FLECS Technology for High-Throughput Screening of Hypercontractile Cellular Phenotypes in Fibrosis: A Function-First Approach to Anti-Fibrotic Drug Discovery

The pivotal role of myofibroblast contractility in the pathophysiology of fibrosis is widely recognized, yet HTS approaches are not available to quantify this critically important function in drug discovery. We develop, validate, and scale-up a HTS platform that quantifies contractile function of primary human lung myofibroblasts upon treatment with pro-fibrotic TGF-β1. With the fully automated assay we screened a library of 40,000 novel small molecules in under 80 h of total assay run-time. We identified 42 hit compounds that inhibited the TGF-β1-induced contractile phenotype of myofibroblasts, and enriched for 19 that specifically target myofibroblasts but not phenotypically related smooth muscle cells. Selected hits were validated in an ex vivo lung tissue models for their inhibitory effects on fibrotic gene upregulation by TGF-β1. Our results demonstrate that integrating a functional contraction test into the drug screening process is key to identify compounds with targeted and diverse activity as potential anti-fibrotic agents.


Introduction 1
Fibrosis is a chronic pathological wound repair process characterized by progressive tissue scarring, resulting 2 in the replacement of normal tissue with thickened and stiffened non-functional scar extracellular matrix (ECM).

3
Central to the disease process that can affect all organs is the activation and maintenance of fibroblasts and 4 other progenitors into myofibroblasts. Myofibroblasts in fibrotic organs exist in a spectrum of activation states 5 that start with the excessive deposition and remodeling of collagen-rich ECM. In the process, myofibroblast 6 develop contractile actin-myosin bundles (stress fibers) that incorporate alpha smooth muscle actin (α-SMA), 7 resulting in a 'super-contractile' phenotype. It is the transmission of high contractile force that leads to ECM 8 stiffening and formation of a scar that impedes and often obliterates organ function (Fig. 1). In addition to 9 compromising normal organ function, stiffened ECM perpetuates fibrosis by driving further activation of 10 mechano-sensitive and mechano-responsive myofibroblasts. Another part of this fibrotic feed-forward loop is the 11 mechanical activation of TGF-β1 from latent complexes in the ECM 1,2 by force transmission via the αv integrins 12 of contracting fibroblastic cells 3,4 . TGF-β1 is a master regulator of fibrosis that, among multiple pro-fibrotic 13 actions, promotes further fibroblast-to-myofibroblast conversion via the canonical Smad2/3 pathway and non-14 canonical signaling pathways 1,5,6 . Finally, myofibroblast contraction-induced strain may protect collagen ECM 15 from enzymatic degradation 7-9 , all resulting in a net accumulation of ECM and overall stiffer fibrotic tissue.

19
Myofibroblasts mechanically activate and release Transforming growth factor-beta 1 (TGF-β1) from its latent stores within the ECM, 20 promoting further fibroblast-to-myofibroblast conversion. (3) Concurrently, strain-mediated protection from enzymatic degradation occurs 21 within fibrotic tissue under stress from myofibroblast contraction, resulting in an imbalance between ECM synthesis and degradation, 22 which ultimately leads to net ECM accumulation and tissue thickening.
Despite the widely recognized role of myofibroblast contractility in the pathophysiology of fibrosis, most in vitro 1 models of fibrosis, focus primarily on molecular markers of activated myofibroblasts , e.g. expression of the 2 contractile protein alpha-smooth muscle actin (α-SMA) or ECM proteins like pro-collagen, collagen, and 3 fibronectin 10 . Functional assessment of hypercontractile myofibroblasts is often neglected, yet enhanced 4 contraction is a defining feature of myofibroblasts 11, 12 . This neglect is partly due to the lack of straight-forward 5 quantitative assays to measure cell contraction. Functional two-dimensional cell growth assays to measure 6 myofibroblast contraction in vitro include traction force microscopy (TFM) 10 and 'wrinkling' assays 11 . Both 7 approaches are based on the principle that the contraction of adherent cells results in deformations of planar 8 deformable culture substrates that can be traced through surface marker displacements in TFM or formation of 9 visible large folds (wrinkles). The most widely used three-dimensional myofibroblast contraction assays are 10 based on the diameter reduction of 3D collagen gels by mixed-in fibroblast populations 13,14 , or measurements 11 with directly attached force transducers 12 . Three-dimensional contraction assays cannot deliver information on 12 contraction force of single cells and none of the two-dimensional assays assessing single cell contraction are 13 suitable for higher throughput screening applications. Thus, there is a pressing need for a quantitative and 14 scalable assay that can assess contractility in a fibrosis-relevant context. Such an assay would enable direct 15 screening for novel anti-fibrotic drugs and genes that selectively modulate contractile cell function.

