Glutaredoxin 1 up-regulates deglutathionylation of α4 integrin and thereby restricts neutrophil mobilization from bone marrow

α4 integrin plays a crucial role in retention and release of neutrophils from bone marrow. Although α4 integrin is known to be a potential target of reactive oxygen species (ROS)–induced cysteine glutathionylation, the physiological significance and underlying regulatory mechanism of this event remain elusive. Here, using in vitro and in vivo biochemical and cell biology approaches, we show that physiological ROS-induced glutathionylation of α4 integrin in neutrophils increases the binding of neutrophil-associated α4 integrin to vascular cell adhesion molecule 1 (VCAM-1) on human endothelial cells. This enhanced binding was reversed by extracellular glutaredoxin 1 (Grx1), a thiol disulfide oxidoreductase promoting protein deglutathionylation. Furthermore, in a murine inflammation model, Grx1 disruption dramatically elevated α4 glutathionylation and subsequently enhanced neutrophil egress from the bone marrow. Corroborating this observation, intravenous injection of recombinant Grx1 into mice inhibited α4 glutathionylation and thereby suppressed inflammation-induced neutrophil mobilization from the bone marrow. Taken together, our results establish ROS-elicited glutathionylation and its modulation by Grx1 as pivotal regulatory mechanisms controlling α4 integrin affinity and neutrophil mobilization from the bone marrow under physiological conditions.

␣4 integrin plays a crucial role in retention and release of neutrophils from bone marrow. Although ␣4 integrin is known to be a potential target of reactive oxygen species (ROS)induced cysteine glutathionylation, the physiological significance and underlying regulatory mechanism of this event remain elusive. Here, using in vitro and in vivo biochemical and cell biology approaches, we show that physiological ROS-induced glutathionylation of ␣4 integrin in neutrophils increases the binding of neutrophil-associated ␣4 integrin to vascular cell adhesion molecule 1 (VCAM-1) on human endothelial cells. This enhanced binding was reversed by extracellular glutaredoxin 1 (Grx1), a thiol disulfide oxidoreductase promoting protein deglutathionylation. Furthermore, in a murine inflammation model, Grx1 disruption dramatically elevated ␣4 glutathionylation and subsequently enhanced neutrophil egress from the bone marrow. Corroborating this observation, intravenous injection of recombinant Grx1 into mice inhibited ␣4 glutathionylation and thereby suppressed inflammation-induced neutrophil mobilization from the bone marrow. Taken together, our results establish ROS-elicited glutathionylation and its modulation by Grx1 as pivotal regulatory mechanisms controlling ␣4 integrin affinity and neutrophil mobilization from the bone marrow under physiological conditions. GSH reaction with the free sulfhydryl (-SH) on certain cysteines of proteins induces protein glutathionylation and forms protein-GSH mixed disulfide adducts (Pr-SSG). Glutathionylation was previously considered to prevent proteins from irreversible oxidizing to sulfenic acid, sulfinic acid, or sulfonic acid (1,2). However, it has become increasingly clear that glutathionylation, an important redox posttranslational modification, regulates cellular functions and signal transduction by affecting target proteins, including both intracellular molecules such as actin, titin, and NF-B (3)(4)(5) and extracellular components such as some cytokines and HMGB1 (6).
Glutathionylation is mostly induced by ROS. 9 ROS are a series of highly reactive chemicals mainly containing superoxide, hydrogen peroxide, hydroxyl radicals, etc. (7). ROS are largely produced by phagocytes to fight against microorganisms in inflammatory responses. Recently, there has been a growing body of evidence showing novel functions of ROS in physiological cellular signaling and functionality. Deglutathionylation is mainly regulated by oxidoreductases such as Grx (8,9). Grx, a family of thioltransferases, removes GSH from Pr-SSG via a thiol-disulfide exchange reaction (10,11). In mammalian cells, there are two main Grx isoforms, Grx1 and Grx2. Grx2 is primarily located at mitochondria, whereas Grx1 is found in many locations, including the cytosol, nucleus, and extracellular matrix (12,13), suggesting that glutathionylation of extracellular proteins might be mediated by Grx1. A recent study has shown that Grx1 regulates interleukin-1␤ deglutathionylation and its bioactivity (14).
A previous studies suggested that glutathionylation might regulate the binding of ␣4 integrin on eosinophil to its ligand VCAM-1 (15). However, whether glutathionylation is essential for ␣4-mediated neutrophil adhesion and related physiological cellular processes is unknown. In this study, we demonstrated that ROS induced ␣4 integrin glutathionylation and subsequently regulated the adhesion of neutrophils to endothelial cells. We further showed that Grx1 reversed the glutathionylation of ␣4 integrin physiologically. Disruption of Grx1 increased ␣4 integrin glutathionylation and inhibited neutrophil binding to endothelial cells. On the contrary, Grx1 overexpression or Grx1 protein addition resulted in opposite effects.
Under rest conditions, most neutrophils exist in the bone marrow. During infection or inflammation, massive neutrophils are mobilized from the bone marrow to the circulation (16). To avoid improper infection or excessive inflammation, it is of great importance to appropriately govern bone marrow neutrophil retention and mobilization. Multiple lines of evidence have demonstrated that ␣4 integrin is the central integrin to regulate marrow neutrophil retention and mobilization (17)(18)(19)(20). As stress conditions induce elevated level of ROS, we speculated that Grx1-and ROS-mediated ␣4 integrin glutathionylation might be involved in neutrophil mobilization from the bone marrow. Our results showed that Grx1 depletion resulted in increased neutrophil egress from the bone marrow and concomitantly increased ␣4 integrin glutathionylation and vice versa. This study reveals that Grx1-mediated and ROS-induced ␣4 integrin glutathionylation are important physiological regulatory mechanisms controlling ␣4 integrin functions in neutrophils and then participate in bone marrow neutrophil mobilization under stress conditions.

