Neuron specificity of the neurofilament light promoter in transgenic mice requires the presence of DNA unwinding elements.

Three reporter genes, the chloramphenicol acetyltransferase (CAT), the lacZ, and the intronless NF-L DNA, were used to test the activity of the proximal promoter region (−292 bp) of the human neurofilament light (hNF-L) gene in transgenic mice. Surprisingly, the hNF-L/CAT construct was highly sensitive to position effect, and its expression was found at low levels in several tissues of adult transgenic mice (Beaudet, L., Charron, G., Houle, D., Tretjakoff, I. Peterson, A., and Julien, J.-P.(1992) Gene (Amst.) 116, 205-214). In [Medline] contrast, the hNF-L/lacZ or the hNF-L/intronless constructs were expressed exclusively in the nervous system during embryonic development and in adult animals. The DNA sequences analysis of the different reporter genes revealed the presence of matrix attachment regions (MARs) within the 3′-untranslated regions of all three transgenes. DNA unwinding elements were found within the MARs of lacZ and hNF-L gene constructs but not in the CAT gene construct. When this element was removed from the lacZ construct, expression of the hNF-L/lacZ transgene became susceptible to position effect and was no longer tissue-specific. These results indicate that DNA unwinding elements are essential for position effect independence conferred by MARs to the hNF-L basal promoter.

A variety of genes such as those coding for luciferase (2), ␤-galactosidase (lacZ) (3), chloramphenicol acetyltransferase (CAT) (4), 1 growth hormone (5), or alkaline phosphatase (6) have been extensively used as reporter genes, because they possess enzymatic activities that are easily detectable either in vitro or in vivo. With the availability of transgenic mouse technology, reporter-containing vectors have been developed for applications as diverse as analysis of enhancer and promoter function (7)(8)(9)(10) and cell lineage studies (11). Transgenic mice have been found to be powerful tools to delimitate regions responsible for tissue-specific (12) developmental (13) or stimulatory (14) pattern of expression.
When a transgene integrates into a host genome, its expression pattern might be affected by elements localized near the insertion site (15). This position effect has been related to the absence of insulator sequences responsible for the establishment of a higher order of chromatin structure called the loop domain (16). Loop domains may represent independent transcription units, with genes located within a loop being subject to inside regulatory environment and insulated from outside environment (17,18). Loops are maintained on the nuclear matrix by MARs located at both extremities and called domain boundaries (19). Even if domain boundaries do not seem to have transcriptional regulatory property of their own, their insulating effect on expression has been demonstrated as new domain boundaries have been characterized (20 -25). It has been shown that not only a position-independent expression but also a finer regulation of a transgene developmental and stimulatory pattern of expression could be achieved when transgenes are flanked by MARs (26). Recently, compilation and comparison of MAR sequences have allowed the assessment of many MAR characteristics (27). MARs range from 250 bp to several kb in length, they are enriched in A/T nucleotides, they bind to nuclear matrix in vitro and in vivo, and they often contain consensus topoisomerase II cleavage sites, singlestranded, kinked, or curved DNA and potential binding sites for homeobox containing DNA-binding proteins. Sometimes MARs colocalize with replication origins (28,29), regions implicated in transcriptional regulation (30,31), and regions of nuclease hypersensitivity (22). Considerable effort has been made to delimitate the minimal requirement for MAR to have an effect on the transcription process. Although binding to nuclear matrix seems to be a prerequisite, this criterion is insufficient by itself to assure an effect on transcription. Using stable transfection and mutagenesis experiments, sequences presenting unwinding capability have been shown to play an important role for transcriptional activation (32). However, their influence on tissue-specific and developmental transcriptional activation has yet to be determined.
The proximal promoter region (positions Ϫ292 to ϩ15 bp) of the human neurofilament light gene (hNF-L) is not sufficient to drive high level and neuron-specific expression of a CAT reporter gene in transgenic mice (1). However, nervous systemspecific expression is obtained with the same promoter linked to a lacZ reporter gene or an hNF-L/intronless gene. 2 These results prompted us to look for functional similarities between the lacZ and the hNF-L/intronless reporter genes that could explain their correct tissue-specific expression. Our results demonstrate that unwinding elements colocalize with MARs found at the 3Ј-untranslated region (UTR) of the lacZ and the hNF-L/intronless reporter genes. Deletion of a DNA segment containing these unpaired DNA regions abolishes tissue-specific expression of hNF-L/lacZ transgene in transgenic mice. We propose that unwinding elements are one of the important features required for MAR to insulate the hNF-L basal promoter from position effect.

