Multiomic mapping of acquired chromosome 1 copy number and structural variants to identify therapeutic vulnerabilities in multiple myeloma

Abstract

Purpose:

Chromosome 1 (chr1) copy number abnormalities (CNAs) and structural variants (SV) are frequent in newly diagnosed multiple myeloma (NDMM) and associate with a heterogeneous impact on outcome the drivers of which are largely unknown.

Experimental Design:

A multiomic approach comprising CRISPR, gene mapping of CNA and SV, methylation, expression, and mutational analysis was used to document the extent of chr1 molecular variants and their impact on pathway utilisation.

Results:

We identified two distinct groups of gain(1q): focal gains associated with limited gene expression changes and a neutral prognosis, and whole-arm gains, which associate with substantial gene expression changes, complex genetics and an adverse prognosis. CRISPR identified a number of dependencies on chr1 but only limited variants associated with acquired CNAs. We identified seven regions of deletion, nine of gain, three of chromothripsis (CT) and two of templated-insertion (TI), which contain a number of potential drivers. An additional mechanism involving hypomethylation of genes at 1q may contribute to the aberrant gene expression of a number of genes. Expression changes associated with whole-arm gains were substantial and gene set enrichment analysis identified metabolic processes, apoptotic resistance, signaling via the MAPK pathway, and upregulation of transcription factors as being key drivers of the adverse prognosis associated with these variants.

Conclusions:

Multiple layers of genetic complexity impact the phenotype associated with CNAs on chr1 to generate its associated clinical phenotype. Whole-arm gains of 1q are the critically important prognostic group that deregulate multiple pathways, which may offer therapeutic vulnerabilities.

Introduction

Copy number abnormalities (CNAs) of chromosome 1 (chr1) are seen in 40% of newly diagnosed myeloma (NDMM) with gain of the long arm [gain(1q)] being seen in 30% and loss of the short arm [del(1p)] seen in 15% of cases. The prognostic importance of gain(1q) has led to it being extensively studied and incorporated into prognostic systems (1–4). Amplification (>3 copies) is observed in 10% of cases and has a greater degree of specificity for poor outcome than simple gain(1q) (3,4). However, gain(1q) is not evenly distributed through the molecular subgroups with an association with t(4;14), t(14;16), and del(13q), as well as an inverse relationship with t(11;14) (3). Consistent with its impact on prognosis, fewer chr1 variants are seen in smoldering MM compared to NDMM (5,6) and 13–19% of patients acquire gain(1q) at relapse (7–9). Despite its widespread use as a prognostic variable there is ongoing debate about its value, likely due to differences in the nature and penetrance of deregulated drivers. Del(1p) has been less extensively investigated but specific loss of gene function has been associated with outcome (2,10,11); however, this variable has only been incorporated into a limited number of prognostic systems (11,12). The frequency and prognostic impact of chr1 variants in MM suggest that they could provide multiple targets for precision medicine if its molecular pathogenesis was clearly understood.

Deregulation of gene expression is critical to the prognostic impact of chr1 molecular variants with up to 30% of the genes making up the GEP70 prognostic score (13) mapping to chr1, Supplemental Table 1. The large size and large number of genes have complicated previous efforts to identify the key drivers on chr1. Potential drivers at 1q21 and 1q32, identified previously, include CKS1B (14), PDZK1 (15), BCL9 (16), ANP32E (1), ILF2 (17), ADAR (18), MDM4, and MCL1 (19) alongside genes with spiked expression such as ZBTB40, LRRIQ3, SLAMF1, ATF6, PBX1, TNFSF4, RABGAP1L, RASAL2, RALGPS2, RGS16, ASPM, PTPRC, IRF6, CENPF, DUSP10, and DISC1 (20). Candidate driver variants on the short arm include CDKN2C/FAF1 at 1p32 (1,21), TENT5C at 1p12, and RPL5/EVI5 at 1p22.1 (22).

Identifying novel drivers on chr1 can be addressed by relating recurrent molecular events to expression changes at these sites. We use a large dataset of NDMM with combined expression and long-insert whole genome sequencing (WGS) to map (23) recurrent SVs and CNAs on chr1. Building on this information, we integrate our findings with information on chromatin compartments, topologically associated domains (TADs), methylation, mutational, and expression status to compile a compendium of recurrent changes on chr1 to determine key drivers and pathways associated with this aggressive clinical variant.

Methods

Long insert WGS and RNA sequencing data

Patients

WGS data from 752 patients with NDMM from the CoMMpass (NCT01454297; IA13) were analyzed after CoMMpass Study Network Institutional Review Board Informed Consent (Copernicus IRB # QUI1–11-217). Samples were taken after informed consent, processed as previously described (24). A subset of 643 samples had RNA sequencing (RNA-seq) data available (25). All research was performed in accordance to the Declaration of Helsinki.

