Combination strategies with PD-1/PD-L1 blockade: current advances and future directions

Antibodies targeting programmed cell death protein-1 (PD-1) or its ligand PD-L1 rescue T cells from exhausted status and revive immune response against cancer cells. Based on the immense success in clinical trials, ten α-PD-1 (nivolumab, pembrolizumab, cemiplimab, sintilimab, camrelizumab, toripalimab, tislelizumab, zimberelimab, prolgolimab, and dostarlimab) and three α-PD-L1 antibodies (atezolizumab, durvalumab, and avelumab) have been approved for various types of cancers. Nevertheless, the low response rate of α-PD-1/PD-L1 therapy remains to be resolved. For most cancer patients, PD-1/PD-L1 pathway is not the sole speed-limiting factor of antitumor immunity, and it is insufficient to motivate effective antitumor immune response by blocking PD-1/PD-L1 axis. It has been validated that some combination therapies, including α-PD-1/PD-L1 plus chemotherapy, radiotherapy, angiogenesis inhibitors, targeted therapy, other immune checkpoint inhibitors, agonists of the co-stimulatory molecule, stimulator of interferon genes agonists, fecal microbiota transplantation, epigenetic modulators, or metabolic modulators, have superior antitumor efficacies and higher response rates. Moreover, bifunctional or bispecific antibodies containing α-PD-1/PD-L1 moiety also elicited more potent antitumor activity. These combination strategies simultaneously boost multiple processes in cancer-immunity cycle, remove immunosuppressive brakes, and orchestrate an immunosupportive tumor microenvironment. In this review, we summarized the synergistic antitumor efficacies and mechanisms of α-PD-1/PD-L1 in combination with other therapies. Moreover, we focused on the advances of α-PD-1/PD-L1-based immunomodulatory strategies in clinical studies. Given the heterogeneity across patients and cancer types, individualized combination selection could improve the effects of α-PD-1/PD-L1-based immunomodulatory strategies and relieve treatment resistance.

Apart from PD-1 signaling, other immune checkpoints, abnormal angiogenesis, immunosuppressive immune cells or cytokines, cancer-associated adipocytes, and hyperactive cancer-associated fibroblasts also modulate cancer-immune set point and promote immune tolerance [15][16][17][18][19][20]. Logically, removing these negative factors could enhance the therapeutic effect of α-PD-1/PD-L1 and relieve drug resistance. On the other hand, some positive factors such as immunogenic cancer cell death, immunosupportive cytokines, and professional antigen presentation cells (pAPCs) contribute to immune clearance [21]. Correspondingly, strengthening these positive elements might boost the cancer-immune cycle, drive the transformation from cold to hot tumors, and improve the response to α-PD-1/PD-L1 therapies [21].

Conventional chemotherapy combined with α-PD-1/PD-L1
Chemotherapy modifying the TME Chemotherapy retards tumor growth mainly by arresting cell cycle, inhibiting DNA replication, disturbing cell metabolism, or suppressing microtubule assembly [32]. Besides, some cytotoxic chemotherapeutic drugs such as anthracycline and oxaliplatin could induce immunogenic cell death and stimulate antitumor immune response [33,34]. Immunogenic cell death is featured with some upregulated damage-associated molecular patterns (DAMPs) such as the secretion of IFN-I, the exposure of endoplasmic reticulum proteins especially calreticulin (CRT, an eat-me signal) on cell membrane, the leak of ATP (a findme signal), and the release of high-mobility group box 1 (HMGB1) [35]. The receptors of CRT, ATP, and HMGB1 are CD91, P2RX7, TLR4 on dendritic cells (DCs). The ATP-P2RX7 signaling recruits DCs into the tumor bed; the CRT-CD91 axis promotes DC to engulf cancer antigens; the HMGB1-TLR4 pathway facilitates the optimal cancer antigen presentation [36]. Collectively, the antigen capture and presentation of DC are enhanced, ultimately motivating adaptive antitumor immune response (Fig. 1a).

