Copy number evolution and its relationship with patient outcome—an analysis of 178 matched presentation-relapse tumor pairs from the Myeloma XI trial

Structural chromosomal changes including copy number aberrations (CNAs) are a major feature of multiple myeloma (MM), however their evolution in context of modern biological therapy is not well characterized. To investigate acquisition of CNAs and their prognostic relevance in context of first-line therapy, we profiled tumor diagnosis–relapse pairs from 178 NCRI Myeloma XI (ISRCTN49407852) trial patients using digital multiplex ligation-dependent probe amplification. CNA profiles acquired at relapse differed substantially between MM subtypes: hyperdiploid (HRD) tumors evolved predominantly in branching pattern vs. linear pattern in t(4;14) vs. stable pattern in t(11;14). CNA acquisition also differed between subtypes based on CCND expression, with a marked enrichment of acquired del(17p) in CCND2 over CCND1 tumors. Acquired CNAs were not influenced by high-dose melphalan or lenalidomide maintenance randomization. A branching evolution pattern was significantly associated with inferior overall survival (OS; hazard ratio (HR) 2.61, P = 0.0048). As an individual lesion, acquisition of gain(1q) at relapse was associated with shorter OS, independent of other risk markers or time of relapse (HR = 2.00; P = 0.021). There is an increasing need for rational therapy sequencing in MM. Our data supports the value of repeat molecular profiling to characterize disease evolution and inform management of MM relapse.


Introduction
Multiple myeloma (MM) is caused by the malignant clonal expansion of plasma cells in the bone marrow [1].
Recent next-generation sequencing projects have provided insight into the sub-clonal complexity of tumor progression in MM [11][12][13][14]. However, few recurrent single nucleotide changes have been reported to be associated with disease relapse, notably sub-clonal mutations of cereblon in immunomodulatory drug (IMiD) treated patients. Knowledge about their prognostic impact remains limited [15]. In contrast, several large-scale chromosomal abnormalities, including gain(1q), del(1p), and del(17p) have been shown to be acquired during tumor progression, but their association with tumor subtype, treatment and outcome is not well understood [11,16]. Recent studies of the relationship between copy number aberrations (CNAs) and progression of MM have largely been based on shortread sequencing or FISH. Importantly, few studies have been performed on patients from a randomized clinical trial and the impact of therapy on acquisition of CNAs remains unclear [15,16]. To gain insight into the relationship between CNAs and progression of MM we performed virtual karyotyping of all chromosomes and targeted hotspot profiling using digital multiplexed ligation-dependent probe amplification (digitalMLPA) in matched diagnosis-relapse tumor pairs in 178 UK NCRI Myeloma XI trial patients [17]. We show that acquisition of CNAs is strongly correlated with MM subgroups. Additionally, a number of acquired changes have independent predictive value for defining patient outcome.

Patients
We studied 178 newly diagnosed MM patients enrolled in UK NCRI Myeloma XI (ISRCTN49407852) for which high quality bone marrow tumor material at both presentation and relapse was available. First outcomes of the trial have been published recently [17]. Briefly, patients were initially randomized to triplet induction with thalidomide (CTD), lenalidomide (CRD), or carfilzomib and lenalidomide (KCRD; transplant-eligible patients only) in combination with cyclophosphamide and dexamethasone. Insufficient responders (partial or minimal response) were randomized to cyclophosphamide, bortezomib and dexamethasone (CVD) vs. no intensification and nonresponders (stable or progressive disease) received CVD. Younger, fitter patients received high-dose melphalan (HDMEL) and autologous stem-cell transplantation, patients were randomized to receive lenalidomide, lenalidomide plus vorinostat, or observation. Maintenance treatment continued until progressive disease in the absence of toxicity ( Supplementary  Fig. 1).
Median time to progression was 20.7 months (range 3.7-71.9 months) and median follow-up 47.0 months. Baseline characteristics are summarized in Supplementary  Table 1. The frequency of chromosomal aberrations detectable in the 178 patients at presentation were representative of the overall trial cohort (Supplementary Table 2) [4].
For all patients CD138-positive tumor cells were immunomagnetically selected at presentation and relapse and quality controlled for tumor cell purity (>95%), DNA and RNA were extracted using QIAGEN (Hilden, Germany) Allprep kits. Additional molecular quality control including longitudinal consistency for IgD deletion status was performed for all cases. Matched identity of presentation-relapse pairs was confirmed using single nucleotide and/or insertion/deletion polymorphisms assayed by digitalMLPA D006-X2 for all samples.
All patients provided written informed consent. The study was approved by the UK National Research Ethics Service, research ethics committees at participating centers and the UK Medicines and Healthcare Products Regulatory Agency.
Targeted genome wide CNA status at presentation and relapse in each patient was assessed with a newly developed research version of D006-X2 Multiple Myeloma digi-talMLPA probemix, as previously described (Supplementary Methods) [19,20]. Details of each probe and their respective genomic positions are provided in Supplementary Table 3. CNAs involving sex chromosomes were not considered. CNAs were called if ≥50% of probes mapping to an individual gene or genomic region deviated from normal diploid pattern.

