The prognostic value of additional copies of 1q21 in multiple myeloma depends on the primary genetic event

Abstract Hyperdiploidy (HRD) and specific immunoglobulin heavy locus (IGH) translocations are primary chromosomal abnormalities (CA) in multiple myeloma (MM). In this retrospective study of 794 MM patients we aimed to investigate clinical features and common CA including gain(1q) in separate subgroups defined by primary CA. In the entire group, we confirmed that gain(1q) was associated with short time to next treatment and adverse overall survival (OS). The impact was worse for four or more copies of 1q21 as compared to three copies. However, in a subgroup of patients with clonal gain(11q) and without known primary IGH translocations (CG11q), already three copies of 1q21 were associated with a poor outcome; in the absence of gain(1q), patients in this subgroup had a remarkably long median OS of more than nine years. These cases were associated with HRD, coexpression of CD56 and CD117, male gender, and IgG subtype. In non‐CG11q patients, four or more copies of 1q21 (but not three copies) had a significant adverse impact on outcome. Several associations with CA and clinical findings were observed for the defined subgroups. As an example, we found a predominance of early tetraploidy, plasma cell leukemia, and female gender in the t(14;16) subgroup. Our results underscore the importance of subgrouping in MM.


| INTRODUCTION
Although novel drugs have improved the management of multiple myeloma (MM), the disease is still characterized by a marked clinical heterogeneity as reflected by overall survival (OS), ranging from less than two years to more than ten years. 1 Various factors such as patient fitness, therapy, microenvironment, and properties of the cancer itself including chromosomal abnormalities (CA) explain, at least in part, this heterogeneity. [2][3][4][5] With CA, MM can be broadly divided into two groups: about half of the cases with primary immunoglobulin heavy locus (IGH) translocations and the remaining with hyperdiploidy (HRD), the gain of odd-number chromosomes. 6 Both IGH translocations and HRD are considered primary genetic events, and as such they are mutually exclusive and present already in asymptomatic precursor stages and in the main clone. 7 These initiating events are followed by secondary events that eventually contribute to tumor progression and relapse. 8 In recent years, high-throughput technologies such as gene expression profiling (GEP) and next-generation sequencing (NGS) have been used to characterize myelomas in more detail in order to improve our understanding of myelomagenesis. 9 Michael Steurer died on March 11, 2019. Although another layer of complexity (eg, by showing clonal heterogeneity or many genes with recurrent mutations at low prevalence) was added, particularly by NGS, these studies also confirmed the importance of primary CA (ie, HRD and primary translocations) that define cytogenetic subgroups and give rise to a non-random accumulation of secondary events. [10][11][12] Fluorescence in situ hybridization (FISH) is implemented in standard clinical workflows for the detection of CA in order to identify high-risk patients. [13][14][15][16] Several CA, namely primary IGH translocations and secondary events, have been associated with adverse prognosis. However, binary risk stratification based on the presence or absence of high-risk CA might be oversimplified, and a possible explanation for heterogeneous survival of high-risk patients. 4 Recently, several new high-risk groups were defined based on additional markers, co-occurrence of adverse CA and weighted CA. 4,[17][18][19][20] Moreover, also the copy number (CN) of chromosomal gains might be associated with prognosis; the negative impact of gain(1q) on survival seems to be driven by the number of additional copies. 18,[21][22][23] To define the impact of CN of common CA on clinical outcome in MM we here provide a detailed analysis of CA in the context of defined subgroups in a series of patients in the Austrian Myeloma Registry that were mainly treated with novel drugs and analyzed by FISH.

| Patients
Between January 2010 and February 2020, 1023 bone marrow (BM) and 4 peripheral blood samples from 794 patients who had a confirmed myeloma diagnosis with a plasma cell infiltration of ≥10% and/or one or more myeloma-defining events 19  10%, numerical abnormalities: 20%). 13 If subsequent patient samples were available, retrospective analyses were performed solely with the result of the first obtained sample, unless otherwise stated. In the text of the article, a chromosomal gain without corresponding CN specification (eg, gain(1q)) is defined as three or more copies. The HRD was defined by a gain of any two of the chromosomal regions 5p15, 9q22, or 15q22. 24 Tetraploidy was predicted if three or more chromosomal regions had four or more copies detected with the standard FISH panel (1p36 or 1p32, 1q21, 11q22, 13q14, 14q32, and 17p13). Cytogenetic cancer clonal fraction of a particular aberration was calculated by dividing the number of affected cells by the number of aberrant cells with the largest detected aberration in the sample. Aberrations were classified as clonal or subclonal using 2/3 as cutoff.

