Introduction

Several transcriptional factors are thought to be key factors directing differentiation of mesenchymal tissues: molecules of the MYOD family for muscle differentiation (Ferrari et al. 1998), PPARG2 for adipose tissue differentiation (Rosen and Spiegelman 2000), and CBFA1 for osteogenic differentiation (Komori et al. 1997). Although SOX9 and CBFA1 are thought to be key transcriptional factors during cartilage differentiation (Bi et al. 1999), other factors are likely to be involved as well (Takeda et al. 2001). Articular cartilage (hyaline cartilage) of the knee, and the knee meniscus (fibrocartilage), are different (Ross et al. 1995). Identification of common cis-elements in genes expressed specifically in these two types of cartilage should yield insight into cartilage-specific transcriptional regulation.

Investigators interested in cartilage development and regeneration have a few markers available for predicting the direction of differentiation, i.e., type II collagen and aggrecan for hyaline cartilage, and type I collagen for fibrocartilage (Burgeson et al. 1982; Ross et al. 1995; Carlberg et al. 2001). Cartilaginous cells are believed to originate from multipotential mesenchymal cells that resemble human mesenchymal stem cells (hMSCs) (DeLise et al. 2000; Carlberg et al. 2001). Since hMSCs are relatively easy to obtain, hMSC cultures are considered to be good sources of cells for modeling regeneration of mesenchymal tissues such as cartilage and bone (Pittenger et al. 1999; Carlberg et al. 2001; Ochi et al. 2003). Thus, genes that are expressed specifically in either hyaline cartilage or fibrocartilage should be useful markers to indicate the direction of differentiation. To identify such genes, we have applied a cDNA microarray containing 23,040 genes in a system that provides high-throughput analysis of expression. In the work reported here, we evaluated the power of this system to detect cartilage-specific molecules.

Materials and methods

Clinical samples

All patients were diagnosed and treated by members of the Department of Orthopedic Surgery, Keio University School of Medicine. "Normal" articular cartilage samples were obtained during total knee replacement surgery performed on eight female patients (age range 56–80 years, median 69 years) who were diagnosed with primary osteoarthritis of the knee. In each case a non-affected site of articular cartilage, such as the patellar–femoral joint, was carefully peeled to provide samples for further investigations. "Normal" meniscus tissues were obtained from samples removed during endoscopic meniscectomy of six patients (five male and one female; age range 13–28 years, median 20 years) who were diagnosed with traumatic meniscal injury; again, non-affected tissue corresponding to each sample was obtained for our investigations. All 14 patients (or their families in some cases) gave informed consent according to the guidelines of both Keio University and the University of Tokyo. Resected samples were snap-frozen in liquid nitrogen immediately and sent to the Human Genome Center, Institute of Medical Science at the University of Tokyo, without any personal identifiers.

Preparation, T7-based amplification, and labeling of RNAs

Total RNA was extracted from each frozen clinical sample and from human mesenchymal stem cells (hMSC) purchased from BioWhittaker, Inc. (Walkersville, MD) to serve as a universal control, using TRIzol reagent (Life Technologies, Inc., Rockville, MD). Total RNAs of 29 normal human tissues (mesenteric adipose tissue, bone marrow, brain, heart, kidney, liver, lung, lymph node, mammary gland, pancreas, placenta, prostate, salivary gland, skeletal muscle, small intestine, spinal cord, spleen, stomach, testis, thymus, thyroid, trachea, uterus, colon, ovary, fetal brain, fetal kidney, fetal liver, and fetal lung) were obtained as described previously (Saito-Hisaminato et al. 2002). Equal amounts of RNA from each of the 29 normal tissues were pooled to serve as another universal control (29Mix). All RNA preparations were digested with RNase-free DNase I (Nippon Gene Co., Tokyo, Japan) according to the manufacturer's recommendations. Two rounds of T7-based RNA amplification and the preparation of cDNA probes were carried out as described elsewhere (Kitahara et al. 2001; Ochi et al. 2002). Amplified RNA (2.5 µg) from each cartilage sample (articular cartilage: a mixture of 8 cases; meniscus, 6 patients) was labeled with Cy5-dCTP (Amersham Biosciences, Uppsala, Sweden); an equal amount of amplified RNA from each universal control (29Mix or a pool of amplified RNA from hMSCs) was labeled with Cy3-dCTP (Amersham Biosciences) for hybridization to the cDNA microarray.

