Imaging Prostate Cancer Invasion with Multi-Nuclear Magnetic Resonance Methods: The Metabolic Boyden Chamber

Abstract

The physiological milieu within solid tumors can influence invasion and metastasis. To determine the impact of the physiological environment and cellular metabolism on cancer cell invasion, it is necessary to measure invasion during well-controlled modulation of the physiological environment. Recently, we demonstrated that magnetic resonance imaging can be used to monitor cancer cell invasion into a Matrigel layer [Artemov D, Pilatus U, Chou S, Mori N, Nelson JB, and Bhujwalla ZM (1999). Dynamics of prostate cancer cell invasion studied in vitro by NMR microscopy. Mag Res Med 42, 277–282.]. Here we have developed an invasion assay (“Metabolic Boyden Chamber”) that combines this capability with the properties of our isolated cell perfusion system. Long-term experiments can be performed to determine invasion as well as cellular metabolism under controlled environmental conditions. To characterize the assay, we performed experiments with prostate cancer cell lines preselected for different invasive characteristics. The results showed invasion into, and degradation of the Matrigel layer, by the highly invasive/metastatic line (MatLyLu), whereas no significant changes were observed for the less invasive/metastatic cell line (DU-145). With this assay, invasion and metabolism was measured dynamically, together with oxygen tensions within the cellular environment and within the Matrigel layer. Such a system can be used to identify physiological and metabolic characteristics that promote invasion, and evaluate therapeutic interventions to inhibit invasion.

Introduction

The invasion of the basement membrane by cancer cells is one of the earliest and most critical steps in the metastatic cascade. Tumor cell nutrients, oxygen, hormones, growth factors, inducers and other regulatory molecules can provide malignant cells with microenvironmental signals that could act through epigenetic cellular modifications such as DNA methylation, and transcriptional, post-transcriptional, translational and post-translational controls or combinations of these [2]. One of the least examined areas in the metastatic cascade is the role of a compromised physiological environment in promoting or triggering invasion and migration of cancer cells. Suboptimal or nutritionally depleted environments must exist in human tumors since necrotic areas are observed even in preinvasive stages such as carcinoma in situ [3,4]. Kato et al. [5] have observed that two human melanoma cell lines secreted a higher level of 90-kDa gelatinase (a type IV collagenase) at an extracellular pH of 6.8 compared to pH 7.3. This acid-induced secretion of gelatinase was blocked by cycloheximide, indicating that the enzyme induction was due to de novo synthesis. Acid-induced secretion of gelatinase increased according to the metastatic potential of mouse melanoma lines. More recently, Rozhin et al. [6] have observed that an acidic pericellular pH induced a redistribution of cathepsin B+ vesicles toward the cell periphery. For the more malignant cells, this resulted in an enhanced secretion of the active form of the lysosomal protease cathepsin B over time.

The term “Boyden Chamber Assay” is usually used to describe the classical assay for measuring cell invasion [7]. This technique, however, was based on two earlier papers, the first by Shelton and Rice [8] and the second by Prehn et al. [9]. Since then, several modifications have been proposed. The invasive potential of tumor cells is usually assayed by determining the penetration of cells into reconstituted basement membrane gel (Matrigel). Invasion is quantitated by counting the number of cells that invade the Matrigel-coated filters over a period of 5 to 72 hours [10–12]. To determine the impact of the physiological environment on cancer cell invasion, it is necessary to measure invasion during well-controlled modulation of the physiological environment. Although significant understanding of the process of cell invasion has been achieved by current assays for invasion, these methods do not permit evaluation of the metabolic state of tumor cells. Also because the cell medium is not continuously perfused over the period of observation and only endpoint measurements are performed, it is not possible to modify physiological conditions during the course of the experiment and monitor invasion dynamically during the modification.

