Genetic landscape of T- and NK-cell post-transplant lymphoproliferative disorders

Post-transplant lymphoproliferative disorders of T- or NK-cell origin (T/NK-PTLD) are rare entities and their genetic basis is unclear. We performed targeted sequencing of 465 cancer-related genes and high-resolution copy number analysis in 17 T-PTLD and 2 NK-PTLD cases. Overall, 377 variants were detected, with an average of 20 variants per case. Mutations of epigenetic modifier genes (TET2, KMT2C, KMT2D, DNMT3A, ARID1B, ARID2, KDM6B, n=11). and inactivation of TP53 by mutation and/or deletion(n=6) were the most frequent alterations, seen across disease subtypes, followed by mutations of JAK/STAT pathway genes (n=5). Novel variants, including mutations in TBX3 (n=3), MED12 (n=3) and MTOR (n=1), were observed as well. High-level microsatellite instability was seen in 1 of 14 (7%) cases, which had a heterozygous PMS2 mutation. Complex copy number changes were detected in 8 of 16 (50%) cases and disease subtype-specific aberrations were also identified. In contrast to B-cell PTLDs, the molecular and genomic alterations observed in T/NK-PTLD appear similar to those reported for peripheral T-cell lymphomas occurring in immunocompetent hosts, which may suggest common genetic mechanisms of lymphoma development.

cell meshworks. One case had a concomitant EBV+ clonal B-cell proliferation. The c-ALCL cases showed diffuse dermal infiltrates of large CD30+ T-cells, many exhibiting prominent nucleoli, and lacked ALK expression. Both ENKTCL were EBV+. One was of cytotoxic T-cell origin involving the breast and the other of NK-cell origin involving the nasal vestibule and lung; both displayed cellular atypia, had high proliferation indices and showed angiodestruction and extensive necrosis. The ATLL presented as lymphomatous disease and the neoplastic cells expressed CD4 and CD25.

clinical characteristics
The detailed clinical features of the 19 T/NK-PTLD are described in Table 1. The median age was 47 (range: 20-83) and 12/19 (63%) patients were male. The most commonly transplanted organ was heart (n = 8), followed by kidney (n = 7), lung (n = 3) and liver (n = 1). Immunosuppressive therapy comprised calcineurin inhibitors (n = 12), corticosteroids (n = 11), azathioprine (n = 10), and mycophenolate mofitil (n = 6); no association between the type of therapy and organ allograft was observed. The median interval between organ transplantation and diagnosis was 7 years (range: 0.5 -18 y). Extranodal disease was seen in 15/19 (79%) cases and 3 (16%) cases were EBV+. The types of therapies for PTLD and response to treatment varied. The patient with lymphomatoid papulosis (case 18) was   Oncotarget 37639 www.impactjournals.com/oncotarget treated with reduction in immunosuppression alone, which led to disease regression. Four patients received a combination of chemotherapy and radiation and 12 received chemotherapy alone. Two patients were referred for palliative care. Clinical remission was achieved in 1 patient with c-ALCL (case 13) who presented with localized disease, after treatment with chemotherapy and radiation; he died of disease unrelated causes 12 years post diagnosis. One patient with EBV+ PTCL (case 6) is in remission 4 years after diagnosis having received chemotherapy with etoposide, prednisolone, vincristine, cyclophosphamide, and doxorubicin (EPOCH) and one individual with PTCL, NOS (case 5) achieved remission after chemotherapy, radiation, and allogeneic stem cell transplantation. The patient diagnosed with ATLL, based on positive serology for HTLV-1, was treated with EPOCH and is alive 2 years after diagnosis. Despite aggressive treatment, however, the median overall survival was 6.6 months (range: 18 days to 12 years).

