T Cells Expressing Checkpoint Receptor TIGIT Are Enriched in Follicular Lymphoma Tumors and Characterized by Reversible Suppression of T-cell Receptor Signaling

This article is featured in Highlights of This Issue, p. 725

Follicular lymphoma (FL) is the most common subtype of indolent non-Hodgkin lymphoma. Although outcomes have improved (1), current chemoimmunotherapy regimens are usually not curative. Additionally, FL patients can transform to more aggressive histology, leading to rapid progression and need for intensive therapy (2). Ongoing clinical trials to improve treatment of FL focus on novel targeted agents and various immunomodulatory regimens, including immunotherapy with checkpoint blockade (3, 4).

Targeting coinhibitory receptors such as PD-1 and CTLA-4 by immune checkpoint blockade can restore the function of exhausted T cells with antitumor reactivity (5, 6). T cells in the FL tumor microenvironment (TME) are considered dysfunctional and associated with disease progression (7–9). However, whereas blockade of PD-1 represents a breakthrough for several solid cancers (10–12) and for Hodgkin's lymphoma (13), the response rate as monotherapy in FL has been lower than anticipated (14), given the high expression of PD-1 in intratumor T cells and presence of PD-L1⁺ histiocytes in the TME (9, 15). However, the influence of different T-cell subsets for lymphomagenesis is complex. While T follicular helper cells (T_FH) display PD-1^hi phenotype and are highly functional by supporting lymphoma B cells through CD40 ligand and secretion of cytokines IL4 and IL21 (16–18), exhausted T cells express intermediate levels of PD-1 (15, 19). A hallmark of T-cell exhaustion is expression of multiple coinhibitory receptors alongside progressive loss of effector functions (20). Therefore, coblockade of several coinhibitory receptors might be necessary to achieve optimal antitumor T-cell responses. T-cell immunoglobulin and ITIM domain (TIGIT) is a recently identified coinhibitory receptor, expressed by natural killer (NK) cells, effector T cells (T_E), T regulatory cells (Treg) and T_FH (21–25). Prior findings suggest TIGIT as a candidate for checkpoint blockade, as TIGIT is frequently found on tumor-infiltrating T cells (TIL) in solid tumors and in acute myeloid leukemia (AML; refs. 26–28), and the TIGIT ligands, CD155 and CD112, are expressed by different cell types, including antigen-presenting cells and tumor cells (21, 22, 24, 29).

Numerous genes are recurrently mutated in FL (30–33), creating tumor antigens, including the lymphoma immunoglobulins, that may trigger T-cell antitumor responses (34). Antigen recognition by the T-cell receptor (TCR) initiates a cascade of tyrosine phosphorylations, and the amplitude and duration of TCR signaling is critical for T-cell effector function (35). Hence, exhausted T cells can be distinguished from functional T cells by low TCR signaling strength. Upon TCR interaction with peptide–MHC, the immunoreceptor tyrosine-based activation motifs (ITAM) of the TCR-associated CD3 subunits become phosphorylated by Src family kinases such as LCK (35, 36). Subsequent recruitment and phosphorylation of the adaptor protein SH2-domain containing leukocyte protein of 76 kDa (SLP76), and linker for activation of T cells (LAT), results in formation of the LAT signalosome, which enables activation of multiple downstream effectors, including activation of the RAS–MEK–ERK, PI3K/AKT and NF-κB pathways. TCR signaling is enhanced by costimulatory receptors such as CD28, but dampened by coinhibitory receptors such as CTLA-4 and PD-1 due to recruitment of phosphatases (37, 38).

The hypothesis underlying this study was that characterizing signaling responses and coinhibitory receptor expression in intratumor T-cell subsets could reveal new targets for immune checkpoint blockade. Based on previous studies, demonstrating the importance of PD-1 for T-cell immunosuppression (9), our approach was to measure functional responses in T cells with differential expression of PD-1, while in parallel screening for coinhibitory receptors that could be of interest for immune checkpoint blockade in combination with PD-1. This approach identified TIGIT as the most frequently expressed coinhibitory receptor in FL T cells, and the expression was associated with T-cell dysfunction. Taken together, our data suggest TIGIT as a promising new target for immune checkpoint blockade in FL.

