gt Efﬁcient gene transfer and long-term expression in

Gene Therapy (1999) 6, 1884–1892 1999 Stockton Press All rights reserved 0969-7128/99 $15.00

http://www.stockton-press.co.uk/gt

Efficient gene transfer and long-term expression inneurons using a recombinant adenovirus with aneuron-specific promoter

V Navarro1, S Millecamps1, M-C Geoffroy1, J-J Robert1, A Valin2, J Mallet1 and G Le Gal La Salle2,3

1Laboratoire de Genetique Moleculaire de la Neurotransmission et des Processus Neurodegeneratifs, Batiment CERVI, Hopital de laPitie-Salpetriere, Paris; and 2Institut Alfred Fessard, CNRS, Gif sur Yvette, France

Adenoviruses are highly efficient vectors for gene transfer cerebellum and striatum. Anatomical and immunohistoch-into brain cells. Restricting transgene expression to specific emical analyses of the Ad-NSE-stained cells showed thatcell types and maintaining long-term expression are major neurons were preferentially transduced. More neuronsgoals for gene therapy in the central nervous system. We were stained in the hippocampus following infection withtargeted gene expression to neurons by constructing an Ad-NSE than with Ad-RSV. Cytotoxicity from Ad-NSE wasadenoviral vector that expressed the E. coli LacZ reporter lower than from Ad-RSV. β-Galactosidase genegene under the control of the rat neuron-specific enolase expression after Ad-NSE infection remained stable for 3.

promoter (Ad-NSE). Expression from Ad-NSE was com- months, and was detectable for 6 months. Thus, the NSE-pared with that from an adenoviral vector encoding the adenoviral vector can be used to transfer potentially thera-same reporter gene under the control of the Rous sarcoma peutic genes into neuronal cells. The use of a cell-specificvirus LTR promoter (Ad-RSV). Both recombinant adeno- promoter also resulted in high in vivo efficiency and long-viruses were injected stereotactically into rat hippocampus, term transgene expression.

Keywords: gene therapy; adenoviral vector; neuron-specific enolase promoter; hippocampus; neurons

IntroductionEfficient gene transfer into brain cells is a major goal forneurobiological studies and for the development of genetherapy for various human neurological diseases whichcurrently have no effective treatment. Various gene deliv-ery systems are available.1 Replication-defective adenovi-ral vectors offer many advantages for gene transfer intothe central nervous system, such as the ability to infect awide variety of brain cells including post-mitotic neu-rons, high transduction efficiency and low pathogen-icity.2,3 Moreover, adenoviral vectors have a cloningcapacity of several kilobases and the adenoviral DNAdoes not integrate into the host genome, avoiding themutagenesis risk associated with retroviral vectors.1

Most gene delivery vectors use viral promoters, parti-cularly the Rous sarcoma virus (RSV) or cytomegalovirus(CMV) promoters, because they have high transcriptionlevels. They drive transgene expression in all infectedcells whatever their type. The development of gene trans-fer strategies for the central nervous system requires theselective targeting of transgene expression to particularcells. The expression of genes encoding neurotrophic fac-tors or substitutive enzymes should be restricted to spe-

Correspondence: J Mallet, Laboratoire de Genetique de la Neurotransmis-sion et des Processus Neurodegeneratifs, Batiment CERVI, UMR CNRS9923, Hopital de la Pitie-Salpetriere, Paris, France3Present address: Laboratoire d’Epileptologie Experimentale et Clinique,Universite Victor Segalem, Bordeaux, FranceReceived 21 December 1998; accepted 8 June 1999

cific cell types, to limit the side-effects caused by theexpression of such genes in healthy cells. Cell-specificpromoters, which drive the expression of genes encodingtissue- or cell-specific proteins, could be used to targettransgene transcription. Moreover, cell-specific pro-moters may give a level of expression better adapted tocellular metabolism, facilitating the longer-termexpression of the transgene. Indeed, the short-termefficiency of viral promoters suggests the down-regu-lation of transcription, probably by cellular defensemechanisms.1 Therefore, transgene expression could alsobe prolonged by improving transcriptional regulation.

Purkinje cell- and oligodendrocyte-specific promotershave been used in adenoviral vectors for specific cellulartargeting.4 In our study, we targeted neurons becausethey play a crucial role in the central nervous system andare the principal cells affected in many neurologicaldiseases.

