IRF1 and NF-kB Restore MHC Class I-Restricted Tumor Antigen Processing and Presentation to Cytotoxic T Cells in Aggressive Neuroblastoma

Neuroblastoma (NB), the most common solid extracranial cancer of childhood, displays a remarkable low expression of Major Histocompatibility Complex class I (MHC-I) and Antigen Processing Machinery (APM) molecules, including Endoplasmic Reticulum (ER) Aminopeptidases, and poorly presents tumor antigens to Cytotoxic T Lymphocytes (CTL). We have previously shown that this is due to low expression of the transcription factor NF-kB p65. Herein, we show that not only NF-kB p65, but also the Interferon Regulatory Factor 1 (IRF1) and certain APM components are low in a subset of NB cell lines with aggressive features. Whereas single transfection with either IRF1, or NF-kB p65 is ineffective, co-transfection results in strong synergy and substantial reversion of the MHC-I/APM-low phenotype in all NB cell lines tested. Accordingly, linked immunohistochemistry expression patterns between nuclear IRF1 and p65 on the one hand, and MHC-I on the other hand, were observed in vivo. Absence and presence of the three molecules neatly segregated between high-grade and low-grade NB, respectively. Finally, APM reconstitution by double IRF1/p65 transfection rendered a NB cell line susceptible to killing by anti MAGE-A3 CTLs, lytic efficiency comparable to those seen upon IFN-γ treatment. This is the first demonstration that a complex immune escape phenotype can be rescued by reconstitution of a limited number of master regulatory genes. These findings provide molecular insight into defective MHC-I expression in NB cells and provide the rational for T cell-based immunotherapy in NB variants refractory to conventional therapy.


Introduction
Neuroblastoma (NB) is a childhood solid tumor derived from precursor cells of the sympathetic nervous system which accounts for 15% of pediatric cancer deaths [1]. Clinical features of the tumor are very heterogeneous, ranging from spontaneous regression to rapid progression. The latter has been shown to be associated with several genomic and functional abnormalities, including MYCN gene amplification and overexpression, allelic loss of chromosome 1p, 3p, 11q and 14q, gains of 1q and 17q, as well as deficient MHC class I (MHC-I) expression and antigen presentation [2][3][4][5][6].
IFN-c has been demonstrated to enhance and reconstitute the MHC-I antigen processing and presentation pathway [25]. Unfortunately, the immunotherapeutic use of IFNs in human solid tumors has met limited success [26,27]. In fact, IFN-c induces negative feedback loops preventing long-lasting biological effects [28,29]. Adverse reactions and negative impact on tumor progression and clinical outcome of the disease have been reported upon systemic administration of this agent in clinical trials [26]. In view of these considerations, it would be desirable to identify which alternative factors are necessary and sufficient to elicit full antigen presentation in NB, without incurring the complication of in vivo treatment with a pleiotropic lymphokine.
IFN-c and TNF-a induce a number of MHC-I transcriptional activators including the interferon regulatory factors (IRF) 1 and IRF2 and the transcription factor NF-kB, respectively [30][31][32]. These, in turn, bind to adjacent cis regulatory sequences shared by MHC-I and APM molecules, such as the enhancer A element and ISRE-like sequences, respectively [33][34][35][36][37][38][39]. Recently, we have shown that NF-kB is the major, direct transacting factor responsible for coordinated regulation of MHC-I and certain APM components in NB, and that reconstitution of this missing transacting function enhances MHC-I in at least some aggressive NB cell lines from stroma-poor lesions [38].
Herein, we demonstrate that wide-range APM reconstitution requires proper selection of additional master MHC-I regulators. Specifically, IRF1 synergizes with NF-kB in reconstituting full MHC-I-restricted tumor antigen processing and antigen presentation to cytotoxic T-cell (CTL) clones. These findings pinpoint a crucial node of immunotherapeutic intervention on NB.

Tumor Cell Lines and Reagents
All human NB and melanoma cell lines were obtained from the American Type Culture Collection and characterized by morphology and HLA class I typing by PCR-SSP sets (Genovision). Cells were grown in different media: ET1 and MSR3-mel cells were maintained in IMDM, SH-SY5Y cells and SK-N-SH cells in DMEM and MEM, respectively, and the other cell lines were maintained in RPMI 1640 medium. All media were supplemented with 10% FCS (HyClone), glutamine, 100 mg/ml penicillin and 50 mg/ml streptomycin. For IFN-c-treatment, NB cell lines were cultured for 48 hours in the presence of 500 U/ml recombinant human IFN-c (R & D Systems).

