Distinct Acute Lymphoblastic Leukemia (ALL)-associated Janus Kinase 3 (JAK3) Mutants Exhibit Different Cytokine-Receptor Requirements and JAK Inhibitor Specificities*

Background: JAK3, a tyrosine kinase associated to cytokine receptors, is frequently mutated in leukemia. Results: JAK3L857P induces constitutive signaling independently of cytokine receptors and JAK1, in contrast to other JAK mutants. Conclusion: Different JAK mutants signal through distinct mechanisms and show different sensitivity to JAK1- or JAK3-specific inhibitors. Significance: Depending on the JAK residue mutated, patients will require different treatments. JAK1 and JAK3 are recurrently mutated in acute lymphoblastic leukemia. These tyrosine kinases associate with heterodimeric cytokine receptors such as IL-7 receptor or IL-9 receptor, in which JAK1 is appended to the specific chain, and JAK3 is appended to the common gamma chain. Here, we studied the role of these receptor complexes in mediating the oncogenic activity of JAK3 mutants. Although JAK3V674A and the majority of other JAK3 mutants needed to bind to a functional cytokine receptor complex to constitutively activate STAT5, JAK3L857P was unexpectedly found to not depend on such receptor complexes for its activity, which was induced without receptor or JAK1 co-expression. Introducing a mutation in the FERM domain that abolished JAK-receptor interaction did not affect JAK3L857P activity, whereas it inhibited the other receptor-dependent mutants. The same cytokine receptor independence as for JAK3L857P was observed for homologous Leu857 mutations of JAK1 and JAK2 and for JAK3L875H. This different cytokine receptor requirement correlated with different functional properties in vivo and with distinct sensitivity to JAK inhibitors. Transduction of murine hematopoietic cells with JAK3V674A led homogenously to lymphoblastic leukemias in BALB/c mice. In contrast, transduction with JAK3L857P induced various types of lymphoid and myeloid leukemias. Moreover, ruxolitinib, which preferentially blocks JAK1 and JAK2, abolished the proliferation of cells transformed by the receptor-dependent JAK3V674A, yet proved much less potent on cells expressing JAK3L857P. These particular cells were, in contrast, more sensitive to JAK3-specific inhibitors. Altogether, our results showed that different JAK3 mutations induce constitutive activation through distinct mechanisms, pointing to specific therapeutic perspectives.

JAK1 and JAK3 are recurrently mutated in acute lymphoblastic leukemia. These tyrosine kinases associate with heterodimeric cytokine receptors such as IL-7 receptor or IL-9 receptor, in which JAK1 is appended to the specific chain, and JAK3 is appended to the common gamma chain. Here, we studied the role of these receptor complexes in mediating the oncogenic activity of JAK3 mutants. Although JAK3 V674A and the majority of other JAK3 mutants needed to bind to a functional cytokine receptor complex to constitutively activate STAT5, JAK3 L857P was unexpectedly found to not depend on such receptor complexes for its activity, which was induced without receptor or JAK1 co-expression. Introducing a mutation in the FERM domain that abolished JAK-receptor interaction did not affect JAK3 L857P activity, whereas it inhibited the other receptor-dependent mutants. The same cytokine receptor independence as for JAK3 L857P was observed for homologous Leu 857 mutations of JAK1 and JAK2 and for JAK3 L875H . This different cytokine receptor requirement correlated with different functional properties in vivo and with distinct sensitivity to JAK inhibitors. Transduction of murine hematopoietic cells with JAK3 V674A led homogenously to lymphoblastic leukemias in BALB/c mice. In contrast, transduction with JAK3 L857P induced various types of lymphoid and myeloid leukemias. Moreover, ruxolitinib, which preferentially blocks JAK1 and JAK2, abolished the proliferation of cells transformed by the receptor-dependent JAK3 V674A , yet proved much less potent on cells expressing JAK3 L857P . These particular cells were, in contrast, more sensitive to JAK3specific inhibitors. Altogether, our results showed that different JAK3 mutations induce constitutive activation through distinct mechanisms, pointing to specific therapeutic perspectives.
