Gliomas, especially GBM, have a unique TME compared to tumors at other sites. The central nervous system (CNS) used to be regarded as an immunologically privileged site. One of the reasons for this understanding is that the lymphatic drainage system of the brain has not been discovered for a long time. Another reason is the existence of the blood-brain barrier (BBB) (51). Recently, the discovery of the glial-lymphatic pathway has proposed a mechanism for connecting the parenchyma and interstitium with the cerebrospinal fluid space (52); meanwhile, the discovery of functional lymphatic vessels in the meninges confirms the existence of a direct drainage pathway for cerebrospinal fluid that contains solutes and immune cells from the brain to the cervical lymph nodes (53, 54). The brain is protected from pathogenic microorganisms by the BBB, consisting of pericytes, astrocyte processes, vascular endothelial cells, and extracellular matrix. Meanwhile, it makes it harder for drugs and peripheral immune cells to enter the CNS, facilitating tumor invasion and growth (55). However, it has been shown in recent studies that T cells can enter the brain and provide immune surveillance (56, 57), challenging the notion that the BBB is sealed to immune cell entry. Simultaneously, damaging the BBB by GBM itself can also limit the BBB’s ability to function (58). In summary, during inflammation, specific antigens are recognized by microglia, then microglia present them to activated lymphoid cells through the glial-lymphoid pathway, and subsequently more immune cells penetrate the BBB, leading to a more intense inflammatory response and following immune responses. Thus, CNS immunity is not so much “privileged” as it is “unique”.

However, compared to other tumor types, CNS tumors have lower levels of tumor-infiltrating lymphocytes (TILs) and other types of immune effector cells (59). This “cold tumor” phenotype has been related to poor response to immunostimulatory therapies such as ICIs (17, 60). Even when inducing T cells to respond to CNS cancer, the number of antigen-specific TILs remains relatively low, and the TILs present often exhibit a depleted phenotype (18, 61). Upon inflammatory stimulation, brain stromal cells produce high levels of classical immunosuppressive cytokines such as TGFβ. These cytokines neutralize inflammatory factors to maintain homeostasis (62). Glioma cells produce high levels of indoleamine-2,3-dioxygenase (IDO), which, besides promoting Treg accumulation, inhibits T-cell activity by depleting microenvironmental tryptophan (63). Microglia and tumor-infiltrating myeloid cells reduce the arginine level in the tissues by producing high levels of arginase, which further suppresses the proliferation and functions of T cells (64). Additionally, the compromised BBB suppresses glioma patients’ adaptive immune response by upregulating programmed death ligand 1/2 (PD-L1/2) expression to prevent effector T-cell from activation (65).

Normally, peripheral circulating DCs reach the vascular-rich compartment through the central lymphatic vessels and are virtually absent in the brain parenchyma (66). However, a recent study found that CD141⁺ cDC1 can infiltrate the region of GBM and present antigens to T cells in deep cervical lymph nodes (dcLNs) (67). Nonetheless, extracranial antigen presentation failed to facilitate tumor eradication in the absence of immunotherapy in a melanoma brain metastasis model (68). This indicates that the presentation of antigens in the periphery is probably not sufficient to induce immunity against brain tumors.

A major barrier to the application of DCs for the treatment of GBM is that DCs must have the capacity to induce anti-tumor immune responses under immunosuppressive conditions. The mechanism of immunosuppression in GBM involves both the glioma cells themselves and the cells in the TME ( Figure 2 ).

A large majority of DC vaccines in clinical trials are based on MoDCs. In particular, DCV trials in GBM now all use MoDCs. The common method is to collect autologous monocytes from patients, induce them to differentiate into immature DCs in vitro, expose them to TAAs after induction of maturation, and then transfuse them into the same patient. In trials that used mDCs, DC maturation was mostly induced by GM-CSF combined with IL-4, PGE2, TNF-α, or IFN-γ (95–99). There are couples of trials that induce DC maturation by using IL-6, IL-1β, TNF-α, or PGE2 without GM-CSF (100–102). Although cDCs may be superior to MoDCs in their ability to stimulate T cells (7, 103), there are currently no established protocols for isolating or differentiating these cells in vitro. Nonetheless, in some diseases, most notably melanoma, the use of cDCs and pDCs as DC vaccines has shown some encouraging early results that may be extended to GBM research in the future (104–107).

In addition, since iDCs are less capable of stimulating T cells than mDCs and may even induce tolerance, mDCs are used in most DCV trials. However, there are trials using iDCs that have reported clinical benefits (108, 109).

By priming CD8+ T cells against TAA, DCs are an important part in antitumor immunity. Thus, the efficacy of DCV is related to the existence of TAAs, also known as neoantigens, in individual tumors. The overall mutational burden of GBM is very low, but patients who relapse after TMZ chemotherapy have an increased mutational burden (110).

