Mechano-redox control of Mac-1 de-adhesion from ICAM-1 by protein disulfide isomerase promotes directional movement of neutrophils under flow

Macrophage-1 antigen or Mac-1 (CD11b/CD18, αMβ2) is a leukocyte integrin essential for firm adhesion of neutrophils, lymphocytes and monocytes against flow when recruited to the endothelium. To migrate to the site of inflammation, leukocytes require coordinated adhesion and de-adhesion for directional movement. The vascular thiol isomerase, protein disulfide isomerase (PDI), was found by fluorescence microscopy to colocalize with high affinity Mac-1 at the trailing edge of stimulated neutrophils when adhered to ICAM-1 under fluid shear. From differential cysteine alkylation and mass spectrometry studies, PDI cleaves two allosteric disulfide bonds, C169-C176 and C224-C264, in the βI domain of the β2 subunit, and in mutagenesis and cell transfection studies, cleavage of the C224-C264 disulfide bond was shown to selectively control Mac-1 dis-engagement from ICAM-1 under fluid shear. Molecular dynamics simulations and binding of conformation-specific antibodies reveal that cleavage of the C224-C264 bond induces conformational change and mechanical stress in the βI domain that allosterically alters exposure of an αI domain epitope and shifts Mac-1 to a lower affinity state. From studies of neutrophil adherence to ICAM-1 under fluid shear, these molecular events promote neutrophil motility in the direction of flow at high shear stress. In summary, shear-dependent PDI cleavage of neutrophil Mac-1 C224-C264 disulfide bond triggers Mac-1 de-adherence from ICAM-1 at the trailing edge of the cell and enables directional movement of neutrophils during inflammation.


Introduction 38
The integrin macrophage-1 antigen or Mac-1 (CD11b/CD18, αMβ2) is essential for the 39 recruitment of leukocytes to sites of infection or injury. It binds to a variety of ligands including 40 intercellular adhesion molecule 1 (ICAM-1), fibrinogen, complement fragment iC3b and CD40 41 ligand (CD40L) to elicit an inflammatory response. To migrate to the site of infection and injury, 42 circulating leukocytes tether and roll on vessel wall by interacting with selectins expressed on 43 endothelial cells, reducing their velocity. Their initial contacts trigger G-protein coupled receptors 44 for inside-out signaling leading to integrin activation and binding to endothelial ICAM-1 ( have been reported to enable cell de-adhesion. For instance, shedding of integrin by 50 metalloproteinases has been described to enable exit of macrophages from the site of 51 inflammation (Gomez et al., 2012). Internalization of integrin by clathrin-mediated endocytosis is 52 another important mechanism that allows disassembly of focal adhesions and detachment of 53 cells from substrata (Bai et al., 2017;Ezratty et al., 2009). These mechanisms rely on removal 54 of functional integrin from the cell. We have described a mechanism of integrin dis-engagement 55 that involves allosteric changes in the integrin binding sites (Passam et al., 2018). 56 De-adhesion of platelet integrin αIIbβ3 (GPIIb/IIIa, CD61/CD41) from fibrinogen occurs via 57 force-dependent cleavage of an allosteric disulfide bond in the integrin binding site (Passam et 58 al., 2018). A member of the vascular thiol isomerase family, ERp5, cleaves a disulfide bond in 59 the β3 subunit to release platelets from fibrinogen. The archetypal thiol isomerase, protein 60 disulfide isomerase (PDI), has been demonstrated to be essential in Mac-1-dependent 61 neutrophil migration during inflammation. Mac-1 becomes upregulated during inflammation to 62 mediate neutrophil adhesion and crawling (Sumagin et al., 2010). Conditional knockout of PDI in 63 murine neutrophils led to their impaired adhesion and crawling on inflamed endothelium (Hahm 64 et al., 2013). 