Inhibition of Cell Motility by Cell-Penetrating Dynamic Covalent Cascade Exchangers: Integrins Participate in Thiol-Mediated Uptake

Integrins are cell surface proteins responsible for cell motility. Inspired by the rich disulfide exchange chemistry of integrins, we show here the inhibition of cell migration by cascade exchangers (CAXs), which also enable and inhibit cell penetration by thiol-mediated uptake. Fast-moving CAXs such as reversible Michael acceptor dimers, dithiabismepanes, and bioinspired epidithiodiketopiperazines are best, much better than Ellman’s reagent. The implication that integrins participate in thiol-mediated uptake is confirmed by reduced uptake in integrin-knockdown cells. Although thiol-mediated uptake is increasingly emerging as a unifying pathway to bring matter into cells, its molecular basis is essentially unknown. These results identify the integrin superfamily as experimentally validated general cellular partners in the dynamic covalent exchange cascades that are likely to account for thiol-mediated uptake. The patterns identified testify to the complexity of the dynamic covalent networks involved. This work also provides chemistry tools to explore cell motility and expands the drug discovery potential of CAXs from antiviral toward antithrombotic and antitumor perspectives.

I ntegrins are cell surface proteins in charge of cell adhesion, cell motility, and bidirectional signaling. 1,2 Because of their involvement in wound healing, thrombosis, 3 and cancer cell migration 4 as well as viral entry 5−10 and drug delivery, 11 they are of great interest as drug targets. 2 In vertebrates, the integrin superfamily consists of 24 heterodimers, all composed of α and β subunits ( Figure 1A). They interact with different components of the extracellular matrix. During cell adhesion and migration, integrins change from a bent to a linear conformation. This change is controlled by one of the most beautiful disulfide tracks in biology, composed of 20 disulfides in the β subunit ( Figure 1B). 12−15 Several of these disulfides are allosteric in nature and in a hook or staple conformation, 3,16 and literature describing dithiol/disulfide redox regulation of integrin function is increasingly emerging. 4,17−20 Admiring their disulfide tracks, we thought that integrins would be perfect exchange partners in thiol-mediated uptake (TMU). TMU 21−25 refers to the cell-penetrating activity provided by thiol/disulfide cascade exchanger (CAX) motifs. CAXs undergo dynamic covalent exchanges with membranebound protein thiols (or disulfides), where each exchange produces a new (or offers another) covalently tethered exchanger, that can continue exchanging until they are delivered into the cytosol. 21 TMU has been realized with many classes of CAX for the cytosolic delivery of small molecules, 21,26 antibodies 22 and other proteins, 27−30 genome editing machinery 31 and other oligonucleotides, 32−37 polymers, 38 liposomes, 39 and nanoparticles 22,40,41 into various cellular targets including deep tissue, 30,33 living animals, 31,42 plant cells, 35 and bacteria. 26 Proteomics data, 43 heatmap patterns, 44,45 and literature on oligonucleotide phosphorothioate 37,46 and viral uptake 8,47−50 all support that multipartner exchange networks are involved in how TMU brings matter into cells. 21 However, the dynamic covalent exchange cascades of TMU are complex and its fleeting intermediates are elusive, which is likely the reason why TMU is not better known and understood. 21 Importantly, CAX-induced TMU can be inhibited by treating cells with surface-thiol-reactive agents. 51,52 Given that thiol-rich integrins are natively poised for reversible thiol modifications, 3,4,16−20 we hypothesized that CAXs which enable or inhibit TMU could also inhibit the cell motility for which integrin dynamics are crucial, which in turn would demonstrate the involvement of integrins as one of the so far essentially unknown cellular exchange partners in TMU.
