Removal of N-Linked Glycosylation Enhances PD-L1 Detection and Predicts Anti-PD-1/PD-L1 Therapeutic Efficacy

Abstract

Reactivation of T cell immunity by PD-1/PD-L1 immune checkpoint blockade has been shown to be a promising cancer therapeutic strategy. However, PD-L1 immunohistochemical readout is inconsistent with patient response, which presents a clinical challenge to stratify patients. Because PD-L1 is heavily glycosylated, we developed a method to resolve this by removing the glycan moieties from cell surface antigens via enzymatic digestion, a process termed sample deglycosylation. Notably, deglycosylation significantly improves anti-PD-L1 antibody binding affinity and signal intensity, resulting in more accurate PD-L1 quantification and prediction of clinical outcome. This proposed method of PD-L1 antigen retrieval may provide a practical and timely approach to reduce false-negative patient stratification for guiding anti-PD-1/PD-L1 therapy.

Keywords: PD-1; PD-L1; antibody-based detection; biomarker; glycosylation/deglycosylation; heterogeneity/homogeneity; immune checkpoint; immunohistochemistry; immunotherapy.

Figure 1.. Removal of N-linked glycosylation enhances anti-PD-L1 signal in human cancer cells in a variety of bioassays.

(A and B) Immunofluorescence confocal microscopy of BT-549 (A) and A549 (B) cells processed with or without deglycosylation by PNGase F (5%) pretreatment stained with DAPI and an anti-PD-L1 antibody (Abcam, ab58810). Bar, 10 μm. Quantification is shown to the right. Data are representative of 3 independent experiments, randomly chosen in 3 different fields. (C and D) ELISA of Con A (C) and PD-L1 (clone 28-8 mAb) (D) levels in BT-549 cells processed with deglycosylation by increasing concentrations of PNGase F (1, 2, 5%) pretreatment for comparison with cells without deglycosylation (PNGase F; 0%). The intensity of Con A and PD-L1 was normalized to that without PNGase F pretreatment and set to 1. (E) ELISA of PD-L1 levels (clone 28-8 mAb) in lung cancer cells processed with deglycosylation by PNGase F (1%) pretreatment for comparison with cells without deglycosylation (0%). Negative control, secondary Ab only control. (F) Left: saturation binding assay of A549 cell lysates binding to anti-PD-L1 clone 28-8 mAb. Right: scatchard plot of cell number binding to anti-PD-L1 antibody transformed from the left. (G) Representative images (left) and quantification (right) of H-score of IHC staining for BLBC (BT-549, BT-20, and MDA-MB-231) and non-BLBC (MCF-7) cancer cell blocks processed with or without deglycosylation by PNGase F (5%) pretreatment. Bar, 50 μm. (H) Representative images (top) and quantification (bottom) of H-score of IHC staining for a panel of lung cancer cell blocks processed with or without deglycosylation by PNGase F (5%) pretreatment. Bar, 50 μm. (A–F) Results are presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, Student’s t test. See also Figure S1.

Figure 2.. Deglycosylation significantly enhances anti-PD-L1 signal in a major population of patient samples from a human tumor tissue microarray.

(A) H-score values representing PD-L1 protein expression from IHC staining of a human multi-organcarcinoma tissue microarray (TMA, n = 200) processed with or without deglycosylation by PNGase F (5%) pretreatment. Results were analyzed by the Wilcoxon signed-rank test. (B) A pie chart highlighting the fold change of H-score after N-linked glycosylation removal through PNGase F treatment from (A). (C) Two representative cases of IHC staining from (A). Bar, 50 μm. (D) H-score values representing PD-L1 protein expression from IHC staining of a human lung cancer TMA (n = 149) processed with or without deglycosylation by PNGase F (5%) pretreatment. Results were analyzed by the Wilcoxon signed-rank test. (E) A pie chart highlighting the fold change of H-score after N-linked glycosylation removal through PNGase F treatment from (D). (F) Two representative cases of IHC staining from (D). Bar, 50 μm. (G) The average population of three individual cohorts of lung cancer patients (total n = 233) expressing PD-L1 positive cells (PD-L1 TPS; %) from the indicated cutoffs without and with deglycosylation (deglyco.). Results are presented as mean ± SD. See also Figure S2.

