Epidermal growth factor prevents APOE4 and amyloid-beta-induced cognitive and cerebrovascular deficits in female mice

Experimental design

All protocols follow the UIC Institutional Animal Care and Use Committee protocols. Breeding and colony maintenance was conducted at the UIC as described in [61]. EFAD mice express five Familial Alzheimer’s disease (FAD) mutations (APP K670N/M671L + I716V + V717I and PS1 M146L + L286V) and human APOE3 (E3FAD mice) or APOE4 (E4FAD mice) i.e. 5xFAD^+/− APOE ^+/+. EFAD mice utilized to start our breeding colonies were kindly provided by Dr. M.J.Ladu.

Behavioral analysis

Behavioral analysis was conducted in the mouse dark cycle, tracked in real time by an overhead camera and videos analyzed using the ANY-Maze software.

Open field. Mice were placed in the center of a white box (l38.5xw30xh30 cm) and allowed to freely move for 10 min. The distance traveled and average speed were measured.

Novel object recognition was conducted as described in [62] with slight modifications. On day 1, mice were habituated in a white testing chamber (38.5x30x30 cm) for 20 min. On day 2, mice were placed in the same testing chamber that contained two identical objects for 7 min, returned to their home cage for 1 hr and placed in the testing chamber with a familiar and a novel object for 7 min. Time spent with each object and the preference index (ratio of time spent with the novel object divided by total investigation time for both objects) were calculated. Mice that did not investigate both objects for longer than 5 s were not included. Data for one E3FADF and one E4FADF were removed (Fig. 1) based on the exclusion criteria.

Sodium fluorescein extravasation

Mice were injected i.p. (200 μl) with 2 % NaFl, sacrificed after 30 min and tissue harvest conducted as described above. Dissected brain regions were weighed, homogenized in PBS, suspended in equal volumes of 60 % tricholoroacetic acid, then vortexed and centrifuged (18,000 × g,10 min, 4 °C). Fluorescence levels were measured using a microplate reader (SpectraMax i3x, Molecular Devices) and cleared volumes calculated as: [1/plasma levels (fluorescence units/μl) x total brain fluorescence]/brain weight.

Biochemical analysis

Dissected brains were weighed and sequentially extracted using the 3-step extraction protocol as described in [61].

Histological analysis

Frozen brains were sectioned at 12 μm and fixed using 10 % Neutral Buffered Formalin (Sigma). Nine nonadjacent sections (108 μm apart) were utilized for quantification per animal.

Statistical analysis

All data are presented as mean +/−S.E.M, and were analyzed using two-way ANOVA followed by either Tukey’s, Fisher’s LSD or Sidak’s post hoc comparisons or using Student’s t-test, as specified in the figure legends using GraphPad Prism version 6.

Female E4FAD mice are cognitively impaired at 8 months of age

Higher CV leakiness and lower CV coverage in E4FADF mice

To determine the role of APOE and sex in CV leakiness, brain levels of NaFl were assessed after i.p. injection (Fig. 1b). NaFl levels were significantly higher in the cortex of E4FADF mice compared to E3FADM and E4FADM mice and were higher in hippocampus compared to all other groups (two-way AVOVA followed by Tukey’s post-hoc analysis). NaFl levels were also higher in E3FADF mice compared to EFADM mice (two-way AVOVA followed by Tukey’s post-hoc analysis) in the hippocampus, whereas in the cortex these effects were only significant by two-way AVOVA followed by Fisher’s LSD test. Prussian blue staining was conducted to assess microbleed coverage as a complimentary measure of CV leakiness (Fig. 1c). In the cortex the area of microbleed coverage followed the order E4FADF > E3FADF > E4FADM = E3FADM (Fig. 1c, two-way AVOVA followed by Tukey’s post-hoc analysis). In general cortical microbleed coverage was 20-fold higher in E3FADF mice and 55-fold higher in E4FADF mice compared to EFADM mice. In the hippocampus microbleed coverage was only higher in E4FADF mice when assessed by two-way AVOVA followed by Fisher’s LSD test.

