In vivo Colocalization of Antigen and CpG within Dendritic Cells Is Associated with the Efficacy of Cancer Immunotherapy

Toll-like receptors (TLR) encompass a specialized set of pathogen recognition receptors that are expressed by cells of the innate and adaptive immune system. When triggered by highly conserved danger-associated molecules, such as lipopolysaccharide, certain lipoproteins, and bacterial DNA or RNA, these receptors initiate signaling cascades leading to protective immune responses (1). The identification of natural agonists for TLRs and the development of synthetic equivalents, showing increased stability and half-life in vivo, have opened novel therapeutic avenues for the treatment of a large range of infectious diseases and cancer.

In the development of cancer vaccines, the TLR9 agonist cytidyl guanosyl (CpG) oligodeoxynucleotide stands out because of its preferential induction of Th1 responses and tumor-specific cytotoxic CD8⁺ T lymphocytes (CTL; refs. 2–4). TLR9 is predominantly expressed by B cells and dendritic cells (DC) that internalize and directly respond to CpG (5). On triggering of TLR9, DCs mature and migrate to draining lymph nodes where they present antigens to T and B lymphocytes (6). Importantly, mature DCs acquire the unique ability to present captured antigens on MHC class I molecules, a process known as cross-presentation (7), which is crucial for efficient priming of tumor-specific CTLs (8, 9). As such, CpG administration has been reported to prevent tumor outgrowth in a prophylactic setting and could also eradicate established tumors in mice (10, 11).

Unfortunately, the clinical application of CpG in humans has thus far been less successful (12). A common argument used is the differential TLR9 expression in DCs in mice and man. TLR9 is abundantly expressed in murine and human plasmacytoid DCs (pDC) and in myeloid DCs (mDC) in mice. The expression in human mDCs is, however, less clear as some studies report weak expression (13, 14), whereas a recent study shows that human mDCs contain TLR9 protein in amounts comparable with pDC (15). Moreover, intensive cross-talk between mDCs and pDCs is essential for CpG-induced immune activation in both mice and man (16, 17), indicating a functional link between the DC subsets in both species.

Another explanation for the divergence in the efficacy of CpG application in preclinical and clinical studies is the regimen of CpG administration. We noted that in murine studies, showing the most potent CpG effects, CpG is provided along with the antigen or in close proximity to tumor antigens (2, 18), whereas in clinical studies CpG is administered independent of antigen release (19).

To investigate the importance of timing of TLR agonist administration relative to the availability of tumor antigens in vivo, we applied a unique in situ tumor destruction model in which the release of tumor antigens can be tightly controlled. Using cryosurgery (20) or radiofrequency ablation of established tumors, we have recently showed that in situ tumor destruction, in contrast to tumor resection, leads to the induction of tumor-specific immunity, which can be greatly enhanced by coadministration of CpG (4). In the present study, we show that the failure or success of CpG as a vaccine adjuvant is closely correlated with the colocalization of antigen/CpG within DCs and thus the timing of CpG administration relative to the release of antigen.

Mice and cell lines. C57BL/6n mice (6–8 wk old) were purchased from Charles River Wiga and the experiments were performed according to the guidelines for animal care of the Nijmegen Animal Experiments Committee. The B16F10 melanoma cell line was obtained from the American Type Culture Collection and maintained in complete medium [Iscove's modified Dulbecco's medium (IMDM) supplemented with 5% fetal bovine serum (FBS; Greiner Bio One), 100 units/mL penicillin G sodium and 100 μg/mL streptomycin (Pen/Strep), MEM sodium pyruvate (1 mmol/L), NaH₂CO₃, MEM vitamins, MEM nonessential amino acids (all from Life Technologies), and 20 μmol/L β-mercaptoethanol (β-ME)]. OVA-transfected murine melanoma cell line B16F10 (B16OVA, clone MO5) was kindly provided by Dr. K. Rock (Department of Pathology, University of Massachusetts Medical Center, Worcester, MA; ref. 21) and cultured in complete medium supplemented with 30 μg/mL hygromycin and 1 mg/mL G418. The D1 cell line, a long-term growth factor–dependent immature splenic DC line derived from B6 mice, was cultured as described before (22). B3Z cells, a T-cell hybridoma specific for SIINFEKL in H-2K^b, which carries a β-galactosidase construct driven by NF-AT elements from the interleukin-2 (IL-2) promotor (23), were cultured in IMDM supplemented with 5% FBS, Pen/Strep, 20 μmol/L β-ME, 0.5 mg/mL hygromycin, and 2 mmol/L l-glutamine. B6-B7.1-sigOVA, a cell line presenting high levels of the MHC class I peptide complex but no OVA class II peptide complexes on its surface (24), was cultured in IMDM, 5% FBS, Pen/Strep, 20 μmol/L β-ME, 0.1 mg/mL hygromycin, and 0.1 mg/mL G418 and used as a positive control.

