Correspondence: Jack D. Bui, Department of Pathology, University of California, San Diego, 9500 Gilman Drive, La Jolla, CA 92093-0612, USA. Email: firstname.lastname@example.org
Senior author: Jack D. Bui
NKG2D ligands are cell surface proteins that activate NKG2D, a receptor used by natural killer (NK) cells to detect virus-infected and transformed cells. When tumour cells express high levels of NKG2D ligands, they are rejected by the immune system. Hence, reagents that increase NKG2D ligand expression on tumour cells can be important for tumour immunotherapy. To identify genes that regulate the NKG2D ligand H60a, we performed a microarray analysis of 3′-methylcholanthrene-induced sarcoma cell lines expressing high versus low H60a levels. A20, an inhibitor of nuclear factor-κB (NF-κB) activation, was differentially expressed in H60a-hi sarcoma cells. Correspondingly, treatment of tumour cells with inhibitors of NF-κB activation, such as sulfasalazine (slz), BAY-11-7085, or a non-phosphorylatable IκB, led to increased levels of H60a protein, whereas transduction of cells with an active form of IκB kinase-β (IKKβ) led to decreased levels of H60a. The regulation probably occurred at the transcriptional level, because NF-κB pathway inhibition led to increased H60a transcripts and promoter activity. Moreover, treatment of tumour cells with slz enhanced their killing by NK cells in vitro, suggesting that NF-κB inhibition can lead to tumour cell rejection. Indeed, when we blocked the NF-κB pathway specifically in tumour cells, there was decreased tumour growth in wild-type but not immune-deficient mice. Our results suggest that reagents that can block NF-κB activity specifically in the tumour and not the host immune cells would be efficacious for tumour therapy.
Cancer immuno-editing describes the process whereby the immune system can interact with developing tumour cells, recognize and destroy especially immunogenic cell clones, and sculpt the tumour cell repertoire to produce clinically evident tumours that display low immunogenicity.[1-3] Hence, one strategy for immune therapy involves increasing the expression of tumour recognition ligands that may be edited, such as NKG2D ligands, to restore immune recognition.[4-8] NKG2D ligands activate NKG2D,[9-11] a recognition receptor on natural killer (NK) cells that is known to participate in tumour surveillance. Indeed, mice that are deficient in NKG2D activity are more susceptible to primary tumour formation,[12, 13] whereas enforced expression of NKG2D ligands in tumour cells leads to their rejection via a mechanism that requires NK cells.[14, 15] Moreover, in humans, soluble inhibitory forms of NKG2D ligands can be detected in the serum of certain cancer patients and can lead to defective immune responses. These studies highlight the potential significance of NKG2D ligands as targets for tumour immune therapy.
H60a (minor histocompatibility antigen 60a) is an NKG2D ligand that binds to NKG2D with high affinity and activates NK cell lysis. We recently found that H60a can act as a tumour antigen in allogeneic but not syngeneic systems, hence activating H60a-specific CD8+ T cells. As a target of both T and NK cells, H60a undergoes cancer immuno-editing such that tumours that arise in immune-deficient mice generally have higher levels of H60a than tumours that arise in wild-type mice. It is not known how tumour cells decrease their level of H60a to evade immune recognition. We recently found that interferon-γ inhibits H60a through a signal transducer and activator of transcription 1-dependent mechanism, but it is unlikely that tumour cells use this mechanism to escape immune recognition because interferons are known to promote immune recognition and tumour surveillance. Hence, understanding how the basal level of H60a on tumour cells is controlled can lead to strategies that boost H60a expression to mediate tumour rejection.
To discover global gene expression pathways that control the basal level of expression of H60a on tumour cells, we used a panel of 3′-methylcholanthrene (MCA) -induced sarcoma cell lines that express varying levels of H60a. We identified the nuclear factor-κB (NF-κB) pathway as a regulator of H60a levels in tumour cells. Importantly, inhibition of NF-κB in tumour cells led to increased NK cell recognition in vitro and decreased tumour growth in vivo. Our results suggest that a mechanism by which NF-κB inhibitors may be useful anti-cancer weapons[21, 22] could involve increasing tumour immunogenicity.
