Library, reagents, buffers, and plasmids

A library of FDA-approved drugs (Screen-Well FDA-approved drug library V2 version 1.0), containing 768 drugs was purchased from Enzo Life Sciences. Chemical reagents used in this study and their commercial sources were as follows: VCP inhibitor CB5083 (Cayman), n-dodecyl-beta-D-maltoside (DDM, Dojindo), itraconazole (TCI), posaconazole (TCI), ketoconazole (TCI), ravuconazole (TCI), terconazole (Sigma-Aldrich), miconazole nitrate (TCI), hydroxyitraconazole (Toronto Research Chemicals), imatinib (TCI), lapatinib (Cayman Chemical), digoxin (TCI), progesterone (TCI). Filipin complex (F9765) and 25-hydroxycholesterol (25HC) were obtained from Sigma-Aldrich.

The following buffers were used: Dulbecco’s phosphate-buffered saline (PBS), PBST (PBS supplemented with 0.05% Tween 20), tris-buffered saline (TBS, 20 mM Tris-HCl pH 7.5 and 137 mM NaCl), TBST (TBS supplemented with 0.05% Tween 20), TNET (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM EDTA, and 1% Triton X-100), TNED (25 mM Tris-HCl pH 7.4, 150 mM NaCl, 5 mM EDTA, and 0.5% DDM).

The plasmid for expression of FLAG-tagged ΔNTD-NPC1-P691S/I1061T-tGFP was prepared by introducing P691S mutation into pCMV-ΔNTD-NPC1-I1061T-tGFP. Site-directed mutagenesis was performed by PCR mutagenesis with KOD plus high-fidelity polymerase, DpnI, and Ligation high ver.2 (all from Toyobo). The following primers were used for the site-directed mutagenesis: forward primer, ATCTCGTTCCTGGTGCTGGCTGTTG (from CCG to TCG); reverse primer, GACTTCAATCACAATGAGGGTCAAG.

Synthetic procedures and characterization data of photo-crosslinking probes (itraAZY, itraACT, itraBIO, and lapaAZY) are available in S1 Appendix.

Cell culture, transfections, and stable cell lines

HEK293 cell lines stably expressing wild-type (WT), I1061T, ΔNTD, ΔNTD-I1061T mutants of FLAG-tagged NPC1-FLAG-tGFP (NPC1-GFPs) under the control of CMV promoter were described previously [20], and were cultured in high-glucose DMEM (FUJIFILM Wako Pure Chemicals) supplemented with 5% heat-inactivated fetal bovine serum (FBS) at 37°C in a humidified incubator (5% CO₂).

To establish HEK293 stably expressing ΔNTD-NPC1-P691S/I1061T-GFP, HEK293 cells at around 80% confluence were transfected with the plasmid using Lipofectamine LTX (Invitrogen) transfection reagents, and selection pressure was applied with 0.4 mg/mL G418 sulfate. Single clones were isolated by limiting dilution and expanded.

Image-based screen

Cells stably expressing NPC1-I1061T-GFP were cultured on poly-D-Lys-coated glass-bottomed 96-well plates (#164588, Nunc) or CELLSTAR μClear 96-well plates (#655090, Greiner), and treated with vehicle (0.1% DMSO) or test compounds (10 μM) at 50–70% confluence. After 20–21 h, cells were fixed with 10% formalin-PBS containing 1 μg/μL Hoechst 33342 for 30 min, and washed with PBS. Images were acquired on an IN Cell Analyzer 2000 (GE Healthcare) with the following settings; 60x (air) objective lens, 2x2 binning, Quad1 dichroic mirror and DAPI/FITC filters, with hardware autofocus, plate heater 30°C, and exposure time for DAPI 0.060 sec and FITC channel 0.450 sec. Six fields of images separated by 200–300 μm were acquired for each well. Processing of the acquired images and extraction of image features were performed with EBImage, an R package for image processing [24]. From a pair of GFP and Hoechst images, we extracted the NPC1-GFP pattern (from GFP images), cell area (from GFP images), and the number and area of nuclei (from Hoechst images) as illustrated in Fig 1C. To extract the NPC1-GFP pattern, we applied the following image processing sequence: gamma transform (gamma = 0.5), subtraction of gaussian-blurred image (sigma = 35) as a background subtraction, adaptive thresholding (kernel size = 25, offset = 0.02), closing operation (disc-shaped kernel with size = 7) to the thresholded binary image, and removal of too small objects (size < 50). From the raw GFP image and the extracted GFP pattern, GFP intensity within the extracted pattern, NPC1-GFP area, and shape-related features, including circularity, were calculated. To extract cell area from the GFP image, we applied the following image processing sequence: gamma transform (gamma = 0.1) to reduce the contribution from high-intensity areas, gaussian blur (sigma = 25), and Otsu’s thresholding to obtain a binary image for overall cell area. To calculate the number and area of nuclei, the following image processing sequence was applied: mean filter (disc-shaped kernel with size = 11), gamma transform (gamma = 0.6), Otsu’s thresholding, segmentation (distance map transform and watershed segmentation), filter objects by size (1,000 < size < 7,000) to remove too large or too small objects, and calculate the number and area of nuclei. The calculated image features along with the treatment conditions are available from the Mendeley Data repository (http://dx.doi.org/10.17632/jr23ccpp46.3). The raw images obtained from the screen are available from the Dryad repository (https://doi.org/10.5061/dryad.fn2z34tpt), and an example R script for the image processing is also available from the Mendeley Data repository (http://dx.doi.org/10.17632/jr23ccpp46.3).

