Sequential intrahost evolution and onward transmission of SARS-CoV-2 variants – Nature Communications

Emergence of a novel BA.1 sublineage through intrahost evolution

We performed genomic analysis of serially collected nasopharyngeal (NP) and anterior nares (AN) samples from an immunocompromised patient (P1) with diffuse B-cell lymphoma and persistent SARS-CoV-2 Omicron BA.1 replication between December 2021 and March 2022. Over a 12-week period, we documented the accumulation of nine amino acid substitutions in the spike protein N-terminal domain (NTD), the receptor binding domain (RBD), and in the S1/S2 furin cleavage site (FCS) (Fig. 1) within the same patient. The first four mutations R346T, K458M, E484V, and A688V were detected simultaneously 40 days after the initial SARS-CoV-2 diagnosis and were fixed. Two weeks later (day 64), L167T and the FCS mutation P681Y were detected in addition to the four initial mutations. During the following weeks additional mutations emerged; samples from this period contained shared (L455W) as well as distinct signature mutations (E96D on day 72, S477D on day 81). Notably, only two mutations emerged outside of the spike gene (Supplementary Fig. 1, P1), suggesting positive intrahost selection of spike protein changes due to competitive replication advantages. The SARS-CoV-2 substitution rate varied throughout the course of infection; after an initial period of three weeks without changes in the consensus sequence, there was a rapid accumulation of substitutions between week 4 and week 12. On average, one substitution per week was observed during this period, corresponding to a rate of 52 substitutions/year – approximately two-fold higher than the global average of 26–27 substitutions/year^21,22.

Forward spread confirms transmission potential of the novel BA.1.23 sublineage

Background health system-wide SARS-CoV-2 genomic surveillance conducted by the Mount Sinai Pathogen Surveillance Program (MS-PSP) during the same time period identified three other patients harboring SARS-CoV-2 Omicron variants that shared the same combination of spike amino acid substitutions found in the index case P1 (E96D, R346T, L455W, K458M, E484V, H681R, A688V), as well as the synonymous mutation T6001C. A query of more than 13.8 million global SARS-CoV-2 sequences deposited in the GISAID (Global Initiative on Sharing Avian Influenza Data) database up to November 2022 revealed only two additional related genomes, both originating from the NYC area (Fig. 2, source data are provided as a Source Data file). Based on the metadata provided, these two genomes were obtained from distinct individuals that differed by age and gender from our patients. Altogether, the presence of the same unique combination of mutations in five additional cases indicates limited local forward transmission of this novel Omicron subvariant, which received the BA.1.23 Pango lineage designation (Fig. 2, Table 1).

a Maximum likelihood (ML) phylogenetic subtree with SARS-CoV-2 (BA.1) sequences from the persistent infection case P1 (red) and the onward transmissions (P2, P3, P4 in cyan, yellow and green respectively), in a global background of sequences available in GISAID. Branches are colored to identify each patient. The number of days after the first SARS-CoV-2 positive specimen of P1 is indicated in brackets. Sibling clusters are collapsed for easier visualization. The x-axis shows the number of nucleotide substitutions relative to the root of the phylogenetic tree. Bootstrap support values above 70% are shown for the un-collapsed branches. The distinct BA.1 subvariant that was transmitted was designated as PANGO lineage BA.1.23. The bottom-right insert shows a time-scaled ML phylogeny and the estimates for the time of the most recent ancestors (TMRCA) and 95% confidence intervals, for the nodes where the transmissions from the index case are positioned. b Frequency of single nucleotide variants (SNV), and amino acid substitutions for SARS-CoV-2 genomic positions with consensus changes from ancestral Wuhan-1 and Omicron BA.1. Positions with mixed nucleotides below consensus levels are also shown for intrahost SNVs (miSNVs) seen in more than one time point. The sequenced specimens are shown sequentially for P1 with prolonged BA.1 infection and transmission cases of BA.1.23 (P2, P3 and P4). The viral genomes from P1 show progressive accumulation of mutations in the N-terminal domain (NTD), receptor binding domain (RBD), and S1/S2 furin cleavage site (FCS); The same constellation of mutations was subsequently detected in three documented transmission cases (P2, P3 and P4). The number of days since the first positive test in P1 is shown on the left, with the number of days after the first positive test for each patient between brackets. Positions with nonsynonymous SNVs are marked by filled circles. Specimen type is indicated on the right by open circles (anterior nares) or filled diamonds (nasopharynx). Positions are numbered according to the reference genome sequence NC_045512.2. Arrows on the left indicate the likely window of transmission from the index patient between days 64–72. Source data is provided in the Source Data file.

