Preventive cognitive protection based on AAV9 overexpression of IGF1 in hippocampal astrocytes

Astrocytes play key roles in the brain. When astrocyte support fails, neurological disorders follow, resulting in disrupted synaptic communication, neuronal degeneration, and cell death. We posit that astrocytes over-expressing neurotrophic factors, such as Insulin Like Growth Factor 1 (IGF1), prevent the onset of neurodegeneration. We overexpressed IGF1 and the reporter TdTomato (TOM) in hippocampal astrocytes with bicistronic Adeno-Associated Virus (AAV) harboring the Glial Fibrillary Acidic Protein (GFAP) promoter and afterwards induced neurodegeneration by the intracerebroventricular (ICV) injection of streptozotocin (STZ), a rat model of behavioral impairment, neuroinflammation and shortening of hippocampal astrocytes. We achieved a thorough transgene expression along the hippocampus with a single viral injection. Although species typical behavior was impaired, memory deficit was prevented by IGF1. STZ prompted astrocyte shortening, albeit the length of these cells in animals injected with GFP and IGF1 vectors did not statistically differ from the other groups. In STZ control animals, hippocampal microglial reactive cells increased dramatically, but this was alleviated in IGF1 rats. We conclude that overexpression of IGF1 in astrocytes prevents neurodegeneration onset. Hence, individuals with early neurotrophic exhaustion would be vulnerable to age-related neurodegeneration.


Introduction
Despite significant advancements to comprehend neurodegenerative mechanisms, available pharmacological interventions primarily address symptoms.The eventual consequence of neurodegeneration manifests as the demise of neural cells, brain shrinkage, and impaired brain function (Selkoe et al., 2012).In this context, the primary role of neuroglia is to maintain the integrity of nervous tissue.
Atrophic alterations in astrocytes appear as a reduction in the size of astroglia territories.The atrophy and functional decline of astrocytes contribute to early cognitive impairments by reducing metabolic support and causing synaptic dysfunction (Brandebura et al., 2023).Moreover, the decreased coverage of synapses by astrocytes may lead to neurotransmitter spillover, resulting in hyperexcitability of neuronal networks frequently observed in neurodegenerative conditions (Blanco-Suárez et al., 2017).
The hippocampus is a pericortical structure critical for memory formation and retrieval particularly susceptible to neurodegeneration (Pardo et al., 2017;Squire et al., 1993).In the STZ model, rats display neuroinflammation in the hippocampus and fail on behavior tasks depending mostly on the hippocampus, such as the Barnes Maze (BM) and the novel object recognition (NOR), but also in tests depending on several brain regions, for instance the marble burying (MB) test (Zappa Villar et al., 2021;Zappa Villar et al., 2018).
Insulin like growth factor 1 (IGF1) is a neuroprotective agent that has been used to implement gene therapy by ICV injecting a Recombinant Adenovirus under the control of the Cytomegalovirus promoter (RAd-CMV-IGF1) in senile rats and in the STZ model (Pardo et al., 2018;Pardo et al., 2016;Zappa Villar et al., 2021).Interestingly, in rodents, the hippocampus is an anatomically delimited structure, potentially allowing for a thorough spread of a viral vector upon a single injection.Hence, overexpressing IGF1 by a high spread vector with the control of an astrocyte-specific promoter would be a simple means to prevent astrocytes from becoming asthenic.
In an outstanding reference paper, Lee and cols.studied the genetic components of the human GFAP promoter and synthesized a truncated version (gfaABC1D) able to drive strong Green Fluorescent Protein (GFP) expression in astrocytes, particularly active for ventral and dorsal hippocampus (Lee et al., 2008).The single-stranded Adeno-Associated vector (AAV) can harbor 4.7 kb of transgenic material.Due to the short sequence of the gfaABC1D promoter, it is possible to accommodate two genes in a single vector separated by an internal ribosome entry site (IRES) sequence.Consequently, we constructed a bicistronic AAV transfer plasmid harboring the gfaABC1D promoter, followed by the IGF1, IRES and TdTomato (TOM) sequences.As a control, IGF1 was replaced by Green Fluorescent protein (GFP).We chose AAV serotype 9 capsid, since it allows for a robust spread and cell entering in the CNS (Issa et al., 2023), and produced the corresponding vectors AAV9-(gfaABC1D)-IGF1-ires-TOM together with its control AAV9-(gfaABC1D)-GFP-ires-TOM.Importantly, TOM (+) reporter endogenous fluorescence would allow for identification of transduced cells.
We posited that a single AAV9 injection would suffice to evenly transduce hippocampal astrocytes along the rostrocaudal axis of this brain region.Thus, astrocyte neuroprotective factors would be evenly secreted throughout the hippocampus in such model.In this study, we present a simple approach to overexpress two genes selectively in hippocampal astrocytes by a single AAV9 injection.
We raise the hypothesis that neurodegeneration onset develops in individuals with asthenic astrocytes.Hence, overexpressing neurotrophic factors from astrocytes would protect the brain from this process.To test our hypothesis, our approach was to overexpress the neuroprotective agent IGF1 in the hippocampal astrocytes by injecting the AAV9-(gfaABC1D)-IGF1-ires-TOM, and to subsequently study behavior tasks depending on hippocampus or other brain regions in the context of the STZ rat model, where neurodegeneration onset takes place when injecting STZ.

