The three-dimensional architecture of the notochordal nucleus pulposus: novel observations on cell structures in the canine intervertebral disc


Dr N. A. Duncan, Department of Civil Engineering, 2500, University Drive NW, Calgary, Alberta T2N 1N4, Canada. E-mail:


Cells from the nucleus pulposus of young (< 2 years) and old (> 5 years) non-chondrodystrophoid dogs were studied using routine histology, confocal laser scanning microscopy and transmission electron microscopy. The architecture of cell structures – from the tissue scale down to subcellular scale – was reported. Clusters of notochordal cells were observed in young nuclei pulposi, ranging from 10 to 426 cells each. These clusters resisted mechanical disruption and showed evidence of cell–cell signalling via gap junctions. Cells (30–40 µm in diameter) within the clusters had a physaliferous appearance, containing numerous large inclusions which ranged from 1 to 20 µm in diameter. The inclusions were surrounded by a dense actin cortex but were not contained by a lipid bilayer. The contents of the inclusions were determined not to be predominantly carbohydrate or neutral lipid as assessed by histochemical staining, but the exact composition of the contents remained uncertain. There were striking differences in the cell architecture of young vs. old nuclei pulposi, with a loss of both cell clusters and physaliferous cells during ageing. These observations demonstrate unique cell structures, which may influence our understanding of the differences between notochordal and chondrocytic cells in the nucleus pulposus. Such differences could have substantial impact upon how we think about development, degeneration and repair of the intervertebral disc.


Intervertebral disc-related low back pain is one of the most significant orthopaedic disorders in industrialized society (Frymoyer, 1996). Proper function of the disc depends upon maintenance by a sparse population of cells, and any change in cell density or function can adversely affect the tissue and lead to pathologic degeneration, resulting in herniation of disc material or collapse of the disc space (Garfin & Herkowitz, 1996).

The intervertebral discs (IVDs) join adjacent vertebral bodies, providing flexibility and integrity to the spine. Each disc consists of three major regions: an outer annulus fibrosus consisting of concentric lamellae of collagen fibres, superior and inferior cartilaginous endplates which mark the transition between the IVD and the vertebral bodies, and an inner, soft and highly hydrated nucleus pulposus (Humzah & Soames, 1988). Cell density in the human disc is extremely low – approximately 6000 cells/mm3 in adults (Maroudas et al. 1975) – yet these cells are essential for maintenance of the tissue (Johnstone & Bayliss, 1995).

The intervertebral discs develop from both the embryonic mesenchyme and the notochord (Walmsley, 1953). During embryogenesis, the notochord is surrounded by mesenchymal cells, which synthesize the fibrocartilaginous annulus fibrosus and the osseous vertebral bodies. The notochord virtually disappears in the osseous sites, but persists inside the primitive annulus fibrosus, where it is believed that the entrapped notochord cells synthesize the nucleus pulposus (Horwitz, 1977; Bell, 1996; Hayes et al. 2001). These cells continue to express genes and proteins characteristic of the embryonic notochord, such as collagen type IIA, vimentin, galectin-3, etc. (Stosiek et al. 1988; Maiorano et al. 1992; Fowlis et al. 1995; Gotz et al. 1997; Zhu et al. 2001). In some species, the notochordal cells persist through most of adult life, while in other species, they gradually disappear during ageing. The cells of these adult nuclei pulposi resemble articular chondrocytes (Walmsley, 1953; Trout et al. 1982a; Maldonado & Oegema, 1992; Errington et al. 1998), being small and spherical, and occurring either singly or in clusters of 4–6 cells (Maroudas et al. 1975). These cells express genes and proteins characteristic of articular cartilage, including aggrecan and collagen type II (Eyre & Muir, 1976; Eyre, 1979; Bayliss et al. 1988; Roberts et al. 1991; Johnstone & Bayliss, 1995). It is presently unclear whether this change in cell population is due to the terminal differentiation of the notochordal cells into the chondrocytic phenotype, or to programmed cell death (apoptosis) and invasion of the nucleus by cells from the cartilaginous endplates or annulus fibrosus (Walmsley, 1953; Butler, 1989; Urban, 1996).

