This study adheres to the principles of the Declaration of Helsinki and was approved by the Brazilian National Research Ethics Committee (number 51535921.2.0000.5505). All participants provided written informed consent before enrollment.

This was a prospective cohort study that took place between October 2021 and November 2023 ( Supplementary Figure S1 in the Supplementary Material ). Patients with an established IEI diagnosis from a Brazilian reference center, the Immunology Clinic of the Federal University of São Paulo, who received three doses of COVID-19 vaccines, were offered the opportunity to join the study. A control group of healthy individuals of similar sex and age distributions were also invited to participate. One-third of patients with IEI and controls had a basic immunization schedule that consisted of two doses of either CoronaVac, BNT162b2 (Pfizer-BioNTech) or ChAdOx1 nCoV-19 (AstraZeneca), followed by a third dose of BNT162b2 (Pfizer-BioNTech). The fourth and fifth doses varied according to the availability of the different COVID-19 vaccine platforms in both the IEI cohort and the control cohort. Patients with IEI had blood samples collected 1 and 6 months after the fourth COVID-19 vaccine dose (the first booster for the IEI cohort) and 1 month after the fifth COVID-19 vaccine dose (the second booster for the IEI cohort). Blood samples were collected from control individuals 1 and 6 months after the third COVID-19 vaccine dose (the first booster for the control cohort) and 1 month after the fourth COVID-19 vaccine dose (the second booster for the control cohort) ( Figure 1 ).

Lymphocyte subsets (CD3⁺ T, CD4⁺ T, CD8⁺ T, CD19⁺ and NK cells) were assessed via a single platform (TruCount and Multitest, BD Biosciences, San Jose, CA) on a BD FACSCalibur™ 4-color flow cytometer using CellQuest software (BD Biosciences, San Jose, CA). Data analysis was performed via MultiSET v3.0.2 software (Becton, Dickinson and Company, New Jersey, USA).

Peripheral blood mononuclear cells (PBMCs) were isolated using Ficoll gradient density. Following isolation, the cells were stored in liquid nitrogen for later ELISpot assays to evaluate peptide-induced cytokine production.

Serum samples were separated on the day of blood collection and stored at -80°C for later analysis. The humoral immune response was assessed via the commercial SARS-CoV-2 NTChip^® Test (V-NTCGOK) (Viramed Biotech AG, Germany). This is an in vitro qualitative and quantitative multiplex microarray microsystem based on an enzyme immunoassay previously validated against the 50% Plaque Reduction Neutralization Test (PRNT₅₀) to determine the level of neutralizing antibodies against purified specific surface antigens from the RBD portion of the Spike protein, in printed spot triplets in a nitrocellulose membrane, from variants and subvariants of SARS-CoV-2, such as Wuhan – wild type, Delta, Omicron BA.1 – B.1.1.529 and BA.2 and BA.5, which are on each well of a 96-well plate ( Supplementary Figure S2 ).

During the development of the NTChip^® (19), in order to validate the PRNT₅₀ – the gold standard to determine the level of neutralizing antibodies for many viral diseases – six groups of sera were evaluated: 1) pre-pandemic (W-O-), 2) previous Omicron variant (W+O-), 3) infected/vaccinated post Omicron (W+O+), 4) naïve Omicron positive (W-O+), 5) Omicron positive 0-2 days and 6) 10 days after infection. Sensitivity and specificity results for RBD-Wuhan, RBD-Omicron BA.1 and Nucleocapsid (N) were: 96.7% and 100.0%; 95.7% and 99.1%, and 95.1% and 94.0%, respectively.

To detect neutralizing antibodies, the assay uses the cellular receptor angiotensin conversion enzyme-2 (ACE-2). Results can be expressed for neutralizing antibodies in percentage of inhibition, which is transformed in IU/mL by the ViraChip^® Software after the results are read by the ViraChip^® Scanner. The NTChip^® test can also assess a previous contact with the virus using Nucleocapsid (N) printed in spots triplets on the nitrocellulose; IgG antibodies from the sera are detected using a conjugated anti-IgG.