65 Here, we report a mechano-redox mechanism that mediates Mac-1 de-adhesion selectively 66 from ICAM-1. PDI colocalizes with high affinity Mac-1 at the trailing edge of neutrophils, and 67 cleaves a disulfide bond in the headpiece of the β2 subunit that changes Mac-1 conformation to 68 a lower affinity state. PDI cleavage of the Mac-1 disulfide bond promotes neutrophil movement 69 in the direction of fluid shear. This mechano-redox regulation by PDI provides a mechanism for 70 neutrophils to de-adhere from ICAM-1 that is essential for directional movement and migration 71 during inflammation. inflamed endothelium and that such migration could be restored by addition of recombinant PDI 78 (Hahm et al., 2013). As neutrophil adhesion and crawling on endothelium is predominantly 79 dependent on Mac-1 binding to endothelial ICAM-1, co-localization of surface PDI and Mac-1 80 was measured in fixed neutrophils adhered to immobilized ICAM-1. Using anti-PDI antibody DL-81 11, low level of PDI was detected on the surface of untreated neutrophils adhered to ICAM-1 82 (Figure 1figure supplement 1). Using anti-CD11b antibody, CBRM1/5, that recognizes an 83 activation-specific epitope in the I domain of αM (αI domain) that is exposed only in Mac-1 at 84 high affinity state (Oxvig et al., 1999), high affinity Mac-1 was hardly detected in resting 85 neutrophils. Upon stimulation with fMLF, PDI and high affinity Mac-1 were readily detected on 86 the neutrophil surface. Increased cell surface PDI was detected in fMLF-stimulated neutrophils 87 in accordance with a previous report (Hahm et al., 2013), which was accompanied by Mac-1 88 upregulation (Kishimoto et al., 1989). PDI predominately colocalized with high affinity Mac-1 89 clusters associated with focal adhesion points for firm adhesion of neutrophils on ICAM-1. 90 To determine if surface PDI colocalizes with high affinity Mac-1 during neutrophil crawling, 91 neutrophils were stained with anti-PDI antibody and CBRM1/5, stimulated by fMLF and perfused 92 onto microfluidic chips coated with ICAM-1 and left to settle. Adhered neutrophils were then 93 subjected to shear stress representing venous or arterial vessel at 0.7 or 5.6 dynes/cm 2 94 (Sakariassen et al., 2015), respectively, and imaged in real-time by confocal microscopy. 95 Adhered neutrophils exhibited polarized morphology in presence of fluid shear with leading and 96 trailing edges clearly defined on DIC images (Valignat et al., 2014). PDI was found to localize in 97 the trailing edge of neutrophils while high affinity Mac-1 also predominately formed clusters in 98 the trailing edge but was also detected in the middle of crawling neutrophils as previously 99 reported (Hyun et al., 2019) ( Figure 1A). The fractions of PDI and Mac-1 that overlapped in 100 leading and trailing edges were measured and expressed as Manders' colocalization 101 coefficients ( Figure 1B and Supplementary File 1  File 1 Table S1). This indicates that PDI and high affinity Mac-were more 110 colocalized in the trailing edge than in the leading edge of crawling neutrophils. 111 Colocalization of PDI and Mac-1 in the trailing edge of crawling neutrophils suggests that PDI is 112 manipulating disulfide bonds in Mac-1. This was measured using differential cysteine alkylation 113 and mass spectrometry. 114

PDI cleaves two disulfide bonds in the β2 integrin 116
Recombinant Mac-1 protein purified from human embryonic kidney cells was incubated with 10-117 fold molar excess of redox active PDI or redox inactive PDI (riPDI). Both catalytic cysteines in 118 the a and a' domains were replaced with alanine in riPDI. Unpaired cysteines in Mac-1 were 119 alkylated with the thiol alkylator 2-iodo-N-phenylacetamide ( 12 C-IPA) followed by reduction of 120 disulfide bonds by DTT and labeling of disulfide cysteines with a carbon-13 isotope of IPA ( 13 C-121 IPA) (Figure 2A). The protein was digested by proteases and the peptides were analyzed by 122 mass spectrometry. Forty-nine cysteine-containing peptides representing 24 of the 28 disulfide  Table S2). Using existing crystal structures of 125 extended αVβ3 (Xiong et al., 2009) and bent αXβ2 (Sen and Springer, 2016) and sequence 126 alignment, a model of extended Mac-1 structure was constructed and the positions of the 28 127 disulfide bonds indicated ( Figure 2C). The four disulfide bonds which were not resolved (C514-128 C537, C519-C535, C581-C590 and C593-C596) occur in the cysteine-rich EGF3 and EGF4 129 domains ( Figure 2C). The redox state of the β2 disulfide bonds ranged from 90-98% oxidized in 130 untreated control, which is in general agreement with the structure of mature β2 integrin where 131 all disulfide bonds were found to be intact (Sen and Springer, 2016). Addition of redox active but 132 not redox inactive PDI resulted in almost complete (>90%) and selective reduction of the C169-133 C176 and C224-C264 disulfide bonds in the βI domain ( Figure 2B). 