A compendium of myeloma-associated chromosomal copy number abnormalities and their prognostic value

Myeloma is thought to result from the transformation of a proliferative “plasmablastic” cell located in the germinal center.¹  The progeny of this cell migrate to specialized niches in the bone marrow where they mature toward terminally differentiated antibody secreting plasma cells. Genetic alterations in the myeloma clone can impact on pathways important in normal plasma cell biology leading to significant changes in cellular behavior, disease progression and changes in clinical outcome. Cytogenetics and positional cloning have given many insights into myeloma pathogenesis and the characterization of abnormalities focused attention on recurrent chromosomal translocations and cyclin D deregulation as a classification system.² 

Additional information about the genetics underlying the deregulation of normal plasma cell functional pathways is being derived from the characterization of the regions of recurrent copy number alteration (CNA), which are frequent in myeloma compared with other hematologic malignancies, and we have previously reported on the importance of deletions of CDKN2C at 1p32 impacting on the G₁/S transition.^3,4  High-resolution single nucleotide polymorphism (SNP)–based gene mapping is a tool that is highly effective in this respect and when combined with expression and mutation analysis can identify candidate genes important in myeloma pathogenesis, as has been shown for other tumor types.⁵  Using this approach we identified the importance of CYLD deletion and mutation while others identified deletion of TRAF2, BIRC2, and BIRC3 that when considered together are consistent with activation of the nuclear factor-κB (NF-κB) pathway in a proportion of myeloma cases.^6-8 

Information on CNAs in myeloma samples have been reported using karyotyping, fluorescence in situ hybridization (FISH), array comparative genomic hybridization (aCGH), and SNP-based arrays.^6,8-13  However, to date there have been no detailed reports of CNAs in myeloma. Here we describe and annotate the major CNAs found in a large series of myeloma samples. We have analyzed gains and deletions in presenting myeloma samples using high resolution SNP-based mapping arrays with which expression array data have been integrated to delineate genes whose expression may be affected by these abnormalities. We have extended this work further by examining the clinical importance of relevant lesions in a large set of uniformly treated clinical samples from the Medical Research Council (MRC) Myeloma IX study.

Bone marrow aspirates were obtained from newly diagnosed patients with multiple myeloma, entered into the UK MRC Myeloma IX study, after informed consent. Plasma cells (PCs) were selected to a purity of > 90% using CD138 microbeads and magnet-assisted cell sorting (Miltenyi Biotec) and analyzed by FISH (n = 1177), expression array (n = 258), and/or mapping array (n = 114).¹⁴  Sample characteristics are summarized in Table 1. RNA and DNA were extracted using commercially available kits (RNA/DNA mini kit or Allprep kit; QIAGEN) according to manufacturers' instructions. Matched germline DNA from 80 patients was extracted from peripheral white blood cells, using the Flexigene kit (QIAGEN).

Interphase FISH analysis was performed on CD138-selected PCs using the micro-FISH technique and probes that have previously been documented.^15-17  Briefly, probes to detect t(4;14), t(6;14), t(11;14), t(14;16), t(14;20), del(1p32.3), gain 1q, del(17p), and hyperdiploidy (essentially defined by gain of any 2 of chromosomes 5, 9, and 15) were used to identify abnormalities. In addition, analysis of 8p was performed using the Provision probes (Abbott Molecular) that interrogate 8p22 (LPL), 8p11.1-q11.1 (centromere), and 8q24.2-q24.3 (MYC).

GeneChip Mapping 500K Array set and U133 Plus 2.0 expression arrays (Affymetrix) were performed as previously described.^3,4,7,18  Nine samples in this study had also been analyzed in a previous 50K mapping array dataset.³  Expression array raw data were normalized in GeneSpring 7.3 (Agilent Technologies) using per chip normalization to the 50th percentile and per gene normalization to the median. Comparisons between samples with a CNA and those without were performed using a 1-way analysis of variance and filtered to remove those with fold change < 1.2. Gene lists were further filtered to solely include those within the region of CNA, thus identifying genes whose expression is directly affected by the alteration. For mapping array data the SNP genotypes and inferred copy number were obtained using GTYPE and dChip, exactly as previously described.^4,7  Data have been deposited into GEO under accession number GSE21349. Homozygous deletions were identified as having an inferred copy number < 0.7, hemizygous deletions between 0.7-1.6 and gains > 2.4.

