Deregulated Expression of Fat and Muscle Genes in B-Cell Chronic Lymphocytic Leukemia with High Lipoprotein Lipase Expression

Deregulated expression of fat and muscle genes in B-cell chronic lymphocytic leukemia with high lipoprotein lipase expression

M Bilban1,2,8, D Heintel3,8, T Scharl4, T Woelfel4, MM Auer2,3, E Porpaczy3, B Kainz3, A Kro¨ber5, VJ Carey6, M Shehata2,3, C Zielinski2, W Pickl7, S Stilgenbauer5, A Gaiger2,3,6, O Wagner1,2,UJa¨ger2,3 and the German CLL Study Group

1Department of Laboratory Medicine, Medical University of Vienna, Vienna, Austria; 2Ludwig Boltzmann Institute for Clinical and Experimental Oncology, Vienna, Austria; 3Department of Internal Medicine I, Division of Hematology and Hemostaseology, Medical University of Vienna; Vienna, Austria; 4Department of Statistics and Probability Theory, Vienna University of Technology, Vienna, Austria; 5Department of Internal Medicine III, University of Ulm, Ulm, Germany; 6Department of Medicine, Harvard Medical School, Boston, MA, USA and 7Institute of Immunology, Medical University of Vienna, Vienna, Austria

Lipoprotein lipase (LPL) is a prognostic marker in B-cell studies was the identification of a novel prognostic marker, the chronic lymphocytic leukemia (B-CLL) related to immunoglo- ZAP-70 protein, which has already entered routine diagnos- bulin VH gene (IgVH)mutational status. We determined gene tics.3,4,9,26–31 However, microarray analysis has identified a expression profiles using Affymetrix U133A GeneChips in two groups of B-CLLs selected for either high (‘LPL þ ’, n ¼ 10) or number of other potential prognostic or therapeutic targets that low (‘LPLÀ’, n ¼ 10) LPL mRNA expression. Selected genes have not yet been validated for their clinical importance. were verified by real-time PCR in an extended patient cohort One of the genes most closely related to IgVH mutation status (n ¼ 42). A total of 111 genes discriminated LPL þ from LPLÀ B- is lipoprotein lipase (LPL). High expression of this central CLLs. Of these, the top three genes associated with time to first enzyme of lipid metabolism was associated with unmutated B- treatment were Septin10, DMD and Gravin (Pp0.01). The CLL in almost all profiling studies, regardless of methodology or relationship of LPL þ and LPLÀ B-CLL gene expression 1,3,4,32,33 signatures to 52 tissues was statistically analyzed. The LPL þ B-cell selection process. We and others have recently B-CLL expression signature, represented by 64 genes was shown that mRNA expression of LPL predicts treatment-free as significantly related to fat, muscle and PB dendritic cells well as overall survival in B-CLL patients.32,33 High intracellular (Po0.001). Exploration of microarray data to define functional levels of LPL protein can be detected in unmutated CLL cells.32 alterations related to the biology of LPL þ CLL identified two Lipoprotein lipase is normally produced and secreted by functional modules, fatty acid degradation and MTA3 signaling, adipocytes, macrophages, cardiac and skeletal muscle cells as being altered with higher statistical significance. Our data and catalyzes the hydrolysis of the triacylglycerol component of show that LPL þ B-CLL cells have not only acquired gene 34–36 expression changes in fat and muscle-associated genes but chylomicrons and very low-density lipoproteins (VLDL). also in functional pathways related to fatty acid degradation The protein is bound to the surface of endothelial cells as well as and signaling which may ultimately influence CLL biology and B-CLL cells.37 Additional non-catalytic binding functions lead to clinical outcome. increased accumulation and cellular uptake of lipoproteins.35 Leukemia (2006) 20, 1080–1088. doi:10.1038/sj.leu.2404220; Interestingly, LPL is regulated by cytokines including tumor published online 13 April 2006 necrosis factor-alpha (TNF-a) which also plays an important role Keywords: B-CLL; lipoprotein lipase; gene expression profiling; 38–40 prognostic markers; septin10; dystrophin in the pathophysiology of B-CLL. Although the function of LPL in B-CLL is still not understood, its effects on metabolism, energy homeostasis and microenvironment suggest a considerable biological relevance for the disease. We hypothesized that microarray analysis of two subsets of B- Introduction CLLs differing greatly in their LPL expression should reveal gene patterns not only associated with prognosis but also with Gene-expression profiling has introduced a new dimension into biology, particularly in regard to fatty acid metabolism and our understanding of biology as well as clinical behavior of adipose tissue. B-cell chronic lymphocytic leukemia (B-CLL).1–10 Initial studies Using this single gene denominator approach we were able to showed that all B-CLL cases have common features resembling identify known as well as novel genes related to IgVH mutational those of a memory B cell.1,3,11,12 Subsequently, considerable status, lipid metabolism, or both. We show that the expression differences in gene expression were linked to various discri- signature of unmutated/LPL-high B-CLL is closely related to that minators between good and poor risk B-CLL: these in- of adipose tissue and muscle cells. Moreover, we determined clude cytogenetic aberrations,13,14 expression of the CD38 the prognostic power of the most prominent factors by real-time protein,15,16 the expression of microRNAs,17,18 and most PCR using RNA from a well-defined patient cohort. These importantly, immunoglobulin VH gene (IgVH) mutational sta- include novel genes as well as several genes, previously tus.15,16,19–25 The first practical result emerging from these identified by expression profiling in the context of mutational status, that have never been validated. Besides the intriguing Correspondence: Professor U Ja¨ger, Department of Internal Medicine I, biological questions regarding the origin, regulation, or plasti- Division of Hematology and Hemostaseology, Medical University of city of B-CLL cells raised in this study, we now provide evidence Vienna, Wa¨hringer Gu¨rtel 18-20, A-1090 Vienna, Austria. for the clinical importance of some of the discriminatory genes E-mail: [email protected] 8These authors contributed equally to this work. commonly found in expression profiling analysis. Received 8 September 2005; revised 21 February 2006; accepted 27 Using this approach it was possible to stratify CLL subgroups February 2006; published online 13 April 2006 by LPL-expression and to recover prognostic factors known from LPL expression profiling in B-CLL M Bilban et al 1081 the studies using IgVH mutational status as well as novel targets from the Weizmann Institute of Science (http://bioinfo.weiz- in high-risk B-CLL. mann.ac.il/bioinfo.html), and a previously published classifica- tion scheme for biological functions.44 Hierarchical clustering was performed using Pearson’s correlation coefficient as Patients, materials and methods distance measure and Ward’s optimization criterion to cluster genes and samples, respectively. To visualize relationships and cRNA synthesis and gene expression profiling to judge the clustering quality, we presented samples and genes Total RNA from peripheral mononuclear cells (PBMC) of 20 in two-dimensional principal component space of the corre- well-characterized B-CLL patients (diagnosed at the Division of sponding variance-covariance matrix. Hematology at the Medical University of Vienna) was isolated Publicly available Affymetrix U133A expression data for 32 as described. Patient characteristics are given in Table 1. 52 tissues were used from SymAtlas (http://symatlas.gnf.org/ Informed consent was obtained from all patients according to SymAtlas/) for a phenotypic analysis. GeneChip files were study protocols approved by the local ethics committee (EK nos. preprocessed as described above for CLL samples. 38/1998, 505/2002, 495/2003). Total RNA was repurified with Gene set enrichment analysis (GSEA)45 is a computational RNeasy MinElute kit as per the manufacturer’s instructions method that determines if a given set of genes (e.g. known (Qiagen Valencia, CA, USA) and checked for integrity pathways, specific areas of the genome or clusters from a cluster by agarose gel electrophoresis. A measure of 5 mg total RNA analysis) shows statistically significant differences between two was then used for GeneChip analysis. Preparation of phenotypic states (i.e. LPL þ or LPLÀ CLLs). Briefly, the GSEA cRNA, hybridization to human U133A GeneChips (Affymetrix, calculation involves three steps: calculation of an enrichment Santa Clara, CA, USA) and scanning of the arrays were carried score (ES) followed by estimation of the significance level of ES out according to the manufacturer’s protocols (https://www. and adjustment for Multiple Hypothesis Testing. All GSEA 41 affymetrix.com; Bilban et al. ). calculations were done in ‘R’.

