Behavioral recovery from spinal cord injury following delayed application of polyethylene glycol

It is both instructive and convenient to consider the catastrophic loss of behavioral function following spinal cord injury (SCI) as two different syndromes: the initial, or acute, phase and the subsequent chronic condition. Theoretical biological approaches to treating each are not mutually exclusive, but must take into account the distinct pathologies characteristic of each phase of the injury.

During the acute phase, a delayed, progressive and self-propagating wave of cell and tissue destruction is initiated, with the injury leading to cell death, the formation of a cicatrix and the irreversible loss of distal segments of axons through Wallerian degeneration (Griffin et al., 1995). This latter dynamic is often referred to as ‘secondary injury’ (Honmou and Young, 1995), and it is the loss of white matter, which cannot be replaced, that frames the more permanent behavioral loss accompanying the long-term, or chronic, phase of the injury.

Techniques aimed at regenerating new and functional connections by promoting nerve regeneration from intact proximal segments show clinical promise in cases of SCI. Inhibition of endogenous inhibitors of central nervous system (CNS) regeneration, application of novel growth factors to the lesioned spinal cord, neurotransplantation of both fetal nervous tissue and activated peripheral nervous system macrophages into the spinal lesion and the application of direct current electrical fields are all thought to facilitate regrowth of spinal cord white matter and/or to facilitate new functional synaptic connections (Schwab et al., 1993; Bregman et al., 1996; Benowitz et al., 1999; Lazarov-Spiegler et al., 1996; Borgens et al., 1999). The last three techniques mentioned have now moved into human clinical study in cases of severe SCI. Another approach, which is aimed at recovering functional deficits irrespective of the time since the original spinal injury, is to restore physiological conduction through intact but non-functional white matter by K⁺ channel blockade. This technique is also at the stage of human clinical testing (Shi and Blight, 1997; Hansebout et al., 1993; Hayes et al., 1993).

It is a matter of debate whether greater success in reducing long-term injury has been achieved in recent years by attacking the initial phase of the insult. One such example, adopted as a means of standard management of acute spinal cord injury, has been the administration of large doses of the steroid methylprednisilone within hours of the injury (Bracken et al., 1990). This approach, often referred to as ‘neuroprotection’, claims to ameliorate behavioral loss by reducing the extent of secondary injury, although the efficacy and safety of this therapy are now under question (Short et al., 2000; Pointillart et al., 1999).

We suggest another approach to treating the acute injury. In this method, the application of a hydrophilic surfactant physically repairs damaged membranes, leading to a rapid (minutes to hours) recovery of cellular structure and function. This treatment is designed to permit immediate recovery of nerve impulse conduction in injured fibers, to reverse the permeabilization of the plasmalemma and immediately to seal breaches in it that would probably progress to dissolution and axotomy.

We have reported that a brief administration of the fusogen polyethylene glycol (PEG) to completely transected, but reapposed, adult guinea pig spinal axons can induce anatomical reconnection of the severed proximal and distal segments and immediate recoveries of compound action potential conduction through fused axons (Shi et al., 1999). We have also shown that a similar application of PEG to severe compression injuries also leads to an immediate recovery of compound action potential conduction through the lesion (Shi and Borgens, 1999). These procedures were repeated in vivo in experiments in which an aqueous solution of PEG was applied for 2 min to a standardized compression injury to adult guinea pig thoracic spinal cord. A swift recovery of both spinal cord conduction, measured by somatosensory evoked potentials (SSEPs), and behavioral function, measured by the recovery of the cutaneus trunci muscle (CTM) reflex (Borgens et al., 1987; Blight et al., 1990), occurred in PEG-treated animals (Borgens and Shi, 2000).

Here, for the first time, we evaluate fully the behavioral character of the recovered CTM reflex produced by a delayed application of PEG and confirm our observations of the physiological recovery of conduction in 100 % of these spinally injured animals.

