Apoptotic bodies in the pancreatic tumor cell culture media enable label-free drug sensitivity assessment by impedance cytometry

Abstract

The ability to rapidly and sensitively predict drug response and toxicity using in vitro models of patient-derived tumors is essential for assessing chemotherapy efficacy. Currently, drug sensitivity assessment for solid tumors relies on imaging adherent cells or by flow cytometry of cells lifted from drug-treated cultures after fluorescent staining for apoptotic markers. Sub-cellular apoptotic bodies (ABs), including micro-vesicles that are secreted into the culture media under drug treatment can potentially serve as markers for drug sensitivity, without the need to lift cells under culture. However, their stratification to quantify cell dis-assembly is challenging due to their compositional diversity, with tailored labeling strategies currently needed for the recognition and cytometry of each AB type. We show that the high frequency impedance phase versus size distribution of ABs determined by high-throughput single-particle impedance cytometry of supernatants in the media of gemcitabine-treated pancreatic tumor cultures exhibits phenotypic resemblance to lifted apoptotic cells and enables shape-based stratification within distinct size ranges, which is not possible by flow cytometry. We envision that this tool can be applied in conjunction with the appropriate pancreatic tumor microenvironment model to assess drug sensitivity and toxicity of patient-derived tumors, without the need to lift cells from cultures.

Graphical Abstract

Drug sensitivity of an in vitro gemcitabine-treated pancreatic tumor model is assessed from supernatants in culture media for shape-based quantification of secreted subcellular bodies within distinct size ranges using their high frequency impedance phase versus size distribution determined by impedance cytometry, which is not possible by flow cytometry.

1. Introduction

Apoptosis or programmed cell death¹ provides important cues to enable cell clearance by phagocytes² and mediate their communication within the broader cellular micro-environment. ³,⁴. Cellular apoptosis starts with the formation of plasma-membrane blebs on the cell surface, followed by the generation of thin membrane protrusions, including microtubule spikes, apoptopodia and beaded structures on the plasma membrane, eventually leading to cell fragmentation and distribution of cellular material into sub-cellular bodies and small microvesicles⁵. All of these sub-cellular (1 to 5 μm), membrane-bound apoptotic extracellular vesicles, with differing size, shape, and internal composition, are herein collectively termed as apoptotic bodies (ABs). The kinetics of cell disassembly at each apoptotic stage and their stratification can provide valuable information on the efficacy of drug treatments and the emergence of drug resistance in cancer cells.⁶,⁷ This is especially of relevance to pancreatic cancer (i.e. pancreatic ductal adenocarcinoma or PDAC), the fourth leading cause of cancer-related deaths, with a 5-year survival rate of < 6%.⁸ Chemotherapy is often the only treatment option, since ~80% of pancreatic cancer patients have inoperable disease at presentation.⁹ However, chemotherapy drugs exhibit a high degree of patient-to-patient variability in drug sensitivity and cytotoxicity.¹⁰ Since these drugs broadly interfere with cell replication or DNA repair pathways, rather than being targeted to interfere with particular cell receptors or proteins, their action cannot be predicted by genetic and transcriptional markers.¹¹ Furthermore, given the short timeframe available to pancreatic cancer patients, who have a median survival duration of just 3–7 months, there is a need for in vitro cell-based assays to screen pre-clinical drug targets to gauge clinical benefit by using patient derived tumor materials.¹²,¹³

Traditional proliferation assays in cell cultures or histology of tissue samples do not provide quantitative information on sub-cellular ABs, while microscopy of adherent cells can image only limited AB numbers.¹⁴ More recently, flow cytometry-based approaches that are capable of measuring statistically relevant numbers of ABs and cells under apoptotic conditions have emerged for stratification based on particle size and fluorescent identification after staining.¹⁵,¹⁶. However, since ABs have distinct sub-cellular contents that are tailored for specific functions in the microenvironment,¹⁷ it is often challenging to identify the appropriate fluorescence stain for each AB type and to minimize dependence of the measurements on differential dye penetration kinetics across each AB type. Hence, there is emerging interest in label-free methods to analyze sub-cellular ABs based on metabolic alterations using Raman spectroscopy¹⁸ and fluorescence lifetime measurements,¹⁹ or based on biophysical alterations measured using optical tomography,²⁰ conductivity²¹ and dielectrophoresis measurements.^22–24 However, these prior label-free studies lack the throughput and sensitivity for measuring the large event numbers at single-particle sensitivity, as needed for statistically relevant stratification of sub-cellular ABs.

Impedance-based flow cytometry can potentially serve as a tool to measure electrical physiology and biophysical properties (loosely termed as electrophysiology), with single-particle sensitivity, for label-free assessment of apoptotic cells and sub-cellular bodies at statistically relevant particle numbers (> 5,000 per sample type). Impedance spectra computed based on disruptions to current flow by single particles can be measured at high throughput (300 – 400 particles/s) and in a non-invasive manner over a wide range of frequencies (0.5 to 50 MHz) to obtain multiparametric sub-cellular information on electrical size (~0.5 MHz), membrane capacitance (2–5 MHz) and interior conductivity (≥10 MHz).²⁵ An early report on its application towards differentiating cells in apoptotic versus necrotic states after drug treatments was carried out using a two-electrode set-up of limited throughput and sensitivity,²⁶ with limited ability to resolve sub-cellular alterations or detect sub-cellular ABs. In more recent work on a lymphoma cell line (U937) induced to apoptosis by etoposide treatment, the impedance phase at high frequency (10 MHz) was used to infer a rise in the number of sub-cellular ABs and for identification of two distinct sub-populations.²⁷ However, kinetic measurements as a function of drug treatment time and level, to identify apoptotic cells and sub-cellular ABs along the treatment progression have not been carried out for characterizing the cell disassembly process.