Flow-cytometric analysis of expression levels of NPC1-GFPs

Cells grown on 12-well plates were treated as indicated at 70–90% confluence, and incubated for 20–21 h. The cells were then detached with 0.25% trypsin-EDTA, and suspended in culture medium, and the GFP fluorescence from 10,000 cells was recorded on a FACS Canto II (BD Biosciences). The data was subjected to logicle transformation for visualization [25]. Data analysis and visualization were performed with R (version 3.3.3) along with RStudio (version 1.0.153) using the flowCore and flowViz packages. The raw data is available from the Mendeley Data repository (http://dx.doi.org/10.17632/jr23ccpp46.3).

Clear-native PAGE analysis of NPC1-GFPs

Clear-native polyacrylamide gel electrophoresis (CN-PAGE) was run as reported [26] with slight modifications. Cells stably expressing the indicated NPC1-GFPs were treated with vehicle (0.1% DMSO) or test compounds for 20–21 h. The cells were washed with PBS, and lysed with TNED buffer supplemented with cOmplete protease inhibitor cocktail (EASYpack EDTA-free, Roche) on ice for 1 h. Cell debris was removed by centrifugation (14,000 g, 10 min, 4°C) and the supernatant was collected. The concentration of the lysate was normalized based on total protein concentration as determined by BCA protein assay. The same amount of the lysate (27 μg total protein / lane) was mixed with 0.5 volume of 3x loading buffer (0.01% DOSS, 0.05% Ponceau S, 60% glycerol, and 50 mM Tris-HCl pH 8.6), and loaded on 5–12% SuperSep Ace gel (Wako). CN-PAGE was performed at 0°C with cathodic buffer (0.02% DDM, 0.005% DOSS, 5% glycerol, 25 mM Tris, and 192 mM glycine) and anodic buffer (25 mM Tris and 192 mM glycine) at constant current (15 mA) for 90–100 min. The gel was briefly washed with water, and GFP fluorescence was imaged on a FLA7000 gel imager (without ND filter, 50 μm grid). Gel images were processed with Image J for visualization (cropping and linear contrast adjustment) and for quantification. For quantification, the lane profile was first extracted from the images after background correction (rolling ball algorithm). The lane profiles were then plotted and peak area was calculated using R in combination with hyperSpec and baseline packages. To stain total protein in the native-PAGE gels, we applied 2,2,2-trichloroethanol (TCE) staining [27]; the native-PAGE gels after GFP detection were soaked in 10% TCE in 50% aqueous methanol for 10 min at room temperature with gentle shaking. The gels were briefly rinsed with water and irradiated with UV (302 nm) for 2 min on a UV transilluminator (UVP dual intensity UV transilluminator). Fluorescence images were acquired on a EzCapture cooled CCD camera imager equipped with UV transilluminator (302 nm) and 560 nm long-pass filter (YA-3) filter. Note that no recognizable GFP signal could be observed at the top of the gels under our experimental condition, implying that terminally misfolded NPC1 proteins were not soluble in the lysis buffer, if any.

Ubiquitination assay

Confluent cells on poly-D-Lys-coated 6-well plates were treated with 3 μM CB5083, which inhibits Valosin-containing protein (VCP; required for extraction and subsequent proteasomal degradation of ubiquitinated membrane proteins in the ER [28]) and the indicated compounds for 6 h. Then the cells were washed once with PBS, and lysed with TNED buffer supplemented with cOmplete EDTA-free protease inhibitor cocktail (Roche) and 10 mM N-ethylmaleimide on ice for 1 h. The lysate was centrifuged to remove cell debris. The total protein concentration was determined by BCA protein assay and adjusted to 2.0 μg/μL. The diluted lysate (300 μL) was mixed with pre-washed anti-FLAG M2 magnetic beads (Sigma-Aldrich, M8823, 4 μL bed volume/sample) and the mixture was rotated overnight at 4°C. The beads were washed twice with 300 μL of TNED buffer, and eluted with 30 μL of 1×Laemmli sample buffer by heating at 60°C for 15 min. The samples (10 μL/lane) were resolved by SDS-PAGE (40 mA, 45 min, in an ice-water bath) using 7.5% SuperSep Ace gel (Wako). The proteins were transferred to PVDF membrane (Immobilon P, Millipore), blocked with 1% BSA-TBST (r.t., 1 h), and stained overnight with anti-ubiquitin antibody conjugated with HRP (1/2,000 dilution, P4D1, mouse monoclonal antibody, sc-8017, Santa Cruz Biotechnology) [29] in signal enhancer, CanGetSignal solution 2 (Toyobo). The membrane was washed with TBST and bands were detected with Immobilon Western chemiluminescence reagent (Millipore) using an EzCapture imaging system (ATTO). The blotted membrane was washed with water, and incubated with a stripping buffer (2% SDS, 62.5 mM Tris-HCl pH 6.8, 100 mM 3-mercapto-1,2-propanediol) at 50°C for 30 min. The stripped blot was washed with water and TBST, and blocked again with 1% BSA-TBST for 1 h. The membrane was washed with TBST, and incubated with anti-FLAG M2 antibody (Sigma-Aldrich, F1804, 1/2,000 dilution) [20] diluted in CanGetSignal solution 1 for 1 h. After five washes in TBST, the membrane was shaken with goat polyclonal anti-mouse IgG conjugated with HRP (Millipore, 12–349) [20] diluted in CanGetSignal 2 (1/2,000 dilution) for 1 h. The membrane was washed five times and chemiluminescence detection was performed as described above. The quantification of the ubiquitination level and NPC1 level was performed as described above for CN-PAGE data analysis.

