Characterization of the human nucleus pulposus cell phenotype and evaluation of novel marker gene expression to define adult stem cell differentiation




Development of stem cell therapies for regenerating the nucleus pulposus (NP) are hindered by the lack of specific markers by which to distinguish NP cells from articular chondrocytes (ACs). The purpose of this study was to define the phenotype profile of human NP cells using gene expression profiling and to assess whether the identified markers could distinguish mesenchymal stem cell (MSC) differentiation to a correct NP cell phenotype.


Affymetrix MicroArray analyses were conducted on human NP cells and ACs, and differential expression levels for several positive (NP) and negative (AC) marker genes were validated by real-time quantitative polymerase chain reaction (PCR) analysis. Novel marker gene and protein expression was also assessed in human bone marrow–derived MSCs (BM-MSCs) and adipose tissue–derived MSCs (AD-MSCs) following differentiation in type I collagen gels.


Analysis identified 12 NP-positive and 36-negative (AC) marker genes that were differentially expressed ≥20-fold, and for a subset of them (NP-positive genes PAX1, FOXF1, HBB, CA12, and OVOS2; AC-positive genes GDF10, CYTL1, IBSP, and FBLN1), differential expression was confirmed by real-time quantitative PCR. Differentiated BM-MSCs and AD-MSCs demonstrated significant increases in the novel NP markers PAX1 and FOXF1. AD-MSCs lacked expression of the AC markers IBSP and FBLN1, whereas BM-MSCs lacked expression of the AC marker IBSP but expressed FBLN1.


This study is the first to use gene expression profiling to identify the human NP cell phenotype. Importantly, these markers can be used to determine the in vitro differentiation of MSCs to an NP-like, rather than an AC-like, phenotype. Interestingly, these results suggest that AD-MSCs may be a more appropriate cell type than BM-MSCs for use in engineering intervertebral disc tissue.

Low back pain is a major public health problem in the Western World, with over 80% of the population estimated to report low back pain during their lifetime (1). This has a serious impact on economies, with total costs related to this condition exceeding $100 billion per year in the US alone (2). One of the principal causes of low back pain is degeneration of the intervertebral disc (IVD), whereby adverse changes in the matrix results in dehydration of the nucleus pulposus (NP), loss of disc height, and loss of mechanical function, leading to traumatic damage and associated pain (3–5).

Current treatments are aimed at relieving symptoms of low back pain, rather than halting progression of the disease by inhibiting the underlying disease mechanisms or by restoring disc structure and function. Thus, there is an urgent need to provide more-effective treatments for IVD degeneration. Since the NP is the region of the IVD that is most affected during degeneration, the focus of many recent studies has been on the use of cell-based tissue engineering to regenerate the NP. Indeed, a number of recent advances in IVD and stem cell biology have demonstrated that both bone marrow–derived mesenchymal stem cells (BM-MSCs) and adipose tissue–derived mesenchymal stem cells (AD-MSCs) are capable of differentiating into chondrocyte-like cells of the NP in vitro (6–10) and in vivo in animal models (11–13). This offers the realistic hope of providing cellular therapies that will restore functional IVD tissue in degenerated discs.

In order to ensure a successful clinical outcome for any stem cell–based tissue engineering strategy, however, it is essential that implanted cells adopt the correct phenotype. Given the similarities in phenotype between NP cells and articular chondrocytes (ACs) (14), many studies have used a small panel of communally expressed markers, including SOX9, COL2A1, and ACAN to identify an NP cell (14–16). However, there are clear morphologic and physiologic differences between articular cartilage and NP tissues, suggesting that there are differences in their cellular phenotypes and that the identification of the exact phenotype for NP cells is essential for ensuring the correct cell type is obtained and that, ultimately, the appropriate extracellular matrix is produced.

There have, however, been surprisingly few attempts to characterize the differences between the molecular phenotypes of NP and AC cells, although a small number of recent studies have used microarray technology to compare the gene expression profiles of IVDs and ACs in rat, canine, and bovine tissues (17–20) or have used proteomic technology to compare proteins in bovine NP cells and ACs (21). Lee et al (18) identified KRT19 and GPC3 as being differentially expressed genes that also stained immunopositive in the NP of discs from young rats but were negative in the anulus fibrosus (AF) and articular cartilage. They also identified A2M, KRT18, and NCAM1 as being enhanced in NP cells as compared with ACs from canine tissues (20). However, when the same group of investigators examined the expression of their rat markers in canine tissues, they found that only 2 AC markers (COMP and MGP) demonstrated a similar pattern, highlighting the importance of identifying gene markers that are both cell-specific and species-specific when defining the NP phenotype.

