|Home | About | Journals | Submit | Contact Us | Français|
Observing drug responses in the tumor microenvironment in vivo can be technically challenging. As a result, cellular responses to molecularly targeted cancer drugs are often studied in cell culture, which does not accurately represent the behavior of cancer cells growing in vivo. Using high resolution microscopy and fluorescently labeled genetic reporters for apoptosis, we developed an approach to visualize drug-induced cell death at single cell resolution in vivo. Stable expression of the mitochondrial intermembrane protein IMS-RP was established in human breast and pancreatic cancer cells. Automated image analysis was then used to quantify release of IMS-RP into the cytoplasm upon apoptosis and irreversible mitochondrial permeabilization. Both breast and pancreatic cancer cells showed higher basal apoptotic rates in vivo than in culture. To study drug-induced apoptosis, we exposed tumor cells to navitoclax (ABT-263), an inhibitor of Bcl-2, Bcl-xL, and Bcl-w, both in vitro and in vivo. Although the tumors responded to Bcl-2 inhibition in vivo, inducing apoptosis in around 20% of cancer cells, the observed response was much higher in cell culture. Together, our findings demonstrate an imaging technique that can be used to directly visualize cell death within the tumor microenvironment in response to drug treatment.
A major obstacle in the development and monitoring of cancer treatments is determining how multiple biological factors interact to control tumor responses to therapeutic drugs. In vivo imaging represents a powerful, non-invasive tool for revealing cellular responses through time and space. A prevailing challenge, however, has been understanding and quantitating the responses of both individual cancer cells as well as whole tumors to therapeutic drugs; namely, determining the distribution of drugs to cells, the extent and timing of cell death, and the impact of these cell-level responses on tumor regression. Recent advances in intravital microscopy have now enabled high resolution in vivo imaging at the single cell level. This technology potentially not only allows measurement of drug distribution and uptake into tumor cells (pharmacokinetics, PK), but also permits evaluation of downstream responses over different time-scales (pharmacodynamics, PD). To visualize and measure in vitro drug responses in cells, a variety of fluorescent protein (FP) reporters have been developed; these include: caspase cleavable FPs (which can function either as fluorescence resonance energy transfer (FRET) or turn-on probes), mitochondrial outer membrane permeabilization (MOMP) probes, senescence markers, and autophagy probes. The majority of these reporters, however, have only been tested in model cell lines and thus often fail to adequately reflect the primary tumors on which specific drugs are being tested in trials. Moreover, upon conversion or cell death, many of these reporters exhibit only subtle changes in fluorescence, which tend to be more difficult to measure in vivo due to motion artefacts, and (often) lower signal-to-noise ratios. In contrast, injectable imaging probes (Annexin V, synaptotagmin) have been used primarily for macroscopic imaging without clear microscopic correlates[6-8]. There is thus a clear need for robust reporters that can consistently and reliably detect cell death in common tumor models undergoing treatment with novel targeted drugs.
To address this need, we reasoned that structural analysis of apoptotic tumor cells should be feasible in live animals provided that the resolution is sufficiently high to structurally resolve mitochondria in the cells of live mice. Using a previously described construct[2, 9] and cloning it into a lentiviral vector for greater efficiency, we created breast (MDA-MB-231) and pancreatic (PANC-1) cancer cell lines that stably express the intermembrane space reporter protein (IMS-RP GFP or mCherry). In steady-state cells, mitochondria are seen to contain IMS-RP. Upon apoptosis and irreversible mitochondrial permeabilization, IMS-RP is rapidly released from the spatially conserved intermembrane mitochondrial space into the cytoplasm, where it can be detected and quantified using automated image analysis or visual inspection. Using these cancer models, we not only demonstrate that apoptosis can indeed be visualized in the tumors of live animals, but that apoptotic rates are markedly different in vitro versus in vivo (both temporally and fractionally). We also show that navitoclax monotherapy induces apoptosis in vivo in both breast and pancreatic cancer.
