Gene Therapy (1999) 6, 616–628  1999 Stockton Press All rights reserved 0969-7128/99 $12.00 http://www.stockton-press.co.uk/gt

Design of a muscle cell-specific expression vector utilising human vascular smooth muscle ␣-actin regulatory elements M-C Keogh1, D Chen1, JF Schmitt1, U Dennehy1, VV Kakkar1 and NR Lemoine1,2 1 Thrombosis Research Institute, London; and 2Imperial Cancer Research Fund, Molecular Oncology Unit, Imperial College School of Medicine, London, UK

The facility to direct tissue-specific expression of therapeutic gene constructs is desirable for many gene therapy applications. We describe the creation of a muscle-selective expression vector which supports transcription in vascular smooth muscle, cardiac muscle and skeletal muscle, while it is essentially silent in other cell types such as endothelial cells, hepatocytes and fibroblasts. Specific transcriptional regulatory elements have been identified in the human vascular smooth muscle cell (VSMC) ␣-actin gene, and used to create an expression vector which directs the

expression of genes in cis to muscle cells. The vector contains an enhancer element we have identified in the 5⬘ flanking region of the human VSMC ␣-actin gene involved in mediating VSMC expression. Heterologous pairing experiments have shown that the enhancer does not interact with the basal transcription complex recruited at the minimal SV40 early promoter. Such a vector has direct application in the modulation of VSMC proliferation associated with intimal hyperplasia/restenosis.

Keywords: Human VSMC ␣-actin; enhancer; gene therapy; tissue-specific expression; luciferase; muscle

Introduction A commonly used and effective treatment for severe symptomatic vessel narrowing (stenosis) induced by a cardiovascular disease such as advanced atherosclerosis is balloon angioplasty. However, in 20–50% of cases the lesion recurs necessitating repeat treatment.1,2 The economic costs of restenosis are substantial: a 33% reduction in the restenosis rate could save $600 million per year in US health care costs,3 with equivalent savings in the rest of the developed world. The exact cause of restenosis is unclear although it has been suggested that exposure of the procoagulant smooth muscle layer by the percutaneous transluminal coronary angioplasty (PTCA) procedure induces a plethora of mitogenic stimuli. The net result of this is the inappropriate migration of vascular smooth muscle cells (VSMCs) from the media to the intima of the damaged vessel where they synthesise extracellular matrix and proliferate, resulting in a gradual reocclusion of the vessel.4–6 A genetically directed therapy that could selectively ablate the proliferating VSMCs in the lesion, to prevent the intimal hyperplasia associated with restenosis while allowing healing of the endothelium, is an appealing concept. The design of specific expression vectors requires the identification of the transcriptional regulatory elements which control a VSMC-specific gene. These can then be employed for directing the expression of genes Correspondence: M-C Keogh at his current address: Bldg C1, Room 210, Department of Biological Chemistry and Molecular Pharmacology, Harvard Medical School, 240 Longwood Avenue, Boston, MA 02115, USA Received 1 September 1998; accepted 11 November 1998

with cytotoxic or cytostatic potential, such as herpes simplex virus thymidine kinase (HSVtk),7 the homeobox gene Gax,8 the cyclin dependent kinase (cdk) inhibitors p21, p27 and p579,10 or constitutively active forms of the retinoblastoma gene product Rb.11 The number of candidate genes which are expressed solely in VSMCs is limited: the VSMC-specific isoforms of myosin light chain, caldesmon, vinculin and meta-vinculin are produced by alternative splicing of genes that are expressed in many cell types.12,13 This makes their transcriptional elements inappropriate candidates for driving the expression of heterologous genes in VSMCs. Other potential candidates are the genes encoding VSMC ␣-actin (also known as aortic type ␣-actin),14 smooth muscle myosin heavy chain,15 h1-calponin,16 telokin,17 and SM-22␣18,19 each of which represent unique gene products. Of these, the VSMC ␣-actin gene encodes the most abundant protein in mature VSMCs. A high level of expression of this contractile protein is necessary for the large force-developing capabilities of the VSMC.20 Although the gene is transiently expressed in developing cardiac and skeletal muscle, and in myofibroblasts within tumours and healing wounds,21,22 in adults it is constitutively expressed solely in VSMCs and smooth musclerelated cells.23 A number of elements controlling the transcription of VSMC ␣-actin have been defined in the chicken,24,25 rat,26,27 mouse28–30 and human31–33 (schematic in Figure 1a). The sequences immediately flanking the VSMC ␣-actin TATA box are highly conserved in human (89% Vs rat), mouse (98% Vs rat), rat and chicken (73% Vs rat) between approximately −255 and +12.27 This region contains a number of highly conserved transcription fac-

Design of a muscle-specific expression vector M-C Keogh et al

617

Figure 1 The VSMC ␣-actin gene 5⬘ flanking region. (a) Schematic figure depicting the elements characterised 5⬘ of the VSMC ␣-actin gene. Areas which have been reported to contain transcriptional activity in transient reporter assays are indicated with the elements supposed to be responsible for this activity, if known. Although the focus of the work in this study examined the human VSMC ␣-actin 5⬘ flanking region, studies from the mouse, rat and chicken genes are also noted. (b) Schematic of −1737hLuc showing restriction enzyme sites employed for pairing analyses. The HindIII and MluI sites are not present in the endogenous gene, but have been engineered into the construct by PCR. C, chicken; Ex, Exon; H, human; M, mouse; R, rat; Luc, luciferase; TSSE, tissue-specific silencer element.

tor (TF) binding motifs, including two CArG boxes, a GArC box and one (human) or two (rat, mouse, chicken) E boxes which have been shown to be essential for activity in the rat.27 Another area of conservation in the sequences flanking the VSMC ␣-actin gene lies in the first intron (+1039/+1147).33 This area contains a consensus CArG box which partially overlaps a polyoma-virus enhancer-like sequence. The syntenic human form has been shown to act as a transcriptional enhancer when linked to an SV40 promoter.33,34 Linkage of the first intron to approximately 600 bp of the 5⬘ flanking region has been reported to drive tissue-specific expression in transgenic mice in a pattern similar to the endogenous gene.34 Transcription analyses of the 5⬘ flanking regions of the murine and rat VSMC ␣-actin genes suggest that domains conferring cellular specificity for these genes lie within approximately 800 bp of the transcription start site.27,28,30

Results The human VSMC ␣-actin 5⬘ flanking region contains two transcriptional repressors and a transcriptional activator The 5⬘ flanking region (−1737/−1) of the human VSMC ␣-actin gene was PCR amplified and ligated 5⬘ of the luciferase reporter gene and a unidirectional deletion series created from this parent clone (−1737hLuc) as in Materials and methods. Since the endogenous VSMC ␣-

actin gene is VSMC specific, the deletion series created was assayed for transcriptional activity in both VSMCs and non-VSMCs. This would allow the determination of which regions conferred not only transcriptional activity, but also specificity. The test cells used were primary VSMCs (human, rat and rabbit: VSMC ␣-actinpos), endothelial cells (human umbilical vein (HUVECs) and bovine aortic (BAECs): VSMC ␣-actinneg), and the HepG2 hepatocyte cell line (VSMC ␣-actinneg). The behaviour of the constructs was essentially the same in HUVECs, HepG2s and BAECs in initial experiments (data not shown). Since the latter two cell populations are easier to grow and transfect, they were used as controls for most experiments. These assays indicate three regions with transcriptional activity: two repressors (−1737hLuc:−1200hLuc (5⬘ repressor) and −522hLuc:−462hLuc (3⬘ repressor)) flanking a region with strong transcriptional activity (−1025hLuc:−950hLuc) (Figure 2). A similar functional element to the 5⬘ repressor has also been described in the 5⬘ flanking region of rat VSMC ␣-actin, albeit located closer to the TATA box by approximately 300 bp.27 The possible function of this element is unknown since it represses transcriptional activity in all cells analysed rather than in non-VSMCs alone, as would be expected if the element functioned to restrict non-VSMC expression of the VSMC ␣-actin gene. The 3⬘ repressor has been previously described.31 A strong transcriptional activator is located between

