Pharmacoproteomics pinpoints HSP70 interaction for correction of the most frequent Wilson disease-causing mutant of ATP7B

Significance

Wilson disease is a disorder of copper homeostasis caused by mutations in ATP7B. The most frequent mutation of ATP7B results in an H1069Q substitution that affects the localization and stability of the protein product. By interrogating the interactome of ATP7B-H1069Q, we found that this mutant shows stronger interaction with HSP70, which drives mutant degradation. Using an HSP70 inhibitor for structural similarity searches, we identified a Food and Drug Administration-approved drug that increases ATP7B-H1069Q stability in cells and thus improves ATP7B function. This pharmacoproteomic strategy provides an effective shortcut from understanding cellular mechanisms operating in Wilson disease to rapid identification of safe pharmacological tools and therefore might be expanded for drug repurposing to counteract other genetic disorders.

Abstract

Pathogenic mutations in the copper transporter ATP7B have been hypothesized to affect its protein interaction landscape contributing to loss of function and, thereby, to hepatic copper toxicosis in Wilson disease. Although targeting mutant interactomes was proposed as a therapeutic strategy, druggable interactors for rescue of ATP7B mutants remain elusive. Using proteomics, we found that the frequent H1069Q substitution promotes ATP7B interaction with HSP70, thus accelerating endoplasmic reticulum (ER) degradation of the mutant protein and consequent copper accumulation in hepatic cells. This prompted us to use an HSP70 inhibitor as bait in a bioinformatics search for structurally similar Food and Drug Administration-approved drugs. Among the hits, domperidone emerged as an effective corrector that recovered trafficking and function of ATP7B-H1069Q by impairing its exposure to the HSP70 proteostatic network. Our findings suggest that HSP70-mediated degradation can be safely targeted with domperidone to rescue ER-retained ATP7B mutants and, hence, to counter the onset of Wilson disease.

A large number of inherited disorders are caused by gene mutations that affect folding of the corresponding protein product and, as a consequence, lead to aberrant localization of the mutant and/or to its rapid degradation by quality control machineries (1–3). Large multispan membrane proteins represent a particular challenge for quality control systems as their luminal, transmembrane, and cytosolic domains require the orchestrated effort of several proteostatic mechanisms. Therefore, these types of protein (that include numerous transporters and receptors) are particularly susceptible to mutations that destabilize their native architecture and cause retention and degradation in the endoplasmic reticulum (ER) (2, 4).

The most frequent mutations that cause Wilson disease result in ER-retained/degraded variants of the copper-transporting ATPase ATP7B (5, 6). ATP7B is expressed mainly in the liver where it supplies copper to secreted proteins in the trans-Golgi network (TGN) or traffics toward the canalicular surface of hepatocytes to transport excess copper into the bile (7). Although ER-retained mutants of ATP7B are capable of transporting copper ions across membranes, they undergo rapid degradation and fail to reach copper excretion sites (5, 6, 8). As a result, copper accumulates inside hepatocytes and causes extensive toxicity culminating in cell death, hepatitis, cirrhosis, and acute liver failure (9). ER-retained variants of ATP7B comprise H1069Q (∼50% in European and North American patients), R778L (∼40% in Eastern Asian patients) and a number of less frequent isoforms, altogether constituting the largest cohort among Wilson disease-causing ATP7B mutants (10). In this context, pharmacological correction of the trafficking defect has emerged as an attractive strategy to rescue ATP7B mutant phenotypes in Wilson disease (5, 8, 11, 12).

Since the interaction landscape of mutant proteins may differ from that of the wild-type (WT) protein, mutant-specific interactions represent potential targets for mutant correction (13–15). For example, the interactome of cystic fibrosis transmembrane conductance regulator (CFTR) with a F508del mutation, the most frequent among cystic fibrosis (CF) patients, contains quality control proteins that can be potentially targeted to treat CF (13, 15). By contrast, the impact of mutations on the interactome of ATP7B remains poorly understood. Only a few interactors (COMMD1, clusterin, and CRYAB) have been documented to be involved in proteostatic regulation of ATP7B (11, 16, 17). However, the value of these interactors for ATP7B mutant correction remains questionable due to their poor druggability. Thus, the demand for interactors suitable for pharmacological correction of ATP7B mutants remains open.

Searching for such interactors, here we used proteomics to identify how the H1069Q mutation affects the ATP7B interactome. Among the mutant-specific interactors, HSP70 emerged as a potential target for correction because HSP70 suppression recovered the stability and trafficking of ATP7B-H1069Q. We subsequently used a specific HSP70 inhibitor to search for structurally similar Food and Drug Administration (FDA)-approved drugs. One of the screen hits, domperidone, affected the interaction of the ATP7B mutant with components of the HSP70-dependent degradation machinery, facilitated export of ATP7B-H1069Q from the ER, and reduced Cu toxicity in mutant-expressing cells. This indicates that systematic analysis of mutant protein proteostatic networks can identify promising candidates for the correction of ATP7B mutants and for drug repurposing in Wilson disease.

