Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Feb 1.
Published in final edited form as: Neurobiol Aging. 2021 Oct 30;110:73–76. doi: 10.1016/j.neurobiolaging.2021.10.015

Expression and proteolytic processing of the amyloid precursor protein is unaffected by the expression of the three human apolipoprotein E alleles in the brains of mice

Mariah J Novy 1, Samantha F Newbury 1, Braison Liemisa 1, Jose Morales-Corraliza 1,2, Melissa J Alldred 1,2, Stephen D Ginsberg 1,2,3,4, Paul M Mathews 1,2,3
PMCID: PMC8758539  NIHMSID: NIHMS1762164  PMID: 34875506

Abstract

The three human apolipoprotein E (APOE) gene alleles modify an individual’s risk of developing Alzheimer’s disease (AD): compared to the risk-neutral APOE ε3 allele, the ε4 allele (APOE4) is strongly associated with increased AD risk while the ε2 allele is protective. Multiple mechanisms have been shown to link APOE4 expression and AD risk, including the possibility that APOE4 increases the expression of the amyloid precursor protein (APP) (Y-W.A. Huang, B. Zhou, A.M. Nabet, M. Wernig, T.C. Südhof, 2019). In this study, we investigated the impact of APOE genotype on the expression and proteolytic processing of endogenously expressed APP in the brains of mice humanized for the three APOE alleles. In contrast to prior studies using neuronal cultures, we found in the brain that both App gene expression and the levels of APP holoprotein were not affected by APOE genotype. Additionally, our analysis of APP fragments showed that APOE genotype does not impact APP processing in the brain: the levels of both α- and β-cleaved soluble APP fragments (sAPPs) were similar across genotypes, as were the levels of the membrane-associated α- and β-cleaved C-terminal fragments (CTFs) of APP. Lastly, APOE genotype did not impact the level of soluble amyloid beta (Aβ). These findings argue that the APOE-allele-dependent AD risk is independent of the brain expression and processing of APP.

Keywords: Alzheimer’s disease, APP, apolipoprotein E

1.1. Introduction

Apolipoprotein E (APOE) participates in the transport of cholesterol-containing lipoproteins between cells, and allelic-variants of APOE contribute to Alzheimer’s disease (AD) risk (Liu et al., 2013). The three APOE alleles differentially impact the risk for AD, with the ε2 allele (APOE2) considered protective against AD, the ε3 allele (APOE3) risk-neutral, and the ε4 allele (APOE4) correlated with increased risk. Recently, Huang et al. (Huang et al., 2019) have suggested that APOE4 drives an increase in β amyloid precursor protein (APP) levels, which could contribute to increased Aβ generation independently of an impact of APOE4 on Aβ clearance. Using neuron-differentiated iPSCs expressing each APOE allele, this group reports that APOE4 genotype, compared to APOE3 and APOE2, increases APP gene and protein expression through activation of the MAP kinase pathway (Huang et al., 2019). Prior studies have reported that APOE genotype did not impact the levels of transgene-derived APP in the brains of mouse models (Bales et al., 2009; Castellano et al., 2011), and some previous in vitro studies have similarly found that no allele-specific effect of APOE on APP processing exists (Biere et al., 1995). Given that these studies have been in vitro or have relied upon transgene expression models, we examined the expression of the endogenous App gene as well as APP expression and processing in the brains of APOE replacement models.

1.2. Materials and Methods

1.2.1. Mouse brain homogenization

All experimental procedures involving animals were approved by the Institutional Animal Care and Use Committee of the Nathan S. Kline Institute. The APOE targeted-replacement mice used in this study are on a C57Bl/6 background and were homozygous for expression of the human APOE2, APOE3 or APOE4 alleles in place of the mouse apoE gene (Sullivan et al., 1997). These animals do not develop brain Aβ plaques or tau pathology (Sullivan et al., 1997; Tai et al., 2011). Hemibrains lacking the cerebellum and olfactory bulbs were dissected and snap-frozen on dry ice and later homogenized as previously described (Morales-Corraliza et al., 2016). Homogenized brain tissue was aliquoted and stored at −80°C until use. Data are from 4 male and 4 female mice of each genotype unless otherwise specified in a figure legend; sex balance was maintained at 4:3 or 3:3 following outlier analysis. Protein biochemistry data are from 18-month-old mice, as we and others have reported the aging-dependence of APOE-driven phenotypes in these mice (Peng et al., 2019). Mice used for mRNA analyses were 12 months of age, 3 males and 3 females (Peng et al., 2019).

