Inhibition of Histone Deacetylases 1, 2, and 3 Enhances Clearance of Cholesterol Accumulation in Niemann-Pick C1 Fibroblasts.

Niemann-Pick disease type C1 (NPC1) is a rare genetic cholesterol storage disorder caused by mutations in the NPC1 gene. Mutations in this transmembrane late endosome protein lead to loss of normal cholesterol efflux from late endosomes and lysosomes. It has been shown that broad spectrum histone deacetylase inhibitors (HDACi's) such as Vorinostat correct the cholesterol accumulation phenotype in the majority of NPC1 mutants tested in cultured cells. In order to determine the optimal specificity for HDACi correction of the mutant NPC1s, we screened 76 HDACi's of varying specificity. We tested the ability of these HDACi's to correct the excess accumulation of cholesterol in patient fibroblast cells that homozygously express NPC1 I1061T , the most common mutation. We determined that inhibition of HDACs 1, 2, and 3 is important for correcting the defect, and combined inhibition of all three is needed to achieve the greatest effect, suggesting a need for multiple effects of the HDACi treatments. Identifying the specific HDACs involved in the process of regulating cholesterol trafficking in NPC1 will help to focus the search for more specific druggable targets.

Introduction

Niemann-Pick disease type C (NPC) is a rare autosomal recessive lysosomal storage disorder. Analysis of multiple exosome sequencing databases predicts an incidence rate of 1:15 000 to 1:20 000 when including late-onset forms of the disease.[1] New screening initiatives and assays will soon improve our understanding of the prevalence of this disease.[2] The characteristics of NPC disease include accumulation of unesterified cholesterol and other lipids in the late endosomes and lysosomes (LE/Ly) of several tissues, with the most serious effects in the central nervous system (CNS). Abnormal lipid accumulation in peripheral organs also contributes to the pathology of this disease.[3] A large fraction of patients experience progressive neurological degeneration and often die before the age of 20.[4−7] Mutations in two genes, NPC1 and NPC2, are the cause of NPC disease, and NPC1 mutations are responsible for approximately 95% of cases.[8−10]

NPC1 is a multipass transmembrane protein in the limiting membrane of the LE/Ly. NPC1 has an N-terminal cholesterol-binding domain and a sterol-sensing domain associated with its transmembrane domains.[11−15] NPC2 is a small, soluble LE/Ly protein that binds cholesterol.[16] NPC2 binds to NPC1 and delivers cholesterol to the N-terminal cholesterol binding domain of NPC1.[13,14,17−19] Recent studies have indicated that NPC1 functions as a sterol tunnel, moving cholesterol past the glycocalyx of the lysosome to the limiting membrane for extraction by sterol transport proteins.[20−22]

More than 650 variants of NPC1 have been described, with over 250 of these variants having a known or likely pathogenic nature (Niemann-Pick Type C Database: https://medgen.medizin.uni-tuebingen.de/NPC-db2/index.php).[3,23−25] The most common mutant allele found in patients is NPC1 (14–18% of disease alleles[26,27]). Cells with this mutation express low levels of the mature protein due to poor folding in the endoplasmic reticulum (ER), followed by ER associated degradation (ERAD).[28] High levels of overexpression of the NPC1I1061T protein rescue the cholesterol storage phenotype,[28] suggesting that the mutant protein is functional if a sufficient amount is delivered to the LE/Ly.

Treatment of NPC patients with Miglustat (Zavesca), an approved treatment in many countries but not in the US,[29,30] slows disease progression by inhibiting glycosphingolipid synthesis. This reduces the production of some lipids that accumulate along with cholesterol in LE/Ly of NPC mutant cells. A clinical trial treatment for NPC has been based on intrathecal injection of VTS-270, a formulation of 2-hydroxypropyl β-cyclodextrin (https://clinicaltrials.gov/ct2/show/NCT02534844).[31] β-Cyclodextrins act by binding cholesterol in LE/Ly, bypassing the requirement for NPC1 and NPC2, to shuttle cholesterol out of LE/Ly.[32−35] Hydroxypropyl β-cyclodextrin treatment has been effective in animal studies,[36−42] and initial human data were promising.[43,44] However, hydroxypropyl β-cyclodextrin does not cross the blood brain barrier (BBB),[45] requiring direct infusion into the CNS. Clinical studies using intrathecal injection of VTS-270, a pharmaceutical hydroxypropyl β-cyclodextrin, have indicated slowing of disease progression.[46,47]

