High-Throughput Cellular Thermal Shift Assay (CETSA) using Acoustic Transfer of Protein Lysates

Abstract

Cellular thermal shift assay (CETSA) is a valuable method to confirm target engagement within a complex cellular environment, by detecting changes in a protein’s thermal stability upon ligand binding. The classical CETSA method measures changes in the thermal stability of endogenous proteins using immunoblotting, which is low-throughput and laborious. Reverse-phase protein arrays (RPPAs) have been demonstrated as a detection modality for CETSA; however, the reported procedure requires manual processing steps that limit throughput and preclude screening applications. We developed a high-throughput CETSA using an acoustic RPPA (HT-CETSA-aRPPA) protocol that is compatible with 96- and 384-well microplates from start-to-finish, using low speed centrifugation to remove thermally destabilized proteins. The utility of HT-CETSA-aRPPA for guiding structure-activity relationship studies was demonstrated for inhibitors of lactate dehydrogenase A. Additionally, a collection of kinase inhibitors was screened to identify compounds that engage MEK1, a clinically relevant kinase target.

Demonstrating a direct interaction between a small molecule and protein target can be difficult to establish in a cellular environment, highlighting the importance of target engagement assays ¹. Small molecule activity in biochemical assays does not always translate to on-target activity within cells due to environmental complexities, including membrane barriers and a crowded protein environment that can hinder a compound’s ability to interact with a target. The cellular thermal shift assay (CETSA) was developed to assess on-target binding within cells by detecting changes in a protein’s thermal stability upon ligand binding ². Compound-induced alterations in thermostability can shift the aggregation temperature (T_agg) of a protein target, which can be measured by quantifying the amount of soluble protein remaining after heating and lysing cells. CETSA enables target engagement studies without requiring modifications to the protein(s) or compound(s) of interest.

The original and most commonly used CETSA detection method is western blot, where soluble proteins are separated by SDS-PAGE and then detected using target-specific antibodies ². This approach, however, is low-throughput and labor-intensive. Several methods have been developed to increase the throughput of CETSA, but each comes with its own limitations ³. The AlphaLISA detection system quantifies endogenous protein levels using antibodies fused to donor and acceptor beads that emit light when brought into close proximity ^4–6. This method is limited by the availability of target-specific antibody pairs with distinct epitopes that are compatible with energy transfer. Reporter-based systems such as NaLTSA, SplitLuc CETSA, and InCELL Pulse Assay rely on the heterologous expression of targets with appended reporter tags that facilitate the detection of soluble protein after heating, typically using a luminescence-based readout ^7–9. While these protein reporter methods improve throughput, the potential effects of the reporter tag on protein function, and the consequences of transgene overexpression, may confound the interpretation of results. Although current high-throughput CETSA options are suitable for some targets, there remains a need for high-throughput CETSA methods that allow detection of endogenous proteins using single, unmodified commercially available antibodies without extensive optimization.

The reverse-phase protein array (RPPA) is a well-established, high-throughput method for protein detection by immunoblotting, in which nanoliter volumes of lysate containing a complex mixture of proteins are transferred directly to a membrane for interrogation with an antibody. Recently, the technology has been coupled with the acoustic transfer of lysates, which permits the replacement of traditionally used nitrocellulose-coated slides with a nitrocellulose membrane immobilized within a reusable 3D-printed holder ¹⁰. Herledan et al. established that the acoustic RPPA (aRPPA) could be used for protein detection in CETSA to increase sample throughput on a single membrane at the protein detection stage ¹¹. However, the majority of the assay was performed in individual PCR tubes to allow for a high-speed centrifugation step after heating and cell lysis to remove aggregated proteins. The requirement to process samples in PCR tubes/strips and subsequent sample transfer manually limits the number of samples that can be processed using the described method. Here, we developed a high-throughput aRPPA-based CETSA assay to measure endogenous intracellular proteins in a 96- or 384-well plate format. The assay was validated using lactate dehydrogenase A (LDHA) and a set of 29 well--characterized small-molecule LDHA inhibitors ¹² that were tested in dose-response. Additionally, we demonstrate that the assay is amenable to quantitative high-throughput screening (qHTS) by probing for MEK1 binders within a collection of 123 kinase inhibitors.