Neuron-specific enolase (NSE), an isoform of the glyco-lytic enzyme enolase, is one of the most abundant pro-teins in the adult mammalian brain, and is found inmature neurons and neuroendocrine cells.5 Transientexpression experiments have identified a region in therat NSE promoter responsible for high-level expressionin neuronal cells.6 A 1.8 kb NSE promoter fragment hasbeen shown to target the expression of various trans-genes to differentiated neurons in transgenic animals.7,8

We constructed a first generation adenoviral vectorcontaining the rat NSE promoter upstream from the LacZreporter gene (Ad-NSE) and compared, in vitro and invivo, the patterns of β-galactosidase gene expression from

Efficient and long-term gene expression in neuronsV Navarro et al

1885

Figure 1 Plasmid constructs. pAd-NSE contains the rat NSE promoterdriving the E. coli LacZ gene, followed by the SV40 virus polyadenylationsignal (PA). This unit was inserted between two adenoviral sequences,containing the inverted terminal repeat (ITR) with the encapsidation sig-nal (χ) and a sequence encoding the adenoviral peptide IX (P IX). pAd-NSE(R) contains the same unit inserted in reverse orientation. pAd-RSVdiffers from pAd-NSE in that it contains the RSV LTR promoter.

Ad-NSE and from an adenoviral vector containing theRSV promoter (Ad-RSV). Ad-NSE gave preferentialexpression of the transgene in brain neurons, with higherefficiency and lower toxicity than Ad-RSV. Ad-NSE alsoshowed capacity of long-term expression.

Results

Neuron-specific expression of Ad-NSEThe 1.8 kb NSE promoter, which has been shown to drivetransgene expression in neurons in transgenic miceefficiently, was inserted into a transcriptional unit to con-trol expression of the LacZ gene. Before generating therecombinant adenovirus, we first assessed the neuron-specific expression of the NSE-LacZ gene unit in theshuttle plasmids. The shuttle plasmids used to generateadenoviruses contain adenoviral sequences implicated inthe partial dysregulation of some exogenous promoters.9We checked that these sequences did not affect theactivity of the NSE promoter by comparing the β-galacto-sidase gene expression of pNSE with that of the shuttleplasmids, pAd-NSE and pAd-NSE(R), which contains theNSE-LacZ gene cassette in reverse orientation (Figure 1).We found that adenoviral sequences did not affect thecell specificity or transcription rates of the adjacent NSEpromoter (Figure 2a and b). The three plasmids weresimilarly expressed in neuronal undifferentiated PC12

Figure 2 Transfection of non-neuronal 293 cells (a) and of neuronalundifferentiated PC12 cells (b) with pNSE and the shuttle plasmids pAd-NSE, pAd-NSE(R) and pAd-RSV. β-Galactosidase activity was expressedin relative activity after removal of the basal endogenous activity and afternormalization based on the luciferase activity of another plasmid simul-taneously transfected. Infection of a primary culture of striatal neurons(c) at various multiplicities of infection (MOI) (p.f.u. per cell). Infectionof non-neuronal 293 cells (d) and HeLa cells (e), and of neuronal undiffer-entiated PC12 cells (f), NGF-differentiated PC12 cells (g), and CHP 212cells (h) by Ad-NSE and Ad-RSV at a MOI of 100. Infection of a primaryculture of sympathetic neurons at a MOI of 100 (i). For the infections,normalization of the β-galactosidase activity was based on the amountof protein. Values are means ± standard errors of the mean (s.e.m.) forthree determinations.


1886 cells and restricted in non-neuronal 293 cells. The orien-tation of the transcriptional unit in the shuttle plasmidhad no effect on β-galactosidase gene expression, basedon comparisons between pAd-NSE and pAd-NSE(R).Therefore, the recombinant adenovirus Ad-NSE was arbi-trarily constructed with the NSE-LacZ cassette in directorientation.

β-Galactosidase expression from Ad-NSE was com-pared in vitro with that from Ad-RSV following the infec-tion of neuronal and non-neuronal cells. Ad-NSE wasshown to give neuron-specific expression. β-Galacto-sidase was not expressed in non-neuronal HeLa cells, asshown by the 1:1000 ratio between the levels ofexpression from Ad-NSE and Ad-RSV (Figure 2e). In thenon-neuronal 293 cells, this ratio was 1:190 (Figure 2d).In the NGF-differentiated neuronal PC12 cells (Figure2g), the undifferentiated PC12 cells (Figure 2f) and theneuronal CHP 212 cells (Figure 2h), the ratios were 1:2,1:4 and 1:6, respectively. In rat primary cultures of sym-pathetic neurons (Figure 2i) and striatal neurons (Figure2c), the ratios were 1:10 and 1:29, respectively. Lineardose-response curves were obtained for both adeno-viruses (Figure 2c).

