A Cyclin D1–Dependent Transcriptional Program Predicts Clinical Outcome in Mantle Cell Lymphoma

Mantle cell lymphoma (MCL) is a mature B-cell lymphoid neoplasm with a heterogeneous clinical behavior that varies from aggressive evolution in most patients, requiring immediate treatment, to indolent course that may be managed with conservative measures. The initial oncogenic event of MCL is the translocation t(11;14)(q13;q32) leading to cyclin D1 overexpression (1–3). In addition, MCL usually carries a high number of secondary genetic and molecular alterations that target mainly genes involved in cell cycle, DNA damage, and cell survival (1). Dysregulation of cell cycle is considered one of the main hallmarks of MCL, and proliferation levels measured either by the Ki67 IHC index or a proliferation gene expression signature predict the outcome of the patients (4–7). Two molecular subtypes of MCL have been recently identified with different biological and clinical characteristics. Conventional MCL (cMCL) is usually very aggressive and the patients present with disseminated, predominantly nodal, disease, whereas the less frequent leukemic nonnodal MCL (nnMCL) subtype often has an indolent clinical evolution with isolated leukemic presentation, although eventually some tumors progress to more aggressive behavior (7–9). Prognostic biomarkers suitable for both subtypes are needed because the MCL proliferation gene signature, which is a robust prognostic parameter when measured in nodal samples, has a lower performance in blood samples (4, 5, 8).

MCL represents a paradigm of a neoplasia with cyclin D1–related oncogenesis. In fact, MCL cells frequently select additional mechanisms besides the t(11;14) translocation to further increase cyclin D1 levels, such as amplification of the translocated allele or mutations and rearrangements involving the cyclin D1 N-terminal fragment or the 3′ untranslated region (3′UTR; refs. 5, 10–13). These secondary genetic events have been associated with increased aggressiveness of the tumors (14). Besides MCL, cyclin D1 is frequently upregulated in other cancers by different genetic alterations. Among them, a subset of multiple myeloma carries the t(11;14) translocation; cyclin D1 amplifications are common in breast, head and neck, lung, pancreatic, and colorectal cancer, and activating point mutations are present in esophageal and endometrial carcinomas (15).

The classical view of cyclin D1–mediated tumorigenesis considers that its oncogenic effect is based on its role in the G₁–S-phase transition through the CDK4-mediated phosphorylation of RB1 (1). However, emerging data have shown that cyclin D1 has additional functions, such as transcriptional regulation through the interaction with several transcription factors and chromatin modifiers (16, 17). Cyclin D1 seems to have a role both as a transcriptional activator and repressor depending on the molecular and cellular context (18, 19). We recently reported that cyclin D1 shows widespread binding to promoters of active genes in malignant lymphoid cells and induces a global transcriptional downmodulation through the interaction with the transcriptional machinery (20). Nevertheless, we also observed that some cyclin D1 target genes were not affected by this global downmodulation and showed a paradoxical upregulation.

Whether cyclin D1–dependent transcriptional dysregulation has an impact on MCL pathogenesis remains poorly understood. In this study, we identified and characterized a transcriptional program activated by cyclin D1 in MCL, whose high overexpression conferred an adverse clinical outcome.

We used two well characterized MCL cell lines (13), JeKo-1 (ATCC, CRL-3006, RRID:CVCL_1865) and Granta-519 (DSMZ, ACC-342, RRID:CVCL_1818), the lymphoblastoid leukemic cell line JVM13 (ATCC, CRL-3003, RRID:CVCL_1318), and HEK293T (ATCC, CRL-3216, RRID:CVCL_0063). Cell lines were cultured at 37°C and 5% CO₂; JeKo-1 and JVM13 in RPMI (Gibco) and Granta-519 and HEK293T in DMEM (Gibco), both supplemented with 10% FBS and penicillin-streptomycin (Gibco). Cell line authentication was performed by qCell Identity (qGenomics) when appropriate. Cells were tested for Mycoplasma on a regular basis. We obtained cyclin D1 knocked-down MCL cell lines by lentiviral infection using the short hairpin RNA (shRNA) plasmids TRCN0000295873 and TRCN0000295874 (Sigma-Aldrich) as described before (20, 21). Cyclin D1–inducible overexpression models in JVM13, with wild type (CycD1^wt) or a highly stable form of the protein with the T286A mutation (CycD1^T286A), were generated previously (20). Cell infections and gene expression analyses were performed with low-passage cells.

