|Home | About | Journals | Submit | Contact Us | Français|
Increasing the level and duration of transgene expression and restricting expression to vascular cells are important goals for clinically useful gene therapy vectors. We evaluated several promoters, enhancers and introns in endothelial, smooth muscle and liver cells in tissue culture and in vivo, comparing local delivery to the carotid artery with intravenous delivery to the liver. A 1800-bp fragment of the oxidized LDL receptor (LOX-1) promoter showed highest in vivo activity in the carotid artery, achieving 39% the activity of the reference cytomegalovirus promoter, with 188-fold greater specificity for carotid artery over liver. An enhancer from the Tie2 gene in combination with the intracellular adhesion molecule-2 promoter improved endothelial specificity of plasmid vectors, increased the expression from adenoviral vectors in cultured endothelial cells and doubled the specificity for carotid artery over liver in vivo. Adding a short intron to expression cassettes increased expression in both endothelial and smooth muscle cells in vitro; however, the eNOS enhancer failed to consistently increase the expression or endothelial specificity of the vector. In conclusion, elements from the LOX-1 promoter and Tie2 enhancer together with an intron can be used to improve vectors for vascular gene transfer.
Atherosclerosis is one of the greatest causes of mortality and morbidity in the developed world. In the early stages, disease progression is promoted by endothelial dysfunction,1 which is therefore a rational target for conventional and gene therapy. In the latter stages, tissue ischaemia caused by occlusive atherosclerotic plaques requires revascularization strategies to re-establish blood flow. For example, 45 000 angioplasties and 28 000 coronary bypass operations are currently performed each year in the United Kingdom alone. Vessel remodelling as a damage/compensatory response following angioplasty or bypass operation leads to smooth muscle cell migration and hyperplasia, that can re-occlude the vessel or act as a seed bed for accelerated atherosclerosis. There is the opportunity to employ genetic therapies for both in-stent restenosis and bypass grafting, to limit the remodelling processes, preventing vessel re-occlusion. Many preclinical studies have identified genes that upon intralumenal delivery can limit the remodelling process, maintaining vessel lumen and patency.2-4 Progress to the clinic has been impeded; however, because of the short duration of expression obtained from first generation adenoviral vectors expressing cDNAs of genes under the control of the cytomegalovirus (CMV) immediate early promoter. The majority of the remodelling process occurs within the first 1–6 months.5,6 For clinical efficacy, expression needs to be maintained for at least the first 1–2 months, much longer than the approximately 2–3 weeks of expression normally observed with CMV-based, first-generation adenoviral vectors. In contrast, expression exceeding 6 months has been documented in the liver, using a first-generation adenovirus with an intron-containing ApoAI promoter/ApoE enhancer expression cassette.7 Hence, there is considerable interest in generating a vascular specific promoter with comparable efficiency to the CMV promoter. The creation of an optimized vascular cassette would have the added advantage of enhancing the safety profile of a vector by reducing transgene expression in nontarget tissues.
Several studies have isolated promoters from vascular-specific genes including fms-like tyrosine kinase-1 (FLT-1),8 intercellular adhesion molecule-2 (ICAM-2),9 angiopoietin-2,10 endothelial nitric oxide synthase (eNOS),11 vascular cell adhesion molecule-1,12 von Willebrand factor,13 tyrosine kinase with immunoglobulin and epidermal growth factor homology domains (Tie),14 kinase-like domain receptor15 and combination of the Tie2 promoter and enhancer (Tshort).16 Other promoters that are active within vascular tissue include the oxidized LDL receptor Lox-117,18 and ICAM-119,20 promoters, which exhibit upregulation upon cytokine stimulation, a possible advantage depending on the application. In addition, enhancers from the eNOS21 and Tie222 genes have also been identified. In this report, several of these promoters, enhancers and introns were compared side by side with the CMV promoter in cultured cells in vitro and in vivo by local delivery into the carotid artery compared with intravenous delivery to the liver.