The nucleus pulposus is critical to normal disc health and function, as disc disease often begins with changes in the structure and composition of the nucleus (DePalma & Rothman, 1970; Holm, 1996). During disc degeneration, the nucleus undergoes organizational and biochemical changes which result in alteration in the mechanical function of the disc and eventually lead to failure of the tissue (DePalma & Rothman, 1970; Urban & Roberts, 1995; Dwyer, 1996). Various mechanisms have been suggested as initiators of disc degeneration, including altered mechanical loads (Lotz et al. 1998; Iatridis et al. 1999), altered biochemical and nutritional factors (Fogelholm & Alho, 2001; Horner & Urban, 2001), and changes in cell survival and metabolism (Urban et al. 2000; Ariga et al. 2001). Interestingly, initial signs of disc degeneration can occasionally be observed in the nucleus pulposus shortly after the loss of the notochordal cells, although the peak incidence occurs much later in life (Butler et al. 1990; Videman & Battie, 1996; Praemer et al. 1999). However, the possible causal connection between loss of notochord cells and disc disease has not been rigorously studied.

The notochordal cells of the young nucleus pulposus may be critical to formation and maintenance of the disc, and the loss of these cells may play be an important initial event in disc ageing and degeneration. Despite the important role these cells play, they have not been extensively studied. Several recent studies have attempted to reconstruct the nucleus pulposus (Sun et al. 2001; Mizuno et al. 2001; Meisel et al. 2002), yet there is little information about the notochordal cells in the nucleus and how they may impact such efforts. If these cells are indeed responsible for formation and early maintenance of the nucleus pulposus, then any attempt to restore or reconstruct the tissue must be rooted in knowledge of how the cells function, how they differ from their chondrocytic successors, and why the shift in cell population occurs. The current study aimed to describe the cellular structure and arrangement of the residual notochordal cells in young canine nuclei pulposi, and furthermore to compare and contrast the notochordal cells to those in nuclei from older, degenerate discs. Here we present the results of a systematic histological, confocal laser scanning microscopic, and electron microscopic examination of cells in the canine nucleus pulposus. We report novel observations on several three-dimensional structures in the tissue, including large interconnected clusters of large-diameter cells which contain inclusions surrounded by densely packed actin filaments.

Materials and methods

Sample preparation

Lumbar IVDs were collected from nine young, skeletally mature, mongrel dogs (age < 2 years, 20–28 kg) and three older mongrel dogs (age > 5 years, 25–27 kg) within two hours of being put-down. The dogs were purpose-bred mongrels from a single breeder using primarily german shepherd and husky stock, and were therefore presumably non-chondrodystrophoid (Hansen, 1952). The lumbar spine was removed en bloc, and whole discs were removed by cutting through the vertebral bodies, leaving the endplates intact. Discs were categorized as either grade I or grade III, based on the Thompson scale, wherein grade I nuclei are translucent and gelatinous with a clear annular–nuclear demarcation, and grade III nuclei contain ‘consolidated fibrous tissue’ with a ‘loss of annular–nuclear demarcation’ (Thompson et al. 1990). Discs exhibiting grade II degeneration were excluded from the study. The different fixation techniques used for the various preparatory methods are summarized in Table 1, and the different stains used are summarized in Table 2. In all treatments, the entire process from removal of the disc to immersion in fixative was less than one hour in order to minimize tissue swelling.

Table 1.  Summary of sample preparation methods
ScaleSample preparationFixation*Stain
  • *

    NBF, neutral buffered formalin; IPA, isopropanol; MeOH, methanol.

