Published Ahead of Print on October 25, 2013, as doi:10.3324/haematol.2013.094474. Copyright 2013 Ferrata Storti Foundation.

Early lesions of follicular lymphoma: a genetic perspective by Emilie Mamessier, Joo Y. Song, Franziska C. Eberle, Svetlana Pack, Charlotte Drevet, Bruno Chetaille, Ziedulla Abdullaev, Jose' Adelaïde, Daniel Birnbaum, Max Chaffanet, Stefania Pittaluga, Sandrine Roulland, Andreas Chott, Elaine S. Jaffe, and Bertrand Nadel Haematologica 2013 [Epub ahead of print] Citation: Mamessier E, Song JY, Eberle FC, Pack S, Drevet C, Chetaille B, Abdullaev Z, Adelaïde J, Birnbaum D, Chaffanet M, Pittaluga S, Roulland S, Chott A, Jaffe ES, and Nadel B. Early lesions of follicular lymphoma: a genetic perspective. Haematologica. 2013; 98:xxx doi:10.3324/haematol.2013.094474 Publisher's Disclaimer. E-publishing ahead of print is increasingly important for the rapid dissemination of science. Haematologica is, therefore, E-publishing PDF files of an early version of manuscripts that have completed a regular peer review and have been accepted for publication. E-publishing of this PDF file has been approved by the authors. After having E-published Ahead of Print, manuscripts will then undergo technical and English editing, typesetting, proof correction and be presented for the authors' final approval; the final version of the manuscript will then appear in print on a regular issue of the journal. All legal disclaimers that apply to the journal also pertain to this production process.

Early lesions of follicular lymphoma: a genetic perspective

Running title: « Follicular lymphoma in situ genomic profiles »

Emilie Mamessier1, Joo Y. Song2, Franziska C. Eberle2, Svetlana Pack2, Charlotte Drevet1, Bruno Chetaille3, Ziedulla Abdullaev2, José Adelaïde4, Daniel Birnbaum4, Max Chaffanet4, Stefania Pittaluga2, Sandrine Roulland1, Andreas Chott5, Elaine S. Jaffe2*, Bertrand Nadel1*.

CIML, Genomic Instability and Human Hemopathies, Marseille, France.

1

National Institutes of Health, Laboratory of Pathology, Bethesda, USA.

2

IPC, Laboratory of Pathology, Marseille, France.

3

IPC, Molecular oncology, Marseille, France.

4

Wilhelminenspital, Institute of Pathology and Microbiology, Vienna, Austria

5

*

These two authors equally contributed to this work.

FCE current affiliation: Department of Dermatology, Eberhard Karls University, Tübingen, Germany

1

key words: B-cells, human, genomic instability, lymphomagenesis, follicular lymphoma, follicular lymphoma in situ, partial involvement by follicular lymphoma, duodenal follicular lymphoma

Correspondence

Elaine S. Jaffe Hematopathology Section, Laboratory of Pathology Center for Cancer Research, National Cancer Institute 10 Center Drive, Room 2B 42 Bethesda, MD 20892 USA @[email protected]

Bertrand Nadel Centre d’immunologie de Marseille-Luminy Laboratory Genomic Instability and Human Hemopathies Parc Scientific et Technologique de Luminy- Case 906 13288 Marseille France @[email protected]

Key points: -

Numerous alterations involving several relevant Follicular lymphoma (FL) (onco)genes are found in all FL precursor entities, including FLIS.

-

This study provides a first set of candidate gene alterations likely involved in the cascade of hits paving FL progression phases

2

Abstract

Follicular lymphoma pathogenesis is a multi-hit process progressing through the accumulation of numerous genetic alterations. Besides the hallmark t(14;18), it is still unclear which and when the other oncogenic hits occur. To address this issue, we performed high-resolution a-CGH on laser-capture microdissected cases of follicular lymphoma in situ (n=4), partial involvement by follicular lymphoma (n=4), and duodenal follicular lymphoma (n=4), assumed to represent the earliest stages in the evolution of follicular lymphoma. Reactive follicular hyperplasia (n=2), uninvolved areas from follicular lymphoma in situ lymph nodes, follicular lymphoma grade 1-2 (n=5) and 3A (n=5) were used as controls. Alterations involving relevant (onco)genes were found in all entities, but in a significantly lower proportion than follicular lymphoma. Nonetheless, while the number of alterations clearly assigns all these entities as precursors, the pattern of partial involvement by follicular lymphoma alterations was quantitatively and qualitatively closer to follicular lymphoma indicating significant selective pressure in line with its higher progression rate. Among the most noticeable alterations, we observed and validated the deletion of 1p36 and gains of the 7p and 12q chromosomes and related oncogenes, which include some of the most recurrent oncogenic alterations in overt follicular lymphoma (TNFRSF14, EZH2, MLL2). By further delineating distinctive and hierarchical molecular and genetic features of early follicular lymphoma entities, our analysis adds to the importance of applying appropriate criteria for differential diagnosis. It also provides a first set of candidate gene alterations likely involved in the cascade of hits paving the various progression phases of follicular lymphoma development.

3

Introduction

Follicular lymphoma (FL) has a widely variable clinical course, with most patients displaying an indolent form of the disease, associated with slow progression over many years. This progression can be followed through histological modifications, in both pattern and cell type, a fact that is reflected in the WHO grading system for FL (1). In this line of events, genomic studies of FL have shown an accumulation of genetic aberrations during the clinical course, correlating with histological transformation (2, 3). One of the earliest events in follicular lymphomagenesis is thought to occur in B-cell precursors developing in the bone marrow, giving rise to B-cells carrying the t(14;18)(q32;q21) translocation (4). However, we know that this translocation is also detected, at extremely low frequencies in B-cells from most healthy adults (5). Furthermore, the tracking of these cells over more than 10-years has demonstrated the persistence and evolution of a major dominant proliferating (14;18)-carrying B-cell clone within a given individual (6, 7). Two anecdotal reports also showed that the t(14;18)-B-cell clone represented in the tumor at diagnosis was already present several years before FL diagnosis (8, 9). Questions regarding the fate and site(s) of these FL-like B-cells remain open, particularly regarding their role as precursors of FL. The phenomenon of localization of FL cells to isolated germinal centers (GC) within a lymph node has been termed FL in situ (FLIS) (10) or intrafollicular neoplasia (1). This pattern may be seen with FL at other sites of disease, or the only manifestation of disease in some patients. In the true “in situ” lesion clusters of B-cells strongly positive for CD10 and BCL2 are localized to GCs in an otherwise reactive lymph node. FLIS often represents an incidental finding, in a lymph node biopsied for other reasons, and in such patients, if there is no other evidence of disease, the risk of progression is very low, in the order of 5% or less. A clinically similar phenomenon is duodenal FL (DFL), which usually presents as isolated mucosal polyps 4

with a low risk of dissemination (11). Both FLIS and DFL have been shown to have the BCL2/IGH translocation, but the presence of other genetic aberrations associated with FL is largely unexplored in these lesions. Given the very low rate of progression, it has been postulated that the BCL2 translocation might be the sole genetic aberration (12, 13). Histological and immunophenotypic criteria have been proposed to distinguish FLIS from partial involvement by FL (PFL). In PFL there is greater architectural distortion, and the diagnosis of lymphoma is more evident histologically. However, interestingly, patients with PFL appear to present with low stage disease, with a relatively low risk of progression and often a very prolonged disease-free interval with limited therapy (12). Thus, PFL may represent a bone fide early stage in the biological evolution of FL, and not just a limited state of lymph node replacement by a fully developed neoplasm (14). Advances in laser capture microdissection (LCM) now permit the isolation of even small numbers of cells representative of the clonotypic population (15), and improvements in genomic methods allow the analysis of genetic aberrations in formalin-fixed paraffin embedded (FFPE)-derived DNA. These new technologies now provide the unique opportunity to explore the genetic landscape of these early lesions, and to better understand their relationship to overt FL.