Results and Discussion

Classical CETSA experiments utilize SDS-PAGE and western blotting, a method that is labor-intensive and low-throughput, which limits utility for SAR and screening campaigns despite the value that early-stage target engagement results can provide. Our goal was to develop a high-throughput CETSA assay with the following characteristics: (1) the capacity to detect unmodified endogenous proteins; (2) compatible with a variety of target classes; (3) high throughput (≥96-well plates) from start-to-finish; and (4) cost-effective. To accomplish this, we focused on traditional CETSA steps that limit throughput (Figure 1A). First, we replaced western blot detection with an acoustic RPPA (aRPPA), which allows quantification of more protein samples on a single immunoblot. By utilizing acoustic droplet ejection (acoustic dispensing), nanoliter volumes of lysate can be transferred directly onto a nitrocellulose membrane secured in a 3D-printed holder with identical dimensions to a microwell plate, enabling its use as a destination plate in acoustic dispensing devices ¹⁰. This modification removes the need to separate proteins by SDS-PAGE and enables up to 1,536 lysate samples to be detected on a single membrane, saving both time and resources.

Figure 1.

High-throughput acoustic RPPA CETSA for target engagement. (A) Schematic overview of the HT-CETSA-aRPPA approach, compatible with 96- and 384-well plate formats across the entire workflow. (B) High-speed centrifugation is not required to detect target aggregation and compound engagement. Endogenous LDHA expression in HEK293 cells was examined after centrifugation at 12,000g for 15 min (high spin) or 2000g for 30 min (low spin). Densitometry of immunoblots following SDS-PAGE was performed. (C) Acoustic RPPA immunoblot probed for LDHA using heated (48–72 °C) or unheated (37 °C columns) samples. A well-characterized LDHA inhibitor (compound 63, 10 μM) was added to cells 1 h before heating. Cell lysates were centrifuged at 2000g for 30 min before acoustic transfer of lysates. (D) Densitometry of the full immunoblot in panel (C) was performed. Heated samples were normalized to 37 °C (mean ± SD, n = 9). Temperature range was not sufficient to determine T_agg for compound 63.

HT-CETSA-aRPPA Assay Development

LDHA is an abundant protein for which NCATS has developed a series of potent small-molecule inhibitors that have undergone testing in various target engagement and enzymatic assays, ¹² providing a rich set of tools for benchmarking the high-throughput CETSA using acoustic RPPA (HT-CETSA-aRPPA). Before transitioning into CETSA optimization, we first verified antibody specificity to LDHA. Unlike traditional SDS-PAGE, antibody specificity in RPPA is paramount because proteins are not separated by size before transferring to nitrocellulose; therefore, antibody specificity cannot be assessed through molecular weight nor is it possible to exclude non-specific bands from quantitative analyses. We utilized the MD Anderson RPPA database ^{13, 14} to select a previously tested LDHA antibody, which we verified for specificity using traditional SDS-PAGE and western blotting. A single band was detectable in both HAP1 and HEK293 cells, but not HAP1-LDHA knockout cells (Figure S1A). The antibody was also tested in the aRPPA format by spotting lysate from HAP1-LDHA KO and HEK293 cells directly onto a membrane to ensure specificity using a non-resolved complex mixture of proteins not subjected to SDS-PAGE (Figure S1B,C). An immunoblotting signal on the aRPPA was quantified using the free ImageJ Protein Array Analyzer macro, which detects a signal in evenly spaced circular regions overlaid on the image. Immunoblotting signal linearity was measured by spotting lysates prepared from serially diluted cells, showing a proportional change in LDHA intensity with increasing cell number between 6.25 × 10⁴ and 4 × 10⁶ cells/mL, indicating that changes in abundance related to altered thermal stability would likely be detectable using a sample input of 1 × 10⁶ cells/mL (Figure S1D,E).