To evaluate the activity of different promoters, the multiple cloning site of pDrive (Qiagen) was used to construct a luciferase-based expression cassette (see Supplementary Figure S1). The CMV, ICAM-2, Flt-1, Tshort, LOX-1 (250 and 1800 bp fragments) and ICAM-1 promoters were cloned in front of intron A and luciferase. Endotoxin-free plasmid preparations were transfected into both human umbilical vein endothelial cells (HUVECs) and porcine aortic smooth muscle cells (PaSMCs) using lipofection, and representative expression levels from three independent experiments are presented in Figure 1. All of the promoters showed measurable activity in both HUVECs and PaSMCs, ranging from 8 to 34% CMV value in plasmid vectors (Tshort to FLT-1) in HUVECs and 0.9–8.2% in PaSMCs.
Using the ICAM-2 promoter, the additional effect of inserting either intron A or B (isolated from either pTARGET vector or a cut-down Δ89 truncated β-globin 2nd intron,23 respectively) was investigated (Figure 2) in both HUVEC and PaSMC. The ICAM-2 promoter was chosen as it had previously been shown to have high activity in both HUVECs and smooth muscle cells.24 The addition of either intron significantly increased the level of luciferase expression from the ICAM-2 promoter in either cell type.
The ability of the eNOS or Tie2 enhancers to augment the expression level was investigated. These were inserted upstream of the ICAM-2 promoter (with intron A). While neither enhancer increased the level of expression in HUVECs, the Tie2 enhancer added a level of endothelial specificity by significantly reducing luciferase expression in smooth muscle cells compared to the enhancerless control. The eNOS enhancer only significantly reduced the expression in PaSMCs in two of the three replicated experiments. The Tie2 enhancer is naturally found in the first intron of the Tie2 gene, so the ability of this enhancer to augment expression in an intronic location was investigated. A PCR-generated Tie2 enhancer fragment was cloned into the unique XmaI site within intron B. From 90 clones tested, it was only possible to isolate a double enhancer insert in the positive orientation (iTie2++), in addition to a single insert in the negative orientation (iTie2−). The iTie2++ enhancer failed to augment expression in HUVECs but again inhibited expression in PaSMCs. The iTie2− enhancer inhibited expression in HUVECs and PaSMCs compared to the intron B control, indicating a possible orientation-specific effect. From these data, the addition of an intron appears to increase expression levels, while the Tie2 enhancer somewhat improved endothelial specificity compared with expression in vascular smooth muscle cells.
Adenoviral vectors were created from the previously generated expression cassettes to allow investigation of the expression cassettes in vitro and in vivo. Figure 3a details the expression cassettes created in addition to the reference CMV promoter construct. As shown, all the non-CMV promoter constructs included an intron. LOX-1 and ICAM-1 promoters were tested without the eNOS enhancer. ICAM-2, Tshort and minimum FLT-1 promoters were tested with the eNOS enhancer. The effect of the intronic Tie2 enhancer (iTie2++) was also tested in the context of the ICAM-2 promoter. The region of the FLT-1 promoter included in the expression cassette was reduced, because the previously published portion of the promoter (−748 to +284) includes a region −748 to −230, which has very few predicted transcription factor binding sites (as identified by MatInspector transcription factor binding matrix—http://www.genomatix.de/products/MatInspector; see Supplementary data), and also includes the first intron splice donor site (+230), which would cause aberrant splicing within the transgene if it contained cryptic splice sites. Because of this, a reduced fragment was included in the viral construct (min FLT-1: −230 to +33), which included the majority of the predicted transcription factor binding sites while avoiding the first intron splice donor site. Unfortunately, it was not possible to rescue an adenoviral construct containing the full-length FLT-1 promoter, as it induced instability within the adenoviral construct, producing constructs with various deletions within the promoter. Hence, it was not possible to directly compare the minimal FLT-1 promoter with the full-length version.
The different adenoviral vectors containing the expression cassettes were used to transduce HUVECs, PaSMCs and the hepatocarcinoma cell line HepG2 cells (Figures 3b–d). All the cassettes promoted transgene expression in all the cell types. However, the ICAM-2 iTie2++ vector generated the highest level of luciferase expression in endothelial cells, while the ICAM-1 promoter was most efficient in smooth muscle and HepG2 cells. The iTie2++ enhancer increased expression in HUVECs but not in either PaSMCs or HepG2 cells, indicating that the iTie2++ continued to promote endothelial specificity in the context of an adenoviral vector. The relative expression of the adenoviral vectors differed from that of the plasmid vectors, the luciferase expression in HUVECs being lower in comparison to CMV (for example, adenoviral expression 2.8% versus plasmid expression 16.5% of the CMV level for the ICAM-2 iTie2++ construct), indicating that the background of the expression cassette affects the expression level.