EXPERIMENTAL PROCEDURES
Construction of Plasmid-The hNF-L basal promoter (positions Ϫ292 to ϩ15 bp) was ligated to a HindIII linker. To obtain the plasmid phNF-L/CAT, the hNF-L basal promoter ligated to the HindIII linker was inserted in the HindIII site of pUC-9 containing the CAT gene. The ␤-galactosidase gene (lacZ) was isolated from the plasmid pCH110 (Pharmacia) with a HindIII digestion followed by an EcoRI partial digestion. The 4,486-bp fragment containing lacZ was subcloned in the Bluescript pSKϩ vector linearized with a HindIII and EcoRI digestion creating the plasmid pSKlacZ. The hNF-L basal promoter ligated to the HindIII linker was inserted in the HindIII site of pSKlacZ to create the plasmid phNF-L/lacZ.
To delete all three introns of the hNF-L gene, we took advantage of the presence of a conserved BglII restriction site in the middle of exon 1 between the human gene and the mouse gene and of another conserved EcoRI restriction site near the polyadenylation signal of both genes. First, a 4.9-kb BamHI-XbaI fragment that includes the hNF-L gene (1) was subcloned in a Bluescript pSKϩ vector to create the plasmid pSKhNF-L. pSKhNF-L was digested to completion by BglII and EcoRI. From the mouse cDNA clone (33), a 1,220-bp BglII-EcoRI fragment was isolated and ligated to the open pSKhNF-L vector, creating the plasmid phNF-L/intronless.
The hNF-L/lacZ unwϪ construct was directly isolated by complete BamHI digestion of the plasmid phNF-L/lacZ and microinjected. Only the basal sequences of SV40 that allow correct polyadenylation were present in this 4,035-bp fragment. This fragment includes no plasmid sequences.
Microinjection into Mouse Eggs-The hNF-L/lacZ unwϪ construct was isolated as described above and diluted to a concentration of 1 g/ml in 10 mM Tris-HCl (pH 7.5), 0.15 mM Na 2 EDTA prior to microinjection in male pronuclei of C57Bl/6J ϫ C3H fertilized eggs. Injected eggs were transferred to the oviduct of pseudopregnant females (34). Identification of transgenic mice was carried out by Southern blot analysis of 10 g of tail genomic DNA (35).
A pair of oligonucleotides specific for the first hNF-L exon were used, their sequences are respectively 5Ј-GCACATCTCCAGCGTGCGT-3Ј (primer 522) and 5Ј-AGATCTGCGCGTACTGGATCTGCGCC-3Ј (primer 784). A second pair of oligonucleotides specific for the mouse and rat G3PDH mRNAs were used to quantify the amount of RNA in each sample. Their sequences are respectively 5Ј-TTGATGGTATTC-GAGAGAAGGG-3Ј (primer 808) and 5Ј-TCCAGGAGCGAGATC-CCGTC-3Ј (primer 809).
The first strand synthesis was performed by adding 8 l of 1.25 mM dNTP mix and 250 ng of primers 784 and 808 in presence of 30 units of RNAguard and 20 units of avian myeloblastosis virus reverse transcriptase. The mixture was incubated at 42°C during 1 h, and then the reaction was stopped at 95°C for 5 min. Each sample was subjected to two PCR reactions, one specific for the hNF-L transcript using primers 522 and 784 and the other for the G3PDH transcript using primers 808 and 809. This last one is an internal control. PCR reactions were done according to the manufacturer (NEB) using 1 ⁄20 of the cDNA mix as template, 0.20 Ci of [␣-32 P]dATP, and 1 unit of VENT DNA polymerase (New England Biolabs). After 25 cycles, 2 ⁄5 of the hNF-L-PCR reaction products and 1 ⁄5 of the G3PDH-PCR products reaction were separated by electrophoresis on a 1% agarose gel, blotted, and autoradiographed.
