Three High Throughput Compatible Cell-Based Assays for Identifying Small Molecule JAG1 Upregulators for Alagille Syndrome

Abstract

Haploinsufficiency disorders arise when loss of function mutations in one allele of a gene reduce gene dosage below the level required for normal physiology. Pharmacologic upregulation of the remaining functional allele represents a promising therapeutic strategy but requires screening assays capable of detecting modest changes in endogenous gene expression. Here we developed and compared three high throughput cell-based assays for identifying small molecule upregulators of JAG1, the gene most frequently mutated in Alagille syndrome (ALGS). The assays measure JAG1 expression at different molecular levels: RNA fluorescence in situ hybridization (RNA FISH) for JAG1 mRNA, immunofluorescence (IF) for endogenous JAG1 protein, and a HiBiT luminescence assay using CRISPR engineered LX-2 hepatic stellate cells expressing HiBiT tagged JAG1. Each assay was optimized in 384-well format and benchmarked using a panel of 32 histone deacetylase inhibitors (HDACi), compounds known to broadly increase gene expression. All three assays detected JAG1 upregulation and identified overlapping sets of active compounds. The homogeneous HiBiT assay showed the most favorable high throughput screening statistics (S/B = 2.7 and Z′ > 0.5) and the lowest well to well variability, whereas the RNA FISH and IF assays provided higher signal to basal ratios and single cell resolution. Entinostat, Mocetinostat, and Chidamide were consistently identified as the most potent JAG1 upregulators across all assays. These complementary assays provide a flexible platform for identifying small molecule modulators of gene dosage and may be broadly applicable to drug discovery efforts targeting haploinsufficiency diseases.

Introduction

Haploinsufficiency is a class of genetic disease wherein a single functional gene copy is insufficient for a normal phenotype. The ClinGen Dosage Sensitivity map annotates 413 genes with sufficient evidence for haploinsufficiency in humans [1]. Recent genome-wide loss-of-function (LoF) CRISPR screens have identified more than 600 haploinsufficient genes in human embryonic stem cells [2]. Due to phenotypic variability, it is difficult to know exactly how many diseases are associated with haploinsufficiency but estimates from the Online Mendelian Inheritance in Man (OMIM) database list more than 1,200 gene-disease relationships with haploinsufficiency as a mechanism. Together, these statistics indicate that haploinsufficient genes are responsible for a significant number of genetic diseases.

Alagille Syndrome (ALGS) exemplifies a haploinsufficiency disorder with significant clinical impact. This rare autosomal dominant disease is caused by heterozygous LoF mutations in JAG1 (93% of cases) or NOTCH2 (2%) [3-6]. Characteristics of ALGS include multisystem developmental abnormalities affecting the hepatic, cardiovascular, skeletal, ocular, craniofacial, and renal systems, with severity ranging from asymptomatic to life-threatening [7]. Cholestatic liver disease, especially in early childhood, is one of the most common phenotypes and can progress to liver fibrosis. Currently, no disease-modifying treatment exists; treatment remains limited to symptom management and supportive care, rather than a definitive cure [8].

High-throughput screening (HTS) offers a powerful tool to identify compounds that restore physiological gene dosage in haploinsufficiency diseases. Several parameters are required for quality HTS assays: sensitivity, reproducibility, stability, accuracy, and affordability [9 10]. In short, an ideal HTS-compatible assay is microtiter plate-based, homogeneous, and has robust and reproducible assay performance. In this study, we directly compare three HTS-compatible assays for their ability to detect compounds that upregulate a gene of interest for a haploinsufficiency disease, using JAG1 upregulation for ALGS as a case study.

For these studies, we selected the immortalized human hepatic stellate cell line LX-2 as a liver-derived model system for evaluating JAG1 upregulation assays. LX-2 cells express endogenous JAG1 and provide a robust, reproducible, and scalable platform that is well suited to microtiter plate-based HTS workflows. In addition, as a hepatic cell line relevant to liver remodeling and fibrosis [11], LX-2 offers a biologically appropriate context for assay development in ALGS.

Using this model system, we evaluated three complementary assay formats for detecting JAG1 upregulation. The first assay employs RNA fluorescence in situ hybridization (RNA FISH) to detect JAG1 mRNA (Figure 1A-B). This method allows detection of single mRNA molecules and additionally provides spatial information about JAG1 expression. Fluorescent probes specific to the JAG1 mRNA sequence are hybridized in fixed LX2 cells. After several rounds of washes and signal amplification with sequential probe hybridization, plates are imaged on a high-content imager. The second assay is an immunofluorescence (IF) stain that also uses high-content imaging but detects endogenous JAG1 protein instead of mRNA (Figure 1C-D). Live LX-2 cells grown in 2D culture are labeled with a fluorescent-conjugated anti-JAG1 antibody. Following fixation and washes, plates are imaged and JAG1 is quantified by the mean fluorescence intensity of each cell. The final assay tags endogenous JAG1 protein with a high-affinity split-luciferase HiBiT system (Figure 1E-F). The 11-amino acid HiBiT tag is inserted at the C-terminus of JAG1 in LX-2 cells via CRISPR to create the LX-2-JAG1-HiBiT cell line. At the assay endpoint, cells are lysed in the presence of substrate and LgBiT, the larger complementary luciferase subunit, which binds HiBiT to form an active luminescent enzyme. Luminescence is read on a plate reader to quantify JAG1 protein expression.

Figure 1. Overview of assays.

RNA fluorescent in situ hybridization (RNA FISH): A) After exposure to compounds, cells are fixed and permeabilized, and hybridized with the target probe. Then, the plates go through several rounds of washes and probe amplification before adding nuclear/cytoplasmic stains and imaging. B) The labeled target probe hybridizes with the mRNA target sequence. Sequential hybridizations with labeled probes amplify the signal. Immunofluorescence (IF) assay: C) JAG1 protein is labeled with PE-conjugated anti-JAG1 antibody in cells grown on collagen-coated plates. D) Compounds are added to plates via acoustic liquid handler and cells are seeded on top. At assay endpoint, antibodies are incubated with live cells, which are then fixed and washed prior to imaging. HiBiT assay: E) An 11-aa HiBiT tag is inserted via CRISPR at the C-terminus of endogenous JAG1. At assay endpoint, cells are lysed in the presence of LgBit protein which binds HiBiT to form a functional luciferase that emits luminescence upon addition of substrate. F) Compounds are spotted to plates via acoustic liquid handler and cells are seeded on top. At assay endpoint cells are lysed in presence of LgBit and substrate, and luminescence is read with a plate reader.

After optimization of these assays to 384-well format, we profiled a panel of 32 HDAC inhibitors, as these compounds are known to generally upregulate global gene expression in cells. This panel allowed us to directly compare the dose-response curves and positive hit identification for each assay. In this paper, we discuss the benefits and drawbacks of each assay and their potential utility for HTS. These assays are broadly applicable for screening small-molecule compounds to identify potential upregulators of genes in haploinsufficiency disease models.

