ICAM-1 overexpression counteracts immune-suppress cell-derived PGE2 to restore CTL function. Journal of Immunological

Tumour-infiltrating cytotoxic T lymphocytes (CTLs) play a key role in tumour killing. However, many cancers adopt various strategies to induce immunosuppression. Priming of naïve CD8 + T cells to become CTLs occurs via cognate interactions of the T cell receptor (TcR) and CD28 with tumour-derived peptide epitopes expressed on major histocompatibility complex (MHC) class I molecules and CD80/CD86 on T cells and antigen-presenting cells (APCs) respectively. Here we report that, in the absence of CD80/CD86 expression by renal carcinoma (Renca) cells, expression of intercellular adhesion molecule-1 (ICAM-1) by Renca cells provides a potent alternative co-stimulation to a tumour-specific CD8 + T cells causing them to produce interferon gamma (IFN-γ) which is crucial for the further up-regulation of ICAM-1 on tumour cells. We have shown that overexpression of cyclooxygenase-2 (COX-2), by Renca cells (Renca-T3), results in increased levels of prostaglandin (PG) E 2 production, which can directly suppress anti-tumour CD8 + T cells resulting in loss of CTL function in vivo and cause metastases to the tumor-draining lymph nodes (TDLNs). Significantly, our data also show that overexpression of ICAM-1 on Renca-T3 cells can counteract the immune-suppressive effect of PGE 2 and restore CTL responses.


Introduction
Anti-tumour CD8 + cytotoxic T lymphocytes (CTLs) are capable of producing interferon gamma (IFN-γ) and cytolytic enzymes and are crucial for cancer regression 1 . Priming naïve CD8 + T cells to become CTLs requires two signals; the first comes from T cell receptor (TcR) interaction with peptide epitope presented by major histocompatibility complex (MHC-I) molecules on antigenpresenting cells (APCs). The second signal was long thought to be solely provided by the interaction of T cell-expressed CD28 with the classical co-stimulatory CD80/CD86 molecules on APCs. However, CD80-and CD86-mediated co-stimulation is not provided by most tumour cells. Thus, the direct interaction between naïve CD8 + T cells and tumour cells often leads to tolerance induction. In the absence of CD80/CD86 2 , intercellular adhesion molecule-1 (ICAM-1) interaction with lymphocyte function associated antigen-1 (LFA-1), expressed by T cells, can provide a co-stimulatory signal resulting in the formation of CTLs 3,4 .
ICAM-1 is a member of immunoglobulin superfamily expressed by many cells including various tumour cells 5 . The main role of ICAM-1 is to firmly arrest leukocytes and facilitate extravasation through the blood vessels to sites of inflammation. However, ICAM-1 also enables T cells to adhere to other cells including APC, endothelial cells, and to normal and tumour cells 6 . In some cancer patients elevated levels of soluble (s) ICAM-1 of around 3 to 5-fold are often found. However, Circulating sICAM-1 is shown to block LFA-1; thus promoting tumour growth and angiogenesis 7 . In contrast, ICAM-1 and/or ICAM-2 negative pancreatic cancer cells are resistant to killing by γδ T cells. However, such resistance to killing can be reversed by re-expressing ICAM-1 or ICAM-2 8 .
Despite the induction of tumour-specific CTL responses by alternative co-stimulation pathways, tumours still develop various strategies to escape antitumour immune responses, such as the over-production of prostaglandin (PG) E 2 as a result of the up-regulation of the cyclooxygenase-2 (COX-2) gene 9,10 . PGE 2 is a known immune modulator which is able to maintain dendritic cells (DC) in an immature state by: increasing the levels of IL-10 within the tumour microenvironment; enhancing Th17 T cell responses, and recruiting myeloid-derived suppressor cells (MDSC) 11 . In addition, coupling PGE 2 with cognate receptors on T cells can increase cyclic adenosine-monophosphate (cAMP) within T cells which, consequently, reduces T cell proliferation and effector function 12 . Moreover, PGE 2 decreases IFN-γ production by T cells. IFN-γ is shown to be involved in enhancing tumour immunogenicity. PGE 2 can indirectly affect the expression of ICAM-1 by preventing IFN-γ mediated up-regulation of ICAM-1. Therefore, we hypothesized that over-expressing ICAM-1 on COX-2-overexpressing Renca-T3 cells would counteract the immunosuppressive effect of PGE 2 on tumour-specific CD8 + T cells.
To test this hypothesis, and to explore the efficacy of ICAM-1 in providing co-stimulatory signals to tumourspecific CD8 + T cells, our lab adapted a well-established murine renal carcinoma (Renca) model to generate cells that not only express the haemagglutinin (HA) protein from influenza virus A/PR/8/H1N1 (PR8) as a neotumour-specific antigen (Renca-HA; 3 ), but which also overexpress COX-2, resulting in the over-production of PGE 2 (Renca-T3; 12 ).

