Nivolumab and Urelumab Enhance Antitumor Activity of Human T Lymphocytes Engrafted in Rag2^−/−IL2Rγ^null Immunodeficient Mice

Activation and expression of the targets for immunostimulatory mAbs on human T lymphocytes engrafted in immunodeficient mice

Rag2^−/−IL2Rγ^null mice are devoid of T, B, and NK lymphocytes. Intraperitoneal injection of human PBMC gives rise to the engraftment of viable CD4⁺ and CD8⁺ T lymphocytes that can be sequentially retrieved by peritoneal lavage over 4 weeks following adoptive transfer. It is well known that human T lymphocytes develop pathogenic xeno-reactivity when they are transferred into mice and become activated (27). hCD3⁺-gated T cells such as hCD4⁺ and hCD8⁺ T lymphocytes underwent blastic activation, which peaked on days 5 to 9 (Fig. 1A and Supplementary Fig. S1A), as shown by enlargement revealed by the forward scatter and side scatter parameters. Such T lymphocytes underwent proliferation as observed by intracellular immunostaining with Ki-67 on gated hCD4⁺ and hCD8⁺ T cells from the sequential peritoneal lavages (Fig. 1B and Supplementary Fig. S1B). The activation/exhaustion markers TIM-3 and LAG-3 have also been studied showing an initial activation during first 5 days and a second peak of coexpression with PD-1 that might reflect T lymphocyte exhaustion after 4 weeks (Supplementary Fig. S1C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Human T cells acquire an activated phenotype and express surface CD137 and PD-1 after being transferred into Rag2^−/−IL2Rγ^null mice. Rag2^−/−IL2Rγ^null mice were injected with 1 × 10⁷ human PBMCs i.p. on day 0. Peritoneal lavages were performed in two mice/day at days 1, 3, 5, 7, 9, 13, 22, and 30 and the cells recovered were analyzed by flow cytometry. A, percentage of hCD3⁺ blasts lymphocytes recovered from the peritoneal lavage at the indicated time points. B and C, level of Ki-67 intranuclear staining (B) and hCD137 (top) and hPD-1 (bottom) surface staining represented by mean fluorescence intensity (MFI) on gated hCD3⁺hCD4⁺ and hCD3⁺hCD8⁺ T cells before transfer and on day 5, 9, and 22 after adoptive transfer (C). Shadowed histograms respresent isotype-matched control stainings.

CD137 and PD-1 are respectively important receptors for costimulation and coinhibition of T cells whose expression on the plasma membrane is known to be induced by antigen stimulation. Figure 1C shows that both receptors were induced on the adoptively transferred human lymphocytes over time. hCD137 was induced mainly on hCD8⁺ T lymphocytes, while hPD-1 was induced on both T-cell subsets with very bright intensity. PD-1 ligand 1 (PD-L1/B7-H1) was also induced in T cells, with a peak on day 5 (Supplementary Fig. S1D).

To analyze whether anti-hCD137 (urelumab) and anti-hPD-1 (nivolumab) mAbs were binding to their corresponding receptors in vivo, flow cytometry of transferred lymphocytes was performed ex vivo. As shown in Supplementary Fig. S2, the fluorescence intensity of CD137 and PD-1 staining was decreased on human T cells recovered from anti-hCD137 and anti-hPD-1–treated mice, but not on T cells from untreated mice.

These data confirm that transferred human T lymphocytes in Rag2^−/−IL2Rγ^null become activated during the first 7 days and consequently express on the cell-surface hCD137 and hPD-1, thus becoming susceptible targets to be manipulated by immunomodulatory mAbs.