There remains a need for effective NPC disease drug therapies, preferably with an easy and safe route of administration. Valproic acid, a weak histone deacetylase inhibitor (HDACi), had been shown to have some benefit in NPC1–/– neural stem cells.[48] We identified more potent and specific HDAC inhibitors as a class of molecules that reversed the NPC1 mutant phenotype in fibroblasts from NPC1 patients.[49] Treatment with broad spectrum HDACi's led to increased expression of the mutant NPC1 protein in patient fibroblasts expressing one or two NPC1 alleles. Specifically, treatment with a broad spectrum HDACi, Vorinostat (SAHA), reduced ERAD of the NPC1I1061T protein, which was correctly delivered to LE/Ly.[50] We also tested the effectiveness of Vorinostat and Panobinostat (Farydak, LBH589) in U2OS-SRA cells engineered to express 1 of 60 different NPC1 mutations found in patients. In these cells, treatment with the HDACi corrected the phenotype for over 80% of the tested mutations.[50] However, mutations in regions required for NPC1 function, such as the N-terminal cholesterol binding domain or the lumenal domain to which NPC2 binds, remained uncorrected with HDACi treatment.

Some studies of HDACi's in mouse models expressing mutant Npc1 have used SAHA (Vorinostat), a HDACi that has relatively poor BBB penetration.[51−53] One study in Npc1 mice, which express a missense D1005G mutation in the Npc1 gene, showed improved liver function and reduced liver pathology. Additionally, there were changes in transcriptional regulation of cholesterol and apolipoprotein B homeostasis consistent with improved cholesterol trafficking. However, SAHA treatment of Npc1 mice at a maximum tolerated dose did not ameliorate disease progression, weight loss or animal morbidity. The very low levels of SAHA found in brain tissue may explain this lack of effect.[53] A study in Npc1 mice reported that a combination therapy with Vorinostat, hydroxypropyl β-cyclodextrin, and polyethylene glycol 400 led to slowed neuronal degeneration and improved lifespan in Npc1 mutant animals.[54] However, in a very similar study using NPC1I1061T mice, the hydroxypropyl β-cyclodextrin component alone proved effective,[55] suggesting that this may have been the case in the earlier study as well. Thus, although HDACi's have been able to rescue most NPC1 mutants in cell culture, there is not yet evidence of a benefit in animal models. It remains possible that other HDACi's would provide a benefit.

Other recent papers have shown effectiveness of HDACi treatments in cells.[56,57] HDACi treatment has also been shown to correct the lysosomal enzyme defect in fibroblasts expressing a common mutation found in Gaucher disease, which is also associated with ERAD of the mutant protein and alterations of molecular chaperone expression.[58] HDACi treatment also has been reported to rescue several CFTR mutations associated with cystic fibrosis[59] as well as Huntington’s and Alzheimer’s disease.[60,61] These results suggest that improving proper folding and exit of mutants out of the endoplasmic reticulum can provide benefit for several inherited diseases.

The most common function of a HDAC enzyme is to catalyze the deacetylation of acetyl-l-lysine side chains of histones and other proteins. HDACs are broadly divided into five classes: Class I (HDACs 1, 2, 3, and 8), IIa (HDACs 4, 5, 7, and 9), IIb (HDACs 6 and 10, which have two HDAC domains), III (sirtuins), and IV (HDAC 11, which shares properties with classes I and II).[62] For each class, the region of homology is the conserved deacetylase domain (reviewed in ref (63)). However, HDACs 4, 5, 7, and 9 have little or no deacetylase activity.[64] The inhibition of HDACs affect NPC disease because they modulate multiple biological processes.[65] HDACs interact with their targeted acetylated proteins and, as members of multiprotein complexes, can alter gene expression and change chromatin structure. Many of the complexes formed include multiple HDACs, indicating that more than one member of an HDAC class may be needed to achieve the desired genetic response. They also influence protein folding through their effect on chaperones and the proteasome by controlling the ubiquitination of lysines.[63]

Our previous work indicated that a series of broad spectrum HDAC inhibitors that are either U.S. Food and Drug Administration (FDA) approved or have been investigated in clinical studies (SAHA, TSA, CI-994, and LBH589) were effective in reducing cholesterol in NPC1 mutant fibroblasts while a HDAC 8 selective inhibitor did not have an effect.[49] Treatment with valproic acid, a weak Class I HDAC inhibitor,[66] corrected cholesterol accumulation in NPC1 fibroblasts and improved lysosomal targeting of NPC1.[56] The dose-limiting toxicity of these HDACi's is thought to be due to the broad-spectrum inhibition, and more selective HDACi's have been proposed to be less toxic.[67] Selective HDACi's that discriminate among these HDACs are now available. We sought to determine if any of the more selective HDACi compounds available to us would optimize restoration of cholesterol trafficking in NPC1 mutant fibroblasts. A more selective compound might not only reduce the adverse side effects associated with HDACi therapy[67,68] but also provide insights into the mechanism of action of HDACi's in NPC by clarifying which HDAC isoform or combination thereof is responsible for the observed effect. We thus examined the effect of 76 known and novel HDACi's with different inhibitor profiles on NPC1 patient-derived fibroblasts. As we will show, inhibition of the combination of HDAC 1, 2, and 3 rather than a narrower subset leads to the best reduction of cholesterol storage in NPC1 mutant fibroblasts. We also confirm that inhibition of HDACs 6 and 8 is not required.

