Whole-genome optical mapping of bone-marrow myeloma cells reveals association of extramedullary multiple myeloma with chromosome 1 abnormalities

Extramedullary disease (EMM) represents a rare, aggressive and mostly resistant phenotype of multiple myeloma (MM). EMM is frequently associated with high-risk cytogenetics, but their complex genomic architecture is largely unexplored. We used whole-genome optical mapping (Saphyr, Bionano Genomics) to analyse the genomic architecture of CD138+ cells isolated from bone-marrow aspirates from an unselected cohort of newly diagnosed patients with EMM (n = 4) and intramedullary MM (n = 7). Large intrachromosomal rearrangements (> 5 Mbp) within chromosome 1 were detected in all EMM samples. These rearrangements, predominantly deletions with/without inversions, encompassed hundreds of genes and led to changes in the gene copy number on large regions of chromosome 1. Compared with intramedullary MM, EMM was characterised by more deletions (size range of 500 bp–50 kbp) and fewer interchromosomal translocations, and two EMM samples had copy number loss in the 17p13 region. Widespread genomic heterogeneity and novel aberrations in the high-risk IGH/IGK/IGL, 8q24 and 13q14 regions were detected in individual patients but were not specific to EMM/MM. Our pilot study revealed an association of chromosome 1 abnormalities in bone marrow myeloma cells with extramedullary progression. Optical mapping showed the potential for refining the complex genomic architecture in MM and its phenotypes.


Materials and methods
Subject enrolment. BM aspirates were obtained from an unselected cohort of 11 newly diagnosed MM patients with EMM presentation (n = 4; median age: 77 years, min-max: 51-79; M/F: 3/1) and without EMM (MM, n = 7; 75 years, 62-82; 5/2). Patients were diagnosed according to the International Myeloma Working Group criteria 15 . The only criteria for patient enrolment were sampling at diagnosis and a sufficient number of sorted cells to perform all genetic analyses (≥ 2 million myeloma cells). In our patients, all EMM sites were bone related, with two in the thoracic spine and two in the pelvis (one in the iliac bone and one in the acetabulum). Patient's clinical and demographic data are summarised in Table 1 and Table S1. For all patients, karyotype, FISH (fluorescence in situ hybridization, Table S2), arrayCGH (Table S3) and next-generation sequencing (NGS) for mutations in the TP53, KRAS, NRAS and BRAF genes (Table S2) were available.
All patients provided written informed consent about the usage of BM for this study, which was performed in accordance with the Helsinki Declaration and approved by the ethics committee of the University Hospital and Palacký University Olomouc.
Isolation of high molecular weight DNA, labelling and analysis. Frozen myeloma cell pellets were processed following the Bionano Prep SP Frozen Cell Pellet DNA Isolation Protocol 16 . High molecular weight (HMW) genomic DNA was isolated using the SP Blood and Cell Culture DNA Isolation Kit (Bionano Genomics, CA, USA, #80030), according to the manufacturer's recommendations. DNA quantification was performed using the Qubit dsDNA BR assay kit (Thermo Fisher Scientific) with a Qubit 2.0 Fluorometer (Thermo Fisher Scientific).
A total of 750-1000 ng of HMW DNA was then labelled using the Bionano Prep Direct Label and Stain DLS DNA Kit (Bionano Genomics, #80005), according to the manufacturer's protocol 17  Data assembly, structural variant calling and the identification of breakpoint regions. All data were analysed using Bionano Access software (v1.5) containing the Bionano Solve tool (v3.5) and featuring both de novo and rare variant bioinformatics pipelines (Fig. 1), according to the manufacturer's recommendations  .