134 The β2 βI domain together with the β-propeller and I domain from the αM integrin subunit form 135 the headpiece of the integrin. PDI cleavage of the βI domain disulfide bonds suggested that this 136 vascular thiol isomerase might influence Mac-1 binding to ICAM-1. To study the effect of cleavage of the βI domain C169-C176 and C224-C264 disulfide bonds on 141 binding of ligands to Mac-1, mammalian cells were transfected with wild-type or disulfide mutant 142 integrins. One or both disulfide bonds were ablated by replacing the disulfide cysteines with 143 serine, baby hamster kidney (BHK) cells stably transfected with wild-type αM and either wild-144 type β2 or mutant β2, and cells selected for comparable expression of the receptors ( Figure  145 3A). Initially, Mac-1 binding to immobilized ICAM-1 in a static cell adhesion assay was 146 assessed. Ablation of either of the two disulfide bonds had no effect on cell adhesion to ICAM-1 147 under static conditions (Figure 3figure supplement 1). As Mac-1 and ICAM-1 interact under 148 shear forces in flowing blood to mediate neutrophil adhesion and crawling on the endothelium, 149 we assessed Mac-1 binding to ICAM-1 under fluid shear. Two different states of PDI were 150 employed in the assays; one where both active site cysteines were fully reduced (reduced PDI) 151 and another where the active site cysteines were fully oxidized (oxidized PDI). BHK cells 152 expressing wild-type Mac-1 were untreated or incubated with the different PDI forms before 153 perfusing over ICAM-1-coated channels and left to adhere. Non-adherent cells were removed 154 by perfusion of buffer at low shear force of 0.175 dynes/cm 2 . De-adhesion of bound cells was 155 then triggered by doubling the shear force every minute until it reached 11.2 dynes/cm 2 . The 156 number of cells remaining adhered at each shear force was measured ( Figure 3B) and 157 expressed as a percentage of the total adherent cells at 0.175 dynes/cm 2 . The area under the 158 curve of the adherent cells as a function of shear force was calculated. Reduced but not 159 oxidized PDI promoted de-adhesion of wild-type Mac-1 expressing cells from ICAM-1 ( Figure  160 3C). The data was fit to a one phase exponential decay model by nonlinear regression to 161 determine the decay constant (K, cm 2 dynes -1 ) and shear force (F 50 , dynes/cm 2 ) at which 50% of 162 the cells were de-adhered from ICAM-1 ( PDI, but not its oxidizing activity, promotes Mac-1 de-adhesion from ICAM-1. 168 To further define whether PDI promotes shear-dependent Mac-1 de-adhesion from ICAM-1 by 169 cleavage of the two β2 disulfide bonds, we subjected BHK cells expressing wild-type Mac-1 or 170 disulfide mutant Mac-1 to the same cell de-adhesion assays. Ablation of the C224-C264 171 disulfide bond but not the C169-C176 bond enhanced the shear-dependent de-adhesion of cells 172 from ICAM-1 (Figure 3D and E). F 50 for cells expressing C224,264S Mac-1 is 0.3392 173 dynes/cm 2 , which is approximately one quarter of the F 50 for cells expressing wild-type Mac-1 174 ( Table 1). This value is half of the F 50 for Mac-1 expressing cells treated with reduced PDI. The 175 difference is possibly due to incomplete PDI cleavage of the C224-C264 disulfide bond in cell 176 surface Mac-1. This finding indicates that PDI cleavage of the C224-C264 disulfide bond is 177 important for Mac-1 dis-engagement from ICAM-1 under shear force. 178 The Mac-1 subpopulation reported to mediate ICAM-1 interaction in activated neutrophils has 179 also been shown to bind fibrinogen (Diamond and Springer, 1993a). To characterize if PDI 180 cleavage of β2 disulfide bonds promotes de-adhesion of Mac-1 from fibrinogen, we subjected 181 BHK cells expressing wild-type or disulfide mutant Mac-1 to de-adhesion assays using 182 fibrinogen-coated channels. Ablation of one or both β2 disulfide bonds had no significant effect 183 on Mac-1 dis-engagement from fibrinogen when compared to wild-type Mac-1 ( Total Mac-1 expression on BHK cells was determined using the H52 monoclonal antibody that 198 recognizes an epitope in the hybrid domain (residues 386-400) of the β2 subunit (Al- Shamkhani  199 and Law, 1998) and is accessible in all Mac-1 conformations (Figure 4figure supplement 1).

