Introduction and Establishment of SARS-CoV-2 Gamma Variant in New York City in Early 2021

Abstract

Background

Monitoring the emergence and spread of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants is an important public health objective. We investigated how the Gamma variant was established in New York City (NYC) in early 2021 in the presence of travel restrictions that aimed to prevent viral spread from Brazil, the country where the variant was first identified.

Methods

We performed phylogeographic analysis on 15 967 Gamma sequences sampled between 10 March and 1 May 2021, to identify geographic sources of Gamma lineages introduced into NYC. We identified locally circulating Gamma transmission clusters and inferred the timing of their establishment in NYC.

Results

We identified 16 phylogenetically distinct Gamma clusters established in NYC (cluster sizes ranged 2–108 genomes); most of them were introduced from Florida and Illinois and only 1 directly from Brazil. By the time the first Gamma case was reported by genomic surveillance in NYC on 10 March, the majority (57%) of circulating Gamma lineages had already been established in the city for at least 2 weeks.

Conclusions

Although travel from Brazil to the United States was restricted from May 2020 through the end of the study period, this restriction did not prevent Gamma from becoming established in NYC as most introductions occurred from domestic locations.

Monitoring the emergence, introduction, and circulation of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) variants of concern (VOCs) is an important public health tool that is essential to guiding an effective pandemic response [1]. Emergence and rapid spread of SARS-COV-2 variants has driven consecutive epidemic waves regionally and around the globe, starting with the Alpha variant (B.1.1.7 lineage) identified in the United Kingdom in late 2020, followed by the Delta variant (B.1.617.2) identified in India in early 2021, and the recent waves of Omicron variants infections, the first of which was identified in South Africa in late 2021. In the United States and elsewhere, the response to the identification of these variants has included restricting travel from locations where the variants were first identified; however, widespread establishment of these variants nonetheless ensued.

Gamma VOC (P.1 lineage) was first identified in late 2020 in Manaus, Brazil, a region that previously experienced a high coronavirus disease 2019 (COVID-19) burden and had an estimated 68% seroprevalence [2]. The rapid spread of Gamma in a region with high preexisting immunity immediately raised concerns about the transmission potential of this variant. On 25 January 2021, the first case of Gamma infection was documented in the United States in the state of Minnesota [3]. Travel from Brazil to the United States was prohibited for non-US citizens starting May 2020, and this was extended indefinitely on 26 January 2021 due to concern about Gamma spread [4]. Nevertheless, by May 2021, Gamma had been detected in at least 41 US states (https://outbreak.info/).

Molecular epidemiology investigations into this lineage characterized 17 novel mutations relative to the wild-type virus (Hu-1 reference strain)—including 10 mutations in the spike protein, 3 of which (K417T, E484K, and N501Y) have been shown to increase spike ACE2 receptor binding affinity [5]—leading to the lineage being designated a VOC by the Centers for Disease Control and Prevention (CDC) [6]. Gamma virus is poorly inhibited by convalescent plasma and sera from vaccinated individuals [7], is partially resistant to monoclonal antibodies used for COVID-19 treatment [8], and might be associated with increased COVID-19 severity and mortality in younger adults and adults without comorbidities, compared to the wild-type virus [9, 10].

New York City (NYC) was among the first locations in North America or Europe to experience a large COVID-19 outbreak: by May 2020, the number of reported cases exceeded 150 000 [11]. The first Gamma VOC case in NYC was identified on 10 March 2021, in the middle of the second COVID-19 wave. The proportion of Gamma-attributed cases in NYC was growing but was outcompeted first by Iota (B.1.526 lineage) and Alpha variants [12], and then by Delta in summer 2021. Importantly, NYC, unlike many other locations in the United States and globally, has a large-scale genomic surveillance system deployed by the New York City Public Health Laboratory (NYC PHL) along with the establishment of the NYC Pandemic Response Lab (PRL), which paved the way for 14 600 SARS-CoV-2 genomes to be sequenced by NYC PHL and PRL between 1 January and 1 May 2021.

Here we used a phylogeographic approach to investigate patterns of introduction and local circulation of the SARS-CoV-2 Gamma VOC in NYC in March–April 2021. Our analysis shows that nearly all circulating Gamma transmission clusters were introduced from other US locations and not from outside the United States, indicating that whereas international travel restrictions might keep the number of introductions from the country where that variant was first detected to a minimum, they do not prevent the lineage from establishing in another location even if they are in place before the variant emerges. We also show that even a comprehensive genomic surveillance system may not be able to detect imported novel variants before they establish persistent transmission clusters.

