We performed a retrospective immunophenotypic analysis of 1,571 consecutive fresh PB samples, corresponding to 1,180 pediatric patients (261 patients were evaluated more than once) collected at the Bambino Gesù Children Hospital in Rome from June 2019 to December 2020. Immunophenotypic analysis were carried out on all PB samples, whereas the correlations with clinical features were only performed on the first flow-cytometric assessment. Seven categories of disorders were identified based on the main clinical findings, diagnosis, or presumptive diagnosis of the patients: immunodeficiency disorders, autoimmune diseases, inflammatory diseases, hematological diseases, infectious diseases, neurological diseases, and other pathologies. Some patients had been diagnosed before the study. We have also included 50 age-matched healthy children control. In the study cohort, 58.1% of the children were males and 41.9% were females with a median age of 6 years (range: 0–18 years); in the healthy cohort, 62% of children were males and 38% were females, with a median age of 5 years (range: 0–18 years).

In order to achieve high levels of standardization, reagents of a B-cell multicolor panel were used in a dried format (BD Bioscience, Table 1) and bulk lysing was performed on the entire blood sample. Briefly, 500 μl of fresh total PB (EDTA) were incubated for 10 min at room temperature with 9.5 ml of the lysing solution Pharm Lyse 1X (BD Biosciences) to lyse red blood cells. Cells were then washed twice with 10 ml of phosphate-buffered saline (PBS) containing 1% of bovine serum albumin (BSA) to wash-off the free IgGs which could interfere with the anti-IgG antibody surface staining, preventing the identification of IgG⁺ memory B cells. Cells were re-suspended in 200 μl of PBS and added to a B-Cell lyotube (BD Biosciences). The CD21 BV605 antibody was also added to the mix. After a 20 min incubation at room temperature in the dark, samples were washed with PBS 1% BSA and re-suspended in 300 μl PBS 1% BSA.

Flow cytometric data were acquired on a BD FACSLyric™ cytometer (BD Biosciences) and analyzed by FlowJo ver. 10.7 (Becton Dickinson and Company). Acquisition criteria were set as follows: 30,000 CD19⁺ events OR max 3 min acquisition at medium rate (60–70 μL/min). Samples with less than 3,000 CD19⁺ events were excluded from the analysis. CD19⁺ B-cells were gated into several subpopulations: transitional (CD24⁺CD38⁺⁺) and plasmablasts (CD24^–CD38⁺⁺), naïve [NOT (transitional or plasmablasts) CD21⁺CD27^–], memory B cells (MBCs) [NOT (transitional or plasmablasts) CD21⁺CD27⁺], activated memory B cells (actMBCs) [NOT (transitional or plasmablasts) CD21^–CD27⁺], and atypical B cells (atBCs) [NOT (transitional or plasmablasts) CD21^–CD27^–].

Selected fcs files from 18 pathological and healthy patients were imported into FlowJo ver. 10.7 and CD19⁺ events were gated. A subset of 20,000 CD19⁺ events for each disease and healthy group were randomly selected for normalization purposes, subsequently concatenated in a 160,000 total events fcs file and used for t-Distributed Stochastic Neighbor Embedding (tSNE) analysis (19). The following parameters were included into the t-SNE calculation: CD19-PE-Cy7-A, CD24-PE-A, CD27-APC-A, CD38-PerCP-Cy5.5-A, IgM-FITC-A, IgG-APC-H7-A, IgD-V450-A, CD21-BV605-A. t-SNE maps were generated by plotting each event by its t-SNE dimensions. Then, the Flow Cytometric Self-Organizing Map (FlowSOM) function was used to automatically identify B-cell subsets (20).

Star charts were obtained from 40 fcs files of stained samples from healthy controls and study patients. After gating B cells and normalizing each sample to 5,000 CD19⁺ events, the concatenated file of 200,000 total events was used for FlowSOM analysis. The following parameters were included into FlowSOM calculation: CD19-PE-Cy7-A, CD24-PE-A, CD27-APC-A, CD38-PerCP-Cy5.5-A, IgM-FITC-A, IgG-APC-H7-A, IgD-V450-A, CD21-BV605-A.