ROS induced elevated ␣4 integrin glutathionylation in neutrophils
The bone marrow extracellular space is highly oxidizing, even stronger than the intracellular compartment (14). Thus, extracellular proteins of neutrophils might be regulated by ROS-induced glutathionylation. First, we performed in vitro glutathionylation assays with biotinylated GSH (BioGEE) to confirm the possible modifications of ␣4 integrin. As expected, recombinant ␣4 integrin was glutathionylated in vitro (Fig. 1A). H 2 O 2 significantly increased the ␣4 -BioGEE mixed disulfide adducts (BioGSS-␣4) signal, whereas treatment with DTT eliminated glutathionylation of ␣4. To determine the modification of ␣4 integrin in neutrophils, we tested ␣4 glutathionylation with a GSH-specific antibody. Although no band was detected in neutrophil-like differentiated HL60 (dHL60) cell lysates, there was a band corresponding to the size of ␣4 integrin after formyl-methionyl-leucyl-phenylalanine (fMLF) stim- Then the solutions were subjected to immunoblotting to detect BioGEE-modified ␣4 with streptavidin-horseradish peroxidase (STR-HRP). Total protein loading was evaluated with Coomassie blue dye. Every treatment used 0.5 g protein. B, neutrophil-like dHL60 cells were treated with the indicated concentrations of fMLF for 5 min. Cells were harvested and subjected to immunoblotting for GSH and ␣4. C, dHL60 cells were left untreated or stimulated with 10 M fMLF for 5 min. Whole-cell lysates were then subjected to ␣4 immunoprecipitation (IP) followed by immunoblotting for GSH-modified ␣4. Blank agarose was tested as a negative control. HMGN2 (without cysteine) antibody-induced pulldown was used as a control. D, dHL60 cells were left untreated or stimulated with 10 M fMLF for 5 min. Whole-cell lysates were then subjected to GSH IP followed by immunoblotting for ␣4. Actin antibody was used as a control. E, murine neutrophils were left untreated or stimulated with 1 M fMLF for 5 min. Whole-cell lysates were subjected to immunoblotting for GSH and ␣4. F, dHL60 cells were pretreated or not pretreated with 50 M DPI for 30 min. Then both groups of dHL60 cells were stimulated with the indicated concentrations of fMLF for 5 min. Cells were harvested and subjected to immunoblotting for GSH and ␣4. G, GSH was docked into the ␣4 integrin crystal structure in complex with the ␣4 integrin ␤-propeller and thigh domain crystal structure (PDB code 4IRZ). The dominant pose of 10 docking runs is represented as sticks, and interacting cysteine is represented as sticks. H, dHL60 cells were transfected with constructs expressing either the WT or the 278 Asp mutant form of FLAG-␣4 integrin. These dHL60 cells were stimulated with 10 M H 2 O 2 for 5 min. Whole-cell lysates were then subjected to FLAG IP followed by immunoblotting for GSH-modified FLAG-␣4. All data are representative of at least three separate experiments.

