Genetic lesions in MYC and STAT3 drive oncogenic transcription factor overexpression in plasmablastic lymphoma.

Plasmablastic lymphoma mutational profile is undescribed. Here we performed a targeted exonic NGS analysis of 30 plasmablastic lymphoma cases with a B cell lymphoma dedicated panel and FISH for the detection of MYC rearrangements. A complete phenotyping of the neoplastic and microenvironment cell populations was also performed. We have identified an enrichment in recurrent genetic events in MYC (69% with MYC translocation or amplification and 3 cases with missense point mutations), PRDM1/Blimp1 and STAT3 mutations. These gene mutations were more frequent in EBV positive disease. Other genetic events included mutations in BRAF, EP300, BCR (CD79A and CD79B), NOTCH pathway (NOTCH2, NOTCH1 and SGK1) and MYD88pL265P. Immunohistochemical analysis showed consistent MYC expression, higher in cases with MYC rearrangements together with phospho-STAT3 (Tyr705) overexpression in cases with STAT3 SH2 domain mutations. Microenvironment populations were heterogeneous and unrelated with EBV, with an enrichment of Tumor Associated Macrophages (TAM) and PD1 positive T cells. PD-L1 was expressed in all cases in the TAM population but only in 5 cases in the neoplastic cells (4 out of 14 EBV positive cases). HLA expression was absent in the majority of PBL cases. In summary, Plasmablastic lymphoma mutational profile is heterogeneous and related with EBV infection. Genetic events in MYC, STAT3 and PRDM1/Blimp1 are more frequent in EBV positive disease. An enrichment in TAM and PD1 reactive T lymphocytes is found in the microenvironment of PBL cases, that express PD-L1 in the neoplastic cells in a fraction of cases.

Introduction

Plasmablastic lymphoma (PBL) is an aggressive type of non-Hodgkin B-cell lymphoma defined as a high-grade large B-cell neoplasm with plasma cell phenotype (i.e., loss of B-cell antigens with downregulation of CD20 and PAX5 expression and overexpression of PRDM1/Blimp1 and XBP1s).[1-4]

Epstein-Barr virus (EBV) infection is found in the majority of cases but is not required for the development of a plasmablastic phenotype since clear-cut PBL can be negative for EBV.[3-5] In addition, recent evidence suggests that EBV or human immunodeficiency virus (HIV) status does not influence the gene expression profile patterns of PBL.[6] However EBV positivity in PBL has been found to be associated with increased expression of the programmed death ligand 1 (PD-L1) protein as well as other immune escape markers,[7,8] and decreased expression of major histocompatibility class II (MHCII)/human leukocyte A (HLA)-DR molecules by the neoplastic cells.[9] This has recently been found to be associated with increased antiviral cytotoxic immunity involving different immune cell populations.[7]

The genetic landscape of somatic mutations in PBL is unclear. So far, MYC-IGH translocations have been the most commonly detected alterations, being present in 60% of cases.[10,11] Concurrent mutations in PRDM1/Blimp1 have been found in half of these cases.[12] Very recently, exome sequencing of a series of HIV-positive cases of PBL showed somatic mutations involving components of the non-canonical NFκB pathway as well as genes involved in immune response,[13] but the data remain limited.

Our aim was to characterize the genetic profile of a series of PBL cases using targeted exonic next-generation sequencing (NGS) and correlate the findings with EBV infection and the expression status of immune checkpoint proteins in both the population of neoplastic cells and cells in the microenvironment. In addition, we quantified the components of the microenvironment and searched for skewed T-cell populations in this tumor. We found that the mutational profile of PBL was related to EBV infection in the tumor cells and identified recurrent genetic events in MYC, STAT3 and PRDM1/Blimp1 that were more frequent in EBV-positive disease. In addition, we identified PD-L1 expression on tumor cells in a subset of cases as well as enrichment of tumor-associated macrophages (TAM) and programmed death 1 (PD1) reactive T cells in the microenvironment of PBL cases.

Methods

Case selection

Twenty-eight new cases were retrieved from the files of the Pathology Department of Universitario Marqués de Valdecilla Hospital (Santander, Spain), ten samples from the files of the University of Texas MD Anderson Cancer Center Hematopathology Department (Houston, TX, USA) and four cases from the Pathology Department of San Bortolo Hospital (Vicenza, Italy). Material transfer agreements were signed by the Instituto de Investigación Marqués de Valdecilla (IDIVAL) and corresponding institutions to share the material in the project. The study and sample collection were approved by the local ethics committee (CEIC Cantabria, Institutional Review Board code 2016.168) and complied with the Declaration of Helsinki. All cases were diagnosed according to the World Health Organization (WHO) classification of Hematolymphoid Neoplasms.[14] All cases had to be negative for pan-B-cell markers (CD20), HHV-8 and ALK in order to be included in the study. The phenotype of the cases was consistent with a plasma cell differentiation program.[4,15] The clinical features of the cases were recorded and a summary is available in Online Supplementary Table S1.

