Engineered scPEX19 variants suppress the toxicity of mHttex1 in yeast
To isolate an scPEX19 mutant gene that suppresses the cellular toxicity of mHttex1 protein, we used the yeast toxicity-based screening method33 (Fig. 1a). Deletion of PRD in Httex1-97Q enhances its polyQ-induced toxicity in yeast19. This mutant is more optimal for screening since it results in a larger difference in cell viability compared to expressing non-toxic Httex1-25Q. To this end, we generated yeast strains carrying chromosomally integrated Httex1 genes (Httex1-25QΔP and Httex1-97QΔP), which encode an N-terminal FLAG tag, the first 17 amino acids of Httex1 (N17 domain), 25 or 97 repeats of glutamine, and a C-terminal GFP gene under the control of the galactose-inducible promoter (Fig. 1a). Expression of the wild-type scPEX19 did not alter the cellular toxicity of Httex1-97QΔP when compared with the empty vector control (Extended Data Fig. 1a). We randomly mutated the entire scPEX19 gene and screened the scPEX19 plasmid library against Httex1-97QΔP toxicity. Among approximately 90,000 transformants, 21 colonies were able to grow on galactose plates. After assessing the cell viability of those colonies, we found that two scPEX19 variants, m1 and m2, effectively suppressed the cellular toxicity of Httex1-97QΔP in yeast (Fig. 1b).
The isolated scPEX19 variants share two common mutation sites, L288F and E292V (Fig. 1a). Therefore, we hypothesized that the mutation of these two sites accounts for the ability of scPEX19 variants to rescue Httex1-97QΔP-induced toxicity in yeast. To test this hypothesis, we generated a double mutant scPEX19-L288F/E292V. The results of the spotting assay showed that scPEX19-L288F/E292V is sufficient to suppress the cellular toxicity of Httex1-97QΔP (Fig. 1b). In contrast, coexpression of the single mutants of scPEX19-L288F or scPEX19-E292V with Httex1-97QΔP did not restore cell viability (Extended Data Fig. 1b), suggesting that scPEX19-L288F/E292V is a minimally mutated suppressor of polyQ-induced toxicity in yeast. In addition, we substituted E292 with other hydrophobic amino acids on the scPEX19 variant. We found that only scPEX19-L288F/E292I suppressed Httex1-97QΔP-induced toxicity to the same degree as scPEX19-L288F/E292V (Fig. 1c), possibly due to the structural similarity between the valine and isoleucine side chains. Therefore, we identified two scPEX19 variants, scPEX19-L288F/E292V (scPEX19-FV) and scPEX19-L288F/E292I (scPEX19-FI), that potently suppress polyQ toxicity in yeast.
Consistent with the results obtained with the spotting assay, microscopy and Western blot analyses showed that scPEX19-FV and scPEX19-FI significantly reduced the aggregation of Httex1-97QΔP proteins compared to scPEX19-WT (Fig. 1d-f). Over 50% of Httex1-97QΔP was found in SDS-insoluble aggregates in scPEX19-WT expressing cells (Fig. 1e, f). In contrast, overexpression of scPEX19-FV and scPEX19-FI drastically reduced the relative amount of SDS-insoluble 97Q aggregate and simultaneously increased SDS-soluble 97Q protein levels (Fig. 1e, f, and Extended Data Fig. 1c-d). This enhancement of Httex1-97QΔP solubility by scPEX19-FV and scPEX19-FI is not due to different expression levels of PEX19 in the cells (Extended Data Fig. 1c, e).