Results

The H1069Q Mutation Promotes ATP7B Interaction with HSP70.

Although the ATP7B-H1069Q and WT protein interactomes have been investigated with traditional proteomics (8), this proteomics approach does not show a difference in the binding of any given protein with either WT or mutant ATP7B (18). As a result, interactors that bind both ATP7B variants but exhibit preference for the mutant risk being ignored as false negatives. To circumvent these drawbacks, we employed a very sensitive stable isotope labeling by amino acids in cell culture (SILAC) proteomics approach (18) to detect the proteins that preferentially bind to the ATP7B-H1069Q mutant (Fig. 1A). HepG2 cells were grown in medium containing “light” (K0R0), “medium” (K4R6), or “heavy” (K8R10) amino acids. The “medium” and “heavy” cells were transduced with adenoviruses carrying complementary DNA (cDNA) of either ATP7B-WT-GFP or ATP7B-H1069Q-GFP, respectively, while empty GFP was expressed in light cells and used as a negative control in immunoprecipitation (IP) and subsequent proteomics procedures (Fig. 1A). Light, medium, and heavy cell lysates were pooled and immunoprecipitated (Fig. 1A) with a GFP-specific antibody tested for IP in control experiments (SI Appendix, Fig. S1), and the individual proteins were identified by mass spectrometry. Quantification of light peptides from cells expressing empty GFP allowed us to threshold all false-positive interactions. For the remaining putative interactors (Dataset S1), the ratio of heavy to medium peak intensities in the mass spectrum was calculated (Fig. 1B) to reveal proteins enriched in either ATP7B-H1069Q or ATP7B-WT interactomes (defined from hereon as “mutant-specific” or “WT-specific” interactors).

Fig. 1.

H1069Q mutation promotes ATP7B interaction with HSP70. (A) Scheme of the SILAC proteomics strategy. (B) The plot shows the outcome of the proteomics analysis with fold change (FC) of heavy to medium peak intensities for each putative interactor and corresponding significance (P value; t test). Red dots correspond to proteins with significant enrichment (>1.3 fold; P value >0.05) in the interactome of the H1069Q mutant, while green dots correspond to proteins enriched in the interactome of the WT protein. HSP70 (HSPA1A) is indicated by an arrow. (C) GO analysis revealed an enrichment in the number of quality control categories associated with protein misfolding (red bars) in the mutant interactome. (D and E) HepG2 cells expressing GFP-tagged ATP7B-WT or ATP7B-H1069Q were treated with 200 μM CuSO₄ for 2 h and processed for IP (D) or proximity ligation assay (PLA) (E). (D) Western blot shows higher levels of HSP70 in ATP7B-H1069Q pull-downs. (E) Efficiency of PLA was confirmed by the presence of the PLA signal (red spots) between ATP7B variants and the known ATP7B interactor p62 (positive control) and lack of PLA signal with TfR (negative control). ATP7B-H1069Q induced a stronger PLA signal with HSP70 than the WT protein. (F) The graph corresponds to the IP experiment in D and shows fold change in HSP70 bound to the H1069Q mutant compared to WT protein (n = 3 experiments; ***P < 0.001, t test). (G) Quantification of PLA experiment (E) shows a higher number of PLA dots in cells expressing the mutant protein (n = 30 cells; ***P < 0.001; t test). (Scale bar: 7 µm.).

To select meaningful candidates for mutant correction, we employed gene ontology (GO) analysis, which revealed strong enrichment of the mutant interactome in quality control proteins (Fig. 1C). In particular, we noted the “HSP70 protein binding” category that suggested preferential interaction of the mutant with a subset of HSP70-associated chaperones/proteins (Fig. 1C and SI Appendix, Fig. S2A). In this group, HSPA1A, also known as HSP70 or inducible HSP70, looked particularly interesting for several reasons. HSP70 had one of the top confidence scores (Materials and Methods), number of peptides, and sequence coverage among ATP7B interactors and exhibited an almost three-fold increase in intensities of peptides that bind to the mutant compared to WT protein (Fig. 1B, arrow, and SI Appendix, Fig. S2A). Furthermore, we overlapped the lists of HSP70-related interactors and of ER genes promoting degradation of ATP7B-H1069Q (12) and found that HSP70 was the only component that belongs to both lists (SI Appendix, Fig. S2B). This suggests that HSP70 might both bind to the H1069Q mutant and drive its degradation. Indeed, direct IP experiments revealed considerably higher amounts of HSP70 in ATP7B-H1069Q pull-downs compared to the WT protein (Fig. 1 D and F). Furthermore, higher affinity of HSP70 for the mutant was confirmed with a complementary proximity ligation assay (Fig. 1 E and G).

HSP70 Suppression Reduces ER Retention and Degradation of the ATP7B-H1069Q Mutant.

