Differential expression level of cytokeratin 8 in cells of the bovine nucleus pulposus complicates the search for specific intervertebral disc cell markers

Low back pain constitutes a major health problem and a huge economic burden [1]. It is highly associated with degeneration of the intervertebral disc [2]. The earliest degenerative changes are seen in the central region of the disc, the nucleus pulposus (NP) [3], and are characterized initially by loss of proteoglycans and finally by loss of matrix integrity [3, 4]. Treatments on offer are still mainly palliative or surgical and do not improve the ability of the disc to regain its original architecture and function. Biological approaches, particularly those that aim to produce a tissue-engineered disc or to insert cells into the damaged NP to regenerate the matrix and restore the disc's biomechanical function, are seen as a potential alternative [5].

Implementation of cell therapies for repairing the disc is limited by lack of an appropriate cell source as healthy disc cells are not available for expansion and treatment [6]. Efforts therefore have concentrated on differentiating stem cells, particularly mesenchymal stem cells (MSCs), into disc cells both in vitro and in vivo [7]. The success of the differentiation protocols used is, however, uncertain as the markers mainly used (such as expression of collagen II, aggrecan, and sox 9 [8]) are expressed by all cartilage cells. The MSCs thus could be differentiating equally well into articular chondrocytes (ACs) as into 'disc-like cells' [8, 9]. It is, however, vital for successful repair that a matrix that is permissive of the disc's biomechanical requirements be produced by the differentiated cells. Although disc cells and ACs express many of the same macromolecules, there are distinct differences in the overall composition and biomechanical properties of the matrix produced. Disc nucleus cells, for instance, produce a loose collagen II network that supports the disc's requirements for flexibility while ACs produce a much more rigid matrix through a tightly cross-linked collagen II network; these differences possibly arise from differences in splice variants and post-translational modifications of the collagen molecules produced by these two different cell types [10‐12]. Specific disc cell markers to ensure that MSCs differentiate into disc cells rather than into some other cartilaginous cell type are thus an essential requirement for success of cell implantation therapies.

Microarray screens have been used to define markers that will distinguish disc cells from other cartilage cells [13‐15]. In addition, expression of HIF (hypoxia-inducible factor) and GLUT (glucose transporter) isoforms have been suggested as markers [13, 16]. In general, these studies, while identifying differences in the level of expression of a number of genes or proteins between annulus cells, nucleus cells, and ACs, have found no specific markers, apart from CD24. Moreover, the studies have been carried out mainly on rats, which because of a difference in NP cell phenotype are not a good model for the human disc nucleus [17]. The NP of all mammals, including humans, originates from the notochord [18, 19] and in early fetal life contains clusters of large vacuolated cells producing a fluid matrix of low collagen content [20, 21]. In some animals such as rats, pigs, and rabbits, these notochordal cells persist well into adulthood and even throughout life [17, 22]. However, in other species, including humans and cattle, the notochordal cell clusters disappear early in life and are replaced by the smaller cells of chondrocyte-like appearance seen in the adult NP; these produce a firmer, more collagenous matrix [20‐23]. Hence, as shown in a recent study [24], the markers produced from studies on rodents may not be specific for human discs and may not be relevant for repair studies on human discs.

Here, we describe another approach to the search for specific disc cell markers. First, as non-degenerate adult human discs are not available to us, we used cells from the nucleus of young adult bovine caudal discs; these cells are thought be phenotypically similar to adult human nucleus cells and produce a matrix similar to that of adult humans [17, 25]. Second, we looked for markers at the protein rather than at the gene expression level. We used the differential in-gel electrophoresis (DIGE) technology to compare disc cells with ACs as these two cell types have a similar morphology and both produce typical cartilaginous markers such as aggrecan and collagen II. Indeed, disc nucleus cells are often referred to as chondrocyte-like cells [26]. We found 14 proteins that were expressed only by disc or only by cartilage cells. Here, we concentrate on findings related to cytokeratin 8 (CK8), an intermediate filament protein strongly expressed by our disc cell preparation and not expressed at the protein level by ACs or cells from the annulus fibrosus (AF). We found that CK8 was differentially expressed in cells of the bovine NP. The apparent cellular heterogeneity raises questions about the search for specific cellular markers to identify cells of the mature NP.

In this study, we identified two distinct cell populations within the bovine NP based on the expression of the intermediate filament protein CK8. We found by proteomic tools (Figure 1), by immunostaining isolated cells (Figure 2a), and by immunostaining tissue sections (Figure 3) that CK8 was exclusively present in the NP and was not seen in the AF or in articular cartilage. However, we also found that only approximately 10% of the NP cells were positive for CK8 (Figure 2b). These CK8-positive cells were not uniformly distributed throughout the tissue but were grouped in small clusters in isolated regions of the matrix which appeared more gelatinous than the surrounding matrix of the nucleus (Figure 3a,b); these clusters were found, independently of stages of maturity, in all discs examined (Figure 6).

