Genetic and epigenetic inactivation of SESTRIN1 controls mTORC1 and response to EZH2 inhibition in follicular lymphoma

See allHide authors and affiliations

Science Translational Medicine  28 Jun 2017:
Vol. 9, Issue 396, eaak9969
DOI: 10.1126/scitranslmed.aak9969

Lymphoma’s loss is a therapeutic gain

Follicular lymphoma is a relatively common and difficult-to-treat hematologic malignancy, for which no specific targeted therapy is available. Knowing that deletions of chromosome 6q are common in this tumor type, Oricchio et al. examined the genes on this chromosome and identified SESTRIN1 as a likely tumor suppressor. The authors examined the mechanism by which the loss of SESTRIN1 contributes to tumorigenesis and identified a mechanistic connection between SESTRIN1 and EZH2, an epigenetic modifier that plays a role in multiple cancer types. The authors demonstrated that the effectiveness of targeting EZH2 depends on SESTRIN1 genetic and epigenetic status and also reported that mutations in EZH2 itself can sensitize cancer cells to additional targeted therapies.


Follicular lymphoma (FL) is an incurable form of B cell lymphoma. Genomic studies have cataloged common genetic lesions in FL such as translocation t(14;18), frequent losses of chromosome 6q, and mutations in epigenetic regulators such as EZH2. Using a focused genetic screen, we identified SESTRIN1 as a relevant target of the 6q deletion and demonstrate tumor suppression by SESTRIN1 in vivo. Moreover, SESTRIN1 is a direct target of the lymphoma-specific EZH2 gain-of-function mutation (EZH2Y641X). SESTRIN1 inactivation disrupts p53-mediated control of mammalian target of rapamycin complex 1 (mTORC1) and enables mRNA translation under genotoxic stress. SESTRIN1 loss represents an alternative to RRAGC mutations that maintain mTORC1 activity under nutrient starvation. The antitumor efficacy of pharmacological EZH2 inhibition depends on SESTRIN1, indicating that mTORC1 control is a critical function of EZH2 in lymphoma. Conversely, EZH2Y641X mutant lymphomas show increased sensitivity to RapaLink-1, a bifunctional mTOR inhibitor. Hence, SESTRIN1 contributes to the genetic and epigenetic control of mTORC1 in lymphoma and influences responses to targeted therapies.


Genetic and epigenetic alterations in cancer point to the biological processes driving tumor development and may reveal opportunities for therapy (1). Follicular lymphoma (FL) is among the most common forms of indolent B cell lymphoma, and cytogenetic studies have identified the chromosomal translocation t(14;18), which activates the anti-apoptotic BCL2 gene, as a hallmark of the disease (2). Recent sequencing studies have cataloged somatic mutations in FL. These include frequent mutations in epigenetic regulators such as KMT2D, CREBBP, and EZH2 (37), immune receptor genes such as TNFRSF14, CD79B, and B2M (810), and somewhat less frequent mutations in mammalian target of rapamycin complex 1 (mTORC1) regulators such as RAGGC and ATP6V1B2 (11).

FLs also harbor recurrent copy number changes, and the exact genetic targets of large chromosomal lesions can be difficult to define. For example, deletions of chromosome (Chr.) 6q occur in about 25 to 30% of FLs, which have been linked to poor prognosis (12, 13). The tumor suppressor genes A20/TNFAIP3 (14) and ephrin receptor A7 (EPHA7) (15) have been identified as targets of the Chr. 6q deletions. However, the size and complexity of Chr. 6q lesions suggest the presence of additional genetic targets. Unbiased genetic screens provide an efficient way to pinpoint candidate tumor suppressor genes within large chromosomal lesions (15, 16). Here, we identified the mTORC1 regulator SESTRIN1 (1720) as a tumor suppressor and treatment response modifier in FL.


SESTRIN1 is a functional target of Chr. 6q deletions in FL

A meta-analysis of copy number data from two large cohorts of indolent FLs [Memorial Sloan Kettering Cancer Center (MSKCC), n = 64; University of Nebraska Medical Center (UNMC), n = 196] (GSE40989) (21, 22) shows copy number losses at Chr. 6q in 34% (89 of 260) of the analyzed cases (Fig. 1A and fig. S1A). GISTIC analysis (23) identifies one peak on Chr. 6p21 and five significant [false discovery rate (FDR) < 0.01] and frequently altered regions on Chr. 6q: Chr. 6q13, Chr. 6q14, Chr. 6q16-21, Chr. 6q23, and Chr. 6q27 (Fig. 1B and table S1). The frequency of homozygous and heterozygous deletions is similar in the different regions, varying between 9 and 13% for the homozygous and between 12 and 16% for the heterozygous (Fig. 1C and table S2). Moreover, we identified individual samples harboring rare but highly focal events (<5 Mb). These focal deletions are larger than commonly described germline variants, and they are not listed in the database of genomic variants (, indicating that they are tumor-specific and may target functionally relevant genes associated with FL development (table S3). For example, we identified focal losses targeting TNFAIP3/A20 and SESTRIN1 (Fig. 1D), suggesting that SESTRIN1 may be a target of the Chr. 6q deletion in FL. However, despite this focal alteration, most of the deletions targeting SESTRIN1 are large and simultaneously also affect other genes including, for example, PRDM1.

