Dendritic Cell–Based Vaccines that Utilize Myeloid Rather than Plasmacytoid Cells Offer a Superior Survival Advantage in Malignant Glioma

Plasmacytoid dendritic cells (pDCs) are a unique subset of DCs that have been traditionally defined as professional IFN-α–producing cells, which upon activation by infectious agents produce large amounts of the powerful proinflammatory cytokine IFN-α. Owing to its ability to effectively stimulate host immune response in the setting of various solid tumors (1–3), IFN-α was the first cytokine to be approved for cancer treatment (4). IFN-α production by pDCs is regulated by downstream signaling via the universal adapter protein MyD88. Upon stimulation of TLR-7 and TLR-9, MyD88 acts via the constitutively expressed transcription factor IFN regulatory factor 7 and the inflammatory transcription factor NF-κB, thereby initiating transcription of type I IFN (5). Recent studies indicate that pDCs also express MHC class II (MHC-II) molecules, undergo a maturation process similar to that of myeloid DCs (mDCs), can function as APCs, and promote T cell–mediated immunity or the maintenance of self-tolerance (6–8). Whether the maturation of pDCs leads to immunity or tolerance depends on the context of Ag presentation. On the one hand, pDCs can promote tolerance by presenting Ags to CD4⁺ T cells and inhibiting their activation (7) or inducing regulatory T cells (Tregs), which confer tolerance to cardiac allografts (9), prevent asthmatic reactions to inhaled Ags (10), and promote tumor progression in several solid tumors (11–13). On the other hand, activated pDCs can produce large amounts of IFN-α (1), induce strong allogeneic T cell responses (14), and prime CD4⁺ and CD8⁺ T cells against viruses (15) or tumor Ags (16) and induce an effective antitumor response (17, 18).

In several solid tumors, pDC infiltration has been associated with disease progression and poor prognosis and has been shown to promote tumor immunosuppression (11, 19, 20). Suggested mechanisms for tumor-induced pDC dysfunction include recruitment of immature pDCs, lack of expression of costimulatory molecules such as CD80, CD83, and CD86 (21, 22), and altered functioning of existing pDCs, such as diminished IFN-α secretion (21), mediated by tumor-associated immunosuppressive cytokines. pDCs also contribute to tumor-associated immunosuppression by mediating mature Treg accumulation (23) and by releasing IDO, which is a powerful promoter of Treg activation (24). Impaired production of IFN-α and accumulation of Tregs significantly impair local immune surveillance, allowing tumors to escape IFN-α–associated immune responses.

Although the role of pDCs as impaired professional IFN-α–producing cells is well characterized in the context of tumors, their role as professional APCs is not. Both in humans and mice, the entire DC pool can be divided into two subsets: a larger subset consisting of mDCs and a smaller subset consisting of pDCs. Murine pDCs are characterized by expression of CD11c and BST-2 (CD317) and mDCs by expression of CD11c and are very distinct in their function and characteristics. Classically, mDCs have been shown to play the critical role in the induction of immunity or tolerance. Because of this decisive role played by mDCs in the induction of immunity, DCs have been widely used as a vehicle for DC-mediated immunotherapy, in which DCs loaded with tumor Ags are injected into patients with cancer to stimulate the antitumor T cell response (25). Because the number of circulating natural DCs is low, virtually all vaccination studies use DCs differentiated ex vivo from monocytes or CD34⁺ progenitors (25). However, recent reports suggest that DCs matured ex vivo are less effective than are their natural counterparts in activating T cells and inducing effective antitumor immunity (26–28). There has been no study so far that compares the capacity of mDCs and pDCs to activate and prime naive T cells.

Malignant gliomas (MGs), consisting of anaplastic astrocytoma (World Health Organization grade III) and glioblastoma multiforme (GBM) (WHO grade IV), are the most common primary brain tumors in adults and are associated with dismal prognosis (29). MGs are associated with a potently immunosuppressive tumor microenvironment and efficiently evade the host antitumor response. We have previously shown that one of the hallmark features of glioma immunosuppression is the presence of Tregs (30–34). Besides the presence of Tregs, several immune modulating mechanisms have been implicated in potentiating the immunosuppressive glioma microenvironment, including the suppression of APC functions via expression of immunosuppressive cytokines, such as IL-10 and TGF-β, which contributes to the inhibition of the effector T cells (30). Even though the paradigm of immune privilege suggests that classical DCs are absent from the brain (35), recent reports have revealed that both pDCs and mDCs are present in human brain, which may contribute to orchestration of the local immune response (36–39).

