Enhanced Neurite Outgrowth in PC12 Cells Mediated by Connexin Hemichannels and ATP*

Gap junctions have traditionally been described as transmembrane channels that facilitate intercellular communication via the passage of small molecules. Connexins, the basic building blocks of gap junctions, are expressed in most mammalian tissues including the developing and adult central nervous system. During brain development, connexins are temporally and spatially regulated suggesting they play an important role in the proper formation of the central nervous system. In the current study, connexins 32 and 43 were overexpressed in PC12 cells to determine whether connexins are involved in neuronal differentiation. Both connexin 32 and 43 were appropriately trafficked to the cell membrane following overexpression and resulted in the formation of functional gap junctions. Connexin overexpression was found to cause enhanced neurite outgrowth in PC12 cells treated with nerve growth factor to initiate neuritogenesis. Surprisingly, however, enhanced neurite outgrowth was found to be the consequence of functional hemichannel formation as opposed to traditional intercellular communication. Additional analysis revealed that ATP was released into the media likely through hemichannels and acted on purinergic receptors to cause enhanced neurite outgrowth. Collectively, the results of the current study suggest that connexins may play an important role in neuronal differentiation by non-traditional mechanisms.

Gap junctions have traditionally been described as transmembrane channels that facilitate intercellular communication via the passage of small molecules. Connexins, the basic building blocks of gap junctions, are expressed in most mammalian tissues including the developing and adult central nervous system. During brain development, connexins are temporally and spatially regulated suggesting they play an important role in the proper formation of the central nervous system. In the current study, connexins 32 and 43 were overexpressed in PC12 cells to determine whether connexins are involved in neuronal differentiation. Both connexin 32 and 43 were appropriately trafficked to the cell membrane following overexpression and resulted in the formation of functional gap junctions. Connexin overexpression was found to cause enhanced neurite outgrowth in PC12 cells treated with nerve growth factor to initiate neuritogenesis. Surprisingly, however, enhanced neurite outgrowth was found to be the consequence of functional hemichannel formation as opposed to traditional intercellular communication. Additional analysis revealed that ATP was released into the media likely through hemichannels and acted on purinergic receptors to cause enhanced neurite outgrowth. Collectively, the results of the current study suggest that connexins may play an important role in neuronal differentiation by non-traditional mechanisms.
Gap junctions form transmembrane channels through the plasma membranes of adjacent cells providing an intercellular link from the cytoplasm of one cell to the next. Traditionally, gap junctions were thought to mediate cell-cell communication through the cytoplasmic exchange of molecules such as Ca 2ϩ and ATP (1). However, recent studies have shown that gap junctions may serve additional functions within the cell including regulating gene expression and the release of small signaling molecules through hemichannels directly into the extracellular environment (2). Gap junctions are composed of oligmerized connexin (Cx) 2 proteins and more than 20 different mammalian Cx genes have been identified to date (3). Cx genes are expressed in most mammalian tissues although not every individual Cx is expressed in every tissue.
In the central nervous system, the expression of several different Cx proteins, including Cx26, 29,30,32,36,43,45, and 47, has been reported (4,5). In the adult brain, specific Cx proteins tend to be enriched in particular cell types. For example, Cx43 is highly expressed in astrocytes, whereas Cx29 and Cx32 have been found in oligodendrocytes (6 -8). Recently, Cx36 and Cx45 were identified as the predominant connexins expressed in adult neurons (9 -11). Deletion of either of the adult neuronal Cx genes, Cx36, or Cx45, results in deficiencies in signaling and transmission properties of neurons (12)(13)(14).
In the developing central nervous system, the role of gap junctions in neurons or cells destined to become neurons is less clear. Studies have shown that Cx genes such as Cx32 and Cx43 are highly expressed in the developing brain and that distinct regional and temporal expression patterns are formed by different Cx proteins (15)(16)(17)(18)(19). For example, in the developing neocortex, Cx32 protein expression is most abundant in the later stages of development and found to be localized in areas of myelinated axons and clusters of developing neurons (16,19). Cx43, in contrast, is expressed throughout cortical development in both neurons and radial glia (16,19,20). Radial glia were classically thought to provide a temporary structural framework for neuronal migration. However, recent evidence suggests that radial glia are in fact neuronal progenitor cells that later become excitatory cortical neurons (21)(22)(23). Interestingly, the extracellular release of ATP via Cx hemichannels in radial glia was shown to promote cortical neural progenitor cell proliferation (24). Cx32 and Cx43 were also found to be temporally regulated in dopaminergic neurons of the midbrain at times that coincided with the development of the forebrain dopaminergic projection system (18). Collectively, the results suggest that Cx proteins may play an important role in proper central nervous system development including the migration and development of neurons.
Consistent with the premise that Cx proteins may play a prominent role in neuronal development, several recent studies conducted in vitro have demonstrated that connexins and gap junctions affect neuronal differentiation. For example, human NT2/D1 and rat P19 cells can be differentiated to obtain neurons using retinoic acid (25,26). Both cell lines express Cx43 and are coupled (27,28). Treatment of NT2 and P19 cells with 18 ␣-glycyrrhetinic acid or carbenoxolone to block connexon channels resulted in a dramatic decrease in the number of mature neurons following retinoic acid differentiation (29,30).
