CTC-derived pancreatic cancer models serve as research tools and are suitable for precision medicine approaches

Abstract

Pancreatic ductal adenocarcinoma (PDAC) poses significant clinical challenges, often presenting as unresectable with limited biopsy options. Here, we show that circulating tumor cells (CTCs) offer a promising alternative, serving as a "liquid biopsy" that enables the generation of in vitro 3D models and highly aggressive in vivo models for functional and molecular studies in advanced PDAC. Within the retrieved CTC pool (median 65 CTCs/5 mL), we identify a subset (median content 8.9%) of CXCR4⁺ CTCs displaying heightened stemness and metabolic traits, reminiscent of circulating cancer stem cells. Through comprehensive analysis, we elucidate the importance of CTC-derived models for identifying potential targets and guiding treatment strategies. Screening of stemness-targeting compounds identified stearoyl-coenzyme A desaturase (SCD1) as a promising target for advanced PDAC. These results underscore the pivotal role of CTC-derived models in uncovering therapeutic avenues and ultimately advancing personalized care in PDAC.

Keywords: CXCR4; circulating cancer stem cells; compound screening; liquid biopsy; metabolism; pancreatic cancer.

Graphical abstract

Figure 1

Highly efficient retrieval of CTCs from patients with pancreatic cancer (A) Comparison of CTC capture rates using EpCAM antibody-coated beads and beads with additional antibodies in PDX-derived primary PDAC cultures (SiC024). EMT induction in PDAC cultures using macrophage-conditioned media (MCM) served two purposes: mimicking clinical heterogeneity and simulating metastatic conditions. During metastasis, cancer cells often downregulate EpCAM expression, rendering EpCAM-only isolation procedures less effective. To model CTC analysis, 100 PDAC cells were spiked into 5 mL of blood obtained from healthy individuals. Beads coated with varying antibody combinations (EpCAM alone, EpCAM + CD47, EpCAM + FAP, and a triple combination of EpCAM, CD47, and FAP antibodies) were employed to isolate CTCs. The graph illustrates the count of isolated spiked-in PDAC cells as a model for CTCs (n = 5). Boxplots show median values as horizontal lines within the boxes, with whiskers representing the 95% confidence intervals (CI). (B) Comparison of CTC retrieval efficiency of 100 cultured PDAC cells (SiC024) spiked into 5 mL of healthy blood. Three methods were used: triple antibody-coated beads, VAR2-coated beads, and immunodensity cell isolation by RosetteSep. The graph shows the number of retrieved CTCs (n = 5). (C) Analysis of CTC retrieval from four patients with PDAC. The same volume of patient-derived blood (5 mL) was used for each isolation. Beads coated with EpCAM or triple antibody combination or VAR2 were compared for effectiveness of CTC retrieval. For one patient, enough blood was available to perform the VAR2 beads-based isolation in duplicate. The chart shows the count of isolated CTCs and the fold increase in VAR2 bead-based isolation compared to EpCAM bead-based isolation. (D) Comparison of triple-antibody cocktail-coated beads and VAR2-coated beads for isolating CTCs from patients with PDAC. The left graph illustrates CTC retrieval numbers, the middle graph presents the count of contaminating white blood cells (WBCs), and the right panel displays representative immunostainings (n = 20). (E) VAR2-based retrieval of CD45^–CK⁺ CTCs for individuals with various conditions, including healthy controls, pancreatitis, intraductal papillary mucinous neoplasms (IPMNs), different stages of PDAC (resectable local, locally advanced, and advanced), neuroendocrine tumors (NETs), and bile duct tumors (cholangiocarcinoma). Boxplots display median values as horizontal lines, with whiskers indicating the 95% CI. (F) Kaplan Meier analysis for overall survival in patients with PDAC patients (n = 71), dichotomized by the “median CTC numbers” (65 CTCs). Statistical analysis by log rank test. (G) Assessment of the sensitivity of the VAR2-based retrieval method versus serum CA 19-9 levels for detecting local PDAC (left) and non-metastatic PDAC (right), respectively. The upper limit of normal (ULN) of 37 U/mL for CA 19-9 is highlighted in green. In one patient, no CTCs were detected (highlighted in orange). (H) Cox proportional hazards survival regression (Cox regression) analysis for overall 1-year survival. Hazard ratios and 95% confidence interval (CI) are shown. Definition of variables: CTC numbers ≥65/5 mL, disease stage I–IV, gender (M/F), age >65 years, CA 19.9 > 37 U/mL, C-reactive protein >3 mg/L. ∗p < 0.05, ∗∗p < 0.01, and ∗∗∗∗p < 0.0001. Please also see Figure S1.