Immunohistochemistry and in situ hybridization

Immunohistochemical reactions were performed following conventional automated procedures.Chromogenic in situ hybridization for EBV and its encoding RNA (EBER) and fluorescence in situ hybridization (FISH) for the detection of MYC rearrangements were also done.

Quantification of the cellular composition of the tumor and transcription factor abundance

The different lymphoid and histiocytic/dendritic subpopulations, identified with CD3, CD8, PD1, CD163, PDL1 and MHCII/HLA DP/DR and the absolute number of nuclei showing expression of MYC and phospho-STAT3 (Tyr705) were quantified.

Next-generation sequencing using amplicon-based library generation

DNA was extracted from formalin-fixed paraffinembedded samples using the PicoPure™ DNA Isolation Kit (ThermoFisher Scientific) and was quantified by an Qbit fluorometer (ThermoFisher Scientific). All samples subjected to NGS analysis were required to have >50% of neoplastic cells, identified by morphology (hematoxylin & eosin).

A TruSeq® Custom Amplicon Low Input Library containing exonic regions of 35 selected genes of interest was used to isolate the DNA for sequencing (Illumina). The selected genes were CARD11, ARID1A, NOTCH1, TCF3, SMARCA4, STAT6, EP300, CREBBP, MLL2, BTK, NOTCH2, TNFRSF14, ATM, FOXO1, B2M, PLCG2, CD79B, TP53, STAT3, BCL2, MEF2B, CD79A, CXCR4, PTPN1, MYD88, FAT2, PRDM1, TNFAIP3, SGK1, CCND3, PIM1, EZH2, BRAF, MYC and NOTHC2. Of note, variants occurring in regions outside the coverage of our targeted design were not explored using this approach. Details about library preparation can be found in the Online Supplementary Material.

Sequencing was performed using a HiSeq instrument (Illumina, paired end, 2x150) at the National Genomic Analysis Center (CNAG, Barcelona, Spain).

Sequencing data interpretation and reporting

Only variants in which both libraries had a coverage ≥300 reads and had the same genotype were selected for downstream analysis. Subsequently only missense, frameshift, and nonsense somatic mutations with a variant frequency >10% were considered (Online Supplementary Table S2). Single nucleotide polymorphisms were filtered out using variant allele frequency criteria, and with comparison with dbSNP and an in-house database of germline variants. Finally, 34 somatic mutations (31 missense, 3 nonsense) in 14 genes were considered (Table 1).

Table 1.

Summary of the mutations found in 18 out 30 cases (60%) of plasmablastic lymphoma analyzed by targeted exonic next-generation sequencing.

Further details on the methods are provided in the Online Supplementary Material.

Results

The mutational profile of plasmablastic lymphoma is heterogeneous and correlates with Epstein-Barr virus infection in the neoplastic cells

After targeted NGS with a lymphoma-dedicated panel, somatic missense and nonsense mutations were identified in 18 out of 30 PBL cases (60%). EBV-negative cases tended to show a higher rate of mutations, as compared to EBV-positive cases (87.5% vs. 54%, respectively; c[2] test, P>0.05) (Figure 1).

Figure 1.

Summary of the mutations found in 18 out of 30 cases (60%) analyzed by targeted exonic next-generation sequencing. Epstein-Barr virus (EBV) positivity of tumor cells and human immunodeficiency virus (HIV) infection by the patient are shown, together with the status of the MYC gene as determined by interphase fluorescence in situ hybridization (FISH). The pattern of somatic mutations is heterogeneous with a trend to a higher rate of mutations in EBV-positive cases. The most common genetic events in plasmablastic lymphoma are mutations (including translocations, amplifications and point mutations) in the MYC gene. Previously undescribed abnormalities in plasmablastic lymphoma such as STAT3 (16% of cases), BRAF, MYD88, NOTCH2 and TP53 mutations were also identified (see details in Table 1).