To help further elucidate the role of Cx genes in neuronal differentiation, Cx32 and Cx43 were overexpressed in PC12 (pheochromocytoma) cells. PC12 cells are neural crest-derived cells that can be differentiated to form neurons following treatment with nerve growth factor (NGF) (31). The advantage of PC12 cells compared with other cell lines is that they only form neurons and not other neural cell types upon NGF differentiation. Furthermore, they do not contain appreciable levels of Cx32 or Cx43 (10,32) and are readily infected with retroviral constructs (33). Overexpressed Cx32 or Cx43 is trafficked appropriately in PC12 cells resulting in functional gap junction formation. In the current study we report that constitutive expression of Cx32 or Cx43 in PC12 cells resulted in enhanced neurite outgrowth upon treatment with NGF. Enhanced neurite outgrowth is the consequence of hemichannel formation and the release of ATP into the extracellular milieu. Collectively the results suggest that Cx genes may play an important role in the process of neuronal differentiation.

EXPERIMENTAL PROCEDURES
Cell Culture-PC12 cells obtained from ATCC were maintained in high glucose DMEM supplemented with 10% horse serum, 5% fetal bovine serum, and penicillin (100 units/ml)/ streptomycin (100 g/ml) in a 37°C humidified incubator containing 5% CO 2 . For all of the differentiation assays, PC12 cells were treated with NGF (50 ng/ml) in defined media (DMEM containing N2 supplement and penicillin/streptomycin). The NGF and N2 media were replaced every 2 days.
All of the media, serum, and supplements used throughout the study were obtained from Invitrogen unless otherwise noted. Cells were grown on Corning tissue culture plates and NGF was obtained from Cedarlane Laboratories (Hornby, Ontario). The 293GPG retroviral packaging cell line (34) was a generous gift from Dr. Richard C. Mulligan (Children's Hospital, Boston, MA). The packaging cells were maintained in high glucose DMEM supplemented with 10% fetal bovine serum, 100 units/ml penicillin, 100 g/ml streptomycin, 1.0 g/ml tetracycline (Sigma), 2.0 g/ml puromycin, and 0.3 g/ml geneticin in a humidified 37°C incubator containing 5% CO 2 .
Constructs and Generation of Retrovirus-The Cx32, Cx32-EGFP, Cx43, Cx43-EGFP, and EGFP constructs were generated through directional subcloning of the cDNAs of interest into the AP2 vector with the original IRES EGFP site excised. Orientation of the inserts was confirmed by bidirectional sequencing and high quality cDNA necessary to generate virus obtained from transformed bacteria using a Qiagen Midiprep Kit (Mississauga, Ontario).
Retrovirus containing Cx, Cx-EGFP, or EGFP was generated as previously described (35). Briefly, 293GPG packaging cells were plated at a density of 6 ϫ 10 6 cells/100-mm cell culture plate overnight and then transfected with the constructs of interest using Lipofectamine 2000 reagent (Invitrogen). OPTI-MEM I media (3 ml total volume) supplemented with 1 g/ml of tetracycline was used during the transfection procedure in place of serum containing medium. After 6 h, PC12 media supplemented with 1 g/ml of tetracycline was added to the plates and the cells were left overnight in the incubator. The following day, the medium was replaced with fresh PC12 medium (without tetracycline) to allow the cells to begin to generate virus. The medium was collected and filtered every 24 h using a 0.45-m syringe-mounted filter (Gelman Sciences, Ann Arbor, MI), then stored at Ϫ80°C. After the final collection, medium was thawed, pooled, aliquoted, tested, and stored at Ϫ80°C until needed.
The AP2 retroviral vector was kindly provided by Dr. J. P. Additional virus containing media prepared using Cx43-G21R and Cx43-G138R mutant constructs was generously provided by Dr. D. Laird. The G21R and G138R mutants are able to traffic to the membrane appropriately but are unable to form functional Cx43 channels (37).
Confirmation of Expression and Infection Efficiency-To estimate infection efficiency, PC12 cells were plated at a density of 5 ϫ 10 6 cells on collagen type I (10 g/ml)-coated 60-mm plates containing glass coverslips the day prior to infection. The following day, cells were infected with EGFP, Cx32, Cx32-EGFP, Cx43, or Cx43-EGFP retrovirus for 48 h and then fixed with 70% ethanol containing 0.15 M NaCl at room temperature for 10 min. Cx32 and Cx43 proteins were detected using immunocytochemistry with anti-Cx32 (Zymed Laboratories Inc., 1:200) or anti-Cx43 (Zymed Laboratories Inc., 1:100) monoclonal antibodies, respectively. Cx32 and Cx43 were visualized using Alexa Fluor-conjugated secondary antibodies (Invitrogen, 1:200). All of the infected cells were counterstained with Hoechst 33342 (100 ng/ml) to visualize nuclei and aid in the quantification of infection efficiency. The coverslips were mounted onto glass slides using Vectashield mounting medium (Vector Laboratories, Burlingame, CA). Images were captured from 5 fields on an Axiovert S100 microscope (Zeiss). The number of cells expressing the retroviral protein was determined and expressed as a percentage of the total number of cells. The experiments were conducted in 5 separate replicates using the same lot of virus. In addition to determining infection efficiency, Cx32-EGFP-or Cx43-EGFP-infected cells were also stained using immunocytochemistry for Cx32 and Cx43, respectively, to determine whether the pattern of immunostaining and EGFP labeling matched.
