Preserved SARS-CoV-2 T-cell responses despite impaired humoral immunity in children with profound B-cell lymphopenia

Abstract

Defining vaccine-induced protection in children with humoral immunodeficiency is essential to guide SARS-CoV-2 vaccination strategies in this high-risk population. We conducted a longitudinal analysis of SARS-CoV-2 immunity at 1, 6 and 12 months after a primary Pfizer-BioNTech mRNA vaccine series in 27 children aged 5-11 years with primary or secondary antibody deficiencies and 48 matched healthy controls. Functional T-cell responses were quantified by IFN-γ and IL-2 ELISpot, and SARS-CoV-2-specific B-cells and T-cells were assessed by spectral cytometry. Systemic and mucosal antibody responses were measured in serum and saliva, and neutralizing activity against ancestral and Omicron BA.5 strains was evaluated through microneutralization. Children with humoral immunodeficiency exhibited impaired systemic antibody responses after two mRNA doses, even after SARS-CoV-2 infection. A third dose improved humoral immunity in children with preserved B-cell compartments but did not rescue neutralizing antibody responses in those with severe B-cell lymphopenia. In contrast, preserved, polyfunctional SARS-CoV-2-specific T-cell responses were observed in children with humoral immunodeficiency, including those with severe B-cell lymphopenia, and were higher in asymptomatic immunocompromised children. These findings reveal a dissociation between humoral failure and preserved cellular immunity in B-cell-deficient children, supporting timely vaccination and integration of T-cell responses into vaccine-response assessment when neutralizing antibodies are absent.

Introduction

Immunocompromised (IC) children, whether due to primary immunodeficiencies arising from inborn errors of immunity (IEI) or secondary to immunosuppressive therapies, often exhibit impaired B‑ and/or T‑cell function, leading to compromised adaptive immunity^1,2. Consequently, they are disproportionately susceptible to infections such as SARS‑CoV‑2, and face increased risks of severe disease, complications, and even death compared to healthy children^3,4. Different degrees of immunosuppression are associated with distinct and graded clinical outcomes in COVID-19. Individuals who are severely IC have a markedly higher risk of adverse outcomes than those who are moderately IC⁵, while patients with isolated antibody deficiencies appear to be less vulnerable than those with combined immune defects or immune dysregulation⁶.

Vaccination remains the most effective strategy to protect children from COVID-19 hospitalization and severe disease^7,8. In healthy pediatric populations, protection against symptomatic COVID-19 correlates with neutralizing antibody titers targeting the SARS-CoV-2 spike protein⁹. However, humoral responses to the standard two-dose regimen are often weak or undetectable in IC children, particularly in those with B-cell defects^10–13. This has led to recommendations for an augmented three-dose primary series in both adult and pediatric IC individuals, although the durability and breadth of this protection is still not well defined¹⁴. Importantly, protection against severe COVID-19 does not rely solely on humoral immunity. T-cell responses play a critical and complementary role in limiting disease progression, as demonstrated in IC adults^15,16. Moreover, vaccine immunogenicity varies depending on the underlying condition or treatment received, underscoring the heterogeneity of this population and the need for tailored immunization approaches^17,18.

In this study, we conducted a comprehensive, longitudinal evaluation of SARS-CoV-2 immunity in children with primary and secondary antibody deficiencies following two or three doses of the Pfizer-BioNTech mRNA vaccine. Building on evidence that humoral responses are frequently blunted in this population, we hypothesized that, even in the context of profound B-cell lymphopenia, vaccine- and infection-induced T-cell immunity may remain intact and contribute to protection from severe COVID-19. To rigorously test this hypothesis, we integrated systemic and mucosal antibody profiling, variant-specific microneutralization, and high-dimensional T-cell analyses, and compared these responses to healthy pediatric controls receiving the standard two-dose regimen. We also assessed whether hybrid immunity enhances protection in immunocompromised children similarly to the effect observed in healthy cohorts.

Results

Participant characteristics

The IRYIS cohort consisted of children with primary or secondary humoral immunodeficiency. Twenty-seven children with a median age of nine years [6,11] (63% male) met the inclusion criteria and had samples suitable for analysis. They were compared with 48 healthy children with a median age of eight years [5,11] (50% male), recruited from three studies (IRYIS, CHOIR and EnCORE) (Table 1). The clinical characteristics and diagnoses of children with antibody deficiency are reported in Supplementary Table S1. Children with primary humoral immunodeficiency (n = 15/27, 56%) had either a combined immunodeficiency or a predominant antibody deficiency (X-linked agammaglobulinemia, hypogammaglobulinemia or common variable immune deficiency) as defined by the 2019 International Union of Immunological Societies classification, or were patients receiving subcutaneous or intravenous immunoglobulin replacement for recurrent infections, low memory B cells counts (less than 8% of total B cells) and/or dissociated vaccine responses with normal IgA and/or IgM levels (unclassified humoral immunodeficiency)¹⁹. Among children with secondary humoral immunodeficiency (n = 12/27, 44%), ten received anti-CD20 therapy (n = 10/12, 83%) with rituximab less than six months before receiving their first dose of COVID-19 vaccine or within the six months after vaccination initiation, with (6/10, 60%) or without (4/10, 40%) steroids or other immunosuppressive therapy. Two children (2/12, 17%) had agammaglobulinemia or hypogammaglobulinemia secondary to CAR-T cell therapy or hematopoietic stem cell transplantation.

Table 1.

Characteristics of children included in the study

	Number (%) of children with humoral immunodeficiency (n = 27)	Number (%) of healthy children (n = 48)
Demographics
Sex
Male	17 (63)	24 (50)
Female	10 (37)	24 (50)
Age (Y) at vaccination
Median [Min, Max]	9 [6, 11]	8 [5, 11]
Number of COVID-19 vaccine doses received as primary vaccination regimen
Two doses	10 (37)	48 (100)
Three doses	17 (63)	0 (0)
Time (M) since second vaccine dose at one-month analysis time point (n = 23)		(n = 46)
	2.2 ± 0.7	1.8 ± 0.7
Time (M) since third vaccine dose at one-month analysis time point (n = 16)
	1.4 ± 0.5	N.A.
Time (M) since last vaccine dose at six-months analysis time point (n = 24)		(n = 38)
	7.0 ± 1.0	6.2 ± 0.9
Time (M) since last vaccine dose at one-year analysis time point (n = 18)		(n = 30)
	13.8 ± 2.7	12.2 ± 1.0

Y Year, M Months, n Number of samples available for analysis at each time point.