Materials and methods
Cell lines and microarray data
The MCA-induced sarcoma cell lines on the 129/SvEv background and microarray data were a gift from Dr Robert Schreiber. The MCA-induced sarcoma cell lines on the C57BL/6 background were generated as described eslewhere and the cell lines were isolated and passaged in vitro as previously described.[3, 24] Cell lines were maintained in RPMI-1640 supplemented with 10% fetal bovine serum (FBS), l-glutamine, non-essential amino acids, sodium pyruvate, sodium bicarbonate, penicillin/streptomycin and β-mercaptoethanol. Microarray data from the cell lines were collected by Dr Hiroaki Ikeda as described (O'Sullivan et al., manuscript submitted). The following cell lines from 129/Sv mice are used (categorized by high, moderate/hi, moderate, moderate/lo, and lo H60a expression): H118-H60ahi; d30m4-H60amod/hi; F244-H60amod/hi; F515-H60amod/hi; d4m3-H60amod; F279-H60amod-lo; d22m1-H60amod-lo; F236-H60alo; d53m-H60alo; d42m1-H60alo; H31m1-H60alo; F221-H60alo. Note that F279 and d22m1 were not included in the microarray analysis because of moderate/lo expression of H60a. The following cell lines from C57BL/6 mice do not express H60a as a result of a deletion in the gene: 9609, 6727. These cells can respond to H60a regulators because the H60a promoter region is intact.
Detection of H60a by flow cytometry or quantitative RT-PCR
Cell lines were administered 1–2 mm sulfasalazine (slz; Sigma, St Louis, MO) or control DMSO overnight and harvested without trypsin using PBS with 2·5 mM EDTA. For flow cytometry, the cells were stained with a monoclonal antibody to H60a from R&D (Minneapolis, MN) and detected using a secondary antibody from Biolegend (San Diego, CA). Staining was conducted for 15–30 min at 4° in FACS tubes containing 0·5–2 million total cells, 0·5–1 μl antibody and 100 μl FACS buffer (PBS + 1% FBS + 0·09% NaN3; Sigma). All analyses were performed on live cells identified by forward and side scatter properties and 7-amino actinomycin D (7-AAD) on a BD FACSCanto. For measurement of transcript, RNA was generated using Trizol Reagent (Invitrogen, San Diego, CA). cDNA was made using the Applied Biosystems (Foster City, CA) protocol. Real-time Taqman PCR reactions (Applied Biosystems) were performed using the following primers: H60a forward, 5′-GAG CCA CCA GCA AGA GCA A; H60a reverse, 5′-CCA GTA TGG TCC CCA GAT AGC T; H60a probe VIC-5′-TTG CCT GAT TCT GAG CCT TTT CAT TCT GCT-TAMRA; glyceraldehdye 3-phosphate dehydrogenase (GAPDH) forward, 5′-CTT AGC ACC CCT GGC CAA G; GAPDH reverse, 5′-TGG TCA TGA GTC CTT CCA CG; GAPDH probe, VIC-5′-CAT CCA TGA CCA CCC CTG GCC AAG-MGB. The H60a primers detect H60a transcripts from both 129/SvEv and C57BL/6 strains of mice.
Transfection of IKKβ-EE mutant and IκB-SR
For transient transfections, control plasmid or plasmid containing the IκB kinase-β-EE (IKKβ-EE) mutant, which displays constitutive activity and leads to sustained activation of NF-κB, were transfected into the F244 cell line using lipofectamine (Invitrogen). A reporter plasmid expressing DsRed fluorescent protein (Clontech Laboratories, Mountain View, CA) was co-transfected to identify the transfected cells. Transfection efficiency was 5–20% based on visualization of DsRed cells. Mock transfection without DsRed did not cause cells to become fluorescent. After 2–3 days, cells were stained for expression of H60a, and data shown are gated on DsRed-positive cells. For production of a stable line with inhibited NF-κB activity, a plasmid containing an unphosphorylatable IκB ‘super repressor’ (IκB-SR) and a puromycin selection marker was transduced into the F244 cell line and selected at 10 µg/ml puromycin. A stable line emerged after 10 days of selection, designated as F244.SR.