DNA Constructs and Transfections
IRF1, IRF2 and NF-kB p65 (kindly provided by A. Battistini, Department of Infectious, Parasitic and Immune-Mediated Diseases, Istituto Superiore di Sanità, Rome, Italy and M. Levrero, Rome Oncogenomic Center, Regina Elena Institute, Rome, Italy, respectively) and the corresponding empty vectors were transfected using Lipofectamine 2000 (Invitrogen). Cells were seeded in tissue culture plates and transfected 24 hours later at an 80% confluence with DNA-lipofectamine complexes in Opti-MEM. Fifteen hours following transfection the culture medium was replaced with fresh medium. For IRF1 and IRF2 silencing, cells (40% confluent) were transfected with 80 nmol/L siRNAs or scrambled siRNA using Lipofectamine 2000. Cell were collected 40 hours after transfection.

Reverse Transcription and Real-time Quantitative PCR
Total RNA was extracted with Trizol Reagent (Invitrogen) and retro-transcribed with SuperScript II (Invitrogen). RT-PCR and real-time PCR analyses were performed as previously described [38,40]. 18S RNA and GAPDH were used for normalization.

Patients and Immunohistochemistry
All tumor specimens were obtained from NB patients diagnosed at the ''Bambino Gesù'' Children Hospital, Rome, (Italy), after obtaining written informed parental consent and approval by the Ethical Committee of the Institution. The histologic features of NB were classified into ganglioneuroblastoma (stroma rich, 10 cases) and undifferentiated NB (stroma poor, 10 cases) according to the percentage and degree of differentiation of the NB cells using the criteria of the International NB Pathology Classification [46]. For immunohistochemical staining, consecutive sections of paraffinembedded tissue blocks were cut at 3 mm. Deparaffinization and antigen retrieval were performed with PT-link (Dako) in Tris/ EDTA (pH 9.0) (Dako) for 15 min at 98uC for HC10 and citrate buffer (pH 6.1) (Dako) for the other Abs. Sections were incubated with the primary mAb for 45 min at room temperature, followed by incubation with the peroxidase-conjugated secondary Ab (Dako) for 20 min at room temperature. Diaminobenzidinetetrahydrochloride (DAB) was used as chromogen. All samples were counterstained with hematoxylin.

Antigen-specific T Cells Generation and Propagation
Peripheral blood mononuclear cells (PBMC) from an HLA-A1 melanoma patient were incubated with the MAGE-A3.A1 peptide EVDPIGHLY (20 mM) [47] for 60 min at room temperature. Then, cells were washed and cultured in complete medium containing 10% human serum, rhIL-2 (20 U/ml) and rhIL-7 (10 U/ml) [48]. After 10 days, the cells were stimulated weekly with autologous MAGE-A3.A1 peptide-pulsed dendritic cells (DCs) in the presence of irradiated (100 Gy) autologous EBV transformed B-cells, in complete medium containing 10% human serum and rhIL-2 (50 U/mL) [48,49]. Generation of CTLs specific to the MAGE-A3.A1 peptide was verified by a standard 5hour 51 Cr release assay and an IFN-c ELISA assay against the MAGE-A3 + HLA-A1 + melanoma cell line ET-1, MAGE-A3 + HLA-A1 2 melanoma cell line 624mel and the HLA class I negative melanoma cell line MSR-3-mel as controls. The presence of natural killer (NK) cell activity was tested by cold inhibition assay using the NK targets K562 cells. Lytic activity of the CTLs against NB cells was evaluated by 51 Cr release assay.

Statistical Analysis
Digital images of Western blots were analysed by Image J (http://rsbweb.nih.gov/ij/index.html) and statistical significance of densitometric values was assessed by the two-tailed unpaired Student's t-test. A P value of ,0.05 was considered to be statistically significant.

