Research ArticleCROHN’S DISEASE

Suppression of p21Rac Signaling and Increased Innate Immunity Mediate Remission in Crohn’s Disease

See allHide authors and affiliations

Science Translational Medicine  23 Apr 2014:
Vol. 6, Issue 233, pp. 233ra53
DOI: 10.1126/scitranslmed.3006763


In inflammatory bowel disease (IBD), large areas of apparently healthy mucosa lie adjacent to ulcerated intestine. Knowledge of the mechanisms that maintain remission in an otherwise inflamed intestine could provide important clues to the pathogenesis of this disease and provide rationale for clinical treatment strategies. We used kinome profiling to generate comprehensive descriptions of signal transduction pathways in inflamed and noninflamed colonic mucosa in a cohort of IBD patients, and compared the results to non-IBD controls. We observed that p21Rac1 guanosine triphosphatase (GTPase) signaling was strongly suppressed in noninflamed colonic mucosa in IBD. This suppression was due to both reduced guanine nucleotide exchange factor activity and increased intrinsic GTPase activity. Pharmacological p21Rac1 inhibition correlated with clinical improvement in IBD, and mechanistically unrelated pharmacological p21Rac1 inhibitors increased innate immune functions such as phagocytosis, bacterial killing, and interleukin-8 production in healthy controls and patients. Thus, suppression of p21Rac activity assists innate immunity in bactericidal activity and may induce remission in IBD.


Crohn’s disease (CD) and ulcerative colitis (UC) are the two major manifestations of inflammatory bowel disease (IBD) and are characterized by chronic, relapsing inflammation of the gastrointestinal tract, which remains clinically challenging (1). The precise etiology of IBD remains unclear, although it is generally accepted that the major contributory factors are genetic susceptibility, environmental triggers, and an ongoing activation of the mucosal immune system against commensal bacteria (2). In UC, the mucosal inflammation is diffuse and extends proximally from the rectum, whereas inflammation in CD is patchy and segmental with affected and macroscopically normal areas lying next to each other in close proximity (skip lesions) (3). The mechanisms responsible for local mucosal remission in CD, allowing tissue to maintain a noninflamed phenotype in an otherwise inflamed environment, remain unexplained.

Genome-wide association studies have identified more than 160 genetic susceptibility loci that underlie IBD (46). Many of the identified loci either harbor genes that are associated with the functioning of the innate immune system or involve genes that seem important for the proper secretion of antibacterial peptides by Paneth cells in the intestinal lumen (7, 8). In agreement, some CD patients are impaired with respect to innate immune responses and controlling bacterial flora in the intestine (911). Furthermore, severe genetic deficiencies in innate immunity, such as lack of nicotinamide adenine dinucleotide phosphate oxidase activity or diminished function of the Wiskott-Aldrich syndrome protein (WASP), are associated with colitis in mice and humans (12). Thus, a picture is emerging where an innate inability to clear intracellular bacteria and subsequent compensatory responses of other branches of the immune system lead to pathogenesis. Nevertheless, how reduced innate immunity in IBD is compatible with its clinical presentation in which remission and ulceration coexist in close physical proximity remains obscure at best. Obviously, an understanding of the mechanisms that sustain the noninflamed intestinal phenotype may prove useful for developing improved therapies.

This consideration prompted us to characterize and compare the kinase signatures of inflamed and noninflamed mucosa of CD patients and to compare these with kinase profiles obtained from biopsies of patients without IBD or with inflamed and noninflamed UC. We used a kinome array technology that allows the comprehensive measurement of kinase enzymatic activities present in whole biopsies toward predefined peptide substrates (1315). The profiles thus generated provide a wealth of information about the signaling pathways active in the colon of IBD patients, but especially highlight a marked suppression of p21Rac signaling in phenotypically normal mucosa of IBD patients, suggesting that active suppression of p21Rac contributes to local remission in IBD. Inhibition of p21Rac1 activity corrects innate immune cell dysfunction in CD patients and may contribute to the maintenance of regional mucosal remission through stimulation of the local innate immune responses.


Generation of kinome profiles of inflamed and noninflamed colonic mucosa

We characterized the mucosal kinase signatures associated with both the inflamed and noninflamed mucosa of the large intestine in CD and UC as well as the noninflamed mucosa of healthy controls. Paired biopsies from inflamed and noninflamed colon were obtained after informed consent (protocol no. 2004.168). (Full details of the nature of the disease and the colonic location of the biopsies obtained are listed in table S1.) Kinome profiles were generated for inflamed and noninflamed biopsies from three CD patients, six UC patients, and six controls by incubating lysates on arrays exhibiting 976 different kinase substrates in the presence of [γ-33P]adenosine triphosphate. Three technical replicates of each biological sample were performed. The arrays incorporated substantial amounts of radioactivity, and the technical quality of the profiles was good (average Pearson product ≥0.85 between the technical and biological replicates for all cell types and conditions). The results reveal the total complement of kinase activities present in inflamed and noninflamed CD, UC, and healthy controls. Comparisons of individual kinase substrates are presented in table S2.