Adoptive transfer of human T lymphocytes causes a xenograft versus host disease that is exacerbated by urelumab and nivolumab and abrogated by CD4⁺ depletion

xGVHD is considered a valuable model to test immunomodulatory strategies, where the engrafted human T lymphocytes are amenable for regulation by therapeutic agents (29, 31). After 3 to 4 weeks we observed that transferred T lymphocytes started to infiltrate the spleen, liver, and lung with signs of serious tissue damage after 4 to 5 weeks (Supplementary Fig. S3). Accordingly, we observed that mice injected with 10⁷ human PBMC (Fig. 2A) started losing weight at 3 to 4 weeks after adoptive transfer and in most instances succumbed 1 week later showing signs of respiratory distress, hunched posture, and/or fur loss.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Administration of anti-hPD-1 mAb (nivolumab) as a single agent or in combination with anti-hCD137 (urelumab) exacerbates xGVHD mediated by human PBMCs in Rag2^−/−IL2Rγ^null mice. A, Rag2^−/−IL2Rγ^null mice were injected with 10⁷ human PBMCs i.p. on day 0. Treatments with anti-hCD137 (urelumab), anti-hPD-1 (nivolumab), combination (Combo; urelumab+nivolumab), or saline control i.v. were injected on days 4, 7, and 10. Survival was monitored up to 100 days. B, the Kaplan–Meier plot depicts overall survival among the four groups of treatment urelumab (n = 22), nivolumab (n = 23), combination (Combo; n = 13), and control (n = 19). Experiments were independently repeated three times and results pooled. *, P < 0.5; ***, P < 0.001. C, microphotographs of H&E and hCD3 IHC stainings in representative tissue sections of spleen, liver, and lung in one representative mouse from each group of treatment. Bar, 400 μm.

Having confirmed the expression of hCD137 and hPD-1 in human T lymphocytes infiltrating the liver and spleen (Supplementary Fig. S4), we tested whether the administration of immunostimulatory mAbs exacerbated xGVHD. We injected mice with 10⁷ human PBMC and treated them with the clinical-grade mAbs urelumab (anti-hCD137), nivolumab (anti-hPD-1), a combination of the two (Combo) or saline (control) on days 4, 7, and 10 after lymphocyte transfer. Mice were followed for lethal xGVHD and a significant decrease in overall survival was observed in the nivolumab (P < 0.001) and Combo (P < 0.05) groups as compared with the control group, while urelumab showed a tendency that did not reach statistical significance (Fig. 2B). Details of the H&E and CD3⁺ spleen, liver, and lung infiltration in a representative mouse of each group of treatment are depicted in Fig. 2C, showing marked perivascular T lymphocyte infiltrations in these organs, more intense when under treatment with the mAbs. To understand the role of specific immune cell subsets, we repeated the xGVHD model injecting hPBMC that were selectively predepleted of specific leukocyte subpopulations by immunomagnetic sorting. These included selective depletions of CD4⁺, Treg, CD8⁺, CD56⁺, and CD14⁺ populations checked to be complete by flow cytometry (data not shown). Transferred mice were treated with Combo (nivolumab and urelumab mAbs). The control group was injected with hPBMC and treated with hIgG4. Notably, mice injected with hPBMC depleted from CD4⁺ T cells did not develop xGVHD and survival, extended over 80 days, was significantly higher than mice injected with hPBMC and treated with Combo (P < 0.01) or hIgG4 (P < 0.01; Supplementary Fig. S5A). Depletion of other immune cells did not significantly affect the survival of mice and only Treg depletion showed a trend to increase the severity of xGVHD, in comparison with mice injected with hPBMC and treated with Combo (Supplementary Fig. S5B). Notably, human-IFNγ plasma levels were almost undetectable in mice transferred without CD4⁺ T cells (Supplementary Fig. S5C).

As a whole, these data indicate that the administration of the mAbs was active, resulting in enhanced functional stimulation of transferred T cells that migrate to visceral organs and exacerbate a xGVHD that is primarily driven by IFNγ-producing CD4⁺ T lymphocytes.