| Statistical analyses
Time to next treatment (TTNT) was defined as time from treatment start to the date of starting second-line therapy, death from any cause, or the last follow-up. A new line of therapy was defined according to current guidelines. 25 The OS was calculated from treatment start until death from any cause or the last follow-up. Both TTNT and OS were estimated with the Kaplan-Meier method. Statistical differences between the survival curves were analyzed using the log-rank test. Univariate and multivariate analyses were performed with Cox regression models. The multivariate Cox regression models were adjusted for age, gender, induction therapy, beta-2 microglobulin (B2M), and high-risk CA. Additional CA with complete data and P values < .1 in the univariate Cox regression analyses were included in the multivariate Cox models. Continuous variables were analyzed using the Wilcoxon rank-sum test. Association between categorical variables was examined with Fisher's exact test, and P values were adjusted for multiple testing using the Benjamini-Hochberg method. The two-sided significance level was set at P value < .05. All computational analyses were performed using R version 3.6.0 (www.r-project. org/). The R packages included ggplot2, survival, and survminer.

| Patient characteristics
Patient characteristics are shown in Table S1. Median age of the 794 myeloma patients was 70 years (range, 34-93 years). Most of the patients (>95%) were treated by an induction with immunomodulatory drugs (IMiD) and/or proteasome inhibitors (PI), 44% underwent a front-line autologous stem cell transplantation (ASCT), and 39% received maintenance therapy.

| Cytogenetic landscape
An overview of the detected CA in the whole cohort is given in Figure 1A. Most common amplified regions (defined as regions with four or more copies) were observed at 1q21 and at 11q22 in 15% (119/794) and 11% (89/794) of the patients, respectively (Table S2).
In untreated patients, 3% of the cases showed multiple amplifications suggesting tetraploidy; the number of predicted tetraploid cases increased to 7% in treated patients ( Figures S1 and S2). Analysis of the cytogenetic clonal cancer fraction confirmed the oncogenic model of primary and secondary CA: IGH translocations t(4;14), t (11;14), and t(14;16) considered as primary events, were almost exclusively clonal and known secondary events such as gain(1q) or del(17p) were more often found to be subclonal ( Figure 1B). Pairwise associations confirmed the cytogenetic subgroups t(4;14) and t(11;14) as mutually exclusive ( Figure 1C). The analysis also showed a negative association between gain(11q) and both t(4;14) and t(11;14) cases. Translocation t (4;14) was associated with del(13q) and gain(1q), both of which are known to be linked to this subgroup. 10 Deletions (eg, del(14q), del (13q), and del(17p)) were associated with each other.

