Overexpression of Lmo2 initiates T-lymphoblastic leukemia via impaired thymocyte competition

T cell acute lymphoblastic leukemia (T-ALL) is an acute, predominantly pediatric cancer caused by the multistep transformation of developing T cell progenitors in the thymus (Belver and Ferrando, 2016; Girardi et al., 2017; Iacobucci and Mullighan, 2017). A hallmark of T-ALL is overexpression of oncogenic transcription factors, with most cases containing chromosomal abnormalities that cause overexpression of transcription factors that are normally silenced during early stages of T cell development. Commonly targeted transcription factors include basic helix-loop-helix factors (TAL1, LYL1, and TAL2) and their binding partners, the Lim-domain only proteins, LMO1 or LMO2, as well as homeobox transcription factors including the HOXA cluster, TLX1, and TLX3. These transcription factors are complemented by additional mutations that commonly affect NOTCH and JAK/STAT signaling, tumor suppressors, protein translation, and epigenetic pathways (Belver and Ferrando, 2016; Girardi et al., 2017; Iacobucci and Mullighan, 2017). However, the molecular and cellular effects of oncogenic transcription factors on T cell development, as well as how they interact with secondary mutations to cause T-ALL, remain poorly understood.

LMO2 is an oncogenic transcription factor that is overexpressed in up to half of all T-ALL cases (Ferrando et al., 2002; Wu et al., 2015) due to chromosomal translocations, loss of insulator boundaries, or non-coding mutations that result in neomorphic promoters (Abraham et al., 2017; Chen et al., 2011; Dik et al., 2007; Goossens and Van Vlierberghe, 2017; Hnisz et al., 2016; Rahman et al., 2017; Van Vlierberghe et al., 2006; Wu et al., 2015). In addition, LMO2 was targeted by retroviral insertion mutagenesis in four out of five cases of T-ALL that arose during gene therapy trials for X-linked severe combined immunodeficiency (X-SCID; Hacein-Bey-Abina et al., 2008; Hacein-Bey-Abina et al., 2003; Howe et al., 2008; McCormack and Rabbitts, 2004) as well as in six cases of T-ALL that occurred following similar gene therapy for Wiskott–Aldrich Syndrome (Braun et al., 2014). LMO2 encodes a small (∼20 kD) protein with no DNA binding capacity, which is almost entirely comprised of zinc finger-containing LIM domains that mediate protein–protein interactions. As such, it functions as a bridging molecule that mediates the assembly of multifactorial transcriptional complexes that play critical roles in hematopoietic stem cells (HSCs), erythroid development, and vascular remodeling, as well as aberrant complexes that drive T-ALL development (Chambers and Rabbitts, 2015; Matthews et al., 2013).

To study the role of LMO2 in T-ALL, transgenic mouse models have been developed that recapitulate its overexpression during T cell development in the thymus (Larson et al., 1994; Ruggero et al., 2016; Smith et al., 2014). Using cell lineage tracing in CD2-Lmo2 transgenic mice, we found that LMO2 induces autonomous thymocyte self-renewal, resulting in a population of transplantable preleukemic stem cells (pre-LSCs) that show serial long-term engraftment capacity and persist for up to 1 yr prior to the development of overt leukemia (McCormack et al., 2010). Subsequent studies have shown that other transcription factors, including the related TAL1 and LMO1 oncoproteins and the fusion protein NUP98-HOXD13, also result in the development of pre-LSCs (Shields et al., 2019; Tremblay et al., 2010). In all three of these models, pre-LSCs share a similar, early T cell precursor (ETP)–like gene expression profile and show an essential requirement for the LMO2 cofactor LYL1, suggesting a common mechanism by which different oncogenic transcription factors cause thymocyte self-renewal (Gerby et al., 2014; Shields et al., 2019). Moreover, recent methylation profiling has provided evidence that most human T-ALL cases derive from self-renewing pre-LSCs (Roels et al., 2020).

Together, the above studies indicate an important role for thymocyte self-renewal in initiating T-ALL. Indeed, thymocytes normally do not self-renew but are maintained by periodic seeding of thymus-settling progenitors (TSPs) from the bone marrow (BM; Boehm, 2012). However, when thymocytes are placed in an environment in which incoming TSPs are defective, for example in mice deficient in the IL-7 receptor common gamma chain, thymus autonomy ensues (Martins et al., 2012; Peaudecerf et al., 2012). In four separate studies (Gao et al., 2019; Ginn et al., 2017; Martins et al., 2014; Schiroli et al., 2017), but not a fifth (Peaudecerf et al., 2016), this situation resulted in T-ALL development, indicating that cell competition is a tumor suppressor mechanism in the thymus. This has prompted speculation that reduced cell competition may have played a role in the development of T-ALL in patients undergoing gene therapy for X-SCID (Ginn et al., 2018; Martins et al., 2014); however, this remains controversial (Peaudecerf et al., 2016; Qasim et al., 2014). Moreover, to date, T-ALL has only been shown in experimental settings of TSP deprivation, and whether cell competition plays a role in oncogene-induced T-ALL is unknown. Here, we investigate this in a model of LMO2-induced T-ALL. We demonstrate that failure of cell competition is critical in the establishment of pre-LSCs and subsequent leukemogenesis. This implies that impaired cell competition is an important mediator of oncogenesis in murine models of T-ALL and may play a key role in human T-ALL pathogenesis.

