Connexin43 and astrocytic gap junctions in the rat spinal cord after acute compression injury

Connexin43 and Astrocytic GapJunctions in the Rat Spinal CordAfter Acute Compression Injury

E. THERIAULT,1 U.N. FRANKENSTEIN,2 E.L. HERTZBERG,3 AND J.I. NAGY2*1Playfair Neuroscience Unit, The Toronto Hospital, Toronto Western Division,

Toronto, Ontario, Canada M5T 2S82Department of Physiology, University of Manitoba, Winnipeg, Manitoba, Canada R3E 0W3

3Departments of Neuroscience and of Anatomy and Structural Biology,Albert Einstein College of Medicine, Bronx, New York 10461

ABSTRACTTo examine the possible role of interastrocytic gap junctions in the maintenance of tissue

homeostasis after spinal cord damage, we initiated studies of the astrocytic gap junctionalprotein connexin43 (Cx43) in relation to temporal and spatial parameters of neuronal loss,reactive gliosis, and white matter survival in a rat model of traumatic spinal cord injury (SCI).Cx43 immunolocalization in normal and compression-injured spinal cord was compared byusing two different sequence-specific anti-Cx43 antibodies that have previously exhibiteddifferent immunorecognition properties at lesion sites in brain.At 1- and 3-day survival times,gray matter areas with mild to moderate neuronal depletion exhibited a loss of immunolabel-ing with one of the two antibodies. At the lesion epicenter, these areas consisted of a zone thatseparated normal staining distal to the lesion from intensified labeling seen with bothantibodies immediately adjacent to the lesion. Loss of immunoreactivity with only one of thetwo antibodies suggested masking of the corresponding Cx43 epitope. By 7 days post-SCI,Cx43 labeling was absent with both antibodies in all regions extending up to 1 mm from thelesion site. Reactive astrocytes displaying glial fibrillary acidic protein (GFAP) appeared by 1day and were prominent by 3 days post-SCI. Their distribution in white and gray mattercorresponded closely to that of Cx43 staining at 1 day, but less so at 3 days whenGFAP-positive profiles were present at sites where Cx43 labeling was absent. By 7 dayspost-SCI, Cx43 again co-localized with GFAP-positive cells in the surviving subpial rim, andwith astrocytic processes on radially oriented vascular profiles investing the central borders ofthe lesion. The results indicate that alterations in Cx43 cellular localization and Cx43molecular modifications reflected by epitope masking, which were previously correlated withgap junction remodeling following excitotoxin-induced lesions in brain, are not responseslimited to exogenously applied excitotoxins; they also occur in damaged spinal cord and areevoked by endogenous mechanisms after traumatic SCI. The GFAP/Cx43 co-localizationresults suggest that during their transformation to a reactive state, spinal cord astrocytesundergo a transitional phase marked by altered Cx43 localization or expression. J. Comp.Neurol. 382:199–214, 1997. r 1997 Wiley-Liss, Inc.

Indexing terms: glial cells; gap junction proteins; neuron survival; subpial white matter;

immunohistochemistry

A likely critical functional feature of astrocytes is theconfiguration of their processes in lamellar and end-feetarrays, forming well-described structural barriers aroundneuronal elements and at brain and vascular surfaces(Mugnaini, 1986). These barriers are presumably designedto separate regions of dissimilar or fluctuating ionic ormetabolic composition (Walz, 1989), although they possess

Contract grant sponsors: Medical Research Council of Canada (MRC),the Spinal Cord Research Foundation (USA), and the National Institutes ofHealth.*Correspondence to: Dr. J.I. Nagy, Department of Physiology, Faculty of

Medicine, University of Manitoba, 770 Bannatyne Avenue, Winnipeg,Manitoba, Canada R3E 0W3. E-mail: [email protected] 29April 1996; Revised 20 December 1996; Accepted 17 January

1997

THE JOURNAL OF COMPARATIVE NEUROLOGY 382:199–214 (1997)

r 1997 WILEY-LISS, INC.

numerous gap junctions (Massa and Mugnaini, 1982;Kettenmann et al., 1983;Waxman andBlack, 1984; Fischerand Kettenmann, 1985; Mugnaini, 1986; Kettenmann andRansom, 1988) and thus can be selectively breached atrestricted locations. The junctionally coupled network ofastrocytes forms what has been referred to as ‘functionalsyncytia‘ (Mugnaini, 1986; Ransom and Carlini, 1986).One proposed role for astrocyte syncytia is ‘spatial buffer-ing‘ (Orkland et al., 1966; Marriero and Orkland, 1996),which is thought to involve uptake of excess extracellularpotassium ions (K1) and their redistribution via astrocyticgap junctions.Gap junctions mediate intercellular electrotonic cou-

pling between neurons and metabolic coupling betweendiverse cell types (Loewenstein, 1981; Dermeitzel et al.,1990; Beyer, 1990, 1993; Bennett et al., 1991). They arecomposed of hexameric arrangements of gap junctionproteins (connexins), which form intercellular channelsthat allow cell-to-cell passage of ions and small molecules(Beyer, 1993; Dermeitzel and Spray, 1993; White et al.,1995; Wolberg and Rohlmann, 1995; Bruzzone et al.,1996). In the central nervous system (CNS), it is wellestablished that astrocytes in vivo express connexin43(Cx43; Dermeitzel et al., 1989; Giaume et al., 1991;Belliveau and Naus, 1994). Punctate immunolabeling forCx43 has been found throughout the brain, and correlativelight and electron microscopic studies have shown that thevast majority of this labeling is localized at astrocytic gapjunctions (Yamamoto et al., 1990a,b; Nagy et al., 1992;Ochalski et al., 1996). Knowledge of how astrocytes altertheir syncytial organization in response to CNS injurymayprovide insight into astrocytic contributions to differentialtissue survival during unusual spatial buffering andmeta-bolic demands. Elucidation of this issue has partly beenachieved through studies of astrocytic Cx43 following bothischemic and excitatory amino acid lesions in the rat brain(Vukelic et al., 1991; Hossain et al., 1994a,c; Ochalski etal., 1995; Sawchuk et al., 1995). These studies indicatethat a reorganization of Cx43 occurs in which certainepitope-specific antibodies no longer bind to the proteinimmunocytochemically while retaining their ability tointeract with denatured protein onWestern blots. We havereferred to this process, for which the molecular basis hasnot been elucidated, as epitope masking. Alterations inCx43 have also been observed in the facial motor nucleusafter lesions of the facial motor nerve (Rohlmann et al.,1993) and in biopsy tissue in human epilepsy (Naus et al.,1991).In this report, we have focused on studies of Cx43 in a

well-characterized model of traumatic spinal cord injury(SCI). The characteristic primary and secondary injuryprocesses, the profile of acute axonal loss, the survival ofsupraspinal projections, and the extent of neurologicalrecovery have been well studied in this compression injurymodel (Tator and Fehlings, 1991; Koyanagi et al., 1993a,b,c;Theriault and Tator, 1994; Anthes et al., 1995, 1996).Following ‘closed‘ spinal cord injuries in both humans andexperimental models (i.e., where the dural or pial mem-brane remains intact), the typical histopathological fea-ture is that of central cavitation and necrosis with asurviving subpial rim (Blight, 1989; Bunge et al., 1993,1995; Theriault, 1995). Although mechanical and vasculartheories have been proposed to explain the pattern ofcentral tissue loss (Young, 1987, 1988; Blight, 1989), thepossible existence of intrinsic neuroprotective factors inthe subpial rim has generally not been considered. In view

of the putative functions of the astrocytic syncytiumdiscussed above and the possibility that such processes asK1 spatial buffering may serve as one of these intrinsicfactors, our aim here was to begin establishing relation-ships between trauma-induced events and the status ofastrocytic gap junctional communication at an SCI site ina clinically relevant experimental model. In particular, wesought to determine whether indicators of astrocytic gapjunctional remodeling, such as alterations in Cx43 cellulardistribution and Cx43 immunoreactivity (Cx43-ir) seen atexcitotoxic brain lesions, also occur following traumaticSCI and whether these indicators can be correlated withthe progression of tissue injury and subsequent survival ata trauma site.