To test the effects of TMU inhibitors on cellular integrin functions, compounds 1−16 were bought or synthesized following reported procedures (Figures 2 and S2). 44,45,51−55 Cell motility assays 4 were adapted to automated high-content high-throughput (AHCHT) imaging methods. 45,51 Before cell seeding, 96-well plates were coated with one of three proteins that interact with different integrins: Collagen I (C), which activates mainly β 1 -containing integrins; fibronectin (F), which activates mainly β 1 -and β 3 -containing integrins; and vitronectin (V), which activates mainly β 3 -and β 5 -containing integrins. 1−4 HeLa Kyoto (HK), MCF-7, and MDA-MB-231 cells were then seeded on the protein-coated wells to test their migration behavior when treated with TMU inhibitors. Among many possibilities, highly aggressive MDA-MB-231 and noninvasive MCF-7 breast cancer cells were selected as established standards to explore motility 4 and HK cells as a link to uptake 45 and to probe the power of AHCHT assays to detect small changes accurately.
The confluent cell monolayers were scraped to create "scratches" with widths of around 1 mm using a homemade device that removes cells without damaging the coating ( Figure S1), as demonstrated by the different intrinsic motilities of different cells on different coatings (e.g., Figure  S14). Cells were then incubated with TMU inhibitor candidates from t = 0, and the cell motility m was determined from the change of cell-free area A from that at t = 0 (m = A 0 − A t ) ( Figures 3A,B and S8−S28).
A complex segmentation pipeline was developed to generate unbiased image masks to estimate A t (Figures S4−S6). This AHCHT motility assay was particularly powerful with slower moving cells that are otherwise not easily distinguishable  Figure 3E) to 0.40 ± 0.01 μm min −1 with 3 μM inhibitor 9 ( Figure 3F). This decrease matched the formal wound healing times to close a 1 mm scratch (17 h against 40 h). Because it was important to avoid misinterpretation of cell death as motility inhibition, 56 CAXs and controls 1−12 were reconfirmed 45,51−53 as nontoxic under the experimental conditions ( Figure S3).
Cell migration inhibition by 1−16 was assessed for the three cell lines on all three surfaces (Figures S8−S28 and Tables  S1−S8). Heatmaps comparing values of IC 50 or MIC (calculated as IC 15 ; see eq S2) were constructed ( Figures 3G  and S40). The difference between IC 50 and MIC tracked to the switching half-window C R , 57,58 a metric that can report on cooperativity (large C R indicates negative cooperativity), or on whether several active sites 58 are involved in an exchange cascade ( Figure S41). Fractional heatmaps ( Figure 3H,I) and two-component correlations ( Figure 3J) were extracted to test for patterns (e.g., inhibitors with similar targets).
The appearance of unique patterns with distinct hotspots supported that CAX inhibitors operate selectively, beyond global reactivity, with specific mechanisms and therefore have a certain drug discovery potential. The most antimigratory CAXs such as 3, 9, and 10 contain highly reactive, fast-exchanging chalcogen-, pnictogen-, and tetrel-centered relays. Those CAXs that are typically excellent for TMU but underperformed in the antimigratory assay were those that (i) produce slowexchanging, "sticky" dynamic covalent networks that excel for cytosolic delivery, like the bioinspired benzopolysulfane (BPS) 54 4, (ii) exchange only in aprotic hydrophobic environments like cyclic thiosulfonate (CTO) 44 6, or (iii) may prefer to exchange with other protein partners, such as asparagusic acid (AspA) 2 with the transferrin receptor. 43 These trends were consistent with the hypothesis that the inhibition of cell motility by the most efficient CAXs operates with fast exchange cascades along thiol or disulfide arrays on cell surfaces.
Good inhibitors of cell motility were overall good inhibitors of thiol-mediated uptake ( Figure 4A,B). Comparison with literature IC 50 for the inhibition of TMU of 17 (ETP 3 with an To close the functional feedback loop, β 1 integrins were knocked down in HK cells, as confirmed by immunofluorescence quantification ( Figure 4C vs E). Cellular uptake of fluorescent ETP 17 into the cytosol and mostly the nucleus, 55 determined under routine conditions, 55 decreased significantly in the absence of β 1 integrins ( Figure 4D,F). As far as we know, this difference provides the first direct experimental support that integrins act as dynamic covalent exchange partners in thiol-mediated uptake. The complementary motility inhibition could not be measured because the motility of knockdown cells was as poor as expected, reappearing only when integrins were starting to be re-expressed (not shown).
