3677Journal of Cell Science 108, 3677-3684 (1995)Printed in Great Britain © The Company of Biologists Limited 1995JCS4041

A 182 bp fragment of the mouse proα1(II) collagen gene is sufficient to direct

chondrocyte expression in transgenic mice

Guang Zhou, Silvio Garofalo*, Krish Mukhopadhyay, Véronique Lefebvre, Chad N. Smith,Heidi Eberspaecher and Benoit de Crombrugghe†

Department of Molecular Genetics, The University of Texas, M. D. Anderson Cancer Center, 1515 Holcombe Boulevard, Houston,Texas 77030, USA

*Present address: Shriner’s Hospital for Crippled Children, Research Unit, 3101 S.W. Sam Jackson Park Road, Portland, Oregon 97201, USA†Author for correspondence

Type II collagen is a major chondrocyte-specificcomponent of the cartilage extracellular matrix and it rep-resents a typical differentiation marker of mature chon-drocytes. In order to delineate cis-acting elements of themouse proα1(II) collagen gene that control chondrocyte-specific expression in intact mouse embryos, we generatedtransgenic mice harboring chimeric constructions in whichvarying lengths of the promoter and intron 1 sequenceswere linked to a β-galactosidase reporter gene. A con-struction containing a 3,000 bp promoter and a 3,020 bpintron 1 fragment directed high levels of β-galactosidaseexpression specifically to chondrocytes. Expression of thetransgene coincided with the temporal expression of theendogenous gene at all stages of embryonic development.Successive deletions of intron 1 delineated a 182 bpfragment which targeted β-galactosidase expression tochondrocytes with the same specificity as the larger intron1 fragment. Transgenic mice harboring a 309 bp Col2a1

promoter lacking intron 1 tester sequences showed no β-galactosidase expression in chondrocytes. Reduction of the182 bp fragment to a 73 bp subfragment surrounding adecamer sequence previously reported to be involved inchondrocyte specificity, resulted in loss of transgeneexpression in chondrocytes. When the Col2a1 promoterwas replaced with a minimal β-globin promoter, the 182 bpintron 1 sequence was still able to target expression of thetransgene to chondrocytes. We conclude that a 182 bpintron 1 DNA segment of the mouse Col2a1 gene containsthe necessary information to confer high-level, temporallycorrect, chondrocyte expression on a reporter gene inintact mouse embryos and that Col2a1 promoter sequencesare dispensable for chondrocyte expression.

Key words: proα1(II) collagen gene, chondrocyte, cartilage,transgenic mouse

SUMMARY

INTRODUCTION

Type II collagen is a major chondrocyte-specific component ofthe cartilage extracellular matrix (Stockwell, 1979; Upholt,1989). Its structural role in cartilage is most clearly demon-strated by the abnormal skeletal phenotypes displayed byhumans and mice carrying mutant proα1(II) collagen chains(Lee et al., 1989; Garofalo et al., 1991). During embryonicdevelopment, the gene for proα1(II) collagen becomesactivated concomitantly with the onset of chondrocyte differ-entiation. This gene is also expressed transiently and at lowlevels in certain non-chondrogenic tissues, includingnotochord, heart, epidermis and discrete areas of the brain(Cheah et al., 1991), but the function of the protein in thesesites is not known. Other studies have identified an isoform ofproα1(II) collagen, which includes sequences encoded by analternatively spliced exon 2. This isoform was reported to beexpressed preferentially in non-chondrogenic tissues (Sandellet al., 1991).

Earlier transient expression experiments in primary chon-drocytes identified a chondrocyte-specific enhancer element ina 620 bp intron 1 sequence of the rat Col2a1 gene (Horton etal., 1987). This element was later reduced to a 260 bp segmentwhich produced a 6-fold enhancement in promoter activity(Wang et al., 1991). In other experiments two silencer elementswere identified in the promoter region of the rat gene between−360 and −460 and between −620 and −700, which reducedthe activity of a heterologous thymidine kinase promoter inHeLa cells in an orientation-independent fashion (Savagner etal., 1990). Since the activity of the intronic enhancer in primarychondrocytes was not affected by the silencer elements, theseelements were postulated to play a role in inhibiting expressionof the gene in cells in which the gene is inactive (Savagner etal., 1990). In these earlier transient expression studies primarychick embryo chondrocytes and other mesenchymal cells wereused since well-differentiated stable chondrocytic cell lineswere unavailable.

Two preliminary experiments on transgenic mice with

3678 G. Zhou and others

Col2a1 chimeric genes were performed earlier. In miceharboring a construction containing 3 kb of 5′ flankingsequences as well as the first intron of the rat Col2a1 genelinked to the chloramphenical acetyltransferase (CAT) reportergene, CAT activity was found specifically in extracts of tissuescontaining cartilage; deletion of intron 1 abolished expressionof the CAT gene in extracts of cartilage tissues (Yamada et al.,1990). Similarly, mice carrying a construction in which theEscherichia coli β-galactosidase (lacZ) reporter gene wasfused in frame to a segment of the mouse Col2a1 geneextending from 3 kb upstream of the transcription start to exon4, showed chondrocyte-specific X-gal staining. Here also,deletion of intron 1 inhibited expression of the transgene (Met-säranta et al., 1995).

