Oncogene (2002) 21, 1989 ± 1999 ã 2002 Nature Publishing Group All rights reserved 0950 ± 9232/02 $25.00 www.nature.com/onc

Gap junction intercellular communication propagates death in cancerous cells

Vladimir A Krutovskikh*,1, Colette Piccoli2 and Horashi Yamasaki3

1Unit of -Environment Interactions, International Agency for Research on Cancer, Lyon, 69372, France; 2Unit of Gene- Environment Interactions, International Agency for Research on Cancer, Lyon, 69372, France; 3Kwansei Gakuin University, 1-1-155 Uegahara, Nishinomiya, 662-8501 Japan

Gap junction intercellular communication (GJIC) or cell Introduction coupling has an important function in maintaining tissue homeostasis and is thus a critical factor in the life and death Gap junctions are tiny aqueous channels which connect balance of cells. While the role of GJIC in cell growth the of contiguous cells, enabling cells to share regulation has been much studied, its involvement in directly small metabolites by the process called gap remains unclear. In this study we elucidated the junction intercellular communication (GJIC) or cell possibility that cell death is propagated via gap junctions, coupling. GJIC has a role in maintaining tissue home- employing the rat bladder carcinoma cell line BC31. BC31 ostasis or its variant morphostasis, which is largely cells proliferate quickly, are tumorigenic, and are well- achieved by regulating the balance between cell gain and coupled via gap junctions that contain the gap junction cell loss. There is a large body of evidence that GJIC are Connexin43 (Cx43). In addition, these cells are implicated in the regulation of cell growth, mostly by predisposed to spontaneous death by apoptosis, particularly negative control of cell proliferation (Loewenstein and upon achieving con¯uency. We found that many dying Rose, 1992). Correspondingly, disorders of GJIC are BC31 cells express Cx43 just as their non-apoptotic strongly associated with aberrant cell growth diseases, counterparts do. Furthermore, Cx43 in apoptotic cells including cancer (Yamasaki, 1990; Krutovskikh and could be functionally competent, supporting coupling of Yamasaki, 1997). Due to their frequent functional these cells with their non-apoptotic neighbors, and as a alteration in tumors, the of gap junction result, clusters of coordinately dying cells were observed. ± (Cx) ± have been classi®ed as tumor The role of Cx43 and GJIC in propagating cell death was suppressors (Chipman, 1995; Yamasaki and Naus, 1996). shown by analysing clones of BC31 cells expressing a In cancer, an increase in cell number is often a result mutant of Cx43 that is a dominant negative inhibitor of of suppressed cell loss (apoptosis) rather than increased GJIC, and by using b-glycyrrhetinic acid to inhibit intrinsic cell proliferation (McDonnell, 1993). In view of this, cell coupling in BC31 cells: in both cases the formation of and the ubiquitous role of cell coupling in the clusters of dying cells was abrogated, and the intensity of maintenance of tissue homeostasis, its implication in cell death was considerably decreased. These results suggest regulation of apoptosis was postulated. that GJIC spreads cell-killing signals initially generated by Previous experiments have found that cell coupling a single cell that spontaneously initiates apoptosis, into could propagate cell injury in cultured (Cortina healthy surrounding cells, thus increasing the level of cell et al., 1998). Furthermore, cell ± cell communication via death. Treatment of BC31 cells with the sleep-inducing lipid gap junctions increases the volume of brain infarct, and gap Oleamide, which selectively restricts gap junction perme- junction blockers e€ectively inhibit this spreading (Raw- ability to Ca2+ , did not abrogate coordinated cell death anduzy et al., 1997). However, until now there has been no by clusters, indicating that Ca2+ ions are the most probable experimental evidence that cell coupling could play a cell-killing signals spread through gap junctions. signi®cant role in regulating cell death in cancerous cells. Oncogene (2002) 21, 1989 ± 1999. DOI: 10.1038/sj/ Here we report experimental evidence that due to cell onc/1205187 coupling, individual dying bladder cancer cells can spread cell death signals into adjacent cells which then also die by Keywords: gap junctions; apoptosis; Connexin43 apoptosis, and that the messenger molecules which pass through gap junctions to kill cells are very likely Ca2+ ions.