Specimens were obtained with informed consent in accordance with the Declaration of Helsinki and with approval from the Regional Committees for Medical and Health Research Ethics (REK S-0749b and 2010/1147a). Malignant LN specimens were obtained at time of diagnosis from FL patients (n = 12) or after treatment (n = 2) at the Norwegian Radium Hospital, Oslo, Norway, and tonsils were obtained from patients undergoing tonsillectomy at Agroklinikken (Asker, Norway). LN and tonsils were processed to single-cell suspensions by mincing and stored as aliquots in liquid nitrogen. Peripheral blood was collected from anonymous, healthy donors at The Blood Bank in Oslo (REK S-03280), processed to mononuclear cells (PBMC) by Ficoll gradient centrifugation (Ficoll-Paque PLUS, GE Healthcare) and cryopreserved in liquid nitrogen.

Stimulation reagents: TCR activation (α-TCR): anti-CD3 biotin and anti-CD28 biotin labeled antibodies were used at 5 μg/mL each and avidin (Thermo Fischer Scientific) was used at 50 μg/mL. Phorbol 12-myristate 13-acetate (PMA) was used at 125 ng/mL and ionomycin was used at 500 ng/mL (Sigma-Aldrich). GolgiPlug was from BD Biosciences. Cells were stained using fluorochrome-coupled antibodies (Supplementary Table S1). Antibody used to detect FoxP3 was added after fixation and permeabilization according to the eBioscience protocol. Brilliant Stain Buffer (BD biosciences) was used as staining buffer. Pacific Blue used for fluorescent barcoding of cells was from Life Technologies, Molecular probes.

Activation of signaling and detection by phospho-specific flow cytometry were performed as described (9, 39, 40). Specimens were thawed, and cells were allowed to rest at 37°C for 4 hours, before redistribution into v-bottomed 96-well plates and given another 20 minutes rest. For functional studies over time, cells were cultured for 48 hours at 37°C, at 2.5 × 10⁶/mL in CellGro DC (CellGenix) supplemented with 5% human serum (Diaserve Laboratories). IL2 (20 U/mL; Chiron) was added in some experiments as specified. Signaling was activated by α-TCR for 1, 4, or 10 minutes (details in Supplementary methods). Signaling was stopped by adding paraformaldehyde (PFA; 1.6%), followed by centrifugation and permeabilization in >90% freezer-cold methanol. After rehydration, the cells were stained with antibodies, or “barcoded” with Pacific Blue prior to staining with antibodies as previously described (9). The samples were collected on a LSR II flow cytometer (BD Biosciences). Data were analyzed using Cytobank Software, https://community.cytobank.org. Relative phosphorylation changes (fold changes) were calculated using arcsinh transformation of median fluorescence intensity (MFI) of the cell population of interest.

The computational tool viSNE (41) was used for visualization of immunophenotype data, see Supplementary Methods.

Samples were incubated for 6 hours in the presence of PMA and ionomycin, with GolgiPlug present for the last 4 hours. PFA (1.6%) was added to stop activity, followed by centrifugation and permeabilization in >90% freezer-cold methanol. At this point, the samples could be stored at −80°C, before staining with antibodies and flow cytometry acquisition.

Gene expression data were obtained from two different datasets; Dave and colleagues (7) and Brodtkorb and colleagues (42), and included pretreatment FL biopsies only, see supplementary methods.

Serial sections of cryopreserved FL tissue were stained with antibodies for CD155 (L95) and CD112 (L14) as previously described (43), in addition to CD21 (2G9).

To explore if PD-1 was more frequently expressed in intratumor T cells from FL than in corresponding subsets from healthy tissues, LN specimens from 14 FL patients were immunophenotyped and compared with 11 tonsillar and 7 PBMC samples from healthy donors. In order to distinguish T_FH from other subsets, distribution of T cells was characterized based on differential expression of PD-1 and ICOS in CD4 (PD-1⁻ICOS⁻, PD-1^intICOS⁻, PD-1^intICOS⁺, and PD-1^hiICOS⁺ (T_FH)) and CD8 (PD-1⁻ICOS⁻ and PD-1^intICOS⁻) T-cell subsets (Fig. 1A). We found that neither the T_FH compartment nor the CD4⁺ PD-1^int T-cell subsets were significantly different between FL tumors and tonsil controls. In contrast, the CD8⁺ PD-1^int subset was markedly increased in FL tumors compared to healthy PBMC or tonsils (P < 0.003 and P < 0.0001, Fig. 1B), suggesting a larger fraction of exhausted CD8 T cells in FL.