Materials and methods
Mice 6 to 8 wk old Thy1.1 +/+ CL4 +/-TcR transgenic BALB/c mice (CL4 mice; 13 ) were bred and housed under specific pathogen-free conditions at the University of Bristol Animal Services Unit. All experimental procedures were conducted in accordance with U.K. Home Office guidelines.

Enrichment and proliferation of CL4 CD8 + T cells
Naïve CL4 CD8 + T cells were purified from peripheral lymphoid tissues isolated from CL4 TcR transgenic mice by magnetic-activated cell sorting (MACS) according to the manufacturer's instructions (Miltenyi Biotec Ltd., Bisley, UK). In some instances, naïve CL4 T cells were labeled with 5 µM carboxyfluorescein succinimidyl ester (CFSE; BioLegend, San Diego, USA). MACS-purified CL4 T cell proliferation was detected either; by a 3 H-thymidine-based proliferation assay, whereby 1 µCi 3 H-thymidine was added in the last 8 hours and the proliferation was measured by thymidine incorporation (counts per minute; cmp) (Amershan Life Science, London, UK), or by a CFSE-based proliferation assay whereby CL4 proliferation was detected by the loss of CFSE expression. Naïve CL4 T cells were activated with; 10 μg/ml of anti-CD3 mAbs plus either; 5 μg/ml of anti-CD28 mAbs, or 3 μg/ml of recombinant (r) ICAM-1, or populations of γ-irradiated Renca cells. Renca cells were irradiated with 9600 RADs using a Cs 137 source of γ-irradiation (RX30/55; Gravatom Projects, Grosport, UK).

Flow cytometry
Activated CL4 T cells were either left untreated, or stimulated with 1 µg/ml K d HA peptide and 1 µg/ ml GolgiPlug (BD Bioscience, San Diego, USA), before staining with live/dead aqua and fluorochrome-conjugated monoclonal antibodies (mAbs) for surface staining or they were first permeabilized and then stained intracellularly for IFN-γ.

Statistical analyses
P values were calculated by one-way ANOVA followed by either Dunnett's or Bonferroni's multiple comparison tests using the Prism 5.03 software (GraphPad Software, Inc.).

Characterization of ICAM-1 and LFA-1 expression by naïve CL4 CD8 + T cells
To investigate the role of CD8 + T cells in anti-tumour immunity, our lab utilized CL4 TcR transgenic mice in which virtually all the CD8 + T cells express the Vα10/Vβ8.2 TcR transgene 13 . These CL4 T cells are able to recognize the dominant K d -restricted epitope of the Influenza virus A/PR/8 haemagglutinin (HA) protein (K d HA [IYSTVASSL] ) expressed by HA-transfected Renca-HA cells 3 . Naïve CL4 CD8 + T cells were purified from the total cells harvested from lymphoid tissues of CL4 mice using the MACS purification technique. This process routinely results in >98% pure CD8 + T cells ( Figure 1A). ICAM-1 has a central role in naïve CD8 + T cell priming 3 , and acts as a CD28-independent costimulator 2 . ICAM-1 interacts with LFA-1 expressed by T cells which is crucial for homotypic T cell aggregation and communication 14 . Therefore, MACS-purified naïve CL4 CD8 + T cells were stained for ICAM-1 and LFA-1 expression.

