IFNγ-Dependent Interactions between ICAM-1 and LFA-1 Counteract Prostaglandin E2–Mediated Inhibition of Antitumor CTL Responses

Strengthening tumor-specific CD8⁺ cytotoxic T lymphocyte (CTL) responses is one of the most promising recent strategies in cancer therapy. However, antitumor CTL activity is widely suppressed within the tumor microenvironment. Understanding mechanisms of suppression and ways to overcome them are key challenges in cancer immunotherapy (1–6).

Overproduction of PGE₂ by tumor cells as a result of overexpression of cyclooxygenase (COX)-2 (7–11) suppresses antitumor immune responses. Studies have shown that PGE₂ suppresses both innate and adaptive immune responses. It alters natural killer (NK)-cell effector function by decreasing TNFα and IFNγ production (12) and reduces phagocytosis by macrophages (13). PGE₂ can also disrupt antigen presentation by dendritic cells (DC), either by inhibiting maturation, which alters the profile of cytokine production and reduces secretion of IL2 and IL12, or by inducing exhaustion, which prevents the induction of CTL, type 1 helper T (Th1) cells, and NK cells in favor of Th2 responses (14–16).

In earlier studies from our laboratory, murine renal cell carcinoma (Renca) cells that expressed the hemagglutinin (HA) protein from influenza virus A/PR/8/H1N1 (PR8) as a tumor-specific neoantigen (RencaHA cells) were generated. We transfected them to overexpress COX-2 (RencaHA/T3 cells), which resulted in elevated PGE₂ production. Overexpression of COX-2 by RencaHA/T3 cells induced metastasis to the tumor-draining lymph nodes (TDLN). Yet, despite tumor growth, when CD8⁺ T cells transgenic for the CL4 T-cell receptor (an H-2K^d/HA–specific TCR) were adoptively transferred into RencaHA/T3 tumor–bearing mice, CL4 T cells in the TDLN remained naïve (15). The lack of CL4 T-cell proliferation and CTL effector function was dependent on COX-2 overexpression by RencaHA/T3. Treatment of RencaHA/T3 tumor–bearing mice with the selective COX-2 inhibitor NS-398 restored both proliferation and CTL effector function to CL4 T cells within the TDLN to the same level found within the TDLN of COX-2–negative RencaHA tumor–bearing mice.

The immunosuppressive effects of PGE₂ are known to be associated with increased intracellular cyclic adenosine monophosphate (cAMP; refs. 13, 17), which modulates the effector function of T cells and inhibits the stabilization of the immunologic synapse formed through LFA-1–ICAM-1 interactions (17–20), which influences subsequent IFNγ-dependent T-cell proliferation. Several studies have shown that cell-surface expression of ICAM-1 by tumor cells can be induced by IFNγ derived from CD8⁺ T cells during productive activation (21, 22). Therefore, we hypothesized that suppression of IFNγ production by CL4 CTLs interacting with metastatic RencaHA/T3 in the presence of high concentrations of PGE₂ would inhibit upregulation of ICAM-1 on tumor cells, which is required to drive proliferation and differentiation of tumor-specific CTL responses.

In this report, we demonstrate that the concentration of PGE₂ during CD8⁺ T-cell/tumor cell interactions plays an essential role in determining the outcome of the response, shifting from productive activation of CTLs at low concentrations, toward antigen-specific tolerance induction at high concentrations. We show that exogenous PGE₂ prevented the direct priming of CL4 CTL responses in vitro by suppressing IFNγ production by CL4 T cells when they initially interacted with RencaHA tumor cells, a vital event for proper priming. We also show that suppression by PGE₂ was temporary and could be mitigated by increasing cell-surface expression of ICAM-1 by tumor cells. ICAM-1 expression not only drove tumor-specific CD8⁺ T-cell proliferation, but also limited tumor growth in vivo.

Thy1.1^+/+ CL4 TCR–transgenic BALB/c mice (6–8-week-old; ref.23) and Thy1.2^+/+ BALB/c mice were maintained under specific pathogen-free conditions at the University of Bristol Animal Services Unit. Some BALB/c mice were injected subcutaneously (s.c.) with 1 × 10⁶ Renca tumor cells, and tumor growth was assessed as previously described (15, 21).

