Mate pair sequencing outperforms fluorescence in situ hybridization in the genomic characterization of multiple myeloma

Fluorescence in situ hybridization (FISH) is currently the gold-standard assay to detect recurrent genomic abnormalities of prognostic significance in multiple myeloma (MM). Since most translocations in MM involve a position effect with heterogeneous breakpoints, we hypothesize that FISH has the potential to miss translocations involving these regions. We evaluated 70 bone marrow samples from patients with plasma cell dyscrasia by FISH and whole-genome mate-pair sequencing (MPseq). Thirty cases (42.9%) displayed at least one instance of discordance between FISH and MPseq for each primary and secondary abnormality evaluated. Nine cases had abnormalities detected by FISH that went undetected by MPseq including 6 tetraploid clones and three cases with missed copy number abnormalities. In contrast, 19 cases had abnormalities detected by MPseq that went undetected by FISH. Seventeen were MYC rearrangements and two were 17p deletions. MPseq identified 36 MYC abnormalities and 17 (50.0% of MYC abnormal group with FISH results) displayed a false negative FISH result. MPseq identified 10 cases (14.3%) with IgL rearrangements, a recent marker of poor outcome, and 10% with abnormalities in genes associated with lenalidomide response or resistance. In summary, MPseq was superior in the characterization of rearrangement complexity and identification of secondary abnormalities demonstrating increased clinical value compared to FISH.


Introduction
Multiple myeloma (MM) is a plasma cell neoplasm (PCN) representing the second most common hematopoietic malignancy and accounts for~20% of all hematologic cancer related deaths in the United States 1 . During the last decade there have been remarkable improvements in the treatment of patients with MM that have resulted in increased survival, including immunomodulatory compounds, proteasome inhibitors, and immunotherapeutic approaches such as monoclonal antibodies 2 . Paralleling the advances in novel therapeutic strategies, characterization of the genomic complexities of MM have significantly improved with the implementation of nextgeneration sequencing (NGS), thus enabling the iden tification of novel single nucleotide variants (SNV), structural rearrangements and copy number abnormalities (CNA) [3][4][5][6][7][8][9][10][11] . Comprehensive genomic characterization studies such as the Multiple Myeloma Research Foundation (MMRF) CoMMpass Trial and other research studies are necessary for the discovery of novel variants of clinical significance that may lead to improved treatment approaches and prognostication strategies 12,13 .
In contrast to the use of genome-wide NGS strategies employed in the research/investigational trial setting, most clinical genomics laboratories rely upon traditional cytogenetic methodologies such as conventional chromosome studies and fluorescence in situ hybridization (FISH) to characterize recurrent cytogenetic abnormalities of prognostic significance. High-risk cytogenetic abnormalities as defined by the Mayo Clinic mSMART 3.0 algorithm 14 include t(4;14), t(14;16), t (14;20) translocations, 17p deletions and 1q gains, while standard-risk cytogenetic abnormalities include hyperdiploidy (gains of odd-numbered chromosomes), t (11;14) and t (6;14) translocations 15,16 . A limited number of laboratories evaluate for MYC and t(6;14) rearrangements, and detection of IGK and IgL rearrangements is not routinely performed in the clinical setting 15 . Although FISH assays have high sensitivity, are relatively inexpensive compared to NGS techniques and provide input for risk stratification 17 , several limitations exist. They allow for the interrogation of only the regions for which FISH probes are available and multiple FISH probes are needed in order to be comprehensive, with each probe requiring a resourceconsuming validation. More importantly, FISH has the potential to miss cryptic abnormalities, including rearrangements that result in a position effect due to juxtaposition of enhancers near oncogenes [18][19][20][21][22] . Since many translocations identified in MM involve a position effect (i.e., IGH and MYC) with heterogeneous breakpoints 10,23,24 and some CNAs may be cryptic, we hypothesize that some clinical FISH probes used in the characterization of PCNs have a high rate of false negative FISH results.
To test this hypothesis, we evaluated the performance of a genome wide mate-pair sequencing (MPseq) assay in comparison to FISH panel testing for MM. Since MPseq utilizes long input DNA (2)(3)(4)(5) followed by circularization and fragmentation to the size of paired-end fragments (200-500 bp) that are sequenced at reduced depth, this assay is designed to detect structural rearrangements and CNAs throughout the genome resulting in a costeffective strategy amenable to a clinical genomics laboratory. Furthermore, as MPseq has higher resolution than FISH and is not limited to specific genomic footprints for interrogation, this assay could provide an alternative technique to comprehensively detect structural rearrangements and CNAs in a single assay. Herein we describe the performance, along with the added clinical utility, of MPseq in 70 samples previously characterized by FISH to detect chromosome rearrangements and CNAs in patients with a PCN.

