Isolation of markers for chondro-osteogenic differentiation using cDNA library subtraction. Molecular cloning and characterization of a gene belonging to a novel multigene family of integral membrane proteins.

To identify novel marker molecules associated with chondro-osteogenic differentiation, we have set up a differential screening system based on a cDNA library subtraction in organ cultures of prenatal mouse mandibular condyles. Differential screening of a cDNA library constructed from in vitro cultured condyles allowed the isolation of a novel gene, named E25. Full-length E25 cDNA is predicted to encode a type II integral membrane protein of 263 amino acid residues. In situ hybridization experiments show that E25 is expressed in the outer perichondrial rim of the postnatal mandibular condyle, which contains the proliferating progenitor cells, but not in the deeper layers of the condyle containing the more differentiated chondroblasts and chondrocytes. Other cartilagenous tissues and their perichondrium were negative. Strong in situ hybridization signals were also detected on bone trabeculae of mature bone in tooth germs and in hair follicles. Northern blot analysis showed strong expression in osteogenic tissues, such as neonatal mouse calvaria, paws, tail, and in skin. This expression profile suggests that E25 could be a useful marker for chondro-osteogenic differentiation. Homology searches of DNA databanks showed that E25 belongs to a novel multigene family, containing three members both in man and mouse. The mouse E25 gene locus (Itm2) was mapped to the X chromosome.

It is well established that mesenchymal stem cells can give rise to committed precursors of several lineages of the connective tissue family, including bone and cartilage (1)(2)(3)(4)(5)(6). The hierarchy of the different lineages and the mechanism that is involved in their differentiation from the multipotential mesenchymal stem cell have not been completely elucidated. Also there are indications that commitment to one lineage is not necessarily irreversible and that transitions between different cell types may be possible (7)(8)(9). In contrast to myogenic (10) and adipogenic (11) commitment, of which the molecular details are gradually being unraveled, the study of the initial stages and molecular control mechanisms of chondro-and osteogenic commitment is less well advanced (12). Although the maturation of preosteoblasts into bone-forming osteoblasts is relatively well known and a model has been proposed, invoking three distinct phases during progressive development of the osteoblast phenotype (13), molecular markers that are specific for the early stages of chondro-/osteogenic commitment and differentiation are scarce (6). It is clear that a better understanding of early osteogenic and chondrogenic differentiation would benefit from the availability of a larger number of these early markers.
One strategy to identify marker molecules is to generate monoclonal antibodies that selectively react with various stages of the osteogenic differentiation cascade (14,15). Other authors have applied subtractive hybridization/differential screening methods (16 -21). Most of the differential screening experiments that have been described have focused on systems containing relatively well differentiated osteoblasts. In order to discover additional markers for the early stages of osteogenic differentiation, we have undertaken a differential cDNA screening analysis of organ cultures of prenatal mouse mandibular condyles. The mandibular condyle is a secondary cartilage that, like most elements of the craniofacial skeleton, arises from neural crest-derived mesenchyme (22). In vivo, mesenchymal-like stem cells in the perichondrium give rise to skeletal precursor cells (skeletoblasts) which enter the chondrogenic differentiation pathway and eventually undergo endochondral ossification. However, when the perinatal condyle is explanted and cultured in organ culture, the perichondrial precursor cells enter osteogenic differentiation, start expressing bone matrix-specific markers, and form a mineralized bone matrix over a period of 2-3 weeks (23)(24)(25)(26). Since the condylar precursor cells develop exclusively along the chondrogenic pathway in vivo and are induced only by culture conditions to switch to osteogenic differentiation, condyle organ cultures constitute a suitable experimental system in which to study and isolate early markers of bone differentiation.
In this study we have subtracted cDNA from 1-day-old condyle explant cultures with an excess driver RNA derived from the fibroblastic cell line Balb/3T3 (27). This subtracted cDNA was used as a probe to differentially screen a cDNA library from 1-day-old explanted condyles. We preferred to use Balb/ 3T3 RNA as a source of subtractor RNA rather than RNA from uncultured condyles (t ϭ 0) as this would allow for simultaneous isolation of novel chondrogenic markers, which can be * This work was supported in part by the "Vlaams Actieprogramma Biotechnologie" (VLAB/034 and TOP-027). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM  assumed to be still present in condylar tissue after 1 day of culture. At the same time the Balb/3T3 cell line is relatively close to the chondro-and osteogenic lineages in order to allow subtraction of a maximum of lineage-nonspecific genes. In view of the limiting amounts of RNA that could be obtained only from fetal condyles and which would preclude performing multiple subtractive hybridizations, we developed a subtractive hybridization/differential screening approach in which subtracted cDNA could be repeatedly generated from target (1day-old condyle explants) and driver (Balb/3T3) cDNA libraries. Rather than converting the subtracted cDNA in a subtracted cDNA library, which is generally of poor quality and needs further subscreening to identify differentially expressed transcripts, the subtracted cDNA was used as a probe in conjunction with unsubtracted cDNA to differentially screen the 1-day-old explant condyle cDNA library. This has the advantage that target specific transcripts which are thus enriched in the subtracted probe can be immediately identified and isolated.
