Original Article

Molecular Psychiatry (2007) 12, 1118–1128; doi:10.1038/sj.mp.4002076; published online 11 September 2007

Central actions of liver-derived insulin-like growth factor I underlying its pro-cognitive effects

Jl Trejo1,4, J Piriz1,4, M V Llorens-Martin1, A M Fernandez1, M Bolós1, D LeRoith2, A Nuñez3 and I Torres-Aleman1

  1. 1Laboratory of Neuroendocrinology, Cajal Institute, CSIC, Madrid, Spain
  2. 2Department of Medicine, Mt Sinai Hospital, New York, NY, USA
  3. 3Department of Neuroanatomy, School of Medicine, Autonoma University, Madrid, Spain

Correspondence: Professor I Torres-Aleman, Laboratory of Neuroendocrinology, Cajal Institute, Avda Dr Arce 37, Madrid 28002, Spain. E-mail: torres@cajal.csic.es

4These authors contributed equally to this work.

Received 2 October 2006; Revised 14 June 2007; Accepted 23 July 2007; Published online 11 September 2007.



Increasing evidence indicates that circulating insulin-like growth factor I (IGF-I) acts as a peripheral neuroactive signal participating not only in protection against injury but also in normal brain function. Epidemiological studies in humans as well as recent evidence in experimental animals suggest that blood-borne IGF-I may be involved in cognitive performance. In agreement with observations in humans, we found that mice with low-serum IGF-I levels due to liver-specific targeted disruption of the IGF-I gene presented cognitive deficits, as evidenced by impaired performance in a hippocampal-dependent spatial-recognition task. Mice with serum IGF-I deficiency also have disrupted long-term potentiation (LTP) in the hippocampus, but not in cortex. Impaired hippocampal LTP was associated with a reduction in the density of glutamatergic boutons that led to an imbalance in the glutamatergic/GABAergic synapse ratio in this brain area. Behavioral and synaptic deficits were ameliorated in serum IGF-I-deficient mice by prolonged systemic administration of IGF-I that normalized the density of glutamatergic boutons in the hippocampus. Altogether these results indicate that liver-derived circulating IGF-I affects crucial aspects of mature brain function; that is, learning and synaptic plasticity, through its trophic effects on central glutamatergic synapses. Declining levels of serum IGF-I during aging may therefore contribute to age-associated cognitive loss.


cognitive decline, trophic factors, insulin-like growth factor I



Blood-borne IGF-I is emerging as an unexpected factor in neurotrophic networks in the adult brain.1 Probably due to its multi-faceted actions on brain cells,2 serum IGF-I has also been related to cognition. Evidence gathered from human studies shows a positive correlation between serum IGF-I levels and mental abilities,3, 4, 5, 6, 7, 8 while cognitive impairment has been found in human patients affected by GH/IGF-I deficiency.9, 10 A more direct relationship between IGF-I and cognition is derived from animal experiments showing that administration of IGF-I restores cognition in aging animals,11, 12 that inhibition of IGF-I action results in cognitive impairment in adult rodents,13, 14 or that old mice with serum IGF-I deficiency show cognitive impairment.15 Based on all this evidence, we recently proposed that serum IGF-I contributes to building a 'cognitive reserve',16 a concept that refers to the availability of functional resources supporting cognition which was originally coined to explain relative resilience to dementia in relation to education level.17 This purported role of serum IGF-I could help to explain its neuroprotective effects against diverse types of insults18 and will theoretically support an active role of this circulating growth factor on higher brain functions.

However, evidence in favor of a role of serum IGF-I in brain mechanisms involved in cognitive processes is merely correlative in human studies and, albeit supportive in animals' experiments, more work is needed before we can establish a plausible relationship between serum IGF-I and cognition. In the present study, we have modulated circulating IGF-I levels using a mouse model of serum IGF-I deficiency combined with systemic administration of IGF-I. Our findings indicate that serum IGF-I modulates cognitive abilities due to its trophic effects on glutamatergic synapses.


Materials and methods


Mice with low-serum IGF-I were generated as described elsewhere by disrupting the liver IGF-I gene (LID mice) with the albumin-Cre/Lox system that starts to be activated at about postnatal day 10.19 Serum IGF-I-deficient mice have normal body and brain weights and do not show developmental defects.19, 20 Mutant mice have normal serum IGF-I levels during the first weeks of life (not shown) but soon thereafter liver IGF-I production starts to decline and by 2 months of age LID mice show a approx60% decrease in serum IGF-I as compared to control littermates.21 Tissue levels of IGF-I mRNA, including the brain, are normal19 (see also below).

General health and gross neurological evaluation of the mutant mice, assessed as published elsewhere,22 showed no difference with controls (Lox+/+Cre- littermates). Mice were genotyped by PCR. Adult 2–6 months old male mice were used throughout (specific ages are given in each experiment). In behavioral tests, age-matched wild-type C57BL/6 mice were included as additional controls together with control littermates because they are partially congenic to the original FVB/N breeders (back-crossed in C57BL/6 for seven generations). Animals were kept under standard diet and light/dark conditions following EU guidelines and handled according to institutionally approved procedures.

To determine the effects of IGF-I administration on LID mice, recombinant human IGF-I (GroPep, Adelaide, Australia) was infused (50 mug kg-1 per day, 2–3 weeks) through either subcutaneous osmotic minipumps (Alzet 2002, Alzet Minipumps, Cupertino, CA, USA) or biodegradable IGF-I microspheres.23 Control groups received saline-filled pumps or blank microspheres, respectively.

Behavioral tests

The ages of the animals used for behavioral testing were 3- to 4-months old animals for Morris water maze, visual discrimination and open-field behavior; 2–3 months for motor learning evaluation. Spatial learning was evaluated using a standard water maze test24 in a 100 cm wide pool. Room/pool cues consisted of four highly contrasted black plastic pieces of geometric forms for intramaze cues (light gray pool), and four highly contrasted white paper pieces of geometric forms for extramaze cues (dark gray curtains rounding the pool). This methodology is described in detail elsewhere.18, 25 Briefly, after 1 day habituation trial (day 1) in which preferences between tank quadrants were ruled out, for the subsequent 6 days the animals learned to find a hidden platform (acquisition). The next day, a probe trial without the platform was performed to evaluate preference for the platform quadrant (retention). On the last day, a cued version protocol was conducted to rule out possible sensorimotor and motivational differences between experimental groups. Swimming speed was found to be similar in all groups on all days. All animals were tested in four trials per day using 20 min inter-trial intervals. Individual sums of the four trials for escape latency were obtained previously to obtain the mean escape latency of each group. Path length and swim speed were measured with software EthoVision (Noldus, The Netherlands).