For human tissue regeneration and engineering purposes, it is essential that genes identified in animals can be applied to human NP cells. A recent study compared rat and canine marker genes identified by microarray analysis to human IVD and cartilage tissues and concluded that only KRT19 has the potential to characterize human NP cells (22), although interestingly, the expression of this marker was lost by the age of 45, suggesting that it is more indicative of a young or nondegenerated NP phenotype. Our group of investigators has also recently identified novel marker genes that were differentially expressed in bovine IVD cells and ACs and examined their suitability for the identification of human NP cells (19). Data showed that the majority of these genes were differentially expressed to a lesser degree in humans, again illustrating species variation. It is therefore imperative that for definitive characterization of a human NP cell phenotype, microarray studies should be undertaken with human tissues to produce a comprehensive molecular signature, which will enable researchers to more accurately distinguish an NP cell from an AC. Thus, the aims of this study were to conduct microarray experiments using human ACs and NP cells to identify NP- and AC-specific markers, and to use these markers to identify the phenotype of both BM-MSCs and AD-MSCs differentiated in vitro in biomaterials for potential use in IVD tissue engineering.


Human tissues and sample preparation.

Human IVD tissues were obtained from tissue donors during postmortem examination, with informed consent from the donor's relatives and with the approval of the North West Research Ethics Committee. All postmortem samples were obtained from donors who died of natural causes (e.g., heart failure), had no history of back pain, and had no other relevant clinical history, such as bone metastases. Representative tissue samples were processed to paraffin wax embedding for grading according to a published 12-point histologic scale (14). The remaining disc tissue was macroscopically dissected.

Human articular cartilage samples were obtained during total knee arthroplasty in osteoarthritis (OA) patients, with the informed consent of the patients and with approval of the South Manchester Research Ethics Committee. Samples with a “normal” macroscopic appearance were harvested, and full-thickness sections (excluding subchondral bone) were processed to paraffin wax embedding for grading according to a published 11-point histologic scoring system adapted from Mankin et al (23), as previously described (24).

Only NP and cartilage samples that were histologically graded as “normal” were used. Each tissue sample was dissected into 2–3-mm3 fragments, placed on an orbital shaker at 37°C, and enzymatically digested for 1 hour in serum-free medium containing 0.5% Pronase (Calbiochem) and then for 2–3 hours in 0.5% type II collagenase (Invitrogen) and 0.1% hyaluronidase (Sigma). Cells were passed through a 40-μm filter, collected by centrifugation at 500g for 5 minutes, and then lysed in TRIzol reagent (Invitrogen).

RNA was extracted according to the manufacturer's instructions and was quantified using a NanoDrop ND-1000 spectrophotometer (NanoDrop Technologies). Quality was checked by means of an RNA 6000 NanoAssay on an Agilent 2100 Bioanalyzer.

Complementary DNA (cDNA) microarrays.

To minimize the effect of biologic variation on differential expression, hybridizations for each cell type from 3 separate individuals were performed in triplicate. Only high-quality RNA with an RNA integrity number of at least 7 taken from 3 histologically “normal” IVD samples (2 male and 1 female donor, ages 46–57 years [mean age 51 years]; histologic grade 1 or 2) and articular cartilage samples (1 male and 2 female donors, ages 55–60 years [mean age 58 years]; Mankin grade 1 or 2) were used for the microarrays.

For each hybridization, 100 ng of total RNA was used in the Affymetrix GeneChip Two-Cycle Target Labeling kit and in the Ambion MEGAscript T7 kit before hybridizing to the GeneChip human genome U133 Plus 2.0 array (Affymetrix) according to manufacturer's instructions. Technical quality control was performed with dChip software (25). Background correction, quantile normalization, and gene expression analysis were performed using the robust multiarray average (RMA) analysis in the Bioconductor software package (26). Differential expression analysis was performed using routine analytical methods (27). Lists of differentially expressed genes were controlled for false discovery rate (FDR) errors using the QValue method (28). Differentially expressed genes chosen for subsequent analysis met the following criteria: an FDR-corrected P value (called the q value) of ≤0.05, a minimum normalized expression level of >20-fold, and a minimum mean fluorescence signal intensity of >100 for either the NP or the AC replicate mean. The microarray data were submitted in Minimum Information About a Microarray Experiment (MIAME)–compliant format to the ArrayExpress database (online at; accession no. E-MEXP-2488).

