Total RNA was isolated from 1 × 10⁶ tumor-tissue derived cells using the RNA Easy Mini Kit (Qiagen) according to the manufacturer’s instructions. The concentration and purity of the samples were determined by spectrophotometry with a NanoDrop© 2000c (Thermo Scientific), and the RNA integrity was assessed using a 2100 Bioanalyzer (Agilent). Complementary DNA was synthesized from 100 ng of total RNA using the High Capacity RNA-to-cDNA Kit (Applied Biosystems). The gene expression of immune response-associated genes was determined using TaqMan low-density array (TLDA) cards according to the manufacturer’s instructions (Applied Biosystems). The TLDA cards (TaqMan® Array Human Immune Panel) were run on a Viia7 instrument (Applied Biosystems) using TaqMan® Universal Master Mix II, no UNG (Applied Biosystems). Ct values were analyzed using GenEx software (MultiD Analyses). Relative gene expression levels were calculated using the ΔΔCt method and were normalized to the expression levels of reference genes GUSB and TFRC, selected by GeNorm from 6 reference genes assessed in total.

Staining was carried out on FFPE sections following deparaffinization and antigen retrieval. Endogenous peroxidase was blocked with 3% hydrogen peroxide. The sections were incubated with protein block (DAKO) and stained with primary antibodies against CD8 (SP16, Spring Bioscience), CD20 (L26, Dako) and DC-LAMP (1010E1.01, Dendritics), followed by the manifestation of enzymatic activity and hematoxylin counterstaining. The images were acquired using a Leica Aperio AT2 scanner (Leica).

Each section was scanned and evaluated for immune cell infiltration in the tumor nest and tumor stroma in 10 representative visual fields at 10× magnification using a Ventana Image Viewer. The cell numbers were related to tumor nest/tumor stroma area assessed by Calopix software (Tribvn). Additionally, a semiquantitative analysis of CD20⁺/CD8⁺ cell-cell interactions was performed (−, negative sections; +, sections positive for B cell/CD8⁺ T cell interactions in 1–5 visual fields; ++, sections positive for interactions in > 5 visual fields). The cell-cell interaction was defined as a direct cell-cell contact of CD20⁺ B cell and CD8⁺ T cell (Fig. 1d) within an aggregate of 20–100 cells (Fig. 1c) or in a distance up to 100 μm from a margin of the aggregate. The quantification was performed by two independent observers and reviewed by an experienced pathologist.Fig1Fig. 1

Differences in frequencies of tumor-infiltrating leucocytes in patients with oropharyngeal squamous cell carcinoma (OPSCC) with respect to the HPV status. a The heat-map expresses z-scores of relative mRNA expression of indicated genes within HPV- (n = 6) and HPV+ (n = 12) tumor samples. Genes with significantly different expression in HPV- and HPV+ tumors are marked in red. b Columns represent the mean (+ standard error of mean, SEM) densities of CD20⁺ B cells, CD8⁺ T cells and DC-LAMP⁺ dendritic cells in tumor nests and tumor stroma of immunohistochemically stained FFPE sections of OPC patients from Cohort 1 (n = 72). c Non-organized CD20⁺ B cell (brown)/CD8⁺ T cell (red) aggregate. d CD20⁺ B cell (brown) – CD8+ T cell interactions. e TLS with germinal center. f Columns show proportions of patients with detected B cell/CD8⁺ T cell interactions in the tumor nest and the tumor stroma of OPC tissue sections (−, interactions not detected; +, interactions detected in 1–5 visual fields; ++, interactions in > 5 visual fields). * p < 0.05 (Mann-Whitney U test)

Fresh tumor tissues were mechanically and enzymatically digested as described previously [6]. Subsequently, the specimens were passed through a 100-μm nylon cell strainer (BD Biosciences) and washed with PBS. Peripheral blood mononuclear cells (PBMCs) were isolated from the peripheral blood samples by centrifugation on a Ficoll-Paque density gradient (GE Healthcare).

Single cell suspensions derived from tumor tissues were labeled using a panel of monoclonal antibodies as listed in Additional file 1: Table S1. For the intracellular detection of cytokines and Ki-67, the cells were fixed and permeabilized with the Fixation/Permeabilization Buffer Set (eBioscience) and intracellularly labeled with primary antibodies. The cells were analyzed on a BD LSR Fortessa (BD Biosciences) and evaluated with FlowJo software (TreeStar).