Bioinformatic analysis RNA signal extraction, normalization and filtering (to eliminate Real-time PCR genes with extremely low expression) was performed as The significance of selected genes was validated by real-time described42 (http://www.bioconductor.org/). We identified sig- PCR in 42 B-CLL PBMNC patient samples (20 mutated, 22 nificant differences between sample groups using a previously unmutated, 25 Binet A, 7 Binet B, 10 Binet C; 17 female, 25 described method:43 we calculated a t-statistic of the two groups male). Total RNA was extracted from CLL and PBM cells and of CLLs for each gene. We addressed the multiple comparisons analyzed by RT-PCR for LPL mRNA expression as described problem by estimating the false discovery rate (FDR) in a simple previously32 using RNA-Beet (Tel-Test Inc., Friendswood, TX, manner as the ratio of the expected number of false positives at a USA) and first strand cDNA synthesis kit (Amersham Pharmacia given P-value threshold to the number of positives actually Biotech, Inc., Piscataway, NJ, USA). Real-time PCR was found. Using this statistical approach comparisons of gene performed with the ABI Prism 7000 Sequence Detector (Applied expression with absolute fold changes of at least 1.5-fold Biosystems, Foster City, CA, USA) according to the manufac- (increase or decrease) were selected at qo0.01. Genes were turer’s instructions. Expression profile of the housekeeping gene annotated based on the putative biological functions of the b-actin showed low variation of DCt values (VICt-labeled pre- encoded proteins, via database searches on PubMed, gene cards developed TaqMans Assay reagent) which was subsequently

Table 1 Patients with low (2–20) and high (35–55) LPL mRNA expression

Patient no. Array ID LPL mRNA expression Clinical stage VH homology (%) VH gene Cytogenetic aberrations

2 L_M_8 0.10 C/IV 93.9 VH1-08 13qÀ 5 L_M_10 0.32 A/II 94.6 VH3-15 Normal 6 L_M_6 0.36 C/IV 97.3 VH3-23 13qÀ 8 L_M_1 0.65 A/0 92.3 VH3-33 14q32, 13qÀ 9 L_M_13 0.75 A/0 90.5 VH4-04 Normal a 11 L_UM_12 1.00 A/I 100 VH3-11 14q32, +12, p53À 12 L_M_9 1.09 C/IV 89.1 VH3-74 +12 15 L_M_3 1.59 A/II 97.9 VH4-34 13qÀ 16 L_M_4 1.65 A/0 93.2 VH3-74 13qÀ 20 L_M_6n 2.47 A/0 93.9 VH3-15 13qÀ 55 H_UM_1 1924.14 A/0 100 VH1-02 13qÀ 54 H_UM_12 1300.63 B/I 99 VH3-8 11qÀ, 13qÀ 53 H_UM_4 858.10 A/0 100 VH1-69 11qÀ 52 H_UM_5 432.03 A/0 100 VH3-30 13qÀ 51 H_UM_11 338.97 B/II 100 VH1-69 11qÀ 49 H_UM_10 200.16 A/0 99.7 VH1-69 14q32À 48 H_UM_9 146.52 B/II 100 VH1-02 13qÀ 46 H_UM_3 141.04 B/II 100 VH3-11 +12, 11qÀ 42 H_UM_2 100.43 A/I 100 VH4-34 11qÀ, 13qÀ a 35 H_M_8 43.41 C/III 96.8 VH3-21 14q32, +12, 13qÀ aDiscordant cases are indicated by asterisks. 13qÀ, deletion in 13q; normal, normal karyotype; 14q32, abnormalities of chromosome bands 14q32; +12 trisomy 12; 11qÀ, deletion in 11q; p53À,loss of p53.