To confirm appropriate expression of Cx, protein was extracted from native PC12 cells, PC12 cells overexpressing EGFP alone, or PC12 cells overexpressing the various Cx constructs. The protein extracts were used to generate Western blots and probed for either Cx32 (Sigma; dilution 1:600) or Cx43 (Sigma; dilution 1:8000) as appropriate. Cell lysis, electro-phoresis, Western blotting, and visualization of the proteins of interest were conducted as outlined previously (33).
Microinjection and Coupling-PC12 cells were plated on collagen-coated 60-mm plates, at a density of 5 ϫ 10 6 cells/plate (2 days post-infection analysis) or 2 ϫ 10 6 cells/plate (96-h post-NGF treatment analysis). Cells were infected and analyzed for functional gap junction coupling 2 days post-infection (EGFP, Cx32, Cx32EGFP, Cx43, and Cx43EGFP) or 4 days following NGF treatment (EGFP, Cx32, and Cx32EGFP). Individual cells infected with Cx32 or Cx43 were pressure injected with 2.5% Lucifer Yellow, whereas cells infected with EGFP, Cx32-EGFP, or Cx43-EGFP were injected with Alexa Fluor 594 dye (Molecular Probes, Eugene, OR) using an Eppendorf microinjection system and an Axiovert S100 microscope (Zeiss). Injection time was set for 10 s. Cells were examined 1 min following dye injection to determine whether dye had passed from the injected cell to neighboring cells. Cells were considered to be coupled if the injected cell passed dye to at least one adjacent cell. The percentage of coupled cells was defined as the [number of cells that passed dye divided by the total number of injected cells] ϫ 100. The order of coupling was determined by counting the number of cells receiving dye from the injected cell, to the farthest cell that received dye.
Neurite Outgrowth of PC12 Cells-Cells were plated at a density of 8 ϫ 10 4 cells/plate (60 mm, collagen coated) and infected the following day with EGFP, Cx32, Cx32-EGFP, Cx43, Cx43-EGFP, Cx43-G21R, or Cx43-G138R retrovirus. Two days postinfection the cells were treated with NGF (50 ng/ml) in N2supplemented DMEM (Invitrogen). NGF and media were refreshed every second day. The number of neurites and neurite length was examined 2 days following infection (time 0) and at 1, 2, 3, and 4 days after the commencement of NGF. After each time point the cells were fixed using 70% ethanol containing 0.15 M NaCl and analyzed using an Axiovert S100 microscope and Northern eclipse software. Neurite length was measured in terms of the number of cell body widths. For quantification of the number of neurites, a neurite was defined as an extension from the cell body equivalent or greater than 1ϫ the cell body diameter (38 -40). Quantification of neurite length and the number of neurites was performed by examining 20 fields of cells, in five separate experimental replicates. Data were analyzed using an analysis of variance followed by Tukey's Multiple Comparison Test.
Results obtained from measuring neurite length suggested that overexpression of Cx proteins alone was sufficient to cause some autodifferentiation of the PC12 cells. Therefore neurite outgrowth in Cx-infected PC12 cells untreated with NGF was further evaluated. Cells were infected with EGFP, Cx32, Cx32-EGFP, Cx43, or Cx43-EGFP retrovirus, and neurite length was measured 48, 72, and 120 h following infection (48 h ϭ time 0 in the NGF experiments).
Co-culture Experiments-Microinjection experiments using the Cx32-EGFP and Cx43-EGFP constructs revealed that only cells expressing the Cx proteins were functionally coupled. However, during analysis of neurite length it was observed that all PC12 cells measured from the Cx-infected culture plates showed enhanced neurite outgrowth although ϳ20% of the cells appeared not to express Cx as determined by the transfec-tion efficiency analysis. To evaluate the possibility that gap junctions might cause enhanced neurite outgrowth by a mechanism other than intercellular communication, PC12 cells were infected with EGFP, Cx32-EGFP, or Cx43-EGFP. Two days post-infection the cells were lifted, counted, and re-plated as co-cultures of EGFP and Cx32-EGFP or EGFP and Cx43-EGFP (1:1). The cells were then treated with NGF for 0, 24, or 96 h and neurite length was measured in EGFP and Cx-EGFP expressing cells.
