Metabolic Heterogeneity in Patient Tumor-Derived Organoids by Primary Site and Drug Treatment

Abstract

New tools are needed to match cancer patients with effective treatments. Patient-derived organoids offer a high-throughput platform to personalize treatments and discover novel therapies. Currently, methods to evaluate drug response in organoids are limited because they overlook cellular heterogeneity. In this study, non-invasive optical metabolic imaging (OMI) of cellular heterogeneity was characterized in breast cancer (BC) and pancreatic cancer (PC) patient-derived organoids. Baseline heterogeneity was analyzed for each patient, demonstrating that single-cell techniques, such as OMI, are required to capture the complete picture of heterogeneity present in a sample. Treatment-induced changes in heterogeneity were also analyzed, further demonstrating that these measurements greatly complement current techniques that only gauge average cellular response. Finally, OMI of cellular heterogeneity in organoids was evaluated as a predictor of clinical treatment response for the first time. Organoids were treated with the same drugs as the patient's prescribed regimen, and OMI measurements of heterogeneity were compared to patient outcome. OMI distinguished subpopulations of cells with divergent and dynamic responses to treatment in living organoids without the use of labels or dyes. OMI of organoids agreed with long-term therapeutic response in patients. With these capabilities, OMI could serve as a sensitive high-throughput tool to identify optimal therapies for individual patients, and to develop new effective therapies that address cellular heterogeneity in cancer.

Keywords: breast cancer; cellular metabolism; heterogeneity; optical metabolic imaging; organoid; pancreatic cancer.

Copyright © 2020 Sharick, Walsh, Sprackling, Pasch, Pham, Esbona, Choudhary, Garcia-Valera, Burkard, McGregor, Matkowskyj, Parikh, Meszoely, Kelley, Tsai, Deming and Skala.

Figures

Figure 1

Patient-derived tumor organoid generation and optical metabolic imaging. (A) Pie charts depicting the success rate for generating viable organoids from patient pancreatic lesions (left), the distribution of PDAC, PanIN, anaplastic cancer, and ampullary cancer among successfully generated organoids (center), and the distribution of previously treated vs. untreated tumors among successfully generated organoids (right). Representative redox ratio (B), NAD(P)H τm(C), and FAD τm(D) images of an untreated pancreatic organoid taken 6 days after surgical resection (Patient PC14). Scale bar is 50 μm (E). Masks of individual cell cytoplasms overlaid onto NAD(P)H intensity image (F). Representative brightfield image of pancreatic organoids (Patient PC14). Scale bar is 200 μm. (G) Pie charts depicting the success rate for generating viable organoids from patient breast tumor biopsies (left), and the distribution of receptor subtypes among successfully generated organoids (right). Representative redox ratio (H), NAD(P)H τm(I), and FAD τm(J) images of an untreated breast cancer organoid taken 30 days after biopsy (Patient BC9). Scale bar is 50 μm. (K) Masks of individual cell cytoplasms overlaid onto NAD(P)H intensity image. (L) Representative brightfield image of breast cancer organoids (Patient BC9). Scale bar is 200 μm.

Figure 2

Sources of baseline metabolic variability in patient-derived organoid cells. (A) The percentage of total metabolic variation among each breast cancer patient's cells which can be explained by variation at the organoid level for each variable. (B) The percentage of total metabolic variation among each pancreatic patient's cells which can be explained by variation at the organoid level for each variable. (C) The percentage of total metabolic variation among cells from all breast cancer patients which can be explained by variation between patients for each variable. (D) The percentage of total metabolic variation among cells from all pancreatic patients which can be explained by variation between patients for each variable. Untreated organoid cells from the first measurement time point were included in this analysis.

Figure 3

Disparities in metabolism and intra-organoid heterogeneity by morphology and cancer type. (A) Representative brightfield images of hollow breast cancer organoids. Scale bar = 100 μm. (B) Representative NAD(P)H intensity images of hollow breast and pancreatic cancer organoids. Scale bar = 100 μm. (C) Representative brightfield images of solid breast cancer organoids. Scale bar = 250 μm. (D) Representative NAD(P)H intensity images of solid breast and pancreatic cancer organoids. Scale bar = 100 μm. (E–H) The intra-organoid standard deviation of the OMI index (E), coefficient of variation of the redox ratio (F), coefficient of variation of mean fluorescence lifetime of FAD (G), and the mean fluorescence lifetime of NAD(P)H (H) for cells within hollow vs. solid organoids. *p < 0.05. Each dot represents one organoid (mean ± SEM).

