Strategy to enhance transgene expression in proximity of amyloid plaques in a mouse model of Alzheimer's disease

Gene therapy can be designed to efficiently counter pathological features characteristic of neurodegenerative disorders. Here, we took advantage of the glial fibrillary acidic protein (GFAP) promoter to preferentially enhance transgene expression near plaques composed of amyloid-beta peptides (Aβ), a hallmark of Alzheimer's disease (AD), in the TgCRND8 mouse model of amyloidosis. Methods: The delivery of intravenously injected recombinant adeno-associated virus mosaic serotype 1/2 (rAAV1/2) to the cortex and hippocampus of TgCRND8 mice was facilitated using transcranial MRI-guided focused ultrasound in combination with microbubbles (MRIgFUS), which transiently and locally increases the permeability of the blood-brain barrier (BBB). rAAV1/2 expression of the reporter green fluorescent protein (GFP) under a GFAP promoter was compared to GFP expression driven by the constitutive human beta actin (HBA) promoter. Results: MRIgFUS targeting the cortex and hippocampus facilitated the entry of rAAV1/2 and GFP expression under the GFAP promoter was localized to GFAP-positive astrocytes. Adjacent to Aβ plaques where GFAP is upregulated, the volume, surface area, and fluorescence intensity of the transgene GFP were greater in rAAV1/2-GFAP-GFP compared to rAAV1/2-HBA-GFP treated animals. In peripheral organs, GFP expression was particularly strong in the liver, irrespective of the promoter. Conclusion: The GFAP promoter enhanced transgene expression in proximity of Aβ plaques in the brain of TgCRND8 mice, and it also resulted in significant expression in the liver. Future gene therapies for neurological disorders could benefit from using a GFAP promoter to regulate transgene expression in response to disease-induced astrocytic reactivity.


Introduction
Recent successes in gene therapy clinical trials include improvements in the vision of patients with leber congenital amaurosis (1), and the first life-saving treatment of neurodegeneration in infants with spinal muscular atrophy (2). These breakthroughs and the advancement of recombinant adeno-associated viruses (rAAVs) have renewed interest in gene therapy for neurological disorders (3)(4)(5). However, for most disorders of the central nervous system (CNS), challenges in translating gene therapy approaches to the clinic include delivery across the blood-brain barrier (BBB) (6,7), and the control of transgene expression (8). Though some more recent rAAVs, such as the AAV9 variant AAV-PHP.B, have been Ivyspring International Publisher shown to overcome the BBB, they cannot be targeted to regions within the brain after systemic delivery (9), which could increase the risk of off-target effects (9). Additionally, the increased brain bioavailability of some of these new capsid variants may be unique to rodents and not observed in non-human primates (10,11) compared to rAAV9.
Recently, ultrasound-mediated BBB permeability has entered clinical trials to establish the safety of the procedure in patients with Alzheimer's disease (AD) (20). When compared to intracranial injections, MRIgFUS delivery of therapeutics to the brain is less invasive, thereby mitigating risks associated with surgical procedures, including infection (21) and tissue damage (22). Additionally, a single MRIgFUS session can cover several areas of the brain or spinal cord with multiple focal points. Intraparenchymal injection of rAAV is associated with limited diffusion and coverage. For example, the cross sectional area of both human hippocampi would require an impractical amount (>50) of intracranial injections (23)(24)(25)(26).
In terms of control following systemic injection, cell-specific promoters can modulate transgene expression in the CNS and in peripheral organs. To that end, the astrocyte-associated, 2.2 kilobase pair (kbp) glial fibrillary acidic protein (GFAP) promoter (27) was tested to control rAAV-mediated green fluorescent protein (GFP) expression. In AD brains where amyloid-beta peptides (Aβ) are present, astrocytes in proximity to plaques and throughout the neuropil contribute to the observed increase in endogenous GFAP immunoreactivity (28). As of three months of age, the TgCRND8 mice demonstrate Aβ deposition in the cortex and hippocampus (29). They likewise demonstrate an increase in astrogliosis measured by GFAP starting at three and half months of age, which progresses with age and Aβ pathology (30). Here, the cortex and hippocampus were targeted with MRIgFUS, in the presence of microbubbles, to facilitate BBB delivery of rAAV1/2-GFP under control of either the GFAP promoter or the constitutive human beta actin (HBA) promoter. GFP expression under the GFAP promoter was significantly higher with respect to fluorescence intensity, as well as volume and surface area of transgene protein distribution in GFAP-positive cells (astrocytes) associated with Aβ plaque, compared to non-Aβ affiliated astrocytes, or astrocytes transduced with rAAV-GFP under control of the HBA promoter. The GFAP promoter permits Aβ-responsive expression, resulting in targeted increases in transgene expression corresponding to increases in Aβ-mediated astrocytic activation. Thus, this expression system could provide a form of therapeutic transgene control that self-modulates with disease progression.