Statistical analyses
Analyses were performed in R version 3.5.2 using "dplyr", "tidyr", "stats", "survival" and "survminer", "ComplexHeatmap" and "ggpubr" packages. The association between categorical variables was examined using Fishers exact test and between continuous variables using the Wilcoxon signed-rank test. Progression-free survival (PFS) was defined as time from induction randomization to progression, according to International Myeloma Working Group criteria, or death of any cause. Overall survival (OS) was time from induction randomization to death of any cause. Cox proportional hazards (PH) regression was used to estimate univariate and multivariable hazard ratios (HRs) and 95% confidence intervals (CI). To examine the predictive value of evolution in CNA status for subsequent therapy, high-risk CNA were considered as time dependent covariates within the multivariate model. The PH assumption was tested to investigate whether time dependent covariate effects persisted irrespective of time of acquisition (Supplementary Methods). Multivariate covariates included t(4;14), t(14;16), del(1p), gain(1q), del(17p) and treatment pathway (transplant eligible vs. not). This is relevant because, although the median follow-up of this cohort is long, the inherent requirement of a relapse biopsy excludes the longest responders from our analyses as they are still on active trial medication or observation with no evidence of disease progression. Differences between Kaplan-Meier survival curves was assessed using the log-rank test. A twosided P value of ≤0.05 was considered statistically significant.

Acquisition of new CNAs is a feature of relapse
Relapse was associated with the acquisition of new CNAs in 87.1% of tumors, most (73.2%) being large-scale chromosome changes. Across all patients, relapse was associated with a higher number of CNAs: median 11.5 (range 0-34) vs. median 12 (range 0-29) (P = 0.0058). As previously well-documented [16,21], IGH translocations were clonal at presentation and their status did not change at relapse.

Evolution of driver copy number aberrations
We next examined for clonal emergence at relapse of subclonal changes at diagnosis, focusing on the most frequent drivers-gain(1q) and del(17p). We and others showed before that calling of CNAs using conservative high confidence MLPA cutoff values detects clonal infiltration equivalent to about >20% by FISH, but that calling of minor sub clones is also feasible [4]. Thirty percent of patients' tumors with clonally detectable 1q gain at relapse had a detectable, potential minor sub-clonal gain(1q) at diagnosis and 50% of del(17p) relapse tumors had potential subclonal del(17p) at diagnosis (Supplementary Fig. 4). Progressive clonal expansion of 1q positive tumors was a feature of 17.5% gain(1q) cases, whereby gain at diagnosis evolved into tetraploidy of 1q (amp(1q)) at relapse. Two of 15 tumors with gain of MYC at diagnosis progressed to amplification at relapse. This was also a feature with gain of 15q (4.4%) and 19p (2.3%), albeit at low frequency.

Impact of subtype on copy number profile
Given the biological heterogeneity of MM we examined the relationship between CNAs and disease progression by subtype: those with IG translocations (t(4;14), t(11;14), t(14;16)/t(14;20)), and those with HRD, sub-grouped into those with and without gain of 11, in analogy to Translocation and Cyclin D (TC) classification (Supplementary  Table 2) [4,5].

Associations with patient outcome
CNAs were stratified by time point of acquisition, i.e., "Gain-Gain" for those with stable gain and "Diploid-Gain" for those with evolution of gain(1q) at relapse. Gain(1q) from presentation and evolution of new gain(1q) at relapse were both associated with significantly shorter OS compared to normal 1q copy number (HR 2.11; P = 0.0040 and HR 2.00; P = 0.021, respectively). Median OS was 44.3 vs. 47.9 vs. 67.1 months for gain(1q) at diagnosis, new gain (1q) at relapse and normal 1q copy number at both time points, respectively (log-rank P = 0.007) (Fig. 5).
Of interest, type of evolutionary pattern was associated with OS; branching evolution was significantly associated with the shortest and linear loss with the longest OS (HR 2.61, P = 0.0048), with median OS being 44.6 vs. 59.6 vs. 62.0 vs. 68.1 months for branching, linear, no change and linear loss respectively (log-rank P = 0.02) (Fig. 5).