16
We here develop and validate a functional-phenotypic assay evaluating myofibroblast contractility as a directly 17 observable and quantifiable endpoint, scaled for high-throughput screening (HTS). We built the assay on a 18 previously established high-throughput cell contractility screening platform, that we coined fluorescent 19 elastomeric contractible surfaces (FLECS) [13][14][15][16][17][18][19] . In a screening campaign conducted over 15 days, we tested 20 40,000 small molecules and identified over a dozen inhibitors of myofibroblast contraction with sub-micromolar 21 potency. Our results illustrate the capacity of the My-FLECS assay (FLECS technology applied to screening 22 myofibroblast contractility) to efficiently identify novel starting points for the development of anti-fibrotic drugs 23 within expansive chemical libraries. The inhibitors identified also demonstrated selectivity toward fibroblasts over 24 smooth muscle cells in both the primary assay and in counter-screens, along with the ability to suppress fibrotic 25 gene upregulation in the human precision-cut lung slice model. Collectively, our study underscores the potential 26 of the new screening assay in uncovering novel anti-fibrotic drug activities, which may not be predictable from 27 molecular readouts alone.  Trypsin-EDTA (0.05%) was used to re-suspend cells at the start of the experiment. For HLF, experiments in the 2 FLECS assay were done in Dulbecco's Modified Eagle Medium (DMEM) supplemented with 10% FBS and 1% 3 penicillin-streptomycin. For HTSM, experiments were done in the same Ham's F12 medium as was used for 4 culturing. All experiments were done using cells at passage 3.

6
My-FLECS contractility assay general protocol 7 The My-FLECS contractility assay was performed using an adaptation of the original FLECS protocol described 8 previously 14 with modifications made specifically to tailor the assay to use with HLF and TGF-β1, for the first 9 time. To initiate the assay, empty 384-well FLECSplates were filled with 25µL of serum-free Dulbecco's Modified 10 Eagle Medium (DMEM) supplemented with 1% penicillin-streptomycin. HLF cells were detached from their flasks 11 and resuspended at a concentration of 50,000 cells/mL in DMEM/10%FBS medium. TGF-β1 (Peptrotech) was 12 optionally added to this cell suspension at desired concentrations prior to seeding. The cells were then seeded 13 into the pre-filled FLECSplates, with 25µL being added to each well. After seeding, plates were left at RT for 1.5 14 hrs to allow for uniform settling of cells to the well bottom after which time plates were moved into a 37 C incubator 15 for 24 hrs.

17
At 24hrs, 10µL of DMEM/10%FBS containing 1µg/mL Hoechst 33342 live nuclear stain was added to each well 18 30 minutes prior to imaging using a Molecular Devices ImageXpress automated microscope. Image analysis 19 was performed using Forcyte's proprietary computer vision algorithms as previously described 14 .

21
It was determined that 70µm sized "X" shaped micropatterns consisting of type IV collagen arrayed over a 22 stiffness of approximately 8 kPa (Forcyte Biotechnologies, product 384-HC4R-QC10) were the optimal assay 23 parameters to support HLF adhesion and contraction both at baseline and following activation with TGF-β1.

