Targeting foam cell formation in inflammatory brain diseases by the histone modifier MS-275

Patients and healthy volunteers

Peripheral blood samples were drawn from 35 healthy volunteers (age 24–61 years, median = 37 years) and from four X-ALD patients with the default disease variant adrenomyeloneuropathy (AMN; age 25–44 years, median = 38 years, Table S1). At the time of blood donation, none of the AMN patients revealed any signs of cerebral involvement on MRI. VLCFA accumulation in plasma and leukocytes of AMN patients was confirmed by measuring the total amount of the fatty acids C26:0, C24:0, and C22:0 by GC-MS or in macrophages by LC-MS/MS as described previously.^{19, 24} The in vitro study of human primary macrophages was approved by the Ethical Committee of the Medical University of Vienna (EK1462/2014), and informed consent was obtained from participating AMN patients and healthy volunteers. Histochemistry was performed using postmortem CNS tissue. Details of the patients’ characteristics and conditions have been summarized previously.⁸ Use of this material was approved by EK729/2010 and EK535/2016.

Isolation of human monocytes

Human CD14⁺ monocytes were isolated from blood by magnetic-activated cell sorting as described previously.¹⁹

Flow cytometry

The purity of isolated CD14⁺ monocytes was determined by flow cytometry as described previously.⁸ To analyze the number of pHrodo Green-positive macrophages, the cells were detached, washed and re-suspended in 250 µL phosphate-buffered saline (PBS) before analysis. Viable macrophages were gated using forward and side scatter. Geometric mean of fluorescence intensity (gMFI) was determined from 5000 macrophages per sample and using a GuavaEasyCyte™6-2L system flow cytometer (Luminex) and guavaSoft3.3 software.

In vitro differentiation of human monocytes to macrophages and activation with lipopolysaccharide (LPS)

CD14⁺ monocytes (1 × 10⁶ cells/well) were cultured in RPMI medium (Sigma Aldrich) containing 1% Penicillin/Streptomycin, 1% glutamine, 1% Fungizone and 10% FCS, supplemented with 50 ng/mL human recombinant M-CSF (PeproTech) for 7 days. For use in β-oxidation assays, M-CSF-differentiated macrophages were polarized with 100 ng/mL IL-4 (Novartis) for 2 days in the presence of 10 ng/mL M-CSF. The HDAC inhibitors SAHA (Cat.No.10009929, Cayman Chemicals) and MS-275 (Cat.No.T6233, Target Mol) were added during the last 48 h. For analysis of pro-inflammatory cytokine gene expression, M-CSF-differentiated macrophages were stimulated with 100 ng/mL LPS (E. coli 055:B5, Cat.no. L4005, Sigma) and treated with DMSO, MS-275, or SAHA in concentrations as indicated for 24 h. For detachment, adherent macrophages were washed with PBS and incubated with 300 µL Gibco™TrypLE™Select(10×) (Gibco, Life Technologies) for 15 min at 37°C, and collected with a cell scraper after adding 300 µL PBS.

Generation of foam cells and determination of neutral lipids and total cholesterol

Human blood-derived monocytes were differentiated using 50 ng/mL M-CSF in the presence of 20 µg/mL myelin for 7 days during differentiation to generate myelin-laden foam cells and the viability was determined by flow cytometry using the fluorescent dye 7-AAD detecting dead cells (Fig. S3).

Then, foam cells were incubated with DMSO or MS-275 as indicated for 24 h before harvesting for RNA analysis of gene expression or 48 h for assays of neutral lipids and total cholesterol content. The myelin was isolated from wild-type C57BL6/J mouse brains and the purity was determined by western blotting with specific antibodies as described and shown in detail previously.⁸ Endotoxin levels were determined by Genscript ToxinSensorTM Chromogenic LAL Endotoxin Assay Kit and were far below the threshold (<0.5 EU/mL) for medicinal products. Part of the myelin was labeled with the amine-reactive pH-sensitive dye pHrodo^®GreenSTP ester (Thermo Fisher Scientific, Cat.No.:P35369) according to the manufacturer’s instructions as described previously.⁸

Neutral lipids were visualized in the cultured cells after fixation in 4% paraformaldehyde by Oil Red-O (C.I. 26125, Merck, Certistain^®, ORO) staining. For analysis of cellular cholesterol content, the harvested cells were washed with PBS before lipid extraction by sonication on ice in chloroform-methanol (v/v 2:1). After centrifugation, the organic phase was transferred to a fresh tube and air dried at 50°C. The dried lipids/cholesterol were re-dissolved in reaction buffer (Component E) of the Amplex Red Cholesterol Assay Kit (Molecular Probes, Invitrogen/Thermo Fisher) and total cholesterol was measured according to the manufacturer’s instructions.