METHODS

Sequence Data

To characterize the emergence of the Gamma variant in NYC, we downloaded all complete high-quality (<5% ambiguous sites) with an available date of sampling (day, month, year) SARS-CoV-2 Gamma genomes available through the GISAID database from 10 March to 5 June 2021; as of 5 June 2021 (n = 15 967) [13]. After removing duplicate sequences from all locations except for the location of interest (NYC), we aligned the remaining Gamma sequences using MAFFT version 7.453 [14], trimmed the 5′ and 3′ ends, and masked all problematic sites in the alignment as suggested before [15]. The institutional review board of UC San Diego waived ethical approval of this work as human subjects exempt.

Phylogenetic Analysis and Downsampling

To identify phylogenetic clusters comprising lineages predominantly circulating in NYC, we first inferred a maximum parsimony phylogenetic tree representing the Gamma global diversity using all available unique sequences and IQTree version 2.0.3 software [16]. We then removed all clades that did not have NYC sequences and were not ancestral to clades that contained NYC sequences. For large polytomies, a random subsampling of genomes was performed. The resulting subsampled dataset was further used to reconstruct a maximum likelihood phylogenetic tree under the GTR+F+I nucleotide substitution model, collapsing polytomies, and enforcing a minimum branch length of 1e-9.

Phylogeographic Analysis

We performed Bayesian phylogeographic reconstruction in BEAST version 1.10.4 [17]. First, we reconstructed time-scaled phylogenetic trees based on the genomes in the subsampled dataset under an HKY85 nucleotide substitution model, assuming a strict molecular clock model (fixed to the value of 8 × 10⁻⁴ nucleotides/site/year) and a Bayesian skyline coalescent tree prior [18]. We ran 3 independent Markov Chain Monte Carlo (MCMC) chains of 1 × 10⁹ generations; for all BEAST analyses, we used Tracer [19] to assess convergence of MCMC chains and to ensure effective sample size (ESS) > 100 for all parameters. We used the LogCombiner package from BEAST to combine the tree distributions from these 3 independent chains and resample them at a lower frequency to obtain a distribution of 1000 trees. The phylogeographic reconstruction of viral lineage movement was then performed using the resulting tree distribution and assigning geographic trait value to the tips of the phylogeny based on the sampling locations. Phylogeographic analysis assumes that location changes follow the same process as sequence evolution; we used an asymmetric model that allows estimating separate incoming and outgoing viral lineage flow rates for each location. We considered 10 locations in our analysis: (1) NYC, (2) other New York locations, (3) California, (4) Florida, (5) Illinois, (6) New Jersey, (7) other US locations, (8) Brazil, (9) other South America Locations, and (10) other locations in the world. The specific states were selected based on the high number of Gamma lineages reported. Brazil was selected as the country where the Gamma VOC was first identified [2]; similarly, other South American countries were included as a separate location because of the close proximity to Brazil and the high number of reported Gamma cases. We then ran another MCMC analysis of 5 × 10⁶ generations, which allowed us to obtain posterior tree distributions annotated with Markov jumps indicating virus migration events between the specified geographical locations.

Gamma clades circulating in NYC were then identified in a phylogeographic context, defined as clades of ≥2 sequences for which the ancestral location was in NYC with 90% posterior probability support. The ancestral location of NYC clades was then identified as the ancestral location of the ancestral node that had the highest posterior probability support.

Analysis of Persisting Lineages

Lineage persistence analysis identifies whether lineages have been introduced into the studied location within a defined time interval or they were already circulating in the given location before that time period (ie, persisting lineages) [20]. To describe the proportion of circulating Gamma strains versus those that were being introduced from other locations, we ran lineage persistence analysis using PersistenceSummarizer as described by Lemey et al [20]. This analysis allowed us to describe the proportion of lineages being introduced to NYC from other locations within each 2-week period and the proportion of lineages that continued circulating in NYC from before the start of the period (those being “reintroduced” from within NYC). We split the time into 2-week intervals going backward from the most recently sampled sequence (5 June 2021). Because the first Gamma sequence was identified in NYC on 10 March 2021, we analyzed the introduction and persistence of circulating lineages for six 2-week time periods starting 2 weeks before the first week with a nonzero number of Gamma sequences (6 February) and until the end of the second epidemic wave in NYC (1 May).

