RRAS2 is frequently overexpressed in human CLL and its overexpression in wild type form drives the development of leukemia
Approximately 13% of all human cancers have alterations in at least one of the classic RAS genes (cBioportal for cancer genomics, www.cbioportal.org). Of those, KRAS is the most frequently altered, affecting 10% of patients. Although there are gene amplifications affecting the KRAS locus, most of the alterations detected consist of point mutations in the coding sequence (Fig. 1a). Mucinous adenocarcinomas of the appendix and pancreatic adenocarcinomas are the cancers with the highest frequency of missense mutations in KRAS, above 80% (Fig. 1a). In contrast, gene alterations affecting the RRAS2 locus have been found in less than 1% of all cancers and those alterations involve mostly gene amplifications (Fig. 1b). Squamous cell carcinomas of the oropharynx are the cancers with the highest rate of missense mutations that however barely reach 9% of the tumors (Fig. 1b). In contrast, RRAS2 is very frequently overexpressed in the wild type form in different types of cancer, being CLL followed by B-cell non-Hodgkin lymphomas the cancers with the highest expression of mRNA for RRAS2 (Fig. 1c; https://dcc.icgc.org/pcawg). In another previous study, CLL was found as the leukemia with the highest expression of RRAS2, peaking at approximately 8-fold more mRNA than blood cells from healthy donors (Fig. S1a, www.oncomine.org [30]).
Given that the wild type form of RRAS2 has a potent transforming activity of NIH-3 T3 fibroblasts [10], we interrogated if overexpression of wild-type RRAS2 could cause CLL. To this end, we followed a genetic approach by generating a mouse line with a cassette containing the wild-type sequence of human RRAS2 under a CAG promoter in the Rosa26 locus, which includes EGFP after an IRES sequence and a LoxP-flanked stop codon (Rosa26-RRAS2fl/fl) at the 5′ end of the construct (Fig. S1b). Crossing this mouse line with different Cre recombinase lines induces removal of the stop codon upon recombination and hence overexpression of R-RAS2 protein in specified tissues. We first set out to study systemic RRAS2 overexpression using Sox2-Cre mice, where Sox2 is an embryonic stem cell transcription factor and therefore will induce deletion of the LoxP-flanked sequence in all tissues. Rosa26-RRAS2fl/flxSox2-Cre mice are viable and fertile in hetero- and homozygosity. In control WT mice, Rras2 mRNA is highly expressed in lymphoid organs (spleen and lymph nodes) compared to other organs like kidneys, skin or liver (Fig. S1c). On top of that, expression of the combined RRAS2 + Rras2 mRNA (human and mouse) in Rosa26-RRAS2fl/flxSox2-Cre mice was found increased in the spleen, lymph nodes, liver, skin and kidneys (Fig. S1c). We did not find evident anomalies in those organs, except in the spleen. Rosa26-RRAS2fl/flxSox2-Cre mice presented marked splenomegaly (Fig. 1d) that was also manifested in terms of organ weight; spleens of Rosa26-RRAS2fl/flxSox2-Cre mice were 3-fold larger on average than spleens from control Rosa26-RRAS2fl/fl littermates (Fig. S1d). Spleen enlargement was paralleled by a net increase in the number of CD19+ B cells, suggesting that splenomegaly was due to B cell lymphocytosis (Fig. 1e). In addition, spleen size increased with age and was concomitant with enlarged follicles and loss of predominant diffuse red pulp areas, as observed by histopathological examination after hematoxylin and eosin staining (Fig. 1f).
The analysis by flow cytometry of marginal zone (MZ) and follicular B cell populations in 8-week-old mice showed the presence of an abnormal and abundant CD19+CD21−CD23− population in Rosa26-RRAS2fl/flxSox2-Cre mice, as well as an abundant IgM+IgD− population (Fig. 1g). Those populations of B cells could correspond to a transitional T1 population of B cells in their differentiation towards follicular or MZ B cells [31]. However, those cells expressed the CD5 marker, constituting a CD19+IgM+CD5+ population which was abundantly detected in spleens of Rosa26-RRAS2fl/flxSox2-Cre mice but not in those of control Rosa26-RRAS2fl/fl mice (Fig. S1e). The abnormal CD19+IgM+CD5+ population was also detected in the bone marrow (Fig. S1f) and was present in high number in the blood (Fig. 1h and i), suggesting that that the lymphoproliferative disease in Rosa26-RRAS2fl/flxSox2-Cre mice is indeed a B cell leukemia.