Materials and Methods

Cell culture

The LX-2 Human Hepatic Stellate Cell Line (Millipore, #SCC064) was used for all assays. Cells were incubated at 37°C with 5% CO2 and maintained in high-glucose Dulbecco’s Modified Eagle Medium (DMEM) with L-glutamine (Gibco, #11965-092) supplemented with 1x Pen/Strep (Gibco, #15140), 1x GlutaMAX (Gibco, #35050-061), and HyClone^™ Characterized Fetal Bovine Serum (FBS; Cytiva, #SH30071.03) at 10% for thawing cells and 2% for routine culture. Medium was changed every 2-3 days, and cells were split at 1:3 to 1:6 every 3-4 days or at 80% confluency. Cells were used at passages 4–8. Recombinant Human TGF-β (R&D Biosystems, #7754-BH) was used as the positive control for JAG1 upregulation.

HDAC inhibitor compounds

A panel of 32 histone deacetylase inhibitors (HDACi) was selected as a set of compounds expected to broadly increase gene expression and thereby upregulate JAG1 in our model system. The compounds are listed in Supplementary Table 1 and span multiple mechanisms of action (MoA) across HDAC classes. Test compounds were cherrypicked from NCATS in-house storage, dissolved in DMSO at 10 mM, and titrated 1:3 in an 11-point dilution series. Compounds were dispensed using a pintool (Hornet NH-TR, Wako) for RNA FISH, or an acoustic liquid handler (Echo 655, Beckman Coulter) for IF and HiBiT assays, as described below.

RNA FISH assay

A 384-well test plate containing PDL, PEI, Collagen I, Fibronectin, Laminin, BME, and PDL-BME coatings was used for initial assay development. For the assays, LX-2 cells were dispensed at 1,300 cells/well in 50 μl of culture medium to collagen-coated black, clear-bottom 384-well Phenoplate microplates (PerkinElmer, #6057700, lot #1770-24235) with a MultiDrop Combi Reagent Dispenser (ThermoFisher Scientific). After 24 hr incubation at 37°C, 46 nL/well of controls or compounds was added by a pintool and cells were incubated for an additional 24 hr. Cells were fixed with 3.4% paraformaldehyde (PFA, Electron Microscopy Sciences, #15714-S) for 30 min at room temperature by adding 20 μL of 12% PFA to 50 μL of cell culture medium. Following fixation, 17 μL of Phosphate Buffered Saline (PBS) + 0.45% Triton-X (PBSTx, 0.1% final) was added to permeabilize cells for 5 min at room temperature. Cells were then washed 3x in PBS with a BioTek Plate Washer (ELx405, Bio-Tek) using settings that left 7 μL per well after aspiration then dispensed 70 μL PBS per well each cycle. Plates were sealed and stored at 4°C until staining.

On the day of the assay, fresh Hybridization Buffer was prepared as 2x SSC (ThermoFisher Scientific J60839.K2), 20% formamide (Millipore, #4650-500mL), and 10% dextran (500 kDa MW; Sigma, #D8906-50G) in nuclease free water and prewarmed at 40°C for >30 min. PBS was removed from plates of cells using a BlueWasher (BlueCat Bio; custom setting 600 rpm, acceleration 500). JAG1 target probe (ThermoFisher, JAG1 ViewRNA Cell Probe Set Type 6, #VA6-15093-VC) diluted 1:100 in hybridization buffer was added at 20 μL/well using an E1-ClipTip^™ Bluetooth^™ Electronic Multichannel Pipette (ThermoFisher, #4671050BT), plates were centrifuged at 500 rpm for 30 sec and incubated 3 hr at 40°C.

Plates were then washed 4x with Wash Buffer (ThermoFisher, #QVT0502) on a BioTek plate washer, with 5 min between washes and residual liquid was removed using a BlueWasher. This wash/removal sequence was repeated after each amplification step. Pre-Amplifier Probe Mix (ThermoFisher, #QVC0001), Amplifier Probe Mix (ThermoFisher, #QVC0001), and Label Probe Mix (ThermoFisher, #QVC0001) were each diluted 1:100 in hybridization buffer, added at 20 μl/well, centrifuged at 500 rpm for 30 sec, and incubated for 1 hr at 40°C. The Label Probe Mix step was performed with overhead lights off. After the final wash, 20 μL/well of PBS with DAPI (1:5,000; ThermoFisher, #62248) and HCS CellMask Green (1:10,000; ThermoFisher, #H32714) was added for 30 min at room temperature. Plates were washed 2x with PBS, filled with a final 50 μL PBS, sealed, and stored at 4°C until imaging.

Images were acquired on an Opera Phenix Plus High-Content Screening System (Revvity) using the DAPI, CellMask Green, and Alexa 647 channels with a 20x water objective over 9 fields per well. Images were analyzed in Signals Image Artist (Revvity). Nuclei were identified in the DAPI channel using the “Find Nuclei” building block and filtered for nuclear area >150 μm². Cell borders were defined using “Find Cytoplasm” in the CellMask Green channel, and cells whose boundaries passed outside of the image were excluded using “Remove Border Objects” filter. JAG1 mRNA puncta were identified using “Find Spots”, and cells exceeding a threshold of spots per cell were classified as positive. The threshold was selected for each experiment to maximize the assay window between TGF-β and DMSO treated wells. The % of positive cells per well (bandpass#, BP#) was calculated and normalized to TGF-β treated wells as 100% and DMSO treated wells as 0%.

IF assay

10 μL of medium was dispensed per well into collagen coated, black, clear-bottom 384-well microplates (Corning, #354667, lot #10624016) using a MultiDrop Combi Reagent Dispenser. Controls and compounds were added at 40 nL/well using an Echo 655 acoustic dispenser. LX-2 cells were then dispensed at 1,000 cells/well in 30 μL using a MultiDrop Combi Reagent Dispenser. After 48 hr at 37°C and 95% humidity, 5 μL of antibodies diluted to 9x final concentration in blocking solution (BioLegend, #420201) was dispensed onto live cells using a BioRAPTR Flying Reagent Dispenser (Beckman Coulter): PE-conjugated mouse anti-human JAG1 (1:200 final ; BD Pharm, #565495), and Hoechst (Life Technologies #H3570; 1:15,000 final). After 1 hr at 37°C, cells were fixed in 2% PFA for 20 min at room temperature (adding 5 μL of 20% PFA to 45 μL cells). Plates were washed 5 times with PBS + 0.1% Tween 20 (PBSTw) using a BioTek Plate Washer, leaving 7 μL of liquid per well after aspiration and dispensing 70 μL PBSTw each cycle. Plates were sealed and stored at 4°C until imaging.

