Introduction
The pathological category of low-grade B cell non-Hodgkin’s lymphoma (NHL) encompasses mature B cell neoplasms with small to medium-sized cells such as follicular lymphoma (FL), marginal zone lymphoma (MZL), B cell chronic lymphocytic leukemia/small lymphocytic lymphoma (B-CLL/SLL), and mantle cell lymphoma (MCL). The histological transformation of low-grade lymphomas is thought to have a major impact on their prognosis, however, the clonal relationship between the two neoplasm types and the pathogenesis underlying the progression of the disease are still controversial.
MCL is considered a well categorized B cell NHL because virtually all cases feature the chromosomal translocation t(11;14)(q13;q32) (IgH/CCND1) which leads to cyclin D1 (CCND1) overexpression. Despite this strong genetic hallmark, MCL presents in a broad spectrum of morphological forms with a high number of additional chromosomal abnormalities. From a practical point of view, we can detect the mantle zone growth pattern, which represents the initial component of MCL, in only 8.6% of cases (1). Aggressive transformation in MCL is more notable, occurring in 32% of living cases and in 70% of autopsy cases (2). Hence, the presence of pathological evidence such as characteristic cytologies (small, classical, blastic and pleomorphic) or proliferating patterns (mantle-zone, nodular and diffuse) at the time of biopsy seems to depend on temporal changes in the accumulation of molecular genetic alterations.
For an accurate understanding of MCL, there are some questions requiring resolution. First, why is this lymphoma characterized by biological heterogeneity despite the characteristic translocation event? Second, what is required for the initial pathogenesis of MCL in addition to the t(11;14) translocation? A recent study has shown that the t(11;14) translocation alone is not sufficient to produce tumors (3) and, though at very low levels, t(11;14)-positive cells have been found in the blood of healthy individuals (4). Thus, key molecular behavior in MCL is thought to be divided into two major stages: initial lymphomagenesis in addition to the t(11;14) translocation (initiation) and the accumulation of variable secondary genomic alterations occurring over time, leading to the evolution into more aggressive forms (transformation).
Recently, microarray profiling studies have identified differential expression of several genes in the progression of MCL (5-7). Some of these studies have used purified mantle zone B lymphocytes sorted from reactive tonsillectomy specimens for gene expression profiling to avoid contamination with stromal cells, T cells and macrophages (6,7). In these studies, several genes identified as overexpressed in aggressive form are known to be involved in cell cycle control or apoptotic cell death, and several genes related to the PI3K/AKT, WNT and TGF-β signaling pathways are reportedly important in the pathogenesis of MCL. Although morphological heterogeneity remains a central issue in the prognosis and treatment of this lymphoma, no gene expression profiling studies to date have attempted to identify the genes and signaling pathways involved in the each event separately. The present study was designed based on the concept that gene expression profiling of morphologically heterogeneous MCL samples would provide insight into the role of aberrant gene expression for both initial lymphomagenesis and transformation events.
Materials and methods
Patients and tissue samples
We performed cDNA microarray experiments using frozen tissues of 19 lymph node biopsies. A total of 15 MCLs were collected from the files of the Departments of Pathology of Kurume University (Fukuoka, Japan), which include a total of 237 MCL patients (1) confirmed by histology as CCND1-positive (Table I). All 15 cases were subjected to cytogenetic and/or FISH studies and found positive for the IgH/CCND1. This study was approved by the Kurume University Institutional Review Board, and patients provided informed consent in accordance with the Declaration of Helsinki.
In order to identify differentially expressed genes and their contribution to each event, we posited four stepwise morphological grades for MCL: MCL in situ, MCL with classical form (cMCL), MCL with aggressive form (aMCL), and MCL with intermediate morphology between classical and aggressive forms at the same site (iMCL) (Fig. 1). Namely, MCL in situ is defined as those samples with a very thin neoplastic mantle zone growth pattern and very little or no spreading of tumor cells into interfollicular areas. cMCL is characterized by a prominent nodular proliferation of atypical medium-sized tumor cells (classical form) without diffuse proliferation. aMCL is designated as a combination of two morphological forms: blastoid and pleomorphic. iMCL contains distinct areas of both classical and aggressive forms.
