LRP3, an apolipoprotein receptor that links reelin signalling and APP expression, is affected in Alzheimer's disease

Background. Members of the low-density lipoprotein (LDL) receptor family are involved in endocytosis and in transducing signals, but also in APP (amyloid precursor protein) processing and β-amyloid secretion. ApoER2/LRP8 is a member of this family with key roles in synaptic plasticity in the adult brain. ApoER2 is cleaved after the binding of its ligand, the reelin protein, generating an intracellular domain (ApoER2-ICD) that modulates reelin gene transcription itself. In this work, we have analysed whether ApoER2-ICD is able to regulate the expression of other members of the LDL receptor family. We focused on LRP3, the most unknown member of the LDL receptor family, whose precise physiological role and potential participation in pathological processes such as Alzheimer’s disease (AD) are still unknown. Methods. The effects of full-length ApoER2 and ApoER2-ICD overexpression on protein levels, in presence of recombinant reelin or Ab42 peptide, were evaluated by a microarray, qRT-PCRs and western blots. The expression of LRP3 was analysed in human frontal cortex extracts from AD and non-demented subjects by qRT-PCRs and western blot; and LRP3 interaction with other proteins was assessed by immunoprecipitation. In CHO cells overexpressing LRP3, protein levels of full-length APP and fragments were evaluated by western blots. Results. ApoER2 E


Introduction
The members of the family of low-density lipoprotein (LDL) receptor are endocytic receptors that mediate the uptake of lipoproteins and have been classically studied for its role in cholesterol transport and metabolism. Robust evidences indicate that LDL receptor family members are involved in synaptic plasticity regulation and neuronal migration (extensively reviewed in [1][2][3][4][5][6]). LDL receptors are related to Alzheimer's disease (AD) pathogenesis as receptors of apolipoprotein E (apoE) [7], being the APOE4 variant the largest known genetic risk factor for late-onset sporadic AD [8,9]. Additionally, several members of the LDL receptors family are able to modulate the amyloid precursor (APP) proteolytic processing, either by regulation of the generation of the β-amyloid peptide (Aβ) or through Aβ clearance [10][11][12][13].
An important member of the LDL receptor family, ApoER2/LRP8, can exert a modulatory effect in transcriptional expression. ApoER2 interaction with its ligand, the reelin protein, drives to a sequential proteolytic processing, resulting in the cleavage of the receptor by α-secretase, that generates a membrane-tethered C-terminal fragment (ApoER2-CTF), followed by the cleavage by the γ-secretase. The action of γ-secretase generates an intracellular domain fragment (ApoER2-ICD) able to decrease the expression of reelin mRNA [14,15]. In brain extracts from AD subjects the generation of ApoER2-CTF appears lower and, accordingly, reelin expression results higher respect to those in control brain extracts [16].
In this study, we have explored further the modulatory transcriptional activity of ApoER2/reelin signalling, and we have observed that this pathway can modulate the expression of the LDL-related protein 3 (LRP3). LRP3 is probably the most unknown member of a new subfamily of LDL receptors [17], whose precise role in the central nervous system is still undetermined. We have explored LRP3 expression in Alzheimer´s disease brain and after overexpression in CHO cells, and we have demonstrated that LRP3 is able to modulate APP expression, as are other members of LDL receptors family.

Human brain samples
This study was approved by the ethics committee of Universidad Miguel Hernández de Elche, Spain, and it was carried out in accordance with the WMA Declaration of Helsinki. Brain samples (frontal cortex) were obtained from the Brain Bank of the Institute of Neuropathology, Bellvitge University Hospital. Lateonset AD cases (12 females/18 males; average age 75±1 years) were categorized according to the Braak stage of neuro brillary tangle (NFT) pathology ( [18]; Braak stage I-II n= 10; Braak stage III-IV, n= 10; Braak stage V-VI, n= 10). Special care was taken not to include cases with combined pathologies to avoid bias in the pathological series. Samples from non-demented (ND) controls (2 females/6 males; average age 52±7 years) corresponded to individuals with no clinical dementia and no evidence of brain pathology.
The mean post-mortem interval of the tissue was ~8 h in all cases, with no signi cant difference between the groups.