Photoaffinity labeling

Photoaffinity labeling of NPC1 was performed as described previously [20] with slight modifications. Confluent cells stably expressing the indicated NPC1-FLAG-GFP construct on a 10 cm dish were washed with PBS (2 mL) and scraped into cold PBS (1 mL). The cells were centrifuged (300×g, 5 min, 4°C) and the pellet was resuspended in TSE buffer (20 mM Tris-HCl pH 7.6, 250 mM sucrose, and 1 mM EDTA) supplemented with cOmplete EDTA-free protease inhibitor cocktail (Roche). The cells were disrupted with a probe sonicator (USP-300, Shimadzu) at 0°C, and nuclei and cell debris were removed by centrifugation (2,000×g, 5 min, 4°C). The supernatant was centrifuged (125,000×g, 1 h, 4°C) to pellet membranes. The pelleted membranes were resuspended in TBS and total protein concentration was determined by BCA protein assay; the result was used to adjust the concentration to 4.0 μg/μL with TBS. The membrane was aliquoted onto a U-bottomed polypropylene 96-well plate (Greiner, 650261) and treated with the indicated photoaffinity labeling probe (1/100 volume, 50 μM stock solution in DMSO) and a competitor (1/100 volume, 5 mM stock solution in DMSO) or DMSO, and incubated for 30 min on ice. The samples were irradiated with UV light (Handy UV lamp, UVP, 365 nm, ~1 cm from the sample) on ice for 20 min. The irradiated membranes were solubilized by adding 10% (v/v) SDS-10% (v/v) Triton X-100 solution (1/10 volume) and incubated at 0°C for 30 min. Separately, a 10×catalyst solution for the click reaction (10 mM CuSO₄, 1 mM TBTA, 50 mM aminoguanidine, and 50 mM sodium ascorbate) was prepared as follows [30]: 100 mM CuSO₄ (12.5 μL), water (11.3 μL), 1.7 mM TBTA in DMSO/tBuOH (4:1) (7.5 μL), 200 mM aqueous aminoguanidine (31.3 μL), and freshly prepared 100 mM aqueous sodium ascorbate (62.5 μL) were mixed and vortexed to obtain a colorless heterogenous solution. To the solubilized membrane was added biotin-PEG3-azide (Tokyo Chemical Industry, 1.25 μL of 5 mM stock solution in DMSO, 50 μM), and then the 10×catalyst solution (12.5 μL) was added. The reaction was performed at room temperature for 1 h. Then, the reaction mixture was diluted 10-fold with TNET, and immunoprecipitated with anti-FLAG M2 magnetic beads (4 μL bed volume / sample, prewashed with TBS and TNED) at 4°C overnight. The magnetic beads were washed twice with TBS (400 μL) and once with TNET (400 μL), and eluted with 1×Laemmli sample buffer (30 μL/sample) at 60°C for 15 min. The eluates were resolved by SDS-PAGE (SuperSep Ace 7.5% gel, Wako) and bands were transferred to PVDF membrane (Immobilon P PVDF membrane, Millipore). The membrane was blocked with 1% (w/v) BSA-TBST (0.05% Tween 20 in TBS) at 4°C overnight, and probed for biotin with ImmunoPure streptavidin HRP (1/2,000 in TBST, Pierce, 21126) [20] for 1.5 h at room temperature. The membrane was washed and chemiluminescence detection was performed as described above for ubiquitination assay. The membrane was stripped and re-probed with anti-FLAG M2 antibody as described above. Quantification of the bands was performed with an ImageJ macro, BandPeakQuantification tool [31].

LAMP1 immunostaining and colocalization analysis

Stable cell lines expressing NPC1-GFPs were grown to 50–70% confluence on single-well glass-bottomed dishes (IWAKI, No. 3911–035) coated with poly-D-Lys, and treated as indicated for 20–21 h. The cells were washed once with PBS and fixed with 10% formalin-PBS at room temperature for 30 min. The fixed cells were permeabilized with 0.1% Triton X-100 in PBS, blocked with 1% BSA-PBST, and stained with mouse monoclonal anti-LAMP1 antibody (Abcam, H4A3, ab25630) [20] at a dilution of 1:100 in PBST for 2 h. The cells were washed with PBST, and incubated with anti-mouse IgG conjugated with Alexa546 (ThermoFisher, A-11003) [20] diluted with PBST at 1:1000 for 1 h. After three washes with PBST, images of NPC1-GFPs and the Alexa548-stained LAMP1 were acquired on an FV3000 confocal laser microscope (×60 oil-immersion objective lens, UPLSAPO 60XS2). Colocalization analysis was performed in the R environment (version 3.3.3) using the EBImage package [24] and visualized with the ggplot2 package. Differences between the groups were evaluated by applying the Kruskal-Wallis rank sum test and Dunn’s multiple comparison with the Benjamini-Hochberg method for adjustment of the false discovery rate, using an R package, dunn.test. The raw images used and the corresponding R script are available from the Mendeley Data repository (http://dx.doi.org/10.17632/jr23ccpp46.3).

