1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
Supplementary Files
Genomic and epigenomic co-evolution in follicular lymphomas
Markus Loeffler1*, Markus Kreuz1*, Andrea Haake2*, Dirk Hasenclever1*,
HeikoTrautmann3*, Christian Arnold4, Karsten Winter5, Karoline Koch6, Wolfram
Klapper6, René Scholtysik7, Maciej Rosolowski1, Steve Hoffmann8, Ole Ammerpohl2,
Monika Szczepanowski6, Dietrich Herrmann3, Ralf Küppers7, Christiane Pott3, Reiner
Siebert2
on behalf of the Haematosys-Project
*
these authors contributed equally to this work
1Institute
for Medical Informatics Statistics and Epidemiology, University of Leipzig, Germany; 2Institute of Human
Genetics, Christian-Albrechts-University Kiel, Germany; 3Second Medical Department, University Hospital
Schleswig-Holstein, Campus Kiel, Kiel, Germany; 4Interdisciplinary Centre for Bioinformatics (IZBI), University of
Leipzig, Germany; 5Translational Centre for Regenerative Medicine (TRM-Leipzig); Germany; 6Hematopathology
Section, Christian-Albrechts-University, Kiel, Germany; 7Institute of Cell Biology (Cancer Research), Faculty of
Medicine, University of Duisburg-Essen, Essen, Germany; 8Transcriptome Bioinformatics, LIFE Research Center
for
Civilization
Diseases,
University
of
Leipzig,
Germany;
23
24
25
26
1. Materials
Supplementary Table 1a: Summary of patient characteristics
All
Number of patients
Number of samples
Number of pair-wise comparisons*
Sex
Diagnosis:
FLI/II
FLIIIa
FL NOS
Age at biopsy (median, range)
27
28
29
30
n=33
(25 pairs; 6 trios;
2 quadruples)
n=76
n=55
n=15 (45%) male
n=18 (55%) female
Core set
(Cases with IGHV-sequences)
n=19
(17 pairs; 2 trios)
n=40
n=23
n=11 (58%) male
n=8 (42%) female
n=58
n=2
n=16
59 [27-88]
n=33
n=1
n=6
54 [27-74]
Time between paired probes in months
(median, range)
IGHV sequencing
Number of samples measured
24 [0-101]**
29 [6-101]
n=40 (53%)
n=40 (100%)
Number of pair-wise comparisons*
n=23 (42%)
n=9 validated using NGS
n=23 (100%)
n=9 validated using NGS
Methylation analysis
Number of samples measured
n=76 (100%)
n=40 (100%)
Number of pair-wise comparisons*
n=55 (100%)
n=23 (100%)
NGS analysis
Number of samples measured
n=69 (91%)
n=40 (100%)
Number of pair-wise comparisons*
n=50 (91%)
n=23 (100%)
SNP 6.0 analysis
Number of samples measured
n=35 (46%)
n=16 (40%)
Number of pair-wise comparisons*
n=19 (35%)
n=9 (39%)
* Patients with 2 samples result in 1 pair-wise comparison, trios in 3 (primary vs. first relapse, primary
vs. second relapse and first- vs. second relapse ) and quadruples in 6 pair-wise comparisons.
** 7 pairs with time between samples less than 4 months were excluded from integrated correlation
analyses
(see
section
4F).
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
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
2. Methods
DNA extraction:
DNA extraction from tissue was done using the QIAamp DNA Mini Kit (Qiagen, Hilden,
Germany) according to the manufacturer´s protocol with minor modifications. For DNA
extraction, 15-20 sections à 20 µm of frozen tissue were used. The modifications are as
follows: For lysis, 360 µl ATL buffer and 40 µl proteinase K and for precipitation 400 µl AL
Buffer and 400 µl ethanol were used. Finally, the DNA was extracted using a total volume of
400 µl AE buffer (200 µl twice). DNA extraction from cells in DMSO was done using the
Gentra Puregene Blood Isolation Kit (Qiagen). According to the manufacturer’s instructions
elution was done in ddH2O. Quality control of the DNA was performed by agarose gel
electrophoresis and showed a discrete band visible at a size  20 kb. Quantification and
determination of purity (A260/280 > 1.8) was carried out using a Nanodrop photometer
(Thermo Scientific, Braunschweig, Germany).
Detection and sequencing of immunoglobulin gene rearrangements:
To identify each patient’s clonal immunoglobulin heavy chain (IGHV) gene rearrangement,
PCR amplification was performed according to the BIOMED-2 IGH Tube A protocol,1
including six consensus forward primers binding to framework region 1 (VH-FR1) in
combination with one consensus reverse primer for all JH-segments. For each reaction,
200 ng DNA from fresh-frozen lymph node specimens were used.
Clonal expansion and Sanger sequencing of clonal VDJ rearrangements:
Clonal IGH VH-JH PCR products from tumor samples were subcloned into pCR4-TOPO-TA
vectors (Life Technologies, Carlsbad, CA) according to the manufacturer´s instructions and
expanded in bacterial colonies. We picked and sequenced between 8 and 59 individual
colonies per tumor sample (median 36 colonies) via colony-PCR using M13 primers on a
3500 Genetic Analyzer (Life Technologies). Sanger sequencing was conducted with the
BigDye Terminator v1.1 Cycle Sequencing Kit (Life Technologies).
454 sequencing of clonal VDJ rearrangements:
To perform 454 sequencing analysis of the rearranged IGHV loci, barcoded amplicons were
prepared for NGS analysis by adding 5’ linker sequences to IGHV-FR1 gene segment family
primers and the consensus JH-primer from the original primer sets published by the
BIOMED-2 / EuroClonality consortium1. All amplifications were performed using a two-step
PCR in a total volume of 50 µl. The first round PCR using 200 ng genomic DNA, 2.5 U
FastStart High Fidelity polymerase (Roche) for 35 cycles was followed by a second
amplification step using a 1/500 dilution of the first round PCR product as a template. During
this second PCR step, adaptors including multiplex-identifier (MID) and sequencing adapter
sequences for emulsion-PCR and 454 sequencing were added to both ends of the
amplicons, applying universal-tailed fusion primers for bi-directional sequencing according to
the manufacturer´s protocol. Parallel pyrosequencing was performed on a GS-Junior (Roche
Diagnostics, Mannheim, Germany) following the manufacturer´s instructions. 1120 to 19311
(median 8600) reads per sample were evaluable. Base calls and quality scores were
extracted using the GS-Data-Analysis Software package (Version 2.5; Roche Diagnostics).
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
Sequencing of candidate genes:
Four different sequencing approaches were taken:
I) To investigate somatic mutations in CREBBP, TNFRSF14, TP53, CDKN2A, EP300, MLL2
and MEF2B, all coding exons of these genes were analyzed.
II) For the genes RHOH, PAX5, IRF4, CIITA, REL and PIM1, which are putative targets of
the SHM machinery, we analyzed the region 2.5 kb downstream of the transcription start
sites (TSS).
III) The genes BCL2, BCL6 und MYC are associated with somatic mutations in coding
regions as well as aberrant SHM, so that both regions of these three genes were
investigated.
IV)Finally, for detection of somatic mutations in EZH2 and MYD88, we analyzed ± 75 bp
around the known mutational hotspots (Tyr641 in EZH2 and L265P in MYD88).
In total, we sequenced 176 regions from 18 genes with high coverage spanning 90,164
bases. The regions analyzed are described in Suppl. Table 2. To achieve a coverage on
target of 1000-10000 reads the target regions were enriched using the RainDance
Technology (http://raindancetech.com/). RainDance amplification and next generation
sequencing was performed as custom service at Atlas Biolabs.
