Immunohistochemical and genomic profiles of diffuse large B-cell lymphomas: Implications for targeted EZH2 inhibitor therapy?

Abstract

Enhancer of Zeste Homolog 2 (EZH2) plays an essential epigenetic role in Diffuse Large B Cell Lymphoma (DLBCL) development. Recurrent somatic heterozygous gain-of-function mutations of EZH2 have been identified in DLBCL, most notably affecting tyrosine 641 (Y641), inducing hyper-trimethylation of H3K27 (H3K27me3). Novel EZH2 inhibitors are being tested in phase 1 and 2 clinical trials but no study has examined which patients would most benefit from this treatment. We evaluated the immunohistochemical (IHC) methylation profiles of 82 patients with DLBCL, as well as the mutational profiles of 32 patients with DLBCL using NGS analysis of a panel of 34 genes involved in lymphomagenesis. A novel IHC score based on H3K27me2 and H3K27me3 expression was developed, capable of distinguishing patients with wild-type (WT) EZH2 and patients with EZH2 Y641 mutations (p = 10-5). NGS analysis revealed a subclonal EZH2 mutation pattern in EZH2 mutant patients with WT-like IHC methylation profiles, while associated mutations capable of upregulating EZH2 were detected in WT EZH2 patients with mutant-like IHC methylation profiles. IHC and mutational profiles highlight in vivo hyper-H3K27me3 and hypo-H3K27me2 status, pinpoint associated activating mutations and determine EZH2 mutation clonality, maximizing EZH2 inhibitor potential by identifying patients most likely to benefit from treatment.

Full text

Oncotarget16712
www.impactjournals.com/oncotarget
www.impactjournals.com/oncotarget/ Oncotarget, Vol. 6, No. 18
Immunohistochemical and genomic profiles of diffuse large
B-cell lymphomas: Implications for targeted EZH2 inhibitor
therapy?
Sydney Dubois1, Sylvain Mareschal1, Jean-Michel Picquenot1,2, Pierre-Julien
Viailly1, Elodie Bohers1, Marie Cornic2, Philippe Bertrand1, Elena Liana Veresezan1,2,
Philippe Ruminy1, Catherine Maingonnat1, Vinciane Marchand1, Hélène Lanic1,3,
Dominique Penther1, Christian Bastard1, Hervé Tilly1,3, Fabrice Jardin1,3
1INSERM U918, Centre Henri Becquerel, Université de Rouen, IRIB, Rouen, France
2Department of Pathology, Centre Henri Becquerel, Rouen, France
3Department of Clinical Hematology, Centre Henri Becquerel, Rouen, France
Correspondence to:
Fabrice Jardin, e-mail: fabrice.jardin@chb.unicancer.fr
Keywords: DLBCL, EZH2, Methylation, Immunohistochemistry, NGS
Received: October 20, 2014 Accepted: January 15, 2015 Published: February 05, 2015
ABSTRACT
Enhancer of Zeste Homolog 2 (EZH2) plays an essential epigenetic role in Diffuse
Large B Cell Lymphoma (DLBCL) development. Recurrent somatic heterozygous gain-
of-function mutations of EZH2 have been identified in DLBCL, most notably affecting
tyrosine 641 (Y641), inducing hyper-trimethylation of H3K27 (H3K27me3). Novel
EZH2 inhibitors are being tested in phase 1 and 2 clinical trials but no study has
examined which patients would most benefit from this treatment. We evaluated the
immunohistochemical (IHC) methylation profiles of 82 patients with DLBCL, as well
as the mutational profiles of 32 patients with DLBCL using NGS analysis of a panel of
34 genes involved in lymphomagenesis. A novel IHC score based on H3K27me2 and
H3K27me3 expression was developed, capable of distinguishing patients with wild-
type (WT) EZH2 and patients with EZH2 Y641 mutations (p = 10−5). NGS analysis
revealed a subclonal EZH2 mutation pattern in EZH2 mutant patients with WT-like
IHC methylation profiles, while associated mutations capable of upregulating EZH2
were detected in WT EZH2 patients with mutant-like IHC methylation profiles. IHC and
mutational profiles highlight in vivo hyper-H3K27me3 and hypo-H3K27me2 status,
pinpoint associated activating mutations and determine EZH2 mutation clonality,
maximizing EZH2 inhibitor potential by identifying patients most likely to benefit
from treatment.
Diffuse large B-cell lymphoma (DLBCL) is the
most common lymphoid malignancy, accounting for 30–
40% of all Non Hodgkin Lymphomas (NHL) [1]. Gene
expression proling has identied two main subtypes:
Germinal Center B-cell like (GCB) and Activated
B-Cell like (ABC), with the ABC subtype having the
most unfavorable prognosis [2, 3]. The development of
immuno-chemotherapy, and most notably rituximab, has
revolutionized the standard-of-care treatment of DLBCL
but a large part of patients still relapses or is refractory to
treatment.
Recently, epigenetic regulation has been shown to
be a crucial element in DLBCL development, and gene
repression mediated by Polycomb Repressive Complexes
1 and 2 (PRC1 and PRC2) has garnered attention.
Enhancer of Zeste Homolog 2 (EZH2), the catalytic
subunit of PRC2 [4], is a histone methyl-transferase
capable of specically mono-, di- and tri-methylating
histone H3 lysine 27 (H3K27me1, H3K27me2, and
H3K27me3) [5].
Recurrent somatic heterozygous gain-of-function
mutations of EZH2 have been identied in DLBCL, most
notably affecting tyrosine 641 (Y641), inducing increased
H3K27me3 [6, 7]. More recently, multiple studies have
shown cell lines with EZH2 mutations to be dependent
on the higher catalytic activity of mutant EZH2 Y641

