1. Top of page
  2. Abstract
  7. Acknowledgements


Back pain is a major cause of disability, affecting millions of people worldwide. One cause of axial back pain is degeneration of the nucleus pulposus (NP) of the intervertebral disc. This study was undertaken to investigate associations of NP cells with cell surface–specific proteins that differ from proteins in closely related cell types, i.e., intervertebral disc anulus fibrosus (AF) cells and articular cartilage (AC) chondrocytes, in order to identify potential surface molecules for directed delivery of therapeutic agents.


We conducted a complementary DNA microarray analysis of 16 human samples from 6 donors, followed by gene list reduction using a systematic approach. Genes that were more highly expressed in NP than AC cells, contained transmembrane domains, and appeared attractive for targeting were assessed by quantitative reverse transcription–polymerase chain reaction (RT-PCR). As a viable candidate, carbonic anhydrase XII (CAXII) was analyzed at the protein level by immunohistochemistry and functional study.


Microarray results demonstrated a clear divide between the AC and AF and between the AC and NP samples. However, the transcriptomic profile of AF and NP samples displayed a greater intersubject similarity. Of the 552 genes with up-regulated expression in NP cells, 90 contained transmembrane domains, and 28 were quantified by RT-PCR. Most intense CAXII labeling was observed in the NP of discs from young subjects and in degenerative tissue.


CAXII may be considered for detection or targeting of degenerating disc cells. Furthermore, CAXII may be involved in pH regulation of NP cells. Its potential for directed delivery of regenerative factors and its functional role in NP cell homeostasis warrant further investigation.

Back pain is a leading cause of disability, and degeneration of the intervertebral disc (IVD) is believed to be a major cause of neck and lower back pain. The IVD is composed of 2 distinct but interdependent tissues: a gelatinous center, known as the nucleus pulposus (NP), and several surrounding coaxial lamellae that form the inner and outer anulus fibrosus (AF). This unique structural feature allows the IVD to constrain motion at high loads and provide flexibility at low loads. Factors such as abnormal mechanical stresses, biochemical imbalances, nutritional factors, and genetic variations are all reported to play a role in degenerative disc disease (1). As the natural aging process continues, the gelatinous NP region of the disc is replaced by a more solid, less flexible, and ultimately fibrous disc. This degeneration has been suggested to cause vascularization and the ingrowth of nerves and Schwann cells (2, 3), which contribute to back pain that in some cases is chronic.

Currently, there is no definitive cure for this cellular transition. Depending on the severity of a patient's symptoms, therapies can include one or a combination of measures, i.e., physiotherapy, pain and antiinflammatory medications, and finally, surgeries such as spinal fusion, dynamic stabilization, or disc arthroplasty. Although the established surgical treatments generally yield good results, especially in the short term, they all have disadvantages. In particular, these treatments can only relieve symptoms; their curative effects are minimal and they do not restore the functional, original disc tissue. Moreover, they are risky and may lead to serious complications (4). Thus, availability of a treatment that could prevent mechanical failure of the segment and restore a functional matrix of the disc using minimally invasive techniques would be desirable.

There are relatively recent and novel approaches to IVD regenerative therapy that avoid the above-described disadvantages, including use of injectable biomaterials, growth factors, and gene therapy (5, 6). An attractive strategy is the use of therapeutic cargo–carrying nanoparticles. The use of this strategy for drug and/or gene delivery would limit the need for surgical procedures because the treatment would be injectable and could be specifically directed toward cells residing in the NP, in order to induce endogenous regenerative activity. Therefore, a primary goal is to identify a suitable and specific target in NP cells with which to tag active compounds in the therapeutic cargo–carrying nanoparticles, for local injection. Yet, current knowledge about the specific expression of surface proteins in human NP cells is limited. The expression of cell surface receptors such as integrin subunits was identified in investigations of porcine and human IVD tissue (7, 8) and in human herniated discs (9); however, these receptors were not located on NP tissue exclusively. CD24, a cell adhesion molecule, was identified as a specific cell surface marker for NP cells in the rat (10), and CD56 (neural cell adhesion molecule 1) was found to be expressed in canine NP cells (11), but gene expression analysis of human disc cells revealed that CD24 expression was not specific for NP cells and that CD56 was expressed only at low levels (12).

In order to gain more comprehensive knowledge about the expression of surface proteins in human NP cells, we investigated gene expression profiles of human NP cells by large-scale microarray analysis, with particular emphasis on cell surface markers. Expression profiles were compared to those of AF cells and articular cartilage (AC) chondrocytes, which are known to exhibit phenotypes similar to NP cells. The information gained from this study may be used not only to develop new curative approaches involving targeted delivery of therapeutic agents to the NP cell via specific surface molecules, but also to acquire a better understanding of the disc cell phenotype in general.


  1. Top of page
  2. Abstract
  7. Acknowledgements

Isolation of NP, AF, and AC cells.

