Resistance analysis following sotrovimab treatment in participants with COVID-19 during the phase III COMET-ICE study

As of February 2023, the global COVID-19 pandemic caused by SARS-CoV-2 has resulted in 755 million documented cases and 6.8 million deaths worldwide [1]. Disease symptoms in individuals can range from asymptomatic to severe disease, resulting in respiratory failure, thromboembolic events, and death. Although global vaccination against SARS-CoV-2 has resulted in decreased disease burden worldwide, continued viral evolution has led to variants of concern (VOCs) that evade humoral immunity from prior infection or vaccination [2], resulting in breakthrough infections and continued disease transmission. In addition, COVID-19 vaccine efficacy is reduced in immunocompromised patient populations who are at increased risk of severe COVID-19 [3]. Other patient populations at higher risk of developing severe COVID-19 include those who are older, have renal disease, diabetes, cancer, or cardiovascular disease [4,5]. Monoclonal antibodies (mAbs) targeting the SARS-CoV-2 spike protein have been utilized clinically both as therapy and as prophylaxis for people who are at high risk of developing severe COVID-19 disease [6–13]. Small molecule antiviral drugs that inhibit viral replication through different mechanisms have also been used in treatment of COVID-19 [14–16].

Sotrovimab (VIR-7831) is an engineered human mAb that binds to an epitope on the spike protein receptor binding domain (RBD) of SARS-CoV-2 that is conserved across sarbecoviruses [17,18]. The SARS-CoV-2 spike protein is located on the surface of the virion and mediates host cell receptor binding and viral entry [19]. Sotrovimab was derived from the parent antibody S309 [17], inhibits SARS-CoV-2 infection in vitro and in vivo, and maintains activity against VOCs circulating at and prior to the enrollment of this study (including Alpha, Beta, Delta, Gamma, Epsilon, Iota or Lambda) [17,18]. The fragment crystallizable (Fc) domain of sotrovimab includes the 2 amino acid LS (M428L/N434S) modification that extends antibody half-life [20–22]. Additionally, sotrovimab retains Fc effector functions, demonstrating FcγR-mediated effector functions both in vitro and in animal models of SARS-CoV-2 infection, which may also contribute to in vivo efficacy [2,18,23].

The safety and efficacy of sotrovimab in high-risk patients with mild to moderate COVID-19 was evaluated in a randomized, double-blind, multicenter, placebo-controlled, phase III clinical study, hereafter referred to as COMET-ICE (ClinicalTrials.gov NCT04545060) [7,9]. Among 1057 patients randomized in the COMET-ICE study (sotrovimab, 528; placebo, 529), treatment with a single 500 mg intravenous (IV) dose of sotrovimab was associated with a 79% reduction in the adjusted relative risk of all-cause hospitalization longer than 24 h or death through day 29 compared with placebo (adjusted relative risk: 0.21 [95% CI: 0.09 to 0.50]; absolute difference: -4.53% [95% CI: -6.70% to -2.37%]; p < 0.001) [9]. This manuscript describes the resistance analyses conducted using SARS-CoV-2 spike gene sequences from nasopharyngeal swabs (NPS) from COMET-ICE participants to evaluate the presence of circulating SARS-CoV-2 VOC or variants of interest (VOIs) and characterize the presence of baseline, post-baseline and treatment-emergent amino acid substitutions in the epitope of sotrovimab for COMET-ICE participants.

Materials & methods

Clinical study design

Sotrovimab was evaluated for the treatment of mild to moderate COVID-19 in nonhospitalized participants with ≥1 risk factor for severe disease in COMET-ICE, a randomized, double-blind, placebo-controlled, phase III study [7,9]. The patients were at high risk for progression of COVID-19 due to at least 1 of the following risk factors: age of 55 years or older, diabetes requiring medication, obesity (BMI >30; calculated as weight in kilograms divided by height in meters squared), chronic kidney disease (estimated glomerular filtration rate <60 ml/min/1.73 m²) [24], congestive heart failure (≥New York Heart Association class II), chronic obstructive pulmonary disease or moderate to severe asthma [25]. Severely immunocompromised patients, including but not limited to patients receiving immunosuppressive chemotherapy or immunotherapy, were excluded. In brief, between August 2020 and March 2021, a total of 1057 participants with mild or moderate SARS-CoV-2 infection within 5 days of symptom onset were randomized 1:1 to receive a single, intravenous infusion of either 500 mg sotrovimab or equal volume of saline placebo at baseline and followed through 24 weeks posttreatment. The primary clinical endpoint of the study was hospitalization >24 h for acute management of illness or death due to any cause through day 29. The secondary clinical endpoints included: 1) progression of COVID-19 as defined by an emergency room (ER) or hospital visit or death due to any cause through Day 29, and 2) progression to develop severe and/or critical respiratory COVID-19. The study was conducted in accordance with the principles of the Declaration of Helsinki and the Council for International Organizations of Medical Sciences international ethical guidelines, applicable Good Clinical Practice guidelines from the International Council for Harmonisation, and applicable laws and regulations. Ethics approval was obtained from institutional review boards and ethics committees at all participating sites. All patients or their representatives provided written informed consent, and the appropriate institutional forms were archived.

