Cytosolic Guanine Nucleotide-binding Protein Rac2 Operates in Vivo as a Component of the Neutrophil Respiratory Burst Oxidase TRANSFER OF Rac2 AND THE CYTOSOLIC OXIDASE COMPONENTS ~ 4 7 ~ ~ " AND p67Phor TO THE SUBMEMBRANOUS ACTIN CYTOSKELETON DURING OXIDASE ACTIVATION*

The respiratory burst oxidase is responsible for 0, production in stimulated neutrophils and B lymphocytes. Components of this oxidase include cytochrome bbeS, a membrane-bound flavohemoprotein; the cytosolic polypeptides p47&OX and ~ ; and one or more small G proteins including Racl, RacZ, and/or RaplA. We found that when normal neutrophils were acti- vated, small percentages of each of the cytosolic proteins p47phox, p67*OX, and RacZ were transferred to the membrane cytoskeleton. However, Rac2 was not transferred to the membrane during activation of p47Ph0"-deficient neutrophils. In normal cells, some p47ph0x also became associated with the non-cytoskel-eta1 portion of the plasma membrane, but ~ 6 7 ~ " , RacZ, and 02-forming activity were restricted to the cytoskeleton. Neutrophil activation also causes the phosphorylation of multiple serines in ~ 4 7 ~ " " . The most heavily phosphorylated forms of p47*OX were found solely in the membrane cytoskeleton. These results suggest that

anti-Racl and anti-Rac2 antibodies were each raised against unique peptides near the C termini of Racl and Rac2, respectively.
Preparation of Neutrophil Fractions-Neutrophils were obtained from normal adult human subjects and from a patient with ~4 7~~deficient chronic granulomatous disease as described previously (25). The neutrophils (IO8 cells/ml) were suspended in Ca2'-and MgZ+free Dulbecco's phosphate-buffered saline, treated with 2.4 mM diisopropyl fluorophosphate (DFP) for 30 min at 4 "C as described elsewhere (26), and then washed and stimulated with 1 pg/ml PMA at 37 "C for 10 min unless otherwise indicated. Neutrophil membranes and cytosol were then prepared by a published method (19).
DNase I Treatment of Membranes-DNase I (2 mg/ml) or carbonic anhydrase (2 mg/ml) was dissolved in cytoskeleton buffer described above and treated with DFP for at least 15 min. Membranes were treated with the DNase I-containing cytoskeletal buffer as described in the previous section, incubating at 4 "C for 2 h unless otherwise noted. Triton-soluble and -insoluble fractions were then separated as described above. Control extractions were carried out with cytoskeleton buffer containing carbonic anhydrase or no additions.
Measurement of 0; Productwn-To measure 0; production, the fractions were suspended in 0.25 M sucrose at 4 'C, and the activity was measured as superoxide dismutase-inhibitable reduction of ferricytochrome c using NADPH as substrate (19,25). For each assay, the reaction rate shown is the maximum rate of cytochrome c reduction observed during the course of the incubation.
Electrophoresis and Zmmurwblotting-Neutrophil membranes were extracted with cytoskeleton buffer as described above. Proteins in the Triton-soluble fractions were precipitated by treatment with methanol/chloroform/H20 as described elsewhere (27). The Triton-insoluble and precipitated Triton-soluble proteins were subjected to SDSpolyacrylamide gel electrophoresis on 12.5% polyacrylamide gels using the Laemmli buffer system (28). For two-dimensional gel electrophoresis, the cytoskeleton buffer used for the extraction was supplemented with phosphatase inhibitors (5 mM EDTA, 5 mM NaF, and 1 mM Na3V04). The first dimension of the two-dimensional procedure consisted of nonequilibrium pH gradient electrophoresis as described by O'Farrell (29), except that before mixing the sample proteins with urea/Nonidet P-40 buffer, they were sonicated in 215 pl of 20 mM Tris. HCl and then dissolved by heating to 100 "C for 3 min after adding 25 pl of SDS solution and 10 pl of aqueous mercaptoethanol to achieve final concentrations of 2% (w/v) and 4% (v/v), respectively. Ampholines consisted of 1% pH 3.5-10 ampholines and 1% pH 7-9 ampholines. The separated proteins were electrophoretically transferred onto a nitrocellulose sheet (30), which was blocked as described (31), probed with ammonium sulfate-purified rabbit antibodies raised against synthetic peptides from ~47~'"'" and ~67~'"'' (22) or with the anti-Rac2 antibody, and finally detected with alkaline phosphatase-labeled goat anti-rabbit immunoglobulin antibodies and the BCIP/NBT detection kit (Bio-Rad) or with horseradish peroxidase-labeled goat anti-rabbit immunoglobulin antibodies and the ECL detection kit (Amersham Corp.). The antibodies were used at the following dilutions: anti-~47~'"'', 1:4000-1:10,000; anti-p67P'"'x, 1:400-1:lOOO; anti-Rac2, 1:200-1:1000; anti-rabbit immunoglobulin, 1:2000. Quantification of proteins on immunoblots was accomplished by scanning with an UltroScan XL laser densitometer.