Microarray analysis
Gene expression was analysed 48 h after transfection with human full-length ApoER2, using microarrays SurePrint G3 Human Microarrays (ID 039494, Agilent Technologies, Spain) and performed by Bioarray SL (http://www.bioarray.es). The concentration and the purity of the total RNA extracted were measured by a NanoDrop spectrophotometer, and RNA quality was determined with the kit R6K Screen Tape (Agilent Technologies, Spain). The estimated RNA integrity number ranged between 9.5 and 9.7. Each sample (four samples and four controls) was labeled with Cy3 using the One-Color Microarray-Based Gene Expression Microarrays Analysis v.6.6 (Agilent Technologies, Spain). Data were imported to the linear models for microarray data Bioconductor software (Limma, Marray, affy, pcaMethods and EMA). Raw data were rst subjected to background subtraction, then to within-array loess normalization. Finally, across-array normalization was performed. Normalized data were tted to a linear model. The signi cance of the gene expression changes was analysed according to the adjusted p value (adj. p 0.05).
qRT-PCR analysis RNA was extracted from human brains, SH-SY5Y cells or CHO-PS70 cells using the TRIzol® Reagent in the PureLinkÔ Micro-to-Midi Total RNA Puri cation System (Life Technologies, Carlsbad, CA, USA) following the manufacturer's instructions. SuperScriptÔ III Reverse Transcriptase (Life Technologies, Carlsbad, CA, USA) was used to synthesize cDNAs from this total RNA (2 µg) using random primers according to the manufacturer's instructions. Quantitative PCR ampli cation was performed on a StepOneÔ Real-Time PCR System (Applied Biosystems, Thermo Fisher Scienti c, Rockford, USA) with TaqMan probes speci c for human LRP3 (assay ID: HS01041220_m1), LDLR (assay ID: HS00181192_m1), (Applied Biosystems, Thermo Fisher Scienti c, Rockford, USA), and human 18S as a housekeeping gene (Applied Biosystems, Thermo Fisher Scienti c, Rockford, USA) for human brain and SH-SYSH cells samples. In CHO-PS70, mRNA expression was measured with primers for human APP (forward: AACCAGTGACCATCCAGAAC; reverse: ACTTGTCAGGAACGAGAAGG), and for glyceraldehyde 3phosphate dehydrogenase (GAPDH, forward: AGAAGGTGGTGAAGCAGGCAT; reverse: AGGTCCACCACTCTGTTGCTGT) to normalize the expression levels of the target gene by the ΔCt method curves.

Recombinant reelin
HEK-293T cells stably transfected with reelin cDNA and GFP (mock) were grown in 175 cm 2 asks at a density of 10×10 6 cells / ask. After 3 days in culture with Optimem, the supernatants were ltered through 0.2 µm pores and then concentrated by an Amicon Ultra size 100 kDa-lter of molecular exclusion (Millipore). For quanti cation, a coomasie gel was loaded with different volumes of the concentrated supernatants as well as with different bovine serum albumin solutions to perform an extrapolation.
Immuno uorescence CHO-PS70 cells overexpressing LRP3-ag were washed with cold Hank-buffered salt solution and xed with 4% paraformaldehyde and 0.1 M EGTA for 10 minutes. Nonspeci c sites were blocked with 10% (w/v) bovine serum albumin for 30 minutes. Cells were incubated with anti-LRP3 Ct (mouse; Sigma-Aldrich) for 1 hour followed by the secondary anti-body (Cy5 anti-rabbit; GE-Healthcare) for 1 hour. After 2 washes with PBS, cells were incubated brie y with Hoechst dye to label nuclei (Invitrogen). Pictures were acquired in a Leica SPEII upright TCL-SL confocal microscope using an oil-immersion 40x objective.

Statistical analysis
The distribution of data was tested for normality using a D'Agostino-Pearson test. The data were analyzed using unpaired Student´s t test (a Welch´s correction was employed in data with different standard deviations) or Mann-Whitney test. The results are presented as the means± SE and all the analyses were performed using GraphPad Prism (Version 7; GraphPad Software, Inc). p value < 0.05 was considered signi cant.