We have previously used a cell lineage tracing approach to show that overexpression of Lmo2 in CD2-Lmo2 transgenic mice blocks the import of BM-derived progenitors and results in self-renewal of thymocytes from a young age, up to 1 yr prior to the onset of T-ALL (McCormack et al., 2010). To determine whether this property of self-renewal is intrinsic to the thymus, we transplanted thymic lobes from neonatal Lmo2-transgenic mice into the kidney capsule of congenic (CD45.1) recipient mice and vice versa. 1 mo later, the percentage of recipient (BM-derived) thymocytes was determined (Fig. 1 A). As expected, WT donor thymi were completely seeded by recipient (BM-derived) cells by 4 wk after transplant (Fig. 1, B and C). Surprisingly, however, Lmo2-transgenic neonatal grafts were also completely seeded by recipient-derived cells in WT recipients but were maintained in Lmo2-transgenic recipients (Fig. 1, B and C). Furthermore, WT thymi also showed substantial autonomy when engrafted into the kidney capsule of Lmo2-transgenic mice, being comprised of mostly donor-derived cells at 4 wk after transplant (Fig. 1, B and C). This suggests a cell non-autonomous effect on thymopoiesis in Lmo2-transgenic mice which results in thymic autonomy.

The above results suggested that CD2-Lmo2 transgenic mice may have a defect in thymic settling progenitors leading to long-term maintenance of downstream T cell progenitors, a situation that has been described previously (Martins et al., 2014; Martins et al., 2012; Peaudecerf et al., 2012). To examine this, we performed competitive BM transplantation using Lmo2-transgenic cells (Fig. 2 A). At 4 wk after transplantation, Lmo2-transgenic BM gave normal BM engraftment, but strikingly showed an almost complete failure to reconstitute the thymus (Fig. 2, B and C). Analysis of individual progenitor populations showed that Lmo2-transgenic progenitors could contribute to BM progenitor populations and early thymocyte stages (ETPs and CD4-CD8 double-negative [DN]2 thymocytes; Fig. 2, D and E). However, there was an almost complete lack of engraftment of Lmo2-transgenic cells beyond the DN2 stage (Fig. 2), an effect that persisted at 12 wk after transplant (Fig. 2, D and E). Thus, Lmo2-transgenic progenitors can settle the thymus but exhibit a profound developmental block at the DN2 stage of T cell development.

Consistent with a developmental block at the DN2 stage of T cell development, analysis of neonatal (1 wk old) Lmo2-transgenic mice showed a clear accumulation of DN2 cells with abnormally high Kit expression that persisted until ∼1 mo of age (Fig. S1, A and B). However, by 2 mo of age, a population of DN3 thymocytes appeared that showed aberrantly high expression of Kit, as previously reported in this model (Larson et al., 1995; McCormack et al., 2010; Fig. S1, A and B). Further analysis showed that these cells were blocked at the DN3a stage, but showed expression of intracellular TCR-β, which is normally expressed from the DN3b stage (Fig. S1, A and C). Analysis of cell cycling status revealed that the accumulating DN2 thymocytes in neonatal Lmo2-transgenic mice were highly quiescent (Fig. S1, D and E). While neonatal Lmo2-transgenic DN3 thymocytes were also abnormally quiescent, by 2 mo of age, these cells showed normal cycling (Fig. S1, D and E). Thus, Lmo2 overexpression initially causes a developmental block at the DN2 stage, which is followed by an accumulation of abnormal DN3a thymocytes by 2 mo of age.

To determine whether Lmo2 directly enables thymocytes to persist and expand in vivo, Lmo2-transgenic lymphoid primed multipotent progenitors (LMPPs) were isolated and their potential for long-term thymic engraftment was examined by injecting them directly into the thymi of sublethally irradiated congenic (CD45.1) mice. At 1 mo after transplant, Lmo2-transgenic cells showed normal numbers of donor-derived DN2 thymocytes but decreased numbers of downstream thymocyte subsets (Fig. 3, A and B). At 3 mo after transplant, both WT and Lmo2-transgenic donor cells showed low levels of donor engraftment (Fig. 3, A and B). However, while the WT donor-derived cells were mostly fully differentiated CD4 T-cells, Lmo2-transgenic LMPP-derived cells still contained significant numbers of donor-derived DN2 thymocytes, with an average of 6,500 DN2 cells remaining (Fig. 3, A and B). As this was ∼10% of the number of donor-derived DN2 cells found at 1 mo after transplant, these results indicate that in this competitive setting, Lmo2-transgenic LMPPs can generate a population of developmentally blocked DN2 thymocytes that persist long term but do not expand in number, suggesting that these cells lack self-renewal capacity.