The MRC Myeloma IX trial recruited 1970 newly diagnosed patients and comprised 2 patient pathways; the first for older and less fit patients and the second for younger, fitter patients. Younger, fitter patients received autologous transplantation after induction with cyclophosphamide, thalidomide, and dexamethasone (CTD) or cyclophosphamide, vincristine, doxorubicin, and dexamethasone (CVAD), deemed the intensive pathway. The nonintensive pathway consisted of a randomization to either attenuated CTD (CTDa) or melphalan and prednisolone (MP). All patients were then randomized to thalidomide maintenance or no thalidomide maintenance. The trial was approved by the MRC Leukaemia Data Monitoring and Ethics committee (MREC 02/8/95, ISRCTN68454111).

Kaplan-Meier overall survival (OS) curves were calculated using Bioconductor and the survival package¹⁹  after a median follow-up of 3.7 years. The difference between curves was tested using the log-rank test within Bioconductor. A threshold of significance was taken as P < .05.

Presenting myeloma cases were studied using the 500K SNP-based mapping GeneChip (n = 114), of which 80 had matched nontumor DNA available to compute accurate CNAs and acquired loss of heterozygosity (LOH) in the tumor (Figure 1 and Table 2). We went on to determine whether these common CNAs cosegregate as well as their impact on prognosis (Table 3). Detailed analysis of the regions of loss and gain can be found in supplemental Figure 1 (available on the Blood Web site; see the Supplemental Materials link at the top of the online article).

In agreement with previous observations²⁰  there was an association between IgH translocations and del(13q) (χ²P = 5.33 × 10⁻²⁴), an association of 1q gain with del(13q) (χ²P = .000429), an inverse association of 1q gain with gain of 11 (χ²P = .000575), and an inverse association of del(13q) with gain of 11 (χ²P = .000184). There was an association of t(4;14) with del(18) (χ²P = .001865) and an inverse association of t(4;14) with del(16q) (χ²P = .00 287).

Gain of the odd-numbered chromosomes is a characteristic of hyperdiploid myeloma and when ranked by the number of full gains chromosome 15 was most often gained (42/53 samples), followed by chromosomes 9 (41 samples), 5 (38 samples), 19 (38 samples), 3 (31 samples), 11 (28 samples), 7 (24 samples), and 21 (14 samples). In a larger dataset of 1004 samples, hyperdiploidy, as determined by FISH, was not associated with a significantly better OS when compared against all nonhyperdiploid samples (49.7 vs 43.7 months, P = .15, supplemental Figure 2). In addition, gain of chromosome 5 did not stratify hyperdiploid cases, because cases with and without the abnormality showed similar OS times (William J. Tapper, L.C., G.P.D., Z.J.K., D.M.S., Alex J. Szubert, W.M.G., Sue E. Bell, Graham H. Jackson, J. Anthony Child, G.J.M., F.M.R., manuscript in preparation). We were able to divide hyperdiploid samples into those with gain of chromosome 11 and those with gain 1q and del(13q). A clear clinical disadvantage was seen in hyperdiploid samples with gain 1q, with or without del(13q), compared with those with normal 1q (supplemental Figure 2), as has been previously reported.²⁰ 

In this dataset 34 cases (29.8%) had any deletion of 1p, of which 29 had an interstitial deletion, and 5 had the whole arm deleted. Copy number–neutral loss of heterozygosity (LOH), or uniparental disomy (UPD), was seen on 1p with 3 cases having complete UPD of 1p and an additional 2 cases with partial UPD extending from 1p36.13 and 1p36.12 to the telomere.

