Leukemia (2004) 18, 841–855 & 2004 Nature Publishing Group All rights reserved 0887-6924/04 $25.00 www.nature.com/leu Identification of whose expression patterns differ in benign lymphoid tissue and follicular, mantle cell, and small lymphocytic lymphoma

SC Schmechel1, RJ LeVasseur2, KH-J Yang1, KM Koehler1, SJ Kussick1 and DE Sabath1,3

1Department of Laboratory Medicine, University of Washington, Seattle, WA, USA; 2RationalDiagnostics, Inc., Seattle, WA, USA; and 3Department of Medicine, University of Washington, Seattle, WA, USA

Improved methods for diagnosing small B-cell lymphomas DNA microarrays make it possible to measure in parallel the (SBCLs) and predicting patient response to therapy are likely to expression of thousands of individual genes8 and represent one result from the ongoing discovery of molecular markers that better define these malignancies. In this report, we identify 120 means to identify new cancer-specific markers. In the area of genes whose expression patterns differed between reactive lymphoma biology, DNA array technology has been used to 9 10–13 lymph node tissue and three types of SBCL: follicular identify novel markers of SBCLs, including FL, MCL, and lymphoma, mantle cell lymphoma, and chronic lymphocytic CLL/SLL.14–16 These analyses have been useful to elucidate leukemia/small lymphocytic lymphoma. Whereas previously specific molecular pathways potentially involved in lymphoma published studies have generally analyzed the expression pathogenesis.17 Further, arrays have been used to identify genes profiles of one type of SBCL, work presented in this paper was intended to identify genes that are differentially expressed whose patterns of expression identify previously unrecognized between three SBCL subtypes. This analysis was performed but clinically distinct subtypes of diffuse large B-cell lymphoma 18–21 13 14,16 using mRNA pooled from multiple specimens representing (DLBCL), MCL, and CLL/SLL. Few published reports each tissue type. Quantitative reverse transcriptase-polymer- have focused on the identification of genes whose expression ase chain reaction (qRT-PCR) was used to validate the profiles differ between currently recognized SBCL sub- differential expression of 23 of these genes. Among the 23 types.13,14,16,22 validated genes were cyclin D1 (CCND1) and B-cell CLL/ lymphoma 2, which have well-known roles in lymphoma Toward the goal of identifying new SBCL markers, we have pathogenesis. The remaining 21 genes have no currently used cDNA microarrays to identify genes whose expression established role in lymphoma development. Using qRT-PCR, patterns differ between benign reactive lymph node tissues (RN) the expression of CCND1 and seven additional genes was and FL, MCL, and CLL/SLL specimens. We have validated the further studied in a panel of individual specimens. Genes expression patterns of 23 of these genes using quantitative identified in this study are of biological interest and represent reverse transcriptase-polymerase chain reaction (qRT-PCR) candidate diagnostic markers. Leukemia (2004) 18, 841–855. doi:10.1038/sj.leu.2403293 analysis, and analyzed the expression of eight genes in a panel Published online 12 February 2004 of individual lymphoma specimens. Importantly, we have Keywords: lymphoma; gene expression profiling; DNA microarray compared our data with publicly available data of other research groups in an attempt to crossvalidate SBCL markers. Genes identified in the course of these studies are of biological interest and represent candidate diagnostic markers. Introduction

Approximately 53 900 new cases of non-Hodgkin’s lymphoma Materials and methods (NHL) are diagnosed in the US annually.1 Together, small B-cell lymphomas (SBCLs), including follicular lymphoma (FL), mantle Clinical material cell lymphoma (MCL), and chronic lymphocytic lymphoma/ small lymphocytic lymphoma (CLL/SLL), comprise one-third of Lymph node and lymphoma specimens were obtained from the all NHL cases.2 SBCLs are indolent but generally not curable.3 University of Washington (UW) Hematopathology Laboratory The time from diagnosis to death is quite variable, ranging from tissue bank. Freshly excised tonsils were obtained from Seattle months to 20 years.4 Advances in understanding the biological Children’s Hospital and Medical Center. All studies were basis and clinical behavior of SBCLs rely on accurate diagnoses. approved by institutional review boards of the University of The currently used Revised European–American Lymphoma Washington Medical Center and Children’s Hospital and (REAL)5 and World Health Organization6 lymphoma classifica- Regional Medical Center. Between 1989 and 1996, lymph tion systems are based on tumor morphology, molecular node and lymphoma specimens were surgically removed from abnormalities, and the immunocytochemical measurement of patients in the course of their medical care at the UW Medical a limited number of leukocyte markers. Using the REAL Center or one of several referral medical facilities in western classification system, there is an approximate 15% interpathol- Washington, Idaho, Montana, and Alaska. Tissues not needed ogist diagnostic variability rate.7 A classification scheme based for diagnostic testing were frozen in a water-soluble tissue- on measuring the expression of a larger number of lymphoma freezing medium (OCT; Tissue-Tek, Naperville, IL, USA) and markers, reflecting the complexity of molecular defects in transferred to a À701C freezer, where they were maintained SBCLs, may yield more accurate diagnoses and may guide until processing. Each specimen was stripped of patient therapy. identifier information with the exception of final diagnosis and anatomic source, and the data were compiled in a computer Correspondence: DE Sabath, University of Washington, 1959 NE database. As a result, information such as patient age, sex, stage, Pacific Street (Box 357110), Seattle, WA 98195, USA; E-mail: [email protected] symptoms, laboratory parameters, performance status, treat- Received 23 May 2003; accepted 2 December 2003; Published online ment, and outcome was not available for analysis. Specimens 12 February 2004 used in this study were identified through a computer-based Gene expression analysis in small B-cell lymphomas SC Schmechel et al 842 search for RN, grade I FL (excluding cases noted to be grade II or transcribed into cDNA labeled with Cy3-dCTP (AP Biotech, III), MCL (excluding cases noted to have blastoid morphologic Little Chalfont, Buckinghamshire, UK) as described pre- features), and SLL (irrespective of the level of CD38 expression). viously;24 in a second reaction, 2 mg mRNA were reverse Cyclin D1 overexpression was validated by immunohistochem- transcribed into cDNA labeled with Cy5-dCTP (AP Biotech). istry for a subset of MCL cases (data not shown). Retrospective Labeled cDNAs were purified as described previously24 and analysis of flow cytometric data collected at the time of dissolved in 100 ml10mM Tris, pH 8.0. The efficiency of Cy3/ diagnosis for a subset of cases revealed that RN specimens Cy5-dCTP incorporation was determined using an HP 8452A comprised B52% of B cells (n ¼ 4, range 40–65%), FL speci- diode array spectrophotometer and the following formulae: mens comprised B64% neoplastic B cells (n ¼ 4, range 50– 75%), MCL specimens comprised B72% neoplastic cells A550Âprobe volume=0:15 ¼ pmol Cy3 probe; (n ¼ 10, range 22–96%), and SLL specimens comprised B75% and neoplastic cells (n ¼ 8, range 50–94%) (data not shown). Flow A Âprobe volume=0:25 ¼ pmol Cy5 probe: cytometric analysis revealed that tonsil specimens comprised 650 60–80% B cells; 60–70% of these B cells showed low CD10- A typical yield for each Cy3-labeled cDNA was 100 pmol and positive staining characteristic of follicle center origin (data not for each Cy5-labeled cDNA was 75 pmol. shown). The spotted cDNA microarrays used in this study were obtained from the University of Washington Center for Expres- sion Arrays (UW-CEA). The arrays contained robotically spotted RNA isolation PCR-amplified gene fragments from 14 976 Homo sapiens sequence-verified Integrated Molecular Analysis of Genomes In all, 18 RN, 21 FL, 25 SLL, and 13 MCL specimens were and their Expression (IMAGE) consortium25 clones (UniGene 1 1 transferred on dry ice from –70 C freezer to a À20 C Tissue-Tek Build 19, plates 1–44) obtained from Research Genetics II Microtome/Cryostat. Using a cryostat, approximately 50 10- (Huntsville, AL, USA).24 Two sets of slides were used. Human B mm tissue sections (representing 250 mg of tissue) were cut HD-1 and HD-2 arrays each contained nearly unique sets of from each specimen and placed in a 50 ml conical tube on ice. 7488 cloned human genes and expressed sequence tags (ESTs) B Fresh tonsil specimens, each in 20 ml RPMI medium (Life spotted in duplicate. A complete list of the genes contained in Technologies, Rockville, MD, USA), were oriented in plastic these arrays is available at http://ra.microslu.washington.edu/ Petri dishes with the epithelial side down. Using a scalpel, the Website/genelist/genelist.html. tissue was finely chopped against the underside of the epithelial Each array was rinsed 10 times in sterile H2O and then layer to free nonadherent cells into the medium. Cells immediately dried using compressed air. Fluorescently labeled suspended in the medium were transferred to 50 ml conical cDNAs were combined as described in the results section, tubes, pelleted by centrifugation for 20 min at 800 Â g in an IEC concentrated by drying, resuspended in 20 ml of hybridization CentraCL centrifuge (International Equipment Company, Need- solution (50% deionized formamide (Sigma, St Louis, MO, ham Heights, MA, USA), and the supernatant was discarded. A USA), 5 Â SSC (0.75 M sodium chloride, 75 mM sodium citrate; sufficient volume (typically 1–5 ml for lymphoma tissue speci- Ambion, Austin, TX, USA), 5 Â Denhardt’s solution (Fisher mens and 5 ml for pelleted tonsil cells) of phenol/guanidine Scientific, Houston, TX, USA), 0.1% sodium dodecyl sulfate isothiocyanate (TRIzol; Invitrogen Life Technologies, Carlsbad, (SDS; Ambion, Austin, TX, USA), 100 mg/ml CotI DNA (Invitro- CA, USA) was added. Samples were vortexed thoroughly and gen), and 20 mg/ml polyA [50-A(75)-30] primer (Invitrogen)), placed on ice for up to several hours. Total RNA was isolated denatured by boiling for 3 min, chilled on ice for 30 s, and according to the TRIzol manufacturer’s instructions and 23 placed at room temperature (rt). Labeled cDNAs were added to quantified by spectrophotometry using a Hewlett Packard each array and covered with 64 mm by 25 mm coverslips. 8452A Diode Array spectrophotometer (Hewlett Packard, Palo Microarrays were hybridized for 14 to 16 h at 421Cina Alto, CA, USA). Poly (A) þ mRNA was purified from total RNA humidified chamber (Genetix Limited, Hampshire, UK). Follow- using oligo(dT)25-linked magnetic beads (Dynal, Oslo, Norway) ing hybridization, the microarrays were washed briefly in according to the manufacturer’s instructions. RNA was labeled 1 Â SSC/0.2%SDS (prewarmed to 541C) to remove the cover- with RiboGreen (Molecular Probes, Eugene, OR, USA) accord- slips. The arrays were transferred to glass dishes in which they ing to the manufacturer’s instructions and quantified using a were washed by gentle rocking in 1 Â SSC/0.2% SDS (pre- Versafluor fluorometer (Bio-Rad Laboratories, Hercules, CA, warmed to 541C) for 10 min, 1 Â SSC/0.2% SDS (prewarmed to USA) by comparison to a standard curve generated using known 541C) for 10 min, 0.1 Â SSC/0.2% SDS (prewarmed to 541C) for concentrations of RNA (Molecular Probes). A typical yield from 10 min, 0.2 Â SSC (rt) for 1 min, and 0.1 Â SSC (rt) for 1 min. each B250 mg tissue specimen was 0.5–4 mg of mRNA. For Finally, the arrays were dipped twice in distilled H2O and dried array analysis, an equal amount of mRNA was pooled from with compressed air. The microarrays were scanned at 532 and multiple specimens identified in a footnote to Table 1. Pooled 633 nm using a Molecular Dynamics Avalanche dual-laser mRNA was analyzed using an Agilent 2100 bioanalyzer (Agilent confocal scanner. Technologies, Palo Alto, CA, USA). In each case, rRNA contamination was p14% and mRNA migrated as a typical population of species predominantly ranging in size from 1.3 to Microarray data analysis 4.4 kb (data not shown). Duplicate human HD1 and HD2 slides were hybridized with the same cDNAs, but with the fluorescent labels reversed to cDNA synthesis, microarray construction, and correct for dye-specific effects.26 Each slide contained two hybridization identical sets of spots. Two images, each corresponding to one set of spots, were obtained for each slide. Using Spot-On Two labeled cDNA populations were prepared from each software developed at the UW-CEA,24 the intensity of each spot mRNA pool. In one reaction, 2 mg mRNA were reverse (subtracted for local background) in both channels was obtained