The result for captured IgG can be expressed in arbitrary units (AU) and in binding arbitrary units (BAU). For quantitative results, the NTChip^® was calibrated against the WHO international standard NIBSC 21/338 (20). The quantitative cutoffs (NIBSC 21/338) for Wuhan, Omicron (BA.1) and Nucleocapsid were: 9.1 IU/mL, 626.5 IU/mL and 16.4 BAU/mL, respectively, which can be translated into the following percentages of inhibition: 9%, 29% and 74 AU, respectively. As only RBD-Wuhan, RBD-BA.1 and Nucleocapsid were validated against PRNT₅₀, there are no cutoffs for RBD-Delta, RBD-BA.2 and RBD-BA.5.

To evaluate the specific T-cell immune response to SARS-CoV-2, the commercial T-SPOT^®COVID kit (Oxford Immunotec, Oxford, England) was used. This standardized method detects CD4⁺ T and CD8⁺ T cells that secrete interferon-gamma (IFN-γ) in response to stimulation with antigens via two specific peptide pools from SARS-CoV-2, one from the Spike protein and the other from the Nucleocapsid. Each assay was read using the AID EliSpot fluorescence microplate reader (Autoimmun Diagnostika GMBH, Germany) and AID EliSpot 7.0 software. The test result is considered positive if either Spike and/or Nucleocapsid have a count of 8 spots or more per 250,000 PBMC ( Supplementary Figure S3 ). The test result is considered negative if both have a count of 4 spots or less. Results of 5, 6, or 7 spots are considered indeterminate according to the manufacturer. A reactive result indicates the presence of SARS-CoV-2-sensitized effector T cells in the sample. A non-reactive result indicates that SARS-CoV-2-sensitized effector T cells were not detected in the sample.

Associations between two categorical variables were assessed using the Chi-square test or Fisher’s Exact test. Comparisons of means between two groups were performed using Student’s t-test. Normality assumption was checked using Kolmogorov-Smirnov test. In case of violation of this assumption, the nonparametric Mann−Whitney test was used alternatively.

Geometric means (GMT) and 95% confidence intervals (CIs) are presented for antibody assessment. Comparisons of variable levels by group and time were conducted via linear models with random effects (21) on log-transformed variables. This model assessed the effects of three components: time, group, and interaction between group and time. The presence of interaction between group and time indicates that group means evolve differently over time.

For qualitative cellular responses, logistic regression models with random effects were used. The linear model with random effects assumes data normality. However, deviation from normality does not bias estimates (22).

A significance level of 5% was used for all the statistical tests. Analyses were performed using SPSS 20.0 and STATA 17.

All the graphs were generated via GraphPad Prism 10.4.3.

A total of 55 patients with IEI and 60 controls were included in the study. The two groups were similar with respect to sex and age (p >0.05). The median age of patients with IEI was 27.3 years (ranging from 13.3 to 61.9 years) and that of the controls was 25.0 years (13.5 to 71.2 years) (p = 0.991). Patients with IEI patients had a median BMI of 21.91 kg/m² (15.15 to 36.51 kg/m²), which was lower than that of the controls (p = 0.026). The groups were comparable regarding the following comorbidities: systemic arterial hypertension, diabetes mellitus, heart disease, and obesity. Among the patients with IEI, 40% (22/55) had pulmonary diseases (asthma, bronchiectasis, pulmonary fibrosis, and lobectomy) and 10.9% (6/55) had other diseases, such as nephropathy, hepatopathy, neuropathy, and myasthenia gravis. Due to the exclusion criteria for the control group, no individuals in this group had diabetes mellitus, heart disease, pulmonary disease, or other comorbidities ( Table 1 ).

The median numbers of CD3⁺ T lymphocytes (p = 0.045), CD4⁺ T lymphocytes (p < 0.001), and B lymphocytes (p = 0.005) were lower in the IEI group than in the control group. No differences were observed for CD8⁺ T lymphocytes or NK cells (p = 0.996 and p = 0.195, respectively) ( Table 1 ).

All controls (n = 60) and 42 patients with IEI completed all 5 study visits ( Figure 1 ). Among the 13 individuals who did not complete the visits, 6 did not want to receive the 4th vaccine dose, and 6 did not want to receive the 5th vaccine dose. One patient with X-linked Agammaglobulinemia had a history of pyoderma gangrenosum in the left lower limb that worsened during the study period; owing to frequent and long periods of hospitalization, he could not receive the 5th vaccine dose; this event was not considered to be caused by the vaccination.