The apparent co-existence of distinct cell populations in the bovine disc raises a number of questions. First, as notochordal disc cells are known to express CK8 [19], are the CK8-positive cells remnants of the original notochordal population of the disc? There are several indications that support this idea. Intermediate filaments such as CK8 are often used to classify the origin of a tissue as differentiated cells usually express only one intermediate filament type; tissues are classified as epithelial when expressing CK8 [36] or as mesenchymal when they express vimentin [33]. Intermediate filaments from two different families can be expressed simultaneously, however, in tumors or in development [37, 38]. Disc notochordal cells, epithelial in origin [19], are CK8-positive as expected but are also positive for vimentin [19, 30, 39] as we saw in the porcine discs (Figures 4 and 5). All bovine NP and also all AF cells were positive for vimentin (Figure 4) as reported previously [35]. Thus, like disc notochordal cells [19, 39], the CK8 subpopulation of the bovine NP is also vimentin-positive (Figures 4 and 5), supporting the idea that these cells could have originated from the original population of notochordal cells. This hypothesis is strengthened by the organization of the bovine CK8-positive cells in clusters and by the more gelatinous texture of the surrounding matrix (Figures 3a,b and 6), both features of the notochordal nucleus. However, notochordal cells of the NP have a morphology very different from that of chondrocyte-like nucleus cells as the former are markedly greater in diameter and contain large vacuoles [20, 28]; indeed, notochordal cells of the disc are identified mostly by their morphological features [31]. Thus, if the CK8-positive cells are notochordal-like, they would have had to undergo a large size reduction since we found that the size distributions of CK8-positive and -negative cell populations were very similar (Figure 2c). Such loss of size of NP cells has been seen experimentally; the diameter of porcine notochordal cells decreased markedly over 15 days in culture with loss of vacuoles and approached that of chondrocyte-like NP cells [40]. In addition, rabbit notochordal cells have been shown to differentiate toward 'chondrocyte-like' cells when maintained in culture [41]. Thus, on balance, it seems likely that the CK8-positive cells are remnants of the original cell population of the bovine disc. The question, however, could be resolved definitively by gene or proteomic profiling of the two cell populations; at present, only gene profiles from rat notochordal cells and mature chondrodystrophoid canine disc tissue are available [14, 15, 42].

Second, if, as we suggest above, the bovine disc contains a notochordal-like cell population, does the adult human disc do so too? The NP of bovine discs is reported to be similar to that of human discs in regard both to matrix composition and to cell phenotype [17, 22]; thus, if notochordal remnants are present in bovine discs, it is possible that they are also present in adult human discs. If, as in bovine discs, the majority of any remaining CK8-positive cells shrink and lose vacuoles (Figure 2a), only the small fraction of larger cells would be identifiable morphologically as notochordal cells; indeed, some few cells resembling notochordal cells have been reported in adult discs [43, 44]. Immunohistochemical studies have, however, identified cytokeratin-positive cells and also the co-expression of CK8 and vimentin in adult human NP [45, 46], but whether all cells or only a subpopulation was immunopositive was not stated in these reports.

If, as these various reports suggest, cells from the original notochordal population are retained in discs regarded as non-notochordal, do these cells have any functional significance? Notochordal cell-conditioned medium or co-culture of notochordal and adult NP cells has been found to stimulate matrix production by adult NP cells; in addition, notochordal cells are reported to retard disc degeneration when inserted into damaged rabbit discs [47‐50]. These results could be explained by the finding that notochordal cells secrete growth factors that stimulate production of extracellular matrix [21, 51]. It has also been suggested that notochordal cell remnants could serve as a stem cell population [20]; indeed, in preliminary experiments, we find that the CK8 population proliferates faster than the CK8-negative population in vitro (data not shown). Functionally distinct non-chondrocytic subpopulations have been identified within the NP of adult human discs [52], indicating the possibility that some notochordal cells survive and remain active. Could it be that the presence of these resident notochordal cells and the growth factors they secrete help in the maintenance of a healthy disc, as suggested by Aguiar and colleagues [47] in a study on dog discs, and that these cells are thus essential for disc homeostasis?

Finally, does the presence of CK8-positive cells provide any information on the origin of the chondrocyte-like cells of the adult human or bovine disc? Studies in rabbits suggest that notochordal cells die off to be replaced by mesenchymal cells originating in the inner annulus or cartilage end-plate [32]. However, it has also been suggested that notochordal cells could differentiate into the adult disc nucleus cell phenotype [20] since chordomas, which arise from embryonic notochordal remnants [13], show chondrogenic potential and can differentiate into cartilage-type cells expressing collagen II and aggrecan [53]. Perhaps the possible differentiation pathway from notochordal to mature NP discs should be revisited.

The finding of a subpopulation of notochordal-like cells in young adult bovine NP is in agreement with a few observations on human discs in the literature [45, 46] and stresses the importance of understanding the cell phenotypes of the intervertebral disc. The bovine disc serves as a model of human disc in many studies [25], and the finding that the nucleus contains at least two subpopulations should be considered in experimental design and analysis. In addition, at present, there are many studies investigating the use of MSCs for disc repair [9, 54, 55]; it is thus essential to understand whether these cells are progenitors of the adult disc population or not [21, 31]. In addition, if, as has been suggested [20, 31], a subpopulation of notochordal cells is required to maintain disc health, strategies for disc cell repair therapies will need to encompass this cell type.