IGF1 Gene Therapy Reversed Cognitive Deficits and Restored Hippocampal Alterations After Chronic Spinal Cord Injury

The hippocampus is implicated in the generation of memory and learning, processes which involve extensive neuroplasticity. The generation of hippocampal adult-born neurons is particularly regulated by glial cells of the neurogenic niche and the surrounding microenvironment. Interestingly, recent evidence has shown that spinal cord injury (SCI) in rodents leads to hippocampal neuroinflammation, neurogenesis reduction, and cognitive impairments. In this scenario, the aim of this work was to evaluate whether an adenoviral vector expressing IGF1 could reverse hippocampal alterations and cognitive deficits after chronic SCI. SCI caused neurogenesis reduction and impairments of both recognition and working memories. We also found that SCI increased the number of hypertrophic arginase-1 negative microglia concomitant with the decrease of the number of ramified surveillance microglia in the hilus, molecular layer, and subgranular zone of the dentate gyrus. RAd-IGF1 treatment restored neurogenesis and improved recognition and working memory impairments. In addition, RAd-IGF1 gene therapy modulated differentially hippocampal regions. In the hilus and molecular layer, IGF1 gene therapy recovered the number of surveillance microglia coincident with a reduction of hypertrophic microglia cell number. However, in the neurogenic niche, IGF1 reduced the number of ramified microglia and increased the number of hypertrophic microglia, which as a whole expressed arginase-1. In summary, RAd-IGF1 gene therapy might surge as a new therapeutic strategy for patients with hippocampal microglial alterations and cognitive deficits such as those with spinal cord injury and other neurodegenerative diseases.


Introduction
The hippocampus is implicated in the generation of memory and learning, processes that involve extensive neuroplasticity [1]. The subgranular zone (SGZ) of the dentate gyrus (DG) of the hippocampus is one of the brain regions where generation of new neurons continues throughout life [2]. Adult-born neurons integrate into hippocampal neural networks providing highly plastic properties to the circuit and participating in the information processing [3][4][5].
Neurogenic niches are modified by surrounding microenvironment and pathological conditions [6,7]. Indeed, neuroinflammation has been extensively reported to impair hippocampal neurogenesis [8,9]. In fact, neurogenesis decreases with aging, chronic stress, and neurodegenerative diseases, all conditions where neuroinflammation is a common landmark [8]. Microglial cells orchestrate CNS immune response, and their activation implies shortening of processes and hypertrophy of the soma [10,11], but it can be supportive or detrimental depending upon changes of microglia phenotype and the kind of released factors. In fact, pro-inflammatory microglia releases cytokines which down-regulate neural stem cell activation, neuronal differentiation, and survival [12][13][14], while anti-inflammatory microglia produces factors that support neurogenesis [15,16].
Neuroinflammation also leads to hippocampal dysfunction altering synaptic plasticity [17,18] and generating cognitive impairment in hippocampal-dependent learning such as retention [19][20][21] and spatial memories [22,23]. Therefore, the fine-tuning polarization from pro-inflammatory towards anti-inflammatory microglial phenotypes represents a potential strategy to treat neurodegenerative diseases or trauma. In this regard, the expression of arginase-1(ARG), an enzyme related to the anti-inflammatory profile, reduces contusion size after traumatic brain injury [24] and promotes retinal neuroprotection after ischemia [25].
Interestingly, most studies regarding spinal cord injury (SCI) have focused on the pathophysiological changes in the spinal cord although cognitive deficits in humans have been described in several reports [26,27]. Few groups have studied in experimental models the effect of SCI on brain regions related to cognitive function such as the hippocampus. Recent evidence has shown cognitive deficits in hippocampal dependent tasks after chronic spinal cord contusion in rodents [28,29]. Concomitantly with the cognitive impairment reported, the authors described that the rodent's brain undergoes diffuse inflammation and neurogenesis reduction after SCI [28,29]. Indeed, our laboratory has recently shown that chronic SCI not only increases the expression of pro-inflammatory cytokines and the activation of astrocytes and microglial cells in the hippocampus but also decreases neurogenesis in the SGZ [30,31].
Since rodents and humans present cognitive deficits related to hippocampal dependent tasks after SCI [28,32,33], it is important to find therapies to reverse the hippocampal abnormalities described after chronic SCI. In this regard, IGF1 is a well-known neurotrophic factor with pleiotropic beneficial effects in the CNS [34][35][36]. This neurotrophic factor increases neurogenesis in the SGZ (Yuan 2014) and polarizes microglial cells towards a neuroprotective and anti-inflammatory phenotype [36].
The administration of IGF1 as a peptide to treat patients implies a constant supply, which is not feasible to apply in the clinic. However, the use of IGF1 gene therapy has been successful in enhancing hippocampal function [37], modifying microglial reactivity [38], and improving cognitive and motor function in aged rats [37,39].
In this regard, the aim of this work was to evaluate whether IGF1 gene therapy could reverse hippocampal alterations and cognitive deficits observed in adult male rats after chronic SCI.

