putting the clinical and biological heterogeneity of non-hodgkin lymphoma into context

2014;20:5173-5181. Clin Cancer Res Owen A. O'Connor and Kensei Tobinai Lymphoma into ContextPutting the Clinical and Biological Heterogeneity of Non-Hodgkin

Updated version

http://clincancerres.aacrjournals.org/content/20/20/5173

Access the most recent version of this article at:

Cited Articles

http://clincancerres.aacrjournals.org/content/20/20/5173.full.html#ref-list-1

This article cites by 25 articles, 14 of which you can access for free at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

[email protected]

To request permission to re-use all or part of this article, contact the AACR Publications Department at

on October 29, 2014. © 2014 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from on October 29, 2014. © 2014 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from

http://clincancerres.aacrjournals.org/content/20/20/5173

http://clincancerres.aacrjournals.org/content/20/20/5173.full.html#ref-list-1

http://clincancerres.aacrjournals.org/cgi/alerts

mailto:[email protected]

mailto:[email protected]

http://clincancerres.aacrjournals.org/


Putting the Clinical and Biological Heterogeneity ofNon-Hodgkin Lymphoma into Context

Owen A. O'Connor1 and Kensei Tobinai2

AbstractThe lymphomas represent one of themost heterogeneous groups ofmalignancies in all of cancer medicine.

Whether one attempts to understand these diseases in the context of their complicated ontogeny, unique

biologic features, or clinical presentation, this heterogeneityhas been amixedblessing.On theonehand, it has

created an ever-changing way to classify these diseases, as classification schemes have been compelled to

reflect the rapidly emerging information that seems to split the disease into smaller and smaller subtypes. On

the other hand, the biologic and clinical dissection of these diseases has allowed for the identification of

unique biologic features—features that have led to novel targets and generated a plethora of new drugs.

Virtually every subtype of non-Hodgkin lymphoma has benefited from these efforts to understand the

biology of the different subtypes. This paradigmhas led to new clinical trials that tailor novel drug regimens to

specific biologic disease subtypes. As a prelude to this CCR Focus section, we attempt to put this evolving

heterogeneity into context, bridging historical andmodern-day views of classification of these diseases. Then,

some of the world’s leading lymphoma researchers share their perspectives on how to formulate new con-

cepts of care in this era of biologic discovery. Over a relatively short time, the advances in lymphoma research

have been nothing short of stunning. There now seems to be little doubt that these recent breakthroughs will

redound favorably on the majority of patients diagnosed with a lymphoproliferative malignancy.

See all articles in this CCR Focus section, "Paradigm Shifts in Lymphoma."

Clin Cancer Res; 20(20); 5173–81. �2014 AACR.

IntroductionAnyone who has tried to keep abreast of the literature, or

attended any national or internationalmeetings in the field,knows we are living in the midst of an era where data andinformation are being revealed at a dizzying pace. Infor-mation, now measured in terabytes, has created new fieldsof study and employment in bioinformatics, a relativelynew interdisciplinary scientific field dedicated to the for-mulation of software for storing, retrieving, organizing, andanalyzing biologic data. It is interdisciplinary as it relies onthe principles of mathematics, statistics, and computerscience to study and process biologic data. It is no longerpossible to comprehensively analyze by hand the largedatasets on disease biology, or the treatment-related pre-dictors of response or toxicity. With the terabytes of dataavailable, the question now becomes, have we becomemore knowledgeable? If information provides answers to

questions of "what," "where," and "when," knowledgeanswers the question of "how." How does one translatethe seemingly infinite amount of information into practicalguidelines? How can we optimally leverage the terabytesof data to improve the outcome of a disease, or to helpa patient navigate a challenging diagnosis of cancer?Although astounding progress has beenmade in generatingdata and information around the sphere of cancer biology,it is clear that the rate-limiting step hindering the realizationof real progress in treating neoplastic disease comes down toanswering the question "how." Translating the basic scienceinto practice is often the rate-limiting step.

So, have we become more knowledgeable? I believe theanswer is yes, absolutely. Although we have just begun tovisualize the possibilities for translational medicine emerg-ing across all of cancer medicine, I believe the lymphoidneoplasms offer a useful paradigm for thinking aboutinnovative ways to tailor new and existing therapies todiscrete biologic entities of disease. This CCR Focus hasbrought together some of the world’s leading authoritiesfocused on trying to make sense of the biologic heteroge-neity that defines the lymphoproliferative malignancies.As information and data emerge about the biologicorigins of lymphoma, we have learned that this disease ischaracterized by remarkable heterogeneity, heterogeneitythat has compelled us to continuously rebuild a "knowl-edge" base for each of the ever-expanding number of sub-types. Every subtype of these complex diseases is being

1Medicine and Experimental Therapeutics, Center for Lymphoid Malignan-cies, Columbia University Medical Center, College of Physicians andSurgeons, The New York Presbyterian Hospital, New York, New York.2Department of Hematology, National Cancer Center Hospital, Tsukiji,Chuo-ku, Tokyo, Japan.

Corresponding Author: Owen A. O'Connor, Medicine and ExperimentalTherapeutics, Center for Lymphoid Malignancies, Columbia UniversityMedicalCenter, 51West 51stStreet, Suite 200,NewYork, NY 10019.Phone:212-326-5723; Fax: 212-326-5725; E-mail: [email protected]

doi: 10.1158/1078-0432.CCR-14-0574

�2014 American Association for Cancer Research.

