PD-L1 expression and TMB landscape across tumour types
In total, 6,668 patients with advanced tumours representing 25 tumour types were enrolled in the study, and samples with paired PD-L1 expression and TMB value were obtained during the course of standard clinical care. Summary PD-L1 qualitative IHC data are shown in Fig. 1A. PD-L1 expression varied widely among the tumour types examined. Nasopharyngeal and thymic carcinoma had the highest frequency of PD-L1 positivity (75%, 68%), whereas small bowel carcinoma had the lowest frequency of PD-L1 positivity (9%). Across all samples, 3.6% of specimens were PD-L1 high-positive (defined as ≥ 50% tumour cells staining positive for PD-L1). Among the distinct tumour types, nasopharyngeal and thymic carcinomas had the highest frequency of PD-L1 high-positive samples (51%, 46%), whereas endometrial cancer had the lowest (0%) (Additional file 1: Supplementary Fig. 2).
The median TMB across all samples was 5.08 mutations/Mb (IQR 1.99–8.89). Summary TMB data for different tumour types are shown in Fig. 1B. The median TMB for each tumour type ranged from 1.27 mutations/Mb in GIST cancer to 34.36 in colorectal (dMMR) cancer. dMMR was detected in 23 tumour types, and tumours with dMMR had the highest TMBs (P < 0.0001) (Additional file 1: Supplementary Fig. 3A-B). Small cell lung cancer had the highest TMBs of the non-dMMR cancers. However, PD-L1 expression did not differ between dMMR tumours and pMMR tumours (P = 0.779) (Additional file 1: Supplementary Fig. 3C). Cancers associated with mutagens (NSCLC-Squamous, urothelial cancer) generally had the higher TMBs of any tumour type.
Relationship of PD-L1 expression and TMB
Figure 2A shows TMB for all PD-L1-negative, PD-L1-low-positive, and PD-L1-high-positive specimens. PD-L1-high-positive specimens had higher TMBs than PD-L1-low-positive and negative specimens (both P < 0.0001). Across all individual specimens examined, there was a small but positive association between the PD-L1 expression and TMB (Spearman R = 0.059, P < 0.0001). However, the relationship between these 2 biomarkers was not consistent across tumour types (Additional file 2: Supplemental Table 2). The strongest association between PD-L1 expression and TMB was within endometrial and neuroendocrine cancers (Spearman R all > 0.3). There was also a weak but positive association between PD-L1 expression and TMB among cervical, gastric (pMMR), HNSCC, NSCLC (non-squamous), NSCLC (squamous) and sarcomas (Spearman R all < 0.3). However, PD-L1 expression and TMB did not correlate among most other tumour types.
Figure 2B shows proportion of samples based on PD-L1-high and TMB-high classifications for different tumour types. Nasopharyngeal cancer had the highest proportion of samples with both PD-L1-high-positive specimens and high TMB (14%). However, endometrial, neuroendocrine, ovarian, gallbladder, breast, urothelial, gastric (dMMR), colorectal (dMMR), and colorectal (pMMR) cancers had no samples with both PD-L1-high-positive specimens and high TMB.
PD-1 + Tils infiltration is related to PD-L1 expression but not to TMB
Figure 3A shows the differences in PD-1+ Tils infiltration between all PD-L1-negative/low-positive and PD-L1-high-positive specimens, and Fig. 3B shows that between all TMB-moderate/low and TMB-high specimens. PD-L1-high-positive specimens had higher PD-1+ Tils infiltration than PD-L1-low-positive and negative specimens (P < 0.0001). However, there was no difference in PD-1+ Tils infiltration between TMB-moderate/low and TMB-high specimens (P = 0.9991). Across all individual samples, there was a positive association between the PD-1+ Tils infiltration and the PD-L1 expression (Spearman R = 0.3056, P < 0.0001) (Additional file 2: Supplemental Table 3). However, the relationship between these two biomarkers was not consistent across tumour types. The PD-1+ Tils infiltration and PD-L1 expression did not correlate within cervical, colorectal (dMMR), endometrial, gastric (dMMR), GIST, glioblastoma, small bowel, thymic, and urothelial cancers. There was a positive association between PD-1+ Tils infiltration and PD-L1 within other tumour types, and the strongest association was within breast, gallbladder, HNSCC, melanoma, and neuroendocrine cancers (Spearman R all > 0.4). There was no correlation between the PD-1+ Tils infiltration and TMB, whether across all individual samples or across tumour types (Additional file 2: Supplemental Table 4).