Data were summarized as medians, 25th, 75th percentiles, and range (min and max). The non-parametric Kruskal–Wallis test was used to compare more than two independent groups, whereas pairwise comparisons were evaluated by the Mann–Whitney U-test. A Wilcoxon matched-pairs signed rank test was performed for matched-paired data, while correlation between two non-parametric continuous variables was assessed by Spearman’s rank correlation. P-values < 0.05 were considered statistically significant and reported as two-sided. All statistics were carried out with the use of STATA/SE 12.0 for Windows.

A flow cytometry staining was performed on 1,571 and 50 fresh PB samples collected from pathological pediatric patients and healthy children, respectively. The combination of antibodies allowed the identification of several B-cell populations: transitional B cells, naïve B cells, MBCs, atBCs, actMBCs, and plasmablasts (Figure 1). The relative percentages and statistics of the B-cell subsets in the two studied cohorts are shown in Table 2. An in-depth age-related analysis of B-cell subsets was also performed comparing percentages between ten different age ranges to assess the age-related changes observed during childhood (Supplementary Figures 1A,B). Transitional and naïve B cells were shown to decrease with age, whereas MBCs significantly increased during the first 3–4 years of age. Percentages of plasmablasts, actMBCs, and atBCs were similar except for the first year of life when they resulted significantly lower.

In the study cohort a positive correlation was observed between percentages of actMBCs and both MBC (ρ = 0.54, P < 0.001) and atBC (ρ = 0.61, P < 0.001) populations, suggesting an inter-relationship among the three types of B-cell subsets. On the contrary, there was a negative correlation between percentages of naïve B cells and MBC (ρ = –0.55, P < 0.001) or actMBC (ρ = –0.63, P < 0.001) populations.

In the study cohort the frequency of atBCs, expressed as percentage of the B-cell population, presented a right-skewed distribution (skewness = 5.69), with 8.9% of the data points (139/1,571) appearing as outliers [> Q3 (upper quartile) + 1.5 × IQR (interquartile range)] (Figure 2A). AtBCs were also observed in the healthy cohort and the comparison with the study cohort did not show any statistically significant difference (medians: 2.4 vs. 2.3%, respectively, P = 0.430) (Table 2). We then focused on the wider percentage range of atBCs observed in the study cohort (between 0.14 and 66.5%) compared to the healthy cohort (between 0.1 and 5.2%), analyzing all samples with atBCs percentages greater than 5%. For this purpose, the study cohort samples were subdivided into four classes based on relative percentages of atBCs (Table 2 and Figure 2). Most of the PB samples (1,325/1,571, 84.3%) presented a percentage of atBCs ≤ 5%, whereas a low increase of atBCs (> 5% and ≤ 10%) was detected in 10.1% (159/1,571) of the samples. A medium increase of atBCs percentages (> 10% and ≤ 20%) was observed in 3.4% (54/1,571) and a high increase (> 20%) was recorded in 2.1% (33/1,571) of the study cohort samples.

Previous immunoglobulins expression analysis on atBCs of healthy and non-healthy adults demonstrated the presence of switched and unswitched B cells (2, 21), whereas very little is known about immunoglobulins expression distribution in children. With this aim, we selected samples with relative atBCs percentages > 5% (246/1,571), we analyzed the expressions of IgM and IgD in this B-cell subset and compared them with MBC population. Four subpopulations with different expression of the two immunoglobulins were discriminated: IgM⁺IgD⁺, IgM⁺-only, IgD⁺-only, and IgM^–IgD^– (Figure 2B). IgM⁺IgD⁺ B cells were the most abundant population among atBCs, with a median frequency of 55.8%. No differences in IgM⁺IgD⁺ B cells frequency were observed between atBCs and MBCs (55.8 vs. 54.3%, P = 0.913). On the contrary, IgM⁺-only and IgD⁺-only B cells were significantly higher in atBCs compared to MBCs (10.7 vs. 5.1%, P < 0.001 and 7.6 vs. 4.0%, P < 0.001, respectively). Moreover, a significant lower frequency of IgM^–IgD^– switched B cells was detected in atBCs compared to the MBCs (13.3 vs. 33.2%, P < 0.001). Furthermore, the expression of IgG was evaluated on IgM^–IgD^– atBC and IgM^–IgD^– MBC subsets. IgG⁺ atBCs were higher than IgG^– atBCs within the IgM^– IgD^– B-cell subpopulation (7.7 and 3.5%, respectively), although significantly lower than IgG⁺ MBCs (7.7 vs. 15.1%, P < 0.001) (Figure 2B). Interestingly, IgM⁺-only atBCs expressed the immunoglobulin isotype with a lower density (dim expression) than MBCs (MFI, 2,185 vs. 6,634, P < 0.001). On the contrary, IgD immunoglobulin was expressed by IgD⁺-only atBCs at a higher level compared to MBCs (MFI, 8,518 vs. 2,781, P < 0.001) (Supplementary Figure 2A).