Interestingly the pattern of mutations was also different between EBV-positive and EBV-negative cases. Recurrent somatic mutations restricted to EBV-positive cases were found in PRDM1/Blimp1 in six cases and in STAT3 in five cases. Notably, a recurrent PRDM1/Blimp1 variant, D203E, was identified in four out of six cases, involving the PR domain of the protein.

STAT3 mutations were found in five out of 30 cases (16%), all EBV-positive. Interestingly all but one (STAT3pD566Y) of the mutations involve the SH2 domain of STAT3 protein (STAT3pY640F, STAT3pM648L, STAT3pG618R, STAT3pN647I) (Figure 2) and lead to phosphoSTAT3 (Tyr705) protein overexpression (see below).

Figure 2.

(A) STAT3 mutations were found in five cases (16%), all of which were positive for Epstein-Barr virus. Interestingly all but one (STAT3pD566Y) of the mutations involved the SH2 domain of the STAT3 protein. (B) The mean phosho-STAT3 expression for SH2 domainmutated cases (2 cases with available mutational and immunohistochemical data) was 249 nuclei per high power field (40x), whereas that for STAT3 wild-type cases was 28 nuclei per high power field. Thus, STAT3 SH2 domain mutations led to phosphoSTAT3 (Tyr705) protein overexpression. (C) Representative microphotographs of phosphoSTAT3 (Tyr705) protein expression in plasmablastic lymphoma.

The majority of PBL cases (16 out of 23 tested, 69%) harbored structural abnormalities at the MYC locus. Fourteen cases were found to have a MYC translocation (60%) using break apart probes. MYC-IGH was confirmed in seven of nine cases tested (77%). MYCwas found to be amplified by FISH in two additional cases (Figure 1). Thus, in cases with MYC rearrangements, MYC-IGH was the most frequent alteration. Although there was a clear trend for an association between EBV positivity and MYC rearrangement the difference was not statistically significant (c[2] test, P=0.06).

Furthermore, MYC was found to be mutated in three cases with all but one of the mutations involving exon 2 and consisting of transversions and transitions at C: G pairs (4 out of 7 mutations) (Table 1). Furthermore, the MYCp79S mutation involves the WRCY consensus motif. All these features are consistent with a mechanism related with aberrant somatic hypermutation, as described in early reports.[16]

Mutations common mutations diffuse large B-cell lymphoma (DLBCL), not otherwise specified (NOS), involving B-cell receptor (BCR) activation, TLR/NFκB, histone-modifying genes and the NOTCH pathway were found in eight cases (Table 1, Figure 1). These mutations involved CD79pAW76*, CD79BpD34N, MYD88pL265P, NOTCH1pP401L, NOTCH2pR2400*, SGK1Kp136*and EP300pM2010I/EP300pR1731H. The NOTCH pathway was affected by somatic mutations in NOTCH2 (1 case), NOTCH1 (1 case) and SGK1 (2 cases). Other mutations found were SMARCA4pR1005Q and TP53pR273H. Of note, two cases, both EBV-negative, had mutations in the BRAF gene, one case with the canonical activating BRAFpV600E mutation and the other with a BRAFpG469A mutation in the ATP binding site.

STAT3 mutations are associated with constitutive phospho-STAT3 (Tyr705) activation and MYC protein overexpression is related to MYC rearrangement status

Expression of phospho-STAT3 (Tyr705) protein was quantified immunohistochemically in 20 cases with available mutational data. Mean phospho-STAT3 expression was 48 nuclei per high power field (HPF; 40x) in these 20 cases. Mean expression for two out of four SH2 domainmutated cases with available immunohistochemical data was 249 nuclei per HPF. Mean phosho-STAT3 expression for STAT3 wild-type cases was 28 nuclei per HPF. Mean phospho-STAT3 expression for the single non-SH2 STAT3- mutated sample was 40 nuclei per HPF. Thus, STAT3 SH2 domain mutations (STAT3pY640F, STAT3pM648L, STAT3pG618R, STAT3pN647I) were associated with overexpression of phospho-STAT3, as determined by immunohistochemistry of tissue samples (Figure 2B).

MYC protein was consistently expressed in all the cases (range, 59-236 nuclei per HPF; mean 236), irrespective of the presence of MYC translocations, as previously reported. [12,17] However, significant differences in the level of MYC expression were found, according to MYC gene status. MYC-translocated (14 cases) and -amplified cases (2 cases) had, as expected, higher MYC protein expression than cases without MYC rearrangements (7 cases). The mean number of positive nuclei per HPF was 109 in nonrearranged cases versus 282 in MYC-rearranged cases (Mann-Whitney test, P<0.0001) (Figure 3).

