T and NK Cells in IL2RG-Deficient Patient 50 Years After Hematopoietic Stem Cell Transplantation

With approval of the institutional review board (protocols P00.068, P01.028 and B17.001), blood (UPN1, healthy donor/controls) and skin samples (UPN1) were analyzed after informed consent was obtained from the hematopoietic stem cell transplant recipient (UPN1) and healthy donor/controls.

The peripheral blood mononuclear cells (PBMC) were isolated by Ficoll density gradient centrifugation and analyzed directly or after cryopreservation. Skin biopsies were directly cut into small pieces and further processed.

To study degranulation, PBMC were cocultured for 4 h with K562 cells. Anti-CD107a (Becton Dickinson (BD), FITC) was added to the PBMC at start of the coculture. Intracellular chemokine (CCL4, XCL1) and cytokine (IFN-γ) production was measured after stimulation with coated anti-CD16 for 4 h or K562 cells for overnight. For the CD16 stimulation, a flat-bottom plate was coated with 5 μg/ml goat anti-mouse antibody (BD) and subsequently with 1 μg/ml anti-CD16 (clone 3G8, Biolegend) in PBS. Mouse IgG1 isotype (1 μg/ml, Biolegend) served as negative control. Cells were cultured in AIM-V medium (Life Technologies, Bleiswijk, the Netherlands) supplemented with 10% FCS (Sigma–Aldrich, St. Louis, MO, USA) and 1% penicillin/streptomycin (Sigma–Aldrich) at 37 °C. For each cell culture condition, after 1 h of culture Golgistop (BD) was added.

PBMC were stained with fluorochrome-conjugated antibodies in PBS supplemented with 0.5% BSA and 2 mM EDTA, for 30 min at room temperature (RT). An eight-tube flow cytometry panel was designed to study the TCR Vβ expression and the T-cell phenotype. Each tube included three TCR Vβ antibodies from the IOTest Beta Mark Kit (Beckman Coulter (BC), Brea, CA, USA) and a backbone panel of the following surface markers: CD3, CD4, CD8, CCR7, CD45RA, CD27, CD28, TIGIT and PD-1. Details of the antibodies used are listed in Table S1A. DAPI (150 ng/ml, Sigma–Aldrich) was added prior to measurement to exclude dead cells. Data were acquired on a LSRII flow cytometer (BD) using FACS Diva software (v8.0, BD).

PBMC were stained with fluorochrome-conjugated antibodies in PBS supplemented with Brilliant Stain buffer plus (BD), 0.5% BSA, 2 mM EDTA and 0.02% NaN3, for 30 min at RT. The TCR Vβ T-cell panel as described for conventional flow cytometry was supplemented with the following antibodies: CXCR6, CD127, CD57, TcRγδ, CD56, CX3CR1, CD16, NKG2A, CD95, DNAM1, KLRG1, CD69, CD103 and CLA (Table S1B). The NK phenotype panel included antibodies against CD69, DNAM1, CD57, CD3, CD45RA, CD62L, CD56, CD16, CX3CR1, CD7, KLRG1, KIRs, NKp46, CD127, NKG2A, NKG2D, TIGIT, NKG2C, CD45 and CD27 (Table S1B). The staining with unconjugated NKp46 antibody was performed prior to incubation with conjugated antibodies. PBMC were incubated with Fc block (eBioscience, San Diego, CA, USA) for 10 min at RT, followed by 30-min incubation with anti-NKp46, washing and 30-min incubation with secondary antibody. 7AAD or DAPI was added prior to acquisition to exclude dead cells. The peptide-stimulated PBMC were stained with Fixable Viability Dye eFluor 455UV to exclude dead cells prior to the extracellular staining. For the intracellular staining of stimulated PBMC, cells were fixed in 4% paraformaldehyde after extracellular staining and permeabilized in 0.1% saponin, as previously described [9]. Next, cells were incubated in Fc block for 10 min at RT. Intracellular staining with antibodies (Table S1B) was performed for 30 min at 4 °C. For the XCL1 staining, 6.7% donkey serum (Jackson ImmunoResearch, West Grove, PA, USA) was added together with unconjugated goat XCL1 to the mix. Finally, cells were incubated with secondary antibody to detect XCL1 for 30 min at 4 °C. Data were acquired on a 3L or 5L Aurora spectral cytometer (Cytek Biosciences, Fremont, CA, USA), using SpectroFlo software (v2.0, Cytek).

