Research ArticleImmunotherapy

Dendritic cell vaccines based on immunogenic cell death elicit danger signals and T cell–driven rejection of high-grade glioma

See allHide authors and affiliations

Science Translational Medicine  02 Mar 2016:
Vol. 8, Issue 328, pp. 328ra27
DOI: 10.1126/scitranslmed.aae0105

DAMPening Glioma

Dendritic cell (DC)–based vaccines have shown promise for treating high-grade glioma (HGG), but efficacy has been limited by antigenic heterogeneity of the tumors. Now, Garg et al. combine DC vaccines with hypericin-based photodynamic therapy–induced immunogenic cell death (ICD) to treat HGG in an animal model. They found that ICD-based DC vaccines improved survival, and that this effect was dependent on the cell-associated reactive oxygen species and release of damage-associated molecular patterns (DAMPs) acting as danger signals. These ICD-based DC vaccines synergized with standard-of-care therapy to further improve survival in HGG-bearing mice and shifted the tumor immune signature from regulatory to TH1/TH17, which is associated with positive outcome in patients.


The promise of dendritic cell (DC)–based immunotherapy has been established by two decades of translational research. Of the four malignancies most targeted with clinical DC immunotherapy, high-grade glioma (HGG) has shown the highest susceptibility. HGG-induced immunosuppression is a roadblock to immunotherapy, but may be overcome by the application of T helper 1 (TH1) immunity–biased, next-generation, DC immunotherapy. To this end, we combined DC immunotherapy with immunogenic cell death (ICD; a modality shown to induce TH1 immunity) induced by hypericin-based photodynamic therapy. In an orthotopic HGG mouse model involving prophylactic/curative setups, both biologically and clinically relevant versions of ICD-based DC vaccines provided strong anti-HGG survival benefit. We found that the ability of DC vaccines to elicit HGG rejection was significantly blunted if cancer cell–associated reactive oxygen species and emanating danger signals were blocked either singly or concomitantly, showing hierarchical effect on immunogenicity, or if DCs, DC-associated MyD88 signal, or the adaptive immune system (especially CD8+ T cells) were depleted. In a curative setting, ICD-based DC vaccines synergized with standard-of-care chemotherapy (temozolomide) to increase survival of HGG-bearing mice by ~300%, resulting in ~50% long-term survivors. Additionally, DC vaccines also induced an immunostimulatory shift in the brain immune contexture from regulatory T cells to TH1/cytotoxic T lymphocyte/TH17 cells. Analysis of the The Cancer Genome Atlas glioblastoma cohort confirmed that increased intratumor prevalence of TH1/cytotoxic T lymphocyte/TH17 cells linked genetic signatures was associated with good patient prognosis. Therefore, pending final preclinical checks, ICD-based vaccines can be clinically translated for glioma treatment.


Anticancer immunotherapy has emerged as an important therapeutic paradigm (1). Of the immunotherapies currently available, remarkable efforts have been invested in the development of dendritic cell (DC)–based vaccines (2). DCs are the main sentinel antigen-presenting cells of the immune system (2). In a therapeutic sense, DC vaccines largely act by stimulating tumor-specific cytotoxic T lymphocytes (CTLs) that recognize and eliminate malignant cells (3). The promise of DC-based immunotherapy is backed by two decades of clinical studies (3) across various cancer types, but mostly in patients with melanoma, prostate cancer, high-grade glioma (HGG), and renal cell cancer (3). A recent meta-analysis showed that of these four, the highest positive objective responses (15.6% of patients) were observed from HGG patients (3, 4), suggesting that HGG might respond to highly efficacious immunotherapy (3, 4). Current multimodal treatments fail to improve HGG patient prognosis (5, 6). Standard-of-care anti-HGG therapies [like temozolomide (TMZ)] partly improve patients’ prognosis but eventually fail because of HGG’s invasion and recurrences (7). This alarmingly negative outlook mandates anti-HGG therapeutic interventions like “next-generation” DC-based immunotherapy (5).

Current anti-HGG DC immunotherapies mostly use specific antigen peptides/RNA for pulsing DCs (8, 9) or whole-glioma tumor cells (10, 11) killed via freeze/thawing (F/T)–based necrosis (5, 12). Although the former methodology might exhibit low efficacy due to the high antigenic heterogeneity of HGG, the latter procedure is associated with poor immunogenic potential (5). Thus, to create the next-generation anti-HGG vaccines, the immunogenicity of the dying/dead cancer cells used to pulse the DCs and the ability to induce a superior T helper 1 (TH1)–mediated immunity need to be improved (3). Hence, we envisaged that a next-generation DC vaccine could be produced by pulsing DCs with HGG cells undergoing immunogenic cell death (ICD) (13, 14).

Cancer cells undergoing ICD exhibit superior immunogenic potential due to exposure/release of damage-associated molecular patterns (DAMPs) acting as potent danger signals (15). Crucial DAMPs for ICD include surface-exposed calreticulin (CRT), surface-exposed heat shock protein (HSP) 70/90, secreted adenosine triphosphate (ATP), and passively released high-mobility group box 1 (HMGB1) protein (15). ICD promotes an efficacious anticancer vaccination effect that is largely biased toward TH1 immunity (13). Although several ICD inducers have been described recently (15), only a few have been thoroughly characterized (14). One such prototypical ICD inducer is hypericin-based photodynamic therapy (Hyp-PDT) (14, 16, 17). Hyp-PDT–treated cancer cells expose/release high amounts/diversity of DAMPs at a rapid pace (16, 18, 19) and elicit fully mature DCs that facilitate proliferation of interferon-γ (IFN-γ)–producing CD4+/CD8+ T cells (16, 18, 19). Moreover, cancer cells undergoing Hyp-PDT–induced ICD act as potent anticancer vaccines in various mice vaccination models (16, 19). Notably, Hyp-PDT has been successfully applied in the past for clinical treatment of patients with nonmelanoma skin cancer (20), cutaneous T cell lymphoma (21), and basal/squamous cell carcinoma (22).

