Evaluation of PD-L1 expression on circulating tumor cells (CTCs) in patients with advanced urothelial carcinoma (UC)

Abstract

Immune checkpoint inhibition (ICI) of the PD-1/PD-L1 axis shows durable responses in a subset of patients with metastatic urothelial carcinoma (UC). However, PD-L1 expression in tumor biopsies does not necessarily correlate with response to PD-1/PD-L1 inhibitors. Thus, a reliable predictive biomarker is urgently needed. Here, the expression of PD-L1 on circulating tumor cells (CTCs) in blood from patients with advanced UC was analyzed. For this purpose, an assay to test PD-L1 expression on CTCs using the CellSearch® system was established using cells of five UC cell lines spiked into blood samples from healthy donors and applied to a heterogeneous cohort of UC patients. Enumeration of CTCs was performed in blood samples from 49 patients with advanced UC. PD-L1 expression in ≥1 CTC was found in 10 of 16 CTC-positive samples (63%). Both intra- and inter-patient heterogeneity regarding PD-L1 expression of CTCs were observed. Furthermore, vimentin-expressing CTCs were detected in 4 of 15 CTC-positive samples (27%), independently of PD-L1 analysis. Both CTC detection and presence of CTCs with moderate or strong PD-L1 expression correlated with worse overall survival. Analyses during disease course of three individual patients receiving ICI suggest that apart from CTC numbers also PD-L1 expression on CTCs might potentially indicate disease progression. This is the first study demonstrating the feasibility to detect CTC-PD-L1 expression in patients with advanced UC using the CellSearch® system. This assay is readily available for clinical application and could be implemented in future clinical trials to evaluate its relevance for predicting and monitoring response to ICI.

Keywords: PD-L1; cellsearch; circulating tumor cells; immune checkpoint inhibition; urothelial carcinoma; vimentin.

© 2020 The Author(s). Published with license by Taylor & Francis Group, LLC.

Figures

Figure 1.

Characterization of the anti-PD-L1 antibody clone E1L3N®. (a) Western blot analysis of PD-L1, EpCAM, pan-keratin and vimentin expression in UC cell lines (RT-4, 647V, 5637, T24, and TCC-SUP) and in the breast cancer cell line MDA-MB-231. Protein loading control: HSC70 (Heat shock cognate 71 kDa protein). The bar chart indicates EMT scores according to Tan et al. (b) Specificity of the anti-PD-L1 antibody was confirmed by Western blot analysis of 5637 cells with the retroviral transfer of the CD274 gene encoding for PD-L1 or the empty vector (EV). Protein loading control: HSC70. (c) FACS (fluorescence activated cell sorting) analysis of PD-L1 expression in UC cell lines (RT-4, 647V, 5637, T24, and TCC-SUP). Cells were stained with the PE-conjugated anti-PD-L1 antibody clone E1L3N® (blue) in comparison to the respective isotype control clone DA1E (gray). Mean fluorescence intensities (MFI) were determined. (d) IF (immunofluorescence) analysis of PD-L1 expression in UC cell line cells (RT-4: PD-L1-negative, 647V: PD-L1-positive). Cells were spiked into whole blood from healthy donors prior to centrifugation. PD-L1 protein was detected by the PE-conjugated anti-PD-L1 antibody clone E1L3N®. The cells were additionally stained with the AlexaFluor488 (AF488)-conjugated anti-keratin antibodies (clones AE1/AE3 and C11) and the APC-conjugated anti-CD45 (clone REA747) antibody. Nuclei were stained by DAPI (4‘,6-Diamidin-2-phenylindol).

Figure 2.