Genome-wide copy number alterations
Copy number alterations were observed in 12/16 cases (75%; complex = 8, simple = 4). Loss of heterozygosity was noted in 7/16 (44%) cases ( Figure 3F; Table 2). No recurrent alterations were seen in 5 cases of PTCL, NOS. Two of three HSTCL showed 7q gain and 7p loss, indicative of isochromosome 7q. One of two ENKTCL showed loss of 6q. Copy number alterations in 1 of 2 cases of c-ALCL analyzed included 1p gain and Oncotarget 37640 www.impactjournals.com/oncotarget losses at 13q and 16q. No copy number changes were detected in the ATLL and lymphomatoid papulosis.
Disruption of TP53 was identified in 6 of 18 (33%) cases ( Figure 3A, 3B). Biallelic inactivation was noted in 3 cases (co-occurrence of mutation with 17p LOH) and monoallelic missense mutations in two cases. All missense mutations (n = 4) are classified as non-functional or partially functional in the IARC TP53 database (Version R17). [5] Deletion of 17p encompassing the TP53 locus alone was seen in one case.
Alterations of JAK/STAT pathway genes were seen in 5 of 18 cases (28%). STAT3 and STAT5B mutations were mutually exclusive ( Figure 3C-3E). STAT5B mutations were present in one case each of HSTCL and PTCL, NOS. STAT3 mutations were present in one case each of c-ALCL, ENKTCL, and the EBV+ PTCL. LOH of the STAT3 locus on 17q was observed in the two EBV+ PTLDs with STAT3 mutations (EBV+ PTCL and ENKTCL), suggesting biallelic gain of function. These two cases also had loss of 6q encompassing FOXO3, PRDM1 and PTPRK. A JAK3 variant (S568P) adjacent to a previously described hotspot was also present in the STAT3-mutated EBV+ PTCL. [6] Subtype-specific genetic alterations were also identified. One of the two cases of angioimmunoblastic T-cell lymphoma demonstrated known pathogenic mutations in RHOA and FYN, proteins involved in T-cell motility and proliferation and TCR signaling. [7] Importantly, mutations in genes not previously implicated in PTCL pathogenesis were also observed. Missense mutations in MED12 were seen in 3 cases (1 PTCL, NOS and 2 HSTCL), including two novel and potentially damaging variants (R516H and R816Q) and one exon 2 mutation (L36P). Alterations of TBX3 were observed in 3 of 5 PTCL, NOS. A missense mutation in MTOR (W1456L) involving the FAT domain was observed in the EBV+ PTCL. A three base pair insertion mutation was observed in FBXW7 in one PTCL, NOS, which was confirmed by Sanger sequencing (Suppl. Figure 1).

Microsatellite instability
Variants in mismatch repair (MMR) genes were detected in 4 of 18 PTLD (cases 1, 2, 8, and 15, Figure 2). However only one (6%) PTLD (case 2) with a heterozygous PMS2 mutation and PMS2 loss by immunohistochemistry showed evidence of high level microsatellite instability (MSI-H); alterations at 5/5 mononucleotide repeats accounting for 7% (1/14) of all cases analyzed with a panel of microsatellite markers.. The remaining 3 cases showed normal immunohistochemical expression of the MMR gene products.