TissueWhole IVDNoneMethylene Blue
 Disrupted nucleus pulposusNoneMethylene Blue
CellParaffin sections, whole IVD10% NBFHematoxylin and Eosin
  10% NBFPeriodic Acid Schiff
 Cryogenic sections, whole IVD60% IPAOil Red O

Nucleus pulposus explant in agarose10% NBFPhalloidin/Propidium Iodide
  10% NBFPhalloidin/Nile Red
  −20 °C MeOHAnti-vimentin/Propidium Iodide
  −20 °C MeOHAnti-connexin-43/Propidium Iodide
SubcellularNucleus pulposus explant2% GlutaraldehydeUranyl Acetate/Lead Citrate
Table 2.  Summary of stains used
Methylene BlueCells
Hematoxylin and EosinNuclei and cytoplasm
Periodic Acid SchiffCarbohydrates
Periodic Acid Schiff with amylase treatmentCarbohydrates other than glycogen
Oil Red ONeutral lipids
Propidium IodideNuclei
Nile RedNeutral lipids

Tissue-scale microscopy

For tissue-scale examination, the vertebral endplates were removed and the discs were briefly washed with 1% methylene blue (an aqueous vital dye) to stain the cells. Alternatively, discs were opened with a transverse cut and the nuclei pulposi were removed en bloc. This material was transferred to DMEM and either left untreated or aspirated twice through an 18G needle on a 60-mL syringe to mechanically disrupt the tissue. The nucleus pulposus samples were then incubated in a 37 °C, 5% CO2 incubator for 24 h, at which point the media was decanted and the samples were briefly stained with methylene blue and washed prior to examination with a Zeiss stereophotomicroscope (Stemi 2000, Zeiss) using transmitted light and oblique illumination.

Cell-scale microscopy

For routine histological analysis (transmitted light microscopy), intact discs were fixed in 10% neutral-buffered formalin for 5 days, decalcified through sequential changes of Cal-Ex (Fisher Scientific), embedded in paraffin (Paraplast), sectioned at a thickness of 5 µm, and stained with haematoxylin and eosin. To assess for carbohydrates, Periodic Acid Schiff staining was performed both with and without amylase pretreatment (positive control: kidney). Alternatively, freshly isolated discs were cryoembedded in OCT compound, sectioned at 5 µm thickness, fixed in 60% isopropyl alcohol, and stained with Oil Red O (positive control: subcutaneous fat).

For confocal microscopy, the nuclei pulposi were removed as before, placed into Petri dishes, and covered with a thin layer of 1% low-melting-point agarose in phosphate-buffered saline, pH 7.2. These samples were fixed in either 10% neutral buffered formalin for 10 min or in −20 °C methanol overnight, as appropriate for the staining method to be used. In cases where the boundary between the nucleus pulposus and annulus fibrosus was not distinct – grade III discs – a 6-mm-diameter biopsy punch was used to remove only the central fibrotic region of the disc.

Agarose-embedded nucleus samples were stained with various fluorescent tags and imaged on a confocal microscope. Samples were incubated in 1.5% Triton X-100 for 5 min, followed by 1.5% normal sheep serum for 30 min, and then either 5 µm Oregon Green-labelled phalloidin (Molecular Probes), or a primary antibody against either connexin-43 (Transduction Laboratories; 1 : 100 dilution) or vimentin (Boehringer Mannheim; 1 : 100 dilution), and an FITC-labelled secondary antibody (Sigma; 1 : 100 dilution). All samples were counter-stained with 2.5 µm propidium iodide for 10 min to visualize the cell nuclei, or 1 µg/mL Nile Red for 30 min to detect neutral lipids (Fowler & Greenspan, 1985).

Three-dimensional, fluorescent image stacks were collected on a Zeiss LSM 510 confocal laser scanning microscope using either a 10× 0.30NA lens or a 63× 1.40NA oil-immersion objective lens (Carl Zeiss Inc.) immediately following staining. The laser power setting (488 nm excitation line) was adjusted as required for the particular specimen (typically 20–30%). Images were collected as stacks of optical sections (z = 10 µm on the 10× objective and z= 1 µm on the 63× objective) at 1024 × 1024 pixel resolution with 2-line averaging. The images are presented both as individual optical slices or as two-dimensional projections of a three-dimensional stack, created using the Zeiss LSM 5 Image Browser software (the specific stack depth is indicated in the caption of each image). No further image processing was performed.