In the current work, cells from a highly tumorigenic patient-derived xenograft (PDX) of pancreatic cancer²⁸ with a KRAS mutant genotype²⁹ are induced towards apoptosis by gemcitabine, which is a common chemotherapy agent. PDXs, due to their ability to preserve the histopathology, genotype, gene expression and heterogeneity of the tumor derived from a patient, can serve as clinically relevant models for studying the kinetics of drug-induced cell disassembly under apoptosis. Our novel finding is that the high frequency impedance phase versus size distribution of ABs determined by impedance cytometry of supernatants in the media of gemcitabine-treated pancreatic tumor cultures, exhibits phenotypic resemblances to lifted apoptotic cells and can be used along with dielectric shell models for size and shape-based stratification. This includes AB subpopulations of small spherical vesicles of <2.6 μm that exhibit low impedance phase (<0.3), mid-sized oblate ABs (3–8 μm) that exhibit high impedance phase (>0.5) and wider sized ABs (3–14 μm) of low impedance phase (<0.3) that can arise from spherical or prolate ABs. Similar stratification is not possible by conventional flow cytometry after Annexin V staining for phosphatidylserine expression. Stratification of ABs is especially of interest, since variations in drug sensitivity kinetics and drug resistance of pancreatic tumors are associated with the presence of particular AB types and their relative proportions.³⁰ Furthermore, since sub-cellular ABs are secreted in large numbers, even from limited numbers of parent cells, we envision impedance cytometry as a tool for assessment of ABs in culture media of the appropriate in vitro pancreatic tumor microenvironment model,³¹ to detect drug sensitivity and toxicity of patient-derived tumors to chemotherapeutic agents, without the need to lift cells from cultures, or perform laborious particle collection and staining steps. This can be particularly significant for patient-derived tumor cells that are fewer in number and have apoptotic transformations with unknown markers for staining or exhibit dependence on dye penetration kinetics.

2. Experimental Section

Patient-derived pancreatic tumor xenografts and cells:

PDAC tumor sample MAD 08–608 was generated from remnant human tumor surgical pathology specimens collected in collaboration with the University of Virginia Biorepository and Tissue Research Facility, with the approval of the University of Virginia Institutional Review Board for Health Sciences Research following written in-formed consent from each patient. Tumor was propagated orthotopically on the pancreata of immunocompromised mice (Figure 1A). As previously described,³²,³³ tumor growth characteristics were measured, genotyped, and a xenograft cell line was established. Cells were transduced with firefly luciferase lentivirus (KeraFAST), selected using puromycin and maintained in RPMI1640 with 10% FBS and 2 mm glutamine (complete medium), with fresh aliquots thawed, propagated, and used for experiments.

Figure 1.

A – From PDAC patient to drug sensitivity assay via the xenograft model. Pancreatic cancer samples collected from PDAC patients through surgical resection and propagated in immunocompromised mice as a xenograft (PDX). PDAC cell line established from PDX and maintained in culture for gemcitabine treatment. B – Schematic representation of the impedance cytometer detection region. PDAC cells and apoptotic bodies in 1×PBS flow through a microchannel which includes the two sets of detection electrodes. AC signals of varying frequency are applied to the top electrodes and the differential current at the bottom electrodes is used to calculate single-cell impedance signals. Depending on the AC signal frequency applied, different biophysical properties can be inferred as different cell components interact with the AC field. C – Exposure to gemcitabine induces apoptotic cell disassembly, with apoptotic cells generating apoptotic bodies of various phenotypes and morphology as disassembly occurs.

Drug Treatment and Cell Proliferation Assays:

PDAC cells were exposed to various doses (0.01 μg mL⁻¹, 0.1 μg mL⁻¹ and 1 μg mL⁻¹) of gemcitabine (University of Virginia clinical pharmacy) for 24 h, 48 h and/or 96 h in complete medium, depending on the specific experiment (Figure 1A). For flow and impedance cytometry, untreated control samples were kept under the same culture conditions and time periods as treated samples, for analysis at end of treatment. Cells were dissociated and processed for flow and impedance cytometry. For the proliferation assays, cells (3 × 10³) were plated in a 96-well plate in complete medium and allowed to attach overnight. Following one day of growth, the cell number was determined to initiate drug treatment and then replenished after 48 h, as needed for each experiment. Upon harvest, the CyQUANT® cell proliferation assay (Invitrogen, ThermoFisher) was used to determine the relative cell number, using a plate reader (Biotek).

Flow Cytometry:

For cell measurements, the cells were harvested using TrypLE Express (Gibco Invitrogen) and centrifuged at 300 g for 5 min. Medium was aspirated, and the cell pellet was resuspended in flow cytometry buffer (1×PBS with 1% BSA) and filtered through a 70 μm cell strainer. Samples were stained with Annexin V (AF647, Biolegend) and 4,6-diamidino-2-phenylindole (or DAPI; 1 μg mL⁻¹, Roche), and immediately analyzed. Flow cytometry was carried out at the University of Virginia Flow Cytometry Core Facility using a FACS Calibur flow cytometer (BD Biosciences). Data were exported as standard FCS files and analyzed using FCS Express (De Novo Software). For measurements of cell culture media supernatants from either untreated or gemcitabine-treated cultures after drug dose of 1 μg mL⁻¹ for 48 h, the supernatant was aspirated and centrifuged at 3000g for 20 min, and the pellet resuspended in 1×PBS. Samples were stained with Annexin V (FITC, Life Technologies), and immediately analyzed. Flow cytometry was carried out using a CytoFLEX (Beckman Coulter) and analyzed using CytExpert (Beckman Coulter).