Fig. 1. Genomic analyses and a genetic screen identify SESTRIN1 as a functional target of Chr. 6q deletions in FL.

(A) Integrative Genome View of Chr. 6q deletions occurring in 34% (89 of 260) of FL tumors. (B) Analysis of recurrent deletions in Chr. 6 in FL using GISTIC algorithm. (C) Frequency of homozygous and heterozygous deletions in Chr. 6q. (D) Integrative Genome View visualization of 10 of 260 cases with focal deletions (<5 Mb) on Chr. 6q. Representative FL case with focal deletion affecting SESTRIN1 and TNFAIP3 locus is shown in detail. (E) Flow cytometry validation of individual shRNAs linked to green fluorescent protein (GFP) showing percentages of GFP-positive cells before and after IL-3 withdrawal in pro–B cells transduced with the indicated constructs.

To identify functionally relevant targets of the Chr. 6q deletions, we performed unbiased RNA interference (RNAi) focused on genes encoded at Chr. 6q. We screened 81 genes [each gene was targeted by one to six short hairpin RNAs (shRNAs)] for enhanced viability in immortalized pro–B cells (FL5-12) that depend on interleukin-3 (IL-3) (fig. S1B) (15, 24). FL5-12 pro–B cells represent a different phase of B cell differentiation than B cells in FL tumors; however, they are a useful in vitro model to screen for genetic modifiers that need to be further explored in a relevant in vivo model of lymphomagenesis (25, 26). In our screen, we identified 29 shRNAs enriched at least 1.5-fold, indicating a survival advantage in IL-3–depleted FL5-12 cells (table S4). As expected, these included shRNAs against the tumor suppressor genes Epha7 (15) and Tnfaip3 (14). In addition, two of three shRNAs targeting Sestrin1 scored positively in our screen, with fold change greater than 1.5 (P = 0.02) (fig. S1C). We confirmed their ability to significantly (P = 0.01) down-regulate Sestrin1 expression (fig. S1D), promote IL-3–independent survival (Fig. 1E), and accelerate proliferation (fig. S1E). Thus, Sestrin1 reduction enhances pro–B cell survival and indicates a potential tumor suppressor function.

Deficiency of Sestrin1 promotes FL development in vivo

We used the vavBcl2 model of FL to directly test the effect of Sestrin1 deficiency on lymphoma development. This model recapitulates key aspects of the genetics and pathology of human FL (15, 27). We transduced vavBcl2 hematopoietic precursor cells (HPCs) with empty vector MSCV-IRES-GFP or with the two shRNAs targeting Sestrin1 pinpointed by the screen, and we transplanted the modified vavBcl2 HPCs into lethally irradiated, syngeneic, wild-type animals (Fig. 2A). In recipient mice, knockdown of Sestrin1 resulted in significant acceleration of lymphoma development (sh-1 Sestrin1: n = 10; median survival, 93 days; log-rank P = 0.041; sh-2 Sestrin1: n = 12; median survival, 82 days; log-rank P = 0.028) (Fig. 2B). The shRNAs selectively down-regulated Sestrin1 expression and did not affect Sestrin2 or Sestrin3 (Fig. 2, C and D). Surface marker and pathological analyses confirmed the similarity of control and Sestrin1-deficient murine lymphomas with human FLs. For example, we found the pathognomonic follicular structure, lack of apoptosis by TUNEL (terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling), expression of the germinal center (GC) markers peanut agglutinin (PNA) and BCL6, somatic hypermutation in the VDJH4 variable region, and a predominant B220+ and CD19+ B cell population with modest infiltration of CD4+ or CD8+ T cells in both controls and tumors expressing shRNAs targeting Sestrin1 (Fig. 2E and fig. S2, A to C). Although the Ki67-positive staining is low as in the human disease, loss of Sestrin1 significantly increases tumor proliferation [average of the number of Ki67-positive cells in control = 8.5%, sh-2 Sestrin1 = 23.1% (P = 0.02), and sh1 Sestrin1 = 15.5% (P = 0.05)] (Fig. 2E and fig. S2D). Consistent with Sestrin’s function as an upstream regulator of mTORC1 (1820), we found increased phosphorylation of eukaryotic translation initiation factor 4E–binding protein 1 (4EBP1) and ribosomal S6 protein by immunohistochemistry (Fig. 2F and fig. S2, E and F) and immunoblots of Sestrin1-deficient lymphoma cells compared to the control (Fig. 2G). Hence, Sestrin1 acts a tumor suppressor that controls mTORC1 activity in an animal model of FL.