JAK3 is a member of the Janus kinase (JAK) family, which comprises three other tyrosine kinases: JAK1, JAK2, and TYK2. These intracellular kinases are associated with a wide variety of cytokine receptors that lack intrinsic kinase activity. Although the three other JAKs are ubiquitously expressed and can bind various receptor chains, JAK3 is exclusively associated with the common gamma chain (␥c), 5 and both genes are preferentially expressed in hematopoietic cells. ␥c partners with a specific alpha (or beta) chain that binds JAK1 to form the heterodimeric receptor complexes for IL-2, IL-4, IL-7, IL-9, IL-15, and IL-21 cytokines, key players of lymphoid development. For the IL-2 receptor complex, IL-2 stimulation revealed JAK1 to be responsible for the phosphorylation of JAK3 and STAT5, with the role of JAK3 being to phosphorylate JAK1 (1).
The JAKs possess three functional domains that are well conserved. The kinase domain is the catalytic domain of the tyrosine kinases, located at the C terminus. Adjacent to this is the pseudokinase domain, which presents a kinase-like sequence yet lacks critical residues required for proper catalytic activity. This domain is supposed to prevent inappropriate activation of the kinase domain in the absence of cytokine stimulation (2). The N terminus contains the FERM (band-4.1, ezrin, radixin, moesin) domain. This domain mediates direct JAK interaction with the intracellular portion of cytokine receptors (3). Mutations disturbing the structural integrity of the extreme N terminus of the FERM domain are interfering with ␥c association, as illustrated, for example, by the Y100C mutation of JAK3 (4).
The JAK3 Y100C mutation was described in patients suffering from severe combined immunodeficiency (SCID) (5). This germ line JAK3 inactivating mutation abrogates the ␥c-dependent cytokine signaling essential for lymphoid cell development, leading to a profound defect in T, B, and NK cells. In contrast to these inactivating mutations, acquired JAK3 activating point mutations have been described in various leukemia types, such as T cell acute lymphoblastic leukemia (T-ALL) (6), acute megakaryoblastic leukemia (7), or juvenile myelomonocytic leukemia (8), resulting in constitutive JAK activation and STAT phosphorylation.
The severe effect of JAK3 signaling impairment specifically impacting on lymphoid development in SCID patients has instigated the development of immunosuppressive drugs targeting this kinase. Tofacitinib, a type I JAK inhibitor targeting JAK3 (1), is currently an approved treatment for rheumatoid arthritis in the United States. This drug is being assessed for use in the treatment of T cell-specific autoimmune diseases, such as psoriasis, multiple sclerosis, and inflammatory bowel disease, as well as for the prevention of organ transplant rejection (9, 10). Although originally described as a selective JAK3 inhibitor, tofacitinib also inhibits JAK1, and its efficacy is likely due to combined inhibition of both kinases (1,11). On the other hand, JAK inhibitors are also currently being investigated for the treatment of hematological malignancies caused by abnormal JAK activation. Ruxolitinib, a JAK1/2 type I inhibitor, is currently administered in the treatment for myeloproliferative neoplasms, which are frequently associated with the V617F activating mutation in JAK2 (12). In line with the results observed with JAK inhibitor use in myeloproliferative neoplasm treatment, these inhibitors represent an appealing therapeutic strategy for leukemia cases associated with JAK3 activation.
This study sought to gain more insight into the potential use of JAK inhibitors in mutated JAK3-driven malignancies. We therefore decided to first determine whether ALL-associated JAK3 mutants need to bind ␥c to be constitutively active and then to test the sensitivity of different mutants to JAK inhibitors.