COX-2–overexpressing RencaHA/T3 cells (9, 21, 24) were transfected with 1.5 μg of ICAM-1–expressing pIREShyg plasmid (a kind gift from Prof. Adrian Whitehouse at Leeds University) to generate RencaHA/T3/ICAM-1. Renca cells were grown in a complete medium (RPMI-1640; Sigma-Aldrich) supplemented with 10% vol/vol fetal calf serum (FCS, Invitrogen); 2 mmol/L l-glutamine (Invitrogen); penicillin/streptomycin (50 U/mL, Invitrogen), 5 × 10⁻⁵ mol/L 2-mercapto-ethanol (Invitrogen). Medium was further supplemented with geneticin (100 μg/mL, Invitrogen) for RencaHA cells, plus puromycin (1 μg/mL; Biomatik) for RencaHA/T3, plus hygromycin B (250 μg/mL, Invitrogen) for RencaHA/T3/ICAM-1. Renca cells are thawed from a large repository of the original frozen “parental” stocks. From each aliquot of parental stock, we generated a second batch of aliquoted and frozen “user” stock cells that provide us with a supply of cells to work with. Our laboratory performs regular Mycoplasma screening, and the cells are validated at the time of thawing in cell-culture medium according to standard operating procedures prior to their use in vitro and in vivo. We monitor phenotypic characteristics such as HA protein expression, class I MHC, ICAM-1, and COX-2/PGE₂ expression. We also revalidate their ability to prime naïve K^d/HA-specific CL4 CD8⁺ cells and to be susceptible to lysis by CL4 CTLs. Experience with these cells over the past 10 years informs us that they retain their phenotype and function, and their growth characteristics in vitro and in vivo throughout the 3 months they are routinely kept in continuous culture. All revalidation is documented prior to replenishment of user stock.

Single-cell suspensions from peripheral lymph nodes and spleen of Thy1.1⁺ CL4 TCR–transgenic mice were enriched for CD8⁺ T cells using anti-CD8 magnetic-activated cell sorted (MACS) midiMACS (Miltenyi Biotec; ref. 25).

MACS-purified naïve CL4 CD8⁺ T cells were pelleted, resuspended in 50 × 10⁶ cells/mL in PBS, and mixed with 5 μmol/L carboxyfluorescein diacetate succinimidyl ester (CFSE; BioLegend), for 15 minutes at 37°C in the dark, and then washed with 45 mL complete medium.

Naïve CFSE-labeled CL4 CD8⁺ T cells were cultured with irradiated tumor cells at a ratio of 1:1 in the presence of different concentrations of PGE₂ (Sigma-Aldrich). In some experiments, RencaHA cells were treated overnight with different concentrations of PGE₂, before being washed and used for priming naïve CL4 cells. In other experiments, CL4 cultures were treated with 10 ng/mL of recombinant (r)IFNγ (Peprotec), or cultured in plates coated with either 1, 3, or 5 μg/mL of rICAM-1 (Sion Biological Inc.), overnight before adding RencaHA and CL4 cells.

Tissue culture plates were coated with 10 mg/mL anti-CD3 (eBioscience) at 4°C. Control wells contained PBS only. MACS-purified, CFSE-labeled naïve CL4 cells (5 × 10⁴ to 25 × 10⁴) were cultured, with or without CD28 monoclonal antibodies (mAbs; 5 μg/mL, eBioScience), for 48 at 37°C with 5% vol/vol CO₂. CL4 cells were then collected and stained with various other mAbs. In other experiments, plates were coated with CD3 mAbs and/or rICAM-1 before the addition of naïve CL4 cells. For the secondary activation of CL4 CD8⁺ T cells preactivated by tumor, CFSE-labeled CL4 cells (1 × 10⁶) recovered from primary cultures were restimulated with 1 × 10⁵ freshly irradiated tumor cells in 24-well plates in the presence or absence of 10^–6 mol/L of PGE₂ for 37°C in 5% vol/vol CO₂.