Plates were imaged on an Opera Phenix Plus High-Content Screening System in the DAPI and PE channels using a 20x water objective over 15 fields per well. Images were analyzed in Signals Image Artist. Nuclei were identified from Hoechst staining in the DAPI channel using “Find Nuclei” and filtered for nuclear area >150 μm². Cell borders were defined using the “Find Cytoplasm” in the PE channel, and cells whose boundaries passed outside of the image were excluded using “Remove Border Objects” filter. Mean fluorescence intensity was calculated for each cell within the identified cytoplasm. Cells exceeding a fluorescence threshold were classified as positive; the threshold was selected to yield <10% positive cells in DMSO treated wells. The % of positive cells per well was calculated and normalized to TGF-β treated wells as 100% and DMSO treated wells as 0%.

HiBiT assay

An 11-amino acid HiBiT tag (GTGAGCGGCTGGCGGCTGTTCAAGAAGATTAGC) was inserted into the intracellular domain at the C-terminus of JAG1 via CRISPR in LX-2 cells to create LX-2-JAG1-HiBiT-KI cells. A single clone was selected by colony purification, and HiBiT knock-in was confirmed by RT-qPCR. LX-2 cells contain four copies of chromosome 20, where JAG1 is located, but only 1 copy was successfully tagged in the clone used here.

For assays, controls or test compounds were dispensed at 40 nL/well into white/opaque 384-well microplates (Greiner, #781073) using an Echo 655 acoustic dispenser. LX-2-JAG1-HiBiT-KI cells were then dispensed in 40 μL/well at the indicated seeding densities using a Multidrop Combi Reagent Dispenser and incubated at 37°C until assay endpoint. The Nano-Glo HiBiT Lytic Detection System (Promega, #N3030) was equilibrated to room temperature and prepared according to the manufacturer’s instructions. Lytic Mixture (Nano-Glo HiBiT Lytic Buffer + LgBit Protein + Nano-Glo Lytic Substrate) was dispensed at 40 μL/well using a BioRAPTR Flying Reagent Dispenser. Plates were shaken for 10 min at 850-950 rpm in the dark, centrifuged at 2,000 x g for 5 min to remove bubbles, and read on a PHERAstar FSX microplate reader (BMG LABTECH; gain = 3600). Data were normalized to TGF-β treated wells as 100% and DMSO-treated wells as 0%.

Cell viability for HiBiT assay was measured using the CellTiter-Glo Luminescent Cell Viability Assay (Promega, #G7572) to quantify ATP content. Parallel plates were seeded in duplicate and incubated alongside HiBiT assay plates. At the assay, 40 μL/well of room temperature CellTiter-Glo was dispensed using a BioRAPTR FDR and incubated for at least 10 min at room temperature in the dark. Luminescence was read on a PHERAstar FSX microplate reader (BMG LABTECH; gain = 2000). Data were normalized to DMSO-treated cells as 100% viability.

Data analysis

Signals were normalized by setting the positive control (TGF-β) to 100% and the negative control (DMSO) to 0% using the equation ${100}^{\ast }((\mathrm{x}-{\text{Mean}}_{\text{DMSO}})∕({\text{Mean}}_{\text{TGF-}\beta }-{\text{Mean}}_{\text{DMSO}}))$. Signal-to-basal ratio (S/B) was calculated as $\mathrm{S}∕\mathrm{B}={\text{Mean}}_{\text{TGF-}\beta }∕{\text{Mean}}_{\text{DMSO}}$. % CV was calculated as $\%\text{CV}={100}^{\ast }(\text{SD}∕\text{Mean})$ where SD is the standard deviation. The Z’ factor was calculated as $\mathrm{Z}’=1\text{-}(3^{\ast }({\text{SD}}_{\text{TGF-}\beta }+{\text{SD}}_{\text{DMSO}})∕({\text{Mean}}_{\text{TGF-}\beta }-{\text{Mean}}_{\text{DMSO}}))$.

Assay data were plotted and half-maximal effective concentrations (EC50) were calculated using GraphPad Prism version 10.3.1 for Windows (GraphPad Software, Boston, MA USA; www.graphpad.com). Graphs show replicate means ± standard deviation (SD). Results are reported rounded to 2 significant figures.

Results

Assay Optimizations

RNA FISH Assay Optimization

To identify optimal seeding conditions, equal numbers of LX-2 cells were plated in 384-well plates coated with poly-D-lysine (PDL), polyethyleneimine (PEI), collagen I, fibronectin, laminin, basement membrane extract (BME), a combination of PDL and BME, or left uncoated. The following day, cells were stained with Hoechst, and nuclei were counted to assess cell attachment, survival, and retention. Among the tested substrates, fibronectin and collagen I supported the highest cell survival, as indicated by the highest nuclei counts (Figure 2A). Collagen I was selected for subsequent imaging experiments because it provided robust cell attachment and consistent performance.

Figure 2. RNA FISH Optimization.

A) LX-2 cells were grown on various plate coatings at 1,000 cells/well in 384-well plates. Cells were fixed the next day, and nuclei were stained and counted. Counts are average nuclei per field-of-view (FOV). 4 technical replicates per condition, data reported as Mean with Standard Deviation. B) Representative images of JAG1 RNA FISH probe stain in LX-2 cells treated with 5 ng/mL TGF-β or DMSO. JAG1 mRNA = magenta (each spot = 1 molecule mRNA), CellMaskGreen = green, Hoechst = cyan. Scale bar = 100um. C-F) JAG1 mRNA expression reported as the % of cells exceeding 20 spots per cell (%BP20). C) Cells were treated with varying dilutions of protease following fixation. 16 replicates. Cell attachment and cell retention measured as number of nuclei per FOV (magenta line). Data are reported as Mean with Standard Deviation. D) Target probe ratios of 1:50 or 1:100 were tested in cells treated with 5 ng/mL TGF-β or untreated controls. Signal-to-basal ratios (S/B) and Z’ scores indicated above bars. 16 replicates. Data reported as Mean with Standard Deviation. E) Cells were incubated with TGF-β for 24, 48, or 72 hr prior to fixation and RNA FISH staining. 3 replicates. Data reported as Mean with Standard Deviation. F) A TGF-β dose-response curve. Top concentration = 16ng/mL, 8-pt 1:2 dilution. 4 replicates. Data reported as Mean with Standard Deviation.

To validate assay responsiveness, TGF-β treatment, previously shown to upregulate JAG1 expression [12], was used as a positive control. The RNA FISH protocol (detailed in Methods) produced distinct fluorescent puncta representing single JAG1 mRNA molecules (Figure 2B), which were quantified on a per-cell basis. Protease treatment, following the manufacturer’s instructions, at the indicated concentrations, was evaluated to enhance probe accessibility. Although protease increased the proportion of cells with >20 JAG1 mRNA spots per cell (%BP20), it also reduced cell viability, as measured by the number of nuclei per field of view (FOV; magenta line, Figure 2C). Thereafter, protease treatment was omitted from subsequent protocols.