Oncotarget16713
www.impactjournals.com/oncotarget
for proliferation, leading to the development of novel
EZH2 inhibitors for therapeutic use, capable of reversing
malignant phenotype [8–11].
Two EZH2 inhibitors are currently being tested
in phase 1 and 2 clinical trials both in patients with and
without EZH2 Y641 mutations (NCT01897571 and
NCT02082977), but no study has specically examined
which patients would be most susceptible to benet from
this treatment and how to screen for them. Patients with
EZH2 gain-of-function mutations have been pinpointed
as ideal EZH2 inhibitor recipients [8–11]; nevertheless, in
today’s targeted therapy era, it seems essential to establish
a method of detecting optimal candidates for EZH2
inhibitor treatment.
In the current study, we examined whether a simple
immunohistochemical (IHC) technique could be used to
distinguish wild-type (WT) -like and mutant-like EZH2
IHC methylation proles, and thus screen for patients
with conrmed overactive EZH2 at the protein level. We
also used Next Generation Sequencing (NGS) analysis to
further detail patients’ genomic proles and to determine
whether associated mutations could justify EZH2 inhibitor
treatment for patients otherwise not considered. We
propose that these methods, used in conjunction, could
serve to better determine candidates most likely to respond
to EZH2 inhibitor treatment.
MATERIALS & METHODS
Patients and biological samples
96 patients with de novo DLBCL at diagnosis
with available tumor DNA and Formalin-Fixed
Parafn-Embedded (FFPE) samples were included
for EZH2 Sanger sequencing analysis and subsequent
immunohistochemistry experiments. To provide a
comprehensive genomic description of DLBCL, targeted
NGS experiments were performed in 32 patients (20/96
and 12 additional cases not in our initial cohort). A
owchart summarizes the experimental methods used
on the entire cohort (Supplementary Figure 1). Table 1
summarizes the patients’ clinical characteristics.
Median follow-ups for overall survival and progression-
free survival were respectively 4.9 and 3.9 years. All
experiments were in accordance with the Helsinki
Declaration and the study was approved by the internal
review board.
Immunohistochemistry
Sections from FFPE tissue samples were used
to build Tissue Microarrays (TMAs). Information
on the primary antibodies used in this study (EZH2,
H3K27me1, H3K27me2 and H3K27me3) is summarized
in Supplementary Table 1. Deparafnization, rehydration,
and epitope retrieval was performed by PT Link
following the manufacturer’s instructions at pH 6
(DAKO, California, USA) and deparafnized sections
were stained using Vectastain kits (Vector Laboratories
Inc, California, USA) according to the manufacturer’s
instructions. The slides were then incubated with DAB+
chromogen for 5 minutes and counterstained with
hematoxyline for 2 minutes. Slides were scored in a
blinded fashion by an experienced anatomopathologist
(JMP). Slides were also scored in a blinded fashion by
a second independent anatomopathologist (ELV) in
order to assess correlation. Cases with lost TMA cores or
non-tumoral tissue were excluded. Tumors were scored
according to staining intensity (1–3, with 1 being weak
and 3 strong) and proportion of tumor cells stained (0–
10, with 0 representing negative staining, 1 representing
1–10% of positive tumor cells and 10 representing 91–
100% of positive tumor cells). For each antibody, a score
that ranged from 0 to 30 was calculated as the product
of staining intensity and proportion of tumor cells stained
[12]. Each tumor was represented 3 times on the TMAs
and the highest score was kept. For each patient, a me3/
me2 score was calculated:
me3
⁄
me2 score = log 2
a
me3 score +1
me2
score +1≤
GCB/ABC cell of origin (COO) subclassication
The GCB/ABC subtype was determined by cDNA-
mediated Annealing, Selection, extension, and Ligation
(DASL) technology based on the expression of 19 genes,
as previously described [13].
Ion torrent personal genome machine (PGM)
sequencing
Genomic DNA was submitted to Next Generat-
ion Sequencing (NGS) using a laboratory-developed
“Lymphopanel” set, designed to identify mutations in 34
genes important for lymphomagenesis (Supplementary
Table 2). This design covers 87 703 bases and generates
872 amplicons. Amplied libraries were submitted to
emulsion PCR with the Ion OneTouch™ 200 Template
Kit (Life Technologies, California, USA) using the Ion
OneTouch™ System (Life Technologies) according to the
manufacturer’s instructions. The generated Ion Sphere™
Particles (ISPs) were enriched with the Ion OneTouch™
Enrichment System and loaded and sequenced on Ion
316™ v2 Chips (Life Technologies).
PGM data analysis
Torrent Suite™ version 4.0 (Life Technologies)
software was used to perform primary analysis, including
signal processing, base calling, sequence alignment to
the reference genome (hg19) and generation of Binary
Alignment/Map (BAM) les. BAM les were used by
Torrent Suite™’s Variant Caller to detect point mutations

Oncotarget16714
www.impactjournals.com/oncotarget
as well as short insertions and deletions using the PGM
Somatic Low Stringency prole. VCF les generated by
Variant Caller were annotated by ANNOVAR [14].
Samples were considered of sufcient quality
when more than 90% of targeted bases were read at
least 20 times with sequencing and mapping precisions
of at least Q20. Only frameshift deletions and insertions,
nonframeshift deletions and substitutions, splicing,
nonsynonymous, stopgain or stoploss Single Nucleotide
Variations (SNVs) were kept. Variants present in dbSNP
(version 138) and absent in COSMIC (version 64)
were discarded, as were variants with a predictive SIFT
score > 0.05 [15]. A normal probability plot dened
thresholds separating true positives (conrmed by
Sanger sequencing, TVC score ≥ 22) from true negatives
(discredited by Sanger sequencing, TVC score < 9.5)
and highlighted a gray zone (9.5 < TVC score < 22) in
which variants must be conrmed by Sanger sequencing
or pyrosequencing.
Further verication by Sanger sequencing was
performed using a BigDye® Terminator v3.1 Cycle
Sequencing Kit (Life Technologies) and an ABI PRISM
3130 analyzer (Life Technologies). Primer sequences are
provided in Supplementary Table 3. Further verication by
pyrosequencing was performed using the PyroMark PCR
kit (Qiagen, France) with internal and sequencing primers
designed using PyroMark software (Qiagen). Bubble
charts to visualize validated variants per patient were
generated usingHighcharts.com (Highsoft AS, Norway).
Karyotyping and uorescent in situ
hybridization (FISH)
Cytogenetic analysis was performed according to
standard techniques. Slides were RHG-banded according
to Sehested [16] and karyotypes were described according
to the International System for Human Cytogenetic
Nomenclature. FISH using the LSI IGH/BCL2 Dual
Color, Dual Fusion Translocation Probe (Vysis, Downers
Grove, USA) was performed on metaphase preparations
according to the manufacturer’s instructions.
Statistical analysis
All statistical analyses except kappa scores were
performed using R software version 3.0.2 [17]. Kappa
scores were calculated using Medcalc software version
10.0.2.0. Overall Survival (OS) was calculated from
beginning of treatment to date of death or last patient
follow-up. Progression-Free Survival (PFS) was calculated
Table 1: Clinical characteristics of patients at diagnosis
Clinical parameter Patients at diagnosis (n = 96)
Gender M/F, n48/48
Age (years), median (range) 66 (17–87)
Adverse prognostic factors, n (%)
Age > 60 years 60 (63)
Ann Arbor stage III–IV 68 (71)
LDH > N 9 (9)
Extranodal sites ≥ 2 37 (39)
Bulky mass ≥ 10 cm 20 (21)
Performance status ≥ 2 26 (27)
IPI, n (%)
0–2 42 (44)
3–5 54 (56)
Treatment, n (%)
R-CHOP 38 (40)
R-ACVBP 17 (18)
R-mCHOP 13 (14)
R-IVA 1 (1)
Abbreviations: LDH, Lactate Dehydrogenase; IPI, International Prognostic Index; R, Rituximab; CHOP,
Cyclophosphamide, Hydroxydaunorubicin, Vincristine and Prednisone; ACVBP, Doxorubicin, Cyclophosphamide,
Vindesine, Bleomycin and Prednisone; mCHOP, miniCHOP; IVA, Ifosfamide, Etoposide and Cytarabine