Immunohistochemical staining of macrophages/microglial cells in active demyelinating lesions of postmortem CNS tissue

Paraffin-embedded postmortem tissue containing active demyelinating brain lesions was available from four CALD cases and six MS cases as described previously⁸ and further outlined in Table S2. Immunohistochemical staining for the pro-inflammatory macrophage/microglia marker CD86 was performed as described before.⁸ To determine the size (cross-sectional area in µm²) of macrophages/microglia in active demyelinating lesions sections were steamed in citrate buffer (pH 6.0) for 1 h and stained using the primary rabbit antibodies against p22-phox (Santa Cruz #sc20781) diluted 1:100 in 10% FCS/Dako buffer and applied over night at 4°C. A secondary biotinylated anti-rabbit antibody (Jackson#711-065-152, diluted 1:1000) was applied, followed by avidin-conjugated horse-radish peroxidase (Jackson#016-030-084, diluted 1:500) for 1 h, both at room temperature. Sections were developed with 3,3′-diaminobenzidine and counterstained with hematoxylin. To quantify the cross-sectional area of p22phox positive macrophages/microglia, the software Image J was used. In addition, histochemical myelin staining was performed using a routine Luxol fast blue-periodic acid Schiff (LFB-PAS) protocol,²⁵ visualizing myelin-derived phospholipids and digested glycolipids.

HAP 1 cells

HAP1 cells and HDAC1, 2 or 3 knockout (KO) HAP1 cell lines were obtained from Horizon Genomics (now, Horizon Discovery). Mutant HAP1 cells expressing FLAG-tagged catalytically inactive enzymes: HDAC1 H141A, HDAC2 H142A, and HDAC3 H135A were generated by CRISPR-Cas9-mediated knock-in of the expression construct under the control of the EF1a promoter into the safe harbor locus AAVS1 of the corresponding KO cell lines (V. Moos et al., manuscript in preparation).

Whole transcriptome analysis

Whole transcriptome analysis of human in vitro-generated foam cells was performed for eight healthy volunteers, and differential expression was compared between vehicle-treated and myelin-laden macrophages. Total RNA (100 ng) was used for the generation of mRNA-focused libraries. The samples were sequenced on a HiSeq2000 sequencing system (Illumina) at the Biomedical Sequencing facility of the Medical University of Vienna, Austria as described previously.⁸ After a global analysis of deregulated genes, a targeted analysis of (1) genes from cholesterol, LXR or PPAR pathways as retrieved from the Molecular Signatures Database (MSigDB) and (2) cytokine-cytokine receptor interaction pathway genes (KEGG pathways:map04060) for pro- and anti-inflammatory gene expression was performed (Dataset S1).

Whole transcriptome analysis of HAP1 cells was performed for three biological replicates each after treatment with 1 µmol/L or 3 µmol/L MS-275 for 24 h, samples were compared to vehicle (DMSO)-treated cells (n = 3). Here, we considered, only significantly changed genes (in comparison to vehicle controls) from the cholesterol, LXR or PPAR pathways, as described for foam cells. The complete dataset from these pathways is shown in Dataset S2.

RNA isolation and reverse transcription-coupled quantitative PCR (RT-qPCR)

RNA isolation and RT-qPCR analysis were carried out as previously described.^{8, 19} Each cDNA was measured in technical duplicates. Relative mRNA levels were detected by SYBRGreen incorporation and calculated by the 2^−ΔΔCq method using HPRT1 as internal reference. In addition, the absolute mRNA abundance of ABCD1, ABCD2 and for normalization, HPRT1 were obtained by the TaqMan ™ method. Sequences of primers are shown in Table S3.

Statistical analysis

For normally distributed data, we performed either two-tailed Student’s t-test or, when more than two groups were analyzed, one-way ANOVA and post hoc Fisher’s LSD test as indicated. The P-values were calculated and the null hypothesis denied for P-values < 0.05. The software GraphPad Prism 7 (GraphPad Software) was used for graphical display and statistical analysis.

Further detailed methods are described in the supplementary materials.