Figure 4

OMI captures non-genetic cellular heterogeneity in pancreatic organoids. (A) Representative images of the redox ratio, NAD(P)H τm, and FAD τm, in organoids generated from Patient PC13 (anaplastic carcinoma of the pancreas), treated with standard chemotherapies and experimental targeted therapies for 72 h. Scale bar is 50 μm. (B,C) Heatmap representation of the OMI index treatment effect size (Glass's Δ) for all patients at 72 h in organoids (B) and fibroblasts cultured with organoids. (C) “∧” indicates the patient lesion was diagnosed as PanIN. “~” indicates the patient lesion was diagnosed as ampullary cancer. *Glass's Δ ≥ 0.75. (D) Normalized density distributions of the OMI index of individual cells contain subpopulations with G+P treatment, but not ABT-263 + T treatment after 72 h in PC13 organoids. Bracketed number indicates number of subpopulations. (E) High-depth targeted cancer gene sequencing of PC13 organoids. Allele frequencies of ~50% for KRAS and 100% for TP53 were found (black bars). Alterations with allele frequencies of 10–30% were detected (gray bars), but none of these alterations were pathogenic. (F) Cleaved caspase-3 staining of PC13 organoids shows differences in apoptosis between treatment conditions after 72 h of treatment. (G) Ki67 staining of PC13 organoids shows differences in proliferation between treatment conditions after 72 h. Each dot represents one organoid (mean ± SEM), and red indicates significant response to treatment. *p < 0.05 vs. control.

Figure 5

Organoid-based predictions of pancreatic cancer patient response to therapy. (A–D) Representative redox ratio, NAD(P)H τm, and FAD τm images of pancreatic organoids from Patients PC1 (A), PC2 (B), PC6 (C), and PC14 (D), who are classified as predicted responders. Left columns indicate control organoids, and right columns indicate organoids treated with the same drugs as the patient adjuvant treatment. Scale bar is 50 μm. (E–H) The effect of the same drugs on the OMI index averaged across all cells in organoids derived from Patient PC1 (E), PC2 (F), PC6 (G), and PC14 (H). Error bars indicate mean ± SEM. *p < 0.0001. (I–L) Single-cell OMI index subpopulation analysis of treatment response in organoids from Patient PCI (I), PC2 (J), PC6 (K), and PC14 (L). (M–O) Representative redox ratio, NAD(P)H τm, and FAD τm images of pancreatic organoids from Patients PC3 (M), PC8 (N), and PC18 (O), who are classified as predicted non-responders. Left columns indicate control organoids, and right columns indicate organoids treated with the same drugs as the patient adjuvant treatment. (P–R) The effect of the same drugs on the OMI index averaged across all cells in organoids derived from Patient PC3 (P), PC8 (Q), and PC18 (R). (S–U) Single-cell OMI index subpopulation analysis of treatment response in organoids from Patient PC3 (S), PC8 (T), and PC18 (U).

Figure 6

Pancreatic cancer patient clinical outcomes while on adjuvant therapy. (A) Swimmer plot indicating the number of months without recurrence following surgical resection of the tumor and adjuvant treatment. Patients are classified as predicted responders and non-responders based on organoid response profiles. Arrows indicate that the patient continues to survive without recurrence at the time of publication. “~” Indicates the patient's lesion was diagnosed as ampullary cancer. (B) Patients with RFS > 12 months had higher OMI index effect sizes at 72 h (Glass's Δ) than patients with RFS <12 months (mean± SEM). Dotted line represents proposed cutoff of Δ = 0.75. (C) Patients with RFS > 12 months show a decrease in wH-index with treatment compared to control organoids. Patients with RFS < 12 months show an increase in wH-index with treatment compared here to control organoids. Error bars not visible. N = 1,000 fits/group.

Figure 7

OMI of single-cell treatment response in patient-derived breast tumor organoids. (A) Representative images of the optical redox ratio, NAD(P)H τm, and FAD τm organoids generated from Patient BC8 after 72 h of treatment. Scale bar = 100 μm. (B) Heatmap representation of the OMI index treatment effect size (Glass's Δ) at 72 h in organoids from each breast cancer patient. *Glass's Δ ≥ 0.75. 4-OOH cyclophosphamide (active metabolite) was used in place of cyclophosphamide. (C,D) Representative images of control (C) and A+C+T treated (D) BC8 organoids stained for Ki67 (green, proliferation), cleaved caspase-3 (red, apoptosis), and DAPI (blue, nuclei) after 72 h of treatment. Scale bar 100 μm. (E) Cleaved caspase-3 staining of organoids shows differences in apoptosis between treatment conditions after 72 h of treatment in BC8. (F) Ki67 staining of organoids shows differences in proliferation between treatment conditions after 72 h in BC8. Each dot represents one organoid (mean ± SEM). *p < 0.05 vs. control. (G) The effect of A+C+T treatment at 72 h on OMI index heterogeneity quantified by the wH-index in Patient BC8, BC17, and BC20 organoids. (H–J) The effect of 72 h A+C+T treatment on the OMI index averaged across all cells in organoids derived from Patient BC8 (H), BC17 (I), and BC20 (J). Error bars indicate mean ± SEM. *p < 0.0001. Single cell OMI index subpopulation analysis of 72 h of A+C+T treatment response in organoids from Patient BC8 (K), BC17 (L), and BC20 (M).