Research ArticleCancer Immunotherapy

Targeting TNFR2 with antagonistic antibodies inhibits proliferation of ovarian cancer cells and tumor-associated Tregs

See allHide authors and affiliations

Science Signaling  17 Jan 2017:
Vol. 10, Issue 462, eaaf8608
DOI: 10.1126/scisignal.aaf8608

Killing ovarian cancer by trapping TNFR2

Signaling by tumor necrosis factor (TNF) family promotes tumor growth and progression. The TNF receptor 2 (TNFR2) is present on the surface of immunosuppressive regulatory T cells and some tumor cells. Torrey et al. (see coverage by Chen and Oppenheim) developed antibodies that bind and lock TNFR2 in an inactive conformation. The antibodies killed ovarian cancer cells in culture. Additionally, the antibodies inhibited the proliferation of regulatory T cells while promoting the proliferation of effector T cells isolated from metastatic sites (ascites) in ovarian cancer patients. The antibodies had less of an effect on T cells isolated from the peripheral blood of normal donors. Thus, these antibodies may be more specific and less toxic than current TNFR antibodies, providing a new path for treating ovarian cancer.


Major barriers to cancer therapy include the lack of selective inhibitors of regulatory T cells (Tregs) and the lack of broadly applicable ways to directly target tumors through frequently expressed surface oncogenes. Tumor necrosis factor receptor 2 (TNFR2) is an attractive target protein because of its restricted abundance to highly immunosuppressive Tregs and oncogenic presence on human tumors. We characterized the effect of TNFR2 inhibition using antagonistic antibodies. In culture-based assays, we found that two TNFR2 antagonists inhibited Treg proliferation, reduced soluble TNFR2 secretion from normal cells, and enabled T effector cell expansion. The antagonistic activity occurred in the presence of added TNF, a natural TNFR2 agonist. These TNFR2 antibodies killed Tregs isolated from ovarian cancer ascites more potently than it killed Tregs from healthy donor samples, suggesting that these antibodies may have specificity for the tumor microenvironment. The TNFR2 antagonists also killed OVCAR3 ovarian cancer cells, which have abundant surface TNFR2. The antibodies stabilized antiparallel dimers in cell surface TNFR2 that rendered the receptor unable to activate the nuclear factor κB pathway and trigger cell proliferation. Our data suggest that, by targeting tumor cells and immunosuppressive tumor-associated Tregs, antagonistic TNFR2 antibodies may be an effective treatment for cancers positive for TNFR2.


Antibody immunotherapy is showing great promise in cancer (1). One component of immunotherapy is to target and eliminate the abnormal numbers of potent host regulatory T cells (Tregs) that mediate the suppression of the immune response (2). A targeted approach to remove or inactivate host Tregs is desirable but has been challenging (35). Many surface receptors of Tregs are found diffusely in the immune system, leading to systemic toxicity in humans, including lethal autoimmunity (6).

A recognized subtype of Tregs is a tumor necrosis factor receptor 2 (TNFR2)–expressing population. Tregs in both mice and humans that express the TNFR2 are potent suppressors of immune responses and are abnormally abundant in human and murine tumors (2, 711). Furthermore, aberrant expression of TNFR2 on tumor cells promotes the growth of colon cancer, multiple myeloma, renal cell carcinoma, Hodgkin’s lymphoma and cutaneous non-Hodgkin’s lymphoma, and ovarian cancer (1217). TNFR2 signaling results in constitutive downstream agonism and augmented cell proliferation through nuclear factor κB (NF-κB) signaling. A single approach to successfully block potent Tregs and also directly inhibit tumor growth through the TNFR2 oncogene is desirable. Such a method would ideally enable the T effector cells (Teffs) to expand and function to kill the tumor as well.

TNFR2 is a TNF superfamily receptor (18). The signaling circuitry of TNFR2 is different from that of TNFR1; TNFR2 is not linked to a death domain but instead promotes NF-κB activation and cell growth. Most TNF superfamily receptors are expressed on all lymphoid and commonly on parenchymal cells. However, TNFR2 has limited expression in the immune system, is induced by the ligands TNF and interleukin-2 (IL-2), and is restricted to minor subpopulations of the lymphoid system, including potent Tregs, myeloid suppressor cells, endothelial cells, and select neurons during growth in mammals (19, 20). This restricted expression of TNFR2 makes it an ideal drug target because systemic toxicity from an antibody-based therapy is less likely to occur. TNFR2 is an attractive candidate for additional reasons. In various human and murine cancers, abundant TNFR2-positive Tregs are found within the tumor microenvironment. Gene duplication and activating mutations in TNFR2 have been found in cancer (13). Studies in baboons show that the known toxicity of high-dose TNF is solely mediated by TNFR1 but not TNFR2 (21, 22). TNFR2 is present at a 10-fold higher density than TNFR1 in naturally occurring Tregs in human blood (19). Mouse models lacking TNFR2 show improved immune system capacity to respond to and kill diverse cancer types without progression to systemic autoimmunity (2331).

Features of tumor immunosuppression are the presence of too many TNFR2-positive Tregs at the tumor site (as seen in ovarian cancer patients) or in the circulation (as seen in lung cancer patients). The greater cell surface abundance of TNFR2 on cancer cells can be due to gene duplications or other modes (as seen in cutaneous non-Hodgkin’s lymphoma patients) (11, 13, 32, 33). In other cancers, the overall abundance of TNFR2-positive Tregs in the tumor is higher than that in the patient’s peripheral blood, but the precise genetic basis for its dysregulation has yet to be defined (11).

TNFR2 is an immune system marker of suppressive T cell types, a master switch for Treg survival and fate (even in healthy adults), and a newly discovered and broadly expressed oncogene (34). In various cancers, the most suppressive Tregs can express excessive amounts of TNFR2 (TNFR2hi Tregs) and exert potent immunosuppressive effects (11), whereas reduced numbers of TNFR2-positive Tregs correlate with better clinical responses in patients with acute myeloid leukemia (35) and lung cancer (36). Colon cancer, multiple myeloma, Hodgkin’s lymphoma, ovarian cancer, and cutaneous T cell lymphomas can also aberrantly express TNFR2 as a growth receptor oncogene on the tumor itself (1216, 37). For noncutaneous T cell lymphomas, the genetic basis of the abnormal TNFR2 expression is associated with constitutive overexpression of TNFR2 from frequent gene duplications or cytoplasmic TNFR2 mutations that confer constitutive agonism (13). These features make TNFR2 an advantageous molecular target for Treg inactivation in the tumor microenvironment and direct tumor targeting as an aberrant surface oncogene.

