Animal experimental procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of San Raffaele Hospital and Italian Ministry of Health. C57Bl/6 wild-type (WT) mice were purchased from Charles River Laboratories (USA). The Rag1^F971L/F971L and the Rag1^R972Q/R972Q mice were previously described (15) and were bred and maintained in the Ospedale San Raffaele (OSR) animal facility (IACUC n. 877, project prot.6EEAF.48, ministry authorization n. 113/2018-PR and IACUC n. 1125, project prot.6EEAF.164, ministry authorization n.5/2021-PR).

Femurs and tibiae of 6- to 12-week-old donor mice were flushed and passed through a 0.4-μm cell strainer. Lineage-negative (Lin-) cells were obtained using the Lineage Cell Depletion kit (Miltenyi Biotec), following manufacturer’s instructions. Lin- cells were transduced as previously described (10) with the pre-clinical GMP batch of the MND-c.o.RAG1 vector. Untransduced Lin- cells from mutant and WT mice were cultured in parallel.

After transduction, cells were divided for in vivo transplantation, liquid culture for VCN determination and colony forming unit (CFU) assay. For in vitro liquid culture, cells were kept in culture in stimulation medium (10) for 10 days and then collected to extract DNA.

For CFU assay, cells were plated in MethoCult GF M3434 (StemCell Technologies) medium at a density of 1000 cells/ml and cultured for 10-12 days. Single cell colonies were counted and picked for VCN determination. In addition, remaining bulk cultures were collected for DNA extraction and VCN determination.

Six- to 10-week-old recipient mice were conditioned with lethal total-body irradiation (TBI, 8 Gy, split dose) at least 2 hours before transplantation and were then injected in the caudal vein with 0.5-1x10⁶ Lin- cells. GT mice received transduced Rag1^F971L/F971L or Rag1^R972Q/R972Q Lin- cells, while BMT WT and BMT UT mice were transplanted with untransduced WT or mutant Rag1 cells, respectively. Gentamicin sulfate (Italfarmaco, Milan, Italy) was administered in drinking water (8 μg/mL) for the first 2 weeks after transplantation to prevent infections.

DNA extraction from liquid cultures, tissue single cell suspensions and bulk CFU cultures was performed with QIAamp DNA micro or mini kits (QIAGEN), following manufacturer’s instructions. DNA extraction from single CFU colonies was performed using QuickExtract DNA extraction Solution, following manufacturer’s instructions. Vector copy number was quantified by digital droplet PCR as previously described (13).

Single-cell suspensions from tissues were obtained smashing tissues through a 0.4-μm cell strainer. Red blood cells of PB were lysed to obtain WBC. Single-cell suspensions and WBC were stained for 15 min at room temperature with the following antibodies from BD Pharmingen, Miltenyi Biotec, BioLegend or eBioscience: CD3, CD4, CD8, CD11b, CD19, CD21, CD23, CD24, CD25, CD43, CD44, CD45.1, CD45.2, CD48, CD62 ligand, CD69, CD117, CD150, B220, IgM, IgD, NKp46, NK1.1, Lin+ cocktail, and Sca1. Viability was determined by using the Live/Dead Fixable Dead Cell Stain Kit (Thermo Fisher Scientific). Samples were acquired on a FACSCanto II (BD Biosciences) and analyzed with FlowJo software.

Four months after GT, mice were injected intravenously with 100 μg of TNP-KLH (Biosearch Technologies), to elicit a T-dependent response. Three weeks later, animals were injected intraperitoneally (i.p.) with 100 μg of TNP-KLH to boost the immune response. Serum was collected weekly. Antigen-specific IgM and IgG antibodies were evaluated in serum by ELISA, as previously described (13).

Alternatively, mice were challenged with 50 μg of TNP-Ficoll (Biosearch Technologies) by i.p. injection and serum was collected after 2 weeks. Antigen-specific IgG3 antibodies were evaluated in serum by ELISA, as previously described (15).

Genomic DNA was extracted with the QIAamp DNA Blood Mini Kit (Qiagen, Hilden, Germany), according to the manufacturer’s instructions. Vector copy number/genome (VCN) was quantified on 10 ng of genomic DNA using the QX200 Droplet Digital PCR System (Bio-Rad Laboratories), as previously described (13).