Real-time quantitative polymerase chain reaction (PCR) analysis of articular cartilage and NP samples.

Real-time quantitative PCR using gene-specific primers (Table 1) was performed on cDNA derived from NP and articular cartilage samples using the SYBR Green method as previously described (19). Five “normal” IVD samples (3 male and 2 female donors, ages 45–60 years [mean age 52 years]; histologic grade 1–3) and 5 “normal” articular cartilage samples (2 male and 3 female donors, ages 50–60 years [mean age 56 years]; histologic grade 1–2) were used for these analyses. The 2math formula method was used to calculate the relative expression of each target gene, as described previously (15, 19, 29). Statistical analysis was performed with GraphPad InStat software using the Mann-Whitney U test. P values less than 0.05 were considered significant.

Table 1. Human oligonucleotide primers used for real-time quantitative polymerase chain reaction analysis
Gene nameGene symbolNCBI RefSeqForward primerReverse primer
Integrin-binding sialoprotein (bone sialoprotein 2)IBSPNM_004967.3CCAGAGGAAGCAATCACCAAAGCACAGGCCATTCCCAAA

BM-MSC and AD-MSC sample acquisition, isolation, and culture.

Human BM-MSCs were isolated from bone marrow obtained from a proximal femur sample removed during hip replacement surgery from a 66-year-old man. The patient's informed consent and approval of the South Manchester Research Ethics Committee were obtained. Mononucleated cells were isolated using a Histopaque 1077 (Sigma) density-gradient method, and cultured in α-minimum essential medium (α-MEM) containing 20% fetal calf serum (FCS) to allow adherence. Nonadherent cells were discarded after 7 days, and the adherent cells were cultured using standard procedures.

Human AD-MSCs were isolated from adipose tissue obtained at surgery from a 38-year-old woman. The patient's consent and approval of the South Manchester Research Ethics Committee were obtained. Adipose tissue was dissected, suspended in Hanks' balanced salt solution containing 20 units of heparin and antibiotics, and centrifuged for 5 minutes at 500g. The resulting cell pellet was resuspended in α-MEM. Mononucleated cells were isolated, and the cells were cultured as described above for the BM-MSCs.

The multipotentiality of the BM-MSCs and AD-MSCs was verified using standard methods to confirm differentiation along the osteogenic, adipogenic, and chondrogenic lineages, in accordance with the guidelines of the International Society for Cellular Therapy (30) (data not shown).

Cells were seeded into type I collagen gels (Arthro-Kinetics) at a final density of 4 × 106 cells/ml. Seeded gels were formed in triplicate for both cell samples in 24-well cell culture inserts at a volume of 200 μl per gel and were cultured for 14 days under standard conditions in high-glucose Dulbecco's modified Eagle's medium (DMEM) containing 1% FCS, 1× ITS-X (insulin–transferrin–selenium X; Life Technologies), 1.25 mg/ml of bovine serum albumin (AlbuMax-1; Life Technologies), 100 units/ml of penicillin, 100 μg/ml of streptomycin, 250 ng/ml of amphotericin B, 100 μM ascorbic acid 2-phosphate, 2 mM L-glutamine, 1 × 107M dexamethasone, and 10 ng/ml of transforming growth factor β1 (TGFβ1) (differentiating medium [15]). The medium was changed every 2–3 days.

Real-time quantitative PCR analysis of BM-MSCs and AD-MSCs.

Following 14 days in culture, gels were digested for 2 hours in a solution of serum-free high-glucose DMEM containing 0.2% type II collagenase. Cells were collected by centrifugation, washed in phosphate buffered saline (PBS), and RNA was extracted and reverse transcribed to cDNA as described above. Real-time quantitative PCR was conducted for the housekeeping gene GAPDH, the standard marker genes COL2A1 and ACAN, as well as a range of novel phenotype marker genes as described above.

Immunohistochemical analysis.