Association of p 4 7 p h o x , p67phbx, and Rac2 with Membrane
Cytoskeleton-Treatment with Triton X-100 separates neutrophil membranes into two fractions: a Triton-insoluble fraction, generally equated with the membrane cytoskeleton, and a Triton-soluble fraction, containing the non-cytoskeleton membrane proteins (21,32). Analysis of cytoskeletal and noncytoskeletal fractions by immunoblotting ( Fig. 1, top panel) showed that cytoskeleton from resting membranes contained no detectable ~4 7~~' or ~6 7~~' .
Stimulation of neutrophils with PMA, however, caused both proteins to move to the membrane. All of the ~6 7~~' that had migrated to the membrane was associated with the cytoskeleton (Fig. 1, bottom  panel); the proportion of membrane-bound ~4 7~~~ associated with the cytoskeleton, however, was relatively small, most appearing in the non-cytoskeletal fraction.
The presence of ~4 7~~" and ~6 7~~' in the Triton-insoluble fraction could reflect either their physiological association with the cytoskeleton or their insolubility in cytoskeleton buffer. To distinguish between these possibilities, the cytoskeleton was dissolved with DNase I (32,33). DNase I will solubilize proteins associated with the cytoskeleton but should not affect proteins that are intrinsically insoluble in cytoskeleton buffer. Fig. 1 (bottom panel) shows that ~4 7~~' and p67p""' were solubilized by DNase I (DNase) but not by carbonic anhydrase (C-Anh), a protein similar in size to DNase I, or by protein-free buffer (-). This result suggests that in activated neutrophils, both p47P""" and ~6 7~~" are associated with the membrane cytoskeleton.
An activated GTP-binding protein is required for 0; pro- Resting ,u to membrane cytoskeleton during neutrophil activation. Neutrophil cytosol and membranes were prepared as described under "Experimental Procedures." Triton-soluble (Sol) and -insoluble (Skl) fractions were prepared from the membranes using cytoskeleton buffer alone (-) or with carbonic anhydrase (C-Anh) or DNase I (DNase). Proteins were separated by SDS-polyacrylamide gel electrophoresis through 10% gels, transferred to nitrocellulose by electroblotting, and detected with antibodies as described under "Experimental Procedures." The cytosol track contains lo6 cell eq of cytosol; the remaining tracks each duction by the respiratory burst oxidase (6,7,11,34,35). In human neutrophils, Rac2 has been shown to fulfill this requirement (6,18). Consistent with its postulated role in oxidase activation, Rac2* migrated from the cytosol to the plasma membrane when neutrophils were stimulated with PMA ( Fig. 2, lanes 1-3). The release of Rac2 by treatment with DNase I confirmed that in stimulated neutrophils this protein, like ~4 7~~' and p67phox, is associated with the membrane cytoskeleton (Fig. 2, lanes 4-7). An earlier study with ~4 7~~~-d e f i c i e n t neutrophils had * The low molecular weight GTP-binding protein Racl has been identified as an oxidase-activating protein in guinea pig macrophages (7,36). We were unable to detect Racl in human neutrophil cytosols (not shown).