Isolation of Nuclear Matrix-All steps used for nuclei isolation were performed at 4°C except where specified. Minced brain and liver tissues from C3H female mice were homogenized in 10% glycerol, 10 mM HEPES (pH 7.4), 25 mM KCl, 1 mM EGTA (pH 8.0), 0.15 mM spermine, 0.5 mM spermidine, 1 mM dithiothreitol, 0.1 mM phenylmethylsulfonyl fluoride, 100 g/ml aprotinin, and 5 g/ml leupeptin supplemented with 2 or 2.3 M sucrose for brain or liver, respectively, by 10 strokes of a B pestle. Homogenates were centrifuged in a SW41 rotor at 25,000 rpm for 45 min. Isolated nuclei in the pellets were resuspended in RSB buffer (10 mM Tris-HCl (pH 7.4), 10 mM NaCl, 3 mM MgCl 2 ) containing 0.25 M sucrose and washed two times in the same buffer.
Nuclear matrix isolation was performed according to Cockerill and Garrard (30), except that EDTA was replaced by EGTA in all the buffers. The nuclei suspension was adjusted to a final concentration of 1 mM CaCl 2 and was digested with 100 g/ml of DNase-1 for 1 h at 24°C. After centrifugation for 10 min at 750 ϫ g, pellets were resuspended in RSB with 0.25% sucrose, and an equal volume of solution containing 20 mM Tris-HCl (pH 7.4), 4 M NaCl, and 20 mM EGTA was added. After a 10-min incubation, the nuclei were centrifuged at 1,500 ϫ g for 15 min. Pellets were extracted twice by suspension in 10 mM Tris-HCl (pH 7.4), 2 M NaCl, 0.5 mM phenylmethylsulfonyl fluoride, and 0.25 mg/ml bovine serum albumin and centrifugation at 4,500 ϫ g for 15 min. The resulting nuclear matrices were washed three times with 10 mM Tris-HCl (pH 7.4), 50 mM NaCl, 1 mM MgCl 2 , and 0.25 mg/ml bovine serum albumin by centrifugation for 30 s at 10,000 ϫ g.
Probe Labeling-Supercoiled containing plasmids (1 g) were digested with restriction enzymes: hNF-L/lacZ with EcoRI, BamHI and HindIII; hNF-L/CAT with BamHI, HindIII, and ScaI; and hNF-L/intronless with BamHI, BglII, XbaI, and EcoRI. After complete digestion, 20 units of calf intestine phosphatase were added, and the mixture was incubated for 1 h at 37°C. Phosphates were inactivated by incubation at 65°C for 15 min, and the reaction mixture was then extracted twice with phenol:chloroform:isoamyl alcohol (25:24:1) prior to ethanol precipitation. DNA pellets were resuspended in water and end-labeled with [␥-32 P]ATP using T4 polynucleotide kinase according to Sambrook et al. (37).
Nuclear Matrix Binding Assays-Nuclear matrix binding reaction was performed according to Cockerill and Garrard (30), except that EDTA was replaced by EGTA in all the buffers. Typically, 1.0 ϫ 10 7 matrices were resuspended in 100 l of binding buffer (10 mM Tris-HCl (pH 7.4), 50 mM NaCl, 2 mM EGTA, 0.25 mg/ml bovine serum albumin) containing 5 ng of [ 32 P]-end-labeled DNA fragments and unlabeled sonicated Escherichia coli genomic DNA (125-1,000 mg/ml). After incubating on a shaker for 1 h at 24°C, 500 l of binding buffer were added, and matrices were recovered by centrifugation at 10,000 ϫ g for 60 s at 4°C. After washing in 1 ml of the same buffer, matrix-bound DNA was solubilized in 0.5% SDS, treated overnight with 0.4 mg/ml Proteinase K, and after addition of 10 g of unlabeled carrier DNA, phenol-extracted and ethanol-precipitated. Resulting purified matrixbound DNA fragments were electrophoretically resolved on 1% agarose gels in TBE, blotted on nylon membrane (Hybond Nϩ), and detected by autoradiography.
