PD-L1 degradation is regulated by electrostatic membrane association of its cytoplasmic domain

Wen, Maorong; Cao, Yunlei; Wu, Bin; Xiao, Taoran; Cao, Ruiyu; Wang, Qian; Liu, Xiwei; Xue, Hongjuan; Yu, Yang; Lin, Jialing; Xu, Chenqi; Xu, Jie; OuYang, Bo

doi:10.1038/s41467-021-25416-7

Download PDF

Article
Open access
Published: 24 August 2021

PD-L1 degradation is regulated by electrostatic membrane association of its cytoplasmic domain

Maorong Wen ORCID: orcid.org/0000-0002-8632-8247¹^na1,
Yunlei Cao^1,2^na1,
Bin Wu³^na1,
Taoran Xiao^1,2,
Ruiyu Cao^1,2,
Qian Wang^1,2,
Xiwei Liu^1,2,
Hongjuan Xue³,
Yang Yu³,
Jialing Lin ORCID: orcid.org/0000-0002-2126-1571^4,5,
Chenqi Xu ORCID: orcid.org/0000-0002-4968-6782^1,2,
Jie Xu⁶ &
…
Bo OuYang ORCID: orcid.org/0000-0002-2877-0635^1,2

Nature Communications volume 12, Article number: 5106 (2021) Cite this article

9171 Accesses
33 Citations
22 Altmetric
Metrics details

Subjects

Abstract

The cytoplasmic domain of PD-L1 (PD-L1-CD) regulates PD-L1 degradation and stability through various mechanism, making it an attractive target for blocking PD-L1-related cancer signaling. Here, by using NMR and biochemical techniques we find that the membrane association of PD-L1-CD is mediated by electrostatic interactions between acidic phospholipids and basic residues in the N-terminal region. The absence of the acidic phospholipids and replacement of the basic residues with acidic residues abolish the membrane association. Moreover, the basic-to-acidic mutations also decrease the cellular abundance of PD-L1, implicating that the electrostatic interaction with the plasma membrane mediates the cellular levels of PD-L1. Interestingly, distinct from its reported function as an activator of AMPK in tumor cells, the type 2 diabetes drug metformin enhances the membrane dissociation of PD-L1-CD by disrupting the electrostatic interaction, thereby decreasing the cellular abundance of PD-L1. Collectively, our study reveals an unusual regulatory mechanism that controls the PD-L1 level in tumor cells, suggesting an alternative strategy to improve the efficacy of PD-L1-related immunotherapies.

Proteome-scale discovery of protein degradation and stabilization effectors

Article 20 March 2024

Juline Poirson, Hanna Cho, … Mikko Taipale

PROTAC targeted protein degraders: the past is prologue

Article 18 January 2022

Miklós Békés, David R. Langley & Craig M. Crews

The JAK/STAT signaling pathway: from bench to clinic

Article Open access 26 November 2021

Xiaoyi Hu, Jing li, … Wei Wang

Introduction

The important discovery that programmed death ligand 1 (PD-L1) expression on tumor cells inhibits programmed cell death 1 (PD-1) on T cells to escape immune surveillance has opened an era of tumor immunotherapy^1,2,3,4. PD-L1 is a type I transmembrane protein⁵ containing an extracellular domain (ECD), a transmembrane domain (TMD), and a cytoplasmic domain (CD). It is highly expressed on the surface of many cancer cells^6,7. Early functional and structural studies have revealed that the canonical immunoglobulin (Ig)-like ECD of PD-L1 binds to PD-1 on T cells to inhibit their tumor-killing activity^5,8. The PD-L1-ECD-targeting drugs that were developed to recover the tumor-killing activity of T cells have achieved great clinical success in treating a wide range of cancers^9,10. However, the therapeutic resistance to these anti-PD-L1-ECD drugs has been increasingly observed in patients^11,12,13, highlighting the need for alternative strategies to overcome the acquired resistance.

The small cytoplasmic domain of PD-L1 (PD-L1-CD; containing residues 260–290) is involved in multiple regulation pathways controlling PD-L1 protein stability and degradation, and hence a potential new target inside the cell for the immune surveillance. PD-L1-CD was reported to bind the cullin 3^SPOP E3 ligase (SPOP) that decreases the cellular abundance of PD-L1 through ubiquitination-dependent degradation¹⁴. Deletion of the cytoplasmic tail disrupts the binding to SPOP and renders PD-L1 resistance to the ubiquitination-dependent proteasome-mediated degradation. Xu lab revealed that PD-L1 undergoes lysosome-dependent proteolysis after huntingtin interacting protein 1 related (HIP1R) binds to its C-terminus and delivers it to the lysosome¹⁵. Xu and Hung lab separately demonstrated that palmitolyation of C272 in the PD-L1-CD regulates PD-L1 stability and trafficking^16,17. Furthermore, PD-L1 was reported to contain a conserved class of sequence motifs in PD-L1-CD that mediate crosstalk with interferon signaling¹⁸. Besides the multiple regulatory roles of the cytoplasmic domain, several cancer-derived mutations were found within this domain, including R260C, R262K, D276Y, and T290M, alluding its critical role in cancer cell survival¹⁸. Interestingly, the peptides designed to target PD-L1-CD reduced the PD-L1 protein levels by enhancing the lysosomal degradation¹⁵ or blocking the palmitoylation¹⁶, resulting in reduced binding to PD-1 on T cells, thereby enhancing their antitumor activity. Therefore, targeting of PD-L1-CD within the cancer cell is a promising route for the development of anticancer immunotherapy^19,20.

Despite of the high biological relevance of PD-L1-CD in regulating PD-L1 levels in tumor cells, the molecular recognition and interaction of PD–L1–CD that control PD-L1 levels remain a mystery. Previous studies showed that PD-L1 is highly expressed on the plasma membrane in tumor cells^6,7, of which the inner leaflet is enriched in acidic phospholipids, including phosphatidylserine (PS) and phosphatidylinositol (PI). The asymmetric distribution of lipids results in negative charges on the cytoplasmic face of the plasma membrane²¹. Moreover, recent studies revealed that PD-L1 is not only located on the cell membrane, but also in extracellular vesicles, such as exosomes, which facilitates tumor cells to evade immune surveillance^12,13,22. Exosomes are also enriched in acidic lipids such as PS²³, indicating that PD-L1 has a preference to locate in an acidic lipid-rich environment. The palmitolyation of C272 stabilizes the membrane insertion of PD-L1-CD and thus suppresses PD-L1 degradation¹⁷, suggesting that the membrane association of PD-L1-CD has a significant effect on stabilizing PD-L1. More broadly, the asymmetric lipid distribution with excess negative charges on the plasma membrane appears to regulate the signaling via several type I transmembrane proteins, such as CD3ε/ζ and CD28 in T cells^24,25,26.

Therefore, in this study we performed structural and functional investigations of the lipid regulation of PD-L1-CD. We found that the membrane insertion of PD-L1-CD is regulated by negatively charged acidic phospholipids. The electrostatic interactions between acidic phospholipids and basic residues in PD-L1-CD proximal to the TM domain are critical for the membrane association. Using nuclear magnetic resonance (NMR) technology, we determined how PD-L1-CD binds to a lipid bicelle, in which three arginine residues are inserted into the lipid bilayer. Sequestration of these arginines into the plasma membrane enhanced the PD-L1 degradation in cells. Interestingly, we observed that metformin can disrupt the interaction between PD-L1-CD and membranes. Further, mutagenesis and cell-based functional assay showed that disruption of the membrane association by metformin reduces PD-L1 levels. And, the PD-L1-CD-membrane interaction controls the abundance of PD-L1 in tumor cells and thereby regulates the immunosurveillance. Thus, our study uncovers a molecular mechanism that regulates PD-L1 levels, providing insights into PD-L1-mediated immune evasion of tumor cells that can be potentially targeted by drugs.

Results

Acidic lipids regulate membrane interaction of PD-L1-CD

To explore whether acidic lipids regulate PD-L1-CD protein–membrane interactions, we expressed the residues 260–290 of human PD-L1 (CD_260-290) (Supplementary Fig. 1a) in E. Coli that was fused to the C-terminus of a His₈-SUMO sequence with a PreScission protease (3 C) cleavage site in between. The fusion protein was purified by nickel affinity chromatography followed by 3 C cleavage and reverse-phase HPLC to remove the His₈-SUMO tag (Supplementary Fig. 1b–d).