5

Methods Samples Most

cases

were

selected

from

FFPE

archived

specimens

submitted

to

the

Hematopathology Section at the National Cancer Institute (NCI) or the Institute of Pathology, at the Medical University of Vienna. Seven cases of FLIS, 5 cases of PFL and 5 DFL were identified based on previously published criteria (11, 12). Patients with FLIS and PFL had no other evidence of disease during the period of follow-up (12). These samples were compared to 5 cases of FL Grade1-2 and 5 FL Grade3A, included as controls of the most frequent alterations occurring in FL. Finally, reactive follicular hyperplasia sections (RFH, n=2) and laser micro-dissected lymphoid cells from uninvolved areas of FLIS#3 (FLIS background) were also hybridized and used as reference for normal cells (TableS1). Only cases with confirmed t(14;18) were included in the study. The Institutional Review Boards of the NCI and the Medical University of Vienna approved the study.

LCM, DNA extraction and QC To obtain sufficient enrichment of lesional cells positive for t(14;18) required for genome-wide array analysis, BCL2+ GCs were LCM from the selected FLIS/PFL/DFL/FL Grade1-2 cases. LCM was performed on hematoxylin stained slides with BCL2+ immunostained slides as a guide for involved follicles using a Leica LMD6000 (Leica Microsystems, Germany) (15). BCL2-negative GCs were microdissected from RFH, without evidence of FLIS. Tumor cell DNA was isolated from FL Grade3A without microdissection, due to the high tumor cells content (>75%). DNA was extracted using a QIAamp® DNA FFPE Kit. The DNA collected from the different cases ranged from 506 to 2820ng. With a calculation of 6.6pg of DNA/cell this amounts to a minimum of ~ 77,000 cells per case needed for appropriate hybridization. DNA integrity was controlled with the Agilent DNA1000 Kit®; the evaluation of amplifiable fragments sizes (from 100 to 800bp) was evaluated with a multiplex-PCR. Only samples displaying amplicon sizes >300bp passed QC for Array Comparative Genomic Hybridization 6

(a-CGH). Using these selective criteria, 4 FLIS, 4 PFL, 4 DFL, 10 FL, 2 RFH and 1 FLIS#3 background provided enough material and reached sufficient DNA quality to undergo a-CGH analysis.

a-CGH Genomic imbalances were analyzed using 244K a-CGH Microarrays (HU-244A); 2µg of sample and reference DNA were labeled with the ULS-labeling kit (Agilent Technologies) and hybridized on the Molecular Oncology platform (CRCM, Marseille), as previously described (16). Briefly, extraction data (log2 ratio) was performed from CGH-analytics, whereas normalized and filtered log2 ratios were obtained from the Feature Extraction™ software (version10.3; Agilent Technologies). Raw data were visualized as graphical figures within the Genomic Workbench v5.0 software. The frequency of alterations was computed for each probe locus as the proportion of samples showing an aberration therein. The gain and loss status for each probe was assigned using a conservative approach with a threshold of |0.4| on at least five consecutive probes across the genome. Alterations found by a-CGH were further validated by FISH on cases with sufficient material (Fig.S1).

See Supplemental Material and Methods for Samples details, FISH analyses (17), Transcription

profiles

public

datasets,

Sanger-sequencing

and

Statistics.

7

Results

Genomic alterations accumulate in early FL pathogenesis To gain insights into the kinetics of genomic aberrations accumulating during the early steps of FL pathogenesis, high-resolution a-CGH was performed on a retrospective collection of biopsies representing putative early stages of FL evolution (12). Genomic alterations were found in all samples from each early-FL group (FLIS: 4/4; PFL: 4/4; DFL: 4/4). Typical hybridization profiles are presented in Fig.1 (in extenso data are summarized in Table 1). As expected, no major genomic alteration (>700kb) was found in normal cells (RFH and FLIS background); in contrast, numerous lesions could be found in positive controls (FL grade 1-2 and 3A). Despite the low sampling, FL grade 1-2 and 3A alterations (Table 1 and Supplemental Fig.S2) were overall comparable to previously reported data (18-20), further validating our technological approach. Similarly, chromosomes most frequently affected in early-FL groups were overall similar to those recurrently observed in overt FL, including chromosome (chr)1 and chr18 (accounting for 17 and 33 % of alterations, respectively). Gain of the BCL2 region on chromosome 18/der14 was particularly apparent (33.5%), and confirmed by FISH (representative examples are shown in Supplemental Fig.S1A), stressing further the crucial role of BCL2 over-expression in early steps of FL development/maintenance (20). Although the mean number of alterations per sample increased with disease progression (from 2.5 in FLIS to 24.4 in FL3A, Fig.2A), alteration sizes were not significantly different in DFL, PFL and FL (Fig.2B), suggesting the occurrence of similar mechanisms of genomic instability. Interestingly the acquisition of very large chromosomal alterations (>10Mb) were rare in FLIS samples. As previously reported in FL (19), gains were more frequently observed than losses in all groups, but only gains were identified in FLIS (Supplemental Fig.S3). Altogether these data indicate that most early FL precursors, including FLIS, already stand in a progression phase in which increased genomic instability is at work.

8

Increased selective pressure with FL progression on a background of established genomic instability We next sought to determine if and how alterations observed in the various early-FL entities related to FL progression and maintenance.

FLIS A total of 259 annotated genes (listed in Supplemental Table 2) were spread over a total of ~52 Mb in 10 amplified regions from the 4 FLIS samples. Among those, only few could be identified as relevant hits recurrently observed in FL. Twelve were known oncogenes (including BCL2 (4), RUNX1 (21), and KDSR (22), and were also found translocated and/or amplified in 100%, 70% and 30% of our series of FL samples, respectively (Supplemental Table S3). Four amplified genes (Tox, BACH2 (23) AFF3 (24), and EBF1 (25), Supplemental Fig.S4) were functionally related to the GC reaction, and were also found amplified in 100%, 90%, 80%, and 70% of FL, respectively. Notably, gain of AFF3 was recurrently found in 2 distinct FLIS samples. Although their functional relevance in FL lymphomagenesis remain to be directly demonstrated, some of the amplified genes were also found to be upregulated compared to normal B-cell subsets in transcriptomes from FL (public microarray dataset GSE12195)(Supplemental Fig.S5 and Table S3). However, only 12 (4%) of the amplified genes were identified oncogenes, while 32 (13%) were gains of genes with tumor suppressive functions (Fig.3A and Supplemental Table S2-S3). Furthermore, and contrary to other entities, no loss of genes with tumor suppressive function could be identified in absence of loss in FLIS. This suggests that a significant fraction of the amplified genes in FLIS might be passenger alterations, rather than selected alterations as observed in more advanced entities.