dIscussIon
Our series of T/NK-PTLD comprised a spectrum of PTCL subtypes that have been reported to occur post organ transplantation and are recognized as monomorphic PTLDs in the 2008 WHO classification [3,4]. Epidemiologic data drawn from national registries have shown a higher frequency of PTCL, NOS, ALCL, HSTCL, and ENKTCL in transplant recipients. [8] Data regarding other PTCL subtypes in the setting of immunosuppression (e.g. AITL, ATLL) are limited. [9,10] Cases of ATLL, however, have been reported in recipients of solid organ allografts in multiple series. [11,12] In fact, immunosuppression has been proposed as a risk factor for neoplastic transformation in HTLV-1 carriers post organ transplantation. [13] Our observations support the broad definition of T/NK-PTLD in the current WHO classification and suggest inclusion of more detailed phenotypic information in PTLD databases, which will allow precise classification of PTCL and help determine the role of immune deregulation in the pathogenesis of different disease subtypes.
The clinicopathological features of T/NK-PTLD are distinct from B-cell PTLDs. In contrast to B-cell PTLDs, which are EBV-driven in 55-65% of cases, only 3 of 19 (16%) T/NK-PTLD in our series showed evidence of EBV infection. [14] Consistent with prior reports, disease presentation was late after transplant, with a median of 7 years between onset of immunosuppression and the development of lymphoma. The time of occurrence of B-cell PTLDs has an apparent bimodal distribution with the largest peak occurring in the first year after transplantation and a smaller peak observed five years after transplantation. [8] While a subset of B-cell PTLDs respond to reduction in immunosuppression, the majority of the T/NK-cell PTLDs do not, and they are associated with poor outcomes despite aggressive treatment. [15] Disease biology, however, appears to mirror the subtype specific differences observed in T-cell lymphomas of immunocompetent individuals. Lymphomatoid papulosis followed a fairly indolent course with survival over 10 years, while ENKTCL and HSTCL were rapidly fatal.
The genomic alterations detected in our cohort were similar to those described in PTCLs occurring in immunocompetent hosts. [16,17] Changes consistent with isochromosome 7q, a characteristic feature of HSTCL, were identified in 2 cases of HSTCL. Loss of 13q, encompassing ING1 and 16q, encompassing the CTCF locus, seen in one c-ALCL, have been reported in nontransplant associated c-ALCL. [18] One of two ENKTCL and the EBV+ PTCL showed loss of 6q. This recurrent abnormality, which is observed in 29-52% of ENKTCL arising in immunocompetent individuals, encompasses loci of the FOXO3 and PRDM1 genes, which have been shown to function as tumor suppressors in ENKTCL and subtypes of PTCL. [16,19,20] In contrast to EBV+ B-PTLD, however, we observed a higher frequency of complex copy number changes in EBV+ T-PTLD. [21] Mutations of epigenetic modifier genes were the most frequent alterations (61%) in T/NK-PTLD and were most commonly observed in cases of PTCL, NOS and AITL. A similar mutational spectrum is frequently observed in PTCL occurring in immunocompetent individuals. [7,22,23] Loss of function mutations in KMT2D/MLL2, a lysine specific histone methyltransferase targeting H3K4 residues that is considered to function as a tumor suppressor, have been reported in PTCL arising in immunocompetent individuals and were seen in a variety of T/NK-PTLD. [24] Missense mutations in TET2, a methylcytosine dioxygenase that plays an important role in active DNA demethylation, are frequent in both PTCL, NOS and AITL (38% and 33-76%, respectively). [23,25,26] Loss of function mutations in DNMT3A, a DNA methyltransferase that is required for genome-wide de novo methylation, are frequent in a variety of malignancies and have been reported in up to 38% of PTCL, NOS and 33%-37% AITL. [23,26,27] In a murine model, the combination of TET2 loss and DNMT3A mutation appeared to cooperate in the development of T-cell lymphomas exhibiting features of AITL. [28] Similar to observations in PTCL, NOS and AITL occurring in immunocompetent individuals, concomitant TET2 and DNMT3A mutations were noted in two cases of PTCL, NOS in our series. [29] Mutations in chromatin modifier genes, including ARID1B, and ARID2, were also common, a finding in line with prior studies of PTCL occurring in immunocompetent hosts. [24] The TP53 pathway is frequently dysregulated in a variety of neoplasms, most commonly by TP53 mutation or loss, and these lesions are associated with aggressive disease. [30] TP53 abnormalities are known recurrent changes in a variety of PTCL, albeit at a low frequency (8-9%) and their prognostic significance is unclear in these neoplasms. [31][32][33] The higher frequency of TP53 aberrations in our series (33%) could reflect the small sample size and future larger studies are warranted to determine if the frequency of this genetic lesion is higher in PTCL arising in states of immune dysregulation.