24
These parameters were used in all experiments herein.

27
The optimal concentration of TGF-β1 was determined by testing the contractile response of HLF over a 24hr 28 period against a row-wise 10-step 2-fold dilution series of TGF-β1 starting at 20ng/mL. The experiment was 29 performed as described above with appropriate amounts of TGF-β1 added to each cell suspension prior to 30 seeding into the FLECSplate. were shaken at 100 RPM for 15 minutes to ensure mixing. HLF cells were seeded onto the FLECSplates without TGF-β1 and incubated with the ALK-5 inhibitors for 30 minutes. Following incubation, TGF-β1 treatment was 1 performed by adding 10 µL of media containing TGF-β1 to columns 2-23, achieving a final concentration of 2.5 2 ng/mL, while column 1 received 10 µL of media without TGFβ as the negative control. Plates were imaged after 3 24hrs as previously described. Percent inhibition was calculated by normalizing measured contraction values 4 against both positive controls (e.g. no TGF-β1) and negative controls (TGF-β1 and vehicle only).

11
FLECSplates were shaken at 100 RPM for 15 minutes to ensure mixing. HLF cells were then seeded as 12 previously described, with TGF-β1 already mixed into the cell suspension, into columns 2-23, while column was 13 seeded with cells not receiving TGF-β1. Plates were imaged after 24hrs as previously described. Percent

32
On day 1, FLECSplates were filled with media, equilibrated to RT, given drug from the library plates (250nL of 33 1mM drug solution yielding a 3.5µM final concentration), then shaken on a plate shaker for 15 minutes. Next,

34
HLF were dissociated and seeded into drug-bearing FLECSplates as previously described with TGF-β1 in plate 35 columns 2-23 and without TGF-β1 in columns 1 and 24. On day 2, following 24 hrs of incubation, cells were 36 given Hoechst 33342 live nuclear stain as previously described and the plates were imaged with a 4x objective 37 at one position per well capturing both the micropatterns (TRITC channel) and stained nuclei (DAPI channel).
Images were subsequently downloaded in .TIFF format and analyzed using Forcyte's proprietary computer vision 1 to derive quantitative contractility data. Compounds were selected as hits for retesting from the primary screen using an "inhibition" cutoff off at least 5 75% but no more than 200% with respect to native HLF contraction. Hits were required to have a z-score on cell 6 count greater than -1.5.

36
Cell viability and proliferation counter-screens 1 HLF and HTSM cells were seeded into black plastic-bottom 384-well plates and treated with hits in dose-2 response format for 72 hours. Cell toxicity was assessed using the LIVE/DEAD Viability/Cytotoxicity Kit (Thermo 3 Fisher Scientific) according to the manufacturer's protocol. The kit employs fluorescent dyes, calcein AM and 4 ethidium homodimer-1 (EthD-1), to label live and dead cells, respectively. After staining, cells were imaged using 5 a fluorescence microscope, and the percentage of live and dead cells was quantified using image analysis 6 software. Immediately following the completion of the imaging of stained cells, CellTiter-Glo® 2.0 Cell Viability 7 Assay solution (Promega) was added to 384-well plates, per the manufacturer's protocol, and incubated for 30 8 minutes. A luminescence plate reader was then used to measure ATP activity corresponding to viable cell count 9 in each well. Also taking into account the live/dead calculation, inhibition of proliferation conferred by drug was 10 calculated as the percentage of the ATP activity signal normalized to negative control and subtracted from 100%.

