PICALM and Alzheimer’s Disease: An Update and Perspectives

Genome-wide association studies (GWAS) have identified the PICALM (Phosphatidylinositol binding clathrin-assembly protein) gene as the most significant genetic susceptibility locus after APOE and BIN1. PICALM is a clathrin-adaptor protein that plays a critical role in clathrin-mediated endocytosis and autophagy. Since the effects of genetic variants of PICALM as AD-susceptibility loci have been confirmed by independent genetic studies in several distinct cohorts, there has been a number of in vitro and in vivo studies attempting to elucidate the underlying mechanism by which PICALM modulates AD risk. While differential modulation of APP processing and Aβ transcytosis by PICALM has been reported, significant effects of PICALM modulation of tau pathology progression have also been evidenced in Alzheimer’s disease models. In this review, we summarize the current knowledge about PICALM, its physiological functions, genetic variants, post-translational modifications and relevance to AD pathogenesis.


Physiopathology of Alzheimer's Disease and Genetic Risk Factors
Alzheimer's disease (AD) is the most common cause of dementia, manifesting itself in cognitive deficits. AD is characterized by two neuropathological hallmarks: extracellular deposition of fibrils made up of amyloid β (Aβ) peptides as senile plaques [1] and intracellular accumulation of hyperphosphorylated tau as neurofibrillary tangles (NFTs) [2] (for review, see [3]).
Aβ is produced by successive cleavages of β-amyloid precursor protein (APP), encoded by APP gene located in chromosome 21, by β-secretase (BACE1) and γ-secretase [4]. APP is cleaved by BACE1 at its N terminal domain to release soluble APP-β and the membrane-bound APP C-terminal fragment (β-CTF or C99). The γ-secretase is a complex of four protein subunits consisting of presenilin (PSEN), Nicastrin, presenilin enhancer (PEN) and APH1 (anterior pharynx-defective 1). Sequential cleavages of APP by BACE1 and γ-secretase lead to production of Aβ peptides of 38-, 40-and 42-amino acids in length. Aβ peptides are principally produced in endosomes and are released from neurons in a synaptic activity-dependent manner [5]. Aβ peptides adopt aggregation-prone conformations that are resistant to proteolysis and form oligomers, protofibrils and fibrils. Due to hydrophobicity at its C-terminus, Aβ42 has greater aggregation properties [6].

PICALM and Clathrin-Mediated Endocytosis
Clathrin-mediated endocytosis, also called receptor-mediated endocytosis, is a process leading to the internalization of plasma membrane-associated cargo molecules and subsequent trafficking through intracellular vesicle compartments. PICALM is necessary Cells 2022, 11, 3994 4 of 26 for clathrin assembly at the plasma membrane by binding to AP2 and clathrin ( Figure 2A). PICALM interacts simultaneously with AP1 and clathrin at the trans-Golgi network in HeLa cells [43]. Several independent studies have reported that the depletion of PICALM leads to an increase in the size of clathrin-coated vesicles (CCVs) and to an increase in heterogeneity in the CCVs shape in cultured cells of rat hippocampus [46] and of HeLa cells [43,47]. CRISPR/Cas9-mediated PICALM disruption also leads to an increase in the number of early endosomes in HeLa cells [48]. A similar observation was obtained by the depletion of uncoordinated protein-11 (unc-11), the ortholog of PICALM in neurons of C. elegans [49]. Another study suggests that PICALM is required for efficient clathrin coat maturation in fibroblasts [50].

PICALM and Clathrin-Mediated Endocytosis
Clathrin-mediated endocytosis, also called receptor-mediated endocytosis, is a process leading to the internalization of plasma membrane-associated cargo molecules and subsequent trafficking through intracellular vesicle compartments. PICALM is necessary for clathrin assembly at the plasma membrane by binding to AP2 and clathrin ( Figure 2A). PICALM interacts simultaneously with AP1 and clathrin at the trans-Golgi network in HeLa cells [43]. Several independent studies have reported that the depletion of PICALM leads to an increase in the size of clathrin-coated vesicles (CCVs) and to an increase in heterogeneity in the CCVs shape in cultured cells of rat hippocampus [46] and of HeLa cells [43,47]. CRISPR/Cas9-mediated PICALM disruption also leads to an increase in the number of early endosomes in HeLa cells [48]. A similar observation was obtained by the depletion of uncoordinated protein-11 (unc-11), the ortholog of PICALM in neurons of C. elegans [49]. Another study suggests that PICALM is required for efficient clathrin coat maturation in fibroblasts [50]. PICALM is involved in clathrin-mediated endocytosis (CME) as a clathrin adaptor by facilitating the formation of clathrin-coated vesicles (CCVs). The interaction between PICALM, adaptor proteins (AP) and clathrin is critical to maintaining the form and sizes of CCVs. PICALM interacts with AP2 at the plasma membrane and AP 1 in the trans-Golgi network to form CCVs. (B) PICALM is also involved in multiple steps of autophagy process. PICALM is involved in the autophagy precursor formation at the autophagosome-lysosome fusion [51], maturation of Cathepsin D and lysosomal function [48].

