|Home | About | Journals | Submit | Contact Us | Français|
A major question for gene therapy in brain concerns methods to administer therapeutic genes in a uniform manner over major portions of the brain. A second question in neuroimmunology concerns the extent to which monocytes migrate to the CNS in degenerative disorders. Here we show that CD11b+ cells (largely monocytes) isolated from the bone marrow of green fluorescent protein (GFP) expressing donors spontaneously home to compacted amyloid plaques in the brain. Injections of these cells as a single pulse show a rapid clearance from circulation (90 minute half-life) and tissue residence half-lives of roughly 3 days. The uptake into brain was minimal in nontransgenic mice. In transgenic mice containing amyloid deposits, uptake was dramatically increased and associated with a corresponding decrease in monocyte uptake into peripheral organs compared to nontransgenic littermates. Twice weekly infusions of the CD11b+ bone marrow cells transfected with a genetically engineered form of the protease neprilysin completely arrest amyloid deposition in an aggressively depositing transgenic model. Exploiting the natural homing properties of peripherally derived blood cells to deliver therapeutic genes has the advantages of access to the entire CNS, expression largely restricted to sites of injury, low risk of immune reactivity, and fading of expression if adverse reactions are encountered. These observations support the feasibility of testing autologous monocytes for application of therapeutic genes in human CNS disease. Moreover, these data support the results from bone marrow grafts that circulating CD11b+ cells can enter the CNS without requiring the use of lethal irradiation.
The role of microglia/macrophages in the amyloid pathology of Alzheimer disease is an area of intense investigation (Wyss-Coray, 2006; Morgan et al., 2005). While inflammation associated with the amyloid deposits in AD brain may contribute to the pathogenesis of the disease (Akiyama et al., 2000), macrophages in vitro avidly digest amyloid fibrils (Paresce et al., 1996; Webster et al., 2000). A number of recent studies present evidence that grafted monocytes home to plaques in mouse models of amyloid deposition (Malm et al., 2005; Stalder et al., 2005) and can restrict amyloid deposition, most probably by phagocytosis and digestion of amyloid fibrils (Simard et al., 2006; El-Khoury et al., 2007; Takata et al., 2007; Town et al., 2008). However, it has been argued that studies labeling CNS infiltrating monocytes with bone marrow grafts may be confounded by CNS damage produced by the irradiation used to deplete the recipient’s bone marrow (Ladeby et al., 2005; Mildner et al., 2007; Ajami et al., 2007). This grafting method of labeling circulating monocytes with grafted donor bone marrow may, thus, exaggerate the true extent of monocyte infiltration.
A separate issue is the challenging problem of exploiting gene therapy to treat neurodegenerative disorders such as Alzheimer’s disease. Even using advanced methods such as convection enhanced delivery (Krauze et al., 2005), it is rare that a gene therapy vector can transfect the entire cerebral cortex of a mouse, a region 1/3000th the size of the human cortex (Burger et al., 2005; Carty et al., 2008; Cearley and Wolfe, 2006). For almost a decade, the potential use of circulating monocytes to home to sites of CNS pathology has been proposed as a solution to this gene delivery problem (Imai et al., 1999). However, to date the utility of this approach to gene therapy for brain disorders has not been demonstrated. The data presented here support the concept that circulating monocytes, and possibly other blood cells, do home to amyloid plaques in the brain and that they can be used as effective gene therapy vectors.
Doubly transgenic Amyloid Precursor Protein (APP) + Presenilin-1 (PS1) mice that are a cross between the mAPP transgenic line Tg2576 (Hsiao et al., 1996) and the mPS1 transgenic line 5.1 (Duff et al., 1996) were used in these studies. This breeding produces both APP+PS1 mice and nontransgenic mice (littermates) used in this study. The GFP transgenic mouse model for bone marrow donors were from Jackson laboratory C57BL/6-Tg(UBC-GFP)30Scha/J[Stock#004353]. These transgenic mice express the green fluorescent protein (GFP) under the direction of the human ubiquitin C promoter. Sixteen mo old transgenic and non transgenic mice were used for the single injection time course study and 9 month old APP+PS1 mice were used for the two month multiple injection study. All mice were bred and maintained in our animal facility according to institutional guidelines.