Previous trials using DCV used tumor lysates, tumor cell apoptotic bodies, irradiated tumor cells, DC-tumor cell fusion, and tumor cell surface eluted peptides as whole tumor cell TAAs. Whole tumor cell-derived TAAs contain numerous TAAs, assuring the diversity of antigens and reducing the risk of TAA-loss variants escaping (111). However, due to the immunosuppressive factors produced by glioma cells, whole tumor cell-derived TAAs may inhibit the DC differentiation and maturation or alter the function of generated DCs (112). Furthermore, whole tumor cell-derived TAA vaccines produced using current methods are poorly immunogenic and difficult to induce potent and durable T cell responses (113).

Some DCV studies use molecularly defined TAAs, including specific peptides, proteins, and DCs transfected with TAA-coding mRNA. The source of molecularly defined TAAs is more standardized and reproducible, making it easier to monitor target-specific responses. In addition, they can be personalized for different individuals (114). However, compared to TAAs derived from whole tumor cells, molecularly defined TAAs lack diversity. Therefore, to reduce the risk of escape of TAA-loss mutants, several molecularly defined TAAs should be used.

To induce a T cell response in a healthy subject, the minimum DC dose is 2×10^⁶ DC/vaccine (115), while no study to date has achieved dose-limiting toxicity. While several clinical studies aiming at determining the best dose of DCV therapy have been conducted previously, and some of them have been completed (e. g. NCT00612001, NCT01171469, NCT00068510, NCT00107185), the relationship between clinical outcomes and DC dose, and the dose-response relationship of the optimal dose have remained inconclusive. Studies have shown that patients receiving lower doses of DC have longer survival (116); while some studies suggest that improving the efficiency of DC migrating to lymph nodes may increase patient survival (117). This may be because the DCV used in these studies was handled differently as well as the status of DC, making it difficult to compare to derive the optimal dose of DC. In the existing clinical trials, almost all patients received multiple vaccinations, mainly using the prime-boost method (18). Several studies have reported a trend toward improved survival in booster recipients (102).

Different routes of injection of DCV result in different distributions of DC in vivo ( 118, 119). Currently, the routes of administration used in clinical trials of DCV include intravenous injection, subcutaneous injection, and nodal injection. Subcutaneous injection is by far the most common route of administration, with up to 4% of DCV reaching the draining lymph nodes. Irrespective of the routes of administration, high numbers of DCs remained at the injection site, lost viability, and were eliminated by infiltrating CD163+ macrophages within 48 hours (120). The intranodal injection may allow more DCs to migrate to the T-cell region, but whether it is more effective in inducing antigen-specific immune responses remains to be determined (120).

Most patients underwent cytoreductive surgery before DCV, while some patients underwent biopsy alone or without surgery. The extent of surgical resection is positively associated with survival (121), while minimal residual disease status also favors DCV therapy (121, 122), which may be related to a reduction in local immunosuppression (123). Yet, other studies have shown that the extent of resection is not related to survival in DCV treatment (124). Therefore, in addition to the absolute volume of the residual tumor, other factors such as the composition of the residual tumor may influence the effect of DCV.

DCV treatment is often combined with radiotherapy or chemotherapy, or both. Tumor cell death after chemoradiotherapy releases tumor antigens, then the brain endothelium presents MHC class I antigens to circulating CD8⁺ T cells, which can enhance the tumor-specific effector CTL homing to brain tumors (125). The most widely used chemotherapeutic agent combined with DCV is TMZ, which has been used in all current DCV-controlled trials. TMZ can improve immunoreactivity by reducing Tregs and interfering with their recruitment to tumors (126). Although TMZ often induces lymphopenia, the lymphocyte zone restored after chemotherapy can still induce an antitumor response (117, 127). The specific efficacy is related to the dose of TMZ: for example, lower doses of TMZ help deplete Tregs, whereas myelosuppressive doses enhance the response to peptide vaccines (128). However, there is also evidence that CD8⁺ T cells expanded by DCV previously may be depleted by TMZ (100, 129). Moreover, only in the absence of TMZ was DCV able to generate IFN-γ-producing effector memory T cells, which was positively related to survival (130). Thus, the effect of TMZ on DCV efficacy remains inconclusive.

By far, no serious vaccine-related adverse events have been observed, except for a few studies that reported severe adverse events (grade ≥3) according to the National Cancer Institute Common Toxicity Criteria (NCI CTC). Some of these adverse events were severe peritumoral edema leading to other neurological symptoms (96, 131); some were allergies following co-injection of DCV and GM-CSF (132).

