TCGA data mining

TCGA data was analyzed using “TCGAbiolinks” Bioconductor package [65] and GEPIA2 (Gene Expression Profiling Interactive Analysis) web server [66]. REST mRNA expression data (in TPM format, transcripts per million) were compared in TCGA-LGG (low-grade glioma) dataset, TCGA-GBM (glioblastoma) dataset, and matching normal samples from TCGA and GTEx databases via GEPIA2. Survival analysis was performed using the same tool. For prediction of REST-target genes, RNA-Seq data (FPKM-UQ format) was downloaded from the open-access part of the TCGA database for 169 GBM patient samples (Primary Solid Tumor/Recurrent Solid Tumor) on 2/13/2021 using “TCGAbiolinks” R package. After standard pre-processing and filtering of low-signal mRNAs across all samples, pair-wise correlation coefficients between REST expression and expression of every transcript from the remainder (n = 42,335) were calculated. Then, the coefficients were ranked in the order of increasing magnitude, and top genes negatively correlated with REST were analyzed in follow-up studies.

Cell culturing and cell treatments

Human embryonic kidney cells (HEK293/HEK293T), human liver cancer cells (HepG2), human fetal glial cells SVGp12, and glioblastoma cell lines (A172, T98G) were purchased from ATCC (Manassas, VA). U251 glioblastoma cells were purchased from Sigma-Aldrich (St. Louis, MO). HEK293, SVGp12, and GBM cells were routinely cultured in minimal essential media (MEM, Sigma-Aldrich) supplemented with 10% Opti-Gold fetal bovine serum, FBS (GenDEPOT, Katy, TX), 1 mM sodium pyruvate (Sigma-Aldrich), and 1% non-essential amino acids (Sigma-Aldrich) at 37 °C in a humidified 5% CO2 atmosphere. HepG2 was cultured in MEM supplemented with 10% FBS. HEK293T cells, CRISPR-edited cell lines, and corresponding CRISPR control cells were cultured in Dulbecco’s modified Eagle’s media (DMEM, Sigma-Aldrich), supplemented with 10% FBS. HyClone penicillin and streptomycin (P/S) mix (Cytiva, Marlborough, MA), at a final concentration of 1%, was added to all media. Prior to the experiments, routinely cultured cell lines were confirmed mycoplasma free by the Mycoplasma qPCR kit (Minerva Biolabs, Skillman, NJ).

Novel small molecule SCP1 inhibitors (T-65, GR-28) were synthesized in Dr. Dionicio Siegel’s lab. Triacsin C/TrC was purchased from Tocris (#2472) or Cayman Chemicals (#10007448). Perhexiline maleate was purchased from MedChemExpress (HY-B1334A). Stock solutions of all the compounds were prepared in DMSO and were stored at − 80°C, avoiding multiple freeze–thaw cycles.

For determination of cytotoxicity, cells at a final density of 8000/well (100 µL) were seeded in black 96-well plates in their corresponding complete media and treated with compound(s) of interest or solvent-control on the next day. For Western blots or RNA extraction, cells were seeded in complete media in flasks and also treated the next day. Wild-type cell treatments were performed for a specified time duration (18, 24, 36, 48, or 72 h) in MEM supplemented with 5% FBS, 1 mM sodium pyruvate, 1% NEAA, and 1% P/S. CRISPR-edited cells and their corresponding controls were treated in DMEM supplemented with 5% FBS and 1% P/S. Drug mixtures with SCP1 inhibitors were replenished daily when treatment duration exceeded 24 h, based on time-course assessment of REST protein recovery under treatment with compounds of this class [14]. DMSO concentrations in the incubation mixtures or solvent-control mixtures never exceeded 0.5% (v/v).

Design and synthesis of SCP1 inhibitors

The syntheses of a large number of analogues with variable targeted covalent functionality has been possible given the design of our route and generality of the reaction used for the syntheses. Amide coupling reactions with the noncovalent amide region were needed for targeting and provided access to improved compounds. Testing of a large number of reactive warheads with structurally diverse, reactively unique, and positional isomers was used to find the optimal reactive group—benzothiophene-1,1-dioxide [67]. The syntheses followed a similar path as that used for GR-28.