Results
ApoER2 overexpression increases expression of LRP3, another lipoprotein receptor.
SH-SY5Y cells were transfected with full-length ApoER2 and after 48 hours, a microarray was performed.
Among the genes affected, we focused in the analysis of LDL receptors and apolipoprotein related genes ( Table 1). The receptors LRP3 and LDLR appeared signi cantly upregulated, both members of the LDL receptors family. Upregulation of LRP3 was con rmed by qRT-PCR, with a signi cant increase in mRNA LRP3 level respect to its expression in non-transfected cells. However, increments in LDLR mRNA expression resulted not signi cant by qRT-PCR (Fig 1a).
Although SH-SY5Y cells secrete reelin to the media and it can act in a paracrine mode, recombinant reelin was employed to treat overexpressing-ApoER2 cells to potentiate the ApoER2 signaling. This treatment induced ApoER2 cleavage and consequently, reduced the amount of full-length ApoER2 and increased the generation of the ApoER2-CTF (Fig 1b). In retinoic acid neuro-differentiated SH-SY5Y cells, reelin treatment was also able to induce an increase in LRP3 protein levels respect to non-stimulated cells (Fig  1c).

Expression of ApoER2-ICD upregulates LRP3 expression
We considered the possibility that increments of LRP3 expression were induced by ApoER2-ICD, a fragment with transcriptional regulatory activity [14], generated by the proteolytic cleavage of ApoER2-CTF. This small fragment was observed in ApoER2-overexpressing cells after treatment with reelin ( Fig  2a). Thus, we overexpressed a chimeric ApoER2-ICD (amino acid residues 728-842) and measured LRP3 expression. LRP3 mRNA expression and protein levels increased signi cantly respect to non-transfected cells (Fig 2b-d), while LDLR mRNA levels were not signi cantly affected by ApoER2-ICD (Fig 2e).
Expression levels of LRP3 in Aβ42-treated cells.
Opposite to the upregulation of LRP3 mRNA and protein that we observed after overexpression of fulllength ApoER2 or ApoER2-ICD, we expected to nd less LRP3 expression in Aβ42-treated cells, due to Aβ treatment reduces the generation of ApoER2-CTF [16]. In agreement with this view, we found that Aβ42 treatment in neuro-differentiated SH-SY5Y cells decreased the LRP3 mRNA expression (Fig 3a) and LRP3 protein levels (Fig 3b), respect to scrambled peptide-treatment.
Expression levels of LRP3 in AD brain Next, we examined LRP3 levels in AD human frontal cortex extracts to explore whether this pathology affects its levels. In AD, LRP3 mRNA expression was lower respect to that in extracts from non-demented (ND) subjects (Fig 4a). When AD subjects were categorized by Braak stages to discriminate among neurodegenerative states, the reduction was signi cant only at Braak stages I-II; resulting close to be statistically signi cant at Braak III-IV and Braak V-VI (Fig 4b). To evaluate LRP3 protein levels in cortex from AD and ND subjects, membrane-enriched fractions were isolated from brain samples and two antibodies were tested (Fig 4c). We found that LRP3 levels were lower in AD at Braak stages I-II in AD, compared to those in ND subjects (Fig 4d).
The estimated expression of LDLR mRNA was not signi cantly different between AD subjects and ND controls when the extracts were compared in overall, but neither when the comparison was performed among Braak stages nor by apoE genotype (Fig 4e).

LRP3 interacts with apoE and APP, but not with reelin
We evaluated by immunoprecipitation assays, whether reelin is as a ligand of LRP3, as it is of ApoER2, in frontal cortex extracts from control subjects and from AD subjects. However, reelin was not coimmunoprecipitated from any brain extracts. We next evaluated whether LRP3 interacts with apoE and APP, as it has been described in many members of the LDL receptors family. After immunoprecipitation, both proteins were co-immunoprecipitated with LRP3 in ND and AD cortex extracts (Fig 5).

LRP3 modulates APP expression levels
To test whether LRP3 is able to in uence APP processing and Aβ generation, we overexpressed LRP3 in CHO-PS70 cells, a cell line that express the wild-type APP770 isoform. LRP3 overexpression was located in the plasma membrane of CHO-PS70 cells, but it was also frequently observed in discrete areas in the soma (Fig 6a). LRP3 and APP co-immunoprecipitated in these cells (Fig 6b).
Interestingly, LRP3 overexpression reduced drastically full-length APP levels, as well as APP-CTF in cell extracts. In the supernatant, the levels of sAPPα, sAPPβ and soluble Aβ decreased in transfected CHO-PS70 cells respect to mock (Fig 6c). The decrease in APP levels was not related to a change in APP mRNA expression (Fig 6d).