The study was approved by the medical ethics committee of the University Medical Center Utrecht and the scientific committee of the Department of Pathology of the University Medical Center Utrecht. Specimens obtained at autopsy from 6 subjects with no known history of IVD disease were studied. Samples were obtained within an average of 16 hours after death (range 6.25–23.0 hours). Between death and tissue collection the subjects' bodies were kept in the mortuary at 4°C. The age of the subjects ranged from 25 to 81 years (mean 54.5 years) (Table 1).

Table 1. Data on autopsy subjects from whom samples were obtained, and sample characteristics*
Subject/age/sex, tissue typeThompson gradeRNA integrity number
  • *

    Articular cartilage (AC) was not available from subject 4. NP = nucleus pulposus; AF = anulus fibrosus; NA = not applicable.

  • Sample had a 3′:5′ ratio of >3 on one of the housekeeping genes, and was therefore excluded from further analysis.

  • Sample was included for reverse transcription–polymerase chain reaction studies, but not for microarray analysis.


The degree of disc degeneration was assessed according to the Thompson scoring system (13). IVD tissue was harvested from segments between L1 and L5 and was separated into NP and AF tissue. To exclude any contamination by AF tissue, only the central gel-like or amorphous NP tissue, which could easily be separated from the lamellar structure of the AF tissue, was harvested to be assigned to the group of NP tissue samples; the transition zone, including part of the inner AF, was entirely excluded from analysis. This is of particular importance with regard to aged discs, in which it can be difficult to clearly distinguish NP and AF tissue. In these cases NP tissue was sampled by conservatively excising tissue from the center of the disc. AC was harvested from the patellae of the same subjects. Chondrocytes were extracted from full-thickness cartilage, which implies the presence of mixed populations of superficial, middle, and deep zone cells.

Tissue was cut into small pieces and cells were enzymatically isolated by sequential Pronase (Roche) and type II collagenase (Worthington Biochemical) digestion, with DNase II (Sigma) added to prevent cell clumping. AC and AF were treated with 0.2% Pronase/0.004% DNase for 1 hour, and then with 200 units/ml collagenase/0.004% DNase overnight. NP was treated with 0.2% Pronase/0.004% DNase for 1 hour, and then with 100 units/ml collagenase/0.004% DNase for 8 hours, with stirring at 37°C in a humidified atmosphere. After enzymatic isolation, cell suspensions were filtered through a 70-μm cell strainer, washed twice with Dulbecco's modified Eagle's medium (DMEM), and lysed in TRI Reagent (Molecular Research Center). Samples were stored at −80°C until RNA isolation.

RNA extraction.

RNA was isolated using a modified TriSpin method (14, 15). Briefly, bromo-chloro-propane (Sigma) was added to the lysate, phases were separated, and ethanol (Merck) added to the aqueous phase. Total RNA was extracted using the SV Total RNA Isolation System (Promega), which includes on-column DNase digestion, and eluted in 100 μl of RNase-free water.

Quality control.

NanoDrop (Thermo Scientific) and Bioanalyzer (Agilent) instruments were used to assess the quantity and the quality of the RNA samples, respectively. RNA quality was assessed for all 17 samples (AC was not available from 1 subject), using the RNA integrity number (Table 1). All samples were considered of high quality, with an RNA integrity number of >7. GeneChip Test3 Arrays (Affymetrix) were used to determine the sample quality in relation to the hybridization of housekeeping genes and the relative quantity of 3′ to 5′ probes. AC from donor 6 had a 3′:5′ ratio on one of the housekeeping genes (AFFX-has) of 15.77, and was therefore removed from further analysis (as recommended by Affymetrix if the ratio is >3). Sixteen samples in total were of suitable quality for microarray analysis.

Microarray data analysis.

Human GeneChip U133 Plus 2.0 Affymetrix arrays were used for all 16 samples. The statistical environment R ( and Bioconductor ( (16, 17) were used for data analysis. Samples were background-corrected, log2-transformed, and quantile-normalized using Robust Multi-array Average within the Bioconductor package Affy (18). Correspondence analysis and hierarchical clustering using the Euclidean distance measure were also performed, with the Made4 and Stats packages, respectively (19). Differential expression analysis was carried out with the Linear Models for Microarray Data package (20, 21), using the false discovery rate to control for multiple testing (22).

Short-listing of candidate genes.

Genes that exhibited higher expression in NP compared to AC cells were ranked to identify potential cell-specific surface marker proteins. Initially, BioMart was used to identify gene products with at least 1 transmembrane domain, after which ExPASy (, InterPro (, and primary literature databases were used to eliminate products associated with nuclear or intracellular organelle membranes. Candidate genes were further eliminated if their products contained only extracellular domains of <30 amino acid residues, if they were known to be glycosylated, if they contained a high number of disulfide linkages, or if they were active as multimers. The latter 3 features represent sources of potential bottlenecks in Escherichia coli expression of recombinant proteins, which would be necessary to validate a protein as a molecular target.