CL4 cells were treated with 1 mmol/L of 3-Isobutyl-1-methylxanthine (IBMX), an inhibitor of cAMP phosphodiesterases, for 1 hour at 37°C followed by 10^–6 mol/L PGE₂ for 10 minutes, then lysed at 4°C. ELISA was performed using the BIOMOLFormat A Cyclic AMP plus EIA kit, according to the manufacturer's instructions. Plates were read at 405 nm with a 590-nm reference using a 3550 microplate reader (BioRad). Data were analyzed using Microplate Manager 4.0 (BioRad), and the graphs were plotted using Prism 4.03 software (GraphPad).

MACS-purified CL4 cells were restimulated in vitro for 4 hours in the presence of 1 μg/mL of both K^d/HA peptide and Golgiplug (BD Bioscience), stained with Zombie Aqua to exclude dead cells (Biolegend), and various fluorescently conjugated mAbs against surface markers CD69, CD62L, and CD44 (BioLegend). CL4 cells were then permeabilized using the BD Perm/fix kit (BD Bioscience) according to the manufacturer's instructions and stained with anti-IFNγ mAbs (Biolegend). Cells were analyzed using an LSRII or FACSCalibur flow cytometer with DiVa or CellQuest software, respectively (BD Cytometry Systems).

To determine if PGE₂ affected the direct priming of naïve tumor-specific CD8⁺ T cells by tumor cells, CFSE-labeled naïve K^d/HA-specific CL4 CD8⁺ T cells were cocultured with RencaHA cells for 72 hours in the presence of increasing concentrations of PGE₂ ([RencaHA + CL4] + PGE₂; Fig. 1A, top). Nontransfected (HA-negative) RencaNT cells were used as a negative control. After coculture, CL4 cells were isolated and analyzed for proliferation using standard ³H-thymidine incorporation (Fig. 1, left), as well as for intracellular IFNγ expression using flow cytometric analyses (Fig. 1, right). In the absence of PGE₂, CL4 cells proliferated and produced IFNγ production, compared with CL4 cells cocultured with RencaNT cells. However, in the presence of 10⁻⁶ to 10⁻¹⁰ mol/L PGE₂, proliferation and IFNγ production by CL4 cells was significantly reduced when compared with untreated cultures, and at 10⁻¹¹ mol/L PGE₂ CL4 proliferation increased in response to RencaHA cells, although this increase was not statistically significant. Importantly, when RencaNT cells were cultured with CL4 cells in the presence of PGE₂, no statistically significant change was observed among CL4 cells, and they maintained their naïve phenotype (data not shown). Correlating with reduced proliferation, among CL4 cells cultured with RencaHA cells in the presence of the highest concentrations of PGE₂ (10⁻⁶ and 10⁻⁷ mol/L), IFNγ expression was significantly reduced compared with expression of PGE₂-untreated cultures. However, when CL4 cells were cultured with RencaHA at lower concentrations of PGE₂ (10⁻¹⁰ mol/L), both proliferation and IFNγ expression increased compared with untreated cultures.

The regulation of naïve CD8⁺ T-cell responses by PGE₂ could be direct, or indirect, by conditioning the tumor cells. To determine whether the effect is indirect, RencaHA cells were pretreated with various concentrations of PGE₂ (Fig. 1A, bottom). However, prior to coculture, RencaHA cells were washed to remove any excess PGE₂ ([RenaHA + PGE₂] + CL4). After 72 hours, CL4 cells were isolated and analyzed for proliferation (left) and intracellular IFNγ expression by flow cytometry (right). The data show that, compared with PGE₂-untreated cultures, pretreatment of RencaHA cells with PGE₂ did not result in a statistically significant reduction in CL4 proliferation even at the highest concentration of PGE₂. Furthermore, IFNγ expression was also unaffected by pretreatment of RencaHA cells with PGE₂, with a slight increase in expression being statistically significant only at 10⁻⁸ mol/L. Together, these data suggest that high concentrations of PGE₂ suppress proliferation and CTL effector function in vitro by acting directly upon CL4 cells.