Design of a muscle-specific expression vector M-C Keogh et al

618

Figure 2 Deletion analyses of human VSMC ␣-actin 5⬘ flanking region. Luciferase activity derived from each deletion construct in VSMCs (rabbit, rat and human) and non-VSMCs (HepG2). Positive control (pGL3 Control) and negative control (pGL3 Basic) are also shown. The scale of all graphs is truncated to 1000 counts of luciferase enzymatic activity to allow direct comparison. The figure ± s.e.m. above the pGL3 Control sample is the activity derived from this construct in the analysis. Cells were transfected and assayed as described. The data show a representative analysis from at least six independent experiments, each experiment performed in triplicate with two preparations per plasmid construct. To facilitate direct comparison between lineages, data are expressed at the observed RLU from each construct relative to that obtained from pGL3 Control, which is given a value of one. The original pGL3 Control RLU for each lineage are: HepG2 1501 ± 236.5, Human VSMC 992 ± 158.9, Rat VSMC 745.6 ± 215.3, Rabbit VSMC 1852 ± 256.7. Luc, luciferase; RLU, relative light units; VSMC, vascular smooth muscle cells.

constructs −950hLuc and −1025hLuc, with most, if not all, contained within the −999hLuc construct. The deletion construct −950hLuc, missing the most 5⬘ 50 bp relative to −999hLuc shows a relative decrease in VSMC transcriptional activity, indicating that this 50 bp contains some transcriptional activity in itself. −999/−890 Behaves as a transcriptional enhancer in VSMCs −999hLuc has maximal transcriptional activity of all deletion constructs studied while the removal of the region −999/−890 to give −890hLuc results in a dramatic reduction in activity (Figure 2). It was investigated whether the region −999/−890 can act as an enhancer. Copies of the region were cloned 5⬘ of the −999hLuc construct (full length (FL) monomer) to derive FL dimer and FL trimer. A pictorial representation of each construct is shown (Figure 3a), with the luciferase activity derived from each in transient transfection transcription assays (Figure 3b). The FL dimer construct supports an increase in the transcription level relative to the FL monomer clone, with the FL trimer clone showing a further increase. This result fulfils criteria for a transcriptional enhancer, namely: (1) copy dependence, since transcription derived from trimer ⬎ dimer ⬎ monomer, and (2) orientation independence, as the orientation of the most 5⬘ copy of −999/−890 in the FL trimer clone is reversed in orientation, but still

supports an increase in transcriptional activity relative to the FL dimer clone. Paired Student’s t tests were applied to the rabbit and rat VSMC data and significance determined as a probability that two means were different at the P ⬍ 0.05 level. Comparison of the data internally (eg ID monomer versus ID dimer or ID trimer, etc) or of the respective multimer across groups (eg FL monomer versus ID monomer) show the difference to be significant in all cases (P ⬍ 0.05). The identification of a transcriptional enhancer is invaluable for the design and creation of expression vectors as it allows the user greater control over the expression level of genes cloned in cis. Thus the gene destined for expression can be cloned into constructs containing increasing enhancer copy number for increased levels of expression.

The 3⬘ repressor (−890/−252) is a tissue-specific silencer element (TSSE) The 3⬘ repressor as observed in the deletion studies has previously been described as a tissue specific silencer element (TSSE).31 The identification and utilisation of such a region is important for the creation of transcriptionally targeted vectors. Removal of the region −890/−252 from the deletion clone −999hLuc (FL monomer) results in the creation of an internal deletion construct (ID monomer) (Figure 3a). The

Design of a muscle-specific expression vector M-C Keogh et al

619

Figure 3 The transcriptional activator (−999/−890) behaves as a transcriptional enhancer while the region −890/−252 is involved in tissue specificity. (a) A pictorial representation of the orientation of the region −999/−890 in each construct is shown for clarity. The Internal Deletion monomer (ID monomer) was created from the parent −999hLuc (Full Length (FL) monomer) by restriction enzyme digestion of internal EcoRI and PstI sites (Figure 1b), followed by the creation of blunt ends with Klenow and religation. The −252hLuc deletion construct, without the −999/−890 region, is also shown. Multimers of the −999/−890 region were created by linking this area as a blunted EcoRI/EcoRV fragment to the linearised and blunted ID monomer construct as appropriate. The orientation of the region −999/−890 is indicated in all multimer clones and identical in the ID and FL monomer, dimer and trimer forms. (b) Luciferase activity derived from each construct in VSMCs (human, rabbit and rat) and non-VSMCs (HepG2) is shown. Experiments were also performed in BAECs which behaved in a similar manner to HepG2 cells (data not shown). Cells were transfected and assayed as described. The data are derived from at least six independent experiments, each experiment performed in triplicate with two preparations per plasmid construct. To facilitate direct comparison between lineages, data are expressed at the observed RLU from each construct relative to that obtained from pGL3 Control, which is given a value of one. The original pGL3 Control RLU for each lineage are: HepG2 2009 ± 237.2, Human VSMC 1027 ± 92.5, Rat VSMC 1738.2 ± 79.2, Rabbit VSMC 1200 ± 123.7. A paired Student’s t test was applied to the rabbit and rat VSMC data and significance determined as a probability that two means were different at the P ⬍ 0.05 level. Comparison of the data internally (eg ID monomer versus ID dimer or ID trimer, etc) or of the respective multimer across groups (eg FL monomer versus ID monomer) show the difference to be significant in all cases (P ⬍ 0.05). FL, full length; ID, internal deletion; Luc, luciferase; ND, not done; RLU, relative light units; VSMC, vascular smooth muscle cells.

addition of the transcriptional enhancer (−999/−890) to ID monomer gives ID dimer and ID trimer with orientation of the region similar to the corresponding FL multimers (arrowed). The ID multimer series supports a progressively increasing level of transcription in VSMCs. However, a concomitant increase is also seen in HepG2 cells which is more pronounced with increasing copy number of the enhancer. These findings have three implications: (3) it supports the proposal that a TSSE is located within the deleted region which functions to repress non-VSMC expression; (2) it demonstrates that these elements are absolutely required for the design of a VSMC-specific expression vector; and (3) it further identifies the −999/−890 region as a transcriptional enhancer as it exhibits position independence; the region is located approximately 638 bp closer to the promoter and proximal 5⬘ flanking region than in the FL opposed to the ID series yet still displays transcriptional activity.