Spinal Cord Injury
Rats were anesthetized with a mixture of ketamine (100 mg/ kg, i.p.) and xylazine (10 mg/kg, i.p.). Once the absence of reflexes had been checked, Tramadol (10 mg/kg sc) was injected to reduce unnecessary pain, and artificial tears were applied to the eyes to prevent corneal abrasion and infection. The clip compression model has provided valuable knowledge about the pathophysiology of SCI [40,41] and produces histopathological changes that are very similar to human SCI due to a combination of contusion and compression injuries [42,43]. A moderate spinal cord compression was performed using 30-g force-closed vascular clip (Kent Scientific Corporation, Torrington, CT) applied for 2 s as previously described [30]. Briefly, after shaving and cleaning the back of the rat with ethanol and iodine, a dorsal midline incision was made in the skin, and the T8 vertebra was exposed. The apophysis, transverse process, and dorsal lamina were removed, and the spinal cord was exposed. Injuries were performed by clipping the spinal cord dorsoventrally, and after compression, the muscle and skin openings were sutured together in layers. After surgery, animals were hydrated and placed in heated blankets for 1 h. Shamoperated controls received the same protocol for laminectomy but without suffering compression. Postoperative care included a subcutaneous injection of cephalexin (20 mg/kg daily for 7 days) to prevent infections, and the bladder was manually emptied until becoming self-voiding. Rats were sacrificed 80 days post injury (dpi), and the hippocampus was processed for several determinations.

Adenoviral Vectors
We employed recombinant adenoviral vectors (RAd) as carriers to deliver either the therapeutic cDNA of IGF1 gene (RAd-IGF1) or the red fluorescent protein from Discosoma sp DsRed (RAd-DsRed) as previously described [44][45][46]. RAd-IGF-I was constructed using a variant of the two-plasmid method, and it can infect all divided and non-divided cells. The cDNA coding for the rat IGF-I gene (kindly donated by Dr. Peter Rotwein, Department of Biochemistry and Molecular Biology, Oregon Health & Science University, Portland, OR), obtained from the mRNA for the IGF-Ib precursor form was placed under the control of the mCMV promoter in order to construct the genome of the desired RAd-IGF-I. The newly generated RAd was rescued from human embryo kidney 293 cell lysates and plaque purified. It was further purified by ultracentrifugation in a CsCl gradient. Final virus stocks were titrated by a serial dilution plaque assay. RAd-IGF1 vector transduces brain ependymal cells with high efficiency releasing IGF1 to the cerebrospinal fluid [44].

Intracerebroventricular (ICV) RAd-IGF1 Injections
Sixty days post injury, rats (n = 10 per group) were anesthetized and placed in a stereotaxic apparatus. Both sham-operated controls and lesioned rats were randomly divided into three groups giving rise to 5 different experimental groups of animals. Sham: rats with mock spinal cord lesion lacking ICV injection; Sham-DsRed: sham rats which received an injection of RAd-DsRed; SCI: injured rats without ICV injection; SCI-DsRed: lesioned rats which received an injection of RAd-DsRed; and SCI-IGF1: lesioned rats injected with RAd-IGF1. Each adenovirus was injected bilaterally in the lateral ventricles. The injection was performed placing the tip of a 26 G needle fitted to a 10-mL syringe at the following coordinates relative to the bregma: − 0.8 mm anteroposterior, − 4.2 mm dorsoventral, and ± 1.5 mm mediolateral [45]. Rats received 4 µL (1 µL/min) per side of a suspension containing 10 10 plaque forming units (pfu) of the appropriate vector. Animals were sacrificed 20 days after the ICV injections.

Behavioral Assessments
Rats (n = 10 per group) were habituated to the room and behavioral chambers after receiving the ICV injection. Behavioral evaluation was performed from 70 to 80 dpi, after which rats were sacrificed.
Basso-Bresnahan-Beattie (BBB) locomotor scale: Locomotor performance was evaluated by BBB locomotor scale [47,48]. Animals were gentled to the open field during several sessions. Then, two blind to treatment trained observers scored rats for 4 min at 7, 10, 15, 21, 30, 40, 50, 60, and 75 dpi. A score of 0 points determines no movement of the hindlimbs and 21 points (the maximum score) determines normal locomotion as observed in sham operated rats.
Locomotor activity (open field test): After 80 dpi, rats (n = 10 per group) were individually placed in a chamber (80 cm × 80 cm) with black walls and floor and allowed to explore it freely for 10 min. The distance and speed traveled were recorded by computer-based Any-Maze automated video tracking system (Stoelting Co).
Novel object recognition (NOR): NOR was performed as previously described from 70 to 80 dpi [49,50]. NOR measures recognition memory based on rodent innate preference to explore novel objects. Once rats were habituated to the arena (40 cm × 40 cm square box), they were placed in the box for 5 min with two identical objects (familiar object) settled near the left and right corners (familiarization phase). Rats were allowed to explore freely, and they were removed and returned to their home cages. After 24 h, a new object (novel object) replaced one of the familiar training objects, and rats were placed in the arena for 5 min, and they were allowed to freely explore both objects (recognition phase). The identity and location of the novel object were counterbalanced across rats. The time of exploration for each object was recorded and expressed as the percentage of the total exploration time (preference %). Exploration was defined as sniffing or touching the object with the nose or forepaws. A criterion of ≥ 10 s of exploration during the familiarization phase was established for animals to be included in the statistical analysis of preference %, since low exploration times may distort encoding processes in this task [51].
Y-maze spontaneous alternation test: The Y-maze test was performed as previously described from 70 to 80 dpi [52,53]. Spontaneous alternation behavior measures spatial working memory as it comprises the tendency for rodents to alternate their choice of Y-maze arms on successive opportunities. Rats were placed at the end of one arm and allowed to freely explore the maze for 10 min. Rats entering three different arms consecutively are defined as alternation. The percentage of alternation is calculated using the following equation: total alternations/(total arm entries-2) × 100.