Fig. 2. Loss of Sestrin1 promotes FL development in vivo.

(A) Graphic representation of chimeric mouse model based on the transplantation of vavBcl2 transgenic hematopoietic progenitor cells. WT, wild type. (B) Analysis of tumor latency in the vavBcl2 chimeric mouse model with sh-1 Sestrin1 (blue; n = 10), sh-2 Sestrin1 (red; n = 12), and vector (black; n = 24). Significant difference in survival was measured by log-rank test between animals expressing empty vector and shRNAs targeting Sestrin1. (C and D) Characterization of the shRNAs against Sestrin1. (C) Sestrin1, Sestrin2, and Sestrin3 mRNA expression in tumors expressing vector or shRNA for Sestrin1. P values were calculated using two-tailed t test. n.s., nonsignificant (P < 0.05). Data are shown as a bar graph corresponding to the mean ± SD. (D) Immunoblot analysis in pro–B cells expressing shRNA for Sestrin1 or vector. The antibody recognized two Sestrin1 isoforms, and they were both targeted by shRNA. (E) Representative images of immunohistochemistry and histopathology analyses of indolent vavBcl2 tumors expressing shRNAs against Sestrin1. Ki67 quantification was performed on three different slides for each tumor type using ImageJ. Scale bars, 100 μm. H&E, hematoxylin and eosin; PNA, peanut agglutinin; BCL6, B cell lymphoma 6. (F) Representative immunohistochemistry analyses of phospho-S6 and phospho-4EBP1 in vavBcl2 tumors expressing vector or shRNA #2 for Sestrin1. Quantification was performed on two different slides for each tumor type using ImageJ. Scale bars, 100 μm. (G) Analysis of phosphorylation of S6 and 4EBP1 by Western blot in B cells expressing vector, short hairpin for Sestrin1, or overexpressing Akt as control.

SESTRIN1 is a direct target of repression by EZH2Y641X in lymphoma

We examined the expression of SESTRIN1 in human B cells and in FL and diffuse large B cell lymphoma (DLBCL) samples. SESTRIN1 expression is similar in naïve and memory B cells, centroblasts, centrocytes, FL, and DLBCL (fig. S3A) (GSE12195). By subdividing the tumor samples based on EZH2 mutation status, we noticed a significant reduction of SESTRIN1 expression in FL tumor samples that carried the EZH2 gain-of-function mutation (EZH2Y641X; n = 5) compared to samples that were EZH2 wild type (n = 52; P = 0.0038) (Fig. 3A). We further observed the same significant association between SESTRIN1 expression and mutant EZH2Y641X in GC-type DLBCL (EZH2Y641X, n = 8; EZH2 wild type, n = 30; P = 0.022) (Fig. 3B and table S5) (28, 29). Expression of SESTRIN2 and SESTRIN3 was not related to EZH2 mutation status in either FL or DLBCL samples (fig. S3, B to E). Chromatin immunoprecipitation–quantitative polymerase chain reaction (ChIP-qPCR) using three separate probes covering the SESTRIN1 promoter revealed a significant (P ≤ 0.01) enrichment of EZH2 on SESTRIN1 promoter (Fig. 3C and fig. S3F) associated with an increase of H3K27me3 in EZH2Y641X mutated compared to EZH2 wild-type cells in the same region (Fig. 3D and fig. S3G). Hence, expression of SESTRIN1 is linked to the activity of the mutant EZH2Y641X polycomb protein in lymphoma samples.

Fig. 3. SESTRIN1 is a direct target of EZH2Y641X silencing in lymphomas.