In traditional CETSA, a high-speed centrifugation of ≥12,000g is performed following cell lysis to pellet insoluble aggregates formed during heating. Because microwell PCR plates cannot be centrifuged at speeds exceeding 2000–3000 RCF, a transfer of samples to individual tubes or tube strips is required to achieve 12,000g, which significantly limits throughput. However, the capacity of lower speed centrifugation to sufficiently remove aggregates from the soluble fraction has not been extensively defined. We explored this by comparing the temperature response profile and ligand-induced thermal shift of LDHA using a high-speed (12,000g) or low-speed (2000g) centrifugation step following cell lysis. HEK293 cells were treated with a vehicle or 10 μM compound 63, a well-characterized LDHA inhibitor ¹², and heated on a gradient from 58 to 82 °C. After separating soluble proteins by SDS-PAGE, the resulting western blots revealed that aggregation profiles of LDHA were comparable between low- and high-force conditions, with a ~9 °C thermostabilization induced by compound 63 under both conditions (Figure 1B, Figure S1F). We next assessed whether a low-speed centrifugation step would also be compatible with the aRPPA method. HEK293 samples were treated with a vehicle or 10 μM compound 63, heated, and lysed in 384-well PCR plates, and then centrifuged at 2000g. Next, lysate was transferred from the top of the sample into a Labcyte Echo 525 acoustic source plate using an Apricot Liquid Handler. Samples were then acoustically transferred to a nitrocellulose membrane. A robust, inhibitor-induced thermostabilization of LDHA was detectable by aRPPA, similar to the traditional CETSA protocol (Figure 1C,D). The HT-CETSA-aRPPA method was also examined using lysates without a spin step, but we observed less consistency in target melting behavior, which may indicate that the antibody can access the epitope despite the target being buried in an “insoluble” aggregate, if not otherwise removed by centrifugation (Figure S1G).

Next, we examined the reproducibility of the HT-CETSA-aRPPA approach under isothermal conditions. First, to assess the contributions of sample-to-sample (biological) and technical (acoustic transfer) variability, we transferred samples from a single 384-well source plate, half of the samples treated with compound 63 and the other half treated with a dimethylsulfoxide (DMSO) control, four times across a membrane, for a total of 768 vehicle-treated and 768 compound-treated lysate spots. Prior to lysis, the cells were heated to 74 °C, a temperature at which the maximum difference in thermal stability between two populations was expected. There was a high degree of similarity between quadrants with comparable signal intensity for each spot that originated from the same source well, suggesting that most of the variability was of biological rather than technical origin (Figures 2A and S2A,B). Blots were analyzed by two densitometry methods: (1) the ImageJ Protein Array Analyzer, a macro tool that is used to analyze dot blots; and (2) an in-house MATLAB script developed to better automate the selection and quantification of spots on high-density immunoblots. With both analysis methods, two distinct populations were observed, with compound-treated wells averaging ~8x greater signal than vehicle-treated wells (Figures 2B and S2C). Signal intensity was highly correlated between the two methods (Figure S2D).

Figure 2.

Ligand-induced thermal stabilization of LDHA under isothermal conditions. (A) Reproducibility of the HT-CETSA-aRPPA was examined by transferring lysate from a 384-well plate repeatedly to four quadrants (technical replicates). Lysates from vehicle- (odd columns) or 10 μM compound 63-treated (even columns) cells were spotted onto the membrane. (B) Densitometry of immunoblot presented in panel A using MATLAB script to identify spots and quantify signal for vehicle-treated (n = 768) or 10 μM LDHA inhibitor compound 63-treated samples (mean ± SD, n = 768). **** p < 0.0001, two-tailed unpaired t-test. (C) Target engagement and thermal stabilization was examined for 29 LDHA inhibitors, in dose-response. Each quadrant represents an independent biological replicate, either unheated (Q1) or heated to 71 °C (Q2-Q4). (D) Quantification of panel (C), showing dose-dependent thermal stabilization of LDHA conferred by small-molecule inhibitors. (E) Potency of thermal stabilization (AC₅₀, Log M) for LDHA inhibitors across three replicates. (F) HT-CETSA-aRPPA potency values (AC₅₀, Log M) for LDHA inhibitors correlated with the previously characterized SplitLuc CETSA method.

The ImageJ Protein Array Analyzer offers a free solution ¹⁵, but because the software overlays a rigid grid, it was not flexible to adapt to misaligned spots or variations in spot size, which sometimes occurs during acoustic dispensing. Therefore, we created a MATLAB script to better handle deviations in spot location or size by automatically segmenting a spot and calculating signal intensity within that region. The MATLAB script was used for subsequent experiments because of its ease of use and adaptability to deviations in spot location and size on a membrane.