The adenoviral expression cassettes were used to transduce both vascular tissue and the liver to investigate their relative expression in both tissues. Adenoviral vectors were instilled into the right common carotid arteries of CD1 mice with a dwell time of 10 min. Seven days post-transduction, the arteries were removed and the expression of luciferase measured. Figure 4a illustrates the luciferase expression from each vector. The LOX-11800 promoter gave the highest expression level equivalent to 39% of the activity with CMV. The addition of the iTie2++ enhancer increased the expression in the carotid (ICAM-2 versus ICAM-2 iTie2++, 12 versus 36% of the CMV activity, P = 0.045), recapitulating the effect observed in vitro. The expression in the liver was assessed by tail vein administration of 1 × 109 PFU of each vector premixed with 1 × 109 PFU CMV-βgal (Figure 4b). All of the vectors showed minimal activity in the liver, averaging 0.37% compared to the reference CMV promoter. Using the CMV promoter to normalize expression levels in the carotid and the liver, the specificity index of the different adenoviral expression cassettes was deduced (Figure 4c). All the tested vectors had improved specificity for the carotid artery by factors that varied from 11-fold for the Tshort to 188-fold for the LOX-11800 promoter. The ICAM-2 promoter construct was 58-fold more active in the carotid than the liver, while the ICAM-2 iTie2++ expression cassette was 106-fold more active in the carotid, although there was insufficient power to show a significant difference between the ICAM-2 promoter and the ICAM-2 promoter with the iTie2++ enhancer.
This study describes the production of several novel expression cassettes that appear to be valuable for vascular gene transfer. Of the promoters tested in this study, the Tshort (7%), ICAM-2 (12%), ICAM-1 (17%) and min FLT-1 (20%) promoters all exhibited reasonable activity in the carotid compared to the CMV promoter. However, the long LOX-1 promoter showed the highest level of luciferase expression in vivo (39%) compared to the CMV promoter. In situations of inflammation, it is possible that the relative transcription from this cytokine inducible promoter would be higher, as previously demonstrated.18,25
Both of the introns tested here increased transgene expression, by approximately fourfold for intron A and threefold for intron B in both HUVEC and PaSMC in vitro. These short synthetic introns both show utility for increasing transgene expression and highlight the importance of splicing for efficient RNA polymerase II-based gene expression.
The eNOS enhancer failed to increase the expression level of the plasmid constructs that contained it, or consistently reduce the expression in PaSMCs. The Tie2 enhancer had differing affects depending on the context of the cassette. In plasmid vectors, the Tie2 enhancer did not increase the level of luciferase expression, but acted to suppress luciferase expression in PaSMCs. This effect was observed regardless of the position of the enhancer (upstream of the promoter or located within an intron); however, a single copy in the reverse orientation inhibited transgene expression in both HUVECs and PaSMCs when placed within intron B. In the context of an adenoviral vector, the intronic Tie2 enhancer specifically enhanced expression in HUVECs, demonstrating that the difference of vector background (plasmid versus adenovirus) affects the action of this enhancer, possible due to the contribution of the adjacent adenoviral sequences on the expression cassette.26
The maximum specificity for vascular tissue compared to the liver was achieved using the LOX-11800 promoter having only 0.21% of the level of the CMV activity in the liver, giving a specificity ratio of 188. The ICAM-2 promoter with the intronic Tie2 enhancer (iTie2++) produced only 0.34% the activity of the CMV promoter in the liver leading to a specificity ratio of 106, while this was greater than the ICAM-2 promoter alone (specificity ratio of 56), significance was not reached with the data collected. It will be interesting to test whether the iTie2 enhancer could further increase the activity of the LOX-11800 construct and increase the specificity ratio. It may be possible also to combine these cassettes with other genomic elements, such as matrix attachment regions, locus control regions or factors that enhance splicing efficiency and nuclear export of mRNA to further enhance expression levels.27
Luminal arterial administration of adenoviral vectors has previously been shown to target endothelial cells28 as apposed to adeno-associated viral vectors that principally transduce smooth muscle cells. The refractory nature of smooth muscle cells to adenoviral transduction in vivo prevents the examination of endothelial specificity in the carotid artery. Within the liver, adenoviral vectors predominantly transduce hepatocytes, as sinusoidal endothelial cells and Kupffer cells are nonpermissive for adenoviral vector transduction.29 The low-level expression of luciferase observed in the liver samples of the non-CMV-driven constructs, is therefore most likely to be a result of many hepatocytes expressing at a low level, rather than high-level expression in a few sinusoidal endothelial cells.