The purpose of the present study was to delineate minimalsequences in the mouse Col2a1 gene that were needed forchondrocyte-specific expression in intact mice and todetermine whether sequences in the promoter were needed forcell-specific expression. We also asked whether the temporalexpression of the chimeric Col2a1 transgene paralleled that ofthe endogenous gene during embryonic development. Trans-genic mice constitute clearly the most physiological system tostudy cell-specific gene expression and the only system todelineate sequences that are capable of conferring a correcttemporal pattern of expression during embryonic development.Because the phenotype of primary chondrocytes in culture isunstable and because they can only be used in transientexpression experiments, we thought it was essential to usetransgenic mice to delineate unambiguously the segmentsneeded for chondrocyte-specific expression of the mouseCol2a1 gene. The advantage of the β-galactosidase reportergene over other reporter genes is that the histochemicalsubstrate for β-galactosidase, X-gal, allows precise histologi-cal identification of the cells in which the transgene is active.This aspect is important, as it is practically impossible toisolate pure cartilages from mouse embryos to measurereporter enzyme activities in tissue extracts.

Here we show that the only requirement for chondrocyteexpression in transgenic mice is a 182 bp intron 1 fragment ofthe mouse Col2a1 gene; this DNA element is also able to targetthe activity of a heterologous β-globin promoter to chondro-cytes.

MATERIALS AND METHODS

DNA constructions and nomenclature of plasmidsAll constructions are represented schematically in Fig. 1. They wereall generated in plasmid pSAβgeo (Friederich and Soriano, 1991)except i(182x4)pgloblacZ, which was generated in plasmid placF(Mercer et al., 1991). In placF the lacZ gene was followed by asequence of the mouse protamine gene, which supplies an intron anda polyadenylation signal.

Three different Col2a1 promoter fragments were used: p3000,p687, and p309. They contained 5′ promoter sequences of ~3 kb, 687bp, and 309 bp, respectively, as indicated by the number followingthe letter ‘p’, plus the 237 bp exon1 and the 70 5′ proximal bp ofintron 1. The Col2a1 translation initiation codon (ATG) was changed(to CTG) by PCR mutagenesis. The number following the letter ‘i’represents the size of intron 1 tester fragments. p309Col2a1 containsno intron 1 tester fragment. The restriction sites used to generate thepromoter and intron fragments are indicated in Fig. 1.

The 182 bp and 73 bp intron fragments were generated by PCRusing two oligonucleotide primers, one containing a SpeI recognitionsite and the other an XbaI site. In order to form multiple repeats, thePCR products were digested with SpeI and XbaI, self-ligated and re-digested with SpeI and XbaI to keep only multimers ligated head totail. Multimers were isolated by electrophoresis in agarose gel and thetandem repeat products were inserted into a SpeI site added by PCRin the vector immediately 3′ of the 70 bp intron 1 fragment.

To generate i(182x4)pgloblacZ, a minimal human β-globinpromoter from −44 to +28 (Lawn et al., 1980) was amplified by PCRusing two primers, one containing an XbaI site and the other a SalIsite. The PCR product was inserted between the XbaI and SalI sitesof placF. Four copies of the 182 bp fragment were cloned in the XbaIsite at the 5′ end of the β-globin promoter. The sequences of all DNAfragments generated by PCR were verified by dideoxy DNA sequenc-ing.

Generation and identification of transgenic mice DNAs were released from the plasmid vector by restriction enzymedigestion, which cleaved immediately upstream of the Col2a1 genesequence and downstream of the polyadenylation signal. The purifiedDNAs were redissolved at a concentration of 2 ng/µl in 5 mM Tris-HCl (pH 7.4), 0.1 mM EDTA and 5 mM NaCl, and were microin-jected into pronuclei of one-cell mouse embryos isolated fromB6D2F1 female mated with B6D2F1 males, and implanted into CD1pseudopregnant foster mothers (Hogan et al., 1986). In some instancestransgenic lines were established; otherwise transgenic founderembryos were sacrificed at day 14.5 post coitum (p.c.). Transgenicembryos were collected by caesarean section and stained by X-gal(Niederreither et al., 1992). Southern analysis (Southern, 1975) wasperformed by hybridizing BamHI-digested genomic DNA extractedfrom either placenta or tail with a 3 kb lacZ-specific probe (obtainedby digestion of pSAβgeo with BamHI).