Clinical sample collection

NPS samples were collected from all participants pretreatment (day 1) and at follow-up visits (days 5, 8, 11, 15, 22 and 29). SARS-CoV-2 RNA was quantified by Covance Central Laboratory (NC, USA) using a reverse transcription quantitative polymerase chain reaction (RT-qPCR) assay with primer pairs and probes based on the CDC's qualitative 2019-nCoV emergency use authorization assay [26]. The lower LOQ in the assay was 2228 copies/ml (3.35 log₁₀ copies/ml) and the limit of detection was 1493 copies/ml (3.17 log₁₀ copies/ml). Viral loads were monitored in a blinded fashion to determine visits that qualified for sequence analysis.

SARS-CoV-2 spike gene sequence analysis

NGS was attempted on RNA isolated from baseline (day 1) and post-baseline (day 5, 8, 11, 15, 22 and 29) NPS samples with SARS-CoV-2 viral load above 4.5 log₁₀ copies/ml. Validation of the SARS-CoV-2 spike sequencing assay determined a viral load cut-off of 4.5 log₁₀ copies/ml as the limit of the sequencing assay. Sequencing was conducted by DDL Diagnostics Laboratory (DDL; Rijswijk, the Netherlands) using the Illumina MiSeq platform (Illumina, Inc., CA, USA). The full-length spike gene was amplified by nested RT-PCR using the primer sets developed by the CDC [26].

Nucleotide changes resulting in amino acid substitutions, deletions, or gaps were reported as a change from reference sequence (Wuhan-Hu-1, GenBank ID#: MN908947.3) using DDL's Athena bioinformatics pipeline [27,28]. Amino acid substitutions detected at an AF ≥5% were evaluated in the resistance analyses. The RBD was defined as amino acids 328 to 532 in the spike protein. For amino acid substitutions in the epitope, amino acid positions 332, 333, 334, 335, 336, 337, 339, 340, 341, 343, 344, 345, 346, 354, 356, 357, 358, 359, 360, 361, 440, 441 and 509 in spike were defined as the epitope positions [17,18]. Treatment-emergent amino acid substitutions were defined as amino acid substitutions detected at ≥5% AF in post-baseline samples that were not present at ≥5% AF at baseline. For determination of VOCs/VOIs, analysis was performed using substitutions defined by the WHO [29] (Supplementary Table 2), as well as evaluation for specific detection of K417N, L452R, S477N, E484K, S494P or N501Y substitutions.

Analysis of linked substitutions at amino acid positions 337 & 340

To evaluate the proportion of linked substitutions at amino acid positions 337 and 340, additional analysis of the NGS data was conducted at Vir Biotechnology, Inc. (CA, USA) for participants with substitutions detected at ≥5% AF at both amino acid positions 337 and 340. Primer sequences were removed, and reads were aligned using 2 alignments steps. Only reads spanning both 337 and 340 amino acid positions, mapped in a proper pair and not containing any ambiguous nucleotides at positions corresponding to amino acid positions 337 and 340, were included in the linkage analysis. Linked substitutions detected at amino acid positions 337 and 340 at ≥1% AF were determined.

Cells & antibodies

Vero E6 cells (ATCC) and Lenti-X 293T cells (Takara) were cultured in Dulbecco's Modified Eagle's medium (DMEM), 10% fetal bovine serum (FBS) (heat inactivated), 1 × penicillin-streptomycin at 37°C, 5% CO₂. Vero E6-TMPRSS2 cells [30] were cultured in DMEM, 10% FBS (heat inactivated), 1 × penicillin-streptomycin, 8 μg/ml puromycin at 37°C, 5% CO₂. Sotrovimab (VIR-7831) was produced at WuXi Biologics (China).