MDA-MB-231 human mammary carcinoma cells (ATCC) were grown in RPMI media supplemented with 10% fetal bovine serum (FBS), 1% L-glutamine, and 1% penicillin/streptomycin (Mediatech). PANC-1 human pancreatic adenocarcinoma (ATCC) cells were grown in Dulbecco’s Modified Eagle Medium (DMEM) supplemented with 10% FBS, 1% L-glutamine, 1% penicillin/streptomycin, and 2% sodium bicarbonate (Lonza). Cells expressing fluorescent reporters were grown in selection medium containing either 3 μg/mL puromycin or 10 μg/mL blasticidin S (Invitrogen). All cells were grown at 37°C and 5% CO2.
Mitochondrial intermembrane space reporter protein (IMS-RP) is an established reporter of mitochondrial outer membrane permeabilization (MOMP); it contains the first 59 amino acids of Smac/DIABLO fused to a fluorescent protein. MOMP is a commitment step during apoptosis, and IMS-RP enables the visualization of this process at single-cell resolution in vitro. pBABE IMS-RP and pBOS H2B-GFP vectors were generously provided by Drs. Peter Sorger and Timothy Mitchison, respectively. IMS-RP and histone H2B were subcloned into the pLVX-AcGFP-N1 or pLVX-mCherry-N1 vectors using In-Fusion cloning (Clontech). Plasmids were verified by DNA sequencing (Genewiz).
For lentivirus production, LentiX-293T cells (Clontech) transfected with pLVX-IMS-RP or pLVX-H2B and Lenti-X HTX Packaging Mix containing lentiviral packaging genes, using the Xfect transfection reagent and following the protocol recommended by the manufacturer (Clontech). Approximately 56-72 hours after transfection, the supernatant was collected, filtered using a 0.45 μm cellulose acetate syringe filter, aliquoted, and then snap frozen in liquid nitrogen for storage at -80°C. All viral supernatants were used immediately upon thawing.
Cells were transduced in 12-well plates using several dilutions of viral supernatant and 4 μg/mL polybrene per well (Invitrogen). Twenty four to forty eight hours following infection, cells were trypsinized and placed into medium containing 3 μg/mL puromycin. When generating double reporter cell lines, double-positive cells were sorted using the FACS Aria II (Becton Dickinson).
To determine the kinetics of apoptosis following drug treatment of cell lines in vitro, we used time-lapse microscopy; this enabled us to visualize apoptosis of single cells, within a population of cells. In vitro time-lapse microscopy was carried out using a DeltaVision imaging system (Applied Precision), which consists of an environmental chamber heated to 37°C, with CO2 bubbled through a water bath, an automated Olympus IX70 inverted microscope and a CoolSnap HQ2 CCD camera. Cells were grown and imaged in ibidi 96-well plates (ibidi). Fields of view were chosen based on the equivalence of fluorescence intensity across the cell population (to avoid under/over-exposure of cells) and on adequate cell density. Once fields of view were chosen, cells were treated with a) 25 or 100 μM ABT-263 (Selleck Chemicals) in 0.5 or 2% DMSO, or b) vehicle only, and time-lapse imaging was initiated. Post-acquisition images were visualized using Fiji/ImageJ 1.45r (National Institutes of Health) and analyzed for markers of apoptosis (MOMP, rounding, blebbing, chromatin condensation) either by eye or using MATLAB (MathWorks).
Nu/nu mice (Cox-7, Massachusetts General Hospital) were housed and handled according to Massachusetts General Hospital Institutional Animal Care and Use Committee guidelines. For procedures involving window chamber and tumor implantation, intravenous injection, and intravital imaging, mice were anesthetized by isoflurane inhalation at a flow rate of 2 L/minute isoflurane : 2 L/minute oxygen. For extended anesthesia, the isoflurane flow rate was gradually reduced to 1-2 L/minute. All surgical procedures were performed in a sterile surgery room using a stereomicroscope. For mice implanted with window chambers, antibiotics were administered via the drinking water. Anesthetized mice were monitored at all times. Titanium dorsal skin fold chambers (APJ Trading Co, Inc.) were surgically implanted in the dorsal skin fold of Nu/Nu mice, as described previously. Tumor cells (1:1 mixture of reporter to non-fluorescent parent cell) in a 1:4 mixture with Matrigel (Becton Dickinson) were injected subcutaneously into the window of the chamber, and allowed to grow for approximately one week. The skin was removed and replaced with a 10 mm-diameter glass coverslip, secured in place with a C-clip. For intravital experiments, mice were treated with 100 mg/kg ABT-263 (12.5% DMSO/PEG-400, 100 μL total volume) i.p. once per day for 2 days.