RESULTS

Genetic Sequence Data and Phylogeographic Analysis Identified 16 NYC Gamma Lineage Clusters

As of 5 June 2021, 15 967 unique Gamma genome sequences have been deposited in GISAID from around the world, including 116 generated by NYC public health surveillance. We reconstructed a maximum parsimony tree using these sequences and found that Gamma lineages circulating in NYC were dispersed throughout the phylogeny, indicating multiple independent introductions (Supplementary Figure 1).

Bayesian phylogeographic analysis performed on a subset of Gamma sequences (n = 1461), selected as relevant to NYC infections based on the maximum parsimony phylogeny, allowed us to identify 16 NYC transmission clusters, defined as clades with ancestral locations estimated to be in NYC (Figure 1). NYC clusters’ sizes ranged between 2 and 108 sequences (median = 4), and included between 1 and 41 NYC sequences (median = 2).

The majority of NYC transmission clusters were introduced from Florida (n = 6), followed by Illinois (n = 4), and other locations in the United States (n = 4). Only 1 NYC Gamma transmission cluster (cluster 10, n = 2) was introduced from abroad, directly from Brazil [2]. The 2 largest clusters, cluster 2 (n = 41) and cluster 7 (n = 108), included 16 and 41 NYC genomes and originated from the states of Florida and California, respectively (Figure 1).

The Majority of Gamma Lineages Were Persisting in NYC Since Early March 2021

To describe patterns of introduction and persistence of Gamma lineages in NYC, we ran a lineage persistence analysis on the full Gamma dataset focusing on the period between 6 February (a month before the first Gamma lineage was detected in NYC) and 1 May (the end of the second epidemic wave in the city) 2021. This analysis allowed us to describe whether local lineages had been circulating in NYC more than 2 weeks prior (persisting lineages) or if they were recently introduced from other locations.

In February 2021, less than 1% of NYC Gamma cases were associated with lineages imported from abroad. Gamma lineages were mainly introduced from other US states (states that were not identified as separate discrete locations in our analyses, 59% of all introductions), followed by the introductions from New Jersey (16%), other parts of the New York state (9%), and Florida (8%), with the proportion of introductions from all other considered locations and local NYC persistence (lineages circulating in the city since 2 weeks prior) <3% each (Figure 2). By 20 March 2021, 57% of all Gamma cases were persisting in NYC for at least 2 weeks; among the lineages that were newly introduced to NYC between 6 and 20 March 2021, the majority came from other US states (22%), New Jersey (15%), other parts of New York State (4%), and Florida (2%). By the end of the second wave of the epidemic, in the second half of April 2021, 43% of Gamma lineages were persisting in NYC, 24% came from other parts of New York State, 9% from New Jersey, 3% from Florida, and 21% from other US states. Although the majority of circulating NYC Gamma clusters identified through the phylogeographic analysis were introduced from Florida and Illinois, only a small proportion (around 5% and 1%, respectively) of Gamma lineages introduced in the observed period were from these states.

Gamma Transmission Clusters Were Established in NYC in February 2021

Phylodynamic analysis of the times to most recent common ancestor (TMRCAs) for the identified Gamma transmission clusters showed that the majority of locally circulating clusters had been established in February 2021. The cluster TMRCAs ranged between 5 January and 18 May 2021 (Figure 3). The TMRCAs for the largest circulating clusters, clusters 2 and 7, were 19 February 2021 (95% highest posterior density [HPD], 9–28 February 2021) and 25 February 2021 (95% HPD, 15 February–9 March 2021), respectively. The TMRCA for the only transmission cluster introduced from Brazil was estimated to be 3 May 2021 (95% HPD, 24 April–5 May 2021).

