Effects of F/G-actin ratio and actin turn-over rate on NADPH oxidase activity in microglia

The human cDNA's coding for amino acids 4-647 of wild type LIMK1 (LIMK1-WT) and kinase dead LIMK1-D406A (LIMK1-DN), and β-actin-YFP (Clontech #6902-1; Mountain View, CA., USA) were inserted into the tetracycline-responsive lentiviral vector pLOX TW [23], and used to superinfect Ra2 045 cells expressing the tetracycline-responsive transactivator protein. For shRNA knock-down of mouse cofilin two DNA sequences containing the target sequences GGAGGACCTGGTGTTCATC (cofilin shRNA 1) and GGTGTTCAATGACATGAAG (cofilin shRNA 2) with loop sequence ACTCGAGA were synthesized, annealed, and cloned MluI/ClaI into lentivector pLVTHM-DsRed [24]. The vector was then used for superinfection of Ra2 tTR-KRAB cells expressing a tetracycline-responsive KRAB repressor protein, which in the absence of doxycycline represses transcription from polymerase I, II and III [24]. Expression of cofilin shRNA in Ra2 cells was induced 4-5 days at 50 ng/ml doxycycline before cells were used for experiments.

FMLP or PMA-induced superoxide release was measured by luminol-enhanced chemiluminescence (luminol E-CL). Ra2 cells in HBSS buffer with 62.5 μM luminol and 2 U/ml HRP II were warmed two minutes in a 37°C water bath before distribution into Wallac isoplate wells (100.000 cells/well). Subsequently, superoxide production was measured as chemiluminescence before and after stimulation with 4 μM FMLP or 100 ng/ml PMA, respectively, delivered through the injector module of a thermostated Synergy HT microplate reader. With the luminol E-CL method used here, ca. 20-25% of the luminal E-CL signal probably derives from the formation of peroxynitrite via reaction of superoxide with NO, as discussed further in [25].

For immunofluorescence Ra2 cells were fixed with 2% paraformaldehyde in phosphate-buffer pH 7.2, washed in PBS, and then incubated with polyclonal rabbit anti-LIMK1 antibodies, anti-Ser3(P)cofilin antibodies, and/or Alexa 488-conjugated phalloidin in blocking buffer consisting of PBS with 5% goat serum and 0.2% saponin (Sigma). Primary antibodies were detected with Alexa488 or Alexa568-conjugated goat-anti mouse or rabbit antibodies (Invitrogen). Images were acquired with a Zeiss LSM510 confocal microscope equipped with a C-Apochromat X63, 1.2 oil immersion objective. Confocal sections (1.0-1.5 μm) were collected and saved as 512 × 512-pixel images at 8-bit resolution before import into Adobe Photoshop for compilation. The same instrument was used for FRAP experiments. Here Ra2 cells contained on thin glass-bottomed 8-chamber slides were allowed to attach for 60 minutes in medium before incubation in HEPES-buffered salt solution at 37°C in atmospheric air for FRAP analysis. A circular ROI of 43 × 43 pixels (diameter 6 μm) was chosen corresponding to either plasma membrane or podosomal F-actin. After obtaining 2 baseline images bleaching of β-actin-YFP was performed with 60 iterations at full power of the krypton/argon laser (514 nm line), and subsequently images (depth 0.8 μm) were collected at 5 second intervals for 2-3 minutes at low laser emission. The fluorescence intensity in the photobleached region was normalized to the fluorescence intensity measured in a non-bleached region at the same post-bleaching time point to correct for bleaching during time lapse (usually less than 1-2%). Data were fitted to a fluorescence recovery curve with formula F(t) = (F₀+(F_inf * t/t_1/2))/(1 + (t/t_1/2))), where F₀ is the fluorescence intensity immediately after bleaching and F_inf is the intensity at infinite times of recovery, to obtain the half-time of recovery t_1/2[26].