Figure 2

CTCs give rise to long-term organoid cultures (A) Schematic of the procedure and culture conditions. (B) Representative high-magnification images showing the initial stage of CTC-derived organoid cultures isolated from four individual patients (SiC002, SiC003, SiC005, and SiC007). The formation of first organoid-like structures was regularly observed within 2–4 weeks (left), followed by further expansion and characterization through H&E staining and immunofluorescent confocal staining (middle). The obtained organoids also produced adherently growing monolayer primary cultures within 6 weeks (right). (C) Number of bulk CTCs or CXCR4⁺ CTCs/5 mL for patients with successful versus unsuccessful growth of organoid cultures. (D) Representative lower-magnification bright-field images of propagated organoid cultures from five representative patients. ∗∗∗∗p < 0.0001. Please also see Figure S2.

Figure 3

CTC-derived cultures are more aggressive compared to PDX-derived cultures (A) Invasion assay with fetal bovine serum as chemoattractant for comparing CTC-derived PDAC models versus PDX-derived cultures. Photographs on the left show invaded cells through a Matrigel-coated Boyden chamber, stained with crystal violet. Quantification is presented on the right. Boxplots show the median as a line within the box and use whiskers to represent the 95% CI. (B) In vivo metastasis assay involved injecting 100,000 luciferase-positive (Luc⁺) CTC- versus PDX-derived cells into the spleen, followed by splenectomy on day 7 and subsequent tracking of Luc⁺ liver metastasis until day 28. Quantification displayed in the upper-left panel. CTC numbers per mouse were assessed using the VAR2-coated beads method (upper-right panel). Ex vivo liver imaging, including quantification and representative ex vivo liver samples, is shown in the lower panel. (C) Immunohistochemical analysis for cytokeratin 19 (brown) to identify metastases in the liver and lungs of mice, injected with CTC-derived and PDX-derived PDAC cells. (D) Response of six representative CTC-derived models to gemcitabine (GEM), nab-paclitaxel (Abraxane; ABX), or their combination, as assessed by in vitro 3D sphere formation. (E) Corresponding patient’s clinical response for their individually allocated treatment regimen. ∗p < 0.05 and ∗∗∗p < 0.001. Please also see Figure S3.

Figure 4

CTC-derived cultures are functionally heterogeneous (A) Gating for FACSorting of SiC002 CTC-derived PDAC cells based on their mitochondrial membrane potential (ΔΨm) using tetramethylrhodamine methyl ester (TMRM; left panel). The right panel shows the oxygen consumption rate (OCR) indicative of mitochondrial respiration in ΔΨm high cells versus low cells after sorting. (B) Mitochondrial membrane potential (ΔΨm) stained with TMRM in CSC-enriched sphere versus adherent cultures derived from a diverse set of CTC- and PDX-derived models, respectively (n = 5). (C) ΔΨm stained with TMRM in CD133⁺ CSCs versus CD133^– differentiated cancer cells; n = 3. (D) qPCR analysis for mRNA levels of stemness-associated genes in ΔΨm high versus low cells sorted from SiC002 and SiC007 CTC-derived models; n = 4 (unpaired two-tailed t test). (E) Representative gating strategy of PDAC cells with low, medium, and high ΔΨm levels for FACSorting. (F) Representative images (left) and quantification (right) of sphere formation capacity on day 7 following ΔΨm sorting in two different CTC-derived PDAC models; n = 6 (unpaired two-tailed t test). (G) In vivo tumorigenicity of graded cell numbers for two different CTC-derived PDAC models after sorting for ΔΨm levels. Images of explanted tumors (left) and extreme limiting dilution analysis (right). (H) ΔΨm stained with TMRM for CXCR4⁺ versus CXCR4^– cells. Representative flow cytometry plots of CXCR4 gating and TMRM staining (left), quantification of mean fluorescence intensity (MFI) of TMRM staining (right); n = 3. (I) Immunofluorescence for CXCR4 (white) and TMRM (red) staining. DAPI serves as a nuclear stain. Representative images are on the left, with quantification on the right for two different CTC-derived PDAC models; n = 41 to 208 cells (horizontal lines indicate the mean values). ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. Please also see Figure S4.