Virus Production, Stable DNA Transfections, and Analysis of Transfected Cells-Retroviral supernatants were generated by transient transfection of HEK293T cells with the packaging plasmid pCL-Eco (coding for gag/pol/env) and biscistronic vectors encoding the gene of interest and human CD4 as marker and used for infection of 0.5 ϫ 10 6 Ba/F3 cells. HEK293 cells were transfected with 3.75 g of pCL-Eco and 3.75 g of the different constructs with 22.5 l of TransIT-LT1 transfection reagent (Mirus). The next day, supernatants were collected and added to Ba/F3 cells for the spin infection (2-h incubation at 1000 ϫ g at room temperature). This last step was repeated the day after to increase the transfection level. Expression of the markers was analyzed by FACS (BD FACSCalibur TM flow cytometer), using a phycoerythrin-coupled antibody against hCD4 (catalog no. 555347; BD Pharmingen) diluted 1/30. For murine bone marrow cell transduction, mutated JAK3 pMSCV-GFP viral vectors were produced in HEK293T cells using an EcoPack packaging plasmid and TurboFect transfection reagent (Fermentas). Virus was harvested after 48 h.
Dual Luciferase Assay-STAT5 transcriptional activity was assessed by measurements of luciferase expression in HEK293 cells upon transient transfection of appropriate cDNA constructs and pLHRE-luc vector harboring tandem copies of the STAT5-inducible lactogenic hormone response element (LHRE) of the rat ␤-casein promoter, inserted upstream from a luciferase gene (14). Another reporter plasmid, Renilla luciferase (pRLTk; Promega), was co-transfected as an internal transfection control. Transient transfection of HEK293 cells by Lipofectamine (Invitrogen) was previously described (15). 24 h after transfection, luciferase assays were performed using the dual luciferase reporter assay system (Promega).
Murine Bone Marrow Transplantation-Balb/c mice were purchased from Charles River Laboratories. Harvest of bone marrow cells from male donor mice, lineage negative cells enrichment, transduction with viral supernatant, and injection into irradiated female recipient mice was performed as previously described (13). After sacrifice of the mice, single-cell suspensions were prepared from peripheral blood, bone marrow, spleen, thymus, and lymph nodes. Cells were analyzed on a FACSCanto flow cytometer (BD Biosciences).

ALL-associated JAK3 Mutants Signal through a Functional
Cytokine Receptor Complex, Except the L857P Mutant-More and more patients suffering from different types of leukemia, such as ALL, acute megakaryoblastic leukemia, or juvenile myelomonocytic leukemia, have been reported to exhibit mutations in the JAK3 gene (6 -8, 18, 19). These mutations were often characterized as activating (7, 18 -20), yet little is known about their activation mechanism. Of the other JAK kinases, JAK2 V617F, found in 95% of polycythemia vera cases, has been shown to be dependent on homodimeric receptors, like the erythropoietin receptor, to become active (20). Similar results were obtained for the activating JAK1 V658F and A634D mutants, which needed to bind to IL-9 receptor alpha or IL-2 receptor beta to form homo-or heterodimeric cytokine receptor complexes able to mediate constitutive signaling (21). Furthermore, the active JAK3 A572V mutant requires heterodimeric complexes formed by IL-9 receptor alpha or IL-2 receptor beta and ␥c (22). We sought to determine whether ALL-associated JAK3 mutants shared similar receptor-dependent mechanisms of constitutive signaling. To this end, we studied four activating JAK3 mutants scattered throughout the kinase and pseudokinase domains, namely M511I, A572V, V674A, and L857P identified in patients with T-ALL. 6 We took advantage of HEK293 cells that are deficient in JAK3 and ␥c, the only cytokine receptor chain known to bind JAK3. WT JAK3 and the four mutants were transiently transfected into HEK293 cells, in combination with a STAT5-responsive reporter construct. This transfection was carried out both in the absence of IL-9R␣ or ␥c and in their presence to reconstitute a functional IL-9R complex. As shown in Fig. 1, the four tested JAK3 mutants induced constitutive STAT5 activation, although for the M511I, A572V, and V674A mutants, this was achieved only in the presence of a functional IL-9R complex (␥c and IL-9R␣). Surprisingly, JAK3 L857P induced STAT5 activation even without co-expression of any receptor chain. This observation suggested a