MRIgFUS facilitates targeted rAAV1/2 delivery to the cortex and hippocampus
Briefly, rAAV1/2-GFAP-GFP or rAAV1/2-HBA-GFP were injected at a dose of 3 x 10 9 vector genomes per gram (VG/g) through a tail vein catheter in TgCRND8 mice. FUS application immediately preceded viral injection, for which the mice were placed in dorsal recumbency over a spherical ultrasound transducer, as previously described (31). MRI images were used to target FUS to the cortex and hippocampus, and contrast-enhanced MRI was used to verify BBB opening and location ( Figure 1A, B). Three, non-overlapping spots were used to target the cortex (1 spot) and hippocampus (2 spots) in each animal ( Figure 1C, D, E). Results show that GFP expression from the rAAV1/2-GFAP-GFP or rAAV1/2-HBA-GFP constructs were concentrated and limited to the FUS-targeted regions 14 days post-delivery (Figure 2A, B). Following delivery of rAAV1/2-GFAP-GFP, GFP-expressing cells in the cortex ( Figure 2C) and hippocampus ( Figure 2E) also expressed GFAP ( Figure 2G), indicating astrocyte specificity of the GFAP promoter. GFP-expressing cells in the cortex ( Figure 2D) and hippocampus ( Figure 2F) of the rAAV1/2-HBA-GFP group did not always co-localize with GFAP ( Figure 2H), as the HBA promoter allows for transgene expression in a variety of cell types (32)(33)(34).

GFAP and HBA promoters result in comparable numbers of GFP-positive cells
Firstly, the percentage of GFAP-positive cells within the population of cells expressing GFP was quantified in both the rAAV1/2-GFAP-GFP and rAAV1/2-HBA-GFP groups ( Figure 2I). This confirmed that the GFAP promoter leads to preferential transgene expression in astrocytes.
After MRIgFUS delivery, the areas of GFP-expression within the cortex and hippocampus of the rAAV1/2-GFAP-GFP and rAAV1/2-HBA-GFP groups were not significantly different (p=0.91) ( Figure 2J), indicating that the focal spots were able to mediate permeabilization of a consistent size. Furthermore, the amount of BBB opening by FUS, as measured by the MRI enhancement from background also confirms that delivery of the rAAV1/2-GFAP-GFP and rAAV1/2-HBA-GFP groups was not significantly different (p=0.87) ( Figure 2K). As an additional confirmation of baseline consistency, the average number of Aβ plaques within the FUS-targeted regions was compared between viral groups and was not significantly different (p=0.69) ( Figure 2L). Collectively, this supports that differences in transgene expression are due to the GFAP versus HBA promoter transcriptional control, and not to differences in FUS delivery or plaque load within the targeted areas.

