Introduction

Peripheral T cell lymphomas (PTCLs) are a group of rare lymphomas originating from mature (i.e., post-thymic or “peripheral”) T lymphocytes and NK cells [1]. PTCLs comprise approximately 10–15% of all non-Hodgkin lymphomas and, with the exception of a few relatively indolent entities, generally are aggressive tumors which carry a poor prognosis [2, 3••]. PTCLs most commonly affect adults and the elderly, and overall are more common in Asia than in the West [4,5,6]. One likely contributor to this geographic variability is different exposures to viral infectious agents that are associated with specific PTCL entities, including Epstein-Barr virus (EBV) and human T-lymphotropic virus 1 (HTLV-1) [7].

The 2008 World Health Organization (WHO) Classification scheme broadly grouped PTCLs into four categories based on the predominant clinical presentation: leukemic (disseminated), nodal, extranodal, and cutaneous (Table 1) [5]. It is recognized, however, that in practice PTCLs can be difficult to subclassify due to evolving diagnostic criteria, multiple subtypes, overall rarity, and the relative paucity of known, specific genetic abnormalities in comparison to B cell lymphomas. Even within established subtypes, PTCLs can exhibit pronounced clinical, histological, immunophenotypic, cytogenetic, and molecular heterogeneity. However, despite these challenges, significant molecular advances have increased understanding of PTCL pathobiology and impacted current classification. In the 2016 revision of the WHO classification of lymphoid neoplasms, mature T and NK cell lymphomas and lymphoproliferative disorders will encompass 27 distinct entities in addition to PTCL, not otherwise specified (NOS), comprising T cell lymphomas not meeting criteria for one of the more specific categories [3••]. The current review discusses the genetic findings and classification of some of the more common PTCLs.

Table 1 Classification of mature T and NK cell neoplasms*
Full size table

Angioimmunoblastic T Cell Lymphoma

Angioimmunoblastic T cell lymphoma (AITL) is one of the most common PTCL subtypes, comprising between 22 and 34% of cases, depending on geographic region [12]. AITL most frequently affects middle-aged adults and the elderly, and is characterized by advanced stage disease, generalized lymphadenopathy, hepatosplenomegaly, skin rash, and polyclonal hypergammaglobulinemia. The clinical course generally is aggressive and often is complicated by infection due to disease-associated immunosuppression. AITL demonstrates a T follicular helper cell (TFH) phenotype and typically is associated with B cells infected by EBV; the neoplastic T cells are EBV-negative. Some patients also develop diffuse large B cell lymphoma, which often is EBV-positive [13].

Recent genetic advances have revealed recurrent abnormalities, including both mutations and structural alterations, in a large proportion of AITLs (Table 2). Somatic mutations in TET2, DNMT3A, RHOA, CD28, and IDH2 have been found in varying frequencies in AITL, offering clues into the pathogenesis of this PTCL subtype [14,20,, 15, 16•, 17•, 18•, 19–21, 46]. RHOA G17V mutations have been identified in both AITL (50–71%) and PTCL, NOS (8–18%), and likely serve a dominant negative function, blocking RHOA GTPase activity [16•, 17•, 18•]. Recently, a novel activating variant, RHOA K18N, has been described [22]. IDH2 mutations are relatively specific for AITL, occurring in 20–45% of cases, and particularly involve IDH2 R172 [19, 21]. IDH2 R172 mutations lead to accumulation of the (R)-enantiomer of 2-hydroxyglutarate, which inhibits TET family and other enzymes and leads to alterations in DNA and histone methylation. In a study of 190 PTCLs, Lemonnier et al. found TET2 mutations in AITL (47%) and PTCL, NOS (38%), but not in other PTCLs except for 2/10 enteropathy-associated T cell lymphomas. Cases of PTCL, NOS with TET2 mutations were more likely to express TFH markers and/or share common histologic features with AITL. TET2 mutations in both AITL and PTCL, NOS were associated with advanced stage disease, thrombocytopenia, high International Prognostic Index scores, and shorter progression-free survival times [46]. Of note, mutations involving TET2, DNMT3A, RHOA, and IDH2 often co-occur in the same case, including PTCLs in which all four of these genes are mutated [16•, 17•, 18•, 22]. Data suggest that among these mutations, those involving DNMT3A may be particularly early events, and sometimes are present in cells of non-T cell lineage from the same patient [14]. Although the molecular pathogenesis of AITL and other TFH-derived PTCLs remains incompletely understood, the relationship of the aforementioned genes to epigenetic regulation suggests that disruption of gene expression via epigenetic mechanisms may be involved in disease development and/or progression [3••, 19, 47], and might in part explain observed clinical responses of AITL to epigenetic modifying drugs such as histone deacetylase inhibitors [48].