Animals
The experiments were conducted on Sprague Dawley rats weighing 260 ± 50 g (INIBIOLP, School of Medical Sciences, National University of La Plata, Argentina).Animals were housed in a temperaturecontrolled room (22 ± 2 • C) on a 12:12 h light/dark cycle with food and water available ad libitum (3 animals/cage).All the experiments were performed according to the Animal Welfare Guidelines of NIH (INIBIOLP's Animal Welfare Assurance No A5647-01) and were approved by our institutional Committee for the Care and Use of Laboratory Animals (Protocol # P01-02-2021).

Molecular cloning
DNA sequences were cloned into the pZac2.1 gfaABC1D-tdTomato AAV transfer plasmid (#44332 Addgene).The plasmid was digested with NheI and EcoRI restriction enzymes and the larger fragment was gel extracted and subsequently used as vector backbone.Afterwards, IGF1 (or GFP) and the IRES sequences were cloned upstream of the TOM sequence.For this aim, the first gene and the IRES sequences were PCR amplified from source plasmids with compatible primers and the 3 DNA fragments were HIFI-assembled (#E2621 New England Biolabs).
The coordinates relative to Bregma were: -3.8 mm AP, 2 mm lateral, and -3.2 mm ventral (Paxinos, 2018).For every time point, one rat was injected bilaterally and the other one unilaterally (N = 3 hippocampi per time point).According to the group, at 1, 2 or 3 weeks, the animals were placed under deep anesthesia and perfused with phosphate-buffered paraformaldehyde 4% (pH 7.4) fixative.The brains were removed and stored in paraformaldehyde 4% (pH 7.4) overnight (4 • C).Finally, brains were maintained in cryoprotectant solution (30% ethylene glycol, 30% sucrose, in PB 0.05 M) at − 20 • C until use.The rat injected bilaterally of the 2-weeks group was used for the photos in Fig. 1B.

Assessment of IGF1 by RT-qPCR
Three 2-month-old rats were randomly selected and used for this experiment.One animal was bilaterally injected in the hippocampus with AAV9-(GFAP)-GFP-TOM, another rat was injected in the right hippocampus with that vector and in the left hippocampus with AAV9-(GFAP)-IGF1-TOM.The remaining animal was bilaterally injected in the hippocampus with the IGF1 vector.With this approach there were 3 hippocampi injected with every virus (N = 3).The injections were performed as described in 2.4.Four weeks later, the animals were euthanized by rapid decapitation and their hippocampi dissected out and snap-frozen until use.For RNA extraction, the hippocampi were homogenized in TRIzol Reagent (#15596026 Thermo Fisher Scientific), and the RNA isolated according to the manufacturer's instructions.

Stereotaxic surgeries and design of the main experiment
Thirty-three 2-month-old animals were randomly allocated to four experimental groups: SHAM, STZ, GFP (N = 8 for every group), and IGF1 (N = 9).On the first day, rats were anesthetized with ketamine hydrochloride (90 mg/kg; ip) plus xylazine (8 mg/kg; ip) and submitted to bilateral intrahippocampal injection at a stereotaxic frame.Animals received artificial cerebrospinal fluid (aCSF: 120 mM NaCl, 3 mM KCl, 1.15 mM CaCl 2 , 0.8 mM MgCl 2 , 27 mM NaHCO 3 , and 0.33 mM NaH 2 PO 4 , pH 7.4) at 2 μl/side (SHAM and STZ groups) or 2 × 10 13 GC in 2 μl of AAV9 expressing GFP (GFP group) or IGF1 (IGF1 group).Hippocampal coordinates to Bregma targeting the Stratum Radiatum (SR) were as follows: -3.8 mm AP, ± 2 mm lateral, and -3.2 mm ventral (Paxinos, 2018).Four weeks later, an animal randomly selected from the IGF1 group was perfused and processed as described in 2.4 to take the photographs of Fig. 2B-C.After that point, every group had N = 8 rats.Then, animals were anesthetized as previously described and submitted to a second stereotaxic surgery for bilateral ICV injection.The SHAM group received aCSF, 2 μl/ventricle, whereas STZ, GFP and IGF1 animals received STZ (Sigma-Aldrich, CAS#18883-66-4) at a dose of 3 mg/kg (Zappa Villar et al., 2018), typically resulting in an injection of 4-6 μl, depending on the animal weight.ICV coordinates were as follows: -0.9 mm AP, ± 1.5 mm lateral, and -4.5 mm ventral.Three weeks later animals were submitted to the behavioral tests.

Open field test
Exploratory and anxiety-associated behaviors were analyzed by the open field (OF) test.The arena consisted in a square box (65 × 45 × 65 cm; W x H x D) divided into 16 equal squares on the floor.Animals were submitted to explore this environment for 5 min.As behavioral readouts, we analyzed distance traveled, number of entries and time spent in the central square.After each performance, the arena was cleaned with ethanol 10% to avoid olfactory cues.

Marble burying test
The marble burying (MB) test relies on the observation that rodents tend to bury non-aversive objects in their bedding, a species-typical behavior (Poling et al., 1981).To perform this test, rats were introduced into a clean housing cage with fresh hardwood chips.Within this environment, a set of 16 glass marbles, thoroughly cleaned and deodorized beforehand, was arranged.Following a 30-min interval, animals were returned to their original cages, and marbles were considered buried if at least two-thirds of their surface area was covered by bedding.