Next we investigated whether HSP70 suppression facilitates export of ATP7B-H1069Q from the ER and, hence, reduces its turnover. To this end, HepG2 cells were used as a Wilson disease-relevant cell type (19), while HeLa cells were employed because RNA interference (RNAi) works very efficiently in this cell type. RNAi-mediated ablation of HSP70 promoted transport of ATP7B-H1069Q to the Golgi in both HepG2 and HeLa cells (Fig. 2 A and B) and increased overall ATP7B-H1069Q levels (Fig. 2 C and D).

Fig. 2.

Silencing of HSP70 decreases ER retention and degradation of ATP7B-H1069Q. (A and B) GFP-tagged ATP7B-WT or ATP7B-H1069Q were expressed in either HeLa or HepG2 cells. Cells expressing ATP7B-H1069Q were incubated with either control (siControl) or HSP70-specific (siHSP70) siRNAs and fixed and stained for TGN46. (A) In both cell types, ATP7B-WT resides mainly in the perinuclear Golgi compartment that is positive for TGN46, while the ATP7B-H1069Q mutant exhibits an ER pattern that poorly overlaps with the Golgi marker. RNAi of HSP70 reduced the ER localization of ATP7B-H1069Q and promoted delivery of the mutant to the Golgi. (B) The graph shows the impact of HSP70 silencing on colocalization (Pearson’s coefficient) of ATP7B-H1069Q with TGN46 (n = 30 cells; *P < 0.05, ANOVA). (C) Western blot reveals a higher amount of ATP7B-H1069Q in HSP70-silenced cells. (D) Quantification showing a significant increase of the ATP7B-H1069Q signal in Western blots (n = 3 experiments; *P < 0.05, t test). (Scale bar: 7 μm.)

To test further the potential of HSP70 as a target for correction of the ATP7B-H1069Q mutant, we used a highly specific allosteric HSP70 inhibitor, HS-72 (Fig. 3A). HS-72 binds C306 in the ATPase domain of two almost identical isoforms of HSP70 (A and B) and reduces their affinity for ATP (20). Fig. 3B shows that vitally important members of the HSP70 family (HSC70; HSP70-2, -5 and -9) do not contain cysteine in the same position and hence are not sensitive to HS-72 (20). HS-72 caused substantial relocation of ATP7B-H1069Q to the Golgi (Fig. 3 C and D) and to LAMP1-positive compartments (Fig. 3 C and D) where ATP7B normally resides upon Cu overload (19, 21). In parallel, we tested whether HS-72 reduces ATP7B-H1069Q degradation. To this end, we investigated the kinetics of ATP7B-H1069Q degradation using cycloheximide, a protein synthesis inhibitor. HS-72 increased the amounts of the mutant ATP7B at both 30-min and 5-h intervals of cycloheximide (CHX) treatment (Fig. 3 F and G), indicating that the HSP70 inhibitor slows down mutant degradation.

Fig. 3.

A specific inhibitor of HSP70 promotes trafficking of ATP7B-H1069Q and reduces its degradation. (A) Chemical structure of the specific HSP70 inhibitor HS-72. (B) Sequence alignment of HSP70 proteins in the region of HS-72 binding. Red font indicates the key position where cysteine 306 in HSP70-1A/B defines the binding site for HS-72. (C) HepG2 cells expressing either WT or H1069Q variants of ATP7B were incubated with 1 μM HS-72 for 24 h and then fixed directly (Upper row) or after an additional 2 h treatment with 200 μM CuSO₄ (+Cu; Lower row). The cells were then labeled for either TGN46 or LAMP1 as indicated. Arrows show LAMP1-positive structures, which received ATP7B in Cu-treated cells (D and E). The graphs show the impact of HS-72 on colocalization (Pearson’s coefficient) of WT or H1069Q variants of ATP7B with either TGN46 in untreated cells (D) or LAMP1 in Cu-treated cells (E) (n = 30 cells; *P < 0.05, ANOVA). (F and G) HepG2 cells expressing ATP7B-H1069Q were treated with 1 μM HS-72 for 24 h or left untreated before incubation with 100 μM CHX for either 30 min or 5 h. (F) Western blot indicates that HS-72 slowed down the decay of the ATP7B mutant protein in CHX-treated cell. (G) The graph shows a slower reduction of ATP7B-H1069Q levels in HS-72–treated cells (n = 3 experiments; *P < 0.05, t test). (Scale bar: 7 μm.)

The HSP70 Inhibitor Facilitates Delivery of ATP7B-H1069Q to the Appropriate Cell-Surface Domain.

Considering that suppression of HSP70 with either RNAi or the chemical inhibitor had an almost identical impact on ATP7B-H1069Q stability/localization, we used HS-72 as a simpler tool to inhibit HSP70 in further experiments. First, we employed surface biotinylation (5, 19) in HeLa cells to test whether HSP70 suppression increases ATP7B-H1069Q amounts at the plasma membrane. Incubation with HS-72 resulted in a significant increase in the amount of the mutant protein in the cell-surface biotinylated fraction (Fig. 4 A and B), demonstrating that the inhibition of HSP70 improves delivery of the mutant protein to the plasma membrane.

