Directional movement of rat prostate cancer cells in direct-current electric field: involvement of voltagegated Na⁺ channel activity

Motility of cancer cells is important for the progression of metastasis because the cells have to migrate from primary sites, intra- or extravasate, and ultimately invade the target tissue to establish secondary tumours (reviewed by Mohler, 1993EF52; Banyard and Zetter, 1999EF3). A number of studies have demonstrated a strong correlation between cell motility (determined by membrane ruffling, pseudopod extensions and cell translocation) and metastatic potential (Mohler et al., 1987EF53; Mohler et al., 1988EF54; Partin et al., 1989EF61). Cancer cells can respond in a motile fashion to many external factors including extracellular matrix components, host-derived motility and growth factors as well as tumour-derived autocrine agents (Levine et al., 1995EF46) and scatter factors stimulating random (chemokinetic) or directional (chemotactic) motility of cancer cells (Aznavoorian et al., 1993EF2).

A variety of motile cells, from protozoan to mammalian, also respond to an externally applied direct-current (DC) electric field by changing the orientation of their movement (Erickson and Nucitelli, 1984EF18; Robinson, 1985EF66; Ferrier et al., 1986EF22; Frank and Gruler, 1990EF24; Nishimura et al., 1996EF55; McCaig and Zhao, 1997EF50). This property, known as `galvanotaxis', is involved in a number of basic biological processes, such as embryonic development (Jaffe and Nucitelli, 1977EF39; McCaig, 1989aEF47; McCaig, 1989bEF48; McCaig and Dover, 1989) and can also manifest itself under pathophysiological conditions, as in wound healing (e.g. Chiang et al., 1992EF11). In the related phenomenon of `galvanotropism', external electric fields can facilitate cellular process extension, again in both normal conditions (e.g. Hotary and Robinson, 1990EF35) and pathological situations (e.g. bone healing; Zhuang et al., 1997EF76) and nerve regeneration (Borgens et al., 1981EF6). However, it is not known whether cancer cells are galvanotactic. Also, it is not known whether there is any relation between galvanotaxis and metastatic potential. Nevertheless, metastasis can involve specific tissue invasion; for example, rat prostate cancer MAT-LyLu cells metastasise specifically to lymph nodes and lungs (Isaacs et al., 1986EF36). Furthermore, metastasis can originate from sites in the body (e.g. epithelial ducts, skin) where local DC electric fields would occur. In the case of the rat prostate gland, a transepithelial potential of some —10 mV has been recorded (Szatkowski et al., 2000EF74). Interestingly, electro-imaging of mammary and cervical tissues has been used in clinical detection of malignancy (Fukuda et al., 1996EF28; Faupel et al., 1997EF21; Cuzick et al., 1998EF15), although its basis is not well understood.

Cells can react to an applied weak DC electric field in different ways, moving towards the cathode or the anode, and it is possible that a variety of membrane mechanisms are involved in such responses (e.g. Nuccitelli, 1988; Robinson, 1985; Soong et al., 1990). It has been observed that, under the influence of externally applied DC electric fields, Ca²⁺ concentration increased significantly and was maintained for the duration of exposure (Onuma and Hui, 1988). In this situation, the cell would move away from the end where intracellular Ca²⁺ rises and contractile activity occurs (Cooper and Keller, 1984). Accordingly, in the presence of Ca²⁺ channel blockers (e.g. D-600, verapamil), the directionality of migration can be disturbed and replacement of external Ca²⁺ by Mg²⁺ can also reverse the electrotactic response (Cooper and Schliwa, 1986; Onuma and Hui, 1988; Nuccitelli and Smart, 1989). Importantly, however, Ca²⁺-independent control of galvanotaxis has also been found (e.g. Brown and Loew, 1994; Palmer et al., 2000), and it is not clear whether other ionic mechanisms also play a role in galvanotactic responses of cells. We have shown previously that there are distinct electrophysiological differences between strongly and weakly metastatic cells of rat and human prostate carcinoma (Grimes et al., 1995; Laniado et al., 1997; Smith et al., 1998; Foster et al., 1999). In particular, functional voltage-gated Na⁺ channels (VGSCs) occurred specifically in the highly metastatic cells (Grimes et al., 1995; Laniado et al., 1997; Grimes and Djamgoz, 1998). Indeed, VGSC expression and invasiveness in vitro were correlated in numerous rat and human prostate cancer cell lines (Smith et al., 1998).