Synthetic procedures

To a stirring solution of 2'-methoxy-[1,1'-biphenyl]-4-carboxylic acid (200 mg, 0.88 µmol, 1.0 equiv) and tert-butyl (3-aminopropyl) carbamate (198 mg, 1.14 mmol, 1.3 equiv) in DMF (5 mL), HATU (666 mg, 1.75 mmol, 2 equiv) and DIPEA (340 mg, 2.63 mmol, 3 equiv) were added, and the reaction was stirred at 23 °C for 14 h. The reaction was then diluted with ethyl acetate (25 mL) and the organic phase was washed with brine (4 × 25 mL). The organics phase was dried over Na2SO4, concentrated, and purified by column chromatography (1:5 EtOAc:Hexanes) to yield tert-butyl (3-(2'-methoxy-[1,1'-biphenyl]-4-carboxamido)propyl)carbamate (302 mg, 90%).

1H NMR (600 MHz, CDCl3) δ 7.88 (d, J = 7.9 Hz, 2H), 7.60 (d, J = 8.2 Hz, 2H), 7.34 (dd, J = 17.3, 8.4 Hz, 2H), 7.22 (s, 1H), 7.04 (t, J = 7.4 Hz, 1H), 7.00 (d, J = 8.2 Hz, 1H), 4.98 (s, 1H), 3.81 (s, 3H), 3.53 (dd, J = 12.2, 6.1 Hz, 2H), 3.36–3.18 (m, 2H), 1.72 (s, 2H), 1.46 (s, 9H).

HRMS: m/z: calcd for C22H28N2O4: 385.2122; found 385.2120.

To a stirring solution of tert-butyl (3-(2'-methoxy-[1,1'-biphenyl]-4-carboxamido)propyl) carbamate (270 mg, 702 µmol, 1 equiv) in MeOH (3 mL), concentrated aqueous HCl (15 mL) was added. The solution was stirred at 23 °C for 15 min. Completion of reaction was monitored by TLC and which was subsequently concentrated in vaccuo to afford 3-(2'-methoxy-[1,1'-biphenyl]-4-carboxamido)propan-1-aminium chloride which was used directly for the next reaction. The solid was dissolved in DMF (2 mL) and benzo[b]thiophene-2-carboxylic acid (162 mg, 912 µmol, 1.3 equiv) was added followed by HATU (533 mg, 1.4 mmol, 2 equiv) and DIPEA (272 mg, 2.1 mmol, 3 equiv). The reaction was stirred at 23°C for 14 h and diluted with ethyl acetate (15 mL). The mixture was washed with brine (4 × 15 mL). The organic extract was dried over Na2SO4, concentrated under vacuum, and purified by column chromatography (1:1 EtOAc: Hexanes) to yield N-(3-(2'-methoxy-[1,1'-biphenyl]-4-carboxamido)propyl) benzo[b]thiophene-2-carboxamide (260 mg, 82%). Rf = 0.5 (silica gel, 0:1 hexanes: EtOAc).

1H NMR (600 MHz, CDCl3) δ 7.93–7.89 (m, 3H), 7.85 (t, J = 8.3 Hz, 2H), 7.62 (d, J = 8.1 Hz, 2H), 7.59 (t, J = 6.0 Hz, 1H), 7.44–7.33 (m, 3H), 7.31 (d, J = 7.4 Hz, 1H), 7.09 (t, J = 6.1 Hz, 1H), 7.05 (t, J = 7.4 Hz, 1H), 7.00 (d, J = 8.2 Hz, 1H), 3.81 (s, 3H), 3.62 (dd, J = 11.8, 6.2 Hz, 2H), 3.57 (dd, J = 11.7, 6.1 Hz, 2H), 1.88–1.82 (m, 2H).

HRMS: m/z: calcd for C26H24N2O3S: 445.1580; found 445.1578.

To a vigorously stirred solution of N-(3-(2'-methoxy-[1,1'-biphenyl]-4-carboxamido)propyl) benzo[b]thiophene-2-carboxamide (30 mg, 67 µmol, 1 equiv) in dichloromethane (15 mL), acetone (5 mL) and aqueous, saturated sodium bicarbonate (100 mL) solution were added. Eight portions of oxone (total amount 8.00 g, 13.0 mmol, 190 equiv) were added to this in 5-min intervals. Upon completion, by TLC, the reaction is diluted with water (50 mL) and dichloromethane (20 mL). The layers were separated and the aqueous layer was extracted with dichloromethane (20 mL). The organics were combined, washed with brine (40 mL), dried with MgSO4, filtered, and concentrated in vaccuo. Crude reaction mixture was purified by column chromatography (1:99 MeOH:DCM) to afford N-(3-(2'-methoxy-[1,1'-biphenyl]-4-carboxamido)propyl) cinnamamide 1,1-dioxide (23 mg, 72%). Rf = 0.3 (silica gel, 9.5:0.5 DCM:MeOH).