Next, laminin (basement membrane protein)-Aβ co-staining was conducted as an indication of total vessel coverage. Visually laminin staining appeared lower in deeper layers of the frontal cortex and the subiculum in EFADF mice compared to males, an effect particularly pronounced in E4FADF mice (Fig. 2a and b). When quantified for % area covered (Fig. 2c), no differences were observed for laminin staining among the groups for the whole cortex. In the deep layer cortex laminin staining followed the order: E4FADM/E3FADM > E3FADF > E4FADF (two-way AVOVA followed by Tukey’s post-hoc analysis). Data for the subiculum followed the same pattern. Overall, laminin staining was 10 % lower in E3FADF mice and 20 % lower in E4FADF mice in the deep layer cortex and subiculum compared to EFADM mice. Extracellular Aβ levels followed the order E4FADF > E4FADM = E3FADF > E3FADM (Fig. 2d).

Plasma EGF levels are lowest in EFADF mice

EGF plays a key role in controlling total vessel coverage and a recent report indicates that low plasma EGF predict the conversion from MCI to AD [33]. We therefore measured plasma EGF levels in EFAD mice (Fig. 2e). E3FADM mice had higher plasma EGF levels than all other groups (2-way ANOVA, followed by Tukey’s post-hoc comparison). There was also a trend of higher EGF plasma levels in E4FADM mice compared to EFADF mice, which was significant by Student’s t-test. One caveat is that NaFl injection may have modulated plasma EGF levels in the EFADF mice, in an APOE genotype specific manner.

Collectively, these data support that 8 month old EFADF mice are cognitively impaired, have low CV coverage, high CV permeability and low plasma EGF levels. Therefore E4FADF mice were selected for treatment with EGF in a prevention paradigm.

EGF prevents cognitive decline in E4FADF mice

Our overarching goal was to test EGF in a prevention paradigm based on our recent data that EGF prevents Aβ-induced disruption of angiogenesis and degeneration of preformed tubes in vitro [31]. E4FADF mice were treated with EGF (300 μg/kg/wk) by i.p. from 6 to 8.5 months and tested by a battery of tests for cognitive ability (Fig. 3a). 300 μg/kg/wk was selected as a low dose that does not promote cancer or overt signs of toxicity in mice [6]. EGF induced a profound protective effect on cognitive decline in E4FADF mice compared to the vehicle control (VC). When assessed by novel object recognition (Fig. 3b), spontaneous alternation (Fig. 3c, Y-maze) and novel arm entry (Fig. 3d, Y maze) the VC control groups progressively declined in cognitive ability from 6 to 8 months (two-way ANOVA followed by Sidak’s post hoc comparisons which takes into account same animal re-testing). In contrast, no age-dependent decline was observed for EGF treated mice. At the mid-point EGF treated mice performed 46 % higher in the novel object recognition and 37 % higher in the number of spontaneous alternations performed (Y-maze) compared to the VC. At the end-point, EGF treated mice performed ~58 % higher in novel object recognition and spontaneous alternation and spent 477 % longer in the novel arm test (Y-maze).

Cognitive ability was further dissected at the end-point in the Morris water maze test (Fig. 3e). In the acquisition phase EGF (day 1 vs 5 and 2 vs 5 were significant) and VC treated mice (day 1, 3 and 4 vs day 5 were significant) learned the location of the platform by day 5. On day 3 the EGF treated mice located the platform earlier than the VC, indicative of improved learning. In the probe trial (Student’s t-test) EGF treated mice located the previous platform area with a lower latency time (VC = 38.6 s, EGF = 5.15 s) and spent a longer time in the target quadrant (VC = 11.6 s, EGF = 30.9 s). EGF did not modulate levels of the pre-synaptic protein synaptophysin in the cortex or hippocampus (Fig. 3f). However, consistent with the behavior data levels of the postsynaptic protein PSD95 were higher in the cortex (35 %) and hippocampus (50 %) in EGF treated mice (Fig. 3g, Student’s t-test). Thus, in a prevention paradigm EGF-induces a beneficial effect on cognition and post-synaptic neuronal markers.