PBMC from UPN1 and an age-matched healthy control were stained with fluorochrome-conjugated antibodies in PBS supplemented with 0.5% BSA and 2 mM EDTA, for 30 min at RT with the following antibodies: CD3, CD4, CD8, CCR7, CD45RA, CD27, CD28 and CD31 (Table S1C). Four populations were sorted per sample: naive CD4 and CD8, and CD27⁻ and/or CD28⁻ memory CD4 and CD8 T cells (Fig. S1). The populations were purified on an Aria I cell sorter (BD).

Skin biopsy pieces from wart enriched skin (wart⁺), developing warts (wart^±) and healthy skin (wart⁻) were incubated for 1 h at 37 °C in IMDM (Lonza, Breda, The Netherlands) supplemented with 10% human AB serum (Life Technologies), 50 µg/ml gentamycin (Life technologies), 25 µg/ml Fungizone (Life Technologies) and 10% penicillin/streptomycin. The skin pieces were cultured in IMDM supplemented with 7.5% human AB serum, 1% penicillin/streptomcyin and 1000 IU/ml IL-2 (Aldesleukin, Novartis, Arnhem, The Netherlands) for 26 days (culture medium). Half of the culture medium was refreshed once per two days with culture medium. When cells were overgrown, cells were divided over two or more wells in the culture medium.

Thirteen 25-mer peptides with ten amino acids overlap (Table S2) from three distinct regions of the L1 protein of the HPV2 virus were synthesized (Peptide 2.0, Chantilly, VA, USA). Peptide stimulation of PBMC revealed two candidate immunogenic peptides. Next, PBMC were cultured with 1 μg/ml single peptides dissolved in DMSO or cultured with 0.002% DMSO only for 26 days. At day 6, brefeldin A (BFA, 1 μg/ml) was added to a fraction of cells for intracellular cytokine/chemokine detection at day 7. At day 13, cells were restimulated with autologous PBMC and 1 μg/ml peptide, and at day 20, 20 IU/ml IL-2 was added to the culture. Cells were harvested at day 26 for TRB sequencing. PBMC were cultured in AIM-V medium supplemented with 10% human AB serum and 1% penicillin/streptomycin at 37 °C.

DNA was isolated from T cells cultured from skin biopsies, T cells sorted from blood, uncultured HPV-induced wart enriched skin tissue and PBMC stimulated with peptides. After resuspension in PBS, DNA from the cultured T cells was isolated using the genomic DNA isolation kit (Sigma–Aldrich), according to manufacturer’s instructions. The sorted blood T-cell subsets were washed in PBS and stored in RLT buffer plus supplemented with 1% β-mercaptoethanol at -80 °C. DNA was isolated from the lysates using the AllPrep DNA/RNA Mini kit (Qiagen, Hilden, Germany), according to manufacturer’s instructions. The HPV-induced wart enriched skin tissue was washed twice with PBS, resuspended in lysis buffer T supplemented with protease K (Sigma–Aldrich) and incubated for 3 h at 55 °C. RNAse and lysis buffer C were added, and the suspension was incubated for 45 min at 70 °C. After centrifugation, DNA from the supernatant was isolated by using the genomic DNA isolation kit according to manufacturer’s instructions. DNA from the peptide-stimulated PBMC was isolated using the QIAamp DNA microkit (Qiagen), according to manufacturer’s instructions. DNA was stored at -80 °C until further processing.

75–100 ng of DNA was amplified through multiplex PCR of TRBV–TRBD–TRBJ gene rearrangements following the BIOMED-2 protocol [10]. Purification of PCR products and library preparation were performed as described previously [11, 12]. Paired-end sequencing was performed using the MiSeq Reagent Kit v2 (2 × 250 bp) on the MiSeq Benchtop Sequencer (Illumina, San Diego, CA, USA). To increase library diversity, PhiX was spiked-in at a 20% concentration. Raw FASTQ files were uploaded to the interactive ARResT/Interrogate immunoprofiler[13] for annotation of the reads and data exploration. QC metrics for the sequencing data are included in Table S3. A mean of 80,607 high-quality sequences was retrieved per sample. Unproductive sequences were removed from further analyses. Additional analyses were performed in R (version 3.5.0, R Foundation for Statistical Computing, Vienna, Austria). For determination of overlap of clonotypes among samples, clonotypes below 50 reads (potential sequencing artifacts) were removed. Shannon’s diversity was calculated using the package Abdiv [14]. The VDJdb database[15] was used, to study the presence of sequences with known antigen specificity. A matching clonotype was defined based on V, CDR3, J, and HLA-type restriction. Although the HLA type of the age-matched healthy control was unknown, the clonotypes of the healthy control that were matching with HLA-A-restricted clonotypes specific for EBV and CMV in the VDJdb database were all restricted to HLA-A*02:01.

