Introduction

The T cell response to colon and pancreatic cancers is associated with improved clinical outcome [1, 2]. Methods for directing T cells against these cancers have included both accines to activate tumor antigen-specific T cell responses [3] and more complex adoptive immunotherapy strategies with ex vivo expanded T cells [4, 5]. Challenges for these therapeutic modalities have included low rates of antigen-specific T cell responses induced by the current generation of tumor vaccines and the complex logistics of in vitro cellular processing and expansion techniques resulting in variable yields of cellular products [6]. One potential strategy for overcoming both these challenges is through in vivo recruitment of T cells using bispecific single-chain antibodies, such as BiTE antibodies, that simultaneously bind CD3+ T cells via one idiotype, and bind tumors via tumor antigen recognized by the second idiotype. For example, blinatumomab, a bispecific CD19/CD3 antibody, has demonstrated activity in clinical trials for CD19-expressing refractory non-Hodgkin’s lymphoma (NHL) [7] and B cell acute lymphoblastic leukemia (ALL) [8].

CD19 targeting would not likely impact colon and pancreatic cancers, as they do not express high levels of CD19. An attractive alternative is to target T cells to recognize antigens known to be expressed at high levels by colon and pancreatic cancer. One such antigen is carcinoembryonic antigen ((CEA, CD66e, and CEACAM5) [9]. CEA BiTE (CEA/CD3-bispecific T cell-engaging BiTE; (MEDI-565; MT111; AMG 211) that is composed of a humanized single-chain antibody recognizing CEA expressed by colorectal, pancreas, and other epithelial malignancies, and a de-immunized single-chain antibody specific for human CD3ε [10]. In preclinical models, a murine BiTE targeting CEA controlled metastatic tumor growth [11]. We had previously reported that CEA BiTE recognized and lysed in vitro metastatic colorectal cancer cell explants that had been isolated from patients with chemotherapy-refractory disease [12]. Nonetheless, some tumor cells were not eradicated despite 5-day exposure to T cells and CEA BiTE in vitro [12]. This observation could be due to ligation of non-primed T cells or more probably, downregulated T cell responses promulgated by tumors using a number of strategies including antigen escape via tumor antigen downregulation [13], secretion of immunoinhibitory cytokines [14], and activation of regulatory T cells (Tregs) [15] or myeloid-derived suppressor cells (MDSCs) [16].

The role of the immune system, in particular T cell-mediated cytotoxicity, in tumor control has been well established. More recently, the immune checkpoint (CTLA-4, PD-1, PD-L1) signaling pathways that temper T cell responses as a means to control autoimmunity [17] have been observed to be coopted by tumors to escape immune surveillance. In particular, expression of programmed death-ligand 1 (PD-L1) by tumors has been shown to modulate the function of activated T cells expressing programmed cell death 1 (PD-1) [18], resulting in T cell “exhaustion” and facilitating an immunosuppressive environment for tumor growth and progression [19]. Impressive clinical responses to therapies that disrupt the PD-1/PD-L1 axis [20] have been observed in melanoma, renal cell carcinoma, and non-small cell lung cancers with elevated PD-L1 expression.

In the current study, we explored whether diminished T cell cytolysis was observed with repeated rounds of CEA BiTE-mediated killing exposure and if so, whether one mechanism was the PD-1/PD-L1 interaction. We also determined whether this effect on T cell function occurred prior to T cell killing of tumor or after T cells had participated in CEA BiTE-mediated killing. Furthermore, we assessed the effect of early PD-1/PD-L1 blockade on T cell cytotoxicity. Finally, we attempted to restore cytolytic activity of T cells that had previously participated in CEA BiTE-mediated attack with anti-PD-1 and anti-PD-L1 antibodies.

