Circulating insulin-like growth factor I is involved in the effect of diet on peripheral amyloid β clearance


 BackgroundObesity is a risk factor for Alzheimer´s disease (AD), but underlying mechanisms are not clear.MethodsWe analyzed peripheral clearance of amyloid β (Aβ) in overweight mice because its systemic elimination may impact on brain Aβ load, a major landmark of AD pathology. We also analyzed whether circulating insulin-like growth factor I (IGF-I) intervenes in the effects of overweight as this growth factor modulates brain Aβ clearance, and is increased in serum of overweight mice. Results Overweight mice showed increased peripheral Aβ clearance by the liver, the major site of elimination of systemic Aβ, but unaltered brain Aβ levels. We also found that Aβ clearance by hepatocytes is stimulated by IGF-I, and that mice with low serum IGF-I levels show reduced peripheral Aβ clearance and unchanged brain Aβ levels. In the brain, IGF-I favored association of its receptor (IGF-IR) with Aβ precursor protein (APP), and at the same time stimulated non-amyloidogenic processing of APP in astrocytes, as indicated by an increased sAPPα/sAPPβ ratio after IGF-I treatment. Since serum IGF-I enters into the brain in an activity-dependent manner, we analyzed in overweight mice the effect of brain activation by environmental enrichment (EE) on brain IGF-IR phosphorylation and its association to APP, as a readout of IGF-I activity. After EE, significantly reduced brain IGF-IR phosphorylation and APP/IGF-IR association was found in overweight mice as compared to lean controls. Conclusions Collectively, these results indicate that diet influences peripheral clearance of Aβ without affecting brain Aβ load. Increased serum IGF-I likely contributes to enhanced peripheral Aβ clearance in overweight mice, without affecting brain Aβ clearance probably because its brain entrance is reduced.


Background
Obesity is considered a risk factor for AD [1][2][3]. However, the relationship between body weight and dementia appears complex [4][5][6], and recent observations even pose a protective role of late-life excess weight in AD [7]. Taking into account the worrying worldwide prevalence of obesity and dementia [8,9], greater knowledge of possible links between the two conditions is imperative. Amyloid β (Aβ) handling may be one such link, as this peptide is considered a major pathogenic factor in AD and obesityassociated in ammation [10] may interfere with its elimination from the brain.
We recently proposed that insulin peptides such as insulin and insulin-like growth factor I (IGF-I), may be involved in the connection between life-style and AD risk [11], although apparently contradictory evidence links IGF-I with AD [12,13]. Signi cantly, the actions of IGF-I on the brain are modulated by diet [14]. Brain IGF-I is in part locally synthesized [15], and in part derived from uptake from the circulation [16]. The entrance of circulating IGF-I into the brain is activity-dependent and tightly regulated [17], probably because it participates in many essential brain functions [18].
IGF-I may not only be involved in brain Aβ clearance [12]; other ndings point to an effect of IGF-I on APP processing towards the non-amyloidogenic pathway, reducing in this way its production [19][20][21].
However, IGF-I has also been shown to favor the amyloidogenic pathway [22,23], while a deleterious effect of IGF-I signaling in proteostasis, favoring Aβ accumulation, has also been reported [13]. Of note, hepatocytes, the main source of circulating IGF-I [24], are the major disposal system for circulating Aβ in mice [25], and previous evidence has shown that insulin, a hormone closely related to IGF-I, favors Aβ uptake by hepatocytes [26].
In the present work we investigated regulation of peripheral Aβ clearance in overweight mice, its impact on brain Aβ levels, and the role of circulating IGF-I.

Materials And Methods
Materials. Human IGF-I was purchased from PeProTec (UK). Primary antibodies were monoclonal anti- Animals. Male adult (3-5 months old) and new-born wild type C57BL6/J mice, and adult liver IGF-I de cient mutant mice (LID mice; bred in-house, congenic with C57/BL6/J) were used. LID mice present low levels of serum IGF-I due to the disruption of the liver IGF-I gene with the albumin-Cre/Lox system [24]. Serum IGF-I de cient mice have normal body and brain weights and they do not show any major developmental defects [24,27] Fig A), developed glucose intolerance (Supp Fig B), together with hyperinsulinemia and insulin resistance (not shown).
Glucose tolerance test (GTT). Mice were fasted for 6 hours and left isolated in individual cages (with water but no food access) for at least 30 min before starting the test to avoid any stress-related effect on glycemia [30]. For the glucose tolerance test (GTT), an overload of glucose (2 g/kg) was injected intraperitoneally. The aqueous solution was left overnight at room temperature so the β-form of glucose was enriched. Blood samples were extracted from the tip of the tail at time 0, 15, 30, 60, and 90 to measure glucose levels with a glucometer (Menarini Diagnosis, Italy). Environmental enrichment. Mice were submitted to environmental enrichment as explained in detail elsewhere [31]. Brie y, animals were placed for 2 hours in a large cage, 10 animals/cage and with different objects (cardboard tunnels, shelters of different materials, a plastic net, toys, chewable and nesting material). Thereafter, they were sacri ced and their brain collected for immunoprecipitation and western blot analysis.

