Address correspondence and reprint requests to Prof. J.-L. Dreyer, Institute of Biochemistry, University of Fribourg, Rue du Musée 5, CH-1700 Fribourg, Switzerland. E-mail: email@example.com
The tetraspanin CD81 is induced in the mesolimbic dopaminergic pathway after cocaine administration. To further investigate its role, a regulatable lentivirus (Lenti-CD81) bearing the CD81 gene under the control of a tetracycline-inducible promoter and lentiviruses expressing short hairpin RNA (shRNA) targeted against CD81 (Lenti-CD81-shRNAs) have been prepared. Infection of HEK293T cells in vitro with Lenti-CD81-shRNAs resulted in 96.5% gene silencing (from quantitative real-time PCR(qRT-PCR) and immunocytochemistry). In vivo delivery of Lenti-CD81-shRNA into the nucleus accumbens or ventral tegmental area resulted in 91.3 and 94% silencing of endogenous CD81, respectively. Stereotaxic injection of Lenti-CD81 into these regions, resulting in CD81 overexpression, induced a four- to fivefold increase in locomotor activity after chronic cocaine administration, which returned to basal levels when Lenti-CD81-shRNA had been coinjected or when CD81 expression was blocked by doxycycline. Furthermore, silencing endogenous CD81 in vivo resulted in a significant decrease in locomotor activity over controls, again suppressing cocaine-induced behaviour.
Administration of drugs of abuse induces strong molecular adaptations and plasticity within the mesolimbic dopamine system, a pathway essential for reward-seeking behaviour. These adaptations underlie a complex rewiring of neural circuitry that results in the behaviours associated with addiction (Robinson and Berridge 1993; Nestler 2000). Addictive drugs (cocaine and amphetamines), depressants (ethanol) and opiate narcotics (heroin and morphine) are powerful reinforcers and produce their rewarding effects of euphoria or pleasure through an interaction with the mesolimbic dopamine system (Nestler 2000). Little is known about the specific targets involved in this neuroadaptation process but it has been suggested that cocaine and other drugs of abuse may alter the morphology of neuronal dendrites and spines, the primary site of excitatory synapses in the brain, by means of inducing expression changes of surface molecules (Nestler 2000; Yue et al. 2002). Complex changes in expression of a number of surface axon guidance molecules have been observed upon cocaine administration, which may underlie important neuroplastic changes in the reward- and memory-related brain centres after drug action (Bahi and Dreyer, 2005). Local expression changes of these cues may mediate plasticity and cytoskeleton rearrangement through mechanisms similar to synaptic targeting during development. In addition, strong induction of a surface tetraspanin protein involved in cell adhesion, CD81, has been described (Halladay et al. 2000; Brenz-Verca et al. 2001; Michna et al. 2001). Using a regulatable lentivirus bearing the rat CD81 gene under the control of a tetracycline-inducible system, previous studies have shown that CD81 expression in the mesolimbic dopaminergic pathway contributes significantly to behavioural changes associated with chronic cocaine administration (Brenz-Verca et al. 2001). Overexpression of CD81 in this pathway induces a four- to fivefold increase in locomotor activity, which can be reversed to normal in the same animal fed doxycycline.
In this study we further characterized the role of CD81 in drug-induced behavioural changes. Cocaine-induced expression changes of CD81 were assessed under different protocols of drug administration. Lentiviruses expressing short hairpin RNA (shRNA) targeted against different regions of the CD81 mRNA were developed and used for silencing endogenous CD81 or overexpressed CD81 in vivo within the mesolimbic dopaminergic pathway. We demonstrate that stereotaxic injection of shRNA-expressing lentiviruses into the nucleus accumbens (NAcc) or ventral tegmental area (VTA) results in 93–95% down-regulation of CD81 in these regions and this effect is associated with suppression of cocaine-induced behavioural changes.
Materials and methods
Animal handling and drug administration
Animals used in this experiment were male Wistar rats weighing 225–250 g (BRL, Fillinsdorf, Switzerland). All animal experiments were carried out in accordance with the guidelines and regulations for animal experimentation of Bundesamt für Gesundheitswesen (Bern, Switzerland). Animals were housed in trios in clear plastic cages with wire grid lids and kept in the animal facility maintained on a 12-h light/12-h dark cycle (lights off at 07:00 h). Access to food and water was unrestricted.
Drug administration protocols
Three different protocols for drug administration were used, as described in Fig. 1: acute paradigm, animals (n = 6) were i.p. injected once with 15 mg/kg cocaine-HCl (Sigma Chemical Co., Buchs, Switzerland); chronic paradigm, animals (n = 6) were i.p. injected daily with 15 mg/kg cocaine-HCl for a period of 15 days; and binge paradigm, animals (n = 6) were i.p. injected with 30 mg/kg cocaine-HCl every 2 h for four injections.
In these three protocols control animals received 0.9% saline i.p. injection instead of drug. At 24 h after the last injection, animals were killed by decapitation and the brain areas were dissected out and used for isolation of total RNA using a RiboPure Kit (Ambion, Huntingdon, UK) followed by RNA amplification according to previous publications (Bahi and Dreyer 2005; Bahi et al. 2004).