Following 14 days in culture in differentiating medium, gels containing either BM-MSCs or AD-MSCs were washed in PBS, fixed in 4% paraformaldehyde, processed to paraffin wax embedding, and 5-μm sections were cut. Immunohistochemical staining was conducted for Forkhead box F1 (FoxF1), paired box protein 1 (Pax-1), carbonic anhydrase XII (CAXII), integrin-binding sialoprotein (IBSP), and fibulin 1, based on previously published protocols (31, 32). Heat-mediated antigen retrieval was performed for 20 minutes using either 0.01M citrate buffer in a pressure cooker or 10 mM Tris, 1 mM EDTA, pH 9.0, in a steamer. Sections were incubated overnight at 4°C with primary antibodies for FoxF1 (rabbit polyclonal, 1:150 dilution; Abcam), Pax-1 (rabbit polyclonal, 1:50 dilution; Novus Biologicals), CAXII (rabbit polyclonal, 1:50 dilution; Sigma-Aldrich), IBSP (rabbit polyclonal, 1:50 dilution; Abcam), and fibulin 1 (mouse monoclonal, 1:25 dilution; Santa Cruz Biotechnology). Biotinylated goat anti-rabbit and goat anti-mouse secondary antibodies were used, and staining was disclosed using a diaminobenzidine chromogen. Negative controls used the appropriate IgG (Dako) in place of the primary antibody at equal protein concentrations. Stained sections were viewed under light microscopy, and images were acquired using an InfinityX camera with DeltaPix software.


Microarray analysis of previously identified “marker” genes.

Typical chondrocyte markers were used to evaluate the ability of the microarrays to detect gene expression in the 2 cell types (Table 2). Expected patterns of expression were seen for COL2A1, SOX9, and ACAN, which were detected at high levels in both NP and AC samples. For comparative purposes, a number of genes that were identified in previous microarray experiments as being phenotype NP markers of murine (18), canine (20), and bovine (19) cells were assessed to determine their expression in human arrays.

Table 2. Expression levels of previously described potential NP cell marker genes*
Gene descriptionGene symbolAC MFINP MFIFold difference (NP/AC)
  • *

    Values are the human microarray data for previously identified markers. For each cell type (nucleus pulposus [NP] cells and articular chondrocytes [ACs]), the mean values for the mean fluorescence intensity (MFI) of expression are shown, along with the calculated fold difference (positive and negative) between the NP cells and the ACs.

Chondrogenic markers
 Type I collagen α1COL1A1201.28218.041.08
 Type II collagen α1COL2A122,740.739,398.03–2.42
 Sex-determining region Y box 9SOX92,618.063,260.231.25
Previously identified murine markers
 Keratin 19KRT1918.4442.292.29
 Glypican 3GPC334.1540.631.19
 Annexin A3ANXA315.7914.66–1.08
 Cartilage oligomeric matrix proteinCOMP26,678.5318,915.58–1.41
Previously identified canine markers
 Matrix Gla proteinMGP17,602.042,356.78–7.47
 Annexin A4ANXA41,482.17953.32–1.55
 Desmocollin 2DSC2306.11540.021.76
 Keratin 18KRT18315.171,637.125.19
 Neural cell adhesion molecule 1NCAM1165.85234.021.41
Previously identified bovine markers
 Keratin 8KRT8140.6995.71–1.47
 Cadherin 2, type 1, N-cadherin (neuronal)CDH218.6233.161.78
 Synaptosomal-associated protein 25 kdSNAP2555.4770.471.27
 Sclerostin domain containing 1SOSTDC121.2624.181.14
 Integrin-binding sialoprotein (bone sialoprotein 2)IBSP7,646.1495.98–79.66
 Fibulin 1FBLN1166.1297.65–1.70
 Tumor necrosis factor α–induced protein 6TNFAIP6954.442,112.002.21
 Forkhead box F1FOXF139.011,837.7847.11
 Forkhead box F2FOXF292.09363.543.95
 Aquaporin 1AQP185.5078.94–1.08

The murine NP cell markers KRT19, GPC3, ANXA3, and PTN were detected at low levels in our human NP cell arrays; VIM and COMP were detected in the arrays at high levels in both NP and AC cells, suggesting their unsuitability as marker genes in human NP cells.

