Protein prenylation restrains innate immunity by inhibiting Rac1 effector interactions

Rho family proteins are prenylated by geranylgeranyltransferase type I (GGTase-I), which normally target proteins to membranes for GTP-loading. However, conditional deletion of GGTase-I in mouse macrophages increases GTP-loading of Rho proteins, leading to enhanced inflammatory responses and severe rheumatoid arthritis. Here we show that heterozygous deletion of the Rho family gene Rac1, but not Rhoa and Cdc42, reverses inflammation and arthritis in GGTase-I-deficient mice. Non-prenylated Rac1 has a high affinity for the adaptor protein Ras GTPase-activating-like protein 1 (Iqgap1), which facilitates both GTP exchange and ubiquitination-mediated degradation of Rac1. Consistently, inactivating Iqgap1 normalizes Rac1 GTP-loading, and reduces inflammation and arthritis in GGTase-I-deficient mice, as well as prevents statins from increasing Rac1 GTP-loading and cytokine production in macrophages. We conclude that blocking prenylation stimulates Rac1 effector interactions and unleashes proinflammatory signaling. Our results thus suggest that prenylation normally restrains innate immune responses by preventing Rac1 effector interactions.

P rotein geranylgeranyltransferase type I (GGTase-I) transfers a 20-carbon geranylgeranyl lipid to a cysteine residue of proteins harboring a carboxyl-terminal CAAX motif, including the Rho family proteins Rac1, RhoA, and Cdc42 1 . Geranylgeranylation, also called prenylation, enhances hydrophobicity and facilitates membrane anchoring of Rho proteins and is believed to be essential for correct subcellular targeting, effector binding, GTP loading, and activation 2,3 .
Geranylgeranylation is an evolutionarily conserved modification that has generated a broad interest for several reasons. First, Rho family proteins contribute to tumor growth and metastasis, which prompted the development of GGTase-I inhibitors (GGTIs) 4 . Several GGTIs exhibit anti-tumor effects in preclinical studies and the rationale for using GGTIs in cancer therapy is supported by mouse gene-targeting experiments 5,6 . Second, Rho proteins regulate phagocytosis, migration, reactive oxygen species (ROS) production, and signaling in inflammatory cells 7 . Thus, targeting GGTase-I has been proposed as a strategy to treat inflammatory and autoimmune disorders, including rheumatoid arthritis and multiple sclerosis [8][9][10] . And third, reduced geranylgeranylation of Rho proteins is frequently suggested to underlie anti-inflammatory properties and other pleiotropic effects of statins 11,12 . Statins lower cholesterol levels by blocking mevalonate synthesis but this also leads to reduced synthesis of geranylgeranylpyrophosphate (GGPP), the lipid substrate of GGTase-I 13 .
The idea that blocking Rho protein geranylgeranylation would inhibit inflammation was challenged by studies into GGTase-Ideficient mice 14,15 . Knockout of GGTase-I's catalytic subunit in macrophages eliminated Rho protein geranylgeranylation, but surprisingly, Rac1, RhoA, and Cdc42 accumulated in their GTPbound active form, and Rac1 remained associated with membranes 14 . Moreover, p38 and NFκB activities were high in GGTase-I-deficient macrophages, which increased proinflammatory cytokine production after lipopolysaccharide (LPS) stimulation; and the mice developed chronic erosive rheumatoid arthritis 14 . Thus, targeting macrophage GGTase-I activates Rho proteins and causes inflammation.
These results dispute the general understanding of the biochemical and medical importance of CAAX protein geranylgeranylation and raise a range of new questions 16 . For example, GGTase-I has more than 60 predicted substrates 17 , and it is not yet known whether one or more of these proteins mediate inflammatory signaling and erosive arthritis development in mice lacking GGTase-I in macrophages. It would also be important to define mechanisms underlying the increased GTP loading of nonprenylated Rho proteins. One potential explanation is that they interact more avidly than prenylated Rho proteins with a guanine-nucleotide exchange factor (GEF) or an adaptor protein such as Ras GTPase-activating-like protein (Iqgap1) [18][19][20] . Alternatively, nonprenylated Rho proteins interact less avidly with a GTPase-activating protein (GAP) 18 or guanine nucleotide dissociation inhibitor (RhoGDI1) 21 . Interestingly, knockdown of Rhogdi1 increases GTP loading of Rho family proteins 22 , and it is possible that a reduced interaction with RhoGDI1 underlies Rhoprotein activation and inflammation in GGTase-I-deficient macrophages. Moreover, it would be important to establish whether statins, by reducing GGPP levels, produce similar effects in macrophages as the knockout of GGTase-I. Statin administration is frequently associated with increased Rho-GTP loading and sometimes with increased LPS-induced cytokine production, including interleukin (Il) 1β, but the underlying mechanism is unknown 23 . In the current study, we used genetic, pharmacologic, and proteomic strategies to address those issues and identify a new potential explanation for the role of prenylation for Rac1 effector interactions and proinflammatory signaling.