Figure 3.

MYC protein expression in plasmablastic lymphoma. (A) MYC protein was consistently expressed in the cases of plasmablastic lymphoma. (A) Mean MYC protein expression in 22 cases with available data was 236 nuclei per high power field (HPF), which was significantly higher than the mean of 48 nuclei per HPF in the case of phospho-STAT3 protein expression (Wilcoxon test, P<0.001). MYC translocated cases (n=14) and MYC amplified cases (n=2) had higher MYC protein expression than cases without MYC rearrangements (n=7) (Mann-Whitney test, P<0.0001). (C) Representative microphotographs of MYC protein expression in plasmablastic lymphoma.

Mean MYC protein expression in 22 cases with available data was 236 nuclei per HPF, which was significantly higher than the mean 48 nuclei per HPF in the cases of phospho-STAT3 protein expression (Wilcoxon test, P<0.001). There was no correlation between the levels of expression of the two proteins (Pearson test, non-significant). Due to the high prevalence of MYC translocations and amplification in PBL and the relatively low levels of phospho-STAT3 expression and absence of correlation between the proteins, it is unlikely that STAT3 activation contributed to MYC overexpression in most cases. However, one of our cases with STAT3 SH2 domain mutations and absence of MYC translocation by FISH showed high levels of both phospho-STAT3 and MYC proteins, without detectable PRDM1/Blimp1 mutations, suggesting that MYC overexpression might be related with STAT3 activation by mutations in rare cases of PBL.

In summary, MYC protein overexpression is due to rearrangements involving MYC in a significant proportion of cases of PBL (69% in our series). Most translocations fuse MYC to IGH and a few cases may show amplifications of the MYC gene. Both alterations lead to MYC protein overexpression. Genetic alterations in the MYC regulatory domains of PRDM1/Blimp1 may also contribute to its overexpression.[12] In addition here we show that a fraction of PBL cases has STAT3 activation due to somatic mutations in the STAT3-SH2 domain that may increase MYC expression, as previously described in DLBCL[18] (Figure 4).

Figure 4.

MYC protein overexpression in plasmablastic lymphoma. MYC protein overexpression is due to rearrangements involving MYC in a significant proportion of cases of plasmablastic lymphoma (69% in these series). Most translocations fuse MYC to IGH and a few cases may show amplifications of the MYC gene. In addition, we found that some cases of plasmablastic lymphoma have STAT3 activation due to somatic mutations in the STAT3-SH2 domain which may increase MYC expression.

Phenotype of the immune microenvironment and neoplastic cells in plasmablastic lymphoma

We quantified the expression of CD163 and PD-L1 in histiocytic/dendritic cells in the cases of PBL. The mean expression of PDL1 was 33 nuclei per HPF (range, 1.67-61) and the mean expression of CD163 was 38 nuclei per HPF (range, 2-84) (Figure 5). The correlation between CD163 and PD-L1 expression was statistically significant (Pearson 0.6, P<0.05), suggesting that PD-L1-positive cells are histio cytes in PBL. There was not a significant difference in the content or distribution of CD163 and PD-L1-positive histiocytes between EBV-positive and EBV-negative cases (Mann-Whitney test, P>0.05).

Figure 5.

Phenotype of microenvironmental and neoplastic cells in plasmablastic lymphoma. (A) Scattergram illustrating the mean and range of expression values after quantification of the immunohistochemical expression of CD8, PD1 in lymphocytes and PD-L1 and CD163 in histiocyte/dendritic cell populations. (B) Representative image of a case with a mean of 36 PD-L1-positive non-neoplastic cells. (C) The same case showed a mean of 37 CD163-positive histiocytes. (D) The mean expression of CD8-positive cells in this representative case was 53. (E) PD1 identified a different T-cell subpopulation (mean of 36 PD1-positive cells in this representative example, case n. 25). (F) PD-L1 expression by neoplastic cells was identified in five out of 24 cases evaluated (20%). (G) MHCII protein/HLA (DP, DR) was, in most cases, restricted to histiocyte and endothelial cell populations. (H) MHCII protein/HLA (DP, DR) expression was identified in the neoplastic cells in three out of 25 cases tested (12%). Two of the three cases showed cytoplasmic granular and membranous staining (as illustrated in the figure) and one case had a membranous pattern.