Leukemia Gene expression analysis in small B-cell lymphomas SC Schmechel et al 843 from each image and exported as a text file. Using GeneSifter. expression as follows: NetTM (VizX Labs, Seattle, WA, USA), the intensity value of each spot in an image was normalized to the mean intensity of all Normalized RNA quantity  2CtGOIÀCtcyclophilin spots in that image. For each IMAGE clone of interest, the gene name, accession number, and functional summary were where GOI is the gene of interest (eg 13cDNA73). The cyclophilin obtained from public databases using GeneSifter.Nett and the gene has been described previously as a useful internal reference 27 Stanford Online Universal Resource for Clones and ESTs. The RNA.30 For the purpose of displaying gene expression on a relative raw microarray data from these experiments are available at scale, the data for each gene were further normalized to the http://expression.washington.edu/public. median expression across all individual lymphoma specimens. The raw qRT-PCR data from these experiments are available at http:// qRT-PCR validation of array data expression.washington.edu/public.

For confirmation of array data using RNA pooled from multiple Results specimens (Figures 1 and 2), an equal amount of mRNA from each specimen was combined as described in the footnotes of cDNA array analysis Table 1. PCR mixtures (40 ml) contained 1  AmpliTaq Gold Buffer (Applied Biosystems, Foster City, CA, USA), 4 mM MgCl2, The goal of this study was to identify genes whose expression 0.025 U/ml AmpliTaq Gold (Applied Biosystems), 0.25 U/ml patterns differed between RN, FL, MCL, and SLL. We Moloney leukemia virus reverse transcriptase (Invitrogen, hypothesized that differentially expressed genes may be Carlsbad, CA, USA), 0.4 U/ml RNase inhibitor (Invitrogen), involved in lymphoma pathogenesis and serve as useful 0.5 mg/ml BSA (Ambion, Austin, TX, USA), 0.33  SYBR Green diagnostic markers. To identify differentially expressed genes, I (obtained as a 10 000  solution from Molecular Probes, we utilized cDNA microarrays constructed at the UW-CEA. Eugene, OR, USA), 0.8 mM passive reference DNA oligohex- These arrays contained duplicate spots of PCR-amplified insert 0 0 amer, 5 -(6-carboxyrhodamine)-GATTAG-PO4-3 (Rox Standard cDNAs from 14 976 IMAGE clones25 representing B13 500 I, Synthegen, Houston, TX, USA), 200 mM dNTPs (Amersham individual UniGene clusters. Messenger RNA was purified from Biosciences, Piscataway, NJ, USA), 5 ng mRNA, and 50 nM archival tissue specimens that had been frozen shortly after their gene-specific primers (Invitrogen). All gene-specific primers surgical removal and maintained at À701C. As our protocol t were designed using the computer program Primer Express 1.5 required a large quantity (2 mg) of mRNA for the synthesis of (Applied Biosystems) using default settings for RT-PCR primer fluorescently labeled target cDNA for array analysis, we chose selection and are listed in Table 2. Intron spanning primers were to pool equal amounts of mRNA from multiple specimens selected if splice junctions could be identified using the Ensembl 28 representing the same tissue type. A similar RNA pooling database. Using an ABI 7700 sequence detector (Applied strategy was shown previously to be useful in identifying genes Biosystems), the reactions were subjected to the following specifically differentially expressed in colon cancer.31 Messen- 1 1 cycling conditions: 30 min at 48 C, 10 min at 95 C, and 40 ger RNA from 17 RN specimens was combined to generate a 1 1 cycles comprising 15 s at 95 C and 1 min at 60 C. The relative single pool of RN mRNA. Similarly, we made three additional amount of mRNA in two samples, for example, sample x pools of mRNA from 21 grade I FL specimens, nine MCL and sample y, were semiquantified based on the following specimens, and 25 SLL specimens. Cy3- and Cy5-labeled cDNA formula: was generated from the lymph node and lymphoma mRNA pools. In addition, Cy3- and Cy5-labeled cDNA was generated ½RNAŠSample x  ECtSample y ÀCtSample x from RNA pooled from several tonsils for use as a reference ½RNAŠSample y control. Sample and reference cDNAs were combined and hybridized to microarrays. All array experiments were per- where Ct is the number of PCR cycles resulting in a threshold formed in duplicate where the labeling scheme was reversed to fluorescence level and E, a measure of PCR reaction efficiency, compensate for potential dye-specific incorporation effects and was assumed to be 2 in the exponential phase of PCR product for dye-dependent nonlinearity of signal intensity.26 For accumulation. For each of the 23 validated genes, array and example, Cy3-labeled lymphoma cDNA and Cy5-labeled tonsil qRT-PCR data were normalized to the level of gene expression cDNA were hybridized to one array, and Cy5-labeled lympho- in the tonsil reference RNA pool. A self-organizing map was ma cDNA and Cy3-labeled tonsil cDNA were hybridized to a generated using the default settings of the Cluster analysis second array. After hybridization, microarray slides were program and data were displayed using the Treeview program; washed under conditions of increasing stringency and scanned both programs were written by Eisen et al.29 For qRT-PCR in the Cy3 and Cy5 channels using a laser confocal scanner. analysis of individual specimens (Figure 3), variable amounts Signal and local background intensities were quantified for (p5 ng) of poly(A) þ RNA were used since 5 ng were not each spot on the arrays. Since the majority of IMAGE clones available for all cases. In all, 10 cases of RN (RN-1, RN-2, RN-3, were spotted in duplicate on the arrays and since two arrays RN-4, RN-5, RN-6, RN-7, RN-8, RN-9, and RN-10), nine cases were used per sample, four measurements were obtained for of FL (FL-1, FL-2, FL-3, FL-4, FL-5, FL-6, FL-7, FL-8, and FL-9), most clones. However, some IMAGE clones were represented nine cases of MCL (MCL-1, MCL-2, MCL-3, MCL-4, MCL-5, by more than one set of spots on the arrays. Further, some genes MCL-6, MCL-7, MCL-8, and MCL-9), and 10 cases of SLL (SLL-1, were represented by more than one IMAGE clone. In these SLL-2, SLL-3, SLL-4, SLL-5, SLL-6, SLL-7, SLL-8, SLL-9, and SLL- cases, statistics for each set of spots were calculated indepen- 10) were studied. qRT-PCR was performed as described above, dently. In all, 120 genes were selected that were 44-fold except that in addition, the expression of cyclophilin was differentially expressed in pairwise comparisons between RN, determined using primers 50-CCACCGTGTTCTTCGACATTG-30 FL, MCL, and SLL tissues, and showed corresponding two-sided (forward) and 50-TCTTTGGGACCTTGTCTGCAA-30 (reverse), P-values of o0.05 (derived from a t-test analysis assuming equal and gene expression levels were normalized to cyclophilin variances and six degrees of freedom). Table 1 lists the IMAGE

