Loss of PIKfyve drives the spongiform degeneration in prion diseases

Abstract Brain‐matter vacuolation is a defining trait of all prion diseases, yet its cause is unknown. Here, we report that prion infection and prion‐mimetic antibodies deplete the phosphoinositide kinase PIKfyve—which controls endolysosomal maturation—from mouse brains, cultured cells, organotypic brain slices, and brains of Creutzfeldt‐Jakob disease victims. We found that PIKfyve is acylated by the acyltransferases zDHHC9 and zDHHC21, whose juxtavesicular topology is disturbed by prion infection, resulting in PIKfyve deacylation and rapid degradation, as well as endolysosomal hypertrophy and activation of TFEB‐dependent lysosomal enzymes. A protracted unfolded protein response (UPR), typical of prion diseases, also induced PIKfyve deacylation and degradation. Conversely, UPR antagonists restored PIKfyve levels in prion‐infected cells. Overexpression of zDHHC9 and zDHHC21, administration of the antiprion polythiophene LIN5044, or supplementation with the PIKfyve reaction product PI(3,5)P2 suppressed prion‐induced vacuolation and restored lysosomal homeostasis. Thus, PIKfyve emerges as a central mediator of vacuolation and neurotoxicity in prion diseases.

We then infected murine Gt1 cells 11 with RML prions. At 75 dpi we observed the appearance of cytoplasmic vacuoles accompanied by a substantial reduction of PIKfyve, but not of FIG4 or VAC14 (Fig. 1E-F). PrP depletion from prion-infected Gt1 cells by RNAi at 70 dpi largely suppressed cytoplasmic vacuolation (Fig.   S1D). Next, we administered RML prions to cerebellar organotypic cultured slices (COCS) from 9-day old tga20 mice 12 . At 45 dpi we observed neurodegeneration accompanied by conspicuous PIKfyve depletion ( Fig.   S1E-F). Finally, we exposed COCS for 72 hours to the antibody POM1 which induces PrP C -dependent neurotoxicity 13 similarly to prion infections 14 . Again, this treatment depleted PIKfyve from COCS (Fig. 1G).
We then asked if PIKfyve is lost in human sporadic Creutzfeldt-Jakob disease (sCJD). Differential proteolysis of the misfolded prion protein (PrP Sc ) discriminates two types of sCJD with distinct clinical courses and histopathological characteristics 15 . We investigated 9 and 8 frontal cortex samples ( Fig. S1G-H) from patients who succumbed from type-1 and type-2 sCJD, respectively, as well as 4 non-CJD controls, collected between 2002 and 2010 (Table S1). PIKfyve levels were reduced in all cases, but the reduction was more profound in type-2 sCJD samples ( Fig. 1H-I) which also showed more extensive vacuolation (Fig. S1H). Along with the earlier onset of disease in Type-2 CJD (average: 57 vs. 63 years) 16 , this suggests that PIKfyve depletion may be a determinant of toxicity. Therefore, PIKfyve depletion is a feature of all investigated instances of prion infections.
We then studied the consequences of prion infection onto the composition of the PIKfyve tripartite complex.
While VAC14 and FIG4 levels were not affected by prion infection, immunoprecipitation of VAC14 captured both FIG4 and PIKfyve from NBH-inoculated brains, but not from prion-infected brains (Fig. 1J). Conversely, the siRNA knockdown of FIG4 or VAC14 in Gt1 cells using RNAi caused a significant PIKfyve reduction. The knockdown of FIG4 reduced VAC14 levels but the VAC14 knockdown did not alter FIG4 (Fig. S1I). Hence the prion-induced PIKfyve depletion destabilizes the association between FIG4 and VAC14 without reducing their concentration.