Probe concentration was optimized by testing the JAG1 ViewRNA Cell Probe Set Type 6 at 1:50 and 1:100 dilutions. The 1:100 dilution was selected based on its superior signal-to-basal (S/B) ratio and Z' factor (Figure 2D). The optimal TGF-β incubation time was determined by treating LX-2 cells with 5 ng/mL TGF-β for 24, 48, or 96 hours (Figure 2E). A 48 hr incubation yielded the highest induction of JAG1 mRNA and was used in all further experiments. Finally, a TGF-β dose-response curve was generated using an 8-point, 1:2 serial dilution starting at 16 ng/mL (Figure 2F). JAG1 mRNA induction had an EC50 of 0.28 ng/mL and plateaued at 2 ng/mL. A concentration of 5 ng/mL TGF-β was selected as the standard positive control for subsequent experiments.

Immunofluorescence Assay Optimization

Immunofluorescence (IF) staining for JAG1 protein was optimized in LX-2 cells seeded onto collagen-coated 384-well plates. Staining with anti-JAG1-PE conjugated primary antibody was first optimized by comparing live-cell and fixed-cell staining across different times and temperatures. Cells were incubated with 5 ng/mL TGF-β as a positive control or DMSO as a negative control before primary antibody labeling under the conditions shown in Figure 3A. S/B ratios were higher with live-cell staining, and the optimal condition was live-cell staining at 37°C for 1 hr (Figure 3B). Antibody dilutions of 1:100, 1:200, and 1:400 were then tested (Figure 3C). S/B ratios exceeded 2-fold for all dilutions and were highest at 1:200 (S/B = 2.4). Next, TGF-β incubation time was tested. Cells were incubated with TGF-β for either 48 or 72 hr to allow time for protein expression changes to occur (Figure 3D). The S/B ratio was slightly higher at 48 hr than 72 hr (33 vs. 28), so 48 hr was selected for subsequent experiments. A dose response curve for TGF-β was then generated using a 15-point, 1:2 serial dilution starting at 16 ng/mL (Figure 3E). JAG1 protein induction showed an EC50 of 0.69 ng/mL, and 5 ng/mL TGF-β was selected as the standard positive control dose for subsequent experiments to match the dose used in RNA FISH.

Figure 3. IF Optimization.

A) Representative images of cells treated with either TGF-β (5 ng/mL) or DMSO and stained with anti-JAG1-PE at various incubation conditions: 30 min vs. 1 hr, 37°C vs. room temperature (RT), and live cell vs. fixed cell staining. Anti-JAG1-PE = yellow, Hoechst = blue. Scale bar 100μm. Quantification of B) Live vs. fixed-cell staining at different antibody incubation times and temperatures and C) Antibody titration. JAG1 expression reported as mean fluorescence intensity per cell at each condition tested. Signal-to-basal ratios (S/B) indicated above graph bars. 3 technical replicates for each condition in (B) and (C). D) TGF-β incubation time in cells treated with DMSO or TGF-β. 16 technical replicates for each condition, data expressed as % of cells exceeding a bandpass of 400 fluorescence mean intensity (BP400 %). E) TGF-β dose-response curve, 6 technical replicates for each condition, data expressed as % of cells exceeding a bandpass of 600 fluorescence mean intensity (BP600). Data are Mean with Standard Deviation.

HiBiT Assay Optimization

Initial optimization of the HiBiT assay examined different seeding densities in 384-well plates. LX-2-JAG1-HiBiT cells were seeded at 500, 1,000, 1,500, 2,000, and 5,000 cells per well in the presence of 5 ng/mL TGF-β as a positive control or DMSO as a negative control. Cells were incubated for 48 hr before addition of HiBiT reagents and luminescence measurement. Increasing cell density raised the raw luminescence signal, but S/B ratios were optimal at 1,500 cells/well, which was used for subsequent experiments (Figure 4A). Incubation time was also tested, S/B ratios were above 2-fold for 24, 48, and 72 hr of incubation with 5 ng/mL TGF-β compared with DMSO (Figure 4B). While 24 hr had a slightly higher S/B ratio, 48 hr incubation was chosen to allow sufficient time for protein expression changes following compound treatment and to match the IF assay timescale, which also measures JAG1 protein levels. A full TGF-β dose-response curve was then generated using a top concentration of 16 ng/mL and a 15-point, 1:2 dilution series at 48 hr. Signal saturation was reached at 1 ng/mL, and the EC50 was 0.17 ng/mL (Figure 4C). A concentration of 5 ng/mL TGF-β was selected as the standard positive control dose for subsequent experiments to match the dose used in RNA FISH and IF assays.

Figure 4. HiBiT Optimization.

A-C) Optimization of the HiBiT assay with LX-2-JAG1-HiBiT cells in 384-well plates. Readout is raw luminescence per well. All data are presented as mean with standard deviation. Signal-to-basal ratios (S/B) are indicated above graph bars. A) Seeding density of cells treated with TGF-β or DMSO controls for 48 hr. 16 technical replicates per condition. B) Optimization of TGF-β incubation time (24, 48, or 72 hr) of 1500 cells/well. 16 technical replicates per condition. C) TGF-β dose response curve, top concentration = 16ng/mL, 16-pt 1:2 dilution. 8 technical replicates. Data reported as Mean with Standard Deviation.

Plate Statistics (Figure 5)

Figure 5. Plate Statistics.

A-C) Assay performance summarized by plate statistics for cells assayed on 384-well plates that include average, standard deviation, % CV, S/B, and Z’. Data are mean with standard deviation. A) RNA FISH assay. TGF-β was loaded into columns 2 and 24, and remaining columns were treated with DMSO. Readout is % of cells exceeding a bandpass threshold of 30 spots per cell. B) IF assay. Column 2 was treated with TGF-β and remaining columns with DMSO. Readout is % of cells exceeding a bandpass threshold of 900 mean fluorescence intensity. C) HiBiT assay. Column 2 was treated with TGF-β and remaining columns with DMSO. Readout is raw luminescence per well. D-F) Plate statistics for control columns (1 column TGF-β, 1 column DMSO per plate) for each replicate plate run in the HDACi screen. Blue squares = Z’ score, Green circles = S/B ratio. D) RNA FISH assay. 3 replicate plates. E) IF assay. 3 technical replicates across 4 plates. F) HiBiT assay. 3 replicate plates.