As expected, marked differences in kinase signatures were detected when kinome profiles of inflamed and noninflamed mucosa were compared. Figures 1 and 2 show the major changes in kinase signaling pathways in inflamed lesions in CD and UC, respectively, and demonstrate that all major inflammatory signaling pathways are activated in affected mucosa. Some of the most prominent changes in the inflamed mucosa of CD patients are the consistent increase in kinase activities of the epidermal growth factor receptor (EGFR), phosphoinositide 3-kinases (PI3Ks), protein kinase B (PKB), focal adhesion kinase (FAK), Janus kinases (JAKs), nuclear factor κBα (NF-κBα), c-Src, the mitogen-activated protein kinase (MAPK) signaling cassettes, LKB1, p21-activated kinase 1/2 (PAK1/2), and Rho kinase. Conventional Western blot analysis was used to confirm some of these kinomic alterations in a cohort of different patients (examples shown in Fig. 3A and fig. S1A). Several of these changes were also observed in the inflamed mucosa of UC patients. However, in clear contrast to CD, inflamed tissue of UC is characterized by decreased EGFR and increased fibroblast growth factor receptor signaling, whereas NF-κB activation seems less increased in this disease. A role for enhanced activity of JAKs (16), Rho kinase (17), and the MAPK signaling cassettes (18, 19) in the colonic mucosa of IBD patients, as well as the dichotomy in EGFR signaling between CD and UC patients, has been reported earlier (20), and thus, our results complement an existing body of biomedical literature. On the other hand, increased activity of PI3K, PKB, FAK, c-Src, and LKB1 in the inflamed mucosa of IBD patients has not been directly documented before and hence shows that unbiased approaches can offer an insight into the pathophysiology of disease.

Fig. 1 Signal transduction in inflamed versus noninflamed mucosa in IBD.

Kinome profiling results of colonic biopsies using peptide arrays exhibiting 976 different kinase substrates. Paired colonic biopsies of inflamed and noninflamed mucosa of three CD patients and six UC patients and biopsies from six individuals without apparent colonic disease were analyzed for kinase activity. Relative intensity of slides from one UC patient was high (because of protein quantification error). Although paired analysis of inflamed and noninflamed regions in this patient yielded similar results, these slides were taken out of the comparison with control biopsies. Results were collapsed on elective signal transduction categories and expressed in gray scale (representing the fraction of peptides phosphorylated of all category peptides). Results for inflamed and noninflamed mucosa were contrasted, providing an overview of the signaling in healthy colon and IBD. Note the difference in p21Rac-dependent signaling between inflamed and noninflamed colon.

Fig. 2 Heat map of kinomic analysis.

Comparison of the elective signal transduction pathways between IBD and controls. Green: Negative Z scores present lower than expected values (shown in black). Red: Positive Z scores present higher values.

Fig. 3 Inhibition of p21Rac1 in noninflamed IBD.

(A) Immunoblot phosphorylation expression levels of a selection of proteins confirm changes in kinase activities observed by kinome arrays. (B) Simplified scheme depicting the signaling pathway of p21Rac1 and related GTPases. (C) Proinflammatory stimuli activate p21Rac1 in phagocytes. Granulocytes were stimulated with 10 μM fMLP, GM-CSF (5 ng/ml) + fMLP, or IL-8 (10 ng/ml), and GTP-p21Rac was determined. Serum-starved monocytic THP1 cells were stimulated with TNFα (10 ng/ml) or 10 μM IL-6. (D) Quantification of immunoblot analysis of expression levels of phosphorylated PAK2 (PAK2pSer20) (n = 10 for CD patients and 8 for controls, **P = 0.0098). (E) IHC for phosphorylated PAK2 on mucosal biopsies. (F) IHC for Rac-GTP on mucosal biopsies. (G) Quantification of Rac-GTP levels in interstitium and epithelium shows increased GTP-Rac levels in inflamed CD biopsies, and decreased GTP-Rac levels in noninflamed CD biopsies, compared to controls (**P < 0.007, ***P = 0.001, unpaired Student’s t test; n = 5 for noninflamed CD, 4 for inflamed CD, and 3 for controls; two to four images analyzed per patient). (H) Quantification of p21Rac1 activation levels by densitometry shows increased GTP-p21Rac1 levels in inflamed CD biopsies and decreased GTP-Rac1 levels in noninflamed CD biopsies, compared to controls (*P = 0.0059, **P = 0.0486, ***P = 0.0005, unpaired Student’s t test; n = 10 for CD patients and 7 for controls). A representative blot is shown in the panel below. (I) GST-PAKcrib protein was used to pull down all PAK binding partners from biopsies. Membranes were probed with p21Cdc42 (upper panel) or p21Rac2 (lower panel) antibodies, followed by Odyssey quantification. Bar graphs below show no differences between CD patients and non-IBD controls (n = 3). See fig. S1B for specificity of antibodies used.

