Rare type 1-like and type 2-like calreticulin mutants induce similar myeloproliferative neoplasms as prevalent type 1 and 2 mutants in mice
Abstract
Frameshift mutations in the calreticulin (CALR) gene are present in 30% of essential thrombocythemia and myelofibrosis patients. The two most frequent mutations are CALR del52 (type 1, approximately 60%) and CALR ins5 (type 2, around 30%), but many other rarer mutations exist accounting each for less than 2% of all CALR mutations. Most of them are structurally classified as type 1-like and type 2-like CALR mutations according to the absence or presence of a residual wild- type calcium-binding motif and the modification of the alpha-helix structure. Yet, several key questions remain unanswered, especially the reason of such low frequencies of these other mutations. In an attempt to investigate specific pathogenic differences between type 1-like and type 2-like CALR mutations and del52 and ins5, we modeled two type 1-like (del34 and del46) and one type 2-like (del19) mutations in cell lines and in mice. All CALR mutants constitutively activate JAK2 and STAT5/3/1 in a similar way in the presence of the thrombopoietin receptor (MPL) and induced cytokine-independent cell growth but to a lesser extent with rare mutants over time. This correlates with reduced expression levels of rare CALR mutants compared to del52 and ins5. Lethally irradiated mice that were engrafted with bone marrow transduced with the different CALR mutations developed thrombocytosis, but to a much lesser extent with ins5 and the type 2-like CALR mutation. In contrast to type 2-like mice, type 1-like mice developed marked myelofibrosis and splenomegaly 10 months after engraftment. Similar to del52, type 1-like CALR mutations induced an expansion at an early stage of hematopoiesis compared to ins5 and type 2-like mutation. Thus, type 1-like and type 2-like CALR mutants structurally and functionally resemble del52 and ins5 mutants, respectively.
Introduction
The acquisition of somatic mutations in the hematopoietic stem cell (HSC) compartment leading to the constitutive activation of the JAK/STAT signaling pathway drives the development of myeloproliferative neoplasms (MPNs) [1, 2].Classical MPNs include three disorders: polycythemia vera (PV) that displays an increased red blood cell count,essential thrombocythemia (ET) and myelofibrosis (MF) both characterized by a megakaryocytic (MK) hyperplasia and an increased platelet count in ET and bone marrow fibrosis and granulocytic proliferation in MF. The gain-of- function mutation in JAK2 (JAK2V617F) is present in the majority of PVs, ETs and MFs. The second most frequent driver mutations inducing approximately 30% of ET and MF are located in the calreticulin (CALR) gene [3, 4].More than 50 different CALR mutations have been reported, the two most frequent ones are a deletion of 52 base pairs (del52) L367fs*46 also called type 1 and an insertion of 5 base pairs (ins5) K385fs*47 known as type 2. They represent together more than 80% of all CALR mutations.These mutants specifically activate the thrombopoietin (TPO) receptor (MPL) in a TPO-independent manner through binding to the N-glycosylated residues of its extracellular domain [5]. This activation requires the new C- terminus domain of CALR mutants [6, 7].In retroviral mouse models, we previously showed that CALR mutants induce thrombocytosis, but to a lesser extent in presence of ins5 compared to del52 [8]. Moreover, del52 mice evolved more frequently into MF than ins5 mice, mimicking what is observed in patients [9, 10]. Indeed, del52 and ins5 are both frequent in ET whereas del52 is largely predominant in MF [11, 10, 12].
Each of the other mutants of CALR represents 0.3–1.7% of CALR mutations in MPNs. Most of them can be classified as type 1-like or type 2-like mutations according to the absence or presence of a residual wild-type calcium-binding motif and the modification of the alpha-helix structure [10, 13]. This classification, based on the structure of the CALR mutants, appears to correlate with the phenotype of the patients since type 1-like mutants, like del52, are more predominant in MF [14]. However, there is no direct demonstration that type 1- like and type 2-like mutants have different pathogenic capa- cities as shown for del52 and ins5 mutants [8] and it could be expected that these rarer mutants might induce a different disease, explaining their very low frequencies in patients compared to del52 and ins5.In this work, we investigated whether the oncogenic activity of two type 1-like mutations, del34 (E369fs*50) and del46 (L367fs*48) and of one type 2-like mutation del19 (K374fs*50) was mediated by MPL, like del52 and ins5. In Ba/F3 cell lines, all mutants activated MPL- dependent JAK/STAT signaling pathway to a similar extent and in a nearly TPO-independent fashion at the exception of del19. Rare CALR mutants induced a smaller cell pro- liferative advantage over time than del52 and ins5, corre- lating with reduced protein levels. We investigated whether the structural-related classification of rarer CALR mutants was confirmed in vivo in mice. We developed retroviral mouse models of del19, del34, and del46. Mice werefollowed for up to 10 months and hematopoiesis was ana- lyzed in bone marrow and spleen and compared to del52, ins5 and wild-type CALR (CALR wt) models. We demon- strate that del34 and del46 mice develop thrombocytosis of similar severity with a progression to MF as del52 mice, whereas del19 phenotype resembles ins5. Unexpectedly, we found no phenotypic differences between rare type 1-like and type 2-like mutants compared to del52 and ins5, respectively.
