Mutations in the neutral sphingomyelinase gene SMPD3 implicate the ceramide pathway in human leukemias

Somatic genetic alterations that arise during cancer progression may constitute incidental so-called passenger mutations, or they may be drivers of malignant proliferation, some of which may be appropriate therapeutic targets.¹  While large-scale nucleotide sequencing efforts hold the promise of comprehensive analysis of the cancer genome, initial studies have focused on known candidate cancer genes within relatively small numbers of tumors.^2-4  To complement such studies, genomic screens for gene copy number alterations have pointed to loci that may harbor recurrent abnormalities.^5,6  In fact, the initial identification of many critical tumor suppressor genes has relied on the study of rare tumors with homozygous genomic deletions targeting the relevant locus, identified by classical strategies, as well as genomic screens. These led to the discoveries of the retinoblastoma gene RB, the Wilms tumor suppressors WT1 and WTX, the melanoma susceptibility gene p16/INK4a, the familial breast cancer gene BRCA2, and the tumor suppressor PTEN.^7-11 

Homozygous genomic deletions may arise by a number of mechanisms, including the overlap of 2 different allelic losses. The relatively small size of homozygous deletions, compared with much larger regions of single allelic deletions detected by loss of heterozygosity (LOH), reflects such coincidental deletion events, as well as potential selection against the homozygous loss of essential genes that may flank the relevant tumor suppressor gene. Strategies to detect gene copy losses include comparative genomic hybridization (CGH), which have resulted in the identification of novel tumor suppressors,^12,13  although the signal-noise ratio in CGH favors detection of gene amplification rather than deletion. A powerful but lower-throughput method designed specifically for identifying homozygous deletions is genomic representational difference analysis (RDA),¹⁴  a polymerase chain reaction (PCR)–based subtractive hybridization of genomic tumor DNA from matching normal tissue. Both PTEN¹¹  and BRCA2¹⁵  deletion loci were first identified using RDA. However, the extreme sensitivity of RDA, which can detect as small as 15 kb (W.J.K. and D.A.H., unpublished data, April, 2003), complicates its application to primary human cancers, since it routinely detects polymorphic deletions, which are now known to be common in the human genome.¹⁶  To maximize the success of RDA screening for homozygous deletions associated with tumorigenesis, we adapted this strategy to analyze cell lines derived from syngeneic mouse tumor models.¹⁷  Mouse tumors initiated by loss of TP53 constitute a particularly interesting model, since their genomic instability may favor chromosomal events over point mutations during progression of tumorigenesis. As such, any initial discovery of candidate tumor progression genes in this model may provide clues to novel modifiers in a variety of human cancers.¹⁷ 

Ceramide is a lipid second messenger that triggers signal transduction pathways in response to cytokines or extrinsic cellular stresses, leading to a variety of cellular responses, including growth suppression and apoptosis.¹⁸  Ceramide may be synthesized de novo or generated from the hydrolysis of sphingomyelin by sphingomyelinases (SMases). The pleiotrophic effect of ceramide-dependent signaling has been attributed to distinct pools of ceramide generated by different catalytic enzymes at different cellular loci.¹⁹  To date, 4 SMases have been identified. SMPD1 encodes lysosomal acidic SMase (aSMase), whose deficiency is associated with the autosomal recessive disorder Niemann-Pick disease.²⁰  A total of 3 neutral SMases (nSMases)—SMPD-2, SMPD-3, and SMPD-4—are localized to different cellular compartments and expressed in different tissues.^21-23  SMPD2 is localized to the endoplasmic reticulum (ER) and expressed in all cell types, but the mouse knockout has no phenotype.²⁴  SMPD4 is localized to the ER as well as to the Golgi, and has been shown to be activated in response to tumor necrosis factor α (TNFα), although its physiologic role has yet to be defined.²³  In contrast to these 2 nSMase genes, Smpd3-null mice have significant developmental defects, including dwarfism and delayed puberty, attributed to hypothalamic pituitary deficiency,²⁵  and a naturally occurring mouse model of noncollagenous osteogenesis imperfecta (fro/fro), associated with skeletal deformities and fractures, results from inactivation of Smpd3.²⁶ SMPD3 is expressed at highest levels in the brain, is activated by TNFα, and contributes to TNFα-induced apoptosis in cultured cells.^27-29 

Here, we identified a gene-specific homozygous deletion of Smpd3 in a mouse TP53-driven osteosarcoma by RDA, and extended this analysis to a panel of human tumors, demonstrating missense mutations of SMPD3 in human leukemias. These observations raise the possibility that the ceramide pathway may be directly implicated in a subset of human cancers.