Oncotarget16715
www.impactjournals.com/oncotarget
from beginning of treatment until disease progression,
relapse, death or last patient follow-up. Log-rank tests
(“survival” R package version 2.37.7) were used to assess
differences in OS and PFS rates calculated by Kaplan-
Meir estimates, as well as to perform univariate analysis.
Multivariate analysis was performed with a Cox regression
model. K-means cluster analysis was performed, with
cluster number set to k = 2. Statistical differences between
all other parameters were determined using χ2, Mann–
Whitney, or Fisher’s exact test when appropriate. p values
< 0.05 were considered statistically signicant.
RESULTS
Patient characteristics according to EZH2
somatic mutation status
Table 2 classies all patients with DASL and
Sanger sequencing data available based on their COO
subtype and EZH2 mutation status, and also highlights
the 82 patients usable for IHC. Of the 49 GCB subtype
patients, 12 were EZH2 Y641 mutant (24%), slightly
higher than the original report of 22% [18]. One EZH2
mutant patient in our 100-patient cohort was of the ABC
subtype, examples of which have been reported in the
literature [19]. IHC-usable WT EZH2 patients were
quite evenly split between ABC (n = 37/70) and GCB
(n = 30/70) subtype, while all IHC-usable EZH2 mutant
patients were of the GCB subtype (n = 12/12), as is
most frequent [18, 20]. EZH2 Y641 mutations showed
signicant association with t(14;18) translocation in
our cohort ( p < 10−4), corroborating previous studies
(Table 2) [20, 21].
Differential methylation levels of H3K27 are
distinguishable by IHC
FFPE samples of DLBCL placed on TMAs were
used for IHC with antibodies targeting EZH2, H3K27me1,
H3K27me2 and H3K27me3 (Supplementary Table 1).
We used breast cancer samples of different histological
subtypes, as well as DLBCL samples, as a guide to
determine primary antibody concentrations and incubation
times in order to observe gradients of EZH2 and H3K27
methylation IHC expression [12]. Figures 1A–1D show
representative images of differential IHC expression
of H3K27me2 and H3K27me3 from samples with WT
or Y641 mutant EZH2. Larger versions of the same
images are shown in Supplementary Figure 2. EZH2 IHC
expression was also able to showcase differential levels
of expression (not shown). H3K27me1 IHC expression
showed high expression levels for all patients, with no
differences observed (not shown).
Patients with EZH2 Y641 mutations present
distinct IHC methylation proles
There was no signicant difference in EZH2 or
H3K27me1 IHC expression between patients with mutant
and WT EZH2 (Table 2). Patients with EZH2 Y641 mutations
presented a signicantly lower H3K27me2 score ( p = 0.005)
and a signicantly higher H3K27me3 score ( p = 0.01) than
patients with WT EZH2 (Table 2). Hyper-trimethylation and
hypo-dimethylation in patients with EZH2 Y641 mutations
is therefore evident at the IHC level. There was no signicant
difference in either EZH2 or H3K27me1/2/3 IHC scores
between ABC and GCB subtypes (data not shown).
Table 2: Patients according to their EZH2 mutation status
Characteristics Total WT EZH2 EZH2 Y641 mutant p-value
Patients, n92 78 14
me3/me2 score usable, n82 70 12 0.65a
EZH2 IHC score, median (range) 18 (0–30) 18 (0–30) 21 (0–27) 0.8b
H3K27me1 IHC score, median (range) 30 30 30 1b
H3K27me2 IHC score, median (range) 27 (0–27) 27 (0–27) 18 (0–27) 0.005b
H3K27me3 IHC score, median (range) 18 (0–30) 18 (0–30) 27 (0–27) 0.01b
me3/me2 score, median (range) 0 (–4.8–4.8) –0.25 (–4.8–3.3) 0.56 (–0.56–4.8) 8.30E–05b
t(14;18), n17 89 3.50E–05a
Age (years), median (range) 66 (17–87) 66 (17–87) 63 (37–77) 0.23b
IPI: 0–2/3–5, n40/52 32/46 8/6 0.38a
GCB / ABC, n49/43 36/39 12/1 0.005a
aFisher’s Exact Test
bWilcoxon Rank Sum Test
Abbreviations: IPI, International Prognostic Index; GCB, Germinal Center B-Cell-like; ABC, Activated B-Cell-like

Oncotarget16716
www.impactjournals.com/oncotarget
We thus decided to implement a score based on the
ratio of me3 and me2 expression levels, in order to take
into account both criteria and gain statistical strength.
A logarithmic approach was used to obtain a wider
distribution (me3/me2 score detailed in methods).
Y641 EZH2 mutant patients had signicantly higher
me3/me2 scores than patients with WT EZH2 (p < 10−4)
(Figure 1E and Table 2) As me3/me2 scores for patients
with WT or mutant EZH2 overlapped at zero, three distinct
IHC methylation proles emerged, centered around zero: a
H3K27me3-high/H3K27me2-low prole (me3/me2 score
> 0, n = 12/82), a H3K27me3-low/H3K27me2-high prole
(me3/me2 score < 0, n = 38/82) and an intermediate
prole (me3/me2 score = 0, n = 32/82). Blinded analysis
by an independent pathologist without prior consultation
rendered a weighted kappa score of 0.55 (Kmax = 0.8,
k = 69% of Kmax).
The me3/me2 score is capable of distinguishing
patients based on their EZH2 mutation status. Indeed,
patients with EZH2 Y641 mutations mostly exhibit a
H3K27me3-high/H3K27me2-low prole (n = 7/12),
with 4/12 exhibiting an intermediate prole and 1/12
exhibiting a H3K27me3-low/H3K27me2-high prole.
On the other hand, patients with WT EZH2 status are
split between intermediate (n = 28/70) and H3K27me3-
low/H3K27me2-high proles (n = 37/70) (p = 10−5).
The maximum accuracy of the me3/me2 score was
88%, demonstrated for a threshold > 0 (Supplementary
Figure 3), leading us to merge patients with me3/me2
scores ≤ 0 into a single WT-like IHC methylation prole
group, compared to the me3/me2 score > 0 mutant-like
IHC methylation prole group.
NGS mutational proles allow more thorough
understanding of IHC methylation proles
In order to better understand the unexpected
IHC methylation proles observed for certain patients
of our cohort, we performed an NGS analysis of their
mutational proles using our Lymphopanel set of genes
Figure 1: Differential IHC H3K27me2/me3 expression can distinguish WT and mutant EZH2 DLBCL. (A–D) All images
are taken at 20× magnication. (A) and (B) are images from the same WT EZH2 tumor sample. (C) and (D) are images from the same Y641
EZH2 mutant sample. IHC scores for images A–D are respectively 27/30, 9/30, 9/30 and 27/30. (E) is a boxplot representation of me3/
me2 score according to EZH2 mutation status, showing signicantly higher score in EZH2 mutant tumor samples. The width of bars in E
is proportionate to sample size. p-values in E were calculated by a Mann–Whitney test.

