Testis Expression of Hormone-sensitive Lipase Is Conferred by a Specific Promoter That Contains Four Regions Binding Testicular Nuclear Proteins*

The testicular isoform of hormone-sensitive lipase (HSLtes) is encoded by a testis-specific exon and 9 exons common to the testis and adipocyte isoforms. In mouse, HSLtes mRNA appeared during spermiogenesis in round spermatids. Two constructs containing 1.4 and 0.5 kilobase pairs (kb) of the human HSLtes gene 5′-flanking region cloned upstream of the chloramphenicol acetyltransferase gene were microinjected into mouse oocytes. Analyses of enzyme activity in male and female transgenic mice showed that 0.5 kb of the HSLtes promoter was sufficient to direct expression only in testis. Cell transfection experiments showed that CREMτ, a testis-specific transcriptional activator, does not transactivate the HSLtes promoter. Using gel retardation assays, four testis-specific binding regions (TSBR) were identified using testis and liver nuclear extracts. The testis-specific protein binding on TSBR4 was selectively competed by a probe containing a SRY/Sox protein DNA recognition site. Sox5 and Sox6 which are expressed in post-meiotic germ cells bound TSBR4. Mutation of the AACAAAG motif in TSBR4 abolished the binding. Moreover, binding of the high mobility group domain of Sox5 induced a bend within TSBR4. Together, our results showed that 0.5 kb of the human HSLtes promoter bind Sox proteins and contain cis-acting elements essential for the testis specificity of HSL.

Hormone-sensitive lipase (HSL) 1 is a triacylglycerol lipase and a cholesterol esterase expressed at high levels in adipo-cytes, testes, and adrenals (1)(2)(3). In adipocytes, HSL catalyzes the rate-limiting step in the hydrolysis of triglycerides into fatty acids and glycerol (4). HSL activation is mediated through phosphorylation by the cAMP-dependent protein kinase (5). In rat testis, HSL mRNA and protein are expressed in the seminiferous tubuli and not in interstitial cells with a stage-dependent pattern corresponding to the appearance of haploid germ cells (3,6). Several isoforms of HSL produced by a single gene have been characterized (2,7,8). Human adipose tissue expresses a 2.8-kb mRNA that encodes an 88-kDa protein (7,9). The mRNA and protein species expressed in testis are larger, 3.9 kb and 120 kDa, respectively (3). Analysis of coding sequences revealed that human adipocyte and testis HSL (HSL tes ) are 775 and 1076 amino acids long, respectively. HSL tes differs from the adipocyte form by a unique NH 2 -terminal region. Elucidation of the HSL gene organization showed that nine coding exons are common to both forms. The additional sequence in HSL tes is encoded by a 1.2-kb-long testisspecific exon (3,7). When a gene is expressed in somatic tissues and in germ cells, tissue-specific expression often results from alternate promoter use (10,11). The promoter of the adipocyte form of HSL (9) is located 13 kb downstream of the HSL tes 5Ј-flanking region suggesting that the expression of the different forms of HSL is controlled by several tissue-specific promoters.
During spermatogenesis, specialized transcriptional mechanisms ensure stage-specific gene expression in the germ cells. The factors controlling gene expression in post-meiotic germ cells are beginning to be elucidated. Several germ cell-specific putative transcription factors have been cloned, but target genes have been identified only for a few of them. CREM is a product of the CREM gene that acts as a transcriptional activator responsive to the cAMP signaling pathway (12). Several target genes for CREM-mediated activation have been identified in haploid germ cells, most notably the gene encoding protamine 1, a nucleoprotein that replaces histones and promotes nuclear condensation (13,14). Moreover, targeted disruption of the CREM gene results in a complete block of germ cell differentiation at the first steps of spermiogenesis (15,16). Thus, CREM may govern a coordinated regulation of gene expression in post-meiotic germ cells.
In this study, we investigated the molecular mechanisms that control the testis-specific expression of HSL tes . During spermatogenesis, HSL tes mRNA was expressed in haploid germ cells concomitantly with protamine 1 mRNA. We show that 0.5 kb of the HSL tes promoter was sufficient to drive testis-specific expression in transgenic mice. In cell transfection experiments, transactivation of the HSL tes promoter was independent of the cAMP signaling pathway. Four regions bound nuclear proteins * 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 /EBI Data Bank with accession number(s) AJ132272.