1.2.2. APP expression and processing

qPCR was performed on reverse transcribed RNA extracted from hemibrain homogenates as previously described (Peng et al., 2019). Samples were assayed in triplicate using TaqMan qPCR primers for APP (ThermoFisher, Waltham, MA; Mm01344172_m1) and Sdha as a housekeeping gene (ThermoFisher; Mm01352360_m1) on a qPCR cycler (7900HT; ThermoFisher) in 96-well optical plates with coverfilm as previously described (Alldred et al., 2012; Ginsberg et al., 2010).

Equal amounts of brain homogenate proteins were determined by BCA (bicinchoninic acid) Protein Assay Kit (Thermo Fisher Scientific), separated on 4–20% Tris-HCL electrophoresis gels (Bio-Rad, Hercules, CA) and transferred to 0.45μm PVDF (ThermoFisher) membranes. Membranes were incubated overnight at 4°C with 1μg/mL of the APP antibodies C1/6.1 (binds the cytoplasmic C-terminus of APP), m3.2 (recognizes murine APP holoprotein, sAPPα, βCTF, and Aβ) (Morales-Corraliza et al., 2009), or 22C11 (Millipore #MAB348, Burlington, MA). Band intensity was analyzed with ImageJ (NIH, Bethesda, MD) and normalized to β-Actin levels (Abcam #ab8227, Cambridge, UK), which we have found to be a reliable reference protein in mouse brain, including APOE models (Peng et al., 2019).

An aliquot of each mouse brain homogenate was treated with diethylamine (DEA) to separate soluble proteins (sAPP and soluble Aβ) from membrane-bound proteins (full length APP and APP-CTFs) as detailed in Schmidt et al. (Schmidt et al., 2012b) and experimentally employed in Morales-Corraliza et al. (Morales-Corraliza et al., 2009). To immunoprecipitate sAPPα, DEA extracted samples were incubated overnight at 4°C with m3.2 antibody (Morales-Corraliza et al., 2009). The remaining flow-through contains sAPPβ; both sAPP forms were then detected by Western blot analysis using 22C11. Aβ was assayed from DEA extracted brain homogenate, which we have reported fully recovers endogenous murine brain Aβ in conditions without β-amyloid deposition (Herzig et al., 2004; Schmidt et al., 2012b). Aβ was quantified by sandwich ELISA as previously described (Schmidt et al., 2012a) and read against synthetic murine Aβ standards. We have previously reported our ability to detect subtle differences in the level of brain-derived endogenous murine Aβ (Mαστρανγελo ετ αλ., 2005; Moραλεσ−Xoρραλιζα ετ αλ., 2009) using in-house sandwich ELISAs sensitive to ~10 fmol Aβ/ml, assays in which the endogenous, DEA-extracted Aβ is typically measured in the range of ~50–200 fmol/ml (Schmidt et al., 2012a).

1.2.3. Statistical analysis

NIH ImageJ (http://rsb.info.nih.gov) was used to quantify Western blots. Images were converted to grayscale (8-bit) and contrast was adjusted for background subtraction. A rectangular box was drawn across the row containing the bands of interest and the lanes were plotted. The base of each peak was enclosed using the straight-line tool to threshold the signal intensity over background. The resulting area under each peak was quantified as the signal intensity. Genotype differences were analyzed with a two-tailed t-test as well as using One-Way ANOVA, with significance judged at p<0.05. A post-hoc comparison of male and female results showed no significant differences between the sexes, although this is inherently limited by the small sample size for each sex (n = 4 prior to outlier analysis); the data presented are pooled from both sexes. To help assure that a lack of an APOE-genotype effect was not due to outliers, an outlier analysis was performed with the criteria for removal being two standard deviations from the mean. This analysis led to the removal of two samples in the quantification of total sAPP and one sample in the quantifications of sAPPα and sAPPβ and m3.2 probed βCTF. Subsequent ROUT (Q=1%) analysis, done using GraphPad Prism version 9.1.2 for Mac (GraphPad Software, San Diego, California USA), identified the same outlier data points.