receptor-independent activity that was specific to JAK3 L857P . An alternative hypothesis would be the involvement of another cytokine receptor chain that could bind JAK3 and would be expressed in HEK293. To exclude the latter hypothesis, we tested the effect of a SCID mutation localized in the FERM domain of JAK3 (Y100C) that inactivates the JAK by disabling its binding to ␥c (4,5). In fact, mutation of this conserved hydrophobic residue in JAK FERM domains disturbs the structural integrity of parts of this FERM domain, which plays an essential role in all JAKcytokine receptor interactions, as also shown for the homologous mutations of JAK1 (Y107A) (3) and JAK2 (Y114A) (23). We introduced the Y100A mutation into a WT, V674A, and L857P JAK3 and evaluated the capacity of each mutant and double mutant to mediate constitutive signaling. As illustrated in Fig. 2A, the JAK3 V674A activity in the presence of the IL-9R complex was abrogated by introducing the Y100A mutation, thereby confirming the essential role of Tyr 100 for receptor binding. This observation contrasted with the results obtained for JAK3 L857P for which introducing the Y100A mutation did 6 M. Tartaglia, unpublished data. not abrogate its activity. On the contrary, increased constitutive signaling in the absence of a functional receptor was observed. Notably, the expression level of each JAK3 mutant was similar under all experimental conditions (Fig. 2B).
To confirm the lack of ␥c requirement for constitutive JAK3 L857P signaling in hematopoietic cells, we down-regulated the expression of endogenous murine ␥c in growth factor-independent Ba/F3 cells obtained after stable transduction with JAK3 V674A , JAK3 L857P , and JAK3 L857P/Y100A or in IL-3-dependent control Ba/F3 cells. Fig. 2C shows that siRNA-mediated knockdown of ␥c resulted in a significant decrease of the proliferation of Ba/F3 cells transformed with JAK3 V674A compared with the proliferation observed after control siRNA treatment, whereas no decrease in proliferation was observed for cells transformed by JAK3 L857P or JAK3 L857P/Y100A . These results confirm the unique capacity of JAK3 L857P to induce cytokinereceptor independent constitutive signaling.
Cytokine receptor complexes are formed of two receptor chains, each associated with one JAK. In ␥c-associated receptor complexes, such as IL-7 or IL-9 receptor, JAK3 and JAK1 are believed to interact and cross-phosphorylate each other to enable signaling. To assess the role of JAK1 in mutant JAK3induced signaling, we took advantage of JAK1-deficient U4C cells. As shown in Fig. 3A, in the absence of JAK1, the JAK3 V674A was unable to activate STAT5, even when cytokine receptor chains were expressed. This activity was restored by co-expression with JAK1 WT , but not with a kinase-dead (KD) JAK1. The results obtained for JAK3 L857P , shown in Fig. 3B, indicated that the activity of this mutant is entirely independent of JAK1, once again supporting the hypothesis of a cytokine receptor-independent activation mode.
To confirm the lack of JAK1 requirement for constitutive JAK3 L857P signaling in hematopoietic cells, endogenous JAK1 knockdown was performed as described for ␥c (Fig. 3C). This knockdown decreased the proliferation of Ba/F3 cells transformed with JAK3 V674A , whereas it did not alter the proliferation of cells transformed by JAK3 L857P or JAK3 L857P/Y100A . Taken together, these results revealed JAK3 L857P as possessing a unique characteristic: the capacity to activate signaling in the absence of binding to a cognate cytokine receptor complex. This contrasts with the other tested JAK3 mutants, as well as all JAK1 and JAK2 activating mutants described in the literature (20,21).  Given that the leucine 857 of JAK3 was conserved among all JAK kinases, we sought to determine whether mutations of the homologous residue in JAK1 (Leu 910 ) and JAK2 (Leu 884 ) could confer a similar receptor-independent constitutive signaling capacity. Those mutations were introduced together with the FERM domain mutations Y107A in JAK1 or Y114A in JAK2 (homologous to Y100A in JAK3). As illustrated in Fig. 4, upon transient expression in HEK293 cells, JAK1 L910P/Y107A and JAK2 L884P/Y114A were capable of activating STAT5, demonstrating that the activating mutations of a leucine residue conserved among JAK kinases share the capacity of inducing STAT5 activation independently of the receptor-FERM domain interaction.