Of the canine markers identified, only KRT18 showed significant differential expression (5-fold) in NP cells as compared with ACs. MGP was differentially expressed in ACs (∼7-fold) as compared with NP cells, but was also detected at significant levels in NP cells; therefore, MGP is unlikely to distinguish the two cell types. None of the other reported canine genes (A2M, ANXA4, DSC2, and NCAM1) showed any significant fold differences in expression between the two cell types.

Similarly, a significant number of the markers identified in the bovine arrays showed relatively low differential expression in human NP cells versus ACs, such as KRT8, CDH2, SNAP25, SOSTDC1, TNMD, and AQP1. Two bovine genes showed particularly high levels of differential expression in the human arrays. The bovine IVD marker FOXF1 showed a 47-fold increased expression in NP cells as compared with ACs, and the bovine AC marker IBSP showed an ∼80-fold increase in expression in ACs as compared with NP cells (Table 2).

Microarray identification of novel cell-specific “marker” genes.

To identify additional novel genes that could be used to distinguish human NP cells from ACs, the average expression levels in NP cells were compared with the average expression levels in ACs (NP mean/AC mean). Differentially expressed genes met the following criteria: an FDR-corrected P value (called the q value) of ≤0.05, a minimum normalized expression level of >20-fold, and a minimum mean fluorescence signal intensity of >100 for either the NP cell or the AC replicate mean. Using these criteria, a total of 64 probe sets (48 genes) were identified, of which 12 genes showed high levels of differential expression in NP cells and 36 genes showed high levels of differential expression in ACs. Table 3 highlights the 24 most differentially expressed genes for both NP cells (12 genes) and ACs (12 genes). Genes that were analyzed by real-time quantitative PCR and genes that were identified in previously published microarray studies (17–20) are indicated in Table 3.

Table 3. Comparison of NP versus AC microarray data*
Gene descriptionGene symbolNP MFIAC MFIFold difference (NP/AC)
  • *

    For each cell type (nucleus pulposus [NP] cells and articular chondrocytes [ACs]), the mean values for the mean fluorescence intensity (MFI) of expression are shown, along with the calculated fold difference (positive and negative) between the NP cells and the ACs.

  • Analyzed by real-time quantitative reverse transcription–polymerase chain reaction.

  • Identified in previously published microarray studies (17–20).

Ovostatin 2OVOS22,426.5827.7087.60
Class II major histocompatibility complex DQβ1HLA–DQB12,803.0446.1960.68
Hemoglobin α1-chainHBA15,284.2292.8256.93
Hemoglobin β-chainHBB2,001.6235.9155.74
Forkhead box F1FOXF11,837.7839.0147.11
Carbonic anhydrase XIICA122,726.9071.3338.23
ATP-binding cassette subfamily G member 1ABCG1739.6820.3036.44
Family with sequence similarity 110 member CFAM110C1711.4548.9834.94
Paired box 1PAX11,638.1751.7131.68
Zinc-finger protein 165ZNF1651,136.7836.9430.77
Double cortin–like kinase 1DCLK1655.1823.2328.21
Annexin A2ANXA2380.7918.4820.61
Matrilin 3MATN313.00481.92–37.08
Tenascin CTNC159.406,436.00–40.38
Cytokine-like 1CYTL1345.5914,174.13–41.01
Spondin 1, extracellular matrix proteinSPON142.781,784.01–41.70
Kallmann syndrome 1 sequenceKAL157.962,612.33–45.07
G protein–coupled receptor 88GPR8814.37677.60–47.16
Complement factor HCFH3,31.5716,879.22–50.91
Collectin subfamily member 12COLEC1247.872,956.07–61.75
Integrin-binding sialoprotein (bone sialoprotein 2)IBSP95.987,646.14–79.66
Epidermal growth factor–containing fibulin-like extracellular matrix protein 1EFEMP171.556,149.25–85.94
Vascular cell adhesion molecule 1VCAM152.665,065.74–96.20
Growth differentiation factor 10GDF1043.324,262.62–98.39

The serine protease inhibitor OVOS2 showed the highest level of differential expression in NP cells and was selected for analysis by real-time quantitative PCR. Other highly expressed genes, including the oxygen-transport metalloprotein gene HBB, the carbon dioxide–transport metalloenzyme gene CA12, the transcription factor gene PAX1, and the previously identified NP marker FOXF1, were also further examined by real-time quantitative PCR analysis. The TGFβ superfamily member gene GDF10, which was the highest differentially expressed gene in ACs, was selected for real-time quantitative PCR analysis, along with the highly expressed CYTL1 gene, and these were analyzed with the previously identified IBSP and FBLN1 genes (19).