Results

BC31 cells as a model to study the role of cell coupling in propagation of cell death *Correspondence: V Krutovskikh; E-mail: [email protected] Received 8 May 2001; revised 26 November 2001; accepted 22 The BC31 rat bladder carcinoma cell line has a high January 2002 coupling capacity due to the expression of only one Gap junctions propagate cell death VA Krutovskikh et al 1990 type of protein, Cx43 (Asamoto et al., 1994; Connexin43 (Cx43), as their non-apoptotic counter- Krutovskikh et al., 1998). Another characteristic of parts do (Asamoto et al., 1994; Krutovskikh et al., these cells is their propensity to die spontaneously. The 1998). The immunoblotting study used an antiserum combination of these features o€ered an opportunity to raised against a Cx43 peptide to test for the presence of employ these cells as a model for investigating the Cx43 in protein prepared from the fraction of dying possible role of gap junction intercellular communica- cells, and revealed levels of Cx43 comparable to those tion in regulating cell death. in the fraction of non-apoptotic BC31 cells. The only di€erence between apoptotic and healthy cells revealed by the immunoblotting analysis of Cx43 expression, BC31 cells die by apoptosis was a less neat electrophoretic separation of individual Examination of the morphology of con¯uent mono- bands corresponding to the di€erently phosphorylated layer of BC31 cells reveals the frequent presence of species of Cx43 i.e. P0,P1,P2, in a sample from the either individual cells, or small clusters of cells with fraction of dying BC31 cells (Figure 2a). obvious signs of death; i.e. they are round in shape, Immunostaining also detected Cx43 in some apopto- often with a `blebbed' membrane, and are poorly tic cells, localizing its expression to parts of lateral attached to the surface of culture dishes (Figure 1a). In plasma membrane that were engaged in intercellular situ staining for apoptotic markers using Annexin V contacts with non-apoptotic neighbors in plaque-like and Propidium Iodide detected substantial number of forms (Figure 2b). This ®nding posed the question individual cells positive either on Annexin V alone or whether these gap junction-like structures between for both these markers of apoptosis, indicating there- apoptotic and non-apoptotic cells could be functionally fore, on apoptotic origin of cell death (Figure 1b). competent. Functional heterologous coupling between In addition, agarose gel electrophoresis of DNA apoptotic and non-apoptotic cells could be visualized prepared from the fraction of dying cells collected by by selectively loading tracers speci®cally permeable for mechanical exfoliation from cell culture, as described in gap junctions, e.g. Lucifer yellow (LY) dye into the `Materials and methods', revealed a `ladder' pattern of of apoptotic cells, and observing the transfer DNA degradation, providing, thus, additional evidence of dye into adjacent non-apoptotic cells in the that individual BC31 cells die by apoptosis (Figure 1c). monolayer. The initial selective loading of dye into apoptotic cells was readily achieved because, as previously reported, metabolically inhibited, in other Expression of Cx43 by apoptotic BC31 cells words dying cells, increase the permeability of their Using this same method of separating dying from plasma membranes and therefore, unlike healthy cells, healthy cells, we were able to collect cells in sucient become permeable to Lucifer yellow dye from outside quantity to examine, by Western blot analysis, whether of cells (Li et al., 1996; John et al., 1999). We apoptotic cells express the con®rmed the validity of this for BC31 cells by brie¯y

Figure 1 BC31 cells frequently die by apoptosis. (a) Numerous individual dying BC31 (indicated by arrows) are morphologically distinguishable and poorly attached to the cell monolayer (bar 100 mm). (b) In situ Annexin V/Propidium iodide staining of non-®xed BC31 cells reveals multiple individual cells which are positive for these markers of apoptosis (bar 200 mm). (c) Agarose gel electrophoresis of DNA samples isolated either from healthy, ®rmly attached cells in the monolayer (left) or from dying cells, which are poorly attached to the monolayer (right). The `ladder' pattern of DNA degradation that is typical of apoptosis is seen for right sample.

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1991 exposing them to medium containing LY and found groups or clusters of non-apoptotic cells that were that LY was indeed selectively taken up from the positive for LY. It was noteworthy that in the middle extracellular medium by numerous individual cells with of each group of such LY-positive non-apoptotic cells morphological signs of apoptosis, such as `blebbing' of there was a single bright cell, or a few particularly the plasma membrane (Figure 3). There were no signs bright cells, with morphological features of apoptosis of direct LY uptake by healthy non-apoptotic cells in (Figure 4). The higher concentration of LY in such monolayer. cells indicated that most probably they are the source from which LY spread into adjacent non-apoptotic cells in the monolayer. The gradual decrease in GJIC between apoptotic and non-apoptotic cells in intensity of LY staining of the cytoplasms of normal monolayer of BC31 cells cells, as one moves from the center of LY-positive However, microscopic examination of a monolayer of focus towards its periphery, resembles the pattern of BC31 cells after exposure to LY occasionally revealed, LY spreading in a cell monolayer after its microinjec- in addition to numerous individual apoptotic cells, tion into a single cell, and therefore indirectly

Figure 2 Expression of Connexin43 in healthy and apoptotic BC31 cells. (a) Western blotting detected comparable levels of Cx43 protein in both healthy (1) and dying (2) BC31 cells. However, sample from dying cells displayed less neatly separated phosphorylated species of Cx43 (P0,P1 and P2). (b) Immunostaining reveals Cx43 plaque-like structures in parts of the lateral plasma membranes involved in cell-to-cell contacts between a single apoptotic cell (in the middle) and several non-apoptotic cells that surround it (shown by arrows). Propidium iodide counterstaining reveals the nuclear fragmentation that is typical of apoptosis, in a cell in the middle of the ®eld (bar 20 mm)