Table 2 Recurrent genetic alterations in major subtypes of peripheral T cell lymphoma
Full size table

Among structural abnormalities, CTLA4-CD28 fusion genes have been identified in AITL and several other PTCL subtypes, though analysis of different PTCL cohorts has uncovered variation in the frequency of this finding [23, 49, 50]. The resultant fusion protein is thought to transform inhibitory T cell signals into activating signals, and may represent a target for anti-CTLA4 immunotherapy [23]. Copy number alterations and loss of heterozygosity also have been demonstrated in AITL and PTCL, NOS; among genes within regions of frequent copy number gains, poor prognosis was associated with overexpression of CARMA1 at 7p22 and MYCBP2 at 13q22 [24].

As noted above, recurrent genetic alterations in AITL also have been found in cases of PTCL, NOS with a TFH phenotype. While the 2008 WHO classification considered these cases within the spectrum of PTCL, NOS, the 2016 revision will introduce two provisional TFH-related entities: follicular T cell lymphoma and nodal PTCL with TFH phenotype [1, 3••, 46].

Anaplastic Large Cell Lymphoma, ALK-Positive

Anaplastic large cell lymphoma (ALCL) is another of the more common PTCL subtypes, accounting for ∼12% of all cases [12]. ALCL is a general heading comprising several distinct WHO entities based on clinical presentation and presence or absence of ALK (anaplastic lymphoma kinase) gene rearrangements: ALK-positive ALCL, ALK-negative ALCL (upgraded from provisional to definite entity in 2016 revision), primary cutaneous ALCL, and breast implant-associated ALCL (new provisional entity in 2016 revision) [3••, 25, 51]. Systemic ALCL often presents with advanced stage disease, frequently including B-symptoms such as high fevers. All ALCL subtypes share common pathological features, including the presence of morphologically distinctive cells designated “hallmark” cells (Fig. 1a) and consistent expression of the lymphocyte activation marker, CD30.

Fig. 1
figure 1

Anaplastic large cell lymphoma, ALK-negative, with DUSP22 rearrangement. a Photomicrograph demonstrating large, pleomorphic tumor cells, including a characteristic “hallmark” cell (H) with a horseshoe-shaped nucleus in the center of the image. b Fluorescence in situ hybridization image of two tumor cells (T) and one smaller, normal cell (N) using a break-apart probe to the DUSP22 gene region. The normal cell has two red-green fusion signals (f), representing two intact alleles. The tumor cells each have one fusion signal and one separated red-green signal (s), indicating a rearrangement of the DUSP22 gene region. This rearrangement has been associated with a favorable prognosis [29•]

Full size image

ALK-positive ALCL affects a younger patient population than ALK-negative ALCL and other PTCLs, occurring most commonly during the first three decades of life [25]. ALK-positive ALCL was the first PTCL in which a recurrent genetic abnormality was recognized [52]. About half of ALCLs subsequently were found to have ALK rearrangements [25]. The ALK gene was characterized from a t(2;5)(p23;q35) translocation in ALCL, resulting in the fusion of nucleophosmin (NPM1) to ALK and encoding NPM1/ALK fusion transcripts [26]. The t(2;5) translocation is present in approximately 75–85% of ALK-positive ALCLs; more than 20 other ALK partners have been found, accounting for the remaining ALK-positive ALCLs as well as now being recognized in other lymphomas and non-lymphoid cancers [25, 51]. The frequency of the t(2;5) compared to variant ALK rearrangements may be slightly higher in pediatric than in adult ALK-positive ALCLs [53], but identifying the ALK fusion partner is not considered necessary in routine clinical practice.