In this study, we show that human grade III MGs have the highest infiltration of pDCs. In the murine model of glioma, intracranial (i.c.) tumor implantation leads to selective maturation of pDCs, characterized by upregulation of MHC-II and B7-H1 (CD 274) on pDCs. Glioma-infiltrating pDCs are deficient in producing IFN-α, and the selective depletion of pDCs during the course of disease in BDCA2-DTR transgenic (Tg) mice (in which injection of diphtheria toxin [DT] in the mice results in selective depletion of pDCs) (40) results in increased median survival of the mice bearing i.c. tumor. pDC depletion also leads to decrease in the number of ICOS⁺ Tregs in the brain of glioma-bearing mice, and the Tregs from BDCA-2 DTR Tg mice are less suppressive compared with the Tregs from wild-type (WT) mice. These finding suggest that in the initial stages of glioma progression, pDCs skew the immune response toward tolerance, rather than the efficient induction of antiglioma immunity. Because our results showed that selectively pDCs undergo maturation in the context of glioma and contribute to glioma-mediated immunosuppression, we compared the immune response generated by pDCs versus mDCs in a DC-based vaccine strategy. We found that mDCs are much better at inducing the antiglioma Th1 immune response when compared with pDCs, and 60% of the mice vaccinated with mDCs survived long-term. In conclusion, our studies indicate that host pDCs promote glioma progression in the murine model of glioma and, in the context of DC-based vaccination pDCs, are less effective than mDCs in generating an antiglioma response.

C57BL/6 (WT) and BDCA-2–DTR (C57BL/6 background) mice were purchased from The Jackson Laboratory (Bar Harbor, ME) and maintained in the University of Chicago Carlson Barrier Facility. BDCA-2–DTR is a transgenic mouse model where the pDC-specific promoter BDCA-2 is under the control of the DT receptor (DTR) promoter. Administration of i.p. DT in these mice results in selective depletion of pDCs (40). All mice were i.c. injected with syngeneic GL261 or GL261-OVA cells (4 × 10⁵ or 2 × 10⁵ cells) between the ages of 6 and 8 wk, as described previously (30). All animal work was reviewed and approved by the University of Chicago Institutional Animal Care and Use Committee. All surgical procedures were completed in accordance with National Institutes of Health guidelines on the care and use of laboratory animals for research purposes. Mice were euthanized by CO₂ and then by cervical dislocation. Following i.c. injection, pDC depletion was carried out by i.p. injection of DT (Sigma-Aldrich, St. Louis, MO) at 100–120 ng/mouse as previously described (40). For vaccination, mice were intradermally injected in the hindleg (41) with pDCs or mDCs in PBS.

GL261 cells were obtained from the National Cancer Institute Frederick National Tumor Repository Laboratory and cultured in DMEM supplemented with 10% FCS, as well as streptomycin (100 mg/ml) and penicillin (100 U/ml), at 37°C in a humidified atmosphere of 95% air/5% CO₂. GL261-OVA cells, a GL261 line stably expressing pCDNA 3.1 containing artificial model Ag (chicken OVA Ag), were generated in-house by selecting the stably transfected cells with media containing 200 μg/ml G418 (Geneticin). Chicken OVA cDNA was amplified by PCR using pAc-neo-OVA as a template. The 5′ primer sequence was 5′-AACGCGGATCCACCATGGGCTCCATCGGCGC-3′, and the 3′ primer sequence was 5′-GAGCACCGCTCGAGTTTTTAAGGGGAAACACATC-3′. OVA expression plasmid was created by ligation of OVA cDNA into pcDNA3.1/Hygro(+) (Invitrogen) vector between BamHI and XhoI sites. To create GL261-OVA cell lines, OVA plasmid was used to transfect GL-261 cell lines and selected using hygromycin B (Invitrogen). OVA expression (45 kDa) in GL261-OVA cell lines was confirmed by Western blotting (data not shown) using anti-OVA mAbs obtained from Sigma-Aldrich. After transfection, stable clones were isolated by a combination of drug selection, flow cytometry sorting with the H-2K^b-SIINFEKL–specific Ab 25.D1.16, and cell cloning at limiting dilution. The OVA expression was validated by immunoblot analysis. All cell culture products were purchased from Life Technologies/Invitrogen (Grand Islandm NY).

