This longitudinal study was carried out on 47 adults with CVIDs diagnosed according to the ESID criteria (37) and regularly followed by the Care Center for adults with IEIs in Rome, Italy. As healthy controls, 22 Health Care Workers (HCWs) were enrolled at the Bambino Gesù Children’s Hospital.

HCWs and patients were considered eligible for the study if they were naïve to SARS-CoV-2 infection at the time of enrollment, had completed the primary schedule with the mRNA BNT162b2 vaccine, and agreed to receive the 3^rd dose.

In patients with CVID, the first booster (3^rd dose) dose was performed in September 2021, six months after completing the full primary immunization schedule. In HCWs, the first booster dose was performed nine months after the last dose of the vaccine. At the time of the booster in Italy, only the monovalent vaccine was available and approved for use.

As prescribed for patients with IEIs by the Italian Agency of Drugs (AIFA), the second booster dose (4^th dose) of the BNT162b2 vaccine was offered in March 2022 only to those participants who remained free from SARS-CoV-2 infection since the last vaccine administration. Patients were tested for SARS-CoV-2 infection by RT-PCR on the nasopharyngeal swab (NPS) every time they attended a hospital site, in case of positive household contacts irrespective of symptoms, and upon onset of symptoms possibly related to COVID-19. The duration of the viral shedding was evaluated by recording the dates of the first positive and first negative NPS. In patients with SARS-CoV-2 infection, we recorded COVID-19 severity [scored according to WHO stage (38)], hospitalization, vaccination status, and SARS-CoV-2 treatments).

Blood samples were obtained 10 days after the 3^rd dose (post 3^rd), six months apart on the day of the 4^th dose (pre 4^th), and 10 days after the 4^th dose (post 4^th). For those who were infected after the 3^rd dose, blood samples were collected one month after recovery and separately analyzed. For those who refused the 4^th dose of vaccine, blood samples were collected six months after the booster immunization and then included in the analysis as pre 4^th values ( Figure 1 ). During the study, participants were allowed to continue their treatments, including IgRT as standard therapy for the underlying antibody deficiency. At the study time, all IgRT brands in use did not contain anti-S1 IgG antibodies (39, 40). For each participant, clinical data were collected, including age, immunoglobulin levels at IEIs diagnosis, and the number of IEI-related disorders. We defined CVIDs with immune-dysregulation phenotype as having at least one of the following: systemic autoimmunity or autoimmune cytopenia, enteropathy, lymphoproliferative disorders, and lymphoid malignancy or solid cancer. Chronic Lung Disease (CLD) was defined as having at least one of the following: Chronic Obstructive Pulmonary Disease diagnosed according to GOLD guidelines (41), Interstitial Lung Disease or Bronchiectasis (based on CT scan imaging), and Granulomatous and Lymphocytic Interstitial Lung Diseases (GLILD) (based on CT scan imaging and histologically confirmed) (42). Immunological laboratory data collected at baseline included complete blood count (CBC) and immunoglobulin serum levels.

Following the actual position of IUIS (1), eight patients displaying complex clinical phenotypes (mainly immune dysregulation) underwent molecular testing to rule out monogenic forms of humoral immunodeficiencies. A custom-targeted Next Generation Sequencing (NGS) panel of 53 genes whose variants are associated with CVIDs, and humoral immunodeficiency was performed ( Supplementary Table 1 ). As targeted genetic testing did not identify known or potential molecular bases for monogenic immunodeficiency in the selected patients, all cases in the present cohort were retained.

The study was approved by the Ethical Committee of the Sapienza University of Rome (Prot. 0521/2020, July 13, 2020) and performed in accordance with the Good Clinical Practice guidelines, the International Conference on Harmonization guidelines, and the most recent version of the Declaration of Helsinki.