To characterize the ability of the HT-CETSA-aRPPA method to capture dose-dependent target engagement and contribute to structure activity-relationship campaigns, we tested a set of 29 NCATS-developed LDHA inhibitors (Table S1). Lysates from four 384-well plates were transferred to a single membrane, corresponding to biological triplicates of heated samples (71°C) and a single plate transfer of an unheated control (Figure 2C). By including samples derived from an unheated plate, we were able to verify that changes in the signal for the heated plates were due to compound-induced thermal shift, rather than confounding factors such as compound cytotoxicity or altered target expression which could impact interpretation. Each heated sample was normalized to its corresponding sample in the unheated plate, and dose-dependent thermal stabilization was observed (Figure 2D). The AC₅₀ values calculated from the three independent heated plates showed similar global rank ordering (Figure 2E, Table S1). The AC₅₀ values also correlated with the potency of target engagement calculated using the previously described SplitLuc CETSA method (Figure 2F), albeit with slightly left-shifted potency values observed for the HT-CETSA-aRPPA. Two datapoints, corresponding to two samples of GSK-2837808A, were noticeably less potent in the HT-CETSA-aRPPA (Figure 2F). Our previous work on GSK-2837808A also showed a discrepant potency in SplitLuc CETSA compared to an orthogonal cell-based lactate assay for this compound (1.7 μM vs 19.2 μM, respectively); ⁷ therefore, the HT-CETSA-aRPPA result appears to be in better agreement with the compound’s effect on lactate production in HEK293T cells.

High-Throughput Screening Using HT-CETSA-aRPPA

Next, we explored the potential for HT-CETSA-aRPPA as a screening platform and performed a quantitative high-throughput screen (qHTS) for target engagement of MEK1, a kinase implicated in a variety of cancers. A MEK1 antibody previously validated for RPPA (MD Anderson) produced a single band by western blot (Figure S3A), and signal linearity in aRPPA format was confirmed using spotted lysate from serially diluted HEK293 cells (Figure S3B,C). Because the expected T_agg of MEK1 is lower than that of LDHA, we considered the possibility that the insoluble aggregates formed during heating would be smaller in size and respond differently to low-speed centrifugation. We confirmed by western blot that low-speed centrifugation (2000g for 30 min) was sufficient to remove insoluble MEK1 aggregates (Figure S3D). The difference in MEK1 T_agg was <0.5 °C between the high- and low-speed centrifugation groups, indicating that the low-speed centrifugation was compatible with performing MEK1 CETSA experiments (Figure 3A). Under both centrifugation conditions, the well-characterized MEK1 inhibitor Selumetinib thermally stabilized its target by approximately 5 °C. Following our previous workflow, we confirmed similar melting behavior and compound-induced stabilization in the HT-CETSA-aRPPA format for samples processed with a low-speed centrifugation step. Under these conditions, MEK1 T_agg was calculated at 54.4 °C, and thermal shift induced by Selumetinib was +4.2 °C (Figures 3B and S3E). While we did not further validate the MEK1 antibody using knockout or knockdown approaches, the similarity in response to Selumetinib by traditional western (where a single band of the correct size was observed) was consistent with specific detection of MEK1 in the aRPPA format.

Figure 3.

MEK1 target engagement screen using a kinase inhibitor-focused small molecule collection. (A) High-speed centrifugation is not required to detect MEK1 target aggregation and ligand-induced thermal shift. Endogenous MEK1 expression in HEK293 cells was examined after centrifugation at 12,000 × g for 15 minutes (high-spin) or 2,000 × g for 30 minutes (low-speed). Densitometry of immunoblots following SDS-PAGE was performed. (B) Temperature response profile of MEK1 by HT-CETSA-aRPPA reveals compound-induced thermostabilization. HEK293 cells were treated with DMSO or 10 μM Selumetinib for one hour before heating (50–74°C). Densitometry signal was normalized to unheated samples (mean ± SD, n = 3). (C) aRPPA immunoblots for kinase inhibitor screen, for MEK1 (top) and total protein (bottom). Compounds were tested in dose-response. U= unheated samples. (D) Active compounds identified in the qHTS of 123 compounds for MEK1 target engagement.

