The Toll like receptor 4 ligand cold-inducible RNA- binding protein as vaccination platform against cancer

ABSTRACT
Tumor infiltrating lymphocytes have been associated with a better prognostic and with higher response rates in patients treated with checkpoint inhibiting antibodies, suggesting that strategies promoting tumor inflammation may enhance the efficacy of these currently available therapies. Our aim was thus to develop a new vaccination platform based on cold-inducible RNA binding protein (CIRP), an endogenous TLR4 ligand generated during inflammatory processes, and characterize whether it was amenable to combination with checkpoint inhibitors. In vitro, CIRP induced dendritic cell activation, migration and enhanced presentation of CIRP-bound antigens to T-cells. Accordingly, antigen conjugation to CIRP conferred immunogenicity, dependent on immunostimulatory and antigen-targeting capacities of CIRP. When applied in a therapeutic setting, vaccination led to CD8-dependent tumor rejection in several tumor models. Moreover, immunogenicity of this vaccination platform was enhanced not only by combination with additional adjuvants, but also with antibodies blocking PD-1/PD-L1, CTLA-4 and IL-10, immunosuppressive molecules usually present in the tumor environment and also induced by the vaccine. Therefore, priming with a CIRP-based vaccine combined with immune checkpoint- inhibiting antibodies rejected established B16-OVA tumors. Finally, equivalent activation and T- cell stimulatory effects were observed when using CIRP in vitro with human cells, suggesting that CIRP-based vaccination strategies could be a valuable clinical tool to include in combinatorial immunotherapeutic strategies in cancer patients.

INTRODUCTION
CD8 T-cells play an important role in tumor recognition and rejection 1-3. Indeed, tumor infiltration by CD8 T-cells is not only associated with a better prognostic 2, but also with a higher response rate to newly developed therapies based on checkpoint inhibiting antibodies 4. Therefore, immunotherapies aimed at enhancing the number and properties of tumor-infiltrating T-cells are needed. Vaccines as monotherapies have achieved limited results in the clinic 5, 6, but with the advent of clinically available immune checkpoint inhibitors, they are gaining new options in combination therapies 7-12. Thus, it is of interest to develop potent immunization strategies amenable to combination with checkpoint inhibitors. Pathogen-associated molecules 13 have been used to activate professional antigen presenting cells, namely dendritic cells (DC), and prime thus T-cell responses through the provision of antigenic stimuli, costimulatory molecules and T-cell polarizing cytokines 14. In addition, there are also endogenous danger signals expressed or released by eukaryotic cells during stress or inflammatory conditions 15 which induce DC maturation. Accordingly, vaccination strategies have been developed based on the administration of antigens of interest together with the DC-activating molecules (adjuvants), avoiding thus the use of microorganisms or cells. In addition to DC activation, efficient antigen capture is a requisite to prime functional T-cell responses which greatly improves vaccine efficacy 16-18. Therefore, besides immunostimulatory properties, targeting of antigen to DC has also been pursued through its covalent binding to the adjuvant molecules 19. With this aim, Toll- like receptor (TLR) ligands have been used as adjuvants, not only for their immunostimulatory properties, but also for their ability to target antigens to DC 20-22. Due to the need of inflammation-inducing molecules for antitumor strategies, we focussed our interest in cold- inducible RNA-binding protein (CIRP), whose expression is upregulated during mild hypothermia 23, ultraviolet irradiation 24 and hypoxia 25. It has been recently reported that CIRP has important pro-inflammatory properties, it is released as a mediator that triggers inflammatory responses during hemorrhagic shock and sepsis, activating macrophage inflammatory properties in a TLR4-dependent manner 26. Since we had previously demonstrated that other protein and peptidic TLR ligands can be used as adjuvants due to their ability to target antigens to DC and induce their activation 22, 27, we hypothesized that CIRP may also behave as an adjuvant with targeting and immunostimulatory capacity. We have thus explored these properties in CIRP, the capacity of CIRP-containing constructs as anti-tumor vaccines and their effect when combined with checkpoint inhibitors, demonstrating the efficacy of this vaccination platform.