To gain more insight about potential mechanisms enabling receptor-independent constitutive activation of JAK3 L857P , we  substituted Leu 857 with different residues and tested the receptor-independent activity of these JAK3 mutants in HEK293 cells, using JAK3 cloned in either pMX-IRES-CD4 (Fig. 5A) or pMSCV-GFP (Fig. 5B). Receptor-independent constitutive STAT5 activation was observed after replacing the leucine with a proline, alanine, or glutamine, but not with a glutamic acid, probably because of this residue's negative charge, even if the mutant was still active in presence of IL-9 (data not shown).
Another Kinase Domain Mutant, JAK3 L875H , Exhibits the Same Receptor-independent Activating Capacity as JAK3 L857P -Among the JAK3 mutants studied here, L857P was the only mutant involving the kinase domain, raising the hypothesis that cytokine receptor independence could be explained solely by the mutation's location in the kinase domain, compared with the pseudokinase mutations that would be receptor-dependent. To address this question, other activating mutations of the kinase domain were investigated. These consisted of one JAK3 mutation associated with T cell pro-lymphocytic leukemia (Y824D) (24), three ALL-associated mutants (L875H, P906S, and E958K), 7 and one mutant described in a murine T cell line (T848A) (16). As illustrated in Fig. 6A, four of these five new mutants were dependent on the co-expression of IL-9R␣ and ␥c for their activity in HEK293 cells. In contrast, JAK3 L875H behaved like JAK3 L857P , exhibiting a receptor-independent activating capacity. The close proximity of L857P and L875H mutants to the ␣-helix C of the kinase domain is illustrated in the predicted protein structure of JAK3 kinase domain, represented in Fig. 6B. By contrast, the receptor-dependent mutants map to the interface between pseudokinase and kinase domains, as previously described (16).
The Difference in Cytokine Receptor Requirement Correlates to a Distinct JAK Inhibitor Specificity-Given that JAK3 L857P and the other JAK3 mutants presented different receptor and JAK1 requirements, we sought to determine whether they would exhibit the same sensitivity to ruxolitinib, a JAK1-2 inhibitor. To assess this, we treated growth factor-independent Ba/F3 cells obtained after stable transduction with JAK3 V674A , JAK3 L857P , and JAK3 L857P/Y100A , with increasing ruxolitinib concentrations and measured their proliferation using a tritiated thymidine incorporation assay. We employed Ba/F3 cells that had been stably transduced with the chronic myeloid leukemia-associated BCR-ABL fusion gene as a negative control, choosing this particular oncogene for its lack of sensitivity to JAK inhibitors. Fig. 7A illustrates that Ba/F3 cells transformed with JAK3 V674A were at least 10-fold more sensitive to ruxolitinib than those transformed with JAK3 L857P and JAK3 L857P/Y100A . The calculated IC 50 values are shown 7 J. Cools, unpublished data.  in Table 1. This difference in ruxolitinib sensitivity between the V674A and L857P/Y100A JAK3 mutants was reproduced in transiently transfected HEK293 cells treated with 0.5 M of ruxolitinib, with STAT5 transcriptional activity being used as readout (Fig. 7B). To complement these observations, we monitored the phosphorylation status of STAT5, JAK1, and JAK3 in growth factor-independent JAK3 V674A and JAK3 L857P/Y100A Ba/F3 cells treated with 0.5, 1, and 2 M of ruxolitinib (Fig. 7C). This JAK1-2 inhibitor abrogated STAT5 phosphorylation induced by the V674A mutant, whereas it had a marginal effect on the L857P/Y100A double mutant, thus confirming the difference in ruxolitinib sensitivity. Although able to inhibit STAT5 phosphorylation by JAK3 V674A , ruxolitinib did not inhibit JAK1 phosphorylation, probably because it is a type I kinase inhibitor, blocking the kinase in an active conformation that can still be phosphorylated by JAK3. JAK3 phosphorylation was only observed in the double mutant cells, with no ruxolitinib-induced inhibition observed, in line with the absence of effect seen with STAT5. Altogether, these results demonstrate that the activity of JAK3 V674A can be blocked with a JAK1specific inhibitor, whereas JAK1-independent JAK3 L857P is much less sensitive to ruxolitinib.
