The serine hydroxymethyltransferase-2 (SHMT2) initiates lymphoma development through epigenetic tumor suppressor silencing

Abstract

Cancer cells adapt their metabolic activities to support growth and proliferation. However, increased activity of metabolic enzymes is not usually considered an initiating event in the malignant process. Here, we investigate the possible role of the enzyme serine hydroxymethyltransferase-2 (SHMT2) in lymphoma initiation. SHMT2 localizes to the most frequent region of copy number gains at chromosome 12q14.1 in lymphoma. Elevated expression of SHMT2 cooperates with BCL2 in lymphoma development; loss or inhibition of SHMT2 impairs lymphoma cell survival. SHMT2 catalyzes the conversion of serine to glycine and produces an activated one-carbon unit that can be used to support S-adenosyl methionine synthesis. SHMT2 induces changes in DNA and histone methylation patterns leading to promoter silencing of previously uncharacterized mutational genes, such as SASH1 and PTPRM. Together, our findings reveal that amplification of SHMT2 in cooperation with BCL2 is sufficient in the initiation of lymphomagenesis through epigenetic tumor suppressor silencing.

Extended Data Fig. 1 |. De novo serine synthesis pathway in human B cell lymphoma.

a, Bar graphs presenting the functional gain (red) or loss (blue) of SHMT2 located on chromosome 12 by DNA copy number analysis of 568 DLBCL and 176 FL tumors. b, The frequency of loss (blue), gain (red) or diploid (black) status of serine synthesis pathway genes in 568 human DLBCL and 176 human FL tumors. c, Diagrams demonstrating the overlaps of amplification (red) and loss (blue) of SHMT2 vs other serine biosynthesis pathway enzymes (PSPH, PHGDH, PSAT1, SHMT1) in 568 DLBCL and 176 FL tumors. d, Bar graph showing the frequency of SHMT2 amplification in different subtype (GC-like, ABC-like or unclassified) of 249 human DLBCL tumors.

Extended Data Fig. 2 |. SHMT2 promotes lymphomagenesis in vivo.

a, Diagram of FL mouse model. Fetal VavPBcl2 HSCs were transduced by MSCV-GFP plasmid carrying SHMT2 cDNA or empty vector and injected to lethally irradiated female mice. b, Representative graphs of flow cytometry analysis comparing GFP⁺ HSCs before injection vs GFP⁺ splenic lymphoma cells from VavPBcl2;vector- and VavPBcl2;SHMT2- induced tumors collected 5 months after injection. c, Dot plot representing the initial GFP⁺ cells in hematopoietic stem cells before injection vs GFP⁺ cells enriched in splenic cells collected from VavP-Bcl2;vector (N = 5 mice) and VavP-Bcl2;SHMT2 (N = 10 mice) tumors. Two-tailed Student’s t-test was used to determine statistical significance; VavP-Bcl2;vector: P(HSCvsLymphoma)= 0.443, NS; VavP-Bcl2;SHMT2: P(HSCvsLymphoma)=0.0006. d, Representative images of histology studies of VavPBcl2; vector and VavPBcl2;SHMT2 lung. The slides were stained with H&E, and antibodies for B220, TUNEL, Ki67, PNA. This experiment was independently repeated three times with similar results. Scale Bars, 500 nm. e, tumor clonality analysis on B220⁺ cDNA collected from VavPBcl2;vector vs VavPBcl2;SHMT2 tumors. Each lane corresponds to one tumor. This experiment was independently repeated two times with four independent samples in each genotype with similar results f, Immunoblot against SHMT2, SHMT1 and ACTIN in DLBCL cell lines carrying two different short hairpins against SHMT2. This experiment was independently repeated two times with similar results. The uncropped images of the original blots are presented in Source Data Extended Data File 12. The numerical data for this figure are presented in Source Data Extended Data File 2.