1.3. Results & Discussion

1.3.1. Brain APP expression is unaffected by APOE genotype.

APOE4, the single most significant genetic contributor to AD, is associated with increased β-amyloid deposition, both in animal models and in humans (Kim et al., 2009; Liu et al., 2013). Multiple mechanisms likely contribute to the AD-risk conferred by APOE4, including the possibility that APOE genotype modulates APP expression and therefore Aβ production (Huang et al., 2019). To determine if APOE genotype affects the expression of APP in vivo, we examined the endogenously expressed murine APP and its processing metabolites in the brains of humanized APOE targeted-replacement mice using antibodies and techniques extensively characterized by our group (Mathews et al., 2002; Morales-Corraliza et al., 2009). Additionally, examining endogenous APP allows for the identification of APP gene expression changes driven by APOE genotype in the context of cell-type appropriate APOE expression, lipidation, and uptake within the brain.

Western blot analysis was used to determine levels of APP holoprotein in APOE2, APOE3, and APOE4 brain homogenates (Fig. 1A; a representative Western blot is shown). Quantitation showed that brain APP protein levels were not significantly different between APOE genotypes. While uniform brain APP protein levels are consistent with APOE genotype not driving changes in APP expression, an APOE4-driven increase in APP expression might be offset by an increase in APP proteolytic processing. Therefore, and given the findings reported by Huang et al. (Huang et al., 2019) where APOE4 genotype was linked to increased APP mRNA in cultured iPSC neurons, we additionally analyzed APP mRNA levels in the brain. Quantitative PCR of APP in APOE2, APOE3 and APOE4 mouse brains showed no differences between genotypes (Fig. 1B). Our in vivo findings in a highly relevant system, in which the cell-type expression and brain metabolism of human APOE is preserved, argue that the AD-risk associated with APOE4 genotype prior to β-amyloid deposition is independent of a brain-wide increase in APP expression.

Figure 1. Endogenous APP protein and mRNA levels in APOE2, APOE3, and APOE4 mouse brain.

Figure 1.

Open in a new tab

(A) Representative Western blot analyses of APP holoprotein in brain homogenates prepared from 18-month-old homozygous APOE2, APOE3, and APOE4 mice using the APP C-terminal mAb C1/6.1(Mathews et al., 2002). Quantification of Western blots shows no differences between APOE genotypes. Equal amounts of proteins were loaded for SDS-PAGE, with Western blot data subsequently normalized to β-actin and expressed as the mean ± SEM relative to APOE3; n = 8 for each genotype. Analysis of APP knockout mouse and BACE1 knockout mouse (which lacks βCTF, see also Fig. 2) are shown as specificity controls. (B) Quantitative PCR analysis of homozygous APOE2, APOE3, and APOE4 mouse brain of APP mRNA levels normalized to the housekeeping gene Sdha; n = 6. No significant change was seen in APP mRNA levels by APOE genotype.