Neutralization of SARS-CoV-2 pseudotyped virus

To evaluate the susceptibility of epitope substitutions to sotrovimab, a luciferase reporter vesicular stomatitis virus (VSV) pseudotyped virus system was utilized as previously described [18]. Briefly, pseudotyped virus was produced by transfecting SARS-CoV-2 spike-D19 plasmid encoding the SARS-CoV-2 spike protein with or without the amino acid substitution(s) of interest into Lenti-X 293T cells using standard methods, followed by infection with VSV-luc (rVSVΔG; Kerafast) 1 day later. The titers of SARS-CoV-2 pseudotyped viruses were determined using VeroE6 or VeroE6-TMPRSS2 cells and methods previously described [18].

Neutralization assays were conducted using Vero E6 or Vero E6-TMPRSS2 cells seeded on 96-well plates and cultured overnight at 37°C. 24 h later, 9-point 1:4 serial dilutions of sotrovimab were prepared in media, with each dilution tested in duplicate or triplicate on each plate. Pseudotyped virus was diluted in media to a final MOI of 0.1 when added to cells and added 1:1 to each antibody dilution. Pseudovirus: antibody mixtures were incubated for 1 hour at 37°C. Media was removed from the cells and pseudovirus: antibody mixtures were added to the cells for 1 hour at 37°C, then additional media was added and cells were incubated overnight at 37°C. Luciferase activity was quantified using Bio-Glo reagent (Promega) and relative light units were read on a plate reader. Percent neutralization was determined by comparing antibody-containing wells to no antibody controls and fold change in EC₅₀ relative to pseudotyped virus encoding wild-type spike was calculated using GraphPad Prism software.

Results

Resistance analysis study population

A total of 1057 participants were enrolled into the study, with 528 randomized to sotrovimab (500 mg IV) and 529 randomized to placebo (Figure 1). NPS samples from baseline (day 1) and post-baseline (day 5 or later) visits from all participants with SARS-CoV-2 RNA above the limit of the sequencing assay (4.5 log₁₀ copies/ml) were analyzed by next-generation sequencing (NGS) of the spike gene. Clinical disposition and viral load data are reported elsewhere [9]. Overall, a total of 1377 samples from 696 participants from COMET-ICE had available sequencing data (702 samples from 358 participants in the placebo arm; 675 samples from 338 participants in the sotrovimab arm) and were included in the resistance analyses. Of the 1057 participants enrolled in COMET-ICE, baseline sequences were available from 616 (58.3%) participants (sotrovimab: 58.1%, 307/528; placebo: 58.4%, 309/529) and post-baseline sequences, defined as day 5 or later, were available from 432 (40.9%) participants (sotrovimab: 38.1%, 201/528; placebo: 43.7%, 231/529). Participants with no sequence data available either had samples that were not available, samples with viral load below the limit of the sequencing assay, or samples where sequencing was attempted but data could not be generated (Table 1).

Prevalence of VOC/VOI in COMET-ICE participants

The prevalence of circulating VOC/VOI in COMET-ICE participants was evaluated for 696 participants with available sequencing data. Overall, 21.7% (151/696) of participants had a VOC/VOI detected (Table 2). The VOCs/VOIs detected in COMET-ICE participants included the Alpha (B.1.1.7) (9.5%, 66/696), Epsilon (B.1.427/B.1.429) (5.5%, 38/696), Gamma (P.1) (2.0%, 14/696), Zeta (P.2) (2.2%, 15/696), Iota (B.1.526) (1.3%, 9/696), Lambda (C.37) (0.1%, 1/696) and B.1.1.519 (1.2%, 8/696) variants. No participants carried the full complement of mutations characteristic of the Beta, Delta, Eta, Kappa, Mu or Omicron variants, consistent with circulating variants at the time of study enrollment (August 2020 to March 2021) and locations where the study was conducted [7]. An additional 40 (5.7%, 40/696) participants did not harbor the full complement of mutations characteristic of a VOC/VOI but did have an occurrence of 1 or more single amino acid substitutions of interest (Table 2). Key single amino acid substitutions of interest due to potential effects on immune evasion and viral fitness [31–36] that were evaluated in this study included K417N, L452R, S477N, E484K, S494P or N501Y. Additionally, no substitutions at the non-epitope position S371, which has been published to result in a moderate decrease in in vitro activity of sotrovimab for some substitutions, were detected in COMET-ICE participants.

