Any new therapy developed for the treatment of intervertebral disc degeneration to alleviate or delay the onset of associated back pain must go through the inevitable series of testing prior to human trials. Models of disc degeneration have severe limitations  with none providing the ideal test environment. Certain properties, for example cytotoxicity, can be carried out on cell cultures. However, disc cells isolated from their extracellular matrix have a completely different morphology and size than when retained within it , such that explant models are more likely to reflect the in vivo situation than isolated cell models. Whilst the use of explant models will not preclude the complete use of whole animal in vivo models, they may be able to answer many of the questions which have to be asked and minimise the number and hence reduce the investment in cost and ethics otherwise required to take a new product to clinical trials and the marketplace.
There are already several in vitro models described in the literature which, although they are characterised by additional techniques (eg cell number and viability, immunohistochemistry, qRT-PCR) to those used in this study, many involve small species, for example mouse , rat [9, 10] or rabbit . These species have inherent limitations of size and notochordal cells within the nucleus pulposus region. More recently larger discs from bigger animals such as sheep and cattle have been cultured as explant models with the possibility of studying mechanisms of disc degeneration [12, 13]. Likewise several models have utilised enzyme digestion of the extracellular matrix (summarised in reviews [4, 14]. Many of these were developed to mimic and understand better the physiological response to the treatment of chemonucleolysis. More recently there is renewed interest in this as a clinical treatment, with chondroitinase ABC as an alternative enzyme. Imai et al  use this in a rabbit model of chemonucleolysis, whilst it has also been suggested to be useful for a model of degeneration in mice. Antoniou et al  have used different enzymes to examine the effect of degradation of individual matrix components (for example, collagen or proteoglycans) on magnetic resonance imaging parameters. Little has been detailed, however, on the GAG profile, which is probably the most important biochemical parameter relating to disc degeneration.
There are real physical limitations of injecting any significant volume into a healthy disc (though small volumes such as might be used for testing growth factor or even cell therapy can obviously be introduced). In contrast, in the clinical setting where it is envisaged that newly designed implants might be injected into a degenerate human lumbar intervertebral disc, it can often be easy to inject volumes of at least 2 ml of a substance (i.e. approximately 10% of its volume). Hence there is a requirement for a model system of a degenerate disc which is larger than the animal discs most commonly used in such tests and which has properties more similar to those of the human disc. The bovine coccygeal disc provides such samples, with similar properties in terms of composition, swelling pressure and metabolism to those seen in human discs . We have shown in this study that breakdown and loss of GAGs can be induced, sufficient to inject a potential nuclear replacement for subsequent testing. (Several nuclear replacements are being investigated by other groups, for example, NuCore® Injectable Nucleus (Spine Wave Inc, Shelton, CT, USA) and the BioDisc™ (Cryolife Inc, Kennesaw, GA, USA)).
There have been some studies examining explant culture with and without the vertebral endplates attached, indicating that bovine discs without endplates and ovine with endplates could maintain viability for up to 1 week [12, 13]. The vertebral bone (and ligamentous tissue) certainly seemed to be an advantage in the present study in that swelling was severely restricted, with no significant or obvious swelling of the whole motion segment, although it is likely that the similar water content seen in the treated and control discs was due to lack of restraint against swelling via musculature, ligaments etc such as would occur in vivo. In addition, the enzymes used to treat the motion segments did not appear to have affected the histology of the remaining surrounding tissue greatly or any differently from that seen in the untreated controls. Both trypsin and papain created a 'space' allowing material to be injected, but trypsin may be preferable on a cost-basis and is certainly much cheaper than other enzymes such as chymopapain or collagenase. Likewise, there was little difference between time-points studied, particularly for papain, hence this would be recommended to use for the shortest interval.
This model could have several uses for testing newly developed products. These could include application systems, mechanical properties of implants or their adhesive and interaction properties. Indeed, preliminary data has been obtained with this model, whereby an injectable gel has been inserted and tested under either hydrostatic load (approximately 4 kg for 2 hrs) or cyclic loading (4 kg, 600 cycles at a frequency of 1 Hz; Dr S Sarit, personal communication). In addition to testing implants, they could also be used to assess biological approaches and, for example, monitor injected cells. When developing new therapies, the choice of the ideal test system depends very much on the questions being asked. In this study we have shown that the use of a general proteolytic enzyme together with a bovine coccygeal explant system can provide a means of producing a cheap, reproducible system suitable for testing some of the new generation of injectable implants which are being developed in the spine field.