Figure 5

Pancreatic CTCs comprise a subset of CXCR4⁺ CSCs (A) VAR2-based retrieval of CXCR4⁺CD45^–CK⁺ CTCs from individuals with various conditions, including healthy controls, pancreatitis, IPMNs, different stages of PDAC (resectable local, locally advanced, and advanced), NETs, and bile duct tumors (cholangiocarcinoma). The content of CD45^–CK⁺CXCR4⁺ CTCs among total CTCs at baseline is shown (left). Boxplots indicate median values with horizontal lines, and whiskers represent the 95% CI. Representative image of a retrieved CD45^–CK⁺CXCR4⁺ CTC (right). (B) Kaplan-Meier analysis for overall survival in patients with PDAC dichotomized for the “median of CXCR4⁺ CTC content” (8.9%). Statistical analysis by log rank test. (C) Cox proportional hazards survival regression (Cox regression) analysis for overall 1-year survival, including the variable “CXCR4⁺ CTC content.” Hazard ratios and 95% CI are shown on a logarithmic scale. (D) Single-cell droplet digital PCR (ddPCR) analysis. Upper panel: schematic illustration of the workflow. Lower panel: ddPCR results of freshly retrieved CXCR4⁺ versus CXCR4^– CTCs for the indicated genes plotted (n = 9 each). (E) Flow cytometry illustrating the content of CXCR4⁺ cells for PDX- versus CTC-derived cultures. (F) Invasive capacity of CXCR4⁺ versus CXCR4^– cells for CTC- and PDX-derived cultures, respectively. Cells sorted for CXCR4 were placed in the upper chamber of a Boyden chamber, and invasion toward 20% fetal bovine serum (FBS) in the lower chamber was recorded (n = 4–8). (G) In vivo tumorigenicity of CXCR4⁺ versus CXCR4^– cells for two representative CTC-derived models. Graded doses (1 to 10³ cells/injection) were transplanted into immunocompromised nude mice. Table providing quantification (upper panel), estimate of stem cell frequency in vivo for CXCR4^– (black) versus CXCR4⁺ (red) CTC-derived PDAC cells by limiting dilution assay (lower panel). Depicted is a log-fraction plot of the limiting dilution model. The slope of the line is the log-active cell fraction. The dotted lines give the 95% CI. Analysis was performed at https://bioinf.wehi.edu.au/cgi-bin/limdil/limdil.pl. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. Please also see Figure S5.

Figure 6

CTC-derived cultures are amenable to drug screening (A) Compound screen in 3D sphere cultures using viability assessment as readout. A stemness library was screened in three different CTC-derived PDAC models. Compounds that decreased average viability by at least 50% were further investigated and are indicated in distinct colors. (B) Viability assessment in sphere cultures for graded doses of selected compounds. Selected compounds from (A), which reduced viability by more than 50%, were subsequently tested in sphere cultures of several CTC-derived models (red) and PDX-derived models (blue). (C) Number of spheres/2 mL in response to treatment with the SCD-1 inhibitor A939572 (10 μM) or GEM or both for a distinct validation set of CTC-derived models and (D) PDX-derived models. Statistical analysis was performed using an unpaired two-tailed t test; n = 3. (E) In vivo drug response tests for A939572 (daily intraperitoneal injections of 20 mg/kg diluted in corn oil for 4 weeks) using three CTC-derived models (n = 8–10 tumors/group). For clarity, in vivo data are shown as means + SEM. ∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001, and ∗∗∗∗p < 0.0001. Please also see Figure S6.