Results
Rare mutants of CALR have been structurally classified as type 1-like and type 2-like based on structural character- istics with the two most frequent CALR del52 (type 1) and CALR ins5 (type 2) prototypes (Fig. 1a) [10]. In order to functionally test this classification, we modeled three CALR mutations, two type 1-like, del34 and del46, and one type 2- like, del19, in Ba/F3 cell lines that expressed MPL. CALR mutants resulted in cytokine-independent growth of the Ba/ F3 cells expressing MPL. Moreover, a MTT-like assay showed that, independently of the CALR mutant, Ba/F3 cells presented a TPO-independent proliferation at 48 h (Fig. 1b). This was confirmed when looking at cell pro- liferation after 2 days of cultured in absence or presence of TPO (Fig. 1c). However, while the different CALR wt- and CALR mutant-expressing Ba/F3-MPL cell lines continued to present similar WEHI-dependent proliferation rates after 3 and 4 days of culture, the rare type 1-like and type 2-like mutants induced a more moderate autonomous proliferation compared to del52 and ins5. Interestingly, the TPO- dependent cell proliferation in presence of del19 or CALR wt was comparable and higher than in presence of other CALR mutants (Fig. 1c). We assessed the activation of MPL signaling induced by CALR mutants in Ba/F3-MPL cells by Western blot (Fig. 1d, Figure S1). Surprisingly, all CALR mutants tested induced similar constitutive activa- tion of JAK2 and STAT1, 3 and 5 compared to CALR wt. There was a slight induction of ERK phosphorylation with CALR mutants, but no significant activation of the PI3K/ AKT pathway. Addition of TPO provided only a minor increase in the phosphorylation of the STATs compared to the strong phosphorylation induced in presence of CALR wt, except in del19-expressing cells that were more sensi- tive to TPO stimulation, like previously observed for cell proliferation (Fig. 1c). In order to assess the expression levels of CALR mutants compared to CALR wt, we mod- eled HA-tagged CALR-expressing Ba/F3-MPL cell lines. Western blot analysis using an anti-HA antibody (Ab) revealed that the rare type 1-like and type 2-like CALR mutants were less expressed than del52 and ins5, that were themselves less expressed than CALR wt (Fig. 1d). Sur- prisingly, unlike del52 and ins5, we were repeatedly unable to detect del19, del34, and del46 using an Ab directed against the mutated C-terminus (Figure S4B). Cycloheximide-based study showed that CALR wt was >10 times more stable than either del52 or ins5 and that the rare CALR mutants were less stable than del52 and ins5 (Figure S2). No major difference in the regulation of rare CALR
We have previously shown that del52 and ins5 CALR mutants recapitulate an ET-like phenotype in vivo in a retroviral mouse model, but with different intensity and progression into MF [8]. To examine whether del19 behaves like ins5 and both del34 and del46 like del52 in mice, we performed bone marrow transplantation (BMT) after retroviral transduction of either CALR wt or the respective CALR mutations into Lin- BM cells. We engrafted 40–60% of GFP+ cell populations representing the percentages of initial CALR+-mutated cells. Type 1-like mutants induce a strong thrombocytosis in mice compared to CALR wt whereas platelet counts only slightly raised above the control in ins5 and del19 mice. No major effects were detected on RBC and WBC (Fig. 2).Type 1-like CALR mutants develop a post-ET myelofibrosisMice expressing type 1-like CALR mutants of ages above9 months post-BMT presented a significant increase in spleen weight (Fig. 3a) associated with a decrease in BMcellularity (Fig. 3b) compared to CALR wt and type 2-like CALR-mutated mice. Histopathology of BM (Fig. 3c) and spleen (data not shown) revealed expanding osteogenesis in BM associated with clusters of MKs in BM and spleen in type 1-like CALR-mutated mice. Fewer MK clusters were also observed for CALR ins5 and del19 mice. Silver staining revealed thickening of the reticulin fiber network in both BM and spleen, while