Anesthetized mice were positioned on a stable platform heated to 37°C with the dorsal window chamber mounted coverslip side facing up. Fluorescein isothiocyanate (FITC)-dextran (Sigma Aldrich), Angiosense 680/750 (Perkin Elmer, Inc), Annexin-Vivo 750 (Perkin Elmer, Inc.), or CLIO VT680 were administered via tail-vein catheter.
Tumors were imaged in vivo using a customized Olympus FV1000 BX61-W1 (Olympus America) confocal/multiphoton microscope with a programmable, motorized X-Y-Z stage. The objectives used were 2× /340 XLFluor (NA=0.14, air), 20× UPlanFL (NA=0.50, water), and 60× LUMFL N (NA=1.10, water). AcGFP1 and mCherry were excited using the 488nm line of an argon ion laser and a 559nm diode laser, respectively. Emitted light was separated using a 405/488/559/635nm dichroic beam splitter and the signal refined by the excitation band pass filters BA505-540 and BA575-605, respectively.
Annexin Vivo was visualized using the Olympus FV1000 and an Olympus IV100 fixed pinhole confocal microscope (Olympus America). The objectives used were the same as those for the FV1000 BX61- W1, and the lasers used to excite H2B-GFP and Annexin Vivo were 488/750nm, respectively. The emitted light was then filtered using a 560/640/750 dichroic mirror and refined with BA505-540 (GFP) and BA770IF (Annexin Vivo) longpass filters. Each of the signals was captured separately to minimize bleedthrough of fluorescent signals into neighboring channels. All images were stored and analyzed as 512 × 512 pixel (0.46mm × 0.46mm) 16-bit images.
Microscopy images from in vivo or in vitro experiments were processed using collaborative digital filtering, and alpha-rooting to simultaneously reduce noise and enhance image details. More details are reported in the supplemental section.
Analysis of time-lapse images was performed using MATLAB (Mathworks) and Fiji/ImageJ 1.45r (National Institutes of Health). Only unprocessed images were used for image analysis. Image analysis for automated MOMP detection was done on individual frames of time-lapse image stacks. Cell detection and segmentation were initially performed using the brightfield images, from which segmentation masks around cells of interest could be created. Regions of interest (ROI) were then automatically selected within the cells, excluding areas of high intensity signal (from mitochondria), to detect subtle changes in cytosolic intensity. Similarly sized ROIs were also selected from the background. Median intensity values for both cytosol and background ROIs were determined and normalized by their areas; background signal was then subtracted from the cytosol signal. From the plotted data, a sharp change in slope is clearly seen over time; this change corresponds to the point at which MOMP occurs.
IMS reporter protein (IMS-RP) is a genetic reporter of apoptosis that was created by fusing an FP to the first 59 N-terminal amino acids of the mitochondrial protein Smac/DIABLO[2, 9]. During apoptosis-associated MOMP, the FP is released from the IMS into the cytosol. While IMS-RP contains the mitochondrial targeting signal peptide, it does not contain the inhibitor of apoptosis protein (IAP) binding motif and is therefore biologically neutral. In cells transduced with IMS-RP green fluorescent protein (GFP), fluorescence was observed within the mitochondria (Figs. 1 and and2),2), co-localizing with both full-length endogenous Smac/DIABLO as well as with apoptosis inducing factor (AIF), another protein that resides within the mitochondrial IMS (Fig. S1). MDA-MB-231 cells typically showed filamentous mitochondria, which localized most abundantly around the nucleus (Fig. 1A). In contrast, PANC-1 cells typically contained shorter and more rounded mitochondria that were scattered throughout the entire cell (Fig. 2A). These mitochondrial phenotypes were consistent throughout all of our experiments, suggesting that the mitochondria of MDA-MB-231 and PANC-1 cells are morphologically distinct. In addition, IMS-RP fluorescence is brighter in the MDA-MB-231 cells, regardless of whether cells are transfected or transduced, which could influence the resultant images. Occasionally, round “swollen” mitochondria were seen in both cell lines, but the presence of these mitochondria did not correlate with apoptosis. In fact, recent findings suggest ill correlation between mitochondrial morphology and apoptosis, such that swollen or fragmented mitochondria do not predict apoptosis.