For evaluation of initiation, we compared samples from the tumor cells of MCL in situ (n=4) with those from normal mantle zone B-lymphocytes from benign lymphadenitis (n=4). Samples were derived from selected specimens by means of laser microdissection (LMD). We hypothesize that transformation into aMCL is a multistep process, similar to the progression of carcinomas. It may be beneficial to subdivide aMCL neoplasms into two major subtypes: ‘aMCL with genetic stepwise process’ which results from the accumulation of many molecular genetic alterations and ‘de novo aMCL’ resulting from fewer but stronger genetic alterations. Therefore, comparing cMCL with aMCL is not sufficient for the detection of transformation specific genes because it is not possible to distinguish them morphologically. Alternatively, we expected that using results from the gene selection method in combination with that of iMCL cases (n=2) would provide the candidate genes restricted within aMCL having genetic stepwise process. Additionally, using iMCL samples as a discovery set can compensate for individual differences between cases. For a concrete explanation of transformation, total RNA from whole tumor tissue samples, cMCL (n=4) and aMCL (n=5) were used [Fig. 2(I)]. We also compared classical areas with aggressive areas in iMCL obtained by LMD [Fig. 2(II)]. Finally, we selected the overlapping genes differentially expressed in both comparisons [Fig. 2(III)].
Laser microdissection (LMD)
The tissue samples were immediately frozen in acetone/dry ice and stored at -80°C for microdissection. The lymph node samples were embedded in an optical cutting temperature (OCT) compound (Sakura Finetek, Tokyo, Japan) and frozen in liquid nitrogen. Cryosections (10-μm-thick) were mounted on 2.0-μm-thick PEN-Membrane slides (MicroDissect GmbH, Herborn, Germany). After fixation in 100% ethanol, the slides were stained rapidly with Toluidine Blue O (Chroma-Gesellschaft Schmid GmbH & Co., Köngen, Germany) and then washed in DEPC-treated water and air-dried with a fan. The frozen sections were microdissected with a Leica LMD6000 laser microdissection system by following the company’s protocol (Leica, Wetzlar, Germany). The sorting regions were micro-dissected from the tissue sections with LMD (Fig. 3), and the dissected cells were collected in 0.5 ml tubes filled with 50 μl lysis buffer for RNA extraction.
RNA extraction and biotinylated cRNA amplification
Total RNA was extracted from the LMD-obtained samples with an RNAqueous-Micro kit (Ambion, Austin, TX, USA) according to the manufacturer’s instructions. For cMCL and aMCL tissues, total RNA was isolated using TRIzol reagent (Invitrogen, Carlsbad, CA, USA). RNA samples were quantified by an ND-1000 spectrophotometer (NanoDrop Technologies, Wilmington, DE, USA) and the quality was confirmed with an Experion System (Bio-Rad Laboratories, Hercules, CA, USA). Both cRNA amplification and labeling with biotin were used to prepare samples for gene expression profiling by microarray analysis. Briefly, 350-500 ng total RNA was amplified overnight (14 h) with the Illumina Total Prep RNA Amplification kit (Ambion) in accordance with the manufacturer’s protocol. Reaction cRNA was biotinylated during in vitro transcription.
Illumina BeadChips microarray
Sentrix Human WG-6 v3.0 Expression BeadChips were purchased from Illumina, Inc. (San Diego, CA, USA). More than 48,000 different bead types, each with a 50-base gene-specific probe, are represented on a single Beadchip. For each probe represented on the array, beads are assembled with an average 30-fold redundancy. A hybridization mixture containing 1.5 μg biotinylated cRNA was hybridized to the beadchips at 58°C overnight (18 h) before being washed and stained with streptavidin-Cy3 (GE Healthcare, Buckinghamshire, UK) according to the manufacturer’s protocol. Beadchips were scanned on Illumina BeadStation 500 and fluorescent hybridization signals were assessed with Illumina BeadStudio software.