After killing, brain regions were dissected out (NAcc and VTA) and total RNA was extracted using a RiboPure Kit (Ambion) including an RNase-free DNase step according to the manufacturer's protocol. Briefly, 1 mL of RNA was added to 50 mg of tissue, homogenized with a Polytron homogenizer, vortexed for 30 s and incubated at room temperature (20°C) for 5 min. Residual protein was removed by the addition of 0.2 mL chloroform, mixing for 30 s, incubation at room temperature for 5 min, vortexing for 30 s and centrifugation for 15 min at 12 000 g and room temperature. The aqueous phase was precipitated with 0.5 mL of absolute ethanol, filtered over a column and washed by centrifugation. The total RNA was then eluted with 0.1 mL of diethylpyrocarbonate (DEPC)-treated water, incubated at 60°C for 10 min and stored at −80°C. RNA was quantified by spectrophotometry and its integrity verified by agarose gel electrophoresis and visualized with ethidium bromide staining.
Lenti-CD81 and Lenti-Green flourescent protein (GFP) lentiviruses
The CD81 gene (GenBank Accession no. NM_013087) was amplified by RT. Briefly, 2 µg of total RNA (prepared from NAcc of cocaine-treated rats) was added to 1 µg Oligo-(dT)12−18 and 2 µL of dNTP Mix at 10 mm each and made up to 12 µL with RNase-free water. These components were mixed and heated at 65°C for 5 min and then kept on ice. To the mixture 4 µL of 5 × First Strand Buffer was added, followed by 2 µL 0.1 m dithiothreitol, 10 U RNAsin (Invitrogen, Abingdon, UK) and 1 µL 200 U/µL Superscript II RNaseH-reverse transcriptase. The mixture was incubated at 42°C for 3 h. To remove RNA–DNA hybrids, 2 U RNase H was added and incubated at 37°C for 30 min. The cDNA was then PCR amplified and 6His-tagged with the two following primers: CGCGGATCCGCGATGGGGTGGAGGGCTGC as forward primer and CCGCTCGAGCGGTTAATGATGATGATGATGATGGTACACGGAGCTGTTCCGG as reverse primer. The forward primer contains a BamHI (Biolabs, Allschwill, Switzerland) restriction site followed by the 5′ rat CD81 cDNA-specific sequence; the reverse primer contains the 3′ rat CD81 cDNA-specific sequence, a 6His-tag, a stop codon and an XhoI (Biolabs) restriction site.
The PCR product was digested with BamHI and XhoI and cloned into similar sites in pTK431 (Fig. 2). The pTK431 is a self-inactivating HIV-1 vector which contains the entire tet-off-inducible system, the central polypurine tract (cPPT) and the woodchuck hepatitis virus post-transcriptional regulatory element. It was generated by ligating a BglII/BamHI DNA fragment containing the tetracycline-regulated transactivator (tTA) (5′) and the tetracycline-inducible promoter (3′) into a BamHI site downstream of a cytomegalovirus (CMV) promoter in a self-inactivating HIV-1 vector. A control vector construct, pTK433, in which GFP expression is regulated by a tetracycline-inducible promoter, was generated by cloning a BamHI/BglII DNA fragment containing the GFP gene into a BamHI site in pTK431. All plasmids were CsCl2 purified.
Construction of Lenti-CD81-shRNAs
To silence CD81 expression in vitro and in vivo, five targets were designed according to the CD81 mRNA sequence (GenBank Accession no. NM_013087). The following targets within the CD81 sequence were selected (Fig. 2), based on Hannon's design criterion (http://katahdin.cshl.org:9331/RNAi/html/rnai.html): first target, 907–927; second target, 290–317; third target, 907–927 (with an extra cytosine compared with the first target); fourth target, 848–866; and fifth target, 676–696. An XhoI restriction site was 3′ added to each oligo. Using the pSilencer 1.0-U6 (Ambion) as a template and a U6 promoter-specific forward primer containing BamHI restriction site GCGGATCCCGCTCTAGAACTAGTGC, each shRNA target was added to the mouse U6 promoter by PCR (each target contains a 3′-specific U6 promoter-specific primer). The PCR conditions were drastic to avoid mutations within the targets. The following PCR programme was performed: 120 s at 94°C (initial denaturation) followed by 94°C for 45 s, 64°C for 45 s and 72°C for 45 s repeated for 35 cycles. The PCR reaction contained 4% dimethyl sulphoxide (Sigma Chemical Co.). The PCR product was digested with BamHI and XhoI, cloned into similar sites in pTK431 and sequenced to verify the integrity of each construct.
The vector plasmids (either pTK431-CD81-6His, pTK433-GFP or pTK431-U6-shRNAs), together with the packaging construct plasmid pΔNRF and the envelope plasmid pMDG-VSVG, were cotransfected into HEK293T cells to produce the viral particles (Naldini et al. 1996; Bahi et al. 2004). The viral titres were determined by p24 antigen measurements (KPL, Lausanne, Switzerland). For the in vivo experiments, the different viral stocks matched for viral particle content were used at 0.4 mg/mL of p24.
In vitro assays
The efficiency of the Lenti-CD81-shRNAs was tested in vitro by infection of HEK293T cells. The day before the infection, 3 × 105 HEK293T cells were plated per well in six-well plates. The next day, 3 µL from lentivirus stocks was added with Polybrene (10 µg/mL final concentration; Sigma Chemical Co.), incubated for 30 min at room temperature and the cells plated at 37°C for 24 h. The next day, the medium was replaced with normal growth medium and cells were left for 24 h. One part of the cells was used for RT-PCR after total RNA isolation and the remaining part was used for immunocytochemistry.
Real-time PCR and CD81 mRNA quantification. Total RNA was extracted from the HEK293T cells using a RiboPure Kit (Ambion), including an RNase-free DNase step, and stored at −80°C according to the manufacturer's protocol. RNA was quantified and its integrity verified. First-strand cDNA was generated from 1 µg of total RNA and Oligo-(dT)12−18 primer with the M-MLV RT kit (Invitrogen) in a total volume of 20 µL according to the manufacturer's protocol. Quantitative real-time PCR (qRT-PCR) was used to quantify the CD81 mRNA level. Primer sets were designed to amplify 100–300-bp products using PRIMER3 software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_http://www.cgi).