Mouse osteosarcoma cell lines derived from TP53 heterozygous mice^30,31  and MDCK cells were maintained in Dulbecco modified Eagle medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin at 37°C in a humidified 5% CO₂ incubator. Transient transfection of SMPD3 expression constructs was performed using Lipofectamine 2000 reagent (Invitrogen, Carlsbad, CA) according to the manufacturer's instruction. Retroviral transduction of SMPD3 expression constructs into adherent F4328 cells was modified from a previously described method for suspension culture.³² 

RDA of F4328 mouse osteosarcoma cell line was carried out as described previously.¹⁷  For genomic Southern blot, genomic DNA was digested by BamHI (New England Biolabs, Ipswich, MA), transferred onto nitrocellulose membrane (GE Healthcare, Little Chalfont, United Kingdom) and probed by radiolabeled RDA fragments cloned and excised from the pCR2.1 plasmid (Invitrogen). Primers for genomic PCR were designed from the sequences in the mouse chromosome 8 genomic contig NT 039467 (National Center for Biotechnology Information [NCBI]) and listed in Table S1A (available on the Blood website; see the Supplemental Materials link at the top of the online article).

For determination of cellular proliferation, cells were plated in 96-well tissue-culture plate, grown, and fixed at appropriate time points by 4% paraformaldehyde in phosphate-buffered saline (PBS). All plates were stained by SYTO60 dye (Molecular Probes, Eugene, OR) simultaneously and quantitated using the Odyssey Imaging System (LI-COR, Lincoln, NE). TNFα-mediated cell viability was measured by MTS assay. In brief, cells in 96-well tissue-culture plates were treated with appropriate concentration of recombinant mouse TNFα (Pierce Endogen, Rockford, IL). After 4 days of incubation, 10 μL MTS solution (Promega, Madison, WI) was added to each well and incubated for 2 hours to develop color. The absorbance was measured spectrophotometrically at a wavelength of 490 nm.

All coding exons of SMPD3 were amplified by genomic PCR from a panel of 67 sporadic human cancer cell lines, including lung (NCI-H460, NCI-522, HOP92, NCI-15C2, NCI-1734, NCI-446, NCI-H128, NCI-H596, and NCI-H2009), ovarian (ES-2, IGROV-1, MDAH2774, OV1063, OVCAR3, OVCAR4, OVCAR5, OVCAR8, SKOV3, and SW626), brain (SF-295, SNB-19, U-251, CCF-STTG1, SW-1088, SW1783, T98G, MO59K, SK-N-DZ, and SK-N-MC), leukemia (CCRF-CEM, MOLT-4, SR, K562, and RPMI8226), prostate (DU-145 and PC-3), colon (COLO-205, HCT-116, HCT-15, HT29, and SW-620), renal (786-O, ACHN, CAKI-1, SN-12C, and UO-31), skin (LOX-IMV1, M14, and SKMEL2), bone (SAOS-2), and breast (BT549, BT483, HS157, UACC893, HS467T, HS496T, HS578T, MCF7, MDAMB175, MDAMB231, MDAMB415, MDAMB436, MDAMB 453, MDAMB468, T47D, BR-412, and MCF7ADR) cancers. For mutation screening of SMPD3 from leukemia samples, genomic DNA from 92 primary acute myeloid leukemia (AML; 43 from Massachusetts General Hospital [MGH] and 49 from Inserm), 95 primary acute lymphoblastic leukemia (ALL), and 33 T-cell ALL (T-ALL) cell lines (ALLSIL, BE13, CCRFHSB2, CMLT1, DND41, EOL1, HBPALL, JMJURKAT, K3P, KARPAS45, KE37, KGI, KOPTK1, LOUCY, Mk-B1, MOLT-13, MOLT-15, MOLT-16, MV411, P12ICHIKAWA, PEER, PF382, REH, REX, SKM, SL-W3, SUPT1, SUPT11, SUPT13, SUPT7, TA527, TALL1, and THP; Dana-Farber Cancer Institute, Boston, MA) were collected and analyzed by genomic sequencing. Epstein-Barr virus (EBV)–immortalized lymphoblastoid cell lines established from 170 healthy blood donors provided for normal population controls. PCR primers for exon amplification and nucleotide sequencing are listed in Table S1B.

