MMPs in the CNS are largely synthesized by endothelial cells, microglia, astrocytes and neurons under normal conditions and can be produced at elevated levels in response to stress . A number of reports have indicated that MMPs are involved in the pathogenesis of a wide range of diseases and disorders of the CNS, including neurodegenerative diseases such as Alzheimer's disease, amyotrophic lateral sclerosis, Parkinson's disease and progressive supranuclear cerebral palsy [10–14] as well as stroke injuries . It has been suggested that this class of extracellular proteases may be appropriate molecular targets to reduce secondary tissue degeneration and improve functional outcome [7, 16]. Although recent studies have demonstrated the involvement of MMPs in experimental models of spinal cord injury, there has, until now, been only one recent correlative investigation regarding the spatio-temporal distribution of MMP-9 in human spinal cord injury . In the present study, the timing and distribution of MMP-1, -2, -9 and -12, as well as their tissue inhibitors TIMP-1, -2 and -3 were investigated in both normal and traumatically injured samples of human spinal cord. In an attempt to obtain tissue samples that were as comparable as possible, an emphasis has been placed on using samples obtained from cases which underwent similar lesion type and severity. Therefore, all patients with traumatic SCI suffered severe injuries of the maceration type (for a detailed description of the morphology of the lesion sites, see ) and were clinically diagnosed as having "complete" injuries. The paucity of human specimens led to the inclusion of cases from different age groups and varying levels of injury. Thus, the present data needs to be interpreted with care and can, by no means, be easily generalised to patients suffering from other types or severities of SCI. Thus, it is clear that future studies will be needed to verify if a similar pattern of post-traumatic MMP and TIMP expression at the lesion site can be found in less severe- as well as other types of SCI (e.g. following laceration and contusion type injuries). The comparison of the present data with previous results demonstrated that the spatio-temporal pattern of microglia/macrophages and astrocytic responses (i.e. the main cell populations demonstrating post traumatic MMP immunoreactivity in the present study) supports earlier investigations in human SCI .
Expression of MMPs and TIMPs in the normal CNS
In the normal, unlesioned human brain and spinal cord, MMP and TIMP immunoreactivity was generally scarce. The data obtained from prior investigations on the expression of MMP-1, -2 and -9 in the human nervous system have been inconsistent, reporting either no immunoreactivity [17, 19–21] or MMP-1 positive microglia , MMP-2 positive microglia, pericytes and blood vessels [7, 14, 22, 23], MMP-9 immunoreactive neurons, microglia and single intravascular monocytes [10, 22, 23] and MMP-12 immunoreactive microglia and astrocytes . No immunoreactivity for TIMP-1 and -3 has been detected. However, TIMP-2 has been found in endothelial cells and occasional neurons and astrocytes .
In the present investigation, neuronal staining for MMP-1, TIMP-2 and -3 was detected as well as individual MMP-9 positive intravascular monocytes. In general, the inevitable delays that occur before human post mortem tissues undergo fixation will result in sub-optimal antigen preservation. It therefore always remains possible that low levels of antigen, below the level of detection by the current immunohistochemical approach, may still be present at pathophysiologically relevant concentrations. All molecules were, nonetheless, clearly detectable in either both control or pathological cases.
Whereas the pattern of MMP-9 immunoreactivity observed in the present investigation supports pervious data , the neuronal expression of MMP-1 has, so far, not been reported. The expression of TIMP-2 and -3 in different populations of neurons, though not previously described in human material, is in line with animal data where both inhibitors were detected in cortical and cerebellar neurons .
The lack of immunoreactivity for MMP-2, -12 and TIMP-1 in the present control cases is unlikely to be attributable to a lack of sensitivity, since a clear signal could be detected in sections of traumatically injured spinal cord. It is possible, however remote, that the earlier described immunoreactivity in astrocytes, microglia and blood vessels may reflect more activated populations of cells in the previous control cases, despite the lack of morphological signs of disease. In contrast to an extensive literature on the function of MMPs and their inhibitors in pathological situations, there is hardly any information on their function under normal conditions apart from those instances where they have been associated with plasticity .
Expression of MMPs and TIMPs following human spinal cord injury
In the present investigation, a post-traumatic up-regulation of MMP-1, -2, -9 and -12 was detected. Following experimental SCI, several studies have reported the involvement of various MMPs and TIMPs in the post-traumatic events at and around the lesion site. In particular, the temporal expression pattern of almost all MMPs has been studied after compression injury to the mouse spinal cord . An up-regulation of multiple MMPs, including MMP-2, -9 and -12 was detected, which occurred in a time-dependent manner, i.e. an early elevation of one group of proteins including MMP-9 and a more delayed elevation of others including MMP-2 and -12. In contrast to the present investigation, MMP-1 could not be detected in mouse spinal cord samples, at least up to 5 days after injury .