The RT reaction was amplified (quantitative real-time PCR was performed) with the following pairs of oligonucleotides specific for rat CD81, glyceraldehyde-3-phosphate dehydrogenase (GAPDH) and β-actin: CD81, 5′-TGATCCTGTTTGCCTGTGAG-3′ and 5′-CAGTTGAGCGTCTCATGGAA-3′; GAPDH, 5′-ATGACTCTACCCACGGCAAG-3′ and 5′-CATACTCAGCACCAGCATCAC-3′; β-actin, 5′-AGCCATGTACGTAGCCATCC-3′ and 5′-CTCTCAGCTGTGGTGGTGAA-3′. GAPDH and β-actin were used as internal controls. The quantification was performed using the real-time PCR iCycler (Bio-Rad, Reinach, Switzerland). For PCR, 5 µL cDNA preparation, 0.5 µm of forward and reverse primers and 10 µL of IQ SYBR Green Supermix (Bio-Rad) in a total volume of 20 µL were applied. The following PCR programme was performed: 3 min at 95°C (initial denaturation), a 20°C/s temperature transition rate up to 95°C for 45 s and 62°C for 45 s, repeated 40 times (amplification). The PCR reaction was evaluated by melting curve analysis following the manufacturer's instructions and checking the PCR products on 2% agarose gel.
Values for CD81 overexpression were calculated against the signal for GAPDH (or β-actin), selected as endogenous, constitutively expressed control mRNAs for normalization. No difference was observed whether GAPDH or β-actin was used for all normalizations. The levels for NAcc or VTA samples from cocaine-injected rats were normalized to the levels for the samples from saline-treated animals within the same experiment. PCR amplifications were performed at least in triplicate from three-pooled samples isolated from separate pools of animals or cells and the PCR cycle number at which each assay target reached the threshold detection line was determined (threshold cycles, Ct value) using the second derivative of the reaction. The Ct of each gene was normalized against that of GAPDH (or β-actin). To determine the linearity and detection limit of the assay, cDNA samples were amplified for successive 10-fold dilutions in a series of real-time PCRs, using duplicate assay on each dilution, so that the correlation coefficient could be calculated from the standard curve of Ct values. Comparisons were made between cocaine and saline groups, significance was calculated using a two-tailed Student's t-test and the level of statistical significance was set at p < 0.05. Data were expressed as mean ± SEM. The ΔCt for each candidate was calculated as ΔCt = [Ct (candidate) − Ct (GAPDH or β-actin)]. The relative abundance of each target in each protocol can be calculated as the ratio between treated and untreated samples (Mühlbauer et al. 2004).
Immunocytochemistry on HEK293T cells. After infection cells were washed twice with 1 × phosphate-buffered saline (PBS), fixed with pre-warmed 4% Performic acid (PFA) for 30 min and washed again with 1 × PBS buffer. Cells were blocked for non-specific binding with 3% normal goat serum diluted in 1 × PBS for 20 min and incubated overnight at 4°C with mouse histidine tag antibody (MCA1396; Serotec, Oxford, UK) diluted 200-fold in 1 × PBS containing 1% normal goat serum and 0.1% Triton-X100. Cells were rinsed three times with 1 × PBS and incubated with the secondary antibody (Texas red-conjugated goat anti-mouse immunoglobulin G; Molecular Probes, Basel, Switzerland) diluted 1000 times in 1% normal goat serum and 0.1% Triton-X100 in the dark at room temperature for 3 h. Cells were washed three times with 1 × PBS and cover-slipped with a medium containing glycerol in 1 × PBS (AF1 mounting solution; Citifluor Ltd, Lucern, Switzerland). In all cases negative controls included omission or substitution of the primary antibody.
Fluorescence microscopy. Stained cells were observed using a multifluorescence microscope (Axioplan 2 imaging; Zeiss, Zurich, Switzerland) with a × 63 objective and photographed using a multichannel camera (Axiocam; Zeiss) linked to acquisition software (Axiovision system 3.1). The fluorophore was detected with the appropriate detecting system (HAL 100). Texas red was exited by a 595-nm beam and detected through a light path ranging from 600 to 660 nm.
Quantification. Cells were photographed with a × 40 objective using a digital camera (Axiocam; Zeiss) attached to a multifluorescence microscope (Axioplan 2 imaging; Zeiss). Special care was taken to place the frame in the same location on each region analysed for each slide. Within each section, frame size was kept constant from slide to slide. The density of CD81-positive cells was calculated for each slide using the Scion software. Each image was transformed into TIFF format, normalized by subtracting the background. A semiautomatic cellular density measurement was then undertaken by calculating total pixel density at a fixed brightness level. Separate exposures were made for each section to ensure that density measurements were reproducible. The high mean level of pixel intensity correlated to high amounts of expression of CD81 in the specific area. anova followed by Scheffe F-tests was used for each set of CD81 immunoreactivity measurements. All results are expressed as mean ± SEM.