Sample Preparation:

For impedance cytometry, cells in complete medium were aspirated, washed in 1×PBS (Thermo Fisher) and exposed to 0.05% trypsin (Thermo Fisher) for 5 min at 37 °C. Cells were resuspended in high glucose DMEM (Thermo Fisher) with 10% FBS (Thermo Fisher) and centrifuged at 300g for 10 min. DMEM was aspirated, and the cell pellet was resuspended in 1×PBS, 500 mm EDTA, and 0.5% BSA. The sample was filtered through a 40 μm cell strainer, and cells were then counted with a hemocytometer. Cells were diluted to ~2 × 10⁵ mL⁻¹ together with 7 μm reference polystyrene beads (Sigma) at ~1.2 × 10⁵ mL⁻¹. For measurements in culture media supernatants, a volume of 250 μL was aspirated from the cell culture media of either untreated or gemcitabine-treated cultures after drug dose of 1 μg mL⁻¹ for 48 h. The aspirated volume was added to 750 μL of 1×PBS, and filtered through a 5 μm cell strainer to remove any suspended cells or larger cell debris. Reference polystyrene beads (5 μm, Sigma) at ~1.2 × 10⁵ mL⁻¹ were then added.

Impedance Cytometry:

The impedance cytometry system (Figure 1B) has been reported previously (SI Section B).²⁸ Samples were introduced in the microfluidic device (detection region ~30 μm tall by ~60 μm wide for cell measurements) at a flow-rate of 60 μL min⁻¹ (neMESYS, Cetoni), with cleaning steps using 1×PBS being done between measurements. AC signals at two discrete frequencies (0.5 and 18 MHz; 2 V_pp at each frequency) were applied to the top electrodes using an impedance spectroscope (HF2IS, Zurich Instruments) and current signal at the bottom electrodes was acquired (sample-rate = 1.15 × 10⁵ samples s⁻¹) and converted using a current amplifier (HF2TA, Zurich Instruments). Lock-in amplification (HF2TA, Zurich Instruments) was used to separate the real and imaginary signal components at each frequency to compute impedance magnitude and phase (SI Section B). For cytometry on culture media supernatants, samples were introduced in the microfluidic device (detection region 30 μm × 30 μm) at a flow-rate of ~10 μL min⁻¹, with impedance measurements (Ampha Z32, Amphasys AG) at 10 MHz, with acquisition settings of modulation, amplification and demodulation levels at 5, 5 and 0, respectively. Processed signal data was then stored in the form of impedance magnitude and phase.

Data Analysis:

Impedance cytometry data was processed and analyzed using custom code written in MATLAB (R2018b, MathWorks). The impedance signal was normalized against the frequency-independent impedance response of the reference polystyrene beads by dividing the impedance data by the mean impedance data of reference beads. Due to normalization, impedance phase is herein reported in arbitrary units. Normalized impedance magnitude is used to compute the metric of electrical diameter by calculating $\sqrt[3]{{\left|\mathrm{Z}\right|}_{0.5\mathrm{M}\mathrm{H}\mathrm{z}}}$, and using the polystyrene beads for size reference. Statistical analyses (MATLAB R2018b) were performed on processed datasets. Significance level was defined at α < 0.05 for all cases. For cell measurements, two sample Students’ t-tests were performed to compare individual datasets, for each exposure time (i.e. 24 h or 48 h) and each parameter (i.e. cell population ratio, impedance phase at 18 MHz and electrical diameter), to assess statistically significant differences between treatment conditions. For supernatant measurements, a two sample Students’ t-test was used to compare the total number of detected sub-cellular events in untreated versus gemcitabine-treated media.

3. Results and Discussion

3.1. Cell proliferation and morphology under gemcitabine

Using gemcitabine treatment on highly tumorigenic PDAC cells of KRAS mutant genotype (T608) derived from a PDX model, the range of gemcitabine levels that lead to drug sensitivity were assessed by cell proliferation assays, on cell cultures exposed to different concentrations of gemcitabine (0.01, 0.1 and 1 μg mL⁻¹) for three different time periods (24 h, 48 h and 96 h). The untreated control sample was maintained in the complete medium for the same durations. Based on a fluorescent DNA intercalating dye to measure cell number, the percentage of maximum cell proliferation is calculated to determine proliferation of treated cells relative to the untreated controls (Figure 2A). The drop-off in cell proliferation with increasing gemcitabine levels confirms its chemotherapeutic effect on this PDAC cell line. While proliferation levels remain positive (declining to ~50% levels) after 24 h of drug treatment, they steeply fall to ~ −50% after 48 h and 96 h of drug treatment at 0.1 and 1 μg mL⁻¹ levels. However, for a gemcitabine concentration of 0.01 μg mL⁻¹, the proliferation is barely affected at the shorter exposure time-points (24 h and 48 h).

Figure 2.