Down-regulation of p21Rac signaling in noninflamed mucosa of IBD patients

Heat map analysis of profiling results reveals that the most prominent changes in kinome signatures are observed in inflamed mucosa from CD and UC patients (Fig. 2). The kinome signature of noninflamed mucosa from UC patients appears relatively comparable to that of noninflamed mucosa from CD patients, suggesting that similar molecular processes maintain the noninflamed state in both manifestations of IBD. Specific differences were observed in the noninflamed mucosa of IBD patients compared to the (noninflamed) mucosa of non-IBD patients, indicating the existence of specific anti-inflammatory mechanisms that are operative in the noninflamed mucosal areas in IBD patients. One of the most marked differences between inflamed and noninflamed mucosa in general was found in the activation of Rac-dependent signal transduction (a schematic overview of the principles of Rac signaling can be found in Fig. 3B). Rac constitutes a family of guanosine triphosphatases (GTPases) that cycle between an active, guanosine triphosphate (GTP)–bound state and an inactive, guanosine diphosphate (GDP)–bound state (21). These proteins play an important role in innate immune cell functions like phagocytosis and activation of the respiratory burst to combat bacterial infection (22). Indeed, stimuli that activate innate immunity, such as granulocyte-macrophage colony-stimulating factor (GM-CSF), N-formyl-Met-Leu-Phe (fMLP), interleukin-6 (IL-6), and IL-8, are potent activators of p21Rac1 GTP loading in phagocytes (Fig. 3C). Intriguingly, tumor necrosis factor–α (TNFα) also potently stimulates p21Rac1 GTP loading, and thus, anti-TNFα medication may conceivably counteract IBD (23) through Rac inhibition. To confirm our observation of aberrant p21Rac1 activation in IBD, we determined the phosphorylation state of the direct p21Rac1 effector PAK2 (24) by Western blot analysis and immunohistochemistry (IHC) (Fig. 3, A, D, and E, and figs. S1A and S2A), determined Rac-GTP levels by IHC (Fig. 3, F and G, and fig. S2B), and performed direct measurement of p21Rac-GTP levels (Fig. 3H). Additionally, analysis of indirect effectors, such as p38 MAPK (25) (Fig. 3A and fig. S1A) and cofilin (fig. S1B), confirmed the suppression of p21Rac1-dependent signal transduction in the noninflamed mucosa and the overactivation of this pathway in the inflamed mucosa of IBD patients. We also performed activity measurements of other GTPases, such as p21Cdc42 and p21Rac2 (Fig. 3G and fig. S1C). The results do not show differences in activity of either Cdc42 or Rac2 between the noninflamed, inflamed, and control mucosa, indicating that inhibition of p21Rac1 in the noninflamed areas of IBD patients is highly specific to this particular GTPase.

The sampling of clinical material always entails certain biases. In our case, controls were usually older than the IBD patients because the former samples are often obtained during screening for colorectal cancer, a disease of the elderly, whereas IBD affects younger people (table S1). Similarly, differences occurred as to the exact location where samples were taken in the colon. However, stratifying samples according to age or colonic location showed that age per se does not affect p21Rac1 activity (fig. S3A), nor did we observe a regional specification with respect to p21Rac1 activation in the colon (fig. S3B). Thus, we concluded that the p21Rac1 effects seen in IBD are a bona fide representation of mechanistic factors acting in the disease.

To pinpoint the cell type responsible for the reduced p21Rac1 activation observed in noninflamed mucosa in IBD, we performed IHC colabeling of Rac-GTP with CD68, CD138, and Ly6G&C/CD66b, markers for monocytes/macrophages, B cells, and granulocytes/monocytes, respectively (Fig. 4, A to C), clearly demonstrating colocalization of these immune cell markers with high-intensity Rac-GTP staining. In addition, quantification reveals that Rac-GTP positivity is higher in the interstitium than in the epithelium compartment (Fig. 3G), highlighting the immune system restriction of the effects on p21Rac1 signaling seen. High-power light microscopy with quantitation of the phospho-PAK2+ cells bearing the CD68, CD138, and Ly6C markers (Fig. 4D and fig. S4) confirms that the Rac-PAK2 signal is highly expressed in cells involved in innate immunity (although we cannot completely rule out some expression in T cells).

Fig. 4 Colocalization of p21Rac1 signaling with immune cells in mucosal biopsies.

(A) Single staining of GTP-Rac (blue), the monocyte marker CD68 (pink), and costaining on consecutive sections. Intense Rac-GTP staining is observed in the interstitium and colocalizes with CD68. (B) Intense staining of GTP-Rac (blue) colocalizes with the granulocyte marker CD66b (pink) (consecutive sections stained). (C) Syndecan 1 (CD138) is present in both the epithelium and interstitium, but colocalizes with intense Rac-GTP staining (blue) in the interstitium. (D) PAK2Ser20 staining colocalizes with immune cells, as determined by CD68, CD138, and CD66b costaining. See fig. S3 for controls and quantification.

Diminished exchange of GDP to GTP and increased p21Rac1-GTP hydrolysis in noninflamed mucosa from IBD patients

Next, we set out to elucidate the mechanisms responsible for the altered p21Rac1 activity in mucosa from IBD patients. One possibility is that p21Rac1 expression itself is differentially regulated, supported by the observation that noncoding polymorphisms in p21Rac1 are associated with increased susceptibility to UC (26). However, expression analysis of a large cohort of IBD patients did not reveal differences in the mRNA levels of RAC1 or RAC2 in either peripheral blood cells (Fig. 5A) or mucosal biopsies (Fig. 5B) from affected areas and noninflamed regions of IBD patients. In addition, polymorphisms in the RAC1 gene itself did not correlate with diminished Rac signaling in noninflamed mucosa of CD patients (table S3), although some evidence was obtained (but did not reach statistical significance) that other CD susceptibility genes may influence the balance between Rac activation in inflamed mucosa and noninflamed mucosa. However, when we analyzed phosphorylation of Vav, a guanine nucleotide exchange factor (GEF) that acts as a major upstream activator of p21Rac in immunity (27), we observed that its activation is diminished in the noninflamed colon of CD patients (Fig. 5C), suggesting that the effect on Rac signaling is of a posttranslational nature and involves differences in GTP loading of this family of immune function relevant GTPases.