Leukemia Gene expression analysis in small B-cell lymphomas SC Schmechel et al 844

Figure 1 Summary flow chart depicting the comparison between gene expression data from microarray and qRT-PCR analyses. The expression of 39 of 120 genes identified by microarray analysis to be 44-fold differentially expressed was quantified by qRT-PCR. Using a threshold of 42- fold differential expression by qRT-PCR analysis, the expression patterns of 23 of the 39 genes were confirmed to be similar both by microarray and qRT-PCR methods. The qRT-PCR product for each of these 23 genes was of expected size by PAGE analysis.

Leukemia Gene expression analysis in small B-cell lymphomas SC Schmechel et al 845

Figure 2 Graphical depiction of expression data for 23 genes whose gene expression profiles were similar by microarray and qRT-PCR analysis. The expression measurements for each gene in reactive lymph node (RN), FL, MCL, and SLL tissues were normalized to the expression level in a reference (tonsil) RNA sample. Each gene (identified at right) is represented by a single row of colored boxes; each tissue type is represented by a single column. Intensity of red indicates the degree of overexpression, whereas intensity of blue indicates the degree of underexpression relative to the expression level in tonsil.

clone numbers, derived names and accession numbers, and fold tion of unintended qRT-PCR amplicons. To investigate these two differential expression data for these 120 genes; 37 genes potential sources of error, we performed qRT-PCR analysis on uniquely differentially expressed in a single lymphoma subtype additional genes and identified a total of 24 genes that gave vs benign lymph nodes are indicated in bolded italics. discrepant array and qRT-PCR results (data not shown). We sequenced both the PCR products representing the 24 cDNA array spots and the corresponding 24 amplicons generated using Validation of array results using qRT-PCR qRT-PCR primers designed to amplify the expected genes. We found that 22 (94%) of 24 qRT-PCR amplicons were of the We used the method of qRT-PCR with SYBR Green I dye expected sequence; on the contrary, only one (4%) of 24 PCR detection32 to quantify the relative RNA expression of 39 of the products representing cDNA spots was of the expected 120 genes. These 39 genes were selected based on their high sequence (data not shown). We concluded that a principal differential expression by array analysis. qRT-PCR was per- source of discrepancy between array data and qRT-PCR formed in duplicate on pooled mRNA from multiple RN, FL, validation data was due to misidentification of cDNA spots on MCL, and SLL specimens as described in the footnotes of the arrays. The occurrence of spot misidentification has been Table 1. Oligonucleotide primers used in this qRT-PCR analysis described by other investigators using cDNA array technol- are listed in Table 2. A flow chart comparing our array and qRT- ogy.33 PCR results is shown in Figure 1. We found that data for four genes (10% of 39) were not informative, since high signal levels were obtained in no-RT control samples (most likely due to qRT-PCR analysis of gene expression in individual amplification from contaminating genomic DNA rather than specimens from mRNA). Of the remaining 35 genes, 23 (66%) were found by qRT-PCR to be 42-fold differentially expressed in the same As we used an RNA pooling strategy to identify differentially direction as the microarray data and to give amplicons that expressed genes, we were next interested in determining the migrated as a single band of expected size by polyacrylamide expression levels of selected validated genes in individual gel electrophoresis (PAGE) (data not shown). We arbitrarily lymphoma specimens. For this analysis, we quantified the level selected a less stringent two-fold cutoff for differential expres- of expression of eight genes (in addition to the cyclophilin gene, sion in qRT-PCR analysis rather than the four-fold cutoff used in which was used as an internal control for normalization) in 10 array analysis. Figure 2 compares our array and qRT-PCR results RN, nine FL, nine MCL, and 10 SLL specimens using qRT-PCR. for these 23 validated genes in pseudocolor graphics where the Data are presented as box plots in Figure 3 to convey the degree expression level of each gene in RN, FL, MCL, and SLL is of variability in gene expression among specimens. In general, normalized to its expression level in the reference RNA (tonsil) data obtained using individual specimens correlated well with pool. qRT-PCR data obtained using pooled RNA (Figure 2). Our degree of concordance between cDNA array data and As expected, we found that CCND1 was overexpressed in qRT-PCR data is in agreement with other investigators, who each case of MCL relative to other specimen types. On average, have found that the expression patterns of approximately two- CCND1 expression was 28-fold higher in MCL than in RN thirds of genes identified using cDNA arrays can be confirmed (P ¼ 0.0003 derived from a t-test analysis assuming equal using qRT-PCR.32 Discrepancies between array data and qRT- variances), 40-fold higher than in FL (P ¼ 0.0005), and 33-fold PCR validation data may be due to the misidentification of higher than in SLL (P ¼ 0.0003) specimens. DNMT3A, ITM3, cDNA molecules spotted on the arrays or due to the amplifica- and KIAA1407 were also generally overexpressed in MCL.

Leukemia Gene expression analysis in small B-cell lymphomas SC Schmechel et al 846

Figure 3 Analysis of gene expression in individual specimens of RN, FL, MCL, and SLL. The expression of 13cDNA73, CCND1, DMNT3A, ITM3, KIAA1407, KIAA1959, MGC40441, and MYBL2 in 10 RN, nine FL, nine MCL, and 10 SLL individual specimens was quantified using qRT- PCR. Expression data for each gene were first normalized to the level of cyclophilin expression and subsequently normalized to the median level of expression in all specimens. Box plots are presented, in which the horizontal lines within each box represents the median expression value across all specimens of a tissue type, the lower and upper boundaries of each box represent the 25th and 75th percentile, respectively, whiskers represent either the minimal and maximal values or 1.5 times the spread between the 25th and 75th percentile (whichever is nearer to the body of the data), and circles represent any values outside of the whiskers. Tissue types are identified on the x-axis; relative expression is indicated on the y-axis.