EXPERIMENTAL PROCEDURES
Northern Blot Analyses-A mouse HSL DNA probe (477 bp) was generated by PCR on mouse genomic DNA with primers located in the first adipocyte coding exon 5Ј-ATG GAT TTA CGC ACG ATG ACA CAG-3Ј and 5Ј-TAG CGT GAC ATA CTC TTG CAG GAA-3Ј. Proenkephalin (227 bp) and protamine 1 (207 bp) cDNA probes were generated by reverse transcription-PCR on mouse testis total RNA using 5Ј-GAC AGC AGC AAA CAG GAT GA-3Ј and 5Ј-TTC AGC AGA TCG GAG GAG TT-3Ј, and 5Ј-AGC AAA AGC AGG AGC AGA TG-3Ј and 5Ј-AGA TGT GGC GAG ATG CTC TT-3Ј primers, respectively. PCR reactions were performed using the proofreading pfu DNA polymerase (Stratagene). PCR products were cloned into pBluescript (Stratagene) using the TA cloning procedure (17). Identity of the amplicon sequences to published sequences was checked by automatic DNA sequencing (Applied Biosystems).
Total testis RNA was prepared from prepuberal and sexually mature mice by a single-step guanidinium thiocyanate phenol/chloroform extraction (18). RNA samples (25 g) were separated on a 1% agarose, 2.2 M formaldehyde gel, transferred, and UV cross-linked to a nylon membrane (Nytran, Schleicher & Schuell). Equal loading of the different lanes was checked by ethidium bromide staining of the gel and by hybridization with a rat ␤-actin probe. Membranes were pre-hybridized for 1 h in hybridization buffer (500 mM Na 2 HPO 4 , 1 mM EDTA, 7% SDS, 1% bovine serum albumin) and then hybridized overnight in 10 ml of the same buffer containing 1.5 10 6 cpm/ml HSL and proenkephalin cDNA probes and 10 6 cpm/ml protamine cDNA probe. After hybridization, membranes were washed twice with 0.3 M NaCl, 30 mM tri-sodium citrate, 0.1% SDS 20 min at room temperature and once with 30 mM NaCl, 3 mM tri-sodium citrate, 0.1% SDS for 30 min at 65°C. Membranes were subjected to digital imaging (Molecular Dynamics).
Plasmid Constructs-A 1.6-kb HindIII/BglII human DNA genomic restriction fragment was isolated from a cosmid clone containing the entire human HSL gene (3). The fragment was subcloned into the HindIII and BamHI sites of pBluescript and sequenced by automatic DNA sequencing (Applied Biosystems). It contained 1.4 kb of the 5Јflanking region upstream of the testis-specific exon. The construct was digested with HindIII and XbaI, and the 1.6-kb fragment was ligated upstream of the chloramphenicol acetyltransferase (CAT) gene into the promoterless pCAT-basic vector (Promega) (p1.4HSLtesCAT). The 1.6-kb HindIII/BglII DNA genomic restriction fragment was also cloned upstream of the luciferase gene into the promoterless pGL3-basic vector (Promega) (p1.4HSLtesLUC). About 900 bp of p1.4HSLtesLUC 5Јflanking sequence was deleted by digestion with SmaI to produce p0.5HSLtesLUC. The SmaI site used to generate p0.5HSLtesLUC and the AvaI site used to generate the microinjected fragment 0.5HSLtes-CAT (see below) are overlapping. The complete 1-kb CREM cDNA (12) was subcloned into the expression vector pSVSport (Life Technologies, Inc.).