We next set out to determine if overexpression of human RRAS2 caused the generation of a B cell leukemia in a B cell-intrinsic manner. To this end, we crossed Rosa26-RRAS2fl/fl mice with mb1-Cre mice which express Cre recombinase specifically in B cells starting at an early precursor phase [23]. Those mice accumulated a large number of CD19+IgM+CD5+ B cells in the blood, showing that the development of the B cell leukemia is B cell-intrinsic (Fig. 2a and b).
The presence of increasingly large numbers of IgM + CD5+ B cells in the blood is a feature of human CLL [3]. We found that the number of leukemic CD19+IgM+CD5+ B cells in the blood progressed with age in nearly all Rosa26-RRAS2fl/flxmb1-Cre mice (Fig. 2c). The time-dependent increase in leukemic cell numbers in the blood may be responsible for the reduced life span of Rosa26-RRAS2fl/flxmb1-Cre mice (t1/2 = 13 months) compared to that of wild type controls (t1/2 = 32 months, Fig. 2d). Another feature of some patients with CLL is the increased concentration of IgM in serum, something that was also detected in Rosa26-RRAS2fl/flxmb1-Cre mice (Fig. S1g). A Giemsa staining of a blood smear showed the presence of enlarged lymphocytes with higher cytoplasmic content in Rosa26-RRAS2fl/flxmb1-Cre mice than those of WT control mice (Fig. S1h). Along these lines, the analysis by flow cytometry of the Forward Scatter parameter showed that the CD19 + CD5+ cells in blood are larger than normal (Fig. S1i and k). However, the analysis of cells entering the cell cycle in a period of 24 h by i.v. administration of the thymidine analogue BrdU showed that only 0.2% of the CD19 + CD5+ cells in the blood had incorporated BrdU (Fig. S1j and k), indicating that, in spite of being large, the CD19 + CD5+ leukemic cells in blood are not progressing through the cell cycle and are not lymphoblastic. The low rate of proliferation in blood correlate with findings in human CLL suggesting that circulating leukemic B cells are not in proliferation [32].
To demonstrate that RRAS2fl/flxmb1-Cre mice develop a CLL leukemia that can be transplanted to normal mice, we carried out adoptive transfer experiments in which sub-lethally irradiated wild-type mice expressing the hematopoietic cell allele marker CD45.1 were transferred with total B cells from young Rosa26-RRAS2fl/flxmb1-Cre mice or Rosa26-RRAS2fl/fl controls. Donor mice bear the CD45.2 allele and therefore transferred cells can be distinguished from endogenous ones according to the expression of the CD45.2 marker (Fig. 2e). We followed the progression of the leukemic CD19+IgM+CD5+ B cell population in the blood of the transferred mice and observed a progressive increase over time of the percentage of leukemic cells within the total (donor plus acceptor) B cell population (Fig. 2f). Those experiments demonstrated that the abnormal B cell population (or its precursors) can be transplanted and expands in non-diseased recipient mice, adding further support to the notion that the CD19+IgM+CD5+ B cell population is a CLL.
We also carried out an analysis of VH family usage in 9 independent Rosa26-RRAS2fl/flxmb1-Cre mice by PCR of genomic DNA. We found that, unlike total bone marrow cells from a wild-type donor (BM, Fig. 2g), which did not result in amplification of any particular band (except for VH family J558), sorted CD19+IgM+CD5+ B cells from mice with RRAS2 overexpression had overrepresentation of some VH families resulting in a pattern of bands that was specific to each individual mouse. Those data indicate that CD19+IgM+CD5+ B cells are oligoclonal and, therefore, additionally support the notion that RRAS2-overexpressing mice develop a CLL leukemia. Interestingly, the oligoclonal BCR repertoire in the analyzed mice included families known to mediate autoimmunity, i.e., VH11 and VH12 [33, 34]. Autoantigens have been shown to select B cells with exacerbated BCR signaling, hence inducing aggressive progression of CLL mouse models [35].