Chemotherapy combined with α-PD-1
Based on the immune-modulatory effect of chemotherapeutic agents, chemotherapy might be an appropriate partner with α-PD-1/PD-L1 to achieve both rapid and long-term cancer control. Nowadays, chemotherapy combined with α-PD-1/PD-L1 has become a standardof-care option for some cancer patients, and there are hundreds of ongoing clinical trials exploring the efficacy and safety of chemotherapy plus α-PD-1/PD-L1 (Table 2). In the clinical trial KEYNOTE-021 (phase 2), nonsquamous non-small cell lung cancer (NSCLC) patients receiving pembrolizumab combined with standard chemotherapy (carboplatin and pemetrexed) had a higher response rate and longer progression-free survival (PFS) than did patients receiving standard chemotherapy [52]. Based on the results of KEYNOTE-021, pembrolizumab plus chemotherapy has been approved by the FDA as the first-line treatment for advanced non-squamous NSCLC, regardless of PD-L1 level [52]. Later, in two phase 3 clinical studies (KEYNOTE-189 and KEYNOTE-407), pembrolizumab combined with standard chemotherapy led to a better overall survival (OS) and PFS in NSCLC patients, relative to chemotherapy monotherapy [53,54]. The results of KEYNOTE-407 engaged the FDA to approve pembrolizumab combined with chemotherapy for squamous NSCLC in 2018. Then, based on a string of successes (KEYNOTE-355, KEYNOTE-590, and KEYNOTE-811), the indication of pembrolizumab plus (See figure on next page.) Fig. 1 The synergistic antitumor efficacies and mechanisms of α-PD-1/PD-L1 in combination with chemotherapy, radiotherapy, or angiogenesis inhibitor. a Chemotherapy synergizes with α-PD-1/PD-L1. Some cytotoxic chemotherapeutic drugs could induce immunogenic cell death and stimulate antitumor immune response. Immunogenic cell death is featured with some upregulated damage-associated molecular patterns (DAMPs) such as calreticulin (CRT), ATP, and high-mobility group box 1 (HMGB1). The ATP-P2RX7, CRT-CD91, and HMGB1-TLR4 pathways facilitate the antigen capture and presentation of DC, ultimately motivating adaptive antitumor immune response. Apart from immunogenic cell death, low-dose chemotherapy depletes regulatory T cells (Tregs) and promotes the repolarization of tumor-associated macrophage (TAM) from M2-like to M1-like phenotype. b Radiotherapy synergizes with α-PD-1/PD-L1. Firstly, radiotherapy could induce immunogenic cell death, enhance antitumor immune response, promote T cell infiltration, expand T-cell receptor (TCR) repertoire in the TME. Secondly, radiotherapy upregulates the expression of PD-L1 on tumor cells, which might be utilized by additional α-PD-1/PD-L1. Thirdly, radiotherapy increases the MHC-I on tumor cells and relieves resistance to α-PD-1/PD-L1. c Angiogenesis inhibitor synergizes with α-PD-1/PD-L1. Angiogenesis inhibitor blocks proangiogenic pathways, promotes vessel normalization, improves tumor perfusion and oxygenation, restores the hypoxic TME, and enhances drug delivery. Also, angiogenesis inhibitor reshapes the TME: promoting T cell infiltration and DC maturation, enhancing the differentiation towards M1-like macrophage, decreasing the ratio of Treg and MDSC, and alleviating hypoxia-induced PD-L1 chemotherapy was expanded to advanced triple-negative breast cancer (TNBC), esophageal cancer, gastroesophageal junction cancer (GEJC) [55-57].
Generally, pembrolizumab has a great advantage on chemoimmunotherapy, with a broad range of indications. The FDA rarely approves chemoimmunotherapeutic strategies with other α-PD-1 drugs (except for nivolumab combined with chemotherapy for gastric cancer and GEJC) [58]. In China, the NMPA approved sintilimab plus pemetrexed and platinum as the first-line treatment for advanced non-squamous NSCLC, based on the results of . In addition, the NMPA approved sintilimab plus gemcitabine and platinum as the first-line treatment for advanced squamous NSCLC, based on the results of ORIENT-12 [60]. In 2020, the NMPA also approved camrelizumab plus carboplatin and pemetrexed as the first-line treatment for non-squamous NSCLC, based on the results of CameL [61]. Later in 2021, the NMPA approved camrelizumab plus gemcitabine and cisplatin (for advanced nasopharyngeal carcinoma) and tislelizumab plus chemotherapy (for NSCLC) [62][63][64].