All transmission cases contained one synonymous change in ORF1a (T6001C) previously only observed at day 72 in the index case (Supplementary Fig. 1). This narrowed the time window of the first transmission to an 18-day period following day 64, two months after the first positive SARS-CoV-2 test of patient P1. During this time, patients P2 and P4 were admitted to the same unit as P1 for 10 and 18 days, respectively. Following a transfer, P4 also overlapped for two days with P3 in a different unit. While this provided a potential opportunity for direct or indirect contact, it is important to note that P3 and P4 were not tested for SARS-CoV-2 until three weeks later, when they were found to be positive. No contact information was available for the two cases identified outside our health system. As of February 2023, the most recent case was detected in mid-April 2022 and the last SARS-CoV-2 positive specimen from a forward transmission case (P3) was collected in September 2022. Although viral isolation failed for specimens available from P1, we successfully cultured SARS-CoV-2 from the initial positive specimens of the subsequent cases (P2–P4), confirming the transmission of replication-competent virus.

Sequential persistent infections drive further intrahost evolution of BA.1.23

All three forward transmissions in our health system were detected in patients with underlying hematologic malignancies. Although patient P2 cleared the BA.1.23 infection, patients P3 and P4 both developed persistent infections (Figs. 2 and 3a). The MS-PSP continued monitoring these patients during follow-up visits and hospitalizations. Patient P4 remained positive by nucleic acid amplification tests (NAAT) for four weeks, with acquisition of an additional mutation in the spike RBD (S:V445A) within one week of the initial positive test (Figs. 2 and 3a). Patient P3 developed a much longer persistent infection lasting for more than four months. Genomic analysis of all 13 serially collected specimens from patient P3 identified several new amino acid substitutions throughout the viral genome (Fig. 2b and Supplementary Fig. 1). The most divergent (i.e. highest number of mutations) specimen collected 131 days after patient P3’s COVID-19 onset contained 11 additional amino acid substitutions compared to the originally transmitted BA.1.23 variant. These included six substitutions in the spike NTD (S254F), RBD (N448S, F456L, reversion of 458 M to K), and S2 (reversion of 981 L to F, S982L); as well as five substitutions in other SARS-CoV-2 proteins (ORF1a nsp3: T1001/183I, K1795/977Q, ORF1ab nsp12: S4398/6 L, G5063/671 S; and ORF7a: T39I/T24I). Notably, two spike amino acid reversions to the Wuhan-1 (S:M458K) or BA.1 (S:F981L) sequence were accompanied by changes at neighboring positions (S:F456L and S:S982L) (Fig. 2b and Supplementary Fig. 1). Interestingly, in the subsequent three swab specimens (days 144, 154 and 165) the number of amino acid substitutions decreased relative to the specimen collected on day 131, with recurrent flips and reversions across specimens, suggesting the presence of competing quasispecies.

a The timeline of BA.1.23 infected patients’ positive (red croses) or negative (open blue circles) nucleic acid amplification test (NAAT) for SARS-CoV-2 (top). The N gene target cycle threshold (Ct) values for respiratory specimens, for different diagnostic methods is shown (bottom panel). The NAAT panel only includes data points from positive tests that reported a Ct value. b Number of nucleotide substitutions in the consensus sequence relative to Omicron BA.1 in the sequenced specimens. c SARS-CoV-2 spike-binding IgG antibody levels for P1–P4. Antibody levels are shown in arbitrary units per mL (Arb. units/mL). Documented vaccine administrations are indicated. d Timeline of SARS-CoV-2 antiviral treatments received by P1–P4. Treatment duration is indicated by the length of the bar. Source data provided in the Source Data file.

