Optical Imaging of Drug-Induced Metabolism Changes in Murine and Human Pancreatic Cancer Organoids Reveals Heterogeneous Drug Response

Abstract

Objectives

Three-dimensional organoids derived from primary pancreatic ductal adenocarcinomas are an attractive platform for testing potential anti-cancer drugs on patient-specific tissue. Optical metabolic imaging (OMI) is a novel tool to assess drug-induced changes in cellular metabolism and its quantitative endpoint, the OMI index, is evaluated as a biomarker of drug response in pancreatic cancer organoids.

Methods

Optical metabolic imaging is used to assess both malignant cell and fibroblast drug response within primary murine and human pancreatic cancer organoids.

Results

Anti-cancer drugs induce significant reductions in the OMI index of murine and human pancreatic cancer organoids. Subpopulation analysis of optical metabolic imaging data revealed heterogeneous drug response and elucidated responding and non-responding cell populations over a 7-day time-course. OMI index significantly correlates with immunofluorescence detection of cell proliferation and cell death.

Conclusions

Optical metabolic imaging of primary pancreatic ductal adenocarcinoma organoids is highly sensitive to drug-induced metabolic changes, provides a non-destructive method for monitoring dynamic drug response, and presents a novel platform for patient-specific drug testing and drug development.

Introduction

Primary tissue derived organoids are an attractive platform for studying disease progression, invasion, and drug response^1-3. Primary ductal adenocarcinoma (PDAC) organoids have the same protein and genetic abnormalities as malignant disease² and better replicate tumor behaviors than 2-dimensional cultures of cells. Additionally, hundreds to thousands of organoids can be generated from a single tissue biopsy, which allows for high-throughput screening of organoid behaviors in response to perturbations. Therefore, organoids provide a relevant system for testing drug response for individualized treatment planning and novel drug discovery.

Optical imaging is well suited for imaging organoids due to its high spatial resolution, functional contrast, and depth of imaging. Optical metabolic imaging (OMI) is a novel, non-destructive method of imaging and quantifying drug-induced changes in cellular metabolism^3-5. OMI probes the autofluorescence intensity and lifetime of NAD(P)H and FAD for robust detection of cellular metabolic states. The OMI endpoint, the optical redox ratio, is the fluorescence intensity of NAD(P)H divided by the intensity of FAD, and measures the redox state of the cell^6,7. The fluorescence lifetime is the time a fluorophore remains in the excited state, and is altered by protein conformation changes, proximity to quenchers, and preferred protein binding. A comprehensive analysis of cellular metabolism can be obtained by probing NAD(P)H and FAD fluorescence intensity and lifetime⁴. Furthermore, OMI is well suited for evaluating heterogeneous drug response due to its single-cell resolution and high sensitivity to changes in cellular metabolism.

The OMI index, a composite endpoint of NAD(P)H and FAD fluorescence properties, was developed to directly correlate metabolism changes with drug response. The OMI index is a linear combination of mean-centered optical redox ratio, NAD(P)H fluorescence lifetime, and FAD fluorescence lifetime data, computed for each cell within a data set. The OMI index allows comparison of metabolic states across experimental groups, with significant decreases in OMI index indicating drug response³. In this study, the OMI index is evaluated as a robust and highly sensitive biomarker for drug response in PDAC organoids, derived from both primary murine and human tumors.

Materials and Methods

Mouse Organoid Generation and Culture

This study was approved by the Vanderbilt University Animal Care and Use Committee and meets the NIH guidelines for animal welfare. Sections of PDAC were resected from 3 PKT (Ptf1a^cre/+;LSL-Kras^G12D/+;Tgfbr2^flox/flox) mice⁸. Tumor sections were washed 3 times with PBS and placed in 0.5 ml RPMI media supplemented with 10% FBS, 1% penicillin: streptomycin, and 10 ng/ml epidermal growth factor receptor (EGF), hereafter referred to as PDAC organoid media. Organoids were generated by mechanical dissociation of the tumor sections with surgical scissors and a scalpel. The resulting solution was filtered to exclude tissue sections larger than 500 μm in diameter and mixed with Matrigel 1:2. Gels were allowed to solidify at room temperature for 30 minutes and in an incubator for 1 hour. Then, media was added to cover each gel. Organoids grew for three days and then were treated with the following drugs and drug combinations: control (DMSO), gemcitabine (G; 25 μg/ml), AZD1480 (A1; 100 nM), AZD6244 (A6; 4 μM), XL147 (X; 25 nM), G+A1, G+A1+A6, and G+A1+A6+X. Media was replaced every 3 days.