The direct effect of high concentrations of PGE₂ on CL4 cell priming was further examined by culturing CFSE-labeled CL4 cells in the presence of immobilized anti-CD3 and anti-CD28 mAb with or without 10⁻⁶ mol/L PGE₂ (Fig. 1B). Although untreated CL4 T cells proliferated and elaborated IFNγ, the presence of high concentrations of PGE₂ resulted in a reduction in proliferation, with many more undivided CFSE-high cells, and a statistically significant decrease in IFNγ expression. Thus, the inhibitory effects of PGE₂ are most likely mediated through direct action on naïve CL4 cells.

To determine whether or not inhibition of CL4 T-cell proliferation and IFNγ production by high concentrations of PGE₂ is permanent, naïve CFSE-labeled CL4 cells were first primed in vitro with RencaHA cells in the presence or absence of PGE₂. After 48 hours, CL4 cells were isolated and washed to remove excess PGE₂, before undergoing a secondary culture for a further 72 hours with fresh RencaHA cells in the presence or absence of more PGE₂. CL4 cells cultured in the presence of nontransfected, HA-negative RencaNT cells were used as a control.

The data show that CL4 cells from primary cocultures without PGE₂ proliferated and produced IFNγ (Fig. 2A; top row), and following secondary coculture with fresh RencaHA cells alone in the absence of PGE₂, resulted in further proliferation with over than 40% of cells being IFNγ⁺. However, when PGE₂ was added to these secondary cocultures, proliferation was less, with fewer than 10% of cells being IFNγ⁺. Critically, when PGE₂-treated CL4 cells from primary cocultures underwent a secondary coculture with fresh RencaHA cells in the absence of PGE₂, nearly half of the CL4 cells had proliferated, and one-tenth of divided cells were IFNγ⁺, as compared with CL4 cells to which PGE₂ was also added in the secondary coculture (Fig. 2A; bottom row). Therefore, suppression by PGE₂ was reversible, and withdrawing PGE₂ from the environment could restore CL4 proliferation and CTL effector function.

The inhibition of effector function among PGE₂-treated CD8⁺ T cells is associated with increased cAMP, which inhibits IFNγ production and the acquisition of CTL effector function (17–20, 26). To establish whether or not PGE₂ suppresses naïve CL4 cells in a cAMP-dependent manner, naïve PGE₂-treated CL4 cells were lysed, and intracellular cAMP concentrations were measured using an ELISA. Following PGE₂ treatment, cAMP expression among CL4 cells rose by around 3-fold compared with expression in untreated CL4 cells (Fig. 2B).

We previously demonstrated that IFNγ treatment of RencaHA cells increased antitumor CTL responses by enhancing MHC class I (H-2K^d) expression by tumor cells (21). To determine whether or not the presence of exogenous IFNγ could directly influence CL4 T-cell responses in the presence of PGE₂, CFSE-labeled naïve CL4 cells were cocultured with RencaHA cells in the presence or absence of PGE₂ and rIFNγ and proliferation (determined by reductions in CFSE) and IFNγ production were assessed by flow cytometry. The proliferation of and IFNγ production by CL4 T cells were inhibited by PGE₂ (Fig. 3A). However, the addition of rIFNγ to these PGE₂-treated RencaHA + CL4 cocultures counteracted the suppression of proliferation and IFNγ production. Addition of rIFNγ also resulted in increased expression of the early activation marker CD69 on proliferating CL4 cells, indicating that the presence of rIFNγ, during the initial priming of naïve CL4 cells, is essential to reverse the suppressive effects of PGE₂ (Fig. 3B). Treatment of RencaNT + CL4 cocultures with rIFNγ did not result in any nonspecific increase in proliferation among CL4 cells (Fig. 3A and B, top rows).