We studied two MCL series of 53 cases from peripheral blood samples and 106 from lymphoid tissues (lymph node and spleen) with gene expression microarray data reported previously (4, 22). In addition, we performed NanoString expression analysis in a third series of peripheral blood samples from 81 patients with MCL, including sequential samples in five of them. All cases had cyclin D1 overexpression and samples were collected before treatment. Cases in the three cohorts were classified as cMCL or nnMCL based on gene expression signatures analyzed by expression microarrays or NanoString and/or SOX11 IHC, depending on the available material, as reported before (23). Tumor cell purification (from peripheral blood samples) and RNA extraction was performed as described previously (24). Molecular characterization was performed as described before (4, 8, 23, 25, 26). To assess the status of cyclin D1 RNA (full-length or truncated 3′UTR) a ratio of 1.5 between the two cyclin D1 probes was used as threshold, both in microarray and NanoString analyses. This study was approved by the institutional review board of the Hospital Clinic of Barcelona, patients provided written informed consent, and the investigation was conducted according to the Declaration of Helsinki.

RNA from cell lines was extracted with the RNeasy Plus Mini Kit (Qiagen). RNA sequencing (RNA-seq) was performed in cyclin D1–silenced and control MCL cell lines (JeKo-1 and Granta-519) shortly after infection and selection. One microgram of total RNA was used to prepare the RNA libraries, which were processed for 50 bp single-end sequencing on Illumina HiSeq 2500 (Supplementary Methods). Affymetrix Human Genome U219 microarrays were used to analyze cyclin D1–overexpressing and control JVM13 cells. Cyclin D1 was induced by doxycycline as described before (20). A total of 150 ng of each RNA sample were processed according to the standard Affymetrix protocol (Supplementary Methods). Data were deposited in Gene Expression Omnibus (GEO) under accession numbers GSE144106 and GSE144095, respectively.

We also revisited several gene expression microarray datasets previously obtained by us and others using Affymetrix Human Genome U133 Plus 2.0 microarrays, and deposited under the following GEO accession numbers: GSE79196 for leukemic MCL samples (n = 53; ref. 22); GSE93291 for lymphoid tissue MCL samples (n = 106; ref. 4); GSE12453 and GSE3526 for normal naïve and memory B cells (n = 10) and normal lymphoid tissues (lymph node and spleen, n = 11), respectively (27, 28); GSE79196 for leukemic B-cell chronic lymphoid neoplasms (n = 101; 54 chronic lymphocytic leukemia, 12 follicular lymphoma, 4 hairy cell leukemia variant, 4 lymphoplasmacytic lymphoma, 23 splenic marginal zone lymphoma, and 4 splenic diffuse red pulp lymphomas; ref. 22); GSE70910 for paired leukemic and nodal MCL samples (29); and GSE19784 for patients with multiple myeloma (30). Microarray data from patients with breast cancer were available online (http://kmplot.com/; ref. 31). Raw microarray data were normalized using the frozen robust multiarray analysis method (32), and differentially expressed genes were identified using Limma (33).

Gene expression of the simplified cyclin D1 signature generated in this study was evaluated on the NanoString platform (NanoString Technologies) together with six housekeeping genes and two cyclin D1 probes (exonic and 3′UTR) described before (4). A total of 100 ng of total RNA from MCL peripheral blood samples were used. Data analysis was performed with nSolver v2.6 software; normalization was done using the geometric mean of both the internal positive controls and the six housekeeping genes. Normalized data were log₂ transformed. Further details are described in Supplementary Methods.

Analyses were performed using R software v3.5.0. The end point of the survival analyses was overall survival (OS). Univariate and multivariate Cox regression models were used to test the association of variables with survival. Martingale residuals of Cox regression models were used to assess the functional form of continuous covariates in relation to survival. The log-rank test was used to compare Kaplan–Meier curves. Unless otherwise specified, survival was calculated from the time of diagnosis. Associations between variables were measured with Student t test, Fisher test, or Pearson r as appropriate; statistical tests are indicated in the figure legends.