Tumor model and cryosurgery. Tumors were induced and treated with cryoablation as previously described (4). In challenge experiments, mice were rechallenged with 2 × 10⁴ B16OVA cells 40 d after ablation treatment. Two months later, mice that survived the first tumor rechallenge received a second rechallenge with 5 × 10⁴ B16OVA or B16F10 cells. Mice were sacrificed when tumor volume exceeded 1,500 mm³.

CpG 1668 (5′-TCCATGACGTTCCTGATGCT-3′) with phosphorothioated backbone (purchased from Sigma-Genosys) was injected in PBS peritumorally (100 μg divided over three injections of 10 μL surrounding the tumor) within 30 min after ablation or 3 or 1 d before or after ablation. We used CpG 1668, which is a type B CpG, similar to the clinical-grade available CpGs currently used in clinical trials.

Enrichment and purification of DCs. Inguinal lymph nodes draining the tumor site were excised and digested in collagenase type II and DNase for 15 min at 37°C. After addition of EDTA and resuspending, cells were filtered and DCs were enriched by positive selection according to the manufacturer's protocol (Miltenyi Biotec B.V.). For confocal experiments, OVA-Alexa488 and CpG-Cy5 double-positive cells were further purified by fluorescence-activated cell sorting (FACS) sorting (Elite cell sorter, Beckman Coulter).

Loading and maturation of DCs. To study the fate of antigen after cryoablation, tumors were injected with fluorescently labeled OVA protein (OVA-Alexa488, 20–40 μg/20 μL). In some studies, Cy5-labeled CpG (CpG-Cy5, 50–100 μg/30 μL) was used to determine the distribution of CpG. Tumor-draining lymph nodes were isolated 48 h after treatment. CD11c-enriched cells were incubated with Fc Block (CD16/CD32; 2.4G2) and stained with anti-CD11c (HL3), biotinylated anti-CD80 (1G10/B7), or isotype controls and streptavidin-phycoerythrin, all obtained from BD Biosciences. Expression of CD80 was analyzed in Alexa⁺ and Alexa⁻ cells and gated CD11c⁺ cells by flow cytometry (FACSCalibur, Becton Dickinson & Co.). As it was reported that uptake of the model antigen OVA by bone marrow–derived DCs largely depended on the mannose receptor (25), we repeated our experiments with bovine serum albumin-Alexa488. No differences in uptake by DCs in vivo were observed between the two antigens (data not shown).

Cross-presentation. DCs were enriched from tumor-draining lymph nodes 48 h after ablation and cocultured with B3Z cells for 24 h at 37°C in 96-well plates. The presentation by DCs of SIINFEKL in H-2K^b results in production of β-galactosidase by B3Z cells (23), which can be detected by adding 0.15 mmol/L chlorophenolred-β-d-galactopyranoside (Calbiochem), 9 mmol/L MgCl₂, 0.125% NP40, and 100 mmol/L β-ME in PBS. Plates were incubated for 3 h at 37°C and absorbance values were measured using a photo spectrometer (595 nm).

Antigen-specific CTL. Ten days after ablation, spleens and tumor-draining lymph nodes were isolated and single-cell suspensions were prepared. Mixed cell suspensions of lymph node and spleen were plated in 24-well plates in complete IMDM supplemented with human IL-2 (10 units/mL). B16OVA cells were treated overnight with recombinant IFN-γ (50 units/mL), harvested and irradiated (1,600 rad), and added to each well (5 × 10⁴ per well). At days 3 to 4 of culture, cells were harvested and dead cell debris was removed by a Ficoll step. Cells were cultured in fresh 24-well plates for another 4 to 5 d. After a second Ficoll step, cells were stained with FITC-conjugated CD4 (L3T4) or CD8 (53-6.7) and allophycocyanin (APC)-conjugated OVA-K_b tetramers (Sanquin) and analyzed by flow cytometry.