Coupling of FL monomer −999h to the SV40 enhancer abrogates specificity The FL monomer construct, −999hLuc, demonstrates a strong level of VSMC-specific transcriptional activity (see

above). It was investigated whether elements within the −999h region were able to confer specificity when paired in the context of the SV40 viral enhancer. These experiments were performed in an attempt to take advantage of the high transcriptional activity supported by the viral enhancer in the context of the specificity and activity supported by the VSMC ␣-actin enhancer. The human VSMC ␣-actin deletion fragments −999hLuc (FL monomer), −128hLuc and the rat VSMC ␣-actin fragment −130rLuc were cloned into the main polylinker of the commercially available pGL3 enhancer vector (contains the SV40 72bp repeat enhancer 3⬘ of the luciferase gene), to derive the constructs −999hELuc, −128hELuc and −130rELuc respectively (Materials and methods) (Figure 4). These constructs were then transfected as before into a panel of VSMCs and non-VSMCs. The resulting chimeric constructs show a general increase in the level of transcription but a loss of VSMC specificity (see HepG2 cells) relative to their respective parents, which is most marked in the case of −999hELuc versus −999hLuc. This implies: (1) that transcription factors (TFs) bound to the VSMC ␣-actin promoter fragments are capable of interacting with TFs bound to the

Design of a muscle-specific expression vector M-C Keogh et al

620

Figure 4 Coupling of FL monomer −999hLuc to the SV40 enhancer abrogates VSMC specificity. To pair with the SV40 enhancer the human VSMC ␣-actin deletions described (−999hLuc, −128hLuc; −130rLuc) were directionally cloned into the main polylinker of the pGL3 Enhancer vector (Promega). Luciferase activity derived from each construct in VSMCs (human, rabbit and rat) and non-VSMCs (HepG2) is shown. Positive control (pGL3 Control), negative control (pGL3 Basic) and pGL3 Enhancer alone are also shown. Cells were transfected and assayed as described. The data are derived from at least six independent experiments, each experiment performed in triplicate with two preparations per plasmid construct. To facilitate direct comparison between lineages, data are expressed at the observed RLU from each construct relative to that obtained from pGL3 Control, which is given a value of one. The original pGL3 Control RLU for each lineage are: HepG2 1440 ± 136.5, Human VSMC 1092 ± 158.9, Rat VSMC 845.6 ± 215.3, Rabbit VSMC 1452 ± 286.7. Luc, luciferase; RLU, relative light units; VSMC, vascular smooth muscle cells.

SV40 enhancer; and (2) any tissue-specific elements contained within the human VSMC ␣-actin region present are incapable of overcoming the strong, non-specific transcription driven by the SV40 enhancer.

Heterologous coupling of VSMC ␣-actin enhancer containing sequences to SV40 promoter shows that no functional interactions occur Enhancers are distinguished by their ability to act over many kb 5⬘ or 3⬘ of the gene they regulate. For this reason, a standard method for the isolation of elements with enhancer activity is by cloning areas of interest in the context of a defined minimal viral promoter. It is then expected that the TF complex recruited at the putative enhancer will interact with the basal transcription complex recruited at the viral promoter, resulting in a concomitant increase in the level of transcription. In our case, these experiments would also determine whether the human VSMC ␣-actin enhancer (−999/−890) could restrict the transcription derived from unrelated promoters to the VSMC lineage. Such a property could have application if pairing to regulatable promoters was per-

Figure 5 Heterologous coupling of VSMC ␣-actin enhancer containing elements to SV40 promoter shows that no functional interactions occur. (a) A linear illustration of the parent pGL3 promoter plasmid is shown, with the major transcriptional or functional units depicted. The elements beneath depict the regions of the VSMC ␣-actin 5⬘ flanking region cloned into the distal MCS of the vector. The orientation of each element is shown (direction of arrow), with the shaded portion indicating the location of the transcriptional enhancer (−999/−890). The naming of the constructs is explained in the text. (b) Luciferase activity derived from each construct in rat VSMCs and non-VSMCs (HepG2) is shown. Cells were transfected and assayed as described. BAECs were also examined with similar findings to HepG2 cells. The data are a representative experiment from the minimum of at least three performed. Each construct was transfected in triplicate with two preparations per plasmid construct. To facilitate direct comparison between lineages, data are expressed at the observed RLU from each construct relative to that obtained from pGL3. Promoter, which is given a value of one. The original pGL3 Promoter RLU for each lineage are: HepG2 67 ± 12.3, Rat VSMC 136.3 ± 16.6. Amp, ampicillinr; BAEC, bovine aortic endothelial cells; Luc, luciferase; MCS, multiple cloning site; RLU, relative light units; VSMC, vascular smooth muscle cells.

formed, such that the resulting expression cassette would be VSMC specific and responsive to an external stimulus. For these reasons, the enhancer region −999/−890 was cloned cis of the minimal SV40 early promoter contained within the commercially available pGL3 Promoter vector (Promega). Subclones were also created with the region −999/−890 and either or both of its 5⬘ or 3⬘ flanking regions to determine the nature of the 5⬘ and 3⬘ sup-

Design of a muscle-specific expression vector M-C Keogh et al

pressors (Figure 5a). pGL3 Promoter contains two multiple cloning sites (MCS), the distal (d-) and proximal (p), so named by their proximity to the minimal SV40 promoter (Figure 5a), and both were utilised for these experiments. The naming of the constructs is as follows: as an example, the construct containing the transcriptional enhancer (−999/−890, EcoRV/EcoRI) in the distal MCS of pGL3 Promoter in the forward orientation is known as d-Prom R5/R1-F. Other constructs are named according to a similar system. The resulting constructs produced a level of transcription similar to the pGL3 Promoter parent clone (Figure 5b). This finding was independent of the nature of the insert (enhancer ± flanking regions), the orientation of the insert (forward or reverse), or the plasmid MCS utilised for cloning (distal or proximal). The presence of the additional 5⬘ or 3⬘ repressor elements, either alone or together, did not modify the transcriptional activity derived from the SV40 promoter. This indicates that the TF complex recruited at the enhancer contained within the region −999/−890 is unable to interact with the basal transcription complex recruited at the SV40 promoter.

Coupling of VSMC ␣-actin enhancer containing sequences to −128h or −130r results in functional interactions The experiments above show that the enhancer-recruited TF complex is unable to interact with the basal transcription complex recruited at the SV40 promoter. A possible reason for this is that the enhancer complex does not interact with the RNA Polymerase II/TFIID basal transcription complex directly, but rather through an intermediate. An ideal candidate for this intermediate are TFs recruited to the upstream regulatory elements (URE) located immediately 5⬘ of the VSMC ␣-actin promoter, or more specifically the CArG boxes contained in this area. To examine this possibility, a similar cloning procedure to that above was undertaken but with either −128hLuc or −130rLuc as the parent construct (Figure 6a). The naming of the constructs is as follows: as an example, the luciferase reporter construct containing the −128h region (−128Luc) paired with the AccI restriction enzyme fragment (−1296/−674) in the forward orientation is known as −128h:AccI-F. Other constructs are named according to a similar system. The resulting constructs, all of which contain the enhancer region −999/−890, support an increased level of transcription relative to the parents −128hLuc or −130rLuc alone in VSMCs (rabbit and rat) (four- to six-fold) (Figure 6b). A similar fold increase in transcription supported by the constructs relative to the parent clones is also observed in non-VSMCs (HepG2s and BAECs), although since these are measured from a lower base line, the actual transcription level is much reduced relative to VSMCs. This transcriptional enhancement is irrespective of the orientation or size of the flanking region. The minimal transcriptional activation in non-VSMCs is not unexpected and recapitulates other findings within this work. The finding that the elements are able to pair with the −128h or −130r promoters, in contrast to their inability to interact with the SV40 promoter is intriguing however. The cloning vector pCI/h999 is muscle specific The experiments above have described the activity of the VSMC ␣-actin elements in a reporter plasmid. An