CCRFOCUS

www.aacrjournals.org 5173

on October 29, 2014. © 2014 American Association for Cancer Research. clincancerres.aacrjournals.org Downloaded from


reclassified and recategorized on the basis of an array ofcellular andmolecular descriptors. The authors have tried topresent a framework for how to interpret the heterogeneityof their disease, and how the field is thinking about usingthis information to generate new concepts aroundtreatment.

The degree of biologic heterogeneity seen in any biologicsystem is highly dependent on the tools available to probeits details. How one sees that heterogeneity also depends onthe lens through which one views the complexity. At apractical level, clinicians confront the heterogeneity oflymphoma every day in the clinic, without a clear under-standing of its basis. Why do some patients with diffuselarge B-cell lymphoma (DLBCL) do better than others?Whyare some patients with gastric MALT cured with antibiotics,and others not? Why can some patients with mantle celllymphoma (MCL) bewatched,while others need a stem celltransplant? Hematopathologists may describe it in terms ofcellular morphology, tissue architecture, patterns of IHCstaining, or presence of translocations or chromosomaladditions or deletion. Basic scientists may see it in the strict

context of any of the emerging –omics (genomics, proteo-mics, epigenomics, metabolomics). Clinicians see it in thecontext of a patient’s experience, with some being cured,while others not.

The heterogeneity we see in lymphoma is a direct productof the natural features of lymphocyte development anddifferentiation (Fig. 1). B- and T- lymphocyte ontogeny iscomplex, and involves a multitude of genetic steps requiredto generate diverse andB- and T-cell populations to producean expanded repertoire of immune cells prepared to con-tendwith diverse antigenic exposures. These cells, at variousstages of their development, undergo complex V(D)J re-combination, IgV, or somatic hypermutation (SHM) andisotypic switching (refs. 1, 2; Fig. 1). SHM in fact represents aunique and intrinsic mechanism to "naturally" generatecellular diversity and heterogeneity. Through the process ofaffinitymaturation, SHMdiversifies B-cell receptors that areimportant in recognizing antigen, producing an armedimmune system able to contendwith a variety of challenges(Fig. 1). The mechanism of BCR diversification involves aprocess of controlled mutagenesis affecting the variable

© 2014 American Association for Cancer Research

V(D)J recombination

Immature

B cells

Naïve

B cellsGerminal-center B cells

B-CLL

DLBCLFLBLMCLALL

Bone marrow

Mantle

Ag

Germinal center

Ig isotype switchIgV hypermutation

Figure 1. B and T lymphocytes naturally undergo "controlled" recombination SHM, leading to immunoglobulin diversity. ALL, acute lymphoblastic lymphoma;BL, Burkitt lymphoma; CLL, chronic lymphocytic leukemia; FL, follicular lymphoma.

CCRFOCUS

Clin Cancer Res; 20(20) October 15, 2014 Clinical Cancer Research5174



regions of the immunoglobulin genes, a process that isrestricted to the immune cells, that is, it does not produceheritable changes in the genome (2). During proliferation,the B-cell receptor undergoes the highest rate of somaticmutation seen anywhere in the body, approximately 105- to106-fold greater than that seen across the rest of the genome(1, 2). Thesemutagenic events primarily involve single basesubstitutions, and to a lesser extent insertions anddeletions,and occur at hypervariable regions of the DNA called "hot-spots." A mistargeting of SHM during the germinal centerreactionhas been called "aberrant somatic hypermutation,"(ASHM) and has been invoked to explain mutations inimportant regulatory genes like tumor suppressor or proto-oncogenes. It has also been implicated in the pathogenesisof DLBCL (3). Thus, in lymphocyte biology, the concept ofheterogeneity has many contexts. While a necessary andcritical component of B- and T-lymphocyte biology, there isalso the heterogeneity that comes with a desire to explaindifferences in behavior of similar diseases and differencesin outcome. It is a process that naturally lends itselfto splitting, and splitting, and more splitting. Althoughthis process has provided deeper insights into pathogenesis(e.g., ASHM), the emergence of heterogeneity at so manydifferent levels has created a conundrum as well. Will it everbe possible to conduct clinical studies on the increasinglysmaller and smaller subsets of an already rare disease?Assuming we had all the right drugs, how long would ittake to prove that treatment of one subtype differentiallycould make a difference?