PICALM, Hematopoiesis, Transferrin-Uptake and Cholesterol Homeostasis
Chromosomal translocations of PICALM have been observed in patients with acute lymphoblastic leukemia and acute myeloid leukemia [52]. In mice, nonsense point mutations of PICALM cause abnormal iron distribution, growth retardation, shortened life span and hematopoietic abnormalities [53].
PICALM is a critical protein, and homozygous PICALM knockout mice are dwarfed and die within 1 month after birth due to a deficiency in erythroid maturation and transferrin uptake [54]. There are controversies in the literature on the effects of PICALM on Figure 2. PICALM regulates clathrin-mediated endocytosis and autophagy. (A) PICALM is involved in clathrin-mediated endocytosis (CME) as a clathrin adaptor by facilitating the formation of clathrincoated vesicles (CCVs). The interaction between PICALM, adaptor proteins (AP) and clathrin is critical to maintaining the form and sizes of CCVs. PICALM interacts with AP2 at the plasma membrane and AP 1 in the trans-Golgi network to form CCVs. (B) PICALM is also involved in multiple steps of autophagy process. PICALM is involved in the autophagy precursor formation at the autophagosome-lysosome fusion [51], maturation of Cathepsin D and lysosomal function [48].

PICALM, Hematopoiesis, Transferrin-Uptake and Cholesterol Homeostasis
Chromosomal translocations of PICALM have been observed in patients with acute lymphoblastic leukemia and acute myeloid leukemia [52]. In mice, nonsense point mutations of PICALM cause abnormal iron distribution, growth retardation, shortened life span and hematopoietic abnormalities [53].
PICALM is a critical protein, and homozygous PICALM knockout mice are dwarfed and die within 1 month after birth due to a deficiency in erythroid maturation and transferrin uptake [54]. There are controversies in the literature on the effects of PICALM on transferrin uptake. While some studies suggest that PICALM overexpression or downexpression did not alter transferrin uptake in HeLa cells or HEK293 cells [55,56], others observed a significant effect on transferrin uptake in the liver, bone marrow and erythroid cells from PICALM-deficient mice [50,54]. It is thus speculated that the deletion of PICALM may be partially compensated by other endocytic proteins in certain types of cells. Indeed, PICALM is dispensable in myeloid and B-lymphoid development [50].
In the C-terminal clathrin adaptor domain of PICALM contains critical motifs that allow PICALM to interact with clathrin and other endocytic machinery proteins. The clathrin binding motif is DLLDLQ (392-397 in the isoform of 652 amino acids). AP2binding motifs are DIF (D375, I376 and F377), DPF (D420, P421 and F422) and FESVF (F489, E490, S491, V492 and F493). The asparagine-proline-phenylalanine (NPF) motif binds to Eps15 homology (EH) domains found in proteins associated with endocytosis and vesicle trafficking. Since exon 13 contains one DPF (AP2 binding) and one NPF (EH domain binding) motif, the splice variant without exon 13 is supposed to have reduced interaction with AP2 and other protein partners owning EH-domains [43].

Post-Translational Modifications of PICALM
PICALM undergoes several post-translational modifications (PTM), such as phosphorylation, methylation, ubiquitination, sumoylation and acetylation [65,66,[68][69][70][71][72][73][74]. These post-translational modifications have been identified on numerous amino acid residues of PICALM: 41 phosphorylation, 15 ubiquitination, 1 acetylation, 1 monomethylation, 1 di-methylation and 1 sumoylation sites have been described so far. Certain factors are known to induce some of the post-translational modifications on PICALM, such as phosphorylation by Aβ treatment in primary neuronal culture [75], by BI2536, a selective inhibitor of protein kinase Polo-Like Kinase 1 (PLK1), inducing phosphorylation (Y138, T146 and S537) [76], by ischemia (S303 and S315) [77] and by nocodazole, an inhibitor of microtubule polymerization (S474 and S645) [78]. The most well-characterized PTM of PICALM are the four phosphorylation sites (S16, Y96, S107 and S303) and two ubiquitination sites (K318 and K324) that have been validated by mass-spectrometry and have been described for more than 5 times [65]. While the phosphorylation on the ANTH domain may influence the interactions of PICALM with PtdIns(4,5)P 2 or R-SNARE proteins, the other phosphorylation residues are located in the C-terminal clathrin adaptor domain exhibiting pivotal functions in protein-protein interaction of PICALM with clathrin and other endocytic machinery proteins. Therefore, it is speculated that phosphorylation of PICALM should play a role in regulating interaction with PtdIns(4,5)P 2 and other protein binding partners. For instance, the R-SNARE binding domain contains several sites that can undergo PTM, such as one phosphorylation (Y237), two ubiquitination (K231 and K252) and one sumoylation (K238), and thus, these PTM may regulate the interaction between PICALM and R-SNARE.
PICALM is a substrate of calpain and caspase [33,79,80]. Nevertheless, it remains largely unknown if the phosphorylation and ubiquitination of PICALM affect its protein stability or degradation.
Given that PICALM contains more than 130 putative phosphorylation sites, PI-CALM phosphorylation may be dysregulated in AD brains not only in terms of its global phosphorylation level but also in the pattern of site-specific phosphorylation. Indeed, Aβ exposure increases PICALM phosphorylation in primary rat neuronal cortical culture [75] though phosphorylation sites and the involved kinases remain to be identified. While it is speculated that phosphorylation may change the interaction of PICALM with its binding partners such as R-SNAREs, PtdIns(4,5)P 2 or clathrin, it remains elusive which kinases are involved in PICALM phosphorylation, whether PICALM phosphorylation is misregulated during the progression of AD, whether PI-CALM phosphorylation has any impact on its protein stability/degradation or its subcellular localization and whether phosphorylation of PICALM may cause a loss of function and/or a gain of toxicity.