Treatment of PC12 with ATP Receptor Agonists and Antagonists-Results of the co-culture experiments indicated that enhanced neurite outgrowth associated with Cx overexpression is likely mediated by a diffusible factor in the media. Because gap junctions are known to allow the passage of ATP and purinergic receptors can regulate neurite outgrowth (41,42), ATP agonists and antagonists were examined to determine whether they could affect neurite outgrowth. PC12 cells were treated with either ATP or 2-methylthioadenosine triphosphate (2-MeSATP, P2Y agonist) to determine whether either purinergic receptor agonists could induce neurite growth in the presence of NGF. For these experiments, PC12 cells were plated as indicated previously and infected with EGFP, Cx32-EGFP, or Cx43-EGFP (for direct comparison). EGFP-infected cells were treated with ATP (10 M), 2-MeSATP, or vehicle (water). Following 48 h of treatment, the medium was replaced with N2-supplemented DMEM containing 50 ng/ml NGF. Neurite length was measured prior to the addition of NGF as well as 24 and 96 h following addition of NGF. For the antagonist experiments, the cells were infected with EGFP, Cx32-EGFP, or Cx43-EGFP as outlined above. One experimental replicate was treated with virus alone, whereas the other two replicates received virus plus 100 M suramin (P2X, P2Y antagonist) or virus plus pyridoxal phosphate-6-azo(benzene-2,4-disulfonic acid) (PPADS, P2 antagonist). Forty-eight hours post-infection, NGF (50 ng/ml) was added to the cells and the length of the neurites was measured prior to the addition of NGF (day 0) and 24 and 96 h following addition of NGF.
ATP Measurements-To determine whether ATP was released from the Cx-infected cells, ATP measurements were obtained from the media of EGFP-, Cx32-EGFP-, or Cx43-EGFP-infected cells. The cells were treated as previously described except that N2 media containing NGF lacked the colored indicator phenol red. Twenty-four hours post-treatment with NGF, the medium was collected, cell debris removed by centrifugation, and the conditioned media stored at Ϫ80°C until analyzed. Because overexpression of Cx32 and Cx43 in the absence of NGF caused some limited autodifferentiation 48 h post-infection (undetectable by 96 h), media was also collected from non-NGF-treated cells 48 and 96 h following infection with the control and Cx constructs. Relative ATP measurements were obtained using an ATP assay kit (Invitrogen) according to the manufacturer's protocols. Statistical analysis was conducted using an analysis of variance followed by Tukey's multiple comparison test (n ϭ 4 ϫ 2 experimental replicates, p Ͻ 0.05).
Conditioned Media and Apyrase Treatment-To further confirm that ATP (as opposed to gap junctions) was responsible for the observed enhancement of neurite outgrowth, cells were grown and infected with EGFP, Cx32-EGFP, or Cx43-EGFP overnight. The virus containing media was removed and the cells were gently washed with DMEM three times to remove any residual virus. Serum-free media (DMEM ϩ N2) was then added and the cells were allowed to condition the media for 48 h. The conditioned media was then collected and stored frozen at Ϫ80°C until needed. Untreated PC12 cells were plated as described for the neurite outgrowth studies but not infected. Instead, conditioned media was mixed 1:1 with fresh media and NGF added. Forty-eight hours following the addition of the conditioned media containing NGF, neurite outgrowth was measured as outlined above. As an additional control, NGF was added to native untreated PC12 cells and neurite length was measured after 48 h in parallel with the conditioned media-treated cells to ensure that media nutrient depletion did not affect neurite outgrowth. In addition, some of the conditioned media was treated with apyrase (Sigma; 0.25 units/ml) to de-grade ATP prior to adding it to the cells. Finally, the cells treated with the conditioned media were carefully examined using a fluorescence microscope to ensure that there was no inadvertent transfer of virus and subsequent infection of the native PC12 cells. Protein was also extracted from some additional cells treated with the conditioned media and tested for Cx32-EGFP and Cx43-EGFP expression by Western blotting (data not shown).
Blocking Cx43 Channels-To demonstrate that functional Cx hemichannels are necessary for enhanced neurite outgrowth, PC12 cells were infected with EGFP, Cx32, or Cx43 as outlined previously. Fortyeight hours post-infection, NGF (50 ng/ml) was added along with either the gap junction channel blocker carbenoxolone (CBX, 100 M) or the control compound glycyrrhizic acid (100 M). Carbenoxolone is known to inhibit Cx hemichannels (43,44). Neurite outgrowth was quantified 24 h post-treatment. PC12 cells were also infected with two mutant Cx43-EGFP constructs (G21R or G138R) and neurite outgrowth was measured 24 h post-NGF treatment and compared with EGFP and Cx43-EGFPinfected cells treated with NGF.