| Associations of subgroups
Based on our finding that the frequent gain(11q) was negatively associated with primary genetic events (ie, t(4;14) and t(11;14)), we introduced for subgroup analysis a subgroup that was defined by clonal gain(11q) and lack of primary IGH translocations (CG11q). Previous studies showed that within the HRD subgroup two clusters can be distinguished according to the presence of a chromosome 11 gain. 10,17 A substantial part of chromosome 11 gains is also found in the t (11;14) subgroup. 10,26 Therefore, cases with both gain(11q) and lack of primary IGH translocations (eg, t(11;14)) might often belong to the HRD cluster that harbors a chromosome 11 gain (HRD11+). We focused on clonal gain(11q), because as an early event the aberration might have a primary impact on pathogenesis. Using Fisher's exact test we investigated the correlation between five subgroups (ie, t(4;14), t(11;14), t(14;16), CG11q, and a group with the remaining cases) and CA, including amplifications, immunophenotypic findings, and clinical features (Tables S2 and S3). As expected, the CG11q subgroup was characterized by an association with HRD (P < .001), and, as reported in studies for HRD, 27,28 the subgroup was associated with IgG (P < .05) and correlated with antigenic coexpression of CD56 and CD117 (CG11q: 50% vs non-CG11q: 11%, P < .001; Tables 1, S2, and S3). About 20% of the CG11q cases showed an amplification of 11q22 (four or more copies). Furthermore, the t(14;16) subgroup was associated with tetraploidy (P < .001), plasma cell leukemia (PCL; P < .05) and lack of CD56 expression (P < .05) (Tables S2 and S3). The distribution of secondary high-risk CA showed that del(17p) was relatively evenly distributed across the different subgroups (6%-14% of the cases), while gain(1q) was enriched (P < .001) in t(4;14) cases (71%) and in the remaining cases (56%). These two groups and t (14;16) cases were also associated (P < .05) with four or more copies of 1q21 in 30%, 20%, and 53% of the cases, respectively. On the other hand, gain(1q) was negatively associated (P < .001) with CG11q and t (11;14) subgroups and present in 29% and 25% of the cases, respectively. Interestingly, del(1p32) was detected in all analyzed subgroups (2%-18% of the cases), whereas del(1p36) was exclusively found in the subgroup with the remaining cases (6% of the cases).
To analyze whether further gender differences were present in our cohort, we studied associations between gender and all CA and clinical characteristics. Additionally, female patients displayed a significantly higher frequency of del(13q) than did male patients (P < .001; Figure S3A, Table S4). This could not be explained by the observation that specific cytogenetic subgroups that co-occur with del(13q) are more prevalent in female patients ( Figure S3B). Furthermore, we found a significantly larger number of cytogenetic aberrations in female patients (P < .05; Figure S3C). Female gender was also positively associated (P < .05) with light chain only myeloma and an increased serum involved/uninvolved free light chain (FLC) ratio (≥100) ( Table S4). The results are shown in Table S5. Several parameters were associated with shorter TTNT and OS: gain(1q) (both three and four or more copies), del(13q), del(17p), B2M of 5.5 mg/L or higher, hemoglobin less than 10 g/dL, creatinine of 2 mg/dL or higher, platelets less than 150 x 10 9 /L, calcium of 2.75 mmol/L or greater, serum involved/uninvolved FLC ratio (≥100), International Staging System stage III (ISS F I G U R E 1 Cytogenetic landscape. (A) Co-segregation of chromosomal abnormalities in 794 myeloma patients detected with FISH probes. Samples were annotated for 1p testing (1p36: blue, 1p32: yellow), stage (treated vs untreated), age group (≥65 years vs <65 years), and gender (female vs male); (B) Percentage of cases in which a cytogenetic aberration is found to be subclonal or clonal is shown across the patient samples subjected to CD138+ plasma cell enrichment (n = 344). Abnormalities with a frequency of ≥2% in the cohort are shown in the panel. The boxplot showing the cytogenetic cancer clonal fraction (CCF) of the chromosomal abnormalities shows the median (thick black horizontal line) and at the vertical extremities of the boxes the 25th and 75th percentiles. Whiskers' ends represent minimum and maximum values; (C) Pairwise associations between the cytogenetic aberrations present in ≥2% of 794 myeloma patients. Associations are defined with Fisher's exact test; blue color indicates a positive association, whereas red color indicates a negative association. Adjustment for multiple testing was done using the Benjamini-Hochberg method and the size of the circle depicts the significance of the q value. Abnormalities of 1p (ie, gains or deletions) relate to either 1p36 or 1p32. Unspecified IGH indicates at least one unspecified IGH abnormality III), 29 revised ISS (R-ISS) III, 16 high-risk CA, 16 double-hit and triple-hit (co-occurrence of two or three adverse lesions, respectively), 17,19,20 four or more copies of 1q21 plus ISS III (defined as Double-Hit myeloma), 18 and PI-based induction. The parameters ISS I, R-ISS I, front-line ASCT, and CD27 were associated with longer TTNT and better OS. Median TTNT was 1.3 vs 1.7 vs 2.8 years (log-rank P < .001) and median OS was 3.1 vs 4.1 vs 6.9 years (log-rank P < .001) for four or more copies, three copies, and two or fewer copies of 1q21, respectively ( Figures S5A,B). In the multivariate Cox analysis, four or more copies of 1q21 (but not three copies), high-risk CA and B2M of 5.5 mg/L or higher were independent adverse prognostic factors for TTNT and OS ( Figures S5C,D). Some features were shown to be associated with survival in univariate analysis, but not included in the multivariate analysis because of their correlation with B2M