N.A. Not applicable.

Children were vaccinated following their individual physicians’ recommendations. Therefore, ten children with humoral immunodeficiency (10/27, 37%) received only two doses of the COVID-19 vaccine as their primary series, while the remaining 17 children (17/27, 63%) received the augmented three-dose primary regimen. Immunologically, both IC subgroups were comparable at the onset of vaccination (Supplementary Table S2). Due to the limited number of samples within the IC group, results from all available samples were included in the analysis. Some children contributed samples at both the one-month post-second dose and one-month post-third dose visits (n = 10/27), whereas others did not have blood drawn at each visit. Furthermore, while serology and neutralizing antibodies were assessed at all collected time points, samples were not always of adequate quality or volume to perform cell-mediated immunity and salivary analyses, leading to varying sample numbers across analyses.

Many children included in the study reported a prior documented SARS-CoV-2 infection or were considered infected based on humoral or cellular anti-N assay results. Among children with humoral immunodeficiency, 74% (20/27) were considered infected with SARS-CoV-2 before or at the time of vaccination, compared with 50% (24/48) of healthy controls (Table 2). Most children in the humoral immunodeficiency (93%, 25/27) and healthy (88%, 42/48) groups were eventually infected with SARS-CoV-2 during the study and were therefore considered to have developed hybrid immunity by the end of follow-up. In both groups, 19% of children were reinfected during the study period. Among children with documented or immunologically inferred infection, 48% reported mild symptoms without hospitalization, while 52% had asymptomatic or unreported infections.

Table 2.

SARS-CoV-2 infections during the study

	Number (%) of children with humoral immunodeficiency (n = 27)	Number (%) of healthy children (n = 48)
SARS-CoV-2 Infection Status
No history of infection	2 (7)	6 (12)
History of infection	25 (93)	42 (88)
Before/at vaccination	20 (74)	24 (50)
After vaccination	5 (19)	18 (35)
Number of SARS-CoV-2 infections
1	20 (74)	33 (69)
2	5 (19)	9 (19)
Symptomatology during infection
Symptomatic*	12 (48)	20 (48)
Asymptomatic	1 (4)	7 (16)
Unreported infection	12 (48)	15 (36)

* Symptomatic children had mild disease requiring no hospitalization. Common COVID-19 symptoms included sore throat, cough, runny/stuffy nose, and headache. Few participants had fever, intestinal symptoms, loss of smell/taste, or fatigue/muscle pain.

Two mRNA vaccine doses elicit suboptimal humoral responses in children with B-cell defects

To delineate the impact of the number of vaccine doses in the primary vaccination series on immune responses in IC children, we first analyzed children without evidence of prior SARS-CoV-2 infection based on diagnostic testing or N-specific responses at inclusion. In this infection-naïve subset, only five of the ten IC children included after the second dose also had a post-third dose visit at which they remained infection-naïve. IC children mounted T-cell responses comparable to healthy controls after two and three doses, as shown by similar frequencies of IFN-γ- and IL-2-secreting T cells (Fig. 1A). In contrast, humoral immune responses were substantially reduced in IC children after two vaccine doses. Two vaccine doses led to reduced serum IgG to S and RBD (Fig. 1B), with a tendency toward reduced IgA titers (Fig. S1A). Similarly, salivary anti-S IgG levels were significantly lower in IC children than in healthy children, likely reflecting systemic IgG responses, while salivary anti-S IgA remained uniformly low across groups, as expected following intramuscular vaccination (Fig. S1B)^20,21. After a third dose, serum antibody titers in IC children increased to levels similar to those observed in healthy controls after the standard two-dose regimen. Neutralizing antibody (NtAb) responses showed an even more pronounced deficit in IC children. Only 40% (4/10) of IC children had detectable NtAb after two doses, compared with 96% (24/25) of healthy controls. Following three doses, all IC children in this infection-naïve subset (5/5) generated detectable NtAb (Fig. 1C), with increased serum IgG levels to S and RBD (Fig S1C). Collectively, these findings indicate that two doses are insufficient to elicit robust humoral immunity in children with antibody deficiencies and that an augmented three-dose primary series improves humoral response compared with two doses.

Fig. 1. Two mRNA vaccine doses elicit suboptimal humoral responses in children with B-cell defects.

A–C Cellular and humoral immunity were assessed one month after the last vaccine dose in children who had no history of SARS-CoV-2 infection. Samples were analyzed for healthy children who received two doses of COVID-19 vaccine (HC, 2 doses, N = 19 to 25, black), for immunocompromised children who received two doses of COVID-19 vaccine (IC, 2 doses, N = 5 to 10, light blue), and immunocompromised children who received three doses of COVID-19 vaccine (IC, 3 doses, N = 4 to 5, dark blue). A Functional T-cell responses assessed by ELISpot following stimulation of PBMCs with ancestral SARS-CoV-2 spike peptides, expressed as IFN-γ (left panel) or IL-2 (right panel) producing cells per 10⁶ PBMCs. B Serum anti-spike IgG (left panel) and anti-RBD IgG (right panel) measured by ELISA. C Neutralizing antibodies for the ancestral strain were measured through live SARS-CoV-2 microneutralization assays on serum samples. Samples that did not neutralize SARS-CoV-2 at the 1:20 serum dilution were represented as a neutralizing titer of 10 for graphic representation and statistical analysis. A–C Dashed black lines indicate the positive threshold value when applicable. The percentage of samples above this positive threshold value is indicated for each group. Boxes indicate mean ± SEM. Statistical significance was established as not significant (ns) P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.