In vivo assays
Stereotaxic surgery and injection of the lentiviral vector
Before the operation, rats were injected with ketamine/xylazine 100 mg/kg/10 mg/kg, i.p.). After 10 min, the surgical field was shaved, cleaned using Hibitane solution (0.5% chlorhexidine gluconate in 70% ethanol) and Lacrilube was applied to the eyes. Animals were placed into a stereotaxic frame (Stoelting, Wood Dale, IL, USA). Stereotaxic injections were performed using a 5-µL Hamilton syringe with a 33-gauge tip needle. The injections were performed in the VTA (anterior, − 6; lateral, ± 0.5; ventral, − 8.2) and NAcc (anterior, + 1.4; lateral, ± 1.6; ventral, − 6.8) as calculated from bregma and the dura mera. The skin was opened with a scalpel incision along the midline and 0.2 mL of Mercain (0.25% bupivacaine hydrochloride) was applied to the wound for local analgesia. Overlying tissue was blunt dissected with sterile Q-tips to visualize bregma and the skull was delicately bored at pre-determined co-ordinates. Heating of the skull was reduced by application of sterile 0.9% saline solution around the drilling site. Male Wistar rats were bilaterally injected into the VTA or NAcc with 4 µL of concentrated lentiviral stock (coinjection of 2 µL of Lenti-CD81 or Lenti-GFP together with 2 µL of Lenti-shRNAs) slowly infused at a speed of approximately 0.2 µL/min. The cannula was then maintained in place for a further 5 min after injection and then withdrawn very slowly to prevent backflow of solution. The wound site was closed with braided silk suture. Rats were then removed from the frame and injected with a mixture of 2.5 mL of 0.9% saline solution and Rimadyl (carprofen 2 mg/kg, s.c.) for hydration and pain relief. Lacrilube was reapplied to the eyes. The animal was returned to the cage from the surgery room and kept warm and off sawdust (as this can be inhaled by rats which are not fully conscious). The post-operative diet for 3 days consisted of normal food pellets soaked in water. Rat recovery was followed up.
Drug treatment and locomotor activity measurement
Locomotor activity was monitored during the dark cycle. Each subject was weighed 1 week after surgery; the animal was brought from the animal facility and placed immediately into the activity-monitoring cage for a 30-min baseline without drug. After a 30-min period, the session automatically paused and, during this interval, each subject received cocaine-HCl (15 mg/kg, i.p.) and was then placed back into the locomotor activity-monitoring cage for 60 min (Bahi et al. 2004).
Locomotor activity was monitored in MED-OFA-RS cages (MED Associates Inc., Hampton, UK). Animals were placed in a square (43.2 × 43.2 × 30.5 cm) PVC retainer. Activity in the monitor was recorded by photobeam interruptions. A ring of 16 sensors, spaced 2.54 cm from each other, measured the X–Y location of the animal four times per second. The activity monitor computed the location of the animal in each of the X and Y dimensions as the middle point between the extreme beam interruptions in this dimension. Two sets of strips were used for the X–Y ambulatory data input while a single set was used for the Z rearing data.
Speed was estimated by computing the SD of the distances of the datapoints to their mean within a sliding time window 0.4 s wide, constructed around each datapoint in turn. The box size was defined with two photobeams of X or Y to be broken before a movement was considered ambulatory. Starting at time 0, the box was centred on the subject. When the animal moved to the outside of the box, it was considered ambulatory and the box recentred on the subject. The subject remained ambulatory until it did not leave the last recentred box in less than the resting delay. The travelled distance was then calculated. Stereotypic counts were defined as the total number of beam breaks inside the box.
Statistical evaluation of behavioural analysis
The distance travelled was summed over each test session and analysed by means of one-way anova, with the data divided into 5-min bins. Appropriate post-hoc comparisons were used to compare groups if a significant interaction was found (p < 0.05). These comparisons were made using Fisher's least significant difference test. Independent Student's t-tests were used when a significant interaction was found. All of these analysis were performed using SPSS (Zurich, Switzerland) software (Okabe and Murphy 2004; Podhorna and Didriksen 2004; Sanchis-Segura et al. 2004). The use of SD is based on the analysis of the coordinates of the path. This is because estimating the speed, measured as travelled distance/min, at a given time-point ti depends on measuring the coordinates at two time-points (e.g. the difference between the coordinates at times ti and ti − 1).
Within the same group, the means and SDs were performed on measurements performed each day. No significant changes were observed in day to day recordings over the same session. The means and SDs from all measurements within one session were then calculated (using values obtained from subjects of the same group at the same session) and plotted (Bahi et al. 2004).
At the end of the session, animals were killed by decapitation. Rat brains were rapidly dissected out, frozen in isopentane upon extraction (at −30°C for 3 min) and kept at −25°C. Coronal sections were cut at 14 µm in a cryostat (Leitz, Aarau, Switzerland), placed on gelatinized glass slides, air-dried at room temperature for 20 min and kept at −25°C until further processing. Antigens were localized using the avidin-biotin-peroxidase technique. Slices were fixed in 4% PFA for 15 min and washed three times with 1 × PBS. Endogenous peroxidase activity was quenched with 2% hydrogen peroxide in water for 40 min at room temperature. Non-specific binding was blocked for 30 min at room temperature in 1 × PBS containing 1% bovine serum albumin, 1% Triton X-100 and 3% normal goat serum. Sections were incubated overnight with mouse anti-histidine antibody (MCA1396; 1 : 12 000; Serotec) diluted in 1 × PBS containing 1% Triton X-100 and 1% normal goat serum. Sections were then washed three times with 1 × PBS and incubated with the biotinylated secondary antibody (goat anti-mouse immunoglobulin G; 1 : 500; Vector Laboratories, Burlingame, CA, USA) for 45 min at room temperature. Sections were rinsed three times for 5 min in 1 × PBS at room temperature, followed by avidin-biotin complex (Vector Laboratories) in 1 × PBS solution. After three rinses in 1 × PBS, all sections were developed in 0.025% 3-3′ diaminobenzidine tetrahydrochloride plus 0.02% hydrogen peroxide for 10–15 min. Sections were then dehydrated, mounted in permanent medium (Eukitt, Freibury, Germany) and examined with a light microscope (Zeiss). GFP-expressing sections were cover-slipped with AF1 mounting solution (Citifluor Ltd) and directly observed using a multifluorescence microscope (Axioplan 2 imaging; Zeiss) mounted with a multichannel camera (Axiocam; Zeiss) linked to acquisition software (Axiovision 3.1; Zeiss). In all cases negative controls included omission or substitution of the primary antibody.