We have previously shown that thymocytes in adult (2 mo old) Lmo2-transgenic mice have self-renewal potential, as evidenced by serial long-term engraftment in transplantation experiments, and that this capacity is restricted to DN3 thymocytes (McCormack et al., 2010). To determine the stage at which long-term engraftment capacity arises in these mice, we transplanted thymocytes from Lmo2-transgenic mice of various ages into sublethally irradiated congenic (CD45.1) recipients (Fig. S2 A). This showed that while thymocytes from 1-wk-old Lmo2-transgenic mice did not engraft long-term (Fig. S2 B), and thymocytes from 1-mo-old Lmo2-transgenic mice showed variable engraftment potential, thymocytes from 2-mo-old Lmo2-transgenic mice showed near-complete long-term engraftment of recipient thymi, as shown previously (McCormack et al., 2010; Fig. S2 B). Further analysis of donor thymocytes from 1- and 2-mo-old Lmo2-transgenic mice showed that they comprised all T cell subsets after the DN3 stage (Fig. S2, C and D), suggesting that DN3 thymocytes are the source of long-term engrafting cells, as shown previously (McCormack et al., 2010).

The presence of long-term engrafting thymocytes from 1–2 mo of age implies that their development is a frequent event in Lmo2-transgenic mice. Accordingly, assessment of TCR-β subfamily usage showed that at 1 wk of age, Lmo2-transgenic T cells were polyclonal; however, by 2 mo of age dominant clones arose, suggesting the possible presence of secondary mutations in enabling thymocyte self-renewal (Fig. S2 E). Moreover, sorted DN3a thymocytes from 2-mo-old Lmo2-transgenic mice showed the presence of multiple clones with distinct TCR-β usage, which were maintained when these cells were transplanted (Fig. S2, F–I). Thus, the acquisition of long-term engraftment potential in Lmo2-transgenic thymocytes is a frequent event, often occurring in multiple clones by 2 mo of age.

Previous studies have shown increased apoptosis in pre-LSCs that arise in Lmo2-transgenic mice from 6 wk of age (Tremblay et al., 2016). To determine whether this also occurs in developmentally arrested DN2 thymocytes in young Lmo2-transgenic mice, we analyzed the apoptosis of Lmo2-transgenic mice at 2 wk of age. We found a marked increase in apoptotic cells specifically within the developmentally arrested DN2 population (Fig. 4, A–C), suggesting that apoptosis is a critical mediator of the defect in cellular competition that drives the development of pre-LSCs in Lmo2-transgenic mice. To test this further, we bred Lmo2-transgenic mice with Vav-BCL2 transgenic mice that express the antiapoptotic factor BCL2 throughout hematopoiesis (Ogilvy et al., 1999). This led to a restoration of thymus size in Lmo2-transgenic mice (Fig. S3 A), and while DN thymocytes still accumulated in Lmo2;BCL2 double transgenic mice, these included not only the DN3a cells that accumulate in Lmo2-transgenic mice but downstream DN3b-c and DN4 subsets as well (Fig. S3, B and C). Accordingly, in competitive BM transplantation experiments, BCL2 expression restored the contribution of Lmo2-transgenic cells to the DN3 and DN4 stages (Fig. S3, D–F). However, development of double-positive (DP) and CD4/CD8 single-positive (SP) T cells remained significantly impaired, in line with previous findings that DP cells are reduced in BCL2-transgenic mice (Ogilvy et al., 1999).

Cell cycle analysis of Lmo2;BCL2 double transgenic thymocytes showed that DN3 cells in these mice were mostly quiescent (Fig. S4 A), and there was no evidence of expanded T cell clones, implying that pre-LSCs do not emerge in the thymi of these mice (Fig. S4 B). Accordingly, in long-term engraftment experiments, while both BCL2-transgenic and Lmo2;BCL2 double transgenic thymocytes showed some engraftment in primary recipients (Fig. S4, C–E), this was not maintained in secondary recipients, unlike Lmo2-transgenic thymocytes, which show serial transplantation capacity (Fig. S4, C, F and G; McCormack et al., 2010). Moreover, onset of T-ALL in these mice was almost completely abrogated (Fig. 4 D), with Lmo2;BCL2 double transgenic mice instead succumbing to B cell neoplasms similar to those seen in BCL2-transgenic mice (Egle et al., 2004; Fig. 4, E–G). Thus, BCL2 rescues the failure of cell competition in the thymus of Lmo2-transgenic mice, thereby preventing the development of pre-LSCs and subsequent leukemogenesis.