Motor learning was assessed using a rota-rod apparatus (Ugo Basile, Italy) under constant acceleration in six consecutive trials of 5 min as described.26 Animals were familiarized with the procedure under constant rod speed, 1 day before. A visual discrimination test was performed as described.27 Briefly, mice were deprived of water for 24 h and trained to distinguish a target cylinder (that is, black vs white non-target cylinders) containing water in a small well, in the absence of olfactory cues. The position of the cylinders is changed randomly during a 10-trial session. The test was performed during 2 consecutive days and results expressed as number of wrong choices before reaching the target. All animals learned the task because they performed significantly better (P<0.05) during the second day, as compared to day 1. Open-field behavior was assessed by measuring deambulatory activity in a computer-controlled actimeter cage (DigiScan, Omnitech Electronics, Columbus, OH, USA) allowing the animals to move freely for 15 min as described.26

In vivo electrophysiology

Initial recordings were performed in 4–5 months old animals. Next, osmotic minipumps were implanted in animals of the same age during 3 weeks; therefore, the subsequent recordings were made in 5–6 months old animals. Recording electrodes were stereotaxically implanted in the dentate gyrus (A: -2,3; L: 2; H: 1.5 mm from bregma) or neocortex (A: -2.3; L: 2; H: 0.5 mm from bregma) according to the Paxinos atlas.28 Twisted bipolar electrodes for electrical stimulation were aimed at the perforant path (A: -2.5; L: 0.5; H: 1.5 mm from bregma) or contralateral neocortex (same coordinates). Baseline recordings were taken with test stimuli (100 muA, 0.3 ms, 0.5 Hz) during 20 min before tetanic stimulation consisting of three pulse trains of 100 muA, lasting each pulse 0.3 ms and at 100 Hz. Trains lasted 500 ms and the inter-train interval was 2 s. Recording was maintained for 1 h after tetanic stimulation. Field potentials were amplified (AC Amplifier, WPI, Stevenage, USA), 0.1 Hz–3 kHz band-pass filtered and digitized at 10 kHz (CED 1401 with Spike 2 software, UK). Field potential averages were calculated using 50 stimuli. Baseline values were calculated as the mean from average responses over a 20 and 10 min period before high-frequency stimulation. Recordings were accepted for analysis when baseline variability was less than 10%.

In vitro electrophysiology

Hippocampal slices were prepared from 2 to 4 months old LID mice and control littermates. Mice were deeply anesthetized (pentobarbital, 35 mg kg-1) and decapitated immediately after disappearance of the pinch reflex. Brain was rapidly removed and submerged in an ice-cold high-sucrose artificial cerebrospinal fluid (aCSF) containing (mM): sucrose 248, KCl 3, NaH2PO4 1.25, NaHCO3 26, Dextrose 10, CaCl2 0.5, MgCl2 6, ascorbic acid 1 and pyruvic acid 3, bubbled with 95% O2–5% CO2, pH 7.4. Sagittal hippocampal slices of 350–400 mum were cut and incubated 1 h with bubbled aCSF of the following composition (mM): NaCl 124, KCl 3, NaH2PO4 1.25, NaHCO3 26, Dextrose 10, CaCl2 2, MgSO4 2, ascorbic acid 1 and pyruvic acid 3, pH 7.4. Slices were transferred to a recording chamber and perfused at a rate of 3–6 ml min-1 with aCSF (32°C with the same composition as above but without ascorbic and pyruvic acid). Extracellular recordings were made with glass pipettes (2–5 MOmega) filled with 150 mM NaCl. Pipettes were located on the outer 2/3 of the molecular layer and field-evoked potentials (fEPPs) were elicited by stimulation of the lateral perforant path. Data were acquired using an amplifier (Axoclamp 2B) and low-pass filter at 3 kHz. fEPPs were elicited at 0.05 Hz using a bipolar nichrome electrode during 20–30 min until baseline fEPPs slope was stabilized. Slices with fEPPs slopes of less than 0.5 mV ms-1 were discarded. Long-term potentiation (LTP) was induced by three high-frequency trains of 1 s at 100 Hz, 20 s apart. GABA-A antagonists biccuculine and picrotoxin (10 and 100 muM, respectively, Sigma, St Louis, MO, USA) were used in some experiments.


Brain tissue for immunohistochemistry was processed from 2 to 4 months old mice immediately after sectioning to obtain confocal images with the quality required for a high magnification. Glutamatergic and GABAergic synapses in the hippocampus were identified by double immunocytochemistry with anti-vesicular glutamate transporter 1 (VGlut1; Chemicon, Millipore, Billerica, MA, USA; 1:5000), a glutamatergic synapse marker,29 and anti-glutamic acid decarboxylase (GAD6; Developmental Hybridoma Bank, Iowa City, IA, USA; 1:1000), a GABAergic synapse marker,30 followed by biotinylated (Vector, USA, 1:1000) and Alexa 594-coupled (Molecular Probes, Invitrogen, Carlsbad, CA, USA, 1:1000) secondary antibodies. Biotinylated antibody was detected with Alexa 468-conjugated streptavidin (Molecular Probes, 1:1000). A dedicated set of 1-in-9 series of 50 mum vibratome sections of mice brains was used for the determination of glutamatergic and GABAergic boutons. Measurements were performed in images recorded with a confocal microscope (Leica TCS 4D), using a times 63 oil objective and a times 9 zoom. Four sections per animal were used comprising the entire frontal cortex or hippocampus, and one random area in the frontal cortex and the hippocampus per section (see Figure 4a) were measured. Image analysis was performed with AIS software (GE Healthcare, Piscataway, NJ, USA). Briefly, boutons with positive immunofluorescence (either VGlut1 or GAD6 because these markers never colocalize, see Figure 4a) were measured separately applying the same threshold to all pictures. Images were previously converted to gray scale to improve the contrast between signal and noise. Areas were measured inside a reference circle with a standard size of 325 mum.2 Reference space was located either in the inner molecular layer of the hippocampal dentate gyrus, lining the most external layer of granule neurons in the GCL, or in the outer portion of layer II of the frontal cortex, lining the most external layer of projection cortical neurons. We first calculated the percentage of reference area occupied by each type of bouton. We then determined the VGlut1/GAD6 ratio, and considering that the two types of boutons do not colocalize we obtained the data of the sum of percentage area occupied by both types of synaptic boutons. Finally, we determined the boutons (VGlut1 only, GAD6 only and the sum)-to-granule neuron number ratio. Data are presented as percentage of control values.