GFP expression under the GFAP promoter is enhanced near Aβ plaque
This study was designed to characterize and quantify the possible increase in transgene expression near Aβ plaque under control of the GFAP promoter, compared to expression unassociated with Aβ plaque, or under control of a constitutive promoter.
Among the astrocytes (GFAP-positive cells) where GFP is expressed under the GFAP promoter, GFP expression shows a distinct distribution pattern in astrocytes with processes overlapping Aβ plaque ( Figure 3A-D), compared to astrocytes unassociated with Aβ plaque ( Figure 3E-H). On the other hand, GFP expression under the HBA promoter do not show a visible difference in GFP distribution pattern in astrocytes with processes overlapping Aβ plaque ( Figure 3I-L), compared to astrocytes unassociated with Aβ plaque ( Figure 3M-P). This suggests that expression of GFP under the GFAP promoter but not the HBA is affected by Aβ pathology. The observed promoter-differences in GFP distribution within Aβ-associated astrocytes were not caused by differences in cell morphology ( Figure 3B and J).
Quantification of mean fluorescence intensity per unit volume shows that GFP expression under the GFAP promoter in astrocytes is significantly higher near Aβ plaque than in astrocytes unassociated with Aβ (p<0.01), or under control of the HBA promoter (Aβ-associated, p<0.001; Aβ-unassociated, p<0.001) ( Figure 4A). The volume of GFP distribution was also significantly higher in astrocytes associated with Aβ plaque relative to astrocytes unassociated with Aβ in the rAAV1/2-GFAP-GFP group (p<0.01), and compared to all GFP-positive astrocytes from the rAAV1/2-HBA-GFP group (p<0.0001) ( Figure 4B). Additionally, the surface area of GFP distribution was significantly greater in astrocytes associated with Aβ plaque in the rAAV1/2-GFAP-GFP compared to astrocytes unassociated with Aβ (p<0.05), and to GFP-positive astrocytes from the rAAV1/2-HBA-GFP group (p<0.001) ( Figure 4C). Fluorescence intensity, volume, and surface area were not significantly different between non-Aβ associated, GFP-positive astrocytes under control of the GFAP promoter, and GFP-positive astrocytes under control of the HBA promoter (p>0.05).

Figure 1. MRI-guided focused ultrasound (MRIgFUS) mediates blood-brain barrier (BBB) opening in the cortex and hippocampus. (A)
A T2-weighted MRI image is used to target brain regions with focused ultrasound (FUS). (B) BBB opening was verified using gadolinium enhancement (arrows: purple targeting the cortex, green targeting the hippocampus), as seen on the T1-weighted MRI image acquired immediately after FUS. (C) FUS was applied using one focal point targeting the cortex (purple circle) and two focal points targeting the hippocampus (green circles). (D and E). The focal spot generated using these parameters is oval in shape, which is demonstrated in the coronal perspective. The focal spots include regions of the cortex (pink) and hippocampus (green), which contain deposits of Aβ plaque in TgCRND8 mice as of 3 months of age. (D and E) Brain atlas images were adapted from the Allen Mouse Brain Atlas.
GFP and GFAP fluorescence intensities in GFP-positive astrocytes strongly correlated in the rAAV1/2-GFAP-GFP group (r=0.75, p<0.0001), but not in the rAAV1/2-HBA-GFP group (r=0.04, p=0.81) ( Figure 4E). This supports that the increase in GFP near plaques are due to an increase in GFAP promoter activity.

GFAP promoter limits transgene expression in periphery, with exception of the liver and kidney
Although expression of rAAV1/2-GFP under the GFAP promoter in the cortex and hippocampus led to almost exclusive colocalization with GFAP-positive cells ( Figure 2I), this was not the case in the liver ( Figure 5A) where GFP expression was seen, even in the absence of detectable GFAP expression. GFP expression was also visible in the liver after delivery of rAAV1/2-HBA-GFP ( Figure 5B). A few GFP positive cells were seen in the kidney in both the GFAP promoter group ( Figure 5C) and in the HBA promoter group ( Figure 5D); GFP expression under control of the GFAP promoter was not detected in the heart ( Figure 5E), muscle ( Figure 5G), spleen ( Figure   5I), or lung ( Figure 5K). In contrast, a few GFP positive cells were seen when rAAV1/2-GFP was expressed under the HBA promoter in the heart ( Figure 5F), muscle ( Figure 5H), and spleen ( Figure  5J), but no positive cells were detected in the lung ( Figure 5L).