The detection of HPV16 E6/E7-specific T cells was performed as described previously [36]. Briefly, freshly prepared tumor-derived single cell suspensions were seeded at a concentration of 3 × 10⁵ cells/ml into 24 well plate and TILs were expanded for two weeks in the presence of IL-2. Monocytes from autologous PBMCs were isolated using the Human CD14 Positive Selection Kit (Stemcell Technologies), loaded with HPV16 E6 and E7 peptide pools (5 μg/ml) (JPT) and added to expanded TILs at a ratio of 1:10. After 6 h of incubation with Brefeldin A (BioLegend), the cells were stained with antibodies for the intracellular detection of IFNγ.

Tumor-derived single cell suspensions were split into halves. One half was depleted of B cells using CD19 MicroBeads (Miltenyi Biotech) according to the manufacturer’s instructions. The second half was subjected to the same procedures without addition of CD19 MicroBeads. After magnetic separation, cell suspensions (6 × 10⁵ cells/ml) were cultured in RPMI 1640 supplemented with 10% heat-inactivated FCS, L-glutamine and penicillin-streptomycin (Invitrogen) in 48 well plates for 6 days without any additional stimuli. The viability of CD4⁺ and CD8⁺ T cells and their capacity to produce cytokines was assessed at day 1 and 6 using LIVE/DEAD™ Fixable Blue Dead Cell Stain Kit (Invitrogen) and intracellular cytokine staining as described above.

Tumor-derived single cell suspensions (1 × 10⁶ cells/ml) were cultured in RPMI 1640 supplemented with 10% heat-inactivated FCS, L-glutamine and penicillin-streptomycin (Invitrogen). For some of the patient samples (n = 3), the B cells were depleted from the cell suspensions using CD19 MicroBeads (Miltenyi Biotech) according to the manufacturer’s instructions. To detect the concentrations of lymphotoxin, IFNγ, TNFα, IL-6, IL-10, IL-12, CXCL9 and CXCL13 released into the culture supernatant, the MILLIPLEX™ Human Cytokine Kit (Merck) was used according to the manufacturer’s instructions.

CD8⁺ T cells were isolated from tumor tissue-derived single cell suspensions and PBMC using the EasySep™ Human CD8 Positive Selection Kit II (StemCell Technologies). Total RNA was isolated from 1 × 10⁶ CD8⁺ T cells using the RNA Easy Mini Kit (Qiagen) according to the manufacturer’s instructions. The concentration and purity of the samples were determined by spectrophotometry with a NanoDrop© 2000c (Thermo Scientific), and the RNA integrity was assessed using a 2100 Bioanalyzer (Agilent). Complementary DNA was synthesized from 100 ng of total RNA using the iScript cDNA Synthesis Kit (BIO-RAD). The gene expression levels of BCL2L1, IL-2, IL-2R, CD27, CD40L, and the β-actin housekeeping gene were evaluated using the CFX 96™ Real-Time System (BIO-RAD). The specificity of the amplified PCR product was assessed using an Agilent DNA 1000 Kit (Agilent). The relative expression of the target genes was normalized to the expression of β-actin.

Statistical analyses were performed using Statistica® 10.0 software (StatSoft). The differences between HPV-positive and HPV-negative tumor samples were analyzed using the Mann-Whitney U test. The prognostic value of tumor-infiltrating immune cells was analyzed using the log-rank test. Additionally, the Cox proportional hazard model was used to perform univariate and multivariate analyses of possible prognostic factors. Only variables with significant differences observed in the univariate analysis were included in the multivariate analysis. The correlation between the presence of B cell/CD8+ T cell interactions and HPV positivity/presence of HPV16 E6/E7-specific CD8⁺ T cells was evaluated using Pearson’s chi-square test. Variability in proportions of Ki-67⁺ cells was detected using Kruskal-Wallis ANOVA. Differences in the B cell phenotype were analyzed using one-way ANOVA, followed by Tukey’s post hoc test. The results were considered statistically significant when p < 0.05.

To evaluate the transcriptional signature of immune response-related genes in HPV-associated and HPV-negative tumors, we assessed the expression of selected genes using TaqMan analysis. Tumor samples with a positive HPV status expressed significantly higher levels of all of the B cell-related genes analyzed, namely, BLK, CD19, CR2, HLA-DOB, MS4A1 and TNFRSF17 (Fig. 1a).