Recently, we demonstrated that magnetic resonance (MR) imaging can be used to monitor dynamically, and with microscopic resolution, cancer cell invasion into the Matrigel layer of standard invasion chambers [1]. Here we have designed and developed an invasion assay (“Metabolic Boyden Chamber”) that combines this capability with the properties of our isolated cell perfusion system, enabling us to control environmental conditions during long-term MR experiments of perfused cells. With this system, we can measure oxygen tensions, cellular metabolism and invasion. Currently, we measure distribution of the cellular volume fraction and concentration of intracellular metabolites along the z-axis with a spatial resolution of 29.3 to 62.5 µm for cellular volume by 1D MR microscopic imaging and 156 to 310 µm for metabolites using 1D chemical shift imaging (1D-CSI). Using this assay, we performed experiments with two prostate cancer cell lines with different invasive characteristics. The results showed invasion into, and degradation of the Matrigel layer, by the highly invasive/metastatic cell line, whereas no significant changes were observed for the less invasive/metastatic cell line.

Materials and Methods

Cell Lines and Culture Conditions

We used two prostate cancer cell lines preselected for differences in invasive and metastatic behavior. MatLyLu is a highly invasive rat prostate cancer cell line, whereas DU-145 is a less invasive human prostate cancer cell line. DU-145 cells were maintained in minimum essential medium (MEM) supplemented with 10% FBS and penicillin (100 U/ml)/streptomycin (100 µg/ml). The medium was supplemented with 10 mM HEPES during the cell perfusion experiments. MatLyLu cells were maintained in RPMI 1640 supplemented with 10% FBS, penicillin (100 U/ml)/streptomycin (100 µg/ml) and 10 mM HEPES. All cells were routinely assayed and found negative for mycoplasma contamination. To immobilize cells for the nuclear magnetic resonance experiments, cells were grown on spherical solid polystyrene microcarrier beads of 160 to 300 µm diameter (Biosilon, NUNC, Denmark). Cells were seeded at a density of 3x10⁶ cells per milliliter of beads in bacteriological petri dishes (0.5 ml beads/dish) and allowed to reach confluency on the bead surface for about 2 to 4 days by which time, the concentration was approximately 2x10⁷ cells/ml of beads. Cells on beads were counted by harvesting 0.5 ml beads in a 15-ml centrifuge tube. Nuclei were stained by incubating beads with cells for 1 hour at 37°C with 0.1 M citric acid containing 0.1% (w/v) crystal violet. Nuclei released by lysis following incubation were counted with a hemacytometer.

Preparation of Perfluorocarbon Emulsion, Alginate Beads and Matrigel Layer

The emulsion was prepared from 0.32 g of egg yolk lecithin (Sigma Chemical Co., St. Louis, MO) mixed with 2 ml of cold Tyrode solution (pH 7.4) [13]. This solution was sonicated twice for 15 seconds (130 W) at intervals of 1 minute, after which 1.33 ml of perfluorotripropylamine (FTPA) (PCR, Inc., Gainesville, FL) was added. The sonication cycle was then repeated several times until the entire solution was transformed to a milky emulsion. The solution was maintained on ice during preparation. The final emulsion was stored in the refrigerator until use. One day before the experiment, FTPA-doped gels (alginate beads and the Matrigel layer, see below) were prepared by mixing the emulsion with the liquid phase of the respective gel (1:100 for the ECM and 1:20 for the alginate).

Alginate beads were prepared from a sterile solution of 2.4% sodium alginate in saline [13]. The solution was gently pushed through a 27-gauge needle into a solution containing 100 mM CaCl₂, which resulted in the formation of alginate beads.

A well-defined Matrigel layer was obtained by pouring the cold liquid mixture onto a polycarbonate membrane that was fixed in a holder made from porous filter material (see scheme in Figure 1). The inner wall of the holder was reinforced with a 0.7-mm-thick ring made from Delrin (McMaster-Carr Supply Company, Chicago, IL). During addition of Matrigel the holder was immersed in growth medium at 37°C. Under this condition, the Matrigel gelled within approximately 5 minutes forming a stable, well-defined layer. The entire procedure was performed under sterile conditions. The perfluorocarbon embedded within the Matrigel provided measurement of oxygen tension directly within the Matrigel using ¹⁹F relaxometry.

Figure 1.

Schematic of the design of the perfusion system. The insert shows a magnified view of the sample structure within the 10-mm NMR tube.

