PDF hosted at the Radboud Repository of the Radboud University
Nijmegen
The following full text is a publisher's version.
For additional information about this publication click this link.
http://hdl.handle.net/2066/53444
Please be advised that this information was generated on 2022-05-30 and may be subject to
change.
Gene-Expression and Immunohistochemical Study of
Specific T-Cell Subsets and Accessory Cell Types in the
Transformation and Prognosis of Follicular Lymphoma
Annuska M. Glas, Laurent Knoops, Leonie Delahaye, Marie José Kersten, Robby E. Kibbelaar,
Lodewyk A. Wessels, Ryan van Laar, J. Han J.M. van Krieken, Joke W. Baars, John Raemaekers,
Philip M. Kluin, Laura J. van ’t Veer, and Daphne de Jong
ABSTRACT
Purpose
Despite the generally favorable clinical course in follicular lymphoma (FL), a minority of patients
have a poor prognosis—with death within 3 years of diagnosis—most often due to transformation
to aggressive disease.
Patients and Methods
In this study, we analyzed the potential of predicting early transformation on the basis of gene
expression and immunologic parameters in FL biopsy samples taken at diagnosis.
Results
At the gene-expression level, FL is a highly uniform disease at the time of diagnosis, precluding the
detection of sufficiently validated prognostic gene-expression profiles suitable for a clinical setting.
Combinations of differentially expressed genes indicate that immunologic mechanisms play a
differential role in the risk of early transformation. Using immunohistochemistry for specific cell
populations, the spatial distribution to neoplastic follicles and the activation of CD4 –positive
T-helper cells (P ⫽ .002) and specifically T-helper 1 (P ⫽ .004) were shown to be highly
discriminatory to predict early transformation. A role for functional modulation of follicular dendritic
cells could also be supported (P ⫽ .04). Other cell populations, including CD68-positive macro-
phages and regulatory T cells, were not differentially present.
Conclusion
These results support the identification of FL as an immunologically functional disease in which an
interaction of the tumor cells and the functional composition of the microenvironment determines
the clinical behavior.
J Clin Oncol 25:390-398. © 2007 by American Society of Clinical Oncology
INTRODUCTION
Follicular lymphoma (FL) is the most frequent in-
dolent B-cell lymphoma in adults. Generally, the
disease responds well to chemotherapy, but it is
characterized by frequent relapses. Despite many
different treatment approaches, survival has not im-
proved over the past decades. The disease is still
considered to be incurable. Transformation to ag-
gressive disease in the form of diffuse large B-cell
lymphoma (DLBCL) or to a lesser extent the devel-
opment of nonresponsiveness to therapy without
transformation form the major causes of disease-
related death.
1,2
Although a long, indolent course of the dis-
ease is considered to be prototypical, the variation
in survival is quite broad with up to 15% of the
patients dying early in the course of their disease.
A combined score of clinical parameters, such as
the Follicular Lymphoma International Prognos-
tic Index, is considered the most successful prog-
nosticator thus far, but does not reveal any
biologic mechanisms that play a role in transfor-
mation and prognosis.
3
Furthermore, to date very
few biologic prognosticators with clinical utility
have been identified. Histologic grading does cor-
relate with clinical outcome both for disease-free
and overall survival.
4
Its usefulness for individual
patient management is hampered, however, by its
subjective nature and poor reproducibility.
5
Most
consistently, a complex karyotype is indicative of
an adverse prognosis.
6,7
This is in line with the
current model of transformation being the result
of an accumulation of genetic alterations and pos-
sibly an inadequate repair process.
8
From the Departments of Pathology
and Medical Oncology/Hematology, the
Netherlands Cancer Institute; Agendia
BV; Department of Hematology,
Academic Medical Center, Amsterdam,
the Netherlands; Department of Pathol-
ogy, Central Laboratories, Friesland;
Department of Pathology, Radboud
University Medical Center Nijmegen;
Department of Hematology, Radboud
University Medical Center Nijmegen,
Nijmegen; Department of Pathology,
University Medical Center Groningen,
Groningen, the Netherlands; and the
Experimental Medicine Unit, Université
Catholique de Louvain, Belgium.
