XLA and healthy donors were informed of the final use of their blood and signed an informed consent document. The study was approved by the local ethics committee, NRC Institute of Immunology FMBA (Moscow, Russia), protocol 6-1, 09 June 2020, and conducted in accordance with the Declaration of Helsinki. The 10 XLA patients (age 18–36) and a cohort of 15 young donors (age 22–35) and 6 old donors (age 49–83) ( Supplemental Table 1 ) were enrolled in the study. All donors were males. Individuals with previously diagnosed cancer or severe autoimmune disease were excluded. The same exclusion criteria were applied to the control cohort. One XLA donor had rheumatoid arthritis affecting both hips, both knees, the right ankle, both wrists, and the metacarpophalangeal joint of the right first finger of the left hand, with ankylosis of the left hip joint, aseptic necrosis of the head of the right femur, and fibrous contracture of the right ankle joint. The other donors had no signs of autoimmune or inflammatory conditions. The baseline treatment for XLA patients included IVIG therapy (0.4 g/kg). We performed immunophenotyping analysis and verified the lack of CD19 expression on XLA PBMCs ( Supplemental Figure 5 ).

Peripheral blood (12–20 ml) was collected into a number of EDTA-treated Vacutainer tubes (BD Biosciences, Franklin Lakes). For cell sorting of CD4⁺ and CD8⁺ (defined as CD3⁺ CD4^-) memory and naïve T cells, RTE, mnCD4+, naïve Treg we used a strategy previously described in Ref (20). Further details see in Supplemental Methods . Number of sorted cells for each population see in Supplemental Tables 2, 3 . All cell populations were collected directly into the RLT buffer (Qiagen) and stored at -70C.

Total RNA was isolated using the RNeasy mini kit (Qiagen) according to the manufacturer’s instructions. In all experiments described here, cDNA libraries were obtained using 5’-RACE with unique molecular identifiers (UMI) (20) using Human TCR profiling kit (MiLaboratory LLC). Libraries were sequenced on Illumina NextSeq 500 using 300 cycle reagent kit, paired-end 150 + 150 nt mode.

Raw sequencing data was analyzed using MIGEC software v.1.2.9 (21). UMI-labeled TCRβ cDNA molecules were obtained per sample with at least ~1 read per UMI ( Supplemental Tables 2 , 3 ). Analysis of the averaged CDR3 characteristics was performed weighted by the abundance of each clonotype. All of the physicochemical characteristics were calculated and averaged for the five amino acid residues located in the middle of CDR3, which are considered to have the highest probability to contact with peptide-MHC complex (20). Further details see in Supplemental Methods .

Statistical analysis was performed on processed datasets with R. Clonotype CDR3 features were calculated for the most frequent V segments extracted from full clone sets to avoid bias related to a particular TRBV segment’s contribution to CDR3 analysis and individual differences in V segment usage. In case of diversity estimation all V segment subsets within the same cell type were down-sampled to the same number of UMI. To exclude potential dependence of clonotype features from the V segment, all features within the same cell type and the same V segment were turned to Z-scores as described in (20). To compare the medians between three groups of samples the Kruskal-Wallis test was used followed by Dunn test with the Benjamini-Hochberg correction for multiple testing, if not mentioned otherwise. To compare the medians between two groups of samples the Wilcoxon rank sum test was applied. Further details see in Supplemental Methods . The data is available by PRJNA752656 in the SRA (NCBI).

RNA was isolated using the RNeasy Micro Kit (Qiagen) and analyzed using a Qubit 2.0 Fluorometer (Thermo Fisher Scientific). cDNA libraries were prepared with SMART-Seq v4 Ultra Low Input Kit for Sequencing (Takara Bio). The samples were sequenced using a HiSeq 4000 (75 bp, 10 million average paired reads per sample). Further details see in Supplemental Methods and Supplemental Table 4 .

We first assessed the relative abundance of naïve T cells overall as well as distinct naïve T cell subsets in PBMCs collected from seven young (18–36 years old) male XLA patients and five healthy sex- and age-matched donors ( Supplemental Table 1 , Figure 1 , Supplemental Figure 1A ). The percentages of naïve CD4⁺ T cells were comparable for the XLA and healthy cohorts ( Figure 1B ). The nCD4⁺ lymphocyte population was subdivided into naïve T_reg (CD4⁺CD27⁺CD45RA⁺CD25⁺), mature naïve CD4⁺ (mnCD4⁺; CD4⁺CD27⁺CD45RA⁺CD25^-CD31^-), and CD31⁺ cells enriched with recent thymic emigrants (RTE; CD4⁺CD27⁺CD45RA⁺CD25^- CD31⁺) (20). Notably, the proportion of naïve T_regs among nCD4⁺ T cells was prominently decreased in the XLA cohort ( Figure 1C ) in contrast to RTEs and mnCD4⁺ cells ( Figures 1D, E ). Low T_reg counts were previously reported in PBMCs from children with XLA (22) and in spleen of the B cell deficient mice (17). Our data indicate that in the condition associated with primary B cell deficiency, the generation of T_reg subsets may be undermined at the level of naïve T_reg production.

Several studies (23, 24) have demonstrated that the naïve T cell subset is not meaningfully affected at the cellular level in XLA patients. However, the impact of a lack of B cells on TCR repertoire selection has not been deeply explored. To investigate this, we performed unique molecular identifier (UMI)-based 5’-RACE TCRβ profiling, and compared repertoire characteristics for the sorted naïve and memory CD4⁺ and CD8⁺ T cells of young XLA patients versus young and old (49-83 y.o.) healthy donors ( Supplemental Figures 1B, C and Supplemental Table 2 ). To increase the statistical power of analysis, we separately compared diversity metrics for the dominant TRBV segments after down-sampling to an equal number of UMI-labeled TCRβ cDNA molecules, as described in (20).

We assessed normalized diversity metrics, which estimate the evenness of clonotype frequency distribution (i.e., normalized Shannon–Wiener) and the richness of the repertoire based on the number of clonotypes that occur once or twice (i.e. Chao1). We observed a decline in Chao1 and normalized Shannon-Wiener diversity values for the nCD4⁺ TCRβ repertoires of young XLA patients compared to sex- and age-matched healthy donors ( Figures 2A, B ). In this respect, the naïve CD4⁺ TCR repertoire of XLA patients reflected a pattern observed in the repertoires of elderly individuals ( Figures 2A, B ). Similar changes in the elderly cohort were previously linked to thymic involution and continuous and biased peripheral proliferation of naïve T cells in the course of cell niche replenishment at the periphery (20, 25). The decrease in TCR diversity for XLA nCD8⁺ cells was less prominent ( Supplemental Figures 2A, B ).

In contrast to the naïve TCR repertoires, we observed more evenly distributed clonotype frequencies in the memory CD4⁺ repertoires of XLA patients compared to matched healthy donors ( Figure 2B ). Previous studies in B cell-deficient mice (26, 27) and clinical data from XLA patients (23) have shown that B cells might be required for appropriate CD4⁺ activation and the generation and maintenance of pathogen-specific memory T cells. In mice, a lack of B cells results in the deeper depletion of antigen-activated CD4⁺ T cells at the contraction phase, followed by the generation of a smaller number of antigen-specific CD4⁺ memory T cells (28). Significant reduction of CD4⁺ memory T cell counts has also been previously reported in XLA donors (23). Thus, the lack of antigen-specific B cell support could contribute to the diminished T cell memory formation in XLA patients. That being said, other evidence indicates that antigen-specific memory T cells can be detected years after immunization of XLA patients, similarly to healthy cohorts (29).

Next, we assessed the physicochemical properties of the amino acid residues forming the central part of the CDR3 ( Figures 2C, D ; Supplemental Figures 2C, D , Figure S3 ), which primarily interacts with the antigenic peptide within pMHC complexes (20). On average, nCD8⁺ and nCD4⁺ CDR3β repertoires of XLA patients contained a higher proportion of “strongly interacting” hydrophobic and aromatic amino acids (F, I, L, M, V, W, and Y; Figure 2C and Supplemental Figure 2C ) and bulky amino acid residues ( Figure 2D and Supplemental Figure 2D ) compared to matched healthy donors. In our previous experience, the average number of strongly interacting amino acid residues is a basic characteristic determining the average features of the CDR3β repertoire (20, 30, 31). Large numbers of strongly-interacting amino acids may be interpreted as higher average affinity and potentially also increased cross-reactivity of a TCR repertoire (32). The observed increased “strength” in nCD8⁺ and nCD4⁺ repertoires of XLA patients could therefore result from increased competition between naïve T cells for tonic signaling in the absence of antigen-presenting B cells (33). Notably, the difference in the number of strongly-interacting amino acids was more prominent for the nCD4⁺ T cells compared to the memory CD4⁺ (mCD4⁺)T cell subset. There was no significant difference in CDR3 length or insertion size among the naïve or memory CD4⁺ or CD8⁺ T cell subsets ( Supplemental Figure 3 ).