Kaluza (v2.1, BC), the OMIQ data science platform (Omiq, Inc., Santa Clara, CA, USA) and R were used for post-acquisition analysis. First, dead cells were excluded and single lymphocytes and monocytes were gated based on forward and side scatter. For spectral cytometry data, unwanted anomalies, based on changes in flow rate and outlier events, were removed by FlowAi [16]. NK cells were further selected as CD3⁻CD56⁺CD45RA⁺ cells. For conventional flow cytometry data, T cells were selected as CD3⁺, and three distinct TCR Vβ⁺ populations and 1 TCR Vβ⁻ population per individual tube were gated and exported as CSV in Kaluza (Figs. S2, S3). Parameters were Arcsinh-transformed (FSC-A and SSC-A were normalized to a 0–5 scale) and exported as FCS, using the FlowCore[17] package in R. The R code is available on GitHub (https://​github.​com/​janinemelsen/​Single-cell-analysis-flow-cytometry) [18]. Individual TCR Vβ populations were proportionally downsampled among tubes to include an equal number of T cells per tube per sample (Table S4). In total, 128 FCS files (4 samples × 8 tubes × 4 populations) were uploaded on the OMIQ platform. The CD4⁻CD3^high T cells were excluded from further analysis, since those T cells are likely to represent the γδ T cells. CD4⁺ and CD4⁻CD8^± were subsampled from each file (Fig. S3) and used as input for opt-SNE [19]. Opt-SNE was performed with default parameters (1000 iterations, 5000 opt-SNE end, 30 perplexity, theta 0.5). Clustering was performed with FlowSOM [20]. The opt-SNE embedding and the FlowSOM clustering were based on the forward scatter, side scatter, and the backbone panel of markers, excluding the TCR Vβ and CD3 expression (Fig. S2). Since the eight tubes were equally distributed among the clusters, we assumed that differences among the tubes were not driving the clustering (Fig. S4). For spectral cytometry data, the complete analysis was performed by using OMIQ, using the same approach except that the γδ parameter was used to exclude the γδ T cells. The inverse Simpson index[21] was calculated by using the package Abdiv [14].

Statistics were calculated in GraphPad Prism software (v9.0.1, GraphPad, La Jolla, CA, USA) by applying a one-way repeated measures ANOVA test. P-values below 0.05 were considered as statistically significant.