To facilitate processing of screening plates without prolonged centrifugation, we tested shorter centrifugation durations following cell lysis. The temperature response profile of MEK1 in HEK293 cells was examined for lysates centrifuged at 2000g for 0, 10, 20, or 30 min before transfer to an acoustic source plate and spotting to a nitrocellulose membrane. The MEK1 temperature response profile was comparable for the 10–30 min samples, indicating that the shorter centrifugation step could be employed to improve throughput (Figure S3F). Notably, a thermal melt was not detectable without centrifugation (Figure S3F). For the high-throughput screen, we selected 15-min centrifugation.

HEK293 cells were screened against a library composed of 123 kinase inhibitors (Table S2) by treating for 1 h in dose-response (final concentration range = 1 nM to 48 μM), heating to 58 °C, and processing for aRPPA (Figure 3C). The membrane was stained for total protein prior to immunoblotting for MEK1 (Figure S3C), and the MEK1 signal was normalized to total protein for each spot. A collection of 15 MEK1 inhibitors (approved, clinical, and pre-clinical phases) and 24 US FDA drugs targeting other kinases were included. All fifteen MEK inhibitors had a positive readout in the MEK1 acoustic-CETSA assay, while only two of twenty-four non-MEK drugs were weakly positive (Fig 3D, Table S2). Of the two unexpected positives, ponatinib has prior literature data demonstrating it is not a MEK1 binder or inhibitor ^{16, 17}. Vandetanib has conflicting MEK1 binding data, showing 1.8 micromolar potency in a KINOMEscan profiling assay ¹⁸ but no detectable binding in a kinobeads profiling assay ¹⁷. The basis of thermal stabilization of MEK1 by these two compounds in our experiments is not understood. The remaining kinase inhibitors in the collection were inactive in the screen (Table S2).

Conclusions

In summary, we developed a high-throughput RPPA-based CETSA that measures unmodified protein in 96- or 384-well formats. Using LDHA as a target, we observed a dose-dependent ligand-induced thermal stabilization that was reproducible across biological replicates and comparable to an alternate high-throughput method, SplitLuc CETSA. We also demonstrated the ability of the method to detect target engagement for MEK1 by screening a kinase inhibitor-focused set, in dose-response. Overall, the HT-CETSA-aRPPA method provides a platform that can be used for a variety of target classes, is easy to implement, and is capable of testing a large number of samples.

The AlphaLISA CETSA offers the most direct methodological counterpart to the HT-CETSA-aRPPA, as both are high-throughput and use an antibody-based system to measure unlabeled endogenous protein stabilization. However, the AlphaLisa detection system relies on the availability of paired, distinct target-specific antibodies capable of energy transfer which can be cost-prohibitive and in some cases requires extensive optimization. HT-CETSA-aRPPA utilizes a single antibody with suitable specificity and sensitivity for the target of interest, making this method adaptable to a variety of targets. HT-CETSA-aRPPA is not restricted to a single cell type, so biologically relevant cell models can be selected based on the target under investigation. In the method described herein, HEK293T and HAP1 cells were heated in suspension, so density could be adjusted to accommodate differences in target protein expression levels and antibody sensitivity. We envision that this method could also be adaptable to adherent cell culture, heating cells in plates through conductive heat transfer from a metal block or heated water bath ^{7, 19, 20}.

We used 384-well plates to maximize throughput, but this method is also compatible with 96-well format. For cell lysis, we utilized a detergent rather than freeze/thaw cycles to streamline the workflow. A concentration of 0.3% NP-40 was preferred, as higher detergent concentrations (1%) exhibited a halo effect in the spots on the membrane, with highest signal intensity at the boundaries of the spots. The described procedure includes a transfer of lysate from PCR plates to acoustic-compatible plates, as PCR plates cannot be used directly for acoustic dispensing. While this transfer step was automated using a liquid handling device, the protocol could be further streamlined by using a hanging drop with a tip- or pin-based method to directly transfer lysate from the top of the PCR-plate well to a membrane after centrifugation. We found acoustic dispensing to be the most reliable method, but observed similar results with other transfer modes, including direct transfer by hand. While we did not try processing the samples in 1,536-well plates due to the limitations of our acoustic dispenser for aqueous samples, this is expected to be achievable and the focus of future investigation. For the immunoblotting surface, a reusable 3D-printed membrane holder and nitrocellulose membrane were selected and performed reliably, but nitrocellulose-coated slides can also be considered. Transfer volume was adjusted depending on the total number of spots on the membrane to avoid overlapping spots (200 nL for 384 and 50 nL for 1,536). Because only a small volume of sample is needed per membrane, multiple copies can be made and if stored properly, can be saved and processed at a later time. When probing with antibody, multiplexing is possible to examine multiple targets or to incorporate a reference control, either a single protein or total protein, for normalization. Including unheated protein controls can also identify rapid effects of small molecules on the target protein’s levels, which can enhance interpretation of isothermal CETSA results.