Submitted February 18, 2006; accepted
November 8, 2006; published online
ahead of print at www.jco.org on January
2, 2007.
Supported by a grant from the Belgian
Federation Against Cancer (L.K.).
A.M.G. and L.K. contributed equally to
this study.
Authors’ disclosures of potential con-
flicts of interest and author contribu-
tions are found at the end of this
article.
Address reprint requests to Daphne de
Jong, Department of Pathology, The
Netherlands Cancer Institute, Plesman-
laan 121, 1066 CX Amsterdam, the
Netherlands; e-mail: d.d.jong@nki.nl.
© 2007 by American Society of Clinical
Oncology
0732-183X/07/2504-390/$20.00
DOI: 10.1200/JCO.2006.06.1648
JOURNAL OF CLINICAL ONCOLOGY
ORIGINAL REPORT
VOLUME 25 䡠 NUMBER 4 䡠 FEBRUARY 1 2007
390
from 131.174.248.56
Information downloaded from jco.ascopubs.org and provided by at UNIVERSITEITSBIBLIOTHEEK on November 12, 2012
Copyright © 2007 American Society of Clinical Oncology. All rights reserved.
Previous immunohistochemical studies, and more recently
gene-expression studies, have lent support to a model of FL as a disease
of immunologically functional cells in which an interaction of the
tumor cells and the microenvironment determines the clinical
behavior.
9-13
It has been hypothesized that the specific genomic alter-
ations in FL cells may actually dictate the functional composition of
the microenvironment.
8
The cellular composition of the nonmalig-
nant infiltrate that is related to prognosis and the mechanisms that
play a role are not elucidated yet.
In this study, we explored models on pathogenesis in FL using
gene-expression profiling and immunohistochemistry concentrating
on the risk of transformation as an objective and disease-related bio-
logic end point.
PATIENTS AND METHODS
Patients
From the files of the departments of pathology of the contributing
institutions, three series of samples from patients with FL were collected for
gene-expression studies: representative frozen tissue of biopsy samples of 31
patients who showed histologically or cytologically proven transformation to
DLBCL within 3 years after diagnosis; representative frozen tissue of samples
of 35 patients with fully documented minimum follow-up of 7 years and
without clinical, histologic, or cytologic signs of transformation; and represen-
tative frozen tissue of biopsy samples of 24 patients with DLBCL as transfor-
mation of previously diagnosed FL.
Biopsy samples of FL were included when taken before treatment or
during follow-up after a varying time of watch and wait. Samples were in-
cluded as well at time of relapse if biopsy site was outside the field of radiother-
apy after local radiotherapy as initial treatment. Time of diagnosis ranged from
1986 to 2004 and treatment protocols were various, including chlorambucil,
combination chemotherapy (cyclophosphamide, vincristin, prednisone), and
radiation therapy. Patients with FL treated with high-dose chemotherapy and
autologous peripheral stem-cell or bone marrow transplantation outside the
context of transformed disease were excluded from the study. None of the
patients received rituximab.
For immunohistochemical analysis, representative formalin-fixed and
paraffin-embedded biopsy samples from 25 patients with transformation
within 3 years using the same selection criteria as discussed earlier were col-
lected. Of these, 15 were also included in the gene-expression study. Thirty-
three instances of patients without signs of transformation during a minimum
follow-up time of 7 years (as described) were included, 15 of which were also
included in the gene-expression analysis. Clinical characteristics of all patients
are listed in Table 1.
Gene-Expression Analysis
RNA isolation, amplification, labeling, and hybridization were per-
formed as previously described.
9
Detailed protocols can be found on the
NKI/AvL Department of Pathology Microarray Projects Web page.