Is ICAM-1 a potent alternative co-stimulatory molecule?
To evaluate the ability of Renca-HA cells, to prime naïve CL4 CD8 + T cells, both the HA low-expressing Renca-HA cells 3 , and control Renca-WT cells were used as APC to prime naïve CL4 T cells in a standard 3 H-thymidine incorporation T cell proliferation assay. Whilst naïve CL4 T cells did not proliferate in response to control irradiated Renca-WT APCs in vitro, significant levels of CL4 T cell proliferation was observed when cultured with Renca-HA cells as APC ( Figure 2A). These data, therefore, suggest that Renca-HA cells are not only able to provide CL4 T cells with signal one but also co-stimulatoty signal two. However, we know that although Renca-HA cells lack CD80/CD86; they do express ICAM-1 ( Figure 2B).
To determine if the co-stimulation in our system occurs solely via ICAM-1, anti-ICAM-1 blocking mAbs were added to the CL4 T cells/Renca-HA co-culture. In Figure 3, bar charts show that not only does blocking ICAM-1 significantly decrease CL4 T cell proliferation in vitro (A), IFN-γ production is also inhibited; as evidenced by the fact that CL4 T cells were unable to produce IFN-γ compared to the non-treated coculture ( Figure 3B: compare 2.7% with 59%). However, the data show that for both CL4 T cell proliferation in response to classical (anti-CD3 mAbs + anti-CD28 mAbs), and alternative (anti-CD3 mAbs + rICAM-1) priming, only the presence of plate-bound rICAM-1 and not soluble (s) rICAM-1 resulted in enhancement of CL4 T cell proliferation. In addition, treating naïve CL4 T cells with rICAM-1 in solution for an hour before the beginning of the culture is not sufficient to induce the proliferation to a level similar to that achieved using platebound rICAM-1 ( Figure 3C). Furthermore, there is a positive correlation between the cell surface expression of ICAM-1 on Renca cells and CL4 T cell proliferation as shown in ( Figure  3D & E); the more cell surface ICAM-1 that is expressed the more CL4 T cell proliferation is detected. Importantly, these differences in proliferation were not due and changes in the level of HA expression, as after overexpressing ICAM-1, HA expression remained the same for all clones; regardless of the levels of ICAM-1 ( Figure 3E).