Once conditions were optimized, DMSO control plates were run for each assay, with positive control 5 ng/mL TGF-β dispensed to 1 or 2 columns and DMSO dispensed to all remaining columns of a 384-well plate. Plates were incubated for 24 hr for RNA FISH and 48 hr for IF and HiBiT assays. Assay readouts were the % of cells above a bandpass of >30 spots per cell for RNA FISH, the % of cells above a bandpass (BP) of >900 mean fluorescence for IF, and raw well luminescence for the HiBiT assay. Plate statistics included average signal, standard deviation, coefficient of variation (% CV), signal-to-basal ratio (S/B), and a Z’ score (Figure 5A-C). RNA FISH showed relatively high inter-well variability (% CV = 19 for TGF-β and 38 for DMSO), but also a 11-fold S/B ratio, representing a Z’ score of 0.27 (Figure 5A). The IF assay showed similar variability (% CV = 16 for TGF-β and 47 for DMSO) but had the highest S/B ratio (14) and a Z’ score of 0.37 (Figure 5B). The HiBiT assay had the lowest inter-well variability (% CV = 6.4 for TGF-β and 10 for DMSO) and the smallest S/B ratio (2.7) but achieved the highest Z’ score of 0.52. Overall, HiBiT assay provided the most robust plate statistics, whereas RNA FISH and IF provided larger assay windows with greater variability.

HDAC Inhibitor validation:

To directly compare the assays, we tested them against a set of small molecule compounds known to broadly increase gene expression. We chose a panel of 32 HDAC inhibitors spanning a range of histone deacetylase class targets (Supplementary Table 1). For each assay, replicate 384-well plates contained 1 column of 5 ng/mL TGF-β positive control and 1 column of DMSO negative control. Plate statistics were calculated from the control wells in all 3 assays (Figure 5D-F). In the RNA FISH assay, Z’ scores ranged from 0.43-0.69 with S/B ratios from 5.5-5.9 (Figure 5D) across 3 replicate plates. The IF assay, which included 3 technical replicates across 4 separate plates, had Z’ scores from 0.27-0.45 and S/B ratios from 15-18 (Figure 5E). The HiBiT assay had Z’ scores from 0.57-0.70 and S/B ratios of 2.8-2.9 (Figure 5F) across 3 replicate plates.

Each HDAC inhibitor was tested in an 11-point 1:3 dilution series with a 10 mM top concentration in the source plate. Compounds were dispensed either by pintool transfer at a dilution of 1:1,087 (RNA FISH; top assay concentration = 9.2 μM) or by acoustic dispense at a dilution of 1:1,000 (IF and HiBiT; top assay concentration = 10 μM). Assays were performed after 24 hr of incubation for RNA FISH or 48 hr for IF and HiBiT. EC50 values were determined for each compound and % efficacy was normalized to the response induced by the positive control TGF-β (Table 1). Compounds with <30% of the TGF-β response were considered inactive. Cytotoxicity EC50 values were also calculated using nuclei counts from the RNA FISH and IF images, or from an ATP-content assay performed in parallel wells for the HiBiT assay (Table 1). Viability was expressed as a % relative to DMSO treated controls. Compounds causing <30% loss of cell viability were considered non-cytotoxic. Full dose-response curves for all compounds are shown in Supplementary Figures 1-3.

Table 1: HDAC Inhibitor Activity and Cytotoxicity.

NCATS in-house compound IDs and commercial names for HDAC inhibitors tested. All data normalized to the response to positive control, TGF-β. EC50 reported for each assay for compound activity and viability. % efficacy is the maximum efficacy for a given compound at any concentration. Compounds that failed to reach at least 30% efficacy were considered not active (N/A). % viability describes the % of cells alive compared to DMSO control at the maximum efficacy concentration. Cells that had <30% cell killing were not considered toxic (NT).

		RNA FISH:				IF:				HiBiT:
		Activity		Viability		Activity		Viability		Activity		Viability
Compound ID	Compound Name	EC50	% Efficacy	EC50	% Viability	EC50	% Efficacy	EC50	% Viability	EC50	% Efficacy	EC50	% Viability
NCGC00165833-22	Entinostat	7.0E-07	94%	4.1E-06	101	9.4E-07	190%	3.3E-06	62	4.7E-07	170%	3.7E-05	52
NCGC00263598-08	Chidamide	7.3E-07	110%	8.7E-04	87	9.7E-07	220%	5.0E-06	54	5.9E-07	160%	2.7E-06	50
NCGC00685364-02	Tucidinostat	5.9E-07	76%	1.3E-06	74	9.6E-07	250%	1.8E-02	44	4.0E-07	160%	2.0E-06	57
NCGC00263182-08	Mocetinostat	2.3E-07	100%	2.7E-06	81	3.4E-07	250%	1.7E-06	27	1.3E-07	150%	5.3E-06	25
NCGC00263639-01	Tacedinaline	2.4E-06	71%	1.7E-09	91	3.1E-06	150%	1.6E-06	78	1.3E-06	140%	3.5E-06	63
NCGC00263220-14	Romidepsin	1.2E-09	86%	6.6E-09	63	7.5E-10	110%	2.7E-09	2.6	1.0E-09	120%	2.4E-09	0.54
NCGC00499575-01	HDACi-02	1.6E-06	67%	1.9E+13	66	4.9E-07	120%	2.3E-06	13	2.1E-07	80%	2.2E-06	5.2
NCGC00384199-03	4SC-202	4.4E-07	72%	1.9E-07	53	6.0E-07	99%	8.9E-07	15	1.2E-04	71%	7.6E-06	20
NCGC00346486-04	Abexinostat	5.9E-08	67%	1.9E-07	72	3.2E-07	110%	4.1E-07	7.6	2.5E-08	69%	4.0E-07	3.4
NCGC00511370-02	ACY-738	1.3E-06	50%	1.3E-06	89	1.9E-06	150%	3.9E-06	34	9.7E-01	66%	−6.8E-01	20
NCGC00263136-13	Pracinostat	8.6E-08	56%	1.9E-07	76	3.1E-07	90%	7.0E-07	7.1	4.6E-07	64%	4.8E-07	3.0
NCGC00381559-03	Resminostat	8.7E-07	54%	2.5E-06	79	2.2E-06	130%	3.2E-06	18	8.6E-03	57%	4.5E-06	6.0
NCGC00346487-03	Quisinostat	1.8E-09	55%	7.9E-09	57	1.6E-09	29%	1.3E-08	3.9	1.5E-09	52%	1.6E-08	1.7
NCGC00168085-21	Vorinostat (SAHA)	1.1E-06	60%	4.0E-06	65	7.6E-06	110%	2.9E-06	17	5.2E-07	48%	2.7E-06	6.6
NCGC00263117-10	Panobinostat	3.6E-09	64%	1.8E-08	68	3.8E-09	83%	1.6E-08	6.4	2.5E-09	47%	2.4E-08	4.3
NCGC00263153-08	AR-42	2.1E+08	46%	4.3E-07	71	4.4E-06	84%	4.8E-07	7.9	3.3E-04	46%	3.9E-07	3.8
NCGC00346743-05	Givinostat hydrochloride	6.1E-08	66%	3.1E-06	60	6.7E-07	150%	3.2E-07	4.7	7.5E-08	45%	3.6E-07	2.4
NCGC00531749-01	ACY-775	N/A	11%	Unstable	101	4.7E-06	59%	3.6E-06	86	4.0E-02	41%	4.4E-06	77
NCGC00481588-03	Citarinostat	1.4E-06	59%	4.6E-01	74	2.0E-06	51%	4.3E-06	25	4.3E-07	35%	1.9E-05	6.1
NCGC00345802-10	Rocilinostat	1.8E-04	64%	1.9E+00	73	7.0E-06	58%	1.3E-01	44	5.0E-07	29%	2.6E-01	28
NCGC00263606-19	Tubastatin A	N/A	16%	Unstable	95	3.9E-05	44%	2.1E-05	74	1.2E-06	21%	9.4E-05	67
NCGC00263155-11	Belinostat	3.8E-07	53%	1.2E-06	84	7.6E-07	68%	1.1E-06	10	8.8E-08	20%	7.9E-07	3.6
NCGC00499576-01	HDACi-01	2.3E-06	95%	1.3E-06	76	1.4E-06	110%	6.4E-06	8.4	N/A	16%	1.6E-06	14
NCGC00386712-02	TMP195	N/A	23%	1.2E-05	93	N/A	9.4%	1.5E-05	100	3.9E-06	12%	4.3E-06	79
NCGC00345492-10	PCI-34051	N/A	10%	1.1E+13	94	N/A	3.3%	2.0E-06	88	N/A	11%	3.3E-06	87
NCGC00687532-01	CHDI 00390576	N/A	21%	3.5E-06	88	N/A	27%	4.3E-06	45	N/A	8.8%	4.4E-06	36
NCGC00346534-04	MC-1568	N/A	6.3%	4.9E-07	110	N/A	7.0%	5.9E-03	73	N/A	7.3%	Unstable	96
NCGC00685385-01	SKLB-23bb	5.3E-07	52%	2.6E-07	61	9.0E-01	38%	2.7E-07	26	N/A	6.5%	1.2E-05	35
NCGC00510505-02	WT-161	2.0E-06	68%	1.7E-04	61	2.2E-06	120%	1.8E-06	5.5	N/A	4.0%	1.0E-06	2.4
NCGC00091149-12	Valproic acid	N/A	6.1%	Unstable	90	N/A	2.9%	7.2E+04	84	N/A	2.8%	3.2E-06	55
NCGC00087387-07	Droxinostat	N/A	10%	1.2E-08	89	N/A	5.9%	3.3E-06	85	N/A	1.3%	4.5E-01	57
NCGC00346840-04	Tubacin	4.8E-06	36%	1.2E-06	92	N/A	14%	4.4E-06	64	N/A	0.70%	9.3E-04	59