Organized Heterogeneity: Development ofClassification SystemsThe evolution of classification systems is medicine’s

attempt to organize heterogeneity. In the late 1960s, thetools were a microscope with dyes staining basic and acidicmacromolecules, which produced a purely morphologicclassification of lymphoma, based on only two descriptors:(i) cells were large or small and (ii) patterns of infiltrationwere either nodular or diffuse. As a result, the Rapportclassification (4), in its simplest form, classified lymphoidmalignancies simply as diffuse small cell, nodular small cell,and the like (Fig. 2). Not all that heterogeneous by today’sstandards. Advances in immunology in the 1970s led to therecognition of cell surface proteins (in particular the clustersof differentiation, or CDmarkers), and the observation thatwith the more markers one could survey, diffuse small celllymphoma (as an example) in fact consisted not merely ofone homogenous entity, but many (5–7). The repertoireof cell surface markers vastly expanded both the number ofpossible diagnoses, and indirectly, the heterogeneity. None-theless, the standardization and routine integration of IHCstaining techniques produced a quantum leap forward. IHCinformed Karl Lennert (Kiel classification) and Lukes andCollins as they formulated a more sophisticated immuno-logic perspective of the lymphoproliferative malignancies,recognizing the importance of cell lineage (B- and T cells) intheir classification system (5, 6). In the mid-1970s, the NCI

convened a panel of experts to make sense of this dauntingbiologic heterogeneity, by focusing less on biology andmore on clinical features. The NCI Working Group formu-lated a clinically based classification system that integratedall the original concepts, producing a more clinically ori-ented set of descriptors, dividing the lymphomas into lowgrade, intermediate grade, and high grade (8, 9). Low-gradelymphoid malignancies like diffuse small cell or nodularsmall cell lymphomas were recognized as very indolent,slow-growing disease (growing over months to years),whereas intermediate-grade lymphomas like nodular largecell or DLBCL were aggressive (growing over weeks tomonths), and if not treated promptly were fatal. Smallnoncleaved cell or lymphoblastic lymphomas were consid-ered as high grade (growing over days to weeks). While notyet a biologically based classification system, the clinicalbehavior descriptor added yet another variable that neededto be considered as new ways to think about these diseasesemerged (Fig. 2). The tools at this stage were comparativelysimple: morphology, IHC, and clinical behavior. However,the insight and ability to synthesize and configure manyapparently diverse lines of data were ingenious. Thesedecades of focus by pathologists and clinicians still formthe foundation upon which all future lymphoma classifi-cation systems are based (Fig. 3).

The recognition that every cancer is fundamentally adisease of aberrant gene expression, a genetic disease, hashad a profound impact on our thinking. The ability to definenew subsets of lymphoma based on very particular geneticevents has become an essential component of all present-dayclassification systems. For example, the 11:14 translocationinvolving the immunoglobulin heavy chain gene promoterand cyclin D1 t(11:14), leads to accumulation of cyclinD1 and in pathognomonic of MCL. Likewise, the pathogno-monic (14:18) translocation leads to accumulation of Bcl-2in follicular lymphoma. These techniques produced yetanother quantum leap in our thinking, lymphoma complex-ity, and as a result, heterogeneity. The routine integration ofnew tools in cytogenetics (FISH, comparative genomichybridization) has helped rewrite the algorithms routinelyused in subclassifying many subtypes of lymphoma.

The World Health Organization (WHO) and REAL(Revised European and American Lymphoma) classifica-tions have emerged as the most current iteration in lym-phoma subtype organization (10–12). These classificationsystems have synthesized all the available information oncellular morphology, tissue architecture, IHC, and cyto-genetic and clinical features (Fig. 2 and Table 1). The WHOclassification now recognizes about 65 different subtypes oflymphoma. In the United States, there are about 70,000cases of non-Hodgkin lymphoma (NHL) per year. If thesediseases were evenly distributed, that would mean therewould only be about 1,000 cases of each type per year. Infact, two subtypes, DLBCL and follicular lymphoma, com-prise about half of all cases, which leaves about 60 subtypesfor the remaining 35,000 cases per year—making manysubtypes of lymphoma incredibly rare orphan diseases.Many subtypes of lymphoma now have an incidence

Putting the Heterogeneity of NHL in Context

www.aacrjournals.org Clin Cancer Res; 20(20) October 15, 2014 5175



measured in 1 to 3,000 cases annually in the United States.Although the heterogeneity in the disease as a whole hasincreased substantially over prior classification schemes,more recently we have seen an expansion of the heteroge-neity seen in individual subtypes (Table 1). For example,DLBCL is now widely recognized as not one, but manydifferent diseases.

With the advent of new and more sophisticated tools tointerrogate the genome, has come new ways to reclassifymany human diseases, and new opportunities to split stillfurther. The most recent era of lymphoma classification isclearly focused on how to interpret gene expression profil-ing. The seminal observations by Alizadeh and colleaguesfrom the NCI (13), as well as Shipp and colleagues from

Boston (14), have opened a new door, changing how weview lymphoproliferative malignancies. Although underthemicroscope, the cells from a patient with DLBCL appearlarge, grow in a diffuse pattern, and all express thesame profile of CD markers on their surface [CD19(þ);CD20(þ); PAX5(þ); CD3(�)], some patients with a diag-nosis of DLBCL will be cured, whereas others will succumbin short order. But how could patients with the samediseasehave suchdivergent outcomes? Yielding vivid differences ona heatmap, what appears indistinguishable under themicroscope, gene expression profiling now reveals as obvi-ously different. Although these techniques are not yetroutinely available for clinical practice (though they willbe soon), every patient and doctor now approach every case


Tissue cytomorphology

Immunohistochemistry

Current chromosome and genetic analysis techniques

FISHCGH

Gene expression profiling

Gen

es

Germinal-center

B-cell likeType 3

Activated

B-cell like

The future:

Next-generation sequencing

Single-nucleotide polymorphisms

GC phenotype

Non-GC phenotype

MUM1CD10 BCL6

Figure 2. A hierarchy of how heterogeneity can be viewed in lymphoproliferative malignancies. CGH, comparative genome hybridization.