Hybrid immunity does not fully restore humoral immunity in children with B-cell defects

Because most children were infected with SARS-CoV-2 during the study period, we next evaluated immune responses in those with hybrid immunity. As observed in infection-naïve participants, IC and healthy children mounted comparable functional T-cell responses, regardless of immunocompromised status or number of vaccine doses (Fig. 2A). Humoral immunity remained impaired in IC children despite infection. IC children who received two doses showed reduced serum IgG and IgA levels to S and RBD compared with healthy children (Fig. 2B and Fig. S2A). Similarly, in saliva, anti-S and anti-RBD IgG levels were significantly lower in IC participants, whereas IgA responses remained uniformly low in most children and did not differ between groups (Fig. S2B). A third vaccine dose increased IgG to levels comparable to healthy controls, whereas anti-RBD IgA responses remained attenuated in IC children. Hybrid immunity did not compensate for defective neutralizing responses in IC children who received only two doses (Fig. 2C). Despite documented SARS-CoV-2 infection, only 38% (6/13) of IC children had detectable NtAb, with a median titer of 10 after two doses, compared with 226 in healthy children (P = 0.0002). A third vaccine dose increased NtAb titers to 50, yet levels remained significantly lower than in healthy controls (P = 0.0043). Whereas hybrid immunity was associated with an approximately twofold increase in NtAb titers in healthy children compared with vaccination alone (2.2-fold; P = 0.0309), this increase was not observed in IC children, whose NtAb titers remained comparable whether they were vaccinated only or had hybrid immunity (IC two-dose and three-dose groups, P > 0.9999). Together, these data show that while hybrid immunity enhances T-cell responses in both healthy and IC children, IC children continue to exhibit compromised humoral immunity and a limited capacity to generate functional neutralizing antibodies, especially after two vaccine doses rather than the recommended augmented three-dose regimen.

Fig. 2. Hybrid immunity only partially enhances humoral responses in children with B-cell defects.

A–C Cellular and humoral immunity were assessed one month after the last vaccine dose in children who were infected by SARS-CoV-2. Samples were analyzed for healthy children who received two doses of COVID-19 vaccine (HC, 2 doses, N = 17 to 20, black), for immunocompromised children who received two doses of COVID-19 vaccine (IC, 2 doses, N = 10 to 13, light blue) and immunocompromised children who received three doses of COVID-19 vaccine (IC, 3 doses, N = 9 to 11, dark blue). A Functional T-cell responses assessed by ELISpot following stimulation of PBMCs with ancestral SARS-CoV-2 spike peptides, expressed as IFN-γ (left panel) or IL-2 (right panel) producing cells per 10⁶ PBMCs. B Serum anti-spike IgG (left panel) and anti-RBD IgG (right panel) measured by ELISA. C Neutralizing antibodies for the ancestral strain were measured through live SARS-CoV-2 microneutralization assays on serum samples. Samples that did not neutralize SARS-CoV-2 at the 1:20 serum dilution were represented as a neutralizing titer of 10 for graphic representation and statistical analysis. Fold increase between children with hybrid immunity and children with no history of SARS-CoV-2 infection is indicated for each group and was calculated based on median neutralizing antibody titers. A–C Dashed black lines indicate the positive threshold value when applicable. The percentage of samples above this positive threshold value is indicated for each group. Boxes indicate mean ± SEM. Statistical significance was established as not significant (ns) P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001.

Long-term humoral immunity remains impaired in immunocompromised children despite hybrid immunity

We next examined the persistence of immunity six and twelve months after vaccination, acknowledging that the high incidence of SARS-CoV-2 infection across the cohort represents a major confounding factor for all long-term immune readouts (Fig. 3A). By six months, most children, both IC and healthy, had already acquired hybrid immunity, while by twelve months all except one healthy control had been infected. This implies that responses at later time points likely reflect the combined effects of vaccination and infection rather than vaccine-derived immunity alone. Within this context, T-cell responses in IC children appeared stable up to twelve months after vaccination, independently of the number of vaccine doses received, and remained reactive to the ancestral Spike and Omicron BA.4/BA.5 and XBB.1.5 variants, with secretion of both IFN-γ and IL-2 (Fig. 3B and Fig. S3A). Consistent with this, activation of S-specific CD4⁺ and CD8⁺ T cells remained stable and comparable between groups (Fig. S3B). These findings indicate that children with primary or secondary B-cell defects can maintain a functional memory T-cell pool over extended periods.

Fig. 3. Humoral immunity remains impaired up to one year after vaccination, while cellular immunity is preserved.

A–E Available samples were analyzed from healthy children who received two doses of COVID-19 vaccine (HC, 2 doses, N = 38), immunocompromised children who received two doses of COVID-19 vaccine (IC, 2 doses, N = 22) and immunocompromised children who received three doses of COVID-19 vaccine (IC, 3 doses, N = 15). Each participant was assigned to a single vaccination group. A Proportion of participants with documented or immunologically inferred SARS-CoV-2 infection at each time point. Numbers inside bars indicate infected participants/participants with available samples, stratified by cohort and time since vaccination. B Functional T-cell responses assessed by ELISpot following stimulation of PBMCs with ancestral SARS-CoV-2 spike peptides, expressed as IFN-γ (left panel) or IL-2 (right panel) producing cells per 10⁶ PBMCs. C Serum anti-spike IgG (left panel) and anti-RBD IgG (right panel) measured by ELISA. D SARS-CoV-2 spike-specific B cells were quantified through spectral flow cytometry. Spike-specific B cells are expressed in absolute count per million of PBMCs. E Neutralizing antibodies for the ancestral strain (left panel) and the Omicron BA.5 variant (right panel) were measured through live SARS-CoV-2 microneutralization assays on serum samples. Samples that did not neutralize SARS-CoV-2 at the 1:20 serum dilution were represented as a neutralizing titer of 10 for graphic representation and statistical analysis. B–E Dashed black lines indicate the positive threshold value when applicable. Boxes indicate mean ± SEM. Statistical significance was established as not significant (ns) P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

In contrast, despite numerous infections, we observed persistently attenuated humoral responses after two vaccine doses in IC children compared with healthy controls. This was particularly evident from the markedly lower IgG and IgA levels to S and RBD at six months (Fig. 3C and Fig. S3C), the substantially reduced pool of Spike-specific B cells (Fig. 3D), and the near absence of neutralizing antibodies to the ancestral strain at this time point (Fig. 3E). A third vaccine dose partially restored these humoral responses, increasing serum S- and RBD-specific IgG and IgA levels, and neutralizing capacity against ancestral and Omicron BA.5 strains toward levels observed in healthy children. Yet, S-specific B-cell frequencies remained lower in IC children at six months, and serum IgA responses declined by twelve months compared with healthy controls. Together, these findings demonstrate that, even in the setting of widespread hybrid immunity, IC children, especially those vaccinated only twice, fail to generate durable and fully functional humoral immunity, underscoring a persistent defect in their capacity to mount and maintain protective antibody responses.