At the age of 22 weeks, UPN1 received a hematopoietic stem cell transplantation (HSCT) from his HLA-identical 7-year-old sister, without being conditioned (Fig. 1A). Since early adolescence the patient presented with lower limb lymphedema combined with recurrent skin infections, general malaise, and slowly progressive cutaneous warts mainly affecting the hands caused by human papilloma virus 2 (HPV2). At adult age, the latter appeared refractory to therapeutic interventions (including interferon treatment and surgical removal). Despite engraftment of the transplant, as demonstrated by 12–15% donor chimerism in the peripheral blood mononuclear cells (PBMC) (Fig. 1B), a gradual decline of natural killer (NK) cells and CD45RA⁺ CD4 and CD45RA⁺ CD8 T cells was observed (Fig. 1C). In addition, T-cell receptor excision circle (TREC) content at 32 years post-HSCT, as previously assessed by TREC analysis, was significantly lower in CD4 T cells (0.16 TRECs per µl blood) and CD8 T cells (0.02 TRECs per µl of blood) compared with his donor (39 yr, CD4 6.29 TRECs and CD8 4.87 TRECs per µl blood) [22]. For these reasons, UPN1 received a peripheral stem cell boost (CD34⁺ positively selected cells by CliniMACS®, 6.8*10⁶ CD34⁺/kg and 0.3*10⁵ CD3⁺/kg) from the same donor without prior conditioning, 37 years after HSCT. No graft versus host disease occurred. Following the boost, an increase in donor chimerism in both the PBMC (from 12% in 2005 to 42% in 2016) and granulocytes (from 1% in 2005 to 7% in 2016) was observed, which is in line with donor stem cell engraftment in the bone marrow (Fig. 1B). Moreover, the NK cells, which were absent prior to the boost, reconstituted after the boost (Fig. 1C). The total CD4 and CD8 T-cell numbers remained stable over 50 years, while the naive CD4 and CD8 T cells increased after the boost (Fig. 1C, S5A). At the age of 51, the CD31 expression on CD4 naive T cells, a hallmark for CD4⁺ recent thymic emigrants[23], was 84.8% compared to 81.7% in an age-matched healthy control (both values are above normal range[24]), indicative of a normal thymic output (Fig. 1D). The B cells steadily increased toward normal levels following the boost, while immunoglobulin levels remained unaffected, in the continued absence of immunoglobulin replacement therapy (Fig. S5A-B). The B-cell donor chimerism of 29%, 49% and 93% in the naive, unswitched memory and switched memory B cells, respectively, after the boost, indicates emergence of functional donor-derived B cells (Fig. S5C). The overall clinical condition significantly improved with less lymphedema and skin infections after the boost. Still, the HPV2 warts persisted although transient expansion and regression of the lesions were observed pointing to a certain level of HPV2 warts-directed immunity. Altogether, the clinical improvement and the increase in donor chimerism as well as the increasing NK cell and naive T-cell numbers after the boost underscore the effectiveness of the stem cell boost. To study in-depth the dynamics and longevity of the circulating NK cells and T cells isolated from blood and skin, we performed phenotypical, functional and repertoire analyses at multiple timepoints, up to 51 years post-HSCT (Fig. 1A,E).

While the initial HSCT did not lead to reconstitution of NK cells, the peripheral stem cell boost did induce NK cell reconstitution. However, we wondered whether the NK cells of UPN1 were phenotypically and functionally comparable to NK cells of healthy control. Fourteen years after the boost, a low frequency of CD56^brightCD16^± NK cells was observed in UPN1 (1.2%) compared with healthy controls (median 6.2%, range 1.8–13.6%, Fig. 2A). Interestingly, CD27, a marker that is usually expressed by only a minor fraction of CD56^dimCD16⁺ NK cells (mean 1.1% controls), was expressed by 27.2% of the CD56^dimCD16⁺ NK cells of UPN1 (Fig. 2B). This CD56^dimCD27⁺ subset was further characterized by higher expression of KLRG1, TIGIT, NKG2A, CD69, CD62L, NKG2D and NKp46 and lower expression of CD57, DNAM1, CX3CR1, KIRs and CD16, compared with the CD56^dimCD27⁻ NK cells (Fig. 2C,D). The fact that cells with exactly this phenotype also exist in low numbers in healthy controls suggests that the CD56^dimCD27⁺ NK cells are preferentially expanded in UPN1. To study the functional capacity of the NK cells, we measured degranulation, as well as the cytokine and chemokine response upon CD16 crosslinking and/or K562 stimulation. Although some differences in response were observed between the CD27⁻ and CD27⁺ CD56^dim NK cells, both subsets were capable of degranulation and production of CCL4, XCL1 and IFN-γ in all samples (Fig. 2E). In conclusion, the peripheral blood NK cells reconstituted by the boost were 14 years after the boost functional and enriched for a CD56^dimCD27⁺ NK cell population.