Fig. 1 |. Genomic amplification of SHMT2 in human B-cell lymphomas.

a, Gain (red) across the genome in the analysis of SNP arrays of human DLBCL (n = 568 tumors). b, rCGH array analysis of human FL (n = 176 tumors). c, Gene expression of SHMT2 in the subtypes of 249 human DLBCL samples with wild-type (diploid) or amplified (gain) SHMT2. A two-tailed Student’s t-test was used to determine statistical significance. GC: P(diploid versus gain) = 0.000016, n = 52 diploid tumors, n = 39 tumors with SHMT2 gain. ABC: P(diploid versus gain) = 0.0001, n = 88 diploid tumors, n = 13 tumors with SHMT2 gain. Unclassified: P(diploid versus gain) = 0.1206, n = 35 diploid tumors, n = 6 tumors with SHMT2 gain. The numerical data for this figure are presented in Source Data File 1.

Fig. 2 |. SHMT2 acts as an oncogenic driver in a mouse model of FL.

a, Kaplan–Meier survival curve of female C57BL/6 mice bearing the VavP-Bcl2; vector (black; n = 43 mice) or VavP-Bcl2; SHMT2 (red; n = 34 mice) tumors. A log-rank test determined the statistical significance of survival between two different groups; VavP-Bcl2; SHMT2 versus VavP-Bcl2; vector, P(log-rank test) = 0.000009. b, Representative SHMT2 immunoblotting of sorted B cells from three VavP-Bcl2; vector and three VavP-Bcl2; SHMT2 tumors. The immunoblotting was performed independently three times with the same results. The uncropped images of the original blots are presented in Source Data File 2. c, Fraction of labeled and unlabeled glycine to total glycine measured by LC–MS in extracts of isolated B cells from VavP-Bcl2; vector (n = 3 mice) or VavP-Bcl2; SHMT2 (n = 2 mice) tumors labeled with ¹³C₃,¹⁵N-serine. C: glycine; C+1: ¹³C₁-glycine; C+2: ¹³C₂-glycine; N+1: ¹⁵N-glycine. d, Ratio of SAM to methionine measured by LC–MS in extracts of isolated B cells collected from VavP-Bcl2; vector (n = 4 mice) or VavP-Bcl2; SHMT2 (n = 5 mice) tumors. A two-tailed Student’s t-test was used to determine statistical significance. P(VavP-Bcl2; vector versus VavP-Bcl2; SHMT2) = 0.1332. e, Representative histological micrographs of spleens collected from VavP-Bcl2; vector and VavP-Bcl2; SHMT2 tumors. The sections were stained with H&E, anti-B220, anti-Ki-67, TUNEL and anti-PNA. This experiment was independently repeated on samples from three different mice in each genotype with similar results. Scale bar, 500 nm. The histological micrographs of replicates are presented in Source Data File 4. f, Representative images of flow cytometry analysis of cellular composition of VavP-Bcl2; vector versus VavP-Bcl2; SHMT2 splenic tumor cells. This experiment was independently repeated on samples from three different mice in each genotype with similar results. The flow cytometry analysis of replicates are presented in Source Data File 4. The numerical data for this figure are presented in Source Data File 3.

Fig. 3 |. MYC regulates the expression of SHMT2 in both human and mice tFL.