Histological Analysis
Tissue processing for immunohistochemistry: Rats were deeply anesthetized (n = 6 rats per group) and intracardially perfused with 0.9% NaCl, followed by ice-cold 4% paraformaldehyde (PFA). Brains were removed from the skull and incubated overnight in 4% PFA. The next day, they were cut in 50-μm-thick serial coronal sections using a vibrating microtome and were stored in a cryoprotectant solution at -20 °C until free-floating immunohistochemistry. The sampled region was the rostro-caudal extension of both halves of the DG, corresponding to Plates 53-73 of a rat brain atlas [45]. Six representative coronal sections separated at least 400 µm from each other were subjected to immunostaining with different antibodies.
Immunohistochemistry and immunofluorescence: Immunostaining was carried out in brain sections as described previously [54] using primary antibodies against Iba-1 (1:1000, rabbit-polyclonal antibody, 019-19,741 Wako) for staining microglia, ARG for non-inflammatory microglia For immunofluorescence, after the incubation with the first antibody, sections were washed with PBS and incubated with fluorochrome-conjugated secondary antibodies (1/1000) (anti-rabbit Alexa 488 and anti-goat Alexa 555, Molecular probes) in PBS for 1 h at room temperature. Finally, slices were rinsed in PBS and mounted on gelatin coated slides with Fluoromount-G (Southern Biotech) and kept in dark at 4 °C until analyzed by confocal microscopy. To avoid inter-assay variations, experiments were run in parallel, and non-specific staining was discarded by incubation of tissue without primary antibodies.

Quantitative Morphometric Analysis
After immunohistochemistry, cross sections were examined under a light microscope, at 400 × or 600 × magnification, and images were captured with a Canon G10 digital camera connected to a Zeiss Axioplan microscope. Concerning immunofluorescence, double-labeled cells were examined under a Nikon Eclipse E 800 confocal scanning laser microscope. Images were acquired using a Nikon EZC1 version 2.1 software. Image analysis was performed using Image J v1.52b, an NIH image analysis software. All slides were assessed blindly with respect to experimental groups.
Stereological DCX counting: The total number of DCX + cells was estimated according to Jure et al. and others [30,55,56]. DCX + cells were counted on every sixth section throughout the entire rostro-caudal extension of both the superior and the inferior blades of the SGZ of DG in the dorsal hippocampus. The same area and number of sections (n = 6 per animal) were studied from each experimental group. The number of DCX + cells counted in SGZ was multiplied by the sampling factor (8) to estimate the total number of DCX-labeled cells in the hippocampus.
Stereological Iba-1 counting: The quantification of the number and phenotypes of Iba-1 + microglia was assessed as previously described [30]. Briefly, the number of the three microglial morphological phenotypes (ramified, hypertrophic, and bushy) in the hilus, molecular layer, and SGZ + GCL (granular cell layer) was quantified using the optical dissector method of unbiased stereology [56]. The optical dissector had a size of 78 × 78 µm in the x-axis and y-axis with a thickness of 25 µm. A total of 30 counting frames were assessed and averaged per animal. Microglial phenotypic classification was based on the length and thickness of the projections, number of branches, and size of cell body, as described previously [57]. Iba-1 + cells with hypertrophic and bushy phenotypes were pooled in one group because both are related to microglial reactivity [10,11]. The estimated number of microglia in each phenotypic class was divided by the volume of the region of interest to obtain the cellular density expressed in cells/mm 3 .
Double positive cell counting: All images captured correspond to 10 μm z-stacks and were acquired following protocols optimized for stereological quantification and quantitative image analysis [31,58,59]. Results were expressed as the number of positive cells per mm 3 or as the percentage of double + cells related to total Iba-1 + cells or total hypertrophic/ bushy Iba-1 + cells. The hippocampal volume measured using stereological quantification remained unchanged between lesioned and sham-operated animals as previously described [30,31].

Statistical Analysis
All the statistical analyses and graphs were performed with "Prism 6.0" software. The number of animals was used as the N number for statistical analysis, a p < 0.05 was considered to be significant, and the data are presented as mean ± standard errors. One-way analysis of variance followed by Newman-Keuls post-test was used to determine statistical differences at 80 days among the different treatments. When microglial cells were analyzed, a two-way analysis of variance followed by Bonferroni post-test was used to determine statistical significances among experimental groups. When appropriate, the Pearson (r) or Spearman (rs) correlation coefficients were calculated to test the relationship between ARG-expressing microglial cells, cognitive function, and neurogenesis.

