Inhibition of heat shock proteins increases autophagosome formation, and reduces the expression of APP, Tau, SOD1 G93A and TDP-43

Introduction

The pathologies Alzheimer’s Disease (AD), familial Amyotrophic Lateral Sclerosis (ALS) and Huntington’s Chorea (HC) are all in part mediated by the aberrant expression and denaturation of disease-specific proteins [1–6]. In AD, amyloid-β and Tau proteins form toxic aggregates inside neurons and in the extracellular environment. The majority of familial ALS is caused by expression of mutant variants of SOD1, e.g., SOD1 G93A. Toxic protein aggregates, such as those containing TDP-43, FUS and SOD1 G93A, are causal in neuronal cell death. An additional major component of brain biology in ALS patients are activated / reactive microglia [7, 8]. The reactive microglia create a toxic inflammatory micro-environment liminal to the dysfunctional neurons. In HC, multiple additional genetically unstable CAG repeats result in translation of a Huntingtin protein which is more readily cleaved by cellular proteases to form insoluble toxic aggregates in neurons.

Protein expression and protein degradation are tightly controlled processes. An important regulator of protein translation is the factor eIF2α [9–12]. Phosphorylation of eIF2α S51 inactivates its function, and bulk protein translation is reduced. Several kinases phosphorylate eIF2α, including PKR-like endoplasmic reticulum kinase (PERK). The bi-functional protein chaperone GRP78 binds to PERK, keeping it inactive [13–16]. GRP78 senses the amount of denatured protein in the ER and cytosol, and as the levels of denatured proteins increase, GRP78 is released from PERK, and then acts as a chaperone to renature the misfolded proteins. PERK becomes active, phosphorylates eIF2α thus stopping translation until GRP78 and other chaperone proteins of the HSP90 and HSP70 families have renatured the misfolded proteins. GRP78 then can re-associate with PERK, inactivating the kinase, and eIF2α is dephosphorylated and activated by protein serine/threonine phosphatase 1.

Autophagy is an evolutionary conserved mechanism by which cells recycle nutrients to maintain viability [17–22]. Proteins that cannot be renatured by chaperones are degraded via autophagy (along with the associated chaperone protein). Although phosphorylation of eIF2α inhibits translation from ~95% of genes, it also stimulates translation of alternate genes including the autophagosome regulatory proteins Beclin1 and ATG5 [23]. Hence ER stress signaling to restore homeostasis acts to both renature proteins and simultaneously to facilitate the break down excess protein load via autophagy. In AD the clearance of amyloid-β plaques can be mediated via autophagy, which correlates with improved in vivo mouse memory function [22, 24]. Similar in vitro and in vivo findings to amyloid-β have been made for ALS aggregates and mutant Huntingtin aggregates [21, 25].

Our prior studies in oncology, specifically developmental cancer therapeutics, have shown how drugs such as neratinib and niraparib, which induce ER stress and cause autophagosome formation through different mechanisms, can be combined to synergistically kill tumor cells, e.g. [26, 27]. Several of these concepts have been translated into the clinic [28, 29]. The drug AR12 was originally developed and translated into the clinic as a cancer therapeutic which killed cells via autophagy and ER stress signaling, however, when its true mechanism of action was discovered, as a pan-chaperone ATPase inhibitor, and particularly as an inhibitor of GRP78, the drug was repurposed into the infectious disease field as an anti-viral drug [NCT00978523, ASCO Abs. #2068, 2014; 30–37]. AR12 prevents the replication of viruses such as SARS-CoV-2 by preventing them from hijacking chaperone functions [38–40]. These attributes suggest to us that AR12 may have utility as an AD therapeutic by causing ER stress and autophagy. In the present studies we have used isogenic HCT116 colon cancer cells as a model system to examine these processes.