After screening a limited number of clones with this approach, we have been able to isolate a novel gene, which shows strong although not exclusive expression in osteogenic (but not chondrogenic) tissue, and belongs to a previously not described multigene family.

Mandibular Condyle Organ Cultures/Cell Culture-Pregnant
Balb/c mice were anesthetized with ether at 18.5 days post coitum and sacrificed by cervical dislocation. Fetuses were removed and cooled on ice. Fetal mandibular condyles were dissected and cleaned from adherent soft tissues. Extra care was taken to remove traces of mandibular or endochondral bone. Condyles were then transferred onto collagen sponges (Porous Collagen Matrices M1531, American Biomaterials Corp., Plainsboro, NJ) which were placed on stainless steel grids at the medium (BGJ/b with 10% fetal bovine serum)/gas (95% air, 5% CO 2 ) interphase of a 37°C incubator (at 100% humidity). Six to eight condyles were cultured per sponge in six-well plates. To construct the t ϭ 1 day cDNA library, approximately 750 condyles were placed in culture, and batches of about 100 condyles were harvested at 1-h intervals between 18 and 24 h and subsequently pooled. Freshly isolated or in vitro cultured condyles were immediately frozen in liquid nitrogen until RNA extraction. Balb/3T3 cells were grown in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum until subconfluency, detached with trypsin-EDTA (Life Technologies, Inc.), and stored as a pellet in liquid nitrogen.
RNA Preparation-RNA was extracted from condyles or from Balb/ 3T3 cell pellets according to the procedure of Chirgwin et al. (28). About 0.25 g of total RNA was obtained per condyle. Poly(A) ϩ RNA was prepared by two sequential rounds of affinity selection using oligo(dT) 25 Dynabeads TM (Dynal A.S., Oslo, Norway), according to the protocol recommended by the supplier.
Construction of the Primary cDNA Libraries-Directional cDNA libraries of in vitro cultured condyles (t ϭ 1 day) and the Balb/3T3 cell line were constructed using the Superscript TM Plasmid system (Life Technologies, Inc.). cDNA was size-fractionated on a Sephacryl S-500 HR drip column, and fragments larger than 500 bp 1 were ligated into NotI/SalI-digested pSPORT1 plasmid and electroporated into Escherichia coli SURE TM (Stratagene) bacteria. For the Balb/3T3 and condyle t ϭ 1 day cDNA libraries, approximately 1.3 ϫ 10 6 and 2.3 ϫ 10 7 independent colonies were collected, respectively. Plasmid DNA was prepared from the pooled bacteria using Qiagen TM columns (Qiagen GmbH, Germany). Insert sizes of 20 -25 random clones from each library were determined. The average insert sizes were 0.9 and 1.8 kb for the Balb/3T3 and condyle t ϭ 1 day libraries, respectively. The number of empty clones, as assessed by blue/white screening in the presence of isopropyl-1-thio-␤-D-galactopyranoside and 5-bromo-4-chloro-3-indoyl ␤-D-galactoside, was only a few percent in both libraries.
Subcloning of the Condyle t ϭ 1 Day cDNA Library-Plasmid DNA from the condyle t ϭ 1 day pSPORT1 library was NotI/SalI-digested and separated on a preparative 0.9% agarose gel. The cDNA smear between 0.5 and 4 kb was purified from the gel using the Geneclean II TM kit (BIO 101 Inc., La Jolla, CA). This cDNA was then subcloned into two different vectors (see Fig. 1). (i) pGEM9Zf Ϫ : purified cDNA was ligated into NotI/SalI-digested pGEM9Zf Ϫ (Promega, Madison, WI) and electroporated into E. coli DH12S (Life Technologies, Inc.). The subcloned pGEM9Zf Ϫ library contained 1.8 ϫ 10 7 independent clones, 91% of which contained an insert. Plasmid DNA was prepared from an aliquot of the collected bacteria. (ii) gt10: for subcloning into the gt10 vector, purified DNA fragments were first blunted with T4 DNA polymerase, ligated to EcoRI adaptors using the RiboClone TM EcoRI adaptor ligation system (Promega) and finally digested with HindIII. This digestion served to render contaminating NotI/SalI linearized pSPORT1 vector molecules unclonable at the expense of a small proportion of the insert fragments which happen to contain an internal HindIII site. Inserts were then cloned into gt10 EcoRI arms. The gt10 library consisted of 50,000 independent clones, 75% of which contained an insert. Preparation of Driver and Target cRNA-Driver and target cRNA were generated by in vitro transcription using the Megascript TM system for large scale RNA synthesis (Ambion, Austin, TX) according to the instructions recommended by the supplier with the exception that spermidine was omitted from the in vitro transcription buffer (it was observed that spermidine caused the cRNA to aggregate). To generate biotinylated driver cRNA (bio-cRNA), UTP (7.5 mM) was replaced with a mixture of UTP (5 mM) and bio-21-UTP (2.5 mM; Clontech). A trace of [␣-32 P]CTP was included in the reaction to allow accurate quantitation of the cRNA produced. The cRNA was separated from free nucleotides by spin chromatography on Sephadex G-50 (Pharmacia Biotech Inc.) in the presence of SDS (0.5%) and stored at Ϫ70°C until needed. Two heterologous cDNA clones (Plasmodium chabaudi clones 100 and 110) were subcloned into pGEM9Zf Ϫ , and sense cRNA was produced according to the same protocol.