13
Primary human lung fibroblasts (HLFs) were seeded in 96-well tissue culture plates pre-coated with a 0.2% 14 gelatin solution. After 48h, culture media was replaced with serum-free DMEM media containing 5uM hit 15 compounds. Cells were pretreated with compounds for 30min at 37℃ before TGF-β1 was added to a final 16 concentration of 2.5ng/ml. Two hours after TGF-β1 addition, cells were processed for Smad2/3 staining. Briefly, 17 cells were washed with PBS, fixed with 4% formaldehyde for 15min, and permeabilized with 0.2% Triton for 18 10min. After blocking with 3% BSA for 1hr, cells were incubated with rabbit anti-Smad2/3 antibody (Cell Signaling

24
Precision-cut lung slice assay 25 Frozen human precision cut lung slices (hPCLS) from healthy donors were purchased from AnaBios. The slices 26 were thawed, punched into 3mm discs using a biopsy puncher and cultured in DMEM/F12 media (supplemented 27 with 0.1% FBS and antibiotics/antimycotics) in 24-well tissue culture plates. 24hrs after thawing, lung slices were 28 pretreated with hit compounds for 1hr before adding TGF-β1 at a final concentration of 5ng/ml. Each condition 29 was run in quadruplicate. 48hrs after treatment, slices were homogenized in RLT lysis buffer on a Bead Mill 4 30 homogenizer using ceramic beads, and RNA was extracted using the RNeasy Mini kit (Qiagen). Reverse  To achieve HTS of contractile cell force, it is necessary to have a scalable and automated assay, where human 6 intervention is minimal and can be replaced with automation workflows and algorithms. This requirement applies 7 to every step of the screening process, from set-up to data output. To address this challenge, we previously 8 developed a scalable cellular force cytometer called FLECS Technology 14 .

20
The original FLECS assay adopts a standard well-plate format but replaces the traditionally rigid surface with a 21 highly compliant and elastic thin film. This film is provided with arrays of covalently micropatterned fluorescently 22 tagged adhesive ECM proteins (Fig. 2a). Non-patterned surfaces within the wells are chemically blocked to limit 23 cellular adhesion solely to the micropatterns. Single cells seeded from suspension self-assemble and adhere 24 over these micropatterns, applying inward contractile forces that cause significant changes in micropattern size.

25
Even minor size changes in the micropatterns (~1 µm) are visible under low magnification microscopy (e.g., 4X) and can be accurately quantified using computer vision algorithms (Fig. 2b). The precision of the micropatterns'    (Fig 3a). A 10-step 2-fold titration of TGF-β1 confirmed 2.5 ng/mL as the optimal dose for activating 27 myofibroblasts in the My-FLECS assay (Fig. 3b). Relative to the low baseline substantial tonic contractile forces 28 of untreated HLFs, TGF-β1-activated HLFs generated between 50% to >80% larger micropattern displacements  reports that fibroblasts are heterogeneous with respect to their spontaneous activation in vitro 21 , as reflected by 34 the wide distribution for untreated HLF, and that TGF-β1 treated fibroblasts become highly contractile 22 . During the optimization of My-FLECS, we observed that TGF-β1 responsiveness was maximized in fibroblast growth 1 factor (FGF)-free conditions. Unique to HLF, compared to other cell types tested with FLECS, was a heightened 2 sensitivity to pH variations in the medium, necessitating the need for precise pH buffering.

12
To validate the ability of our assay to identify inhibitors of HLF activation, we next performed a dose-response 13 experiment characterizing the effects of three known inhibitors of the TGF-β1 receptor I (RI) ALK-5 on HLF 14 contractility following 24 h treatment with TGF-β1. Following a 20-step 3-fold dilution series, the three ALK-5 15 inhibitors (SD-208, Galunisertib, and RepSox) were organized into a single drug source 384-well-plate (Fig. 3C) 16 for simultaneous assessment. As expected from TGF-β1-R1 inhibitors, exposure of HLF to all three ALK-5 17 inhibitors for 30 min prior to adding TGF-β1 inhibited activation of HLF into highly contractile cells. Although exceptionally effective in preventing TGF-β1-driven myofibroblast activation, the direct blockade of 3 TGF-β1 or of its receptor is not a viable clinical approach to therapeutically treat fibrosis due to the systemic 4 importance of TGF-β1 signaling 23 . Thus, we next sought to further validate our assay's ability to detect inhibition 5 activity occurring in pathways downstream of TGF-β1-TGF-β1-RI binding. We selected two inhibitors of the

17
We utilized the FLECS platform to evaluate the effects of these PIKfyve inhibitors on both HLF activation with 18 TGF-β1 and, in parallel, on the basal contractile tone of primary human tracheal smooth muscle (HTSM) cells.