As with all CETSA methods, HT-CETSA-aRPPA is prone to false negatives. Importantly, the absence of a thermal shift does not conclusively demonstrate a compound is not binding, because some binding events do not change the thermal stability of a protein, as has been demonstrated for JAK2 and ABL1 ^{8, 21}. Additionally, cell membrane integrity becomes disrupted at temperatures above ~60 °C ^{7, 9}, so results should be interpreted with caution for targets with a T_agg that require heating above this temperature. There are several other considerations that more specifically apply to HT-CETSA-aRPPA. Scenarios where multiple isoforms of the target are expressed and detected by an antibody can be particularly challenging, and CETSA for these targets may require traditional westerns to resolve the desired isoform using molecular weight information. Because lysates for HT-CETSA-aRPPA are not processed under denaturing conditions, the more native state could also impact antibody binding and specificity. Validating antibodies by SDS-PAGE immunoblotting can provide some confidence that the antibody specifically binds to the intended target, but this will not always translate to RPPA where proteins are not fully denatured. Whenever possible, KO cells should be employed to verify specificity in the context of RPPA.

Overall, the HT-CETSA-aRPPA method provides a cost- and resource-effective approach to screen hundreds to thousands of small molecules, in dose-response, to characterize target engagement for endogenously expressed proteins.

METHODS

Cell culture:

HEK293 cells (ATCC CRL-1573) were grown in DMEM containing high glucose, 1X Glutamax, 110 mg/L sodium pyruvate, 10% FBS (v/v), 100 U/ml penicillin, and 100 U/mL streptomycin (Gibco, Cat #10569010). HAP1 and LDHA KO cells were cultured in IMDM containing L-Glutamine, 25 mM HEPES, and 10% FBS (Gibco, Cat #12440053). Cells were incubated at 37°C with 5% CO₂ and were routinely tested for mycoplasma using a Lonza MycoAlert kit. Cells were not used past passage 20.

Immunoblotting:

0.45 micron nitrocellulose membranes (Amersham Protran, Cat #10600003) were blocked for 30 minutes using 5% milk (w/v) in Tris-buffered saline with 0.1% Tween-20 (TBS-T), then incubated with primary antibody overnight at 4°C with gentle shaking. Primary antibodies were anti-LDHA (Cell Signaling Technology, Cat #3582T, used at 1:1000 for western blot and 1:300 for RPPA) and anti-MEK1 (Abcam, Cat #ab32576, used at 1:2500 for western blot and 1:1500 for RPPA). The following morning, the membranes were washed 3x with TBS-T for 5 minutes, then incubated with a Li-Cor IR800 secondary antibody (used at 1:15,000 for western blot and 1:5000 for RPPA) for 1 hour at room temperature with gentle shaking. Next, the membranes were washed 3x with TBS-T for five minutes and rinsed briefly with TBS. The membranes were imaged on a Li-Cor CLx scanner.

aRPPA quantitation:

For the ImageJ Protein Array Analyzer, the raw TIFF file from the scanner was first imported using the Bio-Formats Importer. The LUT was inverted, then the image was cropped to only include the signal area and contrast was adjusted. Using the Protein Array Analyzer, a linear background subtraction was completed on the image. Next, the corners were selected on the blot and the detection size for each spot was adjusted. The software would then overlay evenly spaced analysis circles and give output values. Digital luminescence images are analyzed using a customized MATLAB (The Mathworks, Inc.) script (public file source). The automated workflow is briefly described in the following section. Prior to running the analysis, the user adjusts script parameters to account for the size of the square bounding box region surrounding each sample in the image (estimated number of pixels, width and height). When executed in the MATLAB computational environment (MATLAB version 2021a) the script functions allow the user to select a specific image file for display, and a graphical user interface prompts the user to locate the positions of the upper left sample, the upper right sample, and the lower left sample. The user also enters the total number of sample rows and columns that are in the selected contiguous block of samples. Based on the above information that is supplied by the user, the script functions adjust the image to correct for any rotation of the samples and applies an evenly spaced analysis grid. Each grid region contains luminescence information from a single sample. An automatic threshold is applied to each region to segment the signal region from the local background region within each grid region. The size (in pixels) of the signal region is reported and can be automatically gated to eliminate irregular signals (for example, small false positive speckle noise regions or large false positive regions when no true signal region is present in the grid). Numerical values for mean signal intensity per signal region, mean local background intensity, and mean pixel intensity region minus local background are reported for each analyzed grid region.