We next measured cytokine production in relation to PD-1 expression in CD4 and CD8 T cells. We observed that reduced percentage of CD8 T cells from FL patients produced IFNγ, compared with healthy individuals. Interestingly, IFNγ production was reduced in PD-1⁻ as well as PD-1^int CD8 T cells (Fig. 1C), indicating that PD-1–negative CD8 FL T cells were also suppressed. This finding suggests presence of other inhibitory mechanisms in PD-1⁻ CD8 FL T cells, leading to reduced functionality. Production of IL4 and IL21 was also measured, but was not significantly different in CD8 T cells from FL LN and healthy donors (Supplementary Fig. S1). In CD4 T cells, IL4 production was low but at comparable levels in FL and tonsillar subsets, whereas IL21 production was reduced in all FL subsets except for PD-1⁻ICOS⁻ cells (Supplementary Fig. S1).

As functional TCR signaling is critical for generation of effective antitumor T-cell responses, including production of IFNγ, we next investigated TCR-induced signaling in T cells from FL tumors (Fig. 2A). TCR signaling was activated using α-CD3 and α-CD28 biotinylated antibodies, followed by avidin crosslinking. To identify optimal time points to detect maximal phosphorylation levels, TCR was activated for 1, 4, and 10 minutes. Whereas p-CD3ζ and p-SLP76 peaked at 1 minute post stimulation, p-ERK signaling was undetectable at 1 minute and reached the maximal level 4 minutes after stimulation (Supplementary Fig. S2). A comparison of TCR-induced signaling responses in FL and healthy individuals revealed that T cells from FL patients were distinguished by highly reduced TCR-induced p-ERK, while p-SLP76 and p-CD3ζ levels were comparable (Fig. 2B). The low levels of TCR-induced p-ERK was evident in CD8⁺PD-1^int FL T cells, with a relative median fold change (FC) of 0.18 as compared with 0.56 and 0.34 in PBMC and tonsils, respectively (Fig. 2C). Strikingly, TCR-induced p-ERK was low in all CD4 FL T-cell subsets (range, 0.2–0.4; Fig. 2C). In contrast, TCR proximal signaling, as determined by p-SLP76, was comparable in FL and tonsillar T cells, with median FC ranges of 1.7–2.0 and 1.9–2.2, respectively (Fig. 2C). Phosphorylation of CD3ζ was also potent in FL, similar to the levels observed in tonsillar T-cell subsets (Fig. 2C). Interestingly, the low TCR-induced p-ERK observed across all T-cell subsets from FL LN indicated a block in the distal part of the pathway. This corresponded with the observed reduction in IFNγ production.

We hypothesized that multiple coinhibitory receptors might play a role in dampening T-cell antitumor responses in FL. We therefore used 11-parameter flow cytometry panels to achieve an in-depth characterization of coinhibitory receptor expression patterns in FL T-cell subsets, and compared patterns with healthy donor samples as before. A viSNE analysis, based on the expression of 6 lineage markers (CD4, CD8, CXCR5, ICOS, CD45RA, and CCR7) was used to visualize the data and to identify conventional T-cell subsets, as well as T-cell subsets identified based on PD-1 and ICOS expression (Supplementary Fig. S3A–S3C). The expression pattern of 9 coinhibitory receptors—PD-1, TIGIT, TIM-3, CTLA-4, LAG-3, BTLA, CD244, LAIR-1, and CD160—was then identified in the conventional T-cell subsets (Fig. 3). Strikingly, TIGIT was an abundant coinhibitory receptor in FL T cells and was expressed by the majority of CD4 and CD8 T effector memory (T_EM) cells (Fig. 3). Furthermore, TIGIT⁺ CD8 T_EM cells from FL coexpressed several exhaustion markers, such as PD-1 and CD244 (Fig. 3; Supplementary Fig. S3D and S3E), suggesting that TIGIT marks exhausted CD8 T cells in FL.

Detailed analysis revealed that TIGIT was expressed at significantly higher levels across all T-cell subsets in FL tumors compared with healthy donor tonsils or PBMC, but with contrasting expression pattern across distinct subsets: low expression in naïve T cells and highest in T_EM and T_FH cells. On average, 80% and 79% of CD8 and CD4 T_EM cells from FL expressed TIGIT (Fig. 4A–C). This is an important finding as T_EM was the major subset of CD4 and CD8 T cells in FL tumors (Supplementary Fig. S4). The majority of FL CD8 and CD4 T_EM cells also expressed PD-1 (80% and 65%, respectively), and some expressed BTLA (10% and 42%; Fig. 4B and C). TIM-3, CTLA-4, LAG-3, LAIR-1, and CD160 were all less frequently expressed in FL CD4 and CD8 T_EM cells (in average <25%; Supplementary Fig. S5). CD244 was expressed by 57% of CD8 T_EM, similar to tonsils (Supplementary Fig. S5). As Tregs also can express several coinhibitory receptors, TIGIT expression was investigated in CD4⁺CD25⁺FOXP3⁺ T_regs from FL LN and from healthy samples. Remarkably, TIGIT was expressed by the vast majority of FL Tregs (range, 92%–99%). This was not tumor specific, as most Tregs from healthy tonsils and PBMC expressed TIGIT (Fig. 4D). The TIGIT⁺ Tregs accounted for in average 25% of the TIGIT⁺ CD4 T cells in FL (Supplementary Fig. S6). Together, these results identified TIGIT and PD-1 as the most frequently expressed coinhibitory receptors in FL T cells.