Only DNA molecules with a minimum length of 150 kbp were used for bioinformatics analysis along with a minimum of nine labels per molecule. Briefly, the de novo pipeline's first assembly of all single molecules was based on the distinct distribution of sequence labels by pairwise alignment. The aligned molecules created consensus maps (contigs) in de novo genome maps, which were compared with the in silico DLE1 labelled human hg38 reference map. This pipeline revealed structural variants (SVs) from 500 bp to tens of Mbp long. In the rare-variant pipeline, all single molecules were pairwise aligned against the hg38 reference assembly; molecules with SVs were clustered, and the obtained maps were locally aligned to the hg38 reference sequence. This pipeline was sensitive enough to detect SVs from 5 kbp to tens of Mbp long at a variant allele frequency (VAF) as low as 5%. SVs were considered subclonal (i.e. low-allele frequency) when VAF was ≤ 25% and clonal (i.e. high-allele frequency) when VAF was > 25%, based on a cut-off value for neutral evolution in MM 11 . Additionally, both pipelines included copy number variation (CNV) analysis to detect the fractional copy number changes and chromosomal aneuploidy events. Specific hg38 masks concealing common structural variation in a human genome, N-base reference gaps and problematic sub-centromeric and sub-telomeric regions were used in both pipelines. To annotate the SV calls  www.nature.com/scientificreports/ that were likely somatic variants, a variant annotation pipeline was applied to filter SVs out of the database of ethnically diverse, mapped control human genomes with no reported disease phenotypes.
In the next step, annotated SVs and CNVs from both pipelines were merged ( Fig. 1), including aberrations sized 500 bp-5 Mbp (deletions, insertions, duplications and inversions) as well as inter-and intrachromosomal aberrations larger than 5 Mbp. The intrachromosomal rearrangements with breakpoints at least 5 Mbp apart, e.g. large deletions (supported by copy number loss), insertions (copy number gains) or inversions (no change in CNVs) were called intrachromosomal translocations by the Bionano software (Fig. S1). Only SVs with VAF > 5% and a minimum of ten self-molecules were further analysed in this study. Identified candidate SVs were confirmed by arrayCGH, FISH, breakpoint-specific PCR amplification and/or long-read whole-genome sequencing (TELL-Seq, Universal Sequencing Technology, CA, USA). For a comparison of optical mapping and long-read sequencing data, we developed our own tool, which is available at http:// olgen. cz/ en/ resou rces 22 .

Ethics declarations.
All patients provided written informed consent about the usage of bone marrow samples for this study, which was performed in accordance with the Helsinki Declaration and approved by the ethics committee of the University Hospital Olomouc and Palacký University Olomouc.

Results
Sample analysis by optical mapping. The infiltration of myeloma cells in BM aspirates based on immunophenotyping was highly variable in enrolled patients (3-36%); more than 10% infiltration of plasma cells was found in the BM smears of all enrolled patients. The inter-individual variability in the myeloma cell infiltration may be linked to patchy or site-varied myeloma cell distribution, haemodilution, aspirate pull order, the aggregation of myeloma cells in aspirated BM, myeloma cell immunophenotypes and time-dependent losses of surface markers 23 , as well as disease heterogeneity itself 27 (Table S5). After filtering only for likely somatic variants, the number of deletions per patient (41, 24-62) dominated over insertions (18,, inversions (3, 1-9) and duplications (3, 0-13) (Table S5, Fig. 2A), reaching high interindividual variability. All detected chromosome translocations and intrachromosomal rearrangements were identified as somatic-like in all samples.
The EMM genome contained more deletions than the MM (median number of 45 vs 34, P = 0.05), particularly small deletions of 500 bp-50 kbp (37 vs 24, P = 0.01) (Fig. 2). The number of inversions and duplications did not differ between EMM and MM (P > 0.05). The spectrum of SVs and affected genes and chromosomes displayed high inter-individual variability. In addition to the deletion of the CCSER1 gene on chromosome 4 found in ~ 45% of our patients, the SVs in two patients covered NKAIN2, and two others covered the EYS gene, both within a commonly affected region, 6q.  (37,, were common in all patients. Except for two MM patients, the majority of patients had a mean of five regions of CN > 3 (range 1-16 per patient) in their genomes. The distribution of CNVs across the genome was highly variable in enrolled EMM and MM patients.
Optical mapping confirmed 98% of SV and CNV changes detected by diagnostic cytogenetic and arrayCGH assessments (Tables S2, Tables S3) and revealed numerous novel rearrangements in all enrolled patients.
Additionally, numerous other translocations were detected across all MM patients, frequently affecting chromosomes 2, 3, 6 and 8 (Table S6). All MM patients carried at least two translocations, except for one MM patient with only t(4;14) (Table S6, Fig. 3). Complex chromosomal rearrangements involving three chromosomes were detected in four (57%) MM patients but not in any EMM patients (Table S6, Fig. 3). The translocations were present at clonal and subclonal levels (VAF 5-43%). The affected genes and putative fusion genes are shown in Table S6.