Gamma Variant in Neighborhoods with High Preexisting Seropositivity in NYC in March–April 2021

Between 6 March and 1 May 2021, NYC PHL and PRL sequenced 11 385 SARS-CoV-2 genomes and identified 116 of them as Gamma using Pangolin [21]. The 2 NYC neighborhoods (as defined by the United Hospital Fund) with the highest number of identified Gamma genomes were in Queens and Brooklyn (West Queens and East New York; Figure 4A). While the average proportion of SARS-CoV-2 sequences identified as Gamma remained low throughout city neighborhoods between 6 March and 1 May 2021 (0.7%; 95% confidence interval, 0.45–0.95 [22]), it was >2% in Chelsea, Greenpoint, Bayside Meadows, Southeast Queens, and East New York in that period. Some of these areas, including Southeast Queens and East New York, also showed some of the highest seropositivity rates observed in NYC as of 11 March 2021 ([22]; Figure 4B), suggesting a more effective Gamma transmission in the presence of other circulating lineages in these areas with preexisting immunity.

The first 3 Gamma lineages were found in Greenpoint, Bayside-Meadows, and Southeast Queens before 19 March 2021 (Supplementary Figure 2). In the second half of March 2021, at which point most Gamma lineages were already established locally and persisting in NYC, they became prevalent in West Queens and Coney Island—the areas where the majority of SARS-CoV-2 lineages were circulating in March–April 2021. In late April 2021, when the number of SARS-CoV-2 infections dropped in NYC, the Gamma lineages were most prevalent in Jamaica. Upon further investigation of the geographical spread of the Gamma lineages identified in the 2 largest circulating NYC clusters, clusters 2 and 7, no localized geographical patterns were observed: infections from both clusters were found throughout the city (Supplementary Figure 2).

DISCUSSION

The emergence and swift spread of SARS-CoV-2 variants have threatened pandemic control efforts. To date, countries with well-developed SARS-CoV-2 genomic surveillance systems that have enabled early identification of newly emergent VOCs, such as South Africa and the United Kingdom, have been subjected to travel restrictions by other countries after alerting the international community of the new variant [4]. Such measures might have substantial sociopolitical and economic consequences, but have not prevented the global dissemination of the variants and the subsequent infections growth [23]. Here we show how the introduction and spread of the Gamma variant in NYC illustrates that although travel restrictions might reduce introductions from the origin locations, they failed to prevent variant establishment, as local NYC transmission clusters were seeded by domestic transmission.

Using a phylogeographic approach, we showed that despite the limited travel from Brazil—the country where Gamma lineage was first identified in late 2020 [2]—being enforced since May 2020, most Gamma lineages arrived in NYC and established locally circulating clusters in February 2021. All but 1 identified Gamma transmission clusters circulating in NYC were introduced from US locations, suggesting that the variant has been widely circulating in the United States at the time and travel restrictions with Brazil did not prevent the variant from establishing in NYC. Only 1 of the Gamma transmission clusters circulating in NYC resulted from an introduction from outside of the United States when the virus was imported from Brazil, but this introduction (in May 2021) was preceded by multiple introductions from domestic locations. Similarly, previous analyses of genomic, epidemiological, and travel data showed minimal evidence of direct SARS-CoV-2 transmission to NYC from China, the country where the virus was first identified, in early 2020; all other identified introductions that resulted in local NYC transmission were from other locations in the United States [11]. International travel restrictions imposed on China in early 2020 likely reduced the number of virus importation events but only modestly affected epidemic trajectory globally; their effect could have been greater if combined with local interventions that reduce transmission [24]. Other evidence suggests that limiting travel might only have an impact on epidemic dynamics in countries with low COVID-19 incidence [25]: the high daily number of COVID-19 cases in the United States in early 2021 [26] might have contributed to a limited role of international introductions in the Gamma transmission. We show that once introduced, a large proportion of Gamma lineages persisted in NYC, also suggesting local mitigation strategies may have a much more substantial impact on variant spread than international travel restrictions.

As expected, in the presence of other dominant SARS-COV-2 strains such as the Iota and the Alpha variants, Gamma spread more effectively in areas with high seroprevalence than in other neighborhoods: NYC areas with some of the highest seroprevalence levels were also those with the highest number of Gamma cases. Vaccine rollout undoubtedly slowed down the spread of Delta [27], but its impact on Gamma spread in NYC was unlikely as pronounced. Even though vaccination rates were accelerating in early 2021 (the number of people who received at least 1 dose of the vaccines grew from 7% on 1 February to 18.8% on 10 March to 45.4% on 1 May [22]), these efforts were delayed relative to the time of Gamma introduction and establishment in the city. Thus, the most successful Gamma lineages that resulted in large locally circulating clusters were widespread in the city and were not constrained to specific neighborhoods.