Results and Discussion

As a model for NPC1 disease, we use NPC1I1061T patient-derived fibroblast cells (Coriell, GM18453). These cells have unesterified (free) cholesterol accumulated in lysosome-like storage organelles (LSOs) that accumulate near the center of the cells. To label free cholesterol in the cells, we use filipin, a fluorescent dye that binds unesterified cholesterol.[69] As we have shown previously,[50] treatment of GM18453 cells with 10 μM SAHA for 3 days greatly reduces the filipin-labeled cholesterol in the cells (Supplemental Figure 1A). To quantify this, we measure the fluorescence power in the bright areas using a threshold and then divide this fluorescence power by the total area occupied by cells in each field. We call this the LSO ratio, and we normalize it for each experiment to the ratio value for DMSO-treated cells (solvent control). We found that 10 μM SAHA, which we use herein as a reference treatment, was nearly as effective as 100 μM hydroxypropyl β-cyclodextrin with 3 days treatment for each.[32,49]

Broad Spectrum HDAC Inhibitors Correct NPC1 Cholesterol Phenotype

Using this filipin assay, we have shown that broad spectrum HDACi's such as SAHA, trichostatin A (TSA), or LBH589 correct the cholesterol accumulation phenotype of several fibroblast cell lines containing a variety of NPC1 mutations.[49] In agreement with previous results,[49] several of the broad spectrum HDACi's tested corrected the cholesterol accumulation phenotype of NPC1I1061T fibroblast cells (Supplemental Figure 1). A 3 day treatment with SAHA (Vorinostat), Chidamide,[70] or AR42[71] efficiently corrected the cholesterol accumulation phenotype of NPC1I1061T fibroblasts (Supplemental Figure 1B). Broad spectrum HDACi's such as LBH589 increase cholesterol esterification as a consequence of release of cholesterol from LSOs,[49] and CI-994 also increases lipid droplet formation in treated cells (Supplemental Figure 1D).

We used SAHA as the reference compound in all subsequent compound screens. SAHA is an effective inhibitor of HDACs 1, 2, 3, and 6 and, to a lesser extent, HDAC 8 (Table ).[49] To determine which HDAC isoform plays the most important role in the correction of cholesterol accumulation in NPC1 fibroblast cells, we tested 76 compounds with a range of potency and selectivity for individual HDACs or groups of HDACs.

Table 2

IC50’s and ED50’s for HDAC 1, 2, and 3 Specific Inhibitors (Figure A Compounds)a

Color coded boxes indicate the specificity of the compound. Green borders, low IC50 values (μM); and red, no specific inhibitory activity. Superscripts indicate the method for IC50 value determination. IC50 values provided by Reaction Biology Corporation (method a) or KDAc Therapeutics (method c). Mean ED50’s in μM (and ranges) for reduction of cholesterol accumulation for each compound.

The broad spectrum HDAC inhibitor TSA has been shown to correct the cholesterol accumulation phenotype in NPC1 patient fibroblasts.[49] The R-enantiomer of TSA is a broad range inhibitor that, like SAHA, inhibits multiple HDACs. The S-enantiomer of TSA is much more specific for HDAC 6 (Table ).[72] Class I HDACs HDAC 1–3 and class II HDAC 10 were all strongly inhibited by (R)-TSA but only weakly by (S)-TSA, whereas other HDACs, including HDAC 6, were inhibited about equally or with only small differences (Table ). Thus, although (R)-TSA is a broad spectrum HDAC inhibitor, (S)-TSA, the unnatural enantiomer, had in vitro moderate selectivity for HDAC 6 (∼20-fold lower IC50 compared to the next closest isoform, HDAC 1). Treating NPC1I1061T patient-derived fibroblasts with the separate enantiomers demonstrated that (R)-TSA corrected the cholesterol accumulation phenotype much more effectively than (S)-TSA (Welch’s ANOVA, P < 0.0001 and P = 0.23, respectively) (Figure ). The racemic TSA is less potent than (R)-TSA at lower doses but still significantly increases cholesterol clearance with the maximal dosage (P < 0.0001, Figure ).