EGF treatment had no effects on locomotion, food intake, anxiety-like behavior and brain EGF levels

A potential confounding factor for the beneficial effects of EGF on cognition is whether EGF modulates anxiety-like behavior. VC treated mice remained steady in weight until week 4 (7 months old) where they gained ~5 % and then remained at the same weight for the rest of the study. EGF treated mice lost ~ 7 % in weight at week 4 and then remained steady (Fig. 3h). The changes in bodyweight were not related to general locomotion, as in the open field test there were no differences between EGF and VC treated mice (Fig. 3i, two-way ANOVA followed by Sidak’s post hoc). At the end-point (8 months) there were no differences in food intake (Fig. 3j over 24 hrs, Student’s t-test) or anxiety like behavior as assessed in the light-dark box test (Fig. 3k, Student’s t-test). The improved cognition with EGF compared to the VC was not related to general locomotion or anxiety like effects.

EGF can exert pleiotropic actions on many cells within the brain and it was important to determine whether brain EGF levels were increased after treatment. 30 min prior to sacrifice mice were treated with EGF and levels assessed by ELISA in the plasma (Fig. 3l, Student’s t-test), cortex, and hippocampus (Fig. 3m). Plasma EGF levels were ~4 pg/ml in VC treated mice and ~ 250 ng/ml after EGF treatment. There were no differences in hippocampal or cortical EGF levels between VC and EGF treated mice. In addition, when assessed by western blot analysis EGF receptor levels were not altered by EGF treatment in the hippocampus or cortex of E4FADF mice (Fig. 3n and Additional file 1: Figure S1d). However, we cannot exclude that there were cell-specific changes in EGF receptor levels after treatment. Therefore, the beneficial effects of EGF on cognition were likely mediated by peripheral rather than CNS levels of EGF.

EGF treatment lowers microbleeds and increases CV coverage compared to the VC

Next the effect of EGF on microbleed coverage was assessed. Microbleeds were lower in EGF treated mice compared to the VC. The percent area covered by microbleeds was 55 % lower in the cortex and 66 % lower in the hippocampus in the EGF treated group (Fig. 4a, Student’s t-test). A sub-goal of this study was to determine whether EGF modulates CV coverage in vivo. Thus laminin-Aβ co-staining was conducted (Fig. 4b, Student’s t-test). Visually EGF treated mice had a higher area covered by laminin staining than the VC in the deep layers of the cortex and the subiculum (Fig. 4b–c). Quantification of the laminin stain validates this finding as the percent area covered was 15 % (deep layer cx) and 25 % (subiculum) higher than the VC (Fig. 4d). To further dissect the role of EGF on CV coverage, quantitative IHC was conducted for CD31, a brain endothelial cell marker. In EGF treated mice CD31 coverage was 36 % higher in the deep layer cortex and 90 % higher in the subiculum compared to the VC (Fig. 4f, Student’s t-test). IHC staining for claudin 5 positive vessels revealed a similar staining pattern as for CD31 (Additional file 1: Figure S1a). Vessel coverage (% area covered) is a combination of density and length. There were no changes in laminin (Additional file 1: Figure S1b) or CD31 density (Additional file 1: Figure S1c) between VC and EGF treated mice. These data support that EGF prevents the lowering of CV length in E4FADF mice. It is important to note that distinguishing CV length and density with the IHC techniques utilized here is not straightforward. The optimal technique for assessing length and density entails three dimensional reconstruction of larger brain volumes with read read-outs for multiple parameters including branching points, major nodes, vessel length between nodes and the distribution of vessel lengths. Therefore, we prefer the term CV coverage for the purpose of this manuscript.