Humoral response to vaccination was assessed quantifying the anti-S/Receptor Binding Domain (RBD)-IgG by a commercial chemiluminescence microparticle antibody assay (ARCHITECT SARS-CoV-2 IgG II Quantitative, Abbott Laboratories, Wiesbaden, Germany) according to the manufacturer’s protocol (Architect^® i2000sr Abbott Diagnostics, Chicago, IL). As reported in the leaflet, anti-RBD-IgG were expressed as binding arbitrary units (BAU)/mL and values were considered positive when ≥7.1. To evaluate the neutralizing activity of vaccine induced antibodies, microneutralization assay (MNA) was performed as previously described, using both SARS-CoV-2 Wuhan (Lineage A, GISAID Accession Number: EPI_ISL_410545) and Omicron (Lineage BA.5, GISAID Accession Number: EPI_ISL_16848123) as challenging viruses (43). Briefly, heat-inactivated sera serially titrated in 8 two-fold dilutions in duplicate (starting dilution 1:10). Equal volumes of serum and medium containing 100 TCID₅₀ SARS-CoV-2 Lineage A or Lineage BA.5 were mixed and incubated at 37°C for 30 min. Serum-Virus mixtures were then added to sub-confluent Vero E6 cell monolayers and incubated at 37°C and 5% CO₂. After 72 hours, microplates were observed by light microscope for the presence of cytopathic effect (CPE). To standardize inter-assay procedures, positive control samples showing high (1:160) and low (1:40) neutralizing activity were included in each assay session. The highest serum dilution inhibiting at least 90% of the CPE was indicated as the neutralization titer and was expressed as the reciprocal of serum dilution (MNA90).

Peripheral blood mononuclear cells (PBMCs) were isolated by Ficoll Paque™ Plus 206 (Amersham PharmaciaBiotech) density-gradient centrifugation and immediately frozen and stored in liquid nitrogen until use. The freezing medium contained 90% Fetal Bovine Serum (FBS) and 10% DMSO.

For the detection of SARS-CoV-2 specific B cells, a biotinylated protein spike was individually multimerized with fluorescently labeled streptavidin as previously described (17, 25, 44). Separate aliquots of recombinant biotinylated Spike or RBD were mixed with streptavidin BUV395, streptavidin PE or streptavidin FITC (BD Bioscience) at 25:1 ratio 20:1 ratio and 2.5:1 ratio respectively. We used two different streptavidin-conjugated fluorochromes, one with a very high brightness index (PE) and the other with a moderate brightness index (BUV395), to be able to distinguish low-affinity MBCs (only visible with a very bright fluorochrome such as PE) from high-affinity MBCs (double positive for PE and BUV395). Streptavidin PE-Cy7 (BD Bioscience) was used as a decoy probe to gate out SARS-CoV-2 antigen non-specific streptavidin-binding B cells. The antigen probes individually prepared as above were then mixed in Brilliant Buffer (BD Bioscience). 4x10⁶ previously frozen PBMC samples were prepared and stained with the antigen probe cocktail containing 100 ng of spike per probe (total 200 ng), 27.5 ng of RBD, and 2 ng of streptavidin PE-Cy7. After one step of washing, surface staining with antibodies was performed in Brilliant Buffer at 4°C for 30 min.

B cell subsets were identified based on the expression of CD19 BV786, CD27 BV510, CD24 BV711, and CD38 BV421 markers by flow-cytometry purchased from BD Bioscience. Naïve B cells were identified as CD19+CD24+CD27-, transitional B cells were gated as CD19+CD24++CD38++, and atypical MBCs as CD19+CD24-CD27-CD38-. MBCs were defined as CD19+CD24+CD27+, and anti-IgM APC (Jackson ImmunoResearch) will be used to discriminate IgM+ form switched (IgM-) MBCs. Spike-specific MBCs were defined as low-affinity (positive for PE, S+) or high-affinity (double positive for PE and BUV395, S++) ( Supplementary Figure 1A ).