In summary, we report that a focused collection of CAXs inhibits the motility of various cells on various surfaces. Their antimigratory activities exceed that of Ellman's reagent by far and correlates globally with their abilities to penetrate cells and deliver substrates into the cytosol. Knockdown experiments support the conclusion that the same proteins are involved in TMU as in the antimigratory effect.
This conclusion is important. Despite the rich collection of candidates, 21 only one TMU partner had previously been robustly identified, i.e., the transferrin receptor, without being general (for AspA, 43 not ETP 55 ). Now we find that the integrin superfamily is the first general exchange partner that has experimental support to participate in TMU. These results thus (a) introduce dynamic covalent cascade exchange chemistry to the control of cell motility, (b) expand the CAX drug discovery space from antiviral toward antithrombotic and antitumor potential, (c) confirm integrins as exchange partners in the dynamic TMU networks that deliver matter into cells, from drugs to pathogens, and thus (d) inspire new design strategies (e.g., multivalent CAXs 30 to benefit from integrin clustering 66 ). These lessons are likely to enable significant and varied future advances.

Automated Cell Motility Inhibition Assay
On a 96-well Black ibiTreat sterile microplate coated with collagen I, fibronectin, or vitronectin (see the Supporting Information), HeLa Kyoto, MCF-7 (6 × 10 4 cells/well), or MDA-MB-231 cells (9 × 10 4 cells/well) were seeded in DMEM + 10% FBS and kept overnight at 37°C under 5% CO 2 atmosphere. The cell monolayers were scraped with a homemade device ( Figure S1) and washed twice with PBS. Then the medium was changed to DMEM (with FBS 0− 7.5%), and the inhibitor candidates were added with an electronic multichannel pipet. With an automated confocal microscope, transmitted light (TL) images were recorded at the center of the wells, and time series of 14−26 h were recorded. The onset of toxicity at high inhibitor concentrations was identified by dead rounded-up cells that detached and accumulated in the center of the scratch ( Figure S3A), and measurements were limited to concentrations below this threshold ( Figure S3B). Automated analysis processed the original time-lapse TL images ( Figure S4A) to generate a relevant mask of the cell layer. The first set of masks determined the cell edge ( Figure  S4C). This was done by top-hat modification of the TL image that highlights the cell boundaries ( Figure S4B). To the segmented image, a size filter was applied to exclude any object below 50 μm 2 ( Figure S4D). Following a similar procedure on a pixel-intensity-inverted image ( Figure S5A,B), the cell body was segmented ( Figure S5C). Finally, the two cell masks, that is, the cell edge ( Figure S4D) and cell body ( Figure S5C), were combined to create the cell layer ( Figure S6A). This cell layer was then slightly grown to give a homogeneous layer ( Figure S6B). Since the growth led to the appearance of some unwanted background objects, a size filter was added to remove all objects below 30 000 μm 2 to give the final cell layer ( Figure S6C). To determine the scratch area, the cell layer ( Figure S6C) was subtracted from the whole image mask (including all pixels of the image, Figure S6D) to give the desired area ( Figure S6E).
In the final image ( Figure S6F), the area of the scratch (A) was deduced from the blue area, the cell layer was labeled in yellow, and the space between cells not caused by the scratch (interstitial space) was labeled in cyan. The motility m was calculated by subtracting the area of the blue layer at a specific time (A t ) from the area at t = 0 (A 0 ), i.e., m = A 0 − A t , and The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/jacsau.3c00113.
Materials and methods; inhibitor synthesis; cell culture; coating procedures; expanded cell motility imaging procedure; expanded AHCHT motility analysis; dependence on inhibitors, concentration, time, cells, surfaces, and serum; pattern generation; single-cell motility inhibition procedure; ITGB1 knockdown; cellular uptake inhibition procedure (PDF) Video of AHCHT motility assay with inhibitor 9 (MPG) Video of AHCHT motility assay without inhibitor (MPG)