ERp57 chaperon protein protects neuronal cells from Aβ‐induced toxicity

Abstract Alzheimer's disease (AD) is a neurodegenerative disorder whose main pathological hallmark is the accumulation of Amyloid‐β peptide (Aβ) in the form of senile plaques. Aβ can cause neurodegeneration and disrupt cognitive functions by several mechanisms, including oxidative stress. ERp57 is a protein disulfide isomerase involved in the cellular stress response and known to be present in the cerebrospinal fluid of normal individuals as a complex with Aβ peptides, suggesting that it may be a carrier protein which prevents aggregation of Aβ. Although several studies show ERp57 involvement in neurodegenerative diseases, no clear mechanism of action has been identified thus far. In this work, we gain insights into the interaction of Aβ with ERp57, with a special focus on the contribution of ERp57 to the defense system of the cell. Here, we show that recombinant ERp57 directly interacts with the Aβ25−35 fragment in vitro with high affinity via two in silico‐predicted main sites of interaction. Furthermore, we used human neuroblastoma cells to show that short‐term Aβ25−35 treatment induces ERp57 decrease in intracellular protein levels, different intracellular localization, and ERp57 secretion in the cultured medium. Finally, we demonstrate that recombinant ERp57 counteracts the toxic effects of Aβ25−35 and restores cellular viability, by preventing Aβ25−35 aggregation. Overall, the present study shows that extracellular ERp57 can exert a protective effect from Aβ toxicity and highlights it as a possible therapeutic tool in the treatment of AD.