The H60a promoter region containing 527 bp of sequence upstream of the transcriptional start was subcloned into a luciferase reporter plasmid (PGL3-basic, Promega, Madison, WI) as described. The NF-κB luciferase reporter plasmid pNF-κB-Luc (which contains two response elements to NF-κB) was obtained from Stratagene (La Jolla, CA). Transfections were normalized using Renilla Luciferase (PRL-TK; Promega). Transfection was carried out through Lipofectamine 2000 (Invitrogen) in triplicate wells in a 48-well plate. All experiments were performed at least twice.
Flow-based killing assay
Natural killer cell cytotoxic activity was assessed using flow cytometry and 7-AAD to detect dead target cells as described previously. Briefly, the F236 target cell line was given DMSO or slz for 1–2 days, and cells were stained with 1 μm carboxyfluorescein succinimidyl ester (CFSE; Molecular Probes, San Diego, CA). Then, 1 × 105 labelled target cells were seeded into a 96-well plate or FACS tube in complete RPMI-1640 medium with 10% FBS and 100 units/ml interleukin-2 (Peprotech, Rocky Hill, NJ). Freshly purified NK effector cells (from a RAG2−/− spleen) were added at various effector : target ratios and incubated for 5 hr at 37° with 5% CO2. As controls, effector and target cells were cultured alone or in the presence of ionomycin and/or 2·5 μm ethylene glycol tetraacetic acid (EGTA). Cells were stained with 7-AAD and acquired on a FACSCanto II. Percentage cell death was assessed by measuring the percentage of CFSE target cells that were 7-AAD positive.
Mice and tumour transplantation
The F244 MCA-induced sarcoma cell line (originally derived from 129/SvEv mice) or F244.SR were transplanted into wild-type or immune-deficient RAG2−/− mice (Taconic Farms, Germantown, NY). Wild-type hosts were F1 (C57BL/6×129) mice generated by breeding 129/SvEv mice (Taconic Farms) with C57BL/6 mice (Charles Rivers, San Diego, CA). We use F1 mice as hosts for both 129/SvEv and C57BL/6 tumour cell lines to ensure that there was no rejection due to strain differences between tumour cells and the host. Tumour cell lines were transplanted into recipient naive mice as described. The mice were monitored for tumour growth. All animal procedures were approved by the UCSD IACUC under protocol #S06201.
Identification of the NF-κB pathway as a putative regulator of H60a expression
Previously, we found that MCA-induced sarcomas expressed a wide range of H60a on the cell surface. For example, the F244 and d30m4 cell lines expressed a 1–2 log shift in H60a staining compared with isotype control, whereas the F236 and d42m1 cell lines displayed minimal H60a expression, as detected by flow cytometry (Fig. 1a). To determine global gene expression signatures that may regulate H60a expression on tumours, we examined the gene expression profiles of H60a-hi versus H60a-lo cell lines (Fig. 1b,c). Nine genes were expressed twofold higher (P < 0·05) in the H60a-hi versus H60a-lo cell lines (Supplementary material, Figure S1). One of these genes was A20 (tumour necrosis factor-α-induced protein 3), an inhibitor of the NF-κB pathway.[29, 30] A20 was expressed more highly in some H60a-hi cell lines but not in others and was poorly expressed in most H60a-lo cell lines (Supplementary material, Figure S2), suggesting that other factors besides NF-κB could regulate H60a levels. Notably, using a transcription factor binding site prediction program found a consensus NF-κB binding site at position −68 of the H60a gene, suggesting that NF-κB can regulate H60a (Fig. 1d).