Twenty-nine adult (300 g) guinea pigs were used in this experiment. They were divided into two groups, PEG-treated (N=15) and sham-treated (N=14). One animal died following surgery in the control group. Guinea pigs were anesthetized with an intramuscular injection of 100 mg kg^–1 ketamine HCl and 20 mg kg^–1 xylazine prior to surgical exposure of the cord by dorsal laminectomy and removal of the dura (Borgens et al., 1986, 1990). A standardized injury was produced between thoracic vertebrae 10 and 11 with a constant-displacement 15 s compression of the cord using specially constructed forceps possessing a détente (Blight, 1991) (see also Moriarty et al., 1998). This lesioning procedure had been calibrated to produce a total loss of compound action potential conduction through the spinal cord injury and behavioral functioning of the CTM reflex (Borgens and Shi, 2000). To sedate animals for behavioral and physiological testing, guinea pigs were injected with 0.1 ml of sodium pentobarbital (50 mg ml^–1). All surgical procedures and testing were carried out under protocols approved by the Purdue University Animal Care and Use Committee in accordance with Federal, State and University guidelines governing the use of animals in research.

An aqueous solution of PEG (M_r 1800, 50 % w/v in distilled water) was applied with a Pasteur pipette directly onto the exposed injury (dura removed) for 2 min and then removed by aspiration. This procedure was performed approximately 7 h after the lesioning procedure (see above). The region was immediately washed with isotonic Krebs’ solution (NaCl, 124 mmol l^–1; KCl, 2 mmol l^–1; KH₂PO₄, 1.24 mmol l^–1; MgSO₄, 1.3 mmol l^–1; CaCl₂, 1.2 mmol l^–1; dextrose, 10 mmol l^–1; NaHCO₃, 26 mmol l^–1; sodium ascorbate, 10 mmol l^–1), which was also aspirated to remove excess PEG. In sham-treated animals, the injury site was re-exposed surgically at approximately 7 h post-lesion, a control application of water (‘vehicle’) was applied for 2 min, followed by a lavage with Krebs’ solution subsequently removed by aspiration. The wounds were closed, and the animals were kept warm with heat lamps until awaking. Guinea pigs were housed individually and fed ad libitum.

The CTM reflex (Fig. 1) is observed as a rippling of the skin on the back of the animal following light tactile stimulation. These contractions can be measured by tattooing a matrix of dots on the animal’s shaved back. When the skin contracts towards the point of tactile stimulation, the dots move in this direction. Practically, a receptive field is determined prior to injury by stimulating the back skin of the sedated animal by lightly touching it with a monofilament probe. When skin outside the receptive field is stimulated, there is no response. By exploring the entire back with the probe, the boundary of the receptive field can be established. This boundary is drawn directly onto the shaved back of the guinea pig using a marker (Fig. 2). Stimulation inside the boundary produces CTM contractions, but stimulation outside it does not. The entire procedure was videotaped by a camera mounted approximately 0.92 m above the examination table. The same procedure was used to determine the region of CTM loss (areflexia) and subsequent recovery, when it occurred. These behavioral tests (and the physiological test described below) were performed approximately 24 h, 3 days, 2 weeks and 1 month post-treatment.

To quantify the CTM behavior, we evaluated four individual components of it by stop-frame analysis of the videotaped recordings. (i) The area of recovery of receptive fields below the level of the lesion was measured by planimetry from videotape images captured to the computer. These data were normalized by dividing the area of CTM loss by the total area of the receptive field prior to injury and expressing the result as a percentage. Similarly, the area of CTM recovery was determined and then divided by the original area of CTM areflexia to give the percentage recovery of the CTM reflex. (ii) The direction of peak skin movement in normally functioning and recovered receptive fields following injury was determined as a vector. (iii) The distance of skin movement during peak skin contraction was measured. (iv) The velocity of skin contraction following tactile stimulation was computed using these data.