To complete the in-depth immunophenotyping of atBCs, the expression of CD38, CD24, and CD19 was then investigated. Due to the extremely low levels or absence of CD38 expression on atBCs, the use of this cell surface marker allowed a clearer discrimination of this specific B-cell subset. Conversely, CD24 was present in a more heterogeneous way on atBCs, with a median frequency of 27% (range from 0.6 to 93.7%) (Figure 2C). Immunoglobulin isotypes distribution of CD24⁺ atBCs reflected the overall atBCs expression frequency: IgM⁺IgD⁺ B cells were the most abundant population (median frequency 68.8%) (Supplementary Figure 2B). Of note, CD19 expression level was significantly higher in atBCs than in MBCs, actMBCs, and naïve B cells (Figure 2D). The cell size across the different B-cell subsets was also analyzed. The atBC population showed a significantly higher FCS-A (forward scatter, Area) than naïve B cells but a lower FCS-A value than both MBCs and actMBCs (Figure 2D).

In order to assess the association between the increase of atBCs in PB samples and pediatric diseases, we performed correlation analyses of the flow cytometric findings with children’s diagnosis or main clinical features. First, patients with atBCs > 5% (181/1,180, 15.3%) were selected and categorized into seven types of disorders, based on the main clinical findings, diagnosis, or presumptive diagnosis. For patients with more than one flow cytometric evaluation during the study time, only the first assessment was included in the analysis. As expected, patients re-evaluated after at least 1 month, did not show any significant change in atBCs percentages [9.4 (T₀) vs. 9.7% (T₁), P = 0.910] (Supplementary Figure 3). Subsequently, the seven clinical categories were associated to a low (> 5% and ≤ 10%), medium (> 10% and ≤ 20%), and high (> 20%) increase of atBCs. Among the selected patients, 132 (11.2%), 31 (2.6%), and 18 (1.5%) children showed a low, medium, and high increase of atBCs, respectively. As shown in Figure 3, all clinical disorders were represented in the group with a low increase of atBCs, with a slight prevalence of the immunodeficiencies (30%). Immunodeficiencies, as a main clinical finding, diagnosis, or presumptive diagnosis, accounted for approximately half of the clinical conditions in children with both medium and high frequencies of the atBC population (55 and 44%, respectively). Autoimmune and inflammatory diseases were more frequently associated with samples with a high increase of atBCs (22 and 17%, respectively). Hematological patients were associated with comparable frequency to all three atBC groups (12% in the low, 13% in the medium, and 17% in the high atBC groups). Infectious diseases, neurological, and other disorders were not observed in the group with the highest percentage increase of atBCs (Figure 3).

The prevalence of the specific disorders in the study cohort, associated to a low (> 5 and ≤ 10%), medium (> 10 and ≤ 20%), and high (> 20%) increase of atBCs are reported in the Supplementary Table 1. The most frequently observed immunodeficiencies were: DiGeorge syndrome (15/64, 23%), IgA deficiency (11/64, 17.2%), combined immunodeficiencies (CID), and severe combined immunodeficiencies (SCID) (11/64, 17.2%), Ataxia-telangiectasia (4/64, 6.3%), common variable immunodeficiencies (CVID) (4/64, 6.3%), Wiskott-Aldrich syndrome (3/64, 4.7%), acquired immune deficiency syndrome (AIDS) (2/64, 3.1%), and primary hemophagocytic lymphohistiocytosis (HLH) (2/64, 3.1%). Three patients with SCID were evaluated after hematopoietic stem-cell transplantation (HSCT).

