A comparability study of natural and deglycosylated PD-L1 levels in lung cancer: evidence from immunohistochemical analysis

The detection of PD-L1 expression status using IHC is the most direct and practicable route for stratification to guide anti-PD-1/PD-L1 therapy [8]. In the current study, we first compared natural PD-L1 expression in LuCa using a panel of antibodies obtained from Abcam, including 28–8, CAL10, 73–10 and SP142. To obtain the best staining effect, we performed IHC at an assay-dependent concentration (Additional file 2: Table S1). The clinicopathological features of LuCa patients represented in the HLugC120PT01 and the array distribution of HLugC120PT01 sections are included in Additional file 3: Table S2 and Additional file 4: Fig. S1. Two paratumor samples exhibited exfoliation of cells, and 3 samples were remarkably infiltrated with tumor cells, which were excluded from this analysis. The representative images exhibited a characteristic PD-L1 staining pattern, which was typified by immunoreactivity mostly in the cytomembrane; besides, the cytoplasm was also partially stained (Fig. 1a). We next compared the PD-L1 expression status in tumor and paratumor tissues. The percentage of PD-L1-positive cells and Histoscore (H-score) of PD-L1 signal intensity in LuCa tissues detected by these 4 mAbs were significantly higher than those in paratumor tissues (Additional file 5: Table S3 and Fig. 1b and c).

To evaluate the agreement among the 4 PD-L1 mAbs, a Venn diagram was generated to compare the percentage of PD-L1-positive tumor cells (tumor proportion score, TPS) in the cohort. Of the 60 cases, five showed concordance below all thresholds of 5% across all mAb combinations, while 17 cases showed expression above all thresholds of 5%. The remaining 38 cases showed a combination of discordant outcomes across the various mAb combinations (Fig. 1d). In addition, we found that the CAL10 mAb had the lowest detection rate (20/60), while the other 3 mAbs had similar detection rates (28–8: 45/60, 73–10: 45/60, SP142: 37/60) in LuCa (Fig. 1d). Using these 4 mAbs, we further analyzed the consistency of PD-L1 with respect to the H-score, and the results revealed high consistency in the detection of PD-L1 expression (Fig. 1e and f). Overall, these 4 PD-L1 mAbs had a similar detection performance in LuCa.

Lee et al. demonstrated a novel strategy for the detection of PD-L1 expression. Briefly, antigen retrieval by protein deglycosylation could enhance the PD-L1 (28–8 mAb) signal intensity [6]. Due to an increasing number of diagnostic PD-L1 antibodies that have become commercially available, whether the deglycosylation strategy is suitable for antibodies with various antigen specificities is worthy of further exploration. Immunofluorescence staining is an effective method that can detect PD-L1 expression with or without deglycosylation [6]. The fluorescence intensity of PD-L1 determined by three antibodies (28–8, CAL10 and SP142) was significantly enhanced after PNGase F treatment in NCI-H1299 compared with no treatment; however, slight inhibition in the fluorescence intensity of PD-L1 was found after PNGase F treatment with the 73–10 antibody (Additional file 6: Fig. S2).

Next, we evaluated the response of different PD-L1 antibodies to deglycosylation in LuCa tissues. The results showed that removal of N-linked glycosylation markedly enhanced PD-L1 detection when the 28–8, CAL10 and SP142 mAbs were used but slightly inhibited PD-L1 detection when the 73–10 mAb was used (Fig. 2a-d and Additional file 7: Fig. S3). When we compared PD-L1 detection after deglycosylation using these 4 mAbs, the PD-L1 staining level obtained with the 28–8 mAb was the highest, while the level obtained with SP142 mAb was close behind (Fig. 2e). In addition, after deglycosylation, PD-L1 expression was not limited to the tumor cell membrane but could also be detected in the cytoplasm (Fig. 2a). Several studies have reported that PD-L1 can also be expressed in the tumor cell cytoplasm [9, 10]. According to the report by Lee et al., PD-L1 detection was significantly enhanced in both the cell membrane and cytoplasm, which was consistent with our results [6].

According to specifications provided by Abcam, the peptides used for the production of the CAL10, SP142 and 73–10 mAbs are generated against the intracellular domain (the exact sequence is not available), and the peptide used for production of the 28–8 mAb is generated against the extracellular domain (Phe19-Thr239). Four glycosylation sites have been previously reported, including N35, N192, N200 and N219, which are all located in the extracellular domain of PD-L1 [5] (Fig. 2f). Thus, the PD-L1 signal intensity could be significantly enhanced when using the 28–8 mAb since the PD-L1 antigenic region is more accessible to antibody binding. However, although the exact glycosylation sites in the intracellular domain were not available, the PD-L1 signal intensity could also be enhanced to some extent when the CAL10 and SP142 mAbs were used. We speculated that the adjacent glycosylation sites, such as N200 and/or N219, reduced the binding affinity. Regardless, our current results showed that sample deglycosylation can not only enhance the 28–8 mAb signal but also improve the detection rate of the CAL10 and SP142 mAbs.

Furthermore, to test whether the deglycosylated PD-L1 level would better predict the response to immunotherapy, 12 LuCa patients who received anti-PD-1 therapy (camrelizumab plus chemotherapy) were recruited (Additional file 8: Table S4). One representative patient who derived a substantial benefit from camrelizumab therapy is shown in Fig. 3a. Besides, the relative change in sum of diameters was exhibited in Fig. 3b. Given the better predictive value of deglycosylated PD-L1 determined by the 28–8 mAb in immunotherapy uncovered by Lee et al. [6], we did not include it in current research. At first, we validated that tissue deglycosylation could enhance the detection rate of the CAL10 and SP142 mAbs in our recruited cohort (Fig. 3c and d). Encouragingly, the relative change in sum of diameters exhibited a stronger correlation with deglycosylated PD-L1 than with natural PD-L1 levels determined by the CAL10 and SP142 mAbs, which suggested that deglycosylated PD-L1 expression might be a better biomarker for predicting efficacy to at least camrelizumab-based therapies but potentially other anti-PD-1 therapies as well (Fig. 3e and f). However, the clinical significance of deglycosylated PD-L1 should be further explored in large-scale clinical trials. Studies should include at least the following three critical points: 1. standard technical protocol for deglycosylated PD-L1 detection, 2. cutoff values of deglycosylated PD-L1 levels for use in clinical practice, and 3. the congruent relationship between different detection antibodies and therapeutic antibodies.