Regulation of ␣4 integrin by glutathionylation
ulation (Fig. 1B). The glutathionylation signals elevated along with increasing concentrations of fMLF (1-10 M) (Fig. 1B) and reached a peak at the 5-min time point, which then decreased along the time course (Fig. S1). To confirm whether the modified protein was ␣4 integrin, lysates of fMLF-treated dHL60 cells were pulled down with ␣4 antibody or GSH antibody and then were tested with ␣4 and GSH antibodies. Both assays showed that ␣4 integrin of neutrophils was indeed glutathionylated by ROS (Fig. 1, C and D). Glutathionylated ␣4 signals were detected after pulldown enrichment, even in nontreated dHL60 cells. Similarly, in murine neutrophils, fMLF triggered the glutathionylation of ␣4 integrin (Fig. 1E). Also, glutathionylated ␣4 was detected even without fMLF treatment. It is well known that ROS are mostly generated by NADPH oxidase in chemoattractant-stimulated neutrophils (13). To investigate whether glutathionylation of ␣4 integrin was directly induced by ROS, dHL60 cells were pretreated with diphenyleneiodonium chloride (DPI), which is an inhibitor of NADPH oxidase. After fMLF treatment, DPI-pretreated cells showed a significant decrease in ␣4 glutathionylation (Fig. 1F). These data indicate that ␣4 integrin is glutathionylated in neutrophils and that physiological ROS drastically augments the glutathionylation signal.
To predict the modified cysteines of ␣4 integrin by glutathionylation, molecular docking was performed by docking GSH into the ␣4 integrin ␤-propeller and thigh domain crystal structure (PDB code 4IRZ). It was shown that the S atom of Cys278 of ␣4 integrin was quite close to the S atom of GSH (4.9 Å) (Fig.  1G), indicating that these two S atoms might be able to form a disulfide bond. To verify the role of Cys-278 in ␣4 integrin glutathionylation, constructs expressing either mutation of Cys278 to Asp (278 Asp ) or WT FLAG-␣4 were transfected in HL60 cells. dHL60 cells with 278 Asp mutant ␣4 integrin displayed much lower H 2 O 2 -elicited glutathionylation of ␣4 integrin (Fig. 1H). The result suggests that Cys278 might be one of the glutathionylated sites of ␣4 integrin.

ROS regulated the binding of neutrophils to HUVECs by ␣4 integrin and VCAM-1
Next we tested the role of ROS-induced glutathionylation of ␣4 integrin in neutrophils adhering to endothelial cells using an in vitro adhesion assay of 5-(and 6)-carboxyfluorescein diacetate succinimidyl ester (CFSE)-labeled neutrophils binding to a human umbilical vein endothelial cell (HUVEC) monolayer. As shown in Fig. 2A, either H 2 O 2 or fMLF treatment elicited apparent increased adhesion of dHL60 cells to HUVECs, whereas ␣4 antibody treatment disrupted the effect. Consistently, fMLF-treated murine neutrophils showed elevated adhesion to HUVECs, and the increased adhesion was inhibited by the ␣4 antibody (Fig. 2B). Also, zaurategrast, an inhibitor of ␣4 integrin (21), displayed a similar effect (Fig. 2C). A membrane-impermeable sulfhydryl blocker, N-ethylmaleimide treatment largely decreased adhering neutrophils (Fig. S2). The adhesion assay was also performed with VCAM-1 antibody or IgG-pretreated HUVECs. VCAM-1 antibody, but not IgG, abolished the effect of ROS-triggered elevated adhesion (Fig.  2D). Similar results were obtained in the adhesion of murine neutrophils to HUVECs (Fig. 2E). We also examined the effect of mutation of Cys278 to Asp on ROS-induced increased adhesion of neutrophils to HUVECs. As shown in Fig. 2F, the adhesion rate of dHL60 cells with 278 Asp mutant ␣4 integrin to

Regulation of ␣4 integrin by glutathionylation
HUVECs was significantly decreased compared with WT dHL60 cells after H 2 O 2 stimulation. These results indicate that ROS-induced glutathionylation promotes the adhesion of neutrophils to endothelial cells through ␣4 integrin and VCAM-1.

Grx1 physiologically catalyzed ␣4 integrin deglutathionylation
Grx is the main reductase of glutathionylation in mammals. In addition to the action in intracellular components, Grx1 has been reported to exist with catalytic activity in some extracellular environments, such as sputum, plasma, and bone marrow extracellular space (14,22,23). Therefore, we speculated that Grx1 might physiologically modulate the glutathionylation of ␣4 integrin. We primarily tested the effect of Grx1 on glutathionylation of ␣4 integrin in vitro. In a very similar fashion as DTT treatment, fusion Grx1 protein addition significantly weakened the glutathionylation signal of FLAG-␣4 (Fig. 3A). The glutathionylated ␣4 integrin signal in dHL60 cells was also decreased when adding Grx1 protein to the culture medium (Fig. 3B).
To further evaluate the role of Grx1 in glutathionylation of ␣4 integrin, we examined the glutathionylation of ␣4 integrin in Grx1-overexpressing dHL60 cells. Grx1 overexpression drastically decreased the glutathionylation of ␣4 in response to fMLF but barely affected the expression of ␣4 integrin (Fig. 3, C and D, and Fig. S3A). With increasing concentrations of fMLF stimulation, the ␣4-SSG signals of Grx1-overexpressing dHL60 cells also elevated but were always weaker than the control. By contrast, a Grx inhibitor, Cd 2ϩ , slightly increased ␣4 glutathionylation in neutrophils (Fig. 3E). To investigate whether Grx1 directly participated in ␣4 integrin glutathionylation, a Grx1 knockout mouse was used (Fig. S3B). Similar to Cd 2ϩ treatment, murine Grx1-deficient neutrophils showed much stronger ␣4 glutathionylation than the WT cells (Fig. 3F). Together, our results demonstrate that ␣4 deglutathionylation is physiologically controlled by Grx1.