A – Proliferation studies on the PDAC cell line. Cell cultures were exposed to varying concentrations of gemcitabine (0.01, 0.1 and 1 μg mL⁻¹) for: 24 h (circle), 48 h (square) and 96 h (triangle). Proliferation (%) is calculated as the relative proliferation under each treated condition compared with untreated for each exposure period and gemcitabine concentration. B – Microscopy images of PDAC cells for the various gemcitabine concentrations, for a 48 h exposure, with key apoptosis features highlighted (63× obj., 2.5× optovar). Flow cytometry assessment of gemcitabine-induced apoptosis on the PDAC cell line. C – Density scatter plots of Annexin V (AV) versus DAPI for gated cells (i-iv) and sub-cellular bodies (v-viii) from samples at varying gemcitabine treatment for 48 h. Sub-population ratios for gated AV−DAPI−, AV+DAPI− and DAPI+ events from gated cells (D) and sub-cellular (E) sub-populations for all drug treatment conditions. Arrows pointing to data clusters in Cvii,viii indicate the differing sub-cellular AB subtypes.

To assess the effects of gemcitabine on PDAC cell morphology, microscopy images (63× obj., 2.5× optovar, Zeiss Observer 7 microscope) taken at various drug concentrations are shown at the 48 h time-point (Figure 2B). Comparison of the phenotypes over the measured gemcitabine concentration range shows cells presenting the hallmarks of apoptosis, including the generation of blebs of varying size on the plasma membrane (at 0.01 μg mL⁻¹), the presence of large protrusions (at 0.1 μg mL⁻¹) and the emergence of beaded structures (at 1 μg mL⁻¹). These structures are associated with apoptosis-induced cell disassembly, which generates ABs of particular composition and structure. In prior work,⁵,⁶ specific stains were used to identify phenotypes of the ABs per Figure 1C as: (i) apoptotic microvesicles released by intact cells; (ii) beaded apoptopodia that form at the cell surface under progressive apoptosis; and (iii) larger apoptotic bodies that indicate cell disassembly. In the subsequent sections, we seek to stratify ABs into these phenotypes using flow cytometry and impedance cytometry methods based on their single-particle sensitivity levels.

3.2. Flow cytometry of gemcitabine-induced apoptosis

PDAC cells treated with gemcitabine (0.01–1 μg mL⁻¹) for 24 h and 48 h were measured by flow cytometry. Plots of forward scatter (FCS) (related to particle size) versus side scatter (SSC) (related to particle “granularity” or internal complexity) shown at 48 h time point (Figure S1A; 24 h time-point in Figure S2A) indicate three distinct data clusters of cell-sized particles, sub-cellular sized particles and cell debris of smallest size. The sub-cellular sized population includes ABs generated due to cell death and due to cell degradation from sample preparation. From their relative proportions (Figure S1B), the cell-sized population forms ~40% of all events, except for samples exposed for 48 h to the higher gemcitabine concentrations (0.1 and 1 μg mL⁻¹), wherein the cell proportion is reduced by half. This observed alteration occurs at the same drug concentration and exposure time at which the proliferation studies (Figure 2A) indicated a substantial decrease in cell proliferation due to the gemcitabine-induced apoptosis.

The same set of samples was measured by flow cytometry after staining with Annexin V (AV) to measure expression of phosphatidylserine (PS) as the apoptotic marker³⁴ and with 4,6-diamidino-2-phenylindole (DAPI) for DNA-based cell viability determination based on membrane function. Scatter plots of AV and DAPI staining for cell-sized and sub-cellular-sized events after 48 h and 24 h of gemcitabine exposure are in Figure 2C and Figure S2B, respectively. Considering the cell-sized events, three sub-populations are expected: AV− DAPI− representing viable cells with no apoptotic markers; AV+ DAPI− representing cells in the early to mid-apoptotic phase that possess an intact membrane; and DAPI+ representing non-viable cells with a permeabilized membrane. However, the identification of cells as simply DAPI+ is a limiting factor in this flow cytometry protocol, since the specific mechanism causing cell death (e.g. necrosis or apoptosis) cannot be discerned, given that AV will be highly expressed by both apoptotic cells (PS exposed on the membrane outer leaflet) and other non-viable cells (PS naturally present on the membrane inner leaflet).

The scatter plots and gating ratios for each sample type and sub-population make two key features evident (Figure 2Ci–iv,D). First, the untreated control presents a relevant percentage of non-viable DAPI+ cells (~40%) at the end of the treatment time-frame (48h), and this percentage remains constant through most treatment conditions. This is attributed to the PDX cell line, which are low passage, non-immortalized cells derived from a primary patient xenograft. This naturally renders them to be much more sensitive to the cell culture conditions (e.g. 48h without media exchange or re-seeding), which accelerates loss of viability along the treatment time-frame. However, since there is virtually no AV+ DAPI− sub-population within the untreated sample, the onset of apoptosis in the studied treatment conditions can still be performed.

Second, the ratios of DAPI+ cells are reduced at the high dose gemcitabine treatment conditions (0.1 and 1 μg mL⁻¹ at 48 h), wherein the ratios of AV+ DAPI− cells (i.e. early/mid apoptotic cells) increase substantially (Figure 2D). Moreover, the AV staining of DAPI+ cells and even AV− DAPI− cells (which, interestingly, cannot be easily discerned, based on AV staining, from cells that are gated as AV+DAPI−; Figure 2Ciii,iv) also increases. This, together with the clear presence of AV+ DAPI− cells, indicates the onset of drug-induced apoptosis under these treatment conditions. Hence, in comparison to the untreated control, for example, the ratios of DAPI+ cells at these high dose treatments show a net reduction, since cells that would have naturally be rendered non-viable under no drug treatment are now further accelerated towards non-viability at earlier time-points and are effectively removed from the adherent cell populations measured with these experiments. Hence, flow cytometry suggests that cells remain largely viable and non-apoptotic at all drug doses at the 24 h time-point (Figure S2), while showing a higher proportion of apoptotic events at the 48 h time-point, only at the 0.1 and 1 μg mL⁻¹ drug dose conditions (Figure 2Ci–iv). However, non-specific staining of PS leads to challenges in discerning the cell death mechanism.