CD11b+ cells were collected using minor variations of previously published protocols (Wu et al, 2006; Wang et al, 2008). Transgenic GFP mice were overdosed with pentobarbital. GFP donor mice were sacrificed and their femurs and tibias removed aseptically. Femur and tibia marrow cavities were flushed with RPMI media containing fetal bovine serum (FBS) and HEPES (pH 7.4) using a 25-gauge needle. Single cell suspensions were prepared by repeat pipetting and the cell preparations passed through a 70 µm nylon mesh to remove particulate matter. Three ml per mouse of 1× RBC Lysis Buffer was added to the cells and incubated at room temperature for 5 minutes before adding cold PBS. Cells were centrifuged, washed twice in RPMI and counted using a hemocytometer.
CD11b+ bone marrow cells were collected using Miltenyi Biotec’s LS columns™ and MidiMacs™ magnet following the manufacturer’s instructions. Briefly, 100–150 million bone marrow cells from transgenic mice ubiquitously expressing GFP were suspended in 2.7 ml PBS+0.5% BSA and incubated for 15 minutes together with CD11b antibody conjugated to magnetic microbeads at 4°C (Miltenyi Biotec, Cat. # 130-049-601). These beads can isolate cells while binding only a fraction of the antigenic sites. The cell suspension was applied to the supplied column in a magnetic field and the CD11b+ fraction separated from the unlabeled cells by washing three times with 3 ml buffer. The column was separated from the magnet and CD11b+ cells were collected. Figure S1, shows freshly isolated GFP+ CD11b cells and its corresponding phase-contrast image.
The purity of immunomagnetically separated and GFP transfected cells were analyzed using a FACScan (Becton Dickinson) equipped with a 488 nm argon laser. The bone marrow cells from nonGFP C57BL/6 mice were harvested as above. To 107 total cells containing 10 µl CD11b microbeads (incubated for 15 minutes at 4°C), we then added 10 µl of anti-CD11b-FITC (a fluorochrome conjugated antibody, Miltenyi Biotec, # 130-081-201) and incubated for 5 minutes at 4°C. Cells were washed by adding 2 ml buffer (PBS+0.5% BSA) and centrifuged at 300–400×g for 10 minutes. Cells were resuspended in 500 µl buffer and transported to Flow Cytometry core for analysis.
Based upon the flow cytometry information, 5 × 106 freshly isolated CD11b+ (GFP) cells were then resuspended in 100 µl saline and injected into the left heart ventricle for the single injection time course studies.
Endogenous NEP is a membrane- bound ectoprotease (native neprilysin or NEP-n). Because Aβ accumulates extracellularly in AD, we hypothesized that a secreted diffusible form of the protease would be most effective in degrading Aβ. Consequently, the membrane binding domain was deleted, leaving bp232–2313, and a GDNF signal sequence triggering secretion was added to the construct (NEP-s). As a control, a construct containing neprilysin with a single-point mutation at the catalytic site (E585V) precluding proteolysis was also generated (NEP-m). A hemaglutinin (HA) tag was appended to each construct.
Proteolytic activity of three different NEP constructs was adapted from a previously published fluorometric assay method (Johnson and Ahn, 2000) for a 96 well plate format with slight modifications. HEK 293 cells were transfected with NEP-n, NEP-m, and NEP-s plasmids and control cells were transfected with a GFP plasmid using lipofectamine 2000 per Invitrogen protocol. Cells were harvested after 72 hrs. and samples were centrifuged at 1,000 × g 4° C for 45 min using Beckman J6- HC Centrifuge (Beckman Instruments, Inc., Palo Alto, CA) to obtain a culture media fraction and a cell pellet. The cell pellets were resuspended in M-PER mammalian protein extraction reagent buffer (Thermo Scientific, Rockford, IL) to obtain a cell lysate. Protein concentration was determined from the cell media and cell lysate from each sample using a general BCA assay (Per Pierce Protocol). Aliquots of the cell media and membrane fractions (100 µg protein) were incubated in NEP ELISA plate containing NEP capture antibody (R&D Systems). The plate was washed with PBS buffer then incubated with 20 µM (final) of the fluorogenic peptide (MOCA- Arg-Pro-Pro-Gly-Phe-Ser-Ala-Phe-Lys(Dnp)-OH), (R&D Systems) in Tris-HCL buffer (sodium phosphate pH 7.4 containing 0.1M NaCl) MOCA- Arg-Pro-Pro-Gly-Phe-Ser-Ala-Phe-Lys(Dnp)-OH is efficiently quenched by resonance energy transfer to the dinitrophenyl group and the continuous fluorescent intensity is increased upon internal cleavage of the peptide (NEP cleavage between the Ala-Phe bond). The increased fluorescence produced from cleavage of the substrate was measured using a Molecular Devices fMax spectrofluorometer plate reader (MDS Analytical Technologies, Sunnyvale, CA) with a 60 min time point to normalize independent experiments. A standard curve of (7-methoxycoumarin-4-yl) acetyl (MOCA) was analyzed along with each assay. Values were calculated and expressed as RFU/ min/ ml protein.