Dye Uptake Assays-To determine whether Cx overexpression (i.e. autodifferentiation) and NGF treatment had a functional effect on hemichannels, propidium iodide (PI) uptake assays were conducted. PC12 cells were treated with EGFP, Cx32, and/or Cx43 viral containing media for 12 h. The infection media was replaced with N2 media for 36 h and then replaced once again with fresh N2 media with or without NGF added for an additional 48 h. Consequently, EGFP-and Cx43-infected cells untreated with NGF were obtained 48 h post-infection and EGFP-, Cx32-, and Cx43-infected cells treated and untreated with NGF were obtained 96 h post-infection (48 h post-NGF treatment). Following treatment, PI (1 mg/ml final concentration) was added directly to the N2 media for 15 min. The cells were fixed with 2% formaldehyde, rinsed, and examined using a Zeiss Axiovert microscope. Images of representative fields were captured using a Sony 3CCD camera and Northern Eclipse Software (Empix Imaging). Changes in PI labeling intensity were quantified using Kodak Molecular Imaging software and analyzed statistically using an analysis of variance followed by Tukey's multiple comparison test (n ϭ 4 ϫ 3 fields/replicate, p Ͻ 0.05). A, immunoblotting revealed that Cx43 and Cx32 are not present in native or EGFP-infected PC12 cells but that Cx32, Cx32-EGFP, Cx43, and Cx43-EGFP are expressed in abundance in PC12 cells following infection. All Cx proteins and Cx fusion proteins were of the expected relative molecular weight. B, native or virus-infected PC12 cells stained using immunofluorescence for Cx32 (top row) or Cx43 (bottom row). No Cx protein was detectable in the native (PC12) or EGFP expressing (EGFP) PC12 although EGFP was readily detectable (EGFP Label). Cells infected with Cx32/Cx32-EGFP and stained with Cx32 or infected with Cx43/Cx43-EGFP and stained with Cx43 revealed robust punctuate cellular labeling (right panels). C, confocal microscopy demonstrated that Cx32 and Cx43 (representative images for Cx43 are shown) proteins were appropriately trafficked to the plasma membrane but that Cx protein could also be found in the Golgi complex.

Cx32 and Cx43 Overexpression and the Formation of Gap
Junctions-Native PC12 cells and cells infected with EGFP alone did not express any detectable levels of either Cx32 or Cx43 proteins. However, PC12 cells infected with Cx32, Cx32-EGFP, Cx43, or Cx43-EGFP retrovirus expressed readily detectable levels of Cx protein of the appropriate M r (Fig. 1A). Cells infected with Cx or Cx-EGFP fusion proteins and examined histologically using immunofluorescence to detect Cx43 or Cx32 revealed abundant labeling and the expected staining pattern (Fig. 1B). For example, cells infected with Cx43-EGFP showed labeling in the cytoplasm and between adjacent cells when viewed using confocal microscopy (Fig. 1C). Cells infected with Cx43-EGFP or Cx32-EGFP and double labeled using immunofluorescence for Cx43 or Cx32 showed an overlapping distribution that was virtually identical (data not shown). Overexpression of the Cx, Cx-EGFP, or EGFP alone did not cause any obvious changes in morphology compared with the native PC12 cells apart from the enhanced neurite outgrowth described in detail later. Cell counts performed based on the presence or absence of EGFP label or positive/negative staining (immunocytochemistry) for Cx32and Cx43-infected cells revealed that the infection efficiency for the retrovirus used in the current study was ϳ80% (77-84%). Therefore, the data collectively suggests that infection with our Cx constructs results in the efficient, appropriate expression and localization of Cx or Cx-EGFP fusion proteins and that the EGFP tag does not substantially alter the distribution of Cx within the cells. Furthermore, the data indicates that the EGFP tag is highly sensitive and accurate for localizing Cx in the infected PC12 cells.
To determine whether overexpression of Cx32, Cx43, or EGFP fusion proteins results in the formation of functional gap junctions, dye injection using Lucifer Yellow (cells infected with non-EGFP-tagged proteins) or Alexa Fluor 594 (cells infected with EGFP or EGFP tagged proteins) was conducted (Fig. 2). The results revealed that native and EGFPinfected cells were not functionally coupled, whereas PC12 cells overexpressing Cx32 or Cx43 passed Lucifer Yellow dye to adjacent cells. Comparable results were obtained using the Cx-EGFP-infected cells and Alexa Fluor 594. The dye injection results obtained from cells infected with Cx32-EGFP or Cx43-EGFP were further subdivided into 2 categories depending on whether the cell injected with dye also expressed the Cx-EGFP fusion protein. Cells that expressed Cx32-EGFP or Cx43-EGFP were functionally coupled, whereas cells that showed no Cx-EGFP labeling were not. PC12 cells were also tested 96 h following NGF treatment for dye passage to determine whether NGF and the process of differentiating PC12 cells influenced functional coupling. No functional coupling was found following treatment of native or EGFP-treated cells. Cx32, Cx43, Cx32-EGFP, and Cx43-EGFP-infected cells showed functional coupling similar to what was observed in the non-NGF-treated cells. However, the coupling order was significantly greater in the NGF versus non-NGF-treated cells (p Ͻ 0.05). Therefore, the overexpression of gap junctions in PC12 cells result in the establishment of functional coupling in a cell line typically devoid of coupling. , and EGFP-expressing PC12 cells show a complete lack of functional dye coupling. Cells that express Cx32, Cx43, or the Cx-EGFP fusion proteins were significantly coupled to neighboring cells. Interestingly, the level of dye coupling (coupling order) was significantly increased (p Ͻ 0.05) following 96 h of NGF treatment in PC12 cells expressing Cx or Cx-EGFP compared with the non-NGF treated Cx-infected cells.