Real-time quantitative PCR validation of cell-specific marker genes in human samples.

Real-time quantitative PCR demonstrated high levels of both COL2A1 and ACAN in NP cells and ACs, although the gene expression levels were lower in NP cells than in ACs (P < 0.0001 for both genes) (data not shown).

The novel NP markers all showed significantly higher expression in NP cells as compared with ACs (Figure 1A). The most significant of these was PAX1, which showed >1,000-fold higher levels of expression in NP cells than in ACs (P < 0.0001). The previously identified transcription factor gene FOXF1 was also found to have significantly higher levels of expression in NP cells than in ACs (>50-fold; P < 0.0001). In addition, CA12 and HBB showed significantly higher levels of expression in NP cells than in ACs (>50-fold [P < 0.0001] and >100-fold [P < 0.0001], respectively). OVOS2 gene expression was significantly lower in ACs than in NP cells (P < 0.0001), although the overall fold difference between NP cells and ACs (∼10-fold) was considerably lower than that detected by the array (87-fold).

Figure 1.

Real-time quantitative polymerase chain reaction analysis of A, the novel positive nucleus pulposus (NP) cell marker genes PAX1, FOXF1, CA12, HBB, and OVOS2, and B, the novel NP-negative cell marker genes GDF10, IBSP, CYTL1, and FBLN1 in human NP cells and articular chondrocytes (ACs). The relative gene expression for all marker genes was normalized to the housekeeping gene GAPDH and plotted on a log scale. Values are the mean ± SEM. ∗ = P < 0.05 versus ACs.

The previously identified NP-negative markers IBSP and FBLN1 showed low expression in NP cells, with significantly higher expression in ACs (>100-fold [P < 0.0001] and >50-fold [P < 0.0001], respectively) (Figure 1B). Expression of GDF10 and CYTL1 genes was significantly higher (>100 fold; P < 0.0001) in ACs than in NP cells, although the expression levels in NP cells were higher than that of either IBSP or FBLN1.

Differentiation of BM-MSCs and AD-MSCs to an NP-like phenotype.

Following culture in type I collagen gels, both AD-MSCs and BM-MSCs demonstrated significant increases in the classic marker genes COL2A1 and ACAN. Expression of the novel NP marker genes PAX1 and FOXF1 was also significantly increased in both AD-MSCs and BM-MSCs as compared with monolayer controls. There were also significant increases in the CA12 gene in both cell types, although basal levels in both AD-MSCs and BM-MSCs were high during monolayer culture as compared with the other genes investigated. Interestingly, AD-MSCs lacked expression of the negative marker gene IBSP, both in monolayer culture and following culture in type I collagen gels, whereas although BM-MSCs expressed this gene in monolayer culture, there was no expression following culture in type I collagen gels. FBLN1 was expressed in both monolayer AD-MSCs and BM-MSCs, with significantly higher expression seen in monolayer BM-MSCs. However, following type I collagen gel culture, although FBLN1 gene expression significantly increased in BM-MSCs as compared with expression in monolayer culture, no expression of this gene was detected in AD-MSCs cultured in type I collagen gels for 14 days in differentiating medium (Figure 2).

Figure 2.

Real-time quantitative polymerase chain reaction analysis of the novel nucleus pulposus (NP) cell and articular chondrocyte (AC) markers COL2A1, ACAN, PAX1, FOXF1, CA12, IBSP, and FBLN1 in monolayer and differentiated human adipose tissue–derived mesenchymal stem cells (AD-MSCs) and bone marrow–derived mesenchymal stem cells (BM-MSCs). Cells were obtained from monolayer cultures or were obtained on day 14 from type I collagen gel cultures in differentiating medium (DM) (see Materials and Methods for details). Relative gene expression was normalized to the housekeeping gene GAPDH and plotted on a log scale. Values are the mean ± SEM. ∗ = P < 0.05.