The expression of markers, the oligoclonality, the transferability, the slow steady increase of non-blastic, non-mitotic, B cells in the blood are all features of human CLL. Therefore, the results presented in Figs. 1 and 2 demonstrate that overexpression of human RRAS2 provokes the development of a CLL in mice, suggesting a cause-effect relationship for RRAS2 overexpression in the human disease.
Time-dependent increase of RRAS2 expression in leukemic B cells
Taking advantage of the GFP marker included in the RRAS2 overexpression cassette in the Rosa26 locus (Fig. S1b), we could track cells in the blood and lymphoid organs of Rosa26-RRAS2xSox2-Cre and Rosa26-RRAS2xmb1-Cre mice. Interestingly, in Rosa26-RRAS2xmb1-Cre mice, two distinct populations of CD19+ B cells with approximately a 10-fold difference in GFP expression were detected in blood: a GFPlow and a GFPhigh population (Fig. 2h). Likewise, the two populations of GFPlow and a GFPhigh B cells were also identified in the lymphoid organs of Rosa26-RRAS2xSox2-Cre mice (Fig. S2a). By contrast, T cells from those mice had a predominant GFPlow population. Interestingly, the presence of the GFPhigh population was more abundant within the leukemic CD5+ cell pool than within the follicular CD23+ B cell population (Fig. S2b). In the leukemic, CD5+ B cells, 80% were GFPhigh. Those results suggested that the enrichment in GFPhigh cells was taking place only within the leukemic B cell population.
To study this phenomenon, we first determined if the levels of GFP expression were correlated with those of RRAS2 expression. We sorted each of the GFP populations from total CD19+CD5+ cells from Rosa26-RRAS2xmb1-Cre mouse spleens and measured RRAS2 mRNA expression by RT-qPCR. We found that RRAS2 was overexpressed in the GFPlow population by 3.8-fold in comparison with follicular B cells from wild-type mice, whereas overexpression reached a mean of 20-fold in the GFPhigh population (Fig. 2i). A Western blot analysis of whole cell lysates of sorted GFPlow and GFPhigh leukemic cells showed that R-RAS2 protein expression was also much higher in GFPhigh than in GFPlow cells (Fig. S2c). Therefore, RRAS2 in CD5+ B cells is overexpressed in two populations, one with moderate and the other with very high (20-fold) levels. Interestingly, we found that the GFPhigh /GFPlow proportion within the CD19+CD5+ B cell population in the blood increases with age when mice are analyzed individually by periodic bleedings (Fig. 2j and Fig. S2d). In some mice, we detected infiltration of non-hematopoietic tissues such as kidneys, lungs and liver by lymphoid cells in old Rosa26-RRAS2xmb1-Cre mice. For instance, in 3 out of 17 one-year-old males, we detected perivascular infiltration of the lungs by lymphoid cells that were not forming follicles (Fig. 2k). The analysis by flow cytometry indicated that those infiltrates were constituted mainly by CD19+CD5+ leukemic B cells (Fig. 2l) with an overrepresentation of GFPhigh cells (Fig. 2m). These data suggest that CD19+CD5+ B cells with the highest expression of RRAS2 are the ones with the highest metastatic potential. Those data suggest that the age-dependent enrichment in CD19+CD5+ B cells with the highest expression of GFP is the result of a selective pressure that favors cells with the highest expression of RRAS2 and metastatic potential. Furthermore, the analysis of B cell markers in GFPhigh and GFPlow cells in spleen shows that GFPhigh are the cells with the highest proportion of CD21-CD23-CD19+ B cells (53%, Fig. S2e) and downregulated IgD, CD21, and B220. Interestingly, B220 downregulation is typical of B1 CD5+ B cells which are suspected to be the origin of leukemic cells in an Eμ-TLC1 mouse model of CLL [36]. In an attempt to determine if CD19 + CD5+ leukemic B cells in Rosa26-RRAS2xmb1-Cre mice derive from B1 CD5+ cells, we quantified the number of B cells in the peritoneum, a site rich in B1a and B1b cells, according to the expression of the CD11b and CD5+ markers. We found that leukemic cells were CD11b + CD5+, markers of B1a cells (Fig. S3a). However, although the number of B cells with those markers was increased 7-fold compared to WT controls, the number of cells with the same markers was increased by 10-fold in spleen. Therefore, it is not possible to assign a B1a origin in the peritoneum or the spleen according to cell numbers. IgM expression was more heterogeneous in the leukemic CD11b + CD5+ population of the peritoneum than in B1a cells of control WT mice (Fig. S3b). On the other hand, B220 was low or negative in both peritoneal WT B1a cells and leukemic cells (Fig. S3c). This population was enriched in the spleen of leukemic mice compared to WT controls. B1 cells emerge early in life from precursors in the liver. We compared the number of CD19 + CD5+ cells and the expression of different markers in 2 week-old mice of Rosa26-RRAS2xmb1-Cre mice with age-matched controls. We found that the liver of leukemic mice was enriched in CD19 + CD5+ negative for B220, CD21, CD23 and expressing intermediate levels of CD24 and high levels of CD38 (Fig. S4). However, the same phenotype and even bigger number of cells were already present in the spleen. Therefore, we cannot conclude if leukemic CD19 + CD5+ in RRAS2-overexpressing mice have a B1a origin in the liver or if they develop from mature B cells in the spleen.
