Brief Communication
Epidermal growth factor prevents oligomeric amyloid-b induced angiogenesis deficits in vitro
Journal of Cerebral Blood Flow & Metabolism 2016, Vol. 36(11) 1865–1871 ! Author(s) 2016 Reprints and permissions: sagepub.co.uk/journalsPermissions.nav DOI: 10.1177/0271678X16669956 jcbfm.sagepub.com
Kevin P Koster, Riya Thomas, Alan WJ Morris and Leon M Tai
Abstract Cerebrovascular dysfunction is a critical component of Alzheimer’s disease (AD) pathogenesis. Oligomeric amyloid-b42 (oAb42) is considered a major contributor to AD progression. However, data are limited on the role of oAb42 in brain endothelial cell vessel degeneration/angiogenesis, including the interaction with angiogenic mediators. Thus, the current study determined the effect of oAb42 on angiogenesis in vitro, utilizing single brain endothelial cell cultures and triple cultures mimicking the microvascular unit (MVU: brain endothelial cells, astrocytes, and pericytes). oAb42 dose-dependently reduced angiogenesis and induced vessel disruption. Critically, epidermal growth factor prevented oAb42-induced deficits, implicating angiogenic pathways as potential therapeutics for AD.
Keywords Alzheimer’s, angiogenesis, blood–brain barrier, cerebrovascular disease, endothelium, pericytes Received 17 March 2016; Revised 3 August 2016; Accepted 8 August 2016
Introduction Cerebrovascular (CV) dysfunction is emerging as a critical component of Alzheimer’s disease (AD).1 AD risk factors and pathogenic pathways can exert detrimental effects on the CV in humans, transgenic models that recapitulate some biological aspects of AD, and in vitro. However, the topic of CV dysfunction in AD is controversial2 including the extent, consequence for neuronal function and the therapeutic significance.3,4 An important question is whether AD pathways induce detrimental effects on the length of the CV, which is dependent on vessel degeneration and/or angiogenesis. Amyloid-b (Ab), particularly oligomeric Ab42 (oAb42), is considered a major contributor to AD progression. Thus, one approach to address the issues in the field is to determine the effect of Ab on vessel length in vitro. Unfortunately, in vitro studies have not produced a clear consensus on the role of Ab in vessel degeneration/angiogenesis, and work focused on the effect of oAb42 using single brain endothelial cell (BEC) cultures or in triple cultures designed to mimic the microvascular unit (MVU: BECs, astrocytes and pericytes) is limited.5 A further important
question is how angiogenic growth factors (GFs) interact with oAb42 to modulate vessel length. Key soluble angiogenic GFs include vascular endothelial growth factor (VEGF), epidermal growth factor (EGF), and fibroblast growth factor-b (FGFb). A number of studies support that these mediators can prevent detrimental Ab-induced changes in ECs; however, data on vessel length are lacking, with the exception of a beneficial effect for VEGF and conflicting results with FGFb.6–9 Further, current evidence for GF-induced effects on angiogenesis is derived from studies utilizing non-BECs, and Ab40 or Ab42 rather than oAb42. Therefore, the goal of this manuscript was to determine the effect of oAb42 and angiogenic GFs on tube formation and disruption in single BEC cultures, and triple cultures mimicking the MVU.
Department of Anatomy and Cell Biology, University of Illinois at Chicago, Chicago, IL, USA Corresponding author: Leon M Tai, Department of Anatomy and Cell Biology, University of Illinois at Chicago, 808 S.Wood St., M/C 512, Chicago, IL 60612, USA. Email:
[email protected]
1866
Materials and methods Cells, oAb42 and GFs hCMEC/D3 cells (BECs) were a kind gift from PierreOlivier Couraud (Inserm) and cultured as described previously.10 Twenty-four hours before experiments, BECs were serum- and GF-starved. Primary human pericytes and astrocytes (Sciencell) were cultured according to manufacturer protocol between passages 2 and 5. oAb42 was prepared as previously described11 and used at the indicated concentrations (0–100 nM). GFs (Shenandoah) were utilized at 100 nM.
Cell viability assays Cell viability was assessed in hCMEC/D3 cells with oAb42 GFs (24 hr) via the MTT assay12 (ThermoFisher).
