The blood-brain barrier (BBB) contains tight junctions (TJs) which reduce the space between adjacent endothelial cells lining the fine capillaries of the microvasculature of the brain to form a selective and regulatable barrier.
The blood-brain barrier (BBB) contains tight junctions (TJs) which reduce the space between adjacent endothelial cells lining the fine capillaries of the microvasculature of the brain to form a selective and regulatable barrier.
Using a hydrodynamic approach, we delivered siRNA targeting the TJ protein claudin-5 to the endothelial cells of the BBB in mice.
We have shown a significant decrease in claudin-5 mRNA levels 24 and 48 hours post-delivery of siRNA, with levels of protein expression decreasing up to 48 hours post-injection compared to uninjected, phosphate-buffered saline (PBS)-injected and non-targeting siRNA-injected mice. We observed increased permeability at the BBB to molecules up to 742 Da, but not 4400 Da, using tracer molecule perfusion and MRI analysis. To illustrate the functional efficacy of size-selective and transient barrier opening, we have shown that enhanced delivery of the small neuropeptide thyrotropin-releasing hormone (TRH) (MW 360 Da) to the brains of mice 48 hours post-injection of siRNA targeting claudin-5 significantly modifies behavioural output.
These data demonstrate that it is now possible to transiently and size-selectively open the BBB in mice, allowing in principle the delivery of a wide range of agents for the establishment and treatment of experimental mouse models of neurodegenerative, neuropsychiatric and malignant diseases. Copyright © 2008 John Wiley & Sons, Ltd.
Many attempts have been made either to break the blood-brain barrier (BBB) or to design delivery systems that enable pharmacological agents to traverse the endothelial cells of brain capillaries 1. However, such attempts have been made without due consideration given to the structure of the tight junctions (TJs), which reduce the space between the plasma membranes of contacting endothelial cells to form a selective and regulatable barrier 2. If transient, reversible opening of the BBB could be achieved, an avenue would be available for experimental delivery to the brain in animals such as mice, of agents which may modulate neuronal function, both in normal animals or in those displaying neurodegenerative or behavioural features.
Recently, genetically engineered proteins termed ‘molecular Trojan horses’ have been described and cross the BBB via endogenous receptor-mediated transport processes 3. In 2007, using a modified, yet similar approach, targeted delivery of proteins across the BBB was reported using a lentiviral vector system, exploiting the binding domain of apolipoprotein B to its receptor ‘low-density lipoprotein receptor’ (LDLR). This report proved feasible for the delivery of proteins via the transcellular pathway, yet these approaches have yet to address peptide cleavage from the engineered binding sites upon delivery to the central nervous system (CNS) 4. It has also been recently reported that it is now possible to deliver small interfering RNA (siRNA) molecules across the BBB in order to modulate gene function in neurons in vivo. Kumar and colleagues have shown that a short peptide derived from the rabies virus glycoprotein (RVG) enables transvascular delivery of siRNA to the brain. The 29-amino-acid peptide described specifically binds to the acetylcholine receptor expressed by BBB endothelial cells and neuronal cells, and, by modifying the carboxy terminus of the peptide, they were able to conjugate siRNAs to the RVG, allowing for delivery across the BBB and subsequent transduction of neuronal cells 5.
While delivery of molecules across the transcellular pathway of the BBB remains an exciting avenue for further research, the alternative strategy of inducing the paracellular pathway to reversibly open may represent a useful and highly versatile alternative for studies of drug-induced modulation of phenotype in experimental animal models of neurodegenerative, psychiatric or malignant disease and possibly clinically relevant therapeutic strategies. In general, transport of components across endothelial cells of the BBB can occur via three routes: a transcellular route, which may be mediated via special transporters as alluded to above, vesicular transport, or a paracellular route which allows for transport between neighbouring endothelial cells 6–8. Brain capillaries exhibit very low rates of fluid phase transcytosis, and the paracellular route between individual cells at the BBB is sealed by TJs that are considerably tighter than in any other microvessels in the body. Therefore, TJs represent a key factor relating to the low permeability properties associated with the BBB 9.
The TJs associated with the BBB are composed of a complex of intracellular and transmembrane proteins including occludin, junctional adhesion molecule (JAM), claudins-1, -5, -12 and ZO-1, -2 and -3 10, 11. Approximately 20 members of the claudin family have also been described, claudin-1, -5 and -12 predominating in TJs of the BBB 12. Claudins, like occludin, span endothelial cell membranes four times and interact with ZO-1 via their C-terminus 13.