Urelumab and nivolumab enhance the functional activation of adoptively transferred human T lymphocytes

A hallmark function of activated T lymphocytes is the production of cytokines. Therefore, we studied the concentrations of human IFNγ (hIFNγ) in sequential plasma samples of mice undergoing treatment with urelumab, nivolumab, the combination (Combo) or the control group. Following the second dose, urelumab-treated mice showed an increase of hIFNγ plasma levels in comparison with the control group (P < 0.05; Fig. 3A). Nivolumab as a single agent did not significantly increase hIFNγ plasma levels, but in combination with urelumab there was a further increase of plasma over control or nivolumab single treatment (P < 0.01). Production of this cytokine might depend on lymphocyte numbers producing it. Indeed, we observed that mice treated with urelumab or Combo, showed higher densities of blastic hCD3⁺ and hCD8⁺ lymphocytes in the liver infiltrates (Fig. 3B).

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Evidence for enhanced xenoreactive human lymphocyte activity upon treatment with anti-hCD137 (urelumab), anti-hPD-1 (nivolumab), and their combination. Experiments performed as in Fig. 2A. A, human IFNγ plasma levels of mice were studied before the first mAb dose and after the first (day 6) and second (day 9) doses of antibodies. B and C, mice were sacrificed at week 3 after human PBMCs transfer and single-cell suspensions were prepared from the lung, liver, spleen, and peritoneal lavage to be studied by flow cytometry. B, absolute numbers of hCD3⁺ (right) and hCD8⁺ (left) per gram of tissue in the liver. C, perforin mean fluorescence intensity (MFI) of hCD3⁺hCD8⁺ (left) and hCD3⁺hCD4⁺ (right) in the lung lymphocyte infiltrates. Data were analyzed by the Mann–Whitney U test. *, P < 0.05; **, P < 0.01. Error bars represent SEM.

It is also possible that antibody treatment may result in functional changes on a per-cell basis. In this regard, we observed that in animals treated with nivolumab lung-infiltrating hCD4⁺ and hCD8⁺ T lymphocytes showed higher intensity of expression of the intracellular cytotoxic effector molecule perforin in comparison with control group (Fig. 3C).

Xenografted human colorectal carcinoma is controlled by allogeneic human PBMC under stimulation with urelumab and/or nivolumab

The experimental setting described in Rag2^−/−IL2Rγ^null mice permits both transfer of human PBMC and the grafting of subcutaneous tumors derived from transplantable human cell lines. We used the HT29 human colorectal carcinoma cell line that gives rise to rapidly progressing subcutaneous tumors in these mice (36). To explore if under these conditions anti-hCD137, anti-hPD-1, or their combination could interfere with tumor progression, we treated these mice with urelumab, nivolumab, Combo (urelumab+nivolumab), or isotype control (irrelevant hIgG4) on days 6, 14, and 20 as depicted in Fig. 4A.

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Immunostimulatory mAbs anti-hCD137 (urelumab) and anti-hPD-1 (nivolumab) alone or in combination show antitumor activity against a xenografted human colon cancer mediated by transferred allogeneic human PBMC. A, Rag2^−/−IL2Rγ^null mice were injected with 7 × 10⁶ human PBMCs i.p. and challenged with 3.5 × 10⁶ HT-29 cells s.c. in the right flank on day 0 of the experiment. Thereafter, the mice were treated with anti-hCD137 (urelumab), anti-hPD-1 (nivolumab), combination (urelumab+nivolumab), or isotype control (hIgG4) i.v. on days 6, 14, and 20. Tumor volumes were measured twice per week until mice were sacrificed on day 22 of experiment. Tumors were removed after sacrifice and studied by flow cytometry and IHC. B, results depict mean ± SEM of the progression of subcutaneous tumor volumes (n = 5 mice per group). Arrows, time of treatment administration. Experiments were repeated four times, rendering similar results. C, human IFNγ plasma levels were studied previous to the first mAb dose, after the first and second antibody dose. D, tumor TILs from each treatment group were studied by flow cytometry and results of number of human CD8 T lymphocytes per mg of tumor (left), number of human regulator T cells per gram of tumor (center graph), and ratio of number of human CD8 T cells per number of human Treg (right) are depicted. *, P < 0.05; **, P < 0.01. Error bars represent SEM. T, tumor; P, hPBMCs.