Here, we performed a preclinical study using a single-agent ICD inducer–based DC vaccine in an orthotopic HGG mouse model and tested its efficacy as a next-generation anti-HGG immunotherapy (12, 23). We also systematically assessed the key molecular/signaling steps required to elicit immunogenicity. Finally, we accounted for various stiff clinical operational/regulatory requirements (24) to produce a “clinically relevant version” of our ICD-based DC vaccines.


Murine glioma cells exhibit ICD-associated molecular determinants after Hyp-PDT

We initially examined the ability of Hyp-PDT to elicit ICD-associated determinants in murine GL261 glioma cells. Hypericin displayed the typical (16, 18) endoplasmic reticulum (ER) localization (Fig. 1, A and B) and, upon light activation, induced oxidative stress (25) (Fig. 1C) along with ER stress signatures culminating into loss of viability and apoptosis (Fig. 1, D to H). Attenuating the induction of reactive oxygen species (ROS), ER stress or apoptosis (19) increased the survival of GL261 cells after Hyp-PDT (Fig. 1I), thereby substantiating the pro-death role of these processes (16, 18).

Fig. 1. Hyp-PDT induces ROS/ER stress–based apoptosis and major ICD-associated DAMPs in murine glioma cells.

(A and B) Localization of Hyp (red) versus ER Tracker Blue-White (blue) via fluorescence microscopy in GL261 cells (A), and the analysis of corresponding colocalization defining coefficients (n = 5; mean ± SEM) (B). (C to H) GL261 cells were treated with Hyp-PDT, followed by (C) estimation of carbonylated proteins [fold change relative to CNTR; n = 3 (two to three technical replicates per n), mean ± SD] at 1 hour after PDT, (D and E) immunoblotting analysis of whole-cell lysate at 1 hour after PDT, (F) percent cellular survival (n = 4; mean ± SD), (G) percent dead SyGr+ cells (n = 3; mean ± SD) at 24 hours after PDT, and (H) immunoblotting of whole-cell lysate at indicated time points. CI, cleaved. (I) Cells were preincubated for 1 hour with NAC (5 mM), cleavedTUDCA (500 μg/ml), or zVAD-fmk (25 μM) followed by percent cellular survival (n = 3; mean ± SEM) at 24 hours after PDT. (J to L) Analysis for surface calreticulin (J), surface HSP70 (K), and surface HSP90 (L) [data presented as mean fluorescence intensity (MFI), n = 3; mean ± SEM] at 1 hour after PDT. (M and N) Analysis for relative amounts of extracellular (M) and intracellular (N) ATP levels (n = 6; mean ± SEM). (O) Immunoblotting analysis of concentrated conditioned media at indicated time points after PDT. Student’s t test was used for statistics; N.S., not significant (see also fig. S13).

On the level of DAMPs, Hyp-PDT caused rapid surface exposure of CRT (Fig. 1J), HSP70 (Fig. 1K), and HSP90 (Fig. 1L) and secretion of ATP (Fig. 1M) in the absence of changes in intracellular ATP levels (Fig. 1N). All this culminated into an extracellular exodus of HSP90/HMGB1 (Fig. 1O). Moreover, murine DCs co-incubated with Hyp-PDT–treated GL261 cells exhibited significant phenotypic maturation (fig. S1).

At similar apoptosis-inducing doses (fig. S2A), the mitochondria-associated photosensitizer 5-aminolevulinic acid (5-ALA; Gliolan), which is frequently used for clinical fluorescence-guided resection of HGG (26), enabled surface HSP70 (fig. S2B) and secreted ATP (fig. S2C). Yet, only Hyp-PDT induced significant surface CRT (fig. S2D), surface HSP90 (fig. S2E), and passively released HMGB1 (fig. S2F). This reiterates (16, 19) that higher ER targeting translates into higher DAMP induction.

ICD-based DC vaccines provide significant protective immunity against HGG

We next tested whether Hyp-PDT–induced ICD-based DC vaccines can induce anti-glioma protective immunity (in prophylactic setup) (Fig. 2A) (27). Immunocompetent, syngeneic C57BL/6 mice were vaccinated, or not (CNTR), twice with either ICD-based DC vaccines or F/T necrosis–based DC vaccines (the latter being a control for non-ICD) intraperitoneally. Thereafter, immunized and nonimmunized mice were inoculated with live GL261 glioma cells intra-axially (that is, in the brain). After this, mice were monitored for HGG-induced neurological deficit symptoms (27). Remarkably, about 70% of mice vaccinated with ICD-based DC vaccines resisted orthotopic glioma challenge and showed significantly higher median survival (versus CNTR or F/T necrosis–based DC vaccinated mice) (Fig. 2B). Similarly, analysis of the HGG-induced neurological deficit scores/grades revealed a considerable delay in the onset of clinically relevant symptoms in mice treated with ICD-based DC vaccines (Fig. 2, C to E). Next, we substantiated these data with magnetic resonance imaging (MRI)–based noninvasive monitoring (Fig. 2, F, J, N, and R) followed by three-dimensional (3D) rendition of the brain (Fig. 2, G, K, O, and S), the gliomas (wherever visible) (Fig. 2, L and P), and the anterior/middle ventricles (Fig. 2, I, M, Q, and U). In comparison to naïve mice (Fig. 2, F to I), CNTR (Fig. 2, J to M) and F/T necrosis–based DC-vaccinated (Fig. 2, N to Q) mice exhibited palpable HGG masses that distorted the ventricles morphology. Instead, most mice vaccinated with ICD-based DC vaccines exhibited the absence of visible HGG masses and normal brain/ventricle morphology (Fig. 2, R to U). Moreover, as compared to CNTR (significantly) and F/T necrosis–based DC-vaccinated mice (to a variable/nonsignificant extent), mice vaccinated with ICD-based DC vaccines did not exhibit any strong increase in brain volume (fig. S3).

