T Cell Acute Lymphoblastic Leukemia as a Consequence of Thymus Autonomy

Normal T lymphocyte development occurs mostly in the thymus and depends on the lifelong seeding of hematopoietic progenitors that originate in the bone marrow. In the thymus, these cells commit to the T lymphocyte lineage and further develop to generate T lymphocytes, which then emigrate and integrate the diverse repertoire that surveys the periphery. The process of T lymphocyte differentiation is highly dynamic and involves cellular turnover at such levels that guarantee complete replacement of all thymocytes every 4 weeks (1). These and other aspects of thymus and thymocyte development can be, and have been, studied using thymus transplantation experiments. Classical experiments have shown that wild type thymi grafted into severe combined immunodeficient (SCID) recipients produced and exported a single cohort of donor-derived T lymphocytes. Thereafter, thymus grafts produced no more T lymphocytes (2). These studies have been extensively reproduced by other laboratories using recipients with other genetic mutations but with the same developmental block (3). The results have been interpreted to indicate that precursors in the thymus (thymocytes) do not have the capacity to self-renew, and this remained a central dogma of thymus and T lymphocyte biology for many years (4). This dogma was disproved by the work of two independent laboratories (5, 6). Performing thymus transplantation experiments, both groups have shown that thymus function can be maintained independently of bone marrow contribution, which indicates that some thymocytes are indeed capable of self-renewal (5, 6). Specifically, it was shown that in the absence of de novo colonization of the thymus by hematopoietic progenitors, the organ maintains the production and export of functional T lymphocytes (5). Furthermore, if (host) hematopoietic progenitors seeding the thymus were deficient for IL-7r, then thymus autonomy occurred from donor-derived thymocytes. This was the case in recipients deficient for common cytokine receptor γ chain (γ_c) and for recipients that were deficient for IL-7rα (5, 6), the two chains that together form the receptor for IL-7. Both studies have shown that the TCR repertoire generated was diverse (5, 6); however, one of the papers showed that after 10 weeks of progenitor deprivation, the thymus grafts already had some degree of clonal dominance (5).

Although the physiological significance of thymus autonomy remains elusive, it became clear that it is a tightly regulated process that must be kept inhibited under physiological conditions. Indeed, follow-up work by one of the laboratories showed that prolonging thymus autonomy was permissive to the development of T cell acute lymphoblastic leukemia (T-ALL) (7). The disease was similar to the human counterpart in immunophenotype, pathologic condition, and genetic and genomic lesions, as well as at the transcriptional level. This included gain-of-function mutations in Notch1, similar to those described in the majority of human T-ALL (8), and expression of Lmo2 and Tal1, which are common to the majority of pediatric T-ALL (9–11). Nonetheless, a review article claimed that T-ALL did not develop in 19 mice used in similar experiments and further suggested that thymus transplants could be used as a therapy for correction of immunodeficiencies caused by cell-autonomous defects in T lymphocyte development (12). This was suggestive that there are important factors that could influence leukemia, and in addition to genetic factors, the animal facilities could also be relevant. Indeed, there is extensive work on the functional relevance of the composition of gut microbiota for immune function, and that identical mice from different sources, i.e., purchased from different suppliers or kept in distinct facilities, can differ in immune function (13–15).

In this study, in a new laboratory, we sought to test whether development of T-ALL from autonomy could be reproduced using newly established animal colonies, thereby excluding potential effects of the animal house on T-ALL. We covered most genotypes used by both groups, and tested whether thymus autonomy could be influenced by the genotype of the hematopoietic progenitors seeding the thymus. Furthermore, we tested whether the genotype of the recipients could impact on the development of T-ALL from wild type thymocytes, on the immunophenotype and/or severity of disease. With that purpose, we transplanted wild type thymi under the kidney capsule of recipients with the following genotypes: Rag2^−/−γ_c^−/−, γ_c^−/−, Rag2^−/−IL-7rα^−/−, and IL-7rα^−/−. They all have in common that they cannot respond to IL-7 but differ in phenotype and/or in their capacity to respond to additional cytokines. These recipients further differed among themselves in microbiota composition and, to some extent, in genetic background (determined as the percentage of C57BL/6J [B6] background). There were differences in the cellularity of thymus grafts in γ_c^−/− versus IL-7rα^−/− hosts during thymus autonomy. Nevertheless, all conditions tested were permissive to autonomy and subsequent development of T-ALL. We could detect no difference in the onset, incidence, pathologic condition, or immunophenotype of T-ALL between all tested conditions.