Real-time reverse transcription–polymerase chain reaction (RT-PCR).

TaqMan reverse transcription reagents (Applied Biosystems) were used for complementary DNA synthesis. PCR was performed with an SDS 7500 real-time PCR instrument using TaqMan Gene Expression Master Mix (all from Applied Biosystems) and standard thermal conditions (10 minutes at 95°C for polymerase activation, followed by 40 cycles of 95°C for 15 seconds and 60°C for 60 seconds). Primer-probe systems, purchased as Gene Expression Array Plates, were from Applied Biosystems. Expression of target genes was normalized to 4 endogenous controls: 18S ribosomal RNA, GAPDH, hypoxanthine phosphoribosyltransferase 1, and glucuronidase β. Gene expression levels were calculated according to the ΔCt method and presented as log2-transformed relative messenger RNA (mRNA) amounts (23). Differences between NP and AF and between NP and AC were assessed in paired samples by 2-tailed t-test, with P values less than 0.05 considered significant.


Human IVD tissue was obtained as part of a standard postmortem procedure in which a section of the lumbar and thoracic spine is removed for diagnostic purposes. IVD between the fourth and fifth lumbar vertebra (spinal motion segment L4–L5), including the adjacent end plates, was obtained from all subjects. The grade of degeneration was scored by 3 individual observers, using the classification system described by Thompson et al (13).

IVD tissue from 21 human subjects ages 3–72 years (mean ± SD 36 ± 21 years, median 35 years) was evaluated. Sagittal slices of the motion segments were fixed in formalin, decalcified for 6 hours with Kristensen's solution (50% formic acid and 68 gm/liter sodium formiate) in a microwave at 150W and 50°C, dehydrated in graded ethanol series, and embedded in paraffin. Sections were deparaffinized, treated with 3% hydrogen peroxide in methanol for 30 minutes, and then treated with heated (95°C) citrate buffer (10 mM sodium citrate, 0.05% Tween 20 [pH 6.0]) for 15 minutes, for antigen retrieval. They were then blocked for 1 hour with 5% normal horse serum, and were incubated with 0.04 mg/liter rabbit anti–carbonic anhydrase XII (anti-CAXII) precursor antibody (catalog no. HPA008773; Prestige Antibodies, Sigma) at a 1:60 dilution overnight at 4°C. Negative control sections were incubated without primary antibody. Biotinylated secondary anti-rabbit antibody diluted 1:200 (Vectastain ABC kit Elite, catalog no. PK-6102; Vector) was applied, followed by addition of ABC complex and chromogen development using diaminobenzidine (DAB Kit, catalog no. SK-4100; Vector). Sections were counterstained with Mayer's hematoxylin. To evaluate the composition of the extracellular matrix (ECM) in the immunostained IVD sections, separate sections were also stained with Alcian blue/picrosirius red and observed both in brightfield and polarized light. The Alcian blue/picrosirius method is known to produce distinctive staining of collagen (red) and proteoglycans, including hyaluronan (blue) (24).

Effect of CAXII inhibition.

NP cells were isolated, as described above, from the tails of 3–4-month-old cows. Cells were seeded in 24-well plates at a density of 50,000 per well and cultured in low-glucose DMEM containing 10% fetal calf serum at an oxygen concentration of either 5% or 21%. After 1 day of culture the medium was changed and 4-aminobenzene sulfonamide (Sigma), a CA inhibitor with specificity for CAXII, was added at concentrations of 1, 10, 100, 1,000, and 10,000 nM (25). The lactate concentration in the culture medium was measured 24 hours after addition of 4-aminobenzene sulfonamide (26) and was normalized to the concentration in the control medium without CA inhibitor treatment. Cytotoxicity of 4-aminobenzene sulfonamide (at concentrations of up to 10,000 nM) toward NP cells was determined using WST-1 Cell Proliferation Reagent according to the protocol of the manufacturer (Roche). A general linear model with multivariate tests and Bonferroni adjustment was used for statistical analysis of the effect of CA inhibitor treatment.


  1. Top of page
  2. Abstract
  7. Acknowledgements

Unsupervised clustering.

Hierarchical clustering and correspondence were used to identify gene expression profiles that differed between the tissue types. The AC samples were clearly separated from the AF and NP samples using both of these unsupervised clustering techniques. However, there was no clear separation between AF and NP samples. Results of the hierarchical clustering analysis are shown in Figure 1. When the data were sorted and viewed by subject, a trend toward an intrasubject gene profile similarity between the AF and NP samples was observed. Correspondence analysis results are available from the corresponding author upon request.

thumbnail image

Figure 1. Hierarchical clustering results, by sample type (articular cartilage [AC], anulus fibrosus [AF], and nucleus pulposus [NP]) (A) and by subject number (B). Each sample type and subject is represented by a different color in the bars below the diagrams. Color figure can be viewed in the online issue, which is available at

Download figure to PowerPoint

Differential expression.