CellSearch® CXC kit analysis of UC cells spiked into the blood from healthy donors. Cells were enriched by anti-EpCAM magnetic beads using the CXC kit. Keratin+ (KER-FLU), CD45- (CD45-APC), DAPI+ cells were identified as tumor cells. (a) PD-L1 expression of RT-4 cells and (b) 647V cells were detected by the PE-conjugated anti-PD-L1 antibody clone E1L3N® (PD-L1-PE, right panel) in comparison to the isotype control clone DA1E (isotype-PE, left panel). Images of PD-L1-specific fluorescent signals were generated using the indicated exposure time. Furthermore, PD-L1 expression of (c) 5637, (d) T24 and (e) TCC-SUP cells were detected by the PE-conjugated anti-PD-L1 antibody clone E1L3N® (PD-L1-PE) using an exposure time of 1.6 sec. Identification of TCC-SUP cells was performed by manually evaluating fluorescent signals of stained cells in the CellSelect mode because of the low keratin expression impeding automatic selection. (f) TCC-SUP identity of the reported cells was confirmed by single-cell mutational analysis of the PIK3CA gene exon 9. Representative sequencing results showed a heterozygous mutation (G1633A) previously reported by Platt et al. that was detected in a recovered single TCC-SUP cell but was absent in a leukocyte.

Figure 3.

Determination of PD-L1 expression on CTCs. (a) CellSearch® analysis of CTCs detected in the blood of an advanced UC patient. Cells were enriched by anti-EpCAM magnetic beads using the CXC kit. Keratin+ (KER-FLU), CD45- (CD45-APC), DAPI+ cells were identified as tumor cells. PD-L1 expression of individual CTCs was detected by the PE-conjugated anti-PD-L1 antibody clone E1L3N® (PD-L1-PE) and categorized into negative (0), weakly positive (1+), moderately positive (2+) and strongly positive (3+). Images of PD-L1-specific fluorescent signals were generated using an exposure time of 1.6 sec. (b) IF analysis of simultaneous binding of the therapeutic anti-PD-L1 antibody atezolizumab and the diagnostic anti-PD-L1 antibody E1L3N® to PD-L1 on tumor cells. PD-L1-negative RT-4 cells and PD-L1-positive 647V cells were spun onto glass slides and incubated in the presence of 1.2 µg/mL atezolizumab, which was detected by AlexaFluor488-labeled secondary antibody (AF488). Cells were additionally incubated with PE-conjugated anti-PD-L1 antibody. Nuclei were stained by DAPI. (c) Detection of atezolizumab in human blood samples. PD-L1-negative RT-4 cells and PD-L1-positive 647V cells spiked into the blood from healthy donors were spun onto glass slides and incubated in presence of 1.2 µg/mL atezolizumab, which was detected by anti-atezolizumab antibody and AF546-labeled secondary anti-mouse antibody. Mouse IgG served as an isotype control to anti-atezolizumab antibody. Keratins, CD45, and nuclei were stained by manually applying reagents of the CellSearch® CXC kit.

Figure 4.

CellSearch® CTC kit analysis of vimentin expression. Cells were enriched by anti-EpCAM magnetic beads using the CTC kit. Keratin+ (KER-PE), CD45- (CD45-APC), DAPI+ cells were identified as tumor cells. Vimentin expression of (a) RT-4, 5637 and MDA-MB-231 cells and of (b) CTCs detected in the blood of advanced UC patients was analyzed by the AlexaFluor488-conjugated anti-vimentin antibody clone V9 (VIM-AF488). Images of vimentin-specific fluorescent signals were generated using an exposure time of 0.8 sec. Vimentin expression of individual CTCs was categorized into negative (0), weakly positive (1+) and moderately positive (2+).

Figure 5.

Disease courses and clinical outcome of patients with advanced UC in relation to CTC-PD-L1 expression. (a) Kaplan-Meier plot and Log Rank test for overall survival of UC patients according to CTC-PD-L1 expression (0/1+ compared to 2+/3+) of CTC-positive samples using the CXC kit. Values of p < .05 were considered statistically significant. (b, c) Disease courses of two individual patients with metastatic UC who progressed under ICI (immune checkpoint inhibition) (B, pretreated; C, first-line) and (d) one patient who responded to ICI (pretreated) accompanied by CTC numbers as well as CTC-vimentin (VIM) and CTC-PD-L1 expression levels analyzed using the CellSearch® system. For each sample (1 and 2), metastatic sites are indicated. CTC analyses were performed using the CTC kit and/or the CXC kit (A and B, respectively, with N/A indicating that the analysis was not applied). Time spans between samples and maximum follow-up times as well as survival information are given. Arrows depict the duration of ICI therapy in relation to the acquisition of samples.