Leukemia LPL expression proﬁling in B-CLL M Bilban et al 1082 used as endogenous control. Polymerase chain reaction (PCR) was carried out in a 25 ml reaction volume using 1 ml of a 1:10 cDNA dilution with gene-speciﬁc primers designed by Applied Biosystem. All samples were run in duplicates.

Statistical analysis Correlation of LPL expression with other genes (Spearman’s correlation) as well as survival analysis was performed using WinStad (Version 3.0).

Results

LPL-associated genes We determined the distinct gene expression signature associated with LPL mRNA expression in B-CLL applying large-scale gene expression profiles using Affymetrix U133A GeneChips inter- rogating 420 000 genes in two groups of unsorted B-CLLs selected for either high (‘LPL þ ’; n ¼ 10, median ¼ 269.7) or low (‘LPLÀ’; n ¼ 10, median ¼ 0.9) LPL mRNA expression (Table 1). A fold change of X1.5/pÀ1.5-fold and a P-value of o0.01 were used as selection criteria in order to detect genes that were altered to a probably biologically significant extent with reasonable confidence. The FDR, that is the probabilities to detect a gene as significantly changed that is not changed in reality, was approximately 2.9% for the comparison LPL þ vs LPLÀ in our experiment yielding 111 genes differentially expressed between LPL þ and LPLÀ (Figure 1). Of these 111 genes, 67 were significantly increased in the LPL þ group, whereas the expression of 44 genes was decreased in this group as compared to LPLÀ CLLs (Supplementary Table 1A and B). Principal component analysis as well as hierachical clustering clearly demonstrated that the expression of these genes distinguishes LPL þ from LPLÀ samples forming an LPL-specific gene expression signature (Figure 1a and b).

Validation of single targets In order to validate this general pattern a selection of important single target genes was tested. A panel of 12 genes was selected for real-time PCR validation on the basis of novelty or their appearance in previous reports indicating their general importance as potential prognostic factors. To this end, mRNA from a group of 42 well-characterized B-CLL patients was analyzed (Materials and methods). High expression of LPL was Figure 1 LPL-associated gene expression profile in CLL. (a) Principal component analysis plot using the LPL þ gene expression signature significantly associated with high mRNA levels of DMD, (n ¼ 111, qo0.01). (b) Hierachical clustering of 111 genes discri- Septin10, PNMA2, SGCE, AKAP12 and PEG10 (Table 2). There minates between ‘LPL þ ’ and ‘LPLÀ‘ CLL samples. Rows and columns was a significant inverse association of LPL with ZNF288, FGL2 represent genes and samples, respectively. The full list of genes is and SORL1. available in Supplementary Table 1. Color scale indicates units of To support the significance of our findings using unsorted standard deviation (s.d.) from the mean expression of each gene. CLLs, we assessed differential expression of these 12 genes in an unrelated microarray data set using CD19 þ sorted CLLs (Shehata M et al., manuscript in preparation; Table 2). As Comparison with IgVH mutational status expexted, sorted CD19 þ CLL cells showed 25-fold higher We re-analyzed the CLL samples according to IgVH mutational expression for the well-characterized ZAP70 marker gene in status. Similar statistical stringency was applied to allow for LPL þ CLL patient cells as compared with LPLÀ CLLs (Table 2). comparison with the LPL discriminatory gene expression More importantly, differential expression of the 12 real-time signature. Applying almost identical FDRs (3.1 as opposed to PCR validated genes was even more pronounced in the CD19 þ 2.9% for LPL þ vs LPLÀ) we identified 11 genes regulated more sorted CLLs (Table 2). Furthermore, peripheral blood monocytes than 1.5-fold (Supplementary Table 2). Of note, the top and T cells altogether express LPL þ genes at fourfold lower discriminatory genes associated with LPL expression or IgVH levels or less than unsorted CLL cells (Table 2). These results mutational status were quite similar. Among the top discriminat- confirm the involvement of several fat- and muscle-related ing genes we identified several candidates that had previously genes. been described in B-CLL, including dystrophin, gravin

Leukemia LPL expression proﬁling in B-CLL M Bilban et al 1083 Table 2 Real-time PCR validation of selected genes differentially expressed between LPL+ and LPLÀ CLLs

Gene name Symbol GenBank ID LPL+/ LPL+/LPLÀ RT-PCRc P-valued Re LPL+/ LPL+/ LPL+/ LPL+/ LPLÀ (unsorted)b CD14f CD8g CD4h WBi (sorted)a

Lipoprotein lipase LPL BF672975 97.4 8.0 NA 9.8 10.9 9.2 9.0 Septin 10 Sept10 BF966021 340.6 7.5 33.4 0.00003 0.59 3.8 5.8 6.3 5.0 Dystrophin DMD NM_004010 46.9 5.1 4.4 0.00002 0.59 11.3 12.4 13.5 13.0 Paraneoplastic antigen MA2 PNMA2 AB020690 1.8 4.3 2.4 0.02 0.32 7.5 8.3 8.2 8.0 Sarcoglycan, epsilon SGCE NM_003919 54.1 2.3 2.4 0.0004 0.5 8.0 8.2 8.3 7.0 Gravin AKAP12 AB003476 27.8 2.4 1.6 0.007 0.38 6.4 6.0 6.1 6.2 Paternally expressed 10 PEG10 BE858180 18.0 2.2 2.4 0.001 0.46 7.1 8.6 9.0 8.3 Seven in absentia homolog 1 SIAH1 NM_003031 1.3 2.0 1.2 0.16 0.16 11.7 8.2 8.4 5.3 Carnitine palmitoyltransferase 11 CPT1A NM_001876 3.4 1.8 1.5 0.02 0.42 1.5 2.4 2.7 1.9 Dual specificity phosphatase 1 DUSP1 NM_004417 À1.9 À4.1 À2 0.4 0.04 0.1 0.2 0.2 0.3 Fibrinogen-like 2 FGL2 NM_006682 À2.3 À3.1 À2.8 0.04 À0.3 0.04 2.3 2.2 0.1 Sortilin-related receptor SORL1 AV728268 À2.4 À4.4 À2.1 0.037 À0.3 0.4 0.2 0.2 0.1 Zinc-finger protein 288 ZNF288 NM_015642 À4.9 À2.3 À4.3 0.01 À0.3 4.4 1.8 1.9 3.0 ZAP70 ZAP70 25.1 1.4 n.d. 2.7 0.1 0.2 0.9 aFold changes of LPL+/LPLÀ of CD19+ (n ¼ 3 or 7 for LPL+ or LPLÀ, respectively). bFold changes of LPL+/LPLÀ of unsorted CLLs (n ¼ 10 for LPL+ or LPLÀ, respectively) as determined by GeneChip analysis. cReal-time PCR-derived mean ratio of 42 CLL samples expressing either high or low levels of LPL mRNA dt-statistic of 42 CLL cells. eCorrelation coefficient for the relationship of LPL expression and these 12 tabulated genes fRatio of mean signals for LPL+ CLL vs peripheral CD14 monocytes. gRatio of mean signals for LPL+ CLL vs peripheral CD8 T-cells. hRatio of mean signals for LPL+ CLL vs peripheral CD4 T-cells. iRatio of mean signals for LPL+ CLL vs peripheral whole blood.

