Brain IGF-I is essential for hippocampal synaptic plasticity

The IGF-I signaling pathway fulfils a crucial role in the activity and plasticity of hippocampal synapses, and in hippocampus-dependent learning and memory (Trejo et al, 2007; Stern et al, 2014; Gazit et al, 2016), although the contribution of locally synthesized IGF-I to these processes remains to be fully elucidated. Here, we first evaluated the impact of local brain IGF-I deficiency on hippocampal synaptic plasticity, monitoring its potential consequences on LTP, a well-known cellular mechanism underlying learning and memory processes (Paulsen & Sejnowski, 2000). We recorded neuronal activity by measuring field excitatory postsynaptic potentials in the dentate gyrus (DG) before and after stimulating the perforant pathway using a high frequency stimulation (HFS) protocol (Fig 1A). HFS evoked robust synaptic potentiation in Igf-I^Ctrl slices relative to the baseline (***P < 0.001, one way-ANOVA, with post hoc Tukey’s test: Fig 1B–D), yet synaptic potentiation was not evident in hippocampal slices from Igf-I^Δ/Δ mice but rather, they displayed a significant depression of synaptic activity after the HFS protocol relative to the baseline (***P < 0.001, one way-ANOVA, with a post hoc Tukey’s test: Fig 1B–D). The difference in the responses to HFS in the hippocampal slices from either mice was significant (^###P < 0.001: Fig 1D) and supports a critical role for brain-synthesized IGF-I in long-lasting synaptic plasticity at the hippocampal DG.

Brain IGF-I regulates spatial memory and sexually dimorphic behavior

Because Igf-I depletion impairs hippocampal LTP, we tested whether two hippocampal-dependent tasks were affected in Igf-I^Δ/Δ mice: spatial learning and memory (Figs 2 and 3). In the Morris water maze (MWM), Igf-I^Δ/Δ mice displayed relatively higher escape latencies in the acquisition trials than the Igf-I^Ctrl/Ctrl mice, yet these increases were only significant on day 4 (*P < 0.05, t test) when the mouse genotype alone was considered in the analysis (Fig 2). The escape latency also increased in Igf-I^Δ/Δ females on day 3 (*P < 0.05, t test), whereas possible differences in Igf-I^Δ/Δ males did not reach statistical significance (P = 0.0509 on day 2: Fig S1A and B).

We observed significant changes in the performance of Igf-I^Δ/Δ mice relative to the Igf-I^Ctrl/Ctrl mice during the transfer test when analyzed independently of the animal’s sex (Fig 3A–C). These differences were evident through reductions in both the relative time spent in the platform (P) quadrant (**P < 0.01, t test: Fig 3B) and the relative distance swum in that quadrant (***P < 0.001, t test: Fig 3C). Moreover, the time spent by Igf-I^Ctrl/Ctrl mice in each quadrant differed significantly (^#P < 0.05, ^###P < 0.001, ^####P < 0.0001, one-way ANOVA followed by Tuckey post hoc test: Fig 3B), whereas Igf-I^Δ/Δ mice spent a similar time in all four quadrants (Fig 3B). In accordance with this, the distance swum by Igf-I^Ctrl/Ctrl mice in each quadrant differed significantly (^##P < 0.01, ^###P < 0.001, ^####P < 0.0001, one-way ANOVA followed by a Tuckey post hoc test: Fig 3C).

The absolute time spent by Igf-I^Ctrl/Ctrl mice in each quadrant also differed significantly (^#P < 0.05, ^##P < 0.001, ^####P < 0.0001, Welch’s ANOVA test followed by Games–Howell post hoc test Fig S2A), whereas Igf-I^Δ/Δ mice spent a similar time in all four quadrants. By contrast, the swimming speed in the pool (Fig S2C) and the mean velocity in each quadrant were largely unaffected by Igf-I deletion (Fig S2C′). Finally, no changes in the total freezing time were observed in the four quadrants (Fig S2D). However, there were significant differences in the Igf-I^Ctrl/Ctrl mice when the relative freezing time was compared between quadrants (^#P < 0.05, ^##P < 0.01, Welch’s ANOVA test followed by Games–Howell post hoc test: Fig S2D′). Together, these findings suggest that brain IGF-I plays a crucial role in the formation of new spatial memories.