To determine whether competition from WT progenitors could similarly prevent Lmo2-driven leukemogenesis, we performed BM transplantation experiments in which Lmo2-transgenic BM cells were injected with or without a 1:3 ratio of WT competitor cells. While mice transplanted with Lmo2-transgenic BM cells in the absence of competitor developed leukemia at a similar rate to Lmo2-transgenic mice (Fig. 5 A, compare with Fig. 4 D) in the presence of WT BM cells, T-ALL initiation was almost completely overcome, with only three of 11 mice analyzed succumbing to T-ALL with long latency (Fig. 5, A, C and D), despite robust hematopoietic engraftment (Fig. 5 B). At endpoint, a further four mice receiving Lmo2-transgenic BM cells plus competitor contained small thymi containing donor cells that were blocked at the DN2 stage (data not shown), while the remainder (four of 11) were athymic. Therefore, impaired cell competition is required for Lmo2-driven T cell leukemogenesis.

To examine the molecular features of developmentally arrested thymocytes in Lmo2-transgenic mice, we used RNA sequencing (RNA-seq) to compare the gene expression profiles of Lmo2-transgenic DN2 and DN3 thymocytes at two separate time points, 2–3 wk of age (hereafter termed “young”), a point at which developmentally arrested DN2 thymocytes are present but pre-LSCs have not yet arisen, and 6–10 wk of age (hereafter termed “old”), at which point self-renewing DN3 thymocytes (pre-LSCs) have developed. Multidimensional scaling showed that the greatest transcriptional changes were in young DN2 thymocytes, with Lmo2-transgenic DN3a thymocytes being more closely related to WT cells, before diverging in older mice (Fig. 6 A and Table S1). Analysis of Lmo2 expression confirmed its overexpression in all three Lmo2-transgenic thymocyte populations analyzed (Fig. 6 B). Gene set analysis showed that genes that are normally upregulated during the transition from ETP to DN2 thymocytes remained repressed in young Lmo2-transgenic DN2 thymocytes (Fig. 6 C), including Ptcra, CD3d, CD3e, CD3g, IL7r, and Rag1 (Table S2). Furthermore, genes normally downregulated during the ETP to DN2 transition remained expressed in these cells (Fig. 6 C), including HSC-associated (Cd34, Hhex) and myeloid-associated genes (Mpo, Gfi1b; Table S2). Moreover, as expected from our finding that these cells are quiescent (Fig. S1, D and E), genes involved in the cell cycle were strongly downregulated (Fig. 6 D). Thus, developmentally arrested Lmo2-transgenic DN2 thymocytes show a gene expression profile that is reminiscent of ETPs.

Similar gene set analyses of old, self-renewing DN3a thymocytes showed that these cells also showed an immature gene expression profile, similar to normal DN2 thymocytes (Fig. 6 E). Moreover, cell cycle genes were upregulated (Fig. 6 F), indicating a high degree of cell cycling in this population, which could account for the acquisition of further mutations that promote overt leukemogenesis.

Our results demonstrate that failure of cell competition in Lmo2-transgenic thymocytes promotes mutations that result in development of pre-LSCs, which are then complemented by further mutations en route to overt leukemogenesis. To explore the nature of these cooperating genetic events, we serially transplanted thymocytes from preleukemic (8–10 wk old) Lmo2-transgenic mice into irradiated recipients, with secondary recipients then left to develop leukemia (Fig. 7 A). Exome sequencing was then performed using tail DNA (as a germline reference), preleukemic thymocytes (n = 4), and leukemia derived from secondary transplants (n = 17 in total from two donors). These data were then analyzed to determine (i) the genomic landscape of Lmo2-induced T cell leukemia and (ii) the ontogeny of secondary mutations to distinguish early- and late-arising events. Notably, secondary transplant recipients developed leukemia at a rate comparable with that arising in Lmo2-transgenic mice (Fig. 7 B).

Analysis of genomic copy number alterations (CNA) revealed trisomy of chromosome 15 and loss of inactive X chromosome in Lmo2-transgenic thymi, which persisted into the derivative leukemias (Fig. 7 C). While these were the only chromosome-level CNAs in Lmo2-transgenic thymocytes, leukemias acquired additional chromosomal changes including trisomy of chromosomes 11 and 14 (Fig. 7 C).