With the use of a strict P value cutoff of P < 0.01, there were no genes that were differentially expressed between NP and AF. Increasing the cutoff to P < 0.05 resulted in the identification of 6,611 genes that were differentially expressed between NP and AF, NP and AC, or AF and AC; however only 3 of these were differentially expressed between NP and AF. These 3 probe sets were annotated to SPARCL1, NFATC4, and one that only had a GenBank accession number (AW003173). In summary, the lack of global gene expression differences between NP and AF, as displayed by the unsupervised clustering techniques, was confirmed by the lack of highly significant P values between these groups.

Cell surface markers.

In order to focus the experiment toward cell surface markers within the NP samples, transmembrane domain–containing genes were identified from genes that were found to be up-regulated in the NP samples using BioMart (27). Starting with a cutoff of P < 0.01, 552 genes were up-regulated in NP samples in comparison to AC samples, 90 of which contained transmembrane domains (Table 2).

Table 2. Top 90 transmembrane domain–containing genes that were found to be up-regulated in nucleus pulposus*
SymbolIDLog fold changeAdjusted PAverage log2 expressionDescription
  • *

    Genes are listed in order of P value.

  • Adjusted for multiple testing.

  • Normalized fluorescence intensity units.

  • §

    Analyzed by reverse transcription–polymerase chain reaction.