The Effects of Cx32 and Cx43 Overexpression on Neurite
Outgrowth-The length of neurites was measured in the Cxand EGFP-infected cells prior to the addition of NGF and at different stages following NGF treatment. Significantly longer neurites were observed in the Cx-or Cx-EGFP-infected cells at all time points following NGF treatment (Fig. 3, A-B). Because it is also possible that overexpression of EGFP alone may have retarded neurite outgrowth, native and EGFP cells were compared in a similar fashion. There was no difference in neurite length between EGFP, empty vector-infected, and non-infected PC12 cells at any time following the addition of NGF up to 96 h post-treatment (data not shown). Enhanced neurite outgrowth was also detected in the Cx-infected cells prior to treatment with NGF although the appearance of neurites was transient ( Fig. 3, B-C). No significant difference was detected between the Cx-and Cx-EGFP-infected cells.
One of the interesting observations that arose from the neurite outgrowth experiments was that almost all of the PC12 cells on the Cx-or Cx-EGFP-infected plates appeared to have longer neurites although only cells that expressed Cx-EGFP (ϳ80% of cells) showed coupling by microinjection. Therefore, neurite outgrowth was measured in Cx-EGFP-infected cells and segregated based on the absence or presence of the EGFP tag (Fig. 4,  A-B). PC12 cells on the Cx-EGFPinfected plates showed significantly enhanced neurite outgrowth compared with PC12 cells infected with EGFP. However, no significant difference was observed in neurite length between Cx-EGFP expressing and non-Cx-EGFP expressing cells following treatment with NGF. The results were confirmed using a co-culture approach because it is conceivable that PC12 cells that apparently lack Cx-EGFP may have actually expressed low levels of Cx protein. PC12 cells were infected with Cx-EGFP or EGFP then co-cultured (1:1) and compared with plates of PC12 cells infected with EGFP alone (control). Cells containing Cx-EGFP or EGFP on the co-culture plates showed significantly enhanced neurite outgrowth compared with cells on culture plates infected only with EGFP. No significant difference was detected between the EGFP expressing and Cx-EGFP expressing cells on the co-culture plates with respect to neurite outgrowth (Fig. 4C).
Cx32, Cx43, and ATP-The neurite outgrowth and co-culture data suggests that expression of Cx in PC12 cells results in the release of soluble factor(s) into the media that are responsible for the enhanced neurite outgrowth. Several recent studies have shown that purinergic receptors can regulate neurite outgrowth and that ATP readily passes through gap junctions (41,42). Therefore, a combination of ATP agonists and antagonists were used to determine whether ATP was responsible for the observed effects on neurite outgrowth. The addition of the purinergic receptor blockers PPADS or suramin to the media abolished the enhanced outgrowth observed in the Cx-infected PC12 cultures (Fig.  5A). Neurite length in Cx-infected PC12 cells containing PPADS or suramin and NGF was indistinguishable from control cells (EGFP infected) treated with NGF alone. Suramin also caused a small but significant reduction in neurite length of EGFP and Cx32-EGFP-infected cultures at 96 h post-NGF treatment compared with EGFP-infected, NGF-treated cells (Fig. 5A). When ATP or 2-MeSATP (purinergic receptor agonists) was added to EGFP-infected cells, enhanced neurite outgrowth was observed (Fig. 5B). Neurite length was indistinguishable between Cx-EGFP-infected cells and ATP agonisttreated cells but significantly longer than EGFP-infected PC12 cells at all time points examined.
To confirm that Cx overexpression leads to a release of ATP into the media, an ATP assay using conditioned media was conducted. The results indicate that overexpression of either Cx32 or Cx43 results in an ϳ500% increase in ATP media levels 24 h after the addition of NGF to the media (Fig. 6A). Interestingly, elevated ATP levels (ϳ2-fold increase) were also detected in the media of Cx32-and Cx43-infected cells prior to the addition of NGF at a time when autodifferentiation was observed. However, a significant decrease in ATP levels were measured in the media of the Cx-infected cells 96 h post-infection (Fig. 6A). By 96 h the spontaneous neurite outgrowth observed 48 h following Cx infection had been lost.
To corroborate the link between ATP released into the media and enhanced neurite outgrowth, conditioned media was col-  . Neuritogenesis of Cx expressing PC12 cells involves ATP release and signaling. A, the ATP receptor antagonists, suramin and PPADS were added to cultures of Cx32, Cx32-EGFP, Cx43, or Cx43-EGFP expressing PC12 cells and neurite formation was measured at baseline, 24 and 96 h following NGF treatment. Suramin and PPADS abolished increased neurite length typically seen in the Cx expressing PC12 cells. At 96 h, suramin treatment also resulted in a small but significant decrease in neuritogenesis of EFGP-expressing and Cx32-EGFP-expressing PC12 cells. Asterisk, neurite length significantly increased compared with controls, p Ͻ 0.05. Caret, neurite length significantly reduced compared with controls. B, the addition of ATP or the purigenic receptor agonist, 2-MeSATP, induced the formation of neurites in EGFP-expressing PC12 cells, mimicking the effects of Cx-EGFP expression in PC12 cells.