RESULTS
CIRP has been described as a TLR4-binding protein which activates macrophages and induces the production of inflammatory cytokines 26. To study its effect on DC as well as its antigen targeting capacity, which would result in presentation to T-cells, we designed a fusion protein containing amino acids 254-276 (QLESIINFEKLTEW) from ovalbumin, which includes the T- cell epitope OVA(257-264) (SIINFEKL) (recognized by CD8 T-cells from H-2b mice) and three amino-acid flanking residues, linked to the N-terminus of murine CIRP (from now on SIIN- CIRP). A 6-His tail was also added to the N-terminus for protein purification (Figure 1A). Protein was expressed in bacteria and highly purified (Figure 1B) until endotoxin levels were below 10 ng endotoxin/mg of protein. Analyses of the capacity of SIIN-CIRP to induce TLR4- mediated activation carried out in HEK-293 cells confirmed that in those cells transfected withTLR4-MD2-CD14, IL-8 was induced by SIIN-CIRP and by LPS, whereas in control cells expressing lacZ, neither SIIN-CIRP nor LPS, induced cytokine production (Figure 1C).Once demonstrated the TLR4 binding capacity of SIIN-CIRP, its effect on murine DC maturation was analyzed. SIIN-CIRP and LPS, but not untargeted CD8 epitope or the original antigen OVA, upregulated expression of maturation markers CD86 and CD54 (Figure 1D). Importantly, treatment of SIIN-CIRP with proteinase K completely abolished marker upregulation, indicating that the stimulatory effect observed was due to the protein and not to potential bacterial endotoxin.To demonstrate that linking of SIIN peptide to a protein moiety was not sufficient for DC activation we included SIIN in a new truncated construct (SIIN-CIRP). This protein contains only amino acids 1-100 from CIRP (Supplementary Figure 1A) and lacks CIRP region 100-125, known to bind the TLR4 partner protein MD2 26, presumably involved in CIRP inflammatory properties. SIIN-CIRP induced much lower activation of TLR4-MD2-CD14-expressing HEK- 293 cells than SIIN-CIRP (Supplementary Figure 1B). Accordingly, SIIN-CIRP did not induce DC maturation, measured as up-regulation of CD86 (Supplementary Figure 1C).Regarding cytokine production, DC treated with graded doses of SIIN-CIRP showed a dose- dependent response (in the range of that induced by 1 g/ml of LPS) in the production of IL-12, TNF- and IL-10 (Figure 1E), confirming the stimulatory capacity of the protein.

However SIIN-CIRP, as occurred with phenotypic maturation, was inactive (Supplementary Figure 1D).Finally, we also demonstrated in migration assays that incubation of DC with SIIN-CIRP greatly increased their migratory capacity in migration assays (Figure 1F).After characterizing its DC activating capacity, SIIN-CIRP was tested in antigen presentation assays. DC pre-incubated with equimolar amounts of SIIN peptide, OVA, CIRP or SIIN-CIRP, were co-cultured with OT-I CD8 T-cells, specific for SIIN peptide, and DC/T-cell interactions visualized by electron microscopy. Strong interactions were observed in SIIN-treated DC, with higher numbers of T-cells in close contact with DC, followed by DC treated with SIIN-CIRP or OVA. Poor interactions were seen in CIRP-treated and in untreated DC (Figure 2A). This was confirmed by quantifying the mean cleft size, the fraction of the synapses occupied by close to intermediate appositions and the mean size of continuous region of close apposition (Figure 2B). In a next series of experiments analyzing functional DC/T-cell interactions, we found that SIIN- CIRP-incubated DC strongly stimulated OT-I T-cell proliferation in a dose-dependent manner (Figure 2C), as opposed to untreated DC or mature DC incubated with LPS but without antigen. Due to the importance of IFN- production by CD8 T-cells in antitumor responses, we measured the production of this cytokine in equivalent experiments. Contrary to image analyses, where SIIN peptide showed the most intense interactions, a high production of IFN- was induced by DC treated with SIIN-CIRP, when compared with DC incubated with peptide SIIN or the whole OVA protein, which induced lower or no production of this cytokine by CD8 OT-I T-cells (Figure 2D). All together, these in vitro results indicate that SIIN-CIRP is able to induce DC activation in a TLR4-dependent manner, favouring antigen presentation to T-cells.Once demonstrated in vitro the activation and targeting properties of CIRP, we next tested the in vivo immunogenicity of SIIN-CIRP. Mice received a single immunization with SIIN-CIRP in saline in the absence of any additional adjuvant.