Effects of RAd-IGF1 Gene Therapy on Locomotor Performance and Cognitive Function After Chronic SCI
Previous studies have shown cognitive deficits in hippocampal-dependent tasks such as recognition and spatial memories after chronic spinal cord contusion in rodents [28,29]. Therefore, we began this study by evaluating the effect of RAd-IGF1 therapy on rat cognitive performance. However, since cognitive tests imply rat displacement, locomotor outcome of the injured rat groups was evaluated before assessing cognitive function.

Locomotor performance
BBB scale was used to determine hind limb rat locomotor performance from 7 to 80 dpi. All rats developed significant bilateral hind limb paralysis at 7 dpi, but at 15 dpi they improved their scores showing consistently uncoordinated weight supported hind limb steps (BBB score = 11). This locomotor performance remained invariable during the evaluated period (p > 0.05, Fig. 1a). Moreover, there were no statistical differences in locomotor outcome between rats with SCI and lesioned rats treated either with RAd-DsRed or RAd-IGF1 (p > 0.05, Fig. 1a).
Additionally, spontaneous locomotor activity at 80 dpi was evaluated analyzing open field locomotion and recording the distance traveled and the maximum speed during 10 min using an automated video tracking system.
There were no statistical changes in both parameters between the two sham groups (Sham and Sham-DsRed, p > 0.05) and the three injured groups (SCI, SCI-DsRed and SCI-IGF1, p > 0.05). Indeed, traveled distance and maximum speed remained unchanged in injured rats treated with RAd-IGF1 (p > 0.05, Fig. 1b, c).
These results indicated that despite the motor deficits (consistently uncoordinated steps) rats with SCI maintained the same locomotor activity, at least for 10 min, as Sham rats. Thus, injured rats were able to perform cognitive tests which implied locomotion.

Cognitive Function
As RAd-IGF1 treatment improved spatial working and recognition memories in other experimental paradigms [37,60], we decided to evaluate whether this therapy improved these memories under our experimental conditions. Spatial working memory was evaluated using the Y-Maze spontaneous alternation test. Sham-DsRed group showed approximately 75% of spontaneous alternation, indicative of functional working memory [61]. Meanwhile, SCI-DsRed group exhibited a significant reduction of this parameter compared to the Sham-DsRed group (p < 0.01, Fig. 2a). On the other hand, RAd-IGF1 treatment of injured rats recovered the percentage of spontaneous alternation (p < 0.01, Fig. 2a). The number of entries among the experimental groups was the same, which validated the obtained data (p > 0.05, Fig. 2b).
There were no statistical changes in both parameters between the two sham groups (Sham and Sham-DsRed, p > 0.05) and between SCI and SCI-DsRed rats (p > 0.05, Fig. 2a, b). In addition, recognition memory was assessed using the NOR test. SCI-DsRed rats spent significantly less time with the novel object during the recognition phase compared to Sham-DsRed (p < 0.01, Fig. 2c). On the contrary, SCI-IGF1 animals spent a longer time exploring the novel object than the familiar one (p < 0.01, Fig. 2c). Rats of all experimental groups explored the familiar object for the same period of time during the familiarization phase (p > 0.05, Fig. 2d). Since all rats had the same opportunity to explore the familiar object, the reduced percentage of preference in the SCI-DsRed group provided evidence that SCI led to recognition memory impairment.
The analysis of this test shows no statistical differences between either Sham and Sham-DsRed groups (p > 0.05) or SCI and SCI-DsRed groups (p > 0.05).

RAd-IGF1 Gene Therapy Enhanced Neurogenesis After Chronic SCI
Previous studies have shown that the reduction of adult neurogenesis impairs hippocampal dependent learning such as recognition and spatial memories [62][63][64]. Moreover, we have described that chronic SCI causes a significant reduction in the production of new-born neurons in the DG [30,31]. Since IGF1 improves recognition and spatial memories in our model and enhances hippocampal neurogenesis in others [37,65], we evaluated the production of immature neurons (DCX + cells) after RAd-IGF1 treatment in the SGZ.
As previously described [30], the number of DCX + cells decreased in SCI rats with respect to Sham animals. The number of DCX + cells remained unchanged between Sham and Sham-DsRed rats (p > 0.05, Fig. 3a). In addition, SCI-DsRed rats did not present any changes in DCX density with respect to the SCI group (p > 0.05, Fig. 0.3a). SCI caused a significant reduction in the number of DCX + cells regarding Sham groups (p < 0.01, Fig. 3a, b vs c). Notably, RAd-IGF1 injection restored the number of DCX + cells reaching the values obtained in the Sham groups (p < 0.05, Fig. 3a, c vs  d). Figure 3b shows representative images of immunohistochemistry for DCX in the SGZ of the DG.
Given that there were no statistical differences between Sham and Sham-DsRed and between SCI and SCI-DsRed to comply with the Institutional Animal Care and Use Fig. 2 RAd-IGF1 gene therapy improved cognitive performance after chronic SCI. a-b Y-maze test. a Percentage of spontaneous alternation. b Number of total entrances in the three different arms. c-d Novel object recognition test (NOR). c Percentage of preference. d Time spent with the familiar object in the familiarization phase. The group acronyms are the same as defined in Fig. 1. All values are expressed as mean ± SEM. N = 10 rats per group. **p < 0.01 vs Sham and Sham-DsRed, + + p < 0.01 vs SCI and SCI-DsRed. One-way ANOVA followed by Newman-Keuls post-test Committee (IACUC), we decided to omit the Sham and SCI groups in the following experiments.