TIGIT exerts inhibitory functions upon interaction with its ligands, CD112 or CD155 (22). Immunophenotypic analysis of FL specimens showed that less than 5% of the tumor cells expressed CD155 or CD112 (Supplementary Table S2). Importantly, although not expressed by malignant B cells, immunohistochemical staining of 6 samples of FL revealed expression of CD155 and CD112 on follicular dendritic cells (FDC) within the neoplastic follicles and on endothelial cells (Fig. 5A). Additionally, transcriptional analysis using two different gene expression profiling datasets (7, 42) showed that CD112 and CD155 are present in FL, and further confirmed TIGIT as the most highly expressed coinhibitory receptor (Supplementary Fig. S7).

Because TIGIT competes for ligand binding with the costimulatory receptor CD226 (24, 44), costaining for the two receptors to identify their relationship in FL T cells was performed. Interestingly, while TIGIT⁺ CD8 T cells from healthy donor PBMC had a substantial proportion of cells that coexpressed both receptors, few TIGIT⁺ CD8 T cells from FL tumors expressed the costimulatory receptor compared with TIGIT⁻ CD8 FL T cells (P < 0.02, by Wilcoxon test; Fig. 5B). Expression of CD226 was particularly low in CD8 T_EM cells (Fig. 5C), which displayed the highest expression of TIGIT (Fig. 4B). In contrast, CD226 was frequently expressed in TIGIT⁺ CD4 T cells in FL, including T_FH cells (Fig. 5B and C). Together, this indicates an imbalance in the TIGIT/CD226 axis in CD8 FL T cells.

To further investigate the relationship between TIGIT expression and dysfunctional TCR-induced signaling, we included detection of TIGIT in our signaling assay. Distinguishing between TIGIT⁻ and TIGIT⁺ cells among the CD8 FL T cells revealed that TIGIT⁺ cells had reduced TCR-induced p-ERK compared with TIGIT⁻ cells (Fig. 6A and B). This contrasted TCR proximal signaling, demonstrated by high levels of TCR-induced p-SLP76 regardless of TIGIT expression (Fig. 6B). These results indicate that TIGIT plays a role in dampening signaling distal to the TCR.

To test whether the dysfunctional TCR signaling could be restored, we studied signaling responses after 48 hours in vitro culture (Fig. 6C). Detection of TIGIT revealed that the percentage of TIGIT⁺ CD8 FL T cells was stable over time (Supplementary Fig. S8). Interestingly, while TCR-induced p-SLP76 was comparable in CD8 T cells at day 0 and after 2 days, we observed a striking increase in TCR-induced p-ERK, from 1.03 to 2.01 FC (Fig. 6D). Importantly, the recovery of TCR-induced p-ERK was highly reproducible and remarkably high in TIGIT⁺ CD8 FL T cells (median fold change from 0.8 at day 0 to 2.1 at day 2) as compared with TIGIT⁻ CD8 FL T cells (from 1.6 at day 0 to 2.7 at day 2; Fig. 6E and F; Supplementary Fig. S9). As TIGIT ligands were expressed by FDC (Fig. 5A), which are tightly adhered to the stroma, these cells were not preserved and are therefore not present in the cryopreserved single cell suspensions used in the functional assays. In conclusion, our results showed that the highly reduced TCR-induced p-ERK in FL could be recovered upon in vitro culture when TIGIT ligand–expressing cells are not present, suggesting that FL T cells receive suppressive signals through TIGIT via ligand⁺ cells in the TME in vivo.