Recent studies have shown that HER2 and HER4 are over-expressed in the hippocampus of human AD brains and that degradation of HER2 enhances autophagic flux, causes degradation of amyloid-β, and improves the cognitive functions of APP/presenilin-1 (PS1) mice [41, 42]. Neratinib down-regulates RAS expression and signaling by RAS proteins has been linked to increased expression of Tau and amyloid-β proteins [43, 44]. Neratinib also inactivates the Hippo pathway and recently Hippo signaling was shown to be activated in amyloid-β-mediated neurodegeneration and that inhibition of Hippo signaling rescues the neurodegeneration effect [45, 46]. As neratinib down-regulates HER2, HER4 and RAS, inactivates Hippo and inactivates mTOR and causes autophagosome formation, we also hypothesized repurposed neratinib could safely reduce misfolded protein levels in AD.

Our oncology studies revealed that the multiple sclerosis drugs fingolimod and dimethyl fumarate interacted to kill glioblastoma cells which was mediated by increased autophagosome formation [47, 48]. The drugs prevented freshly isolated reactive microglia, from a patient GBM tumor, from synthesizing IL-6, TNFα and TGFβ. As the pathologies of AD, ALS and HC all involve sustained expression of denatured toxic proteins, and as our oncology and virology-based drug combinations stimulate autophagosome formation, ER stress and reduced protein translation and autophagic digestion of proteins, we determined whether our repurposed drug combinations using transformed / tumor cell systems, not neurons, could reduce expression of Tau and APP via autophagy.

Materials and Methods

Results

Our initial studies were designed to demonstrate that AR12 and neratinib, and fingolimod and MMF, interacted to cause autophagosome formation and autophagic flux. Treatment of HCT116 ATG16L1 T300 colon cancer cells with AR12 or neratinib increased the activities of the AMPK, ULK1 and ATG13 (Figure 2). The activities of mTORC1 and mTORC2 declined. The drugs combined to cause further activation of PKR-like endoplasmic reticulum kinase (PERK) that was associated with increased phosphorylation, and inactivation of eIF2α.

HCT116 ATG16L1 T300 colon cancer cells and GBM6 cells, a PDX glioblastoma isolate expressing truncated EGFR vIII, were transfected with a plasmid to express LC3-GFP-RFP and with siRNA molecules to knock down expression of AMPKα₁ or eIF2α. In the neutral autophagosome both GFP and RFP fluoresce whereas in the acidic autolysosome GFP is quenched and only RFP fluoresces. Thus, the appearance and disappearance of GFP+/RFP+ and RFP+ vesicles over time permits the detection of autophagosome formation and autophagic flux where autophagosomes fuse with acidic lysosomes to form autolysosomes. AR12 and neratinib, in agreement with the elevated phosphorylation of ATG13, both increased autophagosome formation in the HCT116 cells within 4h of exposure and combined to further promote autophagosome formation (Figure 3A, upper graphs). Autophagosome formation was significantly reduced by knock down of AMPKα₁ or eIF2α. Eight hours after drug exposure, the levels of autophagosomes had declined and the numbers of RFP+ autolysosomes had increased, with this effect also being significantly suppressed by knock down of AMPKα₁ or eIF2α.

In transfected GBM6 cells, fingolimod and MMF both enhanced autophagosome formation, with the combination causing greater levels of autophagy, and these effects were blocked by knock down of AMPKα₁ or eIF2α (Figure 3A, lower graphs). As was observed in the upper graphs, the fingolimod / MMF drug combination caused autophagic flux which was again blocked by knock down of AMPKα₁ or eIF2α. From Addgene we purchased plasmids to express a Tau-GFP fusion protein (“wild type”) and to express a mutant Tau 301L-GFP fusion protein (“mutant”). HCT116 cells were transfected with a plasmid to express “wild type” Tau-GFP and treated with AR12 or neratinib. Treatment with AR12 reduced Tau-GFP levels, as did neratinib (Figure 3B). AR12 did not apparently alter Tau sub-cellular localization. Our prior studies with neratinib had demonstrated that the drug rapidly caused HER family receptors and RAS proteins to become localized in punctate intracellular vesicles, and in the present studies, neratinib, unlike AR12, also caused Tau-GFP to rapidly exhibit a punctate staining pattern [43].

