Pathobiology and Molecular Profiling of Peripheral T-Cell Lymphomas

Malignancies derived from mature (post-thymic) T cells and NK cells, collectively referred to as peripheral T-cell lymphomas (PTCLs), encompass a variety of uncommon and rare diseases, altogether accounting for less than 15% of all non-Hodgkin lymphomas (NHL) worldwide.¹ There is an overall higher frequency of PTCL in Asia and Central/South America than in western countries, especially of those induced by the human T-lymphotropic virus-1 (HTLV1) and by the Epstein-Barr virus (EBV). Irrespective of their pathobiological heterogeneity, PTCLs are, with few exceptions, aggressive diseases with poor prognosis.

The WHO classification of PTCLs—like that of other hematological neoplasms—relies on a combination of morphologic, immunophenotypic, genetic and clinical features, and attempts to correlate disease entities with the normal cellular counterpart.² At variance with B-cell lymphomas where specific translocations have been instrumental in defining lymphoma entities, virtually no recurrent genetic or molecular lesions—with the notable exception of anaplastic lymphoma kinase (ALK) rearrangements in ALK-positive anaplastic large cell lymphoma (ALCL)—have been identified in PTCLs. Moreover, distinct T/NK-cell tumors have a broad range of cellular composition, distinct PTCLs diseases lack distinct immunophenotypic profiles, and there is significant morphologic and immunophenotypic overlap across different entities. Conversely, the clinical features and anatomic location of the disease are critical in defining lymphoma entities, and in the WHO classification PTCLs are listed according to their presentation as disseminated, predominantly extranodal or cutaneous, or predominantly nodal diseases. The list of entities according to the updated WHO classification is given in Table 1 . With respect to the histogenetic derivation, the T-cell system is complex, with only few phenotypic and no genotypic markers of normal T-cell maturation, and numerous functional subsets. Thus, whereas most B-cell lymphomas are assigned to a normal cellular counterpart, the cellular origin of many PTCL entities remains more ambiguous (Figure 1; see Color Figures, page 502).

In addition to the rarity, diversity and heterogeneity of PTCLs hampering the collection of large cohorts of patients, the analysis of tumor samples is often contaminated by numerous reactive T cells, few cell lines are available and good animal models are lacking. Therefore, our understanding of the pathobiology of PTCLs remains far behind the accumulated knowledge on B-cell malignancies, and a comprehensive pathophysiological scheme is lacking for most entities. In recent years, few works have been devoted to the molecular profiling of T-cell neoplasms.^3,–9 Overall, transcriptional profiling studies have evidenced distinct signatures in precursor T-lymphoblastic and mature T-cell lymphomas, supporting the notion of biologically unrelated categories of tumors,^3,5 and have provided some novel insights into the pathogenesis of PTCL subtypes. In this article we will highlight the contribution of these studies and other recent advances relevant to the pathobiology of PTCL entities. The discussion will be restricted to primarily nodal PTCLs which are the most commonly encountered and extensively studied.

Angioimmunoblastic T-cell lymphoma (AITL), otherwise known as angioimmunoblastic lymphadenopathy with dysproteinemia, is one of the most common PTCL in Western countries, accounting for 25% to 30% of cases.¹ 

AITL affects elderly adults at a median age of 60 years or higher.¹⁰ In its full-blown expression, AITL is a systemic disease manifested by generalized peripheral lymphadenopathy, fever and weight loss. Up to half of the patients have skin rash, and arthralgias. Laboratory tests often disclose immunologic abnormalities including polyclonal hypergammaglobulinemia, and Coombs-positive hemolytic anemia. Many patients have concomitant extra-nodal disease, most frequently involving the spleen, bone marrow, skin, liver and lungs, and the disease is stage III or IV in more than 80% of cases. The course is variable, with occasional spontaneous remissions. The prognosis is dismal, with a median survival < 3 years in most studies. However, AITL is not always lethal; 30% of patients experiencing long-term survival.¹⁰ 