One way to inactivate TNFR2-positive Tregs is to develop antagonistic antibodies to TNFR2. Although most receptor-coating TNFR2 antibodies show no effectiveness as TNFR2 antagonists, we identified and developed two potent TNFR2 antagonistic antibodies with similar in vitro kinetics that inactivate human Tregs. The antagonistic antibodies are even more potent inhibitors of TNFR2 Tregs from ovarian cancer patients than from normal donors. These newly identified TNFR2 antagonistic antibodies do not require Fc binding, bind to the same region of the receptor, express dominance over TNF-mediated agonism, and hamper intracellular NF-κB activation and phosphorylation that is obligatory for TNFR2 signaling–mediated cell proliferation. Using linear and three-dimensional (3D) epitope mapping, we established a model of dominant antagonism by these TNFR2 antibodies through the binding and stabilization of a unique antiparallel dimeric conformation of surface TNFR2 that inhibits intracellular signaling, cannot bind TNF, cannot be cleaved to create soluble TNFR2, and has preferential potency against rapidly dividing cancer cells. Even low doses of TNFR2 antagonists rapidly killed TNFR2-positive ovarian cancer cells in culture. Dominant antagonism creates a unique nonsignaling complex from newly appearing surface TNFR2 that has implications for the therapeutic targeting of TNF superfamily receptors especially TNFR2.


TNFR2 antagonists inhibit Tregs and permit Teff proliferation

To identify and characterize the functional activity of possible TNFR2 antagonistic monoclonal antibodies (mAbs), we first used a short-term 48- to 72-hour cell-based assay for Treg expansion or inhibition on freshly isolated T cells from healthy donors. This assay used freshly purified CD4 T cells from peripheral blood and was performed in a dose-response manner to identify killing trends on Tregs or proliferative trends on Teffs (34).

We first confirmed that normal human CD4+ cells incubated with increasing concentrations of TNF, a known agonist of TNFR1 and TNFR2, show a dose-dependent increase in Tregs (Fig. 1A and fig. S1A). We also confirmed that IL-2 is required for Treg proliferation and also aids in Treg expansion (Fig. 1A). TNF at 20 ng/ml induced Treg proliferation (>20% expansion), and TNFR2 agonistic antibody induced even greater Treg proliferation (40% increase) relative to IL-2 expansion alone (Fig. 1B, n = 6 subjects; P < 0.01). Coincubation of TNFR2 agonist with TNF was synergistic, with >60% expansion of Tregs relative to IL-2 expansion alone (Fig. 1B; P < 0.01).

Fig. 1 Selective TNFR2 antibodies inhibit Treg and enable Teff proliferation.

(A) Percentage of Tregs in a culture of freshly isolated peripheral human CD4+ cells in response to TNF and IL-2 (200 U/ml; left) or IL-2 alone (right) for 48 hours. Data are means ± SEM from n = 4 subjects; data on the right are from a representative of the four samples. (B) Fluorescence-activated cell sorting (FACS) analysis for Tregs (CD4+CD25hiFoxP3+) in CD4+ cells from normal donors cultured with IL-2 (200 U/ml) and either TNF (20 ng/ml), a TNFR2 agonist (12.5 μg/ml), TNFR2 antagonists (12.5 μg/ml), or a combination thereof. Data are means ± SEM from n = 6 subjects; P < 0.01 for TNF plus IL-2 versus IL-2 alone and P < 0.01 for TNFR2 agonist plus IL-2 versus IL-2 alone; unpaired t test. (C) Percentage of Tregs in a culture of peripheral human CD4+ cells treated as indicated, testing the dose-dependent effects of TNFR2 antagonists 1 (left) and 2 (right). All conditions included the presence of IL-2 (200 U/ml). Data are means ± SEM from n = 9 subjects; P < 0.05 for TNFR2 antagonist 1 at 12.5, 25, and 50 μg/ml and TNFR2 antagonist 2 at 12.5 and 50 μg/ml compared to no antagonist; unpaired t test. (D) Proportion of Teffs in cultures described in (C). Data are means ± SEM from n = 10 subjects; P < 0.05 for TNFR1 antagonist 1 at 0.1 and 0.5 μg/ml; unpaired t test.

Both TNFR2 antagonistic antibodies suppressed Treg proliferation, resulting in a dose-dependent decrease in the percentage of remaining Tregs even with short incubation times of 48 to 72 hours (Fig. 1, B and C, n = 6 subjects; P < 0.05). The TNFR2 antagonistic effect even overcame the presence of a generous concentration of TNF (20 ng/ml) (Fig. 1B, n = 6 subjects). When the two TNFR2 antagonistic antibodies were studied in the presence of TNF, both overcame TNF agonism in a dose-dependent fashion, decreased Treg expansion, and inverted the TNF agonistic curve (fig. S1B, n = 8 subjects; fig. S1C, n = 4 subjects; fig. S2A, n = 2 representative samples). Together, these results demonstrate the functional ability of TNFR2 antibodies to suppress Treg proliferation in normal human CD4+ cell cultures, with the effect persisting despite the addition of moderate to high concentrations of TNF.

One of the desired functional effects of Treg inhibition is the expansion of Teffs. In normal sample donors treated with TNFR2 antagonism, we studied the samples for reciprocal Teff proliferation. The elimination of Tregs was accompanied by the early expansion of Teffs (Fig. 1, C and D, and fig. S2B) (n = 10 subjects; P < 0.05).

Some inflammatory immune responses are associated with an increase in soluble TNFR2, such as type 2 diabetes and acute myocardial infarctions, but it is not known whether serum-soluble TNFR2 functions as a sink for lowering TNF concentrations (38, 39). We therefore measured the amount of soluble TNFR2 in the culture supernatant after treating CD4 T cell cultures with IL-2, TNF, TNFR2 agonistic antibody, and TNFR2 antagonistic antibodies. TNF and the TNFR2 agonistic antibody increased the amount of soluble TNFR2, whereas TNFR2 antagonistic antibodies reduced it (Fig. 2A, n = 10 subjects). From a treatment perspective, this inhibition of soluble TNFR2 shedding is beneficial because the presence of soluble TNFR2 would effectively sequester TNFR2 antagonistic antibody, reducing the amount available for direct cell binding, thereby decreasing its in vivo potency.