Supplementary Table 2: Candidate genes for mutation analysis
I) Potential driver mutations - all exons were analyzed
gene
No. exons
chromosome
start (Hg19)
end (Hg19)
CREBBP
31
16
3,775,053
3,930,123
TNFRSF14
7
1
2,487,802
2,495,269
TP53
10
17
7,571,717
7,590,865
CDKN2A
4
9
21,967,748
21,975,124
EP300
31
22
41,488,611
41,576,083
MLL2
54
12
49,412,755
49,449,109
MEF2B
13 E
19
19,256,373
19,303,402
II) Aberrant somatic hypermutation – 2.5 kb from transcription start site were analyzed
gene
region
chromosome
start (Hg19)
end (Hg19)
RHOH
approx. 2.5 kb
4
40,198,527
40,201,027
PAX5
approx. 2.5 kb
9
37,031,976
37,034,476
IRF4
approx. 2.5 kb
6
391,752
394,252
CIITA
approx. 2.5 kb
12
10,971,055
10,973,555
REL
approx. 2.5 kb
2
61,108,752
61,111,252
PIM1
approx. 2.5 kb
6
37,137,922
37,140,422
III) Potential driver mutations and aberrant somatic hypermutation
gene
region
chromosome
start (Hg19)
end (Hg19)
BCL6
approx. 2.5 kb
3
187,439,162
187,463,475
BCL2
approx. 2.5 kb
18
60,790,576
60,987,380
MYC
approx. 2.5 kb
8
128,747,765
128,750,815
IV) Known mutation position
gene
107
108
109
position
chromosome
start (Hg19)
end (Hg19)
MYD88
L265P
3
38,179,966
38,184,514
EZH2
Trr641
7
148,504,461
148,581,443
110
111
112
113
114
115
116
For validation of selected SNVs detected by next generation sequencing in CREBBP,
TNFRSF14, TP53, CDKN2A, EP300, MLL2 and MEF2B, Sanger sequencing using an ABI
Sequencer 3100 (Applied Biosystems) was performed using the primers presented in
Supplementary Table 3.
Supplementary Table 3: Primer sequences
Gene
analyzed
region
fwd-primer (Sequence 5'-3')
rev-primer (Sequence 5'-3')
Temp Amplicon Chrom
[°C]
length
(HG19)
Start
(HG19)
End
(HG19)
EZH2*
TYR-641
tttgtccccagtccattttc
tggcaattcatttccaatca
55
267 bp
7
148508598
148508873
TNFRSF14
Exon 1
TCCTCTGCTGGAGTTCATCC
CATGGGGAAGAGATCTGTGG
60
209 bp
1
2488044
2488252
TNFRSF14
Exon 2
ATCTCCCAATGCCTGTCCT
AGAAGGGGGCAAGAGTGTCT
60
202 bp
1
2489135
2489336
TNFRSF14
Exon 3
TAGCTGGTGTCTCCCTGCTT
GGCTGTGCTGGCCTCTTAC
60
250 bp
1
2489677
2489926
TNFRSF14
Exon 4
TCCACGTACCCCTCTCAGC
GAAATGGGAGGGGTGTCC
60
228 bp
1
2491224
2491451
TNFRSF14
Exon 6
CTCCCTGAGGCTGAGTGAAC
GGTGACAGAGCTCCAAGAGG
60
277 bp
1
2493043
2493319
TNFRSF14
Exon 8
AAAATGAACCCGAGAACCTG
AGGTGGACAGCCTCTTTCAG
60
267 bp
1
2494514
2494780
CREBBP
Exon 13
CATCCTCTGGGGTTGTGAAG
CATGAAATGTGCATTCTGGA
55
401 bp
16
3823635
3824033
CREBBP
Exon 14
TCCATTTCTGGTAGGGACAGGT
GC
GGCCCAAAAACAGCAGAGACAG
A
60
463 bp
16
3820539
3821001
CREBBP
Exon 15
TTGTAGGTTGCATGAGCAGC
CAGGGATACCCATGGCAG
55
356 bp
16
3819081
3819436
CREBBP
Exon 2223
GGACGCACACACAGACTTCTAC
AACCAAAGAACAATGGGGAC
60
621 bp
16
3794816
3795436
CREBBP
Exon 25
GGTGTGCAGAAGCACCTTG
GAAGGCTCACAGGCTCCTC
65
306 bp
16
3789484
3789789
CREBBP
Exon 26
aatgacagagcaagaccctg
TTAAAATACCCATTATTTCACGG
55
315 bp
16
3788474
3788788
CREBBP
Exon 27
TAACTCCTTAAAGGCAGGGC
AAAAGGCACACAAATATCCTCC
55
300 bp
16
3786584
3786883
CREBBP
Exon 28
CATGGGACTCTGCCACAC
GACACCACCACAGGAAGGAC
60
388 bp
16
3785931
3786318
CREBBP
Exon 29
TGACCTACTTTGGCCTGAGC
ACTTCCCTCCCACCACAGAC
65
377 bp
16
3781671
3782047
CREBBP
Exon 30
CTATTCTGCAGGCTGGGTG
AAAGGGACAGGATGCTTCG
60
442 bp
16
3781127
3781568
CREBBP
Exon 31
CCTGTACCGGGTGAACATCAAC
GCTGCCTCCGTAACATTTCTCG
60
677 bp
16
3778459
3779135
CREBBP
Exon 31
CCAAGTACGTGGCCAATCAG
ACCGCACCTGGTTACTAAGG
65
717 bp
16
3778015
3778731
TP53
Exon 5-6
TAGTGGGTTGCAGGAGGTG
tcaaataagCAGCAGGAGAAAG
65
594 bp
17
7578076
7578669
TP53
Exon 12
TGGGGTAAGGGAAGATTACG
TTCTGACGCACACCTATTGC
58
399 bp
17
7572815
7573213
CDKN2A
Exon 1
AGTTAAGGGGGCAGGAGTG
GGCTCCTCAGTAGCATCAGC
60
246 bp
9
21994174
21994419
EP300
Exon 4
gaaatagcacattatgactcctacca
tccctggctgtaaaaattgc
60
363 bp
22
41523440
41523802
EP300
Exon 14
ttctgttctgaattgctgtcttg
atggaaatggcccagaagta
55
558 bp
22
41545721
41546278
EP300
Exon 17
tggtaactaatttcaaatgcacttttt
tggctatactgtttggaatgtga
60
243 bp
22
41550963
41551205
EP300
Exon 26
gaactcattatgtgacctgacttttt
tgttacgtaagaactaaaatgaggaaa
60
295 bp
22
41565449
41565743
EP300
Exon 27
caacttgtggtttaaaatgtagcc
ccagatctattgtcagcacctg
65
285 bp
22
41566333
41566617
MLL2
Exon 3
gcgtggtactgatgcttgtg
cagcccttatcccatttcct
60
293 bp
12
49448271
49448563
MLL2
Exon 5
ggctgacactgaggctcttt
tctcatttgccctatgacca
60
235 bp
12
49447723
49447957
MLL2
Exon 6
gcaatgtgctgaggcttaca
tcctgcccttccattcctac
60
247 bp
12
49447239
49447485
MLL2
Exon 10
aggagcatcgtgttgttgtg
GGAGACAGGCGAGATGCT
65
490 bp
12
49445745
49446234
MLL2
Exon 10
CCGCCACCTGAGGAATTG
GTGGGGAAGCAGGTGAGTC
63
463 bp
12
49445338
49445800
MLL2
Exon 10
GTGTCACGCCTGTCTCCAC
GCATAGGCATGGCTCCTC
63
366 bp
12
49445126
49445491
MLL2
Exon 10
TGAGGAGCCGCAACTCTG
CTCCTCAGGGGGCTTTTC
55
424 bp
12
49444856
49445279
MLL2
Exon 11
GGGGACAGTGACCCTGAGT
CCCCCACTACCTTCCCTATG
65
298 bp
12
49444181
49444478
MLL2
Exon 14
tgactctggtcgcaaatcag
attccccagcctacacctct
65
242 bp
12
49441712
49441953
MLL2
Exon 23
ctccttgactgccccaca
ccatcaaataacttgccagctc
65
243 bp
12
49437342
49437584
MLL2
Exon 27
acaggtgggagtggtctgaa
cagatggagggaaaggacaa
65
232 bp
12
49436287
49436518
MLL2
Exon 29
gcctgccaagtcttctctga
cagttcccacgctaatccat
65
152 bp
12
49435663
49435814
MLL2
Exon 31
GTTACCCCTCGCTTCCAGTC
GCCCAAAATGGCTGTTGAT
60
385 bp
12
49433851
49434235
MLL2
Exon 31
TTCACTTTCCCTCAGGCAGT
ggagcgatatagggggctta
60
481 bp
12
49433467
49433947
MLL2
Exon 32
tgggcttattcctcttctctttt
ccactatcccttgccactct
60
242 bp
12
49433192
49433433
MLL2
Exon 33
gggccaggatattgaaggtt
atccatcccccttggtttac
60
234 bp
12
49432959
49433192
MLL2
Exon 34
ttccagGCAACTGGTAGGAG
GTGGGGTGTTGGATGAAGAC
65
493 bp
12
49432286
49432778
MLL2
Exon 34
GCTGCTGATGCCTCTGAAC
CTGAAAGCTGCTGCTTCTTCT
65
496 bp
12
49431337
49431832
Gene
analyzed
region
fwd-primer (Sequence 5'-3')
rev-primer (Sequence 5'-3')
Temp Amplicon Chrom
[°C]
length
(HG19)
Start
(HG19)
End
(HG19)
MLL2
Exon 34
GCATCTGGGGATGAGCTAGA
tggctatgttaccagctgagg
65
575 bp
12
49430884
49431457
MLL2
Exon 35
cgcagatattcactggagca
gggtgtgactgggaaagaaa
58
237 bp
12
49428543
49428779
MLL2
Exon 38
tcctgacacccagcttcttt
tctgggtgctaggctgaagt
60
293 bp