The purified CD_260-290 protein was incubated with 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC) liposomes supplemented with 25% 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine (DMPE), 1,2-dimyristoyl-sn-glycero-3-phosphoglycerol (DMPG), 1,2-dimyristoyl-sn-glycero-3-phospho-L-serine (DMPS), or 5 or 15% cardiolipin (CL), respectively. After an ultracentrifugation, CD_260-290 was found in the pellet associated with the liposomes containing the acidic phospholipid DMPG, DMPS or CL but not the zwitterionic phospholipid DMPE and/or DMPC (Fig. 1a). These results indicate that the acidic phospholipids are required for the membrane binding of CD_260-290.

**Fig. 1: Acidic phospholipids are required for recruitment of PD-L-CD into model and native membranes.**

We further used solution NMR spectroscopy to study the PD-L1-CD-membrane interaction as previously reported²⁴. Isotope-labeled CD_260-290 was reconstituted into the zwitterionic DMPC/1,2-dihexanoyl-sn-glycero-3-phosphocholine (DH⁶PC) bicelles (molar ratio q = 0.8) containing DMPG at different molar ratios (Supplementary Fig. 2a). Two-dimensional (2D) ¹H-¹⁵N transverse relaxation optimized spectroscopy (TROSY) experiments were used to monitor the chemical shift and intensity perturbations. A continuous fashion during the titration was observed on the N-terminal residues of CD_260-290 (Supplementary Fig. 2a-c), i.e., a single resonance peak for each perturbed resonance of CD_260-290 at any titration point, indicating that CD_260-290 is in fast exchange on the chemical shift time scale between membrane-bound and unbound forms. We noticed that CD_260-290 in DMPC/DH⁶PC bicelles (molar ratio q = 0.8) did not have detectable chemical shift differences from the protein in solution (Fig. 1b and Supplementary Fig. 2d), while CD_260-290 in acidic DMPG/DH⁶PC bicelles (q = 0.8) displayed significant chemical shift and peak intensity changes in most N-terminal resonance peaks (Fig. 1b and Supplementary Fig. 2d), suggesting that CD_260-290 specifically binds to acidic phospholipids in the bicelles. Moreover, addition of CD_260-290 to 1-palmitoyl-2-oleoyl-sn-glycero-3-(phospho-rac-(1-glycerol)) (POPG) /DH⁶PC bicelles (q = 0.8) induced chemical shift changes similar to those induced by DMPG/DH⁶PC bicelles (Supplementary Fig. 2e).

To test the effect of acidic lipids in live cells, we used fluorescence resonance energy transfer (FRET) to monitor the membrane binding of PD-L1-CD following an established protocol^24,26, in which the monomeric teal fluorescent protein (mTFP) was the donor and the octadecyl rhodamine B (R18) was the acceptor (Fig. 1c). The mTFP was fused to the C-terminus of full length PD-L1 (PD-L1-mTFP). The PD-L1-mTFP fusion protein was stably expressed in RKO cells whose plasma membranes were labeled with the R18 dye. The FRET efficiency between PD-L1-mTFP and the R18 dyes in the membranes was first measured using the donor dequenching method. The result showed that after a treatment of the cells with PS, which increases the content of acidic lipids in the plasma membrane²⁷, a higher FRET efficiency was detected compared to the solvent-treated cells, indicating stronger interactions between PD-L1-CD and the membrane with more acidic phospholipids (Fig. 1d). This result was further confirmed by the FRET efficiency determined using the donor quenching approach (Supplementary Fig. 3). In addition, a lower FRET efficiency was obtained when the RKO cells were treated with sphingosine (Sph, a membrane-permeant base)²⁷, as expected for the reduced negativity of the plasma membrane by Sph (Fig. 1d).

The effects of PS and Sph on the endogenous PD-L1 level on the RKO cell surface were further examined by confocal imaging (Fig. 1e) and flow cytometry (Fig. 1f). The results showed that the addition of PS to the culture medium significantly upregulated, whereas the addition of Sph significantly downregulated, the cell-surface expression of PD-L1. Consistently, western blot analysis showed that the PD-L1 level increased by the PS treatment but decreased by the Sph treatment (Fig. 1g). In accordance, the ubiquitination of endogenous PD-L1 that was substantially decreased after the PS treatment increased after the Sph treatment (Fig. 1g). Together, the live-cell results show that changing the net charge of the plasma membrane specifically influences the interaction of PD-L1-CD with the membrane as expected from the NMR and biochemical data. And, the membrane interaction in turn controls the cellular abundance PD-L1.

Membrane insertion of PD-L1-CD

We next studied how CD_260-290 interacts with the membrane using solution NMR spectroscopy. As mentioned above, a series of NMR spectra at different ratios of DMPG/DMPC showed that the chemical shifts of the N-terminal resonances moved continuously in straight lines (Supplementary Fig. 2a), indicating that CD_260-290 exists a two-state shift from membrane unbound to bound. To trap CD_260-290 in a membrane-bound state, we chose 100% DMPG/DH⁶PC or 100% POPG/DH⁶PC for the further investigation on protein–membrane interaction. The CD_260-290 protein was first reconstituted in DMPG/DH⁶PC bicelles (q = 0.8). Essentially complete backbone and side-chain resonance assignments were accomplished through a standard set of triple resonance experiments (Fig. 2a). The intramolecular NOE restraints showed that the cytoplasmic domain are largely dynamic with no NOEs characteristic of secondary structures.

**Fig. 2: The NMR characterization of membrane-bound CD_260-290.**

The membrane-bound protein was then characterized using the protein–lipid intermolecular NOEs measured by the protein–lipid NOESY experiments (τ_NOE = 200 ms) (Fig. 2b), in which CD_260-290 was completely deuterated to eliminate any signals from the backbone amides to side chains and reconstituted in the bicelles containing the DH⁶PC with deuterated acyl chains. 100% POPG/DH⁶PC bicelles were used in this experiment to keep the lipid acyl chains with the same length as the native membranes. Three arginine residues (R260/R262/R265) showed strong NOEs (distance of < 5 Å) to lipid acyl chains and headgroups (Fig. 2b, c), suggesting that these residues adopt an orientation toward the interior of the lipid bilayer. Such an orientation can facilitate the ionic interactions between the basic residue side chains and the phosphate groups in the lipid headgroups. V269 showed strong NOE signal to the methyl but weak signal to the methylene of lipid acyl chains, suggesting it is deeply inserted into the lipid bilayer (Fig. 2b).

In addition to NOE-based determination of CD_260-290 membrane insertion, we performed paramagnetic relaxation enhancement analysis for qualitative validation of the membrane-embedded CD_260–₂₉₀ conformation. Two paramagnetic relaxation enhancement (PRE) probes were used to measure residue-specific depth immersion of the protein in the lipid bilayer region of the bicelles, as previously described²⁸ (Supplementary Fig. 4a). The residue-specific PRE amplitudes (PRE_amp) from the water-soluble paramagnetic probe Gd-DOTA indicate that the N-terminal half of CD_260–₂₉₀ is mostly inserted into the lipid bilayer with residues 265–270 displaying the lowest PRE_amp inferring their membrane partition, while the C-terminal half is exposed to water with high PRE_amp (Supplementary Fig. 4b). The results from the lipophilic PRE probe 16-DSA (Supplementary Fig. 4c) showed a clear cutoff of PRE_amp between the membrane-embedded residues 260–274 and the water exposed residues 275–290 (Supplementary Fig. 4d). In short, the PRE data from the paramagnetic probe titrations are consistent with the protein–lipid NOE data, strengthen a model for partially membrane-embedded CD_260-290.

Key arginine residues in PD-L1-CD are critical to the membrane interaction

In the membrane-bound CD_260-290, several basic residues in the N-terminal half of CD_260-290, including R260, R262, K263, R265, K270, and K271, may interact with the lipid headgroups. Sequence analysis of PD-L1-CD shows that this basic region in the juxtamembrane segment is evolutionarily conserved between species, though the exact positions of the basic residues are not well conserved (Supplementary Fig. 5a, b).