Oncotarget16717
www.impactjournals.com/oncotarget
based on literature data obtained from whole exome
sequencing [22]. To this end, we sequenced all Y641
EZH2 mutant patients, as well as all WT EZH2 patients
with mutant-like IHC methylation proles. We also
included 12 additional patients to extend NGS analysis
to a total of 15 Y641 EZH2 mutant patients (13 GCB,
1 ABC, 1 other) and 17 WT EZH2 patients (13 ABC, 2
GCB, 2 other).
NGS results were sorted by quality scores and
Sanger or pyrosequencing when possible, as described
in the methods section (detailed in Supplementary
Table 4). The average overall depth was 215x and
the average depth for EZH2 Y646 codon was 414x.
A total of 127 variants were validated in this fashion
(Supplementary Table 5).
All EZH2 Y641 mutations found by Sanger
sequencing were conrmed by NGS, and their VAFs
as shown were calculated as the percentage of mutant
reads among total number of reads. No additional EZH2
Y641 mutations were found by NGS among our cohort,
and no A677 or A687 mutations were identied either.
The 15 EZH2 Y641 mutants were therefore exclusively
mutated at position Y641 and the 17 EZH2 WT patients
were conrmed to be WT. VAFs for EZH2 mutations
calculated by pyrosequencing were highly correlated with
VAFs calculated by NGS analysis (Pearson’s r = 0.93,
p < 10−5), legitimizing our NGS calculation method
of VAFs for the other genes of the Lymphopanel
(Supplementary Figure 4).
Variants and their VAFs were represented in a
bubble chart format according to their COO subtype
(Figure 2A, 2B).
EZH2 mutations are majoritarily clonal in
DLBCL
The clonal status of EZH2 mutations was
established by comparing VAFs of EZH2 Y641 mutations
with the average of those of the associated mutations
in each patient. Although the direct comparison was
complicated by taking into account all VAFs available
(including those > 50% potentially due to CNVs), we
were able to distinguish two different patterns for EZH2
mutations. The majority of EZH2 mutations (n = 12/15,
80%) represented true clonal events with similar VAFs
for other genes mutated in the same sample (Figure
2A, blue bubbles and Figure 2B, patient 445). Of note,
patient 1687 seems to present a clonal mutation of EZH2
but low tumor content. True subclonal EZH2 mutations,
with lower EZH2 mutation VAFs compared to other
mutations, were found in 3/15 (20%) samples (Figure
2A, orange bubbles). K-means clustering was performed
to separate clonal and subclonal mutations (k = 2) and
successfully segregated these three patients (Figure 3).
The percentages of clonal and subclonal EZH2 mutations
in our cohort are very similar to those found in a cohort
of 43 Follicular Lymphomas (FL) in a recent study by
Bödör et al [23]. A recent study in DLBCL also found
a similar distribution of clonal versus subclonal EZH2
mutations [24].
Interestingly, three EZH2 mutant GCB patients
(1528, 1639 and 1478) also harbor a mutation in MYD88.
While they remain anectodal, given the low sample size,
two of these (1639 and 1478) present similar VAFs in
both EZH2 and MYD88 mutations (38.9% and 36.4%
respectively for 13944 and 23.8% and 27.4% respectively
for 16995), indicating driver/clonal mutation status for
both EZH2 and MYD88. On the other hand, sample 1528
hosts an EZH2 mutation with a low VAF of 6% and a
MYD88 mutation with 38.1% VAF, suggesting a driver
MYD88 mutation with a subsequent EZH2 mutation,
indicative of a secondary EZH2 mutation acquisition in a
de novo case of DLBCL.
Higher number of Lymphopanel variants among
the GCB subtype
On average, patients of GCB subtype (n = 15)
presented 5.2 validated variants among the Lymphopanel
genes (Figure 2A) while patients of ABC subtype
(n = 13) presented only 2.9 validated variants (Figure 2B)
(p = 0.02). Only 1 GCB patient (6.7%) presented no
variants according to our criteria, compared to 3 ABC
patients (23.1%).
Furthermore, there were 11 cases of genes
displaying more than 1 variant in GCB patients (n =
9/15, 60%), and only 4 such cases in ABC patients (n =
3/13, 23.1%). Such genes in GCB patients in our cohort
included KMT2D, GNA13 and CREBBP (respectively 4,
4, and 2 cases each of patients with more than 1 variant).
This mutational prole was very similar to that described
in FL [23]. By contrast, such cases in ABC patients
were evenly distributed among 4 genes (PIM1, PRDM1,
TNFAIP3 and TNFRSF14), with 1 case in each, indicating
no particular variability hotspot.
Subclonal and low-VAF EZH2 mutations may
explain unexpected WT-like IHC proles
Potentially contributing to explain EZH2 mutant
patients with WT-like IHC methylation proles, we noted
that, despite small sample size, the me3/me2 score tended
to correlate with EZH2 mutation VAF (p = 0.09, Pearson’s
r = 0.51). Of the ve EZH2 mutant patients presenting
a me3/me2 score ≤ 0 (304, 494, 1524, 1528 and 1623),
two (1528 and 1623) present low VAFs of 6% and 8.4%
respectively. These two patients also exhibit a subclonal
EZH2 mutation, as determined by NGS and clustering
(Figure 2A and Figure 3), suggesting that EZH2 inhibitor
treatment might be less efcient.

Oncotarget16718
www.impactjournals.com/oncotarget
Figure 2: Genomic proles of patients according to DLBCL subtype. Validated variants for each patient are plotted in a bubble chart,
with bubble size reecting variant VAF, not corrected for CNVs. Patients are represented by Unique Personal Number (UPN). The value of
each sample’s me3/me2 score is shown, with NA corresponding to samples not present in our IHC study. Genes are ordered from most frequent
to least frequent, with EZH2 rst. (A) represents all GCB subtype patients with at least one mutation in our cohort, with clonal and subclonal
EZH2 mutations outlined in blue and orange respectively. (B) Represents all ABC subtype patients with at least one mutation in our cohort.