(A) Differential SESTRIN1 expression in 57 primary FL samples with and without EZH2Y641X mutation. The P value was calculated by Wilcoxon test. Data are shown using the R boxplot function with whiskers extending to minimum and maximum values. FPKM, fragments per kilobase million. (B) Differential SESTRIN1 expression in 38 germinal center B cell–DLBCL (GCB-DLBCL) samples, including 8 samples with the EZH2Y641X gain-of-function mutation. The P value was calculated by Wilcoxon test. Data are shown using the R boxplot function with whiskers extending to minimum and maximum values. (C) ChIP for EZH2 and immunoglobulin G (IgG) in SU-DHL-4 lymphoma cell lines, followed by qPCR (ChIP-qPCR) using three primers spanning SESTRIN1 promoter (P1, P2, and P3), two negative control primers mapping on Chr. 6 upstream of FRK and TNFAIP3 coding sequence (CDS), and two other negative controls mapping on Chr. 7 and Chr. 3. Data are shown as a bar graph corresponding to the mean ± SEM. (D) H3K27me3 ChiP sequencing (ChIP-seq) signal spanning SESTRIN1 genomic locus (Chr. 6: 109,305,000 to 109,420,000) in OCI-LY7 EZH2WT cells (light blue) and Karpas-422 (dark red) and SU-DHL-6 (red) EZH2Y641X mutant cells. Bottom: Zoom-in in SESTRIN1 promoter and coding start site region (Chr. 6: 109,410,000 to 109,420,000). (E) Differential expression analysis focused on genes within the Chr. 6q region in EZH2WT lymphoma cell lines (n = 2), EZH2Y641X-mutated (n = 6), and mutated lines treated with the EZH2 inhibitor GSK126 for 72 hours (n = 6 EZH2Y641X + GSK126). Each cell line has been analyzed in duplicate. SESTRIN1 is significantly differentially expressed between the three groups (FDR < 0.01). (F) Western blot analysis of SESTRIN1 and SESTRIN2 in EZH2Y641X (SU-DHL-6) and EZH2WT (OCI-LY19) lymphoma cell lines treated with the EZH2 inhibitor (1 μM GSK126 for 72 hours). Sestrin1 antibody detects two SESTRIN1 predicted isoforms. Quantification was performed using Image Studio Lite, and the average of the signal intensity was normalized to the control. (G) SESTRIN1 expression in the indicated EZH2WT and EZH2Y641X-mutated lymphoma cell lines treated for 72 hours with dimethyl sulfoxide (DMSO) or 1 μM GSK126. Data are shown as a bar graph corresponding to the mean ± SEM. (H) ChIP for H3K27me3 and histone-3 in SU-DHL-6 lymphoma cell line with and without GSK126 (1 μM for 72 hours), showing loss of SESTRIN1 promoter binding upon EZH2 inhibition. Data are shown as a bar graph corresponding to the mean ± SEM.

We directly tested the effect of EZH2 inhibition on SESTRIN1 gene expression across a panel of cell lines. Gene expression data from lymphoma cell lines of EZH2 wild type (n = 2) or EZH2Y641X mutant (n = 6) confirmed a significant loss of SESTRIN1 expression in EZH2Y641X mutant lines (FDR < 0.01; P = 8.86 × 10−8) (Fig. 3E and table S6) (GSE40971) comparable to the down-regulation observed in previously reported targets of EZH2 (fig. S3H) (30). Global level of H3K27me3 is higher in EZH2Y641X mutant than in the EZH2 wild-type cells, and treatment with the EZH2 inhibitor caused loss of the H3K27me3 mark (fig. S3I). This change in the histone mark corresponded to a significant (P ≤ 0.01) increase in SESTRIN1 mRNA and protein expression in the EZH2Y641X mutant cells (Fig. 3, E and F, and fig. S3J), whereas it had no effect on SESTRIN1 expression in EZH2 wild-type lines (Fig. 3, F and G, and table S7). To allay potential concerns about drug selectivity, we confirmed that RNAi-mediated EZH2 knockdown in EZH2 mutant Karpas-422 cells caused induction of the SESTRIN1 mRNA (fig. S3K) (GSE41239), and we obtained similar results using an additional EZH2 inhibitor, EPZ-6438, in a panel of EZH2 wild-type and mutant lymphoma cells (fig. S3L). ChIP demonstrated a significant (P < 0.05) GSK126-sensitive association of the SESTRIN1 promoter with the H3K27me3 mark (Fig. 3H). We performed ChIP-qPCR for the H3K27me3 mark and histone-3 in EZH2Y641X mutant SU-DHL-6 cells treated with GSK126, showing an association with the SESTRIN1 promoter but not the RPL30 or TNFAIP3 promoter used as controls (Fig. 3H and fig. S3M). An analysis of promoter methylation and SESTRIN1 sequence in 15 primary FL samples and five cell lines did not reveal aberrant CpG island methylation or mutations (fig. S3N). Hence, the mutant polycomb factor EZH2Y641X controls SESTRIN1 expression in lymphoma cells.