19
HTMS cells were chosen since they also reside within the human respiratory system and are phenotypically 20 similar two myofibroblasts. Successful treatments of fibrosis ideally target myofibroblasts without affecting other 21 contractile cells in the target organ like HTMS (lung), vascular smooth muscle (all organs), myocytes (muscle), 22 or cardiomyocytes (heart). PIKfyve inhibition prevented TGF-β1-driven activation of HLF but did not antagonize 23 basal contraction of tracheal smooth muscle (Fig. 4). Hence, the My-FLECS assay is not only able to identify 24 inhibitors of HLF activation downstream of TGF-β1R1 binding but can also differentiate between general effects 25 on cell contractile mechanisms and myofibroblast-specific contraction. With these successful demonstrations, 26 the assay met the criteria for use in anti-fibrotic drug discovery and was ready for scale-up.

28
To achieve true HTS campaigns with the My-FLECS platform, we next developed an automated workflow to 1 screen up to 15,000 compounds per day, equating to approximately 48 384-well plates. Given the relatively low 2 number of reagent additions and wash steps required by our assay, as compared to traditional high-content 3 screening workflows where numerous solution exchange steps are required for fixing, blocking and multi-channel 4 staining of cells 25 , as well as the native well-plate format of the assay, our automation strategy was 5 straightforward (Fig. 5). FLECS 384-wellplates were manufactured in batches by Forcyte Biotechnologies and 6 stored with sterile PBS at 4°C covered with aluminum seals until their use. On the first day of a screening 7 experiment, automation systems were utilized to pre-fill each plate with cell growth medium, add compounds to 8 the plates using a pin tool, and seed HLF cells in medium containing TGF-β1. Following an overnight incubation 9 in the presence of the drug and TGF-β1, the automation system was used to deliver a live nuclear stain and 10 perform high-throughput imaging of all plates.

14
After establishing the necessary automated workflow, we evaluated the performance, variability, robustness, and 15 reproducibility of the assay by screening the complete LOPAC1280 library at a single concentration (3.5 µM) in 16 duplicate across 8 384-well plates in a single batch (Fig. 5). Columns 1 and 24 represented non-activated native negative controls for uninhibited myofibroblast activation. The robust Z-prime values ranged from 0.4 to 0.7 1 across the 8 plates. We defined 'hits' as compounds limiting contraction in TGF-β1-treated HLF by at least 75% 2 but no more than 200% with respect to native HLF contraction. Additionally, to eliminate potentially toxic 3 compounds, hits were required to have a z-score on cell count greater than -1.5. Fig.6 shows the reproducibility 4 of the assay and consistency in hits when run in an automated batch and thus, indicates that the assay is suitable 5 for single compound concentration screening.

12
Rapid primary HTS with >40,000 compound library in primary HLF and counter screening 13 Satisfied with the reproducibility and overall performance from the LOPAC1280 screen, we next implemented 14 the same automation strategy to execute out a HTS of a larger commercial sub-library. The goal was to identify 15 additional compounds that could inhibit TGF-β1-induced HLF activation, thus providing potential starting points 16 for developing anti-fibrotic therapeutics. In total, we screened 40,640 compounds distributed across 5 17 experimental batches, each containing between 10 and 45 assay plates.

18
To ensure the assay performance was robust, the batches were incrementally scaled up, starting with 10 plates  (Fig. 7a).