When protein expression in AD-MSCs and BM-MSCs was analyzed following culture in type I collagen gels, both cell types demonstrated similar levels of immunopositivity for type II collagen α1, aggrecan, and CAXII (data not shown). However, AD-MSCs demonstrated strong cellular staining for Pax-1 and FoxF1 as compared with staining of BM-MSCs under the same conditions (Figure 3). AD-MSCs also demonstrated virtually no cellular staining for either IBSP or fibulin 1, whereas BM-MSCs demonstrated immunopositivity for both negative NP markers. All IgG controls were negative.

Figure 3.

Immunohistochemical staining for novel nucleus pulposus (NP) cell and articular chondrocyte (AC) markers in human adipose tissue–derived mesenchymal stem cells (AD-MSCs) and bone marrow–derived mesenchymal stem cells (BM-MSCs) cultured in type I collagen gels in differentiating medium for 14 days. Photomicrographs represent typical staining patterns for the NP markers paired box protein 1 (Pax-1) in AD-MSCs (A) and BM-MSCs (B) and Forkhead box F1 (FoxF1) in AD-MSCs (E) and BM-MSCs (F), as well as for the AC markers integrin-binding sialoprotein (IBSP) in AD-MSCs (I) and BM-MSCs (J) and fibulin 1 in AD-MSCs (M) and BM-MSCs (N) following culture in type I collagen gels. IgG controls were routinely negative in both AD-MSCs (C, G, K, and O) and BM-MSCs (D, H, L, and P). Bars = 25 μm.


NP cells of the IVD share a common phenotype with ACs, as demonstrated by the expression of the key chondrocyte genes COL2A1, ACAN, and SOX9 in both cell types (14). However, there are clear morphologic and physiologic differences between articular cartilage and NP tissues that suggest differences in the 2 cell phenotypes. In addition to offering a greater understanding of cell biology, the identification of the exact NP cell phenotype is essential for tissue engineering/regenerative therapies for IVD degeneration.

The exact phenotype of the human NP cell is largely unknown, although several recent studies have attempted to elucidate differences in the transcription profiles between ACs and NP cells in rats, cows, and dogs (18–20). Here, we have used Affymetrix MicroArrays and real-time quantitative PCR to identify genes that can be used to distinguish human NP cells from ACs. Furthermore, we have demonstrated that these marker genes can be used to identify the in vitro differentiation of BM-MSCs and AD-MSCs to an NP-like phenotype in biomaterials suitable for tissue engineering/regeneration of the human IVD.

This study is the first to use microarray technology to examine human NP cells and ACs, and a number of differentially expressed novel genes in human NP cells as compared with ACs were identified. Importantly, several differentially expressed genes identified from microarray studies in other animals (e.g., GPC3 and KRT19 from murine, A2M and ANXA4 from canine, and KRT18 and CDH2 from bovine studies) did not translate directly to this human study, with limited differential expression detected between the 2 cell types. This may be due to differences in disc age, size, native cell composition, and the physicochemical environment in these animal species was compared with the human lumbar discs. Furthermore, although macroscopically and histologically normal tissues were used for this study to avoid phenotype changes caused by degeneration, the AC cells were nevertheless derived from “normal” regions of OA joints, and thus, their gene profile could be different from that of non-OA cartilage. Interestingly however, the previously identified bovine NP marker FOXF1 and the AC marker/NP-negative marker IBSP (19) were also highly differentially expressed in the human arrays, and these findings were confirmed by real-time quantitative PCR in human NP cells and ACs. Such data demonstrate the existence of positive (NP) and negative (AC) markers between species and suggest that these 2 genes are potential marker genes for differentiating NP cells from ACs in both bovine and human tissues.

In addition, several novel human markers were identified in the microarray analysis, including the novel NP markers PAX1, CA12, and HBB and were validated by real-time quantitative PCR, with PAX1 having a particularly high differential expression between the 2 tissue types (>1,000 fold; P < 0.0001). The array analyses also identified novel AC markers GDF10 and CYTL1, which were confirmed by real-time quantitative PCR as being expressed >100-fold higher in ACs than in NP cells (P < 0.0001), and thus may serve (together with IBSP and FBLN1 identified from our bovine studies [19]), as negative markers for the NP phenotype.