Chemotherapy combined with α-PD-L1
Besides α-PD-1, α-PD-L1-based chemoimmunotherapy also attracts intensive attention, especially chemoimmunotherapeutic regimens with atezolizumab. IMpower150 is the pioneer of this series of studies, assessing the efficacy of atezolizumab plus angiogenesis inhibitor and chemotherapy in advanced non-squamous NSCLC [65]. Based on the results of IMpower150, the FDA approved atezolizumab plus bevacizumab, paclitaxel, and carboplatin as the first-line treatment for advanced nonsquamous NSCLC [65]. Subsequently, the FDA approved atezolizumab plus chemotherapy for TNBC (atezolizumab plus nab-paclitaxel, based on IMpassion130), SCLC (atezolizumab plus carboplatin and etoposide, based on IMpower133), and non-squamous NSCLC (atezolizumab plus nab-paclitaxel and carboplatin, based on IMpower130) [66][67][68]. Moreover, based on the results of CASPIAN, durvalumab combined with platinum plus etoposide therapy was approved for SCLC in the US [69]. Presently, there are still dozens of chemoimmunotherapeutic regimens with α-PD-1/PD-L1 awaiting approval in the US and China.

Radiotherapy combined with α-PD-1/PD-L1
The mechanisms by which radiotherapy synergizing α-PD-1/PD-L1 Like some chemotherapeutic drugs, radiotherapy could induce immunogenic cell death and enhance antitumor immune response [70]. On the one hand, immunogenic cell death-associated DAMPs and cytokines especially IFN-I recruit immune cells and promote the function of DCs. On the other hand, released tumor antigens could be captured by DCs and presented to T cells [70]. Consequently, radiotherapy not only eliminates local lesions but also stimulates the systemic antitumor immune response (also known as abscopal effects) [71]. Previous preclinical and clinical studies demonstrated that radiotherapy could synergize α-PD-1/PD-L1 in multiple manners. Firstly, radiotherapy promoted T cell infiltration, increased the number of TILs, and expanded T-cell receptor (TCR) repertoire in the TME [72,73]. Secondly, radiotherapy upregulated the expression of PD-L1 on tumor cells, which can be utilized by additional α-PD-1/ PD-L1 [74]. Thirdly, radiotherapy increased the MHC-I on tumor cells and relieved resistance to α-PD-1/PD-L1 ( Fig. 1b) [75]. However, some problems have not been well addressed, including the fractionation, dose, schedule of radiotherapy, irradiated tumor volume, irradiated regional lymph nodes, and the schedule of α-PD-1/ PD-L1 post-radiotherapy [76].
In the phase 2 study NCT02904954, SBRT combined with durvalumab acquired a superior antitumor effect to durvalumab in early-stage NSCLC [83]. In the combination therapy arm, patients received 24 Gy SBRT before durvalumab treatment (given in three consecutive daily fractions of 8 Gy) [83]. The major pathological response rate was significantly higher in the SBRT combined with durvalumab arm than that in the durvalumab arm [83]. Additionally, the results of the phase 3 study NCT02125461 indicated that sequential durvalumab treatment markedly improved the PFS and OS of NSCLC patients undergoing chemoradiotherapy [84]. However, in the phase 2 study NCT02684253, SBRT combined with nivolumab was not superior to nivolumab in response rate, PFS, and OS in advanced HNSCC [85]. Furthermore, in the phase 3 study NCT02952586 exploring the efficacy of avelumab plus standard-of-care chemoradiotherapy in HNSCC, it did not meet the primary endpoint (PFS) [86]. Considering the multiple variants in the combination therapy such as dose, volume, fractionation, sequence, more efforts are needed to explore optimal radioimmunotherapy schemes.