1H NMR (600 MHz, MeOD) δ 7.93 (s, 1H), 7.87 (d, J = 8.3 Hz, 2H), 7.80–7.77 (m, 1H), 7.73–7.69 (m, 2H), 7.66 (m, 1H), 7.58 (d, J = 8.2 Hz, 2H), 7.35 (t, J = 7.8 Hz, 1H), 7.30 (d, J = 7.5 Hz, 1H), 7.09 (d, J = 8.3 Hz, 1H), 7.03 (t, J = 7.5 Hz, 1H), 3.81 (s, 3H), 3.50 (m, 4H), 1.93 (m, 2H).

HRMS: m/z: calcd for C26H24N2O5S: 477.1479; found 477.1483.

Full synthetic scheme for the synthesis of GR-28 Kinetic characterization of GR-series compounds

To determine the potency of GR compounds in inhibiting SCP1 phosphatase activity towards the analog substrate pNPP, we followed the protocols and methods described in [68,69,70,71]. The enzyme concentration was set at 150 nM for SCP1. To study the inhibitory effect, inhibitors were preincubated for varying time durations ranging from 30 min to 23 h at room temperature (RT), and the concentration of DMSO was normalized to 1% in the final reaction volume. In the control group, which did not contain the compound, DMSO was added at a concentration of 1%. The reaction time was set to 3 min at 37 °C. The activity of each phosphatase towards pNPP in the presence or absence of the inhibitor was measured in an assay buffer (50 mM Tris–acetate pH 7.6, 10 mM MgCl2, 0.02% Triton X-100, and 1% DMSO). The amount of released pNP was quantified by measuring the absorbance at 410 nm. The kinact and KI values were calculated using KaleidaGraph software.

Malachite green assay

Initially, SCP1 was preincubated with either DMSO (as a control) or GR28 (20 µM) for 6/18 h. Subsequently, these samples underwent a 20-min incubation at 37 °C with the different concentrations of physiological substrate, REST phospho-peptide containing phosphorylated serine at position 861. Finally, the reaction was terminated by adding 40 μl of malachite green reagent (Biomol, Hamburg, Germany) to each 20 μl sample. This addition resulted in a green color, the intensity of which is directly proportional to the amount of inorganic phosphate released from pREST during the assay.

MALDI-TOF analysis of covalent adducts

SCP1 WT (50 μM) and SCP1 C181A mutant in activity buffer at pH 7.6 (with 5 μM or 500 μM BME) were treated with 500 μM of GR-28 (final 1% DMSO), and control samples were treated with 1% DMSO. The samples were incubated overnight at RT. The pNPP activity was tested the next day by taking a sample from each tube. The samples were then desalted using Ziptip C18 resins (Sigma-Aldrich) following standard protocols. Mass spectrometric analysis of SCP1 treated with GR-28 or DMSO was performed using an Auto-flex Max MALDI-TOF (Bruker Corporation, Billerica, MA) with a 1:1 DHB matrix (ThermoFisher, Waltham, MA).

Molecular docking

The model of GR28 compound was built using MAESTRO v. 13.5.128 from the Schrödinger suite. The compound was positioned manually into the active site of SCP1 in PyMOL v. 2.4.1 (PDB Code: 3PGL). Energy minimization of the protein and the manually positioned GR28 compound was performed in MAESTRO by applying the OPLS_2005 force field. The final model was visualized in PyMOL.

Establishment of CRISPR/Cas9-REST-KO cell lines

To express REST-KO sgRNAs, two pairs of DNA oligos were synthesized: REST-RY1F (caccgGTTATGGCCACCCAGGTAAT) and REST-RY1R (aaacATTACCTGGGTGGCCATAACC); REST-RG6F (caccgGTCTTCTGAGAACTTGAGTA) and REST-RG6R (aaacTACTCAAGTTCTCAGAAGACC) [17]. Annealed double-stranded sgRNAs were cloned into pX330 plasmid [72], and the correct clones were verified by sequencing using U6 promoter primer. Non-neural HEK293 cells and glioblastoma cells T98G were co-transfected with two sgRNA expression vectors (RY1, RG6) and Cas9-2A-GFP plasmid [72] using FugeneR HD (Promega, Madison, WI) transfection reagent according to manufacturer’s guidelines. As a CRISPR-recombination control, we used cells co-transfected with empty pX330 vector and Cas9-2A-GFP plasmid. At 48 h post-transfection, single GFP-expressing cells were sorted using MA900 cell sorter (Sony Biotechnology, San Jose, CA) into 96-well plates (one cell per well) with complete DMEM media for clone expansion. Western blotting was used for screening REST-KO single-cell clones.