Barnes maze test
Spatial learning and memory in rodents are mainly associated with hippocampal function (Squire et al., 1993).We performed a modified Barnes Maze (BM) protocol previously documented (Pardo et al., 2016;Zappa Villar et al., 2021).Briefly, animals were placed in an elevated circular platform (120 cm diameter) with 20 holes at the periphery, one of which was connected to an escape box, named as hole 0 for graphical normalized representation purposes, the remaining holes are numbered 1 to 10 clockwise, and − 1 to − 9 counterclockwise.In the acquisition trials (ATs), animals were exposed to the platform for 2 min or until they found the escape box.The arena was surrounded by visual clues, and during the task animals received light and noise as escape stimuli.After six repeated ATs (2 ATs/day), which is the learning stage, animals were confronted to the probe trial (PT), in which the escape box was removed.In this last trial, we assessed the goal sector exploration as the spatial reference memory of the animals.ATs and PT were recorded for subsequent offline analysis, and after each individual testing, the platform was cleaned with ethanol 10% to avoid olfactory cues.A hole exploration was considered as a rat introduced its head into a hole and passed through the plane of the table.As behavioral readouts, the following parameters were assessed: -Latency (ATs): time (in seconds) spent by an animal from its release on the platform until it enters the escape box.
-Explorations in goal sector (PT): Exploration of escape hole and those immediately adjacent (holes 1 and − 1) during the PT.
The Any Maze Software (v7.1 Stoelting Co) was used to get information from the OF and to produce group average heatmaps in the BM for the PT.

Novel object recognition test
One day following habituation to the arena, the same as that used for the OF, the rats were exposed to two identical objects referred to as familiar objects (FO) for 5 min.After 90 min from the first exposure to the FO, animals were reintroduced to the same arena for another 5-min session, during which one of the items was replaced by a novel object (NO) with similar dimensions but a different shape and color.In this trial, we evaluated exploration preference, based on the premise that animals remembering the FO would exhibit a preference for exploring the NO.Following each individual trial, both objects and the box were cleaned with 10% ethanol.As behavioral readouts, we measured the time spent exploring each object.Then, a discrimination index (D.I.) was calculated by taking the difference in exploration time between NO and FO and dividing this value by the total exploration time: D.I. = (NOt − FOt)/(NOt + FOt), where t = time, a positive score indicates more time exploring the NO, a negative score indicates more time exploring the FO, and a score of zero indicates a neutral preference.Finally, the D.I. values for every group were tested against the hypothetical value D.I. = 0 by one-sample hypothesis testing (Zappa Villar et al., 2021).
Performance at the OF, NOR and BM tests was recorded using a Logitech C922 camera.

Brain processing for the main experiment
Eight weeks after the beginning of the experiment, rats were euthanized by rapid decapitation.Left brain hemispheres were fixed in 4% paraformaldehyde in PBS overnight at 4 • C. Afterwards, paraformaldehyde solution was replaced by 30% sucrose-PB 0.05 M solution for 24 h.This solution was replaced by cryoprotectant solution, and  brains were kept at − 20 • C until use.Finally, the brains were cut coronally in 40-μm-thick sections with a vibratome (Leica) at a 1 in 12 section sampling regime.Subsequently, the section sampling was used as 1/12th or sections were combined to obtain 1/6th series.The sections were stored in cryoprotectant solution at − 20 • C until staining.

Immunohistochemistry
Brain sections were submitted either to 3,3'-Diaminobenzidine (DAB) or immunofluorescence staining by the free-floating method.For every incubation, PBS (pH 7.4) was used as the solution buffer and the sections were washed 3 times for 15 min in PBS between every incubation step.For the DAB protocol, the sections were quenched for 1 h in 3% H 2 O 2 , 10% methanol solution.Then, the sections were blocked in blocking solution (0.25% Triton X-100, 5% secondary antibody species serum) for 1 h.Afterwards the sections were incubated with primary antibody diluted in blocking solution for 48 h at 4 • C.Then, the sections were incubated with biotinylated secondary antibody diluted in blocking solution for 2 h and with Avidin-Biotin complex (#PK6100 Vector Laboratories) for 1 h.Finally, the signal was revealed by DAB Peroxidase Substrate Kit (#SK-4100, Vector Laboratories) according to the manufacturer's instructions.
For the immunofluorescence procedure, the sections were blocked in blocking solution (0.25% Triton X-100, 5% donkey serum) for 1 h and incubated with primary antibodies diluted in blocking solution for 48 h.Afterwards, the sections were incubated with Alexa-Fluor secondary antibodies diluted in blocking solution for 2 h.Finally, the sections were incubated for 5 min with DAPI.Antibodies are listed in Table 2.

Microglial cells analysis
From a 1-in-6 section sampling, to estimate microglial cell density, 3 photos at 60× magnification were acquired in the SR of brain sections spanning the hippocampus.The 3 microscopic fields were located immediately below the CA1 neuron layer peak, one field at the center, one to the left and the remaining one to the right.We chose this region since it is the center point for AAV injection.We produced a composite photo capturing all cells in the field.Then we superimposed a counting frame square of 2.56 × 10 5 μm 2 and manually counted the cells within.Iba1(+) cells were classified as reactive or non-reactive according to previously described criteria (Diz-Chaves et al., 2012;Pardo et al., 2017).Finally, cell density was estimated by dividing the count by the sampled volume.