a associated with LPL only or IgVH mutational status only are indicated in Supplementary Table 2. 1,0 Septin10

0,8 Novel genes and clinical outcome 0,6 Real-time PCR data were analyzed for correlation with clinical outcome by Kaplan–Meier analysis using the median mRNA Low 0,4 expression level as cutoff. Six out of 13 genes were signiﬁcantly associated with time to ﬁrst treatment (TT): the best two were 0,2 High Septin10 and DMD (Figure 2, Pp0.01), followed by AKAP12, % Untreated Patients PEG10, SGCE and CPTA1 (Po0.05). Patients with high 0,0 expression of Septin10, DMD, AKAP12, PEG10, SGCE or CPTA1 required earlier treatment (30 vs 80 months for high vs 0 50 100 150 low DMD expression; 34 months in patients with high Septin 10 Months expression – the median TT was not reached in the low Septin10 b group; Figure 2). 1,0 DMD

0,8 Functional modules associated with LPL þ expression LPL þ B-CLLs are significantly enriched for fat-asso- 0,6 ciated genes. After having established the LPL-associated gene expression in B-CLL cells, we wished to determine in 0,4 Low which tissues this particular pattern was also represented. To this end we compared the expression of 64 LPL þ signature genes in 0,2 52 tissues. Sixty-four genes were selected based on their overall % Untreated Patients expression level in these tissues following normalization and High 0,0 filtering (Supplementary Table 3; see Supplementary material for detailed method). Using a Mann–Whitney rank sum test based 0 50 100 150 upon the ranks of these 64 genes, fat had the most discriminating Months P-value (P ¼ 4.9 Â 10À6) followed by skeletal muscle, heart and À6 À5 Figure 2 Time-to-treatment analysis. Rate of disease progression as peripheral blood dendritic cells (P ¼ 9.8 Â 10 , 4.6 Â 10 and assessed by the treatment-free time interval measured in months from 7.2 Â 10À4, respectively; Figure 3) as the most significantly diagnosis for high (4median) or low (omedian) Septin10 and DMD associated tissues. This indicates a potential biological relation- mRNA expression. These curves refer to the 42 cases. ship of LPL þ (unmutated) B-CLL cells to these tissues in particular to fat and muscle. (AKAP12), Septin10, BCL7A, FGL2, PNMA2, SORL1 or ZNF288 (Supplementary Table 2). Other genes included PEG10, DUSP1, LPL þ B-CLLs are significantly enriched for genes CLIC4, SIAH1, CPTA1, or sarcoglycan epsilon (SGCE). Genes regulating fatty acid degradation and cell plasticity.To

Leukemia LPL expression proﬁling in B-CLL M Bilban et al 1084 Adrenal gland Appendix Atrioventricular node B-Lymphoblasts BM CD105 Endothelial BM CD33 Myeloid BM CD34 BM CD71 Early Erythroid Cardiac Myocytes Cerebellum Ciliary ganglion Cingulate Cortex Colorectal Adenocarcinoma DRG Fetal brain Fetal liver Fetal thyroid Fat HBEC Heart Hypothalamus Islet kidney Liver Lung lymphnode Lymphoma Raji Occipital Lobe Ovary Pancreas Parietal Lobe PB BD CA4 Dentritic Cells PB CD14 Monocytes PB CD19B Cells PB CD4T Cells PB CD8T Cells Pons Prefrontal Cortex Prostate Skeletal Muscle Skin Smooth Muscle Spinal cord Testis Thymus Thyroid Tongue Tonsil Trachea Uterus Uterus Corpus Whole Brain 1 0.1 0.01 0.001 0.0001 0.00001 0.000001 P-Value

Figure 3 Phenotypic analysis of LPL þ B-CLLs. The expression of LPL þ signature genes was tested in 52 normal tissues. Log(base 10) P-values of a Mann–Whitney rank sum test based on the ranks of 64 genes are plotted for each of the 52 tissues.

determine whether any a priori defined gene sets are enriched prognostic value. Additional bioinformatics analysis identified within the LPL þ and LPLÀ microarray data we applied GSEA two functional modules regulating fatty acid degradation and using 1325 gene sets contained within the database MsigDB.45 MTA3 signaling indicating changes in metabolism and signal Gene set enrichment analysis identified two altered pathways transduction. with high statistical significance (Po0.01): the process of fatty acid degradation and MTA3 signaling (Figure 4). Additional five pathways showing significantly altered gene expression Discriminatory genes between LPL þ and LPLÀ CLLs are given in Supplementary A significant enrichment of genes regulating fat metabolism and Table 4. muscle development as well as immune and defense response was observed. As LPL is associated with IgVH mutational status it is not unexpected that several of the genes described in previous Discussion reports investigating mutated and unmutated B-CLL were also discriminating between LPL þ and LPLÀ patients1,3,4 (Table 2). We have determined the gene expression signature of B-CLL This was also confirmed by our own analysis based on cells associated with either high or low LPL mRNA expression. mutational status in which 9 out of 11 discriminatory genes We describe a gene expression signature related to fat and also appeared in the LPL screen (Supplementary Table 2). The muscle in LPL þ B-CLL cells that contains genes of potential top discriminatory genes included genes overexpressed in