12
49427816
49428108
MLL2
Exon 39
GCACACTAATCTCATGGCAGA
GGATTGCCACCTGTCCTAGA
65
500 bp
12
49427228
49427728
MLL2
Exon 39
GAAGCCTCGGACCTGATTC
CCTTGCTGTTGGTGCTGTT
65
484 bp
12
49426885
49427369
MLL2
Exon 39
AGGGCCTTATGGGACACAG
GGCCCATCTGCTGCTGTT
63
396 bp
12
49426559
49426955
MLL2
Exon 39
TCTCCTCAGCAACAACAGCA
AGGCTGATCCCCTAAGGAAA
65
480 bp
12
49426053
49426532
MLL2
Exon 39
GCAGCTAGGCAGTGGATCAT
GTGGGGTCTGGCGTACTG
65
374 bp
12
49425764
49426137
MLL2
Exon 39
AAGGAGTCCTGGCCAAAAAC
GCAGCAGCAGGTGAGACC
60
484 bp
12
49425400
49425883
MLL2
Exon 39
ACCTCAGGGGCCAACCTT
GTTCCTGGTGCCCCTATTG
65
300 bp
12
49425154
49425453
MLL2
Exon 40
ggctctgaggaggagggtag
ctatcctgggatgggaccag
60
233 bp
12
49424632
49424864
MLL2
Exon 48
tacagggcaccctcctacag
ATGTCTCGCGGTACCTTGTC
60
463 bp
12
49420663
49421125
MLL2
Exon 48
CCTTGCGACCTGACAAGGTA
ACAGGGCCCCTTGATCTTAT
60
371 bp
12
49420323
49420693
MLL2
Exon 50
ctttggcctaaccccaaaaa
gaccagaggatccctgtcaa
60
249 bp
12
49418299
49418547
MLL2
Exon 51
cagaggaggtgggtggtatg
gccagctcatacCTGCTCTT
60
368 bp
12
49416361
49416728
MLL2
Exon 5253
agaagggaaaggcaggagaa
aggaggaggagctgctttgt
55
491 bp
12
49415780
49416270
MLL2
Exon 54
gcattgattctgccctcttc
CAATGGCTGCTTCTGTCTGG
60
390 bp
12
49415295
49415684
MEF2B
Exon 5
ggcagacagaggagaggtgt
tcaggtcagtcccttgccta
60
246 bp
19
19261413
19261658
MEF2B
Exon 6
acaccaccccacattcatct
taaagcacgtcagccacaaa
55
389 bp
19
19259911
19260299
MEF2B
Exon 10
gggtgtgggcctcagttt
taaccacccccagtgacagt
55
248 bp
19
19257252
19257499
MEF2B
Exon 11
gaaggcttaaggagatgtccag
gtgcgcagtaccagggatg
60
249 bp
19
19256995
19257243
CREBBP
Exon 31
CACAGCAGCCCAGCACAC
TTGTTGATGTTCACCCGGTA
60
256 bp
16
3779112
3779367
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
* described in (Pellissery et al, 2010)2
Temp indicates the annealing temperature in the PCR.
SNP array analysis:
SNP array experiments were performed according to the standard protocol for Affymetrix
GeneChip SNP 6.0 arrays (Affymetrix). Briefly, a 500 ng sample of DNA was digested with
StyI and NspI, ligated to adaptors, amplified by PCR, fragmented with DNAse I, and biotinlabeled. The labeled samples were hybridized to Affymetrix GeneChip SNP 6.0 arrays,
followed by washing, staining and scanning. The complete dataset comprised 35 FL samples
in total (16 cases included within the core set) and 33 lab-specific euploid samples (17
females and 16 males) for controls.
DNA methylation analysis:
Bisulfite conversion of the DNA was performed using the “Zymo EZ DNA methylation Kit”
(Zymo Research, Orange, CA) according to the manufacturer´s instructions with the
modification described in the Infinium Assay Methylation Protocol Guide (Illumina, San
Diego, CA). All further analysis steps were performed according to the “Infinium II Assay Lab
Setup and Procedures” and the “Infinium Assay Methylation Protocol Guide”. The processed
DNA samples were hybridized to the HumanMethylation 27 BeadChips (Illumina, San Diego,
CA). This array was developed to assay 27,578 CpG sites selected from more than 14,000
genes. Raw hybridization signals were processed using Bead Studio software (version
3.1.3.0, Illumina) applying the default settings.
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
3. Bioinformatic and statistical analyses
Detecting selection by mutation analysis in the IGHV region sequences
The objective of the mutation analysis was to compare the ratio of the observed number of
replacement mutations and the observed number of silent mutations with their expected
ratio, assuming no selection for both the structural regions of the heavy chain known as
„framework regions“ (FWR) and the „complementarity determining regions“ (CDR).
For each tumour sample the analysis consisted of the following steps:
(1) All tumour IGHV sequences were aligned with their most likely germline sequences using
the IMGT/HighV-QUEST online tool.3
(2) The number of different replacement (R) and silent (S) mutations in the set of clonally
related sequences was determined for FWR and CDR resulting in counts Rfwr, Sfwr, Rcdr,
Scdr.
(3) A model4,5 for SHM assuming no selection, accounting for micro-sequence specificity of
SHM targets and transition bias was used to determine expected counts given the total
number of observed mutations resulting in numbers eRfwr, eSfwr, eRcdr, eScdr. Step (2)
and (3) was performed using the web server http://clip.med.yale.edu/selection.4
(4) The web tool provides p-values on the null-hypothesis of no selection (separately for
FWR and CDR) using the so called focused binomial test. P-values do not lend itself to
meaningful meta-analyses. In addition, we wanted quantitative comparisons of strength of
selection within tumour pairs from the same patient. Therefore we defined the
logRSoddsratio=log( (R/S) / (eR/eS) ) as a quantitative measure of selection strength
(compare Yaari et al 20126). This quantity compares the observed R/S ratio to the one
expected under the null-hypothesis of no selection. The logarithm transforms the measure to
its natural scale such that the estimates are approximately normally distributed.
(5) The logRSoddsratio can be estimated using the numbers from step (2) and (3). In line
with the ‘focused binomial test’ outlined in Uduman et al, 20114 and Hershberg et al, 20085
we gain power assuming that silent mutations are neutral concerning selection and thus
Sfwr/Scdr = eSfwr/eScdr (we assume the mutation model that generates the expectations).
Under this assumption logRSoddsratio=log( (R/S) / (eR/eS) )= log( (R/(Sfwr + Scdr)) /
(eR/(eSfwr + eScdr)) ). 95% confidence intervals can be obtained sampling from the
posterior distribution Dirichlet(eRfwr/E+Rfwr, eRcdr/E+Rcdr, (eSfwr + eScdr)/E + (Sfwr +
Scdr)) [with E= eRfwr + eSfwr + eRcdr + eScdr]. P-values dual to these CIs are in good
concordance with the p-values of the focused binomial test.
(6) Standard methods7 for fixed and random effect meta-analyses and forest-plots are used
to analyse logRSoddsratios across samples.