AITL has distinctive pathological features: (1) a diffuse polymorphous infiltrate including variable proportions of medium-sized neoplastic T cells with abundant clear cytoplasm, admixed with small lymphocytes, histiocytes or epithelioid cells, immunoblasts, eosinophils and plasma cells; (2) prominent arborizing blood vessels; (3) perivascular proliferation of follicular dendritic cells (FDCs); and (4) the presence of large B-cell blasts often infected by the EBV, which may morphologically mimic Reed-Sternberg cells. The neoplastic component is often less abundant than the reactive background, and may even comprise only a minority of the T-cell compartment.¹¹ There is a morphologic spectrum. Three architectural patterns have been described: with hyperplastic follicles (pattern I, seen infrequently), with depleted follicles or without follicles (patterns II and III). These patterns associated with increasing numbers of neoplastic cells, are thought to represent progressive stages of the disease.^12,13 According to the cell content, several variants have been recognized, including a variant rich in epithelioid cell clusters, which may be confused with the lymphoepithelioid variant of PTCL not otherwise specified (PTCL, NOS), and a variant rich in large cells. The large cells may be neoplastic T-cells, and/or represent an expansion of B-cell blasts. The cytological grading does not seem to impact the clinical outcome.^10,14 

The neoplastic cells in AITL are mature αβ CD4⁺CD8^– T cells, sometimes with low or heterogeneous CD4 expression.¹⁵ Importantly, several recent studies have provided convergent lines of evidence to indicate that AITL is a neoplasm derived from a peculiar subset of CD4⁺ T cells normally found in reactive germinal centers (follicular helper T cells [TFH]). Normal TFH cells with a CD4⁺/CD57⁺/ CXCR5⁺/CCR7^– immunophenotype are distributed in the light zone of germinal centers where they provide functional help to B cells by inducing expression of the activation-induced cytidine deaminase (AID) critical to the follicular B-cell differentiation.¹⁶ The first hints to suggest a possible relationship to the germinal center were the documentation of BCL6¹⁷ and CD10¹² expression in the majority of AITL cases. Both markers, typically associated with germinal center B cells, have also been documented in a subset of normal T cells with a follicular distribution.¹⁸ Further steps were subsequent to the release of critical information gained from the study of normal T-cell subsets revealing the unique gene expression pattern of CD4⁺/ CD57⁺ germinal center T cells.¹⁶ The most highly up-regulated gene in TFH cells, the CXCL13 chemokine, was found to be strikingly expressed in the majority of neoplastic cells in virtually all AITL cases.^18,19 Additional markers of normal TFH cells, including CXCR5, CD154, programmed death-1 (PD-1), a member of the CD28 costimulatory membrane receptor family, and SLAM-associated protein (SAP), a cytoplasmic adaptor protein, were demonstrated in AITL by immunohistochemistry.^20,21 The ontogenic derivation of AITL from TFH cells was further established, by the demonstration of similar molecular signatures, at a genome-wide level (Figure 2; see Color Figures, page 503).^6,8 

The cellular derivation of AITL from germinal center cells provides a rational model to explain several of the peculiar pathological and biological features inherent to this disease, i.e., the expansion of B cells, the intimate association with germinal centers in early disease stages and the striking proliferation of FDCs. Among the molecular mediators of TFH cells, CXCL13 probably plays a major role. This chemokine, critical in B-cell recruitment into germinal centers and for B-cell activation, likely promotes B-cell expansion, plasmacytic differentiation and hypergammaglobulinemia.