Claudin-5 is considered to be endothelial-cell-specific 14. Claudin 5-/- mice have been reported and the BBB is compromised in these animals. Through a series of tracer molecule experiments and magnetic resonance imaging (MRI), the authors conclude that while removal of claudin-5 compromises the function of the BBB by allowing it to become permeable to molecules of up to approximately 800 Da, the barrier can still form. Furthermore, the authors showed that the barrier remains intact and impervious to larger molecules, showing no evidence of bleeding or oedema. However, these animals die within days of birth 15. Recently, it has also emerged that mice exposed to hypoxic conditions exhibit decreased claudin-5 expression, which selectively disrupted the blood-retinal barrier (BRB) allowing passage of small molecules, similar to the phenotype seen in claudin-5-deficient mice 16. It is thought that claudin-5 may play a role in the formation of paracellular pores or channels that function in mediating selective ion permeability 17, 18. Therefore, it is likely that in the claudin-5 knockout mouse a mechanism may be activated that allows for increases in size-selective paracellular diffusion across the BBB.
Systemic hydrodynamic (high-volume) delivery of siRNA molecules or plasmids via the tail vein has been used by many research teams, including our own, to secure efficient ablation of target transcripts in organs such as the liver in mice 19–23. In a recent paper, Hino et al. have extended such observations by using the same approach to deliver unprotected siRNA molecules to brain capillary endothelial cells in mice. Transcripts derived from the organic anion transporter-3 (OAT3) gene were targeted, ablation of which significantly compromised brain efflux function 24. It is therefore evident that siRNA molecules can reach the endothelial cells of microvessels in the brain using the hydrodynamic inoculation approach.
Here, we report successful in vivo suppression of claudin-5 gene expression at the BBB of C57/BL-6 mice, effected using hydrodynamic tail vein delivery of siRNA targeting claudin-5 at 24 and 48 hours post-injection of claudin-5 siRNA, with levels of expression returning 72 hours post-injection. We also show the permeation of small molecules (molecular weight (MW) 443 and 562 Da) from the microvasculature of mice following delivery of claudin-5 siRNA to the brain endothelial cells, with extravasation of one of these molecules also evident in retinal cryosections showing a compromise also at the BRB. This increase in permeability in brain microvessels was not evident however when we perfused a molecule with a MW of 4400 Da.
The suppression of claudin-5 occurred concomitant with aberrant localisation and expression of this TJ protein in the brain microvasculature. Moreover, we have performed MRI scans on mice following suppression of claudin-5, and following injection of the contrasting agent gadolinium diethylenetriamine N,N,N′,N′,N′-pentaacetic acid (Gd-DTPA; MW 742 Da), we have observed widespread permeation of this agent 24 and 48 hours post-injection of claudin-5 siRNA, with no permeation observed 72 hours or 1 week post-injection of claudin-5 siRNA.
Much effort has been directed toward understanding the TJs of brain capillary endothelial cells in order to identify molecular mechanisms that could be manipulated to enhance drug delivery to the brain, since many drugs are ineffective because they are unable to cross the BBB.
Here, we describe the first report of reversible and controlled RNAi-mediated size-selective opening of the paracellular pathway of the BBB and possibly the BRB, representing a novel approach for delivery of a wide range of small negatively charged molecules to the brain. Indeed, we have shown that the increased permeability of the BBB will allow for the enhanced delivery of a small negatively charged neuropeptide (thyrotropin-releasing hormone (TRH)) to the brain, manifested by a distinct change in behavioural output in mice.
This method of reversible BBB modulation may pave the way for controlled delivery of therapeutic agents to the CNS in a range of behavioural or brain tumour models or of other conditions that currently offer little or no prospect of effective treatment.
All experiments involving the use of C57/BL6 mice were assessed and approved by an internal ethics committee in Trinity College Dublin (TCD) prior to all experimentation. All studies carried out in the Ocular Genetics Unit in TCD adhere to the ARVO statement for the use of Animals in Ophthalmic and Vision Research. C57/BL6 mice were sourced from Jackson Laboratories and bred on-site at the Ocular Genetics Unit in TCD.
In order to minimise complications which may arise owing to possible alternative splicing of target transcripts which could lead to some species escaping RNAi by virtue of the lack of the target sequence, siRNAs were selected targeting conserved regions of the published cDNA sequences. To do this, cDNA sequences from mouse were aligned for the claudin-5 gene and regions of perfect homology subjected to updated web-based protocols (Dharmacon, Ambion, Genescript) originally derived from criteria as outlined by Reynolds et al.32. Initially, four siRNAs targeting claudin-5 were tested and the most efficient used in further studies. In our experience 23, 33 with four different siRNAs targeting the same transcript, at least three will down-regulate by 60%, one of which will give near to 80% suppression. Sequences of the claudin-5 siRNA used in this study were as follows. Sense sequence: CGUUGGAAAUUCUGGGUCUUU. Antisense sequence: AGACCCAGAAUUUCCAACGUU. Non-targeting control siRNA targeting human rhodopsin was used as a non-targeting control since rhodopsin is only expressed in photoreceptor cells in the retina and at low levels in the pineal gland of the brain. Sense sequence: CGCUCAAGCCGGAGGUCAAUU. Antisense sequence: UUGACCUCCGGCUUGAGCGUU 34.