Confocal microscopy. Tumors were injected with OVA-Alexa488 just before cryoablation and CpG-Cy5 was administered peritumorally at the indicated time points. Sixteen hours after ablation, tumor-draining lymph nodes were excised and DCs were enriched by positive selection and OVA-Alexa488 CpG-Cy5 double-positive cells were sorted on a FACS sorting machine. The purified populations were coated on fibronectin-coated or poly-l-lysine–coated glass slides (1.5 h or 15 min), fixed with 2% paraformaldehyde, blocked overnight with PBA, stained with anti-mouse LAMP-1 (eBioscience) or anti-mouse EEA-1 (Abcam) and goat anti-rat/rabbit Alexa568 (Invitrogen), and imbedded in Mowiol (with 1% azide). Images were acquired using a Zeiss LSM510 meta confocal microscope or Bio-Rad MRC1024 operating on a Nikon Optiphot microscope and a Nikon 60× Plan Apo 1.4 immersion lens. Pictures were analyzed with Bio-Rad Lasersharp 2000 and intensity profiles of the merged channel along lines were analyzed using the profile tools of ImageJ software (NIH, Bethesda, MD). Green lines represent the intensity of the antigen marker, whereas the red line represents the intensity of the TLR agonist. Pearson correlation coefficients of total cell images were calculated.

Statistical analyses. Data were analyzed using a two-tailed Student's t test or one-way ANOVA for multiple comparisons with Bonferroni as post hoc test. All analyses were performed using GraphPad Software Prism 4.0 for Macintosh.

Timing of CpG administration determines antitumor immunity. To investigate the importance of the presence of CpG relative to the availability of tumor antigens in vivo, we used a recently developed mouse model in which in situ tumor destruction by cryoablation provides an instant antigen source for DCs in vivo. Hereto, B16OVA tumor-bearing mice were treated with cryoablation and injected with CpG at different time points relative to ablation. Subsequently, tumor-free mice were challenged with a lethal dose of B16OVA cells 40 days after ablation. As reported previously (4), mice treated with cryoablation alone showed prolonged survival, whereas additional CpG injections concomitant with ablation resulted in protective immunity in 100% of the mice (Fig. 1A). Remarkably, treatment of mice with CpG injections either 1 or 3 days before or after ablation was much less effective as only 25% to 50% of the mice survived the challenge (Fig. 1B).

Analyses of OVA-specific CD8⁺ T cells revealed the superior induction of CTL when CpG was administered immediately after ablation (CpG on day 0: 2.4 ± 0.6% versus day −3: 0.8 ± 0.3%, day −1: 1.4 ± 0.7%, day +1: 0.6 ± 0.2%, and day +3: 0.6 ± 0.4%; Fig. 2A–C). These data indicate that the timing of CpG administrations relative to antigen availability is decisive for the efficient induction of tumor-specific CTL and subsequent antitumor immunity.

Timing of CpG injection has minor effects on DC maturation and loading with antigen and CpG. To elucidate the underlying mechanisms of protective CTL induction, we studied the quality of DCs following CpG administration in terms of the up-regulation of the costimulatory molecule CD80. To be able to discriminate between antigen⁺ and antigen⁻ DC in the draining lymph node, tumors were injected with Alexa488-conjugated OVA protein just before tumor destruction. The results show that cryoablation or CpG alone slightly increased CD80 expression on antigen-loaded and unloaded DC, although the expression levels in the latter were markedly lower. When the two treatment modalities were combined, CD80 expression was synergistically enhanced irrespective of the timing of CpG treatment (Fig. 3A). These data show that timing of CpG administrations relative to antigen release does not affect the up-regulation of CD80.

DCs were highly successful in collecting both antigen (10–19% of all DCs) and CpG (66–93% of all DCs) in an in situ tumor ablation–dependent manner. CpG enhanced the percentages of antigen-loaded DCs but only when administered concurrent with ablation (19%) or after ablation (16%). Injection of CpG before ablation did not have any effect on antigen uptake when compared with cryoablation alone (10%; Supplementary Fig. S1A). All antigen-loaded DCs taken from mice receiving any of the combinational treatments also stained positive for CpG, but mean fluorescence levels for CpG were the highest in cells from simultaneously treated mice (Fig. 3B; Supplementary Fig. S1B). Similar results were obtained in in vitro experiments in which all DCs were positive for both CpG and antigen (data not shown).