Discussion
Using a GFAP promoter, we have demonstrated that transgene expression (i.e. GFP) is increased alongside GFAP-positive areas of astrocytic activation. This finding introduces the possibilities of augmenting therapeutic delivery near pathological hallmarks and regulating transgene expression in response to disease progression and therapeutic effects. GFAP expression is increased in several disorders and injuries of the central nervous system including AD (35,36), amyotrophic lateral sclerosis (37), and multiple system atrophy (38), as well as rodent models of traumatic brain injury (39,40). In a mouse model of AD, Vitale et al. found increased efficacy in reducing tau pathology when using a GFAP promoter to express anti-tau antibodies, compared to an ubiquitous promoter (41). Here, enhanced transgene expression near astrogliosis is exemplified using Aβ plaque deposition in the TgCRND8 mouse model of amyloidosis.  rAAV1/2-GFAP-GFP and rAAV1/2-HBA-GFP were delivered from the bloodstream of TgCRND8 mice to the cortex and hippocampus using MRIgFUS. GFP expression under the GFAP promoter led to heightened GFP fluorescence intensity and increased GFP distribution volume and surface area in astrocytes near Aβ plaque, compared to GFP-positive astrocytes unassociated with Aβ plaque, and GFP-positive astrocytes from the rAAV1/2-HBA-GFP delivery group. This demonstrates that transgenic protein quantity and distribution throughout astrocyte processes can be selectively enhanced in the vicinity of Aβ, thereby increasing therapeutic delivery alongside plaque load. Additionally, FUS has been shown to increase GFAP expression at four, and up to fifteen, days following application in the targeted cortex of TgCRND8 mice (42). Together with the results shown here, this suggests that the GFAP promoter can be utilised to both increase transgene expression near Aβ plaques as well as to boost transgene expression in all transduced cells by reapplication of FUS. Gene therapy is traditionally a one-chance treatment which often does not allow for retreatments due to expression of anti-AAV antibodies following the first administration. Future studies will investigate the GFAP promoter as a means to enable a boosting of therapeutic expression by reapplication of FUS to increase GFAP promoter activity.
The increase in endogenous GFAP expression in association with Aβ plaques is approximately half that of GFP expressed under the GFAP promoter. Discrepancies between endogenous GFAP and GFP expressed under a GFAP promoter have been previously reported, and suggested to be caused by the different subcellular localizations of the GFAP and GFP proteins (43). Others have also shown that increased GFAP in TgCRND8 mice correlates with age and Aβ pathology, but that it is also variable and not only found near thioflavin-positive plaques (30,44,45). Regardless of these fluorescence discrepancies, GFP demonstrated a correlation with GFAP fluorescence when under the control of the GFAP promoter, suggesting their regulation is strongly linked.
Protein expression has been correlated in vitro (46), in E. coli (47), and in mice (48) with transgene fluorescence intensity. However, a caveat to using fluorescence intensity to measure protein expression is that its precision is vulnerable to changes in fluorescence background (49), self-aggregation of fluorescent species (50), regional differences in pH (51), photobleaching (51), and pixel saturation (49,52). In order to compliment GFP quantification in a manner that was independent of differences in regional intensity, the volume and surface area of GFP distribution within GFAP-positive cells were also measured. Although volume provides a 3D measurement of space occupied by GFP expression, surface area is more sensitive to distribution within astrocyte processes (53).
In the currently described findings, GFP expression under control of the GFAP promoter was found in the liver, and to a limited extent in the kidney; however, endogenous GFAP expression was not detected ( Figure 5). Previous studies have shown variable results in transgene expression in the liver under control of the same 2.2 kbp GFAP promoter (gfp2) used here, with some results showing transgene expression in the liver (54,55) and others finding no detectable transgenic protein (43,56). The presence of transgene expression under the GFAP promoter in the absence of GFAP could be related to access restriction of highly condensed chromatin containing the genomic GFAP promoter and gene sequence, or variable trans and cis chromosomal interactions, which may not affect transgene expression from an episomal GFAP promoter construct (57). It is also known that hepatic stellate cells representing 5-8% of the human liver cells express GFAP (58). On the other hand, a point of consideration for transgene expression in the liver is duration, as rAAV-mediated transgene expression can be lost after several weeks, whereas expression in the brain has been shown to persist for several years (59,60). The mechanisms of expression loss in the liver (17), and inconsistencies in GFAP promoter-driven expression within off-target organs (43,(54)(55)(56) remain to be fully elucidated. In order to prevent even transient expression in the liver after systemic delivery of rAAV1/2-GFAP, future investigations could utilize organ-specific microRNA inhibition (61).