Immunohistochemistry
The samples were also evaluated for expression of markers such as Wnt3 (Abcam Inc., Cambridge, MA, USA), phosphorylated-β-catenin (Ser552) (pβ-catenin-S552) (Cell Signaling, Boston, MA, USA), and Ki-67 (Dako Cytomation, Glostrup, Denmark). Formalin-fixed, paraffin-embedded tissues were used for all immunohistochemical stains. Antibody dilutions and antigen retrieval procedures were performed as standards.
Data analysis and filter criteria
Pre-processing was performed on the raw signal intensities of all samples by log2-transformation and normalization using the quantile algorithm of the ‘preprocess Core’ library package (8) in Bioconductor (9). We selected the probes (excluding the control probes) where the detection p-value was <0.01 in all samples, and used only these probes to identify differentially expressed genes. For the samples meeting the criteria of initiation (MCL in situ vs. normal mantle zone B lymphocytes) and transformation (I) (cMCL vs. aMCL), we applied Linear Models for Microarray Analysis (limma) package (10) of Bioconductor. We used a cutoff of limma p-value <0.05 and absolute log-fold change (|logFC|) >0.5 to assess differentially expressed genes in each comparison. For the transformation (II) comparison (classical area vs. aggressive area in iMCL), we selected probes that contained a ‘P’ flag in both iMCL samples. To identify up or downregulated genes between aggressive and classical areas, we calculated Z-scores (11) and ratios (non-log scaled fold change) from the normalized signal intensities of each probe. Then we established criteria for regulated genes: absolute Z-score (|Z|) >2. Ingenuity Pathway Analysis (IPA6.0; Ingenuity Systems, Redwood, CA, USA; http://www.ingenuity.com) was used to identify networks of interacting genes. Lists of expressed (up and downregulated) genes were uploaded for IPA. A heat map of the differentially expressed genes was generated by MeV software (12). Array data are available on the Gene Expression Omnibus (GEO) website under accession numbers GSE30189 (http://www.ncbi.nlm.nih.gov/geo/).
Results
Differentially expressed genes in initiaion
In the analysis of MCL initiation, we selected 1,538 genes (851 upregulated and 787 downregulated) which showed significant differences (p<0.05) between normal mantle zone B lymphocytes and MCL in situ samples (http://www.ncbi.nlm.nih.gov/geo/). The identification of CCND1 as the most significantly upregulated gene in ‘initiation’ samples gives us confidence that these array experiments are measuring biologically relevant differences between the sample types. The top 10 up- and down-regulated genes in MCL in situ were: up, CCND1, FCGBP, IL17RB, NULL (probe ID; 5390192), WNT3, D4S234E, FBLN2, CPXM1, DBN1 and TEAD2; down, PLAC8, LOC439949, HVCN1, FGR, LOC651751, IL7R, DEF8, TXNDC8, GZMB and SLAMF1.
For further analysis, we focused on the canonical β-catenin-dependent Wnt pathway (WCP), which is critically involved in cell fate and diffrerentiation (13). As reasons for this, IPA analysis revealed that a significant number of differentially expressed genes (n=23) belong to this category (p=0.016), and none of the genes in this pathway were observed as differentially expressed in the transformation experiments (Table II). Whether the cytosolic pool of β-catenin participates in WCP signaling is dictated by the availability of its binding partners, and these binding interactions are regulated by phosphorylation (14). Phosphorylation of β-catenin at Ser552 by AKT can enhance β-catenin/TCF reporter activation, suggesting that pβ-catenin-S552 is a nuclear-localized form of β-catenin (active form) (15). Microarray results indicate that β-catenin was not significantly upregulated in MCL initiation, but immunohistochemical results revealed that the tumor cells of the MCL in situ samples showed nuclear localization of pβ-catenin-S552 with high levels of cytoplasmic Wnt3 staining (Fig. 4a-d). On the other hand, reactive mantle zone B cells were negative for pβ-catenin-S552 in the nuclear, and cytoplasmic Wnt3 staining was weak (Fig. 4e-h).