Astrocyte analysis
To study branching complexity, GFAP (+) astrocytes were submitted to the Sholl analysis.For this aim, we adapted confocal images to an approach previously implemented on DAB stained sections (Pardo et al., 2016;Zappa Villar et al., 2018).From a 1-in-12 section sampling, we picked hippocampal sections, performed immunofluorescence for GFAP with Alexa Fluor 647 as secondary antibody, and acquired confocal images at 63× on a Z-stack in the SR.We chose to analyze astrocytes in this region since the AAVs were injected there and cells could be individually reconstructed.Then, 20 astrocytes were selected per animal based on maximum intensity projection images.The criteria were that the whole cell must be in the microscopic field, and unambiguously differentiated from neighbor cells.In the case of the SHAM and STZ groups, GFAP (+) cells meeting the criteria were selected, whereas for the GFP and IGF1 groups, only TOM (+) transduced cells were identified based on their endogenous red fluorescence, and then reconstructed based on the GFAP channel signal.
Single astrocytes were manually reconstructed on calibrated images and Sholl analysis performed on them by the ImageJ Sholl Plugin (http://ghoshlab.org/software/).The number of process intersections per ring (i), an index of branching complexity, was computed.Afterwards, the total length of the processes was estimated by the sum of the i values for each ring multiplied by 5 μm (the distance between the concentric rings).Finally, the length of astrocyte processes and their branching complexity at every distance from the soma were averaged for every rat, and these output data were used for the statistical analysis between groups, as previously described (Pardo et al., 2016;Zappa Villar et al., 2018).

Bright field microscopy
Bright field microscopic images from Figs. 1 and 2 were acquired with an Olympus BX61VS microscope, using the Olympus VS-ASW 2.9 software.The images for Fig. 6 were taken with an Olympus BX-51 microscope attached to an Olympus DP70 CCD video camera.

Histological analysis of injected hippocampi for the time course study
The brains were cut coronally in 40-μm-thick sections with a vibratome (Leica) at a 1-in-6 section sampling regime.A small cut was made in the right cortex to differentiate the hemispheres when mounting onto the microscopic slides.To study GFP expression levels at different time points, a series from the 1/6th collected coronal brain sections was randomly used, from where 3 contiguous sections were selected: the section of the injection, one immediately anterior and one posterior.
With this methodology, we studied GFP expression in a 720 μm long rostrocaudal stretch of the brain, encompassing the injection site.The sections were DAB stained for GFP and low magnification images with high resolution (3.47 μm/pixel) were acquired and processed in the ImageJ software (v 2.14).The plugin for color deconvolution was used with the H DAB vector.Afterwards, a binary image was created by adjusting the threshold of the signal (see Fig. 1F).The hippocampi were manually delineated as a region of interest (ROI) and the immunoreactive area within each hippocampus was measured.Thus, from every hippocampus, the transduced area was calculated in 3 contiguous sections, and the transduced volume was estimated based on the Cavalieri principle (Morel et al., 2015) with the formula V = t x ΣA i , where t is the distance between sections (6 × 40 μm) and ΣA i is the sum of the 3 immunoreactive areas for that hippocampus.

Laser scanning confocal microscopy
Confocal microscopic images were acquired using a Leica SP8 microscope.Images were captured using a HyD detector and always with the lasers set to be activated in sequential mode to avoid serial excitation.Solid-state lasers at 405, 448, 552, and 650 nm wavelengths were used to excite their respective fluorophores.A pinhole of 1 AU was always retained during image acquisition.Leica objective 63×/1.40 was used during imaging acquisition.Surface render images were created using a 3D module in the SP8 Leica software from z-stack images acquired with a 63× objective at a resolution of 1024 × 1024-pixel size.

Data analysis and statistics
Immunoreactive area assessment, behavior data analysis, astrocyte analysis and microglial quantification were done by a researcher blinded to the rat group.Data were compiled and analyzed with the GraphPad Prism software (v8).One-way ANOVA followed by Tukey's post hoc test was used for the time course experiment and most of the data of the main study to compare the experimental groups.For the BM latency plot and Sholl analysis Two-Way RM ANOVA followed by Tukey's post hoc test were used.Student's t-test was used to analyze IGF1 expression from qPCR data, and one-sample t-test was used to analyze NOR D.I. against a value D.I. = 0. Pearson correlation was used to perform the correlation study between astrocyte length and Iba1 reactive cells.When the data did not meet the conditions for the mentioned tests, non-parametric tests were conducted as described in the results section for every outcome.Criteria for significant differences were set at the 95% probability level.Data are presented as bar plots showing mean ± SD, with superimposed scatter plot, except for the NOR data, in this case, since D.I. can be positive or negative, mean ± SD with superimposed scatter plot is shown.

An AAV9 bicistronic vector expressing fluorescent proteins in astrocytes along the hippocampal rostrocaudal axis
We constructed an AAV serotype 9 coding for GFP and TOM separated by the IRES sequence under the control of the astrocyte specific promoter gfaABC1D, hereunder named AAV9-(GFAP)-GFP-TOM.We bilaterally injected a rat with the mentioned AAV in the hippocampus and euthanized the animal 2 weeks after the surgery.We observed that even though the injection was performed on the AP coordinate − 3.8, the virus thoroughly spread rostrocaudally until anterior and posterior coordinates − 1.8 and − 5.3, respectively, thus covering hippocampal rerelevant for memory formation and retrieval to a great extent (Fig. 1A-B).Transduced cells expressed GFP and TOM (Fig. 1C-D).To determine the time for transgene expression plateau, we performed a time course study by injecting rats in the hippocampus and euthanizing the animals 1, 2, or 3 weeks later.The quantification of immunoreactive area and estimation of transduced volume revealed that viral expression is significantly higher at 2 and 3 weeks compared to 1 week after injection.Though not statistically different, we observed that at 3 weeks the expression was higher than at the previous time point (One-way ANOVA, F(2,6) = 10.42,p = 0.011, Tukey's post hoc test p = 0.045 and 0.010 for 1-week vs. 2-weeks and 1-week vs. 3-weeks comparisons, respectively, and p = 0.46 for the 2-weeks vs. 3-weeks comparison) (Fig. 1E-F).In summary, the AAV9-(GFAP)-GFP-TOM vector exerts a thorough hippocampal expression of both transgenes in the hippocampus at 3 weeks.