9

PFL A total of 1128 genes were spread over ~268 Mb in 13 amplified regions from the 4 PFL samples; 370 genes were spread over a total of ~85 Mb in 3 losses (listed in Supplemental Table 2). A subset of the alterations was validated by FISH in our cohort (loss of 1p36, 7p, 12q, and BCL2 amplifications, Supplemental Fig S1A-D). The major fraction of amplified genes (99%) was shared with genes also found to be amplified in FL1/2 and/or FL3A (Fig.3B), suggesting selective pressure for their maintenance during FL progression. In agreement with this, most of the shared genes were functionally involved in cell proliferation, expression, transcription, and differentiation. Sixty-seven of them (~6%) were known oncogenes, 6 of which (BCL2, RUNX1, NACA (26), CARD11 (27), ELN (28) & IKZF1 (29)) previously identified as major players in leukemia/lymphoma oncogenesis (Supplemental Table S3). In line with the functional relevance of these candidate hits in FL development, RUNX1, AFF3, and BACH2 were recurrently amplified in FLIS, and/or PFL, and RUNX1, NACA, and IKZF1 were transcriptionally highly up-regulated in overt FL compared to their normal B-cell counterparts (Supplemental Fig.S5 and Table S3). Shared amplified genes in PFL also comprised histone modifiers such as MLL2, EZH2 (30), ARID2 (31), and HDAC7 (32) recently identified as some of the most frequent altered function events of FL pathogenesis (3, 33). Similarly to FL1-2 and 3A, less annotated genes (370) were found in the deleted areas of PFL. Losses included 27 tumor suppressors (TS) (but also 18 oncogenes), and the corresponding functional pathways (cell death, apoptosis, necrosis, cell migration) appeared coherent with loss-of-function in lymphomagenesis. However, and in contrast to amplified regions, only few of these genes were shared with FL1-2 and/or FL3A (60/370 ~16%, Fig.3B). Among these, shared losses comprised deleted genes on chromosome 1p, a chromosome classically affected in FL (34). Several genes related to cytokines and cytokine receptors appeared as potent early loss-of-function hits (Supplemental Tables S2 and S3). This included TNF-family members such as TNFRSF9 and TNFRSF25, and more particularly TNFRSF14, which has recently been identified as one of the most frequent secondary genetic abnormalities in FL (>65%) (35). In this regard, we 10

confirmed that TNFRSF14 was a site of frequent somatic mutations even in early-FL samples (Supplemental Table S4), suggesting early involvement of this molecule in FL pathogenesis. Ten genes in shared losses were recognized tumor suppressor genes (Supplemental Tables S2 and S3), including TP73 (36), also deleted in DFL, 50% of FLs, and transcriptionally downregulated in large cohorts of FL (Supplemental Fig.S6). While several of the amplified oncogenes have been previously associated with FL, to our knowledge, only few of the deleted tumor suppressors have been yet reported as recurrent events or functionally involved in FL pathogenesis and/or lymphomagenesis, and we cannot exclude that some might correspond to passenger alterations. Overall, although the global number of alterations remains quantitatively and qualitatively much lower than in FL (Fig.3A), the large proportion of gains maintained throughout FL progression and the occurrence of functionally relevant gains and losses corresponding to frequent FL hits indicate that PFL already sustained significant selective pressure, in agreement with its reported high progression rate to FL.

DFL A total of 1021 genes were localized in 12 amplified regions (~ 390 Mb) from DFL samples (listed in Supplemental Table S2). While the biological functions of these genes were globally relevant to oncogenesis, tumor suppressors (n=73) were more frequently found than oncogenes (n=45) in amplified regions (Fig.3A). A similar trend was observed for the 348 genes located in 6 deleted regions (~ 40 Mb), with 32 tumor suppressors and 14 oncogenes. In contrast to PFL, a large fraction of both gains and losses were not found in FL grades 1-2 and/or 3A (45 to 83%, respectively, Fig.3C). Among the shared genes, a limited set of relevant amplified oncogenes (BCL2, BCL6FGFR1, EIF4A2 and TFRC) and deleted TS (PTEN, FAS, TP73) could nevertheless be identified; some of them were also shared with the other FLIS/PFL entities (BCL2, BCL6, TP73, Supplemental Table S3), previously

11

reported altered in non-Hodgkin lymphoma (BCL2, BCL6, TFRC, TP73)(33, 37) and/or highly increased transcriptionally in FL samples compared to normal B-cell subsets (EIF4A2, TFRC, Supplemental Fig.S5C and Table S3). Furthermore, recurrent deletion of chromosome 1p was again observed, encompassing the same TNFR family members previously identified in PFL and FL, as well as high mutation load of TNFRSF14 exons (Supplemental Fig.S6B, Table S4). Altogether, this suggests that while DFLs display increased genomic instability and bear some of the classical hits of FL genesis (Supplemental Table S4 & S5), the absence of strong selective forces might have (yet?) limited their evolution to malignant transformation.

12

Discussion Early FL entities have been defined thus far histologically, and beside t(14;18), very little information is available on their molecular status (13, 38). Using high-resolution a-CGH analysis, we provide here the first comprehensive genome-wide genetic viewpoint of FLIS, PFL and DFL in otherwise “healthy” individuals. We first clearly show the presence of genomic alterations in all 3 entities. Considering the extensive subclonal nature of FL and pre-FL (39, 40), the early events uncovered here probably vastly underestimate the multitude of other genetic lesions present in minor and/or non-selected subclones. Furthermore, considering that next generation sequencing data have recently revealed that smaller alterations -including point mutations- are the most frequent events in cancer cells (33), it is highly likely that such events have also extensively accumulated in FLIS, PFL and DFL. In line with this possibility, we found that small copy number variations (between 100 and 700kb) also tended to significantly increase between pre-FL and FL entities, with a dramatic leap between FL grade 1-2 and 3A (data not shown and Supplemental Fig.S7 and S1A). Next generation sequencing on a set of fresh prospective FLIS, PFL, and DFL samples will be necessary to precisely uncover the extent of genomic instability in these entities. Nonetheless, our aCGH analysis first revealed that FLIS exhibited much greater genetic complexity than anticipated, suggesting that increased genomic instability is already at work in these cells, and confirming their status of “early lesions”. However, the diverse functional relevance of the altered genes, and the low proportion of those as recurrent FL hits, suggest that only few could be major actors of malignant progression. As such, FLIS seemingly stand in a crossroad between increased genomic instability and low selective pressure for malignant transformation, consistent with the observation that they rarely progress to overt FL. This adds to the notion that the oncogenic network required for full malignant transformation proceeds through a complex Darwinian-like process of stepwise, multi-hit events of selection/counter-selection.