Fig. 2 TNFR2 antagonists inhibit the secretion of soluble TNFR2 and suppress Tregs.

(A) Enzyme-linked immunosorbent assay (ELISA) quantification of soluble TNFR2 in a culture of freshly isolated peripheral human CD4+ cells in response to IL-2 (200 U/ml) and either TNF (20 ng/ml), a TNFR2 agonist (12.5 μg/ml), or TNFR2 antagonists (12.5 μg/ml) for 48 hours. Data are means ± SEM from n = 10 subjects; ***P < 0.001 for TNF and TNFR2 agonist and *P < 0.05 for TNFR2 antagonist 1; unpaired t test. (B) Suppression of CD8+ T cells by Tregs cultured for 14 to 17 days with IL-2 (200 U/ml) alone or with TNFR2 agonist (12.5 μg/ml), or TNFR2 antagonists (12.5 μg/ml). Responder cells were stained with CFSE, mixed with cultured Tregs, incubated for 4 to 6 days, and then analyzed by FACS. Data are means ± SEM from n = 4 subjects; *P < 0.05 for TNFR2 antagonists; unpaired t test. (C) FACS analysis of CD8+ T cell suppression by various ratios of Tregs treated as described in (B) with IL-2 (50 U/ml) alone or with TNFR2 agonist or antagonist 1 (12.5 μg/ml). Data are representative results from one of four subjects.

One function of Tregs is to suppress the function of cytotoxic CD8+ T cells. Carboxyfluorescein succinimidyl ester (CFSE) assays of suppressor activity found potent suppression of CD8+ cells by TNFR2 antibody agonists and reciprocal release of CD8 suppression by TNFR2 antibody antagonists (Fig. 2B, n = 4 subjects; Fig. 2C, representative sample). These functional data were consistent with the data showing Teff proliferation (Fig. 1D).

TNFR2 antagonistic activity is independent of Fc region or cross-linking

Nonspecific binding by antibody Fc regions can result in functional activity often referred to as antibody-dependent cell-mediated cytotoxicity (ADCC) and is often required for antibody functional activity (40). Antibodies that require ADCC or Fc regions for activity can be more restricted in clinical utility because natural killer cells and monocytes with Fc receptors must be nearby to stabilize antibody binding to the Treg. Another clinical limitation of Fc-mediated antibody binding stems from heterogeneity in Fc receptor genotypes, requiring human treatments to be tailored to human Fc receptor variants. Treatment of CD4+ cells with TNFR2 monoclonal F(ab′)2 fragments from both TNFR2 antagonistic antibodies 1 and 2, with and without TNF, resulted in similar dose-dependent decreases in Treg percentages within the CD4 cell pool. These data were comparable to the full mAbs (fig. S3A, representative sample). This suggests that specific binding by the F(ab′)2 region to TNFR2, rather than nonspecific Fc binding, is likely responsible for the antagonistic activity.

Cross-linking can also result in aberrant functional activity of antibodies. To rule out the possibility that nonspecific cross-linking was involved in the observed functional activity or that cross-linking could aberrantly turn an antagonist into an agonist, we performed a dose-response assay of the TNFR2 antagonistic antibodies 1 and 2, with and without anti-rodent, anti–immunoglobulin G (IgG) cross-linking antibody (an antibody directed to rodent IgG protein). The TNFR2 antagonist dose-dependent suppression of Treg was unaffected by secondary cross-linking (fig. S3B, representative sample). Thus, the functional activity of both TNFR2 antagonistic antibodies was independent of Fc region or receptor cross-linking using exogenous IgG methods.

NF-κB activation pathways and gene expression are inhibited by TNFR2 antagonistic antibodies

NF-κB signaling is required for TNF-mediated cancer cell growth and has been proposed as a potential target for cancer therapy. With the decrease in the total numbers and the percentage of remaining Tregs after treatment with TNFR2 antagonistic antibodies, we expected that the underlying signaling mechanism would involve a reduction in late NF-κB activation steps that result in cell proliferation. To investigate the molecular response to TNFR2 antagonism, we measured the gene expression of eight NF-κB signaling genes commonly stimulated with TNFR agonism and cell growth (CHUK, NFKBIE, NFKBIA, MAP3K11, TRAF2, TRAF3, relB, and cIAP2/BIRC3). As a control, we also monitored cytokine mRNA for TNF and lymphotoxin, as well as two markers of Tregs, FOXP3 and CD25. Treatment with TNFR2 antagonistic antibody 2 resulted in down-regulation of NF-κB–related gene expression compared to treatment with TNF (Fig. 3A). Next, using phosphorylated RelA/NF-κB p65, we demonstrated that treatment of CD4+ cells with TNFR2 antagonistic antibodies reduced NF-κB activation, whereas treatment with TNF increased NF-κB activation (Fig. 3B, representative sample; fig. S6, n = 4 subjects). This suggests that the intracellular signaling effect of TNFR2 antagonists was similar to one another and suppressive. Supportively, kinetic analysis of binding to recombinant human TNFR2 revealed that there was no major difference in association or dissociation rates between the two antibodies (Table 1A).

Fig. 3 Gene expression and signaling pathway analysis of TNFR2 antagonism.

(A) Gene expression analysis by real-time polymerase chain reaction (PCR) of key mRNA encoding cytokines (TNF and lymphotoxin), markers of Tregs (FoxP3 and CD25), and promoters of NF-κB. Freshly isolated peripheral human CD4+ cells were cultured with IL-2 (50 U/ml) alone or with TNF (20 ng/ml) or TNFR2 antagonist 1 (2.5 μg/ml) for 3 hours before RNA isolation. LT, lymphotoxin; cIAP, also known as BIRC3. Data are means ± SEM from n = 12. (B) ELISA quantification of the phosphorylation RelA/NF-κB p65 after 10-min incubation of fresh CD4+ cells with IL-2 (200 U/ml) and various concentrations of TNF (0.2 to 20 ng/ml) or TNFR2 antagonists (0.02 to 25 μg/ml). Data are representative results from one of four subjects. RFU, relative fluorescence units.

Table 1 Binding affinity and linear peptide mapping of TNFR2 antagonists.