Among S+ and S++ MBCs, IgM and RBD expressions were evaluated ( Supplementary Figure 1B ).

Stained PBMC samples were acquired on FACs LSRFortessa (BD Bioscience). At least 2x10⁶ cells were acquired and analyzed using Flow-Jo10.8.1 (BD Bioscience). Phenotype analysis of antigen-specific B cells was only performed in subjects with at least 10 cells detected in the respective antigen-specific gate. Blank was determined in unexposed donors, before vaccination. LOB (limit of blank) was set as the mean of the blank + 1.645xSD. LOD (limit of detection) as the mean of the blank + 2xSD.

PBMCs isolated from CVIDs patients were cultured in 96-well cell plates at 1x10⁶ cells/well concentration in RPMI 1640 culture medium containing 5% AB serum. PBMCs were cultured at 37°C, 5% CO₂, with CytoStim^® (from the SARS-CoV-2 Prot_S PBMC Kit, Miltenyi, Biotech) for a total of six hours. After two hours of incubation, Brefeldin A (SARS-CoV-2 Prot_S PBMC Kit, Miltenyi, Biotech) was added to the cells to inhibit the transport of proteins to the cellular membrane. Cells were fixed, permeabilized, and stained with CD3 APC, CD4 Vio B515, CD8 Vio Green, CD154 (CD40L) APC Vio 770, IFNγ PE, and TNFα PE-Vio770 (SARS-CoV-2 Prot_S PBMC Kit protocol, Miltenyi Biotech) with the addition of CD45RO BUV395(BD Biosciences).

T cells were gated as CD3+ and divided into CD3+CD4+ and CD3+CD8+ T cells. Naïve T cells were gated as CD3+CD4+CD45RO- (or CD8+) and memory T cells were identified as CD3+CD4+CD45RO+ (or CD8+). The CD40L, IFNγ, and TNFα were evaluated on CD4 and CD8 memory T cells. Cells were acquired on a BD FACSymphony A5™ and data were analyzed with FlowJo ver. 10.8.1 (BD Bioscience).

FCS files were analyzed using the FlowJo ™ v10.8.1 software (BD, Biosciences). Prior to analysis, all samples were normalized to equal cell numbers. For each of the four groups of patients, 5 randomly selected samples were concatenated in a single FCS file. Equal numbers of live CD19+ B cells or CD3+CD4+ T cells or CD3+CD8+ T cells were analyzed. Data were partitioned into clusters in high dimensional space by using the X-Shift plugin algorithm (45), and visualized after performing dimensionality reduction using the Optimized t-distributed stochastic neighbor embedding (opt-SNE). Marker expression and frequency across different clusters were then assessed using the heatmaps generated by the Cluster explorer plugin (FlowJo ™ v10.8.1 software, BD, Biosciences).

Sequencing experiments were performed on the Ion Torrent S5 platform (Thermo Fisher Scientific). Sequence data analysis was performed with the Ion Reporter v5.18 software (Thermo Fisher Scientific). Variant validation was performed by Sanger sequencing on the 3730 Genetic Analyzer platform (Thermo Fisher Scientific). Variants with a minor allele frequency >0.01 were filtered out. Only nonsynonymous variants and splice site variants were retained. A visual analysis of the reads including candidate variants mapped against the reference genome (version GRCh37/hg19) was performed on the Integrative Genome Viewer software (IGV, https://software.broadinstitute.org/software/igv/). Variants were classified according to the American College Of Medical Genetics And Genomics (ACMG) guidelines (46). Data are available at http://www.ncbi.nlm.nih.gov/bioproject/967329.

Data were analyzed by GraphPad Prism 9 version 9.3.1 and Statistical Package for Social Sciences version 15 (SPSS Inc). The data were first tested for normality and homoscedasticity using Shapiro-Wilk and Levene’s tests and since the assumptions were violated, non-parametric tests were used for the analysis.