1.3.2. Brain APP proteolytic processing is unaffected by APOE genotype

We additionally analyzed the post-translational processing of APP, which proceeds by either plasma membrane α-cleavage or endosomal amyloidogenic β-cleavage (Zhang et al., 2012; Zheng and Koo, 2011). Both pathways give rise to soluble APP fragments (sAPPα and sAPPβ, respectively), which, due to their stability within the brain, are useful reporters of in vivo α- and β-cleavage rates and more informative than the rapidly degraded, more difficult to detect and less abundant CTFs (Morales-Corraliza et al., 2009). The cell-associated C-terminal fragments resulting from these cleavages (αCTFs and βCTFs, respectively) are less stable, with subsequent γ-cleavage of the βCTF giving rise to Aβ. We determined the brain levels of these APP metabolites in the APOE2, APOE3 and APOE4 mice. Western blot analysis of brain homogenate using two antibodies showed no significant differences between the levels of αCTFs and βCTFs or in the ratio of βCTF to αCTF (Fig. 2A) by APOE genotype. Brain levels of total sAPP were determined by Western blot analysis of soluble APP metabolites using 22C11, while sAPPα and sAPPβ levels were determined by immuno-isolation of sAPPα and subsequent 22C11 Western blot analysis to detect the two sAPP fragments as we have previously described (Morales-Corraliza et al., 2009) (Fig. 2B; representative Western blots shown). Consistent with APOE genotype not impacting APP holoprotein or mRNA levels, brain levels of total sAPP were similar among the three APOE genotypes (Fig. 2B). Moreover, no APOE-genotype differences were seen when sAPPα and sAPPβ were separately analyzed, offering additional support for the idea that APOE genotype does not alter α- or β-cleavage. Finally, soluble brain murine Aβ40 and Aβ42 levels were measured by sandwich ELISA, and no differences were observed across APOE genotypes (Fig. 2c). Multiple studies in mouse models have demonstrated that APOE genotype influences the aggregation and clearance properties of co-expressed human Aβ(Mαηoνεψ−Σανχηεζ ετ αλ., 2016). Murine Aβ, however, is inherently less prone to aggregation and accumulation (Herzig et al., 2004) and has a short half-life in the mouse brain (Morales-Corraliza et al., 2009). Thus, a lack of an APOE-genotype difference in Aβ is consistent with no difference in APP expression and processing.

Figure 2. Proteolytic processing of endogenous brain APP is unaffected by APOE genotype.

Figure 2.

Open in a new tab

(A) Western blot analysis using C1/6.1 showing αCTFs and βCTFs in homozygous APOE2, APOE3, and APOE4 mouse brain homogenates. Quantification of the bands corresponding to αCTFs, βCTFs, and a ratio calculation of βCTFs to αCTFs show no differences between APOE genotypes. Equal proteins were loaded, and Western blot data were normalized to β-actin prior to being expressed as the mean ± SEM relative to APOE3; αCTF: n = 8 for each genotype, βCTF: n = 8 for each genotype (C1/6.1), n = 7 for each genotype (m3.2). (B) Western blot analysis showing levels of the soluble extracellular fragment of APP generated after α- and β-cleavage (sAPPα and sAPPβ). Isolation of sAPPα and sAPPβ was done as previously described (Morales-Corraliza et al., 2009) from soluble brain proteins using mAb m3.2 to immunoprecipitate sAPPα (IP). The flow-through (FT) contains sAPPβ; both fragments were detected by Western blot analysis using mAb 22C11. Quantification of total sAPP, sAPPα, sAPPβ, and the ratio of sAPPα to sAPPβ show no differences across genotype. Data are expressed as the mean ± SEM normalized to APOE3; total sAPP: n = 6 for each genotype, sAPPα and sAPPβ: n = 7 for each genotype; 18-month-old mice of both sexes were analyzed. Analysis of APP knockout mouse and BACE1 knockout mouse (which lacks βCTF) are shown as specificity controls. (C) Soluble Aβ40 and Aβ42 levels were measured by Aβ sandwich ELISA of diethylamine (DEA) extracted-brain homogenates. No significant changes were found across genotypes. Data are expressed as the mean ± SEM normalized to brain weight; n=8 for each genotype; 18-month-old mice of both sexes were analyzed.

Our findings demonstrate that APOE genotype has minimal impact on the expression and proteolytic processing of APP in the brain. The mouse models we studied are highly relevant to the human brain as APOE expression is driven by the endogenous promoter and APOE-mediated effects on APP gene expression should be apparent when examining the endogenous APP gene. Additionally, no changes in brain APP processing were detected. Thus, our results are consistent with the idea that APOE genotype initially influences β-amyloid pathology and AD-risk through mechanisms that are independent of increased APP expression.

Highlights.

  • Impact of the three human apolipoprotein E alleles examined in brain.