We then performed studies to define dose responses for AR12 inhibiting the chaperone ATPase activities of HSP90, HSP70 and GRP78. As an internal control, we observed a concentration-dependent increase in fluorescence from the ATP-lite substrate (Figure 4A). In a dose-dependent fashion AR12 inhibited the ATPase activities of HSP90, HSP70 and GRP78 (Figure 4B–4D). Of note was that the inherent ATPase activity of GRP78 in our analyses was considerably lower than those of HSP70 and HSP90; the basal activity was significantly less as measured by the assay and was generated by twice as long an incubation compared to HSP90 and HSP70. The approximate IC50 inhibitory concentration for HSP70 was ~400 nM, for HSP90 ~300 nM and for GRP78 ~80 nM. In heavily pre-treated cancer patients treated with AR12, the C max of the drug at the RP2D of 800 mg BID varied between 2 μM and 8 μM. We previously published that AR12 crossed the blood-brain barrier, with a brain C max of ~2 μM after administration of a 50 mg/kg bolus [37]. Collectively, our findings support the use of AR12 as a therapeutic anti-chaperone agent in the brain.

We next defined the impact of neratinib and AR12 as single agents and when combined on the expression of protein chaperones. AR12 significantly reduced chaperone expression whether the levels were measured using an antibody directed at the COOH terminus or NH2 terminus of the protein (Figure 5, upper). Notably for GRP78 the reduced staining for the NH2 terminus of the protein did not match that observed for the COOH terminal epitope, recapitulating our prior studies that AR12 binds to the ATP binding domain of GRP78 and alters protein conformation such that the NH2 antibody epitope has become occluded. For all of the tested chaperones, including HSP27 that does not bind ATP, neratinib and AR12 interacted to cause a significant further reduction in chaperone levels. We then determined whether autophagosome formation was responsible for the drug-induced decline in chaperone expression. Knock down of Beclin1 or ATG5 prevented neratinib and AR12 from reducing chaperone expression (Figure 5, lower). Thus, not only does AR12 catalytically inhibit chaperones, but the drug also causes their degradation through autophagy.

We next performed studies to link chaperone proteins to the expression and drug-induced degradation of Tau and APP. Cells were transfected to over-express GRP78, HSP90 or HSP70 and co-transfected to express Tau or APP. Over-expression of GRP78 and to a lesser extent HSP70 both significantly reduced the ability of AR12, neratinib and the drugs in combination to lower the levels of Tau (Figure 6A). Only GRP78 over-expression significantly reduced the abilities of the drugs as single agents or combined to lower APP expression. Cells were then transfected with siRNA molecules to knock down the expression of GRP78, HSP90 or HSP70, and in combination, and co-transfected to express Tau or APP, followed by drug exposure. Surprisingly, under basal conditions, knock down of GRP78 / HSP90 / HSP70 did not significantly alter the basal levels of Tau or APP expression or the basal levels of sIF2α S51 phosphorylation (Figure 6B). Knock down of GRP78 significantly enhanced the ability of AR12 and [AR12 + neratinib] to reduce the expression of Tau and APP. Knock down of GRP78 in combination with knock down of HSP90 or HSP70 caused significantly more protein degradation than GRP78 alone and a significantly greater amount of eIF2α phosphorylation.

Based on the data in Figure 6, we reasoned that if neratinib and AR12 could inhibit chaperone function and rapidly stimulate autophagosome formation, the drug combination would also be able to degrade other proteins including “wild type” and “mutant” Tau and APP proteins. African Green monkey kidney (Vero) cells and human HCT116 ATG16L1 T300 colon cancer cells were transfected to express “wild type” Tau or a “mutant” Tau, Tau 301L, which has been stated to be relatively protease resistant compared to the wild type protein, with both proteins tagged with GFP [53, 54] (Figures 7, 8). Over a time-course, neratinib and AR12 interacted to reduce the expression of wild type Tau or Tau 301L, either via GFP tag fluorescence or by immunostaining (Figures 7, 8). Transfection of the human HCT116 cells to knock down Beclin1 or ATG5 prevented the drug combination from reducing Tau or Tau 301L expression. Hence, we degrade not only the chaperones which associate with Tau proteins, but also the Tau protein itself (Figure 8B).