Reduced memory and Spike-specific B-cell responses are associated with impaired neutralization in IC children

To investigate the immune features associated with impaired humoral responses in IC children, we performed high-dimensional flow cytometry to phenotype and enumerate total and S-specific B cells, CD27⁺ memory B cells, plasmablasts and follicular helper T cells (cT_FH), focusing on the one-month post-vaccination time point to capture peak vaccine-induced responses while avoiding bias from repeated sampling across time. At this time point, IC children exhibited reduced frequencies of Spike-specific B cells and CD27⁺ memory B cells compared with healthy controls (Fig. 4A), whereas total B-cell frequencies were unexpectedly comparable, and plasmablasts and cT_FH frequencies were broadly equivalent between groups (Fig. S4A).

Fig. 4. B-cell and Spike-specific B-cell responses correlate with neutralizing antibody capacity.

A–C Spectral flow cytometry was performed to assess the determinants linked to the absence of neutralizing antibodies in children with humoral immune deficiency. Non-specific B-cell phenotyping as well as SARS-CoV-2-specific staining were done at the closest available time point to vaccination in HC children or immunocompromised children without a history of SARS-CoV-2 infection (black or blue dots, respectively) and in children with prior infection (red dots). Available samples were analyzed from healthy children who received two doses of COVID-19 vaccine (HC, 2 doses, N = 18, black, vaccinated only and N = 4, red, hybrid immune) and immunocompromised children who received 2 or 3 doses of COVID-19 vaccine (IC, 2 or 3 doses, N = 5, light blue, vaccinated only and N = 9, red, hybrid immune). A Frequencies of SARS-CoV-2 spike-specific B cells (left panel) and CD27⁺ memory B cells (right panel) are shown for HC and IC groups. Boxes indicate mean ± SEM. B Correlations between neutralizing antibody titers and total B cells (left panel), spike-specific B cells (middle panel), and CD27⁺ memory B cells (right panel) at one month post-vaccination in the combined HC and IC cohort. C Correlations between neutralizing antibody titers and the same B-cell populations restricted to immunocompromised children (IC) at one month post-vaccination. Statistical significance was established as not significant (ns) P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001. Correlation was assessed by Spearman correlation coefficient. Coefficient R and P-values are shown.

To determine which immune abnormalities best predicted defective antibody responses, we next assessed correlations between NtAb levels and multiple cellular parameters. Across the entire cohort, total B cell counts, Spike-specific B-cell counts, and the frequency of memory B cells showed moderate but statistically significant positive correlations with NtAb titers (Spearman R ≈ 0.40-0.44, P ≈ 0.005-0.010, n = 36, Fig. 4B), while plasmablast and cT_FH frequencies did not correlate with neutralization capacity (Fig. S4B). Importantly, when restricting the analysis to IC children only, these associations were further strengthened (Spearman R ≈ 0.61-0.66, P ≈ 0.006-0.013), despite the smaller sample size (n = 16, Fig. 4C), indicating that both the magnitude of the B-cell compartment and the ability to generate Spike-specific B cells are closely linked to functional neutralization in this subgroup. Together, these analyses indicate that total B-cell counts, CD27⁺ memory B-cell frequencies, and the generation of Spike-specific B cells are closely linked to neutralizing antibody responses in IC children, while cT_FH and plasmablast frequencies did not correlate with neutralization capacity at the one-month post-vaccination time point.

Children with severe B-cell lymphopenia show particularly impaired humoral immune responses to SARS-CoV-2

Because total B-cell counts correlated with neutralizing capacity at the time point closest to vaccination, we next examined whether IC children with the most severely impaired humoral responses were those with the lowest circulating B-cell numbers at the time of vaccination. IC participants were stratified according to the presence (n = 15) or absence (n = 12) of profound B-cell lymphopenia, defined as a CD19⁺ lymphocyte count below the 5^th percentile for age, while other immune parameters were comparable between IC groups, except for IgM levels (Supplementary Table S3). Among the 15 children with severe B-cell lymphopenia, five had primary immunodeficiency, while the other ten children had secondary immunodeficiency (SID). Within the ten SID children, nine were under rituximab therapy with or without other immunosuppressants, and one patient had received anti-CD19 CAR-T cell therapy for leukemia. We then compared SARS-CoV-2-specific immunity across severely B-cell-lymphopenic children, B-cell-sufficient IC children, and healthy controls.

Children with severe B-cell lymphopenia exhibited markedly diminished humoral immunity across all compartments. Serum IgG and IgA levels to S and RBD were profoundly reduced (Fig. 5A and Fig. S5). While B-cell-sufficient IC children generated frequencies of S-specific B cells comparable to healthy controls, those with severe B-cell lymphopenia showed distinctly reduced induction of Spike-specific B cells (Fig. 5B). Consistent with these deficits, neutralizing antibody titers against the ancestral strain were extremely low or undetectable in this subgroup, in contrast to both healthy children and B-cell-sufficient IC peers (Fig. 5C). Furthermore, despite documented infection, children with severe B-cell lymphopenia showed minimal or absent neutralizing activity against BA.5, underscoring an intrinsic inability to mount adequate humoral immunity in the setting of profound B-cell deficiency. In contrast, children with severe B-cell lymphopenia exhibited heightened functional T-cell responses, measured by IFN-γ secretion in response to SARS-CoV-2 Spike, particularly one month after vaccination (Fig. 5D). Together, these results demonstrate that profound B-cell lymphopenia is a major determinant of impaired humoral immunity to SARS-CoV-2 in IC children, whereas preserved and potentially enhanced T-cell responses may contribute to disease control in the absence of functional B-cell immunity.

Fig. 5. Severe B-cell lymphopenia limits humoral responses while T-cell responses are preserved.

