The current study included blood from 23 patients (14 women, nine mem; age: median 50 years ± 13 years) who fulfilled the diagnostic criteria for CVID (www.ESID.org). All patients were classified as CVID smB-21lo according to the EUROclass classification (2) and suffered from additional immune dysregulation, as listed in Supplementary Table S1 . Control blood and tonsils were obtained from healthy donors (HD) without known immunodeficiency. All experiments were performed with ethical approval from local authorities (251/13 and 254/19) in accordance with the Declaration of Helsinki. Written informed consent was obtained from all patients, healthy individuals, or their parents. Tonsillectomies were performed for medical reasons.

The following antibodies and lectins were used: CD3 PE/Cy7 (UCHT1), CD4 BV421 (RPA-T4), sialyl-Lewis X (SLe^X) AF488 (FH6), CD19 APC/Cy7 (HIB19), CD21 PE/Cy7 (Bu32), CD27 BV421 (M-T271), CD38 PerCP/Cy5.5 (HIT2), IgD BV785 (IA6-2), IgM BV605 (MHM-88) were obtained from BioLegend (San Diego, CA, USA). CD3 BUV395 (UCHT1), CD27 BV605 (L128), IgG AF700 (G18-145), and streptavidin APC (70312) for the detection of biotinylated lectins were obtained from BD Biosciences (Franklin Lakes, NJ, USA). CD19 PE/Cy7 (J3-119), and CD21 PE (BL13) were obtained from Beckman Coulter (Brea, CA, USA). IgD PE (polyclonal), IgD FITC (polyclonal), and IgM Cy5 (polyclonal) were obtained from SouthernBiotech (Birmingham, AL, USA). Sambucus nigra agglutinin (SNA) FITC (ZE0328), Lens culinaris agglutinin (LCA) FITC (ZD0104), Aleuria aurantia lectin (AAL) FITC (ZE0628), and Erythrina cristagalli lectin (ECL) biotinylated (ZB042) were obtained from Vector Laboratories (Newark, CA, USA). Lectin-binding properties and symbol nomenclature for graphical representations of glycans (19) are shown in Supplementary Figure S1 . All glycans were created using the web software DrawGlycan (20).

PBMCs were isolated from EDTA blood by Ficoll density gradient centrifugation following standard protocols. For in vitro cultures, naïve B cells were isolated using the naïve B-Cell Isolation Kit II (Miltenyi Biotec, Bergisch Gladbach, Rhineland, Germany) following the manufacturer’s instructions. To determine the purity of naïve B cells, cells were stained with IgD FITC, CD21 PE, CD38 PerCpCy5.5, CD3 PC7, IgM Cy5, CD19 APC-Cy7, CD4 BV421, and CD27 BV605. Purity was above 93%. Tonsil cells were obtained by mechanical dissociation. After isolation, cells were resuspended in IMDM (Life Technologies, Carlsbad, CA, USA) with 10% foetal calf serum (FCS; Biochrom, Cambridge, Cambridgeshire, United Kingdom).

Harvested cells were washed two to three times in phosphate-buffered saline (PBS; Lonza Basel, Switzerland) 1% bovine serum albumin (BSA; Sigma-Aldrich St. Louis and Burlington, MA, USA). Cell surface lectin staining was performed separately for each lectin (SNA FITC, LCA FITC, AAL FITC ECL biotinylated) for 15 min on ice. Cells were washed with PBS containing 1% BSA. For ECL, cells were subsequently stained with streptavidin APC for 15 min on ice. Antibody staining was then performed for 15 min at 4°C. PBMCs and tonsil cells were stained with IgD PE, CD38 PerCpCy5.5, CD21 PC7, CD19 APC-Cy7, and CD27 BV421 to determine the respective B-cell subsets. For flow cytometric analysis of naïve activated B cells, cells were stained with IgD PE and CD19 PC7, and DAPI was added to exclude dead cells prior to data acquisition. Data were acquired using an LSRFortessa (BD Biosciences) and analysed using FlowJo software (Treestar, Ashland, OR, USA).