B6 (CD45.2⁺) mice were bred and kept at the Instituto Gulbenkian de Ciência (IGC) in a colony that is frequently replenished by mice purchased from Charles River Laboratories. B6.SJL-Ptprc^a Pep3^b/BoyJ (CD45.1⁺) mice, stock no. 002014, in this article termed B6/Ly5.1, γ_c^−/− (16), stock no. 003174, and IL-7rα^−/− (17), stock no. 002295, were purchased from The Jackson Laboratory. The strain IL-7rα^−/− was also crossed to Rag2^−/− (18) animals (originally purchased from Taconic Biosciences and kept in a colony at the IGC) to obtain the strain Rag2^−/−IL-7rα^−/−. Rag2^−/−γ_c^−/− mice used in this study were imported from the University of Ulm, Germany, through embryo rederivation. Thymi for thymus transplants were harvested from F1(B6xB6/Ly5.1) newborn donors. All mice were bred and kept in individually ventilated cages in the specific pathogen–free area of the mouse facility of the IGC. All animal experiments were approved by the Ethics Committee of the IGC – Fundação Calouste Gulbenkian and the Direção Geral de Alimentação e Veterinária.

Thymi were isolated from F1(B6xB6/Ly5.1) newborn mice and grafted under the kidney capsule of recipient mice as in previous studies (5, 7). Briefly, the thymus lobes were physically separated and kept in cold PBS. Recipient mice were anesthetized with ketamine (100 mg/kg) and xylazine (16 mg/kg), and each host received one thymus, with each lobe placed in one extremity of the kidney.

Recipients were periodically bled from the submandibular vein from 16 wk posttransplant onwards, and the peripheral blood was analyzed for the presence of T lymphocytes and/or blasts. All animals showing clinical signs of disease and/or presence of lymphoblasts in the blood were euthanized and analyzed.

Organs were harvested and single-cell suspensions prepared in PBS/FBS. Cells were blocked with 114 μg/ml mouse IgG (Jackson ImmunoResearch) for 15 min and stained for 30 min in an appropriate diluted Ab staining solution. Second-step stainings were performed whenever necessary, for 30 min, after washing twice. Abs and streptavidin were purchased from BioLegend and were as follows: CD3ε biotin (145-2C11), CD3ε APC-Cy7 (145-2C11), CD4 PE (GK1.5), CD4 PE-Cy7 (GK1.5), CD8a FITC (YTS169.4), CD11b biotin (M1/70), CD11c biotin (N418), CD19 biotin (6D5), CD25 BV605 (PC61), CD44-PerCP-Cy5.5 (IM7), CD45.1 biotin (A20), CD45.1 PE-Cy7 (A20), CD45.2 PerCP-Cy5.5 (104) and CD45.2 PE (104), CD45.2 PB (104), CD117 APC (2B8), CD127 APC (A7R34), γδTCR PE (GL3), Gr-1 biotin (RB6-8C5), NK1.1 biotin (PK136), Ter119 biotin (TER-119), streptavidin BV605, and streptavidin BV785. Double-negative (DN) thymocytes were defined as CD4 negative, CD8 negative, lineage negative (lineage-positive cells were defined as positive for an Ab mixture, including CD3ε, CD11b, CD11c, CD19, Gr-1, NK1.1, and Ter119). Dead cells were excluded with SYTOX Blue (Molecular Probes), Zombie APC-Cy7, or Zombie Pacific Orange (BioLegend). Sample acquisition was performed in a BD LSRFortessa X-20 cell analyzer using BD FACSDiva 8 software, and analyses were done using FlowJo.