13

From both the genetic and functional perspectives, FLIS clearly appears as a less advanced entity than PFL, supporting the validity of the recently proposed criteria for refined histological stratification (12). The mean number and size of alterations was higher in PFL than in FLIS. The major fraction of amplified genes in PFL (99%) was also found amplified in FL1-2 and/or FL3A and comprised several known FL oncogenes, suggesting selective pressure for their maintenance during FL progression. Nevertheless, it still appears that PFL are genetically not equivalent to overt FL, both in terms of the overall lower number of alterations (3-fold and 6–fold lower compared to FL grades 1-2 and 3A, respectively) and of the small overlap of shared genes in losses (~16%), showing (in contrast to gains) incomplete functional evolution to FL. The observed dichotomy between gains and losses is intriguing, and might suggest a more deleterious/irreversible effect of large chromosomal losses in early stages of FL progression (deletion of amplified regions/sub-regions being possible, but not the reverse), leading to their late acquisition in FL (41). Overall, our data clearly demonstrate that when rigorously applying histological criteria for their identification (12), PFL generally do not constitute partial colonization of the lymph node (LN) by overt FL, but rather represent an earlier stage of tumor evolution (40). Our definition of PFL used for case selection excludes patients with known FL at another site. In the presence of concurrent FL and/or FL by history, we cannot rule out that partial colonization of lymph nodes by overt FL might histologically resemble/be undistinguishable from PFL. Altogether, our results add to the hierarchical model, in which FLIS, PFL, FL grade 12, and FL grade 3A constitute a progression continuum (Supplemental Fig.S8) (42). At the beginning of this spectrum, the relationship between FLIS and FL-like cells remains unclear to date. The report of a circulating t(14;18)+ counterpart in a FLIS case provided a first direct evidence of such an affiliation (9, 43). The extensive dissemination of FL-like clones in lymphoid organs (including bone marrow) and blood, the important dynamics of clonal evolution in these cells, the presence of FL hallmarks of selective pressure (such as the allelic-paradox allowing to maintain a sIgM in spite of active AID activity), and the frequent

14

imprints of AID-mediated genomic instability further argue for a large overlap between these two entities ((39); Sungalee et al. submitted). Our data also revealed that a number of candidate-hits, some of which are frequently seen in FL and/or assumed to play a major role in FL pathogenesis, may be acquired very early in FL genesis. This includes further gain of BCL2 on Chr18 or der14, stressing the crucial role of BCL2 overexpression in early steps of FL development/maintenance. Histone modifiers such as CREBBP, EP300, and EZH2 also recently emerged as some of the most frequent alterations in FL, and it has been proposed that such alterations occur early on in FL natural history (3, 40). In line with this proposal, we show here recurrent gain of MLL2 and EZH2 in both PFL and DFL, and increased mutation rate in CREBBP (Supplemental Table S5). Regarding the recent description of EZH2 as a major player in GC formation, and of the role of EZH2 gain-of-function mutations in enhancing lymphomagenesis (44), we found that a fraction of the alterations in FLIS and PFL functionally related to GC reaction biology, suggesting that selective forces in FLIS might initially converge on events contributing to the GC reaction and/or GC retention. Consistent with the microenvironment dependency of FL, members of the TNFR family located on the frequently deleted 1p36 region also emerge among tangible early candidate hits and were lost both in PFL and DFL. TNFRSF14, in particular has recently been identified as one of the most frequent secondary genetic abnormality in FL (>65%) (35), and TNFRSF9 (4-1BB) deficient mice are predisposed to develop GC-derived (follicular-like) B-cell lymphoma (45). Other members of the TNFR family located on Chr8 and Chr18 (notably TNFRSF11A (46), a major activator of the NF-kappaB and MAPK8/JNK pathways) were amplified in one FLIS sample, further adding to the potential relevance of this gene family. Interestingly, some of the affected regions in PFL and DFL, and particularly the deletion of the 1p region, were previously identified as being significantly associated with poor survival and/or transformation to DLBCL (34). This might suggest that gene deregulations associated with poor prognostic factors may also be

15

acquired early on, but that their adverse effect might be unleashed only at more advanced stages and/or under therapeutic pressure. With confinement to the duodenum and little evidence for FL progression, DFL appears histologically and clinically as a separate entity (11, 47) and it is as yet unclear how it integrates in the FLIS-PFL-early FL scheme. DFL share the t(14;18) with other variants of nodal FL. The propensity to remain localized to the mucosa may relate to the site in which the relevant antigen is first encountered, a feature that these lesions may share with MALT lymphoma. DFL might initiate from the same early founder t(14;18)-only precursor exiting the bone marrow (BM), or from a genetically more advanced one (i.e. that has already experienced antigenic challenge and GC transit before settling in the specific extra nodal environment) (11, 48). From a genetic viewpoint, DFLs appear at increased risk of genomic instability with a large number of already acquired genomic alterations (comparable to PFL), including some of the classical hits of FL genesis. In contrast to PFL, however, only half of the amplified genes were shared with FL (or PFL), suggesting still limited selection pressure to malignant pathways. It has been previously shown that DFL are devoid of AID expression (49). This either suggests that the genomic instability observed is not AID-dependent, or that the observed alterations have been acquired elsewhere in presence of AID (potentially in the frame of a classical GC reaction) before colonization of duodenal GCs. By limiting further genomic instability and selection potential, this scenario might partly explain why this entity rarely evolves to overt malignancy. Interestingly, the low propensity of DFL to progress to overt FL despite the presence and combination of several oncogenic alterations adds to the evidence that such hits are not sufficient for transformation, and that extrinsic factors (such as successive immunological challenges in GCs) might play a key role in the selection process (39). DFL, FLIS and PFL show very distinct progression rates when left untreated (~5%, ~5%, and ~53%, respectively). In the current debate on the clinical management of early and asymptomatic FL patients with low tumor mass, none from a small series of PFL patients

16

treated with Rituximab or radiotherapy progressed to overt FL after 7 years of follow-up (12). By further delineating distinctive and hierarchical molecular and genetic features of FLIS, DFL and PFL, our analysis adds to the critical importance of applying appropriate criteria for differential diagnosis. It also provides a first set of candidate gene alterations likely involved in the cascade of hits paving the early progression phases of FL development, and which could be explored in prospective studies as innovative therapeutic targets.

17

Authorship and disclosures JYS, FCE, SPi, performed LCM and DNA extractions. EM, JA, DB and MC designed and performed the a-CGH assay and analysis. BC: provided FLIS samples and performed IHC staining. EM, and CD performed additional molecular assays and IHC staining. SPa and ZA performed FISH analysis. EM, SR, AC, ESJ and BN designed the study and EM, ESJ and BN wrote the manuscript. The authors have no conflicts of interest to declare.

Acknowledgements This work was funded by grants from the Institut National du Cancer (INCa), the Association pour la Recherche sur le Cancer (ARC), a MedImmune’s Strategic Collaboration to Fund and Conduct Medical Science Research program, Cancéropôle PACA, and institutional grants from the Institut National de la Santé et de la Recherche Médicale (INSERM), the Centre National de la Recherche Scientifique (CNRS) and Intramural Research Program of the Center for Cancer Research, National Cancer Institute, National Institutes of Health. EM and CD were recipients of fellowships from the Institut National du Cancer (INCa), Cancéropôle PACA, and MedImmune’s Strategic Collaboration to Fund and Conduct Medical Science Research program. We thank Shelley Hoover and Dr. Mark Simpson for their support.

18

References 1.

Campo E, Swerdlow SH, Harris NL, Pileri S, Stein H, Jaffe ES. The 2008 WHO

classification of lymphoid neoplasms and beyond: evolving concepts and practical applications. Blood. 2011;117(19):5019-32. 2.

Davies AJ, Rosenwald A, Wright G, Lee A, Last KW, Weisenburger DD, et al.

Transformation of follicular lymphoma to diffuse large B-cell lymphoma proceeds by distinct oncogenic mechanisms. British journal of haematology. 2007;136(2):286-93. 3.

Morin RD, Mendez-Lago M, Mungall AJ, Goya R, Mungall KL, Corbett RD, et al.

Frequent mutation of histone-modifying genes in non-Hodgkin lymphoma. Nature. 2011;476(7360):298-303. 4.

Tsujimoto Y, Cossman J, Jaffe E, Croce CM. Involvement of the bcl-2 gene in human

follicular lymphoma. Science. 1985;228(4706):1440-3. 5.