One week later CD8 responses induced against SIIN were compared with those induced by immunization with the same molar amounts of untargeted SIIN or the SIIN-containing protein OVA. While SIIN-CIRP clearly primed SIIN- specific IFN--producing cells, untargeted SIIN or OVA did not induce appreciable T-cell responses (Figure 3A). Neither did coupling of SIIN to the truncated CIRP protein (SIIN-CIRP construct) (Figure 3B) nor co-administration of SIIN and CIRP as separated molecules (Figure 3C) led to T-cell responses as strong as those induced by SIIN-CIRP, demonstrating that responses induced by SIIN-CIRP depend on DC activating and targeting properties of CIRP.CIRP activates MyD88 and TRIF pathways and induces type I IFN-dependent T-cell responses Since TLR4 ligands may signal through MyD88- and/or TRIF-dependent elements 28, 29 we analyzed which pathways were activated by SIIN-CIRP in DC. Time course experiments showed that both genes dependent on MyD88 (Figure 4A, left panels) and TRIF (Figure 4A, right panels), were upregulated upon incubation of DC with SIIN-CIRP. Moreover, since IFN-beta upregulation by CIRP was maintained in DC, and type I IFN has emerged as an important pathway in T-cell-dependent tumor rejection 30, 31, we tested the IFN dependency of the CIRP- induced T-cell responses. SIIN-CIRP induced a lower phenotypic maturation (Figure 4B) and cytokine production (Figure 4C) in DC from mice lacking type I IFN receptor. Moreover,immunization experiments showed that poorer responses were induced in these animals than in WT mice (Figure 4D), suggesting that T-cell responses induced by CIRP-targeted antigens depend on type I IFN signalling pathway.To test the capacity of SIIN-CIRP as an anti-tumor therapeutic vaccine, mice with 5 mm established E.G7-OVA tumors were vaccinated with SIIN-CIRP, CIRP without any co-expressed antigen, untargeted peptide SIIN, a mixture of SIIN and CIRP or left untreated.

SIIN-CIRP- vaccinated mice, but not control groups with SIIN or CIRP, had a delay in tumor growth, as compared with untreated mice (Figure 5A). Moreover, in accordance with results shown in Figure 3C, co-administration of SIIN + CIRP did not demonstrate antitumor effect. Moreover, 60% of mice vaccinated with SIIN-CIRP survived after treatment, whereas survival rates in remaining groups ranged around 15-25%. These results were confirmed in the MC-38-OVA tumor model, with SIIN-CIRP vaccination inducing tumor rejection also in this setting (Figure 5B).Depletion experiments carried out in combination with the SIIN-CIRP vaccine in mice bearing E.G7-OVA tumors showed that depletion of CD8 cells completely abolished the therapeutic effect of vaccination (Figure 5C), as expected by the presence of CD8 T-cell epitope SIIN peptide as the only antigen in the vaccine, and confirming the essential role of CD8 cells in our vaccination protocol.Since we have shown that vaccine efficacy can be improved by adjuvant combination 32, to enhance responses induced by CIRP-based vaccines, we immunized mice with SIIN-CIRP and adjuvants signalling through different DC receptors and activation routes. Combination with non-TLR adjuvants (agonistic anti-CD40 antibodies), as well as with poly(I:C) (TLR3) and CpG oligonucleotides (TLR9) clearly enhanced CIRP-induced T-cell responses, whereas no enhancement was observed with Imiquimod (TLR7). Finally, a multiple adjuvant combination (MAC) 32 also enhanced CIRP-induced responses (Figure 6A).Together with the presence of tumor-infiltrating lymphocytes, expression of PD-L1 has been intensively studied as a biomarker of efficacy in anti-PD-1/PD-L1 therapies 33. Interestingly, PD- L1 was upregulated in DC after incubation with SIIN-CIRP (Figure 6B), suggesting that CIRP- based vaccines would benefit from inhibition of this pathway. Indeed, combination of SIIN- CIRP with PD-1-blocking antibodies enhanced T-cell responses (Figure 6C). To test the therapeutic efficacy of combined strategies using CIRP-based vaccines in a more challenging tumor model, we carried out experiments in mice bearing 5 mm B16-OVA tumors. SIIN-CIRP vaccination in the presence of control antibodies delayed tumor growth in this poorly immunogenic tumor model, although no animal survived at the end of the experiment. However, combination of the SIIN-CIRP vaccine with antibodies against PD-1 almost abolished tumor growth during treatment period, which was later strongly delayed, resulting in above 30% long term survival (Figure 6D).