As presented in Figure 3, fingolimod and MMF interact to cause autophagosome formation and subsequently to promote autophagic flux. As was observed for AR12 and neratinib, fingolimod and MMF interacted to reduce the expression of wild type and mutant Tau 301L proteins (Figure 9A). Knock down of Beclin1 or ATG5 prevented Tau degradation (Figure 9B). An addition protein, beyond TAU and APP, that has been linked with the biology and pathology of AD is TDP-43 [1–8]. AR12 and neratinib, and MMF and FTY720, interacted to reduce the expression of TDP-43 (Figure 10A). Knock down of Beclin1 or ATG5 prevented either combination from reducing TDP-43 levels (Figure 10B).

The autophagy regulatory protein ATG16L1 is expressed as two isoforms, ATG16L1 T300 and ATG16L1 A300. The ATG16L1 T300 is most commonly found in Africans and their descendants whereas the highest prevalence of ATG16L1 A300 expression is in northern Europeans and their descendants. Expression of ATG16L1 A300 predicts for a higher incidence of Crohn’s Disease in European Americans [49]. In comparison to isogenic HCT116 cells expressing ATG16L1 T300, cells expressing ATG16L1 A300 were significantly less capable of forming autophagosomes or exhibiting autophagic flux when treated with AR12 and neratinib (Figure 11A). This is similar to prior studies using other drugs / drug combinations which induce autophagosome formation. We next determined whether the differential ability of each isoform to facilitate autophagy also impacted the ability of neratinib and AR12 to cause Tau protein degradation. As would be predicted from the data in panel A, cells expressing ATG16L1 A300 were less capable of reducing Tau expression (Figure 11B). We were surprised, however, to observe also that HCT116 ATG16L1 A300 cells were less capable of expressing Tau than were their isogenic HCT116 ATG16L1 T300 counterparts. Tau is being expressed from a plasmid via a constitutive promoter. To determine whether this effect was specific to Tau, we transfected cells with a plasmid and a constitutive promoter to express the SARS-CoV-2 spike protein. The ability of ATG16L1 T300 and ATG16L1 A300 cells to express the spike protein was identical (Figure 11C). And the ability of AR12 cells to cause spike degradation was reduced in the ATG16L1 A300 cells. As African-Americans trend to present with a more rapid onset and severe Alzheimer’s disease when compared to European Americans, we postulate that ATG16L1 isoform expression may play a role.

We extended our analyses of the biologies of HCT116 ATG16L1 T300 and HCT116 ATG16L1 A300 cells to determine whether altered chaperone levels, or drug responses, could provide clues as to why more Tau was expressed in the T300 cells. Regardless of isoform expression, cells expressed similar levels of HSP90 and HSP70, whereas expression of GRP78 was lower in the ATG16L1 A300 isoform cells (Figure 12). This was observed for cell surface levels of GRP78 and for bulk cellular GRP78. Although cells regardless of their isoform exhibited near identical levels of total ATG13 and total eIF2α, the phosphorylation of these proteins caused by either AR12 or neratinib was altered. In cells expressing ATG16L1 A300, the ability of AR12 and neratinib to stimulate ATG13 S318 phosphorylation (autophagy) and eIF2α S51 phosphorylation (ER stress) were reduced. Basal phosphorylation of ATG13 S318 trended towards being lower in the A300 cells (Figure 12). These findings may partially explain why cells expressing ATG16L1 A300 do not produce as many autophagosomes when stimulated as do ATG16L1 T300 cells.

In addition to denaturation of Tau, AD also contains denatured beta-amyloid plaques. Hence, we next determined whether our drug combinations could reduce the expression of wild type amyloid precursor protein (APP) or of the mutant forms of this protein APP692 and APP715. AR12 and neratinib both reduced APP expression and interacted to cause a further reduction in protein levels (Figure 13A). Knock down of Beclin1 or ATG5 prevented APP degradation. As single agents, fingolimod and MMF were apparently less efficacious than AR12 and neratinib at reducing the levels of the tested APPs (Figure 13B). However, fingolimod and MMF did interact to cause a further reduction in APP expression. Knock down of Beclin1 or ATG5 also significantly reduced the ability of fingolimod and MMF to reduce APP expression (Figure 13C). Collectively, our findings in Figures 4–13 support the further evaluation of AR12, neratinib, fingolimod and MMF as repurposed treatments for pathologies such as AD.