The adenoviruses were injected into the hippocampus,cerebellum and striatum of adult rats. In the hippocam-pal dentate gyrus, the patterns of expression of Ad-NSEand Ad-RSV were very different. The Ad-RSV-expressingcells had a diffuse anatomical distribution spread outfrom the neural cell layers (Figure 3b), corresponding tothe staining of neurons and astrocytes. In the granularcell layer, there were many stained nuclei on the innerside, where the astrocytes had their soma. Immunohisto-chemical analyses confirmed that these Ad-RSV-stainedcells colocalized with the glial fibrillary acidic protein(GFAP) marker (Figure 3h), and with the neuron-specificnuclear protein (NeuN) marker (Figure 3e). Quantitativeanalysis showed that β-galactosidase was similarlyexpressed in both cell types after Ad-RSV infection(Figure 4). In contrast, the Ad-NSE-expressing cells in thedentate gyrus were mostly in the granular neuronal celllayer (Figure 3a). Some interneurons and cells in the CA3and CA4 pyramidal cell layers were also stained. Immu-nohistochemical colocalization of the Ad-NSE-stainedcells with the NeuN marker showed that these cells wereneuronal (Figure 3d). Only a few X-gal-stained cells inthe granular cell layer and in the hippocampal fissurewere also stained for the GFAP marker (Figure 3g).Nevertheless, the intensity of X-gal nuclear staining inastrocytes was more pronounced following Ad-RSVinfection and sometimes led to diffusion of the enzymeinto the cytoplasm (Figure 3k; arrow).

Quantitative analysis showed that more than 1000 cellswere transduced by Ad-NSE per 30-µm section of the 1mm rostral–caudal segment of the hippocampus ana-lyzed, and confirmed that neurons were preferentiallystained by Ad-NSE (more than 70% versus less than 30%for non-neuronal cells) (Figure 4). The patterns ofexpression of Ad-NSE and Ad-RSV observed 7 days aftertheir injection into the hippocampus (Figure 3d and e)were unchanged 14 days (data not shown) and 3. monthslater (Figure 5e and f). A two-factor analysis of variance(ANOVA) was used to determine the ‘virus’ effect andthe ‘time’ effect on the number of LacZ-expressing cellsin the injected hippocampus. The analysis revealed a stat-istically significant difference in the number of LacZ-

expressing neurons between the Ad-NSE and Ad-RSVgroups whatever the time (F(1,2) = 26.662; P = 0.0355).The number of stained neurons was two to four timesgreater for Ad-NSE than for Ad-RSV, at each time-pointanalyzed (Figure 4). There was no such difference for thenon-neuronal cells.

In the cerebellum, the Ad-RSV-stained cells had a dif-fuse distribution, whereas the Ad-NSE-stained cells wererestricted to the neuronal layers, 7 days and 3. monthsafter the injection (Figure 5a and b), as observed in thehippocampus. Immunostaining experiments with theNeuN marker confirmed that the LacZ-expressing cellsfollowing Ad-NSE injection were essentially neurons(Figure 6a), whereas the proportion of LacZ-expressingneurons was lower following Ad-RSV injection (Figure6b). In the striatum, the area around the dorso-ventralaxis of the needle track was similarly stained by the twoadenoviruses (Figure 5c and d). Immunostaining experi-ments demonstrated that most of the LacZ-expressingcells following Ad-NSE injection were double-labelled bythe NeuN marker (Figure 6c), whereas the proportion ofdouble-labelling following Ad-RSV injection was clearlylower (Figure 6d). A few cells were labelled by bothadenoviruses in the ipsilateral pars compacta substantianigra, due to axonal retrograde transport (data notshown).

Low toxicity of Ad-NSEThe inflammatory response of the brain to adenoviralinfection was assessed by neutral red counter-stainingand GFAP immunostaining. Ad-NSE infection with lessthan 4 × 107 plaque-forming units (p.f.u.) caused a minorinflammatory reaction similar to that following PBS injec-tion. In the hippocampus, the glial reaction to 1.4 × 107

p.f.u. of Ad-RSV was stronger than that for Ad-NSE (datanot shown). Reactive astrocytes, detected by vimentineimmunostaining were more numerous after Ad-RSVinfection than after Ad-NSE infection (Figure 3j and k).Neuronal loss in the hippocampus was used to assess thecytotoxicity of the adenoviral infections. The granular celllayer narrowed more strongly at the injection site afterAd-RSV infection (Figure 3b; arrow) than after Ad-NSEinfection (Figure 3a), due to a loss of neurons. Similarfindings were observed at 3. months (data not shown).

Long-term expression from Ad-NSEExpressions from Ad-NSE and Ad-RSV were evaluatedby quantitative analysis of the stained cells after intrahip-pocampal injection at 7 days, 14 days and 3. months(Figure 4). No statistically significant decrease of thenumber of stained neurons was observed at these threetime-points, whatever the virus, using a two-factorANOVA (F(2,2) = 7.796; P = 0.1137). Similarly, the differ-ences between the number of non-neuronal stained cellsover time were not statistically significant (F(3,2) = 6.943;P = 0.1285).