Ultrastructural-scale microscopy

For transmission electron microscopy (TEM), the nuclei pulposi were fixed in 2% glutaraldehyde. Samples were then prepared as previously described for meniscal fibrocartilage (Hellio Le Graverand et al. 2001). Briefly, the samples were post-fixed for 1 h in 1% osmium tetroxide in cacodylate buffer (0.1 m, pH 7.4) with added calcium (5 mm), dehydrated to 100% ethanol, and embedded in epoxy resin. Sections were then cut with a diamond knife, mounted on 200 mesh copper grids, and stained with aqueous uranyl acetate and lead citrate. Sections were examined in a Hitachi H-7000 EM operating at 75 kV.

Numerical measures

Various parameters could be quantitatively measured in the samples through analysis of the three-dimensional confocal microscopy image stacks. The number of inclusions per cell and cells per cluster were determined by simple counting of the inclusions or cells in a stack which contained an entire cell or cluster, respectively. The inclusion volume, cluster volume, inclusion volume fraction, and cell density were determined by measuring the area in each optical slice using the Zeiss LSM 5 Image Browser software, multiplying by the slice thickness, and summing over the entire stack (Cavalieri's Principle). The ‘effective radius’, Reff, of a cell was determined as the average radius of n spheres required to fill the cluster volume, or


where Vc is the volume of the cluster and n is the number of cells in the cluster. This radius gives an estimate of the mean volume which a cell occupies within the cluster.


Macroscopic observations

All nine young mongrels exhibited uniform grade I discs with gelatinous nuclei pulposi which could be easily removed from the disc without fibrous tissue contamination (Fig. 1A). The older mongrels exhibited uniform grade III changes with fibrotic nuclei pulposi in all discs of the lumbar spines (Fig. 1B).

Figure 1.

Representative discs from the canine lumbar spines. (A) Young mongrel, grade I; (B) older mongrel, grade III. In the older disc, the gelatinous nucleus had been replaced by fibrous material and there was a loss of annular–nuclear demarcation. No discoloration or fissuring was seen at this stage of degeneration.

Tissue scale

Methylene blue staining of intact grade I discs indicated the presence of cell clusters throughout the nucleus pulposus, as evidenced by discrete dye-retaining bodies (Fig. 2). In contrast, there was no evidence of cell clusters in grade III discs, as the entire disc stained diffusely (not shown). Staining of isolated and cultured nuclei pulposi further supported this observation, as the clusters began to spontaneously separate at the periphery of the sample after 24 h in culture (Fig. 3A,B). These clusters had some structural integrity, as they could be mechanically separated from one another while leaving distinct clusters intact (Fig. 3C,D). Clusters ranged in size from approximately 250 to 1000 µm, and contained from 4 to 100 cells in two-dimensional section. Similar results could be seen with unstained samples viewed on a phase-contrast microscope. Clusters of cells were observed in undisrupted samples (Fig. 3B), and syringe disruption dispersed the clusters from one another (Fig. 3D).

Figure 2.

Grade I disc, methylene blue stain. (A–C) Increasing magnification. Small clusters of dye-retaining material could be seen in the nucleus (arrows), which upon further examination (Fig. 3) appeared to be clusters of cells. Boxes indicate the regions enlarged in the subsequent panels.

Figure 3.

Isolates from a grade I nucleus pulposus after 24 h in culture. (A,C) Phase contrast images and (B,D) methylene blue stained clusters; (A,B) untreated tissue and (C,D) syringe-disrupted tissue. The cell clusters could be seen dispersing from the untreated tissue, and were very clear in the syringe-disrupted sample. The cells are just visible as small spherical bodies (arrows).