To study the transcriptional dysregulation associated with cyclin D1 overexpression in MCL, we initially silenced cyclin D1 using two shRNA constructs in Granta-519 and JeKo-1 MCL cells (Fig. 1A; Supplementary Fig. S1A). As expected, analysis of EdU incorporation showed a noticeable decrease in cell proliferation in cyclin D1–depleted cells (Supplementary Fig. S1B). We then performed RNA-seq experiments using the most efficient shRNA to characterize the transcriptional profile related to cyclin D1 silencing. We identified 556 downregulated and 412 upregulated genes shared by the two cell lines, with a higher overlap in those genes downregulated upon cyclin D1 depletion (Fig. 1B). Because we previously identified that cyclin D1 interacts with active promoters in MCL cells (20), we integrated the RNA-seq results with our previous cyclin D1 chromatin immunoprecipitation sequencing (ChIP-seq) analysis, in which we identified 8,638 cyclin D1 target genes in four MCL cell lines. Interestingly, 448 (81%, P < 2.2 × 10⁻¹⁶) of the genes downregulated upon cyclin D1 silencing had a cyclin D1 peak in their promoters, while only 183 (44%, P = 0.574) of the genes upregulated following cyclin D1 depletion had a cyclin D1 peak (Fig. 1C). These results suggest that cyclin D1 may participate in the activation of a specific set of genes.

We then analyzed the transcriptional profile associated with cyclin D1 overexpression using cellular models established in the cyclin D1–negative lymphoblastoid cell line JVM13. We overexpressed either CycD1^wt or CycD1^T286A (Supplementary Fig. S2A), a mutant that reaches expression levels closer to the high amounts observed in MCL (20). We identified 739 upregulated and 841 downregulated genes common in both models. Concordantly with the cyclin D1–silencing experiments, we observed that the proportion of genes with a cyclin D1 peak in their promoters was markedly higher in the genes upregulated upon cyclin D1 overexpression (63%, P < 2.2 × 10⁻¹⁶) than in the downregulated ones (37%, P = 0.994), confirming the contribution of cyclin D1 in transcriptional activation of a specific gene set (Supplementary Fig. S2B).

We then combined the cyclin D1–silencing and overexpression results to define a robust gene expression program associated with cyclin D1 dysregulation. Remarkably, we observed that 295 (66%) of the 448 genes that were downregulated by cyclin D1 depletion in MCL cells and had a cyclin D1 peak in their promoters were upregulated by CycD1^T286A overexpression in JVM13 cells (Fig. 1D; Supplementary Fig. S2C). This tendency was also observed in the CycD1^wt cells, although the proportion of genes upregulated was slightly lower concordantly with the lower cyclin D1 expression in these cells. Taken together, these results strongly suggest that cyclin D1 promotes the transcriptional activation of a specific subset of its target genes in MCL cell lines. We identified a cyclin D1–dependent transcriptional program of 295 genes that were downregulated upon cyclin D1 silencing, upregulated upon cyclin D1 overexpression, and had a cyclin D1 peak in their promoters (Supplementary Fig. S3; Supplementary Table S1).

Gene Ontology (GO) analysis of the 295 genes activated by cyclin D1 showed a high enrichment of cell cycle and DNA damage, with 182 (62%) of them involved in the four phases of the cell cycle (Fig. 2A and B; Supplementary Table S2). Transcriptional control during cell cycle is based on two major waves of transcription: one occurring during G₁–S phases, whose genes are characterized by E2F motifs in their promoters, and the other during G₂–M, with genes displaying cell cycle genes homology regions (CHR, ref. 34). Both E2F and CHR motifs were highly enriched in the cyclin D1 peaks present in the promoters of cyclin D1–activated genes, reinforcing the idea that cyclin D1 regulates the expression of genes that control the entire cell cycle (Fig. 2C). These results are consistent with the reduced proliferation following cyclin D1 silencing in MCL cells described above.