Sulfasalazine, an anti-inflammatory agent that inhibits the NF-κB pathway, increases H60a expression in tumour cells
Next, we hypothesized that NF-κB inhibited H60a expression, and in cells that express high levels of A20, the diminished NF-κB activity led to higher levels of H60a at the cell surface. To test this hypothesis, we blocked NF-κB activity with slz an anti-inflammatory drug used to treat patients with certain autoimmune diseases such as Crohn's disease or with inflammation-associated cancer.[31, 32] We treated tumour cells with slz or control vehicle and measured H60a expression after 1–2 days of treatment. We found that indeed H60a levels were augmented in cells treated with slz compared with vehicle control, as shown by flow cytometry histograms in the F244 and F236 cell lines (Fig. 2a). This result was reproduced and quantified in a total of four MCA-induced sarcoma cell lines, as shown in Fig. 2(b). Note that the cell lines displayed inductions of H60a from fivefold (F244, P < 0·02) to twofold (F279, P < 0·17), indicating that other pathways probably regulated slz-dependent induction of H60a. This result was also found when another inhibitor of the NF-κB pathway, BAY-11-7085, was used (Supplementary material, Figure S3). Interestingly, when primary fibroblasts were treated with slz, H60a was not up-regulated (Fig. 2c–d), suggesting that slz could be used therapeutically to induce immune recognition of tumour cells without causing attack on normal cells.
Activation of NF-κB by a constitutively active form of IKKβ decreases H60a levels whereas blockade using a non-phosphorylatable IκB mutant increased H60a levels
Next we sought to confirm the regulation of H60a by the NF-κB pathway using genetic gain- and loss-of-function assays. Previously, we found that tumour necrosis factor (TNF), an activator of NF-κB, did not change H60a protein levels after 1–2 days of treatment. We confirmed that TNF could not reduce H60a by measuring transcript levels and also examining early time-points. Supplementary material, Figure S4 shows that TNF did not significantly change H60a transcript levels. In contrast, similar to previous studies, TNF induced the NF-κB target gene A20 within 30 min, suggesting that TNF-induced NF-κB activity may be transient because of feedback inhibition by induction of A20 (Supplementary material, Figure S4). Hence, we used a genetic method to provide sustained NF-κB activation by transfecting IKKβ-EE, a constitutively active mutant form of IKKβ. When the F244 cell line was transduced with a vector encoding the IKKβ gene, H60a was reduced compared with control transduction (Fig. 3a,c). To reduce NF-κB activity genetically, we used a non-phosphorylatable IκB mutant (IκB.SR). IκB.SR anchors NF-κB in the cytoplasm and prevents it from translocating to the nucleus, even when IKK is activated, so functioning as a ‘super-repressor’. Indeed, when IκB.SR was transduced into the F244 cell line, there was a slight increase in H60a levels (Fig. 3b,d). These results were confirmed with another cell line and quantified in Fig. 3(e). Whereas the change in H60a expression compared with control plasmid was not significant, there was a significant difference in the F244 cell line when comparing an activating (F244-EE) versus inhibitory (F244-SR) signal (P < 0·03).
Induction of H60a transcripts by slz
Next, we examine whether NF-κB repression of H60a occurred via its repression of H60a transcripts. Towards this end, we used MCA-induced sarcoma cell lines from 129/SvEv mice (F244, F236), which possessed the intact H60a gene, or from C57BL/6 mice (9609, 6727). The H60a gene in the C57BL/6 background has a fully functional promoter but lacks much of the coding and 3′ untranslated region, so allowing us to test whether NF-κB acted on the promoter or 3′ untranslated region. We treated various MCA-induced sarcoma cell lines with slz or control vehicle and measured H60a transcripts the next day. In all cell lines, we found induction of H60a transcripts from 2× to 15× compared with vehicle treatment (Fig. 4a). The induction occurred in both 129/SvEv and C57BL/6 cell lines, indicating that the H60a downstream regulatory regions, including the 3′ untranslated region, were not involved in slz-dependent regulation of H60a transcripts.
Regulation of the H60a promoter by NF-κB
We next used conventional luciferase reporter assays to confirm the transcriptional regulation of H60a by the NF-κB pathway. We generated a luciferase reporter construct containing the H60a promoter region and measured luciferase activity from this plasmid in cells co-transfected with control constructs or constructs containing the NF-κB activator (IKKβ-EE) or inhibitor (IκB.SR). Figure 4(b) is a control experiment showing that plasmids containing an activator or inhibitor of the NF-κB pathway could indeed activate or inhibit luciferase activity, respectively, when an artificial promoter containing NF-κB response elements was used. Notably, when a fragment of the H60a promoter was used to express luciferase, the opposite results were seen (Fig. 4c). Genetic inhibition of the NF-κB pathway led to increased H60a promoter activity, whereas activation of the pathway led to inhibition of H60a promoter activity.