The 3D NOESY spectra showed that R260, R262 and R265 have the strongest NOE signals to lipid headgroups and acyl chains, whereas K270 and K271 have weak NOE signals to lipids (Fig. 2b). We made R260E/R262E/R265E (CD_260-290^3RE) and K270E/K271E (CD_260-290^2KE) mutants to determine the contribution of these positively charged residues to the membrane interaction. The 2D TROSY spectra showed that the addition of CD_260-290^3RE mutant to the DMPG/DH⁶PC bicelles did not induce the resonance shifts that was induced by the wild type (WT) protein (CD_260-290^WT) (Figs. 1b and 3a). In contrast, obvious chemical shift perturbations were observed in the spectra of CD_260-290^2KE mutant upon the addition of the bicelles (Supplementary Fig. 5c). These spectral data indicate that the CD_260-290^3RE mutant loses the capacity to bind the membrane even it contains the acidic lipids, but the CD_260-290^2KE mutant retains the membrane binding capacity. This conclusion was further supported by the liposome binding assay, in which only CD_260-290^2KE mutant was detected in the DMPG liposome pellet after both mutants were incubated with the liposomes (Fig. 3b and Supplementary Fig. 5d). As a negative control, the R260K/R262K/R265K (CD_260-290^3RK) mutant interact with the lipid bicelles similar to the WT protein as detected by the NMR spectroscopy (Supplementary Fig. 5e) and the liposome binding assay (Supplementary Fig. 5f), implying that the positive charges at these positions instead of the specific residues dictate the ionic protein–lipid interaction.

**Fig. 3: Polybasic residues mediate the membrane interaction of PD-L1-CD in vitro and in cells.**

To validate the role of the positively charged residues in the membrane binding we defined in vitro, we expressed wildtype CD_260-290 with a C-terminal GFP tag (CD_260-290^WT-GFP) in HEK293T cells, and determined the subcellular localization by confocal imaging. CD_260-290^WT-GFP is mainly localized on the plasma membrane (Fig. 3c), while CD_260-290^3RE-GFP mutant is dispersed in the cytoplasm (Fig. 3d). These results were further confirmed by the FRET measurement with the mTFP-tagged full-length PD-L1^3RE mutant (PD-L1^3RE-mTFP) expressed in the RKO cells showing a lower FRET efficiency to the R18 dyes in the plasma membrane than the full-length wildtype PD-L1 (PD-L1^WT-mTFP) in the absence of PS (Fig. 3e), while the presence of PS didn’t affect the FRET efficiency on PD-L1^3RE (Fig. 3e). These mutagenesis data demonstrate that PD-L1-CD binds to the negatively charged membrane via the ionic interaction of the three positively charged juxtamembrane arginine residues, replacement of which with the negatively charged glutamic acid residues can potently disrupt the membrane binding.

PD-L1-CD-membrane interaction regulates PD-L1 stability

As we mentioned above, PD-L1-CD regulates PD-L1 stability and degradation^14,15, and stabilizing the PD-L1-CD-membrane association by palmitoylation¹⁶ suppresses the degradation of PD-L1. To study whether the membrane interaction of PD-L1-CD facilitated by acidic lipids plays a role in the stability of PD-L1, we examined the level of full-length PD-L1^WT and PD-L1^3RE mutant expressed in HEK293T cells. The cycloheximide (CHX)-chase assay²⁹ revealed accelerated reduction of PD-L1^3RE level than PD-L1^WT level (Fig. 4a-c). Cell-surface expression of PD-L1 monitored by flow cytometry showed a more substantial reduction of PD-L1^3RE mutant compared to the wild-type protein (Fig. 4d). MG132 (carbobenzoxyl-L-leucyl-L-leucyl-L-leucine), an inhibitor that can effectively block the proteosome activity³⁰, was introduced to the HEK293T cells expressing PD-L1^WT or PD-L1^3RE mutant to inhibit their degradation. Western blot analysis indicated that MG132 restored the level of both proteins in HEK293T cells with a more significant increase observed for PD-L1^3RE mutant (Fig. 4e).

**Fig. 4: The 3RE mutation reduces cellular PD-L1 protein levels.**

To demonstrate that the proteasome mediated PD-L1 degradation involves ubiquitination, we determined the level of the cellular ubiquitination of PD-L1^3RE mutant relative to PD-L1^WT. The HEK293T cells expressing PD-L1^WT or PD-L1^3RE were first treated with MG132, then subjected to PD-L1 immunoprecipitation (IP), followed by immunoblotting analysis with anti-V5 antibody to detect the ubiquitination of full length PD-L1^WT and PD-L1^3RE mutant. We found more ubiquitination for PD-L1^3RE mutant in the presence of MG132 compared to the wild-type protein (Fig. 4f), indicating that the mutant undergoes more degradation by the proteasome dependent on the ubiquitination. We further performed the CHX-chase assay and the ubiquitination examination on PD-L1^WT and PD-L1^3RE in RKO cells, showing consistent results that PD-L1^3RE degraded faster than PD-L1^WT with higher ubiquitination levels (Supplementary Fig. 6). Collectively, these observations suggest that the impaired membrane-binding capability of PD-L1^3RE mutant promotes its ubiquitination-dependent degradation. Thus, the membrane association of PD-L1-CD enhances PD-L1 protein stability.

Metformin promotes PD-L1 degradation by PD-L1-CD-membrane dissociation

The critical role of PD-L1-CD in PD-L1 stability and degradation prompted us to search for therapeutics that can alter the membrane association of PD-L1-CD and hence control the PD-L1 level. Metformin has been reported to possess antitumor effects against various cancer types in recent years^31,32,33, though it has been a popular drug for the treatment of type 2 diabetes for long time³⁴. However, how metformin acts on cancer cells remains controversial due to the use of supra-pharmacological concentrations in almost all previous studies³⁵. Considering both arginine and metformin contain the same positively charged guanidinium group (Supplementary Fig. 7a), it is possible that metformin may compete with the arginine residues in PD-L1-CD for the lipid headgroups and thereby disrupt the membrane binding of PD-L1-CD, which allows the access to the enzymes that can ubiquitinate and degrade PD-L1.

To test this hypothesis, we first titrated metformin to CD_260-290 in the presence of lipid bicelles and monitored the changes of protein–membrane interaction using NMR spectroscopy. As the concentration of metformin increases, the resonances of CD_260-290 move toward the bicelle-free form in both ¹⁵N and ¹H dimensions, indicating that the conformation of CD_260-290 shifts toward a non-membrane-bound state (Fig. 5a, b). We further performed PRE experiments to verify the role of metformin on the membrane association of CD_260-290. Titration of metformin to CD_260-290 in the presence of the DMPG/DH⁶PC bicelles containing the lipophilic paramagnetic probe 16-DSA increased the recovery of resonance intensities of the membrane-bound residues (Fig. 5c), suggesting that metformin moves these residues away from the membrane and thus shielded this region from the paramagnetic probe. The liposome binding assay confirmed the negative effect of metformin on the membrane association of PD-L1-CD since increasing the concentration of metformin reduced the binding of PD-L1-CD to the liposomes (Supplementary Fig. 7b). Moreover, the {¹H}-¹⁵N HetNOE values of the N-terminal 260-275 residues of CD_260-290 from the samples with bicelles are more positive and less dynamic than those from the samples without bicelles, while in the presence of metformin, the N-terminal half of CD_260-290 in bicelles gave low steady-state {¹H}-¹⁵N HetNOE values as the samples in solution, indicating CD_260-290 is released from the membrane by metformin (Supplementary Fig. 7c).

**Fig. 5: Metformin prevents PD-L1-CD-membrane association to induce PD-L1 degradation.**

Next, we tested the effects of metformin on PD-L1 in cells. In the absence and presence of metformin, the messenger RNA level for endogenous PD-L1 in RKO cells was similar (Supplementary Fig. 7d) but the protein level was reduced in the presence of metformin (Supplementary Fig. 7e). Moreover, the ubiquitinated protein level was higher when metformin was present (Supplementary Fig. 7f), which are consistent with previous studies^31,36. Importantly, the FRET imaging showed a significantly lower FRET efficiency after the metformin treatment of the RKO tumor cells expressing PD-L1^WT-mTFP (Fig. 5d). This direct observation of the membrane-binding status of PD-L1 via the C-terminal fluorescent tag indicates that metformin can induce the dissociation of PD-L1-CD from the plasma membrane.

Finally, we treated HEK293T cells expressing PD-L1^WT or PD-L1^3RE mutant with metformin and examined the PD-L1 protein level by western blot (Fig. 5e) and flow cytometry (Fig. 5f). The PD-L1^3RE mutant showed similar protein levels with or without the metformin treatment, indicating that PD-L1^3RE mutant that is less associated with the plasma membrane is less affected by metformin than PD-L1^WT that is more associated with the membrane. The effects of a structural analog of metformin, moroxydine (Supplementary Fig. 8a), on the membrane association of the cytoplasmic domain in vitro and the PD-L1 protein level in cells were also examined by NMR spectroscopy and western blot, respectively. The same results were observed with this metformin analog as the moroxydine titration shifted the resonances of PD-L1-CD in lipid bicelles toward that of the protein in solution (Supplementary Fig. 8b). And, the level of PD-L1 in RKO cells was reduced by the addition of moroxydine, though a higher amount of moroxydine than metformin was required to achieve the same effect (Supplementary Fig. 8a compared to Fig. 5e).