We further wanted to distinguish evolution in times before tumor onset and after tumor
initiation. Therefore, three computations were performed for each sample, each time using a
different rooting sequence. Supplementary Figure 1 illustrates the three types of reference
sequences:
1. The first rooting sequence was a consensus germline sequence constructed from all
germline sequences assigned to the tumour sequences of the patient using the
IMGT/V-QUEST online tool. Bases which differed among these germline sequences
were substituted by „N“ (any base) in the consensus germline sequence.
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
2. The second rooting sequence consisted of bases common to all sequences from the
primary tumour and the relapse tumour of the same patient. At positions where we
observed different bases among these tumour sequences we inserted the
corresponding base from the consensus germline sequence (the first reference
sequence) into this common rooting sequence.
3. The third rooting sequence was constructed from bases common to the sequences
found in each single tumour sample. Positions which varied among these sequences
were filled with the corresponding base from the germline consensus sequence.
Supplementary Figure 1. An illustration of the chronological position of the three types of
rooting sequences used for detecting selection in each sample.
Using three different rooting sequences allowed us to investigate how strongly selection
acted on the observed sequences
a) at any time since the VHDHJH recombination of the B-cell from which the primary and the
relapse tumours originated (evaluation with respect to the germline rooting sequence),
b) since the time of the last common somatic mutation in the precursor of the primary tumour
and the relapse (evaluation with respect to the common tumor rooting sequence), and
c) since the last somatic mutation which was common to the sequences of the investigated
tumour sample (evaluation with respect to the tumor specific rooting sequence), respectively.
NGS candidate genes
Sequence data of all 69 samples was mapped to hg19 using the segemehl algorithm8 with
default parameters. Samtools mpileup version 0.1.189 was applied to each sample with
parameter “–d” set to 25000, thus allowing a maximum coverage of 25000 reads per
position. For further analysis positions with effective enrichment were selected. Therefore all
positions within enriched genomic regions (see Supplementary Table 2) with coverage >1000
reads and coverage of high quality (HQ) reads >500 (Phred quality score Q≥13) were
selected for further investigation. Median HQ-base coverage over all target positions and
lymphoma samples was 5343 bases (range: 4149-6653). Over all analyzed samples >99.3%
of the enriched genomic positions showed both a coverage >1000 and a HQ coverage >500.
Prior to variant calling for each position the number of reference and alternative alleles were
summarized for forward and backward strand separately. This was repeated for high quality
bases (Q≥13).
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
For the variant calling, the proportion of the most frequent alternative allele was analyzed for
each position in each lymphoma sample. A variant was called if ≥10% of all HQ-bases
showed a concordant alternative allele.
To achieve a high specificity and avoid false positive variant calls, additional quality filters
were applied. Variants showing a proportion of low quality reads of >40% were rejected. In
addition variants with a high allelic imbalance between forward and reverse strand for
reference and mutant alleles were removed. Therefore for each variant the |logOR| of the
number of reference alleles and mutated alleles, each for forward and reverse strand was
calculated. To avoid zeros 0.5 was added to each count. Variants with a |logOR|≥5 were
removed from further analysis.
After filtering, all variants were annotated with dbSNP build 135 for overlap with known single
nucleotide polymorphisms. All variants overlapping positions with known SNPs were
excluded from further analysis. In addition, functional annotation was added to all variants
using vcfCodingSnps version 1.5 (http://genome.sph.umich.edu/wiki/VcfCodingSnps).
a
b
Supplementary Figure 2. A) shows the histogram of the differences of the allele frequencies
for detected mutations on the logit scale. The histogram indicates 3 peaks representing
mutations present in both samples or either in primary (PT) or relapse tumor (RT)
exclusively. A threshold of -2 and 2 (indicated by red vertical lines) is applied to distinguish
these 3 groups. B) displays the allele frequency of mutated alleles for paired primary and
relapse tumors. Red dots indicate mutations selected as differential between primary and
relapse sample using the thresholds described in A).
To compare mutations between paired samples of the same patient, for each variant
identified (see Supplementary Table 5) the frequency of the mutant allele was compared. If a
variant was called only in one sample the allele frequency of the second sample was
calculated from the raw data. When comparing the differences of the allele frequencies on
the logit scale for all non-SNP positions 3 groups appear (see tri-modal histogram in
Supplementary Figure 2a). Using a threshold of 2 respectively -2 allows distinguishing
concordant variants from variants present in only one of the lymphoma samples (see
Supplementary Figure 2b).
The proportion of discordant variants within pairs was determined for candidate genes and
2.5 kb downstream of TSS regions (non-IG SHM targets) separately and used as a summary
measure of divergence. A schematic overview over the analysis pipeline for the sequencing
data is shown in Supplementary Figure 3.
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
Supplementary Figure 3. A schematic overview over the analysis pipeline for the NGS data.
(HQ=high quality)
Methylome data analysis:
“HumanMethylation27 DNA Analysis BeadChip” array data of 76 FL tumor samples were
analyzed in combination with 8 control samples previously measured on the beta-test version
of the same array (described in ref. 10). For the analysis, only CpG positions represented on
both beta and final version of the array were included (n=27568). The raw fluorescent signals
of methylated and unmethylated alleles were normalized across chips using the vsn
method.11 The anti-log of normalized signals was used to calculate beta values as
described.10 CpGs were classified into “hypermethylated in FL”, “hypomethylated in FL”,
“methylated in FL and controls”, “unmethylated in FL and controls” and remnant as outlined.10
To compare methylation between lymphoma samples of the same patient, CpGs were
selected as differentially methylated if the difference of the methylation level between paired
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
samples was >0.25. The proportion of discrepant CpGs within a pair is used as a summary
measure of methylation divergence.
Analysis of chromosomal alterations based on SNP arrays:
Genotyping and copy number analysis was performed using Affymetrix Genotyping-Console
version 4.1.1. Copy number segmentation was calculated using the R package DNAcopy.12
Parameters for prior data smoothing were adapted to data quality. Significant copy number
aberrations were detected using an adapted implementation of histogram entropy
minimization,13 whereas for data sets with bad data quality subsequent manual correction
was necessary. Comparison of paired samples was performed on differences of copy
number profiles from primary and relapse tumor, and carried out in the same manner.
Results of segmented differences from paired samples were depicted for visual evaluation
considering the individual profiles of primary and relapse tumors to distinguish qualitative and
quantitative differences.
A Hidden Markov Model based method14 implemented in dChip (64bit, build date: Apr 13
2010)15,16 was used to infer LOH from tumor samples. The LOH call threshold was set to the
default value of 0.5. Comparison of paired samples was performed on differences of LOH
profiles from primary and relapse tumor. Differences were determined by detecting changes
from heterozygous in primary tumor to homozygous in relapse tumor (AB -> AA | BB) and
vice versa (AA | BB -> AB). Positions with LOH were coded with "1", positions without LOH
were coded with "0" and positions without information were coded with "NA", resulting in
binary profiles. Segmentation of LOH profile differences was calculated using DNAcopy and
its implementation of the circular binary segmentation algorithm for binary data. The
segmentation parameter alpha was adapted manually (ranging from 0.4 to 0.99) to account
for differences in data quality.
Integrative correlation analysis:
The integrated analysis correlated several quantitative measures of divergence between
primary and secondary samples. We address the question whether and to which extent the
biological processes generating divergence in various genomic and epigenomic dimensions
correlated.
A measure of divergence in a specific type of data is designed to capture the overall degree
by which paired samples differ. Measures of divergence are chosen symmetrically in the
samples with the aim to quantify overall drift on a common scale across pairs with small (but
not necessarily zero) values if the samples are close/undistinguishable and large values if
the samples have drifted apart and appear clearly distinct. Measures are scale transformed
such that the distribution of divergences across pairs is approximately normal.
IGHV sequences: IGHV_divergence is calculated as the average Hamming distance
between the observed sequences from the primary and the secondary sample (normalized to
a common sequence length of 100). The Hamming distance between two aligned sequences
is the number of positions in which they differ. The average Hamming distance is log
transformed to achieve an approximately normal distribution. This measure can be calculated
both for clone sequences and NGS results yielding nearly identical results in 9 cases with
both measurements available.
SHM of non-IG genes: For each tumor pair, we observe N mutations found in at least one of
the samples. k out of them will be discordant (i.e. found only in one but not the other sample).