| Survival of the whole cohort
and/or their incompleteness of data (Tables S4 and S5; Figure S6). In addition, high-risk groups that comprised gain(1q) (eg, double-hit) as well as del(17p), t(4;14), and t(14;16) alone, which together defined the already included high-risk CA, were not used as factors. Patients with gain(1q) who underwent front-line ASCT had a longer TTNT (logrank P = .016) and OS (log-rank P = .004) than did patients with gain (1q) who did not receive front-line ASCT ( Figure S7). However, the significance was lost when cases with very short OS (less than six months) were excluded from the analysis. Furthermore, PI-based and IMiD-based induction and maintenance regimens were not associated with a statistically significant better outcome in patients with gain(1q).

| Impact of gain(1q) on survival of patients in the CG11q subgroup
Next, we separated the CG11q subgroup from the remaining cases to perform CG11q subgroup-specific Cox univariate analysis.
Age ≥ 65 years, gain(1q) (both three and four or more copies), creatinine of 2 mg/dL or higher, B2M of 5.5 mg/L or higher, ISS III, doublehit, four or more copies of 1q21 plus ISS III, and PI-based induction were associated with a negative impact on both TTNT and OS (Table S6). Front-line ASCT was associated with longer TTNT and better OS. Median TTNT was 1.6 vs 1.5 vs 3.3 years (log-rank P < .001) and median OS was 2.6 vs 3.3 vs 9.6 years (log-rank P < .001) for four or more copies, three copies, and two or fewer copies of 1q21, respectively (Figure 2A,B). In the multivariate analysis, gain(1q) with three copies was the only parameter that retained its adverse prognostic value for both TTNT and OS ( Figure 2C,D).

| Impact of gain(1q) on survival of patients in the non-CG11q cohort
In the univariate Cox analysis of non-CG11q cases the following parameters were associated with adverse TTNT and OS: four or more copies of 1q21, three copies of 11q22, del(13q), del(17p), high-risk CA, double-hit, triple-hit, four or more copies of 1q21 plus ISS III, B2M of 5.5 mg/L or higher, serum involved/uninvolved FLC ratio (≥100), ISS III, R-ISS III, and PI-based induction. Both R-ISS I and CD27 were associated with better outcome (Table S7). Median TTNT was 1.0 vs 1.9 vs 2.1 years (log-rank P = .018) and median OS was 3.1 vs 4.6 vs 6.0 years (log-rank P = .038) for four or more copies, three copies, and two or fewer copies of 1q21, respectively ( Figure 2E,F).
The survival curves between three copies and two or fewer copies were not significantly different. Gain(1q) with four or more copies and high-risk CA remained adverse prognostic factors in the multivariate analysis for TTNT and OS ( Figure 2G,H). Of note, neither three copies of 1q21 nor four or more copies of 1q21 had a significant impact on survival in the t(4;14) and t(11;14) subgroups alone ( Figure S8A-D). It seems that the observable poor impact of four or more copies of 1q21 in the non-CG11q cases is mainly driven by the remaining cases ( Figure S8E,F).