Transgenic Mice-The two transgenes were prepared by digesting p1.4HSLtesCAT with HindIII and BamHI or with AvaI and BamHI to give, respectively, 1.4HSLtesCAT and 0.5HSLtesCAT. These two fragments were isolated on agarose gel by electroelution and purified using an elutip-d column (Schleicher & Schuell). Transgenic mice were produced by microinjection of the transgenes into the pronuclei of fertilized B6D2/F1 mouse eggs (19). Microinjected embryos were transferred to pseudo-pregnant B6-CBA/F1 female mice and carried to term. Screening of the positive transgenic animals was performed with DNA prepared from tail samples using Southern blot or PCR using as sense primer an oligonucleotide located in the human HSL tes 5Ј-flanking sequence and as antisense primer an oligonucleotide located in the CAT gene. Subsequent generation of heterozygous mice were produced by mating transgenic mice with wild type B6-CBA/F1 mice. The transmission of the transgene was ϳ50% in the progeny of all founders indicating Mendelian transmission. Protein extracts for CAT assays were prepared from hemizygous transgenic mice. Briefly, tissues were rapidly frozen in liquid nitrogen and homogenized in 0.5 ml of 250 mM Tris, pH 7.6, 5 mM EDTA, and 1 mM DTT. Homogenates were heated 7.5 min at 65°C and centrifuged at 4°C for 15 min at 13,000 rpm. Supernatants were kept for CAT and protein analyses (17,20).
RNase H Mapping-Ninety g of RNA from transgenic sexually mature mouse testis or 1 g of human testis poly(A) ϩ RNA (CLON-TECH) were lyophilized and resuspended in 10 l of RNase H buffer (20 mM Tris, pH 7.5, 10 mM MgCl 2 , 100 mM KCl, 0.1 mM DTT, 5% sucrose) containing 10 pmol of the human-specific single strand antisense oligonucleotide 5Ј-GTA GAG TAA CTA AGG AGT TG-3Ј (nt 197 to 179 downstream of the transcriptional start site). After 10 min at 70°C, hybridization was performed for 30 min at 37°C. Then, 40 l of RNase H buffer containing 7 units of RNase H (Amersham Pharmacia Biotech) were added, and digestion was carried out for 45 min (21). The digestion products were separated on a polyacrylamide-urea gel after ethanol precipitation. The gel was washed twice in 7% formaldehyde, 9 mM Tris borate, 0.2 mM EDTA, and RNA was passively transferred onto a nylon membrane. Hybridization was performed as described above using a 32 P-labeled probe corresponding to the 197 bp located downstream of the transcription start site.
Cell Transfection Experiments-JEG3 cells grown in 28-cm 2 plates were transfected using Fugene-6 (Boehringer Mannheim) with 700 ng of p0.5HSLtesLUC or pCRE-LUC (Stratagene), 700 ng of CREM-pSVSport or pSVSport, 700 ng of pFC-PKA, an expression vector encoding the catalytic subunit of the cAMP-dependent protein kinase (Stratagene) or pFC-DBD, the negative control plasmid (Stratagene) and 50 ng of pRL-CMV vector (Promega). The pRL-CMV vector encoding Renila luciferase was used to normalize transfection efficiency. Cells were treated 44 h post-transfection with 1 mM dibutyryl cAMP (Sigma) when specified. Cells were harvested 48 h post-transfection for Firefly and Renilla luciferase activity determinations according to the manufacturer's instructions (Promega).
Preparation of Liver and Testis Nuclear Extracts-Total nuclei extracts were performed as described by Howard et al. (22) with modifications. Four adult rat testis and 10 -15 mg of adult rat liver (perfused with 0.9% NaCl to wash out blood) were washed in ice-cold saline containing 0.1 mM PMSF, decapsulated, and minced with scissors in 40 ml of homogenization buffer (10 mM HEPES, pH 8, 1 mM EDTA, 25 mM KCl, 0.5 mM spermidine, 0.15 mM spermine, 10% glycerol, 0.5 mM DTT, 0.5 mM PMSF, 0.1 mM benzamidine, 1 g/ml leupeptin, 1 g/ml pepstatin, 2 g/ml aprotinin, and sucrose (1.85 M for testis, 2 M for liver)). Tissues were then homogenized in a glass tissue grinder with a motordriven Teflon pestle until cells were broken. The homogenate was then completed to 80 (testis) and 100 ml (liver) with