expression plasmid (termed pCI/h999) was created with these elements linked to a large MCS to facilitate the cloning of heterologous genes for VSMC-targeted expression. The expression plasmid was based on the commercially available pCI expression vector (Promega), containing the CMV promoter and enhancer. The preparation of the pCI/h999 cloning vector and luciferase, ␤-gal and GFP containing forms is described in Materials and methods. A number of different lineages were tested for their ability to support expression derived from the pCI/h999 expression vector containing the luciferase reporter gene (pCI/h999Luc). These include VSMCs (rat and rabbit), hepatocytes (human, Hep3B), ECs (human, ECV304 and bovine, BAEC), cardiac myocytes (rat neonatal), skeletal muscle (rat L6, either in the form of myoblasts or differentiated myocytes) and fibroblasts (human lung, CCD32 and mouse, DAT3). The data are expressed as foldluciferase activity derived from pCI/h999Luc relative to the non-specific CMV-driven pCILuc. The first generation vector pCI/h999 supports expression solely in muscle lineages (Figure 7a). It is thus not VSMC specific (although relative gene expression is supported to a greater extent in VSMCs), with expression also seen in cardiac and skeletal muscle. This is not unexpected as the VSMC ␣-actin 5⬘ flanking cassette utilised contains numerous TF binding motifs known to be involved in the regulation of muscle genes including E boxes and CArG boxes. Gene expression is not supported in the other lines tested. This finding is repeated with the use of the reporter genes GFP (Figure 7b) and ␤-gal (data not shown), showing that the VSMC ␣-actin elements are able to confer muscle cell specificity on heterologous genes other than luciferase.

Discussion The goal of this study was to identify the elements involved in conferring cell-specific transcriptional activity of the human VSMC ␣-actin gene for exploitation in targeted gene therapy constructs for the treatment of intimal hyperplasia. In addition to creating a plasmid vector with possible application in this field, our results have determined that human VSMC ␣-actin transcription requires a complex interplay between three transcription factor complexes: those recruited at the TATA box, the CArG boxes and the enhancer element. Deletion analyses indicate three regions with transcriptional activity: two repressors flanking a region with strong transcriptional activity (Figure 2). The first of these repressors (5⬘ repressor) is a novel finding and its possible function is unknown since it suppresses transcriptional activity in all cells analysed rather than in non-VSMCs alone. It is possible that the silencing activity is not a true biological effect but a function of the method of analysis. Transient transfections such as these do not represent the physiological scenario where histone proteins and architectural elements of tertiary structure, such as scaffold attachment regions (SARs) are involved. An indicator of an SAR is the presence of AT-rich elements,35,36 and such a region is located 5⬘ of −1025hLuc. It is possible that complex secondary structure resulting from internal homology is formed at this region, preventing TF ingress, and thus transcriptional activation. The human VSMC ␣-actin 5⬘ flanking region −999/−890 functions as an enhancer, demonstrating copy

621

Design of a muscle-specific expression vector M-C Keogh et al

622

Design of a muscle-specific expression vector M-C Keogh et al

dependence as well as position and orientation independence when linked to the VSMC ␣-actin promoter. However, the enhancer does not have any strong homology with the syntenic region from the rat, mouse and chicken, implying that the element is not utilised in these species in this form. The experimental analyses detailed here show that the element is functional in human, rat and rabbit VSMCs, indicating that a broadly similar panel of transcriptional factors are present. It is possible that an enhancer similar to that contained within −999/−890 of the human sequence is present in the other species, but in another area. Alternatively, it is possible that the enhancer we have described is specific to human VSMC ␣-actin. Another element with enhancer activity has been described to lie within a conserved area in the first intron of the VSMC ␣-actin gene.33,34 This intronic enhancer contains a conserved CArG box and overlapping polyoma virus enhancer-like sequence,33,37 although these have not been shown to be required for activity. It should be noted that this enhancer can pair effectively with an SV40 promoter, and it is possible that the CArG box in this case is recapitulating the effect of the CArG boxes contained immediately 5⬘ of the VSMC ␣-actin TATA box. Intriguingly, attempts thus far to exploit the VSMC ␣actin elements to direct tissue-specific expression when paired with heterologous viral promoter/enhancer elements have proven unsuccessful (Figure 5). These experiments were performed to determine whether the VSMC ␣-actin enhancer could restrict the transcription derived from unrelated promoters to the VSMC lineage. Pairing experiments with the SV40 enhancer and the FL monomer construct −999hLuc show that while this region from VSMC ␣-actin might confer muscle-specific expression on the reporter gene luciferase when used alone, the recruited TFs which confer specificity are unable to overcome the strong non-specific transcription driven by the SV40 enhancer in cis. It is possible that −999hLuc does not recruit any TFs which act directly as suppressors in non-VSMCs and the absence of expression in non-VSMCs is due to incomplete complex recruitment in these cell types. In this scenario, the SV40-recruited enhancer complex simply interacts with the TATArecruited basal transcription complex in HepG2 cells, causing transcriptional activation. The complex assembled at the SV40 early promoter in Figure 6 Coupling of VSMC ␣-actin enhancer containing elements to −128hLuc or −130rLuc results in functional interactions. (a) A linear illustration of the parents −128hLuc or −130rLuc is shown. The elements beneath depict the regions of the VSMC ␣-actin 5⬘ flanking region cloned into the distal MCS of the vector, cis of the human or rat VSMC ␣actin promoters as appropriate. The orientation of each element is shown (direction of arrow), with the shaded portion indicating the location of the transcriptional enhancer (−999/−890). The naming of the constructs is explained in the body text. (b) Luciferase activity derived from each construct in rat VSMCs and non-VSMCs (HepG2) is shown. Rabbit VSMCs and BAECs were also examined with similar findings. Cells were transfected and assayed as described. The data are a representative experiment from the minimum of at least three performed. Each construct was transfected in triplicate with two preparations per plasmid construct. To facilitate direct comparison between lineages, data are expressed at the observed RLU from each construct relative to that obtained from pGL3 Promoter, which is given a value of one. The original pGL3 Promoter RLU for each lineage are: HepG2 57.1 ± 3.6, Rat VSMC 103.3 ± 11.3. BAEC, bovine aortic endothelial cells; F, forward; Luc, luciferase; MCS, multiple cloning site; RLU, relative light units; R, reverse; VSMC, vascular smooth muscle cells.