Discussion

Recent studies suggest that intracellular expression of PD-L1 and the redistribution to the cell membrane may cause resistance to anticancer drugs and protect cancer cells from T cell-mediated immune surveillance^12,20,37. Thus, alternative anticancer strategies are needed, and a recent focus is on the factors that control the cellular abundance of PD-L1^15,16. The involvement of the cytoplasmic domain in multiple pathways that regulate cellular level of PD-L1 makes it a potentially target for exploring alternative cancer therapeutics in combination with the current immunotherapy. A step forward is to understand how the cytoplasmic domain of PD-L1 interacts with the plasma membrane and how this interaction regulates the PD-L1 level.

In this study, we characterized PD-L1-CD in a model membrane system using NMR spectroscopy. Previous studies showed that some disordered proteins display certain structure features when bound to membranes. For example, α-synuclein is predominantly a random coil in aqueous solution but adopts a helical secondary structure when interacts with negatively charged membranes³⁸. In our case, no obvious secondary structure forms upon the binding to lipid bicelles, which is similar to the observations on CD28 and CD3ε cytoplasmic tails^24,26. We further demonstrated that the electrostatic interactions between a positively charged patch in the cytoplasmic domain and negatively charged phospholipids are critical to the membrane interaction and to the degradation of PD-L1 protein in tumor cells (Fig. 6). In the presence of acidic lipids, the three arginine residues adjacent to the TM domain project their side chains into the membrane (Fig. 6, magnified insert). This orientation facilitates the ionic interactions between the positive guanidinium groups of the arginines and the negative phosphate moieties in the lipid headgroups, when substituting the arginines by glutamates completely abolishes the membrane association (Fig. 3a). In contrast, replacement of the arginines with lysines does not reduce the membrane binding (Supplementary Fig. 5e, f), indicating that the electrostatic attraction is the dominant binding force. K263 within this basic patch may also contribute to the membrane association of PD-L1-CD, though we did not investigate its role due to its relatively weak NOE signals to lipids. In addition, the position of the polybasic residues in PD-L1-CD matters since mutations of K270 and K271 that are farther away from the TM domain do not disrupt the membrane association of PD-L1-CD (Supplementary Fig. 5c, d). Together, only the cluster of basic residues in PD-L1-CD located in the closest proximity to the TM domain is essential for the membrane association.

**Fig. 6: A model for regulation of PD-L1 degradation by the membrane association of its cytoplasmic domain.**

The disruption of PD-L1-CD association with the plasma membrane by the triple arginine mutation leads to a reduction of PD-L1 cellular level. However, it is not clear which of the multiple pathways that regulate PD-L1 level is affected by the PD-L1-CD-membrane dissociation. One possibility is that the membrane partition of the polybasic residues may cause steric hindrance and hence limit the accessibility of themselves or neighboring residues to other downstream proteins, such as SPOP or HIP1R, which promote the ubiquitination-dependent proteosomal or lysosomal PD-L1 degradation in the cells. On the other hand, palmitoylation of PD-L1 has been reported to modulate the protein stability by blocking its degradation¹⁶. PD-L1 is palmitoylated by DHHC3 acetyltransferase at C272¹⁶, seven residues away from the basic membrane-binding patch. Previous crystal structures of the DHHC enzymes showed that the substrate palmitoyl-CoA inserts into the cavity formed by the enzymes within the membrane³⁹. If C272 were located within the membrane as our lipid NOE signal indicated (Fig. 2b), it would be accessible to the palmitoyl-CoA in the DHHC3 enzyme. When PD-L1-CD is dissociated from the membrane, C272 is less accessible to the enzyme and substrate, which can significantly decrease the palmitoylation and hence the stability of PD-L1. Another possibility is that the effects of PD-L1-CD-membrane dissociation on PD-L1 degradation are a combinational outcome from multiple pathways including promoted PD-L1 degradation by increasing SPOP or HIP1R recruitment to PD-L1-CD and reducing DHHC3 palmitoylation. Further investigations are needed to examine these possibilities.

Calcium is important for cellular signaling that exerts regulatory effects on many enzymes and proteins. Previous studies showed that a calcium influx in T cells can disrupt the membrane interactions of the CD3ε^25,40 and CD28²⁴ tails by shielding the acidic lipid head groups, which releases the motifs in the cytoplasmic domains and mediates the downstream signaling. A high dietary calcium intake has been found to reduce the risk of one or more types of cancer, such as colorectal cancer and rectal cancer^41,42, whereas another study suggested that a high calcium intake may actually increase the risk of prostate cancer⁴³. Although the relationships between calcium intake and cancer risk have not been consistent and the exact mechanism about how calcium may help reduce the risk of some cancer types is unclear, we guess one of the possible mechanisms for the reduced risk by the high calcium intake is that the positive charges of calcium can neutralize the membrane charges as well as what was observed in T cell signaling^24,44 and release PD-L1-CD from the membrane, which leads to the reduction of PD-L1 levels in some cancer cells.

The FDA-approved type 2 diabetes drug metformin has been reported to exert substantial antitumor effects in several clinical studies^45,46. However, its mechanism of action is only partially understood. Zhang et al reported that metformin activates Hippo signaling pathway to regulate PD-L1 expression³⁶. Cha and colleagues recently revealed that metformin can activate the AMPK-dependent phosphorylation of PD-L1 at S195 and thus induce the abnormal PD-L1 glycosylation and degradation through an Endoplasmic-reticulum-associated protein degradation (ERAD) pathway⁴⁷. Strikingly, our study uncovered a previously unknown effect of metformin on PD-L1, shedding light on a more complex mechanism. We observed that metformin is able to disrupt the interaction between the polybasic juxtamembrane region of PD-L1-CD and the acidic cytoplasmic surface of the plasma membrane using NMR technology and FRET measurements. Consistent with the notion that metformin disrupts the PD-L1-CD-membrane interaction, we detected a minor FRET reduction from the PD-L1^3RE mutant to the membrane in the presence of metformin since its cytoplasmic domain was already released from the plasma membrane by the mutation. Moroxydine, a chemical analog of metformin, can also cause the dissociation of the cytoplasmic domain from the membrane. The PD-L1-CD-membrane disruption caused by these two compounds further reduced the PD-L1 abundance significantly. A common mechanism for the metformin-kind of molecules to inhibit the PD-L1 level is by electrostatically disrupting the membrane association of PD-L1-CD that is critical to the PD-L1 cellular levels. The above results have led to a unique strategy that the disruption of PD-L1-CD-membrane association can be explored for the cancer therapeutics. However, metformin treatment on cancer cells usually requires high doses. Many previous studies showed that only millimolar concentrations of metformin led to a significant reduction of the cellular abundance of PD-L1^31,47. And a number of clinical studies also used metformin in high doses of 1,500–2,250 mg per day to reduce the risk of cancer⁴⁸. The concentration of metformin used in our study is also high (mM range) due to the nonspecific effects of metformin. Therefore, more specific molecules targeting on PD-L1-CD-membrane interactions should be designed and screened to increase the efficacy in the treatment of cancer.

While many open questions remain regarding how PD-L1-CD-membrane interactions regulate other components and downstream signaling pathways, our study suggests a physiological role for acidic phospholipids in regulating the PD-L1 level in tumor cells. Acidic phospholipids act not only as a membrane scaffold, but also involve in the molecular regulation of tumor cell survival. The NMR studies of the membrane-bound PD-L1-CD showed how basic residues in the cytoplasmic tail interact with a membrane to regulate a receptor protein conformation in the membrane, which in turn regulates the receptor stability and hence its function in a signal transduction. We note that some cancer derived mutations found in this cytoplasmic region might enhance the membrane association to escape the immunosurveillance. For instance, D276Y and T290M can increase the residue hydrophobicity with the replacements and stabilize PD-L1 with stronger hydrophobic membrane interactions. R260C might provide a new position for the palmitoylation and therefore improve the stability. It’s not clear whether R262K enhances the membrane interactions of PD-L1, since arginine has a guanidino group but a shorter acyl chain. Further studies of these mutations in biochemical and cell biological systems are required to confirm or refute the expectations. Interestingly, a broad range of transmembrane receptors also have short cytoplasmic domains that are highly basic. For example, the isoelectric point (pI) for B7 family ligands CD80, CD86 and PD-L2 is 11.47, 9.72, and 10.30, respectively. Therefore, we expect that the electrostatic interactions between the receptors and lipids regulating the protein stability and degradation may occur in a general manner, and our study brings valuable insights into the lipid regulation of these transmembrane receptors.