Human Organoid Generation and Culture

This study was approved by the Vanderbilt University Institutional Review Board. A primary tumor biopsy was obtained from a patient with poorly differentiated pancreatic ductal adenocarcinoma during a Whipple procedure (pancreaticoduodenectomy). The tumor section was placed in PDAC organoid media and transported on ice to the Optical Imaging Laboratory, ~5 min walk. Previous studies of organoid generation and OMI endpoints confirm tissue viability and metabolic endpoint congruity within this timeframe^3,9. The tissue sample was washed 3 times with PBS and placed in 0.5 ml of PDAC organoid media. Organoids were generated by mechanical dissociation with surgical scissors and a scalpel, in the same procedure as the mouse tissue. The tissue sections were filtered remove tissue sections larger than 500 μm in diameter and mixed with Matrigel 1:2. Gels were allowed to solidify at room temperature for 30 minutes and in an incubator for 1 hour. Then, PDAC organoid media was added to cover each gel. Organoids grew for three days and then were treated with the following drugs and drug combinations: control (DMSO), gemcitabine (G; 25 μg/ml), AZD1480 (A1; 100 nM), and G+A1. Media was replaced every 3 days.

Optical Metabolic Imaging

Optical metabolic imaging probes the fluorescence intensity and lifetime of NAD(P)H and FAD. NAD(P)H and FAD are coenzymes used in multiple cellular metabolism processes including glycolysis and oxidative phosphorylation. The endpoints of OMI include the redox ratio, NAD(P)H fluorescence lifetime, FAD fluorescence lifetime, and a combination variable, the OMI index. The redox ratio is the intensity of NAD(P)H fluorescence relative to the intensity of FAD fluorescence and provides information on the relative amounts of electron donors and acceptors in the cell^6,10. The redox ratio is sensitive to shifts in metabolic pathways^4,10. The fluorescence lifetimes report changes in the microenvironment of NAD(P)H and FAD and are especially sensitive to the binding state of the fluorophore, as well as local temperature, pH, and proximity to quenchers such as molecular oxygen¹¹. Both NAD(P)H and FAD fluorescence lifetimes can be either short or long, depending on the binding state of NAD(P)H and FAD (free or bound to an enzyme complex)^12,13. Previous studies have shown that OMI endpoints are sensitive to metabolism differences between cancer subtypes^3,4,14. Additionally, the OMI endpoints provide dynamic readouts of cellular metabolism and detect pre-malignant transformations within tissues^5,15, classify subtypes of breast cancer cells^4,14, and detect response to anti-cancer drugs³.

Fluorescence lifetime imaging was performed on a custom-built multiphoton microscope adapted for lifetime imaging, as previously described^3,4,9. Laser excitation light was provided by a Titanium:Sapphire laser (Coherent, Inc) tuned to 750 nm to excite NAD(P)H (average power ~7.5-8 mW at the sample) and 890 nm to excite FAD (average power ~8.1-8.5 mW at the sample). Excitation and emission light was coupled through a 40X oil immersion objective (1.3 NA) within an inverted microscope, (Nikon; TiE). Customized filter cubes isolated NAD(P)H emission between 400-480 nm and FAD emission between 500-600 nm. Fluorescence lifetime imaging was performed with time correlated single photon counting electronics (Becker and Hickl; SPC-150) and a GaAsP PMT (Hamamatsu, H7422P-40). A 4.6 μs pixel dwell time was used to acquire 256×256 pixel images. Fluorescence lifetime images were acquired for 60s, with the photon count rate maintained above 5×10⁵, ensuring no photobleaching occurred and adequate photon observations for lifetime decay fitting. The instrument response function full width at half maximum was 260 ps as measured from the second harmonic generation of a urea crystal. Daily fluorescence lifetime measurements were validated by imaging of a YG fluorescent bead (Polysciences, Inc). The measured lifetime of the bead, 2.1 ns +/− 0.04 ns, agrees with published values^3,4.