Synthesis of Target First Strand cDNA-5 g cRNA in vitro transcribed from the condyle t ϭ 1 day cDNA libraries (in pSPORT1 for the initial experiments and later in pGEM9Zf Ϫ ) were annealed (10 min at 70°C) to 1 g of random primers (mostly hexamers; Life Technologies, Inc.) and then reverse transcribed for 1 h at 37°C with 1,000 units of Superscript TM (Life Technologies, Inc.) reverse transcriptase in a total volume of 20 l (50 mM Tris-HCl, pH 8.3, 75 mM KCl, 3 mM MgCl 2 , 10 mM dithiothreitol, 500 M each dNTP). 1 Ci of [␣-32 P]dCTP (Amersham Corp., specific activity 3,000 Ci/mmol) was added to the reaction to allow accurate quantification. The reaction was stopped by adding EDTA to a final concentration of 50 mM, and the cRNA template was hydrolyzed by incubation in 0.2 N NaOH (20 min at 65°C). The solution was neutralized by adding 20 l of 1 M Tris, pH 7.5, and 20 l of 1 N HCl, and the cDNA was purified by spin chromatography on Sephadex G-50. First strand cDNA of P. chabaudi clones 100 and 110 was produced from in vitro transcribed cRNA according to the same protocol. These clones served as a positive control in the subtractive hybridization experiments as they should not be subtracted by the Balb/3T3 driver cRNA.
Subtractive Hybridizations-For subtractive hybridization reactions 2-4 g of first strand cDNA, generated from the t ϭ 1 day cDNA library as described above, were spiked with 1-5 ng each of P. chabaudi clones 100 and 110 first strand cDNA (see Fig. 1). A fraction of this cDNA served as unsubtracted probe while the rest was mixed with 100 g of biotinylated Balb/3T3 driver cRNA and equilibrated in 10 mM Hepes, pH 7.5, 0.4 mM EDTA, 0.04% SDS by ultrafiltration in a microcon 100 (Amicon) concentrator to a final volume of 100 l. This solution was then further concentrated to a volume of 18 l by partial lyophilization and finally 2 l of 5 M NaCl were added. The reaction mixture was overlaid with mineral oil, heated for 5 min in a boiling water bath, and immediately transferred to a 65°C waterbath for 15-24 h (achieving a Rot of 800-1200 mol L Ϫ1 s). The final composition of the hybridization reaction was 50 mM Hepes, pH 7.5, 2 mM EDTA, 0.2% SDS, 0.5 M NaCl (ϭhybridization buffer), and 5 g/l driver bio-cRNA. After the hybridization was completed, 80 l of hybridization buffer without SDS were added to the reaction. The mineral oil was removed by CHCl 3 extraction. Subtractions were done by incubating the hybridization mix with 20 g of streptavidin for 15 min at room temperature and then extracting with phenol:CHl 3 (1:1) (29). Subtractions were repeated until no more counts could be detected in the organic phase (usually three to four rounds). The concentration of the remaining cDNA was determined by scintillation counting.
Differential Screening-Three sequential plaque lifts onto Hybond N ϩ membranes (Amersham) were taken. The first lift was discarded, whereas the second and third lift were hybridized to the unsubtracted and subtracted probe, respectively. Preliminary experiments had shown that there was virtually no difference in signal intensity of individual plaques between lifts 2 and 3 when these were hybridized to the same (complex) probe. Transferred DNA was UV-cross-linked, and membranes were hybridized for 24 h to 25 ng of random primed labeled subtracted or unsubtracted cDNA in 5 ml of Rapid Hyb TM solution (Stratagene) at 65°C. Specific activities of subtracted and unsubtracted probes did usually not differ from each other by more than a few percent and were usually greater than 10 9 dpm/g. After hybridization, filters were washed two times for 15 min in 2 ϫ SSC, 0.5% SDS at room temperature and finally two times for 1 h in 0.1 ϫ SSC, 0.5% SDS at 65°C. Differentially hybridizing plaques were isolated, replated, and subjected to a second round of differential screening.