(A) Assessment of TNFR2 antibody affinity. Kon = Ka, association rate constant; Koff = Kd, dissociation rate constant; KD, equilibrium dissociation constant. (B and C) Linear peptide mapping confirmed the conservation of the epitopes for antibody binding to TNFR2. n/a, not applicable.

View this table:

Dominant TNFR2 antagonists bind to identical TNFR2 regions but differ from the binding sites of recessive antagonists

To further investigate the specific binding characteristics of the TNFR2 antagonistic antibodies, we performed epitope mapping. Linear peptide mapping of TNFR2 antagonists A and B identified antibody binding sites in the cysteine-rich domain 2 (CRD2) region of TNFR2. These TNFR2 antagonistic antibodies were designed to prevent TNF binding and TNFR2 activation but had unfavorable characteristics in the Treg assays. They first performed as Treg inhibitors when used in the 48-hour T cell assays with IL-2 on normal donor cells (Fig. 4A). TNFR2 antagonistic antibodies A and B demonstrated dose-dependent Treg antagonism from 0 to 25 μg/ml (Fig. 4A, left). However, when TNF (20 ng/ml) was added to the cultures, the TNFR2-directed antibodies A and B competed poorly with TNF and thus could be characterized as recessive TNFR2 antagonists (Fig. 4A, right). In each case, TNF agonism was dominant over this form of recessive TNFR2 antagonism, demonstrating that the design of antibodies with this intended purpose and to the CRD2 region of TNFR2 was less powerful than predicted (Fig. 4B).

Fig. 4 Recessive TNFR2 antagonistic antibodies to the TNFR2 trimer inhibit Tregs but cannot overcome TNF agonism.

(A) Percentage of Tregs in a culture of freshly isolated peripheral human CD4+ cells in response to recessive TNFR2 antibodies A and B that were made against the TNFR2 receptor in the hopes of preventing the TNFR2 complex from forming with TNF. Tregs in cultured CD4+ cells were defined by FACS analysis as FoxP3+ and CD25hi. Data are representative results from two of at least five antibodies. (B) Surface representation of the TNF-TNFR2 complex and the binding regions of the recessive TNFR2 antibodies. The TNFR2 trimer is shown in blue, the TNF trimer is shown in yellow, and the orange regions indicate the mapped antibody binding sites for the recessive TNFR2 antagonists. The 3D models were created by PyMOL.

Linear peptide mapping of the two dominant TNFR2 antagonists (1 and 2) revealed that, although both mAbs had high affinity to the full-length TNFR2 protein in plate-based assays, only one antagonistic antibody mapped with many overlapping peptides. Specifically, TNFR2 antagonist 1 had low to moderate affinity at amino acids 112 to 139 and strong affinity at amino acids 128 to 155, whereas antagonistic antibody 2 did not bind to any linear peptide region (Table 1B). There was no negative or positive effect on linear peptide binding by TNFR2 antagonistic antibody 1 in the presence of TNF, indicating that TNF was not a competitor for the presumed TNFR2 antagonist binding site (Table 1C). Because linear peptide mapping could not identify the binding by TNFR2 antagonist 2, we conducted 3D binding analysis using Pepscan technology for both antibodies. For both dominant TNFR2 antagonists 1 and 2, the 3D data showed overlapping binding regions from about amino acids 138 to 169 that map to the CRD3 and CRD4 regions of TNFR2 (Fig. 5A).

Fig. 5 TNFR2 conformational epitopes of dominant TNFR2 antagonists: An antiparallel dimeric form of TNFR2 is optimal antagonistic binding.

(A) Mapping of the discontinuous epitopes on TNFR2 recognized by dominant TNFR2 antagonists presented as possible binding regions on different published forms of TNFR2 receptor: antiparallel dimer and the TNF-TNFR2 complex. Conformational epitopes for TNFR2 antagonist (shaded red) are mapped onto surface representation of the antiparallel dimer and the TNF-TNFR2 complex. TNFR2 is shown in blue, and TNF is shown in yellow. The 3D models were created by PyMOL. (B) Mapping as described in (A). Top right: The proposed hexagonal network of antiparallel TNFR2 dimers representing the inhibitory state. Left: A single antiparallel dimer is shown in side view modeled on the cell surface. Bottom left: A single TNF-TNFR2 complex. Bottom right: A model of a dominant antagonist antibody bound to an antiparallel dimer.

Previously, we had unsuccessfully tried to make antagonistic TNFR2 antibodies to prevent TNF binding and TNFR2-driven agonism. The published 3D structure of the TNF-TNFR2 complex shows a central TNF homotrimer surrounded by three TNFR2 receptors as is commonly observed for receptors of the TNF superfamily (Fig. 4B) (41). This conformation had also been observed in many other TNF superfamily members including TNFR1 with lymphotoxin, DR5 with TRAIL ligand, and OX40 receptor with OX40L (4246). The binding sites of the recessive antagonistic antibodies in the TNF-TNFR2 complex structure were to the outside to prevent TNF binding. However, they did not explain the requirement for the obligatory F(ab′)2 structure because a single Fab should be able to displace TNF. This TNFR2 structure for antagonism also did not explain why the dominant antagonistic antibodies preferentially bind in the lower CRD3 and CRD4 regions of TNFR2, whereas the recessive antibodies bind in the upper CRD2 portion (Fig. 5A). Next, we looked at the published and less considered antiparallel and TNF-independent forms of TNFRs, the antiparallel forms that have only been observed in solutions (47, 48). We found that only one model of dominant TNFR2 antagonistic binding was optimal and possible from the functional data: The antiparallel dimer was consistent with the functional properties of the two dominant TNFR2 antagonistic binding regions that were spaced sufficiently apart for the obligatory F(ab′)2 TNFR2 antagonistic binding on a single antiparallel dimer (Fig. 5B, lower right). In the antiparallel dimer complex, the TNF binding region was buried in the dimer interface firmly locked-in by antagonistic antibody 1 or 2, which explains their dominance over added TNF and strong antagonism even with escalating TNF concentrations.

We propose that, on the cell surface, the ligand-free receptor antiparallel dimers (stabilized or formed by TNFR2-dominant antibodies) arrange, forming a hexagonal lattice, but this complex cannot signal because the tight clustering of the TNFR2 trimers and TNF binding is blocked, and intracellular regions are further apart, preventing the intracellular scaffolding (Fig. 5B, upper right). This model also explains why the recessive antagonistic antibodies that bind to the CRD2 region are less effective. The CRD2 regions are too close together in the antiparallel dimer and can only accommodate one Fab binding; the other Fab would have to bind to neighboring antiparallel dimer in the hexagonal lattice, a less efficient arrangement that can be more easily disrupted by TNF at high concentrations. Also, it would be difficult to envision that even a very high affinity recessive TNFR2 antagonist could displace an already bound and active TNF-TNFR2 trimer complex.

