Post-transcriptional Modifications Contribute to the Upregulation of Cyclin D2 in Multiple Myeloma

Gene expression profiling has revealed that expression of CCND1, CCND2, or CCND3 is increased in virtually all multiple myeloma tumors, providing the hypothesis of a potential unifying event in multiple myeloma pathogenesis (1). In addition, D-cyclins have been proposed as molecular therapeutic targets in multiple myeloma (2). Cyclin D2 belongs to the group of D-type cyclin proteins that are cyclically expressed during the cell cycle. In physiologic conditions, cyclins D regulate the transition from G₁–S phase by interaction with cyclin-dependent kinases 4 or 6, which further phosphorylate their substrates (3). One of them is Rb that promotes proliferation by release of E2F transcription factor. CCND1 is directly activated by t(11;14) and by biallelic dysregulation in patients with multiple myeloma with polysomy 11, and CCND3 is overexpressed in patients with t(6;14). High levels of CCND2 are detected in multiple myeloma with t(4;14), t(14;16), and in a set of hyper and nonhyperdiploid multiple myeloma (4). The mechanisms by which CCND2 is upregulated in these cases are not completely deciphered. Apparently, CCND2 is not directly induced by IGH translocations, although the transcription factor Maf involved in t(14;16) transactivates the CCND2 promoter (5). Similarly, a pathway of CCND2 transactivation in multiple myeloma by transcription factor ZKSCAN3 has been described (6).

Nowadays, a large body of evidence indicates that miRNAs are key post-transcriptional regulators of gene expression. In fact, in silico algorithms have identified hundreds of predicted miRNAs targeting CCND2. Previously published data of our group showed a significant correlation between CCND2 upregulation and decreased expression of seven miRNAs in multiple myeloma: miR-15a, miR-19a, miR-19b, miR-20a, miR-135b, miR-196b, and miR-320 (7). In this regard, a downregulation of CCND2 has been demonstrated after transfection of myeloma cells with pre-miRNA-15a and -16, supporting the functional role of both miRNAs in regulating cell cycle in multiple myeloma (8). Because miRNA function requires the presence of regulatory sites in 3′ untranslated regions (3′UTRs) of mRNA, another important mechanism for oncogenic activation is the loss of miRNA sites in the mRNAs of oncogenes (9). This process can occur through the use of proximal alternative polyadenylation (APA) signals, which lead to 3′UTR shortening and subsequent loss of miRNA complementary sites (10). Several genes, such as CCND1, CCND2, IMP-1, DICER, CDC6, and CYP450, have been reported to be regulated by this mechanism in cancer cells of different origin (9, 11, 12). Some mantle cell lymphomas (MCL) express a short CCND1 mRNA isoform, which is correlated with high proliferation of lymphoma cells and decreased overall survival of patients (13). Likewise, a short version of CCND1 mRNA has been detected in multiple myeloma that harbor t(11;14), but it does not correlate with the cell proliferation rate (14).

This background prompted us to investigate the hypothesis that CCND2 overexpression in multiple myeloma could be induced by a more general mechanism based on miRNA regulation, regardless of a CCND2 transactivation by selective transcription factors.

The human myeloma cell lines, NCI-H929, MM1S, MM1R, and U266 were acquired from the ATCC, whereas RPMI-8226, OPM-2, KMS12BM, KMS12PE, JJN3, and HEK923 from Deutsche Sammlung von Mikroorganismen and Zellkulturen (DSMZ). SKMM2 and XG-1 cell lines were kindly provided by Dr Giovanni Tonon from San Raffaele Scientific Institute, Milan, Italy. Cells were routinely checked for the presence of mycoplasma with a MycoAlert kit (Lonza). Cell line identities have been tested and authenticated by short tandem repeat (STR) analysis with a PowerPlex 16 HS System kit (Promega) and online STR matching analysis (DSMZ institute; ref. 15). The human STR profile database includes data sets of 2,455 cell lines from the ATCC, DSMZ, JCRB, and RIKEN.

Forty-five purified CD138-positive myeloma samples from newly diagnosed patients were included in the study. AutoMACS automated separation system (Miltenyi-Biotec) was used to plasma cell isolation. All patients provided written informed consent in accordance with the Helsinki Declaration, and the research ethics committee of the University Hospital of Salamanca approved the study. The systematic screening for IGH translocations using interphase FISH analysis was carried out in all the patients (Supplementary Table S1) as previously described (7).