We next explored the specific features of the cyclin D1 peaks present in the promoters of the 295 genes activated by cyclin D1. For that, we compared the morphology of these cyclin D1 peaks with either the peaks present in all cyclin D1 target genes or with the peaks from a subset of cyclin D1 target genes whose expression remained stable upon cyclin D1 silencing (cyclin D1 nonregulated genes). Cyclin D1 peaks present in cyclin D1–activated genes were significantly larger, with higher tag number and width, suggesting that they represent a specific subgroup among all cyclin D1 peaks (Fig. 2D). We next analyzed whether cyclin D1 peaks overlapped with specific transcription factors in cyclin D1–activated gene promoters compared to cyclin D1 nonregulated genes. We used ChIP-seq data corresponding to multiple transcription factors from the ENCODE project obtained in normal B cells (GM12878; ref. 35). Interestingly, we found a significant overlap between cyclin D1 peaks and several factors involved in transcriptional regulation during cell cycle (34), such as E2F4 and FOXM1 (Fig. 2E; Supplementary Table S3). These results are concordant with the previous observation that most cyclin D1–activated genes are cell cycle related. Accordingly, we observed a clear colocalization of cyclin D1 with both E2F4 and FOXM1 around the transcription start site of cyclin D1–activated genes (Fig. 2F). Motif enrichment analysis showed that E2F4 and FOXM1 binding sites were significantly enriched in the cyclin D1 peaks present in cyclin D1–activated gene promoters (Fig. 2G; Supplementary Table S4). Of note, FOXM1 was one of the genes included in the cyclin D1–dependent transcriptional program. Interestingly, cyclin D1 was previously described to interact with both transcription factors in other cell types (36–40). Co-immunoprecipitation (co-IP) experiments in JeKo-1 cells confirmed that cyclin D1 physically interacted with both E2F4 and FOXM1 (Fig. 2H). In other cellular contexts, cyclin D1 has been shown to activate transcription through the recruitment of the coactivator CREB binding protein (CPB) (41). co-IP experiments also confirmed the interaction between these two proteins in MCL cells (Fig. 2H). Together, these results show that cyclin D1 promotes the activation of an expression program of genes mainly involved in cell cycle control through a mechanism that may include in part the interaction of cyclin D1 with key cell-cycle transcriptional regulators such as E2F4 and FOXM1, and with CBP.

To determine whether the cyclin D1–dependent transcriptional program identified in MCL cell lines was also modulated in primary MCL tumors, we studied the microarray expression data from two independent MCL cohorts of 53 cases from peripheral blood samples (55% cMCL, 45% nnMCL) and 106 from lymphoid tissues (96% cMCL, 4% nnMCL). Hierarchical clustering analysis of the 295 cyclin D1–activated genes in both blood and tissue samples resulted in the distribution of patients in groups with different expression levels (Fig. 3A). Because cyclin D1–activated genes had a homogeneous expression pattern within each patient, we summarized the cyclin D1–dependent program by defining a cyclin D1 signature score based on the mean expression of all genes (Supplementary Fig. S4).

Given that cyclin D1 is specifically overexpressed in MCL, we analyzed whether the cyclin D1 signature was upregulated in primary MCL. We used gene expression microarray data from normal B-cell and lymphoid tissue samples to compare them with blood and tissue MCL samples, respectively (27, 28). Interestingly, we found that the cyclin D1–dependent transcriptional program was highly overexpressed in MCL (Fig. 3B). In addition, by analyzing expression microarrays from a large series of different leukemic cyclin D1–negative B-cell lymphoid neoplasms (22), we found that the cyclin D1 signature was significantly overexpressed in leukemic MCL compared with other lymphoid malignancies (Fig. 3B). Concordantly, a high proportion of the 295 cyclin D1–activated genes were upregulated in MCL compared with the normal lymphoid samples or other lymphoid neoplasms (Fig. 3C; Supplementary Fig. S5A). Furthermore, we found a strong positive correlation between cyclin D1 signature expression and cyclin D1 levels in primary MCL, in both blood and tissue samples (Fig. 3D). Accordingly, the vast majority of cyclin D1–activated genes showed a significant positive correlation with cyclin D1 (Supplementary Fig. S5B). In addition, the expression analysis of two gene sets composed by FOXM1 or E2F4 targets obtained in GM12878 cells (35) displayed a significant correlation with cyclin D1 levels in primary MCL cases, further supporting the functional interplay of cyclin D1 with these two transcription factors (Supplementary Fig. S6). Taken together, these results suggest that cyclin D1 overexpression in primary MCL contributes to the upregulation of the cyclin D1–dependent transcriptional program identified in cell lines.