For B-cell activation, 50,000 isolated naïve B cells were stimulated with anti-IgM (5 µg/ml; SouthernBiotech, Birmingham, AL, USA), IFN-γ (50 ng/ml; BioLegend), CpG oligodeoxynucleotides (2.5 µg/ml, InvivoGen, San Diego, CA, USA), or cultivated without additional stimulus in IMDM supplemented with l-glutamine, l-glutathione, insulin, apo-transferrin (all Sigma-Aldrich), nonessential amino acids (Gibco, Waltham, MA, USA), and 10% FCS for 48 h as described before (5). Cells were harvested and processed for flow cytometry or RNAseq.

After stimulation, B cells were resuspended in RLT buffer (Qiagen, Shanghai, China) and frozen at − 80°C. RNA was isolated using the RNAeasy Micro Kit (Qiagen) according to the manufacturer’s instructions. Samples were processed and sequenced as described previously (5). Alignments and gene count tables were obtained for each sample using STAR (version 2.4.2a) on the GRCh38 genome and annotation (GRCh38.p7). Differential gene expression was analysed using the limma R package. An adjusted p-value (Benjamini and Hochberg) < 0.05 was considered significant. The original raw files are available in the Gene Expression Omnibus data repository of the National Center for Biotechnology Information (21) under the accession number GSE181739 for anti-IgM/IFN-γ in vitro-stimulated B cells or GSE148163 for naïve-like CD21^low B cells ex vivo.

Statistical analyses were performed using GraphPad Prism 10 (GraphPad Software Inc., San Diego, CA, USA). Normal distribution was determined using D’Agostino and Pearson or Shapiro–Wilk test. Data were analysed using repeated measures one-way ANOVA with Tukey’s multiple comparisons test for normally distributed results and Friedman’s test with Dunn’s multiple comparisons test for nonnormally distributed results, as indicated. p-values of less than 0.05 were considered significant: ^* p < 0.05; ^** p < 0.01; ^*** p < 0.001; and ^**** p < 0.0001. Error bars in all figures define the means ± standard deviation (SD).

To address the question of whether the glycome of CVID patients’ B cells, and especially of CD21^low B cells, is altered, we performed lectin staining of B cells from CVID smB-21lo patients and compared these to their CD21^pos B-cell counterparts in patients and HD. Glycan binding properties for SNA (22, 23), ECL (22, 24), AAL (22, 25), LCA (22, 26, 27), and anti-SLe^X are shown in Supplementary Figure S1 . The gating strategy for naïve, IgM memory (IgM mem), cs mem, and CD21^low B cells is depicted in Figure 1A . Due to the low memory cell numbers in CVID patients, we restricted the analysis of CD21^low B cells to the naïve B-cell compartment.

Lectin staining with α2,6 sialic acid-binding SNA (22, 23) revealed significantly increased levels of SNA ligands on the surface of naïve-like CD21^low B cells from CVID patients compared to naïve CD21^pos B cells from patients or HD. In contrast, patients’ naïve CD21^pos B cells were comparable to those from HD ( Figure 1B ). ECL binds terminal β1,4-linked galactose (22, 24), representing an asialylated form of SNA ligands ( Supplementary Figure S1 ). In line with increased SNA binding, ECL staining demonstrated a reduction of ligands on the surface of naïve-like CD21^low B cells from patients compared to naïve CD21^pos B cells from patients and HD. Naïve CD21^pos B cells from patients and HD exposed ECL ligands to a similar extent ( Figure 1C ). Overall, these data reveal α2,6 hypersialylation of naïve-like CD21^low B cells from CVID patients.