In vivo
Aβ40-HiLyte Fluor™ 488 (400 µg/kg) was injected into the tail vein using a 0.38 mm cannula (Intramedic, Spain), and after 90 min mice were sacri ced, blood taken from the heart and liver dissected. Liver tissue was homogenized in Tris-HCl (10 mM) pH 8.0 -guanidine (50 mM). Fluorescence in serum and liver extracts was quanti ed by uorimetry, as above. Values were normalized per ml of serum or mg of protein. The latter was measured in liver samples using the BCA system (Sigma).
Immunoassays. Western blot and immunoprecipitation were performed as described elsewhere in detail [33]. Densitometric analysis of blots was performed using the Odissey system (Lycor Biosciences, USA).
A representative blot is shown from a total of at least three independent experiments. GFAP immunocytochemistry in cultured cells followed previously published procedures [33]. In brief, cultured cells were incubated to block non-speci c antibody binding, followed by incubation overnight at 4 °C with anti-GFAP in phosphate buffer (PB) − 1% bovine albumin − 1% Triton X-100 (PBT). After several washes in PB, sections were incubated with an Alexa-coupled secondary antibody (1:1000, Molecular Probes, USA) diluted in PBT. Finally, a 1:500 dilution (in PBS) of DAPI (Hoechst 33342) was added for 3 minutes. Wells were rinsed several times in PB 0.1 N, mounted with 15 µl of gerbatol mounting medium, and allowed to dry. Omission of primary antibody was used as control. Microphotographs were taken in a Leica (Germany) microscope. Plaque load was determined as explained elsewhere in detail [33].
IGF-I in serum and brain was determined using a species-speci c ELISA (R&D Systems, USA), as described in detail elsewhere [18]. Murine Aβ (Thermo sher, USA), and murine sAPPα and sAPPβ were determined by ELISA in brain lysates and culture supernatants, respectively, following the manufacturer´s instructions. Blood was collected from the heart after pentobarbital anesthesia and thereafter brains were dissected and frozen at -80ºC until used.
Proximity ligation assays (PLA). Assays were performed as described [34]. Amyloid precursor protein (APP) -IGF-IR interactions were detected in astrocytes and neurons grown on glass coverslips using the Duolink II in situ PLA detection Kit (OLink; Bioscience, Sweden). Cultured cells were xed in 4% paraformaldehyde/10 min, washed with PBS containing 20 mM glycine to quench the aldehyde groups, permeabilized with the same buffer containing 0.05% Triton X-100 for 5 min, and washed with PBS. After 1 h/37 °C with the blocking solution in a pre-heated humidity chamber, cells were incubated overnight in antibody diluent medium with primary antibodies: mouse monoclonal anti-APP and rabbit polyclonal anti-IGF-I receptor, and processed following the instructions of the supplier using the PLA probes detecting rabbit or mouse antibodies (Duolink II PLA probe anti-Rabbit plus, and Duolink II PLA probe anti-Mouse minus, diluted 1:5 in antibody diluent), and a DAPI-containing mounting medium.

Statistical analysis. Normal distribution tests were carried out in all experiments and a non-parametric
Wilcoxon test was applied accordingly. For samples with normal distribution, parametric tests include one-way ANOVA followed by a Bonferroni or t-test. A p < 0.05 was considered signi cant.

Diet in uences peripheral Aβ clearance
We examined peripheral Aβ disposal in overweight mice because is a proposed mechanism for central Aβ clearance [35]. We administered uorescently tagged Aβ to mice fed with a high fat diet (HFD) for 10 weeks. Animals became overweight and glucose intolerant (Suppl Figure A,B). Ninety minutes after intravenous injection of Aβ, overweight mice showed signi cantly increased uorescence accumulation in the liver and decreased in serum, suggesting increased disposal of Aβ through hepatocytes (Fig. 1A).
Conversely, brain Aβ levels in overweight mice were not different from those seen in lean ones (Fig. 1B).