pGL3 promoter differs from that assembled on the −128hLuc or −130rLuc constructs in two regards: (1) the SV40 early promoter recruits the basal transcription complex by a grouping of six Sp1 boxes (GGGCGG),38 while the latter two utilise a TATA box; and (2) the latter two promoters have the additional presence of CArG boxassociated transcription factors. Previous work has shown that different promoters, or more specifically promoter sequences, can recruit different basal transcription complexes,39,40 and it is possible that the basal transcription complex recruited to the TATA box in −128h and −130r is different from that recruited at the Sp1 boxes in SV40 promoter. Alternatively, it is possible that the VSMC ␣-actin enhancer mediates its contacts, and thus its effects, with the basal transcription complex through the CArG-associated TFs rather than directly to the complex itself. In this regard it is interesting that the previously described VSMC ␣actin intronic enhancer which contains an associated CArG box can pair effectively with an SV40 promoter.33,37 It is possible that the CArG box in this case recapitulates the effect of the CArG boxes located immediately 5⬘ of the VSMC ␣-actin TATA box. It will be important to determine which of the above proposals (ie either the VSMC ␣-actin promoter recruits a unique basal transcription complex, or the CArG boxes are required for enhancer interactions with the basal transcription complex) are correct for the design of next generation cloning/expression vectors based on VSMC ␣actin. For example, if it is possible to confer VSMC specificity on a regulatable promoter (eg one responsive to cAMP41), then the resulting vector design will have to contain the VSMC ␣-actin enhancer, multimers of cAMP response elements (CREs) and then either the VSMC ␣actin promoter (if the VSMC ␣-actin enhancer interacts with a unique basal transcription complex) or an alternate promoter with attached CArG boxes (if CArG boxes facilitate enhancer–basal transcription complex interactions). For the creation of recombinant expression vectors a number of criteria must be fulfilled: there is a possible size restriction on the transgene dependent on the cloning vector employed (eg in plasmid or retroviral vectors this is limited to approximately 9 kb, in E1/E3-deleted adenoviral vectors to approximately 7.5 kB). Since the transgene includes the gene intended for expression and the elements which mediate this transcription, this usually reduces the permissible size of the transcriptional control elements by approximately 1–2 kB. As the elements which control gene expression in vivo may lie over many kb, identification of the functionally active components and their isolation for utilisation is essential. A tissue-specific expression vector must also support a strong level of transcriptional activity, orders of magnitude over that in unrelated tissues, in order for it to be be useful. The benchmark for the transcription level required is of the order of that derived from a strong ubiquitous viral promoter such as SV40 or CMV although elements have been utilised which drive tissue-specific expression at much lower levels: an adenovirus/myosin light chain (AdMLC) recombinant has been reported to drive myocardial reporter gene expression at 11% that derived from an AdRSV recombinant.42 The pCI/h999 expression vector created fulfils these criteria: the elements to regulate VSMC-specificity occupy

623

Design of a muscle-specific expression vector M-C Keogh et al

624

Figure 7 The pCI/h999 cloning expression vector is SMC specific. (a) The data are expressed as fold-luciferase activity derived from pCI/h999Luc relative to the non-specific CMV-driven pCILuc in each cell line/lineage transfected. Cells were transfected and assayed as described. The data are derived from at least three independent experiments, each experiment performed in triplicate with two preparations per plasmid construct. Lineages assayed include VSMCs (rabbit and rat), hepatocytes (Hep3B), endothelial cells (BAEC, ECV304), cardiac myocytes (neonatal rat), skeletal muscle (L6 myoblasts and L6 myocytes) and fibroblasts (CCD-32 – human lung, DAT3 – mouse). (b) pCI, pCI-GFP-N1 and pCI/h999-GFP-N1 were transfected into VSMCs (rabbit) and non-VSMCs (Hep3B, ECV304) and the resulting GFP activity assayed 48 h later as described. The scale bar in the bottom left panel is relevant for all. BAEC, bovine aortic endothelial cells; GFP; green fluorescent protein; Luc, luciferase; VSMC, vascular smooth muscle cells.

Design of a muscle-specific expression vector M-C Keogh et al

only 1 kB of the plasmid, taking the total size to about 4.2 kB, allowing the cloning of cDNAs or exon/intron combinations up to approximately 6 kB in size. This capacity is sufficient for most genes of interest. The expression level supported is also very strong, being ⭓40% that derived from the CMV promoter enhancer. It should be noted that only a monomer of the identified VSMC ␣-actin enhancer is present in this vector, so it should be possible to drive this level up still further by multimerising the enhancer (Figure 3) without reducing appreciably the acceptable insert size (the enhancer only occupies approximately 110 bp). Finally, the vector is muscle specific, although not VSMC specific: reporter gene expression in non-muscle is detectable above background, but at very low levels (Figure 7). The expression supported in cardiac, skeletal and vascular smooth muscle cells is not unexpected: the VSMC ␣-actin element utilised contains a number of binding motifs for TFs known to be involved in muscle gene regulation, namely CArG boxes and E boxes. However, since the vector is intended for use in transient transfection of a single anatomical site, namely a stenotic lesion, this is not expected to be a drawback. The cell types likely to be encountered in this location include VSMCs, ECs and fibroblasts (along with recruited T cells and macrophages).6 Of the cells tested which represent the former three of these lineages only VSMCs support transcription. In summary, we have succeeded in isolating an element with transcriptional enhancer activity which regulates the expression of human VSMC ␣-actin. This element has been shown to be active in primary VSMCs from a number of species (human, rat and rabbit), cardiac myoblasts, and skeletal myoblasts/differentiated myocytes, while it has minimal activity in non-VSMCs (ECs, hepatoma cells and fibroblasts). We have also shown the functional presence of two elements flanking this region which repress transcription. Through heterologous pairing experiments, we have shown that the transcriptional enhancer is unable to modulate expression derived from the SV40 minimal promoter, while it is able to pair with the human or rat VSMC ␣-actin TATA box with associated CArG boxes. We propose that this difference could be for either of two reasons: (1) the basal transcription complex assembled at the two promoters is significantly different, or (2) the enhancer does not interact directly with the basal transcription complex recruited at the promoter itself, but rather through an intermediate: the TF complex assembled at the URE, or more specifically the TFs bound at CArG-A and CArG-B. Our observation that the identified enhancer is unable to pair with the SV40 promoter is disconcerting as it indicates that a technique commonly used to identify elements with enhancer activity, namely pairing them with minimal heterologous viral promoters, is not always appropriate. The cloning vector pCI/h999 has been created which confers muscle-specific expression on heterologous genes placed under its control, both in vitro (Figure 7) and in vivo.43 This implies that the human VSMC ␣-actin transcriptional control regions could have future application in the field of gene therapy. Further analyses are currently underway to determine the actual TFs involved in the regulation of the human VSMC ␣-actin gene, initially concentrating on those recruited to the enhancer region. We are also examining

whether the enhancer interacts with the promoter via a CArG-recruited intermediate, or if the interaction requires a specific basal transcription complex uniquely recruited to the VSMC ␣-actin TATA box. Further studies will also examine which regions are responsible for the promoter attenuation in response to external stimuli (such as mediated by angiotensin II, insulin, endothelin1 and TGF␤44–46 and encountered in a restenotic lesion), and could possibly lead to their removal without ablation of tissue-specific transcriptional activity.