To ensure spectral isolation of NAD(P)H and FAD, we performed a cyanide experiment⁴ in which cells were exposed to 4 nM NaCN. Cyanide disrupts cellular metabolism and should result in an increased NAD(P)H concentration/fluorescence signal and a decreased FAD concentration/fluorescence signal in the first minutes following exposure¹⁶. FAD images were examined for strong pixelated fluorescence signals with very short lifetimes, as would be apparent from lipofuscin granules, to ensure no fluorescence contributions from lipofuscin in the FAD images.

Organoid Imaging

Images of NAD(P)H and FAD fluorescence lifetimes were acquired on 1, 2, 3, 5, and 7 days of drug treatment. Six to 9 images were acquired for each specific sample drug treatment, with 1-2 images of Type 1 organoids (due to their scarcity), 4-6 images of Type 2 organoids, and 1-2 images of fibroblasts. The data from all three mouse samples was combined for a total of 5-6 images of Type 1 organoids (100-600 cells), 12-18 images of Type 2 organoids (250-800 cells), and 4-6 images of fibroblasts (30-100 cells) per drug treatment. Five to 6 human PDAC organoids were imaged per treatment group, for a total number of 100-300 cells per treatment.

Immunofluorescence

Immunofluorescence of organoid cultures was performed as previously described^3,17. First, the growth media was removed from the gels and the gels were washed with PBS. The PBS was removed and gels were fixed by submerging in 2 ml of 4% paraformaldehyde in PBS. After 10 minutes, the paraformaldehyde was removed and gels were washed with PBS. Then, 0.02% Triton X-100 in PBS was added to cover the gels for 10 minutes. Gels were washed with PBS and overlain with 1% fatty-acid free BSA, 1% donkey serum in PBS overnight. The next day, the blocking solution was removed from the gels and 100 μl of antibody solution (diluted antibody in PBS with 1% donkey serum) was added to each gel and gels were incubated at room temperature for 30 minutes. The gels were washed in PBS 3 times and then incubated with the secondary antibody solution for 30 minutes at room temperature. The gels were then wasted 3 times with PBS and twice with water. The gels were then mounted on slides using 30 μl of ProLong Antifade Solution (Molecular Probes).

Cleaved caspase 3 expression was detected with anti-cleaved caspase 3 primary antibody (Life Technologies). Cell proliferation was detected with anti-Ki67 primary antibody (Life Technologies). Both of these primary antibodies were diluted 1:100. A goat anti-rabbit IgG FITC (Life Technologies) secondary antibody was used for imaging contrast. Images of FITC fluorescence were obtained using the multiphoton microscope with excitation at 980 nm. A minimum of 6 organoids were imaged. Positive staining was confirmed by positive staining of mouse thymus and mouse small intestine for cleaved caspase 3 and Ki67, respectively. Immunofluorescence images were quantified by manual counting the number of positive cells and the total number of cells. Immunofluorescence results are presented as a percentage of positively stained cells, quantified from 6 organoids, approximately 200 cells.