Induction of H60a by slz enhances NK cell killing
Having shown that slz could induce H60a expression, we next examined whether it could augment killing by NK cells. We treated the H60a-lo cell line F236 with slz or vehicle control and used these cells as targets in a flow-based NK killing assay. Using fresh NK cells as effectors, we found that slz increased the killing of F236 cells twofold (Fig. 5), confirming that the quantity of up-regulation of H60a by slz has physiological significance.
Blocking NF-κB in the tumour cells leads to immune rejection
Next we tested whether inhibition of NF-κB in vivo could lead to tumour rejection. To inhibit NF-κB specifically in tumour cells, we generated F244.SR, a stable cell line of F244 that expressed IκB.SR. Notably, we found that F244.SR was rejected when transplanted into wild-type mice compared with the control parental F244 cell line (Fig. 6a). This rejection required the adaptive immune system, because the two cell lines had similar growth kinetics in immune-deficient RAG2−/− mice (Fig. 6b). These results were also seen with another cell line (data not shown).
The NKG2D ligands function to alert the immune system to cellular stress from viral infection and transformation. They display complex regulatory mechanisms,[5, 33] including transcriptional regulation,[34, 35] post-transcriptional regulation involving microRNAs[36, 37] and post-translational regulation involving cleavage, ubiquitination and intracellular retention.[39, 40] H60a is an NKG2D ligand that has one of the highest affinities for murine NKG2D[41, 42] and whose expression on tumour cells is sufficient to induce rejection. Hence, tumour cells actively modulate H60a expression to escape immune recognition. Although H60a can be induced by carcinogen and virus infection, surprisingly little is known about its regulation, because the H60a gene in C57BL/6 mice (which are widely studied in cancer and virus infection models) is not functional because it lacks exons 3, 4 and 5.[17, 25, 33] We have found that H60a is down-regulated by interferons and microRNAs acting on the H60a 3′ untranslated region, but the signals that up-regulate H60a and/or maintain its constitutive expression on tumour cells have not been defined.
In this study, we document that the NF-κB pathway can repress H60a expression, presumably by binding to a canonical NF-κB cis element in the H60a promoter region. We used pharmacological and genetic manipulators of NF-κB activity to show this regulation, which can be demonstrated at the level of H60a protein, transcript and promoter activity. We have not shown that there is direct binding of the H60a promoter by NF-κB so it is possible that the reagents we have used may regulate H60a indirectly. Future studies will address this possibility. Nevertheless, the combination of pharmacological and genetic manipulation of the NF-κB pathway in these studies definitively show that this pathway can regulate H60a expression in tumours, highlighting a potential avenue to regulate tumour immunogenicity. Another group has shown that the NF-κB pathway actually up-regulates the human NKG2D ligand MHC class I chain-like gene A (MICA). They found that primary human T cells activated by signals through CD3 and CD28 induce MICA expression, and this induction was blocked by slz. Moreover, NF-κB proteins were shown to bind to MICA regulatory cis elements, and slz could inhibit expression of MICA on the tumour cell line HeLa. Our preliminary studies have confirmed that slz indeed up-regulates MICA on some but not all human tumour cell lines that we tested (unpublished observations). It is not known why this pathway would up-regulate certain NKG2D ligands and reduce others. Nevertheless, it is clear that NF-κB bindings sites are heavily conserved in the regulatory regions of NKG2D ligand genes, suggesting that this pathway is a bona fide regulator of NKG2D ligand expression.
As NF-κB activation is self-limited, complex and involves transcriptional repression and activation,[46, 47] we speculate that its regulation of NKG2D ligands can vary depending on the ligand, the tissue site, or the type/level of transformation (primary cell versus virus-transformed versus carcinogen-transformed). It should be noted that HeLa cells (ATCC number CCL-2.2), which require NF-κB for expression of MICA, possess integrated viral genes whereas in our study, the carcinogen MCA is used to induce transformation. We conjecture that NF-κB may induce NKG2D ligands in virus infection or virus-associated cancers but may inhibit it in carcinogen-induced tumours.