Next, we investigated the mechanism of PIKfyve suppression by prions. Prion-infected mice (Fig. S1J), as well as prion-infected or POM1-treated COCS (Figs. S1K and S1L, respectively) showed no alterations in PIKfyve, FIG4 and VAC14 brain mRNA levels. The relative expression of PIKfyve mRNA splice variants was similar between RML-infected and control mouse brains ( Fig. S1M-N), and exploration of a splice alteration database during the progression of prion disease 17 did not identify any differential expression of PIKfyve isoforms. Hence PIKfyve suppression occurs post-transcriptionally. PIKfyve levels gradually decreased after 60 dpi whereas eIF2α became phosphorylated from 30 dpi until terminal disease. eIF2α phosphorylation and pPIKfyve depletion in POM1-treated tga20 COCS. Quantification in Figure S2E. E: Gt1 cells were treated with thapsigargin (0.5 µM, 4 hours) or DMSO for control, and metabolically pulsed with [ 35 S]Met/Cys for 20 min at 37 °C. Cells were either immediately lysed or chased without 35 S at 37 °C for ≤3 hours. Proteins were immunoprecipitated with antibodies to PIKfyve and subjected to SDS-PAGE and autoradiography. F: Autoradiographic signals expressed as percentage of the initial PIKfyve levels. Thapsigargin shortened the half-life of PIKfyve from 79 to 34 min. First-order kinetics was assumed. Dotted lines: standard deviation. G: Prion-infected tga20 mice were treated with the polythiophene LIN5044 (LCP) or vehicle. LCP partially restored PIKfyve levels (quantification: Figure S2E). H: tga20 COCS were inoculated with RML or NBH, optionally treated with GSK2606414 starting at 21 dpi, and lysed at 35 dpi. Depletion of PIKfyve in prion-infected samples was largely rescued by GSK2606414 (quantification: Fig. S2G) Prion diseases induce a chronic unfolded-protein response (UPR) in the endoplasmic reticulum (ER) leading to eIF2α phosphorylation and translational suppression 18 , and UPR inhibitors can mitigate the spongiosis 19 .
We first compared the evolution of the UPR and of PIKfyve depletion in various prion models. In prioninfected mouse brains, eIF2α phosphorylation was detectable at 30 dpi, whereas PIKfyve decreased gradually starting at 60 dpi ( Fig. 2A-B). In COCS, eIF2α became phosphorylated 8h after exposure to POM1, yet PIKfyve was depleted only after 72 hours (Fig. 2C, S2A). Hence a chronic UPR precedes downregulation of PIKfyve in vivo and ex vivo. We then induced an acute UPR in Gt1 cells with thapsigargin (0.5 µM) resulting in depletion of PIKfyve substantially but transiently ( If PIKfyve depletion mediates prion neurotoxicity, it may be attenuated by antiprion therapeutics. The conjugated polythiophene LIN5044 binds PrP Sc , antagonizes prion replication 20,21 , delays prion pathogenesis in vivo and reduces spongiosis 22 . Remarkably, prion-infected mice treated with LIN5044 not only showed prolonged survival but also partially restored PIKfyve levels ( Fig. 2G, S2E). We then administered GSK2606414, an inhibitor of the protein kinase RNA-like endoplasmic reticulum kinase (PERK) arm of the UPR 19 , to prion-infected tga20 COCS at 21 dpi (Fig. S2F). GSK2606414 partially restored PIKfyve levels ( Fig.   2H, S2G), strengthening the notion that chronic ER stress is the direct cause of PIKfyve downregulation.
Dynamic acylation and deacylation control protein stability 23 and can be affected by the UPR 24 . Being juxtaposed to endosomal membranes, PIKfyve is a plausible candidate for acylation. Acyl-resin assisted capture (acyl-rac) of mouse brain homogenates showed that PIKfvye, but not the cysteine-less transloconassociated protein-α (TRAPα), was acylated (Fig. 3A). Prion-infected samples, however, showed massively reduced PIKfyve acylation already at 60 dpi, a time point showing vigorous UPR but only minimal depletion of total PIKfyve ( Fig. 2A, 3B, S3A).
An acylation prediction database 25 pointed to cysteine residues 202 and 203 as potential palmitoylation sites in all three isoforms of PIKfyve (Fig. S3B). We therefore expressed PIKfyve-GFP or PIKfyve C202+203A GFP, in which both residues were replaced by alanines, in Gt1 cells. Acyl-rac confirmed that mutagenesis of cysteine residues 202 and 203 prevented PIKfyve acylation (Fig. S3C). We next investigated the function of the acylation machinery in prion infections. The mouse genome encodes 23 acyl transferases characterized by a zinc-finger-domain and an aspartate-histidine-histidine-cysteine tetrapeptide (zDHHC), denoted zDHHC1-24.
Quantitative real-time PCR (qPCR) on terminally scrapie-sick mouse brains failed to identify altered mRNA levels of any zDHHC member (Fig. S3D). We then mined a database of transcriptional changes during the course of prion infection 17 . None of the zDHHC mRNAs were altered at any stage of the disease (Fig. S3E).
Next, we suppressed each zDHHC individually by siRNA. All zDHHC mRNAs were reduced by >80%. After 72 hours, cells were harvested and subjected to acyl-rac. Only the knockdown of zDHHC9 and zDHHC21 modestly reduced the acylation of PIKfyve (Fig. 3D, S3F). We next knocked down zDHHC9 and 21 individually  D). Each dot represents a separate experiment (unpaired t-test). F: Gt1 cells were transfected with siRNA against zDHHC9+21 and stained with LAMP1. Arrows: vacuoles. F-G: Flag-tagged zDHHC9 or zDHHC21 were expressed in Gt1 cells were transfected with plasmids encoding. After 72 hours, zDHHC9/21 (magenta) were localized to the Golgi apparatus (GM130 + , cyan) in NBH-exposed cells but not in prion-infected cells.