TNFR2 antagonistic activity on Tregs from ovarian cancer ascites has enhanced tumor microenvironment sensitivity

The Tregs of cancer are extremely potent immunosuppressors, especially the Tregs from tumor sites compared to the Tregs of peripheral blood of cancer patients or even control subjects (11). To begin to understand the potency of TNFR2 antagonistic antibody on a tumor-residing Treg, fresh ovarian cancer Tregs from ascites fluid were isolated and compared to normal Tregs isolated from a normal blood donor. With TNFR2 antagonism, the ovarian cancer Tregs were exponentially more sensitive to TNFR2 antagonism than were the Tregs from a normal blood donor (Fig. 6A). This pattern of heightened tumor microenvironment Treg death with TNFR2 antagonist was highly reproducible. Pooled data from the Tregs from ovarian cancer ascites were compared to the Tregs from normal donors (Fig. 6C, n = 3 subjects). Killing of tumor-residing Tregs occurred at lower doses and to a more complete degree even with short-term 48-hour assays. Also, as expected, if the Tregs of cancer are potent and killed, then the reciprocal Teff response should also be prominent. The Teffs of the cancer subject proliferate with TNFR2 antagonism and do so more than the Teff of normal peripheral blood (Fig. 6, B and D). We conclude that TNFR2 antagonists have specificity for the Tregs of the tumor microenvironment.

Fig. 6 Impact of TNFR2 antibodies on Tregs from the ovarian cancer microenvironment compared to Tregs from normal donors.

(A) Percentage of Tregs in a culture of freshly isolated CD4+ cells from peripheral blood of normal donors (black) or from ascites fluid of the ovarian cancer cell microenvironment (blue), in response to TNFR2 antagonist 1 for 48 hours. Data are representative results from one of three subjects. (B) Proportion of Teffs in cultures described in (A). (C) Relative change in the proportion of Tregs in cultures described in (A). Data are means ± SEM from n = 3 subjects; P < 0.05 for Tregs of cancer with TNFR2 antagonist 1 at 25 and 50 μg/ml; P < 0.05 for Tregs from normal donor with TNFR2 antagonist 1 at 50 and 125 μg/ml; unpaired t test. (D) Relative change in the proportion of Teffs in cultures described in (A). Data are means ± SEM from n = 3 subjects; P < 0.05 for Teff from normal donors with TNFR2 antagonist 1 at 5 μg/ml; unpaired t test. Tregs were defined by FACS analysis as FoxP3+ and CD25hi, and Teffs were defined as FoxP3 and CD25hi. All culture conditions included IL-2 (200 U/ml).

We sought in a preliminary manner to understand why the TNFR2 antagonistic antibodies might be more selective and more potent on the Tregs of the cancer site. The structural biology data suggest that dominant TNFR2 antagonists capture newly synthesized and recently appearing membrane forms of TNFR2 on the cell surface; therefore, inhibition of cell growth might prevent heightened Treg killing in the tumor microenvironment. Freshly isolated CD4 T cells were either treated with mitomycin C (50 μg/ml) or not treated before IL-2 with or without TNF, TNFR2 agonist, or the TNFR2 antagonistic antibodies. The data show that TNFR2 antagonist killing of Treg was blocked by inhibition of cell division through mitomycin C (fig. S4). This supports the concept that the tumor microenvironment specificity might be in part driven by faster proliferation of the Treg at the tumor site and the capture by the antagonists of only newly synthesized TNFR2 proteins.

Potent TNFR2 antagonistic activity is observed on TNFR2 oncogene–expressing tumor cells

TNFR2 is rapidly becoming appreciated as a broadly expressed oncogene on many human tumors (1217). We ordered through American Type Culture Collection (ATCC) the well-studied OVCAR3 ovarian cancer cell line. Staining of OVCAR3 with TNFR2 antibodies rapidly identified TNFR2 oncogene surface expression (fig. S5B). Ovarian adenocarcinoma cell line OVCAR3 was treated with a range of TNFR2 antibody concentrations (0 to 50 μg/ml) for 7 days and studied in a live/dead cell assay. In the live/dead cell assay, Fig. 7A shows dose-dependent killing with the TNFR2 antagonistic antibody with the first OVCAR3 tumor killing at 2.5 μg/ml (20% dead) and maximal killing at 12.5 μg/ml (80% dead). Next, time-dependent and dosing experiments of OVCAR3 with TNFR antagonistic antibody 1 either at 2.5 μg/ml or at 12.5 μg/ml are presented (Fig. 7B). At moderate doses of TNFR2 antibody (12.5 μg/ml), killing of OVCAR3 was complete by day 3. At very low doses of TNFR2 antagonistic antibody (2.5 μg/ml), killing by the TNFR2 antagonist at a rate of 50% was observed at day 3 and complete by day 7. Last, the relative change in the proportion of live cells remaining after a 7-day incubation with TNFR2 also showed OVCAR3 cell death (Fig. 7C). The TNFR2 antagonistic antibody was an efficient killer of the TNFR2 oncogene–expressing OVCAR3 cell line.

Fig. 7 TNFR2 oncogene–expressing tumor cells are directly eliminated by TNFR2 antagonism.

(A) Ovarian adenocarcinoma cell line, OVCAR3, which expresses the TNFR2 oncogene, was treated with TNFR2 antagonist 1 (0 to 50 μg/ml) for 7 days, and survival was monitored by live/dead staining and FACS analysis. Data are means ± SEM from n = 3 independent experiments; P < 0.05 for TNFR1 antagonist 1 at 5, 12.5, 25, and 50 μg/ml; unpaired t test. (B) Time-dependent killing of OVCAR3 cancer cells by TNFR2 antagonist 1 at 2.5 and 12.5 μg/ml for up to 17 days determined by direct live cell counts. Data are means ± SEM, results of one of at least five independent experiments measured in triplicate. (C) Relative change in the proportion of live OVCAR3 cells as described in (A). Data are means ± SEM from n = 3 independent experiments; P < 0.05 for TNFR1 antagonist 1 at 5 to 50 μg/ml; unpaired t test.