Single nucleotide variant (SNV) analysis revealed few mutations in the Lmo2-transgenic thymi (median 18, range 7–25) and an increased frequency in derived leukemias (median 33, range 3–62; for example, Fig. 7 D). This also revealed interesting details regarding the timing of Notch mutations in this model. Subclonal insertion mutations in exon 34 of Notch1 were found in three of four Lmo2-transgenic thymocyte samples (Fig. 7 E) and increased in penetrance in leukemias (Fig. 7, D and E). These caused frameshift mutations in the PEST domain, loss of which causes stabilization of NOTCH1 protein but not constitutive activation (Koch and Radtke, 2011). Moreover, donor 4 contained two clones with distinct mutations, each giving rise to a subset of derivative leukemias (Fig. 7 E). In addition to exon 34 mutations, four mutations were detected in exon 26/27 of Notch1, which encodes the heterodimerization domain, across six leukemias (Fig. 7 E). These mutations are thought to cause constitutive receptor activation (Koch and Radtke, 2011), and in contrast to exon 34 mutations, these were absent in preleukemic thymocytes.

In addition to Notch1, two other genes were recurrently mutated: Trp53 and Pten, both of which are established tumor suppressors in human T-ALL (Liu et al., 2017). In all, 12 different Trp53 mutations were detected across 10 leukemias, and in many cases, the variant allele frequency exceeded 50%, indicating loss of the WT allele (Fig. 7, D and E). Pten mutations were detected in three leukemias, and a further two had evidence of focal copy number loss (Fig. 7, E and F). Like Notch1 exon 26/27 mutations, Trp53 and Pten mutations were absent in donor thymocytes and mostly unique, implying that they were acquired late during oncogenesis.

These results imply that exon 34 Notch1 mutations are a feature of pre-LSCs in this model. To further analyze this, we studied our previously acquired RNA-seq data from Lmo2-transgenic pre-LSCs (Fig. 6). This showed that exon 34 Notch1 mutations were also seen in pre-LSCs from three separate mice and increased in penetrance upon transplantation (Fig. S4 H). In contrast, Notch mutations were not seen in DN3a thymocytes from two Lmo2;BCL2 double transgenic mice (Fig. S4 H). These results imply a model whereby impaired cell competition promotes the selection of stabilizing mutations in the Notch1 PEST domain that facilitates the development of pre-LSCs (Fig. 7 G). These are then complemented by further mutations in tumor suppressors (in particular Pten and Trp53) to facilitate the development of overt leukemia.

Cellular self-renewal is a key driver of cancer progression. Whilst LMO2 is one of a group of oncogenic T-ALL–causing transcription factors that can transform thymocytes into self-renewing pre-LSCs (Gerby et al., 2014; McCormack et al., 2010; Shields et al., 2019), the cellular and molecular events leading to the development of pre-LSCs in Lmo2-transgenic mice were unclear.

In our previous studies, we used a lineage-tracing approach to show that thymic import from the BM was impaired from an early age (McCormack et al., 2010). Together with our findings that Lmo2-transgenic thymocytes possessed serial transplantation capacity from a young age, this led us to conclude that LMO2 directly induced self-renewal of thymocytes and that these self-renewing pre-LSCs then blocked the import of thymus-settling progenitors (Curtis and McCormack, 2010; McCormack et al., 2010). In contrast, here we have used a combination of neonatal thymus transplantation, competitive BM transplantation, and intrathymic transplantation of LMPPs to show that Lmo2-transgenic thymocytes do not initially show long-term engraftment/self-renewal capacity, but rather show a severe block at the DN2 stage of T cell development. These developmentally arrested cells are quiescent and show an ETP-like gene expression profile, but do not expand in vivo nor show long-term engraftment capacity. However, the developmental arrest of these DN2 cells results in a failure of downstream cell competition, which promotes outgrowth of cells with genomic abnormalities, including Trisomy 15 and Notch1 PEST domain mutations, resulting in the emergence of self-renewing pre-LSCs (Fig. S5). This selection/mutation process is remarkably rapid, inevitably leading to the development of one or more populations of clonally expanded pre-LSCs that are developmentally arrested at the DN3a stage, show long-term engraftment capacity, and are marked by rearranged TCR-β subunits by 2 mo of age. These pre-LSCs then acquire further mutations, in particular loss or mutation of tumor suppressor genes including Pten and Trp53, leading to overt leukemia.

Our results are reminiscent of studies showing that impaired cell competition in the thymus can result in thymocyte self-renewal (Boehm, 2012; Martins et al., 2012; Peaudecerf et al., 2012) and that this can cause T cell leukemogenesis (Martins et al., 2014). More recent studies have examined the consequences of failed thymocyte competition in the setting of impaired progenitor import (Ramos et al., 2020), identifying an abnormal population that is the putative cell-of-origin of leukemia in this setting (Paiva et al., 2021). It will be interesting to determine whether a similar cell population arises in the thymi of Lmo2-transgenic mice to give rise to pre-LSCs.