The microvascular ports (Kent Scientific) are 9×9×3 mm polyurethane ports that were placed subcutaneously in the nape of the neck of a mouse and were connected via catheter to the right jugular vein. A skin incision was made over the area of the right jugular, a sterile trocar tunneled to the soft tissue between the scapula and an incision placed at the tip of the trocar. The vessel was dissected free from surrounding tissue. Proximal and distal ligatures were placed around the vessel. The distal ligature was closed, and a venotomy was performed and the catheter advanced into the jugular vein 5 mm. The proximal suture was placed around the catheter and secured. A subcutaneous pocket was be made between the scapulae, and the port was placed into the pocket. Both incisions were closed with suture. The catheter was flushed with 100unit per ml heparin in saline solution, approximately three times the dead space of catheter (30 µl). On the day of injection, mice were anesthetized with isoflurane, the catheter was cleared by inserting a 27 g needle through the skin into the port and injecting 30 µl of 100U/ml heparin in saline, followed by 100 µl of monocyte cell suspension, followed by another 30 µl of the heparin solution.
Mice were anesthetized with pentobarbital and 1 mL of blood was drawn from each mouse via cardiac puncture. 10 ml of RBC lysis buffer (eBioscience, cat # 4333) was immediately added to the blood and incubated for 5 minutes. Each mouse was used for a single determination (we did not serially bleed the same animal). To stop the lysis reaction 30 ml of 1×PBS was added the cells were centrifuged at 300×g at 4°C for 10 minutes. Cells were resuspended in PBS and evaluated for GFP labeled cells by flow cytometry analysis.
On the day of sacrifice, mice were overdosed with pentobarbital (100mg/kg). The brain, spleen, liver and lung were removed, bisected, the right half collected for flow cytometry (see below) and the left half was immersed in freshly prepared 4% paraformaldehyde in 100 mM PO4 buffer (pH 7.4). The organs were postfixed in paraformaldehyde for 24 hours. The brain, liver and spleen tissue were cryoprotected in a series of sucrose solutions, frozen, sectioned in the horizontal plane at 25 µm using a sliding microtome and stored at 4°C in Dulbecco’s phosphate buffered saline for immunocytochemistry and histology. The lung tissue was paraffin embedded before being sectioned using a rotary microtome.
Immunohistochemistry was performed on free floating sections as described in detail previously (Gordon et al., 1997). A series of 8 sections spaced approximately 600µm apart were incubated with primary antibody overnight at 4°C, then incubated in the biotinylated secondary antibody (2h) followed by streptavidin-peroxidase. Peroxidase reactions consist of 1.4 mM diaminobenzidine with 0.03% hydrogen peroxide in PBS for 5 minutes.
Single and multiple immunofluorescent labeling: after incubation with the primary antibody, the free floating sections were incubated for 2 hours with the appropriate fluorophore coupled secondary antibodies [AlexaFluor 594 (1:1500), AlexaFluor 488 (1:1500), AlexaFluor 350 (1:1500) (Molecular Probes, Eugene, OR]. Sections were rinsed in Delbecco’s PBS and coverslipped with VECTASHIELD Mounting Medium.
Primary antibodies used for immunohistochemistry: CD11b (rat monoclonal anti CD11b, Serotec, Raleigh, NC); GFP (Chicken anti-GFP; AbCam), 6E10 (mouse monoclonal; Covance, Emeryville, CA); CD68 (rat monoclonal; Serotec, Raleigh, NC); CD45 (rat antimouse, Abd Serotec); F4/80 (rat anti-mouse, Abd Serotec); Iba-1 (rabbit anti IBA-1, WAKO). haemagluttinin (mouse anti-HA rhodamine; Roche, Indianapolis, IN).