RAd-IGF1 Gene Therapy Modified Microglial Response After Chronic SCI
There is compelling evidence showing that hippocampal neuroinflammation decreases both recognition and spatial memories [19]- [23]. Our laboratory has recently described glial long-term activation in the hippocampus after SCI [30] [31]. As RAd-IGF1 treatment modifies microglial number and reactivity in aged rats [38], we decided to investigate whether RAd-IGF1 therapy could reverse microglial changes in the DG after chronic SCI.
Surveillance microglia presents ramified cellular morphology, while activated forms exhibit cellular hypertrophic or bushy morphology [10,11]. Figure 4 shows representative images of Iba-1 immunohistochemistry indicating the morphological criteria used for microglial classification as previously described [66]. Iba-1 + cells with small cell bodies and elongated thin projections were considered as ramified microglial cells (Fig. 4a). Cells with large cell bodies and thicker projections were identified as hypertrophic microglial cells (Fig. 4b). Finally, cells with enlarged cell bodies with short or no processes were considered as bushy microglial cells (Fig. 4c). The activated hypertrophic and bushy phenotypes were pooled in one group called hypertrophic microglia as hypertrophic cells are much more abundant than bushy cells.
The total number of Iba-1 + microglia remained unchanged across treatments in all hippocampal studied regions (p > 0.05, Fig. 4j-l). In the hilus and ML, there was a significant decrease in the number of ramified microglial cells in the SCI-DsRed with regard to the Sham-DsRed group (Fig. 4j, k, p < 0.05). This reduction coincided with an increase in the number of active or hypertrophic microglia (Fig. 4j, k, p < 0.05). Notably, RAd-IGF1 therapy restored the distribution of microglial phenotypes after injury, enhancing the number of ramified microglia and decreasing the number of hypertrophic cells in SCI-IGF1 rats (p < 0.05, Fig. 4j, k).
Regarding the SGZ and granular cell layer (GCL), ramified microglial cells decreased their density, while activated microglia increased their number in SCI-DsRed rats (p < 0.01, Fig. 4l). Peculiarly, SCI-IGF1 animals presented an even lower density of ramified cells with regard to the Sham-DsRed and SCI-DsRed groups (p < 0.01, Fig. 4l). Furthermore, RAd-IGF1 treatment enhanced the number of hypertrophic cells reaching higher values than those observed in the Sham-DsRed and SCI-DsRed groups (p < 0.01, Fig. 4l).
Representative images of Iba-1 immunohistochemistry are also shown in Fig. 4 in which white arrows indicate ramified microglia and black arrows hypertrophic ones. Images illustrate the aforementioned results.
To further study microglial response, we analyzed the number of cells which also expressed ARG, an enzyme which is normally associated with a non-inflammatory microglial profile. Colocalization studies of Iba-1 and Representative immunohistochemistry for DCX showing immature neurons 80 days post injury (dpi). The group acronyms are the same as defined in Fig. 1. All values are expressed as mean ± SEM. N = 6 rats per group, **p < 0.01 vs Sham and Sham-DsRed, + p < 0.05 vs SCI and SCI-DsRed. One-way ANOVA followed by Newman-Keuls post-test. Scale bar is 25 µm ARG revealed that the percentage of ARG-expressing microglia (Iba-1 + , ARG + cells) respect to total Iba-1 + cells decreased in SCI-DsRed compared to Sham-DsRed rats in the hilus (p < 0.01, Fig. 5a), ML (p < 0.01, Fig. 5b ), and SGZ + GCL (p < 0.05, Fig. 5c ). However, SCI-IGF1 rats showed the opposite effect, presenting a higher percentage of double-positive cells with respect to SCI-DsRed animals in all regions (p < 0.001 for hilus and ML, p < 0.05 for SGZ + GCL, Fig. 5a-c). Figure 5 shows representative images of the immunofluorescence for Iba-1 and ARG in the hilus (Fig. 5 d-f) and SGZ + GCL (Fig. 5g-i). In Sham-DsRed rats more ARG + microglial cells can be appreciated (yellow arrows) than in SCI-DsRed animals both in the hilus (Fig. 5 d vs e) and SGZ + GCL (Fig. 5g vs h). ARGmicroglial cells predominated in the in SCI-DsRed group (Fig. 5e, h white arrows). On the other hand, the number of ARGexpressing microglia (yellow arrows) was restored in SCI-IGF1 animals both in the hilus (Fig. 5f) and SGZ + GCL (Fig. 5 i).
Furthermore, the number of Iba-1 + ARGhypertrophic cells increased, while the number of double-positive hypertrophic cells decreased in the SCI-DsRed with respect to Sham-DsRed animals in the hilus (p < 0.001 for Iba-1 + ARG cells and p < 0.05 for double-positive cells, Fig. 6a) and the ML (p < 0.001 for Iba-1 + ARGcells and p < 0.01 for double-positive cells, Fig. 6b). RAd-IGF1 treatment reversed this distribution and restored the values to those obtained in the Sham-DsRed group in the hilus (p < 0.001 for Iba-1 + ARGcells and p < 0.01 for double-positive cells, Fig. 6a) and the ML (p < 0.001 for Iba-1 + ARGcells and p < 0.05 for double-positive cells, Fig. 6b). Regarding the SGZ + GCL, Iba-1 + ARGhypertrophic microglia increased their number after SCI (p < 0.05, Fig. 6c) while RAd-IGF1 treatment decreased the number of Iba-1 + ARGcells (p < 0.05, Fig. 6c) and increased the number of Iba-1 + ARG + hypertrophic microglial cells reaching values even higher than those obtained in the Sham-DsRed group (p < 0.01 vs SCI-DsRed and p < 0.01 vs Sham-DsRed, Fig. 6c).
In summary, the analysis of microglial population showed that there was a decrease in the number of ramified microglia concomitant with an increase in the number of ARG-hypertrophic microglial cells after SCI in all studied regions. However, RAd-IGF1 therapy showed a region dependent effect in the regulation of microglial activity. In the hilus and ML, RAd-IGF1 therapy restored the distribution of microglial phenotypes, increasing the number of ramified ARG + cells. On the other hand, in the SGZ + GCL, RAd-IGF1 injection increased the number of ARG + hypertrophic microglia.
Finally, correlation analyses were performed to explore the possible relationships between ARG-expressing microglial cells and cognitive function. Linear correlation analysis showed that cognitive function (spatial working memory and recognition memory) and ARG-expressing microglia exhibited a significant positive correlation (Pearson r 2 = 0.822; Spearman rs = 0.847 p < 0.001 for Fig. 7A and Pearson r 2 = 0.760; Spearman rs = 0.7587 p < 0.001 for Fig. 7B ). Figure 7 shows that values corresponding to alternation and preference percentages of Sham-DsRed and SCI-IGF1 animals were superimposed and coincided with higher values of Iba1 + Arg + cells. Instead, the values corresponding to Sham-DsRed rats did not overlap with those obtained by the other groups and coincided with the lower values of doublepositive microglia.