Enlarged foam cells in actively demyelinating CNS lesions are associated with inflammation

To assess foam cell formation in acute MS and CALD lesions, we first measured the size of enlarged macrophages/microglia in actively demyelinating lesions of postmortem brain tissue from patients using the lipid-staining LFB-PAS (Fig. 1A and B) and the immunohistochemical staining for the plasma membrane-accentuated p22phox (Fig. 1C and D). We observed comparable enlarged LFB-PAS positive phagocytes in CALD and MS and obtained similar results when assessing enlarged p22phox-positive foam cells (Fig. 1E–G). Because the smallest sizes in our data unlikely reflect the true cell size, which also depends on the plane at which the cell was cut, merely the highest 75% of the values were included in violin plots of the cell size distribution (Fig. 1E and F). In both diseases, especially enlarged foam cell sizes (3^rd quartile of the cell size) coincided with a high number of CD86-positive, pro-inflammatory macrophages/microglia (Fig. 1H–J, Fig. S2), and the presence of a phagocytic, inflammation-associated phenotype evidenced by the marker CD68 (Fig. 1K–M, Fig. S2). Together, these results show that profound myelin debris uptake by macrophages and microglial cells leads to formation of enlarged foam cells in actively demyelinating CNS lesions of both MS and CALD. In addition, especially enlarged foam cells appear to coincide with a pro-inflammatory, disease-promoting environment.

Genome-wide analysis of human myelin-laden macrophages

In contrast to previous studies,^{5, 26-30} we particularly aimed to investigate the effect of a prolonged myelin accumulation on human macrophages, and thus, performed a genome-wide transcriptome analysis of in vitro-differentiated macrophages, which were derived from healthy donors (n = 8) and exposed for 7 days to myelin debris (Fig. 2).

We found that myelin phagocytosis especially downregulated genes involved in cholesterol and fatty acid synthesis, whereas genes involved in intracellular lipid shuttle and export were increased (Fig. 2A). Specifically, myelin-loading prominently induced NR1H3 and PPARG encoding the nuclear receptors LXRα and PPARγ, respectively (Fig. 2A). Whereas some pro-inflammatory genes were significantly downregulated upon myelin ingestion, the expression of genes encoding growth and neurotrophic factors or cytokines and their cognate receptors involved in homeostatic functions was significantly upregulated in myelin-laden macrophages (Fig. 2B). Using RT-qPCR and an enlarged sample set (n = 13), we confirmed the differential expression patterns of selected key genes in lipid homeostasis. Furthermore, we analyzed the expression of the characteristic anti-inflammatory genes CCL18 and CD83, which were significantly upregulated after myelin phagocytosis (Fig. 2C–H). Together, these findings indicate that sustained myelin loading of human healthy macrophages in a non-inflammatory environment induce gene expression programs possibly enabling not only efficient lipid degradation and efflux, but also tissue regeneration.

Class I HDACs regulate lipid metabolism-associated genes in human macrophages

Next, we tested whether the specific inhibition of class I HDACs, which was previously shown to regulate the lipid metabolism in an atherosclerotic mouse model,²¹ could be beneficial to prevent foam cell formation in human macrophages. As an initial screening tool, we used the human cell line HAP1, modified for production of isogenic variants carrying mutations in selected genes.³¹ First parental HAP1 cells that had been treated with 1 or 3 µmol/L MS-275 for 24 h were compared to vehicle control in a whole transcriptome analysis. We queried these gene sets selectively for pathways associated with lipid metabolism (LXR, PPAR, and cholesterol) and found significant upregulation of key target genes (Fig. 3A and B, Dataset S2) from the PPAR and/or cholesterol pathways and to a lower extent from LXR and LXR/cholesterol-overlapping pathways. To obtain a better understanding of the contribution of the individual histone deacetylases inhibited by MS-275, we used isogenic HAP1 cell lines with either complete knock-out of HDAC1, HDAC2 or HDAC3 or the corresponding knock-in mutants producing stable but catalytically inactive variants (HDAC1 H141A, HDAC2 H142A, and HDAC3 H135A) instead of the wild-type enzymes, resulting in the integration of non-functional HDACs into co-repressor complexes. The latter strategy was applied to prevent compensatory upregulation and integration of other HDACs into co-repressor complexes, thereby more closely mimicking the pharmacological inhibition with HDAC inhibitors (Moos et al., manuscript in preparation). The RT-qPCR analysis revealed that the lack of distinct class I HDACs results in upregulation of selected key genes in macrophage lipid metabolism, like LXR-regulated cholesterol efflux transporters, ABCA1 and ABCG1 (Fig. 3C and D) and key players in lipid uptake and clearance (CD36 and FABP4) linked to PPARγ pathways (Fig. 3E and F). However, this observed altered expression might be independent of changes in acetylation patterns at specific gene regions.

Intriguingly, when comparing the dosage effects of the pan-HDAC inhibitor SAHA and the class I-specific HDAC inhibitor MS-275 on the regulation of the same lipid metabolism-associated genes in human macrophages, we found that treatment with MS-275 showed a higher potency than SAHA for upregulating these genes (Fig. 4A–E). These results indicate that in human macrophages, especially class I HDACs play an important role in regulating lipid homeostasis and, thus, constitute a potential target for prevention of foam cell formation.