The expression data generated in the hepatocarcinoma cell line HepG2 ranged between 1 and 23% of the expression from the CMV promoter, whereas in vivo, the promoters appeared much less active in the liver. This illustrates the limitation of using HepG2 as a predictor for in vivo promoter activity in the liver. In addition, adenoviral expression levels in HUVECs and PaSMCs were much lower than those observed in the carotid, suggesting that these data underplay the in vivo activity.
Use of nonviral promoters has been demonstrated to increase the length of expression in a range of different tissues (reviewed by Papadakis27), so there is a good precedent for expecting the same to be found in vascular tissue. Experiments to determine enhancement of long-evity of expression will probably require a larger animal model, as the quantity of tissue available in the mouse model could render the expression levels unmeasurable if they decreased to less than 10% of the ICAM-2 promoter level, hence a rat or rabbit model would be preferential. However, this study has identified the LOX-11800 and ICAM-2 iTie2++ cassettes and potentially worthy of additional study.
Promoters, enhancers and intron-containing fragments were generated as described in Supplementary Table S1 and cloned into the multiple cloning site of pDrive (Qiagen, Crawley, UK). Enhancers were cloned into the MluI/SnaBI sites, promoters were inserted into the TA cloning site, or the flanking EcoRI sites and introns between the SalI/HindIII sites. A copy of luciferase was excised from pXCX-CMV-luciferase (see below) (HindIII, NotI-blunted) and inserted into the HindIII, XbaI-blunted sites of pDrive. A blunted SmaI fragment from pHM5/hFIX30 containing the bGH polyA was inserted into the blunted SacI site of pDrive (see Supplementary Figure S1). A copy of the Tie2 enhancer (PCR-generated with XmaI ends) was inserted into the central XmaI site of intron B (to recapitulate the natural position of the Tie2 enhancer). Despite screening 90 clones, it was not possible to isolate a single copy of the enhancer, inserted in the positive orientation. A double copy (positive orientation) and single copy (negative orientation) were isolated and investigated. The Tie2 enhancer located within intron B is denoted by ‘iTie2’ ++ or −, depending on the orientation of the enhancer.
pXCX-CMV-luciferase (SW1) was created by insertion of a proofreading polymerase generated PCR fragment of luciferase (Supplementary Table S1) into the HindIII/NotI sites. The other viral expression cassettes were generated by removing the CMV from the plasmid using restriction enzymes MluI/HindIII and inserting an MluI/HindIII fragment from the expression cassettes generated above. Adenoviral vectors were then created using standard techniques31 by co-transfection with adenoviral backbone pJM17 into HEK 293 cells.
Pooled donor, HUVECs were purchased from TCS cell works (Parkleys, UK) and grown in large vessel endothelial growth media (TCS) supplemented with 20% fetal calf serum. For transfection, HUVECs were seeded into fibronectin-coated 48-well plates at 1.3 × 104 cells per well. Just prior to transfection, HUVECs were changed to media containing 2% serum, and enough transfection mix to transfect four wells per sample (1 μg test plasmid, 0.5 μg pSV-βgal and 12 μl of JetPEI-HUVECs) was prepared according to the manufacturer’s instructions (PolyPlus Transfection, Autogen Bioclear, Calne, UK). This was added to the cells and incubated for 4 h, after which the transfection solution was removed and replaced with fresh complete media. Expression analysis was performed after 48 h. HUVECs (seeded at 4 × 104 cells per well of a 24-well plate) were transduced by adenoviral vectors at 200 PFU per cell (test virus) with 100 PFU per cell CMV-βgal overnight. The virus was then removed and the cells incubated for a further 24 h before analysing gene expression. Transfections and adenoviral transductions were performed three times with different batches of pooled donor HUVECs, and representative results are presented. PaSMCs were isolated by explant culture and cultured in Dulbecco’s modified Eagle’s medium (4.5 g l−1 glucose) and 10% fetal calf serum. For transfection, cells were split into 24-well plates (4 × 104 cells per well). The following day, sufficient transfection reagent was prepared for three wells: 9.4 μl of Mirus LT1 (Mirus) was added to 150 μl of Optimem, after 10 min, the DNA was added (1 μg test plasmid, 0.5 μg pSV-βgal) and incubated for 20 min before adding to the cells. Media were changed after an overnight incubation, and gene expression analysis was performed after a further 24 h. PaSMCs were transduced by adenoviral vectors as described for HUVECs. Transfections and adenoviral transductions were performed three times with different batches of PaASMCs, and representative results are presented. HepG2s, grown in minimal essential media supplemented with 10% fetal calf serum, were transduced by adenoviral vectors as described for HUVECs, except that 50 PFU per cell (test virus) with 25 PFU per cell CMV-βgal was used.