Oncotarget16719
www.impactjournals.com/oncotarget
Associated mutations may explain unexpected
IHC methylation proles
The idea behind establishing mutational proles for
patients was to identify associated mutations, which might
give reasonable cause to accept or deny treatment options,
including EZH2 inhibitors, for a patient.
In our cohort, ve WT EZH2 patients (1768, 1342,
1631, 773 and 478) presented a mutant-like IHC methylation
prole. Four of these patients are of the ABC subtype,
suggesting a potential EZH2 mutation bypass in ABC
patients. Furthermore, two of these patients (1768 and 1631)
showed remarkably similar mutational proles (Figure 2B),
both of them harboring mutations in TP53, MYD88, and
PRDM1, whereas no other ABC-subtype patient in our
cohort exhibited an association of either of these mutated
genes. An additional mutation in PIM1 (patient 1768)
proved interesting as well, as these were the only cases of
mutations in PRDM1 and/or PIM1 in our cohort.
Low EZH2 IHC expression is associated
with better prognosis in ABC-DLBCL
Survival analysis was performed on the 70 patients
treated with R-chemotherapy, as detailed in Table 1. The
median follow-up for OS and PFS was 5.1 and 4.5 years
respectively.
Following the thresholds dened by a previous
study [25], low EZH2 IHC expression (< 70% of tumoral
cells stained) was observed in 36% of patients (55% ABC
and 39% GCB), whereas high EZH2 IHC expression
(≥ 70% of tumoral cells stained) was observed in 64% of
patients (41% ABC and 53% GCB). In univariate analysis,
low EZH2 IHC expression was signicantly associated
with superior OS ( p = 0.035, OS = 77% at 3 years versus
35%) and PFS ( p = 0.02, PFS = 77% at 3 years versus
29%) in ABC patients treated with R-chemotherapy
(Figure 4A, 4B). However, in a multivariate analysis
including IPI and EZH2 IHC expression in this ABC-
DLBCL subgroup, neither low EZH2 IHC expression
nor IPI was a statistically signicant prognostic factor,
with low sample number potentially responsible for
this drawback. Of note, the prognostic impact of
EZH2 expression was not observed in GCB patients
(Figure 4C, 4D). Furthermore, no correlation was found
between prognosis and IHC methylation prole in our
cohort (data not shown) [23, 26].
DISCUSSION
We have analyzed EZH2, H3K27me1, H3K27me2
and H3K27me3 IHC expression in relation to EZH2 somatic
mutation status in a cohort of patients with DLBCL and
shown that a simple IHC experiment is able to distinguish
Figure 3: Clustering by EZH2 mutation VAF relative to associated mutation VAFs enables subclonal mutation
detection. The log ratio of EZH2 VAF and average of associated mutation VAFs was calculated for each patient. K means clustering (k = 2)
was performed and isolated patients 1528, 1251 and 1623 as a unique group with subclonal EZH2 mutations. Horizontal lines indicate
means for each cluster and vertical dotted lines represent each point’s distance to the cluster’s mean.

Oncotarget16720
www.impactjournals.com/oncotarget
patients with WT EZH2 and patients with EZH2 Y641
mutations according to their me3/me2 score in the majority
of cases. This result conrms the accumulation of steady
state levels of H3K27me2 in WT EZH2 patients and the
increase in H3K27me3 levels with lower H3K27me2 steady
state levels in patients with EZH2 Y641 mutations at the
IHC level. To our knowledge, this is the rst such study in
DLBCL. A previous study showed variable H3K27me3 and
EZH2 IHC expression regardless of EZH2 mutation status in
FL and H3K27me2 IHC expression was not analyzed [26].
We have also shown that no signicant difference
exists between patients with WT or mutant EZH2 in either
EZH2 or H3K27me1 IHC expression. Lower H3K27me1
expression could have been expected in EZH2 mutant
samples; however, decreased H3K27me1 in EZH2 mutant
cell lines is not always observed [6] and H3K27me1
formation can also be catalyzed by noncanonical PRC2
complexes containing WT EZH1 [27]. The lack of
difference in EZH2 IHC expression between patients
with WT or mutant EZH2, previously shown in FL [26],
conrms that the mutation mostly affects EZH2 activity,
although a recent study has identied a mechanism by
which it also affects EZH2 stability [28].
Most importantly, our me3/me2 score highli-
ghts patients with “mutant-like” and “WT-like” IHC
methylation proles. In patients with DLBCL, our IHC
assay should be carried out alongside Sanger sequencing
for EZH2. We propose that when both parameters are
concordant, no further testing would be necessary: EZH2
mutant patients with mutant-like IHC methylation proles
would be recommended for EZH2 inhibitor treatment,
whereas WT EZH2 patients with WT-like IHC methylation
Figure 4: Low IHC EZH2 expression is a positive prognostic indicator in ABC-DLBCL. Survival was calculated on ABC-
subtype and GCB-subtype patients with R-chemotherapy treatment (n = 30 and n = 31 respectively), divided into EZH2-low (< 70%) and
EZH2-high (≥ 70%) groups. (A) and (B) show OS and PFS respectively, calculated for ABC subtype patients. (C) and (D) show OS and
PFS respectively, calculated for GCB subtype patients. Low EZH2 expression is associated with signicantly higher OS and PFS in ABC-
DLBCL patients, whereas no difference is observed in GCB-DLBCL patients.

Oncotarget16721
www.impactjournals.com/oncotarget
proles would not. For patients with discordant IHC assay
and Sanger results, NGS sequencing should be performed
in order to detect EZH2 mutation VAF or associated
mutations which might justify accepting or denying
EZH2 inhibitor treatment (Figure 5). Thus, a fast and
readily accessible combination strategy including Sanger
sequencing and an IHC assay would serve an initial
ltering purpose, successfully singling out patients most
likely to benet from EZH2 inhibitor treatment, while
restricting the number of patients screened by NGS for
EZH2 inhibitor treatment approval.
Further comforting our hypothesis that immunohisto
chemistry is a valuable tool in the determina tion of
patients apt for EZH2 inhibitor treatment, Mccabe et al’s
study showed that among EZH2 mutant cell lines,
H3K27me3 Western Blot levels were signicantly higher
in transcriptionally responsive cell lines, indicating
that the association of EZH2 mutation status and
hypertrimethylation might be a more sensitive marker
for EZH2 inhibitor treatment than EZH2 mutation status
only [8]. Additionally, a study showed that cell lines
presenting low H3K27me2 levels in association with high
H3K27me3 levels in Western Blot were more respon-
sive to the anti-proliferative effects of EZH2 inhibitors,
highlighting the importance of a mutant-like methylation
prole in prospective patients [11]. IHC assays do present
drawbacks in terms of inter-laboratory reproducibility,
although differences could be reduced by using pixel
analysis software to score staining for instance [29].
We found a minority of patients with unexpected
WT-like or mutant-like IHC methylation proles,
given their mutation status, potentially predicting a
respectively impaired or improved response to EZH2
inhibitor treatment. One explanation comes in the form
of EZH2 mutation clonality analysis, and associated
mutations might point to explanations for the remaining
cases. Overall, our NGS study revealed similar mutation
frequencies in genes previously analyzed in large DLBCL
genomic studies [22, 30, 31] Interestingly, GCB-DLBCL
with EZH2 mutations in our cohort showed genomic
proles similar to those previously described for FL, with
frequent associated mutations in CREBBP, KMT2D and
TNFRSF14, potentially indicating a common genetic
history between GCB-EZH2 mutant-DLBCL and FL [23].
Similar data was obtained in a large DLBCL genomic
study, where 5 of 7 patients with EZH2 mutations
presented associated mutations of TNFRSF14 and 4
presented associated mutations of KMT2D [30].
NGS analysis has highlighted cases of interesting
associated mutations in either patients with WT or Y641
mutant EZH2. Although these patients represent anecdotal
evidence only at this time, they lay the groundwork for the
premise that associated mutations should also be taken into
account when deciding which patients to treat with EZH2
inhibitors. For instance, we detected unique mutations in
PIM1 and PRDM1 in patients 1768 and 1631 with WT
EZH2 but mutant-like IHC methylation proles. These
genes are part of the gene network heavily affected by
Figure 5: An IHC/Sanger combination approach as a decision aid for EZH2 inhibitor treatment. By using an initial
combination approach at time of diagnosis, three patient groups emerge, potentially simplifying EZH2 inhibitor treatment guidelines.
Further analysis by NGS would thus be restricted to patients with discordant Sanger sequencing and IHC methylation prole results.