Fig. 4.

The HSP70 inhibitor stimulates delivery of ATP7B-H1069Q to the cell surface. (A) HeLa cells expressing ATP7B-WT or ATP7B-H1069Q were treated with 10 μM HS-72 for 24 h and then with 200 μM CuSO₄ for 2 h. The cells were subsequently processed for cell-surface biotinylation. (B) Quantification of the ATP7B signal in the biotinylated fraction reveals an increase in the amount of ATB7B-H1069Q in HS-72–treated cells (n = 3 experiments; *P < 0.05, ANOVA). (C–E) Polarized HepG2 expressing either WT (C) or H1069Q (D and E) were exposed to CuSO₄ for 4 h and stained with the canalicular marker MRP2. HS-72 (10 µM) was added to the ATP7B-H1069Q–expressing cells before exposure to Cu (E). Arrows and arrowheads in all images indicate an MRP2-positive canalicular vacuole (cyst). (F) The graphs show intensities of ATP7B (green) and MRP2 (red) signals along the lines 1, 2, and 3 drawn through the canalicular vacuole in C, D, and E, respectively. (Scale bars, C–E: 4.8 μm.)

In hepatocytes, ATP7B has to be targeted to the canalicular (apical) surface to support excretion of Cu into the bile (19, 22). To test whether the HSP70 inhibitor facilitates delivery of ATP7B-H1069Q to the canalicular surface, HepG2 cells were grown to achieve maximal polarization that is manifested by the development of the canalicular vacuole, which is formed by apical membranes of two neighboring cells and recapitulates the biliary duct (23). Polarized HepG2 cells were exposed to Cu for activation of ATP7B transport and then stained for MRP2 to reveal the canalicular surface. Fig. 4C shows ATP7B-WT in the canalicular domain, while ATP7B-H1069Q failed to reach the canalicular vacuole (Fig. 4D). However, exposure of these cells to the HS-72 inhibitor resulted in an increase in the amounts of ATP7B-H1069Q in the canalicular region (Fig. 4 E and F). This suggests that HS-72 facilitates delivery of the mutant to the Cu excretion site at the apical surface of hepatic cells.

A Three-Dimensional Structure Similarity Search Reveals the FDA-Approved Drug Domperidone as a Corrector of Several ATP7B Mutants.

Despite the effectiveness of HS-72 in correcting the ATP7B-H1069Q, its translation to clinical use might be lengthy due to the significant time/costs of clinical studies. To circumvent this problem, we investigated whether HS-72 resembles any FDA-approved drug, which could be rapidly repurposed for clinical use in Wilson’s disease. We found three different FDA-approved drugs using HS-72 as a query structure for three-dimensional (3D) ligand-based screening (24) of a collection of 1,309 FDA-approved drugs (Fig. 5A). Among these hits, only domperidone increased ATP7B-H1069Q levels (Fig. 5B), promoted ATP7B-H1069Q export from the ER to the Golgi (Fig. 5C), and facilitated delivery of the mutant to the canalicular domain in polarized HepG2 cells (SI Appendix, Fig. S3).

Fig. 5.

The FDA-approved drug domperidone rescues localization and function of ATP7B-H1069Q. (A) HS-72 was used as a query structure for 3D ligand-based screening of a collection of FDA-approved drugs (Materials and Methods). Three hits were retrieved with a structural similarity score above 0.5. Overlapping of the query 3D structure poses (wireframe in gray) based on the superimposition of the hydrophobic (solid green; DRY probe) and polar interaction fields (solid cyan, OH2 probe) of the query and the three FDA-prioritized drugs (green mesh for hydrophobic and cyan mesh for polar interaction fields). (B and C) HepG2 cells expressing ATP7B-H1069Q were treated with domperidone (at the indicated concentrations) for 24 h and further processed for Western blot (B) or confocal microscopy (C). Percentage in C indicates colocalization (Pearson’s coefficient) of ATP7B-H1069Q with TGN46 (n = 30 cells; P < 0.001, t test). (D) Survival of WT, ATP7B-KO, ATP7B-H1069Q, and ATP7B-R778L HepG2 cell lines after treatment with 1 mM CuCl₂ alone or in combination with either 1 µM domperidone or 1 µM HS-72 was analyzed using MTT assay. The MTT signal in the treated cells was normalized to those in control cells (treated with dimethylsulfoxide only) and expressed as percentage of viability (n = 3 experiments; *P < 0.05, ANOVA). (Scale bars: 6 μm.)