Figure 3 Spontaneous uptake of Lucifer yellow (LY) dye by apoptotic cells from extracellular medium (see Materials and methods). Several individual bright cells of round shape are seen (bar 100 mm). Insertion: An individual cell that has taken up LY shows the membrane blebbing that is typical of apoptosis (bar 100 mm)

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1992

Figure 4 Functional GJIC between apoptotic and non-apoptotic BC31 cells as revealed by LY uptake. Groups of cells, ranging in size from a very few (a) to several (b) cells contain LY dye, as the apoptotic cells which were permeable to the LY added to the medium, spread this dye into their non-apoptotic neighbors. Note a gradient of intensity LY from the center of such coupled foci towards their periphery. (bars 100 mm)

corroborates the suggestion that LY was initially taken achieved, particularly big clusters of dying cells up by an apoptotic cell in the center of such focus and detached from the cell monolayer simultaneously, then spread into the cell monolayer. It is noteworthy which eventually resulted in the formation of defects that Propidium Iodide (PI), which we used to stain the in the cell monolayer (Figure 6b). At that stage of nuclei of cells to reveal the DNA fragmentation culture a signi®cant decrease in the number of cells in characteristic of apoptosis, is known to be another monolayer was detected (Figure 6c). This formation of dye to which gap junctions are speci®cally permeable. particularly big clusters of dying cells and/or defects in Double vital staining of BC31 cell monolayer by cell monolayer coincided with the highest coupling Annexin V and Propidium Iodide, showed that some- capacity of BC31 which is observed soon after times a single apoptotic cell that stains positive with con¯uence, prompting us to associate these events. both Annexin V and Propidium Iodide is surrounded by several non-apoptotic cells positive for Propidium Inhibition of intrinsic GJIC in BC31 cells by Iodide only (Figure 5). Again we assume that PI could dominant-negative mutant of Cx43 decreases the level of get into the cytoplasms of healthy cells only through a cell death single Annexin V-positive apoptotic cell in the middle of such focus. In order to verify whether cell coupling is indeed a means Taken together, we consider that our observations of propagating cell death in a cell monolayer, we obtained by the spontaneous LY uptake assay and by measured the e€ect of abrogating the intrinsic GJIC the Annexin V/ Propidium Iodide staining experiments between BC31 cells, on the total level of cell death at can be interpreted as an evidence of GJIC between di€erent stages of cell growth. Previously we demon- apoptotic and non-apoptotic cells. The limited number strated that several Cx mutants are strong dominant of apoptotic cells found coupled with non-apoptotic negative inhibitors of GJIC in di€erent cell lines (Du¯ot- counterparts, compared to the total number of Dancer et al., 1997; Krutovskikh et al., 1998, 2000). individual apoptotic cells permeable for LY, suggested Speci®cally, we found that Cx43 lacking seven residues that GJIC between apoptotic and healthy cells in from the internal loop at positions 130 ± 136, e€ectively monolayer is a very short-term event during apoptosis. abrogated GJIC in BC31 cells (Krutovskikh et al., 1998) Consecutive brief exposures of BC31 cell monolayer to by occluding the channels of established gap junctions in ®rst Propidium Iodide and then LY, showed that with the lateral plasma membrane of cells. Clones of BC31 a 20-min interval between the two exposures there are cells de®cient in cell coupling due to expression of this no cells positive for both these dyes. Cx43D mutant were used in this study. In situ staining of apoptosis markers using Annexin V and Propidium Iodide did not reveal an obvious Cell death due to spreading of killing messengers from di€erence in the number of individual apoptotic cells apoptotic cells to non-apoptotic neighbors between communication-de®cient clones of BC31 and Observation of BC31 cell culture during its exponential the well-coupled parental cell line. This indicates that phase of growth revealed the progressive formation of inhibiting intrinsic GJIC has no apparent impact on morphologically distinguishable clusters of dying cells the frequency with which `primary' apoptosis sponta- (Figure 6a). Subsequently, when con¯uency was neously occurs in individual BC31 cells.