Cocaine-induced CD81 expression
CD81 was found initially in a screening for genes up-regulated after cocaine administration following a binge protocol. In order to further characterize this observation, up-regulation of CD81 was confirmed in different paradigms of drug administration, summarized in Fig. 1, i.e. acute, binge and chronic treatments. The total RNA from the NAcc and VTA of animals treated under the different paradigms (Fig. 1, n = 6) was prepared, reverse transcribed and CD81 expression was assessed by means of quantitative real-time PCR analysis after normalization against GAPDH, β-actin and Heat shock protein 70 (HSP70) considered as endogenous controls (no difference was observed between all normalizations). The results are shown in Fig. 3. The strong up-regulation observed after binge treatment in these two brain regions could be confirmed; CD81 displays a 6.2-fold up-regulation in the NAcc and a 3.9-fold up-regulation in the VTA, in agreement with previous studies (Brenz-Verca et al. 2001). Furthermore, CD81 was also 2.9–4.0-fold up-regulated after chronic treatment in these regions. On the other hand, acute treatment induced small but significant overexpression of CD81, yielding a 1.4-fold up-regulation in the VTA and NAcc. Subsequent studies (locomotor activity monitoring) were then performed according to the chronic protocol but the silencing effect of the endogenous CD81 in the NAcc and VTA was followed-up after binge treatment, which induces the highest expression of endogenous CD81.
shRNA lentiviruses and in vitro CD81 silencing
In order to locally knock down CD81 expression in selected brain areas, lentivirus-based RNA interference was used. Five different shRNAs were designed, targeted against different regions of the CD81 mRNA, as displayed in Fig. 2 (as described in Materials and methods). Constructs bearing a U6 promoter were inserted into the transfer plasmid of the lentivirus system. Lenti-CD81-His6 (Bahi et al. 2004), lenti-GFP (Bahi et al. 2004) and lenti-CD81-shRNAs all share the same vector backbone; however, lenti-CD81-shRNAs are not regulatable due to the presence of the U6 promoter. Incorporation of the tetracycline-inducible system into the lentivirus vectors conferred the ability to control transgene expression in vivo (Kafri et al. 2000; Reiser et al. 2000). The tetracycline-inducible system is based on two components: (i) tTA, which is a fusion protein of the tet repressor, and the transactivator domain of the herpes simplex virus protein VP16 and (ii) the inducible promoter, which contains a minimal promoter and seven copies of the Tet operon. In the absence of tetracycline, the tTA binds and activates the inducible promoter. Binding of tetracycline (or its more potent analogue doxycycline) to the tTA induces conformational changes that render the tTA incapable of binding to the Tet operon and thus abolish transgene expression (Haack et al. 2004). A CMV promoter located in the middle of the vector genome constitutively expresses the tTA, while the inducible promoter at the 3′ end of the vector regulates the expression of the transgene of interest (Kafri et al. 2000).
To assess the efficiency of the five different targets at silencing CD81, in vitro assays were performed in HEK293T cells coinfected with Lenti-CD81. As controls, cells coinfected with Lenti-GFP (instead of Lenti-CD81-shRNAs) and Lenti-CD81 were used. At 48 h after infection, cells were harvested, total RNA was extracted and CD81 mRNA measured by means of quantitative RT-PCR. As shown in Figs 4a and b, shRNA-3 and -5 induced 88 and 85% decreases in CD81 mRNA levels, respectively. The other targets, shRNA-1, -2 and -4, were less efficient at silencing CD81 in vitro, resulting in 62, 56 and 45% expression of CD81 mRNA, respectively. However, when all targets were coinfected together, CD81 silencing was > 96.5% (Figs 4a and b). It should be noted that Lenti-CD81-shRNA3 and Lenti-CD81-shRNA1 target the same region of the mRNA yet differ only very slightly in their sequence; in Lenti-CD81-shRNA3 one C has been deleted compared with Lenti-CD81-shRNA1. It is interesting to observe that this very minor change in the sequence results in much more efficient silencing of Lenti-CD81-shRNA3 compared with Lenti-CD81-shRNA1 and makes it the best among all candidates tested, whereas the other may serve as a very good negative control.
Protein expression was also tested by means of immunocytochemistry; cells cotransfected with the Lenti-CD81-shRNAs (all targets together) and Lenti-CD81 display 96–97% silencing of CD81 (Figs 4c and d). For in vivo assays, only the two most efficient Lenti-CD81-shRNAs, i.e. Lenti-CD81-shRNA-3 and Lenti-CD81-shRNA-5, were used and coinfected.