Cell scale

Histological sections indicated clusters of large cells in grade I nuclei pulposi, containing from 4 to over 30 cells in two-dimensional section, distributed throughout the nuclei (Fig. 4A). The clusters appeared to be contained by a thin layer of matrix material that physically separated them from one another (Fig. 4B). Throughout the grade I nuclei pulposi, cells containing bubble- or vacuole-like inclusions surrounded by fibrous material were identified (Figs 4C and 5). A small central region of intensely eosinophilic cytoplasm was typically present around the nucleus, while the bubble-like inclusions did not stain with eosin, PAS, or Oil Red O (not shown). The cells were generally large (∼30–40 µm in diameter), and the inclusions took up a substantial volume fraction of the cell. There were very few small, chondrocyte-like cells in the grade I nuclei (< 10% of the total cell count). In contrast, grade III nuclei pulposi contained relatively few cells, usually alone or in small clusters (4–6 cells in two-dimensional section). These cells generally resembled chondrocytes from hyaline cartilage, with smaller diameters (∼10 µm) and no inclusions (Fig. 4D–F).

Figure 4.

Grade I (A–C) and grade III (D–F) nuclei pulposi, haematoxylin- and eosin-stained sections. While clusters of large cells were evident in the grade I nucleus (arrows, A), very few cells could be found in the grade III nucleus (arrows, D). Cell clusters in grade I nuclei appeared to be surrounded by a thin capsule of matrix material (arrow, B), and generally exhibited vacuole- or bubble-like inclusions in the cytoplasm (arrow, C). The extracellular matrix of the grade I tissue was generally homogeneous, while the matrix of the grade III tissue was fibrous and showed regions of stronger staining (dark bands in D–F).

Figure 5.

Hematoxylin and eosin stain of inclusion-containing cells in grade I nucleus pulposus. Small regions immediately around the cell nuclei stained with eosin (arrows), but the inclusions did not stain positive for eosin, Periodic Acid Schiff, or Oil Red O (not shown).

Confocal microscopy further indicated the existence of inclusion-containing cells. Actin staining of cells from grade I discs demonstrated large, densely packed cells (∼30–40 µm in diameter) containing numerous large inclusions which took up approximately 55.8% of the cell volume (Table 3). Actin filaments were densely packed around the inclusion bodies and the nucleus. In cell clusters, bundles of actin often converged at membrane sites in common with adjacent cells, and gave the actin cytoskeleton a bridge-like appearance (Fig. 6A, arrows). Occasionally a cell was found which appeared to contain two nuclei inside the cytoplasm (Fig. 6A, lower right corner). In contrast, cells from grade III discs were small and sparse, and exhibited a morphology similar to articular chondrocytes, with a thin cortex of actin around the nucleus and very little cytoplasmic volume (Fig. 6B). Staining with antibodies against connexin-43 and vimentin indicated the presence of both proteins in the grade I nuclei. Connexin-43 staining was generally punctate, with some concentration in the vicinity of cell–cell junctions (Fig. 7A). Vimentin fibres ran throughout the cytoplasm in many cells, although some cells did not appear to contain any vimentin (Fig. 7). Both connexin-43 and vimentin were found in grade III nuclei pulposi, but at lower levels than in the grade I cells (not shown). Nile Red stained the cell membranes intensely, but did not stain the contents of the inclusions (not shown).

Table 3.  Descriptive statistics for various cell and tissue structures found within grade I canine nuclei pulposi, based upon confocal microscopy image stacks (see text for complete descriptions)
ParameterMean ± SEMMin./Max.MedianN
Inclusions/cell       7.2 ± 0.693/15  621
Inclusion volume (µm3)     1870 ± 531160/621694613
Inclusion volume fraction per cell (%)     55.8 ± 3.332.5/79.7 58.115
Cells/cluster     109 ± 1710/426 77.536
Cluster volume (µm3)1.06 × 107 ± 2.60 × 10−68.50 × 10−5/3.68 × 10−7  6.12 × 10-615
Numerical density (cells/cm3) 1.31 × 105 ± 1.21 × 10−66.78 × 10−6/2.29 × 10−5  1.37 × 10−515
Effective radius (µm)       27.1 ± 0.921.8/32.6 25.915
Figure 6.