shown that p47p"'" is required for the transfer of ~67~'"" from the cytosol to the plasma membrane during oxidase activation (37). We have found that ~4 7~~' is also needed for the transfer of Rac2 to the plasma membrane. When normal neutrophils were activated with PMA, ~4 7~~" and Rac2 migrated as expected from cytosol to the plasma membrane (Fig. 3, left). When ~47~"O~-deficient neutrophils were similarly treated, however, Rac2 remained in the cytosol (Fig. 3, right). Kinetics and Extent of Transfer of Cytosolic Proteins to Membrane Cytoskeleton during Oxidase Activation-We compared the kinetics of activation of the respiratory burst (15) with the rate of transfer of ~47~'"'", p67p""x, and Rac2 from cytosol to the cytoskeleton in PMA-stimulated neutrophils. Fig. 4 shows that the transfer of all three proteins was well under way by 2 min and was virtually complete by 5 min. A good temporal correlation, therefore, exists between protein transfer and the activation of 0; production in PMA-treated neutrophils.
At the concentration of PMA employed in these studies, the neutrophil respiratory burst is fully activated (15). Table  I shows, however, that 4 5 % of the cytosolic oxidase components had migrated to the cytoskeleton. The role of the components remaining in the cytosol during oxidase activation is not clear. They may provide a reserve to maintain 0; production over a long period as previously activated oxidase molecules lose their function or to reactivate oxidase molecules that have been reversibly turned off.
Association of Respiratory Burst Oxidase Activity with Membrane Cytoskeleton-When incubated with NADPH, plasma membranes from PMA-treated neutrophils produce 0; via the respiratory burst oxidase. Table I1 shows the distribution of oxidase activity between the cytoskeleton and membrane. Confirming earlier results (21), we found that the membrane cytoskeleton contained >90% of the 0;-forming activity of whole membranes from PMA-activated neutrophils.
The association of p47p"Ox, p67Ph"', and RacP with the membrane cytoskeleton may help establish a productive interaction between these subunits and other oxidase components (e.g. cytochrome b5m). It was, therefore, of interest to examine the relationship between oxidase activity and cytoskeletal integrity. Dissolution of activated neutrophil cytoskeletons with DNase I resulted in a loss of oxidase activity (Table 11) that was nearly complete after 2 h (Fig. 5) with NADPH, indicating that DNase I does not interfere with the catalytic activity of the oxidase. The effects of DNase I appeared to be related to its F-actin depolymerizing ability since competition with G-actin greatly decreased the DNase I-induced losses of p47P"Ox, ~67~"O", and RacP (Fig. 6), as well as oxidase activity (Table 111), from the membrane cytoskeleton.
The foregoing results suggest that cytoskeletal integrity is important for oxidase activity and that cytoskeletal ~47~"O*, ~67~"O', and Rac2 are part of the active oxidase, whereas the non-cytoskeletal portion of p47P""" is not. These findings are compatible with the idea that one or both of those proteins and non-cytoskeletal membrane proteins were prepared and assayed for 0; production as described under "Experimental Procedures." Each assay mixture contained 5 X lo6 cell eq of cellular fraction. DNase I and carbonic anhydrase were used a t 2 mg/ml. Extracted membranes were incubated with DNase I, carbonic anhydrase, or cytoskeleton buffer alone for 2 h before separating the Triton-soluble and Triton-insoluble fractions. Intact membranes were incubated with DNase I or carbonic anhydrase for 5 min before assay. The results are presented as the mean f S.D. of four exDeriments. a Designated as "Triton-insoluble" rather than "cytoskeleton" because the cytoskeletons had been solubilized by the DNase I treatment. associates directly with an actin binding protein or with actin itself.

Phosphorylated Forms of p 4 7 " in Membranes and Membrane Cytoskeletons from Resting and Activated Neutrophils-
The activation of the respiratory burst oxidase in intact neutrophils is associated with the phosphorylation of 6-8 serine residues in p47p'"" (14-16). Both partly and fully phosphorylated forms of ~47~'"'' are routinely observed in activated neutrophils, the variously phosphorylated forms appearing as  Table 111. After incubation, the mixtures were centrifuged for 2 h at 180,000 X g, and 5 X 10' cell eq of membranes were subjected to electrophoresis and immunoblotting as described under "Experiment a l Procedures." The results shown are representative of four separate experiments. Rates of 0; production in the control (100%) incubations were 2.1 f 0.3 nmol/min/5 X lo6 cell eq. C-anh, carbonic anhydrase.