Limpens J, Stad R, Vos C, de Vlaam C, de Jong D, van Ommen GJ, et al.

Lymphoma-associated translocation t(14;18) in blood B cells of normal individuals. Blood. 1995;85(9):2528-36. 6.

Liu Y, Hernandez AM, Shibata D, Cortopassi GA. BCL2 translocation frequency rises

with age in humans. Proc Natl Acad Sci U S A. 1994;91(19):8910-4. 7.

Roulland S, Lebailly P, Lecluse Y, Heutte N, Nadel B, Gauduchon P. Long-term

clonal persistence and evolution of t(14;18)-bearing B cells in healthy individuals. Leukemia. 2006;20(1):158-62. 8.

Weigert O, Kopp N, Lane AA, Yoda A, Dahlberg SE, Neuberg D, et al. Molecular

ontogeny of donor-derived follicular lymphomas occurring after hematopoietic cell transplantation. Cancer Discov. 2012;2(1):47-55. 9.

Bretherick KL, Bu R, Gascoyne RD, Connors JM, Spinelli JJ, Brooks-Wilson AR.

Elevated circulating t(14;18) translocation levels prior to diagnosis of follicular lymphoma. Blood. 2010;116(26):6146-7. 10.

Cong P, Raffeld M, Teruya-Feldstein J, Sorbara L, Pittaluga S, Jaffe ES. In situ

localization of follicular lymphoma: description and analysis by laser capture microdissection. Blood. 2002;99(9):3376-82. 11.

Schmatz AI, Streubel B, Kretschmer-Chott E, Puspok A, Jager U, Mannhalter C, et al.

Primary follicular lymphoma of the duodenum is a distinct mucosal/submucosal variant of follicular lymphoma: a retrospective study of 63 cases. J Clin Oncol. 2011;29(11):1445-51. 12.

Jegalian AG, Eberle FC, Pack SD, Mirvis M, Raffeld M, Pittaluga S, et al. Follicular

lymphoma in situ: clinical implications and comparisons with partial involvement by follicular lymphoma. Blood. 2011;118(11):2976-84.

19

13.

Bonzheim I, Salaverria I, Haake A, Gastl G, Adam P, Siebert R, et al. A unique case

of follicular lymphoma provides insights to the clonal evolution from follicular lymphoma in situ to manifest follicular lymphoma. Blood. 2011;118(12):3442-4. 14.

Adam P, Katzenberger T, Eifert M, Ott MM, Rosenwald A, Muller-Hermelink HK, et al.

Presence of preserved reactive germinal centers in follicular lymphoma is a strong histopathologic indicator of limited disease stage. The American journal of surgical pathology. 2005;29(12):1661-4. 15.

Eberle FC, Hanson JC, Killian JK, Wei L, Ylaya K, Hewitt SM, et al. Immunoguided

laser assisted microdissection techniques for DNA methylation analysis of archival tissue specimens. J Mol Diagn. 2010;12(4):394-401. 16.

Adelaide J, Finetti P, Bekhouche I, Repellini L, Geneix J, Sircoulomb F, et al.

Integrated profiling of basal and luminal breast cancers. Cancer Res. 2007;67(24):11565-75. 17.

Pack SD, Zhuang Z. Fluorescence in situ hybridization : application in cancer

research and clinical diagnostics. Methods in molecular medicine. 2001;50:35-50. 18.

Cheung KJ, Delaney A, Ben-Neriah S, Schein J, Lee T, Shah SP, et al. High

resolution analysis of follicular lymphoma genomes reveals somatic recurrent sites of copyneutral loss of heterozygosity and copy number alterations that target single genes. Genes Chromosomes Cancer.;49(8):669-81. 19.

Cheung KJ, Shah SP, Steidl C, Johnson N, Relander T, Telenius A, et al. Genome-

wide profiling of follicular lymphoma by array comparative genomic hybridization reveals prognostically significant DNA copy number imbalances. Blood. 2009;113(1):137-48. 20.

Leich E, Salaverria I, Bea S, Zettl A, Wright G, Moreno V, et al. Follicular lymphomas

with and without translocation t(14;18) differ in gene expression profiles and genetic alterations. Blood. 2009;114(4):826-34. 21.

De Braekeleer E, Douet-Guilbert N, Morel F, Le Bris MJ, Ferec C, De Braekeleer M.

RUNX1 translocations and fusion genes in malignant hemopathies. Future Oncol. 2011;7(1):77-91. 22.

Czuchlewski DR, Csernus B, Bubman D, Hyjek E, Martin P, Chadburn A, et al.

Expression of the follicular lymphoma variant translocation 1 gene in diffuse large B-cell lymphoma correlates with subtype and clinical outcome. American journal of clinical pathology. 2008;130(6):957-62. 23.

Alinikula J, Nera KP, Junttila S, Lassila O. Alternate pathways for Bcl6-mediated

regulation of B cell to plasma cell differentiation. Eur J Immunol. 2011;41(8):2404-13. 24.

Hiwatari M, Taki T, Taketani T, Taniwaki M, Sugita K, Okuya M, et al. Fusion of an

AF4-related gene, LAF4, to MLL in childhood acute lymphoblastic leukemia with t(2;11)(q11;q23). Oncogene. 2003;22(18):2851-5.

20

25.

Vilagos B, Hoffmann M, Souabni A, Sun Q, Werner B, Medvedovic J, et al. Essential

role of EBF1 in the generation and function of distinct mature B cell types. J Exp Med. 2012;209(4):775-92. 26.

Hammerle K, Shayan P, Niemeyer CM, Flotho C. Expression analysis of alpha-NAC

and ANX2 in juvenile myelomonocytic leukemia using SMART polymerase chain reaction and "virtual Northern" hybridization. Cancer Genet Cytogenet. 2003;142(2):149-52. 27.

Chan W, Schaffer TB, Pomerantz JL. A quantitative signaling screen identifies

CARD11 mutations in the CARD and LATCH domains that induce Bcl10 ubiquitination and human lymphoma cell survival. Mol Cell Biol. 2013;33(2):429-43. 28.

Bousquet M, Broccardo C, Quelen C, Meggetto F, Kuhlein E, Delsol G, et al. A novel

PAX5-ELN fusion protein identified in B-cell acute lymphoblastic leukemia acts as a dominant negative on wild-type PAX5. Blood. 2007;109(8):3417-23. 29.

Zhang J, Ding L, Holmfeldt L, Wu G, Heatley SL, Payne-Turner D, et al. The genetic

basis of early T-cell precursor acute lymphoblastic leukaemia. Nature. 2012;481(7380):15763. 30.

Lohr JG, Stojanov P, Lawrence MS, Auclair D, Chapuy B, Sougnez C, et al.

Discovery and prioritization of somatic mutations in diffuse large B-cell lymphoma (DLBCL) by whole-exome sequencing. Proc Natl Acad Sci U S A. 2012;109(10):3879-84. 31.

Li M, Zhao H, Zhang X, Wood LD, Anders RA, Choti MA, et al. Inactivating mutations

of the chromatin remodeling gene ARID2 in hepatocellular carcinoma. Nature genetics. 2011;43(9):828-9. 32.

Lemercier C, Brocard MP, Puvion-Dutilleul F, Kao HY, Albagli O, Khochbin S. Class II

histone deacetylases are directly recruited by BCL6 transcriptional repressor. J Biol Chem. 2002;277(24):22045-52. 33.