We next investigated whether domperidone overcomes ER retention of other disease-causing mutants of ATP7B and of other membrane transporters such as CFTR and SLC30A10. Among ATP7B mutants with significant catalytic activity (5, 6, 25), domperidone stimulated ER export of the D765N, L776V, and, to some extent, R778L mutants (SI Appendix, Fig. S4 A and B) although a significant fraction of treated cells still contained ATP7B-R778L in the ER (SI Appendix, Fig. S4 A and B). Then, we tested the ability of domperidone to rescue the CF-causing F508del-CFTR mutant (4, 13, 15, 26, 27). A halide-sensitive yellow fluorescent protein assay (26) revealed that the established F508del-CFTR corrector VX-809 (27) induced a nearly three-fold increase in anion transport, while domperidone failed to rescue F508del-CFTR function (SI Appendix, Fig. S4C). Finally, we tested domperidone for correction of the ΔA105-107P mutant of the manganese transporter SLC30A10 that causes toxic manganese overload (28), but failed to detect any significant impact on ER retention of the SLC30A10 mutant (SI Appendix, Fig. S4 D and E).

Domperidone Rescues the Ability of ATP7B-H1069Q to Counteract Copper Toxicity.

The beneficial impact of domperidone on the stability and localization of ATP7B mutants is expected to reduce Cu toxicity. To test this, we used HepG2 lines lacking ATP7B (knockout [KO] cells) or expressing only mutant ATP7B variants (H1069Q cells and R778L cells). These ATP7B-deficient cells poorly tolerate Cu (29, 30), and their death upon Cu overload provides a reliable readout of the loss of ATP7B function (Fig. 5D). Domperidone and HS-72 failed to improve the tolerance of KO cells to Cu but increased the survival of H1069Q cells (Fig. 5D and SI Appendix, Fig. S5). This suggests that both drugs require the ATP7B mutant protein product to confer resistance to Cu and, hence, rescue function of ATP7B-H1069Q. Dose–response experiments revealed a 1-µM concentration of domperidone to be optimal for further use due to high efficacy and low toxicity (SI Appendix, Fig. S5). Notably, none of the tested concentrations significantly improved Cu tolerance in R778L cells (Fig. 5D and SI Appendix, Fig. S5), likely due to the failure of the drugs to support R778L delivery to the canalicular domain of the cells (SI Appendix, Fig. S3). This indicates the specificity of domperidone for the H1069Q mutant.

Domperidone Inhibits HSP70 Activity and Affects ATP7B-H1069Q Interaction with Components of the HSP70 Machinery.

To investigate the mechanisms that might underlie the ability of domperidone to rescue ATP7B-H1069Q, we first investigated whether domperidone affects the catalytic activity of HSP70. Fig. 6A shows that domperidone reduced the activity of HSP70 even more strongly than HS-72 at the same concentration. Then, we tested the ability of domperidone to impact the interaction of ATP7B-H1069Q with HSP70. IP experiments revealed a striking reduction of HSP70 binding to the mutant in domperidone-treated cells (Fig. 6B).

Fig. 6.

Impact of domperidone on the interaction of ATP7B-H1069Q with HSP70 and its cochaperones. (A) HSP70 ATPase activity assay reveals that HS-72 and domperidone (both at 1 μM concentration) significantly reduce the activity of HSP70 (n = 3 experiments; *P < 0.05, ANOVA). (B) ATP7B-H1069Q was immunoprecipitated from control and domperidone-treated HepG2-H1069Q cells and the pull downs (IPs) were immunoblotted for HSP70, DNAJA1 and BAG2. (C) Graph shows protein fold change in IP fractions (n = 3 experiments; **P < 0.01, t test). (D) DNAJA1 was silenced in HepG2-H1069Q cells with pooled siRNAs. Western blot revealed an increase in the ATP7B-H1069Q signal in DNAJA1-silenced cells. (E) Graph shows ATP7B-H1069Q fold change in DNAJA1-silenced cells (n = 3 experiments; **P < 0.01, t test). (F) HSP70 was immunoprecipitated from control and domperidone-treated HepG2-H1069Q cells and the pull-downs (IPs) were immunoblotted for DNAJA1. (G) A scheme of HSP70-mediated degradation [adapted from Young (4)] with ATP7B-H1069Q as a client protein with the following steps: 1—recognition of ATP7B-H1069Q by DNAJA1; 2—recruitment of HSP70 to the DNAJA1/ATP7B-H1069Q complex; 3—hydrolysis of ATP by HSP70; 4—release of DNAJA1 from the HSP70/ATP7B-H1069Q complex and BAG2-mediated ADP to ATP exchange on HSP70; 5—release of HSP70 and delivery of ATP7B-H1069Q for proteasomal degradation. Domperidone blocks step 2.