Taking the available evidence together, the possibility arises that metastatic cells could be galvanotactic and that membrane ion channel (VGSC) activity could play a role in this process. The present study aimed to evaluate this overall hypothesis. In order to test the possible involvement of VGSC activity in galvanotactic reaction, tetrodotoxin (TTX), a highly specific blocker of VGSCs, was used. The working concentration of TTX was 1 μM, which would effectively block the TTX-sensitive VGSCs present in the strongly metastatic MAT-LyLu line (Grimes and Djamgoz, 1998; Diss et al., 2001). A further, `opposite' test was carried out using veratridine (a VGSC `opener'), which potentiates VGSC activity by lowering the threshold of the activation voltage (Eskinder et al., 1993). Finally, possible involvement of voltage-gated Ca²⁺ channels was tested using verapamil as a general blocker (Cooper and Schliwa, 1986).

Experiments were carried out on two well characterised rat prostate cancer (`Dunning') cell lines having markedly different metastatic ability: MAT-LyLu and AT-2 cells which metastasise in >90% and <10% of cases, respectively, when injected into Copenhagen rats (Isaacs et al., 1986). The cells were cultured at a density of ∼100,000 cells cm^-2, as described previously (Grimes et al., 1995), except that the concentration of fetal calf serum (FCS; Sigma) was increased to 5% to facilitate cell attachment. Cultures were always used 24 hours after plating, when single cells were abundant.

Migration was assayed using a galvanotaxis apparatus described in detail by Korohoda et al. (Korohoda et al., 2000). Essentially, this was made up from a glass observation chamber comprising a sandwich of two glass coverslips (with the cells free to move in between) mounted in a Plexiglas holder. Direct current was applied for 6 hours, through Ag/AgCl reversible electrodes (of 6 cm² surface area) immersed in wells filled with cultured medium (RPMI + 5% FCS). Each well was connected by an agar bridge to a neighbouring pool, which was continuous with the observation chamber. The cover-glasses composing the latter measured 60 mm×5 mm×0.2 mm. The electric current flowing through the chamber was measured continuously with a milliampermeter; the voltage gradient was calculated using Ohm's Law and confirmed by measuring with a high-input impedance voltmeter (Korohoda and Kurowska, 1970; Cho et al., 1996). 24 hours before starting an experiment, the cells were replated onto one of the glass covers. At the beginning of the experiment, this was sandwiched onto a second plate (with a 0.2 mm gap, the cells still being bathed in their normal growth medium), sealed with silicone grease and mounted in the Plexiglas apparatus. Observations were carried out on an inverted microscope (Olympus, IMT-2). Cells were observed using phase contrast optics at a total magnification of 285× (corresponding to a field of view of 710×710 μm). All experiments were done at 37°C (the temperature of the chamber was monitored and found to remain constant). Analysis of cells' motility during the experiment (6 hours) confirmed that the recording conditions were very stable because there was no observable change in any parameter under the control conditions. All experiments were carried out for 6 hours unless otherwise stated.

Cell images were recorded with a Hitachi CCD camera, digitized and processed with the computer programs as described previously (Korohoda and Madeja, 1997EF41; Korohoda et al., 1997aEF42; Korohoda et al., 1997bEF43). For cell motion tracking, a series of time-lapse images of cells were acquired using a frame-grabber device and a `slow motion VCR' program (VCR 1.0). The cell trajectories were constructed from 72 successive cell centroid positions recorded over 6 hours with a time interval of 5 minutes (30 seconds in some experiments). For each experimental condition, the trajectories of at least 50 cells were analysed. Usually, 15-20 cells were examined in one experiment and data from three or four experiments were pooled. Detailed analyses of the data were performed using the program `Mathematica' (Wolfram Research Inc., Champaign, IL; Korohoda and Madeja, 1997EF41; Korohoda et al., 1997aEF42).

The following parameters characterising different aspects of cell locomotion were computed and analysed for each cell or cell population (Korohoda et al., 2000EF44).

(1) Total length of cell trajectory (TLT). This was essentially the `true' length of the path (in μm) travelled by the cell. The value of TLT was calculated from the sequence of n straight-line segments, each corresponding to a cell-centroid translocation between two successive images.