MATERIALS AND METHODS

Animal surgery and tissue preparation

Atotal of 38 femaleWistar rats weighing 200–250 g wereused to examine Cx43 in the normal spinal cord (n 5 8),and at each survival time (n 5 6-8) of 1, 3, and 7 days afterSCI. Acute, traumatic SCI was performed as previouslydescribed (Theriault and Tator, 1994). Briefly, animalswere given anesthesia with 2% halothane and nitrousoxide:oxygen (2:1), a laminectomy was made to expose C6to T2, and a modified Kerr-Lougheed aneurysm clip with aclosing force of 35 g was used to produce a moderatelysevere compression injury at the C8-T1 level. The clip wasplaced extradurally, released rapidly to compress the cordfor 60 seconds, and then removed. The surgical site wasclosed, 5 ml of dextrose saline was injected subcutaneouslyfor volume repletion, and animals were allowed to recoveron heated pads (37°C). Animals with SCI received manualbladder expression three times daily and were individuallyhoused with free access to food and water. All protocolswere reviewed and approved by the Toronto HospitalAnimal Care Committee and found to be in accordancewith the Canadian Council of Animal Care guidelines.After various survival times, animals were anesthetizedwith equithesin and transcardially perfused with ice-cold0.1 M sodium phosphate buffer pH 7.4 (PB) containing0.9% saline (PBS), followed by 400 ml of 0.1 M PBcontaining 4% paraformaldehyde. The spinal cords wereremoved, postfixed for 2 hours in the same fixative, and a3-mm block, comprising 1.5 mm both rostral and caudal tothe lesion epicenter, was cryoprotected in 50 mM PBcontaining 15% sucrose. The remainder of the rostral cordwas cut into three blocks (1.5–3.5, 4.5–7.5, and 7.5–10.5mm from the lesion epicenter) and cryoprotected in 50 mMPB containing 25% sucrose and 10% glycerol.

Antibodies and immunohistochemistry

Two polyclonal anti-Cx43 antibodies were generatedagainst bovine serum albumin (BSA)-conjugated peptidescorresponding to amino acids 346–363 or 241–260 of theCx43 sequence (Beyer et al., 1987) and designated 18A-8and 16A-8, respectively. These antibodies have been exten-sively used by us, and their specificity has been wellestablished (Yamamoto et al., 1990a,b; Hossain et al.,1994b). Both antibodies produce similar punctate immuno-labeling in sections of normal brain where Cx43-ir waslocalized to astrocyte gap junctions and intracellularly atsites near these junctions (Yamamoto et al., 1990b). Preab-sorption of the antibodies with the peptides against whichthey were raised was shown to eliminate all staining(Yamamoto et al., 1990a,b; Hossain et al, 1994a,b). In the

200 E. THERIAULT ET AL.

present experiments, no immunoreactivity was observedin spinal cord sections processed after omission of theprimary antisera. A mouse monoclonal antibody againstglial fibrillary acidic protein (GFAP; Boehringer Mann-heim, Laval, Quebec, Canada) was used to identify astro-cytes.Sections were cut in transverse or parasagittal planes at

a thickness of 20 µm. Cryostat sections of the lesionepicenter were collected on slides. Frozen sections of allremaining tissues were obtained on a sliding microtomeand collected free-floating in PBS. Sections were washed inPBS containing 0.3% Triton X-100 (PBST) and stainedwith creysl violet, or processed by either the peroxidaseantiperoxidase (PAP) method or by double immunofluores-cence.All antibodies were diluted in PBST, primary incuba-tions were conducted at 4°C for 48–72 hours, and washesand incubations with secondary immunoreagents wereperformed at room temperature. Primary antisera wereused at the following dilutions: anti-Cx43 antibody 18A-8,1:2,000; anti-Cx43 antibody 16A-8, 1:1,000; anti-GFAP,1:25. To eliminate endogenous peroxidase activity at thelesion epicenter, sections were incubated for 30 minutes inPB containing 0.6% hydrogen peroxide prior to antibodyapplication. After incubation with anti-Cx43 antibody,sections processed by the PAP method were incubated for1.5 hours with goat anti-rabbit antibody (SternbergerMonoclonals) diluted 1:100, washed for 1 hour, and thenincubated with rabbit PAP (Sternberger Monoclonals)diluted 1:500. After a final wash in PBST and two washesin 50 mM Tris-HCl buffer, pH 7.4, sections were reacted inTris-HCl buffer containing 0.02% 3,3-diaminobenzidine(DAB) and 0.005% hydrogen peroxide and then mountedonto slides from gelatin and alcohol, dehydrated in alcohol,and coverslipped with Lipshawmounting medium.For double immunofluorescence, sections were simulta-

neously incubated with either anti-Cx43 (18A-8) and anti-GFAP or with anti-Cx43 (16A-8) and anti-GFAP, washed,and then incubated for 1.5 hours with indocarbocyanine(Cy3)-conjugated sheep anti-rabbit antibody (Sigma, Oak-ville, Ont., Canada) diluted 1:250 and fluorescein isothio-cyanate (FITC)-conjugated horse anti-mouse antibody (Vec-tor, Mississauga, Ont., Canada) diluted 1:50. After washesin PBST, sections were coverslipped with antifade medium(Valnes and Brandtzaeg, 1985) and viewed with a LeitzDialux 20 microscope equipped with Ploempak filter cubesL3 for FITC and N2 for Cy3.For electron microscopy (EM), transverse or sagittal

sections through the lesion site were cut at a thickness of30–40 µm on a vibratome, collected in PB, and processedwith anti-Cx43 antibody 18A-8 with slight modifications ofthe above PAP procedure as previously described (Yama-moto et al., 1990a,b). Sections were then flat-embedded inTAAB-812 (Marivac), and areas of labeling or those devoid

of immunostaining were identified by light microscopy(LM), trimmed from sections, and glued onto resin blocks.Sections were stained with lead citrate for 1–2 minutes.