Oncotarget16722
www.impactjournals.com/oncotarget
EZH2 binding and are involved in GC reaction [32, 33].
Interestingly, PIM1 was mutated in only one patient in
our cohort, whereas previous genomic studies showed
signicantly higher mutation frequencies [30, 31]. While a
previous study showed ABC-DLBCL cells to be refractory
to EZH2 inhibitor treatment, patient-specic associated
mutations such as these might modify their response and
should be evaluated [34].
Furthermore, associated mutations are essential
information when deciding on individual targeted
therapeutic cocktails. Patients with several targetable
mutations, such as patient 304 with mutations in both
EZH2 and MYC, might greatly benet from an inhibitor
combination approach [35, 36]. Indeed, in a recent
mouse model, it was shown that only the association of
an EZH2 Y641 mutation and MYC overexpression, and
not the EZH2 Y641 mutation alone, led to lymphoma
development [37].
Four of the ve WT EZH2 patients with mutant-
like IHC proles were of the ABC subtype. While this
may not be relevant for clinical trials which administer
EZH2 inhibitor treatment to GCB subtype patients
exclusively, it is indeed pertinent for clinical trials
where the main inclusion criterion is the presence of
EZH2 gain-of-function mutations. Although rare, EZH2
mutations in ABC subtype patients do exist, either linked
to misclassication or a change in subtype during disease
progression [19]. In any case, this result adds to the
still-open question of the extent to which EZH2 mutant
ABC subtype patients will benet from EZH2 inhibitor
treatment.
Our me3/me2 score was not correlated with prognosis,
although this was not unexpected, given previous studies
showing no correlation between EZH2 mutation status and
prognosis in FL [23, 26]. On the other hand, we showed
that low IHC EZH2 expression is correlated with superior
OS and PFS among ABC-DLBCL patients, identifying
a prognostic impact of our assay, although not present in
multivariate analysis, potentially due to low sample size. A
previous study in breast cancer also showed that low EZH2
expression is correlated with better Distant Disease Free
Survival (DDFS) [12], corroborating our ndings. On the
contrary, Lee et al recently analyzed EZH2 IHC expression
in a cohort of DLBCL patients of similar size and showed
that high EZH2 expression was associated with superior OS,
with EZH2-high ABC patients being the subgroup with the
highest OS, although this nding was not quite statistically
signicant in multivariate analysis [25]. Compared to
Lee et al, our cohort was marginally older, with a larger
percentage of patients over 60 years old or with Ann Arbor
stage III–IV at diagnosis. The molecular characteristics
of DLBCL have indeed been shown to be age-dependent
[38, 39]; however, although this might be a contributing
factor, the reasons for our discrepant ndings are still unclear.
EZH2 inhibitors are currently being tested in
clinical trials in DLBCL as novel and promising weapons
in clinicians’ therapeutic arsenal. This study has shown
that IHC and genomic proles can identify patients who
are most likely to benet from treatment with EZH2
inhibitors by highlighting a specic in vivo H3K27me3-
high/H3K27me2-low prole, determining EZH2 mutation
clonality and pinpointing associated activating mutations.
Immunohistochemistry could thus serve as a convenient,
fast, and easily accessible method to pre-screen patients
exhibiting high me3/me2 scores for sequencing for
associated mutations, thus reducing time and expenses
before determining optimal, patient-specic treatment. As
such, analyzing these parameters could maximize EZH2
inhibitor benet and potentially serve to grant access to
patients who would otherwise not have been considered.
ACKNOWLEDGMENTS
This study was funded by grants from the Ligue
Contre le Cancer (Comité de la Seine Maritime, France)
and from the Institut National du Cancer.
REFERENCES
1.
Swerdlow SH, Campo E, Harris NL, Jaffe ES, Pileri SA,
Stein H, Thiele J, Vardiman JW. WHO classication of
tumours of haematopoietic and lymphoid tissues. Lyon:
IARC; 2008.
2.
Alizadeh AA, Eisen MB, Davis RE, Ma C, Lossos IS,
Rosenwald A, Boldrick JC, Sabet H, Tran T, Yu X,
Powell JI, Yang L, Marti GE, et al. Distinct types of diffuse
large B-cell lymphoma identied by gene expression prol-
ing. Nature. 2000; 403:503–11.
3.
Wright G, Tan B, Rosenwald A, Hurt EH, Wiestner A,
Staudt LM. A gene expression-based method to diagnose clini-
cally distinct subgroups of diffuse large B cell lymphoma. Proc
Natl Acad Sci U S A. 2003; 100:9991–6. [Internet] Available
from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?art
id=187912&tool=pmcentrez&rendertype=abstract
4.
Margueron R, Reinberg D. The Polycomb complex PRC2
and its mark in life. Nature. 2011; 469:343–9. [cited 2014
Jul 9] [Internet] Available from: http://www.pubmedcentral.
nih.gov/articlerender.fcgi?artid=3760771&tool=pmcentrez
&rendertype=abstract
5.
Viré E, Brenner C, Deplus R, Blanchon L, Fraga M,
Didelot C, Morey L, Van Eynde A, Bernard D,
Vanderwinden J-M, Bollen M, Esteller M, Di Croce L,
et al. The Polycomb group protein EZH2 directly controls
DNA methylation. Nature. 2006; 439:871–4. [cited 2014
Jul 10] [Internet] Available from: http://www.ncbi.nlm.nih.
gov/pubmed/16357870
6.
Sneeringer CJ, Scott MP, Kuntz KW, Knutson SK,
Pollock RM, Richon VM, Copeland RA. Coordinated activ-
ities of wild-type plus mutant EZH2 drive tumor-associated
hypertrimethylation of lysine 27 on histone H3 (H3K27) in