Can ICAM-1 counteract the inhibitory effect of PGE 2 ?
Although the basal level of ICAM-1 expression by Renca-HA cells is low, such low level expression is enough to provide sufficient alternative co-stimulation to prime naïve CL4 CD8 + T cells in vitro. However, despite this, Renca-HA cells continue to grow in vivo. Overexpression of COX-2 by Renca-T3 cells resulted in elevated PGE 2 Co Co Vβ unts unts 8 production; which has been shown to inhibit anti-tumour responses in vitro and in vivo 12,15 . To compare the effect of PGE 2 on the proliferation and the production of IFN-γ by CD8 + T cells under classical (anti-CD28), and alternative (rICAM-1), co-stimulation pathways, CFSE-labeled CL4 T cells were primed with either; anti-CD3mAbs + anti-CD28 mAbs, or with anti-CD3 mAbs + plate-bound rICAM-1, in the presence or absence of 1 µM PGE 2 . In the absence of any PGE 2 , the FACS plots shown in Figure 4A&B; top row, reveal that around 75% of anti-CD3 mAb-treated CL4 T cells receiving alternative co-stimulation through plate-bound rICAM-1 underwent two or three rounds of division; (each peak showing successive loss of CFSE refers to one round of division). This is compared with only 55% of anti-CD3mAbs-treated CL4 T cells receiving classical costimulatory signaling through anti-CD28 mAbs. This decrease in the proliferation is also associated with a reduction in the number of IFN-γ producing CL4 T cells ( Figure 4A&B; bottom row). However, in the presence of PGE 2 , there is a decrease in IFN-γ-producing CL4 T cells of nearly 4-fold amongst those cells receiving classical anti-CD28 costimulation; compared to a 2-fold decrease in IFN-γ production among CL4 T cells receiving alternative costimulation via rICAM-1 ( Figure 4A&B; bottom row).   It has been shown that some molecules such as matrix metalloproteinase 9 (MMP-9) can cause ICAM-1 to be shed from the surface of tumour cells and is involved in tumour evasion from the immune surveillance. ICAM-1 provides a docking site for pro-MMP-9 which proteolytically cleaves the extracellular domain of ICAM-1 leading to its release from the cell surface 16 . To determine whether or not PGE 2 directly affect the priming of naïve CL4 cells by causing ICAM-1 to be shed from the surface of tumour cells, Renca-HA cells were cultured in the presence of 1µM PGE 2 for 72 hours then stained for HA and ICAM-1 expression. Data presented in Figure 5A show that Renca-HA cells maintained the same levels of both HA and ICAM-1 following PGE 2 treatment suggesting that PGE 2 does not cause modulation or shedding of these molecules from the cell surface.
IFN-γ is able to exert its anti-tumour effects by directly acting upon the tumour cells themselves 17 ; as evidenced by the fact that tumour cells lacking IFN-γR expression grow in mice despite the presence of effective anti-tumour immune responses, which would otherwise kill IFN-γR-sufficient tumour cells 18 . One mechanism through which IFN-γ is thought to reduce the growth of tumour cells in vivo is by enhancing tumour cell immunogenicity 19 . Indeed, IFN-γ has been shown to increase antigen presentation to CD8 + T cells by increasing MHC class I and ICAM-1 expression by tumour cells 3 . Similarly, IFN-γ treatment of Renca-HA cells also resulted in a marked increase in ICAM-1 expression ( Figure 5B).
To determine whether or not the up-regulation of ICAM-1 is instrumental in abrogating the effect of PGE 2 , we generated an ICAM-1 overexpressing Renca-T3 cell line (Renca-T3/ICAM-1), for co-culture with CFSE-labeled naïve CL4 T cells in vitro. The data show that co-culture with this cell line resulted in a significant increase in proliferation of CFSE-labelled naïve CL4 T cells, whereby greater than 65% of cells undergo at least three rounds of divisions compared with around 18% of CL4 T cells co-cultured in the presence of either un-transfected Renca-T3 or Renca-T3 empty-vector control cells ( Figure 5C; top row). Moreover, co-culture with Renca-T3/ICAM-1 also resulted in a large increase in CTL effector function as evidence by the fact that around 60% of CL4 cells expressed IFN-γ, compared with only 18% of CL4 cells that were co-cultured in the presence of control Renca-T3 or Renca-T3 emptyvector cells ( Figure 5C; bottom row). Taken together, these