IGF-I and brain-derived neurotrophic factor (BDNF) were measured in 2–4 months old mice. Endogenous IGF-I levels were assayed with an enzyme-linked immunosorbent assay (ELISA) for murine IGF-I using anti-mouse IGF-I antibodies (MAP791 and BAF791, R&D, Minneapolis, MN, USA) and mouse IGF-I (IBT, Germany) as standard (sensitivity of the assay is 0.5 ng ml-1). Using this assay, we did not detect human IGF-I in treated animals: mice had identical endogenous serum IGF-I values after human IGF-I treatment (486plusminus7 ng ml-1 before and 486plusminus4 ng ml-1 after treatment; n=5–7 per group). Exogenously administered IGF-I (human recombinant IGF-I) was measured after 7–10 days of infusion with a commercial ELISA for hIGF-I (DSL, Webster, TX, USA) following the manufacturer's instructions (sensitivity of 13 ng ml-1). In this assay, control wild-type mice receiving blank microspheres had undetectable serum levels of hIGF-I. Hippocampal BDNF levels were measured by a commercial ELISA (Chemicon). The sensitivity of the assay is 7.8 pg ml-1. Western-blot and immunoprecipitation were performed as described21 to determine TrkB and Tyr-phosphorylated TrkB. Anti-TrkB antibodies (Santa Cruz, CA, USA) and anti-pTyr (PY20 clone) were used at 1:1000–2000 dilution. Levels of pTyrTrkB were expressed relative to TrkB load in each lane. Levels in the ELISA assays were expressed relative to total protein levels assessed with the Bradford reagent (Bio-Rad, Hercules, CA, USA) in parallel aliquots. Densitometric analysis was performed using Quantity One software (Bio-Rad). Horseradish peroxidase-coupled secondary antibodies were from Bio-Rad.

RNA quantification

IGF-I mRNA was quantified by qPCR using an Applied Biosystems (Foster City, CA, USA) kit as specified by the manufacturer in an ABI Prism 7000 Sequence Detector (Applied Biosystems, Weiterstadt, Germany) in a 20 mul volume reaction. Total RNA was extracted from 3-month-old mouse hippocampi using the Rnaqueous-4PCR kit (Ambion, Austin, TX, USA). cDNA synthesis was performed by using 100 ng of RNA with High-capacity cDNA reverse transcription kit (Applied Biosystems, Foster City, CA, USA). TaqMan probes and primers for IGF-I and for the control housekeeping gene, rRNA 18 s, were from the Assay-on-Demand gene-expression products (Applied Biosystems). All TaqMan probes were labeled with 6-carboxy fluorescein (FAM). All reactions were performed in triplicates using the TaqMan Universal PCR Master Mix. The standard curves for IGF-I and rRna 18 s were generated with serial dilutions of the cDNAs from hippocampus (n=6) of wild-type mice, control littermates and LIDs. Gene expression was normalized for rRNA 18 s levels.


Statistical analysis was performed with a t-test when comparing two groups and a two-way analysis of variance (ANOVA) for comparing multiple groups. Post-hoc comparisons were made with a Newmans–Keuls' test. Results are shown as meansplusminuss.e.m. Water maze results were analyzed with two methods. In the learning curves, escape latencies of all animals across days and groups were compared with repeated measures of ANOVA and post-hoc tests. This allows to analyze if time is a factor influencing the scores across groups. In Figure 1a, representing escape latencies, significant differences between groups are indicated with asterisks. In the probe trials, escape latency scores of the different groups were compared with a one-way ANOVA and post-hoc Tukey's test.

Figure 1.
Figure 1 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Spatial learning is impaired in mice with low-serum IGF-I levels (LID mice). (a) LID mice learned significantly worse the position of a hidden platform in the water maze as compared to control littermates (control) or wild-type (WT) mice. Treatment of LID mice with IGF-I significantly improved learning scores although animals remained partially impaired (**P<0.0001 vs controls). Mean escape latencies along the training days are shown (F=104.7, *P<0.0001 vs all other groups). (b) Probe trials (no platform) indicate that control mice spent significantly longer time in the platform quadrant than in the rest of quadrants as compared to LID mice (*P<0.05), and that treatment of LID mice normalized their performance. LID mice were swimming in each quadrant the same time (no significant preference for the platform quadrant). Wild-type C57BL/6 animals memorized the platform quadrant significantly better than all other groups (**P<0.05); n=5–8 per group. Error bars are s.e.m. in this and following figures.

Full figure and legend (75K)



Serum IGF-I and learning and memory

To determine a possible relationship between serum IGF-I and cognitive performance, we tested learning and memory status in low-serum IGF-I (LID) mice. LID mice showed normal motor learning in the rota-rod test (Table 1), while spatial learning was impaired. Young adult LID mice learned significantly worse than age-matched control littermates or wild-type mice in the water maze (F=104.7, P<0.0001) as determined by significantly different mean escape latencies (Figure 1a; P<0.0001 vs all other groups). LID mice also showed significantly worse memory of the location of the hidden platform as evidenced by a significantly decreased preference for the platform quadrant when animals were tested in the water maze in the absence of platform (Figure 1b, P<0.05 vs control littermates, probe trials). Probe trials also evidenced a significant difference between wild-type C57BL/6 mice and control littermates in retention of the platform position (Figure 1b, P<0.05 vs control littermates). Worse spatial memory in control littermates may be attributed to slight differences in genetic background with wild-type mice.

To rule out possible interferences in learning due to sensorimotor impairments, we checked the performance of LID mice in visual discrimination tests and in the cued version of the water maze. In both tasks, LID mice show normal scores as compared to control littermates (Table 1). To assess the role of circulating IGF-I in cognitive deficits, we chronically treated LID mice with systemic hIGF-I, a procedure that results in increased serum and brain hIGF-I in normal animals.26 Confirming previous observations, LID mice showed significantly lower endogenous serum levels (P<0.001 vs wild-type controls and littermates) but normal brain levels (P=0.06 vs wild types and P=0.08 vs littermates) of IGF-I as compared to controls (Table 2). After prolonged infusion with human IGF-I, both LIDs and controls had similar increases in serum and brain hIGF-I levels (Table 3).

We speculated that spatial learning deficits in LID mice could be related to low-serum IGF-I levels because these mice have normal levels of IGF-I mRNA as compared to wild-type controls and littermates (Table 2). Also, hippocampal IGF-I protein levels in LIDs were normal: 0.55plusminus0.05 pg mug-1 protein in LID vs 0.64plusminus0.03 pg mug-1 protein in wild types (n=5 per group; P=0.22). Notably, after IGF-I treatment, learning scores determined as mean escape latencies were significantly improved in LID mice (Figure 1, P<0.001 vs untreated LIDs).