The overall pattern of skin movement can be quite complex in response to focal stimulation. Thus, we chose to restrict our quantitative evaluation of back skin movement to the peak contractions in response to stimulation, i.e. to the region of skin where markers were displaced by the greatest distance. When the region of peak skin contraction had been determined, the videotape was reversed to a time just prior to stimulation and skin movement. The videotape was then advanced at intervals of 1/24th of a second so that a time point prior to, and just at the peak of skin contraction, could be captured to the computer. These frames were superimposed over images of the animals, and the distance of peak contraction was divided by the time required to produce it. This provided a measure of the velocity of skin contraction (Fig. 2). The character of skin movement following tactile stimulation was determined at the pre-injury evaluation for all but four animals, and for all animals at 1 day, 3 days, 2 weeks and 1 month post-treatment. When the peak contraction was determined for any one animal, a protractor was used to measure the angle at which the skin was pulled towards the monofilament probe relative to an imaginary line perpendicular to the long axis of the animal at the midline. The peak contraction of the skin was recorded as a positive angle when the skin pulled towards the probe and as a negative angle in the infrequent cases in which the skin pulled away from the probe. We also recorded whether this peak response occurred on the same side of the midline as the point of stimulation or on the other side (a contralateral response).

Subdermal electrodes stimulated nerve impulses from the tibial nerve of the hindleg (stimuli trains in sets of 200 at 3 Hz; stimulus amplitude ⩽3 mA square wave, 200 μs in duration) (Fig. 3). Evoked potentials, more properly termed somatosensory evoked potentials (SSEPs), were conducted through the spinal cord to the sensory cortex of the brain. To record SSEPs, a pair of subdermal electrodes, located above the level of the contralateral cortex, was used with a reference electrode usually located in the ipsilateral pinna of the ear. The stimulation and computer management of evoked potential recordings utilized a Nihon Kohden Neuropak 4 stimulator/recorder and PowerMac G3 computer. In all animals, failure to record an SSEP was confirmed to be due to the absence of evoked potential conduction through the lesion by a control test carried out at the same time. The medial nerve of the forelimb was stimulated, initiating evoked potentials in a neural circuit above the level of the crush injury. To perform this test, recording electrodes were left in place while stimulating electrodes were relocated to stimulate the median nerve of the foreleg.

Video images were acquired to an Intel Dual Pentium Pro computer. Superimposition of images, the coloring of receptive field boundaries made on the back of the animal during CTM testing and the general management of video images were performed using Adobe Photoshop software. The final figures were constructed with Microsoft PowerPoint software and printed on an Epson Stylus Color 800 printer. Quantitative planimetry of the area of receptive fields or of the regions of behavioral loss and recovery from these video images was carried out using IP Lab Spectrum software.

The Mann–Whitney two-tailed test was used to compare the means of the data derived from experimental and sham-treated groups. To compare the proportions between groups, Fisher’s exact test was used. All tests were performed using Instat software.

Only one animal died (after surgery) during the course of this study. The loss of the receptive fields subsequent to lesioning of the spinal cord resulted in a region of areflexia that was similar in all animals (mean loss of the total receptive field in sham-treated animals, 59.2±5.0 %; in PEG-treated animals, 52.2±2.4 %, means ± s.e.m.; P=0.19; Mann–Whitney two-tailed test). Thus, the standardized injury technique produced a mean CTM loss that was similar in both groups.

As will be discussed below, since none of the control animals recovered an SSEP, 13 of the 15 experimental animals with the best electrical recordings were evaluated quantitatively. Similarly, a full evaluation of CTM functioning by stop-frame videographic analysis was carried out on the three controls that recovered the reflex and on 13 of the 15 recovered PEG-treated animals for comparison.

Application of PEG produced a very rapid recovery of CTM function in 73 % of treated animals within the first 24 h compared with a complete lack of spontaneous recovery in sham-treated animals at this time (Fig. 4). Some spontaneous recovery of the CTM reflex in controls began to appear on day 3, resulting in three recoveries out of a total of 13 animals (23 %) by 1 month (Table 1). In marked contrast, 11 of 15 PEG-treated animals recovered the reflex activity within the region of areflexia during the first day post-treatment (Table 1) (Fig. 4), and another three animals by 1 month (total number of animals showing 93 %; P⩽0.0003; Fisher’s exact test, two-tailed) (Fig. 4). The area of recovered areflexic back skin was 27.6±8.6 % in the 15 PEG-treated animals and 18.3±3.4 % (means ± s.e.m.) in the three controls. Thus, the total area of PEG-mediated recovery was not statistically different from that occurring spontaneously, although infrequently (P=0.28; Mann–Whitney, two-tailed test).