Fig. 5 Altered GEF and GAP activity in CD mucosa.

(A and B) Expression of p21Rac1 and p21Rac2 mRNA in peripheral blood mononuclear cells (PBMCs) (A) and colon biopsies (B). The results derive from the GDS1615 and GDS1642 data depositories and show Rac expression in peripheral blood (42 healthy controls, 25 UC, and 59 C D) and colonic mucosa (4 healthy controls, 4 noninflamed colonic UC, 5 inflamed colonic UC, 13 noninflamed CD, and 7 inflamed CD). Percentile rank scores were used. (C) Decreased Vav phosphorylation in CD patient noninflamed mucosa. (D) GAP activity toward p21Rac1 in CD. Top: GTP-loaded human recombinant His-tagged p21Rac1 was incubated with biopsy lysate. A marked reduction in GTP-Rac1 levels was observed (compare left versus right lane). Bottom: GAP assay on noninflamed and inflamed biopsies from CD patients (n = 3) (also see fig. S4A). (E) GDP→GTP exchange activity toward p21Rac1 in CD. Top: GDP-loaded human recombinant His-tagged Rac1 was incubated with lysates from noninflamed and inflamed CD patient mucosa. The inflamed region shows more exchange (compare left and right lane). Bottom: GEF assay quantification showing the significant increase in exchange activity toward Rac1 in inflamed CD mucosa (*P = 0.0092, n = 3).

The activities of small GTPases such as p21Rac1 are regulated by the actions of both GEFs and GTPase activating proteins (GAPs) (22). Whereas GEFs stimulate GTP loading, and hence activity, of Rac proteins, GAPs inactivate Rac by activating its intrinsic GTPase activity, resulting in GTP hydrolysis to GDP (see Fig. 3B). We subsequently investigated GAP and GEF activity in lysates from mucosal biopsies using recombinant His-tagged p21Rac1 to avoid possible issues with endogenous p21Rac1 present in the lysates. First, we confirmed efficient loading of recombinant p21Rac1 with GTP (fig. S5A) and showed that GAP activity assays worked well by demonstrating that incubation of recombinant GTP-loaded Rac with lysates from biopsies results in a clear decrease in its GTP loading (Fig. 5D, upper panel). Subsequent comparisons of Rac-GAP activity between inflamed and noninflamed biopsies from CD patients indicate a higher activity of Rac-GAPs in the noninflamed regions (Fig. 5D, lower panel). Furthermore, in biopsies taken from inflamed regions, GAP activity is counteracted by an increased GEF activity toward p21Rac1 (Fig. 5E). The lower GEF activity in noninflamed CD mucosa is in apparent agreement with the reduced phosphorylation of Vav in these areas (Fig. 5C), although other Rac GEFs may be involved as well. Thus, a deregulation of GAP and GEF activities and especially down-modulation of exchange activity toward p21Rac1 explain the suppression of p21Rac1-dependent signal transduction in the noninflamed mucosa of CD patients.

Effect of p21Rac1 inhibition on innate immune function

It is becoming clear that IBD, and especially CD, has many characteristics of innate immune deficiency, at least in a subgroup of patients (911). We therefore decided to investigate the effects of p21Rac1 inhibition on innate immune effector functions. First, monocytes from CD patients and age- and gender-matched healthy controls were subjected in parallel to Escherichia coli phagocytosis assays. Figure 6A shows that monocytes from (unmedicated) CD patients are significantly less efficient at phagocytosis than those from healthy controls (P = 0.0313). Expression of a large panel of cell surface markers involved in bacterial uptake (Fc receptors CD32 and CD64, and phagocytosis receptors CD11b and CD44) was measured but did not provide evidence that differences in phagocyte phenotype may explain the differences in innate immune effector function (Fig. 6B).

Fig. 6 p21Rac1 and innate immunity.

(A) E. coli phagocytosis by CD patient monocytes is significantly lower than that in age- and gender-matched controls. (B) Cell surface expression of monocyte markers CD32, CD11b, CD44, and CD64 does not differ between CD patients and controls. MFI, mean fluorescence intensity. (C) Treatment with 5 μM NSC23766 leads to an increase in both the fractions of monocytes displaying successful phagocytosis (that is, full engulfment) and the number of E. coli phagocytosed per monocyte. (D) Phagocytes were treated with increasing concentrations of NSC23766, and p21Rac1-GTP was determined. Lanes cropped from the same blot are shown (original file: fig. S4B). Densitometry represents mean of two independent experiments. (E) Phagocytosis in monocytes isolated from controls and quiescent CD patients, untreated or treated with 5 μM NSC23766. (F) Increased PAK2ser20 levels in phagocytes from CD patients compared to healthy controls (n = 12). See also fig. S4. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (G) Bacterial killing in monocytes isolated from controls and quiescent CD patients, untreated or treated with 5 μM NSC23766. (H) Bacterial killing was corrected for phagocytosis. Inhibition of p21Rac1 restores bacterial killing of CD monocytes to the control level.