Non-neoplastic cells typically represent a quantitatively major component of AITL. Accordingly, by comparing the expression profiles of highly enriched tumor cell samples to tissues, nearly 90% of the AITL signature was contributed by non-neoplastic cells.⁶ Clinically, the manifestations of the disease mostly reflect a deregulated immune and/or inflammatory response rather than direct complications of tumor growth,¹⁰ supporting the concept of a paraneoplastic immunological dysfunction. Moreover, AITL patients have defective T-cell responses, linked to both quantitative and qualitative perturbations of T-cell subsets.²² The complex pathways and networks and the mediators linking the various cellular non-neoplastic and neoplastic components are only partly deciphered. Lymphotoxin beta demonstrated in AITL tumor cells²³ and potentially released by B cells under CXCL13 stimulation might be involved in inducing FDC proliferation. Vascular endothelial growth factor (VEGF) is overexpressed in AITL and probably acts as a key mediator of the prominent vascularization observed in the disease.^6,8 By immunostaining, both neoplastic cells and endothelial cells are positive for both VEGF and its receptor, suggesting the possibility of some paracrine and/or autocrine loop.^8,24 

EBV⁺ cells are detected in most cases of AITL,²⁵ and it is now established that these EBV-infected cells are B cells, indicating the virus is unlikely to play a primary role in lymphomagenesis.^25,26 Zhou and colleagues recently found that higher EBV viral loads in tissue biopsies correlated with progression of histological patterns, and with B-cell clonality.²⁵ In the latter study, PCR showed the presence of HHV6B in almost half of the cases. The involvement of other herpesviruses, in particular HHV8, appears very unlikely.²⁵ Although viral infection/reactivation likely occurs as a consequence of the underlying immune dysfunction, EBV, and potentially also HHV6B, may through the modulation of cytokines, chemokines and membrane receptors play a role in the development of the tumor microenvironment, ultimately favoring disease progression. Interestingly, HHV6B also has immunosuppressive properties.

The molecular alterations underlying the neoplastic transformation of TFH cells remain unknown. In that respect, genetic studies have provided fairly deceptive information.^27,–30 Clonal aberrations are detected in up to 90% of the cases (reviewed in Dogan et al³¹). The most common recurrent events are trisomies of chromosomes 3, 5 and 21, gain of X, and loss of 6q.^29,31 Surprisingly, results obtained by matrix-based CGH analysis confirmed frequent aberrations but showed little overlap with classical cytogenetic data.³⁰ Chromosomal breakpoints affecting the T-cell receptor (TCR) gene loci appear to be extremely rare: only 1 out of 54 AITLs analyzed in two FISH-based studies had a TCR break, involving the TCR α/β locus.^27,28 

Mutations of p53 are infrequent, and mutations in the 5′ region of BCL6 have not been detected.³² A role for the c-maf transcription factor has been suggested, because its overexpression in transgenic mice induces the development of T-cell lymphomas, and high levels of c-maf have been detected in human AITL tissues.³³ 

In addition to monoclonal TCR rearrangements which can be detected in > 95% of AITL cases, a clonal or oligoclonal expansion of B cells is also evidenced in up to one-third of patients,³⁴ reflecting the expansion of EBV-infected B-immunoblasts with a restricted repertoire.²⁶ Most EBV-infected cells carry hypermutated IG genes with destructive mutations, suggesting alternative pathways for their survival.²⁶ A subset of AITL patients go on to develop an EBV-associated B-cell lymphoproliferation, in most instances an EBV⁺ diffuse large B-cell lymphoma.^13,35,36 Transformation into a large cell PTCL appears to be rare.¹³ Occasionally, AITL may also be complicated by EBV B-cell or plasma cell neoplasms.^13,36 Irrespective of the EBV status, deregulated somatic hypermutation activity may play a role in the pathogenesis of the supervening B-cell lymphoproliferations.²⁶ 

The designation “anaplastic large cell lymphoma” (ALCL) originally applied to lymphomas composed of large anaplastic lymphoid cells, strongly positive for CD30 with a tendency for a sinusoidal and cohesive growth pattern, was later restricted to tumors of T-cell or null lineage, and two major clinical forms were recognized, according to presentation as a primary systemic versus primary cutaneous disease (reviewed in Stein et al³⁷). Primary systemic ALCL shows one main peak of incidence in children and young adults, and another in later adulthood. It accounts for approximately 3% of adult NHL and 10% to 20% of childhood lymphomas, and represents about 12% of PTCLs.³⁸ 

Patients with systemic ALCL usually present with lymphadenopathy, but involvement of extranodal sites, including the skin, bone and soft tissues, is frequent. Systemic symptoms are common, especially fever. More than half of the patients present with stage III or IV disease.