Abnormal angiogenesis hampering the antitumor immune response
Hyperactive metabolism and incommensurate blood supply contribute to the hypoxic and acid TME [87]. As the feedback on hypoxia, the levels of some pro-angiogenic cytokines such as vascular endothelial growth factor (VEGF) and angiopoietin 2 (ANGPT2) are upregulated, driving angiogenesis [88]. The disorganized angiogenesis promotes the formation of the immunosuppressive TME [16]. Firstly, the immature and leaky vessels lead to increased interstitial fluid pressure, which hinders blood perfusion and immune cell infiltration [89]. Secondly, VEGF could inhibit the maturation of DC, induce the exhaustion of T cells, promote the proliferation of Tregs, and increase the ratio of MDSCs [90][91][92][93]. Thirdly, despite without direct influence on T cells, ANGPT2 recruits Tie-2-expressing monocytes, enhances the differentiation towards M2-like macrophages, and upregulates the expression of IL-10 [94][95][96][97]. Moreover, other proangiogenic cytokines such as placental growth factor (PLGF) and TGF-β also contribute to immunosuppression [98].
In 2019, pembrolizumab combined with axitinib was approved by the FDA as the first-line treatment for advanced RCC, based on the results of KEYNOTE-426 (Table 4) [109]. At a median follow-up time of 30.6 months, the median OS and PFS were longer in the pembrolizumab combined with axitinib arm compared to those in the sunitinib arm [109]. Moreover, pembrolizumab plus lenvatinib was also approved for advanced endometrial carcinoma [110]. Additionally, as mentioned above, the FDA approved atezolizumab plus bevacizumab and chemotherapy as the first-line treatment for advanced non-squamous NSCLC based on the results of IMpower150 [65]. Then, in 2020, the FDA approved atezolizumab combined with bevacizumab for advanced HCC based on the data of IMbrave150 [111]. Besides pembrolizumab plus axitinib, nivolumab plus cabozantinib (based on CheckMate-9ER) [112] and avelumab plus axitinib (based on JAVELIN Renal 101) [113] were also approved by the FDA as the initial-line treatment for RCC.

α-CTLA-4 plus α-PD-1/PD-L1
CTLA-4 is primarily expressed on activated T cells and Tregs, as a negative regulator for T cell activation [121]. On the one hand, CTLA-4 could competitively suppress the binding of CD28 to CD80/CD86, halting the secondary signal of T cell activation [122]. On the other hand, CTLA-4 engagement with CD80/CD86 counteracts TCR-induced downstream signaling and suppresses PI3K-Akt pathway (vital signaling of T cell activation), via SHP-2 and protein phosphatase 2A (PP2A) [123,124]. Additionally, CTLA-4 could capture, remove, and degrade its ligands CD80/CD86 from nearby APCs by trans-endocytosis, further hampering the co-stimulatory signal [125]. It is commonly believed that CTLA-4 signaling mainly undermines T cell priming in secondary lymphoid organs [126]. Ipilimumab (developed by Bristol-Myers Squibb) is the first approved α-CTLA-4 drug, initially used for advanced melanoma [127]. So far, the mechanism of antitumor activity of ipilimumab is still unclear. Theoretically, ipilimumab blocks the binding of CTLA-4 to CD80/CD86, removes the immunoinhibitory signal, and promotes T cell priming. However, multiple studies have been confirmed that antibody-dependent cell-mediated cytotoxicity of Treg also substantially contributes to the antitumor activity of ipilimumab [128][129][130].
Apart from efficacy, it is concerned that dual PD-1/ PD-L1 and CTLA-4 blockade might lead to serious immune-related adverse events (irAEs) such as colitis, hypophysitis, pneumonitis, and thyroiditis [153]. Therefore, ipilimumab is commonly administrated at a reduced dose [154], which might weaken the efficacy of combination therapy. A preclinical study found that prophylactic TNF blockade could dissociate the efficacy and toxicity of α-CTLA-4 plus α-PD-1/PD-L1 therapy [155]. Further clinical investigations are needed to improve the safety and strengthen the efficacy of dual PD-1/PD-L1 and CTLA-4 blockade.