Northern Blot Analysis and 5Ј-RACE-RNA samples were separated on denaturing formaldehyde gels, blotted, and hybridized according to standard procedures (30). Blots were washed to a final stringency of 1 ϫ SSC at 42°C.
Determination of the 5Ј-end of the E25 transcript was done using the 5Ј-RACE system (Life Technologies, Inc.), using the nested E25 antisense primers complementary to positions 1397-1416 and 1099 -1118, respectively (see Fig. 3).
In Situ Hybridization (ISH)-Sense and antisense digoxigenin (DIG)-labeled cRNA probes were generated by in vitro transcription from an E25 subclone containing a BamHI/EcoRI restriction fragment (positions 435-1603 in Fig. 3) in the presence of 10 mM CTP, GTP, UTP, and DIG-11-UTP. Eight-m sections were cut from frozen newborn mice or their dissected mandibular condyles, spread on poly-L-lysinecoated glass slides, and dried on a 50°C hot plate. Sections were then fixed in freshly prepared 4% paraformaldehyde in phosphate-buffered saline for 30 min at room temperature and dehydrated by passage through graded ethanol/H 2 O baths. After drying at room temperature, slides were stored at Ϫ70°C. When needed, sections were briefly thawed and serially treated with phosphate-buffered saline, 0.1% Triton X-100 (5 min; room temperature), 0.2 N HCl (20 min; room temperature), H 2 O and 2 ϫ SSC (30 min at 70°C). The tissue was subsequently deproteinized by incubation with 2 g/ml proteinase K (in 20 mM Tris-HCl, pH 7.4, 2 mM CaCl 2 ; 15 min at 37°C), followed by treatment with 2 mg/ml glycine (5 min at room temperature). After postfixation in 4% paraformaldehyde in phosphate-buffered saline for 15 min at room temperature, and acetylation (10 min incubation in 0.1 M triethanolamine, pH 8.0, containing 0.25% acetic anhydride), the sections were rinsed with H 2 O and dried. Tissues were prehybridized for 3-4 h at 42°C with 100 l of preheated hybridization solution (50% formamide, 4 ϫ SSC, 10% dextran sulfate, 1 ϫ Denhardt's solution, 0.5% SDS, and 500 g/ml herring sperm DNA). The sections were then hybridized for 16 -24 h at 42°C in 100 l of hybridization buffer containing a heat-denatured sense or antisense DIG-labeled cRNA probe at a final concentration of 0.5 ng/l. Sections were then briefly washed with 2 ϫ SSC, 0.1% SDS (30 min at 42°C), 0.1 ϫ SSC, 0.1% SDS (30 min at 42°C), and twice with 2 ϫ SSC at room temperature. Next they were treated with 2 ϫ SSC containing 10 g/ml RNase A (15 min at 37°C) and briefly rinsed in 2 ϫ SSC. Hybridized cRNA probes were immunostained with alkaline phosphatase-conjugated anti-DIG antibodies using the Boehringer Mannheim DIG immunodetection kit according to the protocol recommended by the manufacturer. Sections were not counterstained.
Chromosomal Mapping-We have found a BglII restriction fragment length polymorphism for the E25 probe between the murine strains C57BL/6 and SEG. We have used this restriction fragment length polymorphism to score 29 backcross progeny between the C57BL/6 and SEG strains (32). The segregating polymorphisms were analyzed with the help of GENE-LINK (33). Fig. 1 illustrates the subtractive hybridization/differential screening approach that was used to isolate transcripts, specifically expressed in 1-day-old explants of mouse mandibular condyles. Our initial screening strategy relied on differentially screening the condyle t ϭ 1 day pSPORT1 library with unsubtracted and subtracted (with Balb/3T3 bio-cRNA) probes derived from the same library. It was discovered, however, that the complex probes exhibited strong hybridization to the pSPORT1 vector (and other related plasmids), probably as a result of the presence of pSPORT1derived sequences in these probes. These contaminating vector sequences most probably originated from a small fraction of undigested template plasmid molecules during the in vitro transcription reaction, resulting in readthrough of the T7 polymerase.

Establishment of a Novel Subtractive Hybridization/Differential Screening Protocol-
Since radiolabeled pSPORT1 DNA did not hybridize to phage gt10 DNA, the condyle t ϭ 1 day library was subcloned into the gt10 vector. We also realized that the presence of the 60-nucleotide long vector-derived sequence at the 5Ј end of all in vitro transcripts from the pSPORT1 cDNA library posed the problem of aspecific subtraction as this sequence would be common to driver and target. We therefore simultaneously subcloned the condyle t ϭ 1 day library from the pSPORT1 into the pGEM9Zf Ϫ vector in order to generate an improved target cDNA library (in pGEM9Zf Ϫ the start of T7 transcription occurs only 10 nucleotides in front of the cDNA insert). We preferred to subclone the existing library rather than constructing a new library in the appropriate vectors because of the preciousness of the starting material and because we first wanted to demonstrate the feasibility of the approach.