Conclusion
Our results provide proof-of-concept for a novel approach using MRIgFUS to facilitate non-surgical delivery of a gene vector containing a GFAP promoter, hereby enhancing transgene expression in astrocytes surrounding amyloid plaques. In GFP-positive astrocytes associated with Aβ plaque in the rAAV1/2-GFAP-GFP group, fluorescence intensity, as well as volume and surface area of GFP distribution was increased, compared to astrocytes unassociated with Aβ plaque, or transgene-positive astrocytes from the rAAV1/2-HBA-GFP delivery group. This data illustrates the potential of the GFAP promoter to target and increase transgene expression alongside astrocyte activation and pathology, and future studies will evaluate the efficacy of therapeutic molecules expressed under the control of a GFAP promoter to decrease Aβ pathology. Provided that astrogliosis occurs in cases of neurodegeneration, neuroinflammation, stroke and other types of injuries of the central nervous, the use of promoters responding to astrogliosis could be beneficial in curbing disease progression.

Animals
TgCRND8 mice were used at 15 weeks of age, with an average mass of 26 grams. The animal procedures carried out in these experiments complied with the Canadian Council on Animal Care and the Animals for Research Act of Ontario guidelines, and were approved by the Sunnybrook Research Institute Animal Care Committee. The rAAV1/2-GFAP-GFP and rAAV1/2-HBA-GFP delivery groups each included four mice.

Virus Preparation
rAAV1/2 expressing enhanced GFP was generated under control of either the 2,210 base pair human GFAP promoter, or the HBA promoter as previously described (27,33). Briefly, rAAV1 and rAAV2 packaging plasmids were used at a 50:50 ratio to generate mosaic rAAV1/2 particles, which were purified using iodixanol gradient centrifugation and fast protein liquid chromatography on heparin affinity columns. To increase expression, the cytomegalovirus enhancer sequence was included directly upstream of the HBA promoter. The woodchuck hepatitis virus posttranscriptional regulatory element was included after the GFP sequence to enhance mRNA stability, along with the bovine growth hormone polyadenylation sequence. rAAV virus was injected at a dose of 3 x 10 9 VG/g through a 22-G angiocatheter in the tail vein for MRIgFUS delivery.

Magnetic Resonance Imaging-Guided Focused Ultrasound (MRIgFUS)
Isofluorane inhalation was used to anesthetize the mice, and depilatory cream was applied to remove hair from the head and neck. The mice were positioned in dorsal recumbency over an MRI radiofrequency surface coil as previously described (31).
A 7T MRI (Bruker BioSpin MRI GmbH, Ettlingen, Germany) was used to generate images of the brain and target regions of the cortex and hippocampus ( Figure 1A). Unilateral targeting of the cortex and hippocampus was done using one or two FUS spots, respectively (Figure 1). A 1.68 MHz spherically focused transducer (aperture: 7 cm, F-number: 0.8) was used to generate ultrasound, and was driven using a function generator and radio frequency power amplifier. FUS sonications were applied using 10 msec bursts, at a repetition frequency of 1 Hz, for 120 seconds. To control acoustic pressures, a 4.8 mm diameter wideband polyvinylidene fluoride hydrophone was used as previously described (62). For all sonications, the acoustic pressure amplitude was increased in a step-wise manner, while the hydrophone was used to detect sub-harmonic acoustic emissions. When a 840 kHz sub-harmonic emission was detected by the hydrophone, the pressure amplitude level was dropped to 50% of the value at which the subharmonic had been detected, and maintained for the duration of the sonication. An injection of Definity microbubbles (0.02 ml/kg), followed by saline (200 µL) through the tail vein catheter was given immediately before FUS application. Subsequently, virus was injected (3 x 10 9 VG/g), followed by saline (200 µL), Gadodiamide MRI contrast agent (0.2 ml/kg, Omniscan, GE Healthcare Canada, Mississauga, ON, Canada), and additional saline (200 µL). Following FUS application, contrast-enhanced T1-weighted MRI images were acquired at a resolution of 0.25 x 0.25 x 1.5 mm in the X x Y x Z axis to visualize the 1 mm 2 BBB permeability, as demonstrated by regions of enhancement ( Figure 1B, arrowheads). Upon recovery from anesthesia, the mice were returned to their cages.