In cell culture, apoptosis occurred in ~0.6 % of MDA-MB-231 cells and in ~1.4 % of PANC-1 cells. Fig. 1B summarizes a typical and representative time series both before and after MOMP. The morphological changes (disappearance of distinct intracellular foci/appearance of diffuse cytosolic fluorescence) were readily visible, and were used to identify MOMP via image analysis algorithm (Figs. 1B and and2B).2B). The sequence of events leading up to irreversible cell death (MOMP, rounding, blebbing, nuclear fragmentation) was defined for both cell lines. However, as previously observed in HeLa cells, biological variability was prevalent throughout the cell population. To better compare time scales of events and to provide statistics, we quantified these cell processes in real-time. For MDA-MB-231, ABT-263-induced MOMP at 135±85 minutes (mean ±SD), rounding occurred at 61±83 minutes, blebbing at 134±77 minutes, and chromatin condensation at 173±67 minutes; for PANC-1 cells, MOMP occurred at 363±37 minutes, rounding at 291±131 minutes, and blebbing at 386±83 minutes.
While image analysis of sparsely plated cells in culture was relatively simple, in vivo analysis proved more complex due to cellular crowding. To enable easier identification of individual cells, a second fluorescent marker was introduced. For this, we selected H2B-FP, a nuclear histone protein, which when fused to an FP can be used to identify chromatin condensation, a later step in the apoptotic pathway. We subsequently determined whether these tumors could grow in vivo, and if so how their growth and apoptotic rates compared to those in the in vitro environment.
Both MDA-MB-231-IMS/H2B and PANC-1-IMS/H2B cells grew well in xenografts, demonstrating similar growth characteristics to the parental cell lines (Fig. 3). Both cell lines also grew readily in window chambers. At 60x magnification, individual mitochondria could be easily discerned within individual cells in vivo (Fig. 3). We were thus able to verify that mitochondria have a similar appearance in vivo as in vitro. Furthermore, cells undergoing apoptosis in vivo could be easily identified using the dual fluorescent marker. One significant difference between our in vitro and in vivo measurements, however, was in the basal apoptosis rates of cells. For both cell lines tested, we observed higher apoptosis rates in vivo than in vitro (MDA-MB-231-IMS/H2B: 0.6% in vitro versus 11.6% in vivo; PANC-1-IMS/H2B: 1.4% in vitro versus 8.6% in vivo). These differences could be explained by the presence of selective pressures in the tumor microenvironment, which are not encountered in culture conditions.
The availability and use of both MOMP and chromatin condensation/fragmentation reporters was useful not only for identifying individual cells but also for temporarily assessing the apoptotic process (Fig. 4). Under basal state conditions (~10% of cells in apoptosis), 35% of apoptotic cells were MOMP positive only, 25% only showed chromatin condensation, and 40% underwent both MOMP and chromatin condensation. In drug treated tumors (~20% of cells in apoptosis), apoptotic cells were seen to shift slightly in favor of chromatin condensation states.
We next determined whether drug treatment effects could be monitored directly in vivo at the single cell level (“single cell pharmacodynamics”). To do this, we treated established tumors with the Bcl-2 inhibitor, navitoclax (ABT-263), a drug currently undergoing clinical trials both as a combination treatment (with Docetaxel, Paclitaxel, Erlotinib, Rituximab, Gemcitabine, among others) and as a monotherapy. In parallel, and as a comparison, we performed similar experiments in cell culture. We found that while a single dose of navitoclax (100 μM) in vitro induced apoptosis in nearly all MDA-MB-231-IMS/H2B cells within a 5 hour period, it only induced apoptosis in ~20% of PANC-1-IMS/H2B cells under the same conditions (Fig. 5). These results are consistent with earlier findings, which showed that while EC50 values for navitoclax ranged from 0.11-39 μM in small cell lung cancer and hematologic cancer lines, most other solid tumors are not as responsive to navitoclax treatment alone[14, 15].
