Presence of Epstein–Barr virus (EBV) antigens detected by sensitive methods has no influence on local immune environment in diffuse large B cell lymphoma

A total of 48 DLBCL patients, previously characterized [6], were enrolled in this study, 26 pediatrics and 22 adults, collected retrospectively, based on the availability of sufficient material, from the archives at Pathology Division, at the Ricardo Gutierrez Children’s Hospital, and at the Marie Curie Hospital in Buenos Aires, Argentina. The samples were reviewed independently by 4 different pathologists: DDSM, GL, CS, and DME, according to the WHO classification for lymphoid neoplasms criteria [20]. The age range was 0–73 years (total median: 16 years, pediatrics median: 9 years, and adults: median 55 years). Institutional guidelines regarding human experimentation were followed, in accordance with the Helsinki Declaration of 1975. The Ricardo Gutierrez Children’s Hospital Ethics Committee (CEI) approved the study, and all the patients or patients’ guardians gave informed consent for the study.

EBERs in situ hybridization (ISH) was performed in formalin fixed paraffin-embedded (FFPE) tissue sections, using fluorescein isothiocyanate (FITC)-conjugated EBERs oligonucleotides as probes (Dako). A monoclonal antibody anti-FITC labeled with alkaline phosphatase was used to detect hybridized sites. In each hybridization run, an EBV-positive Hodgkin lymphoma FFPE paraffin-embedded tissue block was used as a control slide. A cutoff 20% of positive tumor cells we considered to define EBV+ DLBCL, NOS, was described [21].

Double in situ hybridization with ViewRNA ISH Tissue 2-Plex Assay (Thermo Fisher, formerly Affymetrix) was performed in 36 FFPE DLBCL cases with good quality material, as previously reported [6]. Custom-specific probes sets against LMP1 (probe type 1) and EBNA2 (probe type 6) transcripts were customized by the manufacturer (Thermo Fisher, formerly Affymetrix), based on LMP1 and EBNA2 mRNA sequences. In order to define the presence of viral traces, the tumor cells that expressed LMP1 and/or EBNA2 transcripts in tumor cells were observed, and counted 100 × objective lenses in ten fields selected on the basis of the best-preserved tissue areas that contained positive cells, expressed per area as cells + /mm².

Total RNA was isolated from FFPE DLBCL biopsies with RNeasy FFPE kit (Qiagen, Valencia, MA, EE. UU.), according to the manufacturer’s instructions, and quantified with NanoDrop 2000/2000C (Thermo Fisher Scientific, Waltham, MA, USA). The integrity and quality of the RNA were determined using TapeStation 4200 RNA ScreenTape kit (Agilent Technologies, Santa Clara, CA, USA), following the manufacturer’s instructions.

The Research Use Only Version of the NanoString XT assay along with the nCounter Flex Analysis System (NanoString Technologies, Seattle, WA, USA) was used to analyze gene expression. A customized 208 gene panel, listed in Supplementary Table 1, was used to evaluate gene expression. The gene panel included genes expressed by various components of the stroma and tumor cells in neoplastic diseases, as well as numerous genes known as therapeutic targets, in addition to a set of 8 housekeeping genes to normalize the values ​​of gene expression.

The probes were hybridized with 200 ng of total RNA for 16 h at 65 °C. The excess was removed, and immobilization of the probe transcription complexes was performed on a streptavidin-coated cartridge in the nCounter Preparation Station. Gene expression values ​​were normalized with housekeeping genes. Normalized count data were log2-transformed, and agglomerative hierarchical clustering of gene expression was performed. The data were analyzed using nSolver Analysis Software 4.0 (NanoString Technologies).

Antibodies against TIM3 (polyclonal, Abcam), LAG3 (clone [11E3], Abcam), CD68 (clone KP-1, Roche Ventana) and CD163 (clone MRQ-26, Roche Ventana) were used, as described [10, 14], to characterize microenvironment composition and to validate gene expression analysis. IHC detection of antibodies was carried out using a universal streptavidin–biotin complex-peroxidase detection system (UltraTek HRP Anti-Polyvalent Lab Pack, ScyTek, UT) according to the manufacturer’s instructions. Visualization of positive cells was performed using diaminobenzidine as chromogen. Appropriate positive controls were immunostained for each antibody. The counting of TIM3, LAG3, CD68 and CD163 positive cells was performed as follows: using the 100 × objective lenses and counting ten fields selected on the basis of the best-preserved tissue areas that contained immunopositive cells. The positive cells in tumor cells and in tumor-infiltrating lymphocytes (TILs) were counted, and the results were expressed as positive cells per area unit (cells + /mm2). Cells partly included in the fields were not counted.