Oncotarget 37644 www.impactjournals.com/oncotarget Alterations of the JAK/STAT pathway are frequent in diverse subtypes of PTCL of immunocompetent hosts, including ALCL, HSTCL, T-PLL, ENKTCL, and also in LGL. [6,[34][35][36][37] Genetic aberrations activating the JAK-STAT pathway were identified in 28% of T/ NK-PTLD, including PTCL, NOS, HSTCL, c-ALCL, and ENKTCL, and an EBV+ PTCL. As previously observed in PTCL, STAT3 and STAT5b alterations were mutually exclusive in T/NK-cell PTLD as well. [38] We observed an association between EBV infection and STAT3 activation. STAT3 mutations and LOH of the STAT3 locus were seen in 2 of 3 EBV+ T/NK-PTLDs with concomitant loss of 6q encompassing PTPRK. Loss of PTPRK has been associated with constitutive STAT3 activation in ENKTCL. [39] Activation of STAT3 by EBV, even in the absence of genetic lesions, has been shown to induce lymphoproliferative disorders of T-and NK-cells. [40] EBNA-2 is a coactivator of STAT3 transcription and LMP1 is known to constitutively activate STAT3 in B-cells. [41,42] Of interest, IFN-gamma induced by LMP1 has been shown to enhance STAT1 expression in EBV+ B-PTLD. [43] Whether analogous mechanisms are operational in EBV-associated T/NK-PTLD is not yet known. Nonetheless, our findings suggest that targeting the JAK-STAT pathway with currently available drugs or more selective agents might be of therapeutic benefit in a subset of T/NK-PTLD.
Mutations in genes not previously described in PTCL were also observed. Novel and potentially damaging variants in MED12 were observed in two cases and a MED12 exon 2 mutation (L36P), recently described in chronic lymphocytic leukemia, was seen in one case. [44] MED12 is part of the CDK8 subcomplex, which includes CDK8 kinase and cyclin C, of Mediator, a component of the preinitiation complex, with integral roles in transcriptional regulation. [45] The MED12 protein has been shown to regulate JAK/STAT signaling in mouse embryonic fibroblasts [46]. The role of MED12 in normal and neoplastic T-cells, however, is unclear. The functional consequence of mutations of the transcription factor T-box 3 (TBX3), a transcriptional repressor, important for myriad developmental processes, [47] noted in three cases, is also not known at present. A novel missense mutation in MTOR was detected in an EBV+ PTCL. Similar mutations, which have been shown to activate both mTORC1 and 2 signaling in other neoplasms, may contribute to T-cell growth, survival and proliferation, in part by metabolic reprogramming of lymphocytes. [48] Activation of the mTOR pathway has been reported in T-cell lymphomas, but the presence of mTOR mutations has not been systematically investigated in these malignancies. Of interest, rapamycin analogues, which disrupt the mTORC signaling cascade have shown promise in early phase clinical trials of PTCL patients. [49][50][51] Azathioprine use has been associated with an increased risk of hematological malignancies in allograft recipients and individuals with autoimmune diseases. [52] Thiopurines have been implicated in MMR pathway defects possibly due to epigenetic silencing of the MMR enzymes. [53] Borie et al. reported microsatellite instability (MSI-H) in 9% of immunosuppressionrelated lymphomas, including 33% (6/18 cases) of PTCL post solid organ transplantation. [52] Only 1 of 10 patients in our series treated with an azathioprine-based immunosuppressive regimen developed a MSI-H tumor. Although the observed frequency of MSI-H was lower in our series (1/14 cases; 7%) than a prior study, this difference was not statisically significant (p = 0.1).
In summary, our study represents the first survey of the genetic landscape of T/NK-PTLDs and adds to the growing list of genes altered in T-cell neoplasms that affect a variety of T-cell functions, including activation, proliferation, epigenetic regulation and metabolism. The observed clinicopathologic features and genetic aberrations appear similar to those reported for PTCL subtypes occurring in "immunocompetent" individuals, suggesting shared pathogenetic mechanisms. Our findings raise the possibility of novel therapeutic options for these aggressive lymphomas that need to be explored further in pre-clinical studies.