CCRFOCUS




of DLBCL wanting to know where they fit in the lexicon ofexpression profiling. In the case of DLBCL, the NCI expe-rience has emerged as the one most commonly referenced,though it is clear there are many valid ways to sort andclassify gene expression profiles. Dunleavy and colleagues(15) report on the NCI gene expression profiling (GEP)experience inDLBCL,whichhas provided abiologic contextthat nicely separates differences in outcome seen in patientswith DLBCL, based on ontogeny. Their approach, definingthe disease based on the cell of origin (COO), has describedall cases ofDLBCL asbeing derived from the germinal center(GC), or postgerminal center, or activatedB-cell type (ABC).A third type represents a smaller fraction, now known torepresent primary mediastinal large B-cell lymphoma, ahighly favorable subtype of DLBCL genetically closer toHodgkin lymphoma than DLBCL. Now, through the filterof GEP, those patients with lymphomas derived from theGC exhibit a more favorable outcome, and have disease

prominently defined by enrichment of cells with dysregu-lated Bcl-6, and other features indicative of marked epige-netic dysregulation like EZH2 (16). Those patients withDLBCL of the ABC subtype appear to have an inferioroutcome, and have disease enriched for dysregulation ofNF-kB signaling. This molecular classification of DLBCL isalready leading to new translational efforts, allowing us toleverage the GEP data to inform us how to treat patientsdifferently. Drugs such as ibrutinib, proteasome inhibitors,and lenolidomide appear to be more active in the ABCphenotype, and are now being integrated into upfrontR-CHOP–based chemotherapy regimens in randomizedclinical studies. In contrast, other studies have optimizedhistone deacetylase inhibitors to target the Bcl-6–p53 axis inGC-derived DLBCL with promising results (17). In thisparticular example, we have become more knowledgeable,as data and information from efforts in quantitative exper-imentation are translated into clinical studies. Although we


Morphologic and clinical

Era of conventional cytotoxic therapies

Milestones in the approval of new drugs for the treatment of lymphoma

Milestones in the classification and characterization of lymphoma

Era of novel targeted therapies

1960 1970 1980 1990 2000 2005 2010 2014

1974: Kiel and Lukes and

Collins classifications—

utilization of IHC

1960s: Rappaport

classification—

morphologic distinctions

1982: NCI Working

Formulation—groupings

by clinical features

1973: development

of ABVD for

Hodgkin lymphoma

1976: development

of CHOP for

aggressive lymphoma

Late 1980s: HDT/ASCT

for relapsed

aggressive lymphoma

1997: FDA approval of

rituximab for R/R low-grade

or follicular B-cell NHL

2014: FDA approval

of ibrutinib for

R/R CLL and MCL;

idelalisib for R/R CLL

and FL; belinostat for

R/R PTCL; siltuximab

for multicentric

Castleman disease

2013: FDA approval

of lenalidomide

for R/R MCL

2011: FDA approval of

brentuximab

for R/R HL and ALCL;

romidepsin for R/R PTCL

2009: FDA approval

of pralatrexate

for T-cell NHL

2007: FDA approval

of vorinostat for

advanced CTCL

2006: FDA approval

of bortezomib

for MCL

2004: identification

of distinct stromal

signatures in FL by GEP

2012: whole-genome

sequencing of Burkitt

lymphoma performed

2001: WHO

classification (1st ed.)

1994: REAL classification—

integration of pathologic

and clinical characteristics

2000: identification of distinct

DLBCL subtypes by GEP

2008: latest edition of WHO

classification—subdivision by

morphologic, immunophenotypic,

genetic, and clinical features

Integration of genomics

Figure 3. Milestones in the classification of lymphoma and novel agents approved in the past decade. FL, follicular lymphoma.





still need the results of the randomized studies under way,inmy opinion, the process is likely to represent amajor stepforward in our quest to cure all patients with DLBCL.