An AAV9 vector overexpressing IGF1 and the reporter TOM in hippocampal astrocytes
We posited that a bicistronic AAV overexpressing a neuroprotective agent like IGF1 in astrocytes would achieve a hippocampal neuroprotective milieu to prevent neurodegeneration.Transduced cells would be visualized by endogenous TOM fluorescence.Hence, we constructed an AAV serotype 9 coding for IGF1 and TOM separated by the IRES sequence under the astrocyte specific promoter gfaABC1D, hereunder named AAV9-(GFAP)-IGF1-TOM, and let it express for 4 weeks (Fig. 2A), reasoning that one week after the maximum observed immunoreactive volume would be an adequate time to secure astrocyte-mediated neuroprotection.We observed a strong TOM signal from transduced GFAP (+) astrocytes in the hippocampus (Fig. 2 B-C and Suppl Video 1).To corroborate IGF1 expression, we injected rats in the hippocampus with AAV9-(GFAP)-GFP-TOM or AAV9-(GFAP)-IGF1-TOM and assessed IGF1 levels by RT-qPCR.The IGF1 group displayed significantly higher IGF1 transgene levels (Unpaired t-test, F(2,2) = 2.603, p-value = 0.024) (Fig. 2D).We concluded that the rat model obtained 4 weeks after hippocampal AAV9-(GFAP)-IGF1-TOM injection is a preventive astrocyte neuroprotection paradigm to probe against neurodegeneration.

Hippocampal IGF1 overexpression in astrocytes does not prevent STZ-mediated impairment in an integrative behavior task
We probed our neuroprotective paradigm against ICV STZ neurodegeneration, a toxic challenge for the brain involving behavioral deficit and reactive microglia-mediated neuroinflammation (Zappa Villar et al., 2018).Thus, animals were injected with AAVs or aCSF in the hippocampus, and after 4 weeks they were challenged with ICV aCSF or STZ, to be subsequently tested for behavior 3 weeks after the ICV injection (Fig. 3A, see M&M 2.6).First, the rats were challenged in the OF task.By assessing distance covered, we excluded any locomotor impairment in the animals (One-way ANOVA, F(3,28) = 0.48, p = 0.70).Then, we quantified the time spent by the animals in the center region of the arena and the number of entries, we did not find any difference between the groups (Kruskal-Wallis test, H(3) = 2.65, p = 0.45 and One-Way ANOVA F(3,28) = 0.17, p = 0.92, for time spent in the center and entries, respectively), which suggests the STZ injected animals did not suffer from anxiety (Fig. 3B).Thus, the animals did not show any significant differences in behavior at the OF.
The MB paradigm tests for species-typical behavior.As such, this involves several brain regions.We asked whether IGF1 astrocyte overexpression in the hippocampus would prevent STZ-associated deficit in this test.We recorded a significant decrease in the number of buried marbles for the STZ group, compared to SHAM, albeit this was not reverted by GFP or IGF1 (Kruskal-Wallis test, H(3) = 10.21,p = 0.017, Dunn's post hoc test p = 0.018 for SHAM vs. STZ) (Fig. 3C).This shows that IGF1 treatment did not prevent cognitive impairment in complex tasks involving several brain regions.

Spatial learning and memory in hippocampus-dependent tasks improved by astrocyte overexpression of IGF1
Next, we reasoned that even though astrocyte IGF1 neuroprotection did not prevent behavior impairment in complex tasks such as MB, it would do on behavior paradigms largely involving hippocampal function, such as the BM and the NOR paradigms.
Subsequently, the animals were submitted to the NOR test.When rats were exposed to two identical objects, no preference was detected for any of them (D.I. was not significantly different from 0 for any group).However, when the animals were presented the novel object, we found that SHAM and IGF1 groups displayed a significantly high D.I., whereas D.I. of STZ and GFP groups did not significantly differ from what would be expected for an equal exploration of both objects (One sample Student's t-test against D.I. = 0, t = 5.96, 2.00, and 2.77, p < 0.001, 0.084 and 0.028 for SHAM, STZ, and IGF1 groups, respectively, one sample Wilcoxon test against D.I. = 0, W = 9.0, p = 0.25 for the GFP group) (Fig. 4D).

