Challenges and opportunities for checkpoint blockade in T-cell lymphoproliferative disorders
Genomic complexity and neoantigen load
In addition to PD-L1 expression itself, the burden of nonsynonymous mutations and neoantigens has emerged as an important biomarker in CPB-treated patients. The frequency of mutations is highly variable across tumor types (and within a given tumor type). Carcinogen-associated tumors, most notably melanoma and non-small cell lung cancer (NSCLC), are associated with both a relatively high frequency of somatic mutations (?10/Mb) and superior response rates to CPB [28] that is likely explained by immune-mediated destruction of neoantigen-expressing tumors [29–32]. For example, in melanoma patients treated with CTLA-4 CPB, a high mutational load was associated with clinical benefit from CPB [31, 32]. The overwhelming majority of patients who derived clinical benefit from CPB had 100 missense mutations, whereas patients who failed to benefit had a significantly lower mutational burden. A similar association between mutational burden and response to PD-1 CPB has been observed in NSCLC [29]. Despite the highly significant association between mutational and neoantigen load and response to CPB, this relationship is not absolute.
In contrast to melanoma and NSCLC, most hematologic malignancies (e.g. acute myelogenous leukemia, chronic lymphocytic leukemia, multiple myeloma) are associated with a lower frequency of somatic mutations (?1/Mb) [28]. Recently performed next-generation sequencing studies highlight the genomic complexity associated with many T-cell NHL. Somatic copy number variants (SCNV), many of which are focal deletions/amplifications, and novel fusion events, are common in CTCL [33–35]. Catastrophic genomic rearrangements in which a chromosomal region(s) is subject to multiple double-stranded DNA breaks followed by random reassembly to form a complex mosaic chromosome, while infrequently observed in most cancers [36, 37], are highly prevalent in CTCL [33]. These studies have also implicated ultraviolet B (UVB) radiation in CTCL pathogenesis [33–35, 38, 39], as a high frequency of C T transitions have been observed, in contrast to the T G transversions observed in B-cell lymphproliferative disorders [40]. Many of these occurred at NpCpC trinucleotides, a signature associated with UVB exposure in melanoma [40]. In addition to this global genomic complexity, nonsynonymous point mutations occur in CTCL at a rate of ? 3 mutations/Mb, a rate that is substantially higher than many other hematologic malignancies [28]. An average of ? 50–100 somatic nonsynonymous mutations are observed in CTCL, but considerable variability with mutation rates exceeding 300/tumor in some cases have been appreciated [33–35, 38, 39, 41]. By comparison, the mutation rate (?25/tumor) appears to be lower in PTCL [42, 43]. While not uniformly reported, these studies suggest that the mutational load varies by histology and disease stage. For example, a nonsynonymous mutation rate exceeding 5/Mb was observed in 29% (5/17) of CTCL tumors that had undergone large-cell transformation [38]. This mutation rate exceeds that observed in most lymphoproliferative disorders, and approaches that observed in many melanomas [28]. Therefore, distinct subsets of T-cell NHL likely possess a mutational burden that rivals that associated with clinical benefit from CPB in solid tumors like melanoma and NSCLC.
In addition to mutational burden and the presentation of clonal neoantigens [30], specific mutations, including those that are prevalent in selected T-cell NHL, may confer susceptibility to CPB. In an effort to understand the mechanisms involved in acquired resistance to CPB, Zaretsky et. al. performed whole-exome sequencing in paired initial and relapsing biopsies from four melanoma patients. Loss-of-function mutations in JAK1 and JAK2, associated with diminished responsiveness to interferon-? induced PD-L1 expression, were observed in two patients [44]. Conversely, amplification of the JAK2 locus and enhanced JAK/STAT signaling promotes PD-L1 expression in selected B-cell malignancies [6, 45]. Therefore, gain-of-function mutations and other genetic alterations that promote JAK/STAT signaling in selected PTCL subtypes may confer susceptibility to CPB [reviewed in [46]]. Of course, the incorporation of genomic data in future and ongoing CPB trials will be required to address this question and may improve the ability to identify “exceptional responders”.
In contrast to other tumor types, including the B-cell lymphomas, the T-cell lymphomas are derived from the very cell types (e.g. effector T cells) that are required for CPB-mediated tumor eradication. Consequently, when considering CPB within this context, “the world turned upside down”. For example, recurrent mutations in both the CD28 extracellular and intracytoplasmic domains have been observed in a minority of PTCL [42, 47], and more recently in MF/SS [33, 38]. Mutations in the extracellular domain increase the binding affinity to CD28 ligands, particularly CD86 [33]. Whether CTLA-4 blockade, by increasing the availability of CD28 ligands within the tumor microenvironment, may promote the growth of malignant T cells harboring these rare mutations is unknown. Similarly, a rare in-frame fusion between the extracellular and transmembrane domain of CTLA-4 and the intracytoplasmic domain of CD28 has been observed, and presumably exploits the high-affinity ligand-binding domain of CTLA-4 to activate CD28 signaling [38, 39, 48]. A rapid, transient response was observed in one patient harboring this novel translocation following therapy with ipilimumab [48]. While PD-1 is highly expressed is many CTCL and TFH-derived PTCL [25, 49], it is also recurrently deleted in a subset of CTCL [38], but the implications of these observations for PD-1 CPB are uncertain.