EGF has no effect on Aβ levels

As EFAD mice are a model of APOE modulated Aβ levels and induced dysfunction, we next assessed whether EGF modulates Aβ levels. There were no differences in extracellular amyloid (ThioS, Fig. 5a, Student’s t-test) or extracellular Aβ (Fig. 4b–e and Fig. 5b, Student’s t-test) staining between the EGF and VC treated mice. EGF also had no effect on soluble, detergent soluble or insoluble Aβ42 levels in the cortex or hippocampus when assessed biochemically (Fig. 5c and d). More recent data suggest that soluble oligomeric Aβ is more AD-relevant than extracellular Aβ, however EGF treatment did not modulate soluble oligomeric Aβ levels (Fig. 5e, Student’s t-test). Further, there were no differences in apoE levels between EGF and VC treated mice in any extracted fractions (Fig. 5f–g). Therefore, the protection of EGF on cognitive and CV length was unlikely to be mediated by lowering Aβ42 or changing apoE levels.

Sex and APOE4-induced cognitive and CV dysfunction in EFAD mice

Our data support that sex and APOE4 impact cognition and CV dysfunction in EFAD mice. In E4FADF the CV dysfunction was accompanied by cognitive dysfunction. In E3FADF mice CV leakiness was higher, CV coverage lower and plasma EGF levels lower compared to EFADM mice. As there were no changes in cognitive function in E3FADF mice compared to EFADM mice, the CV effects may represent early changes that later manifest as cognitive decline. These findings raise important questions on the relationship to human and other in vivo data and the mechanisms underlying sex-induced APOE modulated changes in cognition and CV coverage.

In AD patients APOE4-induced risk is greater in females [2, 19, 20, 41, 45] consistent with our finding that E4FADF are cognitively impaired compared to all other groups tested at 8 months. In partial contrast a few reports have demonstrated impaired cognition in male APOE4-TR mice (which do not overproduce human Aβ) compared to APOE3-TR mice [46, 56]. However, in a separate detailed analysis of APOE-TR mice the deficits in memory were primarily observed in female APOE4-TR mice [21]. Data on CV coverage is lacking in human patients stratified for sex and APOE genotype. In 9 and 12 months old APOE-TR mice lower microvessel coverage has been demonstrated [1, 4], although the effect of sex is unclear. A detailed comparison of EFAD and APOE-TR mice may reveal whether Aβ accelerates the APOE4 and female sex driven cognitive deficits and changes in CV coverage. In AD microbleeds are associated with male sex, higher blood pressure, lower CSF Aβ42 and APOE4 [5]. These data contrast with the female bias observed in EFAD mice [12, 20]. A confounding factor is whether additional AD-risk factors are present in male AD patients assessed for microbleeds. For example, level of exercise, hypertension, hyperglycemia, hyperinsulinemia and insulin resistance observed in type 2 diabetes all modulate AD-risk and the CV. Further, men are at a higher risk of developing cardiovascular events. The controlled laboratory conditions utilized here may be more relevant for studying the interaction among sex, APOE and Aβ in the absence of additional risk factors. Further research could focus on introducing additional AD-risk factors in vivo.