The advances in classification, while remarkable, repre-sent only a portion of the advances in lymphoma therapy inrecent years (Fig. 3). In thisCCR Focus, fiveworld-renownedgroups provide insights on how our concepts in basicbiology are being translated into practical treatment strat-egies, approaches that represent marked departures fromthe "one-size-fits-all" model of the past. Using the GEPprofiling experiences from the NCI as a platform, Dunleavyand colleagues (15) outline a path in which the molecularclassification of DLBCL based on the cell of origin (COO)model is being used to define molecular phenotypes of thedisease, phenotypes that appear to exhibit a differentialsensitivity tomany of the newer drugs now in development.The ABC subtype of DLBCL, for example, appears to besensitive to agents targeting the Bruton tyrosine kinase(BTK) and immunomodulatory drugs, which whencombined with upfront treatment platforms like R-CHOP,produce remarkably high response and complete responserates, concepts now being vetted in international random-ized phase III studies. In a similar manner, Dreyling andcolleagues (18) have shown that a disease like MCL, his-torically thought of as one of the most challenging lym-phomas, is in fact a disease that exhibits a broad spectrumofbehavior. This behavior is tightly linked to its proliferationindex, with patients having the higher proliferation rateexhibiting the worse outcome, whereas those with lowproliferation indices exhibit a highly favorable outcome.Remarkably, patients with a low index may require nothingmore than a "watch-and-wait" approach, whereas patientswith a high index could require an autologous stem celltransplant. This is a perfect example of how the heteroge-neity within a particular subtype is influencing treatmentdecision making. Patients with the same histologic subtypeof the disease can be managed by two extremes in care.Zucca and colleagues (19) and Tsukasaki and Tobinai (20)introduce the notion that many subtypes of lymphoma areactually caused by infectious agents, and that controlling oreradicating the infectious etiology canbe a very effectivewasto treat the malignant disease. Zucca and colleagues (19)underscore the casual relationship between the manyknown etiologic agents linked to marginal zone lympho-magenesis, and the heterogenous nature of that diseasebased on the infectious agent, and the underlying geneticsof the lymphoma. It is the underlying genetic nature of thedisease that defines those patients likely to benefit fromantibiotic therapy. In similar fashion, human T-cell lym-photropic virus type I (HTLV-1) causes adult T-cell leuke-mia–lymphoma (ATL). Tsukasaki and Tobinai (20) providea detailed update on the epidemiology ofHTLV-1, and sharedata demonstrating that measures to control HTLV-1 havealso led to a reduction in the malignant disease. Just asimportantly, CC chemokine receptor 4 (CCR4), a proteinexpressed on the surface of ATL cells, can nowbe targeted bya newly developed defucosylated humanized anti-CCR4monoclonal antibody (mogamulizumab), which has

Tab

le1.

Com

mon

lympho

mas

andtheirch

arac

teriz

ation

Disea

seTissu

earch

itec

ture

and

cellu

larmorpho

logy

IHC

profile

Cytogen

etic

features

Rec

entgen

etic

or

biologic

features

DLB

CL

Diffus

eefface

men

tof

nodal

arch

itecture

CD20

þ;C

D10

þ;C

D19

þ;

CD79

aþ;C

D4�

;CD8�

,CD5þ

/�COO

divides

disea

seinto

GC,

ABC

subtypes

,and

PMLC

L

Follicu

larlympho

ma

Nod

ular

replace

men

tof

node

bysm

allo

rlargeBce

llsCD20

þ,C

D10

þ;C

D19

þt(1

1:14

)—Bcl

2ov

erex

press

ion

Strom

almicroen

vironm

entl

correlates

with

progn

osis

MCL

Diffus

eor

nodular

bymed

ium

size

dBce

llsCD20

þ;C

D5þ

;CD23

�;

CD10

�;c

yclin

D1�

t(11:14

)—cy

clin

D1

overex

press

ion

Proliferationindex

(Ki67su

rrog

ate)

correlates

with

survival

Margina

lzon

elympho

ma

Diffus

einvo

lvem

entby

smallc

ells

CD20

þ;C

D10

�;C

D19

þ;

CD5�

;CD2þ

Chron

iclympho

cytic

leuk

emia

Smallc

ells

CD20

þ;C

D10

�;C

D5þ

;CD23

þ;C

D19

þ11

q,12

q,a

nd17

pde

letio

ns

Periphe

ralT

-celllym

pho

mas

Diffus

elargeTce

llsCD4or

8þ;C

D20

�Non

ech

arac

teris

tic

CCRFOCUS




recently been approved in Japan for the treatment ofpatients with ATL. This article highlights how the uniquebiology of a lymphoid neoplasm can be successfully trans-lated into a disease-specific treatment. In a similar fashion,the article by O’Connor and colleagues points out wherewe are in understanding PTCL biology and treatment,highlighting the many breakthroughs over the past fewyears that one hopes will change the natural history of thesechallenging diseases. Finally, building on the remarkableadvances in the treatment of indolent diseases like follicularlymphoma, and the major contribution of immunologi-cally targeted therapies for thedisease, Bachy andSalles (21)provide a clear vision for how a focus on the future man-agement of this disease could circumnavigate the need formore conventional cytotoxic therapies. By combining newagents targeting the BTK level and the PI3K pathway,immunomodulatory drugs with monoclonal antibodies,or combinations of "newer" and "older" drugs, compara-tively less toxic and potentially more efficacious treatmentplatforms are emerging. These new paradigms carry thehope that they will turn follicular lymphoma into a highlymanageable chronic disease that is less predicated on con-ventional chemotherapy approaches for its management.Each of these experiences underscores and highlights theprofound advances occurring across the larger panoply oflymphoid neoplasms, and how emerging data from thebasic science side is informing a focused translational effort.However, despite the dazzling advances, some might see

the job as far from complete. Although we may effectivelycure more patients by combining the novel agents with up-front treatment programs such as R-CHOP, there are anumber of important new considerations that we need toacknowledge inmoving forward. First, if we are to optimallyleverage the latest advances in basic science and the latestbreakthroughs in highly targeted therapy, will it ever bepossible to consider the development of chemotherapy-freetreatments of some subtypes of lymphoma? Second, howdo we move away from the "one-size-fits-all" model oflymphoma treatment, and develop strategies that move ustoward more of a personalized or precision treatment ofgood andpoor risk subgroups of patients? Finally, given our50-year history in classifying and reclassifying lymphoma ina fashion that has magnified the heterogeneity, are thereapproaches in bioinformatics that might allow us to see thelymphomas in a different perspective, allowing us to build aview based on common themes rather than uncommondifferences?First, it is important to acknowledge that we are still