Total IgA, IgG, and IgM levels were quantified on serum or plasma samples obtained at termination by using the MILLIPLEX MAP Mouse Immunoglobulin Isotyping Magnetic Bead Panel and MAGPIX System (Merck Millipore), according to the manufacturer’s instructions.

B cell–activating factor (BAFF) levels were analyzed with the mouse BAFF Quantikine ELISA (R&D Systems), according to the manufacturer’s instructions.

Protein arrays were used to screen for a broad panel of IgG autoantibodies (University of Texas Southwestern Medical Center, Genomic and Microarray Core Facility). Serum obtained 6 or 12 months after transplant was used. The raw signals were background subtracted and normalized to generate Normalized Signal Intensities (NSI). The matrix of NSI data was used to create heatmaps, where each value represents the fold-change on the log2 scale, using the row-wise average of BMT WT as a reference.

Total RNA was obtained from spleen cells with the RNeasy Mini Kit (Qiagen), according to the manufacturer’s instructions. TCR beta chain clonotypes were identified by RACE-PCR (rapid amplification of cDNA ends-PCR) followed by deep sequencing (ProtaGene). Briefly, between 6.6ng and 720ng RNA input were used for cDNA synthesis using a TCR constant region specific primer followed by the addition of an adaptor oligo molecule at the 5’cDNA end. cDNA was AMPure XP purified and applied to two subsequent nested PCR reactions using primers binding to the beta chain constant region and the introduced adaptor. In the second amplification step, fusion primers enabling sequencing on Illumina MiSeq platform were used and PCR products were purified using AMPure XP beads. Finally, products were pooled and sequenced in asymmetric 400bp+100bp paired-end mode.

Sequencing data were analyzed using MiGEC tool (https://migec.readthedocs.io/en/latest) in order to sort the sequences by the sample specific barcodes introduced during library preparation followed by clustering based on unique molecular identifiers. Sorted sequences were then analyzed with MiXCR (https://mixcr.readthedocs.io/en/master/index.html) for the identification of the TCR beta clonotypes and their respective relative frequencies. The top 10 clonotype frequencies were taken into account to receive an estimation on the clonality and the TCR beta repertoire of the samples. Further downstream analyses were performed using VDJtools (https://vdjtools-doc.readthedocs.io/en/master/) that allowed the evaluation of V- and J-gene usage of the single clonotypes and the display of the gene recombination, as well as CDR3 sequence length distribution.

Standard S-EPTS/LM-PCR (shearing extension primer tag selection ligation-mediated PCR) and deep sequencing were performed to identify lentiviral vector flanking genomic sequences (ProtaGene) (16–18). Inputs between 16ng and 400ng of genomic DNA were used.

Briefly, input genomic DNA was sheared to a median length of 400 to 500 bp using the Covaris M220 instrument. Sheared DNA was purified, and primer extension was performed using a long terminal repeat (LTR) specific biotinylated primer. The extension product was again purified, followed by magnetic capture of the biotinylated DNA for at least 60 minutes and two washing steps with 100 µl H₂O. The captured DNA was ligated to linker cassettes including a sample barcode. The ligation product was divided into 2 aliquots and amplified in a first exponential PCR using biotinylated vector- and sequencing adaptor-specific primers. Biotinylated PCR-products were magnetically captured, products were pooled, washed and 1/2 of this eluate served as template for amplification in a second exponential PCR step with primers allowing deep sequencing by MiSeq technology (Illumina) after purification. Preparation for deep sequencing was previously described (17, 18). DNA double barcoding was applied to allow parallel sequencing of multiple samples in a single sequencing run while minimizing sample cross-contamination.

IS were recovered using the S-EPTS/LM-PCR protocol (16) and analyzed using the GENE-IS tool suite (19). Briefly, raw sequence data were trimmed according to sequence quality (Phred 30). Only sequences showing complete identity in both sample barcodes, linker cassette barcode and sequencing barcodes, were further analyzed.

Modern sequencing technologies like Illumina MiSeq allow for semi-quantitative estimation of clonal size by the determination (‘counting’) of the number of retrieved sequences (retrieval frequency) for individual vector-genome junctions (ISs). The relative sequence count of all detected ISs was calculated in relation to all sequences which could be mapped to a definite position in the genome. The ten most prominent ISs (Rank 1: highest sequence count; Rank 2: 2^nd highest sequence count; Rank 10: 10^th highest sequence count) were analyzed for each sample.