Grx1 inhibited the ligand binding activity of ␣4 integrin in neutrophils to VCAM-1
As our data show, glutathionylation modulated integrin ␣4/VCAM-1dependent neutrophil adhesion by modifying ␣4. Grx1, the main regulator of ␣4 integrin deglutathionylation, might also be involved in this process. Therefore, the effect of Grx1 on ␣4-mediated neutrophil binding to immobilized VCAM-1 was tested first. According to a previous study (15) and our pretest, 1 g/ml was chosen as the concentration of recombinant ligands. The effects of Grx1 on neutrophil binding to immobilized Fc were also examined as controls. Grx1 overexpression reduced dHL60 cell adhesion to VCAM-1 but not Fc (Fig. 4, A and B). Then the binding activity in murine Grx1 Ϫ/Ϫ neutrophils was detected. In contrast, Grx1 deficiency activated murine neutrophils adhering to VCAM-1 but not Fc (Fig. 4, C and D). Interestingly, even without H 2 O 2 treatment, Grx1 already displayed an impairing effect in ligand binding activity of ␣4 integrin in this assay (Fig. 4, A and C). Our results suggested that Grx1 might inhibit neutrophil binding to VCAM-1. To further evaluate the function of extracellular Grx1, murine WT and Grx1 Ϫ/Ϫ neutrophils were pretreated with Grx1 protein. The result showed that Grx1 pretreatment suppressed the adhesion of WT neutrophils to VCAM-1 and restored the After adding 1 g of Grx1 protein and being incubated for another 15 min, the solutions were subjected to immunoblot analysis for BioGEE-modified ␣4 with streptavidin-horseradish peroxidase (STR-HRP), and total ␣4 loading was evaluated with FLAG and ␣4 antibodies. B, dHL60 cells were preincubated with 1 g of Grx1 protein or not for 30 min and then treated with 1 M fMLF for 5 min. Cells were harvested and subjected to immunoblot analysis for GSH and ␣4. C, control dHL60 cells (Phage-dHL60) and Grx1-overexpressing dHL60 cells (Grx1-dHL60) were harvested and subjected to immunoblot analysis for ␣4 and actin detection. D, phage-dHL60 and Grx1-dHL60 cells were treated with the indicated concentrations of H 2 O 2 for 5 min. Cells were harvested and subjected to SDS-PAGE followed by immunoblot analysis for GSH and ␣4. E, Grx1-dHL60 cells were pretreated with 2 mM Grx1 inhibitor Cd 2ϩ or not for 30 min and then stimulated with the indicated concentration of H 2 O 2 for 5 min. Cells were harvested and subjected to SDS-PAGE followed by immunoblot analysis for GSH and ␣4. F, WT and Grx1 Ϫ/Ϫ mouse neutrophils were treated with the indicated concentrations of H 2 O 2 for 5 min. Cells were harvested and subjected to SDS-PAGE followed by immunoblot analysis for GSH and ␣4. All data are representative of at least three separate experiments.
In addition, we performed a ligand-binding assay to soluble VCAM-1. Soluble VCAM-1 was prelabeled with FITC and then detected by flow cytometry. The results showed that more murine Grx1 Ϫ/Ϫ neutrophils bound to FITC-labeled VCAM-1 than the WT after H 2 O 2 treatment, whereas, under rest conditions, cells displayed a similar binding ability to VCAM-1 (Fig.  4G). These results also support the inhibiting effect of Grx1 on the ligand-binding activity of ␣4 integrin to VCAM-1.

Grx1 suppressed the adhesion of neutrophils to endothelial cells
Next, the role of Grx1 in neutrophils adhering to endothelial cells was investigated. Primary results showed that recombinant Grx1 pretreatment decreased the adhesion of dHL60 cells to the HUVEC monolayer in response to H 2 O 2 (Fig. 5A), and fewer Grx1-overexpressing dHL60 cells adhered to HUVECs (Fig. 5B). Then Grx1-overexpressing dHL60 cells were pretreated with Cd 2ϩ , which rescued the impairing effect of Grx1 on dHL60 cells adhering to HUVECs after H 2 O 2 stimulation (Fig. 5C). As Cd 2ϩ is membrane-permeant, it may inhibit both intracellular and extracellular Grx1 activities (24). We thus performed the experiment pretreated with a Grx1-specific antibody to block the extracellular Grx1 activity. Similarly, Grx1 antibody-treated dHL60 cells displayed significantly increased adhesion to HUVECs (Fig. 5D). A Wright-Giemsa staining assay was also explored to directly observe the adhesion of neutrophils to HUVECs. As Fig. 5E shows, the morphology of HUVEC was much larger than that of dHL60 cell. Therefore, it is very clear to distinguish dHL60 cells from HUVECs. It can be observed that much fewer Grx1-overexpressing dHL60 cells attached to HUVECs than the control (Fig. 5, E and F).
In murine neutrophils, Grx1 pretreatment also impaired the adhesion of neutrophils to HUVECs (Fig. S4). In contrast, Grx1 abolishment promoted murine neutrophils adhering to the HUVEC monolayer (Fig. 5G). Then murine Grx1 Ϫ/Ϫ neutrophils were treated with recombinant Grx1 protein. As expected, the elevated adhesion after H 2 O 2 stimulation was attenuated (Fig. 5H). It is noteworthy that sometimes Grx1 showed an impairing effect on adhesion of neutrophils to HUVECs under rest conditions, which might be due to the basic ROS level under rest conditions that has been detected in a previous study (25). Consistently, we observed a visible glutathionylated ␣4 signal in murine Grx1 Ϫ/Ϫ neutrophils without stimulation (Fig.  3F). Collectively, these data show that Grx1 might negatively modulate the adhesion of neutrophils to endothelial cells through ␣4/VCAM-1 signaling in response to ROS.