^*N/A = not applicable (<30% efficacy threshold)

^**NT = non-toxic (<30% cell killing not considered cytotoxic)

In general, the RNA FISH assay revealed less compound cytotoxicity, likely reflecting its shorter incubation time compared with the other assays (24 vs. 48 hr). Compound efficacy was also generally lower in RNA FISH, with few compounds reaching the activity level as TGF-β. In contrast, IF detected 15 compounds that reached or exceeded the TGF-β-induced level of JAG1 expression, whereas HiBiT detected 6. The same top 3 JAG1 upregulators were independently identified by all assays: Entinostat, Mocetinostat, and Chidamide (Figure 6A-C). All 3 assays measured an EC50 value in the nanomolar range for these compounds, with low cytotoxicity at effective concentrations.

Figure 6. HDAC Inhibitor – Top JAG1 upregulators.

Comparison of top hits (Entinostat, Mocetinostat, and Chidamide) of JAG1 upregulation across all assays. A) Assay readout for RNA FISH is % of cells with >25 spots per cell (% BP25) normalized to positive control, TGF-β. % cell viability measured by counting # nuclei per field of view and expressed as a % of DMSO-treated control. B) Assay readout for IF is mean intensity of anti-JAG1-PE bandpass >900 (BP900) normalized to positive control, TGF-β. % cell viability measured by counting # nuclei per field of view and expressed as a % of DMSO-treated control. C) Assay readout for HiBiT is luminescence normalized to positive control, TGF-β. % cell viability measured with an ATP-content assay and expressed as a % of DMSO-treated control. Black/left y-axis = assay readout, Blue/right y-axis = cell viability. 3 technical replicates. Results are reported as mean with standard deviation.

However, some discrepancies in hits were observed across assays, as highlighted in Figure 7 A-C. RNA FISH and IF identified HDACi-01 as an active compound, with efficacy approaching that of the positive control in the low μM range, but this compound showed no activity in HiBiT. Similarly, WT-161 showed >100% efficacy in IF, but only 68% efficacy in RNA FISH and <0% efficacy in HiBiT. Viability curves for these compounds indicate that both HDACi-01 and WT-161 are highly cytotoxic by 48 hr, as indicated in the IF and HiBiT assays, and begin to show toxicity even by 24 hr in the RNA FISH assay.

Figure 7. HDAC Inhibitor Profiling – Assay Disagreements.

Comparison of hits (HDACi-01 and WT-161) of JAG1 upregulation that differ between assays. A) Assay readout for RNA FISH is % of cells with >25 spots per cell (% BP25) normalized to positive control, TGF-β. % cell viability measured by counting # nuclei per field of view and expressed as a % of DMSO-treated control. B) Assay readout for IF is mean intensity of anti-JAG1-PE bandpass >900 (BP900) normalized to positive control, TGF-β. % cell viability measured by counting # nuclei per field of view and expressed as a % of DMSO-treated control. C) Assay readout for HiBiT is luminescence normalized to positive control, TGF-β. % cell viability measured with an ATP-content assay and expressed as a % of DMSO-treated control. Black/left y-axis = assay readout, Blue/right y-axis = cell viability. 3 technical replicates. Results are reported as mean with standard deviation.

Discussion

The ideal HTS assay is sensitive, reproducible, stable, accurate, and cost-effective. In practice, trade-offs among these characteristics are inevitable. For example, assays with high sensitivity but modest Z' scores can still be acceptable for HTS if run in replicate or in concentration–response formats [9]. Assay selection therefore requires balancing and prioritizing these factors in the context of specific project goals and available equipment and reagents. Table 2 summarizes the main advantages and limitations of the three assays evaluated in this study, and below we discuss how each aligns with HTS criteria.

Table 2: Assay Comparison.

Comparison of the pros and cons of each assay for high-throughput screening.