Serum IGF-I and synaptic plasticity

To explore possible mechanisms underlying cognitive disturbances in low-serum IGF-I mice, we determined development of LTP in brain synapses of adult LID mice and control littermates. Various forms of synaptic plasticity have been proposed to underlie cognitive processes, and LTP is one of the best characterized (although its ultimate relationship to learning and memory processes remains to be established). We analyzed LTP of the perforant path-granule cell synapse in the hippocampus and found a complete absence of this synaptic form of plasticity in LID mice. Electrical stimulation of the perforant path at 0.5 Hz evoked a field potential in the granule cell layer with an initial positivity corresponding to the population excitatory postsynaptic potential (EPSP) followed by a negative wave representing the population spike potential (Figure 2a, inset). In control littermate mice (n=8), tetanic stimulation (three trains of 500 ms at 100 Hz) elicited a significant potentiation of the population EPSP slope (P<0.001 vs baseline levels for all points after tetanic stimulation), peaking at 10–30 min and lasting for at least 1 h after stimulation (Figure 2a; 191.3plusminus10.6% over baseline levels, 60 min post-tetanus). In LID mice (n=8), the same LTP protocol failed to produce an increase in the EPSP slope: 76.2plusminus10% of pre-stimulus levels after 60 min (Figure 2a). Further increasing the intensity of the stimulation to 120–150 muA was ineffective (n=4, not shown). Similarly, tetanus stimulation frequencies of 10–50 Hz did not produce LTP in LID mice either.

Figure 2.
Figure 2 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Synaptic plasticity is altered in the hippocampus of LID mice. (a) Plot of the populational excitatory postsynaptic potential (EPSP) slope before and after tetanic stimulation of the perforant path shows long-term potentiation (LTP) of the EPSP in control littermates (closed circles) and wild-type mice (closed triangles). However, LTP is absent in LID mice (open circles), but is restored after treatment with IGF-I (squares). Inset: raw data of the population hippocampal EPSPs before (left) and 10 min after (right) tetanic stimulation of the perforant path in a control case; n=8 per group. (b) Plot of the neocortical population EPSP amplitude before and after contralateral neocortical tetanic stimulation. LTP is evoked in both control and LID mice (closed circles and open circles, respectively). Inset: raw data of neocortical field potentials before (left) and 10 min after (right) stimulation of contralateral neocortex in a control case. Horizontal line in A and B indicate mean baseline values, which were taken as 100%; n=6 per group. Recordings were performed in 4–6 months old mice.

Full figure and legend (61K)

Lack of LTP in LID mice is specific for this hippocampal circuitry because neocortical LTP was evoked in LID as well as in control mice. A similar stimulation protocol was applied to callosal fibers and the evoked field potential was recorded in the contralateral neocortex. In this case, electrical stimulation elicited a field potential which consisted in an initial negativity followed by a positive wave (Figure 2b, inset). Neocortical field potential amplitude in either control littermates (n=6) or LID mice (n=6) increased after LTP induction before reaching a stable potentiated level (that is, values similar to those obtained 60 min post-tetanus; Figure 2b). However, LID mice reached a steady LTP level only after 30 min of tetanus while controls achieved it within 10 min of tetanic stimulation. When LTP was induced in the neocortex and the hippocampus of the same animal, LID mice exhibited LTP in the neocortex but not in the hippocampus (n=3).

We next determined whether hippocampal LTP in LID mice can be restored by systemic IGF-I administration. Indeed, after chronic treatment (3 weeks) with subcutaneous IGF-I, LID mice exhibited hippocampal LTP: EPSP slopes were larger after tetanic stimulation as compared to baseline values (P<0.005 vs untreated LID; n=4, Figure 2a).

Loss of hippocampal LTP in LID mice: underlying mechanisms

We analyzed possible mechanisms underlying the lack of hippocampal LTP in LID mice using hippocampal slices. We first corroborated the absence of LTP in the perforant path-granule cells pathway in hippocampal slices obtained from LID mice, while control hippocampal slices showed LTP (Figures 3a–c). We next observed that LTP was brought about in LID hippocampal slices when GABA-A blockers (bicucculine+picrotoxin) are included in the perfusion medium (Figures 3a, d and e). As shown in Figure 3d, in the absence of GABA blockers tetanic stimulation did not elicit LTP in LID slices (black dots) but when GABA inhibitors were included in the perfusion bath a second identical tetanus elicited LTP similar to control slices (white dots).

Figure 3.
Figure 3 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Lack of long-term potentiation (LTP) in hippocampal slices of IGF-I-deficient mice is due to excess GABAergic inhibition. (a) Following tetanic stimulation of the perforant pathway, field evoked potentials (fEPSP) obtained from slices of control littermates (dashed bar, n=15) and wild-type C57BL/6 (dotted bar, n=6) displayed LTP: after 60 min of the tetanus, a mean 129.6plusminus3.7% and 141.2plusminus10% increase, respectively, over the baseline (blank bar) fEPSP slope was obtained. In contrast, LID mice (gray bar) slices fail to develop LTP. A mean 104.3plusminus2.42% of the baseline fEPSP slope was seen (n=20). This lack of LTP was reversed by blocking GABA-A receptors (solid bar): the mean fEPSP slope was raised to 141.4plusminus4.7% of baseline values (n=13). **P<0.001 compared to baseline. Baseline fEPSP slope values are taken as 100% (note that basal fEPSP values are similar in LID and control mice, see panel C). (b) Representative fEPSP traces before (black trace) and after (green trace) tetanus in slices from control (n=5) and LID mice (n=6), respectively. Note that slices from LID mice do not display potentiation. (c) Tetanic stimulation elicited LTP in slices from control (white dots), but not from LID mice (black dots). While fEPSP values before tetanus were similar in both groups, after tetanus only control mice showed sustained enhanced fEPSP values. Tetanus is indicated by the arrow. (d) Perfusion of GABA-A antagonists (bicuculline+picrotoxin) unmasked the capacity of LID slices to undergo LTP. In the absence of GABA inhibitors LID slices (black dots, n=4) did not show LTP after tetanic stimulation (first arrow to the left). However, after GABA inhibition, a second tetanus (second arrow) elicited LTP in LID hippocampal slices similar to that obtained in slices from control littermates (white dots, n=5). (e) Representative fEPSP from LID mice slices at the indicated times (as shown in d): one, before tetanus; two, after the first tetanus; three, before the second tetanus in the presence of GABA-A antagonists and four, after the second tetanus in the presence of GABA-A antagonists. Note that slices from LID mice display potentiation only in the presence of GABA blockers. Total number of animals analyzed in each group is indicated in panel a. (For color figure see online version.)

Full figure and legend (109K)

Because no delayed potentiation was observed in the presence of GABA blockers (Figure 3d, stage 2), it appears that the first tetanus did not produce any form of non-expressed LTP and that GABAergic transmission was not potentiated either. No differences in pulse-paired facilitation between LID and control slices were found (Supplementary Figure 1A). Application of IGF-I to the bath did not affect the slope of fEPPs in LID or control slices (Supplementary Figure 1B and C). Moreover, in the presence of IGF-I, tetanus stimulation did not induce LTP in LID slices, indicating that IGF-I does not acutely modulate LTP (not shown).