During protein targeting to peroxisome membranes, farnesylation of the C-terminal cysteine residue in the PEX19-CaaX motif increases the binding affinity of PMPs, effectively preventing PMP aggregation in the cytosol34. In addition, the PMP-bound PEX19 is recruited to the peroxisomal membrane by PEX3, which interacts with an N-terminal αa helix in PEX1935–37. Noticeably, the levels of farnesylated PEX19 were significantly reduced in scPEX19-FV and scPEX19-FI overexpressing cells (Extended Data Fig. 1c, f). Therefore, we tested whether two major aspects of PEX19-mediated PMP targeting, farnesylation and interaction with PEX334–36,38, are crucial for ameliorating polyQ-induced cellular toxicity in yeast. To this end, we introduced two further mutations, a farnesylation-defective mutation scPEX19-C339S and a PEX3 binding-defective mutation scPEX19-ΔN, into scPEX19-WT and the toxicity-reducing variants (Extended Data Fig. 2a). The results of a spotting assay showed that coexpression of scPEX19-FV/C339S, scPEX19-FV/ΔN, scPEX19-FI/C339S, and scPEX19-FI/ΔN with Httex1-97QΔP did not alter cell growth compared to scPEX19-FV and scPEX19-FI (Extended Data Fig. 2b, c). Therefore, both farnesylation and recruitment of PEX19 to the peroxisomal membrane by binding to PEX3 are dispensable for scPEX19-FV and scPEX19-FI to suppress the cellular toxicity of Httex1-97QΔP in yeast.
hs PEX19 variants suppress mHttex1 aggregation
Both mutated residues (L288F/E292V and L288F/E292I) are located in the α4 helix of PEX19 protein (Fig. 2a, green highlighted box). Sequence alignment analysis showed that these residues are highly conserved from Human (M255/Q259) to Arabidopsis (M202/Q206) (Fig. 2a). In addition, the M255 residue of hsPEX19 directly interacts with the farnesyl group in its C-terminal end34 (Fig. 2b), suggesting that this residue could be important for substrate recognition. Due to their highly homologous sequences, we hypothesized that introducing identical mutations (M255F/Q259V or M255F/Q259I) into hsPEX19 could also enhance suppression of mHttex1 aggregation.
To test whether both purified scPEX19 and hsPEX19 variants directly prevent Httex1-51Q aggregation in vitro, we used the well-established filter trap assay that detects heat-stable, SDS-insoluble aggregates16,39. In this assay, the N-terminal Httex1-51Q can be exposed by cleaving off a GST-tag using TEV protease, thus initiating polyQ aggregation (Extended Data Fig. 3a). In the absence of a chaperone, Httex1-51Q readily formed SDS-insoluble aggregates at 3 h (Fig. 2c, d). In contrast, scPEX19-FV and scPEX19-FI effectively suppressed aggregation of the purified Httex1-51Q protein, while scPEX19-WT was insufficient to prevent Httex1-51Q aggregation (Fig. 2c). Similar to scPEX19 variants, hsPEX19 variants effectively prevented aggregation of Httex1-51Q protein in vitro (Fig. 2d). This enhanced chaperone activity of hsPEX19 variants was not due to different TEV cleavage efficiency caused by their mutations (Extended Data Fig. 3b). In addition, our negatively stained transmission electron micrograph (TEM) analysis showed that hsPEX19-WT was not sufficient to prevent Httex1-51Q fibril formation (Fig. 2e). In contrast, hsPEX19-FV completely suppressed the formation of Httex1-51Q fibrils at 15 h (Fig. 2e). Consistent with the results of the filter trap assay (Extended Data Fig. 3c), hsPEX19-FI also prevented fibril formation by Httex1-51Q (Fig. 2e, lower), although in some cases, TEM analysis of hsPEX19-FI revealed both larger Httex1-51Q aggregates and small fibril fragments (Fig. 2e, upper). Indeed, hsPEX19 variants were unable to redissolve preformed Httex1-51Q aggregates when added at 3 h, suggesting that they do not have a disaggregase activity (Extended Data Fig. 3d). Therefore, we conclude that hsPEX19 variants function as a holdase that prevents the initial aggregation process of Httex1-51Q.