The present post mortem investigation of human material revealed a lesion-induced bi-phasic pattern of raised MMP-1 levels at and around the lesion site. At survival times of up to 8 days, MMP-1 was expressed in macrophages and microglia within the lesion epicentre, however, at the later survival times of 4 months to 1 year, activated astrocytes at the border of the glial scar became strongly MMP-1 immunoreactive. The delayed up-regulation of MMP-1 in astrocytes in the spinal cord parenchyma around the lesion site might be a consequence of earlier pro-inflammatory cytokine production since in vitro investigations have demonstrated the release of MMP-1 from astrocytes when stimulated by TNF-alpha or IL-1β . Furthermore, cytokines regulate the activity of both gelatinases A and B (MMP-2 and -9) in cultured rat astrocytes [25, 26]. The early induction of MMP-1 in the present investigation may have been associated with the further pathological breakdown of the blood-spinal cord barrier (see later), however, the role of the later induction MMP-1 in reactive astroglia expression is less clear. Experimental studies on wound repair after skin lesions have demonstrated an increased expression of MMPs, including MMP-1 in the scarless healing process of fetal injuries [27, 28]. Furthermore, an investigation into regeneration-associated factors in the adult rat optic nerve revealed an increased post-traumatic expression of MMP-1 in peri-lesional astrocytes . Therefore, it might be possible that the presence of MMP-1 in the population of peri-lesional astrocytes at the border of the evolving glial scar, might play a role in limiting the extension of this residual tissue barrier.
Several reports have demonstrated post-traumatic and post-ischaemic increases in MMP-2 [6, 30–32]. Increased MMP-2 signals were detected at both protein and mRNA levels, with elevated expression levels lasting several weeks following experimental rat spinal cord injury . A recent investigation using MMP-2 knock-out mice demonstrated a more severe outcome following traumatic SCI in mice lacking MMP-2 . Knock-out mice demonstrated a reduced white matter sparing, a more widespread reactive astrogliosis as well as an impairment in spontaneous locomotor recovery compared to wild-type littermates. In the wild-type animals undergoing SCI, MMP-2 was mainly expressed in astrocytes and some macrophages for up to 2 weeks at the borders of the lesion site. In the present human material, there was an early and brief induction of MMP-2 in macrophages and microglia which lasted from 2 to 8 days after injury. By 24 days, when the first clear morphological indication of scar formation was visible, MMP-2 expression was on the borderline of detectability. In contrast to the situation in experimental animals, it is likely that MMP-2 expression after human traumatic SCI is more involved with the early vascular and inflammatory events than with reactive astrocytosis. In contrast to SCI, previous studies using post mortem human tissue of multi-infarct induced dementia, ALS and Parkinson's disease demonstrated the expression of MMP-2 at the border of the evolving astroglial scar or in astroglia in the cerebral cortex or substantia nigra [12, 13, 30]. The cellular distribution of MMP-2 expression in human pathology therefore appears to differ significantly in relation to the particular disease or disorder under investigation.
Contused mouse spinal cord demonstrated substantial blood vessel wall and astrocytic MMP-9 immunoreactivity within 3 days of injury . Furthermore, the use of MMP-9 knock-out mice following SCI demonstrated a favourable functional outcome compared to wild type animals, with reduced blood-spinal cord barrier permeability after the lesion and increased spared white matter . In the present study of severe traumatic SCI of the maceration type of injury, MMP-9 was mostly expressed in macrophages/microglia, where levels rose progressively from 1 week to 3 weeks after injury. However, some neutrophils were also found to be MMP-9 immunopositive at survival times of up to 8 days after injury. Although delayed, this spatio-temporal expression pattern is similar to that observed in experimental investigations [6, 31, 32]. However, a recent investigation of human traumatic SCI, using a heterogeneous sample of patients, found an early increase in MMP-9 at the lesion site for up to 10 days after injury and showed neutrophils to be the only cellular source . One possible explanation for the differences observed between the present investigation and that of Fleming and colleagues may be the differences in severity and mechanism of injury for the cases chosen. Therefore, additional studies which assess the influence of different mechanisms of injury (e.g. maceration or laceration) on the spatio-temporal patterns of cellular invasion and protein expression would be useful. This is of particular importance when experimental strategies are transferred to the clinical domain.
The up-regulation of MMP-12 following human SCI was delayed for up to 24 days, at which time an abrupt increase in the number of immunoreactive microglia/macrophages could be detected. Macrophages have already been described as a principal source of MMP-12 . The use of PCR following experimental mouse SCI revealed that, of all MMPs induced following traumatic injury, the induction of MMP-12 was by far the most striking, being approximately 189-fold greater than the basal levels. Furthermore, the importance of MMP-12 expression was demonstrated by using knock-out mice, in which an improved functional outcome was observed following SCI. This beneficial effect was reported to be due to a reduced permeability of the blood-spinal cord barrier and hence reduced infiltration by neutrophils and macrophages, as well as a lack of direct MMP-12-induced toxicity .
In contrast to the MMPs investigated in the present study, the detection and distribution of their inhibitors TIMP-1, -2 and -3 was limited. Only occasional TIMP immuno-positive macrophages could be detected at survival times of 2 – 24 days post injury. Neuronal TIMP-2 immunoreactivity was qualitatively reduced in comparison to control cases. The imbalance between MMPs and their respective TIMPs in certain situations may contribute to the development of pathology. Such an imbalance has already been described following experimental SCI in mice and also in the cerebral cortex of patients suffering PSP [7, 14]. The minor up-regulation of TIMPs detected in the present study is largely in line with experimental studies using mice, in which a short term, transient up-regulation of TIMP-1 was detected following spinal cord compression injuries . Therefore, similar to animal investigations, the strong induction of multiple MMPs after human SCI was not accompanied by a concomitant expression of their inhibitors, allowing these proteins to exert their effects in the lesioned spinal cord. The only clear increase in TIMP immunoreactivity was detected for TIMP-3 at survival times of 8 months and 1 year. In sections from these cases, the same peri-lesional activated astrocytes which expressed MMP-1 were also immunoreactive for TIMP-3. It may be that the up-regulation of TIMP-3 acts to limit the effects of MMP-1, but this suggestion remains speculative.