Leukemia LPL expression proﬁling in B-CLL M Bilban et al 1085

Figure 4 Functional modules enriched in LPL þ CLL. Lipid metabolism (a, b; P ¼ 0.00965) and MTA3 signaling (c, d; P ¼ 0.00881) pathways are signiﬁcantly altered for the comparison of LPL þ vs LPLÀ CLL. (a and c) GSEA plots for the fatty acid degradation and MTA3 pathway, respectively. Shown is a plot of the running sum vs the gene list rank (detailed description of the GSEA algorithm is given in Subramanian et al., 2005). (b and d) Corresponding heatmap showing relative expression of the fatty acid degradation pathway gene members across the LPL þ and LPLÀ CLLs. Color bar indicates units of standard deviation from the mean expression.

unmutated B-CLLs like Septin10 and DMD as well as genes malignancies (data not shown). The role of septins in LPL þ repressed in these cases like Znf288 (ZBTB20). Novel genes CLLs seems intriguing: Septins are an evolutionarily conserved discovered by the LPL screen included DUSP1, CLIC4, SIAH1, group of at least 12 distinct genes encoding GTP-binding and CPTA1, PEG10 and SGCE. Importantly, validation of our results filament-forming proteins implicated in cellular polarity deter- in a set of 10 CD19 þ sorted CLL samples confirmed our mination, cytoskeletal reorganization, membrane dynamics, findings. vesicle trafficking, and exocytosis.47 The (actin) cytoskeleton organizes surface proteoglycans that bind LPL.48–50 Therefore, Septin10 could direct LPL to the cell surface by establishing a Clinical value link between proteoglycans (the cell surface LPL receptor) and Only a few of the genes discovered by microarray studies have the actin cytoskeleton.47 Several septins have altered expression been validated for clinical use by real-time PCR (ZAP70,9,46 in lymphoid, breast, skin, ovary, endometrium, kidney, lung, LPL,32,33 and ADAM2933). We and others have recently shown liver, brain, pancreas, colorectal and urological tumors.47,51,52 that assessment of LPL constitutes an excellent alternative for Involvement of the cytoskeleton is further underlined by the fact ZAP70 measurement.32,33 Here, we have investigated 12 genes that DMD was found in most previous CLL microarray associated with LPL for their diagnostic and prognostic power studies1,3,4,53 and proved to be of prognostic value in our (Table 2, Figure 2). In general, microarray and real-time PCR patients. data correlated extremely well although real-time PCR was extended to include 22 more patients (n ¼ 42 in contrast to n ¼ 20 for microarray analysis). We speculate that variability in Biological significance gene expression from these additional patients resulted in a lack Our non-supervized hierarchical clustering analysis was able to of significant association for DUSP1, CLIC4 and SIAH1 with separate the LPL þ from the LPLÀ group (suggesting that LPL þ LPL þ gene expression signature. Interestingly, Septin10 and and LPLÀ CLLs constitute two variants of the same disease). We DMD turned out to be the best clinical markers in time-to- reasoned that combined measurement of subtle (o50%) but treatment analysis. Furthermore, the expression of both genes coordinated changes across multiple members of a pathway maintained the same prognostic impact also in Binet stage A could help explain the striking differences in clinical outcomes CLL of this cohort for this analysis (data not shown). This is of these two variants. Such ‘gene set enrichement analysis’ underlined by their highly restricted expression in B-cell allows for the discovery of gene sets that might function in an