DNMT3A expression in MCL specimens was 4.5-fold higher (P ¼ 0.04) higher than in RN, FL, and MCL specimens, than in RN (P ¼ 0.0007), 13-fold higher in FL (P ¼ 0.0002), and respectively. In agreement with pooled RNA qRT-PCR results, eight-fold higher than in SLL (P ¼ 0.0001) specimens. ITM3 was 13CDNA73 was overexpressed in FL and SLL specimens relative expressed in MCL specimens 2.5-fold (P ¼ 0.008), 9.2-fold to RN and MCL specimens, although there was high variability (P ¼ 0.0003), and 4.4-fold (P ¼ 0.001) than in RN, FL, and SLL in the expression of this gene product in FL and SLL tissues. specimens, respectively. KIAA1407 was variably expressed in 13CDNA73 was expressed in FL specimens 4.6-fold (P ¼ 0.007) MCL, with some MCL specimens expressing very little of this and six-fold (P ¼ 0.006) higher than in RN and MCL specimens, gene product. On average, KIAA1407 was 21-fold (P ¼ 0.02), respectively; 13CDNA73 was expressed in SLL 16-fold 37-fold (P ¼ 0.02), and 3.1-fold (P ¼ NS) overexpressed in MCL (P ¼ 0.0003) and 21-fold (0.0004) higher than in RN and MCL, specimens than in RN, FL, and SLL specimens, respectively. respectively. MYBL2, on the contrary, was more highly KIAA1959 was generally overexpressed in SLL, although the expressed in RN and MCL specimens than in FL and SLL. expression of this gene in individual SLL specimens was MYBL2 expression in RN specimens was 10-fold (P ¼ 0.0001) variable. On average, KIAA1959 was expressed in SLL speci- and 4.8-fold (P ¼ 0.0002) higher than in FL and SLL; MYBL2 mens five-fold (P ¼ 0.02), nine-fold (P ¼ 0.02), and 4.6-fold expression in MCL specimens was 7.1-fold (P ¼ 0.0006) and 3.4-

Leukemia Table 1 Genes differentially expressed among RN, FL, MCL, and SLL tissuesa

Fold changeb

IMAGE clone GenBank accession no. Name Description FL/RN MCL/RN SLL/RN FL/MCL FL/SLL MCL/SLL

1. Apoptosis 232714 NM_000633 BCL2e B-cell CLL/lymphoma 2 4.4 712049 NM_006850 IL24f Interleukin-24 À8.1 592125 NM_003804 RIPK1 TNFRSF-interacting serine-threonine kinase 1 À4.2 345586 AB018263 TNFRSF12 Tumor necrosis factor receptor superfamily member 12 4.3 2. Cell adhesion 897910 NM_006475 OSF-2 Osteoblast-specific factor 2 4 3. Cell cycle regulation 841641 NM_053056 CCND1c Cyclin D1 13.8 À10.1 17.1 898286 NM_001786 CDC2 Cell division cycle 2 À7.7 244058 NM_031942 CDCA7 Cell division cycle-associated 7 8.8 4. Cell signaling 813279 NM_001640 APEH N-acylaminoacyl-peptide hydrolase À6.5 6 950690 NM_001237 CCNA2c,h Cyclin A2 À5 509823 NM_002483 CEACAM6 -related cell adhesion molecule 6 À4.4 4.7 843139 NM_003915 CPNE1f Copine I 7.1 À4.5 681906 NM_005248 FGR Gardner–Rasheed feline sarcoma viral oncogene homolog À5.8 À4.5 131318 NM_018842 LOC55971 tyrosine kinase substrate À8.4 782141 NM_147180 PPP3R2 phosphatase 3 regulatory subunit B À4.4 5. Cytoskeleton 140951 NM_004924 ACTN4 Actinin alpha 4 4.5 810131 NM_002276 KRT19d Keratin 19 À5 lymphomas B-cell Schmechel small SC in analysis expression Gene 592540 NM_000424 KRT5e Keratin 5 À5.3 À4 138496 NM_003480 MAGP2 Microfibril-associated glycoprotein-2 4.2 6. Extracellular matrix 280567 NM_152890 COL24A1 Collagen type XXIV alpha 1 5.1 4.9 4.9 al et 42373 NM_001888 CRYMc Crystallin mu 8.5 À7.3 9.3 811162 NM_002023 FMOD Fibromodulin 4.4 À4.6 À4.8 246722 NM_006216 SERPINE2 Serine (or cysteine) proteinase inhibitor, clade E 2 À4.1 4.5 7. Gene expression regulation 949971 NM_001675 ATF4 Activating transcription factor 4 4.4 4.3 296483 NM_021813 BACH2 Basic leucine zipper transcription factor 2 5.5 810057 NM_003651 CSDA Cold-shock domain protein A À4.4 240367 NM_006565 CTCFd CCCTC-binding factor À4.1 9.3 10.2 202514 AF331856 DNMT3Ae DNA (cytosine-5-)-methyltransferase 3 alpha À8 5.2 810061 NM_005537 ING1 Inhibitor of growth family member 1 4.5 358531 NM_002228 JUNc,g v-jun sarcoma virus 17 oncogene homolog -4.6 188232 NM_004235 KLF4f EST 5 347036 NM_016269 LEF1 Lymphoid enhancer-binding factor 1 À5 À4.4 815526 NM_002466 MYBL2c v-myb myeloblastosis viral oncogene homolog 2 5.1 884438 NM_006164 NFE2L2 Nuclear factor (erythroid-derived 2)-like 2 4.1 809648 NM_004630 SF1d,g Splicing factor 1 À4.2 À7 725680 NM_003222 TFAP2Cd Transcription factor AP-2 gamma À8.2 À7.2 8. Immunity/inflammation 179890 NM_000698 ALOX5 Arachidonate 5-lipoxygenase 5.8 208718 NM_000700 ANXA1 Annexin A1 À4.1 136919 NM_019846 CCL28 Chemokine (C–C motif) ligand 28 5.2 Leukemia 847 Leukemia 848

Table 1 Continueda

Fold changeb

IMAGE clone GenBank accession no. Name Description FL/RN MCL/RN SLL/RN FL/MCL FL/SLL MCL/SLL

205633 NM_002984 CCL4c,g Chemokine (C–C motif) ligand 4 À5.6 À8.7 704459 NM_001781 CD69e CD69 antigen À4.1 50214 NM_006889 CD86d CD86 antigen 7.9

290749 NM_032738 FREB Fc receptor homolog expressed in B cells 4.8 lymphomas B-cell small in analysis expression Gene 289337 BC019046 IGHMc Immunoglobulin heavy constant gamma 3 À6.6 80948 NM_144646 IGJd Immunoglobulin J polypeptide À4.4 À5.6 488019 NM_002189 IL15RA Interleukin-15 receptor alpha À4.1 809776 NM_004513 IL16 Interleukin-16 4.9 À7.8 714453 NM_000418 IL4Re,g Interleukin-4 receptor 5.4 À7.1 9. Metabolism 288736 NM_004827 ABCG2 ATP-binding cassette subfamily G (WHITE) member 2 À8.4 À4.6 À7.2 809523 NM_000483 APOC2e Apolipoprotein C-II 5.6 159608 NM_001647 APOD Apolipoprotein D 5.5 À8.3 À5.8 782758 NM_014257 CD209Lf CD209 antigen like À5.1 À4.8 363058 NM_001830 CLCN4d Chloride channel 4 À4.9 Schmechel SC 796297 NM_020944 GBA2 Bile acid b-glucosidase À5.1 795173 NM_000405 GM2Ae GM2 ganglioside activator protein À4.4 813426 NM_021643 GS3955f,h GS3955 protein À5.4 7 4.1

135608 NM_017784 OSBPL10 Oxysterol-binding protein-like 10 4.7 al et 840493 NM_002933 RNASE1d RNase A family 1 À5.4 51406 NM_022829 SLC13A3 Solute carrier family 13 member 3 4.3 4.9 121981 NM_006931 SLC2A3e,h Solute carrier family 2 member 3 À4.3 773617 NM_003339 UBE2D2 Ubiquitin-conjugating enzyme E2D 2 4.4 10. Plasma membrane protein 784910 NM_005277 GPM6A Glycoprotein M6A 5.8 11. Unknown 46284 NM_023037 13CDNA73c Hypothetical protein CG003 À7.6 754479 NM_006820 C1ORF29 1 open-reading frame 29 À4.3 305302 BM906531 C3ORF6 Chromosome 3 open-reading frame 6 4.5 À7.8 309499 NM_024728 C7ORF10 Chromosome 7 open-reading frame 10 4.5 60201 NM_025263 CAT56 CAT56 protein 4.3 258242 BE786990 FLJ21195 Hypothetical protein FLJ21195 4.1 212772 NM_025113 FLJ21562c Hypothetical protein FLJ21562 4.8 À5.3 501778 NM_024713 FLJ22557 Hypothetical protein FLJ22557 6.5 15.7 11.9 205049 NM_014365 H11 Protein kinase H11 À4.2 47151 R48935 IMAGE:47151 EST 4.5 53092 BG284034 IMAGE:53092 Putative L-type neutral amino-acid transporter 4.3 110582 T90074 IMAGE:110582 EST 6.8 6.4 121977 T97780 IMAGE:121977 EST À4.1 122702 BC034319 IMAGE:122702 EST À5 122723 AA777690 IMAGE:122723 EST À5.2 127710 AA579610 IMAGE:127710 EST À4.1 À5.2 130742 H13708 IMAGE:130742 EST 5 À4.3 À4.5 133613 R30836 IMAGE:133613 EST À4.4 136909 BU162571 IMAGE:136909 EST 4.5 193771 BQ322085 IMAGE:193771 EST À4.1 201981 BC025340 IMAGE:201981e EST 8.4 À13.6 6.3 203114 BF431502 IMAGE:203114 EST À4 204740 H57305 IMAGE:204740 EST 5.9 Table 1 Continueda