Samples were fixed by immersion in 10% buffered formalin for ∼48 h at room temperature, with exception of the brain (1 wk). Bone samples were decalcified with an EDTA solution changed daily during 1 wk. Samples were routinely processed, embedded and sectioned into 3-μm-thick sections, and stained with H&E. Liver, brain, and male reproductive organs were submitted to gross examination and trimming to maximize the area of analysis. The liver lateral left lobe was macroscopically sectioned in four to five transverse fragments. The brain was macroscopically sectioned in four to five consecutive coronal sections, ∼2.5 mm apart, starting in the middle of the olfactory bulbs. All histology was performed by the Histopathology Unit at the IGC. Slides were analyzed with a DMLB2 microscope (Leica), and images were acquired with a DFC320 camera (Leica) and NanoZoomer-SQ Digital slide scanner (Hamamatsu Photonics).

Statistical analysis of the results was conducted in Prism 7. The survival curves in Fig. 2A were compared using a log-rank (Mantel–Cox) test, and no statistically significant difference between the curves was found. Sample groups in Fig. 2C were subjected to the D’Agostino and Pearson normality test, revealing a nonparametric distribution. Samples were compared with a Kruskal–Wallis test followed by a Dunn multiple comparison post hoc test, and no statistically significant difference was found.

Fresh fecal pellets were collected from five to seven cages per genotype, with the aim of representing the whole colony (avoiding sampling cages that came from the same breeders). DNA extraction from the mouse fecal content was performed using a QIAamp DNA Stool Mini Kit (Qiagen), according to the manufacturer’s instructions. Amplification and sequencing of the 16S rRNA gene was carried out at the IGC’s Genomics Unit using primers specific to the V4 region of 16S rRNA, as described (19), and paired-end sequencing (2 × 250 bp) on an Illumina MiSeq Benchtop Sequencer following Illumina recommendations.

Bioinformatic processing of raw demultiplexed reads was done using QIIME2 v.2018.4 (https://qiime2.org) with default parameters. DADA2 was used for quality filtering, denoising, paired-end merging, and amplicon sequence variant (ASV, i.e., sub–Operational Taxonomic Units) calling using qiime dada2 denoise-paired method (20). A threshold was defined at a Phred quality score of 20 for trimming before merging (parameters: –p-trunc-len-f and –p-trunc-len-r), defining 235-bp length for forward reads and 220 bp for reverse reads. After filtering, average sample size was 32,576 reads (minimum: 12,245, maximum: 62,457), and 627 ASVs were detected. ASVs were aligned using the qiime alignment mafft method (21). The alignment was used to calculate the phylogenetic distances between ASVs using qiime phylogeny fasttree (22). ASV tables were subsampled without replacement to even sample size for diversity analysis using qiime diversity core-metrics-phylogenetic pipeline. The smallest sample size was chosen for subsampling. Unweighted and weighted UniFrac distances were calculated to compare community structure (23). Taxonomic assignment of ASVs was performed using a Bayesian classifier trained with the Greengenes database (i.e., 16S rRNA sequences aligned at 99% and truncated at positions 515–806) using the qiime feature-classifier classify-sklearn method (24). UniFrac distance matrices and ASV tables were used to calculate principal coordinates and construct ordination plots using R software package version 3.4.3 (http://www.R-project.org). The significance of groups in community structure was tested using PERMANOVA. BiodiversityR version 2.8-4 and vegan version 2.4-5 packages were used, and ggplot version 2.2.1 was used for plotting. Bioinformatic processing and analysis of the data were performed at the Bioinformatics Unit of the IGC.

Enrichment of B6 genomic DNA was determined by genotyping each individual mouse for 159 informative single nucleotide polymorphisms (SNPs) evenly distributed across the genome (1 SNP/9 cM on average). Genotype detection was done using the mass spectrometry–based Agena iPlex Gold technology, which allows for multiplexing of up to 40 SNPs per reaction, totalling five multiplex reactions. In brief, a set of three primers was designed for each SNP: two to be used in a regular PCR reaction to amplify the region of the SNP of interest, and the third to be used in the iPlex reaction. This is a single base extension PCR with mass-modified nucleotides, in which the primer anneals immediately before the position of the SNP. The extension product is analyzed by time of flight, and genotypes were attributed according to the m.w. difference of the extended base relatively to the primer mass.