Table 1

IC50’s and ED50’s of TSA Enantiomersa

Broad spectrum HDAC inhibitors have varying specificity for individual HDACs. IC50 values (μM) for each HDAC indicated are shown in color-coded boxes. Green border boxes indicate low IC50 values and high efficacy for the indicated HDAC. The IC50’s for these compounds were measured by Reaction Biology Corporation (method a). Mean ED50s in μM (and ranges) for reduction of cholesterol accumulation for each compound.

Figure 1

Enantiomers of TSA demonstrate the importance of HDAC 1, 2, and 3 inhibition for cholesterol clearance and that inhibition of HDACs 6 and 8 is not necessary. Racemic TSA was separated into R- and S-enantiomers. NPC1I1061T fibroblasts were treated with these enantiomers for 72 h at concentrations of 4 nM to 1.10 μM. R-TSA produced by University of Notre Dame (UND; green line) or racemic TSA (produced by UND) (blue line) specifically inhibits HDACs 1, 2, 3, 6, and 8 (see Table ). Both dose-dependently increased cholesterol clearance from treated NPC1I1061T fibroblasts. S-TSA produced by UND (red line) more specifically inhibits HDACs 6 and 8 (see Table ) and did not increase cholesterol clearance in treated NPC1I1061T fibroblasts. N = 3 independent cell culture experiments. Bars are SEM ****, P < 0.0001; n.s., not significant.

To categorize the compounds studied, they were binned in three classes according to their IC50’s. An IC50 value below 2 μM was considered to be an effective inhibitor of the target HDAC (green borders in Tables –4). An IC50 between 2 and 10 μM (yellow) was considered low inhibitory activity, while a compound was considered inactive if the IC50 was above 10 μM (red). The IC50 values are from published sources when available or the provider of the compounds. (See Methods section for details on compound providers a–c as indicated by superscripts in the tables.)

Table 4

IC50’s for Inhibitors of HDACs 1, 2, or 3 (Figure Compounds)a

Color coded box borders indicate the specificity of the compound. Green, low IC50 values (μM); yellow, residual inhibitory activity; and red, no specific inhibitory activity. IC50 values provided by Dana-Farber for UND (method b) or KDAc therapeutics (method c).

Testing HDAC Inhibitor Isoform Specificity For Correction of NPC1 Cholesterol Phenotype

To test the hypothesis that inhibition of only HDACs 1–3 was responsible for increased cholesterol clearance, we used two different approaches. In the first approach, we investigated inhibitors that were specific for HDACs 1, 2, and 3 (Table ) but not HDAC 6 or 8. Most of these HDAC 1, 2, and 3 inhibitors corrected the cholesterol accumulation phenotype, and the dose–response curves of the compounds are shown in Figure . Several of the compounds tested that have low IC50 values for HDAC 1, 2, and 3 led to cholesterol clearance comparable that for to SAHA (CI-994, P = 0.0007; K-1,2,3-S, P = 0.0042; K-1,2,3-V, P = 0.0033; Figure A and Supplemental Figure 2).

Figure 2

Inhibitors specific to HDACs 1, 2, and 3 increased cholesterol clearance in NPC1I1061T fibroblasts. (A) Inhibitors of HDACs 1, 2, and 3 dose-dependently increase cholesterol clearance similar to SAHA. N = 3 independent cell culture experiments. Bars are SEM. ***, P < 0.001; **, P < 0.005. (B) Compounds that do not inhibit HDACs (negative controls) do not increase cholesterol clearance in treated cells. N = 3 independent cell culture experiments. Bars are SEM. (C) Structures of HDAC inhibitors and negative controls.

Metal Chelation is Required for HDACi Cholesterol Clearance

Several HDAC inhibitors are also active against a variety of other targets. For example, valproic acid is a HDACi that was found to correct NPC1 defects in cells,[48,56] but it also inhibits a number of other targets.[73] Furthermore, nonmetal chelating binding modes of TSA have been found in crystal structures.[74] In our second approach, we tested the importance of metal chelation in the HDAC active site for increased cholesterol clearance. We designed and tested three different HDACi analogues that do not contain a metal-chelating moiety and therefore do not inhibit any HDACs (Table ), but are otherwise structurally very similar, as negative controls. For the first, the hydroxamic acid, which binds zinc in the HDAC catalytic site,[71] is replaced by a much weaker metal binder in suberoylanilide carboxylic acid (SACA, see structures in Figure C). As a complementary negative control of benzamide-type HDAC inhibitors, we designed N-NC-A, which is a protected precursor of N-1,2-C (Figure A,C) in which the benzamide group required for zinc binding is blocked. To exclude the possibility that modification of the zinc binding moiety would lead to a negative result for reasons other than the disruption of metal binding, we designed BRD4097,[75] a modified form of CI-994 that leaves the zinc binding moiety intact but disrupts binding through steric repulsion via an additional methyl group. None of these compounds led to cholesterol clearance in treated NPC1I1061T fibroblasts (Figure B), strongly suggesting that metal chelation by HDACi's is required for increased cholesterol clearance.