Aberrant_SHM_divergence is defined as the proportion of mutations found which are
discordant. Since we deal with varying denominators some may be small and we want to
avoid zeros, we use the non-informative Bayesian estimate (k+0.5)/(N+1). The proportion is
logit transformed to achieve an approximately normal distribution across pairs on the real
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
line. Besides overall aberrant_SHM_divergence we also consider sub-divergences in BCL2
sites only and transcription sites outside BCL2 only.
Mutations in candidate genes: Candidate_Gene_Mutation_divergence
analogously to aberrant_SHM_divergence.
is
defined
Differentially methylated CpG islands: DNA_Methylation_divergence is calculated as the
proportion of CpG islands in which we observe a difference in methylation degree of >25%
between the pairs. The cut-value at 25% has been used before9 and cut-points between 20%
and 35% do not change results. The proportions are logit transformed to achieve an
approximately normal distribution across pairs on the real line.
SNP chip data: Chromosomal_divergence is defined as the proportion of SNPs
investigated which differed between primary and secondary sample in copy number or LOH
status. The proportions are logit transformed to achieve an approximately normal distribution
across pairs.
Time difference: The number of days elapsed between obtaining the samples is log10
transformed to achieve an approximately normal distribution across pairs.
Correlation methods: For pairs of measures of divergence Pearson linear correlation
coefficients were estimated with 95% BCa-bootstrap confidence intervals.17 Correlations
were
considered
“significant”
if
the
confidence
interval
excluded
zero.
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
4. Supplementary Results
A. Detecting selection by mutation analysis in the IGHV region sequences
As described in section 3E, the observed ratio of the number of replacement and silent
mutations was compared with the expected ratio, separately for „framework regions“ (FWR)
and „complementarity determining regions“ (CDR). To distinguish selection in times before
tumor onset and after tumor initiation, three computations using different rooting sequences
were performed (see Suppl. Figure 4 and Suppl. Table 4a and 4b).
a
b
c
d
e
f
Supplementary Figure 4. LogRSoddsratios quantifying selection together with their highest
credibility intervals. Each interval corresponds to one sample. (a, b, c) Results for the FWRs
using the reference sequences 1, 2 and 3, respectively, (d, e, f) Results for the CDRs using
the reference sequences 1, 2, and 3, respectively. The result of a meta-analysis is shown in
blue at the bottom of each plot.
Supplementary Table 4a. Results of a meta-analysis of selection in the IGHV regions.
Shown are the estimates of the logRSoddsratios, their confidence intervals, values of the Q
statistic testing for heterogeneity of the effects and the corresponding P-values for
heterogeneity (see also Figure 2).
Germline rooting sequence
estimate (95% CI)
397
398
399
400
401
402
Common tumor rooting sequence
Q
P
estimate (95% CI)
Q
P
Single tumor rooting sequence
estimate (95% CI)
Q
P
FWR
-0.77
(-0.88; -0.67)
43
0.229
-0.89 (-1.02; -0.75)
36 0.517
-0.91 (-1.06;-0.76)
41.3 0.290
CDR
-0.32 (-0.45; -0.19)
46
0.147
-0.58 (-0.76; -0.39)
55.4 0.026
-0.65 (-0.87; -0.43)
43.9 0.202
Supplementary Table 4a provides the summary statistics and Supplementary Table 4b
(external Excel file) the results for each sample. The following observations can be made
regarding the FWRs:
1. The FWRs are clearly preserved (selection against replacement mutations).
2. There is no heterogeneity in the logRSoddsratios between samples (Q-test).
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
3. These results are valid independent of the choice of the rooting sequence.
4. There are fewer numbers of mutations with respect to later rooting sequences and
thus the corresponding confidence intervals are wider.
Hence we conclude that FL depends on a functional BCR.
Regarding to the CDRs, we observe that:
1. The estimates of the logRSodds ratios show higher variability due to smaller number
of mutations than in the FWRs.
2. Similar to the FWRs, there is no evident heterogeneity in logRSoddsratios between
samples (Q-test).
3. Global meta-analytic estimate shows overall significant preservation but
a. The effect for CDR is quantitatively (significantly) less pronounced than that for
the FWRs.
b. The estimates tend to become stronger using later rooting sequences
Hence, these results are consistent with ongoing dependence of FL on affinity against a
tumor specific but unknown antigen.
Taken together, the data strongly suggest that the malignant clone is affected by ongoing
SHM during tumor evolution. The data also indicate that in most cases the SHM process is
not only working before the primary tumor is detected but also during the period until relapse.
B. IGH mutation analysis – comparison of VH clone sequencing and NGSsequencing
Visual inspection of the phylogenetic trees (see section 3D) resulted in 8, 5, 6 and 4 pairs of
samples classified to the “no evolution”, “sequential evolution”, “divergent evolution” and
“complex evolution” categories, respectively. Examples of trees from each category are
shown in Figure 1 of the manuscript.
We also quantified the evolutionary divergences between the primary and the matching
relapse specimen using a measure IGHV_divergence which we developed for this purpose
(section 3D). We observed that the IGHV_divergence values were small in the “no evolution”
category, but progressively increased in the “sequential evolution” and the “divergent
evolution” categories (Supplementary Figure 5). This indicated that the measure reflected the
biologic intuition expressed in the visual classification of the phylogenetic trees while being
more objective and quantitative. Using this, we were able to correlate the data from the
sequencing with other biological and clinical parameters.
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
Supplementary Figure 5. The IGHV_divergence values conforms well with the diversity
seen in pedigrees regarding “no evolution”, “sequential evolution” and “divergent evolution”
categories (points jittered to avoid overlaps).
To validate the results obtained from Sanger sequencing, we performed NGS of the IGHV
rearrangements in a subset of our cases (paired measurements from 9 patients). The
measures of divergence between the primary and the relapse samples using the Sanger and
the NGS data were highly correlated (Supplementary Figure 6).
This measure can be calculated both for clone sequences and NGS results yielding nearly
identical results in N=9 cases with both measurements available.
Supplementary Figure 6. Correlation between the measures of divergence of the IGHV
sequences of the primary and the corresponding relapse tumors measured by Sanger
sequencing (x-axis) and NGS (y-axis).
We further asked whether the evolutionary divergence of the paired samples of tumors could
be explained by the time elapsed between the two biopsies. However, we found no
correlation (Figure 3e of the manuscript).
C. Methylation analysis
Classification of CpGs in “hypermethylated in FL”, “hypomethylated in FL”, “methylated in FL
and controls”, “unmethylated in FL and controls” showed a high concordance to previously
analyzed DLBCL.10 Of 26604 CpGs analyzed in both datasets, 66.58% showed a concordant
classification. The majority of discordantly classified CpGs (98.85%) were classified as
remnant in either FL or DLBCL. As previously reported10,18 hypermethylated CpGs were
highly enriched among known polycomb target genes. Hypermethylated CpGs were enriched
for loci repressed by PcG marks in embryonic stem cells.19 663/1458 (45.5%) of all
hypermethylated CpGs were polycomb target genes compared to 2247/21826 (9.3%) in all
other groups (OR=8.1/p<0.001). Similar results were observed for embryonic fibroblasts 20
with 785/1560 (50.3%) polycomb targets among hypermethylated CpGs compared
4935/25679 (19.2%) in all other classes (OR=4.3/p<0.001).
Unsupervised cluster analysis of CpG array data for all samples showed clustering of the
vast majority of paired samples from the same patient rather than separation of initial and
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
relapse samples (Supplementary Figure 7). This indicates that each tumor has its unique
DNA methylation profile which is largely conserved over evolution, although permitting for
minor variations as seen below.
Supplementary Figure 7. Correlation heatmap of the methylation dataset. All samples are
ordered according to the patients. Patients are indicated by color bars for rows and columns,
i.e. neighbored samples marked by the same color belong to the same patient. Each point in
the heatmap represents the correlation (R²) of 2 samples. For the majority of patients all
related samples show high correlation indicated by light green squares along the diagonal.
The average degree of methylation was higher in relapses compared to primary samples. On
average the ratio “increased methylation in relapse / increased methylation in primary
sample” was 1.43 (Supplementary Figure 8).
Supplementary Figure 8. Histogram of pair-wise differences of the proportion of
hypermethylated CpGs in the relapse sample – hypermethylated CpGs in the primary tumor
on the logit scale. On average the CpGs showed slightly higher methylation levels for the
relapse samples. Included were all pairs from the complete dataset with exclusion of pairs
with time between biopsies <4 months. It is, however, also noticeable that both gains and
losses of methylation are found simultaneously.