Recently, a similar mechanism of impaired cellular competition leading to leukemia was found to occur in a mouse model of gene therapy for X-SCID (Ginn et al., 2017; Ginn et al., 2010; Ginn et al., 2018). As LMO2 was preferentially targeted by retroviral insertion in gene therapy trials for X-SCID (Hacein-Bey-Abina et al., 2008; Hacein-Bey-Abina et al., 2003; Howe et al., 2008; McCormack and Rabbitts, 2004), our results imply that the resultant LMO2 overexpression may have exacerbated the accumulation of developmentally arrested progenitors in the thymi of these patients, as well as promoting self-renewal of downstream progenitor populations in conjunction with NOTCH1 mutations.

These findings have implications for the ontogeny of T-ALL in humans, which in the majority of cases appear to evolve from pre-LSCs (Roels et al., 2020). They indicate that LMO2 is insufficient to initiate self-renewal of pre-LSCs, as assumed previously (Cleveland et al., 2013; Curtis and McCormack, 2010; McCormack and Curtis, 2010), but can facilitate as well as cooperate with mutations in Notch1 to do so (Fig. S5). As such, overexpression of LMO2 can contribute to leukemogenesis either as an initiating event, as in Lmo2-transgenic mice, or insertional activation during X-SCID gene therapy trials or as a secondary event, as was seen in the setting of progenitor deprivation (Martins et al., 2014), in which LMO2 was upregulated late during leukemogenesis following thymus autonomy.

Notch1 mutations have been previously reported in transgenic mouse models of T-ALL (Göthert et al., 2007; Lin et al., 2006), and it has been suggested that these occur in the preleukemic phase (Tatarek et al., 2011), in line with leukemia-associated NOTCH1 alleles being weak tumor initiators (Chiang et al., 2008). Using exome sequencing, we show that in Lmo2–transgenic mice, Notch1-stabilizing (PEST domain) mutations occur in the preleukemic phase; however activating (transmembrane) mutations occur late during oncogenesis. In human T-ALL however, NOTCH1 mutations have been found to occur either as early (prenatal) or as late events (Eguchi-Ishimae et al., 2008; Mansour et al., 2007). Irrespective of the order of mutation, our studies indicate that cooperation of the LMO2 and NOTCH pathways is essential to facilitate the development of self-renewing pre-LSCs that can initiate T-ALL.

We have recently used a repressible model of LMO2 overexpression to demonstrate that continued expression of LMO2 is essential for survival and self-renewal of pre-LSCs in vivo (Abdulla et al., 2021). Thus, LMO2 drives a self-renewal program that is essential for pre-LSC self-renewal but requires additional mutations, perhaps to allow cell survival. In contrast to pre-LSCs, we and others have found that overt leukemia stem cells (LSCs) in LMO2-driven T-ALL models often do not require LMO2 for sustained self-renewal, indicating that subsequent mutations or epigenetic events overcome the requirement of LMO2 (Abdulla et al., 2021; García-Ramírez et al., 2018). In particular, we found loss of the tumor suppressor Ikaros can do so (Abdulla et al., 2021). In this study, no mutations were found in Ikaros; however, additional mutations in tumor suppressors (Pten and Trp53) were found. As loss of either of these tumor suppressors could not directly enable the transition to LMO2 independence (Abdulla et al., 2021), it will be interesting to determine whether leukemias in the CD2-Lmo2 transgenic model used here are also LMO2-independent and if so, to define the molecular mediators of this process.

An interesting finding of this study is that the presence of WT progenitors in the thymus can potently inhibit Lmo2-driven T cell leukemogenesis (Fig. 5). This implies that Lmo2 overexpression in a majority of T cell progenitors is necessary to impair cell competition and promote T-ALL (Fig. S5). Indeed, ubiquitous expression of a transgene within one or more lineages is a general feature of transgenic mouse models, which may impair cell competition and promote oncogenesis as shown here. In contrast, during retroviral transduction/transplantation studies, transduction is incomplete, thus it is interesting that LMO2 can also cause leukemogenesis in this setting (Sincennes et al., 2016; Treanor et al., 2011). However, retroviral vectors can cause retroviral insertion mutagenesis, which can itself cause a developmental block at the DN2 stage (Zhou et al., 2016). Thus, either increased expression of LMO2, caused by retroviral overexpression, or activation of cooperating genes by insertion mutagenesis may facilitate leukemogenesis in these studies.

In addition to WT progenitors, we found that overexpression of the antiapoptotic factor BCL2 potently inhibited the emergence of pre-LSCs containing Notch1 mutations and subsequent T-ALL development. This correlated with complete restoration of the DN3-4 compartment in competitive BM transplant assays, supporting previous studies that showed that DN3 is the major stage at which cell competition occurs in the thymus to prevent T cell leukemogenesis (Martins et al., 2014). Together with our finding of increased apoptosis in DN2 thymocytes in Lmo2-transgenic mice, this implies that a key role of secondary mutations, in particular in the Notch1 pathway, is to overcome LMO2-induced thymocyte apoptosis and allow the development of self-renewing pre-LSCs that are the origin of leukemia in this model (Fig. S5). Indeed, thymocyte apoptosis is also induced by the related transcription factors TAL1 and LMO1 as a prelude to leukemia (Herblot et al., 2000; Shank-Calvo et al., 2006) and is also a key mediator of thymic lymphoma caused by γ-irradiation (Michalak et al., 2010). These findings also imply that therapies that induce thymocyte apoptosis may inadvertently promote T-ALL.