Demographics were summarized with descriptive statistics (median and IQR for continuous values). Immunological and clinical variables were compared between the different study times. A univariate analysis assessed the impact of variables of interest. The tests of normalcy variables did not follow normal distribution according to the results of the Kolmogorov–Smirnov and Shapiro–Wilks tests. As a result, a nonparametric approach was used to analyse the data, the Wilcoxon matched pair signed-rank test or the two-tailed Mann–Whitney U-test were used. Categorical variables were compared by Chi-square test. Differences were deemed significant when P < 0.05).

Among 179 patients with CVIDs, 50 were randomly selected and agreed to participate to the study. Three patients dropped-out and 47 CVIDs patients [median age 53.5 years (IQR 45.7-66), females 28 (58.3%)], immunized with the 3^rd dose of the mRNA BNT162b2 vaccine, were included in the analysis. Six months apart, the 4^th dose was administered to 25/47 patients, because 12 participants were infected by SARS-CoV-2 and 10 patients refused the additional booster doses ( Figure 1 ).

After the 3^rd dose, all HCW controls (median age 54.8 years, IQR 43.8-61.9) had measurable levels of serum anti-S1 IgG Abs. In contrast, specific IgG were above the cut-off value in 30/47 (63.8%) CVIDs patients (median 617.6 BAU/ml, IQR 99.44-967.8) ( Figure 2A ). We separately show the antibody levels in CVIDs patients that exceeded (Responders, R) or not (Non-Responders, NR) the cut-off value (7.1 BAU/ml). CVIDs patients able to seroconvert had a highly variable response with a significantly lower antibody concentration than control HCWs ( Figure 2A ). The data suggest that, even when we exclude Non-Responders CVIDs patients from the analysis, our cohort includes individuals with different capacities to mount an antibody response.

High specificity and affinity are the most important characteristics of MBCs, generated by the adaptive immune system in response to infection or vaccination (25, 44). In control HCWs, low-affinity (S+) MBCs can be detected before vaccination, whereas high-affinity (S++) MBCs are produced only after the 2^nd vaccine dose (44). S+ MBCs are mostly of IgM isotype and S++ are switched MBCs ( Supplementary Figure 1B ). The RBD specificity is absent among S+ MBCs. In contrast, 20-40% of the S++ MBCs bind RBD and are of switched isotype ( Supplementary Figure 1B ).

We measured S+ ( Figure 2B ) and S++ MBCs ( Figure 2C ) ten days after the 3^rd dose in patients who were defined as Responders or Non-Responders based on the level of specific Abs. Overall, as compared to control HCWs, CVID patients had significantly lower frequencies of S+ and S++ MBCs. S+ MBCs were detectable in 45% (21/47) of the CVIDs patients ( Figure 2B ) and S++ MBCs in 30% (14/47) of the cases ( Figure 2C ).

Thus, depending on the biological read-out of efficacy used, a different percentage of patients can be classified as Responders or Non-Responders (63.8%, based on antibody levels, vs 45%, based on S+ MBCs vs 30%, based on S++ MBCs) ( Figure 2D ).

Among CVIDs patients who could be considered Responders based on antibody levels, we identified three different types of immune responses ( Figure 3A ; Table E2 ), which were used to define functional groups of patients. Fourteen patients (group 1) generated both S+ and S++ MBCs, 7 patients (group 2) generated only S+ MBCs and 9 patients (group 3) had no spike-specific MBCs. The 17 Non-Responder patients, lacking both Abs and spike-specific MBCs, were assigned to group 4 ( Figure 3A ).

High-affinity MBCs are generated in the germinal centers, with the help of T cells, thanks to the combined mechanisms of somatic hypermutation and antigen selection (47). Only a minority of CVIDs patients have high-affinity S++ MBCs suggesting that only patients of this group can perform all the functions required to complete the adaptive immune response.