Leukemia LPL expression profiling in B-CLL M Bilban et al 1086 orchestrated manner that would not be revealed otherwise.45,54 References The two top-scoring pathways identified suggest metabolic (fatty acid degradation) as well as cell plasticity (MTA3 signaling) 1 Rosenwald A, Alizadeh AA, Widhopf G, Simon R, Davis RE, Yu X related adaptations of LPL þ CLL (Figure 4 and Supplemental et al. Relation of gene expression phenotype to immunoglobulin Table 4). Further evidence for extensive reprogramming of gene mutation genotype in B cell chronic lymphocytic leukemia. J Exp expression in LPL þ CLL comes from analysis of the LPL þ Med 2001; 194: 1639–1647. 2 Falt S, Merup M, Tobin G, Thunberg U, Gahrton G, Rosenquist R signature in 52 tissues: we reasoned that a tissue’s phenotype is et al. Distinctive gene expression pattern in VH3-21 utilizing B-cell related to the LPL þ phenotype if the expression levels of the chronic lymphocytic leukemia. Blood 2005; 106: 681–689. LPL þ signature genes in the (potentially related) tissue 3 Klein U, Tu Y, Stolovitzky GA, Mattioli M, Cattoretti G, Husson H significantly differ from the mean expression over all tissues.55,56 et al. Gene expression profiling of B cell chronic lymphocytic Based on the LPL þ signature analyzed in 52 tissues, we leukemia reveals a homogeneous phenotype related to memory B demonstrate a significant relationship to genes specifically cells. J Exp Med 2001; 194: 1625–1638. 4 Haslinger C, Schweifer N, Stilgenbauer S, Dohner H, Lichter P, regulated in fat and muscle (Figure 3). Kraut N et al. Microarray gene expression profiling of B-cell Several explanations for this phenomenon may be envisioned: chronic lymphocytic leukemia subgroups defined by genomic (i) Existence of a common master switch regulating the LPL þ aberrations and VH mutation status. J Clin Oncol 2004; 22: signature on the transcriptional level. We are currently 3937–3949. analyzing the promoter regions of a number of genes involved. 5 Schroers R, Griesinger F, Trumper L, Haase D, Kulle B, Klein- (ii) Selection of the malignant clone on the basis of the Hitpass L et al. Combined analysis of ZAP-70 and CD38 expression as a predictor of disease progression in B-cell chronic lymphocytic biological function exerted by fat and muscle genes, that is fat 37 leukemia. Leukemia 2005; 19: 750–758. metabolism (LPL, CPTA1), cell structure and motility (Sep- 6 Ferrer A, Ollila J, Tobin G, Nagy B, Thunberg U, Aalto Y et al. tin10,47 DMD, SGCE). Higher activity of the fatty acid Different gene expression in immunoglobulin-mutated and im- degradation pathway may point towards augmented energy munoglobulin-unmutated forms of chronic lymphocytic leukemia. consumption of LPL þ CLLs. Differential activity of lipid Cancer Genet Cytogenet 2004; 153: 69–72. metabolism has recently been described for adenocarcinomas 7 Gaiger A, Heintel D, Jager U. Novel molecular diagnostic and therapeutic targets in chronic lymphocytic leukaemia. Eur J Clin of the stomach.54 It is well recognized that the cytoskeleton is 57 Invest 2004; 34 (Suppl 2): 25–30. involved in cell transformation. (iii) Different stages of 8 Davis RE, Staudt LM. Molecular diagnosis of lymphoid malig- differentiation at the moment of malignant transformation or a nancies by gene expression profiling. Curr Opin Hematol 2002; 9: certain degree of plasticity or cell reprogramming.58 The MTA3 333–338. pathway regulates epithelial to mesenchymal transition in breast 9 Wiestner A, Rosenwald A, Barry TS, Wright G, Davis RE, cancer and the (de)differentiation state of B cells.59,60 We Henrickson SE et al. ZAP-70 expression identifies a chronic lymphocytic leukemia subtype with unmutated immunoglobulin speculate that higher activity of the MTA3 pathway in the genes, inferior clinical outcome, and distinct gene expression unmutated, LPL þ CLL cells may cause them to ‘dedifferentiate’ profile. Blood 2003; 101: 4944–4951. towards a more naive germinal B-cell phenotype. 10 Vasconcelos Y, De VJ, Vallat L, Reme T, Lalanne AI, Wanherdrick We also noted an association with dendritic cells. Throughout K et al. Gene expression profiling of chronic lymphocytic leukemia dendritic cell maturation, a portion of LPL þ signature genes can discriminate cases with stable disease and mutated Ig genes was downregulated with LPL being upregulated more than 50- from those with progressive disease and unmutated Ig genes. Leukemia 2005; 19: 2002–2005. fold (Bilban et al., manuscript in preparation). B-cell chronic 11 Chiorazzi N, Rai KR, Ferrarini M. Chronic lymphocytic leukemia. lymphocytic leukemia cells are among the most potent antigen- N Engl J Med 2005; 352: 804–815. presenting cells as recently demonstrated for Rh-autoantigen 12 Stevenson FK, Caligaris-Cappio F. Chronic lymphocytic leukemia: presentation.61 In this context it is interesting that several of the revelations from the B-cell receptor. Blood 2004; 103: 4389–4395. discriminatory genes are known as autoantigens including 13 Dohner H, Stilgenbauer S, Benner A, Leupolt E, Krober A, LPL,62 AKAP1263 and Septin10.47 Bullinger L et al. Genomic aberrations and survival in chronic lymphocytic leukemia. N Engl J Med 2000; 343: 1910–1916. In conclusion we show a distinct molecular pattern for LPL þ 14 Dohner H, Stilgenbauer S, James MR, Benner A, Weilguni T, Bentz and LPLÀ B-CLL subtypes and have identified and validated a M et al. 11q deletions identify a new subset of B-cell chronic number of potential target genes for diagnosis and treatment. In lymphocytic leukemia characterized by extensive nodal involve- addition, this study raises intriguing questions regarding the ment and inferior prognosis. Blood 1997; 89: 2516–2522. origin and biology of B-CLL. 15 Hamblin TJ, Orchard JA, Ibbotson RE, Davis Z, Thomas PW, Stevenson FK et al. CD38 expression and immunoglobulin variable region mutations are independent prognostic variables in chronic Acknowledgements lymphocytic leukemia, but CD38 expression may vary during the course of the disease. Blood 2002; 99: 1023–1029. We are greatly indebted to G.Kostner and R. Zechner for helpful 16 Damle RN, Wasil T, Fais F, Ghiotto F, Valetto A, Allen SL et al. Ig V discussions. This study is supported by grants from the Austrian gene mutation status and CD38 expression as novel prognostic National Bank (Grant no. 9964), the Austrian Human Genome indicators in chronic lymphocytic leukemia. Blood 1999; 94: 1840–1847. Project (‘C.h.i.l.d’) (UJ), the Center of Molecular Medicine 17 Calin GA, Liu CG, Sevignani C, Ferracin M, Felli N, Dumitru CD (CeMM) of the Austrian Academy of Sciences (no. 20010 to UJ et al. MicroRNA profiling reveals distinct signatures in B cell and no. 20030 to WP), Corixa Co., Seattle (AG), DFG (STI 296/1), chronic lymphocytic leukemias. Proc Natl Acad Sci USA 2004; Sander (2001.04.02) and Fresenius (CLL4) (DK, AK, HD, SS). 101: 11755–11760. 18 Chen CZ, Lodish HF. MicroRNAs as regulators of mammalian Note added in proof: hematopoiesis. Semin Immunol 2005; 17: 155–165. 19 Hamblin TJ, Davis Z, Gardiner A, Oscier DG, Stevenson FK. In a recently published report,64 the authors confirm our Unmutated Ig V(H) genes are associated with a more aggressive 32 form of chronic lymphocytic leukemia. Blood 1999; 94: previous data on LPL as a predictive factor in B-CLL. In 1848–1854. addition, the paper shows that Septin10 and DMD are also 20 Oscier DG, Gardiner AC, Mould SJ, Glide S, Davis ZA, Ibbotson predictors of survival and that unpurified mononuclear cells RE et al. Multivariate analysis of prognostic factors in CLL: clinical from B-CLL samples can be used for real-time PCR in this setting. stage, IGVH gene mutational status, and loss or mutation of the

