The Breadth of the Neutralizing Antibody Response to Original SARS-CoV-2 Infection is Linked to the Presence of Long COVID Symptoms

Long COVID (LC), a form of post-acute sequelae of SARS-CoV-2 infection (PASC), is characterized by ongoing symptoms affecting quality of life. Its mechanisms are incompletely understood and may include autoreactive immunity, microvascular dysregulation, herpesvirus reactivation, and persistent SARS-CoV-2 antigens. Individuals with LC often have higher Spike-specific antibody levels, alongside signs of adaptive immune dysregulation and exhaustion. While neutralizing antibody levels wane faster than total Spike IgG, the relationship between the longitudinal dynamics and breadth of neutralization and specific LC phenotypes has been unclear. The study aims to determine whether the breadth and durability of neutralizing antibody responses after original SARS-CoV-2 infection (pre-vaccination) are associated with LC and distinct LC symptom clusters.

Prior work shows higher SARS-CoV-2 Spike antibody levels and immune dysregulation in LC compared to full recovery, with persistence of viral RNA/protein in tissues of some LC patients. Neutralizing responses decay over time, and weaker responses have been linked to LC in some studies. Emerging evidence suggests broader coronavirus immune responses (e.g., OC43) may associate with LC, implying antibody breadth might be relevant. Omicron variants exhibit immune escape relative to ancestral/Alpha/Delta infection, motivating evaluation of cross-neutralization breadth in relation to LC phenotypes.

Design and cohort: Longitudinal study of 184 participants recovering from nucleic acid–confirmed SARS-CoV-2 infection during early pandemic waves, prior to Delta emergence and before widespread vaccination (most samples collected before March 2021). A total of 384 plasma samples were analyzed at approximately 1, 2, and 4 months after symptom onset (median windows: ~33, 59, and 120 days). LC was defined as any new or worsened symptom since acute infection not attributable to another cause. Ethics: UCSF IRB-approved; informed consent obtained. Neutralization assays: PhenoSense (Monogram Biosciences) HIV-1 pseudovirus neutralization assay using HEK293 producer cells co-transfected with HIV-1 luciferase reporter and SARS-CoV-2 Spike expression vectors; neutralization quantified by inhibition of luciferase in ACE2/TMPRSS2-expressing HEK293 target cells. ID50 calculated for: ancestral (original) SARS-CoV-2, Alpha (B.1.1.7), D614G, Delta (B.1.617.2), Omicron BA.1 (B.1.1.529), and Omicron BA.5 (B.1.1.529). Subset analysis (N=16) expanded to BA.2, BA.4.6, BQ.1.1, and XBB.1.5, selecting samples with high or low BA.5 responses across timepoints. Data handling: Antibody data generated blinded to participant information. Statistics: Between-group comparisons of ID50 used nonparametric Mann–Whitney or Friedman tests with Dunn multiple-comparison corrections (Prism v8, SPSS v29). Categorical data via Fisher’s exact test; correlations via Spearman rank. Longitudinal modeling used linear mixed-effects models (lme4 in R) on log10-transformed ID50 with time and covariates (age, sex, hospitalization, BMI≥30, diabetes, autoimmune disease, LC symptoms) as fixed effects and participant as random effect. Cross-sectional logistic regression models (SPSS v29) assessed associations between neutralization (continuous log ID50 or binary top-15% thresholds by sensitivity analysis) and LC outcomes (any LC; phenotypes: neurocognitive, cardiopulmonary, gastrointestinal, musculoskeletal, fatigue) at 1, 2, and 4 months, adjusting for demographics and clinical covariates. Only one sample per participant per time window was included to avoid oversampling. Study approval and IRB oversight were noted.

