Introduction

It is well-established that poly(ADP)ribose-polymerase (PARP) assumes major importance during ischemic brain damage [1], and that PARP inhibitors attenuate tissue injury after transient cerebral ischemia [2–6]. Furthermore, the PARP inhibitor PJ34 reduced spinal cord injury in a murine model of thoracoabdominal aortic ischemia-reperfusion injury [7]. We have recently demonstrated in a clinically relevant porcine model of ischemia/reperfusion (I/R) injury induced by thoracic aortic cross-clamping that the novel, highly selective PARP-1 inhibitor INO1001 facilitated hemodynamic stabilization during the early reperfusion period without affecting DNA integrity or repair [8]. Therefore in the present study we tested the hypothesis that INO1001 would reduce spinal cord injury [8, 9]. The data presented were obtained in a subset of animals that had been studied in the experimental series investigating the hemodynamic and renal effects of INO1001 [8].

Material and methods

The experiments were performed in adherence to the National Institute of Health Guidelines on the Use of Laboratory Animals. The experimental protocol was approved by the University Animal Care Committee and the federal authorities for animal research (Tübingen, Germany). Ten domestic pigs of either sex with a mean body weight of 48 g (range 43–58) were used. The anesthesic procedure, surgical preparation, placement of catheters, physiological measurements [8, 9], and measurement of INO1001 [10] and tumor necrosis factor-α [11] plasma concentrations have been described in detail previously.

Postmortem spinal cord analysis

In addition to hematoxylin-eosin staining, postmortem spinal cord cross-sections were analyzed after nuclear cresyl violet staining (Nissl staining) [12]. Immunohistochemistry for p21 and p27 gene expression and the determination of the number of apoptotic nuclei were performed using specific antibodies and the terminal deoxynucleotidyl transferase mediated nick end labeling (TUNEL) assay as described previously [8, 9, 13, 14].

Experimental protocol

After 120 min of postsurgery recovery time animals were randomly assigned to receive either the vehicle (glucose 5%) or the INO1001 starting 90 min before clamping. Animals received a total of 4 mg kg–1 INO1001: after an initial bolus (2 mg kg–1 h–1 over 30 min) INO1001 (1 mg kg–1 h–1) was infused for 60 min until immediately before clamping, stopped during the clamping period, and restarted again (0.5 mg kg–1 h–1) after declamping during the remaining 4 h of the experiment. This approach was chosen so that plasma levels were higher immediately prior to clamping in order to provide a sufficient “loading up” of the tissues. Furthermore, total doses of 1 mg kg–1 and 4 mg kg–1 of INO1001 in dogs [15–17] and pigs [18] had virtually completely abolished poly(ADP-ribose)-staining in the heart, lung, and intestine after cardiopulmonary bypass. After baseline data collection the aorta was occluded for 45 min, which was verified by the disappearance of the blood pressure signal distal (MAPdist) to the clamping. This clamping period was chosen to avoid both the large spinal cord infarction over several segments reported in pigs after a clamping period of 60 min or longer [19] and the fairly mild histological damage observed after only 30 min of clamping [9, 12]. During the clamping period intravenous nitroglycerin (1.7 mg min–1), esmolol (16.5 mg min–1), and adenosine-5′-triphosphate (2–10 mg min–1) were adjusted to maintain mean blood pressure proximal of the aortic tourniquet (MAPprox) at 80–120% of the baseline value. In addition to 10 ml kg–1 h–1 Ringer's solution infused throughout the experiment, 1500 ml hydroxyethylstarch was infused during the clamping period to optimize preload prior to the declamping and during the first 30 min of reperfusion to prevent declamping-associated hypotension. Continuous intravenous norepinephrine was incrementally adjusted as long as needed to maintain MAPprox higher than 80% of baseline. Further data sets were obtained 120 and 240 min after declamping. At the end of the experiment the spinal cord was taken for hematoxylin-eosin and Nissl staining as well as for immunohistochemistry.

Statistical analysis

All data are presented as median and range. After exclusion of normal distribution using the Kolmogorov–Smirnov test, Friedman's repeated measures analysis of variance on ranks with post-hoc multiple comparison procedure (Dunn's method) was used for data analysis within the experimental group. The Mann–Whitney rank sum test was performed to analyze intergroup differences. Differences with a p value less than 0.05 were regarded as statistically significant.

Results

Hemodynamic, gas exchange, and acid-base data are summarized in the accompanying Electronic Supplementary Material (ESM; S.T1). Except for a clinically negligible, albeit statistically significant lower MAPprox and MAPdist in the INO1001-treated animals at the end of the experiment, none of hemodynamic, gas exchange, or acid-base variables or tumor necrosis factor-α levels showed any intergroup difference. The duration of the norepinephrine infusion and the total norepinephrine dose administered tended to be lower in the INO1001-treated animals, without, however, reaching statistical significance. S.F1 shows that the INO1001 plasma concentrations were always at least three times higher than 0.1 mM.