IRF1 and IRF2 Expression Correlates with MHC-I Expression in NB Cell Lines and Primary NB Samples
Flow cytometry and Western blotting of NB cell lines revealed that the 3 cell lines expressing higher levels of cell surface MHC-I (SH-EP, SK-N-AS and SK-N-SH) also expressed higher levels of IRF1 and IRF2, whereas the 5 cell lines expressing lower levels of cell surface MHC-I displayed lower levels of IRF1 and IRF2 (Figs. 1A, 1B and S1). Similar results were obtained in nuclear extracts ( Fig. 1B and S1) and at mRNA level (Fig. 1C).
IRF1, IRF2, and the NF-kB p65 subunit were clearly detected by immunohistochemistry in the nuclei of MHC-I-positive, mature ganglion cells, in 10/10 primary stroma-rich NB lesions. Conversely, the three transcription factors were hardly seen in the nuclei of MHC-I-negative neuroblastic cells from 10/10 stroma-poor (undifferentiated) NB samples. Representative staining (Figs. 2 and S2) reveals expression levels and subcellular localization. Thus, a significant correlation was detected, both in vitro and in vivo, between MHC-I, NF-kB, IRF1 and IRF2.

IFN-c Enhances the Expression of IRFs, MHC-I and APM Components in NB Cell Lines
NB cell lines were treated with IFN-c, a major enhancer of MHC-I, APM components, IRF1 and IRF2. Western blot analysis revealed coordinated expression and significant enhancement of IRF1, MHC-I and almost all the APM gene products tested in all the cells, although the magnitude of the enhancement tended to be greater in the 5 cell lines expressing lower baseline levels, grouped at the right-hand side of the panel (Fig. 3A, S3 and S4). These cells also displayed preferential up-regulation of IRF2, a known mediator of IFN-c signaling extinction that competitively inhibits the binding of IRF1 to the ISRE (Fig. 3A, S3 and S4). Consistent with polypeptide enhancement, cell surface MHC-I expression was strongly increased in all NB cell lines (Fig. 3B).
Thus, constitutive and IFN-c-mediated IRF1/IRF2 expression correlates with constitutive and IFN-c-mediated MHC-I and APM expression, suggesting that IRFs are involved in regulating MHC-I in NB.

Synergistic Enhancement of MHC-I and APM Component Expression by IRF1 and NF-kB in NB Cell Lines
Combined with our previous study [38], the above observations strongly indicate that not only a lack of NF-kB, but also a lack of IRF1 and/or IRF2, is responsible for the MHC-I-low NB phenotype. To obtain direct evidence for this, the three NB cell lines SH-SY5Y, SK-N-BE(2) and SK-N-BE(2)c, which are particularly depressed in MHC-I expression, were single-transfected and double-transfected with expression vectors encoding IRF1, IRF2, and the NF-kB p65 cDNAs (Fig. 4A). IMR-32 and LAN-5 could not be included in this analysis because they were refractory to transfection.
Transfection-mediated overexpression of either IRF1 or IRF2 did not affect NF-kB p65, MHC-I or APM components in any of the three tested cell lines, the only exceptions being IRF1- Remarkably, these two cell lines are MYCN amplified and representative of very aggressive NB tumors [38]. Also noteworthy is the observation that only in fully responsive SH-SY5Y cells was p65 capable of promoting substantial IRF1 enhancement, suggesting that efficient correction of APM-low phenotypes requires not only a direct effect of p65 on target genes, but also an indirect effect through IRF1    figure 3, in which IFN-c also failed to up-regulate ERAP2 in SK-N-BE(2) and SK-N-BE(2)c, and suggests a gene-specific upregulation defect in these cell lines.
In agreement with the above experiments, optimal recovery of cell surface MHC-I expression could only be observed in double p65/IRF1 transfectants (Fig. 4B), conclusively showing critical dependence of MHC-I/APM reactivation on NF-kB/IRF1 synergy, particularly in certain NB cells with hard-to-rescue APM defects.