To further compare the GFPlow and GFPhigh B cell populations, and these with leukemic CD19 + CD5+ B cells and normal follicular B cells, we carried out an RNAseq analysis of the transcriptome. We compared the following cell populations: follicular B cells sorted from spleens of six wild-type C57BL/6 mice; sorted CD19 + CD5+ leukemic B cells from spleens of six independent Rosa26-RRAS2xmb1-Cre mice; sorted normal B follicular B cells from the spleens of two Rosa26-RRAS2xmb1-Cre mice; sorted CD19 + GFPlow and sorted CD19 + GFPhigh B cell populations from two Rosa26-RRAS2xmb1-Cre mice. Gene expression data is summarized in Table S1. Principal component analysis shows a clustering of follicular B cells from Rosa26-RRAS2xmb1-Cre mice together with follicular B cells from wild-type C57BL/6 mice and of GFPlow and GFPhigh B cell populations with leukemic B cells (Fig. S5a). A comparison of the most differentially expressed genes in GFPlow versus GFPhigh B cell populations with their expression in leukemic (CD19 + CD5+) and normal follicular B cells populations shows that GFPhigh B cells are more closely related to leukemic cells than GFPlow B cells (Fig. S2f).
Overexpression of wild-type R-RAS2, but not expression of an oncogenic mutant of R-RAS2, induces CLL
Unlike for KRAS, oncogenic mutations in RRAS2 are rarely found in human cancer (Fig. 1a and b). However, the oncogenic mutation Q72L in RRAS2 has been found as a hotspot [37]. Indeed, of the 218 cancer samples (out of 45,604) with gene alterations in RRAS2, the missense mutation Q72L was the most frequent (www.cbioportal.org and Fig. S6a). To determine if mice expressing the Q72L mutant of Rras2 also developed CLL, we crossed knock-in Rras2(Q72L)fl/fl mice bearing a repeated and inverted exon 3 in the Rras2 locus with mb1-Cre mice to specifically exchange wild-type exon 3 for the mutant exon in B cells. The transcription of the Rras2 locus bearing the Q72L mutation was equivalent to that of B cells from WT control mice and much lower than that of B cells from Rosa26-RRAS2xmb1-Cre mice (Fig. S6b). We analyzed 14-month-old mice and their control littermates (negative for mb1-Cre) for evidences of a CLL and found that the number of B cells in spleen and blood (Fig. S6c), the presence of CD19 + CD5+ cells in blood (Fig. S6d), and the distribution of B cells among follicular, marginal zone and CD21−CD23− B cells (Fig. S6e) were within normal limits. These data suggested that overexpression of wild-type RRAS2 and not an oncogenic mutation drives the development of CLL in mice.