Gene Location and Genetic Variants of PICALM
In humans, the PICALM gene is located on chromosome 11q14 ( Figure 4). Since several independent GWAS have reported the involvement of PICALM SNPs in LOAD susceptibility, many other studies have replicated and confirmed the association of PICALM genetic variants with LOAD in diverse cohorts (see review [44]). There have been several LOAD-associated SNPs identified within or near the PICALM gene so far (Table S1). A recent large GWAS (n = 788,989) suggests that the most significant SNP associated with LOAD susceptibility is rs3851179, located in the intergenic regions between PICALM and EED [17].
susceptibility, many other studies have replicated and confirmed the association of CALM genetic variants with LOAD in diverse cohorts (see review [44]). There have be several LOAD-associated SNPs identified within or near the PICALM gene so far (Tab S1). A recent large GWAS (n = 788,989) suggests that the most significant SNP associat with LOAD susceptibility is rs3851179, located in the intergenic regions between PICAL and EED [17]. The most significant LOAD-associated SNP is rs3851179, located proximately 88 kb upstream of PICALM in the 5′ intergenic region [17]. rs10792832 is 765 bp fr rs3851179 and in full linkage disequilibrium (LD) with rs3851179 (r 2 = 1). rs1237999, located 34 upstream of PICALM, is localized in a CCCTC binding factor in a regulatory region [81]. rs5616 is the lead SNP of PICALM locus in [82], located 20 kb upstream of PICALM gene, a putative tr scription factor binding site [15] and in LD with rs3851179 (r 2 = 0.76). rs541458, located in 8 kb u stream of PICALM gene, is also genome-wide significant [15] and is in LD with rs3851179 (r 2 = 0.6 rs592297 is located in exon 5 and is an exonic splice enhancer [29]. rs588076 is located in the intr 17 of PICALM and is associated with the allelic expression imbalance (AEI) in PICALM isofo lacking exons 18 and 19 (D18-19) [83]. rs3851179 T is a protective allele reported to reduce AD risk by 9-29% among vario cohorts [81]. rs3851179 is located in a region 88 kb upstream of the PICALM gene (Figu 4) and is likely involved in transcription factor binding [84].

PICALM and Quantitative Trait Loci (QTL)
Analyses on expression quantitative trait loci (eQTL) suggest that there may be a cell-type specific effect of the PICALM variant, notably in microglia. Several recent PI-CALM eQTL analyses have consistently suggested that there is a significant colocalization between AD GWAS signals and PICALM eQTL in microglia [90][91][92][93]. PICALM rs10792832 (in almost full LD with rs3851179) is located in an open chromatin region (OCR), and the non-protective variant was associated with both lower OCR signal and gene expression by regulating chromatin accessibility [91].
Microglia may not be the only cell type affected by PICALM variants. The protective rs3851179 T allele is linked to increased PICALM expression in the brain [81,94], especially implicated in PICALM expression in endothelial cells [30,34]. The rs3851179 T allele is associated with increased PICALM mRNA and protein expression in induced pluripotent stem cells (iPSC) differentiated into endothelium compared to isogenic cells bearing non-protective rs3851179 C [34]. The non-protective rs3851179 C allele is associated with a significant reduction in PICALM isoform 2 mRNA expression in the frontal and in temporal cortex [94].
Alternative splicing of PICALM is significantly changed in AD brains in relation to LOAD-susceptibility variants [41]. Differential excision of introns in the PICALM gene was observed in the dorsolateral prefrontal cortex (DLPFC) of AD brains in correlation with the GWAS signal [41]. Raj and colleagues observed such alteration of alternative splicing in iPSC-derived neurons overexpressing tau, suggesting that accumulation of tau drives abnormal alternative splicing of key AD genes, including PICALM.
It has to be noted that the nearest coding gene of rs3851179 is EED [17]. A colocalization of the EED function association has been reported [95]. It remains elusive if EED is also involved in AD pathogenesis.
These data support the hypothesis that PICALM is the risk gene in the locus and that the potential effect for AD risk is, at least partially, via modulation of microglial PICALM expression.