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
D. Analysis of chromosomal alterations based on SNP arrays
The average call rate for all tumor samples was 98.42% [96.39%–99.19%] and 98.33%
[96.39%–99.19%] for the core set. Genotypes of paired samples showed a concordance of
98.99% [94.63%–99.93%]. In contrast, genotype concordance of randomly paired samples
from different patients was typically about 61.7%, thus confirming that all paired samples
originate from the same patient.
Supplementary Figure 9 shows the profile of copy number aberrations and uniparental
disomies (UPD) for the analyzed dataset. Most notable gains were detected at chromosomes
1q, 2p, 5p, 7, 11, 12, 16q, and 18 while notable losses can be seen at chromosomes 1p, 6q,
7p, 17p, 19 and 22 concordant with previous copy number analyzes by arrayCGH.21
Supplementary Figure 9. Copy number and UPD profile of the analyzed SNP-6.0 dataset.
The proportion of gains (green), losses (red) and UPD (cyan) is displayed in genomic order.
Supplementary Figure 10 shows the changes of structural aberrations between primary and
relapse tumors in 19 tumor pairs. For each analyzed pair changes in copy number or UPD
are shown in genomic order. Green color indicates an aberration only present in the primary
tumor, red indicates an aberration in the relapse that is lacking in the primary tumor.
Sup
plementary Figure 10. Heatmap of discrepant copy number / LOH regions. Each column
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
shows discrepant copy number / LOH regions within a sample pair of the same patient.
Green regions are only aberrant in the primary tumor, regions shown in red are exclusively
aberrant in the relapse sample.
E. Mutation analysis of coding genes and aSHM target regions
Variant calling in potential lymphoma driver genes and aSHM target regions for all 69
lymphoma samples resulted in 3043 variants (see Supplementary Table 5, external Excel
file) of which 1134 showed no overlap with known SNP positions. Of these 1134, 954 (84%)
were up to 2.5 kB downstream of TSS (non-IG SHM targets) and 180 (16%) within the
candidate genes. This can be taken as a strong indication of genetic damage inflicted by an
active somatic hypermutation machinery. Within the TSS regions, 28/954 (3%) mutations
were annotated as non-synonymous, affecting splice sites or affecting/generating stop
codons. For the candidate gene regions, 139/180 (77%) were annotated as nonsynonymous, affecting splice sites or affecting/generating stop codons. 219 mutations were
validated using Sanger sequencing. In 215 (98%) validations, the mutations were confirmed,
2 showed wild-type by Sanger sequencing and 2 were n.a. (see Supplementary Table 5). In
addition, 118 mutations rejected by our quality filters were validated by Sanger sequencing,
resulting in 26 (22%) confirmed mutations, 85 (72%) were disproved and 7 n.a. These results
illustrate a high specificity for the applied variant calling, but indicate on the other hand a
reduction in the sensitivity due to the quality filtering.
In a second step the mutations were compared within paired samples. To this end the allele
frequency of each mutation (SNP positions excluded) that was present in one of the paired
samples was analyzed in the remaining sample. The allele frequency difference was
analyzed as outlined in section 3F and each mutation was classified as present in both
samples or as discrepant (i.e. only present in one of the samples). Within 5' regions of genes,
379/904 (41.9%) mutations were discrepant between paired samples, mutations in coding
regions were discrepant in 32/145 (22.1%) compared variants (see Supplementary Table 6,
external Excel file). Mutations not detected in single case analysis but detected as present in
both samples within the paired analysis were considered as mutated in both samples for
further analyses. Thus the number of mutations in non-IG SHM target regions increased from
954 to 1059 and from 180 to 197 for the candidate genes.
Distribution of mutations in candidate genes and TSS regions are summed up in
Supplementary
Tables
7a
and
7b.
560
561
562
563
564
565
566
567
568
569
Supplementary Table 7a. Mutations in candidate genes. The table shows the number of
mutations and the number of protein-changing mutations for each analyzed gene locus. In
addition the number and percentage of samples affected by at least 1 protein-changing
mutation is shown. As indication for mutations occurring early in tumor development the
allele frequency of the mutated alleles was analyzed. For each locus the mean allele
frequency of detected mutations is displayed (in case of multiple mutations in the same gene
for single samples only the mutation with the highest allele frequency was included). In the
last column the number of patients affected by at least 1 single discrepant mutation in the
related gene locus is shown.
0.28
0.37
0.45
0.22
0.35
0.41
0.25
0
7
4
2
1
0
1
0
0
2 / 69
2.90
0.31
1
0
0 / 69
0.00
-
0
0
0 / 69
0.00
-
0
BCL2
MLL2
CREBBP
TNFRSF14
EZH2
EP300
MEF2B
BCL6
MYC
10
81
55
19
15
8
3
2
2
0
54
52
15
15
4
1
0
2
0 / 69
39 / 69
43 / 69
15 / 69
11 / 69
4 / 69
1 / 69
0 / 69
2 / 69
TP53
2
2
CDKN2A
0
MYD88
0
% samples
affected by
proteinchanging
mutations
Supplementary Table 7b. Mutations in a 2.5 kB region downstream of TSS (non-IG SHM
targets). The table shows the number of mutations for each locus analyzed. In addition, the
number and percentage of samples affected by at least 1 mutation is shown.
Region
CIITA_SHM
BCL2_SHM
RHOH_SHM
BCL6_SHM
PAX5_SHM
REL1_SHM
IRF4_SHM
MYC_SHM
PIM1_SHM
574
575
576
577
578
579
580
581
582
0.00
56.52
62.32
21.74
15.94
5.80
1.45
0.00
2.90
Number of
mutations
Number of
proteinchanging
mutations
Gene
570
571
572
573
mean allele
frequency of
mutations
Number of
patients
with ≥1
discordant
proteinchanging
mutations
Number of
samples
affected by
proteinchanging
mutations
Number of
mutations
Number of
samples affected
% samples affected
41
718
104
144
17
5
14
11
24 / 69
68 / 69
54 / 69
37 / 69
7 / 69
5 / 69
9 / 69
11 / 69
34.78
98.55
78.26
53.62
10.14
7.25
13.04
15.94
5
4 / 69
5.80
The majority of the protein-changing mutations within the exons of candidate genes affect
CREBBP (52 mutations in 43/69 (62%) samples) and MLL2 (54 mutations in 39/69 (57%)
samples). This indicated a strong mutation effect on genes regulating histone modifications
and transcriptional control.
An overview over the mutations affecting the non-IG SHM targets is shown in Supplementary
Table 7b. The majority of the mutations affect the BCL2 locus. This is in line with the fact that
the BCL2 locus is translocated to the IGH region by t(14;18) translocation and therefore
frequently targeted by the SHM machinery.
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
The characteristic WRCY/RGYW motif was highly enriched among mutations in the non-IG
SHM targets. When considering each mutation only once per patient, 104/618 mutations
were overlapping the C/G base of the motif. Considering the genomic sequence of the
analyzed regions, the number of expected overlaps is 35/618 (OR=3.4; binomial test
p<0.001). In comparison, no enrichment for mutations overlapping C/G of the WRCY/RGYW
motif could be shown for the mutations affecting the candidate genes. Only 4/100 mutations
(again each mutation was only considered once per patient) overlapped the C/G base.
Considering the genomic sequence, one would expect 7/100 overlapping mutations
(OR=0.55; binomial test p=0.33).
We
compared
mutation
status
of
the
analyzed
candidate
genes
and
DNA_Methylation_divergence as well as Aberrant_SHM_divergence (see Figure 4). Pairs in
which at least one of the samples was affected by a CREBBP mutation were compared to
unaffected pairs and showed significantly higher median DNA-methylation-divergence
(p=0.008) and aSHM-divergence (p=0.024). Thus, CREBBP may be involved in generating
evolutionary divergence particularly with regard to DNA methylation patterns.