Next, we studied the dynamics of the T-cell compartment in PBMC isolated prior to the boost, 4 months after the boost (2005), and 13 years after the boost (2018) by flow cytometry (Fig. S2). The panel consisted of eight tubes, each including markers specific for T-cell lineage (CD3, CD4, CD8), differentiation (CCR7, CD45RA, CD27, CD28), activation or exhaustion (TIGIT, PD-1) and receptor repertoire (3 TCR Vβ antibodies per tube, Fig. S2). As a control, PBMC isolated from the donor around the time of the boost, at the age of 40, were included. Within the blood CD4 T-cell compartment, nine clusters were assigned and visualized by opt-SNE (Fig. 3A). Based on the phenotype, clusters were subdivided into three groups: CCR7⁺CD45RA⁺CD27⁺CD28⁺ naive T cells (cluster 1), CD27⁺CD28⁺ early memory T cells (cluster 5,8) and CD27⁻ and/or CD28⁻ late memory T cells (cluster 2–4,6,7,9) (Fig. 3A,B). After the stem cell boost, there was a gradual increase of absolute numbers of naive T cells (Fig. 1C), while the frequency of naive T cells was not altered in the long term (11.3% pre-boost versus 11.2% post-boost 2018, Fig. 3B). In contrast, memory CD4 T cells that had lost CD27 and/or CD28 were nearly absent in the donor (5.7% of total CD4, sum of clusters 2–4, 6,7, 9), while they were abundant before and at the two time points after the stem cell boost in UPN1 (49.8%, 60.4%, 39.0% of total CD4, respectively, Fig. 3B).

After identifying the different subsets by clustering, we projected the TCR Vβ expression on the clusters and calculated the distribution of the TCR Vβs in the total CD4 T cells and within each cluster (Fig. 3C). Apart from an enrichment of TCR Vβ4 in UPN1, the TCR Vβ frequencies within the total CD4 T cells of the patient and donor were within normal or slightly below ranges of the reported reference values as measured in at least 46 healthy controls (Fig. S6A). When calculating the frequency of each TCR Vβ family within each CD4 T-cell subset, it became evident that in the naive and CD27⁺CD28⁺ early memory CD4 T cells, but not in the CD27⁻ and/or CD28⁻ memory population, each TCR Vβ family is present in UPN1 at a constant frequency over time and resembles the frequency in the donor (Fig. 3D). Strikingly, neither in the patient nor in the donor differences in the TCR Vβ family distribution were observed between the CD27⁺CD28⁺ memory and naive CD4 T cells (Fig. 3D). Therefore, the TCR Vβ repertoire of both the naive and CD27⁺CD28⁺ memory CD4 T cells is diverse and comparable to the healthy donor.

Within the memory CD4 T cells that lost CD27 and/or CD28, a highly restricted TCR Vβ repertoire was detected in the patient, as indicated by, for instance, the high frequency of TCR Vβ2 and Vβ4 and the low inverse Simpson index (Fig. 3D). In contrast, in the healthy donor the inverse Simpson index of these cells was not reduced compared with the naive cells (Fig. 3B,D). We therefore zoomed in on the CD27⁻ and/or CD28⁻ memory T cells, generated an opt-SNE and subclustered the data (Fig. 4A). Only one, two or three dominant TCR Vβs were observed within all the CD28⁻ subclusters from the patient, which is indicative of an oligoclonal expansion (Fig. 4A,B). To quantitatively determine which TCR Vβs are enriched, we calculated the frequency of each TCR Vβ within each cluster and compared the frequencies to the Vβ distribution of the naive CD4 T cells of the UPN1 post-boost 2018 sample (Fig. 3A-C). The clusters that did not cluster with the naive CD4 T cells in the heatmap were considered an oligoclonal expansion (Fig. 4B). For instance, in cluster 5, TCR Vβ13.2 was almost completely dominant. In this cluster, 10% of all the cells were represented by TCR Vβ13.2⁺ cells (the maximum is 12.5% since for each TCR Vβ⁺ cell, 7 Vβ⁻ cells from the other seven tubes are included in the cluster). Although CD4 memory T-cell differentiation is assumed to be associated with a loss of CD27, followed by a loss of CD28, we identified two CD28⁻ clusters (1 and 5) still expressing CD27 in which one TCR Vβ was dominant (Vβ18 and Vβ13.2, respectively). Notably, CD28⁻ CD4 T cells were barely present in the healthy donor (Fig. 3B, 4A), explaining the high diversity index of the CD27⁻ and/or CD28⁻ memory CD4 T cells. Together, these data indicate that the abundantly present CD28⁻ CD4 T cells have a skewed Vβ repertoire and are not derived from an adoptively transferred antigen-experienced CD28⁻ T-cell pool of the donor.