Filipin staining

NPC fibroblasts (I1061T/I1061T, GM18453, obtained from the NIGMS Human Genetic Cell Repository at the Coriell Institute for Medical Research) were grown on glass-bottomed 96-well plates (#164588, Nunc) and treated as indicated. Filipin staining for evaluation of cholesterol accumulation was performed as previously described [20], and images were acquired on an IN Cell Analyzer 2000 using the following settings: 60x (air) objective lens, 2x2 binning, Quad1 dichroic mirror and DAPI/DIC filters, with hardware autofocus, plate heater 30°C, and exposure time for DAPI (filipin) 0.100 sec and bright field 0.100 sec. Nine fields of images separated by 300 μm were acquired for each well. Quantification of relative filipin intensity, defined as filipin intensity within high-intensity lysosome-like organelles per total filipin intensity within cells, was performed using R and the EBImage package [24]. Details of the image processing are summarized in S1 Fig, and an example R script for the image analysis is available from the Mendeley Data repository (http://dx.doi.org/10.17632/jr23ccpp46.3).

Development of an image-based screen to identify compounds that improve the folding efficiency of the NPC1-I1061T mutant

To evaluate the chaperone-like activity of test compounds towards the NPC1-I1061T mutant, we exploited the subcellular localization of NPC1-I1061T-GFP. This folding-defective NPC1 mutant is recognized by the cellular quality control machineries and retained in the ER for eventual degradation via the ERAD pathway [13, 14, 20]. Thus, at the steady state, the NPC1-I1061T-GFP mutant shows an ER localization pattern. In the presence of proteostasis regulators, including PCs, the folding intermediate of the NPC1 mutant is stabilized to a more native-like conformational state, and is allowed to traffic to the LE/L compartment, similarly to wild-type (WT) NPC1-GFP. Our previous method to evaluate this localization change involved colocalization analysis with immunostained lysosomal-associated membrane protein 1 (LAMP1), a marker of the LE/L compartment [20], or simply visual phenotypic classification [21]. Although both methods offered a reliable measure for evaluation of PC activity, their signal-to-noise ratio and throughput were insufficient for large-scale screening.

To overcome these problems, we developed an automatable, image-based screen that quantifies the localization pattern of NPC1-I1061T-GFP in terms of morphological features. In this screen, time-consuming and costly immunostaining is not required, and only NPC1-I1061T-GFP images and Hoechst (nuclear stain) images are acquired. The distribution pattern was extracted from the GFP images by application of image processing techniques, and the morphology of the pattern was quantified in terms of several selected features, which were confirmed to reflect the ER-to-LE/L change of NPC1-I1061T localization in our preliminary experiments (Fig 1C). In addition, the approximate cell area was extracted from the GFP images by applying a lower threshold, and the number of nuclei and average nuclear size were also obtained from the Hoechst images as measures of cytotoxicity (Fig 1C).

Among the image features we tested, “circularity” and “smallObjRatio” gave high sensitivity and S/N ratio. The circularity is defined as the average ratio of the area of each object to the area of a circle with the same perimeter (4*π*area/perimeter²) for each NPC1-GFP-positive object, and this feature gives higher values for circular objects and lower values for longer, branched objects. A smallObjRatio is defined here as the area ratio of NPC1-GFP objects with less than a specified threshold size with respect to the total NPC1-GFP-positive area. The good performance of these parameters is intuitively understandable, considering that the localization of NPC1-I1061T-GFP changes from network-like, larger patterns of ER localization, to vesicular, smaller patterns of LE/L localization. We also confirmed that the expression level of the mutant protein, measured as GFP fluorescence intensity per cell area (GFPInt), exhibited a dose-dependent increase, as previously reported [20]. As these three features capture different aspects of the chaperone-induced localization change, we envisioned that multiparametric selection of hit compounds would increase the sensitivity and accuracy of the hit selection process. Of the three features, the circularity feature generally gave the highest Z’ factor of >0.5 for samples treated with 0.3–3 μM mo56HC, a previously developed NPC1 chaperone, indicating suitability of the screen for large-scale screening and for quantitative comparison of potency [32]. Although all three features showed a clear dose-response relationship for mo56HC, GFPInt was much noisier and less sensitive than the other morphological features (Fig 2B). Thus, we established an image-processing protocol that yields several good measures of chaperone activity towards the NPC1-I1061T mutant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Screening of an FDA-approved drug library identified several classes of potential NPC1 chaperones.