| INTRODUC TI ON
Alzheimer's disease (AD) is a chronic disease that is estimated to affect about 47 million people worldwide (Tiwari et al., 2019). AD is a late-onset disease (80-90 years of age) and is the leading cause of dementia, beginning with memory loss. In the past years, the diagnosis of genetic forms of AD has been greatly improved, but the etiology of the sporadic forms remains debated (Armstrong, 2013). Even if symptoms can be ameliorated, to date there is no treatment that can change the outcome of the disease (Sheppard & Coleman, 2020;Tiwari et al., 2019). AD is characterized by neuritic extracellular amyloid plaques in the brain, with loss of neurons and synapses (Cabral-Miranda & Hetz, 2018;Masters et al., 2015). Neuronal damage in AD is induced by the aberrant accumulation of amyloid aggregates and neurofibrillary tangles, which consist of amyloid beta peptide (Aβ) and phosphorylated tau protein, respectively (Tiwari et al., 2019).
Aβ is generated from the Amyloid Precursor Protein (APP) by the sequential action of two secretases, that is, the β-and γ-secretase (Sheppard & Coleman, 2020) and many studies show that its accumulation and a conformational change into forms with a high βsheet structure is key in the pathogenesis of AD (Zhao et al., 2012;Zou et al., 2019). Even though the predominant forms of Aβ in the human brain are Aβ 1−40 and Aβ 1−42 , other Aβ peptides can be present. Among them, the Aβ 25−35 fragment can be found in the senile plaques (Kaminsky et al., 2010). Aβ 25−35 is produced in the brain of the elderly because of the cleavage of racemized Aβ 1−40 . It has been hypothesized that Aβ 25−35 represents the biologically active region of Aβ, being the shortest fragment retaining the toxicity of the full-length peptide and which tends to aggregate and form fibrils (Clementi et al., 2005;Frozza et al., 2009;Millucci et al., 2010).
Administration of Aβ 25−35 recapitulates the pathological features of AD neurodegeneration and has been shown to lead to amnesia and memory loss in rats (Stepanichev et al., 2003) and to oxidative stress in the hippocampus of mice (Lu et al., 2009), suggesting its involvement in the pathogenesis of AD and its potential mechanism of toxicity. Because of this evidence, Aβ 25−35 has often been chosen as a model in structural and functional studies of Aβ-induced pathogenesis.
ERp57, also known as PDIA3, is a thiol oxidoreductase with Protein Disulfide Isomerase (PDI) activity. The main function of ERp57 is to promote the proper folding and to carry out quality control of newly synthesized glycoproteins in the lumen of the Endoplasmic Reticulum (ER). This enzyme works in conjunction with the lectin chaperones Calnexin and Calreticulin to ensure that secretory and membrane proteins are in the correct conformation before leaving the ER (Leach et al., 2002;Williams, 2006). ERp57 consists of four thioredoxin-like domains, named a, b, bʹ, and aʹ. The a and aʹ domains are catalytically active, with Cys-Gly-His-Cys active site motifs, and bind unfolded proteins substrates, while the b and bʹ domains contain a positively charged calnexin binding site, that assists in the substrate recruitment process (Dong et al., 2009a;Kozlov et al., 2006). Activities of ERp57 were reported in different cellular compartments (Turano et al., 2011). It binds STAT3 in the cytoplasm and ER (Coe et al., 2010), in the nucleus it has DNA binding ability (Chichiarelli et al., 2007;Coppari et al., 2002), and it has been identified on the plasma membrane as the cell surface receptor for the metabolite 1,25-dihydroxycholecalciferol (Boyan et al., 2012;Nemere et al., 2004). Intriguingly, ERp57 has been detected in the extracellular space of different cell types, for example, to suppress complement activation or as an early sign of renal fibrosis onset (Dihazi et al., 2013;Eriksson et al., 2019), suggesting that ERp57 might have different, currently unknown, extracellular functions. Moreover, ERp57 is a stress-responsive protein which is up-regulated in response to ER stress (Nundlall et al., 2010;Turano et al., 2011). ER stress happens when the equilibrium between folding capacity and protein load in the cell is altered and misfolded proteins accumulate in the ER. As a result, the cell initiates a signaling network called Unfolded Protein Response (UPR) in an attempt to reestablish the cellular homeostasis (Ghemrawi & Khair, 2020). Precisely, ERp57 is one of the players that enhance the folding capacity of the cell (Bargsted et al., 2016). Chronic ER stress has been linked to different neurodegenerative diseases, which share the common feature of accumulation of protein aggregates in the central nervous system, like AD, Parkinson's disease (PD), Huntington disease, and Amyotrophic Lateral Sclerosis (ALS). It has been proven that the expression of the misfolded protein characteristic of the disease is sufficient to induce UPR in cellular and animal models (Cabral-Miranda & Hetz, 2018;Gerakis & Hetz, 2018).
ERp57 has been previously linked to key processes in neurodegeneration and particularly in AD pathogenesis. It has been demonstrated that cerebrospinal fluid of healthy individuals contains Aβ peptide complexed with ERp57 and calreticulin (Erickson et al., 2005) and that ERp57 interacts with full-length APP in the early secretory pathway of this protein (Selivanova et al., 2007). This evidence suggests that ERp57 interacts with Aβ and may play a role in preventing its aggregation. As evidence of this, activation of ERp57 with diosgenin significantly improves performance of object recognition memory and reduces amyloid plaques and neurofibrillary tangles in the cerebral cortex and hippocampus in the AD 5XFAD mice model (Tohda et al., 2012). Moreover, involvement of ERp57 in other neurodegenerative diseases has been demonstrated. Indeed, it was found that the expression and function of ERp57 could be modulated under conditions of oxidative stress in neuronal cells in culture and in a PD animal model (Aureli et al., 2014;Giamogante et al., 2018), suggesting a feedback loop mechanism, and that ERp57 has a protective role to motor function in early stages of ALS progression, preserving the neuromuscular junction (Rozas et al., 2021) and in prion neurotoxicity (Hetz et al., 2005;Thapa et al., 2018).
In this work, we show how ERp57 is involved in the cellular response to Aβ-induced stress through a previously unreported mechanism. In this respect, we employed SH-SY5Y neuroblastoma cells treated with micromolar amounts of Aβ 25−35 , a wellestablished AD cellular model (Xicoy et al., 2017)