Fig. 2. ICD-based DC vaccines elicit efficient HGG-rejecting immunity in an orthotopic setup.

(A) GL261 cells were treated with Hyp-PDT and co-incubated with DCs to produce ICD-based DC vaccine. Alternatively, DC vaccines were produced with F/T-based necrosis-treated GL261 cells. Thereafter, prophylactic vaccination was carried out. (B) Kaplan-Meier curve depicts survival [n = number of mice as indicated; log-rank (Mantel-Cox) test; P value depicts comparison with CNTR; bar indicates comparison between F/T necrosis and Hyp-PDT–based DC vaccines]. (C to E) Graphs indicate progressive HGG-induced neurological deficit symptoms (grade 1 to 3) before mice were sacrificed (grade 4). (F to U) Representative MRI scans for naïve/normal mice (F) and day 29 after intra-axial GL261-inoculated mice (J, N, and R). 3D renditions of the whole brain (G, K, O, and S), visible HGG tumor mass wherever available (H, L, P, and T), and morphology of the anterior brain ventricles (I, M, Q, and U) are depicted.

HGGs exhibit high heterogeneity on multiple levels, including major histocompatibility complex (MHC)–based immunogenicity (28, 29). GL261 tumors are partially immunotherapy-susceptible because of GL261’s detectable MHC-based immunogenicity (28, 29). Thus, we tested the potential of ICD-based DC vaccine against CT2A, a well-known immunotherapy-evasive glioma model (29). Compared to GL261 cells, CT2A cells exhibited a different morphology [fig. S4A (29)] and expressed significantly less MHC-I levels (fig. S4B) and almost negligible MHC-II levels even after IFN-γ stimulation (fig. S4C). These observations substantiate that CT2A has low immunogenicity (29). Despite this disparity, Hyp-PDT–treated CT2A cells exposed/released all the main ICD-related DAMPs efficiently (figs. S4, D to H and S12). Also, mice vaccinated with ICD-based DC vaccines and orthotopically challenged with live CT2A cells showed significantly higher median survival as compared to CNTR mice (fig. S4I).

Antitumor immunity induced by ICD-based DC vaccines depends on ROS, danger signals, DCs, intact adaptive immune system, and CD8+ T cells

The immunogenic impact of major ICD-associated molecular/cellular determinants has never been comparatively tested in an orthotopic HGG model (14). So, we used relevant blockade strategies to assess the relevance of cancer cell–associated danger signals/DAMPs for the efficacy of DC immunotherapy (Fig. 3A). The presence of antioxidants, NAC/L-Hist (25), significantly attenuated the immunogenic potential of the ICD-based DC vaccine (Fig. 3, B and C). Similarly, blocking different DAMPs reduced the vaccine’s immunogenic potential in a hierarchical manner (extracellular HMGB1 > extracellular ATP > surface CRT) (Fig. 3, B and C). Extracellular HMGB1 ablation significantly reduced both percent median survival (Fig. 3B) and percent long-term survivors (Fig. 3C), whereas degradation of extracellular ATP only significantly ablated the latter (Fig. 3, B and C). Surface CRT blockade reduced both, albeit insignificantly (Fig. 3, B and C). Remarkably, concurrent ablation of all three DAMPs additively reduced the mice median survival (30 days), more than individual DAMP blockade, to nearly the level of CNTRs (26 days) (Fig. 3, B and C).

Fig. 3. Blockade of ROS, danger signals, DCs (or DC-associated MyD88), intact adaptive immune system, and CD8+ T cells abrogates the efficacy of ICD-based DC vaccines.

(A) Molecular (ROS/CRT/ATP/HMGB1) and immunological determinants (DCs/DC-based HMGB1-MyD88 axis/adaptive immunity especially CD8+ T cells) were targeted through chemical/antibody/genetic means. (B to G) Respective vaccines were injected twice in C57BL/6 mice. The respective vaccines were pretreated/neutralized by NAC/L-Hist, anti-CRT or anti-HMGB1 antibodies (Ab), apyrase, and a combination of the latter three (B and C), or the DCs were either absent or derived from Myd88−/− mice (D and E) or mice receiving the vaccine were Rag1−/− (F) or depleted for CD8+ cells (G). Thereafter, mice were intra-axially inoculated with live GL261 cells. Kaplan-Meier curves depict survival (B, D, F, and G); percent glioma-free long-term surviving mice are depicted in (C) and (E). Log-rank (Mantel-Cox) test (B, D, F, and G) or χ2 test (C and E) was used; in the absence of bars, P values depict comparison with CNTRs; n = number of mice as indicated.

Cellular adjuvants like DCs are often considered dispensable for ICD (13), because the cancer cells dying via ICD are considered to be sufficiently immunogenic (16). However, in our setup, vaccination with dead/dying cancer cells undergoing ICD alone, without DCs, drastically reduced the vaccine’s immunogenic potential (Fig. 3, D and E). Because extracellular HMGB1 emerged as a predominant DAMP, we ablated its Toll-like receptor (TLR) agonist function (30) conveyed through the TLR adaptor, myeloid differentiation primary response gene 88 (MyD88), for innate immune signaling (31). Notably, using Myd88−/− DCs for making ICD-based DC vaccines ablated both percent median survival and percent long-term survivors (Fig. 3, D and E), thus indicating the crucial adjuvant role of DCs in these settings.

