Characterising protective immune responses to SARS-CoV-2 in urban and rural Malawi between February 2021 and April 2022

Categories: Disease & Virus

October 29, 2025

Participant characteristics

Data from 1,876 participants were included: Karonga (rural), n = 958 (51.1%, CI 48.8–53.3); Lilongwe (urban), n = 918 (48.9%, CI 46.7–51.2). The median age was higher in Lilongwe (26.7 years, IQR 14.8–41.1) than Karonga (21.2 years, IQR 11.8–39.8) (p = 0.00028, Wilcoxon) (Table 1). In terms of biological sex, Lilongwe had more females (61.8%, CI 58.6–64.9) compared with Karonga (50%, CI 46.8–53.2) (p p = 0.016, chi-sq), hypertension (p = 0.00079, chi-sq), and people living with HIV (p = 0.0090, chi-sq) (Table 1). Asthma (p = 0.0010, chi-sq) and heart disease (p = 0.0070, Fisher) were more prevalent in Karonga (Table 1). The median follow up time between study surveys was 106 days (minimum – 44 days, maximum – 222 days). COVID-19 vaccination coverage was low at Survey 1 (Karonga, 3.4% (CI 2.3–4.9); Lilongwe 9.2% (CI 7.3–11.5)) with participants having only received one dose of the AZ vaccine (Supplementary Fig. 1). By Survey 4, this increased to 23.6.% (CI 20.6–27.0) in Karonga and 30.8% (CI 26.8–35.3) in Lilongwe, with participants having received AZ-one or two doses and J&J-one dose, with one dose of J&J being considered fully vaccinated with a primary course. Vaccination coverage was consistently higher in Lilongwe (Survey 1, p p = 0.021; Survey 3, p = 0.0003; Survey 4, p = 0.0085; chi-sq) (Supplementary Fig. 1).

Table 1 Participant characteristics at baseline. For continuous variables, the p-value was determined using the Wilcoxon rank-sum test. For categorical variables where both cell counts >=5, a chi-squared test was used, while a Fisher’s exact test was used if at least one cell count was

Increasing complexity of SARS-CoV-2 neutralisation profiles over time

Neutralisation responses to SARS-CoV-2 evolved greatly over the course of the study, indicating changes in protective immunity in the cohort. To accurately assess the nAb responses in the cohort, the following analyses were restricted to HIV-uninfected participants (n = 1,780), as an alternative assay was required for measurement of neutralisation in HIV-infected participants due to interference with the HIV base of the pseudotypes. Overall, nAb prevalence increased over time, being consistently higher in Lilongwe than Karonga (p

In both study sites, when looking at participants sampled across the entire study period, the SARS-CoV-2 neutralisation profiles became increasingly complex as vaccination coverage increased, and variants emerged (Fig. 1). At Survey 1, infected participants had ancestral B.1 virus- and Beta-dominant responses. By Survey 2, Delta-dominant responses were detected – Karonga 2.0% (CI 1.1–36), Lilongwe 3.5% (CI 2.0-6.2) (including individuals with dominant responses to Delta and a variant that occurred before the Delta wave). By Survey 4, Omicron BA.1/BA.2 had emerged. In Karonga, 7.5% (CI 5.6–10.0) were Omicron BA.1-dominant and 2.7% (CI 1.5–4.3) were Omicron BA.2-dominant, while in Lilongwe, 13.9% (CI 10.1–18.2) were Omicron BA.1-dominant and 4.2% (CI 2.5-7.0) were Omicron BA.2-dominant (including individuals with dominant responses to BA.1/BA.2 and a variant that occurred before the Omicron wave).

Hybrid immunity was determined by looking at those vaccinated who were positive when tested with the Nucleocapsid (N) ELISA and therefore had also been naturally infected. Rates of hybrid immunity increased over time, reaching 16.0% (CI 13.1–19.3) in Karonga and 26.8% (CI 22.2–32.0) in Lilongwe at Survey 4 (p = 0.0004; chi-sq) (Fig. 1). Rates of hybrid immunity were consistently higher in Lilongwe (Survey 1, p = 0.0022; Survey 2, p p = 0.0035; chi-sq), (Fig. 1). The proportion of vaccinated individuals who had no detectable nAbs peaked at Survey 2 (Karonga, 9.5% (CI 7.4–12.3); Lilongwe 6.1% (CI 4.0-9.4)) then decreased by Survey 4 (Karonga, 4.4% (CI 3.0-6.5); Lilongwe, 1.6% (CI 0.1–3.7)), as many developed hybrid immunity. Among vaccinated participants in each survey who were HIV-uninfected (not just those with complete sample series displayed in Fig. 1), the proportion who were N ELISA positive generally increased, with a substantial rise in infections at Survey 4 in particular (Survey 1: 36% (34/94); Survey 2: 41% (80/195); Survey 3: 33% (79/239); Survey 4: 73% (194/267)).