Rapid high-pressure, high-volume tail vein injections were carried out essentially as previously successfully used at this laboratory 23. Wild-type C57/Bl6 mice of weight 20–30 g were individually restrained inside a 60-ml volume plastic tube. The protruding tail was warmed for 5 min prior to injection under a 60-W lamp and the tail vein clearly visualised by illumination from below. Twenty micrograms of targeting siRNA, or non-targeting siRNA made up with phosphate-buffered saline (PBS) to a volume in ml of 10% of the body weight in grams or PBS alone was injected into the tail vein at a rate of 1 ml/s using a 26-gauge (26G 3/8) needle. After 24, 48, 72 hours and 1 week, protein was isolated from total brain tissue by crushing brains to a fine powder in liquid N2 and subsequently homogenising in lysis buffer containing 62.5 mM Tris, 2% SDS, 10 mM dithiothreitol, 10 µl protease inhibitor cocktail/100 ml (Sigma Aldrich, Ireland). The homogenate was centrifuged at 10 000 g for 20 min at 4 °C, and the supernatant was removed for claudin-5 analysis.
Briefly, protein samples were separated by sodium dodecyl sulphate/polyacrylamide gel electrophoresis (SDS-PAGE) on 12% gels and transferred to a nitrocellulose membrane overnight using a wet electroblot apparatus. Efficiency of protein transfer was determined using Ponceau-S solution (Sigma Aldrich, Ireland). Non-specific binding sites were blocked by incubating the membrane at room temperature with 5% non-fat dry skimmed milk in Tris-buffered saline (TBS) (0.05 M Tris, 150 mM NaCl, pH 7.5) for 2 hours. Membranes were briefly washed with TBS, and incubated with polyclonal rabbit anti-claudin-5 (Zymed Laboratories, San Francisco, CA, USA) (1 : 500), polyclonal rabbit anti-claudin-1 (Zymed Laboratories) (1 : 500), polyclonal rabbit anti-occludin (Zymed Laboratories) (1 : 500), monoclonal mouse anti-Tie-2 (Chemicon) (1 : 500), polyclonal rabbit anti-β-actin (Abcam, Cambridge, UK) (1 : 1000). Antibodies were incubated with membranes overnight at 4 °C. Membranes were washed with TBS, and incubated with a secondary anti-rabbit (IgG) or anti-mouse (IgG) antibody with horse radish peroxidase (HRP) conjugates (1 : 2500) (Sigma-Aldrich, Ireland) (A-6154), for 3 hours at room temperature. Immune complexes were detected using enhanced chemiluminescence (ECL). All Western blots were repeated a minimum of three times (i.e. ‘n = 3 mice per treatment’).
At the same time points post-delivery of siRNA total RNA was isolated from brains using Trizol (Invitrogen). RNA was then treated with RNase-free DNase (Promega, Madison, WI, USA) and then chloroform extracted, isopropanol precipitated, washed with 75% RNA grade ethanol and re-suspended in 100 µl RNase-free water.
RNA was analysed by real-time RT-PCR using a Quantitect Sybr Green Kit as outlined by the manufacturer (Qiagen–Xeragon) on a LightCycler (Roche Diagnostics, Lewes, UK) under the following conditions: 50 °C for 20 min; 95 °C for 15 min; 38 cycles of 94 °C for 15 s, 57 °C for 20 s, 72 °C for 10 s. Primers (Sigma–Genosys, Cambridge, UK) for the sequences amplified were as follows CLDN5 (forward 5′-TTTCTTCTATGCGCAGTTGG-3′, reverse 5′-GCAGTTTGGTGCCTACTTCA-3′), β-actin (forward 5′-TCACCCACACTGTGCCCATCTA-3′, reverse 5′-CAGCGGAACCGCTCATTGCCA-3′). cDNA fragments were amplified from claudin-5 and β-actin for each RNA sample a minimum of four times. Results were expressed as a percentage of those from the similarly standardised appropriate control experiment. The reciprocal values compared to the non-targeting control siRNA gave percentage suppression of claudin-5 mRNA levels. Mean values, standard deviations, and analysis of variance analyses (ANOVA) with a Tukey-Kramer post-test were calculated using GraphPad Prism©. Differences were deemed statistically significant at P < 0.05. (All RT-PCR analyses were performed using ‘n-values’ ranging from 3–5 mice.)