Grx1 inhibited neutrophil mobilization from the bone marrow via ␣4 integrin
Increasing evidence has shown that integrin ␣4/VCAM-1 is critical adhesion signaling in the mobilization of neutrophils from the bone marrow. We explored neutrophil mobilization from the bone marrow during inflammation using an intraperitoneal Escherichia coli-induced acute peritonitis model in live mice. As expected, mice pretreated with zaurategrast displayed impaired neutrophil egress from the bone marrow (Fig. 6A).
As our in vitro results showed that Grx1 modulated integrin ␣4/VCAM-1 adhesion signaling, we speculated that Grx1 might take part in neutrophil mobilization from the bone marrow by regulating ␣4 integrin. Preintravenous injection of

Regulation of ␣4 integrin by glutathionylation
recombinant Grx1 protein attenuated the glutathionylated ␣4 integrin of bone marrow neutrophils (Fig. 6B). Grx1-pretreated mice showed fewer neutrophils in peripheral blood, only half compared with control mice, at the 1-h time point after challenge (Fig. 6C), and even before peritonitis, administration of Grx1 impaired the glutathionylation of ␣4 integrin and decreased the number of neutrophils in peripheral blood (Fig.  S5).
On the contrary, we compared bone marrow neutrophil mobilization between WT and Grx1 Ϫ/Ϫ mice. We also checked the glutathionylation of ␣4 integrin and neutrophil number in peripheral blood 1 h after intraperitoneally injecting live E. coli. The Western blot result showed that Grx1 deficiency induced a dramatically elevated level of glutathionylated ␣4 integrin signal (Fig. 6D), and many more neutrophils in the circulation were detected in Grx1 Ϫ/Ϫ mice (Fig. 6E). We also measured the alternation of neutrophil number in the bone marrow at the 1-h time point after the challenge by flow cytometry. The variation tendencies of neutrophil numbers in the bone marrow were contrary to those in peripheral blood (Fig. S6), but no significant differences were observed in zaurategrast pretreatment, Grx1 pretreatment, or Grx1 Ϫ/Ϫ mice compared with the control. As noted previously, Grx1 knockout mice were similar to their WT littermates in differential leukocyte counts in the bone marrow and peripheral blood (13). Furthermore, ␣4 integrin inhibitor (zaurategrast) pretreatment lessened the effects of Grx1 depletion. Zaurategrast-treated Grx1 Ϫ/Ϫ mice showed similar neutrophil number as WT mice in the circulation (Fig. 6F).
Immunohistochemistry was used to examine the location of neutrophils in the bone marrow after inflammation. The slides showed that more Grx1-attenuated neutrophils appeared near the marrow venous sinusoidal endothelium than WT neutrophils ( Fig. 6G and Fig. S7, red). These results suggest that Grx1 might prevent neutrophil egress from the bone marrow via ␣4 integrin in response to stimuli.

Regulation of ␣4 integrin by glutathionylation
found that ROS-induced ␣4 integrin glutathionylation and its reversal by Grx1 were crucial mechanisms controlling binding of VCAM-1 by ␣4 integrin in neutrophils and then promoted neutrophil mobilization from the bone marrow in inflammation (Fig. 7), providing some new clues for the treatment of exaggerated infections or inflammation-related diseases.