Materials and methods Materials Tfx-50 and the luciferase assay system were purchased from Promega UK (Southampton, UK). Lipofectamine was purchased from Gibco BRL (Renfrewshire, UK). Calcium Phosphate transfection kit was purchased from InVitrogen (Leek, The Netherlands). The pGL3 Luciferase vectors (Basic, lacks promoter and enhancer; Control, contains SV40 promoter and enhancer; Promoter, contains SV40 promoter only; Enhancer, contains SV40 enhancer only) were purchased from Promega UK, transformed into E. coli (JM109) and purified on Qiagen-500 columns (Qiagen, Surrey, UK) prior to transfection. Restriction and modifying enzymes were purchased from Boehringer Mannheim (East Sussex, UK) and Promega UK. Sequenase T7 polymerase V2.0 for dideoxy DNA sequencing was purchased from Amersham Pharmacia Biotech (Buckinghamshire, UK). The thermostable DNA polymerase Ultma was purchased from Perkin-Elmer Applied Biosystems (Warrington, Cheshire, UK). Human transferrin was purchased from Sigma Chemicals (Poole, UK). The Protein DC assay system was purchased from Biorad Laboratories (Hertfordshire, UK). Heat-inactivated FCS was purchased from Harlan Sera-labs (Crawley Down, UK). Human arteries were provided by King’s College Hospital Liver Transplant Unit (Denmark Hill, London, UK) or Royal Brompton Hospital (Kensington, London, UK) from organ transplant donors with the consent of relatives. Human umbilical cords were provided by King’s College Hospital Maternity Unit, Denmark Hill, London. Neonatal rat cardiac myoblasts were provided by Dr N Brand (NHLI, London, UK). Cell culture and transfection Primary cells rather than established cell lines were used for these studies wherever possible. Cells were isolated and primary cultures set up as previously described.47 VSMCs from rabbit and rat aorta, or human arteries (aorta, inferior mesenteric or splenic), endothelial cells from human umbilical veins (HUVECs) or bovine aorta (BAECs), rat neonatal cardiac myoblasts (isolated as described48,49), the HUVEC cell line ECV304 (ATCC CRL1998),50 the rat skeletal myoblast line L6 (ATCC CRL1458), the human lung fibroblast line CCD-32Lu (ATCC CRL-1485), the mouse fibroblast line DAT3, and the human HepG2 or Hep3B hepatoma lines were routinely grown in the appropriate medium as described (Table 1). Cultures were maintained in the exponential growth phase at 37°C in a humidified incubator with 5–10% CO2. Cells were transfected using the cationic liposomes Tfx50 (Promega) or lipofectamine (Gibco BRL) or calcium

625

Design of a muscle-specific expression vector M-C Keogh et al

626 Table 1 Optimised transfection and culture conditions for cells used Cell type

Transfection conditions (DNA – ratio – time – serum)

Rabbit VSMCs

Tfx-50: 0.5 ␮g – 4:1 – 1 h – (−)

Rat VSMCs

Tfx-50: 0.5 ␮g – 4:1 – 1 h – (−)

Human aortic VSMCs

Tfx-50: 0.5 ␮g – 4:1 – 1 h – (−)

HUVEC

Tfx-50: 0.5 ␮g – 4:1 – 2 h – (−)

Hep G2/Hep3B

Tfx-50: 0.5 ␮g – 4:1 – 2 h – (+)

BAECs

5 ␮l Lipofectamine, 1 ␮g DNA/ml SF medium, 4 h 5 ␮l Lipofectamine, 16 ␮g transferrin, 1.25 ␮g DNA/ml SF medium, 5 h (10 ␮g DNA, 18 ␮l CaCL2 to 300 ␮l with ddH2O) mixed 1:1 with 300 ␮l 2× HEPEsbuffered saline, O/N As cardiac myoblasts

ECV304 Rat neonatal cardiac myoblast L6 myoblasts L6 myocytes CCD-32Lu DAT3

Transfected as myoblasts, then proliferation replaced by differentiation medium 5 ␮l Lipofectamine, 1.0 ␮g DNA/ml Optimem, 5 h 5 ␮l Lipofectamine, 1.0 ␮g DNA/ml Optimem, 5 h

Culture conditions (medium and passage) M199, 10% FCS, 1% NEAA, 5% CO2. Split 1:2 every 10–14 days DMEM, 10% FCS, 10% CO2. Split 1:5 twice weekly DMEM, 10% FCS, 10% CO2. Split 1:2 once every 14 days, p2–8 M199, 20% FCS, 20 ␮g/ml ECGF, 80 ␮g/ml Heparin, 10% CO2. Split 1:3 twice weekly DMEM, 10% FCS, 10% CO2. Split 1:3 twice weekly DMEM, 10% FCS, 10% CO2. Split 1:3 twice weekly M199 (+ HEPES), 10% FCS, 10% CO2. Split 1:5 twice weekly 4:1 DMEM:M199 (+HEPES), 10% Horse serum, 5% FCS, 5% CO2 Proliferation medium: DMEM 10% FCS, 5% CO2 Differentiation medium: DMEM 2% horse serum, 5% CO2 DMEM, 10% FCS, 5% CO2 DMEM, 10% FCS, 5% CO2

Culture and optimised transfection conditions for the lineages/lines used in this study. All culture media contain penicillin/streptomycin and l-glutamine as standard. All cells were transfected in 24-well plates except rat cardiac myoblasts and L6 cells which were transfected in six-well plates, and the fibroblast lines which were transfected in 12-well plates. DMEM, Dulbecco’s modified Eagle’s medium; ECGF, endothelial cell growth factor; HUVEC, human umbilical vein endothelial cell; NEAA, non-essential amino acids; SF, serum free; VSMC, vascular smooth muscle cells.

phosphate (InVitrogen) as indicated (Table 1). Triplicate cultures transfected with Tfx-50 were treated according to optimal conditions previously described.47 Cells for liposome transfection were plated at 5 × 104 cells per well in 24-well plates 2 days before transfection with the exception of the fibroblast lines which were seeded at 106 cells per well in 12-well plates. On the day of transfection the medium was aspirated and the cells washed twice in the appropriate serum-free medium. After the incubation period indicated, 1.5 ml of complete medium was added per well and cells cultured for 48 h before assaying for luciferase activity. Cells for calcium phosphate transfection (cardiac myoblasts and L6) were plated at 0.4 × 106 cells per well in 6-well plates the day before transfection. Cells were then transfected in triplicate with 10 ␮g DNA per well according to the manufacturer’s instructions. The next day, transfection complexes were removed, and cells washed with complete medium and incubated for a further 48 h before luciferase analysis. L6 myoblasts were differentiated to myocytes following transfection by replacing the proliferation culture medium with differentiation medium (low serum) and incubating for 48 h before analysis. Progression to myocytes was visualised by the appearance of multinucleate cells.

Assay of luciferase activity Luciferase activity was measured according to the manufacturer’s instructions (Promega) as previously

described.47 Data are expressed as the mean of triplicate measurements of luciferase enzymatic activity (in light units) with the standard error of the mean also shown. A minimum of two clones for each construct was prepared and transfected in all the transfection experiments described. This was to reduce the possibility that the transcriptional activity observed was a result of the preparative process. Cotransfection experiments utilising another plasmid-encoded marker were not used due to the possibility that they might compete for limited trans-factors which regulate VSMC ␣-actin transcription. Normalisation to total protein content as assayed by the Protein DC assay system was performed however. Each clone of each construct was assayed enzymatically on a minimum of six separate occasions with an observed experimental variability of less than 20%. Where indicated, statistical analyses were performed on raw data from transfections performed in sextuplet. Paired Student’s t tests were applied to the data with the aid of the Sigmaplot program (Jandel Scientific) and significance was determined as a probability that two means were different at the P ⬍ 0.05 level.

GFP expression Cells 5 × 104 were grown on coverslips in 12- or 24-well plates and transfected as above. Forty-eight hours after transfection with the appropriate GFP vector (pCI-GFPN1 or pCI/h999-GFP-N1), the coverslips overlaid with a

Design of a muscle-specific expression vector M-C Keogh et al

cell monolayer were rinsed briefly in PBS and mounted on a layer of 2% agarose/PBS to prevent drying. Samples were observed using a laser confocal microscope (Biorad MRC 600) with the FITC excitation filter (␭488 nm).