Tube formation and disruption assays BEC single cultures. MatrigelÕ was added to 96-well angiogenesis m-plates (Ibidi) and allowed to set for 1 h at 37 C. At t ¼ 0 h, BECs preloaded with CellTrackerTM green (30 min 37 C, ThermoFisher) were added at 10,000 cells/well. Paradigm 1, simultaneous – All treatments; GFs, oAb42, vehicle controls (VCs) and combinations were added to the cells immediately following the addition of cells to the plate (t ¼ 0 h). Paradigm 2, spike/protection – GFs or VCs were added to the cells immediately following the addition of cells to the plate (t ¼ 0 h). After 4 h incubation at 37 C, oAb42 or VCs were added (t ¼ 4 h). Paradigm 3, reversal – oAb or VC were added to the cells immediately following the addition of cells to the plate (t ¼ 0 h). After 4 h incubation at 37 C, GFs or VCs were added (t ¼ 4 h). MVU triple cultures. Angiogenesis plates were prepared as described above. At t ¼ 0 h, BECs preloaded with CellTrackerTM green (ThermoFisher) were plated GFs or VCs. The following were added sequentially: at t ¼ 4 h, pericytes (2000 pericytes/well) preloaded with CellTrackerTM blue (ThermoFisher), at t ¼ 7 h, astrocytes (10,000 astrocytes/well) preloaded with CellTrackerTM orange (ThermoFisher), and at t ¼ 11 h, oAb42 or VC. The BEC:pericyte:astrocyte cellular ratio was 5:1:5.
Immunocytochemistry and imaging analysis All angiogenesis plates were fixed in 4% paraformaldehyde at t ¼ 24 h. Fluorescent images of BECs in each well were captured at 1.6 magnification and
Journal of Cerebral Blood Flow & Metabolism 36(11) analyzed using the ImageJ angiogenesis plug-in.13 Immunocytochemistry (ICC) was then performed as described previously10 using the Ab specific antibody MOAB-2 (1:400, Biosensis). To image BECs, astrocytes and pericytes, high magnification images were captured on a Zeiss Axio Imager M1 under identical capture settings at 5 or 10 magnification. Ab and pericyte coverage were assessed using 5 magnification images in imageJ (Figures 1 and 2).
Statistical analysis All data are presented as mean þ/ s.e.m and were analyzed using two-way AVOVA, one-way ANOVA, or Student’s t-test, as specified, in GraphPad Prism version 6 (Table 1).
Results EGF prevents direct oA42-induced disruption of tube formation in single BEC cultures (paradigm 1) Initially, the effect of non-toxic oAb42 concentrations (identified by MTT assay, Figure 1(a)) on tube formation was assessed in single cultures of immortalized human BECs (hCMEC/D3 cells) by simultaneously adding cells þ/oAb42 (VC, 25 nM, 50 nM, 100 nM) for 24 h to a MatrigelÕ matrix (Figure 1(b)). At nontoxic doses, oAb42 dose-dependently disrupted tube formation (Figure 1(c)). Compared to the VC, total vessel length was 20% and 16% lower with 50 and 100 nM oAb42, respectively. Further, the number of meshes was 10% (50 nM) and 40% (100 nM) lower with oAb42. Next, the effect of simultaneous oAb42 (0–100 nM) and GF treatment was determined (Figure 1(b)). EGF (100 nM), but not VEGF, FGFb, or IGF (data not shown) ameliorated oAb42-induced disruption of tube formation (Figure 1(c) and (d)). For example, compared to the VC, total tube length was 110% for combined oAb42-EGF treatment (Figure 1(c)). Interestingly, EGF-induced protection was not due to changes in oAb42 accumulation in the BECs (Figure 1(e)). These data support that direct oAb42 signaling prevents tube formation and EGF ameliorates this effect.
EGF prevents oA42-induced tube disruption in single BEC cultures (paradigm 2) To determine whether oAb42 causes tube degeneration, BECs were incubated for 4 h to form vessel networks, and oAb42 was spiked into the media for a further 20 h (Figure 1(f)). oAb42 reduced total vessel length by 20% and the number of meshes by 40% at both 50 and 100 nM (Figure 1(g)). EGF prevented the
Koster et al.