The TNFR2 protein is a cell surface protein for the expansion of potent Tregs and an aberrantly expressed cell surface oncogene on diverse human tumors (1217). Here, we identified and characterized dominant TNFR2 antibody antagonists that inhibited Tregs, with a greater potency in cultures of cancer-associated Tregs than in those of normal, peripheral Tregs. We also found that low doses of TNFR2 antagonists killed a TNFR2-positive cell line, OVCAR3. The TNFR2 antibodies exhibited the ability to shut down early intracellular phosphorylation events downstream of TNFR2 that precede NF-κB–dependent cell proliferation and to suppress soluble TNFR2 secretion. The TNFR2-dominant antagonists succeeded even in the presence of high concentrations of TNF, putatively stabilizing a nonsignaling cell surface complex of TNFR2 antiparallel dimers. The dose- and time-response curves established that the TNFR2-expressing Tregs of the tumor microenvironment were more sensitive than normal Tregs to TNFR2 antagonism. Thus, TNFR2 antagonism provides a two-pronged approach for targeting the Tregs of the tumor microenvironment and the TNFR2 oncogene–expressing cancers.

The identification of TNFR2 as a broadly expressed human oncogene has been largely ignored. On the basis of the published literature, TNFR2 is known to promote the proproliferative NF-κB signaling pathway, thus making it a prime candidate to direct cellular expansion in tumors as well as infiltrating Tregs of tumors that express this receptor. Abnormally increased expression of TNFR2 is common in various tumors, including but not limited to breast cancer, carcinoid, cervical cancer, colorectal cancer, endometrial cancer, glioma, head and neck cancer, liver cancer, lung cancer, and lymphoma (17). Gradually, the polymorphisms of this oncogene are being defined with the most advanced data on cutaneous T cell lymphoma; now, additional data on non–small cell carcinoma, esophageal squamous cell carcinoma, and others are accumulating (33, 49, 50). Moving forward, it will be important to ensure that TNFR2 antagonism through these antibodies can kill all oncogenic versions of the TNFR2 proteins expressed on cancer cells. To date, we have not seen variability in the death of TNFR2-positive Tregs in fresh human ovarian ascites, nor have we identified an ovarian cell line with TNFR2 that is not susceptible to the antibodies in this report.

The shared receptor binding regions of the two TNFR2 antagonists lead us to a different possible model of antibody binding to the TNFR2 that causes stable and dominant antagonism. The commonly studied trimeric form of TNFR2 would not allow the newly identified antibody antagonists to bind to their amino acid sequences or explain how these antagonistic antibodies dominate over TNF or IL-2 and especially act as dominant antagonists on cancer-residing Tregs (Fig. 6) (41). Our years of work generating TNFR2 antibodies to the binding region of TNF in the trimer TNFR2 structure or to the exterior surface of the TNF binding opening of the trimer did not generate potent antagonists that dominated over TNF or TNF agonism. The TNFR2 trimer–directed antibodies designed to compete with TNF were never dominant, albeit neutral at times when challenged with the inflammatory environment of TNF or IL-2.

Previous studies have proposed that the ligand-free form of TNF superfamily members on the cell surface can exist as antiparallel dimers, especially as it relates to the closely matched TNFR1. The antiparallel dimers are of special interest because they occlude the TNF binding site and prevent TNF-driven trimerization for proper intracellular signaling (47). If TNFR2 antagonistic antibodies stabilized the antiparallel dimer, then this explains not only the dominance of antagonism over TNF but also the requirement for at least F(ab′)2 structures and the failure of the Fab antibody fragment. Also, no benefit was observed by adding a cross-linking reagent to either augment antagonism or, worse yet, conversion of antagonism to agonism. The antiparallel dimeric form of TNFR2 fits best with the accessible binding site for antagonistic antibodies and the functional assays. It also provides a way to reconcile data and provide a unified model of TNF superfamily signaling. To maintain the required threefold symmetry for signaling, the antiparallel receptor dimers may arrange on the cell surface in a stable hexagonal lattice. On the intracellular side, the cytoplasmic TNFR2 domains would be distant, and this would prevent soluble receptor cleavage. The dimer is formed by receptor monomers from two different signaling units that occlude the TNF ligand–binding site, maintaining a nonsignaling inhibitory state. This state can most effectively be stabilized by the newly characterized dominant antagonistic antibodies that firmly lock in the antiparallel dimer.

The Tregs in the tumor microenvironment are extremely potent suppressors of the immune response compared to the Tregs from normal donors or even the Tregs from the peripheral blood of the same donor (11). We observe with TNF or TNFR2 agonism the same potency of the cancer Tregs to expand rapidly in culture with IL-2, TNF, or TNFR2 agonism. Remarkably, we also see heightened inhibition of ovarian cancer Tregs with TNFR2 antagonism mediated through dominant TNFR2 antibodies. This is favorable in light of recent findings about the identity of the TNFR2 gene sequence that allows extra TNFR2 potency in cancer. In cutaneous T cell lymphoma, the gene for TNFR2 can be duplicated many times or can even be mutated in the intracellular region to confer constitutive agonism directly on cancer cells. This genetic pressure to constantly maintain TNFR2 growth signals is in line with exaggerated growth of the tumor itself as well as the associated expansion of the potent Tregs (13).

The antagonistic TNFR2 antibodies identified here bind to a restricted region of the receptor that confers Treg inhibition even in the presence of TNF. These antagonists are different from the recessive antibody antagonists also identified here that can mildly inhibit Tregs but cannot overcome the agonism of high doses of TNF and IL-2. Together, our findings provide a new way to create TNFR2-targeted growth inhibition whether the receptor is on potent Tregs or is expressed directly on the tumor cell surface. Dominant TNFR2 antagonism might be effective in vivo at cancer sites where abundant TNF and IL-2 are commonly found. The TNFR2 antagonists are more potent against the proliferation of Tregs from T cells isolated from ovarian cancer ascites than those isolated from the blood of normal donors, suggesting that rapidly dividing cells that have abundant and newly expressed surface TNFR2 are more susceptible to this approach. We also provide a novel model of TNFR2 antagonism that is achieved through the stabilization of the receptor inhibitory state represented by a hexagonal lattice of antiparallel dimers. The hexagonal lattice model may be broadly applicable to all TNF superfamily receptors with similar extracellular structures and in settings for required receptor inhibition.