PICALM Variants and Phenotypes Relevant to AD
Studies have shown significant associations of some of the SNPs in the PICALM locus with phenotypes relevant to AD, such as the age of onset, cognitive functions, information processing speed, spontaneous brain activity, hippocampal atrophy and Aβ or tau levels in cerebrospinal fluid (CSF) ( Table 1). Some SNPs are not genome-wide significant to LOAD (rs510566, rs592297, rs17148741) but showed statistically significant association with hippocampal atrophy or Aβ and tau levels in cerebrospinal fluid (CSF) [81,96].   PICALM rs541458 T risk allele is associated with decreased Aβ in CSF [98]. A potential association of rs541458 was reported with the spontaneous brain activity of amnestic MCI [99] and with the cognitive performance of non-demented Chinese in an agedependent manner [109].
Xu et al., reported that rs1237999, located 34 kb upstream of PICALM, is localized in a CCCTC binding factor in a regulatory region (Figure 4). rs1237999 in LD with rs3851179 and is associated with the levels of Aβ, tau and the ratio of phosphotau/Aβ in CSF [81].
There is a significant association of PICALM rs3851179 T protective allele with the change rate of adjusted hippocampal volume [103,104] and entorhinal cortex thickness [103,108] measured by magnetic resonance imaging (MRI). rs3851179 T is associated with better cognitive functions in the oldest old [106], metabolic connectivity [110], and default mode network function in mild cognitive impairment (MCI) [111]. Non-protective rs3851179 C allele is associated with the decline in the information processing speed [112] and earlier onset of AD [105].
It has to be noted that some studies have a rather small sample size (Table 1) and may be underpowered. Further replication of these studies in larger cohorts is warranted. Although correlation does not imply causation, these reports support the hypothesis that PICALM variants are associated with phenotypes relevant to AD.

PICALM and APP Processing
Given that the generation of Aβ peptides from APP and its accumulation are pivotal events in AD pathogenesis, researchers have sought to identify the role of PICALM on Aβ pathology. Multiple lines of evidence suggest that PICALM regulates APP processing [35,[113][114][115] and Aβ transcytosis [34].
Tian et al., have reported a rather protective role of PICALM in APP metabolism. PICALM forms a complex with APP-CTF together with AP2 and microtubule-associated protein 1 light chain 3 (LC3), an autophagy marker [115]. The interaction of APP-CTFs, PICALM, AP2 and LC3 results in transporting APP-CTFs into the autophagosome in which APP-CTFs are degraded by autophagy. The interaction between APP-CTFs and PICALM consequently leads to a reduction of Aβ production [114]. This critical role of PICALM in autophagy is strengthened by another study showing that loss of exon 1 of PICALM in HeLa cells leads to an augmentation of lysosomal enzymes in the endosome-enriched fraction and accumulation of autophagy substrates such as APP-CTFs [48].
PICALM regulates APP processing via γ-secretase internalization due to a direct interaction between the PICALM ANTH domain and Nicastrin, one of the components of the γ-secretase complex [116]. Intriguingly, PICALM itself modulates APP-CTF production via controlling BACE1 expression [94].
In addition, PICALM has been reported to be a regulator of PICALM/clathrindependent internalization of Aβ bound to the low-density lipoprotein-receptor-related protein-1 (LRP1), a major Aβ receptor, leading to Aβ transcytosis and clearance [34].
As there are multiple implications of PICALM in the regulation of Aβ pathology (autophagy, APP processing, transcytosis), it is not surprising that there are some contradictory observations in transgenic models of amyloid pathology with reduced expression of PICALM. One study has shown a significant increase in amyloid load in PICALM +/− crossed with APP sw/0 and observed a 3.5-4-fold increase in Aβ load in the 9-month-old APP sw/0 PICALM +/− mouse brain [34]. On the other hand, Kanatsu et al., observed a significant reduction of Aβ load in 12-month-old A7/PICALM +/− mouse brains. Further, Xiao et al., have reported that PICALM is colocalized with APP and PICALM expression level has a significant impact on Aβ level in APP/PS1 mice injected with adeno-associated virus (AAV)-mediated modulation of PICALM levels. While PICALM overexpression increases APP internalization and Aβ production, PICALM knockdown decreases amyloid load [35]. Furthermore, drosophila PICALM orthologue LAP overexpression did not alter the Aβ42 level in a drosophila model overexpressing Aβ42 [117]. The discrepancy of the observation on Aβ production may be related to the different transgenes with distinct mutations and the differences in amyloid doses in the models analyzed. These contradictory data imply that the effect of PICALM reduction can differ depending on the progression of Aβ pathology. Further studies are necessary to determine the function of PICALM in APP processing by using a model close to the physiological conditions of human LOAD to achieve clinical relevance.