14
All sam-
ples were cohybridized with a standard reference of pooled and amplified RNA
isolated from five routine tonsillectomy specimens, chosen and opti-
mized to identify small changes in expression levels between the tumor
groups of interest.
cDNA microarray slides, with a complexity of 19,200 spots per slide, were
prepared at the Central Microarray Facility at the Netherlands Cancer Institute.
15
The description of this study follows the minimum information about a
microarray experiment (MIAMI) guidelines issued by the Microarray Gene
Expression Data group.
16
Immunohistochemistry
Immunohistochemical stainings were performed on complete paraffin
sections of FL samples according to standard antigen microwave/citrate-based
retrieval techniques. Antibodies used are listed in Table 2.
For CD20, the pattern of interfollicular tumor cell distribution was
scored. MIB1 staining was assessed semiquantitatively as a proportion of total
number of tumor cells. T-cell subpopulations were semiquantitatively assessed
in a four-tiered fashion (0% to 1%; 1% to 5%; 5% to 10%; ⬎ 10%) relative to
total numbers of T-cells and spatial distribution was scored as predominantly
intrafollicular, predominantly interfollicular, or diffuse. CD69 as a marker for
T-cell activation was scored relative to the total number of T cells based on
pattern recognition of the infiltrate. CD69 positivity on tumor cells was not
taken into account. CD21 and CD23 as markers for follicular dendritic mesh-
works were scored in a four-tiered system (absent, minority of the neoplastic
follicles with disrupted meshworks, majority of neoplastic follicles with well-
developed meshworks, uniformly well-developed meshworks). CD68-
positive macrophages were counted as absolute cell numbers per three
representative follicular high-power fields (at magnification ⫻600). All scor-
ings were performed in a blinded fashion for clinical outcome.
Statistical Analysis
Both for the gene-expression study and the immunohistochemical anal-
ysis, we chose to compare the instances of clinical spectrum extremes, rather
than random samples of FL reflecting an unbiased clinical distribution.
9,17
Because the far majority of FL has a favorable prognosis and only approxi-
mately 20% of the patients showed early transformation and/or a poor prog-
nosis, an unbiased series would show a significant under-representation of
poor prognosis patients. This would potentially inhibit the recognition of
minor but biologically significant information in this group. The presently
used choice, however, precludes formal survival and Cox regression analysis.
As a consequence of the protracted clinical course in the majority of the
patients and the dominant elderly age group, patients often die of other causes
than of lymphoma. Therefore, we concentrated on disease-related survival
rather than overall survival.
Unsupervised Analysis
Unsupervised analysis was performed on a selection of the total patient
series consisting of 24 patients with early proven transformation; 22 patients
without transformation; and 24 patients with DLBCL as transformation of
previously diagnosed FL.
Gene clustering was performed on 9,441 significantly differentially
expressed genes, based on the Rosetta error model.
18
Agglomerative hier-
archical clustering was performed using complete linkage similarity met-
rics in Genesis.
19
Supervised analysis of the FL gene-expression data was approached using
several computational strategies based on the Shrunken Centroids method
in the significance analysis of microarrays (SAM) software on a selection of
24 patients with rapid transformation to DLBCL and 22 patients without
transformation
20
and on signal-to-noise ratios on a similar selection of
patients groups.
For construction of a formal predictive gene-expression profile, genes
that optimally separated the two classes best were determined by using
different classifier methods in a train and validation protocol as described
by Wessels et al.
21,22
RESULTS
Unsupervised Analysis Can Distinguish FL
From DLBCL
As a first step, unsupervised two-dimensional hierarchical clus-
tering was performed on 46 FL samples including both clinical groups
and 24 DLBCL samples. This showed a clear separation between FL
and DLBCL, with one cluster containing 75% of DLBCL samples (18
of 24) and only 7% of FL samples (three of 45; Fig 1). Unsupervised
clustering was not able to distinguish clinically favorable FL (n ⫽ 22)
and unfavorable FL (n ⫽ 24), underlining the biologic homogeneity of
this disease.