Human subjects

Human blood samples from more than 100 donors were collected according to a human studies protocol approved by the Massachusetts General Hospital (MGH) Human Studies Committee (MGH-2001P001379). All the donors provided written informed consent. Blood was collected into BD Vacutainer EDTA Tubes (BD Diagnostics) and processed within 2 hours of phlebotomy. Ovarian cancer Treg samples from ascites were obtained from women with newly diagnosed ovarian cancer before irradiation and chemotherapy. These human studies were approved by the MGH Human Studies Committee (MGH-2015P002489).

Blood and cell culture

Fresh human blood was processed within 2 hours of venipuncture. Blood was washed twice with 1× Hanks’ balanced salt solution (HBSS) (Invitrogen) plus 2% fetal bovine serum (FBS) (Sigma-Aldrich), and CD4+ cells were isolated using Dynabeads CD4 Positive Isolation Kit (Invitrogen) or Direct Human CD4+ T Cell Isolation Kit (Stemcell Technologies). Isolated CD4+ cells were resuspended in RPMI GlutaMAX (Life Technologies) plus 10% FBS (Sigma-Aldrich) and 1% penicillin-streptomycin (Life Technologies). Because isolated and cultured human T cells die without IL-2 in the media, all culture conditions in all experiments contained a low amount of IL-2 (200 U/ml) to prevent IL-2 withdrawal from influencing the data. Tregs from sterile ovarian cancer ascites were isolated by first concentrating the cells in 50-ml conical tubes and then suspending the cell pellets in ascites supernatant before CD4+ cell isolation. Ovarian adenocarcinoma cell line OVCAR-3 [OVCAR3] (ATCC HTB-161) was cultured in ATCC-formulated RPMI 1640 (Life Technologies) supplemented with 20% non–heat-inactivated FBS (Gibco), 1% penicillin-streptomycin, and human insulin (0.01 mg/ml; Sigma-Aldrich). All cell cultures were incubated at 37°C with 5% CO2.

Treg assays

For all Treg assays, freshly isolated CD4+ cells were seeded in 96-well round-bottom plates at a concentration of 0.2 × 106 to 1 × 106 cells per well, treated with TNFR2 antagonists and various reagents, and incubated for 48 to 72 hours or up to 17 days for expanded Tregs. Medium was renewed every 2 to 3 days for long-term cultures. After incubation, cells were collected and stained for FACS analysis.

Direct cancer-killing assays

OVCAR3 cells were cultured in 96-well flat-bottom plates at a concentration of 0.1 × 106 cells per well in 200 μl of media. Cells were treated directly with TNFR2 antagonistic antibodies and incubated for up to 21 days with half of the medium renewed every 2 to 3 days. After incubation, cells were detached from the plate with 0.25% trypsin-EDTA (Gibco), collected, and stained for FACS analysis or with trypan blue (Sigma-Aldrich) to count viable cells.

Reagents and flow cytometry

mAbs against human TNFR2 were produced internally or obtained from external commercial vendors as previously described for their functional significance for agonism versus antagonism (31). Recombinant human TNF was purchased from Sigma-Aldrich, and recombinant human IL-2 was purchased from Life Technologies. F(ab′)2 fragments of mAbs were prepared using Pierce F(ab′)2 Preparation Kit (Life Technologies). Antibody MAB2261 (R&D Systems) was used for measuring TNFR2 cell surface expression. Cross-linking antibodies against rodent IgG (ab9165 and ab99670) were purchased from Abcam. To inhibit cellular division, CD4 cells were treated with mitomycin C (M4287, Sigma-Aldrich) at 50 μg/ml for 1 hour and washed three times in culture medium before cell culture with the TNFR2 antagonist. Cells were prepared for flow cytometry using Human Treg Flow Kit (BioLegend) according to the manufacturer’s instructions. Fluorescently stained cells were resuspended in 1× HBSS (Invitrogen) and analyzed using a BD FACS Calibur flow cytometer machine (Becton Dickinson). Antibodies used for FACS analysis of Tregs included Alexa Fluor 488 anti-human FOXP3 (clone 259D; BioLegend) for intracellular staining of FOXP3 and phycoerythrin (an antibody directed to FOXP3) and human CD25 (clone BC96; BioLegend) for cell surface staining of CD25. Treg populations were assessed by FACS with FL2 (red) versus FL1 (green) and defined as CD25hi and FoxP3-positive, whereas Teff populations were defined as CD25hi and FoxP3-negative (fig. S2B). For cell suppression assays, responders were stained with CFSE (BioLegend) and CD8–allophycocyanin (APC) (clone SK1; BD Biosciences) and analyzed by FACS with FL4 (far red) versus FL1 (green) (Fig. 2C). For OVCAR3 assays, cells were stained with LIVE/DEAD Fixable Green Stain (Molecular Probes). The proportion of live OVCAR3 cells was assessed by histogram analysis of event count versus FL1 (green) (fig. S6A). FACS data were processed using FlowJo software (version 10.1).

Protein gel electrophoresis

Protein samples were run alongside Precision Plus (Bio-Rad) or Perfect Protein (EMD Millipore) markers on NuPAGE 4 to 12% Bis-Tris gels with MOPS SDS Running Buffer (Life Technologies) at 200 V for 1 hour. Gels were stained for 24 hours with SimplyBlue SafeStain (Invitrogen).

Binding affinity measurement

The affinity of antibody binding to recombinant human TNFR2 was measured by Biacore Analysis Services (Precision Antibody). Briefly, the antibody was biotinylated at 5:1 stoichiometry with biotinyl-LC-LC-NOSE (Thermo-Fisher) in phosphate-buffered saline. Excess biotinylation reagent was removed by centrifugation chromatography, and the biotinylated antibody was captured on 3000 RU of streptavidin surface to a level of 100 RU. Theoretical maximum of signal with TNFR2 with that level of antibody capture was 26 RU, and that signal was reached with a preliminary experiment using 500 nM TNFR2 in the running buffer. Analysis of the kinetics of antigen binding was performed at a flow rate of 60 μl/min with 2-min injections. Antibodies were injected at a concentration of 1 ng/ml to the final capture of 100 RU. The instrument used was Biacore 3000 with the BioCap chip (GE Healthcare). The double reference method was used for analysis. Reference channel contained an identical level of streptavidin.