MR Cell Perfusion System

For the MR studies, immobilized cells grown adherently on the polystyrene microcarrier beads were transferred into a custom built 10 mm screw cap NMR tube (Wilmad, Ltd., Buena, NJ) that had a reinforced open base to fit the out flow lines of the perfusion system. The sample structure is shown in Figure 1 and was prepared by loading the NMR tube according to the described protocol. First, a filter disc made from polycarbonate was pushed gently to the bottom of the tube. Then, beads with cells attached were transferred to the tube forming a layer of approximately 1.5 cm. The holder containing the Matrigel layer was placed on top of the beads. The cup-like structure of the holder minimized convection at the interface between microcarrier beads and Matrigel. A second layer of cell-covered beads, approximately 1.5 cm, was placed on top of the Matrigel cup followed by a single or double layer of alginate beads with perfluorocarbon. The alginate beads were covered by another layer (0.5 cm thickness) of beads with cells providing a homogeneous surface for the polycarbonate filter that was placed at the top of the sample. For some experiments, the bottom cell layer also contained a layer of perfluorocarbon-doped alginate beads. Medium was removed by the out flow line at the bottom of the sample at a rate of 1 ml/min during the time required to load the NMR tube. This provided continuous removal of excess medium added with the beads, in addition to perfusion of sample during the preparation period. Once the loading was complete, the sample was closed with a screw cap and perfusion was switched to well-defined media with a flow rate of 1 ml/min. Oxygen tension of the medium reservoir was maintained by stirring the reservoir while flushing the reservoir bottle continuously with a well-defined gas mixture (95% O₂/5% CO₂). Changes in oxygen tension of the influx medium, caused by diffusion exchange of oxygen across the wall of the tubing, was minimized using viton tubing (Cole-Parmer Instrument Company, Vernon Hills, IL). To achieve low oxygen tensions, it was crucial to compensate even for the remaining small changes. This was achieved by guiding the influx lines through a jacket circulating with water before they entered the NMR tube (see Figure 1). The circulating water in the jacket was saturated with the desired gas mixture and maintained at a temperature of 37 ± 1°C. To promote maximum gas exchange in the jacket, thin-walled silicon tubing (Silicone Platinum tubing, 0.8 mm i.d., 2.4 mm o.d., Cole-Parmer Instrument Company) was used for this portion of the lines.

MR spectra were acquired on a GE Omega 400 Spectrometer equipped with shielded gradients of up to 130 G/cm in three directions. A home-built probe consisting of a broadband coil (used for ³¹P spectroscopy) and a ¹H coil tunable to ¹⁹F was used to obtain ¹H, ¹⁹F and ³¹P MR spectra from the sample. The sample temperature was maintained at 37°C by blowing warm air, guided through the probe, against the bottom of the sample tube. The temperature of the incoming perfusate was adjusted to 37 ± 1°C during its passage through the water jacket. The perfusate reservoir was heated to 43°C to produce a thermal gradient between the reservoir and the NMR probe, thereby preventing air bubble formation in the perfusion lines [14]. Cell viability in the system, evaluated by the exclusion of trypan blue by cells recovered from the perfusion system after completing the experiments, was usually of the order of 95% or higher.

Oxygen Tension Measurements

Oxygen tension was monitored by localized ¹⁹F MR relaxometry of perfluorocarbons in alginate beads and/or the Matrigel (Figure 2). A layer of alginate beads doped with perfluorocarbon can be seen in Figure 2. An inversion recovery ¹⁹F spectroscopy sequence, spatially localized within this layer, provided T₁ relaxation rates of the embedded perfluorocarbons enabling determination of oxygen concentration at the position of the layers [13]. Calibration curves relating T₁ values to oxygen tension were obtained from inversion recovery experiments with perfluorocarbon-alginate beads in sealed tubes performed for oxygen tensions of 0%, 20.95% and 95%.

Figure 2.

¹H MR image of sample (right), ¹⁹F MR image (center), and two panels (left) showing the recovery of the inverted ¹⁹F signal of perfluorocarbons at the top and within the Matrigel layer of the sample.