To test whether hsPEX19 variants are also effective in reducing mHttex1 aggregation in a mammalian HD model, we coexpressed hsPEX19 variants with Httex1-19Q-GFP or Httex1-134Q-GFP in HEK293T cells40. Overexpression of hsPEX19-FV and hsPEX19-FI at ~ 3-fold over endogenous PEX19 levels strongly prevented the aggregation of Httex1-134Q, as demonstrated by both fluorescence microscopy analysis and the filter trap assay (Fig. 2f-h and Extended Data Fig. 4a, c). In contrast, overexpression of hsPEX19-WT reduced the Httex1-134Q aggregates by ~ 50% on average, suggesting that hsPEX19-WT itself exhibits a mild chaperone activity toward polyQ proteins as supported by the in vitro aggregation assay (Fig. 2g, h). The difference in rescuing effects observed in hsPEX19 variants relative to their wild-type protein was not due to different expression levels of exogenous PEX19 or Httex1-134Q (Extended Data Fig. 4a-c). Critically, overexpression of hsPEX19 variants did not perturb the peroxisomal localization of the peroxisomal membrane protein PMP70, suggesting that this approach is unlikely to interfere with peroxisome biogenesis (Extended Data Fig. 4d, e). Therefore, these data demonstrate that the substitution of two conserved residues on the α4 helix of hsPEX19 significantly increases its chaperone activity toward mHttex1.
PEX19 variants bind the N17 domain of mHttex1
The N17 domain of Httex1 has an amphipathic helical property, which contributes to the initiation and acceleration of mHttex1 aggregation16,41 (Fig. 3a). Furthermore, a recent study suggested that structural coupling between the N17 and polyQ repeat domains stabilizes the helical content of Httex1 and accelerates aggregation42. Deletion of the N17 domain of Httex1-51Q (Httex1-51Q-ΔN) delays the kinetics of Httex1-51Q aggregation16. Given that hsPEX19 variants generate a more hydrophobic environment at their C-terminal domain (CTD) than hsPEX19-WT, we hypothesized that they bind to the hydrophobic amino acids in mHttex1, possibly at the N17 domain of mHttex1. Thus, we tested whether hsPEX19 variants also suppress Httex1-51Q-ΔN aggregation in vitro. In contrast to Httex1-51Q-WT, hsPEX19 variants were unable to suppress the aggregation of Httex1-51Q-ΔN (Fig. 2d vs 3b). Furthermore, hsPEX19 variants did not suppress the aggregation of another polyQ repeat protein, Ataxin3 (Extended Data Fig. 5a). Since Ataxin3-78Q has only the polyQ-repeat domain in common with Httex143,44, it is plausible that hsPEX19 variants do not target this polyQ-repeat domain (Extended Data Fig. 5a). Taken together, these results suggest that the N17 domain could be the primary recognition site of hsPEX19 variants within the Httex1-51Q protein.
To check whether the mutated hydrophobic residues at the hsPEX19 variants directly interact with Httex1-51Q, we used the Bpa crosslinking assay that uses a photocrosslinker, p-benzoyl-l-phenylalanine (Bpa) (Fig. 3c). We site-specifically incorporated Bpa into F255 at the hsPEX19 variants using amber suppression45. Suppression of Httex1-51Q aggregation by hsPEX19-FVBpa and hsPEX19-FIBpa was also observed at 3 and 6 h, albeit to a lesser extent than the unincorporated hsPEX19 variants (Extended Data Fig. 5b vs Fig. 2d). At 3 h incubation, hsPEX19-FVxHttex1-51Q or hsPEX19-FIxHttex1-51Q crosslink at ~ 70 kDa was readily detectable, whereas there was no observed crosslinked band in the presence of Httex1-51Q-ΔN (Fig. 3d, e). Therefore, these results indicate that the F255 residue in the hsPEX19 variants specifically binds to the N17 domain of Httex1-51Q.
