Supplement: Genetic landscapes of relapsed and refractory DLBCLs

Supplemental Figures

Supplemental Figure S1.Overview of somatic copy number alterationsin rrDLBCLs.

Supplemental Figure S2.Consistent low coverage in the first exon of FOXO1.

Supplemental Figure S3.Overview of FOXO1 mutations in rrDLBCLs.

Supplemental Figure S4. VAFs corrected using purity estimates.

Supplemental Figure S5.Somatic copy number alterations affecting lymphoma-related genes.

Supplemental Figure S6: Recurrent deletions in NFKBIE.

Supplemental Tables

Supplemental Table S1.Clinical details and sample processing of 38 patients with relapsed or refractory DLBCL and TLy

Supplemental Table S2. Characteristics of samples used for exome sequencing and targeted sequencing of selected genes.

Supplemental Table S3.Somatic variants detected in rrDLBCL exomes.

Supplemental Table S4.Variants detected by targeted sequencing of rrDLBCL cases.

Supplemental Table S5.Overview ofSTAT6 variants detected by targeted sequencing a validation cohort of rrDLBCL and FL.

Supplemental Table S6. Comparison of VAF in diagnostic samples using an alternative approach.

Supplemental Figure S1.Overview of somatic copy number alterationsin rrDLBCLs.

Somatic copy number alterations detected from the paired exome data in each patient using TITAN are shown above. Copy number increases (shades of red) and losses (blue) were common and included many events known to be common in diagnostic DLBCL. The locations of the recurrently mutated genes shown in Figure 1 are indicated.

Supplemental Figure S2.Consistent low coverage in the first exon of FOXO1.

The first exon of FOXO1 is poorly covered in most Illumina exome and genome sequencing libraries due to very high GC content in the region. Shown above is exome data for the first exon from three of the patients (QC2-07, QC2-19 and QC2-39), each of which were found to harbor a FOXO1 mutation using targeted sequencing.

Supplemental Figure S3.Overview of FOXO1 mutations in rrDLBCLs.

The locations of somatic FOXO1 mutations detected within the rrDLBCLcohort by targeted sequencing are shown relative to the exon and protein domain structure. The variants shown above include those identified in five additional clinical trial (QCROC2) patients biopsied at relapse not included in Figure 1. Mutations largely affected known hot spots near the N-terminus including the initiator codon (M1) and around one of the phosphorylation sites (T24) as well as S205 in the DNA binding domain (red). One patient (QC2-015) harbored two mutations (M1V and I10V) and another patient (MT-260) had three mutations (P16T, S22P and S205N). Interestingly, the S205N mutation in this patient was not found in the matched diagnostic tumor and it appeared sub-clonal at relapse (Figure 2).

Supplemental Figure S4. VAFs corrected using purity estimates.

In tumors comprising 100% tumor cells, VAFs for clonal heterozygous mutations are expected to be 0.5 whereas LOH can cause a VAF of 1.0. For some samples, the VAFs were clearly affected by tumor purity causing proportionately lower VAFs for all mutations. The above examples show the VAFs following correction for tumor purity, which was estimated from the maximal VAFs observed and known copy number state of individual mutations. In QC2-39, we estimated 0.26 cellularity for the primary tumor and 0.80 for the relapse tumor. Following this correction, it appears in this sample that only mutations in FOXO1, B2M an SOCS1 underwent clonal expansion between diagnosis and relapse. In the second example (QC2-32), all mutations were stable across samples with the exception of the EZH2 mutation, which underwent a clonal expansion, and the MLL3 mutation, which presumably existed in a sub-clone that was largely extinguished. In QC2-34, the data from a separate capture-based experiment is shown. In both experiments, many of the SNVs had very low VAFs in the diagnostic tumor sample. An additional mutation in GNA13, which was not assessed in the amplicon sequencing experiment, demonstrated a higher VAF and was used to estimate the purity of the tumor sample. After adjusting the VAFs from the diagnostic specimen for purity, it was evident that the clone bearing the EZH2 mutation underwent a contraction and a sub-clone harboring a STAT6 mutation and several BCL2 mutations became dominant. This clone also gained mutations in MEF2B and TBL1XR1 as well as three new mutations in BCL2, all of which were undetectable in the primary tumor. In QC2-35, the purity of the relapse tumour was estimated to be 0.285 based on the abundance of the heterozygous TP53 mutation, which was interpreted as being present in the dominant clone in each tumour. Adjustment of the VAFs in this case reveals clonal expansion of mutations affecting STAT6, TNFRSF14 and CARD11(Table 2; Supplemental Table S6).