Kridel R, Sehn LH, Gascoyne RD. Pathogenesis of follicular lymphoma. The Journal

of clinical investigation. 2012;122(10):3424-31. 34.

Cheung KJ, Johnson NA, Affleck JG, Severson T, Steidl C, Ben-Neriah S, et al.

Acquired TNFRSF14 mutations in follicular lymphoma are associated with worse prognosis. Cancer Res. 2010;70(22):9166-74. 35.

Launay E, Pangault C, Bertrand P, Jardin F, Lamy T, Tilly H, et al. High rate of

TNFRSF14 gene alterations related to 1p36 region in de novo follicular lymphoma and impact on prognosis. Leukemia. 2012;26(3):559-62. 36.

Duhoux FP, Ameye G, Lambot V, Herens C, Lambert F, Raynaud S, et al.

Refinement of 1p36 alterations not involving PRDM16 in myeloid and lymphoid malignancies. PloS one. 2011;6(10):e26311. 37.

Morishige S, Oku E, Takata Y, Kimura Y, Arakawa F, Seki R, et al. A case of 8p11

myeloproliferative syndrome with BCR-FGFR1 gene fusion presenting with trilineage acute 21

leukemia/lymphoma, successfully treated by cord blood transplantation. Acta haematologica. 2013;129(2):83-9. 38.

Carbone A. Does "in situ lymphoma" occur as a distinct step in the development of

nodular lymphocyte-predominant Hodgkin lymphoma? Cancer. 2012;118(1):15-6. 39.

Roulland S, Faroudi M, Mamessier E, Sungalee S, Salles G, Nadel B. Early steps of

follicular lymphoma pathogenesis. Adv Immunol. 2011;111:1-46. 40.

Green MR, Gentles AJ, Nair RV, Irish JM, Kihira S, Liu CL, et al. Hierarchy in somatic

mutations arising during genomic evolution and progression of follicular lymphoma. Blood. 2013;121(9):1604-11. 41.

Hoglund M, Sehn L, Connors JM, Gascoyne RD, Siebert R, Sall T, et al. Identification

of cytogenetic subgroups and karyotypic pathways of clonal evolution in follicular lymphomas. Genes Chromosomes Cancer. 2004;39(3):195-204. 42.

Krijgsman O, Gonzalez P, Balague Ponz O, Roemer MG, Slot S, Broeks A, et al.

Dissecting the grey zone between follicular lymphoma and marginalzone lymphoma using morphological and genetic features. Haematologica. 2013. [Epub ahead of print] 43.

Cheung MC, Bailey D, Pennell N, Imrie KR, Berinstein NL, Amato D, et al. In situ

localization of follicular lymphoma: evidence for subclinical systemic disease with detection of an identical BCL-2/IGH fusion gene in blood and lymph node. Leukemia. 2009;23(6):1176-9. 44.

Beguelin W, Popovic R, Teater M, Jiang Y, Bunting KL, Rosen M, et al. EZH2 Is

Required for Germinal Center Formation and Somatic EZH2 Mutations Promote Lymphoid Transformation. Cancer cell. 2013;23(5):677-92. 45.

Middendorp S, Xiao Y, Song JY, Peperzak V, Krijger PH, Jacobs H, et al. Mice

deficient for CD137 ligand are predisposed to develop germinal center-derived B-cell lymphoma. Blood. 2009;114(11):2280-9. 46.

Compagno M, Lim WK, Grunn A, Nandula SV, Brahmachary M, Shen Q, et al.

Mutations of multiple genes cause deregulation of NF-kappaB in diffuse large B-cell lymphoma. Nature. 2009;459(7247):717-21. 47.

Takata K, Okada H, Ohmiya N, Nakamura S, Kitadai Y, Tari A, et al. Primary

gastrointestinal follicular lymphoma involving the duodenal second portion is a distinct entity: a multicenter, retrospective analysis in Japan. Cancer Sci. 2011;102(8):1532-6. 48.

Bende RJ, Smit LA, Bossenbroek JG, Aarts WM, Spaargaren M, de Leval L, et al.

Primary follicular lymphoma of the small intestine: alpha4beta7 expression and immunoglobulin configuration suggest an origin from local antigen-experienced B cells. Am J Pathol. 2003;162(1):105-13. 49.

Takata K, Sato Y, Nakamura N, Tokunaka M, Miki Y, Yukie Kikuti Y, et al. Duodenal

follicular lymphoma lacks AID but expresses BACH2 and has memory B-cell characteristics. Mod Pathol. 2013;26(1):22-31. 22

23

Table 1. Results from a-CGH analysis in FL3A, FL 1-2, PFL, DFL and FLIS samples compared to negative controls (2 RFH and 1 FLIS background).

Samples

Nb of samples

FL3A FL1/2 PFL DFL FLIS Controls

5 5 4 4 4 3

Samples Total nb of with Alterations alterations 5 5 4 4 4 0

122 61 16 16 10 0

Mean Recurrency* Mean Alterations Alteration Nb (%) size / sample 47.0 49.0 25.0 33.0 27.0 0.0

24.4 12.2 4.0 4.0 2.5 0.0

2088 3366 3004 2242 845 0

* Recurrency represents the percentage of alterations that are found in at least two samples.

24

kbp kbp kbp kbp kbp kbp

Figure legends

Figure 1. CGH-a profiles by entities from laser capture microdissection (LCM) of germinal center regions. To enrich for cells of interest, LCM of BCL2+ GCs from the selected FLIS/PFL/DFL cases was performed on hematoxylin stained slides with BCL2+ immunostained slides as a guide for involved follicles. LCM was also performed on FL 1-2 biopsies to select for neoplastic follicles. Lymphoid cells from cases of FL 3A containing >75% tumor cells were utilized without microdissection, due to the high tumor cell content.

Finally, reactive GCs were

captured by LCM from cases of RFH and utilized as negative controls. DNA from all of the relevant lesions above was hybridized on 244k a-CGH chips. Representative profiles of alterations found in the different entities are shown. Controls samples (RFH) are in grey, early-FL and FL samples are in colors).

Figure 2. Quantitative and qualitative characteristics of alterations found in the various progression stages of follicular lymphomagenesis. A. Number of alterations observed in each sample, classified by entity. B. Length (in base pairs) of each alteration observed in the various subtypes. * : p < 0.05 Controls represent reactive GCs isolated by LCM from RFH, and FLIS background as defined in the methods.

Figure 3. Summary of alterations observed in early-FL and FL grade 1-2 and FL grade 3A. A. Oncogene (Red) and Tumor suppressor (Blue) genes located in the gains and losses, represented by entities. The total number of altered genes in each category is indicated at the right of the bars, and some of the most relevant genes are listed above the bars. B. Venn

25

diagram showing the number of genes found in the gained and lost regions of PFL, FL grade 1-2 and FL grade 3A samples. The dotted area indicates the percentage of genes located in amplified or lost regions of the PFL samples, which are also amplified or lost in FL grade 1/2 samples. An example of a region (gain) shared by one PFL (in blue) and several FL samples (in grey), containing several genes (notably the NACA oncogene) is shown. RFH controls are in black. C. Venn diagram showing the number of genes found in the gained and lost regions of DFL, FL grade 1-2 and FL grade 3A samples. The dotted area indicates the percentage of genes located in amplified or lost regions of the DFL samples, which are also amplified or lost in FL 1-2 samples. An example of one alteration observed in a DFL sample. DFL samples are in purple, FL grade 1-2 and FL grade 3A are in grey, RFH in black.