Thymus autonomy has been previously reported by two independent laboratories (5, 6), and extending the period of thymus autonomy led to T-ALL (7). The first two studies had relatively short observation times posttransplantation, and only one of the groups referred to a small number of recipients that were followed for 7 mo (6). The reduced number of mice followed for that time could account for failure in the detection of leukemia. However, a more recent article refers to one cohort of Rag2^−/−γ_c^−/−–transplanted mice that was followed for 9 mo posttransplantation by the same group (12). The genotype of the recipients used for the studies on thymus autonomy did not fully overlap between the laboratories (Table I), and we reasoned that this could be one aspect impacting on autonomy. To address this point, we transplanted wild type thymi into the following recipients: Rag2^−/−γ_c^−/−, γ_c^−/−, Rag2^−/−IL-7rα^−/−, and IL-7rα^−/−. The cells in γ_c^−/− mice cannot respond to IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21. In IL-7rα^−/− mice, cells cannot respond to IL-7 and to thymic stromal lymphopoietin. In the thymus, both γ_c^−/− (16, 25) and IL-7rα^−/− (17) mice have marked reduced cellularity, caused by the inability of thymocytes to respond to IL-7 at the DN2 stage (CD4⁻ CD8⁻ double negative, CD3 negative CD25 positive, CD44 positive). As a consequence, the next stage, the DN3 (CD25 positive, CD44 negative), is severely reduced, and very few thymocytes progress through differentiation. Rag2^−/−γ_c^−/− and Rag2^−/−IL-7rα^−/− thymocytes have, in addition, a complete developmental block at the DN3 stage, caused by the Rag2 deficiency (18), which leads to an even more severe phenotype. Thymus grafted into these mice and analyzed 9 wk posttransplantation could maintain double-positive thymocytes of donor origin in recipients from all tested genotypes (Fig. 1A). Similar to one of the reports (5), no population of thymus-donor origin resembling the DN3, according to immunophenotype, was detected at this stage (Fig. 1B). The recipients with at least one productive thymus graft were over 80% in every group (Fig. 1C). The only difference detected was a statistically significant difference between the number of CD4⁺ CD8⁺ double-positive thymocytes of donor origin in thymi grafted into γ_c^−/− versus IL-7rα^−/− (Fig. 1D) recipients.

Even if no major differences were detected 9 wk posttransplantation, it could still be that there were differences in the development of T-ALL. Therefore, we used the same transplantation setting and extended the time of observation following thymus transplantation to 70 wk posttransplantation. T-ALL developed with onset at 15–17 wk posttransplantation and reached an incidence of 74–85% in all recipients (Fig. 2A). No statistical difference was found between groups. Blast cells were detected in the blood of leukemic animals (data not shown). Organs of mice with T-ALL were analyzed by flow cytometry (bone marrow, grafted and endogenous thymi, and spleen) and were mostly composed of leukemic blast cells that had an immature surface phenotype (Fig. 2B). Although there was some variability in the immunophenotype of the different leukemias, they were most commonly immature single positive to (CD4 and CD8) double positive. In line with one previous report (7), animals had large thymus grafts and developed hepatosplenomegaly (Fig. 2C). H&E–stained sections show multiorgan neoplastic cell infiltration of monomorphous medium- to large-sized neoplastic lymphocytes, with scant cytoplasm and irregular-edged nuclei with coarsely clumped chromatin, and with numerous mitotic figures (Fig. 3A). The heart was often not affected, but when it was, it exhibited multifocal, perivascular to interstitial neoplastic lymphoid cell infiltrates. The lungs presented predominantly perivascular and peribronchial infiltrations. The liver showed perivascular and periportal cell infiltrations with invasion into the adjacent liver parenchyma. The kidney exhibited interstitial cell infiltrations, denser in the renal cortex, with invasion of the retroperitoneal and perirenal adipose tissue. The brain displayed, in most cases, a multifocal cell infiltration in the subarachnoid and pia mater and in the choroid plexus in the lateral and third ventricles. Less frequently, neoplastic cells were also identified in the olfactory bulbs and in the median eminence of the hypothalamus (Fig. 3A). The spleen showed total obliteration of the normal architecture, with white and red pulp structure loss, due to the proliferation of lymphoid cells. The bone marrow was replaced by a diffuse monomorphous neoplastic lymphoid population. In male mice, both testis and accessory glands presented an interstitial cell infiltration, whereas in female mice, there was total obliteration of the ovaries and uterus architecture due to the neoplastic diffuse cell infiltrates (Fig. 3B). Those mice that did not develop T-ALL were analyzed at the end of the experiment and, out of 15 recipient mice, across all genotypes, only one still had productive thymus grafts (had CD4⁺ CD8⁺ double-positive thymocytes of donor origin), whereas all others were exhausted, showing no sign of thymopoiesis of donor origin. This might indicate that the mice that failed to develop leukemia did so mostly because autonomous thymopoiesis stopped previously or did not take place. Overall, no differences could be determined between experimental groups in type, degree of infiltration, or in the organs affected by T-ALL (Fig. 3).