200
The monoclonal antibody, MEM48, recognizes the EGF3 domain (residues 534-543) of β2 201 integrin that is only exposed when Mac-1 is extended (Sen and Springer, 2016 To elucidate how the β2 βI C224-C264 bond could influence ligand affinity, we conducted MD 216 simulations of the effect of C169-C176 and C224-C264 redox state on the conformational 217 dynamics of the βI domain in complex with the β-propeller ( Figure 5A). 218 219 Cleavage of the C224-C264 bond perturbs inter-residue contact and mechanical stress in 220 the βI domain of the β2-integrin 221 As the structure of Mac-1 has not yet been determined, we took initial atomic positions from the 222 X-ray structure of the highly close homolog, LFA-1 (ɑLβ2, CD11a/CD18) (Sen and Springer, 223 2016). LFA-1 has an identical βI domain and highly similar β-propeller to Mac-1 ( Figure 5A).

224
The dynamics of the complex was monitored in multiple molecular dynamics simulation replicas 225 and for different redox states of the C169-C176 and C224-C264 disulfide bonds. During the 226 simulations, the complex was found to be very stable, with a backbone root mean square 227 deviation from the initial positions smaller than 0.45 nm. Inside the βI domain, the loop 228 connecting the strands B2 and B3 (L B2-B3 ) displayed the largest conformational variations, 229 although the redox state of the bonds of interest did not favor any preferential position of this 230 loop ( Figure 5B). We also analyzed the change in residue-residue contacts induced by 231 reduction of either disulfide bond ( Figure 5C and The question arises how this allosteric effect originating from C224-C264 reduction impacts the 246 complex structurally. Calculation of the root mean square fluctuation (RMSF) also displayed 247 changes in the dynamics of several residues distant to C224-C264, even close to C169-C176 248 ( Figure 5E). In addition, the solvent accessible surface area (SASA) of βI was found to shift 249 towards larger areas when the C224-C264 bond was reduced, but not in the other situations 250 ( Figure 5F). This increment in SASA is attributed to a more exposed surface area of the region 251 R2 near C224-C264 rather than the distant region R1 for which changed connectivity and stress 252 was also observed (compare regions R2 and R1 in Figure 5F). In summary, our MD simulations 253 demonstrate that reduction of the C224-C246 bond, and to a minor extent the reduction of 254 C169-C176 or both, allosterically alters the internal connectivity and mechanical stress and 255 modulates the surface area of βI. 256 To demonstrate how PDI and force are essential to regulate neutrophils de-adhesion from 257 ICAM-1, we measured neutrophil crawling as a function of cell adhesion and de-adhesion 258 events under fluid shear.  Table S4). In contrast, when subjected to 5.6 dynes/cm 2 fluid shear, 281 there was a significant increase of neutrophils migrating in the direction of flow when treated 282 with reduced PDI when compared to neutrophils treated with control oxidized PDI ( Figure 6B).

283
Percentage of neutrophils migrating in the direction of flow at 5.6 dynes/cm 2 was 42% for cells 284 treated with oxidized PDI and increased to 70% for cells treated with reduced PDI ( Table S4). Displacement of neutrophils in the Y-286 direction which is perpendicular to the direction of flow was also determined ( Figure 6C). 287 Neutrophils treated with oxidized or reduced PDI had no significant difference in their migration 288 in the Y-direction, indicating that PDI has no effect on neutrophil movement in the direction 289 perpendicular to fluid shear. 290 Crawling speed of neutrophils treated with control oxidized or reduced PDI was also determined 291 by measuring the total distance traveled by each neutrophil and dividing it by the total time of 292 migration. Neutrophils treated with reduced PDI exhibited significantly slower crawling speeds at 293 0.7 and 5.6 dynes/cm 2 fluid shear compared to neutrophils treated with control oxidized PDI 294 (Figure 6D), and the speeds were comparable at both shear force (Supplementary File 1 295 Table S5). This result indicates that PDI-mediated decrease in crawling speed of neutrophils 296 treated with is independent of shear force. 297 Together, our data indicates that reduced PDI but not control oxidized PDI slows down 298 neutrophil crawling under shear force and promotes migration in the direction of flow. 299 300

Discussion 301