We next tested the effect of a JAK3-specific inhibitor (NIBR3049) (17) on the different JAK3 mutants by performing the same experiments as for ruxolitinib. As illustrated in Fig. 8A, the Ba/F3 cells transformed with JAK3 L857P and JAK3 L857P/Y100A proved more sensitive to NIBR3049 than those transformed with JAK3 V674A . The calculated IC 50 values are shown in Table 1. Similar results were obtained for STAT5 activity in HEK293 cells (Fig.  8B). In Ba/F3 cells, the NIBR3049 inhibitor completely blocked STAT5 and JAK3 phosphorylation in the JAK3 L857P/Y100A cells, confirming their sensitivity to this molecule (Fig. 8C). By contrast, NIBR3049 barely affected STAT5 and JAK1 phosphorylation in JAK3 V674A cells and even increased JAK3 phosphorylation because this type I inhibitor blocked JAK3 in the active conformation without affecting the kinase activity of the JAK1 partner. Altogether, these results show that JAK3 L857P activity can be blocked by a JAK3-specific inhibitor, whereas the JAK3 V674A is much less sensitive to NIBR3049. Finally, when an inhibitor active against  both JAK1 and JAK3 was used (tofacitinib), both receptor-dependent and -independent mutants exhibited a similar sensitivity, as illustrated by a decrease in the proliferation of Ba/F3 cells and in STAT5 activity in HEK293 cells (Fig. 9).

JAK3 Mutants Cause Distinct Leukemia Phenotypes in a Murine Hematopoietic Stem Cell Transplant
Model-Degryse et al. (13) reported that mice transplanted with bone marrow  progenitor cells expressing T-ALL-associated JAK3 mutants developed a disease resembling T-ALL, characterized by an accumulation of immature CD8 positive T cells. This phenotype can be explained as the result of constitutive activation of ␥c-associated receptors, which are essential for T cell lymphoproliferation. Given that JAK3 L857P and JAK3 L857P/Y100A exhibit no cytokine receptor restriction in vitro, we wondered whether their expression in hematopoietic stem cells would be leukemogenic and produce the same phenotype in vivo. Hematopoietic progenitor cells were transduced with retroviral vectors expressing JAK3 V674A , JAK3 L857P , or JAK3 L857P/Y100A , together with GFP, and injected into irradiated Balb/c mice. As shown in Table 2, almost all the mice exhibited increased white blood cell counts ranging from 20,000 to 300,000 cells/microliter at sacrifice, within 50 -200 days post-transplant. JAK3 V674A mice displayed an accumulation of CD8 single positive cells in the spleen and bone marrow ( Table 2, bracket A, and Fig. 10A), as well as in the peripheral blood, thymus, and lymph nodes (data not shown). These mice homogenously developed a disease resembling T-ALL, similar to the one described by Degryse et al. By contrast, JAK3 L857P and JAK3 L857P/Y100A were found to induce at least four different types of lympho-or myeloproliferative diseases, characterized by an accumulation of either CD4/8 double positive T-lymphocytes (Table 2, bracket B, and Fig. 10B), Gr1 positive myeloid cells (Table 2, bracket C, and Fig. 10C), CD8 single positive T-lymphocytes (Table 2, bracket D, and Fig. 10D), or B220 positive B-lymphocytes (Table 2, bracket E, and Fig. 10E). These results indicated that JAK3 L857P and JAK3 L857P/Y100A exhibit an oncogenic potential in vivo but induced a more heterogeneous phenotype compared with cytokine receptor-restricted JAK mutants.