A–D Cellular and humoral immunity were assessed at one, six, and twelve months after the last vaccine dose in children without a history of SARS-CoV-2 infection (black dots) and in children with prior infection (red dots). Available samples were analyzed for healthy children who received two doses of COVID-19 vaccine (HC, 2 doses, N = 38), for immunocompromised children with B cell levels over the 5^th percentile for their age (IC, B cells sufficient, N = 10) and immunocompromised children with B cell levels under the 5^th percentile for their age (IC, severe B cell lymphopenia, N = 15). A Serum anti-spike IgG (left panel) and anti-spike IgA (right panel) measured by ELISA. B SARS-CoV-2 spike-specific B cells were quantified through spectral flow cytometry. Spike-specific B cells are expressed in absolute count per million of PBMCs. C Neutralizing antibodies for the ancestral strain (left panel) and the Omicron BA.5 variant (right panel) were measured through live SARS-CoV-2 microneutralization assays on serum samples. Samples that did not neutralize SARS-CoV-2 at the 1:20 serum dilution were represented as a neutralizing titer of 10 for graphic representation and statistical analysis. D Functional T-cell responses assessed by ELISpot following stimulation of PBMCs with ancestral SARS-CoV-2 spike peptides, expressed as IFN-γ-producing cells per 10⁶ PBMCs. A–D Dashed black lines indicate the positive threshold value when applicable. Boxes indicate mean ± SEM. Statistical significance was established as not significant (ns), P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

Functional T-cell responses are associated with reduced COVID-19 symptom severity

Vaccine efficacy is often reported in terms of its capacity to protect against severe infection and hospitalization. Although this small cohort of high-risk patients does not allow for proper statistical analysis of vaccine efficacy, we observed that although most IC children were infected during the study period (25/27), none reported severe disease nor required hospitalization (Table 2). Children were either mildly symptomatic or had asymptomatic infections. As T-cell responses have been shown to restrict COVID-19 symptoms¹⁵, we asked whether a similar association was observed in our IC pediatric cohort. We therefore compared T-cell responses in children with or without symptoms, in both healthy and IC groups. IC children without symptoms had higher levels of IFN-γ secreting cells in response to both ancestral S peptides and Omicron BA.4/BA.5 S peptides than IC children who reported symptoms, whereas this difference was not observed in healthy controls (Fig. 6A). In contrast, NtAb levels against ancestral Spike or Omicron BA.5 did not differ according to infection symptomatology (Fig. 6B). These findings suggest that, in the absence of robust humoral immunity, S-specific functional T-cell responses may contribute to limiting COVID-19 symptom severity in children with humoral immunodeficiency.

Fig. 6. SARS-CoV-2-specific T-cell responses are associated with reduced COVID-19 symptom severity in children with humoral immunodeficiency.

A, B Cellular and humoral immunity were assessed at all available post-vaccination time points in healthy children (HC) and in children with humoral immune deficiency (IC) who experienced an asymptomatic or undeclared SARS-CoV-2 infection (green dots, N = 7 to 21 HC and N = 7 to 13 IC) compared to children who experienced mildly symptomatic COVID-19 (yellow dots, N = 7 to 20 HC and N = 6 to 12 IC). A Functional T-cell responses assessed by ELISpot following stimulation of PBMCs with SARS-CoV-2 spike peptides from the ancestral strain (left panel) and Omicron BA.4/BA.5 strain (right panel), expressed as IFN-γ-producing cells per 10⁶ PBMCs. B Neutralizing antibodies for the ancestral strain (left panel) and the Omicron BA.5 variant (right panel) were measured through live SARS-CoV-2 microneutralization assays on serum samples. Samples that did not neutralize SARS-CoV-2 at the 1:20 serum dilution were represented as a neutralizing titer of 10 for graphic representation and statistical analysis. A, B Dashed black lines indicate the positive threshold value when applicable. Boxes indicate mean ± SEM. Statistical significance was established as not significant (ns), P > 0.05, *P < 0.05.

Discussion

In this study, we provide a comprehensive analysis of adaptive immune responses to SARS-CoV-2 vaccination and infection in children with primary or secondary humoral immunodeficiencies. Our data reveal a clear dissociation between humoral and cellular immunity in this high-risk pediatric population. While antibody responses varied according to B-cell status, SARS-CoV-2-specific T-cell responses were preserved across immunocompromised subgroups, maintained through time and were associated with reduced COVID-19 symptomatology. These findings highlight the limitations of serum IgG and neutralizing antibody titers as universal correlates of protection and support the relevance of evaluating cellular immunity in children with humoral immune defects.

We show that a standard two-dose mRNA vaccination regimen is insufficient to induce robust and durable humoral immunity in children with humoral immunodeficiency, even in the context of SARS-CoV-2 infection. In contrast, a three-dose primary series increased binding and neutralizing antibody responses to levels comparable to healthy controls when the B-cell compartment was preserved. This supports the value of an augmented primary series in IC children, while also showing that vaccine responsiveness remains constrained by B-cell availability. Our findings extend recent pediatric studies evaluating SARS-CoV-2 vaccine responses in immunocompromised children. Tagarro et al. reported that a three-dose BNT162b2 regimen induced humoral responses in immunosuppressed children aged 5-11 years comparable to those observed in healthy children after two doses, although their cohort was heterogeneous and primarily included children receiving biologic therapy, chemotherapy or radiotherapy, or solid organ transplantation²². Similarly, Porto et al. showed in a multicenter study of children with IEI that, six months after a third vaccine dose, binding and neutralizing antibody responses remained attenuated compared with healthy controls, particularly among children with predominant antibody deficiency, whereas Spike-specific T-cell responses were preserved²³. Our study complements these data by focusing on children with primary or secondary humoral immunodeficiency, stratifying responses according to B-cell availability, and showing that profound B-cell lymphopenia identifies children who remain unable to generate effective neutralizing antibodies despite vaccination and infection. In this setting, preserved SARS-CoV-2-specific T-cell responses reveal a clinically relevant dissociation between humoral and cellular immunity in children with B-cell defects.

Importantly, children with severe B-cell lymphopenia remained unable to mount effective IgG, IgA, and neutralizing antibody responses despite repeated vaccination and infection. This is consistent with adult studies showing improved immunity after three vaccine doses in most IC groups, but persistently impaired responses in individuals receiving anti-CD20 therapy^24,25. In our cohort, B-cell lymphopenia predicted poor antibody and neutralization responses independently of whether children had primary or secondary immunodeficiency. These findings support B-cell availability as a key constraint on vaccine-induced humoral immunity, even after repeated antigen exposure.

By integrating immune phenotyping with functional antibody analyses, we found that neutralizing antibody capacity was linked to total B-cell counts, CD27⁺ memory B-cell frequencies and Spike-specific B-cell responses. These associations support B-cell immunophenotyping as a useful tool to interpret vaccine responsiveness in IC children. In contrast, cT_FH and plasmablast frequencies did not correlate with neutralization at the one-month post-vaccination time point, although earlier sampling would be required to fully assess their contribution to early vaccine-induced humoral responses.