PICALM and Aβ-Mediated Neurotoxicity via Glutaminergic Neurotransmission
In the yeast model, PICALM orthologue was identified as a protective modifier of Aβ toxicity [118]. Overexpression of YAP1802, the yeast orthologue of PICALM, partially rescues Aβ-induced toxicity without affecting the secretory pathway. Treusch et al., confirmed the rescuing effect of PICALM in C. elegans and rat cortical neuronal culture: overexpression of unc-11, the C. elegans orthologue of PICALM, rescues Aβ-induced glutaminergic neuron loss in C. elegans and lentiviral PICALM overexpression diminished Aβ-induced neuron loss in primary culture of rat cortical neurons. On the contrary, D'Angelo and colleagues observed that mammalian PICALM enhanced Aβ-induced toxicity in a yeast model overexpressing Aβ tagged with GFP [119]. These inconsistent observations are partly due to the heterogeneity of the Aβ species and the distinct expression level of Aβ. In addition, Aβ overexpression in these bio-engineered yeast models was driven by transgene encoding Aβ, and thus, the nature of overexpressed Aβ should differ from the Aβ peptides derived from successive cleavages of APP in a physiological/pathophysiological condition in human brains.
It is noteworthy that Aβ is not co-immunoprecipitated with PICALM in human brains, and PICALM immunoreactivity is never observed in the center of amyloid plaques [32][33][34][35]. Thus, PICALM may presumably play roles in Aβ-induced toxicity in an indirect manner. One of the candidate intermediates between PICALM and Aβ is N-methyl-d-aspartate receptor (NMDAR). Glutaminergic neurons are the most vulnerable subtype of neurons in AD, and soluble Aβ enhances glutamate release, leading to excessive activation [120,121]. Whilst NMDAR is critical for synaptic plasticity and neuronal survival, excessive NMDAR activation leads to excitotoxicity and cell death [122]. A recent study in the drosophila model of AD has suggested that overexpression of LAP, the drosophila PICALM orthologue, rescues Aβ42-induced toxicity without affecting Aβ42 level [117].

PICALM Expression and Localization in AD Brains
Different lines of evidence suggest that PICALM expression is dysregulated during the progression of AD. Under a physiological condition in the central nervous system (CNS), PICALM is ubiquitously expressed in various cell types such as endothelial, neuronal and microglial cells [30,[32][33][34][35][36]123]. In AD postmortem brains, however, PICALM may undergo an alteration of expression levels and protein localizations in distinct cell types.
During the progression of AD, PICALM mRNA expression levels are increased in the brain [32,124] and in the blood [125] of AD patients. In contrast, the PICALM protein level in the total bulk tissue is decreased in AD brains [33]. Another study has shown that PICALM protein level is decreased in the isolated microvessel fraction of AD brains but is not significantly altered in the microvessel-depleted fraction of AD brains [34]. In line with this discordant observation, Tasaki et al., have recently reported that there is a general divergence between the levels of transcripts (mRNA) and protein expression in postmortem AD brain tissue. By mass-spectrometry, they observed a significant reduction of PICALM protein in a correlation with 14.3% faster cognitive decline [124].
Most of the analyses made on PICALM expression were conducted using human postmortem tissues of the affected brain regions that contain different cell types. Singlecell transcriptomic studies have shown that groups of GWAS hit genes, including PI-CALM, are generally upregulated in microglial cells in correlation with neuronal AD pathology [126]. While PICALM immunoreactivity in microglial cells is increased in LOAD brains [33], PICALM immunolabelling is decreased in endothelial cells [30], and PICALM protein level assessed by Western blot is decreased in the microvessel fraction of postmortem AD brain tissues [34]. Neuronal PICALM labeling is increased in AD brains [33,34] and is associated with NFTs ( Figure 6A-C) [33]. PICALM and hyperphosphorylated tau are co-immunoprecipitated in AD brain lysate [33]. Since pathological tau in NFTs is ubiquitinated [127], and the ANTH domain of PICALM has an affinity to ubiquitin [67], the interaction between PICALM and tau needs to be further analyzed in non-pathological and pathological conditions to determine the role of ubiquitin to form tau-PICALM complex. Other studies have shown immunoreactivities of PICALM interactors such as PtdIns(4,5)P 2 [123,128] and AP2A1 [129] in association with phosphotau in NFTs and dystrophic neurites. Of note, PICALM is not the only LOAD GWAS hit protein detected in NFTs: some other proteins encoded by GWAS hits genes are also linked with tau pathologies such as APOE4 [130], CLU [131], BIN1 [132], TPK2B [133] and AP2A1 [129].

PICALM in Primary Tauopathy Brains
To our knowledge, there is no report showing alteration of PICALM mRNA expression levels in primary tauopathy brains; however, PICALM protein localization is altered in primary tauopathies with no amyloid pathology [134]. In postmortem human tissues, PICALM immunoreactivity is associated with pathological tau in the AD brain [33] and in other tauopathies such as PSP and Pick disease [134]. In NFTs, PtdIns(4,5)P 2 and PICALM are colocalized [123,128]. PICALM protein level is decreased in the affected brain areas of various tau-related neurodegenerative diseases, including AD, FTLD-tau-MAPT, corticobasal degeneration (CBD) and Lewy body dementia (LBD) diffuse type with tau pathology [33,134,135]. A significant negative correlation was observed between PICALM protein level and hyperphosphorylated tau or autophagy initiation marker Beclin1 in the postmortem human frontal cortex of these neurodegenerative diseases [134]. Such reduction of PICALM seems related to the fact that PICALM is a substrate of both calpain and caspase-3 [33,79,80] that are activated in the brains of AD and other tauopathies [136][137][138][139][140].