We describe here a mechano-redox event controlling the function of Mac-1 on the trailing edge 302 of neutrophils. PDI in the presence of fluid shear from 0.17-11 dynes/cm 2 selectively regulates 303 Mac-1 de-adhesion from endothelial ICAM-1 by cleaving the C224-C264 allosteric disulfide 304 bond in the β2 βI domain ( Figure 7A). Cleavage of this bond induces mechanical stress in the 305 βI domain and allosterically perturbs residue contacts between the βI and β-propeller domains. 306 We suggest that this conformational change in Mac-1 results in suboptimal binding to ICAM-1 307 that leads to detachment of ICAM-1 in the fluid shear encountered in the circulation. As a 308 consequence of Mac-1 de-adhesion at the trailing edge of the cell, PDI promotes neutrophil 309 migration in the direction of flow. In our studies, PDI has no effect on Mac-1 adhesion to ICAM-1 in static conditions, which 339 suggests there may be separate mechanisms regulating neutrophil migration under static 340 versus shear conditions. 341 We previously reported that platelet surface thiol isomerase ERp5 cleaves the βI domain C177-342 C184 disulfide bond in the β3 subunit of platelet αIIbβ3 and this cleavage changed the positions 343 of residues critical for metal ion coordination resulting in release of fibrinogen. The β2 subunit 344 C169-C176 disulfide bond in Mac-1 is homologous to the β3 subunit C177-C184 disulfide bond 345 ( Figure 7B) and is close to the metal ion binding sites and αM I (or αI) domain involved in 346 ICAM-1 binding. We, therefore, anticipated that the redox state of this bond would be a critical 347 determinant of binding of Mac-1 to ICAM-1. Although our mass spectrometry analysis showed 348 that PDI cleaves both βI-domain disulfide bonds equally well, our functional data supports that 349 only the C224-C264 disulfide bond controls Mac-1 affinity for ICAM-1 in fluid shear. This finding 350 is surprising since the C224-C264 disulfide bond is distant from known epitopes in the αI, β- common LAD-1 alleles is G284S (or G262S in the mature β2 integrin) that is two residues from 367 C264. G284 (or G262) is precisely at the region that displayed the largest change in area These findings support our conclusion that residues influenced by the redox state of the C224-372 C264 disulfide bond are important for ICAM-1 binding. 373 Mechano-redox regulation of Mac-1 by PDI controls ICAM-1 but not that of fibrinogen binding. 374 Among the integrin family, Mac-1 is considered the most promiscuous that can bind to over 30 375 extracellular ligands . Distinct epitopes in Mac-1 have been identified to be 376 important for binding to specific ligands (Diamond et al., 1993). For example. a motif in the αI 377 domain M7 (E162-L170) specifically interacts with the inflammatory ligand CD40L. Inhibitory 378 anti-M7 antibody blocks Mac-1 binding to CD40L but has no effect on binding to ICAM-1 or 379 fibrinogen (Wolf et al., 2018). On the other hand, the αI domain epitopes for ICAM-1 and 380 fibrinogen binding are overlapping as demonstrated by blocking of Mac-1 adhesion to both 381 ICAM-1 and fibrinogen by anti-αM CBRM1/5 antibody (Diamond and Springer, 1993a). PDI, 382 therefore, selectively controls Mac-1 promiscuity. Notably, the βI-domain C169-C176 and C224-383 C264 disulfide bonds are conserved in 7 of the 8 β integrins (Figure 7B) suggesting that other β 384 integrins might be also subject to mechano-redox regulation. 385 In conclusion, we have identified a mechano-redox mechanism that selectively controls Mac-1 386 binding to ICAM-1 in fluid shear conditions. This mechanism allows the trailing edge of 387 neutrophils to detach from ICAM-1 and enables movement in the direction of flow. Importantly, 388 this informs studies on the development and optimization of PDI inhibitors as therapeutic agents 389 to attenuate neutrophil migration to subdue inflammation. 390

Neutrophil Isolation 393
All procedures involving human whole-blood were collected from healthy human volunteers in 394 accordance with the Human Research Ethics Committee of the University of Sydney (2014/244) 395 and the declaration of Helsinki. Human whole-blood was collected from healthy human 396 volunteers into plastic syringes containing Clexane at 20 U/mL (Sanofi). Neutrophils were 397 isolated from human whole-blood by Histopaque density gradient centrifugation. A 48 mL 398 density gradient was created by layering Histopaque-1077 on top of Histopaque-1191 (Sigma-399 Aldrich), followed by a layer of whole blood at a volumetric ratio of 2:1:1 in a 50 mL conical 400 centrifuge tube at 25°C. The tube was centrifuged at 600 G for 20 min with no brake, and the 401 pink neutrophil buffy coat that formed above the