Patient samples
All samples were referred to the Mayo Clinic Genomics Laboratory as part of routine clinical testing and further evaluated by MPseq as part of a Mayo Clinic Institutional Review Board approved study. There were multiple sources of samples obtained either from fresh or frozen whole bone marrow (BM), or from fixed cell pellets (FCP) from an abnormal BM chromosome study. Some specimens had undergone plasma cell enrichment from fresh whole BM that was either flow sorted or subjected to CD138+ magnetic-enrichment from patients that had an abnormal plasma cell FISH result. Additional methodology, including conventional chromosome analysis and flow cytometry are included in supplemental data.

Fluorescence in situ hybridization
Plasma cell proliferative disorder FISH (PCPDF) of immunoglobulin (cIg)-stained positive PCs studies were performed as previously described 25 Table 1 with footprints  and probe source shown in supplemental Table 2.

DNA extraction and library preparation
DNA extraction and mate pair library preparation methods have been previously described 18,27,28 . Briefly, DNA was isolated using either the Qiagen Puregene extraction kit (for samples < 2 mL), Autopure LS Automated high quality DNA extraction (for samples > 2 mL) or the QIAmp Tissue kit for fixed cell pellet samples. DNA was processed using the Illumina Nextera Mate Pair library preparation kit and sequenced on the Illumina HiSeq 2500 in rapid run mode as described in Aypar, et al. 18 . Pooled libraries were hybridized onto a flow cell (2 samples per lane) and sequenced using 101-basepair reads and paired end sequencing.

Structural variant bioinformatics pipeline and visualization
The sequencing data was analyzed for the detection of structural variants (SVs), which are large genomic changes (>30Kb) that involve breakpoint junctions and/or CNAs. The sequencing data was mapped to the reference genome (GRCh38) using BIMA 29 and the output was analyzed using SVAtools. This set of algorithms can detect and report the breakpoint locations of both junctions and CNAs at high resolution and accuracy (Schematic in supplemental Fig. 1) 18,27,28 . Junctions and CNVs were graphically illustrated using genome, junction and region plots as previously described 18,27,30 .

Patient characteristics
A total of 70 cases referred to the Mayo Clinic Genomics Laboratory for routine clinical PCN FISH testing were selected for further evaluation by MPseq (Tables 1, 2  and supplemental Table 3). Criteria for inclusion included the type of primary cytogenetic abnormality to ensure representation of each recurring rearrangement and sample source to evaluate various methods of sample attainment, including PC enrichment (Tables 1, 2 had undefined primary abnormalities including one case of tetrapoloidy with a relative 1q gain, one case with monosomies 13 and 14, two cases with monosomy 15 by FISH and a single case with normal FISH results in a patient with a diagnosis of amyloidosis (Tables 1, 2). Conventional chromosome studies were performed on 42 (60.0%) cases and an abnormal PC clone was identified in 33.3% of the cases with chromosome studies performed (Supplemental Table 3).
We have previously determined tumor content requirements for MPseq requiring 10% tumor for the detection of structural rearrangements and 25% tumor for the detection of CNAs 18 . Since variable and sometimes low clonal PC percentages can be identified in the BM aspirates of patients with NDMM 31 , we performed two enrichment strategies for samples with low PCs including magnetic enrichment or flow sorting. For some samples, no enrichment was performed. Thirty-nine (55.7%) samples with a median 23.0% PCs were subjected to either flow sorting (N = 24) or CD138 + magnetic bead for PC enrichment (N = 15). For the remaining 31 samples (44.3%) with a median 58% PC, no PC enrichment was performed.