H929, MM1S, and RPMI cell lines were transfected using the nucleofector II system (Lonza) with C-16 program for H929, G-16 for MM1S, and RPMI, and X-001 for JJN3 cells. mirVana miRNA mimic of specified miRNA or mirVana miRNA mimic negative, nontargeting control#1 (Ambion) at 50 nmol/L concentration were used. CCND1-on-targeting smart pool siRNA at 100 nmol/L was used to silence CCND1 and compared with nontargeting smart pool control siRNA (Dharmacon). JJN3 cells were transfected with 1 pmole of each of the plasmids. pCMV6-XL5 control empty plasmid and pCMV6-XL5-Cyclin D3 were purchased from Origene. Plasmid to overexpress cyclin D1, pRc/CMV cyclin D1, was a gift from Bob Weinberg (Whitehead Institute for Biomedical Research, Cambridge, MA) (Addgene plasmid #8962; ref. 16). Transfection efficiency was assessed as previously described (17).

RNA was extracted using an Allprep DNA/RNA kit (Qiagen). The RNA integrity was assessed using Agilent 2100 Bioanalyzer (Agilent Technologies). Mature miRNA expression levels were detected with TaqMan qRT-PCR miRNA assays (Applied Biosystems) and normalized to RNU43 endogenous control using the 2^−ΔCt method. Total RNA (1 μg) was reverse transcribed to cDNA using an iScript cDNA synthesis kit (Bio-Rad). CCND2 and CCND1 mRNA expression were also evaluated by Taqman assay qRT-PCR using the respective GAPDH Taqman assay as a control. The CCND2 short and long 3′UTR isoform expression was evaluated by qRT-PCR. The expression of the isoforms was normalized against GAPDH. The efficiency of PCR was evaluated by 10-fold change dilutions. qPCRs were performed using iQSYBR Green Supermix and iQ5 real-time PCR system (Bio-Rad). The long 3′UTR versus short 3′UTR CCND2 mRNA expression was calculated as the percentage of isoform expression. Percentage of LONG 3′UTR = ((2^−ΔCt 3′UTR)/(2^−ΔCt CDS))*100% and percentage of SHORT = 100% − LONG. Sequences of all custom primers used in the study are provided in Supplementary Table S2.

For the validation of the APA of CCND2 mRNA, 3′RACE was performed using the FirstChoice RLM-RACE Kit (Ambion). RACE-specific cDNA was synthesized from 1 μg of total RNA and 3′RACE anchor adaptor. Outer 3′RACE adaptor primer, complimentary to the anchored adapter, and the outer-CCND2 primer were used for first amplification, followed by nested PCR with the inner-3′RACE adaptor primer and the inner CCND2-specific primer (Supplementary Table S2). 3′RACE PCR was performed using Accu-PrimeTaq DNA Polymerase High Fidelity (Invitrogen). Following gel extraction of correct size bands with the GeneJet Gel Extraction Kit (Thermo Scientific), the PCR product was ligated into plasmid using a pGEMT-Easy kit (Promega). The vector was then sequenced using Applied Biosystems 3130 × l Genetic Analyzer. The sequences were aligned to CCND2 mRNA sequence (NM_001759.3) using BLASTN algorithm.

Whole cell lysates were collected using RIPA buffer (Santa Cruz Biotechnology) containing protease inhibitors and phosphatase inhibitors (Roche). Protein samples (50 μg/lane) were subjected to SDS-PAGE electrophoresis and transferred to 0.45 μm polyvinylidene difluoride membrane using iBlot Dry Blotting System (Invitrogen). The primary antibodies used for immunoblotting were anti-cyclin D2 (Cell Signaling, #3741), anti-cyclin D1 (Abcam ab134175), anti-cyclin D3 (Abcam ab52598), and anti–β-actin-HRP (Sigma-Aldrich) as control for protein loading. The chemiluminescence from horseradish peroxidase-linked anti-rabbit IgG antibody (PierceNet) was detected using Clarity Western ECL Substrate (BioRad). The density of bands was determined using the Image J program.

HEK293 cells were transfected with 1 μg of the pGFP and 0.65 pmoles of plasmid with the 3′UTR cloned into pIS1 Cyclin D2 long UTR (Addgene Plasmid 21645) or pIS1 Cyclin D2 short UTR (Addgene Plasmid 21644) and cotransfected with 50 nmol/L miRNA precursor molecule by nucleofection using HEK293 cell line program in the nucleofector II system (LONZA). Cells were collected 24 hours after transfection and Renilla luciferase activity was measured using Dual-Glo Luciferase Assay System (Promega) according to the manufacturer's protocol. GFP was measured using the fluorescence microplate reader Tekan Infinite F500. Renilla luciferase activity was normalized to GFP.

Total RNA (10 μg) was glyoxylated and fractionated in 1.4% agarose gels in 10 mmol/L Na₂HPO₄ buffer (pH 7). After electrophoresis, the gel was blotted onto Hybond-N (Amersham), UV light-cross-linked, and hybridized to ³²P-radiolabeled CCND2 probe that was developed by PCR product gel extraction. Northern blot probe was generated with the same primers used to detect both the short and long CCND2 isoforms by qPCR (see scheme in Fig. 2A). Loading was monitored by reprobing the filters with β-actin cDNA.

mRNA FISH was performed using Stellaris FISH probes (Biosearch Technologies) as described previously (18). Forty-eight probes against CCND2 3′UTR and 25 against CCND2 CDS were designed using the Stellaris FISH Probe Designer (Supplementary Table S3). 3′UTR and CDS probes were coupled to Quasar-670 and Quasar-570 dyes, respectively. Cells were visualized with Olympus IX70 FV 500 confocal microscope equipped with 60× oil immersion objective. The microscope settings were adjusted as to observe no signal from the negative control and were kept identical for all the samples measured. The whole sample was examined and the most representative image was taken for analysis, which included only black and white levels adjustments.

Genomic DNA (500 ng) was processed using EpiTect Bisulfite Kits (Qiagen) to obtain bisulfate-treated DNA. To achieve sequencing data concerning DNA methylation, PCR products obtained from bisulfate-treated DNA were ligated into pGEM-Easy plasmid (Promega) and sequenced with the ABI 3100 DNA sequencer using M13 primers. PCR primers to amplify parts of the CCND2 promoter are provided in Supplementary Table S2.

The two-sided Student t test was used to analyze differences in experiments. Data are reported as mean values ± SD of at least triplicate determinations. The Mann–Whitney U test was used to identify statistically significant differences between the two groups of multiple myeloma. All statistical analyses were conducted using the SPSS 21.0 statistical package (SPSS).

We evaluated CCND2 and CCND1 mRNA expression level by qRT-PCR in a set of 45 newly diagnosed multiple myeloma. We found that in all the samples with high expression of CCND1, the expression levels of CCND2 were much lower, particularly in those samples with t(11;14), in which CCND2 transcript was undetectable or barely expressed. On the contrary, most of the samples with the highest levels of CCND2 displayed lower expression levels of CCND1, between 10 and 100 time less (Fig. 1A). The simultaneous expression of both cyclin D genes was an uncommon observation. The predominant overexpression of a single CCND gene was also observed in the cell lines (Fig. 1B). Cyclin D2 protein levels assessed by Western blot were noticeable in all the cell lines (Fig. 1C).

We reported in a previous study that the high expression of CCND2 mRNA in the multiple myeloma samples was correlated with a significant downregulation of some of the following seven miRNAs, miR-15a, miR-19a, miR-19b, miR-20a, miR-135b, miR-196b, and miR-320, depending on the cytogenetic abnormality (7). We updated these miRNA-CCND2 predictions using six databases (DIANAmT, miRanda, miRWalk, RNAhybrid, PICTAR4, and Targetscan). Because CCND2 gene shares high homology with CCND1, miRNAs targeting CCND1 were also explored. After selecting the miRNA-target interactions predicted by three out of six databases, 88 and 225 miRNAs were predicted to target CCND1 and CCND2, respectively. Moreover, 164 miRNAs were common for CCND1 and CCND2 targets (Supplementary Table S4). Among the seven miRNAs correlated with CCND2 expression, miR-135b, miR-196b, and miR-320 were predicted to exclusively target CCND2, whereas miR-15a, miR-19a, miR-19b, and miR-20a were predicted to target both CCND1 and CCND2.

Next, the level of miRNAs correlated with CCND2 expression was evaluated by qRT-PCR in six human myeloma cell lines. Additionally, miR-214 and miR-375, which also targeted CCND2 and whose expression was downregulated in multiple myeloma (7), were included in qRT-PCR analysis. No expression of miR-15a, 135b, 196b, 214, and 375 was observed in almost all cell lines, although expression of miR-19a, 19b, and 20a was detected in all the cell lines (Supplementary Table S5).

According to the prediction algorithms, miR-15a contains three; miR-19a, miR-19b, and miR-320 have two; and miR-20a, miR-135b, miR-196b, miR-214, and miR-375 contain one putative binding site in the CCND2 3′UTR. The locations of seed sites in the 3′UTR of CCND2 is presented in Fig. 2A. To determine whether CCND2 was a direct target of those miRNAs, we carried out luciferase reporter assays in HEK293 cells with a reporter plasmid vector containing the 3′UTR of CCND2. Data from our experiments revealed that luciferase activity of CCND2 3′UTR luciferase reporter in the cells cotransfected with miR-15a, 19a, 19b, 20a, 135b, 196b, 214, 320, and 375 was significantly lower than that of the cells transfected with nontargeting control miRNA or with miR-155, which does not target CCND2 (Fig. 2B).

Given the validation of miRNAs-CCND2 interaction for the miRNAs explored, we analyzed the effect of ectopic expression of the miRNAs downregulated in the multiple myeloma cell lines on cyclin D2 protein. We found a significant decrease of cyclin D2 protein level in miR-15a, 135b, 196b, 214, and 320 transfected MM1S cells. A dramatic downregulation of cyclin D2 level was observed in RPMI cells transfected with miR-214, and to a small extent with miR-15a and miR-196b. This effect on cyclin D2 protein was correlated with the downregulation of mRNA observed in RPMI for miR-15a and miR-214 (Supplementary Fig. S1). Conversely, we only noted an attenuated decrease of cyclin D2 in the H929 cells transfected with miR-214 and miR-135b (Fig. 2C and D). The ectopic expression of miRNAs could result in nonphysiologic concentrations inside the cells, which could induce off-target effects leading to false positive results. In order to evaluate the miRNA overexpression after transfection, we quantified the level of miR-15a and miR-196b in the H929 and MM1S cell lines (Supplementary Fig. S2). The range of miR-15a and miR-196b overexpression was not higher than the expression levels of other miRNAs physiologically present in the multiple myeloma cells (Supplementary Table S5). In spite of the lower miR-15a expression in MM1S than in H929 cell line, the decrease of cyclin D2 protein was higher in MM1S cells. Although the upregulation of miR-196b after ectopic transfection was similar in both cell lines, the decrease of cyclin D2 protein was stronger in MM1S than in H929 cell line. To further rule out the possibility that off-target effects were influencing the results, the concentration of miRNA in each transfection was lowered to 10 nmol/L in MM1S cells. The effect was similar to 50 nmol/L concentrations (Fig. 2C and D for densitometric analysis).

Although the aforementioned results suggest the possible role of miRNA downregulation in cyclin D2 overexpression in multiple myeloma, some of the findings were, to a certain extent, contradictory. Thus, miR-19a, 19b, and 20a that were shown to interact with CCND2 3′UTR in vitro did not induce downregulation of cyclin D2 in the multiple myeloma cell lines expressing these miRNAs. On the other hand, the ectopic expression of miRNAs 15a, 135b, 196b, 214, and 320 downregulated in multiple myeloma cell lines and whose interaction with CCND2 at its 3′UTR was demonstrated by luciferase assays, did not provoke a decrease in cyclin D2 mRNA and protein level in all the cell lines explored. These unexpected findings prompted us to explore the possibility that CCND2 post-transcriptional repression induced by miRNAs could be interfered. It has been reported in several cancer cell lines that CCND2 was subject to changes in 3′UTR length by means of APA with the consequent modification of 3′UTR repressive elements (9).

In order to investigate the presence of a full-length CCND2 mRNA and a shorter isoform in myeloma cells, two qRT-PCR assays were designed: one for the detection of both the long and the short isoforms, and the second for targeting the sequence at 3′UTR that was lost in the truncated form (Fig. 2A). The percentage of the two CCND2 mRNA isoforms was calculated in 31 multiple myeloma primary samples that expressed CCND2 mRNA. Most of them exhibited both CCND2 isoforms, although only long form was detected in five samples (Fig. 3A). The short mRNA of CCND2 was also detected in the six multiple myeloma cell lines expressing CCND2 (Fig. 3B).

To validate the PCR results, Northern blot was carried out in multiple myeloma cell lines. As shown in Fig. 3C, the presence of two isoforms of CCND2 mRNA, the full-length mRNA and its shorten isoform, was observed in multiple myeloma cell lines with cyclin D2 protein. The same approach was performed in three multiple myeloma samples with adequate amount of RNA. Both CCND2 isoforms were detected in one patient (MM#1), although only the full-length isoform was present in the other (Fig. 3C). Nested 3′RACE PCR using CCND2-specific forward primer and oligo-T anchor reverse primer generated the expected size product for the short 3′UTR isoform (Fig. 3D). Sequencing of this PCR product verified the presence of proximal poly(A) site (AATAAA) and the polyA tail, which demonstrated that this isoform was generated by polyadenylation. Moreover, the sequence was uniquely aligned to CCND2 mRNA 3′UTR (NM_001759.3; Supplementary Fig. S3).

The quantification of full-length and short mRNA of CCND2 by qRT-PCR approach showed that most of the patients and all the cell lines displayed both the transcripts. In order to find out if the presence of both short and long CCND2 3′UTR occurs in the same cell, or rather there are subpopulations of cells with short and long 3′UTR, two-color mRNA FISH was performed. The coding sequence of CCND2 and the 3′UTR fragment lost in the short isoform were differentially stained (Supplementary Fig. S4). Consistent with the qRT-PCR results, H929 cells showed brighter fluorescence of CCND2 CDS than JJN3 cells, indicating the predominance of short 3′UTR in H929 cells. Apparently, all the cells showed uniform and homogenous staining and different subpopulations with more abundance of one of the isoforms were not identified. Although the number of cells in patients was much lower, the two multiple myeloma samples showed similar results. As it was expected, KMS12BM cells did not express detectable levels of CCND2 mRNA.

To determine if the short 3′UTR isoform of CCND2 could be responsible for higher expression of CCND2 coding region in multiple myeloma cells, we analyzed the CCND2 expression by qRT-PCR in all multiple myeloma samples. Then, the 31 multiple myeloma samples expressing detectable level of CCND2 were grouped regarding predominant expression of one out of the two isoforms different in length. Thus, those multiple myeloma samples whose percentage of long isoform was detected above 50% were considered as samples with predominant long 3′UTR, and those whose long isoform percentage was lower than 50% were grouped as samples with predominant short 3′UTR. The statistical analysis revealed that multiple myeloma samples with predominant short 3′UTR expressed significantly higher levels of CCND2 than those with long 3′UTR (Fig. 4).

Because one of the functional consequences of the shortening of CCND2 3′UTR is the loss of miRNA sites (9), we investigated if this possibility could be the reason of the escape of CCND2 from miRNAs validated in the present study. To confirm that the form of cyclin D2 with truncated 3′UTR was not to be regulated by miRNAs that repress the full-length form, luciferase reporter assays in HEK923 cells with a plasmid containing the short 3′UTR isoforms of CCND2 was performed. The results showed that none of the miRNAs downregulated the luciferase activity at the same level as for the luciferase reporter assay with long isoform, suggesting that the higher level of CCND2 in multiple myeloma with shorter 3′UTR was due to inefficient downregulation by miRNAs (Supplementary Fig. S5)

Among the 10 multiple myeloma with intermediate CCND1 levels and detectable CCND2 expression, the prevalent form of the latter was the long isoform in 80% of the cases (8/10), although only 20% (2/10) of multiple myeloma with high CCND1 expression exhibited short isoform of CCND2 mRNA. Additionally, by using U266 cells that coexpress CCND1 and CCND2, we confirmed previous observations indicating that the CCND1 silencing led to increased cyclin D2 protein levels (Fig. 5A and Supplementary Fig. S6A for densitometric quantification; ref. 19). Likewise, CCND2 mRNA levels were found to be increased 72 hours after CCND1 silencing (Supplementary Fig. S6B). One possibility raised by these findings was that polyadenylation/cleavage reaction of CCND2 mRNA could be modulated by cyclin D1. Accordingly, we found a shortening of CCND2 mRNA after CCND1 silencing (Fig. 5B).

In order to confirm this finding and to find out whether CCND3 could also be involved in the regulation of CCND2 expression, the JJN3 cell line was nucleofected with plasmids containing CCND1 and CCND2. We have observed that both cyclin D1 and cyclin D3 overexpression resulted in the decrease of cyclin D2 at the protein and mRNA levels (Fig. 5C and Supplementary Fig. S6C–S6E). Moreover, this effect was accompanied by an increase in the proportion of long CCND2 mRNA isoform (Fig. 5D). Overall, these results suggest a mechanism of partial cross-regulation of D-cyclins in multiple myeloma.

Although CCND1 might have a role in the downregulation of CCND2 through APA-mediated shortening of CCND2 3′UTR, this mechanism is unlikely to be active in those multiple myeloma with t(11;14), which lack CCND2 expression. Therefore, we hypothesized that epigenetic silencing, such as DNA methylation, could be involved in the abrogation of CCND2 expression in multiple myeloma cells with high level of CCND1 because of t(11;14). To test this possibility we analyzed two regions of CCND2 promoter that contain CpG islands located −642 nt (25 CpG) and −19 nt (20 CpG) before transcription start site (TSS). A clear methylation pattern of the CpG island more proximal to the TSS of CCND2 was identified in cell lines carrying t(11;14) (Fig. 6A). Thus, KMS12BM and KMS12PE, which are derived from the same patient, displayed a similar pattern of methylation with 54% and 44% of the CpG island methylated, respectively (Fig. 6B and C). For the remaining two cell lines, SKMM2 and XG-1, methylation of the distal part of CpG island was observed (26% and 19%). In contrast, cells with strong CCND2 expression, like H929 or MM1S, displayed 3.6% and 2.1% of methylated CpG island, respectively, and U266 coexpressing CCND1 and CCND2, only had 3.6% of methylation.

CCND2 expression has previously been shown to be upregulated in most of multiple myeloma samples without CCND1 or CCND3 overexpression (1). The mechanisms leading to CCND2 dysregulation are not well characterized in some multiple myeloma cases. In contrast to other D-cyclins, no aberrations involving the CCND2 locus have been detected that could explain its overexpression in multiple myeloma. Up to now, only transcriptional mechanisms have been described to enhance CCND2 expression in multiple myeloma. Thus, MAF and MAFB transcription factors directly target CCND2 in about 7% of tumors (5). In addition, the transcription factor ZKSCAN3 has also been described to participate in CCND2 regulation in multiple myeloma (6). The present study shows that other mechanisms like modifications in the normal post-transcriptional regulation of CCND2 could participate in the upregulation of CCND2 in a group of multiple myeloma.

Our results confirmed that ectopic transfection of multiple myeloma cells with several miRNAs (miR-15a, 135b, 196b, 214, and 320) downregulated in multiple myeloma, decreased the level of cyclin D2 by directly targeting CCND2 3′UTR. However, this effect was not observed in all the cell lines. This fact together with the inability of miR-19a, 19b, and 20a to downregulate CCND2 in those multiple myeloma cell lines expressing these miRNAs, suggested the possible disruption of miRNA target sites. In fact, we detected the presence of short CCND2 mRNA, both in multiple myeloma cell lines and primary cells, using four different methodological approaches: qRT-PCR, Northern blot, mRNA FISH, and 3′RACE PCR with product sequencing.

The short CCND2 isoform was observed by qRT-PCR in the majority of patients and in all multiple myeloma cell lines expressing CCND2. This finding was confirmed by Northern blot results. Interestingly, the H929 cells in which the longer CCND2 3′UTR was less predominant compared with other multiple myeloma cell lines, barely displayed modification of cyclin D2 protein expression after miRNAs transfection. The level of cyclin D2 protein was higher in the H929 cell line than in the other cell lines, despite the CCND2 mRNA expression was lower than in MM1S, MM1R, or RPMI-8226. It is possible that other regulatory mechanisms, like translational and post-translational modifications participate in this surprising finding. The abundance of each CCND2 mRNA isoform was also assessed by two-color mRNA FISH designed to discriminate the two mRNA different in length. This approach also enabled us to notice that no subpopulation of cells distinguishable by the load of one isoform with respect to another was present. Apparently, clonal heterogeneity identified in other genomic context of multiple myeloma was not observed in the abundance of short and long CCND2 isoforms (20, 21). It has been shown for several genes, such as CCND1, CCND2, IMP-1, DICER, CDC6, or CYP450, the shortening of their 3′UTR region (9, 11, 12, 22). Genomic deletions of the 3′UTR or point mutations in the 3′UTR region are genetic events that can cause the expression of truncated transcripts through premature cleavage and polyadenylation signals (9). Nevertheless, a large fraction of human genes use alternative cleavage and polyadenylation (APA) to generate mRNA transcripts that differ in the length of their 3′UTR (23). The results obtained by RACE experiments in the multiple myeloma cell lines support the idea that changes in CCND2 3′UTR length are explained by APA. Moreover, the recent genome-wide analyses of hundreds of multiple myeloma samples using massively parallel sequencing failed to detect genomic deletions or mutations in CCND2 (24–26). The functional consequences of 3′UTR shortening is the mRNA stabilization due to the loss of miRNA sites and regulatory elements located in the 3′UTR (9, 10, 23). Accordingly, the luciferase assays using plasmids harboring the truncated CCND2 mRNA strongly confirmed the loss of miRNA sites in the shorter CCND2 mRNA isoform. The short 3′UTRs lacking miRNA-binding sites have been associated with increased expression of different genes at both the mRNA and the protein level (9, 11, 12, 22). In fact, the shorter CCND2 mRNA has been found to increase protein expression in cell lines from various tissues (9). However, there are evidences that support the thesis that increased transcription of genes is positively correlated to the relative usage of proximal poly(A) site (27). Here, we observed a significant higher level of overall CCND2 mRNA expression in those multiple myeloma with greater abundance of the shorter 3′UTR isoform, although it is still an open question whether CCND2 mRNA shortening is the cause or the consequence.

The short form of CCND1 3′UTR has been previously reported in multiple myeloma with t(11;14) (1, 14). The existence of this isoform was correlated with higher CCND1 mRNA level but not with differences in survival (14). Short CCND1 mRNAs are also observed in MCL whose genetic hallmark is t(11;14). The mechanisms underlying the truncation of CCND1 3′UTR in MCL are both genomic deletions of the 3′UTR region and premature polyadenylation signal creation by point mutations (28). In contrast to multiple myeloma, the presence of short CCND1 3′UTR was correlated with poor survival of MCL patients (13).

D-type cyclins are highly homologous proteins and there is a growing body of evidence that the functions of the D cyclins are mostly exchangeable. During mouse development, the three D-type cyclins are expressed following an often mutually exclusive pattern and their function may be tissue specific (29, 30). In keeping with this observation, simultaneous overexpression of CCND1 and CCND2 in multiple myeloma is an infrequent finding (1, 31). Because the 3′UTR length can be considered as a mechanism of gene expression regulation, we investigated if it can be involved in a potential cross-regulation between cyclin D1 and D2. It was previously observed in U266 cells, which express both cyclins D, that when cyclin D1 was silenced the level of cyclin D2 increased (19) and our results ascertain these data. Similar results have also been reported in MCL (32). Although the explanation of this effect is probably unknown, our results showing a CCND2 mRNA shortening after CCND1 silencing suggest that cyclin D1 could downregulate CCND2 level by modification of polyadenylation/cleavage reaction. Moreover, the increased proportion of long CCND2 isoform after cyclin D1 or cyclin D3 overexpression along with simultaneous decrease of cyclin D2 protein level in the JJN3 cell line further supports the significance of this mechanism in the regulation of cyclin D2 expression. The signals that provoke mutual exclusion of cyclins D during development and in adult tissues remain elusive. Studies with mouse models point out the existence of a negative feedback loop, in which the dominant D-type cyclin represses the expression of remaining D-cyclins. These studies revealed that the molecular mechanism leading to cyclins D repression functions differently in distinct tissues, in some of them by modification of mRNA level and in others by possible post-transcriptional mechanisms (30). Here, we provide for the first time data about a possible mechanism that can induce the downregulation of CCND2 in CCND1 and CCND3 overexpressing multiple myeloma cells. Different signals provoking 3′UTR shortening have been described in several cell types: for example, activation of T and B lymphocytes results in increased usage of proximal polyadenylation sites (10, 33); 17-β estradiol(E2) induces APA to activate proto-oncogenes in breast cancer cells (12); and E2F transcription factors mediate enhanced APA usage in human cellular models (34).