A highly conserved pattern of somatic mutations accompanies the development of RRAS2-driven CLL
The comparison of the transcriptome of six wild-type follicular B cell samples and seven independent leukemic samples using the Ingenuity Pathway Analysis (IPA) software revealed a strong signature of pathway alterations related to molecular mechanisms of cancer (Fig. S5b), thus confirming the malignant nature of the CD19 + CD5+ B cells expanding in Rosa26-RRAS2xmb1-Cre mice. In addition, there is a signature of increased mTOR activity which could be expected given the capacity of R-RAS2 to activate this pathway in B cells (Fig. S7, and [14].
Nonetheless, the most striking finding resulting from the analysis of RNAseq data was that CD19 + CD5+ leukemic cells from the spleens of seven independent mice showed the consistent presence of somatic mutations in 270 genes in all mice. Those mutations corresponded mostly to missense mutations and less so to frameshifts or nonsense mutations (Table S2). The mutated genes concentrated on chromosomes 7 and 4, followed by chromosomes 8, 9 and 17 (Fig. 3a). Chromosomes 2, 6, 12, 14, 16, 18 and both sex chromosomes contained no mutations in mRNA-encoding genes. IPA analysis of the 270 mutated genes identified a signature of genes related to immunological and hematological neoplasia (Fig. 3b) and to immunological development (Fig. S7); including genes such as Atm, Kmt2a, Macf1, Tet2, Akap13, Cd19, Cd22 and Polk. Finally, mutations in genes associated with retinoblastoma, cyclins and Atm, suggest that leukemic cells have unlocked the G1-S checkpoint (Fig. S7).
Out of the 270 genes found mutated in our RNAseq analysis of murine CLL cells, a total of 107 were found mutated within a cohort of 1094 human CLL patients and 8 within a cohort of 54 human MBL patients (www.cbioportal.org). Of those, the most frequently mutated gene in human CLL that is also found in our murine CLL model is ATM, with 114 cases out of 1094 patients (Fig. 3c). A table with the list of genes identified in four or more human CLL patients shows that, in addition to ATM, other known tumors suppressors such as ARID1A, AKAP13, PRDM2, SPEN, SMARCA2 and TET2 are also mutated (Fig. 3D). Within this set of genes, there are the epigenetic regulators ARID1A and SMARCA2, which form part of the chromatin remodeling complex, SWI/SNF, and the DNA damage repair genes ATM and HERC2. Most of the genes mutated in Rosa26-RRAS2 CLL leukemias and human CLL leukemias are also found mutated in other hematological cancers of B and T cells (Fig. 3d). A striking observation of our RNAseq analysis is that the vast majority of the 107 genes mutated in R26-RRAS2 CLL, that are also found mutated in human CLL, were mutated in homozygosity (100% of the sequences) or heterozygosity (50% of the sequences) in leukemic cells from each of the seven independent mice (Fig. 3e). Those data suggest a strong selective pressure for a combination of accompanying gene mutations that drive CLL development in mice that overexpress wild-type RRAS2, thus reinforcing the idea previously established that B cells with the highest RRAS2 expression (GFPhigh, Fig. 2 and Fig. S2) are selected during evolution of the leukemia. Another piece of evidence supporting the idea of selection is that many genes contain multiple missense mutations in their mRNAs in 50% or even 100% of the sequences, suggesting an evolution of B cells with a progressive number of mutations in the same genes. An example is shown for the genes Spen, Arid1a and Akap13 that have multiple mutations in their coding sequences (Fig. 3f and Table S2). Those positions are also found mutated in human CLL.
R-RAS2 is complexed with the BCR in leukemic cells and is required for proliferation and formation of tumors in xenografts by human CLL
Human CLL is characterized by the presence of recurrent “stereotyped” BCRs often with similar or identical sequences in the IgHV chain [6, 38]. This indicates that BCR signaling is fundamental for the development of CLL with the existence of stereotyped BCRs, suggesting the existence of common antigens (probably autoantigens) as initial triggers of the expansion of leukemic clones. Indeed, expression of the CD5 marker, characteristic of CLL, is induced by BCR signaling [39]. Our IgHV usage data (Fig. 2g) shows oligoclonality in the BCR repertoire of CLL cells emerging in Rosa26-RRAS2xmb1-Cre mice. In addition, IPA analysis of gene transcription in leukemic versus normal follicular B cells (Table S1) shows a strong signature of active BCR signaling (Fig. 4a), suggesting that the BCR is actively signaling in leukemic cells with activation of the PI3K-Akt-mTOR, the NFκB and the NFAT pathways, among others.