Rac1-GTP may activate inflammatory signaling pathways through multiple effectors including ROS and p38. Pharmacological inhibition of ROS-reduced phosphorylation of Stat3, Src, and Ikk in Pggt1b Δ/Δ macrophages, but did not influence p38 phosphorylation (Supplementary Fig. 2E and F). Moreover, a p38 inhibitor and a ROS inhibitor prevented the increased cytokine production observed in LPS-stimulated Pggt1b Δ/Δ macrophages ( Fig. 1g and Supplementary Fig. 2G and H).
GGTase-I knockout increases Rac1-GTP, but reduces total Rac1. Nonprenylated Rac1, RhoA, and Cdc42 in Pggt1b Δ/Δ macrophages accumulate in their GTP-bound active forms 14 ( Fig. 2a and Supplementary Fig. 3A and B); gene-expression levels were unaffected (Fig. 2b). Further analyses revealed that the total levels of Rac1 were reduced by~50% (Fig. 2a). Moreover, the half-life of Rac1 was 65-70% shorter in Pggt1b Δ/Δ than in control cells (Fig. 2c, d). To determine whether ubiquitin-mediated proteasomal degradation contributed to the reduced Rac1 levels, we first immunoprecipitated ubiquitin and performed western blots for Rac1; and found higher amounts of ubiquitin-associated Rac1 in Pggt1b Δ/Δ than in control lysates (Fig. 2e). Although we   Fig. 1 Rac1 haploinsufficiency rescues arthritis and inflammatory signaling in Pggt1b Δ/Δ mice. a Left, Western blots showing steady-state levels of GTPbound and total Rac1 in BM macrophages isolated from Pggt1b Δ/+ , Pggt1b Δ/Δ , and littermate Rac1 Δ/+ Pggt1b Δ/Δ mice. Actin was used as a loading control. Right, Bar graphs showing mean Rac1-GTP levels determined by densitometry (n = 2 per genotype). b Synovitis and erosion score in joints of 12-week-old Pggt1b +/+ (n = 4), Pggt1b Δ/Δ (n = 12), and Rac1 Δ/+ Pggt1b Δ/Δ (n = 9) mice. c Cytokine concentrations, 8 h after LPS (10 ng/ml) stimulation, in medium of primary bone marrow (BM) macrophages isolated from Pggt1b +/+ (n = 3), Pggt1b Δ/Δ (n = 4), and Rac1 Δ/+ Pggt1b Δ/Δ (n = 3) mice. d Western blots showing levels of mature Il-1β and caspase-1 in supernatants (Sup), and pro-Il-1β and pro-caspase-1 in lysates (Lys) of LPS (200 ng/ml) stimulated BM macrophages; tubulin in lysates was used as a loading control. The antibiotic nigericin (28 mM) was used as a positive control for inflammasome-mediated caspase-1 activation and Il-1β production. e Western blot showing levels of Mmp13 in medium of LPS-stimulated BM macrophages; Actin in lysates was used as a loading control. f Western blots showing phosphorylated (p) and total levels of intracellular signaling mediators in lysates of BM macrophages isolated 0, 15, and 30 min after LPS stimulation. g Concentration of Il-1β in medium of LPS-stimulated BM macrophages (n = 3/genotype) that had been pre-incubated for 1 h with inhibitors of p38 (SB203580; 1 and 5 µM) and ROS (DPI; 500 nM and 5 µM). For c-e, g, similar results were observed in two to three independent experiments. Error bars presented as s.e.m. when n is equal to or more than three. Significance between groups were calculated with two tailed Student's t test (c, g) and one-way ANOVA with Tukey's post hoc test (b). n.s. not significant, *P < 0.05, **P < 0.01, ***P < 0.001 only detected Rac1 conjugated with three Ubiquitins under the current experimental