Fold changeb

IMAGE clone GenBank accession no. Name Description FL/RN MCL/RN SLL/RN FL/MCL FL/SLL MCL/SLL

234376 AK097411 IMAGE:234376 EST À5.2 258118 N27108 IMAGE:258118f EST 6.2 265294 N20848 IMAGE:265294 EST À8 278944 AL121338 IMAGE:278944 EST 4 284584 N59450 IMAGE:284584 EST 4.4 287721 N79323 IMAGE:287721 EST À5.2 4 294647 BE971364 IMAGE:294647 EST 9.4 4.7 325024 BC022095 IMAGE:325024 Hypothetical protein DKFZp434P0531 4.8 325247 BU630466 IMAGE:325247 EST À5.8 À4.3 382773 AA065090 IMAGE:382773 EST 4.5 418185 NM_031305 IMAGE:418185 Hypothetical protein DKFZp564B1162 5.6 4.4 429165 BF677678 IMAGE:429165 EST 4.8 429569 AI248013 IMAGE:429569 EST 5 4.2 503051 BU630466 IMAGE:503051 EST 4.1 À4.8 564567 AA127395 IMAGE:564567 EST À8 À6.8 626199 BG283145 IMAGE:626199 EST À6.4 471196 NM_030926 ITM3c Integral membrane protein 3 4.6 812975 D79994 KANK Kidney ankyrin repeat-containing protein 4.2 4 742904 BQ070901 KCP2 Keratinocytes-associated protein 2 À5.5 210368 NM_014792 KIAA0125d KIAA0125 gene product À4.4 À5 784104 NM_014686 KIAA0355 KIAA0355 À4.5 810621 AB029034 KIAA1111 KIAA1111 protein À10.8 À5.8 417637 BQ722784 KIAA1276 KIAA1276 protein À5 lymphomas B-cell Schmechel small SC in analysis expression Gene 321886 AB037771 KIAA1350c KIAA1350 4.2 121475 AF509494 KIAA1407c KIAA1407 protein 10 4.1 489047 NM_032873 KIAA1959c NM23-phosphorylated substrate À5.3

341096 BM546103 LOC145758 Hypothetical protein LOC145758 À6 al et 259902 NM_016570 LOC51290f,h CDA14 5.8 À8.8 781088 BC001077 LOC87769 Hypothetical protein BC004360 À4.9 À4.9 202315 NM_138379 LOC91937 Hypothetical protein BC008988 À4.1 À4.3 293005 NM_152785 MGC40441c Germinal center expressed transcript 2 À10.7 À8.7 11.9 9.6 126450 NM_024319 MGC4174 EST À4.4 À5 123735 AU142060 NUDT4P2 EST 4.2 128506 NM_015670 SENP3 Sentrin/SUMO-specific protease 3 4.1 784218 NM_015436 ZNF363 Zinc-finger protein 363 4.3 aArray analysis performed using cDNA synthesized from pooled polyA(+) RNA from 17 RN specimens (RN-1, RN-2, RN-3, RN-4, RN-5, RN-7, RN-8, RN-9, RN-10, RN-11, RN-12, RN-13, RN-14, RN-15, RN- 17, RN-18, and RN-19), 21 FL specimens (FL-1, FL-2, FL-3, FL-4, FL-5, FL-6, FL-7, FL-8, FL-9, FL-10, FL-11, FL-12, FL-13, FL-14, FL-15, FL-16, FL-17, FL-18, FL-19, FL-20, and FL-21), nine MCL specimens (MCL-1, MCL-2, MCL-3, MCL-6, MCL-7, MCL-8, MCL-9, MCL-10, and MCL-11), and 25 SLL specimens (SLL-1, SLL-2, SLL-3, SLL-4, SLL-5, SLL-6, SLL-7, SLL-8, SLL-9, SLL-10, SLL-11, SLL-12, SLL-13, SLL-14, SLL-15, SLL-16, SLL-17, SLL-18, SLL-19, SLL-20, SLL-21, SLL-22, SLL-23, SLL-24, and SLL-25). bDownregulation is indicated by negative numbers. cGene expression pattern confirmed by qRT-PCR analysis performed on pooled polyA(+) RNA from 12 RN specimens (RN-1, RN-2, RN-3, RN-4, RN-5, RN-9, RN-10, RN-11, RN-13, RN-15, RN-16, and RN- 17), 12 FL specimens (FL-2, FL-5, FL-6, FL-8, FL-9, FL-11, FL-14, FL-15, FL-16, FL-18, FL-20, and FL-21), 11 MCL specimens (MCL-1, MCL-2, MCL-3, MCL-4, MCL-5, MCL-6, MCL-7, MCL-8, MCL-9, MCL- 12, and MCL-14), and 12 SLL specimens (SLL-2, SLL-3, SLL-5, SLL-6, SLL-7, SLL-9, SLL-10, SLL-13, SLL-14, SLL-18, SLL-19, and SLL-20). dGene expression pattern not confirmed by qRT-PCR analysis as described abovec. eGene expression pattern confirmed by qRT-PCR analysis performed on polyA(+) RNA derived from 11 RN specimens (RN-1, RN-2, RN-3, RN-4, RN-5, RN-9, RN-10, RN-11, RN-13, RN-15, and RN-17), nine FL specimens (FL-2, FL-5, FL-6, FL-8, FL-9, FL-11, FL-15, FL-16, and FL-20), 11 MCL specimens (MCL-1, MCL-2, MCL-3, MCL-4, MCL-5, MCL-6, MCL-7, MCL-8, MCL-9, MCL-12, and MCL-14), and 11 SLL specimens (SLL-2, SLL-3, SLL-5, SLL-6, SLL-7, SLL-9, SLL-10, SLL-13, SLL-14, SLL-19, and SLL-20). fGene expression pattern not confirmed by qRT-PCR analysis as described abovee. g IMAGE clones corresponding to these genes were represented by more than one set of spots on the micorarrays; data obtained from only one set of spots gave significant (44-fold, Po0.05) results. h

Leukemia Multiple IMAGE clones corresponding to these genes were represented on the microarrays; data obtained from spots corresponding to only one IMAGE clone gave significant (44-fold, po0.05) results. Genes uniquely over- or underexpressed in one type of lymphoma vs RN tissue shown in bold italics. 849 Gene expression analysis in small B-cell lymphomas SC Schmechel et al 850 Table 2 Primers used in qRT-PCR analysis