homogenization buffer, and 28-ml aliquots were layered over 10-ml cushions of the same buffer in SW28 tubes. The tubes were centrifuged at 27,000 rpm for 1 h at 4°C. The supernatants were carefully removed, and the tube walls were washed with water and dried. Nuclei pellets were resuspended in 3 (testis) or 5 ml (liver) of buffer A (20 mM HEPES, pH 8, 100 mM EDTA, 8.8 mM MgCl 2 , 25% glycerol, 0.5 mM spermidine, 0.15 mM spermine, 0.14 M NaCl) using an all-glass Dounce homogenizer (pestle B). An aliquot was diluted 100 times in 0.5% SDS, and the absorbance at 260 nm was measured. The nuclear suspension was diluted at 40 A 260 units per ml. An equal volume of buffer B (20 mM HEPES, pH 8, 100 mM EDTA, 8.8 mM MgCl 2 , 25% glycerol, 0.5 mM spermidine, 0.15 mM spermine, 0.7 M NaCl) was added dropwise, and the extract was gently shaken for 45 min. The viscous lysate was then centrifuged at 35,000 rpm for 1.5 h at 4°C to pellet the chromatin. Solid (NH 4 ) 2 SO 4 was progressively added (0.4 g/ml) to the supernatant and dissolved by gentle mixing. After incubation 45 min on ice, the precipitated proteins were centrifuged in an SW60 rotor at 37,000 rpm for 30 min at 4°C. The pellets were resuspended in 200 l of dialysis buffer (20 mM HEPES, pH 8, 1.2 mM EDTA, 60 mM KCl, 25% glycerol, 1 mM DTT, 0.5 mM PMSF) for 40 A 260 units of nuclear lysate (see above). The protein extract was dialyzed twice for 2 h against 200 ml of the dialysis buffer without DTT and PMSF. The precipitate, formed during dialysis, was discarded by a 10-min centrifugation at 10,000 rpm at 4°C. The protein extract was frozen in small aliquots in liquid nitrogen and stored at Ϫ80°C. Protein concentrations ranged between 5 and 10 mg/ml.
Gel Retardation Assays-Single strand oligonucleotides (35 bp) covering 0.5 kb of the testis HSL promoter were gel-purified. Other oligonucleotides used were as follows: mSRY/Sox, 5Ј-GTA GGG CAC CCA TTG TTC TCT-3Ј (23); signal transducer and activator of transcription, 5Ј-CTG ATT TCC CCG AAA TGA CGG-3Ј (24); HNF3, 5Ј-CTA GAA CAA ACA AGT CCT GCG T-3Ј (25); C/EBP, 5Ј-GAT CCG CGT TGC GCC ACG ATG-3Ј (26). 100 ng of single strand oligonucleotides were 5Ј-end-labeled using T4 polynucleotide kinase (Eurogentec) and [␥-32 P]ATP (Ͼ4000 Ci/mmol). After heat inactivation of the kinase, the labeled oligonucleotides were annealed to 300 ng of the complementary strand oligonucleotides. Labeled double strand oligonucleotides were purified with the QIAquick nucleotide removal kit (Qiagen). 32 P-Labeled DNA (1 ng at approximately 100,000 cpm/ng) was incubated on ice for 30 min with testis (10 g) or liver (8 g) nuclear extracts in a total reaction buffer volume of 25 l containing 10 mM HEPES, pH 7.9, 60 mM KCl, 0.1 mM EDTA, 1 mM DTT, 4 mM spermidine, 5 mM MgCl 2 , 12% glycerol, and 1 g of poly(dI-dC) (Amersham Pharmacia Biotech) or 0.5 g of poly(dI-dC) and 0.5 g of poly(dG-dC) (Amersham Pharmacia Biotech). DNA-protein complexes were resolved on 6% nondenaturing polyacrylamide gels at 10 V/cm for 3 h in a 23 mM Tris borate, pH 8, 0.5 mM EDTA migration buffer. Polyacrylamide gels were dried under vacuum and subjected to digital imaging (Molecular Dynamics). Gel retardation assays were also performed with Sox proteins. A truncated form of the trout orthologue of Sox6 deleted of the leucine zipper region was produced in a reticulocyte lysate-coupled transcription-translation system (Promega) using the pCMV/Sox-LZ(D105-356) vector and the empty pRc/CMV vector (27). Six l of the reaction mixture were used in gel retardation assays. A peptide containing the high mobility group (HMG) box of mouse Sox5 (28) was produced in Escherichia coli as a glutathione S-transferase fusion protein and purified using glutathione-Sepharose beads and bovine thrombin. The purity of the peptide was checked on SDS-polyacrylamide gel electrophoresis. Fifty ng of purified protein and 0.5 ng of 32 P-labeled double strand oligonucleotide were used in gel retardation assays.