	Pros	Cons
RNA FISH (RNA)	Single-cell resolution Cell viability data gathered simultaneously (nuclei stain) High signal:basal ratio	Not homogeneous (20+ washes) Highly temperature sensitive → inter-plate variability Requires low-base imaging plates Requires high-powered microscope Time-consuming (assay & imaging) Cell viability data susceptible to cells lifting off during washes
IF (protein)	Single-cell resolution Cell viability data gathered simultaneously (nuclei stain) High signal:basal ratio	Not homogeneous (5 washes) Need a highly specific fluorescence-conjugated primary antibody Requires special collagen-coated low-base imaging plates Requires high-powered microscope Time-consuming imaging High inter-well variability (low Z’ score)
HiBiT (protein)	Homogeneous assay Quick reads (plate reader) Low inter-sample variability (high Z’ score) Highly cytotoxic compounds weeded out of hits Can multiplex cell viability assay	Time for derivation of HiBiT-tagged cells No data on cell morphology / heterogeneity Whole-well phenotypic readouts only

Both IF and RNA FISH assays yielded high S/B ratios, with 11- to 18-fold increases in signal in TGF-β–treated wells compared with DMSO controls. RNA FISH is well known for its sensitivity and its ability to detect single mRNA molecules, making it particularly suitable for genes with low endogenous expression or for detecting subtle transcriptional changes. In contrast, the success of the IF assay depended on access to a high-quality, target-specific, primary antibody against JAG1. The specificity of the JAG1-PE antibody used in this study was previously validated by another group using CRISPR knockout of JAG1 in A673 Ewing Sarcoma cells [13]. Antibody-based assays remain dependent on reagent quality; not every protein target is amenable to immunostaining, and not all antibodies are sufficiently sensitive or specific. This should be considered when choosing between assays for a given project.

Relative to IF and RNA FISH, the HiBiT assay had a lower S/B ratio, although it still exceeded 2-fold, which is generally sufficient for HTS. This may reflect limitations intrinsic to our gene of interest, JAG1, in terms of basal expression levels in quiescent hepatic stellate cells. According to the Human Liver Cell Atlas [14], JAG1 is expressed at low but detectable levels in stellate cells of the healthy adult liver and is enriched in endothelial cells and cholangiocytes. Additionally, in our implementation, the sensitivity of the HiBiT assay was likely limited by the genetic configuration of the LX-2 cell line: the chromosome harboring JAG1 is present in four copies, and only one allele was tagged with HiBiT in the clone used here. Accordingly, the dynamic range observed here may underestimate the performance that could be achieved in a clone with additional tagged alleles or in a cell type with higher endogenous JAG1 expression. Anecdotally, in other HiBiT cell lines used in our lab, we observe clone-to-clone differences in signal even at similar HiBiT copy number, suggesting that some variability is inherent to the cellular background rather than the assay chemistry alone.. A further limitation of the HiBiT technology is that it provides a whole-well readout and cannot resolve per-cell phenotypes. For disease models or studies requiring more complex or spatially resolved phenotypic readouts, assays with single-cell resolution, such as IF or RNA FISH, may therefore be preferable.

Plate-level statistics such as the coefficient of variation (%CV) and the Z' factor are useful metrics of assay precision and robustness. Among the three assays, HiBiT was the only one that consistently achieved Z' > 0.5, the generally accepted cutoff for cell-based HTS assays. This reflects the relatively low inter-well variability of the HiBiT readout and resulted in %CV values compatible with HTS. In contrast, %CV was high for RNA FISH and IF, largely because the mean signal in DMSO-treated wells was very low (Figure 5A, 5B). When the mean approaches zero, %CV becomes less informative, as small absolute fluctuations translate into large relative changes. In the IF assay, we also observed edge effects with certain lots of collagen-coated plates, which contributed to variability in the untreated wells of the DMSO plates. More broadly, the relatively modest Z' values for RNA FISH and IF suggest that, in their current form, these assays may be better suited as secondary, orthogonal, or lower-throughput screening tools rather than as stand-alone primary assays for very large campaigns.

Variability in TGF-β–treated wells was also higher in the RNA FISH and IF assays than in HiBiT. For RNA FISH in particular, this variability was strongly influenced by hybridization temperature. For example, replicate plates 1 and 3 from the HDACi screen were incubated on different shelves of the same hybridization oven, which we later found differed by 1–2°C. This small temperature difference was sufficient to alter Z' scores, even though the S/B ratio between TGF-β and DMSO controls remained stable (Figure 5D). Similar minor temperature variations between runs or between positions in the oven likely contributed to the differences in Z' scores observed between the DMSO plate (Figure 5A) and the HDACi screen plates (Figure 5D). This highlights the sensitivity of imaging-based assays to environmental and workflow variables and reinforces the importance of strict process control if these formats are used in larger screening campaigns.

There was substantial overlap between hits across the three assays. Using an arbitrary threshold of 60% efficacy, 9 of the 32 compounds were identified as hits by all three assays, and another 7 compounds were called hits by at least two assays. Only one compound met this threshold in RNA FISH alone and three in IF alone, whereas HiBiT did not identify any hits that were not corroborated by at least one of the imaging assays. With this hit criterion, the IF assay had a hit rate of 59%, RNA FISH 47%, and HiBiT 34%. On this basis, HiBiT appears to be the most stringent assay, at least when efficacy is the primary criterion for hit identification. Because the present study was designed primarily as an assay development and comparison study, this comparison should not be interpreted as providing new biological insight into JAG1 regulation beyond demonstrating responsiveness to known modulators. Our benchmarking set was limited to 32 HDAC inhibitors, most of which belong to a related pharmacologic class, and therefore does not fully model the chemical and mechanistic diversity encountered in a large-scale screening campaign.

Interpreting these hits in the context of cell viability is essential. Viability can be assessed in imaging assays by counting nuclei, or more generally by ATP-content assays. For whole-well readouts such as HiBiT, the signal is intrinsically coupled to cytotoxicity: if most of the cell population is killed, the overall luminescence will be low. In contrast, single-cell assays such as RNA FISH and IF are biased toward surviving cells. If the cells that escape cytotoxicity exhibit elevated expression of the gene of interest, a compound may appear to be a hit even though its toxicity makes it a poor candidate for further development. This effect is illustrated in Figure 7, where HDACi-01 and WT-161 show high apparent efficacy in RNA FISH and IF, but no effect in HiBiT, because cell viability drops steeply at higher concentrations. For whole-well assays, cytotoxic compounds will generally decrease signal, which could appear as a false negative in agonist screens or a false positive in inhibitor screens. Non-homogeneous assays involving multiple wash steps also carry the risk of washing cells off the plate, artificially selecting for certain cell populations. Incubation time is another important factor: cytotoxicity is more pronounced with longer compound exposure in the HiBiT and IF assays than in the 24-hour RNA FISH assay.