The expression kinetics of Ad-NSE were evaluated inthe hippocampus at various time-points for up to 6months (Figure 7). The total number of stained cellsslowly decreased up to 3. months, and a few cells werestill labelled 6 months after the injection (last timeanalyzed) (Figure 5g and h). The intensity of staining ofthese cells was not changed.


1887

a b c

d e f

g h i

j k l

Figure 3 Histochemical staining for β-galactosidase in the rat hippocampus 1 week after infection with 1.4 × 107 p.f.u. of Ad-NSE on the left side (a,d, g, j) and 1.4 × 107 p.f.u. of Ad-RSV on the right side (b, e, h, k). Neurons were immunostained by the NeuN marker (a–f) and astrocytes by theGFAP marker (g–i). Most Ad-NSE-infected cells, whose nuclei are blue stained, were localized in the dentate gyrus granular cell layer, gl (a) and werecolabelled by the nuclear neuron-specific NeuN marker (detected by immunoperoxidase) giving a dark nuclear staining (d). A few stained cells in thegranular cell layer were also colabelled by the cytosolic glial GFAP marker (detected by nickel reinforced immunoperoxidase) (g; arrows). The Ad-RSV-stained cells are neurons (e) and astrocytes, the soma of which lie on the inner side of the granular layer (h; arrows). Diffusion of the X-gal stainingto the cytoplasm sometimes allows morphological identification of astrocytes (k; arrow). A narrowing of the granular cell layer was detected at theinjection site of Ad-RSV (b; arrow). Glial reaction evaluated by vimentine immunostaining is weaker after Ad-NSE infection (j) than after Ad-RSVinfection (k). Control images of uninfected and immunostained hippocampal sections for which a X-gal reaction was performed (c, f, i, l). Scale bar: 100µm (a–c, j–l), 50 µm (d–i). hf, hippocampal fissure; h, hilus.

Figure 4 Number of neuronal (black) and non-neuronal (grey) X-gal-stained cells, 7 days, 2 weeks and 3. months after injection with 1.4 ×107 p.f.u. of Ad-NSE and Ad-RSV into the rat hippocampus. Values aremeans ± s.e.m. of the number of cells per 30 µm-thick section (n = 6sections).

Discussion

Improvements of current gene transfer strategies arebased, among others, on the transcriptional regulation ofthe transgene. A recombinant adenovirus able to directtransgene expression to neurons may be a particularlyuseful vector for gene therapy in the central nervous sys-tem. In this study, we constructed an adenoviral vectorcarrying the cell-specific NSE promoter (Ad-NSE), anddemonstrated the neuron-specific expression of the trans-gene in this vector, whereas the transgene in a vector car-rying the viral RSV promoter was expressed in a ubiqui-tous manner. Ad-NSE was also low in toxicity and gavelong-term expression.

The fact that the NSE promoter fragment has beenfound to give specific expression in neurons in transgenicmice, does not guarantee a similar expression pattern inan adenoviral vector. The E1A gene enhancer, includedwithin the packaging sequence, has been implicated inthe partial dysregulation of the rat albumin and mouseβ-major globin promoters in an adenoviral vector.9 Our


1888

a b

c d

e f

g h

Figure 5 Histochemical staining for β-galactosidase in the rat cerebellum (a, b), striatum (c, d) and hippocampus (e, f) 3. months after infection with1.4 × 107 p.f.u. of Ad-NSE (a, c, e) and Ad-RSV (b, d, f) on opposite sides of the same animal. The Ad-NSE-stained cells were mostly in the Purkinjecell layer, pl, and in the granular cell layer, gl, counterstained with neutral red (a) whereas the Ad-RSV-stained cells were mostly in the molecular celllayer, ml (b). Persistence of β-galactosidase expression in the rat hippocampus 6 months after Ad-NSE injection (g, h). Scale bar: 300 µm (a–d), 100µm (e–h). CA3 and CA4, pyramidal cell layers.

transfection experiments demonstrated that the E1A geneenhancer did not disrupt the activity of the adjacent NSEpromoter. In vitro studies with Ad-NSE showed that theNSE promoter in the adenoviral backbone did indeedgive neuron-specific expression. The high level ofexpression obtained with Ad-NSE in NGF-differentiatedPC12 cells demonstrates that a cellular promoter may beas efficient as a viral promoter in specific cells, and thatthe level of expression from Ad-NSE in neuronal cellsmay be related to the maturity of these cells. These resultsare in accordance with the expression of the NSE proteinin the PC12 neuroendocrine cells5 and with a higher con-tent of endogenous NSE mRNA in NGF-differentiated