We subsequently speculated that the reduction in p21Rac1-GTP loading in noninflamed regions of an otherwise inflamed intestine might improve the imbalance between Rac activation and inactivation and thus facilitate local phagocytosis. To test this hypothesis, we treated monocytes with a low dose of a pharmacological inhibitor of p21Rac GTP loading, NSC23766 (28). Enhanced phagocytosis of fluorescein isothiocyanate (FITC)–labeled E. coli was observed with respect to both the fraction of monocytes displaying successful phagocytosis (that is, full engulfment) and the number of E. coli phagocytosed per monocyte (Fig. 6C). This low dose of NSC23766 was effective in partially blocking p21Rac1 activity (Fig. 6D). It is important to note, however, that NSC23766 only inhibits a subset of GEFs for p21Rac1, because p21Rac1 activation elicited in phagocytes by the combination of GM-CSF and fMLP was unaffected by this compound (fig. S5B). Monocytes from patients with CD also strongly respond to moderate Rac inhibition with enhanced phagocytosis (Fig. 6E). Moreover, upon such inhibition of the NSC23766-sensitive GEFs for p21Rac1, statistical differences in phagocyte function between healthy controls and CD patients were abrogated, providing further evidence that these GEFs mediate functional differences between CD patients and healthy controls in phagocyte function. Because phagocytosis is decreased in CD patients, and improves upon Rac inhibition, we tested whether endogenous Rac signaling is enhanced in CD monocytes. Indeed, we detected higher levels of PAK2 phosphorylation in peripheral blood phagocytes from CD patients compared to healthy controls (Fig. 6F and fig. S6).

We next tested the effect of p21Rac1 inhibition on bacterial killing. First, we observed a reduced number of bacterial colonies obtained from phagocytes from CD patients compared to healthy controls (Fig. 6G), which, upon correction for the number of bacteria taken up by these cells, corresponds to a decreased bacterial killing (Fig. 6H). Treatment of phagocytes with low doses of the p21Rac1 inhibitor strongly enhances the capacity of such cells to kill E. coli (Fig. 6H), evident from a decrease in the number of colonies (Fig. 6G), despite an increase in bacterial phagocytosis (Fig. 6E). Thus, exaggerated p21Rac1 activity impairs the capacity of the immune system in CD patients.

Bell-shaped dose response of p21Rac1 inhibitors in innate immunity

Our results show that moderate p21Rac1 inhibition increases innate immune function. However, because p21Rac1 is essential for phagocytosis per se, full p21Rac inhibition should again be associated with reduced phagocytosis. Indeed, we observed that exogenous introduction of either constitutively active or dominant negative p21Rac1 proteins could impair phagocytosis (Fig. 7A), which is consistent with previous reports that p21Rac1 activity needs to be finely tuned in order for this process to be performed efficiently (29). To further elucidate p21Rac1 dependency of phagocyte immune function, we performed concentration curves with NSC23766. The dose-response relationship between Rac inhibition and phagocytosis is of a bell-shaped nature, with higher levels of Rac inhibition provoking a complete shutdown of phagocytic capacity (Fig. 7B). In contrast, increasing concentrations of NSC23766 result in inhibition of monocyte migration (Fig. 7C), suggesting that different Rac-dependent functions may be regulated through different GEFs. Together, these data imply that the p21Rac1 activity needs to be balanced within a proper dynamic range and needs to cycle between an active and inactive state for optimal phagocyte activity, providing a direct possible link between Rac1 inhibition in noninflamed patches and function of the innate immune system.

Fig. 7 6-TG, p21Rac1 inhibition, and innate immunity.

(A) Effect of Rac mutants on uptake of fluorescently labeled catalase. Murine macrophages were injected with fluorescent immunoglobulin G and, where appropriate, with either constitutively active Rac (RacV12) or dominant negative Rac (RacN12). After incubation with fluorescent catalase (6.6 mg/ml) for 45 min, anti-catalase antibodies were used to assay Fc receptor–mediated uptake. (B) Monocytes from three healthy individuals were preincubated with NSC23766 and exposed to FITC-labeled E. coli. Bacterial phagocytosis by monocytes was enhanced by NSC23766 at lower concentrations but inhibited at higher concentrations. (C) Monocyte migration was inhibited by increasing concentrations of NSC23766 (n = 3). (D) Stimulation of phagocytosis by 6-TG corresponds to moderate p21Rac1 inhibition as evaluated by the effects of 6-TG on PGN-elicited PAK2 autophosphorylation. In turn, inhibition of phagocytosis by higher 6-TG concentrations correlates with severe repression of PAK2 autophosphorylation. (E) Effects of 6-TG on p21Rac1-dependent signaling are much more pronounced in monocytes compared to lymphocytes. The effect of a 30-min incubation of 30 μM 6-TG was investigated using PAK2 autophosphorylation as a readout. (F) As in (C), bacterial phagocytosis by monocytes was enhanced by 6-TG at lower concentrations but inhibited at higher concentrations. (G) Phagocytosis by monocytes was increased by 10 μM 6-TG in both CD patients and healthy controls. (H) Effect of 6-TG treatment (48 hours) of monocytes on IL-8 and IL-10 production (n = 7). PGN stimulation (30 ng/ml) drastically increased IL-8 production. 6-TG treatment (10 μM) significantly enhanced PGN-stimulated IL-8 production (*P = 0.019). PGN incubation of monocytes induced a significant increase in IL-10 production, which was not further increased by cotreatment with 6-TG.