There were 4 minimally altered regions on 1p. We have previously identified the region at 1p32.3, containing FAF1 and CDKN2C, which is associated with adverse overall survival (OS)⁴  and had a deletion in 18 samples. In line with these observations, 6 cases in this dataset had homozygous deletion of both CDKN2C and FAF1 and could not delineate which gene was the target (Figure 2A); however, others have proposed CDKN2C as the sole target.²¹  Twelve additional cases had hemizygous deletion. Confirming our previous study of this region we expanded this analysis further using FISH to examine 510 newly presenting cases in the intensive path of the trial and show a significant effect on OS (Figure 3A, P < .001, median OS 34.5 vs > 70 months). It is not clear whether the effect on survival is solely due to deletion of 1p32.3 as cases may also have additional deletions affecting other regions and genes on 1p.

The most commonly deleted region on 1p was at 1p22.1-p21.3 that contains 35 genes and was deleted in 25 cases (22%). Comparison of expression data from deleted and nondeleted samples showed that 2 adjacent genes, MTF2, and TMED5, were down-regulated in this region. Another deleted region lies at 1p12, and was hemizygously deleted in 20 cases, with homozygous deletion of the FAM46C locus in an additional 2 cases. A fourth deleted region lies within 1p31.1 and encompasses 6.6 Mb containing 14 genes with only 1 gene underexpressed in deleted samples, USP33, a ubiquitin-specific protease.

Gain of 1q was identified in 39/114 (34.2%) samples of which 2 had complete LOH. In myeloma, gain of 1q has an adverse effect on OS²²  and in this data series, studying 531 samples in the intensive arm of the trial using a FISH probe specific for CKS1B, this was confirmed (P < .001, median OS 52.1 vs > 70 months, Figure 3B). This effect was examined further by removing cases with other adverse cytogenetic factors (del(17p) and translocations involving FGFR3/MMSET, MAF, and MAFB) after which gain of 1q still retained prognostic significance (P = .01, median OS 57.8 vs > 70 months, supplemental Figure 2), suggesting that gain of 1q21 is an independent prognostic factor.

One case had gain of all of 1q but with acquired LOH between 1q21.1-q25.2, and an additional 5 cases had UPD of all of 1q (2 of which were UPD of the entire chromosome). A minimally amplified region was identified between 1q21.1-q23.3 containing 679 genes. When expression array data were clustered according to copy number status (2 vs > 2 copies), 9 genes had a greater than 2-fold increase in expression including IL6R, CKS1B, KCNN3, ANP32E, FCRL2, S100A4, HAX1, PEX11B, and C1orf43. Of these, IL6R, KCNN3, and CKS1B had a greater than 2.5-fold increase in expression. The expression data for these genes was ranked in 258 samples and the lowest expressing quartile compared with the highest expressing quartile and Kaplan-Meier survival curves generated for each gene. From this analysis, only ANP32E had a statistically significant effect on OS (P = .004, median survival 28.9 vs 58.9 months), and validated in another dataset (GSE2658, supplemental Figure 3).²³ 

Deletion of 17p was seen in 8 samples (7%) with UPD in a further 2 samples. One sample had a small deletion of 0.55 Mb containing 30 genes centered on TP53. Expression of TP53 in the del(17p) samples was significantly lower than in nondeleted samples. Four further genes within the minimally deleted region (MDR) were expressed at lower levels in the deleted samples: TMEM107 (2.2-fold), MGC10744 (1.8-fold), SAT2 (1.4-fold), and EIF5A (1.3-fold). Loss of 17p (using a TP53 FISH probe, n = 501) was associated with an adverse OS (median OS 40.9 vs 67.8 months, P < .001, Figure 3D) in the intensive arm of the trial. We did not identify homozygous deletions of TP53, but mutation analysis of cases with LOH identified TP53 mutations in 27% of the samples (K.D.B., F.M.R., L.C., William J. Tapper, G.P.D., R.K.M.P., D.M.S., Z.J.K., W.M.G., B.A.W., Christopher P. Wardell, D.G., F.E.D., G.J.M., manuscript in preparation), directly implicating TP53 as the crucial gene at this site.