Congo red histology was performed using sections mounted on slides and air dried. Rehydrated sections were incubated in an alkaline alcoholic saturated sodium chloride solution (2.5mM NaOH in 80% alcohol, freshly prepared) for 20 minutes, then incubated in 0.2% Congo red in alkaline alcoholic saturated sodium chloride solution (freshly prepared and filtered) for 30 minutes. Sections were rinsed through three rapid changes of 100% ethanol, cleared through three changes of xylene, then coverslipped with DPX.
Organs were dissociated by enzymatic digestion followed by centrifugation over a percoll gradient. GFP fluorescence intensity was measured using a FACScalibur (BD) flow cytometer and analyzed using CellQuest software.
Immunohistochemistry and Congo red staining were quantified with Image Pro Plus (Media Cybernetics) image software. Segmentation of positive stain was performed using RGB identification. Sample numbers are randomized before the start of the tissue processing, and the code is broken only after the analysis is complete. Data was collected from both frontal cortex and hippocampus. All values obtained from a single mouse were averaged together to represent a single value for that animal. Statistical analysis was performed using ANOVA followed by Fischer’s LSD post hoc means comparison test (Statview software from SAS).
Stereological analysis was performed as described by West et al., 1991. We used a rare event protocol in which all positive cells in the region analyzed were counted after confirming they were not visible on the upper surface of the section.
We purified CD11b+ bone marrow monocytes from donor mice ubiquitously expressing a GFP marker using magnetic cell sorting. Cells were injected into recipient mice using either intracardiac puncture for single injections, or subcutaneous vascular ports inserted into the jugular vein near the left atrium for repeated injections. In all cases we injected 5 × 106 cells in a volume of 100 µl. In figure 1a, blood was collected at multiple times from 5 minutes to 24 hours after a single pulse of cells were injected. GFP+ cells were counted using flow cytometry. We found that the injected CD11b+ cells cleared rapidly from the circulation with a half life of 90 minutes. Virtually all injected cells were cleared from the circulation by 24 hours after the injection.
To identify if the injected CD11b+ cells migrated to the CNS, we compared the tissue distribution of GFP+ cells after the infusion into both nontransgenic mice and mice with amyloid deposits in their brain (16 mo old APP+PS1 transgenic mice; Holcomb et al, 1998). Following exsanguination with saline, the numbers of GFP labeled cells in liver, spleen, lung and brain were estimated at 1, 3 and 7 days both by flow cytometry of cell suspensions and by histological cell counts using stereology (Figs 1b and 1c). Nontransgenic mice had few GFP labeled cells in brain measured either by flow cytometry or stereology (Figs 1d and 1e), yet APP+PS1 transgenic mouse brain had concentrations of GFP labeled cells in the similar numbers as peripheral organs (liver, spleen, lung). In most peripheral organs (Figs1b–c, f–h), the migration of labeled CD11b+ cells in the transgenic and nontransgenic mice were comparable, with the exception of the liver where a slight but statistically significant reduction of infiltrating cells was detected in transgenic mice, possibly due to increased competition for the circulating cells by the brain tissue. In all tissues, the half-life of the labeled cells found within organs was several days, with only a few cells detected one week after the injection.
In the brain, the presence of amyloid deposits greatly enhanced the infiltration of GFP+ cells into the CNS. We found the GFP labeled cells homed to the immediate vicinity of the compacted amyloid deposits. The majority of plaques in mice administered twice weekly injections for 2 mo had multiple GFP labeled cells in the surrounding region. We found very few, if any, labeled cells in brain regions that lacked amyloid deposits. Virtually all of these GFP+ cells expressed CD11b (Fig 2a–c). A second marker, CD68, expressed by phagocytic cells, labeled largely those microglia/macrophages in the immediate vicinity of the plaque (Fig 2e). Double labeling studies found that some, but not all of the GFP labeled cells co-expressed CD68, in particular those GFP+ cells that were less ramified tended to co-express this marker (Fig 2f). No GFP signal could be found in APP+PS1 mice that were not injected with labeled monocytes (Fig 2g–i).
To further identify the cell type expressing GFP we evaluated several other markers of brain microglia in double labeling studies. Figure 3 shows cells labeled with CD45, IBA-1 or F4/80 in conjunction with GFP labeling. As for CD68, not all cells expressing these microglial markers were labeled with GFP, but there were some cells that co-expressed both proteins. Not all GFP cells co-expressed CD45 Fig 3c) nor did all GFP cells co-express F4/80 (Fig 3i), but to a first approximation it appeared that most GFP cells also expressed IBA-1 (Fig 3f). Combined with the data from CD68 co-localization it is certainly the case that the GFP cells assume multiple phenotypes, and the IBA-1 data suggest that to a large extent these are microglial phenotypes.