Supplemental Figure S5.Somatic copy number alterations affecting lymphoma-related genes.

The above image illustrates the overlap of somatic CNAs and genes shown in Figure 1. Cells marked with a backslash indicate the presence of one or more SNVs in the gene and those marked with an X indicate samples bearing an indel in the gene (* = both). Many of the genes that were recurrently affected by somatic SNVs and indels were also affected by copy number alterations in multiple patients. CNAs predicted to increase Jak/Stat signaling were common in this cohort including amplifications of STAT6 (7 cases) and deletions affecting SOCS1 (2 cases). With only two exceptions, mutations and CNAs affecting these two loci were mutually exclusive from cases harboring SNVs in these genes.

Supplemental Figure S6: Recurrent deletions in NFKBIE.

Of the 38rrDLBCLs sequenced, two separate patients (QC2-40 and QC2-25) harbored the same 4-nt deletion in NFKBIE. Targeted sequencing of NFKBIE in untreated DLBCLs revealed two additional patients with the same 4-nt deletion. A third patient in the rrDLBCL cohort (QC2-39) contained two separate insertions resulting in a frameshift and each of these variants was absent from the diagnostic biopsy (Figure 2).

1

Supplemental Tables

Supplemental Table S1.Clinical details for rrDLBCL patients.

Sample ID / Age¶
sex / Path T1 date of Dx / # Rx / Path T2
date of Bx / COO / % tumor
T2 / Exome sequencing / Targeted sequencing
QC2-3 / 79M / FL
03/09/2008 / 3 / TLy
03/02/2011 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-4 / 52M / DLBCL
13/04/2006 / 5+RT / DLBCL
11/05/2011 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-5 / 49M / DLBCL
05/10/2009 / 4+RT / DLBCL
24/05/2011 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-7 / 59M / DLBCL
15/04/2010 / 6+RT / DLBCL
15/06/2011 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-9 / 72F / DLBCL
08/02/2011 / 2+RT / DLBCL
07/12/2011 / GCB / 70 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-11 / 56M / FL-G2
24/02/95 / 6+RT / TLy
09/12/2012 / GCB / <20 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-12 / 54M / FL
25/08/94 / 6+RT / TLy
22/12/2011 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-13 / 53F / DLBCL
14/04/2010 / 5+RT / DLBCL
23/01/2012 / ABC / 60 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-15 / 75/F / DLBCL
07/10/2010 / 1 / DLBCL
16/02/2012 / GCB / 100 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-17 / 49F / DLBCL/FL
13/03/2006 / 2+RT / TLy
22/02/2012 / GCB* / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-18 / 61F / DLBCL
05/07/2011 / 2+RT / DLBCL
07/05/2012 / ABC / 90 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-19 / 68F / DLBCL
01/11/2008 / 2 / DLBCL
25/06/2012 / ABC / 90 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-20 / 60F / DLBCL
01/08/2010 / 6 / DLBCL
04/07/2012 / GCB / 90 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-21 / 71F / DLBCL
13/07/2011 / 3 / DLBCL
23/07/2012 / NA / <20 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-22 / 67F / DLBCL
15/03/2007 / 2 / DLBCL
23/07/2012 / ABC / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-23 / 59M / DLBCL
15/11/2007 / 3 / DLBCL
10/10/2012 / ABC / 100 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-25 / 65F / DLBCL
12/04/2010 / 4 / DLBCL
26/11/2012 / ABC / 100 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-26 / 63M / DLBCL
23/09/2011 / 1 / DLBCL
04/12/2012 / ABC / <20 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-30 / 41M / FL
24/12/2008 / 8 / TLy
13/07/2013 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-32 / 47F / FL
06/01/2007 / 3+RT / TLy
11/07/2013 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-33 / 56F / DLBCL
15/04/2011 / 3 / DLBCL
17/07/2013 / GCB / 100 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-34 / 68M / FL
18/02/2010 / 2 / TLy
23/07/2013 / GCB / 70 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-35 / 73F / FL
03/01/2006 / 4+RT / TLy
23/09/2013 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-36 / 75M / DLBCL
03/09/2012 / 1 / DLBCL
25/09/2013 / ABC / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-39 / 29F / DLBCL
29/04/2013 / 3+RT / DLBCL
11/10/2013 / GCB / 80 / T2, B cell rich
Germline, PBMC / T2, B cell rich
T1, no enrichment
QC2-40 / 75M / WM
07/12/2007 / 1 / TLy
10/10/2013 / ABC / 100 / T2, B cell rich
Germline, PBMC / T2, B cell rich
QC2-42 / 63F / DLBCL
15/12/2010 / 4 / DLBCL
13/12/2013 / GCB / 100 / T2, B cell rich
Germline, PBMC / T2, B cell rich
CH-50 / 69F / DLBCL
05/09/2007 / 1 / DLBCL
07/07/2009 / ABC / 80 / T2, no enrichment
Germline, PBMC / Not performed
CH-52 / 57M / DLBCL
12/07/2007 / 1+RT / DLBCL
23/12/2012 / ABC / >95 / T2, no enrichment
Germline, PBMC / Not performed
CH-54 / 65M / DLBCL
18/07/2005 / 1 / DLBCL
17/08/2009 / GCB / >90 / T2, no enrichment
Germline, PBMC / Not performed
CH-56 / 22F / DLBCL
30/11/2002 / 1 / DLBCL
12/12/2003 / ABC / NA / T2, no enrichment
Germline, PBMC / Not performed
CH-60 / 83M / DLBCL/FL
04/12/2003 / 1+RT / TLy
25/03/2011 / GCB / >90 / T2, no enrichment
Germline, PBMC / Not performed
CH-103 / 60M / DLBCL
14/11/2011 / 1+RT / DLBCL
13/05/2012 / GCB / 100 / T2, no enrichment
Germline, PBMC / Not performed
CH-109 / 61F / DLBCL
28/10/2010 / 1 / DLBCL
17/05/2011 / ABC / 100 / T2, no enrichment
Germline, PBMC / Not performed
CH-144 / 48M / DLBCL
12/05/2009 / 1 / DLBCL
20/08/2009 / ABC / NA / T2, no enrichment
Germline, PBMC / Not performed
MT-198 / 44M / FL
01/05/2007 / 2 / TLy
10/11/2009 / GCB / NA / T2, no enrichment
Germline, PBMC / T2, no enrichment
T1, no enrichment
MT-260 / 62F / FL
23/12/2005 / 1 / TLy
12/01/2010 / GCB / NA / T2, no enrichment
Germline, PBMC / T2, no enrichment
T1, no enrichment
MT-439 / 71F / MZL
31/10/2008 / 1 / TLy
30/06/2009 / GCB / 90 / T2, no enrichment
Germline, PBMC / T2, no enrichment
T1, no enrichment