red blood cell layer was collected, diluted in 402 Hank's Buffered Salt Solution (HBSS, 1 mM CaCl 2 , 1 mM MgCl 2 , 5.4 mM KCl, 0.44 mM 403 KH 2 PO 4 , 136.9 mM NaCl, 0.34 mM Na 2 HPO 4 , 5.5 mM D-glucose, 0.5% (w/v) BSA, pH 7.2), and 404 centrifuged at 600 G for 5 min. The cell pellet was then resuspended in HBSS and cleared of 405 red blood cells by lysis in ice cold 0.2% (w/v) NaCl for 20 secs, neutralized with an equivalent 406 volume of 1.6% (w/v) NaCl and centrifuged at 250 g, 4°C for 6 min. This process was repeated 407 up to 2 times until all red blood cells were cleared. Neutrophils were stored at 4°C, brought up to 408 25°C prior to use in assays, and used within 4 h of blood collection. 409 410

Colocalization of Mac-1 and PDI on neutrophils 411
For colocalization of Mac-1 and PDI under static conditions, wells of an 8-well microslide were 412 coated with 10 μg/mL of ICAM-1/Fc (R&D systems) for 2 h at room temperature, washed with 413 PBS and blocked with 0.5% (w/v) BSA for 30 min. After a final wash, neutrophils (1x10 6 414 cells/mL) were added to each well in the presence of APC conjugated anti-CD11b antibody 415 CBRM1/5 (BioLegend) at 1 μg/mL, anti-PDI antibody DL11 (Sigma-Aldrich) at 2 μg/mL, and 416 Alexa Fluor 488 conjugated goat anti-rabbit IgG (Thermofisher) at 2 μg/mL. Neutrophils were 417 added to the wells without or with 10 μM of fMLF and left to adhere for 30 min at 37°C. After 418 incubation, neutrophils were fixed with 4% (w/v) PFA for 1 h, washed with PBS, and covered 419 with ProLong gold Antifade reagent (Thermofisher) according to manufacturer's instructions 420 before imaged on a Zeiss LSM880 confocal microscope with a 63x oil objective (NA 1.4). 421 For colocalization of Mac-1 and PDI under fluid shear, neutrophils were incubated with 1 μg/mL 422 of an APC-conjugated anti-CD11b CBRM1/5 and 2 μg/mL of the rabbit anti-PDI DL-11 for 1 h on 423 ice. Neutrophils were then washed and stained with 0.5 μg/mL of an Alexa Fluor 488-424 conjugated goat anti-rabbit IgG for 1 h on ice. After a final wash with Hank's buffered saline 425 solution, neutrophils were primed with 1 μM of fMLF, perfused through microfluidic devices 426 coated with 10 μg/mL of ICAM-1/Fc, and left to settle for 5 min. Neutrophils were then exposed 427 to 0.7 dynes/cm 2 (100 s -1 ) or 5.6 dynes/cm 2 (800 s -1 ) of shear by perfusing Hank's buffered 428 saline solution containing 5 μM of fMLF and imaged using a 63x oil objective (1.4 NA) on a 429 Zeiss LSM880 confocal microscope. 430 Devices were then perfused with PBS for 1 min at a shear stress of 0.175 dynes/cm 2 (25 s -1 ) to 517 wash off non-adherent cells. Tile scan images were taken on a Zeiss LSM 880 confocal 518 microscope and the shear force was doubled every min until 11.2 dynes/cm 2 (1600 s -1 ) was 519 reached. Adherent cells were then quantified and normalized to the number of cells adherent at 520 0.175 dynes/cm 2 . 521 The area under each curve was calculated in GraphPad Prism 9. Data was also best fitted using 522 nonlinear regression for one phase exponential decay in GraphPad Prism 9 to calculate the 523 decay constant (K, cm 2 dynes -1 ) and shear force (F 50 , dynes/cm 2 ) at which 50% of BHK cells 524 were de-adhered from ICAM-1. Initial coordinates of the complex were taken from the X-ray structure of the Leukocyte integrin 541 ɑLβ2 (PDB id. 5E6U) (Sen and Springer, 2016). Note that the βI domains of Mac-1 and LFA-1 542 (ɑLβ2) are identical, while the β-propeller units have a 73% sequence similarity (with 41% 543 residues being identical). The β-propeller consisted of the segments 1-122 and 320-591 of the 544 ɑ domain sequence, while the βI domain corresponded to the amino acids 101-344 of the β2 545 sequence. Four situations were considered: (i) with both C169-C176 and C224-C264 disulfide 546 bonds formed ("wt"), (ii) with C169-C176 bond reduced, (iii) with C224-C264 bond reduced, and 547 (iv) with both disulfide bonds reduced ("both"). The complex (in any of the four forms) was 548 inserted in a dodecahedral simulation box and solvated by ~35165 water molecules. Four 549 Calcium and one Magnesium ions were observed to be bound to the protein in the 550 crystallographic X-ray structure. These ions were considered in the simulation. Surrounding 551 crystallographic water molecules were also considered. Sodium and Chloride ions were added 552 at a concentration of approximately 0.15 M, with an excess of the earlier to ensure an 553 electrically neutral system. The system contained ~116835 atoms in total. 