CD8-positive and PD1-positive T-cell subpopulations were quantified. The mean number of CD8-positive lymphocytes was 52 nuclei per HPF (range, 1-117) and the mean number of PD1-positive lymphocytes was 32 nuclei per HFP (range, 0-76). There was a significant difference in the distribution of CD8 and PD1-positive cell subsets (Wilcoxon test, P<0.001) consistent with different cell populations. The Pearson correlation value was however statistically significant (Pearson 0.59, P<0.05). There was no significant difference in the content and distribution of CD8 or PD1-positive lymphocytes between EBV-positive and EBV-negative cases (Mann-Whitney test, P>0.05) (Figure 5).

PD-L1 was expressed by tumor cells in five out of 24 (20%) cases evaluated (mean 59 nuclei per HPF; range, 25-98). Four out of five PD-L1-positive cases (in the neoplastic cells) were EBV-positive. Fourteen EBV-positive PBL cases were negative for PD-L1 in the tumor cells. Thus four out of 18 (22%) EBV-positive PBL cases were PD-L1-positive, while one out of six (16%) EBV-negative cases was PD-L1- positive. Thus, there was no association between EBV infection by tumor cells and PD-L1 expression, since most of the EBV-positive cases were PD-L1-negative (P=nonsignificant) (Figure 5). Interestingly one case with STAT3 SH2 mutations showed concurrent PD-L1 and phospho- STAT3 (Tyr705) expression. PD-L1 expression data were not available for the other STAT3 SH2-mutated cases to test this association.

Consistent with previously published data,[9]MHCII protein/ HLA (DP, DR) was virtually absent in PBL. Only three cases out of 25 tested were positive (12%, mean 349 nuclei per HPF; range, 284-440). Two cases showed a membranous and cytoplasmic granular pattern and the other a membranous pattern. All three cases were EBVpositive. The other 22 cases were completely negative for HLA expression in tumor cells (Figure 5).

Discussion

In this study we characterized the genetic profile of a series of cases of PBL using targeted exonic NGS, any correlations with EBV infection and the expression of immune checkpoint proteins in both the neoplastic population and tumor microenvironment. We found that genetic abnormalities (including translocations, amplifications and point mutations) in the MYC gene were the most common genetic event in PBL. In addition to previously described translocations, involving IGH and MYC,[10,11] here we found that a few cases may have MYC amplification, confirming our previous observations.[12] Both MYC translocations and amplifications lead to a significantly increased expression of MYC protein. Interestingly we also identified MYC point mutations, mainly consisting of transversions and transitions at C:G pairs and involving exon 2 and, in the case of MYCp79S mutation, the WRCY consensus motif. All these features are consistent with a mechanism related to aberrant somatic hypermutation.[16] The oncogenic effect of these point mutations does, however, remain unclear.

We also found that 16% of our cases (5 cases) carried recurrent somatic mutations in the oncogene STAT3, preferentially involving the SH2 domain of the protein. Interestingly these mutations were restricted to EBV-positive PBL. Here we demonstrate that these mutations led to increased expression of phospho-STAT3 (Tyr705).

STAT3 mutations and phospho-STAT3 overexpression have been found very rarely in DLBCL NOS (6% according to Ohgami et al.[19]). In cases of ALK-positive large B-cell lymphomas, which commonly show a plasmablastic phenotype, phospho-STAT3 expression has been found to be associated with the presence of ALK rearrangements and overexpression.[20] Importantly, STAT3 activation, due to somatic mutations in the STAT3-SH2 domain may contribute to MYC overexpression, as previously described in DLBCL.[18] In addition, one case in our series showed concurrent STAT3 SH2 mutations and phospho-STAT3 (Tyr705) expression and PD-L1 overexpression, confirming previous results in other lymphoma types suggesting that STAT3 activation triggers PD-L1 overexpression.[21]

STAT3 somatic mutations in PBL have not been previously described so far and may have therapeutic implications for the clinical testing of STAT3 inhibitors in these patients.