Discussion
Integrins are a series of transmembrane ␣ and ␤ subunitformed adhesion molecules mediating cell adhesion in many processes (29). In this study, we revealed that Grx-and ROSmediated glutathionylation regulated ␣4 integrin affinity in neutrophils and further established a role for Grx1 in acute bone marrow neutrophil mobilization through suppressing integrin ␣4/VCAM-1 adhesion signaling.
ROS-induced glutathionylation, as a reversible redox modification, controls the activities of target proteins and further modulates cellular function and signaling. Accumulating studies have reported that some extracellular molecules might be regulated by glutathionylation (6). Our data showed that neutrophil surface ␣4 integrin was glutathionylated by ROS both in vitro and in vivo, and demonstrated that ROS-triggered glutathionylation promoted ␣4 integrin-derived neutrophils adhering to the endothelium. Other studies on ␣4 integrin in eosinophils, RAW264.7 cells, or melanoma B16 cells also supported our findings (15,30,31). They found that a high concentration Figure 6. Grx1 inhibited neutrophil mobilization from the bone marrow via ␣4 integrin. A, mice were pretreated with zaurategrast (5 mg/kg) or PBS for 1 h. Mice were then challenged with E. coli (injected intraperitoneally, 1 ϫ 10 7 /mouse) or PBS. Neutrophils in peripheral blood were counted 1 h after challenge (n ϭ 3/group). *, p Ͻ 0.05; **, p Ͻ 0.01 versus control. B and C, mice were pretreated with Grx1 (injected intravenously, 1 g/mouse) for 1 h. Mice were then challenged with E. coli (injected intraperitoneally, 1 ϫ 10 7 /mouse). B, murine neutrophils purified at the 1-h time point and subjected to immunoblot analysis for GSH and ␣4 (n ϭ 3/group). C, neutrophils in peripheral blood were counted 1 h after challenge (n ϭ 3/group). *, p Ͻ 0.05 versus PBS pre-administrated mice. D and E, WT and Grx1 Ϫ/Ϫ mice were challenged with E. coli (injected intraperitoneally, 1 ϫ 10 7 /mouse). D, murine neutrophils purified at the 1-h time point and subjected to immunoblot analysis for GSH and ␣4 (n ϭ 3/group). E, neutrophils in peripheral blood were counted 1 h after challenge (n ϭ 3/group). **, p Ͻ 0.01 versus WT mice. F, WT and Grx1 Ϫ/Ϫ mice were pretreated with zaurategrast (5 mg/kg) or PBS for 1 h. Mice were then challenged with E. coli (injected intraperitoneally, 1 ϫ 10 7 /mouse). Neutrophils in peripheral blood were counted 1 h after challenge (n ϭ 3/group). *, p Ͻ 0.05 versus PBS-pretreated WT mice; n.s., not significant. Data are the mean Ϯ S.D. of three mice. One representative of three experiments is shown. G, the bone marrows of WT and Grx1 Ϫ/Ϫ mice after treatment as in E was stained with myeloperoxidase antibody (MPO, red) and 4Ј,6-diamidino-2-phenylindole (DAPI, blue). Representative fluorescence images are shown. KO, knockout.

Regulation of ␣4 integrin by glutathionylation
of H 2 O 2 (100 M) suppressed the adhesion of eosinophil surface ␣4 integrin to VCAM-1. Similarly, in our study, 100 M H 2 O 2 also inhibited the adhesion of neutrophils to endothelial cells and decreased glutathionylation of ␣4 integrin (Fig. 3D). It is possible that overdose of ROS might induce irreversible oxidization, but not glutathionylation. However, recent studies showed that the level of ROS was less than 10 M in the bone marrow extracellular space under stress conditions (14,25). We also examined the ROS level of the bone marrow extracellular space under both rest and inflammatory conditions. The value was no more than 20 M even after inducing inflammation. In short, ROS may rarely reach 100 M under physiological conditions. Therefore, we chose no more than 10 M of H 2 O 2 to perform experiments to investigate the role of physiological ROS-triggered glutathionylation on ␣4 integrin.
Glutathionylation is reversed by some reducing enzymes, which is called deglutathionylation. There are three main groups of molecules acting for deglutathionylation, including Grx, thioredoxin (Trx), and sulfiredoxin (Srx) (32,33). They oxidize their own free thiols and then reduce protein thiols. Classically they are found to mediate deglutathionylation of intracellular proteins. Some groups have reported activity of Grx1 in the extracellular area (6,14,22,23,34), implying that Grx1 might play reducing roles in the extracellular space. Structural analysis of ␣4 integrin shows that all 24 cysteines are located at its extracellular domains (35). Therefore, we assumed that deglutathionylation of ␣4 integrin might be modulated by extracellular Grx1. As expected, we found that recombinant Grx1 protein treatment or Grx1 overexpression attenuated the glutathionylation of ␣4 integrin and, thus, drastically inhibited ␣4-derived neutrophils adhering to VCAM-1 or endothelial cells. By contrast, an inhibitor of Grx (Cd 2ϩ ), neutralizing Abs against Grx1, or Grx1 depletion significantly elevated the level of glutathionylated ␣4 integrin and thus promoted the adhesion of neutrophils to VCAM-1 or endothelial cells. Moreover, the increasing adhesion was prevented by preculture with Grx1 protein in murine Grx1 Ϫ/Ϫ neutrophils. To the best of our knowledge, these results are the first to show the effect of extracellular Grx1 on membrane-anchored proteins via glutathionylation. However, the roles of Trx and Srx in glutathionylation of ␣4 integrin need to be studied in the future.
As reported previously (36), the ␣4 subunit participates in two integrins, ␣4␤1 (very late antigen 4, VLA-4) and ␣4␤7 (lymphocyte Peyer's patch adhesion molecule, LPAM). LPAM is expressed at very low levels on neutrophils and appears to have few effects on neutrophil mobilization from the bone marrow (19). Therefore, the effects of Grx1 on adhesion of neutrophils to endothelial cells might be contributed by regulating VLA-4 specifically. Although the ␤1 subunit has been proven not to be modified by glutathionylation (30), Grx1 and ROS might mainly target the ␣4 subunit. A previous study showed that Cys278 is important for ␣4 integrin ligand-binding (15,37). A docking analysis of GSH and ␣4 integrin predicted that the sites of glutathionylation might include Cys278, and mutation of Cys278 to Asp showed reduced glutathionylation of ␣4 integrin and impaired ROS-induced elevated adhesion of neutrophils to HUVECs, suggesting that glutathionylation is crucial to increase ␣4 integrin affinity. ␣4 integrin contains some other