Differentially expressed genes in transformation
With the Illumina BeadStudio software, we detected 710 genes that showed significant differences in expression between cMCL and aMCL groups [Fig. 2(I)] (http://www.ncbi.nlm.nih.gov/geo/). Gene expression profiling in iMCL showed that 220 genes were differently expressed between microdissected classical area and aggressive area samples [Fig. 2(II)] (http://www.ncbi.nlm.nih.gov/geo/). Finally, we identified 60 overlapping genes [Fig. 2(III)] (Table III), both upregulated (n=58) and down-regulated (n=2), which showed significant differences in pre- vs. post-transformation with a p-value of <0.05. These genes are visualized as a heat map in Fig. 5.
As may be expected, IPA analysis of the filtered 60 genes revealed that most of these genes (42/60; 70%) were classified in the following categories, all of which are thought to be involved in transformation: 36 genes belonged to the category of ‘Cell cycle progression’ (UBE2C, BIRC5, CDCA5, TYMS, KIFC1, TOP2A, FOXM1, CCNF, E2F2, KIF2C, TPX2, KIF11, CDC2, CENPF, NEK2, CDKN3, PTTG1, DLGAP5, CENPE, CDC25A, AURKA, CCNB2, ASPM, KIF20A, UHRF1, CEP55, NCAPG, MKI67, KIF23, H2AFX, GPSM2); 28 ‘DNA replication, recombination and repair’ genes (AURKA, BIRC5, CENPE, DLGAP5, KIF2C, KIFC1, NCAPG, TOP2A, CCNB2, GPSM2, NEK2, PTTG1, KIF11, TPX2, CDCA5, CDK1, CDC25A, E2F2, FEN1, FOXM1, HMGB2, KIAA0101, H2AFX, POLQ, TYMS, KIF23, CENPA, UHRF1); 10 ‘cell death’ genes (A4GALT, AURKA, BIRC5, CCL2, CDC25A, CDCA2, CDK1, E2F2, NEK2, PTTG1, TOP2A, TYMS, UBE2C, CENPF, FEN1, FOXM1, HMGB2, KIF2C). As seen in Fig. 6, these genes recapitulate much of the p53 interaction network, a network centrally involved in cell cycle progression. Most of these genes (AURKA, BIRC5, CDC2, CDC25A, CENPF, CENPE, CCNB2, FOXM1, NEK2, PTTG1, TPX2, ASPM, TOP2A, DLGAP5, KIF2C, KIF23, UBE2C) were considered important players in MCL transformation because of their known function as mitotic regulators, which mediate cell cycle progression during the G2/M transition. CDC2, FOXM1 and BIRC5 in particular interact with many other highly expressed genes in aMCL, which suggests that they may play a critical role in transformational events along with p53. Of these transformation-associated genes, only CDCA5 and MKI67 were also significantly changed in initiation, while the remaining 58 genes were aberrantly expressed specifically in transformation.
Discussion
Biological heterogeneity has become an important position in the current understanding of MCL. To meet this need, we performed cDNA microarray experiments stratifying MCL samples into four morphological grades based on the files including a large series of patients. In order to investigate the molecular mechanisms underlying these differences, we compared initial component samples or transformational morphology samples and identified genes associated with the Wnt signaling pathway and several known mitotic regulators as differentially regulated.