| Associations of high-risk myeloma
In our cohort, the number of patients with multiple amplified regions increased during disease course, mirroring clonal evolution. We stud- , and t(14;16)) and high-risk CA were common cytogenetic features of both tetraploidy (P < .05) and PCL (P < .05 and P < .001, respectively). Baseline characteristics such as del(17p), hemoglobin less than 10 g/dL, and platelets less than 150 x 10 9 /L were associated with primary PCL (pPCL) in 38% (P < .05), 80% (P < .05), and 89% (P < .001) of cases, respectively. Secondary EMM (sEMM) was associated with patients who received three or more lines of therapy (76% of the patients; P < .001) and acquired a tetraploid clone (23% of the patients; P < .05) during their course of disease. No other specific association with clinics or cytogenetics (eg, subgroup) was observed for EMM. Median time between therapy initiation and diagnosis of secondary PCL (sPCL), sEMM, and late tetraploidy was 1.6, 1.8, and 2.8 years, respectively. Primary and sPCL, sEMM, and late tetraploidy were all associated with very poor survival after detection (less than one year), whereas pEMM and early tetraploidy were associated with a relatively better outcome ( Figure 3A-C).

| DISCUSSION
Our results highlight the importance of defining cytogenetic subgroups in MM as this has an impact on the course of the clinical disease. It may be of particular importance to integrate a more advanced  (Table S4), indicating that gain(1q) does not impact the expression of these antigens, which was previously associated with good prognosis (CD117) 31 and dependence on the BM microenvironment (CD56). 32 In contrast to the CG11q subgroup, the negative impact of gain(1q) was less pronounced in the non-CG11q cohort and restricted to four or more copies of 1q21. Amplifications of 1q21 (four or more copies) are known to be accompanied by high-risk states. 33 In line with this, we observed an association between four or more copies of 1q21 plus ISS III (double-hit patients) 18 and tetraploidy, which has been correlated with genomic instability, advanced disease, and poor prognosis. 34,35 The Overall survival (OS) after detection of early tetraploidy vs late tetraploidy, (B) OS after detection of primary plasma cell leukemia (pPCL) and secondary PCL (sPCL), (C) OS after detection of primary extramedullary multiple myeloma (pEMM) and secondary EMM (sEMM) is shown. The log-rank test was used to perform group comparisons adverse prognosis of CG11q patients harboring gain(1q), already observable with three copies of 1q21, might be linked to an increased expression of both CCND1 (D1; associated with gain(11q)) and CCND2 (D2; associated with gain(1q)). 17 patients with gain(1q) may also play a role in the observed adverse phenotype. 37 We analyzed several high-risk groups with cytogenetic markers such as high-risk CA (ie, del(17p), t(4;14), and t(14;16)), R-ISS III, double-hit as well as triple-hit, and four or more copies of 1q21 plus ISS III, which in the entire group had an incidence of 20%, 15%, 11%, 1%, and 9%, respectively ( Table S2). All of these parameters were significantly associated with adverse survival for the entire cohort (Table S5). However, four of five of these high-risk definitions were associated with specific non-CG11q cases (ie, t(4;14) and/or t(14;16) subgroups) and therefore might be less appropriate for the identification of high-risk patients in other subgroups such as CG11q.
In our cohort, similar to findings recently reported, 34 about 30% of the t(14;16) cases had an early tetraploidy, indicating that whole genome-doubling is a relatively early event in this MM subgroup.
Additionally, our data suggest that tetraploidy acquired in a late phase of the disease is associated with a prognosis that is similarly poor as for sPCL and sEMM. Interestingly, we found gender discrepancies for CA consistent with a previous analysis of the MRC Myeloma IX dataset. 38 Gender discrepancies, which have been observed in other cancers including hematological malignancies, could also comprise molecular lesions. 39,40 Analyzing the gender discrepancy in more detail in future studies will contribute to our understanding of the pathobiology of MM. This study is limited by the retrospective nature of data collection. As clinical annotation was very comprehensive in We conclude that cytogenetic subgroups in MM differ in various aspects in our cohort and that evaluation of secondary genetic events on the basis of cytogenetic subgroups might further improve MM risk stratifications. Our data suggest that already three copies of 1q21 are associated with an adverse outcome in patients of the CG11q/ HRD11+ subgroup. The 1q21 testing in this subgroup might enable patients to be stratified in a group with adverse prognosis as well as a group with a very favorable outcome. In non-CG11q patients, four or more copies of 1q21 (but not three copies) were associated with a significant adverse impact on the outcome.

ACKNOWLEDGMENTS
We would like to thank the patients in the Austrian Myeloma Registry and the staff at the Institute of Human Genetics of the Medical University of Innsbruck.