Detection of DNA Single-stranded Region-The reaction with phNF-L/CAT, phNF-L/intronless, and phNF-L/lacZ plasmids was performed by resuspending 25 g of each supercoiled plasmid in 50 l of 50 mM NaOAc (pH 5.0) containing from 0 to 10% of chloracetaldehyde. The mixture was incubated for 1 h at 37°C. The modified DNAs were purified by two ETOH precipitations and were resuspended in 50 mM Tris-HCl (pH 8.0), 10 mM MgCl 2 , and 100 mM NaCl. The plasmid phNF-L/CAT was digested with HindIII, the phNF-L/intronless was digested with BamHI, and the plasmid phNF-L/lacZ was digested with XhoI to linearize each plasmid at the 5Ј end of the promoter region. After ETOH precipitation, 2 g of each plasmid was treated with 0.1 unit of SI nuclease in 30 l of 30 mM NaOAc (pH 4.5), 150 mM NaCl, and 2 mM ZnCl 2 for 10 min at 25°C, followed by ETOH precipitation. Purified DNAs were separated by electrophoresis on agarose gel and blotted. For the hNF-L/lacZ and the hNF-L/intronless transgenes, a 307-bp hNF-L promoter DNA fragment was labeled with [␣-32 P]ATP using an oligolabeling kit (Pharmacia) and was used as probe for Southern hybridization, whereas a 225-bp HindIII-EcoRI DNA fragment corresponding to the 5Ј end of the CAT gene was used for the hNF-L/ CAT transgene. The positions of labeled bands was revealed by autoradiography.

RESULTS
Using three different reporter genes (Fig. 1A), very different conclusions regarding the strength, the tissue-specificity, and the developmental expression driven by the basal hNF-L promoter were reached. 2 Fig. 1B summarizes the expression pattern generated in transgenic mice with each hNF-L/reporter gene construct. The hNF-L/lacZ and the hNF-L/intronless transgenes were correctly expressed in neuronal cells; examples are shown in Fig. 1, C and D, respectively. However, the same promoter was not sufficient to direct CAT expression in those cells (1). The striking finding from those studies is that the different reporter genes exerted some influence on the activity of the hNF-L promoter.
A/T-rich Regions in the Three Reporter Gene Constructs-The primary DNA sequence of each reporter gene construct was analyzed for its A/T content, as depicted in Fig. 2. Different patterns are observed for each construct. The CAT gene, for instance, is slightly A/T-rich at 55.4%, the lacZ gene is G/C-rich with an average of 44.1% A/T nucleotides, and the hNF-L/ intronless gene is composed of a 5Ј A/T-depleted region at 35%, followed by a region nearly equal in A/T versus G/C nucleotides at 52.5%. The three constructs have, at their 3Ј-UTRs, a DNA region enriched at more than 60% in A/T nucleotides; this ranges from 61.9% in the lacZ construct and 66.5% in CAT construct to 68.9% in hNF-L/intronless construct. The 849-bp A/T-rich region of the CAT construct is composed of SV40 sequences including the small t-antigen intron. In the lacZ construct, the 450-bp A/T-rich region is localized at the junction between the 3Ј-UTR of lacZ gene and SV40 sequences. For the hNF-L/intronless, the 434-bp A/T-rich region is present in the natural 3Ј-UTR. In summary, each construct has a single A/Trich region localized at the 3Ј-flanking region.