The majority of ALCLs (55% to 85% of cases) express a chimeric protein containing the cytoplasmic portion of anaplastic lymphoma kinase (ALK) as a consequence of a translocation involving the ALK gene, and these have an overall good response to therapy and good prognosis in both adults and children. A smaller subset of cases have comparable morphologic and phenotypic features to ALK-positive ALCL, but lack ALK protein expression and, importantly, have a less favorable prognosis. In the 2001 WHO classification, ALK⁺ and ALK^– ALCLs were listed as two variants of the same disease entity. Since that time, however, several additional distinctive features have been evidenced that oppose ALK⁺ and ALK^– ALCL, and in the 2008 WHO classification they are listed as separate disease entities.² 

ALK⁺ ALCL represents a well-delineated entity, defined by a characteristic genetic alteration consisting of rearrangement of the ALK gene on chromosome 2p23. A variety of translocations involving the ALK gene have been reported (summarized in Table 2A ). The most common is the t(2;5)(p23;q35), which fuses the ALK to the nucleophosmin gene (NPM) on chromosome 5. All translocations juxtapose the cytoplasmic catalytic domain of ALK to a partner protein, forming a chimeric fusion protein that induces constitutive activation of the tyrosine kinase ALK (reviewed in Pulford et al³⁹ and Amin and Lai⁴⁰). The subcellular distribution of upregulated ALK as evidenced by immunohistochemistry depends on the type of translocation (Table 2A  ). ALK translocations are not entirely specific for ALCL, since they are also occasionally encountered in a rare large B-cell lymphoma subtype, and in inflammatory myofibroblastic tumors (reviewed in Pulford et al³⁹).

ALK⁺ ALCL comprises several morphologic variants (Table 2B ), bearing no correlation to the genetic variants of the ALK translocation. All contain so-called “hallmark cells” typical of ALCL, characterized by an eccentric horseshoe-shaped nucleus and a prominent eosinophilic golgi region. The common variant (75%) comprises sheets of large cells, with numerous hallmark cells. In the small cell variant (5% to 10%), the neoplastic population comprises small lymphoid cells with irregular nuclei, and fewer larger hallmark cells, which tend to cluster around vessels. In the lymphohistiocytic variant (10%), closely related to the small cell variant, the neoplastic cells are scattered within a predominant population of reactive histiocytes. More than one variant may be seen in one biopsy (mixed variants), and relapses may reveal morphologic features different from those seen initially.

In addition to expression of CD30 and ALK, the vast majority of cases are also positive for EMA. The tumor cells exhibit an aberrant T-cell immunophenotype with defective expression of many T-cell antigens, and often an apparent “null” immunophenotype. By molecular analysis, however, ALCL can usually be shown to be of T-cell origin with monoclonal rearrangement of the TCR beta and gamma chains.³⁴ CD3 is negative in > 75% of cases, CD5 and CD7 are often lost as well, CD8 is usually negative, while CD2, CD4 and CD45 are positive in a significant proportion of cases. Most cases exhibit positivity for cytotoxic associated antigens (TIA-1, granzymeB, perforin) and accordingly ALCL is thought to derive from activated cytotoxic T cells. Intriguingly, the pattern of expression of chemokine/cytokine receptors in ALCL has features of a Th2 phenotype.⁴¹ 

Numerous studies have proven that NPM-ALK is oncogenic (for review, see Amin and Lai⁴⁰ and Chiarle et al⁴²). In vitro, constitutively activated ALK chimeras induce cellular transformation, enhance cell proliferation and survival, and lead to cytoskeletal rearrangements and changes in cell shape. Oncogenic ALK transformation is mediated by interaction with downstream molecules that engage intracellular signaling pathways, the most relevant and better characterized being the ERK pathway, the JAK3-STAT3 pathway, and the PI3K-Akt pathway. NPM-ALK transgenic mice have been established, which spontaneously develop precursor T-cell lymphomas, and in this in vivo model lymphomagenesis is dependent on STAT3. Interestingly, enforced expression of NPM-ALK in transgenic models also leads to B-cell transformation and B-cell lymphomas or plasma cell neoplasms.