In vivo knock-down of endogenous CD81
To test the in vivo silencing efficiency of the Lenti-CD81-shRNAs, the Lenti-CD81-shRNA-3 and Lenti-CD81-shRNA-5 were stereotaxically coinjected into the NAcc or VTA of animals (n = 6). Control animals were injected with Lenti-GFP (instead of Lenti-CD81-shRNAs) into the same areas. Endogenous CD81 is expressed at low levels in the rat brain under normal conditions but is strongly induced in the mesolimbic dopaminergic pathway upon cocaine administration, especially after a high-dose cocaine treatment (binge protocol, Fig. 3). Therefore, in order to optimally induce the expression of endogenous CD81 in the targeted areas, animals were treated according to the binge protocol (Fig. 1; 120 mg/kg cocaine in a single day), which strongly promotes CD81 expression in the mesolimbic dopaminergic pathway (Fig. 3; Brenz-Verca et al. 2001; Bahi et al. 2004). Seven days after surgery, animals were injected with 30 mg/kg of cocaine i.p. every 2 h for four injections and killed 24 h after the last injection. Control animals, i.e. animals treated with Lenti-GFP, received cocaine injections under the same schedule. After killing, the brain regions of interest were dissected out and the total RNA prepared. Quantitative real-time PCR was then performed on cDNA from these RNA preparations and the mRNA levels were quantified and normalized against GAPDH considered as an endogenous control (normalization against β-actin and HSP-70 was also performed and gave practically the same results, data not shown). As shown in Fig. 5, coinjection of Lenti-CD81-shRNA-3 and -5 together resulted in 91.3 and 94% decrease in endogenous CD81 mRNA levels in the VTA and NAcc, respectively, compared with CD81 mRNA levels observed in control animals (infected with Lenti-GFP in the same brain areas).
Behavioural changes induced upon CD81 expression
Four groups of animals (n = 6) were used for assessing the effects of Lenti-CD81-shRNAs on cocaine-induced behavioural changes. One group of animals was injected with the doxycycline-regulatable Lenti-CD81 into the NAcc. After surgery, animals were fed without doxycycline (normal water), enabling full expression of CD81 in the targeted area. Seven days after surgery the animals were receiving chronic cocaine i.p. (15 mg/kg) and the locomotor activity was monitored daily after drug administration. As shown in Fig. 6(a) (Session A, left panels), after cocaine administration a large rise in locomotor activity was observed with a peak at 19 192 mm/min observed 15 min after drug administration, in agreement with previous observations (Bahi et al. 2004). After 5 days the same animals were fed doxycycline in the drinking water (inducing down-regulation of lentivirus-expressed CD81 in the targeted area) and their behaviour upon chronic cocaine administration was further evaluated for five consecutive days. Under these conditions locomotor activity dropped back almost to control levels after cocaine injection, with a peak at 6302 mm/min (Fig. 6a, Session B). Five days later, doxycycline was finally removed (enabling re-expression of lentivirus-mediated CD81 in the NAcc). This restored locomotor activity to its initial levels, with a peak of 18 852 mm/min at 45 min (Fig. 6a, Session C). The difference between peaks in both sessions A and C was not significant.
Another set of control animals was infected with the doxycycline-regulatable control lentivirus, Lenti-GFP, into the same brain area (NAcc) and submitted to the same cycles of chronic drug administration, with regimens consisting initially of 5 days in the absence of doxycycline (Session A) followed by 5 days under doxycycline (Session B) and finally 5 days under removal of the doxycycline (Session C). As shown in Fig. 6(a) (rightmost panels), cocaine administration induced a small raise in locomotor activity, with a peak of 3780 mm/min at 40 min. However, this behaviour was independent of GFP expression. Another control group consisting of naive animals, not treated with lentivirus, behaved like the group infected with Lenti-GFP when receiving the same doses of cocaine (data not shown). The observed locomotor activity was not significantly affected by the doxycycline regimen; locomotor activity displayed a peak at 3647 mm/min at 40 min in the presence of doxycycline and 3888 mm/min when doxycycline was again removed.
A third group of animals was treated by coinjection of the regulatable Lenti-CD81 (same concentration as the first group) together with (non-regulatable) Lenti-CD81-shRNA-3 and Lenti-CD81-shRNA-5. Animals were then submitted to the same regimen and drug treatment as the other groups; 7 days after surgery, chronic cocaine administration was initiated in the absence of doxycycline, i.e. under conditions where lentivirus-mediated CD81 expression was achieved. As shown in Fig. 6(a) (middle left panels), in the absence of doxycycline, locomotor activity was low, with a peak of 5105 mm/min at 45 min, corresponding to only 8% of the locomotor activity observed under the same conditions in the absence of silencing Lenti-CD81-shRNAs, i.e. almost 92% of the behavioural effect was abolished with the Lenti-CD81-shRNAs when basic locomotor activity was subtracted (Lenti-GFP-infected animals) (Fig. 6a, session A, left panel). After 5 days, doxycycline was added to the regimen (session B) inducing down-regulation of exogenous, lentivirus-mediated CD81 expression but not of lentivirus-mediated shRNA because Lenti-CD81-shRNAs are not regulated by doxycycline (due to the presence of two distinct promoters in the lentivirus construct). As shown, locomotor activity decreased further under these conditions and reached a peak of 3100 mm/min at 45 min, i.e. the locomotor activity under these conditions was significantly lower (40%) than in the corresponding GFP-treated animals. This low activity probably corresponds to silencing of the endogenous CD81 which, in the absence of silencing shRNAs (first two groups of animals), would be strongly induced under these conditions (see Fig. 3, chronic treatment). In fact, the level of exogenous CD81 was decreased drastically by doxycycline. Upon removal of doxycycline 5 days later (session C), this group of animals displayed no significant rise in locomotor activity; under these conditions lentivirus-mediated CD81 overexpression would be restored but, at the same time, it was clearly silenced by the expression of Lenti-shRNAs.