Figure 3.. Improved PD-L1 detection after deglycosylation is associated with response to anti-PD-1/PD-L1 therapy.

(A) H-score values representing PD-L1 protein expression from IHC staining of the archived FFPE tumor tissue blocks before treatments from patients with different types of cancer who received or are undergoing anti-PD-1/PD-L1 immunotherapy (n = 95) processed with or without deglycosylation by PNGase F (5%) pretreatment. Results were analyzed by the Wilcoxon signed-rank test. (B) A pie chart highlighting the fold change in H-score after N-linked glycosylation removal through PNGase F treatment from (A). (C) Representative cases of IHC staining from (A). Bar, 50 μm. (D) Pearson correlation test between H-score representing PD-L1 protein expression in patient tissue slides processed with or without deglycosylation and the corresponding PFS from anti-PD-1/PD-L1 therapy from (A). (E) Pearson correlation test between the percentage of PD-L1 positive cells (TPS) in patient tissue slides processed with or without deglycosylation and the corresponding PFS from anti-PD-1/PD-L1 therapy from (A). (F and G) Pearson correlation test between PD-L1 H-score (F) or PD-L1 TPS (G) in patient tissue slides processed with or without deglycosylation and the corresponding OS from anti-PD-1/PD-L1 therapy from (A) (n = 49 with the OS available). See also Figures S3 and S4.

Figure 4.. Increased PD-L1 signal after deglycosylation is beneficial to therapeutic selection.

(A and B) Pearson correlation test between PD-L1 H-score (A) or PD-L1 TPS (B) in lung cancer patient tissue slides (n = 44) processed with or without deglycosylation from Figure 3A and the corresponding PFS from anti-PD-1/PD-L1 therapy. (C) The PFS of lung cancer patients expressing PD-L1 TPS in the indicated cutoffs without or with deglycosylation. n = 12 for group (1), n = 7 for group (2), n = 25 for group (3). (D) The PFS of lung cancer patients expressing PD-L1 TPS from < 1% in the indicated cutoffs without or with deglycosylation. n = 10 for group (4), n = 5 for group (5), n = 3 for group (6). (C and D) Results are presented as mean ± SD. *p < 0.05, **p < 0.01, ***p < 0.001, NS, not significant, Student’s t test.

Figure 5.. Deglycosylation improves predictive ability of PD-L1 as a biomarker for immunotherapy.

(A and B) The PFS of cancer patient samples processed without (A) or with (B) deglycosylation by PNGase F (5%) pretreatment. Cases with H-score equal to or higher than the median value of total 95 cases (H-score = 57.5) were considered as high expression and those with H-score less than the median value as low expression. (C and D) The PFS of cancer patient samples processed without (C) or with (D) deglycosylation by PNGase F (5%) pretreatment. Cases with H-score equal to or higher than the median value of individual group [H-score = 40 in the group of without glycosylation (C) and H-score = 90 in the group of with glycosylation (D), respectively] were considered as high expression and those with H-score less than the respective median value as low expression. (E and F) Representative images of computed tomography (CT) scan and chest X-ray from case 6 (E) and case 11 (F) from Figure 3A, pre- and post-anti-PD-1 (nivolumab) immunotherapy. (G) A proposed model of PD-L1 antigen retrieval through sample deglycosylation. In brief, the glycan structure of PD-L1 hinders antibody-based detection targeting the PD-L1 antigen. Sample deglycosylation more accurately assesses PD-L1 expression to allow better estimation of PD-L1 levels to prevent false-negative readouts in clinical settings. (A–D) Cohort size for each group is indicated. p values were determined by Log-rank (Mantel-Cox) test. Hazard Ratio (HR) and 95% confidence interval (CI) were determined by Mantel-Haenszel method. See also Figure S5.