We next investigated whether the cyclin D1–dependent transcriptional program was associated with clinicopathologic features. We observed a significant association between high levels of cyclin D1 signature and common genetic alterations in MCL such as 17p/TP53 alteration, truncated 3′UTR cyclin D1, and high number of copy-number alterations (CNA; Supplementary Figs. S4 and S7). Interestingly, analysis of the cyclin D1 signature score as a continuous variable using Cox regression showed a very strong direct correlation between cyclin D1 signature expression levels and the death risk, in both leukemic [HR = 5.51(1.96–15.47); P = 0.0012] and tissue [HR = 4.69(2.60–8.47); P = 2.98 × 10⁻⁷] MCL cohorts (Fig. 4A). Accordingly, a high proportion of the 295 cyclin D1–activated genes had a positive correlation with death risk (Supplementary Fig. S8A). Moreover, the division of patients in two subgroups based on the cyclin D1 signature score distribution (Supplementary Fig. S8B) showed that patients with high score had a significant shorter OS than patients with low score, in both leukemic (median survival 2.7 vs. 8.7 years; P < 0.001) and tissue (median survival 0.8 vs. 4.4 years; P < 0.001) samples (Fig. 4B). Together, these results demonstrate that increasing expression levels of cyclin D1–dependent program are associated with more aggressive features and shorter survival in patients with MCL.

We next wanted to validate the association of the cyclin D1–dependent transcriptional program with clinical outcome using an approach more adaptable to the clinical practice, such as digital gene expression profiling by NanoString technology. We initially applied several filters to reduce the full cyclin D1 program to a 37-gene signature (Supplementary Fig. S9; Supplementary Methods). For the validation, we focused on MCL blood samples, so that the assay could be applied to all patients with leukemic MCL, either with cMCL or nnMCL. We first analyzed the simplified 37-gene signature by NanoString in 28 cases previously evaluated by microarray and we found that it showed a very high correlation with the 295-gene signature (r = 0.9525; Supplementary Fig. S10A), confirming the adequacy of this approach. To validate the clinical impact of this simplified cyclin D1 signature, we studied an independent cohort of 53 patients with MCL (validation series) together with sequential samples from 5 patients. The analysis of the sequential samples showed that the cyclin D1 signature levels changed over time (Supplementary Fig. S10B), and therefore we evaluated the OS from the sampling time. In this validation series, we could confirm the previous findings observed with the full cyclin D1–dependent gene program, including the clustering of patients in different expression groups, the positive correlation of this simplified signature with cyclin D1 expression levels, and the association of increasing cyclin D1 signature with adverse prognosis (Fig. 5A–C; Supplementary Fig. S11). Moreover, by dividing the patients into five groups of increasing cyclin D1 signature expression by equal width binning, we found a significantly shorter OS as cyclin D1 signature levels increased (Fig. 5D).

We further studied the biological and clinical features associated with the cyclin D1 signature in the full set of 81 patients with leukemic MCL analyzed by NanoString (Fig. 5E). We found that cMCL had higher expression of cyclin D1 than nnMCL and, concordantly, cyclin D1 signature levels were also higher in cMCL (Supplementary Fig. S12A). Interestingly, cyclin D1 signature correlated with cyclin D1 expression and with survival in both MCL subtypes independently (Supplementary Fig. S13). Furthermore, higher levels of both cyclin D1 and cyclin D1 signature were significantly associated with some of the most common genetic alterations in MCL such as aberrations in 17p/TP53 and 9p/CDKN2A, presence of truncated 3′UTR cyclin D1 and high number of CNA (Supplementary Fig. S12B–S12F), in line with the observations in the previous MCL microarray series. Finally, patients classified as “high” and “standard” proliferation according to the MCL35 proliferation signature assay defined in tissue samples (4, 8), showed higher cyclin D1 signature expression than those classified as “low” proliferation, although only 15% of patients were classified in “high” or “standard” groups (Supplementary Fig. S14A). Multiple two-variable Cox regression models in the 53 cases from the validation series showed that cyclin D1 signature was a better predictor of survival compared not only with the MCL proliferation signature but also with most of the molecular factors analyzed previously (Supplementary Table S5). This analysis also showed that cyclin D1 signature had prognostic value for OS independently of TP53, CDKN2A, and total number of CNA.