When tested in vivo, drug induced apoptosis was readily identifiable. In order to approximately match the drug dose encountered by cells in vitro, 100 mg/kg of navitoclax was injected intraperitoneally (i.p.) into mice. Previous studies have determined this dose to be efficacious without inducing major thrombocytopenia[14, 17]. While the pancreatic and breast cancer cell lines showed divergent behavior in vitro, apoptosis rates of the cell lines in vivo were remarkably similar. Furthermore, apoptosis rates in vivo plateaued at ~20%, a 100% increase from basal levels. Our studies therefore indicate that both tumor types respond to navitoclax monotherapy in vivo. Apoptotic cells did not demonstrate any preferential distribution, either in their distance from the tumor microvessels or within the tumors themselves.
Drug induced apoptosis is often measured via non-genetic means using fluorescent Annexin V or other affinity ligands (e.g. NucView). While the Annexin V approach is particularly well established in vitro, its validity in in vivo settings has not yet been explored at the single cell level. We resolved to determine whether systemic injection of Annexin V in mice would result in its accumulation in apoptotic cells (defined by MOMP and chromosome condensation). Interestingly, we observed that nearly all fluorescently labeled Annexin V co-localized with tumor associated macrophages (TAM) outside of the tumor rather than with apoptotic tumor cells (Fig. 6). Indeed, TUNEL-stained apoptotic cells failed to show significant colocalization with Annexin V in vivo (Fig. S3). We hypothesize that dying cancer cells recruit monocytes to the tumor site, where they differentiate into TAM, and result in both cell types accumulating the imaging agent. Assessment of TAM may therefore represent an indirect means for identifying cell death for some tumor types, reporting on the macrophage response rather than on apoptosis directly.
The use of high spatial resolution imaging has become widespread in cell biology. However, the attainment of similar resolutions in vivo has remained challenging due to motion, limited achievable signal-to-noise ratios and restricted tissue penetration of light . In view of the findings showing that many cellular processes and responses can behave differently in vivo compared to in vitro, it is now clear that the ability to image at subcellular resolution in vivo would have important consequences. Here, we not only show that it is practically feasible to image apoptotic events within individual cells over time, but also that it is possible to measure drug induced effects in the living animal. These results are important for several reasons. Firstly, such technological advances will likely broaden the study of apoptosis and its regulation at both the single cell and population level in vivo. Secondly, the use of such a method could be applied as a tool in systems pharmacology for measuring the therapeutic effect (or failure) of drugs on individual cells, cell populations or cell niches. Thirdly, if and when combined with fluorescent drugs[19, 20], this approach could serve to improve our ability to map both the pharmacokinetics and pharmacodynamics of drugs. Similarly, the toxicity of drugs on non-malignant cells could be determined in transgenic mice that express the reporters. Finally, our findings suggest the possibility that other cellular structures and pathways could likewise be imaged at high spatial resolutions in vivo.
Several studies have used intravital microscopy to study tumor immunology [21-23], microvascularity, invasion and metastases[25-27], as well as cell cycle progression[1, 18]. Many of these studies, however, used lower resolutions to visualize cell-cell interactions or cellular movements over time. In our investigation, because our aim was to resolve intracellular details (not generally needed for cell tracking studies), we required much higher spatial resolutions. We achieved this by creating highly fluorescent dual reporter lines and then carefully optimizing the acquisition conditions. IMS-RP is a biologically neutral reporter that lacks the IAP binding motif. As a result, there are no biological side-effects from over-expression of the Smac/DIABLO peptide (as there would from expression of full length proteins fused to FPs). Lentiviral constructs provide a rapid method for labeling cells, including primary cells. A major technical challenge in intravital microscopy, however, is overcoming artifacts caused by physiological motion. In our study, to reduce motion during prolonged imaging sessions, the window chambers of isoflurane-anesthetized mice were immobilized on a heated stage, and the mice were monitored carefully throughout each experiment. Scan speeds were likewise adjusted to reduce artifacts. To obtain a three-dimensional representation of the field of view, all images were collected as z-stacks using a high magnification water-immersion objective.