Circular Permutation Assay-Annealed synthetic oligonucleotides containing the TSBR4 region (Fig. 4) were cloned into the XbaI site of the circular permutation vector pBend2 (29). Circularly permuted DNA fragments were made by cleavage with the restriction enzymes indicated in Fig. 8A, dephosphorylated using calf intestine phosphatase (Eurogentec), and gel-purified. The DNA fragments were 5Ј-end-labeled using T4 polynucleotide kinase and [␥-32 P]ATP, and gel-purified. Binding reactions were performed for 20 min at room temperature with 10 ng of Sox5 peptide, 50 000 cpm of DNA in a reaction buffer containing 10 mM HEPES, pH 7.9, 60 mM KCl, 1 mM EDTA, 1 mM DTT, and 12% glycerol. The reactions were electrophoresed on 8% nondenaturing polyacrylamide gels at 10 V/cm for 4 -5 h. Bend parameters were calculated according to Thompson and Landy (30).

HSL tes Expression in Germ Cells-
The developmental expression of HSL tes mRNA was examined by Northern blot analysis of testis total RNA from mice at different ages. In rodents, the time at which a transcript appears during the first wave of spermatogenesis in prepuberal animal can be used to identify the spermatogenic cell type in which transcription initiates (31). The levels of proenkephalin and protamine 1 mRNA were therefore determined. In rodents, somatic and spermatogenic cells expressed a 1.4-and a 1.7-kb proenkephalin mRNA, respectively. The testis-specific proenkephalin mRNA is expressed at high levels in late pachytene spermatocytes, and protamine 1 mRNA expression appears in round spermatids (32)(33)(34). The proenkephalin germ cell form was detected from day 21 on (Fig. 1). HSL tes and protamine 1 mRNAs appeared on day 24. Densitometric analyses of the bands showed that the kinetics of HSL tes and protamine mRNA expression were very similar (data not shown), suggesting an expression of both genes in haploid round spermatids.
Analysis of Tissue-specific Expression in Transgenic Mice-To investigate whether the 5Ј-flanking region of the human HSL tes specific exon contained cis-acting sequences involved in tissue-specific expression, we generated transgenic mice with 1.4HSLtesCAT and 0.5HSLtesCAT constructs. Three (A, B, and C) and two (D and E) lines were generated from the large and small constructs, respectively. High levels of CAT activity were detected in testis and epididymis from sexually mature mice (between 60 and 90 days old) for the two transgenes ( Table I). The activity in epididymis was ascribed to sperm since, when mature sperm is washed from the epididymis, CAT activity was between 200 and 800 cpm/min/mg protein in the collected fluid. HSL enzymatic activity and protein was also detected in the collected fraction indicating the expression of HSL in sperm after spermiation (data not shown). No apparent variation in CAT activity was observed in the offspring of the founders and subsequent generations (data not shown). These data provided evidence that 0.5 kb of the 5Јflanking region are sufficient to drive expression of the CAT gene in testis. Next, we sought to determine if the 5Ј-flanking regions conferred tissue-specific expression. In males, CAT activity levels were very low in all non-gonadal tissues. In females, the low level of CAT activity seen in all tissues was comparable to the level detected in tissues of non-transgenic male and female mice (data not shown). Therefore, the sequences present in the first 0.5 kb of the human HSL tes promoter are critical for specific expression in testis. CAT activity was also determined in testis of 25-and 60-day-old mice from lines A and D. Four animals were analyzed per line at both ages. In young mice, the levels of CAT activity were 21 Ϯ 3 cpm/min/mg protein for line A and 18 Ϯ 1 cpm/min/mg protein for line D. In older mice, CAT activity levels were 818 Ϯ 36 and 862 Ϯ 43 cpm/min/mg protein, respectively. The marked increase in CAT activity showed that the transgenes were expressed in post-meiotic germ cells.