Identification of recurrent, primary cytogenetic abnormalities using MPseq
To determine the accuracy of MPseq in comparison to our PCN FISH panel (Supplemental Table 1) in the detection of recurrent, primary MM abnormalities (IGH rearrangement and/or hyperdiploidy), we analyzed DNA extracted from either a fixed cell pellet (FCP) from a chromosome study (n = 8), from fresh (n = 18) or frozen (n = 5) BM aspirates or from fresh BM specimens that had been flow sorted (n = 24) or subjected to CD138 + magnetic enrichment (n = 15) (Supplemental Fig. 1, Tables 1, 2). For samples 1-65, MPseq confirmed the primary abnormality identified by FISH in each case demonstrating 100% concordance between both assays for the classification of recurrent, primary cytogenetic abnormalities (Fig. 1). For those cases without evidence of a recurrent, primary abnormality (samples 66-69), MPseq did not identify tetraploidy in case 66 and monosomy 15 in cases 68-69, but identified monosomies 13 and 14 in case 67 and confirmed no recurrent abnormality in case 70 with normal FISH results. As a negative control, no recurrent primary MM abnormalities (MM specific IGH rearrangements and/or hyperdiploidy with gains of odd numbered chromosomes) were identified by MPseq in a previously described cohort of 88 patients with a reason for referral of acute myeloid leukemia (data not shown) 18 .

Comparison of MPseq to FISH for detection of recurrent, secondary abnormalities
For each primary and secondary abnormality that was identified by either MPseq or FISH, 40 cases (57.1%)

Increased detection rate of MYC rearrangements by MPseq
From 70 total cases, we identified 36 cases (51.4%) that displayed a MYC rearrangement by MPseq (Fig. 1). Of these 36 cases, 34 had FISH data evaluating the MYC locus. Seventeen cases (50.0% of MYC abnormal group with FISH results) displayed a false negative MYC FISH result where a MYC rearrangement was identified by MPseq, but was negative by FISH (Fig. 1). The most common partner gene/ enhancer segment identified were IGH (n = 7), FAM46C (n = 5), IGK (n = 4), NSMCE2 (n = 4), TXNDC5 (n = 4) and IGL (n = 4) ( Table 3). Of the 36 MYC rearrangements, multiple mechanisms resulting in positioning of MYC near enhancer sequences including small insertions, inversions, simple, balanced or complex translocations were identified ( Table 3). The most common method of rearrangement identified in 15 cases included a small insertion of enhancer sequences near the MYC gene or, alternately, the insertion of MYC near enhancer sequences. These insertions typically involve the duplication of genetic material of similar size at both the source location and the insertion location, whereby the source DNA is inserted between flanking duplications at the insertion location 32 (Fig. 2b). Thirteen of these 15 insertion cases co-occurred with hyperdiploidy (hyperdiploidy only or hyperdiploidy with IGH separation) and two of these cases were identified by FISH studies. Of the 17 MYC cases that were missed by FISH, 11 represented these small insertions ( Table 3).