Generation of deletion series The 5⬘ flanking regions of the human or rat VSMC ␣actin genes were generated by PCR. Oligonucleotide primers with MluI (5⬘ primers) or HindIII (3⬘ primers) restriction sites were supplied deprotected and desalted by Pharmacia UK. The fragments (−1737h; 5⬘ flanking region, −1737/−1 of human VSMC ␣-actin gene: −128h; 5⬘ flanking region, −128/−1 of human VSMC ␣-actin gene; −130r 5⬘ flanking region, −130/−1 of rat VSMC ␣actin gene) were PCR amplified with the high fidelity thermostable DNA polymerase Ultma (Perkin-Elmer) from either human HepG2 or rat L6 skeletal myoblast genomic DNA as appropriate. PCR conditions used were 94°C 60 s, 60°C 60 s, 72°C 120 s for 35 cycles. After purification with Wizard columns (Promega) according to the manufacturer’s instructions, the PCR products were digested with MluI and HindIII, purified on a 1% agarose gel and directionally cloned into the promoterless luciferase vector pGL3 basic (Promega) that had been similarly digested. Transformation and screening for positive recombinants were by standard techniques.51 The fidelity of each clone was verified by sequencing. The resulting clones were grown in large scale culture and the plasmid DNA purified using Qiagen-500 columns. The naming of the recombinant plasmid constructs described in this paper is based on a number of criteria: in the case of the deletion series the naming denotes the 5⬘ extent of the clone followed by the letter h (denoting human) or r (rat) and followed by Luc, designating the marker gene Luciferase; thus the large human clone isolated as above is designated −1737hLuc. All clones contain as their 3⬘ extent the nucleotide −1 immediately before the untranslated first exon of VSMC ␣-actin. The heterologous pairing constructs described are named by an amalgam of the two paired fragments as indicated. Unidirectional deletion analysis of −1737hLuc towards the transcription start site was performed at 25°C for a series of time points using ExoIII, with the 5⬘ end protected by generating a 3⬘ overhang using the restriction enzyme SacI (site present in the pGL3 basic plasmid polylinker), while the 3⬘ ExoIII sensitive end was generated with an internal VSMC ␣-actin PvuII site (Figure 1b). The lag strand was removed by digestion with S1 nuclease, followed by treatment with Klenow to fill in the ends, religation and transformation into competent JM109 E. coli. The resulting deletion series was sized by PCR with primers to the pGL3 basic vector. Creation of plasmid constructs Internal deletion constructs were created by restriction enzyme digestion and religation of the indicated parent plasmid using standard techniques.51 The location of the restriction enzyme sites employed is shown in Figure 1b. The promoter pairing experiments utilised either −128hLuc, −130rLuc or pGL3 Promoter as the parent vector, into which was cloned the appropriate element. For the sake of clarity, a more in-depth explanation of the creation of each element is found in the appropriate section, and a diagrammatic representation depicting the cloning procedure is found in the relevant figure.

The pCI/h999 expression vector was created by excising the CMV promoter/enhancer from the pCI expression vector (Promega) using flanking unique BglII and PstI restriction sites, polishing the ends with Klenow and blunt-end ligation of the immediate 1 kb of flanking sequence from the human VSMC ␣-actin gene (as in deletion construct −999hLuc, excised with flanking MluI and EcoRV sites from −1737hLuc, and blunt-ended) in the correct orientation. The luciferase gene cDNA (Luc+) was removed from the pGL3Basic vector (Promega) by restriction enzyme digestion of flanking NcoI and XbaI sites and the ends polished with Klenow. The resulting fragment was bluntend ligated in the correct orientation into the SmaI site of pCI or pCI/h999 to give pCI/Luc and pCI/h999-Luc, respectively. The bacterial LacZ gene was removed from the pSV–␤-galactosidase vector (Promega) by restriction enzyme digestion of flanking HindIII and DraI sites and the ends polished with Klenow. The resulting fragment was blunt-end ligated in the correct orientation into the SmaI site of pCI or pCI/h999 to give pCI-␤-Gal and pCI/h999-␤-Gal, respectively. The GFP (green fluorescent protein) gene was removed from the p-EGFP-N1 N-terminal protein fusion vector (Clontech) by restriction enzyme digestion of flanking XhoI and NotI sites and the ends polished with Klenow. The resulting fragment was blunt-end ligated in the correct orientation into the SmaI site of pCI or pCI/h999 to give pCI-GFP-N1 and pCI/h999GFP-N1, respectively.

Acknowledgements These studies were supported by funding from the Garfield Weston Foundation. We would like to thank the Liver Transplant Unit, Kings College Hospital and the Royal Brompton Hospital for supplying human arteries, the maternity unit for supplying umbilical veins and Dr Nigel Brand, NHLI, for supplying rat cardiac myoblasts. We would like to thank the staff of the Immunohistochemistry lab, TRI for their aid and technical expertise. Finally we thank Drs Helen Hurst, Richard Vile and Nigel Brand for critical reading of the manuscript.

References 1 CABRI. First-year results of CABRI (Coronary angioplasty versus bypass revascularisation investigation). Lancet 1995; 346: 1179–1183. 2 The Bypass Angioplasty Revascularisation Investigation (BARI) Investigators. Comparison of coronary bypass surgery with angioplasty in patients with multivessel disease. New Engl J Med 1996; 335: 217–225. 3 Califf RM. Restenosis: the cost to society. Am Heart J 1995; 130: 680–684. 4 Forrester JS et al. A paradigm for restenosis based on cell biology: clues for the development of new preventative therapies. J Am Coll Cardiol 1991; 17: 758–769. 5 Mattsson E, Clowes AW. Current concepts in restenosis following balloon angioplasty. Trends Cardiovasc Med 1995; 5: 200–204. 6 Casscells W, Engler D, Willerson JT. Mechanisms of restenosis. Texas Heart Inst J 1994; 21: 68–77. 7 Guzman RJ et al. In vivo suppression of injury induced vascular smooth muscle cell accumulation using adenovirus-mediated transfer of the herpes simplex virus thymidine kinase gene. Proc Natl Acad Sci USA 1994; 91: 10732–10736. 8 Maillard L et al. Percutaneous delivery of the gax gene inhibits vessel stenosis in a rabbit model of balloon angioplasty. Cardiovasc Res 1997; 35: 536–546.