Construction and analysis of the cDNA microarray

Fabrication of our cDNA microarray slides has been described previously (Ono et al. 2000; Kitahara et al. 2001). Slides containing a duplicate set of 23,040 cDNA spots were used for each analysis of expression profiles, to reduce experimental fluctuation. Hybridizations were performed to compare hyaline cartilage (Cy5) with 29Mix (Cy3), fibrocartilage (Cy5) with 29Mix (Cy3), hyaline cartilage (Cy5) with hMSC (Cy3), and fibrocartilage (Cy5) with hMSC (Cy3). Hybridization, washing, and scanning were performed as described elsewhere (Ono et al. 2000; Kitahara et al. 2001). The intensity of each duplicated signal was evaluated photometrically using the ArrayVision computer program (Imaging Research, Inc., St Catharines, Ontario, Canada). Intensities of the 29 individual normal tissues (test) and 29Mix (control) were evaluated as described elsewhere (Saito-Hisaminato et al. 2002). The fluorescence intensities of Cy5 (test) and Cy3 (control) for each target spot were adjusted so that the mean Cy5/Cy3 ratio of the 52 housekeeping genes was equal to one. We assigned a cut-off value to each microarray slide, using variance analysis. If both Cy3 and Cy5 signal intensities were less than the cut-off values, the expression level of the corresponding gene in that sample was assessed as low or absent. For other genes we calculated Cy5/Cy3 as a relative expression ratio. A two-dimensional hierarchical clustering method was performed as described before (Saito-Hisaminato et al. 2002).

Semi-quantitative RT-PCR

A 3-µg aliquot of each RNA to be used for microarray analysis was reverse-transcribed to single-stranded cDNA using oligo(dT)12–18 primer and Superscript II (Invitrogen, Carlsbad, CA). Semi-quantitative reverse transcription PCR (RT-PCR) was carried out for seven selected genes as described previously (Ono et al. 2000), using the following primer sets: for UBA52, 5′-GAT CTT TGT GAA GAC CCT CAC TG-3′ and 5′-CAG ATC ATC TTG TCG CAG TTG TA-3′; for RPL23, 5′-CTG TAT ATC ATC TCC GTG AAG GG-3′ and 5′-AAG TCT GCA CAC TCC TTT GCT AC-3′; for RAB13, 5′-CAA CAA CAC TTA CAT CTC CAC CA-3′ and 5′-GAG CAC TTG TTG GTG TTC TTC TT-3′; for IGF2, 5′-CTT GGA CTT TGA ATC AAA TTG G-3′ and 5′-CCT CCT TTG GTC TTA CTG GG-3′; for IGL@, 5′-CAA CTG TAC ACC TAA AGG CTC TC-3′ and 5′-ATC TGA GAG GAG AGG AGA GTG A-3′; for PDLIM1, 5′-CCA TAC AAG ATG AAT TTA GCC TCT G-3′ and 5′-GTC AGA CAC GTT ATA TT TGA TTG GG-3′; and for COL1A1, 5′-CAT GTT CGG TTG GTC AAA GAT A-3′ and 5′-AAT ACA AAA CCA CCA AGA CCT CC-3′. PCR reactions were optimized for the number of cycles to ensure product intensity within the linear phase of amplification.

Results and discussion

Identification of genes that were highly expressed in both hyaline cartilage and fibrocartilage

Comparison of expression profiles in cartilage with those of other normal human tissues should indicate genes that are preferentially expressed in cartilage. We selected 6707 genes for which data were present for hyaline cartilage; 585 of them, including five collagen genes, showed hyaline cartilage/29Mix ratios > 10, and 209 also showed hyaline cartilage/hMSC ratios > 10 (data not shown). Similarly, we selected 6436 genes for which data were present for fibrocartilage; 605 showed fibrocartilage/29Mix ratios > 10 and 145 showed fibrocartilage/hMSC ratios > 10 (data not shown).

Identification of "cartilage-specific" genes

Articular cartilage and meniscus are both considered to be cartilaginous tissues (Ross et al. 1995). A hierarchical clustering of the 31 normal human tissues we examined, including the two types of cartilage, indicated that the latter are very closely related in gene expression profiles (data not shown). Thus we assumed that some phenotypes common to hyaline cartilage and fibrocartilage should be maintained by the same "cartilage-specific transcriptional regulators."