To determine if the presence of VOCs/VOIs impacted viral kinetics, median viral load kinetics through day 29 were evaluated for participants with or without VOC/VOI or single amino acid substitutions of interest. Median viral loads in participants with or without VOC/VOI demonstrated viral load decline and reached the limit of quantification (LOQ; 3.35 log₁₀ copies/ml) of the assay by day 15 in both treatment arms (Figure 2).

Clinical outcomes for COMET-ICE participants with VOC/VOI

Clinical outcomes were assessed for the 696 participants with or without a VOC/VOI to determine the impact of VOC/VOI and single amino acid substitutions of interest on clinical progression in the COMET-ICE study. The primary clinical endpoint for progression of COVID-19 was defined as clinical progression to hospitalization >24 h or death due to any cause through day 29. Additional secondary endpoints for clinical progression of COVID-19 of 1) progression of COVID-19 as defined by an ER or hospital visit or death due to any cause through day 29, and 2) progression to develop severe and/or critical respiratory COVID-19, were also evaluated. Of the participants carrying VOCs/VOIs or at least 1 substitution of interest, 5.2% (10/191) met the primary endpoint for clinical progression (sotrovimab: 2.1%, 2/94; placebo: 8.2%, 8/97). In the sotrovimab arm, 1 participant carrying the Alpha variant and 1 participant carrying the Epsilon variant met the primary clinical endpoint for progression. In the placebo arm, 8 participants with VOC/VOI met the primary endpoint for progression (3 Gamma variant, 2 Alpha variant, 2 Zeta variant, 1 B.1.1.519 variant). In addition, 1 placebo participant carrying the Zeta variant met the secondary endpoint for clinical progression of COVID-19 due to an ER visit, and 1 participant in the sotrovimab arm carrying the B.1.1.519 variant met the secondary endpoint for progression due to development of severe COVID-19. Of the participants not carrying VOCs/VOIs or single substitutions of interest, 4.0% (20/505) met the primary endpoint for clinical progression (sotrovimab: 0.4%, 1/244; placebo: 7.3%, 19/261). Overall, infection with VOCs/VOIs or SARS-CoV-2 carrying key substitutions of interest (L452R, S477N, E484K, S494P or N501Y) was not associated with clinical progression in sotrovimab-treated participants.

Frequency of SARS-CoV-2 spike amino acid substitutions detected in the Epitope of Sotrovimab in COMET-ICE participants

The presence of baseline and post-baseline amino acid substitutions in the spike protein, RBD, and epitope of sotrovimab was characterized in 696 COMET-ICE participants with available sequence data (Table 3). SARS-CoV-2 spike sequences were evaluated for the presence of amino acid substitutions using the Wuhan-Hu-1 (GenBank ID#: MN908947.3) sequence as the wild-type reference. Among 696 participants with sequence data available, 99.9% (695/696) had a substitution in the spike protein, which includes 4.2% (29/696) of participants where the only substitution detected was D614G. This is consistent with the expected variability in the spike protein. Among participants treated with sotrovimab, epitope substitutions were detected in 3.9% (12/307) of participants at baseline and in 23.9% (48/201) of participants at post-baseline visits.

Baseline epitope substitutions were detected in 2.3% (7/309) of participants in the placebo arm and 3.9% (12/307) of participants in the sotrovimab arm (Table 3, Table 4). The majority of epitope substitutions were detected in 1 participant each and occurred at allelic frequencies (AF) of <28%. Among sotrovimab-treated participants, the most common baseline epitope substitutions were P337H (0.98%, 3/307; <7% AF) and C361T (1.3%, 4/307; <6% AF). In 1 sotrovimab-treated participant, E340A was detected at high AF (>99%) at baseline, and this participant achieved viral suppression (defined as viral load below the limit of detection) by day 8.

Post-baseline epitope substitutions were detected in 3.5% (8/231) of participants in the placebo arm and 23.9% (48/201) of participants in the sotrovimab arm with available post-baseline sequences (Table 3, Table 4). Among participants treated with sotrovimab, the predominant post-baseline epitope substitutions were P337L/R (5.0%, 10/201; 9.090%–97.917% AF) and E340A/K/V (12.9%, 26/201; 9.883%–82.642% AF; Table 4), which confer reduced susceptibility to sotrovimab in vitro [18]. Post-baseline substitutions at epitope amino acid positions 337 and/or 340 were detected in 13.9% (28/201) of sotrovimab-treated participants, and among these 28 participants, substitutions at position P337 alone, E340 alone, or both positions 337 and 340 were detected in 2, 18 and 8 participants, respectively.