Brain cryosections were blocked with 5% normal goat serum (NGS) in PBS for 20 min at room temperature. Primary antibodies; rabbit anti-claudin-5 (Zymed), rabbit anti-occludin (Zymed), rat anti-claudin-1 (RnD Systems, Abingdon, UK), were incubated on sections overnight at 4 °C. Following incubation, sections were washed three times in PBS and subsequently blocked again with 5% NGS for 20 min at room temperature. Secondary rabbit IgG-Cy3, rabbit IgG-Cy2 and rat IgG-Cy3 (Jackson-Immuno Research, Europe) or rabbit IgG-Alexa 568 (Molecular Probes, UK) antibodies were incubated with the sections at 37 °C for 2 hours followed by three washes with PBS. All sections were counterstained with 4′,6-diamidine-2-phenylindole dihydrochloride (DAPI, Sigma Aldrich, Ireland) for 30 s at a dilution of 1 : 5000 of a stock 1 mg/ml solution. Analysis of stained sections was performed at room temperature with an Olympus FluoView TM FV1000 confocal microscope with integrated software.
Following RNAi-mediated ablation of transcripts encoding claudin-5, the biotinylated reagent EZ-Link TM Sulfo-NHS-Biotin (Pierce) (1 ml/g body weight of 2 mg/ml EZ-Link TM Sulfo-NHS-Biotin, 443 Da) was perfused at 37 °C for 5 min through the left ventricle of the beating heart of anaesthetised mice 24, 48, 72 hours and 1 week post-hydrodynamic delivery of claudin-5 siRNA. Following perfusion with the tracer molecule, the whole brain was dissected and placed in 4% PFA (pH 7.4) overnight at 4 °C and subsequently washed four times for 15 min with PBS. Following cryoprotection using a 10%, 20% and 30% sucrose gradient, 12 µm frozen sections were cut on a cryostat at − 20 °C and incubated with streptavidin conjugated to the fluorescent probe FITC. This allowed for the assessment of leakage of the biotinylated reagent of MW 443 Da from the microvessels of the brain. All sections were counterstained with DAPI for 30 s at a dilution of 1 : 5000 of a stock 1 mg/ml solution, and sections were visualised using an Olympus FluoView TM FV1000 confocal microscope.
In order to determine the permeability of brain and retinal microvessels to a molecule of 562 Da, mice were perfused through the left ventricle of the beating heart and at 37 °C with 500 µl/g body weight of PBS containing 100 µg/ml Hoechst stain H33342 (Sigma Aldrich, Ireland) and 1 mg/ml FITC-dextran-4 (FD-4) 24, 48, 72 hours and 1 week post-hydrodynamic delivery of claudin-5 siRNA. Following perfusion, the whole brain was dissected and placed in 4% PFA (pH 7.4) overnight at 4 °C and subsequently washed four times for 15 min with PBS. Brains were then embedded in 4% agarose and 50 µm sections were cut using a Vibratome®. Whole eyes were removed and fixed with 4% PFA, and, following washing with PBS and cryoprotection using a sucrose gradient, 12 µm cryosections were cut using a cryostat. Following analysis of retinal cryosections with an Olympus FluoView TM FV1000 confocal microscope, images were oriented correctly using Adobe® Photoshop®. Permeability assays were performed up to five times for each experimental group. All separate time points were analysed by confocal microscopy on the same day, using constant settings to allow for comparison of tracer perfusions at each time point.
Following injection of siRNA and using appropriate controls, BBB integrity to a molecule of 742 Da was assessed via MRI, using a dedicated small rodent Bruker BioSpec 70/30 (i.e. 7 T, 30 cm bore) with an actively shielded USR magnet. Mice were anaesthetised with isofluorane, and physiologically monitored (ECG, respiration and temperature) and placed on an MRI-compatible support cradle, which has a built-in system for maintaining the animal's body temperature at 37 °C. The cradle was then positioned within the MRI scanner. Accurate positioning is ensured by acquiring an initial rapid pilot image, which is then used to ensure the correct geometry is scanned in all subsequent MRI experiments. Upon insertion into the MRI scanner, high-resolution anatomical images of the brain were acquired (100 µm in-plane and 500 µm through-plane spatial resolution). BBB integrity was then visualised in high-resolution T1-weighted MR images before and after injection of a 0.1 mM/l/kg bolus of Gd-DTPA, administered via the tail vein. Following injection of Gd-DTPA, repeated 3-min T1-weighted scans were performed over a period of 30 min, and images shown are representative of the final scans of this 30-min period. Statistical analysis of all densitometric results of combined regions of the cerebellum, hippocampus and cortex was performed using ANOVA, with significance represented by a P value of ≤0.05, and results are presented both graphically and in a quantitative image depicting the rate of Gd-DTPA deposition within the brain. All MRI scans were performed on two mice from each experimental treatment.