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1993

Figure 5 GJIC between apoptotic and non-apoptotic BC31 cells shown by double Annexin V/Propidium iodide staining of live cells. A single apoptotic cell, positive on Annexin V and permeable for Propidium iodide dye, is shown in the middle of ®gure (a) and outlined by a dotted line on ®gure (b). Propidium iodide has spread into the cytoplasms of non-apoptotic Annexin V-negative neighbors (b). Note the normal shape of the nuclei stained by Propidium iodide in the non-apoptotic cells (b). Pictures A and B have been taken from the same ®eld with di€erent ®lters (bar 100 mm)

Figure 6 Coordinated death by clusters in communication-pro®cient BC31 cells. (a). Morphologically-distinguishable clusters of dying cells (bar 500 mm) (b). Defect in monolayer as a result of cell death by clusters (bar 500 mm)

The Lucifer yellow uptake assay demonstrated that di€erent stages of cell culture (Figure 7a). As a expression of Cx43D dominant negative mutant in result, the intensity of cell loss after con¯uency in BC31 cells did not prevent uptake of this dye from communication-de®cient clones of BC31 cells was extracellular medium by single apoptotic cells. considerably less than in the well-coupled control Meanwhile, as we expected, there was no spread- (Figure 7b). A similar selective inhibitory e€ect, on ing of LY from individual apoptotic cells into only `bystander', or secondary cell death, was contiguous non-apoptotic counterparts in the mono- observed when intrinsic GJIC in BC31 cells was layer of communication-de®cient clones of BC31 cells suppressed by another dominant-negative mutant of which expressed the dominant negative mutant of Cx43 (E166G) (not shown). Unlike Cx43D, the Cx43D. Furthermore, no clusters of dying cells were E166G mutant abrogates intrinsic GJIC by retaining observed in non-coupling clones of BC31 cells at of Cx43 complexes intracytoplasmically (Krutovskikh

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1994 et al., 2000). Thus we conclude that the protection The nature of cell death signals spread through gap a€orded against cell death by dominant negative junctions mutants of Cx43 should be attributed to their capacity to inhibit cell ± cell coupling. We then addressed the critical question what is/are the killing messenger(s) that are generated in individual apoptotic cells and then spread through gap junctions The inhibition of GJIC by the blocking agent into the cytoplasms of healthy adjacent cells in the 18b-glycyrrhetinic acid also decreases the level of cell monolayer. The permeability of gap junctions (about death 1 kD) limits the list of putative candidates to Ca2+, In order to verify our interpretation of the potassium, IP3, and ATP. Previous studies indicate experiments using dominant-negative Cx43 mutants that, at least in neural tissue, Ca2+ ions are the most to abrogate cell coupling, and to exclude any likely messengers of death (Badd and Lipton, 1998). possible e€ects of enforced expression of Cx43 In order to verify whether Ca2+ is the most likely protein on cell growth per se, we conducted cell-killing signal in BC31 cells, rather than other experiments using the GJIC inhibitor 18b-glycyrrhe- molecules of bigger size that spread through gap tinic acid (bGA). This compound was shown to junctions, we studied the e€ect of selectively restricting block cell coupling fully (Davidson et al., 1986), GJIC permeability in BC31 cells using the sleep- presumably by disassembling gap junctional channels inducing agent cis-9-Octadecenamide (Oleamide). This (Guan et al., 1996); the inhibitory e€ect on GJIC con®nes the permeability of gap junc- lasts for many hours and apparently has no direct tions to calcium waves (Guan et al., 1997). Our e€ect on cell proliferation. experiments showed that treatment of BC31 cells with Unlike the well-coupled control BC31 cells, BC31 100 mM of Oleamide e€ectively blocks the spreading of cells treated with bGA at a dose (70 mM), which was LY and that this e€ect lasts at least 24 h. Further not toxic, but e€ective in disrupting cell coupling, studies showed, however, that treatment of BC31 cells did not form either clusters of dying cells, or defects by Oleamide at this dose did not prevent the in the cell monolayer (Figure 8a,b). A quantitative formation of either clusters of dying cells or defects analysis revealed that the bGA-treated cells remained in the cell monolayer. Furthermore, estimation of the attached to the monolayer after con¯uency a lot number of cells remaining in the cell monolayer did better than non-treated control cells (Figure 8c). not reveal any signi®cant di€erence between Oleamide-

Figure 7 Expression of the dominant-negative Cx43 mutation abrogates cell coupling and a€ects BC31 cell growth and death. The cell-coupling capacity of parental BC31 cells (a) is shown by the spread of LY dye after microinjection of LY into a single cell (see insertion, bar 500 mm). By contrast, clones of the BC31 cell line that express a dominant-negative mutant of Cx43 (b) show a lack of cell coupling after LY microinjection into a single cell (see insertion, bar 500 mm). After con¯uency, defects arise in the monolayer of parental BC31 cells due to intensive coordinated cell loss by clusters (a). Such defects are not seen after con¯uency in a culture of BC31 cells that express the dominant-negative mutant of Cx43 (b). A quantitative assessment of cell growth (c, see Materials and methods) shows that cell loss of parental BC31 cells occurs from day 4 to day 6 of cell culture; the cells shown in (a) and (b) correspond to day 6 of the experiment shown in (c). (bars 500 mm)