Single-cell suspensions were made from brain, cervical lymph node (cLN), or spleen as described in our previous publication (30). For cellular staining, cells were incubated with: anti-CD4 (RM4-5; eBioscience, San Diego, CA), anti-CD3 (145-2C11; eBioscience), anti-CD8 (53-6.7; eBioscience), PD1 (J43; eBioscience), CD45 (30-F11; eBioscience), BST-2 (eBio927; eBioscience), ICOS (C398-4A; eBioscience), TLR-9 (M9.D6; eBioscience), MHC-II (M5/114.15.2; eBioscience), Lag-3 (C9B7W; eBioscience), TIM (8B.2C12; eBioscience), CD11c (N418; eBioscience), B7-H1 (MIH5; eBioscience), CD45.2 (104; eBioscience), TGF-β (TW7-20B9; BioLegend, San Diego, CA), and CD45.1 (A20; eBioscience) in PBS plus 2% BSA (Sigma-Aldrich) for 30 min on ice. For intracellular cytokine analysis, T cells were treated with cell stimulation mixture with protein transport inhibitors (PMA/ionomycin with brefeldin A and monensin) (eBioscience) for 5 h, and pDCs were stimulated with type A CpG with protein transport inhibitors for 7 h at 37°C in a humidified atmosphere of 95% air/5% CO_2. Cells were then fixed and permeabilized overnight at 4°C using Fix/Perm buffer (eBioscience) according to the manufacturer’s instructions and stained with intracellular Abs: anti–Foxp3-FITC/allophycocyanin (FJK-16s; eBioscience), IFN-α (RMMA-1; PBL Interferon Source, Piscataway, NJ), Tbet (eBio4B10; eBioscience), IL-10 (JES5-16E3; BD Biosciences, San Jose, CA), and TGF-β (TW7-20B9; BioLegend, San Diego, CA) for 30 min on ice. All flow cytometric analyses were done using a BD FACSCalibur (BD Biosciences) flow cytometer. For in vitro T cell suppression assays, single-cell suspensions were made from LN and spleen and then were sorted into CD4⁺CD25⁺ and CD4⁺CD25⁻ cells.

pDCs (CD3⁻CD45⁺CD11c⁺BST-2⁺) and mDCs (CD3⁻CD45⁺CD11c⁺BST-2⁻) were directly isolated from pooled spleen and LN of mice using the fully closed BD FACSAria cell sorter (BD Biosciences). This procedure resulted in clinically applicable purified pDCs, which had an average purity of 85%. Following isolation, pDCs and mDCs were cultured overnight at a concentration of 10⁶ cells/ml in X-VIVO 15 (Lonza, Hopkinton, MA) containing 2% FBS, supplemented with 10 ng/ml recombinant murine IL-3 (PeproTech, Rocky Hill, NJ) or GM-CSF (PeproTech). For the vaccination, pDCs and mDCs were subsequently activated for 6 h by addition of type A CpG (oligodeoxynucleotide 1585; InvivoGen, San Diego, CA) or LPS-EB (tlrl-eblps; InvivoGen). During the last 3 h of activation, pDCs and mDCs were loaded with the OVA_257–264 peptide (Anaspec, Fremont, CA) (42). The peptide-loaded pDCs (5000 cells) and mDCs (5000 cells) were administered intradermally in the hindleg of the mouse. This procedure gave rise to mature pDCs and mDCs meeting the following release criteria: >50% viability, IFN-α secretion, and high expression of MHC class I, MHC-II, CD83, CD80, and CD86 as previously described by Tel et al. (42).

In vitro suppression assays were carried out in RPMI 1640/10% FCS in 96-well V-bottom plates (Costar, Corning, NY) with 1 × 10⁶ CD4⁺ responder cells, titrated amounts of FACS-sorted CD4⁺CD25⁺ cells, and 4 × 10⁶ irradiated (2500 rad) sorted APCs (CD4⁻ splenocytes). CD4⁺ cells were labeled with CellTrace Violet (C34557; Life Technologies, Carlsbad CA). Stimulation was carried out with plate-bound anti-CD3 (145-2C11; 1 μg/ml) and after 72 h at 37°C, proliferation was determined using FACS.

mDC and pDC populations were sorted from mice 1 wk after GL261 i.c. tumor implantation. DCs were then pulsed with SIINFEKL (2 μm) for 4 h. MACS-purified OT-1 T cells were then labeled with CTV proliferation dye and cocultured with the pulsed DCs for 48 h. Cells were then harvested and assessed via flow cytometry for number of cell divisions.

A human glioma microarray with normal brain control (25 tumor samples plus normal brain control) obtained from US Biomax (Rockville, MD) was deparaffinized in xylene and then rehydrated. After deparaffinization and rehydration, the microarray was treated with Ag retrieval buffer (S1699; DAKO, Carpenteria, CA) in a steamer for 20 min. Human anti–BDCA-2 mAb (clone 10E6.1) was obtained from Millipore (Billerica, MA). The Ab was applied (1:100) on the tissue microarray for 1 h at room temperature. Normal human tonsil tissue was used as positive control (as suggested by the manufacturer’s instructions). The Ag/Ab binding was detected by a Bond polymer refine detection system (DS9800; Leica Microsystems, Buffalo Grove, IL).