Table 3

Specificity of Negative Control Compounds for HDAC Inhibitors (Figure B Compounds)a

IC50 values (μM) for each HDAC indicated are shown in color-coded boxes (red for no inhibitory activity against listed HDAC). IC50 values provided by KDAc Therapeutics (method c) and Reaction Biology Corporation (method a).

Figure 3

Inhibitors specific for individual HDACs (1, 2, or 3) do not substantially increase cholesterol clearance from NPC1I1061T fibroblasts. (A) Compounds that more specifically inhibit HDAC1 do not increase cholesterol clearance in treated NPC1I1061T fibroblasts. N = 3 independent cell culture experiments. Bars are SEM. *P < 0.05; n.s., not significant. (B) Compounds that inhibit HDACs 1 and 2 only partially increase cholesterol clearance in treated NPC1I1061T fibroblasts. N = 3 independent cell culture experiments. Bars are SEM. (C) Compounds that inhibit HDAC3 alone do not significantly increase cholesterol clearance in treated NPC1I1061T fibroblasts. N = 3 independent cell culture experiments. Bars are SEM. (D) Structures of HDACi compounds in (A)–(C).

Testing Individual HDAC Isoforms

We studied the question of whether one HDAC isoform was solely responsible for the observed effect. While there are isoform selective inhibitors available that inhibit HDAC 1 and 2 over HDAC 3, HDACs 1 and 2 are 83% identical[76] and the sequence differences are far removed from the active site. Therefore, compounds that inhibit HDAC 1 also inhibit HDAC 2 with some level of potency (e.g., Tables and 4). We therefore used an approach in which we tested compounds with varying specificity for the individual HDACs. The results were correlated with the overall effect on cholesterol clearance to determine if inhibition of one HDAC alone could result in the cholesterol clearance response. The HDAC inhibitor that most specifically targeted HDAC 1 (BRD3227,[77] see Table ) did not correct the cholesterol accumulation phenotype in NPC1I1061T fibroblast cells (Figure A). Compound N-1,2-C, which had a lower IC50 for inhibiting HDAC3 (Table ) was also less effective at reducing cholesterol storage in the NPC1I1061T fibroblasts (Figure A). We included cells treated with SAHA to confirm that the cells were responding to treatment (Figure B).

Compounds (N-1,2-A and N-1,2-D) with very weak activity against HDAC 3 compared to HDAC 1 and 2 did not lead to cholesterol clearance. These compounds arose from a computationally designed modification of a previously reported HDAC 1 and 2 inhibitor that exploits the mutation of a serine at the bottom of the 14 Å side pocket adjacent to the active site in HDAC 1/2 to a tyrosine in HDAC 3 (Supplemental Figure 3).[78,79] The interactions of BRD3227 in the active site of HDAC1/2 vs 3 have been discussed elsewhere[80] and served as the inspiration for the computational design of N-1,2,-C and related compounds. The basic design idea for these compounds is to position hydrogen bond donors and acceptors into the 14 Å pocket[78] that is different between HDAC1/2 and 3 to achieve selectivity. Suitable modification of the benzamide was predicted to alter the HDAC 3 activity while maintaining the HDAC 1/2 activity in closely related analogs. If only HDAC 1/2 activity would be relevant for the observed phenotype, then these two compounds should be very similar in their efficacy. Conversely, significant differences in the observed phenotype correction should be due to the differences in HDAC 3 activity. Treatment with the inhibitors more specific to HDAC 3 (Figure C), K-3-A and K-3-B, resulted in minimal cholesterol clearance. Overall, these compounds failed to show the robust correction that we observed with inhibition of HDACs 1, 2, and 3.