EMM genomes were associated with fewer translocations than MM; two EMM patients had no translocations, one EMM patient had one translocation and the only EMM patient that reached complete response after first-line therapy had four translocations. The translocations were present at clonal and subclonal levels (VAF 5-49%) (Fig. 3).

Intrachromosomal rearrangements in MM and EMM.
Large chromosomal rearrangements encompassing regions longer than 5 Mbp on chromosome 1 were detected in all EMM genomes but not in any MM genomes (Fig. 4, Table 2). The large rearrangements, together with the small SVs (predominantly deletions), affected various regions across chromosome 1, often involving deletions and inversions accompanying the CNV changes. EMM1 had one large intrachromosomal rearrangement of 14.5 Mbp, encompassing 230 genes in the 1p36 region, and five deletions; EMM2 had three large intrachromosomal rearrangements of 47.5 Mbp, 57.9 Mbp and 21.5 Mbp, encompassing 1093 genes in the 1p35-p31, 1p32-p12 and 1p22-p13 regions, and an additional six deletions and one insertion. EMM3 had four rearrangements on chromosome 1 of 7.6 Mbp, 7.5 Mbp, 12.6 Mbp and 12.8 Mbp, encompassing 794 genes in the 1p35-p34, 1p22-p21 and 1p21-p13 regions, and two deletions. EMM4 had two large rearrangements of 36.1 Mbp and 12.0 Mbp, encompassing 564 genes in the 1p34-p31 and 1p34-1q23 regions, three deletions and five insertions (Fig. S1). The majority of the affected genes by intrachromosomal rearrangement across chromosome 1 in EMM were associated with cancer (~ 35%), cell cycle (~ 30%) and inflammation (~ 10%); very few affected genes were associated with MM (~ 10%) (Table S7).
Additionally, intrachromosomal rearrangements were distributed across other chromosomes in both MM and EMM (Table S8, Fig. S2). The typical patterns of intrachromosomal translocation were large deletions with partial inversion, accompanied by copy number loss. Multiple rearrangements within the same chromosome   (Table S9). Also, somatic-like SVs within the IGK and IGL loci were detected: deletions in five patients (three EMM and two MM), insertions in two (two EMM) and duplication in one (MM) were identified (Table S9, Fig. S3).
Regarding TP53 disruption, diagnostic analysis by FISH and NGS did not detect any abnormalities in enrolled patients. Nevertheless, optical mapping revealed copy number loss (CN = 1) in the region overlapping the TP53 gene in two EMM patients (Fig. 5).
Optical mapping confirmed del(13q14) identified by FISH in one EMM and three MM patients. Additionally, optical mapping detected a 1.1 Mbp deletion affecting the RB1 gene, supported by copy number loss in the 13q14.2 region in one MM patient, which was not detected by FISH (Fig. 5).
Regarding the 8q24 locus, one EMM patient carried a deletion and one MM a duplication, detected by both mapping and cytogenetics. Optical mapping revealed additional changes within this locus associated with MYC gene amplification in three patients: one MM patient had a 0.6 Mbp insertion and three translocations, t(6;8), t(8;17) and t(8;22); one MM had an inversion; and one EMM patient had a novel 0.2 Mbp insertion (Table S10).
Regarding high-risk regions on chromosome 1 commonly affected in MM, we confirmed 1q21 gain in six patients (two EMM and four MM) and del(1p32) in one EMM patient (Table S11). On chromosome 1, 1.4 times more SVs within/outside the high-risk 1q21/1p32 regions were found in EMM than in MM. In EMM, deletions (50%) and intrachromosomal rearrangements (31%) were the most frequent, not duplications and translocations.

Discussion
This study characterised genomes of BM myeloma cells in newly diagnosed EMM and MM patients using nextgeneration optical mapping. When comparing the EMM and MM genomes, EMM was associated with large intrachromosomal rearrangements across chromosome 1, fewer interchromosomal translocations and more deletions across the entire genome compared with MM. For high-risk loci, optical mapping revealed copy number  To date, the complex genomic architecture in MM and EMM has been poorly characterised, probably due to its complexity, heterogeneity and multiple levels of somatic mosaicism 29,30 . Therefore, we analysed EMM and MM genomes using innovative optical mapping that can detect small SVs and CNVs as well as complex large genomic rearrangements or chained fusions 14,31-33 , which are not recognisable by NGS and/or cytogenetics. The utility of this approach has been recently shown in leukaemia samples, where optical mapping confirmed the results of whole-genome sequencing and/or cytogenetic analysis and additionally revealed a large number of SVs not previously recognisable in analysed samples 14,33 . In this study, we used optical mapping for the first time to study the genome architecture of isolated myeloma cells from BM from newly diagnosed EMM and MM patients.