Imaging

Two dimensional images (proton and fluorine) of the sample were obtained using a spin-echo imaging sequence with a field of view (FOV) of 40 mm, repetition time (T_R) of 1 second and an in-plane resolution of 78 µm. Proton images were obtained with an echo time (T_E) of 30 msec for a 2-mm-thick central slice running through the length of the tube. Fluorine images were obtained with a T_E of 20 msec for a 4-mm-thick central slice. The geometry of the sample, inhomogeneity in the Matrigel and/or bead packing, and potential macroscopic structural changes within the sample were readily evaluated from these images. The profile of intracellular water along the z-axis of the sample was obtained by diffusion-weighted ID ¹H MR imaging [15] using gradient pulses of 3 msec duration with 18 G/cm gradient strength and employing diffusion weighting time of 100 msec. This profile was obtained for the entire sample with a spatial resolution of 62.5 µm (FOV = 64 mm). A “zoomed-in” region near the Matrigel holder was measured with a spatial resolution of 29.3 µm (FOV = 30 mm) using selective excitation of a 10-mm slice positioned at the region of interest.

Spectroscopy

Energy metabolism of cells within the sample were measured by global ¹H and ³¹P MR spectroscopy. ¹H MR spectra were obtained by acquiring a stimulated echo using CHESS water suppression pulses [16] applied before and during the mixing interval of the stimulated echo pulse sequence [15]. Diffusion weighting for complete suppression of extracellular metabolites was achieved using the parameter settings described earlier for obtaining the intracellular water profile (diffusion time = 100 msec, duration of gradient = 3 msec, gradient strength = 18 G/cm) [14]. A total of 128 scans were accumulated with a repetition time of 1 second and an echo time of 11 msec. The temperature within the sample was determined periodically over the experimental time, from the chemical shift difference between the choline and water signals [17,18]. For this purpose, a diffusion-weighted proton spectrum was acquired without water suppression. ³¹P MR spectra, acquired by accumulating 2000 scans using a 45° rf pulse and a repetition time of 1 second, served to evaluate the stability of the energy levels of the cells throughout the entire period of the experiment.

Spectroscopic Imaging

1D Spectroscopic imaging of cell metabolism and growth in each layer along the z-axis was performed using a 1D ¹H CSI pulse sequence derived from the diffusion-weighted ¹H stimulated echo sequence described above [19,20]. To perform 1D-CSI, incremented phase encoding gradients in the z-direction were applied during the last T_E/2 interval of the sequence. Experiments with a spatial resolution of 156 µm or 310 µm (256 or 128 phase encoding steps, 40 mm FOV, T_R = 2 seconds, T_E = 11 msec) were performed within 9 hours of experimental time accumulating 64 or 128 scans per phase encoding step, respectively. Corresponding diffusion-weighted CSI data acquired without water suppression (two scans per phase encoding step) were obtained to normalize metabolites to the cellular water signal.

Spectroscopic imaging data were processed with an exponential window in the time domain and optional sine multiplication in the phase encoding (spatial) domain. Following 2D Fourier transformation, phase corrections were applied to the data set to obtain pure absorption spectra with narrow lines (compared to spectra processed by magnitude calculation). The temperature variation across the sample, evaluated from 1D-CSI spectra acquired with and without water suppression as described earlier, was found to be less than 1°C.

Experimental Protocol

The initial setup of the MR experiment including transfer to the magnet, shimming, adjustment of temperature, required approximately 2 hours. MR data were acquired consecutively according to the following protocol: 1) T₁-weighted ¹H image; 2) profile of total water; 3) zoomed profile of diffusion-weighted water (cellular water); 4) profile of diffusion-weighted water (cellular water); 5) global ³¹P spectrum; 6) global ¹H metabolite spectrum (diffusion-weighted); 7) global diffusion-weighted water spectrum (no water suppression); 8) 1D-CSI metabolite spectra; 9) 1D-CSI diffusion-weighted water without water suppression. At the end of each set, ¹⁹F experiments were performed to determine oxygen tensions.

The spatial resolution, derived from the FOV divided by the block size (number of data points along the z-axis) for the 1D MR microscopic imaging, and FOV divided by the number of phase encoding steps for the 1D-CSI were as follows: For the 1D MR microscopic imaging, the FOV was 64 mm for the full profile and 30 mm for the zoomed profile. With 1024 data points, this resulted in a resolution of 62.4 µm for the full profile and 29.3 µm for the zoomed profile. For the 1D-CSI, the FOV was 40 mm with either 128 or 256 phase encoding steps resulting in a spatial resolution of 156 or 310 µm.