To understand the prognostic difference between the cyclin D1 signature and the previously defined MCL proliferation signature (4) in this leukemic MCL cohort, we analyzed the genes from both signatures. While most genes in the proliferation signature belonged to G₂–M, the cyclin D1 signature contained a similar proportion of G₁–S and G₂–M genes (Supplementary Fig. S14B). In addition, by evaluating both cyclin D1 and MCL proliferation signatures in a set of blood and tissue samples from the same patients (29), we found that the genes from both signatures were expressed at similar levels in tissue, but the ones from the proliferation signature displayed significantly lower expression than those of the cyclin D1 signature in blood samples (Supplementary Fig. S14C).

We investigated whether the cyclin D1–dependent transcriptional program identified in MCL could be present in other cyclin D1–overexpressing cancers. We analyzed the gene expression profiling of a multiple myeloma series reported previously (30), focusing on a subgroup characterized by cyclin D1 overexpression. Hierarchical clustering analysis led to the distribution of patients with multiple myeloma in groups with different expression levels of the cyclin D1–dependent program (Fig. 6A). More importantly, we found a significant positive correlation between cyclin D1 levels and those of the cyclin D1 signature (Fig. 6B). We next analyzed a large series of patients with breast cancer published previously (31). Because cyclin D1 was shown to play different roles in estrogen receptor (ER)-positive versus ER⁻ breast cancers and to be overexpressed mainly in the first (42–44), we analyzed the two groups independently. Hierarchical clustering analysis of the cyclin D1–dependent program in ER⁺ breast cancer showed distinct groups characterized by different expression levels (Fig. 6C). Remarkably, cyclin D1 signature expression showed a significant positive correlation to cyclin D1 levels in ER⁺ but not in ER⁻ breast cancer (Fig. 6D; Supplementary Fig. S15A). Concordantly, when splitting the patients by the median cyclin D1 signature level, we found that high expression of cyclin D1 signature was associated with poorer survival in ER⁺ but not in ER⁻ breast cancer (Fig. 6E; Supplementary Fig. S15B). Together, these data suggest that cyclin D1 overexpression in a subset of patients with multiple myeloma and breast cancer may contribute to the upregulation of the specific cyclin D1–dependent transcriptional program identified in MCL.

In this study, we have identified a cyclin D1–dependent transcriptional program in MCL cells composed of 295 genes mainly related to cell-cycle regulation. Cyclin D1–dependent program was overexpressed in primary MCL directly proportional to cyclin D1 levels and increased expression of this program was associated with poorer survival. This transcriptional program was also present in multiple myeloma and breast cancer with cyclin D1 overexpression. Our data suggest that transcriptional dysregulation mediated by cyclin D1 overexpression might play an important role in MCL pathogenesis thus impacting on clinical outcome.

The relevance of cyclin D1 overexpression in MCL is stressed by the finding that tumors with higher cyclin D1 levels have increased proliferation and aggressiveness (14, 45, 46). These observations are concordant with our finding that the cyclin D1–dependent transcriptional program is highly enriched in cell cycle–related genes and high levels of this program are associated with shorter survival of the patients. Interestingly, half of the genes from the previously defined MCL proliferation signature are included in the cyclin D1–dependent program. Therefore, gene activation by cyclin D1 may be, in addition to other common MCL genetic alterations, one of the underlying mechanisms dysregulating MCL proliferation. The association between the cyclin D1 signature and the presence of TP53 and CDKN2A alterations could be the result of a selective pressure to inactivate cell-cycle regulatory pathways in cases with high cyclin D1.