Sequencing data were further evaluated for the 8 sotrovimab-treated participants with a substitution detected at both amino acid positions 337 and 340 at 1 or more post-baseline visits. For 2 of these participants, the P337 and E340 substitutions did not occur together in the same sample/visit. Six sotrovimab-treated participants had substitutions detected at both amino acid positions 337 and 340 concurrently in the same sample/visit (4 participants at a single visit, 1 participant at 2 visits, 1 participant at 3 visits). Given the close proximity of P337 and E340 amino acid positions in the spike gene, it was possible to evaluate linkage of P337 and E340 substitutions in these 9 samples using FASTQ sequences with coverage spanning both P337 and E340 amino acid positions. Linked substitutions at amino acid positions 337 and 340 were not detected at ≥5% AF for any of the 9 samples included in the linkage analysis. For all 9 samples, the AFs of the individual P337 or E340 substitutions, or wild-type sequence, comprised the majority of the viral population and added up to approximately 95%, while each of the linked double P337 and E340 substitutions were detected in <2.7% of the relevant sequencing reads (Supplementary Table 1).

Among participants in the placebo arm, the predominant epitope substitution detected post baseline was C361T (2.1%, 5/231; <17% AF; Table 4). Substitution C361T was also detected post baseline in 4.5% (9/201; 5.040%–15.696% AF) of participants in the sotrovimab arm, 1 of whom also had E340K detected post-baseline (Table 4). In total, substitution C361T was detected among 16 participants across baseline and post-baseline visits, including 5 from the placebo arm and 11 from the sotrovimab arm. Spike amino acid position C361 is >99.99% conserved in the Global Initiative on Sharing All Influenza Data (GISAID) database among >13 million SARS-CoV-2 spike sequences deposited as of 26 January 2023, and substitution C361T is very rare with a prevalence of 0.0006% (83/13,782,939). The cysteine to threonine amino acid substitution identified in COMET-ICE participants also requires 3 nucleotide changes (TGT to ACA) from the reference sequence. Given the rarity of these events, an investigation into the quality of the sequencing data in the region of the spike protein encompassing amino acid position C361 was conducted. Additional quality control analysis of sequence reads containing the C361T substitution determined that these reads were artifacts generated during library preparation. Palindromic sequences have been reported to generate sequence artifacts during NGS library preparation and/or cluster generation on Illumina systems [37]. Quality control analysis of FASTQ mapped reads containing the C361T substitution identified these as hybrid sequence reads, which contained sequence matching the genome as well as sequence matching the reverse complement of the genome surrounding a palindromic region containing the C361T substitution (Supplementary Figure 1). The hybrid sequence reads containing the C361T substitution were determined to be artifacts generated during library preparation (Supplementary Figure 2), and the reported C361T substitutions from these hybrid reads do not represent authentic amino acid changes.

Frequency of treatment-emergent epitope substitutions in COMET-ICE participants

Treatment-emergent amino acid substitutions were defined as any substitution that was detected in a post-baseline sample at ≥5% AF but was not detected in the corresponding baseline sample at ≥5% AF. Overall, 352 participants had sequencing results available from paired baseline and post-baseline samples to evaluate the presence of treatment-emergent substitutions (Table 5). Among the 352 participants with paired baseline and post-baseline sequencing results available, 71.4% (130/182) of participants in the placebo arm and 81.2% (138/170) of participants in the sotrovimab arm had a treatment-emergent substitution in the spike protein. Overall, 3.8% (7/182) of participants in the placebo arm and 23.5% (40/170) of participants in the sotrovimab arm had treatment-emergent epitope substitutions (Table 5).

The predominant treatment-emergent substitutions in participants treated with sotrovimab were P337L/R and E340A/K/V, detected as treatment-emergent in 14.1% (24/170) of sotrovimab-treated participants with paired baseline and post-baseline sequences (Table 5). Other treatment-emergent substitutions were detected in ≤2 participants each and occurred at AF of <20%, with the exception of C361T, which was detected as treatment-emergent in 2.7% (5/182; 5.993%–16.055% AF) and 3.5% (6/170; 5.04%–15.696% AF) of participants in the placebo and sotrovimab arms, respectively. Given that the prevalence of treatment-emergent C361T was similar in both placebo and sotrovimab-treated participants, and that C361T was determined to be an artifact generated during NGS library preparation, it was concluded that C361T was not a sotrovimab treatment-emergent substitution in the COMET-ICE study. When C361T was excluded from the treatment-emergent analysis, 20.6% (35/170) of participants in the sotrovimab arm had a treatment-emergent epitope substitution.