Next, we explored the genetic relationship between SESTRIN1 deletions, mutation status of EZH2, and recently reported RAG C guanosine triphosphatase (GTPase) (RRAGC) mutations that also act to increase mTORC1 activity (11). The analysis of 172 FL samples revealed a significant mutually exclusive relationship between alterations in EZH2, RRAGC, and SESTRIN1 (P = 0.03), which affect 47% of the cases (fig. S3O). Moreover, an analysis focused only on the relationship between SESTRIN1 loss and EZH2 mutations was performed across multiple data sets, including 172 FL and 88 DLBCL (31, 32). Thirty-five percent of the cases (92 of 260) harbor either EZH2 mutations or SESTRIN1 loss, and only six cases show concurrent alteration of both genes (P = 0.06) (fig. S3P). Together, these data indicate that SESTRIN1, RRAGC, and EZH2 lesions are alternate events implicated in mTORC1 control in lymphoma.

Mutant EZH2Y641X activates mTORC1 and mRNA translation in lymphoma

SESTRIN family of proteins control mTORC1 activity in response to p53 activation and reduce mRNA translation in response to DNA damage (1719, 33). Therefore, we examined the cellular response to DNA damage in the presence of the EZH2Y641X gain-of-function mutation. We treated OCI-LY19 and SU-DHL-5 cells (wild type for p53 and EZH2) and SU-DHL-10 and VAL cells (p53 wild type and EZH2Y641X) (table S8) with the topoisomerase-2 inhibitor doxorubicin (DXR; 200 nM). This treatment failed to induce SESTRIN1 expression in EZH2Y641X mutant cells, whereas both EZH2 wild-type and mutant cell lines responded with induction of SESTRIN2 and p21 mRNAs (Fig. 4A and fig.S4A). Consistently, DXR caused an inhibition of mTORC1 signaling, as indicated by loss of phosphorylation of 4E-BP and ribosomal S6 in wild-type cells but not in EZH2Y641X mutant cells (Fig. 4B and fig. S4, B and C; quantified in fig. S4, D and E). Hence, the EZH2Y641X gain-of-function mutation can override mTORC1 inhibition caused by genotoxic stress in lymphoma cells by inhibiting SESTRIN1.

Fig. 4. SESTRIN1 determines the effects of EZH2 inhibition in lymphoma.

(A) SESTRIN1 and SESTRIN2 expression in OCI-LY19 and SU-DHL-5 lymphoma cells (EZH2WT/p53WT) and SU-DHL-10 and VAL cells (EZH2Y641X/p53WT) treated with DXR or vehicle for 6 hours. Two-tailed t test was used to define significant (P < 0.05) or nonsignificant (P > 0.05) difference between EZH2WT and EZH2Y641X mutant cells. Data are shown as a bar graph corresponding to the mean ± SEM. Each experiment had two or three technical replicates. (B) Immunoblot with the indicated antibodies in OCI-LY19 and SU-DHL-5 lymphoma cells (EZH2WT/p53WT) and SU-DHL-10 and VAL cells (EZH2Y641X/p53WT) treated with DXR or vehicle for 6 hours. (C) Analysis of S6 and 4EBP1 phosphorylation in EZH2Y641X-mutated lymphoma cell lines SU-DHL-10 and SU-DHL-6 treated for 72 hours with DMSO (vehicle), or 1 or 2 μM GSK126. (D) Immunoblot to assess S6 and 4EBP1 phosphorylation in SU-DHL-10 cells with mutations in SESTRIN1 genomic locus (CRISPR-SESTRIN1) or expressing shRNA targeting SESTRIN1 (sh-SESTRIN1) treated with vehicle (DMSO) or GSK126 (1 or 2 μM for 72 hours). (E) Flow cytometry analysis to measure methionine analog (AHA) incorporation in lymphoma cells treated with the EZH2 inhibitor (GSK126) or rapamycin with or without SESTRIN1. (F) Quantification of the methionine synthetic analog incorporation. Data are shown as a bar graph corresponding to the mean ± SD. (G and H) Tumor growth curves of xenografted SU-DHL-10 expressing vector (G) or CRISPR-SESTRIN1 (H), treated for 14 days with vehicle or GSK126. The mean tumor volumes ± SD are expressed in cubic millimeter. P values were calculated by two-tailed t test. n.s., nonsignificant difference (P > 0.05). (I) Tumor weight in animals xenografted with SU-DHL-10 cells expressing vector or single-guide RNA (sgRNA) targeting SESTRIN1 (CRISPR-SESTRIN1) and treated with the EZH2 inhibitor (75 mg/kg ip) for 14 days. P value was calculated by paired t test. Data are shown as single values with mean ± SD. (J) Representative images of immunohistochemistry and histopathology analyses of xenografted tumors expressing vector or sgRNA targeting Sestrin1 (CRISPR-SESTRIN1) and stained for cleaved caspase3, phospho-S6, and phospho-4EBP1. Quantification was performed with ImageJ. Scale bars, 200 μm. (K) Graphic representation of RapaLink-1 molecular structure and the FAT (FRAP-ATM-TTRAP), FRB, and kinase mTOR domains targeted by the inhibitor. (L) Percentage of viable cells in EZH2Y641X mutant and EZH2WT cells treated with the mTOR inhibitor (12.5 nM RapaLink-1) for 48 hours. Each cell line was analyzed in triplicate, and two-tailed t test was used to assess the significance of the difference in response between EZH2Y641X mutant and EZH2WT cells. Data are shown as a bar graph corresponding to the mean ± SD.