(A) Three image features (circularity, smallObjRatio, and GFPInt) capture the chaperone-mediated localization change of NPC1-I1061T-GFP. Data points are from a series of mo56HC-treated wells on 11 screening plates (duplicate dose-response analysis per plate). The image features are presented in a set of 2D plots and shown as robust S/N ratio (effect size normalized with median absolute deviation) after plate-wise normalization with respect to plate medians. (B) Dose-dependency of mo56HC examined by employing each of the three image features. Circularity gave the best separation of negative and positive data. Data points and error bars represent mean ± SD (n = 11 experiments, each performed in duplicate). (C) Screening results represented in a set of 2D plots (circularity, smallObjRatio, and GFPInt). Screening was performed at 10 μM. The dashed lines denote thresholds for each of the image features, and compounds that exceeded all the thresholds were selected as hits (red circles). (D) Chemical structures of the hits. (E) Heatmap representation of the screening results for the selected hits and closely related compounds in the library. The heatmap colors encode the robust S/N ratio of image features.

https://doi.org/10.1371/journal.pone.0243746.g002

Screening of a library of approved drugs

With the validated image-analysis protocol in hand, we used it to perform drug-repurposing screening of 768 approved drugs in order to identify compounds that affect the folding efficiency or proteostasis status of the NPC1-I1061T mutant. The image features obtained from the screen were first normalized to plate medians in a plate-by-plate manner, and transformed into a robust Z score, which is defined as the effect size (values–plate median) divided by the median absolute deviation (MAD, a non-parametric version of SD) of the plate [33]. The positive control mo56HC, placed on each plate, gave a reproducible dose-dependent effect in terms of circularity, smallObjRatio, and GFPInt (Fig 2A and 2B), and again the circularity measure gave the clearest separation from negative controls. Note that cells began to round up at 10 μM mo56HC due to cytotoxicity, resulting in a lower smallObjRatio, while circularity was less affected.

For hit selection, we used multiparameter thresholding of the circularity, smallObjRatio, and GFPInt. The thresholds were determined by taking median values for the features (circularity 4.03; smallObjRatio 4.01; GFPInt 1.50) from a marginally active positive control group (0.03 μM mo56HC). With this threshold, no negative controls (vehicles on each plate and blank wells on the library plates) were identified as hits (0 of 76 negative controls), and 5 hit compounds were identified: itraconazole, posaconazole, ketoconazole, lapatinib, and progesterone (Fig 2C and 2D and S2 Fig).

Three of the hit compounds (itraconazole, posaconazole, and ketoconazole) are azole antifungals, which preferentially inhibit fungal lanosterol demethylase, a cytochrome P450 enzyme, along with human CYP enzymes. Lapatinib is a kinase inhibitor that primarily targets endothelial growth factor receptors (EGFRs). The other compound was a steroidal hormone, progesterone, which we previously identified during our SAR study of oxysterol-based chaperones, so we excluded this from further analysis [21]. Notably, the chaperone-like activity was specific for a subset of the azole antifungals and kinase inhibitors, indicating that the activity is not mediated through CYPs or kinases (Fig 2E). Also, comparison of the azole antifungals in the library suggested the importance of a specific substructure, N,N-diaryl piperazine, for chaperone activity against the NPC1 mutant.

In addition to these hit compounds, we noted that the proteasome inhibitor bortezomib and the Na⁺,K⁺-ATPase inhibitor digoxin exhibited significantly higher circularity values, despite their high toxicity (lower cell area and fewer nuclei) under the screening conditions (10 μM treatment) (S2 and S3 Figs). In contrast to mo56HC or the hit compounds, which gave high smallObjRatio values, bortezomib gave a low value for this parameter. Inspection of the raw images revealed that bortezomib treatment resulted in accumulation of NPC1-I1061T-GFP protein in an aggresome-like structure, consistent with the potent proteasome-inhibitory activity of bortezomib (S2 Fig) [34]. Although images from digoxin-treated cells contained only small numbers of cells due to significant toxicity (S3 Fig), the localization of NPC1-I1061T-GFP within the remaining cells implied a distribution pattern distinct from that of untreated cells. Therefore, we included digoxin for further validation.

Azoles and lapatinib behave similarly to oxysterol-based pharmacological chaperones

To validate the obtained hit compounds, we first performed dose-response analysis of the hits and related compounds (Fig 3A and 3B). Among the azoles tested, itraconazole (and its primary metabolite hydroxyitraconazole) and posaconazole exhibited potent activity, comparable to that of mo56HC. Ketoconazole and ravuconazole were only weakly active, and the other azoles, including terconazole and miconazole, were inactive. Lapatinib, a kinase inhibitor, was confirmed to be active at 10 μM, but was much less potent than the azoles, while the other kinase inhibitors, including imatinib and erlotinib, were inactive. Interestingly, digoxin exhibited an activity profile distinct from those of itraconazole, lapatinib, or mo56HC; while these three compounds increased the GFP intensity in addition to the circularity feature, digoxin reduced the GFP intensity in a dose-dependent manner, implying that digoxin acts differently from mo56HC, azoles, or lapatinib.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Validation of the hits by dose-response analysis and flow-cytometric analysis.