To obtain candidates for these putative "cartilage-specific transcriptional regulators," we searched for genes whose relative expression ratio for either hyaline cartilage/29Mix or fibrocartilage/29Mix were >5 while (i) expression ratios in the 29 non-cartilaginous tissues were <2 or (ii) expression was absent in 90% of the normal tissues, and identified 69 genes. Among them, 24 revealed expression ratios of >5 for both hyaline cartilage/29Mix and fibrocartilage/29Mix; but 22 of the 69 were highly expressed only in hyaline cartilage and 23 were highly expressed only in fibrocartilage (Fig. 1a, b).

Fig. 1a, b.
figure 1figure 1

Schema of "cartilage-specific genes." Relative expression ratios were (i) >5 in cartilaginous tissues, while ratios in a mixture of cDNAs from 29 other normal tissues were <2 or (ii) >5 in cartilaginous tissues, while ratios were missing for 90% of individual normal tissues compared with the 29Mix. a Twenty-four of the 69 genes were highly expressed in both hyaline cartilage and fibrocartilage (overlapping), while 22 were highly expressed in hyaline cartilage and 23 were highly expressed in fibrocartilage. Each cell in the matrix represents the expression level of a single transcript in a single tissue, with red and green indicating transcript levels respectively above and below the median for that gene across the 29Mix. Black represents unchanged expression; gray indicates slight expression or none (intensities of both Cy3 and Cy5 less than the cut-off value). Abbreviations: H, relative expression ratio in hyaline cartilage; F, relative expression ratio in fibrocartilage. b Semi-quantitative RT-PCR of three selected genes using RNAs from two types of cartilage and six other normal tissues. The integrity of each template was controlled through amplification of β2MG

Full size image

Identification of separate marker genes for hyaline cartilage and fibrocartilage

Comparison of expression profiles of articular cartilage (hyaline cartilage) and meniscus (fibrocartilage) with the profile obtained from a mixture of 29 other normal tissues showed a specific character for cartilaginous tissue after final differentiation, but this approach was not sufficient for identifying genes important for understanding development and regeneration of mesenchymal tissues. Thus, we re-examined the expression profiles to investigate genes that showed different levels of expression between each type of cartilage and hMSCs. We first selected 4837 genes for which data were obtained for either hyaline cartilage/hMSC or fibrocartilage/hMSC and then looked for candidate marker genes that could satisfy the criteria of both "hyaline cartilage/hMSC ratio of >2" and "fibrocartilage/hMSC ratio of <0.5," or genes that would satisfy the criteria of both "fibrocartilage/hMSC ratio of >2" and "hyaline cartilage/hMSC ratio <0.5." We detected 22 genes that were, by this definition, differentially expressed in one or the other of the two types of cartilaginous cells (Fig. 2a, b).

Fig. 2a, b.
figure 2

Schema of marker genes for hyaline cartilage or fibrocartilage. Relative expression ratios were >2 in either type, and the ratios in the other cartilage were <0.5 compared with hMSC, for all 22 genes listed. a Eleven of the 22 genes meeting the strict differential criteria were highly expressed only in hyaline cartilage (H), while 11 others were highly expressed only in fibrocartilage (F). The color scale was constructed from expression ratios of cartilaginous tissue (Cy5):hMSC (Cy3) as described in Fig. 1. b Semi-quantitative RT-PCR of four selected genes using RNAs from two types of cartilage and from hMSCs. The integrity of each template was controlled through amplification of β2MG

Full size image

In all, we examined 92,160 gene expression patterns in two cartilaginous tissues, in comparison with 668,160 measurements of genes expressed in 29 other normal tissues, by means of a cDNA microarray. To our knowledge this is the first study to apply such a large-scale approach for analyzing expression profiles of cartilaginous tissues. This work has established an expression database for sets of genes that are expressed specifically in cartilage.