AAV9 transduction impacts on astrocyte length
The STZ model displays astrocyte shortening in the SR.Since we injected the viral vectors in this region, we hypothesized that IGF1 cells would maintain an undisturbed average length.Thus, from confocal images of brain sections stained for GFAP, we reconstructed astrocytes residing in the SR.In the SHAM and STZ groups we reconstructed GFAP (+) cells, whereas in the GFP and IGF1 groups we analyzed GFAP (+) TOM (+) cells only (Fig. 5A-C).Sholl analysis showed that SHAM astrocytes displayed significantly higher branching at 15 μm radius compared to STZ and at 20 μm when comparing it with STZ and GFP animals.Interestingly, the GFP group had a significantly higher branching at 10 μm than the STZ group, and gradually decayed to match STZ levels at longer radii.On the other side, the IGF1 group did not differ significantly from any other group (Two-way RM ANOVA: F interaction (24, 224) = 1.94, p = 0.0070; F radius (2.060, 57.69) = 587.1,p < 0.001; F experimental group (3, 28) =3.40, p = 0.031, Tukey's post hoc test for 10 μm p = 0.020 for STZ vs. GFP, Tukey's post hoc test for 15 μm p = 0.024 for SHAM vs. STZ, Tukey's post hoc test for 20 μm p = 0.017 and p = 0.0018 for SHAM vs. STZ and GFP groups, respectively) (Fig. 5D).The estimation of cell length revealed that astrocytes of the SHAM group were significantly longer than STZ cells.On the other side, astrocyte length of GFP and IGF1 animals did not differ significantly from any group (Kruskal Wallis for astrocytes length, H (3) =9.26, p = 0.026, Dunn's post hoc test p = 0.031 for SHAM vs. STZ) (Fig. 5E).Hence, AAV9 transduced astrocytes resisted STZ toxicity regarding cell length, albeit they did not maintain SHAM levels.

Overexpression of IGF1 by astrocytes in the hippocampus prevents microglial reactivity in the context of neurodegeneration
The STZ prompts neuroinflammation in the hippocampus, manifested in high microglial reactivity (Zappa Villar et al., 2018).Given documented literature on IGF1-mediated anti-inflammatory action (Falomir-Lockhart et al., 2022), we hypothesized that astrocytes overexpressing IGF1 would prevent microglial reactivity in the STZ model.Thus, we DAB stained for the microglial marker Iba1 and quantified microglial cells in the SR and classified them as non-reactive or reactive based on morphology.Total Iba1 cell density did not differ among groups (One-Way ANOVA F(3,28) = 1.64, p = 0.2).However, there was an overt increase in hippocampal microglial reactivity upon STZ injection in the STZ and GFP groups, whereas no significant difference was found between the IGF1 animals and their SHAM counterparts (Kruskal-Wallis test for Iba1 reactive cells percentage, H (3) =14.00, p = 0.0029, Dunn's post hoc test p = 0.025 and 0.0054 for SHAM vs. STZ and GFP groups, respectively) (Fig. 6A-B).We observed that the IGF1 group did not display significantly fewer reactive microglia because of a rat with 33% reactive Iba1(+) cells, which was indeed detected as outlier by the interquartile range method.Of note, this animal displayed an average astrocyte length of 207 μm, which was the second largest in the group.This finding suggested that there could be a biological relation between astrocyte length in transduced cells and microglial reactivity in the injected region.To investigate this, we performed separate correlation analysis for GFP and IGF1 animals between reactive Iba1(+) percentage and astrocyte length.We found a positive significant correlation for the GFP animals and no correlation for IGF1 animals (Pearson correlation coefficient 0.90, p = 0.0021 for GFP animals, Spearman correlation coefficient 0.28, p = 0.49 for the IGF1 group) (Fig. 6C).In sum, these results suggest that IGF1 from transduced astrocytes prevents microglia reactivity in the hippocampus in the context of neurodegeneration; however, AAV9 vector injection induces a cellular response, manifested for instance as astrocyte extension, correlated to a microglial reactivity state in the hippocampus.This is counteracted when IGF1 is overexpressed.