554 The GROMACS MD package (version 2020.3) was employed (Abraham et al., 2015). The 555 CHARMM36 force-field was used for the protein, (Best et al., 2012) the CHARMM TIP3P model 556 for the water molecules, and default CHARMM parameters for the ions. Electrostatic 557 interactions were computed with the Particle mesh Ewald method 558 Essmann et al., 1995). Short-range non-bonded interactions were modelled with a Lennard-559 Jones potential, within a distance of 1.2 nm. Neighbor searching was carried out by using the 560 Verlet buffer scheme (Pall and Hess, 2013). Bonds involving protein hydrogen atoms were 561 constrained using the LINCS algorithm (Hess et al., 1998). Both angular and bond stretching 562 internal motions of water molecules were also constrained by using SETTLE (Miyamoto and 563 Kollman, 1992). Equations of motion were integrated using the Leap-Frog algorithm at discrete 564 time steps of 2 fs. Temperature was maintained constant at 310 K by using the Nose-Hoover 565 thermostat ( (Berendsen et al., 1984;Nose, 1884) for the equilibration steps), using a coupling 566 constant of 1 ps. Pressure was also kept constant at 1 bar by coupling the system to the 567 isotropic Parrinello-Rahman barostat (coupling constant 5 ps) (Parrinella and Rahman, 1981). 568 Before molecular dynamics, the potential energy of the system was minimized by using the 569 steepest descent method. Subsequently, the solvent was equilibrated around the protein, during 570 500 ps at constant volume followed by 1000 ps at constant pressure. During these equilibration 571 steps the protein was maintained position-restrained (elastic constant of 1000 kJmol -1 nm -2 ). For 572 the subsequent production runs the protein restraints were released. N=10 independent 573 simulation replicas were carried out for each system (n=8 when C169-C176 was reduced). The The fraction of simulation time C i,j in which the residue pair (i,j) was found in contact was 587 computed using CONAN (Mercadante et al., 2018). A contact was assumed to be established if 588 the residues came closer than 0.35 nm. F i,j was obtained separately for the four different data 589 sets concatenating all replicas corresponding to each set: C i,j (wt), C i,j (C169-C176), C i,j (C224-590 C246), and C i,j (both). Accordingly the change in contact probability was quantified as the 591 difference ΔC ij (X) = C i,j (X) -C i,j (wt), with X=C169-C176, C224-C246, or both. To assess the 592 statistical significance of the change in contacts ΔC ij (X), C ij was computed separately for each 593 individual replica: C ij r (with r=1,...,N replicas with non-negligible C ij r values). Only pairs with r≥5 594 were considered for this analysis. The following normalized difference function was computed 595 z=: [ <C ij (X)> r -<C ij (wt)> r ] / [ σ 2 r (C ij (X)) + σ 2 r (C ij (wt) ] ½ , with <> r and σ 2 r denoting average and 596 standard deviation squared over the r replicas, respectively. Residue pairs with z>0.5 were 597 considered (main text) and the dependency on z was monitored by setting z> 0.25, 0.5, 0.75, 598 and 1 (Figure 5figure supplement 1). In all cases pairs with non-negligible change were 599 selected, meaning |ΔC ij |>0.4% of the total simulation time. 600 The non-bonded pair-wise force F i,j between the residues i and j was extracted from the 601 simulations by using force distribution analysis (version 2.10.2) (Costescu and Grater, 2013). 602 Analogously to the change in contacts, pair-wise force differences, with respect to the fully 603 oxidized system, were computed: ΔF ij (X) = <F i,j (X)> -<F i,j (wt)>, with X=C169-C176, C224-C246, 604 or both, and with <> denoting time-average over the concatenated trajectory. Similarly as with 605 the contacts, the z normalized difference was determined by computing per-replica time-606 averages of the pair-wise forces. z> 0.25, 0.5, 0.75, and 1 and |ΔF ij |>1 pN threshold values 607 were applied. 608 Neutrophils (1x10 6 cells/mL) were stained with 1 μg/mL of calcein AM for cell tracking and 620 primed with 1 μM fMLF. Neutrophils were added onto microfluidic devices, allowed to settle and 621 adhere for for 5 min. Adherent neutrophils were then exposed to either 0.7 dynes/cm 2 (100s -1 ) 622 or 5.6 dynes/cm 2 (800s -1 ) of shear by perfusing Hank's buffered saline solution containing 5 μM 623 fMLF. Neutrophil crawling was imaged every second using a 40x oil objective (1.2 NA) on an 624 Olympus IX81 fluorescent microscope. 