To differentiate the expression of PDL1 in tumor cells or at the microenvironment immunohistochemical double staining with PAX5 was carried out, to differentiate PDL1 + tumor cells (PAX5 +) from PDL1 + cells at the microenvironment (PAX5-), adapting the technique previously described [14]. 5-μm-thick FFPE tissue sections for all cases were used. Slides were incubated with anti-PAX5 (clone SP34, Dako) rabbit monoclonal antibody for 1 h at room temperature. PAX5 was detected with the Vectastain-ABC- Peroxidase kit, using diaminobenzidine (DAB) as chromogen. Then, the slides were incubated with a second primary antibody anti-PDL1 (clone 210,934, Abcam) diluted in 1/100 TBS-BSA 2%. Finally, the second primary antibody PDL1 was incubated overnight, and antibody signal amplification was performed using Alexa 594 anti-mouse, followed by Hoechst.

For double staining analysis, ImageJ imaging program was used. First, the positive cell counts of individual markers were made (PAX5 and PDL1). Then, pictures were merged and double-positive or single-positive cells were counted (PAX5 + PDL1 + and PAX5-PDL1 +). Afterward, positive cells were averaged and expressed per area unit (cells + /mm2). We defined double-positive (PAX5 + PDL1 +) for tumor cells expressing PDL1, PDL1t, and single-positive (PAX5-PDL1 +) for microenvironmental cells, PDL1m, in DLBCL, as described [8].

Data were analyzed with GraphPad Prism5 (GraphPad Software Inc., San Diego, California, USA). Normality test was performed using Shapiro–Wilks’s test. Comparison of the means of cell counts by IHC was assessed by one-way ANOVA or Kruskal–Wallis tests for more than 2 groups, or by t test or Mann–Whitney, for 2 groups, based on normality test results between the groups according to EBV+ and EBV− cases, and cases with and without viral traces (positive or negative for LMP1 and/or EBNA2 transcripts). Outliers were defined using the robust test to compare data median absolute deviation (Mad) in Excel. For gene expression analysis, to test for statistical difference in medians, Mann–Whitney U tests were performed. To test if medians were different to zero, one-sample Wilcoxon signed rank tests were performed. Both types of tests were performed using wilcox.test function from the stats package in R [22]. p < 0.05 was considered statistically significant.

Forty-eight DLBCL cases were included in this analysis. Twenty-six cases were pediatric and 22 were adults, and median age at diagnosis was 16 years. There was a male prevalence (52%), with 25 males and 23 females. On the basis of Hans’ IHC classification [23], 24 cases (50%) were classified as non-GC subtype, 20 (42%) as GC subtype, while in the remaining 4 cases (8%), the data were not available.

EBERs expression by ISH is summarized in Supplementary Table 2. Of the 48 DLBCL cases, 11 were positive, with ≥ 20% EBERs + cells (Supplementary Fig. 1A), as previously established in our population [21, 24]. The remaining 37 cases were negative by EBERs ISH, the gold standard to define EBV association with lymphomas.

Since traces of EBV infection were detected by sensitive methods in lymphomas [3, 4, 6], but its involvement in lymphomagenesis is still under discussion, double ISH to detect single copies of viral LMP1 and EBNA2 transcripts was assessed in 36 cases with good quality material. Of those 36 cases, 32 cases (89%) were defined as EBV− (< 20% EBERs + cells), whereas 4 cases (11%) were EBV+ (≥ 20% EBERs + cells) by ISH. In order to confirm the presence of traces, remarkably in the 32 EBV− cases, viral transcripts in tumor cells were detected in 17 cases (47%) (13 cases without EBERs expression), indicating the presence of viral traces (Supplementary Fig. 1 B and C). The presence of LMP1 transcripts, the most important viral oncogene, was confirmed in 6/13 cases positive for transcripts without EBERs expression by LMP1 immunohistochemical staining, as described [6].

According to those results, DLBCL cases were separated into different groups for further analysis. In order to evaluate the effect of traces of EBV infection in the expression of immune genes, initially, three groups were defined: i) EBV+ cases, with ≥ 20% EBERs + cells which also were positive for EBV transcripts, ii) EBV− Traces+ cases, including EBV− cases with < 20% EBERs + cells but expressing LMP1 and/or EBNA2 transcripts, and iii) Traces− cases, negative for both approaches. Given that only two groups are allowed to be compared by gene expression assay analysis, the three groups were clustered into two groups and mean expression was compared as follows: EBV+ (≥ 20% EBERs + cells) vs EBV− (< 20% EBERs + cells); Traces+ (positive for LMP1 and/or EBNA2 transcripts, regardless of EBV status by ISH) vs Traces− (negative for both transcripts); and EBV− Traces+ (cases that expressed viral LMP1 and/or EBNA2 transcripts but were EBV− by ISH) vs EBV− Traces− (cases that were negative by both methods). In contrast, specifically for microenvironment markers, the three initially defined groups were compared, in addition to a comparison between EBV+ vs EBV− cases to evaluate microenvironment composition with conventional EBERs cutoff [21].