Up-regulation of Cell Surface MHC-I Expression with IRF1 and NF-kB p65 Renders NB Cells Susceptible to Antigenspecific CTLs
NB cells are not susceptible to tumor antigen-specific CTLs because of their very low MHC-I expression [23]. Hence, we sought to determine whether an increase in cell surface MHC-I expression by IRF1 and p65 would render NB cells susceptible to antigen-specific T-cell cytotoxicity.
To this end, the three NB cell lines tested above, i.e. SH-SY5Y, SK-N-BE(2) and SK-N-BE(2)c, were evaluated by RT-PCR for the expression of the well-known tumor-antigen coding genes MAGE-A1, MAGE-A3, MART-1 and NY-ESO-1. Three antigens (MAGE-A1, MART-1 and NY-ESO-1) were not detectable. Absence of MART-1 in NB is not surprising in light of a previous study [8].
Only MAGE-A3 was highly expressed in the HLA-A1 + SH-SY5Y cell line (Fig. 5A). Therefore, SH-SY5Y cells treated with IFN-c, or transfected with IRF1 and/or p65 were tested in a standard 51 Cr release assay as targets of HLA-A1-restricted CTL specific for MAGE-A3-encoded peptide EVDPIGHLY [48]. CTLs did not lyse the HLA class I-negative melanoma cell line MSR3-mel (not shown) and SH-SY5Y cells transfected with an empty vector, but lysed efficiently, and approximately to the same extent, the MAGE-A3 + HLA-A1 + melanoma cell line ET-1, SH-SY5Y cells treated with IFN-c, and SH-SY5Y cells co-transfected with IRF1 and p65 (Figs. 5B and S6B). Conversely, SH-SY5Y singletransfected with IRF1 or p65 were lysed at lower efficiency as compared to SH-SY5Y double-transfected cells (Figs. 5B and S6B). These lytic effects are entirely due to MHC-I rescue, since IRF1 and/or p65 transfections did not alter the expression of MAGE-A3 (S6A).
Altogether, these data provide the proof of principle that the efficacy of T cell-based immunotherapy of NB may be improved by enhancing cell surface MHC-I expression with IRF1 and NF-kB p65.

Discussion
NB is the human tumor in which MHC-I-low phenotypes have the highest prevalence and most closely correlate with molecular determinants of aggressiveness [4]. In this report, we used NB cell lines representative of distinct NB molecular classes to identify impaired steps of MHC-I antigen presentation, and to correct them.
We confirm that NF-kB is the master regulator of constitutive MHC-I expression in NB, but, in addition, we show that it is a poor inducer of some APM components in NB cells with aggressive features and low MHC-I expression. However, these cells are particular susceptible to IFN-c up-regulation as demonstrated by the elective IFN-c-mediated MHC-I up-regulation and by the induction of both IRF1 and its inhibitor IRF2. These results suggest a full-fledged activation-extinction signaling loop. Accordingly, IRF1 is absolutely necessary to obtain a full rescue of antigen-presenting molecules in these MHC-I-low cells.
MHC-I reconstitution requires synergy between IRF1 and NF-kB, since IRF1 (either alone or in combination with IRF2) is substantially inactive. This finding places IRF1 on a hierarchically lower position than NF-kB. Nevertheless, IRF1 also appears to be crucial. Without it, large sections of the antigen processing and presentation cascade are impaired. IRF1 has long been known to be a major IFN-c-induced mediator of MHC-I up-regulation in tumors, including NB [21]. However, a direct effect on MHC-I and APM has never been demonstrated, to our knowledge, by direct reconstitution with IRF1. Herein, we identify IRF1 as a crucial coordinator of APM up-regulation. In our hands, IRF1 and NF-kB synergize in enhancing MHC-I, b2m, TAP1, TAP2, ERAP1, ERAP2, and tapasin, which are structural MHC-I subunits, dedicated peptide translocators, trimmers, editors, and chaperones.
Along the same line, IRF1, NF-kB p65, and MHC-I are coordinately expressed in vivo in tumor lesions, their nuclear absence and presence neatly segregating between high-grade and low-grade NB, respectively. Altogether, these findings suggest that linked suppression of genes collaborating in antigen processing and presentation in aggressive NB depends, in vivo as well as in vitro, on epistatic repression mainly acting on IRF-1 and NF-kB. Downstream correction of the stoichiometric insufficiency of IRF1 and NF-kB restores the presentation of tumor antigens in a model NB cell line, providing the proof of principle that complex immune escape phenotypes can be rescued by a limited number of master genes, and that NB can be made sensitive to CTLs that otherwise would be unable to recognize expressed, endogenous NB tumor antigens. To our knowledge, this is the first complete reconstitution of the MHC-I antigen-processing pathway in a tumor model by reintegration of a limited number of transcription factors, and not through treatment with pleiotropic cytokines or a mixture of undefined constituents. It is remarkable that at least in some NB cell lines MHC-I/APM defects are fully reversible, and that, in the limited cell panel studied by us, NF-kB and IRF1 are as good as IFN-c in rescuing low expression.
Our inability to correct ERAP2-low phenotypes in certain cells also suggests that extending the panel of NB cell lines and tissues is likely to reveal additional requirements. Further studies are necessary to test the percentage of NB tumors that can be rescued by double NF-kB/IRF1 reconstitution. It will be of interest to identify the master regulators that are still missing to achieve stepby-step, full reconstitution of the antigen presentation pathway in as many NB tumors as possible.
ERAP1 and ERAP2 trim peptide antigens to a size that fits the MHC-I antigen-binding groove. Whereas interference with the mouse homologue of ERAP1 has been shown to drastically alter tumor recognition by both T and NK cells [50], ERAP2 has no known mouse homologue, is less tightly coordinated with MHC-I [51,52] in vitro and in vivo, and is believed to have a more specialized role than ERAP1, i.e. ERAP2 only trims a limited subset of peptide antigens. Therefore, whereas of interest in light of the specialized role and regulation of ERAP2, the selective nonrescuability of this aminopeptidase in some NB cell lines is not expected to affect the processing of most antigens, including tumor antigens.
The present results support the original suggestion that MHC-I and APM components act as a coordinome with core and peripheral (gene-specific) control features, and a hierarchy of control steps [41]. Rescue of tumor antigen presentation on the whole by a limited number of transactivators will hopefully provide impetus to search for NB antigens to be used as immunotherapeutic targets, and to design small molecules to selectively manipulate the expression of NF-kB and IRF1 transcription factors, and target their activity to tumor cells. Indeed, antineoplastic effects of small molecule inhibitors targeting a single transcription factor (the late SV40 factor) have been preliminary reported [53]. It is hoped that antigen-specific immunotherapy (altogether one of the safest therapeutic approaches in oncology) may be more successful in early childhood, in the unexplored context of the very plastic immune system of the infant.

