Immunotherapy and tumor microenvironment

2.1.1 Checkpoint blockade antibodies

Immune checkpoints refer to a series of pathways that can regulate T cell activity as either co-inhibitory or co-stimulatory signals [14], and they function to protect the host against autoimmunity under normal conditions [15, 16]. Increasing evidence suggests that tumors use many of these pathways as important mechanisms to escape antitumor immune responses [6, 17, 18]. Among them, inhibitors targeting programmed cell death protein 1 (PD-1) and its ligand, PD-1 ligand (PD-L1 or B7H1), have shown the most impressive efficacy in clinical trials [3, 4]. PD-1 is mainly expressed on activated T cells [19]. Although PD-L1 expression is limited in normal tissues, it is greatly increased on some tumor cells [20]. Interestingly, PD-L1 expression can be upregulated on many cells if they are stimulated by inflammatory cytokines, especially interferons (IFNs) [20]. PD-L1 engagement of PD-1 on T cells inhibits their activation and induces exhaustion [21]. A paradigm has been proposed suggesting that tumor-expressed PD-L1 inhibits T cells located within the tumor, which leads to a failure of the host rejecting the tumor. This idea is supported by the initial observation that patients with PD-L1–positive tumor cells are more likely to respond to anti–PD-1 therapy [3]. With the growing number of patient samples, however, some patients with PD-L1–negative tumor cells have also been observed to respond to PD-1/PD-L1–blockade therapies [22]. Additionally, recent retrospective clinical studies show a high correlation between responses to PD-L1 blockade and PD-L1 expression on tumor-infiltrating immune cells [23]. These studies raised the possibility that PD-L1 expression on cells in the TME besides the tumor cells may also play important roles for immune evasion.

In order to increase the response rate to checkpoint blockade therapy, several combination therapies have been developed [24]. Among them, the combination of anti–PD-1 with anti–cytotoxic T-lymphocyte-associated antigen 4 (CTLA-4) has shown the best improvement in clinical trials [25–27]. CTLA-4 is another checkpoint molecule mainly expressed on regulatory T cells (Tregs) [28]. Blocking this pathway by anti–CTLA-4 depletes Tregs in tumor tissues, results in the expansion of antigen-specific cytotoxic T lymphocytes. The synergistic effects are mainly restricted to the local TME, as blocking lymphocyte trafficking after tumor is established has no effect on the synergy [29].

2.1.2 Immunostimulatory mAbs

In contrast to PD-1 and CTLA-4 that inhibit immune responses upon activation, the CD40/CD40 ligand (CD40L) pathway represents a different group of immune checkpoints that promote immune responses [30]. CD40 is expressed on B cells, monocytes, and dendritic cells (DCs). CD40L is primarily expressed on activated T cells and platelets [31]. Signaling through CD40 activates antigen-presenting cells and induces co-stimulatory and MHC molecule expression, resulting in increased antigen presentation and T-cell priming [32]. Agonist antibodies to CD40 have been developed to mimic CD40L engagement and can activate immune response both in vitro [33–35] and in vivo [36–38]; furthermore, several of these antibodies have entered clinical trials and show promising results [39, 40]. Interestingly, although the original paradigm suggests that T-cell priming is increased downstream of CD40 activation, later studies indicate that T cells may be dispensable, at least in some tumor models [39]. Specifically, tumor regression induced by anti-CD40 depends on macrophages, but not T cells, in a mouse model of pancreatic ductal adenocarcinoma. Anti-CD40 activates macrophages, which then translocate to tumor tissues and induce the depletion of tumor stroma, leading to tumor regression [39]. The idea that stromal cells may play a more important role for the efficacy of CD40 agonist antibody treatment is not only supported in tumor models but also in other disease models. Indeed, Bouchlaka et al. has found in a recent study that systemic injection of CD40 agonist antibody together with IL-2 induces a cytokine storm and lethality in aged, but not young, mice [41]. A follow-up study shows that CD40 signaling induces macrophage activation within the visceral adipose tissues [42]. These activated macrophages produce very high TNF levels, leading to a cytokine storm. Interestingly, one possible reason to explain why lethality is found only in aged mice is that they tend to accumulate more fat in their organs than young mice. These possible adverse effects should be taken into account when applying the various CD40 therapies to cancer patients, as the incidence of many cancers tend to be higher in aged people [43]. The CD40/CD40L story is an interesting case demonstrating that the TME can sometimes be the major target of immunotherapy rather than just the tumor cells.