Materials and methods

Reagents

PerCP-conjugated anti-CD4 mAb, APC-conjugated anti-CD8 mAb, PE-conjugated anti-CD69, anti-CD25, anti-CD28, anti-CTLA4, anti-PD-1, anti-PD-L1 mAbs, and streptavidin-APC were purchased from BD Biosciences (San Jose, CA). FITC-labeled and PE-labeled anti-CEA mAbs from Sanquin (Amsterdam, Netherlands), and 7-AAD and annexin V-biotin kit were purchased from Immunotech (Marseille, France, cat# PN IM3422). For the analysis of CEACAM5 expression, anti-CEACAM5 mAb (clone 26.5.1, GENOVAC, Germany) was used. Anti-PD-1 (clone J116), anti-PD-L1 (clone MIH1), and mouse IgG1κ isotype control (P3.6.2.8.1) were purchased from EBioscience (San Diego, CA).

CEA BiTE and the control MEC14 BiTE (Cont BiTE) were constructed as described elsewhere [10]. CEA BiTE is composed of a humanized anti-CEA single-chain antibody [17], and a ‘de-immunized’ human CD3e-specific single-chain antibody derived from the mouse monoclonal antibody L2K. MEC14 BiTE, kindly provided by Amgen (Thousand Oaks, CA) [21], is composed of a murine anti-Mecoprop (a herbicide) single-chain antibody linked to the same anti-CD3e single-chain antibody used to construct CEA BiTE.

Tumor cell lines

CEA-expressing colon carcinoma cell lines, SW1463 (ATCC CCL-234), Colo205 (ATCC CCL-222), HT29 (ATCC HTB-38), and a CEA-positive pancreatic adenocarcinoma cell line AsPC-1 (ATCC CRL-1682) were purchased from ATCC (Manassas, VA). The cell lines were cultured in RPMI1640 medium supplemented with 10 % fetal bovine serum (Invitrogen Life Technologies, Carlsbad, CA). Colorectal cancer cells (CRC057 and CRC096) were isolated from metastatic lesions of cancer patients and propagated in NOD.CB17-Prkdc scid/J (NOD/SCID) mice as described elsewhere [12].

Flow cytometry-based cytotoxicity assay

T cells were negatively isolated from normal donor PBMCs using a T cell isolation kit (Invitrogen Dynal AS, Oslo, Norway, cat#113.11D). In all experiments, purity of CD3-positive cells exceeded 95 % of the CD45-positive leukocyte population after isolation procedures. For the cytotoxicity assays, 1 × 106 tumor cells and 5 × 106 T cells were added to T75 flasks with CEA BiTE or Cont BiTE (100 ng/mL). After 5-day incubation, all cells were harvested with 0.05 % trypsin/EDTA and spun down by centrifugation. Cells were then stained with anti-CD3-FITC or anti-CD45-FITC, 7-AAD, and annexin V-APC, and CD3 negative or CD45 negative tumor cells were analyzed for expression of annexin V as a marker of apoptosis using a FACSCalibur™ or LSR II flow cytometer (BD Biosciences).

Selection of tumor cells escaping CEA BiTE-mediated T cell killing

In our previous published experience with CEA BiTE, we found an effector/target ratio of 5:1 and 5-day incubation was optimal for the study of CEA BiTE-mediated killing of tumor cells [12]. Therefore, in the current study, tumor cells (5 × 106 cells) were incubated with negatively isolated T cells (25 × 106 cells) in T75 flasks in the presence of CEA BiTE (100 ng/mL). After a 5-day incubation (1st round), floating cells were discarded and tumor cells adherent to the flask were harvested by trypsin/EDTA treatment. Cytotoxicity was analyzed by flow cytometry based on annexin V/7-AAD staining as described above. Tumor cells were centrifuged with Ficoll-Paque, and viable tumor cells were isolated and transferred into the flask (5 × 106 cells/T-75 flask) together with T cells isolated from the same donors’ PBMCs of the first round co-incubation. After this second-round co-incubation, tumor cell viability was analyzed by flow cytometry again. To monitor CEA expression level of tumor cells, cells were labeled with anti-CEACAM5 mAb (clone 26.5.3), incubated for 30 min at 4 °C, and then washed and stained with PE-labeled goat anti-mouse IgG antibody for 30 min.