The immunogenic potential of ICD/DC vaccines relies upon an intact adaptive immune system and especially on CD8+ T cell activity (1, 3). Remarkably, Rag1−/− mice lacking adaptive immune system completely failed to resist intra-brain glioma inoculation despite immunization with ICD-based DC vaccines (Fig. 3F). Even in CNTR settings, Rag1−/− mice (17 days) showed significantly reduced median survival as compared to Rag1+/+ mice (22 days) (Fig. 3F), an observation reminiscent of anti-glioma immunosurveillance (27). Similarly, CD8+ T cells’ depletion (confirmed in various immunological compartments; fig. S5) also significantly ablated the efficacy of ICD-based DC vaccines (Fig. 3G). Notably, vaccinated mice depleted of CD8+ T cells (median survival, 21 days) (Fig. 3G) survived only marginally better than similarly vaccinated Rag1−/− mice (median survival, 17 days) (Fig. 3F).

ICD-based DC vaccines can withstand the rigid clinically relevant operational parameters

Two clinically relevant operational parameters pose a significant challenge to the immunogenicity of anticancer vaccines—that is, avitalization (absolute absence of living cells) and availability of tumor tissue material required for vaccine preparation in frozen state (24). To enable future translation toward the clinic, we generated a clinically relevant version of Hyp-PDT–induced ICD-based DC vaccine accounting for these operational parameters. We used, in combination with Hyp-PDT, several avitalization strategies often applied in clinical protocols, that is, high-dose radiotherapy (RT), mechanical necrosis [achieved here via GentleMACS (GM) system, that is, GM necrosis], and F/T necrosis. Hyp-PDT + F/T necrosis and Hyp-PDT + RT + F/T necrosis, but not other combinations, achieved complete avitalization (Fig. 4, A and B). Moreover, only DC vaccines based on Hyp-PDT + RT, Hyp-PDT + F/T necrosis, and Hyp-PDT + RT + F/T necrosis induced survivals that were not significantly different than mice immunized with Hyp-PDT–induced ICD-based DC vaccines (Fig. 4C). Notably, these avitalization strategies did reduce the percent long-term surviving mice (Fig. 4A). The minimal combination achieving both avitalization (Fig. 4, A and B) and significant anti-HGG immunity (Fig. 4C) was Hyp-PDT + F/T necrosis–based DC vaccines. These were used subsequently as a clinically relevant version along with the original biologically relevant vaccine.

Fig. 4. ICD-based DC vaccine + F/T necrosis achieves absolute avitality coupled with efficient HGG-rejecting immunity.

(A and B) GL261 cells were left untreated or treated with Hyp-PDT and then exposed to avitalization methodologies like F/T necrosis, GM necrosis, RT (60 Gy), RT + F/T necrosis, and RT + GM necrosis. This was followed by (A) determination of percent cellular viability (n = 4; mean ± SEM) or (B) clonogenic assay (n = 3; mean ± SD) 10 days after treatment. (C) Respective vaccines were injected twice in C57BL/6 mice followed by intra-axial GL261 inoculation and depiction of Kaplan-Meier survival curve. (D) Frozen and then thawed (frozen cells) or live GL261 (live cells) were treated with Hyp-PDT or Hyp-PDT + F/T necrosis followed by depiction of Kaplan-Meier survival curve. In (A) and (B), Student’s t test was used, whereas log-rank (Mantel-Cox) test was used in (C) and (D). In the absence of bars, P value depicts comparison with CNTRs/medium-only/background; n = number of mice as indicated.

Next, we tested the effect of the start-up cellular material having undergone a cycle of snap freezing (overnight) and thawing (that is, 1 × F/T cycle, so called “frozen” cells) on vaccine immunogenicity. This F/T step drastically reduced the survival of the residual cancer cells (fig. S6). Hyp-PDT and Hyp-PDT + F/T necrosis treatment on these frozen cells further reduced the cellular survival gradually toward avitalization (fig. S6). For Hyp-PDT + F/T necrosis–based DC vaccine, state of the start-up cellular material did not strongly matter, because mice immunized with respective vaccines exhibited similar efficacy (median survival, live: 48.5 days versus frozen: 60 days; as compared to CNTR: 26 days) (Fig. 4D). On the other hand, for Hyp-PDT alone, the state of the start-up cellular material made a significant difference (Fig. 4D).

ICD-based DC vaccines induce a prophylactic shift in the brain immune contexture from regulatory T cells (Tregs) to TH1/CTL/TH17 cells

Next, we tested whether the ICD-based DC vaccine’s immunogenicity was associated with an immunostimulatory shift in the brain immune contexture. We analyzed the brain-infiltrating immune cells in mice that were prophylactically vaccinated (or not) and inoculated intra-axially with live GL261 glioma cells. We found that glioma-inoculated mice prophylactically vaccinated with Hyp-PDT or Hyp-PDT + F/T necrosis–based DC vaccines exhibited a significant increase (as compared to CNTR mice) in brain infiltration of CD3+ T lymphocytes (Fig. 5A), CD4+ T lymphocytes (Fig. 5B), CD8+ T lymphocytes (Fig. 5C), TH1 cells (Fig. 5D), CTLs (Fig. 5E), and TH17 cells (Fig. 5F). Whereas CNTR mice showed higher intra-brain infiltration of Tregs, mice vaccinated with respective ICD-based DC vaccines showed a significant reduction in Tregs (Fig. 5G). However, mice vaccinated with Hyp-PDT + F/T necrosis–based DC vaccines showed a trend toward slightly higher Tregs (Fig. 5G) and reduced TH17 levels compared to Hyp-PDT–based DC vaccines (Fig. 5F).