Abbreviations: M denotes male sex, F denotes Female sex; Path, pathology; T1, diagnostic biopsy; T2, biopsy at relapse used for study; Dx, diagnosis; FL, follicular lymphoma; DLBCL, diffuse large B cell lymphoma; WM, waldenstrommacroglobulinemia; MZL, marginal zone lymphoma; TLydenotes a DLBCL that arose in the context of histological transformation from an indolent lymphoma; Bx, biopsy; # Rx, number of treatment regimens administered to the patient before the acquisition of T2; RT; radiotherapy; disease; NA, not available; COO, molecular subtype by cell of origin; IHC, immunohistochemistry; ABC, activated B cell type; GCB, germinal center B cell type; PBMC, peripheral blood mononuclear cells

¶ age is at the time of relapse. * GCB phenotype based on EZH2 mutation (unclassifiable by GEP and non-GCB by IHC).

%T2 denotes the percentage of tumor cells by immunohistochemistry. In the cases of the QCROC samples, this was the percentage of tumor before B cell enrichment.

Supplemental Table S2: Characteristics of samples used for exome sequencing and targeted sequencing of selected genes including STAT6

PRE-THERAPY
number of cases with STAT6 mutations / POST-THERAPY
number of cases with STAT6 mutations
EXOME SEQUENCING
TLy
DLBCL
ABC
GCB
Not available / Total: 138
publically available exomes, no COO provided
5/138 (4%) / Total: 38
5/13 (38%)
4/25 (16%)
0/13 (0%)
4/11 (36%)
0/1 (0%)
TARGETED SEQUENCING ONLY
TLy
DLBCL
ABC/Non-GCB
GCB
Not available
Initial clinical characteristics of patients with DLBCL:
Age ≥ 60
Stage III or IV
LDH > normal
ECOG PS ≥ 2
ENS ≥ 2
IPI 0-1
2-3
4-5 / Total: 49
4/22 (18%)
4/27 (15%)
0/10 (0)
3/10 (30%)
1/7 (14%)
17/27 (63%)
17/27 (63%)
12/27 (44%)
2/27 (7%)
6/26 (23%)
11/27 (40%)
12/27 (44%)
4/27 (15%) / Total: 38
Results confirm exome sequencing, no additional mutations detected.
Frequency of STAT6 mutations
TLy
DLBCL by targeted sequencing
DLBCL by exome sequencing / 4/22 (18%)
4/27 (15%)
5/138 (4%) / 5/13 (38%)
4/25 (16%)
4/25 (16%)