  • Apolipoprotein E4 does not alter amyloid precursor protein (APP) expression.

  • Apolipoprotein E4 does not alter amyloid precursor protein (APP) processing.

  • Apolipoprotein E4 Alzheimer’s disease risk is not due to increased APP expression.

Acknowledgements

We thank Dr. Monika Pawlik for her expert assistance with our mouse colonies. This work was supported by the National Institute of Health [P01 AG017617 and R01 AG057517] to P.M.M. and [R01 AG043375 and P01 AG014449] to S.D.G.

abbreviations:

APP

β-amyloid precursor protein

APOE

apolipoprotein E

αCTF

α-cleaved C-terminal fragment of APP

βCTF

β-cleaved C-terminal fragment of APP

sAPPα

α-cleaved soluble APP ectodomain

sAPPβ

β-cleaved soluble APP ectodomain

Aβ

amyloid-β

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Author Credit Statement

M.J.N.: Experimentation, Data analysis, Writing- Original draft preparation; S.F.N and B. L.: Additional experimentation; J.M.-C.: Methodologies and ELISA; M.J.A and S.D.G.: Experimentation for qPCR and associated data analysis; P.M.M.: Funding, Data analysis, Writing- Reviewing and editing.

We confirm that this brief communication was not published previously nor is under consideration for publication elsewhere. All authors have approved submission of this manuscript.

Competing Interests

These authors report no competing interests.