Because CCND2 expression is undetectable in myeloma cells with t(11;14), we extended our investigation to explore if DNA methylation might play a role in abolishing CCND2 expression. We observed that the CpG island more proximal to CCND2 TSS was highly methylated just in multiple myeloma cell lines with t(11;14). Several studies have indicated that methylation of CCND2 is responsible for absence of its mRNA and protein in breast, prostate, lung, and hepatocellular cancers (35–38).

In summary, this study demonstrates that post-transcriptional modifications play a role in CCND2 expression regulation in multiple myeloma. Downregulation of specific miRNAs directly targeting CCND2 contributes to overexpression of CCND2 in a set of multiple myeloma. Moreover, the shortening of CCND2 3′UTR by APA with the consequent loss of miRNA binding sites is also participating in CCND2 upregulation. In fact, this mechanism seems to play a decisive role in the regulatory network between CCND1 and CCND2 in multiple myeloma. Further studies are needed to unravel the molecular basis of some of these findings, and to define among the mechanisms reported so far, which are the most critical that determine the expression of CCND2 in myeloma.

J.F. San-Miguel is a consultant/advisory board member for Bristol-Myers Squibb, Celgene, Janssen, Millennium, MSD, Novartis, and Onyx. No potential conflicts of interest were disclosed by the other authors.

Conception and design: I. Misiewicz-Krzeminska, P. Krzeminski, J.L. García, R. García-Sanz, J.F. San Miguel, N.C. Gutiérrez

Development of methodology: I. Misiewicz-Krzeminska, M.E. Sarasquete, C. Vicente-Dueñas, P. Krzeminski, K. Wiktorska, D. Quwaider, E.A. Rojas, J.L. García, I. Sánchez-García, R. García-Sanz

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): I. Misiewicz-Krzeminska, P. Krzeminski, K. Wiktorska, R. Corral, A.A. Martín, F. Escalante, A. Bárez, I. Sánchez-García, N.C. Gutiérrez

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): I. Misiewicz-Krzeminska, C. Vicente-Dueñas, P. Krzeminski, L.A. Corchete, D. Quwaider, E.A. Rojas, R. García-Sanz, J.F. San Miguel, N.C. Gutiérrez

Writing, review, and/or revision of the manuscript: I. Misiewicz-Krzeminska, C. Vicente-Dueñas, P. Krzeminski, I. Sánchez-García, R. García-Sanz, J.F. San Miguel, N.C. Gutiérrez

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.A. Martín

Study supervision: J.F. San Miguel, N.C. Gutiérrez

The authors thank Isabel Isidro, Teresa Prieto, and Vanesa Gutiérrez for their technical assistance in the purification of multiple myeloma samples and María Angeles Hernández for the help in mRNA FISH assays.

This study was partially supported by Instituto de Salud Carlos III (Fondo de Investigaciones Sanitarias: PI08/0568 and PI13/00111), Asociación Española Contra el Cancer (AECC, GCB120981SAN), “Gerencia Regional de Salud, Junta de Castilla y León” (GRS 702/A/11, BIO/SA57/13 and BIO/SA35/14) grants, the Spanish Myeloma Network Program (RD12/0036/0058) and the INNOCAMPUS Program (CEI10-1-0010). I. Misiewicz-Krzeminska was supported by Instituto de Salud Carlos III (PS09/01897). M.E. Sarasquete was supported by Contrato Miguel Servet (CP13/00080). Research in the I. Sánchez-García group was supported by FEDER and MINECO (SAF2012-32810, and Red de Excelencia Consolider OncoBIO SAF2014-57791-REDC), Instituto de Salud Carlos III (PIE14/00066), NIH (R01 CA109335-04A1), “Junta de Castilla y León” (BIO/SA32/14 and CSI001U14) grants, the Fundacion Inocente Inocente, the German Carreras Foundation (DJCLS R13/26) and by the ARIMMORA project (European Union's Seventh Framework Programme (FP7/2007-2013) under grant agreement no. 282891). Research in the C.V. Dueñas group was supported by “Miguel Servet” Grant (CP14/00082 AES 2013-2016) from the Instituto de Salud Carlos III (Ministerio de Economía y Competitividad). I. Sánchez-García is an API lab of the EuroSyStem project and a partner within the Marie Curie Initial Training Network DECIDE.

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.