Discussion
This study demonstrates that IGF1 gene therapy reversed permanent hippocampal alterations in the chronic phase after SCI. Our results show that IGF1 gene therapy not only improved cognitive deficits, which involve Fig. 4 RAd-IGF1 gene therapy modified microglial morphology after chronic SCI. a-c Representative Iba-1 immunohistochemical images displaying different microglial morphologies: ramified (a), hypertrophic (b), and bushy (c). d-i Representative images of Iba-1 immunohistochemistry in Sham-DsRed (d,g), SCI-DsRed (e,h), and SCI-IGF1 (f,i) rats in the Hilus (d-f) and SGZ + GCL (subgranular zone and granular cell layer) (g-i). White arrows point out ramified microglia (d,f,g,h) and black arrows show hypertrophic microglia (d,e,f,g,h,i). j-l Stereological quantitative assessment of microglial phenotypes among the experimental groups in the hilus (j), molecular layer (k), and SGZ + GCL (l). The group acronyms are the same as defined in Fig. 1. All values are expressed as mean ± SEM. N = 6 rats per group, *p < 0.05 vs Sham-DsRed, + p < 0.05 and + + p < 0.01 vs SCI-DsRed. Two-way ANOVA followed by Newman-Keuls post-test. Scale bar is 10 µm ◂ hippocampal-dependent learning after SCI but also modulated microglial cells and neurogenesis.
There is compelling evidence demonstrating that ICV IGF1 gene therapy is an effective route of IGF1 administration. RAd-IGF1 vector transduces brain ependymal cells with high efficiency releasing IGF1 to the cerebrospinal fluid [44]. We chose the ependymal route for IGF1 gene delivery to the brain based on the fact that most peptides used by the CNS come from the circulation [67,68] are actively transported through the choroid plexus to the CSF, from where they reach specific areas of the brain by unknown mechanisms [67,69]. Therefore, the ependymal route appears to be a suitable strategy to perform IGF1 gene delivery to the brain. An adenoviral vector was chosen for ICV gene therapy based on the fact that these vectors are highly selective for ependymal cells, which are efficiently transduced  Fig. 1. All values are expressed as mean ± SEM. N = 6 rats per group, *p < 0.05 and **p < 0.01 vs Sham-DsRed, + p < 0.05 and + + + p < 0.001 vs SCI-DsRed. One-way ANOVA followed by Newman-Keuls post-test. Scale bar is 10 µm Fig. 6 RAd-IGF1 gene therapy increased arginase-1 (ARG)-positive hypertrophic microglia after chronic SCI. a-c Stereological quantitative assessment of Iba1 + ARG + and Iba1 + ARG − hypertrophic cells in the hilus (a), molecular layer (b) and SGZ + GCL (c). The group acronyms are the same as defined in Fig. 1. All val-ues are expressed as mean ± SEM. N = 6 rats per group, *p < 0.05, **p < 0.01 and ***p < 0.001 vs Sham-DsRed, + p < 0.05, + + p < 0.01 and + + + p < 0.001 vs SCI-DsRed, ##p < 0.01 and ###p < 0.001 Iba1 + ARG + vs Iba1 + ARG − . Two-way ANOVA followed by Newman-Keuls post-test The group acronyms are the same as defined in Fig. 1. N = 6 rats per group by RAd-IGF1 and substantially increase IGF1 levels in CSF [44]. Several reports have shown that this route of IGF1 administration is effective to modify neuron and glial modifications in brain parenchyma [35,38,45] Moreover, it is known that RAd-IGF1 gene therapy improves spatial and recognition memories in different models of neurodegeneration [35,37]. In this scenario, we decided to assess cognitive function after SCI and evaluate whether RAd-IGF1 gene therapy could reverse these deficits.
Our model of SCI generates a moderate spinal cord lesion and locomotor performance similar to those obtained by Fehlings [70,71]. In accordance with our previous results, injured rats performed uncoordinated consistent steps after chronic SCI [30]. Neither SCI nor ICV injections to injured rats modified locomotor activity (maximum speed and traveled distance) achieving values similar to Sham rats. Thus, these results validated cognitive evaluation as the tests performed imply rat locomotion.
We found that injured rats decreased the percentage of both novel object preference and spontaneous alternation, suggesting retention and spatial memory impairments after chronic SCI. The hippocampus is involved in the formation of memory and the representation of space [1], processes which require neuroplasticity. In this regard, immature neurons display higher excitability and plasticity making them functionally unique [72]. Moreover, several studies have demonstrated that adult-born neurons are required for hippocampus-dependent forms of memory such as recognition and spatial memory [62,64]. In this regard, neurogenesis reduction could be also related to cognitive failure during chronic SCI. Hippocampal neurogenesis is involved not only in plasticity but also in memory retention during spatial processing and encoding pattern separation [64,73].