Oncotarget16723
www.impactjournals.com/oncotarget
human B-cell lymphomas. Proc Natl Acad Sci U S A. 2010;
107:20980–5. [cited 2014 Jul 11] [Internet] Available from:
http://www.pubmedcentral.nih.gov/articlerender.fcgi?artid=
3000297&tool=pmcentrez&rendertype=abstract
7.
Yap DB, Chu J, Berg T, Schapira M, Cheng S-WG,
Moradian A, Morin RD, Mungall AJ, Meissner B, Boyle M,
Marquez VE, Marra M a, Gascoyne RD, et al. Somatic
mutations at EZH2 Y641 act dominantly through a mech-
anism of selectively altered PRC2 catalytic activity, to
increase H3K27 trimethylation. Blood. 2011; 117:2451–9.
[Internet] Available from: http://www.pubmedcentral.nih.
gov/articlerender.fcgi?artid=3062411&tool=pmcentrez&re
ndertype=abstract
8.
McCabe MT, Ott HM, Ganji G, Korenchuk S, Thompson C,
Van Aller GS, Liu Y, Graves AP, Della Pietra A, Diaz E,
LaFrance L V, Mellinger M, Duquenne C, et al. EZH2 inhi-
bition as a therapeutic strategy for lymphoma with EZH2-
activating mutations. Nature. Nature Publishing Group.
2012; 492:108–12. [Internet] [cited 2013 Jan 29] Available
from: http://www.ncbi.nlm.nih.gov/pubmed/23051747
9.
Knutson SK, Kawano S, Minoshima Y, Warholic NM,
Huang K-C, Xiao Y, Kadowaki T, Uesugi M, Kuznetsov G,
Kumar N, Wigle TJ, Klaus CR, Allain CJ, et al. Selective
inhibition of EZH2 by EPZ-6438 leads to potent antitumor
activity in EZH2-mutant non-hodgkin lymphoma. Mol
Cancer Ther. 2014; 13:842–54. [Internet] Available from:
http://www.ncbi.nlm.nih.gov/pubmed/24563539
10.
Qi W, Chan H, Teng L, Li L, Chuai S, Zhang R, Zeng J,
Li M, Fan H. Selective inhibition of Ezh2 by a small mol-
ecule inhibitor blocks tumor cells proliferation. Proc Natl
Acad Sci. 2012; 109:21360–5.
11.
Copeland RA, Richon VM, Scott MD, Sneeringer CJ,
Knutson SK, Pollock RM. US 8691507. 2014. [Internet]
Available from: http://www.google.com/patents/US8691507?
12.
Holm K, Grabau D, Lövgren K, Aradottir S, Gruvberger-
Saal S, Howlin J, Saal LH, Ethier SP, Bendahl P-O, Stål O,
Malmström P, Fernö M, Rydén L, et al. Global H3K27
trimethylation and EZH2 abundance in breast tumor sub-
types. Mol Oncol. 2012; [cited 2014 Jul 28]; 6:494–506.
[Internet] Available from: http://www.ncbi.nlm.nih.gov/
pubmed/22766277
13.
Bohers E, Mareschal S, Bouzelfen A, Marchand V,
Ruminy P, Maingonnat C, Ménard A-L, Etancelin P,
Bertrand P, Dubois S, Alcantara M, Bastard C,
Tilly H, et al. Targetable activating mutations are very fre-
quent in GCB and ABC diffuse large B-cell lymphoma.
Genes Chromosome Cancer. 2014; 014 [cited 2013 Dec
19]; 53:144–53. [Internet] Available from: http://www.ncbi.
nlm.nih.gov/pubmed/24327543
14.
Wang K, Li M, Hakonarson H. ANNOVAR: functional
annotation of genetic variants from high-throughput
sequencing data. Nucleic Acids Res. 2010; [cited 2014
Jul 15]; 38:e164. [Internet] Available from: http://www.
pubmedcentral.nih.gov/articlerender.fcgi?artid=2938201&
tool=pmcentrez&rendertype=abstract
15.
Kumar P, Henikoff S, Ng PC. Predicting the effects of cod-
ing non-synonymous variants on protein function using
the SIFT algorithm. Nat Protoc. 2009; [cited 2014 Jul 14];
4:1073–81. [Internet] Available from: http://www.ncbi.nlm.
nih.gov/pubmed/19561590
16.
Sehested J. A simple method for R banding of human chro-
mosomes, showing a pH-dependent connection between
R and G bands. Humangenetik. 1974; 21:55–8.
17.
Foundation for Statistical Computing: R: A language and
environment for statistical computing. Vienna, Austria.
[Internet] Available from: http://www.r-project.org/
18.
Morin RD, Johnson NA, Severson TM, Mungall AJ, An J,
Goya R, Paul JE, Boyle M, Woolcock BW, Kuchenbauer F,
Yap D, Humphries RK, Grifth OL, et al. Somatic muta-
tions altering EZH2 (Tyr641) in follicular and diffuse large
B-cell lymphomas of germinal-center origin. Nat Genet.
Nature Publishing Group. 2010; 42:181–5. [Internet] [cited
2013 Jan 29] Available from: http://www.pubmedcentral.
nih.gov/articlerender.fcgi?artid=2850970&tool=pmcentrez
&rendertype=abstract
19.
Pasqualucci L, Trifonov V, Fabbri G, Ma J, Rossi D,
Chiarenza A, Wells V a, Grunn A, Messina M, Elliot O,
Chan J, Bhagat G, Chadburn A, et al. Analysis of the cod-
ing genome of diffuse large B-cell lymphoma. Nat Genet.
Nature Publishing Group. 2011; 43:830–7. [cited 2013
Feb 5] [Internet] Available from: http://www.pubmedcen-
tral.nih.gov/articlerender.fcgi?artid=3297422&tool=pmcen
trez&rendertype=abstract
20.
Ryan RJH, Nitta M, Borger D, Zukerberg LR, Ferry JA,
Harris NL, Iafrate AJ, Bernstein BE, Sohani AR, Le LP.
EZH2 codon 641 mutations are common in BCL2-
rearranged germinal center B cell lymphomas. PLoS One.
2011; [cited 2014 May 6]; 6:e28585. [Internet] Available
from: http://www.pubmedcentral.nih.gov/articlerender.fcgi
?artid=3237460&tool=pmcentrez&rendertype=abstract
21.
Guo S, Chan JKC, Iqbal J, McKeithan T, Fu K, Meng B,
Pan Y, Cheuk W, Luo D, Wang R, Zhang W, Greiner TC,
Chan WC. EZH2 mutations in follicular lymphoma from
different ethnic groups and associated gene expression
alterations. Clin Cancer Res. 2014; [cited 2014 Jul 28];
20:3078–86. [Internet] Available from: http://www.ncbi.
nlm.nih.gov/pubmed/24634383
22.
Bohers E, Mareschal S, Bertrand P, Viailly PJ, Dubois S,
Maingonnat C, Ruminy P, Tilly H, Jardin F. Activating
somatic mutations in diffuse large B-cell lymphomas: les-
sons from next generation sequencing and key elements in
the precision medicine era. Leuk Lymphoma. 2014; [cited
2014 Dec 4]; (May):1–10. [Internet] Available from: http://
www.ncbi.nlm.nih.gov/pubmed/25130477
23.
Bödör C, Grossmann V. EZH2 mutations are frequent and
represent an early event in follicular lymphoma. Blood.
2013; [cited 2014 Jul 31]; [cited 2014 Jul 31]; 122:3165–8.
[Internet] Available from: http://bloodjournal.hematologyli-
brary.org/lookup/pmid?view=long&pmid=24052547