Ra2 cells were collected in PBS without Ca²⁺ and Mg²⁺ and then seeded in full growth medium in tissue culture dishes for one hour at 37°C. Subsequently, cells were washed once in ice-cold PBS before lysis with actin stabilization buffer (0.1 M PIPES, pH 6.9, 30% glycerol, 5% DMSO, 1 mM MgSO₄, 1 mM EGTA, 1% TX-100, 1 mM ATP, and protease inhibitor) on ice for 10 minutes. Cells were dislodged by scraping and the entire extract centrifuged at 4°C for 75 minutes in a tabletop centrifuge at 16.000 g. The supernatant containing G-actin was recovered, and the pellet containing F-actin was solubilised with actin depolymerization buffer (0.1 M PIPES, pH 6.9, 1 mM MgSO₄, 10 mM CaCl₂, and 5 μM cytochalasin D). Aliquots of supernatant and pellet fractions were separated on 12% SDS-PAGE gels and then western blotted with monoclonal anti-β-actin antibody. Signal was detected by ECL in a digital dark room and integrated optical band density was used to estimate the cellular F/G-actin ratio.

Latrunculin causes F-actin depolymerization by sequestration of monomeric G-actin, and is known to increase the respiratory burst in neutrophils [8, 27]. In Figure 1 we compared the effect of latrunculin with that of jasplakinolide, an actin filament-stabilizing toxin, on the FMLP and PMA-induced superoxide production of the murine microglia cell line Ra2 [22], which express all the subunits of the phagocyte NADPH oxidase [28]. We used FMLP because it is a physiological and commonly used phagocyte stimulant. However, any perturbation of the actin cytoskeleton may affect FMLP receptor signaling in addition to effects on the NADPH oxidase complex, for which reason we have also included stimulation with the strong agonist PMA, which works solely on a post-receptor level by PKC activation. Both single well traces (Figure 1A,C,E, and 1G) and mean peak chemiluminescence of at least three independent experiments (Figure 1B,C,F, and 1H) are shown. Latrunculin at a concentration of 40-50 ng/ml induced a 2-3 fold increase in superoxide release measured by luminol-enhanced chemiluminescence (luminol E-CL) for both FMLP and PMA-stimulated cells (Figure 1A,B,E, and 1F). Conversely, jasplakinolide dose-dependently blocked the FMLP-induced response (ca. 90% inhibition at 2 μM), while jasplakinolide in PMA-stimulated cells exerted an enhancing effect on superoxide production (maximum of 60% increase at 1 μM), in the concentration range from ca 0.25-4 μM before becoming inhibitory at concentrations above 4 μM.

It has been proposed that cytosolic phox proteins have to be released from an actin-sequestered state for NADPH oxidase assembly [8, 16]. We therefore tested if the effects of latrunculin and jasplakinolide on NADPH oxidase activation could be overruled by over-expression of p47phox through titration of actin-related binding sites. Ra2 cells over-expressing p47phox produce five-fold more superoxide than control cells (not shown in Figure 2; see [25]). Figure 2A and 2C shows that latrunculin at concentrations, which caused PMA-stimulated Ra2 cells to produce ca. two-fold more superoxide, only had a modest enhancing effect in Ra2 cells over-expressing wild type p47phox. No further enhancement could be gained by expressing the p47phox(S303D/S304D/S320D) mutant, which mimics phosphorylation and partial activation of p47phox, and is known to induce phosphoinositide and p22phox binding [29], and a basal superoxide production in un-stimulated Ra2 cells [25]. The p47phox over-expressing cells were also partially insensitive to the effects of jasplakinolide (Figure 2B and 2D) although concentrations of more than 0.5 μM caused a delayed peak response of superoxide production.

We next analyzed the effects of over-expression of wild type and a dominant negative mutant of LIM kinase-1 (LIMK1) as an alternative means of modulating the actin cytoskeleton. Activated LIMK1 shapes the actin cytoskeleton by phosphorylating and thereby inactivating cofilin, which in its active form severs and depolymerizes actin filaments. We created Ra2 cell lines conditionally expressing wild type LIMK1 (LIMK1-WT) or the kinase-dead LIMK1-D406A mutant (LIMK1-DN) at different expression levels (80, 150, and 200 virus dose equivalents). Expression of LIMK1-WT dose-dependently increased phosphorylation of serine-3 in cofilin as expected (Figure 3A, B and 3C) and increased F/G-actin ratios in Ra2 cells (Figure 3E). Curiously, while LIMK1-DN decreased F/G-actin ratios, at all levels of expression it slightly increased the levels of phosphorylated cofilin above the levels of control cells (Figure 3B). We show in Additional File 1 that in fact introducing dominant negative mutants of any of the proteins in the important VAV1, Rac1, PAK1, LIMK1 signaling axis slightly increases p-cofilin levels above control cells, possibly because altered PAK1 or RhoA activation (which is negatively regulated by Rac1) changes the tonus of cofilin phosphatases. For comparison F/G-actin ratios in Ra2 045 control cells treated with select concentrations of latrunculin and jasplakinolide are shown in Figure 3F.