The most frequently deleted region on chromosome 4 was of the telomere of the short arm as a consequence of the unbalanced t(4;14) translocation that abrogates FGFR3 expression. Of the 17 samples with t(4;14), 4 had deletion of the tip of the short arm (4p16.3) at the FGFR3 locus (Figure 4A-B). An IgH-MMSET cDNA polymerase chain reaction (PCR) showed that all but 1 sample with del(4p16.3) had an MMSET translocation product (Figure 4C). The remaining sample did not give an IgH-MMSET cDNA product but the mapping data indicate that the breakpoint in this sample is within FGFR3, consistent with a longer cDNA product. A copy number breakpoint at the FGFR3 locus resulted in loss of expression of FGFR3 (Figure 4D).

We found deletion of the reciprocal translocation chromosome 14 in all cases with deletion of FGFR3. However, the IGH locus to the telomere retained a normal copy number, confirming that the der(14) chromosome is lost after the translocation occurs (Figure 4E). This results in loss of the FGFR3 allele that is under the control of the IGH enhancer.

Deletion of 8q was present in 14% and included 3 samples with deletion of all of chromosome 8, but most deletions were small and interstitial. There were 10 samples (8.7%) with a copy number breakpoint at the MYC locus, 5 of which were small regions of gain and may be due to a translocation or gene duplication. Expression of MYC in these samples was high compared with those without breakpoints, indicating that the CNAs directly affect expression of the oncogene.

Also on chromosome 8, between 75% and 100% of 8p was deleted in 29 of 114 (25%) samples. The use of 2 of the interstitial deletions allowed the identification of a 23.1 Mb MDR between 8p23.1-p12, containing 237 genes. By differential expression analysis these 237 genes were limited to 36 genes that are underexpressed (> 1.2-fold, P < .05) and of these 14 are underexpressed > 1.5-fold. The most underexpressed genes are ZDHHC2 (1.9-fold), FDFT1 (1.8-fold), CNOT7 (1.8-fold), PPP2R2A (1.7-fold), and PPP2CB (1.6-fold). FISH for del(8p) was performed on 152 presenting cases, and while it was deleted in 29% of samples it did not impact on survival in this dataset (supplemental Figure 2).

Gains of chromosome 11 were common (47/114, 41%), and 31 cases had gain of the whole chromosome. 7 samples had copy number increases at the CCND1 locus, of which 6 had a t(11;14) suggesting they occur during the translocation event. Loss of der(14) was not seen in these samples and suggests that CNAs in this region occur through more complex chromosomal rearrangements compared with CNAs surrounding the t(4;14) breakpoint.

Additional abnormalities on chromosome 11 center on the BIRC2/BIRC3 loci, where 8 samples (7%) had deletion and of these 6 had homozygous deletions. Analysis of the 7 genes (ANGPTL5, KIAA1377, C11orf70, YAP1, BIRC3, BIRC2, TMEM123) within the homozygously deleted region indicates that BIRC2, BIRC3, and TMEM123 are the likely targets as KIAA1377, C11orf70 and YAP1 were not expressed in any myeloma sample (Figure 2B). Recent data focusing on the NF-κB pathway suggest that BIRC2 and BIRC3 are the most likely targets of the deletion.⁸ 

The whole of chromosome 14 was deleted in 14 samples (12%), with an additional 18 samples having smaller deletions, and a further 15 samples with deletion of 14q32.33-qter due to loss of the IGH locus. An 11.86 Mb MDR existed at 14q24.1-q31.1 containing 103 genes between WDR22 and C14orf145. Of these 103 genes, 33 were underexpressed in the 27 samples with deletion including ACOT2, ZNF410, SFRS5, SNW1, PSEN1, WDR21A, C14orf156, and ALKBH1 that were under-expressed > 1.5-fold. Homozygous deletion of TRAF3 and AMN, outside of the MDR, was seen in 2 samples and another 17 samples had hemizygous deletion but had no effect on OS.

The t(14;16) translocation results in over-expression of MAF and disruption of WWOX that is located at the fragile site FRA16D. The WWOX locus was deleted in 33 samples with UPD in 10 additional samples. Differential expression analysis between deleted and nondeleted samples showed that deletion of 16q had the largest effect on WWOX with a 2.4-fold reduction in expression. CHD9, MAF, and CDH1 were also affected > 1.6-fold.