AAL binds to α1,3-, α1,4-, and α1,6-fucosylated glycans (22, 25). Lectin staining revealed an increased expression of AAL ligands on the surface of naïve-like CD21^low B cells from CVID patients compared to naïve CD21^pos B cells from patients or HD ( Figure 2A ). Similarly, lectin staining with LCA, which binds selectively to N-glycans with α1,6-linked core fucose (22, 26, 27), was increased on naïve-like CD21^low B cells from CVID patients compared to naïve CD21^pos B cells from patients and HD ( Figure 2B ). The mean fluorescence intensity (MFI) of both lectins was similar in naïve CD21^pos B cells from patients and HD. Given the prominent role of SLe^X, an α1,3 fucose-containing selectin ligand, on leukocyte trafficking, we analysed its expression on naïve B cells (28–33). Consistent with increased AAL staining, anti-SLe^X staining demonstrated elevated levels of SLe^X on patients’ naïve-like CD21^low B cells compared to patients’ and HD’s naïve CD21^pos B cells. Naïve CD21^pos B cells from patients tended to have increased SLe^X expression compared to naïve CD21^pos B cells from HD, although this did not reach statistical significance ( Figure 2C ). The fucosylation pattern of naïve-like CD21^low B cells supports increased α1,3-, α1,4-, and/or α1,6-linked fucose, increased core fucosylation of N-glycans, and increased SLe^X epitopes relative to naïve CD21^pos B cells.

We further tested whether external factors might influence the glycosylation of B cells in our patients. We did not observe any correlation between age or sex and the glycosylation of CD21^low or CD21^pos B cells in our patients ( Supplementary Figures S2A, B ). Genetic alterations (chromosome 22 aberrations, CTLA4 or KMT2D mutation) ( Supplementary Figure S2C ) and therapeutic treatments, such as low-dose steroids or sirolimus ( Supplementary Figure S2D ), did not affect the surface glycosylation pattern of naïve-like CD21^low and CD21^pos B cells. Lastly, we tested if noninfectious complications, which are common in CVID patients with an accumulation of the CD21^low B-cell subset, were associated with glycome aberrations. We separated patients according to the presence of splenomegaly, granuloma formation, granulomatous lymphocytic interstitial lung disease (GLILD)/interstitial lung disease (ILD), lymphoproliferation (LP), autoimmune organ manifestation (AIO), autoimmune cytopenia (AIC), enteropathy, and hepatopathy. None of these clinical manifestations was specifically associated with significant changes in the surface glycome ( Supplementary Figure S3 ). Overall, these results indicate: (A) that the glycosylation pattern of B cells in CVID is per se not strongly influenced by these factors and (B) that the alterations are characteristic for the naïve-like CD21^low B-cell subset, in contrast to the CD21^pos B-cell compartment.

Next, to determine whether the altered glycome observed in CD21^low B cells from CVID patients is unique to these cells, we compared the glycome of different B-cell subsets derived from peripheral blood and secondary lymphoid organs of HD. Due to the altered composition of CD21^low B cells in HD, including classical or atypical memory B-cell subsets and only very low proportions of naïve-like CD21^low B cells (34), the analysis in HD refers to “total” CD21^low B cells, without differentiating CD27 and IgD expression. Also, in HD, increased SNA and reduced ECL staining corroborated hypersialylation of total CD21^low B cells compared to CD21^pos naïve, IgM mem, and cs mem B cells ( Figures 3A, B ), indicating that this glycosylation pattern is characteristic of the CD21^low B-cell population and not limited to CVID. The degree of α2,6 surface sialylation was similar between the different blood-derived CD21^pos B-cell subsets ( Figure 3A ) and, as previously published (16, 35), also between tonsillar B-cell populations ( Supplementary Figures S4A, B ). The level of terminal β1,4-linked galactose was significantly reduced in circulating cs mem B cells compared to naïve and IgM mem B cells, yet still increased compared to CD21^low B cells ( Figure 4B ). This difference was not seen for the B-cell populations from secondary lymphoid tissue ( Supplementary Figure S4C ).