The gene expression profile for immune response genes using customized NanoString Technologies platform in 48 DLBCL cases was assessed, in order to evaluate if traces of EBV infection could be involved in microenvironmental immune response. As previously mentioned, among those 48 DLBCL cases, 11 were considered EBV+ (according to ≥ 20% EBERs + cells used as cutoff) and 37 EBV− and included the cases with traces of EBV infection that were studied by double ISH for the detection of viral transcripts. As a first approach, the comparison in the expression of genes between the EBV+ vs EBV− cases was clustered as mentioned and analyzed (Fig. 1a). Even though significant differences in the gene expression between EBV+ and EBV− patients were not proved (FDR > 0.050 and log2 FC < 2), in the EBV+ cases, there was an increase in genes expression such as CD8A, CD68, LAG3, CD274 (PDL1), among others. Furthermore, the EBV+ DLBCL cases displayed an enhanced gene expression for cell types such as CD8 T and cytotoxic cells (p = 0.022 and p = 0.046, respectively, Mann–Whitney U test) (Fig. 2a–c).

In contrast, when the gene expression was compared in the second cluster, between the cases with viral traces (including EBERs + cases) vs the cases without viral traces, the slight increase in gene expression like LAG3 and CD274 (PDL1) previously observed for EBV+ vs EBV− cases was lost (Fig. 1b). The significant increase observed for gene expression signature for CD8 T cells in EBV+ cases shifted to a trend (p = 0.077, Mann–Whitney U test) in the cases that expressed viral transcripts. In addition, only a trend to an increase in exhausted CD8 T cells was observed in DLBCL cases with LMP1 and/or EBNA2 transcripts (Traces +) (p = 0.096, Mann–Whitney U test). (Fig. 2d–f). To exclude the influence of EBV+ cases by EBERs ISH, the gene expression for a third analysis, EBV− Traces+ vs EBV− Traces− cases, was assessed (Fig. 1c). As expected, the previous trend observed for CD8 T and exhausted CD8T cells was lost when EBV+ cases were excluded (p > 0.05, Mann–Whitney U test) (Fig. 2g–i).

It was previously demonstrated, on the one hand, that EBV+ DLBCL may exhibit a tolerogenic microenvironment [10]. On the other hand, a slight increase in PDL1 and CD68 genes in EBV+ cases and a trend to a higher expression of exhausted cells in cases with LMP1 and EBNA2 transcripts (including EBV+ ones) were observed in this study. Therefore, several markers of tolerance, such as PDL1, TIM3, and LAG3, were evaluated in tumor cells and at the TME by single and double IHC in 47 DLBCL cases with available material for analysis, in order to confirm gene expression findings (Fig. 3a–e). Positive cell count was analyzed in relation to EBV presence and viral traces. For immunohistochemical analysis, at first three groups were compared: EBV+ (≥ 20% EBERs + cells), EBV− Traces+ (positive for LMP1 and/or EBNA2 transcripts), and Traces− (negative for both transcripts). No significant differences in the TIM3 + , LAG3 + , and PAX5-PDL1 + expression, that reflects protein expression PDL1 at the microenvironment (PDL1m), were observed among these three groups (p > 0.05; Kruskal Wallis, p > 0.05; Mann–Whitney) (Fig. 4, first column, a–c). Remarkably, when PDL1 protein expression in tumor cells (PAX5 + PDL1 + , PDL1 + t) was analyzed, a significant increase in PAX5 + PDL1 + cells was observed in EBV+ cases (p = 0.0024, Kruskal Wallis, p = 0.0120 Mann–Whitney) (Fig. 4d, first column).

Then, cases were grouped into EBV+ vs EBV− cases, and into Traces+ (including EBV+ cases) vs Traces− cases. No significant differences in TIM3 + , LAG3 + , and PDL1 + m expression were proved (p > 0.05, Mann–Whitney test) (Fig. 4a–c, second and third columns). However, PDL1 expression in tumor cells displayed significantly higher expression in both clustered scenarios (p = 0.0048, p = 0.0106; Mann–Whitney, respectively) (Fig. 4d, second and third columns), but no differences were proved when cases EBV− Traces+ were compared with EBV− Traces—(Fig. 4d, first column). Finally, since a slight increase in CD68 gene expression was observed in EBV+ cases, and in order to explore the macrophage polarization markers, CD68 and CD163 were evaluated for each group defined. In addition, the polarization profile was defined by the CD68/ CD163 ratio in our cohort, as previously suggested [25]. Of the 47 cases available for immunohistochemical analysis, 57% showed an intermediate pattern (CD68/CD163 < 1.5), most of the EBV+ and EBV− Traces− cases (60% and 64%, respectively) and 34% showed M1 pattern (CD68/CD163 > 1.5), half of EBV− Traces+ cases (50%). Only 9% of DLBCL cases displayed the M2 polarization pattern (CD163/CD68 > 1.5). The CD68 + cell count was significantly higher in the EBV+ cases in comparison with EBV− Traces+ and EBV− Traces− cases (p = 0,0036; Kruskal Wallis, p = 0.0007, Mann–Whitney) (Fig. 4e, first column). Furthermore, when EBV+ vs EBV− cases were compared, there was also a significant increase in the former (p = 0.0034; Mann–Whitney) (Fig. 4e, second column). However, this difference was lost comparing Traces+ (including EBV+) vs Traces− cases (p > 0.05; Mann–Whitney) (Fig. 4e, third column). No differences were observed for CD163 + cell count (p > 0.05, Kruskal Wallis and Mann–Whitney test) (Fig. 4f).