Discussion
Our results suggest that reelin signaling, through the cleavage of its receptor ApoER2, can ultimately in uence expression of other liporeceptors, such as LRP3. Many LDL receptor family members, as ApoER2, LDLR, LRP1, LRP1b, LRP6 and SorLA (LRP11), as well as other alternative apoE receptors such as Trem2, are γ-secretase substrates [21,22]. For many of these receptor the nuclear translocation of the respective ICDs and their transcriptional functions have been demonstrated or inferred [13,[23][24][25]. Here, we demonstrate that reelin-induced generation of ApoER2-ICD, as well as ApoER2-ICD overexpression, increase LRP3 expression. This supports a link between ApoER2 processing and the regulation of LRP3 protein levels.
In frontal cortex extracts from AD, where ApoER2/reelin signaling is impaired and ApoER2 processing is lessened (reviewed in [26]), we have found lower LRP3 protein and mRNA levels. LRP3 expression was mainly affected at early Braak stages of AD (stages I-II), in which the transentorhinal region show neuro brillary tangles and neuropil threads [18]. However, since the same decreasing trend was determined in later Braak stages, additional studies are needed to determine whether LRP3 decrease is only an early phenomenon associated to AD progression.
In the microarray, after overexpression of full-length ApoER2, the expression of another LDL receptors family member, LDLR, was also upregulated. Interestingly, both LRP3 and LDLR are encoded by genes located on chromosome 19, locus 19q13 [27,28]. The APOE gene also maps in chromosome 19, on locus 19q13.32 [29], in a cluster together with apolipoprotein C1 and C2 genes. Genetic linkage studies suggest the presence of AD risk genes on chromosome 19, that would act in an independent manner from apoE, such as ABCA7 (19p13.3) and CD33 (19q13.41) [30]. Indeed, LDLR was analyzed as a potential AD risk factor, but the study concluded that the genetic variants in LDLR did not make a signi cant contribution to AD risk in the general population [31]. Interestingly, recent multiplex proteomics studies have identi ed that LDLR levels are modestly decreased in CSF from early AD patients, suggesting that this receptor can represent a new speci c biomarkers of AD [32]. Other genes encoding LDL receptors family members, such as LRP1, LRP1b, LRP2, LRP4, LRP6, and SorLA have been associated to AD risk (reviewed in [13]), even ApoER2 [33]. Despite the results from the microarray study, the qRT-PCR failed to corroborate the modulation of LDLR by ApoER2 and did not nd changes on LDLR expression in AD extracts.
The reelin receptors ApoER2 and VDLR are core members of the LDL family that share the same extracellular domains structure, the ligand binding-type repeat domains (LBDs) and the EGF-precursor homology domains. The intracellular domain of each of the core members contain at least one NPxY (Asn-Pro-X-Tyr) motif, that functions in protein interaction/signal transduction [34][35][36] and endocytosis [37]. In comparison, LRP3 is smaller than the core members of the LDL receptors family. LRP3 belongs to a subfamily, together with LRP10 (murine LRP9), LRP12 and Lrad3 (ST7/Mig13). These subfamily members are characterized by the sole presence of LBDs and CUB-domains (which binds Complement, Uegf and Bmp1) in their extracellular domain, and lacks the EGF-like repeats [13]. The short LBD in LRP3 is likely the domain responsible of the co-immunoprecipitation of apoE, as this is the competent region to bind several ligands [38]. However, reelin was not co-immunoprecipitated, in the same manner as RAP (receptor-associated protein), another ApoER2 ligand, does not bind LRP3 neither [17,27,39,40]. In the intracellular domain, LRP3 lacks the NPxY motifs, but instead, contains a similar tyrosine-based sequence (EDFPVY) [27,41]. It is yet to be determined the domain by which APP is able to interact with LRP3. In vitro data showed that the extracellular domain of LRP10 interacts with APP [42], while Lrad3, the LDL receptor family member with the shortest extracellular domain [43] is also able to interact with APP and to modulate APP processing pathways. ApoER2 and APP are linked extracellularly by binding different domains of F-spondin [44] and intracellularly through the adaptor proteins Dab-1 and Fe65, that interact with the NPxY-motif of ApoER2 and APP [45][46][47]. Therefore, more studies are necessary to explore the interaction, either direct or indirect, between LRP3 and APP.
Core members of the LDL receptor family have been related to APP tra cking and internalization and, therefore, they could determine the APP proteolytic processing and Aβ production, likely playing a role in AD pathogenesis [48][49][50][51]. For example, LRP1 increases APP endocytosis and generation of Aβ [52][53][54], while LRP1B retains APP at the cell surface [55]. ApoER2 is able to alter APP subcellular distribution, increasing the generation of Aβ; this effect depends on the integrity of the NPxY motif in ApoER2 [56]. Whereas in a mouse ApoER2 isoform that lacks three LBDs, predominates the non-amyloidogenic processing of APP [57]. Here, we observed that overexpression of LRP3 decreased the levels of full-length APP and APP fragments generated after amyloidogenic and non-amyloidogenic processing pathways, opening questions about the possible mechanism involved. A direct down-regulation of APP mRNA would be likely discharged as results of our qRT-PCR data. However, as with other LDL receptors [55], endocytic mechanisms could be involved in the alteration of APP processing, as LRP3 has been described as an endocytosis receptor [27]. In this line, LRP1 endocytosis impairment favours non-amyloidogenic processing of APP due to a reduced internalization, resulting in less extracellular Aβ [59,60]. Additionally, mechanisms related to the APP secretory pathway are also possible, as for LRP1 again, which retention in the endoplasmic reticulum by the expression of a speci c motif, leads a decrease in full length APP and CTF levels at the plasma membrane as well as in Aβ secretion [1,61].
LRP3 highest expression is in skeletal muscle and ovary, but it is also present at relatively high levels in brain and heart, among other tissues [17]. LRP3 has been involved so far in osteogenic and adipocytic differentiation [62] and systemic use of steroids has been associated with site-speci c differential methylation of the LRP3 gene [63], but its role in neuronal activity is still unknown. LRP3 has been identi ed as a gen up-regulated for a short window of two hours, exclusively following learning, in the rat dentate gyrus [64]. To clarify LRP3s biological functions it is essential to de ne the signi cance of LRP3 reduction in AD frontal cortex that we have observed. An alteration in the expression of LRP3 may in uence the processing and expression of APP, affecting its synaptic function, and therefore, contributing to this AD pathology.