CD21^low B cells from HD also revealed hyperfucosylation with increased expression of AAL, LCA, and anti-SLe^X ligands ( Figures 3C–E ). Unlike the sialylation profile, fucosylation differed between the various CD21^pos B-cell populations. We observed increasing levels of AAL ligands from the naïve to the cs mem B-cell compartment, still not reaching the level seen in CD21^low B cells ( Figure 3C ), while LCA binding was slightly increased on IgM mem B cells compared to naïve and cs mem B cells ( Figure 3D ). Reduced expression of SLe^X was observed in cs mem B cells ( Figure 3E ). Interestingly, in the secondary lymphoid tissue, the strongest AAL and LCA staining was detected in pre-germinal center cells and plasmablasts ( Supplementary Figures S4D, E ). Overall, HD CD21^low B cells exhibited the same surface glycosylation pattern as naïve-like CD21^low B cells from CVID patients, with the highest levels of α2,6 sialic acids of all investigated B-cell subsets, reduced levels of terminal β1,4 galactose, and increased levels of α1,3, α1,4, and/or α1,6 fucose, N-glycan core fucose, and SLe^X. The direct comparison of total CD21^low B cells in HD and CVID patient-derived naïve-like CD21^low B cells did not reveal significant differences ( Supplementary Figure S5 ). Minor changes were observed for AAL, which tended to be higher in HD, or LCA, which was slightly lower in HD ( Supplementary Figure S5 ). These differences are most likely caused by the changes also observed in the CD21^pos memory B-cell subsets compared to naïve, and thus reflect the different composition of the underlying CD21^low B-cell subsets in HD and CVID ( Figures 3C, D ). Thus, the altered glycome of CD21^low B cells is a general, yet unique and characteristic feature of this B-cell population.

To investigate the impact of B-cell activation on the glycome, naïve CD21^pos B cells from HD were cultured for 48 h in vitro with CpG, anti-IgM, anti-IgM/IFN-γ, or without stimulation. These conditions were chosen to distinguish the impact of BCR versus Toll-like receptor signals and, especially, the effect of the combined stimulation with anti-IgM/IFN-γ, which has been shown to provide an important signal for increased expression of the transcription factor T-bet and the differentiation of CD21^low B cells (3, 5, 36).

All investigated stimuli increased the expression of SNA ligands on the B-cell surface after 48 h compared to unstimulated cells, but anti-IgM/IFN-γ stimulation induced significantly more α2,6-sialylated epitopes ( Figure 4A ). Unlike our observation of CD21^low B cells ex vivo, B-cell activation also generally upregulated ECL ligands on the surface of activated cells, indicating a higher overall expression of terminal β1,4-linked galactose, irrespective of increased sialylation ( Figure 4B ). Analysis of fucose residues showed upregulation of AAL and LCA ligands upon each type of activation, reaching the highest levels after anti-IgM/IFN-γ stimulation ( Figures 4C, D ). In conclusion, all investigated conditions of B-cell activation led to the upregulation of SNA, ECL, AAL, and LCA ligands compared to unstimulated cells. A more potent role of anti-IgM/IFN-γ stimulation was visible for the upregulation of SNA, AAL, and LCA ligands, indicating a significantly stronger α2,6 hypersialylation, α1,3, α1,4, and/or α1,6 hyperfucosylation, and N-glycan core fucosylation under this condition.

The transcription of glycosylation-related genes is a key factor in the cellular regulation of expressed glycans (37). Given the prominent role of anti-IgM/IFN-γ stimulation in the differentiation of the naïve-like CD21^low B-cell phenotype (5), we compared DEGs from anti-IgM/IFN-γ-activated and unstimulated B cells from HD to DEGs of naïve-like CD21^low B cells from CVID patients and naïve CD21^pos B cells from HD ex vivo. A comparative analysis of selected genes involved in sialylation ( Figure 5 ) and fucosylation ( Figure 6 ) was performed. We focused on the expression of genes encoding for α2,6 sialyltransferases and α1,3, α1,4, or α1,6 fucosyltransferases, which are responsible for the generation of SNA, AAL, or LCA ligands, as well as genes contributing to the synthesis of the nucleotide sugars cytidine monophosphate (CMP)-sialic acid and guanosine diphosphate (GDP)-fucose as substrate for the respective glycosyltransferases.