Fig. 5. ICD-based DC vaccines cause increased brain infiltration of TH1/TH17/CTLs.

Respective ICD-based DC vaccines were injected in a prophylactic manner. (A to H) At day 16 after GL261 inoculation, the brains were isolated and processed for fluorescence-activated cell sorting (FACS)–based immunophenotyping for (A) CD3+ T cells, (B) CD4+CD3+ T cells, (C) CD3+CD8+ T cells, (D) TH1 cells, (E) cytotoxic T cells/CTLs, (F) TH17 cells, and (G) Tregs. In (A) to (G), n = 3 to 6 mice; mean ± SEM. At the same time, T cell–enriched splenocytes (H) were recovered and co-incubated with DCs pulsed with “naïve” GL261 lysates for 5 days followed by estimation of IFN-γ production in respective supernatants (n = 4 to 5 mice; mean ± SEM). Mann-Whitney statistical test was used; in the absence of bars, P value depicts comparison with CNTRs.

Next, we analyzed whether the immunostimulation observed above was only brain-localized or also visible on a systemic level (crucial for long-term immunity) by investigating the ability of mice splenocytic T cells to exhibit “immune memory” responses upon antigen reexposure. Splenocytic T cells were isolated from glioma-inoculated CNTR mice or mice vaccinated with respective ICD-based DC vaccines (fig. S7) and restimulated with naïve (that is, untreated) GL261 lysates. We observed that splenocytic T cells isolated from respective vaccinated mice showed significantly higher IFN-γ production upon restimulation as compared to CNTR mice (Fig. 5H).

ICD-based DC vaccines in combination with chemotherapy exhibit the ability to cure HGG-bearing mice

Prophylactic vaccination, although critical for mechanistic analysis, does not represent the clinical status quo where vaccination is given in a curative setup (often in combination with standard-of-care chemotherapeutics like TMZ) (32). Thus, we tested the immunogenicity of ICD-based DC vaccines in a curative setup, either alone or in sequential combination with previous TMZ-based chemotherapy in HGG-bearing mice (Fig. 6A). TMZ did not induce any overt signs of toxicity as evident by maintenance of a healthy weight by the treated mice (fig. S8). We found that both versions of the ICD-based DC vaccines (that is, Hyp-PDT and Hyp-PDT + F/T) alone increased the median survival of glioma-inoculated mice by about 34 to 41% (albeit statistically insignificantly) (Fig. 6, B and C), but no long-term “cured” survivors were observed (Fig. 6C). TMZ-based chemotherapy alone significantly increased the median survival of glioma-inoculated mice by about 95% (Fig. 6, B and C) but failed to produce long-term cured survivors (Fig. 6C). Remarkably, combination of TMZ treatment followed by immunization with the two respective ICD-based DC vaccines not only strongly increased the median survival of glioma-inoculated mice by about 302 to 306% (Fig. 6, B and C) but also produced about 50% of long-term cured survivors (Fig. 6C). These remarkable results were also confirmed by an MRI-based noninvasive analysis of the brain (Fig. 6, D to K). At day 23, CNTR mice showed large glioma masses (Fig. 6D), yet mice that received vaccination only (Fig. 6, E and F) or TMZ only (Fig. 6G) exhibited relatively smaller but palpable tumors. Conversely, long-term cured survivors belonging to the “TMZ + vaccination” cohorts (Fig. 6, H to K) showed the absence of palpable glioma masses on days 55 (Fig. 6, H and J) and 90 (Fig. 6, I and K) after glioma inoculation. Further analysis (33) revealed that TMZ and the respective ICD-based DC vaccines synergized with each other to achieve this anti-glioma efficacy (fig. S9). Notably, long-term cured survivors from the respective TMZ + vaccination cohorts survived significantly better (than age-matched CNTR mice) after further rechallenge with glioma/GL261 cells (fig. S10), thus proving the induction of long-term anti-HGG immune memory by ICD-based DC vaccines.

Fig. 6. ICD-based DC vaccines, in combination with TMZ-based chemotherapy, provide strong survival benefit in therapeutic HGG setup.

(A) C57BL/6 mice were inoculated with live GL261 cells (day 0), intra-axially and randomly divided into six “cohorts.” Respective ICD-based DC vaccines and TMZ were administered as depicted. (B and C) Median survival data and percentage change (%Δ) in median survival (compared to CNTR) (B) and Kaplan-Meier curve (C) are depicted [log-rank (Mantel-Cox) test; in the absence of bars, P values depict comparison with CNTR]. (D to K) Representative MRI scans for day 23 after intra-axial GL261-inoculated mice of theindicated cohorts. Scans are also shown for day 55 (H and J) and day 90 (I and K) after intra-axial GL261-inoculated mice from TMZ + Hyp-PDT–based DC vaccine (H and I) and TMZ + Hyp-PDT + F/T necrosis–based DC vaccine (J and K) cohorts.