Genotyping of CRISPR/Cas9 repair outcomes

We used Sanger sequencing to identify REST protein sequence in REST-KO clones after double-nicking CRISPR/Cas9 recombination. Briefly, we extracted genomic DNA from target cells using Monarch kit (NEB, Ipswich, MA). Next, we designed primers flanking the regions of sgRNA-guided double-stranded breaks (RY1, RG6) and performed PCR reactions with subsequent Sanger sequencing of PCR products where applicable (Additional File 9: Table S8). For PCR, we used GC buffer and Phusion DNA-polymerase (ThermoFisher) and dNTP mixture from NEB. PCR reactions were run on MasterCycler nexus (Eppendorf, Enfield, CT), and resulting agarose gels were visualized using GelDoc XR + imager (Bio-Rad, Hercules, CA). PCR products were purified using PCR/gel purification kits following manufacturer’s instructions (Qiagen, Germany) and submitted for sequencing with original primer sequences (RY1-F, RY1-R, RG6-F, RG6-R). Sequences were translated into protein using Expasy online tool (www.expasy.org).

REST transient overexpression

For the rescue experiments, REST-null glioblastoma cells were seeded at 800,000 cells per t-25 flask. On the next day, cells were transiently transfected with 500 ng either REST-WT-expressing pLPC-vector (Addgene, #41903) or empty pLPC vector (Addgene, #12521) with Fugene HD transfection reagent (Promega) following manufacturer’s instructions. At 24 h post-transfection, cells were collected and used for proliferation assay and Western blotting. For the rescue experiments on wild-type glioblastoma cells, transient transfection was performed in 6-well plates with 50 ng of REST-pLPC-vector or empty pLPC-vector per well.

Western blots

Briefly, cells were lysed in RIPA buffer (10 mM Tris–HCl, pH 7.5, 150 mM NaCl, 1% sodium deoxycholate, 0.1% SDS, 1% Triton x-100, 5 mM EDTA) supplemented with 1 × Halt protease and phosphatase inhibitor cocktail (ThermoFisher) followed by centrifugation for 15 min at 12,000 rpm. Protein concentration in supernatants was measured with BCA assay. Typically, proteins (50 µg) were separated by Novex 4–12% Tris–Glycine gels (Invitrogen, Waltham, MA), transferred to PVDF membranes (Bio-Rad), followed by membrane blocking at room temperature for 1 h in 1% Tween-20-TBS buffer containing 5% BSA (bovine serum albumin, ThermoFisher) or non-fat milk. Membranes were incubated at 4 °C overnight with primary rabbit antibodies against REST (#22242–1-AP, Proteintech, Rosemont, IL) at 1:500 dilution, SCP1 (#ab136038, Abcam, Cambridge, MA) at 1:500 dilution, or β-tubulin (#ab6046, Abcam) at 1:4000 dilution. On the next day, membranes were washed and incubated with 1:15,000 diluted goat anti-rabbit secondary IRDye 680RD antibody (LI-COR, Lincoln, NE) for 1 h at room temperature. After washing, membranes were visualized on LI-COR Odyssey CLx image reader.

RNA isolation, library preparation, and Taq-Sequencing

Total RNA was isolated from cells using DirectZol RNA Miniprep kit (Zymo Research, Irvine, CA, product number #R2050). 3’ Tag-Seq was performed by the University of Texas Genomic Sequencing and Analysis Facility, based on the protocols from Lohman et al. [26] and Meyer et al. [73]. Libraries were quantified using the Quant-it PicoGreen dsDNA assay (ThermoFisher) and pooled equally for subsequent size selection at 350–550 bp on a 2% gel using the Blue Pippin (Sage Science, Beverly, MA). The final pools were checked for size and quality with the Bioanalyzer High Sensitivity DNA Kit (Agilent, Santa Clara, CA) and their concentrations were measured using the KAPA SYBR Fast qPCR kit (Roche, Basel, Switzerland). Samples were then sequenced on the NovaSeq 6000 (Illumina, San Diego, CA) instrument with single-end, 100-bp reads.

