Molecular profiling of LGL leukemia reveals role of sphingolipid signaling in survival of cytotoxic lymphocytes

Abstract

T-cell large granular lymphocyte (LGL) leukemia is characterized by clonal expansion of CD3(+)CD8(+) cells. Leukemic LGLs correspond to terminally differentiated effector-memory cytotoxic T lymphocytes (CTLs) that escape Fas-mediated activation-induced cell death (AICD) in vivo. The gene expression signature of peripheral blood mononuclear cells from 30 LGL leukemia patients showed profound dysregulation of expression of apoptotic genes and suggested uncoupling of activation and apoptotic pathways as a mechanism for failure of AICD in leukemic LGLs. Pathway-based microarray analysis indicated that balance of proapoptotic and antiapoptotic sphingolipid-mediated signaling was deregulated in leukemic LGLs. We further investigated sphingolipid pathways and found that acid ceramidase was constitutively overexpressed in leukemic LGLs and that its inhibition induced apoptosis of leukemic LGLs. We also showed that S1P(5) is the predominant S1P receptor in leukemic LGLs, whereas S1P(1) is down-regulated. FTY720, a functional antagonist of S1P-mediated signaling, induced apoptosis in leukemic LGLs and also sensitized leukemic LGLs to Fas-mediated death. Collectively, these results show a role for sphingolipid-mediated signaling as a mechanism for long-term survival of CTLs. Therapeutic targeting of this pathway, such as use of FTY720, may have efficacy in LGL leukemia.

Figure 1

Preparation of samples and experimental procedure. (A) Freshly isolated PBMCs from healthy individuals (naive normal PBMCs) were enriched for CD8⁺ T cells using negative isolation (naive normal enriched CD8⁺ cells) as described in “Methods.” Cells (5 × 10⁷) were activated using 1 μg/mL PHA for 3 days followed by 500 IU/mL IL2 for 7 days (activated normal PBMCs). Following activation, 10⁸ cells were enriched for CD8⁺ cells (activated normal enriched CD8⁺ cells). Percentage of CD3⁺CD8⁺ cells is shown in parentheses. (B) Experimental procedure: The samples were used for RNA isolation and subjected to microarray analysis. LGL leukemia PBMCs were obtained fresh from the patients and were not cultured, sorted, or activated. The samples with similar phenotype were either pooled (pooled sample analysis) or analyzed without pooling (unpooled sample analysis). Genes differentially expressed in both analyses were considered to be differentially expressed.

Figure 2

Microarray profiling of naive normal PBMCs, activated normal PBMCs, and leukemic LGLs using gene- and theme-based approaches. (A) Differential expression of genes in normal PBMCs following activation. A total of 775 genes are expressed differentially (at least 2-fold change and 1% FDR) following activation. EASE analysis of 2 phenotypes shows statistically significant up-regulation of genes belonging to GO categories “regulation of programmed cell death” and “positive regulation of apoptosis.” (B) Constitutive gene expression signature of leukemic LGLs compared with naive normal cells. A total of 174 genes were differentially expressed in leukemic LGLs compared with naive normal cells (at least 2-fold change and 1% FDR). EASE analysis of leukemic LGLs compared with naive normal cells shows significant up-regulation of genes belonging to GO categories such as immune system process, immune response, viral life cycle, viral infectious cycle, and viral genome replication. (C) Constitutive gene expression signature of leukemic LGLs compared with activated normal cells. A total of 1492 genes were differentially expressed in leukemic LGLs compared with activated normal cells (at least 2-fold change and 1% FDR). EASE analysis of leukemic LGLs compared with activated normal cells shows up-regulation of immune response–related signaling and cytotoxicity-related GO categories to be significantly up-regulated in leukemic LGLs compared with activated normal cells. Each row in the cluster image represents an individual gene, and each column represents an individual sample from LGL leukemia patient or healthy donor. The relative transcript abundance of each gene is color coded. A red color indicates high expression, black indicates intermediate expression, and green indicates low expression.