Methods

Reagents and cells

Lipids and detergents (DMPC, DMPG, DMPE, DMPS, POPG, DH⁶PC, Sph and PS (Brain, Porcine)) were from Avanti Polar Lipids. Stable isotopes for NMR spectroscopy experiments were from Cambridge Isotope Laboratories. Anti-human PD-L1 antibodies (ab213524), Alexa488-conjugated anti-IgG (ab150077) and anti IgG-HRP (ab97051) were from Abcam. Anti-V5 antibodies (96025) were from Thermo and FITC anti PD-L1 antibodies (393606) were from Biolegend. MG132 (HY-13259) and metformin (HY-B0627) were from MedChem Express (MCE). The HEK293T cell line was a generous gift from Liming Sun (SIBCB). The RKO cell lines for expressing PD-L1 were generous gifts from Jie Xu (Fudan University), other cell lines used in the paper were from Cell Bank of Chinese Academy of Sciences.

Expression and purification of CD_260-290 and its mutants

The cytoplasmic domain of human PD-L1 (residues 260–290), named as CD_260-290, was fused to an N-terminal 8 × His tag, a Small Ubiquitin-like Modifier (SUMO) protein and a PreScission protease (3 C) in the pET28a vector. The construct was transformed into Escherichia coli BL21 (DE3) cells and grown at 37 °C in M9 minimal media until the culture reached an optical density at 600 nm (OD₆₀₀) of 0.6–0.8. One or more stable isotopes were supplemented to the growth media according to the NMR experimental requirements. Cells were cooled to 25 °C before the induction with 0.5 mM isopropyl β-D-thiogalatopyranoside (IPTG) at 25 °C overnight. Full deuteration of CD_260-290 was achieved with the growth in 99.9% D₂O (Sigma Aldrich) with deuterated glucose (Cambridge Isotope Laboratories).

The expressed fusion protein was extracted and purified by nickel affinity resins (Thermo Fisher) followed by 3 C cleavage at 4 °C for 14 h to remove the His and SUMO tag. The CD_260-290 protein was further purified by reverse-phase HPLC with Zorbax 300SB-C18 PrepHT column (Agilent) using an elution gradient from 10% (v/v) acetonitrile with 0.1% (v/v) trifluoroacetic acid (TFA) to 80% (v/v) acetonitrile and 0.1% (v/v) TFA. The fractions corresponding to pure CD_260-290 peptide were identified by MALDI-TOF mass spectrometry and SDS-PAGE analysis. All the mutants were expressed and purified following the same protocols.

Reconstitution of CD_260-290 into bicelles

The NMR samples without bicelles were prepared by directly dissolving CD_260-290 and variants (0.8–1 mg) in 25 mM MES pH 6.5, 10% D₂O. To reconstitute CD_260-290 and variants in bicelles, 0.8–1 mg of purified and lyophilized proteins were mixed with 10 mg POPG, DMPG (protonated or deuterated from Avanti Ploar Lipids) or DMPC and dissolved in hexafluoroisopropanol. The mixture was slowly dried to a thin film under nitrogen stream, followed by overnight lyophilization. The dried thin film was redissolved in 0.5 mL of 25 mM MES pH 6.5 buffer containing 37 mM DH⁶PC (protonated or deuterated from Avanti Polar Lipids). The POPG:DH⁶PC, DMPG:DH⁶PC or DMPC:DH⁶PC ratio was measured by 1D NMR to verify the q value. The final NMR sample contained 0.4–0.5 mM CD_260–₂₉₀ or variants, ~ 30 mM POPG, DMPG or DMPC, ~ 37.5 mM DH⁶PC, 25 mM MES pH 6.5 and 10% D₂O.

Assignment of NMR resonances in bicelles

All NMR spectra were acquired at 30 °C by Topspin on Bruker Avance III 600 MHz or 900 MHz spectrometers equipped with cryogenic probes and by VnmrJ Biopack on Agilent DD2 800 or 700 MHz spectrometer equipped with triple-resonance cold probes. Detailed parameters for the NMR experiments are given in the Supplementary Table 1. NMR data were processed using NMRPipe⁴⁹ and analyzed by Sparky⁵⁰ and XEASY⁵¹. Sequence specific assignment of backbone chemical shifts was accomplished by performing a suite of standard triple resonance experiments, including the TROSY version of HNCA, HN(CO)CA, HN(CA)CO, HNCO and HNCACB on a (¹⁵N, ¹³C, 85% ²H)-labeled sample at ¹H frequency of 600 MHz. These spectra employed a non-uniform sampling scheme in the indirect dimensions and were reconstructed using Sparse Multidimensional Iterative Lineshape-Enhanced (SMILE) algorithm interfaced with NMRPipe. Aliphatic side chain assignments relied on (H)CCH-TOCSY and H(C)CH-TOCSY spectra to measure the homo- and heteronuclear J couplings from quantitative J correlation^52,53 on a ¹⁵N/¹³C-labeled sample. 98% of the C and H resonances for all backbone and side chains have been assigned.

The backbone dynamics experiments were performed using an Agilent DD2 800 MHz spectrometer on the ¹⁵N/¹³C-labeled sample⁵⁴, in which the NOE spectrum was acquired with a 7 s recycle delay followed by a 5 s saturation and the reference spectrum was collected with no saturation and a 12 s recycle delay. The intensities of the NOE peaks were estimated by the Sparky program⁵⁰, and the NOE ratios of the two states were calculated using the peak intensities in the presence and absence of proton saturation.

NMR characterization of membrane-bound CD_260-290

Protein intramolecular distance restraints for membrane-bound CD_260-290 were derived from cross-peaks with a simultaneous ¹⁵N and ¹³C-NOESY-HSQC (τ_NOE = 120 ms) experiments^55,56 on a ¹⁵N/¹³C-labeled sample. Peak analysis of the NOESY spectra was generated by interactive peak picking with the program Sparky⁵⁰.

To position the dynamic PD-L1-CD in the lipid bilayer, the protein–lipid distance restraints were obtained from a 3D ¹⁵N-edited NOESY-TROSY-HSQC spectrum (τ_NOE = 200 ms) recorded with a ¹⁵N, ²H-labeled protein sample in bicelles made of regular POPG and deuterated DH⁶PC (q = 0.8), which allowed the measurement of exclusive protein–lipid NOEs between the peptide backbone amide protons and POPG aliphatic protons. The protein–lipid NOEs were then categorized into four groups according to the protein positions to the lipid. Three parallel planes were used to represent a leaflet of the POPG bilayer: the POPG head-group glycerol (y = −22 Å), the hydrophobic acyl chain (y = −11 Å) and the methyl tail (y = 0 Å).

Lipophilic PRE and solvent PRE analysis of membrane-bound CD_260-290

We cross-validated the NOE-derived membrane insertion using paramagnetic probe titration (PPT) method²⁸. The solvent PRE measurements were performed using a 0.4 mM ¹⁵N-labeled CD_260-290 in DMPG bicelles with q = 0.8. The water-soluble and membrane inaccessible paramagnetic agent, Gd-DOTA (Macrocyclics, Inc.) was titrated into the bicelle sample at different concentrations, including 0, 0.1, 0.2, 0.4, 0.6, 1, 1.5, 2, 3, 6, 10, 15, 20, and 30 mM. At each titration point, a 2D ¹H-¹⁵N TROSY-HSQC spectrum was recorded at a 700 MHz Agilent spectrometer. The recovery delay was set to 3 s. The residue-specific PRE is defined as the ratio of peak intensity in the presence (I) and absence (I₀) of the paramagnetic agent. Peak intensities were measured at peak local maxima using quadratic interpolation to identify peak centers. For individual peaks, Origin was used to fit I/I₀ vs. Gd-DOTA concentration to the exponential decay to derive the residue-specific PRE amplitude (PRE_amp) by the following equation, in which τ is the decay constant²⁸.

$$\frac{I}{{I}_{0}}=1-{{{{{{\rm{PRE}}}}}}}_{amp}\Big(1-{e}^{-\frac{[Gd-DOTA]}{\tau }}\Big)$$

(1)

PRE_amp is the indicator of PRE effects of the probe on the protein, as the higher PRE_amp values indicate the protein is more close to the PRE probe. The lipophilic PRE measurements were performed using a 0.4 mM ¹⁵N-labeled CD_260-290 in DMPG bicelles with q = 0.8. A stock solution of lipophilic paramagnetic agent 16-DSA (Sigma-Aldrich) was prepared at 40 mM concentration in the same NMR sample buffer to prevent changes of q value in the bicelles upon addition of the titrant. The progress of the titration was monitored by measuring a set of 2D ¹⁵N-TROSY-HSQC spectra at each of the following 16-DSA concentrations: 0, 0.1, 0.25, 0.5, 1, 2.5, 3, and 5 mM. The residue-specific PREamp was determined by fitting the peak intensity decay as a function of [16-DSA].