Survival was defined as the time from injection of GL261/GL261-OVA cells to day 150 of the time course. Kaplan–Meier curves were generated and the survival distributions were compared using a log-rank test. All other data are presented as mean ± SEM. Comparisons between two groups were conducted using Student t test or Mann–Whitney U test as appropriate, and differences between more than two groups were assessed using ANOVA with a Tukey post hoc test. All analyses were conducted using GraphPad Prism version 4.0 (GraphPad Software). All reported p values are two-sided and are considered to be statistically significant at p < 0.05.

pDCs have been implicated to play a critical role in many solid tumors (11, 13, 43); however, the role of pDCs in glioma progression is unknown. Hussain et al. (44) analyzed DC infiltration in human GBM sample and normal brain by flow cytometry and showed that there were fewer pDCs in GBM (grade IV) specimen than in normal brain. To establish the correlation between pDC infiltration and all glioma grades, human glioma tissue microarray, containing 25 samples of all representative glioma grades, was stained with anti–BDCA-2 Ab. The level of staining was initially graded by the author and then by an independent, blinded neuropathologist on a scale of 0–5, where 0 signifies no staining and 5 signifies the highest amount of staining (positive control). In human glioma, pDC infiltration steadily and significantly increased from grade I to III (Fig. 1A–D); however, in accordance with previous observation, no identifiable pDC staining was observed in grade IV (Fig. 1E). There was a statistically significant difference in pDC staining between low-grade (grade II) and high-grade anaplastic (grade III) glioma (p < 0.05) (Fig. 1F).

The presence of pDCs and their phenotype has been well characterized in murine bone marrow (BM) and various other organs, and their role has been implicated in the pathophysiology of several solid tumors (19, 45, 46). To delineate the DC population in the brain of naive mice and compare that to other organs, DCs from brain, cLN, and spleen were compared. Most (>80%) DCs in the cLN and spleen were CD11c⁺BST-2⁻ mDCs. Interestingly, in the brain, DCs were split almost equally between CD11c⁺BST-2⁺ pDCs and CD11c⁺BST-2^- mDCs (Fig. 2A). The phenotype of the DCs from each tissue was analyzed for the presence of activating (MHC-II, CD80) or inhibitory markers (B7-H1). There was no statistically significant difference noted in the expression of the three molecules between pDCs and mDCs from the cLN and spleen. In the brain, there was no statistically significant difference between MHC-II and CD80 expression; however, there was a significantly higher expression of the inhibitory molecule B7-H1 (mean fluorescence intensity [MFI] of 2636 ± 500.5 for pDCs versus 1300 ± 185.1 for mDCs; p < 0.01) on pDCs, when compared with the mDCs (Fig. 2B). Thus, our data show that there are resident pDCs and mDCs in the naive mice brain, and at baseline, brain-resident pDCs express higher levels of the inhibitory marker B7-H1 when compared with mDCs.