References

  1. Alldred MJ, Duff KE, Ginsberg SD, 2012. Microarray analysis of CA1 pyramidal neurons in a mouse model of tauopathy reveals progressive synaptic dysfunction. Neurobiol Dis 45(2), 751–762. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Bales KR, Liu F, Wu S, Lin S, Koger D, DeLong C, Hansen JC, Sullivan PM, Paul SM, 2009. Human APOE isoform-dependent effects on brain beta-amyloid levels in PDAPP transgenic mice. J Neurosci 29(21), 6771–6779. [DOI] [PMC free article] [PubMed] [Google Scholar]
  3. Biere AL, Ostaszewski B, Zhao H, Gillespie S, Younkin SG, Selkoe DJ, 1995. Co-expression of beta-amyloid precursor protein (betaAPP) and apolipoprotein E in cell culture: analysis of betaAPP processing. Neurobiol Dis 2(3), 177–187. [DOI] [PubMed] [Google Scholar]
  4. Castellano JM, Kim J, Stewart FR, Jiang H, DeMattos RB, Patterson BW, Fagan AM, Morris JC, Mawuenyega KG, Cruchaga C, Goate AM, Bales KR, Paul SM, Bateman RJ, Holtzman DM, 2011. Human apoE isoforms differentially regulate brain amyloid-beta peptide clearance. Sci Transl Med 3(89), 89ra57. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Ginsberg SD, Alldred MJ, Counts SE, Cataldo AM, Neve RL, Jiang Y, Wuu J, Chao MV, Mufson EJ, Nixon RA, Che S, 2010. Microarray analysis of hippocampal CA1 neurons implicates early endosomal dysfunction during Alzheimer’s disease progression. Biol Psychiatry 68(10), 885–893. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Herzig MC, Winkler DT, Burgermeister P, Pfeifer M, Kohler E, Schmidt SD, Danner S, Abramowski D, Sturchler-Pierrat C, Burki K, van Duinen SG, Maat-Schieman ML, Staufenbiel M, Mathews PM, Jucker M, 2004. Abeta is targeted to the vasculature in a mouse model of hereditary cerebral hemorrhage with amyloidosis. Nat Neurosci 7(9), 954–960. [DOI] [PubMed] [Google Scholar]
  7. Huang YA, Zhou B, Nabet AM, Wernig M, Sudhof TC, 2019. Differential Signaling Mediated by ApoE2, ApoE3, and ApoE4 in Human Neurons Parallels Alzheimer’s Disease Risk. J Neurosci 39(37), 7408–7427. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Kim J, Basak JM, Holtzman DM, 2009. The role of apolipoprotein E in Alzheimer’s disease. Neuron 63(3), 287–303. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Liu CC, Liu CC, Kanekiyo T, Xu H, Bu G, 2013. Apolipoprotein E and Alzheimer disease: risk, mechanisms and therapy. Nat Rev Neurol 9(2), 106–118. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Mahoney-Sanchez L, Belaidi AA, Bush AI, Ayton S, 2016. The Complex Role of Apolipoprotein E in Alzheimer’s Disease: an Overview and Update. J Mol Neurosci 60(3), 325–335. [DOI] [PubMed] [Google Scholar]
  11. Mastrangelo P, Mathews PM, Chishti MA, Schmidt SD, Gu Y, Yang J, Mazzella MJ, Coomaraswamy J, Horne P, Strome B, Pelly H, Levesque G, Ebeling C, Jiang Y, Nixon RA, Rozmahel R, Fraser PE, St George-Hyslop P, Carlson GA, Westaway D, 2005. Dissociated phenotypes in presenilin transgenic mice define functionally distinct gamma-secretases. Proc Natl Acad Sci U S A 102(25), 8972–8977. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Mathews PM, Jiang Y, Schmidt SD, Grbovic OM, Mercken M, Nixon RA, 2002. Calpain activity regulates the cell surface distribution of amyloid precursor protein. Inhibition of calpains enhances endosomal generation of beta-cleaved C-terminal APP fragments. J Biol Chem 277(39), 36415–36424. [DOI] [PubMed] [Google Scholar]
  13. Morales-Corraliza J, Mazzella MJ, Berger JD, Diaz NS, Choi JH, Levy E, Matsuoka Y, Planel E, Mathews PM, 2009. In vivo turnover of tau and APP metabolites in the brains of wild-type and Tg2576 mice: greater stability of sAPP in the beta-amyloid depositing mice. PLoS One 4(9), e7134. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Morales-Corraliza J, Wong H, Mazzella MJ, Che S, Lee SH, Petkova E, Wagner JD, Hemby SE, Ginsberg SD, Mathews PM, 2016. Brain-Wide Insulin Resistance, Tau Phosphorylation Changes, and Hippocampal Neprilysin and Amyloid-beta Alterations in a Monkey Model of Type 1 Diabetes. J Neurosci 36(15), 4248–4258. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Peng KY, Perez-Gonzalez R, Alldred MJ, Goulbourne CN, Morales-Corraliza J, Saito M, Saito M, Ginsberg SD, Mathews PM, Levy E, 2019. Apolipoprotein E4 genotype compromises brain exosome production. Brain 142(1), 163–175. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Schmidt SD, Mazzella MJ, Nixon RA, Mathews PM, 2012a. Abeta measurement by enzyme-linked immunosorbent assay. Methods Mol Biol 849, 507–527. [DOI] [PubMed] [Google Scholar]
  17. Schmidt SD, Nixon RA, Mathews PM, 2012b. Tissue processing prior to analysis of Alzheimer’s disease associated proteins and metabolites, including Abeta. Methods Mol Biol 849, 493–506. [DOI] [PubMed] [Google Scholar]
  18. Sullivan PM, Mezdour H, Aratani Y, Knouff C, Najib J, Reddick RL, Quarfordt SH, Maeda N, 1997. Targeted replacement of the mouse apolipoprotein E gene with the common human APOE3 allele enhances diet-induced hypercholesterolemia and atherosclerosis. J Biol Chem 272(29), 17972–17980. [DOI] [PubMed] [Google Scholar]
  19. Tai LM, Youmans KL, Jungbauer L, Yu C, Ladu MJ, 2011. Introducing Human APOE into Abeta Transgenic Mouse Models. Int J Alzheimers Dis 2011, 810981. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Zhang H, Ma Q, Zhang YW, Xu H, 2012. Proteolytic processing of Alzheimer’s beta-amyloid precursor protein. J Neurochem 120 Suppl 1, 9–21. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Zheng H, Koo EH, 2011. Biology and pathophysiology of the amyloid precursor protein. Mol Neurodegener 6(1), 27. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES