Regulation of PD-L1 expression in the tumor microenvironment

In physiological conditions, the activities of T cells are intricately regulated. T cell immunity selectively eliminates pathogens and abnormal cells but avoids attacking normal cells, termed immune homeostasis [ 1]. Programmed cell death-1 (PD-1, which is encoded by PDCD1) and programmed death-ligand 1 (PD-L1, which is encoded by CD274) are vital proteins in maintaining immune homeostasis [ 2]. The PD-1/PD-L1 pathway restrains the hyperactivation of immune cells and prevents autoimmune diseases [ 3]. However, in the tumor microenvironment (TME), the PD-1/PD-L1 axis is hijacked by cancer cells to escape immune surveillance [ 4]. The overexpressed PD-L1 on cancer cells binds to the PD-1 on tumor-infiltrating lymphocytes (TILs), which counteracts the TCR-signaling cascade by phosphorylating SHP-2 [ 5, 6]. As a result, T cell activation is impaired. Apart from cancer cells, some other types of cells in the TME, such as macrophages, dendritic cells (DCs), activated T cells, as well as cancer-associated fibroblasts, also express PD-L1 [ 7]. These components orchestrate an immunosuppressive microenvironment, supporting tumor growth.

Inhibiting the PD-1/PD-L1 signaling is a feasible strategy to normalize the dysregulated TME [ 8]. Up to now, α-PD-1/PD-L1 treatments have exhibited potent antitumor activities in various cancers, such as melanoma, non-small cell lung cancer (NSCLC), gastric cancer, liver cancer, urothelial cancer, lymphoma, and all MSI-high cancers [ 2, 9– 19]. Commonly, the PD-L1 protein level is the primary standard to select patients who are more likely to respond to α-PD-1/PD-L1 treatments [ 20, 21]. However, the PD-L1 level is determined by several factors, which results in the different significances of PD-L1 positivity or negativity. The PD-L1 positivity might result from immune response-induced PD-L1 expression or oncogenic constructive PD-L1 upregulation [ 22]. For the latter, in the absence of pre-existing immune response, patients with PD-L1 positive tumors commonly are resistant to α-PD-1/PD-L1 therapies [ 20].

On the contrary, patients with PD-L1 negative tumors might respond to α-PD-1/PD-L1 treatment when undergoing combination therapies that promote T cell infiltration [ 22]. Therefore, an in-depth understanding of PD-L1 regulation is valuable for efficacy prediction and patient selection. In this review, we summarized the latest advances of PD-L1 regulation, including genomic alterations, epigenetic modification, transcriptional regulation, post-transcriptional modification, and post-translational modification. Moreover, we discussed the potential applications of these findings in the clinic.

In some cancers such as classical Hodgkin lymphoma and small-cell lung cancer, the copy number of chromosome 9p24.1 (where CD274 resides) was increased [ 23, 24]. The chromosome rearrangement caused CD274 amplification without influences on the open reading frame (Fig.  1) (Table 1) [ 24]. Besides, in mediastinal large B-cell lymphoma, the increased transcriptional expression of CD274 was related to an adjacent ectopic promoter or enhancer by translocation [ 25]. These findings indicated that genomic alterations contributed substantially to cancer immune escape, which might be a potential biomarker for patient selection.

Epigenetic regulations such as methylation and histone acetylation determine the PD-L1 expression as well (Fig.  1). Tri-methylation of histone H3 on lysine 4 (H3K4me3) is generally believed as a histone modification promoting gene transcriptions [ 26]. In pancreatic cancer, MLL1 protein could bind to the CD274 promoter to catalyze H3K4me3, leading to the increased expression of PD-L1 [ 27]. In agreement, the MLL1 inhibitor had a synergistic effect with α-PD-1/PD-L1 therapy [ 27]. On the contrary, tri-methylation of histone H3 on lysine 27 (H3K27me3) relates to transcription suppression [ 28]. In hepatocellular carcinoma, enhancer of zeste homolog 2 negatively regulated PD-L1 expression by promoting H3K27me3 [ 29].

Besides the methylation of histone, the methylation of DNA at CpG islands regulated PD-L1 expression [ 30]. Inhibiting methylation of DNA by DNA methyltransferase inhibitors (DNMTis) increased PD-L1 level in cancer cells [ 31, 32]. The authors assumed that DNMTis elevated the expression of hypermethylated endogenous retroviruses in cancer cells, which might activate the innate immune response and lead to IFN-γ-stimulated PD-L1 expression [ 30]. Moreover, the methylation of some specific CpG loci in the CD274 promoter determined the level of CD274 mRNA [ 33– 35]. In NSCLC, TGF-β1 impaired the activity of DNMTs, demethylated the CD274 promoter, and increased PD-L1 expression [ 36]. Notably, in patients with recurrent gastric cardia adenocarcinoma, PD-L1 expression was reduced after α-PD-1/PD-L1 treatment [ 37]. Further investigations indicated that the CD274 promoter was more hypermethylated in the relapsed tumors than in the primary tumors without α-PD-1/PD-L1 treatment [ 37]. In murine tumor models, the combination therapy of hypomethylating agent azacytidine and α-PD-1 showed a more significant antitumor effect than α-PD-1 monotherapy [ 37].

Histone acetylation is an epigenetic modification enhancing gene transcription [ 38]. In some drug-resistant cancer cells, hyperactivated JNK/c-Jun signaling suppressed the histone deacetylase 3 (HADC3) expression, thereby elevating the histone H3 acetylation of the CD274 promoter [ 39]. The HADC inhibitor had a synergistic effect with α-PD-1 in the B16F10 tumor model [ 40]. Furthermore, HADC inhibitor-mediated PD-L1 upregulation was observed in other types of cancers [ 41, 42]. These findings provide a rationale to combine α-PD-1/PD-L1 treatments with HDAC inhibitors.

Post-translational modifications, including ubiquitination, phosphorylation, glycosylation, palmitoylation, and SUMOylation, play a vital role in regulating protein stability, activation, localization, as well as interaction [ 137]. Aberrant post-translational modification patterns participated in PD-L1 upregulation in the TME (Fig.  1) [ 138].

A growing body of evidence suggests that it is inaccurate to select patients merely by PD-L1 abundance. Understanding the difference between inflammation-induced PD-L1 and oncogenic signal-mediated constitutive PD-L1 is helpful to patient selection. For instance, for EGFR mutant NSCLC patients, α-PD-1 therapy's efficacy was poor despite the high level of PD-L1 [ 153]. The EGFR mutation-driving NSCLCs commonly harbor lower mutation burdens, and the lower immunogenicity leads to the resistance to α-PD-1 treatments [ 43]. This oncogenic EGFR-mediated PD-L1 expression could not reflect the real status of the TME. Alternatively, a comprehensive framework containing multiple surrogate markers such as tumor mutation burden would be valuable for selecting patients and predicting outcomes.

Besides, agents regulating PD-L1 expression might have a synergistic effect with the current immune checkpoint inhibitors (Fig.  2). Targeting therapies such as CDK4/6 inhibitor upregulated PD-L1 expression and promoted immune escape [ 141, 146]. This treatment-induced immune evasion could be overcome by combination therapies containing α-PD-1/PD-L1. Besides, adjuvant treatment regulating PD-L1 expression might elevate the sensitivity to α-PD-1/PD-L1 or other immune checkpoint inhibitors [ 144, 145]. For example, metformin downregulated PD-L1 by promoting endoplasmic-reticulum-associated degradation, and the combination therapy of metformin and α-CTLA-4 exhibited a synergistic antitumor activity [ 144].

Generally, in the TME, the expression of PD-L1 is regulated by numerous factors, including inflammatory stimuli and oncogenic pathways at the levels of transcription, post-transcription, and post-translation. Exploring potential PD-L1 regulators helps select patients and overcome resistance to α-PD-1/PD-L1 treatments. Besides, the agents regulating PD-L1 expression might be possible adjuvant therapies for the current immune checkpoint inhibitors.