Rates of reversion to seronegative status (when an individual who previously tested nAb positive subsequently tested negative, i.e. seroreverted) peaked at Survey 3 (Karonga, 5.7% (CI 4.0–8.0); Lilongwe, 5.2% (CI 3.2–8.2)) (Fig. 1). Seroreversion was highest among Delta-dominant individuals (Fig. 1). Of those who seroreverted at Survey 3, 29% (CI 16.1–46.6) in Karonga and 38% (CI 18.5–61.4) in Lilongwe had previous Delta-dominant responses at Survey 2.

Trajectory of nAb responses over the study period

Many individuals showed similar responses to multiple variants, complicating dominant variant assignment. To explore if this stemmed from multiple infections, we examined nAb titres over time in infected participants who contributed a complete sample series, grouped by the survey at which individuals seroconverted. Post initial exposure, median titres declined by the following survey then increased at Survey 4, likely due to Omicron infections (Fig. 2). Individuals who were seropositive at Survey 1 (ancestral B.1/Beta-dominant) generally exhibited the strongest and most sustained nAb responses, with the magnitude of boosting being particularly high after Omicron reinfection (Survey 4) (Fig. 2).

We then looked at the individual-level changes in nAb titres over time. We included all individuals with a complete sample series who were nAb positive within the study timeframe, and grouped them by exposure type (infected, vaccinated, and infected and vaccinated), and survey of seroconversion. A large degree of heterogeneity in the nAb responses was observed, with both the magnitude of boosting and waning varying substantially between participants (Supplementary Fig. 3). Additionally, within-individual responses varied greatly between surveys, though as observed with Fig. 2 those seroconverted prior to Survey 1 had more consistent titres. When comparing the responses to different variants, there was a high degree of cross-reactivity – for example, individuals infected at Survey 4 were most likely infected with an Omicron variant, but also displayed responses to the other SARS-CoV-2 variants. Notably, seroconversions at Survey 3 were dominated by those vaccinated, with many then becoming infected and vaccinated at Survey 4. In contrast, Survey 4 seroconversions were dominated by infections.

Vaccination leads to higher nAb responses than infection

To examine the strength of participants’ protective, neutralisation responses, we focused on the titrated samples. We compared nAb titres in those solely infected, solely vaccinated, and both infected and vaccinated. Individuals both infected and vaccinated had higher nAb responses than those infected for all variants (ancestral B.1, p p p p p = 0.0016; Omicron BA.2, p = 0.0016; Wilcoxon) (Fig. 3A). For ancestral B.1, Alpha, Beta, and Delta, the median nAb titres for those vaccinated were significantly higher compared with those infected (ancestral B.1: 196.7, IQR 84.1-447.3 vs. 89.1, IQR 50.0-209.1, respectively, p p = 0.0016; Beta: 121.4, IQR 56.1-312.4 vs. 69.2, IQR 50.0-175.2, p p = 0.0006; Wilcoxon) (Fig. 3A).

To study vaccination-induced nAbs solely, we analysed vaccinated (not infected) individuals. Alpha was excluded as it was only tested against at Survey 1 when participants had received just AZ-one dose. Vaccinated individuals displayed their highest nAb activity against ancestral B.1 (AZ-one dose: 240.3, IQR 83.6-586.6; AZ-two doses: 203.8, IQR 95.1-449.1; J&J-one dose: 109.1, IQR 65.8-282.2; Wilcoxon) (Fig. 3B). Neutralisation was lower against other variants, decreasing as variants diverged from the ancestral B.1 virus. Omicron BA.1 and BA.2 were poorly neutralised (AZ-one dose, BA.1: 50.0, IQR 50.0–63.0; BA.2: 50.0, IQR 50.0-132.0) (AZ-two doses, BA.1: 50.0, IQR 50.0-92.8; BA.2: 59.6, IQR 50.0-85.5) (J&J-one dose, BA.1: 50.0, IQR 50.0-58.6; BA.2: 50.0, IQR 50.0–50.0) (Fig. 3B). Next, nAb responses generated by different vaccination types were assessed. No differences in SARS-CoV-2 nAb titres were detected when comparing individuals who had received AZ-one dose, AZ-two doses, and J&J-one dose (Supplementary Fig. 4).