Figure 3

Differentially regulated apoptosis-related genes in leukemic LGLs compared with activated enriched normal CD8⁺ cells. (A) Apoptosis-related genetic signature unique to leukemic LGLs. The genes that were differentially regulated between leukemic LGLs and activated enriched normal CD8⁺ cells were identified (at least 2-fold change and 1% FDR). A total of 128 genes belonging to GO category “apoptosis” were differentially expressed in leukemic LGLs compared with activated enriched normal CD8⁺ cells (AcCD8⁺). Each row in the cluster image represents an individual gene, and each column represents an individual sample from LGL leukemia patient or healthy donor. The relative transcript abundance of each gene is color coded. A red color indicates high expression, black indicates intermediate expression, and green indicates low expression. (B) Differentially regulated genes between leukemic LGLs and activated enriched CD8⁺ cells were imported in GenMAPP software for visualization. Known apoptosis-related genes and their relation are shown. The genes constitutively up-regulated in leukemic LGLs compared with activated normal enriched CD8⁺ cells are shown in red, genes constitutively down-regulated in leukemic LGLs are shown in blue, and those in white show no change in expression between 2 phenotypes. The number accompanying each gene indicates fold change in expression.

Figure 4

Differential expression of TNFAIP3 (A20) mRNA in leukemic LGLs compared with activated normal PBMCs. Total RNA (10 μg) from 9 LGL patients and from 5 healthy individuals was analyzed by Northern blot hybridization. The blot was probed with 800-bp fragment of TNFAIP3 or housekeeping gene (GAPDH). Northern blot analysis shows that TNFAIP3 is constitutively expressed in naive normal PBMCs (N PBMCs, lanes 1-5) and that the expression was down-regulated following PHA+IL-2 activation of normal PBMCs (AC PBMCs, lanes 6-7). In contrast, the expression of TNFAIP3 gene transcripts was much higher in RNA from LGL leukemia patients (TLGLs, lanes 8-16) compared with activated normal PBMCs. (White spaces have been inserted to indicate realigned gel lanes.)

Figure 5

Sphingolipid metabolism and signaling pathway is differentially regulated in leukemic LGLs. The expression profile of LGL leukemia PBMCs (n = 30) was compared with that of activated normal PBMCs (n = 3) using Gene Set Enrichment Analysis (GSEA). Two of the 13 pathways enriched (FDR ≤ 15%) in leukemic LGLs are shown. The expression profile of the components of (A) EDG_Pathway (P = .006) and (B) ST_GA_12_pathway (P = .001) gene sets in leukemic LGLs compared with activated normal PBMCs. Each column represents individual sample from a LGL leukemia patient (gray) or healthy donor (yellow). Each row represents a gene. Red shows high expression, white denotes intermediate expression, and blue denotes low expression. (C) Microarray analysis of ASAH1 mRNA expression. The expression of ASAH1 mRNA in naive normal PBMCs (N PBMCs, n = 4), and activated normal PBMCs (AC PBMCs, n = 3) compared with LGL leukemia PBMCs (TLGLs, n = 30). Each bar represents mean relative fluorescence units, and error bars represent standard error of mean (SEM). (D) Differential expression of ASAH1 in LGL leukemia PBMCs. Expression of α-subunit of acid ceramidase in naive normal PBMCs (N PBMCs, n = 3) and activated normal PBMCs (AC PBMCs, n = 4) compared with LGL leukemia PBMCs (TLGLs, n = 6). Samples were subjected to SDS-PAGE followed by membrane transfer. The blot was probed with antibody to α-subunit of acid ceramidase or β-actin and developed using chemiluminescence. Western blot analysis suggests that acid ceramidase is expressed in naive PBMCs constitutively (N PBMCs, lanes 1-3). Following activation of lymphocytes, α-subunit of acid ceramidase is down-regulated to almost undetectable levels (AC PBMCs, lanes 4-7), whereas it is significantly up-regulated in all LGL leukemia PBMC samples (TLGLs, lanes 8-13). (E) Relative abundance of S1P receptors in naive enriched normal CD8⁺ cells (white dots, n = 4) and leukemic LGLs (black dots, n = 5) as analyzed by real-time PCR. The figure shows that S1P₅ is the predominant S1P receptor for leukemic LGLs. (F) Relative expression of S1P₁ (□), S1P₄ (), and S1P₅ (■) in leukemic LGLs (n = 5) compared with normal phenotypes (n = 3-5). A positive value indicates up-regulation, whereas a negative value indicates down-regulation of a gene in leukemic LGLs. Error bars represent standard deviation of expression. S1P₅ is up-regulated in LGL leukemia PBMCs compared with all normal phenotypes. S1P₁ is up-regulated in naive phenotypes compared with both activated phenotypes and leukemic LGL. (*P < .05; **P < .001.)