We previously demonstrated that R-RAS2 is constitutively associated to both the BCR and the TCR of normal B and T cells, respectively [12], and that it is an activator of PI3K and mTOR pathways [12, 14]. Therefore, we investigated if R-RAS2 is also physically associated to the BCR in leukemic cells. To this end, we biotin-labelled all plasma membrane proteins of purified CD19 + CD5+ leukemic cells from spleens of Rosa26-RRAS2xmb1-Cre mice using a membrane-impermeable biotinylation reagent and immunopurified R-RAS2 protein using an anti-HA antibody. The immunoprecipitates were subjected to two-dimensional SDS-PAGE under non-reducing/reducing conditions and immunoblotting was carried out with streptavidin-peroxidase for identification of total membrane proteins co-purifying with R-RAS2. Two biotinylated proteins corresponding in size to the IgH μ chain and the IgL chain were detected (Fig. 4b). Those two proteins were specifically detected with an anti-IgM (H + L) reactive antibody, thus demonstrating that R-RAS2 is constitutively interacting with the BCR in leukemic cells. By interacting with the BCR, R-RAS2 could be placed just downstream of this receptor in the activation of the PI3K-Akt-mTOR pathway and other canonical BCR signaling pathways (Fig. 4c), as suggested by the RNAseq data (Fig. 4a). Of note, we previously placed R-RAS2 as a direct BCR effector important for the transmission of tonic BCR signals required for survival and homeostatic cell proliferation in the absence of antigen [12]. This property of R-RAS2 aligns with a proposed antigen-independent signaling role for the BCR in CLL [8].
In order to highlight the causal relationship between RRAS2 overexpression and appearance of CLL, we next used the human CLL cell line MEC-1 to determine if R-RAS2 overexpression is required for the activation of BCR signaling pathways and for proliferation and survival of human CLL cells. MEC-1 cells overexpress RRAS2 mRNA at values 12-fold higher than those of B cells from peripheral blood of healthy human donors (Fig. 4d). Transducing MEC-1 cells with a lentiviral construct expressing a shRNA for human RRAS2 reduced the expression of this gene to levels close to those of normal B cells (Fig. 4d). The RRAS2 knockdown MEC-1 cells proliferated more slowly than their control counterparts in vitro (Fig. 4e) and, more importantly, produced smaller tumors than control MEC-1 cells when transplanted subcutaneously into lymphopenic mice (Fig. 4f). These results suggest that high R-RAS2 expression is required for human CLL cell proliferation in vitro and in vivo. Furthermore, the analysis by phosflow cytometry of knockdown and control cells showed that high R-RAS2 expression in MEC-1 cells is required to activate the PI3K-Akt-mTOR (pAkt and p4EBP1) pathway (Fig. 4g), as well as the proximal BCR signaling (pBlnk, pVav1 and pBtk) and other canonical pathways (MAPK/ERK) (Fig. 4h). Therefore, those data suggest that R-RAS2 expression is mediating BCR signaling in a human CLL cell line, as well as proliferation in vitro and in vivo.
To determine if R-RAS2 is also mediating BCR signaling in leukemic cells from Rosa26-RRAS2xmb1-Cre mice, we first analyzed those cells by phosflow cytometry in parallel to B cells from WT controls and to B cells from Rras2(Q72L)fl/fl xmb1-Cre mice, which do not develop CLL (Fig. S6a-f). We found that leukemic B cells overexpressing R-RAS2 present significantly higher activation of the PI3K-Akt-mTOR pathway (pAkt, pS6 and pEBP1) and also higher activation of proximal BCR-signaling (pBlnk) and Raf-ERK pathway (pERK) than normal control follicular B cells (Fig. S6f). Activation of the PI3K-Akt-mTOR, Raf-ERK and BCR signaling pathways was also higher in leukemic B cells than in B1a, B1b and marginal zone (MZ) normal B cells (Fig. S6g), being MZ the B cell population with the highest constitutive activation of those pathways. In addition, the results of Fig. S6 suggest that Rras2(Q72L)fl/fl xmb1-Cre mice expressing a constitutively active “oncogenic” mutant of R-Rras2 do not develop CLL because they do not activate those BCR signaling pathways. Rras2(Q72L)fl/fl xmb1-Cre mice do not develop any detectable malignancy, whereas mice that express the Rras2(Q72L) mutant in all tissues (tamoxifen-regulated Cre-ERT2-regulated mouse strain [iCre-Rras2Q72L]) do develop a form of T-cell acute lymphoblastic leukemia (T-ALL) in 100% of mice, but not CLL (Fernandez-Pisonero et al., under revision).