RESULTS

General observations

The lesion site comprising approximately 1.5 mm bothrostral and caudal to the lesion epicenter contained theregion of maximal tissue damage with hemorrhage andcentral necrosis evident by 1 day and further developmentof necrotic changes and cavitation by 1 week (Theriaultand Tator, 1994; Theriault, 1995). Because the borders ofthe acute lesion are difficult to define, the histopathologi-cal profile of the resulting compression lesion is presentedin a three-dimensional reconstruction of an 8-week post-SCI specimen showing the relative distributions of sparedtissue and neuronal loss or cavitation (Fig. 1A–C). Thevolume of the stabilized lesion can be evaluated in severalways, one of which is by calculating ratio of lost overspared tissue, resulting in a ‘cavitation index.‘ Typically,the region of the spinal cord rostral to the lesion epicenterdisplays the greatest volume of tissue loss (i.e., has thegreatest cavitation index).Parasagittal sections were examined to evaluate the

relative temporal and spatial progression of the lesion(Table 1). At 1 day post-SCI, neuronal loss extended anaverage of 710 µm rostrally and 440 µm caudally. At 3days, neuronal loss was increased, now extending anaverage of 830 µm rostrally and 500 µm caudally from thelesion epicenter. At 7 days, the extent of neuronal loss nowextended as far as 1,280 µm rostrally and 920 µm caudally.These progressive necrotic changes in the acute stage arereflected in the characteristically greater cavitation indexof spinal cord tissue rostral to the lesion epicenter inchronic survival animals. As assessed in Nissl-stainedtransverse sections, neuronal loss at 1 day post-SCI wasalready extensive at the lesion epicenter in laminae III–X(Fig. 2A) and became progressively less pronounced atgreater rostral and caudal distances with laminae V–VIIbeing most affected (Fig. 2B). At distances of up to 1.5 mmfrom the lesion, a ‘patchy‘ distribution of neuronal lossoccurred in intermediate laminae, whereas lamina Xremained severely affected (Fig. 2C). At 3 days survivaltime, presumptive macrophages and reactive microgliapopulated the enlarged lesion epicenter and the centrallylocated cavity. By 7 days survival, these cells filled theepicenter (Fig. 2D) and occupied more restricted region ofgray and white matter at farther distances from theepicenter (Fig. 2E,F).

TABLE 1. Characteristics of Neuronal Loss and Cx43 Immunolabeling at Various Distances from the Lesion Epicenter1

Survivaltime

Neuronal loss Intensified Cx43 staining2Width of zone withloss of Cx43 staining

Rostral Caudal Rostral Caudal Rostral Caudal

1 day 710(630–750)

440(330–500)

1,000(630–1,380)

520(380–630)

290(160–440)

200(50–350)

3 day 830(750–880)

500(380–640)

1,060(750–1,380)

710(380–1,000)

210(100–350)

290(130–450)

7 day 1,280(1,250–1,350)

920(875–1,000)

None None 1,110(880–1,250)

690(630–750)

1Values are means in micrometers with ranges in parentheses.2Distances from the lesion epicenter in which tissue exhibited intensified labeling for Cx43.

ASTROCYTE GAP JUNCTIONS AFTER SCI 201

Cx43 distribution after SCI

The overall pattern of Cx43 staining in gray matterfollowing SCI was best revealed in sagittal sections (Fig. 3,Table 1). At 1 day post-SCI, Cx43 immunoreactivity (Cx43-ir) with antibody 18A-8 displayed a distinct ‘banded‘pattern along the rostrocaudal axis (Fig. 3A). The lesionepicenter was totally devoid of Cx43 staining, whereasimmediately adjacent regions both rostrally and caudallyexhibited intensified staining (Fig. 3A,B). The intensifiedregion was bordered by a zone showing patchy loss ofCx43-ir, which, in turn, was bordered by normal staining.Similar results were obtained at 3 days post-SCI.At 7 dayspost-SCI, all staining in a widemargin of tissue adjacent tothe lesion had disappeared, and this was flanked rostrally

and caudally by normal Cx43-ir (Fig. 3C). No Cx43 label-ing was seen in sections processed after omission of theprimary antibody (Fig. 3D). Similar alterations in Cx43-irwere seen in transverse sections (Fig. 3E–G). Comparedwith relatively homogeneous staining in the graymatter ofnormal tissue (Fig. 3E), the epicenter of the injury at 1 and3 days post-SCI was completely devoid of Cx43-ir (notshown). Adjacent to the injury site, intensified Cx43-ir indeeper regions of gray matter were flanked by patchesdevoid of staining, as seen in sagittal sections, whereasareas of normal Cx43-ir persisted in the tips of the dorsaland ventral horns (Fig. 3F). Farther caudally and rostrally,patches of Cx43-ir loss were seen within areas of normalstaining (Fig. 3G). At 7 days post-SCI, all intensifiedCx43-ir had disappeared. Cx43-ir in gray matter at farther

Fig. 1. Three-dimensional reconstruction (3-D) of the spinal cordlesion site at the cervical level taken from an 8-week postspinal cordinjury (SCI) specimen. The epicenter of the lesion at C8-T1 is indicatedby the arrow in (A). The area of neuronal loss (i.e., cavitation) along the

rostrocaudal axis is shown inAand is illustrated separately beside the 3-Dspinal cord reconstruction (B). The distribution of surviving tissue in thesubpial rim is comprisedmainly of axons andglia and is representedby thedotted area in (C). r, rostral; c, caudal. Scale bars5 1mm.


distances from the lesion site (1.5–4.5 mm) was relativelyunaffected.To compare areas of neuronal loss with Cx43 labeling,

adjacent sections were taken for Nissl staining (Fig. 2B,C)and Cx43 (18A-8) immunofluorescence (Fig. 4A,B). At 1day post-SCI, intensified Cx43-ir of the kind shown in Fig.3F near the lesion was consistently associated with neuro-

nal loss. Zones exhibiting an absence of Cx43-ir fartherfrom the injury site (Fig. 4A,B) also displayed neuronaldepletion (Fig. 2B,C), whereas normal Cx43-ir persisted inareas where neurons were preserved.Although no attemptwas made to quantitate the degree of neuronal loss thatwas associated with loss of Cx43-ir, it appeared that thisrelationship was region dependent such that Cx43 per-

Fig. 2. Nissl-stained transverse sections of spinal cord showingextent of injury at rostrocaudal distances from the lesion epicenter.A–C: 1 day post-SCI. Neuron loss (asterisks) occurs in nearly all graymatter areas adjacent to the compression (A), is most prominent inintermediate laminae at distances of 0.5–1.0 mm from the lesion (B)

and is limited to patches at distances of 1–1.5 mm (C). D–F: 7 dayspost-SCI. Areas of neuron loss at levels corresponding to those shownin A and B, respectively, are occupied by inflammatory cells (arrows inD,E). At farther distances, these cells also invade the ventral aspect ofthe dorsal column (arrows in F). Scale bar 5 250 µm.


Fig. 3. Parasagittal (A–D) and transverse (E–G) sections of spinalcord showing connexin43 immunoreactivity (Cx43-ir) around thelesion site. A,B: Low and higher magnifications at 1 day post-SCI.Cx43-ir in gray matter is intense in areas adjacent to the lesion(arrowheads). Labeling is reduced or absent along a narrow dorsoven-tral zone (asterisks) and normal beyond these zones (stars). C: 7 dayspost-SCI, Cx43-ir is absent in a wide zone (asterisks) extending from

the crush. D: Section comparable with that shown in A. No labeling isseen after omission of primary antibody. E–G: Transverse sectionsthrough a region of normal staining (E corresponds to an areaindicated by stars in A), a region of increased staining intensity (Fcorresponds to an area indicated by arrowheads in A) and a region atthe transition between reduced (asterisks) and normal staining (G).Scale bar 5 400 µm.


sisted, despite substantial neuronal dropout in some ar-eas, but not in others.