In contrast to humoral immunity, SARS-CoV-2-specific T-cell responses were preserved across IC children, including those with severe B-cell lymphopenia. These responses were functional and cross-reactive, with IFN-γ and IL-2 secretion after stimulation with ancestral and Omicron BA.4/BA.5 and XBB1.5 Spike antigens. Stronger functional T-cell responses were observed in IC children with asymptomatic infections compared with those reporting symptoms, a relationship not observed in healthy controls. None of the IC children required hospitalization despite high infection rates and marked humoral impairments, supporting a role for cellular immunity in limiting disease severity when antibody responses are absent. Similar heightened T-cell responses have been described in adults treated with anti-CD20 antibodies and individuals with agammaglobulinemia, although their relationship with symptom severity has not been clearly established^25–30.

Although frequent infections during follow-up limit conclusions about vaccine-induced persistence, these exposures may have boosted T-cell responses and contributed to their maintenance over time, contrasting with reports of accelerated waning of immunity in IC children³¹. Together, these findings support prompt vaccination even in children with severe B-cell lymphopenia, as cellular immunity remains inducible and may contribute to disease control when humoral responses are absent.

These findings have direct implications for the clinical management of children with humoral immunodeficiencies. They support an augmented primary vaccination strategy while emphasizing that vaccine responses should not be evaluated solely through serum IgG and neutralizing antibody titers. Instead, B-cell immunophenotyping may help identify children unlikely to mount effective humoral responses and guide the use of additional protective strategies, including closer monitoring, monoclonal antibody prophylaxis, updated boosters or mucosal vaccine approaches. This framework is particularly relevant for children with profound B-cell lymphopenia or recent B-cell-depleting therapy, in whom cellular immunity remains inducible despite limited antibody responses.

Because currently circulating SARS-CoV-2 variants remain largely Omicron descendants and updated vaccines continue to be recommended for moderately to severely IC individuals, defining cellular immunity in children with profound B-cell lymphopenia remains relevant to ongoing booster strategies. Our findings support prompt vaccination of IC children, even when humoral responses are expected to be limited, because cellular immunity remains inducible. They also support continued evaluation of vaccine platforms and schedules that optimize cellular immunity in populations with impaired B-cell responses³². Although T-cell immunity is generally more resistant to viral evolution than neutralizing antibodies, future studies should determine how updated boosters and ongoing viral evolution affect the durability and breadth of cellular protection in children with profound B-cell lymphopenia³³.

This study has several limitations. The high incidence of SARS-CoV-2 infection during follow-up limited our ability to isolate vaccine-induced immunity, particularly for long-term analyses, and infection timing was often inferred retrospectively using anti-nucleocapsid responses. In addition, the absence of pre-vaccination baseline samples limited assessment of pre-existing immunity. Immunoglobulin replacement may also have contributed to serum binding IgG measurements in some participants, although this would be expected to overestimate rather than underestimate humoral immunity. Blood volume constraints, incomplete sampling, and small IC subgroups reduced statistical power for some immune phenotyping analyses. Finally, T-cell analyses focused primarily on Spike peptide pools, limiting assessment of broader non-Spike responses after infection. Future studies with prospective infection surveillance, baseline sampling, and standardized longitudinal follow-up will be essential to refine correlates of protection in this population.

In summary, our findings show that vaccine responses in children with humoral immunodeficiency are strongly shaped by B-cell availability. A three-dose mRNA vaccination regimen can improve humoral responses when the B-cell compartment is preserved, whereas children with severe B-cell lymphopenia remain unable to generate effective neutralizing antibodies despite vaccination and infection. In this setting, preserved SARS-CoV-2-specific T-cell immunity is associated with mild or asymptomatic infection and may contribute to disease control when humoral responses are absent. These results support immune monitoring beyond antibody titers, prompt vaccination without waiting for B-cell reconstitution, and vaccine strategies that sustain functional cellular immunity in immunocompromised children.

Methods

Study design and data collection

We performed an observational study of SARS-CoV-2 mRNA vaccine responses in children with primary and secondary antibody deficiency. The “Immune Response in Young ImmunoSuppressed children to COVID-19 vaccination” (IRYIS) study cohort included children aged 5-to-11-years with primary or secondary antibody deficiency, as well as healthy control children recruited and followed in the Province of Quebec, Canada. In total, 27 children with antibody deficiency were included between November 2021 and July 2022. They were compared with 48 healthy children recruited between November 2021 and October 2022 through three studies: IRYIS, the “Children and Older Teens Immune Response to SARS-CoV-2 in Montreal” (CHOIR) study, and the “Children and COVID-19: Seroprevalence study” (EnCORE) study, corresponding to the Montreal cohort of the Pediatric Network Expansion study “Understanding Immune Responses to SARS-CoV-2 Infection and Vaccination in Children Aged from Five to Eleven Years”.

Participants were vaccinated according to standard public health recommendations between December 2021 and September 2022 and were followed for up to one year after completion of their primary vaccination series. All participants received the Pfizer-BioNTech Comirnaty COVID-19 vaccine at the 10 µg dose indicated for children aged 5-11 years. None of the participants included in the analyses received a booster dose during the study period. At the time of their vaccination, the National Advisory Committee on Immunization (NACI) recommended an enhanced three-dose primary vaccination regimen for moderately to severely IC individuals³⁴. Because this was an observational study, vaccination was not assigned by protocol; children received vaccines through provincial public health programs according to their treating physician’s recommendations. Consequently, some IC children received a two-dose primary series, whereas others received the recommended three-dose primary series.

Clinical characteristics and baseline immunological data of IC children

Detailed clinical characteristics of the IC cohort are provided in Supplementary Table S1. Baseline immunological parameters of IC children who received two or three vaccine doses are reported and compared in Supplementary Table S2. Baseline immune characteristics of participants with or without severe B-cell lymphopenia are reported in Supplementary Table S3. Severe T- or B-cell lymphopenia was defined as an absolute cell count lower than the 5^th percentile for age³⁵.

Determination of SARS-CoV-2 infection status

SARS-CoV-2 infection status was determined using complementary clinical and immunological approaches. At each visit, documented SARS-CoV-2 infection confirmed by rapid antigen testing was assessed, and the presence or absence of COVID-19-associated symptoms was documented. At each time point, immune responses to SARS-CoV-2 nucleocapsid (N) protein were evaluated using serum anti-N IgG titers and IFN-γ⁺ T-cell responses to N peptide pools. Children were considered to have been infected with SARS-CoV-2 if they had a positive COVID-19 test, detectable anti-N IgG, and/or a detectable T-cell response to N peptides, even in the absence of reported symptoms.