Immune checkpoint blockers have shown impressive clinical benefits in several tumor types. Despite frequent expression of PD-1 in intratumor T cells in FL (9, 15), a significant proportion of patients do not respond to anti–PD-1 blockade (14, 45). Tumor genomic landscape, mutational load and tumor specific neoantigens are potential determinants of the response to immune checkpoint blockade, as well as characteristics of the TME (34, 46–49). As T-cell exhaustion might relate to coexpression of several coinhibitory receptors, identification of the most relevant types as targets for immune checkpoint blockade in FL patients will be important in order to fully unleash the antitumor response. In this study, we performed a multidimensional functional and phenotypical characterization of intratumor T cells from FL patients and compared with tonsils and PBMC from healthy donors to identify relevant targets for immune checkpoint blockade in FL. This approach identified TIGIT and PD-1 as the most frequently expressed coinhibitory receptors. In FL CD8 T cells, we observed reduced production of IFNγ as well as highly reduced TCR-induced p-ERK, which correlated with TIGIT expression and could be fully restored by in vitro culture in absence of TIGIT ligands CD155 and CD112. The TIGIT ligands were expressed by FDC and endothelial cells in FL tumors. Together, these results indicate that TIGIT is a relevant target for immune checkpoint blockade in FL.

Strikingly, our results showed that TIGIT in average was expressed in more than 80% of CD8 and CD4 T_EM cells from FL tumors, which accounted for 50% and 60% of CD8 and CD4 T cells, respectively. Furthermore, more than 95% of T_regs and T_FH cells from FL LN expressed TIGIT. Importantly, TIGIT might potentially have divergent functions in different T-cell subsets. Agonistic anti-TIGIT antibody had direct inhibitory effects on T-bet expression and IFNγ production in CD4⁺ T_E cells (50), and loss of TIGIT in vivo increased T-cell proliferation and proinflammatory cytokine production (25). In contrast to the unresponsive phenotype of TIGIT⁺ T_E cells, TIGIT⁺ Tregs are highly functional cells. Several studies have demonstrated that TIGIT⁺ Tregs have increased expression of effector molecules and are more potent suppressors of T_E proliferation than TIGIT⁻ Tregs (28, 51, 52). As the frequency of Tregs is increased in FL LN, TIGIT⁺ Tregs are likely to contribute to sustained immune suppression in FL. In addition, TIGIT is frequently expressed by tonsillar T_FH (53), and we observed that the majority (>95%) of T_FH from FL LN as well as tonsils from healthy donors expressed TIGIT. Previous studies suggest that TIGIT mediates adhesion of T_FH to FDC in germinal centers (23), and TIGIT is required for efficient B-cell helper function of peripheral blood circulating T_FH (54). Furthermore, TIGIT can outcompete the costimulatory receptor CD226 due to its higher affinity for the same ligand and by blocking dimerization of CD226, thus preventing its costimulatory function (21, 44, 50, 55). Our results revealed that TIGIT⁺ CD8 FL T cells rarely expressed the competing costimulatory receptor. This indicated an imbalance between costimulation and coinhibition in these cells, further suggesting that TIGIT plays a role in dampening CD8 T-cell antitumor responses in FL. Altogether, this suggests that immune checkpoint blockade targeting TIGIT should enable highly potent T-cell antitumor responses in several ways, including restoring antitumor potential of T effector cells, dampening the Treg immunosuppressive effect and by reducing the tumor supporting effects of T_FH cells. In addition to the direct effects of TIGIT in T cells, TIGIT can directly restrain NK cell activity (56) and indirectly exert inhibitory effects by activating immunoregulatory dendritic cells upon ligand interaction (21). Hence, blocking TIGIT in these cells may also be pivotal for efficient immunotherapy responses.

