Sequencing‐based analysis of clonal evolution of 25 mantle cell lymphoma patients at diagnosis and after failure of standard immunochemotherapy

Our knowledge of genetic aberrations, that is, variants and copy number variations (CNVs), associated with mantle cell lymphoma (MCL) relapse remains limited. A cohort of 25 patients with MCL at diagnosis and the first relapse after the failure of standard immunochemotherapy was analyzed using whole‐exome sequencing. The most frequent variants at diagnosis and at relapse comprised six genes: TP53, ATM, KMT2D, CCND1, SP140, and LRP1B. The most frequent CNVs at diagnosis and at relapse included TP53 and CDKN2A/B deletions, and PIK3CA amplifications. The mean count of mutations per patient significantly increased at relapse (n = 34) compared to diagnosis (n = 27). The most frequent newly detected variants at relapse, LRP1B gene mutations, correlated with a higher mutational burden. Variant allele frequencies of TP53 variants increased from 0.35 to 0.76 at relapse. The frequency and length of predicted CNVs significantly increased at relapse with CDKN2A/B deletions being the most frequent. Our data suggest, that the resistant MCL clones detected at relapse were already present at diagnosis and were selected by therapy. We observed enrichment of genetic aberrations of DNA damage response pathway (TP53 and CDKN2A/B), and a significant increase in MCL heterogeneity. We identified LRP1B inactivation as a new potential driver of MCL relapse.


| INTRODUCTION
Mantle cell lymphoma (MCL) is a heterogeneous disease characterized by a chronically relapsing clinical course. 1,2 Despite the implementation of several innovative drugs and adoptive immunotherapy with genetically modified autologous T-lymphocytes into clinical practice, the standard front-line therapy for patients with newly diagnosed MCL is still based on immunochemotherapy with or without highdose therapy, and autologous stem cell transplantation. 3 In the last decade, whole exome sequencing (WES) of large unbiased cohorts of newly diagnosed MCL patients enabled not only the identification of driver genetic lesions (variants, copy number variants) but also genetic profiles associated with clinical outcome. [4][5][6][7][8][9][10][11][12][13][14][15] In contrast, our knowledge on the clonal evolution of MCL after the failure of standard immunochemotherapy remains limited. 7,16 In this study, we used WES to study mutational profiles of 25 patients with MCL, who experienced relapse or progression after standard front-line immunochemotherapy.

| Next-generation exome sequencing and copy number variant analysis
Genomic DNA was extracted using DNeasy Blood & Tissue Kit (Qiagen, Germany) according to the manufacturer's protocol. Samples were sequenced on the NextSeq 500 instrument (Illumina, San Diego, CA) according to the manufacturer's protocols with sequencing libraries prepared using SureSelectXT Human All Exon V6 + UTR kit (Agilent Technologies, Santa Clara, CA). The resulting reads were then aligned against the human reference genome (build GRCh37). All alignments were performed by BWA. 17 Mean coverage per sample was 66 reads.
Genomic variants were called with samtools and VarScan 2. 18,19 Variant annotation was performed using SnpEff. 20 Only nonsynonymous variants in the gene coding regions with coverage of at least 10 reads with mapping quality and base quality scores higher than 20 in related samples from each patient were compared together based on their frequency. Variants present in the patient's germline DNA at a frequency higher than 0.05 were excluded from analysis in all cases. We compared variants with an allele frequency ≥0.1 in at least one of the compared samples that were present in at least three reads in both, diagnostic and relapsed sample. Both the filtering of variants and plotting of counts and frequencies of variants was done in RStudio. The variants were manually reviewed in Integrative Genomics Viewer to exclude sequencing artifacts or variants present but not called in the germline sample. 21,22 "Oncoprint" function from the ComplexHeatmap R package was used to visualize detected variants. 23 Resulting plots show variants present in at least two patients. Mutations of TP53 were plotted on a lollipop plot using the maftools package. 24 A complete list of variants filtered out are available at the Table S1.
Copy number variants (CNVs) were predicted using CNVkit and a pooled reference from normal control samples. 25,26 Genomic segments were identified as amplifications or deletions based on their allelic copy numbers. We used the "weight" value from the CNVkit output as a metric that combines both segment length and sequencing coverage. The "Genomic Identification of Significant Targets in Cancer" (GISTIC 27 ) was used for assigning the statistical significance to the amplified and deleted segments identified by CNVkit. The broad GISTIC algorithm was used to find significantly altered copy numbers on the levels of chromosome arm regions and of individual genes. To generate color-coded heatmaps, these genes were filtered out using a CNV gene list of 22 genes (Table S1). [4][5][6][7]9,11,[13][14][15][16][28][29][30][31][32] An overview of data processing pipelines is shown in Figure S1.

| Targeted next-generation sequencing
For details, see Supplementary Materials-Methods.