Effect of 6-thioguanine on innate immunity via p21Rac1 inhibition

Azathioprine (AZA) is an important drug in the treatment of IBD (1). After oral administration and absorption, the prodrug AZA undergoes conversion to 6-thioguanine (6-TG) via various metabolic steps (30). AZA is a slow-acting compound in CD, suggesting that its activity lies more in remission maintenance than in remission induction (31). Recent studies have shown that AZA and its metabolites impair lymphocyte proliferation and induce apoptosis of human CD4+ T lymphocytes through inhibition of p21Rac1 after 5 days of treatment (32), but a link to innate immune cells and their function has not been made. In otherwise unchallenged monocytes, a 30-min incubation with 6-TG strongly reduces Rac1-GTP levels (fig. S7A), monocyte spreading (in which Rac1-mediated actin rearrangement plays a key role; fig. S7B), and either basal or stimulated autophosphorylation of PAK2 (Fig. 7, D and E). In contrast, exposure of T lymphocytes to the same short-term 6-TG treatment left the levels of basal or stimulated PAK2 autophosphorylation unchanged (Fig. 7E), suggesting that 6-TG is more efficient in targeting p21Rac1 in innate immunity compared to adaptive responses, an effect possibly due to the slow p21Rac1 GDP/GTP cycle in lymphocytes. Thus, 6-TG is a potent Rac1 inhibitor in the monocyte compartment, and it is tempting to link its action in maintaining remission in IBD to increased innate function mediated through moderate p21Rac1 inhibition.

To test this hypothesis, we analyzed the effect of 6-TG treatment on phagocyte immune function. We observed that if monocytes were incubated with various concentrations of 6-TG for 30 min and subsequently challenged with FITC-labeled E. coli, they exhibit a dichotomous response to the medication; for example, phagocytosis is enhanced at 6-TG concentrations between 10 and 60 μM (Fig. 7F and fig. S7, C and D) but is strongly inhibited at higher concentrations. The concentrations at which 6-TG is effective on phagocytosis correspond to the therapeutic window of the drug in IBD (31) and also to moderate inhibition of p21Rac1 signaling in monocytes (Fig. 7, D and E). The stimulation of monocyte phagocyte function by 10 μM 6-TG was confirmed by analyzing monocytes from 18 healthy controls and 7 CD patients. Compared to vehicle condition, 6-TG significantly stimulated both the number of monocytes exhibiting phagocytosis and the number of bacteria phagocytosed per monocyte in both CD patients and controls (Fig. 7G).

Another monocytic immune function is their production of IL-8, a cytokine that is important for the innate immune-associated granulocyte recruitment. Because monocytes of CD patients show impaired IL-8 production (10), we first investigated the effect of 6-TG on this immune response. In contrast to IL-10, 6-TG increased IL-8 production in monocytes even under unstimulated conditions (+45%, n = 4, Fig. 7H). Upon stimulation with peptidoglycan (PGN), a strong increase in production of both IL-8 and IL-10 by monocytes was observed. Notably, 6-TG also provoked a superstimulation of PGN-dependent IL-8 production (+50%, n = 7, P < 0.05), whereas no differences were observed in IL-10. Thus, the cytokine profile induced by 6-TG–mediated p21Rac1 inhibition in monocyte cultures is consistent with a stimulation of innate immunity. Next, we analyzed IL-8 production in biopsies and demonstrated that although IL-8 production was lower than that at inflamed sites, noninflamed mucosa in an otherwise inflamed environment was characterized by higher levels of IL-8 production compared to either controls or CD patients in remission (fig. S8), indicating that a limited amount of IL-8 production through p21Rac1 inhibition may be required for maintaining local remission. These data suggest that suppression of p21Rac1-mediated signaling may contribute to the noninflamed phenotype in patients through enforcement of innate immunity.


In IBD, and especially CD, noninflamed mucosae lie adjacent to inflamed areas. How the inflammation keeps from spreading to noninflamed regions remains unanswered, but may provide valuable insight into disease pathology and clues as to how remission in these patients may be gained and retained. The most marked observation that comes forward from our kinome signatures is the suppression of p21Rac signaling in non-IBD tissues of CD and UC patients compared to both the mucosa of non-IBD controls and the diseased areas of the CD and UC patients. This suggests that suppression of p21Rac signaling is instrumental for maintaining a noninflamed histotype during IBD. Rac inhibition in the noninflamed mucosa of CD patients appears restricted to the colon because we were unable to validate it in the ileum of such patients (fig. S1B). This may fit with the current idea that ileal CD has a fundamentally different etiology compared to colonic CD (3335), or may be the result of different immunological processes contributing to disease at these locations. Our data suggest that in some respects, colonic CD and UC may be more closely related than ileal and colonic CD because the former pair seem to share a common mechanism aimed at keeping inflammation at bay in noninflamed mucosa. WASP deficiency, which causes a CD-like colitis in mice (12), results in increased p21Rac1 activation (36), suggesting that deregulation of Rac activity is not only a characteristic of IBD, but that other types of enterocolitis, such as Wiskott-Aldrich syndrome, may relate to disturbed Rac functioning and subsequent impaired innate immunity.