Furthermore, we hypothesized that, being similar to the allosteric HSP70 inhibitor, domperidone might alter the conformation of HSP70, thus affecting the efficacy of its cochaperones that regulate ATP7B-H1069Q degradation. To test this hypothesis, we selected from the list of putative ATP7B interactors (Dataset S1) all chaperones with a documented role in HSP70-mediated degradation (DNAJs A1, A2, B1; HSPB1; BAG2) (4) and validated their binding to ATP7B-H1069Q in direct IP experiments. An ability to bind the mutant was clearly observed for DNAJA1 and BAG2 (Fig. 6 B and C). DNAJA1 coordinates binding of misfolded client proteins to HSP70 (4, 31). Importantly, we also found that suppression of DNAJA1 reduced the degradation of ATP7B-H1069Q (Fig. 6 D and E). IP experiments revealed that domperidone significantly increased the amount of DNAJA1 bound to ATP7B-H1069Q (Fig. 6 B and C) while impairing the interaction of both ATP7B-H1069Q and DNAJA1 with HSP70 (Fig. 6 B, C, and F). These findings suggest that domperidone prevents HSP70 recruitment to the ATP7B-H1069Q/DNAJA1 complex and, hence, blocks further progression of the ATP7B mutant through the HSP70-dependent degradation pathway (4, 31) (see scheme in Fig. 6G). Indeed, domperidone reduced the interaction of the ATP7B mutant with BAG2 (Fig. 6 B and C), which acts downstream of DNAJA1 in the HSP70-mediated degradation cascade (4, 31) (Fig. 6G). Thus, domperidone affects the ability of the HSP70 proteostatic network to trap ATP7B-H1069Q in the ER, thereby increasing the chances of the mutant to avoid ER degradation.

Domperidone Attenuates Degradation of ATP7B-H1069Q in Hepatocyte-like Cells Derived from Patients with Wilson Disease.

A growing body of evidence indicates that validation of pharmacological correctors requires reliable cell/animal systems with endogenous expression of the disease-causing mutant (21, 32). Such an animal system is not available for ATP7B-H1069Q, but the generation of hepatocyte-like cells (HLCs) derived from homozygous patients with the H1069Q mutation was reported recently (21). These HLCs endogenously express ATP7B-H1069Q, which, however, undergoes rapid and extensive degradation in the ER. Only a small fraction (20%) of the mutant escapes from the ER and reaches the TGN (21). Thus, we tested the potential of both domperidone and HS-72 to protect the mutant from degradation in patient’s HLCs. Fig. 7A shows that the amount of the ATP7B protein product in patient HLCs was very low compared to HLCs from control subject expressing ATP7B-WT. Both domperidone and HS-72 treatment resulted in a significant increase in ATP7B-H1069Q levels (Fig. 7 A and B), indicating that these drugs attenuate mutant degradation in patient HLCs. Finally, we checked whether protection of ATP7B-H1069Q by domperidone (or by HS-72) supports mutant trafficking to the Golgi. In untreated cells, low amounts of the mutant might be detected in the Golgi and in the ER (Fig. 7C and SI Appendix, Fig. S6A). We found that domperidone and HS-72 improved colocalization of the mutant with a Golgi marker (SI Appendix, Fig. S6 B and C) and increased ATP7B-H1069Q levels in the Golgi compartment (Fig. 7C and SI Appendix, Fig. S6D), indicating that both drugs allow higher amounts of the mutant to be correctly transported in patient cells.

Fig. 7.

Domperidone and HS-72 rescue endogenous ATP7B-H1069Q in hepatic cells obtained from patients with Wilson disease. HLCs from patients homozygous for the H1069Q mutation in ATP7B were treated with 1 μM HS-72 or 1 μM domperidone and prepared for either Western blot (A and B) or immunofluorescence (C). HLCs from control subject expressing ATP7B-WT were used as control. (A) Western blot shows an increase in ATP7B-H1069Q levels in patient HLCs, which were treated with HS-72 or domperidone. (B) Quantification of Western blots reveals a significant increase in the ATP7B-H1069Q signal upon HS-72 and domperidone treatment (n = 3 experiments; *P < 0.05, ANOVA). (C) Control and patient HLCs were labeled for ATP7B and the TGN marker Golgin 97. Confocal images showing high levels of the ATP7B signal in the TGN of control HLCs and low ATP7B levels in the TGN of patient HLCs (arrows). Both HS-72 and domperidone increase ATP7B-H1069Q amounts in the TGN of patient HLCs. (Scale bar: 10 μm.)

Discussion

The interactomes of misfolded pathogenic mutants have been extensively explored to find targets for mutant protein rescue (13–15). Despite the large number of mutant-specific interactors (13–15), only a few of them were further advanced to preclinical studies with pharmacological antagonists or agonists. In part, this is because the remaining mutant interactome was either poorly druggable or lacked a specific inhibitor that could be used without the risk of significant side effects (4).

Here, we present a pharmacoproteomic strategy that identifies an FDA-approved corrector of an ATP7B mutant starting from a comparison of the H1069Q- and WT-ATP7B interactomes. To find a possible target for mutant correction, we prioritized the translational value of mutant-specific interactors. Apart from the strength of interaction, the druggability and safety was considered for each specific binding partner of ATP7B-H1069Q. Using these criteria, HSP70 emerged as an attractive and realistic candidate for mutant correction. HSP70, being an ATPase, can be easily suppressed with pharmacological tools (20). Furthermore, despite being strongly involved in proteostasis, HSP70 activity does not seem to be indispensable since mice with knockout of both HSP70 isoforms (A and B) are viable and fertile (33). This contrasts with other ATP7B interactors from the HSP70 family, BIP and HCS70, whose depletion results in lethal phenotypes (33).