Epidermal growth factor receptor-tyrosine kinase inhibitor (EGFR-TKI) plus α-PD-1/PD-L1
EGFR is a member of ErbB family driving the initiation and development of multiple types of cancers [181]. Upon the engagement with its ligands (such as epidermal growth factor, transforming growth factor-alpha, amphiregulin), EGFR would be homodimerized or heterodimerized [182]. Then, the cytoplasmic tyrosine kinases domain of EGFR is phosphorylated, triggering the activation of PI3K-AKT and MAPK pathways [182]. Some cancers especially NSCLC are addicted to the hyperactive EGFR pathway [183]. Therefore, agents targeting EGFR could effectively suppress the growth of these EGFR-addictive cancers. Generally believed, the efficacy of α-PD-1/PD-L1 is modest in EGFR-mutated patients [184,185], which might be attributed to the lack of concurrent TIL and PD-L1 expression, low tumor mutation burden, or increased Tregs in the TME [186]. Recent studies demonstrated that EGFR-TKI could promote T cell infiltration, decrease the ratios of tumor-infiltrating Treg and M2-like macrophage, and improve the responsiveness to α-PD-1/PD-L1 in EGFR-mutated models [17,187]. Besides, activated EGFR signaling contributes to the upregulated PD-L1 on cancer cells, and EGFR-TKI might cooperate with α-PD-1/PD-L1 to attenuate immune evasion [188]. Collectively, EGFR-TKI plus α-PD-1/PD-L1 therapy would maximize the efficacy of immunotherapy in patients with EGFR-mutated cancers (Fig. 2b).
In the phase 1 trial CheckMate-012, nivolumab combined with erlotinib showed potent and durable antitumor activity in EGFR-mutated NSCLC patients, with tolerable adverse events (no grade 4/5 adverse event reported) ( Table 6) [189]. Moreover, in the phase 1 study NCT02013219, EGFR-mutated NSCLC patients received erlotinib (150 mg QD for 7 days), followed by erlotinib (150 mg QD) plus 1200 mg atezolizumab (1200 mg, q3w) [190]. The ORR of combination therapy was as high as 75% in the expansion-stage group, and tumor-infiltrating CD8+ T cell was increased in 8/13 paired biopsies after 7-day erlotinib treatment [190]. No pneumonitis and dose-limiting toxicity were reported in this study [190]. However, a retrospective study found that patients receiving nivolumab plus erlotinib might have a higher risk of treatment-associated interstitial pneumonitis (Odds ratio: 4.31, P < 0.001), relative to patients undergoing EGFR-TKI monotherapy [191]. Additionally, in the phase 1 study TATTON, the incidence rate of interstitial lung disease in the osimertinib (a third-generation EGFR-TKI) plus durvalumab arm was unexpectedly high (22%), leading to the termination of patient enrollment [192]. Because of the increased risk of treatment-associated interstitial lung disease, a phase 3 clinical trial CAU-RAL was stopped early [193]. Although the mechanisms of combination therapy-caused irAEs are still unclear, it has been confirmed that treatment sequence and timing are closely associated with the incidence of irAE. PD-1/ PD-L1 blockade followed by osimertinib led to a higher incidence rate of irAE, while osimertinib followed by PD-1/PD-L1 blockade decreased the risk of irAE [194]. This phenomenon appears to be unique to osimertinib [194]. The efficacy and toxicity of EGFR-TKI plus α-PD-1/PD-L1 should be further valuated in patients harboring EGFR-mutations.