Our results suggest that cyclin D1 may facilitate transcriptional activation of cell-cycle genes through the binding to their promoters in MCL cells. We cannot rule out that the induction of these genes may also reflect in part the activation of the CDK4/6–RB1 pathway mediated by cyclin D1 overexpression that triggers the G₁–S-phase transition. However, previous genome-wide studies in mouse embryonic fibroblasts and in breast cancer models showed that cyclin D1 binds to the promoter of genes that regulate cell division and activates their expression independent of CDK4, stressing the importance of the transcriptional role of cyclin D1 (36, 47). Moreover, the complete inactivation of RB1 by mutations and deletions in some MCLs (48), would support the idea that cyclin D1 may play additional oncogenic roles besides its classical cell-cycle function. Similarly to other cellular models (16–18), in MCL cells cyclin D1 might facilitate transcriptional activation through the interaction with transcription factors, including E2F4 and FOXM1. Remarkably, cyclin D1 was previously described to interact with E2F4 and stimulate the expression of E2F motif–containing genes (36–38), and cyclin D1-CDK4 complex was reported to activate FOXM1 thus inducing the expression of cell-cycle genes (39, 40), in different murine and human models. These two proteins are important regulators of the two transcriptional waves occurring during cell cycle: E2F4 mainly regulates G₁–S genes while FOXM1 mostly works in G₂–M phase (34). This is consistent with our observation that the cyclin D1–dependent transcriptional program contains genes that belong to the different cell-cycle phases. Moreover, we found that cyclin D1 interacted with the coactivator CBP in MCL cells as previously reported in other cellular models (41), reinforcing the idea that cyclin D1 would play a role in transcriptional activation in MCL cells. Overall, our results suggest that cyclin D1–mediated cell cycle activation in MCL would rely not only on its function as a regulator of the G₁–S-phase transition through CDK4 binding but also in its role facilitating transcriptional activation of G₁–S and G₂–M cell-cycle genes.

The transcriptional role of cyclin D1 in MCL was not addressed until recently when we described that cyclin D1 shows widespread binding to promoters of active genes in MCL cells, inducing a global transcriptional downmodulation (20). However, we also observed that some cyclin D1 target genes were not affected by this effect. Here, we have found that, in parallel to the large-scale transcriptional downregulation, cyclin D1 also promotes the activation of a specific subset of its many targets. We observed that cyclin D1–activated genes had larger cyclin D1 peaks and presented specific DNA motifs compared to other cyclin D1 targets, suggesting that these genes may represent a distinct functional subgroup. This behavior of cyclin D1 is similar to that reported for the oncogene MYC, which works as a global transcriptional amplifier and in parallel activates or represses specific gene sets (49–51). Further studies should clarify in more detail the mechanisms of cyclin D1–mediated transcriptional regulation, not only in lymphomagenesis but also in other cancers.

The fact that a subset of patients with MCL, particularly with nnMCL, have only leukemic manifestation of the disease, jeopardize the use of prognostic biomarkers such as the MCL proliferation signature, that is a robust prognostic parameter when measured in nodal samples but has a lower performance in leukemic samples (8). In contrast, the cyclin D1–dependent transcriptional program described here is consistently associated with the survival of the patients in both tissue and leukemic MCL samples. Moreover, our 37-gene cyclin D1 signature assay uses a gene expression platform compatible with clinical setting. The different performance between the proliferation and the cyclin D1 signatures could be due to the fact that the former was developed through the identification of genes associated with survival in MCL nodal samples (4, 5), and we have seen here that the expression of these genes is significantly lower in MCL blood samples. Since cyclin D1 signature is highly expressed in both blood and tissue tumor cells, it is likely that it may represent an intrinsic biological feature of MCL cells, independent of microenvironment interactions. Moreover, because it contains genes from both G₁–S and G₂–M phases, it might better capture the proliferation state in both tumor compartments. In any case, the fact that measurement of the cyclin D1–dependent transcriptional program, including the simplified 37-gene signature, could be applied to predict the clinical outcome in both MCL subtypes is relevant and it may help to identify patients with nnMCL with adverse clinical outcome similar to cMCL.

The cyclin D1–dependent transcriptional program identified in MCL cells was also observed in specific molecular subsets of multiple myeloma and breast cancer with cyclin D1 overexpression. The moderate correlation observed between the cyclin D1 signature and cyclin D1 levels may reflect that dysregulation of the proliferation genes that belong to the signature could also occur by different oncogenic mechanisms active in these tumors. Concerning multiple myeloma, although cases with cyclin D1 overexpression generally show a more favorable prognosis (52), increased cyclin D1 signature might confer shorter survival in this particular subset of patients as we have observed in MCL. Regarding breast cancer, high cyclin D1 signature was associated with decreased survival in ER⁺ but not ER⁻ tumors. Accordingly, cyclin D1 overexpression was previously associated with high proliferation and increased risk of mortality exclusively in ER⁺ breast cancers (42–44). Further research should clarify to which extent cyclin D1–mediated transcriptional dysregulation contributes to oncogenesis in these and other cyclin D1–overexpressing tumors.