The gated sub-cellular events are analyzed based on similar scatter plots (Figure 2Cv–viii) and ratios (Figure 2E). For low gemcitabine concentrations (Figure 2Cv,vi) or shorter drug exposure (24 h; Figure S2), the majority of events have low AV staining and high DAPI staining. This DAPI+ sub-population could be related to the larger cell debris that contain the remains of nucleus or other DNA content from non-viable cells. There is also a smaller sub-population of AV− DAPI− sub-cellular bodies, indicating the presence of a different type of cell debris with no relevant DNA or PS quantities. However, for samples treated with 0.1 and 1 μg mL⁻¹ for 48 h (Figure 2Cvii,viii), there is a sharp increase in the ratio of particles with the AV+ DAPI− phenotype, with their proportion growing from ~10–30% to ~60% of all sub-cellular events (Figure 2E). Due to the specific combination of AV+ and DAPI− staining, we infer that these sub-cellular particles likely correspond to larger ABs generated during apoptotic cell disassembly. Furthermore, based on the different levels of AV and DAPI staining in that sub-population and the shift in AV staining within the DAPI+ gated particles, as is apparent from the distinct data clustering (arrows in Figure 2Cvii,viii), we suggest the presence of ABs of differing phenotypes (per Figure 1C) that likely present different genetic content and membrane conformations. However, distinguishing these AB phenotypes by flow cytometry requires specialized staining protocols that could be strongly dependent on dye penetration kinetics, which leads us to explore alternate label-free methods.

3.3. Impedance cytometry of apoptotic events

The electrophysiology of PDAC cells lifted from the culture after various gemcitabine treatment conditions was measured by impedance cytometry, to determine impedance magnitude (|Z|) and impedance phase (ϕZ) at a low (0.5 MHz) and a high frequency (18 MHz) for analyzing individual cellular and/or sub-cellular events (Figure 3). The lipid cell membrane screens the field at low frequencies in high conductivity media (e.g. 1×PBS) to cause insulator-like behavior, which can be used to estimate their electrical diameter from |Z| at 0.5 MHz, as given by: $\sqrt[3]{{\left|\mathrm{Z}\right|}_{0.5\mathrm{M}\mathrm{H}\mathrm{z}}}$. ³⁵ At increasing frequencies, capacitive coupling across the cell membrane renders cells to become more conductive. At a high enough frequency (e.g. 18 MHz), the impedance signal is effectively determined by the dielectric properties of the cell interior,³⁵ rather than the cell exterior. Hence, the ϕZ_{18 MHz} for biological particles with a conductive interior will differ from that of co-flowing polystyrene beads of standard size (7 μm), as computed from |Z|_{0.5 MHz}. This allows for normalization of impedance signals to account for any temporal variations during the measurement and to enable quantitative comparison between measurements. This feature also allows for the measurement of each event based on its size (using |Z|_{0.5 MHz}) and its interior electrophysiology (using ϕZ_{18 MHz}) to analyze the phenotypes in a label-free manner. Figure 3 shows density scatter plots of single-cell data plotted based on this metric of ϕZ_{18 MHz} versus electrical diameter, with exemplary untreated versus treated doses (0.01, 0.1 and 1 μg mL⁻¹) at the 48 h time-point (Figure 3A–D; Figure S3 presents data for the 24 h exposure period). The data are classified at each treatment condition using size-based gates to compare the proportions of cellular and sub-cellular events for impedance (Figure 3E) and flow cytometry (Figure 3F) data, wherein a steady drop is apparent after 48 h drug treatment. When analyzing the median cell population ratio in impedance cytometry ratio, this drop in the proportion of cellular to sub-cellular events with increasing levels of drug exposure becomes statistically significant (*p < 0.05) after 48 h of gemcitabine exposure at the 0.1 and 1 μg mL⁻¹ drug levels versus the untreated control (Figure 4A).

Figure 3.

Density scatter plots of electrical diameter (i.e. $\sqrt[3]{{\left|\mathrm{Z}\right|}_{0.5\mathrm{M}\mathrm{H}\mathrm{z}}}$) versus impedance phase at 18 MHz (i.e. ϕZ_{18 MHz}) for samples exposed to gemcitabine for 48 h at varying drug concentration: (A) untreated, (B) 0.01 μg mL⁻¹, (C) 0.1 μg mL⁻¹ and (D) 1 μg mL⁻¹ (impedance data from exemplary individual runs). Gates for the cells and the sub-cellular populations and location of reference polystyrene beads “B” are indicated. Proportions of cellular and sub-cellular events for: (E) impedance and (F) flow cytometry after size-based gating. Impedance cytometry data from n = 3 (24 h - 0.01 μg mL⁻¹; 48 h - 1 μg mL⁻¹), n = 4 (24 h - 1 μg mL⁻¹; 48 h - untreated and 0.1 μg mL⁻¹) and n = 5 (24 h - untreated and 0.1 μg mL⁻¹; 48 h - 0.01 μg mL⁻¹) biological repeats merged together (total number of gated events between 1 × 10⁵ to 2.5 × 10⁵ per drug treatment).

Figure 4.