PICALM and Tau in Cellular and Animal Models
PICALM modulates tau degradation via both autophagosome formation and autolysosomal fusion in several experimental models, such as cultured cells, drosophila and zebrafish [51]. Knockdown of PICALM caused accelerated accumulation of phosphotau in HeLa cells transiently transfected with DsRed-Tau 4R (four repeats) by impairing autophagy. Nonetheless, they observed in the same study that PICALM overexpression drove tau-mediated neurodegeneration in zebrafish [51]. Thus, they observed that both knockdown and overexpression of PICALM accelerated neurodegeneration in the presence of tau pathology. interaction between PICALM and tau needs to be further analyzed in non-pathological and pathological conditions to determine the role of ubiquitin to form tau-PICALM complex. Other studies have shown immunoreactivities of PICALM interactors such as PtdIns(4,5)P2 [123,128] and AP2A1 [129] in association with phosphotau in NFTs and dystrophic neurites. Of note, PICALM is not the only LOAD GWAS hit protein detected in NFTs: some other proteins encoded by GWAS hits genes are also linked with tau pathologies such as APOE4 [130], CLU [131], BIN1 [132], TPK2B [133] and AP2A1 [129].  FAD with PSEN1 mutations at R35E and E120D (C,F). In non-demented control brain, a perinuclear immunolabelling of PICALM was detected in neurons (black arrows) (A). A PICALM immunoreactivity was detected in NFTs in neuronal perikarya (blue arrows) in the affected brain areas of LOAD (B) and FAD (C). PICALM-immunoreactive microglial cells were faintly detected in control brains (arrowheads) (D), and the number and the intensity of PICALM-stained microglia were increased in LOAD (D) (arrowheads). In FAD (F) brains, the increase in PICALM immunostained microglial cells was less prominent than in LOAD. More detailed data can be found in our article [33]. Control (Ctrl), Late-onset Alzheimer's disease (LOAD), Familial Alzheimer's disease (FAD). Scale bar 20 µm.
Studies in PICALM +/− mice suggest that 50% of PICALM reduction did not trigger a clear motor phenotype or any detectable tau pathology for at least up to 12 months [54,135]. On the contrary, PICALM heteroinsufficiency aggravates tau pathology in a tau transgenic mouse model. By crossing Tg30 overexpressing human double mutant tau (P301S and G272V) transgenic mice developing an age-dependent tau pathology [141] with PICALM +/− mice [54], we generated a novel transgenic mouse line Tg30xPICALM +/− [135]. Tg30xPICALM +/− exhibited an aggravation of motor deficits and tau pathology, suggesting that a reduction of PICALM in the presence of tau pathology may induce deficits in the autophagy pathway [135]. It should be emphasized that this model develops a tau pathology at least partially in a cell-autonomous manner due to a tau transgene bearing double FTLD-related mutations [141]. Since MAPT gene mutations are not linked to AD, further studies are required to decipher the role of PICALM in the development and propagation of non-mutant tau.

PICALM and Tau Pathology Propagation
In AD brains, tau pathology develops in a stereotypical manner across neuroanatomically connected networks [142]. Given that tau pathology propagates, at least partially, via clathrin-mediated endocytosis (CME) [143], a clathrin adaptor such as PICALM may be involved in clathrin-mediated tau internalization.
In addition to the prominent role of LRP1 in Aβ clearance [144], recent studies suggest the involvement of LRP1 in regulating tau seeding and internalization [145]. LRP1 is expressed in neuronal, glial and endothelial cells [146]. The affinity of the tau microtubulebinding domain to LRP1 is high under physiological conditions, while hyperphosphorylation of tau leads to a reduction of its binding affinity to LRP1 [147]. Therefore, their interaction may be weaker in AD brains. Since PICALM directly binds to LRP1 to regulate the internalization of LRP1 [34], it can be speculated that PICALM modulates pathological tau spreading via LRP1-mediated pathological tau internalization.
PICALM also modulates cholesterol biogenesis and lipoprotein uptake (see Section 2.1.2). Cholesterol is implicated in amyloid pathology [148] and is detected in amyloid plaques [149]. Since cholesterol depletion promotes tau uptake and aggregation [150], PICALM-mediated misregulation of cholesterol homeostasis may be related to both amyloid and tau pathology progression.
PICALM is involved in iron uptake via transferrin (see Section 2.1.2). Iron in Fe 3+ (redox-inert state) is detected in NFTs of AD [151] and is implicated in tangle formation [152] and in APP processing [153]. Given PICALM is misregulated in AD brains, the PICALMmediated iron uptake may be involved in AD pathogenesis.
A study using the HEK tau biosensor cell model [154] has suggested no effect of deletion of GWAS hit genes, including PICALM, on the FRET-based signal of tau-tau interaction [155]. Since a lipofectamine-based method was used for PHF internalization, Kolay et al., focused on the effect of a GWAS hit gene depletion on FRET-based tau-tau interaction but not on the effect on the "physiological" uptake of PHF. It remains elusive if PICALM overexpression modulates PHF uptake or tau-tau interaction. Further studies are necessary to assess the role of PICALM in modulating the prion-like tau pathology propagation in in vitro and in vivo models.