Serum IGF-I and glutamatergic/GABAergic synapses

Absence of LTP in LID mice may be related to an altered excitatory-inhibitory balance, as suggested by normal LTP in LID hippocampal slices after GABAergic inhibition was blocked.31 Indeed, assessment of the number of GABAergic (GAD6+ profiles) and glutamatergic (VGlut1+) boutons in the hippocampal molecular layer of LID mice (Figure 4a) showed a significantly decreased number of VGlut1+ boutons (Figure 4b), without changes in GAD6+ profiles (Figure 4c), leading to a significantly reduced ratio in glutamate/GABA synapses (Figure 4d, P<0.05). Notably, the VGlut1+/GAD6+ ratio was not altered in the cortex of LID mice (P=0.6, layer II, the site where LTP recordings were performed). Other alterations that could account for disrupted hippocampal LTP in LID mice, such as presynaptic changes in the cortico-hippocampal circuitry analyzed or altered levels of BDNF or its receptor (TrkB) in the hippocampus were ruled out. Thus, the number of cells in layer II of the entorhinal cortex, the cortical area projecting to granule neurons through the perforant path was not altered. LID mice have 13 931plusminus1048 cells and control littermates 14 609plusminus1411 cells (n=6 per group). Similarly, hippocampal BDNF levels were comparable in LID mice and controls (Table 4), while levels of phosphorylated TrkB (the active form of the BDNF receptor) were in LID mice 85–94% of wild type and control littermate levels, respectively (not significantly different, see Table 4 and Supplementary Figure 1D). Administration of IGF-I did not produce any significant alteration in hippocampal BDNF and pTrkB levels as compared to baseline levels in any of the groups (Table 4).

Figure 4.
Figure 4 - Unfortunately we are unable to provide accessible alternative text for this. If you require assistance to access this image, please contact help@nature.com or the author

Hippocampal glutamatergic synapses are reduced in LID mice. (a) Left panel: representative photomicrograph of GAD6 (red, GABAergic synapses) and VGlut1 (green, glutamatergic synapses) double immunocytochemistry. Immunopositive boutons in the inner molecular layer and around the somas of granule cells (asterisks) of the hippocampal dentate gyrus of a control animal are shown. Right panels: representative confocal micrographs of the same brain section used to quantitate the number of VGlut1 and GAD6 synaptic boutons. The surface covered by either VGlut1 (upper panel) or GAD6 boutons (lower panel) within a given area (marked in this example with a blue circle) was quantified with image analysis software using a gray scale over a fixed threshold. Asterisks in the two panels show the cell soma of the same granule neuron: note that VGlut1+ and GAD6+ boutons do not colocalize. Mol, molecular layer; GCL, granule cell layer. (b, c) Percentage of total immunopositive surface occupied by GAD6 and VGlut1 boutons. In the inner molecular layer LID mice had significantly less VGlut1 density (b) than controls, while GAD6 density (c) did not vary. *P<0.05 vs all other groups. (d) As a result, the VGlut1/GAD6 ratio was significantly decreased in LID mice. Systemic treatment with IGF-I resulted in normalization in the density of VGlut1+ boutons (b) and VGlut1/GAD6 ratio (d) in LID mice. Wild-type controls have a similar VGlut1/GAD6 ratio than control littermates. *P<0.05 vs all other groups (F=4.03; P=0.006). Controls (n=20), Control+IGF-I (n=7), LID (n=22), LID+IGF-I (n=7), wild-type C57BL/6 (n=5).

Full figure and legend (278K)

Thereafter, we treated LID mice with systemic IGF-I for 2 weeks and counted the number of GAD6+ and VGlut1+ profiles. We found that IGF-I treatment significantly increased VGlut1 boutons in LID mice leading to a normalization of the glutamate/GABA synapse ratio. No changes in the total number of synapses or in the synapse-to-neuron ratio was observed in LID mice, nor after IGF-I treatment.



The present results suggest that the trophic actions exerted by circulating IGF-I on glutamatergic synapses modulate synaptic LTP within hippocampal circuitries and in this way affect hippocampal learning and memory. These observations agree with previous suggestions of a role of IGF-I on cognition in rodents11, 13, 14 and provide a mechanistic frame to help understand parallel changes in serum IGF-I levels and cognitive status in humans.3, 4, 5, 6, 7, 8 If these observations are translated to the human situation, we would anticipate that both in aged humans and in GH/IGF-I-deficient subjects, the number of hippocampal GABAergic and glutamatergic synapses may be imbalanced as compared to young and healthy individuals, which in turn may relate to cognitive disturbances known to occur in both cases.4, 9 This possibility warrants further study.

These observations add to the wide array of actions exerted by blood-borne IGF-I on brain cells, such as protection against injury,18, 26, 32, 33 modulation of neuronal excitability34, 35 brain angiogenesis,36 hippocampal neurogenesis37, 38 or amyloid clearance.21 An important conclusion that is emerging from these findings is that serum IGF-I exerts both relatively fast (within minutes to hours) and slow (within days and weeks) effects on brain function. For instance, neuronal excitability is modulated within minutes after systemic injection of IGF-I,35 while increases in new hippocampal neurons37, 38 brain vessels36 or glutamate synapses (present findings) take weeks. Naturally, we should envisage the biological role of serum IGF-I on brain function as a continuum ranging from acute to long-term effects on brain cells. In this regard, although we cannot yet establish a mechanistic link between acute and long-term actions of IGF-I on glutamatergic transmission, it is tempting to relate its reported acute effects on glutamate receptor function39, 40, 41, 42, 43, 44, 45 or in intraneuronal Ca2+ mobilization46, 47, 48 to long-term increases produced by IGF-I on glutamatergic receptor levels.49, 50 At any rate, since pulse-paired facilitation was not modified in LID hippocampal slices, presynaptic alterations may be ruled out, leaving postsynaptic actions of IGF-I as the most likely mechanism involved in its trophic action on glutamatergic transmission.

Other, non-exclusive explanations of the synaptic disturbances seen in serum IGF-I-deficient mice may be inferred from previously characterized long-term actions of this circulating growth factor. For instance, serum IGF-I controls brain amyloid levels.21 Since amyloid (Abeta) reduces glutamatergic transmission51 through, at least in part, increased internalization of glutamate NMDA receptors without affecting GABA receptors,52 high levels of brain Abeta due to reduced serum IGF-I levels may also lead to reduced glutamatergic synapses and altered LTP. Indeed, the perforant path-granule cell LTP examined in the present work depends on NMDA transmission.53 This type of process may help explain impaired cognition in aging because age is associated to low-serum IGF-I levels,54 and to impaired LTP at these synapses.53 It also may help to explain why successful aging in rats is associated to an increased use of non-NMDA-dependent LTP.55 In this regard it is important to note that neocortical LTP, albeit slower to develop, was normal in LID mice, suggesting that the effects of serum IGF-I are more evident in given brain areas. However, slower induction of cortical LTP in LID mice could indicate dysregulation in NMDA receptors and/or Ca2+ homeostasis at these synapses. Since the number of neocortical glutamatergic boutons was not significantly changed in LID mice, it is probable that altered Ca2+ handling in LID neocortical synapses may underlie this delay. At any rate, unveiling the cellular and molecular mechanisms underlying the trophic actions of serum IGF-I on glutamatergic synapses require further studies.