1867
Figure 1. oAb42 inhibits tube formation and induces vessel disruption in BEC cultures, prevention by EGF. (a–e) Treatment Paradigm 1 – Simultaneous. (a) hCMEC/D3 viability in response to oAb42 at 0, 100 nM, and 5mM with and without GFs, as assessed by MTT assay (****p < 0.0001 by two-way ANOVA, main effect of concentration). (b) Schematic representation of the workflow. BECs were plated and treated with oAb42 (0–100 nM) alone or in combination with VEGF (100 nM), FGFb (100 nM), or EGF (100 nM). (c) Total vessel length and number of meshes formed by BECs quantified using 1.6 magnification images (n ¼ 3 with 2–4 replicates, *p < 0.05, ****p < 0.0001 by standard two-way ANOVA, main effect of treatment group; p < 0.05 by Dunnet’s multiple comparison (continued)
1868 oAb42-induced vessel disruption to BECs (Figure 1(g) and (h)). Indeed, compared to 50 and 100 nM oAb42, total vessel length values were 20% and 35% higher with combined oAb42-EGF treatment, respectively. Further, with combined EGF and oAb42 treatment, the number of meshes were 88% (50 nM oAb42) and 61% (100 nM oAb42) higher compared to oAb42 alone (Figure 1(g)). As for tube formation, the beneficial effects of EGF were not related to alterations in oAb42 uptake or degradation (Figure 1(i)). For other GFs, there was a non-significant trend for VEGF to prevent oAb42-induced tube disruption and interestingly, FGFb in combination with oAb42-induced more robust vessel disruption compared to oAb42 alone (Figure 1(g)).
EGF does not reverse oA42-induced disruption of tube formation in single BEC cultures (paradigm 3) To determine whether EGF could reverse oAb42induced tube formation disruption, BECs were incubated oAb42 from 0–4 h and GFs spiked into the media for a further 20 h (Figure 1(j)). As in paradigm 1, oAb42 (0–100 nM) disrupted tube formation (Figure 1(k)). However, with the exception that EGF appeared to reverse oAb42-induced disruption in total vessel length at 50 nM oAb42, EGF was largely unable to reverse oAb42-induced tube disruption in this paradigm (Figure 1(k) and (l)).
EGF prevents oA42-induced tube disruption in MVU cultures Although our data demonstrates that oAb42 induces detrimental effects on tube formation directly in BECs, it is important to conduct complementary analysis using the main cell types of the MVU, which can
Journal of Cerebral Blood Flow & Metabolism 36(11) respond to oAb42 to collectively determine vessel length. Therefore, we next investigated the effects of oAb42 and GFs in cultures containing BECs, primary human pericytes and primary human astrocytes. We focused on the oAb42-induced effects on preformed vessels to mimic the in vivo scenario of oAb42 accumulation with an established CV (Figure 2(a)). oAb42 treatment resulted in tube disruption only at the highest dose (100 nM) causing a 20% reduction in number of meshes and 13% reduction in total tube length (Figure 2(b)). As 50 nM oAb42 caused disruption in single BEC cultures, these data imply that BECs are more resistant to insult within the MVU (proposed in Itoh et al.14). At high magnification, pericyte coverage at tube junctions and tube projections appeared lower in oAb42-treated triple cultures (Figure 2(c)). Indeed, when quantified, 100 nM oAb42 reduced the percent area occupied by pericytes by 49% compared to VC (Figure 2(d)). As observed in BEC cultures, EGF protected against 100 nM oAb42-induced tube disruption, with a 20% higher total tube length and a 50% greater number of meshes compared to oAb42 treatment alone (Figure 2(b)). Interestingly, treatment with 25 and 50 nM oAb42 in combination with EGF resulted in higher in total length and number of meshes, despite oAb42 treatment alone at the same concentrations not causing any direct disruption. Further, EGF prevented the 100 nM oAb42-induced decrease in the number of and percent area covered by pericytes, while VEGF and FGFb had no effect (Figure 2(d)). These data suggest a direct link between pericyte health/coverage and the ability of BECs to form viable networks that is disrupted in the presence of oAb42. Mirroring data in single BECs, VEGF trended to protect against the oAb42-induced disruption at 100 nM.