Nuclear Matrix Attachment of the A/T-rich Regions-A/Trich regions are one of the characteristics of MAR. We thus tested whether the A/T-rich regions present in the constructs were capable of binding to the nuclear matrix. Nuclear matrices were derived from brain and liver tissues of adult mice. The nuclear matrix fraction was prepared by sequential extraction of purified nuclei with low salt buffer, DNase-1, high-salt buffer, and nonionic detergent. DNA from each plasmid-containing construct was digested by appropriate restriction enzymes and end-labeled with [␥-32 P]ATP. The binding assay was accomplished by mixing nuclear matrices from either brain or liver with labeled probes in the presence of increasing concentration of nonspecific competitor DNA. DNA fragments bound to the nuclear matrix were isolated and separated by agarose struct is derived from plasmid pCH110 (Pharmacia). Its 3Ј-UTR contains a SV40 DNA fragment that comprises the polyadenylation signal region. The hNF-L/intronless transgene has been generated using conserved restriction sites between the human NF-L gene and the mouse cDNA. We deleted as a block from middle of exon 1 up to the first polyadenylation signal of the human NF-L gene and replaced it by corresponding coding sequences from the mouse cDNA NF-L making a hybrid NF-L/intronless transgene (detailed under "Experimental Procedures"). No SV40 DNA sequence has been added to this hNF-L/ intronless ,construct and polyadenylation occurs from the first endogenous hNF-L poly(A) signal (71). B, BamHI; Bg, BglII; Hd, HindIII; RI, EcoRI; Sc, ScaI; Xb, XbaI. B, relative expression pattern of the three hNF-L/reporter gene constructs in transgenic mice. Expression was judged to be in the nervous system when it colocalized with nervous system structures. C, X-gal staining on a whole mount hNF-L/lacZ embryo. ␤-galactosidase (lacZ) staining of one transgenic embryo expressing the transgene. Embryo was analyzed 13.5 days after microinjection, and pictures were taken from side, back, and front views. Positive tissues are identified by an X. D, reverse transcription PCR expression analysis of the hNF-L/intronless transgene. RNAs from various tissues of transgenic mouse line 38 were reverse transcribed prior to the PCR. Two sets of oligonucleotides were used: Set 1 corresponds to oligonucleotides specific for the first exon of the hNF-L gene, and the second set corresponds to the mouse G3PDH gene and is used as an internal control. Upper panel, amplified products from a hNF-L/intronless transgenic mouse mRNA; lower panel, amplified products from G3PDH mRNA. The hNF-L/CAT construct is constituted of the TN9 chloramphenicol acetyltransferase gene followed at its 3Ј end by the SV40 intron from the small t-antigen gene, which is flanked on each side by coding sequences. This segment is linked to the SV40 polyadenylation signal-containing region. The hNF-L/lacZ con-gel electrophoresis prior to autoradiography. Identification of bound fragment(s) was eased by selection of restriction enzymes digesting the plasmids in fragments of different sizes. Vector sequences present in the assay served as negative control because they possess no A/T-rich region.
Every DNA region of the hNF-L/CAT construct did bind to the brain nuclear matrix but with different affinities (Fig. 3A). The SV40 polyadenylation sequence (band 3) presents the strongest binding signal, followed by the CAT gene itself (band 2) and the hNF-L promoter (band 1), which is only weakly bound. Poly(dA⅐dT) is a better competitor than unrelated E. coli genomic DNA (Fig. 3A, lanes 8 and 15), indicating that A/T-rich sequences are responsible for specific nuclear matrix DNA attachment as expected (30,38). Fig. 3C shows the binding pattern of the hNF-L/lacZ construct DNA fragments. Three regions are attached to the nuclear matrix in the presence of nonspecific competitor DNA. These regions correspond respectively to the hNF-L promoter (band 1), the lacZ 3Ј-UTR (band 2), and the SV40 flanking sequences (band 3). The lacZ gene itself was not attached to the nuclear matrix. The lacZ 3Ј-UTR has the highest affinity for the nuclear matrix, and its attachment is not competed by high quantity of nonspecific competitor DNA, whereas the promoter region and the SV40 flanking sequences were competed away. For the hNF-L/intronless construct (Fig. 3E), two regions attached to nuclear matrix. The first one corresponds to the promoter and first hNF-L exon (band 1), whereas the second one corresponds to the 3Ј-UTR (band 2). This latter band has the highest affinity for nuclear matrix, and binding is efficiently competed by poly(dA⅐dT). For all these constructs, identical results were obtained with brain or liver nuclear matrices. Fig. 3 (B, D, and F) summarize the results for each construct. In conclusion, the 3Ј-UTR regions of all three reporter genes are bound to the nuclear matrix with high affinity.