Supplemental Methods

General characteristics of relapsed or refractory DLBCL

The clinical characteristics and outcome varied in our discovery cohort. The 27rrDLBCL samples obtained in the context of a clinical trial had been exposed to more lines of therapy compared to the other 11 archived samples that were obtained at first progression after RCHOP (mean number of therapies 3.5 vs 1, p<0.001) (Supplemental Table S1 for details).

Sample preparation

In 12 patients, formalin fixed paraffin embedded tissue (FFPET) on the original diagnostic biopsy was available for analysis. The DNA from these cases was extracted using the Recover All Total Nucleic Acid Isolation kit (Ambion). Germline DNA from peripheral blood mononuclear cells and DNA from fresh frozen lymphoma tissue that were not part of the QCROC2 cohort were isolated using the All-Prep RNA/DNA kit (Qiagen).

Deep amplicon sequencing for mutation screening and assessment of clonal evolution

Amplicon sequencing was performed on diagnostic and relapse DLBCL sample pairs from 12 patients. Genes known to be recurrently mutated in DLBCL and found by exome sequencing to be mutated in our discovery sequencing cohort were selected for targeted amplicon sequencing. The selected mutated genes were: FAS, MYD88, CREBBP, FOXO1, HIST1H1E, MLL2, SGK1, STAT6, TNFRSF14, TP53, BTG1, ARID1, CSPP1, PIM1, DUSP2, GNAI2, MYC, RB1, BTG2, CD79B, DST, MLL3, NFKBIE, NFKBIZ, TMSB4X, TBL1XR1, EZH2, B2M, and SOCS1. The aim was to investigate the clonal expansion of these mutations between diagnostic and relapse. The amplicon libraries were prepared using a nested PCR protocol. For the first round (locus specific amplification) primer pairs flanking the mutated regions were designed. Primers were 16-22 bp long with an average of 58-60ᵒC annealing temperature. Two indexed sequences CGCTCTTCCGATCTCTGNNNN and TGCTCTTCCGATCTGACNNNN were added to each forward and reverse primers, respectively. Thermal cycling condition for the first PCR started with initial incubation at 98°C for 30 seconds, followed by 35 cycles at 98°C for 10 seconds, 63°C for 30 seconds, and 72°C for 30 seconds. Additionally, a final extension step at 72°C for 2 min followed the last cycle.Between 3 and 7 primer pairs were multiplexed in PCR reactions containing 5 ng of DNA extracted from FFPE or fresh frozen tumor samples. For each multiplex reaction, a negative control with no template DNA was prepared, ensuring for absence of any contamination with other DNAs or PCR amplicons from the environment. The products of PCR1 were run on 3% agarose gels to confirm the presence of PCR products and absence of any contamination in negative control reactions. After validation, PCR products (PCR1) were cleaned up using AgencourtAmpure XP beads (Beckman Coulter). Second PCR (PCR2) was performed using the forward primer AATGATACGGCGACCACCGAGATCTACACTCTTTCCCTACACGAC and reverse primer CAAGCAGAAGACGGCATACGAGATXXXXXXGTGACTGGAGTTCAGACGTGTGCTCTTCCG. For each reaction the XXXXXX sequence varied based on the selected primer for multiplexing the final libraries. The thermal cycling condition for PCR2 was as follows: initial incubation at 98°C for 30 seconds, followed by 6 cycles of 98°C for 10 seconds, 65°C for 30 seconds, and 72°C for 30 seconds, and then a final extension step at 72°C for 5 minutes. 10 ng of purified PCR1 product were used as template for the second PCR. The Products of PCR2 were again purified as described above. PCR reactions in both steps were prepared following the Q5® High-Fidelity DNA Polymerase (M0491) protocol. The final amplicon libraries varied in size from 150 to 250 bps. For targeted sequencing of the STAT6 DNA binding domain, we prepared 3 sets of primers that would amplify the exons encoding this region of the protein(residues 295 to 535).The amplicon libraries were sequenced using an Illumina MiSeq on a 300 cycle version 2 kit. The sequencing reads were aligned to hg19 using BWA-mem and SAMtools. The aligned reads were visualized and the variant allele fraction (VAF) wasextracted using the Integrative Genomics Viewer (IGV). Where shown, VAFs were corrected by dividing by tumor purity estimates.