ICD-based DC vaccines partially overcome the immune-ablating effects of chemotherapy and favor a shift in the brain immune contexture from Tregs to TH1/CTL/TH17 cells in curative settings

TMZ has been traditionally found to exert immune-ablative effects, leading to myelosuppression and/or lymphopenia (34). In line with this, TMZ treatment severely reduced the absolute levels of intra-brain mononuclear immune cells (Fig. 7A). However, combined treatment with respective ICD-based DC vaccines rescued the TMZ immune-ablating effect (Fig. 7A). TMZ treatment reduced the overall levels of intra-brain T cells (Fig. 7B); however, these lympho-ablative effects were not targeted toward CD4+ T cells (Fig. 7C) but rather toward CD8+ T cells (Fig. 7D). TMZ-mediated intra-brain ablation of CD8+ T cells was so severe that the vaccines could not significantly rescue it (Fig. 7D). But combination of TMZ with ICD-based DC vaccines still reduced intra-brain levels of Tregs (Fig. 7E), whereas TMZ alone failed to do so (Fig. 7E). We further analyzed the shift in the brain immune contexture by estimating immune cell ratios (35) of immunostimulatory T cells to Tregs (that is, TH1/Tregs, CTLs/Tregs, and TH17/Tregs) (36). TMZ alone failed to increase the levels of TH1 (Fig. 7F), CTL (Fig. 7G), or TH17 (Fig. 7H) cells relative to Tregs in the brain. However, combining the respective vaccines with TMZ caused a significant increase in the TH1/Treg (Fig. 7F), CTL/Treg (Fig. 7G), and TH17/Treg (Fig. 7H) ratios.

Fig. 7. ICD-based DC vaccines cause a Treg-to-TH1/TH17/CTL shift in brain immune infiltrates after TMZ-based chemotherapy.

C57BL/6 mice were inoculated with live GL261 cells (day 0), intra-axially and randomly divided into four cohorts, that is, untreated (CNTR), TMZ-treated (40 mg/kg), and two that received TMZ in sequential combination with respective ICD-based DC vaccines (Fig. 6A). Thereafter, mice were sacrificed within 10 to 20 days after the end of respective treatments/challenge, and the brains were isolated. (A to H) Initially (A), total mononuclear immune cells were counted. Thereafter, these were processed for FACS-based immunophenotyping for (B) T cells, (C) CD4+ T cells, (D) CD8+ T cells, (E) Tregs, (F) TH1-to-Treg ratio, (G) cytotoxic T cell/CTL-to-Treg ratio, and (H) TH17-to-Treg ratio. For all data, n = 5 to 6 mice, mean ± SEM, and Mann-Whitney statistical test were used.

Increased prevalence of TH1/CTL/TH17 cell–associated genetic signatures, but not Tregs, correlates with good prognosis in glioblastoma patients

An immunostimulatory tumoral immune contexture is associated with improved prognosis in cancer patients (35). Although clinical translation of our vaccines is beyond the scope of this study, we were interested to ascertain the prognostic impact of TH1/CTL/TH17-based immune contexture in HGG/glioblastoma (GBM) patients. Two technologies can be used to analyze this, that is, the “classical” immunophenotyping of tumor-infiltrating lymphocytes or (preestablished) lymphocyte subpopulation-specific mRNA signatures within the tumor (37, 38). Here, we used the latter, because this allowed us to analyze a very large, standardized, and publicly available cohort of >500 GBM patients [that is, The Cancer Genome Atlas (TCGA) cohort of 541 newly diagnosed GBM patients] (39, 40).

Broad lymphocyte subtype-specific mRNA signatures available from previous studies (37, 38) were tailored for GBM by delineating groups of genes within these signatures, showing strong coexpression [that is, “metagenes” (41)] centered on standard/specific T cell markers (37). Through this methodology (37), we produced gene coexpression profiles of different T cell subpopulation–associated mRNA signatures (fig. S11, A and B, and Fig. 8, A to D). We then delineated a collection of GBM-tailored, T cell subtype–associated metagenes (centered on specific markers in case of high coexpressional heterogeneity), for example, CD4+ T cells (uncentered; fig. S11A), CD8+ T cells (CD8A/CD8B-centered; fig. S11B), Tregs (CD3D/CD3G/CD2-centered; Fig. 8A), TH1 cells (IFNG-centered; Fig. 8B), CTLs (CD8B-centered; Fig. 8C), and TH17 cells (IL17A/RORC-centered; Fig. 8D).

Fig. 8. Increased tumoral expression of TH1/CTL/TH17-associated metagenes is associated with prolonged OS in GBM patients.

(A to D) T cell type–specific metagenes were established through gene coexpression profiles (coexpression scored by Pearson’s correlation coefficient). Metagene profiles are delineated by yellow lines. (E to I) TCGA GBM patient cohort (n = 541) was stratified into “high-expression” (red; n = 136) or “low-expression” (black; n = 405) groups, followed by Kaplan-Meier plotting of patient’s OS versus follow-up duration in days for metagenes associated with Treg (E), TH1 (F), CTLs (G), TH17 (H), and combination of TH1/CTLs/TH17 (I). In (E) to (I) graphs, log-rank (Mantel-Cox) test Pvalues and hazard ratios (HR) are displayed. Two dotted lines represent approximate point of 2- or 5-year survival (y.s.). (J) Analysis of 20 articles studying the prognostic impact of T cell immune infiltrates in glioma/GBM patients (table S1) is represented. “Good” = increased infiltrates associated with favorable prognosis, “Poor” = increased infiltrates associated with poor prognosis, and “None” = lack of conclusive correlation.

Next, we estimated the prognostic impact of these metagenes’ differential expression on patient overall survival (OS) in the TCGA GBM cohort. This was further complemented with calculation of percent change in median survival (%ΔMS) between the high- and low-expression groups, that is, [(MShigh − MSlow)/MShigh] × 100, as described previously (3). High expression of CD4+ T cell metagene was associated with slightly reduced OS (fig. S11C), as evident from negative %ΔMS of −8.37% and a hazard ratio of more than 1. Further analysis of CD4+ T cell subtypes showed that whereas high expression of Treg metagene was associated with reduced OS (Fig. 8E) (%ΔMS of −7.36%), high expression of TH1-associated metagene was correlated with prolonged OS (Fig. 8F) (%ΔMS of +8.73%). Conversely, high expression of CD8+ T cell metagene was associated with slightly prolonged OS (fig. S11D) (%ΔMS of +4.47%), which further improved on the CTL level (Fig. 8G) (%ΔMS of +10.94%). Although the above survival trends were not significant, high expression of TH17 metagene alone (Fig. 8I) (%ΔMS of +18.76%) or combined high expression of TH1/CTL/TH17 metagenes (Fig. 8H) (%ΔMS of +17.56%) was associated with significant prolongation of OS.