IGF-I promotes Ab uptake by hepatocytes
To try to clarify the discrepancy between increased peripheral disposal of Ab and normal brain Ab load, we analyzed a possible role of IGF-I, that is mostly produced by the liver [24], and is increased in overweight mice (Figure 2A). We determined whether IGF-I modulates uptake of Ab by hepatocytes, as a read-out of its clearance by this liver cell. As shown in Figure 2B, in the presence of IGF-I (10 nM), hepatocytes accumulated signi cantly more uorescence, suggesting a stimulatory action of IGF-I on Ab uptake by these cells. Moreover, liver IGF-I de cient (LID) mice with a 70% reduction in circulating IGF-I [36], showed reduced liver accumulation of tagged Ab after intravenous injection, while blood levels were increased, as compared to controls, indicating reduced liver clearance ( Figure 2C). However, as in overweight mice, LID mice did not show changes in brain Ab levels ( Figure 2D).
Cell-speci c actions of IGF-I in APP metabolism by brain cells IGF-I has been reported to promote either amyloidogenic [22], or non-amyloidogenic [20] APP processing pathways in neuronal cell lines. To clarify its role in primary cells, we analyzed the actions of IGF-I on amyloidogenic and non-amyloidogenic APP processing by astrocytes and neurons, the primary sources of Ab in the brain [37,38]. Using the soluble APP metabolites sAPPb and sAPPa as markers of the amyloidogenic and the non-amyloidogenic pathway, respectively, we found that IGF-I modulates their production in a cell speci c fashion. In astrocytes, secretion of both soluble forms of APP was stimulated by IGF-I, whereas in neurons IGF-I inhibited their secretion ( Figure 3A). However, the APPa/sAPPb ratio was increased in both cell types, indicating that the net action of IGF-I is to promote non-amyloidogenic processing of APP ( Figure 3B).
Since both IGF-IR and APP associate to LRP1, and APP processing depends on its subcellular localization [39], we assessed whether IGF-IR and APP interact with each other. Indeed, IGF-IR and APP coimmunoprecipitated in astrocytes, whereas in neurons the interaction was negligible ( Figure 3C). Proximity ligation assays (PLA) con rmed a robust interaction of APP with IGF-IR in astrocytes ( Figure  3D), while in neurons the interaction was negligible (not shown). Treatment of astrocytes with IGF-I resulted in a signi cantly increased interaction between both proteins, as determined by a stronger PLA signal ( Figure 3D).
Since IGF-I promotes Ab uptake by hepatocytes [40], we examined whether it can exert similar action in brain cells. In this organ, the main cell types involved in Ab clearance are microglia and astrocytes through its uptake and degradation [41,42], and endothelial cells at the blood-brain-barrier (BBB), through e ux of brain Ab into the circulation [43]. We found that IGF-I promoted Ab uptake by astrocytes ( Figure  4A), while decreased it in microglia ( Figure 4B). In brain endothelial cell cultures mimicking the BBB architecture [17], IGF-I not signi cantly inhibited Ab e ux from the "brain" side to the "blood" side of the double chamber ( Figure 4C).

Reduced brain IGF-I activity in overweight mice
Since serum IGF-I levels are increased in overweight mice (Figure 2A), we determined whether brain IGF-I is correspondingly higher, as serum IGF-I crosses the BBB [16]. However, overweight mice showed normal brain IGF-I levels ( Figure 5A). To explain this discrepancy between peripheral and central IGF-I levels, we determined whether passage of serum IGF-I into the brain is reduced in overweight mice. To this end, we took advantage that exposure to environmental enrichment (EE) stimulates the passage of serum IGF-I into the brain [17]. We tested whether overweight mice show altered passage of IGF-I after EE by measuring Tyr-phosphorylation of brain IGF-I receptors as a proxy of their activity. After EE, overweight mice showed reduced IGF-IR phosphorylation ( Figure 5B), pointing to impaired entrance of circulating IGF-I. In addition, since systemic administration of IGF-I to lean mice housed under standard conditions showed enhanced APP/IGF-IR interaction ( Figure 5C), corroborating in vitro observations ( Figure 3D), we used this interaction as an additional indicator of the entrance of IGF-I into the brain of EE-stimulated overweight mice. Signi cantly, whereas in lean mice EE produced enhanced brain APP/IGF-IR interactions, in overweight mice, this interaction was signi cantly smaller ( Figure 5D), pointing to reduced entrance of IGF-I.