Heterogeneous mnCD4⁺, RTE, and naïve T_reg lymphocyte populations are maintained on the periphery through different mechanisms, and contribute differently to the cumulative landscape repertoire of naïve CD4⁺ Т lymphocytes (20, 34). To explore the impact of each naïve CD4⁺ subset, we analyzed TCR repertoires of sorted mnCD4⁺, RTE, and naïve T_reg cell fractions from XLA patients and healthy donors ( Figure 3 ). At least 5,000, 28,000, and 13,589 cells were sorted for naïve T_reg, mnCD4⁺, and RTE subsets, respectively ( Supplemental Table 3 ), and UMI-based analysis was performed as above.

For the elderly donors, and to a higher extent for the XLA patients, we observed increased convergence as measured by the relative number of distinct nucleotide sequence variants for each amino acid CDR3 sequence. This effect was observed for all subsets but was more prominent for the naive T_reg cells ( Figure 3A ). The latter subset was also characterized by increased clonality in XLA patients, reflected by a decreased normalized Shannon-Wiener index ( Figure 3B ) but a smaller proportion of shared public clonotypes ( Figure 3C ). Such features of XLA patients’ naïve T_reg cells may indicate more focused production and peripheral proliferation with narrowed antigenic specificities in an individual MHCII context. This could be a possible consequence of the lack of B cells presenting a wide range of self and commensal antigens both in the thymus and on the periphery for naïve T_reg positive selection and further tonic signaling, which is strongly required for T_reg homeostasis (35). Altogether, the decline in the naïve T_reg proportion among nCD4⁺ cells ( Figure 1C ), increased repertoire convergence ( Figure 3A ) and decreased diversity suggest impaired thymic T_reg selection in XLA patients, accompanied by biased peripheral proliferation.

The average CDR3β length was shorter in all nCD4⁺ T cell subsets (mnCD4⁺, RTE and naïve T_reg cells) of XLA individuals compared to healthy young donors ( Figure 4A ). Additionally, the average number of random nucleotides inserted between Vβ-Dβ and Dβ-Jβ segments in CDR3β was smaller in RTEs and naïve T_reg cells ( Figure 4B ). It indicated limited capacity for conformational changes of CDR3β in XLA nCD4⁺ repertoires, potentially reflecting more stringent selection. Shortening of CDR3β length in naïve conventional repertoires also occurred in aged healthy individuals ( Figures 4A, B ) (20). In this respect, the repertoires of the XLA cohort resembled those of elderly individuals.

We next evaluated the physicochemical landscape of TCR repertoires for the nCD4⁺ T cell subsets. The amino acid properties of XLA CDR3β differed substantially from age-matched healthy donors, although the extent of the differences depended on the cell type ( Figures 4C–F ). In particular, the CDR3β repertoires of both RTE and mnCD4⁺ subsets from XLA patients were characterized by increased numbers of strongly interacting amino acid residues ( Figure 4C ), supporting the results obtained for the entire pool of nCD4⁺ T cells ( Figure 2C ). Thus, the difference in strength between mnCD4⁺ cells and naïve T_reg cells was decreased in XLA individuals ( Figure 4D ) potentially reducing the average capacity of T_regs to suppress activated conventional CD4⁺ T cells. High-affinity TCRs allow T_reg cells to compete efficiently and in an antigen-specific fashion with conventional T cells for binding to peptide-MHC complexes presented by APCs (36, 37). Recent data have demonstrated that TCR-specific interactions may be one of the possible suppression mechanisms exerted by T_regs (37–39).

Repertoires from all XLA nCD4⁺ subsets differed from their counterparts in young healthy individuals in having an increased average number of bulky amino acid residues in the central region of CDR3β ( Figure 4E ). Additionally, mnCD4⁺ T cells have a more charged CDR3β region in XLA patients than in healthy donors ( Figure 4F ). In contrast, naïve T_regs in this group have decreased mean charge within the CDR3β region. The abovementioned shortening of CDR3β might be partially compensated at the expense of a high frequency of bulky amino acid residues within CDR3β in the XLA patient repertoires. Notably, we have shown previously that CDR3β domains from mnCD4⁺ T cells tend to have smaller numbers of strongly interacting and bulky amino acid residues in aged individuals (20), a finding that we have confirmed here ( Figure 4 ). However, we observed the opposite in XLA patients, suggesting an association with B cell deficiency rather than general immunosenescence. In other words, the differences in the nCD4⁺ TCR repertoire indicate that the specific process of T cell selection—but not the early exhaustion of naïve T cell pools by pathogen burden—triggers these differences in XLA patients.