Data Analysis

Cell invasion was determined from changes in the profile of intracellular water with time, which reflected invasion of cells from the polystyrene beads into the Matrigel layer. A quantitative measure of cell invasion, the invasion index (I), at time t was obtained as:

$$I(t\text{) = [}{I}_{\text{pz}}(t\text{)/}{I}_{\text{p}}(t\text{)] - [}{I}_{\text{pz}}(t\text{= 0)/}{I}_{\text{p}}(t\text{= 0)]}$$

where I_p is the integral value for the entire diffusion-weighted profile at time point t. I_pz describes integral values obtained by integrating the cellular signal from the zoomed profile covering a 7-mm region that started at the base of the Matrigel holder and included the Matrigel layer. Normalization to the integral of the entire profile, which represents the total cellular signal within the sample, accounted for cell proliferation in the Matrigel layer throughout the time course of the study.

Six separate sets of experiments were performed for MatLyLu cells and five for the DU-145 cell line. Statistical analysis was performed using repeated analysis of variance (ANOVA).

Results

Figure 3 (a and b) demonstrates the ability of the assay to measure cell metabolism and invasion within the same system. Proton spectra obtained from slices localized within a sample region containing cells attached to beads show well-resolved cellular signals from total choline (3.24 ppm), creatine (3.0 ppm), lactate, and triglycerides (1.3 ppm) as well as signal from beads (0.13 ppm) that serves as a specific marker for the presence of these microcarriers. Spectra from the Matrigel layer in the holder show hardly any cellular metabolites for less invasive DU-145 cells. In contrast, spectra from the same region for the MatLyLu cells, reveal signals from choline, creatine, lactate plus triglycerides with very little signal from beads. These data demonstrate that the assay is measuring invasion while providing metabolic information from the invading cells. Analysis of metabolites normalized to the cellular water signal did not indicate any significant metabolic differences between cells at the invading front and cells further away from the Matrigel layer for the MatLyLu cells.

Figure 3.

Expanded ¹H MR image of sample showing the Matrigel layer in the filter cup for (a) DU-145 cells after 58.5 hours and (b) MatLyLu after 56.5 hours. The Matrigel forms a well-defined bright layer in the filter cup. Images were obtained with TR = 1 second; TE = 30 msec; FOV = 40 mm; slice thickness = 2 mm; in-plane resolution of 78 µm. The ¹H MR spectra on the left side of the ¹H MR images are from localized slices (310 µm thick) from within the sample.

A characteristic feature of the MatLyLu cells was their ability not only to invade but also to degrade the Matrigel layer during the course of the experiment. As demonstrated in Figure 4, most of the Matrigel is degraded by 72 hours for the experiment with the MatLyLu cells, whereas no significant changes were observed for the DU-145 cells. This degradation appeared to be largely due to enzymes secreted by the MatLyLu cells. Since MatLyLu cells acidify their growth medium by extrusion of lactate at a much higher rate than DU-145 cells, we performed control experiments with a cell-free bead system with acidic medium to rule out the possibility of an acidic pH in combination with continued flow causing this effect. No significant degradation was observed in these control experiments.

Figure 4.

Comparison of T₁ weighted ¹H MR images from MatLyLu and DU-145 cells showing continuous degradation of Matrigel by MatLyLu compared to DU-145.

Zoomed cellular profiles of the sample, taken at different time points through a period of 2 to 3 days, for DU-145 and MatLyLu cells are shown in Figure 5. This figure further confirms the highly invasive behavior of MatLyLu cells compared to DU-145 cells. Cellular profiles, corresponding to the experiment represented in Figure 5, but covering the entire sample are shown in Figure 6. These profiles clearly demonstrate an increase in cell number with time for both cell lines and demonstrate the capability of cells to grow and proliferate in the cell perfusion system. The quantitative index of cell invasion (I) obtained from the zoomed and total profiles, averaged for five DU-145 experiments and six MatLyLu experiments is shown in Figure 7. A significant difference (P<.05, ANOVA, Fisher protected least significant difference (PLSD) test) was observed between the invasion indices (I) for the MatLyLu and the DU-145 line.