Gene Forward primer Reverse primer

13CDNA73 GATGACGACAGGCCGATGATT TGACCAGGACTGCGTTCCATT APOC2 CCCGCTGTAGATGAGAAACTCA TCTCCCTTCAGCACAGAAAGAA BCL2 ATGACTGAGTACCTGAACCGGC CAGAGACAGCCAGGAGAAATCA CCL4 CCAGCTGTGGTATTCCAAACCA TGAGCAGCTCAGTTCAGTTCCA CCNA2 GCTGGCCTGAATCATTAATACG GCATGCTGTGGTGCTTTGA CCND1 AGGTCTGCGAGGAACAGAAGTG TGCAGGCGGCTCTTTTTCA CD209L TGCTGCAACTCCTCTCCTTCAT CGTCTTGCTCGGATTGTTCCT CD69 CATGGTGCTACTCTTGCTGTCA CCCTGTAACGTTGAACCAGTTG CD86 GGAAAAGACATCAACCCCCATA TCTGGTTGTGGTCTCTGGTGTT CLCN4 GCGGCACTGCAGGTGTAATTA TTCCCTTAGCCAGTCGATGGT CPNE1 CTGCCTCGCAATACTTCATGCT CCACACCCACAATGATCACTGA CRYM GGCAGGTGCAGATGTGATCAT TGGCTCCAACAGCATTGATG CTCF CACACAGGTACTCGTCCTCACA TCGCACATGGAACACTTGAA DNMT3A CCATTCCTGGTCACGCAAAAC TCCTGTGTGGTAGGCACCTGAA FLJ21562 CAGCTGGCTCGATAGTCGTAAA TCTAGGAGGAGCCCAGTCTTCA GM2A AAAAGCCATCCCAGCTCAGTAG CACATTTCCAGGAACGACGAT GS3955 AGGAGCTGGTGTGCAAGGTGTT CCCCATAGCTTCGCTCAAAGAA IGJ TCCCATGGCAAGTCCTAAAGC CCATGACACAGCCAAACAGAAA IGHM GCAGCCGGAGAACAACTACAAG TGCATCACGGAGCATGAGAA IL24 TCTCATCGTGTCACAACTGCAA GAGCTGCTTCTACGTCCAACTG IL4R CAGCGTTTCCTGCATTGTCATC GACCCCTGAGCATCCTGGATTA IMAGE:201981 CCGTCTGTCTCCTTTCCTTCTG TCCTGTCCTCTGCTCTGTGGAT IMAGE:258118 TGCTCCCCTGTTTTTGTGACA TCCTGGAAGTAATGCCAACTCA ITM3 GGAGCTCCTCATGAACGTGAA AGGTGTCTTTCCCGTTGCA JUN CTAACGCAGCAGTTGCAAACA TCTCCGTCGCAACTTGTCAA KIAA0125 ATGGCTCCTGCTGTACCTCAAG GTGAAGCGGTGGACAAGAAACT KIAA1350 CGAAGCTGTTGTTCGGAATC GGCTGGTGTAGCAGATCATACC KIAA1407 AACCTGCCAGATGCTTGTGAAT CGGTGTCATCAATTGCTTTGG KIAA1959 GGTGGATCTGTCAGCTGCCATA GCCTGTCACCTCAGAACTCCAA KLF4 GCTCCATTACCAAGAGCTCATG GTGCCTGGTCAGTTCATCTGA KRT19 GCATGAAAGCTGCCTTGGAA CCTGATTCTGCCGCTCACTATC KRT5 CAGAAGCCGAGTCCTGGTATCA TGGCGCACTGTTTCTTGACA LOC51290 AGCAGAAAGAGTGGCAGAGGAT TTGGTGGAAGAGCTGTTGATGT MGC40441 GGCCTAGAGCCTCTTGATTCAA TTGCTCCTCTCACTCCATGTGT MYBL2 CCCATCAAGAAAGTCCGGAAGT GCAGTTGTCGGCAAGGATAGA RNASE1 TCCACTGCATCATTCAGCTTTC TCTCCAAAGCGAGGTCTTCCT SF1 AGCTCAGAGACCCGCAGCATTA ACTGAGGATCACCAGGCCTTTG SLC2A3 GCCCATCATCATTTCCATTGTG TGAACACCTGCATCCTTGAAGA TFAP2C TCGCAAAGGTCCCATTTCC CGTAGAGCTGAGGAGCGACAAT

fold (P ¼ 0.004) higher than in FL and SLL specimens, data set that were significantly (Po0.05) differentially expressed respectively. Finally, MGC40441 was expressed in RN and FL between tissue types. Genes showing similar patterns of specimens more highly than in MCL and SLL specimens. differential expression in our data set and in publicly available MGC40441 was expressed in RN 8.1-fold (P ¼ 0.0007) and data are shown in Table 3. six-fold (P ¼ 0.0007) higher than in MCL and SLL specimens, respectively; MGC40441 expression in FL was 6.5-fold (0.001) and 4.8-fold (P ¼ 0.002) higher than in MCL and SLL, Discussion respectively. Using cDNA microarray analysis, we have identified 120 genes whose expression patterns differed among RN, FL, MCL, and Comparison with array data obtained by other research SLL. The differential expression patterns of 23 of these genes groups were validated using the complementary approach of qRT-PCR. We found that for over one-third of genes analyzed, gene It is increasingly appreciated that crossvalidated genes, identi- expression patterns obtained using microarrays could not be fied as differentially expressed by independent laboratories verified using qRT-PCR, primarily due to cDNA array spot performing similar experiments, may represent particularly misidentification. The problem of spot misidentification has also useful diagnostic markers and therapeutic targets.34,35 We thus been reported by other investigators, stressing the importance of compared expression data obtained in this study with publicly sequencing verifying cDNA clones as a final step prior to available data obtained by other research groups who have used printing microarray slides.33 arrays to examine SBCL gene expression.9–15 For these cross Among genes identified in this study are genes previously comparisons, we used a feature of GeneSifter.NetTM software known to be differentially expressed in SBCL, genes known to be that allowed us to cross reference genes in other data sets that involved in cancer types other than lymphoma, genes not were identified by names or accession numbers that differed previously associated with malignancy, and partially character- from the IMAGE clone designations in our data set. In these ized genes/ESTs of unknown function. Much work remains to comparisons, we did not set a minimal fold-differential investigate the potential role of these genes in lymphoma expression threshold. Rather, we included all genes from our pathogenesis and their potential diagnostic utility.

Leukemia Table 3 Genes showing similar expression patterns in this study and in published data

IMAGE clone Gene Name Description Ratio derived from publicly Ratio determined in identifier in available dataa–g this study (P-value) published reference

1. Published results comparing FL vs tonsil enriched FL cells vs enriched tonsillar germinal center B cellsa 232714 M14745 BCL2 B-cell CLL/lymphoma 2 Infinity 2.3 (0.00004) 714453 X52425 IL4R Interleukin-4 receptor 21.1 1.9 (0.006) 814636 D26155 SMARCA2 SWI/SNF-related, matrix- 12.1 1.5 (0.05) associated, actin-dependent regulator of chromatin 666377 D28118 ZNF161 Zinc-finger protein 161 7.5 1.6 (0.0005) 731648 M59079 NFYA Nuclear transcription factor Y, 6.1 2.4 (0.000001) alpha 730555 D49547 DNAJB1 DnaJ (Hsp40) homolog, subfmaily 5.3 1.4 (0.003) B, member 1 753234 X59738 ZFX Zinc-finger protein, X-linked 4 1.6 (0.03) 812965 V00568 MYC v-myc myelocytomatosis viral 2.6 1.3 (0.004) oncogene homolog (avian) 305606 M18391 EPHA1 EphA1 2.5 1.5 (0.02) 280882 M93255 FLI1 Friend leukemia virus integration 1 2.5 1.6 (0.05) 810039 D15057 DAD1 Defender against cell death 1 0.4 0.6 (0.03) M26708 M26708 EST Prothymosin alpha 0.3 0.4 (0.02)

898062 U05340 CDC20 CDC20 cell division cycle 20 0.1 0.6 (0.02) lymphomas B-cell Schmechel small SC in analysis expression Gene homolog

2. Published results comparing MCL vs RN microdissected MCL specimens vs hyperplastic lymph nodesb al et 841641 X59798 CCND1 Cyclin D1 410 13.8 (0.002) 258589 X75042 REL v-rel reticuloendotheliosis viral 6 1.4 (0.007) oncogene homolog 815526 X13293 MYBL2 v-myb myeloblastosis viral Not provided (up in MCL) 1.6 (0.02) oncogene homolog-2 854401 Y07512 PRKG1 Protein kinase, cGMP-dependent, o0.2 0.4 (0.03) type I

3. Published results comparing MCL vs tonsil enriched MCL cells vs enriched tonsillar B-cell subpopulationsc 841641 M73554 CCND1 Cyclin D1 137.5 15.4 (0.002) 345616 X76534 GPNMB Glycoprotein transmembrane nmb 18.2 1.7 (0.005) 949938 AI362017 CST3 Cystatin C 16.0 3.6 (0.02) 24642 L35594 ENPP2 Ectonucleotide pyrophosphatase/ 16.0 3.9 (0.02) phosphodiesterase 2 260200 M81750 MNDA Myeloid cell nuclear differentiation 15.4 2.7 (0.01) antigen 840865 D10522 MARCKS Myristoylated alanine-rich protein 11.6 4.4 (0.01) kinase C substrate 232714 M14745 BCL2 B-cell CLL/lymphoma 2 8.1 2.8 (0.0001) Leukemia 851 Leukemia 852

Table 3 Continued

IMAGE clone Gene Name Description Ratio derived from publicly Ratio determined in identifier in available dataa–g this study (P-value) published reference

753301 X16354 CEACAM1 Carcinoembryonic antigen-related 6.1 2.5 (0.01) cell adhesion molecule 1

175727 Y09392 TNFRSF25 Tumor necrosis factor receptor 5.2 2.3 (0.02) lymphomas B-cell small in analysis expression Gene superfamily member 25 204335 L33930 CD24 CD24 antigen 2.9 4.5 (0.0009) 287687 J03171 IFNAR1 Interferon alpha, beta, and omega 0.6 0.4 (0.0007) receptor 1 824547 U18288 MHC2TA MHC class II transactivator 0.5 0.4 (0.001) 823691 U47414 CCNG2 Cyclin G2 0.4 0.7 (0.03) 290230 M91196 ICSBP1 Interferon consensus sequence 0.3 0.5 (0.009) binding protein 1 181831 J04164 MGC27165 Hypothetical protein MGC27165 0.3 0.3 (0.004) 714453 X52425 IL4R Interleukin-4 receptor 0.2 0.3 (0.000002) 812965 V00568 MYC v-myc myelocytomatosis viral 0.1 0.8 (0.04) Schmechel SC oncogene homolog 4. Published results comparing MCL vs tonsil MCL specimens vs purified tonsillar