addicted to a backbone of treatment that is predicated onnonspecific cytotoxic drugs for some types of lymphoma.For the aggressive diseases, it is understandable, as thesebackbones produce cures in many, if not most patients.Discarding an effective backbone puts cure rates at risk.Froma regulatory perspective, it is also the only strategy thatallows one to design a randomized study against a controlarm acceptable to regulatory and ethics committees review-ing such studies. However, as seen in the article by Bachyand Salles (21), there may in fact be subtypes of lymphoma

in which chemotherapy may not be required. Of course, indiseaseswhere there is little consensus on a standard of care,it is easier to design such studies. In the case of follicularlymphoma, the second most common subtype of lympho-ma, Bachy and Salles (21) provide interesting insights intohow the treatment of this disease will change over the nextfew years. The emergence of new monoclonal antibodies,antibody drug conjugates, immunomodulatory drugs, andmore pathway-targeted agents such as BTK inhibitors, PI3Kinhibitors, and BH3 only mimetics, could create successivelines of therapy that allow the practitioner to managefollicular lymphoma as a chronic illness, with drugs exhi-biting a safety profile that might allow longer periods ofadministration. Although there may be important syner-gistic interactions with these newer agents and classic DNA-damaging drugs or topoisomerase inhibitors, new regimenswith new components are moving us away from the stan-dard CHOP backbone. Certainly, in such diseases as dou-ble-hit lymphomas and many subtypes of peripheral T-celllymphoma, it is likely that new non–CHOP-predicatedbackbones will emerge here first.

Second, how do wemove toward a more personalized orprecision focused model of lymphoma care? Irrespective ofthe incremental improvement thatmay resultwhen thedatafrom the randomized studies reported, not everyone in themore favorable treatment cohort will be cured, and noteveryone in the unfavorable treatment cohortwill succumb.Not every patient with GC-derived DLBCL does well, nordoes every patient with ABC-derived DLBCL do poorly. Notevery patient with low Ki-67 MCL has indolent disease.About 20% of patients with follicular lymphoma do nothave indolent disease, and exhibit patterns of continuedrapid relapse and succumb within a couple of years ofdiagnosis. The exceptions to the treatment outcome ornatural history reflect what we do not know about thesignaling networks, what we do not appreciate when wetry to resolve the heterogeneity by creating "cleaner" sub-groups. To solve the exception, one approach would be tocontinue the splitting, refining at smaller and smaller levelsthe subset of the subset, until we figure it out for everypatient. This would be a time-consuming endeavor. Alter-natively, we can try alternative views of complex data, bydistilling out irrelevant heterogeneity in an effort to identifyunifying concepts.

So, finally, what unifying concepts might allow us tomove forward without addressing every pathogenetic eventseparately? Just how does the tumor cell respond to theimposed therapy, andhowdo the responses of the cell to thenew stressor differ between favorable-risk and unfavorable-risk patients? The notionof "one disease (or phenotype of)–one target–one drug" ignores the idea that the cell is adynamic complex system of many interacting pathways,pathways that may respond differently to new stressors,depending on the intrinsic genetic subtleties of that partic-ular cell. Inhibiting BTK does not produce an effect on thecell in a vacuum; it stimulates changes in a host of pathwaysthat interface with the targeted pathway, responses that canvary in different cells from the same tumor. In some cases,





those responses may mediate the emergence of a drug-resistant clone, whereas in others, it may result in cell death.These types of descriptors (BTK level, PI3K activity, etc.) thatwe are now targeting in a therapeutic fashion are all diag-nostic; they represent characterizations of the disease beforewe treat. They importantly do not reflect the phenotype ofdisease left after initial treatment, nor does it provide insightinto the biologic response of the disease to the imposedtreatment. Obtaining serial biopsies shortly after treatmentand interrogating the cellular responses in the context of thelarger cellular interactome are likely to shed light on themechanisms of resistance or sensitivity, and are likely torefine our ability to more precisely ascribe the most appro-priate therapy to the right patient population.

The phenotypes of cancer cell behavior we see in patientsdo not arise from single genes or isolated molecular events.The phenotypes we see in patients arise from the function ofhundreds or thousands of macromolecules like proteins,DNA, RNA, microRNA, carbohydrates, which themselvesall interact in complex regulatory networks, some of whichundergo complex and critical posttranslational modifica-tions. It is the cascade of biologic functions mediated byinteracting networks that produces the behavior of cancer inany given patient. The identification of a singular geneinfluence or impact of the latest mutation, albeit contrib-uting factors, obviously should not be viewed in isolation. Itmay seem an insurmountable enterprise, but how does oneunderstand the impact of all these interactions at the level ofthe protein:protein or protein:nucleic acid level, at thegenomic level, or at the epigenetic level? The answer is totake a broader "systems" level view of the biology (22–24).Systems biologists provide a holistic view of cancer byintegrating genetics, signaling networks, epigenetics, prote-omics, metabolomics and the like, they attempt to synthe-size all the information, which to the naked eyemight seemto be disjointed information. They attempt to reducethe heterogeneity, or perhaps reconfigure it, by trying tofind the similarities. Systems biology provides us with theunique ability to integrate detailed data stemming fromquantitative experimental analysis through the use ofmath-ematicalmodels and computational analysis. These holisticapproaches leveragemany seemingly divergent lines of dataand reorganize the information in a fashion that places anemphasis on defining larger regulatory networks, reducingthe heterogeneity, and defining the principles of the cells’larger interactome. An approach that may take us full circle.