Using this modified strategy, subtracted probe was prepared by subtracting cDNA from the condyle t ϭ 1 day pGEM9Zf Ϫ library with biotinylated cRNA generated from the Balb/3T3 pSPORT1 library. The quality of the subtracted probes was evaluated by differential hybridization on Southern blots containing a number of test sequences. Bands containing the test sequences were cut from the hybridized blots and quantitated via scintillation counting. Table I shows that the subtractive hybridization was specific. Sequences that are clearly known to be expressed in both fibroblastic and condylar tissue (e.g. collagen I, osteonectin, glyceraldehyde-3-phosphate dehydrogenase (GAPDH), ␤-actin) showed decreased hybridization with the subtracted probe, indicating that they had indeed been partially removed. In contrast, sequences that are known to be specific for the chondrogenic tissue (collagen II, collagen IX) or exogenously added unrelated DNA sequences (P. chabaudi clones 100 and 110) showed increased hybridization signals with the subtracted probe and thus had been selectively enriched. For transforming growth factor-␤, the difference in hybridization signal between subtracted and unsubtracted probe was less pronounced than for the other test sequences. The pattern illustrated in Table I was consistently observed in several independent subtraction experiments, illustrating the reliability and reproducibility of the subtraction procedure.
Differential Screening of the Condyle t ϭ 1 Day cDNA Library-Approximately 5,000 plaques of the condyle gt10 t ϭ 1 day cDNA library were screened with both unsubtracted and subtracted probes on duplicate plaque lifts. 25 plaques (E1-25) that hybridized more strongly to the subtracted than to the unsubtracted probe were isolated. 10 clones could be reconfirmed after a second round of differential screening and were partially sequenced. Comparison of the obtained sequence information with DNA databanks showed that all sequences except E22 and E25 had been described previously. The known sequences consisted of H19 (three clones), collagen II (three clones), bone sialoprotein (one clone), and placental alkaline phosphatase (one clone). Although clone E22 did not match any sequence present in DNA databanks, its high degree of similarity to human collagen IX suggested that it encoded mouse collagen IX. The presence of collagen II, collagen IX, and bone sialoprotein among the clones that were enriched after condyle minus Balb/3T3 subtraction is not unexpected as all of these genes are known to be expressed in cartilagenous tissue such as the condyle and not in the Balb/3T3 cell line (34,35). In addition we confirmed by Northern blot analysis that the H19 transcript was indeed expressed in condylar RNA (1-day explants) but was absent in Balb/3T3 RNA (data not shown). Likewise, a radiolabeled E25 probe strongly hybridized to a band of Ϯ1.8 kb on condyle RNA but did not hybridize to Balb/3T3 RNA ( Fig. 2A). It was thus clear that all clones isolated after differential screening were present in the target tissue and absent in the Balb/3T3 driver population, further demonstrating the reliability of the subtractive screening procedure. We decided to characterize the novel clone E25 in greater detail.
Northern Blot Analysis-Northern blot analysis of different mouse neonatal tissues showed very strong expression of E25 in calvaria and slightly weaker expression levels in front paws, skin, and tail (Fig. 2B). Expression in brain and heart was very weak, whereas there was no detectable transcript in liver and lung.
Analysis of the E25 Primary Sequence-Two additional clones were isolated from the condyle pSPORT1 cDNA library, which extended the known E25 sequence by 200 bp at the 5Ј end. 5Ј-RACE analysis on mRNA from freshly isolated condyles led to the isolation of five more clones, again extending the E25 sequence at the 5Ј end with 50 bases. The composite nucleotide sequence obtained from these overlapping clones was 1654 bp in length and contained a polyadenylation signal followed by a poly(A) tail at its 3Ј end (Fig. 3). The cDNA contained a single long open reading frame (nucleotides 1-939). The presumed initiator methionine occurred at position 151-153. Several arguments underscored this view. (i) All five clones obtained by 5Ј-RACE analysis stopped within a window of 5 base pairs at their 5Ј end, suggesting that the reverse transcriptase had indeed reached the mRNA start point. (ii) The size of the obtained E25 cDNA was very close to the size of the transcript The complex probe consisted of first strand cDNA which was generated from the condyle t ϭ 1 day pGEM9Zf Ϫ library and "doped" with heterologous first strand cDNA (P. chabaudi clones 100 and 110). Part of this probe was used as the unsubtracted control, whereas the rest was subtracted with an excess of Balb/3T3 bio-cRNA. The Southern blots contained different recombinant plasmids which had been restriction digested to release the inserts. Hybridized insert bands were cut from the blot and quantitated via scintillation counting. For each test insert the ratio of the hybridization signal obtained with the subtracted probe (average of two Southern blots using the same probe preparation) to that of the unsubtracted probe is indicated.