We individually followed tumor growth over time. In a series of three experiments, it became clear that control mice without hPBMC experienced a faster tumor growth than those mice with concomitant adoptive transferred of hPBMC. Under these conditions, treatment with the immunostimulatory mAbs led to a clear stabilization of the experimental tumors in most instances that was of comparable efficacy to each antibody administered separately or in combination (P < 0.001; Fig. 4B). Indeed, the plasma of the mice under treatment showed significantly higher concentrations of hIFNγ after the second antibody dose as compared with the control group (P < 0.05; Fig. 4C). After the first dose a significant increase (P < 0.05) in TNFα plasma levels was also observed in the Combo-treated group in contrast with the control (irrelevant hIgG4) group (data not shown).

As a control, mice bearing HT29 human tumors but without hPBMC transfer did not show any improvement upon Combo treatment when compared with the hIgG4-treated control group (Supplementary Fig. S6).

To evaluate the role of the different immune cells controlling tumor growth, we tested the combination of anti-mCD137 (clone 1D8) and anti-mPD1 (clone RMP1-14) mAbs in a murine syngeneic model bearing murine colon cancer (MC38). We found that depletion of CD8⁺ cells significantly abrogate the antitumor effect of the combinatorial treatment, while CD4, NK, or macrophage depletions did not affect the antitumor effect with the combinatorial treatment (Supplementary Fig. S7).

Studying cell suspensions of tumor-infiltrating lymphocytes

To address the mechanism of action of urelumab and nivolumab either alone or in combination, the human CD4⁺ and CD8⁺ T lymphocyte density was analyzed in tumors excised on day 22 from the experimental groups of mice, as depicted in Fig. 4A. The Combo treatment led to more dense infiltrates of hCD8⁺ T lymphocytes in comparison with control group (P < 0.001) and also in comparison with each antibody given alone (urelumab, P < 0.01; nivolumab, P < 0.05; Fig. 4D, left).

The role of regulatory T lymphocytes is known to be important in dampening antitumor immunity (37). We observed that treatment with urelumab alone or in combination with nivolumab showed a significant improvement in the intratumoral hCD8⁺ T cells/hCD4⁺FOXP3⁺ ratio when compared with the control group (P < 0.01; Fig. 4D, right). In the group in which urelumab was administered as a single agent, it seems that this improvement is dependent of a significant decrease in Treg number per gram of tumor tissue (Fig. 4D, center), while in the Combo group it is more dependent on an increase in total CD8 T cells over Tregs (Fig. 4D, left). Of note, tumor-infiltrating lymphocytes (TIL) under treatment with urelumab or Combo showed a tendency to express higher protein levels of the transcription factor EOMES in hCD4⁺ and hCD8⁺ T cells on a per cell basis, when analyzed by intracellular staining (data not shown).

Pathologic examination of tumors excised on day 22 (Fig. 5) showed denser infiltrates of human T cells upon treatment with the immunostimulatory mAbs that progressed inwards from the peripheral rim of the tumors and sparsely reached to the core of the round tumor lesions. Indeed, both the densities of human hCD4⁺ and hCD8⁺ T cells were increased. In the case of mice treated by the Combo treatment, the infiltrate was denser than in the single-agent treatment tissue samples, and human T cells infiltrated more clearly the central part of the tumor lesions (Fig. 5).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Increased density of TILs in anti-hCD137 and/or anti-hPD-1 treatment groups. Rag2^−/−IL2Rγc^null mice were treated as described in Fig. 4. On day 22, mice were sacrificed and tumor studied by IHC. Representative histology and IHC images of lymphocyte tumor infiltrate at the invasive margin highlighted by hCD3, hCD4, and hCD8 immunostaining. Microphotographs of tumors subjected to the indicated treatments are presented with two different magnifications (smaller square, ×50; larger square, ×200). Bar, 400 μm.