TRH was purchased from Sigma Aldrich, Ireland, and was received in lyophilised form. The neuropeptide was reconstituted with sterile PBS, and, 48 hours post-injection of mice with non-targeting siRNA or siRNA targeting claudin-5, mice were administered a tail vein injection of a solution containing 20 mg/kg body weight TRH. Following administration of TRH, mice were placed in a clear Perspex box and filmed using a Sony® DCR-PC8E/PAL digital video camera for up to 10 min each. Immobility of mice was assessed as the inability of the animal to move in a coordinated manner within its surrounds. The length of time of immobility of mice was determined and plotted on a bar chart comparing mice injected with a non-targeting siRNA 48 hours prior to TRH injection to mice injected with siRNA targeting claudin-5 48 hours prior to TRH injection. Statistical analysis was performed using Student's t-test, with significance represented by a P value of ≤0.05 (n = 5 mice per treatment).
The hydrodynamic approach for delivery of siRNAs to endothelial cells of the brain microvasculature appeared to be highly efficient in suppressing claudin-5 expression (Figures 1A and 1B). This method of delivery caused no apparent harm and was well tolerated in mice. Our Western data showed that we achieved maximum suppression of claudin-5 48 hours after delivery of the siRNA, with levels of expression of claudin-5 returning to normal between 72 hours and 1 week after injection when compared to the corresponding β-actin levels in the same lane. Levels of claudin-5 mRNA were determined by RT-PCR analysis and showed a significant (*P = 0.0427) decrease 24 and 48 hours post-injection (*P = 0.0478) of siRNA targeting claudin-5 compared to the control groups (up to almost 80% with respect to the control groups at the 24 hour time point). Levels of claudin-5 mRNA, 72 hours (P = 0.0627) and 1 week (P = 0.2264) post-injection were not significantly changed, as statistical analyses revealed P-values > 0.05, representing insignificant changes. ‘n-values’ ranged from 3–5 mice per treatment (Figure 1B).
In endothelial cells of the brain microvasculature, levels of claudin-5 expression appeared strong and continuous in the microvasculature upon immunohistochemical analysis of all the control groups employed (Red = Claudin-5; Blue-DAPI = nuclei). However, when claudin-5 was targeted, the continuous appearance of expression became discontinuous and fragmented, with levels appearing dramatically reduced 48 hours after injection of claudin-5 siRNA (Figure 2). The pattern of staining observed was similar to that previously reported and reflected the decreased levels of claudin-5 protein observed in the Western data of Figure 1A 15. An enlarged image of claudin-5 staining 48 hours post-injection of claudin-5 siRNA clearly shows the fragmented pattern of staining pertaining to decreased levels of claudin-5.
In order to determine levels of expression of other TJ-associated proteins following suppression of claudin-5, we analysed the levels of expression of the transmembrane proteins claudin-1 and occludin. Moreover, in order to ascertain levels of an endothelial-cell-specific protein, we analysed levels of expression of the receptor tyrosine kinase Tie-2. Levels of expression of claudin-1 which shares almost 80% homology with claudin-5 appeared to remain unchanged following all experimental treatments. Normalisation was completed by analysing levels of expression of β-actin and Tie-1 (Figure 3A). Indeed, levels of expression of the four-transmembrane TJ-associated protein occludin were also shown to remain unchanged following injection of siRNA targeting claudin-5 at all time points post-siRNA injection (Figure 3B).
In order to ascertain any potential off-target effects on claudin-1 expression and localisation following injection of siRNA targeting claudin-5, we analysed claudin-1 and claudin-5 expression simultaneously on brain cryosections. It was evident that as the dramatic decrease in claudin-5 expression (Green staining) and localisation previously observed 48 hours post-siRNA injection occurred, claudin-1 expression (Red staining) remained similar to the control groups (Figure 4B). Levels of expression of claudin-5 at 1 week post-injection resemble staining observed in control groups, while claudin-1 expression remains constant throughout all time points (Figures 4A–4D).
Following perfusion of mice with the biotinylated molecule EZ-Link TM Sulfo-NHS-Biotin for 5 min, we observed a significant compromise in barrier function up to and including 72 hours post-delivery of siRNA targeting claudin-5. EZ-Link TM Sulfo-NHS-Biotin has a MW of 443 Da, and will normally not cross the BBB if the TJs are intact as observed in the control groups. Interestingly, 1 week after delivery of claudin-5 siRNA, this molecule no longer crossed the BBB (Figure 5), suggesting that, consistent with real-time PCR and Western analyses, the compromise in BBB function is a transient and reversible process. Analysis of brain cryosections was in the hippocampus region (for ease of recognition), yet increases in permeability were evident throughout the brain.