Although previously identified in the IVDs of young adult mice (33) and in the human fetal vertebral column (34), this is the first time that PAX1 expression has been demonstrated in the adult human IVD. PAX1 encodes a transcription factor that regulates pattern formation during embryogenesis in vertebrates (35). Its expression is regulated by SHH, which has been shown to be expressed in the embryonic sclerotome and in the postnatal mouse NP (36, 37); however, its role in the adult human NP remains to be elucidated. Interestingly, FOXF1 expression has also been shown to be activated by SHH, and while the role of FOXF1 in cell growth, proliferation, differentiation, and longevity (38–40) has been established, its role in the adult human IVD has yet to be understood.

The high levels of expression of hemoglobin chains (HBB and HBA1) and CA12 in NP cells suggests that the harsh, hypoxic environment of the IVD significantly influences gene expression of resident cells. The oxygen-transport metalloprotein hemoglobin, which is found in red blood cells, has been recently detected in a number of nonerythroid cells, including dopaminergic neurons, macrophages, alveolar cells, and mesangial cells of the kidney (41–43). It is thought to act by storing oxygen, and it thereby provides a homeostatic mechanism under hypoxic conditions for metabolically active cells. Similarly, the zinc-containing metalloenzyme CAXII, which catalyzes the reversible hydration of carbon dioxide and is responsible for the maintenance of body ion and pH homeostasis (42), is regulated by hypoxia-inducible factor 1α (HIF-1α) under hypoxic conditions (44), and other CA family members have been identified in the fetal human IVD (45). It is therefore likely that these genes are highly expressed in NP cells in order to maintain homeostasis in the avascular, hypoxic, nutrient-deprived, and acidic environment of the adult human IVD.

There is increasing interest in the potential use of adult stem cells in tissue engineering strategies for regenerating the degenerated human IVD. Both BM-MSCs and AD-MSCs are known to be capable of chondrogenic differentiation (46, 47), and recent studies have demonstrated their potential for use in IVD tissue regeneration (11–13). Such studies have routinely used chondrocyte marker genes such as SOX9, COL2A1, and ACAN to identify the differentiation of BM-MSCs or AD-MSCs to NP-like cells. However, evidence suggests that implantation of chondrocyte-like cells, which do not have the correct NP cell phenotype, will result in the formation of a tissue that does not mimic the highly hydrated NP, and therefore lacks the morphologic or biomechanical properties, to ensure successful regeneration of a functional IVD (48). To overcome this problem, we used the novel marker genes identified in this study in a small proof-of-principle study to assess differentiation to an appropriate NP phenotype.

While both BM-MSCs and AD-MSCs have similar cell surface marker profiles, during monolayer culture, differences in the gene expression profiles between the 2 cell populations suggested that there was a difference in the initial phenotype that may affect their subsequent differentiation to NP cells.

Following differentiation in type I collagen gels, both BM-MSCs and AD-MSCs demonstrated significant up-regulation in expression of the conventional NP marker genes. Importantly, both cell types also demonstrated large and significant increases in expression of the novel NP marker genes PAX1 and FOXF1 as well as small but significant increases in CA12, while immunohistochemical staining suggested more marked and widespread expression of both Pax-1 and FoxF1 in AD-MSCs than in BM-MSCs. Significantly, differentiated AD-MSCs did not express the negative NP markers IBSP or FBLN1 at either the gene or the protein level, whereas BM-MSCs expressed FBLN1 at both the gene and the protein levels and expressed the IBSP protein. Expression of IBSP at the protein level may be due to the high levels of IBSP RNA observed in monolayer culture, which may not be immediately down-regulated under differentiating conditions.

These preliminary results, albeit on a small number of samples, suggest that the novel marker genes identified here may be used to characterize NP-specific differentiation of adult stem cells. Interestingly, while both BM-MSCs and AD-MSCs are capable of chondrogenic differentiation, AD-MSCs differentiate to a more NP-like phenotype than do BM-MSCs and may therefore be more suited to IVD tissue engineering strategies.


All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Prof. Hoyland had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Study conception and design. Minogue, Richardson, Zeef, Freemont, Hoyland.

Acquisition of data. Minogue, Richardson, Zeef.

Analysis and interpretation of data. Minogue, Richardson, Zeef, Hoyland.


We would like to acknowledge Mr. Adam Dickinson, who assisted with the stem cell cultures, and Miss Francesca Ludwinski, Miss Sonal Patel, and Mrs. Pauline Baird, who helped with antibody optimizations. Array hybridizations and initial analyses were performed by staff at the Microarray and Bioinformatics Core Facilities, Faculty of Life Sciences, University of Manchester.