we constructed a genotype network of all sequenced Omicron viruses; nodes represented nucleotide genotypes of viruses and edges between nodes represented pairs of nucleotide genotypes with the highest genetic similarity. We visualized this network using Gephi version 0.9.2 (20). We also constructed a phylogenetic tree of those Omicron viruses in NYC using Nextstrain SARS-CoV-2 workflow version 3.0.6 (21) and visualized it as timescaled using Auspice version 2.33.0 (https://auspice.us), which is part of the Nextstrain workflow. We then identified different clades of Omicron viruses based on the genotype network and the phylogenetic tree.

To investigate the introductions of the Omicron variant in NYC, we downloaded all global Omicron sequences collected before December 11, 2021 and their metadata from GISAID (https://www.gisaid.org) (22). We removed sequences with incomplete information such as collection date or location. We performed mutation calling of these contextual sequences using Nextclade version 1.10.1 (19). We applied the same quality control standards for our sequenced samples as we did for GISAID sequences. Sequences that were <29,000 nt long or had Nextclade-assessed QC.overallStatus value below good were considered low quality and removed. To identify the genetic relationship between viruses clustered into different clades from NYC and the rest of the world, we constructed a phylogenetic tree using local viruses and global contextual viruses. We defined the viruses in NYC clustered into these clades as local viruses. For each clade, global viruses detected before the time at which we detected the virus within the clade in NYC were selected as contextual viruses. We used Nextstrain SARS-CoV-2 workflow version 3.0.6 (21) to construct this phylogenetic tree and Auspice to visualize it as divergence-scaled.

To investigate the genetic relationship between viruses from travelers and locals, we reanalyzed the same phylogenetic tree that was used to investigate the genetic relationship between Omicron viruses in NYC and highlighted travelers. To reveal the detailed transmission pattern of Omicron in NYC, we analyzed the mutational profiles of Omicron viruses in 2 local districts. In the mutational profiles, we presented only the substitutions that were not Omicron-defining substitutions.

We have provided GISAID accession numbers and metadata of Omicron sequences generated in this study (Appendix Table 1) and the GISAID global Omicron sequences used in this study (Appendix Table 2). The source code used to generate the figures has been released at GitHub (https://github.com/wuaipinglab/sarscov2-omicron-nyc).

Since the COVID-19 pandemic started in March 2020, we tested 185,870 specimens for SARS-CoV-2 RNA by diagnostic multiplex RT-PCR. A total of 17,058 (9.18%) specimens were positive. These specimens were collected from 87,616 unique persons who were tested once or multiple times. Of those persons, 12,858 had SARS-CoV-2 infection; average incidence rate was 14.68%.

A total of 9,516 specimens were run through NGS; 7,237 specimens passed our quality control, of which 392 specimens were identified as Omicron (Figure 1, panel A). Those Omicron viruses were collected during November 25–December 11, 2021 (Figure 1, panel B); they were widespread throughout NYC, and densities were higher in the boroughs of Manhattan and Brooklyn.

We performed a phylogenetic analysis of these 392 Omicron isolates. We found that these isolates could be divided into 4 main clades based on the genotype network, phylogenetic tree, and nucleotide substitutions (Figures 2, 3, 4). Most (n = 262; 67%) of these isolates clustered into clade A; those isolates shared a G5515T nucleotide substitution. A total of 65 isolates clustered into 3 smaller clades: clade B (n = 26; 7%), with clade-defining substitution G5924A; clade C (n = 25; 6%), with 3 clade-defining substitutions T10135C, C25708T, and A29301G; and clade D (n = 14; 4%), with 2 clade-defining substitutions C2470T and G22599A. Of note, clade D was consistent with the Omicron subclade BA.1.1 because of its G22599A nucleotide substitution (spike, R346K).

We also noted that some Omicron viruses in regions other than NYC had the same substitutions as those 4 clades (Figure 5). To investigate the origins of these 4 clades, we clustered 14,817 global Omicron viruses downloaded from GISAID (22) into the 4 clades based on their clade-defining substitutions. Among the global Omicron viruses, 861 (6%) clustered into clade A, 3,563 (24%) clustered into clade B, 1,716 (12%) clustered into clade C, and 1,686 (11%) clustered into clade D.