ODZ3§219523_s_at−2.170.00000156.77Odz, odd Oz/10-m homolog 3 (Drosophila)
DNER§226281_at−3.50.000001510.23Delta/notch-like epidermal growth factor repeat–containing
FLRT2§204358_s_at−2.430.00000876.69Fibronectin leucine-rich transmembrane protein 2
PERP217744_s_at−1.590.00001649.1p53 apoptosis effector related to peripheral myelin protein 22, TP53 apoptosis effector
CHPT1221675_s_at−1.20.000036110.83Choline phosphotransferase 1
CLEC2B§209732_at−4.070.00009968.02C-type lectin domain family 2, member B
JPH3229294_at−2.530.00013395.5Junctophilin 3
SLC14A1229151_at−3.630.00013488.16Solute carrier family 14 (urea transporter), member 1 (Kidd blood group)
CLDN11228335_at−3.40.00015737.04Claudin 11
PTPLA219654_at−1.940.00016226.91Protein tyrosine phosphatase–like (proline instead of catalytic arginine), member A
RNF128219263_at−2.050.00016375.71Ring finger protein 128
BACE2217867_x_at−1.910.00016377.55β-site APP-cleaving enzyme 2
DSG2§1553105_s_at−2.260.00019567.52Desmoglein 2
ZDHHC9222451_s_at−0.80.00020718.73Zinc finger, DHHC-type–containing 9
ATRNL1§1569796_s_at−1.970.0003456.49Attractin-like 1
WFS11555270_a_at−1.260.00045817.33Wolfram syndrome 1 (wolframin)
SLC44A1§228486_at−1.720.00047447.11Solute carrier family 44, member 1
CRTAP1555889_a_at−0.860.000494710.57Cartilage-associated protein
PTGFR207177_at−2.030.00053386.24Prostaglandin F receptor FP
CDS1205709_s_at−1.630.00063956.66CDP-diacylglycerol synthase (phosphatidate cytidylyltransferase) 1
CA12§203963_at−3.10.00068087.09Carbonic anhydrase XII
RNF130§217865_at−0.820.00068089.88Ring finger protein 130
ORAI3221864_at−1.240.00069349.01Calcium release–activated calcium modulator 3
TMEM106A§1552302_at−2.160.00071585.28Transmembrane protein 106A
INSIG1201625_s_at−2.020.00075397.29Insulin-induced gene 1
LRRN4CL§1556427_s_at−1.880.00077528.27Leucine-rich repeat neuronal 4, C-terminal–like
LAMP2200821_at−0.660.00082729.76Lysosomal-associated membrane protein 2
AGPAT9224480_s_at−1.960.00084538.561-acylglycerol-3-phosphate O-acyltransferase 9
KCNS3205968_at−2.110.00090385.67Potassium voltage–gated channel, delayed-rectifier, subfamily S, member 3
CDH19§206898_at−2.150.00090918.97Cadherin 19, type 2
TMEM27§223784_at−2.160.00102354.8Transmembrane protein 27
MXRA7§212509_s_at−1.040.001149311.91Matrix remodeling–associated 7
F2RL1213506_at−2.370.00126994.25Coagulation factor II (thrombin) receptor–like 1
BCL2L13217955_at−0.830.00131998.9Bcl-2–like 13 (apoptosis facilitator)
IGF1R§203627_at−1.30.00134937.97Insulin-like growth factor 1 receptor
NETO2§218888_s_at−2.280.00135615.23Neuropilin and tolloid–like 2
KCNMB4222857_s_at−1.470.00138646.13Potassium large conductance calcium-activated channel, subfamily M, beta member 4
SLC38A7218727_at−0.680.00147355.85Solute carrier family 38, member 7
TYRO3§211432_s_at−1.060.00149626.21TYRO3 protein tyrosine kinase
PDIA4208658_at−1.40.00152338.45Protein disulfide isomerase family A, member 4
REEP1204364_s_at−1.290.0016694.86Receptor accessory protein 1
SLC27A2205769_at−1.610.00169255.17Solute carrier family 27 (fatty acid transporter), member 2
SLC7A7204588_s_at−1.320.00183817.04Solute carrier family 7 (cationic amino acid transporter, y+ system), member 7
GCNT1205505_at−1.640.00188327.98Glucosaminyl (N-acetyl) transferase 1, core 2 (β-1,6-N-acetylglucosaminyltransferase)
CRHR1§214619_at−0.550.00200585.08Corticotropin-releasing hormone receptor 1
SEL1L202064_s_at−0.560.00205256.72Sel-1 suppressor of lin 12–like (Caenorhabditis elegans)
ABCG1204567_s_at−2.50.00205465.05ATP-binding cassette, subfamily G, member 1
SLC7A1206566_at−0.830.00205464.86Solute carrier family 7 (cationic amino acid transporter, y+ system), member 1
ARMCX6§214749_s_at−0.920.0021710.13Armadillo repeat–containing, X-linked 6
GPR175218855_at−0.730.00228587.62G protein–coupled receptor 175
NPAL3214579_at−0.550.00242596.48Nuclear interacting partner of anaplastic lymphoma kinase–like domain–containing 3
SSR2200652_at−0.730.002544811.26Signal sequence receptor, beta (translocon-associated protein β)
LRRC15213909_at−2.870.00257086.76Leucine-rich repeat–containing 15
HS6ST3232276_at−1.010.00283935.13Heparan sulfate 6-O-sulfotransferase 3
FZD5221245_s_at−1.730.00311438.5Frizzled homolog 5 (Drosophila)
SLC22A17218675_at−1.070.00350637.3Solute carrier family 22, member 17
MYADM224920_x_at−1.520.0035147.96Myeloid-associated differentiation marker
DUOX1219597_s_at−0.550.00356064.34Dual oxidase 1
NRXN3§229649_at−1.210.00416624.52Neurexin 3
TSPAN31§203227_s_at−0.830.00416639.57Tetraspanin 31
KCNE3227647_at−1.420.00418036.24Potassium voltage–gated channel, Isk-related family, member 3
YME1L1201352_at−0.70.004469510.73Yeast mitochondrial escape–like 1 (Saccharomyces cerevisiae)
BMPR1B229975_at−1.320.00448064.91Bone morphogenetic protein receptor, type 1B
FAM62B§1558511_s_at−0.810.00448929Family with sequence similarity 62 (C2 domain–containing) member B
RELL1226430_at−1.170.004725810.36Receptor expressed in lymphoid tissues–like 1
FKBP11219117_s_at−1.450.00504978.88FK-506 binding protein 11, 19 kd
SGCG§207302_at−1.830.00506134.55Sarcoglycan, gamma (35-kd dystrophin-associated glycoprotein)
UBE2V1201001_s_at−0.50.00511138.95Ubiquitin-conjugating enzyme E2 variant 1
NHLRC3227040_at−1.390.0052536.81NHL repeat–containing 3
ITGB8§205816_at−1.890.00531655.46β8 integrin
TOMM20200662_s_at−0.790.005717711.29Translocase of outer mitochondrial membrane 20 homolog (yeast)
TMEM71238429_at−2.060.00618265.76Transmembrane protein 71
GPR133232267_at−2.660.00623816.18G protein–coupled receptor 133
FNDC3B§229865_at−1.080.00634787.49Fibronectin type III domain–containing 3B
SLC9A1209453_at−0.720.00652878.38Solute carrier family 9 (sodium/hydrogen exchanger), member 1 (antiporter, Na+/H+, amiloride sensitive)
PPAPDC1B226150_at−1.020.00693899.67Phosphatidic acid phosphatase type 2 domain–containing 1B
C1orf9§203429_s_at−0.80.00729989.45Chromosome 1 open-reading frame 9
SLC24A1206081_at−0.860.00730225.83Solute carrier family 24 (sodium/potassium/calcium exchanger), member 1
MFAP3L§210493_s_at−1.330.00732996.51Microfibrillar-associated protein 3–like
NRM225592_at−0.620.0075287.08Nurim (nuclear envelope membrane protein)
TFR2210215_at−0.710.00795355.36Transferrin receptor 2
SLC22A23223194_s_at−1.420.00860947.35Solute carrier family 22, member 23
IL31RA243541_at−0.760.00916683.51Interleukin 31 receptor A
FZD3§219683_at−1.110.00946525.39Frizzled homolog 3 (Drosophila)
MAGT1221553_at−0.960.00966168.77Magnesium transporter 1
WNT11206737_at−0.930.00970595.49Wingless-type murine mammary tumor virus integration site family, member 11
POR208928_at−0.710.00996557.39Cytochrome P450 oxidoreductase

Gene list reduction.