Thymus samples were formalin-fixed and paraffin-embedded. Sections (1.5-μm) were stained with hematoxylin and eosin, cytokeratin 5 or CD3 antibodies. Digital images were acquired with an Olympus DP70 camera mounted on an Olympus BX60 microscope by using CellF Imaging software (Soft Imaging System GmbH, Munster, Germany). Morphometric analysis of the medullary-to-cortical ratio was evaluated by using Image-Pro software.

Statistical analyses were performed with GraphPad Prism software (GraphPad Software) by using Spearman correlation or nonparametric 1-way ANOVA Kruskal-Wallis test. Dunn’s multiple comparisons test was performed after ANOVA, comparing GT groups to the correspondent mutant controls and BMT WT mice, unless otherwise stated. Statistical differences between autoantibody profiles were assessed by global test (20) using the GlobalAncova R package. For all tests, a p value of less than 0.05 was considered significant. Further details on significance levels are provided in figure legends.

Lentiviral vector (LV) gene therapy has been shown to obtain effective RAG1 expression and immune reconstitution in Rag1^-/- mice using a MND-c.o.RAG1 vector (10). We extensively tested the same LV in the setting of 2 mouse models carrying hypomorphic Rag1 mutations and recapitulating the CID phenotype, the Rag1^F971L/F971L and the Rag1^R972Q/R972Q mice (15). To this purpose, we isolated Lineage-negative (Lin-) cells from the bone marrow (BM) of donor animals and transduced them with the MND-c.o.RAG1 LV. Transduced Lin- cells showed clonogenic potential similar to untreated Lin- cells (Supplementary Figure S1A) indicating that LV transduction does not affect stemness potential. Vector copy number/genome (VCN) evaluated in transduced bulk population ranged from 1.2 to 4.0 (Supplementary Figure S1B), while VCN in single colony forming units had a median of 2.1 in Rag1^F971L/F971L and 1.4 in Rag1^R972Q/R972Q gene therapy (GT) cells (Supplementary Figure S1C), with a mean of transduced colonies of 87.1% and 76.8%, respectively (Supplementary Figure S1D).

For the in vivo experiments, GT-treated animals received transduced cells after lethal TBI. A total of 23 Rag1^F971L/F971L and 29 Rag1^R972Q/R972Q mice received the corresponding GT cells, while controls were transplanted with wild-type (WT) Lin- cells (BMT WT) or untransduced mutant Lin- cells (BMT UT). Four experiments were terminated 6 months after transplant and two experiments 12 months after transplant. The survival curve of 6-month experiments showed 84% survival in Rag1^F971L/F971L GT and 90% in Rag1^R972Q/R972Q GT mice. Early loss (2 weeks after transplant) of BMT UT Rag1^R972Q/R972Q mice was observed, probably due to lack of engraftment of transplanted cells (Supplementary Figure S2A). In experiments terminated at 12 months, we experienced long-term loss of BMT UT animals in both mutant strains, probably due to disease progression exacerbated by the irradiation procedure. Survival of 100% Rag1^F971L/F971L GT and 90% Rag1^R972Q/R972Q GT mice was observed, in line with the experiments terminated 6 months post-treatment (Supplementary Figure S2B).

In the peripheral blood (PB), analysis of immune reconstitution over time showed a partial increase in the percentage of B cells in Rag1^F971L/F971L GT mice compared to Rag1^F971L/F971L untreated control and BMT UT mice. However, the relative frequency after GT remains significantly lower than that achieved in Rag1^F971L/F971L BMT WT mice, up to 22 weeks post treatment (Figure 1A). On the other hand, the low frequency of T cells found in untreated control mice did not improve after GT and remained significantly lower than mice transplanted with WT cells (Figure 1B). At later time points, no statistically significant differences were found in the frequency of B and T cells, probably due to the myeloid skewing of aging mice and the lower number of mice tested starting from 35 weeks post-GT. Similar results were found in the relative frequency of B and T cells of Rag1^R972Q/R972Q GT mice (Figures 1A, B). Concomitantly, GT mice from both strains had significantly higher proportion of myeloid cells at earlier time points (Supplementary Figure S3A). However, B, T and total white blood cell absolute counts remained comparable to untreated Rag1 mutated controls and significantly lower than WT controls and BMT WT mice (Figures 1C, D and Supplementary Figure S3E). No major changes were observed in myeloid and natural killer (NK) absolute counts (Supplementary Figures S3B-D).