conditions, we can't rule out the existence of mono-, di-, and polyubiquitinated forms. Second, we incubated macrophages with proteasome inhibitors and found that total Rac1 levels were restored (Fig. 2f); proteasome inhibition also increased Rac1-GTP levels ( Supplementary Fig. 3C). In contrast with Rac1, total RhoA, and Cdc42 levels were increased in Pggt1b Δ/Δ cells (Supplementary Fig. 3A and B). Thus, blocking prenylation reduces Rac1 stability but appears to increase that of RhoA and Cdc42.    RhoGDI1 is not involved in phenotypes of GGTase-I deficiency. The GGTase-I knockout reduces interactions between RhoGDI1 and Rac1 and Cdc42, but not RhoA 14 . We hypothesized that a reduced interaction between RhoGDI1 and Rac1 might underlie increased GTP loading and cytokine production and tested this by inactivating RhoGDI1 expression. Consistent with a previous study 22 , suppressing RhoGDI1 expression with small-interfering (si) RNAs increased Rac1 GTP loading and reduced total Rac1 levels-a result that was associated with increased LPS-stimulated Il-6 and Tnf production (Supplementary Fig. 4A). However, we only observed this result in the RAW264.7 macrophage cell line; it was neither observed in primary macrophages incubated with siRNAs, nor in immortalized macrophages where the Arhgdia1 gene had been inactivated with CRISPR/CAS9 (Supplementary Fig. 4B and C). Moreover, RhoGDI1 inactivation never increased Il-1β production, including in RAW264.7 cells (Supplementary Fig. 4A-C).

Iqgap1 binds nonprenylated Rac1 and mediates arthritis.
To determine whether blocking Rac1 prenylation influences other effector interactions, we immunoprecipitated Rac1 from Pggt1b Δ/ Δ and control macrophage lysates and performed isobaric tagging for relative peptide quantification with mass spectrometry 30 . From a list of 717 proteins whose levels differed significantly in Pggt1b Δ/Δ and control cells, we identified five known Rho family effector proteins ( Table 1). The top hit was GTPase-activatinglike protein 1 (Iqgap1)-an adaptor protein that binds and stabilizes GTP-bound Rho family proteins but lacks GAP and GEF activity 20,31,32 . Immunoprecipitation (IP) and western blot analyses of macrophage lysates revealed that the Rac1-Iqgap1 interaction was two to threefold higher in Pggt1b Δ/Δ than control cells (Fig. 3a). The interaction between Iqgap1 and RhoA and Cdc42 in Pggt1b Δ/Δ macrophages was also increased (Supplementary Fig. 5A and B).
Tiam1 binds nonprenylated Rac1 and stimulates GTP loading. Iqgap1 has no GEF or GAP activity 20,31 . Thus, we asked whether an increased interaction with a known GEF-or a reduced interaction with RacGAP1-could explain the increased GTP loading of nonprenylated Rac1. IP-western blot analyses revealed a two to sixfold higher Rac1-Tiam1 interaction in Pggt1b Δ/Δ than in control macrophages (Fig. 4a). Interactions between Rac1 and the GEFs Vav1, Vav2, β-Pix, and Dock1, and the GAP RacGAP1 were similar in Pggt1b Δ/Δ and control macrophages ( Supplementary Fig. 6A). Inhibiting Tiam1 expression with siRNAs reduced GTP-bound and increased total Rac1 (Fig. 4b). Tiam1 inhibition also reduced cytokine production of LPS-stimulated Pggt1b Δ/Δ macrophages ( Fig. 4c and Supplementary Fig. 6B and C).