B-cellsd al et 841641 BCL1 CCND1 Cyclin D1 Up in MCL 15.4 (0.002) 503617 MIG CXCL9 Chemokine (C–X–C motif) ligand 9 Up in MCL 7.6 (0.00001) 713145 CD44H CD44 CD44 antigen Up in MCL 1.8 (0.003) 140574 CX3C CX3CL1 Chemokine (C–X3–C motif) ligand Up in MCL 1.4 (0.04) 1 742143 CD5 CD5 CD5 antigen Up in MCL 1.2 (0.005) 302591 ARHH ARHH Ras homolog gene family member Down in MCL 0.8 (0.03) H 344589 LCP1 LCP1 Lymphocyte cytosolic protein 1 Down in MCL 0.5 (0.009) 824547 CIITA MHC2TA MHC class II transactivator Down in MCL 0.4 (0.001) 5. Published results comparing MCL vs CLL/SLL MCL specimens vs CLL/SLLe 898218 IGF binding IGFBP3 Insulin-like growth factor binding 4.8 1.5 (0.04) prot. 3 protein 3 840865 MARCKS MARCKS Myristoylated alanine-rich protein 4.3 2.8 (0.02) kinase C substrate 306013 X07203 MS4A1 Membrane spanning four domains 4.2 3.0 (0.002) subfamily A member 1 753301 CEA-related CEACAM1 Carcinoembryonic antigen-related 3.4 1.9 (0.03) adhesion 1 cell adhesion molecule 1 119914 AA211803 EST Clone MGC:9515 2.7 1.5 (0.004) 204335 CD24 CD24 CD24 antigen 2.6 2.2 (0.01) 305302 H61552 C3orf6 Chromosome 3 open-reading 2.6 2.9 (0.02) frame 6 6. Published results comparing CLL/SLL vs tonsil CLL/SLL vs a mixture of normal human tissuesf 714453 X52425 IL4R Interleukin-4 receptor 7 2.5 (0.0009) Gene expression analysis in small B-cell lymphomas SC Schmechel et al 853 We validated the differential expression of several genes FL previously known to be involved in SBCLs. Our finding that vs CCND1 was significantly overexpressed in MCL was not

ession in FL surprising since the overexpression of this cell cycle regulatory 36,37 on in CLL/SLL protein is important in MCL pathogenesis We found that

CLL/SLL BCL2 was significantly overexpressed in SLL relative to RN. this study (P-value) Ratio determined in Since the antiapoptotic BCL2 protein is known to be over- expressed in both FL and SLL,38,39 we had expected that BCL2 would also be overexpressed in FL relative to RN and other specimen types. Closer inspection of our array data revealed that BCL2 was 2.3-fold overexpressed in FL relative to RN (P ¼ 0.002) and 1.9-fold overexpressed in FL relative to MCL

a–g (P ¼ 0.001). BCL2 was also expressed at a 2.3-fold higher level in SLL than in MCL (P ¼ 0.01). Although these differential 5.3 3.1 (0.02) expression levels fell below our arbitrarily chosen four-fold stringency threshold, our results agree with the known role of BCL2 in FL and SLL. We found that c-jun was significantly available data overexpressed in MCL relative to FL, RN, and SLL. c-Jun is an AP-1 transcription factor component known to be overexpressed in the malignant Reed–Sternberg cells that characterize Hodg- kin’s disease.40 Lastly, we found that the EST MGC40441, which has also been alternatively referred to as HGAL21 and GCET2,41 was overexpressed in RN and FL relative to MCL and CLL. These data complement recent reports that this gene product is relatively overexpressed in FL, purified germinal center B cells, and germinal center B-cell-like (GCB) DLBCL specimens, relative to SLL and MCL specimens.21,41 Sequence analysis revealed that MGC40441/HGAL/GCET2 shares 51% identity with the murine germinal center-expressed M17 gene.21,42 Interestingly, Lossos et al21 found that overexpression of this single gene product appears to identify the favorable prognosis GCB subset of DLBCLs.18,21 Several of the genes that we identified have well-established

chemoattractant factor) roles in cancer types other than lymphoma. We found that c-myb was significantly underexpressed in MCL vs SLL. c-Myb is a member of the myb family of transcription factors that regulates the proliferation, differentiation, and apoptosis of hematopoietic cells and is frequently overexpressed in human myeloid and lymphoid leukemias.43 To our knowledge, c-myb expression in SBCL lymphomas has not been studied previously. We found that DNA methyltransferase 3A (DNMT3A) was significantly overexpressed in MCL relative to FL and CLL. Name Description Ratio derived from publicly DNMT3A and DNMT3B are thought to establish cytosine methylation patterns that influence the expression of genes containing upstream CpG islands.44 DNA from malignant cells often shows global hypomethylation but localized CpG island hypermethylation resulting in the downregulated expression of tumor suppressor genes.45 DNMT overexpression may contri- bute to altered DNA methylation patterns in cancer, and CLL cells were recently shown to have increased DNMT3A identifier in published reference expression relative to normal lymphocytes.45 However, no published studies have compared methylation patterns between FL, MCL, and SLL. The expression patterns of several of the genes that we ratios were determined from publicly available data by calculating the ratio of the median expression in CLL/SLL specimens divided by the median expr ratios were determined from publicly available data by calculating the ratio of the median expression in MCL specimens divided by the median expressi identified have been studied in cancer, but have no known role 14 13 ; ;

10 in carcinogenesis. We found that CD69 is overexpressed in MCL . 15 . 9 .

et al relative to RN. CD69 is expressed by activated T lymphocytes as et al 46 et al 11 et al

. well as by malignant FL, MCL, and SLL cells. However, a role et al 12 . for CD69 in lymphoma is unclear and our findings may be et al et al g explained by the expression of this gene product in T cells FL Continued contained within MCL specimens. Finally, a large number of

vs identified genes are ESTs of completely unknown function. A large amount of work remains to investigate the biological role and potential diagnostic utility of genes identified in the course 179890809776 J03571 U82972 ALOX5 IL16 Arachidonate 5-lipoxygenase Interleukin-16 (lymphocyte 6.4 2.7 (0.03) 742143232714 CD5 BCL-2 CD5 BCL2 B-cell CD5 CLL/lymphoma antigen 2 3.5 6.3 1.9 (0.02) 1.6 (0.0006) Reference Hofmann Reference Zhu Reference Rosenwald Reference Husson Reference Ek Reference Rosenwald Reference Stratowa Table 3 IMAGE clone Gene g b c d e a f 7. Published results comparing CLL/SLL specimens. specimens. of this study.