However, targeting of HSP70 would be of limited use without a specific inhibitor. In recent years, a few inhibitors of HSP70 have been explored for F508del-CFTR correction, but their specificity was questionable as they also suppressed HSC70 (4). Recently, a specific inhibitor, HS-72, was found to target HSP70 on the C306 residue (20) that is not shared by other vitally essential HSP70 proteins. We demonstrated that HS-72 was effective in reducing degradation, rescuing trafficking, and correcting localization of ATP7B-H1069Q. Considering that the inhibitor is well tolerated in vivo (20), it might become a promising candidate for therapy.

Unfortunately, translation from preclinical studies to the patient bedside frequently takes an excessively long period for any new molecule (such as HS-72). In this context, we reasoned that already approved drugs with a chemical structure resembling HS-72 might correct ATP7B-H1069Q and thus provide a shortcut for transition into clinical use. Using in silico screening, we identified the FDA-approved drug domperidone as an efficient replacement of HS-72 for inhibition of HSP70 and for rescue of the ATP7B-H1069Q mutant. Domperidone allowed the mutant to avoid recognition by the HSP70 proteostatic network and promoted ATP7B-H1069Q export from the ER into the secretory pathway. As a result, tolerance of ATP7B-H1069Q–expressing cells to Cu was significantly improved by domperidone treatment. Domperidone is a dopamine-2 receptor antagonist with an excellent safety profile and is widely used against gastric pain and vomiting (34). Unfortunately, it was impossible to test this drug in vivo due to the lack of an appropriate animal model expressing the mutant. In this context, hepatic cells from ATP7B-H1069Q homozygous patients represent a unique reliable system since they endogenously express the ATP7B mutant and might be used for preclinical drug testing (21). Importantly, the effectiveness of domperidone was validated in this patient-derived cell system, indicating that the drug has the potential to be rapidly repurposed for clinical use.

Could domperidone be used to rescue other disease-causing mutants? We found that domperidone was unable to correct ER-locked variants of CFTR and SLC30A10 and exhibited a high degree of selectivity for different ATP7B mutants. It seems that such selectivity was not defined by location of the mutation in a particular domain of ATP7B, indicating that domperidone does not inhibit a domain-specific ER degradation (ERAD) pathway, such as ERAD-M or ERAD-C (35). For example, of the variants with mutations in the fourth transmembrane domain of ATP7B, D765N and L776V were recovered from the ER by domperidone treatment, while rescue of the R778L mutant was fairly poor. Similar differences in sensitivity to domperidone were noted between the H1069Q and L1083F mutants, which both carry a substitution in the nucleotide-binding domain of ATP7B. This suggests that recognition/degradation of different mutants relies on diverse and specific sets of proteostatic molecules and that domperidone targets only some of these sets.

Indeed, our data indicate that the sets of HSP70/HCS70 cochaperones operating in the degradation of ATP7B-H1069Q differ from those that have been reported to drive degradation of F508del-CFTR. We found that DNAJA1 recognizes ATP7B-H1069Q and directs the mutant into the HSP70-dependent degradation pathway. Indeed, domperidone inhibits the progression of the ATP7B-H1069Q/DNAJA1 complex through this pathway by blocking the interaction of both the ATP7B mutant and DNAJA1 with HSP70 (Fig. 6G). By contrast, F508del-CFTR degradation is promoted by different DNAJ proteins (B12 and C5) and requires HCS70 instead of the HSP70 (36, 37). Thus, domperidone is unlikely to impair the HCS70 proteostatic network driving degradation of F508del-CFTR.

On the other hand, such selectivity indicates that domperidone should not significantly affect overall protein quality control surveillance. Indeed, the drug did not impact trafficking and delivery of both secreted and membrane proteins in hepatic cells (SI Appendix, Fig. S7). Therefore, the eventual risks of using domperidone in patients are expected to be minimal and consistent with its excellent biosafety properties (34). Existing Wilson disease therapies (zinc salts and Cu chelators) may cause severe side effects and intolerance or even be ineffective in a substantial cohort of patients (9, 38). Therefore, domperidone offers an option for reduction of dosage of these drugs (or even their substitution) in patients expressing ATP7B-H1069Q or similar domperidone-sensitive ATP7B mutants. This does not mean that the correction approach would not work in patients with other missense mutations of ATP7B. A significant effort should be made to analyze the interactomes of these ATP7B mutants to detect specific binding partners, which in turn could be targeted for mutant rescue and further identification of bio-safe correctors.

Materials and Methods

Antibodies, cDNAs, Adenoviruses, and RNAi.