To supplement the results of gene expression, we immunohistochemically analyzed the density of CD20⁺, CD8⁺ and DC-LAMP⁺ cells in 72 OPSCC tumor tissue sections (Cohort 1). Compared to HPV-negative tumors, HPV-associated tumors showed significantly higher infiltrates of CD20⁺ B cells in the tumor nest and significantly higher levels of CD8⁺ T cells in both the tumor nest and the tumor stroma. No differences were observed in DC-LAMP expression (Fig. 1b). Additionally, we observed that tumor-infiltrating CD20⁺ B cells and CD8⁺ T cells create non-organized aggregates in both the tumor nests and the tumor stroma (Fig. 1c) with CD20⁺ B cells and CD8⁺ T cells in a direct cell-cell interaction (Fig. 1d). The proportion of these cell-cell interactions was markedly higher in HPV-associated tumors than in HPV-negative tumors (Fig. 1f). In contrast to direct CD20⁺ B cell/CD8⁺ T cell interactions, no differences between HPV-associated and HPV-negative samples were observed in the density of tertiary lymphoid structures (TLS) with germinal centers (Fig. 1e). Well-defined TLS with germinal centers were detected in 29.8% of HPV-associated samples and in 25.0% of HPV-negative samples.

To evaluate the prognostic impact of tumor-infiltrating CD20⁺ B cells, CD8⁺ T cells, DC-LAMP⁺ DCs and B cell/CD8⁺ T cell interactions in both intratumoral and stromal compartments of OPSCC samples, we investigated overall survival (OS) upon stratifying the patient cohort based on the median of positive cells per 1 mm² of the tumor nest and the tumor stroma area. The presence of abundant intratumoral CD20⁺ B cells and CD8⁺ T cells was associated with significantly improved OS (p < 0.001 and p = 0.013, respectively; Fig. 2a, b). Furthermore, the presence of abundant intratumoral and stromal CD20⁺ B cell/CD8⁺ T cell interactions was also positively correlated with OS. This correlation was highly statistically significant (p = 0.001 and p = 0.009, respectively; Fig. 2c). Surprisingly, the density of direct CD20⁺ B cell/CD8⁺ T cell interactions stratified the patients better than the concurrent presence of both CD20⁺ B cells and CD8⁺ T cells (Fig. 2d).Fig2Fig. 2

Prognostic value of tumor-infiltrating CD20⁺ B cells (a), CD8⁺ T cells (b), CD20⁺B cell/CD8⁺ T cell (B/Tc) interactions (c) and combination of CD20⁺ B cells and CD8⁺ T cells (d) in patients with OPSCC (n = 70). Kaplan-Meier curves show overall survival of patients according to densities of the indicated cells in the tumor nests (left) and in the tumor stroma (right). P values were determined using the log-rank test

In addition to the differences detected between HPV-positive and HPV-negative tumors, we observed substantial variability in the density of tumor-infiltrating lymphocytes and CD20⁺ B cell/CD8⁺ T cell interactions within the group of patients with HPV-associated tumors, splitting HPV-positive samples into “hot” and “cold” subgroups. Therefore, to assess whether the interactions between CD20⁺ B cells and CD8⁺ T cells might be important for the HPV-specific T cell response in HPV-driven tumors, we correlated the presence and density of B cell/CD8⁺ T cell interactions in the FFPE tumor sections with the proportions of HPV16 E6/E7-specific CD8⁺ T cells detected in TILs expanded from matched native HPV-positive OPSCC samples (Cohort 2). Indeed, 81.8% of patients with detected HPV16 E6/E7-specific CD8⁺ T cells had a high density of B cell/CD8⁺ T cell interactions in the tumor stroma and 61.5% of these patients also had high density of these interactions in the tumor nests. In contrast, it was only 42.8 and 14.3%, respectively, in patients without detected HPV16 E6/E7-specific CD8⁺ T cell responses (Fig. 3a). Moreover, the proportion of HPV16 E6/E7-specific CD8⁺ T cells was significantly positively correlated with the density of B cell/CD8⁺ T cell interactions in the tumor nests (Fig. 3b), indicating that patients with low levels of direct B cell – CD8⁺ T cell interactions also had low levels of HPV16 E6/E7-specific CD8⁺ T cells. On the contrary, the presence of HPV16-specific CD8⁺ T cells was neither correlated to the density of CD8⁺ T cells in general nor to the density of CD20⁺ B cells (Fig. 3c).Fig3Fig. 3