To determine the reproducibility of this approach using an alternative strategy, we performed capture-based sequencing of our gene panel using the methods described below for three samples. These data are included in Supplemental Table S6 and Supplemental Figure S4. Although the VAFs detected by amplicon sequencing were typically uniformly lower than those obtained by this approach, the relative levels of individual mutations were reproduced and overall the VAFs showed a strong correlation (Pearson’s correlation coefficient = 0.8958508, P=0.0001909). Owing to the breadth of this complementary strategy, additional mutations not assessed by amplicon sequencing were found to result from clonal selection.

Targeted re-sequencing of candidate genes using hybridization capture

For most samples, we sequenced the genes described in the main text using the method described. In samples for which this identified no mutations, we performed a second sequencing using baits targeting a larger set of genes: STAT6, MYD88, CD79B, CCND3, EZH2, BCL2, MEF2B, TP53, ID3, TCF3, FAS, B2M, HIST1H1E, SOCS1, FOXO1, MS4A1, CD58, NFKBIE, MYC, PIM1,TNFAIP3,TNFRSF14,MLL2, ETS1,TMEM30A, IL4R,CARD11,TBL1XR1,TMSB4X,CREBBP,GNAI2,HIST1H1C,SGK1,NOTCH1,P2RY8,GNA13,IRF8,BTG2,EBF1,KLHL6,CCND1, BCL10, EP300, and ZFP36L1.

Approximately 25 ng of genomic DNA from each tumor was first sheared to 300 bp fragments using a Covaris M220. We constructed libraries from each using the KAPA LTP library preparation kit for Illumina platforms (KAPA Biosystems, Inc.). Specifically, DNA fragments were end-repaired and A-tailed in the presence of Agencourt XP magnetic beads (Beckman Coulter, Inc.) according to the manufacturer’s protocol before ligation to a 30-fold molar excess of Y-shaped adapters. The ligation reaction was cleaned up in a single step using a 0.8x volume of PEG/NaCl SPRI solution (KapaBiosystmes, Inc). Libraries were PCR enriched with indexed primersaccording to the manufacturer’s protocol during 7 amplification cycles. Enriched libraries were cleaned up using magnetic beads, quantified in Qubit (Life Technologies, Inc.) and analysed in an Agilent 2100 Bionalyzer instrument (Agilent Technologies, Inc.) for quality control. Between 8 and 11 libraries we pooled in roughly equimolar concentrations and enriched using a pool of 173 XGen Lockdown probes (for most samples)(Integrated DNA Technologies, Inc.) targeting the exons of the candidate genes or a pool of roughly 1,000 probes targeting an expanded set of lymphoma-related genes for a small number of samples in which the smaller pool yielded no mutations (see Table S5 for details about the pools of probes used). The XGen Lockdown protocol for DNA probe hybridization and target capture (version 2.1) was followed with a series of minor modifications. One of the blocking oligonucleotides was custom designed to efficiently work with these libraries. We also used a more diluted pool of probes than recommended by the manufacturer (0.72 pmol/µl, 2.88 pmol total) and the hybridization reaction was incubated during 4 hours or overnight. Hybridized targets were bound to either Dynabeads M-280 or M-270 streptavidin beads (Life Technologies, Inc.).

To detect mutations, we created an artificial reference in Geneiousver 7.1 (Biomatters, Ltd) that was generated by the concatenation of the targeted genes. These genes kept the original annotations from the reference from which they were extracted. We paid special attention to include in this reference any potential paralog or pseudogen to avoid misalignment issues. Local Blast searches for each gene were performed in Geneious to evaluate whether the exons of the targeted genes had one or multiple hits in the human genome.Illumina-gerenatedFastQ files from each sequencing primer were imported into Geneious and set as lists of paired reads. Reads were mapped against the reference in Geneiousallowing for a maximum of 6% mismatches and a maximum gap size of 100 nucleotides. PCR duplicates were removed using the MarkDuplicates tool from the Picard package, from which we installed a functional plugin in Geneious, and the resulting BAM files were screened for variants. We then searched for variations along the exons and splicing sites of the targeted genes in Geneious by setting the minimum coverage to 25 reads, minimum variant frequency to 0.05 and maximum P-value for the variant (i.e. probability to detect a variant by chance) to 10-9. From all the variants reported, those showing strand bias values higher than 90% were carefully inspected and those with average Phred quality scores below Q25 were discarded. Variants were annotated via the Oncotator server hosted at