625 To measure neutrophil crawling, the centroid of each neutrophil was defined as the cell's 626 position and was determined by thresholding calcein fluorescence on ImageJ. The cell positions 627 between each frame were then used to determine displacement. The net displacement was 628 calculated by the difference of position at the beginning and end of the time as described by 629 Buffone et al. (Buffone et al., 2018(Buffone et al., , 2019. Crawling speed (µm/min) was calculated from the 630 total distance traveled by a neutrophil from its initial position as determined by cell tracking and 631 dividing it by the total time of migration. The migration index in X-direction is defined as the ratio 632 of the difference between the initial and final X-displacement over the total distance traveled by 633 a cell (X end -X initial )/Distance total . When migration is near 0, there is no preferred direction in cell 634 migration. When migration index for X-displacement is near -1, it indicates cell migration against 635 the direction of flow, whereas migration index for X-displacement is near +1, it indicates cell 636 migration in the direction of flow. The migration index in Y-direction is defined as the ratio of the 637 difference between the initial and final Y-displacement over the total distance traveled (Y end -638 Y initial )/Distance total . Only single cells that remained in the field of view in the duration of 639 experiment were analyzed. Dividing cells and clusters of cells were excluded from analysis. The mass spectrometry data is available via ProteomeXchange with the dataset identifier 664 PXD032688. All other data generated or analyzed are included in the manuscript and 665 supporting files, and are also available from the corresponding authors.  Alexa-Fluor 488-conjugated goat anti-rabbit IgG (green), and high affinity active Mac-1 was 847 detected using an APC conjugated anti-CD11b antibody CBRM1/5 (red). After staining with 848 antibodies, neutrophils were stimulated with 1 µM fMLF, perfused onto ICAM-1 coated 849 microfluidic channels at 0.175 dynes/cm 2 and left to adhere. Adhered neutrophils were 850 subjected to fluid shear at 0.7 dynes/cm 2 or 5.6 dynes/cm 2 for 1 min before being imaged by 851 confocal microscopy. Scale bar represents 10 µm.  adhered to ICAM-1 in static condition. Neutrophils were isolated from human blood and 860 stimulated with fMLF. Surface PDI was detected using Alexa fluor 488 conjugated anti-PDI 861 antibody DL-11 (green) on resting or fMLF-stimulated human neutrophils. Active Mac-1 was 862 detected using an APC conjugated anti-CD11b antibody CBRM1/5 (red). After staining with 863 antibodies, neutrophils were added to ICAM-1 coated surface and allowed to adhere before 864 fixing with 4% paraformaldehyde and imaged by confocal microscopy. Scale bar represents 10 865 µm. 866   or with PDI to immobilized ICAM-1 under increasing shear force. Calcein-stained BHK cells 892 incubated without or with 1µM reduced or oxidized PDI were perfused and allowed to adhere to 893 ICAM-1 coated microfluidic channels for 15 min. Cells were subjected to shear force at each 894 defined shear rate for 1 min (0.175, 0.35, 0.7, 1.4, 2.8, 5.6 and 11.2 dynes/cm 2 ) to allow cell de-895 adhesion. Images were acquired and the number of adhered cells remained at each shear rate  adhere on an ICAM-1/Fc coated 96 well plate at 37°C for 2 h. Cells were washed, stained with 907 calcein AM, and fluorescence at 488/520 nm measured. Fluorescence intensity is shown as the 908 mean ± SEM of 4-5 independent experiments. 909 perfused and allowed to adhere to fibrinogen coated microfluidic channel for 15 min. Cells were 912 subjected to shear force at each defined shear rate for 1 min (0.175, 0.35, 0.7, 1.4, 2.8, 5.6 and 913 11.2 dynes/cm 2 ) to allow cell de-adhesion. Images were acquired and the number of adhered 914 cells remained at each shear rate was quantified. (B) Area under each curve was from panel B. 915 Data represent mean ± SEM of three biological replicates. P<0.05; N.S.=non-significant by one-916 way ANOVA with Dunnett's post-hoc multiple comparisons. 917  conjugated anti-CD18 antibody H52 at 1 µg/mL in the presence or absence of 1 mM MnCl 2 or 933 10 µM fMLF for 30 min at 37°C. Samples were then read immediately by flow cytometry on a 934 BD Accuri C6 flow cytometer. The binding of H52 antibody to β2 integrins on the surface of 935 isolated neutrophils is shown at resting (red), upon integrin extension induced by incubation with 936 Mn 2+ (blue), and upon upregulation of β2 integrins induced by fMLF-stimulation (green) 937 compared to unstained neutrophils (black). 938 Figure 5. Cleavage of the C224-C264 disulfide bond perturbs inter-residue contact and