To further investigate this, we looked for positions with intrahost single nucleotide variants (iSNVs) that were present in only a minority of the SARS-CoV-2 viral population within each specimen (referred to as miSNVs). We identified miSNVs within or outside the spike gene in each of the three persistent infection cases (P1, P3, and P4) with several instances in which miSNV were present in earlier specimens, prior to their fixation in subsequent specimens from the same patient (Fig. 2b and Supplementary Figs. 1, 2). We also observed numerous positions with miSNVs that persisted across multiple sequential specimens without ever becoming dominant. This was most notable for patient P3, where the number of positions with miSNVs increased from 0 to 32 between days 101 and 266 – outnumbering the consensus sequence changes by a factor of three (Fig. 2b and Supplementary Fig. 3). Approximately half of these mutations occurred within the spike gene, where nonsynonymous miSNVs were clustered in the S1 (NTD:S254F, RBD:K356T, S371F, P384L, F456L, K458M, and FCS:N679K) and S2 (D936Y, V963F, N978D, L981F, S982L, E1195G) domains (Fig. 2b and Supplementary Fig. 4). Mutations at these positions are rare at the consensus level, with a maximum prevalence of 0.1% in the GISAID database [as of 2022-09-10]. Although we found no clear evidence of compartmentalization of miSNVs by specimen source across patients, there was a notable decrease in miSNVs and increase in consensus mutations in the AN specimen collected from patient P3 on day 131, compared to the earlier and later NP specimens. Thus, intrahost adaptation occurs with different dynamics pointing to bottlenecks and competing selection pressures, resulting in the appearance and disappearance of specific mutations.

To examine a potential role for recombination in the diversification of BA.1.23 we used RIPPLES²³ to assess the phylogenetic placement of genome segments from the most divergent consensus genotypes of BA.1.23 within the global SARS-CoV-2 diversity. We further queried if the observed miSNV patterns were present in contemporary lineages using covSPECTRUM²⁴ and considering all patterns in a sliding window of 3 consecutive miSNVs. Neither analysis yielded evidence for recombination, providing further support for diversification through the accumulation of mutations during intrahost evolution.

Impact of treatment and host immunity on BA.1.23 evolution

To determine the potential impact of SARS-CoV-2 antiviral treatments on intrahost evolution we examined the medication histories of patients P1 (index), P2, P3 and P4 (Fig. 3, source data are provided as a Source Data file). We also assessed SARS-CoV-2 spike binding antibody levels using available serological data from clinical tests. The index patient P1 was vaccinated at the time of admission with two doses of an unspecified vaccine administered six months and one week before hospitalization (Fig. 3c). A third vaccine dose was administered two days after admission. Additional SARS-CoV-2 antiviral treatments included a course of remdesivir on days 1-6 and a dose of Gamunex IgG on day 9 (Fig. 3D). Moderate titers of SARS-CoV-2 spike binding antibodies were detected on day 38, around the time the first intrahost mutations were found (Figs. 2b and 3b). These antibody titers could be due to residual antibodies from the Gamunex IgG treatment, an immune response to vaccination and/or infection, or a combination of treatment and host response. Notably, the detection of the first three spike mutations followed a rapid increase in virus detected in nasal secretion (NAAT mean cycle thresholds (Ct) decreased from 35 on day 30 to 28 on day 40), suggesting a potential selection bottleneck around this time (Figs. 3a and 3c).

Patient P2 received four doses of Moderna mRNA vaccine, three of them at least four months prior to their first positive SARS-CoV-2 test and had high titers of SARS-CoV-2 antibodies when assayed three months before their single positive SARS-CoV-2 NAAT (Fig. 3c). This patient also received a one-month course of nirmatrelvir/ritonavir (Paxlovid™) (Fig. 3d). The combination of high levels of spike binding IgG antibodies and antiviral treatment may explain why this patient successfully cleared the BA.1.23 infection.