We have recently demonstrated that blockade of adjuvant-induced IL-10 (a cytokine induced by CIRP; Figure 1E) greatly improves anti-tumor efficacy of therapeutic vaccines 34. Although vaccination with SIIN-CIRP combined with single blockade of IL-10R or PD-1 enhanced vaccine potency, simultaneous blockade of these molecules in combination with vaccine induced the strongest T-cell responses (Figure 6E). Accordingly, therapeutic vaccination plus double blockade resulted in greatly delayed tumor growth in parallel with tumor rejection in more than 40% of mice (Figure 6F).Finally, we also considered blockade of CTLA-4, another immune checkpoint clinically targeted. As occurred with IL-10R/PD-1 blockade, combined CTLA-4/PD-1 blockade also enhanced T- cell responses (Figure 6G), which resulted in a strong antitumor effect, clearly improving the therapeutic efficacy over that induced by the vaccine or by antibodies (Figure 6H).Homology between human and murine CIRP is above 95% at the amino acid level. Moreover, we had seen that murine CIRP could activate HEK293 cells expressing human TLR4 (Figure 1C). We thus took advantage of this high similarity across species and tested the effect of murine CIRP on human DC. Incubation of human monocyte-derived DC with CIRP induced clear phenotypic upregulation of maturation-associated markers CD86 and CD54 (Figure 7A) as well as production of IL-12, TNF- and IL-10 (Figure 7B). Regarding the T-cell stimulatory capacity of CIRP-treated human DC, we observed in MLR assays that human DC treated with CIRP stimulated allogeneic T-cell proliferation more efficiently than untreated DC, confirming the mature phenotype of these cells (Figure 7C).

DISCUSSION
Increased levels of tumor-infiltrating lymphocytes have been associated to higher responses to checkpoint inhibitors 4, suggesting that strategies promoting tumor inflammation may enhance response rates to these inhibitors. In this setting, although therapeutic vaccination clinical trials have yield limited clinical results 5, 6, combination of both strategies may offer better results. Indeed, tumor PD-L1 expression, which is associated to higher therapeutic efficacy, may result as a consequence of adaptive resistance induced by IFN- produced by infiltrating T-cells. Conversely, PD-L1 may be absent in tumors lacking lymphocytes, suggesting that vaccination combined with checkpoint blockade would be a suitable strategy for these patients 35. Thus, it is of paramount importance the design of new vaccines containing molecules with capacity to stimulate DC and target antigens to this cell population, criteria needed for proper T-cell activation 36. With this aim, we tested in vitro and in vivo properties of CIRP, a TLR4-binding endogenous danger molecule released during different stress conditions, with inflammatory properties 26. In vitro characterization of a fusion protein containing CIRP and a CD8 T-cell epitope demonstrated its immunostimulatory activity on DC as well as its antigen targeting properties, which improved in vitro antigen presentation and resulted in stronger in vivo T-cell responses. These experiments showed that both properties were needed for CIRP activity, since co-administration of antigen with uncoupled CIRP or antigen linkage to a truncated CIRP molecule lacking inflammatory properties yielded much poorer responses. Important efforts have been carried out to characterize DC receptors suitable for antigen targeting 19. However, in some
cases, despite efficient antigen capture and payload delivery to DC, these pathways do not result in DC maturation, requiring additional signals for efficient activation and proper antigen presentation 16. In the case of CIRP, both properties are present in the same molecule.

As previously reported and confirmed in our work, CIRP-induced cell activation is related to TLR4 binding. Peptides belonging to amino acids 100 to 125 have been reported to bind the TLR4-partner protein MD2 26. However, the full CIRP sequence seems to be required for their functional properties, since as characterized here, the truncated CIRP(1-100) protein lacks the capacity displayed by full CIRP. Moreover, SIIN antigen bound to the CIRP region 100-125, responsible for MD2-binding, is unable to induce T-cell responses (our unpublished results). TLR4 is a receptor whose ligation leads to activation of two signalling pathways, mediated by TRIF and MyD88 molecules, related to its capacity to induce type I IFN and pro-inflammatory mediators 28. There are strong adjuvants which induce both pathways, whereas in other cases, molecules and microorganisms with selectivity for one of them have been described 29, 37, 38. We show here that CIRP induces the expression of genes corresponding to both pathways, although our results do not formally demonstrate that both are actually needed. In this sense, we show that IFN-beta, a TRIF-dependent molecule, plays an important role on the DC activating capacity of CIRP and the ensuing T-cell responses. In this regard, as an endogenous danger molecule which induces TLR4-dependent DC activation, migration and release of type I IFN, CIRP induces some of the features of immunogenic cell death 39, a death modality associated with the immunogenic properties of many chemotherapeutic agents used in the clinic.