In summary, we have found that impaired cell competition is critical for the emergence of pre-LSCs and subsequent leukemogenesis in a transgenic mouse model of T-ALL, implying that this process may play a role in human T-ALL leukemogenesis. Notably, the effects of impaired competition on downstream thymocyte populations are starting to be appreciated (Martins et al., 2014; Paiva et al., 2021; Ramos et al., 2020), and it will be interesting to determine how these processes cooperate with oncogenic transcription factors to cause leukemia in transgenic mouse models as well as human T-ALL.

CD2-Lmo2 (Larson et al., 1994) and Vav-BCL2 (Ogilvy et al., 1999) transgenic mice were maintained on a C57BL/6 background. WT C57BL/6 mice (CD45.1 or CD45.2) were supplied by Walter and Eliza Hall Institute (WEHI) Bioservices. All procedures were approved by WEHI Animal Ethics Committee.

For thymocyte transplantation assays, C57BL/6 (CD45.1) congenic recipient mice were sublethally irradiated (6.5 Gy) and injected via the tail vein with thymocytes equivalent to one-quarter of the thymus as described previously (McCormack and Curtis, 2010).

For BM transplantation assays C57BL/6 (CD45.1), congenic recipient mice were lethally irradiated (9.5 Gy) and injected intravenously with 10⁷ BM cells (CD45.2). For competitive BM transplantation assays, mice received test BM and WT BM (CD45.1) at a 3:1 ratio.

For intrathymic transplantation assays, C57BL/6 (CD45.1) recipients were sublethally irradiated (6.5 Gy), anesthetized, and intrathymically injected with 5,000 sorted LMPPs.

Cells were stained using fluorophore-conjugated antibodies purchased from either eBiosciences, Biolegend, BD Pharmingen, or WEHI Monoclonal Antibody Production Facility. For lineage depletion, BM cells were stained with biotinylated anti-CD2, CD3, Thy1.2, Gr1, Ter119, B220, CD19, and Mac1 and depleted using MACS LS columns (Miltenyi Biotec). To distinguish apoptotic/necrotic cells in stained samples, FITC Annexin V and propidium iodide (final concentration of 2.5 μg/ml) were added 1–10 min prior to acquiring the data. For cell cycle analysis, cells were fixed and permeabilized using Cytofix/Cytoperm (BD Pharmingen) and then stained with anti-Ki67 FITC (B56; BD) and 10 µg/ml DAPI. TCR Vß-subfamily usage was assessed using the Mouse Vß TCR Screening Panel (BD Pharmingen). Cells were analyzed on an LSR Fortessa or FACSymphony A3 (BD) and sorted on a FACS Aria II (BD). Data were analyzed using FlowJo (Treestar; version 9.9.5).

RNA was extracted using RNeasy Plus Micro (Qiagen) and analyzed using a 2200 Tapestation Analyser (Agilent). Libraries were then generated using TruSeq RNA Library Prep v2 (Illumina). Sequencing was performed on a HiSeq 2000 (Illumina) to produce 100-bp single-end reads. Between 7 and 12 million reads were generated for each sample, and reads were aligned to the Mus musculus genome (mm10) using Rsubread (Liao et al., 2019). The number of reads overlapping mouse Entrez genes was summarized using featureCounts and Rsubread’s built-in National Center for Biotechnology Information gene annotation. Low expressed genes were filtered if they failed to achieve at least 0.5 counts per million in at least three libraries. Genes without annotation were also filtered. Differential expression analyses were undertaken using the edgeR (Robinson et al., 2010) and limma (Ritchie et al., 2015) software packages. Library sizes were normalized using the trimmed mean of M-values method (Robinson and Oshlack, 2010). Log2 fold-changes were computed using Voom (Law et al., 2014). Differential expression was assessed relative to a fold change threshold of 1.5 using TREAT (McCarthy and Smyth, 2009). The false discovery rate was controlled below 0.05 using the method of Benjamini and Hochberg. Barcode plots illustrating the enrichment of interested pathway genes were drawn using limma’s barcode plot function. Gene set enrichment tests used the camera method (Wu and Smyth, 2012).

Analysis of Notch1 mutations was performed using Integrative Genomics Viewer (Broad Institute) version 2.4.