Further reduction of the gene list was required in order to obtain the most worthy candidates for laboratory validation and for potential use as an NP target using phage display and cloning. Gene list reduction according to the methods described above for short-listing of candidate genes generated a final short list of 28 candidates from the original 90, of which 23 were “independent” candidates and the other 5 were made up of 3 genes containing fibronectin type III domains and 2 containing Ig-like C2 type I domains. While the latter groups may not provide the level of cell specificity required, they were retained for RT-PCR analysis to quantify their expression level.

RT-PCR results.

Due to small sample volume, real-time PCR was carried out on NP, AF, and AC samples from only 3 donors (subjects 4, 6, and 7). Since there was no AC available from subject 4, the AC sample from subject 5 was used, in order to obtain 3 representative values. Of the 28 genes analyzed, 27 were detectable in all samples (MFAP3L was undetectable), although some of them were expressed at very low levels, especially in the AC. Genes were normalized to the average of 4 housekeeping genes. For comparison of NP and AC an average from the 3 AC samples was taken for subject 4, while the corresponding AC was considered for subjects 6 and 7. For most of the genes, the expression differences between NP and AC cells found in the microarray analysis was mirrored. Significantly higher expression in NP versus AC cells was found for CLEC2B, ATRNL1, SLC44A1, CA12, TMEM27, TSPAN31, and SGCG (all P < 0.01) and for DNER, DSG2, CDH19, MXRA7, NETO2, CRHR1, and TYRO3 (all P < 0.05), while the differences between NP and AC cells were close to significant for ODZ3, FLRT2, and ITGB8 (P < 0.07). Of these genes, CLEC2B and CA12 (P ≤ 0.01), and SGCG and TYRO3 (P < 0.05) were additionally found to be expressed more highly in NP as compared to AF cells (Figure 2).

thumbnail image

Figure 2. Transmembrane domain–containing proteins exhibiting significantly higher relative gene expression in nucleus pulposus (NP) cells compared to articular cartilage (AC) cells. Gene expression was normalized to the average of 4 housekeeping genes and is presented as log2-transformed values. Data are shown as box plots, where the boxes represent the 25th to 75th percentiles, the lines within the boxes represent the median, and the lines outside the boxes represent the 10th and 90th percentiles. ∗ = genes more highly expressed in NP cells than in anulus fibrosus (AF) cells as well as AC cells.

Download figure to PowerPoint

To determine potential correlations between the relative mRNA expression and the age or disc degeneration grade of the donor, the most significantly differentially expressed gene, CA12, was chosen for assessment by real-time PCR in samples from all 7 donors. CA12 mRNA expression in NP cells was found to be strongly negatively correlated with both age (r = −0.923) and Thompson grade (r = −0.981), as shown by Pearson correlation analysis (P < 0.01). In accordance with previous reports (13), there was a strong positive correlation between age and Thompson grade (r = 0.74, P < 0.01). These results indicated an obvious reduction in CA12 mRNA expression with degrading NP cells. This significant result suggested that CAXII could possibly be used as a sensitive biomarker of degradation and as a target for NP cells, as per the primary aim of this study.

Immunohistochemistry results.

CAXII was selected for immunohistochemical evaluation because of its relatively high mRNA expression level in the NP and its markedly stronger expression in NP compared to both AF and AC cells. The results of CAXII immunolabeling of NP tissue sections are summarized in Table 3. Although CAXII is known to be specifically expressed at the cell surface (28, 29), its transmembrane localization cannot be recognized using conventional immunohistochemical methods, as the appearance of the staining is highly dependent on the exact sectioning plane of the sample.

Table 3. Immunohistochemical results on intervertebral disc sections*
Section no. (donor age/sex)Thompson gradeCAXII staining of NP
  • §, *

    Nucleus pulposus (NP) tissue of discs between L4 and L5 was assessed. (+) = 1–2 positive cells per field of view; + = 3–4 positive cells; ++ = 5–10 positive cells; +++ = >10 positive cells. CAXII = carbonic anhydrase XII; NA = not available.