Detection of Unwinding Elements in Constructs Expressed in the Nervous System-Some DNA sequences like those from bacterial and yeast origin of replication have the capability to be stably base-unpaired (39 -42). The presence of sequences with unwinding properties can be detected by treating supercoiled plasmids containing sequences of interest with chloracetaldehyde (43). This chemical reacts with unpaired bases and blocks reannealing once the plasmid is relaxed by restriction enzymatic digestion. The reaction leaves single-stranded regions that can be digested with S1 nuclease (44). To detect single-stranded regions in all three hNF-L/reporter gene constructs, the plasmids phNF-L/intronless, phNF-L/lacZ, and phNF-L/CAT were incubated with chloracetaldehyde. In this experiment, the bacterial replication origin ColE1 present in the Bluescript vector served as internal positive control for the reaction, this sequence being known to unpair easily under stress caused by supercoiling (39). With this mapping procedure, all three plasmids' replication origins have reacted with the chemical agent (Fig. 4). For the plasmid phNF-L/CAT, the unpaired sequences visualized as a band of 1.6 kb are localized outside the CAT construct region used to generate transgenic mice. No band is visible within the construct region (Fig. 4, left  panel). Within the plasmid phNF-L/lacZ, the presence of another single-stranded region was revealed by a clear band migrating at 4.1 kb (Fig. 4, central panel). The single-stranded region is localized in the SV40 sequences not present in the CAT reporter gene construct (Fig. 1). The plasmid phNF-L/ intronless also possess unpaired sequences, because bands migrating at 2.2 and 2.4 kb are observed. Those sequences are localized 300 bp before the end of the intronless NF-L gene. We conclude from these results that the lacZ and hNF-L/intronless reporter constructs do contain unwinding elements, unlike hNF-L/CAT, and that for both, these elements are localized within the associated MAR.
Deletion of lacZ Reporter Gene Unwinding Element Abolishes Tissue-specific Expression in Transgenic Mice-We were interested in testing if the DNA unwinding element present in MAR had any influence on the tissue-specific expression of the reporter gene driven by the hNF-L basal promoter. A new lacZ construct was generated by isolating a 4,035-bp BamHI fragment from the hNF-L/lacZ construct (see map in Fig. 1) generating construct hNF-L/lacZ unwϪ . The BamHI digestion allowed deletion of the SV40 flanking sequences harboring the unpaired sequences but maintained the 3Ј-UTR of lacZ and the minimum fragment assuring an efficient polyadenylation. With the hNF-L/lacZ unwϪ construct, only two out of seven transgenic embryos expressed the transgene at 13.5 days postfertilization. This proportion is quite different from the hNF-L/lacZ construct where six out of seven transgenic embryos correctly expressed the transgene (Fig. 1B). In the first transgenic embryo, lacZ staining occurred in the liver, limbs, and somites (Fig. 5, a, b, and c). The second transgenic embryo presents punctuated lacZ staining of tissues surrounding the telencephalon and myelencephalon; some nervous tissues are also faintly stained (Fig. 5, d, e, and f). Results indicate that deletion of the unwinding element of the MAR abolishes the tissue-specific expression of the hNF-L/lacZ fusion gene.

DISCUSSION
The hNF-L basal promoter does contain regulatory elements to confer neurone specificity. As demonstrated in transgenic mice, a region extending at the very 5Ј end of the promoter between Ϫ292 and Ϫ190 bp is necessary for neuronal expres- sion (45). This Ϫ292-bp basal promoter was found to be sufficient to allow expression of the lacZ reporter gene and the hNF-L/intronless gene in a correct tissue-specific and developmental manner (Fig. 1B). However, the CAT reporter gene expression under the same basal hNF-L promoter is highly susceptible to insertion site influence and is not expressed in a neuron-specific fashion (1).
All three transgenes have the same basal hNF-L promoter, and therefore their pattern of expression cannot be totally explained by the influence of this promoter. Clearly, other unknown elements within the reporter genes themselves do interfere with the expression of these transgenes. The presence of regulatory elements such as enhancers and silencers in the CAT or the lacZ reporter genes is unlikely because a neutrality on transcriptional activation of these genes has been inferred from numerous studies. In fact, both CAT and lacZ genes have been extensively used as enhancer and silencer trap vectors (review in Refs. 46 and 47). MARs represent likely candidates to explain apparent discrepancies in our analysis of NF-L promoter (23,48).