A fourth group of animals was treated by coinjection of the Lenti-CD81-shRNA-3 and Lenti-CD81-shRNA-5 (same doses as previous group) together with Lenti-GFP and submitted to the same cycles of regimen and drug administration (Fig. 6a, middle right panels). This control group was aimed at testing the effects of Lenti-CD81-shRNAs on expression of the endogenous protein in the absence of exogenous CD81. Under all conditions and regimens, locomotor activity was slightly but significantly lower than the activity observed with the second group of (control) animals which were treated with Lenti-CD81 only. In the initial 5-day cycle (i.e. in the absence of doxycycline, session A), a peak of 3347 mm/min was observed, significantly lower (p < 0.08) than in the second group under the same conditions. In the presence of doxycycline (session B), a peak of 3103 mm/min was observed, identical to the level reached by the third group of animals (3100 mm/min) but significantly lower (p < 0.08) than the control group, which had been injected with Lenti-GFP only (3647 mm/min). Finally, when doxycycline was removed (session C), these animals displayed no change in locomotor activity compared with the third group (3424 vs. 3209 mm/min, respectively) but the observed activity was still significantly lower than the second group (treated with Lenti-GFP only) under the same conditions (3888 mm/min). This decrease in activity over the control groups again should be related to silencing of endogenous CD81 and further confirms the previous observations.
A similar series of experiments was conducted on four groups of animals where the viruses were injected into the VTA instead of the NAcc. The same results were observed, as shown in Fig. 6(b). Cocaine-induced locomotor activity upon CD81 expression was lower, with a peak of 13 343 mm/min at 45 min compared with 19 192 mm/min when CD81 was expressed in the NAcc. However, the decreases in locomotor activity in the presence of the Lenti-CD81-shRNAs were of the same proportion under all conditions.
Figure 6(c and d) summarizes the changes observed under these conditions at peak activity (40 min after drug injection). As shown, under all conditions statistically significant differences (p < 0.08) were observed between the second (corresponding to silencing endogenous CD81) and the fourth groups, i.e. where endogenous CD81 was not silenced in these brain areas. Due to the low levels of endogenous CD81, the differences are small but nevertheless significant. Together these data show that lentivirus-mediated gene silencing of CD81 expression in the mesolimbic dopaminergic pathway resulted in suppression of CD81-induced locomotor activity.
Expression of CD81 was assessed by immunohistochemistry at the end of each session, as shown in Fig. 7. Clearly, lentivirus-mediated gene delivery with Lenti-CD81 results in very local gene expression in the NAcc or VTA (Figs 7a and c, top panels). This expression is supressed with doxycycline (Figs 7b and d, top panels), as expected. In addition, when the CD81 silencer viruses, Lenti-CD81-shRNAs, are coinjected with Lenti-CD81 (Figs 7a–d, lower panels), full gene knock-down also results in suppression of protein expression in the infected brain area. These controls further establish the good functionality of the tools developed in this study.
For the present study we developed lentiviral-mediated delivery of shRNA in vivo and demonstrated the efficacy of this approach in locally reducing target CD81 expression in vivo. The method has been validated by several means, including in vitro assays and evaluation of behavioural changes in vivo. The ability of lentiviral vectors to transduce cells efficiently in vivo coupled with the efficacy of virally expressed shRNA shown here extend the application of shRNA to viral-based therapies and in vivo targeting experiments that aim to define the function of specific genes.
In vivo delivery of shRNAs in the brain
The efficiency of shRNA in silencing gene expression in vivo centres around four primary challenges responsible for the downfall of the approach: in vivo delivery of shRNA, choice of the appropriate target, stability of the shRNA and long-term expression of the shRNA (Shi 2004). So far, very few studies have reported the use of shRNA in vivo because of these limitations. Its efficacy has been demonstrated in vivo in inhibition of expression of some disease-related targets: the vascular endothelial growth factor (Filleur et al. 2003; Usman 2004), caspase 8 (Zender et al. 2003), hepatitis B (McCaffrey et al. 2003) and C viruses (Yokota et al. 2003; Usman 2004), β-catenin (Verma et al. 2003), tumour necrosis factor-α (Sorensen et al. 2003), agouti-related peptide (Makimura et al. 2002), β-glucuronidase (Davidson and Paulson 2004), the receptor tyrosine kinase muscle-specific receptor kinase of neuromuscular junction (Kong et al. 2003) and the endogenous Fas gene expressed in adult mouse liver (Song et al. 2003). Delivery has been cited as one of the most important barriers to RNA-based therapy (Davidson and Paulson 2004). The effectiveness of siRNA is often lost when siRNAs are injected into animals, due to rapid clearance and degradation (Usman 2004). Recently, several breakthroughs have highlighted viruses as excellent vehicles for siRNA delivery (Zhang et al. 2003; Devroe and Silver 2004). Retroviruses, the transgene delivery vector of choice for many experimental gene therapy studies, have been engineered to deliver and stably express therapeutic siRNA within cells both in vitro and in vivo (Rubinson et al. 2003; Jain 2004). Delivery of siRNAs into the CNS adds an additional challenge. RNAi targeting neurological disease genes can be accomplished in vitro, with profound effects on cellular phenotypes, but the difficult challenge lies in the translation of this technology to the CNS, primarily the problem of delivery to the brain. Therefore, there has been no study so far relating the efficiency of siRNA in treating pathophysiological disorders or mood disorders in vivo and studies so far have been limited to target evaluation in vitro (Beckman et al. 2003; Kong et al. 2003). Recently, another study also successfully demonstrated, for the first time, the use of this technology in assessing pathophysiological conditions in the CNS, by intrathecal administration of shRNAs to down-regulate P2X3 gene expression in vivo, and evaluation of the roles of putative pain genes in pain models (Dorn et al. 2004).