Neprilysin is an ectoprotease that may be the primary enzyme responsible for Aβ degradation within the brain (Iwata et al., 2000; Marr et al., 2003; Leissring et al., 2003; Hemming et al., 2007). To enhance the likelihood this enzyme would reach amyloid deposits from transfected microglia, we engineered a secreted form of the enzyme (NEP-s) by deleting the transmembrane domain and substituting the signal peptide for glial derived neurotrophic factor. We further appended a haemagluttinin (HA) tag to the C-terminus of the enzyme to enable specific detection of the recombinant enzyme. As a control gene, we transfected cells with a mutant form of neprilysin containing a single amino acid substitution (E585V) known to inhibit enzyme activity (NEP-m).
The protease activity of these neprilysin constructs was measured by transfecting HEK293 cells and measuring the neprilysin activity in the medium and in the cell homogenates. Cells transfected with GFP showed little activity and were identical to cells transfected with the NEP-m construct (Fig 4a) While increased activity was observed in both the NEP-n and NEP-s treated samples. As expected, the NEP-n had more activated associated with the cell fraction than NEP-s. In the culture media fraction (Fig 4b) the only condition to demonstrate enzyme activity was cells transfected with the NEP-s construct. Thus, both the NEP-m and NEP-s constructs behave as intended with no activity present in either fraction of NEP-m and only the NEP-s construct demonstrating activity in the culture media.
For gene therapy studies, we isolated the CD11b+ monocytes from the bone marrow of GFP donor mice and transfected them by electroporation. A major challenge in working with monocytes is a difficulty in efficiently transfecting them. Transfection efficiencies less than 10% have been accomplished using traditional transfection techniques, including electroporation, liposomal transfection reagents and DEAE-dextran (Mayne et al., 2003). By performing electroporation with the Amaxa macrophage protocol, we obtained a 54% transfection efficiency measured by flow cytometry using a GFP expressing plasmid (in monocytes from non-GFP mice) (Fig 5a–b). Staining CD11b+ cells from GFP expressing mice for HA expression after transfection with NEP plasmids, we show that only the transfected cells are positive for HA when compared to control transfected cells (Fig 5c–d). Cells were then infused into the jugular vein of 9 mo old APP+PS1 mice twice weekly using a subcutaneous port. Tissues were collected 2 mo later when the mice were 11 mo of age. In addition to the mice administered NEP-s and NEP-m monocytes, we also collected tissue from 11 mo old mice that were untreated and from 9 mo old untreated APP+PS1 mice to assess the degree of amyloid deposition when the treatment was initiated.
In untreated APP+PS1 mice, there was a dramatic increase in the amyloid load detected with Congo red staining between 9 and 11 mo, indicating this was an accelerating phase of amyloid deposition in this model (Fig 6). Treatment of the mice with NEP-s transfected cells completely blocked the increase in amyloid deposition found between 9 and 11 mo in these mice. However, treatment with CD11b+ cells transfected with the inactive form of the enzyme, NEP-m, showed the same degree of amyloid deposition as the untreated animals.
When amyloid loads were measured with a polyclonal antibody to Aβ, a similar 2 fold rise in Aβ deposition was obtained between 9 and 11 mo of age (Fig 6). Again, this elevation was abrogated by the administration of cells containing NEP-s, but not NEP-m. There is a trend in the results for the administration of the monocytes with the inactive NEP-m to have less Aβ immunostaining than untreated mice, but this was not statistically significant unless the 9 mo control group was excluded from the analysis. Nonetheless, this observation suggests some minor benefit of excess circulating monocytes in slowing Aβ deposition.
These results are the first demonstration that circulating CD11b+ cells derived from bone marrow can be used to deliver therapeutic genes to the CNS. The effect of the cells transfected with NEP-s cannot be simply due to increasing the macrophage population, as the NEP-m transfected cells had minimal effects on amyloid deposition. Similarly the NEP-m group also controls for the isolation and transfection procedures and expression of an heterologous protein. One minor difference is the the NEP-s is secreted while the NEP-m remains membrane bound, Thus it cannot be decisively excluded that secretion of a heterologous protein accounts for the results, irrespective of proteolytic activity but we believe this outcome is unlikely.