Tag-Seq data analysis

Quality of raw reads was assessed using FastQC read quality reports (https://usegalaxy.org [74]). Adaptor trimming, deduplicating, and quality filtering were performed using a published pipeline (https://github.com/z0on/tag-based_RNAseq). Next, trimmed reads were aligned to human reference genome, GRCh38 version, using HISAT2 fast aligner v.2.2.1 [75] with default parameters, except Forward (F) –rna-strandedness. Gencode v38 gtf file was used as annotation gtf. Lastly, mapped fragments were quantified by featureCounts v.2.0.1 [76] in Galaxy. Differential expression was analyzed using DESeq2 v.1.30.1 [77] in R; genes with adjusted p-value < 0.05 and FC cutoff of 1.5 were considered as differentially expressed. Tag-Seq data was deposited in Gene Expression Omnibus/GEO under the accession number GSE234912. GO-enrichment analysis of gene clusters was performed using Bioconductor R package “clusterProfiler” v.3.18.1 [78] and STRING v.12 [79]. For every gene network based on protein–protein interactions (PPI), PPI enrichment p-value or FDR was recorded. For correlation with PDT (proliferation doubling time), raw counts were converted to FPKM in R.

qPCR

Total RNA was isolated from cells using either DirectZol RNA Miniprep kit (Zymo Research) or Trizol reagent with subsequent isopropanol precipitation; 0.5 µg of total RNA was used for reverse transcription using the AzuraQuant™ cDNA Synthesis Kit, #AZ-1995 (Azura, Raynham, MA, USA) using manufacturer’s guidelines. Relative gene expression was measured using AzuraQuant™ Green Fast qPCR Mix, Lo-Rox (Azura) and normalized to ACTB gene expression. Amplification was performed using the ViiA 7 Real-Time PCR System (Applied Biosystems, Waltham, MA). Specificity of amplification was controlled with melting curves/primer efficiency calculation. Analysis of qPCR data was performed using the ∆∆Ct method. Primer sequences (designed to span exon-exon junctions or to be separated by a relatively large intron) and qPCR conditions are shown in Additional File 9: Table S8.

Cytotoxicity assays (resazurin reduction) and drug combination landscapes

After treatment (72 h), 20 µl 0.15 mg/ml resazurin solution (ThermoFisher) was added to each well of a 96-well plate. After 3 h incubation, fluorescence was recorded using a 560-nm excitation/590-nm emission filter set on an Infinite F200 microplate reader (Tecan, Switzerland) [80]. Cell viability was normalized to that of control wells after background subtraction. LD50s of selected compounds were fitted based on cell survival data using the “drc” (dose–response curves) R package [81]. To assess synergetic effects of GR-28 with Triacsin C, we built 5 × 5 drug combination landscapes using the Bioconductor package “synergyfinder” and its Bliss model [54]. For every landscape, cells were seeded in plates and treated the next day with serial twofold dilutions of drug(s), and the maximal dose of single drugs corresponded to 40–50% or higher mean viability. HepG2 cells were treated at the doses corresponding to the A172 cell line which was more sensitive to SCP1 inhibitors, HEK293 were treated at T98G doses. Maximal synergy coefficients were extracted from each landscape and recorded.

Migration (wound scratch) assay

On day 0, glioblastoma cells were seeded at a density of 400,000 per well in complete media in 6-well plates. On the next day, after reaching 70–80% confluency, cell monolayer was scraped in a straight line with a pipette tip. After scratch, monolayer was gently washed with 1 × PBS to remove detached cells, and media was replenished. Images were taken using EVOS FL fluorescence microscope (Invitrogen) at 4 × magnification 24 h and 48 h post-wound. Images were analyzed using ImageJ software and the Wound_healing_size_tool plugin [82].

Cell proliferation assay

For the proliferation assay, cells were seeded at a density of 50,000 cells per well in complete media in 24-well plates. Then, cells were counted every 24 h for four subsequent days using Trypan Blue exclusion assay (0.4%) on automated Luna-II automated cell counter (Logos Biosystems, Annandale, VA). Population doubling time (PDT) was estimated with the following formula, PDT = (72 h × ln2) / ln(N4/N1), where N1 and N4 are cell counts in every well on first and fourth days, respectively.

Statistical analyses

Statistical analyses were performed using RStudio v.4.0.5 and GraphPad Prism v.9.5. One-tailed or two-tailed, unpaired or paired (where applicable) t-test was used for comparing two groups. ANOVA was used when comparing several groups vs control. p < 0.05 values were considered as significant. Correlations were assessed using two-tailed Pearson r coefficients. Protein bands were quantified and compared using ImageJ software. Illustrations were created using BioRender software. The statistical details of experiments can be found in the figure legends.