Immunohistochemistry showed extremely rare staining for apoptotic nuclei (data not shown), and staining for the cyclin-dependent kinase inhibitor (CDKI) genes p21 and p27 did not differ between the two groups (ESM, S.T2). Hematoxylin-eosin staining showed only minor inflammation as detected by the presence of lymphocytes, which also did not differ between the groups. By contrast, there was marked neuronal damage which presented as vacuolization and cytoplasmic swelling. Some vacuoles were surrounded by membranes, indicating that these vacuoles were due to lysosomal swelling as occurs during cellular edema. In the INO1001-treated animals this neuronal damage was markedly less pronounced, particularly in the central and anterior gray matter of the spinal cord. Typical examples are demonstrated in Fig. 1. This result was underscored by the Nissl staining (Fig. 2), the quantitative analysis showing a 70% reduction in the number of damaged neurons as a fraction of the total number of neurons evaluated (see Table 1).

Fig. 1
figure 1

Hematoxylin-eosin staining in an INO1001-treated animal (left) and an animal of the vehicle group (right). Note the well-maintained nerve cells in the central and anterior gray matter of the spinal cord after treatment with INO1001, while the control animal (right) showed strongly swollen nerve cell with cytoplasmic vacuolization and a disintegrating nerve cell, with greatly reduced staining

Full size image
Fig. 2
figure 2

Cresyl-violet (Nissl) staining in a control animal showing overview of the spinal cord near the central canal with basally contiguous sections of the anterior gray matter at × 5 (a) and a frontal section of the central and anterior gray matter of the spinal cord × 20 (b) magnification, the latter demonstrating disintegrating nerve cells, with greatly reduced staining

Full size image
Table 1 Proportion of damaged neurons in the total number of neurons; data are medians of the mean of all fields of view in each animal (parentheses range)
Full size table

Discussion

The present study tested the hypothesis that the protective effect of the selective PARP-1 inhibitor PJ34 reported in a murine model could be confirmed using another PARP-1 inhibitor, INO1001, in a clinically relevant porcine model of thoracic aortic cross-clamping-induced I/R injury. Swine were investigated because of their striking similarity with humans with respect to both the arterial vascularization of the spinal cord [20–22] and their susceptibility to oxidative stress and tissue antioxidant profiles [23, 24].

Our data that INO1001 reduces neuronal damage in the spinal cord confirm previous data using PJ34 in mice [7]. However, no difference was observed in the expression of the CDKI genes p21 and p27 or in the number of apoptotic nuclei. In fact, we found hardly any apoptotic cells, regardless of the treatment. This observation seems to be in contrast with literature findings reporting a reduced number of TUNEL-positive cells 24 h postischemia in rats after pretreatment with the PARP inhibitor 3-aminobenzamide prior to a 2-h middle cerebral artery occlusion [5]. Other authors, however, have not found different density of TUNEL labeling despite a decreased total number of TUNEL-positive cells reflecting the reduced infarct size in PARP–/– mice and animals pretreated with 3-aminobenzamide [2]. Equivocal reports are available on the occurrence of apoptotic cells after spinal cord I/R injury: Short-term aortic occlusion of 15–40 min was affiliated with neuronal apoptosis in rodents [25–28], but more recent studies have not confirmed these findings [29, 30]. Our observation of marked neuronal damage without TUNEL-positive neurons well agrees with recent data by Papakostas et al. [31] who also did not find TUNEL-positive neurons after 45 min of aortic occlusion in pigs, which was referred to motor neuron death from necrosis. Finally, our findings are further underscored by the similar p21 and p27 expression: albeit the enhanced expression of p21 in neurons surviving transient forebrain ischemia was referred to as an adaptive response to cerebral ischemia and reperfusion [32] and spinal cord demyelination [33, 34], p21 and p27 expression in other tissues subjected to I/R injury [35–37] was directly related to the amount of DNA damage and/or apoptosis. Since the increased expression of the cell cycle regulatory CDKI genes is referred to facilitate DNA repair [38] rather than being a marker of neuronal cell death [39], it is consequently unlikely that different CDKI gene expression is observed when only negligible apoptosis is present.

Limitations of the study

One limitation of our study certainly is that we do not have a direct confirmation of efficient PARP blockade since we performed neither tissue poly(ADP-ribose) staining nor direct measurements of PARP activity. The INO1001 blood levels, however, were always at least three times higher than those previously resulting in near-complete PARP inhibition [10, 40, 41]. Furthermore, the INO1001 dose chosen in our experiment had abolished tissue poly(ADP-ribose)-staining after cardiopulmonary bypass [15–18], hence suggesting efficient PARP inhibition. Unfortunately, we were unable to assess spinal cord function using neurological scoring or evoked motor potentials. Thus we can only speculate whether the reduced spinal cord morphological damage was associated with improved neurological outcome. It should be noted, however, that several previous reports from porcine models with aortic occlusion times of 30–60 min showed a close correlation between the degree of histomorphological damage in the spinal cord and the severity of neurological impairment using the Tarlov scale [12, 19, 31, 42–44] and/or evoked motor potentials [19]. Finally, recent experiments showed that after 45 min of aortic occlusion resulted in complete loss of evoked motor potentials [19] and paraplegia (unpublished data). Thus it is likely that the marked reduction in neuronal damage was affiliated with improved spinal cord function.

Conclusion

In a clinically relevant porcine model of aortic cross-clamping the selective PARP-1 inhibitor INO1001 markedly reduced the I/R-induced morphological spinal cord injury. Given the close correlation reported in the literature between spinal cord morphological damage and impaired function, PARP inhibition may be a promising approach to improve spinal cord recovery after aortic cross-clamping.