We subjected the expression profiles of two types of cartilaginous tissue and 29 normal tissues to a hierarchical clustering analysis. The dendrogram showed that articular cartilage and meniscus fell into the same, closely related terminal branch, indicating that both hyaline cartilage and fibrocartilage were very similar in gene expression patterns. Among several mesodermal tissues included in the 31-tissue panel, only bone marrow and nerve tissues yielded expression patterns that were relatively similar to that of cartilage (data not shown). Our data imply that significant alterations in expression level must occur for a large number of genes during the differentiation process.

Transcriptional regulation of genes in a given tissue is controlled by a combination of mechanisms, for example changes in chromatin structure mediated by the association of histone acetylases and deacetylases (Wade 2001), binding of transcriptional regulators to an enhancer and/or a suppressor (Luscher and Larsson 1999), binding of transcription factors to a promoter region, and methylation of CpG islands in promoter sequences (Ballestar and Wolffe 2001). Although the precise mechanisms of cooperation among these elements are not well understood (Cha et al. 2000), identification of common cis-elements that exist only in cartilage-specific genes should yield insight into the physiology of chondrocytes. Such core sequences might be useful for (a) identifying novel transcription-related factors in cartilage, and (b) establishing cartilage-specific expression vectors for genes that are necessary for regenerating and/or maintaining phenotypes of cartilaginous tissues.

In our experiments, 24 genes were expressed specifically in both hyaline cartilage and fibrocartilage. Among them, WNT7A is member of the WNT gene family, which consists of structurally related genes encoding secreted signaling molecules. This gene is implicated in the regulation of limb chondrogenesis (Rudnicki and Brown 1997; DeLise et al. 2000; Tufan et al. 2002). COMP is normally expressed at high levels in the territorial matrix of chondrocytes and is believed to belong to the family of thrombospondin genes (Newton et al. 1994); mutations of COMP have been found in patients with pseudoachondroplasia or multiple epiphyseal dysplasia (Ikegawa et al. 1998), indicating its important role in maintaining cartilaginous structure (Thur et al. 2001). GLG1 encodes a conserved membrane sialoglycoprotein present in the Golgi apparatus of most cells; this protein binds with basic fibroblast growth factor (FGFB) and is expressed during chondrogenesis (Stieber et al. 1995). GPS2 encodes a 327-amino-acid polypeptide with no similarity to other proteins or known motifs; however, this molecule potently regulates RAS- and mitogen-activated protein kinase (MAPK)-mediated signals and interferes with JNK activity in yeast and mammalian cells (Spain et al. 1996). Considering that WNT genes, COMP, FGFB, MAPK, and JNK are known to play important roles in chondrogenesis (Thur et al. 2001), their products might well be critical to the physiology of chondrocytes. SOX9, believed to be a key transcriptional factor in chondrocytes (Bi et al. 1999), is a member of the sex-determining region Y-type HMG (high mobility group) box protein family; it is required for chondrocyte-specific gene expression and chondrogenesis (Bi et al. 1999). In our experiments the SOX9 gene was highly expressed in both hyaline cartilage and fibrocartilage (relative ratios > 10), but its expression was relatively ubiquitous (relative ratio > 2 in seven of the 29 normal non-cartilaginous tissues examined).

Several known markers are commonly invoked to predict the direction of differentiation in studies of cartilage development and regeneration, specifically collagen type II and aggrecan for hyaline cartilage and type I collagen for fibrocartilage (Burgeson et al. 1982). Our genome-wide cDNA microarray analyses revealed 22 additional genes that were specifically expressed in cartilaginous tissue: 11 in hyaline cartilage and 11 others in fibrocartilage. RT-PCR experiments also supported differences in gene expression patterns among hyaline cartilage, fibrocartilage, and hMSC, indicating that this set of 22 genes can include candidates to serve as markers during chondrogenesis from multipotential mesenchymal cells.

Our data indicated high expression of type I collagen in fibrocartilage, but expression of this gene was very low in hyaline cartilage. On the other hand, type II collagen was extremely abundant in hyaline cartilage (ratio of >100), and relatively so in fibrocartilage (ratio of 15). The consistency of our data with previous reports verifies their quality (Tufan et al. 2002). One of the main goals of regenerative medicine is to replace damaged tissue with regenerated tissue that can function normally. To realize this goal it is important to establish methods to generate the required tissue as well as to maintain the physiological function of the regenerated tissue. The data documented here should provide useful information for both approaches and contribute to research toward regeneration of cartilage tissues in clinical settings.