Supporting Information
Figure S1 Specificity, densitometric and statistical analyses of IRF1 and IRF2 expression. (A) Immunoblot analysis of IRF1 and IRF2 in SH-EP cells transfected with either IRF1 siRNA (siIRF1), IRF2 siRNA (siIRF2) or control scrambles (scr). The Jurkat cell line was included as a positive control. Equal amounts of whole-cell extracts were resolved by SDS-PAGE, immunoblotted and probed with IRF1 and IRF2 antibodies. ERp57 was used for normalization. Silencing of IRF1 and IRF2 by RNA interference in SH-EP cells affects the levels of IRF1 and IRF2 polypeptides as compared with the controls demonstrating the specificity of the antibodies. (B) Densitometric and statistical analyses of the Western blot bands shown in Figure 1B. Western blot bands in Fig. 1B were submitted to densitometric analysis. IRF1 and IRF2 values of the 3 MHC-I-expressing NB cells (to the left) and the 5 MHC-I-low NB cells (to the right) were separately averaged, and statistical significance (P value) of the differences between the two series was calculated as described in Materials and Methods.  Figure 3A. Densitometric values of the IFN-c-treated (+IFNc) or left untreated NB cells were separately averaged, and statistical significance (P value) of the differences between the two series was calculated as described in Materials and Methods. In addition, densitometric values of the 3 MHC-I-expressing NB cells and the 5 untreated MHC-I-low NB cells both untreated were separately averaged, and statistical significance (P value) of the differences between the two series was calculated as described above. (TIF) Figure S5 Densitometric and statistical analyses of the immunoblots shown in Figure 4A. Western blot bands were submitted to densitometric analysis. Densitometric values of cells transfected with pcDNA3, NF-kB p65 subunit, or IRF1 and NF-kB p65 subunit were separately averaged, and statistical significance (P value) of the differences between the three series was calculated as described in Materials and Methods. (TIF) Figure S6 A, RT-PCR analysis of MAGE-A3 in the SH-SY5Y either untransfected (none) and transfected with IRF1 and/or the NF-kB p65 subunit and the control empty vector (pcDNA3). Total mRNA was extracted from the transfected cells, reverse transcribed and cDNAs amplified with specific primers for MAGE-A3. GAPDH gene was used for normalization. B, flow cytometry analysis of surface MHC-I expression of SH-SY5Y cells treated as described in Figure 5 using W6/32 mAb. Data are shown as mean of fluorescence. (TIF)