Beyond the genotype of the recipients and the consequent causes in T lymphocyte development of host origin, additional factors could have influenced the difference in T-ALL between the two laboratories (7, 12). These include the health status of the animals, the sanitary conditions in which they were kept, their microbiota composition, and even the genetic background of the recipients. Of note, the studies were performed in different research institutions, i.e., one group kept their mouse colonies in the Central French breeding facilities (6, 12), the other at the University of Ulm, in Germany (5, 7). In this study, we have established the mouse colonies newly at the IGC, Portugal. Rag2^–/–γ_c^–/– were imported through embryo rederivation from Germany and correspond to the strain used in some of the previous studies (5, 7). IL-7rα^–/– were imported from The Jackson Laboratory through embryo rederivation. After rederivation, IL-7rα^–/– were crossed to animals from a Rag2^–/– colony that was already kept at our institute to generate the strain Rag2^–/−IL-7rα^–/–. γ_c^–/– were imported from The Jackson Laboratory and tested clean upon arrival and hence were not rederived. Because all animals used in this study were imported newly into a new animal facility, they inevitably differed from the animals used in former studies. In addition, gut microbiota composition could also differ between the strains used in this study as a result of rederivation (or its absence), which was involved in the import of some of the mouse strains. To determine whether the gut microbiota composition differed between the four recipient strains used in this study, we collected fecal pellets from animals of five to seven different cages per strain and sequenced the V4 region of the 16S rRNA gene. The composition of the gut microbiota was analyzed using unweighted and weighted UniFrac phylogenetic distances. Principal coordinate analyses show that the gut community of γ_c^–/– mice differed from that of the other strains (Supplemental Fig. 1A). These results were consistent for both the unweighted UniFrac distance (PERMANOVA, p ≤ 0.05), which is based on the presence/absence of bacterial groups, and for the weighted UniFrac distance, which also considers the relative abundance of bacterial groups (PERMANOVA, p ≤ 0.05). The differences in community structure were further evidenced by analysis of taxonomic composition, depicted in this article at the level of phyla (Supplemental Fig. 1B). Specifically, the phylum Deferribacteres was absent in the gut microbiota of the strain γ_c^−/− (Supplemental Fig. 1B). Further differences could be detected at the level of genera and include the absence of Prevotella, Odoribacter, AF12 (phylum Bacteroidetes), Sutterella, and Bilophila (phylum Proteobacteria) in the gut microbiota of γ_c^−/−. These genera were, however, present in Rag2^−/−IL-7rα^−/−, IL-7rα^−/−, and Rag2^−/−γ_c^−/− (Supplemental Fig. 1C). The only genotype with a clearly different microbiota composition was γ_c^−/−, consistent with these mice being the only ones not rederived upon import into the current facility. Nevertheless, the fact that γ_c^−/− mice developed T-ALL similarly to the other genotypes tested indicates that T-ALL was not influenced by the microbiota.