Crucially, at 120 hours the zDHHC9/21 siRNA mix induced not only a downregulation of PIKfyve but also the appearance of vacuoles (Fig. S3G, 3E) similar to those of prion-infected cells.
Next, we expressed in Gt1 cells zDHHC9 and zDHHC21 fused to a C-terminal flag-tag. While zDHHC9 and zDHHC21 localized to Golgi complexes of NBH-treated cells, in prion-infected cells both enzymes became drastically mislocalized to the cell periphery (Fig. 3F-G). The mislocalization of zDHHC9/21 may explain the loss of PIKfyve acylation and its degradation thereby driving spongiogenesis.

Cellular consequences of prion-induced spongiosis
Since PIKfyve is necessary for the maturation of late endosomes into lysosomes, prion-associated spongiosis may result from a blockade of endosomal maturation. We found that LAMP1, a marker for late endosomes/lysosomes, was associated with vacuoles in both prion-infected Gt1 cells and in Gt1 cells depleted of PIKfyve using shRNA (Fig. 4A). We wondered whether the association of LAMP1 with vacuoles can also be observed in other model systems of prion disease. We infected COCS generated from 9-day old tga20 mice with RML. At 50 dpi, we detected LAMP1 upregulation in neurons (Fig. 4B, S4A). As in Gt1 cells, vacuoles in COCS were lined by LAMP1 (Fig. 4C). Treatment of tga20 COCS with POM1 (10 days) also resulted in microvacuolation and LAMP1 in neurons (Fig.4D, S4B). These results suggest that spongiosis originates from late endosomal/lysosomal compartments. To differentiate between these two possibilities, we stained PIKfyve-depleted Gt1 cells (72 hours after siRNA transfection) with the ratiometric pH probe Lysosensor Yellow/Blue (5 µM, 5 min). Flow cytometry revealed no enhanced acidification in vacuolated cells (Fig. 4E), suggesting that vacuoles represent stalled prelysosomal compartments.
Increased activity of LAMP1 and other lysosomal enzymes is a frequent feature of lysosomal dysfunction.
Indeed, we found a three-fold increase in the enzymatic activity of lysosomal hexosaminidase-β in brains of prion-infected mice (Fig. 4F). In addition, we monitored the RNA expression levels of cathepsin-D, cathepsin-A, glucocerebrosidase, α-galactosidase and mucolipin-1, previously shown to be upregulated in lysosomal diseases 26 , and found all of them to be upregulated in brains of terminally scrapie-sick mice (Fig. 4G).  Fig. S4A). C: Higher magnification showing LAMP1 + vacuoles (arrowheads) in prion-infected COCS. D: tga20 COCS treated with POM1 or control IgG (10 days) and immunostained with NeuN and LAMP1. Nuclei: DAPI. POM1-treated COCS showed increased LAMP1 expression (quantification: Fig. S4B). NeuN + cells exhibited vacuoles (arrowheads). E: Gt1 cells were transfected with control siRNA (scrambled) or siRNA targeting PIKfyve for up to 72 hours and stained with LysoSensor Yellow/Blue. Cells were gated based on emission spectra, and mean fluorescence intensities (MFIs) per sample were quantified. Increased 530/450 MFI ratio indicates enrichment of acidic compartments. No acidic compartment expansion in PIKfyve-depleted cells. F: β-hexosaminidase A activity in brain lysates from terminally scrapie-sick and NBH-inoculated mice. Control: NBH inoculated mice. Prion-infected brain lysates showed 2-2.5 fold increased activity. Panels show independent triplicates. Statistics: unpaired t test. G: Lysosomal enzymes, but not β-actin, were elevated in brains of terminally scrapie-sick mice. Panels show independent triplicates. Statistics: ANOVA.
Lysosomes undergo cycles of fission and fusion, which are impaired by the absence of PIKfyve, resulting in coalesced endolysosomes 27 . The consequences include nuclear translocation of TFEB and depletion of TRPML1, a channel controlling lysosomal size [28][29][30] . TRPML1, whose lysosomal association relies on PIKfyve 31 and was abolished by shRNA-mediated PIKfyve knockdown in Gt1 cells (Fig. S4C), was co-precipitated by anti-LAMP1 antibodies in NBH-inoculated but not in scrapie-sick mouse brains (Fig. S4D). Total TRPML1 protein levels were unaltered.