Oncotarget16724
www.impactjournals.com/oncotarget
24.
Jiang Y, Redmond D, Nie K, Eng KW, Clozel T, Martin
P, Tan LH, Melnick AM, Tam W, Elemento O. Deep
sequencing reveals clonal evolution patterns and muta-
tion events associated with relapse in B-cell lympho-
mas. Genome Biol. 2014; [cited 2014 Nov 18]; 15:432.
[Internet] Available from: http://www.pubmedcentral.nih.
gov/articlerender.fcgi?artid=4158101&tool=pmcentrez&r
endertype=abstract
25.
Lee HJ, Shin DH, Kim K Bin, Shin N, Park WY, Lee
JH, Choi KU, Kim JY, Lee CH, Sol MY. Polycomb pro-
tein EZH2 expression in diffuse large B-cell lymphoma
is associated with better prognosis in patients treated with
rituximab, cyclophosphamide, doxorubicin, vincristine and
prednisone. Leuk Lymphoma. 2014; [cited 2014 May 5];
55:2056–63. [Internet] Available from: http://www.ncbi.
nlm.nih.gov/pubmed/24304372
26.
Bodor C, O’Riain C, Wrench D, Matthews J, Iyengar S,
Tayyib H, Calaminici M, Clear A, Iqbal S, Quentmeier H,
Drexler H, Montoto S, Lister A, et al. EZH2 Y641 muta-
tions in follicular lymphoma. Leukemia. 2011; 25:726–9.
27.
Shen X, Liu Y, Hsu Y-J, Fujiwara Y, Kim J, Mao X,
Yuan G-C, Orkin SH. EZH1 mediates methylation on his-
tone H3 lysine 27 and complements EZH2 in maintaining
stem cell identity and executing pluripotency. Mol Cell.
2008; [cited 2014 Jul 15]; 32:491–502. [Internet] Available
from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
artid=2630502&tool=pmcentrez&rendertype=abstract
28.
Sahasrabuddhe A, Chen X, Chung F, Velusamy T,
Lim MS, Elenitoba-Johnson KSJ. Oncogenic Y641 muta-
tions in EZH2 prevent Jak2/β-TrCP-mediated degradation.
Oncogene. Nature Publishing Group. 2014; [cited 2014 Oct
1];(August 2013) 1–10. [Internet] Available from: http://
www.ncbi.nlm.nih.gov/pubmed/24469040
29.
Rizzardi AE, Johnson AT, Vogel RI, Pambuccian SE,
Henriksen J, Skubitz AP, Metzger GJ, Schmechel SC.
Quantitative comparison of immunohistochemical stain-
ing measured by digital image analysis versus patholo-
gist visual scoring. Diagn Pathol. 2012; 7:42. [Internet]
Available from: http://www.pubmedcentral.nih.gov/arti-
clerender.fcgi?artid=3379953&tool=pmcentrez&renderty
pe=abstract
30. Lohr JG, Stojanov P, Lawrence MS, Auclair D, Chapuy B,
Sougnez C, Cruz-Gordillo P, Knoechel B, Asmann YW,
Slager SL, Novak AJ, Dogan A, Ansell SM, et al. Discovery
and prioritization of somatic mutations in diffuse large
B-cell lymphoma (DLBCL) by whole-exome sequenc-
ing. Proc Natl Acad Sci U S A. 2012; [cited 2014 Oct
4]; 109:3879–84. [Internet] Available from: http://www.
pubmedcentral.nih.gov/articlerender.fcgi?artid=3309757&
tool=pmcentrez&rendertype=abstract
31.
Morin RD, Mungall K, Pleasance E, Mungall AJ, Goya R,
Huff RD, Scott DW, Ding J, Roth A, Chiu R, Corbett RD,
Chan FC, Mendez-lago M, et al. Mutational and structural
analysis of diffuse large B-cell lymphoma using whole-
genome sequencing. Blood. 2013; 122:1256–66.
32. Turner CA, Mack DH, Davis MM. Pillars article: Blimp-1, a
novel zinc nger-containing protein that can drive the matura-
tion of B lymphocytes into immunoglobulin-secreting cells.
1994. J Immunol. 2010; 185:5–14. [Internet] Available from:
http://www.ncbi.nlm.nih.gov/pubmed/20562267
33.
Velichutina I, Shaknovich R, Geng H, Johnson NA,
Gascoyne RD, Melnick AM, Elemento O. EZH2-mediated
epigenetic silencing in germinal center B cells contrib-
utes to proliferation and lymphomagenesis. Blood. 2010;
[cited 2014 Oct 28]; 116:5247–55. [Internet] Available
from: http://bloodjournal.hematologylibrary.org/ content/
116/24/5247.short
34.
Béguelin W, Popovic R, Teater M, Jiang Y. EZH2 is
required for germinal center formation and somatic EZH2
mutations promote lymphoid transformation. Cancer Cell.
2013; 23:677–92. [Internet] Available from: http://www.
sciencedirect.com/science/article/pii/S1535610813001797
35.
Zhao X, Lwin T, Zhang X, Huang a, Wang J, Marquez VE,
Chen-Kiang S, Dalton WS, Sotomayor E, Tao J. Disruption
of the MYC-miRNA-EZH2 loop to suppress aggressive
B-cell lymphoma survival and clonogenicity. Leukemia.
2013; [cited 2014 Sep 4]; 27:2341–50. [Internet] Available
from: http://www.pubmedcentral.nih.gov/articlerender.fcgi?
artid=4015113&tool=pmcentrez&rendertype=abstract
36.
Sander S, Bullinger L, Klapproth K, Fiedler K, Kestler HA,
Barth TFE, Möller P, Stilgenbauer S, Pollack JR, Wirth T.
MYC stimulates EZH2 expression by repression of its nega-
tive regulator miR-26a. Blood. 2008; [cited 2014 Sep 4];
112:4202–12. [Internet] Available from: http://blood-
journal.hematologylibrary.org/content/112/10/4202.short
37.
Berg T, Thoene S, Yap D, Wee T, Schoeler N, Roste P,
Lim E, Bilenky M, Mungall AJ, Oellerich T, Lee S, Lai CK,
Umlandt P, et al. A transgenic mouse model demonstrating the
oncogenic role of mutations in the polycomb-group gene EZH2
in lymphomagenesis. Blood. 2014; 123:3914–25. [Internet]
Available from: http://bloodjournal. hematologylibrary.org/
content/early/2014/05/06/blood-2012-12-473439.short
38.
Mareschal S, Lanic H, Ruminy P, Bastard C, Tilly H,
Jardin F. The proportion of activated B-cell like subtype
among de novo diffuse large B-cell lymphoma increases
with age. Haematologica. 2011; [cited 2014 Oct 15];
96:1888–90. [Internet] Available from: http://www.haema-
tologica.org/content/96/12/1888
39.
Klapper W, Kreuz M, Kohler C, Burkhardt B,
Szczepanowski M, Salaverria I, Hummel M, Loefer M,
Pillizzery S, Woessmann W, Schwänen C, Trümper L,
Wessendorf S, et al. Patient age at diagnosis is associated
with the molecular characteristics of diffuse large B-cell
lymphoma. Blood. 2012; [cited 2014 Sep 1]; 119:1882–7.
[Internet] Available from: http://bloodjournal.hematology-
library.org/content/119/8/1882.short