Upon perfusion of the nuclear stain Hoechst H33342 (562 Da) and the FITC-labelled dextran, FD-4 (4400 Da), extravasation of Hoechst was observed up to and including 48 hours post-delivery of siRNA targeting claudin-5; however, unlike EZ-Link TM Sulfo-NHS-Biotin, extravasation was not evident 72 hours post-siRNA delivery, suggesting a restoration of barrier integrity to a molecule of 562 Da, and implying a time-dependent and size-selective opening of the BBB. Hoechst H33342 dye extravasation from the brain microvessels was manifested by nuclear staining of surrounding neural and glial cells in the parenchyma. FD-4 remained within the microvessels of the brain vasculature and no extravasation was evident at any time point post-injection of siRNA (Figure 6A).
Moreover, upon analysis of retinal cryosections, we observed that Hoechst H33342 extravasated from the retinal microvessels, staining the inner nuclear layer (INL) and outer nuclear layer (ONL) of the retina up to 48 hour post-delivery of siRNA targeting claudin-5 (Figure 6B).
Using the MRI contrasting agent Gd-DTPA (742 Da), we observed extremely large quantities of Gd-DTPA deposited in the brain 24 and 48 hours post-delivery of claudin-5 siRNA. This BBB breakage to a molecule of 742 Da was a transient event, as 72 hours and 1 week post-injection of siRNA targeting claudin-5 there appeared to be no infiltration of Gd-DTPA (Figure 7A). It was also apparent from the MR images that a large amount of contrasting agent was present in the eye, at 48 hours post-delivery of siRNA targeting claudin-5, further suggesting that concomitant to BBB compromise, we also observed a distinct and transient increase in permeability at the BRB (Figure 7A).
There was a significant increase in contrasting within selected regions of the cerebellum, hippocampus and cortex of the brain at 24 hours (**P < 0.05) and 48 hours (**P < 0.05) post-injection of claudin-5 siRNA when compared to a non-targeting control siRNA as revealed by densitometric analysis of these specific areas (Figure 7B).
The quantitative image in Figure 7C is a representation of the slope of the linear fit, determined for every pixel in the MRI scans of mice receiving a non-targeting control siRNA compared to mice 24 and 48 hours after injection of siRNA targeting claudin-5 (i.e., the time points at which increased contrasting of Gd-DTPA is observed within the brain). The ventricular region is a mean of 602 (out of just over 16 000) pixels. The red end denotes very little change, with green areas showing some change and blue denoting a large change. In terms of the slopes involved, the units are arbitrary, but the most intense (bluest) areas on the graph had a slope of d(y)/dx = d(ln(intensity))/dt (time in seconds) = 0.1, and scaling down to approximately 0.0001 for red.
The graph below the quantitative image in Figure 7C shows the change in intensities in the left ventricle over a 28-min time course after Gd-DTPA injection. The data are plotted as the natural logarithm (ln) of the signal intensity (y-axis) against time in minutes on the x-axis (each unit on the x-axis is 128 s long). The red line represents the non-targeting control siRNA-injected mouse; the yellow line represents the 24-hour time point post-injection of siRNA targeting claudin-5; while the green line represents the 48-hour time point post-injection of claudin-5 siRNA. It is clear that there is an increased rate of transport of contrasting agent into the brains of mice in the selected region of the brain around the left ventricle at the 24 and 48 hour time points post-injection of siRNA targeting claudin-5.
Forty-eight hours post-delivery of siRNA targeting claudin-5 or a non-targeting siRNA, we injected 200 µl of a solution containing 20 mg/kg TRH (360 Da). Following suppression of claudin-5 protein expression, we observed a distinct and statistically significant increase in the length of time C57/Bl6 mice remained immobile after systemic injection of 20 mg/kg TRH. This behavioural output was significantly different from the behaviour observed in the non-targeting control mice (**P = 0.0041), and clearly suggested that delivery of TRH was significantly enhanced when the BBB was compromised (Figure 8).