No impact of epitope substitutions on clinical progression in COMET-ICE participants

Clinical outcomes were assessed for the 15 placebo and 57 sotrovimab-treated participants with epitope substitutions detected at baseline and/or post-baseline visits to determine the impact of epitope substitutions on clinical progression. Of the 19 participants with baseline epitope substitutions in the placebo or sotrovimab arms, none met the primary clinical endpoint for progression of hospitalization >24 h or death due to any cause through day 29. In addition, none of the 7 placebo participants or 40 sotrovimab-treated participants with treatment-emergent epitope substitutions met the primary endpoint for clinical progression.

In the placebo arm, 8 participants had post-baseline epitope substitutions detected and none met the primary clinical endpoint for progression, while 1 participant met the secondary endpoint for clinical progression of COVID-19 due to an ER visit. Of the 48 participants in the sotrovimab arm with post-baseline epitope substitutions detected, only 1 participant met the primary clinical endpoint for progression to hospitalization >24 h or death due to any cause, and none met the secondary endpoints for clinical progression of COVID-19. The sotrovimab-treated participant who met the primary clinical endpoint for progression had E340K detected post-baseline at day 11 and day 15 visits (≥99.8% AF) and was infected with the Epsilon variant of SARS-CoV-2. Baseline sequencing was not possible due to low viral load at day 1, so the presence of E340K at baseline is unknown. Overall, 2.1% (1/48) and 0.7% (1/153) of participants in the sotrovimab arm with or without post-baseline epitope substitutions, respectively, met the primary endpoint for clinical progression. Of the 32 sotrovimab participants who had an epitope substitution detected at spike amino acid positions 337 or 340 at baseline or post-baseline visits, only 1 (3.1%) met the progression criteria for the primary endpoint of the study. Overall, the presence of baseline, post-baseline, or treatment-emergent epitope substitutions was not associated with clinical progression in sotrovimab-treated participants.

Comparison of viral load kinetics for participants with or without epitope substitutions

Median viral load kinetics for participants with or without epitope substitutions were evaluated to determine the impact of epitope substitutions on viral load decline or viral rebound during sotrovimab treatment (Figure 3). Median baseline viral load was >1 log higher in participants with an epitope substitution at baseline compared with those without an epitope substitution at baseline. Although baseline viral loads were higher in participants with epitope substitutions, median viral load continued to decline below the LOQ (3.35 log₁₀ copies/ml) through day 29 for participants with or without epitope substitutions (Figure 3). Median viral load for participants with epitope substitutions at position 337 or 340 declined through day 29 in both the placebo and sotrovimab arm, with a median viral load at Day 29 below the LOQ for participants in either arm.

Viral rebound, defined as an increase of >1 log₁₀ copies/ml in viral load or as having quantifiable viral load after having been below the LOQ, was evaluated for participants with or without epitope substitutions. Of the 130 placebo participants who experienced viral rebound, 114 had sequences available and 6.1% (7/114) of these participants had an epitope substitution detected at ≥5% AF at any visit, while 93.9% (107/114) did not have an epitope substitution. Of the 123 sotrovimab-treated participants who experienced viral rebound, 95 had sequences available and 20% (19/95) of these participants had an epitope substitution detected at ≥5% AF at any visit, while 80% (76/95) did not have an epitope substitution. Overall, among 338 participants in the sotrovimab arm with available sequence results, 33.3% (19/57) and 27.0% (76/281) of participants with or without an epitope substitution at any visit, respectively, experienced viral rebound. Of the sotrovimab-treated participants with or without epitope substitutions who experienced viral rebound, 1 or 0 participants, respectively, met the primary clinical endpoint for progression. Overall, the detection of baseline or post-baseline epitope substitutions did not appear to be associated with viral rebound in sotrovimab-treated participants. In addition, although the numbers of participants are small, there was no evidence that viral rebound in participants with or without epitope substitutions was associated with clinical progression.

Phenotypic analysis of SARS-CoV-2 VOCs/VOIs, single amino acid substitutions of interest, & epitope substitutions detected in COMET-ICE participants

Seven distinct VOCs/VOIs were detected in COMET-ICE participants, including the Alpha, Gamma, Epsilon, Iota, Lambda, Zeta and B.1.1.519 variants. Sotrovimab retained activity (<three-fold change in EC₅₀ relative to Wuhan-Hu-1 wild-type) against pseudotyped virus expressing the Alpha (B.1.1.7), Gamma (P.1), Epsilon (B.1.427/B.1.429), Iota (B.1.526), Lambda (C.37), Zeta (P.2) and B.1.1.519 spike variants [2,18]. Sotrovimab also retained activity against pseudotyped virus expressing spike amino acid substitutions L452R (0.76-fold change in EC₅₀), S477N (1.99-fold change in EC₅₀), E484K [18], S494P [18] or N501Y (0.60-fold change in EC₅₀).