We further tested whether the hydrophobic amino acid residues in the N-terminal amphipathic helix of mHttex1 also bind the hsPEX19 variants (Fig. 3a). To minimize structural perturbation, we incorporated Bpa at the F11 residue on Httex1-51Q among seven hydrophobic amino acid residues (Fig, 3a). Bpa incorporation on Httex1-51Q did not alter the aggregation kinetics (Extended Data Fig. 5c vs Fig. 2d). Similar to Httex1-51Q-WT, both hsPEX19-FV and hsPEX19-FI suppressed the aggregation of Httex1-51Q-F11Bpa more efficiently than hsPEX19-WT (Extended Data Fig. 5c). In the presence of hsPEX19 variants, two distinct hsPEX19-Httex1-51Q crosslink bands at ~ 70 and ~ 80 kDa were observed (Fig. 3f, g), suggesting that Httex1-51Q -F11Bpa binds hsPEX19 variants, possibly with two different conformations. In contrast, hsPEX19-WT resulted in a distinct crosslinked band at ~ 80 kDa and a weak diffuse band at ~ 70 kDa (Fig. 3f, g). Consistent with its mild chaperone activity, hsPEX19-WT also binds to Httex1-51Q-F11Bpa, but likely with one dominant conformation (Fig. 3f, g). These observed differences in the aggregation and Bpa crosslinking assays are not due to different TEV cleavage efficiency (Extended Data Fig. 5d). Therefore, our data demonstrate that the F11 hydrophobic residue on Httex1-51Q directly interacts with hsPEX19 and consistent with the results in Fig. 2d, its variants increase this interaction.
The α4 helix of hsPEX19 variants serves as a specific binding site for the N17 domain of mHttex1
PEX19 binds to the moderately hydrophobic transmembrane domains (TMDs) of peroxisomal and mitochondrial membrane proteins30,37,46,47 (Fig. 4a). In addition, PEX19 interacts with TMDs located in diverse topologies of membrane proteins, multi-spanning PMPs, tail-anchored membrane proteins (TAs), and N-terminal signal-anchored membrane proteins (Fig. 4a). Since hsPEX19 binds to these moderately hydrophobic TMDs, we hypothesized that hsPEX19 variants might also interact with the isolated N17 domain of Httex1 (Fig. 4a). To test this, we fused the N17 domain of Httex1 to the N-terminus of the Maltose binding protein (MBP) (Fig. 4b). The hsPEX19-FVBpa and hsPEX19-FIBpa proteins readily crosslinked to the N17-MBP protein, whereas no crosslinked band appeared in the presence of wild-type MBP protein (Fig. 4c). These results suggest that the N17 domain of Httex1 is a minimum recognition motif for hsPEX19 variants that allows suppression of the mHttex1 aggregation. In contrast to hsPEX19 variants, SGTA, a cytosolic co-chaperone that binds highly hydrophobic TMDs of ER TAs48,49, was unable to suppress the aggregation of Httex1-51Q (Fig. 4d). Collectively, our results suggest that mutations on the α4 helix of hsPEX19 enable binding to the relatively low hydrophobic N17 domain of mHttex1.
Several studies suggested that the α1 helix of PEX19-CTD serves as a binding site of PMPs32,34,50. Given that the Httex1-51Q binds to the α4 helix of hsPEX19 variants, we checked whether the hsPEX19 variants also interact with a bona fide PEX19 substrate, the peroxisomal TA, PEX2631 (Fig. 4e, f). At an approximately 3-fold excess concentration of the endogenous hsPEX1951, the amounts of PEX26 loaded onto hsPEX19 variants were comparable to hsPEX19-WT, indicating that these mutations on the α4 helix of hsPEX19 do not largely alter the overall binding capacity of PEX26 (Fig. 4e, f). In contrast to Httex1-51Q, both hsPEX19-FVBpa and hsPEX19-FIBpa did not crosslink to PEX26 (Fig. 4g-i). These results suggest that F255 and V259/I259 mutations on hsPEX19 could create a specific binding site for the N17 domain of Httex1-51Q, eventually resulting in robust suppression activity of Httex1-51Q aggregation.