It is a remarkable accomplishment tobe able to say thatwehave changed the natural history of not just one, but severalmalignant diseases in such a short period of time. In somecorners, our community continues to strugglewithpersistentchallenges in lung, colorectal, and pancreas cancer, as well asacute myeloid leukemia, looking to demonstrate even mod-est improvements in progression-free or overall survival.Some drugs in these settings have been approved based onsurvival benefits measured in weeks to months. What hasoccurred in the lymphoid malignancies over this period hasbeen extraordinary (Fig. 3). Over the past 20 years, we havewatched MCL move from a disease with a life expectancy

measured in 3 to 4 years, to one that might in fact be curablein 2014. Through appropriate patient selection, many ofthese patientsmay be able to avoid intense cytotoxic therapyand their cancermay bemanageable as a chronic disease.Wehave seen some forms of marginal zone lymphoma curedwith antibiotic therapy, and nowwe recognize themolecularphenotypes associated with eradication of a malignant dis-ease by treating an underlying infectious etiology. We havewatched a disease like follicular lymphoma, a disease with amedian life expectancy of 7 or so years when the originalStanforddatawere releasedbyDr.Horning,becomeadiseasewhere we do not know, with over 15 years plus of follow-up,where the median life expectancy might end up in our newera of treatment. A disease changed by essentially one drug, aCD20-targeted monoclonal antibody. And now, most dis-cussions around futuremanagement of follicular lymphomadonot includeconventional chemotherapy components.Wehave watched as gene expression profiling has distilled theheterogeneity seen in a disease like DLBCL. Clinical trialsexploring themerits of addingnew targeteddrugs to standardbackbones are reporting overall response rates between 90%and100%,andappear tobemore active in theprognosticallyworsephenotypesof thedisease.Randomized studies explor-ing R-CHOP � ibrutinib, new monoclonal antibodies withimproved ADCC, immunomodulatory drugs such as leno-lidomide, andNF-kB–targeted drugs suchas bortezomib, areappearing at a rapid rate. Even in a forgotten disease likePTCL, fournewdrugshavebeenapprovedover thepast4 to5years, where none had been approved over the previous 50years (Fig. 3; refs. 25–27). Despite its rarity, four new ran-domized studies are under way in PTCL for the upfrontsetting. We are more knowledgeable, and new scientificconcepts are being translated at a remarkably efficient pace.Yes, there are still important areas for improvement. We candobetter by trying to identify biomarkers of response, so thatwe can carefully select our patients for a given therapy. Wecan do better in understanding the network responses of acancer cell to a drug therapy, and use that information toprevent the emergence of drug resistance. And, of course, wecan do better by collaborating with colleagues across thedisciplines of biology, chemistry,mathematics, and statistics,identifying fresh new ways to interpret the diverse lines ofinformation. Although the progress has been no less thanstunning, our job is done only when each and every patientwe see can be rest assured that every lymphoma is a goodlymphoma.

ConclusionsClearly, dissecting the similarities and differences in

"lymphoma" behavior both among patients and betweendiscrete entities has led to a natural splitting of subtypes.This splitting has allowedus to seemore clearly those factorsdriving the misbehavior of these malignant cells, and howbest to target its molecular roots. Our succeeding articleswill share the latest perspectives on how biologic discoveryin particular subtypes of lymphoma is being translated intonovel concepts in care, paradigms that are changing thelong-standing natural history of many lymphoma entities.

CCRFOCUS




Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: O.A. O’Connor, K. TobinaiDevelopment of methodology: O.A. O’ConnorAcquisitionofdata (provided animals, acquired andmanagedpatients,provided facilities, etc.): O.A. O’ConnorAnalysis and interpretation of data (e.g., statistical analysis, biosta-tistics, computational analysis): O.A. O’Connor, K. TobinaiWriting, review, and/or revision of the manuscript: O.A. O’Connor,K. Tobinai

Administrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): O.A. O’ConnorStudy supervision: O.A. O’Connor

Grant SupportO.A. O’Connor was supported by the Lymphoma Research Fund of

Columbia University Medical Center. K. Tobinai was supported by theNational Cancer Center Research and Development Fund (23-A-17 and26-A-4).

Received August 11, 2014; accepted August 26, 2014; published onlineOctober 15, 2014.

References1. Li Z, Woo CJ, Iglesias-Ussel MD, Ronai D, Scharff MD. The generation

of antibody diversity through somatic hypermutation and class switchrecombination. Genes Dev 2004;18:1–11.

2. Odegard VH, Schatz DG. Targeting of somatic hypermutation. Nat RevImmunol 2006;6:573–83.

3. Pasqualucci L, Migliazza A, Fracchiolla N, William C, Neri A, Baldini L,et al. BCL-6 mutations in normal germinal center B cells: evidence ofsomatic hypermutation acting outside Ig loci. Proc Natl Acad Sci U SA1998;95:11816–21.