Staining and histochemistrylacZ activity was detected by staining with X-gal (5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside; Boehringer-Mannheim) asdescribed (Mercer et al., 1991). Embryos were fixed for 30 minuteswith 0.2% glutaraldehyde, 5 mM EGTA, 2 mM MgCl2 and 2%formalin in 0.1 M sodium phosphate buffer at pH 7.3. They were thenwashed three times for 30 minutes each with a rinse solution con-taining 2 mM MgCl2, 0.2% NP-40 and 0.1% sodium deoxycholate in0.1 M sodium phosphate buffer at pH 7.3. Staining was performed for1-24 hours at room temperature in the rinse solution supplementedwith 1 mg/ml X-gal, 5 mM potassium ferrocyanide and 5 mMpotassium ferricyanide. After staining the embryos were washed withPBS and stored in 80% ethanol at −20°C. After photography embryoswere dehydrated, embedded in paraffin and used to generate 8 µmthick sections, which were counterstained with eosin.

RESULTS

Chimeric Col2a1 constructions used to generatetransgenic miceThe different constructions that were used in our experimentsare shown in Fig. 1. The βgeo reporter gene specifies a fusionpolypeptide which has both β-galactosidase and neomycinresistance activities. In these constructions, the βgeo gene ispreceded by a splice acceptor (SA) and followed by apolyadenylation signal. Varying lengths of 5′ flankingpromoter sequences of the mouse Col2a1 gene were linked toexon 1 (237 bp) and the 5′ end of intron 1 (70 bp), followedby varying lengths of intron 1 and the SA-βgeo bpA reportercassette. Given the arrangement of regulatory elements in these

3679A 182 bp Col2a1 enhancer

A

-687AAAAAAAAAAAAEcoRI

+2345 +2164182bp

DraISmaI

p687i182Col2a1

AAA-309

ApaI

+2345 +1878468 bp

BamHISmaI

p309i468Cola2a1

AAA-309

ApaI

+2164 +2345182bp (2x)p309i(182x2)Col2a1

-687AAAAAAAAAAAA

EcoRI

+2345 +1878468 bp

BamHISmaI

p687i468Col2a1

i(182x4)pgloblacZ

B

human β-globin promoter

lacZ-polyA

XbaI SalI-44 +28

182bp(4x) A

p3000i3020Col2a1

ScaI

ATG->CTG

NotI

-3000 +308

XbaI

SA βgeo bpA

Asp718

+3327exon 1AAAAAAAA

+227373

+2345AAAAAA-309

ApaI

(3x)p309i(73x3)Col2a1

p309i(116x2)Col2a1 116bp+2164 +2279

(2x)AAAAAA-309

ApaI

p309Col2a1 AAA-309

ApaI

Fig. 1. (A) Schematic representation of Col2a1 DNAconstructions used to generate transgenic mice. Hatched boxesrepresent Col2a1 promoter sequences and stippled boxesrepresent Col2a1 intron 1 tester sequences. The sizes of intron 1tester fragments and the position of their first and last nucleotidesrelative to the Col2a1 transcription start site are indicated. Eachconstruction contained in addition to tester promoter and intron 1sequences, exon 1 of the mouse Col2a1 gene, the 5′ 70 proximalbp of intron 1 of the mouse Col2a1 gene and the βgeo reportergene preceded by a splice acceptor (SA) sequence and followedby the bovine growth hormone gene polyadenylation signal(bpA). p309Col2a1 contains no tester intron 1 sequences. Intron1 fragments were cloned in their 5′ to 3′ natural orientation,except the 468 bp intron 1 fragment and the single copy of the182 bp fragment which were cloned in the opposite orientation.(B) Schematic representation of i(182x4)pgloblacZ fusion gene.Four tandem copies of the 182 bp Col2a1 intron 1 fragment from+2164 to +2345 were placed upstream of the human β-globinsequence from −44 to +28; this construction was linked to thelacZ gene followed by a polyadenylation signal.

constructions, which follows the order in which they appear inthe endogenous gene, the splice acceptor preceding the βgeogene is required for correct splicing of intron 1 RNA sequencesand lacZ expression. In the first seven constructions of Fig. 1,the translation initiation codon of the proα1(II) collagenpolypeptide was mutated from ATG to CTG to inhibit transla-tion initiation at this site and favor translation initiation at theAUG of βgeo RNA.