To determine whether Tiam1 interacts with Iqgap1, we performed IP-western blot analyses and found a low basal interaction in control Pggt1b +/+ macrophages and a high degree of interaction in Pggt1b Δ/Δ macrophages (Fig. 4d). These experiments also revealed consistently higher Tiam1 levels in Fig. 2 GGTase-I knockout increases Rac1-GTP loading, but reduces Rac1 total levels. a Left, western blots showing steady-state levels of GTP-bound and total Rac1 in BM macrophages. Nonprenylated RAP1A was used as marker of GGTase-I-deficient cells; actin was used as a loading control. Middle and right, amount of GTP-bound (n = 5/genotype) and total Rac1 (n = 17/genotype) in BM macrophages determined by densitometry of protein bands. b Quantitative polymerase chain reaction (QPCR) data showing levels of Rac1, Rhoa, and Cdc42 expression in cDNA of BM macrophages (n = 3 per genotype). c Left, western blots showing levels of Rac1 and Actin that remain in BM macrophages at various time points after incubation with cycloheximide (20 µg/ml) to stop protein synthesis. Equal amounts of total proteins were loaded. Right, densitometry of protein bands normalized to timepoint 0 within each genotype. d Similar experiment as in (c) except twice the amount of total proteins from the Pggt1b Δ/Δ lysates were loaded compared to Pggt1b Δ/+ to obtain similar Rac1 levels at time-point 0. e Left, immunoprecipitation (IP) of Ubiquitin (Ub) followed by western blots for Rac1. Direct western blots were performed on the same lysates (input) to quantify total levels of Rac1 and Actin. The molecular weight of the main band was~45 kDa which corresponds to Rac1 conjugated with three Ubs. Two independent experiments are shown. Right, amount of ubiquitin-bound Rac1 determined by densitometry of protein bands (n = 4/genotype). f Upper panel, western blots showing total Rac1 levels in lysates of BM macrophages incubated for 10 h with proteasome inhibitors MG-132 (15 µM) and lactacystin (15 µM). Lower panel, quantification of protein bands normalized to Actin and expressed as percent of control (Pggt1b Δ/+ ). Similar results were obtained three times. Error bars represent s.e.m. Significance between groups were calculated with two tailed Student's t test. *P < 0.05, **P < 0.01, ***P < 0.001  Synovitis score   Iqgap1 binds nonprenylated Rac1 and mediates GTP loading, cytokine production, and arthritis. a Left, immunoprecipitation (IP) of Rac1 in BM macrophage lysates followed by western blots for Iqgap1. Direct western blots were performed on the same lysates (Input) to quantify total Iqgap1 and Rac1 levels. Nonprenylated Rap1A was used as marker of GGTase-I-deficient cells; actin was used as a loading control. Middle and right, levels of Rac1bound Iqgap1 was determined by densitometry of IP-western blots of BM macrophages from six mice/genotype. b Synovitis and erosion scores in joints of 12-week-old Pggt1b Δ/+ (n = 4), Pggt1b Δ/Δ (n = 9), and littermate Iqgap1 -/-Pggt1b Δ/Δ (n = 7) mice. c Cytokine concentrations, 8 h after LPS stimulation, in medium of BM macrophages isolated from Pggt1b Δ/+ , Pggt1b Δ/Δ , and Iqgap1 -/-Pggt1b Δ/Δ mice (n = 3/genotype). d Left panel, Western blots showing levels of GTP-bound Rac1, total Rac1, and Iqgap1 in BM macrophages isolated from Pggt1b Δ/+ , Pggt1b Δ/Δ , and Iqgap1 −/− Pggt1b Δ/Δ mice. Right panels, levels of GTP-bound and total Rac1 determined by densitometry (n = 5/genotype). e Western blots of macrophage lysates isolated at base-line and 15 and 30 min after LPS stimulation. f Confocal microscopy images showing F-Actin staining of BM macrophages. Scale bars: 10 μm. Error bars represent s.e.m. The graph shows adhesive area of 20-30 macrophages in ten randomly selected fields analyzed in cells from two mice/genotype. Significance between groups were calculated with two tailed Student's t test (a, c, d, f) and one-way ANOVA with Tukey's post hoc test (b). n.s. not significant, *P < 0.05, **P < 0.01, ***P < 0.001 Pggt1b Δ/Δ than in control macrophages (Fig. 4a, b, d); the Iqgap1 knockout normalized Tiam1 levels (Fig. 4d).