Having confirmed that LIMK1 expression regulates cofilin and F-actin in Ra2 cells, we then examined the effect on FMLP or PMA-elicited superoxide production. Expression of LIMK1-DN dose-dependently increased the FMLP-induced respiratory burst but only expression of LIMK1-WT protein at low levels (Ra2 LIMK1-WT80 cells) significantly induced a two-fold increase in superoxide release (P < 0.05, one sample t-test; Figure 4A and 4C). After PMA stimulation, Ra2 LIMK1-DN80 cells had a significantly (P < 0.05) increased NADPH oxidase activity compared to control cells, and again expression of low levels of LIMK1-WT (WT80) caused an increased superoxide production, while higher levels in WT150 and WT200 Ra2 cells significantly (P < 0.05) decreased the superoxide output with up to 40% (Figure 4D-F). Therefore, despite opposing effects of transgene on F/G-actin ratio both LIMK1-WT80 and LIMK1-DN80 Ra2 cells increased superoxide production up to two-fold following FMLP or PMA stimulation. PAK1 increases FMLP and PMA-stimulated superoxide generation in Ra2 microglia through phosphorylation (partial activation) of the cytosolic p47phox subunit [25], but we also considered that PAK1 through activation of LIMK1 [30] might additionally contribute to NADPH oxidase activity by modulating actin dynamics. However, analysis of superoxide production in a series of Ra2 cell lines co-expressing PAK1 and LIMK1 mutants indicates a parallel rather than sequential organization of these two kinases with respect to NADPH oxidase activity (see Additional File 2).

We also targeted cofilin directly using shRNA knock-down as described in materials and methods, which decreased the level of cofilin protein in Ra2 cells expressing cofilin-shRNA-1 and shRNA-2 with 75% and 55% relative to control (pLVTHM vector) transduced cells, respectively (Figure 5A and 5B). A knock-down of cofilin would be expected to increase actin polymerization, which was confirmed (Figure 5C). The F/G-actin ratio achieved in Ra2 cofilin-shRNA-1 cells was even higher than that of LIMK1-WT200 cells (10 ± 1 and 6.9 ± 0.9, respectively). The degree of knockdown of cofilin obtained with either cofilin shRNA correlated with a similarly decreased superoxide production (Figure 5, D-G). Cofilin shRNA-1 reduced the respiratory burst following FMLP stimulation with roughly 80% and following PMA stimulation with 70%.

F/G-actin ratios derived by end point measurements of F- and G-actin says little or nothing about the dynamics of the actin cytoskeleton, and we therefore considered that turn-over rates could be a better correlate of NADPH oxidase activity than F/G- actin ratios (compare Figure 3F with 4C, F; Figure 3G with 1B, D, F, H). We therefore performed FRAP analysis of β-actin-YFP-expressing Ra2 cells to obtain the actin recovery half time (recovery t_1/2), which is a measure of actin off- and on-rates combined (Figure 6). We chose two different types of regions of interest (ROI) to work with: podosomal F-actin and cortical, plasma membrane-associated F-actin (see representative examples in Figure 6A and Additional Files 7, 8, 9). FRAP measurements were performed on Ra2 045 control cells, Ra2 LIMK1-WT80/200 and -DN80/200 cells, Ra2 cofilin shRNA-1 cells, and Ra2 cells treated with concentrations of latrunculin and jasplakinolide resulting in maximal positive or negative effect on superoxide production as determined previously (see Figure 1). As seen in Figure 6C, recovery t_1/2 of actin-YFP fluorescence was depressed slightly by latrunculin or expression of LIMK1-DN at low levels (LIMK1-DN80 cells), which both increase the respiratory burst. Conversely, jasplakinolide at 1 μM, which increases the PMA-stimulated superoxide release to 150% of control levels, caused a modest increase in recovery t_1/2 in plasma membrane ROI's, while a higher concentration of 8 μM (50% reduction in superoxide release) increased recovery t_1/2 more than fourfold relative to controls. Expression of LIMK1-WT in WT80 cells distinguished itself from other conditions of enhanced superoxide production (latrunculin 100 ng/ml, LIMK1-DN80, and 1 μM jasplakinolide) by a greatly increased half-time of recovery (62 ± 18 seconds and < 20 seconds, respectively). Of note, cofilin shRNA-1 expression and jasplakinolide treatment both antagonized the formation of podosomes and caused distribution of F-actin to the dorsal plasma membrane.