We tested whether hsPEX19-FV also prevents aggregation of a non-polyQ protein, TDP43, which is associated with another neurodegenerative disease, ALS. To this end, we performed an established in vitro aggregation assay using the purified TDP43-TEV-MBP-His6 protein52. The addition of TEV protease enables initiation of TDP43 aggregation (Extended Data Fig. 6a, black). In contrast to Httex1-51Q aggregation in Fig. 2d, incubation with hsPEX19-WT or hsPEX19-FV exhibited only a minor delay in TDP43 aggregation kinetics (Extended Data Fig. 6a, blue and red). To further monitor TDP43 aggregation in cells, we generated a stable HEK293 cell line (TDP43-BiFC) that expresses both TDP43-VN and TDP43-VC. Given that phosphorylation and acetylation on TDP43 promote its aggregation53–56, we used Forskolin as a phosphorylation activator and Apicidin as an acetylation-inducing agent for TDP4357,58. Treatment with either Forskolin or Apicidin significantly increased the fluorescence intensities of TDP43-BiFC in the cytosol (Extended Data Fig. 6b, c). Overexpression of hsPEX19-WT or hsPEX19-FV showed at most a minor rescue of Forskolin or Apicidin-induced cytosolic TDP43 aggregation in HEK293 cells (Extended Data Fig. 6d-g). Together with Fig. 2, we conclude that hsPEX19-FV selectively suppresses the aggregation of mHttex1 in vitro and in mammalian cells.
hs PEX19-FV rescues HD-associated phenotypes
To test whether hsPEX19-FV protects striatal neurons from mHttex1 proteotoxicity, we coexpressed Httex1-134Q-GFP with hsPEX19-WT or hsPEX19-FV at 7 days in vitro (DIV) in primary striatal neurons (Fig. 5a). In contrast to Httex1-19Q-GFP- and vector control- coexpressing striatal neurons, at 48 h post-transfection, we observed largely fragmented neurites in the striatal neurons when coexpressed with Httex1-134Q-GFP and vector control, suggesting that mHttex1 induces neuritic degeneration59,60 (Fig. 5a). Striatal neurons coexpressing Httex1-134Q-GFP and hsPEX19-FV exhibited unfragmented healthy neurites, while partially fragmented neurites were observed in the Httex1-134Q-GFP-and hsPEX19-WT-coexpressing neurons (Fig. 5a). These results suggest that hsPEX19-FV effectively protects neuritic degeneration in mHttex1-expressing mouse striatal neurons.
We next tested whether the hsPEX19-FV variant could rescue HD-associated phenotypes in Drosophila HD models. To this end, we generated transgenic fly lines expressing pACU2 empty vector (vector control), hsPEX19-WT, or hsPEX19-FV and coexpressed Httex1-20Q or Httex1-93Q under the control of Elav-GAL4 (pan-neurons) or D42-GAL4 (motor neurons) drivers (Supplementary Table 1). As a negative control, we used the W1118 fly line which does not carry a Httex1 transgene. Compared to W1118/vector control and Httex1-20Q/vector control flies, motor- or pan-neuronal Httex1-93Q overexpression in Httex1-93Q/vector control flies led to a significant defect in their locomotion capacities (Fig. 5b, c). In contrast to hsPEX19-WT, hsPEX19-FV expression partially restored the impaired climbing ability of flies overexpressing Httex1-93Q in motor- and pan-neurons. Consistent with these results, the numbers of Httex1-93Q-positive puncta in Httex1-93Q/hsPEX19-FV flies were significantly reduced in both motor- and pan-neurons compared to Httex1-93Q/vector control flies (Extended Data Fig. 7a-f). Furthermore, despite the exclusive cytosolic localization of hsPEX19-FV, Httex1-93Q/hsPEX19-FV flies displayed nuclear-localized soluble Httex1-93Q in both motor- and pan-neurons (Extended Data Fig. 7g, h). Overexpression of hsPEX19-FV significantly increased the lifespan of flies expressing Httex1-93Q, whereas it did not affect the W1118 and Httex1-20Q flies (Fig. 5d-f). Taken together, hsPEX19-FV provides effective neuroprotection in both mouse striatal neurons and Httex1-93Q-expressing flies.