2.3.1 Targeting chemokines

The fact that leukocyte trafficking and migration are tightly regulated by chemokines raises the interesting idea that manipulating the TME chemokine profile may be able to recruit sufficient numbers of effector cells into tumor tissues for tumor destruction [54–56]. Among all chemokines, those that mediate T cell chemoattraction, such as Chemokine (C-C motif) ligand 21 (CCL21) and C-X-C motif chemokine 10 (CXCL10), are of special interest [57].

CCL21, a CCR7 ligand, is an important chemokine that mediates DC and T cell homing to both lymphoid and non-lymphoid tissues [58]. In an early study, intratumoral injection of CCL21 completely eradicates tumors in lung cancer models [59]. Moreover, tumor regression mediated by CCL21-expressing DCs is accompanied by a significant increase in tumor-infiltrating CD8⁺ T cells, leading to tumor eradication [60]. In another study, Shields et al. knocked down CCL21 in a melanoma model [61] and surprisingly found that tumor cells expressing high CCL21 levels established an immune tolerant TME, which then facilitated tumor progression. In contrast, CCL21-deficient tumors grow much slower and induce tumor-specific T cell responses [61]. The conflicting conclusions from these studies may be due to different TMEs, such as the differing types of tumor models or CCL21-expressing cells. Collectively, these studies suggest that the effector and target cells should be chosen with care when using chemokines to modify the TME.

Another chemokine that recruits T cells is CXCL10, a CXCR3 ligand [62]. In a recent study, da Silva et al. demonstrate that dipeptidylpeptidase 4 (DPP4) is the enzyme responsible for inactivating CXCL10 in tumor tissues [63]. Inhibition of DDP4 enzymatic activity preserves active CXCL10, which increases CXCR3⁺ lymphocyte trafficking to tumor tissues. The number of tumor-infiltrating T cells, which express high CXCR3 levels, is significantly increased after DPP4 inhibition. This study further shows that the DDP4 inhibitor can be used in combination with other immunotherapies, including adjuvant-based immunotherapy and checkpoint blockade molecules [63]. The idea that increasing tumor-infiltrating lymphocytes (TILs) can improve current immunotherapies is attractive, even though the improvements shown in this study are not dramatic. It remains possible that more than one chemokine may be involved.

2.3.2 Targeting cytokines

The overall efficacy of chemokines in tumor immunotherapy are less impressive than other immunotherapies, implying that targeting a single chemokine may not be sufficient to recruit enough lymphocytes to completely control the tumor. In fact, when comparing chemokine profiles in tumors with different sensitivities to immunotherapy, our lab found that several chemokines related to T-cell trafficking were significantly different (unpublished data). In contrast to chemokines, cytokines are a broad group of proteins that can modulate immune responses [64]. Cytokines can directly activate immune effector cells or stimulate tumor stromal cells, leading to the production of chemokines and adhesion molecules for lymphocyte recruitment [65]. These functions suggest that targeting cytokines may also be efficient approaches for tumor immunotherapy depending on various tumor microenvironments.