Next, we wondered whether the epigenetic control of mTORC1 strictly depends on SESTRIN1 or whether there is a backup mechanism. Treatment with the EZH2 inhibitor, GSK126, blocked mTORC1 signaling in EZH2Y641X mutant cells (SU-DHL-10, SU-DHL-6, and Karpas-422) as indicated by loss of ribosomal S6 and 4E-BP phosphorylation (Fig. 4C and fig. S4F). By contrast, SESTRIN1 isogenic deficient cells generated using clustered regularly interspaced short palindromic repeats (CRISPR)/Cas9 or shRNAs showed no effect on mTORC1 activity (Fig. 4D and fig. S4, G and H). Moreover, metabolic labeling of nascent protein production with a methionine synthetic analog [l-azidohomoalanine (AHA)] revealed that SESTRIN1-proficient lymphoma cells readily reduced translation rates by 23%, whereas SESTRIN1-defective lymphoma cells were unable to reduce translation in response to EZH2 inhibition; rapamycin was used as control, and, as expected, it showed equal translation inhibition independent of SESTRIN1 (Fig. 4, E and F). Hence, SESTRIN1 is an important node that links EZH2 and the control of mTORC1 and mRNA translation.

Therapeutic EZH2 inhibition depends on SESTRIN1

The mutant EZH2Y641X protein can potentially control a large number of genes, and it is not clear which of these are required for response to an EZH2 inhibitor. We assessed the relative importance of SESTRIN1 by treating isogenic SESTRIN1-competent and SESTRIN1-deficient lymphoma cells with GSK126. Treatment of EZH2Y641X mutant SU-DHL-10 cells with GSK126 readily induced cell death in the parental line, whereas SESTRIN1 inactivation in the same cells prevented cell death (fig. S4, I and J). Next, we directly compared the in vivo response of isogenic pairs of SESTRIN1-proficient (SU-DHL-10 vector) and SESTRIN1-deficient (SU-DHL-10 CRISPR-Sestrin1) lymphomas to EZH2 inhibition. We xenografted lymphomas into the right and left flanks of immunodeficient (NOD.CB17-Prkdcscid) mice. We initiated treatment with EZH2 inhibitor (GSK126) [75 mg/kg intraperitoneally (ip) daily for 14 days] or vehicle once tumors had reached a size of 5 to 10 mm3. In SESTRIN1-proficient tumors, GSK126 caused a significant (P = 0.013) growth reduction compared to vehicle (Fig. 4G); in contrast, the SESTRIN1-deficient tumors failed to respond to the EZH2 inhibitor (Fig. 4, H and I). Tumor analyses by immunohistochemistry and immunoblots readily showed differential induction of apoptosis by cleaved caspase3 and differential inhibition of mTORC1 activity by phosphorylation of ribosomal S6 and 4E-BP1 proteins between SESTRIN1-proficient and SESTRIN1-deficient tumors (Fig. 4J and fig. S4, K and L). These results indicate that mutated EZH2 controls mTORC1 activity through SESTRIN1 in lymphoma.

We have seen that mutant EZH2 can override the p53-mediated induction of SESTRIN1. We now wondered whether EZH2 inhibition and SESTRIN1 expression also contribute to DNA damage response induced by chemotherapeutic agents. We pretreated SU-DHL-10 lymphoma cells with the EZH2 inhibitor (GSK126, 2 μM) for 1 day, and then we added DXR for an additional day (DXR, 50 nM). Pretreatment with the EZH2 inhibitor reduced mTORC1 activity, as indicated by a decrease in phospho-S6, but did not significantly enhance the antiproliferative effect of DXR in vitro (fig. S4, M and N).

EZH2 mutant lymphomas are sensitive to a bifunctional mTOR inhibitor

EZH2 mutant lymphoma cells show down-regulation of SESTRIN1 and mTORC1 activation. We wondered whether these tumors were also more sensitive to pharmacological mTORC1 inhibition. RapaLink-1 is a third-generation mTORC1 inhibitor that binds both the FRB (FKBP-rapamycin binding) and the mTORC kinase domains and therefore combines the mTORC1 selectivity of rapamycin with the better target inhibition of mTOR kinase inhibitors (Fig. 4K) (34). We treated a panel of EZH2 wild-type (OCI-LY19, Toledo, and DoHH2) and EZH2 mutant (SU-DHL-10, SU-DHL-4, and SU-DHL-6) lymphoma cell lines with low (12.5 nM) and high (50 nM) dose of RapaLink-1 for 48 hours. EZH2 mutant cells were more sensitive to this inhibitor than EZH2 wild-type lines. Specifically, we observed cell death induction in 80 to 95% of EZH2Y641X mutant cells compared to 10 to 70% of EZH2WT cells for both low and high drug concentrations (Fig. 4L and fig. S4O). These results indicate specific pharmacological vulnerabilities of EZH2Y641X mutant lymphomas.