CD8 + T cell infiltration is related to PD-1+ Tils infiltration and PD-L1 expression but not to TMB
We labelled CD8 and PD-1 proteins on the same slides from 347 NSCLC samples by multiplex immunohistochemistry and analysed the content of CD8+ T cells, PD-1+ Tils, and CD8+PD-1+ T cells (Fig. 4A). The frequency of PD-1+ Tils, CD8+ T cells, and CD8+PD-1+ T cells across all individual samples varied, and their median frequency was 5.5% (IQR 2.9%-8.9%), 1.7% (IQR 0.6%-4.2%), and 0.1% (IQR 0%-0.5%) respectively. There was a positive association between CD8+ T cells and PD-1+ Tils (Spearman R = 0.4117, P < 0.0001) (Fig. 4B). However, the median proportion of CD8+PD-1+ T cells in total PD-1+ Tils and total CD8+ T cells was not high, 7.8% (IQR 3.6%-14.7%) and 2.4% (IQR 0.5%-7.6%), respectively (Fig. 4C-D). Further analysis also found a positive association between CD8+ T cells and PD-L1 expression (Spearman R = 0.2045, P = 0.0007), but there was no correlation between CD8+ T cells and TMB (Spearman R = 0.0007, P = 0.9138) (Fig. 4E-F).
PD-L1 expression, TMB, PD-1 + Tils, and CD8+ T cell infiltration are related to the response of anti-PD-1 therapy in NSCLC
In the 347 NSCLC patients mentioned above, 33 of them received the anti-PD-1 therapy or anti-PD-1 therapy plus chemotherapy. All patients were EGFR/KRAS wild-type patients, including 14 cases of lung adenocarcinoma and 20 cases of lung squamous cell carcinoma, of which 10 received PD-1 inhibitor monotherapy and 23 patients received a combination of PD-1 inhibitor plus chemotherapy. Detailed clinical information on those cases is provided in Additional file 2: Supplementary Table 5. The efficacy of treatment was evaluated as objective tumour response; 1 case achieved CR (complete response), 12 cases achieved PR (partial response), 9 cases had SD (stable disease), and 11 patients had PD (progressive disease). There was no significant difference in objective response rates (ORR = (CR + PR)/(CR + PR + SD + PD)) between the PD-1 inhibitor monotherapy group and the PD-1 inhibitor plus chemotherapy group. The ORR of the monotherapy group was 40.0%, and the ORR of the PD-1 inhibitor plus chemotherapy group was 39.1%. The heat map result for the cluster analysis of each case showed that the CR/PR group had higher levels of PD-L1 expression, TMB, PD-1+ Tils infiltration, and CD8+ T cell infiltration, and most patients in this group exhibited elevated levels of multiple biomarkers. However, in the SD/PD group, the four markers were at a lower level (Fig. 5A). Generally, the levels of the three biomarkers, including TMB, PD-1+ Tils infiltration, and CD8+ T cell infiltration, were significantly higher in the CR/PR group than in the SD/PD group (P = 0.0017, P = 0.0466, and P = 0.0396, respectively) (Fig. 5B-D). The PD-L1 expression was also higher in the CR/PR group than in the SD/PD group, but this difference was not significant (P = 0.7364) (Fig. 5E).