Differences in immune cell composition in inflamed and noninflamed biopsies may account for some of the differences in kinome signatures observed. Although the technical limitations of this study precluded measuring p21Rac1 activity directly in different subsets of mucosal innate immune cells, the finding that inhibition of the Rac1/PAK2 signaling pathway, by either 6-TG or NSC23766, leads to enhanced peripheral blood monocyte phagocytosis and IL-8 production suggests that this inhibition would also coincide with a local enhancement of innate immune function. This notion fits well with the emerging concept that diminished innate immunity is the primary defect in at least part of the IBD patients. Thiopurines are the most effective drugs in CD for maintaining remission, and all act through the common metabolite 6-TG (30). Data by Tiede and colleagues showed that 6-TG specifically inhibits the activation of p21Rac GTPases in adaptive immunity (32), but did not link these effects to the innate immune system. In vivo, AZA and 6-TG are metabolized to thioguanine monophosphate and thence to thioguanine triphosphate by hypoxanthine-guanine phosphoribosyltransferase, and it is conceivable that these metabolites are the causative agents of Rac inhibition in our in vitro experiments. 6-TG has also been shown to be effective in colitis mouse models that do not rely primarily on immune cell imbalance (37). Thus, although our data show that expression of Rac-GTP in the mucosa is mostly localized to the immune compartment, and our in vitro data show a clear effect of 6-TG on phagocytes, it is conceivable that Rac signaling in the epithelium is also susceptible to 6-TG and its metabolites and may contribute to maintaining local remission. In this sense, it is important to note that the clinical effects of thiopurines often take 1 month or more to manifest, suggesting that the medication does not so much act on achieving remission in existing lesions and ulcers but rather prevents the occurrence of new ulcers. Thus, more specific inhibitors of p21Rac signaling may have significant clinical promise as they would be expected to yield fewer side effects (thiopurines target other GTP-mediated processes, in particular cell proliferation, as well). Together, our results define a new paradigm in which rectifying an imbalance in p21Rac1 activity mediates both spontaneous local remission and AZA-mediated clinical benefit in IBD by enhancement of innate immune function.


Study design, patients, and specimens

The predefined aim of this study was the generation of a comprehensive comparison of signal transduction in inflamed and noninflamed colonic mucosa of IBD patients and contrast results to colonic mucosa of healthy controls. To avoid biases, we used a cohort of sequential new presentations of IBD at the endoscopy ward at the Department of Gastroenterology of the University Medical Center Groningen (UMCG) in Groningen (collected between 26 March 2009 and 15 December 2009), whereas concomitantly a control cohort was collected from patients undergoing endoscopy on suspicion of colorectal cancer but without disease. CD was distinguished from UC using the Lennard-Jones criteria, with transmural inflammation, fibrosis, and discontinuous inflammation as histological parameters. Validation of kinome results and their possible physiological interpretation were performed on ad hoc recruited patients and controls, and experiments were conducted until statistical significance was reached or the H1 hypothesis was rejected. Because all experiments are internally controlled, no blinding was deemed necessary. This study received Institutional Review Board approval from the UMCG and Erasmus MC Rotterdam, and informed consent was obtained in accordance with the Declaration of Helsinki and the National Institutes of Health Belmont Report (protocol no. 2004.168). For all patients, paired biopsies from inflamed (if relevant) and noninflamed colon or ileum were obtained after informed consent. Full details of the nature of the disease and the localization in the colon of the biopsies obtained are listed in table S1.

Kinome array analysis

Kinome array analysis was done as described earlier (38, 39). Full biopsies were weighed and lysed with mPER lysis buffer in a final concentration of 0.4 mg biopsy/μl lysis buffer (average biopsy weight was 20 mg) with the addition of HALT protease (1:100) and phosphatase inhibitors (1:100). Detailed description of the method can be found in supplementary materials. ScanAlyze software was used ( Using grid tools, we analyzed spot density and individual background intensities. Data from three technical replicates were exported to an excel sheet and further analyzed as described in supplementary materials. Significance analysis was performed with a minimally modified algorithm developed for microarray analysis ( Z scores were calculated for the 59 predefined signal transduction pathways, and a heat map was created with CIMminer (

Western blotting

Western blotting was done as described earlier (40). Blots were incubated with the primary antibody overnight at 4°C. Phosphorylated PKAThr197, phosphorylated p38MAPKThr180/Tyr182, phosphorylated GSK3βSer9, phosphorylated SMAD1Ser463/465/5Ser463/465/8Ser426/428, phosphorylated PKBSer473, phosphorylated PAK2Ser20, Rac2, and Cdc42 were purchased from Cell Signaling; phosphorylated VavTyr174 was purchased from Epitomics; phosphorylated N-WASPTyr256 was from Abcam; and phosphorylated cofilinSer3 and anti-His were from Signalway Antibody. β-Actin was purchased from Santa Cruz Biotechnology. Total p21Rac1 antibody was from BD Biosciences.

p21Rac activation assay

Activated p21Rac was precipitated using the Cdc42 and Rac interactive binding domain of PAK (amino acids 56 to 272) as described previously (41). Further detailed description of the method can be found in supplementary materials.