In order to check if the transcriptional start site of the chimeric genes expressed in transgenic mice and of the endogenous human HSL tes gene were identical, we performed RNase H mapping analyses with human-specific oligonucleotides on RNAs from human and transgenic mouse testis (Fig. 2). In both tissues, a band of ϳ175 nucleotides was detected. The data show that the human HSL tes promoter in transgenic mice used the same initiation site as the endogenous human promoter. Moreover, the length of the 5Ј-noncoding region deduced from RNase H mapping corresponded to the size (277 nucleotides) found using 5Ј-rapid amplification of cDNA ends PCR (3).
HSL tes Promoter Activation Is cAMP-and CREM-independent-It has been shown that CREM binds to cAMP-responsive elements (CREs) and stimulates transcription of several germ cell-specific genes (12,13). CREM functions as a transcriptional activator after phosphorylation by cAMP-dependent protein kinase. Computer-based (35) and visual analyses did not reveal apparent consensus sequences for CREs in the HSL tes promoter. Since functional CRE-like sites can substantially diverge from the palindromic sequence TGACGTCA (36), we wished to determine whether CREM and cAMP had an effect on HSL tes transcriptional activity. Because of the lack of haploid germ cell lines, cotransfection experiments were performed in JEG3, a human choriocarcinoma cell line bearing an efficient cAMP-dependent transduction pathway (13). To ensure that FIG. 1. HSL tes mRNA expression during development in mice. RNA blots were prepared with 25 g of testis total RNA from 14-, 17-, 21-, 24-, 28-, 35-, and 56-day-old mice. The blot was hybridized with HSL, proenkephalin, and protamine 1 cDNA probes. gProenk, germ cell proenkephalin; sProenk, somatic proenkephalin.
our transfection system was valid to study cAMP-dependent transactivation, we used a control CRE-LUC vector containing four copies of CREs upstream of a minimal promoter. This reporter construct was strongly cAMP-inducible whether the cells were treated with dibutyryl cAMP, a stable and permeable analogue of cAMP, or cotransfected with an expression vector for the catalytic subunit of the cAMP-dependent protein kinase. This result was predictable because JEG3 cells contain endogenous CREB. CREM coexpression resulted in a further increase of cAMP-induced luciferase activity. The results obtained with 0.5HSLtesLUC were strikingly different. No significant increase in luciferase activity was observed showing that the HSL tes promoter does not represent a cellular target of CREM/CREB transregulatory function.
Testis Nuclear Protein-binding Sites within the Human HSL tes Promoter-Transgenic analyses demonstrated that the 0.5-kb region located upstream of the transcriptional start site was sufficient to confer testis-specific expression. To assess directly whether sequences within the human HSL tes promoter bound nuclear proteins present in testis, a series of in vitro DNA binding studies was performed. The strategy used consisted in designing 20 overlapping double strand oligonucleotides spanning the entire region (Fig. 4). Each of the 20 oligonucleotides was used to map interaction sites for factors present in nuclear extracts prepared either from rat testis or from rat liver, an organ that does not express HSL. Four probes bound nuclear proteins expressed in testis but not in liver (Fig.  5). Analysis of the sequences of three testis-specific binding regions (TSBR) revealed no binding motifs for known testis transcription factors. TSBR4 contained a sequence AACAAAG (Fig. 4) that has been shown to bind members of the SRY/Sox protein family (37). The testis-specific binding on TSBR4 (Fig.  6) was competed by mSRY/Sox oligonucleotide but not by signal transducer and activator of transcription and HNF3 oligonucleotides. The mSRY/Sox oligonucleotide contains an AACAAT sequence with high affinity for mouse SRY, Sox5, and Sox6 (27, 28, 38 -40). An efficient competition was observed that was maximal with a 30-fold excess of mSRY/Sox oligonucleotide. Binding of the testis-specific nuclear proteins to TSBR4 was increased when poly(dG-dC) was added as nonspecific competitor (Fig. 6), a feature suggesting interaction of HMG domain proteins such as Sox proteins with A-T pairs in the minor groove of the DNA helix (41,42). A short form of Sox5 and Sox6 is expressed in mice in post-meiotic germ cells (27,38). Binding   A  7  11  1  2  2  5  1  10  1  21  750  450  B  12  5  