Cytotoxic function of T cells after longer-term co-incubation with tumor cells in the presence of CEA BiTE

Negatively isolated T cells from normal donor PBMCs were incubated with tumor cells (SW1463, AsPC-1) at a 5:1 ratio in the presence of CEA BiTE or Cont BiTE (100 ng/mL) for 5 days. Floating cells were collected, and density gradient centrifugation with Ficoll-Paque was performed to obtain viable T cells only. As controls, fresh T cells isolated from the same normal donors, and in some experiments, 5-day incubated T cells without tumor cells/BiTE antibody, were used. Soon after isolation of viable T cells, co-incubation of these T cells and tumor cells at a 5:1 ratio was initiated in the presence of CEA BiTE or Cont BiTE (100 ng/mL). After a 5-day incubation, tumor cells were harvested and the percentage of annexin V-positive tumor cells was measured by a flow cytometry-based cytotoxicity assay.

Expression of co-stimulatory molecules on T cells and tumor cells following CEA BiTE-T cell-mediated killing

SW1463 cells were co-incubated with T cells derived from a normal donor at a 1:5 ratio, in the presence of CEA BiTE or Cont BiTE (100 ng/mL) for 1, 3, 5, or 7 days. Controls consisted of T cells or tumor cells incubated alone in the presence of CEA BiTE or Cont BiTE. Cells were harvested with trypsin/EDTA, stained with anti-CD45-APC and PE-conjugated antibodies for CD28, CTLA4, PD-1, or PD-L1. PE-conjugated mouse IgG was used for negative control staining. Cells were acquired via FACSCalibur and analyzed with CellQuest software. Tumor cells were identified as CD45- and T cells as CD45+.

Restoration of cytolytic T cell activity with anti-PD-1 and anti-PD-L1

T cells isolated from PBMCs of a normal donor were incubated with SW1463 cells in the presence of CEA BiTE or Cont BiTE (100 ng/mL) for 7 days after which the viable T cells were isolated by density gradient centrifugation with Ficoll-Paque, and used as effector cells for a second round of co-incubation with SW1463 cells in the presence of either CEA BiTE or Cont BiTE. To assess the role of PD-1 and PD-L1 in decreased T cell cytolytic killing of tumor, anti-PD-1 blocking mAb, anti-PD-L1 blocking mAb, or a combination of both antibodies were added to the culture (final concentration: 5 µg/mL). Controls consisted of isotypic IgG (final concentration: 5 µg/mL) added to parallel cultures. After 5-day incubation, the flow cytometry-based cytotoxicity assay was performed as described above.

Results

A subset of CEA-expressing tumors resists killing despite repeated exposure to T cells and CEA BiTE in vitro

We first confirmed CEA-expressing tumor cells survived despite repeated exposure to T cells and CEA BiTE in vitro. We tested the viability of the CEA-expressing colorectal cell line SW1463 following 2 cycles of co-culture with CEA BiTE and negatively isolated T cells from normal donor PBMCs. A new isolate of T cells from the same donor was used for the second cycle of killing in order to study the survival of the tumor cells in the absence of their effects on the T cells. In a representative experiment, we observed that cytotoxicity was 47.8 % (vs. 2.9 % for tumor alone) after one round of attack. Viable tumor cells were isolated and re-exposed to CEA BiTE and T cells isolated from a second aliquot of the same donor PBMCs. After this second round of co-culture with fresh T cells and CEA BiTE, 40.6 % (compared with 14.9 % in the absence of CEA BiTE) of the SW1463 tumor cells were killed (Fig. 1a). A similar trend was seen with AsPC-1 (Fig. 1b). Tumor cells that survived the first round of CEA BiTE-mediated T cell killing showed similar levels of survival following the second round of T cell/CEA BiTE exposure. These data suggest that tumor cells not killed in the first exposure remained susceptible to immune attack.