Regulation of ␣4 integrin by glutathionylation
unique cysteine residues at positions 717, 767, and 828 in nonanalyzed domains (30,38). Thus, there might be other modified cysteine residues in ␣4 integrin. The exact sites of glutathionylation still need to be determined in a future study.
Integrins are of importance in many immunological, physiological, and pathological processes, especially leukocyte extravasation from blood to tissue and progenitor cell/leukocyte release from the bone marrow (39). It is quite clear that ␤2 integrins are the key integrins driving the adhesion of neutrophil migration from blood to tissue (40,41). ␣4 integrin seems to play few roles in this process, with controversial expression on peripheral blood neutrophils (42)(43)(44). A wide range of studies revealed that VLA-4/VCAM-1 adhesion signaling is essential in homeostatic stem cell/progenitor cell egress from the bone marrow (45,46). Although the effect of VLA-4 on neutrophil release from the bone marrow under rest condition is debated, VLA-4 has been proven to mediate marrow neutrophil mobilization under stress conditions (17,19). Bone marrow neutrophils have been shown to express VLA-4 (17), and a specific blocking antibody or antagonist of ␣4 inhibits MIP-2induced bone marrow neutrophil mobilization (17). Our results also demonstrated that inhibition of ␣4 integrin decreased the number of neutrophils in the circulation after inflammation.
As we know, ␣4 integrin is also expressed in some other cells, such as monocytes, macrophages, and T cells, and in some tumor cells, such as melanoma cells. Several studies have shown that oxidative stress is involved in some ␣4-dependent cellular functions. Low-dose ionizing radiation-induced ROS promoted the avidity of VLA4 of RAW264.7 and VCAM-1 (30). Treatment with the ROS scavenger N-acetylcysteine elevates peripheral blood mononuclear cell (PBMC) surface thiol and regulates VLA4-dependent adhesion and Jurkat aggregation (47). Grx1 has been reported to participate in some related processes or diseases. Recent studies showed that overexpression of Grx1 inhibited monocyte chemotaxis induced by MCP-1 (48), and Grx1 KO mice are more susceptible to pulmonary fibrosis (49). Elevated Grx1 activity diminishes the development of Parkinson's disease (50). However, few studies have reported interactions of Grx1 and VLA4 in those processes, which need further investigation.
Given that Grx1 regulates the affinity of ␣4 integrin in neutrophils, Grx1 might be involved in stress-induced bone marrow neutrophil mobilization. Here we uncovered Grx1-modulated neutrophil mobilization from the bone marrow in inflammatory conditions. We examined the neutrophil number in peripheral blood at the 1-h time point after induction of peritonitis. Mice preadministered recombinant Grx1 protein showed significantly decreased neutrophil numbers in the circulation and suppressed glutathionylated ␣4 integrin in bone marrow neutrophils. By contrast, compared with WT mice, Grx1 Ϫ/Ϫ mice displayed many more neutrophils in peripheral blood and enhanced glutathionylation of ␣4 integrin of neutrophils in the bone marrow. Similarly, we observed that Grx1 Ϫ/Ϫ mice egressed more neutrophils from the bone marrow after MIP-2 stimulation.
In conclusion, our study identified Grx1-mediated and ROSinduced glutathionylation as a novel regulating mechanism of bone marrow neutrophil mobilization under stress conditions. By mediating ␣4 integrin glutathionylation, Grx1 suppressed the adhesion of neutrophils to the marrow venous sinusoidal endothelium and thus impaired neutrophil migration from marrow to blood in response to infection or stress.

Experimental procedures
Cell culture HL60 cells were purchased from the ATCC (CCL-240 TM ) and cultured in RPMI 1640 medium (Gibco) supplemented with 2 mM L-glutamine, 15% heat-inactivated fetal bovine serum (Gibco), and 1% penicillin/streptomycin. dHL60 cells were treated with 1.3% DMSO for 6 days in Iscove's modified Dulbecco's medium to be differentiated to a neutrophillike phenotype. HUVECs were purchased from the ATCC (CRL-1730 TM ) and maintained in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin. HEK293T cells were obtained from the ATCC (CRL-11268 TM ) and cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% penicillin/streptomycin.