Cx43-ir with antibodies 18A-8 and 16A-8

In normal spinal cord, antibody 16A-8 produced stainingpatterns similar to those obtained with antibody 18A-8,except that staining intensity was characteristically lighteras occurs elsewhere in the CNS (Hossain et al., 1994a).Adjacent to the compression injury (e.g., 0.5–1.0 mm),antibody 16A-8 produced the same intensified staining as18A-8 (not shown). A notable difference was evident,however, at farther distances from the lesion epicenter(e.g., 1.0–1.5 mm) where Cx43-ir was absent with 18A-8but persisted with antibody 16A-8, which gave the sameincreased staining intensity as seen with both antibodiescloser to the compression site. Thus, at 1 and 3 dayspost-SCI, patches of 18A-8 loss in both dorsal and ventralhorns (Fig. 5A,C) corresponded exactly to areas of in-creased 16A-8 staining (Fig. 5B,D), indicating that Cx43was still present in these areas, although undetectablewith antibody 18A-8.

Comparisons of Cx43-ir and GFAP

Patterns of GFAP-ir in the spinal cord at both 1 and 3days post-SCI resembled those of Cx43-ir with antibody18A-8. Loss of GFAP-ir at the lesion epicenter was bor-dered both rostrally and caudally by zones of increasedstaining, which were then flanked by regions exhibitingnormal patterns of GFAP. By 7 days post-SCI, thereremained a large central area devoid of GFAP-ir (notshown). In sections stained for both GFAP and one of thetwo anti-Cx43 antibodies, patches of GFAP loss (Fig. 5E,F)corresponded very closely to areas devoid of staining withantibody 18A-8 (Fig. 5A,C) and to areas displaying intensi-fied staining with 16A-8 (Fig. 5B,D). In unaffected tissuebeyond the borders of the lesion, Cx43-ir (Fig. 6A) andGFAP-positive astrocytes (Fig. 6B) appeared normal. Atthese borders, slightly reduced but otherwise normal

Cx43-ir (Fig. 6C) was accompanied by the presence ofGFAP-ir reactive astrocytes (Fig. 6D).Alterations in Cx43-ir and GFAP-ir near the lesion

epicenter also occurred in white matter, and these exhib-ited a dependence on survival time as shown in Figures7–9. In transverse sections of the dorsal columns fromnormal animals, fine Cx43-ir was distributed among cross-cut myelinated fibers (Fig. 7A), and some accumulations ofpuncta were co-localized with GFAP-positive processes(Fig. 7B). At 1 day post-SCI, the ventral half of the dorsalcolumns was devoid of both GFAP-ir and Cx43-ir, thedorsal half contained GFAP-positive reactive astrocytes,and a good correlation existed between the presence andabsence of staining for these two proteins (Fig. 7C,D). At 3days post-SCI, this correlation was no longer evident, sothat ventral areas that were devoid of Cx43 (Fig. 7E) nowcontained GFAP-positive reactive astrocytes (Fig. 7F).Farther removed from the lesion epicenter, white matter-reactive astrocytes were only evident in the subpial rim ofdorsal columns where Cx43 staining persisted.In ventral (Fig. 8) and lateral (not shown) white matter

of the normal cord, Cx43-ir was seen along radially ori-ented bundles of myelinated fibers (Fig. 8A) that wereladen with GFAP-ir (Fig. 8B). At 1 day post-SCI, there wasa correspondence between Cx43 and GFAP distribution inroughly the outer half of these white matter areas and aloss of both in the inner half (Fig. 8C,D). At 3 dayspost-SCI, some peripheral margins of the spinal cordexhibited intense GFAP-positive cell bodies and processes(Fig. 8F). Regions adjacent to the injury site often con-tained radially oriented GFAP-positive fibers originatingfrom immunoreactive cell bodies located in the subpial rimof white matter; some of these cells were also positive forCx43 (Fig. 8E).By 7 days post-SCI, there was again a very good

correlation between the distribution of Cx43 and GFAPalong peripheral margins of white matter at distances ofover 300 mm from the injury site (Fig. 9). These peripheralareas were starkly demarcated from central regions devoidof Cx43 and GFAP (Fig. 9A,B). In contrast, the entirespinal cord at the lesion epicenter was devoid of both Cx43and GFAP. Blood vessels penetrating radially into the cordfrom the subpial rim at 7 days post-SCI were invariablyladen with Cx43-ir and GFAP-ir fibers (Fig. 9C–E). Bloodvessels within the lesion epicenter were never associatedwith GFAP-positive elements but were occasionally sur-rounded by Cx43-ir puncta (not shown).At farther distances from the lesion epicenter (1.5–4.5

mm), the only effects noted in white matter, evident at 3and 7 days post-SCI, was a reduction in Cx43-ir associatedwith radially oriented myelinated fiber bundles in lateraland ventral funiculi. The processes of radially orientedglial cells were also less intensely stained (not shown). Inaddition, the ventral portion of the dorsal columns occasion-ally displayed an absence of both Cx43 and GFAP staining.

Ultrastructural observations

Sites of Cx43-ir in normal gray matter were numerousand intensely labeled (Fig. 10A), occurring in the cyto-plasm of astrocytic processes, at gap junctions betweenthese processes (not shown), and occasionally along theirnonjunctional membranes. Finer caliber labeled processeswere usually seen partly ensheathing synaptic complexes(Fig. 10B). Cx43-ir was intermittent, rarely seen along theentire length of these circumscribing processes, and many

Fig. 4. Distribution of Cx43-ir as related to extent of neuronal lossat 1 day post-SCI. A,B: Cx43 immunofluorescence in magnified areasof sections adjacent to the Nissl-stained sections illustrated in Figure2B and C, respectively. At 0.5–1.0 mm from the lesion site (A) and at1.0–1.5 mm (B), areas with an absence of Cx43-ir (asterisks) corre-spond to those regions exhibiting neuronal loss as indicated byasterisks in Figure 2B and C. Scale bar 5 200 µm.


were devoid of staining (Fig. 10B). At 1 day post-SCI,intense and extensive Cx43-ir was seen on nonjunctionalmembranes of astrocytic processes that surrounded axons,dendrites, and cell bodies in the neuropil (Fig. 10C,D). Insome areas, these processes were swollen and electronlucent (Fig. 10C), whereas in others they were much finerand often appeared as two closely apposed and densely

stainedmembranes with little or no intervening cytoplasm(Fig. 10D,E).The labeling of astrocyte processes in injured spinal cord

was less intermittent than seen in normal cord, althoughunlabeled processeswere often observed (Fig. 10E). Stainedastrocytic gap junctions were observed (Fig. 10F), anddiffuse labeling was often present in astrocyte cell bodies

Fig. 5. Immunofluorescence labeling of Cx43 and glial fibrillaryacidic protein (GFAP) 1.0–1.5 mm from a lesion site at 1 day post-SCI.A–D:Adjacent sections of the same field in the dorsal horn (A,B) and inthe ventral horn (C,D) showing Cx43-ir with antibody 18A-8 (A,C) and16A-8 (B,D). Areas with normal labeling with both Cx43 antibodies(stars) surround regions that are devoid of Cx43-ir with 18A-8

(asterisks) but exhibit intensified staining with 16A-8 (arrows). E,F:Labeling for GFAP (E) in the same field as in the dorsal horn (A,B) andin the same field (F) as in the ventral horn (C,D). Areas exhibiting lossof GFAP-ir in gray and white matter (asterisks) correspond to thosewith altered Cx43-ir. Scale bar 5 40 µm inA,B,E, 100 µm in C,D,F.