lected from Cx32-EGFP-, Cx43-EGFP-, and EGFP-infected cells and used to treat native uninfected PC12 cells. Cells treated with the conditioned media showed no EGFP fluorescence (microscopy) and did not express detectable levels of Cx32 or Cx43 protein (Western blotting), indicating that there was no transfer of virus and infection of the native cells (data not shown). Native PC12 cells treated with conditioned media obtained from the Cx32-EGFP-and Cx43-EGFP-infected cells showed significantly increased neurite length compared with the conditioned media obtained from the EGFP-infected cells (Fig. 6B). Pre-treatment with apyrase (an ATP diphosphohydrolase) completely abolished the enhanced neurite outgrowth (Fig.  6B). There was no difference in neurite length between PC12 cells treated with EGFP-conditioned media or fresh unconditioned media, indicating that media nutrient depletion was not a confounding factor (data not shown).
Connexin Hemichannels and Neurite Outgrowth-The neurite outgrowth and ATP experiments suggested that Cx hemichannels were responsible for the release of ATP in the media. To confirm the involvement of Cx channels, the pharmacological blocker CBX or mutant Cx43 constructs were used. The addition of CBX, an agent known to block Cx hemichannels (43,44), to NGF-treated, Cx overexpressing PC12 cells, abolished enhanced neurite outgrowth. EGFPinfected PC12 cells were unaffected by CBX as were EGFP and Cx expressing PC12 cells treated with the control compound glycyrrhizic acid (Fig. 7B). PC12 cells infected with mutant Cx43 constructs that are incapable of forming functional Cx channels, including hemichannels (37,45), also failed to show enhanced neurite outgrowth following the addition of NGF. Overexpression of wild-type Cx43 fused to EGFP, in contrast, caused a significant increase in neurite length (Fig. 7, A and B).
To examine the functional state of Cx hemichannels, PI dye uptake assays were performed. Overexpression of Cx32 or Cx43 followed by treatment with NGF caused a significant increase in the amount of dye taken up by the PC12 cells through Cx hemichannels. Cells untreated with NGF or cells infected with EGFP and treated with NGF only showed a baseline level of labeling (Fig. 8). The exception was Cx43 overexpressing PC12 cells untreated with NGF and tested 48 h following infection (Cx32 was not tested). Significant increases in staining intensity were observed at this time point but not 96 h post-infection. Interestingly, spontaneous neurite outgrowth is observed 48 h post-infection but neurites had retracted by 96 h post-infection (Fig. 8). Additional experiments were conducted using CBX and the mutant Cx43 constructs confirming that blocking hemichannels through either method reduced the level of dye uptake to baseline (data not shown).

DISCUSSION
Connexin Expression Enhances Neuritogenesis-Infection of PC12 cells with Cx32 or Cx43 resulted in cells that appropriately express the desired Cx protein. Both Cx32 and Cx43 were properly trafficked to the site of cell-cell contacts and generated functional gap junctions. The extension of neurites from native PC12 cells treated with NGF proceeds steadily for several days but eventually begins to taper off. Expression of Cx32 or Cx43 in PC12 cells substantially enhanced neurite outgrowth resulting in cells with significantly longer neurites than EGFP-expressing or native PC12 cells. Cx-infected PC12 cells even showed some limited autodifferentiation prior to the addition of NGF that is not typically observed in native PC12 cultures. However, by 96 h post-infection, neurites in the Cx expressing cells had retracted. Extension followed by retraction of neurites indicates that autodifferentiation of the Cx expressing cultures is transient and requires NGF to maintain Cx hemichannels in an open state. The increased neurite length observed following infection with Cx32 or Cx43 over the course of the 96-h experiment was not likely the result of autodifferentiation providing the cells with a head start because the difference in neurite length between the Cx-and EGFPinfected cells continued to widen over time.
Connexins Are Downstream Targets of Nerve Growth Factor-Although Cx32-and Cx43-infected PC12 cells were able to pass dye to adjacent cells in the presence or absence of NGF, the coupling order was significantly increased following the addition of NGF. NGF binds to and activates TrkA receptors located in the plasma membrane of PC12 cells (46). PC12 cells treated with NGF show decreased proliferation and begin the process of neuritogenesis (31). Previous studies have shown that sustained activation of several downstream targets of the TrkA receptor, including mitogen-activated protein kinase, is critical for neurite outgrowth and maintenance (47)(48)(49)(50). Recently, we demonstrated that TrkA receptors also cause increased Cx43 phosphorylation via activation of the mitogen-activated protein kinase pathway leading to enhanced cell-cell communication (33). The current study extends our previous observations by showing that NGF can also stimulate the opening of traditional gap junction channels chronically as well as Cx hemichannels suggesting that Cx may play an active role in neuritogenesis in the developing central nervous system. Connexin Hemichannels Release ATP Enhancing Neuritogenesis in PC12 Cells-There is increasing evidence that Cx contribute to cellular signaling through the release of small molecules such as ATP or NAD from hemichannels at sites of non-junctional membranes (2). The results of the current study support this supposition suggesting that enhanced neurite outgrowth observed in Cx32-and Cx43-infected cells was the consequence of a soluble factor released into the culture media through Cx hemichannels. PC12 cells that expressed Cx32-EGFP or Cx43-EGFP showed significant increases in neurite outgrowth whether they actually expressed the Cx protein or not. Similar results were obtained using multiple approaches including co-culturing and conditioned media.