Fig. 1
figure 1

Surviving tumor cells in first round of CEA BiTE-mediated T cell killing did not acquire resistance for T cell killing and had enhanced CEA expression. Tumor cells were incubated with T cells and CEA BiTE as described in the “Materials and methods” section. After 5-day incubation, floating dead cells were discarded and washed with PBS twice, and alive adherent tumor cells were harvested and plated to 12-well plate for second-round CEA BiTE/T cell attack. For control, tumor cells alone or tumor cells with control BiTE and T cells were incubated for the same period. Cytotoxicity was analyzed by staining cells with FITC-conjugated anti-CD3, 7-AAD and annexin V. Tumor cells (CD3-negative) were analyzed for their annexin V positivity (% shown in each dot plot). a SW1463 cells. b AsPC-1 cells. c Change of CEA expression level after CEA BiTE-mediated T cell attack. CEA-expressing colorectal cancer cell lines (SW1463, Colo205, HT29, AsPC-1) were incubated with CEA BiTE/T cells for 5 days. As control, tumor cells were incubated alone or with Cont BiTE/T cells for the same period. Alive cells from the first round CEA BiTE-mediated T cell killing were put into second-round incubation with fresh T cells and CEA BiTE, incubated for another 5 days. The change in CEA expression level was analyzed by flow cytometry. Indirect staining method was used for detection of CEA expression with PE-conjugated goat anti-mouse IgG as a secondary antibody. FITC-conjugated anti-CD3 was added to eliminate T cells from the analysis. Filled histograms: anti-CEACAM5 mAb. Open histograms: isotype control. Percentage of positive cells and median fluorescence intensities are shown

Full size image

To determine whether the mechanism of tumor cell survival was due to downregulation of the target antigen, tumor-specific CEA expression was measured by mean fluorescence intensity (MFI) and showed to have increased at the time of the second co-culture. This increase in CEA expression level was not observed with incubation of tumor cells alone or with co-incubation of T cells and tumor cells without CEA BiTE (Fig. 1c). Furthermore, this effect was not unique to established colorectal cancer cell lines and was observed in primary culture of human colorectal cancer cells as well. In explants CRC057 and CRC096, the MFI for CEA increased by 109 % (1545–3227 MFI) and 48 % (2500–3700 MFI), respectively, after the second incubation of CEA BiTE + T cells (Fig. 2). Despite an increase in CEA expression and T cell-mediated cytolysis, a fraction of the cells remained alive after repeat exposure to CEA BiTE plus T cells, suggesting partial resistance of tumor to CEA BiTE-mediated T cell killing.

Fig. 2
figure 2

Metastatic cancer cells from colorectal cancer patients did not acquire resistance for T cell killing and had enhanced CEA expression. The cytotoxicity assay was done as in Fig. 1 for cancer cells derived from metastatic colorectal cancer lesions in patients. a CEA-positive CRC057 and CRC096 were used. Alive tumor cells from the first round CEA BiTE-mediated T cell killing were put into second-round incubation with or without fresh T cells and CEA BiTE, incubated for another 5 days. b Change in CEA expression level after CEA BiTE-mediated T cell killing was analyzed by flow cytometry. Representative staining with CRC057 cells is shown. Filled histograms: anti-CEACAM5 mAb. Open histograms: isotype control. Median fluorescence intensities are shown

Full size image

Diminished T cell activity over time after repeated CEA BiTE exposure

Another potential mechanism for tumor cell escape from CEA BiTE-mediated T cell killing could be upregulation of the PD-1/PD-L1 axis [22]. To determine the viability and function of T cells previously co-cultured with CEA BiTE and tumor cell lines SW1463 and AsPC-1, viable T cells were harvested and used as effector cells in repeat incubations with fresh tumor cells to assess for possible diminished T cell cytolytic activity with repeated CEA BiTE-mediated killing. Using matched PBMC for the same donor, fresh T cells induced 79.7 % killing after a 5-day co-incubation with SW1463 cells in the presence of CEA BiTE, whereas only 34.7 % killing was noted using T cells that had previously participated in CEA BiTE-mediated attack. With AsPC-1 cells under similar conditions, fresh T cells induced 63.0 % killing versus 27.5 % in T cells that had previously been cocultured with CEA BiTE (Fig. 3). Thus, viable T cells previously exposed to tumor cells showed diminished killing activity compared to fresh T cells. One potential explanation for this effect is the induction of immunoregulatory mechanisms.