Cell cycle alterations resulting in unscheduled proliferation are strongly associated with the evolution of malignant tumors. Most of these changes alter pathways involved in G1 progression or the G1/S transition (16). With regard to MCL, the ‘Lymphoma and Leukemia Molecular Profiling Project (LLMPP)’ demonstrated that the length of survival of MCL patients depends upon quantitative differences in progression from G1/S phase of the cell cycle (17). Sander et al also showed that a subset of MCL tumors with low levels of the long CCND1 transcript is highly proliferative and some of its related genes have homology to the group of cell cycle (G1/S) promoting E2F transcription partners (18). Indeed, there is a strong correlation between these reported genes and the 60 we identified in this study as involved in transformation (Table III), which emphasizes the importance of G1/S regulators in MCL evolution. The morphological classifications used in this study were based on our previous findings that MCL presents in three morphological evolutions: classical, intermediate and aggressive, and that the aggressive form is considered an important clinical signal of poor prognosis (1). Therefore, it might be natural to expect that dysregulation of G1/S proliferation signature genes is necessary to determine both the length of MCL patient survival and the time to neoplasm transformation.
Although the dysregulation of progression through mitosis does not directly promote proliferation, a few centrosomal and mitotic proteins (such as AURKA, PLK1 and PTTG1) have been reported to act as oncogenes (19). Interestingly, our results highlight a large number of mitotic regulators (CDC2, AURKA, BIRC5, CDC25A, CENPE, CCNB2, FOXM1, NEK2, PTTG1, TPX2, DLGAP5, KIF2C, KIF23 and UBE2C) in addition to G1/S regulators (CDC25A, E2F2, FOXM1 and TYMS). Of these genes, CDC2, FOXM1 and BIRC5 interact with many other highly expressed mitotic regulators within the p53 interaction network, suggesting that specific mitotic regulators facilitate the transformation of MCL into more aggressive form. In support of this hypothesis, Blenk et al identified significantly differentially expressed genes between samples with good and poor prognosis in MCL using exploratory analysis of gene expression values and CGH data (20). Surprisingly, the majority of the genes highlighted in that study (BIRC5, ASPM, MKI67, UHRF1, CDC2, CENPF and KIF23), including their survival predictor of genes for MCL patients (CENPE, CDC2, BIRC5 and ASPM), overlap with our current findings (Table III). These observations imply that both CDC2 and BIRC5 expression play central roles in MCL transformation and thus reflect prognosis.
CDC2 (also known as CDK1) is one of the master regulators of mitosis, as it is involved in the centrosome cycle and early mitotic events. Hui et al reported that elevated protein levels of CDC2 are correlated with the expression of proliferation marker, and represent a useful and simple method in evaluating the prognosis of MCL patients (21). Although BIRC5 (also known as survivin) is known as a member of inhibitor of apoptosis proteins gene family (22), several studies have shown that its role in cancer is not limited to apoptosis inhibition (23). Accordingly, BIRC5 expression levels are higher in aMCL samples than cMCL samples and are associated with proliferative activity and survival of the patients (24).
The role of FOXM1 in lymphoma has not been reported in the literature. However, FOXM1 was also significantly upregulated genes in MCL transformation (FC; 1.01 and p<0.005) in our results. FOXM1 is a typical proliferation-associated transcription factor, and is intimately involved in mitosis regulation (25). Namely, FOXM1 directly or indirectly (via MYC) regulates genes that control G1/S transition, S-phase progression, G2/M transition and M-phase progression (25), supporting that it is involved in tumorigenesis. It is of interest that its expression level is known to be increased in several other tumor grades, such as prostate carcinoma and glioblastoma (26,27). Thus, all of these genes, as well as the other mitotic regulators we found to have differential expression in MCL transformation, are potentially attractive therapeutic targets or strong diagnostic tools.