NYC dramatically increased its sequencing capacity in January–February 2021 with the implementation of sequencing at PRL: in February 2021, in the middle of the second epidemic wave, 4%–6% of all SARS-CoV-2 infections in NYC were genotyped [22], which is much higher than the average (1.75%) in North America [1]. Although these efforts allowed for timely identification of the rapidly growing Iota variant that originated in NYC [12], they did not help to substantially reduce a lag between the introduction and identification of Gamma lineages, which were present at much lower frequencies, in NYC. This lag was similar to the one recorded in the early days of the pandemic in the absence of available widespread testing when the first SARS-CoV-2 case in NYC was registered on 29 February 2020 but phylodynamic analysis later revealed that the virus has likely been introduced to NYC as early as January [11, 28]. Importantly, laboratory surveillance includes a necessary lag from the sample collection to laboratory analysis and quality checks before a newly emerging variant can be reported. In NYC, the public health laboratory is continuously improving systems and processes to minimize this lag and encourages timely reporting of genomic data by hospitals and academic laboratories that led to substantial improvement in volume and timeliness of genomic data in each subsequent SARS-CoV-2 pandemic wave.

Patterns of introduction from within the United States showed that a small number of introductions can result in a large number of locally circulating clusters. Even though <5% of the circulating NYC Gamma lineages, which includes circulating Gamma clusters and singletons, were introduced from the state of Florida in February–April 2021, introductions from Florida seeded the majority of circulating NYC Gamma transmission clusters. Similarly, even though the overall proportion of Gamma lineages introduced from Illinois was <1% at all times during the observation period, these introductions seeded 4 of the 16 (25%) identified circulating transmission clusters. Such differences might stem from certain behavioral patterns displayed by travelers from Florida and Illinois that resulted in more onward transmission upon arrival. These patterns are similar to those previously reported for Alpha VOC (B.1.1.7): the lineage first identified in the United Kingdom was initially introduced into 3 locations in the United States (California, Florida, and Georgia) and then spread through unmitigated transmission between different US locations, with Florida state playing an important role in that transmission [29].

Similar to any such analysis, our work is affected by the disproportional sampling and sequencing efforts in various US locations and abroad. As our analysis only focused on the introductions and establishment of Gamma into NYC, we did not investigate importations of Gamma into other domestic locations. Furthermore, our subsampling approach was based on maximum parsimony phylogenetic reconstruction, which could have affected the representation of different geographic locations in our final dataset but was necessary to create a dataset that could be analyzed in a Bayesian framework. Both the development of methods that can analyze larger datasets to avoid downsampling, and further improvement in genomic surveillance capacity proportionally across the US locations and worldwide can help improve similar investigations in the future.

The analysis of the introduction and spread of Gamma lineages in NYC, a location with large-scale genomic surveillance, shows that even in the presence of effective genomic monitoring efforts, low-prevalence variants with potential public health importance can circulate undetected amidst an ongoing epidemic. Even though US travel restrictions might be effective at reducing the number of direct introductions from the international locations where the variant was first identified, they do not preclude the establishment of local transmission clusters if the variant has already established itself elsewhere domestically.

Supplementary Data

Supplementary materials are available at The Journal of Infectious Diseases online (http://jid.oxfordjournals.org/). Supplementary materials consist of data provided by the author that are published to benefit the reader. The posted materials are not copyedited. The contents of all supplementary data are the sole responsibility of the authors. Questions or messages regarding errors should be addressed to the author.

Notes

Acknowledgments. We thank Mustapha Mustapha of the New York City Department of Health and Mental Hygiene for his useful comments on the manuscript text.

Disclaimer. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the Centers for Disease Control and Prevention (CDC). Use of trade names is for identification only and does not imply endorsement by the CDC.

Financial support. This work was supported by the Branco Weiss Society in Science Fellowship (to T. I. V.); the National Institutes of Health (grant numbers AI135992 to J. O. W., AI154824 to G. W. H., and U01AI151812 and U19AI135995 to K. G. A.); CDC Broad Agency Anouncement (grant number 75D30120C09795); and CDC support to the Epidemiology and Laboratory Capacity (ELC) for Infectious Diseases Cooperative Agreement (grant number ELC DETECT 6NU50CK000517-01-07).

References

Author notes