Fig. 3
figure 3

Decreased killing activity of T cells after incubation with CEA BiTE and cancer cells. Tumor cells were incubated with T cells and CEA BiTE for 5 days, then floating cells were harvested and washed with PBS twice, and single cell suspension was obtained by repeated gentle pipetting. Alive cells were isolated by density gradient centrifugation with Ficoll-Paque. Harvested alive cells were resuspended in the medium, put into flasks to make tumor cells adhere. After 2 h incubation, non-adherent cells were harvested and used as T cells from CEA BiTE/Tumor cell culture. Fresh T cells were isolated from frozen stocked PBMCs of the same normal donor, using flask adherence technique and negative T cell isolation beads kit. T cells incubated alone in the flask for 5 days were also used. cytotoxic function of T cells was compared by staining cells with FITC-conjugated anti-CD3, 7-AAD, and annexin V. Tumor cells (CD3-negative) were analyzed for their annexin V positivity (% shown in each dot plot). a SW1463 cells as target cells. b AsPC-1 cells as target cells

Full size image

Enhanced PD-1 and PD-L1 expression in CEA BiTE-mediated T cell killing of tumor cells

A major mechanism of immune regulation is the induction of immune co-inhibitory pathways such as CTLA-4 and the PD-1/PD-L1 axis (expression of PD-1 by T cells and PD-L1 by tumor cells), leading to decreased T cell cytolytic activity. T cells were assayed for CTLA-4 and PD-1 expression after 7 days of CEA BiTE-mediated T cell attack of SW1463 (Fig. 4). While CTLA-4 expression was minimally changed, PD-1 expression on T cells increased in concert with CD28 and CD69, markers of T cell activation.

Fig. 4
figure 4

Increased PD-1 expression on T cells and PD-L1 expression on cancer cells are specific to co-incubation in the presence of CEA BiTE. a T cells were incubated with tumor cells (SW1463 or AsPC-1) in the presence of CEA BiTE or Cont BiTE (100 ng/mL) for 5 days, harvested and stained with APC-anti-CD45, and PE-conjugated anti-CD28, CTLA4, PD-1, PD-L1, or CD69. After 30-min incubation, T cells were washed twice with PBS and analyzed for CD45+ (T cell) and CD45− (tumor) cell populations. Numbers in each histogram show median fluorescence intensity of staining. b, c T cells were incubated with SW1463 cells in the presence of CEA BiTE or Cont BiTE (100 ng/mL) for up to 7 days, and overtime changes in PD-1 expression on T cells and PD-L1 expression on cancer cells were analyzed. After 1-, 3-, 5-, or 7-day incubation, cells were harvested and stained with APC-anti-CD45, and PE-conjugated anti-PD-1, or anti-PD-L1 mAb. CD45+ T cells (b) and CD45− (tumor) cells (c) were analyzed for their PD-1 and PD-L1 expression, respectively. Gray open histograms: PE-labeled isotype control IgG. Filled black histograms: Staining with PE-anti-PD-1 mAb (T cell) or with PE-anti-PD-L1 mAb (tumor cell)

Full size image

In addition to the increase in PD-1 expression on T cells, there was increased PD-L1 expression on tumor cells during the 7-day period (Fig. 4). Another CEA-expressing colorectal cancer cell line (HT29) also experienced induction of PD-L1 expression by day 7 (Fig. 6a). Co-incubation in the presence of CEA BiTE induced secretion of multiple Th1- and Th2-type cytokines [12], especially IFN-gamma that might be the trigger for PD-L1 upregulation [23]. We found that upregulated PD-1 on T cells and PD-L1 on tumor cells is an early event that occurs in the combined cultures within days of the first cycle of CEA BiTE-mediated T cell attack (Fig. 4b, c). These changes in PD-1 and PD-L1 expression did not occur following co-incubation of T cells and tumor cells in the absence of CEA BiTE, suggesting that this may be a potential mechanism of immune modulation, and a possible opportunity to reverse this dysfunction.