Secondary genetic alterations are common according to a recent CGH-based study,⁴³ with a similar profile in the morphologic variants and irrespective of the type of the translocation partner. Among several aberrations identified, gain of 17p (including a gain of TP53, but with no apparent correlation with protein expression) and loss of 4q seem to be the most characteristic of ALK⁺ ALCL.

Gene expression profiling studies have revealed peculiar molecular features of ALK⁺ ALCL in comparison to ALK^– ALCL.^44,45 Lamant and colleagues analyzed 25 ALK⁺ and 7 ALK^– samples⁴⁵ and found 117 genes defining the ALK⁺ ALCL signature. Among the most discriminant genes were BCL6, CEBPB (CAAT/enhancer binding protein beta) and SERPINA1. Expression of CEBPB has indeed been validated at the protein level in most ALK⁺ ALCL and appears to be a critical ALK-regulated target gene necessary to induce cell transformation and sustain the growth and survival of ALK⁺ ALCL cells in vitro.^46,47 SerpinA1 is also a plasma serine protease inhibitor regulated by ALK and may have an invasion-promoting effect in ALK⁺ ALCL.⁴⁸ 

Overall, ALK^– ALCL has comparable morphologic and immunophenotypic features to ALK⁺ ALCL, except for lack of ALK expression. The expression of pan–T-cell antigens tends to be more frequent in ALK^– ALCL, whereas the expression of cytotoxic markers and of EMA tends to be less frequent than in ALK⁺ ALCL.³⁸ As the morphological variants of ALCL cannot reliably be identified without ALK expression, ALK^– ALCL comprises exclusively cases with classical morphology. ALK gene rearrangements are also absent, and the oncogenic alterations are unknown. There is no association with EBV infection. Chromosomal aberrations are detected by genomic profiling in about two-thirds of the cases and differ from the secondary aberrations seen in ALK⁺ ALCL.⁴³ Transcriptional profiling has not provided a meaningful insight into the pathobiological mechanisms of this lymphoma.⁴⁵ 

ALK^– ALCL tends to occur in older individuals, extranodal involvement is less common, and the clinical course and prognosis are worse in comparison to patients with ALK⁺ tumors, but more favorable when compared to PTCL, NOS patients.³⁸ 

PTCL, NOS is an exclusion diagnosis for cases not fulfilling the criteria for one of the “specific” PTCL types. As such, it is the most common and also most heterogeneous category of PTCL. So far, the numerous attempts to individualize meaningful variants of PTCL, NOS have failed to identify clinically relevant subtypes. Presentation is usually nodal but any site can be affected and extranodal involvement is common. The median age of patients is in the seventh decade, and 65% of the patients have stage IV disease. Blood eosinophilia, pruritis and hemophagocytic syndromes may occur. The clinical course is aggressive, with frequent relapses, and poor overall outcome (20% to 30% 5-year survival).⁴⁹ 

PTCL, NOS typically contain a mixture of small and large atypical cells. The presence of cells with clear cytoplasm, increased vascularization and eosinophilia are frequent features. T-cell-associated antigens are variably expressed with frequent loss of CD7 or, more rarely, of CD3, CD5 and/or CD2. Most cases are CD4⁺CD8^– and are usually non-cytotoxic. EBV is detected in a variable proportion of tumor cells in a small subset of cases.