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1995

Figure 8 18 b-glycyrrhetinic acid abrogates cell coupling and reduces cell loss in post-con¯uent cultures of BC31 cells. Cultures of non-treated control BC31 cells (a) and BC31 cells treated with b-glycyrrhetinic acid (bGA) (b) are shown after 6 days of culture, 2 ± 3 days after reaching con¯uency (bars 500 mm). The insertions illustrate the inhibitory e€ect of bGA on cell coupling, as revealed by LY injection (bar 500 mm); the main panels show that bGA treatment during exponential phase of growth and after con¯uency (days 2 ± 6 of experiment) abolishes the formation of the defects in the cell monolayer that are seen everywhere in communication pro®cient non-treated control BC31 cells. Growth curves (c) show that bGA treatment did not a€ect the rate of cell growth during the exponential phase, but decreased both the level of con¯uency (days 3 and 4 of culture) and intensity of cell loss afterwards (days 4 ± 6 of culture). (Vertical arrows indicate changes of medium to ensure a supply of active bGA) treated cells and non-treated coupled BC31 cells noteworthy that coupling between apoptotic and (Figure 9). non-apoptotic cells is a two-way process: after microinjecting LY into an individual healthy cell that was situated next to a morphologically-distinguishable Discussion single apoptotic cell in the monolayer, we found that LY could spread into the apoptotic cell (not shown). In this study we obtained several lines of experi- Apparently, a transient functional coupling between mental evidence that gap junction intercellular dying and healthy cells nevertheless suces to trigger communication in cancer cells is a means of irreversible processes, which eventually result in the propagating cell death. Speci®cally, using the rat death of these adjacent, previously healthy, cells. Our bladder carcinoma BC31 cell line, we found that and others' (Cortina et al., 1998) observation that only individual cells, which spontaneously commit to die, communicating cells die in culture by clusters, could be coupled with non-apoptotic counterparts. corroborates this suggestion. It also appears, however, Apparently early apoptotic events do not alter gap that the spread of killing signals through gap junctions junctions; Cx43, which is the only connexin to is somehow limited by protective mechanisms which provide coupling in these cells (Asamoto et al., determine the size of the clusters of dying cells: even in 1994; Krutovskikh et al., 1998), was found in very well-coupled cells, the propagation of death never apoptotic cells and was neither degraded nor became totally generalized. translocated from its normal location in the lateral Ca2+ has a role in both the initiation and regulation plasma membrane into the cytoplasm. Furthermore, of apoptosis (McConkey and Orrenius, 1997). Very the spontaneous loading of apoptotic cells by Lucifer high concentrations of Ca2+, e.g. due to its in¯ux from yellow dye con®rmed that dying cells could be extracellular space as a result of metabolic inhibition or functionally coupled with their healthy neighbors. oxidative insult, or after treatment of cells with Ca2+ Apparently, communication between apoptotic and ionophores (Martikainen and Isaacs, 1990), may non-apoptotic cells is a short-lived event, as only a trigger apoptosis. Thus, it is very likely that the few of the numerous individual dying cells, which spontaneous apoptosis of individual BC31 cells is 2+ were able to take up LY from the medium during associated with a large increase in [Ca ]i concentra- 5 min of exposure to this dye, could then spread it tion. Due to coupling with neighboring cells, a single into the normal cells surrounding them. It is cell that spontaneously dies by apoptosis, could spread

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1996

Figure 9 Partial inhibition of GJIC of BC31 cells by Oleamide treatment does not prevent cell loss after con¯uency. Oleamide treatment from days 2 ± 6 of BC31 cell culture e€ectively inhibited cell coupling for Lucifer yellow (b, insertion, bar 500 mm), but did not eliminate the formation of defects in cell monolayer (b), or modulate the rate of cell growth during days 1 ± 4 of culture or prevent cell loss after achieving con¯uency (days 4 ± 6) (c). (bars 500 mm)