Observations made during FRAP trials prompted us to perform detailed analysis of F-actin distribution by Z-sectioning using fluorophore-conjugated phalloidin to visualize F-actin. Figure 7A shows for each condition a main panel with a XY-view of the most ventral plane of collected Z-stacks, and on the sides of this panel is shown XZ (top) or YZ (side) projections of the reconstructed Z-stack. This analysis confirmed that LIMK1-WT strongly induces podosome-like structures (see movies in Additional Files 3 and 4) and that 8 μM jasplakinolide and cofilin shRNA-1, which most effectively inhibited NADPH oxidase activity, both caused a marked redistribution of F-actin associated with the ventral aspect of the cell membrane to lateral and dorsal membrane (see Additional Files 5 and 6). To allow easier comparison of the effects of different perturbations of the actin cytoskeleton on superoxide production, Figure 7B depicts the normalized values for F/G-actin ratios, actin recovery half-times and superoxide production derived from PMA-stimulated cells overlaid in the same graph (absolute values derived from experiments shown above).

In resting cells p40phox, p47phox, and p67phox are thought to be associated in a cytosolic complex [33‐35]. In neutrophils [16] and in COS cells [8] binding of p40phox to moesin exerts a restraining effect on phox protein translocation and NADPH oxidase activity. This suggests that cytosolic phox proteins may be tied to F-actin at rest, and is in line with our observation that over-expression of p47phox is sufficient to overcome effects of latrunculin or jasplakinolide on NADPH oxidase activity, likely by increasing the soluble pool of activated phox proteins available to cyt b₅₅₈. A constant exchange of cyt b₅₅₈-associated Rac1/2 and cytosolic phox proteins with new counterparts from cytosol is required to sustain NADPH oxidase activity [10, 11].

The autonomy of the translocation of cytosolic phox proteins to the membrane is such that redistribution of p40/p47/p67phox proteins does not necessarily require either cyt b₅₅₈[12] or inositol lipids [4, 36, 37], which are otherwise powerful determinants of membrane recruitment of p47phox or p40phox expressed alone [3, 7]. At least in mesenchymal cell types cortactin and moesin are important for guiding cytosolic phox proteins to the membrane [15, 18, 36]. Depending on the relative binding affinities of cytosolic phox proteins and actin-associated proteins for membrane attractors, cytosolic phox proteins may be facilitated in membrane translocation by 'piggy-backing' on coronin, moesin, or cortactin, which bind p40phox and p47phox and are themselves recruited to the plasma membrane following cell activation, incidentally, by intracellular mediators that also participate in NADPH oxidase assembly [16, 38, 39]. It is not in general known how phosphorylation of p40phox and p47phox relates to binding of actin-associated proteins, but notably, phosphorylated p47phox binds TRAF4 with increased affinity [40]. TRAF4 in turn binds the focal adhesion scaffold protein Hic5 [13]. At the membrane F-actin, actin-regulatory proteins, inositol lipids, and cyt b₅₅₈ present binding sites for newly recruited phox proteins, but it is currently unknown if there is a sequence to binding. However, the failure of p47phox to translocate to the membrane in the absence of WAVE1 [15] or moesin [36], and the premature loss of p47phox, concurrent with the normal loss of F-actin and coronin (within minutes), from newly formed phagosomes in CGD neutrophils devoid of cyt b₅₅₈[12], suggest that actin-binding proteins may serve as the initial docking station for cytosolic phox proteins before contacts with cyt b₅₅₈ are established. Once the NADPH oxidase holoenzyme is assembled and superoxide production commences, at least in the reconstituted, cell-free system, ongoing actin polymerization increases duration of NADPH oxidase activity [41]. Additionally, in macrophage-like U937 cells cofilin inactivation through LIMK1 activity or anti-sense RNA increases superoxide production following phagocytosis [42‐44]. These studies may agree with the positive effect on superoxide production of actin polymerization induced by 1 μM jasplakinolide or LIMK1-WT at low expression levels in Ra2 microglia (158 ± 8% and 135 ± 25% of control values, respectively) as well as the prolonged response of FMLP-stimulated Ra2 cells expressing LIMK1-WT (Figure 4A). However, when we specifically examined if PAK1 could contribute to NADPH oxidase activity by LIMK1-directed actin reorganization, we found no evidence of such a role for PAK1 and LIMK1 (see Additional File 2). It is not known if NADPH oxidase disassembly relates to the state of the immediately surrounding actin cytoskeleton, but in the reconstituted cell-free system actin depolymerization decreases NADPH oxidase activity [41].