NPC1 Cholesterol Phenotype Correction Requires Inhibition of HDACs 1, 2, and 3

The effects of specificity when inhibiting HDACs 1–3 for reversal of the cholesterol storage phenotype of NPC1 fibroblasts are illustrated in a 3D graph in which the ED50 for lowering cholesterol (i.e., the dose at which it led to 50% clearance of cholesterol in NPC1I1061T fibroblasts) is color coded and plotted versus the log of the IC50 value for each of the three HDACs (Figure ). Violet represents the lowest ED50 values (0.15 μM), and red represents the highest ED50 values (10 μM). In general, the compounds with the lowest ED50 values (blue and violet) are clustered at the lowest IC50 values for all three HDACs. This 3D correlation plot is rotated to focus on each HDAC individually and then oriented to show all three with equal emphasis. This plot demonstrates the relationship between specificity for the three HDACs and the ability of each compound to exert a corrective effect on cholesterol clearance within NPC1I1061T fibroblasts. The very high similarity of HDACs 1 and 2 is reflected in the better correlation between the specificities for these two enzymes and the concentrations at which cholesterol is cleared from the cells. Although the correlation is less clear for HDAC 3 than for HDACs 1 and 2, the clearance of cholesterol depends on inhibition of all three enzymes. HDAC 3 is a member of the same family as HDACs 1 and 2 (Class I), but it has less similarity to 1 and 2 as they do to each other.[62]

Figure 4

Correlation between HDACi specificity (IC50) and dosage (ED50) for cholesterol clearance with compounds that inhibit HDACs 1, 2, and 3. (A) HDAC1 focused view. The (log) IC50 values for HDAC1 on the front facing axis show that points of similar ED50’s (color-coded) cluster in separate halves (reds to left, blue/purple to right). The ED50 values range from 0.15 μM (purple) to 10 μM (red). (B) HDAC2 focused view. The (log) IC50 values for HDAC2 on the front facing axis show a pattern that is a mirror image of the HDAC1 focused view. (C) HDAC3 focused view. The (log) IC50 values for HDAC3 on the front facing axis show an ED50 cluster pattern that is less distinct with red/pink clustered at the top while blue/purple is on the bottom. (D) Equal weighted plot for all three HDACs. The specificity of the inhibitor for HDACs 1, 2, and 3 is correlated with the experimentally determined ED50.

We found that the specificity of the inhibitor for HDACs 1, 2, and 3 is correlated with the concentration at which it leads to correction of the cholesterol accumulation phenotype in NPC1I1061T fibroblasts. This finding overlaps with findings in other disease models including Alzheimer’s disease and cystic fibrosis.[60,61,80,81]

Broad spectrum HDAC inhibitors and compounds that inhibit HDACs 1, 2, and 3 increase cholesterol clearance in NPC1I1061T patient derived fibroblasts (Coriell, GM18453). Although broad spectrum HDACi's such as SAHA have been FDA approved for a range of diseases, and SAHA has been evaluated in a clinical trial as a potential treatment for NPC, it has undesirable side effects that have been attributed to its limited selectivity.[82−84] The therapeutic and/or toxic effects seen in clinical trials are thought to be due to a lack of HDAC isoform-specific inhibition, and the development of more selective inhibitors is predicted to result in better clinical outcomes.[67,68] Determining which HDACs are important for this process would avoid unnecessary side effects resulting from the inhibition of HDACs not involved in cholesterol clearance. Furthermore, the mechanism by which SAHA leads to improved cholesterol clearance is not well understood. This is similar to many other epigenetic treatments where the downstream effects of the inhibition of the epigenetic target are unclear. We did find that the broad spectrum HDAC inhibitory activity of SAHA was necessary for improving the cholesterol clearance in NPC1 fibroblasts, similar to results seen with valproic acid.[56] It is therefore important to determine which HDAC isoform(s) are responsible for the effect observed in the cellular assay.

While the broad spectrum HDAC inhibitors SAHA, Chidamide, AR42, and TSA all show correction of the cholesterol phenotype in NPC1I1061T fibroblast cells, we aimed to identify specific HDACs as the necessary targets for inhibition. By testing a library of compounds with different HDAC isoform selectivity profiles, we have identified that it is necessary to simultaneously inhibit HDACs 1, 2, and 3 together to increase cholesterol clearance from NPC1I1061T fibroblasts. The combined inhibition of 1, 2, and 3 explains all the observed effects of broad spectrum HDAC inhibitors.

It was not surprising that inhibition of both HDACs 1 and 2 was necessary in our studies. They appear to have overlapping functions in complexes and are 83% identical.[76] Interestingly, we found that HDAC 3, which has some significant differences to HDAC 1/2 that can be exploited for isoform selectivity,[85] must also be inhibited to see the optimal effect on cholesterol clearance. This result can be rationalized by the fact that HDACs, like many epigenetic modulators, exert their biological effect in a variety of protein–protein complexes. For example, HDACs 1 and 2 can exist as homodimers or heterodimers within different complexes.[76] Inhibition of HDACs 1–3 has recently been shown to downregulate transcription of core regulatory transcription factors and lead to more open chromatin architecture,[80] potentially allowing for transcription of other genes involved in the trafficking of cholesterol. HDACi's selected for cancer treatments reflect changes in transcription within rapidly growing cells.[86] In some diseases, HDACi treatment alters proteostasis due to activation of the chaperone-mediated refolding of pathway specific proteins.[87] However, the cellular basis by which HDACi's rescue the activity of mutated NPC1 is not yet known. Identifying which HDACs play the largest part in the process is a crucial step forward.