Both NPM1/ALK and variants lead to expression of ALK fusion proteins with constitutive ALK tyrosine kinase activity [25, 27, 54, 55]. Central to the function of ALK fusion proteins is activation of the downstream oncogenic transcription factor, STAT3 [28•]. ALK-induced STAT3 activation plays a variety of roles in cancer cells, including regulation of availability of hypoxia-inducible factors under hypoxic conditions [56]. ALK tyrosine kinase activity can be targeted by clinically available inhibitors, and pharmacologic ALK inhibition is effective in a variety of clinical scenarios, including relapsed/refractory ALK-positive ALCL [47,58,, 57–59]. Even with conventional cytotoxic chemotherapy, ALK-positive ALCL has a more favorable prognosis than ALK-negative ALCL as a whole (see section below) and most other PTCL subtypes, with overall 5-year survival rates of 70 to 85% [29•, 60, 61]. The ALK partner gene (NPM1 or variant) does not appear to be a significant prognostic factor [62].

Anaplastic Large Cell Lymphoma, ALK-Negative

Systemic PTCLs that are morphologically compatible with ALCL and express CD30, but lack ALK rearrangements, now are considered a separate subtype [3••, 63]. ALK-negative ALCL affects an older patient population than ALK-positive ALCL, and the molecular genetics remain incompletely understood [29•, 51, 63]. Although, as a whole, ALK-negative ALCL generally has been found to have a prognosis inferior to that of ALK-positive ALCL, some studies have found similar overall survival (OS) rates after adjusting for age [60, 64].

Distinguishing ALK-negative ALCL from CD30-positive PTCL, NOS is diagnostically challenging, but is of critical importance due to the potential prognostic and therapeutic implications. Gene expression profiling studies have shown not only that ALK-negative ALCL and ALK-positive ALCL share a common signature, but also that the signature of ALK-negative ALCL is distinct from that of PTCL, NOS [3••, 65•, 66, 67]. A three-gene model (TNFRSF8, BATF3, and TMOD1) has been proposed to help distinguish ALK-negative ALCL from PTCL, NOS in tissue samples, but is not in routine clinical use [68].

Recently, Crescenzo et al. reported that the JAK/STAT3 pathway was constitutively activated via multiple genomic mechanisms in ALK-negative ALCL [28•]. The JAK1 and STAT3 genes each were recurrently mutated in a subset of cases, but were not found in other PTCLs, including ALK-positive ALCL and PTCL, NOS (including 16 CD30-positive cases). In addition, RNA sequencing identified a small subset of ALK-negative ALCLs that lacked these mutations but harbored gene fusions involving non-ALK tyrosine kinase genes, including ROS1 and TYK2, leading to constitutive activation of STAT3. Thus, STAT3 may drive oncogenesis in the majority ALCLs, independent of ALK status.

Recurrent rearrangements of non-tyrosine kinase genes also have been identified in ALK-negative ALCL, notably including rearrangements of the DUSP22-IRF4 locus on 6p25.3 (DUSP22 rearrangements, Fig. 1b) or of TP63 on 3q28 [30, 31]. Parrilla Castellar et al. have shown that these rearrangements have prognostic significance in ALK-negative ALCL [29•]. DUSP22 rearrangements were seen in 30% of ALK-negative ALCLs and were associated with a favorable prognosis (90% 5-year OS), similar to that observed in ALK-positive ALCL (85%). TP63 rearrangements were seen in 8% of ALK-negative ALCLs and were associated with a 5-year OS rate of only 17%; ALCLs lacking rearrangements of ALK, DUSP22, and TP63 (so-called triple-negative cases) had an intermediate 5-year OS rate of 42%. For comparison, PTCL, NOS has been found to have a 5-year OS rate of 32% (and only 19% for cases with CD30 expression in ≥80% of cells) [12].

Scarfo et al. recently identified ERBB4 and COL29A1 as gene expression outliers in 24% of ALK-negative ALCLs [32]. They found that ERBB4 deregulation resulted from expression of two truncated transcripts with intronic transcriptional start sites. Truncated ERBB4 showed in vitro and in vivo oncogenic potential, and an ERBB4-positive patient-derived xenograft model showed sensitivity to pharmacologic ERBB4 inhibition. A genome-wide copy-number study using single nucleotide polymorphism arrays recently reported frequent losses at 6p21 and 17p13, corresponding to the PRDM1 (encoding the tumor suppressor, BLIMP1) and TP53 genes, respectively [33]. PRDM1 was found to be inactivated via multiple mechanisms in 56% of ALK-negative ALCLs and only 6% of ALK-positive ALCLs. Additionally, losses in 6q21 and/or 17p were associated with inferior OS among ALK-negative ALCLs.