Figure 5.

Zoomed cell density profiles of MatLyLu and DU-145 cells. The MatLyLu cells invade the Matrigel, whereas DU-145 cells remain at the surface of the Matrigel layer. Each profile corresponds to an individual time point.

Figure 6.

Cell density profiles of the entire sample for MatLyLu and DU-145 cells. The profiles, representing the diffusion-weighted cellular water signal (see Materials and Methods section), were taken at consecutive days starting from day 0. Cell growth is represented by increased signal intensity. The lines in the center mark the border between cells, filter (F) and Matrigel (M). Each profile corresponds to an individual time point.

Figure 7.

Invasion index I for the highly invasive MatLyLu cells and the less invasive DU-145 cells. Values are mean ± SEM.

Representative global ³¹P spectra obtained through the experiment to evaluate the stability of cellular energy levels are shown in Figure 8. No significant changes were observed in these spectra, demonstrating the efficiency of the cell perfusion system in maintaining the energy levels over this period of time.

Figure 8.

Representative ³¹P MR spectra obtained from the samples 24 and 48 hours after transferring the cells to the perfusion system. The spectra demonstrate the stability of the energy levels of the cells while in the perfusion system.

Discussion

Here we have developed and characterized an assay which, for the first time, can measure invasion and metabolism dynamically. We recently obtained preliminary data for the invasive human prostate cancer cell line, PC-3, which further confirm the ability of the assay to measure invasion; the invasion index for the PC-3 cell line was significantly higher than the less invasive DU-145 cell line [21]. Coupled to this is the unique ability to measure oxygen tensions online within the extracellular environment and within the Matrigel layer, as well as the ability to alter, on-line, the extracellular physiological environment.

Invasion of the basement membrane, by cancer cells, is a necessary component of the metastatic cascade. Therefore, understanding the mechanisms, and identifying the stimuli, for cancer cell invasion is critical to prevent metastasis. Glucose starvation, acidosis and hypoxia have been shown to enhance the metastatic potential of murine tumor cells [22,23]. An association between tumor hypoxia and distant metastasis has also been observed in advanced cancer of the uterine cervix and in human soft tissue sarcoma [24,25]. Similarly, Schwickert et al. [26] have found that patients with cervical cancers containing high lactate were associated with a high risk of metastases. Increased metastasis also occurs when tumors grow in stressed or injured stroma such as a pre-irradiated tumor bed [27]. Although histological evidence suggests that metastasis is linked to high vascular density in a range of clinical tumors [28–31], it is likely that a fraction of these histologically identified vessels are not functional and the resultant hostile environment (low pH, oxygen, and glucose) increases the invasive behavior of a subpopulation of cells. The environmental stimulation and promotion of invasion may occur within a subpopulation of deprived cells. The probability of extravasation may be enhanced by an increased amount of vasculature available for these cells to spread from, especially for tumors where vascular drainage is the predominant mode for dissemination.

The “Metabolic Boyden Chamber” assay described here can probe metabolism during alteration of the physiological environment and simultaneously measure the invasion or migration of cells through reconstituted basement membrane. It can be used to identify physiological and metabolic characteristics that promote invasion, and equally, it can evaluate depletion of energy levels, or the chemotherapeutic alteration of metabolism, to inhibit cell motility and invasion. Although it is possible that microenvironmental physiological and metabolic heterogeneity may occur, especially for studies performed at low oxygen tensions, an analysis of localized proton spectra normalized to the cellular water signal did not reveal cellular metabolic gradients in the system, for both DU-145 and MatLyLu cell experiments. Another factor to consider is the difference in attachment properties of cells to beads. We did not detect any difference between the attachment properties of MatLyLu and DU-145 cells to the beads while growing them in culture. However, decreased cell attachment following physiological trauma, or increasing cell attachment as a form of therapy, would be attractive interventions to further understand the invasion and metastasis cascade.

Here we have established that the technique shows reproducible differences in invasion for two cell lines preselected for differences in invasive and metastatic characteristics. In the next stage of studies, which are currently underway, the effects of reduction in oxygen tensions on the invasion index are being investigated.