no fibrosis was observed in CALR wt, ins5 or del19 mice.We engrafted approximately 40–60% of retrovirally- transduced GFP+ cells into recipient mice. After9 months, 100% of the HSC-enriched cell population (SLAM) were GFP+ in BM of type 1-like CALR-mutated mice while CALR wt and type 2-like CALR-mutated mice remained at levels similar to the initial graft (Fig. 4a). Accordingly, there was an increase in frequency of SLAMs (Fig. 4b) confirmed in early hematopoietic progenitor LSKs (Fig. 4c) in BM of type 1-like CALR-mutated mice. These data suggest that type 1-like mutations provide a strong competitive advantage to the mutated clone in wild-type hematopoiesis at the HSC level. This cell advantage was visible in blood with WBCs of almost exclusively virus- transduced origin, especially for del52 (Fig. 4d).Type 1-like CALR mutations induce an early amplification of the MK lineage while type 2-like mutations induce late megakaryocytic/platelet amplificationAs expected, type 1-like CALR mutations provided a competitive advantage to the MK lineage with 100% of the megakaryocytic progenitors (MkPs) detected in BM being GFP+ at 9 months post-BMT (Fig. 5a). This was accom- panied with an increased frequency of MkPs in BM of type 1-like CALR-mutated mice (Fig. 5b).
Concomitantly, there was also a visible increase in the frequencies of MK pro- genitors (CFU-MKs) and MKs (CD41+CD42+) in both BM and spleen (Fig. 5c, d, left and right panels, respectively) associated with a significant increase in the percentages of GFP+ platelets detected in blood of type 1-like CALR- mutated mice (Fig. 5f). For type 2-like CALR-mutated mice, a mild increasing trend of the frequencies of MkPs, MKs and in the GFP+ platelets was visible, especially when looking at the total number of MKs in mice (Fig. 5e). This suggests that type 2-like mutants induce a very late ampli- fication in megakaryopoiesis that is only visible at the late stages. In BM from type 1-like CALR-mutated and del52 mice older than 9 months post-BMT, cells were mostly all GFP+ in contrast to BM cells from del19 and ins5 mice.(Figs. 4b, 5b). Thus, we examined CALR wt and mutant protein levels in del52, del34 and del46 mouse BM cells that all express the CALR mutants, by Western blot (Figure S4A). Using the Ab directed against the mutated C-termi- nus, del34 and del46 were undetectable compared to del52. This suggests that either rare type 1-like CALR mutants were less expressed than del52 in BM, like what we observed in Ba/F3-MPL cells expressing HA-tagged pro- teins (Fig. 1e) or/and that the Ab had less specificity for del34 and del46 (and del19) compared to del52 (and ins5) since it does not detect the rare CALR mutants in cell lines either (Figure S4B).The extent of platelet increase induced by type 2- like CALR mutations in the retroviral mouse model is independent of the initial amount of CALR-mutated cellsIn order to clarify whether a minimal level of CALR ins5 clones might be necessary to induce a more robust pheno- type in mice we retrovirally transduced Lin- cells with CALR ins5 and CALR del19, as well as CALR wt and CALR del52 as controls, and sorted the CALR+ (GFP+) cells. We engrafted 100% of transduced CALR+ clones in recipient mice and verified that the percentages of GFP+ platelet and GFP+ WBC remained at 100% after BMT (Fig. 6a). Sur- prisingly, the increase in platelets developed by mice car- rying 40% (Fig. 2) and 100% of type 2-like CALR mutations was similarly minor, to barely significant (Fig. 6b). In contrast, when dominance is achieved, del52 mice develop a stronger thrombocytosis than mice with a lower number of initially engrafted mutated clones (at 4 months,nearly 3,000 103/µL platelets with 100% GFP+ cells engrafted vs. 2,000 103/µL platelets with 60% GFP+ cells engrafted). Together, these results indicate that the slight increase in platelets induced by ins5 and type 2-like CALR mutations is independent of the amount of mutated clone, while it enhanced the thrombocytosis phenotype induced by del52 mutation.