| TaqMan copy number assay
TaqMan Copy Number Reference Assay RNase P (ThermoFisher scientific) was performed according to the manufacturer's instructions.
Briefly, DNA was diluted to the concentration of 5 ng/μL in nuclease-free water and stored at À20°C. The reporter dye of the Copy Number Assay primers was FAM w/o quencher, and the reporter dye of the Copy Number Reference Assay was VIC with TAMRA probe quencher. The assay was completed by a QuantStudio 7 Pro PCR analyzer (ThermoFisher Scientific) and had 40 cycles with the denaturing steps at 95°C for 15 0 and the annealing steps at 60°C for 60 0 .
Signal from the samples was thresholded to the signal of RNAse P and cycle threshold difference (dCT) was calculated as (CT sample À CT control ).
Allele copy number was then calculated as 2 Â 2 ÀdCT . found to be 5% (mean ± 3SD) for losses (deletions, monosomies).   Table 1. Briefly, 75% of patients were men (median = 68 years, range 47-81 years), 84% had high-risk disease according to MIPI, 59% had adverse morphology (pleomorphic or blastoid), and 56% high proliferation index by Ki-67 (≥30%). Fifty-six percent of patients were treated with R-CHOP-like regimens, and 44% received intensified treatments (for details, see Table 1). 33,34 The best response to front-line therapy is shown in Table 1. Briefly, response assessed by CT or PET-CT (i.e., complete, or partial remission = CR or PR) was observed in 23 out of 25 patients, stable disease (SD) was observed in two patients. The median EFS and OS for the whole cohort were 10 months (1.3-90.6 months) and 29 months (9.5-109.5 months), respectively.

| Genetic alterations identified by WES
A rigorous filtration process of the dataset revealed 922 nonsynonymous variants (referred to as "variants"), 90% of which were single-nucleotide variants (SNVs), the rest was represented by small indels. Overall, 616 variants were detected both at the diagnostic (DG) and relapsed samples (REL) (i.e., shared variants). In addition, 236 and 70 variants were newly detected (N/D), and newly undetected (N/U), respectively, at relapse compared to diagnosis. There were 24.6 ± 1.7 (mean ± SEM) shared variants per patient between the diagnostic and relapsed sample. There were 9.8 ± 1.9 newly detected variants at relapse (N/D) and only 3.5 ± 0.8 newly undetected variants at relapse (N/U), which is significantly lower (p < .0001; Figure S2A). Consequently, the total mean count of variants per patient was significantly higher at relapse (34.1 ± 2.9) compared to diagnosis (27.4 ± 1.8, p < .001) ( Figure S2B). Detailed overview of counts of shared, N/D and N/U variants for each patient is shown in a stacked bar plot at Figure S3A.
The mutational pattern between the diagnostic and relapsed samples remained similar and comprised predominantly missense variants (82.7% at diagnosis and 83.1% at relapse), stop-gained variants (7.1% at diagnosis and 7.0% at relapse), and frameshift variants (6.9% at diagnosis and 6.6% at relapse). The most frequent substitutions at both diagnostic and relapsed samples were transitions ( Figure S2C,D). Prediction of variant's impact by SnpEff toolbox revealed 97 high-impact mutations (i.e., frameshift and stop-gained variants) at diagnosis and 117 at relapse, respectively (20.6% increase at relapse). Most of the variants were predicted to have moderate impact (i.e., missense variants, inframe deletions), namely 583 at diagnosis, and 727 at relapse (24.7% increase at relapse; Figure S2E).
The majority of TP53 variants were localized in the p53 DNA binding domain (9/12, 75%), two in the p53 tetramerization motif (2/12, 17%) and one in the untranslated region ( Figure 2D). Of note, mutations of TP53 retained a trend of unfavorable prognostic significance even in this preselected cohort of relapsed MCL patients ( Figure S5, part 5A).

| Predicted CNV profile of MCL patients at diagnosis
Significant arm-level CNVs at diagnosis (i.e., frequency score > 0.15,   gene; Figure 3A). As previously shown for other malignancies, mutations of LRP1B at relapse significantly correlated with higher total mutation burden in MCL and at diagnosis had a trend toward shorted OS ( Figure S5, part 5B and Figure S8). Other newly detected variants comprised mutations of KMT2D, HOXD9, CDC27, RYR2, and FLNA genes (each newly mutated in two patients) ( Figure S3B). Complete mutational profile of variants found at relapse in at least two patients is shown in Figure S9. Of note, from the list of significantly mutated genes at diagnosis (i.e., genes mutated in ≥2 patients), five genes were newly mutated in one additional patient at relapse including ATM, CASP5, ETNK1, LRRIQ1, and NOTCH2.

| Predicted CNV profile of MCL patients at relapse
Compared to diagnosis, arm-level events consisted of one newly amplified chromosome arm (7p) and three newly deleted chromosome arms (6q, 8p, and 21p). All chromosomal arms significantly altered at diagnosis (q < 0.01) were significantly altered at relapse (Figure 2A).
The majority of CNV changes detected at diagnosis were also observed at relapse. However, deletions and amplifications detected at relapse were more frequent, and were larger in scale (i.e., had bigger weight, p < 0.05; Figure 2B). The most prevalent alterations at lymphoma relapse were deletions of CDKN2A and CDKN2B genes, newly predicted in 8 patients (in total 18 out of 24 analyzed patients at relapse, 75%; Figure 3B). In univariate analysis, the predicted  (Table S2). 4,5,7,10,16,28 The data suggest that the  Table S2).