Effect of anti-PD-1 and anti-PD-L1 on T cell activity

In order to determine the importance of PD-1 and PD-L1 in T cell cytolytic inhibition and whether this effect could be abrogated, we tested inhibition of the PD-1/PD-L1 axis in tumor cells exposed to T cells and CEA BiTE using blocking antibodies specific to either PD-1, PD-L1, or the combination (Figs. 5, 6, Supplementary Figure 1). Dual inhibition of PD-1 and PD-L1 resulted in increased tumor cytolysis compared with anti-PD-1 or anti-PD-L1 monotherapy, and resulted in greater cytolytic activity compared with CEA BiTE alone. Blockade of PD-1 on fresh T cells with anti-PD-1 antibody at the initiation of the cultures led to increased CEA BiTE-mediated cytolysis of SW1463 (49.9 %) relative to fresh T cells exposed to CEA BiTE only (39.7 %), while dual blockade of PD-1 and PD-L1 induced greater cytolytic function of fresh T cells (74.1 %) (Fig. 5a).

Fig. 5
figure 5

Blocking effect of PD-1 and PD-L1 on CEA BiTE/T cell-mediated killing with pre-exposed T cells. a SW1463 cells (0.5 M cells/well) and T cells (2.5 M/well) from a normal donor were put into to the well with/without BiTE (100 ng/mL). Cells were incubated for 5 days in the presence of anti-PD-1 mAb, anti-PD-1/anti-PD-L1 mAbs or isotype control IgG (final 5 µg/mL). Tumor cells were harvested with trypsin/EDTA and stained with anti-CD3-FITC/7-AAD/annexin V-APC. Percentages of annexin V-positive cells in CD3-negative tumor cells are shown in each dot plot. b T cells from a normal donor were incubated with SW1463 tumor cells at 5:1 effector/target ratio in the presence of CEA BiTE or Cont BiTE (100 ng/mL) for 7 days. Floating cells were harvested, single cell suspension was obtained by gentle pipetting, and alive cells were isolated with gradient density centrifugation. Isolated T cells were used again for co-incubation with SW1463 cells in the presence of CEA BiTE (100 ng/mL) and anti-PD-1 mAb, anti-PD-1/anti-PD-L1 mAbs or isotype control IgG (final 5 µg/mL). After 5-day incubation, tumor cells were harvested and stained with anti-CD3-FITC/7-AAD/annexin V-APC. Percentages of annexin V-positive cells in CD3-negative tumor cells are shown in each dot plot

Full size image
Fig. 6
figure 6

Blocking effect of PD-1 and PD-L1 on CEA BiTE/T cell-mediated killing against HT-29 cells. a PD-1 expression on T cells and PD-L1 expression level on HT-29 cell line were analyzed by flow after 7-day co-incubation in the presence of CEA BiTE (100 ng/mL). Gray-filled histogram: T cells or Tumor cells alone stained with PE-conjugated isotype-control IgG. Gray open histogram: T cells or tumor cells alone incubated with CEA BiTE for 7 days, stained with PE-anti-PD-1 or PE-anti-PD-L1 antibody, respectively. Black line open histogram: T cells or tumor cells co-incubated for 7 days in the presence of CEA BiTE, stained with PE-anti-PD-1 or PE-anti-PD-L1 antibody, respectively. b Effect of anti-PD-1, anti-PD-L1, and combination of these antibodies on CEA BiTE-mediated killing with pre-exposed T cells was analyzed. T cells, incubated with HT-29 cells in the presence of CEA BiTE (100 ng/mL) for 7 days, were harvested and used as effector cells in the second-round incubation with HT-29 cells and CEA BiTE. Anti-PD-1 mAb, anti-PD-L1 mAb, combination of anti-PD-1/anti-PD-L1 mAbs, or isotype control IgG (final 5 µg/mL) was added. After 5-day incubation, cells were stained with anti-CD3-FITC/7-AAD/annexin V-APC. CD3-negative FSC large tumor cells were analyzed for annexin V and 7-AAD labeling. Percentages of annexin V-positive cells in tumor cells are shown in each dot plot. c Enhanced CEA BiTE-mediated T cell cytotoxicity with PD-1/PD-L1 blockade is summarized. Percentages of annexin V-positive cells in Cont BiTE condition were subtracted from each CEA BiTE+ conditions as a background. Fold increase compared to CEA BiTE+ IgG condition is shown for each condition for both cell lines (SW1463, HT-29)