A segment of the mouse Col2a1 gene containing a 3kb promoter and a 3 kb intron 1 targets β-galactosidase activity in chondrocytesThree transgenic mouse lines harboring constructionp3000i3020Col2a1, which contained a 3 kb promoter and a 3,020

Table 1. Ratios of transgenic mouse embryos positive by Xanal

Ratios in transgenic Ratios in Constructions mouse lines founde

p300i3020Col2a1 3/4 –p687i468Col2a1 2/3 2/5p687i182Col2a1 – 2/7p309i468Col2a1 1/2 2/3p309i(182x2)Col2a1 – 4/8p309i(73x3)Col2a1 – 2/4p309i(116x2)Col2a1 – 0/4p309Col2a1 1/2 2/2

i(182x4)pgloblacZ 1/1 4/5

bp intron 1 fragment, showed an identical cartilage-specific X-gal staining pattern whereas a fourth line showed no X-galstaining (Table 1). In embryos of these three lines the earliest X-gal staining appeared at day 10.5 p.c. (Fig. 2A). Whole-mountembryos showed segmental X-gal staining in discrete dorsal areasof the trunk. Staining decreased in intensity in a rostral-to-caudaldirection. Histological analysis of this region showed strong X-gal staining in the notochord and in surrounding mesenchymalcells, which likely represent sclerotome cells in which the chon-drocytic program is just beginning to be activated (Fig. 2B).Staining was also observed in the first branchial arch, whichcontains precursors of the mandibular and maxillary skeleton,and more faintly in the second branchial arch, which containsprecursor cells for the hyoid and thyroid cartilages (Fig. 2A).

-gal staining versus embryos positive by genomic Southernysiskilled Pattern ofrs Total expression

3/4 All cartilages4/8 All cartilages2/7 All cartilages3/5 All cartilages4/8 All cartilages2/4 Some dermis and tongue0/4 No staining3/4 Some dermis, tongue,

and some tendons

5/6 All cartilages

3680 G. Zhou and others

Fig. 2. Cartilage-specific β-galactosidase expression intransgenic embryos from ap3000i3020Col2a1 line atdifferent stages ofdevelopment. (A) 10.5-day p.c.whole-mount embryo stainedwith X-gal. I, mandibular arch;II, second branchial arch.(B) Transverse section of thesame 10.5-day p. c. embryo. nt,neural tube; nc, notochord.(C) 12.5-day p.c. embryo.(D) Sagittal section of the same12.5-day p.c. embryo. ve,vertebra; in, intervertebralmesenchymal cells.(E-F) Lateral and dorsal viewsof a 14.5-day p.c. embryo.(G) Sagittal section of the same14.5-day p.c. embryo. Bars,100 µm.

Fig. 3. β-Galactosidaseexpression in transgenicembryos carrying shorterCol2a1 promoter and intron 1fragments. (A) Lateral view of14.5-day p.c. whole-mount X-gal-stained embryo carryingp687i468Col2a1. (B) Lateralview of 14.5-day p.c. embryoharboring p309i468Col2a1.(C) Lateral view of 14.5-dayp.c. transgenic embryos from ap309Col2a1 line. Only veryfaint staining of areas in thebrain are seen and no stainingof cartilages. (D) Sagittalsection of a 14.5-day p.c.founder embryo transgenic forp309Col2a1, showing X-galstaining of tongue (to) and ofdermis (dm), but not inchondrocytes (ch). (E) Lateralview of 14.5-day p.c. embryoharboring p309i(182x2)Col2a1. (F) Sagittal section ofthe same embryo as shown inE. X-gal staining is restricted tochondrocytes. Boxed area isrepresented at highermagnification in G. di, digit; hl,hindlimb; ve, vertebra. Bars: (Dand F), 100 µm; (G), 25 µm.

3681A 182 bp Col2a1 enhancer

At day 12.5 p.c., X-gal staining was more extensive (Fig. 2C).In the cephalic region, X-gal staining was observed in Meckel’scartilage, otic capsule and primitive nasal septum, and in thecartilage precursors at the lateral edges of the parietal and frontalbones. Staining was also evident at the base of the skull, whereprecartilaginous condensations first appear and form variouscomponents of the chondrocranium. X-gal staining was alsoobserved in the cartilage anlages of long bones in limbs, pelvicand shoulder girdles, and ribs. Parasagittal sections showed X-gal staining in skeletal structures derived from sclerotomes, i.e.in the vertebral body precursors and at somewhat lesser intensityin the zones where intervertebral discs were forming (Fig. 2D).Except for the notochord (data not shown), there was no X-galstaining in other non-chondrogenic tissues.

At day 14.5 p.c., X-gal staining was found exclusively in thecartilaginous primordia of all long bones in the limbs, ofvertebrae, ribs, scapula and ischia, of the exooccipital, basooc-cipital and mandibular bones, and in nasal cartilages (Fig.

Fig. 4. β-Galactosidase expression pattern inembryos containing the 182 bp intron 1 fragment ofCol2a1 and a heterologous promoter. (A) 14.5-dayp.c. whole-mount X-gal-stained transgenic embryocarrying i(182X4)pgloblacZ. Staining pattern isvery similar to the stage-matched transgenic embryocarrying p3000i3020 shown in Fig. 1E.(B-C) Sagittal sections of the same embryo showingstaining in chondrocytes but not in other tissues. ad,adrenal gland; ki, kidney; ve, vertebra. D, Sagittalsection of a 12.5-day p.c. transgenic embryocarrying i(182x4)pgloblacZ showing staining of ribchondrocytes. Bars, 100 µm.