Prognostic Biomarkers in Follicular Lymphoma
www.jco.org 391
from 131.174.248.56
Information downloaded from jco.ascopubs.org and provided by at UNIVERSITEITSBIBLIOTHEEK on November 12, 2012
Copyright © 2007 American Society of Clinical Oncology. All rights reserved.
Discriminatively Expressed Genes Between Clinically
Distinct Groups of FL Are Comparable to Genes
Expressed in Activated Lymphoid Tissue
To focus on gene-expression differences between both clinical
groups, reporter genes were selected using SAM as well as signal-
to-noise ratios.
SAM analysis of nontransforming FL (n ⫽ 22 patients) versus
rapidly transforming FL (n ⫽ 24) resulted in 86 genes with signifi-
cantly different expression between the two groups (with a false dis-
covery rate of 8%). Signal-to-noise ratios (SNRs) were calculated for a
selection of patients representing the clinical extremes of the spectrum
(Fig. 2).
Gene ontology analysis revealed that both the classical SNR
approach and the SAM approach showed a significant over-
representation between both clinical FL groups of the genes involved
in immunologic functions (30% overall of SAM and SNR identified
genes). Detailed information on these genes and functional grouping
is listed in Table 3. The majority of these immune response–related
genes were expressed at relatively higher levels in rapidly transforming
FL than in nontransforming FL (25 of 30 markers, excluding the
various immunoglobulins). Moreover, the relative expression levels in
the rapidly transforming FL were very similar or relatively increased to
normal lymphoid tissue with follicular hyperplasia, which was used as
a hybridization reference, while their expression was relatively down-
regulated in nontransforming FL. Therefore, the rapidly transforming
group showed an overall higher similarity of gene expression to follic-
ular hyperplasia.
Table 1. Clinical Characteristics of the Patients in the Gene Expression and Immunohistochemical Studies
Study
FL With Transformation
Within 3 Years
FL Without
Transformation Within
⬎ 7 Years P
Transformed FL
(n ⫽ 24)
Gene Expression
No. of patients 31 35
Interval to transformation
Median 15
Range 3-29 Not applicable 0-169
Follow-up
Median 30 120 57
Range 8-115 89-304 7-180
DOD ⫹ DOOC, % 70.9 14.2 ⬍ .0001 79.1
DOD, % 67.7 8.6 ⬍ .0001
Age, years .027
Median 52 44 53
Range 30-84 30-76 29-74
Stage III or IV, % 67.7 50 NS
“B” symptoms present, % 29.0 5.9 NS
Poor PS (WHO 2-4), % 0 0 NS
No. extranodal sites ⬎ 1, % 34.6 6.1 .018
LDH level elevated, % 22.7 0 .032
IPI 0-1 v 2 v 3-5, % 51.7 v 20.7 v 27.5 90.9 v 9.1 v 0
Immunohistochemical
No. of patients 25 33
Interval to transformation, months
Median 14
Range 3-36 Not applicable
Follow-up, months
Median 69 149
Range 7-218 96-276
DOD ⫹ DOOC, % 56 24.2 NS
DOD, % 56 16.6 .002
Age, years
Median 55.4 46.2
Range 30-74 30-75 .04
Stage III or IV, % 72 54.5 NS
B symptoms present, % 32 6.2 .04
Poor PS (WHO 2-4), % 0 4.1 .0001
No. extranodal sites ⬎ 1, % 27.7 4.1 NS
LDH level elevated, % 25 0 .03
IPI 0-1 v 2 v 3-5, % 50 v 27 v 23 97 v 3 v 0
NOTE. An imbalance exists between the two clinical groups of FL known clinical prognostic factors, IPI risk groups, and survival. This is fully in line with rapid and
absent transformation as a selection criterion for the groups and indeed supports that the series is not biased.