23
Nonetheless, the performance of the assay was still satisfactory for a cell-based functional-phenotypic screen.
The performance metrics for each batch also highlight the batch size (Fig. 7b). These hits represented potential 1 starting points for further development as anti-fibrotic therapeutics.   shown. We screened 127 384-well library plates (corresponding to 40.6k compounds) at a single dose of 3.5µM across 5 batches as 3 previously described over a 5-week period inclusive of completing automated data analysis on a cloud server. We next cherry-picked and 4 retested 166 initial hits in both HLF and HTSM cells 3 times each at multiple doses. Hits exhibiting consistent and selective activity without 5 obvious structural liabilities were repurchased as fresh powder. Upon receipt, confirmed hits were re-tested in a dose range in the core 6 assay and in several counter-screens. The overall process, including waiting periods, took a total of 15 weeks and ultimately yield 19 7 potent hits with extremely attractive phenotypic profiles.

8
Concentration-response testing of resupplied hits was conducted in both the original HLF assay and in a panel 9 of phenotypic counter-screens reporting on cell viability (live/dead), proliferation (ATP assay), and off-target 10 effects on HTSM tonic contraction over a broader dose-range (FLECS assay). Moreover, given that cultured 11 HTSM respond to TGF-β1 like HLF by becoming significantly more contractile 26 , we also counter-screened the 12 hits resupplied as powder in a modified HTSM contractility assay where TGF-β1 was used as a pro-contractile 13 trigger. The goal of this counter-screen was to determine if the primary hits, which were originally found in HLF,

14
inhibited broad TGF-β1 signaling pathways shared by multiple cell types or if they were selective to pathways 15 within HLF. Selectivity to pathways within HLF was considered more attractive and novel, given the safety issues

14
To address potential mechanisms of action of our advanced hits, we probed whether they acted upstream or 15 downstream of the TGF-β1-mediated nuclear translocation of Smad2/3 by employing fluorescent staining and 16 assessing the Smad signal ratio in the nuclear and perinuclear regions 24 h following TGF-β1 exposure.

17
Inhibition of Smad nuclear translocation would imply interference with TGF-β1 receptor engagement or processes immediately following it while actions independent of Smad signaling would indicate interference with 1 non-canonical pathways. In contrast to our control ALK5 inhibitor (RepSox), none of our hits affected TGF-β1-2 stimulated Smad nuclear accumulation (Fig 9a-4). Thus, our compounds act downstream of the TGF-β1-RI or 3 through non-canonical pathways with potential selectivity towards HLF-to-myofibroblast activation.

4
Next, we assessed chemical structure resemblances of our advanced hits with that of annotated compounds 5 with established activities by reviewing the CHEMBL database, scientific literature, and patents. Our analysis 6 served to evaluate the novelty of our findings and to generate potential target hypotheses based on the identified 7 structural similarities. Most of our hits contained a prevalent structural motif common to nitrogen-containing 8 heterocycles found in diverse phosphodiesterase (PDE) inhibitors. To test whether PDE inhibition is a potential 9 mechanism of preventing fibroblast activation, we performed a concentration response analysis of a panel of tool 10 inhibitors of PDE4, PDE5, a pan-PDE inhibitor, and a dual inhibitor of PDE3/4 (Fig. 9a). Among the tested PDE 11 inhibitors, only the dual-selective PDE3/4 inhibitor Zardaverine exhibited activity in our My-FLECVS assay, 12 reaching and exceeding 100% inhibition above 100 µM. The calculated IC50 for this response was approximately 13 45 µM, which is nearly three orders of magnitude higher than the IC50 values reported for its known PDE targets 14 (170 nM and 580 nM to PDE3 and PDE4, respectively). Zardaverine also did not exhibit an obvious plateau in 15 concentration-response curve even at high doses above 100 µM. Further considering that two other PDE4 16 inhibitors failed to affect TGF-β1-induced HDF contraction, PDE3/4 inhibition is unlikely the mechanism through 17 which Zardaverine inhibiting fibroblast activation and contraction. Collectively, these results suggest that our 18 advanced hits act through a potentially novel mechanism of action, unrelated to inhibition of ALK5 binding, 19 pSmad shuttling, PDE activity, and may therefore represent valuable new therapeutic opportunities.