We have identified the inhibition of HDACs 1, 2, and 3 as necessary for the clearance of excess cholesterol accumulation in the lysosomes of NPC cells. In contrast, we found that it was not necessary to inhibit HDACs 6, 8, and 11, which are also targeted by the broad spectrum HDAC inhibitors previously shown to correct the phenotype. Based on a recent study in NPC1I1061T mice,[55] the HDAC inhibitors tested have not been effective treatments for NPC1 disease in mouse models. However, the identification of newer HDACi's and the pathways and genes regulated by HDACs 1, 2, and 3 that yield phenotypic improvements in NPC1 mutant cells may offer new targets for improving cholesterol trafficking in NPC1 patients. Effective treatments will also require efficient entry into the CNS, and only some of the HDACi's we used have been tested for this action.[66]

Materials and Methods

GM18453 cell line, NPC1I1061T patient-derived fibroblast cells from Coriell Institute for Medical Research (RRID:CVCL_DA78). 384 well plates, Corning #3712 (2018). Chidamide, Cayman #13686 (RRID:SCR_008945). FBS, LifeTechnologies #10437028 (2018). DMSO, Sigma #34869 (2018), dried with Fluka Chemika, Molecular Sieves #69839 (2018). Filipin, Sigma #F9765 (2018). DRAQ5, ThermoFisher #DR51000/62252 (2018). Modified Eagle Medium (MEM), ThermoFisher #11095098 (2018). LipidTox Red, Life Technologies #H34476 (2021).

Compound Synthesis

Procedures for synthesis of the Notre Dame compounds are provided in the Supporting Information. The compounds provided for screening by KDAc are extensively described in US patent US9365498B2.[88] Custom made materials will be shared upon reasonable request.

Software

ChemDraw 16.0, PerkinElmer, Waltham MA.

Microsoft Excel, Office 365, Redmond WA.

MetaMorph and MetaXpress software, Molecular Devices LLC, San Jose, CA.

Prism 8.0, GraphPad Software, San Diego, CA.

PyMOL Molecular Graphics System, ver. 2.0, Schrödinger, LLC.

SigmaPlot 14.0, Systat Software, San Jose, CA.

Equipment

AiryScan Confocal microscope, Zeiss.

ImageXpress Micro Imager (automated wide-field fluorescence microscope), MultiDrop384 Dispenser, and BioTek Plate Washer.

Methods

No ethical approval was required for this study. To test HDAC inhibitors for their ability to correct the cholesterol accumulation phenotype of NPC1 patient derived fibroblast cells (GM18453), we used a high-throughput filipin assay, previously described in detail.[89] Briefly, GM18453 fibroblast cells homozygously express the I1061T mutant form for the NPC1 protein. These cells were grown in Modified Eagle Medium with 5.5% FBS on a 384-well plate and maintained at 37 °C in a humidified 5% CO2 incubator. On the second day of plating, the cells were treated with (dry) DMSO (control) or with a HDAC inhibitor compound dissolved in (dry) DMSO at concentrations ranging from 40 nM to 10 μM for 72 h. Following treatment, the cells were fixed with 3% paraformaldehyde (PFA) for 20 min, stained with 50 μg/mL filipin for 45 min, followed by a nuclear stain of 2 μM DRAQ5. All staining was done at room temperature. Images of the cells were acquired on an ImageXpress automated wide field microscope at 10× magnification using UV and CY5 filters. Each compound concentration was tested in triplicate on each plate. The results plotted are the average of three separate cell culture experiments. Filipin stains the outer membrane of the cell and the “lysosomal storage organelles” (LSOs). The LSO ratio value is the filipin intensity in the lysosomal storage organelles above the intensity of the cell membrane. DMSO-treated control cells are included in each plate, and the average value from these cells is used to set the “normal value” (1.0) for each plate. The normalized LSO ratio plotted for each treatment condition represents the LSO ratio value for the treated cells divided by the DMSO-treated cells in the same plate, allowing a comparison to be made for experiments performed on separate days. All data are plotted as mean ± standard error measurement (SEM). GM18453 fibroblast cells were treated with DMSO or 10 μM CI-994 for 48 h. After treatment, cells were stained with LipidTox Red for 30 min (1:2000). Cells were imaged with an AiryScan confocal microscope using a 40× oil immersion objective (1.3 NA) and a 561 laser line. Total fluorescence intensities in lipid droplets per image were measured with the granularity feature in the MetaXpress software. Data are shown from three independent experiments; error bars are ± SEM.