Twenty-nine epitope amino acid substitutions were detected in COMET-ICE participants. Of the 22 epitope amino acid substitutions with available data, sotrovimab retained activity against amino acid substitutions L335F, L335S, G339C, N343S, N343Y (0.20-fold change in EC₅₀), A344P (0.27-fold change in EC₅₀), A344V, R346I, R346G, N354K, K356N, K356R, R357I, I358V, S359G and N360D (0.93-fold change in EC₅₀); fold-change values not listed have been published in Cathcart et al. [18]. Substitution P337H demonstrated a 5.13-fold change in EC₅₀ relative to Wuhan-Hu-1 wild-type [18]. Substitutions P337L, P337R, E340A, E340K and E340V resulted in significant EC₅₀ shifts (>100-fold) indicating reduced susceptibility to sotrovimab in vitro [18]. Amino acid substitutions C336R, C336S, N343del, E340STOP, C361T, C361F and R509I could not be evaluated due to decreased infectivity of the pseudotyped virus or poor expression of the spike protein containing these substitutions.

Discussion

In this resistance analysis, we evaluated the prevalence of circulating SARS-CoV-2 VOCs/VOIs and characterized the presence of baseline, post-baseline, and treatment-emergent amino acid substitutions in the epitope of sotrovimab for COMET-ICE participants. In the COMET-ICE study, a single dose of 500 mg intravenous (IV) sotrovimab administered in adults with symptomatic, mild to moderate COVID-19 and at least 1 risk factor for progression significantly reduced the risk of all-cause hospitalization or death through day 29 compared with placebo [7,9]. The resistance analysis conducted for the COMET-ICE study is a comprehensive evaluation of clinical resistance for sotrovimab 500 mg IV and represents an important characterization of resistance for a COVID-19 treatment in clinical use as of the writing of this manuscript.

The enrollment period of the COMET-ICE study was during the early phase of the pandemic (August 2020–March 2021) when VOC/VOI were just beginning to emerge. In this analysis of SARS-CoV-2 spike sequences from COMET-ICE, the majority of participants were infected with the Wuhan-Hu-1 wild-type virus, with or without D614G. Additionally, the VOC/VOI detected in COMET-ICE participants included the Alpha, Epsilon, Gamma, Iota, Lambda, Zeta and B.1.1.519 variants, which corresponds with the circulating strains during that time [29,38–40]. Clinical progression in participants treated with sotrovimab did not appear to be associated with infection of any identified VOC/VOI. This is in line with in vitro data where sotrovimab retains activity against pseudotyped virus expressing Alpha, Epsilon, Gamma, Iota, Lambda, Zeta or B.1.1.519 spike variant proteins [18]. One limitation of this study is that few participants infected with VOC/VOIs were enrolled. This may limit the ability to detect any potential associations with clinical progression or rare events [9].

In this resistance analysis, there was no evidence that the presence of baseline, post-baseline, or treatment-emergent epitope substitutions impacted clinical outcomes in sotrovimab-treated participants. Among participants who received 500 mg IV sotrovimab, 3.9% (12/307) and 23.5% (40/170) had substitutions in the epitope of sotrovimab detected at baseline or as treatment emergent. The predominant treatment-emergent substitutions included P337L/R and E340A/K/V, which confer reduced susceptibility to sotrovimab in vitro [18]. Although baseline (1.3%, 4/307) and treatment-emergent (14.1%, 24/170) substitutions at amino acid positions 337 and 340 were observed clinically, they were not associated with clinical disease progression in these participants in the COMET-ICE study. Overall, 9.5% (32/338) of sotrovimab-treated participants had an epitope substitution detected at amino acid positions 337 or 340 of the spike protein at ≥5% AF at any visit. Despite this observation, very few (1/32) sotrovimab-treated participants with these amino acid substitutions experienced clinical progression and the clinical significance of these substitutions remains unknown. One limitation of the present study is that it did not include immunocompromised patients. The limited endogenous immune response in these patients could impact the development and persistence of resistance. However, in a small study of immunocompromised patietns treated with sotrovimab, mutations at amino acid positions 337 or 340 were detected in ∼12.8% of patients, consistent with the rates of resistance in the present study in immunocompetent individuals [41]. This is consistent with clinical studies and metanalyses that have found sotrovimab to be safe for the treatment of COVID-19 in immunocompromised patients [42,43].