Sample collection and processing

Saliva and blood samples were collected from participants at multiple time points: one month after their second and third vaccine doses, and at six and twelve months post-vaccination. Detailed time points are presented in Table 1. Saliva samples were collected using the Salivette system (Sarstedt, Numbrecht, Germany). Blood samples were collected into serum separation tubes (SST^TM, BD) and acid-citrate-dextrose tubes (ACD, BD). Following collection, saliva, serum, plasma, and peripheral blood mononuclear cells (PBMCs) were isolated and cryopreserved according to standard operating procedures at the Mother-Child Institutional Biobank at the Sainte-Justine University Hospital and Research Center (CR-CHUSJ) as previously reported by our team^20,36,37. PBMCs were stored in liquid nitrogen, and all other samples were stored at -80 °C. Assays were performed using all available samples for each participant at each visit. However, biological material was sometimes insufficient to perform every assay, resulting in variation in sample numbers across analyses.

Ethic approvals

The IRYIS and the CHOIR/EnCORE protocols were conducted in line with the Declaration of Helsinki and approved by the Research Ethics Board (REB) at the CR-CHUSJ under study numbers MP-21-2022-3909 and MP-21-2021-3105. The CHOIR/EnCORE study was also approved by the University of Montreal REB under study number CERSES-20-103-P. Participants were enrolled after obtaining written informed consent from one of their parents or legal guardians.

Cell-mediated immunity

PBMCs isolated from blood samples were rapidly thawed and rested overnight. Each sample was assessed for viability. The number of assay(s) performed was determined based on the number of cells with adequate viability for each participant at each of their visits.

Enzyme-linked immunospot assay (ELISpot) for the quantification of functional cellular immunity

ELISpot assay was used to detect interferon-gamma (IFN-γ) or interleukin-2 (IL-2) secreting cells. PBMCs were stimulated for 20 hours with 1 µg/mL of a spike (S) or nucleocapsid (N) SARS-CoV-2 peptide mega pool from the ancestral strain, as well as an S peptide mega pool from the Omicron subvariants BA.4/BA.5 and XBB.1.5 (JPT Peptide Technologies, JPT, Berlin, Germany) as previously described^20,36,37. Culture media without peptide (AIM-V® Medium (1X), Thermo Fisher Scientific, Waltham, MA, USA) was used as a negative control, and the number of spots detected in the negative well of each sample was subtracted from all stimulated wells of the same sample. CytoStim^TM (Miltenyi Biotec, MA, USA) was used as positive control. CTL ImmunoSpot® SS UV Analyzer (Cellular Technology Ltd., OH, USA) was used to quantify spots with a detection limit of one secreting cell per 200,000. The positive threshold of 25 spot-forming units per million PBMCs was previously validated on pre-pandemic samples³⁶.

Activation Induced Markers (AIM) assay and circulating T follicular helper cells (cT_FH)

The AIM assay was used to detect activated CD4⁺ T cells and CD8⁺ T cells in response to stimulation with ancestral SARS-CoV-2 S peptides. PBMCs were stimulated for 20 hours with 1 µg/mL of a SARS-CoV-2 S peptide mega pool (JPT Peptide Technologies, Berlin, Germany). Following stimulation, cells were stained with a panel of fluorescent monoclonal antibodies for AIM and cT_FH identification, as described in Supplementary Table S4. cT_FH cells were defined as CD3⁺ CD4⁺ CXCR5⁺ PD1^hi T cells. Spike-specific CD4⁺ T cells were defined as CD3⁺ CD4⁺ CD134⁺(OX-40⁺) CD137⁺(4-1BB⁺) non naïve cells (excluding CD45RA⁺ CCR7⁺ T cells). Spike-specific CD8⁺ T cells were defined as CD3⁺ CD8⁺ CD45RA^- CD69⁺ CD137⁺(4-1BB⁺) non naïve cells (excluding CD45RA⁺ CCR7⁺ T cells). Data were acquired using the SONY Spectral Cell Analyzer ID7000. Flow cytometry data were analyzed using FlowJo software version 10.10.0. Fluorescence minus one (FMO) controls, unstimulated controls (negative) and CytoStim^TM (positive control) were used to define the gates. Dead cells were excluded using viability dye, and doublets were removed prior to analysis. Results are represented as the frequency of activated CD4⁺ or CD8⁺ T cells following peptide stimulation, after subtraction of the corresponding unstimulated condition for each sample.

B cell immunophenotyping

PBMCs were stained directly after overnight resting to assess SARS-CoV-2 S-specific B cells. The Miltenyi Biotec SARS-CoV-2 Spike B Cell Analysis Kit for humans was used to detect B cells binding to the recombinant SARS-CoV-2 S protein of the ancestral strain. Additional antibodies were added to the Miltenyi Biotec panel to identify plasmablasts as well as switched and unswitched memory B cells. Plasmablasts were defined as CD3^- CD19⁺ CD27^hi CD38⁺ cells. Switched memory B cells were defined as CD3^- CD19⁺ CD27⁺ IgD^- cells and unswitched memory B cells as CD3^- CD19⁺ CD27⁺ IgD⁺ cells. The complete panel of fluorescent monoclonal antibodies is described in Supplemental Table S4, and the gating strategy is shown in Supplementary Fig. S6. Data were acquired using the SONY Spectral Cell Analyzer ID7000 and analyzed using FlowJo software version 10.10.0. FMO and unstained controls were used to set the gates. Results are presented as the number of S-specific B cells per million PBMCs or as frequency of each subpopulation among total B cells.

Mucosal Immunity

Enzyme-linked immunosorbent assay on saliva

As previously published by our team, enzyme-linked immunosorbent assays (ELISA) were performed to detect IgG binding to S and RBD and IgA binding to S in saliva^20,21,38. Plates were coated with biotinylated S or RBD, incubated overnight, and subsequently blocked. Samples were then added to each well for two hours. Horseradish peroxidase (HRP)-conjugated IgG or IgA detection antibodies were then added for one hour. After a final wash, 3,3’,5,5’-Tetramethylbenzidine substrate (Thermo Fisher Scientific) was added for color development. Optical density (OD) was read at 450 nm using a Thermo Multiskan FC spectrophotometer. Samples were analyzed as previously described by subtracting the raw OD450 of phosphate-buffered saline (PBS) controls for each sample from their antigen-specific OD450 at each dilution^20,21. These background-subtracted values were used to calculate the area under the curve (AUC) for each sample, which was then adjusted to pre-selected standards consisting of pooled saliva from acute and convalescent COVID-19 patients, as described previously^20,21. This accounted for inter-plate variability. These adjusted AUCs were then divided by the average AUC of the positive control run on all plates, resulting in a final value expressed as the percentage of the positive control. The detection range and positivity threshold for each antigen are presented in Supplementary Table S5.