Limitations
The main limitation of this study is the scarce knowledge of the physiological function of LRP3 in the brain, as there are few reports about this as a neuronal receptor. We employed a well characterized brain collection, but it would be interesting to validate our nding with an alternative collection of post-mortem cortex samples from non-demented and AD subjects. Development of in vivo knockouts or knockdowns of LRP3 would contribute to the understanding of the mechanism that links this receptor and APP, given that, for example, knockdown of Lrp10 led to increased processing of APP to generate Aβ [65].

Conclusions
ApoER2/reelin signalling is able to regulate LRP3 expression, and LRP3 reduces APP protein levels, including sAPP fragments and Aβ peptide. The mechanism involved is yet to be determined, but this study could contribute to nd new strategies in AD research, considering that LRP3 modulation could participate in the rising Aβ levels. Declarations Funding (information that explains whether and by whom the research was supported) This work was supported by grants from the Fondo de Investigaciones Sanitarias (PI15/00665 and PI19-01359, cofunded by the Fondo Europeo de Desarrollo Regional, FEDER "Investing in your future"), and through CIBERNED (Instituto de Salud Carlos III, Spain). We also acknowledge nancial support from the Spanish Ministerio de Economía y Competitividad, through the "Severo Ochoa" Programme for Centres of Excellence in R&D (SEV-2017-0723).

Con icts of interest/Competing interests: Not applicable
Availability of data and material (data transparency) All data and materials support their published claims and comply with eld standards.  Table 1. Genes associated with lipid binding and transport, and cholesterol metabolism, whose transcripts were upregulated in ApoER2 overexpressing SH-SY5Y cells compared with control cells transfected with an empty vector. The expression of the genes was analyzed on DNA microarrays. The fold change (logFC) in gene expression between samples and controls, as well the adj p (p-value adjusted for multiple testing) is indicated.