Weaker nAb responses in younger participants (

Among SARS-CoV-2 infected (not vaccinated) participants, those aged p p p = 0.043; Wilcoxon) (For Beta, p = 0.0005; 40–59 years: 93.3, IQR 50.0-215.7, p = 0.0003; ≥60 years: 78.3, IQR 50.0-323.6, p = 0.031; Wilcoxon) (Fig. 4). For Delta, those p = 0.031) and 40–59 years (67.5, IQR 50.0-177.8, p = 0.014; Wilcoxon). Among those vaccinated (not infected) (≥15 years only), no differences by age group were observed (Supplementary Fig. 5). For those infected and vaccinated, the 15–39 year olds had lower nAb titres than the 40–59 year olds (for ancestral B.1, p = 0.044; Beta, p = 0.0036; Delta, p = 0.044; Wilcoxon) and the ≥60 year olds (for ancestral B.1, p = 0.012; Beta, p = 0.0036; Delta, p = 0.044; and Omicron BA.2, p = 0.044; Wilcoxon) (Supplementary Fig. 5).

Comorbidity reporting did not influence nAb titres

We determined whether comorbidity status influenced SARS-CoV-2 nAb titres. Among SARS-CoV-2 infected (not vaccinated) participants we observed no significant difference between those reporting no comorbidities vs. those reporting at least one comorbidity (for ancestral B.1, p = 0.89; Alpha, p = 0.97, Beta, p = 0.65; Delta, p = 0.65; Omicron BA.1, p = 0.97; Omicron BA.2, p = 0.69, Wilcoxon) (Supplementary Fig. 6). The same was observed among vaccinated participants (combining both those infected and vaccinated and solely vaccinated under “COVID-19 vaccinated” due to limited participants when stratifying) (Supplementary Fig. 6).

Lower nAb titres in HIV-infected individuals than HIV-uninfected individuals

The effect of HIV status on nAb titres was investigated using the VSV-based pseudotype system, comparing all HIV-infected participants with a representative subset of HIV-uninfected participants that were tested with this assay. As above, due to limited samples from people living with HIV, those infected and vaccinated, and solely vaccinated were combined under “COVID-19 vaccinated”. Alpha was again excluded as it was only tested at Survey 1, when few HIV-infected participants (n = 4) were SARS-CoV-2 nAb positive. Among SARS-CoV-2 infected (not vaccinated) individuals, people living with HIV had lower nAb titres to ancestral B.1 and Omicron BA.2 than HIV-uninfected participants (ancestral B.1: 59.1, IQR 50.0-88.6 vs. 105.6, IQR 73.6-161.1, respectively, p = 0.012; Omicron BA.2: 50.0, IQR 50.0–60.0 vs. 75.8, IQR 61.6-130.9, p = 0.012; Wilcoxon) (54). No significant differences were observed for Beta, Delta, and Omicron BA.1 (Beta, p = 0.67; Delta, p = 0.14; Omicron BA.1, p = 0.89) (Fig. 5). Among COVID-19 vaccinated individuals, those living with HIV had lower nAb titres across variants compared with HIV-uninfected participants (ancestral B.1: 61.1, IQR 50.0-82.4 vs. 225.8, IQR 121.1-542.5, respectively, p p = 0.0004; Delta: 50.0, IQR 50.0-59.3 vs. 75.9, IQR 52.3-212.9, p = 0.0001; Omicron BA.1: 50.0, IQR 50.0-73.7 vs. 68.9, IQR 50.0-165.4, p = 0.010; Omicron BA.2: 53.8, IQR 50.0-79.4 vs. 115.6, IQR 56.0-184.6, p = 0.016; Wilcoxon) (Fig. 5).