PC12 as compared with this content in undifferentiatedPC12 cells.6

The neuronal cell-specificity of the Ad-NSE vector wasconfirmed by the staining observed in the rat hippocam-pus and cerebellum, in which neuronal cell layers areeasy to recognize. Immunostaining experiments perfor-med in the hippocampus demonstrated that more than70% of the Ad-NSE-stained cells were neurons, whereasneurons accounted for only 50% of the Ad-RSV-stainedcells. Moreover, the number of neurons expressing β-galactosidase in the hippocampal granular cell layer wasthree times higher on average after Ad-NSE infectionthan after Ad-RSV infection. The staining of more than


1889

a b

c d

Figure 6 NeuN immunostaining and histochemical staining for β-galactosidase in the rat cerebellum (a, b) and the striatum (c, d) 1 week after infectionwith 1.4 × 107 p.f.u. of Ad-NSE (a, c) and Ad-RSV (b, d). In the cerebellum, the Ad-NSE-stained cells colocalize with the neurons of the granular celllayer, gl. In the striatum, more colabelling with neurons (arrows) were observed following Ad-NSE infection than following Ad-RSV infection. Scalebar: 50 µm; ml, molecular cell layer.

Figure 7 Expression kinetics of the Ad-NSE-stained cells in the hippo-campus. Values are means ± s.e.m. for the total number of cells per 30µm-thick section (n = 6 sections). Analysis was performed at 7 days,2 weeks, 1. months, 3. months, 4. months and 6 months following theinjection.

800 neurons per section in the hippocampus followinginjection of Ad-NSE is particularly striking in view of thesmall number of virus particles used (1.4 × 107 p.f.u.).These major in vivo results conflict with those in vitro, inwhich the β-galactosidase activity after Ad-NSE infectionwas lower than that with Ad-RSV. This difference maybe due to the activity of the NSE promoter which is betteradapted to mature brain neurons than to fetal striatalneurons, newborn sympathetic neurons or undifferen-tiated PC12 and CHP 212 cells, given the lower levels ofendogenous expression of the NSE protein in these cells.

The weak non-neuronal expression observed after Ad-NSE infection may be due to the characteristics of the ratNSE promoter. Its cell-specific activity is due to neuronalenhancers driving the activity of a basal promoter,6 thebasal promoter having ‘house-keeping gene’ features,

generally associated with ubiquitous expression.10 Thisminimal activity should be detectable in vivo in somenon-neuronal cells following overinfection. In vitro, thedetection of an expression in the 293 non-neuronal cellsafter Ad-NSE infection was not observed after NSE-con-taining plasmids transfection. The high level of adenovi-ral replication allowed in these E1 gene-containing cellscould explain this difference in β-galactosidaseexpression. Preferential, rather than specific, neuronalexpression has previously been observed following ratbrain injection with naked plasmids containing the NSE-LacZ cassette11 or adeno-associated virus (AAV) vectorcontaining the NSE-green fluorescent protein cassette.12

Strict cell-specific expression of the LacZ gene fromrecombinant adenoviruses carrying Purkinje cell- or oli-godendrocyte-specific promoters has also been reported.4However, in the absence of further immunohistochemicalanalysis, the possibility of non-specific expression cannotbe excluded. Our results with Ad-NSE demonstrate forthe first time, high level of transgene expression in theneurons of the hippocampus, cerebellum and striatum.

Another important finding was the low cytotoxicity ofAd-NSE. This may be due to the physiological level oftranscription from the NSE promoter, which shouldavoid the exhaustion of the protein synthesis machineryof infected cells. The overaccumulation of β-galactosidasein the nucleus of Ad-RSV-infected cells is probably toxic.β-Galactosidase may also induce an immune responsemediated by antigen-presenting proteins (MCH I andMCH II), which are only present on glial cells.13 The tar-geting of β-galactosidase to neurons by Ad-NSE mayhave attenuated this immune response. Further analysesneed to be performed to study the differences betweenthe brain inflammation caused by both adenoviruses.


1890 Finally, we found that Ad-NSE gave long-termexpression of the transgene. Transgene expression wasstill stronger in neurons at 3. months with the NSE pro-moter than with the RSV promoter, and it remaineddetectable for 6 months. Cellular promoters may be lesssensitive to inactivation than viral promoters, as observedwith retroviral vectors.14 Episomal adenoviral DNA maybe diluted and lost during cell division in reactive astro-cytes following local brain infection, but this cannot occurin neurons. Adenoviral vectors containing viral pro-moters have been shown to express the β-galactosidasegene for 2 months after their injection into the brain,despite a decline in the number of stained cells.2,3,15 Ourresults showed that β-galactosidase expression followingAd-RSV infection persisted for up to 3. months.