(Fig. 10D) but was rare in normal spinal cord. Similarresults were obtained at 3 days post-SCI (not shown),although ultrastructural disruption was more severe nearthe lesion.At 7 days post-SCI, areas devoid of Cx43-ir wereprominent adjacent to the lesion, and these were flankedby regions of moderate staining (not shown). In Cx43-depleted zones, microglia and macrophages had invadedthe neuropil (not shown). Astrocytic somata were not seen,and the few astrocytic processes that remained displayedno intermediate filaments, were unlabeled for Cx43 (Fig.11A), and were devoid of gap junctions. In contrast, inregions distal from the lesion where staining resumed,astrocytic somata and processes were frequently hypertro-phied, electron lucent, and laden with intermediate fila-ments, and Cx43-ir was seen in these elements (Fig.11B-D), frequently in association with polyribosomes (Fig.11C,D) and at astrocytic gap junctions.

DISCUSSION

The present results demonstrate that alterations inCx43-ir occur in white and gray matter areas of the spinalcord at the site of a mechanically induced traumatic injury.A consistent observation was a close spatial correspon-dence between areas of neuronal loss and areas exhibitingalterations in Cx43 immunolabeling. We have shown thatthe response of astrocytic Cx43, as revealed by epitope

masking and antibody detection, differs in damaged re-gions of tissue adjacent to a compression site, comparedwith damaged regions adjacent to normal tissue. Further-more, during their transformation to a reactive state,astrocytes appear to undergo transitional states character-ized by the degree to which Cx43-ir correlates with GFAP-ir. Both normal and reactive astrocytes in the subpial rimof white matter ultimately express Cx43. We speculatethat interastrocytic gap junctions may contribute to repairand homeostatic processes in this spinal cord region andthus contribute to its survival after SCI.We show a progressive temporal increase in post-SCI

gray matter injury and corroborate findings in an openmodel of SCI (Dusart and Schwab, 1993) that showrelatively greater neuronal loss (i.e., larger cavitationindex) in spinal cord to the lesion epicenter. Other closedinjury models result in various lesion morphology as seenin three-dimensional reconstructions, depending on sever-ity and location of the injury (Bresnahan et al., 1991). Wehave shown that areas of Cx43 loss correspond to areas ofneuronal loss, both adjacent to and farther away from thelesion epicenter. A consistent finding in excitotoxin injurymodels has been a loss of Cx43 immunorecognition withsome anti-Cx43 antibodies in areas of tissue exhibitingincreased detection with others (Hossain et al., 1994a,c;Ochalski et al., 1995; Sawchuk et al., 1995). We havepreviously shown that these antibody recognition patterns

Fig. 6. Comparison of Cx43-ir and GFAP-ir in intermediate lami-nae of normal spinal cord gray matter with an area bordering normaland injured tissue at 3 days post-SCI. A,B: In normal tissue, denseCx43-ir (A) is seen throughout areas containing GFAP-positive astro-

cytes (B). C,D: Cx43-ir is reduced but, otherwise, has a normalappearance (C) in areas occupied by reactive astrocytes exhibitingintense GFAP-ir (D). Scale bar 5 40 µm inA,B, 50 µm in D,C.


occur at intact astrocytic gap junctions and in conjunctionwith remodeling of junctions and cellular redistribution ofCx43 (Hossain et al., 1994a,c; Ochalski et al., 1995). Inspinal cord immediately adjacent to the compression site,intensified Cx43-ir was evident with both antibodies andcorresponded ultrastructurally to extensive labeling ofnongap junctional astrocytic membranes with commensu-

rately fewer gap junctions present thanmight be expected,given the widespread Cx43-ir observed at both the LM andEM levels. Thus, where seen with both antibodies, intensi-fied staining reflects injury-induced gap junction disassem-bly and Cx43 dispersal in plasma membrane leadingpresumably to an uncoupling of astrocytes immediatelyadjacent to the site that sustained the greatest injury.

Fig. 7. Pairs of micrographs (A,B; C,D; E,F) show the same fieldsimmunolabeled for Cx43 (A,C,E) and GFAP (B,D,F) in the dorsalcolumns. A,B: In normal dorsal column, punctate Cx43-ir is uniformlydistributed (A) and occasionally associated (arrows) withGFAP-positiveastrocytes (B). C,D:At 1 day post-SCI, labeling for Cx43 (C, asterisks)and GFAP (D, asterisks) in ventral areas of the dorsal column is

largely absent up to 1.5 mm from the lesion epicenter but is stillpresent in the subpial rim where there is a correspondence of labeling(arrows). E,F: At 3 days post-SCI, Cx43-ir is still seen in the subpialwhite matter and is absent in ventral regions of the dorsal columns (E,asterisks). These Cx43-negative areas now contain reactive astrocytes(F, arrows). Scale bar 5 65 µm inA,B,E,F, 100 µm in C,D.


A different astrocytic response to injury is suggested byobservations of zones farther from the injury site whereintensified staining due to epitope unmasking is seen withantibody 16A-8, concomitant with complete masking of the18A-8 epitope. Because of difficulties in obtaining ad-

equate vibratomed sections of lesioned tissue, areas withmasking of 18A-8 but intensified 16A-8 were not examinedby EM. In any case, severity of tissue injury might beexpected to be less far from the injury site (i.e., in theischemic penumbra) where masking of the 18A-8 epitope

Fig. 8. Pairs of micrographs (A,B; C,D; E,F) show the same fieldsimmunostained for Cx43 (A,C,E) and GFAP (B,D,F) in ventral andlateral white matter. A,B: Normal ventral white matter showingassociation of Cx43-ir (A, arrows) with GFAP-positive (B, arrows) fiberbundles exiting the ventral horn (vh). C,D: Ventral white matter at 1day post-SCI at a distance of 0.5 mm from the crush site. Cx43-ir (C,asterisks) and GFAP-ir (D, asterisks) are absent in the central half of

white matter but correspond in the subpial rim (arrows). E,F: Lateralwhite matter 0.5 mm from the crush site at 3 days post-SCI showingreduced Cx43-ir (E) in a field of intensely GFAP-positive reactiveastrocytes (F). Cx43-ir (E, arrows) is seen in some astrocytes (F,arrows) at the central margin of the subpial rim. Scale bar5 100 µm inA-D, 50 µm in E,F.