Discussion
In previous studies, we have employed ICV injection of a RAd-IGF1 in a restorative approach to treat age-related and STZ-mediated neurodegeneration in rats (Pardo et al., 2016;Zappa Villar et al., 2021).In those works, IGF1 therapy rescued spatial memory in the case of the senile rats, and, in addition to that, restored the species-typical behavior in the STZ animals.In that approach, the vector transduced ependymal cells lining the brain ventricles, from where IGF1 was secreted to the CSF and impacted on anatomically accessible brain regions.
Astrocytes play a critical role in the CNS by maintaining the bloodbrain barrier, adjusting local blood flow, providing neurons with energy substrates, and regulating synaptic transmission and neuronal excitability (Filosa et al., 2016;Gavrilov et al., 2018;Magistretti and Allaman, 2018;Sweeney et al., 2019).Their atrophy can disrupt these functions, leading to impaired calcium signaling, compromised synaptic plasticity, and weakened blood-brain barrier integrity, ultimately fostering neurodegeneration (Liu et al., 2018;Simon et al., 2018;Tanaka et al., 2013;Tarantini et al., 2017).Given the crucial roles of astrocytes in the brain, we hypothesized that hippocampal astrocytes overexpressing IGF1 could prevent the onset of neurodegeneration.To test our hypothesis, it was crucial to produce a virus expressing transgenes specifically in hippocampal astrocytes.Our findings demonstrated that the AAV9 bicistronic system effectively targeted astrocytes throughout the rat hippocampus, as evidenced by robust expression of both GFP and TOM proteins and high IGF1 transgene levels.
In this work, we used the hippocampus of the STZ neurodegeneration model to show that astrocytes overexpressing IGF1 prevent memory deficits in tasks related to the hippocampus and alleviate microglialmediated neuroinflammation.Previous studies have explored the neuroprotective effects of IGF1 overexpression in astrocytes using similar AAV systems.For instance, Chen et al. (2019) applied IGF1-AAV transfer to astrocytes in rodents treated with kainic acid, showing reduced brain cell death and improved motor and cognitive functions (measured by the modified limb preference test and the Y-maze, respectively).This study suggests that delivering IGF1 to astrocytes could be a promising therapeutic strategy for mitigating excitotoxicity-induced brain damage and restoring motor and cognitive function (Chen et al., 2019).Our findings align with this, showing that IGF1 overexpression in hippocampal astrocytes can mitigate neurodegenerative effects and improve cognitive outcomes.Okoreeh et al. (2017) provided evidence that IGF1 overexpression via an AAV GFAP-IGF1 construct in a stroke model resulted in neuroprotection and a reduction in microglial reactivity (Okoreeh et al., 2017).Our results support these findings, as we observed that IGF1 overexpression decreased the reactive Iba1(+) population in the hippocampus.This reduction in microglial activation may contribute to the observed neuroprotective effects.
We studied astrocyte morphology using Sholl Analysis.In line with previous studies (Zappa Villar et al., 2018), we observed that the STZ group displayed significantly shorter astrocytes than the SHAM animals.We hypothesized that IGF1 astrocytes would be comparable or longer than SHAM cells, whereas GFP astrocytes would retain a STZ profile.However, GFP and IGF1 groups did not differ significantly from the STZ nor from the SHAM groups regarding astrocyte length.Interestingly, the IGF1 animals displayed larger variability in astrocyte length than the GFP cells (mean ± standard deviation, 161 ± 43.53 μm and 133 ± 10.6 μm, respectively).Since the only difference between the viruses is that GFP replaces IGF1 in the control vector, there would be a non-linear effect of IGF1 interacting with various sources of biological variability.A significant correlation between Iba1 reactive cells percentage and astrocyte length was observed for GFP animals but not for their IGF1 counterparts, suggesting astrocyte resistance to STZ-mediated shortening as a general, highly variable response to the AAV treatment.On the other hand, there is a strong effect of IGF1 overexpression in decreasing brain neuroinflammation and altering astrocyte morphology.
In a transgenic mouse model overexpressing IGF1 in astrocytes (IGF-1Tg), Madathil et al. (2013) observed increased size and thicker processes in astrocytes and higher GFAP levels after traumatic brain injury, suggesting that astrocyte-derived IGF1 exerted both autocrine and paracrine effects.In that study, IGF1 overexpression stimulated Akt phosphorylation, reduced hippocampal neurodegeneration, and improved motor and cognitive function in brain-injured mice (Madathil et al., 2013).These findings align with our results, where IGF1 prompted cognitive benefits.In a previous work on the ICV-STZ model, we recorded a significant increase in hippocampal phosphorylated Akt 18 days post RAd-IGF1 treatment (Zappa Villar et al., 2021).It was suggested that IGF1 exerts a protective action on astrocytes, bolstering their resilience against oxidative stress and contributing to its overall neuroprotective effects.This protective role is linked to the ability of IGF1 to activate the pro-survival kinase Akt within astrocytes (Dávila et al., 2016).In this regard, ICV STZ injection increased both oxidative stress (Agrawal et al., 2009;Lu et al., 2017) and inducible nitric oxide synthase (iNOS) expression (Rai et al., 2013;Zappa Villar et al., 2021) in the hippocampus, and ICV IGF1 gene transfer reduced these iNOS levels (Zappa Villar et al., 2021).Hence, we propose that IGF1 exerts a protective effect on astrocytes through autocrine and paracrine mechanisms, contributing to the resilience of these glial cells against STZinduced damage, such as oxidative stress.The Akt pathway might be partially involved in this protective effect.Activation of Akt by IGF1 appears to be crucial for preserving its cytoprotective effect in astrocytes.This mechanism may represent part of the brain defense system against injuries, playing a key role in neuroprotection and modulating neuroinflammation.Further investigation into the precise molecular mechanisms underlying IGF1 effects on Akt signaling within astrocytes could provide valuable insights into enhancing therapeutic strategies aimed at neuroprotection in various neurodegenerative conditions.
Neuroprotection by IGF1 is associated with microglial modification and reduced neuroinflammation (Serhan et al., 2020).Other interesting studies show that IGF1 overexpression has significant potential in modulating neuroinflammation in neurological disorders.In amyotrophic lateral sclerosis (ALS) models, AAV9-hIGF1 delivery to the subarachnoid space led to hIGF1 expression in the brain and spinal cord, reducing motor neuron loss, improving motor function, extending lifespan, and alleviating neuroinflammation (Hu et al., 2018).Similarly, in epileptic models, transgenic mice with IGF1 overexpressed specifically in neural stem cells exhibited a decrease in seizure severity and hippocampal inflammation (Wu et al., 2022).These findings align with our results, which revealed that IGF1 overexpression did not alter the overall microglial population in the hippocampus but reduced the percentage of reactive Iba1 (+) cells, suggesting that IGF1 alleviates brain inflammation in the STZ model and dampens neuroinflammatory responses in microglia.
Importantly, our initial postulation was that astrocyte-mediated neuroprotection would be mediated by their morphology, as asthenic astrocytes would lead to poor trophic assistance to neurons.However, our data showed that astrocytes overexpressing IGF1 prevented memory deficits by mechanisms independent of astrocyte length, for example, by decreasing neuroinflammation in the hippocampus.We posit that a cocktail of neuroprotective factors secreted from astrocytes would be a potent approach to prevent age-related neurodegeneration.For instance, HIFI technology allows for multiple neurotrophic factor genes to be cloned in a single AAV library.Therefore, we propose the production of a large therapeutic genetic cocktail from a single AAV9 library targeting astrocytes as a future step.