The microvasculature plays an essential role in supplying the high-energy-demanding brain with oxygen-enriched blood. The endothelial cells that line these fine capillaries have evolved ‘tight junctions’ (TJs), which form a selective and regulatable barrier. However, oxygen can still diffuse from these cells, and other essential materials can be delivered to the brain by special transporters located in the membranes of the endothelial cells. Complete breakdown of the BBB would have disastrous consequences for overall brain function. However, if transient, reversible opening of the barrier could be achieved, an avenue would be available for experimental delivery to the brain, in animals such as mice, of agents which may modulate neuronal function. This enhanced delivery could also potentially be used both in normal animals and in those displaying neurodegenerative, neuropsychiatric disorders or indeed those presenting with brain injury or trauma. The human brain contains approximately 100 million capillaries comprising a total surface area of about 12 m225. Nearly every neuron in the brain has its own capillary, with an average distance from capillary to neuron of 8–20 µm 26. Therefore, delivery of a neuromodulatory agent to neurons across the capillary membrane would represent an immensely powerful experimental tool. Since evidence is now available to indicate that siRNA molecules can access the endothelial cells of brain capillaries in mice 24, we hypothesised that such molecules could, in principle, be directed toward controlled down-regulation of transcripts encoding TJ proteins associated with the endothelial cells of the brain microvasculature. Our target TJ protein was claudin-5, a transmembrane protein implicated in the maintenance of BBB integrity with regard to the passage of molecules across the paracellular pathway. Indeed, the claudin-5 knockout mouse displayed size-selective loosening of the BBB without affecting overall TJ morphology or normal brain histology, although these animals died neonatally 15.
The hydrodynamic approach for delivery of siRNAs to endothelial cells of the brain microvasculature appeared to be highly efficient in suppressing claudin-5 expression (Figures 1A and 1B). This method of delivery caused no apparent harm and was well tolerated in mice. Our Western data showed that we achieved maximum suppression of claudin-5 48 hours after delivery of the siRNA, with levels of expression of claudin-5 returning to normal between 72 hours and 1 week after injection.
Following these initial observations, we endeavoured to determine if, similar to the claudin-5 knockout mouse, the BBB became compromised to small molecules when claudin-5 expression was suppressed. When claudin-5 was targeted, its microvessel-associated expression became discontinuous and fragmented, with levels appearing dramatically reduced 48 hours after injection of claudin-5 siRNA (Figure 2). In order to determine whether RNAi-mediated suppression of claudin-5 elicited changes in expression of other TJ-associated proteins, we analysed levels of expression of the transmembrane TJ-associated proteins claudin-1 and occludin. Both of these TJ proteins have been implicated in mediating changes in the paracellular permeability of TJs, and it appeared that their levels of expression and localisation remained largely unchanged following injection in mice of siRNA targeting claudin-5 (Figures 3A, 3B, and 4A–4D). These data highlight the specificity of claudin-5 siRNA used in this approach. Immunohistochemical analysis of occludin expression also showed no aberrancies at any time post-injection of siRNA targeting claudin-5 (Supplementary Figure 6, Supplementary Material).
Following perfusion of mice with the biotinylated molecule EZ-Link TM Sulfo-NHS-Biotin for 5 min, we observed a significant compromise in barrier function up to and including 72 hours post-delivery of siRNA targeting claudin-5. EZ-Link TM Sulfo-NHS-Biotin has a molecular weight of 443 Da, and will normally not cross the BBB if the TJs are intact, as observed in the control groups. Interestingly, 1 week after delivery of claudin-5 siRNA, this molecule no longer crossed the BBB (Figure 5), suggesting that consistent with real-time PCR and Western analyses, this compromise in BBB function is a transient and reversible process. Upon perfusion of the nuclear stain Hoechst H33342 which has a molecular weight of 562 Da and the FITC-labelled dextran FD-4 (4400 Da), extravasation of Hoechst was observed up to and including 48 hours post-delivery of siRNA targeting claudin-5 in both the brain and the retina; however, unlike EZ-Link TM Sulfo-NHS-Biotin, this extravasation was not evident after 72 hours post-siRNA delivery, suggesting a restoration of barrier integrity to a molecule of 562 Da, and implying a time-dependent and size-selective opening of the BBB, as FD-4, which has a molecular weight of approximately 4400 Da, does not cross the BBB at any time points following suppression of claudin-5 (Figure 6). This observation was noted throughout the brain, and lends further proof to the finding that, in tandem with being a transient event, RNAi of claudin-5 expression appears to cause a size-selective increase in BBB permeability.
Following MRI analyses, we observed extremely large quantities of Gd-DTPA deposited throughout the brain 24 and 48 hours post-delivery of claudin-5 siRNA. BBB breakage to a molecule of 742 Da was a transient event, as 72 hours and 1 week post-injection of siRNA targeting claudin-5, there appeared to be no infiltration of Gd-DTPA as observed in T1-weighted images. Interestingly, infiltration of Gd-DTPA was also evident in the eye at 24 and 48 hours post-injection of siRNA targeting claudin-5, showing a compromise in the BRB as levels of claudin-5 decrease (Figure 7A).