a, Scatter plot of genes differentially expressed in mouse and human tFLs compared to mouse and human FL. The red dots label genes differentially expressed in both mouse and human samples. The red box outlines the genes overexpressed in both mouse and human tFLs. n = 6 human FL; n = 6 human tFL; n = 8 VavP-Bcl2; vector; n = 8 VavP-Bcl2; MYC. b, Scatter graph representing the correlation of SHMT2 expression with MYC expression in 249 human DLBCLs. Pearson’s r = 0.3839, two-tailed P(SHMT2 versus MYC) = 0.000001. c, Relative mRNA expression of serine-glycine biosynthetic pathway genes in the P493-6 cell line. Cells (n = 2 independent experimental replicates) were treated for 72 h with DOX (MYC^off); subsequently, the DOX was removed (MYC^on) and mRNA expression was measured at several different time points after DOX treatment. d, Heatmap showing the expression of serine-glycine biosynthetic pathway genes in spleen (n = 4 mice), VavP-Bcl2; vector (n = 8 mice) and VavP-Bcl2; MYC (n = 8 mice) B cells. e, Bar graphs representing the viability of DLBCL cell lines following SHMT2 knockdown by shRNA (shSHMT2_1 or shSHMT2_2). A two-tailed Student’s t-test was used to determine statistical significance. SU-DHL-4: P(vector versus shSHMT2_1) = 0.0246; P(vector versus shSHMT2_2) = 0.0343; SU-DHL-8: P(vector versus shSHMT2_1) = 0.0034; P(vector versus shSHMT2_2) = 0.00007; Toledo: P(vector versus shSHMT2_1) = 0.0216; P(vector versus shSHMT2_2) = 0.000039. *P < 0.05, **P < 0.01, ***P < 0.001. n = 8 cell culture replicates for all experimental groups presented in this panel. Eight replicates are the combination of three independent experiments. f, Fraction of labeled and unlabeled glycine to total glycine measured by LC-MS in extracts of ¹³C₃,¹⁵N-serine labeled (0.8 mM) SU-DHL-4 cells carrying control vector or shSHMT2. C: glycine; C+1: glycine ¹³C₁; C+2: glycine ¹³C₂. Vector: n = 2 experimental replicates; shSHMT2_1: n = 2 experimental replicates; shSHMT2_2: n = 2 experimental replicates. g, SAM/methionine measured by LC-MS in extracts of SU-DHL-4 carrying vector or short hairpin against SHMT2 (shSHMT2_1 or shSHMT2_2). A two-tailed Student’s t-test was used to determine statistical significance; n = 2 independent experimental replicates. h, Bar graphs showing ATP content in human B-cell lymphoma cell lines after overexpression and knockdown of SHMT2. n = 2 vector and n = 5 SHMT2 technical cell culture replicates of the P493-6 cell line; n = 3 technical cell culture replicates of SU-DHL-8. This experiment was repeated independently three times with similar results. i, Survival curve of C57BL/6 mice carrying VavP-Bcl2; MYC (n = 27 mice) and VavP-Bcl2; MYC; shShmf2 (n = 28 mice) tumors. A log-rank test was used to determine the statistical significance of survival between two different groups; P(VavP-Bcl2; MYC versus VavP-Bcl2; MYC; shShmt2) = 0.0033. j, Immunoblot against SHMT2 and MYC in B cells collected from VavP-Bcl2; MYC and VavP-Bcl2; MYC; shShmt2 tumors. The immunoblot was performed twice on three samples from three different mice with the same genotype with similar results. The uncropped images of the original blots are presented in Source Data File 5. The numerical data for this figure are presented in Source Data File 6.

Fig. 4 |. Targeting SHMT2 activity or expression for lymphoma therapy.

a, Bar graphs presenting the viability of DLBCL cell lines in response to SHIN1 on day 4 compared to day 0 at different concentrations of the drug (n = 3 technical cell culture replicates). This experiment was repeated independently three times with similar results. b, Representative bivariate density plot (upper panel) and scatter plot (lower panel) of cell death analysis (annexin V and 7-ADD) of the SU-DHL-4 cell line treated with SHIN1 (10 μM) or dimethyl sulfoxide (DMSO) for 48 h. This experiment was repeated independently three times with similar results. The flow cytometry analysis of the replicates is presented in Source Data File 9. c, Bar graphs showing the percentile of viable cells in SHIN1-treated versus DMSO-treated DoHH2 cells carrying vector, catalytic dead SHMT2 or SHMT2 construct; n = 3 technical cell culture replicates for all the treatment groups presented. This experiment was repeated independently twice with similar results. d, Relative mRNA expression of serine biosynthesis pathway enzymes in Karpas-422 cells after treatment with 5 μM BIX-01294 for 24 h. n = 3 technical cell culture replicates for all the treatment groups presented. This experiment was repeated independently with three different cell lines with similar results. e, Bar graphs demonstrating the viability of DLBCL cells after 3 d of treatment with different concentrations of BIX-01294. n = 4 technical cell culture replicates for all the treatment groups presented. This experiment was repeated independently three times with similar results. f, Representative scatter plot of cell death analysis (annexin V and 7-ADD) of the SU-DHL-4 cell line treated with BIX-01294 (2.5 μM) or ABT-199 (250 nM) or both for 72 h. This experiment was repeated independently three times with similar results. The flow cytometry analysis of the replicates is presented in Source Data File 9. g, Cell death analysis by immunoblotting against total and cleaved caspase-3 in DLBCL cell lines after treatment with BIX-01294 (2.5 μM), ABT-199 (250 nM) or their combination. This experiment was repeated independently twice with similar results. The uncropped images of the original blots are presented in Source Data File 7. The numerical data for this figure are presented in Source Data File 8.