| GST protein production
To express GST, the plasmid pGEX-4 T-1 (catalog number #28954549, GE healthcare), bearing the GST gene, was transformed into the E. coli BL21 (DE3) strain. The recombinant bacteria were grown in 1-liter LB medium (catalog number #L3397) in the presence of ampicillin (50 μg/ml) (catalog number #A9393). When the culture had reached an optical density (OD600) of 0.6, isopropylβ-D-thiogalactopyranoside (IPTG) (catalog number #PHG0010) 1 mM was added to induce protein expression and then growth was continued for 3 h. The cells were spun down at 3000 g, resuspended in 50 mM HEPES +150 mM NaCl +0.1 mM EDTA at pH 7.4 (buffer A) + 1 mM PMSF (catalog numbers #H3375, #S9888, #E9884, #PMSF-RO), sonicated and centrifuged at 15 000 g for 45 min. The supernatant was loaded onto a GSTrap FF affinity column (catalog number #11340212 GE healthcare), equilibrated with buffer A, washed extensively with buffer A, and eluted with elution buffer (50 mM Tris-HCl, 10 mM reduced glutathione, pH 8.0) (catalog numbers #10812846001, #G4251) at 1 ml/min flow rate. Optical density was monitored at 280 nm using the optical unit of a liquid chromatography system (AKTA P-900, GE Healthcare BioScience AB). A putative peak containing the recombinant GST protein was identified at 20% elution buffer, collected for analysis, and dialyzed in 20 mM HEPES at pH 7.5. The total protein concentration of each fraction was determined using a Micro BCA kit (catalog number #23252, Pierce) using bovine serum albumin as the reference protein. The purity of the protein sample was analyzed using aliquots of the fractions by 12% SDS-PAGE (catalog number #5678044, Biorad) and western blotting with appropriate antibodies.
The cell line is not listed as a commonly misidentified cell line by the ICLAC (http://iclac.org/datab ases/cross -conta minat ions/). Therefore, we did not perform any authentications of these cell lines.

| Surface plasmon resonance
SPR experiments were performed with a SensiQ Pioneer apparatus. ERp57 immobilization was carried out essentially as previously reported (Genovese et al., 2020;Poser et al., 2017).

| Bioinformatic analysis
The FASTA sequence of the Human ERp57 protein was retrieved from the Uniprot server (UniProt) (Bateman et al., 2021).

| Statistical analysis
Experiments were repeated at least in triplicate and all the results are expressed as the mean value ± standard deviation (SD).
No formal randomization procedures were applied when allocating treatments to different experimental groups, and no blinding was performed during data analysis. No exclusion criteria were pre- determined. An assessment of the normality of the data was not performed. No corrections were applied. No sample calculation was performed. No test for outliers was conducted.
Post hoc power analysis was performed using G*Power 3.1 software (Faul et al., 2007) which verified that the sample size was sufficiently powered (input parameters: two tails, 0.05 α error probability, 0.95 (1β error) power and effect size d was calculated using mean and standard deviation (σ values of groups).
p-values were calculated using a two-tailed Student's t-test or oneway or two-way analysis of variance (ANOVA) with Tukey's multiple comparison post-tests. p values <0.05 were regarded as significant. All statistical analyses were performed using Prism 6 software (GraphPad) (RRID: SCR_002798). Full statistical report is included in supplementary.
In western blot analyses, two extracellular samples (1-3 h) were excluded because it exceeded ±2 SD. One sample (24 h, intracellular and extracellular) was lost.

| Ethics approval
No ethical approval was required for the study.

| Aβ 25−35 treatment of neuronal cells in culture decreases intracellular and increases extracellular protein levels of ERp57
Increasing

| ERp57 cellular localization upon Aβ 25−35 treatment
ERp57 is localized in different cellular compartments, though it is prevalently found in the ER where it exerts its disulfide-isomerase and redox activities (Turano et al., 2011). Since we observed an ERp57 secretion upon Aβ 25−35 treatment, we evaluated ERp57 intracellular localization by immunofluorescence. As expected, in control conditions, ERp57 appears as a widespread intracellular signal, mainly ascribable to the ER compartment ( Figure 2).