Anti-IgM/IFN-γ in vitro-stimulated B cells revealed diverse deviations in the expression of sialyltransferases that generate SNA epitopes by linking α2,6 sialic acid to galactose or N-acetyl galactosamine (38, 39). Only ST6GALNAC3 tended to be upregulated in the activated B cells, while ST6GALNAC4 and ST6GALNAC5 were not differentially expressed, ST6GAL2 tended to be downregulated, and ST6GAL1, ST6GALNAC1, and ST6GALNAC6 were significantly downregulated ( Figure 5Ai ). Genes for sialidases (also known as neuraminidases) that remove sialic acid from sialic acid-containing glycans (40, 41), such as NEU1 and NEU3, were not differentially expressed in vitro ( Figure 5Aii ). Genes regulating the de novo pathway of CMP-Neu5Ac synthesis (NANS, NANP, and CMAS but not GNE) were significantly upregulated in anti-IgM/IFN-γ-activated B cells ( Figure 5Aiii ) (42, 43). NPL, involved in the degradation of sialic acid, was significantly downregulated ( Figure 5Aiv ) (42, 44).

The analysis of differentially expressed genes in naïve-like CD21^low B cells compared to naïve HD B cells ex vivo revealed a significant upregulation of ST6GALNAC4 and increased ST6GALNAC6 expression, though not reaching significance. The expression of other α2,6 sialyltransferases, such as ST6GAL1, ST6GAL2, ST6GALNAC2, ST6GALNAC3, and ST6GALNAC5 were unremarkable or tended to be downregulated ( Figure 5Bi ). The sialidase NEU1 tended to be upregulated in naïve-like CD21^low B cells, while others, such as NEU3 and NEU4, were not differentially expressed ( Figure 5Bii ). In contrast to the in vitro- activated B cells, NANS was the only significantly upregulated gene related to the de novo synthesis of CMP-sialic acid in naïve-like CD21^low B cells ex vivo (42, 43), while GNE, NANP, and CMAS were not differentially expressed ( Figure 5Biii ). NPL was not differentially expressed ex vivo ( Figure 5Biv ).

Overall, the results suggest an increased de novo synthesis of CMP-sialic acid in anti-IgM/IFN-γ-activated B cells and reduced sialic acid degradation potentially leading to increased substrate for sialyltransferases as the basis for α2,6 hypersialylation. In contrast, the α2,6 hypersialylation in naïve-like CD21^low B cells may reflect a multifactorial basis involving the de novo and salvage synthesis of CMP-sialic acid and upregulation of individual sialyltransferases ( Figure 5C ).

None of the fucosyltransferases FUT3 to FUT11, linking fucose α1,3, α1,4, and/or α1,6 to N-acetyl glucosamine (45–47) and thus elicit AAL or LCA ligands, were differentially expressed in anti-IgM/IFN-γ-stimulated B cells ( Figure 6Ai ). In contrast, the FUCA1 and FUCA2 which remove fucose from fucosylated glycans, were significantly downregulated ( Figure 6Aii ) (40, 41). While activated B cells significantly upregulated FUOM, further genes of the fucose salvage pathway (FPGT, FCSK) were not differentially expressed ( Figure 6Aiii ) (48, 49). However, with GMDS and GFUS, all genes involved in the de novo synthesis of GDP-fucose were significantly upregulated in vitro ( Figure 6Aiv ) (48, 49).

Our analysis of naïve-like CD21^low B cells ex vivo exhibited diverse expression of fucosyltransferases. FUT7, one of the α1,3 fucosyltransferases with the highest potency in creating the SLe^X epitope (47, 50), and FUT8, the only known fucosyltransferase that can link fucose α1,6 to form N-glycan core fucose (45), were significantly upregulated ( Figure 6Bi ). The significant upregulation of both fucosidases, FUCA1 and FUCA2, in naïve-like CD21^low B cells ( Figure 6Bii ) and the trend toward upregulation of FUOM suggests increased recycling of fucose in naïve-like CD21^low B cells, in the presence of regular expression of the other genes involved in the fucose salvage pathway, FPGT and FCSK ( Figure 6Biii ). Genes of the de novo synthesis of GDP-fucose were not differentially expressed ( Figure 6Biv ).

As seen for sialylation, in vitro anti-IgM/IFN-γ-activated B cells upregulate genes involved in the de novo synthesis of GDP-fucose, potentially resulting in increased substrate for fucosyltransferases as basis for hyperfucosylation. In contrast, the basis in naïve-like CD21^low B cells ex vivo seems to involve the upregulation of individual genes of fucosyltransferases and the fucose salvage pathway ( Figure 6C ).