The distribution of patients with IEI according to the first two COVID-19 vaccine doses administered is depicted in Table 2 . Immunoglobulin replacement therapy (IGRT) was used by 48/55 (87.3%) patients with IEI. No differences were noted between the IEI and control groups with respect to the distribution of the types of the first two vaccine doses (p = 0.245). For the first and second doses, 47.3% (26/55) of the patients received CoronaVac, 30.9% (17/55) received BNT162b2 (Pfizer-BioNTech), and 21.8% (12/55) received ChAdOx1 nCoV-19 (AstraZeneca). The controls had an equal distribution of 33.3% for each vaccine ( Figure 1 ).

In both groups, all participants received original strain of BNT162b2 (Pfizer-BioNTech) as the third vaccine dose, according to the study protocol. For the fourth dose, following vaccination campaign priorities, patients with IEI predominantly received original strain of BNT162b2 (93.9%; 46/49). One patient received ChAdOx1 nCoV-19 (AstraZeneca), one received Ad26.COV2.S (Janssen), and one received bivalent mRNA (Original/Omicron BA.1, Pfizer-BioNTech) vaccine. Among controls, the fourth dose was predominantly with ChAdOx1 nCoV-19 (AstraZeneca) (49.2%; 29/59), followed by Ad26.COV2.S (Janssen) (25.4%; 15/59), CoronaVac, and original strain of BNT162b2 (11.9% each; 7/59), with one person receiving bivalent mRNA (Original/Omicron BA.1, Pfizer-BioNTech) (1.7%) ( Figure 1 ). One control did not receive the fourth dose as he was under 18 years of age, and this vaccine was not yet available for that age group during the study period. Since patients with IEI were prioritized in the vaccination campaign, the mean interval between the third and fourth doses was smaller among patients with IEI compared to controls (150.2 days vs. 234.8 days; p < 0.001, Mann-Whitney test). This also occurred between the second and third doses: 95.5 days in patients with IEI and 174.3 days in controls (p < 0.001, Student’s t test).

Patients with IEI predominantly received the fifth dose (second booster) with original strain of BNT162b2 (54.8%; 23/42), followed by bivalent mRNA (Original/Omicron BA.1, Pfizer-BioNTech) (31%; 13/42). Three patients received CoronaVac (7.1%), two received ChAdOx1 nCoV-19 (AstraZeneca) (4.8%), and one received Ad26.COV2.S (Janssen) (2.4%) ( Figure 1 ). As the bivalent mRNA (Original/Omicron BA.1, Pfizer-BioNTech) vaccine was not yet available at the start of the second booster campaign for patients with IEI, some individuals did not receive it. The mean interval between the fourth and fifth doses was 247 days, ranging from 123 to 514 days.

We assessed neutralizing antibodies to RBD-BA.1, RBD-BA.2, and RBD-BA.5 recognizing that there is some cross-reactivity and some antibodies related to natural infection (23, 24) ( Supplementary Figure S1 ). We noted lower levels of neutralizing antibodies at first assessment for most epitopes tested. Changes in the GMTs of antibodies over time were evaluated via a linear regression model with random effects ( Figure 2 ; see Supplementary Table S1 for seropositivity data and Supplementary Table S2 in the Supplementary Material for detailed data). Compared with controls, lower antibody levels for all variables analyzed were observed in patients with IEI (p < 0.05).

The rate of change of antibodies to RBD-Wuhan (p = 0.201) and RBD-Delta (p = 0.343) was similar over time when patients with IEI and controls were compared. Close to significant differences were also observed for antibodies to RBD-BA.1 (p = 0.078) and RBD-BA.5 (p = 0.056) variants. Over time, the two cohorts also showed a significantly distinct response to RBD-BA.2 neutralizing antibodies (p = 0.008). IgG-Nucleocapsid antibodies clearly differed over time with less increase over time apparent in patients with IEI (p = 0.010).