occurred. This masking may underlie a physiologicallyrelevant mechanism in astrocytes to regulate gap junc-tional communication before either their disposal or thereestablishment of coupling, depending on other contingen-cies surrounding the tissue injury. Thus, masking of Cx43epitopes in a zone adjacent to healthy tissue could reflectreduced junctional coupling and an attempt by astrocytesto sever communication with their neighbors, therebyrestricting the outward flow of potentially damaging me-tabolites. This possibility remains to be tested by func-tional analysis of interastrocytic junctional communica-tion at the spinal cord lesion site, e.g., by analyzing dyetransfer in slice preparations from the injured zone.The functional correlates of andmolecular events under-

lying Cx43 epitope masking and unmasking reported hereand observable under other pathophysiological manipula-

tions remain to be elucidated. The generation of alterna-tively spliced message for Cx43 such that different epi-topes are expressed (or not) can be excluded because thecoding sequence for Cx43 and for other connexins thus faranalyzed occurs in a single exon (Lang et al., 1991;Willecke et al., 1991; Sullivan, 1993; Goodenough et al.,1996). Although epitope masking may result from Cx43interactions with other proteins, there is as yet no evi-dence for such an interaction.Altered binding of 18A-8 dueto posttranslational modification, including phosphoryla-tion (Saez, 1993), remains a possibility. This appearsunlikely, however, because the 18A-8 antibody recognizesCx43 on Western blots even when masked immunocyto-chemically in tissue sections. In contrast, the 16A-8 epi-tope in Cx43 has been identified to contain a serinephosphorylation site (Bruzzone et al., 1996). Thus, it is

Fig. 9. Pairs of micrographs (A,B; C,D) show the same fieldsimmunostained for Cx43 (A,C) and GFAP (B,D) in lateral white matter0.7 mm from the crush site at 7 days post-SCI. A–E: All gray matterand central regions of white matter are devoid of Cx43 and GFAPlabeling (asterisks). In the surviving subpial rim of white matter,

punctate Cx43-ir (A) is uniform in areas containing GFAP-ir (B) and isalso associated with radially oriented blood vessels (C, arrows; magni-fied in E) invested with GFAP-positive fibers (D, arrows). Scale bar 5100 µm inA-D, 40 µm in E.


Fig. 10. Electron micrograph of Cx43 localization in spinal cord.A,B: Normal dorsal horn. Cx43-ir comprises cytoplasmic and nongapjunctional membrane labeling of astrocytic processes (A, arrows).Labeled processes circumscribe neuronal elements and display inter-mittent cytoplasmic staining (B, arrows) with unlabeled segments (B,arrowheads) containing intermediate filaments (i). C–F: 1 day post-SCI. Extensive Cx43 labeling of astrocytic nongap junctional cytoplas-mic membranes (C, arrows) is seen in swollen processes in neuropil

and surrounding a cell body (So). Astrocytes (As) contain cytoplasmicCx43-ir (D, asterisks) and thin, labeled astrocytic processes (D, arrow-heads) envelop neural elements. Intense labeling is evident in thinastrocytic processes, including regions where their nongap junctionalmembranes are separated by little or no cytoplasm (E, arrowheads)and at gap junctions between some processes (F, arrows). Scale bar5 3µm in A, 0.5 µm in B, 2.3 µm in C, 0.8 µm in D, 200 nm in E, 165 nmin F.

possible that intensified staining with 16A-8 arises fromCx43 dephosphorylation at this site and results in greaterimmunohistochemical detection of this dephosphorylatedepitope.Because opening or closure of the gap junction channels

is influenced by a variety of factors, including calcium, pH,

and phosphorylation of junctional proteins (Spray andBennett, 1985; Bennett et al., 1991; Beyer, 1993; Wolburgand Rohlmann, 1995), elevations in extracellular calciumand acidosis (Young and Koreh, 1986) following SCI wouldlikely serve to close off sites of severely damaged spinalcord, as proposed for other tissues (Yee and Ravel, 1978;

Fig. 11. Electron micrograph of Cx43-ir in spinal cord at 7 dayspost-SCI. A: Area corresponding to the zone devoid of Cx43-ir withantibody 18A-8 as seen by light microscopy near the lesion site. NoCx43-ir was seen in damaged tissue or in swollen astrocytic processes(asterisks) surrounding a blood vessel (Bv). B: Area of spinal cordadjacent to this zone showing labeled astrocytic processes surrounding

a blood vessel (arrows), cytoplasmic (asterisks) labeling in a swollenprocess containing numerous intermediate filaments (i), and nonjunc-tional membrane labeling (arrowheads). C,D: Cytoplasmic labeling islargely associated with polyribosomes (C, arrows) shown at highermagnification in D. Scale bar5 4 µm inA, 2.4 µm in B, 0.9 µm in C, 300nm in D.


Dermeitzel et al., 1987). If evoked to excess, this processmight compromise astrocytic spatial buffering and contrib-ute secondarily to, rather than limit, neuronal damage.Indeed, Largo et al. (1996) have recently shown thatdepression of astrocyte function in the rat brain in situleads to K1 accumulation and blocked synaptic transmis-sion within minutes and massive neuronal cell death inhours. Their results strongly suggest that the extent of theischemic penumbra might depend on the failure of glialneuroprotection. Notably, at all time points in our studyand in all tissue sections through the lesion epicenter,GFAP-ir persisted in the surviving regions of white matterimmediately below the pial surface encompassing theintact subpial rim. In zones flanking the SCI lesionepicenter (e.g., in the ischemic penumbra), there may be apotential for rescue, depending on the patency of thevascular supply and the degree of intactness of the glialimitans. In previous studies we have shown that the pialvasculature is remarkably preserved following SCI inclinical and experimental material (Koyanagi et al.,1993b,c; 1995) both at the epicenter of the lesion and atdistances farther removed. Postmortem analyses of clini-cal material reveal that over two-thirds of human SCI areclosed injuries, i.e., with the dura or pia remaining intact(Bunge et al., 1994, 1995). Interestingly, in these cases,there is a distinct preservation of the subpial rim of tissueas well. Whether the presence of astrocytes in the subpialrim of the spinal cord and the effectiveness of theirbuffering functions contribute to local tissue survivalfollowing traumatic closed SCI requires further investiga-tion.Lesion-induced loss of GFAP-ir has been described with

some anti-GFAP antibodies, but this protein remainsdetectable with other anti-GFAP antibodies (Graeber andKreutzberg, 1986; Schmidt-Kostner et al., 1993; Eddlestonand Muske, 1993) within astrocytes remaining at lesionsites. Thus, the absence of GFAP-ir in some regions ofspinal cord at 1 and 3 days post-SCI cannot be taken asconclusive evidence for absence of astrocytes at these sites.At 7 days post-SCI, however, gray matter areas of spinalcord appear devoid of astrocytes as seen by EM. Alter-ations in GFAP labeling, transformation of astrocytes to areactive form, and masking of the Cx43 18A-8 epitope allappear to occur in a consistent temporal sequence atborders between damaged and normal tissue. Initially,there is a loss of both GFAP and Cx43 labeling. This isfollowed by the reappearance of GFAP in reactive astro-cytes that continue to show little Cx43 detectability with18A-8 and some detection with 16A-8. Finally, Cx43 detect-ability with 18A-8 returns as well. Thus, in the presentinjury model, reactive astrocytes possess gap junctions, asreflected by their decoration with Cx43-ir puncta, but theirtransformation to a reactive state coincides with maskingof the 18A-8 epitope. We speculate that this masking maybe related to gap junction remodeling, which might benecessary during reorganization of the astrocyte syncy-tium in a region undergoing reactive gliosis. However, anydirect relationships between patterns of Cx43 immunode-tection and cytoskeletal functions associated with GFAP inintermediate filaments remain to be demonstrated.