3  7  1  0  4  5  8  15  830  ND  C  1  4  10  3  6  3  8  2  7  6  of recombinant Sox5 HMG box peptide and in vitro translated Sox6 protein on TSBR4 was studied using linker scan mutagenesis (Fig. 7). The HMG domain of Sox5 bound the mSRY/ Sox probe containing the AACAAT sequence and TSBR4 but not a C/EBP recognition motif (Fig. 7B). Mutation of the AA-CAAAG sequence of TSBR4 strongly decreased the binding. Because Sox6 homodimers do not bind DNA (27), a truncated form of Sox6 deleted of the leucine zipper region was produced by in vitro translation. As previously reported (27), incubation of the unprogrammed lysate (data not shown) and of the programmed lysate of the empty vector pRc/CMV with labeled double strand oligonucleotides resulted in retardation of the probe (Fig. 7C). The binding appeared to be due to endogenous DNA binding factor of the reticulocyte lysate. When the pCMV/ Sox-LZ(D105-356) vector was used, an additional binding complex was detected. This binding was abolished by mutation of the AACAAAG sequence. These data show that testis Sox proteins produced in vitro and from nuclear extracts bind TSBR4. The HMG Box of Sox5 Bends a Testis-specific Binding Region-Sox proteins as other proteins containing HMG domains induce a marked bend within the DNA (23,37). Therefore, we investigated whether the HMG box of Sox5 was able to modify TSBR4 DNA curvature using a circular permutation assay. TSBR4 DNA was cloned into the pBend2 vector (29). Digestion of the resulting plasmid with various restriction endonucleases gave DNA probes of almost identical sizes and base composition but with TSBR4 at variable distances from the end of the probe (Fig. 8A). Fig. 8B shows the result of a gel retardation assay with recombinant Sox5 HMG box peptide and the different DNA probes. The retarded complexes migrated with a mobility that was inversely correlated to the distance between the binding site and the end of the probe, a relationship characteristic of proteins that bend their target DNA (43). The ratio between the fastest and the slowest migrating species was used to estimate the extent of DNA distortion (Fig. 8C). The center of the bending mapped to the AACAAAG motif of TSBR4. Using the empirical equation proposed by Thompson and Landy (30), the angle of DNA bending was estimated between 65 and 70°. DISCUSSION In this paper, we show that the proximal 5Ј-flanking region of the HSL gene functions as a testis-specific promoter and binds testis nuclear proteins. HSL tes mRNA appears in round spermatids concomitantly to protamine 1 mRNA (Fig. 1). This result is in agreement with in situ hybridization data obtained in rat which showed that HSL tes mRNA was detected in stages X-XIV of spermatogenesis (6). As shown for many genes expressed during spermatogenesis, the HSL tes protein accumulation is delayed to stages XIII-VIII corresponding to late spermatids (3,44). The similar stage-specific expression pattern observed for HSL tes and protamine 1 mRNAs and other transcripts suggests the presence of common regulatory mechanisms. Since CREM plays an important role during the first steps of spermiogenesis as a transcriptional activator, we checked whether this transcription factor transactivates the HSL tes promoter. Cell transfection experiments (Fig. 3) similar to the ones performed with CREM-activated promoters (11,13,45) do not support a direct role for CREM and members of the CREB family in HSL tes promoter transactivation. An indirect role of CREM that is essential for a complete differentiation of haploid germ cells (15,16) cannot, however, be ruled out, e.g. through the control of expression of a transcription factor activating the HSL tes promoter.

1.4HSLtesCAT
The lack of appropriate male haploid germ cell line led us to use transgenic mice to investigate the transcriptional regulation of HSL tes . We demonstrate here that 0.5 kb of the region flanking the HSL tes -specific exon govern testis expression in transgenic mice (Table I). Analysis of a large number of tissues in male and female transgenic mice showed the strict testis expression of the transgene. The testis form of HSL that is characterized by larger mRNA and protein species than the other isoforms has only been detected in testis (1)(2)(3). Moreover, the 25-day-old transgenic mice showed very little CAT activity compared with the 60-day-old animals. The data in transgenic mice are therefore in agreement with the pattern and timing of expression of HSL tes .