Full size image

To determine whether T cells that had previously engaged tumor could be rescued from the effects of PD-1 upregulation, we compared the cytolytic function of T cells that had been exposed to tumor cells + CEA BiTE with T cells that had been exposed to tumor cells + Cont BiTE. Cytolytic function of these T cell populations was measured during a second round of exposure to CEA BiTE and tumor cells in the presence of anti-PD-1 antibody or the combination of anti-PD-1 and anti-PD-L1. T cells previously incubated with CEA BiTE and tumor in the absence of PD1/PD-L1 blockade and then re-incubated for a second round of tumor cytolysis, had decreased cytolytic efficacy even when PD-1/PD-L1 blockage was instituted. For example, addition of anti-PD-1 during the second round of tumor cytolysis was not associated with improvement in T cell killing of tumor (37.2 vs 35.9 %) for T cells that had previously participated in CEA BiTE-mediated tumor killing. The combination of anti-PD-1 and anti-PD-L1 during the second round of tumor cytolysis was associated with an improvement in killing (50.0 %), but not to the level seen with the T cells that had not participated in CEA BiTE-mediated killing of tumor during the first round of exposure (75.3 %) (Fig. 5b). To confirm the findings, the experiment was repeated using the same SW1463 cell line and another CEA-expressing colon cancer line, HT-29, with the additional condition of anti-PD-L1 blockade alone (Fig. 6 and Supplementary Figure 1). In Fig. 6c, the enhancement (fold increase) of CEA BiTE-mediated T cell cytolytic function in the presence or absence of PD-1/PD-L1 blockade is shown for both cell lines. Percentages of annexin V-positive cells in the Cont BiTE condition were subtracted as a background. We confirmed mild enhancement of CEA BiTE-mediated cytotoxicity by PD-1 and PD-L1 blockade, and dual blockade induced stronger cytotoxicity compared with single blockade with anti-PD-1 or anti-PD-L1.

Overall, tumor killing by the T cells that had previously been exposed to tumor remained persistently inferior to killing with fresh T cells (Figs. 5, 6, Supplementary Figure 1). These data suggest that once PD-1 upregulation has already occurred on T cells, cytolytic activity is diminished despite intervention with anti-PD-1. In addition, these data suggest that the effects of PD-1 and PD-L1 on T cell function may be partly inhibited by immune checkpoint blockade, particularly combinatorial blockade, but immunomodulation cannot easily be reversed fully once it has occurred in this in vitro assay, suggesting that early or prophylactic intervention is required to maximize T cell-mediated cytolysis of tumor.

Discussion

We demonstrated in an in vitro model where all effector T cells have the potential to mediate anti-tumor activity that the tumor may remain partially resistant to multiple rounds of T cell-mediated killing. A commonly reported mechanism for tumor immune escape, loss of the target antigen, in this case CEA, did not occur. This result is clinically relevant, as there has been concern that antigen escape mutants would minimize the clinical impact of any antigen-specific type of immunotherapy. Rather, upregulation of the PD-1/PD-L1 axis was observed, suggesting that T cell suppression is the cause of incomplete tumor cell killing. Blockade of this immune checkpoint, particularly exposure to the anti-PD-1 and anti-PD-L1 mAbs at the start of the T cell/CEA BiTE/tumor co-cultures, was an effective strategy for partially restoring T cell function.

An important observation is that the tumor cells, which did not express PD-L1 at baseline, rapidly upregulated PD-L1 after exposure to CEA BiTE and T cells. This is likely due to interferon-gamma generated by T cells initiating their attack against the tumors [23]. We previously reported the robust increase of interferon-gamma in co-cultures of tumor, T cells, and CEA BiTE [12]. Further, we observed rapid upregulation of PD-1 expression on T cells after exposure to CEA BiTE and tumor. This is not unexpected as the BiTE molecules had previously been optimized so that they only stimulate T cells to express surface activation markers and cytokines when they are co-cultured with tumor cells that express the antigenic target [24]. As PD-1 is upregulated during T cell activation [25], it is expected that PD-1 would be upregulated during BiTE engagement of tumor and T cells. The result is coordinated expression of PD-L1 and PD-1, the interaction of which impairs the ability of the T cells to eradicate tumor [17, 26, 27] and leads to tumor immunologic escape [28, 29].