The most extensively studied cytokines in terms of tumors are the type I IFNs, which are expressed by almost all cells and play important roles in antitumor responses [66]. Although systemic delivery of recombinant type I IFN proteins has been used in the clinic to treat some tumors, such treatment has been associated with high toxicity due to the broad distribution of IFN receptors [66]. Several groups have tested a second generation of IFNs that uses an antibody to guide IFNs specifically to tumor tissues. In one study, Xuan et al. generated a fusion protein linking IFN-α to an anti-CD20 antibody [67] and showed that it induced tumor regression in established tumors with higher efficacy and less toxicity than free IFN-α. This antitumor effect is attributed mostly to direct killing and can be independent of adaptive immunity, as this fusion protein can control xenograft tumors grown on nude mice. Knocking down the IFN receptor on tumor cells greatly reduced the antitumor effect, suggesting that the anti-CD20–IFN-α fusion protein mediates its antitumor effect mainly by directly inducing tumor cell apoptosis [67]. However, T cells might contribute to the overall antitumor activity when T cells are available. In another study, Yang et al. used an anti-EGFR antibody to guide IFN-β (anti-EGFR–IFN-β fusion protein) into tumors for tumor immunotherapy [68]. Similar to the previous study, the anti-EGFR–IFN-β fusion protein also showed potent antitumor activities. Interestingly, in contrast to the anti-CD20–IFN-α fusion protein that directly induces tumor cell apoptosis at a high dose, the data in the Yang et al. study clearly shows that the antitumor effect of the anti-EGFR–IFN-β fusion protein depends on adaptive immunity. Specifically, the anti-EGFR–IFN-β fusion protein targets DCs to increase antigen cross-presentation, leading to increased T-cell priming in the TME; in support of this, a conditional knockout of the IFN receptor on DCs completely abrogates the antitumor effect [68]. The discrepancy between these studies is not surprising. In fact, a similar contradictory phenomenon was observed when free type I IFNs were used for tumor immunotherapy [66], wherein free IFNs directly acted on tumor cells in some cases, while host immunity was likely required to control tumors in others.

Whether host immunity is required for the IFN effect seems to depend on the level of IFNs within the TME. Since a lower dose of IFN-β cannot occupy all the IFN receptors in the tumor tissue, host immunity is thus required for optimized tumor control [68]. In contrast, a high IFN-β dose delivered locally into the tumor tissues can induce destruction of the tumor vasculature, leading to tumor elimination [69]. The latter process is independent of T, B, or NK cells. Taken together, the interesting lesson learned from these studies is that the immunomodulator dose can result in completely different interactions between immunotherapy and the TME.

Sufficient lymphocyte infiltration into tumor tissues is critical for the success of many cancer therapies [70]. In fact, the presence of TILs correlates with a better prognosis to various antitumor therapies in a multitude of cancers [71–73]. An intriguing property of chemokines and cytokines for cancer immunotherapy is their potential to recruit leukocytes, especially T cells, into tumor tissues and change the TME from immunosuppressive to immunosupportive [13]. IFNs have been considered as such candidates [68]. However, preliminary studies in our lab showed that increased TILs were only observed in some tumor models, likely due to multiple downstream effects of IFNs (unpublished observations). Another candidate that can potently increase TILs in the TME is LIGHT (also known as Tumor Necrosis Factor Superfamily member 14 [TNFSF14]), a co-stimulatory molecule that can regulate T-cell activation [74]. Full-length, membrane-bound LIGHT can be secreted as a cytokine through protease digestion. Studies in our lab have shown that forced expression of LIGHT in tumor cells promotes the formation of lymphoid-like structures for direct T-cell sequestration and activation, leading to tumor regression [75, 76]. LIGHT is a ligand protein that binds to two different receptors, Herpesvirus Entry Mediator (HVEM) and Lymphotoxin beta Receptor (LTβR). While LIGHT binding to HVEM delivers a co-stimulatory signal to T cells [74], LIGHT binding to LTβR, which is commonly expressed on non-lymphoid cells, is critical for the formation of secondary and tertiary lymphoid structures [77, 78]. LTβR plays a pivotal role in lymph node formation and in the organization of distinct T-cell and B-cell zones in secondary lymphoid organs. Signaling via LTβR regulates the expression of various chemokines and adhesion molecules that control the migration and positioning of DCs and lymphocytes in the spleen [79]. Thus, in contrast to targeting just one specific chemokine, LIGHT activation of LTβR signaling in tumor stromal cells induces the expression of multiple chemokines and adhesion molecules for effective T cell recruitment to the TME. This induction of multiple chemokines makes LIGHT more efficient for lymphocyte recruitment and activation than single cytokine or chemokine. Taken together, extensive studies targeting the TME with cytokines that can induce an immunostimulatory environment support the idea that they may be an effective approach for cancer therapy [80].