RNA-seq

RNA-seq data were aligned to GRCh38 using STAR (v2.5.1b) (26) and quality controlled using QoRTS (v1.2.42) (27) with alignment and quantification of gene read count with Salmon (v0.7.2) (28). Normalization of counts used DESeq2 (v1.14.1 RRID:SCR_000154) (29) with differential expression plotted using EnhancedVolcano (30).

WGS and mapping CNAs and SVs

Long-insert low coverage WGS was carried out as previously described (23). GISTIC2.0 defined copy number peaks (23) and hotspots (https://github.com/evenrus/hotspots/tree/hotornot-mm), were used to define recurrent regions. CNAs were filtered to include patients with at least one segment of copy number gain (copy number >2) or copy number deletion (copy number <2), and recurrent gained/deleted regions were identified using BEDtools intersect (v2.30.0 RRID:SCR_006646) (31). For co-occurrence, a Phi correlation coefficient was calculated in a pairwise fashion for all gained/deleted regions. Analysis of per patient occurrence/co-occurrence was performed using R (v4.1.0) using Ringo (32) for binary heatmaps with plotBM, sjstats for phi coefficient correlation with phicoef, and pairwise correlation heatmap with corrplot.

The PCF algorithm was used to define regions of chr1 enriched with breakpoints associated with copy number oscillations related to SVs such as TI and chromoplexy. Manual inspection of calls was used for chromothripsis (CT) (22) to effectively remove sequencing artifacts.

Methylation, Hi-C chromatin confirmation, and super-enhancer analysis

Methylation array data (Gene Expression Omnibus accession number GSE21304, RRID:SCR_005012) (33) was obtained from peripheral blood derived B cells (n=6), plasma cells from non-malignant donors (PC, n=3), patient samples of monoclonal gammopathy of undetermined significance (MGUS) (n=4), NDMM (n=161), plasma cell leukemia (PCL) (n=7), and 9 different human myeloma cell lines (HMCLs) (RPMI-8226, U266, H929, KMS11, KMS12, KMS26, LP1, MM1r, and MM1s). Methylation analysis was performed with RnBeads R package (RRID:SCR_010958) (34), probes annotated against GRCh38 using GENCODE v36 (35).

Hi-C data was obtained for normal B-cell states including peripheral blood derived naïve B-cells (NBC), memory B-cells (MBC), germinal-center B-cells (GCBC), and tonsillar plasmacells (PC) from EGAD00001006485dataset (36). Additional data was included for three HMCLs: U266, RMPI8226, and KMS11 (GSE87585 and EGAD00001003597). All Hi-C data was processed using hic-bench (v0.1) (37). Paired-end reads were mapped to GRCh38 using BWA (v0.7.17, RRID:SCR_010910) and low mapping quality (MAPQ < 20) were removed. Hi-C contact matrices were generated with Juicer (v1.5). Active and inactive chromatin compartments were assigned and scored with HOMER (v4.10) at 100kb resolution. The Hi-C contact matrix was then normalized using the ICE method and, subsequently, TADs were called using TopDom (38). Structural events in cell lines were called using HiNT (39).

Super-enhancer elements were characterized using H3K27ac signal peaks across 10 NDMM patients and U266, RPMI8226, and KMS11 cell lines as previously described (40). To summarize, the H3K27ac peaks were evaluated using ROSE2 to determine which regions could be considered super-enhancers. The resulting super-enhancer elements were then ranked within each sample based on increasing H3K27ac signal.

Multiomic maps and transcriptional compartments

Multiomic maps were generated using pyGenomeTracks (v3.6) (41), with tracks including: genomic coordinates of GRCh38, cytobands, Hi-C contact maps with TAD domains overlaid, A/B compartments and scores, peak signal for H3K27ac, H3K4me1, and H3K27me3 of PCs, U266, and KMS11 from BLUEPRINT (42), and ENCODE4 (43) ChIP-seq datasets, and gene bodies were annotated with GENCODE (v34) (44). Transcription data for PCs, U266, RPMI8226, and KMS11 were aligned to GRCh38 using STAR (RRID:SCR_004463) and normalized read coverage was quantified for visualization.

Data Availability

No original data was generated. All datasets are publicly available and assertion number provided in the methods section.

Additional CRISPR dependency analysis methods may be found, Supplemental Methods.

Results

Mutational events in chr1 genes

Recurrently mutated driver genes on chr1 were studied in 1273 NDMM exomes (45). A total of 1708 genes were identified on chr1 with 1478 having non-synonymous mutations more than once and 95 mutated in more than 1% of cases, Supplemental Figure 1. The mutated genes were correlated with the nine copy number (CN) clusters typical of MM (45). Using 63 known MM driver genes, 8% (5/63) mapped to 1p and none mapped to 1q (46). NRAS at 1p13.2 was the most frequently mutated (17%) and negatively correlated with CN cluster 3 that encompasses del(1p), gain(1q), del(13), del(14), and del(16q) (46). Inactivating mutations of TENT5C at 1p12 were seen in 12% of cases (47) and associated with CN cluster 1 (hyperdiploidy and gain(1q)) (46). Mutations in ARID1A at 1p36.11, CDKN2C at 1p32.3, and FUBP1 at 1p31.1 were each seen in 1% of cases. We identified bi-allelic inactivation of tumor suppressor genes located in regions of deletion such as. FUBP1 (50% of mutated cases), TENT5C (26%), and CDKN2C (62%) consistent with them being key tumor suppressor genes. Other frequently mutated genes such as TTN and OBSCN are classically considered passenger mutations given their size. Finally, the role of other genes such as RYR2 is unclear as it has previously been associated with immune regulation in non-small cell lung cancer (48).