2E,F,G). No staining was detected in the epidermis, the brainor any other non-chondrogenic tissue. Previous in situ hybrid-ization analysis revealed, however, transient low-levelexpression of the proα1(II) collagen gene in certain areas ofthe brain and epidermis at the same developmental stage(Cheah et al., 1991).

Newborn mice harboring p3000i3020Col2a1 also exhibiteda cartilage-specific X-gal staining pattern (data not shown). Asreported elsewhere (Lefebvre et al., 1994), primary rib chon-drocytes isolated from these mice survived in the presence ofG418 (neomycin) and expressed high levels of β-galactosidase.

In summary, the 3 kb promoter and 3,020 bp of intron 1targeted expression of the βgeo gene to chondrocytes or pre-chondrocytes in a tissue-specific and stage-specific manner(Table 1).

Col2a1 promoter and intron 1 deletionsWhen the length of the Col2a1 promoter was reduced to 687

3682 G. Zhou and others

bp in a construction containing the same 3 kb intron 1 fragmentas that of construction p3000i3020Col2a1, the same chondro-cyte-specific X-gal staining pattern was maintained in 14.5-dayp.c. embryos (data not shown). Thus promoter sequencesupstream of position −687 were not necessary for chondrocyte-specific expression in intact mouse embryos. In a parallelapproach to the one used in the present study we also used anew highly differentiated chondrocytic cell line established inculture from a rat chondrosarcoma tumor to perform transientexpression experiments in order to delineate a chondrocyte-specific enhancer in the mouse Col2a1 gene. These experimentsindicated that a 546 bp segment of intron 1 acted as an enhancerin chondrosarcoma cells but not in 10T1/2 fibroblasts(Mukhopadhyay et al., in press). Using the results of thesetransient expression experiments as guide, we then reduced theintron size to a 468 bp BamHI-SmaI fragment while maintain-ing the 687 bp promoter (p687i468Col2a1). This 468 bpfragment was included in the 546 bp fragment. We obtainedeight different transgenic mice which were positive for this con-struction by Southern analysis. X-gal staining was positive infour of these mice and showed a cartilage-specific expressionpattern at day 14.5 p.c. that was identical to that of embryoscarrying the larger intron sequence (Fig. 3A and Table 1).

To test whether the two previously reported silencerelements (Savagner et al., 1990) located in the promoter playeda role in tissue specificity, the segment between −687 and −309 in the promoter was deleted, in a construction containingthe 468 bp intron 1 sequence (p309i468Col2a1). The mousepromoter segment between −687 and −309 contains sequencesthat are homologous to the silencer elements of the rat gene(Vikkula et al., 1992). In 14.5-day p.c. transgenic mouseembryos the same cartilage-specific X-gal staining pattern wasfound as in embryos harboring a longer promoter (Fig. 3B).Histochemical analysis showed that only chondrocytes werestained (data not shown). Hence, the previously describedsilencer elements were not needed for cartilage-specificexpression in vivo.

Deletion of intron 1 prevents cartilage-specificexpressionTo better define the respective roles of promoter and intron 1sequences in the chondrocyte-specific expression pattern, wegenerated transgenic mice carrying the construct p309Col2a1in which the βgeo reporter gene was driven by the 309 bpproximal promoter. No intron 1 tester sequences were presentin this construction except for the 70 5′ proximal bp. In twodifferent transgenic lines the 14.5-day p.c. embryos showedfaint X-gal staining in the mesencephalon and metencephalonareas, but no cartilage staining (Fig. 3C). In two other 14.5-day p.c. transgenic founder embryos, there was X-gal stainingin the dermis covering the lower jaw and on the surface of thetongue (Fig. 3D), but again no expression was detected in car-tilages. Since X-gal staining did occur, we assume that theremaining intron 1 RNA sequences were removed correctlyduring the splicing reaction. Hence, deletion of the 468 bpintron 1 fragment abolished cartilage-specific expression.

Further deletion of intron 1 segment The 468 bp intron 1, which confers chondrocyte specificity, wasfurther dissected in shorter segments. When a 182 bp long DraI-SmaI fragment located at the 3′ end of the 468 bp fragment was

used in the βgeo vector together with a 687 bp Col2a1 promoter(p687i182Col2a1), this construction retained the ability totarget expression of β-galactosidase specifically in chondro-cytes (data not shown). With this construction only 25 per centof transgenic mice that were positive by Southern analysisexpressed the transgene. The sequence of this fragment ishighly conserved between human, rat and mice (Metsäranta etal., 1991; Vikkula et al., 1992). In transient expression experi-ments of rat chondrosarcoma cells and 10T1/2 cells, we hadobserved that two tandem copies of a 231 bp intron 1 fragmentof the mouse Col2a1 gene, which included the 182 bp fragment,produced a much stronger chondrosarcoma cell-specificenhancing activity than a single copy of this 231 bp fragment(Mukhopadhyay et al., in press). We, therefore, generated micecarrying the p309i(182x2)Col2a1 construction in which the 182bp fragment was tandemly repeated. In 14.5-day p.c. embryosharboring this construction chondrocyte-specific X-gal stainingwas more pronounced and occurred with a higher frequencythan in embryos carrying a single copy of the 182 bp enhancer(see Table 1). In these embryos, X-gal staining was locatedexclusively in chondrocytes in all cartilages (Fig. 3E,F,G).Therefore, this construction targeted β-galactosidase activity ina cartilage-specific pattern that was the same as that ofp3000i3020Col2a1.