As noted previously, other lethal neurological diseases are also caused by accumulation of denatured toxic proteins including SOD1 G93A in ALS and Huntingtin in HC. Thus, based on our data with APP and Tau we determined whether our drug combinations could reduce the protein expression of SOD1 G93A and CAG 145 repeat Huntingtin. AR12 and neratinib interacted to reduce the expression of SOD1 G93A and Huntingtin, with Huntingtin being relatively less capable of being down-regulated (Figure 14A). In contrast to AR12/neratinib, and different to our findings with APP and Tau, fingolimod and MMF were ineffective at reducing protein expression. Knock down of Beclin1 or ATG5 prevented the drug-induced degradation of SOD1 G93A (Figure 14B). Additional studies will be required to understand why fingolimod and MMF did not cause significant amounts of degradation for these specific proteins.

Discussion

Toxic misfolded proteins are key drivers of AD, ALS, HC and other neurodegenerative diseases. In order to make progress against these diseases we need drugs that target these toxic proteins. In this paper we examined using isogenic colon cancer cells several existing drugs that function by increasing autophagy and degrading misfolded proteins.

Our initial studies tested AR12, neratinib, fingolimod, and MMF looking at autophagosome formation and autophagic flux. We demonstrated that AR12 and neratinib increased autophagosome formation in HCT116 cells and when these two drugs were combined showed even greater autophagosome formation. In GBM6 cells fingolimod and MMF enhanced autophagosome formation and the combination caused enhanced levels of autophagy. The increased autophagosome formation generated by these different drugs was reduced when AMPKα₁ or eIFα were knocked down, suggesting eIF2α dependent autophagosome formation.

After confirming these drug combinations led to increased autophagosome formation we looked more specifically at their ability to degrade Tau and APP. Tau and APP levels were decreased in HCT116 cells when treated with neratinib and AR12 or fingolimod and MMF. The autophagy dependent nature of Tau and APP degradation was shown by knocking down Beclin1 or ATG5. When Beclin1 or ATG5 were knocked down using siRNA the levels of Tau and APP recovered to the same levels as the control. Similarly, in GBM6 cells we showed autophagosome degradation of Tau and APP. While many but not all researchers postulate that an aggregated form of the amyloid-beta peptide is involved in the development and potential treatment and prevention of Alzheimer’s disease and recent trial findings have raised the possibility that amyloid-beta plaque-reducing antibodies may slow clinical progression in symptomatic people, the roles of amyloid-beta and APP in the development and potential treatment and prevention if AD needs to be clarified and confirmed in additional studies.

Aberrant expression of chaperone proteins is found in many human pathologies including cancer, in virology and in AD, ALS and HC. Chaperones fall into two broad families, the HSP70 family and the HSP90 family. Thirty-years ago the chaperone HSP90 was found complexed to protein kinases such as RAF-1, where it played a key role in maintaining RAF-1 stability and its interactions with RAS proteins and MEK1/2 [55, 56]. Pharmaceutical companies have made multiple attempts to drug HSP90, however all approaches have as of the present, failed in the clinic [57, 58]. This is due to the overlapping activities of the HSP90 and HSP70 families in terms of the proteins each can chaperone. Thus, inhibition of HSP90 caused a compensatory survival response with cells increasing HSP70 levels to maintain viability. Furthermore, high efficacy drugs which bound to the ATP binding site in HSP90 did not block the ATPase activities of HSP70 family chaperones. As a therapeutic agent AR12 would have been discarded by most drug companies because it is not a low-nanomolar inhibitor of chaperone ATPases. However, AR12 has similar IC50 inhibitory efficacy against both HSP90 and HSP70; cells cannot compensate for chaperone inhibition as AR12 reduces the functions of both families. The relative efficacy of AR12 as a chaperone inhibitor may also explain why the drug is not toxic in non-transformed cells; because even at concentrations above 1 μM, AR12 does not completely abolish chaperone ATPase activity which very likely provides for its therapeutic window for its safe use in patients.