SESTRIN1 is a known regulator of mTORC1 that is transcriptionally induced by p53 upon DNA damage and mediates cell growth inhibition upon genotoxic stress (1719). Our study identifies the SESTRIN1 gene as a bona fide tumor suppressor in lymphoma and a biologically relevant target of the frequent Chr. 6q deletions that are observed in ~20% of human lymphomas. These chromosomal deletions are typically large and encompass several genes, including previously reported tumor suppressors in lymphoma such as TNFAIP3, EPHA7, and PRDM1 (14, 15, 35). SESTRIN1 is rarely focally deleted and could thus be missed when considering genomic data alone. Nonetheless, through complementary functional analyses using a chimeric mouse model of lymphoma, we could demonstrate the importance of SESTRIN1 in FL development.

In addition to chromosomal loss, SESTRIN1 is epigenetically silenced by the lymphoma-specific mutant EZH2Y641X, resulting in mTORC1 activation in lymphoma. SESTRIN1 is not a target of mutations, in contrast to the RRAGC, which is mutated in 7% of FLs (11). Mutations in EZH2 and RRAGC and loss of SESTRIN1 show a mutually exclusive relationship, indicating that they are alternate paths to maintain mTORC1 activity under conditions that would otherwise shut down cell growth. The combined frequency of these lesions (47% of examined cases of FL and DLBCL) indicates that maintaining active mTORC1 signaling is an important biological requirement in a large group of lymphomas. There are functional differences between SESTRIN1 and RRAGC. Although SESTRIN1 controls the Rag GTPases RAG A and RAG B in response to genotoxic stress (19, 20), RRAGC encodes for the RAG C protein that, together with RAG D, mediates nutrient-sensitive mTORC1 control (36). However, mutual exclusivity between RRAGC and SESTRIN1 alterations indicates that either lesion is sufficient to maintain mTORC1 in lymphoma cells.

The polycomb factor EZH2 is a target of activating mutations in 7 to 20% of human FLs, promoting increase in H3K27me3 genome-wide and thus targeting several genes. We showed that SESTRIN1 is an important EZH2Y641X target, and it may influence therapeutic response to EZH2 inhibitors like GSK126. Our findings show that EZH2 controls mTORC1 through SESTRIN1, and this epigenetic control can override p53 and DNA damage–induced mTORC1 blockade. We further find that SESTRIN1 reactivation and mTORC1 control are important downstream effects of EZH2 gain-of-function mutations. This is supported by the genetic relationship between these lesions and experimental data showing that SESTRIN1 loss protects against EZH2 inhibition. Although we do not have direct clinical evidence and it is difficult to test the efficacy of EZH2 inhibitors on primary FL samples because of their limited ability to grow in vitro, these results suggest that SESTRIN1 deletions may predict reduced efficacy of EZH2 inhibitor treatment in lymphoma. Conversely, our results reveal that EZH2-mediated epigenetic down-regulation of SESTRIN1 increases the dependency of lymphomas on mTORC1 and induces sensitivity to mTORC1 inhibitors in EZH2 mutant lymphomas. Together, our study establishes an epigenetic mechanism of mTORC1 control and suggests that SESTRIN1 status should have an impact on the response to lymphoma therapeutics.


Study design

This study was designed to identify tumor suppressor genes in FL that are targeted by chromosomal deletions and/or epigenetically silenced. Initially, we analyzed array comparative genomic hybridization data (aCGH) for 260 FL patients and used an shRNA screen in murine pro–B cells (as a surrogate model) to identify genes that control B cell proliferation. We confirmed the screen results in pro–B cells and in a murine model of FL (vavP-Bcl2) (Figs. 1 and 2). We calculated the minimum number of animals needed to observe significant differences in the survival curve by power analysis. In the second part of the study (Figs. 3 and 4), we focused on epigenetic regulation. We analyzed expression data in FL and GC-DLBCL patient samples that were EZH2WT or EZH2Y641X. We selected patients harboring EZH2Y641X mutations; we excluded a cell line with UTX mutations and patients harboring different mutations for EZH2. To confirm a direct effect of mutated EZH2 on H3K27me3, we compared H3K27me3 ChIP at the gene-specific locus in mutant or wild-type EZH2 cells. To compare different responses to the EZH2 inhibitor (GSK126), we used genetically modified cells. Syngeneic cells were transplanted subcutaneously, and tumor size was determined by direct measurement of the tumors. We selected animals with similar tumor size and randomly assigned them to untreated and treated groups. The number of replicates and the statistical analyses are reported in the figure legends and below.