Leukemia LPL expression profiling in B-CLL M Bilban et al 1087 p53 gene are independent prognostic factors. Blood 2002; 100: evidence that TNF-alpha is involved. Am J Physiol Endocrinol 1177–1184. Metab 2004; 286: E711–E717. 21 Krober A, Seiler T, Benner A, Bullinger L, Bruckle E, Lichter P et al. 40 Ziouzenkova O, Asatryan L, Sahady D, Orasanu G, Perrey S, Cutak V(H) mutation status, CD38 expression level, genomic aberrations, B et al. Dual roles for lipolysis and oxidation in peroxisome and survival in chronic lymphocytic leukemia. Blood 2002; 100: proliferation-activator receptor responses to electronegative low 1410–1416. density lipoprotein. J Biol Chem 2003; 278: 39874–39881. 22 Vasconcelos Y, Davi F, Levy V, Oppezzo P, Magnac C, Michel A 41 Bilban M, Ghaffari-Tabrizi N, Hintermann E, Bauer S, Molzer S, et al. Binet’s staging system and VH genes are independent but Zoratti C et al. Kisspeptin-10, a KiSS-1/metastin-derived decapep- complementary prognostic indicators in chronic lymphocytic tide, is a physiological invasion inhibitor of primary human leukemia. J Clin Oncol 2003; 21: 3928–3932. trophoblasts. J Cell Sci 2004; 117 (Part 8): 1319–1328. 23 Thunberg U, Johnson A, Roos G, Thorn I, Tobin G, Sallstrom J et al. 42 Irizarry RA, Bolstad BM, Collin F, Cope LM, Hobbs B, Speed TP. CD38 expression is a poor predictor for VH gene mutational status Summaries of Affymetrix GeneChip probe level data. Nucleic and prognosis in chronic lymphocytic leukemia. Blood 2001; 97: Acids Res 2003; 31: e15. 1892–1894. 43 Tusher VG, Tibshirani R, Chu G. Significance analysis of 24 Stilgenbauer S, Bullinger L, Lichter P, Dohner H. Genetics of microarrays applied to the ionizing radiation response. Proc Natl chronic lymphocytic leukemia: genomic aberrations and V(H) Acad Sci USA 2001; 98: 5116–5121. gene mutation status in pathogenesis and clinical course. 44 Ashburner M, Ball CA, Blake JA, Botstein D, Butler H, Cherry JM Leukemia 2002; 16: 993–1007. et al. Gene ontology: tool for the unification of biology. The Gene 25 Guarini A, Gaidano G, Mauro FR, Capello D, Mancini F, De Ontology Consortium. Nat Genet 2000; 25: 25–29. Propris MS et al. Chronic lymphocytic leukemia patients with 45 Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, highly stable and indolent disease show distinctive phenotypic and Gillette MA et al. Gene set enrichment analysis: a knowledge- genotypic features. Blood 2003; 102: 1035–1041. based approach for interpreting genome-wide expression profiles. 26 Chen L, Widhopf G, Huynh L, Rassenti L, Rai KR, Weiss A et al. Proc Natl Acad Sci USA 2005; 102: 15545–15550. Expression of ZAP-70 is associated with increased B-cell receptor 46 Kienle DL, Korz C, Hosch B, Benner A, Mertens D, Habermann A signaling in chronic lymphocytic leukemia. Blood 2002; 100: et al. Evidence for distinct pathomechanisms in genetic subgroups 4609–4614. of chronic lymphocytic leukemia revealed by quantitative expres- 27 Crespo M, Bosch F, Villamor N, Bellosillo B, Colomer D, Rozman sion analysis of cell cycle, activation, and apoptosis-associated M et al. ZAP-70 expression as a surrogate for immunoglobulin- genes. J Clin Oncol 2005; 23: 3780–3792. variable-region mutations in chronic lymphocytic leukemia. 47 Hall PA, Russell SE. The pathobiology of the septin gene family. J N Engl J Med 2003; 348: 1764–1775. Pathol 2004; 204: 489–505. 28 Rai KR, Chiorazzi N. Determining the clinical course and outcome 48 Martinho RG, Castel S, Urena J, Fernandez-Borja M, Makiya R, in chronic lymphocytic leukemia. N Engl J Med 2003; 348: Olivecrona G et al. Ligand binding to heparan sulfate proteogly- 1797–1799. cans induces their aggregation and distribution along actin 29 Durig J, Nuckel H, Cremer M, Fuhrer A, Halfmeyer K, Fandrey J cytoskeleton. Mol Biol Cell 1996; 7: 1771–1788. et al. ZAP-70 expression is a prognostic factor in chronic 49 Fernandez-Borja M, Bellido D, Vilella E, Olivecrona G, Vilaro S. lymphocytic leukemia. Leukemia 2003; 17: 2426–2434. Lipoprotein lipase-mediated uptake of lipoprotein in human 30 Orchard JA, Ibbotson RE, Davis Z, Wiestner A, Rosenwald A, fibroblasts: evidence for an LDL receptor-independent internaliza- Thomas PW et al. ZAP-70 expression and prognosis in chronic tion pathway. J Lipid Res 1996; 37: 464–481. lymphocytic leukaemia. Lancet 2004; 363: 105–111. 50 Fernandez-Borja M, Bellido D, Makiya R, David G, Olivecrona G, 31 Rassenti LZ, Huynh L, Toy TL, Chen L, Keating MJ, Gribben JG Reina M et al. Actin cytoskeleton of fibroblasts organizes surface et al. ZAP-70 compared with immunoglobulin heavy-chain gene proteoglycans that bind basic fibroblast growth factor and mutation status as a predictor of disease progression in chronic lipoprotein lipase. Cell Motil Cytoskeleton 1995; 30: 89–107. lymphocytic leukemia. N Engl J Med 2004; 351: 893–901. 51 Tanaka M, Tanaka T, Kijima H, Itoh J, Matsuda T, Hori S et al. 32 Heintel D, Kienle D, Shehata M, Krober A, Kroemer E, Characterization of tissue- and cell-type-specific expression of a Schwarzinger I et al. High expression of lipoprotein lipase in poor novel human septin family gene, Bradeion. Biochem Biophys Res risk B-cell chronic lymphocytic leukaemia. Leukemia 2005; 19: Commun 2001; 286: 547–553. 1216–1223. 52 Scott M, Hyland PL, McGregor G, Hillan KJ, Russell SE, Hall PA. 33 Oppezzo P, Vasconcelos Y, Settegrana C, Jeannel D, Vuillier F, Multimodality expression profiling shows SEPT9 to be over- Legarff-Tavernier M et al. The LPL/ADAM29 expression ratio is a expressed in a wide range of human tumours. Oncogene 2005; novel prognosis indicator in chronic lymphocytic leukemia. Blood 24: 4688–4700. 2005; 106: 650–657. 53 Zheng Z, Venkatapathy S, Rao G, Harrington CA. Expression 34 Mead JR, Irvine SA, Ramji DP. Lipoprotein lipase: structure, profiling of B cell chronic lymphocytic leukemia suggests deficient function, regulation, and role in disease. J Mol Med 2002; 80: CD1-mediated immunity, polarized cytokine response, altered 753–769. adhesion and increased intracellular protein transport and proces- 35 Merkel M, Kako Y, Radner H, Cho IS, Ramasamy R, Brunzell JD sing of leukemic cells. Leukemia 2002; 16: 2429–2437. et al. Catalytically inactive lipoprotein lipase expression in muscle 54 Gomes LI, Esteves GH, Carvalho AF, Cristo EB, Hirata Jr R, Martins of transgenic mice increases very low density lipoprotein uptake: WK et al. Expression profile of malignant and nonmalignant lesions direct evidence that lipoprotein lipase bridging occurs in vivo. of esophagus and stomach: differential activity of functional Proc Natl Acad Sci USA 1998; 95: 13841–13846. modules related to inflammation and lipid metabolism. Cancer 36 Mamputu JC, Renier G. Differentiation of human monocytes to Res 2005; 65: 7127–7136. monocyte-derived macrophages is associated with increased 55 Lim LP, Lau NC, Garrett-Engele P, Grimson A, Schelter JM, Castle J lipoprotein lipase-induced tumor necrosis factor-alpha expression et al. Microarray analysis shows that some microRNAs down- and production: a process involving cell surface proteoglycans regulate large numbers of target mRNAs. Nature 2005; 433: and protein kinase C. Arterioscler Thromb Vasc Biol 1999; 19: 769–773. 1405–1411. 56 Wang HW, Trotter MW, Lagos D, Bourboulia D, Henderson S, 37 Strauss JG, Frank S, Kratky D, Hammerle G, Hrzenjak A, Knipping Makinen T et al. Kaposi sarcoma herpesvirus-induced cellular G et al. Adenovirus-mediated rescue of lipoprotein lipase-deficient reprogramming contributes to the lymphatic endothelial gene mice. Lipolysis of triglyceride-rich lipoproteins is essential for high expression in Kaposi sarcoma. Nat Genet 2004; 36: 687–693. density lipoprotein maturation in mice. J Biol Chem 2001; 276: 57 Pawlak G, Helfman DM. Cytoskeletal changes in cell transforma- 36083–36090. tion and tumorigenesis. Curr Opin Genet Dev 2001; 11: 41–47. 38 Ghia P, Caligaris-Cappio F. The indispensable role of microenvi- 58 Streubel B, Chott A, Huber D, Exner M, Jager U, Wagner O et al. ronment in the natural history of low-grade B-cell neoplasms. Adv Lymphoma-specific genetic aberrations in microvascular endothe- Cancer Res 2000; 79: 157–173. lial cells in B-cell lymphomas. N Engl J Med 2004; 351: 250–259. 39 Wu G, Brouckaert P, Olivecrona T. Rapid downregulation of 59 Staudt LM. Cancer: negative feedback for B cells. Nature 2004; adipose tissue lipoprotein lipase activity on food deprivation: 431: 919–920.