Fig. 5 |. Epigenetic studies of VavP-Bcl2; SHMT2 B cells.

a, Bar graphs representing the relative abundance of mono-, di- and trimethylation of H3K4 in VavP-Bcl2; vector (n = 5 mice) and VavP-Bcl2; SHMT2 (n = 3 mice) B cells measured by LC-MS. A two-tailed Student’s t-test was used to determine statistical significance. Monomethylation: P(VavP-Bcl2; vector versus VavP-Bcl2; SHMT2) = 0.0175; dimethylation: P(VavP-Bcl2; vector versus VavP-Bcl2; SHMT2) = 0.5692; trimethylation: P(VavP-Bcl2; vector versus VavP-Bcl2; SHMT2) = 0.2318. b, Genomic distribution of H3K4me2 and H3K4me3 in VavP-Bcl2; SHMT2 (dimethylation: n = 2 mice; trimethylation: n = 3 mice) B cells compared to control (VavP-Bcl2; vector; dimethylation: n = 2 mice, trimethylation: n = 2 mice). c, Dot blot of DNA (100 nM, 50 nM and 25 nM) from SU-DHL-4 cells carrying vector or short hairpin against SHMT2 (shSHMT2). Methylene blue staining was used as the DNA loading control. This experiment was repeated independently three times with similar results. d, Dot blot of DNA (100 nM, 50 nM and 25 nM) from VavP-Bcl2; vector (n = 2) or VavP-Bcl2; SHMT2 HSCs (n = 2). Methylene blue staining was used as the DNA loading control. This experiment was repeated independently twice with three independent samples in each genotype with similar results. e, Dot blot of DNA (100 nM, 50 nM and 25 nM) from VavP-Bcl2; vector and VavP-Bcl2; SHMT2 lymphoma B cells. Methylene blue staining was used as the DNA loading control. This experiment was repeated independently twice with three different mice in each genotype with similar results. f, Pie charts showing the genomic distribution of hypo- and hyper-DMCs (upper diagrams) within CpG islands, shores and oceans (lower diagrams) in VavP-Bcl2; SHMT2 (n = 2 mice) versus VavP-Bcl2; vector (n = 2 mice) tumors. g, Scatter plot of genes downregulated in B cells from VavP-Bcl2; SHMT2 tumors versus control (VavP-Bcl2; vector) tumors with hyperDMCs on their promoters. The data from f were compared to the expression data (RNA-seq: n = 2 VavP-Bcl2; vector tumors versus n = 3 for VavP-Bcl2; SHMT2 tumors). The dotted line further indicates genes with >20 hyperDMCs. The uncropped images of the original dot blots in c,d,e are presented in Source Data File 11. The numerical data for this figure are presented in Source Data File 10.

Fig. 6 |. Epigenetic silencing of tumor suppressor genes contributes to the oncogenic action of SHMT2.