Previous DNA-binding experiments by another laboratoryhad identified a decamer sequence in the rat Col2a1 intron 1as a binding site for a protein that was enriched in nuclearextracts of primary chondrocytes compared to nuclear extractsof fibroblasts (Wang et al., 1991). To examine the function ofthis decamer we cloned three tandem copies of a 73 bpsequence located at the 3′ end of the 182 bp fragment, whichincludes the mouse homologue of this decamer sequence, inthe βgeo vector in conjunction with the 309 bp Col2a1promoter and generated transgenic mice haboring the con-struction p309i(73x3)Col2a1. In 14.5-day p.c. embryos X-galstaining was detected in the eyes, vibrissae, around the mouth,between digits, in skin and spinal cord but not in cartilages(data not shown). Histological analysis also showed X-galstaining in some non-chondrogenic tissues such as dermis andtongue (data not shown). We concluded that the tandemlyrepeated 73 bp fragment was unable to confer chondrocyte-specific expression of the βgeo transgene (Table 1).

To test whether sequences at the 5′ end of the 182 bp intron1 segment could direct chondrocyte-specific expression, weinserted two tandem copies of a 116 bp sequence at its 5′ endin the βgeo vector containing the 309 bp Col2a1 promoter. Infour transgenic mice carrying this construction no X-galstaining was detected (Table 1). Since both a 73 bp fragmentand a 116 bp fragment were unable to confer cartilage-specificlacZ expression in transgenic embryos, it is possible that boththe 5′ part and the 3′ part of the 182 bp fragment are neededtogether for efficient chondrocyte-specific X-gal staining intransgenic embryos.

The 182 bp intron 1 DNA sequence targets activityof the human β-globin promoter to chondrocytesTo investigate whether Col2a1 promoter sequences wereneeded for chondrocyte-specific enhancer activity in intactmice, we generated a transgene in which the Col2a1 promoterwas replaced by a 72 bp human β-globin promoter element(from −44 to +28). Previous work had shown that a 133 bp

3683A 182 bp Col2a1 enhancer

fragment of the myogenin gene, placed upstream of this humanβ-globin promoter fragment, conferred skeletal muscleexpression to a lacZ reporter gene in transgenic mice,mimicking the expression pattern of the endogenous myogeningene (Yee and Rigby, 1993). Four tandem repeats of the 182bp intron 1 enhancer fragment of the mouse Col2a1 gene wereligated upstream of the −44 to +28 β-globin promoter, whichwas linked to the lacZ gene (i(182x4)pgloblacZ).

Fourteen-and-a-half-day p.c. embryos harboring this construc-tion showed the same high-level X-gal staining in cartilages asdid p3000i3020Col2a1 transgenic mice at the same develop-mental stage (Fig. 4A,B,C). At earlier stages of embryogenesis,this construction targeted β-galactosidase expression to all chon-drogenic tissues in a pattern that was similar to that ofp3000i3020Col2a1. At day 10.5 p.c. X-gal staining was observedin sclerotomes and also in the first and second branchial arches.Histological section showed X-gal staining in the notochord andsurrounding mesenchymal cells (data not shown). At 12.5 daysp.c. X-gal staining was found in the cartilages of the head, trunkand limbs, and histological analysis showed that X-gal stainingwas present in chondrocytes (Fig. 4D).

In these embryos promiscuous X-gal staining was detectedin some areas of the brain but not in other non-chondrogenictissues. It is possible that given the very short size of the humanβ-globin promoter in the transgene, the promiscuous brainexpression was due to site-of-integration effects. We cannotexclude, however, the possibility that the mouse Col2a1promoter could contain sequences that prevent this promiscu-ous expression. These experiments indicate, nonetheless, thatno Col2a1 promoter sequences were needed to targetexpression of the lacZ transgene to chondrocytes.

DISCUSSION

In previous preliminary experiments we have shown that asegment of the mouse Col2a1 gene, spanning a region of DNAfrom 3 kb upstream of the transcription start to part of exon 4,was able to confer chondrocyte-specific expression to the β-galactosidase reporter gene in transgenic mice (Metsäranta etal., 1995). Successive deletions allowed us to delineate a 182bp fragment in intron 1 which was sufficient for chondrocyteexpression in intact mouse embryos. Moreover, Col2a1promoter sequences were dispensable for chondrocyteexpression since transgenic mice harboring the 182 bp intron1 fragment placed upstream of a minimal β-globin promoterexpressed β-galactosidase in chondrocytes.