In general, back skin contracts towards the point of stimulus when the CTM reflex is activated. The largest response usually occurs ipsilateral and rostral to the point of stimulation because the reflex occurs bilaterally. Recall that the spinal lesion produces a bilateral injury eliminating the entire CTM receptive field below the level of injury on both sides (Fig. 4). In the normal reflex, there is a minor contralateral contraction in response to ipsilateral stimulation of the skin, so we provide details of the region of peak skin contraction, no matter which side of the midline on which it occurred, with reference to the region of stimulation producing it (Table 2). However, we observed only three occasions where the peak contraction occurred contralateral to the point of stimulation (normal or recovered CTM) and only in PEG-treated animals. In the three control recoveries, the peak contraction was always ipsilateral, as expected. Furthermore, we observed only six examples out of 52 separate comparisons (experimental and control; left and right sides) where the peak contraction was directed away from the focal stimulus in the uninjured animal, emphasizing that this is normally an infrequent occurrence. When the direction of skin contraction, its angle and velocity were compared between the pre-injury and post-injury data in PEG-treated animals, there were no statistical differences; thus, the recovered reflex was faithfully reproduced by PEG treatment (Table 2).

We also evaluated the change in direction of skin contraction on both the left and right sides of the animals by paired comparison of the angle of skin contraction in PEG-treated animals only (as will be emphasized below, control CTM behavior did not change following its spontaneous recovery in just three animals). The mean angle of skin contraction following ipsilateral CTM stimulation was not significantly different after PEG-mediated recovery from the normal CTM in the same animals prior to injury (P=0.43; Mann–Whitney, two-tailed paired comparison). This was also true for the contralateral responses (P=0.44, same test).

In the three spontaneous recoveries in control animals, the peak distance of contraction (1 mm) and its velocity (25 mm s^–1) were identical in two of them and the reflex was unchanged after recovery. In the third case, a reduction in the velocity of CTM contraction was measured, while the angle of contraction and the peak distance of contraction remained unchanged (Table 2).

As in previous studies, SSEPs usually segregated into two peaks following tibial stimulation in the uninjured animal: an early-arriving peak (approximately 20–35 ms) and a late-arriving peak (40–50 ms). Table 1 shows the proportion of animals recovering an SSEP in the experimental population, which was 87 % for the first day after PEG treatment and by week 4 had reached 100 %. Not one sham-treated animal recovered conduction over the same period (Table 1). Fig. 5A shows a typical example of an SSEP recorded from a sham-treated animal, as well as a median nerve control procedure. Such control procedures were undertaken for any measurement that failed to demonstrate a repeatable SSEP and, in every case, demonstrated that the absence of evoked potentials was due to a failure to conduct them through the lesion. In Fig. 5B, a typical process of PEG-mediated recovery is shown. Note that the latency of the recovering evoked potentials was greater than normal in the early stages of recovery, but gradually declined with time (depicted for the early-arriving peaks). Fig. 6 shows that normal latency was not reached during the 1 month of observation and also plots the magnitude of the early-arriving evoked potentials, which recover to more than 50 % of their pre-injury values.

This report both confirms and extends an earlier one that PEG treatment can, within hours of its application, reverse the behavioral and conduction losses that follow severe spinal cord injury. Here, we have focused exclusively on a 7 h delayed administration of PEG to a standardized spinal cord injury.