An important additional point that will need further clarification is the role of endocrine and metabolic disturbances associated with low-serum IGF-I levels found in LID mice, such as insulin resistance or high-GH levels.56 Although IGF-I administration ameliorated learning disturbances and its associated synaptic disturbances, we cannot entirely rule out whether metabolic/endocrine abnormalities associated with IGF-I deficiency may contribute or even be the cause of cognitive impairment in LID mice; even more if one considers the connection between insulin insensitivity and cognitive disturbances,57 or the possible actions of GH on hippocampal function.58 However, a direct correlation between serum IGF-I levels and insulin sensitivity,59 and a feedback effect of IGF-I on GH levels, makes it difficult to separate metabolic disturbances due to IGF-I deficiency from impaired tropism. In addition, because IGF-I is able to stimulate, albeit at lower affinity, the insulin receptor and insulin has pro-cognitive effects of its own,60 we cannot entirely discard a possible insulin-mimicking activity of IGF-I therapy on LID mice.

The present study also leaves open two intriguing points. Firstly, the observed differences between control littermates and wild-type C57BL/6 mice in the retention phase of the water maze test, while other parameters potentially associated to learning mechanisms such as hippocampal LTP were similar in both strains. As explained above, the strain where the LID mouse is bred is partially congenic with the C57BL/6 strain, and for this reason we included the latter as an additional control group in our study. Inter-strain differences in water maze performance and LTP have been documented in laboratory rodents61 and these have been started to be characterized at the protein level,62 which may help explain our observations, although more studies are warranted to clarify these differences. Second, while in the present experiments we did not find any difference between basal BDNF/pTrkB levels and those after prolonged administration of IGF-I, a functional association between IGF-I and BDNF has been found. Thus, basal BDNF mRNA levels in the hippocampus,63 as well as exercise-induced increases in BDNF protein are abrogated by infusion of an anti-IGF-I blocking antibody,64 while acute injection of IGF-I stimulates forebrain BDNF mRNA in a fashion reminiscent of an acute bout of exercise.34 The most parsimonious explanation for these apparent inconsistencies is that chronic low availability of IGF-I impinges on basal levels of brain BDNF, while an abrupt increase in IGF-I availability, perhaps resembling an exercise bout, is also able to produce an abrupt increase in BDNF. Conversely, an adaptive mechanism may likely impede that a prolonged increase in IGF-I availability will result in a permanent increase in BDNF levels. However, to determine this putative homeostatic process would require additional studies.

In summary, our findings add a new neuroactive role for circulating IGF-I and provide an explanation for previous observations correlating serum IGF-I levels and cognitive status in humans.