Editing the Rac1 CAAX sequence increases GTP loading. Levels of GTP-bound and total Rac1 in GGTase-I-deficient cells could conceivably be influenced by the accumulation of other nonprenylated CAAX-proteins. To define biochemical consequences of blocking Rac1 prenylation in cells expressing normal GGTase-I activity, we edited the CAAX sequence of endogenous Rac1 in human embryonic kidney (HEK) cells with a nonintegrating homologous recombination-based CRISPR/CAS9 approach. We isolated three CAAX-mutant clones (Rac1CM1-3), sequenced their DNA and cDNA (Fig. 5a), and found in their lysates that Rac1 exhibited a reduced electrophoretic mobility indicating that the protein had not been prenylated by GGTase-I (Fig. 5b). Rac1-GTP levels were two to fourfold higher in CAAX-mutant than in control cells; and Rac1 total levels were two to threefold lower (Fig. 5c). Further analyses revealed that the distribution of Rac1 in cytosol and membrane fractions was similar in CAAX-mutant and control cells (Fig. 5d). However, nuclear Rac1 levels were lower in the edited cells (Fig. 5d). Similar to the findings with Pggt1b Δ/Δ macrophages, Rac1 total levels increased following lactacystin and MG-132 administration (Fig. 5e, f). Moreover, Rac1 ubiquitination and Iqgap1 association was higher in Rac1 mutant than control cells (Fig. 5g, h).
Statins mimic effects of GGTase-I deficiency. To determine whether statins can produce similar cellular effects as GGTase-I deficiency, we incubated three different macrophage types with Atorvastatin, Rosuvastatin, and Simvastatin. Statins increased Rac1-GTP levels and reduced total Rac1; and increased GTPbound and total RhoA (Fig. 6a, b). Moreover, statins facilitated Il-1β maturation and secretion in response to LPS, and potentiated Il-6 and Tnf production (Fig. 6c, d, and Supplementary  Fig. 7A-C). Pre-incubating the macrophages with GGPP abolished the statin effect on cytokine production in most, but not all cases (Fig. 6c, d, and Supplementary Fig. 7A and B). To determine if Iqgap1 underlies the ability of statins to increase cytokine production, we incubated primary Iqgap1-knockout macrophages with Simvastatin and found that LPS-stimulated cytokine production was either reduced or normalized (Fig. 6e). Inhibiting Iqgap1 with siRNAs produced similar results ( Supplementary  Fig. 7D).

Discussion
In this study, we identified three components of the mechanism underlying inflammatory phenotypes of mice with macrophagespecific GGTase-I deficiency. First, GGTase-I prenylates at least 60 substrates 17 , and our data indicate that one of them, Rac1, mediates the majority of the robust innate immune responses. Second, nonprenylated Rac1 becomes hyperactivated through an Pggt1b IP: RAC1 TIAM1 increased interaction with Tiam1 and Iqgap1. And third, hyperactive Rac1 activates inflammasome-, ROS-, and p38-driven signaling pathways that enhance LPS-induced Il-1β, Il-6, and Tnf production.
Prenylation is believed to have evolved as a strategy to target proteins to membranes, promote effector interactions, and thereby stimulate activation and signaling 34 . Several results indicate that this is not the case for Rac1. First, Rac1 membrane/ cytosol partitioning was unaffected in GGTase-I-deficient macrophages 14   This reasoning is particularly relevant for the control of Il-1β production. Due to its potent pleiotropic inflammatory effects, Il-1β secretion by macrophages is tightly controlled and requires two different events: a priming step through cell-surface receptors that leads to Il1b transcription; and a second signal that activates caspase-1 and leads to pro-Il-1β cleavage and secretion of the mature protein 35,36 . Our data suggest that nonprenylated Rac1 can act as the second signal in Il-1β production. This argument is supported by the finding that the robust caspase-1-mediated Il-1β production in GGTase-I-deficient macrophages was abolished by normalizing Rac1-GTP levels-which was accomplished by knocking out one copy of Rac1 or both copies of Iqgap1.
GGTase-I deficiency increased basal phosphorylation of Src, Stat3, and Ikkα/β-and to some extent p38-in a Rac1-dependent fashion; LPS-induced phosphorylation of all four proteins was also increased. Rac1-induced activation of Src, Stat3, and Ikk was mediated by ROS, whereas p38 activation was likely a direct consequence of Rac1/PAK activity. These findings are consistent with previous reports that Rac1 triggers ROS production by NADPH oxidases which activates Stat3 37,38 ; although other studies show a physical interaction between Rac1-GTP and Stat3 39 . Interestingly, Stat3 contributes to the progression of chronic inflammation and joint destruction in mouse models of rheumatoid arthritis 40,41 . Src and p38 are also involved in Ikkdependent activation of NFκB which stimulates transcription of cytokines including Il-1β, Il-6, and Tnf [42][43][44] . Importantly, knockout of one copy of Rac1 normalized Rac1-GTP levels in GGTase-I-deficient macrophages and reduced or normalized signaling of the entire pathway.