Despite numerous technological advances in intravital imaging, there remain a number of unresolved practical issues. Firstly, to obtain the highest possible spatial resolution, capable of resolving mitochondria and chromatin within cells, the present study required the use of window chamber models. In future research, it would be interesting to extend this type of imaging to the study of orthotopic cancers in genetically engineered mouse models. For this to be possible, however, better immobilization strategies and optics are needed. We nevertheless remain confident that with continuing advances in miniaturized optics[28, 29], fast scanning and newer immobilization techniques, such studies will ultimately be feasible. A second area in need of improvement is in the design of reporter constructs. For example, the use of more red shifted or brighter FPs would increase the signal-to-noise ratio and thus facilitate image analysis. Alternatively, the creation and use of turn-on FP reporters that only fluoresce during apoptosis would be another interesting means of obtaining information about the signaling pathways involved in this process. Several caspase-activatable probes have recently been described for detection of cell death in vitro; but these have yet to be tested in vivo. Another area demanding significant attention is in the development of automated image analysis and quantitation, particularly for use in unsupervised paradigms[31-33]. Finally, there is an obvious need for deeper imaging into tumors. It is possible that this issue could be addressed through the use of multiphoton approaches, more red shifted constructs and perhaps newer illumination schemes.
The in vivo studies presented here not only complement earlier investigations in HeLa cells[2, 13, 35], but also serve to further validate these studies in breast and pancreatic cancer cell lines. Several important differences should nevertheless be highlighted. For example, Spencer et al. primarily used a combination of TRAIL (tumor necrosis factor-related apoptosis-inducing ligand) and cycloheximide treatment to induce cell death. In cases where TRAIL was used alone, however, it appears the researchers required five-fold greater concentrations of TRAIL to induce cell death, which is comparable to the concentration used in our study to induce cell death within a similar time frame. In addition, we also found that the duration of MOMP in both PANC-1 and MDA-MB-231 cells was similar to that seen in HeLa cells (for both, it occurred in minutes). In the present study, we used navitoclax rather than TRAIL to induce apoptosis on the basis that this drug is currently under investigation in clinical trials. Navitoclax (ABT-263) is a small molecule inhibitor of the pro-survival Bcl-2 family of proteins, which include Bcl-2, Bcl-xL, and Mcl-1. From studies of small-cell lung carcinoma and acute lymphoblastic leukemia, the reported effective dose range for navitoclax is approximately 110 nM to 39 μM. In our own experiments, while 25 μM ABT-263 was sufficient to cause cell death in approximately 50% of MDA-MB-231 cells, it did not cause significant apoptosis in PANC-1 cells. We thus increased the in vitro dose of ABT-263 to 100 μM.
In summary, we believe that by continuing to make advances in microscopic imaging and apoptotic reporter constructs, these tools could soon have broad applications for biology and drug development. We expect that the use of IMS-RP, together with other genetic reporters, will not only allow us to visualize multiple components of the apoptotic signaling pathways (which will likely differ depending on drug treatment/cell type), but will also be useful for drug efficacy profiling. Naturally, such reporters will need to be applied to the study of other forms of drug-induced cell death, e.g. by the creation of in vivo reporters for DNA damage, senescence, autophagy, and necrosis, among others.
We would like to thank Drs. Kohler, Sergeyev, Fisher-Jeffes, Thurber, Yoshiko Iwamoto, and all members of the Pittet Lab for valuable technical suggestions, review and reagents. Our special thanks go to our collaborators for their invaluable discussions: Drs. Timothy Mitchison (H2B constructs), Peter Sorger (IMS-FP constructs), Mikael Pittet, Anthony Letai, Gaudenz Danuser and Charles Lin. This work was supported in part by NIH grants: 2P50CA086355, 2R01EB006432, 1P01CA139980 and 1R01CA164448.