ACKNOWLEDGMENTS

The authors thank P.A.Y. Ochalski, B. Boguski, and S.Mortin-Toth for excellent technical assistance. This work

was supported by grants from the Medical ResearchCouncil of Canada (MRC) to J.I.N., the Spinal CordResearch Foundation (USA) to E.T., and the NationalInstitutes of Health to E.L.H. Studentship support toU.N.F. was provided by the Rick Hansen Man in MotionLegacy Fund.

LITERATURE CITED

Anthes, D.L., E. Theriault, and C.H. Tator (1995) Characterization ofaxonal ultrastructural pathology following experimental spinal cordcompression injury. Brain Res. 702:1–16.

Anthes, D.L., E. Theriault, and C.H. Tator (1996) Ultrastructural evidencefor arteriolar vasospasm following spinal cord trauma. Neurosurgery39:804–814.

Belliveau, D.J., and C.C. Naus (1994) Cortical type 2 astrocytes are not dyecoupled nor do they express the major gap junction genes found in thecentral nervous system. GLIA 12:24–34.

Bennett, M.V.L., L.C. Barrio, T.A. Bargiello, D.C. Spray, E.L. Hertzberg,and J.C. Saez (1991) Gap junctions: new tools, new answers, newquestions. Neuron 6:305–320.

Beyer, E.C. (1990) Connexin family of gap junction proteins. J. Membr. Biol.116:187–194.

Beyer, E.C. (1993) Gap junctions. Int. Rev. Cytol. 137:1–37.Beyer, E.C., D.L. Paul, and D.A. Goodenough (1987) Connexin43: a protein

from rat heart homologous to a gap junction protein in liver. J. Cell Biol.105:2621–2629.

Blight, A.R. (1989) Mechanical factors in experimental spinal cord injury. J.Am. Paraplegia Soc. 11:26–34.

Bresnahan, J.C., M.S. Beattie, B.T. Stokes, and K.M. Conway (1991)Three-dimensional computer-assisted analysis of graded contusionlesions in the spinal cord of the rat. J. Neurotrauma 8:91–101.

Bruzzone, R., T.W. White, and D.L. Paul (1996) Connections with connex-ins: the molecular basis of direct intercellular signaling. Eur. J.Biochem. 238:1–27.

Bunge, R.P., J.D. Puckett, W.R. Becerra, A. Marcillo, and R.M. Quencer(1993) Observations on the pathology of human spinal cord injury. InF.J. Seil (ed): Advances in Neurology, Vol. 59. New York: Raven Press,pp. 75–89.

Bunge, R.P., R.M. Quencer, W.R. Becerra, J.D. Puckett, and J.D. Guest(1995) Clinical/pathological correlates in damage to the human spinalcord. J. Neurotrauma 12:354–360.

Dermietzel, R.O., and D.C. Spray (1993) Gap junctions in the brain: where,what type how many and why? Trends Neurosci. 16:186–192.

Dermietzel, R.O., S.B. Yancey, O. Traub, K. Willecke, and J.P. Revel (1987)Major loss of the 28-kD protein of gap junction in proliferatinghepatocytes. J. Cell Biol. 105:1925–1934.

Dermietzel, R.O., T.K. Traub, R.T.K. Hwang, E. Beyer, M.V.L Bennett, D.C.Spray, and K. Willecke (1989) Differential expression of three gapjunction proteins in developing and mature brain tissues. Proc. Natl.Acad. Sci. USA 86:10148–10152.

Dermietzel, R.O., R.T.K. Hwang, and D.S. Spray (1990) The gap junctionfamily: structure, function, and chemistry. Anat. Embryol. 182:517–528.

Dusart, I., andM.E. Schwab (1993) Secondary cell death and the inflamma-tory reaction after dorsal hemisection of the rat spinal cord. Eur. J.Neurosci. 6:712–724.

Eddleston,M., and L.Mucke (1993)Molecular profile of reactive atrocytes—implications for their role in neurological disease. Neuroscience 54:15–36.

Fischer, G., and H. Kettenmann (1985) Cultured astrocytes form a syncy-tium after maturation. Exp. Cell Res. 159:273–279.

Giaume, C., C. Fromaget, A.E. Aoumari, J. Cordier, J. Glowinski, and D.Gros (1991) Gap junctions in cultured astrocytes: single channelcurrents and characterization of channel-forming protein. Neuron6:133–143.

Goodenough, D.A., J.A. Goliger, and D.L. Paul (1996) Connexins, connexonsand intercellular communication. Annu. Rev. Biochem. 65:475–502.

Graeber, M.B., and G.W. Kreutzberg (1986) Astrocytes increase in glialfibrillary acidic protein during retrograde changes of facial motorneurons. J. Neurocytol. 15:363–373.

Hossain, M.Z., M.A. Sawchuk, L.J. Murphy, E.L. Hertzberg, and J.I. Nagy(1994a) Kainic acid induced alterations in antibody recognition of


Connexin43 and loss of astrocytic gap junctions in rat brain. GLIA10:250–265.

Hossain, M.Z., L.J. Murphy, and J.I. Nagy (1994b) Phosphorylated forms ofconnexin43 predominate in rat brain: demonstration by rapid inactiva-tion of brain metabolism. J. Neurochem. 62:2394–2403.

Hossain, M.Z., J. Peeling, G.R. Sutherland, E.L. Hertzberg, and J.I. Nagy(1994c) Ischemia-induced cellular redistribution of the astrocytic gapjunctional protein connexin43 in rat brain. Brain Res. 652:311–322.

Kettenmann, H., R.K. Orkand, and M. Schachner (1983) Coupling amongidentified cells in mammalian nervous system. J. Neurosci. 3:506–516.

Kettenmann, H., and B.R. Ransom (1988) Electrical coupling betweenastrocytes and between oligodendrocytes studied in mammalian cellcultures. Glia 1:64–73.

Koyanagi, I. (1995) The vascular mechanisms in the pathophysiology ofacute spinal cord injury. J. Neurotrauma 12:357–367.

Koyanagi, I., C.H. Tator, and P.J. Lea (1993a) Three dimensional analysis ofthe vascular system in the rat spinal cord with scanning electronmicroscopy of vascular corrosion casts. 1. Normal spinal cord. Neurosur-gery 33:277–284.

Koyanagi, I., C.H. Tator, and P.J. Lea (1993b) Three dimensional analysis ofthe vascular system in the rat spinal cord with scanning electronmicroscopy of vascular corrosion casts. 2. Acute spinal cord injury.Neurosurgery 33:285–292.

Koyanagi, I., C.H. Tator, and E. Theriault (1993c) Silicone rubber microan-giography of acute spinal cord injury of the rat. Neurosurgery 32:260–268.

Lang, L.M., E.C. Beyer, A.L., Schwartz, and J.D. Gitlin (1991) Molecularcloning of a rat uterine gap junction protein and analysis of geneexpression during gestation. Am. J. Physiol. 260:787–793.