MAterIAls And MetHods case selection
The archives of the Departments of Pathology, Columbia University Medical Center, New York and University Hospitals of KU, Leuven were searched for cases of T/NK-PTLD. Nineteen cases were identified (Columbia University Medical Center, n = 14; University Hospitals of KU, Leuven, n = 5) that had adequate formalin-fixed, paraffin-embedded (FFPE) tissue for analysis ( > 50% tumor, except for one case of lymphomatoid papulosis with a tumor burden of 20%). We defined T/NK-PTLD as T-or NK-cell lymphomas occurring in individuals after organ transplantation, in accordance with the WHO 2008 guidelines. Morphologic and phenotypic data were reviewed and the lymphomas were classified according to current WHO criteria. [1] Patient demographic and clinical data were collected from the electronic medical record. This study was approved by the Institutional Review Board of Columbia University and performed in accordance with the Declaration of Helsinki.

Whole genome copy number and loH analysis
Genome-wide copy number and LOH analysis was performed for 16 cases using the Affymetrix OncoScan FFPE assay (Affymetrix, CA, USA). Briefly, the Oncoscan Oncotarget 37645 www.impactjournals.com/oncotarget assay has been optimized to work with DNA from FFPE tissue, utilizing molecular inversion probes to identify changes in copy number and loss of heterozygosity. [54] DNA extraction, sample preparation, hybridization, and scanning were performed according to the manufacturer's specifications. Analysis was performed using the Affymetrix Chromosome Analysis Suite 2.0 (ChAS) and Nexus Copy Number 7.5 software (Biodiscovery, Inc. CA, USA).
All copy number alterations and regions of LOH called by the software were verified manually to determine erroneous calls and identify clonal gains and losses undetected by the software(s). Analysis was restricted to gains and losses > 1 Mb in length. Genomic alterations are reported based on the NCBI build 37 (hg19) of the human genome and cancer-associated genes were curated from the Cancer Gene Census (COSMIC v61 Release; http:// www.sanger.ac.uk/genetics/CGP/Census/).

Next generation sequencing analysis
Targeted next generation sequencing of 18 tumors and 9 matched normal samples was performed using a panel comprising 465 cancer-associated genes (Suppl. Table 3). Fifty to 250 ng of DNA, extracted using the Qiamp mini kit or the Qiamp FFPE kit (Qiagen, Germantown, USA) was fragmented to a median of 150-200bp, by sonication. Following end-repair and 3' adenylation of the fragments, and ligation of doublestranded sequencing and indexing adaptors to both ends, target capture and enrichment was performed with the Sure Select Hybrid Capture system (Agilent Technologies, Santa Clara, USA), using custom designed probes. Libraries were then quantified using qPCR, diluted to 2nM and pooled, prior to cluster generation and analysis on Illumina HiSeq2500, using Illumina TruSeq v3 chemistry (San Diego, USA) and 100bp paired-end reads (up to 9 indexed samples per run). Fastq files (more than 70% of reads with > Q30) were demultiplexed with CASAVA, and samples with at least 6Gb of sequencing data were used for mapping and variant calling using NextGene Software (Softgenetics, State College, USA). Reads were aligned to the hg19 (GRCh37) reference genome. At least 3 variant reads were required to call a variant. All calls with ambiguous alignments were excluded, including calls corresponding to regions with large indels. Candidate variants at very low allele frequency ( < 10% of variant reads) were eliminated.
After annotation, the variants were cross referenced with those in the 1000 Genomes Project, OMIM, dbSNP, and the Exome Variant Server. Variants with an allele prevalence > 1% in the 1000 genomes project were excluded. Common variants present in our departmental database of variants identified in prior constitutional exome analysis, non-pathogenic variants reported in dbSNP, and low quality calls were filtered out. The remaining variants were submitted for manual curation and variant prioritization with visual review of alignments. Synonymous variants and intronic variants greater than 2 bp from the coding sequence were excluded. Variants were manually cross-referenced with the Catalog of Somatic Variants in Cancer (COSMIC) and those that were not known "hot spot" mutations or had not been previously reported as potential driver variants were analyzed by PROVEAN and SIFT algorithms. For cases with matched normal specimens, variants present in the normal samples were excluded.

Assessment of microsatellite instability
Microsatellite instability (MSI) testing was performed in 14 cases using a fluorescent PCR-based assay (MSI Analysis system, Version 1.2, Promega, Madison, WI). Briefly, this test assessed 5 mononucleotide repeats (BAT25, BAT26, NR-21, NR-24, and MONO-27) and 2 pentanucleotide repeats (PentaC and PentaD) on genomic DNA and matched normal, where available. The fluorescently labeled PCR products were analyzed by capillary gel electrophoresis. MSI was determined if the tumor alleles showed a size difference ≥3 bp. Tumors with 2 or more microsatellite unstable markers were classified as MSI-H; the remainder were classified as microsatellite stable (MSS). Immunohistochemistry was performed for MLH1 (clone M1, predilute, Ventana), MSH6 (clone 44, predilute, Ventana), MSH2 (clone G219-1129, predilute, Cell Marque) and PMS2 (clone EPR3947, predilute, Cell Marque) using an autostainer (Benchmark Ultra TM Ventana, Tuscon, AZ) after heat induced antigen retrieval (EDTA, pH 7.3). The ultraView TM Universal DAB Detection kit (Ventana) was used for visualization. Absence of nuclear staining was evaluated in lesional tissue; adjacent stromal cells were used as an internal control.
Methods for DNA extraction, immunohistochemistry, flow cytometry, and Sanger sequencing are described in the supplemental methods.