Tissue Processing
14 days after MRIgFUS application, mice were anesthetized using an intraperitoneal injection of ketamine (75 mg/kg) and xylazine (10 mg/kg). Transcardial perfusion using 0.9% saline and 4% paraformaldehyde solution in 0.1M PO 4 was performed. The brain and peripheral organs were collected and post-fixed in 4% paraformaldehyde solution for 24 hours, transferred to 30% sucrose solution and then stored at 4°C. The brains and peripheral organs were mounted in Tissue-Tek OCT (Sakura, Torrance, CA, USA), frozen with dry ice and cut in 40 μm-thick sections on a sliding microtome. Sections were kept at -20°C in cryoprotective glycerol solution.

Immunohistochemistry
Free floating brain sections were rinsed in phosphate-buffered saline (PBS, pH 7.4) for five minutes three times before antigen retrieval, which was done using incubation in 70% formic acid in PBS at room temperature for 5 minutes. The sections were rinsed three times before incubation for 1 hour at room temperature in blocking solution (PBS++) composed of 2% donkey serum (Wisent Bioproducts, Peripheral organ sections were stained as described above without antigen retrieval in formic acid, and in blocking solution that consisted of 10% donkey serum and 1% TX-100 in PBS. The primary antibodies used were the same rabbit anti-GFP (1:500), donkey anti-GFAP (1:300), and DAPI as described above.

Cell Counting
The numbers of Aβ plaques, GFP-positive and GFAP-positive cells, and GFP-positive and GFAP-negative cells within the FUS-targeted areas of the brain were quantified using Stereo Investigator software on a Zeiss AxioImager M2 microscope. GFP expression associated with Aβ plaque was defined as the occurrence of plaque and GFP expression from the GFAP-positive cell body or projections overlapping in space. For coronal brain sections, six 40 μm-thick sections from the FUS-targeted regions were used at an interval of one in six for quantification. For axial brain sections, five 40 µm-thick sections were used at an interval of one in three. Quantification was done using an optical fractionator probe and 63x oil objective (NA= 1.4) on an exhaustive grid covering all regions of the hippocampus and cortex with visible GFP expression. The final number of plaque and cell counts was extrapolated from the section interval. The Cavalieri estimator probe within the Stereo Investigator software was used to estimate area of GFP expression. The mean number of GFP-positive cells, area of GFP expression and number of Aβ plaques were evaluated between the rAAV1/2-GFAP-GFP and rAAV1/2-HBA-GFP experimental groups using a two-tailed, unpaired t-test. The difference in number of GFAP-positive and GFAP-negative cells within the rAAV1/2-GFAP-GFP and rAAV1/2-HBA-GFP groups was evaluated using a two-way ANOVA and Bonferroni post-hoc comparison. A p value of less than 0.05 was considered statistically significant for all analyses. All statistical analyses were performed using GraphPad Prism 5 (GraphPad Software, La Jolla, CA, USA).

Fluorescence Quantification
The GFP and GFAP fluorescence intensity per unit volume, as well as GFP expression volume and surface area of a GFP and GFAP-positive cell, either associated or unassociated with Aβ plaque, was compared using the 3D Measurement module of Nikon Elements software (Nikon Instruments). This analysis was done using 0.1 µm Z-stack images comprising the entire volume of the GFP-positive cell, taken with a 60x objective (NA 1.4). A sample of 12 Z-stack images of GFP and GFAP-positive cells associated with Aβ plaque was used per rAAV1/2-GFAP-GFP or rAAV1/2-HBA-GFP group. A sample of 20 Z-stack images of GFP and GFAP-positive cells that were unassociated with Aβ plaque were used per rAAV group. The mean fluorescence per unit area, as well as volume and surface area of GFP expression were compared between GFP and GFAP-positive cells either associated or unassociated with Aβ plaque from the rAAV1/2-GFAP-GFP or rAAV1/2-HBA-GFP delivery group using a Kruskal-Wallis one-way analysis of covariance (Gaussian distribution not assumed) and Dunn's multiple comparison test post hoc (GraphPad Prism 5, GraphPad Software).

MRI Enhancement Quantification
Enhancement at each FUS focal spot was measured from an average of a 3X3 pixel area of the MRI image and expressed as a percentage increase from background enhancement using Matlab (MathWorks, Natick, MA, USA). A two-tailed, unpaired t-test was used to compare the enhancement of all focal spots between the experimental groups (GraphPad Prism 5, GraphPad Software).