Comparison between samples at varying gemcitabine treatments for ratio of cellular population (A) and ϕZ_{18 MHz} of cells (B) and sub-cellular (C) populations (boxplots show the median, the 25^th and 75^th percentiles, and the maximum and minimum values (non-outliers) from n = 3 (24 h - 0.01 μg mL⁻¹; 48 h - 1 μg mL⁻¹), n = 4 (24 h - 1 μg mL⁻¹; 48 h - untreated and 0.1 μg mL⁻¹) and n = 5 (24 h - untreated and 0.1 μg mL⁻¹; 48 h - 0.01 μg mL⁻¹) biological repeats (*p < 0.05). Comparison between ϕZ_{18 MHz} normalized distributions of untreated control and gemcitabine-treated samples at 1 μg mL⁻¹ drug at 24 h and 48 h exposure for cells (D) and sub-cellular (E) populations. ϕZ_{18 MHz} distributions represent exemplary data of 1 × 10⁴ cells (D) and 4 × 10⁴ sub-cellular (E) events from individual runs (total 5 × 10⁴ events per treatment condition). F – Difference between the mean impedance phase at 18 MHz ($\overline{{\varphi \mathrm{Z}}_{18\mathrm{M}\mathrm{H}\mathrm{z}}}$), of cell and sub-cellular populations, for samples at varying gemcitabine treatment conditions and the $\overline{{\varphi \mathrm{Z}}_{18\mathrm{M}\mathrm{H}\mathrm{z}}}$ of untreated controls. Data from n = 3 (24 h - 0.01 μg mL⁻¹; 48 h - 1 μg mL⁻¹), n = 4 (24 h - 1 μg mL⁻¹; 48 h - untreated and 0.1 μg mL⁻¹) and n = 5 (24 h - untreated and 0.1 μg mL⁻¹; 48 h - 0.01 μg mL⁻¹) biological repeats.

The rise in sub-cellular events is attributed to release of ABs under gemcitabine-induced apoptosis that is linked to lowering of cell viability and cell proliferation (Figure 2A). These sub-cellular events arise from a combination of cell debris and ABs, including microvesicle ABs, clusters of ABs of varying size and morphologies, and larger apoptotic bodies (> 5 μm) generated later during cell disassembly⁵. Indeed, based on electrical diameter evolution of the gated sub-cellular particles, there is an increase in the size range at the higher drug exposure times and concentrations (Figure S4A), confirming the increasing levels of beaded aggregates and larger ABs. Another key feature is the downward shift in ϕZ_{18 MHz} as a function of gemcitabine treatment dose, for the cellular and sub-cellular populations (Figure 4B–F and Figure S5), which suggests systematic apoptosis-induced alterations in electrophysiology of the cell interior. This is apparent based on the small shift in ϕZ_{18 MHz} between the untreated control and gemcitabine-treated samples at 1 μg mL⁻¹ after a 24 h exposure (Figure 4D), which is especially clear in the distribution of ϕZ_{18 MHz} for sub-cellular particles (Figure 4E); while statistically significant differences in ϕZ_{18 MHz} (*p < 0.05) are apparent for the cellular and sub-cellular populations between the untreated control and gemcitabine-treated samples at 1 μg mL⁻¹ for a 48 h exposure (Figure 4B,C). The drop in mean ϕZ with drug treatment dose suggests a decrease in interior conductivity (σ_int), as also observed for other cell lines under apoptosis using dielectrophoresis.²²,²³

This effect is likely related to an ionic efflux that causes a dramatic decrease in intracellular ions during apoptosis, primarily K⁺ and Na⁺.^36–38 This can lead to a net increase in the insulating material inside the cytoplasm to cause a drop in the overall σ_int. It is noteworthy that this impedance phase alteration is observed at lower drug doses (concentration and/or time-point) than is possible to discern with flow cytometry (Figure 4D,E versus Figure 2C,D), highlighting the greater sensitivity of impedance cytometry to apoptotic processes.

Using the difference between the mean phase value ($\overline{{\varphi \mathrm{Z}}_{18\mathrm{M}\mathrm{H}\mathrm{z}}}$) for each treatment condition of the cellular and sub-cellular populations compared to their respective untreated control sample as the baseline, a differential metric can be calculated (Figure 4F), which confirms the trend of decreasing $\overline{{\varphi \mathrm{Z}}_{18\mathrm{M}\mathrm{H}\mathrm{z}}}$ with drug treatment dose (concentration and/or time). The ϕZ_{18 MHz} distribution shifts down toward levels of 0.2–0.4, with an increasing majority of events within this region for increasing gemcitabine treatment doses (Figure 4D,E). The similarities in the trend of the ϕZ_{18 MHz} distributions of the cellular and sub-cellular populations indicate that this impedance phase metric can be used to infer phenotypic similarity between the two populations due to apoptosis. Furthermore, the sharper drops observed for the ϕZ_{18 MHz} distributions of the sub-cellular populations suggest their greater sensitivity to the onset of apoptosis. However, since the sub-cellular gate of the impedance data includes various sub-populations of the ABs,⁵ such as smaller microvesicles, beaded aggregates of ABs and larger ABs generated by cell disassembly (Figure 1C), the culture supernatant media comprising the size fraction of ABs (< 5 μm) is studied to stratify their phenotypes.