We next applied principal component analysis to the nCD4+ TCRβ repertoires of young healthy donors and XLA patients that revealed major contributing parameters, including the CDR3 length and average number of bulky amino acid residues, and illustrated the profound divergence of CDR3β repertoire characteristics between healthy donors and XLA patients ( Figure 4G ).

To get insights on the functional alterations of mnCD4⁺ and naïve T_reg cell subsets, we performed transcriptomic analysis of these subsets from XLA and healthy young donors ( Figure 5 ). Despite the fact that the mnCD4⁺ cells showed a significant change in the physicochemical properties of their TCR repertoires, we observed no essential changes in their transcriptome ( Figure 5D ).

We evaluated classical T_reg signatures of gene expression (e.g., FOXP3, IL2RA, TIGIT, CEACAM4, RTKN2) in the sorted naïve T_reg cells in comparison to mnCD4⁺ cells from XLA and healthy donors ( Figures 5A, B ). XLA naïve T_regs maintained comparable expression of T_reg phenotype-specific genes to healthy donors ( Figure 5C ) as well as T-cell-specific genes ( Supplemental Figure S4 ). Functionally distinct T_reg subsets that can be identified based on expression of certain transcription factors (i.e., Bcl6, Stat4, Stat3, T-bet, RORC) (40) and chemokine receptors (i.e., CCR6, CXCR3) (41) were also not overrepresented in XLA naïve T_reg relative to healthy donors.

Nevertheless, we found several atypical transcriptomic features of XLA naïve T_reg in comparison to healthy donors. Among 39 gene transcripts enriched in XLA naïve T_reg were genes that are abundantly expressed by antigen-presenting cells (APCs) and myeloid cells, including IL1b, SECTM1, CD33, LILRB3, PILRA, NCF1, and TACI ( Figures 5E–H ). TACI, also known as TNFRSF13B, is a receptor for a proliferation-inducing ligand (APRIL) and B cell-activating factor of the tumor necrosis factor family (BAFF) (42), which is predominantly expressed on B cells. Recent evidence indicates that TACI expression in T_reg cells promotes their survival and proliferation (43), and the upregulation of TACI in XLA naïve T_reg cells suggests a mechanism of homeostatic naïve T_reg expansion in B cell-deficient conditions. Several genes enriched in XLA cell populations (e.g., SECTM1, CD58) encode co-stimulatory molecules involved in T cell activation (44, 45). CD58 is widely expressed by both hematopoietic and non-hematopoietic cells, including B lymphocytes (46). The functional outcome of CD58-CD2 interaction in CD4⁺ T cells remains poorly understood, but this interaction with CD2 on NK and effector T cells has been shown to contribute to T cell proliferation and NK cell activation (47). Importantly, it can also induce rapid differentiation to an antigen-specific T_reg cell subtype, Tr1, that is characterized by high IL-10 production (48). Thus, CD58 upregulation in XLA might provide additional positive feedback stimulation of T_reg cells.

Interestingly, we also found enrichment of chemokine genes in the XLA naïve T_reg transcriptome, including CCL20, CXCL4 (PF4), and PPBP (CXCL7) ( Figures 5E, F ). CXCL4 has been shown to exert Th17 induction in autoimmune diseases (49). Expression of CCL20 was previously reported in Tfh and Th17 cells and at very low levels in T_regs, where the latter is mediated synergistically by TGF-β/IL-6 (50).

Our data may indicate that peripheral maturation of XLA naïve T_reg cells occur under pro-inflammatory conditions that promote aberrant expression of proinflammatory markers and mediators (e.g., IL1b, CXCL4, CCL20 and S100A11) (49, 51). Interestingly, we also detected several atypically expressed genes that are normally specific to dendritic cells and granulocytes ( Figures 5G, H ). For example, IL1b expression by lymphoid cells is not typical, although it has been described previously for CCR5⁺CD4⁺ T cells (52). Our data assume that enhanced S100A11 and IER3 gene expression might be related to IL1b levels in XLA patients. IL1b-dependent up-regulation of IER3 has been reported to increase T cell lifespan (53), hinting at possible changes in naïve T_reg homeostasis in XLA donors.