Cells from (A) grade I and (B) grade III nuclei pulposi. Green: actin; red: cell nuclei. Note the large inclusions in the cells from grade I nuclei, which were surrounded by actin filaments. These inclusions took up a very large volume fraction of the cell. Also note the apparent bridges of actin between the cells (arrows). In contrast, cells from mature grade III discs were small and dispersed, and the volume fraction of the cytoplasm was low. Single 1-µm optical sections from confocal microscopy.

Figure 7.

Cells from grade I nuclei pulposi: (A) Connexin-43 (green) and (B) vimentin (green) staining (red: cell nuclei). Connexin-43 was scattered over the entire cell surface, with a concentration in the vicinity of cell–cell junctions (arrows). Vimentin formed fibres which ran throughout the cytoplasm of many cells, although some cells did not appear to contain any vimentin (lower left). Three-dimensional projection from 32 1-µm optical sections from confocal microscopy.

Distinct clusters of inclusion-containing cells were also observed on confocal microscopy. Many of the clusters were larger than the three-dimensional field of view available using the 63× objective and contained over 100 cells each (as determined using the 10× objective) (Table 3), but some clusters were small enough to capture in a z-stack (Fig. 8). This image also clearly demonstrated the intercellular actin bridges.

Figure 8.

In situ cell cluster in a grade I nucleus pulposus. The entire cluster was approximately 100 × 80 × 50 µm and contained over 30 cells connected by bands of actin (arrows). Actin staining, 3D reconstruction from 69 1-µm optical sections from confocal microscopy.

Based upon the confocal image stacks, the cells contained an average of 7.2 ± 0.69 inclusions with mean volume 1870 ± 531 µm3 per inclusion, taking up 55.8 ± 3.3% of the total cell volume. The cell clusters contained an average of 109 ± 17 cells, and the clusters took up 1.06 × 107 ± 2.60 × 106 µm3 each, with an average cell density of 1.31 × 105 ± 1.21 × 10−6 cells/cm3. The ‘effective radius’ of a cell was determined to be 27.1 ± 0.9 µm (mean ± SEM, Table 3).

Ultrastructural scale

Transmission electron microscopy revealed numerous unusual features within the cells. The cells in grade I nuclei pulposi contained numerous large vacuous inclusions, which were surrounded by a dense fibrous network – presumably actin filaments (Fig. 9A). The inclusions ranged in size from 1 to 20 µm in diameter, and were generally circular in cross-section. Some inclusions were observed with TEM that were too small to have been detected with confocal microscopy. The inclusions were not membrane-bound, and the contents were generally electron-lucent with a very fine granular material, although some inclusions appeared to contain coarse extracellular matrix material. These ‘mature’ inclusions generally appeared collapsed, i.e. non-convex, as opposed to the electron-lucent, nearly circular inclusions (Fig. 9B). Some inclusions contained membrane-bound ‘subinclusions’ which appeared to contain a different material (Fig. 9C). Occasional multinucleated cells were observed, with no evidence of mitotic bundles or cell membranes between the nuclei. Most of the nuclei contained distinct nucleoli (Fig. 9D). The endoplasmic reticulum was sparse and generally difficult to locate. Very few mitochondria and only small amounts of glycogen were observed.

Figure 9.

Transmission electron micrographs of cells from grade I nuclei pulposi. (A) The electron-lucent inclusions (i) were surrounded by densely packed filaments (f) and lacked a lipid bilayer. (B) Inclusions varied in size and geometry, with some inclusions exhibiting a collapsed appearance and containing what appeared to be extracellular matrix material (arrow). (C) Some inclusions contained membrane-bound ‘subinclusions’. (D) Most cell nuclei contained distinct nucleoli. Occasional examples of multinucleated cells were found, with no mitotic bundles or membranes between the nuclei.