This chondrocyte expression also followed the sametemporal pattern during embryonic development as that of theendogenous Col2a1 gene. The earliest time of transgeneexpression occurred at day 10.5 p.c., a time when prechondro-cytic mesenchymal cells first appear in sclerotomes and pre-sumably also in the first two branchial arches, clearly indicat-ing that Col2a1 is an early marker of chondrocytedifferentiation during embryonic development. Our experi-ments imply that the 182 bp sequence of intron 1 containsbinding sites for proteins which control the chondrocyteexpression of the Col2a1 gene. Our cartilage-specificexpression vectors could also be used to target other genes tochondrocytes in vivo so as eventually to study the role of thesegenes in cartilage differentiation and skeletal development.

The non-homogeneous X-gal staining pattern in mesenchy-mal cells surrounding the notochord in 10.5-day p.c. embryossuggests the hypothesis that the earliest differentiation of mes-enchymal cells in chondrocytes in this region may be influencedby extracellular molecules originating from the notochord, aknown source of inducing agents based on graft experiments(Pourquié et al., 1993). Except for the notochord, no X-galstaining was detected in our transgenic embryos in extrachon-drocytic sites in which the endogenous gene is active, such asbrain, heart and epidermis. This could mean either that DNAelements outside those in the 3 kb promoter or the first intronare responsible for this extrachondrocytic expression or that thelevel of expression of the transgene in these sites was below thethreshold of X-gal detection. Although the level of expressionof the endogenous gene in these extrachondrocytic sites has notbeen precisely determined, in situ hybridization suggested amuch reduced expression level compared to what is observed inchondrocytes (Cheah et al., 1991). Promiscuous expression inthe brain was observed only with the construction containing theminimal β-globin promoter and the 182 bp intron 1 segment.

Experiments from another laboratory identified a decamersequence in the rat gene, which was proposed to be involvedin chondrocyte-specific expression based on transientexpression experiments in primary chondrocytes and whichwas shown to be a binding site for a protein that is enriched inthese cells (Wang et al., 1991). The mouse homologue of thissequence is located towards the 3′ end of the 182 bp segmentin intron 1. Three tandem copies of a 73 bp fragment whichsurrounds this decamer sequence were unable to confer chon-drocyte-specific expression in transgenic mice. Furthermore, 2copies of a 116 bp fragment corresponding to the 5′ end of the182 bp segment were also unable to direct chondrocyte-specificexpression in mice. Constructions containing fragments similarto the 73 bp and 116 bp subfragments were also unable to actas chondrocyte-specific enhancers in transient expressionexperiments of a highly differentiated chondrosarcoma cellline (Mukhopadhyay et al., in press). It is possible thereforethat sequences in both the 5′ and 3′ parts of the 182 bp intron1 segment are required together for efficient chondrocyteexpression in mice and in transient transfection experiments ofrat chondrosarcoma cells. Alternatively, it is possible that theelement necessary for chondrocyte expression in the 182 bpsegment is located around the short overlapping segment of the116 bp and 73 bp subfragments, and that this element wasdisrupted in both subfragments. However, this explanation isunlikely, since the overlapping segment of the two subfrag-ments used in transient transfections of rat chondrosarcomacells was 29 bp long. Further experiments will be needed toidentify chondrocyte-specific proteins binding to this 182 bpsequence.

We are grateful to Patricia McCauley for editorial assistance. Thiswork was funded by the grants NIH AR 40335 and AR 42909 (toBenoit de Crombrugghe). S. G. and V. L. were supported by post-doctoral fellowships from the Arthritis Foundation.

REFERENCES

Cheah, K. S., Lan, E. T., Au, P. K. and Tam, P. P. (1991). Expression of themouse α1(II) collagen gene is not restricted to cartilage during development.Development 111, 945-953.

3684 G. Zhou and others

Friedrich, G. and Soriano, P. (1991). Promoter traps in embryonic stem cells:a genetic screen to identify and mutate developmental genes in mice. GenesDev. 5, 1513-1523.

Garofalo, S., Vuorio, E., Metsäranta, M., Rosati, R., Toman, D., Vaughan,J., Lozano, G., Mayne, R., Ellard, J., Horton, W. and de Crombrugghe,B. (1991). Reduced amount of cartilage collagen fibrils and growth plateanomalies in transgenic mice harboring a glycine-to-cysteine mutation in themouse type II procollagen α1-chain gene. Proc. Nat. Acad. Sci. USA 88,9648-9652.