GEF and GAP assays

His-tagged recombinant Rac1 protein (Cytoskeleton) was loaded with either GTP or GDP. The loaded recombinant protein was added to lysates of biopsies (as prepared for PepChip analysis). The presence of GTP-loaded recombinant Rac1 being reduced by endogenous GAPs present in the lysate is a measure of GAP activity. In contrast, exchange of GDP for GTP on GDP-loaded recombinant Rac1 is a measure of GEF activity present in the lysates. GTP-Rac was determined by p21Rac activation assay described above, with one exception: precipitated GTP-Rac was not determined by Rac1 antibodies but through His antibodies to distinguish between endogenous Rac1 and recombinant Rac1. For full description of method, see supplementary materials.

p21Rac expression analysis

For analysis of Rac1 and Rac2 transcript numbers in blood and colon of patients with IBD or healthy controls, we exploited the GDS1615 and GDS1642 data depositories as provided through Gene Expression Omnibus. To allow comparison of results, we used percentile rank scores for expression. Expression scores of Rac genes were high, allowing us to make meaningful comparisons. Possible differences between comparable groups were analyzed for possible statistical significance through one-sided heteroscedastic t tests, but none was detected.


IHC was performed on paraffin-embedded tissue sections with phosphorylated PAK2Ser20 (U.S. Biologicals) at a dilution of 1:200 or Rac-GTP (Biomol) at 1:500 as per the manufacturer’s instructions (see supplementary materials).

Monocyte isolation from human peripheral blood

Heparinized blood was obtained from healthy volunteers and patients with CD after informed consent. Patients were not receiving immunomodulatory medication, and all had quiescent disease according to the Harvey-Bradshaw score. Human PBMCs were isolated with Ficoll-Hypaque gradients and further purified with CD14 monoclonal antibodies conjugated to microbeads (Miltenyi Biotec). Cell cultures were maintained in RPMI 1640 supplemented with 10% heat-inactivated fetal calf serum and gentamicin (10 μg/ml) at 37°C in 5% CO2 humidified air. 6-TG pretreatment of monocytes is 10 μM/30 min unless otherwise indicated.

Phagocytosis assays

Phagocytosis assays were performed as described earlier (42) using monocytes pretreated with or without 6-TG (Sigma-Aldrich) or NSC23766 (Calbiochem) for 30 min and then challenged with FITC-labeled E. coli at a 1:5 cell/bacteria ratio for 5 min or other time points where indicated. At least 300 cells were counted for each slide.

Cytokine measurement

Monocytes (2 × 105/ml) were plated in 96-well plates and pretreated with 6-TG (10 μM) for 30 min. Subsequently, cells were stimulated with PGN (30 μg/ml) or vehicle. Culture supernatants were collected at 48 hours and analyzed by enzyme-linked immunosorbent assay (R&D Systems) for IL-8 or IL-10 production and normalized to the numbers of viable cells in each well using MTT assay. In addition, IL-8 mRNA production was analyzed by real-time polymerase chain reaction with the Assays-on-Demand kit on the ABI Prism 7700 (Applied Biosystems). 18S levels were used as endogenous control.


Isolated monocytes were allowed to migrate toward IL-8 (10 ng/ml) in a 4-nm Transwell assay for 4 hours, after which cells were counted by fluorescence-activated cell sorting analysis referenced to beads.

Statistical analysis

Significance analysis was done using the Wilcoxon ranked sign test for comparing two distributions or Kruskal-Wallis one-way analysis of variance for multiparameter analysis using an α of 0.05. For exchange and GTPase assays, heteroscedastic paired t tests were used. Testing was done with SPSS 16.0. Error bars represent the SEM.



Fig. S1. Confirmation of signaling differences in CD identified by PepChip analysis.

Fig. S2. Magnification of PAKSer20 and Rac-GTP IHC.

Fig. S3. Age of patients and location of biopsy in colon do not affect Rac activity.

Fig. S4. Macrophages, plasma cells, and neutrophils show changes in PAK phosphorylation.

Fig. S5. Increased PAK2Ser20 staining in peripheral blood phagocytes from CD patients.

Fig. S6. GEF assay and effect of NSC23766 on p21Rac1 activity.

Fig. S7. 6-TG reduces GTP-loaded Rac1 in monocytes from healthy individuals.

Fig. S8. IL-8 mRNA expression in colonic biopsies.

Table S1. Patients used for the biochemical experiments.

Table S2. Kinome results in CD UC.

Table S3. Genotyping information for patients.


  1. Funding: G.M.F. is financially supported by the Dutch Cancer Society (EMCR 2010-4737). M.P.P. is supported by the Netherlands Organisation for Scientific Research (840.12.001). C.d.H. and M.P.P. are also supported by personal European Crohn’s and Colitis Organisation grants. M.A. is supported by the Foundation Michelle. R.K.W. is supported by a VIDI grant (016.136.308) from the Netherlands Organisation for Scientific Research. L.Z. is supported by the National Science Foundation of China, grant 81200282. Author contributions: K.P., R.S., L.Z., G.M.F., J.J.D., A.R., and T.B. planned the experiments, performed the experiments, and wrote the manuscript. A.R., L. Vogelaar, L. Visser, M.A., and C.d.H. helped with experiments. R.K.W. and K.K.K. analyzed the data. V.J.N., C.J.v.d.W., E.J.K., and G.D. identified patients, planned the experiments, and wrote the paper. K.N.F. and M.P.P. planned the experiments and wrote the paper. Competing interests: There are no financial interests or any awarded or filed patents that could be perceived as being a conflict of interest.
View Abstract

Stay Connected to Science Translational Medicine

Editor's Blog

Navigate This Article