Although group 1 patients generated an immune response composed of S+, S++, and RBD+ MBCs ( Figures 3B–D ), all frequencies were significantly lower compared to HCWs ( Figure 3E ).

In group 2, only S+ IgM MBCs were detectable ( Figure 3B ), whereas patients of groups 3 and 4 failed to generate any spike-specific MBCs.

Serum anti-S1 IgG levels differed in the four groups, being significantly higher in group 1 than in groups 2 and 3 (p=0.0016 and p=0.0012, respectively) ( Figure 3F ). In group 1, anti-S levels were significantly lower than in HCWs and had a reduced neutralizing activity against wild-type SARS-CoV-2. Neutralizing Abs titers were significantly lower in group 2 and 3 and absent in group 4 ( Figure 3F ).

Six months after the 3^rd dose, the 4^th dose was administered to 25 patients. Twelve patients did not receive the second booster because they were infected by SARS-CoV-2 after the 3^rd dose ( Figure 1 ) and were not included in the longitudinal study.

After the 4^th vaccine dose, in group 1, the frequency of S+, S++ RBD-specific MBCs significantly increased (p=0.0386, p=0.0122, p=0.0245, respectively) ( Figures 4A–C ). In group 2, the frequency of S+ MBCs did not change, but few S++ and RBD-specific MBCs were generated in four patients ( Figures 4B–D ). In one patient of group 3, S+ MBCs became detectable after the 4^th dose ( Figure 4A ).

In the pre-4^th dose serum sample, anti-S1 IgG levels were still detectable in all patients who had produced Abs before (groups 1, 2, and 3) ( Figure 4D ). In response to the 4^th dose, anti-S1 IgG increased in groups 1 and 2 (p=0.0500), but not in groups 3 and 4 (median 5.35 BAU/ml, IQR 2.67-10.7) ( Figure 4D ). Thus, only patients who generated spike-specific MBCs were able to secrete Abs in response to the booster.

Virus neutralization after the 4^th dose confirmed that patients of group 1 had a more effective response than patients of groups 2, 3, and 4, resulting in the generation of Abs able to prevent viral infection in vitro (p=0.0152, p=0.0095, and p=0.0006 respectively) ( Figure 4E ). In immunocompetent subjects, administration of the 3^rd and 4^th dose induces the production of Abs able to also neutralize the omicron variants, independently of previous exposure to the novel strains (48). Broadening of the immune response did not occur in CVIDs patients as they were all unable to produce Abs with neutralizing activity against the omicron variant BA.5 ( Figure 4E ).

We compared the isotype of MBCs generated in response to vaccination in HCWs and CVIDs patients of groups 1 and 2. In HCWs, S+ MBCs were half of IgM and half of switched isotypes, whereas S++ MBCs were almost all switched ( Figures 5A, B ).

Class switching appeared to be less effective in CVIDs patients since both in groups 1 and 2 S+ MBCs were mostly of IgM isotype ( Figure 5A ). S++ MBCs, detectable only in patients of group 1 at all-time points, were in the majority switched MBCs but remained significantly lower than in the controls ( Figure 5B ).

We asked the question of whether the four groups of CVIDs patients, defined by the response to SARS-CoV-2 vaccination, correspond to distinct B-cell phenotypes. On samples obtained after the 3^rd dose, we performed an unbiased analysis of CD19+ B cells using optimized t-stochastic neighbor embedding (Opt-SNE) to reduce dimensionality, followed by graph-based clustering analysis (X-shift). We identified 14 B-cell clusters in the four concatenated groups ( Figure 6A ). The B cell clusters were analyzed by marker expression to identify the different B cell populations resulting in the definition of seven major clusters ( Figure 6B ). The relative X-Shift cluster sets were calculated for each of the four groups ( Figure 6C ). The main differences among the groups were caused by the distribution of MBCs.