We found that slz does not influence the expression of H60a in primary fibroblasts, even though H60a is expressed to a high level in these cells. The expression of NKG2D ligands in primary tissues is documented. For example, H60c can be induced in primary keratinocytes via ‘culture shock’, but the pathways that regulate this are not known. Our results suggest that in mesenchymal tissue, the regulation of H60a by NF-κB is complex and may require a co-stimulus that is acquired during transformation. It should be noted that even in transformed cells, the induction of H60a protein by slz ranges from twofold to fivefold, indicating that other pathways could influence the NF-κB-dependent repression of H60a. These unknown pathways may have differing levels of activities in different tumour cells, thereby influencing the degree of up-regulation of H60a by slz. This model is consistent with the idea that the many pathways that are perturbed during transformation have different capabilities of regulating NKG2D ligands, which explains the diverse array of NKG2D ligands that may be expressed on tumour cell lines.
When we blocked NF-κB activity specifically in tumour cells using an unphosphorylatable IκB mutant, we found that there was tumour rejection. This rejection was seen in wild-type mice but not in immune-deficient RAG2−/− mice, indicating that adaptive immunity was required for complete rejection. We conclude that the decreased tumour growth upon inhibition of NF-κB is not simply a result of lack of survival, because the cells with and without NF-κB blockade grew equivalently in immune-deficient mice (Fig. 6). As such, we suggest that the NF-κB pathway can control tumour immunogenicity to some extent. Although we suggest that the enhanced immunogenicity is a result of H60a expression, we have not shown this, and believe that multiple pathways are responsible for the increased immunogenicity of tumour cells expressing IκB.SR. One intriguing possibility is that IκB.SR might induce tumour rejection/death via an immunogenic cell death pathway, which is known to boost adaptive immune responses. Indeed, a recent study has found that calreticulin exposure could be regulated by induction of NF-κB pathways.
Targeting NF-κB for cancer therapy has been extensively reviewed.[53-55] Reviews have described the possible beneficial effect of NF-κB blockade in cancer therapy, including promoting tumour cell apoptosis, blockade of cell cycling, diminishing cell survival, blocking angiogenesis and invasion, suppressing immune inflammatory cytokines, and reducing genomic stress from free radicals.[53-55] Notably, despite these extensive reviews, it has not been emphasized that NF-κB blockade can promote tumour cell immunogenicity. Our finding that NF-κB can suppress tumour cell immunogenicity therefore provides another key mechanism of action for the in vivo effects of NF-κB inhibition. Moreover, our tumour-specific blockade of NF-κB corroborates the suggestions from these reviews that global blockade of NF-κB in vivo could potentially have the untoward effect of promoting tumour formation by blocking tumour immune surveillance. Hence, our study emphasizes the need to block NF-κB activity in specific cell types and with defined kinetics.
Our results are consistent with a previous study in which the mutant IκB super-repressor transfected into mammary and colon carcinoma cell lines led to their regression in syngeneic mice. In this study, the rejection was initiated by injection of lipopolysaccharide into the mice to activate the immune response. In our study, the mice were naive and not treated with any form of immune therapy. In conclusion, our findings suggest that blocking NF-κB activity in tumour cells may be an effective means to enhance their immunogenicity. We suggest that to achieve long-term remission in cancer patients, this form of therapy should be combined with conventional chemotherapy or other forms of immune therapy to maximize anti-tumour immune responses.
We thank Robert Schreiber for the MCA-induced sarcoma cell lines from 129/Sv mice and the microarray data. We thank Michael Karin for providing us with the IKKβ-EE mutant and Alex Hoffmann for providing us with the IκB-super repressor. This work was supported by grants to J.D.B. from the American Cancer Society (ACS-IRG #70-002), a grant from the Cancer Research Coordinating Committee (6-444951-34384), the V Foundation Scholar Award, the Concern Foundation, the Hartwell Foundation, and NIH-CA128893 and CA157885.
The authors have no conflicts of interest to disclose.