Interestingly, each cluster containing a dominant TCR Vβ family in UPN1 was characterized by a consistent expression pattern of TIGIT and PD-1. For instance, all the TCR Vβ13.2⁺ CD4 memory T cells in cluster 5 were positive for TIGIT and negative for PD-1, while the TCR Vβ4⁺ cells in clusters 8, 10, 22, 2, 21, 26, 11, 12, 16 and 18 expressed neither TIGIT nor PD-1 (Figs. 4A-B, S7). In contrast, some subclusters with TCR Vβ4 (11, 12, 16, 18), Vβ2 (9) and TCR Vβ17 (7, 17) dominance expressed a gradient of CD45RA, suggesting the presence of a differentiation gradient (Figs. 4A-B, S7). Similarly, a gradient of CD8 expression was observed on a fraction of all TCR Vβ dominant subclusters (5, 22, 21, 26, 13, 1, 12, 16), except for the TCR Vβ17⁺ family, ranging from 38% of TCR Vβ2⁺ cells to 92% of TCR Vβ13.2⁺ cells (Figs. 4A-B, S7). Therefore, it is important to consider CD4⁺CD8^dim T cells as conventional CD4 T cells, to prevent underestimation of the oligoclonal compartment.

Next, we evaluated the persistence of each TCR Vβ dominant expansion by determining the frequency in 2005 (pre-boost, 4 months post-boost) and 2018. The frequency of the clustered TCR Vβ8- and Vβ2-expressing cells was the highest in the pre- and post-boost 2005 samples and was decreased to a low level in 2018. The clustered TCR Vβ13.2⁺ and Vβ17⁺ cells were only present in the 2018 sample (Fig. 4C). The clustered TCR Vβ4⁺ cells represented more than 30% of the CD27⁻ and/or CD28⁻ memory CD4 T cells in 2018 and were present already in 2005 (Fig. 4C). The clustered TCR Vβ18⁺ and Vβ13.6⁺ cells were continuously detected at a level below 2% of the CD27⁻ and/or CD28⁻ memory CD4 T cells (Fig. 4C). Taken together, UPN1 has a dynamic T-cell compartment, characterized by CD28⁻ memory CD4 T-cell expansions, of which each has a dominant TCR Vβ expression and a unique surface marker phenotype.

To longitudinally study the CD8 and CD4⁻CD8⁻ compartment up to 50 years post-HSCT, we applied the same analysis approach and identified by clustering the naive CD8 T cells (cluster 10), CD27⁺CD28⁺ early memory CD8 T cells (cluster 1), CD27⁻ and/or CD28⁻ late memory CD8 T cells (cluster 2, 3, 6- 9), naive CD4⁻CD8⁻ T cells (cluster 5) and memory CD4⁻CD8⁻ T cells (cluster 4) (Fig. 5A,B). As for the CD4 T cells, the proportion of naive CD8 T cells was decreased, while the CD27⁻ and/or CD28⁻ memory CD8 T cells were enriched in UPN1 compared with his healthy donor. CD4⁻CD8⁻ T cells were present at a low frequency in both UPN1 and the healthy donor, varying from 1.7–4.7% (Fig. 5B).

Next, the TCR Vβ distribution was calculated among the total CD8 T cells (Fig. 5C). Except for the high frequency of TCR Vβ12 (18.3% sample 1) and Vβ5.2 (4.8% sample 3), the TCR Vβ frequencies of UPN1 and his donor were within the normal range for each Vβ (Fig. S6B). When zooming in on the naive CD8 T cells of UPN1, it became evident that the distribution of the TCR Vβs was solely comparable to the healthy donor 13 years after the stem cell boost, while the diversity index remained stable over time (Fig. 5D,E). Of note, the input naive CD8 TCR Vβ⁺ cells of the first two patient samples (39 and 164 cells) were fairly low compared with the naive CD4 TCR Vβ⁺ cells (965 and 1365 cells), probably resulting in a less accurate assessment of the TCR Vβ repertoire. The TCR Vβ distribution and diversity index of the CD27⁺CD28⁺ memory T cells were fairly similar between the donor and the patient in all samples, with the higher frequency of TCR Vβ22 in the UPN1 post-boost 2005 sample as exception (Fig. 5D,E). In the donor, the TCR Vβ distribution of the CD27⁺CD28⁺ memory T cells was reminiscent of the distribution of the naive T cells. Given the fact that the TCR Vβ distribution of the CD27⁺CD28⁺ memory T cells mirrors the naive T cells, as also shown in the CD4 population, we speculate that the TCR Vβ distribution of the naive CD8 T cells in the patient was probably before the boost already similar to the donor and therefore not altered by the boost.