Supplementary Table 7c.Comparison of the observed mutation frequency affecting
candidate genes for lymhomagenesis between our cohort and the FL / tFL cohorts of Okosun
et al. 201422 (% affected samples)
Gene
CREBBP
MLL2
TNFRSF14
EZH2
EP300
MYC
TP53
MEF2B
BCL6
BCL2
CDKN2A
602
603
604
605
606
607
608
Our series on FL
(primary and
relapse pooled)
62%
57%
22%
16%
6%
3%
3%
1.5%
0%
0%
Okosun et al.
for FL*
64%
82%
35%
20%
18%
-
Okosun et al.
for tFL**
70%
73%
40%
24%
12%
-
0%
-
-
MYD88
0%
2%
* according to Fig. 3 of Okosun et al. 201422
**according to Suppl. Fig. 7B of Okosun et al. 201422
12%
Supplementary Table 7d. Comparison of the observed frequency of mutations affecting
aSHM targets between our cohort and the FL / tFL cohort of Pasqualucci et al. 201423 (%
affected samples)
Region
Our series on FL
(primary and relapse
pooled)
Pasqualucci et al.
FL*
Pasqualucci et al.
transformed FL*
CIITA_SHM
BCL2_SHM
RHOH_SHM
BCL6_SHM
PAX5_SHM
REL1_SHM
IRF4_SHM
MYC_SHM
35%
99%
78%
54%
10%
7%
13%
16%
7%
87%
7%
60%
0%
7%
46%
87%
36%
64%
38%
23%
13%
46%
PIM1_SHM
6%
* according to Fig. 6 of Pasqualucci et al. 201423
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
F. Integrated correlation analysis
We correlated the measures of divergence (see section 3I) for all analyzed genetic and
epigenetic read outs with the time elapsed between samples to examine whether the
observed divergence is increasing over time. However, we found no support for this
expectation. Supplementary table 8 summarizes the correlation coefficients and confidence
intervals from Supplementary Figure 11.
Correlation with time between samples
IGHV_divergence
DNA_Methylation_divergence
Aberrant_SHM_divergence
Candidate_Gene_Mutation_divergence
Chromosomal_divergence
rhoEst
0.010
0.033
0.066
0.173
0.319
Lower
[-0.35
[-0.23
[-0.22
[-0.16
[-0.03
Upper
; 0.44]
; 0.28]
; 0.33]
; 0.44]
; 0.57]
Supplementary Table 8: Correlation of measures of divergence with time between samples
(log10TimeDiff) with 95% confidence intervals
21
627
628
a
b
629
630
c
d
631
632
633
634
635
636
637
e
Supplementary Figure 11. There is no evidence for a correlation of measures of divergence
with time. In particular, SHM related measures and methylation appear to be completely
uncorrelated with time. Note that the grey dots label observations for whom no IGHV
sequences and hence no pedigrees were available.
22
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
We also investigated the correlation between IGHV-divergence and divergence in the other
genetic and epigenetic readouts. There were pronounced correlations in all comparisons
(see Supplementary Table 9 and Supplementary Figures 12-14).
Correlation with IGHV-divergence
rhoEst Lower Upper
Aberrant_SHM_divergence
0.724 [ 0.40 ; 0.88]
Aberrant_SHM_divergence_BCL2only
0.723 [ 0.41 ; 0.88]
Aberrant_SHM_divergence_BCL2excluded
0.461 [-0.03 ; 0.76]
DNA_Methylation_divergence
0.516 [ 0.24 ; 0.72]
Candidate_Gene_Mutation_divergence
0.372 [-0.03 ; 0.67]
Chromosomal_divergence
0.475 [ 0.07 ; 0.70]
Supplementary Table 9: Correlation of measures of divergence and IGHV-divergence with
95% confidence intervals
a
b
c
Supplementary
Figure
12.
IGHV-divergence
is
strongly
correlated
with
Aberrant_SHM_divergence. This is expected since SHM is a common biological process
behind both measures of divergence. This correlation is still strong when restricted to
23
661
662
663
664
665
666
667
668
669
670
671
672
673
674
aberrant SHM in BCL2 only, which is translocated near the IGHV locus. Excluding BCL2
sites reduces the observed correlation (not significant anymore).
Supplementary Figure 13. IGHV-divergence is also clearly correlated with
DNA_Methylation_divergence. Note that the correlation estimate is attenuated due to two
outliers.
a
b
Supplementary
Figure
14.
IGHV-divergence
is
also
correlated
to
Chromosomal_divergence, while the correlation with discrepant mutations in candidate
genes
is
not
significant.
24
675
676
677
678
679
680
681
G. Display of all data levels for prototypic cases
The following graphics illustrate the correlation between primary and relapse samples for all
analyzed omics levels. Each figure represents a prototypical case for the respective class of
the phylogenetic IGHV tree. Clinical characteristics of each sample are available via
MPI/SYS identifiers in Suppl. Table 1b.
a
b
c
d
682
683
684
685
686
Supplementary Figure 15: Integrative display of a case with “No evolution”. The sample IDs
are MPI-775 (primary tumor), MPI-776 (relapse). a) Phylogenetic tree of the Ig heavy-chain
sequences (Sanger sequencing). Blue (red) leaves indicate sequences from the primary
(relapse) tumor. b) Allele frequency of mutations in both samples. Shown are mutations in the
transcription start sites and in the protein coding regions of the sequenced genes. Red dots
indicate discrepant mutations. c) Scatter plot of methylation in both samples. Red dots indicate
differentially methylated CpGs. d) Copy number alterations and loss of heterozygosity (LOH)
events as measured by SNP arrays. Two upper panels show the copy number in the primary
and the relapse tumor, respectively. The third panel shows the difference between the two
profiles. The fourth panel depicts LOH regions for each sample in red. The bottom panel shows
25
regions where heterozygous calls in the primary tumor change to homozygous genotypes in the
relapse (brown) and vice versa (purple).
a
687
688
689
690
691
692
693
694
695
b
c
Supplementary Figure 16: Integrative display of a case with “Sequential evolution”. The
sample IDs are MPI-871 (primary tumor), MPI-891 (relapse). a) Phylogenetic tree of the Ig
heavy-chain sequences (Sanger sequencing). Blue (red) leaves indicate sequences from the
primary (relapse) tumor. b) Allele frequency of mutations in both samples. Shown are
mutations in the transcription start sites and in the protein coding regions of the sequenced
genes. Red dots indicate discrepant mutations. c) Scatter plot of methylation in both
samples. Red dots indicate differentially methylated CpGs.
(No copy number data available for relapse sample)
26
a
b
c
d
696
697
698
699
700
701
702
703
704
705
706
707
708
709
Supplementary Figure 17: Integrative display of a case with “Divergent evolution”. The
sample IDs are MPI-772 (primary tumor), MPI-771 (relapse). a) Phylogenetic tree of the
IGHV sequences (Sanger sequencing). Blue (red) leaves indicate sequences from the
primary (relapse) tumor. b) Allele frequency of mutations in both samples. Shown are
mutations in the transcription start sites and in the protein coding regions of the sequenced
genes. Red dots indicate discrepant mutations. c) Scatter plot of methylation in both
samples. Red dots indicate differentially methylated CpGs. d) Copy number alterations and
loss of heterozygosity (LOH) events as measured by SNP arrays. Two upper panels show
the copy number in the primary and the relapse tumor, respectively. The third panel shows
the difference between the two profiles. The fourth panel depicts LOH regions for each
sample in red. The bottom panel shows regions where heterozygous calls in the primary
tumor change to homozygous genotypes in the relapse (brown) and vice versa (purple).
27
a
b
c
d
710
711
712
713
714
715
716
717
718
719
720
721
722
723
Supplementary Figure 18a: Integrative display of a case with “Complex evolution”. The
sample IDs are SYS-016 (primary tumor), SYS-017 (first relapse) and SYS-018 (second
relapse). a) Phylogenetic tree of the Ig heavy-chain sequences (Sanger sequencing). Blue,
red and brown leaves indicate sequences from the primary tumor, the first and the second
relapse tumor, respectively. b) Allele frequency of mutations in both samples. Shown are
mutations in the transcription start sites and in the protein coding regions of the sequenced
genes. Red dots indicate discrepant mutations. c) Scatter plot of methylation in both
samples. Red dots indicate differentially methylated CpGs. d) Copy number alterations and
loss of heterozygosity (LOH) events as measured by SNP arrays. Two upper panels show
the copy number in the primary and the relapse tumor, respectively. The third panel shows
the difference between the two profiles. The fourth panel depicts LOH regions for each
sample in red. The bottom panel shows regions where heterozygous calls in the primary
tumor change to homozygous genotypes in the relapse (brown) and vice versa (purple).
