Cell State and Cell Type: Deconvoluting Circulating Tumor Cell Populations in Liquid Biopsies by Multi-Omics

Abstract

Bi-directional crosstalk between the tumor and the tumor microenvironment (TME) has been shown to increase the rate of tumor evolution and to play a key role in neoplastic progression, therapeutic resistance, and a patient's overall survival. Here, we set out to use a comprehensive liquid-biopsy analysis to study cancer and specific TME cells in circulation and their association with disease status. Cytokeratin+, CD45- circulating rare cells (CRCs) from nine breast and four prostate cancer patients were characterized through morphometrics, single-cell copy number analysis, and targeted multiplexed proteomics to delineate cancer cell lineage from other rare cells originating in the TME. We show that we can detect epithelial circulating tumor cells (EPI.CTC), CTCs undergoing epithelial-to-mesenchymal transition (EMT.CTC) and circulating endothelial cells (CECs) using a universal rare event detection platform (HDSCA). Longitudinal analysis of an index patient finds that CTCs are present at the time of disease progression, while CECs are predominately present at the time of stable disease. In a small cohort of prostate and breast cancer patients, we find high inter-patient and temporal intra-patient variability in the expression of tissue specific markers such as ER, HER2, AR, PSA and PSMA and EpCAM. Our study stresses the importance of the multi-omic characterization of circulating rare cells in patients with breast and prostate carcinomas, specifically highlighting overlapping and cell type defining proteo-genomic characteristics of CTCs and CECs.

Keywords: breast cancer; circulating endothelial cells; circulating tumor cells; epithelial–mesenchymal transition; liquid biopsy; prostate cancer.

Figure 1

Schematic overview of HDSCA platform.

Figure 2

Longitudinal assessment of circulating rare cells in a patient with metastatic prostate cancer (PC1). (A) Representative cells organized by their Vim and AR expression. (B) CNA profiles together with IF images of the representative copy number of altered Vim− EPI.CTCs, copy number of altered Vim+ pEMT.CTCs and non-altered Vim+ CECs. (C) Copy number alterations of rare cells across four blood draws. (D) Multiplex protein expression of rare cells and CD45+ white blood cells across four blood draws. Cells with no Vim score available are color-coded grey in the top heatmap annotation. (E) CK intensity measured as standard deviation over the mean (SDOM). (F) Vim intensity and (G) cellular eccentricity from immunofluorescence image analysis of all cells with downstream NGS or multiplex proteomic data. Cells were scored as either clonal or non-altered based on the CNV profiles or as CTC or CEC based on the results from targeted proteomics. The Kruskal–Wallis test with Dunn’s correction for multiple comparisons was used to test for differences between each group. p-value annotations: 0.1234 (ns), 0.0332 (*), 0.0021 (**), 0.0002 (***), <0.00001 (****). (H) Longitudinal assessments across four blood draws. Cells were grouped by EPI.CTC (clonal, Vim−), pEMT.CTC (clonal, Vim+), and CEC (genomically non-altered, morphologically consistent with endothelial cell). Tick marks on the x-axis were set to 4-week intervals. Percentages of cell types might differ slightly between the total cells found per draw by the imaging microscope and those sequenced, as not all cells can be sequenced, and the sequenced cell population is, hence, a subset of the total cells detected. (AD = Active Disease; SD = Stable Disease). (I) Cellular eccentricity of CTCs and CECs per draw of cells that underwent IMC. (J) CK SDOM, (K) Vim intensity and (L) cellular eccentricity of CTCs and CECs across draws analyzed by multiplex proteomics. Scale bars are 10 μm.

Figure 3

Representative single-cell whole-genome copy number alterations of breast and prostate cancer patients. CNA profiles are displayed as ratio to median. Vim status, as determined by immunofluorescence, is annotated on top of heatmaps, where Vim− cells = light orange and Vim+ cells = dark orange. Copy number gains are defined as >1.25 above median, losses as <0.75 below median and copy number neutral between 0.75–1.25. Copy number gains = red, copy number loss = blue, copy number neutral = white. (A–F) Single-cell CNV heatmaps of breast cancer patients (BC 1–6). (G) Single-cell CNV heatmap of a prostate cancer patient (PC 2).

Figure 4

Multiplex proteomics of circulating rare cells and WBCs in patients with metastatic breast or prostate cancer. Multiplex proteomics of (A) breast cancer (BC 2, BC 7–9) and (B) prostate cancer (PC 1–4) patients. Grey = N/A; Not in panel.

Figure 5

Morphometrics and multi-omics (A) CNA profiles together with IF images of representative EPI.CTC, pEMT.CTC and non-altered Vim+ cells from breast cancer patients BC 1–6. (B) Protein expression of representative EPI.CTC, pEMT.CTC and CEC by IF and targeted proteomics from patient BC 2. (C) CK SDOM, (D) Vim intensity and (E) cellular eccentricity of rare cells from all breast and prostate cancer patients (BC 1–9 and PC 1–4) assessed by CNV and IMC. Eccentricity is determined on a scale of 0–1, where 0 = circle and 1 = ellipse. The Kruskal–Wallis test with Dunn’s correction for multiple comparisons was used to test for differences between each group. p-value annotations: 0.1234 (ns), 0.0332 (*), 0.0021 (**), 0.0002 (***), <0.00001 (****). Scale bars are set to 10 μm.

Figure 6

Comparison of endothelial cell lines with cells detected in MI and cancer patient blood samples. (A) Representative immunofluorescent images of HPAEC and HUVEC cell-line cells spiked into a NBD. (B) Rare cell enumeration of HPAECs and HUVECs compared to rare cells detected in NBD controls. HPAECs were spiked at approximately 430 cells/mL and had a recovery of 103%. HUVECs were spiked at approximately 100 cell/mL and had a recovery of 113%. (C) CK SDOM and (D) Vim signal intensity of spiked HPAECs and HUVECs cell line cells, CECs detected in MI patients, and CTCs and CECs detected in cancer patients. The Kruskal–Wallis test with Dunn’s correction for multiple comparisons was used to test for differences between each group. p-value annotations: 0.1234 (ns), 0.0332 (*), 0.0002 (***), <0.00001 (****). Scale bars are 10 μm.