a, Mutation map of PTPRM and SASH1 (cBioPortal). b, Bar graph presenting the loss of SASH1 located on chromosome 6 by DNA copy number analysis in fractions of DLBCL (n = 568 patient samples) and FL (n = 176 patient samples) tumors. c, Analysis of SNP arrays of DLBCL (n = 568 patient samples) and rCGH array analysis of FL (n = 176 patient samples) tumors showing deletions between chromosomes 4 and 8. SASH1 is located on 6q24.1. d, Relative mRNA expression of selected target genes after 5-Aza treatment in DLBCL cell lines for 24 h. SASH1: SU-DHL-6: n = 2 untreated samples, n = 2 5-Aza-treated samples; SU-DHL-4: n = 3 untreated samples, n = 3 5-Aza-treated samples. PTPRM: SU-DHL-6: n = 2 untreated samples, n = 3 5-Aza-treated samples. DoHH2: n = 2 untreated samples, n = 2 5-Aza-treated samples. The number of samples presented are technical cell culture replicates. This experiment was repeated independently in at least three different DLBCL cell lines with similar results. e, Scatter graph representing the correlation of SHMT2 expression with SASH1 expression in human B-cell lymphoma (n = 249 human samples). Pearson’s r = −0.4295, two-tailed P(SHMT2 versus SASH1) = 0.000001. f, Relative mRNA expression of Ptprm and Sash1 in VavP-Bcl2; vector versus VavP-Bcl2; SHMT2 mouse B cells. A two-tailed Student’s t-test was used to determine statistical significance. Ptprm: P(VavP-Bcl2; vector versus VavP-Bcl2; SHMT2) = 0.0208, n = 2 VavP-Bcl2; vector mice, n = 3 VavP-Bcl2; SHMT2 mice. Sash1: n = 2 VavP-Bcl2; vector mice, n = 2 VavP-Bcl2; SHMT2 mice. The numerical data for this figure are presented in Source Data File 12.

Fig. 7 |. Characterization of candidate tumor suppressor genes in vivo.

a, Kaplan–Meier survival curve of female C57BL/6 mice bearing VavP-Bcl2; vector (black; n = 17 mice) or VavP-Bcl2; shSash1 (purple; n = 38 mice) or VavP-Bcl2; shPtprm (blue; n = 19 mice) tumors. A log-rank test determined the statistical significance of survival between three different groups; P(VavP-Bcl2; shSash1 versus VavP-Bcl2; vector) = 0.0167; P(VavP-Bcl2; shPtprm versus VavP-Bcl2; vector) = 0.0703. b, Representative images of flow cytometry analysis of cellular composition of VavP-Bcl2; shSash1 mouse splenic tumor cells. This experiment was repeated independently three times with similar results. c, Representative histological micrographs of spleens collected from VavP-Bcl2; vector and VavP-Bcl2; shSash1 tumors. The sections were stained with H&E and antibodies for B220 to stain B cells, and TUNEL and Ki-67 to analyze apoptosis and proliferation. This experiment was repeated independently three times with similar results. Scale bar, 500 nm. d, Relative mRNA expression of Ptprm and Sash1 in VavP-Bcl2; MYC (n = 2 mice) and VavP-Bcl2; MYC; shShmt2 (n = 3 mice) mouse B cells. A two-tailed Student’s t-test was used to determine statistical significance. Ptprm: P(VavP-Bcl2; MYC versus VavP-Bcl2; MYC; shShmt2) = 0.04; Sash1: P(VavP-Bcl2; MYC versus VavP-Bcl2; MYC; shShmt2) = 0.008. e, Bar graphs presenting the relative mRNA expression of PTPRM and SASH1 measured by qRT-PCR in human DLBCL cell lines after treatment with 10 μM of SHIN1 for 48 h. PTPRM: SU-DHL-6: n = 2 DMSO, n = 2 SHIN1; SU-DHL-4: n = 2 DMSO, n = 3 SHIN1. SASH1: SU-DHL-4: n = 3 DMSO, n = 3 SHIN1. The number of samples presented are technical replicates. This experiment was repeated independently three times with similar results. f, Bar graph demonstrating the viability of SU-DHL-4 cells with and without short hairpin against PTPRM (shPTPRM) or SASH1 (shSASH1) after treatment with SHIN1 for 4 d. Vector: n = 3 DMSO; n = 3 SHIN1. shSASH1: n = 2 DMSO; n = 2 SHIN1. shPTPRM-1: n = 2 DMSO; n = 3 SHIN1. shPTPRM-2: n = 2 DMSO; n = 3 SHIN1. The number of samples presented are technical replicates. This experiment was repeated independently three times with similar results. g, Schematic diagram of the mechanism of SHMT2 oncogenesis in lymphoma. The flow cytometry analysis and histological micrographs of the replicates is presented in Source Data File 13. The numerical data for this figure are presented in Source Data File 14.