Fine mapping of CNAs and SVs along Chr1

Using the WGS data, we mapped chr1 specific SVs, and CNAs (23). This analysis identified we 307 gains, 232 deletions, 54 chromothripsis, 100 template-insertion, and 18 chromoplexy events involving chromosome 1 that clustered into nine regions of gain, seven of deletion, three of chromothripsis (CT) and two of templated-insertion (TI), Figure 1. Regions of gain were predominantly located on the long arm and deletions on the short arm. Subsequently we mapped chromatin compartment status, super-enhancer locations, Hi-C, and activated and repressed chromatin marks (H3K27ac, H3K4me1, and H3K27me3) (49) to these regions.

Figure 1: Ideogram of chromosome 1.

The figure highlights regions of copy number gain, deletion, templated-insertion and chromothripsis. Broad chromosomal bands are shown for the long and short arms. Chromosomal sites are given in millions of base pairs. Recurrent sites of gain (G), deletion (D), template-insertion (TI) and chromothripsis (CT) are marked, chromosome sites are given and the potential oncogenes at that locus are given (Created with BioRender.com).

Regions of gain.

The GISTIC2.0 defined boundaries of gain, ranged in length from 51kb-3.3Mb. Several potential drivers including MCL1 (G3), SLAMF7 (G5), POU2F1 (G6), BTG2 (G8) and NLRP3 (G9) lay within regions of gain (16). Of note, some previously identified drivers lay outside these identified regions: BCL9 (2Mb downstream of G2), ANP32E (306kb upstream of G3) (1), CKS1B and ADAR1 (1.2Mb and 1.5Mb upstream of G4, respectively) (14,18), ATF6 (941kb downstream of G5), and PBX1 (2.3Mb upstream of G6) (20). These data suggest that although important, these regions do not drive the entire spectrum of genes upregulated as a consequence of gain(1q), Figure 2A, Supplemental Table 2A. The regions of gain tend to fall within active chromatin regions with G3, G4, G5, G6, and G8 lying exclusively within A-compartments (100%) across B-cell differentiation states and in U266, RPMI8226, and KMS11. The G1 region also followed this pattern except for KMS11 (A-compartment representing 78% only). The G2 region showed variability between the normal and malignant plasma cells, with moderate switching to B-compartments of 24%, 21%, and 37% in U266, RPMI8226, and KMS11, respectively, Supplemental Figure 2. The compartment profile of the G7 region is interesting as each sample comprised a considerable B-compartment percentage (range 16–82%) but across all samples ABL2 was found to lie within an A-compartment and was consistently expressed suggesting it may be a candidate driver in this region, Figure 2B–C, Supplemental Figure 3. The G8 region overlapped with a templated insertion region, Figure 2D. The G9 region showed the greatest variability across all samples and was shifted more towards inactive chromatin (range 45–97%).

Figure 2: Landscape of chromatin state across 1q.

A. Copy number changes and compartment scores across chr(1q). Regions of gain and templated-insertion occur in predominantly active compartments and overlap with known driver genes. B.,C. The compartment profile of the G7 region highlighting that ABL2 is contained in an active compartment and significantly expressed in HMCL in comparison to normal PC (C). D. The TI2 region overlaps with a number of known oncogenes and multiple super-enhancers. NBC=Naïve B-cell, MBC=Memory B-cell, GCBC=Germinal Center B-cell, PC=plasma cell, CNV=copy number variation, Mb=Megabase. G=gain, CT=Chromothripsis, TI=Templated-Insertion.

Regions of deletion.

We identified seven regions of deletion ranging in size from 367kb-10.7Mb, Figure 3A. D6 at 1p22.1 (74%) encompassed RPL5, EVI5, GFI1 and MTF2. Other regions and the downregulated genes contained within them are shown, Figure 3A, Supplemental Table 2B. The D1 region had identical A-compartment percentage (56%) across all samples and was associated with the loss of a sub-telomeric region that included TNFRSF4, a candidate driver of adverse prognosis, which resides within the A-compartment across samples. The D2 region was consistently in an A-compartment across all samples (range 96–100%) with the exception of KMS11 (47%), where this switch to a B-compartment occurred in a gene-sparse region. The D3 region had a similar active chromatin profile (range 44–77%) across all samples except in RPMI8226. This region encompassed the tumor suppressor genes CDKN2C, which was in an A-compartment in all samples and FAF1 that is also in an A-compartment in GCB-cells, PC, and KMS11 but in a partial A-compartment in NBC, MBC, and U266. The D4 and D5 regions lay completely (100%) and near completely (>87%) in inactive compartments, respectively. Interestingly, the small A-compartments in D5 seen in U266 and RPMI8226 correspond to the tumor suppressor FUBP1. The D6 region was seen in an A-compartment across the samples specifically at the RPL5 and MTF2 loci, while EVI5 and GFI1 were contained within a region that showed switching from B-A across the B-cell differentiation stages, Supplemental Figure 4. The D7 region was centered on TENT5C and its associated super-enhancer which lies within an A-compartment across all samples except for KMS11, Figure 3B, Supplemental Figure 5. These observations are consistent with a critical role for the A-compartment in association with structural events as drivers as well as an important role for regions with strong enhancers that also involve structural rearrangements to multiple regions of the genome where genes at receptor loci may be potentially overexpressed.

Figure 3: Landscape of chromatin state across 1p.

A. Copy number changes and compartment scores across chr(1p). Regions of deletion and chromothripsis occur in variable compartments and overlap with known tumour suppressor genes. B. The compartment profile of the CT3, TI1, and D7, highlighting that all but one cell line and B-cells place the gene TENT5C and its associated super-enhancer in an A-compartment. NBC=Naïve B-cell, MBC=Memory B-cell, GCBC=Germinal Center B-cell, PC=plasma cell, CNV=Copy number variation, Mb=Megabase. G=gain, CT=Chromothripsis, TI=Templated-Insertion.

Recurrent SVs.

We identified two regions of templated-insertion, TI1 at 1p13.1–1p11.2, covered a 5.0Mb pericentromeric region and overlapped D7, and was involved with recurrent chromosomal translocations (5,50). The TENT5C super-enhancer was frequently disrupted by structural rearrangement with variable proportions of the super-enhancer being rearranged to the site of potential drivers at varying receptor sites. The relevance of inactivation of TENT5C was suggested by its pattern of mutation but gene upregulation at receptor sites as a result of rearrangement of its super-enhancer likely deregulates genes at its receptor site and may explain the frequency of events at this locus as a result of their active selection. The TI1 region lay predominantly in B-compartment (range 63–73%) with U266 and RPMI8226 having slightly more active compartment (24% and 47% B-compartment, respectively), Supplemental Figure 6. Furthermore, analysis of H3K27ac signal (40) identified a super-enhancer associated with NOTCH2 that was found in U266, RPMI8226, KMS11 and 2/10 NDMM patients.

A second region, TI2 at 1q32.1–1p32.2 covered 6.4Mb, encompassed the oncogenes BTG2, FMOD, UBE2T, ELF3, MDM4, ELK4, overlapped the G8 region, and was associated with a super-enhancer linked to KDM5B (51), Figure 2D. Pathologically, this region is in an almost entirely active compartment (range 91–100%) and contains a high concentration of per patient super-enhancers (mean 8.8, range 6–11) and all patients had at least 1 super-enhancer linked to BTG2 and MDM4, consistent with their expression being deregulated as a consequence. These data suggest that regions of TI may be driven by regions being actively transcribed and containing super-enhancers.

We identified three regions of recurrent CT, all occurring on the short arm, that were also associated with regions of deletion. The CT1 region at 1p36.11–1p34.3 overlapped with the D2 region and comprised of greater than 74% A-compartment (range 74–100%); CT2 at 1p33–1p32.3 overlapped the D3, the site of the FAF1 and CDKN2C, and comprised both inactive (range 35–59%) and active chromatin (range 41–65%) and, CT3 at 1p13.3–1p12, overlapped with the D7 region and TI1, had greater variability with 86% A-compartment in U266 while only 1% in KMS11, Figure 3B, Supplemental Figure 7, Supplemental Table 3. This could be explained using Hi-C data where we were able to identify a translocation in KMS11 (breakpoint in 98.4–98.6MB window and involving chrX) that appears to involve the rest of the 1p pericentromeric region. No such translocation was seen in U266 (data not shown).

Fusion genes.

As previously described, we identified four cases with a translocation of the Ig gene locus to 1q21.2 (OTUD7B, SF3B4, MTMRII) or 1q41 (HHIPL2). Recurrent NTRK1 (1q23.1) fusions were identified in 0.5% of samples (52).

The relationship between regions of loss and gain

For the nine regions of gain, 40% (302/752) of patients had gained at least one region. Gains along the 1q arm in the G2 to G9 regions occurred at a comparable frequency (range 79–87%). The G1 region, located at the sub-telomeric border of 1p, was seen in only in 6% (18/302) of patients, Supplemental Figure 8A–B. There were 219 patients with whole-arm gain of 1q (i.e., gains of all region from G2 to G9, which also includes patients with gains G1 to G9) and 83 with focal gains defined as any cases with less than gain of G2-G9 (the largest groups involved gain of seven regions n=22, two regions n=14, and one region n=32). These results are consistent with the observation that there are two broad prognostic groups of gain(1q); the majority with whole-arm gains (73%, gain of all regions from G2 to G9), and the remainder with focal gains, Figure 4A.

Figure 4: Pattern of 1q changes and impact on outcome.

A. By analysing the structure of chromosome 1q changes we identified two broad groups. The first (green) was characterised by a gain of 8–9 regions that we defined as whole-arm gain and the second was associated with focal or smaller gains (pink). B.,C. The association of whole-arm and focal gains on progression free and overall survival

Patients in the group with whole-arm gains had a worse outcome, PFS (HR 1.3, 1.02–1.6, p=0.04) and OS (HR 1.6, 1.15–2.2, p=0.004) than patients with no gain(1q); focal gains did not appear to confer a significant adverse prognosis, Figure 4B–C. Whole-arm gains were associated with a complex genomic background including with t(4;14) (corr=0.10, BF=1.06), loss of acrocentric chromosomes (del(13q):RB1 corr=0.23, BF=118623 and del(14q):ABCD4 corr=0.12, BF=2.9) and other deletions (del(4p) corr=0.11, BF=1.89 and del(16p):CYLD: corr=0.12, BF=2.3) and negatively correlated with trisomies and t(11;14) (corr=0.16, BF=184); whereas focal gains strongly associated with TI (corr=0.34, BF=6e14), MYC translocation (corr=0.24, BF=3034538), and hyperdiploidy (corr=0.13, BF=4.9), Supplemental Spreadsheet 1–3. We show that the common associations with CKS1B amplification are mostly found in the whole-arm gains group (83%), and there is a trend suggesting a worse prognosis than those with focal gains and amplification, Supplemental Figure 9. Interestingly patients with whole-arm gains had significantly shorter telomeres and a higher mean age, Supplemental Figure 10. Patients with critically short tumor telomeres were also over-represented in this group (corr=0.11, BF=1.2).

For the seven regions of deletion, a total of 232 (31%) patients had at least one deleted region with 14% (33/232) involving all 7 (D1-D7), 9% (22/232) D7 only, and 9% (21/232) D3-D7, Supplemental Figure 8C–D. A total of 61% (141/232) had between 1–3 regions with loss of D5, D6, and D7 being more common (range 60–74%) than D1-D4 (range 19–48%). Each region of deletion strongly correlated with deletion of others (range corr=0.32–0.86, BF: 8.7×10¹²-2.24×10⁹²). Some of these regions co-occurred, specifically whole-arm gains of 1q correlated with deletion of D3 and D4 (corr=0.13, BF=3.2 and corr=0.14 BF=150, respectively). Other strong correlations included D1 and CREBBP mutation (corr=0.14, BF=3.97), D5 and D6 with PTPN11 mutations (corr=0.15, BF=28 and corr=0.14, BF=19, respectively), and D6 with MYC inversion (corr=0.14, BF=8.3), Supplemental Spreadsheet 1–3. In terms of outcome, del(1p) was not associated with an adverse outcome in this dataset and we were unable to confirm that patients with biallelic 1p32 losses did worse as there were none in this dataset (53).

Transcriptional rewiring and gene expression changes

In order to define the relationship between chr1 structural rearrangements with their expression patterns we performed differential gene expression analysis. We used copy number gain of CKS1B at 1q12, a standard clinically useful marker of gain(1q), to identify 1q changes in CoMMpass patients with associated RNA-seq. This analysis identified 173 patients with a gain(1q) and 376 without and identified 2687 differentially expressed genes (p-adj<0.05) of which 608 were located on chr1, Supplemental Figure 11–Table 4, Spreadsheet 4. The 608 gene set contained the majority of the previous suggested drivers associated with 1q, including 124 transcription factors. NES at 1q23.1 was the most upregulated gene, based on fold change (54,55). The oncogenes S100A4 and NTRK1 were significantly upregulated (log2(FC)>1 and q<0.05). Pathway analysis of the 608 differentially expressed genes on chr1 identified four potential e-pathways using g:Profiler including positive regulation of metabolic process (GO:0009893, q=8.4e-3) and biosynthetic processes (GO:0009058, q=9.7e-7). Similarly, using the total of 2687 genes we identified as deregulated over 100 upregulated pathways including the regulation of cellular processes (GO:0050794, q=5.5e-19) l and metabolic processes (GO:0031323, q=6.6e-9), Supplemental Figure 12A.

To examine the expression of genes on 1q in more detail and to remove the expression changes related to the molecular subgroups with which 1q CNAs associate, we performed the same analysis within molecular subgroups. For t(4;14), 37 of 74 patients had a gain(1q), and we identified 94 significantly deregulated genes (p-adj<0.05) with 87 located on 1q, Supplemental Spreadsheet 5. A pathway analysis identified an enrichment for several pathways pertaining to the mitochondrion (GO:0005739 , q=9.3e-4; GO:0005742, q=2.6e-2; GO:0098798, q=6.1e-3). For the t(11;14) subgroup, 19 of 112 patients had a gain(1q), we identified 149 significantly deregulated genes (p-adj<0.05) with 109 located on 1q, Supplemental Spreadsheet 6. In a pathway analysis, we identified pathways pertaining to the protein metabolism (GO:0044249, GO:0006406). Among the genes that were not on 1q, PAPD4 and CARD18 were overexpressed on chromosome 5 and 11, respectively. We identified a limited set of genes (n=47) that overlapped between the t(4;14) and t(11;14) groups (p-adj<0.05), including two transcription factors ZBTB7B and CREB3L4. Many of these genes were associated with metabolism and mitochondria stability, proliferation and protein homeostasis consistent with gain(1q) in these subgroups having a more aggressive disease phenotype ), Supplemental Figure 12B–C.

To address differences in expression between whole-arm or focal gains, we compared patients with no evidence of gain(1q) (n=297) with patients with focal gains (n=56); this analysis identified 20 genes of which the majority (n=12) were located on chr1, Supplemental Spreadsheet 7. We then compared patients with no evidence of gain(1q) (n=297) with patients with whole-arm gains; the number of deregulated genes on chr1 was higher (n=585). Notably, there were also significantly more deregulated genes overall (n=2409) suggesting that generalized gene deregulation is part of a more complex genome wide mechanism of gene deregulation in the group with whole-arm gains, Supplemental Spreadsheet 8. Interestingly of the total set (n=2409), 101 were transcription factors and involved upregulation of BACH2, NR5A1, MYBL1, GLMP, and E2F2 (the last two of which are on chr1) and downregulation of PAX5, SMAD1, TBX21, SIX4, EGR3, and GF1 (the last on chr1), with most of these being involved in histone H3K9me methylation, intracellular signaling, and regulation of transcription. We directly compared patients with focal gains to patients with whole-arm gains and identified 6 genes associated with focal gains and 129 associated with whole-arm gains of which 119 genes were on chr1. When comparing normal and focal gains, we note that most upregulated genes in focal gains were associated with protein handling (TSACC, RABIF, TOMM40L1 DAP3, UBQLN4, SLC25A44, TMEM183, ZC3HIIA). Overall, these data show that the extent of gene deregulation is associated with the underlying genetic events and that whole-arm gains deregulate both more and a different set of genes in comparison to focal gains.

To identify genes altered in expression on 1p, we interrogated a set of patients defined by either del(1p12):TENT5C or del(1p22): RPL5, Supplemental Table 5A–B. We used this set to identify changes in expression with the recurrent deleted regions. The analysis identified RPL5 as the most downregulated gene across both patient sets (FC=−0.76, q=3.84×10⁻³ and FC=−1, q=7.55×10-¹⁷,respectively) (22,56).

Transcriptional dependencies on chr1 genes

The functional genome wide essentiality screen (DepMap) took account of gene dependency levels as well as accounting for copy number specific effects (CERES dependencies) (56) on 19 MM cell lines and 15 random other cell lines . A total of 1701 genes were analyzed. Cell lines did not cluster according to their copy number profile but tended to clutter according to disease type. Common dependencies genes with gene dependencies< −1 clustered together and comprised proteasome subunit coding genes (PSMA5, PSMB2, PSMB4), three regulators of ubiquitin-protein transferase activity (RPL5, RPL11, CDC20), splicing factors (SF3B4, SF3A3, SFPQ, RNPC3, SRNPE, PRPF38A, PRPF38B) and DTL. Pathway analysis revealed an enrichment for pathways essential to the processes of normal cell survival including protein metabolism, cell cycle and cell division, Supplemental Figure 13, Spreadsheet 10.

The second clade could be further subdivided into two clades (n2.1=125 and n2.2=1594). The second of these two clades could further be divided into two clades (n2.2.1=105 and n2.2.2=1489). Clades 2.1 and 2.2.1 included genes with significant dependencies in most (n2.1) and some (n2.2.1) cell lines but not the others. Genes in these clusters include suspected drivers such as ADAR and MCL1. These genes were also enriched for key cellular processes, Spreadsheet 10. The tumor suppressor genes commonly found in deleted regions on 1p such as CDKN2C, or FAF1 were commonly found in clade with no dependency.

We grouped the 19 myeloma cell lines according to their CNAs nine had gain(1q) and six had del(1p); all that had del(1p) also had gain(1q). Differential dependency analysis comparing gains vs normal, identified 14 genes that met criteria. Nine were located on 1q (ZNF281, ZBTB37, PIGN, CD1D, HAX1, UCHL5, POU2F1, LAMTOR2. CCDC190), Supplemental Figure 14. When comparing loss vs normal, 32 genes met the criteria, 13 of which localized to 1p. Three of these genes were associated with increased dependency (ID3, GLMN, TRNP1) constituting potential targets, and 8 genes associated with reduced dependency (NRAS, NFIA. PDIK1L, RPL11, ZFP69B. THAP3, AMY2A, NFYC) suggesting they would not constitute effective therapeutic targets for patients with del(1p). Of note, the dependency on ALG14 and RPL5 that are already close to −1, were significantly lower in del(1p) cell lines, Supplemental Figure 15. None had a delta greater than 0.4, suggesting the overall dependencies may not just be the resultant of the CNA on chromosome 1 but the resultant of many other factors such as cytogenetic background or associated structural event.

Interestingly, PSMD4 encodes the protein subunit that caps the proteasome and is easily targetable, thus has been explored as a therapeutic target in MM. Recently is has been shown as a promising target to overcome proteasome resistance occurring as a consequence if gain(1q) (57). Also, MDM4 amplification could lead to DNA instability via p53 inhibition which opens the possibility of using MDM4 inhibition specifically in cases where this gene is amplified (58).

Methylation based gene deregulation

We used methylation array data derived from 161 patient and normal PC controls to identify the role played by methylation in deregulation of genes on chr1. To accomplish this goal we used PCA analysis of methylation features along chr1 (probes=2498) using malignant B and plasma cells reference patterns. Tumor samples were clustered by translocation subgroup and disease aggression, Figure 5A–B (33).

Figure 5: The 1q methylation landscape of MM samples.

A. Tumor samples clustered by disease aggressiveness. B. By translocation subgroup. C. By differentially overexpressed genes in gain(1q) where a total of 62 genes were both differentially hypomethylated and overexpressed. D. Heatmap representation of the beta value for all 383 differentially methylated probes across chromosome 1 between MM and normal PC, the probes are presented based on their genomic coordinates. The additional column to the right displays the difference in the group mean beta value, where positive values indicate a change towards hypomethylation and negative values indicate a change towards hypermethylation in MM. Of note, is a distinct ribbon of hypomethylated probes in the MM group near the middle of the heatmap corresponding to a region of 150MB-162MB (1q21.2–1q23.3).

A total 383 probes on chr1 were significantly differentially methylated in MM, compared to normal plasma cells, of which 378 were hypomethylated (genes=360), Supplemental Figure 16 and Spreadsheet 14. The methylation pattern seen is marked by hypomethylation in a near homogenous profile across chr1, which is distinct from that seen normal plasma cells. In particular we draw attention to a band of extreme hypomethylation at 1q21.3–1q23.1 which potentially contributes to the over-expression of genes within the region. Interestingly, of the GEP70 genes and previously implicated drivers on chr1, only EVI5, SLAMF7, and PDZK1 were hypomethylated in MM, Supplemental Figure 17. When restricted to gain(1q), a total of 62 differentially overexpressed genes were both differentially hypomethylated and overexpressed, Figure 5C, Supplemental Spreadsheet 12. Cross referencing to COSMIC (59) identified 2 of these to be oncogenes, DDR2 and NTRK1. Using the same 383 differentially methylated probes, we generated a further dendrogram sorted by chr1 genomic location and identified a consistent region of demethylation at 1q21.3–1q23.1, Figure 5D. This region had a high concentration of the 62 hypomethylated overexpressed genes associated with gain(1q). Most of these probes did not fall within the boundaries of the recurrently gained regions and, therefore, provide an alternate mechanism for gene overexpression, Supplemental Figure 18.

Discussion

We provide an in-depth analysis of the molecular events associated with acquired genetic variation on chr1. These data provide important insights into the potential therapeutic vulnerabilities associated with whole-arm copy number gain of 1q. This information provides important new insights into the biology of the clonal cells carrying gain(1q), which come to dominate the bone marrow based on their capacity to metabolically adapt to their environment. Further, as this biology provides therapeutic vulnerabilities, optimizing biomarker detection of chr1 variants based on the greater resolution of the data provided by this study could enhance their use in precision medicine strategies.

On the long arm we resolve CNAs into two broad groups characterized by their structure into either a “focal” or “whole-arm” group which moves us away from a prognostic system based on a simple dichotomy of chr1 CNAs into gain or no gain groups. This increased understanding moves us towards a prognostic approach based on an understanding of the molecular mechanisms underlying the copy number changes present. Importantly, distinct patterns of gene expression characterize these groups, and are likely important in driving the biological behavior of each subtype opening the way for more tailored therapeutic strategies.

We show that it is gain of the entire chr1 long arm that is associated with adverse outcome making it important to distinguish cases with this abnormality from the group characterized by focal copy number gains, which is prognostically neutral. Mechanistically, whole-arm gains are mediated by “jumping translocations” that are the result of centromeric fragility associated with de-condensation of centromeric heterochromatin that leads to persistence of somatic pairing and formation of multibranched chromosome arms and whole-arm duplication. We show that whole arm gains constitute 60% of gain(1q) events, associate with an adverse prognosis, occur on complex genomic backgrounds, and are over-represented in high-risk patient subgroups. The biology of the whole-arm gains group is associated with the overexpression of many drivers on 1q rather than any single gene and in addition many genes are overexpressed at other loci that potentially contribute to the biology of these cases. This observation suggests that there is generalized deregulation occurring as a consequence of whole-arm gains of 1q, the nature of which is currently uncertain. The gene set enrichment analysis (GSEA) analysis, however, shows that whole-arm gains of 1q focus on the deregulation of a limited number of specific gene pathways particularly those involving metabolism and deregulation of cell growth and maintenance. In addition, we identified upregulation of a set of transcription factors that include E2F2, GLMP, and GFI1 on chr1 that are associated with deregulation of methylation pathways and regulation of transcription pathways, Figure 6.

Figure 6: Diagrammatic representation of study findings.

There are two types of 1q gain. The first are focal, usually small in size and associated with the deregulation of the protein handling. The second are large, encompass all 8 regions of gain and are associated with deregulation of pathways involved in metabolism, proliferation, and transcription regulation (Created with BioRender.com)

Focal gains of 1q, by contrast, are seen in 11% of NDMM cases and have a neutral impact on prognosis. Data presented here defines several distinct regions of gain on 1q providing greater resolution than has been possible previously. Focal events arise via TI on a background that is frequently hyperdiploid and are presumably selected because of their biological impact. The focal gene deregulation induced by these events, while pathologically important is not associated with an adverse clinical outcome. Interestingly, GSEA analysis of this group of patients points to a focus on the deregulation of genes involved in protein homeostasis that may be targeted by proteosome inhibition, a widely used treatment in MM.

The region of copy number gain at 1q21.1–1q25.2, occurs in a transcriptionally dense region but has not been characterized previously to a depth which can fully resolve the key drivers. Here, we identify at least 6 distinct foci of gain providing significantly improved resolution allowing us to map them to key chromatin features. The regions G3, G4, G5, G6, and G8 occur in the transcriptionally active A-compartments containing multiple differentially overexpressed genes, including SLAMF1, SLAMF7, MCL1, BGLAP, and BTG2 all of which could drive cell proliferation, migration, invasion, and apoptosis. The recurrent upregulation of these genes implies that targeting them is likely to be particularly active in this group of cases.

We show that not all of the candidate genes located at 1q21.1–1q25.2 are impacted by GISTIC2.0 peaks consistent with there being additional mechanisms that are important in deregulating driver genes. We note that significant global hypomethylation of 1q occurs during disease progression (33,60,61). Genes impacted by this include members of the S100 calcium binding protein family, which have recently been associated with poor outcome in MM and they may contribute to aggressive behavior through cross-talk with monocytes in the microenvironment (62–64). Furthermore, we identify a set of hypomethylated and overexpressed oncogenes within the 1q21 region with DDR2 being particularly important as we have shown it is mutated in 0.22% of cases (3) and has been associated with resistance to IMiD treatment and could be targeted by tyrosine kinase inhibitors (65).

Loss of copy number on the short arm of chr1 provides a counterpoint to gains on the long arm and is broadly associated with closed chromatin and the loss of tumor suppressor gene function. In addition to copy number loss, mutational inactivation of gene function leading to bi-allelic loss of gene function is common as part of these events. Oncogene upregulation can also occur via chromosomal rearrangement of super-enhancers to receptor sites as seen on at TENT5C locus. Loss of function of this gene, a non-canonical cytoplasmic poly(A)-polymerase that inhibits proliferation, is the key target at this locus (66). We propose that it is the impact of these multiple mechanisms, involving both down and upregulation of gene function, deliver the selective advantage associated with this variant. While we show that 1p events associate with a significant impact on prognosis, the key prognostic event remains uncertain as there is significant overlap between the individual regions of copy number loss.

Additional gene loci where multiple mechanisms of gene inactivation are noted suggesting they are potential drivers include the sites of FUB1, CDKN2C and NTRK1. The mutational profile at these genes is inactivating, supporting a tumor suppressor role for them. An important alternate mechanism is highlighted by NTRK1 (52), which is differentially hypomethylated, overexpressed and is associated with the formation of fusion genes in 0.5% of NDMM (67). A novel region on 1q which has these features is G8 where there is a concurrence of a region of gain and TI, and, analogous to D7, it is the site of multiple super-enhancers linked to known oncogenes BTG2 and MDM4.

By providing a comprehensive analysis of genetic and epigenetic changes present on chr1 we identify convergent mechanism leading the deregulation of potentially actionable drivers, Supplemental Table 6. We dichotomize chr1 gains into two groups, focal, with no adverse outcome and whole-arm, that encompasses most of the amp(1q) associated with an adverse outcome.

Translational statement.

Fine mapping of recurrent copy number abnormalities and structural variants identified seven regions of deletion, nine of gain, three of chromothripsis, and two of template-insertion, which contain a number of potential drivers. We resolve chromosome 1 copy number abnormalities, seen in up to 40% of newly diagnosed multiple myeloma, into two major subtypes: either focal-changes associated with a neutral outcome or whole-arm gains associated with a complex genetic background and adverse outcome. The major impacts of whole-arm gains are deregulated transcription, apoptotic resistance, upregulated metabolism and increased signaling via the MAPK pathway opening opportunities for targeted therapy of this aggressive disease subset.