Figure 1. Continued vs. oAb42 group. #p < 0.05 by ANOVA for both oAb42 50 and 100 nM vs. VC, $p < 0.05 by ANOVA for oAb42 100 nM vs. VC). (d) Representative 10 images of VC, VEGF, and EGF treatments (from left to right, top) and oAb42, oAb42 plus VEGF, and oAb42 plus EGF (from left to right, bottom. All images at 100 nM oAb42. Green ¼ BECs; red ¼ Ab42). (e) Percent area occupied by Ab42 (100 nM, n ¼ 3 with 2–4 replicates). (f–i) Treatment paradigm 2 – spike/protection. (f) Schematic representation of the workflow. In the spike/prevention treatment paradigm, BECs were treated with VEGF (100 nM), FGFb (100 nM), or EGF (100 nM) and oAb42 (0– 100 nM) spiked into the culture. (g) Quantification of total vessel length and number of meshes formed by BECs quantified using 1.6 magnification images (n ¼ 4 with 2–4 replicates, ****p < 0.0001 by standard two-way ANOVA, main effect of treatment group; p < 0.05, p < 0.001 by Dunnet’s multiple comparison vs. oAb42 group. #p < 0.05 by ANOVA for both oAb42 50 and 100 nM vs. VC). (h) Representative 10 images of VC, VEGF, and EGF treatments alone (from left to right, top) and oAb42, oAb42 plus VEGF, and oAb42 plus EGF (from left to right, bottom. All images at 100 nM oAb42. Green ¼ BECs; red ¼ Ab42). (i) Percent area occupied by Ab42 (100 nM, n ¼ 4 with 2–4 replicates). (j) Schematic representation of the workflow for the reversal paradigm. (k) Total vessel length and number of meshes formed by BECs quantified using 1.6 magnification images (n ¼ 5 in triplicate, #p < 0.05 by ANOVA for both oAb42 50 and 100 nM vs. VC, $p < 0.05 by ANOVA for oAb42 100 nM vs. VC, þ p < 0.05 by Student’s t-test vs. oAb42 50 nM). (l) Representative 10 images of oAb42, and oAb42 plus EGF (from left to right. All images at 100 nM oAb42. Green ¼ BECs). All data are expressed as mean s.e.m compared to oAb42 VC. Scale bar represents 100 mm.
Koster et al.
1869
Figure 2. oAb42 induces vessel disruption in MVU cultures, prevention by EGF. (a) Schematic representation of the workflow for the treatment paradigm in MVU cultures, treated with oAb42 (0–100 nM) alone or in combination with VEGF (100 nM), FGFb (100 nM), or EGF (100 nM). (b) Total vessel length and number of meshes formed by BECs quantified using 1.6 magnification images (n ¼ 3 in triplicate, **p < 0.01, ****p < 0.001 by standard two–way ANOVA, main effect of treatment group; p < 0.001, p < 0.0001 by Dunnet’s multiple comparison vs. oAb42 group. $p < 0.05 by ANOVA for oAb42 100 nM vs. VC). (c) Representative 10 images of VC, VEGF, and EGF treatments alone (from left to right, top) and oAb42, oAb42 plus VEGF, and oAb42 plus EGF (from left to right, bottom. All images at 100 nM oAb42. Green ¼ BECs; red ¼ astrocytes; blue ¼ pericytes). (d) Quantification of pericyte count and percent area occupied (n ¼ 3 in triplicate, &p < 0.05 compared to oAb42 VC; %p < 0.05 compared to oAb42 100 nM by Student’s t-test). All data are expressed as mean s.e.m compared to oAb42 VC. Scale bar represents 100 mm.