Next, to further outline which DC population acts as major APC in the orthotropic GL261 cell–based murine model of glioma, both DC cell populations were analyzed from the brain and cLN of tumor bearing mice at 1 and 3 wk after tumor implantation (wpo) (Fig. 3). There was no significant difference in the pDC or mDC frequency in the brain at 1 wpo. However, the frequency of pDCs in the tumor draining cLN increased at 1 wpo (6.42 ± 1.5% for naive versus 13.45 ± 1.67% for 1 wpo; p < 0.01). By week 3 of tumor progression, the frequency pDCs decreased drastically in the brain compared with 1 wpo (62.83 ± 3.283% for 1 wpo versus 3.09 ± 2.09% for 3 wpo; p < 0.01) and the frequency of mDCs increased (34.36 ± 3.087% for 1 wpo versus 95.10 ± 2.34% for 3 wpo; p < 0.001). The same trend in pDC (13.46 ± 1.7% for 1 wpo versus 6.6 ± 1.1% for 3 wpo; p < 0.01) and mDC (83.97 ± 2.0% for 1 wpo versus 90.43 ± 1.25% for 3 wpo; p < 0.05) frequency was observed in the cLN (Fig. 3A). There was a marked increase in MHC-II expression by pDCs in the brain (3,000 ± 249 for naive versus 17,509 ± 3,181 for 1 wpo; p < 0.01) and cLN (3,094 ± 1,539 for naive versus 17,293 ± 1,767 for 1 wpo; p < 0.01) at 1 wpo compared with naive control; however, there was no difference in MHC-II expression by mDCs in the brain and a slight increase in MHC-II expression by mDCs in the cLN (884.7 ± 67.7 for naive versus 5794 ± 744.6 for 1 wpo; p < 0.01). There was also a statistically significant increase in B7-H1 expression on tumor infiltrating pDCs (3261 ± 187 for naive versus 4712 ± 387.7 for 1 wpo; p < 0.05) in the brain and cLN (929.7 ± 473.7 for naive versus 2507 ± 214.2 for 1 wpo; p < 0.05). Both in naive mice (3261 ± 187 pDCs versus 1557 ± 132.1 mDCs ; p < 0.01) and 1 wpo (4712 ± 387.7 pDCs versus 2062 ± 143.1 mDCs ; p < 0.001) pDCs in the brain had higher expression of B7-H1 compared with mDCs (Fig. 3B). Collectively, these in vivo data show that in the context of the orthotopic murine model of glioma, pDCs undergo maturation by upregulating MHC-II along with upregulation of inhibitory molecule B7-H1. To analyze which group of DCs is better at Ag presentation and stimulate immune response versus which group is more inhibitory, we compared the ability of pDCs and mDCs to present Ag and stimulate proliferation of OT-1 T cells in vitro. Our results show that in vitro–pulsed mDCs from tumor-bearing mice are significantly better at presenting Ag to cause T cell proliferation than are pDCs (Fig. 3C). Thus, we concluded that pDCs are more inhibitory when compared with mDCs and most likely contribute to the profound immunosuppression associated with malignant glioma.

In response to a variety of pathogenic stimulation, pDCs are known to produce large amounts of IFN-α via TLR-9 and TLR-7 signaling (40, 47). Their ability to produce IFN- α varies depending on tissue location. In mice, BM-resident pDCs are the major source of IFN-α; however, the pDCs from LN or spleen are not (45). To analyze IFN-α production by brain pDCs in naive mice and in tumor-bearing mice, pDCs isolated from the brain of tumor-bearing mice and naive mice were stimulated with type A CpG and stained for IFN-α. In accordance with the previously published literature (45), in naive mice, a high fraction of pDCs from BM (42.6 ± 2.7%) produced IFN-α when stimulated with CpG, although pDCs from cLN (1.8 ± 0.33%) did not. Interestingly, in naive mice, most pDCs from the brain (57.6 ± 15.77%) also produced a high level of IFN-α. One week after tumor implantation, there was a significant decrease in IFN-α production by pDCs from the brain (57.63 ± 15.77 for naive versus 10.9 ± 6.4% for 1 wpo; p < 0.05), BM (42.55 ± 2.7% for naive versus 1.4 ± 0.16% for 1 wpo; p < 0.001), and cLN (1.8 ± 0.33% for naive versus 0.5 ± 0.2% for 1 wpo; p < 0.05), compared with the naive mice (Fig. 4A). To assess whether the observed decrease in IFN-α secretion by tumor-infiltrating pDCs is due to change in TLR-9/TLR-7 expression, pDCs from the brain, cLN, and BM of naive and 1 wpo mice were analyzed for TLR-9/TLR-7 expression. There was no difference in TLR-7 expression by pDCs between naive mice and tumor-bearing mice; however, compared with naive pDCs, brain tumor-infiltrating pDCs had significantly lower expression of TLR-9 (75.6 ± 4.4% for naive versus 3.7 ± 1.5% for 1 wpo; p < 0.001). The presence of brain tumor also decreased TLR-9 expression by pDCs from cLN (3.13 ± 0.7% for naive versus 2.61 ± 0.61% for 1 wpo) and BM (46.6 ± 7.6% for naive versus 30.5 ± 8.8% for 1 wpo) (Fig. 4B). These results indicate that presence of i.c. tumor leads to decreased expression of TLR-9 and decreased production of IFN-α by pDCs. In several solid tumors, tumor-derived inhibitory cytokines, such as TGF-β and IL-10, have been shown to influence TLR-9 expression and decrease IFN-α production by pDCs (46). We have previously shown that glioma-infiltrating immune cells are major source of IL-10 and TGF-β (30). We thus conclude that in the setting of malignant glioma, glioma-derived suppressive cytokines decrease TLR-9 expression and IFN-α production by pDCs.

IFN-α secreted by pDCs has been known to be an effective antiglioma therapeutic agent (1). In the study by Candolfi et al. (1) the authors show that pDCs can be recruited in the glioma microenvironment by gene therapy using cytokine Flt3L, which effectively secreted IFN-α and led to increased survival in animals bearing intracranial glioma. Our results show that the presence of brain tumor inhibits the capability of brain-resident pDCs to produce IFN-α during the early priming phase of the glioma. Thus, we hypothesized that presence of pDCs deficient in their ability to secrete IFN-α contributes to the tolerogenic glioma microenvironment. To test this hypothesis, pDCs were selectively depleted by injection of DT in the BDCA-2 DTR Tg mice bearing i.c. glioma, and animals were followed for survival. Compared to the control group, pDC depletion in BDCA-2 DTR Tg mice with glioma resulted in a significant increase in median survival from 19 to 30 d (p < 0.05) (Fig. 5A). Besides their ability to produce a large amount of IFN-α, pDCs are also capable of internalizing, processing, and presenting Ags to CD4⁺ T cells and cross-present Ags to CD8⁺ T cells (48–50).

To understand the consequence of pDC depletion and the observed prolongation of survival, T cells from brain, cLN, and spleen were analyzed at 2 wpo from WT mice and BDCA-2 DTR Tg mice bearing tumor. pDC depletion resulted in significant decrease in Treg numbers in the brain of the BDCA-2 DTR Tg mice compare with the WT mice (40,310 ± 8,937 for WT versus 11,606 ± 4,178 BDCA-2 for DTR; p < 0.05) (Fig. 5B). pDCs expressing the ICOS molecule ligand have been reported to strongly favor ICOS⁺ Treg proliferation (51). Phenotypic analysis of the glioma-infiltrating Tregs showed that there was a statistically significant decrease in ICOS⁺ Tregs numbers in BDCA-2 DTR mice compared with WT mice (3954.5 ± 996.2 for WT versus 1265 ± 376.7 for BDCA-2 DTR; p < 0.05) (Fig. 5C). In terms of critical cytokine production, CD4⁺ cells from the brain of BDCA-2 DTR mice with glioma were found to produce significantly less immunosuppressive cytokine IL-10 compared with WT mice (13.24 ± 0.44% for WT versus 6.84 ± 0.74% for BDCA-2 DTR; p < 0.001) (Fig. 5D). Through an in vitro Treg suppression assay, where CD4⁺ cells and Tregs from the tumor-bearing WT and BDCA-2 DTR mice were cocultured and CD4⁺ cell proliferation was measured, we showed that Tregs from BDCA-2 DTR mice were significantly less suppressive and CD4⁺ cells were more proliferative compared with WT mice at every CD4/Treg ratio (Fig. 5E). Collectively, our data from the selective pDC depletion experiment in the setting of glioma shows that pDC depletion results in increased median survival of the mice bearing i.c. tumor. pDC depletion in the context of glioma results in decreased immunosuppressive cytokine, IL-10, and production by CD4⁺ T cells, decreased Treg and ICOS⁺ Treg numbers in the brain, and decreased suppressive capabilities of Tregs from the BDCA-2 DTR tumor-bearing mice. Hence, in the context of glioma, pDCs deficient in IFN-α production help to maintain a large number of highly suppressive Tregs in the tumor.

DCs have been harvested and used in anticancer therapeutic strategy for their unique capacity to process Ags and to present them to naive T cells and activate targeted antitumor response. Our data collectively showed that in the context of glioma, pDCs set a tolerizing glioma microenvironment, which helps promote glioma-associated immunosuppression. Thus, we hypothesized that pDCs and mDCs process Ag differentially and activate different arms of the immune system. Using a GL261-OVA Ag murine model of glioma, mice bearing GL261-OVA tumor (2 × 10⁵ cells) were immunized with stimulated mature pDCs or mDCs. To analyze the migratory ability and distribution of DCs following immunization of tumor-bearing mice, mature pDCs and mDCs were harvested from spleen and LN of WT (CD45.2) mice, and DCs were stimulated overnight and loaded with SIINFEKL peptide (42). Stimulated and loaded DCs were injected in the hindleg of CD45.1 tumor-bearing mice at days 1, 2, and 6 after i.c. tumor implantation. Analysis of tissue distribution of injected CD45.2 DCs 1 wk after immunization showed that both pDCs and mDCs migrated systemically to brain, cLN, spleen, and inguinal LN (Fig. 6A). A relative higher percentage of both pDCs and mDCs migrated to the brain compared with other lymphatic organs (Fig. 6B). Vaccination with mDCs resulted in significant survival advantage in tumor-bearing GL261-OVA tumor, where median survival of mice vaccinated with PBS was 26 d, pDCs was 36.5 d, and 60% of the mice vaccinated with mDCs were alive at 60 d after tumor implantation (p < 0.001; mDC versus pDC and PBS) (Fig. 6C). These results show that both Ag-loaded pDCs and mDCs, when injected intradermally, can migrate systemically and preferentially localize in the glioma microenvironment. Vaccination of tumor-bearing mice with pDCs results in worse survival outcome compared with the mice vaccinated with mDCs. Thus, in the setting of brain tumor, DC-based immunotherapy might benefit from selecting out pDCs and using only mDCs for Ag loading. However, further studies are necessary to solidify our findings.

To further dissect the difference between the abilities of pDCs and mDCs in directing a glioma-specific immune response and define the type of immune response generated by each group of DCs, mice bearing i.c. GL261 tumor-expressing OVA Ag were vaccinated with stimulated and Ag-loaded pDCs, mDCs, or PBS as per the previous protocol, and T cell analysis was done at 2 wpo. T cell analysis at 2 wpo showed that mice vaccinated with mDCs had significantly higher frequency of CD8⁺ T cells in the brain compared with the pDC or PBS group (0.9 ± 0.23% for PBS versus 0.96 ± 0.13% for pDCs versus 2.01 ± 0.16% for mDCs; p < 0.01) (Fig. 7A). Staining for SIINFEKL tetramer showed that there was a significantly higher frequency of Ag-specific CD8⁺ T cells in the brain of mice vaccinated with mDCs, compared with pDCs or PBS (6.75 ± 0.56% for PBS versus 9.73 ± 2.27% for pDCs versus 14.82 ± 1.24% for mDCs; p < 0.001) (Fig. 7B). To investigate the ability of the DCs to induce effective antitumor Th1 immune response, T cells were analyzed for Th1-specific transcription factor (T-bet) (52). Compared to mice vaccinated with pDCs and PBS, mice vaccinated with mDCs had a significant increase in CD4⁺T-bet⁺ (4.29 ± 1.06% for PBS versus 4.87 ± 2.28% for pDCs versus 35.98 ± 4.42% for mDCs; p < 0.001) and CD8⁺T-bet⁺ (2.15 ± 0.73% for PBS versus 1.63 ± 0.86% for pDCs versus 30.48 ± 8.13% for mDCs; p < 0.001) T cells in the brain (Fig. 7C). Previous reports have shown that tumor-infiltrating CD8⁺ T cells express high levels of inhibitory receptors and are incapable of tumor clearance (53). In our vaccination model, we show that CD8⁺ glioma-infiltrating T cells have lower expression of the inhibitory molecules in the group vaccinated with mDCs compared with the group vaccinated with pDCs (Fig. 7D). Collectively, the data suggest that mDCs are much more effective than pDCs in mounting effective antitumor Ag-specific immune responses, which translates in significant survival benefit.

One of the major barriers to the development of effective antiglioma immunotherapy is the profoundly immunosuppressive glioma microenvironment (54). Even though several processes have been implicated, the exact mechanism by which the presence of i.c. glioma leads to profound immunosuppression is not well understood. Self or foreign Ag processing and presentation by DCs is the critical step in determining whether activated DC and naive T cell interaction will result in T cell activation and induction of immunity or T cell anergy and induction of tolerance. For a long time it was thought that the CNS is an immune-privileged site and this immune privilege was attributed to lack of presence of DCs in the CNS. Recently, however, this notion has been challenged and proven wrong (39). DCs have been shown to be present in human brain and thus may contribute to orchestration of the local immune response (38). Therefore, better understanding of the immune milieu of the glioma microenvironment is essential for development of effective antiglioma immunotherapy. In this study we investigated the role of pDCs in glioma progression and studied the T cell immune profile generated by the different subsets of DCs in the context of DC-based vaccine. Our results show that pDCs are present in human glioma samples, with the highest level of staining being observed in grade III glioma, and no staining was seen in grade IV glioma. Interestingly, in line with the observation in the human glioma, in the GL261-based murine model of glioma we noticed the highest frequency of pDCs during the first week of glioma progression and almost no pDCs by week 3 of glioma progression. Even though both pDCs and mDCs are present in the brain of naive mice, presence of i.c. tumor selectively induces maturation of pDCs, with upregulation of MHC-II and B7-H1, but not mDCs. Thus, they are potentially the major APCs in the brain tumor where they present Ag in the context of an inhibitory costimulatory molecule B7-H1 during the early priming phase of the glioma progression, thus setting the stage for a tolerogenic glioma microenvironment that contributes to glioma progression. This observation that Ag presentation by pDCs in glioma leads to polarization of glioma microenvironment toward tolerization instead of immune activation needs to be further investigated to understand the exact mechanism of Ag presentation.

Phenotypic analysis of the tumor-infiltrating pDCs showed that presence of i.c. tumor drastically impairs the ability of tumor-infiltrating pDCs to make IFN-α. This impairment in IFN-α secretion was noted systemically and not just in tumor-infiltrating pDCs. Along with the decreased ability to make IFN-α, there was also a decrease in TLR-9 expression by pDCs mostly from the brain, but to some extent there was a systemic decrease in TLR-9 expression by pDCs. In connection of several solid tumors, pDCs are known to be defective in their ability to effectively produce IFN-α (13, 55). In accordance with these reports, our results also suggest that presence of glioma leads to decreased production of IFN-α by pDCs, which potentially could be due to downregulation of TLR-9 in the presence of glioma. The observed downregulation of TLR-9 and impairment of IFN-α production were not just localized to the glioma microenvironment but were systemic, suggesting that this is most likely mediated by soluble factors. This phenomenon of tumor-induced suppression of IFN-α has been previously described in several other cancers. In head and neck cancer, it has been demonstrated that tumor cells actively suppress IFN-α production by pDCs by decreasing TLR-9 expression and secreting inhibitory cytokine IL-10 (19). Breast cancer–derived TGF-β and TNF-α have been shown to directly compromise IFN-α secretion by pDCs by influencing TLR-9 downstream signaling (46). The presence of inhibitory cytokines such as TGF-β and IL-10 is a hallmark feature of malignant glioma (30). Thus, the defect in IFN-α production observed in glioma pDCs is most likely due to glioma-generated inhibitory cytokine–induced downregulation of TLR-9.

Our data collectively showed that, in the context of brain tumor, tumor-infiltrating pDCs are 1) present during early priming phase of the tumor progression, 2) selectively undergoing maturation, and 3) incapable of producing proinflammatory cytokine IFN-α. We investigated the effect of selective pDC depletion in the setting of glioma. Using BDCA-2 DTR Tg mice, we show that pDC depletion during the course of the disease progression provides a significant survival advantage. In the setting of several solid tumors, such as breast, ovarian, and head and neck tumors, pDCs have been shown to promote tolerance by inducing Tregs (11–13). In line with the present literature, our studies also show that pDC depletion results in decreased numbers of Tregs and ICOS⁺ Tregs in the brain tumor. Additionally, in the absence of pDCs, Tregs are also less immunosuppressive, as quantified by the in vitro suppression assays. Thus, we show that in the murine model of glioma, pDCs promote tumor tolerance by inducing Treg number and function. The observed survival advantage was present when pDCs were depleted during the first 3 wk after tumor implantation starting with day 0. However the survival advantage disappeared when the pDCs were depleted starting at week 2 or week 3 (data not shown). This suggests that pDCs are critical during the early priming stage of glioma progression and needs to be depleted before the priming stage; however, once pDCs set the stage for immunosuppression, their depletion is inconsequential for the course of disease progression. In the clinical setting it is impossible to manipulate the priming phase of the tumor, because the priming stage has passed by the time tumor is diagnosed. Hence, clinically targeting or manipulating pDCs in MG is not an effective therapeutic strategy. However, following glioma resection, use of DC-based immunotherapy might provide an attractive opportunity to reset and reprime the immune response. Thus, we further went on to characterize the immune response generated by the two different subset of DCs.

DCs, used as vehicles in current immunotherapy regimens, dictate the effectiveness of the antitumor T cell response. Because pDCs and mDCs act differentially in the presence of glioma, we decided to study the Ag-specific T cell response generated by the two subsets of DCs when used for DC-based vaccination. We show that both pDCs and mDCs effectively migrated to inguinal LN, cLN, spleen, and brain. Animals vaccinated with mDCs mounted a robust Ag-specific Th1-type T cell response that translated into survival of 60% of the animals beyond 60 d. However, pDCs were less effective in mounting an effective antiglioma response, and pDC vaccination did not provide any survival advantage compared with vaccination with PBS. Natural mDCs isolated from WT mice can be effectively stimulated to induce an antiglioma immune response and are more effective than pDCs when used in DC-based immunotherapy. This observation has tremendous clinical implications because virtually all current DC-based vaccines use ex vivo–differentiated cells that are considered to be less effective than their natural counterparts (27). Thus, using natural mDCs and selectively depleting pDCs from the DC pool can potentially enhance the efficacy of DC-based immunotherapy. Our results have significant impacts on the development of DC-based antiglioma vaccine strategies as we show that the type of DCs used for the vaccination significantly influence the antitumor immune response. Although this observation needs further testing and optimization in terms of Ag and timing of administration, it provides the essential groundwork for further investigation and maximization of antiglioma vaccine therapy.

The authors have no financial conflicts of interest.