Discussion
In this study we demonstrate that the structurally-based classification of rare mutants of CALR, related to the two most frequent mutations found in patients, del52 and ins5, is functionally valid in vivo in mice. However, we found subtle differences in cell lines with less proliferative advantage induced by the rare CALR mutants and reduced protein levels although they present TPO-independent growth and similar constitutive signaling than del52 and ins. We showed that ectopic expression of the different CALR mutants in mice is sufficient to drive an ET-like phenotype evolving into MF for type 1-like CALR-mutated mouse models. The thrombocytosis was much more severe in presence of del34 and del46 than with del19. Although both type 1-like and type 2-like mutations are present in the HSC compartment, only type 1-like CALR mutations pre- sent a clonal dominance at this early stage of hematopoiesis. In this study, amplification of the MK lineage was moder- ately seen at the platelet level for ins5 and del19. The increase in MK frequency in BM observed in previous work
[8] was not significant here, suggesting a very mild the reticulin network. Images were obtained using a DM2000 Leica microscope and a DFC300FX Leica camera with Leica Application Suite v.2.5,OR1 acquisition software (2.5×, 10×, and 25× magnifications) amplification at the latest stage of megakaryopoiesis. One of our hypothesis was that a threshold level of CALR ins5 cells might be necessary to induce a more robust phenotype in mice. Indeed, when CALR ins5 patients develop a disease, allele burden has often reached a 50% level (100% het- erozygous mutated clones). However, when we engrafted 100% of mutated ins5 (GFP+) cells in mice to test whether the clonal dominance-like was important to develop a stronger thrombocytosis, it was not the case, in contrast to del52 cells, suggesting that a more robust disease might be due to higher protein expression levels rather than increased number of ins5-mutated cells. Since it is difficult to control the level of GFP intensity (proportional to CALR protein expression level) in a retroviral model, this hypothesis would be easier to test in knock-in mice generating het- erozygous and homozygous animals.
Very intriguingly, we found in Ba/F3 cell lines that rare mutants display subtle differences compare to del52 and ins5. They present a reduced cytokine-independent proliferative advantage over time compared to del52 or ins5 that correlated with lower protein expression levels due to decreased protein stability. With such properties, the rare CALR mutations, especially the type 1-like mutations, were expected to present a dif- ferent phenotype in mice than the del52 and ins5, which was not the case. The reasons are still unclear but we speculate that if they were expressed at similar level than del52 or ins5, they might trigger a stronger disease. More- over, since rare CALR mutants seem to induce lower pro- liferative advantage than del52 and ins5, we might hypothesize that the cells harboring atypical CALR mutants are less efficient to outcompete normal cells and are thus less prone to initiate the disease. This could explain why type 1-like and type 2-like CALR mutations are found so rarely in human MPNs compared to the del52 and ins5. A very intriguing question is why the two types of mutations induce different phenotypes in mouse. The first hypothesis could be that it activates differentially the JAK2/ STAT pathway. Using Ba/F3 cell line models, we showed that MPL is essential for autonomous cell growth and for the activation of the JAK2/STAT signaling axis whatever the mutant used, further suggesting that the C-terminus is the common feature responsible for MPL activation [6, 7, 15]. In addition, all CALR mutants activate the MPL/ JAK2 pathway to a similar extent in this cell model. This may suggest that the differences between type 1-like and type 2-like mutants are not related to the level of MPL/ JAK2 activation, or might only be visible with endogenous proteins levels. As the CALR mutations were modeled with MPL in Ba/F3 cells using retroviral transduction, if subtle differences exist they may be minimized by the retroviral- induced overexpression (of CALR and MPL) compared to endogenous protein levels. Alternatively, the JAK2/STAT pathway could be differentially activated by the different mutants in other cell types such as HSCs or megakaryocytic cells. A second likely hypothesis is that the phenotypic differences between del52 and ins5 might be related to calcium binding and signaling as suggested in patient MKs [10]. Indeed, the main structural difference between the two types of mutants is the retention of a wild-type residual calcium-binding motif in ins5 that is absent in del52 [3, 4, 10]. One major finding of our study is that del19, which holds only 5 out of the 12 negative charges from the wild-type sequence retained by ins5, exhibits a similar phenotype than
ins5. This suggests that the N- terminal negative charges in exon 9 are key to ins5 and the type 2-like CALR mutant phenotype.