Abbreviations: FL, follicular lymphoma; DOD, dead of disease; DOOC, dead of other causes; NS, not significant; PS, performance status; LDH, lactate
dehydrogenase; IPI, International prognostic index.
Glas et al
392
JOURNAL OF CLINICAL ONCOLOGY
from 131.174.248.56
Information downloaded from jco.ascopubs.org and provided by at UNIVERSITEITSBIBLIOTHEEK on November 12, 2012
Copyright © 2007 American Society of Clinical Oncology. All rights reserved.
Translation of Gene-Expression Patterns to Lymphoid
Cell Populations Using Immunohistochemistry
To further define the cell types involved in prognosis in FL and to
analyze the variations in relative cell numbers and the activation state
of specific cell populations, immunohistochemical studies were per-
formed. The results are summarized in Table 4 and illustrated in
Figure 3. There were no significant differences in tumor cell distribu-
tion and total numbers of T cells. Most importantly, in FL with rapid
transformation, CD4-positive T cells were predominantly found
within the neoplastic follicles with or without an interfollicular com-
ponent. In contrast, the CD4-positive T cells were predominantly
found between the neoplastic follicles in the nontransforming pa-
tients. No differences were found for CD8- and CD57-positive popu-
lations. Virtually no CD56 positive natural killer cells were seen in FL
samples. Specific cell populations of T-helper1 cells and regulatory T
cells were studied using T-bet and forkhead box protein P3 (FOXP3)
immunostainings, respectively. T-helper1 cells were present as a mi-
nor interfollicular population with significantly higher densities in
samples from the rapidly transforming patients, but never exceeding
1% of the total T-cell numbers. Regulatory T cells were present at
varying frequencies in approximately half of the cases containing less
than 5% positive cells per total number of T cells (26 of 53). Cases with
distinctly denser infiltrates of FOXP3-positive cells were not signifi-
cantly different between the two clinical groups.
The activation state of the T cells was significantly higher
(P ⫽ .017) in the rapidly transforming patients with uniform
expression of CD69 as an activation marker. In three patients,
tumor B cells uniformly expressed CD69. In all remaining patients,
FL without transformation, minimum follow-up 7 years
FL with transformation within 3 years after diagnosis
DLBCL as transformation of FL
log
2
expression ratio
2-2
Fig 1. Unsupervised two-dimensional hierarchical clustering on 45 follicular lymphoma (FL) samples and 24 diffuse large B-cell lymphomas (DLBCL). FL and DLBCL
can be separated, but not the two clinical categories of FL, patients with transformation within 3 years after diagnosis and those without signs of transformation with
a follow-up of at least 7 years.
Table 2. Antibodies Used in This Study
Antibody Clone Source and Location
Ki-67 MIB1 DAKO, Glostrup, Denmark
CD20 L26 DAKO
CD3 CD3 DAKO
CD4 4B12 Novacastra, Newcastle Upon Tyne, United Kingdom
CD8 C8/144B DAKO
T-bet anti-T-bet Zymed, Invitrogen, Breda, the Netherlands
CD57 leu-7 Becton Dickinson, Franklin Lakes, NJ
CD56 123C3.D5 NeoMarkers
FOXP3 236A/E7 AbCam, Cambridge, United Kingdom
CD21 1F8 DAKO
CD23 1B12 Novacastra
CD68 KP1 DAKO
CD69 CH11 Lab Vision, Fremont, CA
Abbreviation: FOXP3, forkhead box protein P3.
Prognostic Biomarkers in Follicular Lymphoma
www.jco.org 393
from 131.174.248.56
Information downloaded from jco.ascopubs.org and provided by at UNIVERSITEITSBIBLIOTHEEK on November 12, 2012
Copyright © 2007 American Society of Clinical Oncology. All rights reserved.