Patient P3 was also vaccinated with two doses of Pfizer mRNA vaccine at least five months prior to their first positive test and received a three-week course of nirmatrelvir/ritonavir (Paxlovid™). However, no SARS-CoV-2 spike binding IgG antibodies were detected during the first four months of the persistent infection (Fig. 3c). Notably, during this time we also did not detect new substitutions in the BA.1.23 virus, despite indications of active viral replication based on consistently low NAAT Ct values (less than 25) in NP and AN samples (Fig. 3a). Active BA.1.23 replication was confirmed by isolation of replication-competent SARS-CoV-2 from five specimens collected between study days 101–147. The appearance of new intrahost substitutions during the last month of persistent infection occurred around the time of detection of SARS-CoV-2 antibodies at levels close to the limit of detection of the serological assay used (COVID-SeroKlir, Kantaro Semi-Quantitative SARS-CoV-2 IgG; Fig. 3). Since patient P3 did not receive any biologicals during their hospitalization, it is likely that the observed seroconversion was the result of a delayed and weak host immune response.

For patient P4 no prior vaccinations were registered. Bebtelovimab was administered after their first positive SARS-CoV-2 PCR test, followed by a course of nirmatrelvir/ritonavir (Paxlovid^TM). P4 had low SARS-CoV-2 antibody titers at the time of the first positive PCR test prior to the monoclonal antibody treatment (Fig. 3c). Of note, a single S:V445A mutation emerged in this patient. Viral variants carrying this mutation have been associated with reduced susceptibility to bebtelovimab in pseudotyped virus-like particle (VLP) neutralization assays²⁵. We did not find SARS-CoV-2 mutations associated with remdesivir^{26,27,28,29,30,31} or nirmatrelvir/ritonavir resistance³² in any of the patients treated with these drugs.

Serum neutralization profile of the transmitted persistent Omicron BA.1.23 variant

Neutralization of Omicron BA.1 isolates by sera from convalescent or vaccinated individuals is strongly reduced compared to the levels obtained for neutralization of the ancestral strains^33,34,35. Therefore, we assessed the degree of neutralizing resistance of the transmitted BA.1.23 variant (BA.1 + S:E96D, R346T, L455W, K548M, E484V, P681R, A688V) compared to Omicron BA.1 (B.1.1.529), as well as the ancestral SARS-CoV-2 (USA-WA1/2020, WA). We used a multi-cycle microneutralization assay in which human serum is present continuously to best mimic physiological in vivo conditions^33,36. We selected sera from two subsets of PARIS study participants representing distinct levels of immunity. The first panel of sera tested was collected before and after booster vaccination, while the second set of samples was obtained before and after Omicron BA.1 breakthrough infection in vaccinated participants (see Supplemental Table 1 for details).

Sera collected prior to booster vaccination with monovalent SARS-CoV-2 RNA vaccines (18 matched samples), neutralized BA.1 and BA.1.23 less well than WA1/USA (geometric mean titer; GMT WA1: 71; GMT BA.1: 12; GMT BA.1.23: 6; Fig. 4a, source data are provided as a Source Data file), with the majority of samples failing to display any neutralization activity against BA.1 (5/9) and BA.1.23 (7/9), respectively. Matched sera collected after mRNA booster vaccination from the same study participants showed the booster vaccination increased the neutralization titers for all the viral isolates, but the loss in neutralization for BA.1.23 was greater than 12-fold compared to the >6-fold reduction for BA.1 relative to WA1/USA (geometric mean titer; GMT WA1: 1,689; BA.1: 189; GMT BA.1.23: 43; Fig. 4a). Of note, the loss of neutralization activity for BA.1 relative to the WA1/USA ancestral variant measured here is less than what we and others have previously reported which could be due to assay variation in combination with the limited number of samples tested^33,34,37.

