It Takes a Village to Unmask HSTL

The 2016 World Health Organization classification divides mature T-cell and natural killer (NK)–cell lymphomas into 28 different subtypes (1). Most subtypes are extremely rare, which has limited efforts to understand their biology and develop targeted therapeutic approaches. In this issue of Cancer Discovery, McKinney and colleagues (2) report the genetic landscape of a poor prognostic subtype of T-cell lymphoma called hepatosplenic T-cell lymphoma (HSTL). The authors collected 68 HSTL specimens from 24 institutes in 8 countries on 3 continents, a gargantuan task of cat-herding that should be enthusiastically commended.

HSTL has several unique features that distinguish it from nearly all other lymphomas. First, HSTLs typically present with hepatosplenomegaly due to infiltration by lymphoma, but not with lymphadenopathy. The bone marrow is almost always involved, but other extranodal lesions are uncommon. Second, most T-cell lymphomas derive from αβ T cells, but the majority of HSTLs derive from γδ T cells. Third, over one half of HSTLs have a characteristic isochromosome 7q that is uncommon in other T-cell and NK-cell lymphomas. Finally, HSTL is associated with immune suppression, particularly among young males treated with thiopurines or TNFα-blocking drugs for inflammatory bowel disease.

McKinney and colleagues (2) first performed whole-exome sequencing (WES) on a discovery set of 20 HSTLs as well as paired bone marrow cells, with the latter used as “germline” comparators. WES was then performed on the 48 HSTLs that lacked bone marrow comparators, but only genes that were found to be somatically mutated in the discovery set were considered in the other 48 tumors. Of note, TET2 and DNMT3A are known to be mutated in hematopoietic progenitors from patients with T-cell lymphomas (3). Because McKinney and colleagues used bone marrow cells as germline comparators, it is possible that mutations in TET2, DNMT3A, or other genes were identified in the “germline” material and thus filtered out in some cases.

McKinney and colleagues (2) focused on 13 somatically mutated genes that may be drivers in HSTL. These included STAT3 and STAT5B, which are known to be mutated in HSTLs (4, 5), as well as genes not previously implicated in this disease. Discerning truly somatic mutations with functional consequence, especially in the absence of germline material, can be quite difficult. For example, several ARID1B mutations reported by McKinney and colleagues were either synonymous mutations, which are likely to have no functional effect, or in-frame insertions/deletions, which can be sequencing errors when found in trinucleotide repeats. Similarly, none of the mutants in isocitrate dehydrogenase 2 (IDH2) reported by McKinney and colleagues involved codons encoding R140 or R172. Mutant proteins with substitutions at these codons are known to generate the oncometabolite 2-hydroxyglutarate (2-HG). In contrast, other mutants that were previously identified in T-cell lymphomas, including codon substitutions and truncated proteins, have been assayed in vitro and do not generate 2-HG (6). Thus, their functional significance, if any, is undefined.

The most notable findings in this new landscape of HSTL genetics were frameshift, nonsense, and clustered missense mutations in SETD2. These were identified in about 25% of cases, many of which had more than one SETD2 mutation. SETD2 is believed to be the only mammalian methyltransferase that trimethylates lysine 36 on histone H3. The resulting H3K36me3 mark has a variety of effects on transcription, splicing, DNA damage repair, DNA methylation, and heterochromatin stability (reviewed in ref. 7). Interestingly, a subset of SETD2 mutations clustered in the C-terminal SET2 RBP1 Interacting (SRI) domain (2). These mutations could confer a neomorphic function such as dominant-negative activity. McKinney and colleagues (2) showed that shRNA-mediated knockdown of SETD2 in a SETD2–wild-type HSTL cell line markedly depleted H3K36me3, promoted a transcriptional signature enriched for gene sets associated with cell-cycle progression, and actually made the cells grow faster.

The combination of mutations in SETD2 and STAT5B (as well as the upstream kinase JAK3) was also recently reported in cases of monomorphic epitheliotropic intestinal T-cell lymphomas (MEITL; previously known as enteropathy-associated T-cell lymphomas; ref. 8). Like HSTL, MEITL primarily involves extranodal sites, has a highly aggressive clinical course, and typically derives from γδ T cells. Thus, both clinical and genetic aspects of HSTL and MEITL suggest that these two diseases share a distinct biology and could even derive from the same precursors.