The means of injury we chose was a constant-displacement injury in which each spinal cord received a severe lateral compression for approximately the same distance and for the same duration. This is a different approach from previous studies of impact injuries (usually dropping a weight, usually 20 g from 20 cm, onto the dorsal aspect of the exposed spinal cord). These different approaches have been discussed and described elsewhere (Blight, 1988, 1991; Collins and Kauer, 1979). We looked for a method that would produce a repeatable injury among animals, at least with respect to the anatomical consequences of the technique. In our hands, the constant-displacement injury method produces lesions that are not statistically different among animals, as determined by three-dimensional computer reconstruction of the vertebral segment containing the injury (Moriarty et al., 1998; Duerstock et al., 2000). The resulting hemorrhagic lesion is fairly typical of clinical injuries in that it spreads in severity from the center of the spinal cord (destroying most gray matter) to the margins, leaving a subpial rim of spared white matter over a longitudinal distance of approximately one vertebral segment (Moriarty et al., 1998) (see also Duerstock and Borgens, 2002). Moreover, the technique does not require special laboratory equipment, such as stereotaxic frames and impactors, and is simple to carry out. Nevertheless, all attempts at standardizing an injury to the spinal cord are still associated with some variation in functional recovery among animals. This is believed to be due to biomechanical differences between the cords, to the variable level of endogenous sealing of mechanically damaged axons (Krause et al., 1994; Shi et al., 2000) and to other factors that cannot be controlled (Blight, 1991). This is evident in this report from the modest number of control animals showing CTM recovery. A more precise anatomical and behavioral outcome can be obtained by specific transection techniques; however, severing tracts, or the spinal cord itself, is generally not a clinically meaningful injury model (Borgens, 1992).

Other conventional experimental methods have also been vitiated, or found to be unnecessary, by the extremely rapid recovery produced by PEG in prior published experiments and unpublished pilot trials. Thus, blinding of evaluators to the treatment used was not performed in this study since, within a matter of a hours, the experimentally treated animals could be identified with greater than 90 % accuracy in CTM studies or with 100 % accuracy when evaluating SSEP recovery. The most important precaution in these series of experiments was the determination that the measures of damage are statistically similar between animals after injury and before PEG treatment.

The skin movement is dependent on sensory afferents projecting as a long tract of axons in each ventral funiculus of the spinal cord (just lateral to the spinothalamic tract) to nuclei of CTM motor neurons located at the cervical/thoracic junction left and right of the midline (Blight et al., 1990; Thierault and Diamond, 1998) (Fig. 1). There are some contralateral afferent projections within the cord, but these appear to play only a very marginal role in the normal expression of the reflex. This mainly local response to stimulation is further represented by the fact that there are no contralateral projections between the nuclei of CTM motor neurons (Blight et al., 1990). The reflex is bilaterally organized into segmentally arranged receptive fields, displays little supraspinal control and is lost following spinal injury, producing a region of areflexia below the level of the lesion (Borgens et al., 1990; Blight et al., 1990; Thierault and Diamond, 1988) (Fig. 1). Following transection, recovery of the CTM reflex within this region of areflexia is not usually observed for the rest of the life of the animal and is infrequent (⩽20 %) following severe crush lesions to guinea pig spinal cord (Blight et al., 1990; Borgens and Shi, 2000; Borgens and Bohnert, 2001).

Although these injuries to the spinal cord also produce locomotor deficits in the animals, we have ignored these deficits and their apparent changes over time. This is because standing, stepping and overall locomotion in small animal models of SCI are locally controlled and generated within the spinal cord and so have little dependence on supraspinal control and organization (for a review, see Borgens, 1992). We also note that the recovery of the CTM and the recovery of SSEP conduction should be considered as independent outcomes. The long tracts associated with the CTM are bilateral and located in the ventrolateral funiculus, while SSEP conduction (initiated by stimulation of the tibial nerve) is more a measure of dorsal column activity (see below).