From these data, we conclude that the differences in the induction of antitumor immunity cannot solely be explained by quantitative differences in antigen/CpG loading or expression of costimulatory molecules by DCs.

Timing of CpG administrations crucially affects cross-presentation. The priming of CTL depends on the unique feature of DCs to present captured antigens on MHC class I, a process known as cross-presentation (26). We next tested the cross-presenting abilities of lymph node DCs directly ex vivo on coculture with B3Z cells. As B3Z is a MHC class I–restricted OVA-specific T-cell hybridoma, it is readily activated by K_b-OVA complexes independent of costimulatory molecules or cytokines (23). The data show that DCs from mice treated with cryoablation alone were unable to activate B3Z cells and hence to cross-present antigen (Fig. 4). Remarkably, additional CpG injections day(s) before or after ablation failed to improve cross-presentation by these lymph node DCs. In contrast, DCs isolated from mice injected with CpG at the time of antigen release were able to activate B3Z cells and hence were capable of cross-presentation. To exclude the possibility that a quantitative rather than a qualitative effect was responsible for the differences in cross-presentation, we increased the numbers of CD11c⁺ cells in B3Z cocultures. Again, cross-presentation was uniquely induced by lymph node DCs from mice exposed to antigen and CpG at the same time (data not shown). Thus, the ability of DCs to effectively cross-present antigen is highly dependent on the timing of CpG administration.

Cross-priming ability of DCs depends on intra-DC colocalization of antigen and CpG. As these in vivo and in vitro data imply that DCs that have encountered antigen and TLR agonist at the same time are qualitatively distinct, we determined the fate of antigen and TLR agonist within DCs. As a consequence of cryoablation, tumor cells are opened up and antigenic material, among which soluble antigens, is released (27). To mimic the situation of ablated soluble antigens as much as possible, tumor-bearing mice received an intratumoral injection with OVA-Alexa488 just seconds before ablation and were additionally treated with Cy5-labeled CpG either before or concurrent with tumor destruction.

OVA⁺CpG⁺ DCs from tumor-draining lymph nodes were isolated and purified by FACS sorting and then analyzed for the localization of antigen and CpG inside the DCs. Strikingly, lymph node DCs from mice treated with ablation and concurrent CpG injections showed very clear colocalization of antigen and CpG in essentially all DCs (Fig. 5A). In contrast, 20% of the DCs derived from mice treated with CpG before ablation showed only partial colocalization, whereas in 80% of the DCs no colocalization was observed (Fig. 5B). Recent in vitro studies showed that antigen present in EEA-1⁺ endosomes is efficiently cross-presented, whereas pinocytosed material is directed to LAMP-1⁺ compartments for presentation to CD4⁺ T cells (28). We found that colocalization of antigen and CpG was present in EEA-1⁺ and in LAMP-1⁺ compartments (Fig. 5C), indicating that antigen may be internalized via different endocytic routes.

By calculating the correlation coefficients for antigen/CpG in multiple cells, we observed that the colocalization of antigen/CpG in cells from mice treated with the combinational therapy on day 0 was significantly higher than when the treatments were separated in time (Fig. 6). In fact, cells with colocalization from mice treated with CpG before ablation displayed significantly lower correlation coefficients than cells taken from mice treated simultaneously with cryoablation plus CpG. We also compared the CTL responses in mice that were treated with exogenous soluble OVA and CpG at the same time and mice that were treated with CpG before OVA injection. Simultaneous exposure resulted in higher numbers of antigen-specific IFN-γ–producing CTL in spleen and draining lymph nodes when compared with separate exposures in time (data not shown).

These results thus show that colocalization of antigen and CpG in EEA-1 or LAMP-1⁺ DC compartments in vivo highly correlates with the ability of DCs to cross-present antigen to CD8⁺ T cells resulting in increased protective immunity against tumors.

Here, we show that CpG administration synergizes with cryoablation of solid tumors to induce long-lasting antitumor immunity provided that CpG is administered concurrent with ablation. Remarkably, the differences in the efficacy of treatments closely correlate with the colocalization of antigen and CpG within endosomal DC compartments. These data are supportive of the view that colocalization of antigen and CpG within DCs in vivo is a prerequisite for DC cross-presentation and the induction of CTL.