PD-L1/B7-h1, hPD-1, and hCD137 are present in the xenografted tumors

PD-L1/B7-H1 is known to be induced on malignant cells by inflammatory lymphokines such as IFNγ. Hence, it is considered an adaptive mechanism of resistance to immune attack (38). Indeed, HT-29 cells cultured in the presence of hIFNγ readily upregulate surface PD-L1/B7-H1 (data not shown). Analyzing the xenografted tumors under treatment using multiplexed immunofluorescence, which highlight the mAb targets as well as the tumor cells, it became clear that carcinoma cells were indeed expressing PD-L1/B7-H1 (Fig. 6 and Supplementary Fig. S8). PD-1 signal was allocated predominantly in lymphoid cells at the periphery of the neoplasms and PD-L1 staining was higher toward the tumor center/core. Both markers showed predominant membranous staining pattern (Supplementary Fig. S8). The proportion of cytokeratin-positive tumor cells was prominently lower and the PD-L1 signal, in these cells, was higher in the Combo treatment than in the control condition (Fig. 6 and Supplementary Fig. S9).

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

hCD137, hPD-1, and PD-L1/B7-H1 are expressed in the xenografted tumors. Representative multiplexed immunofluorescence microphotographs showing the protein expression of hCD137 (red fluorescence channel, left), hPD-1 (red channel, middle), and PD-L1 (red channel, right) in the tumor tissues (cytokeratin positive, green channel) from mice subjected to treatment with hIgG4 (control), anti-hCD137 (urelumab), anti-hPD-1 (nivolumab), or the combination (Combo). hCD137 and hPD-1 positivity was distributed predominantly toward the periphery of the tumor surrounding the cytokeratin-positive tumor areas. In contrast, PD-L1 was allocated predominantly toward the tumor center/core, in close association with the tumor nests. Nuclei were stained with DAPI (blue fruorescence channel). Bar, 100 μm.

It is of note that the TILs in those tumors under treatment showed as well membrane expression of hCD137 (Fig. 6). Taken together, these results demonstrate that the hPD-1–PD-L1 pathway and hCD137 are available targets in the tumor infiltrate to be modulated by the exogenously provided mAbs. In addition, the selective expression of PD-L1/B7-H1 toward the center of the tumor, right beneath a peripheral rim enriched in hCD3⁺/hCD137⁺/hPD-1⁺, suggests that these T lymphocytes are in an activated state but may not be able to gain access to the PD-L1 enriched tumor core. Thus, such T cells are conceivably amenable to be both unrepressed and costimulated to further increase tumor cell destruction.

Therapeutic control of a xenografted gastric cancer by syngenic PBMC upon treatment with urelumab and nivolumab