As IFNγ receptors are expressed on a variety of cell types, including naïve CD8⁺ T cells and tumor cells (27, 28), it is possible that in the RencaHA + CL4 cocultures, IFNγ may be acting on either or both cell types. To address this issue, naïve CFSE-labeled CL4 cells were primed with mAb to CD3 and CD28 in the presence or absence of PGE₂ with or without rIFNγ. As anticipated, control mAb-primed CL4 cells elaborated IFNγ after 72 hours, which was further enhanced by the addition of rIFNγ (Fig. 3C). However, although after PGE₂ treatment, IFNγ production by CL4 cells primed with mAbs to CD3 and CD28 was inhibited, addition of rIFNγ did not restore IFNγ production by CL4 cells. Therefore, these data clearly show that reversal of the PGE₂-mediated suppression of CTL effector function by IFNγ can only occur in the presence of RencaHA cells.

To test whether or not rIFNγ counteracts PGE₂-mediated suppression by acting directly on RencaHA cells, cocultures were set up, in which RencaHA cells were treated with rIFNγ at different time points in the presence of PGE₂ (Fig. 4). Addition of 10 ng/mL of rIFNγ to RencaHA + CL4 cocultures at 0 and 20 hours resulted in an increase in proliferation, as both CFSE and IFNγ were reduced (Fig. 4; middle two rows). However, the addition of rIFNγ to the cocultures after 40 hours of exposure to PGE₂ did not enhance IFNγ production by CL4 cells (Fig. 4; bottom row). This suggests that the reversal of PGE₂-mediated suppression by rIFNγ can occur in the very early stages of CL4 T-cell priming.

Not only is ICAM-1 interaction with T cell–expressed LFA-1 crucial for the homotypic T-cell aggregation that is required for T-cell communication and exchanging information (29), but the LFA-1–ICAM-1 interaction can transduce downstream costimulatory signals and drive T-cell proliferation (6, 27). We have shown that the cell-surface expression of ICAM-1 by RencaHA cells is crucial for direct priming of naïve CL4 T cells. Moreover, we showed that upregulation of ICAM-1 expression is induced by IFNγ derived from CL4 cells during their early activation by RencaHA cells. This reinforces further proliferation and the induction of CTL effector function (21, 22). Based upon these findings, we wished to compare classical anti-CD28 costimulation with alternative ICAM-1–mediated costimulation through LFA-1 (Fig. 5A). In the presence of rICAM-1, proliferation of naïve CL4 T cells was significantly greater compared with the proliferation in the presence of anti-CD28 mAbs. Whereas some CL4 cells proliferated in response to CD3 mAb alone, rICAM-1 alone did not induce any proliferation (Fig. 5A). These results indicated that ICAM-1 acts as a highly potent alternative costimulatory molecule to drive naïve CL4 T-cell proliferation.

It is known that following priming, T cells decrease surface expression of CD62L and increase CD44 expression, enabling extravasation through blood vessels into inflamed tissues and the formation of effector memory T (T_em) cells (28). To test whether or not classical (CD3 mAb + CD28 mAb) or alternative (CD3 mAb + rICAM-1) priming of naïve CL4 cells in vitro gave rise to T_em cells, CFSE-labeled naïve CL4 cells were primed accordingly, and the expression of CD62L, CD44, and IFNγ was assessed by flow cytometry. Treatment of CFSE-labeled CL4 cells with anti-CD3 + rICAM-1 consistently produced more IFNγ-expressing cells that divided two or more times, than when CL4 cells were primed with anti-CD3 + anti-CD28 (Fig. 5B). Therefore, the induction of CTL effector function is at least comparable. Given the high expression of CD44 and low expression of CD62L under both conditions, the cells are T_em CTLs (Fig. 5B).

Our previous studies showed that, although low expression of ICAM-1 by RencaHA cells is sufficient to prime naïve CL4 cells (21), PGE₂ has the ability to prevent priming of CL4 cells in vitro and in vivo (15). Therefore, we wished to compare the effect of PGE₂ on classical (CD3 mAb + CD28 mAb), with alternative (CD3 mAb + rICAM-1) priming of CL4 cells in vitro. Furthermore, we also wished to determine whether or not increasing ICAM-1–mediated costimulation, by the addition of rICAM-1, could counteract PGE₂-mediated inhibition of naïve CL4 T-cell priming.