As summarized in Figure 7, conditions that resulted in higher than control values of superoxide production were characterized by widely different actin recovery half times (13.0 ± 1.4 to 62.0 ± 18) and F/G-actin ratios (0.15 ± 0.03 to 10.4 ± 1.0). However, at critical thresholds defined in Ra2 microglia by the transition from LIMK1-WT80 to -WT200 expression levels (WT80/200: recovery rates 62 ± 18/68 ± 12 and F/G-actin ratios 4.1 ± 0.9/6.9 ± 0.9) and in the lower end by LIMK1-DN200 (7.3 ± 0.8/1.5 ± 0.3) superoxide production is repressed to ca. 50% of that of control Ra2 cells (16.8+-3.4/2.6+-0.6) and any further increases/decreases in actin recovery rate progressively abolishes superoxide production. The most effective block of superoxide production was under conditions of enforced actin polymerization by either cofilin knock-down (70-80% inhibition) or high concentrations (> 8 μM) of jasplakinolide (80% inhibition), which yielded actin recovery half times of 94 ± 16.9 and 88.6 ± 12.2 seconds, respectively. These conditions however also uniquely caused the redistribution of F-actin from ventral to lateral and dorsal cell surfaces, which could maybe interfere with NADPH oxidase function by disturbing any substratum directed signaling required for oxidase assembly.

Therefore we conclude that at least in microglia actin polymerization and depolymerization can both enhance NADPH oxidase activity. This apparently contradictory result is probably reconciled by the cyclic nature of NADPH oxidase assembly during superoxide production: cytosolic phox proteins and Rac1/2 bound to cyt b₅₅₈ are constantly exchanged with new subunits [10, 11], which require release from F-actin (mobilization), translocation to membrane (actin or actin-associated protein facilitated), and tethering to membrane (initial contacts with cortical F-actin or associated proteins?) as illustrated in Figure 8. During cell activation these processes may be aided preferentially by induced F-actin depolymerization. Superimposed on this subunit cycling is the mechanic influence of F-actin on the assembled and functional NADPH oxidase whose activity is increased and prolonged by active actin polymerization [41]. Therefore, increasing the concentration of an actin polymerizing or depolymerizing factor may increase NADPH oxidase activity, but only to a point, where after one or more steps in the continuous assembly/disassembly cycle of cytosolic phox proteins with cyt b₅₅₈, facilitated by the opposite actin fate, will be inhibited to a degree where it negatively affects superoxide release. Our results also imply that NADPH oxidase function in stimulated phagocytes always will be in an enhanced state of activity regardless of the very different alterations of the actin cytoskeleton required to effect different immune cell tasks such as chemotaxis, invasion, phagocytosis, redox signaling, or immunological synapse formation. Such a capacity of the respiratory burst to accommodate the plasticity of the actin cytoskeleton may be particularly relevant for microglia, which undergoe dramatic changes in morphology following cell activation going from a highly elongated and ramified morphology to a rounded, amoeboid shape.