Knockout of Iqgap1 normalized Rac1-GTP loading, abolished proinflammatory signaling and cytokine production, and markedly reduced arthritis in GGTase-I-deficient mice. Although it is not known whether Iqgap1 encounters nonprenylated Rho proteins in activated wild-type macrophages, it is possible that targeting Iqgap1 might be useful in the therapy of some inflammatory conditions. One example would be patients with mevalonate kinase deficiency (MKD)-an autoinflammatory disease associated with high Il-1β production, fever episodes, lymph node enlargment, and joint pain. MKD leads to reduced synthesis of GGPP which reduces prenylation 23,[45][46][47] , and could therefore enhance cytokine production through an increased Rac1-Iqgap1 interaction. Targeting Iqgap1 would likely be associated with few side-effects as Iqgap1-deficient mice are viable and exhibit only mild phenotypes late in life 33 ; and their macrophages responded normally to LPS. Targeting Rac1 might be an alternative strategy and has been proposed earlier 23,46 ; but Rac1 deficiency is lethal in mice and produces a range of cellular and tissue phenotypes.
The knockout of Iqgap1 rescued most of the robust proinflammatory phenotypes of GGTase-I-deficient macrophages but did not influence LPS-induced cytokine production of GGTase-I wild-type macrophages-despite reducing Rac1-GTP levels. The simplest explanation for these observations is that both the levels of Rac1-GTP and the affinity of Rac1 for Iqgap1 were markedly higher in GGTase-deficient than wild-type macrophages; thus Iqgap1's role in controlling Rac1-GTP levels and its downstream signaling could simply be comparatively more important in the GGTase-I-deficient cells. Whether or not Iqgap1 influences Rac1-GTP signaling and cytokine production in arthritis and other inflammatory diseases in the setting of wild-type GGTase-I remains to be determined.
Statin administration produced similar effects as the knockout of GGTase-I: Rac1-GTP levels increased, as did LPS-induced cytokine production, in a GGPP-dependent fashion. Part of those results are in line with previous studies 23,48 , but here we also provide evidence that the proinflammatory statin effect requires Iqgap1. It, therefore, seems reasonable to speculate that statins' pro-inflammatory effects on macrophages stem from reduced Rac1 prenylation and increased interaction with Iqgap1. However, statin therapy is more often associated with antiinflammatory than pro-inflammatory effects. A potential explanation is that anti-inflammatory statin effects are the result of the drug's action on lymphocytes rather than macrophages. For example, statins inhibit T-cell proliferation and differentiation, and immune synapse formation 12 . Another speculation would be that some side-effects of statin therapy, such as myositis and rhabdomyolysis, are caused by hyperactivation of a nonprenylated Rho protein. However, these speculations should be interpreted with caution because there is little evidence that statins inhibit prenylation in vivo. Daily statin doses used by patients range from 5 to 80 mg/day, resulting in plasma concentrations of 1-15 nM 49 . The doses used in vitro in the present study (i.e., 1-5 μM) are lower than those of many other studies 50-52 , but they are likely higher than those that cells in vivo are exposed to.
Our results show clearly that prenylation is not required for GTP loading and activation of Rho proteins. Prenylation actually inhibits GTP loading. Why then are Rho proteins prenylated? One potential role of prenylation is to fine-tune the targeting of Rho proteins to specific subcellular membrane domains. Another possibility is that prenylation is required for rapid cycling between GTP and GDP-bound states during specific functions of Fig. 5 Editing the endogenous Rac1 CAAX-motif increases GTP loading and reduces total Rac1 levels. a Predicted amino acid sequence from sanger sequence results of PCR-amplified Rac1 DNA and cDNA fragments. The DNA/cDNA was from HEK-293 clones whose Rac1 gene had been edited at the CAAX sequence by CRISPR/Cas9-facilitated homologous recombination to prevent geranylgeranylation by GGTase-I. CM1-3, CAAX mutant clones; Ctr, unedited control clone. b Left, western blot showing electrophoretic mobility of wild-type Rac1 in Rac1Ctr lysates, and nonprenylated (np) Rac1 in CM1-3 lysates. c Left, western blots showing amounts of GTP-bound Rac1, total Rac1, and the loading control Actin in lysates of the gene-edited cells. Middle and right, levels of GTP-bound and total Rac1 determined by densitometry; data are mean of three independent experiments for each cell line. d Left, western blots showing the distribution of Rac1 in cytosol, membrane, and nuclear fractions (designated C, M, and N, respectively) of the gene-edited cells. Rho-GDI1, integrin-α5, and histone H3 were used as markers for cytosol, membrane, and nuclear fractions, respectively. Right, levels of nuclear Rac1 determined by densitometry data of three independent experiments and expressed as percent of Rac1Ctr. e, f Western blots showing levels of total Rac1 and the loading control Actin in lysates of gene-edited cells isolated after incubation with 15 µM lactacystin for 12 h (e); and 50 µM MG-132 at different time points (f). g Left, immunoprecipitation (IP) of Rac1 in gene-edited cells followed by western blot for Ubiquitin. Direct western blots were performed on the same lysates (Input) to quantify total Rac1 levels. Actin was used as a loading control. Right, Bar graphs showing levels of Rac1-Ub 3 . h Immunoprecipitation (IP) of Rac1 in gene-edited cells followed by western blot for Iqgap1. Direct western blots were performed on the same lysates (Input) to quantify Iqgap1 and the cell such as during proliferation or certain types of cell movement. These issues will be possible to address in the future.