Discussion
The originality of our observations resides in the demonstration that cancer-associated activating mutations of different residues in the same JAK kinase induce different functional properties, both in vitro and in vivo. In vitro, different ALLassociated JAK3 mutants were associated with distinct sensitivity to JAK inhibitors, whereas the in vivo transduction of these mutants in hematopoietic precursors produced distinct leukemia phenotypes. This can, for the most part, be explained by the unexpected behavior of JAK3 L857P , which did not require cytokine receptor binding to be active. These findings contrast sharply with what has previously been published about other cancer-associated JAK mutants (20,21), including JAK3 V674A , which required binding to a functional receptor complex to be active.
The different phenotypes observed in vivo can also be explained by the different cytokine receptor requirements of the various mutations tested. The transduction of murine hematopoietic cells with JAK3 V674A led to homogenous lymphoblastic leukemias. Because the oncogenic activity of this mutant is restricted to cells expressing ␥c, downstream constitutive signaling mimics the signaling induced by ␥c-associated receptors like IL-2 receptor, IL-7 receptor, or IL-9 receptor, which are essential for lymphoproliferation. In contrast, transduction with JAK3 L857P induced various types of lymphoid and myeloid leukemias in mice, revealing the existence of a broader oncogenic potential resulting from the absence of cytokine receptor restriction.
The second key observation reported here is that different activating mutations of JAK3 dictate specific pattern of sensitivity to JAK inhibitors. Such different sensitivities to JAK inhibitors could be explained by the model presented in Fig. 11. Cytokine receptor complexes are formed of two chains, each associated with one JAK, namely JAK1 and JAK3 for the ␥c cytokine receptor family (Fig. 11A). JAK3 L857P , although still able to bind to ␥c, does not require this binding to be active. This explains its JAK1-independent activity and accounts for the inefficacy of JAK1 inhibitors, such as ruxolitinib (Fig. 11B). In contrast, a JAK1 inhibitor blocks the activity of classical JAK3 mutants, such as JAK3 V674A , because these mutants are dependent on a functional cytokine receptor complex for signaling. A JAK3-specific inhibitor, such as NIBR3049, effectively inhibits the activity of JAK3 L857P (Fig. 11C). Its relative inefficacy against JAK3 V657A is in line with the role of the kinase activity of JAK3 in the context of IL-2 signaling, as reported by Haan et al. (1). Indeed, these authors demonstrated that although signal transduction by the IL-2 receptor complex required JAK3 expression, it was not abrogated by a kinasedefective JAK3, suggesting that JAK1 kinase activity was suffi-

Expression of JAK3 mutants in bone marrow cells of Balb/c mice leads to distinct leukemia phenotypes
The table shows the days post-transplant at sacrifice, the white blood cells count at sacrifice, and the diagnosis of Balb/c mice receiving a bone marrow transplantation of lineage negative cells expressing JAK3 V674A , JAK3 L857P or JAK3 L857P/Y100A . Five different diagnosis were observed: CD8ϩ T-lymphoproliferation in JAK3 V674A mice (Bracket A), CD4/8ϩ T-lymphoproliferation in JAK3 L857P and JAK3 L857P/Y100A mice (Bracket B), Gr1ϩ myeloproliferation in JAK3 L857P and JAK3 L857P/Y100A mice (Bracket C), another type of CD8ϩ T-lymphoproliferation in JAK3 L857P and JAK3 L857P/Y100A mice (Bracket D), and B220ϩ B-lymphoproliferation in a JAK3 L857P mice (Bracket E).
cient for the complex to be active, with JAK3 serving as a scaffold protein. One hypothesis is therefore that classical mutants, such as JAK3 V674A , modify the conformation of JAK3 as a scaffold in the receptor complex to enable JAK1 constitutive kinase activity.