  1. Torres-Aleman I. Serum growth factors and neuroprotective surveillance. Mol Neurobiol 2000; 21: 153–160. | Article | PubMed | ISI | ChemPort |
  2. D'Ercole AJ, Ye P, Calikoglu AS, Gutierrez-Ospina G. The role of the insulin-like growth factors in the central nervous system. Mol Neurobiol 1996; 13: 227–255. | Article | PubMed | ChemPort |
  3. Aleman A, Verhaar HJ, de Haan EH, de Vries WR, Samson MM, Drent ML et al. Insulin-like growth factor-I and cognitive function in healthy older men. J Clin Endocrinol Metab 1999; 84: 471–475. | Article | PubMed | ISI | ChemPort |
  4. Arwert LI, Deijen JB, Drent ML. The relation between insulin-like growth factor I levels and cognition in healthy elderly: a meta-analysis. Growth Horm IGF Res 2005; 15: 416–422. | Article | PubMed | ISI | ChemPort |
  5. Dik MG, Pluijm SM, Jonker C, Deeg DJ, Lomecky MZ, Lips P. Insulin-like growth factor I (IGF-I) and cognitive decline in older persons. Neurobiol Aging 2003; 24: 573–581. | Article | PubMed | ISI | ChemPort |
  6. Kalmijn S, Janssen JA, Pols HA, Lamberts SW, Breteler MM. A prospective study on circulating insulin-like growth factor I (IGF-I), IGF-binding proteins, and cognitive function in the elderly. J Clin Endocrinol Metab 2000; 85: 4551–4555. | Article | PubMed | ISI | ChemPort |
  7. Okereke O, Kang JH, Ma J, Hankinson SE, Pollak MN, Grodstein F. Plasma IGF-I levels and cognitive performance in older women. Neurobiol Aging 2007; 28: 135–142. | Article | PubMed | ISI | ChemPort |
  8. Rollero A, Murialdo G, Fonzi S, Garrone S, Gianelli MV, Gazzerro E et al. Relationship between cognitive function, growth hormone and insulin-like growth factor I plasma levels in aged subjects. Neuropsychobiology 1998; 38: 73–79. | Article | PubMed | ISI | ChemPort |
  9. Bulow B, Hagmar L, Orbaek P, Osterberg K, Erfurth EM. High incidence of mental disorders, reduced mental well-being and cognitive function in hypopituitary women with GH deficiency treated for pituitary disease. Clin Endocrinol (Oxf) 2002; 56: 183–193. | Article | PubMed |
  10. Lijffijt M, van Dam PS, Kenemans JL, Koppeschaar HP, de Vries WR, Drent ML et al. Somatotropic-axis deficiency affects brain substrates of selective attention in childhood-onset growth hormone deficient patients. Neurosci Lett 2003; 353: 123–126. | Article | PubMed | ISI | ChemPort |
  11. Markowska AL, Mooney M, Sonntag WE. Insulin-like growth factor-1 ameliorates age-related behavioral deficits. Neuroscience 1998; 87: 559–569. | Article | PubMed | ISI | ChemPort |
  12. Trejo JL, Carro E, Lopez-Lopez C, Torres-Aleman I. Role of serum insulin-like growth factor I in mammalian brain aging. Growth Horm IGF Res 2004; 14(Suppl A): 39–43. | Article | ChemPort |
  13. Lupien SB, Bluhm EJ, Ishii DN. Systemic insulin-like growth factor-I administration prevents cognitive impairment in diabetic rats, and brain IGF regulates learning/memory in normal adult rats. J Neurosci Res 2003; 74: 512–523. | Article | PubMed | ISI | ChemPort |
  14. Castro-Alamancos MA, Torres-Aleman I. Learning of the conditioned eye-blink response is impaired by an antisense insulin-like growth factor I oligonucleotide. Proc Natl Acad Sci USA 1994; 91: 10203–10207. | Article | PubMed | ChemPort |
  15. Svensson J, Diez M, Engel J, Wass C, Tivesten A, Jansson JO et al. Endocrine, liver-derived IGF-I is of importance for spatial learning and memory in old mice. J Endocrinol 2006; 189: 617–662. | Article | PubMed | ISI | ChemPort |
  16. Carro E, Torres-Aleman I. Serum insulin-like growth factor I in brain function. Keio J Med 2006; 55: 59–63. | Article | PubMed | ChemPort |
  17. Fratiglioni L, Paillard-Borg S, Winblad B. An active and socially integrated lifestyle in late life might protect against dementia. Lancet Neurol 2004; 3: 343–353. | Article | PubMed | ISI |
  18. Carro E, Trejo JL, Busiguina S, Torres-Aleman I. Circulating insulin-like growth factor I mediates the protective effects of physical exercise against brain insults of different etiology and anatomy. J Neurosci 2001; 21: 5678–5684. | PubMed | ISI | ChemPort |
  19. Yakar S, Liu JL, Stannard B, Butler A, Accili D, Sauer B et al. Normal growth and development in the absence of hepatic insulin-like growth factor I. Proc Natl Acad Sci USA 1999; 96: 7324–7329. | Article | PubMed | ChemPort |
  20. Sjogren K, Liu JL, Blad K, Skrtic S, Vidal O, Wallenius V et al. Liver-derived insulin-like growth factor I (IGF-I) is the principal source of IGF-I in blood but is not required for postnatal body growth in mice. Proc Natl Acad Sci USA 1999; 96: 7088–7092. | Article | PubMed | ChemPort |
  21. Carro E, Trejo JL, Gomez-Isla T, LeRoith D, Torres-Aleman I. Serum insulin-like growth factor I regulates brain amyloid-beta levels. Nat Med 2002; 8: 1390–1397. | Article | PubMed | ISI | ChemPort |
  22. Crawley JN, Belknap JK, Collins A, Crabbe JC, Frankel W, Henderson N et al. Behavioral phenotypes of inbred mouse strains: implications and recommendations for molecular studies. Psychopharmacology (Berl) 1997; 132: 107–124. | Article | PubMed | ChemPort |
  23. Carrascosa C, Torres-Aleman I, Lopez-Lopez C, Carro E, Espejo L, Torrado S et al. Microspheres containing insulin-like growth factor I for treatment of chronic neurodegeneration. Biomaterials 2004; 25: 707–714. | Article | PubMed | ISI | ChemPort |
  24. Morris R. Developments of a water-maze procedure for studying spatial learning in the rat. J Neurosci Methods 1984; 11: 47–60. | Article | PubMed | ISI | ChemPort |
  25. Carro E, Spuch C, Trejo JL, Antequera D, Torres-Aleman I. Choroid plexus megalin is involved in neuroprotection by serum insulin-like growth factor I. J Neurosci 2005; 25: 10884–10893. | Article | PubMed | ISI | ChemPort |
  26. Fernandez AM, de la Vega AG, Torres-Aleman I. Insulin-like growth factor I restores motor coordination in a rat model of cerebellar ataxia. Proc Natl Acad Sci USA 1998; 95: 1253–1258. | Article | PubMed | ChemPort |
  27. Myhrer T, Johannesen TS. Learning and retention of a visual discrimination task in rats with various combinations of lesions in the temporal-hippocampal region. Brain Res Bull 1995; 36: 499–503. | Article | PubMed | ISI | ChemPort |
  28. Paxinos G, Watson CR. The Rat Brain in Stereotaxic Coordinates. Academic Press: Sydney, 1982.
  29. Alonso-Nanclares L, De Felipe J. Vesicular glutamate transporter 1 immunostaining in the normal and epileptic human cerebral cortex. Neuroscience 2005; 134: 59–68. | Article | PubMed | ISI | ChemPort |
  30. Cossart R, Dinocourt C, Hirsch JC, Merchan-Perez A, De Felipe J, Ben Ari Y et al. Dendritic but not somatic GABAergic inhibition is decreased in experimental epilepsy. Nat Neurosci 2001; 4: 52–62. | Article | PubMed | ISI | ChemPort |
  31. Kleschevnikov AM, Belichenko PV, Villar AJ, Epstein CJ, Malenka RC, Mobley WC. Hippocampal long-term potentiation suppressed by increased inhibition in the Ts65Dn mouse, a genetic model of Down syndrome. J Neurosci 2004; 24: 8153–8160. | Article | PubMed | ISI | ChemPort |
  32. Fernandez AM, Gonzalez de la Vega AG, Planas B, Torres-Aleman I. Neuroprotective actions of peripherally administered insulin-like growth factor I in the injured olivo-cerebellar pathway. Eur J Neurosci 1999; 11: 2019–2030. | Article | PubMed | ISI | ChemPort |
  33. Pulford BE, Whalen LR, Ishii DN. Peripherally administered insulin-like growth factor-I preserves hindlimb reflex and spinal cord noradrenergic circuitry following a central nervous system lesion in rats. Exp Neurol 1999; 159: 114–123. | Article | PubMed | ISI | ChemPort |