We subsequently investigated the spatiotemporal distribution and phylogenetic relationship of global and NYC viruses within these 4 clades (Figures 5, 6). Clade A and its corresponding substitutions were initially detected in NYC on November 29, 2021; clade B on November 25, clade C on December 6, and clade D on December 1. Of note, we found viruses from other regions clustered into these 4 clades that had been collected earlier than our sequenced viruses in NYC (Figure 5): a total of 12 clade A viruses, 71 clade B viruses, 233 clade C viruses, and 83 clade D viruses. The earlier detection of these viruses in other regions suggests independent introductions into NYC.

To further identify the potential introductions, we constructed a phylogenetic tree using viruses clustered into clades A, B, C, and D, including 327 viruses from NYC and 399 viruses from around the world (Figure 6). We found that the early viruses clustered into clade A were detected in North America, Africa, and Europe. Of note, 3 of them (GISAID accession nos. EPI_ISL_7129868, EPI_ISL_7782594, and EPI_ISL_7908023) were detected in other laboratories in New York state. In the phylogenetic tree, early global clade A viruses were located near the base of clade A, and some viruses in NYC were distributed along main branches. Combining the phylogenetic and spatiotemporal distribution of all clade A viruses, we inferred that this clade was introduced into NYC and then spread through local transmission.

Most clade B early viruses were detected mainly in Africa, with sporadic detection in Europe, North America, and Asia. The viruses in NYC clustered closely with those viruses in the early stage of the wave. Clade B viruses were detected only in Africa before November 21, 2021, suggesting that clade B had spread outside of Africa after early local transmission. Thus, we believe that clade B viruses in NYC were the result of another independent introduction. Similarly, clade C and clade D were distributed in regions including Europe, North America, and Africa before our detections in NYC. The close genetic relationship within these 2 clades suggests 2 additional independent introduction events.

Our phylogenetic analysis shows that the Omicron variant outbreak in NYC likely resulted from multiple independent introductions. We found that, among the 392 sequenced Omicron viruses in NYC, 13 of them were obtained from domestic travelers from the following states: California (2), Florida (2), Georgia (1), Maryland (1), Maine (1), North Carolina (2), Oregon (1), Rhode Island (1), Texas (1), and Utah (1) (Figure 7). To investigate the genetic relationship between viruses from travelers and locals, we analyzed the 392 Omicron viruses in a timescaled phylogenetic tree, which we constructed using Nextstrain workflow and visualized as timescaled using Auspice (21) (Figure 7). The viruses from travelers were distributed across the phylogeny: 8 of them fell into clade A and 1 into clade C; the remaining 4 did not fall into any of the 4 main clades being discussed. Considering the collection time and genetic similarity, we did not find irrefutable evidence that the sequenced viruses from domestic travelers were the origin of the Omicron variant in NYC.

The Omicron variants were distributed in various areas in NYC (Figure 3). To investigate if introductions from outside NYC led to local transmission, we constructed the genotype network and timescaled phylogenetic tree of these 392 viruses (Figures 2, 4). We constructed this phylogenetic tree using Nextstrain workflow and visualized it as timescaled using Auspice (21). In addition to clade-defining substitutions, we found many novel heritable substitutions added to the subsequent progenies. The close genetic relationship within different clades showed that multiple clades spread in NYC with a co-circulation pattern after introductions.

We also discovered several instances of Omicron community transmission in NYC. We noted the viral distribution in 2 postal (ZIP) codes that had the highest detected number of viruses (Figure 8). In the 11201 ZIP code, we found the 2 viruses (nyomi222 and nyomi335) shared the G2398A substitution that belonged to 2 patients from the same family (family 1). Two viruses (nyomi200 and nyomi352) from patients living in the same household shared the C16596T substitution (household 4). These viruses with shared genome sequences from patients within the same family or the same household address suggest local household transmission (Figure 8, panel A; Figure 9). In the 11220 ZIP code, 8 viruses were from patients working in the same building (workplace 1), and 4 of them (nyomi198, nyomi228, nyomi337 and nyomi358) shared identical genetic substitutions. This epidemiologic evidence suggests community transmission in the workplace (Figure 8, panel B; Figure 9). We also found that even viruses isolated from patients at the same working address might fall into different clades, implying complex and possibly cryptic transmission.