Females are at a higher risk for AD, which is often attributed to the loss of sex hormones during the menopause [45]. In the mouse model utilized here, with no ovariectomy, there may be inherent sex-differences that modulate CV and AD-like pathology. Extracellular Aβ levels from highest to lowest were: E4FADF > E4FADM = E3FADF > E3FADM consistent with previous reports [12, 51, 61]. Thus, higher levels of Aβ in E4FADF mice could induce or accelerate signaling cascades that cause CV degeneration. As described above CV coverage is lower in APOE4-TR mice which supports an acceleration rather than primary causation of CV deficits caused by Aβ. The low CV coverage and EGF levels in E3FADF mice compared to E4FADM mice are not associated with higher Aβ levels, supporting an Aβ-level independent mechanism. Females prior to menopause are at an increased risk for some autoimmune diseases and there are a number of proposed mechanisms underlying these effects [28]. For example, females may have a greater immune/inflammatory response in the periphery and the brain [28]. Although potentially beneficial for fighting infection an enhanced or prolonged immune/inflammatory response in females may predispose, or potentiate Aβ-induced direct and indirect damage to the CV. It is also noteworthy that lower vessel coverage was associated with low plasma EGF levels in EFADF mice. Low EGF plasma appear to increase the conversion of MCI to AD [33] but a sex bias has not been reported. Early reports in mice have demonstrated lower plasma EGF levels in females compared to males [42]. Further, plasma EGF levels in female mice are altered by the circadian rhythm, pregnancy, [53] and are increased with testosterone treatment [52]. Although 17-β estradiol is protective against brain endothelial damage in vitro [38], perhaps the male sex hormones exert a greater protective effect. A possibility is that low EGF levels in females result in lower protection of the CV to damage in AD, which is potentiated by APOE4 and accelerated by Aβ. Speculating further, the low plasma levels of EGF in female mice may be indicative general growth factor suppression or a form of accelerated aging/senescence.

The effects of APOE in AD are multifactorial, complex at the structural and mechanistic level and an area of intense research (reviewed extensively in [23, 34, 37, 43, 55, 59]). On a simplified level apoE4 can effect CV length through modulating Aβ levels and via signaling to cells at the blood-brain barrier (BBB) to alter Aβ-dependent and independent effects. As discussed, there are higher Aβ levels in APOE4 AD patients and E4FADF mice in this study. Aβ can cause vessel disruption by directly signaling to brain endothelial cells and indirectly via neuronal dysfunction and activation of astrocytes, microglia and pericytes. ApoE also signals to different cells types in the brain to disrupt vessel length. In astrocytes and microglia apoE4 is associated with a detrimental stress-induced (including Aβ) neuroinflammatory response [49]; mediators described as inflammatory can affect brain endothelial function (e.g. cytokines, matrix metalloproteases). In pericytes apoE4 is less effective at suppressing motility [15] and preventing MMP9 production [4, 22], both of which disrupt total vessel length. ApoE4 also directly disrupts brain endothelial cell tight junctions and effects peripheral lipid metabolism to cause brain endothelial dysfunction.

Overall, the female sex-induced, APOE4 modulated, Aβ accelerated changes in homeostatic signaling at the CV may ultimately converge to predispose and/or amplify stress-induced brain endothelial damage. Our ongoing studies are focused on delineating the role of APOE, aging, sex and peripheral risk factors on CV length.

EGF prevents cognitive and CV dysfunction in E4FADF mice

An important finding here is that EGF prevented CV dysfunction in E4FADF mice. These data are in agreement with the lower EGF plasma levels observed in humans and the beneficial effects of EGF in models of stroke and traumatic brain injury. In contrast improved cognition after blocking the tyrosine kinase activity of the EGF receptor (EGFR) with gefitinib has been reported in drosophilia and AD-Tg mice [57]. Aside methodological differences that commonly impact data (e.g. mouse model, time of treatment), the specificity of gefitinib for the EGFR has been drawn into question. Recently, gefitinib was demonstrated to antagonize a number of G-protein coupled receptors including adrenoreceptors, chemokine receptors, histamine receptors and other neuronal receptors [60]. Gefitinib is reported as a brain penetrant and therefore its beneficial effects in AD-Tg mice could be mediated by non-EGFR targets that play a role in AD-like pathology.