Discussion
These results indicate that in adult normal mice, diet in uences peripheral Aβ clearance through the liver and that circulating IGF-I may play a role in this effect. Of note, either increases or decreases in peripheral Aβ clearance did not impact on brain Aβ levels. A lack of correlation between peripheral and central Aβ clearance agrees with observations that reducing peripheral Aβ does not affect brain Aβ levels, or with no correlation of central and peripheral Aβ levels in AD patients [44][45][46][47]. Compartmentalization of Aβ clearance may be re ecting multiple sources of this circulating peptide, as under normal physiological circumstances, Aβ is produced not only in the brain, but throughout the body.
However, other observations do not support compartmentalization of Aβ clearance. Thus, increased peripheral Aβ levels after anti-Aβ treatment was reported to parallel a decrease of brain Aβ; reducing peripheral Aβ was su cient to reduced brain Aβ, and recent studies favor a diagnostic utility of the relationship between plasma and CSF Aβ 1 − 42 [35,48,49]. Thus, the relationship between peripheral and central Aβ is still under debate [50]; indeed, a substantial part of brain Aβ clearance in humans takes place in the periphery [51].
In turn, normal brain levels of Aβ in overweight mice agree with previously reported similar observations [52], but not with increased brain Aβ load found by others [53,54]. Conversely, enhanced elimination of circulating Aβ in overweight mice favors the notion that higher body mass index may be protective rather than detrimental for AD risk [4], However, unaffected brain Aβ load does not t with a protective effect of increased body mass, unless still unde ned systemic changes contribute to AD, as recently postulated [55].
Among the latter, we considered circulating IGF-I as a probable systemic factor in uencing the connection of obesity with AD. IGF-I is involved in brain Aβ clearance [12] -although this has been questioned [56], and shows diet-sensitive actions in the brain [14]. Indeed, several observations favor an involvement of circulating IGF-I in the systemic pro-clearance actions of a high fat diet. 1) IGF-I levels are increased in overweight mice, 2) IGF-I, as previously seen with insulin [26], stimulates uptake of Aβ by hepatocytes, and 3) LID mice with low serum IGF-I show reduced peripheral Aβ clearance. Thus, higher serum IGF-I levels in overweight mice may contribute to reduce peripheral Aβ levels, without affecting brain Aβ load. The latter disagrees with our previous observation of increased brain Aβ levels in LID mice [12]. The use of an in-house ELISA and formic acid extraction of total brain Aβ (prior results) vs a commercial ELISA with guanidinium HCl extraction of brain Aβ (current results), or changing housing conditions of LID mice along time in our animal facility (i.e.: sterile food pellets) affecting their microbiome [57], that shows strong interactions with IGF-I function [58], may explain this discrepancy, but we do not have a straightforward explanation of this important difference with our previous results.
As indicated by an increased sAPPα/sAPPβ ratio in IGF-I-treated neurons and astrocytes, the net action of IGF-I on the main cell types producing Aβ in the brain is to favor non-amyloidogenic processing of APP, contributing in this way to lower its brain levels and enhance neuroprotection, as sAPPα is neuroprotective acting in part through IGF-IR [59]. Thus, the overall action of IGF-I in the brain may be antiamyloidogenic. Intriguingly, insulin favors Aβ secretion in neurons [60], suggesting a complex interplay of these hormones in regulating brain Aβ levels. At the same time, reduced IGF-I entrance in the brain of overweight mice may hamper its anti-amyloidogenic actions. Indeed, overweight mice showed not only reduced entrance of serum IGF-I in response to EE stimulation, as determined by reduced brain IGF-IR phosphorylation, but also reduced APP/IGF-IR interaction. In previous work we documented an inhibitory effect of triglycerides (TGLs) in BBB entrance of IGF-I across the choroid plexus [14]. It is possible that high serum TGLs as a result of the high fat diet interfere also with the BBB entrance of IGF-I across brain endothelial cells in overweight mice, as previously seen for other circulating hormones such as insulin [61] or leptin [62].
Reduced IGF-I entrance would affect its pro-clearance actions on brain Aβ [12,63]. Also, we cannot discard that the inhibitory actions of IGF-I on Aβ uptake by microglia may also counteract its actions on astrocytes (but see below). Alternatively, brain Aβ levels may not be affected by peripheral Aβ clearance or other factors may also contribute to it, such as the recently postulated vascular drainage [64,65]. Interestingly, insulin also enhances the degradation of Aβ and its clearance in astrocytes [66], and hepatocytes [26], respectively. Thus, these two closely related hormones may modulate Aβ disposal in a concerted manner, as previously reported for glucose handling [67].
IGF-I stimulates Aβ uptake by astrocytes, while inhibits it in microglia. Whereas astrocytes appear critical to determine Aβ load [68], and increased clearance of Aβ by astrocytes may result in reduced Aβ plaques [42], inhibition of Aβ uptake by microglia may also reduce plaques [69], as the role of microglial uptake of Aβ in plaque formation may be detrimental [70,71]. In accordance with a stimulatory effect of IGF-I on astrocytes, previous observations suggested that astrocyte-derived IGF-I protects neurons against Aβ toxicity through a mechanism involving its uptake [72].
The observed astrocyte-speci c interaction of APP with IGF-IR and on sAPPα and sAPPβ levels may be related to a differential processing of APP by IGF-I in these cells, since its processing depends on its intracellular localization [39]. Of the different isoforms of APP, the major one expressed in neurons is APP 695 , that lacks the extracelular protein-protein interaction domain KPI. This domain is present in the longer isoforms, APP 751 and APP 770 , that are the most abundant types in glial cells [73]. It is possible that KPI is involved in the observed interactions with IGF-IR in astrocytes, but this requires further analysis. In turn, a trend of IGF-I to inhibit brain e ux of Aβ through BBB endothelial cells, would favor its accumulation in brain parenchyma [74]. We previously reported that IGF-I stimulates Aβ e ux through the choroid plexus BBB [12], an observation supported by the reducing effects of in vivo IGF-I administration on brain Aβ levels [12,63]. Thus, IGF-I may show site-speci c effects on Aβ e ux through BBB cells.