The genetic background of the recipients could also have impacted on the results. Rag2^−/−γ_c^−/− mice used by one of the groups had been crossed into a pure B6 background (12), whereas the other was crossed to B6 for fewer generations. To determine the genetic background of the strains used in this study and define the degree of genetic backcross into the B6 background, we analyzed the genomic DNA of our strains and screened for SNPs to determine the percentage of B6 background in each chromosome (Fig. 4). As predicted, every strain had one to two SNPs positive for the strain 129 and not B6, consistent with the loci that were targeted in those strains, i.e., Rag2 in chromosome 2, IL-7rα in chromosome 15, and γ_c in chromosome X (Fig. 4). Beyond these loci, the only strain for which the B6 background was incomplete is the Rag2^−/−γ_c^−/−, which was the same strain used in previous work (5, 7). Nevertheless, T-ALL developed identically in the four strains, three of which were in a pure B6 background. Hence, our results show that the pure B6 background of the recipients does not protect from leukemogenesis (Fig. 4).

In this article, we show that transplantation of wild type thymi into recipients that are Rag2^−/−γ_c^−/−, γ_c^−/−, Rag2^−/−IL-7rα^−/−, or IL-7rα^−/− is followed by a period of thymus autonomy, and that is permissive to malignant transformation and development of T-ALL. Although thymus autonomy was characterized by a difference in cell numbers of donor double-positive thymocytes between the thymi transplanted into γ_c^−/− versus IL-7rα^−/− hosts, no other differences were detected among recipients, including incidence and phenotype. Furthermore, no differences for T-ALL in terms of onset, incidence, pathology, or immunophenotype of T-ALL were found among the four groups. This was so despite the small difference in genetic background (percentage of B6 background) of the Rag2^−/−γ_c^−/− recipients and the variability in composition of the gut microbiota in γ_c^−/− recipients. These data are concordant with the concept of cell competition in the regulation of thymus turnover under physiological conditions, in which “young” precursors, with a shorter time of thymus residency, lead to the clearance of the “old” precursors, with a longer dwell time in the thymus (7). Taken together, our data support that IL-7 is a limiting factor in early T lymphocyte development, which determines the capacity of the young precursors to outcompete the old, thereby promoting cellular turnover. Although the mechanism of IL-7 regulation of thymus turnover remains elusive, several hypotheses can be considered. Because IL-7 is limiting, the differential access to the cytokine could determine the outcome of the cells: survival and expansion of the young versus clearance of the old. This outcome could be reached if the young are closer to the source of IL-7 than the old precursors, or if young and old cells express different levels of the receptor. Future work will be essential in defining the molecular mechanism of cell competition and testing which of the hypotheses is correct. Furthermore, the regulation of thymus autonomy as an outcome of the competitive disadvantage of the progenitors seeding the thymus will be crucial in understanding whether any cell that enters the thymus has the potential to become capable of sustaining thymopoiesis autonomously, or if this property is acquired de novo in specific conditions. Finally, it will be interesting to identify the molecular changes occurring that cause malignant transformation of healthy, wild type thymocytes.

The difference of results between two laboratories is rather intriguing. Although both reported on thymus autonomy (5, 6), the work that followed in the Rodewald laboratory showed that autonomy is permissive to T-ALL (7). The latter finding was challenged by Rocha’s group (12). This difference has important implications, as thymus transplants were proposed as a potential therapy to be used in the clinics for correction of immunodeficiencies caused by cell-autonomous defects in thymocyte development (12). Furthermore, the discrepancy suggests that there might be factors beyond the genotypes tested that can affect leukemogenesis in thymus transplantation experiments. Whereas Rocha’s former study did not follow enough mice for long enough time to detect T-ALL (6), the review article refers to a group of 19 animals followed for over 6 mo (12). It would have been interesting to know whether thymopoiesis was still active in all those animals at the end of the observation period. Although the authors suggested that inefficient backcrossing of recipients onto B6 background might have triggered disease (12), we now show that the pure B6 background per se is not sufficient to prevent leukemia. Alternatively, differences in microbiota composition could account for the dissimilar results between laboratories, but we consistently observed identical T-ALL development independently of differences in the gut microbial diversity and composition of the recipient mice. Moreover, results were reproduced in a new animal facility, likely to differ significantly in gut microbiota. Because the leukemias originate in the cells of donor origin, it would be very interesting to test the donor strain that was used by Rocha’s laboratory. It cannot be fully excluded that the strain used might have undergone genetic drift and lost the genetic determinants involved in leukemogenesis, as the authors themselves discuss (12). However, their mouse colonies have been “refreshed” by newly imported animals, and the original mice are no longer available (12). In this study, we addressed all sources of variability that we could assess experimentally, including genotypes, genetic backgrounds, and the possible influence of the microbiome of the mice used. Although we detected some differences, we could show that they played no role in the two relevant phenotypes for our study: thymus autonomy and leukemia. Under the new environmental conditions intrinsic to a new laboratory, it was possible not only to verify the data from the Rodewald laboratory but also to extend those findings to genotypes that had not been tested previously for thymus autonomy and/or leukemia. We covered several mouse strains that had not been covered in former studies both during the earlier stage of thymus autonomy and for T-ALL.