excess Ca2+ into the cell monolayer, i.e. generate a However, our data from the experiments using the Ca2+ wave. The possibility that Ca2+ waves spread sleep-inducing lipid Oleamide, show that restricting gap through gap junctions may kill cells has been shown junction permeability to molecules of the size of Ca2+ before in a study on the development of neural tissue ions, does not prevent the spreading of cell death (Walszon et al., 1994). It is likely that the limitation of signals through gap junctions, thus indirectly implicat- the size of the focus of dying cells is, at least partially, ing Ca2+ ions in this process. These data also exclude 2+ due to the relatively slow speed with which Ca waves the implication of IP3 in mediating of spreading of spread through gap junctions (Jorgensen et al., 1997). Ca2+ waves, which kill cells. In addition, the passage of a Ca2+ wave through gap Apparently gap junctions are not the only means of junctions does not provoke Ca2+ release from propagating cell death in cell cultures. Thus, gap intracellular stores. Therefore, one may suggest that junction-independent clustering of apoptotic cells, due after passing through several cells the intensity of a to the release of toxic compounds from an initially Ca2+ wave is greatly decreased. In addition, it is single dying cell into the culture medium, has been known that gap junction gating is Ca2+ dependent described, and hydrogen peroxide was suggested as the (Pfahnl and Dahl, 1999). Thus cells that are a few cells death signal (Reznikov et al., 2000). away from the original, spontaneously-apoptosing cell, The capacity of cells to kill each other through gap may receive an attenuated Ca2+ wave and have time to junctions has been shown before in co-culture experi- respond by closing their gap junctions before receiving ments and was named `bystander death'. For instance, enough Ca2+ to trigger their own apoptosis. It should glioma cells that resisted apoptosis due to the enforced be noted, however, that cell coupling has been shown expression of Bcl2 could, nevertheless, be killed by to have a role in protecting against cell injury, by di€erent injuries (metabolic inhibition, oxidative reducing the concentrations of harmful molecules in stress), when they were coupled via gap junctions with 2+ cytosol or by the maintenance of [Ca ]i homeostasis vulnerable non-transfected counterparts (Lin et al., (Blanc et al., 1998). 1998). This phenomenon is being pursued for its A direct demonstration of the cell killing power of potential bene®t in cancer therapy. Despite a low Ca2+ would be to show that introducing a high eciency of transfecting cancer cells with the herpes concentration of these ions into the cytoplasm of a simplex virus thymidine kinase (HSV-tk)-gene, which single healthy cell indeed kills it. Unfortunately, all our converts the anti-tumor compound ganciclovir from a attempts to demonstrate this by microinjecting a highly prototoxin into an active form, it has been shown that concentrated solution of Ca2+ have been, so far, the therapeutic e€ect of this compound could be unsuccessful, probably due to the limitations of essentially exacerbated by cell coupling, as the toxic microinjection manipulation, which does not allow metabolite of ganciclovir produced in individual HSV- the introduction of Ca2+ in amounts optimal to kill the tk gene-bearing cells spreads via gap junctions into recipient cell. numerous cells which themselves lack the HSV-tk gene,

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1997 and eventually kills them (Mesnil et al., 1996; Estin et Immunostaining al., 1999). Similar `bystander cell death' or `fratricide' phenom- After removing the culture medium, cells were brie¯y washed 2+ ena have been observed during the development of foci with Ca Hanks solution, ®xed in 4% paraformaldehyde on of ischemia-induced brain damage, when cell death was PBS for 10 min at room temperature, washed with PBS several times and permeabilized with 0.5% Triton X-100 found to spread slowly away from the core of infarct solution for 10 min. After incubation with primer antibodies (Cortina et al., 1998), presumably by means of gap against Cx43, immunoreactivity was revealed using the TSA junction intercellular communication, but again, up to (Tyramid signal ampli®cation)-direct technique from NEN1 certain limits (Badd and Lipton, 1998). Furthermore, Life Science Products, according to the manufacturer's inhibiting GJIC by administrating the gap junction instructions. Counterstaining with Propidium Iodide helped inhibitors, octanol and hepthane, signi®cantly reduced to reveal cells with morphological alterations of their nuclei infarct volume during experimental focal brain stroke that are typical of apoptosis. (Rawanduzy et al., 1997). There is a large body of evidence that gap junction Immunoblotting intercellular communication plays an essential role in Apoptotic and non-apoptotic cells, collected as described tumor suppression. Until now, the tumor suppressive above, were lysed in a bu€er containing sodium dodecyl e€ect of the gap junction proteins known as Connexins sulfate, 2%; dithiothreitol, 0.1 M; Tris HCl, 60 mM, 6.8 pH; was attributed to their capacity to inhibit cell growth, phenylmethylsulfonyl ¯uoride, 1 mM and glycerol, 10%. The which in some cases directly correlates with cell ± cell samples were sonicated for 30 s, centrifuged, resolved by communication (Zhang et al., 1998), or is, at least, electrophoresis on a 10% polyacrylamide gel and blotted contingent upon the formation of gap junctions (Mehta onto a nitrocellulose membrane. Preincubation of the et al., 1999). Now we show that in cancer cells membrane and its subsequent exposure to the primary and intercellular communication via gap junctions propa- secondary antibodies and washing were done with bu€er gates cell death and thereby contributes to the tumor containing 0.1% Tween 20 in 100 mM Tris, 0.9% sodium suppression mediated by Connexins, particularly when chloride, pH 7.5. A positive reaction with antibody was revealed by NiCl2-DAB staining after incubation with their direct inhibitory e€ect on cell proliferation is not peroxidase-conjugated goat anti-rabbit antiserum (Sigma evident. Co., St. Louis, MO, USA).