26

Control CGH profiles

CGH profiles by entity

Non-tumoral controls-K18

FLIS - K8 -1

LCM

0

+1 FL3A controls- K8

DFL - K18 -1

LCM

0

+1 PFL- K13 -1

LCM

0

+1 FL Grade 1/2 - K8 -1

0

Figure 1

+1

LCM

PF L

FL 1/ 2 FL 3A

FL 1/ 2

FL 3A

10 5

PF L

10 6

DF L

10 7

D FL

10 8

FL IS

10 9

FL IS

Co nt ro ls

B.

C on tr ol s

Alteration length (bp)

Nb of Alterations/ per sample

Figure 2

*

A. 55

50

25

20

15

10

5

0

*

Figure 3

A.

FLIS Oncogenes

AFF3 BCL2 BACH2 CARD11 EGFR ELN EPHA1 EZH2 HOXA11 IKZF1 KDSR MALT1 NCA RUNX1 WNT1 MDM2 HDAC7 HOXA5

Tumor suppressors

Nb of genes

500

AFF3 BCL2 BACH2 KDSR RUNX1 TNFRSF10A TNFRSF10B

BCL2 KDSR MALT1 PIM2 TNFRSF10A TNFRSF10B

FL 1/2

PFL

DFL

NFKB2 PAX2 TLX1 DFFB FAS PTEN TP73

CDC25A PARK7 DFFB RB1 TP73

AFF3 ALK BCL2 BACH2 BCL11A CARD11 EGFR EPHA1 EZH2 HOXA11 IKZF1 KDSR KRAS MALT1 MYC REL RUNX1 WNT1 HDAC7 HOXA5 MDM2 MLH1 TNFRSF10A

FL 3A

EPHA2 PARK7 RUNX3 SERPINB2 DFFB EPHB2 TP73

400 300 100

45 73

12 32

0

Gains

B.

Losses

Gains

67 59

14 32

Losses

Gains

Gains

-1

FL 1/2

PFL

99%

827

349

5 1

773

310

8 0

1755

677

DFL 516 0

53

Losses

0

+1

16%

155

52

256

2476

PFL FL

FL 3A

FL 1/2

452

Gains

798

FL 3A

55%

FL 1/2

PFL PFL

Losses

Typical profiles

LOSSES

GAINS

173 255

16 47

18 27

Losses

ALK CBL CD79a CDC42 EP300 EPHA2 ETS1 GATA2 MALT1 MAX MLL MTOR TLX1 ARID3B BAX DFFB EPHB2 FAS PTEN TP73 TP53

177 237

177 227

200

C.

AFF3 AKT1 ALK BCL2 BACH2 BCL11A CARD11 EGFR EZH2 HOXA11 IKZF1 KDSR MLH1 MYC NOTCH REL RHOA RUNX1 SETD2 WNT1 HDAC7 HOXA5 MDM2 MLH1 TNFRSF10A

660

DFL 246

Chr12

FL 1/2

16%

8 42

52

155 256

673

756

FL 3A

FL 3A

DFL FL

Chr10

Supplemental Material and Methods Online Supplemental Revised tables 1-6 uploaded as Excel files Samples Most

cases

were

selected

from

FFPE

archived

specimens

submitted

to

the

Hematopathology Section at the National Cancer Institute (NCI) for consultation, from 1999 through 2011. Seven cases of FLIS and 5 cases of PFL (all Grade 1-2), with suitable material were identified based on previously published criteria (12). Patients with FLIS and PFL had no other evidence of disease during the period of follow-up; 3 of 7 FLIS (#1, #2, and #4) were included in a previously published series (12). Five cases of DFL were obtained from the archival files of the Institute of Pathology, Medical University Vienna, Austria, collected from January 1997 to April 2008, as previously reported by Schmatz et al (11). The principal cohort of FLIS, PFL and DFL was compared to 5 cases of FL Grade 1-2 and 5 cases of FL Grade 3A included as controls of the most frequent alterations occurring in low grade and presumptively more advanced FL. Finally, reactive follicular hyperplasia sections (RFH, n=2) and laser micro-dissected lymphoid cells from uninvolved areas of one FLIS (FLIS background) were also hybridized and used as reference for normal (non-neoplastic) cells (Supplemental Table 1). Only cases with confirmed t(14; 18) were included in the study. The Institutional Review Boards of the National Cancer Institute and the Medical University of Vienna approved the study.

FISH validation at the 1p36, 7p and 12q loci, and FISH analysis of the BCL2 gene rearrangements on paraffin-embedded tissue sections Deletion in the 1p36, amplifications in the 7p and 12q loci, and BCL2 gene rearrangement status were assessed using Fluorescence In Situ Hybridization (FISH). For BCL2 interphase FISH analysis, BCL2 Split Signal FISH DNA Probe, (DAKO, Denmark, Code Y5407) was used. FISH for 1p36, 7p12, 12q 13.3-14.1 chromosomal regions was performed using Vysis probes Vysis LSI 1p36 / LSI 1q25 Dual-Color Probe covering MEGF6 and TP73 genes for Chr. 1p36 region, Vysis LSI EGFR SpectrumOrange/ CEP 7 SpectrumGreen Probe (Abbott Molecular, Chicago, IL) for Chr 7p12, and custom-made probe for Chr. 12q13.3-14.1 region

using RP11-571M6 BAC clone (that includes CDK4 locus and 11 more genes) labeled with Red-5-ROX dUTP combined with the control centromeric probe for chromosomes 12 labeled with Green-5-Fluorescein ( Empire Genomics, Buffalo, NY). FISH assays were performed on 5 micron Formalin-Fixed-Paraffin-Embedded (FFPE) tumor sections using laboratory standardized protocol with slight modifications (17). In brief, 5 µm thick sections of FFPE tissue blocks were de-paraffinized, and rehydrated. Antigen retrieval was performed with IHC-Tek Epitope Retrieval Solution (IHC World, Woodstock, MD) with steaming for 25 min. After cooling, slides were subjected to 50 µkg/ml pepsin treatment at 37oC, rinsed in PBS solution followed by dehydration in ethanol series. Co-denaturation of the probe and target DNA at 73oC in HYBrite (Abbott Molecular, Chicago, IL) for 5 min was followed by overnight hybridization at 37oC. The next day, slides were washed at 72oC in 0.4X SSC/0.3% Tween-20 for 2 min and then in 2X SSC/0.1% Tween-20 at room temperature for 1 min. The slides were counterstained, mounted with DAPI/Antifade (Vector Laboratories, Burlingame, CA) and analyzed on the BioView Duet-3 fluorescent scanning station using 63X-oil objective and DAPI/FITC/Rhodamine single-band pass filters (Semrock, Rochester, NY). At least 100 tumor cell nuclei were scored for each specimen. The cutoff >10% was applied for the cells with the split Red and Green fluorescence signals to be scored as positive for BCL2 gene rearrangement. For 1p deletion, ratio of 1p-Red/1q-Green signals 3 signals in more than 40% of cells, as well as ratio Red/Green signal>1.3 was scored as copy number gain, while ratio >2 was indicative of focal amplification.