As EGF induced a pronounced effect on preventing cognitive decline in E4FADF mice, it is important to consider potential mechanisms of action. Our data support that the primary mode of action is not on cells within the brain, since brain EGF levels were not altered after EGF treatment. One caveat is that the technique used to measured EGF (ELISA of whole cortex or hippocampus) may have failed to detect regional high concentrations of EGF. Further, EGF did not modulate Aβ levels unlike VEGF in AD-Tg mice [58]. Although VEGF lowers Aβ levels in slice cultures [11] a role for EGF in this process has not been assessed. As EGF did not change Aβ levels, one hypothesis is that EGF acts directly on brain endothelial cells to prevent disrupted signaling induced by the interactive effects of female sex, APOE4 and Aβ in E4FADF mice. This is consistent with our in vitro data [31], the beneficial effect of EGF in wound healing, the promotion of angiogenesis in cancers that produce high EGF levels or that contain a mutation in erbb2 (a dimerization receptor partner of the EGFR), the protective effect of EGF-induced angiogenesis in stroke models and the general protective function of EGF in epithelial tissue (reviewed in [16]). Further, during the course of treatment EGF may have prevented a loss of key proteins involved in the homeostatic functions of brain endothelial cells, prior to vessel degeneration. For example key proteins involved in the nutrient transport/signaling molecule supply, waste product removal and efflux/metabolic/structural barrier functions of brain endothelial cells. For microbleeds a key question remains on whether EGF prevented global leakiness deficits in capillaries (e.g. via preventing tight junction changes or basement membrane degradation) and/or dysfunction at the post capillary venule level (the site of cellular trafficking).

An alternative hypothesis is that EGF activated receptors outside of brain endothelial cells in the CNS and/or the periphery to cause functional effects that prevent changes in CV coverage. For example, subtle increases in brain EGF levels may induce a beneficial response in pericytes, astrocytes, microglia or neurons to modulate levels of soluble molecules that maintain CV length. In the periphery EGF can signal in cells of virtually every organ in the body [15] and is associated with changes in metabolism and insulin signaling, both of which may protect the CV. An indicator of the peripheral effects is the lower body weight in EGF treated mice. EGF may also prevent cognitive decline independent of any CV changes through the same CNS and peripheral processes described above, in addition to promoting neurogenesis. Our data provide the basis for further studies dissecting if brain endothelial cells are the primary target of EGF.

CV length and AGFs as potential therapeutic targets in AD

What do the data mean in the context of CV length in AD and therapeutic options for AD patients? There is no clear in vitro or in vivo consensus on whether there is higher or lower CV in AD. One proposal is that lower CV density contributes to AD progression by disrupting cerebral blood flow and the complex transport and metabolic systems of the CV [18, 25, 29, 39]. The counter argument is that AD-pathways induce angiogenesis, causing hypersprouting and tight junction disruption, which increase CV permeability [7, 9, 13]. Traditional angiogenic signaling, typically observed over the short-term is likely not applicable to a chronic neurodegenerative condition such as AD. Pathological angiogenesis, described as the angiogenesis hypothesis [3] may be more applicable for AD: AD-processes not only cause direct vessel degeneration but also induce signaling cascades consistent with increased angiogenesis, however there is a failure to complete maturation and vessel formation. AD-risk factors both in the CNS and periphery could predispose and amplify the disrupted angiogenic signaling in brain endothelial cells.

There is a drive to develop novel strategies for AD as an alternative or complimentary approach for Aβ or tau-based therapeutics. AGF-like drugs may meet this criteria, yet are frequently overlooked in drug development programs. Indeed, an alternative focus is proposed of using therapeutics to block angiogenesis-like processes [54]. Our data with EGF and published data with other factors that increase total vessel coverage support an alternative approach [3, 26, 27, 32], which is to stabilize brain endothelial cells and maintain or increase vessel length. Further research is required to fully dissect the use of AGF-based treatments for AD. Key research areas include the potential for AGFs to reverse AD-like pathology in vivo, distinguishing CV versus peripheral effects and dissecting whether EGF-like molecules act on cells within the brain (i.e., would a brain penetrant AGF-like drug or treatment protocol be more efficacious than non-CNS penetrant). Further research is needed to address the concerns that agonists for AGF receptors could increase cancer risk. Cancer cells themselves are often the origin of high AGF levels; therefore finding a dosing strategy that negates this concern could be important in the elderly AD population.