Wild-type RRAS2 is overexpressed in human CLL and correlates with parameters of poorer prognosis
Considering that Rosa26-RRAS2xmb1-Cre and Rosa26-RRAS2xSox2-Cre mice develop CLL, that the MEC-1 human CLL cell line overexpresses RRAS2 and that these cells require RRAS2 expression for proliferation and tumor formation, we set out to assess RRAS2 mRNA levels in our own cohort of previously untreated CLL patients to reinforce the metadata (Fig. 1c and S1a) and further analyze the relationship between RRAS2 overexpression and human CLL. In the cohort (n = 178), the average age was 69 years and 60% were males (Table S3). A 63% were in early stages of the disease, i.e., Binet stage A or Rai stage 0, and a total of 52 were classified as having a premalignant mononuclear B cell lymphocytosis (MBL) condition, whereas 38% were IGHV-UM CLL with worse prognosis [40]. Importantly, in line with repository data and correlating with our RRAS2-overexpressing mice, we found a significant increase in RRAS2 mRNA levels in CLL patients compared to healthy subjects, with a mean of 5.3-fold higher expression in CLL patients (Fig. 5a). A classification of patients according to the number of total lymphocytes in blood (Fig. 5b) and the percentage of malignant CD19 + CD5+ B cells (Fig. 5c) showed a direct correlation with overexpression of RRAS2, i.e., the group of patients with highest lymphocytosis and malignant cell content has the highest overexpression of RRAS2. The classification of patients that present premalignant MBL versus full-blown CLL is based on the degree of lymphocytosis (threshold at 5 × 106 lymphocytes/mm3). Both conditions occurred upon overexpression of RRAS2, though overexpression in CLL was significantly higher (5.9 fold) than in MBL (3.4 fold, Fig. 5d). Advanced age is another condition of poorer prognosis [40]. We found increased median and mean overexpression of RRAS2 with age in full-blown CLL, but not in MBL (Fig. 5e).
The analysis of RRAS2 expression according to the unmutated vs mutated status of the IgHV region showed a higher mean for the IGHV-UM than for the IGHV-M CLL, although those differences did not reach statistical significance (Fig. 5f). In the subject cohort, most of the samples with unmutated IgHV classified as CLL, not MBL (Fig. 5g). The grouping of leukemias according to MBL, CLL, IGHV-UM and IGHV-M showed that RRAS2 expression correlated with CLL, regardless of IgHV status (Fig. 5h). Another risk factor in CLL is sex, for male patients carry a worse prognosis than female ones [40]. We found that leukemias from male patients had significantly higher expression of RRAS2 than those from female patients (Fig. 5i). A classification of leukemias by fold expression intervals showed that 82% of all samples overexpressed two-fold or more RRAS2 mRNA than B cells from healthy controls, although the distribution by sex was unequal: 13.6% of leukemias from male patients expressed lower than 2-fold levels compared to 25.4% for female patients (Fig. 5j). Conversely, 65.7% of leukemias from male patients overexpressed 4-fold or more RRAS2 mRNA compared with 55.2% for female patients. Altogether, these data show that RRAS2 mRNA is overexpressed in the large majority of CLL and that higher expression is associated to factors of worse prognosis such as CLL versus MBL, higher lymphocytosis, advanced age and male sex. The results of human RRAS2 mRNA expression in our cohort of samples (Fig. 5), in repository data (Fig. 1c and S1a), and the effect of RRAS2 overexpression in mouse B cells driving the development of CLL strongly suggest that overexpression of unmutated RRAS2 also drives the development of human CLL. A comparative RT-qPCR expression analysis of RRAS2 expression in human CLL and in the two mouse models studied here shows that overexpression attained in the mouse systems is within the range of RRAS2 expression levels detected in human leukemias (Fig. 5k), thus reinforcing the idea that the driver role in human CLL is possible.