Liposome-binding assays

Mixed phospholipids were dissolved in chloroform with indicated compositions (10 mg). The solvent was evaporated under a stream of nitrogen to achieve a thin film. Followed by an overnight lyophilization, 500 μL extrusion buffer (20 mM MES, pH 6.5) was added to the dried lipid mixture. Liposomes were generated by extrusion of the hydrated lipids through a 0.1 μm polycarbonate filter (610005, Avanti Polar Lipids Inc.) 50 times using the Mini-Extruder device (Avanti Polar Lipids Inc.). CD_260-290^WT or mutants (50 μM) were incubated with the indicated liposomes (5 mM lipids) at room temperature for 30 min in a total volume of 200 μL, respectively. Samples were centrifuged at 4 °C for 1 h at 100, 000 g. The supernatant and pellet fractions were analyzed by SDS–PAGE followed by Coomassie blue staining.

Confocal imaging analysis

For immunofluorescence, RKO cells were seeded at approximately 50% confluence in a glass bottom dish (150680, Thermo). After removal of culture medium, the dish was washed twice with PBS. Then, cells were fixed with 4% Paraformaldehyde (158127, Sigma) for 20 min and washed three times with PBS. Cells were then permeabilized with 0.2% Triton X-100 in PBS and blocked in 5% BSA in PBS for 1 h at room temperature. FITC anti-PD-L1 antibodies were diluted in the blocking buffer (1:100) and incubated cells overnight at 4 °C. After rinsed by PBS three times, cells was treated by SlowFade Diamond Antifade reagent (S36968, Thermo) and sealed by a coverslip. Cells were observed with Zeiss LSM710 confocal microscope (Carl Zeiss) fitted with a 100× oil immersion objective. Micrographs were captured by means of confocal software (ZEN system 2012 Black Edition, Zeiss). All confocal images are representatives of three independent experiments.

CD_260-290^WT-GFP and mutants expressed in HEK293T cells were plated on coverslips in glass bottom dish in complete media. Cells were fixed with 4% paraformaldehyde for 20 min at room temperature. Cells were washed three times with PBS for each step. After treated by SlowFade Diamond Antifade reagent, cells were observed on Zeiss LSM710 confocal microscope fitted with a 100 × oil immersion objective.

Western blot and ubiquitination assays

For western blot analysis, cells were lysed in RIPA (50 mM Tris-HCl pH 7.4, 150 mM NaCl, 1% NP-40) buffer supplemented with protease inhibitors cocktail (11697498001, Merck) after PBS washing. Protein concentrations were measured by bicinchoninic acid (BCA) assay. Equal amounts of protein were resolved by SDS–PAGE and immunoblotted with antibodies (PD-L1 antibody, 1:5000; GAPDH antibody, 1:8000) overnight at 4 °C. The membrane was then washed three times with TBS-T (50 mM Tris, 1.37 mM NaCl, 2.7 mM KCl, pH 8.0; 0.1% Tween 20) buffer. The membrane was incubated with HRP-conjugated goat anti-rabbit IgG antibodies (1:10000) for 1 h at room temperature. After washed five times with TBS-T, the western blot bands were detected by using an ECL western blotting substrate (T7101A, Takara). The protein bands from western blot were quantified by ImageJ.

For ubiquitination assay, HEK293T cells were transfected with the indicated plasmids of PD-L1 or mutants and V5-ubiquitin via lentiviral transduction and stable clones were isolated by puromycin and single-cell sorting. Cells were treated with 10 μM MG132 for 6 h before harvesting. Cells were then lysed in RIPA lysis buffer and subjected to Co-IP with anti-PD-L1 antibodies and protein G-conjugated agarose beads (10003D, Thermo) followed by immunoblotting analysis with anti-V5 antibody (1:3000).

Flow cytometry for detection of cell-surface PD-L1

For the detection of cell-surface PD-L1, cells were harvested and suspended in 200 μL PBS buffer after washing twice and incubating with anti-human PD-L1 antibody (1:100) at 4 °C for 30 min. After washing twice by PBS, the cells were stained by Alexa488-conjugated secondary antibodies (goat-anti-rabbit IgG) were diluted in PBS (1:500) at 4 °C for 30 min. After washing twice by PBS, stained cells were collected by BD FACSDiva software in flow cytometry (BD, LSRFortessa). Data were analyzed by FlowJo Software. The gating strategy is shown in Supplementary Fig. 9.

Fluorescence resonance energy transfer measurements and analysis

Human PD-L1 containing a C-terminal mTFP tag (a gift from Prof. Chenqi Xu’s lab) with a GSS linker was constructed. The constructs were transfected into RKO cells via lentiviral transduction and stable clones were isolated by single-cell sorting. Cells were injected into a flow chamber set-up treated with poly-lysine. After staining cells with 1 µM octadecyl rhodamine B (R18, O246, Thermo) membrane-labeling dye for 5 min on ice, FRET efficiencies between the C-terminal mTFP (the donor) and the membrane dye R18 (the receptor) were measured by fluorescence microscopy (Zeiss LSM 710) using the dequenching method that mTFP was excited at 458-nm and visualized by detection at 470–550 nm; R18 was excited at 561-nm and visualized by detection at 570–650 nm. Fully photobleached R18 signal of whole cells by the 561-nm laser set at 100% efficiency for 60 s. The mTFP fluorescence signals were measured before and after R18 photobleaching three times and FRET efficiencies were calculated using the following formula:

$${{{{{\rm{FRET}}}}}}\,{{{{{\rm{efficiency}}}}}}=({{{{{{\rm{FL}}}}}}}_{{{{{{\rm{AP}}}}}}}-{{{{{{\rm{FL}}}}}}}_{{{{{{\rm{BP}}}}}}})/{{{{{{\rm{FL}}}}}}}_{{{{{{\rm{AP}}}}}}}$$

(2)

where FL_AP and FL_BP represent mTFP fluorescence after and before R18 photobleaching, respectively.

In the FRET detected lipid effects on PD-L1-CD-membrane interaction assay, 1 × 10⁶ RKO/PD-L1-mTFP cells were seeded in a 6-well plate and cultured in RPMI-1640 medium treated with dimethylsulfoxide (DMSO, negative control), 5 μM phosphatidylserine (PS) or 2 μM sphingosine (Sph) for 2 h, respectively. Cells were collected and fixed into a flow chamber treated with poly-lysine. After staining cells with 1 µM R18, fluorescent images of the cells were taken to detect the mTFP or R18 dye emission before or after photobleaching (BP or AP) the R18 dye. FRET efficiencies were calculated by Eq. 2.

Metformin and moroxydine titration assay

The metformin titration experiments were used to detect metformin effects on CD_260–₂₉₀ in DMPG/DH⁶PC bicelles. 0.4 mM ¹⁵N-labeled CD_260-290 was reconstituted in q = 0.8 DMPG/DH⁶PC bicelles. A stock solution of 2 M metformin was added to the NMR sample at the molar ratios of [metformin]:[DMPG] = 0, 0.05, 0.1, 0.2, 0.4, 0.6, 1.0, 2.0. A 2D ¹H-¹⁵N TROSY-HSQC spectrum was recorded at each concentration of metformin. Moroxydine was titrated to CD_260–₂₉₀ following the same protocol. To further verify the effect of metformin on CD_260-290, PRE recovery experiment was performed using 0.4 mM ¹⁵N-labeled CD_260-290 reconstituted in DMPG/DH⁶PC bicelles with 2.5 mM 16-DSA. Metformin was then titrated using the same method mentioned above at molar ratios of [metformin]:[DMPG] = 0, 0.2, 0.4, 0.6, 0.8, 1.0, 2.0.

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability

¹H, ¹³C, and ¹⁵N chemical shifts have been deposited in the Biological Magnetic Resonance Bank under accession number BMRB 36293 (Membrane-bound human PD-L1 cytoplasmic domain), related to the PDB code 6L8R. Source data are provided with this paper.