b Transforming growth factor-␤. observed on Northern blot (Ϯ1.8 kb; see Fig. 2A). (iii) The presumed initiator codon was within a perfect Kozak consensus context (ACCATGG) (36). (iv) Inspection of four EST sequences (R01594, R50284, T93146, and R57511) from the human E25 equivalent (see below) that are available in this region shows that their consensus sequence contains an in-frame stop codon in front of the initiator codon in question. It is therefore likely that the coding sequence starts at position 151. This would produce a protein of 263 amino acids with a M r of 30,000. Hydropathy analysis of the predicted E25 protein sequence according to the Kyte and Doolittle (37) and Klein et al. (38) algorithms (SOAP, PCGENE, Intelligenetics) demonstrated the presence of a single membrane spanning domain at amino acid positions 52-76. This suggested that E25 was a type II integral membrane protein (39). There was a single potential N-glycosylation site (amino acid 166). Apart from a potential leucine zipper motif, motif analysis did not reveal any structural features of known protein families. Overall the E25 protein has a high content of leucine (10.2%) and isoleucine residues (8.0%). The E25 mRNA has a rather long 3Ј-untranslated region of Ϯ700 bp, containing five "ATTTA" sequence motifs, which have been shown to mediate RNA destablization in certain lymphokine and immediate early genes (40,41).
In Situ Hybridization Experiments-In order to study expression of E25 mRNA in more detail, in situ hybridization experiments were carried out on frozen sections of neonatal mice (Fig. 4). In the neonatal mouse mandibular condyle strong ISH signals were observed only in the outermost layers, corresponding to the perichondrial and subperichondrial region. There was a marked absence in the deeper layers of the condyle, which contain the more differentiated chondroblasts and chondrocytes. This could not be due to an artefactual loss of mRNA from the deeper layers of the condyle as a result of section preparation, because under the same conditions, we were able to obtain strong ISH signals throughout the condylar tissue with another clone derived from the mandibular condyle. 2 Prominent ISH signals were also detected on the trabeculae of various bones including the ossified parts of ribs. In contrast, E25 ISH signals were markedly absent in the cartilagenous parts of ribs, vertebrae, and long bones. Interestingly, strong expression of E25 was also observed in mature odontoblasts of germinating teeth but not in the the less differentiated mesenchyme of the dental papilla. Hair follicles and stratum corneum of the skin also stained strongly positive. E25 was strongly expressed in the acini of several exocrine glands such as parotid gland, lacrimal gland, and seromucous glands in nose mucosa. Finally weak ISH signals were found in the brain (choroid plexus), renal cortex, and in the crypts of the small intestine. Striated muscle, liver, and lung showed no E25 expression. Sense probes were completely negative on all sections (data not shown).
Establishment of the E25 Multigene Family-Homology searches of DNA databanks using the BLAST algorithm showed that more than 80 ESTs of human and murine origin had been deposited, of which the open reading frame showed significant homology to the E25 protein sequence. Detailed analysis of these ESTs revealed that E25 belongs to a multigene family which contains at least three members (Fig. 5). We will refer to the mouse gene isolated in this study as E25AMM (E25A Mus musculus). Via alignment of a subset of the human ESTs (nine sequences) we were able to obtain the complete coding sequence of the human equivalent of E25AMM, which we named E25AHS (E25A Homo sapiens). The amino acid sequences of E25AMM and E25AHS are 94% identical. The majority of the human ESTs (63 sequences) could be aligned to yield a full-length protein sequence, which was 38% identical to E25AHS and was named E25BHS. The deduced amino acid sequence of murine EST L26775 was nearly identical to E25BHS (93%) suggesting that it represents the murine counterpart, which we named E25BMM. Finally nine human EST sequences could be aligned to yield a partial protein sequence (131 amino acids) which was related to but distinct from the E25AHS and E25BHS sequences. We refer to this third variant of E25 as E25CHS. The deduced assembled amino acid sequence of murine ESTs R75511, R74770, and R75509 was 91% identical to the E25CHS partial amino acid sequence, suggesting that these ESTs encode the murine counterpart of E25CHS, referred to as E25CMM.
Conservation was less pronounced between different isoforms within the same species. For instance E25AHS was 38% identical to E25BHS and 49% to E25CHS (it should be noted that in contrast to E25AHS and E25BHS, the E25CHS sequence information was only partial). The ESTs representing E25AHS were derived from fetal brain, heart, and liver/spleen, 2 G. Hong, unpublished observations. The putative transmembrane domain is marked with crosses, whereas the single N-linked glycosylation site is indicated by *. The (iso)leucine residues in the potential leucine zipper motif are in bold and underlined. The "ATTTA" sequence motifs in the 3Ј-untranslated region have been underlined, and the polyadenylation signal is in bold. and from adult lung and breast. E25BHS ESTs originated from a variety of tissues including fetal brain and liver/spleen and from aorta, placenta, lung, prostate, adrenal gland, white blood cells, and endothelial cells. E25CHS ESTs were derived from brain, placenta, and fetal liver/spleen.