The coding sequence of human SMPD3 was amplified by PCR from cDNA and cloned into pcDNA6/V5-HIS or pcDNA4/Flag (Invitrogen) or Flag-tag version of MSV-SV-GFP³²  plasmids to generate cytomegalovirus (CMV) promoter- or retrovirus-driven expression constructs, respectively. Both V5 and Flag tags were located at the C-terminus of SMPD3. To generate SMPD3 mutant alleles, PCR mutagenesis was used as described.³³ 

F4328 cells were disrupted by repeated freezing and thawing in lysis buffer (25 mM Tris-Cl [pH 7.4], 5 mM EDTA, 10 mM DTT, and protease inhibitor cocktail [Roche, Indianapolis, IN]). Crude membrane fractions were sedimented by centrifugation for 2 minutes at 2500g and extracted with lysis buffer containing 0.2% Triton X-100. After centrifugation, 10 μg of supernatants were incubated with 20 nM [choline-methyl-¹⁴C] sphingomyelin (PerkinElmer, Waltham, MA) for 30 minutes at 37°C in 100 μL of 100 mM Tris-Cl (pH 7.4), 10 mM MgCl₂, 0.2% Triton X-100, and 10 mM dithiothreitol. Reactions were terminated by the addition of 1 mL chloroform/methanol (2:1), and the aqueous phase was separated by the addition of 200 μL water. The radioactivity of the upper phase containing phosphorylcholine was measured using a liquid scintillation counter.

F4328 cells in a 6-well plate format were transfected with 1 μg V5-tagged wild-type or D358G SMPD3 in pcDNA6. At 48 hours after transfection, cells were treated with 100 μg/mL cycloheximide for various times, and total cell lysates were prepared for Western blot analysis.

MDCK cells grown on coverslips were transiently transfected with either Flag- or V5-tagged SMPD3 or mutant SMPD3 in pcDNA vector using Lipofectamine 2000 (Invitrogen). After 48 hours, cells were fixed with 4% paraformaldehyde in PBS for 10 minutes, permeabilized with 0.5% Triton X-100 in PBS for 10 minutes, and blocked with blocking solution (5% goat serum, 5% horse serum, 0.2% fish gelatin, and 0.2% Tween 20 in PBS) for 30 minutes at 37°C. Cells were then incubated with anti-V5 (1:250) or anti-Flag (1:250) monoclonal antibodies in blocking solution for 1 hour at 37°C, washed, probed with anti-mouse Alexa 488 (1:500; Invitrogen), washed and mounted with DAPI. A Nikon Eclipse 90i fluorescence microscope (Nikon, Melville, NY) was used for visualization of staining. NIS-Element AR 2.30 software (Nikon) was used for the acquisition of images.

Anti-V5, anti-Flag, and antitubulin antibodies were purchased from Invitrogen, Sigma-Aldrich (St Louis, MO), and Santa Cruz Biotechnology (Santa Cruz, CA), respectively.

We undertook an RDA screen of 3 osteosarcoma cell lines derived from mouse tumors with a heterozygous germ-line TP53 deletion.³⁰  No homozygous deletions were identified in 2 cell lines, while the third (F4328) harbored homozygous deletions at a number of loci, including the known tumor suppressor gene PTEN. Of 4 deletion loci that did not harbor a known tumor suppressor, we selected a small homozygous deletion on chromosome 8q for detailed analysis. The chromosome 8q deletion produced 4 nonredundant RDA fragments (Figure 1A), which were used to confirm the homozygous deletion by genomic Southern blot (Figure 1B). Based on the sequence of mouse chromosome 8 genomic contig NT 039467 (NCBI), we mapped the boundary of the deletion by genomic PCR, defining a deletion of 100 kb. This gene-poor region contains the first 2 5′ exons of the Smpd3 gene, which is localized to the centromeric border of the deletion. The coding exons of the neighboring gene Zfp90, which are localized at the telomeric border of the deletion, are not affected. However, since the deletion affects sequences upstream of this gene, we confirmed expression of Zfp90 in F4328 cells by reverse transcription (RT)–PCR. In contrast, Smpd3 expression is absent (Figure 1C).