Upon dephosphorylation , transcription factor EB (TFEB) translocates to the nucleus and promotes coordinated expression of lysosomal genes 26,32,33 . TFEB became progressively dephosphorylated in prioninfected mouse brains whereas total TFEB levels remained constant (Fig. 5A). We then examined a subset of TFEB-regulated lysosomal genes in terminally prion-sick brains. The mRNA levels of all investigated genes were conspicuously increased, whereas the non-lysosomal gene STAT3 was not (Fig. 5B). These findings could reflect the vivacious proliferation of activated microglia in terminal prion disease. We therefore investigated the lysosomal status of Gt1 cells (Fig. 5C) and COCS (Fig. 5D). We found that prion infection led to TFEB dephosphorylation in both models. Moreover, prion-infected Gt1 cells (75 dpi) displayed nuclear translocation of TFEB (Fig. 5E) and prion-infected Gt1 cells (75dpi), as well as Gt1 cells depleted of PIKfyve by shRNA, showed upregulation of TFEB responsive genes (Fig. 5F-G). This disproves that lysosomal gene upregulation merely indicates distorted tissue composition, and suggests that TFEB-mediated reactions are a cell-autonomous consequence of PIKfyve depletion. Finally, we tested the relationship between TFEB and lysosomal homeostasis by treating prion-infected Gt1 cells with siRNA against TFEB. 72 hours post siRNA transfection, the expression of TFEB-responsive genes was normalized, yet the vacuolation of prion infected cells was unaltered (Fig. S4E-G), suggesting that TFEB-related lysosome pathologies may be a consequence rather than a cause of spongiosis.
The following hierarchy of events emerges from these findings. Chronic activation of the UPR leads to mislocalization of acyl transferases. This, in turn, compromises PIKfyve acylation, reduces its half-life, and destabilizes its association with VAC14 and FIG4. The resulting reduction of PI(3,5)P2 stalls lysosome maturation and induces inappropriate TFEB-dependent transcriptional responses. We challenged this model by interfering with each of its predicted checkpoints. Firstly, we asked whether overexpression of zDHHC9 and zDHHC21 might restore PIKfyve levels and abrogate vacuole generation in prion-infected Gt1 cells (75 Figure 5: Prions induce lysosomal dysregulation via TFEB. A: Time course of TFEB phosphorylation in prion-infected C57BL/6 mice. Total TFEB was unchanged but reduced TFEB phosphorylation was evident at ≥90 days. B: TFEBcontrolled transcripts in brains of terminally scrapie-sick C57BL/6 mice were measured by qPCR. Prion infection led to upregulation of these genes, but not of STAT3. Control: NBH-inoculated mice. ANOVA on independent triplicates C: Prion-infected Gt1 cells (75dpi) showed reduced TFEB phosphorylation. Total levels of TFEB remain unchanged. D: COCS prepared from tga20 mice were infected with RML prions and cultured for 5 weeks. Cell lysates were prepared at 4 wpi followed by Western blot analysis using anti-phospho-TFEB antibody. Activated (dephosphorylated) TFEB was increased in RML infected COCS. Total levels of TFEB remain unchanged. E: Gt1 cells (as in D) were fixed and stained with anti-TFEB antibody. Prion-infected cells showed nuclear translocation of TFEB. F: TFEB responsive genes were upregulated in Gt1 cells chronically infected with RML prions (75 dpi). Panels depict independent triplicates. Statistics: ANOVA. G: TFEB responsive genes were upregulated in cells transfected with shPIKfyve for 5 days compared to shCtrl (shRNA against luciferase). Panels depict independent triplicates. Statistics: ANOVA. dpi). Indeed, transfection (72 h) with both zDHHC9 and zDHHC21 partially restored PIKfyve levels and reduced the number of vacuolated cells (Fig. 6A-B). Secondly, we attenuated the UPR in prion-infected Gt1 cells (75 dpi) by lentiviral transduction of GADD34, which dephosphorylates eIF2α 34 . Again, we found restoration of PIKfyve levels and a significant decrease in the frequency of vacuolated cells (Fig. 6C-D).
The above evidence shows that neuronal vacuolation is tightly coupled to PIKfyve depletion in disparate experimental models, and the virulence of human prion diseases correlates with the extent of PIKfyve depletion. Prion infection causes a sustained UPR, as do other neurodegenerative diseases 35 including granulo-vacuolar degeneration of Alzheimer's disease 36 and inclusion-body myopathy which features prominent myofiber vacuolation 37 . This begs the question whether PIKfyve may play a role in vacuolating diseases other than prion infections. Competing Interest declaration: The authors declare no competing financial interests.

Data availability
All original data has been included in the article. Raw data and reagents will be made available by the corresponding author upon request.