a Absolute neutralization titers (left) and fold-reduction (right) for WA−1, BA.1 and BA.1.23 variants by paired sera from 9 study participants collected before and after booster vaccination (n = 18). The number of samples with titers below the limit of detection of the serological assay (dashed line) is indicated at the bottom of the graph. The error bars represent the geometric mean with 95% confidence intervals. A one-way RM ANOVA with Tukey’s multiple comparison test was used to compare the neutralization titers before and after booster vaccination. ***p-values < 0.001, *p-value: 0.02. b Absolute neutralization titers (left) and fold-reduction (right) for WA-1, BA.1 and BA.1.23 isolates by sera from study participants who experienced breakthrough infection with BA.1. Data for paired sera from 11 participants collected before and after BA.1 infection (n = 22) are shown. The error bars represent the geometric mean with 95% confidence intervals. A one-way RM ANOVA with Tukey’s multiple comparison test was used to compare the neutralization titers before and after BA.1 breakthrough infection. *p-values: 0.02, ns not significant. c Absolute neutralization titers for each of the isolates (WA-1, BA.1 and BA.1.23) by sera collected from patient P2 before and after BA1.23 infection. The first vertical dotted line (left) represents the time of the third vaccine dose as days relative to the index case. The second vertical dotted line (right) indicates the time of infection with the BA.1.23 variant. d Comparison of neutralization fold-change in inhibitory dilution 50% (ID50) measured before and after booster vaccination (left panel, based on 4a) as well as before and after BA.1 (middle panel, based on 4b) and BA.1.23 (right panel, patient 2, based on 4c) breakthrough infection for each of the three viruses tested (WA-1, BA.1 and BA.1.23). Each dot represents IC50 or fold-change data for a specific serum specimen tested by serial dilution in a single replicate experimental setting. GMT denotes mean geometric mean of the inhibitory dilution 50% (ID50) values. The horizontal dotted line represents the limit of detection (10). Samples with neutralization titers below the level of detection were assigned the neutralization value of 5 (equaling to half of the limit of detection) in the ID50 plots. Source data for this figure is provided in the Source Data file.

We next tested how sera collected from the same study participants before and after breakthrough infection with the BA.1 Omicron variant (22 matched samples) would neutralize BA.1.23. We noted a significant difference in neutralization activity for BA.1.23 and BA.1 compared to WA1 in the samples collected after BA.1 breakthrough infection (Fig. 4b). Prior to BA.1 infection, the difference was >17-fold with 3/11 samples having undetectable neutralization activity for BA.1.23 and >14-fold difference with 2/11 of the samples failing to neutralize BA.1 (GMT WA1: 346; GMT BA.1: 26; GMT BA.1.23: 22; Fig. 4b). After infection, neutralization titers increased for all three viruses reducing the difference to four-fold for BA.1 and five-fold for BA.1.23 (GMT WA1: 1,913; GMT BA.1: 503; GMT BA.1.23: 399; Fig. 4b).

Then, we analyzed sera collected from patient P2 at different time points (i.e., before booster vaccination, after booster vaccination but before BA.1.23 infection, and at two time points after infection with BA.1.23). P2 mounted neutralizing antibodies after booster mRNA vaccination against WA1, but not against BA.1 or BA.1.23 (Fig. 4c). Serum collected one month after the breakthrough infection with BA.1.23 showed a sharp increase in neutralization activity for BA.1 and BA.1.23. (Fig. 4c). Lastly, we computed the fold changes for each viral isolate before and after booster vaccination or breakthrough infection with either BA.1 or BA.1.23 (Fig. 4d). These data show that monovalent booster vaccination increases neutralization titers for the ancestral variants, while breakthrough infections with Omicron variants yielded a boost in neutralization of these contemporary variants. Interestingly, we observed large variation in the extent to which BA.1 infection boosted neutralization of the two different Omicron variants (Fig. 4d, middle panel, fold changes range from less than 1 to more than 100-fold difference).

In conclusion, the transmitted persistent BA.1.23 isolate is more neutralization-resistant than the parental BA.1. However, infection with either Omicron variant BA.1 or BA.1.23 induced cross-reactive humoral responses that were capable of neutralizing both variants.