Our study was designed to overcome two additional difficulties, on the one hand, we aimed at silencing a gene very locally in the CNS to prevent effects in unrelated areas and, on the other hand, long-term expression of the shRNA was desired to assess behavioural changes over several weeks. In this respect, delivery of retroviral self-inactivating lentiviruses is ideal as they transfect post-mitotic cells, including neurones, with permanent incorporation of the elements encoding for shRNA into the host genome and will not spread away from the injection site, thus yielding a very localized effect. This enabled the use of shRNA in combination with behavioural evaluation to assess changes induced upon gene expression in vivo.
Another challenge in the methodology lies in the design of shRNA. It is crucial that siRNA must not cause any effects in vivo other than those related to the knock-down of the target gene. This issue is particularly important in therapeutic applications where unwanted side-effects would be undesirable (Shi 2004). Although the actual substrate specificity of individual siRNAs appears to be very high (Brummelkamp et al. 2002), recent studies have indicated that siRNAs can tolerate single mutations located in the centre of the molecule and up to four mutations are necessary for complete inactivation (Holen et al. 2002; Jacque et al. 2002). Thus, some mismatches can be tolerated, in particular non-canonical base pairs that are frequently found in double-stranded microRNA precursors (Mourelatos et al. 2002). In the present study, we found, however, that a single mutation in shRNA greatly affects its activity, as shown by the efficacy of shRNA3 compared with shRNA1. A further complication in the method has been pointed out from global gene expression using microarray technology to examine the specificity of siRNAs, with the interesting finding that a large number of non-targeted genes containing as few as 11 continuous nucleotides with identity to the siRNA may be affected by siRNA treatment (Jackson et al. 2003). Using the same approach, the overall cellular effects of siRNAs on transcription levels have been investigated (Semizarov et al. 2003). Similar assays of validation remain to be performed on our targets.
Application of lentiviral-based shRNA delivery in vivo to assess the putative role of CD81 in addiction
In this study we also confirmed earlier observations establishing that the tetraspanin CD81 is induced in the mesolimbic dopaminergic pathway after cocaine administration. CD81 expression causes behavioural changes associated with drug intake which can be reversed by means of local gene knock-down. Applying our strategy of lentiviral-based delivery of shRNA in vivo to reduce CD81 expression in the mesolimbic dopaminergic pathway, we were able to demonstrate that this effect significantly reduces drug-induced locomotor activity in the rat. Stereotaxic injection of a regulatable Lenti-CD81 into the NAcc or VTA, resulting in CD81 overexpression in these regions, induced a four- to fivefold increase in locomotor activity after chronic cocaine administration which returned to basal activity when Lenti-CD81-shRNA had been coinjected or when CD81 expression was blocked by doxycycline. Furthermore, in vivo delivery of Lenti-CD81-shRNA into these areas resulted in > 90% silencing of the endogenous CD81 which, under these conditions, would be induced by drug administration, yielding a significant decrease in locomotor activity as compared with controls.
The role of the tetraspanin CD81 in drug addiction is not surprising. Proteins of the tetraspanin superfamily are scaffold proteins which participate in the formation of plasma membrane-signalling complexes and regulate signal transduction and association with extracellular matrix. Growing evidence indicates that, in the nervous system, they enable synaptic adaptations and plasticity, which are key processes induced by the drug. Recent evidence implicates neuronal tetraspanins in axon growth and target recognition (Fradkin et al. 2002). CD81 regulates neurone-induced astrocyte cell cycle exit (Kelic et al. 2001) and plays a central role in the dynamic regulation of complex formation between specific G-coupled protein receptors and their G-proteins (Little et al. 2004). CD9, a major molecular partner of CD81 (Charrin et al. 2001), forms extensive complexes with other tetraspanins, integrins and other proteins (Stipp et al. 2001), building the tetraspanin web beneath the plasma membranes (Terada et al. 2002). CD9 is a paranodal component regulating paranodal junctional formation in myelinated axons (Ishibashi et al. 2004) and is localized to neurones early in development, coincident with a period of active axon growth for many neuronal populations (Schmidt et al. 1996). Tetraspanins form complexes with integrins, mainly β1 integrins (Hemler 1998); these integrins are known to mediate axon growth induced by numerous extracellular matrix proteins. Enhanced integrin-mediated neurite outgrowth in response to CD9 activation suggests a functional relationship between integrins and tetraspanins (Banerjee et al. 1997). Furthermore, integrin association with extracellular matrix proteins, e.g. laminin-5, is regulated by EWI-2 (Stipp et al. 2003), a major binding partner of CD81 (Stipp et al. 2001). Recent evidence suggests that tetraspanins may serve as a link between integrin subunits and various intracellular signalling molecules, such as phosphatidylinositol 4-kinase and protein kinase C (Berditchevski et al. 1997; Hemler 1998). CD81 is also linked to extracellular signal-regulated kinase/MAPK signalling (Carloni et al. 2004). Tetraspanin activation can also result in increases in intracellular calcium and enhanced tyrosine kinase and phosphatase activity (Carmo and Wright 1995), each of which has been implicated in the regulation of neurite outgrowth (Keynes and Cook 1995; Desai et al. 1997).
As drugs of abuse induce strong plasticity and molecular adaptations which underlie a complex rewiring of neural circuitry and result in the behaviours associated with addiction, it is not surprising that a scaffolding protein such as CD81 may play a role in these processes, as clearly shown in this study. Further studies will be devoted to clarifying its function.
Supported by a Swiss National Foundation grant 3100-059350 and 3100AO-100686 (J-LD). The authors are very grateful to Dr G. Wagner for critical reading of the manuscript and discussions and Mrs C. Deforel-Poncet and Mr H.-H. Wang for skilful technical assistance.