Conclusion
Preventing brain damage by overexpressing neurotrophic factors in astrocytes is an attractive approach.AAV9 vectors can, with a single injection, overexpress neuroprotective peptides, such as IGF1, in the mammalian hippocampal astrocytes.In our approach, this prevented memory deficit and alleviated microglial reactivity prompted by STZ injection.In summary, our work describes a straightforward strategy to overexpress neurotrophic factors in hippocampal astrocytes, which has huge translational potential since advancement in safe vectors and the

Fig. 1 .
Fig. 1.Sustained expression of an AAV9 bicistronic system in hippocampus.A. Experimental design for the hippocampal injection to validate gene expression.B. DAB staining images showing GFP expression in the hippocampus along the rostrocaudal axis.C. Higher magnification image showing GFP expression in the hippocampus on a DAB-stained brain coronal section.Scale bar at the left inset applies to the right one.D. High magnification image of a hippocampus injected with the vector stained by double immunofluorescence for GFP and TOM.Scale bar applies to all photos.E. Schematic of the experimental design for the time course study.F. Example of the method to make the images binary and bar plot displaying mean ± SD, with superimposed scatter plot to show the estimation of the transduced volume for the time course study.N = 3 hippocampi/group.*p < 0.05.SR: Stratum Radiatum.DH: Dentate Hilus.

Fig. 2 .
Fig. 2. Hippocampal expression of the AAV9-(GFAP)-IGF1-TOM vector.A. Schematic for the experimental design to assess the expression of the vector.B. Left: pictures of a hippocampal coronal section DAB-stained for TOM from an AAV injected animal.Scale bar at the left inset applies to the right one.Right: low magnification pictures of coronal brain sections DAB-stained for TOM in animals injected in the hippocampus with the AAV.C. High magnification images of a hippocampus injected with the vector stained by double immunofluorescence for GFAP and TOM.Scale bar applies to all photos.D. Schematic of the experimental design for the qPCR study to assess IGF1 levels in injected animals and the corresponding bar plot showing mean ± SD, with superimposed scatter plot.Data are referred to the GFP sample with lowest detection.N = 3 hippocampi/group.* p < 0.05.SR: Stratum Radiatum.DH: Dentate Hilus.

Fig. 3 .
Fig. 3. Behaviorally testing for the prevention of neurodegeneration onset by IGF1 overexpressing astrocytes in the hippocampus.A. Schematic for the design of the main experiment of this study.The behavioral test pictures represent the OF, MB, BM, and NOR tests.B -C. Bar plots showing mean ± SD, with superimposed scatter plots displaying the performance of the animals at the OF and the buried marbles for every experimental group.N = 8/group.* p < 0.05.

Fig. 4 .
Fig. 4. Memory deficit prevention by the hippocampal IGF1 gene therapy on astrocytes. A. Line plot showing group means with SD bars displaying the latency for the BM ATs.B. Depiction of the BM and bar plot showing mean ± SD, with superimposed scatter plot displaying goal sector exploration.C. Experimental group average heat map of the rats in the BM PT.Orientation of the maze is the same as in B at the left.D. Plot showing mean ± SD, with superimposed scatter plot displaying the discrimination index at the NOR paradigm at the training and test phases.N = 8/group.AT: Acquisition trial, PT: Probe trial.# p < 0.05 for STZ vs IGF1 and p < 0.01 for GFP vs IGF1.$ p < 0.05 for GFP vs IGF1.*p < 0.05, *** < p < 0.001, NS=Not significant.

Fig. 5 .
Fig. 5. Microscopic analysis of the hippocampal astrocytes.A'.High magnification image of a representative astrocyte stained for GFAP from the SHAM group and cell reconstruction.A".Image of a representative cell from the STZ group.B′.Representative image of an astrocyte from the GFP group.The cell was identified with Tom fluorescence and then reconstructed based on the morphology observed in the GFAP channel.B″.A representative image of an astrocyte from the IGF1 group.Scale bar at the end of A" applies to A'-B″.C. Depiction of the Sholl analysis using as example the cell from B″. D. Line plot showing group means with SD bars displaying the intersections of astrocytes from the experimental groups and concentric rings with radius increasing 5 μm at a time.# p < 0.05 for STZ vs GFP.$ p < 0.05 for SHAM vs STZ.& p < 0.05 for SHAM vs STZ and p < 0.01 for SHAM vs GFP.E. Bar plot showing mean ± SD, with superimposed scatter plot displaying the mean astrocyte length for the animal groups.N = 8/group.*p < 0.05, NS=Not significant.

Fig. 6 .
Fig. 6.Iba1(þ) microglial cells quantification in the hippocampus and classification according to reactivity. A. Representative images showing the hippocampus of brain coronal sections DAB stained for Iba1.Scale bars apply to every picture of the same magnification.B. Bar plots showing mean ± SD, with superimposed scatter plots displaying the total Iba1 density and the percentage of reactive cells in the hippocampal SR. C. Line plots with superimposed scatter plots illustrating the correlation between reactive microglial percentage in the SR and astrocyte length for the GFP and IGF1 groups.N = 8/group.*p < 0.05, **p < 0.01.SR: Stratum Radiatum.DH: Dentate Hilus.

Table 2
Antibody information.