The high-level expression obtained in neurons with theNSE promoter and the advantages of the adenoviral vec-tor represent major progress in the development ofimproved vectors for gene therapy in the central nervoussystem. However, the NSE promoter has performed lesswell in some other vectors. Naked plasmids containingthe NSE-LacZ cassette appeared less efficient wheninjected into rat brain11 than Ad-NSE. Recombinant her-pes simplex virus, which is naturally tropic for neurons,was found to express the NSE-LacZ cassette inefficientlyin vivo due to the vector high toxicity and transcriptionaldysregulation by adjacent viral regions.16 The NSE pro-moter was also tested in an adeno-associated virus (AAV)vector.12 However, our study with an adenoviral vectorgave higher efficiency in terms of the number of granularand pyramidal neurons stained in the hippocampus.Moreover, the adenovirus has a larger cloning capacitythan the AAV vector, rendering our system more attract-ive for future applications.1

This study demonstrates that recombinant adeno-viruses containing the NSE promoter are the vector ofchoice for highly efficient and long-term expression oftransgenes in brain neurons. The improvements obtainedwith the use of the NSE promoter, coupled with new gen-erations of adenoviral vectors in which unnecessary,17 orall18 viral genes are deleted to prevent viral protein pro-duction, have potential for use in human clinical trials.Most animal experiments investigating gene transfer withadenoviral vectors have used viral promoters to drive thetransgenes. It would be of interest to test whether theneuronal targeting of therapeutic agents, such as trophicfactors, anti-oxidative agents19 or specific enzymes20

improves their neuroprotective or restorative effects.In conclusion, the NSE-adenoviral vector offers the

possibility of transferring potentially therapeutic genesinto neurons within brain areas affected by variousneurodegenerative diseases, such as the hippocampus inAlzheimer’s disease, the striatum and the substantianigra in Parkinson’s disease, and in other neurologicaldisorders such as focal refractory epilepsy.

Materials and methods

Plasmid constructionpNSE contains the 1.8-kb upstream sequence and theuntranslated leader sequence of exon I from the rat NSEgene driving the E. coli LacZ gene. The reporter gene isinserted between the nuclear localizing signal and thepolyadenylation signal from SV40. Le Bert (personal

communication) demonstrated the pan-neuronalexpression of this plasmid in transgenic mice. pAd-RSV,a gift from M Perricaudet, contains the RSV LTR pro-moter driving the same reporter gene.21 This transcrip-tional unit is inserted between the adenoviral packagingsequence and a viral sequence facilitating homologousrecombination.

The SpeI/XbaI fragment of pNSE containing the NSEpromoter-LacZ gene was inserted into pAd-RSV in placeof the SpeI/SalI fragment containing the RSV promoter-LacZ gene. Shuttle plasmids, pAd-NSE and pAd-NSE(R),in which the transcriptional unit was in the same (pAd-NSE) and reverse orientation (pAd-NSE(R)) to pAd-RSV,were isolated.

Recombinant adenovirus generationThe construction of Ad-RSV has been described else-where.21 A similar method was used to obtain the recom-binant (E1-, E3-deleted) adenovirus Ad-NSE. Homolo-gous recombination between the XmnI-linearized pAd-NSE and the large ClaI fragment of adenovirus 5 variantDNA was achieved by cotransfection of the 293 cell lineby the calcium phosphate-DNA precipitation method.Viral stocks were prepared by propagation of the virusesin 293 cells, which have the E1 gene integrated into theirgenome to transcomplement the adenoviral genes. Stockswere purified by ultracentrifugation through two suc-cessive CsCl gradients, and dialyzed on a SephadexG25M (Pharmacia, Uppsala, Sweden) column. Titers ofviral stocks were determined by measuring optical den-sity22 and were checked by the plaque assay method.Functional quantification of adenoviral stocks by genetransfer assay23 in 293 cells demonstrated the absence ofa wild-type virus and a similar bioactivity of both adeno-viruses according to the same number of X-gal-stainedplaques for equivalent dilution of Ad-NSE and Ad-RSV.

In vitro transfection with plasmidsThe rat pheochromocytoma undifferentiated PC12 cellswere grown in RPMI 1640 (Gibco, Life Technologies,Cergy-Pontoise, France) containing 15% fetal calf serum(FCS). The human embryonic kidney 293 cells weregrown in MEM containing 10% FCS and 1% non-essentialamino-acids. The cells were transfected by electropor-ation using a BioRad Gene Pulser (Ivry sur Seine, France)at 190 V and 960 µF. One picomole of each plasmid testedwas used, along with 0.5 µg of a plasmid encodingluciferase. Two days after transfection, cells were placedin a lysis solution containing 25 mm trisphosphate pH7.8, 1 mm dithiothreitol, 8 mm MgCl2, 1 mm EDTA, 1%Triton X-100. β-Galactosidase activity was determined bychemiluminescence assay (Clontech, Palo Alto, CA, USA)in a Lumat LB 9501 luminometer (Berthold, Wildbad,Germany). Luciferase activity was assessed by adding0.08 mm luciferin and 0.1 mm ATP to 50 µl of the lysateand determining luminescence. All the in vitro experi-ments were repeated at least twice.