Addition of exogenous PGE₂ has a greater inhibitory effect upon classical CD28-mediated costimulation compared with alternative ICAM-1-LFA-1 costimulation (Fig. 6A). When CL4 cells were cocultured with RencaHA cells, addition of PGE₂ resulted in reduced CL4 proliferation by about two-thirds. However, in the presence of rICAM-1, CL4 proliferation, in response to coculture with RencaHA cells, was unaffected by PGE₂ (Fig. 6B). Thus, although PGE₂ has a greater inhibitory effect upon “classical” costimulation compared with “alternative” costimulation pathway, suppression is significantly abrogated by the addition of rICAM-1.

It is evident that IFNγ exerts its antitumor effects by directly (27, 30) enhancing immunogenicity through elevated ICAM-1 expression (10, 31). To determine whether or not upregulation of ICAM-1 is instrumental in abrogating PGE₂-mediated suppression, COX-2–overexpressing RencaHA/T3 cells (15) were further transfected with a cDNA plasmid expressing full-length murine ICAM-1. The resulting RencaHA/T3/ICAM-1 cell line expressed much more ICAM-1 compared with conventional RencaHA/T3 cells, which increased further after treatment with rIFNγ (Fig. 7A). Despite having equivalent amounts of HA protein and PGE₂ production (data not shown), overexpression of ICAM-1 by RencaHA/T3/ICAM-1 cells increased proliferation of naïve CL4 cells in vitro compared with RencaHA/T3 cells (Fig. 7B). Whereas s.c. injection of RencaHA/T3 cells into BALB/c mice resulted in the formation of solid tumors after 2 weeks, s.c. injection of RencaHA/T3/ICAM-1 cells did not result in tumor growth (Fig. 7C and D). Together, these data clearly show that overexpression of ICAM-1 can counteract PGE₂-mediated immunosuppression, restoring CTL effector function and profoundly preventing tumor growth in vivo.

Overexpression of COX-2 by RencaHA/T3 cells not only results in abortive activation of CL4 CD8⁺ T cells in the TDLN of RencaHA/T3 tumor-bearing mice, but also facilitates metastasis of RencaHA/T3 cells to the TDLN (15). However, while metastases of RencaHA/T3 cells to the TDLN would allow direct priming of naïve CL4 cells, overproduction of PGE₂ prevented the induction of antitumor CTL responses in these mice (32). The experiments described in this report set out to determine the mechanisms of PGE₂-mediated immunosuppression.

PGE₂ exhibits various and sometimes opposing effects on the immune responses. For example, not only does PGE₂ stimulate activation of mast cells (33, 34), it also inhibits cytokine release by macrophages (35). We have shown that the concentration of PGE₂ is a major factor in determining its overall affect upon CD8⁺ T-cell responses. At physiologic concentrations of ≤20 ng/mL, produced by constitutive expression of COX-1 (36), PGE₂ is able to enhance productive activation of CL4 cells. However, at high concentrations found in COX-2–overexpressing tumor microenvironments in vivo (37), PGE₂ prevents proliferation and IFNγ production of CL4 cells in response to RencaHA cells. We show that PGE₂ mediates its effects through direct action on CL4 cells. In the absence of RencaHA cells, proliferation and IFNγ production by PGE₂-treated CL4 cells primed by anti-CD3 + anti-CD28 mAb was also reduced, and pretreatment of RencaHA cells with PGE₂ did not reduce proliferation or IFNγ production by CL4 cells.

The fact that the PGE₂-mediated suppression of CL4 cell responsiveness could not be reversed by the addition of IL2 (Ahmadi and Morgan, unpublished data) suggests that PGE₂ does not induce a state of functional unresponsiveness due to anergy. Our findings correlate with other data showing that PGE₂ directly inhibits CTL function, such as IFNγ production (20). However, importantly, PGE₂-mediated suppression of CL4 cells is reversible, as secondary restimulation of PGE₂-conditioned CL4 cells, with RencaHA cells in the absence of PGE₂, restored proliferation and IFNγ production. Furthermore, irrespective of the primary culture conditions, the presence of PGE₂ in the secondary cultures renders CL4 cells refractory to proliferation and IFNγ production.