20
Identified hits suppress fibrotic gene upregulation in human precision-cut lung slices 21 Finally, we performed initial translational testing on our advanced hits by evaluating their ability to suppress TGF-22 β1-driven upregulation of fibrotic gene expression in human precision-cut lung slices (hPCLS) (Fig 10a). Because 23 our best hits inhibited TGF-β1-induced rather than baseline contractile function, we hypothesized that they exert 24 broader effects on markers associated with TGF-β1-driven myofibroblast activation. To pursue this idea, we    Unexpectedly, one of the tested hits, originally identified for its activity in suppressing TGF-β1-driven HLF contractility, was also found to 6 suppress the upregulation of COLA1. Each data point comprises a unique biological replicate (e.g. a different 3mm tissue punch treated 7 in separate well on a plate). For each biological replicate, qPCR reactions were performed using four technical replicates.

9
We observed a substantial increase in the transcript levels of both ACTA2 and COL1A1 in hPCLS treated with 10 TGF-β1 for 48h, compared to untreated hPCLS (Fig. 10a). This upregulation was effectively abrogated by co-11 incubation with the ALK5 inhibitor RepSox (Fig. 10a). Of the four tested hit compounds identified in our My-12 FLECS screens, three exhibited significant inhibition of ACTA2 upregulation (Fig. 10b). Moreover, one of our hits 13 also substantially inhibited upregulation of COL1A1 expression relative to controls (Fig. 10c), despite having no 14 prior evidence predicting this activity.

Discussion 16
The high contractility of activated myofibroblasts plays a pivotal role in the advancement of fibrosis, signifying its 17 importance when formulating therapeutic strategies. Nevertheless, HTS for anti-fibrotic compounds has 18 traditionally prioritized cellular protein expression or proliferation, often sidelining myofibroblast contractility until 19 later stages of hit validation. To address this gap, we have developed a HTS functional assay specifically tailored 20 to assess cellular contractility, specifically in primary human fibroblasts. This innovation enables us to 21 concentrate on the crucial contractile phenotype in fibrosis from the very onset of the discovery process, 22 enhancing our ability to pinpoint functionally active compounds right from the primary screening stage, providing 23 a more targeted and efficient approach to anti-fibrotic drug discovery.

24
This My-FLECS assay was built on the previously reported FLECS Platform for single-cell contractility, an 25 approach presenting several distinctive capabilities that are essential for high-throughput implementation of this 26 phenotypic readout, including normalization of cell behavior, deployment in a 384-wellplate format, automated image analysis, and the generation of statistically significant data throughput per test condition. Using the My-1 FLECS, we successfully demonstrate that TGF-β1-driven contractile function can be quantified and used as a 2 functional biomarker of myofibroblast activation in primary HLF, with excellent dynamic range and reproducibility 3 both intra-and inter-plates. We further demonstrated the potential of the My-FLECS to be scaled to industrial 4 HTS of >17K wells per day (45 384-well-plates), and its ability identify promising phenotypic hits from expansive 5 drug libraries. Through comprehensive counter-screening of these hits in other cell types and assay paradigms 6 within the core FLECS assay, we demonstrate the ability to rapidly profile additional relevant phenotypic 7 activities, enabling data-driven hit prioritization strategies to guide the advancement of compounds with the most 8 promising profiles. Additionally, our follow-up experiments using precision-cut lung slices provide compelling 9 evidence that our approach, focused on isolating activity on TGF-β1-driven contractile function in activated 10 myofibroblasts, encompasses a wide range of potential targets. As a result, it enables the identification of 11 compounds with broad activities across various TGF-β1-driven phenotypes, highlighting the versatility and 12 efficacy of our methodology.