However, WWOX and MAF are not the sole targets of CNAs on chromosome 16. Chromosome 16 was fully deleted in 4 samples with the entire long arm deleted in a further 21 samples and interstitial deletion of 16q in another 15 samples. UPD of 16q was present in 10 samples of which 3 also had UPD of 16p. These data are consistent with alterations of 16q being present in 43% of the total cases. Several regions of interest on 16q include 16q12.1 that was deleted in 33 samples with UPD present in 8 additional samples. The smallest region of LOH at this locus was a hemizygous deletion of 3.05 Mb that contained a homozygous deletion of 442 kb containing 4 genes: NKD1, SNX20, NOD2, and CYLD, of which only CYLD was differentially expressed (1.6-fold). Another sample had a homozygous deletion of 212 kb containing only NOD2 and CYLD. These data are consistent with involvement of CYLD, a tumor suppressor gene and negative regulator of the NF-κB pathway, as the critical gene within this region and is mutated in 21% of samples with LOH.^6-8 

The second region is present at 16q22.1, comprises 1.05 Mb of DNA, and was deleted in 31 samples with UPD in 10 samples. This region contained 25 genes, including E-cadherin, and of these genes 9 were differentially expressed between 1.21- to 1.5-fold lower in deleted samples: THAP11, NUTF2, EDC4, PSMB10, DPEP2, NFATC3, SLC7A6, SLC7A60S, and ZFP90.

Deletions on chromosome 6 were observed in 40/114 cases (35%) and predominantly found on 6q (38/40, 33.3%) with only 2 samples having deletion of the entire chromosome. The most commonly deleted region centered at 6q25.3 (155.35-161.74 Mb) that was deleted in 34 samples. This region contains 32 genes between TIAM2 and PARK2, the location of the fragile site FRA6E. Differential expression of these samples shows 17 genes within this region had significant differences (P < .05) with a fold change > 1.4. Of the 17 genes, the 3 with the greatest reduction in expression are IGF2R (1.9-fold), TFB1M (1.68-fold), and WTAP (1.7-fold).

Gain on chromosome 6 was seen in 26/114 cases (22.8%) and was predominantly limited to 6p (21/26) with gain of the whole chromosome in 5 samples. Gains of 6p mostly extended over the entire short arm, but a minimally amplified region of 3.1 Mb was identified, at 6p22.3, and was gained in 20 samples. This region contains 9 genes, but expression analysis of these samples showed no significant overexpression of any of the genes. The next commonly gained region extended from 6p22.3-p21.31, was 20.5 Mb and was gained in 19 samples. This larger region contains 657 genes and extends from JARID2 through histone locus 1 to SCUBE3 within the MHC region; 34 of these genes are up-regulated in samples with gain of this region compared with those without.

On chromosome 12, 4 samples had a gain, 2 samples had UPD, and 25 samples had a deletion. The most frequently deleted region (16/114, 14%) was at 12p13.1-p13.2 and contained 49 genes including CDKN1B, a negative regulator of cell cycle. However, analysis between samples with and without a deletion highlighted only 3 differentially expressed genes, CREBL2, ETV6, and MANSC1, with > 1.5-fold decrease in expression. In this limited dataset no prognostic value of del(12p) was seen. The second most frequently deleted region (9/114, 7.9%) lies at 12q23.2 and contained 3 genes, IGF1, PAH and ASCL1, within a minimally deleted region of 696 kb, but none were underexpressed in the deleted samples. Unlike a previous study¹⁰  we found no prognostic significance of del(12p) in a cohort of 803 samples examined by FISH (William J. Tapper, L.C., G.P.D., Z.J.K., D.M.S., Alex J. Szubert, W.M.G., Sue E. Bell, Graham H. Jackson, J. Anthony Child, G.J.M., F.M.R., manuscript in preparation).