Microinjection dye transfer assay In order to estimate cell coupling quantitatively, microinjec- tions of 5% Lucifer yellow CH solution in 0.33 lithium Materials and methods M chloride were performed essentially as described previously (Fitzgerald et al., 1983). BC31 cells culture The rat BC31 cell line, established from a primary Functional assay to reveal heterologous coupling between chemically-induced rat bladder transitional cell carcinoma apoptotic and non-apoptotic BC31 cells in monolayer (Asamoto et al., 1994), was maintained in culture in Dulbecco's modi®ed Eagle's medium supplemented with To estimate the possibility of functional heterologous 10% fetal bovine serum on plastic under 5% CO2/95% air coupling between individual apoptotic and surrounding at 378C in a humidi®ed incubator. non-apoptotic BC31 cells, the former must be speci®cally loaded by a tracer dye speci®c for gap junctions such as Lucifer Yellow. We achieved this by exposing the monolayer Preparation of fractions of apoptotic and non-apoptotic cells of BC31 cells to PBS with 5% of Lucifer yellow dye for It was possible to collect selectively the fraction of dying cells 5 min and then washing the cell monolayer three times in from cultures of BC31 cells because one of the features of PBS. The increased permeability of the lateral membranes of these apoptotic cells is their poor attachment to the surface of dying cells allowed LY to get into their cytoplasms; by the culture dish, and they could be detached from the contrast no signs of spontaneous penetration of LY into the monolayer of healthy cells by intensive pipetting. Microscopic cytoplasms of healthy cells in the monolayer were seen. examination revealed that intensive pipetting e€ectively Immediately after this exposure the cell monolayer was removed only those cells which were poorly attached to the examined by inverted ¯uorescence microscope. surface of the culture, and did not a€ect the cellular monolayer. The cells left attached to the culture dishes after Detection of apoptosis this manipulation represented a healthy population of BC31 cells and were collected separately by mechanical scraping. ± by in situ staining Non-®xed BC31 cells were stained in DNA and protein samples were extracted from each of these situ by AnnexinV/Propidium Iodide using the Annexin-V- cell fractions and were analysed by electrophoresis. FLUOS Staining kit (Boehringer Mannheim GmbH, Ger- many) according to the manufacturer's protocol. Immediately after this procedure the stained cells were examined under an Antibodies inverted microscope with epi-¯uorescent equipment. Polyclonal primary anti-Cx43 antibodies, raised in a rabbit against a synthetic peptide corresponding to residues 363 ± ±by the pattern of DNA degradation DNA samples from 382 of native Cx43, were provided by Dr E Rivedal, Oslo, di€erent cell fractions, collected from cell culture as described Norway. before, were resolved in agarose gel and stained by Ethidium

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1998 Bromide to reveal whether the DNA was degraded in the ± by cis-9-Octadecenamide (Oleamide) The sleep-inducing ladder pattern typical of apoptotic cells. lipid cis-9-Octadecenamide (Oleamide) has been described as a compound which restricts the permeability of gap junctions to Ca2+ ions (Guan et al., 1997). To obtain this e€ect on Inhibition of GJIC BC31 cells, Oleamide was dissolved in DMSO and supplied ± by dominant negative mutant of Cx43 As we described to the culture medium at a ®nal concentration of 100 mM;in previously (Du¯ot-Dancer et al., 1997, Krutovskikh et al., an initial experiment this concentration of Oleamide was 1998, 2000), several mutations of Cx proteins have been found to be non-toxic for BC31 cells, but e€ective in blocking characterized as dominant negative because on being co- the intercellular passage of Lucifer yellow dye at least for expressed with wild type Cx these communication-defective 24 h. To obtain stable long-lasting inhibition of cell ± cell Cx mutants are able to interfere with the functional normal communication the medium with freshly supplied Oleamide Cx protein and deactivate its function. One such Cx43 was changed daily. mutant with a strong dominant negative property has seven residues deleted from its cytoplasmic loop at positions 130 ± Growth rates of cells in vitro 136 (Krutovskikh et al., 1998). Communication-de®cient clones of BC31 cells stably transfected with this Cx43 mutant Cells (56104) were seeded on a 35-mm culture dish with were used in the study. supplemented medium, as described before, at day 0 of the experiment. Cells from three dishes were trypsinized and ±by 18 b-glycyrrhetinic acid Bearing in mind that enforced counted independently every next day of experiment. expression of Cx protein could have an e€ect on cell growth per se, independent of its modulation of cell coupling Statistical analyses capacity, we also inhibited intrinsic GJIC in BC31 cells using the known gap junction blocking agent ± 18 b-glycyrrhetinic All in vitro experiments were repeated three times. Statistical acid (Davidson et al., 1986). An initial experiment showed analyses were performed using a commercial statistical that a 70 mM concentration of 18 b-glycyrrhetinic acid analysis package (GraphPad Prizm 3.0). e€ectively inhibits GJIC between BC31 cells, without being cytotoxic. We also found, that exposure of BC31 cells to the medium supplied with 70 mM of 18 b-glycyrrhetinic acid e€ectively inhibited their coupling for approximately 24 h. In order to achieve inhibition of cell coupling throughout the several days required to assess the e€ects on growth rate, the medium was replaced with fresh medium containing 18 b- Acknowledgments glycyrrhetinic acid on a daily basis. We thank Dr J Cheney for editing this manuscript.