| Surface plasmon resonance experiments demonstrate a direct interaction between ERp57 and Aβ 25−35
Although evidence from the literature reports the biological link The dissociation phase is biphasic, an indication that Aβ 25−35 can bind two ERp57 sites, that is a high-affinity site and a low-affinity site.
Our SPR experiment does not indicate which one of the ERp57 domains is involved in the binding of Aβ 25−35 , so we decided to perform an in silico docking in order to elucidate the putative sites of interaction.

| ERp57 rescues the reduction in SH-SY5Y cellular viability induced by Aβ 25−35
To expand our findings, we aimed to investigate whether recom-

| ERp57, inside and outside the cell
ERp57 is an endoplasmic reticulum disulfide isomerase which modulates folding of newly synthesized glycoproteins by recognizing and interacting with peptides or incorrectly folded proteins, thereby promoting formation of disulfide bonds and correct protein folding (Torres et al., 2015). Conversely, Aβ is a small peptide that, because of its intrinsic characteristics, tends to aggregate and to escape the ER-associated degradation process (ERAD) (Holtzman, 2013).
Eventually, monomeric Aβ is released out of the cell and begins its slow aggregation into fibrils that will form the template for subsequent oligomers (Törnquist et al., 2020). undertaken to test the hypothesis that ERp57 is released outside the neuronal cell following stress induced by amyloid aggregation, thus binding Aβ and blocking its toxic effects. This hypothesis stems from the first studies by Erickson (Erickson et al., 2005), who found ERp57

| ERp57-Aβ interaction and biological significance
Many literature reports indicate that ERp57 is a multifunctional protein with a high propensity to interact with peptides and proteins prone to misfolding, such as PrP (Hetz et al., 2005;Thapa et al., 2018;Torres et al., 2015) and α-synuclein (Serrano et al., 2020). Moreover, other members of the PDI family are able to bind tau protein (Xu et al., 2013) and proteins without disulfide bonds (Cai et al., 1994).
Additionally, Irvine has demonstrated that PDI binds more effectively the unfolded proteins than the folded ones (Irvine et al., 2014).
As ERp57 belongs to the PDI family, this could imply that it is able to recognize specific structural features responsible for the aggregation process. Here, we demonstrated that ERp57 binds Aβ 25−35 with micromolar affinity, and that ERp57 impairs Aβ 25−35 peptide aggregation. SPR experiments indicated that ERp57 is able to bind to Aβ 25−35 peptide, which is a short fragment of the full-length Aβ 1−42 .
The dissociation phase was found to be biphasic, an indication that Aβ 25−35 can bind two ERp57 sites, that is a high-affinity site and a low-affinity site.
Our in silico study reinforced these data and showed that the first pocket is placed on the domain, in the proximity of the CHGC catalytic site, in line with the description of Xu and coworkers (Xu et al., 2013), who demonstrated that PDI protein is able to reduce the formation of tau fibrils within the cell, by interacting with the thioredoxin-like catalytic domain a and aʹ; on the other hand, the second pocket is between the bʹ and b domains, in line with Kozlov, who proposed that the bʹ domain of ERp57 provides the majority of the binding site with substrates, while the b domain affords additional contacts (Kozlov et al., 2006), strengthening the interaction.
The thioflavin T experiments demonstrated that the interaction between Aβ 25−35 and ERp57 strongly reduced the aggregation of the peptide but was not able to disassemble pre-aggregated Aβ 25−35 .
We thus hypothesize that the mechanism by which ERp57 exerts its neuroprotective action resides in ERp57 impairment of peptide aggregation, rather than in disaggregation of existing fibrils. The inhibition of the aggregation was induced by low amounts of ERp57 protein, present in the reaction mixture in a very low molar ratio compared to the peptide (1:250), in line with the experiments conducted by Serrano and coworkers, who demonstrated that ERp57 is able to reduce the aggregation of α-synuclein in a very efficient manner, in a 1:50 molar ratio (Serrano et al., 2020).