The chromatin of interphase nuclei is organized into topological constrained loops averaging 80 -90 kb in length that are attached to the nuclear matrix (19,49). This DNA organization seems to be important not only in the compaction of the chromatin fiber but also for the utilization of genetic information. Each domain can define an independent unit of gene activity insulated from the regulatory influences of adjacent domains (18). MARs have been found in the 5Ј-and 3Ј-flanking regions of fushi-tarazu, sgs-4, and alcohol dehydrogenase genes of Drosophila (19,21,50), the J-C intron of the mouse immunoglobulin gene (30,51), the first intron of the human HPRT gene, the chicken lysozyme gene, the human interferon-␤ gene, the human and murine ␤-globin gene, the chicken ␣and ␤-globin gene, and the apolipoprotein B gene (reviewed in Ref. 27). MARs enhance general promoter functions in an orientationand partially distance-independent manner, and their effect is restricted to the integrated state of transfected templates (23). These findings strongly suggest that some MARs might serve as a crucial control point for gene regulation.
Our results are indicative of nuclear matrix attachment through A/T stretches as described for many well characterized MAR sequences (38). For the hNF-L/CAT transgene, the MAR coincide to SV40 sequences already reported by Pommier et al. (52) as having the highest in vitro affinity for nuclear matrix in the entire SV40 genomic DNA. The SV40 sequences present in the hNF-L/lacZ transgene bind to nuclear matrix with a lower affinity (52). In this construct, the region having the highest affinity for nuclear matrix is localized at the 3Ј-UTR of the lacZ FIG. 3. In vitro nuclear matrix binding assays. Binding assays were performed by incubating labeled DNA probes with nuclear matrices purified from adult C3H mouse brain and liver. Increasing concentration of E. coli genomic DNA (125-1,000 g/ml) were used as unspecific competitor with or without poly(dA⅐dT) (250 g/ml). Isolated DNA from the pellet fraction was run on agarose gel in parallel with 20% of input DNA probe (lanes 1) and autoradiographed (A, C, E). Under each panel, a schematic representation of the constructs summarizes the binding pattern. The thickness of the band underlining the construct region is indicative of the relative attachment in the binding assays (B, D, F). A, the plasmid phNF-L/CAT construct was digested with HindIII, ScaI, and BamHI restriction enzymes. DNA fragments were labeled and served as probes. This digestion pattern generated bands at 320 bp (band 1), 657 bp (band 2), and 972 bp (band 3) and 2.8 kb corresponding respectively to the hNF-L promoter, the CAT gene, the SV40 polyadenylation signal, and the pUC-9 plasmid. C, the plasmid phNF-L/lacZ was digested with HindIII, EcoRI, and BamHI restriction enzymes prior to end-labeling. Bands at 320 bp (band 1), 3.7 kb, 450 bp (band 2), 750 bp (band 3), and 2.95 kb correspond respectively to the hNF-L promoter, the lacZ gene, the 5Ј segment of the 3Ј-UTR, the 3Ј segment of the 3Ј-UTR, and the Bluescript plasmid. E, the plasmid phNF-L/intronless was digested with EcoRI, BglII, and XbaI restriction enzymes prior to end-labeling. This generate bands at 1,120 bp (band 1), 1,220 bp, 485 bp (band 2), and 2.95 kb corresponding respectively to the hNF-L promoter and first exon, mNF-L exons 2, 3, and 4, the hNF-L 3Ј-UTR, and the Bluescript plasmid.
reporter gene. In addition, these experiments enabled us to identify a novel MAR element in the 3Ј-UTR of the hNF-L gene. By sequence comparison between the hNF-L/lacZ and hNF-L/ intronless MARs, we have tried to highlight similarities that could not be found in the hNF-L/CAT transgene. No highly conserved sequence motif was found that could explain their common neuronal expression pattern, except for their A/T-rich sequences.