Immune stimulation by TLR agonists have been intensively studied but the mechanisms to efficiently enhance tumor-specific CTL responses in vivo remain elusive. TLR triggering of DCs programs the expression of costimulatory molecules and cytokines (29) and provides signals that affect the membrane vacuolar system, the cytoskeleton, and the machinery of protein translation and degradation (30). Our data show that the timing of CpG administration does not affect the expression of costimulatory molecules but influences the internalization of antigen and CpG. DCs encountering CpG in vivo before ablation internalize less antigen compared with DCs from mice treated with CpG concurrent with or after ablation. These findings are consistent with previous studies showing that TLR triggering first increases the endocytic capacity and then decreases phagocytosis by DCs (31, 32).

Applying fluorescently labeled CpG at different time points in vivo, we additionally show that the majority of antigen-loaded DCs are capable of internalizing CpG irrespective of timing. Internalization of antigen and CpG seems, however, not sufficient to induce efficient DC cross-presentation in vivo and in vitro. Only lymph node DCs isolated from mice treated with CpG concurrent with ablation are able to cross-present antigen to the OVA-specific T-cell hybridoma B3Z. In contrast, CpG administration after ablation also enhances antigen internalization but fails to induce cross-presentation in lymph node DCs. Moreover, in vitro experiments in which 100% of the DCs are loaded with antigen confirmed that DCs simultaneously exposed to antigen and CpG are most effective in cross-presentation (data not shown). Together, these data imply that the expression of costimulatory molecules and the level of antigen and/or CpG internalization are not decisive for cross-presentation in this setting.

Recently, Blander and Medzhitov (33) suggested from in vitro data that the presence of TLR agonists and antigenic cargo in the same phagosome is crucial for efficient antigen presentation and subsequent priming of CD4⁺ T cells by DCs. This finding points to a plausible mechanism in which efficient T helper cell induction is dependent not only on the expression of costimulatory molecules and cytokines by DCs but also on phagosome maturation. Whether this also holds true for antigen cross-presentation has been suggested but has never been unequivocally shown. We found that colocalization of antigen and CpG was present in EEA-1⁺ compartments, which were recently identified to enable antigens to access the cross-presentation pathway in vitro (28). Hence, our study provides in vivo evidence that the presence of antigen and TLR agonist within the same compartment not only is a pivotal checkpoint for the induction of MHC class II presentation (33) but also allows the antigen present in the same endosomal compartment to gain access to the MHC class I pathway.

Tumor debris after cryoablation contains a complex mixture of numerous different antigens; some will behave as soluble antigens and others will behave as cell-associated antigens. In our studies, we used soluble OVA as a representative marker for soluble antigens that are rapidly released from the tumor following cryoablation. For instance, Huang and colleagues (27) reported a steep rise in serum levels of carcinoembryonic antigen, a cell surface 200-kDa glycoprotein, after cryotherapy in patients with hepatic metastasis from colorectal cancer. Comparable observations were reported by Ghanamah and colleagues (34) after laparoscopic radiofrequency of hepatic metastasis. Whether similar mechanisms are at play in the processing of cell-associated antigens remains to be assessed.

Based on the present results, one would thus predict that TLR-based immunotherapy only occurs in optima forma when DCs internalize antigen and TLR agonist simultaneously. Indeed, antigens covalently linked to TLR agonists induce a stronger immune response than simultaneously but separately administered compounds in mice (35, 36). In addition, intratumoral (2), peritumoral (4, 37), or intralymphatic (18) administrations of CpG in vivo were efficient in improving T-cell–mediated immunity, whereas systemic (i.v. or i.p.) application of CpG in vivo has been shown to result in T-cell suppression rather than immune activation (38–40). In this context, the timing and location of CpG administration will be an important element in the design of future clinical trials.

We note that the possibility to control the release of antigen by in situ tumor ablation as described herein may represent the ideal setting to combine with CpG as this will maximize loading of DCs with both tumor antigens and TLR agonist. As both treatment modalities are already applied in the clinic, it is of great interest to determine the efficacy of CpG in combinational treatment with tumor-ablative strategies in cancer patients.

No potential conflicts of interest were disclosed.

Grant support: Dutch Cancer Society grants KWF-2003-2893, KWF-2004-3137, and KWF-2005-3325. S. Nierkens is a research fellow of the Dutch Cancer Society.

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

We thank Ben Joosten, Liza Toonen, and Rob Woestenenk for technical assistance.