(A) Dose-response analysis of the hits and related compounds. (B) Chemical structures of the compounds used in (A) and not shown in Fig 2. (C) Stabilizing effect of the hits on the NPC1 mutant, examined by flow-cytometric analysis. Cells stably expressing NPC1-I1061T-GFP were treated as indicated for 21 h, and processed for flow-cytometric analysis to quantify the steady-state expression level of the mutant protein. Consistent with the stabilizing effect of mo56HC (upper panel), lapatinib (upper panel) and itraconazole (middle panel) both stabilized the mutant protein. (D) Dose-dependent increase in the mean GFP fluorescence intensity from the flow-cytometric data shown in (C). Itraconazole (left panel) increased the NPC1-I1061T-GFP level approximately 270% with an EC₅₀ of 0.15 μM, and digoxin (right panel) reduced the expression level to 45% with an EC₅₀ of 0.31 μM.

https://doi.org/10.1371/journal.pone.0243746.g003

To further validate the hit compounds, we examined their effect on the steady-state expression level of NPC1-I1061T-GFP, expression of which is driven by the CMV promoter. As previously demonstrated, pharmacological chaperones upregulate NPC1-I1061T-GFP, probably via stabilization of the destabilized mutant protein [20]. Flow-cytometric analysis clearly confirmed an up-regulatory effect of the azoles (itraconazole, OH-itraconazole, posaconazole, and ketoconazole) and lapatinib, and the dose-response relationships were consistent with those obtained from the image-based assay, supporting the idea that these compounds act as pharmacological chaperones for NPC1 (Fig 3C and 3D). Consistent with the idea, the up-regulatory effect was also observed in the presence of emetine, a protein synthesis inhibitor, supporting our hypothesis that the hits attenuate the degradation rate of the NPC1 mutant (S4 Fig) [28, 35, 36]. Digoxin was also confirmed to decrease the expression level of NPC1-I1061T-GFP at sub-micromolar concentrations. This different profile of digoxin clearly supports a distinct action mechanism for digoxin from those of the other hit compounds.

Clear-native PAGE analysis supports improved folding status of NPC1-I1061T mutant in the presence of the hit compounds

In the above experiments, the instability of the NPC1-I1061T mutant was captured in terms of the steady-state expression level of the mutant protein in the cells. We also examined this point by clear-native polyacrylamide gel electrophoresis (CN-PAGE), which allows evaluation of the hydrodynamic status of GFP-fused proteins under non-denaturing conditions by means of in-gel fluorescence detection, similarly to fluorescence-detection size-exclusion chromatography [26, 37, 38]. Cells expressing either WT- or NPC1-I1061T-GFP were treated as indicated, and their lysates were subjected to CN-PAGE analysis, which utilizes a mixture of a mild anionic detergent, dioctyl sulfosuccinate (DOSS), and a nonionic detergent, dodecyl maltoside (DDM). Note that we also tested another CN-PAGE protocol that relies on a colorless derivative of Coomassie Brilliant Blue (mCBB) and obtained comparable results, but we selected the former condition for routine use due to the instability of mCBB [39]. CN-PAGE analyses revealed different electrophoretic patterns for wild-type NPC1-GFP and NPC1-I1061T-GFP (Fig 4A and 4C). While wild-type NPC1-GFP migrated mainly as the faster migrating band, less of this faster band was observed for the I1061T mutant and a significant portion of the mutant protein migrated as slower migrating, high molecular-weight (MW) species. Please note that the CN-PAGE is reported to give larger apparent molecular weight for membrane proteins compared to the soluble molecular weight markers, due to bound detergent micelles, and we could not deduce the size of the observed species [26]. Considering that thermal denaturation of proteins results in a reduced monodisperse band/peak [40, 41], a plausible explanation would be that the slower migrating band is possibly arisen from oligomerization due to partial denaturation and the faster/slower band ratio reflects the stability and/or folding status of WT and the I1061T mutant. Consistent with this idea, we observed a clear increase of the faster band upon treatment of the mutant with mo56HC or itraconazole (Fig 4C). Lapatinib also exhibited a similar effect, albeit less potent than mo56HC or itraconazole. These data suggest that itraconazole, posaconazole, and lapatinib increase the stability of the mutant protein, probably by enhancing the folding efficiency, in a similar manner to mo56HC. Digoxin was again confirmed to reduce the NPC1-GFP expression level, but a slight increase in the faster/slower ratio was observed, which may imply slightly improved proteostasis of the NPC1 mutant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Effect of itraconazole and lapatinib on the hydrodynamic status on native-PAGE gels and on the ubiquitination status.

(A) High-resolution clear native PAGE (fluorescence-detectable native PAGE) analysis of NPC1-GFP from cells treated as indicated. Cells stably expressing the indicated NPC1-GFP were treated with vehicle, mo56HC (3 μM), itraconazole (3 μM), or lapatinib (10 μM) for 21 h, and DDM-solubilized lysates were subjected to native-PAGE analysis after normalization with respect to total protein concentration. The native-PAGE gel was imaged for GFP fluorescence (upper panel), and total protein was also visualized by TCE staining (lower panel). (B) Ubiquitination status of NPC1-GFPs upon treatment with itraconazole or lapatinib. Cells stably expressing the indicated NPC1-GFP were treated as shown for 18 h followed by treatment with CB5083 (3 μM) for 6 h. The ubiquitination status was first probed with anti-Ub antibody after immunoprecipitation of the FLAG-NPC1-GFP, and then the amount of the FLAG-NPC1-GFP was re-probed with anti-FLAG antibody. The extent of ubiquitination was quantified and normalized to the amount of detected FLAG-NPC1-GFP. The raw, uncropped blot images are available from Mendeley Data repository (http://dx.doi.org/10.17632/jr23ccpp46.3). (C) Quantification of the high-resolution clear native PAGE analysis from three independent experiments. The filled circles, error bars, and open circles represent mean, SEM, and raw data points from three independent experiments. Statistical significance was assessed by ANOVA and Tukey-Kramer multiple comparison test (***, p<0.001). The raw, uncropped gel images are available from Mendeley Data repository (http://dx.doi.org/10.17632/jr23ccpp46.3). (D) Quantification of the ubiquitination assay data collected from three independent experiments. Data was normalized with the ubiquitination level of WT for each experiment. The filled circles, error bars, and open circles represent mean, SEM, and raw data points. Statistical significance of the treatment was assessed similarly as in (C). *, p<0.05; p<0.1; ns, not significant (p = 0.24).