Alterations in the DNA damage response pathway and mitotic checkpoints through p53 are major additional genetic events in MCL, as indicated by the high rate of tetraploidy found in aMCL (28). In fact, p53 inactivation in MCL as a consequence of deletion or mutation occurs more frequently in the aggressive form than in the classic form (29). However, inactivating mutations of p53 are found in only 38% of aMCL cases (30,31), suggesting that several other genetic alterations also contribute to inactivation of the p53 pathway. Currently, ATM deletions are thought to be strongly associated with the dysregulation of the DNA damage response pathway through p53, and are probably present in the early phase of MCL (17,32-34). Our microarray experiments have revealed that ATM was significantly downregulated in initiation with strong fold change (FC=-0.73), while there was no significant change observed in transformation. In the results of our previous study using immunohistochemistry, higher cell positivity of p53 (DO7) was observed in iMCL than in cMCL, while there was no significant difference between iMCL and aMCL (1). This is due to the fact that there is a high concordance between p53 (DO7) nuclear overexpression and gene mutation in human carcinomas (35). These results strongly indicated that dysregulation of the ATM/p53 pathway in MCL would occur at undetectable levels as a relatively early phenomenon. After that, the disruption of central mitotic regulators (CDC2, BIRC5 and FOXM1) is responsible for the induction of chromosomal instability. Combined, these factors play an important pathogenic role in the evolution of MCL, perturbing the regulation of tumor cell cycling at the G2/M transition.
The mechanisms of the initial lymphomagenesis of MCL in addition to the IgH/CCND1 remain unclear, and new approaches are urgently needed to elucidate which genes and signaling pathways contribute to this event. Our results support the hypothesis that aberrant Wnt3 signaling is required for the MCL lymphomagenesis, because a significant number of WCP associated genes were aberrantly expressed in the initiation and not significantly changed in the transformation. Moreover, immunohistochemical findings revealed the special activation of WCP.
The significance of WCP signaling in tumor initiation may be straightforward from the view point of the adenoma-carcinoma sequence (36). Deletion of the APC gene resulting in the activation of WCP is a consistent finding among the earliest events in both de novo and sporadic colon carcinomas. Also, several studies support the notion that WCP can influence both lymphopoiesis (37,38) and hematological malignancies (39). In our initiation experiments, especially Wnt3 and LRP5 were genes with a high average increase (2.65- and 1.04-fold, respectively), and are reportedly highly expressed in MCL cases (40).
The binding of Wnt proteins to their respective cell surface receptors, including seven of the transmembrane fizzled (Fz) receptors and low-density lipoprotein receptor-related protein (LRP5 or 6), activates disheveled (Dvl). Activated Dvl can inhibit the degradation of β-catenin by the destruction complex, which is composed of adenomatous polyposis coli (APC), axin, casein kinase 1 (CK1) and GSK-3β. Consequently, accumulation of β-catenin in the nucleus regulates gene expression in cooperation with T cell factor (TCF)/lymphocyte enhancer factor (LEF) transcription factors, resulting in the activation of the WCP target genes such as CCND1 and c-Myc (13).
Using gene expression profiling, Rosenwald et al identified MCL signature genes in CCND1-negative lymphoma cases classified as MCL by both morphology and IHC. Wnt3 is one of these MCL signature genes, suggesting that Wnt3 is more intrinsic to MCL than CCND1 (17). According to Gelebart et al, Wnt 3 is highly and consistently expressed in MCL as detected by WCP-specific oligonucleotide arrays (40). Lako et al showed that Wnt3 protein can enhance haematopoietic commitment during in vitro differentiation of embryonic stem cells (41). From these studies, we speculate that aberrant expression of Wnt3 can emerge as potential activator of lymphomagenesis in MCL because CCND1 itself is an oncogenic target of activated WCP. In addition, several studies have shown that Wnt3 is highly expressed in B cell CLL (42,43) as well as in MCL.
To summarize, this study shed light on the mechanisms of initiation and evolution in MCL. The resulting patterns of gene dysregulation in these evens strongly indicate that the Wnt signaling pathway plays a critical role in initial lymphomagenesis, and that specific mitotic regulators facilitate transformation into more aggressive forms. Our unique approach may contribute to future understanding of various mature B cell lymphomas. These data hint at a novel system for the classification of low-grade B cell neoplasms using the expression levels of WCP genes and specific mitotic regulator genes as markers for disease stage and predicted outcomes.