At euthanasia 6 months after treatment, we assessed the immune reconstitution and VCN in hematopoietic organs. B cell lymphopoiesis in bone marrow (BM) revealed the appearance of immature B cells and mature recirculating B cells in both hypomorphic mutants after GT, although absolute counts remained lower than BMT WT mice (Figure 2A and Supplementary Figure S4). In the Rag1^R972Q/R972Q GT mice, statistically significant higher counts of immature and mature recirculating B cells than untreated control mice were observed, whereas no significant differences were found in Rag1^F971L/F971L GT when compared to both untreated and BMT WT groups (Supplementary Figures S4F-J). Mean VCN in BM was 2.0 and 2.8 in Rag1^F971L/F971L and Rag1^R972Q/R972Q GT mice, respectively (Figure 2B). In the spleen, some improvements of B and T cell counts were observed in GT mice, but no statistically significant differences were found between GT groups and BMT WT or untreated control mice (Figure 2C and Supplementary Figures S5A-D). GT induced a modest increase in the proportion of follicular B cells, especially in the Rag1^R972Q/R972Q model, and a slight decrease of marginal zone B cells, which were elevated in untreated or BMT UT mutant mice (Figure 2D). No significant differences were found, except for Rag1^R972Q/R972Q GT mice that had increased proportion of marginal zone B cells when compared to BMT WT counterpart (Supplementary Figures S5E-H). In addition, we observed a trend to increase the white pulp area after GT, in line with that of BMT WT controls (Supplementary Figure S5I). To assess the functionality of B cells we quantified immunoglobulin (Ig) levels in the plasma. IgM serum levels were increased in a fraction of mutant mouse, but no statistically significant differences were detected (Figure 2E). Similarly, no major differences were observed in IgA and IgG levels in the Rag1^F971L/F971L strain, while in the Rag1^R972Q/R972Q GT mice a statistically significant increase of IgA after GT was observed (Supplementary Figures S5J, K).

In the thymus, normalization of total thymocyte counts was observed after GT (Figure 3A). GT-treated mice displayed partial correction of the block at double negative (DN) 3 stage, especially in the Rag1^R972Q/R972Q GT mice (Figure 3B and Supplementary Figures S6A-D). VCN was variable and ranged from 0.01 to 12.3, although comparable in the 2 mutant GT groups (Figure 3C). Histological and immunohistochemistry evaluation confirmed partial thymic immune reconstitution, with modest improvement in the cortical/medullary ratio of some GT mice (Figure 3D and Supplementary Figures S7A, B). However, statistically significant correlation with the VCN of thymocytes was found only in the Rag1^R972Q/R972Q strain (Supplementary Figure S7C). In the periphery, phenotypic analysis of splenic T cells demonstrated no significant changes in the proportion of naïve, effector and memory T cells in GT treated mice as compared to BMT UT mice, as shown by persistence of an elevated fraction of effector T cells and a reduced proportion of naïve T lymphocytes (Figure 3E and Supplementary Figures S6E-G). We assessed the T cell receptor (TCR) clonality from the splenocytes of treated mice, 6 months post-treatment. In Rag1^F971L/F971L strain, the lowest number of clonotypes and diversity values could be seen in the untreated control mice, whereas mice from the BMT WT group showed the highest V- and J-gene recombination. In the GT group, variable recombination was observed ranging from low to high TCR clonality. In Rag1^R972Q/R972Q mice, similar results were obtained, although BMT WT mice showed an oligoclonal repertoire. One GT mouse showed low V/J recombination, while 2 out of 3 GT mice had high diversity values (Figure 3F).

To understand if this incomplete immune reconstitution, compared to mice treated with WT stem cells, would suffice to induce adaptive immune responses, we assessed antibody responses upon challenge with T-dependent (TNP-KLH) and T-independent (TNP-Ficoll) antigens. The TNP-KLH-specific response was evaluated by specific IgG production after the second boosting dose. In Rag1^F971L/F971L mice, we did not detect differences in TNP-KLH-specific IgG serum levels among the various experimental groups (Figure 4A). In Rag1^R972Q/R972Q treated mice, the improvement in specific IgG response was comparable to WT or BMT WT animals, although no significant differences were found, also due to the variability of the assay (Figure 4B). In parallel, another cohort of animals were immunized with the T-independent antigen. Levels of TNP-specific IgG₃ were similar in GT mice, and WT controls (Figures 4C, D).