Positive correlation of direct CD20⁺ B cell/CD8⁺ T cell interactions with HPV16 E6/E7-specific CD8⁺ T cells. a Columns show the proportions of patients with low (interactions detectable in 0–5 visual fields) and high (interactions detectable in > 5 visual fields) densities of B cell/CD8⁺ T cell interactions with respect to the presence or absence of tumor-infiltrating HPV16 E6/E7-specific CD8⁺ T cells. b Columns represent the mean (+ SEM) proportions of tumor-infiltrating HPV16 E6/E7-specific CD8⁺ T cells with respect to the densities of B cell/CD8⁺ T cell interactions within the tumor nests. c Columns represent the mean (+ SEM) densities of CD20⁺ B cells, CD8⁺ T cells and DC-LAMP⁺ dendritic cells in tumor nests and tumor stroma of patients without/with detected HPV16-specific T cells. *, p < 0.05 (Pearson’s chi-square test and Mann-Whitney U test)

To characterize the phenotype and function of TIL-Bs in HPV-associated tumors with a “hot” versus “cold” phenotype, we analyzed intratumoral and blood-derived B cell subsets by flow cytometry (Cohort 3). Tumor suspensions were divided according to the proportions of TIL-Bs into “cold” B^lo samples (B cell proportions < 0.5% of total cells; mean = 0.11 ± 0.05%) and “hot” B^hi samples (mean = 4.22 ± 5.96%). In all samples, CD19⁺ B cells were divided into five subtypes based on the expression level of IgD and CD38, namely, IgD⁻CD38⁺⁺ plasma cells, IgD⁻CD38⁺ germinal center B cells, IgD⁻CD38⁻ memory B cells, IgD⁺CD38⁻ naive B cells and IgD⁺CD38⁺ pre-germinal center B cells (Fig. 4a). Memory B cells represented the major B cell subtype in the tumor tissue (Fig. 4b). There was no difference in the B cell subtype composition between B^lo and B^hi samples.Fig4Fig. 4

Flow cytometric analysis of tumor-infiltrating B cells and B cells derived from patients’ PBMC divided according to the proportion of TIL-Bs into B^lo (% of TIL-Bs < 0.5 of total cells) and B^hi samples. Data are expressed as (a) representative dot plots and (b) mean + SEM of the proportions of B cell subsets within total CD19⁺ B cells. IgD⁻CD38⁺⁺, plasma cells; IgD⁻CD38⁺, germinal center B cells; IgD⁻CD38⁻, memory B cells; IgD⁺CD38⁻, naive B cells; IgD⁺CD38⁺, pre-germinal center B cells. c Columns represent mean + SEM of the MFI of B cell surface markers assessed on total CD19⁺ B cells. d, e Histograms show a representative expression of indicated B cell surface markers in B^lo (upper line) and B^hi (lower line) patient. Gray-filled areas represent isotype-matched controls, red line represents peripheral blood B cells and blue line represents tumor-infiltrating B cells of the same patient. *, p < 0.05; **, p < 0.01 (ANOVA followed by Tukey’s post-hoc test)

To assess the impact of TIL-Bs on survival and functional capacity of T cells, we cultivated B^hi tumor-derived cell suspensions and analyzed the viability and cytokine production of CD4⁺ and CD8⁺ T cells after B cell depletion (n = 4). In B cell-depleted suspensions, the viability of both CD4⁺ T cells and CD8⁺ T cells did not differ at day 1, but was markedly lower compared to bulk suspensions after 6 days of cultivation without any additional stimuli (15.1 ± 7.8% vs. 11.0 ± 4.5% for CD4⁺ T cells; p = 0.068) and 22.4 ± 10.6% vs. 14.4 ± 8.4% for CD8⁺ T cells; p = 0.068) (Fig. 5a, b, c). Despite the impaired viability, we did not observe any substantial differences in the proportions of IL-2 and IFN-γ producing CD4⁺ and CD8⁺ T cells with respect to the presence or absence of B cells in the cell cultures.Fig5Fig. 5