PICALM and Glial Cells
Evidence suggests that PICALM immunoreactivity in microglial cells is increased in the postmortem hippocampus of LOAD compared to non-demented controls. Such a striking increase was not evident in human FAD brains ( Figure 6D-F) [33]. These observations are in line with recent studies showing the potential involvement of glial cells in AD pathogenesis [156]. Indeed, single cell transcriptome from postmortem AD brains suggests that a group of genes, including PICALM, APOE, TREM2, MEF2C and the MHC class II genes HLA-DRB1 and HLA-DRB5 are upregulated in microglial cells in correlation with AD pathology [126].
APOE4 carriers exhibit endocytic defects before significant amyloid deposition [157]. PICALM is described as a modifier of APOE4-induced endocytic defects in human iPSCderived astrocytes and in yeast. PICALM and its yeast orthologue YAP1802 overexpression rescued APOE4-induced endosomal abnormalities in human iPSC-derived astrocytes and yeast, respectively [158]. Nonetheless, PICALM immunoreactivity is hardly detected in astrocytes in postmortem human brain tissues [33], and thus, the implication of PICALM in astrocytic endocytosis might be an adjunct contributor to AD pathogenesis.
Elevated levels of reactive oxygen species (ROS) induce neuronal lipid synthesis that leads to the sequestration of glial lipid droplets (LD) to delay neurotoxicity [159]. PICALM modulates this neuroprotective process of neuron-to-glia lipid transport. PICALM orthologue LAP knockdown in drosophila glia leads to a significantly reduced LD formation. PICALM knockdown in rat astrocytes leads to reduced uptake of neuronderived fatty acids [159]. PICALM and other proteins encoded by LOAD GWAS hits (APOE, BIN1, CD2AP, AP2A2 and RIN3) are thus involved in LD formation and/or neuron-glial lipid transfer pathway, and knockdown of PICALM leads to blunted lipid transfer.

Difficulties in Interpreting Expression Levels of PICALM in AD Brains
Cautions need to be taken to interpret the data about PICALM expression in human autopsy tissues. The results of PICALM expression are often based on the bulk tissue of human postmortem brains [32,33,124]. Since the cell type composition of brain tissue undergoes alterations during the progression of neurodegenerative disease, such bulk data may confound the interpretation of differences in gene expression between AD patients and controls.
In general, there is a widespread heterogeneity between mRNA and protein levels in postmortem human control and AD brains, including PICALM [124]. For instance, increased PICALM transcripts [32] and decreased PICALM protein level [33,34] was observed in AD brains. Due to such an obvious contradiction of PICALM mRNA and protein levels in AD, mRNA data cannot be simply used to predict the protein expression of PICALM. The first possible hypothesis of such divergence is a cellular autoregulatory mechanism with a negative feedback loop [160]. The second possible hypothesis is microRNA (miRNA)related gene expression regulations. Post-transcriptional repression of mRNA by miRNA has been well-studied in mammalian cells [161]. Studies suggest an important regulatory role of miR-155 in neuroinflammatory conditions, including AD [162]. Interestingly, several independent studies suggest the upregulation of miR-155 in various brain regions of AD patients compared to controls as well as in the brain of transgenic mouse models of Aβ and tau pathologies [163,164]. Indeed, it has been shown that PICALM is regulated by miR-155 in mouse brains [165]. Although there is no obvious effect of miRNA-155 knockout on the expression level of PICALM transcripts in the mouse brains of these models [166], the effect of miRNA-155 overexpression on PICALM (transcript and protein) expression in human AD brains remains elusive. While further study is necessary to better understand the potential regulatory mechanisms of PICALM expression, miR-155 may provide a key pathway to untangle the divergence of mRNA and protein levels of PICALM. It would be intriguing to find correlation between the PICALM level and miR-155 in further studies. In addition, post-transcriptional regulatory effects of other non-coding RNAs, such as long non-coding RNA and circular RNA on PICALM, need to be investigated in AD models. The last but not least hypothesis to explain the divergence of mRNA and protein levels is that PICALM may undergo abnormally accelerated proteolysis in AD brains. Several lines of evidence suggest that PICALM is cleaved by both calpain and caspase [33,79,80], which are activated in AD brains [138]. Moreover, post-translational modifications of PICALM might affect its degradation, stabilization, protein-protein interaction or subcellular localization, yet such effects remain largely elusive. Further studies are thus necessary to decipher the potential abnormalities of PICALM post-translational modifications using 2D Western blot analyses and mass spectrometry from control and AD brain samples to search for AD-specific post-translational modifications of PICALM and their implications in pathology.

Challenges in Deciphering the Effect of PICALM Variants on Its Expression
Several independent studies have recently reported a significant association of the LOAD-susceptibility PICALM risk variants (rs10792832 and rs3851179 that are in full LD) on the decreased expression in microglial cells [90,91]. It should be highly informative to create AD models in which microglial PICALM expression is modulated in order to unravel the effect of reduced or increased PICALM expression in microglia.
Additionally, in iPSC-derived endothelial cells, the protective rs3851179 T allele is associated with increased mRNA and protein expression of PICALM [34] and in bulk tissues when analyzed relative to cell-specific mRNA markers for microvessels and neurons [30]. Given that the effect of rs3851179 T is rather modest when normalized to astrocyte marker GFAP (glial fibrillary acidic protein), it becomes more remarkable when normalized to endothelial cell markers or neuronal markers [30]. Therefore, it is possible that PICALM variants may have distinct effects on its expression in different cell types.
To decipher the inconsistencies in mRNA and protein levels of PICALM observed in postmortem human AD brain samples [124], the PICALM levels in each cell type (e.g., microglia, neuron, astrocytes, oligodendrocytes, etc.) should be analyzed in large scale diverse cohorts by high throughput single cell sequencing and further validated by RT-qPCR, Western blotting, mass-spectrometry (in distinct cell-type fractions) and by in situ hybridization and/or immunohistochemistry (in sections) of postmortem human AD brain samples [167]. As shown by Zhao et al., on the functional effect of rs3851179 on PICALM expression in iPSC-derived endothelial cells, it should be interesting to differentiate iPSC isogenic for rs3851179 T/C into various cell types to examine the expression level of PICALM in distinct cell types. More work is necessary to increase the number of samples from different cohorts to improve the power of statistical analyses and pursue proteomic analyses considering heterogeneous cell populations in the brain.