In line with the high degree of somatic genomic mosaicism and multiple levels of genetic variation in MM 29,30 , long-fragment mapping revealed simple and complex genomic rearrangements and CNVs in all samples. More interchromosomal translocations were detected in MM patients than EMM. Except for one patient with a high-risk 14q32 translocation, a common primary event in MM 34 , all MM patients had at least two other translocations. These often involved chromosomes 2, 3, 6 and 8, and many of them led to gene disruptions or the creation of putative gene fusions with at least one partner associated with cancer. Moreover, interchromosomal translocations in MM were often accompanied by intrachromosomal rearrangements located in the same chromosomal loci. On the contrary, three EMM patients had one or zero translocations, and one EMM patient with good treatment response had three translocations; the translocations occurred in our patients at subclonal and clonal levels (5-49%). Although the impact of clonal status on the prognostic value of SVs is unclear for most cancers, recent NGS studies in MM have shown that the clonality status of mutations does not influence survival but does impact the disease phenotype 35 . Experimental evidence also suggests that MM progression, both spontaneous in asymptomatic stages and at relapse after treatment, is linked to its heterogeneous subclonal composition 36 ; thus, the direct measures of the clone size and its intrinsic biological features deserve future investigation. Optical mapping also revealed numerous complex translocations, involving three chromosomes in about half of the MM patients but not in the EMM patients. There are already reports about large chromosomal rearrangements, called chromothripsis, in MM 29,30 . Such complex structural changes, often accompanied by loss of heterozygosity 37 , are difficult to identify by other techniques and may escape attention. The presence of unusual rearrangements of numerous chromosomes in MM, but not EMM, deserves future investigation.
In addition to interchromosomal translocations, we detected numerous intrachromosomal rearrangements, which are rearrangements that involve loci located on the same chromosome. To date, few cancer types harbour both interchromosomal and intrachromosomal rearrangements; one of them is MM 38 . It has been suggested that the occurrence of intra-or interchromosomal recombinations depends on the spatial proximity between recombinogenic partners within the chromosome territories (CTs), a non-randomly formed, distinct space where each chromosome decondenses 39,40 . When loci are situated near the surface of their CTs, interchromosomal translocations occur, and when they are located deep in the CTs, intrachromosomal rearrangements occur 41 . There is already evidence that chromosomes involved in commonly occurring translocations -t(4;14), t(14;16) Importantly, we detected EMM-specific intrachromosomal rearrangements encompassing several Mbp-long regions within chromosome 1, commonly including combinations of deletions and inversions and affecting hundreds of genes. These rearrangements were located across the whole of chromosome 1 and led to changes in the copy number of genes on large regions of this chromosome. The intrachromosomal rearrangements on chromosome 1 have already been reported in progressive, multi-drug refractory EMM 10 and EMM with soft tissue involvement at the time of MM diagnosis 44 . Interestingly, 80% (8/10) of patients with soft tissue EMM had chromosome 1 abnormalities, and an association between chromosome 1 abnormalities and soft tissue EMM was suggested 44 . Furthermore, 1p deletion and/or 1q gain were associated with the extramedullary plasmablastic transformation of MM in both BM and matched extramedullary tissue 45 . Other studies reported an association of chromosome 1 abnormalities in MM with the relapsed disease 46 . The affected patients have an exceedingly poor prognosis, short progression-free survival and overall survival, even in the era of novel therapies 44,47,48 . A recent study showed that the adverse impact of chromosome 1 abnormalities on survival is of similar magnitude to other high-risk chromosomal abnormalities 47 . The crucial role of chromosome 1 in MM pathogenesis is also www.nature.com/scientificreports/ supported by the significant overrepresentation of genes derived from chromosome 1 in the high-risk signature in MM 48 . The occurrence of fewer interchromosomal translocations and more intrachromosomal rearrangements in EMM, particularly on chromosome 1, suggests that recombinations within loci deep in CTs may play a crucial role in MM pathogenesis, particularly influencing the phenotype of the disease. Furthermore, the observed chromosome 1 abnormalities may play a role on the required events that allow the dissemination of myeloma cells from BM to blood and distant tissues; this also deserves future investigation. In addition to translocations and intrachromosomal rearrangements, we also detected tens of SVs in every EMM and MM genome. The most common were deletions distributed across all chromosomes. In particular, deletions ranging in size from 500 bp to 50 kbp occurred more frequently in EMM than in MM. An increased number of deletions in MM has already been associated with MM progression, as shown by comparing MM genomes at diagnosis and relapse 49 . Future studies should investigate the relationship of a higher deletion load in EMM compared with MM as well as prognosis. We also detected numerous novel SVs and CNVs within highrisk loci associated with MM not previously detected by sequencing and cytogenetics.