It should be noted that systemic ALCL needs to be distinguished from primary cutaneous ALCL (pcALCL), which is an indolent entity with excellent disease-specific survival rates; pcALCL presents in the skin, and although it occasionally involves regional lymph nodes, it rarely disseminates [51, 69, 70]. The reasons that pcALCL is more indolent than systemic ALCL are incompletely understood, but may relate to specific profiles of cytokines and chemokine receptors, partly shared with breast implant-associated ALCLs, that could limit dissemination and suggest distinct initiating events related to chronic antigenic stimulation [71, 72]. Because systemic ALCL can involve the skin secondarily, patients with skin biopsies showing ALCL should undergo staging procedures. Though pcALCL is not specifically covered in this review, it should be noted that it shares some genetic features with systemic ALK-negative ALCL, including DUSP22 rearrangements in 28% [73], TYK2 rearrangements in 13% [74], STAT3 mutations in 5% [28•], and occasional TP63 rearrangements [31]. pcALCL generally is considered an ALK-negative entity, but rare, otherwise similar cases with ALK rearrangements have been reported [75].

Extranodal NK/T Cell Lymphoma, Nasal Type

Extranodal NK/T cell lymphoma, nasal type (NKTCL) is an aggressive lymphoma that most often originates from natural killer (NK) cells, but in some cases is derived from cytotoxic T cells [76, 77]. NKTCL primarily affects adults within Asian and Central and South American populations. The nasal cavity and adjacent structures are the most common sites of presentation, but other extranodal locations may be involved. Patients often present with symptoms of nasal obstruction, epistaxis, or with midfacial destructive lesions (“lethal midline granuloma”). Histologically, NKTCL features an angiocentric and angiodestructive growth pattern. The neoplastic cells are associated with EBV in virtually all cases, suggesting a likely pathogenic role of the virus. The prognosis of NKTCL is generally unfavorable but can vary. Tumors occurring outside the nasal cavity have been found to be highly aggressive and generally do not respond well to chemotherapy. Of note, chemoresistance of NKTCLs is partly attributable to expression of P-glycoprotein, encoded by the multidrug resistance gene, MDR1; drug combinations not affected by P-glycoprotein, particularly L-asparaginase-based regimens, have been associated with improved outcomes [78].

A variety of somatic mutations has been identified in NKTCL, most commonly involving the RNA helicase gene DDX3X (20% of cases) and JAK3 (35% of cases) [36, 37]. Mutations of STAT5B and STAT3 also have been identified in NKTCL and PTCLs derived from γδ-T cells, further emphasizing the role of JAK/STAT signaling in this and related diseases [38]. As discussed above, CTLA4-CD28 fusions may be identified in NKTCLs [23]. Other structural abnormalities include copy number losses at 6q21, including the PRDM1 gene [40].

Peripheral T Cell Lymphoma, Not Otherwise Specified

PTCL, NOS is an intentionally heterogeneous entity, sometimes referred to as a “wastebasket” category, comprising nodal and extranodal PTCLs that do not fulfill criteria for any of the more specific PTCL subtypes [79]. The fact that PTCL, NOS accounts for approximately ∼30% of all PTCLs, making it the most common entity, attests to the need for new molecular and other data to increase understanding of these cases. PTCL, NOS most often presents in the lymph nodes; however, any site can be involved. The most frequent extranodal sites of involvement are the gastrointestinal tract and the skin [79]. As noted above, PTCL, NOS has a generally poor prognosis (5-year OS rate, 32%) [12].

The genetic findings in PTCL, NOS are heterogeneous and demonstrate significant overlap with other PTCL subtypes. As discussed above, this overlap may guide development of criteria to place cases in a more specific entity, such as the new provisional categories for non-angioimmunoblastic TFH-derived PTCLs [3••]. Recent gene expression profiling studies have identified at least two major subgroups of PTCL, NOS that are characterized by high expression of either GATA3 or TBX21, with the former associated with inferior prognosis [3••, 65•, 80]. Poor clinical outcomes also were observed when TBX21-associated cases expressed a cytotoxic gene signature.