In conclusion, our results suggest that cyclin D1 oncogenic effect in MCL, and possibly in other cancers, is based not only on its canonical cell-cycle function but also on its role in transcriptional regulation. We have identified a cyclin D1–dependent transcriptional program that is overexpressed in MCL and can be applied to predict clinical outcome in both MCL subtypes.

E. Giné reports grants and nonfinancial support from Janssen; personal fees and nonfinancial support from Gilead; and grants from Roche outside the submitted work. D.W. Scott reports grants from NanoString Technologies and Roche; personal fees from AbbVie, AstraZeneca, and Celgene; and grants and personal fees from Janssen outside the submitted work. In addition, D.W. Scott has a patent for molecular subtyping of lymphoma including mantle cell lymphoma pending and licensed to NanoString. L.M. Rimsza reports a patent related to NanoString technology that has been issued and licensed. A. López-Guillermo reports grants and other from Roche, Gilead/Kite, and Celgene, and other from Novartis during the conduct of the study. E. Campo reports grants from NIH, Spanish Ministerio de Ciencia, Innovación y Universidades, and Generalitat de Catalunya Suport Grups de Recerca AGAUR during the conduct of the study; in addition, E. Campo has a patent for NanoString Technologies licensed. P. Jares reports grants from Instituto de Salud Carlos III-Fondo de Investigaciones Sanitarias during the conduct of the study. No disclosures were reported by the other authors.

S. Demajo: Conceptualization, data curation, software, formal analysis, validation, investigation, visualization, methodology, writing-original draft, writing-review and editing. R. Albero: Formal analysis, investigation, methodology, writing-review and editing. G. Clot: Software, formal analysis, methodology, writing-review and editing. G. Castellano: Software, formal analysis, methodology. A. Navarro: Data curation, formal analysis, writing-review and editing. C. Capdevila: Investigation. A. Enjuanes: Formal analysis, writing-review and editing. F. Nadeu: Software, writing-review and editing. E. Giné: Resources, data curation. M. Pinyol: Formal analysis, writing-review and editing. E.S. Jaffe: Resources, writing-review and editing. G. Ott: Resources, writing-review and editing. L.M. Staudt: Resources, writing-review and editing. A. Rosenwald: Resources, writing-review and editing. D.W. Scott: Resources, writing-review and editing. L.M. Rimsza: Resources, writing-review and editing. A. López-Guillermo: Resources, data curation. S. Beà: Data curation, formal analysis, writing-review and editing. E. Campo: Conceptualization, resources, formal analysis, supervision, funding acquisition, writing-original draft, project administration, writing-review and editing. P. Jares: Conceptualization, formal analysis, supervision, funding acquisition, validation, writing-original draft, project administration, writing-review and editing.

This study was supported by Fondo de Investigaciones Sanitarias-Instituto de Salud Carlos III (grant PI14/00355 and grant PI19/01838, to P. Jares); NIH “Molecular Diagnosis, Prognosis, and Therapeutic Targets in Mantle Cell Lymphoma” (grant P01CA229100, to E. Campo); the Spanish Ministerio de Ciencia, Innovación y Universidades (grant RTI2018–094274-B-I00, to E. Campo); Generalitat de Catalunya Suport Grups de Recerca AGAUR (grant 2017-SGR-1142, to E. Campo); FEDER: European Regional Development Fund “Una manera de hacer Europa”; and CERCA Programme/Generalitat de Catalunya. S. Demajo was supported by a Juan de la Cierva fellowship (Spanish Ministerio de Economía y Competitividad). R. Albero was supported by a grant from the Leukemia and Lymphoma Society. E. Campo is an academia researcher of the Institució Catalana de Recerca i Estudis Avançats (ICREA) of the Generalitat de Catalunya. This work was mainly developed at the Centre Esther Koplowitz (CEK), Barcelona, Spain. The authors thank the Hematopathology Collection registered at the Biobank of Hospital Clínic—IDIBAPS for sample procurement. They are indebted to the Genomics core facility of the IDIBAPS and the Genomics Unit of the CRG for technical help. They are grateful to N. Agell for providing reagents and N. Garcia and S. Ruiz for their technical and logistic assistance.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.