Discussion
Most clinical laboratories employ FISH analysis of CD138 enriched plasma cells as the preferred methodology in order to identify recurrent primary and secondary genomic abnormalities of prognostic and therapeutic significance in patients with PCNs 15 . The majority of these laboratories utilize only a limited FISH panel with many focusing on high risk abnormalities defined by the revised International Staging System (R-ISS) including 1q gain, t(4;14), t(14;16) or 17p deletion 15 . Some laboratories have incorporated the use of chromosomal microarray analysis in the detection of CNAs such as hyperdiploidy, 17p deletions and 1q gains, however microarray studies are unable to identify balanced structural rearrangements necessitating the use of other methodologies in the detection of IGH rearrangements 15 . It has also become increasingly apparent that some FISH probes, such as those targeting MYC rearrangements, display evidence of false negative results [18][19][20][21][22] . In addition, FISH panels for PCNs are variable between individual laboratories, provide a limited view of the whole genome and may not always reflect genomic complexity. Given that multiple research studies and investigational trials have used NGS based techniques to identify CNAs, SNVs along with structural rearrangements, we sought to explore the feasibility of employing an NGS technique in the detection of CNAs and structural rearrangements as a FISH replacement assay within a clinical genomics laboratory.
We describe the performance and added utility of a whole genome NGS based strategy, MPseq, in comparison to the current gold standard FISH approach in the evaluation of patients with PCNs. While MPseq and FISH displayed equal performance in the ability to classify the presence or absence of a recurrent, primary cytogenetic subtype (i.e. hyperdiploidy or specific IGH   For each case where a secondary alteration involving MYC was found, the relevant genomic information is provided for the junction(s). The case column is the case number. The FISH column indicates whether or not the MYC FISH test detected the secondary alteration (dark gray-detected by both FISH, light gray-detected by MPseq but not tested by FISH and red-detected by MPseq only). The type column is the type of alteration involved with MYC classified as either a balanced event, a tandem duplication, a translocation, an inversion, part of an amplification, part of a small insertion motif, a complex event, or ND where it was not possible to definitively classify the alteration. The Junction column is the number of junctions involved directly in the alteration, either 1 or 2. The Chr Partner and Pos Partner columns are the chromosome and position location (GRCh38) of the partner breakpoints that are part of alteration. The Pos MYC and MYC Loc columns give the position of the breakpoint in the MYC locus and whether the alteration is to the left, right, or encompassing (L, R, or E) the MYC gene, respectively. The Gene Pair column is the gene that is found at or near the partner breakpoint location. The Primary column is the primary alteration for the case rearrangement), MPseq was superior compared to FISH in the characterization of rearrangement complexity, identification of secondary abnormalities, resolution of atypical FISH results and identification of novel abnormalities of prognostic significance not targeted by traditional FISH panels. Many samples chosen for this study had a high plasma cell burden (median 36% PCs) and~33% of cases were obtained from fresh or frozen samples that did not require enrichment. An advantage to using a whole genome NGS technique like MPseq is the ability to identify rearrangements using an unbiased approach. Other laboratories have developed and validated NGS methodologies utilizing targetenrichment approaches for PCNs allowing a custom target pull down of limited genomic regions 13,[35][36][37] . While these targeted approaches have reduced cost and simplified analysis workflows, a genome wide approach utilizing long-insert whole genome sequencing employed by the MMRF CoMMpass Study in their Seq-FISH analysis has demonstrated improved sensitivity with similar specificity in relation to clinical FISH testing 38 . Although MPseq is similar to Seq-FISH with regard to a whole genome sequencing approach, a significant limitation to the current MPseq strategy is the inability to identify SNVs. This limitation can be resolved with deeper and faster sequencing, coupled with reduced sequencing costs. An integrated genomic analysis incorporating structural variation, CNAs, and SNVs together may lead to enhanced prognostication 13 . Of practical consideration is the~twofold increased cost and "turn-around-time" of reporting of clinical grade testing for MPseq compared to a comprehensive FISH panel; although we anticipate over time the cost and time of reporting for NGS approaches will continue to be reduced.
Another limitation to the use of MPseq is the inability to identify rearrangements in highly repetitive regions of the genome containing constitutive heterochromatin such as those involving telomeres, centromeres, and in regions near the centromeres of chromosomes 1, 9, and 16 and in the Y chromosome 27 . This limitation may be reflected by the inability of MPseq to identify apparent trisomies in 2 cases (cases 5 and 42) with evidence of hyperdiploidy. Case 5 displayed a gain of a structurally abnormal chromosome 3 by conventional chromosome studies. Since the centromere regions that are targeted by the FISH probes are not covered by MPseq, it is unclear whether a small gain or presence of a polymorphism of these regions are present without evidence of a bona fide trisomy or whether the trisomy was present at a subclonal level below the limit of detection by MPseq (<25% for CNAs) 18 . Polymorphisms of the acrocentric chromosome 15 have also been reported 39   MPseq. Since MPseq does not rely on detection of only the centromere region like FISH, analysis of copy number changes throughout the whole chromosome can be useful to interpret the presence or absence of a trisomy. On the other hand, any missed hyperdiploid cases may be of less relevance as hyperdiploidy can be detected by flow cytometric methods 40 . MPseq is not currently used in the detection of copyneutral LOH or ploidy. As a result, MPseq did not identify any cases as tetraploid, an abnormality found in approximately 6% of patients with MM that has been associated with high-risk genomic abnormalities and with poor prognosis in NDMM 41 . Six patients in our cohort had evidence of tetraploidy and 5 of these cases had tetraploidy in combination with high risk cytogenetics (Fig.  1), which were identified by MPseq. For case 39, it is unclear why MPseq failed to identify a 1q gain. No evidence of a duplication involving chromosome 1q (chr. 1:155122503-155571708) was identified by MPseq and based on FISH data, the 1q gain did not appear to be subclonal, however this sample was extracted from unsorted bone marrow with 20% clonal plasma cells which may have contributed to this missed abnormality. Subclonal CNAs and cases with low tumor load have a risk of being missed by MPseq, a risk that also exists when performing FISH.
Of the 30 discordant cases, 19 cases had abnormalities identified by MPseq that were missed by FISH, the majority involving insertional events near the MYC gene region. In total, MYC rearrangements were found in 51.4% of the cases in our cohort which includes NDMM and RRMM. This is consistent with previous reports identifying MYC rearrangements in 35% of NDMM with increased frequency in RRMM 10,42 . MYC rearrangements have been reported as subclonal events and are associated with disease progression 3,11,43 . Approximately 66% of MYC rearrangements have been found in association with non-Ig partners resulting in juxtaposition to enhancer Total abnormalities 1  2  3  4  5  6  7  8  9  10  11  12  13  14  15  16  17  18  19  20  21  22  23  24  25  26  27  28  29  30  31  32  33  34  35  36  37  38  39  40  41  42  43  44  45  46  47  48  49  50  51  52  53  54  55  56  57  58  59  60  61  62  63  64  65  66  67  68  69   ND or RR  sequences promoting aberrant MYC gene expression, which may be targeted by BRD4 inhibitors in MM 10 . Identification of MYC rearrangements using a break-apart probe strategy resulted in a 50.0% false negative rate in our patient cohort. Whether these false negative insertion cases have the same prognostic implication as other MYC rearrangements remains unknown.