Transcription profiling public dataset Public Affymetrix U133+2 data sets from purified Resting, Naïve, GC, memory and FL B cells were retrieved from the public GEO datasets GSE12195 (http://www.ncbi.nlm.nih.gov/gds). We used Robust Multichip Average (RMA) with the non-parametric quantile algorithm as normalization parameter. RMA was applied to the raw data, and quantile normalization and Loess’ correction were done in R using Bioconductor and associated packages. The probesets corresponding to the gene of interest were retrieved from the normalized data sets and the corresponding log values were linearized for graphical representation. Ontology and pathway analyses were submitted to Ingenuity and DAVID’s database for annotation, visualization, and integrated discovery.

Sanger Sequencing

DNA from FLIS (n = 7), DFL (n=4), PF (n=3)L, FL (n = 6) and RFH (n=5 ) samples was used to analyze all exons of TNFRSF14 genes, exons 25-30 of the CREBBP gene, and Exon 15 of the EZH2 gene by PCR amplification. Sequencing was performed by Eurofins MWG Operon (Ebersberg, Germany). Mutations were sequenced twice from different PCR reactions to verify the aberration. Primers pairs are listed below: TNFRSF14 Exon 1: TGAGCTAGGTCTGGGTTGACCCGAGA/ CTCTGCTGGAGTTCATCCTG TNFRSF14 Exon 2: CCCAGGGCTTCATCACAC / CCCAAGTGCAGTCCAGGTAG TNFRSF14 Exon 3: CACTGTGGCCATGGAGAGAG / GGCCATTTGAGTCCCCTTAG TNFRSF14 Exon 4: TCTCTGTCCTGGAGCAGTTC / CTCCCAGGACCTTCCTGCAA TNFRSF14 Exon 5: TGCTGGAGAAGACGGTCAT / GGTCTAGAAGCTCACAGACA TNFRSF14 Exon 6: GCTTTCTTCTTTTCACACATATGATTAG/ CTGGGACCTGTCTTCACTG TNFRSF14 Exon 7: TGAGCTACCGTGGCTGTACT / GTGTATTGCAGGTGATGTAG TNFRSF14 Exon 8: GAAAACAGGAGCCGAATTC / TGGGTTGGCTGCAGTGTG CREBBP Exon 25: GGACACTTAAGAGCCCTGGTC/ CATTCACAGAGGTGCAGTTCC CREBBP Exon 26: CACCTGGAAAGAGGAGCTTTG/ CAGGGTGTTGTTTGTTGCTTG CREBBP Exon 27: CTCCAACTGTGCTGCTCTCAG/ TCCTGGCTTTAGTCCTTGCTC CREBBP Exon 28: AGGACCTAACAGTCGACACGC/ CACACATGCATGGGACTCTG CREBBP Exon 29: ACTTCCCTCCCACCACAGAC/GTGACCTACTTTGGCCTGAGC CREBBP Exon 30: CAGCCACCATCAGGTACAGAC/CTCAGCCACCTGCCTATTCTG EZH2_Exon 15: TCTCAGCAGCTTTCACGTTG/ CAGGTTATCAGTGCCTTACCTCTCC Statistics

Groups were compared using non parametric unpaired Mann & Whitney t-test. Only p-values inferior to 0.05 were considered as significant. * : p < 0.05

Figure S1. Alterations found in the various progression stages of follicular lymphomagenesis. The "% of total alterations” corresponds to the number of alteration per chromosome expressed as percentage.

Figure S2. FISH validation for BCL2 gene rearrangements in early FL entities with and without copy number variations. A. Dual-color Interphase FISH on FFPE tissue sections using BCL2 split signal probe showing BCL2 gene rearrangement that results in split red and green fluorescence signals. Cells with rearranged BCL2 gene are marked by arrowheads. Cells having BCL2 gene rearrangement and CNV are depicted by long arrows (i, ii) i. PFL #4 (left panel). Cell at lower left contains two fused signals, two red and two green split signals 2F/2R/2G, indicative of additional copies of wild –type and rearranged BCL2. ii. PFL #2 (center panel). One normal cell in center of the field contains two fused signals, 2F (normal diploid), whereas PFL cells contain 3-4 fused signals and one red, one green split signals 3-4F/1R/1G (arrow). iii. FLIS #4 (right panel). FLIS cells without CNV contain one normal fused signal, one red and one green split signals, 1F/1R/1G. Abbreviations: CNV, gain of BCL2 copy number; F, fused; R, red; G, green. B-D. Amplifications of the 7p14 region were confirmed on the PFL#2 sample, as well as on all available FL 1-2 (n = 1) and FL 3A (n = 4) samples displaying this alteration. A representative example of FISH result is shown in B. Amplifications of the 12q12-13 and 12q15 regions were confirmed on the PFL#1 sample, as well as on available FL 1-2 (n = 3) and FL 3A (n = 3) samples displaying this alteration. A representative example of FISH result is shown in C. Losses of the 1p36 region were confirmed on the PFL#2 sample, as well as on available FL 3A samples displaying this alteration (n = 2) . A representative example of FISH result is shown in D.

A. i.

ii.

iii.

B.

7p14

C.

12q12-15

12p11 CEP12 Green-Fluorescein 12q13.3-q14.1 Red-Rox

12

D.

1p36red-1qgreen

Figure S3. Mean number of gained and lost segments per sample in early-FL entities and FL samples.

Controls

FLIS

DFL

PFL

FL 1/2

FL 3A

Figure S4. Selected examples of focal gains in microdissected FLIS samples (orange and pink) versus their backgrounds (grey). Some of these alterations involved GC-related genes (AFF3, EBF1, BACH2, TOX). -4

0 -3

-2

-1

+4 +1

+2

+3

-4

0 -3

-2

-1

+4 +1

+2

+3

EBF1

AFF3

K2

K5

TOX

BACH2

K6

K11

Figure S5. mRNA expression of candidates (identified by CGH) in overt FL samples. A. Genes found in altered regions from PFL, FLIS and FL samples. B. Oncogenes (previously involved in NHL lymphomagenesis) found in PFL. C. Oncogenes amplified in DFL samples.

A

B

C

Figure S6. CGH profile of the chromosome 1, showing the loss at the 1p36 locus in PFL and DFL samples. In total, loss of the 1p36 locus are found in 17% of early-FL samples. PFL (A) and DFL (B) samples are in colors (Blue and purples shades respectively). Genes of interested covered in this region (TNFRSF14, TP73 and TNFRSF25) are indicated with black arrows

A

-1

-1

0

0

+1

+1

TNFRSF14

TP73

TNFRSF25

K1

B

-1

-1

0

0

+1 TNFRSF14

TP73

TNFRSF25

K1

+1

Figure S7. Mean number of alterations between 100 and 700kb (gains and losses) in early-FL and FL

Gains

Losses

FL 1/2

FL 3A

Mean number of alteration / sample

120 100 80 60 40 20 0 Controls

FLIS

DFL

PFL

Figure S8. t(14;18)+ frequency

1 10-1

10-2 10-3 10-4

10-5 10-6 10-7

FL-like B cells

?

?

FLIS

Weak

Strong

GC-(environment) dependency

Allelic paradox

Frozen GC features

Lymph node architectural distortion

Gains

PFL

t(14;18) Detection limit

FL

Major Genomic Alterations  Losses

CNV (GC related genes)

On going AID-mediated activity (including off-target mutations)

Progression phase of t(14;18)+ cells to FL diagnosis BCL2 staining intensity Antigen dependency Proliferation Index

Normal GC-experienced t(14;18)+ Memory B cells

t(14;18)+ B cells