(2) Average speed of cell locomotion (ASL). This was defined as the total length of cell trajectory (TLT) ÷ time of recording (6 hours).

(3) Total length of cell displacement (TLD). This was the distance (in μm) from the starting point direct to the final position of the cell.

(4) Average rate of cell displacement (ARD). This was defined as the total length of cell displacement from the starting point to the final cell position (TLD) ÷ time of recording (6 hours).

(5) Coefficient of movement efficiency (CME). This was defined as the ratio of total cell displacement (TLD) to total cell trajectory length (TLT). The value of CME would be 1 for cells moving persistently along one straight line in one direction, and 0 for random movement (Friedl et al., 1993EF27; Korohoda et al., 1997bEF43).

(6) Average directional cosine γ (ADCγ). The angle γ was defined as the directional angle between the x axis (parallel to the electric field) and a vector AB, A and B being the original and each subsequent positions of the cell, respectively. This parameter would equal +1 for a cell moving towards the cathode, -1 for a cell moving in the direction of the anode and 0 for random movement (Gruler and Nuccitelli, 1991; Korohoda et al., 1997b). This parameter was used generally to quantify the directionality of movement.

(7) Average directional cosine β (ADCβ). The angle β was defined as the directional angle between the x axis (parallel to the electric field) and a vector AB, A and B being two successive positions of the cell, respectively. This parameter would equal +1 for a cell moving towards the cathode, -1 for a cell moving in the direction of the anode and 0 for random movement (Gruler and Nuccitelli, 1991; Korohoda et al., 1997b). This parameter was particularly useful in testing the reversibility of effects of electric field or a given pharmacological agent.

In experiments involving pharmacological treatment of the cells under investigation, a given test compound was introduced into the observation chamber and recording started immediately. In some experiments, it was possible to re-perfuse the cells with the control medium while recording to check the reversibility of the effect, if any. The following compounds were used: TTX (citrate free), veratridine and verapamil. All chemicals were obtained from Sigma-Aldrich.

Data are presented as means ± s.e.m. Statistic significance was determined using the Mann-Whitney test. Values of P < 0.01 were assumed to represent significance.

Under the culture conditions used, there were some noticeable differences between the morphologies of the strongly metastatic MAT-LyLu and the weakly metastatic AT-2 cells (Fig. 1). The MAT-LyLu cells were more rounded and the population was more heterogeneous (Fig. 1A,B). By contrast, the AT-2 cells were much flatter, with processes that spread out and attached to the surface (Fig. 1C,D). The quantitative motility data obtained for MAT-LyLu and AT-2 cells are summarised in Tables 1 and 2, respectively. In the absence of any external electric field, the MAT-LyLu and At-2 cells migrated laterally in culture with comparable speeds (52-57 μm hour^-1) and random directionality (ADCγ=-0.04 to -0.05) (Fig. 2A,C). A dramatic change occurred in the MAT-LyLu cells following application of an electric field of 3 V cm^-1, whereby their movement became strongly directional (ADCγ=0.82), cells turning to move almost linearly towards the cathode (Fig. 2A,B). Thus, although the average speed of movement was not affected, the rate of cell displacement increased nearly fourfold (Table 1). This effect was very rapid, microscopically detectable changes occurring within 30 seconds (i.e. between successive frames in the fast time-lapse recordings) after applying the field. By contrast, the AT-2 cells showed no such response (Fig. 2C,D; Table 2). In fact, there was a tendency for the AT-2 cells to move in the opposite direction but this was not studied further.

The electric field effects seen on the motility of the MAT-LyLu cells were enhanced as the applied voltage gradient was made stronger (Fig. 3A,B). Thus, as the electric field was increased from 0.1 V cm^-1 to 4.0 V cm^-1, both the directionality of movement and the overall length of cell displacement increased steadily. These effects were reversible. Thus, upon extinguishing or reversing the field, the cells' corresponding responses were rapidly abolished or reversed, respectively (Fig. 4). On the whole, however, there was some heterogeneity in the cells' response to the electric fields, with small, rounded cells appearing more sensitive than larger cells with prominent extensions.