3.4. Detection of secreted apoptotic bodies in media

PDAC cell cultures under drug treatment shed ABs into the media that mediate communication within the cellular microenvironment and prompt cell clearance by phagocytes. Impedance cytometry is used to study the ABs in the supernatant of the PDAC cell culture media (size sensitivity ~ 2 μm using 30 μm × 30 μm channels³⁹), under apoptosis due to treatment with 1 μg mL⁻¹ gemcitabine for 48 h. This sample can be used to associate the onset of apoptosis with the electrophysiology phenotypes of the collected ABs from the culture media, thereby potentially enabling rapid analysis of apoptosis, without the need to lift cells or centrifuge the sample, analogous to our prior work to detect sub-micron scale aggregates from bacterial cultures.³⁹ To optimize the signal-to-noise ratio for single-particle detection, a single frequency of 10 MHz was used for the impedance phase (ϕZ_{10 MHz}) and electrical size ($\sqrt[3]{{\left|\mathrm{Z}\right|}_{10\mathrm{M}\mathrm{H}\mathrm{z}}}$) measurements. In this manner, we can discern differences in interior electrophysiology, while enabling size normalization using co-flowing 5 μm polystyrene reference beads. We estimate that electrical diameter determination at 10 MHz rather than at 0.5 MHz, introduces < 10% error levels in size computation (Figure S6). Based on this, the 3D histogram plots of ϕZ_{10 MHz} versus electrical diameter show a clear difference in the number of detected events between the untreated (Figure 5A) and treated (Figure 5B) conditions, with a significant difference (**p < 0.01) being found between the two media types (Figure 5C). Another striking feature is the presence of three distinct subpopulations: small vesicles of low impedance phase (<0.3), mid-sized sub-cellular bodies of high impedance phase (>0.5) and wider sized sub-cellular bodies of low impedance phase (<0.3). Due to this clustering, the median electrical diameter of all sub-cellular events (~2.6 μm) was used as a size threshold for further data analysis and fitting. The plots of flow cytometry data (Figure S7) that are analogous to Figure 5A–C show that while differences in number of apoptotic bodies are apparent for the supernatants from the untreated versus apoptosis causing drug-treated cultures, a similar correlation to size, shape and compositional distributions is not possible with flow cytometry based on the metrics of FSC, SSC or Annexin V staining. The Annexin V stained data shows clustering of a smaller-sized and a larger sized subpopulation (Figure 5D,E), with the majority of events being found in the smaller sub-population, similar to our findings with impedance cytometry (events <2.6μm in Figure 5B). However, the larger size range events in flow cytometry seem to cluster within a single subpopulation, without clear differences in size or other markers, such as Annexin V staining (Figure S7B); whereas the impedance phase distribution exhibits clustering of ABs into two distinct subpopulations (events >2.6μm in Figure 5B). Finally, a significant difference in the total number of sub-cellular events is also observed for flow cytometry data (Figure 5F), confirming again the results obtained by impedance cytometry (Figure 5C).

Figure 5.

Impedance and flow cytometry of culture media supernatants from the PDAC cell line. Density 3D distributions of electrical diameter versus impedance phase at 10 MHz (ϕZ_{10 MHz}) for untreated (A) and 1 μg mL⁻¹ gemcitabine treated (B) cell culture media after 48 h exposure (impedance data from exemplary individual runs). Dotted lines delineate the median electrical diameter of all sub-cellular events (~2.6 μm). Density plots of FSC versus SSC for untreated (D) and 1 μg mL-1 gemcitabine treated (E) cell culture media after 48 h exposure (flow cytometry data from exemplary individual runs). Main clusters of detected events are highlighted in B and E (colored ellipses). Comparison between the number of detected sub-cellular events between untreated and gemcitabine treated media for impedance (C) and flow (F) cytometry (n = 3 biological repeats; **p < 0.01).

To stratify the phenotypes of the ABs in the supernatant, we identified the peak positions of the impedance phase versus electrical size distribution, by fitting a single Gaussian distribution for the <2.6 μm events (Figure 6A), while the >2.6 μm events were fit using two Gaussian distributions (Figure 6B) due to the two subpopulations apparent in Figure 5B in this size range. It is noteworthy that the ϕZ_{10 MHz} level for the <2.6 μm events occurs in the 0.2–0.4 range, which is similar to the respective downshifted levels observed previously for apoptotic cells (Figure 4D) and subcellular bodies (Figure 4E). The same is true for one subpopulation of the >2.6 μm events that exhibits a histogram position of ϕZ_{10 MHz} levels in the 0.2–0.4 range, suggesting a degree of similarity in compositional phenotype of these ABs to apoptotic processes apparent from apoptotic cells (Figure 4D) and sub-cellular bodies. There is also one subpopulation of >2.6 μm events in the supernatant exhibiting high ϕZ_{10 MHz} levels (Gaussian distribution centered at ~0.59), which is closer to respective levels of untreated viable cells (Fig. 4D) and their sub-cellular events (Fig. 4E) or to non-spherical ABs as explored later. Sub-cellular bodies are known to exhibit spherical, prolate and oblate shapes (Figure 1C), with the subtypes varying between cell lines⁵ and being influenced by drug treatments to inhibit specific molecular factors (e.g. inhibition of the plasma membrane channel pannexin 1 inhibits the formation of apoptopodia or beaded structures on Jurkat T cells).¹⁵ While these distinct shape features are challenging to discern by flow cytometry, they can potentially be stratified due to the strong influence of shape on electrical polarization, as discerned in impedance data using dielectric multi-shell model simulations. ⁴⁰,⁴¹,⁴². These represent biological particles as a series of concentric shells with well-defined dielectric properties determined by fitting to the impedance data (Figure 6C; refer to SI Section B for further details).⁴³,⁴⁴,⁴⁵ While such models have been implemented to study, for example, PDAC cell lines of different tumorigenicity,²⁸ human neural progenitor cells,⁴⁶ healthy⁴⁷ and malaria-infected erythrocytes,⁴⁸ T-lymphocytes,⁴⁹ and human pathogenic bacteria,⁵⁰ there are no reports on their application to sub-cellular bodies.