Further studies are needed to clarify whether SETD2 mutations confer a targetable vulnerability in patients with HSTL. In fact, inhibitors of the WEE1 kinase can selectively kill H3K36me3-deficient cells through deoxynucleoside triphosphate starvation (9). Cells with partial loss of SETD2 function may also be susceptible to inhibitors of the H3K36 demethylase KDM4A, which should increase global H3K36 trimethylation.

With this newfound understanding of genetic mutations in T-cell and NK-cell lymphomas, how do we help patients? The answer, unfortunately, is not so clear. Even for a promising target like mutated STAT5B, chemical inhibition in an HSTL cell line that harbors a clonal activating mutation in STAT5B only partially reduced cell viability (2), so the therapeutic benefit from inhibiting STAT5B signaling remains unclear. Compounding this issue, nearly all recurrently mutated genes in T-cell and NK-cell lymphomas are mutated in less than 30% of cases from any given subtype, and commonly less than 10%. From the perspective of a pharmaceutical company or clinical trialist, there simply are not enough patients to robustly evaluate response rates if only a small subset of those with a rare lymphoma harbor a mutation that may predict therapeutic sensitivity. Finally, most mutations identified in SETD2, STAT5B, INO80, and other potentially “targetable” genes by McKinney and colleagues (2) were subclonal. Even if targeted agents were able to eradicate the mutated subclones, it is hardly guaranteed that the patient would experience any clinical benefit.

On the other hand, recurrent mutations in a subset of cases could indicate that a particular pathway is activated across a broader swath of lymphomas (or subclones of a lymphoma) even in the absence of mutations. As an example, activating mutations in PI3K isoforms occur in less than 10% of T-cell and NK-cell lymphomas, including a subset of HSTLs. Yet, the clinical response rate in a trial of duvelisib, a selective inhibitor of PI3Kγ and PI3Kδ isoforms, was approximately 50% among patients with relapsed/refractory T-cell and NK-cell lymphomas (10). One can speculate that signals within the lymphoma microenvironment drive PI3K signaling through these isoforms and that gain-of-function mutations simply “turn up the gain.” Studies using in vivo models of these lymphomas, such as patient-derived xenografts, may be helpful for teasing out this biology and establishing biomarkers that predict pathway addiction.

In the absence of biomarkers, a certain degree of empiricism remains the only option for most physicians enrolling patients with T-cell and NK-cell lymphomas on clinical trials of investigational agents. As depressing as that reality is, there are several examples where biomarkers can at least identify patients who are very unlikely to respond to a particular agent and therefore should probably be excluded from trials testing that agent. For example, MDM2 inhibitors are largely inactive against tumors with TP53 mutations. Patients receiving therapeutic antibodies are very unlikely to respond if their tumors lack immunohistochemical or flow cytometric evidence of target expression (e.g., CD30 for brentuximab or PD-L1/PD-L2 for inhibitors of this pathway). Functional biomarkers that predict resistance ex vivo may be particularly useful for identifying patients who are unlikely to respond to BCL2 or MCL1 inhibitors.

In conclusion, the genetic landscape of HSTL provides important insights into the complexity of this disease and its similarity to MEITL. The rarity of lymphomas like HSTL, the difficulty in obtaining fresh specimens, and the lack of faithful cell line and in vivo models have undoubtedly dissuaded many researchers from focusing on these diseases. Collaborations like the one published by McKinney and colleagues (2) are essential for defining the breadth of alterations in rare lymphomas like HSTL and will hopefully drive new efforts to understand and target these diseases. Innovative approaches for collecting fresh specimens, such as rapid autopsies and academic–community partnerships, are needed to facilitate the development of cell line and xenograft models. Once those tools are available, we can start defining the particular roles of individual mutations (and isochromosome 7q), their relationships to HSTL phenotypes, the extent to which mutations define biologically distinct subsets of HSTL, the downstream effects of reduced H3K36me3, the relationship between inflammatory bowel disease, immune suppression, male sex, and HSTL, and the translational significance of all of these outstanding issues for patients with HSTL.

D.M. Weinstock reports receiving commercial research grants from AbbVie, Novartis, Aileron, AstraZeneca, and Roche; is a consultant/advisory board member for Novartis, Infinity, and Roche; and has given expert testimony for Monsanto. No potential conflicts of interest were disclosed by the other author.