But based on the current study, we conclude that a major role of GGTase-I-mediated prenylation in macrophages is to limit Rac1 activation and pro-inflammatory signaling, prevent Rac1dependent Il-1β maturation, and thereby restrain innate immune responses.

Methods
Mouse breeding and genotyping. Mice harboring conditional knockout alleles of the beta subunit of GGTase-I (Pggt1b fl/fl ) were bred with lysozyme M-Cre (LC) knock-in mice to produce offspring lacking GGTase-I in macrophages as described 6 ; these mice were designated Pggt1b Δ/Δ (Δ = delta, deleted allele). Pggt1b Δ/Δ mice were bred with mice harboring conditional knockout alleles of Rac1, Rhoa, and Cdc42, and conventional knockout alleles for Iqgap1 33,53-55 . Genotyping was performed by polymerase chain reaction (PCR) on genomic DNA from ear or tail biopsies 6 . Mice were housed in a specific pathogen-free facility monitored by routine testing of sentinel mice, and were given free access to water and chow. Mice were on a mixed genetic background (129Ola/Hsd-C57BL/6) and control mice in all experiments were gender-matched littermates. Animal experiments were approved by the research animal ethics committee in Gothenburg, Sweden.
Histology and quantification of arthritis in joints. Joints from 12-week-old mice were fixed in 4% paraformaldehyde and stored in 70% ethanol. The joints were decalcified, embedded in paraffin, sectioned, and stained with hematoxylin and eosin. The slides were analyzed in a Zeiss Axioplan 2 microscope (Carl Zeiss AG). Synovitis and erosion scores were evaluated in sections of knee, ankle, metatarsal, elbow, wrist, and metacarpal joints by an observer blinded to genotype. An arbitrary scale from 0 to 3 was used: 0-is a healthy joint, 1-hypertrophy/mild proliferation of synovia constituting more than two intimal lining cell layers, and mononuclear cell infiltration is visible; 2-hypercellularity, multiple inflammatory foci in the sub-lining layer, and proliferation of synovia; and 3-massive influx of inflammatory cells scattered throughout the synovial tissue, and growth of granulation tissue (pannus). Cartilage and bone erosion were evaluated with a separate 0-3 scale: 0 is a healthy joint; 1-reduction or uneven cartilage thickness; 2-compromised cartilage integrity and formation of erosions on the cartilage surface under the growing pannus; and 3-loss of articular surface, obliteration of the joint cavity, and loss of joint or bone shape. The cumulative arthritis index for each mouse was constructed by adding the scores from the fore and hind paws 56 .