When comparing relative percentages of atBCs in patients suffering from immunodeficiencies, we observed higher median values in patients with ataxia-telangiectasia and CID/SCID. However, the highest relative percentage of atBCs was detected in a patient with CVID (Figure 4). In contrast, DiGeorge Syndrome and IgA deficiency patients had the lowest median percentages of atBCs. No significant differences were observed neither for immunoglobulin isotypes nor for CD24 expression on atBCs across all immunodeficiencies (Supplementary Figure 4).

Among the autoimmune diseases with increased percentages of atBCs, we were able to observe Fisher-Evans syndrome (2/11, 18.2%), immune thrombocytopenia (2/11, 18.2%), ANCA glomerulonephritis, HLH associated with connective tissue disease, and autoimmune hemolytic anemia (all 1/11, 9%) (Supplementary Table 1).

As part of the inflammatory diseases, secondary HLH (3/18, 16.7%), Crohn’s disease (3/18, 16.7%), juvenile idiopathic arthritis (3/18, 16.7%), A20 haploinsufficiency (2/18, 11.1%) were the most frequently observed among patients with atBCs > 5%. Patients with secondary HLH were only observed within the high atBCs increase group (> 20%). Most of the inflammatory diseases were detected in the group with the lowest increase of atBCs while in the medium atBCs increase group we observed only one patient with A20 haploinsufficiency (Supplementary Table 1).

Among the hematological patients with atBCs > 5%, 14/23 (60.9%) children suffered from acute lymphoblastic or myeloid leukemia, 2/23 (8.7) from non-Hodgkin lymphoma, 2/23 (8.7%) from Fanconi anemia, and 2/23 (8.7%) from iron-deficiency anemia. Children with acute lymphoblastic or myeloid leukemia (11/14, 78.6%) and the two patients with Fanconi anemia were evaluated after HSCT. Hematological patients were present in all three atBC groups with a comparable frequency (Supplementary Table 1).

The majority of patients with infectious diseases had a modest increase of atBCs (> 5% and ≤ 10%). In most cases, they had recurrent infections not associated to previous disorders (15/25, 60%). Pulmonary tuberculosis (3/25, 12%) and Malaria (2/25, 8%) were also found to expand the atBC subset (Supplementary Table 1).

In the context of neurological diseases, early infantile epileptic encephalopathies (EIEE) (6/11, 54.5%) and cerebellar ataxia (2/11, 18.2%) were the most frequent in patients with atBCs > 5%. No patient with neurological disorders was found in the high increase group, whereas three cases (West syndrome, X linked lissencephaly, GABRA1/EIEE-19) were observed in the group with a medium increase of atBCs (Supplementary Table 1).

When focusing on all other diseases in our study cohort, cardiovascular diseases (9/29, 30%), genetic disorders (5/29, 16.7%), metabolic disorders (2/29, 6.7%), and carcinoma (2/29, 6.7%) were the most frequent in patients with atBCs between 5 and 20%. One patient with hypoplastic left heart syndrome, one with polycystic kidney disease and kidney failure and another patient who underwent surgical splenectomy showed a medium increase of atBCs percentages (Supplementary Table 1).

In order to appreciate high-dimensional similarities of cells and to visualize the proportions of the atBC population in patients with different diagnosis, data from B-cell phenotype have been represented as a t-distributed Stochastic Neighbor Embedding (tSNE) plot after data integration and batch correction. Separated tSNE plots for samples from representative diseases and healthy children are shown in Figure 5. Ten clear immunophenotypic clusters were automatically identified and recognized using Flow cytometric Self-Organizing Map (FlowSOM). The two identified subsets of atBCs (IgM⁺IgD⁺ and IgM⁺-only) were visually and quantitatively increased in representative diseases compared to the healthy cohort. Differences in other B-cell subsets were easily evaluable (Figure 5).

A star chart visualization by FlowSOM was applied to the immunophenotypic analysis of selected patients from both study and healthy cohorts in order to investigate cluster relationships among all B-cell subsets (Figure 6A). As shown in Figure 6B, atBCs established a well-defined and divergent cluster compared to MBCs. Moreover, a high increase of atBCs percentages (> 20%) corresponded to a marked reduction of unswitched and, more in particular, switched MBCs.