Largo, C., P. Cuevas, G.G. Somjen, R. Martin del Rio, and O. Herreras(1996) The effect of depressing glial function in rat brain in situ on ionhomeostasis, synaptic transmission, and neuron survival. J. Neurosci.16:1219–1229.

Loewenstein, W.E. (1981) Junctional intercellular communication: thecell-to-cell membrane channel. Physiol. Rev. 61:829–913.

Marrero, H., and R.K. Orkland (1996) Nerve impulses increase glialintercellular permeability. GLIA 16:285–289.

Massa, P.T., and E. Mugnaini (1982) Cell junctions and intramembraneparticles of astrocytes and oligodendrocytes: a freeze-fracture study.Neuroscience 7:523–538.

Mugnaini, E. (1986) Cell junctions of astrocytes, ependyma, and relatedcells in mammalian central nervous system, with emphasis on thehypothesis of a generalized functional syncytium of supporting cells. InS. Fedoroff and A. Verndakis (eds): Astrocytes, Vol. 1. New York:Academic Press, pp. 329–371.

Nagy, J.I., T. Yamamoto, M.A. Sawchuk, D.M. Nance, and E.L. Hertzberg(1992) Quantitative immunohistochemical and biochemical correlatesof connexin43 localization in rat brain. GLIA 5:1–9.

Naus, C.C.G., J.F. Bechberger, and D.L. Paul (1991) Gap junction geneexpression in human seizure disorder. Exp. Neurol. 111:198–203.

Ochalski, P.A.Y., U.N. Frankenstein, E.L. Hertzberg, and J.I. Nagy (1996)Connexin43 in rat spinal cord: localization in astrocytes and identifica-tion of heterotypic astro-oligodendrocytic gap junctions. Neuroscience(in press).

Ochalski, P.A.Y., M.A. Sawchuk, E.L. Hertzberg, and J.I. Nagy (1995)Astrocytic gap junction removal, connexin43 redistribution and epitopemasking at excitatory amino acid lesion sites in rat brain. GLIA14:279–294.

Orkland, R.K., J.G. Nicholls, and S.W. Kuffler (1966) Effect of nerveimpulses on the membrane potential of glial cells in the central nervoussystem of amphibia. J. Neurophysiol. 29:788–806.

Ransom, B.R., and W.G. Carlini (1986) Electrophysiological properties ofastrocytes. In S. Federoff and A. Vernadakis (eds): Astrocytes, Vol. 2.NewYork: Academic Press, pp. 1–49.

Rohlmann, A., R. Laskawi, A. Hofer, E. Dobo, R.O. Dermeitzel, and J.R.Wolff (1993) Facial nerve lesions lead to increased immunostaining of

the astrocytic gap junction protein (connexin43) in the correspondingfacial nucleus of rats. Neurosci. Lett. 154:206–208.

Saez, J.C., V.M. Berthoud, A.P. Moreno, and D.C. Spray (1993) Gapjunctions. Multiplicity of controls in differentiated and undifferentiatedcells and possible functional implications. In S. Shenolikar and A.C.Nairn (eds): Advances in Second Messenger Phosphoprotein Research.NewYork: Raven Press, pp. 163–198.

Sawchuk, M.A., M.Z. Hossain, E.L. Hertzberg, and J.I. Nagy (1995) In situtransblot and immunocytochemical comparisons of astrocytic con-nexin-43 responses to NMDA and kainic acid in rat brain. Brain Res.683:153–157.

Schmidt-Kastner, R., K. Weitasch, H. Weigel, and U.T. Eysel (1993)Immunohistochemical staining for glial fibrillary acidic protein (GFAP)after deafferentation or ischemic infarction in the rat visual system:features of reactive and damaged astrocytes. Int. J. Dev. Neurosci.11:157–174.

Spray, D.C., and M.V.L. Bennett (1985) Physiology and pharmacology ofgap junctions. Annu. Rev. Physiol. 47:281–303.

Sullivan, R., C. Ruangvoravat, D. Joo, J. Morgan, B.L. Wang, X.K. Wang,and C.W. Lo (1993) Structure, sequence and expression of the mouseCx43 gene encoding connexin43. Gene 130:191–199.

Tator, C.H., and M.G. Fehlings (1991) Review of the secondary injurytheory of acute spinal cord trauma with emphasis on vascular mecha-nisms. J. Neurosurg. 75:15–26.

Theriault, E. (1995) Neuroprotective role of astrocytes following traumaticspinal cord injury. J. Neurotrauma 12:364–372.

Theriault, E., and C.H. Tator (1994) Persistence of rubrospinal projectionsfollowing spinal cord injury in the rat. J. Comp. Neurol. 342:249–258.

Valnes, K., and P. Brandtzaeg (1985) Retardation of immunofluorescencefading during microscopy. J. Histochem. Cytochem. 33:755–761.

Vukelic, J.I., T. Yamamoto, E.L. Hertzberg, and J.I. Nagy (1991) Depletionof connexin43-immunoreactivity in astrocytes after kainic acid-inducedlesions in rat brain. Neurosci. Lett. 130:120–124.

Walz, W. (1989) Role of glial cells in the regulation of the brain ionmicroenvironment. Prog. Neurobiol. 33:309–333.

Waxman, S.G., and J.A. Black (1984) Freeze-fracture ultrastructure of theperinodal astrocyte and associated glial junctions. Brain Res. 308:77–87.

White, T.W., R. Bruzzone, and D.L. Paul (1995) The connexin family ofintercellular channel forming proteins. Kidney Int. 48:1148–1157.

Willecke, K., H. Hennemann, E. Dahl, S. Jungbluth, and R. Heynkes (1991)The diversity of connexin genes encoding gap junctional proteins. Eur.J. Cell Biol. 56:1–7.

Wolburg, H., and A. Rohlmann (1995) Structure-function relationships ingap junctions. Int. Rev. Cytol. 157:315–373.

Yamamoto, T., A. Ochalski, E.L. Hertzberg, and J.I. Nagy (1990a) On theorganization of astrocytic gap junctions in rat brain as suggested by LMand EM immunohistochemistry of connexin43 expression. J. Comp.Neurol. 302:853–883.

Yamamoto, T., A. Ochalski, E.L. Hertzberg, and J.I. Nagy (1990b) LM andEM immunolocalization of the gap junctional protein Connexin43 in ratbrain. Brain Res. 508:313–319.

Yee, A., and J.P. Revel (1978) Loss and reappearance of gap junctions inregenerating liver. J. Cell Biol. 78:554–564.

Young, W. (1987) The post-injury responses in trauma and ischemia:secondary injury or protective mechanisms. Cent. Nerv. Syst. Trauma4:27–52.

Young, W. (1988) Spinal cord injury. Curr. Opin. Neurol. Neurosurg.1:611–622.

Young, W., and I. Koreh (1986) Potassium and calcium changes in injuredspinal cords. Brain Res. 365:42–53.

Zhang, J.T., and B.J. Nicholson (1989) Sequence and tissue distribution of asecond protein of hepatic gap junctions, Cx26, as deduced from itscDNA. J. Cell Biol. 109:3391–3401.


Documents

Connexin43 and astrocytic gap junctions in the rat spinal cord after acute compression injury