Due to their effects on DC, vaccination with CIRP-containing immunogens induced antitumor responses able to reject established tumors, as demonstrated in several models. These experiments showed again that both properties (DC stimulation and antigen targeting) are responsible for its antitumor efficacy, since the single administration of CIRP into the tumor, even in the presence of tumor-provided endogenous antigens, does not have any meaningful effect. Although we do not discard the potential capacity of CIRP-based immunogens to activate CD4 T-cell responses which would support activation of CD8 T-cells, we show here that even in the absence of CD4 epitopes, CIRP-based vaccines may induce functionally competent CD8 T- cells which are responsible for tumor rejection.Therapeutic vaccination may bypass the lack of immunogenicity of tumor cells, resulting in the induction of potent T-cell responses. However, there are other mechanisms which preclude tumor rejection, such as the immunosuppressive tumor microenvironment and immunomodulatory molecules induced by the vaccine to avoid excessive T-cell activation. PD- L1, IL-10 and CTLA-4 are examples of these elements which may operate at the level of T-cell activation during antigen presentation by DC and at the effector phase of tumor cell recognition 40, 41. Indeed, preclinical 41 and recent clinical results (in the case of PD-1/PD-L1) 42 obtained after blocking these pathways have shown the pertinence of this approach. Thus, the expression of these molecules induced by vaccination suggests that those effects observed when using monotherapies based on inhibitory antibodies could be improved if combined with CIRP-based vaccines. Accordingly, we have demonstrated that combinations of CIRP-based vaccines with blocking antibodies result in higher therapeutic effect, due to superior effect at the induction phase (as we have demonstrated) and enhanced activity at the effector phase. Indeed, monotherapies based on checkpoint inhibiting antibodies seem to display a superior activity in those tumors with already primed immune responses, whereas “cold” tumors lacking immune effectors show poor responses to these therapies 35. Thus, already available anti-PD-1 therapies may benefit from priming of antitumor immunity with CIRP-based vaccines.Finally, an important step to consider when translating these findings to the clinical setting is the activity of this molecule in human cells. Although not specifically tested with the human CIRP protein, the high sequence homology of the murine version at the amino acid level (> 95%) allowed us to demonstrate that this approach is also feasible with human cells. Hence, human DC were efficiently activated when incubated with CIRP, results which demand further characterization before being used in humans.

In summary, we have shown that the endogenous danger molecule CIRP has DC activating and targeting properties, which facilitates antigen presentation for efficient T-cell activation. These properties allow the use of CIRP as a vaccination platform which can be included in combinatorial strategies containing other immunostimulatory approaches, such as antibodies blocking inhibitory pathways. These new combinations will help to provide better results than those currently obtained by using immunological monotherapies.pET14b plasmids encoding CIRP and CIRP-bound antigens and a 6x His tail (Genscript) were used to transform either BL21(DE3) (Novagen) or ClearColi BL21 (DE3) E. coli cells (Lucigen). After overnight induction with 0.4 mM IPTG and bacterial lysis, proteins were harvested from inclusion bodies and resuspended in buffer containing 8 M urea and 20 mM HEPES pH 7.2. They were purified by affinity chromatography (HisTrap; Pharmacia) using a fast protein liquid chromatography platform (AKTA; Pharmacia). Before protein elution, endotoxin was removed by extensive washing with buffer UTT (8 M urea, 20 mM HEPES, 0.4% Tween 20, 0.4% Triton X-100, pH 7.2), and then eluted with 500 mM imidazol. The eluted protein was desalted using HiTrap desalting columns (Pharmacia). Endotoxin levels were always below 10 ng endotoxin/mg of protein as tested by Quantitative Chromogenic Limulus Amebocyte Lysate assay (Lonza). Recombinant endotoxin-free OVA protein (Endograde) and peptide OVA(257-264) SIINFEKL were purchased from Hyglos (Germany) and from Genecust (Luxemburg), respectively.Experimental work with C57BL/6 (Harlan), OT-I (C57BL/6-Tg(TcraTcrb)1100Mjb/J; Jackson Laboratories) and IFNAR KO mice (C57BL/6-IFN-a/bRo/o; a gift from Mathew Albert, Institute Pasteur, Paris, France) was conducted according to relevant national and international guidelines, after approval by the institutional review board.