Mass spectrometry analysis of Rac1-interacting proteins. Macrophages were lysed in buffer containing 25 mM Tris-HCl, 150 mM NaCl, 1 mM EDTA, 5% glycerin, and 1% CHAPS. Rac1 was immunoprecipitated with the 05-389 antibody (Millipore) using the Dynabeads Protein G Immunoprecipitation Kit (10007D, Thermo Fisher Scientific). The protein complex was eluted with 50 mM triethylammonium bicarbonate and 4% SDS, and digested with trypsin using the filteraided sample preparation method, as described 30 . The digested peptides were labeled with 10-plex isobaric tandem mass tag (TMT) reagents (Thermo Scientific). The labeled samples were combined into one TMT-set; purified with trifluoroacetic acid precipitation and HiPPR Detergent Removal Resin (Thermo Scientific). The purified sample was fractionated into twelve fractions using the Pierce High pH Reversed-Phase Peptide Fractionation Kit (Thermo Scientific), and the fractions were dried in a vacuum centrifuge and reconstituted in 20 μl of 3% acetonitrile, 0.1% formic acid for analysis. Peptides were then injected onto an Acclaim Pepmap 100 C18 trap column (2 cm × 100 μm, particle size 5 μm, Thermo Fischer Scientific) using an Easy-nano-LC 1000 liquid chromatography system and separated with an acetonitrile gradient on a 75-μm Reprosil-Pur C18-AQ column with a 300 nl/min flow rate. Fractions were analyzed on Orbitrap Fusion Tribrid mass spectrometer (Thermo Fisher Scientific). Precursor ion mass spectra were acquired at 120.000 resolution and mass spectrometry (MS)/MS analysis was performed in a datadependent multinotch mode where collision-induced dissociation spectra of the most intense precursor ions were recorded in ion trap at collision energy setting of 30 for 3 s. Charge states 2-7 were selected for fragmentation, dynamic exclusion was set to 30 s. MS 3 spectra for reporter ion quantitation were recorded at 60,000 resolution with higher energy collisional dissociation fragmentation at collision energy of 55 using synchronous precursor selection. In a second nano-LCMS analysis a list containing theoretical peptides from Iqgap1 and Tiam1 were included in the analysis.
The data files for the set were merged for identification and relative quantification using Proteome Discoverer version 1.4 (Thermo Fisher Scientific). The search was against the Mus musculus Swissprot Database version November 2017 (Swiss Institute of Bioinformatics, Switzerland) using Mascot 2.5 (Matrix Science) as a search engine with precursor mass tolerance of 5 ppm and fragment mass tolerance of 0.5 Da. Tryptic peptides were accepted with zero missed cleavage and variable modifications of methionine oxidation, fixed cysteine alkylation, and TMT-label modifications of N-termini and lysines were selected. The sum of the control samples was used as denominator and for calculating ratios. The detected peptide threshold in the software was set to a minimum quantification threshold value of 1000 and a 1% false discovery rate by searching against a reversed database and grouping identified proteins by shared sequences to minimize redundancy. Only peptides unique for a given protein were considered for identification; peptides common to other isoforms or proteins of the same family were excluded.
F-actin staining. Macrophages were fixed with ice-cold methanol in chamber slides and stained with Alexa Fluor 488 Phalloidin (A12379, Thermo Fischer Scientific) for 30 min. The slides were mounted with Prolong Gold Antifade Mounting reagent with DAPI (P36935, Thermo Fischer Scientific) and analyzed with confocal microscopy (LSM700, Zeiss).
Generating Rac1-CAAX mutants with CRISPR/CAS9 editing. We used a single plasmid approach to edit the CAAX sequence of Rac1. A CRISPR (cr) RNA template oligonucleotide 5′-GCTGAGACATTTACAACAGC-3′ and its complementary fragment targeting exon 6 of Rac1 were annealed and cloned into the GeneArt-CRISPR Nuclease vector (Life Technology) downstream of a human U6 promoter. A ∼1-kb DNA fragment for homologous recombination-based editing was synthesized and cloned into unique MfeI and SpeI restriction sites of the vector; the fragment was composed of a 15-18-bp sequence encoding the mutant CAAX-motif (i.e., 5′-CGAAAGAGAAAATCTTTA-3′ for C-L-L-L to S-L-L-L editing, and 5′-CGAAAGAGAAAATGC-3′ for C-L-L-L to C-L-STOP editing) and 0.5 kb flanking genomic sequences. In the resulting all-in-one gRNA-CAS9-editing plasmid the U6 promoter drives gRNA transcription; a cytomegalovirus promoter drives bicistronic expression of CAS9 and orange fluorescent protein linked by the 2A self-cleaving peptide; and the donor DNA sequence can replace the endogenous sequence by homologous recombination after CAS9 cleavage.
Statistics. Values are mean and SEM unless stated otherwise. Differences between groups were determined with Student's t test or one-way ANOVA with Tukey's post hoc test, and were considered significant when P < 0.05.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.