Leukemia LPL expression proﬁling in B-CLL M Bilban et al 1088 60 Fujita N, Jaye DL, Geigerman C, Akyildiz A, Mooney MR, systemic lupus erythematosus. Arthritis Rheum 2002; 46: Boss JM et al. MTA3 and the Mi-2/NuRD complex regulate 2957–2963. cell fate during B lymphocyte differentiation. Cell 2004; 119: 63 Sasaki H, Kunimatsu M, Fujii Y, Yamakawa Y, Fukai I, Kiriyama M 75–86. et al. Autoantibody to gravin is expressed more strongly in younger 61 Hall AM, Vickers MA, McLeod E, Barker RN. Rh auto- and nonthymomatous patients with myasthenia gravis. Surg Today antigen presentation to helper T cells in chronic lymphocytic 2001; 31: 1036–1037. leukemia by malignant B cells. Blood 2005; 105: 2007– 64 van’t Veer MB, Brooijmans AM, Langerak AW, Verhaaf B, 2015. Goudswaard CS, Graveland WJ et al. The predictive value of 62 Reichlin M, Fesmire J, Quintero-Del-Rio AI, Wolfson-Reichlin M. lipoprotein lipase for survival in chronic lymphocytic leukemia. Autoantibodies to lipoprotein lipase and dyslipidemia in Haematologica 2006; 91: 56–63.

Supplementary Information accompanies the paper on the Leukemia website (http://www.nature.com/leu)

Leukemia