https://doi.org/10.1371/journal.pone.0243746.g004

Notably, itraconazole treatment resulted in a monomer band of higher electrophoretic mobility. As this effect was not observed when itraconazole was added after NPC1 solubilization, the difference is likely due to some perturbation of NPC1 protein synthesis. The most likely explanation would be the previously reported inhibitory effect of itraconazole on N-glycoside processing [42].

Itraconazole and lapatinib reduce ubiquitination of the NPC1-I1061T mutant

Next, we examined the effect of itraconazole on the ubiquitination status of NPC1-I1061T. The I1061T mutant protein is primarily degraded through the ubiquitin-proteasome pathway [13, 14], and chaperone compounds should reduce ubiquitination of the mutant by helping it to escape the ER quality control machinery. So, we examined whether treatment with the hit compounds could reduce the extent of ubiquitination of the I1061T mutant. Consistent with the stabilizing effect of itraconazole, lapatinib, and mo56HC, these compounds also clearly reduced the ubiquitination level relative to the I1061T protein level (Fig 4B and 4D).

Design & synthesis of chemical probes based on the hit compounds

So far, our data supported the idea that itraconazole and lapatinib serve as pharmacological chaperones for NPC1-I1061T. As pharmacological chaperones exert their stabilizing effect via direct binding to the mutant protein, we next aimed to confirm direct binding of the compounds to NPC1-I1061T by means of photoaffinity crosslinking. For this purpose, we designed and synthesized photo-crosslinking probes based on itraconazole and lapatinib (Fig 5A; see S1 Appendix for details of the synthesis of the probes). We confirmed that the synthesized probes retained their activity towards the NPC1 mutant (Fig 5B). ItraAZY, a minimally modified photo-crosslinking probe with aryl azide as the photoreactive group and an alkyne moiety as a click chemistry handle [20, 43], was found to fully retain its activity towards the NPC1 mutant. In contrast, ItraACT, which was designed using the recently developed photo-crosslinker aryl-5-carboxytetrazole [44], was completely inactive. We also designed a pull-down probe, itraBIO, for pull-down binding assay [45], but this probe was barely active. Unfortunately, lapaAZY, a photo-crosslinking analogue of lapatinib, showed no detectable activity in our assays, possibly in part due to the weak activity of the parent compound itself. So, we focused on itraAZY for the following experiments.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. Direct binding of itraconazole to NPC1-I1061T mutant protein.

(A) Structures of photoaffinity probes and pull-down probes for itraconazole and lapatinib. (B) Chaperone activity of the probes on NPC1-I1061T mutant protein, examined by image-based chaperone assay. (C) Photoaffinity labeling of FLAG-tagged NPC1-I1061T-GFP or NPC1-P691S/I1061T-GFP with itraAZY probe. Membrane fractions were prepared from the cells stably expressing the indicated NPC1-GFP, and photo-crosslinking was performed in the presence or absence of the indicated competitors. After conjugation of biotin to the alkyne via click chemistry, the NPC1 was immunoprecipitated and the presence of biotin was probed with streptavidin-HRP conjugate, followed by re-probing with anti-FLAG antibody. The raw, uncropped blot images are available from Mendeley Data repository (http://dx.doi.org/10.17632/jr23ccpp46.3). (D) Quantification of the labeling. The extent of labeling was quantified and normalized with respect to the amount of the FLAG-NPC1-GFPs. Data is from independent experiments (n = 4 for I1061T and n = 3 for P691S/I1061T), and expressed as relative labeling respect to I1061T labeled with itraAZY. The filled circles, open circles, and error bars represent mean, raw data, and SEM, respectively. Statistical significance was assessed by the exact Wilcoxon rank sum test with correction of multiple comparisons by Benjamini-Hochberg method (*, p = 0.05; ns, not significant).

https://doi.org/10.1371/journal.pone.0243746.g005

Itraconazole directly binds to NPC1-I1061T competitively with respect to oxysterol-based chaperones

Using the highly active photo-crosslinking probe itraAZY, we performed photoaffinity labeling of FLAG-tagged NPC1-I1061T-GFP according to a protocol similar to the one we previously described [20]. Briefly, membrane preparation from HEK293 cells stably expressing FLAG-NPC1-I1061T-GFP was treated with itraAZY (and competitors) and irradiated with UV for 20 minute on ice. The irradiated membrane was solubilized, and a biotin-azide was introduced onto the alkyne moiety of the itraAZY via copper-catalyzed azide-alkyne cycloaddition reaction [30, 46]. After immunoprecipitation with anti-FLAG beads, itraAZY-labeled NPC1 mutant was quantitated. As shown in Fig 5C and 5D, itraAZY crosslinked the NPC1-I1061T mutant in a UV-irradiation-dependent manner, and competition with excess itraconazole or mo56CFA attenuated the crosslinking. This result indicates that itraconazole directly binds to NPC1-I1061T, in accordance with the reported interaction of itraconazole or posaconazole with WT NPC1 [47, 48], at the same site, or an adjacent or overlapping site, as the sterols. These data support our hypothesis that itraconazole directly binds to NPC1-I1061T, acting as a pharmacological chaperone to enhance the folding and stability of the mutant protein.