The class I-specific HDAC inhibitor MS-275 reduces human foam cell formation

Based on our findings that class I HDACs regulate genes associated with lipid metabolism in human macrophages, we analyzed the efficacy of MS-275 to improve lipid homeostasis using myelin-laden human macrophages. In vitro-differentiation in the presence of myelin debris for 7 days resulted in enlarged macrophages with foam-like morphology (Fig. 5A and C), as evidenced by ORO staining for neutral lipids including cholesterol esters. Without myelin exposure the macrophages showed an elongated morphology, typical for non-activated, anti-inflammatory cells in vitro, and only scarce lipid inclusions (Fig. 5B). Treatment of the lipid-laden macrophages with MS-275 (Fig. 5D and E) significantly reduced the number of foam cells and increased the frequency of cells with an elongated shape in comparison to vehicle-treated cells (Fig. 5F). Whereas myelin uptake elevated the amount of total cellular cholesterol (Fig. 5G), mainly containing free cholesterol, treatment with MS-275 significantly lowered the intracellular total cholesterol content of macrophages upon prolonged myelin loading. Interestingly, even the basal cholesterol levels of non-loaded macrophages could be downregulated with MS-275 (Fig. 5G). To further confirm that MS-275 interferes with lipid-enriched myelin accumulation, macrophages were incubated with pHrodo Green-labeled myelin and analyzed by flow cytometry. Here, we consistently found a significant reduction of pHrodo Green-positive foam cells indicating a possible increased net efflux of myelin-associated lipids/cholesterol upon 2-day-treatment with 2.5 µmol/L MS-275 (Fig. 5H).

Next, we assessed whether the observed lowered lipid accumulation in foam cells and, hence, possibly also tissue clearance, after MS-275 treatment is mediated by the induction of genes involved in lipid uptake and export. Indeed, RT-qPCR analysis revealed that MS-275 treatment resulted in a dose-dependent upregulation of ABCG1, CD36, FABP4, and SOAT (Fig. 6A–F). In addition, protein expression of the lipid exporters ABCA1 and ABCG1 and the scavenger receptor CD36 crucial for lipid uptake were further confirmed by flow cytometry analysis (Fig. S3).

Taken together, our findings indicate that inhibition of class I HDACs by MS-275 treatment leads to reduced lipid accumulation with concurrent upregulation of genes involved in cholesterol and fatty acid transport and adoption of an anti-inflammatory morphology in human foam cells generated in vitro.

MS-275 improves impaired lipid homeostasis and immune functions in macrophages from X-ALD patients

In the pathological context of X-ALD, we explored whether class I HDAC inhibition by MS-275 would improve the metabolic and immunological phenotype of ABCD1-deficient macrophages, which accumulate high levels of VLCFAs. Firstly, we found that the selective inhibition of class I HDAC members in HAP1 cells leads to a significant upregulation of the ABCD2 gene (Fig. 7A), which upon overexpression can compensate for ABCD1 deficiency in X-ALD cells.^32-35 We recently found that pan-HDAC inhibition with Vorinostat (SAHA) stimulates ABCD2 expression in human macrophages.²⁰ Interestingly, the class I-specific inhibitor, MS-275 not only caused a greater ABCD2 induction when compared to SAHA (Fig. 7B) but also efficiently induced ABCD2 expression at low doses (Fig. 7C). When analyzing the peroxisomal β-oxidation of the VLCFA C26:0, MS-275, and SAHA stimulated the rate of VLCFA degradation to a similar extent (Fig. 7D). A 5-day-treatment with 0.5 µmol/L MS-275 or 2.5 µmol/L SAHA significantly downregulated the accumulation of C26:0-lysophosphatidylcholine (C26:0-LPC) (Fig. 7E), albeit the levels were still higher than those observed in healthy control cells.

We have previously demonstrated that the disturbed lipid homeostasis in X-ALD macrophages results in pro-inflammatory skewing and impaired plasticity⁸ and recently found that SAHA treatment partially rescued this phenotype.²⁰ Strikingly, MS-275-treatment almost completely abolished the IL12p40 secretion in LPS-stimulated X-ALD macrophages (Fig. 7F). Finally, we analyzed the effect of MS-275 on the ability of LPS-stimulated X-ALD macrophages to attract monocytes in a Boyden chamber. For this, we applied supernatants harvested from X-ALD macrophages, which had been LPS-activated in the presence of either MS-275 or vehicle. Although statistical significance was not reached, we found a trend for lower numbers of migrating monocytes with supernatants derived from pro-inflammatory X-ALD macrophages treated with MS-275 in comparison to vehicle control (Fig. 7G, P = 0.057).