The CD27⁻ and/or CD28⁻ memory CD8 T cells, analyzed separately, were characterized by a skewed TCR Vβ repertoire. The highest frequency observed was attributed to TCR Vβ12 in the pre-boost sample, which represented 22.3% of the CD27⁻ and/or CD28⁻ memory CD8 T cells (Fig. 5D). Half of the CD27⁻ and/or CD28⁻ memory CD8 T-cell clusters were considered an oligoclonal expansion, since they grouped separately from the naive CD8 T cells of the patient (Fig. 6B). In these clusters, multiple dominant TCR Vβ families were found, indicating an overlapping phenotype among clonally distinct T-cell populations (Fig. 6A,B). Each TCR Vβ family had a homogeneous phenotype concerning TIGIT and PD-1, while the CD45RA expression varied within one family. For instance, TCR Vβ12⁺ cells were enriched in CD45RA⁻ clusters 39, 32 and CD45RA⁺ cluster 17, but were all TIGIT⁻PD-1⁺ (Figs. 6B, S8). TCR Vβ8⁺ cells were enriched in CD45RA^dim clusters 16, 23 and CD45RA^high clusters 9, 5, 6, 11, 4 and 10, but overall had a TIGIT⁺PD-1⁻ phenotype. In contrast, within the TCR Vβ22⁺-enriched clusters two phenotypes were found, TIGIT⁻PD-1^high (cluster 8) and TIGIT⁺PD-1^low (cluster 28, 36, 38) (Fig. S8). This could indicate the presence of two clonally distinct CD8 T-cell expansions which express the same TCR Vβ. Within the CD4⁻CD8⁻ memory T cells (cluster 2, 3), a gradient expression of CD27 and CD28 was observed, and no dominant TCR Vβ families were identified.

When studying the kinetics, it became evident that also multiple dominant TCR Vβ families in the CD8 memory pool (Vβ12, Vβ17, Vβ14) persisted for at least 13 years, albeit at different frequencies Fig. 6C). One expansion was potentially transferred with the boost, since the clustered TCR Vβ3⁺ T cells were present in the donor at the same frequency as in the 2018 sample in the patient (3.5%, Fig. 6C). Notably, we found only one donor-specific dominant TCR Vβ family (Vβ11, part of cluster 31 and 33), representing 2.5% of CD27⁻ and/or CD28⁻ memory CD8 T cells (Fig. 6A,C). Taken together, the CD27⁻ and/or CD28⁻ memory CD8 T-cell population of UPN1 is characterized by the presence of multiple dominant TCR Vβ families, indicative of oligoclonal expansions.

To confirm our interpretations about the diversity of the naive T cells and to provide evidence that the cells expressing one of the dominant TCR Vβ families in the CD4 and CD8 CD27⁻ and/or CD28⁻ memory T-cell clusters are oligoclonal, we performed next-generation sequencing of the T-cell receptor β-chain. We sorted naive and CD27⁻ and/or CD28⁻ memory T-cell populations from blood (2019) from UPN1 and an age-matched healthy control (Fig. S1). Of note, in contrast to the donor of UPN1, the contribution of CD28⁻ cells to the CD27⁻ and/or CD28⁻ memory T cells of the age-matched control was comparable to UPN1 (Fig. S1A-C). The Shannon diversity score of the naive and memory T-cell populations was equal between UPN1 and the healthy control (Fig. 7A). The CDR3 length distribution plots with the top 50 most frequent clonotypes colored revealed that within the memory population of UPN1 and the healthy control some clonotypes represented 15–20% of all sequences (Fig. 7A). Altogether, these findings confirm that UPN1, similar to the healthy control, has a diverse naive and restricted memory CD4 and CD8 TCR repertoire.