A single-nucleotide polymorphism (SNP) in the 3′ UTR region of RRAS2 mRNA genetically links overexpression of unmutated RRAS2 with human CLL and more aggressive disease
Although mutations in the coding sequence of RRAS2 have rarely been found in human cancer, we decided to sequence the RRAS2 mRNA in our cohort of 178 patients. We did not find any missense mutation in the coding sequence. However, in some patients we found a C nucleotide in position 124 after the stop codon in the 3’UTR, whereas the canonical sequence contains a G in that position (Fig. 6a). The C nucleotide at position 124 of the 3’UTR was the previously cataloged SNP rs8570 (from now on, 124C, in this paper). This SNP was identified within a group of SNPs of 26 genes in the MAP kinase pathway to be associated with risk of cutaneous melanoma [41]. The 124C SNP was first detected in homozygosity or heterozygosity by Sanger DNA sequencing (Fig. 6b) but later was reassessed by a RT-qPCR method using primers with mismatching positions (Fig. 6c, Methods). Using this double procedure, we calculated the frequency of GG homozygotes at position 124 of the 3’UTR in 51% of the CLL samples, 33% for GC heterozygotes and 15% for CC homozygotes. Thus, the frequency of the non-canonical 124C allele in the CLL cohort is of 32%, slightly above the 22% frequency measured in the general European population (ALFA project; https://www.ncbi.nlm.nih.gov/snp/rs8570#frequency_tab). Within our cohort of CLL patients, and according to the frequencies of GG, GC and CC genotypes at position 124, we observed a higher-than-expected frequency of CC homozygotes and lower frequency for the GC heterozygotes (Fig. 6d). The difference between Observed and Expected values indicates that the distribution of alleles is not in a Hardy-Weinberg equilibrium [42] (χ2 experimental = 9.62 > χ2 theoretical = 3.84; p < 0.005). The CC genotype correlated with the highest expression of RRAS2 mRNA, whereas expression in leukemias of the GG phenotype was significantly lower than the ones of the CC genotype (Fig. 6e). Since the size of the CC group is small and mean RRAS2 mRNA expression was very similar between the GC and CC groups, we decided to combine both groups for further analysis. The combined CC + GC group expressed significantly higher levels of RRAS2 mRNA than the GG one (Fig. 6f). This correlated with indicators in CLL patients’ blood of more aggressive disease, such as total number of lymphocytes (Fig. 6g), higher percentage of B cells (Fig. 6h), higher percentage of leukemic CD19 + CD5+ cells (Fig. 6i) and lower number of platelets (Fig. 6j). The combined CC + GC group was significantly more represented in patients with full-blown CLL at diagnosis than in those with MBL (Fig. 6k). Several genetic alterations detected by FISH are prevalent among CLL patients, reflecting varying degrees of association with disease prognosis [43]. The presence of one or two C alleles was more common within the group of patients with chromosomal alterations determined by FISH than those with no alterations (Fig. 6l). Taking together the frequencies of deletions found in chromosome arms 11q and 17p, there is a significant association of those alterations with the presence of one or two 124C alleles (Fig. 6m). Chromosome arm 11q encodes for ATM and chromosome arm 17p encodes for TP53. Loss of any of those tumor suppressor genes is associated with poorer prognosis. We also searched for associations between the expression of the 124C allele and CLL bearing mutated or unmutated IgHV. We did not find any association between the frequency of IGHV-UM in the GG and CC groups of the cohort. However, the presence of one 124C allele (GC heterozygotes) compared to two 124G alleles (GG homozygotes) is also significantly associated to IGHV-UM (Fig. 6n). Finally, we found a clear association between male sex and the frequency of GC and CC (Fig. 6o). Expression of an unmutated IgHV and male sex are two additional factors of poorer prognosis [40]. Therefore, the SNP at position 124 of the 3′ UTR of RRAS2 mRNA emerges as a novel prognostic factor of CLL progression and genetically proves a cause-effect relationship between RRAS2 overexpression and human CLL.