Limitations
An important limitation of this study is that we determined peripheral Ab clearance indirectly. Measuring dynamic changes in circulating levels of Ab in overweight mice would be necessary to rmly establish that peripheral Ab clearance is enhanced. However, available methods of quanti cation of serum Ab are not sensitive enough to reliably detect decreases in non-transgenic mice.

Conclusions
In summary, diet in uences peripheral, but not central Aβ clearance. A lack of correlation between peripheral clearance and central Aβ load further support a non-linear relationship between both compartments. Actions of IGF-I on Aβ handling may be related to diet in uences on AD pathology; therefore cellular sites of IGF-I interaction may constitute new druggable targets, through, for example, potentiating the passage of circulating IGF-I into the brain across the BBB.

Ethical approval
This work was carried out in mice after receiving the approval of the Ethics Committee, as indicated in the manuscript Consent for publication

Not applicable
Availability of data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Competing interests experiments and analyzed results. JAZV and SDP helped with experiments. AMF performed experiments and analyzed results. ITA designed the study, analyzed results, and wrote the manuscript. Diet modulates peripheral A uptake. A, Overweight mice show enhanced A uptake by the liver and lower serum A levels (n=4). B, Brain A levels remain unaltered in overweight mice (n=6-8). *p<0.05, **p<0.01, and ***p<0.001 in this and following gures.  IGF-I modulates brain A uptake in a cell-speci c manner. A, A uptake by astrocytes is signi cantly increased by IGF-I (n= 8). Representative photomicrograph showing uptake by cultured astrocytes of uorescently labeled A (green). Cell nuclei stained with Hoescht. B, A uptake by microglia is signi cantly reduced by IGF-I (n= 7). C, IGF-I did not signi cantly affect brain-to-blood e ux in an in vitro system mimicking the blood-brain-barrier (cartoon in the left). Amount of A in the upper chamber was quanti ed 15 h after adding it to the lower chamber in the presence or absence of IGF-I (n= 6).

Figure 5
Reduced entrance of serum IGF-I in overweight mice. A, Brain levels of IGF-I were normal in HFD-fed overweight mice (n= 10 per group). B, In response to environmental enrichment (EE), overweight mice show lower brain IGF-IR phosphorylation than lean mice receiving a standard diet. Representative blot is shown, together with quanti cation histograms (n= 10 EE/ 6 Control; for each diet). C, Brain IGF-IR/APP co-immunoprecipitation is increased after systemic IGF-I administration. D, Interaction of APP with IGF-IR in the brain of mice submitted to EE stimulation was signi cantly decreased in overweight mice (n=10 per group).

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.