Since incomplete or ectopic RAG1 expression could possibly lead to autoimmune manifestations typical of the CID-G/AI phenotype, we assessed the production of autoantibodies in the serum. Autoantibodies were tested at 6 months post-transplant, by screening a panel of 123 autoantigens using a high-throughput autoantigen microarray platform. No increase of IgG autoantibodies was observed after GT, when values were normalized to the correspondent BMT WT controls. Statistical analysis, despite the low number of analyzed mice, showed a statistical difference between Rag1^F971L/F971L controls and GT animals, while no differences were found between GT and BMT WT groups of both strains (Figure 5A). We also measured the levels of the B cell-activating factor (BAFF), a survival and maturation factor for B cells, which is usually upregulated in lymphopenic conditions and autoimmune diseases (21). We observed high BAFF levels in untreated mutant mice, as well as in BMT UT animals, as expected. We detected lower BAFF levels in most Rag1^F971L/F971L and Rag1^R972Q/R972Q GT mice (Figure 5B). To exclude late onset of autoimmunity we also measured autoantibody production and BAFF levels 12 months after transplant. We did not observe the increase in autoantibody production in neither Rag1^F971L/F971L nor Rag1^R972Q/R972Q GT mice (Figure 5C), in line with those obtained at 6 months post GT. No significant differences between groups were found in BAFF levels (Figure 5D), probably due to the general reduction of B cells and the concomitant myeloid skewing in the BM that is expected in aged mice (Supplementary Figure S8). It is worth mentioning that mice living in specific pathogen-free condition are not regularly exposed to new antigens and may not efficiently predict the onset rate of autoimmunity in humans.

At 12 months post GT, we confirmed a partial overcoming of the DN3 block in thymocytes of GT animals (Supplementary Figure S9). In the spleen, comparable B and T cell counts were observed in GT animals when compared to BMT WT mice, but also to untreated Rag1^F971L/F971L or Rag1^R972Q/R972Q controls, suggesting a general flattening of differences due to the age (Supplementary Figures S10A-F). In line with the previous results, TCR repertoire showed lower clonality than at 6 months not only in GT mice, but also in transplanted or untreated controls (Supplementary Figure S10G).

To distinguish donor (GT) and recipient cells, we generated hypomorphic Rag1^R972Q/R972Q CD45.1 mice. We transduced donor CD45.1 cells with the MND-c.o.RAG1 vector and transplanted them into CD45.2 mutant Rag1 recipients, similarly to previous experiments. The mismatched transplants allowed to study the immune reconstitution of GT cells in the setting of hypomorphic RAG1 expression that may favor the persistence of residual T cells after irradiation.

We analyzed the engraftment of donor GT cells in PB over time, in comparison to BMT WT and BMT UT mice. We observed full donor chimerism of B, myeloid and NK cells over time in all treated groups (Supplementary Figures S11A-C). However, mixed chimerism of T cells was observed, especially in GT and BMT UT treated mice (Figure 6A). At 6 weeks, CD4+ T cells had a mean donor engraftment of 69% and 33% in GT and BMT UT, respectively, while BMT WT mice showed 95% chimerism. At 20 weeks, engraftment increased up to 94% and 78% in GT and BMT UT respectively, reaching full chimerism (99%) in BMT WT mice (Figure 6B). A similar trend was observed in CD8+ T cells with a mean donor engraftment of 59% and 18% at 6 weeks in GT and BMT UT, respectively, while BMT WT mice showed 94% chimerism. Engraftment increased up to 84% and 50% in GT and BMT UT respectively, reaching full chimerism (99%) in BMT WT mice at 20 weeks (Figure 6C).