Discussion
Our analysis demonstrates that progression of MM is characterized by emergence of clones with additional largescale chromosomal changes, commonly involving 1q. Gain or amp(1q) frequently co-evolves with other CNAs, either implicating 1q directly in their genesis as per "jumping 1q" hypothesis, or highlighting 1q as a region providing clonal advantage for genetically instable tumors [23]. While a number of genes on 1q have been proposed as drivers, our study does not provide data to make a specific inference [8,[24][25][26][27]. Our findings also suggest progression of MM is associated with positive selection of sub-chromosomal loss of 1p, implicating CDKN2C, and thereby indirectly CDK4/6 in conferring clonal survival benefit at progression. We also identified recurrent sub-chromosomal gain at relapse of an area to which MYC maps (chr8q24.21), in line with its driver role in B-cell malignancies [28,29]. By considering MM molecular subtype, we highlight differences in evolution trajectories, particularly between HRD and IG translocated MM, but also between t(4;14) and t (11;14) and/or between CCND2 and CCND1 driven tumors [5]. Of note, there is a marked enrichment of acquisition of del(17p) at relapse in CCND2 driven tumors which has, to our knowledge, not been described before. These differences are features of subtypes irrespective of therapy. Our findings are in keeping with published observations for evolution of pre-therapeutic mutational patterns but extend these in context of therapy and relapse [30][31][32]. We also describe, to our knowledge for the first time, an association between CNA evolution pattern and subsequent outcome. Whether individual CNA evolution pattern has independent clinical relevance and could be included in management considerations at relapse requires further investigation.
We identified only a low frequency of CNAs to which IMiD response genes map. In conjuncture with published somatic SNV data, results suggest mutation of these genes per se is not the major determinant of acquired resistance to lenalidomide, or other members of this class of agent in MM [12,[14][15][16][33][34][35].
HRD with gain (11) HRD no gain (11)  We demonstrate, to our knowledge for the first time, in a randomized controlled trial that acquisition of gain(1q) at relapse is independently and strongly associated with inferior OS [36]. Although findings are also indicative for del(17p), del(1p/CDKN2C) and gain(8q/MYC), our power to robustly assert clinical relevance of acquired lesions at relapse is inevitably limited by the lower frequency of these CNAs [37]. Clinical evaluation of these markers at relapse is technically feasible in most developed healthcare systems, but currently not widely recommended [38][39][40].
It is unknown whether sub-clones residing outside of the standard iliac crest bone marrow sampling area at diagnosis are the main source of clonal evolution detected at relapse or whether these predominantly emerge through ongoing genomic instability [12]. Multi-region bone sampling in MM is associated with significant risk and morbidity, making its implementation as part of standard care very challenging. Functional bone marrow soft tissue imaging techniques such as diffusion-weighted whole body MRI and/or molecular techniques including circulating tumor DNA profiling are promising methods in development which may contribute to diagnostic assessment of spatial MM heterogeneity in the future. However, sensitivity, precision and clinical relevance of minor sub-clone detection from a single time point for these methods remains to be established, in particular if results differ from those of parallel bone marrow sampling, before informing clinical management [41][42][43][44].
All patients received uniform trial treatment, in particular providing insight into thalidomide and lenalidomide associated CNA changes. However, patients also received lowdoses of the oral alkylator cyclophosphamide during induction, which may have specific impact on CNA changes, potentially limiting generalizability of our findings.
In addition, molecular information generated with a targeted tool like digitalMLPA is inherently focused and does not discriminate complex processes potentially underlying identified CNAs, such as chromotripsis or chromoplexy, which are detectable by discovery tools such as whole genome sequencing, or single nucleotide variants captured by myeloma specific CNA/sequencing panels [29,[45][46][47]. Our data also cannot reflect accompanying changes of the tumor microenvironment, which have been implicated in myeloma progression [48]. However, digitalMLPA is in development for clinical diagnostic application, offers sensitive CNA HRD with gain (11) HRD no gain (11)    information specific to myeloma from low tumor DNA quantities at high throughput. It requires limited computational infrastructure and uses standardized analysis algorithms, thus overcoming the significant limitations of conventional FISH analysis [20]. As our study demonstrates, digitalMLPA is suitable for longitudinal intra-individual tracking and provides a standardized and accessible method for cross-study validation and putative implementation in molecularly stratified prospective clinical trials. Genetic re-profiling at relapse in MM is currently not widely recommended and, as a consequence, often not reimbursed [10,39,49]. Whilst treatment options for relapsed disease were until recently very limited, a range of therapeutic regimens with varying intensity are now approved and available, requiring better tools for clinical decision making at relapse [50][51][52][53]. Adapting first-line therapy and its intensity to individual tumor risk markers in MM is now a key focus of research, with multiple genetically stratified prospective clinical trials currently ongoing [10,54,55]. The debate about optimal tailoring of second line treatment is likely to intensify with increasing therapeutic options. Our study demonstrates not only the frequency of emerging high-risk CNAs at MM relapse, but also the unmet clinical need of patients with chromosomally evolving MM. These patients should ideally be recognized early during relapse and prospective clinical trials investigating longitudinal management strategies adapted to dynamic risk profiling, designed [56].
Our findings strongly support repeated tumor molecular analysis in MM in context of modern treatment, even in circumstances where only selected markers such as 1q can be tested, as a means of tailoring patient treatment beyond first-line therapy.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons. org/licenses/by/4.0/.