From a practical standpoint, the HiBiT assay is homogeneous, involves the fewest steps, and has the shortest hands-on and incubation times of the three assays. These operational advantages are offset by reagent cost: HiBiT detection reagents cost approximately $0.78 per well, with an additional ~$0.19 per well if using the CellTiter-Glo Assay for ATP-content. These costs are in addition to the up-front investment required to generate a HiBiT-tagged cell line. IF staining is more labor-intensive, requiring five rounds of washes, but is the least expensive on a per-well basis. The costs of blocking solution, Hoechst, and paraformaldehyde are negligible per well; in our hands the main expense was the conjugated antibody at about $0.26 per well. The RNA FISH kit components and probe set cost ~$0.84 per well. However, the RNA FISH protocol requires more than 20 wash steps and typically one to two full days to complete, so reagent savings must be weighed against the time and effort required from research staff.

Instrumentation requirements further influence assay choice and cost. Depending on antibody quality and target abundance, IF signals can be detected with a standard fluorescence microscope. For HTS applications, however, the microscope must be able to read 384-well plates and ideally provide automated acquisition and analysis. Even among high-content imaging systems, sensitivity differs. For example, we were unable to reliably resolve JAG1 mRNA puncta in the RNA FISH assay using an ImageXpress HCS.ai high-content imager (Molecular Devices) and required the higher optical performance of the Opera Phenix (Revvity) to obtain clear images. By contrast, whole-well readouts such as the HiBiT assay can be measured with a luminescence plate reader, which is more common in laboratories. Plate readers also offer faster acquisition than high-throughput imaging, which helps offset the higher per-well reagent cost of HiBiT.

Miniaturization beyond 384-well plates would be the next step toward making these assays fully compatible with ultra-high-throughput screening. RNA FISH is currently the most time-consuming and labor-intensive, but with appropriate automation for incubation, liquid-handling, and washing it could be feasible at higher plate densities and would provide a powerful orthogonal assay for hit confirmation. It is also feasible to further miniaturize the IF assay, especially with automation, because of the low relative number of wash and incubation steps. For HiBiT, the raw luminescence from the JAG1-HiBiT clone used here was not sufficient to detect differences between DMSO- and TGF-β–treated wells in a 1,536-well format (data not shown). Successful miniaturization of HiBiT in this disease context will likely require a cell type that expresses JAG1 more robustly or a clone with HiBiT tags on additional alleles.

Several additional limitations should be considered when interpreting this study. First, all assays were developed and benchmarked in a single immortalized hepatic stellate cell line, LX-2. Although this cell type provided a tractable and biologically relevant system for assay development, the performance characteristics reported here may not fully translate to other JAG1-expressing cell types or more disease-relevant models. Second, assay responsiveness was characterized by using TGF-β as the principal positive control and a focused panel of HDAC inhibitors. As a result, the present study does not establish how these assay formats will perform across a wider range of mechanisms of JAG1 regulation. Third, we did not perform orthogonal confirmation of top compounds in this manuscript using independent methods such as RT-qPCR or Western blotting. The current study was intended to compare screening-compatible assay formats rather than to validate individual chemical matter. Accordingly, the overlapping activity of several compounds across platforms should be viewed as supportive of assay concordance, but not as a substitute for independent confirmation. Finally, we did not conduct a pilot screen of a larger and more chemically diverse library, and therefore the real-world performance of these assays in a full-scale discovery campaign remains to be established.

Conclusion

The assays described here provide a useful framework for developing screening strategies to identify small-molecule modulators of endogenous JAG1 expression. More broadly, the principles compared here may inform assay development for other dosage-sensitive targets, but the applicability of these specific formats to other haploinsufficiency disorders, targets, or disease contexts will need to be established experimentally on a case-by-case basis. Likewise, although these assays could in principle be adapted to identify downregulators in other settings, that application was not tested here. Because haploinsufficient genes may be sensitive not only to underexpression but also to overexpression, therapeutic development for these disorders requires attention to gene dosage and pharmacologic window [15]. Strategies to restore function of the wild-type allele to physiologic levels may include increasing transcription, preventing degradation of the target protein, or enhancing target protein function[16]. In that context, cell-based assays capable of quantifying endogenous target expression may be valuable components of discovery workflows. However, the present study should be viewed primarily as a methodological comparison rather than as proof that any single assay format is broadly optimal for all targets or screening campaigns.

Overall, our comparison highlights that no single assay satisfies all HTS criteria optimally. In the LX-2/JAG1 system studied here, the homogeneous HiBiT assay provided the most robust plate statistics and greatest operational simplicity, making it an attractive option for primary screening, if cytotoxicity is monitored in parallel, and if sufficient signal can be achieved in the chosen cellular model. IF and RNA FISH offered higher spatial and single-cell resolution and greater flexibility for complex phenotypes, at the cost of increased variability, labor, and equipment demands. In this specific experimental context, a combined strategy—for example, HiBiT for primary hit triage, and IF or RNA FISH for orthogonal confirmation and mechanistic follow-up—is a practical approach. Future work in additional cellular models, with more diverse compound libraries and orthogonal validation methods, will be important to determine how broadly this workflow can be applied.

Supplementary Material

1

Supplementary Table 1. HDAC Inhibitor List

List of the HDAC inhibitors included in the screening panel. NCATS internal sample ID, commercial compound name, gene target, and primary Mechanism of Action (MoA).

Supplementary Figure 1. RNA FISH Assay HDAC Inhibitor Panel Results

Complete results from the RNA FISH assay screen of the HDAC inhibitor panel. Assay readout for RNA FISH is % of cells with >25 spots per cell (% BP25) normalized to positive control, TGF-β. % cell viability measured by counting # nuclei per field of view and expressed as a % of DMSO-treated control. Black/left y-axis = assay readout, Blue/right y-axis = cell viability. 3 technical replicates. Results are reported as mean with standard deviation.

Supplementary Figure 2. Immunofluorescence Assay HDAC Inhibitor Panel Results

Complete results from the IF assay screen of the HDAC inhibitor panel. Assay readout for IF is mean intensity of anti-JAG1-PE bandpass >900 (BP900) normalized to positive control, TGF-β. % cell viability measured by counting # nuclei per field of view and expressed as a % of DMSO-treated control. Black/left y-axis = assay readout, Blue/right y-axis = cell viability. 3 technical replicates. Results are reported as mean with standard deviation.

Supplementary Figure 3. HiBiT Assay HDAC Inhibitor Panel Results

Complete results from the HiBiT assay screen of the HDAC inhibitor panel. Assay readout for HiBiT is luminescence normalized to positive control, TGF-β. % cell viability measured with an ATP-content assay and expressed as a % of DMSO-treated control. Black/left y-axis = assay readout, Blue/right y-axis = cell viability. 3 technical replicates. Results are reported as mean with standard deviation.