13
During the scale-up of our assay to >24 plates per day, we encountered bottlenecks that potentially impact plate 14 quality and overall screening outcomes. First, we observed that HLF, compared to other cell types, displayed 15 heightened sensitivity to minor changes in medium pH before adhesion. Inadequately buffered medium led to 16 compromised cell adhesion to the adhesive micropatterns, resulting in incomplete spreading, reduced overall 17 contraction, and ultimately, decreased long-term cell viability. To mitigate this sensitivity, it is crucial to prepare 18 fresh and appropriately buffered cell medium for each screening batch and minimize exposure of the medium to 19 the ambient environment. Even a brief delay, such as holding medium in narrow tubing of a liquid dispenser for 20 any period longer than it takes to prime the tubing, could induce observable colorimetric changes in pH, affecting 21 cell behavior. Therefore, strict adherence to these protocols is necessary to ensure reliable and accurate results 22 in HTS.

23
Secondly, we observed a significant adverse effect on the magnitude of HLF contractility with prolonged 24 incubation using live nuclear stain, such as Hoechst 33342. Specifically, contractility measurements taken later 25 than approximately 6 h after stain exposure exhibited a decrease in overall magnitude compared to earlier time 26 points under identical conditions. As the post-exposure time continued to increase, overall contraction diminished 27 to an extent that began to comprise the separation between controls, thereby reducing Z-primes and impacting 28 the interpretability of the assay. To mitigate this interference, screening batches should be appropriately sized 29 or divided into sets where the addition of nuclear stain is done set-wise, preventing prolonged exposure and 30 preserving the integrity of the assay results.

31
Third, we occasionally observed the occurrence of "edge-effects," commonly seen in cell-based assays, which 32 can manifest as non-uniform cell distributions in the outer wells or increased contraction along the outer edges 33 compared to the interior wells. To ensure consistent and uniform cell adhesion across all wells, it is crucial to 34 seed plates that have fully equilibrated to room temperature. Temperature disparities across the plates can induce micro-flows that drive cells towards one side of the wells, leading to non-uniformity in cell adhesion 1 distributions. By allowing plates to reach equilibrium with the ambient temperature, we can mitigate the 2 occurrence of these edge-effects and maintain uniform cell behavior throughout the assay. Other limitations to 3 consider are that the assay measures contractility in an isolated system, which may not fully replicate the 4 complex in vivo environment where myofibroblasts interact with other cells and the extracellular matrix. We also 5 note that while the assay showed consistency and reliability in our hands, further validation in other laboratories 6 would be beneficial to establish its robustness. In particular, our screens and follow-up studies were performed 7 in well-characterized cells derived from only a single donor for each unique cell type used in the study.

8
In conclusion, this study introduces the first ever HTS paradigm that offers the ability to identify effectors of 9 myofibroblast contractility from large-scale primary screening campaigns, in an efficient and scalable manner 10 that can produce promising functionally validated starting points for drug discovery. Our hPLCS results highlight 11 the potential drawbacks of depending solely on molecular readouts, such as the quantification of ACTA2 protein 12 levels, which are commonly employed in HTS for anti-fibrotic compounds. As demonstrated with FYTE-0066, a 13 lack of observable activity in these molecular readouts does not definitively equate to an absence of functional 14 efficacy in inhibiting fibrotic processes. This discrepancy indicates that a singular focus on molecular markers 15 may inadvertently exclude compounds that could serve as promising starting points for deeper investigation. Our 16 results further highlight the value of incorporating a functional readout into compound screening, specifically one 17 that gauges the multi-faceted hypercontractile phenotype triggered by myofibroblast activation. As seen with, 18 e.g., FYTE-0070, this approach facilitates the discovery of compounds that not only exhibit vital functional activity 19 in impeding the mechanical mechanisms underlying fibrosis, but it may also reveal broader inhibitory effects on 20 a variety of phenotypes driven by TGF-β1, extending beyond the contractile pathways. By integrating a functional 21 viewpoint into the screening process, it's possible to identify compounds with diverse activity, offering significant 22 promise as potential anti-fibrotic agents.

23
Ultimately, this functional assay can potentially be applied to various types of fibrosis, since myofibroblast 24 contractility plays a significant role in many fibrotic diseases in many tissues. It can also extend to other 25 modalities including screens for biologics as well as fibrotic genes via arrayed genome-wide knock-out screens.