Consistent with the inhibition of CL4 CTL responses was an observed increase in intracellular cAMP following treatment with PGE₂. Increased cAMP inhibits IFNγ production by T cells as well as inhibits stabilization of the interaction between LFA-1 and ICAM-1 (17, 18), thus preventing T-cell activation due to the inability of the T cells and antigen-presenting cells to form a stable synapse (18, 19). Yet, we showed that CL4 cells increased cell-surface expression of the early activation marker CD69, possibly as a result of cognate TCR signaling (38). We suggest that, although some initial cognate interactions may occur that result in the upregulation of CD69, these interactions are not sustained sufficiently enough to trigger IFNγ production to promote further interactions. Our finding that rIFNγ reversed PGE₂-mediated suppression of naïve CL4 priming by directly increasing the immunogenicity of tumor cells correlates with findings from other studies that suggest that IFNγ enhances antigen presentation by tumor cells through increased MHC class I expression (21, 22).

Previously, we showed that in the absence of classical CD28-mediated costimulation, the interaction of LFA-1 with ICAM-1 provides sufficient costimulatory signals to prime naïve CD8⁺ T cells and induce CTL function (8, 10). We now show that costimulation provided through LFA-1 is more potent than costimulation through CD28, which may arise from that fact that the type and quantity of the signals induced within the T cell by these two costimulatory receptors are distinct from each other (29).

In the presence of rICAM-1, CL4 cells produced more IFNγ and were CD44^high and CD62L^low, facilitating T-cell migration from lymph nodes to the site of inflammation, a typically T_em CTL phenotype (30). Such differences were also observed when priming naïve CL4 cells with RencaHA tumor cells, in which costimulation occurs solely through the ICAM-1–LFA-1 interaction. Blocking LFA-1 or ICAM-1 causes a reduction in CTL production of IFNγ and lysis (1, 2), therefore enhancing tumor invasion and metastasis. In addition, studies have shown in breast cancer that silencing of the ICAM-1 gene by siRNA decreased tumorigenicity in vitro (39).

We showed that exposure to PGE₂ during priming with both classical and alternative costimulation pathways gave rise to a reduction in CL4 proliferation and IFNγ production. However, PGE₂-mediated suppression of priming with classical costimulation was greater than that with alternative rICAM-1–mediated costimulation. In the presence of PGE₂, addition of rICAM-1 together with RencaHA-expressed ICAM-1 restored CL4 proliferation. Thus, in the presence of PGE₂, LFA-1–ICAM-1 signaling is essential to fully initiate CL4 T-cell priming, presumably due to the stabilizing effect of rICAM-1 on T-cell/tumor cell interactions, increasing TCR signaling.

Increased expression of ICAM-1 among melanoma cells gives rise to an increase in the lysis susceptibility of melanoma cells by lymphokine-activated killer cells (LAK; ref. 40). Other studies also showed that transfecting gastric cancer cells with ICAM-1 causes significant increases in both adhesion to peripheral blood mononuclear cells (PBMC) and subsequent lysis (41). Induction of ICAM-1 by artificial transcription factors results in decreased growth of ovarian cancer cells (42). Studies in mice show that siRNA-mediated silencing of hepatic ICAM-1 in vivo prior to injecting C26 murine colon carcinoma cells alters the liver microenvironment. In ICAM-1^−/− mice, reduced numbers of C26 cancer cells were obtained from the liver as well as myeloid suppressor cells, and the numbers of TILs were increased compared with those in controls (43). In our study, we found that overexpression of ICAM-1 in RencaHA/T3 cells restored their ability to induce CL4 cell proliferation in vitro, despite the presence of PGE₂. However, although these RencaHA/T3/ICAM-1 cells express abundant ICAM-1 after transfection, treatment with exogenous IFNγ could further increase ICAM-1 expression. Unlike conventional RencaHA/T3 cells, RencaHA/T3/ICAM-1 cells did not form tumors in BALB/c mice, suggesting that potent antitumor CTLs had been primed, which eradicated the tumors.