In order to determine testis-specific DNA-protein interactions on the HSL tes promoter 0.5-kb region, gel retardation assays were performed using overlapping double strand oligonucleotides (Fig. 4). Four regions were shown to bind testis nuclear proteins absent in liver nuclear extracts (Fig. 5). One of them, TSBR4, contained a DNA sequence motif AACAAAG recognized by the HMG domain of SRY/Sox proteins (23,39). Competition experiments revealed that TSBR4 bound a testis nuclear protein that shows properties of a Sox protein (Fig. 6). Two members of the family, a short form of Sox5 and Sox6, are expressed in male germ cells at the round spermatid stage (27,28,38,46). The role of these proteins in spermatogenesis has not been documented. The HMG domain of Sox5 and a leucine zipper region-deleted Sox6 was shown to bind TSBR4 (Fig. 7). This observation raises the possibility that the short form Sox5 and/or Sox6 may directly or indirectly participate to the trans-activation of the HSL tes promoter. The HSL gene would therefore represent the first target gene of these proteins.
Cooperation of several Sox proteins and other transcription factors is often necessary to promote target gene expression (41, 46 -49). In teratocarcinoma cells, Sox2 and the POU domain transcription factor octamer 3 bind adjacent sites and participate together to the transactivation of the fibroblast growth factor 4 gene through protein-protein interaction (41,49,50). Either factor alone is ineffective. Three different Sox proteins, a long form of Sox5, Sox6, and Sox9, are coexpressed in chondrocytes and cooperatively activate the chondrocytespecific enhancer of the type II collagen gene (46). The activation is facilitated by the dimerization of the long form of Sox5 and Sox6. Sox6 contains a leucine zipper motif that allows dimerization of the protein, and homodimers fail to bind DNA (27). These data suggest that, in testis, Sox6 may bind to another protein as heterodimers to show transactivation properties. The short form of Sox5 expressed in testis does not contain the coiled-coil domain present in the long form. This domain is necessary for homo-and heterodimerization (46). In HeLa cells, expression of the short form Sox5 alone or coexpression of the short form Sox5 and Sox6 did not activate the HSL tes promoter (data not shown). Other uncharacterized testis Sox proteins might be involved in the activation of HSL tes . In addition, the cooperation between Sox proteins and other transcription factors might be necessary. The identification of the interacting partners will require extensive investigation since, except Sox binding to TSBR4, the other TSBRs do not show significant sequence homology with consensus binding motifs for known testis transcription factors.
Sox proteins could indirectly modulate transcription activity by organizing local chromatin structure. Binding of Sox proteins occurs in the minor groove and results in a bend within the DNA (37). It has been reported that Sox5 HMG box induces a bend to the AACAAT motif with an estimated angle of 74° ( 28). The nature of the recognition sequence and of the flanking nucleotides influence the angle of the bend (51). Here, we show that the HMG domain of Sox5 induces an estimated flexure of 65-70°through binding to the AACAAAG sequence of TSBR4 (Fig. 8). The data demonstrate that Sox5 can induce a strong bend in DNA in the context of a natural testis-specific pro-moter. Testis Sox proteins may act through an alteration of local chromatin structure around the AACAAAG site in TSBR4 to facilitate the interaction of distant enhancer nucleoprotein complexes (e.g. on the other TSBRs) with the basal transcription machinery.
To conclude, we have identified a testis-specific promoter that contains four regions binding testicular nuclear proteins. The HSL tes promoter provides a molecular basis to characterize new cis-acting elements and transcription factors responsible for the transactivation of genes in post-meiotic germ cells.
Acknowledgments-We are grateful to Ghislaine Hamard (INSERM U380, Paris, France) for help with oocyte microinjection; Dr. Michel   FIG. 8. Bending of the testis-specific binding region TSBR4 by Sox5 using a circular permutation assay. A, probes used in gel retardation analysis. TSBR4 represented as a box was cloned between the two direct repeats of pBend2 vector. Cleavage of the resulting plasmid at the restriction sites shown generated DNA fragments with permuted positions of TSBR4. B, gel retardation analyses were performed with the high mobility group domain of Sox5 and the circularly permuted DNA probes. C, relative mobility of the Sox5-DNA complexes according to the position of the AACAAAG motif of TSBR4 within the DNA probes. The mobilities of the protein-DNA complexes (R bound ) were normalized to those of the free probes (R free ). The flexure displacement was defined as the distance between the AACAAAG motif to the 5Ј end of the probe divided by the total length of the probe.