Immunohistochemistry

Gels were scraped and mixed with agarose to create a solid gel for IHC processing. The agarose was placed in buffered formalin, paraffin embedded, sliced, and stained with hematoxylin and eosin stains. Additional slides were stained for cytokeratin AE1/AE3. Optical Metabolic Imaging Data Analysis

Fluorescence lifetime and intensity data was extracted from the fluorescence lifetime images. At each pixel, the fluorescence lifetime decay curve was fit to a two component model,

$$I\left(t\right)={\alpha }_1\exp \left(-\frac{t}{{\tau }_1}\right)+{\alpha }_2\exp \left(-\frac{t}{{\tau }_2}\right)+C$$

where I(t) is the fluorescence intensity at time t after the laser excitation pulse, α₁ and α₂ are the fractional contributions of the free and bound molecules (i.e. α₁ + α₂ = 1), τ₁ and τ₂ are the fluorescence lifetimes of the short and long lifetime components, and C is a constant that accounts for background light. A two component model is used because NAD(P)H and FAD can exist in two conformation states, bound or unbound, which correspond to short (quenched) and long lifetimes^5,11,12.

Cytoplasms of cells were segmented from the nuceli and background of autofluorescence organoid images using an automated image segmentation algorithm¹⁸. Fluorescence lifetime endpoints, including photon count, α_1, τ_1, and τ₂ are extracted for each cell cytoplasm. Three additional endpoints, the optical redox ratio (intensity of NAD(P)H to intensity of FAD), the mean NAD(P)H fluorescence lifetime, and the mean FAD fluorescence lifetime were also computed and recorded for each cell cytoplasm. The mean NAD(P)H and mean FAD fluorescence lifetimes were computed from the components, τ_m = α₁ τ₁ + α₂ τ₂. A composite endpoint, the OMI index, was computed from the redox ratio, NAD(P)H τ_m and FAD τ_m. The OMI index is computed for each cell, $OMIIndex=\frac{R{R}_i}{\langle RR\rangle }+\frac{NAD\left(P\right)H{\tau }_{mi}}{\langle NAD\left(P\right)H{\tau }_m\rangle }-\frac{FAD{\tau }_{mi}}{\langle FAD{\tau }_m\rangle }$, and is a linear combination of mean centered redox ratio, NAD(P)H τ_m, and FAD τ_m^3,19. The “OMI index normalized difference” is defined as $OM{I}_{ND}=\frac{(OM{I}_T-OM{I}_C)}{OM{I}_C}$, where OMI_ND is the OMI index normalized difference, OMI_C is the OMI index of the control-treated organoids, and OMI_T is the OMI index of the drug-treated organoids.

Subpopulation Analysis

Heterogeneity of cellular metabolism was assessed with subpopulation analysis. Each cell population was modeled as a Gaussian mixture distribution model^3,4,19,20,

$$f(y;{\Phi }_g)={\sum }_{i=1}^g{\pi }_i\varphi (y;{\mu }_i,V_i),$$

where g is the number of components, ϕ(y; μ_i, V_i) represents a normal probability density function with mean μ_i and variance V_i, and π_i is the mixing proportion. Φ_g represents the unknown parameters, (π_i, μ_i, V_i): i =1...g in a g-component model. The mixture model was fitted by maximum likelihood using the expectation maximization algorithm to determine the optimum parameters, (π, μ, V), (Matlab). Cell populations were modeled three times as Gaussian mixture distribution models with 1-3 components (g = 1, 2, or 3). The Akaike information criteria (AIC) is a measure of model goodness of fit and is minimized in the optimal model²¹. The most representative model of the data was selected as the model with the lowest AIC. Probability density functions were normalized to have an area under the curve equal to 1.

Statistical Analysis

Time course drug response OMI data was analyzed with one-way ANOVA with a Dunn correction for multiple comparisons between control and drug treated organoids at each time point. A non-parametric rank sum test with a Bonferroni correction for multiple comparisons was used to assess differences in immunofluorescence results between drug treatments. For all statistical comparisons, an alpha level of 0.05 was used for significance. The number of cells per group varied between 30 and 800. A non-parametric Spearman's correlation was performed for correlation analysis between OMI index and immunofluorescence endpoints for all time points, treatment groups, and cell subtypes pooled together.