The factor most likely responsible for the enhanced neurite outgrowth is ATP. Previous studies have shown that functional hemichannels exist (51)(52)(53) and that hemichannels formed by Cx43 in astrocytes and C6 glioma cells can release ATP (42). ATP, in turn, can have significant functional consequences on Ca 2ϩ signaling. For example, Stout and colleagues (42) demonstrated that the Cx hemichannel activator quinine can stimulate intercellular Ca 2ϩ waves in Cx43-expressing C6 cells but not native communication-incompetent C6 cells. The addition of apyrase reduced the degree of Ca 2ϩ propagation suggesting that the observed Ca 2ϩ waves were stimulated by the release of ATP through hemichannels (42). Previous studies have also shown that purines, such as ATP, can induce neurite outgrowth in PC12 cells and synergistically enhance neurite elongation with NGF (39,41).
To explore ATP as a potential mechanism responsible for the enhanced neurite outgrowth observed in the current study several different experimental approaches were used. For example, EGFP-infected PC12 cells were treated with 10 M ATP or 2-MeSATP either in the presence or absence of NGF. The ATPtreated cells showed autodifferentiation prior to NGF addition comparable with that seen in the Cx32 and Cx43 expressing cells. The addition of NGF to cells treated with ATP or 2-Me-SATP also resulted in increased neurite length comparable with what was observed following the expression of Cx32 or Cx43 in PC12 cells treated with NGF.
ATP binds to purinergic receptors and ATP and purinergic receptors have been shown to help regulate PC12 cell neurite outgrowth (39,41). Two classes of purinergic receptors have been identified (P2X and P2Y) and both are known to be expressed by PC12 cells (54). In the current study, blockade of P2 receptors using suramin or PPADS (P2 antagonists) abolished Cx-enhanced neurite outgrowth following NGF as well as Cx-mediated transient autodifferentiation. ATP levels were found to be substantially higher in the culture media of Cx-infected PC12 cells treated with NGF versus empty vector-infected cells and increased levels of ATP in the media correlated with enhanced Cx hemichannel activity. Native PC12 cells grown in conditioned media obtained from Cx-infected cells displayed enhanced neurite outgrowth that was abolished by apyrase pretreatment to degrade the ATP present in the media. Collectively, the data obtained in the current study supports the hypothesis that enhanced neuritogenesis in the Cx expressing cells is mediated by ATP release through Cx hemichannels.
There are a number of potential signaling cascades activated by purinoreceptors that may be responsible for the enhanced neurite outgrowth observed in the current study. For example, binding of ATP to the P2Y receptor (G protein-coupled receptor) activates phospholipase C, which hydrolyzes phosphatidylinositol 4,5-bisphosphate to 1,2-diacylglycerol and inositol triphosphate (55). Inositol triphosphate binds to receptors on the endoplasmic reticulum facilitating the release of Ca 2ϩ , whereas 1,2-diacylglycerol activates protein kinase C (55). Both protein kinase C and Ca 2ϩ are also activated by NGF and the TrkA receptor and are responsible for initiating PC12 cell differentiation and neurite outgrowth (56). Therefore, NGF and ATP likely act synergistically on common signaling pathways to cause the enhanced neurite outgrowth observed in the current study.
Connexins and Development-Connexins are present during all aspects of neural development and maturation although they may be differentially expressed throughout the process. Both Cx32 and Cx43 are present during embryonic development and have been localized to developing neocortical neurons (57), dopaminergic neurons of the midbrain (18), and spinal motor neurons (58). A study by Bannerman et al. (59) showed that blocking gap junction communication in neural crest cells resulted in decreased cell survival and migration, suggesting that gap junctions play an important functional role in central nervous system development. Studies using neuronal precursor cell lines also support the notion that gap junctions are important in differentiation. For example, Bani-Yaghoub et al. (29) examined the effects of gap junction blockers on the differentiation of human NT2/D1 cells. Cells treated with gap junction blockers along with retinoic acid (induces differentiation) resulted in a dramatic decrease in the number of mature neurons (7% that of control cultures) compared with control cultures treated with retinoic acid alone. The results suggest that important morphogenic or differentiation factors are shared between coupled cells or released through Cx hemichannels. Finally, Weissman et al. (24) have demonstrated that radial glial cells in the developing neocortex release ATP via Cx hemichannels that stimulate calcium release from inositol triphosphate-sensitive internal stores through a phospholipase C-dependent pathway. Disruption of the inositol triphosphate pathway decreases neocortical progenitor proliferation in the cortex (24). In conclusion, the results of the current study suggest that Cx proteins may contribute to neuronal differentiation in the developing central nervous system through the release of molecules such as ATP in the local microenvironment.