We have shown previously that only strongly metastatic cells express functional VGSCs (Grimes et al., 1995; Laniado et al., 1997; Smith et al., 1998). Therefore, the possible role of VGSC activity in the cells' response to electric field was investigated by pharmacological manipulation of VGSC activity (Grimes and Djamgoz, 1998; Fraser et al., 2000; Figs 5,6,7; Tables 1, 2). An additional parameter, P_c (percentage of cells crossing a criterion boundary of 200 μm from the position occupied just before application of the field during the 6-hour recording period) was used to compare the effects of the drugs on the extent of the cells' directional migration (Fig. 7). Under control conditions, only very few (∼2%) MAT-LyLu cells crossed the barrier; this was greatly facilitated by the electric field (P_c=24.0±2.9%; P=0.001) (Fig. 7), consistent with the demonstration above. None of the AT-2 cells migrated to this extent (Fig. 2C; Fig. 5E). Treatment of the MAT-LyLu cell cultures with verapamil (1-10 μM) produced no effect (not illustrated). However, the specific VGSC blocker TTX at 1 μM noticeably reduced the percentage of MAT-LyLu cells migrating over the 200 μm criterion distance (P_c=7.6±4.3%), whereas 2.5-5 μM TTX blocked the effect of the field completely (P_c=0%) (Fig. 5A,B; Fig. 7). The effects of TTX were reversed by washing the cells with the control medium (Fig. 6). TTX had very little effect on the average speed of movement of the MAT-LyLu cells (Table 1) or any of the measured parameters of the AT-2 cells (Table 2; Fig. 5E,F).

Potentiating VGSC activity with veratridine (VER) had the opposite effect to TTX, enhancing the specific actions of the electric field treatment on the MAT-LyLu cells (Fig. 5C,D; Fig. 7). Thus, in the presence of 1 μM VER, the coefficient of movement efficiency increased significantly from the normal value of 0.34 to 0.62 without any effect on the average speed of cell movement (Table 1), and 30.0 ± 10% of cells crossed the 200 μm criterion barrier (Fig. 7). The effects of VER were dose-dependent (Table 1; Fig. 7). The value of ADCγ was enhanced significantly at 5 μM, the cells increasing their directional response further than with application of the field alone. At the highest concentration of VER used (10 μM), P_c=44.0±2.0% (P=0.002 compared with electric field without VER) (Fig. 7). Cell viability, assessed by trypan blue staining, was not affected by VER (10 μM) or TTX (5 μM). However, VER treatment again had relatively little effect on the AT-2 cells (Table 2), consistent with functional VGSCs being absent in these cells.

The results of the present study show for the first time that prostate cancer cells can be galvanotactic and that VGSCs are involved in galvanotactic response, both aspects being correlated with metastatic potential (Grimes et al., 1995EF31; Laniado et al., 1997EF45; Smith et al., 1998EF71). Thus, the hypothesis under investigation has been confirmed.

The galvanotactic responses of various cell types, including neurons and epithelia, have been found to be cathodal (McCaig and Zhao, 1997EF50; Palmer et al., 2000EF60), as shown here for the MAT-LyLu cells. The effect is unlikely to be due to any asymmetry in the distribution of VGSC protein over the cell surface because this was found to be uniform on MAT-LyLu cells (not shown). Furthermore, any gross redistribution of VGSCs (or other protein) in the membrane by electrophoresis would also seem to be unlikely because this would take hours (Jaffe, 1977EF38), or at least several minutes (Fang et al., 2000), whereas the electric field produced an observable morphological change and directional response in the cells within 30 seconds. The following possibilities can be considered as underlying mechanisms.

It has been observed that, under the influence of externally applied DC electric fields, intracellular Ca²⁺ concentration increased significantly and was maintained for the duration of exposure (Onuma and Hui, 1988; Perret et al., 1999). In the simplest case, of a cell devoid of voltage-gated (depolarization-activated) Ca²⁺ channels (VGCCs), passive influx of Ca²⁺ and contraction would on the anodal side and the cell would move towards the cathode. In cells with VGCCs, however, Ca²⁺ influx would (also) occur on the cathodal, depolarised, side; thus, the direction of movement might depend on the balance between the two sites of intracellular Ca²⁺ rise. These possibilities are unlikely to be directly applicable here for several reasons. (1) Any simple passive Ca²⁺ influx should have affected the two cell types similarly but this was not seen (Fig. 2B,D). (2) Verapamil (a general blocker of Ca²⁺ channels) was found not to affect the galvanotactic response of the cells. (3) The membrane depolarisation caused directly by the electric field should have facilitated Ca²⁺ channel activation directly without involving VGSCs; that is, TTX should not have had any effect. (4) Our preliminary patch-clamp recordings and intracellular Ca²⁺ measurements suggest, in fact, that the MAT-LyLu cells do not possess voltage-gated Ca²⁺ channels (or significant Na⁺-Ca²⁺ exchanger activity).