A series of analyses were then performed to determine whether the animal’s sex might influence the spatial memory of Igf-I^Ctrl/Ctrl and Igf-I^Δ/Δ mice. Indeed, female mice appeared to spend relatively longer in the P quadrant than males, although only the difference between Igf-I^Δ/Δ female and male mice was significant (*P < 0.05, t test: Fig 3D). In addition, the Igf-I^Δ/Δ females spent relatively less time in the P quadrant than the Igf-I^Ctrl/Ctrl females (^#P < 0.05, t test: Fig 3D). Igf-I^Ctrl/Ctrl female mice swam a greater relative distance in quadrant P than the Igf-I^Ctrl/Ctrl males (*P < 0.05, t test: Fig 3E). Moreover, the relative distance swum in quadrant P by both Igf-I^Δ/Δ male and female mice was significantly less than that swum by Igf-I^Ctrl/Ctrl male and female mice (^##P < 0.01, t test: Fig 3E).

When the relative time spent by the mice in the different quadrants was plotted according to sex, more significant changes were observed among the Igf-I^Ctrl/Ctrl mice in females than in males (^#P < 0.05; ^###P < 0.001; ^####P < 0.0001, one-way ANOVA followed by Tuckey post hoc test: Fig S2A′). In addition, Igf-I^Δ/Δ females spent longer in the P quadrant than Igf-I^Δ/Δ male mice (*P < 0.05, t test: Fig S2A″). Some similarities were observed in the relative distance swum in the different quadrants, with the most significant changes observed in the female as opposed to the male Igf-I^Ctrl/Ctrl mice (^#P < 0.05; ^####P < 0.0001, one-way ANOVA followed by Tuckey post hoc test: Fig S2B), although these significant differences were not evident in the Igf-I^Δ/Δ mice (Fig S2B′). Moreover, Igf-I^Ctrl/Ctrl female swam longer distances in the P quadrant than Igf-I^Ctrl/Ctrl male mice (*P < 0.05, t test: Fig S2B). By contrast, no significant changes in the total velocity (Fig S2C″) or in the total freezing time (Fig S2D″) were observed between the male and female mice. These results indicate that the effect of brain IGF-I on spatial memory formation is at least in part sex-dependent.

Additional phenotyping of the IGF-I-deficient mice (Igf-I^Δ/Δ) showed a loss of sexual dimorphism in various behaviors (Fig 4). Horizontal and vertical deambulatory activity was determined in the open field test (OFT) and it was similar in both sexes of Igf-I^Δ/Δ mice, whereas it differed significantly in male and female Igf-I^Ctrl littermates (*P < 0.05, two way-ANOVA, with post hoc Bonferrani’s test: Fig 4A and B). The OFT can also be used to assess stress by measuring stereotypic movements and the time spent in the center of the arena. Igf-I^Δ/Δ mice did not show sexual dimorphism for either of these parameters, whereas this dimorphism was evident in control mice (*P < 0.05 and **P < 0.01, two way-ANOVA, with post hoc Bonferrani’s test: Fig 4C and D). A similar loss of sexual dimorphism between male and female Igf-I^Δ/Δ mice was evident in terms of their motor coordination measured in the rota-rod test (Fig 4E). By contrast, no significant changes were observed in the elevated plus maze that assesses anxiety-related behavior (Fig S3).