28
a
b
c
724
725
726
727
728
729
730
731
732
733
734
735
Supplementary Figure 18b: Integrative display of a case with “Complex evolution”. The
sample IDs are SYS-016 (primary tumor), SYS-018 (second relapse). a) Allele frequency of
mutations in both samples. Shown are mutations in the transcription start sites and in the
protein coding regions of the sequenced genes. Red dots indicate discrepant mutations. b)
Scatter plot of methylation in both samples. Red dots indicate differentially methylated CpGs.
c) Copy number alterations and loss of heterozygosity (LOH) events as measured by SNP
arrays. Two upper panels show the copy number in the primary and the relapse tumor,
respectively. The third panel shows the difference between the two profiles. The fourth panel
depicts LOH regions for each sample in red. The bottom panel shows regions where
heterozygous calls in the primary tumor change to homozygous genotypes in the relapse
(brown) and vice versa (purple).
29
a
b
c
736
737
738
739
740
741
742
743
744
745
746
747
Supplementary Figure 18c: Integrative display of a case with “Complex evolution”. The
sample IDs are SYS-017 (first relapse), SYS-018 (second relapse). a) Allele frequency of
mutations in both samples. Shown are mutations in the transcription start sites and in the
protein coding regions of the sequenced genes. Red dots indicate discrepant mutations. b)
Scatter plot of methylation in both samples. Red dots indicate differentially methylated CpGs.
c) Copy number alterations and loss of heterozygosity (LOH) events as measured by SNP
arrays. Two upper panels show the copy number in the primary and the relapse tumor,
respectively. The third panel shows the difference between the two profiles. The fourth panel
depicts LOH regions for each sample in red. The bottom panel shows regions where
heterozygous calls in the primary tumor change to homozygous genotypes in the relapse
(brown)
and
vice
versa
(purple).
30
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
Supplementary References:
1. van Dongen JJ, Langerak AW, Brüggemann M, Evans PA, Hummel M, Lavender FL,
et al. Design and standardization of PCR primers and protocols for detection of clonal
immunoglobulin and T-cell receptor gene recombinations in suspect
lymphoproliferations: report of the BIOMED-2 Concerted Action BMH4-CT98-3936.
Leukemia 2003; 17: 2257-317
2. Pellissery S, Richter J, Haake A, Montesinos-Rongen M, Deckert M, Siebert R.
Somatic mutations altering Tyr641 of EZH2 are rare in primary central nervous
system lymphoma. Leuk Lymphoma 2010; 51: 2135-6.
3. Alamyar E, Giudicelli V, Li S, Duroux P, Lefranc MP. IMGT/HighV-QUEST: the
IMGT® web portal for immunoglobulin (IG) or antibody and T cell receptor (TR)
analysis from NGS high throughput and deep sequencing. Immunome Res 2012; 8:
26
4. Uduman M, Yaari G, Hershberg U, Stern JA, Shlomchik MJ, Kleinstein SH. Detecting
selection in immunoglobulin sequences. Nucleic Acids Res 2011; 39: W499-504
5. Hershberg U, Uduman M, Shlomchik MJ, Kleinstein SH. Improved methods for
detecting selection by mutation analysis of Ig V region sequences. International
Immunology 2008; 20: 683-694
6. Yaari G, Uduman M, Kleinstein SH. Quantifying selection in high-throughput
Immunoglobulin sequencing data sets. Nucleic Acids Res 2012; 40: e134
7. Whitehead A, Whitehead J. A general parametric approach to the meta-analysis of
randomized clinical trials. Statistics in Medicine 1991; 10: 1665-77
8. Hoffmann S, Otto C, Kurtz S, Sharma CM, Khaitovich P, Vogel J, et al. Fast mapping
of short sequences with mismatches, insertions and deletions using index structures.
PLoS Comput Biol 2009; 5: e1000502
9. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, et al. The Sequence
Alignment/Map format and SAMtools. Bioinformatics 2009; 25: 2078-9
10. Martín-Subero JI, Kreuz M, Bibikova M, Bentink S, Ammerpohl O, Wickham-Garcia E,
et al. New insights into the biology and origin of mature aggressive B-cell lymphomas
by combined epigenomic, genomic, and transcriptional profiling. Blood 2009; 113:
2488-97
11. Huber W, von Heydebreck A, Sueltmann H, Poustka A, Vingron M. Parameter
estimation for the calibration and variance stabilization of microarray data. Stat Appl
Genet Mol Biol 2003; 2
12. Venkatraman ES, Olshen AB. A faster circular binary segmentation algorithm for the
analysis of array CGH data. Bioinformatics 2007; 23: 657-63
13. Kapur JN, Sahoo PK, Wong AKC. A new method for gray-level picture thresholding
using the entropy of the histogram. Comput Vision Graph 1985; 29: 273-85
14. Beroukhim R, Lin M, Park Y, Hao K, Zhao X, Garraway LA, et al. Inferring loss-ofheterozygosity from unpaired tumors using high-density oligonucleotide SNP arrays.
PLoS Comput Biol 2006; 2: e41
15. Lin M, Wei LJ, Sellers WR, Lieberfarb M, Wong WH, Li C. dChipSNP: significance
curve and clustering of SNP-array-based loss-of-heterozygosity data. Bioinformatics
2004; 20: 1233-1240
16. Zhao X, Li C, Paez JG, Chin K, Jänne PA, Chen TH, et al. An integrated view of copy
number and allelic alterations in the cancer genome using single nucleotide
polymorphism arrays. Cancer Research 2004; 64:3060-3071
17. DiCiccio TJ, Efron B. Bootstrap Confidence Intervals. Stat Sci 1996; 11: 189-212
18. O'Riain C, O'Shea DM, Yang Y, Le Dieu R, Gribben JG, Summers K, et al. Arraybased DNA methylation profiling in follicular lymphoma. Leukemia 2009; 23: 1858-66
19. Lee TI, Jenner RG, Boyer LA, Guenther MG, Levine SS, Kumar RM, et al. Control of
developmental regulators by Polycomb in human embryonic stem cells. Cell 2006;
125: 301-13
31
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
AUTHOR CONTRIBUTIONS
821
822
823
824
825
826
827
828
829
830
831
832
WK, CP and AH provided patient samples and clinical data; AH, MS, KK and WK provided
the respective quality controlled DNA analytes and immunohistological staining; CP and HT
performed IGHV sequencing (Sanger + NGS); CP, DHe and HT provided bioinformatic
analysis of IGHV sequencing; CA performed phylogenetic analyses; DH, MR and RK
performed analysis of R/S ratio in IGHV regions; AH and OA performed methylation array
experiments; MK and DH provided biometric analysis of methylation arrays; RSc and RK
provided SNP-6.0 array data; KW and MK provided bioinformatic analysis of SNP6.0 arrays;
AH coordinated NGS analyses for non-IG SHM and candidate genes; MK, SH and DH
provided bioinformatic analysis of NGS data; DH performed the integrative biometric
analysis; RS and ML designed the project and the grant application (PIs); RS, RK, MK, DH,
ML interpreted data and wrote the manuscript; all authors read and approved the final
manuscript.
20. Bracken AP, Dietrich N, Pasini D, Hansen KH, Helin K Genome-wide mapping of
Polycomb target genes unravels their roles in cell fate transitions. Genes Dev 2006;
20: 1123-36
21. Schwaenen C, Viardot A, Berger H, Barth TF, Bentink S, Döhner H, et al. Microarraybased genomic profiling reveals novel genomic aberrations in follicular lymphoma
which associate with patient survival and gene expression status. Genes
Chromosomes Cancer 2009; 48: 39-54
22. Okosun J, Bödör C, Wang J, Araf S, Yang CY, Pan C, et al. Integrated genomic
analysis identifies recurrent mutations and evolution patterns driving the initiation and
progression of follicular lymphoma. Nat Genet. 2014; 46:176-81
23. Pasqualucci L, Khiabanian H, Fangazio M, Vasishtha M, Messina M, Holmes AB, et
al. Genetics of follicular lymphoma transformation. Cell Rep. 2014; 6:130-40
32