4. Rappaport H. Classification of neoplastic diseases of the reticularsystem. In: Roulet FC, editor. Symposium on lymphoreticulartumours in Africa; 1964 Sept 12; New York: S. Karger; 2004.p. 394–404.

5. Lukes RJ. The immunologic approach to the pathology of malignantlymphomas. Am J Clin Pathol 1979;72 Suppl 4:657–9.

6. Lennert K, Stein H, Kaiserling E. Cytological and functional criteria forthe classification of malignant lymphomata. Br J Cancer Suppl1975;2:29–43.

7. Dorfman RF. Letter: classification of non-Hodgkin's lymphomas. Lan-cet 1974;1:1295–6.

8. Classification of non-Hodgkin's lymphomas. Reproducibility of majorclassification systems. NCI non-Hodgkin's Classification ProjectWrit-ing Committee. Cancer 1985;55:91–5.

9. National Cancer Institute sponsored study of classifications of non-Hodgkin's lymphomas: summary and description of a working formu-lation for clinical usage. The Non-Hodgkin's Lymphoma PathologicClassification Project. Cancer 1982;49:2112–35.

10. Harris NL, Jaffe ES, Stein H, Banks PM, Chan JK, Cleary ML, et al. Arevised European-American classification of lymphoid neoplasms: aproposal from the International Lymphoma Study Group. Blood1994;84:1361–92.

11. Melnyk A, Rodriguez A, Pugh WC, Cabannillas F. Evaluation of theRevised European-American Lymphoma classification confirms theclinical relevance of immunophenotype in 560 cases of aggressivenon-Hodgkin's lymphoma. Blood 1997;89:4514–20.

12. Swerdlow SH, Campo E, Harris NL, Jaffe ES, Pileri SA, Stein H, et al.,editors.WHOclassification of tumours of haemaopoietic and lymphoidtissues. 4th ed. Lyon (France): IARC Press; 2008.

13. Alizadeh AA, EisenMB, Davis RE,MaC, Lossos IS, Rosenwald A, et al.Distinct types of diffuse large B-cell lymphoma identified by geneexpression profiling. Nature 2000;403:503–11.

14. Shipp MA, Ross KN, Tamayo P, Weng AP, Kutok JL, Aguiar RC, et al.Diffuse large B-cell lymphoma outcome prediction by gene-expres-

sion profiling and supervised machine learning. Nat Med 2002;8:68–74.

15. DunleavyK, RoschewskiM,WilsonWH.Precision treatment of distinctmolecular subtypes of diffuse large B-cell lymphoma: ascribing treat-ment based on the molecular phenotype. Clin Cancer Res 2014;20:5182–93.

16. Amengual JE, O'Connor OA. Manipulating the epigenome in germinalcenter lymphomas: is it getting easier and ezier? Clin Cancer Res2014;20:3047–9.

17. Amengual JE, Clark-Garvey S, Kalac M, Scotto L, Marchi E, Neylon E,et al. Sirtuin and pan-class I/II deacetylase (DAC) inhibition is syner-gistic in preclinical models and clinical studies of lymphoma. Blood2013;122:2104–13.

18. DreylingM, Ferrero S, VogtN,KlapperW.Newparadigms inmantle celllymphoma: is it time to risk-stratify treatment based on the proliferativesignature? Clin Cancer Res 2014;20:5194–206.

19. Zucca E, Bertoni F, Vannata B, Cavalli F. Emerging role of infectiousetiologies in the pathogenesis ofmarginal zoneB-cell lymphomas. ClinCancer Res 2014;20:5207–16.

20. Tsukasaki K, Tobinai K. Human T-cell lymphotropic virus type I–associated adult T-cell leukemia–lymphoma: new directions in clinicalresearch. Clin Cancer Res 2014;20:5217–25.

21. Bachy E, Salles G. Are we nearing an era of chemotherapy-free man-agement of indolent lymphoma? Clin Cancer Res 2014;20:5226–39.

22. Basso K, Margolin AA, Stolovitzky G, Klein U, Dalla-Favera R, CalifanoA. Reverse engineering of regulatory networks in human B cells. NatGenet 2005;37:382–90.

23. Floratos A, Honig B, Pe'er D, Califano A. Using systems and structurebiology tools to dissect cellular phenotypes. J Am Med Inform Assoc2012;19:171–5.

24. Mani KM, Lefebvre C,Wang K, LimWK, Basso K, Dalla-Favera R, et al.A systemsbiology approach to prediction of oncogenes andmolecularperturbation targets in B-cell lymphomas. Mol Syst Biol 2008;4:169.

25. O'Connor OA. Update in T-cell lymphoma. Clin Adv Hematol Oncol2010;8:604–6.

26. Dunleavy K, Piekarz RL, Zain J, Janik JE, Wilson WH, O'Connor OA,et al. New strategies in peripheral T-cell lymphoma: understandingtumor biology and developing novel therapies. Clin Cancer Res2010;16:5608–17.

27. O'Connor OA, Bhagat G, Ganapathi K, Pedersen MB, D'Amore F,Radeski D, et al. Changing the paradigms of treatment in peripheral T-cell lymphoma: from biology to clinical practice. Clin Cancer Res2014;20:5240–54.





putting the clinical and biological heterogeneity of non-hodgkin lymphoma into context

Documents