TABLE 111
Actin prevents DNase I from destroying the 0;-generating activity of activated neutrophil CytoskeZetonS DNase I (2 mg/ml in cytoskeletal buffer) was incubated for 15 min a t 4°C with actin or carbonic anhydrase (each a t 4 mg/ml) or with buffer only and then added to activated neutrophil membranes and incubated for 2 h a t 4 "C. Control samples were not treated with DNase I. After incubation, the mixtures were centrifuged for 2 h at 180,000 X g, and 5 X lo6 cell eq of each supernatant and pellet was assayed for 0; production. The results are presented as the mean f  (3) ~4 7~~' and its attachment to the cytoskeleton, cytoskeletal and non-cytoskeletal fractions from neutrophil membranes were subjected to two-dimensional gel electrophoresis, detecting ~4 7~~" by immunoblotting with an antibody against its Cterminal decapeptide.
To verify that the antibody recognized all the phosphorylated forms of p47ph"', neutrophils were loaded with 32P, stimulated with PMA, and analyzed by two-dimensional gel electrophoresis, visualizing the p47P"' isoforms by both autoradiography and immunoblotting. The immunoblot showed that activated neutrophils contained multiple ~4 7~~' forms spanning a PI range of =4 to >8 (Fig. 7, top band). Every 32Plabeled ~4 7~~' isoform (Fig. 7, bottom band) was detected by the antibody. The antibody also detected some ~4 7~~' isoforms that were not labeled. Those isoforms presumably contained no 32P, either because they were not phosphorylated at all or because they contained phosphate that did not turn over during the course of labeling. Alternatively, they could  FIG. 7. Isoforms of ~4 7~"~ from "P-loaded neutrophils as detected autoradiographically and by immunoblotting after electrophoretic transfer of proteins from two-dimensional gels onto nitrocellulose. Activated neutrophils that had been loaded with 32Pi (1.25 X lo7 cells in each sample) were subjected to two-dimensional gel electrophoresis. The separated proteins were transferred electrophoretically to a nitrocellulose sheet, and the p47Pho' isoforms on the nitrocellulose sheet were visualized, first by immunoblotting (top band) and then by autoradiography (bottom band), as described under "Experimental Procedures." The locations of the PI markers are indicated above the immunoblot. cell eq) were prepared from activated neutrophil membranes and analyzed by two-dimensional gel electrophoresis followed by immunoblotting as described under "Experimental Procedures." The results shown are representative of three separate experiments.
have contained 32P in quantities too small to detect. In any case, phosphorylation did not prevent the antibody from recognizing p47P""'.
Using this antibody, we examined the state of phosphorylation of ~4 7~" " ' molecules in the cytoskeletal and non-cytoskeletal fractions from activated neutrophil membranes. The results (Fig. 8) showed that the locations of t h e ~4 7~" " ' isoforms varied with their levels of phosphorylation. Acidic isoforms (i.e. the more completely phosphorylated isoforms) were principally found in the membrane cytoskeleton (those with the most acidic PI values residing exclusively in the cytoskeleton where the 0;-forming activity is located). In contrast, the most basic isoforms were found only in the noncytoskeletal membrane fraction, while isoforms with intermediate PI values were located in both fractions. These data suggest that the expression of respiratory burst oxidase activity in intact neutrophils requires the relatively complete phosphorylation of ~47~""'.

DISCUSSION
Our experiments showed that when neutrophils are activated by PMA, the oxidase components ~4 7~" " " and ~6 7~" " ' and the GTP-binding protein Rac2 migrate from the cytosol to the membrane cytoskeleton. Membrane-bound ~6 7~" " " and Rac2 are found exclusively in the membrane cytoskeleton, but membrane-bound ~4 7~" " ' is distributed between the membrane cytoskeleton and the non-cytoskeletal fraction of the plasma membrane. When neutrophils deficient in ~4 7~" " " were activated with PMA, neither ~6 7~" " " nor Rac2 was transferred to the membrane; this finding indicates that in whole neutrophils, Rac2 behaves like a dedicated component of the respiratory burst oxidase. Antisense experiments suggested that Rac2 serves a similar function in B lymphocytes (38).