Humoral immunity

Serum Enzyme-linked immunosorbent assay

IgG and IgA binding antibody levels against the full-length SARS-CoV-2 S trimer, its RBD, and N were measured in serum samples using chemiluminescent ELISA as previously described³⁹. Briefly, LUMITRAC 600 high-binding white polystyrene 384-well microplates (Greiner Bio-One, Thermo Fisher Scientific, #78107425) were coated overnight with 10 µL/well of antigen (50 ng of S, 20 ng of RBD, or 7 ng of N), all supplied by the National Research Council of Canada (NRC). Plates were centrifuged at 1000 rpm for one minute to ensure even coating, then incubated overnight at 4 °C. The next day, the assay was performed at room temperature with two washes in 115 µL/well of PBS with Tween® detergent (PBS-T) before each of the following four steps. First, the wells were blocked for one hour in 115 µL/well of 5% Blocker BLOTTO (ThermoFisher Scientific, #37530). Second, 10 µL of diluted serum in 1% Blocker BLOTTO in PBS-T was added and incubated for two hours. Serum samples were diluted 1:160 and 1:2560 to assess anti-N IgG antibody levels, at 1:160, 1:2560 and 1:40960 to assess anti-RBD and anti-S IgG antibody levels, and at 1:160 and 1:2560 to assess IgA antibody levels. Third, 10 µL of an HRP-fused anti-human IgG (IgG#5 by NRC, 0.9 ng/well) or HRP-conjugated anti-human IgA (Jackson ImmunoResearch, #109-035-011, 0.8 ng/well) diluted in 1% final Blocker BLOTTO in PBS-T was added, followed by a one-hour incubation. Fourth, 10 µL/well of ELISA Pico Chemiluminescent Substrate (ThermoFisher Scientific, #37069, diluted 1:4 in MilliQ distilled H₂O) was added, and the plates were centrifuged at 1000 rpm for one minute. After five minutes of incubation, chemiluminescence signals were read on an EnVision 2105 Multimode Plate Reader (Perkin Elmer) at 100 ms/well using an ultra-sensitive detector.

Raw chemiluminescent values were normalized to a blank-subtracted reference point from the synthetic standard included on each assay plate, to create relative ratios. The synthetic standards used were as follows. For IgG, VHH72-Fc was used for S and RBD (supplied by NRC) and an anti-N IgG for N (clone HC2003, Genscript, #A02039). For IgA, anti-S CR3022 (Absolute Antibody, #Ab01680-16.0) and anti-N CR3018 (Absolute Antibody, #Ab01690-16.0) were used. For IgG and IgA results, relative ratios within the linear range of the assay were further converted to binding antibody units (BAU)/mL using the World Health Organization (WHO) International Standard 20/136, as previously described³⁹. Linear ranges for the sample dilutions performed and positivity thresholds are presented in Supplementary Table S5. Positivity thresholds were determined for the 1:160 dilution using three standard deviations from the mean of control samples (blanks, pre-pandemic negative sera, commercially purified IgG or IgA) as previously described, or two standard deviations for anti-N IgG .

Microneutralization assay

Neutralizing antibody (NtAb) titers were measured using a live SARS-CoV-2 microneutralization assay, as previously described by our team^36,40. Two strains were used: the ancestral Wuhan-1-like SARS-CoV-2 isolated from a clinical sample in March 2020 in Québec City and the Omicron variant BA.5 provided by the National Microbiology Laboratory (NML), Public Health Agency of Canada. Heat-inactivated serum was diluted by two-fold starting with 1:20 dilution. Serum and virus were mixed at equal volumes and incubated for one hour at room temperature. The residual infectivity of the virus-serum mixture was assessed in quadruplicate wells of African green monkey kidney E6 cell line (Vero ATCC^® CRL-1586^TM) for ancestral SARS-CoV-2 and in Vero cells overexpressing transmembrane protease serine 2 (TMPRSS2, JCRB) for Omicron BA.5. NtAb titers were defined as the reciprocal of the serum dilution that completely neutralized the infectivity of the 50% tissue culture infectious dose (TCID₅₀) of SARS-CoV-2 multiplied by 100, which was determined by the absence of cytopathic effect on cells after four days⁴¹. NtAb titers were then calculated using the Reed/Muench method⁴². These microneutralization assays were performed in the Containment Level 3 laboratory at the Centre Hospitalier Universitaire de Québec – Université Laval Research Center.

Statistics

Mann-Whitney or Kruskal-Wallis unpaired nonparametric tests were performed using Prism 9, version 9.2.0 (2021 GraphPad Software, LLC) to assess statistical significance. Simple linear regression is presented with Spearman correlation (R) and P-value (P). Significance was set as *P < 0.05, **P < 0.01, ***P < 0.001 and ****P < 0.0001. Time since infection is presented as mean ± SD, while experimental data are displayed as median [interquartile range, 25^th-75^th percentile].

Declaration of AI and AI-assisted technologies in the writing process

During the preparation of this work, the author(s) used ChatGPT 5.1 for language editing. After using this tool/service, the authors reviewed and edited the content as needed and take full responsibility for the content of the publication.

Author contributions

Conceptualization: H.D. Methodology: S.N., B.B., H.R., F.Q, M.D.B., L.S., L.Ward, S.S.M., G.C., K.C., Y.L., J.M.R., J.G., A.C.G., M.B., C.Q., H.D. Investigation: H.D. Visualization: S.N., B.B. Funding acquisition: H.D., K.Z. Project administration: B.B., M.B.D.T., L.Wang, S.Ni., K.A., Z.L, L.S. Subject recruitment: H.D., K.Z., C.Q. Supervision: H.D., C.Q., K.Z. Writing – original draft: S.N., S.S., B.B., H.D. Writing – review & editing: S.N., S.S., B.B., G.S., F.Q., L.Ward, K.C., M.B., C.Q., H.D.

Data availability

The datasets generated and/or analyzed during the current study are not publicly available due to patient confidentiality and institutional ethical restrictions but are available from the corresponding author upon reasonable request.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41541-026-01525-6.