Figure 6

Differential regulation of sphingolipid metabolism and signaling in leukemic LGLs. (A) Sphingolipid metabolism and signaling pathway and its inhibitors. The inhibitors of the pathway are underlined. The genes shown in reversed color are core enriched as analyzed by GSEA. Naive normal PBMCs (N PBMCs, 1 representative sample of 2 independent experiments showed) or leukemic LGLs (TLGLs, n = 3) were either left untreated or treated with vehicle or indicated concentrations of myriocin (B), fumonisin (C), desipramine (D), and GW4869 (E) for 24 hours. Induction of apoptosis was assessed using flow cytometry. There was no differential apoptosis of leukemic LGLs compared with normal naive PBMCs using each of these inhibitors. (F) Inhibition of acid ceramidase induced differential apoptosis in leukemic LGLs. Naive normal PBMCs (N PBMCs, white dots, n = 5), activated normal PBMCs (AC PBMCs, gray dots, n = 6), or leukemic LGLs (TLGLs, black dots, n = 6) were either left untreated or treated with vehicle (methanol) or 100 μM NOE for 6 hours. Induction of apoptosis was assessed using flow cytometry. NOE induced approximately 30-fold higher apoptosis in LGL leukemia PBMCs compared with normal PBMCs (*P < .001). (G) Inhibition of S1P-mediated signaling by FTY720-induced differential apoptosis in leukemic LGLs. Naive normal PBMCs (N PBMCs, white dots, n = 5), activated normal PBMCs (AC PBMCs, gray dots, n = 4), or leukemic LGLs (TLGLs, black dots, n = 5) were either left untreated or treated with vehicle (DMSO) or 5 μM FTY720 for 6 hours. Leukemic LGLs showed approximately 13-fold higher apoptosis compared with naive normal PBMCs (*P < .001). (Each open dot [○] represents mean percentage of apoptosis ± SEM of 3 separate experiments in an individual sample; marker (•) represents the mean of all samples in a given treatment.)

Figure 7

S1P-mediated signaling plays an important role in survival of leukemic LGLs. (A) PBMCs from healthy donors (N PBMCs 1,2), normal activated PBMCs (AC PBMCs), or LGL leukemia patients (TLGLs 1,2) were treated with DMSO (□) or 5 μM FTY720 (■) as described. Flow cytometric analysis for induction of apoptosis was done. Cells were gated for CD3⁺CD57⁺ double-positive cells (LGL cells) and further analyzed for apoptosis. Leukemic LGLs showed 20-fold higher apoptosis compared with naive normal LGLs. (B) Treatment with FTY720 sensitizes LGL leukemia PBMCs to Fas-mediated apoptosis. PBMCs from LGL leukemia patients were incubated with vehicle or 5 μM FTY720 for 1 hour. Each treatment group was further divided into 2 and left untreated or treated with 1 μg/mL CH11 and further incubated for 6 hours. In addition to inducing spontaneous apoptosis, treatment with 5 μM FTY720 further sensitizes leukemic LGLs to Fas-mediated apoptosis (*P = .001). (Results shown are representative of 3 independent experiments.) (C) Activated PBMCs from healthy donors were cultured in RPMI-1640 supplemented with 1% FBS for 18 hours. The cells were treated with either vehicle or indicated concentrations of S1P for 1 hour. CH11 (1 μg/mL) was added to the wells and cells were further incubated for an additional 3.5 hours. The graph shows that increasing amounts of S1P in culture protects cells from Fas-mediated apoptosis in a dose-dependent manner. At 0.5-μM and 0.05-μM concentrations, S1P inhibits Fas-mediated apoptosis by more than 55% and 30% (*P < .03), respectively. The results shown are a representative of 1 of the 3 individual experiments performed.