Next, we substantiated the above results with a literature meta-analysis of papers using direct immunophenotyping of HGG/GBM-infiltrating lymphocytes (table S1) as described previously (35). This meta-analysis revealed that high tumor infiltration of CD4+ T cells/CD8+ T cells was associated with good prognosis in HGG/GBM patients (Fig. 8J). A similar trend was implied for GBM/HGG infiltration of TH1, CTL, and TH17 cells, albeit covered by fewer studies (Fig. 8J). On the other hand, high tumor infiltration of Tregs was associated with poor prognosis (Fig. 8J).


ICD-based vaccines have only been applied either in heterotopic subcutaneous cancer models (14, 42) or in an orthotopic model of non–vital organ–associated cancer (43). This study combines a single-agent ICD inducer with DC immunotherapy against a highly immunosuppressive and vital organ–associated cancer (HGG) in a preclinical, orthotopic model and directly compares the immunogenic impact of four different ICD-essential molecular determinants (ROS/CRT/ATP/HMGB1) (27, 44). Considering the previous application of Hyp-PDT for clinical cancer treatment and of hypericin for clinical fluorescence-guided HGG resection (45), we believe that Hyp-PDT–based anticancer vaccines can be translated toward the clinic in the near future. We show that Hyp-PDT–induced ICD-based DC vaccines have a high potency in both biologically and clinically relevant formats.

Remarkably, in prophylactic/curative setups, both biologically and clinically relevant versions of the ICD-based DC vaccines provided considerable survival benefit against HGG. Especially in the latter case, in combination with the standard-of-care chemotherapeutic TMZ, both ICD-based DC vaccines synergized to increase the median survival by more than 300%. The combination with TMZ-based chemotherapy was essential to provide strong survival benefit, because the vaccines alone only increased median survival by 34 to 40%. This is a known phenomenon, that is, DC immunotherapies tend to function suboptimally in the presence of preexisting tumors, owing to the high tumor–induced immunosuppression (3, 46). Also, clinically, an inverse association has been observed between tumor burden and immunotherapy effectiveness (46). These observations have paved the way for combining DC immunotherapy with cyto/tumor-reductive anticancer chemotherapies like TMZ (3)—a rationale that we successfully applied here. It has also been postulated that chemotherapy preceding DC immunotherapy can make tumor cells more susceptible to immunotherapy (5, 47, 48), thereby providing synergistic survival benefit (49).

Although, in the curative setup, there was no significant difference between the efficacies of Hyp-PDT–based and Hyp-PDT + F/T necrosis–based DC vaccines, in prophylactic setup, there was a distinct difference. Such differences could originate from the nature of prophylactic vaccination. Although this approach is adequate for immunogenicity and mechanistic testing (13, 19, 50), it does not take into account the preestablished tumor burden–based immunosuppression (32) and systemic tolerogenicity toward overexpressed cancer antigens (32, 50), factors that can change immunological responses induced by DC immunotherapy (3, 5, 47). Although our DC immunotherapies show potency in both prophylactic and curative setup, the latter is much more relevant.

Our study shows that ROS and danger signals (extracellular HMGB1 > extracellular ATP > surface CRT) are major cancer cell–autonomous determinants of the vaccine’s immunogenicity. Clearly, in the case of Hyp-PDT or PDT in general, ROS generation is the most apical signaling event triggering cell death–associated signals crucial for ICD, including ER stress and exposure/release of some key DAMPs (16, 18). Beyond PDT, ROS might have very broad implications for all “oxidative” anticancer therapies with immunogenic potential (25). Our studies further show that concurrent blockade of all three ICD-relevant danger signals almost completely ablates the in vivo immunogenic potential of the vaccine, and outline the predominant immunogenic role of HMGB1-MyD88 axis in this setup. It is possible that, here, HMGB1 has a predominant role, because the vaccines are largely composed of post-apoptotic cells (16, 17, 51). Thus, depending on the vaccination or therapeutic context, the immunogenic potential of one, or a few, DAMP may prevail on the others (14). Irrespective of this, our study supports the decisive role of ICD-associated DAMPs in generating powerful DC-based vaccines.

The efficacy of ICD-based DC vaccines was highly dependent on an intact adaptive immune system and CD8+ T cells as the predominant adaptive immune mediators, although this does not rule out the role of other immune cells. ICD-based DC vaccines also induced an immunostimulatory shift in the brain immune contexture, that is, increased intra-brain infiltration of TH1/CTL/TH17 cells relative to reduction in Tregs in both prophylactic and curative setups. TMZ exerted specific (intra-brain) immune-ablative effects that were corrected to a certain extent by the vaccines to sustain a Treg-to-TH1/CTL/TH17 shift. Hence, whereas on the level of OS, TMZ- and ICD-based DC vaccines might exhibit synergism, on the level of brain immune contexture, these two strategies engage in partially antagonistic interactions. This further substantiates the largely cytotoxic or tumor-reductive role of TMZ in this setup. Thus, as expected from a next-generation DC immunotherapy (3), Hyp-PDT–induced ICD-based DC vaccines produced a TH1-biased immunity. Our TCGA GBM patient analysis and literature meta-analysis showed that increased incidence of molecular signatures of TH1/CTL/TH17 relative to Tregs can have a significant positive prognostic impact. Remarkably, across preclinical and clinical data, brain-infiltrating TH17 cells (or corresponding molecular signatures) were associated well with good prognosis. Because the role of TH17 in GBM/HGG remains largely enigmatic, in the future, it would be necessary to carry out detailed studies on the anti-HGG immunological effects of TH17 cells. The analysis of the prognostic impact of TH1/CTL/TH17 cell–associated metagene in GBM/HGG patients is of great significance, because it now joins few other such analyses accomplished in other cancer types (melanoma, colorectal, breast, and lung cancer) (35). Such analysis may also be crucial in the future for patient stratification exercises striving to delineate GBM/HGG patients most likely to respond to ICD-based DC vaccines (14).