All experiments were performed according to the Home Office guidelines and approved by the Local Ethics Committee for Animal Experimentation. Adenoviral transduction of carotid arteries was performed by luminal incubation of each vector for 10 min as described.32 Viruses were diluted to 1 × 1010 PFU ml−1 using the dialysis buffer used to prepare the adenoviral vector stocks (10 mm Tris pH 7.5, 135 mm NaCl, 1 mm MgCl2, 10% glycerol) to ensure that all transduction were performed under the same conditions. Carotids were removed 7 days after transduction and analysed for gene expression. This technique predominantly transduces endothelial cells28 (Supplementary Figure S2). Liver transduction was performed by tail vein delivery of 1 × 109 PFU of test virus premixed with 1 × 109 PFU of Ad CMV-βgal.33 Livers were removed 5 days after transduction and analysed for gene expression. The LOX-1250 construct was not tested in vivo as it was not possible to produce a stock of sufficiently high-titre vector.
Gene expression was assessed using the Dual-Light combined reporter gene assay system for detection of luciferase and β-galactosidase (Tropix, Applied Biosystems, Warrington, UK) according to the manufacturer’s instructions. Results are presented as luciferase expression normalized to the β-galactosidase expression, to control for both transfection and lysis efficiency. For assaying luciferase expression in the carotid arteries, the carotids were individually ground in cell culture lysis reagent buffer and assayed using the luciferase assay reagent (Promega, Southampton, UK). Data are presented as raw relative light units per carotid and not normalized to protein, because when removing the carotids, it is not possible to determine the precise limit of transduction, hence, a larger length of carotid is removed to ensure all the transduced tissue is removed. In addition, inclusion of any surrounding connective tissue would increase the protein concentration in the lysates and so increase variability. Because these two compounding factors have a potentially large effect on the analysis of the very small piece of tissue that is removed, data are presented as relative light units per carotid.
Initial examination of the data suggested a skewed distribution with the variability increasing with the mean. To stabilize the variance and induce normality, the data were transformed to the logarithmic scale for analysis. The results were back-transformed to the original measurement scale and are reported as geometric means with 95% confidence intervals. Estimates were obtained using multiple linear regression and viral vectors were compared using a χ2-test. To obtain robust estimation of the s.e., 10 000 bootstrap samples were generated and used in the analysis, with confidence intervals being corrected for bias. Bootstrap samples where one or more parameters could not be estimated were excluded from the calculations. As triplicate samples from each mouse liver was analysed; the three measurements taken from the same mouse were replaced by the geometric mean prior to group analysis. All analyses were carried out using Stata version 9.2.
We thank Ray Bush (University of Bristol) for help with animal work; Jill Tarlton and Melanie Roberts (University of Bristol) for help with virus production and technical assistance; James Uney (University of Bristol) for plasmid pXCX-CMV;34 Mike Antoniou (Kings College, London) for the provision of Cβ88 plasmid, containing the Δ89 truncated β-globin 2nd intron23 (intron B); Alessandra Tessitore (University of L’Aquila) for plasmid LC120 containing full ICAM-1 promoter; Anja Ehrhardt (Stanford University) for plasmid pHM5/hFIX.30 This work was performed at the Bristol Heart Institute, the University of Bristol. SJW was supported by British Heart Foundation intermediate fellowship (FS/03/096/16318).
Conflict of interest
The authors have no conflicts of interest to declare.
Supplementary information accompanies the paper on Gene Therapy website (http://www.nature.com/gt)