Commonly observed adverse reactions attributed to DCV are generally mild (≤ grade 2), including induration, pain, pruritus, and erythema in injection sites, as well as meningeal irritation, lymph node swelling, flu-like symptoms, edema, etc (95, 97, 98, 101, 102, 116, 133–143). These symptoms may be caused by disease progression or other concomitant therapies as well. All in all, DCV therapy was well tolerated as a therapeutic method (for a brief introduction of DCV clinical trials registered in clinical trials.gov, see Table 2 and Table 3 ).

Although most DCVs use MoDCs that have been induced to differentiate in vitro, long-term in vitro culture can result in decreased MoDC migration and functional loss (7). Therefore, MoDC is probably not the best DC subpopulation for vaccine manufacturing, and the development of vaccines based on naturally circulating DC subtypes such as cDC, pDC, or Langhans cell may achieve better results. Among these DC subsets, it has been proven that cDCs have a stronger ability to induce CD8⁺ T cell response (7, 103). To date, the difficulty of producing cDC1/2 in large quantities from patients remains an obstacle to cDC-based DCVs. Therefore, future efforts should focus on solving the technical and cost issues of generating large numbers of cDCs.

ICIs have achieved clinical success in effectively treating various cancers, which are related to specific immune biomarkers to guide application. Immune biomarkers such as tumor mutation burden and PD-L1 positivity provide accurate and non-invasive means for patient preselection (13, 15, 144), which are of great value to the success of antitumor immunotherapy. Unfortunately, the lack of strong patient-preselected biomarkers immensely limits the guide of application of DCV; therefore, there is a surge in urgency to screen out biomarkers that are most likely to predict a positive patient response to DCV. The selection of patient subgroups by specific biomarkers that improve the likelihood of a subject’s response to DCV will help guide the design of clinical trials.

The glioma microenvironment is composed of various immunosuppressive cells, all of which are of great importance in disease progression. Targeting only one type of cell is not sufficient to modify the entire TME. Therefore, to improve DC function in the TME, it may be necessary to combine it with a variety of other immunotherapy methods to get over the negative effects of immunosuppression and immune checkpoint modulation.

In preclinical models, anti-CD25 antibodies are commonly used to deplete Tregs (145–147) or limit their immunosuppressive function by blocking molecules like PD-1, CTLA-4, and Tim-3 or enzymes such as IDO (146, 148, 149). However, it has been reported in murine models that high-dose unfractionated radiotherapy or low-dose TMZ or cyclophosphamide chemotherapy can deplete Tregs (150, 151). An encouraging approach is reducing the effect of Tregs by combining radiotherapy with anti-IDO, which eventually improves the survival of mice (152). When anti-CD25 therapy is combined, beneficial effects on survival caused by DCV have been reported by several other studies, especially when Tregs are depleted before vaccination (145, 153).

In vitro, paclitaxel promoted MDSC differentiation into DCs in a TLR4-independent manner (154). Docetaxel induces the transformation of MDSCs into M1-like macrophages and selectively enhances CTL responses (155). All-trans retinoic acid can promote MDSC maturation (156). In addition, low doses of 5-FU (157), capecitabine (158), etc. can deplete MDSCs. Pexidatinib reduced MDSCs and M2-like TAMs by blocking CSF-1 receptor signaling (159), while STAT3 inhibitors reduced MDSCs and impaired their function (160).

Blocking the CSF-1/CSF-1R axis prevents monocyte differentiation, thereby reducing the number of TAMs, while also reducing the survival of existing TAMs (161, 162). Blockade of the CCL2/CCR2 axis inhibited monocyte recruitment but did not affect the TAMs formed (163, 164). CD47 is a “don’t eat me” signal, and blocking CD47/SIRP enhances TAM phagocytosis of tumor cells (165). Oncolytic virotherapy repolarizes M2-like TAMs to M1-like TAMs (166, 167).

Although ICIs have achieved impressive results in various tumors, and immune checkpoint inhibitors have improved survival in a mouse GBM model (146, 148, 149, 168), ICIs alone have not been effective in the treatment of GBM (17, 169). However, the combination of ICI and DCV was more effective than DCV alone. Currently, the most commonly used target of ICIs is PD-1, followed by CTLA-4 (10).

To date, there haven’t been any reported clinical trials for glioma using intratumoral injection of DCV yet. It has been shown that in an orthotopic GL261 glioma murine model, compared with subcutaneous injection of GL261 lysate-loaded DCs, intratumoral injection is less effective; however, combining these two administration routes is more effective than subcutaneous injection alone (170). Intratumoral injected DCs could be detected in the tumor parenchyma while not in the cervical lymph nodes. Therefore, intratumoral injection of DC may have a distinct mechanism to improve survival. This may be because intratumoral DC injection enhances the anti-tumor immune response induced by subcutaneous injection of DC by pro-immunomodulating cytokines in the TME, reducing Treg cells, and directly inhibiting tumor proliferation by TNF (170, 171). Therefore, combining the two in clinical trials may lead to better results.