Potential Role of PICALM in Cellular Homeostasis
PICALM is involved in various pathological processes, such as Aβ production and transcytosis, as well as tau pathology progression (Figure 7). Given that PICALM plays versatile roles in iron homeostasis, glutaminergic neurotransmission [117], endocytosis, and autophagy, PICALM may as well be implicated in the "cellular phase" of AD proposed by Bart De Strooper and Eric Karran [168]. In this hypothesis, they propose that the accumulation of amyloid and tau pathologies at the biochemical phase is rather tolerated by neuronal cells. However, the clinical signs manifest only when the cellular homeostatic mechanisms fail along with dysfunction of the neurovascular unit, abnormal neuronal network activity and dyshomeostasis of astrocytes and microglial cells. PICALM and other endocytic proteins encoded by AD susceptibility genes may be implicated in such steps that lead to dysfunction and breakdown of finely tuned cellular physiological functions. PICALM may be a small piece of complex endocytic machinery that may be vulnerable to neurodegeneration [169]. differentiate iPSC isogenic for rs3851179 T/C into various cell types to examine the expression level of PICALM in distinct cell types. More work is necessary to increase the number of samples from different cohorts to improve the power of statistical analyses and pursue proteomic analyses considering heterogeneous cell populations in the brain.

Potential Role of PICALM in Cellular Homeostasis
PICALM is involved in various pathological processes, such as Aβ production and transcytosis, as well as tau pathology progression (Figure 7). Given that PICALM plays versatile roles in iron homeostasis, glutaminergic neurotransmission [117], endocytosis, and autophagy, PICALM may as well be implicated in the "cellular phase" of AD proposed by Bart De Strooper and Eric Karran [168]. In this hypothesis, they propose that the accumulation of amyloid and tau pathologies at the biochemical phase is rather tolerated by neuronal cells. However, the clinical signs manifest only when the cellular homeostatic mechanisms fail along with dysfunction of the neurovascular unit, abnormal neuronal network activity and dyshomeostasis of astrocytes and microglial cells. PICALM and other endocytic proteins encoded by AD susceptibility genes may be implicated in such steps that lead to dysfunction and breakdown of finely tuned cellular physiological functions. PICALM may be a small piece of complex endocytic machinery that may be vulnerable to neurodegeneration [169].

Figure 7.
Effects of PICALM on AD pathogenesis pathways and neurodegeneration. PICALM is involved in multiple pathways of AD pathogenesis. PICALM regulates pathways related to APP processing and Aβ-related neurodegeneration (shown in pink). PICALM plays critical role related to both tau and Aβ pathologies, such as cholesterol dyshomeostasis, inflammation, iron dyshomeostasis and autophagy (yellow). PICALM is also crucial for neuronal toxicity via glial lipid transport, synaptic functions, glutaminergic transmission and modifying Aβ-toxicity (blue).

Conclusions
Multiple lines of evidence suggest that human PICALM polymorphism modulates AD pathogenesis. PICALM is involved not only in Aβ production and clearance but also in tau-mediated neurodegeneration and in endocytic homeostasis. The exact role of the Figure 7. Effects of PICALM on AD pathogenesis pathways and neurodegeneration. PICALM is involved in multiple pathways of AD pathogenesis. PICALM regulates pathways related to APP processing and Aβ-related neurodegeneration (shown in pink). PICALM plays critical role related to both tau and Aβ pathologies, such as cholesterol dyshomeostasis, inflammation, iron dyshomeostasis and autophagy (yellow). PICALM is also crucial for neuronal toxicity via glial lipid transport, synaptic functions, glutaminergic transmission and modifying Aβ-toxicity (blue).

Conclusions
Multiple lines of evidence suggest that human PICALM polymorphism modulates AD pathogenesis. PICALM is involved not only in Aβ production and clearance but also in tau-mediated neurodegeneration and in endocytic homeostasis. The exact role of the polymorphism on PICALM expression may vary in different cell types and contribute to multiple pathways in AD pathological mechanisms. Further studies are necessary to decipher the AD-specific post-translational modifications and the functional roles of LOAD-related susceptibility alleles to provide mechanistic insights into the pathological dysregulation of PICALM.
Author Contributions: All the coauthors participated in constructing the concept and writing the manuscript. K.A. and J.-P.B. contributed conception and design of this article. All authors contributed to the manuscript revision. All authors have read and agreed to the published version of the manuscript.