Chromosome 13 has been extensively investigated both as a prognostic factor and as a location for tumor suppressor genes. In this dataset deletions were seen in 67/114 samples (59%) of which 87% (58/67) were of the complete long arm. UPD of chromosome 13 was seen in 2 samples and only 1 sample had gain of the chromosome. An MDR defined by 4 samples with interstitial deletion lay between 13q14.11-q14.3 (8.8 Mb) and contained 68 genes. Of these, 16 were significantly underexpressed in the deleted samples, and RB1 was the most underexpressed (1.9-fold), followed by EBPL (1.7-fold), RNASEH2B (1.6-fold), and RCBTB2 (1.6-fold). This region also contained the microRNA genes mir-16-1 and mir-15a, which may also be critical in the pathogenesis of the disease. Homozygous deletions were seen in 4 samples, but only 1 of these fell within the MDR containing RB1, P2RY5, and RCBTB2, and was 213.5 kb in length. The 3 remaining homozygous deletions contained no known genes. Using FISH probes for the RB1 and D13S319 loci in 1058 presenting samples we could not define an impact on survival for del(13q) when del(17p) and t(4;14) samples were excluded (Figure 3C).

Deletions of chromosome 20 were most often seen in the short arm (14 samples) and were deletions to the telomere. Deletions of the long arm were mostly interstitial with no discernible MDR. Loss or gain of the whole chromosome was seen in 4 samples and an additional 13 samples had varying sizes of interstitial gain.

Chromosome 22 was completely deleted in 12 cases with an additional 9 cases having interstitial deletions. The IGL locus was also deleted in another 15 cases indicating rearranged Ig light chains. Of those with deletions at IGL, 4 had a homozygous deletion. One sample had UPD of chromosome 22 and 1 sample had gain of the chromosome. Another 11 samples had interstitial gains.

Changes in the X chromosome are of interest because of the obvious sex differences and the potential for imprinting. In this context UPD of X was common both in males and females and is a potentially important mechanism of gene inactivation. One copy of X was lost in 19 female samples, but 4 of these had regions of gain on the remaining copy, resulting in UPD. In contrast, 18 male samples had UPD, resulting from duplication, extending from the telomere of the long arm and 2 samples had UPD of the whole chromosome (1 of each sex). Of the females, 9 samples had gain of X that also extended from the telomere of the long arm. Deletion of the telomeric end of the short arm was seen in 6 female samples. Providing further evidence for the importance of genes carried on the X chromosome, homozygous deletions were seen affecting UTX (histone demethylase), CNKSR2 (regulator of Ras signaling), and HDHD1A (halogenase).

This study has analyzed the myeloma genome to a resolution of 2.5 kb and identified CNAs including gains, deletions and acquired UPD. The common CNAs were found at 1p, 1q, 6q, 8p, 11q, 13q, 14q, 16q, 17p, 20, and 22, along with gain of the odd numbered chromosomes (Figure 1).

UPD is a novel pathogenic mechanism affecting gene copy number and function and was frequently found at 1q (8%), 16q (9%), and X (20%) in myeloma. There are potential candidate oncogenes or tumor suppressor genes at 1q and 16q, and the frequency of UPD on Xq suggests it too may contain relevant genes. The frequency with which UPD is seen in these regions suggests that they must be under some degree of positive selection. In contrast, UPD is rarely seen in other regions of deletion such as 8p, 6q, and 13q, suggesting 2 distinct cellular mechanisms within myeloma.

UPD at 16q is frequent and is also the site of recurrent hemizygous deletions. We hypothesize that an active tumor suppressor gene allele on 16q is deleted and the remaining mutated or inactive allele is duplicated. Consistent with this hypothesis we have identified 2 tumor suppressor genes; CYLD and WWOX⁷ and found mutations in the residual CYLD allele, present in 21% of cases with LOH (data not shown). WWOX is underexpressed in samples with LOH and can also be inactivated by the t(14;16) translocation.