1870
Journal of Cerebral Blood Flow & Metabolism 36(11)
Table 1. Summary of statistics from all paradigms. Paradigm
Measure
Paradigm 1-simultaneous
Total length [3, 144] Number of meshes [3, 144] Total length [3, 144] Number of meshes [3,144] Total length [3, 224] Number of meshes [3, 224] Total length [3, 128] Number of meshes [3, 128]
Paradigm 2-spike
Paradigm 3-reversal
MVU paradigm T1-spike
p: treatment effect (Two-way ANOVA)
p: oAb42 vs VEGF (Dunnet’s)
p: oAb42 vs FGFb (Dunnet’s)
p: oAb42 vs EGF (Dunnet’s)
p < 0.0001**** p ¼ 0.0154*
p ¼ 0.4750 p ¼ 0.9762
p ¼ 0.0965 p ¼ 0.09891
p ¼ 0.0195 p ¼ 0.0466
p < 0.0001**** p < 0.0001****
p ¼ 0.9811 p ¼ 0.4336
p ¼ 0.0006 p ¼ 0.0161
p ¼ 0.0249 p ¼ 0.0003
p ¼ 0.1370 p ¼ 0.6259
p ¼ 0.1456 (DNS) p ¼ 0.6714 (DNS)
p ¼ 0.9847 (DNS) p ¼ 0.9847 (DNS)
p ¼ 0.3575 p ¼ 0.9996
p ¼ 0.0013** p < 0.0001****
p ¼ 0.5260 p ¼ 0.0747
p ¼ 0.6041 p ¼ 0.3438
p ¼ 0.0006 p < 0.0001
Note: Standard two-way ANOVAs were used to analyze across groups for treatment effect and concentration effect (no significance by two-way ANOVA). Post hoc multiple comparisons were made using Dunnet’s multiple groups correction, in which each treatment condition (GF, column effect) was compared to the oAb42 group. Degrees of freedom for each statistical test are displayed in brackets with the format [df1, df2]. *p < 0.05, **p < 0.01, ****p < 0.0001 by standard two-way ANOVA, main effect of treatment group. p < 0.05, p < 0.001, p < 0.0001 by Dunnet’s multiple comparison vs oAb42 group. DNS, data not shown.
Discussion Data are conflicted on whether vessel length is increased or decreased in AD and the mechanistic implications. There is evidence of reduced vessel density in AD patients15 and in AD-mice, data supported in select in vitro studies.16 Therefore, 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.17 The counter argument is that Ab induces angiogenesis, causing hypersprouting, tight junction disruption, and increased CV permeability.18 An important contributor to the in vitro discrepancies on the angiogenic effects of Ab may be related to the method of preparation and use of varying species of Ab. In the present study, oAb42 treatment consistently reduced vessel formation and caused vessel disruption in both BEC and MVU cultures. These data are particularly significant as oAb42 is considered the most active form of Ab in AD. Although the mechanisms are currently unknown, oAb42-induced vessel/BEC disruption is likely the result of both direct and indirect signaling. Directly to BECs, Ab can bind to a number of receptors, accumulate intracellularly, and induce senescence16 to impact angiogenic signaling. Indirectly, oAb42 activates proinflammatory signaling in astrocytes,19 dysregulating the angiogenic signaling in BECs required to form and maintain healthy tubes. This process may also translate to pericytes (Figure 2). Thus, although complex and more research is required, one potential explanation is: Ab induces signaling cascades consistent
with increased angiogenesis; however, rather than inducing angiogenesis, Ab halts the process, resulting in a failure to complete vessel formation in addition to causing direct damage. Dissecting the cellular signaling processes modulated both directly and indirectly by oAb42 may lead to a deeper understanding of CV dysfunction in AD. Pro-angiogenic GFs have shown promise for AD pathogenesis in vivo and in vitro. Our results demonstrate that EGF protects against oAb42-induced disruption of vessel formation and vessel disruption in BEC cultures and protects against oAb42-induced vessel and pericyte disruption in MVU cultures. Data are conflicted on whether EGF levels are higher or lower in AD. For example, higher plasma EGF levels have been reported in AD patients,20 whereas recent data demonstrate that low EGF levels predict conversion from amnestic mild cognitive impairment to AD.21 Potentially, lower EGF levels induce disrupted angiogenic signaling early in AD, causing an adaptive increase in EGF levels in a subset of patients and/or BECs to become unresponsive to EGF with age and pathological progression. Our ongoing research is focused on further dissecting the reversal potential of EGF in paradigms that more closely mimic the in vivo scenario. Evaluating the activity of EGF in AD-models and the cell-specific signaling cascades may lead to novel therapeutic approaches for AD. Funding The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this
Koster et al. article: Leon M. Tai is supported by University of Illinois at Chicago start-up funds.
Declaration of conflicting interests The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Authors’ contributions KPK conducted experimental procedures and was involved in the preparation of the manuscript. RT contributed to experimental procedures and experimental setup. AWJM conducted experimental procedures and was involved in the preparation of the manuscript. LMT was responsible for project oversight, experiment planning, data interpretation, and manuscript submission.