Results

Primary ductal adenocarcinoma (PDAC) tumors were mechanically dissociated and grown as organoids in Matrigel. Three distinctive morphologies were observed in murine PDAC organoids (Fig. 1A), including spherical organoids (Type 1), asymmetric organoids (Type 2), and fibroblasts. Immunohistochemistry analysis revealed that Type 1 and 2 organoids were positive for Cytokeratin AE1/AE3 (Fig. 1B) and negative for vimentin, CD34, and CD45, indicating an epithelial lineage in agreement with other studies of PDAC organoids². The basal metabolic state varied across the organoid subtypes, with Type 1 having the greatest OMI index and Type 2 having the smallest OMI index (Fig. 1C). Significant differences in the optical redox ratio indicated a lower redox state in Type 2 organoids compared to Type 1 and fibroblasts (Fig. 1D). Furthermore, NAD(P)H and FAD fluorescence lifetime analysis revealed significantly shorter NAD(P)H lifetimes in Type 2 organoids (Fig. 1E) and longer FAD fluorescence lifetimes (Fig. 1F) compared to Type 1 and fibroblasts.

Figure 1.

(A) Representative Redox Ratio, NAD(P)H τ_m, and FAD τ_m images of murine organoids and fibroblasts. (B) Cytokeratin AE1/AE3 staining of murine PDAC Type 1 and 2 organoids. (C) OMI Index, (D) Optical Redox Ratio (NAD(P)H/FAD), (E) NAD(P)H mean lifetime, and (F) FAD mean lifetime of untreated Type 1, Type 2 organoids and fibroblasts;*p<0.05, **** p<0.0001.

Organoids were treated with gemcitabine, AZD1480 (JAK2 inhibitor), AZD6244 (MEK inhibitor), XL147 (PI3K inhibitor), and combinations to evaluate drug-induced changes in cellular metabolism. OMI was performed over 7 days of drug treatment. The time-course of drug-induced metabolism changes (Fig. 2A) in Type 1 organoids demonstrated an initial reduction in the OMI index on day 1 of AZD1480 treatment (p<0.05), and an increase in the OMI index by days 5 and 7. Gemcitabine induced a metabolic response with a progressive reduction in the OMI index over the entire time-course (p<0.05; Fig. 2A). The therapies AZD6244 and XL146, and the combination therapies, A1+G, A1+A6+G, and A1+A6+G+X, induced a significant reduction in the OMI index that persisted over the time-course (p<0.01; Fig. 2A), with the greatest reductions induced by the combination therapies.

Figure 2.

(A) Time-course of drug response for Type 1 organoids. (B) Subpopulation analysis of AZD1480 treated Type 1 organoids. (C) Time-course of drug response for Type 2 organoids. (D) Time-course of drug response for fibroblasts. (E) Subpopulation analysis of A1+A6+G+X treated fibroblasts.

Subpopulation analysis of OMI index data can reveal cell populations with heterogeneous metabolic responses¹⁹. This analysis revealed that Type 1 organoids treated with AZD1480 initially had a bimodal metabolic response to treatment (Fig. 2B). However, this bimodal metabolic response disappeared at later time points, day 3 and 7, revealing the emergence of a distribution that closely matches control.

The time course of drug-induced changes for Type 2 organoids showed significant reductions in OMI index due to drug treatment (Fig. 2C). By day 7, the OMI index was significantly reduced due to all single and combination drug treatments (p<0.05; Fig. 2C).

We also assessed drug-induced metabolism changes of the fibroblasts that grew with the organoid cultures. The time-course of drug-induced metabolism changes within fibroblasts indicated significant metabolism changes with AZD1480, AZD6244, A1+G, A1+A6+G, and A1+A6+G+X treatments (Fig. 2D). Both gemcitabine and XL147 induced a decrease in the OMI index initially at days 1 and 2 (p<0.05) but the OMI index reverted to control levels by days 3, 5, and 7 (p>0.05 vs. control). Subpopulation analysis of the A1+A6+G+X treated fibroblasts revealed two populations of cells on days 1 and 3 of treatment (Fig. 2D) and a single population of cells on day 7. Over the time course, the A1+A6+G+X population shifts to OMI index values significantly reduced from control cells.