Risk stratification MPseq S H H H H H H H S H S S S S H H H H H H H H H H H H H H H H H H H H S H S H S H H S S H H H H H H S H H S H H S H H H S S S S S S H H S S
Two cases had deletions of the TP53 gene region that were not identified by FISH. For cases 4 and 21, MPseq identified a deletion of TP53 (5.6 Mb in case 4, 2.7 Mb in case 21). Interestingly, for case 21, MPseq also identified a translocation involving TP53 (to 4q32.1). Both cases were scored as having two copies of TP53 by FISH and represent false negative results due the location of the deletion in relation to the FISH footprint in case 4 and the TP53 translocation in combination with the deletion in case 21. Although these cases had missed high risk abnormalities, the mSMART risk did not change since those cases also had additional high risk abnormalities [1q gain for cases 4 and t(4;14) for case 21]. For case 4, a separate NGS assay analyzing SNVs identified a pathogenic TP53 mutation [Chr17(GRCh37):g.7577111 G > T; NM_001126113.2 (TP53):c.827 C > A; p.Ala276Asp] located in the DNAbinding domain and in vitro functional data predicts that this variant results in non-functional p53 protein 44 . TP53 has been found to be mutated in 3-16% of NDMM 6,45-47 with a higher frequency in RRMM [48][49][50] . TP53 mutations in combination with 17p deletions are associated with double hit MM with reduced overall, progression-free and relapse-free survival 51 . Therefore, this missed TP53 deletion fails to identify the presence of a likely double hit MM in patient 4. The combination of a rearrangement and deletion also likely represents a double hit MM abnormality in case 21.
The CCND1/IGH dual color, dual fusion probe set is used to identify CCND1/IGH rearrangements. However, three copies of CCND1 in the absence of IGH fusion can indicate trisomy 11 or non-IGH CCND1 rearrangements. MPseq identified three CCND1 rearrangements including case 4 (IGL/CCND1), case 43 (IGK/CCND1) described more fully in Peterson, et al. 52 and case 57 (BRINP3/ CCND1). FISH also identified amplification of CCND1 in case 4, three copies of CCND1 in case 43 and a normal signal pattern for CCND1 in case 57. The CCND1 rearrangement identified in case 57 was a complex translocation between 1q31.1 and 11q13.3 consisting of four junctions and deletions of~100 kb at both ends. Through this complex event, CCND1 is brought into close proximity to the 3′ end of BRINP3, while the balancing set of junctions brings the 5′ end of BRINP3 near 11q24.3. Additionally, the derivative chromosome containing CCND1 has been copied. Overall this would result in three copies of CCND1, two of which have been translocated near the 3′ end of BRINP3. This case demonstrates how MPseq is able to determine complex rearrangements involving important genes without prior knowledge of the junction partner location.
Although immunoglobulin lambda rearrangements have been recently reported in association with poor prognosis 3 , light chain rearrangements are typically not evaluated in the diagnostic work up of MM in most clinical genomics laboratories. Using MPseq data, we identify 10 cases (14.3% of entire cohort) with IGL rearrangements with five of these cases with standard risk cytogenetic results. IGL rearrangements and other focal deletions of clinical significance are typically not evaluated by FISH. Given the high rate of false-negative MYC rearrangements and inability to appreciate all abnormalities of clinical significance, we demonstrate that MPseq has increased clinical value compared to FISH in characterizing genomic abnormalities in PCNs.