At 6 months post-transplant, we verified the engraftment of transplanted cells in the hematopoietic organs. In thymus, we found full chimerism in most of GT mice in DN, as well as in double positive (DP) cells (Figures 6D-E). In single positive CD4+ and CD8+ cells, lower chimerism was observed only in BMT UT group, while GT animals showed full donor chimerism (Figures 6F-G). In the spleen, we confirmed significantly lower chimerism of T cells in BMT UT mice and variable frequency of CD45.1 in GT mice (Figures 6H). T cell phenotype confirmed that the majority of naïve T cells were derived from the transplanted cells in all treated groups (Figures 6I-L and Supplementary Figures S11D-I). Regarding the B cell progenitors in the BM, we found statistically significant lower chimerism in BMT UT in the most primitive pre-proB cells, while GT mice showed variable proportions of donor cells (Supplementary Figure S11J). No differences in chimerism were observed in pro, pre, immature or mature recirculating B cells (Supplementary Figures S11K-N). Full chimerism of the hematopoietic stem and progenitor cells Lin- Sca1+ cKit+ (LSK) was retrieved in all transplanted animals (Supplementary Figure S11O). In the spleen, B, myeloid and NK cells had full donor chimerism (Supplementary Figures S11P-R).

In order to dissect the kinetic of T cell reconstitution, we set up a short-term mismatched experiment in both mouse strains, terminated 3 weeks post-transplant. Peripheral blood showed partial T cell chimerism in BMT WT mice and even lower in BMT UT and GT mice from both mutant strains (Supplementary Figures S12A-C), although we do not expect T cell reconstitution at this early time point. We found full donor chimerism in all transplanted groups in the DN CD4- CD8- cells of the thymus (Supplementary Figure S12D), suggesting that the reduced T cell chimerism in PB is not due to incomplete depletion and/or competition of residual host T cell progenitors in the thymus. On the other hand, partial chimerism was seen in DP and single positive thymocytes, especially in the BMT UT group (Supplementary Figures S12E-G). Variability in chimerism is not correlated with VCN (Supplementary Figure S12H). These results points to slower T cell differentiation and thymic output when RAG1 expression is reduced or unregulated. Moreover, this suggests an important difference from the Rag1^-/- setting, that benefits from strong selective advantage of GT cells. Chimerism in the spleen was lower in BMT UT and GT mice than in BMT WT, similarly to PB (Supplementary Figure S12I). Rag1^F971L/F971L GT mice showed a higher VCN in the spleen than Rag1^R972Q/R972Q GT mice, despite the lower proportion of donor T cells (Supplementary Figure S12J). Notably, we also detected high proportion of CD4- CD8- T cells in the spleen of transplanted mice (Supplementary Figures S12K-N), suggesting the presence of immature T cells at this early time-point.

At termination, 6- or 12-months post-transplant, a total of 4 out of 20 (20%) Rag1^F971L/F971L and 7 out of 28 (25%) Rag1^R972Q/R972Q GT mice showed an abnormally enlarged thymus (Figure 7A). In addition, 1 Rag1^F971L/F971L GT mouse showed a cystic formation in a thymus of normal size. To characterize these abnormalities, we performed histological, cellular (flow cytometry) and molecular (vector copy number and integration site analysis) assays (Table 1). Flow cytometry analyses showed different expanded populations (Figure 7B), with a disorganized thymic tissue being a recurrent histological finding in most of them (Figure 7C). VCN detected in these tissues (Table 1), although variable, was in line with those observed in normal thymi (Figure 3C and Supplementary Figure S9F), suggesting that the LV is not involved in transformation. This observation is also supported by the very low VCN retrieved in 3 samples (Table 1, mouse ID: 39.3, 127.1 and 127.2 mice).

To better dissect the contribution of LV to tissue transformation, integration site (IS) analysis was performed. Results demonstrated an oligoclonal composition because Top10 IS accounted for most of total IS in both macroscopically abnormal and normal thymic tissues (Supplementary Figure S13). Similar profiles, although with a more polyclonal profile (lower cumulative frequency of Top10 IS), was observed in the corresponding spleen cells. IS in proximity to MECOM and LMO2 genes, involved in severe adverse events of previous GT clinical trials (22–29), were found only in 6 samples from 4 mice, with individual retrieval frequencies well below 1% in all cases (Supplementary Table S1). Therefore, no direct correlation could be established between abnormal thymic tissue and vector integration, although a formal toxicology study would be required in hypomorphic murine models, where the immune dysregulated context may influence the onset of thymic hyperplasia.