As siRNA was administered via the tail vein, and given the fact that claudin-5 is expressed in microvascular endothelial cells of the lung and the heart, we wished to assess whether siRNA targeting claudin-5 would adversely affect endothelial cell morphology in the liver, lung, kidney or heart. Cryosections of each of these organs were prepared at all time points following injection of siRNA targeting claudin-5 and incorporating the appropriate controls. Sections were stained with HRP-conjugated Griffonia simplicifolia-isolectin B4, which binds to intact endothelial cells, and showed that endothelial cell morphology appeared similar at all time points and in all major organs following siRNA injection when compared to the control groups (Supplementary Figures 1–4, Supplementary Material). The role of claudin-5 in organs other than the brain and eye has not been well characterised, and it is important to note that it does not appear to be fundamental in maintaining the size-selective properties of the TJs associated with these other organs 27.
It has been hypothesised that claudin-5 may play a very important role in development, since the gene is frequently deleted in velo-cardio-facial/DiGeorge syndrome patients 28. Therefore, given its documented role in BBB function, this may explain why fully developed adult mice display no obvious side effects when claudin-5 is transiently suppressed following systemic delivery of siRNA. Also, rather than ‘knocking out’ claudin-5, a transient suppression of its expression appears to cause little adverse effects in mice. It should be noted that during all time points employed in this study, we observed no distinct or noticeable behavioural changes in these mice, while the gross histology of vibratome, cryosections and high-resolution anatomical MRI scans of the brain appear normal under all experimental conditions (results not shown).
In order to determine if the BBB modulation approach could be used for the delivery of active agents to the brains of mice, we tested the efficiency of delivery of TRH. TRH is a tripeptide hormone that stimulates the release of thyroid-stimulating hormone from the anterior pituitary. TRH is produced by the hypothalamus in the brain and has a molecular weight of approximately 360 Da. It has been proposed to possess numerous neuroprotective effects 29. Concomitantly, its passage across the BBB is extremely slow, producing short-lived behavioural outputs in mice 30. Previous studies using TRH have strongly suggested that its passage across the BBB is via passive paracellular diffusion 31. As it is also highly unstable when injected systemically, the majority of TRH will be degraded before it reaches the brain. Due to its small size, negatively charged properties, and behavioural output, TRH was an ideal candidate for this particular set of experiments. Our results strongly suggest that increased quantities of TRH are passively diffusing across the BBB as the paracellular pathway has been compromised in a size-selective manner.
A striking difference that we observed in mice following injection of TRH 48 hours after delivery of siRNA targeting claudin-5 was that the animals were significantly (**P = 0.0041) immobilised for periods of up to 4.5 min when compared to mice that had received a non-targeting siRNA 48 hours prior to TRH injection (Figure 8). These data suggest that the delivery of TRH to the brain was significantly enhanced following ablation of claudin-5 at the BBB to allow for the passage of small molecules (see also supplementary data in video file format, Supplementary Figure 5, Supplementary Material).
Overall, these data clearly demonstrate that it is possible to systemically deliver siRNA molecules to the endothelial cells of the BBB. Targeted suppression of the TJ protein claudin-5 causes both a transient and size-selective increase in paracellular permeability of the barrier, which will allow for the delivery of molecules which would otherwise be excluded from the brain. Unlike the transient opening of the BBB observed with mannitol infusion, this opening of the paracellular pathway appears to be size-selective in nature. Further work will elucidate a more accurate size selectivity that is being observed and any broad transcriptional changes occurring in the brain following suppression of claudin-5 will be analysed using a ‘BBB genomics approach’. Both behavioural analyses and functional MRI studies will also allow for the identification of any characteristic changes in behaviour and neuronal function associated with opening of the BBB. Furthermore, an exact turnover rate of claudin-5 will be elucidated by incorporating a large range of time points between 24 and 72 hours post-injection of siRNA targeting claudin-5.
The potential to exploit transient compromises in BBB integrity has far-reaching implications for the development of experimental animal models of neurodegenerative and neuropsychiatric diseases, for therapeutic applications involving drug delivery/drug screening in mice with xenografted or spontaneously occurring brain tumours, and for advancing our understanding of BBB function in general.
The Ocular Genetics Unit at TCD is supported by the Wellcome Trust; Fighting Blindness Ireland; Enterprise Ireland; European Union Evi-GenoRet, LSHG-CT-2005-512036; European Union RETNET Project (MRTN-CT-2003-504003); The British RP Society; the Health Research Board of Ireland, and is a member of the Applied Neurotherapeutics Research group (ANRG, Science Foundation Ireland).
The supplementary electronic material for this paper is available in Wiley InterScience at: http://www.interscience.wiley.com/jpages/1099-498X/suppmat/.