Statistical analysis

Significant differences in mRNA expression measured by RNA sequencing (fragments per kilobase of transcript per million) were determined by Wilcoxon rank test. Differential expression analyses on microarray data were performed using the Limma R package. Significant differences in expression measured by qPCR were determined by one- or two-tailed t test as reported in each figure legend. Statistical significance of mutual exclusivity was determined by random permutation tests (n = 1000). The bar graphs show the mean and SD or SEM as indicated in the figure legends. The exact P values are reported in each figure, and n.s. stands for non–statistically significant (P > 0.05).


Materials and Methods

Fig. S1. Genomic analysis of frequently altered regions in 260 FLs.

Fig. S2. Characterization of vavBcl2 lymphoma mouse model.

Fig. S3. Characterization of epigenetics and expression of SESTRIN1, SESTRIN2, and SESTRIN3 in FL and DLBCL.

Fig. S4. Regulation of mTOR signaling by SESTRIN1 in lymphoma.

Table S1. Copy number analysis of 260 indolent FL samples.

Table S2. Quantification of homozygous and heterozygous deletions on Chr. 6q in 260 FL.

Table S3. Analysis of focal deletions on Chr. 6q.

Table S4. shRNA screen analysis.

Table S5. Sestrin1 expression in EZH2WT or EZH2Y641X GCB-DLBCL patient samples.

Table S6. Differential expression analysis of genes on Chr. 6q in EZH2WT or EZH2Y641X cell lines.

Table S7. Differential expression analysis of cell lines treated with the EZH2 inhibitor (GSK126) or DMSO.

Table S8. List of lymphoma cell lines with relevant genomic lesions.

References (3743)


  1. Acknowledgments: We thank the laboratories of M. Li and M. Peng (MSKCC), as well as D. Hanahan [Swiss Institute for Experimental Cancer Research–École Polytechnique Fédérale de Lausanne (ISREC-EPFL)], for sharing reagents, F. Martinon (University of Lausanne) and F. Bertoni (Istituto Oncologico della Svizzera Italiana) for sharing lymphoma cell lines, the Memorial Sloan Kettering and EPFL research animal facilities, and the pathology and flow cytometry core facilities. Funding: This work is supported by grants from the ISREC Foundation (to E.O.), Anna Fuller Fund (to E.O.), Swiss National Science Foundation (to E.O.), and Giorgi-Cavaglieri Foundation (to G.C.). A.V.B. is supported by the National Cancer Institute grant (CA172660). H.-G.W. is supported by the Lymphoma Research Foundation; Mr. William H. Goodwin and Mrs. Alice Goodwin and the Commonwealth Foundation for Cancer Research; the Center for Experimental Therapeutics at MSKCC; NIH grants RO1CA183876-01, R01CA207217-01, R01CA19038-01, and P50CA192937-01A1; the Starr Cancer Consortium; the Geoffrey Beene Cancer Research Center; a Leukemia and Lymphoma Society (LLS) SPORE grant; and the MSKCC Core Grant (P30 CA008748). H.-G.W. is a scholar of the LLS. R.K. is supported by the Princess Margaret Cancer Centre start-up fund. Author contributions: E.O. and N.K. conducted experiments and were involved in the experimental design and data analysis. M.J. and E.B. assisted with experiments. M.C.D. performed the Western blots for histone methylation marks. S.S. analyzed the ChIP-seq data. J.P.P. conducted the treatment experiment with mTOR inhibitor. V.R.S. designed and cloned CRISPRs. W.B. analyzed the GCB-DLBCL patient expression data. Y.J. and M.T. analyzed the expression data in FL patients. R.K., F.C.C., and S.P.S. provided FL samples for mutually exclusive analysis. A.M.M. supervised the expression analysis in primary DLBCL and FL patients and discussed the results. A.V.B. and A.P. provided Sestrin1 antibody and analyzed Sestrin1 expression by Western blot. G.C. designed and performed computational analyses. E.O. designed the study. E.O. and H.-G.W. wrote the paper. Competing interests: A.M. is a consultant for Epizyme. The other authors declare that they have no competing interests. Data and materials availability: The data for this study are available in the Gene Expression Omnibus database at accession numbers GSE40989, GSE12195, and GSE40971.

Stay Connected to Science Translational Medicine

Navigate This Article