Chromosomal Mapping-The radiolabeled E25AMM probe hybridized with a 4.5-kb BglII DNA fragment from strain C57BL/6 and a 4.3-kb BglII DNA fragment from the SEG strain. This size polymorphism was used to localize the E25AMM locus (which we named Itm2, integral membrane protein 2, as a successor to Itm1 the gene for another integral membrane protein, that we characterized previously) (42) by linkage analysis. We found that all the DNA samples derived from the male offspring of the backcross were all of one parental type, either C57BL/6 or SEG. Such an observation indicated X-linkage. Checking the allellic distribution of E25 for the 18 females, we found no recombinant with two X-linked loci: DXMit8 and Xist. DISCUSSION We have used differential cDNA screening of organ cultures of prenatal mouse mandibular condyles to identify and isolate a gene encoding a novel integral membrane protein, E25AMM, that is differentially expressed in bone and cartilage tissue and belongs to a multigene family.
We developed an alternative subtractive hybridization/differential screening protocol, which allowed us to directly use library derived subtracted cDNA as a probe in differential screening, a strategy which to our knowledge has not been described so far in the literature. We found that the major difficulty in using library derived cDNA as a probe for differential screening, was that cDNA probes, which are generated from libraries by sequential in vitro transcription and reverse transcription are contaminated with vector derived sequences. We solved this problem by generating subtracted probes from a plasmid library and using these probes in the differential screening of a cDNA library made in the unrelated gt10 vector.
The reliability of the subtractive hybridization protocol was investigated. We found that cDNAs known to be expressed in both condylar tissue and fibroblasts (GAPDH, ␤-actin, and ␣1(I) collagen) were indeed subtracted. Quantitative analysis showed that the hybridization signal of the subtracted genes in the subtracted probe was around 30% of that in the unsubtracted probe. It should, however, be noted that equal amounts of both subtracted and unsubtracted cDNA were used in the differential screening procedure. Because one round of subtraction typically removed 60 -80% of the cDNA, the hybridization signal of subtracted species was artificially increased. When one corrects for these effects, it becomes clear that over 90% of each of the above-mentioned cDNAs has been removed. On the other hand, genes which are known to be specifically expressed in neonatal mouse condyles such as the cartilage specific ␣1(II) and ␣1(IX) collagen cDNAs as well as the two exogenously added P. chabaudi cDNA species had an increased abundance in the subtracted probe. The increase in abundance ranged from a factor 2.0 (P. chabaudi clone 100) to 5.16 (␣1(IX) collagen). Theoretically the maximal increase in abundance is dictated by the total amount of cDNA that can be subtracted and thus by the degree of similarity between the two RNA populations. The fact that typically 60 -80% of the cDNA is removed after subtraction is in agreement with the experimentally observed increase in hybridization intensity of unsubtracted FIG. 4. In situ hybridization of E25 expression in different tissues of neonatal mice. A, isolated mandibular condyle of newborn mice. E25 is strongly expressed in the outer apical and lateral layers of the condyle (thick arrow). Expression fades toward the deeper layers. E25 is also expressed in the articular disc (arrowhead) and in the area of the mandibular bone (thin arrow). B, section through rib. There is intense staining in the ossified region of the rib (arrow), but no staining of the rib cartilage (arrowhead). C, developing tooth germs. There is strong E25 expression in the area of mature odontoblasts (long thin arrow) but not in the less differentiated mesenchyme of the dental papilla (arrowhead). Also note the strong staining of the bone trabeculae of the jaw (short thick arrow). D, section through skin. Hair follicles (thick arrow) as well as stratum corneum (thick arrow) stain strongly. cDNA species. It is not clear why some cDNAs are more enriched than other ones. Possibly there is some degree of aspecific subtraction of cDNA based on accidental low level homology between cRNA in the driver and some target cDNA molecules. The enrichment factor we observe is comparable to that reported in other systems (43)(44)(45)(46). Although substantially higher enrichment factors have been reported by some authors it is not always clear whether these results had been corrected for yield losses during the subtractive hybridization procedure or whether other artefacts could explain the high subtraction efficiencies, as remarked in a recent review (47).
The identity of the 10 differential clones that were isolated further demonstrated the specific character of the subtractive hybridization/differential screening protocol. Bone sialoprotein, collagen II, and collagen IX are indeed known to be specifically expressed in the cartilagenous tissue of the mandibular condyle but not in Balb/3T3 fibroblasts. Furthermore, Northern blot analysis showed that the H19 and E25 transcripts were completely absent in Balb/3T3 fibroblasts, but strongly expressed in mandibular condyles. The H19 transcript is a noncoding RNA molecule that is expressed in many fetal tissues and which was shown to have tumor suppressor activity (48).