The current study has some limitations that need urgent attention in the near future. For instance, HGG-tumor microenvironment/stroma–associated factors that facilitate immunosuppression or cause T cell exclusion/anergy, including hypoxia/acidosis, immune checkpoints’ status, effects of prostaglandin E2/transforming growth factor–β (TGF-β), and macrophages/myeloid-derived suppressor cells, were not studied (1, 5, 12). On the level of DC vaccine, further investigations on the impact of the route of injection (5) and/or enrichment of particular DC subsets [for example, CD8α+ DCs (14)] over others on the vaccine’s efficacy are required. Furthermore, we have not addressed here interpatient heterogeneity that can affect the translation of ICD-based DC vaccines, for example, inactivating polymorphisms/mutations in certain immune receptors (13, 14), cancer cell–autonomous mechanisms disrupting danger signaling or immunogenic phagocytosis (17, 41), and HGG stem cell–associated immunosuppression (29). Such clinically relevant barriers need to be identified to allow better patient stratification and to design “smart” combinatorial therapies. Our results also show that responsiveness to ICD-based DC vaccines could also be affected by MHC-level heterogeneity, thereby urging studies addressing strategies to overcome this resistance. Nevertheless, the current study using the HGG model consolidates the strength of Hyp-PDT–based anticancer vaccines observed previously using other cancer models (16, 19). Whether (or not) Hyp-PDT–induced ICD-based DC vaccines have similar effectiveness in other cancer types needs to be tested.


Detailed materials and methods are available in the Supplementary Materials.


Materials and Methods

Fig. S1. DCs co-incubated with Hyp-PDT–treated GL261 cells exhibit increased phenotypic maturation.

Fig. S2. Hyp-PDT induces superior enrichment of DAMP exposure/release than 5-ALA–PDT.

Fig. S3. Mice treated with ICD-based DC vaccines maintain normal brain volume despite HGG challenge.

Fig. S4. The low immunogenic, immunotherapy-resistant, CT2A glioma can be significantly rejected by Hyp-PDT–induced ICD-based DC vaccine.

Fig. S5. Anti-CD8 antibody depletes CD8+ T cells (but not CD4+ T cells) in various immune compartments.

Fig. S6. Single freezing step does not completely abrogate the survival or clonogenic potential of murine glioma cells.

Fig. S7. Splenocytes derived from mice are functionally competent.

Fig. S8. Treatment of mice with the chemotherapeutic drug TMZ does not lead to general toxicity.

Fig. S9. ICD-based DC vaccines synergize with the chemotherapeutic drug TMZ in providing survival benefit in a therapeutic HGG setup.

Fig. S10. Long-term survivors immunized previously by ICD-based DC vaccines tend to significantly reject rechallenge with orthotopic HGG.

Fig. S11. Increased tumoral expression of CD8+ T cell–associated metagenes, but not CD4+ T cell–associated metagene, is associated with prolonged OS in GBM patients.

Fig. S12. Source data for figs. S2F and S4H.

Fig. S13. Source data for Fig. 1, D, E, H, and O.

Table S1. Literature meta-analysis of the prognostic impact of intratumoral T cell infiltration in HGG or GBM patients.

References (5282)


  1. Acknowledgments: We thank S. Seys for help with CBA (cytometric bead array) assay and J. Belmans for his general help. GL261 cell and Rag1−/− mice were received from I. Y. Eyupoglu (University of Erlangen) and G. Boeckxstaens (KU Leuven), respectively. Funding: A.D.G. and T.V. are postdoctoral fellows, whereas S.W.V.G. is a senior clinical investigator of the FWO (Fonds Wetenschappelijk Onderzoek), Vlaanderen. L.V. is the recipient of a Strategic Basic Research grant from the IWT (Innovatie door Wetenschap en Technologie), Vlaanderen. This work is supported by FWO (G0584.12N, K202313N, and GA01111N) and KU Leuven (C16/15/073) grants to P.A. and by Olivia Hendrickx Research Fund, the James E. Kearney Memorial Fund, the Herman Memorial Research Fund, the Belgian Brain Tumor Support, and individual donors to S.W.V.G. This paper represents research results of the IAP 7/32 funded by the Interuniversity Attraction Poles Programme initiated by the Belgian State. Author contributions: A.D.G. and L.V. jointly carried out most of the experiments/experimental design/data analysis and statistical analysis and did the final data representations. L.V. standardized various methods/protocols. A.D.G. did the TCGA patient analysis/literature survey and wrote the manuscript/made the figures. T.V. helped with various experiments/experimental design. C.K. helped with various experiments/MRI scanning. L.B. provided the purified CD8+ T cell–depleting antibodies. S.W.V.G. and P.A. jointly supervised the study and participated in the experimental design/manuscript writing/critical reading/revision. Competing interests: The authors declare that they have no competing interests.
View Abstract

Stay Connected to Science Translational Medicine

Navigate This Article