Based upon our findings, we propose the following model to explain the role of alternative LFA-1–ICAM-1 costimulatory interactions in restoring CTL function in the presence of PGE₂ (Supplementary Fig. S1). In the steady state, naïve CL4 T cells express LFA-1 with low affinity for ICAM-1, and although coculture of naïve CL4 cells with RencaHA cells enables CL4 TCR interactions with H-2K^d/HA peptide complexes, this interaction is transient and may be terminated if a synapse does not form between the interacting cells. However, this initial contact is sufficient to stimulate Ca²⁺ influx inside CL4 cell, which provides the necessary stimuli to increase the affinity of LFA-1 for ICAM-1 (Supplementary Fig. S1A). When expression of ICAM-1 on RencaHA cells is too low to maintain a stabilized interaction between CL4 cells and RencaHA cells (Supplementary Fig. S1B), increased cAMP levels within CL4 cells will interfere with Ca²⁺ influx, resulting in a reduction in the affinity of LFA-1 for ICAM-1. Eventually, interactions between the CL4 and RencaHA cells are terminated and they disassociate without any CL4 proliferation and/or IFNγ production. However, production of IFNγ in the vicinity of CL4 cells and RencaHA cells coming into contact with one and another, increases ICAM-1 expression, thus promoting further interactions between LFA-1 and ICAM-1, the formation of the synapse, and stabilized binding between these cell types. In contrast to signals mediated through CD28 ligation, the signals associated with LFA-1–ICAM-1 interactions appear to be essential for antitumor-specific CD8⁺ T cells to overcome the inhibitory effects of PGE₂ and generate mature CTLs. When expression of ICAM-1 on RencaHA cells is high (Supplementary Fig. S1C), the increased affinity of LFA-1 for ICAM-1 serves to maintain the contact between CL4 and RencaHA cells, enabling the formation of a stable synapse to be formed between the two cell types. Following synapse formation, the signals provided by both TCR–K^dHA and LFA-1–ICAM-1 interactions further increase Ca²⁺ influx, inducing a high-affinity state of LFA-1 for interaction with ICAM-1. In this situation, the inhibitory signals on Ca²⁺ influx in CL4 cells, induced by PGE₂-dependent increases in cAMP levels, cannot override the stimulatory signals maintained by stable TCR–K^d/HA and LFA-1–ICAM-1 interactions, such that the net result is the expansion of antitumor CTLs.

In conclusion, the data presented in this report support the thesis that overproduction of PGE₂ by tumor cells that have metastasized to the TDLN favors tumor progression in the presence of an otherwise competent immune system by preventing productive activation of tumor-specific CTL responses within the TDLN. Thus, the use of PGE₂-specific inhibitors to reduce or inhibit PGE₂ production at the tumor site could promote antitumor-specific CTL responses from both naïve and preactivated CD8⁺ T cells. Our data also show that IFNγ-dependent upregulation of ICAM-1 expression by tumor cells protects tumor-specific CTLs from the inhibitory effect of PGE₂, by sustaining CD8⁺ T-cell activation, proliferation, and induction of CTL effector function. Therefore, drugs that can increase cell-surface expression of ICAM-1 by tumor cells could provide us with a powerful immunotherapeutic tool to counteract the CTL-inhibitory action of tumor-derived PGE₂ and ultimately control tumor growth.

No potential conflicts of interest were disclosed.

Conception and design: F.S. Basingab, M. Ahmadi, D.J. Morgan

Development of methodology: F.S. Basingab, M. Ahmadi, D.J. Morgan

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): F.S. Basingab, M. Ahmadi, D.J. Morgan

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): F.S. Basingab, M. Ahmadi, D.J. Morgan

Writing, review, and/or revision of the manuscript: F.S. Basingab, M. Ahmadi, D.J. Morgan

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): D.J. Morgan

Study supervision: D.J. Morgan

D.J. Morgan was supported in part by Cancer Research UK project grant C1484. F.S. Basingab is the recipient of a scholarship from King Abdulaziz University, Jeddah, Kingdom of Saudi Arabia.

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