Proportions of dead cells in cultures of bulk and B cell-depleted tumor-derived single cell suspensions. a, b Box plots show the mean proportion of dead CD4⁺ and CD8⁺ T cells in bulk (B+) and B cell-depleted (B-) tumor-derived cell suspension after 1 and 6 days of cultivation. c Histograms show differences in the LIVE/DEAD Blue Stain positivity on day 6 in a representative patient. d The heat-map expresses z-scores of relative mRNA expression of indicated genes within B^lo (n = 53) and B^hi (n = 52) samples extracted from TCGA databases. e Box plots show the mean expression of indicated genes in tumor tissues and matched PBMCs of B^hi OPSCC patients (n = 4). The boundaries of the box indicate the standard error of the mean and the squares in the box represent the mean. Whiskers indicate the standard deviation. *, p < 0.05; **, p < 0.01 (t test and Mann-Whitney U test)

To estimate the expression levels of a wide spectrum of costimulatory molecules, we analyzed the data extracted from TCGA databases using Statistica® 10.0 software (StatSoft). HNSCC patients with defined p16 status were divided into B^hi and B^lo subgroups according to the median expression of CD19. With the exception of BCL2L1, TNFSF9 and CD86, the B^hi samples expressed significantly higher levels of all costimulatory molecules and molecules associated with activation of the TNFR family signaling pathways tested (Fig. 5d).

In mouse models of viral infections, CD27 was assessed as the key factor in directing the autocrine production of IL-2 that is required for the long-term survival of CD8⁺ T cells in nonlymphoid tissues [35]. Therefore, we analyzed the levels of expression of IL-2, IL-2RA and CD27 together with CD40LG and the anti-apoptotic regulator BCL2L1 on CD8⁺ TILs isolated from peripheral blood and tumor tissues of B^hi OPSCC patients (n = 4; Cohort 3). Indeed, significantly higher levels of IL-2 and IL-2R were expressed in tumor-derived CD8⁺ T cells than in matched peripheral blood CD8⁺ T cells (Fig. 5e).

Regulatory B cells (Bregs) are characterized by the production of IL-10. To assess the proportion of Bregs in the OPSCC tumor microenvironment, we analyzed the level of IL-10 secreting TIL-Bs after 5 and 24 h of stimulation with CpG ODN 2006 and CD40L in the presence of PMA, ionomycin and brefeldin A using flow cytometry (Cohort 3).

After 5 h of stimulation, the proportion of Bregs, which were found to be predominantly CD5⁺CD24^hi, was slightly higher in the tumor tissue (0.98 ± 0.78%) compared to matched peripheral blood B cells (0.46 ± 0.12%) and control tonsils (0.41 ± 0.09). Surprisingly, the proportion of IL-10-secreting Bregs after 24 h of in vitro maturation with CpG ODN 2006 and CD40L was significantly lower in the tumor samples than in matched peripheral blood B cells (2.74 ± 0.53% vs. 8.01 ± 1.75%, respectively; p = 0.039), but similar to the levels of Bregs in control tonsils (2.16 ± 1.51%). During the longterm stimulation by TLR ligands and CD40L, Breg progenitors mature into IL-10 producing Bregs [37]; therefore, both Bregs and Breg progenitors were detected after 24 h of cultivation in vitro. Interestingly, the proportion of Bregs was negatively correlated to the frequency of CD19⁺ B cells in general (r = − 0.69; p = 0.085). Due to a limited number of cells, IL-10 production was assessed in B^hi samples only.

To estimate the impact of TIL-Bs on cytokine production in the tumor microenvironment, we analyzed the spontaneous production of cytokines and chemokines in B^lo and B^hi tumor-derived cell suspensions and in B^hi suspensions depleted of CD19⁺ B cells (Cohort 3). B^hi cell suspensions produced markedly higher levels of CXCL9 than B^lo cell suspensions (Additional file 4: Figure S2A). In accordance with these results, we observed significantly lower levels of CXCL9 in B cell-depleted samples than in whole cell suspensions (579.6 ± 262.9 vs. 1238.8 ± 290.6 pg/ml, respectively; p = 0.025; Additional file 4: Figure S2B), indicating that TIL-Bs are an important source of this chemokine.