An intriguing finding was the presence of placental alkaline phosphatase in mandibular condyles. Whereas a potential contamination of the fetal condyle tissue with placental material during isolation, although unlikely, cannot be excluded, these results raise the possibility that fetal condyles express the placental isotype of alkaline phosphatase instead of the liver/ bone/kidney isotype which is generally found in cartilagenous tissue.
The aim of the current experiments was to identify novel markers of osteogenic differentiation. Our results indicate that E25 could be a useful marker in this respect. In situ hybridiza-tion experiments showed that E25 expression in neonatal mandibular condyles was clearly limited to the outer rim of the condyle, which contains the prechondroblastic mesenchymal cells. All other cartilagenous tissues that were examined by in situ hybridization (ribs, long bones, vertebrae) were negative for E25. By contrast there was strong staining of bone trabeculae in mature bone. A strong association of E25 with osteogenic tissue was also suggested by the Northern blot analysis of various neonatal mouse tissues. By far the strongest expression was observed in calvaria and to a lesser extent also in front paws and tail. These tissues, especially calvaria, are actively involved in osteogenesis. With the exception of skin, expression in nonosteogenic tissues was weak (brain and heart) or absent (lung and liver).
We hypothesize that expression of the E25 marker in the precursor cells of the mandibular condyle reflects the ontogeny of this secondary cartilage. Avian secondary cartilages develop from the periosteal cells of membrane bones (49), whereas mammalian mandibular condyles originate as a separate cellular blastema and later associate with the membranous bone of the mandible (50,51). The initial blastemal type of development is thus succeeded by a periosteal/perichondrial type of development. The E25 marker which we showed to be strongly expressed in mature osteoblasts would thus be gradually turned off in the differentiating periosteum/perichondrium of the condyle during secondary chondrogenesis. This would also explain why the perichondrium of primary cartilages (e.g. rib) is negative, as it does not differentiate from membranous bone.
Although E25 expression is strongly associated with mature osteoblasts and the early stages of (secondary) chondrogenesis in the mandibular condyle it appears not to be uniquely restricted to these tissues. This is shown by the weak Northern blot signals that were detected in (fetal) heart and brain and the relatively strong signal in skin. Strong in situ hybridization signals were indeed found in hair follicles and stratum corneum of the skin. The acini of several exocrine glands also exhibited strong ISH staining. Weaker ISH signals were also detected in brain, renal cortex and in the small intestine.
It is unlikely that the positive ISH signals are due to crosshybridization to other members of the E25 family. Although complete sequence information is not available on all three mouse variants, it is clear that even in the more conserved C-terminal part the difference in sequence between different members is large enough for the melting temperature of crosshybrids to be substantially lower than that of the specific hybrid. Also if there was some cross-hybridization, this would have been detected on the Northern blots, where additional bands or at least a more diffuse band would have been present for the embryonic tissues that were investigated. However, only a single discrete band was present on the Northern blots (Fig. 2).
The E25AMM predicted amino acid sequence contains a single putative transmembrane domain between amino acids 52-76 and no amino-terminal signal peptide, suggesting that it is a type II integral membrane protein with the C-terminal part being extracellular. This topology would fit with the position of the sole potential N-glycosylation site (Asn-166). The significance of the leucine zipper motif is unclear. Whereas leucine zipper motifs are well known as protein-protein interaction domains in transcription factors, they presumably also mediate multimerization in other classes of proteins. Alternatively it could be argued that the observed leucine zipper is purely accidental, due to the high leucine/isoleucine content of the E25 protein.
DNA databank searches yielded a multitude of human and murine EST sequences, which were homologous to the FIG. 5. Alignment of human and murine members of the E25 multigene family. Amino acid residues that are identical in at least four sequences are indicated by black boxes (conservative substitutions by gray boxes). The asterisks (*) in the E25CHS sequence denote a gap due to the fact that no overlapping EST sequence was available in this area. Dots denote gaps introduced to maximize homology. Amino acid residue no. 58 of E25BHS could not be unambiguously identified in the different ESTs (denoted X). E25AMM sequence. From these sequences a E25 multigene family, containing three members both in mouse and man, emerged. Individual family members exhibited a very high degree of conservation between man and mouse, whereas the homology between the different family members was around 40% within the same species. All three members appear to be integral membrane proteins since they all contain a single transmembrane domain (roughly between amino acid 50 and 70). Individual members show the highest degree of conservation in the C-terminal half of the protein, corresponding to the extracellular part of E25. They do not show any significant degree of conservation in the amino-terminal (presumedly cytoplasmic) part. The leucine zipper motif present in E25A (both in man and mouse) was not retained in the E25B and E25C members. Like in E25A, no distinct motifs of known protein families could be detected in the available parts of the E25B and E25C protein sequences.
Judging from the origin of the cDNA libraries from which E25 related ESTs have been isolated, members of the E25 gene family are expressed in a wide variety of tissues and some tissues appear to be capable of expressing all three members (e.g. brain). Whether the E25B and E25C family members are also associated with osteogenic/chondrogenic differentiation will need further experimentation.