- Neutralization patterns: Neutralization ID50s were highest against the original strain and early variants (Alpha, Beta, Delta), declining from 1 to 4 months post-infection. Cross-neutralization against Omicron BA.1 was very low; a greater proportion of participants exhibited measurable neutralization to BA.5 than to BA.1 or XBB.1.5 up to 4 months. Across all time points, 12.5% of titers were below assay positivity for the original strain, 78.4% for BA.1, and 67.4% for BA.5. In a subset (N=16), BA.5 neutralization resembled BA.4.6 and BQ.1.1, while XBB.1.5 resembled BA.1 (low/negative); BA.2 responses were more evenly distributed. - Demographic/clinical associations: Neutralization decreased over time for all strains (p<0.01) except BA.1 (p=0.16). Male sex associated with higher neutralization for original, Delta, and BA.5 (not BA.1). Hospitalization during acute infection associated with higher neutralization across strains except BA.1. Associations with BMI and diabetes varied by variant, with differential baseline levels and decay rates. - Cross-sectional LC associations: At ~4 months, neutralization to ancestral, Alpha, and Delta was not associated with LC overall or specific phenotypes. In contrast, higher BA.5 cross-neutralization (continuous ID50) was independently associated with higher odds of LC, particularly neurocognitive and gastrointestinal phenotypes; not significant for fatigue or cardiopulmonary phenotypes. Using a binary variable for top 15% ID50s, robust BA.5 cross-neutralization remained significantly associated with any LC and with neurocognitive and gastrointestinal LC, while high ancestral neutralization was not associated. - Longitudinal LC associations: Mixed-effects models showed higher ancestral-strain neutralization across all time points among participants with gastrointestinal and cardiopulmonary LC symptoms compared with those without LC. Despite higher levels, neutralization decayed faster among participants with cardiopulmonary and musculoskeletal symptoms; musculoskeletal, cardiopulmonary, and neurocognitive phenotypes exhibited more rapid decay relative to those without these specific symptoms.

Findings indicate that the breadth of neutralizing antibody responses after original SARS-CoV-2 infection, particularly cross-neutralization to Omicron BA.5 measured several months post-infection, is linked to the presence of Long COVID overall and to gastrointestinal and neurocognitive phenotypes. Neutralization to the infecting strain itself did not predict LC in cross-sectional analyses, underscoring that breadth rather than magnitude to the original strain may be more relevant. Longitudinal analyses revealed both higher overall neutralization and a faster decline among certain LC phenotypes, suggesting complex temporal immune dynamics. One hypothesis is that persistent antigen presentation in tissues could drive broader and higher neutralization responses that subsequently wane more quickly. The observed higher BA.5 versus BA.1 cross-neutralization may reflect Spike mutation reversion patterns affecting antigenicity. Together, results support a multifactorial LC pathogenesis where antibody breadth and decay kinetics are associated with symptom persistence and phenotype-specific manifestations.

Broad cross-neutralization to Omicron BA.5 approximately four months after primary SARS-CoV-2 infection is associated with higher odds of Long COVID and with neurocognitive and gastrointestinal symptom phenotypes, whereas neutralization to the original infecting strain is not. Longitudinally, participants with certain LC phenotypes exhibit higher neutralization levels coupled with faster decay. These data implicate antibody response breadth and kinetics as correlates of LC and motivate further studies to determine causality and mechanisms, including roles of persistent antigen and autoreactivity. Future work should validate findings in larger, diverse cohorts, including vaccinated and reinfected populations and infections with more recent variants, and should integrate multi-omic immune profiling with clinical phenotyping.

- Cohort derives from the pre-vaccine, pre-Delta/Omicron era; findings may not generalize to later variants, vaccinated, or reinfected individuals. - Convenience sample may limit generalizability beyond the study population. - Neutralization measured using pseudovirus assays rather than live, replication-competent virus, though prior work supports comparability. - Limited evaluation of very recent Omicron sublineages (expanded panel tested in a small subset of 16 participants). - Cross-sectional analyses at defined windows may miss temporal nuances; residual confounding is possible despite adjustments.