We found that switched MBCs were present only in group 1, whereas IgM MBCs were identified in groups 1 and 2 ( Figures 6C, D ). Atypical MBCs which have been shown to be increased in autoimmune disease, chronic viral infection, and in a group of CVIDs patients (49, 50), were detected at low frequencies in all groups. Plasmablasts were found only in group 1 and transitional B cells were increased in group 4 ( Figures 6C, D ).

Based on the populations identified by cluster analysis, we calculated the frequency and absolute numbers of the identified populations in all patients of the four groups. Frequency and absolute numbers of MBCs and IgM MBCs were within the normal range in groups 1 and 2 but switched MBCs were significantly reduced in group 2 compared to group 1 (frequency: p=0.0159; absolute numbers: p=0.0297, Figures 6E, F ). Patients of group 3 had significantly fewer MBCs, both of IgM and switched isotypes, compared to groups 1 and 2 ( Figures 6E, F ). Patients of group 4 were lymphopenic and CD19+ B cells were significantly reduced ( Supplementary Figures 2A, B ). Despite the reduction of total B cells, transitional B cells, the most immature cell type found in the peripheral blood, were present at normal frequency and numbers indicating a normal output from the bone marrow. In contrast, mature-naive B cells and MBCs were strongly reduced, both in frequency and absolute numbers ( Supplementary Figures 2A, B , 6E, F ). In groups 1, 2 and 3 transitional and mature-naive B cells were within the normal values ( Supplementary Figures 2A, B ). The frequency and the absolute number of atypical MBCs were comparable in the four groups of patients ( Supplementary Figures 2A, B ).

In summary, group 1 patients had a B cell phenotype similar to that of control HCWs, and all B-cell populations were detected at normal frequencies. Patients of group 2 had normal numbers of IgM MBCs but were characterized by a strong reduction of switched MBCs. Patients in group 3 had no MBCs. Patients of group 4 had low total B-cell numbers with a significant reduction of mature-naive B cells and the absence of MBCs.

In line with the description of the B-cell phenotype, IgM serum levels measured at the study time were significantly reduced in group 4 compared with groups 1 and 2 (p=0.0001 and p=0.0238, respectively) and IgA levels were significantly reduced in groups 3 and 4 compared to group 1 (p=0.0052, 1 vs 3; p=0.0018, 1 vs 4) and group 2 (p=0.0428, 2 vs 3; p=0.0500, 2 vs 4) ( Figure 6G ).

T cells in groups 1 and 2 were present in normal frequency and absolute numbers, while total CD4+, CD4+ memory (CD45RO+) and naïve CD4+ (CD45RO-) were low in group 4 ( Supplementary Figure 3 ). One patient of group 3 (11%) and 6 of group 4 (35%) had a severe reduction in CD4+ (<200 cells/mm3) and/or naïve CD4+ naive (20 cells/mm3), thus fulfilling the criteria for the diagnosis of Late-Onset Combined Immunodeficiency (LOCID) (51). The frequency and the absolute number of total CD8+ and both memory and naïve CD8+ were comparable in all groups ( Supplementary Figure 3 ).

The adaptive response to vaccination requires the collaboration of B cells with CD4+ T cells in the GC. We have demonstrated before that switched MBCs can be only generated in individuals with functional GCs (33). IgM MBCs, instead, have been also detected in individuals with severe immunodeficiencies, lacking T cells or the costimulatory molecule CD40 Ligand (CD40L) (33). Because of their importance in the GC reaction and generation of immune memory, we analyzed the function of CD4+ T cells activated by polyclonal stimulation in samples obtained after the 3^rd dose. We evaluated their phenotype and ability to respond with upregulation of the CD40L and secretion of IFNγ and TNFα. The Opt-SNE map identified 7 clusters corresponding to naïve and memory CD4+ T cells, that expressed or not the CD40L and secreted TNFα and/or IFNγ ( Figures 7A, B ). The relative X-Shift cluster sets were calculated for each of the four groups of CVID patients ( Figure 7C ). As shown by the heatmap, the function of naïve and memory T cells was maintained in group 1, slightly altered in group 4, but severely impaired in groups 2 and 3 ( Figure 7D ). All the data obtained with unbiased analysis were confirmed by a bias analysis (see Supplementary Figure 4 ). CD40L was significantly diminished in all groups compared to group 1. In groups 2 and 3, also the mean fluorescence intensity (MFI) for the CD40L, which is proportional to the number of expressed molecules, was lower than in group 1. Among CD4+ T cells, those that express the CD40L after stimulation also secrete IFNγ and/or TNFα and were significantly reduced in patients of groups 2 and 3 ( Supplementary Figure 4 ).