Anaplastic lymphoma kinase (ALK)-TKI plus α-PD-1/PD-L1
ALK is a receptor tyrosine kinase belonging to insulin receptor superfamily [195]. EML4-ALK fusion is the most common ALK arrangement variant in NSCLC patients [196]. The constitutively activated ALK fusion gene promotes cancer development by initiating some oncogenic pathways including MAPK, PI3K-Akt, JAK-STAT, and PLCγ [197]. ALK-TKI has dramatically prolonged the survival of ALK-arranged patients [198]. Similar to EGFR-mutation, ALK rearrangement is also related to the poor response to α-PD-1/PD-L1 [199]. A retrospective analysis showed that the co-expression of PD-L1 and CD8 was rare in ALK-arranged tumors, which might contribute to the lower response rate to α-PD-1/PD-L1 [200]. Overexpressed ALK fusion protein increased PD-L1 level, promoting the apoptosis of tumor-infiltrating T cells [201]. Besides, ALK inhibition induced immunogenic cell death in ALK-arranged cancer cells and conferred the protection of tumor rechallenge in the mouse model [202]. Combination therapy of α-PD-1 and ceritinib had an enhanced antitumor efficacy in NPM1-ALK + R80 model [202].
It should be noted that ALK-TKI combined with α-PD-1/PD-L1 might increase treatment-associated hepatotoxicity. In the phase 1/2 study CheckMate-370, 38% of patients receiving nivolumab plus crizotinib developed severe hepatic toxicities, leading to the termination of the enrollment [203]. Moreover, pembrolizumab plus crizotinib also showed intolerable hepatotoxicity in NSCLC [204]. Conversely, some other combination strategies such as atezolizumab plus alectinib and avelumab plus lorlatinib had a manageable safety profile, indicating the hepatotoxicity might be ALK-TKI specific [205,206]. Additionally, the timing and sequence of combination therapy also influence treatment toxicity, which should be further validated in clinical studies [186,207].

RAS-targeted therapy plus α-PD-1/PD-L1
RAS family (KRAS, NRAS and HRAS) is frequently mutated in cancer cells. Mutated KRAS is a well-established driver gene of NSCLC, colorectal cancer, and pancreatic cancer [208]. In normal cells, RAS is activated by growth factor receptors such as EGFR. RAS is a small G protein, toggling between GTP-bound state (active) and GDP-bound state (inactive). In active state, RAS triggers several downstream pathways including MAPK and PI3K-AKT [209]. In tumor cells, mutations in RAS disturb this switch between GTP-bound state and GDPbound state. As a result, RAS is locked in GTP-bound state, leading to the hyperactive downstream pathways and tumor growth [209,210]. Recent studies have shown that RAS and its downstream pathways participated in cancer immune escape: negatively regulating MHC-I expression on cancer cells, increasing the cell-intrinsic PD-L1 level, elevating immune suppression-associated cytokine production [211,212]. RAS-targeted therapy abrogated RAS-MAPK/PI3K-AKT-involved immune evasion, synergizing with α-PD-1/PD-L1 [213,214].
In the phase 1 study NCT01988896, atezolizumab plus cobimetinib (MEK inhibitor) had a manageable safety profile and clinical activity in advanced solid tumors, regardless of KRAS/BRAF status [215]. However, in the phase 2 study NCT02322814, atezolizumab plus cobimetinib and taxane had no improvement in ORR in TNBC, relative to cobimetinib plus taxane [216]. Moreover, in the phase 3 study NCT02788279 exploring the efficacy of atezolizumab plus cobimetinib in metastatic colorectal cancer, the primary endpoint of improved OS (atezolizumab plus cobimetinib vs. regorafenib) could not be reached [217]. At present, other combination strategies including α-PD-1/PD-L1 plus AMG 510 Table 6 The clinical trials exploring the efficacy of α-PD-1/PD-L1 combined with targeted therapy (except for angiogenesis inhibitor) Abbreviations: NSCLC non-small cell lung cancer, PFS progression-free survival, OS overall survival, RP2D recommended phase 2 dose, DOR duration of response, DCR disease control rate, CBR clinical benefit rate