Numerous reports describe that hippocampal failures are associated with low NOR and Y-Maze performances [74][75][76][77]. Notably, hippocampal neuroinflammation results in cognitive impairments deteriorating retention and spatial memory evaluated by the NOR [19][20][21] and Y-Maze tests [22,23] respectively. Therefore, our results are consistent with previously reported data which describe cognitive failure in hippocampal-dependent tasks associated with microglial activation after SCI [28,29].
RAd-IGF1 therapy applied long term after SCI succeeded in reversing hippocampal cognitive impairments. Indeed, injured rats treated with RAd-IGF1 recovered recognition and functional working memory as they presented a better performance in NOR and Y-Maze tests. These results are in line with two reports which demonstrated that ICV IGF1 gene therapy improves spatial memory performance in aged rats [35,37].
It is important to clarify that although both recognition and spatial memory involved the hippocampus, these functions also depend on other cortical regions such as the prefrontal or the perirhinal cortex [74,78]. Further studies would contribute to better understand the role of different brain areas after SCI.
In accordance with our previous results and other author publications, SCI led to neurogenesis reduction at chronic stages [30,31,79]. Notably, RAd-IGF1 therapy applied long term after SCI restored neurogenesis. Our laboratory has recently shown that neurogenesis reduction after chronic SCI is due to the inhibition of NSCs activation and proliferation [31]. In this regard, it has been described that IGF1 enhances neural stem cell (NSC) proliferation and differentiation in adult SGZ [65,80,81]. Since NSCs express the IGF1 receptor [81], this growth factor could be acting directly on these cells rescuing them from quiescence. Our result coincided with the increase in the number of DCX + cells induced by IGF1 gene therapy, which was described in the aged hippocampus [37].
In line with Faden's group, our laboratory has previously described glial abnormalities and neuroinflammation after chronic SCI [28][29][30]. In the present work, we found a reduction in the number of ramified microglial cells corresponding to surveillance microglia in the neurogenic niche and surrounding regions after chronic SCI. This reduction could be due to the activation of microglia given that the number of hypertrophic microglia increased and the number of total Iba-1 + cell remained unchanged after SCI. In addition, the percentage of ARG-expressing microglia, marker of noninflammatory phenotype [82], decreased about 50% after SCI.
The analysis of activated microglial population showed that there was an increase in the number of ARGhypertrophic microglial cells after SCI. Given that the number of total double-positive cells and the number of ramified microglia decreased, ARG + ramified microglia might have turned into ARGactivated microglia, increasing their number. However, the fact that some hypertrophic double cells became hypertrophic ARGmicroglia could not be discarded. ARGhypertrophic microglia is normally associated with neuroinflammation, aging, and neurodegenerative diseases [82,83]. Our laboratory has recently shown that SCI not only increases the number of hypertrophic microglial cells but also upregulates pro-inflammatory cytokines in the hippocampus [30,31]. In this regard, a higher proportion of ARGactivated microglia in the SGZ + GCL could be affecting neurogenesis as shown by the reduction in DCX + cells.
Although RAd-IGF1 therapy increased the expression of ARG throughout the hippocampus, we observed a region dependent effect in the regulation of microglial activity, as previously described by Falomir-Lockhart et al. [38]. In regions surrounding the neurogenic niche such as the hilus and ML, RAd-IGF1 therapy restored the distribution of microglial phenotypes, increasing the number of surveillance cells and decreasing the density of activated or hypertrophic microglia. As the total number of microglia remained unchanged among treatments, activated microglia could become ramified cells after RAd-IGF1 administration. RAd-IGF1 gene therapy also enhanced the percentage of total double-positive cells and recovered the number of ARG + hypertrophic microglia.
Since the number of ARGactivated microglia decreased concomitantly with the increase in the number of ramified cells, total double-positive cells, and ARG + activated microglia, RAd-IGF1 could be polarizing microglia to an antiinflammatory, neuroprotective phenotype, stimulating ARG activated microglia to switch to ARG-expressing ramified cells. We have recently published that chronic SCI induces neuronal loss [31]. Therefore, the recovery of ramified microglia could be protecting hippocampal neurons since ramified microglia exerts neuroprotective effects in response to excitotoxicity [84].
Therefore, our data supports that IGF1 acting on microglial cells promotes the anti-inflammatory microglial phenotype but inhibits the pro-inflammatory one [36]. The reduction of ARGhypertrophic microglia is associated with neuroprotection after traumatic brain injury [24] and in a model of hypertensive rats [85].