In vitro infection with adenovirusesThe human cervix carcinoma HeLa cells and the humanneuroblastoma CHP 212 cells were grown in DMEM(Gibco, Life Technologies) containing 10% FCS and inDMEM-F12 (Gibco, Life Technologies) containing 15%FCS, respectively. Differentiated PC12 cells were grownin DMEM containing 10% horse serum, 5% FCS, 50


1891ng/ml nerve growth factor-β (Sigma, St Quentin-Falla-vier, France) and 1% non-essential amino acids for 10days. Primary cultures of superior cervical ganglia andstriatal neurons were obtained from 1- to 2-day-old Wis-tar rats (Iffa-Credo, L’Arbresle, France) and from E15Sprague–Dawley rat embryos (Iffa-Credo) respectively,as described elsewhere.19,24 Adenoviral infections wereperformed 8 days after plating.

β-Galactosidase activity was determined 2 days afterinfection. Protein in the lysate was determined by theBioRad Laboratories’ assay.

In vivo brain injection of adenovirusesViral stocks were diluted to the appropriate concen-tration in PBS and one microliter of each dilution wasslowly injected into the hippocampus (A: 4 mm, L: 2.5mm, V: 7 mm), striatum (A: 8 mm, L: 3 mm, V: 6 mm)and cerebellum (A: −4.5 mm, L: 3 mm, V: 3.5 mm) ofadult male Wistar rats (Iffa-Credo) according to the atlasof Albe-Fessard et al.25 For the expression kinetics study,1.4 × 107 p.f.u. of Ad-NSE were injected into the hippo-campus, and β-galactosidase activity assayed from 7 daysto 6 months (n = 6). This number of adenoviral particleswas previously shown to give high efficiency and only aminor inflammatory response by testing the injection ofvarious dilutions of the viral stock into the hippocampus(data not shown). Ad-NSE was compared with Ad-RSVby injecting 1.4 × 107 p.f.u. of Ad-NSE into the left sideof the hippocampus, and striatum or cerebellum (n = 4).The same amount of Ad-RSV was injected contralaterally(n = 4). Animals were anaesthetized with sodium pento-barbital and were perfused intracardially with 4% par-aformaldehyde. Brains were post-fixed for 2 h, cryopro-tected by incubation for 2 days in 30% sucrose at 4°C, andsectioned (30 µm thick sections) on a freezing microtome.

For β-galactosidase detection, sections were incubatedfor 1 h at 37°C with the chromogenic X-gal substrate(0.1%), in a solution containing 5 mm potassium ferro-cyanide, 5 mm potassium ferricyanide, and 2 mm MgCl2.Immunostaining of E. coli β-galactosidase with rabbitanti-β-gal antibodies (1:800; Amersham, Les Ulis, France)on adjacent tissue sections confirmed the specificityobtained with the X-gal reaction (data not shown). Forimmunostaining, mouse anti-NeuN (1:500; Chemicon,Temecula, CA, USA), anti-vimentine (1:500; Dako,Trappes, France), or rabbit anti-GFAP (1:500; Dako) anti-bodies were used. The secondary antibodies were anti-rabbit antiserum (1:250) coupled to rabbit peroxidase-antiperoxidase for β-galactosidase and GFAP, and biotin-ylated anti-mouse IgG (1:250) coupled to streptavidin-biotinylated-peroxidase complex for NeuN and vimen-tine. Diaminobenzidine with H2O2 was used to detect thesecondary antibodies.

For in vivo quantitative analysis, the number of cellsstained positively for X-gal was counted every five cor-onal sections along a 1 mm segment of the hippocampus.Ad-NSE (n = 4) and Ad-RSV (n = 4) were compared at 7days (n = 1 per virus), 2 weeks (n = 1 per virus) and 3.

months (n = 2 per virus) after injection. From two to fourphotographs were done at a high magnification for eachanalyzed hippocampal section. The number of X-gal cellsthat colocalized or not with the NeuN marker wasassessed by manual counting on each hippocampal sec-tion on blind-coded prints.

AcknowledgementsThe authors thank M Barkats for critical reading of themanuscript, M Le Bert for generously providing pNSE,the NSE promoter fragment of which was obtained fromG Sutcliffe. We also thank V Ridoux for assistance. Thiswork was supported by grants from the Centre Nationalde la Recherche Scientifique, Rhone-Poulenc Rorer, theAssociation Francaise contre les Myopathies, the Associ-ation Francaise Retinis Pigmentosa, and the Institut deRecherche sur la Moelle Epiniere. VN was supported bythe Assistance Publique-Hopitaux de Paris, SM by theMinistere de l’Education Nationale de l’EnseignementSuperieur et de la Recherche.