Six out of seven dominant TCR Vβ families (Vβ17, Vβ18, Vβ2, Vβ8, Vβ4 and Vβ13.6) previously identified by flow cytometry were included in the top ten most frequent clonotypes in CD4 memory T cells, confirming that the earlier discussed Vβ dominant clusters are mostly clonal T-cell expansions (Fig. 7B). In the CD8 memory T cells, seven of the nine dominant TCR Vβ families (Vβ17, Vβ3, Vβ5.2, Vβ14, Vβ12, Vβ8 and Vβ22) were matched to the top 15 clonotypes (Fig. 7B). We repeated the TCR Vβ analysis in a patient sample (2019), by using spectral cytometry to provide a more detailed phenotype of the clonal T-cell expansions. Seven dominant CD4 and seven dominant CD8 TCR Vβs were still present with the same TIGIT and PD-1 phenotype as earlier described (Fig. 7C, D). Multiple CD4 and CD8 clusters with only TCR Vβ’neg’ cells were identified, which is in line with the occurrence of clonotypes in the top 10 and top 15 that are not detected by the TCR Vβ antibodies present in the cytometry panel.

Although UPN1 and his donor are seronegative for cytomegalovirus (CMV), the majority of the CD4 and CD8 clonal expansions in the patient expressed CD57, KLRG1 and CX3CR1, a phenotype which is associated with CMV infection (Fig. 7C, D). To validate this, we checked the CD4 and CD8 memory sequences of UPN1 and the age-matched healthy control in the VDJdb database which contains TCR sequences of known antigen specificity such as CMV and EBV. None of the clonotypes matched the clonotypes in the database, indicating that they are not associated with a currently known antigen specificity. In contrast, in the healthy control multiple memory clonotypes matched clonotypes specific for CMV and EBV, explaining the abundance of CD27⁻ and/or CD28⁻ memory T cells (Fig. S1, data not shown).

Since the clonotypes of UPN1 could not be linked to CMV, we wondered whether another chronic virus caused the phenotypic imprinting on the CD4 and CD8 memory T-cell compartment of the patient. Since the patient suffers from chronic HPV2-induced cutaneous warts, we aimed to study the link with HPV. First, the T-cell receptor β-chains of cultured T cells from wart⁺, wart^± and wart⁻ skin and of uncultured T cells from wart⁺ skin were sequenced. Overall, a skewed repertoire was found, also in the uncultured T cells of which the diversity index was comparable to the CD27⁻ and/or CD28⁻ memory CD4 T cells (Figs. 7A, 8A). We reasoned that if the observed expansions of clonotypes in blood would be related to recognition of HPV, there should be an overlap in TRB sequences between blood and affected skin. The overlap in sequences between blood and skin revealed that a few clonotypes observed in the memory pool were present in wart⁺ skin (Fig. 8B). In total, 17 clonotypes of the 180 clonotypes (with a cut-off value of 50 reads) of the CD4 CD27⁻ and/or CD28⁻ memory T cells in blood were observed in cultured and/or uncultured T cells from wart⁺ skin (Fig. 8B, red rectangles). Four out of these 17 were included in the top 50 most frequent CD4 clonotypes in blood; however, none of them matched the major clonal expansions observed by flow cytometry (Figs. 7C, 8C). In contrast, all the five overlapping CD8 memory clonotypes were present in the top 50 clonotypes of the blood memory CD8 T cells, of which one was detected by flow cytometry as well (TCR Vβ17, Fig. 8B, purple rectangles, 8C, 7D). The most frequent CD8 clonotype (representing 20.6% in blood) might match one of the CD27⁻CD28⁻ clusters, not recognized by any of the TCR Vβ antibodies in the panel (Figs. 7D, 8C). The frequency of the overlapping CD8 clonotypes in wart⁺ skin tissue was fairly low, ranging from 0.13% to 2.07% of all sequences. In line with this, the CD8 T cells, usually representing the vast minority of T cells in healthy skin[25], also had a low frequency in cultured (5.9%) and uncultured T cells (8.0%) from wart⁺ skin (Fig. 8D). To study whether the clonal expansions are specific for HPV2, we cultured PBMC from UPN1 in the presence of single peptides derived from the L1 protein of HPV2 (Table S2). After 7 days of stimulation, two peptides induced CCL4 and TNF-α production in CD8 T cells, suggesting the presence of HPV2-directed immunity (Fig. S9A). However, by TRB sequencing after an additional 19 days of culture, we did not observe expansion of the highly frequent clonotypes observed in blood (Fig. S9B). Thus, although we could not confirm the specificity of the major clonal expansions using our selection of HPV2-derived peptides, the presence of clonotypes from the circulating CD8 CD27⁻ and/or CD28⁻ memory pool in HPV-induced warts suggests a link with HPV.