Hogan, B. L. M., Costantini, F. and Lucy, E. (1986). Manipulating the MouseEmbryo. A Laboratory Manual. Cold Spring Harbor Laboratory Press, NY.

Horton, W., Miyashita, T., Kohno, K., Hassell, J. R. and Yamada, Y.(1987). Identification of a phenotype-specific enhancer in the first intron ofthe rat collagen II gene. Proc. Nat. Acad. Sci. USA 84, 8864-8868.

Lawn, R. M., Efstratiadis, A., O’Connell, C. and Maniatis, T. (1980). Thenucleotide sequence of the human β-globin gene. Cell 21, 647-651.

Lee, B., Vissing, H., Ramirez, F., Rogers, D. and Rimoin, D. (1989).Identification of the molecular defects in a family with spondyloepiphysealdysplasia. Science 244, 978-980.

Lefebvre, V., Garofalo, S., Zhou, G., Matsäranta, M., Vuorio, E. and deCrombrugghe, B. (1994). Characterization of primary cultures ofchondrocytes from type II collagen/β-galactosidase transgenic mice. MatrixBiol. 14, 329-335.

Mercer, E. H., Hoyle, G. W., Kapur, R., Brinster, R. L. and Palmiter, R. D.(1991). The dopamine β-hydroxylase gene promoter directs expression of E.coli LacZ to sympathetic and other neurons in adult transgenic mice. Neuron7, 703-716.

Metsäranta, M., Toman, D., de Crombrugghe, B. and Vuorio, E. (1991).Mouse type II collagen gene. Complete nucleotide sequence, exon structureand alternative splicing. J. Biol. Chem. 266, 16862-16869.

Metsäranta, M., Garofalo, S., Leppanen, P., Smith, C., Niederreither, K.,de Crombrugghe, B. and Vuorio, E. (1995). Developmental expression ofa type II collagen/β-galactosidase fusion gene in transgenic mice. Dev.Dynam. (in press).

Mukhopadhyay, K., Lefebvre, V., Zhou, G., Garofalo, S., Kimura, J. H.and de Crombrugghe, B. (1995). Use of a rat chondrosarcoma cell line todelineate a 119 bp chondrocyte-specific enhancer element and to defineactive promoter segments in the mouse proa1(II) collagen gene. J. Biol.Chem. (in press).

Niederreither, K., D’Souza, R. N. and de Crombrugghe, B. (1992). MinimalDNA sequences that control the cell lineage-specific expression of theproα2(II) collagen promoter in transgenic mice. J. Cell Biol. 119, 1361-1370.

Pourquié, O., Caltey, M., Teillet, M., Ordahl, C. and Le Douarin, N. M.(1993). Control of dorsoventral patterning of somitic derivatives bynotochord and floor plate. Proc. Nat. Acad. Sci. USA 90, 5242-5246.

Sandell, L. S., Morris, N., Robbins, J. R. and Goldring, M. B. (1991).Alternatively spliced type II procollagen mRNA define distinct populationsof cells during vertebral development: differential expresssion of the amino-propeptide. J. Cell Biol. 114, 1307-1319.

Savagner, P., Miyashita, T. and Yamada, Y. (1990). Two silencers regulatethe tissue-specific expression of the collagen II gene. J. Biol. Chem. 265,6669-6674.

Southern, E. (1975). Detection of specific sequences among DNA fragmentsseparated by gel electrophoresis. J. Mol. Biol. 98, 503-517.

Stockwell, R. A. (1979). Biology of Cartilage Cells. Cambridge UniversityPress, Cambridge.

Upholt, W. B. (1989). The type II collagen gene. In Collagen, vol. IV (ed. B.R.Olsen and M. E. Nimni), pp. 31-49, CRC Press, Boca Raton, FL.

Vikkula, M., Metsäranta, M., Syvänen, A.C., Ala-Kokko, L., Vuorio, E.and Peltonen, L. (1992). Structural analysis of the regulatory elements of thetype II procollagen gene. Biochem. J. 285, 287-294.

Wang, L., Balakir, R. and Horton, W. E. Jr (1991). Identification of a cis-acting sequence in the collagen II enhancer required for chondrocyteexpression and the binding of a chondrocyte nuclear factor. J. Biol. Chem266, 19878-19881.

Yamada, Y., Miyashita, T., Savagner, P., Horton, W., Brown, K. B.,Abramczuk, J., Xie, H., Kohon, K., Bolander, M. and Bruggeman, L.(1990). Regulation of the collagen II gene in vitro and in transgenic mice. InStructure, Molecular Biology, and Pathology of Collagen (ed. R.Fleischmajer, B. R. Olsen and K. Kühn) vol. 580, pp. 81-87, The New YorkAcademy of Sciences, New York.

Yee, S. and Rigby, P. W. J. (1993). The regulation of myogenin geneexpression during the embryonic development of the mouse. Genes Dev. 7,1277-1289.

(Received 1 August 1995 - Accepted 25 September 1995)