Discussion
The requirement of ICAM-1-mediated co-stimulation in T cell priming has remained controversial. Some data clearly demonstrate that overexpression of ICAM-1 by RencaT3/ICAM-1 cells is able to circumvent the PGE 2mediated suppression of tumour-specific CTL responses.
reports suggest that ICAM-1 is not sufficient for CD4 + T cell activation. In contrast, other reports propose the essential need of ICAM-1 for T cell priming but not for cytokine secretion, whereas others suggest that ICAM-1 can stimulate both activation and cytokine secretion 20 . However, studies using ICAM-1 knockout mice suggests that ICAM-1 is not essential in T cell-mediated tumour rejection if sufficient numbers of T cells are transferred 21 . In our study, Renca-HA tumour cells, that lack expression of both CD80/CD86, provide K d HA-specific CL4 CD8 + T cells with sufficient co-stimulation through ICAM-1 interactions with T cell-expressed LFA-1 that not only results in proliferation but also in IFN-γ production. This was evidenced by the clear finding that the presence of both anti-ICAM-1 and anti-LFA-1 mAbs during priming significantly reduces CL4 T cell proliferation and IFN-γ production 3 . Although the interaction of LFA-1 on T cells with soluble or plate-bound rICAM-1 can prime naive CD8 + T cells, only plate-bound rICAM-1 enhances markedly the CL4 T cell response. This finding is consistent with a study in which T cells from mice lacking full-length cell-surface expression of either cell surface ICAM-1 or LFA-1 had significantly impaired T cell function 22 . It also supports the findings from another study in which over-expression of soluble rICAM-1 blocked LFA-1 on T cells preventing subsequent T cell activation due to the inability of LFA-1 to bind cell-surface ICAM-1. We have shown that overexpressing cell surface ICAM-1 on Renca-HA cells increased CL4 T cell proliferation and IFN-γ. This finding correlates with other studies in which the up-regulation of ICAM-1 on melanoma cells induced by retinoic acid, or ICAM-1 gene transfection, was found to improve the susceptibility of melanoma cells to lysis by lymphokine activated killer cells (LAK) 23 . Nevertheless, Renca-HA tumour cells continue to grow in vivo when they are injected into BALB/c mice either alone, or with large numbers of naïve CL4 cells. (CN Janicki and DJ Morgan; unpublished data). Following adoptive transfer, CL4 T cells undergo productive activation within the TDLNs, but they lose their effector function once they reach to the tumour site due to the immunosuppression created within the tumour microenvironment 24 . Many solid tumours create an immunosuppressive microenvironment through various immune escape strategies such as the production of PGE 2 . Over-expression of COX-2 not only results in high levels of PGE 2 production, but also tumour cell metastasis by Renca-T3 cells and suppression of activation of naive CL4 CD8 + T cells. Such abortive activation is evidenced by lack of proliferation and the absence of IFN-γ production due to the maintenance of low level ICAM-1 expression 12 . Whereas, exposure to IFN-γ was shown to markedly upregulate ICAM-1. These findings correlate with other studies, which suggest that IFN-γ enhances antigen presentation by tumour cells through the increase of MHC class I expression 18 . We showed that disabling PGE 2 production by COX-2 inhibitor has shown to restore IFN-γ production by effector T cells 25 . Significantly, over expression of ICAM-1 on Renca-T3 cells restores CL4 T cell proliferation as well as IFN-γ production even in the PGE 2 -rich microenvironment. Our findings clearly show that PGE 2 is able to inhibit both the direct priming of naïve tumour-specific CD8 + T cells, as well as the effector function amongst a tumour-specific CTL, this effect is temporary and may be mitigated by increasing ICAM-1 expression on tumour cells which therefore enable the ligation of LFA-1 on CL4 with ICAM-1 on tumour cell which then allows the formation of a stable synapse between the two cell types and prolongs this interaction. Along with LFA-1/ICAM-1 binding, the interactions of TcR with K d HA further increase Ca 2+ influx and maintain LFA-1 in a high affinity state. In this state, the inhibitory effects, induced by PGE 2 -mediated up-regulation of cAMP levels, on Ca 2+ influx within CL4 T cells fail to override the stimulatory signals through stable LFA-1/ICAM-1 and TcR/K d HA interactions. As a result, tumour-specific CD8 + T cells are forced to proliferate more and produce high levels of IFN-γ and therefore may have greater potential to prevent tumour growth in vivo. These findings clearly indicate that drugs which can increase the cell-surface expression of ICAM-1 by tumor cells could provide us with a powerful immune-therapeutic tool to counteract the immunosuppressive action of tumorderived PGE 2 to greatly enhance anti-tumor CTL responses, which may ultimately control tumor growth.