1 (3/F)1+++
2 (6/F)1(+)
3 (14/M)1++
4 (14/F)1(+)
6 (18/M)1
7 (19/F)1+
8 (21/M)1+
9 (22/F)1+
5 (17/M)2++
10 (35/F)2+
11 (35/M)NA+
12 (36/M)2+
13 (41/M)2(+)
15 (51/F)2(+)
14 (47/F)3++
19 (63/M)3++
16 (54/F)4++
17 (57/M)4+
18 (62/M)5+++
20 (67/F)5+++
21 (72/M)5+++

Strong immunoreactivity for CAXII was observed in the large notochord-like cells of the NP of the 3-year-old female donor (Figures 3A and B). Also in this section, some cells of the inner AF stained positive, whereas CAXII immunoreactivity was not noted in either AF or the cartilaginous end plate of any other IVD section probed. In healthy discs (Thompson grade 1–2) from adolescent and young adult donors, only a few positive cells were identified. Immunoreactive cells were either located as single cells or occasionally arranged in cell clusters (Figures 3C and D). In discs with more degeneration (Thompson grade 3–4) from older donors, the cells that stained for CAXII were more abundant (Figures 3E and F). Finally, the most intense immunolabeling was visible in the tissue of severely degenerated discs (Thompson grade 5) (Figures 3G and H). To verify the specificity of the reaction, corresponding negative control sections for all positive areas were examined, and staining was not found in any.

thumbnail image

Figure 3. Immunolabeling of nucleus pulposus (NP) cells for carbonic anhydrase XII (CAXII). A and B, NP of a 3-year-old female subject. Fine granular cytoplasmatic CAXII labeling is seen. C and D, NP of a 17-year-old male subject. Clusters of CAXII-positive cells are evident. E and F, NP of a 47-year-old female subject with Thompson grade 3 disc degeneration. Perinuclear CAXII labeling is seen. G and H, NP of a 62-year-old male subject with Thompson grade 5 disc degeneration. Intense positive labeling for CAXII can be detected in cells of diseased tissue. B, D, F, and H are higher-magnification views of the boxed areas in A, C, E, and G, respectively. Bars in A, C, E, and G = 50 μm; bars in B, D, F, and H = 10 μm.

Download figure to PowerPoint

In Alcian blue/picrosirius red–stained sections, mainly blue labeling of proteoglycans was observed in Thompson grade 1–2 discs from young subjects, red staining was apparent in grade 3 moderately degenerated discs, and only pathologic tissue was detected in severely degenerated discs. Observations with polarized light confirmed the increasing amount of collagenous fibers with increased degeneration. In degenerated discs, a detailed image of the collagen fibers could be obtained under polarized light only (images available from the corresponding author upon request).

Effect of CAXII inhibition.

In low concentrations (1 nM and 10 nM), 4-aminobenzene sulfonamide did not affect the lactate level in NP cell–conditioned media. However, at concentrations of 100, 1,000, and 10,000 nM, respectively, the CA inhibitor significantly (P < 0.01) reduced the relative lactate level to 0.891 ± 0.045, 0.880 ± 0.060, and 0.867 ± 0.059 (mean ± SD values relative to the concentration in control medium [set at 1]) in the medium with NP cells cultured under 5% oxygen conditions; the lactate level in NP cells cultured at 21% oxygen tension was affected by CA inhibitor in high concentration only (10,000 nM). No cytotoxic effect of 4-aminobenzene sulfonamide on NP cells was observed at the concentrations studied.


  1. Top of page
  2. Abstract
  7. Acknowledgements

Conceptually, as NP cells are considered both a unique and a contained cell type as well as the major, constant contributor to IVD degeneration, localized targeting is a viable option for IVD therapy. This approach requires cell surface–specific markers, and we hence adopted a systematic approach to carry out their detection. DNA microarrays were used to facilitate a global transcriptomic approach to identifying transmembrane domain–containing genes that are differentially expressed between NP cells and AF/AC cells. The unsupervised clustering transcriptomic results revealed similarity between the NP and AF cell types and a distinctive difference between the NP/AF and AC samples. The former result was unexpected for 2 reasons: 1) the different origins of the tissue types, with the ongoing (but nonexclusive) hypothesis that NP is the remnant of the notochord (30) and AF is from the mesoderm (31), and 2) a clear morphologic distinction, as visible by early Thompson grades.

The NP contains mostly type II collagen and proteoglycans in a random orientation, whereas the AF consists of lamellae composed of highly orientated types I and II collagen and proteoglycans (32). Looking specifically at our data set, this profile is consistent with COL1A1 down- and COL2A1 up-regulation in the NP in comparison to AF; however, for both genes, the difference in expression between NP and AF did not reach statistical significance (data not shown). While there may be clear phenotypic differences between NP and AF cells that are healthy or only mildly degenerated, this distinction becomes less clear in older, more highly degenerated discs. However, the observed lack of differences between the 2 IVD tissue types cannot be attributable solely to the similarity between the tissues during later-stage disc degeneration, since differences between NP and AF were also not observed in samples with early degeneration. It rather indicates the similarity of the AF and NP cell phenotypes in spite of different cell morphology and ECM composition of the tissues, which may be related to the different mechanical environment to which the cells are exposed. More subtle and individual gene differences must be attributable to the morphologic differences between the IVD cell types.