Several variants of PTCL, NOS have been described according to distinctive morphologic features, but without clear translation into distinct clinico-pathological entities. The lymphoepithelioid variant of PTCL, NOS (Lennert’s lymphoma) has a characteristic background made up of numerous epithelioid histiocytes and consists of small cytotoxic T-cells, most commonly CD8^+.50

The follicular variant of PTCL, NOS named after a pattern of growth intimately related to follicular structures, has been recently described. It comprises cases with a truly follicular pattern, mimicking follicular lymphoma,⁵¹ T-cell lymphomas with a perifollicular growth pattern,⁵² or involving expanded mantle zones.⁵³ In view of the TFH phenotype, the possible relationship to AITL is questionable.⁵¹ A novel recurrent chromosomal translocation t(5;9)(q33;q22) involving ITK and SYK tyrosine kinase has been described in association with follicular PTCLs,⁵⁴ but is otherwise rarely incriminated in non-follicular PTCLs, which may support the concept of a distinct subtype. Interestingly, however, overexpression and activation of SYK have been recently reported as a feature common to most PTCLs, which potentially represents a novel therapeutic target.⁵⁵ 

There have been several attempts to subclassify PTCL, NOS according to immunological features (CD4 versus CD8 subsets, Th1 versus Th2 helper function, cytotoxic phenotype, differentiation stage); these approaches, however, are technically challenging and hampered by the difficulty in assessing marker expression in the tumor cells frequently admixed with a benign immunologically reactive component, and so far have not emerged as providing clinically meaningful groups.^56,57 

In PTCLs, NOS complex karyotypes are common, especially in cases with larger cells. The genetic imbalances observed in PTCL, NOS differ from those observed in AITL and ALCL. By CGH,^29,30,58 recurrent chromosomal gains have been observed in chromosomes 7q (targeting cyclin-dependent kinase 6),⁵⁹ 8q (involving the MYC locus),³⁰ 17q and 22q, and recurrent losses in several chromosomes. In one study, deletions in chromosomes 5q, 10q and 12q were associated with a better prognosis.⁵⁸ 

Gene expression profiling has now been applied to several small series of PTCL, NOS.^4,–7,9 Although the results are somewhat difficult to interpret and compare because of different hybridization platforms and analytical tools, they have been providing some hints for better understanding PTCL, NOS. As expected, PTCLs, NOS are also heterogeneous at the molecular level. Compared to normal T cells, PTCLs, NOS appear to be most closely related to either CD4⁺ or CD8⁺ activated T cells and are characterized by deregulation of genes related to proliferation, apoptosis, cell adhesion, and matrix remodeling, which is similar to many other malignancies.⁷ Importantly, gene expression profiling has identified overexpression of the PDGFRα a in PTCL, NOS, which was confirmed by immunohistochemistry and might have therapeutic implications given the potential sensitivity of this tyrosine kinase to imatinib.^7,60 

Gene profiling has identified distinct gene clusters, reflecting biological features related to the neoplastic cells and/or the reactive cell populations, that may be relevant to the delineation of molecular subgroups. Classification according to differential expression of NF-κB pathway genes may be clinically useful, since overexpression of this pathway was found to correlate with a better outcome.⁴ Conversely, overexpression of a proliferation signature correlated with an adverse prognosis and appears to be inversely related to signature clusters related to inflammatory response and reflective of an abundant reactive background in the tumor.^5,9 

It should also be highlighted that clustering of PTCLs according to their molecular profiles variably correlated with the pathological classification, indicating some overlap between PTCL, NOS, AITL and ALCL, which may reflect the influence of non-neoplastic elements or the existence of common tumor-associated pathways.⁷ More specifically, traces of the TFH signature have been identified amongst CD30^– PTCL, NOS, suggesting that the AITL spectrum may be wider than suspected, while conversely molecular similarities are observed between CD30⁺ PTCL, NOS and ALK^– ALCL (Figure 3; see Color Figures, page 503).⁶¹ 

Acknowledgments

Laurence de Leval is a senior research associate of the Belgian National Fund for Scientific Research (F.N.R.S.).