HEK293/human TLR4 (hTLR4)-MD2-CD14- or HEK293/LacZ-expressing cells (Invivogen) were grown in complete DMEM medium (supplemented with 10% FCS, 2 mM glutamine, 1% Penicillin/Streptomycin) plus 5 g/ml blasticidin and 25 g/ml hygromycin. E.G7-OVA thymoma cells (ATCC), MC38-OVA colorrectal carcinoma cells (a kind gift of Dr. I Melero; Pamplona, Spain) and B16-OVA melanoma cells (a kind gift of Dr. G. Kroemer; Paris, France) were cultured in complete RPMI medium (RPMI 1640 supplemented with 10% FCS, 2 mM glutamine and 1% Penicillin/Streptomycin).Bone marrow-derived DC were generated as described 43. Human DC were differentiated from monocytes 44 obtained from buffy coats from the Blood and Tissue Bank of Navarra. Samples were obtained after informed consent and all investigation was conducted according to the principles of the Declaration of Helsinki after approval by the institutional ethical review board. In both cases, DC were collected and cultured in 96-well plates (2 x 105 cells/well) with different concentrations of CIRP-containing protein, LPS (1 g/ml) or left unstimulated. One day later supernatants were harvested to determine cytokine production and DC were analyzed by flow cytometry. In some experiments CIRP-containing proteins were previously digested for 30 min with proteinase K (20 mg/ml).Murine DC were stained with FITC-labelled anti-CD54, PE-anti-CD11c, PE-Cy5-anti-CD86 and PE-anti-PD-L1 antibodies, whereas for human DC, FITC-labelled anti-CD86 and PE-labelled anti-CD54 antibodies and their corresponding control isotypes (all from BD-Biosciences; San Diego, CA) were used. After 30 min, cells were washed and surface expression of the different molecules was analyzed by using a FACScanto flow cytometer (BD-Biosciences).DC migration assays were carried out in 24-well plates with transwell inserts of 5 m pore size using a Transwell chamber (Costar Corning, Cambridge, MA). DC (2 x 105) harvested one day after CIRP stimulation were cultured in the upper compartment, whereas the lower compartment contained culture medium with or without 30 ng/ml of CCL21 (Peprotech). After 2 hours, inserts were removed and migrated cells were counted as described 45. Results are expressed as chemotactic index, fold increase of migrated cells in the presence vs. absence of CCL21.

For analyses of DC/T-cell interactions by electron microscopy, 106 murine DC previously incubated for 12 h with antigens (proteins or peptides at 1 M) were co-cultured on coverslips for 30 min with 5×105 murine splenic CD8 T-cells from OT-I mice positively selected (Miltenyi). Next, cells were fixed and processed as described 46. For in vitro T-cell stimulation assays, purified murine CD8 OT-I T-cells mice (104 cells/well) were cultured in 96-well plates with graded numbers of DC previously incubated for 12 h with antigens. After 24 h, supernatants were harvested to measure IFN- production and cells pulsed overnight with 0.5 Ci of tritiated thymidine to determine cell proliferation. For human DC, MLR were carried out as described 44 by co-cultivating allogeneic lymphocytes (105 cells/well) with graded numbers of monocyte- derived DC previously incubated with CIRP.For ultrastructural studies, cells from each treatment were adhered to poly-L-lysine-coated coverslips. Then, the samples were processed following the protocol described previously with modifications 46. Briefly, the cells were treated with a mixture of 2% formaldehyde (Ultra Pure EM Grade, Polysciences Inc., Philadelphia, USA) and 2.5% glutaraldehyde (EM Grade, TAAB Laboratories Equipment Ltd., Berks, UK) in PBS for 1 h at room temperature. The cell monolayer on the coverslips was then washed with PBS and distilled water, post-fixed for 45 minutes with 1% osmium tetroxide (TAAB Laboratories Equipment Ltd.) in PBS, washed with distilled water, treated during 45 minutes with 1% aqueous uranyl acetate (Electron Microscopy Sciences, Hatfield, USA), washed again and dehydrated with increasing quantities (50%, 75%, 95% and 100%) of ethanol seccosolv (Merck KGaA, Darmstadt, Germany). The samples were maintained in coverslips throughout the process and finally embedded in epoxy resin 812 (TAAB Laboratories Equipment Ltd.) contained in gelatine capsules (Electron Microscopy Sciences). The epoxy resin was polymerized for 2 days at 60°C. Resin was detached from the coverslips by successive immersions in liquid nitrogen and hot water. Ultrathin, 70-nm-thick sections were obtained with an Ultracut UCT ultramicrotome (Leica Microsystems), transferred to 200 mesh Nickel EM grids (Gilder, Lincolnshire, UK) and stained with 3% aqueous uranyl acetate (10 minutes) and lead citrate (2 minutes) (Electron Microscopy Science). Sections were visualized on a JEOL JEM 1200 EXII electron microscope operating at 100 kV (JEOL Ltd., Tokyo, Japan).