With regard to the difference between the NP and AC profiles, NP cells are morphologically similar to AC cells, both having a round, chondrocytic morphology (33), from which one might assume a functional similarity. However, it has been reported that the ECM in which they reside differs in terms of proteoglycan and collagen content (34–36), and Nettles et al (8) suggested that there is a strong physical interaction between cells and ECM and hypothesized that the composition of the ECM may contribute to the differing cellular responses to mechanical loading between anatomic zones (37). This difference is obvious in our analyses and may partially explain the number of differentially expressed genes between these cell types. Only 3 probe sets were found that fit (albeit marginally) under our significance threshold of P < 0.05 for the difference between AF and NP, with each having a P value of 0.049, further illustrating the heterogeneous nature of the samples from different donors. Pursuing a search for transmembrane-associated genes meant treating the NP and AF populations as one and searching for significant differences from the cartilage samples. CAXII was the most viable candidate, as it had a high level of gene expression relative to the housekeeping genes and was more abundantly expressed in the NP than AF and AC cells. In fact, CA12 was recently described by Minogue et al as a marker gene for human NP cells (38).

The mRNA and protein data for CAXII did not directly correlate. Moreover, mRNA levels in Thompson grade 5 samples could not be assessed to corroborate the immunohistochemistry results because the low cell density did not allow for gene expression profiling; thus, it is not clear to what extent the disparity is present at this stage of degeneration. However, the finding of decreasing mRNA and protein levels with early aging combined with a reexpression of the protein in degenerated tissue parallels observations on other potential NP markers, such as keratin 19 (12). Carbonic anhydrases catalyze the reversible hydration of carbon dioxide into protons and bicarbonate. In tumor cells, hypoxia-inducible CAIX and CAXII have been shown to promote cell survival and growth through maintenance of the intracellular pH (39). There is little available information about the pH regulation in IVD cells. Ion transport pathways have been investigated (40), and distinct molecules that allow for survival in a low-pH environment have been identified (41). We found reduced lactate release upon CA inhibition in samples cultured under low oxygen conditions similar to those in IVD cells in vivo, and this suggests that intracellular pH regulation may be disturbed, leading to intracellular acidosis. At higher concentrations the CAXII inhibitor used in this study may also affect other CAs (25), leading to dysregulation of the cell metabolism under normoxic conditions as well.

CAXII may thus function to counteract cellular acidosis in the hypoxic NP, which results from lactic acid accumulation following glucose metabolism and lack of waste product removal. While CAXII expression in cells of young growing NP may be necessary for cellular pH regulation during high metabolic activity during development, the induction of this molecule in degenerative discs may indicate a response to increased metabolic activity following inflammation or catabolic changes. Such cells with enhanced CAXII expression could be targeted for delivery of anabolic, anticatabolic, and/or antiinflammatory agents. Considering the intense staining in 6 of 7 Thompson grade 3–4 discs in the present study, CAXII may be suitable for targeting discs with moderate to severe degenerative changes.

Although at present it is still not clear which degenerating discs will actually result in back pain, treatment at adjacent levels after spinal fusion will be among the first clinical targets, as this surgical procedure commonly leads to adjacent-level degeneration. In addition, the application of therapeutic cargo–carrying nanoparticles in patients with NP protrusion (another condition that invariably leads to disc degeneration) is envisaged. Studies to verify whether cells in the NP of these patients also produce CAXII are ongoing.

In summary, we have applied standard molecular biology tools to identify cell surface proteins that are specific for IVD cells. We used a sample set of human tissue rather than tissue from other mammals, as interspecies variations are common and the research question directly aims to advance toward application in the human patient. Ultimately, we created a short list of candidate cell surface markers that can be targeted in future studies, with CAXII showing promise as an intermediate-to-late–stage degeneration target, which may be of crucial importance in the preventive treatment of disc degeneration leading to chronic low back pain.


  1. Top of page
  2. Abstract
  7. Acknowledgements

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. Dr. Grad 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. Power, Grad, Wall, Alini, Pandit, Gallagher.

Acquisition of data. Power, Grad, Rutges, Creemers, van Rijen, Pandit.

Analysis and interpretation of data. Power, Grad, O'Gaora, Wall, Alini, Pandit, Gallagher.


  1. Top of page
  2. Abstract
  7. Acknowledgements

The authors thank Dr. Stefan Milz and Christoph Sprecher for their assistance with immunohistochemical analysis and Dr. F. Cumhur Oner for support with sample scoring.


  1. Top of page
  2. Abstract
  7. Acknowledgements