For RBD-BA.2, controls had lower GMT antibodies 6 months after the 1st booster (4266 IU/mL; 95% CI, 3124-5827), but antibodies increased 1 month after the 2nd booster (8402 IU/mL; 95% CI, 6513-10838), as did levels 1 month after the 1st booster (6603 IU/mL; 95% CI, 5159-8451). In the IEI cohort, the GMT was greater only 1 month after the 2nd booster (3002 IU/mL; 95% CI 1690-5329) than it was 1 month after the 1st booster (1975 IU/mL; 95% CI 1217-3203).

For Nucleocapsid, controls had higher GMT antibodies after the 2nd booster (87.8 BAU/mL; 95% CI, 68.1-113.1; p < 0.001) than 1 month and 6 months after the 1st booster, with similar GMTs between these time points [30.4 BAU/mL (95% CI, 23.1-39.9) and 49.5 BAU/mL (95% CI, 35.0-69.8), respectively]. In contrast, patients with IEI had similar GMTs across all evaluated time points [30.7 (95% CI, 22.6-41.7), 32.4 (95% CI, 21.8-48.3) and 42.2 BAU/mL (95% CI, 30.3-58.9)].

The RBD-BA.1, RBD-BA.2, RBD-BA.5 and Nucleocapsid antibodies were significantly increased 1 month after the second booster. For Nucleocapsid, IEI and control cohorts had similar GMTs 1 month after the 1st booster (30.7 vs 30.4 BAU/mL); however, 6 months after the 1st booster and 1 month after the 2nd booster, the IEI cohort had significantly lower GMTs than the controls did (32.4 vs 49.5 BAU/mL and 42.2 vs 87.8 BAU/mL, respectively). No time effect was observed for RBD-Wuhan (p = 0.173) or RBD-Delta antibodies (p = 0.376).

No significant differences were observed in the humoral response between IEIs and controls who received CoronaVac, ChAdOx1 nCoV-19 (AstraZeneca), or BNT162b2 (Pfizer-BioNTech) as their first two doses of COVID-19 vaccines ( Supplementary Table S3 in the Supplementary Material ).

T cell and humoral responses may be discordant and T cell responses are thought to offer some level of protection in COVID-19. We therefore assessed T cell responses in patients and controls. A similar evolution of cellular response to Nucleocapsid (p = 0.763) and Spike (p = 0.695) was observed over time for IEI and control cohorts ( Figure 3 , Supplementary Table S4 in the Supplementary Material ). The IEI cohort presented a greater T-cell response to Spike in terms of the mean number of spots per million (SFC/10⁶) PBMCs than did the control cohort (p = 0.002), but a similar Nucleocapsid response (p = 0.180).

Within group analysis of change related to time revealed that the 1 and 6 months post-1st booster responses were lower than those at one month post-2nd booster (p = 0.017) for Nucleocapsid. This pattern was not observed for Spike (p = 0.445).

The vaccines received in the first two doses (CoronaVac, ChAdOx1 nCoV-19 or BNT162b2) did not affect the mean number of spots for Nucleocapsid (p = 0.554) and Spike (p = 0.554) ( Supplementary Table S5 in the Supplementary Material ).

The goal of vaccination is to reduce mortality and hospitalizations. We therefore assessed the severity of infection among vaccinated individuals. Among the 115 participants, 71 were found to have COVID-19 before and during the study, as assessed by SARS-CoV-2 RT−PCR or antigen test detection.

Before immunization, 9/55 (16.3%) of patients with IEI had confirmed COVID-19: 6/55 (10.9%) did not need hospitalization and 3/55 (5.4%) required hospitalization. Between the 1st and 3rd vaccine dose, 3/55 (5.4%) had confirmed COVID-19, of which 2/55 (3.6%) did not need hospitalization and 1/55 (1.8%) had to be hospitalized. Between the 3rd vaccine dose and the 4th dose, during the Omicron circulation, 21/55 (38.2%) developed SARS-CoV-2 infection and only 1/55 (1.8%) was admitted to hospital. After the 4th dose and up to 16 months, 7/49 (14.3%) had confirmed SARS-CoV-2 infection, but none had to be hospitalized ( Figure 4 ).

In the control group, 3/60 (5.0%) individuals developed SARS-CoV-2 infection before immunization and 2/60 (3.3%) between the 1st and 3rd vaccine dose; 23/60 (38.3%) were infected between the 3rd dose and the 4th dose, and 8/60 (13.3%) were infected after that up to 16 months of follow-up. None of them needed hospitalization ( Figure 4 ).