Control LacZ-expressing HEK293 cells or hTLR4-MD2-CD14-transduced cells (5 x 104 cells/well) were cultured in 96 well plates with different protein antigens or LPS. Next day supernatants were harvested and cell activation was determined by measuring IL-8 production.Total RNA extraction from DC and real-time PCR were performed as described 32, using primers shown in Supplementary Table 1. Results were normalized according to -actin. The amount of each transcript was expressed by the formula: 2ΔCt (ΔCt = Ct(β-actin)-Ct(gene)).Cytokines produced by murine or human DC (IL-12, TNF- and IL-10), HEK293 cells (IL-8) or CD8 T-cells (IFN-) were determined by ELISA sets (BD-Biosciences).C57BL/6 or IFNAR KO mice were immunized s.c. with equimolar amounts (2 nmoles) of CIRP- containing protein antigens, OVA protein, peptides or unbound mixtures of CIRP protein and antigenic peptide. In some experiments, mice received i.p. 500 g of anti-IL-10R (1B1.3A;BioXcell), 50 g of anti-PD-1 (RMP1-14; BioXcell), 100 g of anti-CTLA4 (9D9; BioXcell) or the corresponding isotype control antibodies (BioXcell) on day 0. In adjuvant combination experiments mice also received Imiquimod cream (Meda Aldara™; topical application; 2.5 mg/mouse), poly(I:C) (Amersham; 50 g/mouse s.c.), CpG 1668 (Sigma; 50 g/mouse s.c.), agonistic FGK45.5 anti-CD40 antibodies (BioXcell; 50 g/mouse s.c.) or a multiple adjuvant combination (MAC) 32 containing Imiquimod, poly(I:C) and anti-CD40. One week later animals were sacrificed and T-cell responses were measured by enumerating IFN--producing cells by ELISPOT as described 32 using a kit from BD-Biosciences. For these experiments splenocytes were stimulated with 1 g/ml of OVA(257-264) or left unstimulated.C57BL/6 mice were injected s.c. with 105 tumor cells (E.G7-OVA, MC38-OVA) or intradermically (B16-OVA cells).

One week later, when the tumor diameter was about 5 mm, they were treated for 3 weeks with 2 weekly i.t. injections of CIRP-containing immunogens, CIRP, peptide antigen, the unbound mixture or PBS. In combination experiments, vaccine was accompanied by three weekly i.p. injections of antibodies (500 g of anti-IL-10R, 100 g of anti-PD-1, 100 g of antiCTLA4) or the corresponding isotype control. Tumor volume was calculated according to the formula: V= (length x width2)/2. For depletion experiments, mice received i.p. injections of 200 g of anti-CD8 (H35.17.2; a kindly gift of Dr. C. Leclerc; Institute Pasteur, Paris, France) or isotype control (BE0088, BioXcell) antibodies on days -1, 0, 1 and 6, being 0 the day when treatment starts. Mice were killed when tumor diameter reached 17 mm.Survival curves of animals treated with different protocols were plotted according to the Kaplan– Meier method and were compared using the log-rank test. Immune responses were analyzed using nonparametric Kruskal-Wallis and Mann-Whitney U tests. P<0.05 was taken to represent statistical AUNP-12 significance.