| ERp57 in neurodegenerative diseases
As previously reported (Williams, 2006), it is known that ERp57 together with calreticulin and calnexin forms a complex capable of interacting with glycoproteins, including APP, at the Golgi level, and therefore it can be hypothesized that the contact regions between ERp57 and APP are maintained at least in part also by Aβ, despite the cuts suffered by the secretases. Indeed, Selivanova (Selivanova et al., 2007) has demonstrated that ERp57 in permeabilized cells interacts with full-length APP during the process of O-glycosylation, whereas it does not interact with the c99 fragment of APP generated by the β-secretase activity; nevertheless, the same author admits that the interaction observed by Erickson (Erickson et al., 2005) could be explained by a different interaction between ERp57 and Aβ 1−42 .
Holtzman hypothesizes that a decline in the ER ability to catalyze post-translational changes on the protein APP could be at the base of AD. Indeed, the N-glycosylation of the APP is a fundamental step for the subsequent cleavage by the secretases and the folding assisted by the calreticulin-ERp57 complex. If the APP does not undergo the correct post-translational changes and is not folded correctly, the resulting Aβ peptides aggregate rapidly and become too bulky to undergo the process of ER-associated degradation (ERAD).
Thus, aggregated Aβ remains in the lumen of the ER and is eventually secreted outside the cell (Erickson et al., 2005;Holtzman, 2013).
In line with this theory, Chun demonstrated that the correct glycosylation of the residue T576 of the APP is fundamental for the correct cleavage of the APP (Chun, Kwon, et al., 2015). Correct glycosylation of APP decreases its endocytosis and increases trafficking from Trans-Golgi-Network to the cell surface, resulting in an increase in non-amyloidogenic processing and a decrease in Aβ production .
ERp57 is able to guarantee the correct folding of APP after glycosylation, but it could also be able to perform a regulatory function outside the cell, binding Aβ and possibly facilitating its phagocytosis and digestion by the microglial cells.
In this context, our work shows how ERp57 is able to interact with Aβ 25−35 probably by two different sites, with micro-molar affinity. In vitro experiments show that ERp57 is able to bind Aβ 25−35 probably in an oligomeric form (considering the sub-stoichiometric ratio of the assay), but fails to disaggregate a fibril, while cell experiments show that Aβ 25−35 bound by ERp57 no longer leads to toxicity.
It could be speculated that in AD patients this regulation is lost, possibly because less ERp57 is secreted or the oxidative environment is not favorable, because of the aging process (Ghosh & Brewer, 2014) and therefore the Aβ, released outside the cell, is no longer digested by enzymes such as neprilysin (Campos et al., 2020) and Insulin Degrading Enzyme (Bulloj et al., 2010), triggering a fibrillation process that will eventually lead to the formation of amyloid aggregates (Erickson et al., 2005;Zhao et al., 2012). To date, it was reported that neuronal  and SH-SY5Y cells are able to internalize Aβ (Ida et al., 1996) and pre-formed alpha-synuclein fibrils (Pantazopoulou et al., 2021) through endocytosis and clear them, so it could be hypothesized that ERp57, which is able to bind alpha-synuclein (Serrano et al., 2020), is released from the neuronal cells in order to bind the Aβ, prevent its aggregation and favor the endocytosis of this complex.

| Future perspectives
To date, there are no cures for AD, and the currently approved drugs can only mitigate symptoms such as depression and agitation (Arvanitakis et al., 2019). Hence, Aβ production and clearance are key pathways in the development of therapeutic strategies for AD. Finding protein interactors able to bind Aβ and reduce its aggregation could represent a possible therapeutic approach to slow down the onset and progression of the disease, especially in cases of familial AD (Magzoub, 2020;Soares et al., 2021). ERp57 has been studied extensively in the literature, since it is an important chaperone, but only recently attention was paid to the functions that it can perform in the extracellular space in relation to protein aggregates. Our work provides clues on the possible role of ERp57 in AD pathogenesis and suggests recombinant ERp57 peptides as a possible therapeutic approach for this pathology.
Interestingly, new therapeutic approaches focus on recombinant proteins which interact with Aβ hampering its aggregation (Magzoub, 2020), hence ERp57 could represent a possible therapeutic tool useful in counteracting neurodegeneration induced by amyloid aggregation.

ACK N OWLED G M ENTS
This work was financially supported by funds from Ateneo Sapienza.

CO N FLI C T O F I NTE R E S T
The authors declare no competing financial interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.