To explore these mechanisms in a tumor-lymphocyte autologous setting, we xenografted tumor pieces (7 mm × 7 mm) derived from a surgical specimen of a gastric carcinoma patient in Rag2^−/−IL2Rγ^null mice and preinfused these mice with 7 × 10⁶ PBMC from the same patient. Mice were treated intravenously with urelumab, nivolumab, Combo (urelumab+nivolumab), or isotype control (hIgG4) on days 5, 12, and 19 of the experiment. As shown in Fig. 7A, repeated antibody treatment resulted in a significantly slower progression in tumor size in all the treatment groups in contrast with the control group (urelumab P < 0.001; nivolumab P < 0.05; Combo P < 0.01). Analysis of hIFNγ levels in plasma showed a progressive increase during treatment with a tendency to increased levels in the urelumab monotherapy and Combo groups in contrast with control group and nivolumab monotherapy (Fig. 7B). We confirmed the same result in a second experiment (Supplementary Fig. S10A and 10B). Analysis of the excised tumors using multiplexed and compartment-specific QIF for human TIL subpopulations indicated the presence of human lymphocytes in the tumor microenvironment (Fig. 7C) and an increase in hCD3⁺ cells in Combo-treated group in contrast to control group that reached statistical significance only within the tumor compartment (P < 0.05; Fig. 7D, left) but not in the stromal compartment (Fig. 7D, right), suggesting more penetration of TILs into the tumor bed. No significant change in the human TIL subpopulations was noted with the other treatment modalities (Fig. 7D). We could not detect CD20⁺ human B cells in any of the samples analyzed. It is of note, that in tumors from mice treated with combination therapy showed extensive necrosis, a phenomenon not seen in the other treatment groups (data not shown). Moreover, in the urelumab and Combo-treated groups the number of Treg per gram of tumor decreased (Fig. 7E, right) while the CD8⁺:Treg cells ratio was increased (Fig. 7D, left). In this line, a study at day 22 of human lymphocytes in the peripheral blood from treated mice showed a higher percentage of hCD3⁺ lymphocytes in the Combo group when compared with the isotype-matched antibody control (hIgG4) group (Supplementary Fig. S10C). Finally, we measured PD-L1/B7-H1 expression using QIF and found positive immunoreaction in both the tumor and stromal compartments (Supplementary Fig. S10D), with a tendency to express higher levels of PD-L1 in tumor than in stromal compartment, particularly in groups treated with anti-hPD1 alone or in combination with anti-hCD137, a finding compatible with a higher penetration of the TILs in the tumor bed as can be seen in Fig. 7C (Combo group).

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

Anti-hCD137 (urelumab) and anti-hPD-1 (nivolumab) treatment alone or in combination show antitumor effects against a xenografted human gastric cancer transferred with autologous lymphocytes of the patient. Rag2^−/−IL2Rγ^null mice were injected with 7 × 10⁶ human PBMCs from a gastric cancer patient on day 0 of the experiment. Explants (7 mm × 7 mm) from the tumor of the same patient were implanted s.c. into the flank of Rag2^−/−IL2Rγ^null mice on day 3 of the experiment. Mice (n = 6 per group) were treated i.v with anti-hCD137 (urelumab), anti-hPD-1 (nivolumab), combination (nivolumab + urelumab), or isotype control (human IgG4). Tumor volumes were measured twice per week. A, results depict mean ± SEM of tumor growth curves. Arrows, time of treatment administration. T, tumor; P, hPBMCs. B, human IFNγ plasma levels were studied previous to the first mAb dose, after the first and third dose. C, representative multiplexed immunofluorescence microphotographs showing the architecture of tumor xenografts (cytokeratin-positive, green channel) and the presence of TIL subsets (hCD3, red; hCD8, yellow; hCD20 white) from mice treated with hIgG4 (control), anti-hCD137 (urelumab), anti-hPD-1 (nivolumab), or the combination (Combo). Nuclei were stained with DAPI (blue fluorescence). Insets, image segmentation defining tumor (yellow) and stromal compartments (red) as well as the automated cell phenotyping based on multispectral fluorescence analysis. Bar, 100 μm. D, hCD3 T cells/total number cells ratio in tumor (left) and the stroma compartment (right). E, tumor TILs from each treatment group were studied by flow cytometry and results of number of human Tregs per gram of tumor (left) and ratio of human CD8 T cells and human CD4⁺ CD25⁺ FOXP-3⁺ Tregs (right) are depicted. *, P < 0.05; **, P < 0.01; ***, P < 0.001.

These results confirm the antitumor activity of the mAbs observed in previous experiments and provide indications of a more pronounced effect with the combination treatment. In addition, the findings open the possibility of studying the mechanisms of action and pharmacodynamics of immunostimulatory antibodies against human targets in an autologous tumor-lymphocyte experimental setting.