We found 9 samples (8%) with UPD of 1q, even though this region has gains in copy number, which is unusual in cancer cells as UPD is normally associated with regions of deletion. Extra copies of 1q are frequently involved in translocations, although in the case of UPD it may be that isochromosomes are formed in a fashion similar to follicular lymphoma where gains, UPD and isochromosome 6p are detected.^24-26  Isochromosome 1q is known to occur in other cancer types and has been proposed as a marker for aggressive tumors.^27-29 

On the X chromosome, UPD occurs through deletion of 1 copy in females and duplication of the remaining copy and in males the sole copy is duplicated. However, translocations involving the X chromosome have been reported in myeloma and the UPD that extends from the telomere of the long arm may be a result of unbalanced translocations of the region (data not shown).

A further common and important mechanism contributing to myeloma pathogenesis are translocations involving the immunoglobulin locus. In this dataset CNAs are frequently associated with translocations including 4p, 11q, 16q, and 20q. Deletion of 4p occurs in ∼ 25% of t(4;14) samples resulting in loss of the over-expressed FGFR3 allele.³⁰  Deletion and loss of expression correlates exactly with loss of 1 copy of chromosome 14 (with the exception of 14q32.2-qtel) confirming that the method of loss is through deletion of the der(14) chromosome. This is further evidence that the IgH translocation is the primary event and subsequently der(14) is deleted. We have identified evidence for an impact of IGH translocation on chromosomal abnormalities with the MYC locus at 8q affected in 9% of samples, CCND1 at 11q (6%), CCND3 at 6p (4%), and FGFR3/MMSET at 4p (6%), identifying 25% of samples with common CNAs associated with translocations. Potentially, non-Ig translocations may affect progression of multiple myeloma, but solely using SNP mapping arrays we cannot identify these and further information on these await alternate technologies.

Our analysis of the regions with recurrent CNAs reveals that not all regions are of prognostic importance. Those with a prognostic impact on OS include del(1p) (FAF1, CDKN2C), 1q+ (ANP32E, CKS1B), and del(17p) (TP53), as well as the translocation groups t(4;14), t(14;16) and t(14;20) (Table 3). No prognostic impact on OS was found due to del(8p), del(11q), del(12p), del(13q), or del(16q). However, the importance of these latter regions should not be discounted, as some of the datasets were relatively small and were not studied extensively by FISH. The identification of novel genes, which may be important in identifying patients with a poor prognosis, merits further screening of samples for mutational and methylation abnormalities.

In this dataset, gain of the region around CKS1B on 1q by FISH is associated with shorter OS. This region is large and we identified ANP32E as a potentially interesting oncogene, whose increased expression was associated with poor OS. ANP32E is a member of the acid nuclear protein family that has been implicated in histone acetyltransferase inhibitory activity with a role in chromatin remodeling and transcriptional regulation, having similar properties to the more fully characterized and pathogenically important MMSET.³¹ 

Homozygous deletions are important genetic events as they can identify relevant genes within deleted regions as they, by definition, fully inactivate genes contained within them. Homozygous deletions of at least 100 kb in single cases were identified at FAM46C (1p), TSPYL4 (6q), PARK2 (6q), TLR4 (9q), RB1 (13q), WWOX (16q), CDH1 (16q), keratin locus (17q), GSK3A and neighboring genes (19q), UTX (Xp), CNKSR2 (Xp), and HDHD1A (Xp). Frequently occurring homozygous deletions were located at 1p32.3 (FAF1/CDKN2C), 11q (BIRC2 and BIRC3), 14q (TRAF3 and AMN), and 16q (CYLD). The majority of these genes have known relevance to myeloma biology, such as CDKN2C and RB1 (cell-cycle regulation), TRAF3, BIRC2, BIRC3 and CYLD (NF-κB regulation),^6-8 WWOX (apoptosis),^7,32 GSK3A (Wnt signaling),^14,33  and CDH1 (frequently methylated).^34,35 