The chaperone whose ATPase activity is most potently inhibited by AR12 is the HSP70 family member GRP78. GRP78 plays two key roles in cell biology; it acts both as a chaperone which uses ATP to renature proteins and it acts as a sensor which assesses the levels of denatured protein in a cell. Under normal resting conditions GRP78 inhibits ER stress signaling, e.g., by blocking PERK activity and the phosphorylation of eIF2α S51. When high levels of denatured protein exist in a cell, GRP78 dissociates from PERK and associates with denatured proteins. PERK activates and phosphorylates eIF2α which prevents translation from 95% or genes and enhances translation of genes whose gene products are components of the autophagic machinery. As a more generic inhibitor of all ATP binding chaperones, AR12 also reduces mTORC1 activity, promoting autophagosome formation, flux, and the degradation of denatured proteins. As we have shown, this approach can reduce the intracellular expression of Tau and APP.

GRP78 is not only located in the cytoplasm and ER; it is also a cell surface protein, localized on the outer leaflet of the plasma membrane. For example, inhibition of plasma membrane localized GRP78 causes inactivation of the PI3K/AKT/mTOR pathway in cancer cells [59]. Plasma membrane GRP78 acts to stabilize membrane receptors, for example, acetyl choline esterase 2 (ACE2), the receptor for the SARS-CoV-2 virus [40, 60]. Our data, hence, also suggest the possibility that plasma membrane localized GRP78 may be capable of sensing the presence of denatured amyloid-β in the extracellular space liminal to the plasma membrane. Extracellular GRP78 has been shown to promote amyloid-β uptake by microglia however it is known that microglia in AD are less capable of phagocytosing denatured amyloid-β [61–65]. It has also been shown that amyloid-β induces cells to over-express GRP78 and GRP78 is over-expressed in neurons from APP/PS1 mice. The exogenous GRP78, once ingested with the amyloid-β protein, translocated to the ER of the microglia where, presumably, it would act to block ER stress signaling and the autophagic digestion of denatured amyloid-β protein. Thus, our findings raise the intriguing possibility that AR12, by blocking GRP78 function in both populations of the protein, will facilitate amyloid-β degradation in microglia as well as in neurons.

ATG16L1 is a key protein required for autophagosome formation. African-Americans trend to express the ATG16L1 T300 isoform compared to Europeans who express the ATG16L1 A300 isoform and this is causal in European Americans presenting with higher levels of Crohn’s Disease than African Americans. African Americans appear to be at greater risk of developing and presenting with a more severe form of Alzheimer’s Disease. Our initial findings were counter-intuitive to this information as we discovered that the T300 isoform is permissive for a greater ability of cells to form autophagosomes and a greater ability of immune cells to take up extracellular material, causing its degradation, thus reducing inflammation. Hence this information would argue that expression of ATG16L1 T300 would predict for microglia who will more readily ingest extracellular denatured material, e.g., amyloid-β, which would predict for less inflammation in such an Alzheimer’s brain. However, our studies also demonstrated that ATG16L1 T300 cells expressed significantly greater levels of plasma membrane associated GRP78 than did isogenic ATG16L1 A300 cells. This implies that GRP78 bound to amyloid-β will be ingested and the intracellular levels of GRP78 and denatured amyloid-β in a cell expressing ATG16L1 T300 will be relatively greater than in a cell expressing ATG16L1 A300. The levels of intracellular GRP78 will now be elevated to such an extent that this prevents ER stress signaling and autophagic degradation of the denatured amyloid-β. Furthermore, in contrast to every other protein we have previously expressed from a constitutive plasmid promoter in isogenic T300/T300 and A300/A300 cells, the protein expression of “wild type” Tau or “mutant” Tau 301L is 25% greater in T300/T300 cells compared to A300/A300 cells. These findings would a priori predict that persons homozygote for T300/T300 would present with a more severe form of AD.

The drugs tested in this manuscript have been used preclinically and clinically in several anti-cancer studies. Our present studies were performed in non-neuronal cells and as a caveat, it is possible that our data in HCT116 and Vero cells will not be reflective of the same processes in neuronal cells. As the mechanism of drug-action became clearer it was apparent that these agents should also be tested in neurodegenerative diseases. The entire neurodegenerative field needs rapid translational methods that target the underlying cause of disease, toxic misfolded protein. The findings from this work warrant further testing with a focus on clinical utility.