However, the scenario was a little different in the neurogenic niche. There was a significant reduction of ramified microglia together with an increment in hypertrophic cells. Since the total number of microglial cells remained unchanged among treatments, the increase of hypertrophic microglia may be due to the decrease of ramified microglial. Rad-IGF1 injection also increased the number of ARG + hypertrophic microglia. Therefore, both ARG − and ARG + ramified microglia are likely to become ARG + hypertrophic microglia. Microglial activation is not inevitably neurotoxic, and several neuroprotective actions have been demonstrated [15,86]. In fact, IGF1 gene therapy increases the number and the percentage of reactive microglia in the striatum and reduces motor impairments in aged rats [38]. In line with these results, ARG + activated microglia could release pro-neurogenic factors to neural stem cells and recover neurogenesis.
The biological actions of IGF1 are mediated through IGF1R, which is shown to be widely expressed throughout the brain, and concentrated in neuronal-rich areas [87]. IGF1R is expressed by different types of brain cells [88] and its expression varies with brain status, such as age and neurodegenerative diseases [89]. The different effects observed with IGF1 gene therapy could be related with a differential pattern of expression of IGF1R in microglial cells. Further work must be performed in order to discern the role and effects of IGF1 gene therapy in the different hippocampal regions under study.
Microglia activation and neurogenesis impairments affect hippocampal cognitive performances in several models [19][20][21][22][23][62][63][64]. Considering the present results, we postulate that RAd-IGF1 gene therapy could restore hippocampal cognitive performances by recovering the production of adult-born neurons and regulating microglial cells. In this regard, correlations studies showed a positive correlation between the number of ARG + microglia, cognitive function, and neurogenesis. These studies indicate a relationship between ARG + microglia and the improvement of hippocampal functions.
In addition, the restoration of ARG expression is involved in the regeneration of the CNS as it reduces contusion size after traumatic brain injury [24] and promotes retinal neuroprotection after ischemia [25]. ARG metabolizes L-arginine, which is the substrate for nitric oxide synthase, and it is known to regulate oxidative stress in various degenerative diseases by modulating nitric oxide and neuroinflammation [24,90]. Further studies should be performed in order to demonstrate that there is a causal relationship among IGF1 therapy, ARG + microglia, and the recovery of cognitive function after SCI.
Regarding IGF1 mechanism of action, IGF1 induces the activation of the PI3K/Akt signaling giving rise to NSCs differentiation after hippocampal denervation [86] and decreasing NSCs apoptosis after hypoxia [87]. Moreover, several reports have described that IGF1 is necessary for the expression of ARG and full adoption of the microglial antiinflammatory phenotype via the activation of STAT6/Akt signaling pathway [91,92]. Since IGF1R is expressed not only in microglia and NSCs but also in neurons and many other cells [93], pleiotropic mechanisms of action could not be discarded. As microglia and neural precursor cells in the spinal cord expressed IGF1R, it will be interesting to explore the effect of intracisternal IGF1 gene delivery on these cells. In this regard, intracisternal Rad-IGF-1 therapy abrogated later spinal cord damage and reduced the glial response induced by kainic acid [94]

Conclusions
In summary, chronic SCI caused neurogenesis reduction, microglial activation of the neurogenic niche, and neighboring regions. Injured rats also displayed both recognition and working memories impairments. Microglial alterations could explain neurogenesis reduction and cognitive deficits related to hippocampal dysfunction observed in rodents and even humans after SCI.
On the other hand, RAd-IGF1 gene therapy applied long term after SCI regulates microglial activation and increased neurogenesis. RAd-IGF1 modulated differentially hippocampal regions. In the hilus and ML, IGF1 increased the number of surveillance microglia, but in the neurogenic niche, IGF1 enhanced the number of activated microglia. The reversion of hippocampal abnormalities restored recognition and working memory impairments.
RAd-IGF1 gene therapy might surge as a new therapeutic strategy for patients with hippocampal microglial alterations and cognitive deficits such as those with spinal cord injury and other neurodegenerative diseases. Based on these findings, we will consider an alternative route of RAd-IGF1 administration in future experiments. Herrera et al. demonstrated that less invasive, feasible, and controllable route of IGF1 gene delivery, such as intramuscular injection, was effective for the treatment of the depressive phenotype in old mice [95]. Therefore, this route of administration will be considered in future experiments to evaluate hippocampal alterations after SCI.