Drug response of the organoids and fibroblasts was confirmed by immunofluorescence staining of cleaved-caspase-3 and Ki67 on day 3 (Fig. 3). Correlation analysis revealed the OMI index significantly correlated (p<0.05) with both cleaved-caspase-3 and Ki67 staining in the pooled data of both types of organoids and fibroblasts (Fig. 4A), indicating that the OMI index is a robust indicator of drug response. The correlations between immunofluorescence and the OMI index within the type 1 organoids, type 2 organoids, and fibroblasts is also shown (Fig. 4 B-D).

Figure 3.

(A-B) Immunofluorescence staining of cleaved caspase 3 and Ki67 for murine PDAC Type 1 organoids, (C-D) Type 2 organoids, and (E-F) fibroblasts, treated with anti-cancer drugs for 72hr. Image inserts (A-D) are representative images of control (left) and A1+G (right) organoids. Scale bar is 50 μm. * p<0.05, ** p<0.01

Figure 4.

(A) Correlation analysis of the OMI index versus cleaved caspase 3 and Ki67 staining for both types of organoids and fibroblasts pooled together. Correlation between (B) Type 1, (C) Type 2, and (D) Fibroblast OMI index versus Cleaved Caspase 3 and Ki67. ns = not significant, p>0.05

To translate this organoid and OMI approach to human tumors, organoids were generated from a primary human PDAC biopsy. Organoids grew robustly from the human PDAC biopsy (Fig. 5A); however, a broad spectrum of organoid morphologies grew that were not easily classified into subtypes. The cells within these organoids stained positive for cytokeratin AE1/AE3, indicating epithelial origin (Fig. 5B). These organoids were treated for 24hr with gemcitabine, AZD1480, and A1+G. A significant reduction in the OMI index was detected with gemcitabine and A1+G treatment (p<0.01; Fig. 5C). AZD1480 failed to induce a significant reduction in OMI index in these primary human PDAC organoids at 24hr.

Figure 5.

(A) Representative NAD(P)H τ_m images of human PDAC organoids. (B) Cytokeratin AE1/AE3 staining of human PDAC organoid. (C) OMI Index of human PDAC organoids treated for 1 day. ** p <0.01; *** p<0.001

Discussion

OMI of PDAC organoids is an attractive platform for studying drug response of novel drugs and for testing specific drugs on patient tissue for individualized treatment decisions. This study investigated the use of an organoid-OMI screen for detecting drug-induced metabolism changes in murine and human PDAC organoids.

We achieved 100% organoid generation rates with mechanical dissociation, with thousands of organoids generated per tissue sample, suggesting robust representation of the majority of cancerous cells in the organoids. Furthermore, the isolation of two different morphologies of malignant organoids as well as the growth of fibroblasts indicates isolation and culture of multiple cell types. While some cell types inherent in the initial tumor tissue may not be represented in the organoids, the organoids morphology matches that of a prior study of PDAC organoid generation². The Type 1 organoid morphology matches that of organoids derived from primary PDAC while Type 2 matches that of organoids derived from metastatic lesions suggesting different protein expression and genetics between the two organoid types². Our results showed that organoids of differing morphologies could be grown from murine PDAC biopsies (Fig. 1A). Furthermore, in addition to morphological differences, these organoids had varying basal metabolic redox states⁷ (Fig. 1D) and differing NAD(P)H and FAD lifetimes indicating differing coenzyme environments and preferred protein binding (Fig. 1E-F).