  34. Carro E, Nunez A, Busiguina S, Torres-Aleman I. Circulating insulin-like growth factor I mediates effects of exercise on the brain. J Neurosci 2000; 20: 2926–2933. | PubMed | ISI | ChemPort |
  35. Nunez A, Carro E, Torres-Aleman I. Insulin-like growth factor I modifies electrophysiological properties of rat brain stem neurons. J Neurophysiol 2003; 89: 3008–3017. | Article | PubMed | ISI | ChemPort |
  36. Lopez-Lopez C, LeRoith D, Torres-Aleman I. Insulin-like growth factor I is required for vessel remodeling in the adult brain. Proc Natl Acad Sci USA 2004; 101: 9833–9838. | Article | PubMed | ChemPort |
  37. Aberg MA, Aberg ND, Hedbacker H, Oscarsson J, Eriksson PS. Peripheral infusion of IGF-I selectively induces neurogenesis in the adult rat hippocampus. J Neurosci 2000; 20: 2896–2903. | PubMed | ISI | ChemPort |
  38. Trejo JL, Carro E, Torres-Aleman I. Circulating insulin-like growth factor I mediates exercise-induced increases in the number of new neurons in the adult hippocampus. J Neurosci 2001; 21: 1628–1634. | PubMed | ISI | ChemPort |
  39. Carroll RC, Beattie EC, von Zastrow M, Malenka RC. Role of AMPA receptor endocytosis in synaptic plasticity. Nat Rev Neurosci 2001; 2: 315–324. | Article | PubMed | ISI | ChemPort |
  40. Gonzalez de la Vega V, Buno W, Pons S, Garcia-Calderat MS, Garcia-Galloway E, Torres-Aleman I. Modulatory effects of insulin-like growth factor I on cerebellar glutamate receptors. October, New Orleans, USA. 1998. Society for Neuroscience: New Orleans, 1998, pp 1784.
  41. Gonzalez de la Vega AV, Buno W, Pons S, Garcia-Calderat MS, Garcia-Galloway E, Torres-Aleman I. Insulin-like growth factor I potentiates kainate receptors through a phosphatidylinositol 3-kinase dependent pathway. NeuroReport 2001; 12: 1293–1296. | Article | PubMed | ChemPort |
  42. Man YH, Lin JW, Ju WH, Ahmadian G, Liu L, Becker LE et al. Regulation of AMPA receptor-mediated synaptic transmission by clathrin-dependent receptor internalization. Neuron 2000; 25: 649–662. | Article | PubMed | ISI |
  43. Ramsey MM, Adams MM, Ariwodola OJ, Sonntag WE, Weiner JL. Functional characterization of des-IGF-1 action at excitatory synapses in the CA1 region of rat hippocampus. J Neurophysiol 2005; 94: 247–254. | Article | PubMed | ISI | ChemPort |
  44. Skeberdis VA, Lan J, Zheng X, Zukin RS, Bennett MV. Insulin promotes rapid delivery of N-methyl-D-aspartate receptors to the cell surface by exocytosis. Proc Natl Acad Sci USA 2001; 98: 3561–3566. | Article | PubMed | ChemPort |
  45. Wang YT, Linden DJ. Expression of cerebellar long-term depression requires postsynaptic clathrin-mediated endocytosis. Neuron 2000; 25: 635–647. | Article | PubMed | ISI | ChemPort |
  46. Blair LA, Marshall J. IGF-1 modulates N and L calcium channels in a PI 3-kinase-dependent manner. Neuron 1997; 19: 421–429. | Article | PubMed | ISI | ChemPort |
  47. Kanzaki M, Zhang YQ, Mashima H, Li L, Shibata H, Kojima I. Translocation of a calcium-permeable cation channel induced by insulin-like growth factor-I. Nat Cell Biol 1999; 1: 165–170. | Article | PubMed | ISI | ChemPort |
  48. Torres-Aleman I, Pons S, Arevalo MA. The insulin-like growth factor I system in the rat cerebellum: developmental regulation and role in neuronal survival and differentiation. J Neurosci Res 1994; 39: 117–126. | Article | PubMed | ChemPort |
  49. Sonntag WE, Bennett SA, Khan AS, Thornton PL, Xu X, Ingram RL et al. Age and insulin-like growth factor-1 modulate N-methyl-D-aspartate receptor subtype expression in rats. Brain Res Bull 2000; 51: 331–338. | Article | PubMed | ISI | ChemPort |
  50. Zona C, Ciotti MT, Calissano P. Human recombinant IGF-I induces the functional expression of AMPA/kainate receptors in cerebellar granule cells. Neurosci Lett 1995; 186: 75–78. | Article | PubMed | ISI | ChemPort |
  51. Kamenetz F, Tomita T, Hsieh H, Seabrook G, Borchelt D, Iwatsubo T et al. APP processing and synaptic function. Neuron 2003; 37: 925–937. | Article | PubMed | ISI | ChemPort |
  52. Snyder EM, Nong Y, Almeida CG, Paul S, Moran T, Choi EY et al. Regulation of NMDA receptor trafficking by amyloid-beta. Nat Neurosci 2005; 8: 1051–1058. | Article | PubMed | ISI | ChemPort |
  53. Barnes CA, Rao G, Houston FP. LTP induction threshold change in old rats at the perforant path–granule cell synapse. Neurobiol Aging 2000; 21: 613–620. | Article | PubMed | ISI | ChemPort |
  54. Breese CR, Ingram RL, Sonntag WE. Influence of age and long-term dietary restriction on plasma insulin-like growth factor-1 (IGF-1), IGF-1 gene expression, and IGF-1 binding proteins. J Gerontol 1991; 46: B180–B187. | PubMed | ISI | ChemPort |
  55. Lee HK, Min SS, Gallagher M, Kirkwood A. NMDA receptor-independent long-term depression correlates with successful aging in rats. Nat Neurosci 2005; 8: 1657–1659. | Article | PubMed | ISI | ChemPort |
  56. Yakar S, Liu JL, Fernandez AM, Wu Y, Schally AV, Frystyk J et al. Liver-specific igf-1 gene deletion leads to muscle insulin insensitivity. Diabetes 2001; 50: 1110–1118. | Article | PubMed | ISI | ChemPort |
  57. Craft S, Watson GS. Insulin and neurodegenerative disease: shared and specific mechanisms. Lancet Neurol 2004; 3: 169–178. | Article | PubMed | ISI | ChemPort |
  58. Donahue CP, Kosik KS, Shors TJ. Growth hormone is produced within the hippocampus where it responds to age, sex, and stress. Proc Natl Acad Sci USA 2006; 103: 6031–6036. | Article | PubMed | ChemPort |
  59. Dunger D, Yuen K, Ong K. Insulin-like growth factor I and impaired glucose tolerance. Horm Res 2004; 62(Suppl 1): 101–107. | Article | PubMed | ISI | ChemPort |
  60. Biessels GJ, Kamal A, Urban IJ, Spruijt BM, Erkelens DW, Gispen WH. Water maze learning and hippocampal synaptic plasticity in streptozotocin-diabetic rats: effects of insulin treatment. Brain Res 1998; 800: 125–135. | Article | PubMed | ISI | ChemPort |
  61. Nguyen PV, Abel T, Kandel ER, Bourtchouladze R. Strain-dependent differences in LTP and hippocampus-dependent memory in inbred mice. Learn Mem 2000; 7: 170–179. | Article | PubMed | ISI | ChemPort |
  62. Pollak DD, Scharl T, Leisch F, Herkner K, Villar SR, Hoeger H et al. Strain-dependent regulation of plasticity-related proteins in the mouse hippocampus. Behav Brain Res 2005; 165: 240–246. | Article | PubMed | ISI | ChemPort |
  63. Koopmans GC, Brans M, Gomez-Pinilla F, Duis S, Gispen WH, Torres-Aleman I et al. Circulating insulin-like growth factor I and functional recovery from spinal cord injury under enriched housing conditions. Eur J Neurosci 2006; 23: 1035–1046. | Article | PubMed | ISI |
  64. Ding Q, Vaynman S, Akhavan M, Ying Z, Gomez-Pinilla F. Insulin-like growth factor i interfaces with brain-derived neurotrophic factor-mediated synaptic plasticity to modulate aspects of exercise-induced cognitive function. Neuroscience 2006; 140: 823–833. | Article | PubMed | ISI | ChemPort |


We thank Dr J de Felipe (Cajal Institute) for his advice with synaptic markers, Dr MA Arevalo for his help with the pPCR and Dr J Torrado (Universidad Complutense, Madrid) for his generous help with IGF-I microspheres. We also appreciate the invaluable help of J Sancho, I Alvarez and M Garcia. This work was supported by Grants SAF2001-1722 and 2004-0446.

Supplementary Information accompanies the paper on the Molecular Psychiatry website (http://www.nature.com/mp)