Our findings provide further evidence that although up to 30% of ALL cases might harbor activating mutations of the IL-7 receptor-JAK1-JAK3 pathway, working with JAK inhibitors to treat such hematological neoplasms will not be a simple task. The selection and dosing of the inhibitor must not only be based on the kinase affected but also on the specific mutation observed. We previously showed that activating mutations affecting Phe 958 and Pro 960 of JAK1 not only promoted autonomous cell proliferation but also conferred resistance to ATPcompetitive inhibitors, suggesting that mutations of the same kinase could require different JAK inhibitor dosages (25). This also proved true for mutations of the JAK2 Tyr 931 residue (25) and, furthermore, applied to the homologous mutation of the Tyr 904 residue of JAK3 (16). Our current study has demonstrated that different JAK inhibitors are required to block dif- ferent JAK3 mutants, depending on their mechanisms of signal transduction activation. Although JAK1 (L910P) and JAK2 (L884P) mutations that are homologous to L857P also induce cytokine receptor-independent constitutive signaling, their effect on JAK inhibitor sensitivity should be less critical, because many JAK2-associated receptors are homodimeric, and the kinase activity of JAK1 appears essential for ␥c-associated receptors.
The mechanisms enabling the receptor-independent constitutive activation of JAK3 L857P are still subject to speculation. For JAK3, receptor-independent constitutive activation is observed after replacing the leucine with different residues (Fig.  5), except for glutamic acid. JAK3 L857Q has been described in ALL patients (13) and is also receptor-independent. The L910Q mutation of JAK1 has already been described as activating in a random mutagenesis screening analysis (26) and introducing the Y107A mutation (homologous to the JAK3 Y100A mutation) in the FERM domain of JAK1 L910Q has been found to not completely abolish STAT5 constitutive phosphorylation. The diversity of these residues suggests that the loss of leucine 857 is the key factor inducing constitutive activation. Interestingly, the leucine 857 of JAK3 is not only conserved among all JAK kinases, but also in other tyrosine kinases, such as c-Src, LCK, c-ABL, EGFR, FGF receptor, and PDGF receptor. The relevance of this residue for kinase activity regulation is illustrated by the EGFR exon 19 insertions leading to the L747P mutation described in patients with lung adenocarcinoma or by the L747S mutation of the EGFR, reported in patients acquiring secondary resistance to EGFR inhibitors. It is believed to shift the equilibrium toward the receptor's active conformation (27)(28)(29). Other arguments point toward the importance of the JAK ␣C helix in the regulation of the cytokine receptor dependence. As illustrated in 4B, the Leu 857 residue is situated in front of the ␣C helix of the N-lobe of the JAK3 kinase domain. This alpha helix is conserved among all kinases and constitutes a key mediator of conformational change between the active and inactive states of the kinase (30). Additionally, another mutation located within the ␣C helix of JAK3, L875H, is present in patients with ALL, exhibiting the same characteristics as L857P, namely constitutive activation and receptor independence.
A further puzzling observation is that when the Y100A mutation was added to the active JAK3 L857P mutant, STAT5 activation was seen to increase in HEK293 cells. A preliminary explanation for this could be that Y100A increases L857P-induced constitutive kinase activity. Previous studies have, in fact, suggested that the FERM domain fulfills a regulatory role in kinase activity through interaction between the kinase and FERM domains (2,31,32). Moreover, electron microscopy JAK1 imaging revealed this protein as exhibiting a high flexibility, able to shuttle from open to closed states (32). In the compact conformation, the FERM domain comes in close proximity to the kinase domain. The Y100A mutation could change these interactions and increase the kinase activity induced by L857P. Another explanation could reside in the subcellular location of the kinase. JAK3 V674A needs to bind to a cytokine receptor to be active, with a kinase activity located close to the plasma membrane. The localization of JAK3 L857P/Y100A is unknown, yet cytokine-receptor independent. The constitutive kinase activity could therefore take place in a location where the activation of downstream signaling is more efficient, such as the cytoplasm. Further studies will be needed to address these issues. Because of their particular mode of action, the cytokine receptor-independent active mutants of JAK kinases described here open new perspectives, not only for treating patients with alterations of the JAK-STAT pathway but also for a better understanding of this pathway.