The full list of antibodies, cDNAs, and small interfering RNA (siRNAs) are provided in SI Appendix, Supplementary Materials and Methods. Adenoviruses carrying cDNAs of ATP7B-WT-GFP or ATP7B-H1069Q-GFP (8, 19) were used at different values of multiplicity of infection to transduce HepG2 or HeLa cells. For RNAi, HepG2 or HeLa cells were transfected with individual or pooled siRNAs using Dharmafect4 (Dharmacon) or Oligofectamine (Invitrogen) according to the manufacturer’s instructions. Silencing efficiency was evaluated using Western blot.

Cell Culture.

HeLa and HepG2 cells were grown in HepG2 cells were grown in Dulbecco’s Modified Eagle Media (DMEM). ATP7B-deficient HepG2 cells (KO, H1069Q, and R778L) were provided by Hartmut Schmidt (Medizinische Klinik B für Gastroenterologie und Hepatologie, Universitätsklinikum Münster, Münster, Germany) and grown in RPMI medium 1640. HLCs from patients carrying the H1069Q mutation and from control subjects were differentiated from corresponding induced pluripotent stem cell clones according to published protocols (21). The bronchial epithelial cell line CFBE41o- expressing F508del-CFTR and halide-sensitive yellow fluorescent protein (HS-YFP) (26) was kept in Minimum Essential Media (MEM). All cell lines were supplemented with 10% fetal calf serum, 2 mM L-glutamine, penicillin, and streptomycin.

SILAC Proteomics.

HepG2 cells were grown in DMEM containing medium R6K4, heavy R10K8, or light R0K0 isotopes (Dundee Cell Products). Medium, heavy, and light cells were then transduced with ATP7B-WT-GFP, ATP7B-H1069Q-GFP, or GFP (control) adenoviruses, respectively, and incubated with lysis buffer. The GFP signal was analyzed with Western blot in lysates from HepG2 cells expressing ATP7B-H1069Q-GFP, ATP7B-WT-GFP, or GFP alone. The samples were then pooled in a 1:1:1 ratio according to the GFP quantification, immunoprecipitated with anti-GFP antibody, and processed with sodium dodecyl sulfate/polyacrylamide gel electrophoresis. The lanes were excised from the gel and sent to Oxford University Proteomics Facility, where they were subsequently digested and subjected to nanoscale liquid chromatography coupled to tandem mass spectrometry (nano LC-MS/MS; see SI Appendix, Supplementary Materials and Methods, for details). The nano LC-MS/MS data were acquired in the Orbitrap using Xcalibar v2.1 software (Thermo Fisher Scientific). The raw data files were processed and quantified using Proteome Discoverer software v1.2 (Thermo Fisher Scientific). All data were filtered to satisfy a false discovery rate of <5%. For each putative interactor the score was calculated by Proteome Discoverer (using the MudPIT scoring method) as a measure of confidence in the identification of the protein. Several parameters such as score, coverage, and number of peptides were considered for interactor ranking. Most importantly, the ratios between heavy/light, medium/light, and medium/heavy peak intensities in the mass spectrum were calculated for each putative interactor to avoid false-positive hits and to reveal enrichments in either ATP7B-WT or ATP7B-H1069Q interactomes.

Drug Treatments.

HS-72, a specific chemical inhibitor of HSP70 (HSPA1A/B), was provided by Timothy Haystead (Duke University School of Medicine, Durham, NC) (20). To investigate the impact of HS-72 and the structurally similar FDA-approved drug domperidone (see chemical similarity search in SI Appendix, Supplementary Materials and Methods) on stability and localization of ATP7B-H1069Q, different cell types expressing the mutant were treated with different concentrations of the drugs (ranging from 0.5 to 100 µM) for 24 h and then prepared for immunofluorescence or Western blot. Minimal effective concentrations of the drugs were than tested for their ability 1) to rescue copper-dependent trafficking of ATP7B-H1069Q, 2) to reduce Cu toxicity in ATP7B-H1069Q–expressing cells, and 3) to correct other ER retained mutants of ATP7B, CFTR, or SLC30A10. To analyze the dynamics of ATP7B-H1069Q stabilization, HS-72–treated and control cells were incubated with 100 μg/mL cycloheximide for different time intervals and then were processed for Western blot.

Other Methods.

Gene ontology enrichment analysis, chemical similarity search, Western blot, immunoprecipitation, cell-surface biotinylation, proximity ligation assay, immunofluorescence, confocal microscopy as well as assays for Cu-dependent trafficking of ATP7B, cell viability, and activities of HSP70 and CFTR are described in detail in SI Appendix, Supplementary Materials and Methods.

Statistical Analyses.

Data are expressed as mean ± SD, collected from multiple independent experiments performed on different days. Statistical significances for all data were computed using Student t tests or one-way ANOVA (for all figures, *P < 0.05, **P < 0.01, and ***P < 0.001 indicate statistical significance).

Data Availability.

The mass spectrometry proteomics data have been deposited in the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD016816 (39).