Signaling analysis

NF-κB activation was measured using the Human Phospho-RelA/NF-κB p65 Cell-Based ELISA Kit (R&D Systems). Briefly, fresh CD4+ cells were cultured in 96-well flat-bottom plates (0.2 × 106 cells per well) in the presence of IL-2 (200 U/ml) alone or in the presence of TNF or TNFR2 antagonists at the indicated concentrations for 10 min at 37°C. Cells were adhered to the plate by centrifugation and fixation and then stained according to the manufacturer’s instructions. Fluorescence was read using the EnVision Multilabel Plate Reader (PerkinElmer), and normalized relative fluorescence units were calculated.

Measurement of secreted TNFR2

Secreted human TNFR2 was measured from cell culture supernatants using Quantikine ELISA (R&D Systems) with some modifications. Briefly, supernatants were collected after 24 to 42 hours of incubation of CD4+ cells with IL-2 (200 U/ml) alone or with TNF (20 ng/ml) or TNFR2 mAbs (12.5 μg/ml) and incubated on either the commercial plates or custom plates coated with 2 μg per well of TNFR2-directed antibodies. ELISA was performed according to the manufacturer’s instructions. Absorbance was measured using the SpectraMax 190 Absorbance Plate Reader and analyzed with SoftMax Pro 6.3 (Molecular Devices).

RNA isolation and gene expression analysis

Isolated CD4+ cells were incubated for 3 hours in the presence of IL-2 (50 U/ml) and TNF (20 ng/ml) or TNFR2 antagonist mAb (2.5 μg/ml). Cells were collected, and total RNA was isolated using RNAqueous-4PCR Kit (Ambion). Total RNA was reverse-transcribed using High Capacity Complementary DNA Reverse Transcription Kit (Applied Biosystems). Real-time PCR was performed using the TaqMan Array Human NF-κB Pathway 96-well Plate with TaqMan Gene Expression Master Mix and the ABI Prism 7000 Sequence Detection System (Applied Biosystems).

Cell suppression assays

For CFSE suppression assays, peripheral blood mononuclear cells (PBMCs) were used as responders. PBMCs were isolated on the day of venipuncture using a Ficoll-Plaque Plus (GE Healthcare) density gradient and cryopreserved at −80°C. Cells were thawed the day before mixing with Tregs and rested overnight in RPMI 1640 and IL-2 (10 U/ml). The next day, responder cells were stained with 1 μM CFSE. Responder cells (5 × 104 cells) were then mixed at various ratios (0:1, 4:1, 2:1, and 1:1) with Tregs that had been expanded for 14 to 17 days, and the mixtures stimulated with either anti-CD3 mAb, an antibody directed to human CD3 protein (HIT3a, BD Biosciences), or Dynabeads Human T-Activator CD3/CD28 (Gibco) at a ratio of 2:1 (cells/beads) and IL-2 (50 to 200 U/ml). Cells were collected after 4 to 6 days and stained with CD8-APC (clone SK1; BD Biosciences), and suppression of cell division was assessed by FACS analysis of CD8 counts versus CFSE (Fig. 2C).

Epitope mapping

ELISA was used for linear epitope mapping of TNFR2 antagonists on the TNFR2 external membrane protein sequence. Peptides were purchased from GenScript, diluted in coating buffer (REF), and placed on Immulon 4HBX Flat Bottom Microtiter Plates (Thermo Scientific) at a concentration of 1 μg per well. Primary TNFR2 mAbs (0.1 μg per well) were incubated with substrates. Secondary antibodies against rodent IgGs were used to label the primary mAbs. Absorbance was measured using the SpectraMax 190 Absorbance Plate Reader and analyzed with SoftMax Pro 6.3 (Molecular Devices).

3D peptide mapping was performed using Chemically Linked Peptides on Scaffolds (CLIPS) technology (Pepscan). Briefly, the target protein was converted into a library of up to 10,000 overlapping linear peptides, which were then bound to a solid support and shaped into a matrix of CLIPS constructs. The affinity of the antibody to the various peptide confirmations was used to determine the precise discontinuous epitopes.

Statistical analysis

Data analysis was performed by Student’s t test (unpaired, type 3) using Excel (Microsoft) or GraphPad Prism 5 software (GraphPad Software). Significance was determined by P < 0.05.


Fig. S1. TNFR2 antagonists inhibit Treg proliferation in the presence of IL-2 and TNF.

Fig. S2. TNFR2 antagonists inhibit Treg expansion by TNF.

Fig. S3. Functional activity of TNFR2 antagonists is independent of Fc region or receptor cross-linking.

Fig. S4. Inhibition of cell division limits the efficacy of TNFR2 antagonist inhibition of Tregs.

Fig. S5. Representative FACS dot plot of OVCAR3 live/dead staining analysis and TNFR2 oncogene abundance.

Fig. S6. Treatment of CD4+ cells with TNFR2 antagonistic antibodies reduces NF-κB activation.


Acknowledgments: We thank the Cutaneous Biology Research Center at MGH for use of the fluorescence plate reader. We thank M. Davis of our department for her editing of the manuscript. Funding: Y. Okubo was supported in part by a travel grant from the Manpei Suzuki Diabetes Foundation. H. Torrey was supported in part by a fellowship from the American Autoimmune Related Disease Association. The Advanced Medical Research Foundation supported the collection of ascites from ovarian cancer subjects. Author contributions: H.T. and J.B. performed the 48- to 72-hour Treg assays and the ELISAs. H.T. prepared the data, formatted the paper, and performed the assays related to cancer cells. T.M. performed the transcription assays and created the mAbs. Y.O. performed the CFSE assays. L.W. performed the peptide mapping studies to identify the antibody binding sites. D.B., S.W., S.P., and D.H. performed subject recruitment, obtained human studies approval, consented subjects, and obtained the blood for these studies. D.B. also performed the Treg assay on normal donors. A.D. performed the OVCAR assays. E.V. and L.W. used the newly identified selective amino acid binding regions to map the binding sites onto the 3D dimer and trimer structures of TNFR2. R.F. identified, consented, and obtained the ovarian ascites fluid from ovarian cancer patients. D.L.F. was the principal investigator on this study and thus guided the study, wrote the paper, and provided oversight of the data. Competing interests: All work in this paper was supported by nonprofit funding sources, and no author accepted consulting money for performance of the studies. The authors declare that they have no competing interests. Data and materials availability: The antibodies require a material transfer agreement from MGH, USA.
View Abstract

Navigate This Article