Cells in grade III nuclei pulposi exhibited a markedly different appearance in TEM. The cells were significantly smaller, and did not contain large inclusions (Fig. 10). No multinucleated cells were observed, and the cells did not generally contain distinct nucleoli. The rough endoplasmic reticulum was clearly defined. Few mitochondria were observed, although they were more prevalent than in cells from grade I nuclei pulposi (Fig. 10). Small amounts of glycogen were found within the cytoplasm (not shown).

Figure 10.

Transmission electron micrograph of a cell from grade III nucleus pulposus. Note the significant difference in size and the lack of inclusions. RER: rough endoplasmic reticulum, Mit: mitochondria.


The cells of the embryonic notochord are responsible for synthesis and early maintenance of the nucleus pulposus of the intervertebral disc. Loss of these cells with age may be the first stage of a process which can lead to pathological changes in the discs. This study aimed to describe comprehensively the cellular architecture in both young and older canine nuclei pulposi, ranging from the macroscopic tissue to the ultrastructural scale.

To our knowledge, this study is the first to report an examination of the three-dimensional structure of cell clusters in the nucleus pulposus. Previous authors have reported evidence of cell clustering in the nucleus pulposus based upon two-dimensional sections, and have indicated that the clusters contained up to six cells in section (Maroudas et al. 1975; Roberts et al. 1991). The present study confirms and extends these observations: it appears that young mongrel nuclei pulposi contain cell clusters which are substantially larger than previously reported, ranging from 4 to 100 cells per cluster. The clusters had some physical integrity, as they resisted mechanical disruption. This integrity may have been partly due to an encapsulating membrane, as suggested by the histological sections, and may also have been the result of cell–cell bonds, as suggested by the presence of connexin-43 and dense actin networks between the cells. As connexin-43 is a gap junction protein, it is possible that the cells within a cluster were in direct communication with one another. It is also worth noting that in a separate study, we demonstrated that the size of the cell clusters appeared to vary with species, suggesting a certain variability of cell behaviour and functionality (Hunter et al. 2003). The functional significance of these cell clusters is unclear, and warrants further investigation.

This study is also the first to investigate the three-dimensional structure of the individual notochordal cells and their inclusions. It is interesting to note that the inclusions took up a substantial volume fraction of the cell (55.8%), and that they were apparently surrounded by a dense network of actin filaments. The inclusions were present in a wide range of sizes from less than 1 µm to almost 20 µm in diameter, and seemed to be generally spheroidal in shape. They were not observed in the older grade III discs, which suggests a substantial change in cell function during ageing and/or degeneration. These inclusion-containing cells resemble the ‘physaliferous’ cells reported in both the embryonic notochord and the notochord-derived tumour, chordoma (Horwitz, 1977). However, inclusions were not reported in a previous electron microscopy study of human nuclei pulposi (Trout et al. 1982b), nor were they reported in a confocal microscopy study of embryonic rat nuclei pulposi (Hayes et al. 1999), but similar cells were reported in a histological study of cat nucleus pulposus (Butler & Smith, 1967), and in TEM studies of rabbit, chicken and frog embryonic notochords (Leeson & Leeson, 1958; Jurand, 1962; Bruns & Gross, 1970). Together, these data suggest that the presence of inclusions inside the cells may be a species-dependent phenomenon, or that the natural history of the changes in cell population may vary substantially with species. Furthermore, we speculate that cells from non-chondrodystrophoid canine nuclei pulposi are closer to an embryonic state than cells from human tissue.

At present, we can only speculate as to the exact contents of the inclusions. They did not stain with eosin, PAS, Oil Red O, or Nile Red, and they were generally electron-lucent. This suggests that the contents were not typical cytoplasm, and may not have been predominantly carbohydrate or neutral lipid. Alternatively, the contents may have been soluble in the histological solvents, and were lost during preparation. They did not require a lipid bilayer to partition the contents from the cytoplasm, but they may have produced considerable swelling pressure, as evidenced by the dense cortical actin filaments around them. The lack of a lipid bilayer makes it doubtful that the inclusions were involved in membrane-mediated exocytosis, despite the observed phenomenon of ‘collapsed’ inclusions which appeared to contain matrix material.