Methods for IC50 Value Determination

Method a: Reaction Biology Corp. IC50 Profiling Service (reactionbiology.com)

Compounds were tested by Reaction Biology Corp. in singlet 10 dose IC50 mode with 3-fold serial dilution starting at 100 μM against HDACs 1, 2, 3, 6, and 8. The substrate for HDACs 1, 2, 3, and 6 was a fluorogenic peptide from p53 residues 379–382 (RHKK(Ac)AMC). The substrate for HDAC8 was a fluorogenic peptide from p53 residues 379–382 (RHK(Ac)K(Ac)AMC). The IC50 values were calculated using the GraphPad Prism 4 program based on the sigmoidal dose–response equation. The blank (DMSO) value was entered as 1.00 × 10–12 of concentration for curve fitting.

Method b: Biochemical HDAC assay for UND Produced Compounds (Performed at Dana-Farber Cancer Institute)

HDAC activity was determined using an in vitro optimized assay performed in 384-well plates as previously described.[90−92] Briefly, each compound was incubated with recombinant full-length HDAC protein (BPS Biosciences, San Diego, CA) for 3 h prior to enzymatic reactions. Reactions were initiated by the addition of fluorophore-conjugated substrates, MAZ1600 and MAZ1675, at a concentration equal to the substrate KM for each individual HDAC (MAZ1600: 9 μM for HDAC 1, 10 μM for HDAC 2, 8 μM for HDAC 3; MAZ1675: 202 μM for HDAC 8). The assay was carried out in assay buffer (50 mM HEPES, 100 mM KCl, 0.001% (v/v) Tween 20, 0.05 (w/v) bovine serum albumin, pH 7.4) and followed for fluorogenic release of 7-amino-4-methylcoumarin from substrate upon deacetylase and trypsin enzymatic activity. Fluorescent measurements were obtained every 5 min using a multilabel plate reader and plate stacker (Envision; PerkinElmer, Waltham, MA). The first derivative of the fluorescent intensity over time within the linear range of the data was taken (Spotfire DecisionSite) and imported into analytical software (GraphPad Prism) for analysis. Replicate experimental data (n = 4) from incubations with inhibitor were normalized to DMSO ([DMSO] < 5%). IC50 values were determined by logistic regression.

Method c: KDAc/BRD

Purified HDACs were incubated with a FAM-labeled fluorescent substrate. For HDACs 1–3, compounds were preincubated for 3 h; all other HDACs had no preincubation and were tested at room temperature for 1 h. The fluorescence intensity of electrophoretically separated substrate and product was measured, and the percent inhibition was plotted against the compound concentration. The IC50 values were determined by curve fitting with Origin 8.0 software.[77]

Racemic TSA Synthesis and Chiral Separation of (R)-TSA and (S)-TSA

Racemic TSA was obtained via synthetic methods described in.[93] Racemic TSA was purified using a Waters XBridge Prep C18 5 μm OBD column (19 × 50 mm) and a water/acetonitrile gradient (10–90% ACN in water in 10 min, pH = 7; 14 min run, flow rate = 20 mL/min; racemic TSA RT = 7.0 min). With analytically pure racemic TSA in hand, a method was developed for the chiral separation. Racemic TSA was determined to be separable using a Daicel ChiralPAK AD-H 5 μm column (amylose tris(3,5-dimethylphenylcarbamate)-coated 5 μm silica gel, 4.6 × 250 mm) and heptane/isopropanol gradient (10–90% IPA in heptane over 15 min, no additives; total 25 min run, flow rate = 1.0 mL/min). (S)-TSA was eluted at 14.64 min. (R)-TSA was eluted at RT = 15.77 min. These samples were collected, and comparison of (R)-TSA to the commercially available (R)-TSA was performed on the same chromatography column. Commercially available (R)-TSA was confirmed to have >95% enantiomeric purity under the same mobile phase conditions ((S)-TSA is not commercially available at the time of this report). Co-injection during chiral HPLC of commercially available (R)-TSA and the isolated (R)-TSA sample derived from the separated racemic mixture confirmed that a single enantiomer was isolated and matched known material. 1H NMR and LCMS data conform to previously reported structural data in ref (66). Commercially available (R)-TSA was purchased from TCI AMERICA (Lot No. CZU7I-GC) for use in HPLC analysis.

Statistics

One-way Welch’s ANOVA or Mann–Whitney, as indicated in figures, using GraphPad Prism 8.0 software.