The most critical genetic factors that portend a poor prognosis for MM are translocations within the IGH/ IGK/IGL loci 50 . Our study confirmed the common translocations t(4;14) and t (11;14) in three MM patients and revealed additional interchromosomal translocations involving chromosomes 2, 14 and 22, where immunoglobulin genes are located, in a majority (6/7) of MM patients. The functional consequences of the translocations outside the IGH/IGK/IGL loci needs to be clarified, as they may influence antibody expression and function and the mediation of disease phenotypes. Interestingly, we did not detect any translocation on the previously mentioned chromosomes in EMM, where deletions and inversions were predominantly found. Differences between MM and EMM in genetic rearrangements on immunoglobulin-associated chromosomes should be further investigated.
Next, we were interested in the disruption of the 17p13 locus overlapping the TP53 gene, a driver aberration associated with EMM 51,52 , poor prognosis and low treatment response rates in MM patients 53 . The loss of TP53 and other genetic aberrations may additionally occur in the extramedullary mass due to regional clonal evolution, as shown by comparing extramedullary tumours with their BM myeloma cells 7,13 . Although no TP53 disruption was detected in myeloma cells from BM aspirates of enrolled patients by diagnostic cytogenetic and mutational analyses, optical mapping revealed copy number loss in the 17p13 region in two EMM patients. Our data further support the key role of TP53 in EMM and emphasise the need to routinely incorporate SVs and CNVs, the major forms of genetic alterations in cancer, at many length scales to understand the MM genome more comprehensively.
Optical mapping also confirmed rearrangements at the MYC/8q24 locus, a late tumour progression event associated with an increased expression of MYC and poor prognosis 54 , in about a third of patients. One MM patient had three translocations within this region, and two others had SVs within the MYC/8q24 locus. Whether the changes at the MYC/8q24 locus were EMM specific, as reported by others 55 , needs further investigation using larger cohorts.
This study has several limitations. First, we did not investigate extramedullary tumour mass because invasive biopsy was not feasible in enrolled patients. Second, due to the moderate number of patients included in this exploratory study, a sub-analysis based on clinical and laboratory parameters was not performed. Third, the proportion of the IgA subtype was higher in the MM cohort. However, there is growing evidence that adverse prognosis in patients with IgA MM versus non-IgA MM subtypes is more likely to be caused by the misclassification of disease response or the delayed detection of disease due to an underestimation of tumour burden 56 than changes in expression profile or cytogenetics [57][58][59] . Future studies on larger patient cohorts enabling a subanalysis of patients with particular clinical characteristics and stages of disease and the investigation of extramedullary tissue sites are warranted.
There is a growing body of evidence on the utility of optical mapping for comprehensive SV detection in haematology and solid tumours 14,[31][32][33] . The aberrations detected by mapping have been confirmed by cytogenetics 14 or NGS 33,60 , particularly by long-read sequencing 61,62 , as also shown in our study. Optical mapping thus provides an ideal complement to sequencing for resolving complex genomic architecture in cancers 31 .

Conclusion
Our pilot study using next-generation optical mapping revealed that in addition to known high-risk cytogenetic factors, chromosome 1 abnormalities in BM myeloma cells are associated with extramedullary progression. The detection of numerous novel, distinct genetic aberrations associated with EMM and MM shows the potential of optical mapping for the refinement of complex genomic architecture in MM and its phenotypes. The methodology and results described here represent a significant advance that may accelerate the introduction of genomics at long-length scales into clinical decisions for MM.

Data availability
The data of this study are available from the corresponding author on reasonable request.