Bioinformatic studies have also shown that the del52 mutant has more structural properties that are altered than ins5 mutant [13]. Phenotypic differences between the two types of CALR mutants may thus also be related to differences in their structures that may modify their direct interaction with MPL. Indeed, del52 mutant is capable of activating G-CSFR in vitro to a much higher extent than the ins5 mutant [5]. Finally, a third hypothesis could be that there might be a subtle difference in the interaction of human type 2 CALR mutant with the murine MPL vs. human MPL tested in Ba/F3 cell lines, while murine and human CALR mutants exhibit similar activity [15]. This would imply that the ins5 model under- represents the phenotype compared to the del52. The present retroviral mouse models have some simila- rities, but also some divergences with the human diseases. In both human and mouse, the del52 and type 1-like CALR mutations lead to an ET with a frequent progression to MF whereas ins5 and type-2 like mutants are mainly responsible for a thrombocytosis, barely detectable in mice. In contrast to mouse, the human type 2 disease is associated with an important thrombocytosis and a similar clonal dominance at the level of HSC compared to type 1 mutations. It remains to be determined whether these differences between mouse and human models are related to the species or to the use of retrovirus (i.e., decreased CALR wt/mutated CALR ratio). Results obtained with CALR del52 transgenic mice [16] as well as with CALR del52 knock-in mice [17] suggest that overall CALR mutants induce a milder myeloproliferative disorder in mice than in human. However, the present ret- roviral mouse models strongly suggest that del52 and type 1-like mutants have stronger oncogenic activities than ins5 and type 2-like mutants. This seems to be in contradiction with the human disease where only type 1 and type 1-like MF have an improved prognosis [14]. As it has been demonstrated that the prognosis of MF strongly depends on the number of driver mutations and also on their types, our results suggest that the favorable prognosis of type 1 and type 1-like CALR mutations may be due to the natural progression of the disease and not to a transformation of the disease whereas type 2 or type-2 like mutants may require several additional mutations to induce MF. In a similar manner, ins5 and del19, as JAK2V617F, may be associated with other mutations (for instance in epigenetic regulators) for the emergence and dominance of the CALR-mutated clone, in contrast to type 1-like mutations. This hypothesis can be easily tested in the future.
Future functional and biochemical studies will be required to precisely understand the mechanisms respon- sible of the phenotypic differences between type 1 and type 2 mutations. Moreover, understanding why most of the CALR mutations are so rarely found in patients compare to del52 and ins5 remains to be fully elucidate.pMSCV-IRES-GFP retroviral vector. We also constructed the N-terminal HA-tagged version of CALR wt and mutants inserting the HA sequence on the 3′ of the signal peptide sequence. Similarly, the human receptor of TPO (MPL) harboring a N-terminal FLAG tag was cloned into the pMEGIX-IRES-CD4 retroviral vector. Vesicular stomatitis virus glycoprotein (VSVG) pseudotyped viral particles were produced into 293EBNA cells. Ba/F3 cells were grown in DMEM supplemented with 10% fetal bovine serum (Stem Cell Technologies, Ozyme) and 5% WEHI-conditioned medium as a source of murine IL-3. Cells were retrovirally transduced to stably express MPL as well as the various human CALR constructs (non-tagged and HA-tagged). Viability assay using the Premix WST-1 Cell Prolifera- tion Assay System (Clontech) was conducted to measure dose-dependent cell proliferation to TPO at 48 h. Cell pro- liferation was evaluated by numbering cells in culture without cytokine or with either 5% WEHI or 10 ng/mL TPO. Western blot served to assess MPL signaling using antibodies (Abs) against the phosphorylated forms of JAK2 (Tyr 1054/1055), STAT1 (Tyr701), STAT3 (Tyr705), and STAT5 (Tyr694), extracellular signal-regulated kinase 1/2 (ERK1/2 Thr202/Tyr204) and AKT (Thr308) and against the pan forms (Cell Signaling Technology, Ozyme). MPL was revealed using an Ab from Upstate (Millipore), anti- total CALR Ab was from Abcam, anti-HA was from Covance (Ozyme) and anti-β-Actin Ab was from Sigma and serves as a loading control. All Gandotinib procedures were approved by the Gustave Roussy Ethics Committee (protocol 2016-066-7171). Bone marrow (BM) was collected from donor C57bL/6 mice (Janvier). Lineage-negative (Lin-) cells were purified and cultured for 2 days in presence of a cocktail of recombinant cytokines (SCF, IL-3, TPO, FLT3L, IL-6) before infection with the different viruses. Approximately 4 × 105 transduced cells were injected intravenously into lethally irradiated (9.5 Gy) recipient mice.