Various recurrent mutations have been identified in subsets of PTCL, NOS. Palomero et al. have identified novel mutations in the FYN gene, which encodes a tyrosine kinase that plays an important role in T cell activation [16•]. FYN kinase activity was successfully targeted by a multikinase inhibitor in vitro, thereby displaying promise for a therapeutic approach in the small subset of PTCL patients with these mutations. Several additional recurrent genetic mutations in PTCL, NOS were mentioned in the section on AITL, including the epigenetic modulator genes, TET2, DNMT3A, and RHOA [14, 16•, 17•, 19]. Similarities between AITL and other TFH-derived PTCLs have been noted at the gene expression profiling level as well, though differences also exist [81, 82]. Interestingly, a recurrent t(5;9)(q33;q22) translocation leading to ITK/SYK fusion and overexpression and activation of spleen tyrosine kinase (SYK) has been identified in a subset of PTCLs with a follicular growth pattern and a TFH phenotype, and rarely in AITL [8, 44, 45, 83]. This fusion appears targetable with SYK inhibitors. Overall, these molecular similarities and differences have contributed to the introduction of the aforementioned provisional TFH-derived PTCL categories, which separate these entities from the broader category of PTCL, NOS but also keep them distinct from AITL [3••].

As discussed above, CTLA4-CD28 fusions occur in a subset of cases of PTCL, NOS [23]. Boddicker et al. recently discovered additional gene fusions in PTCL, NOS by integrating mate-pair DNA sequencing and RNA sequencing [34]. Of particular interest were recurrent VAV1 fusions with varying partners, which were found not only in PTCL, NOS (11%) but also in ALK-negative ALCL (16%) and a single case of ALK-positive ALCL. Cells overexpressing a VAV1-GSS fusion protein showed increased growth and were sensitive to azathioprine, an inhibitor of RAC1 activity. Of note, mutations of VAV1 also have been reported recently in various PTCL subtypes [22, 84].

As mentioned above, Fujiwara et al. have identified various loci with copy number alterations in PTCL, NOS and AITL, some of which were found to have prognostic implications [24]. Similar to AITL, overexpression of CARMA1 was associated with poor prognosis in PTCL, NOS. Additionally, a recurrent loss at 9p21 was associated with reduced expression of CDKN2A, CDKN2B, and MTAP. Reduced levels of CDKN2A were found in 9% of PTCL, NOS samples and were associated with inferior outcomes.

Future Directions and Perspectives

The most recent update to the WHO classification highlights several emergent themes with regard to genetics and high-throughput molecular profiling. First, these approaches, particularly analysis of mutational landscapes and gene expression profiling, have identified areas of commonality between subsets of PTCL, NOS and other PTCL subtypes, as in the case of TFH-derived PTCLs and AITL [3••]. As molecular understanding of PTCLs deepens, this trend is likely to continue, hopefully leading to a decrease in the number of patients given the “wastebasket” diagnosis of PTCL, NOS, albeit likely at the price of an increasing number of overall categories. Second, we are likely to deepen our understanding of the molecular heterogeneity within established entities (e.g., ALK-negative ALCL [29•]), giving rise to new prognostic biomarkers and/or tools for risk-stratifying patients for clinical research studies. Third, we are likely to see the introduction of new molecular tools that help refine the borderlands between related entities and aid clinical diagnosis [65•, 68]; digital gene expression analysis [85] is one technology that may find application here. New “omics” methodologies also may impact classification as they are applied more comprehensively, including epigenomics, proteomics, and metabolomics. Finally, increasing understanding of the scope and targetability of genetic alterations will help determine the role of individualized medicine approaches in clinical practice [86]. These approaches may be particularly applicable to complement the evolving molecular classification of PTCLs (compared to B cell lymphomas, for example), given the overall rarity and molecular heterogeneity of this constellation of disorders.

Conclusions

The recent advances described above have provided insight into the genetic heterogeneity of PTCLs. These discoveries have shed light on the classification, prognosis, and potential targeted therapy of several of the more common PTCL subtypes, but there is still much progress to be made. Treatment outcomes for PTCLs have not yet shown significant improvement despite increased molecular understanding, partly due to a relative paucity of candidate therapeutic targets. As knowledge of the genetic drivers of PTCL increases, cooperative group studies, innovative clinical trial design, and biomarker investigations to guide therapy will be critical to make a tangible impact on patient outcomes.