We have not previously evaluated the character of the CTM behavioral recovery by comparing its characteristics with those of the pre-injury reflex. Here, we have shown that the recovered reflex activity is statistically similar in terms of the direction, distance and velocity of CTM contractions when compared with the normal reflex. The entire receptive field lost after injury was not restored, however. The largest area of back showing recovery after PEG treatment approached 50 % of the original area of areflexia. Given that we evaluated only one specific sensorimotor reflex, which is dependent on an identified white matter tract, this amount of appropriately organized recovery was surprising. The recovery of the receptive fields on the back was also variable in location and was sometimes revealed as islands of CTM-receptive flank that stood out within a still-areflexic region of back skin. This variability in the way in which recovery was re-established is believed to result from the variable re-recruitment of afferent axons into conduction through the lesion (see also Blight et al., 1990; Borgens et al., 1987, 1990). In summary, direct application of this hydrophilic polymer to the site of a spinal cord injury can rapidly reverse behavioral loss, restoring an appropriately organized behavioral response as well as nerve impulse conduction through the lesion within a clinically useful time frame.

The most striking effect of PEG application in all SCI studies in this series, in vitro and in vivo, has been the uniform recovery of conduction through the lesion in response both to transection (and reattachment) and to compression injury (Shi and Borgens, 1999; Shi et al., 1999; Borgens and Shi, 2000; Shi and Borgens, 2000; Borgens and Bohnert, 2001). This recovery of conduction, documented by the recovery of either compound action potential in vitro or of SSEP conduction in vivo, occurred in 100 % of the spinal cords treated with PEG. It is very clear that the fusogenic action of PEG (Nakajima and Ikada, 1994; Davidson et al., 1976) alters the dynamics of axolemma injury to permit a very rapid re-establishment of excitability in the region of conduction block. The magnitude of recovered compound action potentials in spinal cord strips in isolation can be nearly doubled by the addition of the fast K⁺ channel blocker 4-aminopyridine, suggesting that the ‘repaired’ region of membrane is still somewhat leaky to K⁺ and not yet a perfect seal (Shi and Borgens, 1999).

We have also shown that the PEG-restored membrane is repaired sufficiently to exclude the uptake of a large-molecular-mass intracellular tracer, horseradish peroxidase (HRP) (Shi and Borgens, 2000). This label is normally taken up into breaches of the nerve membrane, a common method of loading neurons or their processes with this intracellular ‘dye’ (Borgens et al., 1986; Malmgren and Olsson, 1977). Compression-injured spinal cords showed a dramatic uptake of HRP into white matter axons at the epicenter of the injury when evaluated only 15 min post-injury. This uptake was greatly reduced by PEG treatment, and the effect was independent of axon diameter. Thus, the dye-exclusion test (Asano et al., 1995) proved that the polymeric ‘seal’ is indeed just that and is sufficient to interfere with the ability of even large-molecular-mass compounds to cross the membrane after damage (Shi and Borgens, 2001). Physical reconnection of axons within completely severed strips of guinea pig spinal cord ventral white matter also demonstrated the capability of a brief PEG treatment to reunite membranes sufficiently to seal in small-molecular-mass fluorescent intracellular markers (rhodamine- and fluorescein-decorated dextrans; M_r 8000) (Shi et al., 1999).

Finally, the companion paper (Duerstock and Borgens, 2002) demonstrates that this anatomical sealing of single cells can also be represented at the level of the whole tissue. The histopathological character of the in vivo spinal lesion is also reduced by PEG treatment.

This report is one of a series that has explored the ability of a cell fusogen, PEG, to reconnect severed mammalian spinal cord axons and to seal the axolemma of severely compressed/crushed spinal axons. We have previously discussed what are believed to be the mechanisms of action of PEG, in particular, as well as the mechanisms that may be shared with non-ionic triblock polymers such as the poloxamines and poloxamers discussed below (for a review, see Borgens, 2001) (see also Borgens and Shi, 2000; Lee and Lentz, 1997; Lentz, 1994; Marks et al., 2001). Briefly, sealing of membrane breaches by high-molecular-mass molecules such as PEG may involve a dehydration of the plasmalemma where closely apposed regions of the bilayer resolve into each other, i.e. the structural components of the plasmalemma are no longer partitioned by the polar forces associated with the aqueous phase. The lipid core of the membrane, exposed across a breach in it, would be expected to flow together in the absence of an aqueous ‘barrier’ (following dehydration of the membrane). Subsequent to the removal of the polymer and rehydration, the now-continuous phase undergoes spontaneous reassembly of its structural elements, including protein components and complex lipid components. This reorganization of cellular water is believed to result from the strongly hydrophilic structure of PEG.