References

Francisco, L. M., Sage, P. T. & Sharpe, A. H. The PD-1 pathway in tolerance and autoimmunity. Immunological Rev. 236, 219–242 (2010).
Article CAS Google Scholar
Keir, M. E., Butte, M. J., Freeman, G. J. & Sharpel, A. H. PD-1 and its ligands in tolerance and immunity. Annu. Rev. Immunol. 26, 677–704 (2008).
Article CAS PubMed Google Scholar
Azuma, T. et al. B7-H1 is a ubiquitous antiapoptotic receptor on cancer cells. Blood 111, 3635–3643 (2008).
Article CAS PubMed PubMed Central Google Scholar
Tumeh, P. C. et al. PD-1 blockade induces responses by inhibiting adaptive immune resistance. Nature 515, 568–571 (2014).
Article ADS CAS PubMed PubMed Central Google Scholar
Zak, K. M. et al. Structural biology of the immune checkpoint receptor PD-1 and its ligands PD-L1/PD-L2. Structure 25, 1163–1174 (2017).
Article CAS PubMed Google Scholar
Mu, C. Y., Huang, J. A., Chen, Y., Chen, C. & Zhang, X. G. High expression of PD-L1 in lung cancer may contribute to poor prognosis and tumor cells immune escape through suppressing tumor infiltrating dendritic cells maturation. Med. Oncol. 28, 682–688 (2011).
Article CAS PubMed Google Scholar
Hino, R. et al. Tumor cell expression of programmed cell death-1 ligand 1 is a prognostic factor for malignant melanoma. Cancer 116, 1757–1766 (2010).
Article PubMed Google Scholar
Freeman, G. J. Structures of PD-1 with its ligands: sideways and dancing cheek to cheek. Proc. Natl Acad. Sci. USA 105, 10275–10276 (2008).
Article ADS CAS PubMed PubMed Central Google Scholar
Garon, E. B. et al. Pembrolizumab for the treatment of non-small-cell lung cancer. N. Engl. J. Med. 372, 2018–2028 (2015).
Article PubMed Google Scholar
Brahmer, J. et al. Nivolumab versus docetaxel in advanced squamous-cell non-small-cell lung cancer. N. Engl. J. Med. 373, 123–135 (2015).
Article CAS PubMed PubMed Central Google Scholar
Gong, B. et al. Secreted PD-L1 variants mediate resistance to PD-L1 blockade therapy in non-small cell lung cancer. J. Exp. Med. 216, 982–1000 (2019).
Article CAS PubMed PubMed Central Google Scholar
Yang, Y. et al. Exosomal PD-L1 harbors active defense function to suppress T cell killing of breast cancer cells and promote tumor growth. Cell Res. 28, 862–864 (2018).
Article CAS PubMed PubMed Central Google Scholar
Chen, G. et al. Exosomal PD-L1 contributes to immunosuppression and is associated with anti-PD-1 response. Nature 560, 382–386 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Zhang, J. et al. Cyclin D-CDK4 kinase destabilizes PD-L1 via cullin 3-SPOP to control cancer immune surveillance. Nature 553, 91–95 (2018).
Article ADS CAS PubMed Google Scholar
Wang, H. et al. HIP1R targets PD-L1 to lysosomal degradation to alter T cell-mediated cytotoxicity. Nat. Chem. Biol. 15, 42–50 (2019).
Article CAS PubMed Google Scholar
Yao, H. et al. Inhibiting PD-L1 palmitoylation enhances T-cell immune responses against tumours. Nat. Biomed. Eng. 3, 306–317 (2019).
Article CAS PubMed Google Scholar
Yang, Y. et al. Palmitoylation stabilizes PD-L1 to promote breast tumor growth. Cell Res. 29, 83–86 (2019).
Article PubMed Google Scholar
Gato-Canas, M. et al. PDL1 signals through conserved sequence motifs to overcome interferon-mediated cytotoxicity. Cell Rep. 20, 1818–1829 (2017).
Article CAS PubMed Google Scholar
Lefrancois, S. Tumour suppression by blocking palmitoylation. Nat. Biomed. Eng. 3, 255–256 (2019).
Article PubMed Google Scholar
Wu, Y., Chen, W., Xu, Z. P. & Gu, W. PD-L1 distribution and perspective for cancer immunotherapy-blockade, knockdown, or inhibition. Front Immunol. 10, 2022 (2019).
Article CAS PubMed PubMed Central Google Scholar
Harayama, T. & Riezman, H. Understanding the diversity of membrane lipid composition. Nat. Rev. Mol. cell Biol. 19, 281–296 (2018).
Article CAS PubMed Google Scholar
Poggio, M. et al. Suppression of exosomal PD-L1 induces systemic anti-tumor immunity and memory. Cell 177, 414–427 (2019). e413.
Article CAS PubMed PubMed Central Google Scholar
Skotland, T., Sandvig, K. & Llorente, A. Lipids in exosomes: current knowledge and the way forward. Prog. lipid Res. 66, 30–41 (2017).
Article CAS PubMed Google Scholar
Yang, W. et al. Dynamic regulation of CD28 conformation and signaling by charged lipids and ions. Nat. Struct. Mol. Biol. 24, 1081–1092 (2017).
Article CAS PubMed Google Scholar
Guo, X. et al. Lipid-dependent conformational dynamics underlie the functional versatility of T-cell receptor. Cell Res. 27, 505–525 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Xu, C. et al. Regulation of T cell receptor activation by dynamic membrane binding of the CD3epsilon cytoplasmic tyrosine-based motif. Cell 135, 702–713 (2008).
Article CAS PubMed PubMed Central Google Scholar
Yeung, T. et al. Membrane phosphatidylserine regulates surface charge and protein localization. Science 319, 210–213 (2008).
Article ADS CAS PubMed Google Scholar
Piai, A., Fu, Q., Dev, J. & Chou, J. J. Optimal bicelle size q for solution NMR studies of the protein transmembrane partition. Chemistry 23, 1361–1367 (2017).
Article CAS PubMed Google Scholar
Kao, S. H. et al. Analysis of protein stability by the cycloheximide chase assay. Bio-protocol 5, e1374 (2015).
Lim, S. O. et al. Deubiquitination and stabilization of PD-L1 by CSN5. Cancer Cell 30, 925–939 (2016).
Article CAS PubMed PubMed Central Google Scholar
Xue, J. et al. Metformin suppresses cancer cell growth in endometrial carcinoma by inhibiting PD-L1. Eur. J. Pharm. 859, 172541 (2019).
Article CAS Google Scholar
Cantrell, L. A. et al. Metformin is a potent inhibitor of endometrial cancer cell proliferation–implications for a novel treatment strategy. Gynecol. Oncol. 116, 92–98 (2010).
Article CAS PubMed Google Scholar
Dowling, R. J., Zakikhani, M., Fantus, I. G., Pollak, M. & Sonenberg, N. Metformin inhibits mammalian target of rapamycin-dependent translation initiation in breast cancer cells. Cancer Res. 67, 10804–10812 (2007).
Article CAS PubMed Google Scholar
Thomas, I. & Gregg, B. Metformin; a review of its history and future: from lilac to longevity. Pediatr. Diabetes 18, 10–16 (2017).
Article PubMed Google Scholar
He, L. & Wondisford Fredric, E. Metformin action: concentrations matter. Cell Metab. 21, 159–162 (2015).
Article CAS PubMed Google Scholar
Zhang, J. J., Zhang, Q. S., Li, Z. Q., Zhou, J. W. & Du, J. Metformin attenuates PD-L1 expression through activating Hippo signaling pathway in colorectal cancer cells. Am. J. Transl. Res. 11, 6965–6976 (2019).
PubMed PubMed Central Google Scholar
Burr, M. L. et al. CMTM6 maintains the expression of PD-L1 and regulates anti-tumour immunity. Nature 549, 101–105 (2017).
Article ADS CAS PubMed PubMed Central Google Scholar
Ulmer, T. S., Bax, A., Cole, N. B. & Nussbaum, R. L. Structure and dynamics of micelle-bound human alpha-synuclein. J. Biol. Chem. 280, 9595–9603 (2005).
Article CAS PubMed Google Scholar
Rana, M. S. et al. Fatty acyl recognition and transfer by an integral membrane S-acyltransferase. Science 359, eaao6326 (2018)
Gagnon, E., Schubert, D. A., Gordo, S., Chu, H. H. & Wucherpfennig, K. W. Local changes in lipid environment of TCR microclusters regulate membrane binding by the CD3epsilon cytoplasmic domain. J. Exp. Med. 209, 2423–2439 (2012).
Article CAS PubMed PubMed Central Google Scholar
Han, C. et al. Dietary calcium intake and the risk of colorectal cancer: a case control study. BMC cancer 15, 966 (2015).
Article PubMed PubMed Central CAS Google Scholar
Wu, K., Willett, W. C., Fuchs, C. S., Colditz, G. A. & Giovannucci, E. L. Calcium intake and risk of colon cancer in women and men. J. Natl Cancer Inst. 94, 437–446 (2002).
Article PubMed Google Scholar
Wilson, K. M., Shui, I. M., Mucci, L. A. & Giovannucci, E. Calcium and phosphorus intake and prostate cancer risk: a 24-y follow-up study. Am. J. Clin. Nutr. 101, 173–183 (2015).
Article CAS PubMed Google Scholar
Li, L. et al. Ionic CD3-Lck interaction regulates the initiation of T-cell receptor signaling. Proc. Natl Acad. Sci. USA 114, E5891–E5899 (2017).
CAS PubMed PubMed Central Google Scholar
Deng, J. et al. Novel application of metformin combined with targeted drugs on anticancer treatment. Cancer Sci. 110, 23–30 (2019).
Article CAS PubMed Google Scholar
Laskov, I. et al. Anti-diabetic doses of metformin decrease proliferation markers in tumors of patients with endometrial cancer. Gynecol. Oncol. 134, 607–614 (2014).
Article CAS PubMed Google Scholar
Cha, J. H. et al. Metformin promotes antitumor immunity via endoplasmic-reticulum-associated degradation of PD-L1. Mol. Cell 71, 606–620 (2018). e607.
Article CAS PubMed PubMed Central Google Scholar
Saraei, P., Asadi, I., Kakar, M. A. & Moradi-Kor, N. The beneficial effects of metformin on cancer prevention and therapy: a comprehensive review of recent advances. Cancer Manag. Res. 11, 3295–3313 (2019).
Article CAS PubMed PubMed Central Google Scholar
Delaglio, F. et al. NMRPipe: a multidimensional spectral processing system based on UNIX pipes. J. Biomol. NMR 6, 277–293 (1995).
Article CAS PubMed Google Scholar
Lee, W., Tonelli, M. & Markley, J. L. NMRFAM-SPARKY: enhanced software for biomolecular NMR spectroscopy. Bioinformatics 31, 1325–1327 (2015).
Article PubMed Google Scholar
Bartels, C., Xia, T. H., Billeter, M., Guntert, P. & Wuthrich, K. The program XEASY for computer-supported NMR spectral analysis of biological macromolecules. J. Biomol. NMR 6, 1–10 (1995).
Article CAS PubMed Google Scholar
Bax, A. et al. Measurement of homo- and heteronuclear J couplings from quantitative J correlation. Methods Enzymol. 239, 79–105 (1994).
Article CAS PubMed Google Scholar
Kay, L. E. NMR methods for the study of protein structure and dynamics. Biochem. cell Biol. 75, 1–15 (1997).
Article CAS PubMed Google Scholar
Farrow, N. A. et al. Backbone dynamics of a free and phosphopeptide-complexed Src homology 2 domain studied by ¹⁵N NMR relaxation. Biochemistry 33, 5984–6003 (1994).
Article CAS PubMed Google Scholar
Pascal, S. M., Muhandiram, D. R., Yamazaki, T., Formankay, J. D. & Kay, L. E. Simultaneous acquisition of ¹⁵N-edited and ¹³C-edited NOE spectra of proteins dissolved in H₂O. J. Magn. Reson B 103, 197–201 (1994).
Article CAS Google Scholar
Bellstedt, P. et al. Sequential acquisition of multi-dimensional heteronuclear chemical shift correlation spectra with (1)H detection. Sci. Rep. 4, 4490 (2014).
Article PubMed PubMed Central CAS Google Scholar