Antibody specificity by WB and aRPPA:

HEK293, HAP1, and LDHA KO cells were lifted using 0.25% trypsin and resuspended in their respective phenol-, antibiotic-, and serum-free media. Cells were lysed by adding NP-40 and protease inhibitors to a final concentration of 0.3% and 1X, respectively, incubating at room temperature for 30 mins, then centrifuging at 12,000 × g for 15 mins. The supernatant was collected, total protein concentration was determined using a DC Protein Assay (Bio-Rad), and 60 μg of each sample in 1X reducing agent (Thermo) and 1X LDS sample buffer (Thermo) was separated on a 4–12% Bis-Tris NuPage gel (Thermo) in 1X MOPS buffer (Thermo). The gel was transferred to a nitrocellulose membrane using an iBlot2 transfer device (Thermo), and immunoblotting was performed as described above. Lysate from the western blot specificity experiment was used to confirm specificity by aRPPA. 10 μL per well of LDHA KO and HEK293 lysate was loaded into a Labcyte Echo LDV source plate, then 200 nL (~1 μg) per sample was transferred to a nitrocellulose membrane using an Echo 525. Protein spots were allowed to dry for at least 10 mins at room temperature. Once dry, the membrane was rehydrated with TBS then proceeded through the immunoblotting steps described above.

Antibody linearity:

HEK293 cells were lifted using 0.25% trypsin and resuspended in a phenol-free high glucose DMEM + Glutamax. Cells were serially diluted in tubes then lysed with NP-40 (0.3% final) and protease inhibitors (1X final) for 30 mins. Next, 10 μL of lysate per well was transferred to a Labcyte Echo LDV plate, and 200 nL per sample of lysate was spotted onto a nitrocellulose membrane using an Echo 525. The membranes were allowed to dry, then were rehydrated with TBS before going through the immunoblotting steps described above.

High-throughput CETSA using aRPPA:

HEK293 cells were lifted using 0.25% trypsin then resuspended (1 × 10⁶ cells/mL for LDHA and 2.5 × 10⁶ cells/mL for MEK1) in phenol-free DMEM + Glutamax. DMSO or compound was acoustically transferred to 384-well PCR plates (Roche LightCycler 480 Multiwell plate, white, Cat #04729749001) using an Echo 555, after which 20 μL of cells were dispensed into each well by a Multidrop Combi. Cells were then incubated at 37°C for 1 hour. Following compound incubation, cells were heated at a single temperature or on a gradient for 3.5 minutes using a 384-well thermocycler. Once the cells cooled to room temperature, they were lysed for 30 minutes with 0.3% NP-40 and 1X protease inhibitor final concentrations that was dispensed into the plate using a Multidrop Combi, then mixed using an Apricot Personal Pipettor. After lysing, plates were centrifuged at 2,000 × g for 30 minutes, then the top 10.5 μL of lysate was transferred to an Echo LDV source plate using an Apricot Personal Pipettor. Next, 200 nL (384 format) or 50 nL (1,536 format) of lysate was transferred from the source plate to a nitrocellulose membrane using an Echo 525. After drying for at least 10 minutes, the membranes were hydrated with 1X TBS before immunoblotting.

Classical CETSA:

Lysate from corresponding aRPPA experiments were used. Lysate was centrifuged for 30 min at 2,000 × g (low-speed spin) or for 30 min at 2,000 × g followed by 15 mins at 12,000 × g (high-speed spin). The supernatant was collected and loaded onto a 4–12% Bis-Tris NuPage gel and run in 1X MOPS buffer. Next, the gel was transferred to a nitrocellulose membrane using an iBlot2, then the membrane was processed using the immunoblotting steps described above.