Binding of itraconazole and oxysterol-based chaperones is sensitive to P691S mutation in the sterol-sensing domain

A mutation in the sterol-sensing domain (SSD), P691S (Fig 1A), interferes with the interaction of NPC1 with small molecules, including cholesterol [49], oxysterol-based chaperones [50], U18666A [51], posaconazole [48], and itraconazole [47]. Consistent with those reports, the extent of labeling with itraAZY was significantly reduced in P691S/I1061T double mutant, implying that P691S mutation directly or indirectly interferes with itraconazole binding to NPC1-I1061T (Fig 5C and 5D), as seen with the sterols.

We previously showed that the oxysterol-based chaperones exert their effect through sterol-binding site(s) other than the N-terminal domain (NTD), a well-characterized sterol-binding domain of NPC1 [52–55]. To see if the azoles also act independently of NTD, we tested the chaperone activity of itraAZY on NTD-deleted NPC1-GFP harboring the I1061T mutation (ΔNTD-I1061T) [20] or P691S/I1061T double mutations (ΔNTD-P691S/I1061T) [50]. For ΔNTD-I1061T, both the oxysterol-derived chaperone (mo56AZK) and the azole (itraAZY) clearly corrected the ER-localization to LAMP1-positive LE/L localization (see Fig 6A for representative images and Fig 6B for colocalization analysis). This result indicates that NTD is not required for the chaperone activity, again implying that the oxysterol-derived chaperone and the azole work in similar ways. Consistent with the sensitivity of itraAZY binding to the P691S mutation, the ΔNTD-P691S/I1061T mutant was refractory to both mo56AZK and itraAZY, confirming the importance of the intact sterol-sensing domain.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. Chaperone effect of mo56HC and itraconazole is not affected by deletion of NTD but is abrogated by P691S mutation.

(A) Colocalization of ΔNTD-NPC1-I1061T-GFP and LAMP1. Cells stably expressing ΔNTD-NPC1-I1061T-GFP were treated as indicated (1 μM mo56AZK or 3 μM itraAZY for 21 h). After LAMP1 immunostaining, images were acquired on a confocal microscope FV-3000 (Olympus) equipped with x60 objective lens. Scale bar, 50 μm. (B) Quantitative colocalization analysis of ΔNTD-NPC1-I1061T-GFP and LAMP1. The filled black circles and error bars denote mean ± SD from 10 images, and gray points represent raw correlation coefficients obtained for each image. Statistical significance was evaluated by applying the Kruskal-Wallis rank sum test along with Dunn’s multiple comparison with the Benjamini-Hochberg-adjusted p value presented as *p<0.05 and **p<0.01 (two-sided). (C) Colocalization of ΔNTD-NPC1-P691S/I1061T-GFP and LAMP1, performed as in (A). (D) Quantitative colocalization analysis of ΔNTD-NPC1-P691S/I1061T-GFP and LAMP1 performed as in (B).

https://doi.org/10.1371/journal.pone.0243746.g006

Effect of itraconazole on cholesterol accumulation in NPC patient-derived fibroblasts

Finally, we examined whether itraconazole could reduce the accumulation of free cholesterol in the LE/L compartment of fibroblasts harboring I1061T mutation on both alleles, by means of filipin staining (Fig 7) [56]. As previously demonstrated [20], mo56HC treatment reduced cholesterol accumulation at 1 μM, while at 10 μM apparent inhibition of the NPC1 protein was observed; such a biphasic response is commonly observed with pharmacological chaperones [18, 57]. Itraconazole treatment, however, reduced cholesterol accumulation much less efficiently than mo56HC treatment, and only a slight reduction was only observed at 10 μM. Thus, although itraconazole was slightly more potent than mo56HC in correcting the folding defect of the I1061T mutant, it seemed to be less efficient for the functional rescue of NPC1-I1061T, for reasons that are currently unclear.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Effect of itraconazole on cholesterol accumulation in NPC fibroblasts bearing I1061T mutation.

NPC fibroblasts (GM18453, NPC1^{I1061T/I1061T}) were treated with the indicated compounds for 48 h and cholesterol accumulation was evaluated by means of filipin staining. The extent of cholesterol accumulation was quantified as fractional filipin intensity within filipin-positive puncta over total filipin staining intensity within cells. Statistical significance was evaluated by applying Dunnett’s multiple comparison test (two-sided) with adjusted p value (single-step method) presented as *p<0.05, **p<0.01, and ***p<0.001 versus vehicle, after confirming normality and variance homogeneity by means of the Shapiro-Wilk test and Bartlett test. Error bars represent means ± SD from n = 3 (for LBH589) or n = 4 (vehicle, mo56HC, and itraconazole) independent experiments.

https://doi.org/10.1371/journal.pone.0243746.g007