Download references

Acknowledgements

We thank the staffs from Large-scale Protein Preparation System/Nuclear Magnetic Resonance System/Mass Spectrometry System/Integrated Laser Microscopy System/Molecular Imaging System at National Facility for Protein Science in Shanghai, Zhangjiang Laboratory (NFPS, ZJLab), China for providing technical support and assistance in data collection and analysis. This work was supported by grants from National Natural Science Foundation of China (31861133009 and U1732125), National Key R&D Program of China (2017YFA0504804), Key Research Program of Frontier Sciences, CAS (QYZDB-SSW-SMC043), and CAS Major Science and Technology Infrastructure Open Research Projects to B.O. and Institutional Development Award from the National Institute of General Medical Sciences of US National Institutes of Health (P20GM103640) to J.L.

Author information

These authors contributed equally: Maorong Wen, Yunlei Cao, Bin Wu.

Authors and Affiliations

State Key Laboratory of Molecular Biology, Shanghai Institute of Biochemistry and Cell Biology, Center for Excellence in Molecular Cell Science, Chinese Academy of Sciences, Shanghai, China
Maorong Wen, Yunlei Cao, Taoran Xiao, Ruiyu Cao, Qian Wang, Xiwei Liu, Chenqi Xu & Bo OuYang
University of Chinese Academy of Sciences, Beijing, China
Yunlei Cao, Taoran Xiao, Ruiyu Cao, Qian Wang, Xiwei Liu, Chenqi Xu & Bo OuYang
National Facility for Protein Science in Shanghai, ZhangJiang lab, Shanghai Advanced Research Institute, Chinese Academy of Sciences, Shanghai, China
Bin Wu, Hongjuan Xue & Yang Yu
Department of Biochemistry and Molecular Biology, University of Oklahoma Health Sciences Center, Oklahoma City, OK, USA
Jialing Lin
Stephenson Cancer Center, Oklahoma City, OK, USA
Jialing Lin
Institutes of Biomedical Sciences, Fudan University, Shanghai, China
Jie Xu

Authors

Maorong Wen
View author publications
You can also search for this author in PubMed Google Scholar
Yunlei Cao
View author publications
You can also search for this author in PubMed Google Scholar
Bin Wu
View author publications
You can also search for this author in PubMed Google Scholar
Taoran Xiao
View author publications
You can also search for this author in PubMed Google Scholar
Ruiyu Cao
View author publications
You can also search for this author in PubMed Google Scholar
Qian Wang
View author publications
You can also search for this author in PubMed Google Scholar
Xiwei Liu
View author publications
You can also search for this author in PubMed Google Scholar
Hongjuan Xue
View author publications
You can also search for this author in PubMed Google Scholar
Yang Yu
View author publications
You can also search for this author in PubMed Google Scholar
Jialing Lin
View author publications
You can also search for this author in PubMed Google Scholar
Chenqi Xu
View author publications
You can also search for this author in PubMed Google Scholar
Jie Xu
View author publications
You can also search for this author in PubMed Google Scholar
Bo OuYang
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

B.O. and M.W. conceived of the study; Y. C. and Q. W. prepared NMR samples; M.W. and Y.C. performed NMR experiments and data collection with the help of H.X.; M.W. and B.W. analyzed the NMR data; M.W., Y.C., T.X., and R.C. performed the biochemical and imaging experiments with the help of X.L. and Y.Y.; J.L., J.X. and C.X. provided extensive discussions; B.O. and M.W. wrote the paper, and all authors, especially J.L., contributed to editing of the manuscript.

Corresponding authors

Correspondence to Maorong Wen or Bo OuYang.

Ethics declarations

Competing interests

The authors declare no competing interests.

Additional information

Peer review information Nature Communications thanks Michiyuki Matsuda and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Peer Review File

Reporting Summary

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Wen, M., Cao, Y., Wu, B. et al. PD-L1 degradation is regulated by electrostatic membrane association of its cytoplasmic domain. Nat Commun 12, 5106 (2021). https://doi.org/10.1038/s41467-021-25416-7

Download citation

Received: 17 September 2020
Accepted: 06 August 2021
Published: 24 August 2021
DOI: https://doi.org/10.1038/s41467-021-25416-7

This article is cited by

Metformin and cancer hallmarks: shedding new lights on therapeutic repurposing
- Yu Hua
- Yue Zheng
- Ai Zhuang
Journal of Translational Medicine (2023)
Interface synergism and engineering of Pd/Co@N-C for direct ethanol fuel cells
- Jinfa Chang
- Guanzhi Wang
- Yang Yang
Nature Communications (2023)
Harnessing epithelial-mesenchymal plasticity to boost cancer immunotherapy
- Yuanzhuo Gu
- Zhengkui Zhang
- Peter ten Dijke
Cellular & Molecular Immunology (2023)
Molecular mechanisms of action of metformin: latest advances and therapeutic implications
- Hong Zhu
- Zhenquan Jia
- Igor Danelisen
Clinical and Experimental Medicine (2023)
Noninvasive evaluation of PD-L1 expression in non-small cell lung cancer by immunoPET imaging using an acylating agent–modified antibody fragment
- Yuan Cheng
- Dai Shi
- Dengfeng Cheng
European Journal of Nuclear Medicine and Molecular Imaging (2023)

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.