We also performed an unsupervised analysis of CD8+ T cells phenotypes following polyclonal stimulation. The Opt-SNE map identified 7 clusters corresponding to naïve and memory CD8+ T cells, that secreted or not TNFα and/or IFNγ and expressed or not the CD40L ( Supplementary Figure 5A ). The relative set of X-Shift clusters for each group were calculated ( Supplementary Figure 5B ).

The heatmap shows that also on CD8+ cells the expression of the CD40L was severely impaired in groups 2 and 3, thus confirming the alteration of this pathway ( Supplementary Figure 5C ). Analyzing the data with a classical gating strategy we confirmed that CD8+ cells expressing CD40L were significantly diminished in groups 2, 3, and 4 compared to group 1. In groups 2 and 3, also the MFI was lower than in group 1 ( Supplementary Figure 5D ).

IFNγ and TNFα are not co-expressed with CD40L in CD8+ cells. The frequency of memory CD8+ cells able to secrete IFNγ was significantly upregulated in patients of groups 2, 3, and 4 compared to those of group 1 and the frequency of CD8+TNFα+ cells were significantly increased in patients of groups 2 and 4 compared to group 1 ( Supplementary Figure 5E ).

Differences between groups are mirrored by the clinical and immunological characteristics of participants ( Table 1 ; Figure 8A ). Groups 1 and 2 had a less severe clinical phenotype than groups 3 and 4, with a significantly lower frequency of patients with signs of immune dysregulation, lower prevalence of chronic lung disease ( Figure 8A ), lower cumulative number of IEI-related disorders, and absence of lymphopenia (<1000 cells/mm3) ( Table 1 ). These observations confirm that the presence of MBCs is associated with a less severe course of clinical disease (52). The inability of patients of group 2 to generate high-affinity MBCs, supports the notion that IgM MBCs have the function of systemic and mucosal first-line protection (34, 53), but do not represent the selected product of adaptive immunity.

For 10/47 patients who presented with complex phenotype, monogenic forms of CVIDs were investigated by targeted gene sequence analysis ( Table E1 ). No Pathogenic or Likely Pathogenic Variants were identified, and no molecular diagnosis was achieved. The NGS analysis identified one or more constitutional heterozygous variants classified as Variants of Uncertain Significance (VUS) in four patients ( Table E3 ).

After receiving the 3^rd dose of the vaccine, 12/47 (25%) patients were infected by SARS-CoV-2. Infected patients were previously classified as being part of groups 1, 3, or 4. One month after recovery, the frequency of S+, S++, and RBD+ MBCs increased in group 1. One patient of group 3 generated S+ MBCs of IgM isotype that were able to bind RBD ( Figure 8B ). No response was detectable in patients of group 4.

Nine out of 25 (36%) of CVIDs patients were infected by SARS-CoV-2 after receiving the 4^th dose of vaccine, and 6/10 (60%) were infected after refusing the 4^th dose (p=0.436).

The vast majority (24/27) of infected patients received a SARS-CoV-2 specific treatment (monoclonal Abs or antivirals). Twenty-six patients (96%) had mild COVID-19, whereas one patient from group 4 who refused to receive the 4^th dose died due to the progression of lymphoma ( Table 2 ).