Leukemia Gene expression analysis in small B-cell lymphomas SC Schmechel et al 854 Acknowledgements chemotherapy for diffuse large-B-cell lymphoma. N Engl J Med 2002; 346: 1937–1947. This work was supported by Washington Technology Center grant 19 Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS, Rosenwald A number F01-B10 and by research funding from RationalDiagnos- et al. Distinct types of diffuse large B-cell lymphoma identified by gene expression profiling. Nature 2000; 403: 503–511. tics, Inc. to DES. SCS was supported in part by NIH training Grant 20 Shipp MA, Ross KN, Tamayo P, Weng AP, Kutok JL, Aguiar RC 5T32HL007312-25. et al. Diffuse large B-cell lymphoma outcome prediction by gene- expression profiling and supervised machine learning. Nat Med References 2002; 8: 68–74. 21 Lossos IS, Alizadeh AA, Rajapaksa R, Tibshirani R, Levy R. HGAL is a novel interleukin-4-inducible gene that strongly predicts 1 Jemal A, Thomas A, Murray T, Thun M. Cancer statistics, 2002. CA survival in diffuse large B-cell lymphoma. Blood 2003; 101: Cancer J Clin 2002; 52: 23–47. 433–440. 2 Ries L, Miller B, Hankey BE. SEER Cancer Statistics Review, 1973– 22 Thieblemont C, Chettab K, Felman P, Callet-Bauchu E, 1991: Tables and Graphs. Bethesda, MD: National Cancer Traverse-Glehen A, Berger F et al. Identification and validation Institute, 1994, Report No.: NIH Pub. No. 94-2789. of seven genes, as potential markers, for the differential diagnosis 3 Voliotis D, Diehl V. Challenges in treating hematologic malig- of small B cell lymphomas (small lymphocytic lymphoma, nancies. Semin Oncol 2002; 29 (Suppl 8): 30–39. marginal zone B cell lymphoma and mantle cell lymphoma) by 4 Horning SJ. Follicular lymphoma: have we made any progress? cDNA macroarrays analysis. Leukemia 2002; 16: 2326–2329. Ann Oncol 2000; 11 (Suppl 1): 23–27. 23 Sambrook J, Fritsche E, Maniatis T. Molecular Cloning: A 5 Harris NL, Jaffe ES, Stein H, Banks PM, Chan JK, Cleary ML et al. A Laboratory Manual, 2nd edn Plainview, NY: Cold Spring Harbor revised European–American classification of lymphoid neoplasms: Laboratory Press, 1989. a proposal from the International Lymphoma Study Group. Blood 24 Geiss GK, An MC, Bumgarner RE, Hammersmark E, Cunningham 1994; 84: 1361–1392. D, Katze MG. Global impact of influenza virus on cellular 6 Harris NL, Jaffe ES, Diebold J, Flandrin G, Muller-Hermelink HK, Vardiman J et al. World Health Organization classification of pathways is mediated by both replication-dependent and neoplastic diseases of the hematopoietic and lymphoid tissues: -independent events. J Virol 2001; 75: 4321–4331. report of the Clinical Advisory Committee Meeting – Airlie House, 25 Lennon G, Auffray C, Polymeropoulos M, Soares MB. The IMAGE Virginia, November 1997. J Clin Oncol 1999; 17: 3835–3849. Consortium: an integrated molecular analysis of genomes and their 7 Harris NL, Jaffe ES, Diebold J, Flandrin G, Muller-Hermelink HK, expression. Genomics 1996; 33: 151–152. Vardiman J. Lymphoma classification – from controversy to 26 Kenworthy AK. Imaging protein–protein interactions using fluor- consensus: the REAL and WHO classification of lymphoid escence resonance energy transfer microscopy. Methods 2001; 24: neoplasms. Ann Oncol 2000; 11 (Suppl 1): 3–10. 289–296. 8 Schena M, Shalon D, Davis RW, Brown PO. Quantitative 27 Diehn M, Sherlock G, Binkley G, Jin H, Matese JC, Hernandez- monitoring of gene expression patterns with a complementary Boussard T et al. SOURCE: a unified genomic resource of DNA microarray. Science 1995; 270: 467–470. functional annotations, ontologies, and gene expression data. 9 Husson H, Carideo EG, Neuberg D, Schultze J, Munoz O, Marks Nucleic Acids Res 2003; 31: 219–223. PW et al. Gene expression profiling of follicular lymphoma and 28 Hubbard T, Barker D, Birney E, Cameron G, Chen Y, Clark L et al. normal germinal center B cells using cDNA arrays. Blood 2002; The Ensembl genome database project. Nucleic Acids Res 2002; 99: 282–289. 30: 38–41. 10 Hofmann WK, de Vos S, Tsukasaki K, Wachsman W, Pinkus GS, 29 Eisen M, Spellman P, Brown P, Botstein D. Cluster analysis and Said JW et al. Altered apoptosis pathways in mantle cell lymphoma display of genome-wide expression patterns. Proc Natl Acad Sci detected by oligonucleotide microarray. Blood 2001; 98: 787– USA 1998; 95: 14863–14868. 794. 30 Bjarnason R, Wickelgren R, Hermansson M, Hammarqvist F, 11 Zhu Y, Hollmen J, Raty R, Aalto Y, Nagy B, Elonen E et al. Carlsson B, Carlsson LM. Growth hormone treatment prevents the Investigatory and analytical approaches to differential gene decrease in insulin-like growth factor I gene expression in patients expression profiling in mantle cell lymphoma. Br J Haematol undergoing abdominal surgery. J Clin Endocrinol Metab 1998; 83: 2002; 119: 905–915. 1566–1572. 12 Ek S, Hogerkorp CM, Dictor M, Ehinger M, Borrebaeck CA. Mantle 31 Agrawal D, Chen T, Irby R, Quackenbush J, Chambers AF, Szabo cell lymphomas express a distinct genetic signature affecting M et al. Osteopontin identified as lead marker of colon cancer lymphocyte trafficking and growth regulation as compared with progression, using pooled sample expression profiling. J Natl subpopulations of normal human B cells. Cancer Res 2002; 62: Cancer Inst 2002; 94: 513–521. 4398–4405. 32 Rajeevan MS, Vernon SD, Taysavang N, Unger ER. Validation of 13 Rosenwald A, Wright G, Wiestner A, Chan WC, Connors JM, array-based gene expression profiles by real-time (kinetic) RT-PCR. Campo E et al. The proliferation gene expression signature is a J Mol Diagn 2001; 3: 26–31. quantitative integrator of oncogenic events that predicts survival in 33 Taylor E, Cogdell D, Coombes K, Hu L, Ramdas L, Tabor A et al. mantle cell lymphoma. Cancer Cell 2003; 3: 185–197. Sequence verification as quality-control step for production of 14 Rosenwald A, Alizadeh AA, Widhopf G, Simon R, Davis RE, Yu X cDNA microarrays. Biotechniques 2001; 31: 62–65. et al. Relation of gene expression phenotype to immunoglobulin 34 Rhodes DR, Barrette TR, Rubin MA, Ghosh D, Chinnaiyan AM. mutation genotype in B cell chronic lymphocytic leukemia. J Exp Meta-analysis of microarrays: interstudy validation of gene Med 2001; 194: 1639–1647. expression profiles reveals pathway dysregulation in prostate 15 Stratowa C, Loffler G, Lichter P, Stilgenbauer S, Haberl P, cancer. Cancer Res 2002; 62: 4427–4433. Schweifer N et al. cDNA microarray gene expression analysis of 35 Nelson PS. Identifying immunotherapeutic targets for prostate B-cell chronic lymphocytic leukemia proposes potential new carcinoma through the analysis of gene expression profiles. Ann prognostic markers involved in lymphocyte trafficking. Int J Cancer NY Acad Sci 2002; 975: 232–246. 2001; 91: 474–480. 36 Tsujimoto Y, Yunis J, Onorato-Showe L, Erikson J, Nowell PC, 16 Klein U, Tu Y, Stolovitzky GA, Mattioli M, Cattoretti G, Husson H Croce CM. Molecular cloning of the chromosomal breakpoint of et al. Gene expression profiling of B cell chronic lymphocytic B-cell lymphomas and leukemias with the t(11;14) chromosome leukemia reveals a homogeneous phenotype related to memory B translocation. Science 1984; 224: 1403–1406. cells. J Exp Med 2001; 194: 1625–1638. 37 Tsujimoto Y, Jaffe E, Cossman J, Gorham J, Nowell PC, Croce CM. 17 Davis RE, Brown KD, Siebenlist U, Staudt LM. Constitutive nuclear Clustering of breakpoints on chromosome 11 in human B-cell factor kappaB activity is required for survival of activated B cell- neoplasms with the t(11;14) chromosome translocation. Nature like diffuse large B cell lymphoma cells. J Exp Med 2001; 194: 1985; 315: 340–343. 1861–1874. 38 Hockenbery D, Nunez G, Milliman C, Schreiber RD, Korsmeyer 18 Rosenwald A, Wright G, Chan WC, Connors JM, Campo E, Fisher SJ. Bcl-2 is an inner mitochondrial membrane protein that blocks RI et al. The use of molecular profiling to predict survival after programmed cell death. Nature 1990; 348: 334–336.

Leukemia Gene expression analysis in small B-cell lymphomas SC Schmechel et al 855 39 Vaux DL, Cory S, Adams JM. Bcl-2 gene promotes haemopoietic 42 Christoph T, Rickert R, Rajewsky K. M17: a novel gene expressed cell survival and cooperates with c-myc to immortalize pre-B cells. in germinal centers. Int Immunol 1994; 6: 1203–1211. Nature 1988; 335: 440–442. 43 Davies J, Badiani P, Weston K. Cooperation of Myb and Myc 40 Mathas S, Hinz M, Anagnostopoulos I, Krappmann D, in T cell lymphomagenesis. Oncogene 1999; 18: Lietz A, Jundt F et al. Aberrantly expressed c-Jun and JunB 3643–3647. are a hallmark of Hodgkin lymphoma cells, stimulate prolifera- 44 Jones PA, Laird PW. Cancer epigenetics comes of age. Nat Genet tion and synergize with NF-kappa B. EMBO J 2002; 21: 1999; 21: 163–167. 4104–4113. 45 Melki JR, Clark SJ. DNA methylation changes in leukaemia. Semin 41 Pan Z, Shen Y, Du C, Zhou G, Rosenwald A, Staudt LM et al. Two Cancer Biol 2002; 12: 347–357. newly characterized germinal center B-cell-associated genes, 46 Erlanson M, Gronlund E, Lofvenberg E, Roos G, Lindh J. Expression GCET1 and GCET2, have differential expression in normal and of activation markers CD23 and CD69 in B-cell non-Hodgkin’s neoplastic B cells. Am J Pathol 2003; 163: 135–144. lymphoma. Eur J Haematol 1998; 60: 125–132.

Leukemia