Given the present data and based upon a process of elimination, we can speculate that the inclusions are charged lipid deposits, which may provide an energy store for the cells. The nucleus pulposus is an avascular tissue, and therefore experiences a very poor nutrient supply. It seems possible that the inclusions provide a secondary source of energy for the cells. When these stores are used up, either due to further reduction in the external nutrient supply (Bibby et al. 2001) or due to excess energy utilization to repair damage, the cells starve, thus explaining the loss of notochordal cells during ageing. This functionality is at present purely hypothetical, but certainly these curious structures deserve further study to determine their functional role in the nucleus pulposus.

The cell clusters may be important for proper cell function in the tissue. Maldonado & Oegema (1992) previously reported a study wherein cells were isolated from young non-chondrodystrophoid canine lumbar nuclei pulposi and cultured in alginate beads. The cells were isolated using pronase/collagenase digestion followed by filtration through a 70-µm filter, so it is doubtful that any large cell clusters could have remained intact. Interestingly, they reported that the large notochordal cells began to die after 10 days in culture. We speculate that disruption of the cell clusters may have had an adverse effect upon cell survival. If so, these results could have important implications for any attempt to culture notochordal cells – particularly in attempts to tissue-engineer an intervertebral disc with cells from a notochord-retaining species.

The canine model is relatively uncommon in IVD studies, and could be considered a limitation of the current work. However, IVD disorders are commonly seen in clinical veterinary practice, so the canine model may shed some light upon human disc disease. The Thompson grading scale was designed for human IVDs, and has limited applicability to canine discs. This scale was chosen to emphasize the changes that occurred in the canine discs with a familiar grading scale, rather than to establish absolute comparisons to human tissues. Alternatively, the discs could be graded as Stage 1 (grade I discs) or Stage 2 (grade III discs) on the scale proposed by Bray & Burbidge (1998).

Non-chondrodystrophoid dogs are reported to lose their notochordal cells relatively late in life, around 5 years of age (Hansen, 1952), whereas humans generally lose these cells during the first decade of life (Trout et al. 1982b). However, the remarkable age-related degenerative changes that occur in the dog disc (Fig. 1) compare well with early morphological changes in the human disc. Therefore, the canine model may prove useful in studying the relationship between notochordal cells and disc degeneration. Subsequent studies with chondrodystrophic breeds (which lose their notochordal cells during the first year of life; Hansen, 1952) will further clarify this relationship.

The nucleus pulposus is one of the few tissues in the body to undergo such a dramatic shift in cell population during ageing. The cells of the young nucleus differ from those of the adult tissue in morphology, gene expression and protein synthesis. These changes may play an important role in the organizational and compositional changes that occur in the intervertebral disc during ageing and degeneration. An appreciation for the tissue and cellular architecture of the young nucleus pulposus may have implications for the development, maintenance, degradation and potential repair of degenerative intervertebral discs. This study presented unique three-dimensional structures within the canine nucleus pulposus, and revealed new aspects of cellular and cytoplasmic organization that have not been previously described. However, we have only begun to understand even the most basic aspects of cell behaviour during the transition from young to adult to degenerative tissue. Further study of the notochordal cells and the events that trigger their transformation into or replacement by chondrocytic cells may provide useful insights into disc degeneration and potential means of treating the damaged disc.


C.J.H. is supported by a post-doctoral fellowship from the Canadian Institutes of Health Research and the McCaig Centre for Joint Injury and Arthritis Research at the University of Calgary, J.R.M. is a senior scholar of The Arthritis Society, and N.A.D. is supported by a Canada Research Chair in Orthopaedic Bioengineering. We would like to thank Michael Kimm for his assistance in collecting tissue samples and preparing histological sections, Leona Barclay for preparation of the electron microscopy sections and Dr J. B. Rattner for his input.