Poly (ADP-ribose) polymerase (PARP) inhibitor plus α-PD-1/ PD-L1
Normal cells preferentially repair double strand break (DSB) via homologous recombination (HR). However, some HR-deficient (e.g. BRCA1/2 mutant) cancer cells only repair DSB by nonhomologous end joining, which is a low fidelity repair pathway [219]. As a result, chromosomal rearrangements are accumulated in cancer cells, eventually leading to cell death [220]. Therefore, intact single-strand break (SSB) repair pathway is essential to these HR-deficient cancer cells. Based on this synthetic lethality theory, interfering SSB could destroy HR-deficient cancer cells [221]. As the core of SSB repair, PARP is the ideal target for drug development [222]. Besides synthetic lethal effect, PARP inhibitor (PARPi) modulates the TME and promotes the antitumor immune response [223]. Firstly, PARPi activates cGAS-STING pathway in cancer cells and increases T cell recruitment [224]. Moreover, PARPi upregulates PD-L1 expression by inactivating GSK3β signaling, which attenuates antitumor immunity [225]. Inspired by the results of preclinical studies, numerous clinical studies are ongoing to evaluate the efficacy of PARPi combined with α-PD-1/PD-L1 [219].
SHP2 is an oncogenic protein belonging to protein tyrosine phosphatases family [250]. As the convergent node of MAPK, PI3K-AKT, JAK-STAT, and PD-1 pathways, SHP2 widely regulates multiple cancer-associated processes such as cell survival and immune escape [251]. SHP2 inhibition increased PD-L1 and MHC-I expression by augmenting intrinsic IFN-γ in cancer cells [252]. SHP2 inhibitor enhanced the efficacy of α-PD-1 in murine tumor models [252][253][254]. A clinical study exploring SHP2 inhibitor combined with α-PD-1 is still ongoing (NCT04000529), and the final data of this combination study are not yet available [255].

STING pathway and STING agonist
Cytosolic chromatin fragments and micronuclei are commonly accumulated during malignant transformation, increasing the probability of cytosolic DNA leakage in cancer cells or tumor-derived DNA uptake in DCs [256]. cGAS-STING pathway is a cytosolic DNA sensing signaling. Cytosolic dsDNA binds to cGAS, catalyzing the generation of cyclic GMP-AMP (cGAMP). Stimulated by cGAMP, STING changes from monomer to dimer and translocates from ER to perinuclear microsome. Then, STING recruits and phosphorylates TBK1, which further activates downstream IRF3 and upregulates IFN-I [257][258][259]. Besides, STING also increases IFN-I by activating NF-κB pathway [260]. IFN-I is a versatile immune stimulator that could enhance the functions of DC, NK, and T cells [261]. Given the critical role of cGAS-STING pathway in bridging innate and adaptive immunity, STING is the potential target for cancer immunotherapy.
Dimethyloxoxanthenyl acetic acid (DMXAA) is the first STING agonist which failed in the clinical trials [262]. Further investigation has identified that DMXAA is a mouse-specific STING agonist, with a subtle influence on human STING pathway [263,264]. Sharing similar structures and biological characteristics with cGAMP, some natural and artificially synthetic cyclic dinucleotides (CDNs) are developed as STING agonists for cancer immunotherapy [265][266][267]. Generally, CDNs have two main flaws: poor transmembrane capability and depending on intratumor injection. Recently, some novel STING agonists such as diABZI and MSA-2 have been developed which could be systemically administrated [268,269]. Besides, manganese is also identified as a natural STING agonist, playing an important role in antitumor immunity [270,271].

Bispecific/bifunctional antibody targeting PD-1/ PD-L1
Dual targeting by bispecific/bifunctional antibodies has emerged as an option for combination therapy. Bispecific/bifunctional antibody simultaneously blocks two molecules with one drug, having a strategic advantage over the combination therapy (Table 7) [277].

Metabolic modulators plus α-PD-1/PD-L1
The engagement of adenosine 2A receptor (A2AR) with adenosine elicits immunoinhibitory effects: suppressing the activities of tumor-infiltrating CD8+ T cells and hampering the function and differentiation of DCs [333,334]. The accumulated adenosine in the TME promotes cancer immune evasion, and A2AR blockade rescues immune cell function [335,336]. The results of a phase 1 study showed that ciforadenant (A2AR inhibitor) combined with atezolizumab effectively prolonged PFS and OS in RCC patients [337]. Besides, other metabolic modulators such as glutaminase inhibitor also had a synergistic effect with α-PD-1/PD-L1 in murine tumor models [338].