Poloxamers are complex polymers of a PEG–propylene glycol–PEG structure. They too can seal membrane breaches (see below), but perhaps through a slightly different mechanism. It is proposed the hydrophobic head group inserts itself into the membrane breach, sealing it by ‘plugging’ it (Marks et al., 2001). We have preliminary evidence that Poloxamer 188 can also induce axolemma sealing and induce recovery from adult guinea pig SCI in a manner similar to PEG (D. Bohnert and R. B. Borgens, unpublished observations). It is also possible that these non-ionic detergents, like PEG, may seal membrane breaches initially as a surfactant film, although this does not explain the permanent nature of polymer-mediated seals, even after the polymers have been removed.

In all studies, we have used an application of an aqueous solution of PEG (50 % w/v in distilled water) for 2 min. We have detected no difference in response using PEG solutions prepared with polymers with a relative molecular mass of 400 to approximately 3000 (D. Bohnert and R. B. Borgens, unpublished observations), but believe that the viscosity of the solution may be more important to PEG-mediated repair than the molecular mass of the polymer. Using fluorescently decorated PEG, we have traced the distribution of PEG after various routes of administration in guinea spinal cord injury. PEG clearly targets the hemorragic lesion and is only faintly detected in uninjured spinal cord following a direct application to the injury site (Borgens and Bohnert, 2001). More importantly, subcutaneous or intravenous application labels the lesion just as well. This may be more important to an eventual clinical use where intravascular injection of PEG may be beneficial during emergency care and later applied directly to the lesion during surgical management of the injury. At present, we are testing the application of hydrophilic polymers administered both topically and through the blood supply in veterinary cases of severe, acute paraplegia in dogs secondary to disc herniation and fracture dislocation (see also Blight et al., 1991; Borgens et al., 1999).

As mentioned above, non-ionic detergents, so-called triblock polymers, are mainly composed of PEG and may share mechanisms of action in reversing cell permeabilization. Their structure usually incorporates a high-molecular-mass central hydrophobic core with hydrophilic PEG side chains. Poloxamer 188 (P188) has been shown to reverse muscle cell death subsequent to high-voltage insult (Lee et al., 1992). Isolated rat skeletal muscle cells were labeled with an intracellular fluorescent dye that leaked out of the cells after high-voltage trauma. This insult was sufficient to disrupt muscle membranes, allowing the leakage of the marker in 100 % of the control preparations. Treatment of skeletal muscle cells in vitro with P188 reduced or even eliminated leakage of dye following the injury. Further in vivo tests extended these results: an intravenous injection of P188 produced physiological and anatomical recovery of rat muscle following electric shock (Lee et al., 1992). This approach has also been tested for its ability to reverse cell death in a testicular reperfusion injury model in rats (Palmer et al., 1998). P188 can also seal heat-shocked muscle cells in vitro, as shown by an inhibition of calcein dye leakage from cells induced by elevated temperature (Padanlam et al., 1994). P188 rescues fibroblasts from lethal heat shock (Merchant et al., 1998). Another biocompatible detergent (Poloxamer 1107; administered intravenously) was used in an in vivo testicular ischemia/reperfusion injury model in rats as well as for inhibiting the leakage of hemoglobin from irradiated erythrocytes (Palmer et al., 1998; Hannig et al., 1999). These studies demonstrate that non-ionic biocompatible detergents and large hydrophilic molecules can reverse the permeabilization of cell membranes and also that they can be administered through the vascular system to reach damaged target cells.

We wish to thank Kimberly Harrington for the artwork and Erin Case, Carie Brackenbury and Charlyce Patterson for preparation of the manuscript. Financial support from research funds came from the Institute for Applied Neurology, Center for Paralysis Research, Purdue University, a generous gift from Mrs Mari Hulman George, AASERT DAAHO4-93-G-101, the Department of Defense and NIH R01 NS39288-01 to R.B.B.