Mice
Grx1 Ϫ/Ϫ mice were obtained from Cyagen (KOCMP-23992-Glrx). In all experiments, we used age-matched C57BL/6 mice as WT controls. 6-to 8-week-old C57BL/6 mice were obtained from DaShuo and housed in the animal facilities of the West China School of Basic Medical Sciences and Forensic Medicine at Sichuan University. All procedures that involved mice were approved by the ethics committee of the West China School of Basic Medical Sciences and Forensic Medicine at Sichuan University.

Cell adhesion assay
Here we performed an in vitro adhesion assay based on a previous study (15). For adhesion to immobilized proteins, soluble VCAM-1 or Fc at 10 g/ml in PBS was cultured overnight Regulation of ␣4 integrin by glutathionylation at 4°C in 96-well plates. We treated the plate wells with 0.5% BSA to block nonspecific binding. Cells were labeled with 5 M CFSE and cultured for 20 min at room temperature. The CFSElabeled dHL60 cells were pretreated with H 2 O 2 or fMLF and then adhered to immobilized proteins for 60 min at 37°C. For adhesion to the HUVEC monolayer, HUVECs were seeded at 5 ϫ 10 3 cells/well into a 96-well plate and cultured overnight. CFSE-labeled cells were pretreated with H 2 O 2 or fMLF and adhered to the HUVEC monolayer for 30 min at 37°C before washing and detection of cell adhesion. Next, we removed nonadherent cells and detected the fluorescence. The adhesion ratio refers to the ratio of adherent cell fluorescence to input fluorescence.
Alternatively, we stained both attached dHL60 cells and HUVECs via Wright-Giemsa staining and then counted these two types of cells by light scope. The cell adhesion ratio was defined as the ratio of dHL60 cells to HUVECs.

FITC-labeled recombinant protein binding assay
The binding assay was modified from a previous study (15). Briefly, recombinant proteins were labeled with FITC using a FITC labeling kit (Gbiosciense). 1 ϫ 10 6 cells/ml were pretreated with or without H 2 O 2 in Hank's balanced salt solution (HBSS) for 10 min and then cultured with 1 g of FITC-VCAM-1 in 0.1 ml HBSS at room temperature for 10 min. Finally, we tested the intensity of FITC by flow cytometry.

Molecular docking
Possible glutathionylated sites of ␣4 were predicted by docking GSH into the ␣4 integrin ␤-propeller and thigh domain crystal structure (extraction from the ␣4␤7 crystal structure in complex with Fab natalizumab, PDB code 4IRZ) via CDOCKER, which was implemented in Accelrys Discovery Studio Client (DS) version 3.1 (Accelrys Software Inc., San Diego, CA) (51).

Immunohistochemistry
Marrow plugs were obtained from mouse femora after decalcification and embedded with paraffin before sectioning and mounting. For staining, slides were permeabilized with 0.1% Triton X-100 in PBS and blocked with 1% BSA followed by 10% goat serum, and then incubated with rabbit anti-CD31 (GB11063-3, 1:100) overnight at 4°C, followed by Alexa phosphatidylethanolamine goat anti-rabbit IgG (P9795, 1:400) for 2 h. Slides were then washed, counterstained with 4Ј,6-diamidino-2-phenylindole (Sigma-Aldrich) before scanning with Pannoramic MIDI (3D HISTECH) with the attached Quant center software (Optical Analysis) and associated CaseViewer software. Marrow for MPO staining was prepared as above, except for imaging with an Olympus BX50 inverted microscope (Olympus America) with an attached Optronics MagnaFire digital camera (Optical Analysis) and associated MagnaFire software (version 2.0) when the images were magnified about ϫ100.

Marrow neutrophil acute mobilization model, glutathionylation, and other neutrophil functional assays
Peritonitis was induced by intraperitoneal injection with 1 ϫ 10 7 E. coli (strain 19138, ATCC) in WT and Grx1 Ϫ/Ϫ mice. One hour after the challenge, peripheral blood was collected through the fundus after anesthetizing mice with isoflurane and tested using a hematology analyzer. Then mice were sacrificed, and bone marrow neutrophils were purified. Purified neutrophils were probed with GSH and ␣4 integrin antibodies. For recombinant Grx1 treatment, 1 g of Grx1 protein was intravenously injected into the tail vein 1 h before inducing peritonitis. For ␣4 inhibitor treatment assay, zaurategrast was intraperitoneally injected at a concentration of 5 mM/kg 3 h in advance. Some other related assays, including neutrophil isolation, immunoprecipitation, and Wright-Giemsa staining, have been described in a previous publication (13).

Statistical analysis
All comparisons were performed with two-tailed Student's t test using Microsoft Excel. p Ͻ 0.05 was considered significant *, p Ͻ 0.05, **, p Ͻ 0.01, ***, p Ͻ 0.001. All experiments were performed at least three times.