References 1. Love S and Miners JS. Cerebrovascular disease in ageing and Alzheimer’s disease. Acta Neuropathol 2016; 131: 645–658. 2. Bien-Ly N, Boswell CA, Jeet S, et al. Lack of widespread BBB disruption in Alzheimer’s disease models: focus on therapeutic antibodies. Neuron 2015; 88: 289–297. 3. Ambrose CT. A therapeutic approach for senile dementias: neuroangiogenesis. J Alzheimer’s Dis 2015; 43: 1–17. 4. Ambrose CT. Neuroangiogenesis: a vascular basis for Alzheimer’s disease and cognitive decline during aging. J Alzheimer’s Dis 2012; 32: 773–788. 5. Paris D, Townsend K, Quadros A, et al. Inhibition of angiogenesis by Abeta peptides. Angiogenesis 2004; 7: 75–85. 6. Patel NS, Mathura VS, Bachmeier C, et al. Alzheimer’s beta-amyloid peptide blocks vascular endothelial growth factor mediated signaling via direct interaction with VEGFR-2. J Neurochem 2010; 112: 66–76. 7. Cantara S, Donnini S, Morbidelli L, et al. Physiological levels of amyloid peptides stimulate the angiogenic response through FGF-2. FASEB J 2004; 18: 1943–1945. 8. Solito R, Corti F, Fossati S, et al. Dutch and Arctic mutant peptides of beta amyloid(1-40) differentially affect the FGF-2 pathway in brain endothelium. Exp Cell Res 2009; 315: 385–395. 9. Religa P, Cao R, Religa D, et al. VEGF significantly restores impaired memory behavior in Alzheimer’s mice by improvement of vascular survival. Sci Rep 2013; 3: 2053.
1871 10. Tai LM, Reddy PS, Lopez-Ramirez MA, et al. Polarized P-glycoprotein expression by the immortalised human brain endothelial cell line, hCMEC/D3, restricts apicalto-basolateral permeability to rhodamine 123. Brain Res 2009; 1292: 14–24. 11. Dahlgren KN, Manelli AM, Stine WB, Jr., et al. Oligomeric and fibrillar species of amyloid-beta peptides differentially affect neuronal viability. J Biol Chem 2002; 277: 32046–32053. 12. Tai LM, Holloway KA, Male DK, et al. Amyloid-betainduced occludin down-regulation and increased permeability in human brain endothelial cells is mediated by MAPK activation. J Cell Mol Med 2010; 14: 1101–1112. 13. Carpentier G MM, Courty J and Cascone I. Angiogenesis analyzer for imageJ. In: 4th imageJ user and developer conference proceedings, 198–201, Mondorf-les-Bains, Luxembourg, 24–26 October 2012. 14. Itoh Y, Toriumi H, Yamada S, et al. Astrocytes and pericytes cooperatively maintain a capillary-like structure composed of endothelial cells on gel matrix. Brain Res 2011; 1406: 74–83. 15. Kitaguchi H, Ihara M, Saiki H, et al. Capillary beds are decreased in Alzheimer’s disease, but not in Binswanger’s disease. Neurosci Lett 2007; 417: 128–131. 16. Donnini S, Solito R, Cetti E, et al. Abeta peptides accelerate the senescence of endothelial cells in vitro and in vivo, impairing angiogenesis. FASEB J 2010; 24: 2385–2395. 17. Bell RD and Zlokovic BV. Neurovascular mechanisms and blood-brain barrier disorder in Alzheimer’s disease. Acta Neuropathol 2009; 118: 103–113. 18. Biron KE, Dickstein DL, Gopaul R, et al. Amyloid triggers extensive cerebral angiogenesis causing blood brain barrier permeability and hypervascularity in Alzheimer’s disease. PLoS One 2011; 6: e23789. 19. Tai LM, Ghura S, Koster KP, et al. APOE-modulated Abeta-induced neuroinflammation in Alzheimer’s disease: current landscape, novel data and future perspective. J Neurochem 2015; 133(4): 465–488. 20. Biella G, Franceschi M, De Rino F, et al. Multiplex assessment of a panel of 16 serum molecules for the differential diagnosis of Alzheimer’s disease. Am J Neurodegen Dis 2013; 2: 40–45. 21. Lim NS, Swanson CR, Cherng HR, et al. Plasma EGF and cognitive decline in Parkinson’s disease and Alzheimer’s disease. Ann Clin Transl Neurol 2016; 3: 346–355.