Hippocampal proteome and the structure of the GCL rely on brain IGF-I

To examine the consequences of IGF-I deficiency on hippocampal homeostasis, we applied shotgun proteomics to obtain novel information about the site-specific molecular signature of male and female Igf-I^Δ/Δ mice in the hope that this may help explain the electrophysiological and behavioral phenotypes observed. As such, hippocampal proteostasis was monitored using tandem mass tags (TMTs) coupled to tandem mass spectrometry (MS), detecting 65 differentially expressed proteins (DEPs) in female Igf-I^Δ/Δ mice and 58 DEPs in male Igf-I^Δ/Δ, relative to their respective Igf-I^Ctrl mice (Fig 5A and Table S1). No proteins were deregulated uniformly across both cohorts. However, most of the DEPs mapped to vesicle-related compartments, such as the TGN, presynapses, vesicle tethering complexes, and the early endosome (Fig 5B), with membrane trafficking the only biological activity disrupted at the protein level in both the female and male Igf-I^Δ/Δ hippocampi. This conclusion was also reached by performing a bioinformatics analysis that highlighted different functional profiles between the sexes of these mice (Fig S4).

Interlocking the differentially expressed proteomes with the SYNGO and Ingenuity Pathway Analysis repositories revealed alterations to synaptic proteins and to proteins involved in LTP. Specifically, Igf-I deficiency in males modulated a synaptic protein module comprised of FLOT1, RAB8A, RAB2A, BAIAP2, PLCB3, RPL7A, LGI1, PENK, CTNNA2, ADGRL1, VPS45, RHEB, FBXO2, PICK1, NF1, and ATP6AP1, whereas Igf-I^Δ/Δ female mice displayed changes in expression of a quite different set of synaptic proteins that included PRKCE, RAB6A, CASK, VAMP7, ERC1, GPHN, RPL6, MYO6, CTNND2, KCND2, SH3GL3, LYN7C, and NBEA (Table S2).

These alterations were accompanied by changes in protein intermediates of LTP and long-term depression, such as PAFAH1B2 and PRKCE in Igf-I^Δ/Δ females, and BAIAP2/IRSp53 together with PPP2R5A in Igf-I^Δ/Δ males. We analyzed the relationships between the altered hippocampal proteomes and the estrogen/androgen pathways or neurogenesis using the STRING and BIOGRID tools (Szklarczyk et al, 2019; Oughtred et al, 2021). Proteins involved in estrogen and androgen metabolisms were functionally connected to some of the deregulated hippocampal proteins (Fig 6). Furthermore, according to the BioGrid repository, 25% of the altered hippocampal proteins in female Igf-I^Δ/Δ mice have been previously characterized as proteins that physically interact with the oestrogen receptors ESR1 (16 proteins—MYO6, AGFG1, KHSRP, CEP170B, CKAP5, COX4I1, LIN7C, LUZP1, MINK1, PDK3, TPP1, RPL6, CASK, CTNND2, PSMC1, and NDUFB3) or ESR2 (17 proteins—MYO6, AGFG1, KHSRP, CEP170B, CKAP5, COX4I1, LIN7C, LUZP1, MINK1, PDK3, TPP1, CTSB, EZR, MCCC2, UGGT1, SFXN1, and PSMA5). Significantly, none of the hippocampal proteins altered in male Igf-I^Δ/Δ mice is currently considered to interact with androgen receptors.

To complement our study, we wanted to determine whether we could obtain functional evidence of a relationship between genes involved in neurogenesis (Nieto-Estevez et al, 2016b) and the dyshomeostatic proteome detected at the level of the hippocampus in Igf-I^Δ/Δ mice. As such, both the differential proteomes were interlocked with HES5, NEUROG2, CALB1, and DSCALML1, and although no functional relationships were observed in male Igf-I^Δ/Δ mice, deregulated proteins were identified in female Igf-I^Δ/Δ mice that were linked to CALB1, such as CASK1, GPHN, and KCND2 (Fig 7).

To search for specific targets affected by Igf-I deletion, we decided to investigate changes in individual hippocampal cells and proteins in Igf-I^Δ/Δ mice. Because we previously found that Igf-I deletion affected the positioning of Prox1⁺ neurons in the GCL of postnatal day 49 Igf-I^Δ/Δ mice (Nieto-Estevez et al, 2016b), we sought to investigate whether this phenotype was maintained in old mice (6–16 mo old). Immunohistochemistry (IHC) with an antibody against Prox1 identified more ectopic Prox1⁺ neurons (arrowheads) in the molecular layer (ML, 2.5-fold, *P < 0.01, t test: Fig 8A and B) and hilus (Hi, 2.5-fold, *P < 0.05, t test: Fig 8A and C) of Igf-I^Δ/Δ than in Igf-I^Ctrl/Ctrl mice. Therefore, local synthesis of IGF-I is necessary for the correct positioning of granule cells in the GCL throughout the animal’s life.