Figure 6.

Dielectric models for ABs phenotypes from gemcitabine-treated culture media. Impedance phase at 10 MHz (i.e. ϕZ_{10 MHz}) distribution (solid line) for 1 μg mL⁻¹ gemcitabine treated cell culture media after 48 h exposure for sub-cellular events smaller (A) and larger (B) than the median electrical diameter of all sub-cellular events (~2.6 μm). Gaussian models were fit to the distributions (dashed lines; A - 1 Gaussian model R² = 0. 9686, and B - 2 Gaussian model R² = 0.9852) and overlaid (shaded areas), with their mean ϕZ_{10 MHz} value highlighted. C – Dielectric model results for the variability of ϕZ_{10 MHz} based on variations in the sphericity and electrical diameter of ABs. The colormap shows the estimated ϕZ_{10 MHz} for a particle with a specific combination of size and shape.

To understand irregularly shaped ABs phenotypes, such as microvesicles, beaded apoptopodia and large oblate-like bodies that are expected to occur during apoptosis (Figure 1C), we explore the application of an ellipsoidal model (Figure S9),⁴⁰ wherein the size and sphericity (i.e. ratio of particle volume to surface area) of the ABs was varied (Figure 6C). For this modelling approach, the dielectric properties of membrane capacitance (C_mem) and internal conductivity (σ_int) were fixed (C_mem = 15.6 mF m⁻²; σ_int = 0.6 S m⁻¹) based on previous modelling results for this specific PDAC cell line.²⁸ The sphericity of ABs was then iteratively altered, with the estimated ϕZ_{10 MHz} being calculated for each specific combination of size, morphology and dielectric properties. Based on the modelled ABs phenotypes, the overall ϕZ_{10 MHz} increases with decreasing sphericity, with oblate-like ABs presenting medium size and high ϕZ_{10 MHz}, while prolate-like ABs presenting broader size distribution and lower ϕZ_{10 MHz} values. The phenotypes from the oblate-like ABs suggest that ϕZ_{10 MHz} is highly sensitive to particle shape based on its sharp rise with the loss of sphericity in the ABs. With electrical diameter varying in the 2 and 4 μm range and the sphericity varying from 1 (sphere-like) to ~0.3 (prolate-like), the ϕZ_{10 MHz} variation ranges from ~0.2 to 0.7. Hence, size and sphericity likely act together to cause the wide ϕZ_{10 MHz} variation observed for ABs in gemcitabine-treated media (Figure 5B), and indicate that these irregularly shaped oblate-like ABs (i.e. with a high surface area-to-volume ratio) make up the distinct sub-population of events at high ϕZ_{10 MHz} (Figure 5B and Figure 6B). Hence, based on the observed impedance distributions (Fig. 5B) and simulations (Fig. 6C), we infer that the low impedance phase levels (<0.3) obtained from < 2.6 μm particle events can be attributed to spherical shaped particles and those at high impedance phase levels (>0.5) obtained from > 2.6 μm particle events can be attributed to oblate-shaped particles. The third subpopulation at low impedance phase levels (<0.3) obtained from the broadly sized particle events > 2.6 μm is likely due to spherical and prolate-shaped particles, with prolate particles making up a greater proportion of the higher size events.

4. Conclusion

The vision of label-free drug sensitivity assessment based on ABs in the culture media supernatant, rather than based on cells lifted from the culture, is particularly appealing due to the elimination of laborious sample collection tasks and due to the fewer cells present in patient-derived tumors that could present apoptotic transformations with less reliable markers for staining. However, ABs present a wide diversity in size, morphological and compositional characteristics, which motivates the need to identify electrophysiology-based metrics for label-free stratification of ABs. Hence, the apoptotic transformations in the ABs obtained from the media of gemcitabine-treated PDAC cell cultures are assessed based on their phenotypic similarity to the apoptotic markers on lifted cells, as determined by single-particle techniques of flow cytometry after Annexin V labeling and of label-free impedance cytometry. Specifically, the high frequency impedance phase level of single ABs and apoptotic cells, which is strongly related to their interior conductivity, exhibits systematic drops as a function of gemcitabine dose (drug concentration and time), with a sharper drop-off for sub-cellular versus cellular events. This impedance phase alteration is observed at lower drug doses (concentration and/or time-point) than changes in apoptotic markers observed with flow cytometry, highlighting the greater sensitivity of impedance cytometry. While drug sensitivity can be predicted based on the rise in numbers of ABs in the culture media below a size cut-off (< 5 μm), the shape of the ABs within distinct size ranges can be stratified based on their impedance phase levels using multi-shell dielectric models. Based on fitting the high frequency impedance phase distribution to three distinct AB subtypes, we infer that the events at low impedance phase levels (<0.3) obtained from < 2.6 μm particles can be attributed to spherical-shaped particles and the events at high impedance phase levels (>0.5) obtained from > 2.6 μm particles can be attributed to oblate-shaped particles, while the third subpopulation of low impedance phase levels (<0.3) obtained from the broadly sized particle events > 2.6 μm is attributed to arise from spherical and prolate-shaped particles. Similar classification was not possible by flow cytometry of the culture media supernatants. This stratification will be studied in future work to follow ABs using different tumor microenvironment models and drug types, to identify phenotypes capable of gauging drug sensitivity and toxicity.