Chimeric antigen receptor-T (CAR-T) cell therapy plus α-PD-1/PD-L1
CAR-T cells are genetically engineered T cells, which could recognize and bind cancer antigen in an MHCindependent manner [339]. CAR-T cell therapy provides numerous cancer-reactive T cells and overcomes MHC downregulation-mediated cancer immune evasion [339]. However, the efficacy of CAR-T cell therapy is modest in most solid tumors, which is partly attributed to the immunosuppressive TME [340]. α-PD-1/PD-L1 enhanced CAR-T cell therapy by rescuing CAR-T cell exhaustion [341][342][343]. The results of phase 1 study demonstrated CAR-T cell therapy combined with α-PD-1/ PD-L1 had confirmed antitumor activity in patients with malignant pleural diseases [344]. Moreover, modified CAR-T cells, which secret PD-1-blocking single-chain variable fragments (scFv), had improved antitumor activity by an autocrine and paracrine manner [345]. This combination strategy protects CAR-T cells from immune exhaustion and optimizes CAR-T cell efficacy.

Perspective and conclusion
Although dozens of combination regimens exhibit potent antitumor activities in preclinical studies, some positive preclinical findings could not be validated in the clinic. At present, only combinations of α-PD-1/PD-L1 with chemotherapy, angiogenesis inhibitor, or α-CTLA-4 are approved by the FDA or NMPA. For most combinations, the striking antitumor activities are limited in animal tumor models. Therefore, how to select an optimal preclinical model is a grand challenge to identify the activities of combination regimens. Relative to widely used syngeneic murine models, humanized patient-derived models could provide a more precious efficacy evaluation. Besides, combination therapy increases the risk of irAEs and the cost of health care. Inappropriate combination treatments will expose patients to significantly higher toxicities. How to optimize administration regimen, including dosage, timing, and sequence, is another challenge for the development of combination therapy. Lastly, it is still unclear how to select appropriate combination therapy and find biomarkers predicting treatment response. Considering the heterogeneity and evolution of tumors, liquid biopsy could dynamically monitor the immune landscape of the TME and provide a real-time biomarker for guiding precision immunotherapy [346].
We believe individualized combination therapy should be provided based on patient's immune profiling and other predictive biomarkers. A comprehensive framework integrating genome, transcriptome, immune profiling, microbiome could be adopted to select patients benefiting from combinations. For patients with non-inflamed tumors, α-PD-1/ PD-L1 monotherapy scarcely provides clinical benefits, and a personalized combination is needed to overcome drug resistance. In the background of immune-excluded, Fig. 3 Therapies regulating the cancer-immunity cycle. The cancer-immunity cycle starts with cancer antigen release and ends with cancer cell-killing by immune cells. Each step in the cycle is regulated by various factors. The stimulatory factors (shown in green) enhance antitumor immunity, while the inhibitory factors (shown in red) undermine antitumor immunity. These factors provide potential therapeutic targets to promote antitumor immunity. The figure presents some of therapies regulating the cancer-immunity cycle. Abbreviations: CAF, cancer-associated fibroblasts; PARP, Poly (ADP-ribose) polymerase; DSB, double-strand break; STING, stimulator of interferon genes; A2AR, adenosine 2A receptor therapies such as TGF-β blocker could rescue the restrained T cell penetration by inhibiting CAF activities and reducing peritumoral collagen deposition. In the context of immune-desert, therapies such as radiotherapy, chemotherapy, and STING agonist could overcome low immunogenicity-mediated immune tolerance by inducing immunogenic cell death, increasing cancer antigen release, and promoting the function of APC. Combining these therapies with α-PD-1/PD-L1 simultaneously boosts multiple processes in the cancerimmunity cycle, reshapes the TME, and substantially promotes the transformation from non-inflamed to inflamed tumors (Fig. 3). Besides, with the development of next-generation α-PD-1/PD-L1 drugs such as bifunctional or bispecific antibodies, the indication of α-PD-1/ PD-L1 therapies would be greatly extended, and more patients could benefit from the updated α-PD-1/PD-L1 treatments.