Having seen that Igf-I deletion altered the hippocampal proteome, we assessed whether this deletion produces changes in the relative expression of specific proteins in Western blots and using IHC. There was less RAB6A in the hippocampus of Igf-I^Δ/Δ mice, a small GTPase involved in LTP regulation (Gerges et al, 2005; Hausser & Schlett, 2019), although the change relative to Igf-I^Ctrl/Ctrl mice did not appear to be significant (P = 0.0554, t test: Figs 9A and S5A). A similar trend was observed for RAB2A (Gerges et al, 2005; Hausser & Schlett, 2019), gephyrin (Choii & Ko, 2015; Ravasenga et al, 2022) and VAMP7 (Kandachar et al, 2018), although the data dispersion was high and the small reductions observed in Igf-I^Δ/Δ mice were not significant (Figs 9B–D and S5A–C). Moreover, no differences were detected between males and females (Figs 9A–D and S5A–C).

To investigate whether RAB6A might be specifically modulated in neurons rather than in the hippocampus as a whole, we immunostained sections with antibodies against RAB6A and MAP2ab, the latter a general marker of neurons (Vicario-Abejon et al, 1998). We evaluated the co-localization of both these proteins in neurons located in the Hi, as they were easily identified (Fig S6). The fluorescence intensity was expressed in relative fluorescence units (RFU), and it was similar in both Igf-I^Δ/Δ and Igf-I^Ctrl/Ctrl mice. Moreover, no differences were found in the proportion of neurons immunoreactive for both MAP2 and RAB6A relative to the total number of MAP2⁺ neurons in the Hi. Similarly, no change in the RFU was evident when sections from Igf-I^Δ/Δ and Igf-I^Ctrl/Ctrl mice were immunostained with an antibody against the postsynaptic protein gephyrin (Fig S7), and against Synapsin-I and PSD95, a pre and a postsynaptic protein, respectively (Fig S8) (Vicario-Abejon et al, 2002), or an antibody against aromatase (an enzyme that catalyses oestrogen synthesis from androgens, Fig S9) (Garcia-Segura et al, 2010). By contrast, we detected significant changes when hippocampal sections from Igf-I^Δ/Δ and Igf-I^Ctrl/Ctrl mice were immunostained with antibodies against GAD65 and VGLUT1, two markers of inhibitory and excitatory neurons, respectively (Vicario-Abejon et al, 2002). In fact, GAD65 RFU was 36% higher in Igf-I^Δ/Δ than in Igf-I^Ctrl/Ctrl mice (*P < 0.05, t test: Fig 10A and B), whereas VGLUT1 RFU was a 12% higher in Igf-I^Δ/Δ than in Igf-I^Ctrl/Ctrl mice (*P < 0.05, t test: Fig 10A and C), suggesting that Igf-I deletion could provoke an imbalance in the inhibitory/excitatory ratio in the DG. Because parvalbumin (PVA) expressing neurons in the Hi drive synaptic inhibition in DG granule cells (Afrasiabi et al, 2022), we assessed the distribution of PVA in sections using an antibody raised against this protein (Fig 11). However, although the average number of PVA⁺ neurons in the Hi, ML, and GCL were all higher in Igf-I^Δ/Δ than Igf-I^Ctrl/Ctrl mice, these increases were not significant.

Altogether, these findings suggest that Igf-I deletion alters the structure of the GCL, the inhibitory/excitatory ratio, and hippocampal protein modules, which could ultimately lead to impairing LTP, producing deficits in spatial memory formation and affecting sexual dimorphic behaviors.