At the end of the study (1 day after the final injection) HA expression was detected in brain, but as expected, none could be found in plasma, arguing for a central site of action. Moreover, by measuring the amyloid loads at the beginning of the experiment, we demonstrated that the NEP-s activity totally prevented further amyloid deposition. Few studies have included this treatment initiation control group, and those that have included this group, show only slowing of deposition rather than total prevention (Best et al., 2007). The APP+PS1 mouse is an aggressive model of amyloid deposition, with first deposits around 3–4 mo of age and amyloid loads at least 3 times greater than the parental singly transgenic APP (Tg2576) mouse at most ages (Gordon et al., 2002; Holcomb et al., 1998). Hence we anticipate that AD patients and other transgenic mouse models with slower rates of Aβ accumulation could benefit from even less exposure to the therapeutic gene than that administered here.
These results add to the increasing evidence that peripheral CD11b+ cells, presumed to be largely monocytes, migrate to sites of neurodegeneration within the brain. In this instance, the migration cannot be accounted for by irradiation damage to the CNS (or other tissues), excess numbers of stem cells in circulation or systemic illness brought about the irradiation. The argument that Aβ deposits stimulate recruitment (Fiala et al, 1998; Giri et al, 2000; Humpel, 2008) is consistent with our observation that very few labeled cells were found in brain regions that lacked amyloid deposits. Moreover, these studies may be the first to estimate the half-life of a circulating CD11b+ cell population and the half-lives of the tissue infiltrating CD11b+ cells after a pulsed administration. Both half-lives were shorter than anticipated. The injected CD11b+ cells initially increase the numbers of these cells in blood by a factor of 2 or more. Hence an elevated monocyte concentration may contribute to the trafficking we find into the CNS of transgenic mice. However, it is unlikely that mechanical or chemical aspects of the intravenous infusion (5% of blood volume in buffered physiological saline) account for the CNS infiltration observed, as very few cells entered the brains of nontransgenic mice. One caveat is that the cells used here were derived from marrow, and thus are immature compared to CD11b+ cells already in circulation. It is conceivable that this immaturity contributes to their CNS migration. We interpret these data as supporting the argument that circulating CD 11b+ cells contribute a meaningful fraction of the activated myeloid cells in the vicinity of compacted amyloid plaques, likely in response to chemotactic factors secreted in the vicinity of the deposits.
This approach to ex vivo gene therapy has multiple applications for demonstrating proof of concept for specific therapeutic genes in human conditions. For trials of human gene therapy, a patient’s own monocytes could be elutriated from blood, transfected ex vivo, then reintroduced in a single visit. Given the short half lives of the monocytes, this would require repeated injections before therapeutic effects might be observed. However, this time dependent fading of therapeutic gene expression would spontaneously reverse unforeseen adverse effects (much like clearance of a small molecule drug) that could be challenging to treat using more permanent gene therapy approaches. For neurodegenerative disorders, this approach would permit distribution of the therapeutic gene throughout the brain, and deliver the gene precisely at locations undergoing degeneration. Compared to viral vectors, there would be no expression of the gene in unintended cells, or a gradient of expression around the injection site. Moreover, there would be no need for intracranial surgery, the hazards of which were evident in the first gene therapy trial for Alzheimer’s disease (Tuszynski et al., 2005). The absence of integration into the host genome by the plasmid, and the use of postmitotic cells would diminish the risk for inadvertent transformation into cancerous cells. As a long term goal, once the efficacy and safety of a gene therapy approach is demonstrated using monocytes, the gene could be transfected into bone marrow stem cells. With appropriate promoters to drive expression of the gene in differentiated tissue macrophages, as opposed to circulating monocytes, this approach might provide a safe, effective and permanent treatment for some CNS degenerative diseases.
This work is supported by grants from NIH (AG-25509, AG-15490, AG04418), from the Johnnie B Byrd Research Institute and from the Alzheimer’s Association. LL is the Benjamin Scholar in Alzheimer’s Research. We thank the staff of the Division of Comparative Medicine, especially Margaret Baldwin for her assistance with the surgical insertion of microvascular ports. We thank Barbara Nicholson and Bridget Shields for clerical assistance in conducting these studies. A disclosure has been filed with the USF Office of Patents and Licensing for possible patent protection of this approach to gene therapy. Morgan, Lebson and Kamath are inventors