It has been suggested that T-ALL developing as the consequence of thymus grafting could explain the onset of leukemia in X-linked SCID (SCID-X1) patients treated by gene therapy (7). SCID-X1 patients carry loss-of-function mutations in the γ_c gene and require a transplant of healthy bone marrow to survive. Being a monogenetic disease that affects the hematopoietic system, SCID-X1 is a good candidate for correction by gene therapy in the absence of a compatible donor. Following this rationale, SCID-X1 has pioneered the first clinical trials of gene therapy for correction of a primary immunodeficiency (26). In total, 20 patients lacking a suitable donor enrolled in a trial for correction of the genetic defect and showed good signs of recovery and immune reconstitution, including thymus activity with de novo production of T lymphocytes for several years (27). However, 5 out of the 20 patients developed T cell leukemia as a consequence of the treatment (28, 29). This unfortunate event followed a period of years in which the patients continuously produced T lymphocytes de novo, independently of bone marrow contribution (30). We consider likely that the cellular constellation in the patients was identical to that of thymus autonomy in the thymus-transplanted mice. Nevertheless, the favored explanation for leukemia in the treated patients has been that genotoxicity due to integration of the retroviral vectors close to oncogenes led to their ectopic activation and consequent neoplastic transformation of the corrected cells (28, 29), and extensive work has followed on the improvement of better vector design (31, 32). Although the vector design is of unquestionable importance, namely for the efficiency of transduction of the target cells and safety for the patients, it is plausible that this was not the sole factor driving T-ALL in the gene therapy trials. In support of this view is the case of success for correction of adenosine deaminase–SCID (ADA-SCID) by gene therapy (33, 34). ADA-SCID is caused by deficiency in adenosine deaminase, and although the genetic and molecular defects differ from those in SCID-X1, they are both characterized by impaired lymphocyte development. Furthermore, the original vectors used in both trials were very similar (26, 35). Opposite to the treatment of SCID-X1, however, ADA-SCID patients are always myeloablated prior to the infusion of the corrected hematopoietic stem and progenitor cells, thereby effectively engrafting the corrected cells in the bone marrow. This difference suggests that the efficiency of bone marrow correction is an essential factor in the prevention of leukemogenesis emerging during treatment of SCID by gene therapy. In this line, two independent studies recently showed that the relative proportion of γ_c-corrected (or wild type) versus γ_c-deficient cells in the bone marrow is fundamental to guarantee that immune reconstitution is complete and no lymphoid malignancy emerges as a side effect (36, 37). It is becoming progressively evident that conditions that enable thymus autonomy are permissive for, and precede, leukemogenesis (38, 39). It will be important to extend these studies and understand which non–cell autonomous mechanisms regulate normal turnover and inhibit thymus autonomy, as well as understanding the cellular and molecular alterations occurring during autonomy that enable leukemogenesis. Finally, it is fundamental to limit thymus autonomy to prevent the risk of T-ALL development, and this is of particular relevance in the context of correction of immunodeficiencies.

We thank K. Xavier, J. Barata, J. Demengeot, J. Carneiro, and H.J. Fehling for critical reading of the manuscript. We acknowledge D. Zanatta and C. Alves for technical support and H.J. Fehling for the Rag2^−/−γ_c^−/− mice. We thank the Animal House Facility, the Genomics Unit, and the Histopathology Unit of the IGC for outstanding support of this work.

The authors have no financial conflicts of interest.