The `direct' effect of the electric field on the membrane potential could be compounded by the presence of a negative surface charge, which is known to occur in cancer cells and to increase in line with metastatic potential (Abercrombie and Ambrose, 1962; Carter and Coffey, 1988; Carter et al., 1989; Price et al., 1987). It is therefore possible that, in the presence of the exogenous electric field, the charge `cloud' would be asymmetrical, and this could generate a spatial variation in membrane potential (Gross, 1988; Heberle et al., 1994; Scherrer, 1995) that is different between the two cell types. In the case of the MAT-LyLu cells, the negative charges could accumulate on the anodal surface, thereby depolarising the cell membrane. This could then simply increase the passive influx of Ca²⁺, leading to contraction on the anodal side and propulsion of the cells towards the cathode, as has been observed (Bedlack et al., 1992; Olivotto et al., 1996). In addition, the VGSCs could interact directly with the cytoskeleton (Sheng and Kim, 1996), possibly by the membrane depolarisation inducing repeated conformational changes in the VGSC proteins. Furthermore, VGSC β-subunits possess a cell-adhesion motif (e.g. Isom et al., 1995) and this could facilitate interaction with the extracellular matrix. We should also note that control of galvanotaxis by Ca²⁺ released from internal stores (Rosado and Sage, 2000) and other more elaborate secondary messenger have also been proposed (McCaig and Zhao, 1997). However, how these could involve VGSC activity is not clear at present. We also do not know whether electric field treatment could have any long-term effect. Further work is required to evaluate these possibilities.

The hitherto unknown role of VGSC activity in galvanotaxis that we have demonstrated could have important implications in both cellular physiology and pathophysiology. First, as regards normal biological functioning, a variety of basic processes do involve directional or patterned growth, including target-specific axonal migration and patterning of regional synaptic connectivity. These processes have frequently been shown to be blocked by TTX treatment, so would appear to depend on VGSC activity (Dubin et al., 1986EF17; Catalano and Shatz, 1998EF10; Meyer, 1982EF51; Penn et al., 1998EF62; Shatz, 1990EF68). Accordingly, VGSC expression and activity in cells exposed to endogenous electric fields could facilitate directional growth in vivo.

Second, assuming that our findings are also applicable to the situation in vivo (Fig. 8), especially because VGSC protein is produced in clinical tumours at levels correlated with pathological grading (Stewart et al., 1999), it would follow that the directional migration of prostate cells during the early stages of metastasis could be influenced significantly by endogenous transepithelial potentials (TEPs). We have recently found that rat prostate epithelia have a lumen potential of about -10 mV (Szatkowski et al., 2000). Such a lumen potential would correspond to transepithelial voltage gradient of 5 V cm^-1, assuming that the cellular thickness of the prostatic ducts is 20 μm (Fig. 8). Such a voltage gradient is comparable to the DC electric field strengths used to induce galvanotaxis in the present study. If a similar situation occurs in the human prostate epithelium in vivo then it would follow that presumed premetastatic cells with VGSC activity would tend to migrate into the lumen and be detectable in the semen (Gardiner et al., 1996; Barren et al., 1998). Subsequently, as metastatic behaviour progressed, accompanied by deformation of the epithelia, the negative TEP would degrade, cellular migration into lumen would slow down and might even reverse, encouraging invasion of the surrounding tissue. Furthermore, the transendothelial potential (Revest et al., 1993) could similarly influence extra- and intravasation of circulating metastatic cells, which might be a critical step in metastasis (Wyckoff et al., 2000).

This study has been supported by the Prostate Cancer Research Fund (UK) and the Committee for Scientific Research (KBN, Poland). We thank Clare Isacke and Jimmy Diss for previewing the manuscript, and Andrew Goldsworthy and Marek Szatkowski for many useful discussions.