Other studies have also indicated the emergence of epitope substitutions at positions 337 or 340 in sotrovimab-treated patients [44,45]. The P337L/R and E340A/K/V substitutions confer reduced susceptibility to sotrovimab in vitro using a pseudotyped virus system (>100-fold change in EC₅₀) [18]. Notably, substitutions at positions 337 and 340 continue to be rare and are cumulatively found in <0.04% of sequences out of >13,970,000 sequences deposited as of 26 January 2023 in the GISAID database. The lack of increased prevalence in these substitutions across global populations or geographies despite ongoing global spread and selection pressure may suggest that substitutions at these amino acid positions result in reduced viral fitness or transmissibility. The linkage analysis of sequencing data from COMET-ICE found that while substitutions at amino acids P337 and E340 can additively comprise a majority (>90%) of the viral population in some participants, substitutions at these positions do not exist in combination on the same viral genomes (linkage detected in <2.7% of the relevant sequencing reads). These data suggest that the presence of specific substitutions at positions P337 and E340 may be mutually exclusive and further support the hypothesis that these substitutions may confer a fitness cost.

Despite the lack of evidence for an association between development of sotrovimab resistance and reduction in clinical efficacy, as well as the rarity of substitutions that confer reduced susceptibility to sotrovimab in the GISAID database, a potential concern around anti-SARS-CoV-2 pharmacotherapy-driven resistance exists since sotrovimab is administered as monotherapy. Combination antibody therapy may reduce the emergence of single amino acid substitutions that confer resistance to a single component of the combination therapy, as was observed for the REGN10933/REGN10987 combination [46]. However, the increased barrier to resistance observed with combination antibody therapies does not preclude viral escape from the entire combination due to emerging VOCs [47].

Sotrovimab demonstrates two antiviral mechanisms of action, including neutralization of virus particles as well as antibody-mediated killing of infected cells via Fc effector functions, both of which have been documented to occur in vitro and in vivo [18,23]. Sotrovimab demonstrated antibody dependent cellular cytotoxicity (ADCC) and antibody dependent cellular phagocytosis (ADCP) activity in vitro using SARS-CoV-2 spike expressing CHO target cells and primary human NK (F/F158 and V/V158 donors) or monocyte effector cells, respectively [18,23,48]. In addition, a study by Case et al. demonstrated that Fc effector functions contribute to the antiviral activity of the sotrovimab parent antibody S309-LS in vivo using a prophylactic model of antibody protection in K18-hACE2 mice infected with SARS-CoV-2 Omicron BA.1 or BA.2 variants [23]. Indeed, prophylactic administration of S309-LS resulted in significant reductions in total and infectious viral load in the lung or nasal turbinates compared with isotype control for all SARS-CoV-2 variants evaluated (D614G, BA.1, BA.1.1, BA.2); however, a version of S309 containing Fc mutations that abrogate effector function demonstrated no significant differences in lung or nasal viral load compared with isotype control in BA.1- or BA.2-infected animals [23]. These results suggest that sotrovimab has the potential to induce NK cell-mediated killing and phagocytosis of cells expressing SARS-CoV-2 spike protein and hence, ADCC and ADCP are possible additional contributions to the mechanism of action for sotrovimab in the prevention of severe COVID-19. To date, sotrovimab demonstrates some in vitro neutralization activity against all evaluated VOCs in the VSV pseudotyped assay through XBB.1.5, however shifts in susceptibility vary from <five-fold for pre-omicron strains to as much as >700-fold reduction in activity for BN.1 [49]. The Fc-mediated functions demonstrated by sotrovimab in vitro and in animal models in vivo may contribute to the mechanism of action in humans and may be able to compensate for some decreases in neutralizing activity due to less susceptible amino acid substitutions or viral variants, however these hypotheses remain to be confirmed clinically.

Conclusion

Results from this study suggest that infection with circulating VOC/VOI in participants treated with sotrovimab are not associated with clinical progression. Additionally, sotrovimab epitope substitutions were rare at baseline and occurred more frequently post-baseline in sotrovimab-treated participants, but there was no evidence that their presence was associated with clinical progression. Overall, findings from sequence and phenotypic analysis of the COMET-ICE study suggest that emergence of sotrovimab resistance substitutions during treatment is not associated with a reduction in clinical efficacy.