References1 Anderson WF. Human gene therapy. Nature 1998; 392: 25–30.2 Le Gal La Salle G et al. An adenovirus vector for gene transfer

into neurons and glia in the brain. Science 1993; 259: 988–990.3 Akli S et al. Transfer of foreign genes into the brain using aden-

ovirus vectors. Nat Genet 1993; 3: 224–228.4 Hashimoto M et al. A neural cell-type-specific expression system

using recombinant adenovirus vectors. Hum Gene Ther 1996; 7:149–158.

5 Marangos PJ. Neuron-specific enolase, a clinically useful markerfor neurons and neuroendocrine cells. Ann Rev Neurosci 1987;10: 269–295.

6 Sakimura K et al. Upstream and intron regulatory regions forexpression of the rat neuron-specific enolase gene. Mol Brain Res1995; 28: 19–28.

7 Forss-Petter S et al. Transgenic mice expressing β-galactosidasein mature neurons under neuron-specific enolase promoter con-trol. Neuron 1990; 5: 187–197.

8 Martinou JC et al. Overexpression of BCL-2 in transgenic miceprotects from naturally occurring cell death and experimentalischemia. Neuron 1994; 13: 1017–1030.

9 Babiss LE, Friedman JM, Darnell JE. Cellular promoters incor-porated into the adenovirus genome: effects of viral regulatoryelements on transcription rates and cell specificity of albuminand β-globin promoters. Mol Cell Biol 1986; 6: 3798–3806.

10 Sakimura K, Kushiya E, Takahashi Y, Suzuki Y. The structureand expression of neuron-specific enolase gene. Gene 1987; 60:103–113.

11 Hannas-Djebbara Z et al. Transgene expression of plasmidDNAs directed by viral or neural promoters in the rat brain.Mol Brain Res 1997; 46: 91–99.

12 Klein RL et al. Neuron-specific transduction in the rat septohip-pocampal or nigrostriatal pathway by recombinant adeno-asso-ciated virus vectors. Exp Neurol 1998; 150: 183–194.

13 Byrnes AP, Wood MJA, Charlton HM. Role of T cells ininflammation caused by adenovirus vectors in the brain. GeneTherapy 1996; 3: 644–651.

14 Palmer TD, Rosman GJ, Osborne WRA, Miller AD. Geneticallymodified skin fibroblasts persist long after transplantation butgradually inactivate introduced genes. Proc Natl Acad Sci USA1991; 88: 1330–1334.

15 Davidson BL et al. A model system for in vivo gene transfer intothe central nervous system using an adenoviral vector. Nat Genet1993; 3: 219–223.

16 Andersen JK, Frim DM, Isacson O, Breakefield XO. Herpesvirus-mediated gene delivery into the rat brain: specificity andefficiency of the neuron-specific enolase promoter. Cell Mol Neu-robiol 1993; 13: 503–515.

17 Dedieu JF et al. Long-term gene delivery into the livers of immu-nocompetent mice with E1/E4-defective adenoviruses. J Virol1997; 71: 4626–4637.

18 Kochaneck S et al. A new adenoviral vector: replacement of allviral coding sequences with 28 kb of DNA independentlyexpressing both full-length dystrophin and β-galactosidase. ProcNatl Acad Sci USA 1996; 93: 5731–5736.


1892 19 Barkats M et al. An adenovirus encoding CuZnSOD protects cul-tured striatal neurons against glutamate toxicity. NeuroReport1996; 7: 497–501.

20 Robert JJ et al. Adenovirus-mediated transfer of a functionalGAD gene into nerve cells: potential for the treatment of neuro-logical diseases. Gene Therapy 1997; 4: 1237–1245.

21 Stratford-Perricaudet LD, Makeh I, Perricaudet M, Briand P.Widespread long-term gene transfer to mouse skeletal musclesand heart. J Clin Invest 1992; 90: 626–630.

22 Maizel JV, White DO, Scharff MD. The polypeptides of aden-

ovirus. I. Evidence for multiple protein components in the virionand a comparison of types 2, 7A, and 12. Virology 1968; 36:115–125.

23 Mittereder N, March KL, Trapnell BC. Evaluation of the concen-tration and bioactivity of adenovirus vectors for gene therapy.J Virol 1996; 70: 7498–7509.

24 Hawrot E, Patterson PH. Long-term culture of dissociated sym-pathetic neurons. Meth Enzymol 1979; 58: 574–584.

25 Albe-Fessard D, Stutinsky F, Libouban S. Atlas Stereotaxique duDiencephale du Rat Blanc. CNRS: Paris, 1971.

Documents

gt Efﬁcient gene transfer and long-term expression in