References

Asamoto M, Takahashi S, Imaida K, Shirai T and Krutovskikh V, Yamasaki H, Tsuda H and Asamoto M. Fukushima S. (1994). Carcinogenesis, 15, 2163 ± 2166. (1998). Mol. Carcinogen., 23, 254 ± 261. Blanc EM, Bruce-Keller J and Mattson MP. (1998). J. Krutovskikh V, Troyanovsky S, Piccoli C, Tsuda H, Neurochem., 70, 958 ± 970. Asamoto M and Yamasaki H. (2000). Oncogene, 19, Badd SL and Lipton SA. (1998). Nat. Neurosci., 1, 431 ± 432. 505 ± 513. Chipman JK. (1995). Hum. Exp. Toxicol., 14, 539 ± 540. Li H, Liu TF, Lazrak A, Peracchia C, Goldberg GS, Lampe Cortina ML, Kang J, Lin JH-C, Bueno E, Hansen TW, He L, PD and Johnson RG. (1996). J. Cell Biol., 143, 1019 ± Liu Y and Nedergaard M. (1998). J. Neurosci., 18, 2520 ± 1030. 2537. Lin JH, Weigel H, Cotrina ML, Liu Sh, Bueno E, Hansen AJ, Davidson JS, Baumgarten IM and Harley EH. (1986). Hansen TW, Goldman S and Nedergaard M. (1998). Nat. Biochem. Biophys. Res. Commun., 143, 29 ± 36. Neurosci., 1, 494 ± 500. Du¯ot-Dancer A, Mesnil M and Yamasaki H. (1997). Loewenstein WR and Rose B. (1992). Semin. Cell Biol., 2, Oncogene, 15, 2151 ± 2158. 59 ± 79. EstinD,LiM,SprayDandWuJK.(1999).Neurosurgery, Martikainen P and Isaacs J. (1990). Prostate, 17, 175 ± 187. 44, 361 ± 368. McConkey DJ and Orrenius S. (1997). Biochem. Biophys. Fitzgerald DJ, Knowles SE, Ballard FJ and Murray AW. Res. Comm., 239, 357 ± 366. (1983). Cancer Res., 43, 3614 ± 3618. McDonnell TJ. (1993). Mol. Carcinogen., 8, 209 ± 213. Guan X, Cravatt BF, Ehring GR, Hall JE, Boger DL, Lerner MehtaPP,Perez-StableC,NadjiM,MianM,AsotraKand RA, and Gilula NB. (1997). J. Cell Biol., 139, 1785 ± 1792. Roots BA. (1999). Dev. Genet., 24, 91 ± 110. Guan X, Wilson S, Schelender KK and Ruch RJ. (1996). Mesnil M, Piccoli C, Tiraby G, Willecke K, and Yamasaki Mol. Carcinogen., 16, 157 ± 164. H. (1996). Proc.Natl.Acad.Sci.USA,93, 1831 ± 1835. John SA, Kondo R, Wang SY, Goldhaber JI and Weiss JN. Pfahnl A and Dahl G. (1999). Pfugers Arch., 437, 345 ± 353. (1999). J. Biol. Chem., 274, 236 ± 240. Rawanduzy A, Hansen A, Hansen TW and Nedergaard M. Jorgensen NR, Geist ST, Civitelli R and Steinberg TH. (1997). J. Neurosurg., 87, 916 ± 920. (1997). J. Cell Biol., 139, 497 ± 506. Reznikov K, Kolesnikova L, Pramanik A, Tan-no K, Gileva Krutovskikh V and Yamasaki H. (1997). Histol. Histo- I, Yakovleva T, Rigler R, Terenius L and Bakalkin G. pathol., 12, 761 ± 768. (2000). FASEB J., 14, 1754 ± 1764.

Oncogene Gap junctions propagate cell death VA Krutovskikh et al 1999 Walszon LR, Rehder V, Kater SB and Macagno ER. (1994). Yamasaki H and Naus CC. (1996). Carcinogenesis, 17, 1199 ± 1213. J. Neurosci., 14, 3437 ± 3448. Zhang ZQ, Zhang W, Wang NQ, Bani-Yaghoub M, Lin ZX Yamasaki H. (1990). Carcinogenesis, 7, 1051 ± 1058. and Naus CC. (1998). Carcinogenesis, 19, 1889 ± 1894.

Oncogene