Data are available that pertain to specific interactions between individual components of the respiratory burst oxidase. Mutual interactions among ~47~""', ~67~""', and cytochrome bss8 are indicated by the findings that in cytochrome bsw-deficient neutrophils neither ~4 7~" " " nor ~6 7~" " ' migrates from the cytosol to the plasma membrane when the cells are activated and that ~6 7~" " ' does not translocate when ~4 7~" " "deficient neutrophils are activated (37). Kinetic studies have suggested that the GTP-binding protein of the respiratory burst oxidase interacts only with p67p""x (39). Our results show that ~4 7~" " " is required for the transfer of RacP from the cytosol to the cytoskeleton when neutrophils are activated with PMA but do not indicate whether Rac2 interacts with p47P"OX, ~67~""', or both.
In the activated neutrophil plasma membrane, the most fully phosphorylated forms of ~4 7~" " ' are restricted to the membrane cytoskeleton, while the less phosphorylated forms are located only in the non-cytoskeletal region of the membrane. These results suggest that the transfer of ~4 7~" " ' from cytosol to plasma membrane occurs when a certain level of phosphorylation is achieved but that association of ~4 7~" " ' with the membrane cytoskeleton requires further phosphorylation. These conclusions support earlier evidence for a multistage phosphorylation of ~4 7~" " ' during oxidase activation A prior study suggested that in the cytosol from resting neutrophils, all of the ~6 7~" " " ' was present in a complex of M , = 240,000 that also contained some, but not all, of the ~4 7~" " ' (19). The present study showed that in activated neutrophils, all of the membrane-bound p67p""x, but only some of the p47Pho', was associated with the cytoskeleton. It may be that only the 240,000 complex, which contains both ~4 7~" ' " and p67phox, is translocated to the membrane cytoskeleton where the active oxidase is located and that free ~47~""', though susceptible to phosphorylation and transferable to the plasma membrane, is not incorporated into the active oxidase. Measurements showing three times as much ~4 7~" " ' as ~6 7~" " ' in neutrophil cytosol (43) are consistent with this interpretation.
Previous work from this laboratory suggested t h a t ~6 7~" " ' was attached to the cytoskeleton in both resting and activated neutrophils, while ~4 7~" " ' was never associated with the cytoskeleton (22). With regard to ~67~""", the present findings are consistent with the earlier observations because the earlier study dealt with the entire cellular cytoskeleton, while the present study is concerned only with the membrane cytoskeleton. The failure to detect cytoskeletal ~4 7~" " ' in the earlier study was probably due to the very limited amount of ~4 7~" " ' that is transferred to the cytoskeleton during oxidase activation and to the blending of the 32P signals from the most acidic isoforms of ~4 7~" "~ into the background of the autoradiogram. Our present results agree with those of Clark and co-workers (23) who reported that some ~4 7~" " ' is transferred to the neutrophil cytoskeleton during oxidase activation.
All of the 0;-generating activity of activated neutrophil membranes was found in the membrane cytoskeleton. This activity was eliminated, and ~4 7~" " " , p67phox, and RacP were released by dissolving the cytoskeleton with DNase I. These results suggest that the activity of the respiratory burst oxidase depends on the integrity of the membrane cytoskeleton and indicate that the cofractionation of ~4 7~" " " , ~6 7~" " ' , and Rac2 with the cytoskeleton was due to an association between these proteins and the cytoskeleton rather than to the insolubility of these proteins in the Triton-containing cytoskeleton buffer.

Respiratory Burst Oxidase
effect on the interaction between the respiratory burst oxidase and the cytoskeleton (21, 22). Cytochalasins depolymerize some of the F-actin in neutrophils but spare the cortical Factin that lies directly beneath the plasma membrane (44). Those earlier observations are, therefore, compatible with our finding that the respiratory burst oxidase is associated with the membrane cytoskeleton.