Inhibition of the PD-L1/PD-1 interaction with anti-PD-1 or anti-PD-L1 antibody therapy or both improved but did not completely restore T cell cytolysis in our experiments. In view of our observation of early expression of PD-1 and PD-L1 following CEA BITE and T cell exposure, we were gratified to find that early intervention with therapeutic immune checkpoint inhibition has the greatest impact on abrogating tumor resistance. Our data suggest that T cell exhaustion is not an irreversible process, but it is less readily reversed once there has been direct exposure of T cells to tumor cells allowing sufficient time for the upregulation and interaction of PD-1 and PD-L1 to have occurred. This observation is somewhat different than observed by Ahmed’s group [30] during anti-PD-1 therapy for chronic viral infections in which the T cell exhaustion is more readily reversible by PD-L1/PD-1 blockade and is associated with downregulation of PD-1 expression; however, they observed this downregulation while the viral titer simultaneously decreased which would be expected to reduce immunosuppressive stimuli. In our model system, the continued direct engagement of the T cells with tumor cells may maintain the immunosuppressive interaction so that it is not as readily ameliorated. In this regard, it is interesting to note that it has been reported that CD8+ T cells in the tumor-free lymph nodes of colorectal cancer patients, despite expressing PD-1, retain the ability to produce IL-2, IFN-γ and perforin, while the PD-1-positive tumor-infiltrating T cells are exhausted [31]. In this scenario, tumor-infiltrating lymphocytes continue their direct exposure to tumors resulting in ongoing immunosuppression.

We observed that dual inhibition of PD-1 and PD-L1 was slightly superior to anti-PD-1 monotherapy (as per Fig. 5). One potential explanation is that PD-L1 and PD-1 have other partners that may influence their overall impact on immune response [32]. PD-L2 may bind to PD-1, affecting T cell function despite anti-PD-L1 blockade [33]. T cell-expressed CD80 may bind to PD-L1, impairing T cell function despite anti-PD-1 therapy [34]. As anti-PD-1 and anti-PD-L1 dual blockade can only partially restore T cell cytolysis against tumor, other immune checkpoints or mechanisms of resistance to T cell killing may be concomitantly upregulated with PD-1. Further characterization of other immune checkpoints that may interfere with CEA BiTE-mediated tumor cytolysis is warranted.

One important aspect of our study is the use of BiTE antibodies to arm all the T cells in culture against the target antigen. This permits analysis of influences on T cell cytolytic activity without the potential confounding effects of poor T cell stimulation or trafficking to tumor; however, a limitation of our assay system is that it may not mimic the microenvironment of solid tumors where T cell to tumor ratios may differ substantially and stromal elements and other immune cells may interfere with tumor/T cell interactions. We chose the assay system because we had previously used it to study CEA BiTE-mediated killing of tumor cells and because it allowed us to test directly whether PD-1/PD-L1 interactions could be one reason that there was not continued ability of T cells to kill after exposure to tumor.

CEA BiTE is currently being studied in a phase I clinical trial, and immune correlates of tumoral protection from T cell-mediated cytolysis are being assessed. Blinatumomab, a CD19-targeting BiTE being studied in refractory hematologic malignancies, has shown promising activity with durable molecular remissions lasting 3 years in B cell ALL [35]. To date, mechanisms of resistance to this BiTE antibody, as well as profiling of PD-1/PD-L1 expression on patients treated with blinatumomab, have not been published. Finally, based on the demonstration of clinical activity of single agent T cell immune checkpoint inhibitors targeting CTLA-4 (ipilimumab) [36], PD-1 [37], and PD-L1 [38], clinical trials of combination therapies with other immune therapies are ongoing. Future studies of CEA BiTE should incorporate concomitant targeting of immunomodulatory molecules based on assessment of immune inhibitory feedback mechanisms.