To test for functional consequences of the Smpd3 deletion in F4328 cells, we reconstituted human SMPD3 expression in these cells using retroviral infection with a bicistronic green fluorescent protein (GFP)–linked vector, allowing monitoring of infected cells by flow cytometry. The more than 80% infection rate allowed analysis of pooled infected F4328 cells, avoiding clonal selection bias. SMPD3-reconstituted cells displayed a modest suppression of proliferation (Figure 1D) and showed no obvious difference in contact inhibition-induced growth arrest (data not shown). However, cell viability following treatment with TNFα was significantly decreased in SMPD3-reconstituted cells compared with parental F4328 cells lacking the endogenous gene (Figure 1E). While the functional consequences of restoring SMPD3 expression and ceramide biogenesis in cancer cells with a deletion of the endogenous gene may be complex, TNFα-induced reduction of cell viability provides an initial measure of ceramide signaling.

To determine whether SMPD3 is also targeted by mutations in human cancers, we first screened a cohort of 67 sporadic human cancer cell lines from various tissue origins, including lung, ovary, brain, leukemia, prostate, colon, renal, skin, bone, and breast by bidrectional nucleotide sequencing. A total of 2 missense mutations in SMPD3 were identified in the leukemia cell lines CCRF-CEM (ALL) and MOLT-4 (ALL). We therefore expanded our analysis to a panel of 92 cases of primary AML, 95 cases of primary ALL samples, and 33 ALL cell lines, identifying SMPD3 mutations in an additional 11 cases. Thus, in total, SMPD3 mutations were present in 8 (6%) of 131 ALLs and 5 (5%) of 92 AMLs (Table 1). All mutations led to nonsynonymous missense changes. Of 7 cell lines with SMPD3 mutations, 4 are heterozygous, and 2 are compound heterozygotes. One cell line has 3 missense mutations including one homozygous mutation (E499G). All primary leukemia specimens displayed both mutant and wild-type sequences (Figure S1), presumably due to normal cell contamination. A total of 2 SMPD3 mutations were recurrent: D194N was present in 5 independent cases and D358G was found in 3 cases. Although remission samples from leukemia cases were not available for analysis, all mutations identified in these leukemias were absent from 170 healthy population controls (340 alleles), and from the NCBI single nucleotide polymorphism database.³⁴  Amino acid alignment of SMPD3 from various vertebrate animals showed that of the 11 mutations, 9 mutations occurred in amino acids that are conserved in all vertebrate animals for which SMPD3 sequence is known, while 5 of them are conserved in mammals (Figure 2). Thus, most of the SMPD3 mutations in human leukemias affect highly conserved amino acids. The overall amino acid identity in SMPD3 from these species compared with human is 91% (murine), 89% (bovis and canine), 72% (chicken), 69% (xenopus), and 56% (zebrafish). In addition, computational prediction of potential impact of an amino acid substitution on the structure and function of a human protein using the Polyphen program (http://genetics.bwh.harvard.edu/pph) resulted in 4 benign, 3 possibly damaging, and 4 probably damaging mutations that are relatively well correlated with their conservation among species (Table S2).

To study the effects of leukemia-associated mutations, we generated retroviral expression constructs of wild-type SMPD3 and the mutant alleles. Following infection of F4328 cells, nSMase activity of cellular lysates was quantified by measuring hydrolysis of sphingomyelin into phosphorylcholine and ceramide. nSMase activity was minimal in parental F4328 cells, indicating that SMPD3 is the primary contributor to sphingomyelin hydrolysis activity in these cells. As expected, cells infected with wild-type construct had increased nSMase activity that was appropriately restricted to the crude membrane fraction, with minimal enzymatic activity in cytosolic fractions. Cells infected with the various SMPD3 mutants showed no gross defect in nSMase activity with the exception of the recurrent mutation D358G, which showed no increase in enzymatic activity compared with the empty vector control. (Figure 3A). While similar infection rates were observed for each construct by flow cytometry, Western blot analysis showed strikingly decreased protein expression of D358G SMPD3. The consistently low levels of D358G mutant expression raised the possibility of an inherently unstable protein. Indeed, transient transfection of a CMV-driven construct into F4328 cells showed comparable levels of RNA expression for wild-type and the D358G mutant, but a marked reduction in D358G protein expression (Figure 3B,C). To determine whether the D358G mutation leads to SMPD3 protein instability, we compared the half-life of wild-type and mutant proteins by measuring protein levels following treatment with cycloheximide to block new protein synthesis. Whereas protein turnover, as measured by cycloheximide/Western blot analysis, was undetectable for wild-type SMPD3 (t_1/2 over 12 hours), the half-life of the D358G mutant was as short as 8 hours (Figure 3D,E). Thus, the recurrent D358G mutation appears to encode a grossly unstable SMPD3 protein.