Interestingly the pattern of somatic mutations in EBVnegative disease was more heterogeneous. Mutations involving BCR activation, TLR/NFκB, histone modifying genes and the NOTCH pathway were found in eight cases (Table 1, Figure 1). MYD88pL265P mutation, involving the TIR domain of the MYD88 gene, has been previously described in activated B-cell-type DLBCL, in primary central nervous system lymphoma and in other DLBCL in immune privileged sites[22,23] as well as in lymphoplasmacytic lymphoma/Waldenström macroglobulinemia[24] and leads to downstream activation of the IRAK4/IRAK1/TRAF6 complex and NFκB activation. The pattern of mutations in CD79A/B in PBL cases was distinct from that found in DLBCL NOS. Mutations in CD79A/B were found located outside the ITAM domains related with constitutive BCR activation in activated B-cell-type DLBCL.[25] NOTCH pathway genes that were mutated were NOTCH2, NOTCH1 and SGK1. NOTCH2pR2400 *is a nonsense mutation that truncates the PEST domain of the NOTCH2 protein and has already been described in Bcell non-Hodgkin lymphomas, including DLBCL NOS.[26] PEST domain-truncating mutations have been found in multiple tumor types and functional studies suggest that this class of mutations can be targeted with Notch inhibitors including γ secretase inhibitors.[27] NOTCH1pP401L was reported in chronic lymphocytic leukemia in a previous study[28] and lies within the calciumbinding EGF-like domains repeat. Mutations in SGK1 involved the SGK1pS451F and SGK1pA380V point mutations and the SGK1pK136* truncating mutation. These mutations have not been previously described in DLBCL NOS.[26] SGK1 has been suggested to be a negative regulator of NOTCH signaling, enhancing NOTCH protein degradation and reducing its activation by γ-secretase.[29] Other mutations found were SMARCA4pR1005Q and TP53pR273H.

Of note MAPK/ERK pathway-activating mutations involving BRAF (BRAFpV600E, BRAFpG469A) were found in two cases, both EBV-negative. BRAF mutations have been observed, rarely, in related neoplasms such as multiple myeloma. Previous studies found BRAF mutations in 4% of cases of multiple myeloma;[30] they were associated with aggressive clinical features, a plasmablastic phenotype and clonal evolution,[31,32] with obvious clinical implications for targeted therapy.

In addition to the genetic profile of the cases, we also explored the composition of the tumor microenvironment and the expression of immune-checkpoint markers in both the neoplastic and other lymphoid and histiocytic/ dendritic populations. Our results confirm those of previous studies showing an enrichment in TAM that express CD163 and PD-L1. The PBL also had a significant population of CD8-positive T cells, irrespective of the almost absent expression of MHCII/HLA by the neoplastic cells.[9] Importantly, together with CD8-positive T cells, there was a distinct population of PD1-positive T cells. In the PBL cases that we studied, EBV did not influence the immune populations, with regards to the content of TAM and CD8-positive and PD1-positive T cells quantified in the tissue. Furthermore, in our series, PD-L1 expression by the neoplastic cells was found in 20% of the cases analyzed, similarly to previously published series,[8] and there was no association between EBV infection by tumor cells and PD-L1 expression, since PD-L1 was found in both EBV-positive and EBV-negative variants and most of the EBV-positive cases were PD-L1-negative. These findings are in agreement with previously published data on PBL, with variable expression of PDL1 ranging from 20 to 44%, by the neoplastic population.[8,33] In our series, however, we did not confirm an association between EBV infection and PD-L1 expression, suggested by others.[8] This difference may be due to a combination of factors, including different clones used for the detection of PD-L1 expression (22C3 clone in this study, SP142 in others[8]) and different quantification and statistical methods used. In addition another biological factor related to the uncommon PD-L1 expression in PBL cases could be related to the usual latency pattern found in these cases, since PD-L1 expression in EBV-positive post-transplant lymphoproliferative disorder has been strongly associated with EBV latency patterns 2 and 334 while PBL cases usually have EBV latency pattern 1.[7] Notably one of our cases points to STAT3 activation as a potential cause for PD-L1 overexpression in PBL. Collectively our results on the microenvironment and immune-checkpoint expression in PBL indicate a potential for immune checkpoint interference in patients with this type of lymphoma.

In summary, in this study we found that the mutational profile of PBL was related to EBV infection in the tumor cells and identified recurrent genetic events in MYC, STAT3 and PRDM1/Blimp1 that were associated with EBVpositive disease. MYC genetic alterations (including translocations and amplification) and SH2 domain STAT3 mutations led to MYC and phospho-STAT3 (Tyr705) protein overexpression, respectively. Other somatic mutations including BRAFpV600E, MYD88pL265P, NOTCH2pR2400 *and TP53pR273H, appeared in EBV-negative disease, suggesting an overlapping mutational profile with both multiple myeloma and DLBCL NOS. Furthermore, the tumor microenvironment in PBL was characterized by an enrichment in PD-L1-positive TAM and PD1 reactive T lymphocytes with expression of PD-L1 by the neoplastic tumor cells in a fraction of cases. Novel molecular targets derived from the present study include MYC and STAT3 activation, MAPK/ERK and NOTCH2 pathway mutations and immune-checkpoint interference.