One intrinsic property of A/T-rich sequences is that they unpair easily (53). DNA single-stranded regions are formed by opening of the double helix under torsional stress in different paranemic structures (reviewed in Ref. 54). Sequences composed of a minimum of 15-20 consecutive purines on one strand have a tendency to present that state (55)(56)(57)(58)(59)(60)(61). In vivo, unpaired sequences can be identified by digestion of nuclei with S1 nuclease. This generates a pattern of single-strand nucleasesensitive sites. This mapping assay done on different gene loci has revealed that single-strand nuclease-sensitive sites are generally found in promoter regions (62,63), within genes (59,64), or at the 3Ј end (65). Their occurrence on chromatin coincides with a region of transcriptional activity (66 -68). It has been demonstrated that DNA strand unpairing could be induced by supercoiling and that this phenomenon is independent of external protein factor for its initiation (66). This characteristic has allowed us to directly test the presence of unpaired sequences in supercoiled plasmids containing the different MARs. Sequences with unwinding properties were present only in plasmids containing NF-L fusion constructs correctly expressed in the nervous system of transgenic mice. It was found within the associated MAR and colocalized with polypurine sequences.
Position Effect-Position effect is characterized by the influence on transgene expression of regulatory elements placed at the vicinity of the insertion site. Using an enhancer trap vector, it has been demonstrated that insertion near such elements is a rare event because only a weak proportion of transgenic mice lines expressed the reporter gene (69). Recently there has been much evidence supporting the idea that MARs can establish a domain of independent gene activity and insulate a reporter gene from chromosomal position effects in transgenic animals (18,26).
The hNF-L/intronless and the hNF-L/lacZ transgenes are basically not susceptible to insertion site interference for their tissue-specific expression, 2 whereas the hNF-L/CAT construct is highly sensible to position effect (1). A conceivable explanation for the different expression patterns of these hNF-L/reporter gene constructs in transgenic mice is that the different transgenes are in different chromatin conformations, the hNF-L/intronless and the hNF-L/lacZ transgenes being in independent loop domains while transcription of the hNF-L/CAT transgene is dependent on the integration site because it does not carry the information for self-modulation of chromatin structure.
Unwinding elements within the MARs were found only in the two NF-L transgenes correctly expressed in the nervous system. Our results show that when this element is deleted from the lacZ reporter gene, only two out of seven transgenic embryos did express the transgene but in a nonspecific fashion in contrast to 6 out of 7 transgenic embryos expressing correctly the complete hNF-L/lacZ construct. 2 This indicates that removal of the unwinding element made expression of the lacZ transgene insertion-dependent. Using a lacZ transgene with a deleted 3Ј end, Allen et al. (69) have shown that one out of six transformed embryonal stem cells expressed a hsp/lacZ transgene, this weak proportion of expressing versus unexpressing cells being caused by rare insertion events near activating sequences.
Tissue Specificity-Expression patterns observed with the hNF-L/lacZ and hNF-L/intronless transgenes were characteristic of the endogenous NF-L pattern of expression. Without the unwinding element of an MAR, the hNF-L promoter is probably too weak to overrule chromatin-induced repression as sug- FIG. 4. Localization of unwinding elements in each hNF-L/ reporter gene construct. Plasmids phNF-L/intronless, phNF-L/lacZ, and phNF-L/CAT were tested for presence of unwinding elements. 25 g of supercoiled plasmid was treated with 0, 1, 5, and 10% of chloracetaldehyde (lanes 1-4). After linearization of each vector at the 5Ј of the promoter region with proper restriction enzyme, plasmids were digested with the SI nuclease to remove single-stranded regions. Digestion products were separated by electrophoresis, hybridized with a 307-bp probe recognizing the hNF-L promoter, and autoradiographed.  1 (a, b, and c) and embryos 2 (d, e, and f) were analyzed 13.5 days after microinjection, and pictures were taken from the side (a and d), the back (b and e), and the front (c and f ). Positive tissues are identified by an X. These two embryos correspond to different insertion events. gested by expression of the hNF-L/CAT transgene. We suggest that MARs with unwinding elements can function as domain boundary and generate an independent domain in which a weak promoter, such as the hNF-L basal promoter, is excluded from the environment influence. In this chromosomal context, a promoter is able to drive its reporter gene expression according to regulatory elements contained within this domain.
The same phenomenon is likely to be observed with other basal tissue-specific promoters. A good strategy in assessment of a promoter transcriptional characteristic would be to test this promoter with and without MAR with unwinding elements. The lacZ reporter gene from pCH110 is a good choice because the unwinding element influence can be removed easily by a BamHI digestion. The addition of MARs with DNA unwinding elements should be considered when designing transgene or viral vectors because it might circumvent some of the expression problems usually encountered.