Biochemical activity for nSMase has been observed primarily in the plasma membrane, and relocalization of SMPD3 from Golgi to the plasma membrane has been observed upon its activation by contact inhibition in MCF7 cells,²⁷  by oxidative stress in lung epithelial cells,³⁵  and at baseline in Hep-G2 cells.³⁶  To further analyze functional defects of SMPD3 mutants, we determined the subcellular localization of SMPD3 and each mutant form in the well-characterized MDCK kidney epithelial cell line. SMPD3 was localized to the plasma membrane in confluent MDCK cells (Figure 4D). Of the 11 mutants tested, 2 SMPD3 mutants, D358G and G248S, failed to localize to the plasma membrane under these conditions, showing diffuse staining throughout the cytoplasm (Figure 4G,J).

We have demonstrated that a subset of human leukemias have mutations in the nSMase gene SMPD3, which encode the major catalytic enzyme responsible for the generation of ceramide in response to various cellular stresses. Of the 11 mutations identified, 2 mutations showed gross defects in either protein stability and/or subcellular localization. The functional significance of the other 9 mutations is uncertain, and may require more physiologic analyses of hematopoietic differentiation. Matched normal tissues or remission specimens are not available for analysis, a common problem in mutational studies in leukemia, and we therefore cannot exclude the possibility that at least some of the other SMPD3 variants represent rare germ-line mutations. However, their identification as likely somatic mutations is supported by their absence in large number of population controls, the absence of SMPD3 mutations in other tumor types analyzed, and the fact that most mutations affect highly conserved residues across a different evolutionary species. Distinguishing true “driver” versus “passenger” mutations in human cancers may require detailed functional studies, as recently demonstrated in an analysis of FLT3 mutations in leukemia.³⁷  Consistent with that study, the presence of clearly defective SMPD3 alleles in a subset of leukemias supports that other SMPD3 mutations may have more moderate effects that are less readily defined in vitro. Together, these observations point to the ceramide pathway as being targeted by mutations in human leukemias.

Ceramide has been proposed as an antioncogenic lipid second messenger in multiple cancer types, where decreased levels of ceramide or increased levels of the counterbalancing sphingolipid sphingosine-1-phosphate or its metabolic enzyme sphingosine kinase 1 have been observed in cancers (reviewed in Ogretmen and Hannun³⁸ ). However, genetic evidence supporting a specific role of the ceramide pathway in human cancer has been lacking. It is of interest that the sphingomyelin cycle was first discovered in HL-60 human myelocytic leukemia cells, where induction of differentiation along the monocytic lineage using vitamin D₃, TNFα, and γ-interferon, but not along other lineages, activates cellular nSMase, triggering sphingomyelin hydrolysis and ceramide accumulation.^19,39-41  Thus, ceramide metabolism may play a particularly important role in hematopoetic differentiation, and its dysregulation may contribute to a subset of human leukemias. In addition to the point mutations observed here, we note the chromosomal localization of SMPD3 on chromosome band 16q22.1, a locus rich in leukemia-associated chromosomal abnormalities.^42-44  Taken together, the finding that a significant subset of human leukemias harbors mutations in SMPD3 provides the first genetic evidence for alterations in ceramide metabolism in human cancers.

We are grateful to Drs D. Gary Gilliland, Vijay Yajnik, Charles Paulding, and Ryn Miake-Lye for helpful discussions, reagents, and critical reading.

This work was supported by grants from the National Foundation for Cancer Research (D.A.H.), the Doris Duke Distinguished Clinical Scientist Award (D.A.H.), and the National Insitutes of Health (CA68484 and CA109901 to A.T.L.)

National Institutes of Health

Contribution: W.J.K., R.A.O., S.M.H., L.E.P., M.G., and F.D. performed experiments; A.I.M. provided mouse osteosarcoma cell lines; O.A.B., and A.T.L. provided leukemia DNA samples; W.J.K., D.A.S., D.W.B., D.T.S., and D.A.H analyzed data; W.J.K and D.A.H designed the research and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

Correspondence: Daniel A. Haber, Cancer Center, Massachusetts General Hospital, 149 13th St, Charlestown, MA 02129; e-mail: haber@helix.mgh.harvard.edu.