Genomic DNA was extracted using the AllPrep DNA/RNA Mini kit (Qiagen) and analyzed with the Bioanalyzer DNA 1000 chip (Agilent 2200; TapeStation instrument). DNA libraries were prepared from 100 ng DNA using SureSelectXT2 Reagent kit (Agilent) with purifications between steps carried out with Agencourt AMPure XP beads. Library yields were estimated with the Qubit dsDNA HS kit (Invitrogen). Hybridization of pooled libraries to capture probes and removal of nonhybridized molecules were then carried out. Targeted regions were amplified, producing sequence-ready libraries that were sequenced on an HiSeq 2000 sequencer (Illumina) at the Australian Genome Research Facility.

The sequenced libraries were aligned to mm10 using Burrows-Wheeler Alignment (Li and Durbin, 2009), resulting in a mean read depth of 29× (range 6–59×) over the capture regions for the cohort. Preliminary variant calling was carried out with Varscan2 (Koboldt et al., 2012). SuperFreq (Flensburg et al., 2020) was used for somatic SNV and CNA calling and for tracking mutations across samples. Two normal control (tail DNA) samples were used to designate germline and somatic variants. Somatic mutation profiles were used to match leukemia samples with donors. Leukemias were designated Donor 3 (D3 L1–L8) or Donor 4 (D4 L1–L9) based on clustering, with one sample removed due to poor quality. To assess mutation load, we considered variants with SomaticP scores >0.5, with lower stringency used to assess subclonal mutations in Notch1, Trp53, and Pten. Somatic variants within 20 Mb of the Ptprc gene (CD45.1) were removed. To visualize CNAs, we plotted calls larger than 20 Mb, and any smaller segments were merged with their largest neighbor.

Graphs and survival curves were plotted using GraphPad Prism with Log-rank Mantel-Cox test used to compare survival significance.

Fig. S1 shows phenotypic evolution, TCR-β expression, and cell cycle analyses of developmentally arrested thymocytes in Lmo2-transgenic mice. Fig. S2 shows kinetic studies detailing the emergence of transplantable, clonal thymocyte populations in Lmo2-transgenic mice. Fig. S3 shows thymocyte analyses and competitive BM transplantation studies demonstrating that BCL2 overexpression restores T cell development and competitive fitness of T cell progenitors in Lmo2-transgenic mice. Fig. S4 shows cell cycle, TCR-β expression, thymocyte transplantation, and Notch1 mutation analyses, demonstrating that pre-LSCs do not develop in Lmo2;BCL2 double transgenic mice. Fig. S5 shows a model detailing the proposed role of impaired cell competition at the DN3 stage in the emergence of pre-LSCs in Lmo2-transgenic mice, as well as the inhibition of this process in the presence of BCL2 or WT lymphoid precursors. Table S1 shows a summary of the number of differentially expressed genes between Lmo2-transgenic and WT thymocyte populations. Tables S2, S3, and S4 show the differentially expressed genes between Lmo2-transgenic and WT young (2–3 wk old) DN2 thymocytes, young (2–3 wk old) DN3 thymocytes, and old (6–10 wk old) DN3 thymocytes, respectively.

RNA-seq data are available in NCBI Gene Expression Omnibus (accession GSE188225) and exome sequencing data are available in NCBI Bioproject (accession PRJNA909315).

We thank Phillippe Bouillet (Walter and Eliza Hall Institute of Medical Research, Parkville, Australia), for Vav-BCL2 transgenic mice, Adam Uldrich (Peter Doherty Institute for Infection and Immunity, Parkville, Australia) for technical assistance, and Samantha Ginn and Ian Alexander (Children’s Medical Research Institute, Westmead, Australia) for discussions.

This work was supported by an Alan W. Harris Scholarship and an Australian Postgraduate Award (to H.D. Abdulla) from the Australian Government, project grants (1003391 to M.P. McCormack, 1187367 to D.H.D. Gray, and 1104145 to M.P. McCormack and I.J. Majewski), a Program grant (1113577 to W.S. Alexander), Fellowships (1154970 to G.K. Smyth, 1158024 to D.H.D. Gray, and 1058344 to W.S. Alexander), and the Independent Research Institute’s Infrastructure Support Scheme from the National Health and Medical Research Council of Australia, a grant-in-aid from the Cancer Council of Victoria, a Future Fellowship from the Australian Research Council (to M.P. McCormack), and a Victorian State Government Operational Infrastructure Support grant.

Authorship contributions: H.D. Abdulla, R. Alserihi, W. Abeysekera, M.X. Luo, and C. Flensburg performed research; H.D. Abdulla, R. Alserihi, W. Abeysekera, M.X. Luo, and C. Flensburg analyzed data; X. Liu, D.H.D. Gray, G.K. Smyth, W.S. Alexander, I.J. Majewski, and M.P. McCormack supervised research; and H.D. Abdulla, C. Flensburg, X. Liu, D.H.D. Gray, G.K. Smyth, W.S. Alexander, I.J. Majewski, and M.P. McCormack wrote the paper.