To determine the potential biologic significance of the CNAs, a pathway analysis of differentially expressed genes within the regions of interest, including homozygously deleted regions, was performed. Using the KEGG pathway analysis tool within DAVID³⁶  we identified the Wnt signaling pathway as being enriched (P = .0026), along with the Apoptosis pathway (P = .046). Genes belonging to the Wnt signaling pathway were NFATC3, PPP3CC, FZD3, CTBP1, SIAH1, PSEN1, GSK3B, CSK22, and CSNK2B. Those involved in apoptosis were PPP3CC, BIRC2, TNFRSF10A, TNFRSF10B, and TRADD. In addition, some of the genes may be more relevant to other pathways, such as BIRC2 and TRADD with the NF-κB pathway. The genes identified in this analysis point toward several biologic networks relevant to myeloma biology. Cell-cycle regulation genes such as CDKN2C, CDKN2A, CDKN2B, CDKN1B, and RB1, are deleted and underexpressed and would be predicted to result in abnormalities of cell-cycle progression, as has been suggested by the role of cyclin D overexpression.

Wnt gene signaling is important in stem cell biology but in this context may be relevant to G₁/S deregulation via the impact of GSK3A directly onto the function of cyclin D1, whereas loss of negative regulators of the NF-κB pathway would facilitate a ligand-independent survival advantage for cells. FAF1, which has also been implicated in the negative regulation of NF-κB and FAS-associated cell death signaling and apoptosis pathways, may also be important. The other genes involved in apoptosis include TNFRSF10A/D potentially implicating TRAIL signaling deregulation in the immortalization of the myeloma clone. TP53 abnormalities are also important in both clonal immortalization and survival after treatment and deletion/mutation of this gene are one of the most prognostically important acquired genetic alterations in myeloma.

Histone methylation or acetylation also plays a key role in the pathogenesis of myeloma. Of particular relevance to this process is MMSET, which is overexpressed due to the t(4;14) translocation.³⁷ MMSET is a histone methyltransferase and is thought to act as a transcriptional repressor through methylation of histone lysine residues.^38-40  In addition to MMSET, UTX has been shown to be deregulated in up to 10% of myeloma samples. UTX is a histone demethylase, performing the opposite function to MMSET, and is either mutated or deleted in myeloma resulting in histone methylation in the myeloma epigenome.⁴¹  We have also identified ANP32E as a potential partner in the regulation of histones in myeloma. ANP32E is a histone acetyltransferase inhibitor (INHAT)³¹  and is over-expressed in samples with gain of 1q. Over-expression of this gene would potentially result in increased histone methylation, therefore adding weight to the theory that histone modification and epigenetics in myeloma are key to the pathogenesis of the disease.

The chromosomal alterations in presenting myeloma cells are diverse and include balanced and unbalanced translocations, UPD, deletions and gains of regions and whole chromosomes, and homozygous deletions of genes important in the pathogenesis of the disease. It is clear that although NF-κB has been implicated in the disease, it is not the sole contributing factor and we have shown other genes and pathways that have an important prognostic impact on the future treatment of myeloma patients.

We thank the staff at the Haematological Malignancy Diagnostic Service, Leeds, the LRF UK Myeloma Forum Cytogenetics Group, Salisbury, and the Clinical Trials Research Unit, Leeds.

Research grants and financial support were received from the Leukaemia Research Fund, Cancer Research UK, the Bud Flanagan Research Fund, the Kay Kendall Leukaemia Fund, the United Kingdom Department of Health, and the Royal Marsden Hospital National Institute for Health Research Centre.

Contribution: B.A.W. performed research, analyzed data, and wrote the paper; P.E.L., L.C., K.D.B., D.C.J., G.P.D., R.K.M.P., Z.J.K., and D.M.S. performed research and analyzed data; N.J.D., M.W.J., and D.G. analyzed data; W.M.G. collated clinical information; F.E.D. designed research; F.M.R. designed and performed research and analyzed data; and G.J.M. designed research and wrote the paper.

Conflict-of-interest disclosure: The authors declare no competing financial interests.

The current affiliation for P.E.L. is Banco Andaluz de Cellulas Madre, Centro de Investigaciones Biomedicas, Granada, Spain.

Correspondence: Dr Brian Walker, Section of Haemato-Oncology, The Institute of Cancer Research, 15 Cotswold Rd, London, SM2 5NG, United Kingdom; e-mail: brian.walker@icr.ac.uk.