By combined detection of TIGIT, T-cell markers and phosphorylation of signaling effectors post TCR activation, we identified a clear correlation between TIGIT expression and TCR signaling dysfunction in CD8 FL T cells. However, TIGIT needs to be ligated to exert its suppressive function. Our results showed that less than 5% of FL tumor cells expressed the TIGIT ligands CD155 and CD112. Instead, immunohistochemical staining revealed the presence of CD155 and CD112 on FDC and on endothelial cells in FL tumors. These cells are tightly adhered to the stroma and were not detectable in the cryopreserved samples used for immunophenotyping by flow cytometry. However, the ligand⁺ FDC are likely to interact with TIGIT-expressing T cells in vivo, thereby preventing potent antitumor T-cell responses in FL. We were not able to provide direct proof for this hypothesis, but further support comes from the in vitro cultures of FL T cells. When cultured in the absence of CD155⁺ or CD112⁺ cells, CD8 TIGIT⁺ FL T cells could regain their TCR signaling capacity. Based on this, we cannot exclude the possibility that culture of FL derived CD8 T cells over time also removes other suppressive signals, as demonstrated by effectiveness of TIL therapy in FL (57). Although the mechanisms underpinning how TIGIT modulates T-cell–intrinsic signaling is poorly understood, studies in NK cells suggest that TIGIT upon ligation recruits the inositol 5-phosphatase SHIP1 to attenuate signaling downstream of SLP76, leading to dephosphorylation of ERK and subsequent inhibition of IFNγ production (58, 59). This is in agreement with what we observed in FL T cells; that TIGIT expression correlated with highly reduced TCR-induced p-ERK that translated into reduced IFNγ production in CD8 FL T cells, while phosphorylation of CD3ζ and SLP76 remained unaffected and similar to healthy control T cells. Our hypothesis, that low TCR-induced p-ERK marks dysfunctional T cells in FL, is further supported by the current understanding that impaired activation of ERK is an indicator of T-cell anergy. This is based on observations showing that uncoupling of the ERK pathway is an important underlying mechanism in antigenic unresponsiveness of T cells. Antigen recognition under suboptimal conditions, such as lack of costimulation or upregulation of coinhibitory receptors, can act to disrupt TCR-induced p-ERK, hence resulting in poor T-cell effector function (60, 61).

While TIGIT can recruit SHIP1 to modulate cell function, PD-1 blocks signaling events downstream of the TCR by recruiting the protein tyrosine phosphatases SHP1 and SHP2. These phosphatases can inhibit phosphorylation of signaling effectors both proximal and distal to the TCR (62–64). In fact, we found low levels of TCR-induced p-ERK to be associated with TIGIT as well as PD-1 expression. This indicates that TIGIT and PD-1 may both contribute to the dysfunctional TCR-induced signaling observed in FL, potentially by recruitment of different phosphatases. In context with the finding that TIGIT and PD-1 were the two major coinhibitory receptors, and often coexpressed by FL T cells, this provides a rationale for coblockade of these receptors to improve T-cell activity and tumor killing. Although not yet explored in lymphoma, coblockade of TIGIT and PD-1 has generated promising results from preclinical studies in other cancer types. Combined blockade of the two receptors led to complete responses in tumor mouse models of breast and colorectal cancers, while blocking only one receptor had little effect (27). Furthermore, coblockade of PD-1 and TIGIT led to increased IFNγ production in CD8 TILs from melanoma patients, and TIGIT blockade was able to restore cytokine production in CD8 T cells from AML patients (26, 65).

In conclusion, our results provide new insights into mechanisms that may contribute to immune suppression in FL. In-depth mapping of coinhibitory receptor expression and functional assessment in distinct T-cell subtypes will enhance our biological understanding for the complex regulation of antitumor T-cell responses, and exploiting this further in relation to immune checkpoint blockade is needed to further enhance the precision of this therapy.

No potential conflicts of interest were disclosed.

Conception and design: S.E. Josefsson, K. Huse, E.M. Inderberg, E.B. Smeland, R. Levy, J.M. Irish, J.H. Myklebust

Development of methodology: S.E. Josefsson, K. Huse, R. Levy, J.H. Myklebust

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S.E. Josefsson, K. Huse, A. Kolstad, K. Beiske, B. Østenstad, R. Levy

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): S.E. Josefsson, K. Huse, K. Beiske, C.B. Steen, O.C. Lingjærde, E.B. Smeland, R. Levy, J.M. Irish, J.H. Myklebust

Writing, review, and/or revision of the manuscript: S.E. Josefsson, K. Huse, A. Kolstad, K. Beiske, C.B. Steen, E. M. Inderberg, B. Østenstad, E.B. Smeland, R. Levy, J.M. Irish, J.H. Myklebust

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): S.E. Josefsson, K. Huse, A. Kolstad, R. Levy

Study supervision: K. Huse, J.H. Myklebust

Other (application to regional ethics committee): A. Kolstad

Other (supply of reagents generated by herself): D. Pende

We thank Eva Kimby for critical review of the manuscript. This work was supported by the Research Council of Norway (FRIMEDBIO 230817/F20; S.E. Josefsson) and Centre of Excellence (Centre for Cancer Biomedicine; J.H. Myklebust and E.B. Smeland), the Norwegian Cancer Society (162948; K. Huse, 163151; E.B. Smeland, and 162844; J.H. Myklebust), and Associazione Italiana per la Ricerca sul Cancro (AIRC IG-16764; D. Pende).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.