The incidence of COVID-19 diagnosis was greater after the 3rd dose, when the Omicron variant was circulating in Brazil ( Supplementary Figure S1 in the Supplementary Material ). However, although 77.1% (27/35) of the patients with IEI and 86.1% (31/36) of the controls were diagnosed with COVID-19 after the 3rd vaccine dose, all cases were mild, except for one patient with IEI who still required hospitalization but did not need oxygen therapy or ICU admission. Thus, vaccination achieved its overall clinical goal in our IEI cohort.

As there were only 5 patients who were hospitalized in this cohort, with 2 occurring after the start of the study, it is not possible to analyze specific details of the humoral or cellular responses that might have led to a higher probability of hospitalization compared to other patients. However, a qualitative analysis was performed.

After the start of the study, one 38-year-old female patient with CVID, who had a confirmed SARS-CoV-2 infection 12 days after her first dose and very likely without antibodies in her IVIg (June 2021) required oxygen therapy and ICU admission. She had a mild reinfection 8 months later. At study admission (one month after the 3rd vaccine dose, CoronaVac/CoronaVac/BNT162b2 scheme), her immunological status was CD4⁺ T = 664 cells/mm³, CD8⁺ T = 849 cells/mm³, CD19⁺ = 213 cells/mm³, with neutralizing antibodies detected against all SARS-CoV-2 variants (e.g., NAb anti-RBD-BA.5 = 20,925.0 IU/mL) and a robust cellular response (Nucleocapsid T-cell response = 53 SFC/10⁶ PBMC, Spike T-cell response = 367 SFC/10⁶ PBMC).

The other hospitalized patient was a 29-year-old male with AT, who tested positive for SARS-CoV-2 RT-PCR in June 2022, 153 days after receiving his 3rd vaccine dose. Differently from the other patient, he did not require oxygen therapy or ICU admission. At admission to study (one month after the 3rd vaccine dose, ChAdOx1 nCoV-19/ChAdOx1 nCoV-19/BNT162b2 scheme, and 4 months before COVID-19 hospitalization), he had CD4⁺ T = 202 cells/mm³, CD8⁺ T = 79 cells/mm³, CD19⁺ = 6 cells/mm³. He developed a robust cellular response (Spike T-cell response = 163 SFC/10⁶ PBMC), despite lymphopenia and was receiving IVIg, likely containing antibodies (as he had low B cell numbers, albeit lower antibody levels compared to other patients, e.g., NAb anti-RBD-BA.5 = 79.7 IU/mL). He was infected 4 months after vaccination during Omicron circulation (June 2022). Hospitalization was mainly for medical observation, without need for oxygen therapy or ICU admission.

When cellular and humoral immune responses were compared between patients with IEI and controls at three different time points ( Figure 5 ), control individuals displayed a more homogeneous response for both antibodies and the cellular response. In contrast, patients with IEI had a much more variable humoral immune response, and some did not even reach the cutoff level for the antibody response.

The results presented in Figure 5 reveal two distinct subgroups of patients with IEI regarding their humoral response at the time points evaluated. At one month post-1st booster, patients with negative antibodies (below the threshold of 9.1 IU/mL) included one case of Hyper-IgM, two of CVID, one of XLA and one of Combined Immunodeficiency. Among those with a borderline response (antibody levels at threshold of 9.1 IU/mL), there were three patients with CVID, one with XLA, one with Hyper-IgM and one with Combined Immunodeficiency.

At one month post-2nd booster, one patient with CVID had negative antibodies (the same who had 24.4 IU/mL after the 1st booster). Among those with a borderline response (antibody levels at threshold of 9.1 IU/mL), there were four patients with CVID and two with XLA. Except for one XLA patient, all the other five patients were previously described at one month post-1st booster.

These patterns suggest heterogeneity in the humoral response, even among patients receiving similar vaccine regimens, with a potential impact of individual characteristics, such as the type and severity of IEI, on the ability to generate neutralizing antibodies and the presence of antibodies in the commercial immunoglobulin preparations administered to these patients.