627

Design of a muscle-specific expression vector M-C Keogh et al

628

9 Chang MW et al. Adenovirus-mediated over-expression of the cyclin/cyclin-dependent kinase inhibitor, p21 inhibits vascular smooth muscle cell proliferation and neointima formation in the rat carotid artery model of balloon angioplasty. J Clin Invest 1995; 96: 2260–2268. 10 Ohno T et al. Combination gene transfer to potentiate tumor regression. Gene Therapy 1997; 4: 361–366. 11 Chang MW et al. Cytostatic gene therapy for vascular proliferative disorders with a constitutively active form of the retinoblastoma gene product. Science 1995; 267: 518–522. 12 Sobue K, Sellers J. Caldesmon, a novel regulatory protein in smooth muscle and nonmuscle actomyosin systems. J Biol Chem 1991; 266: 12115–12118. 13 Byrne B et al. Chicken vinculin and meta-vinculin are derived from a single gene by alternative splicing of a 207-base pair exon unique to meta-vinculin. J Biol Chem 1992; 267: 12845–12850. 14 Owens GK, Thompson MM. Developmental changes in isoactin expression in rat aortic smooth muscle cells in vivo. Relationship between growth and cytodifferentiation. J Biol Chem 1986; 261: 13373–13380. 15 Nagai R et al. Identification of two types of smooth muscle myosin heavy chain isoforms by cDNA cloning and immunoblot analyses. J Biol Chem 1989; 264: 9734–9737. 16 Maino JM, Olson EN. Expression of the smooth muscle cell calponin gene marks the early cardiac and smooth muscle cell lineages during mouse embyrogenesis. J Biol Chem 1996; 271: 7095–7103. 17 Herring BP, Smith AF. Telokin expression is mediated by a smooth muscle cell-specific promoter. Am J Physiol 1996; 270: C1656–C1665. 18 Nishida W, Kitami Y, Hiwada K. cDNA cloning and mRNA expression of calponin and SM22␣ in rat aorta smooth muscle cells. Gene 1993; 130: 297–302. 19 Moessler H et al. The SM22 promoter directs tissue-specific expression in arterial but not in venous or visceral smooth muscle cells in transgenic mice. Development 1996; 122: 2415–2425. 20 Murphy R. Muscle cells of hollow organs. News Physiol Sci 1988; 3: 124–128. 21 Gabbiani G. The biology of the myofibroblast. Kidney Int 1992; 41: 530–532. 22 Ruzicka DL, Schwartz RJ. Sequential activation of alpha-actin genes during avian cardiogenesis: vascular smooth muscle alpha-actin gene transcripts mark the onset of cardiomyocyte differentiation. J Biol Chem 1988; 107: 2575–2586. 23 Vandekerckhove J, Weber K. Actin typing on total cellular extracts: a highly sensitive protein-chemical procedure able to distinguish different actins. Eur J Biochem 1981; 113: 595–603. 24 Blank RS et al. Elements of the smooth muscle ␣-actin promoter required in cis for transcriptional activation in smooth muscle. Evidence for cell type-specific regulation. J Biol Chem 1992; 267: 984–989. 25 McNamara CA et al. Nuclear proteins bind a cis-acting element in the smooth muscle alpha-actin promoter. Am J Physiol 1995; 268: C1259–C1266. 26 Simonson MS et al. Two proximal CArG elements regulate SM alpha-actin promoter, a genetic marker of activated phenotype of mesangial cells. Am J Physiol 1995; 268: F760–F769. 27 Shimizu RT et al. The smooth muscle ␣-actin gene promoter is differentially regulated in smooth muscle versus non-smooth muscle cells. J Biol Chem 1995; 270: 7631–7643. 28 Foster DN et al. Positive and negative cis-acting regulatory elements mediate expression of the mouse vascular smooth muscle ␣-actin gene. J Biol Chem 1992; 267: 11995–12003. 29 Cogan J et al. Plasticity of vascular smooth muscle alpha-actin gene transcription. Characterization of multiple, single-, and double-strand specific DNA-binding proteins in myoblasts and fibroblasts. J Biol Chem 1995; 270: 11310–11321. 30 Min B, Foster DN. The 5⬘-flanking region of the mouse vascular smooth muscle ␣-actin gene contains evolutionary conserved sequence motifs within a functional promoter. J Biol Chem 1990; 27: 16667–16675.

31 Bushel P et al. Two serum response elements mediate transcriptional repression of human smooth muscle ␣-actin promoter in ras-transformed cells. Oncogene 1995; 10: 1361–1370. 32 Reddy S et al. Structure of the human smooth muscle ␣-actin gene. Analysis of a cDNA and 5⬘ upstream region. J Biol Chem 1990; 265: 1683–1687. 33 Nakano Y et al. Transcriptional regulatory elements in the 5⬘ upstream and first intron regions of the human smooth muscle (aortic type) ␣-actin-encoding gene. Gene 1991; 99: 285–289. 34 Owens G, Vernon S, Madsen C. Molecular regulation of smooth muscle cell differentiation. J Hypertens 1996; 14: S55–S64. 35 Adachi Y, Kas E, Laemmli UK. Preferential, co-operative binding of topoisomerase II to scaffold associated regions. EMBO J 1989; 8: 3997–4006. 36 Izaurralde E, Kas E, Laemmli UK. Highly preferential nucleation of histone H1 assembly on scaffold associated regions. J Mol Biol 1989; 210: 573–585. 37 Owens GK. Regulation of differentiation of vascular smooth muscle cells. Physiol Rev 1995; 75: 487–517. 38 Courey AJ, Tijan R. Analysis of Sp1 in vivo reveals multiple transcriptional domains involving a novel glutamine rich activation motif. Cell 1988; 55: 887–898. 39 Aso T, Conaway J, Conaway R. Role of core promoter structure in assembly of the RNA polymerase II preinitiation complex. A common pathway for formation of preinitiation intermediates at many TATA and TATA-less promoters. J Biol Chem 1994; 269: 26575–26583. 40 Buratowski S. Multiple TATA-binding factors come back into style. Cell 1997; 91: 13–15. 41 Suzuki M, Singh RN, Crystal RG. Regulatable promoters for use in gene therapy applications: modification of the 5⬘-flanking region of the CFTR gene with multiple cAMP response elements to support basal, low-level gene expression that can be upregulated by exogenous agents that raise intracellular levels of cAMP. Hum Gene Ther 1996; 7: 1883–1893. 42 Franz WM et al. Analysis of tissue-specific gene delivery by recombinant adenoviruses containing cardiac-specific promoters. Cardiovasc Res 1997; 35: 560–566. 43 Schmitt J et al. Dominant-negative proteins for the regulation of vascular smooth muscle cell proliferation. Gene Therapy (submitted). 44 Andrawis NS, Ruley EH, Abernethy DR. Angiotensin II regulates human vascular smooth muscle ␣-actin gene expression. Biochem Biophys Res Commun 1993; 196: 962–968. 45 Andrawis NS, Wang E, Abernethy DR. Endothelin-1 induces an increase in total protein synthesis and expression of the smooth muscle alpha-actin gene in vascular smooth muscle cells. Life Sci 1996; 59: 523–528. 46 Hautmann MB, Madsen CS, Owens GK. A transforming growth factor ␤ (TGF␤) control element drives TGF␤-induced stimulation of smooth muscle ␣-actin gene expression in concert with two CArG elements. J Biol Chem 1997; 272: 10948–10956. 47 Keogh M-C et al. High efficiency reporter gene transfection of vascular tissue in vitro and in vivo using a cationic lipid–DNA complex. Gene Therapy 1997; 4: 162–171. 48 Iwaki K et al. Alpha- and beta-adrenergic stimulation induces distinct patterns of immediate early gene expression in neonatal rat myocardial cells. fos/jun expression is associated with sarcomere assembly; Egr-1 induction is primarily an alpha 1mediated response. J Biol Chem 1990; 265: 13809–13817. 49 Bhavsar PK et al. Isolation and characterisation of the human cardiac Troponin I gene (TNN13). Genomics 1996; 35: 11–23. 50 Takahashi K et al. Spontaneous transformation and immortalization of human endothelial cells. In Vitro Cell Dev Biol 1990; 25: 265–274. 51 Sambrook J, Fritsch EF, Maniatis T. Molecular Cloning. A Laboratory Manual, 2 edn. Cold Harbor Laboratory Press: Cold Harbor, NY, 1989.