The drug response studies indicated that both Type 1 and Type 2 organoids were responsive to the anti-cancer drugs and combinations (Fig. 2A, C), with the greatest reductions in OMI index due to combination therapies, supporting the use of combination treatments in the clinic. The OMI results suggest similar drug-response behavior between the two organoids, however, Type 2 shows a response to AZD1480 at days 5 and 7 while Type 1 develops a resistance to this drug. Furthermore, the subpopulation analysis highlights the utility of OMI and demonstrates a dynamic behavior of Type 1 organoids which initially responded to AZD1480 and later regressed by day 7 (Fig. 2B). The immunofluorescence results and correlation analysis (Fig. 3 and 4) demonstrate that the OMI index correlates with cell proliferation and death and is an accurate measure of drug response in these PDAC organoids. Notably, no spatial patterns of Ki67 or cleaved caspase 3 staining were observed indicating penetrance of a therapeutic dose of drugs throughout the organoids, in agreement with prior studies of nutrient and drug diffusion in organoids^22,23. When pooled, the Type 1, 2 and fibroblast IHC and OMI index correlated, yet the individual correlations were not all significantly correlated due to the small sample size. OMI can measure dynamic changes in cellular metabolism within intact, living samples over a time-course of drug treatment, which is not possible with destructive immunofluorescence measurements.

Due to the fibrotic nature of PDAC, drug delivery remains a challenge²⁴. While our results demonstrate drug-induced metabolism changes in epithelial tumor cells, these drugs may have limited efficacy in vivo due to hindrance of drug delivery by the extracellular matrix. Several new therapeutic strategies are being investigated to directly target the extracellular matrix and the malignant tumor compartment concurrently^8,24. Therefore, we also evaluated the metabolism of drug-treated fibroblasts that grew with the PDAC organoids (Fig. 2D). The fibroblasts also showed response to the anti-cancer drugs with decreased OMI indices and decreased proliferation (Fig. 3F), although the drugs failed to induce significant increases in cell death (Fig. 3E). The lack of fibroblast cell death due to anti-cancer drugs highlights the challenges of drug delivery to the pancreas; and, the organoid-OMI drug screen provides a novel platform to investigate new stromal-targeting drugs.

Organoid drug response matched in vivo drug response of PDAC for a sampling of drugs. Previous studies have found a significant reduction (p<0.05) in tumor volume and increased survival was observed in PKT mice treated with the combination treatment, AZD1480+gemcitabine, compared to control mice or monotherapy mice⁸. For both Type 1 and Type 2 organoids, the OMI index had greater reductions due to combined A1+G treatment than A1 or G alone (Fig. 2A, C), at all time points, in agreement with in vivo data⁸ indicating PKT tumors are more responsive to A1+G treatment than the monotherapies. Furthermore, these previous studies⁸ observed enhanced drug delivery in AZD1480 monotherapy and A1+G treated PKT tumors. Correlative analysis of the OMI organoid data indicates a significant response of fibroblasts to AZD1480 monotherapy and the combination A1+G (Fig. 2D), in agreement with the in vivo PKT results⁸, and suggesting OMI measured response of fibroblasts can be used to predict drug delivery of anti-cancer drugs. Altogether, these results and previously published results in breast cancer^3,4, confirm that OMI measures of organoid drug response provide a robust surrogate measure of in vivo drug response.

Finally, organoids were derived from a human PDAC biopsy (Fig. 5). These results demonstrate that organoids can be grown from primary human PDAC tumors, these organoids exhibit metabolic changes due to anti-cancer drug treatment, and an organoid-OMI screen can be used to determine drug response with various agents over a short time course (24h).

Organoids are an attractive platform for testing drugs on patient-specific tumor tissue for individualized treatment planning and novel drug development. In organoids, malignant epithelial cells can be grown in concert with the tumor-associated fibroblasts and immune cells, which provides a relevant platform to study cell-cell interactions and drug effects based on the entire tumor microenvironment. This is particularly important for PDAC as drug efficacy can be impaired by the tumor stroma. Here, we present novel findings of differential drug responses in different subpopulations of tumor cells and in tumor-associated fibroblasts within the same cultures. Additionally, we demonstrate a novel imaging technique, OMI, which is highly sensitive to detecting early, drug induced changes in cellular metabolism within intact, living samples. We have also shown that these changes in cellular metabolism correlate with changes in cell proliferation and cell death, indicating that cellular metabolism can be a surrogate measure of drug response. These findings have significant implications for rapidly assessing therapeutic response in patient pancreatic tumors.