ARF6 inhibition stabilizes the vasculature and enhances survival during endotoxic shock

Abstract

The vascular endothelium responds to infection by destabilizing endothelial cell-cell junctions to allow fluid and cells to pass into peripheral tissues facilitating clearance of infection and tissue repair. During sepsis, endotoxin and other pro-inflammatory molecules induce excessive vascular leak, which can cause organ dysfunction, shock, and death. Current therapies for sepsis are limited to antibiotics and supportive care, which are often insufficient to reduce morbidity and prevent mortality. Previous attempts at blocking inflammatory cytokine responses in humans have proven ineffective at reducing the pathologies associated with sepsis, thus highlighting the need for a new therapeutic strategy. The small GTPase ARF6 is activated by a MYD88-ARNO interaction to induce vascular leak through disruption of endothelial adherens junctions. Here we show that the MYD88-ARNO-ARF6 signaling axis is responsible for lipopolysaccharide (LPS)-induced endothelial permeability and is a destabilizing convergence point utilized by multiple inflammatory cues. We also show that blocking ARF6 with a peptide construct of its N-terminus is sufficient to reduce vascular leak and enhance survival during endotoxic shock without inhibiting host cytokine response. Our data highlight the therapeutic potential of blocking ARF6 and reducing vascular leak for the treatment of inflammatory conditions such as endotoxemia.

Introduction

The innate immune system is the first line of defense against pathogenic microbes. It facilitates the recognition of microbial components such as endotoxin and initiates an inflammatory response that clears the invading organism and promotes reconstruction of damaged tissues. People with sepsis often have a frenetic inflammatory response and associated excess vascular leak that leads to tissue edema, organ failure, shock, and often death (1-3). Current treatment options are limited to supportive care and antibiotic therapies (4). Unfortunately, even with these options mortality still occurs in more than 25% of septic patients and occurs with even greater incidence in patients whose condition progresses to septic shock (5, 6).

The vast majority of therapeutic interventions for sepsis outside of antibiotic therapies and supportive care have focused on reducing the inflammatory and cytokine responses (7). These approaches, which include immunosupression by steroids (8), inhibition of the inflammatory Toll-Like receptor 4 (TLR4) with eritoran (9), and in particular, direct inhibition of cytokines such as TNF-α (10-12) have been successful in some animal models but have produced conflicting or negative outcomes in human phase III clinical trials. Because of this, alternative inflammatory pathways important to the pathology of sepsis need to be identified to exploit their potential as therapeutic targets.

The recognition of microbial components by the TLRs is critical to the inflammatory response during sepsis. TLRs are expressed in many cell types and stimulate a MYD88-mediated cascade, which leads to activation of the inflammatory transcription factor NF-κB and to the subsequent cytokine storm observed during sepsis (13, 14). We recently identified an association between MYD88 and the guanine nucleotide exchange factor ARNO. IL-1β requires this association to activate ARF6 and to induce vascular leak in a process independent of MYD88’s canonical role in NF-κB-mediated inflammatory gene expression (15). This MYD88-ARNO-ARF6 cascade promotes enhanced vascular permeability through the internalization of Vascular Endothelial-cadherin (VE-Cadherin). MYD88 is a critical adapter protein utilized by numerous other inflammatory pathways including the interleukin-18 receptor and most of the described TLRs (14). We therefore hypothesize that TLR stimulation may induce vascular permeability independent from cytokine expression, that this permeability is mediated by a MYD88-ARNO-ARF6 cascade, and that blocking ARF6 would enhance survival in models of sepsis.

Materials and Methods

Reagents

Human dermal microvascular endothelial cells (HMVEC-d) were purchased at passage 0 from Lonza and experiments were performed at passages 3-6. IKKβ/NF-κB inhibitor, SC-514, was purchased from Calbiochem. IKK-ε/IRF3 inhibitor, BX-795 was purchased from EMD Millipore. siRNAs were purchased from QIAGEN. Salmonella enterica serotype enteritidis lipopolysaccharides were purchased from Sigma. Catalogue number L7770 was used forin vitro experiments and L6011 was used for in vivo experiments. Endothelial cells were cultured in EGM-2 MV from Lonza.

siRNAs

siRNAs were diluted in 12.5% by volume HiPerFect Transfection Reagent (Qiagen) in Opti-MEM (Invitrogen) and incubated for at least 10m at room temperature. Passage 3-4 endothelial cells were trypsinized and resuspended in growth media, then combined with siRNAs such that the final concentration of siRNA was 30nM for all targets. Cells were then plated and media was changed the following morning. Three days after the initial transfection, the cells were transfected a second time using HiPerFect/siRNA concentrations as described above. All siRNA treatments were compared to the All-Stars Control siRNA (Qiagen).

Gene	Catalog Number	Target sequence (5’-3’)
ARF6	SI02757286	CAACGTGGAGACGGTGACTTA
MYD88	SI00300909	AACTGGAACAGACAAACTATC
ARNO	SI00061299	CACGCTGTTGGTAATCTTATT

Peptide synthesis

Peptides were synthesized in the University of Utah DNA/Peptide Core Facility using standard Fmoc synthesis protocol on an ABI 433 synthesizer (Life Technologies /Applied Biosystems, Carlsbad, California). In each cycle, a threefold or fourfold excess of the Fmoc-amino acid was activated with a mix of HBTU (2-(1H-benzotriazole-1-yl)-1,1,3,3-tetramethylaminium hexafluorophosphate), Oxyma Pure (ethyl cyano(hydroxyimino)acetate), and DIEA (diisopropylethylamine (1:1:2). Fmoc was removed in each cycle with 20% piperidine. In peptides with N-terminal myristoyl the last coupling was myristic acid. The peptide was cleaved and deprotected in a 2 hr treatment with TFA (trifluoroacetic acid)/ water / TIS (triisopropylsilane) (95:2.5:2.5) followed by precipitation in diethyl ether. The crude peptides were purified by reversed phase chromatography using a Hamilton PRP-1 column (Sigma/Aldrich) for myristoylated peptides or a Higgins Proto 200 C18 column (Higgins Analytical, Inc., Mountain View, Ca.) for peptides without myristoyl. In either case, a gradient of water/acetonitrile containing 0.1% TFA was used. The purest fractions of the eluate were combined, lyophilized, re-dissolved in 0.1M HCl and lyophilized again. Final purity was greater than 95% as measured by reversed phase chromatography monitored at 214nm. Molecular weight was confirmed within 0.1 Da by MALDI Mass Spectrometry.

Name	Sequence
MyrARF6 2-13	Myristoyl- GKVLSKIFGNKE
MyrSCR 2-13	Myristoyl- GNFKVILKGKES
ARF6 2-13	GKVLSKIFGNKE

Transwell permeability

Permeability assays were adapted from previous protocols (15-17). Twenty-four-well Costar™ cell-culture inserts (1.0 μm pore, 0.3cm² total growth area) coated with 100 μL of 15μg/mL human fibronectin solution were seeded with 30,000 HMVEC-D cells per insert for a total of 105 cells per cm². The media was replaced the following day and the assay was performed after a stable confluent monolayer was formed, 48 to 72 hr after seeding. LPS (100 ng/mL) or Mock) were added to both apical and basolateral sides of the cell-culture insert chamberin phenol red-free media and allowed to stimulate for 4 hr. HRP was added at a final concentration of 1mg/mL to the apical side of the cell-culture insert three hours after stimulation and allowed to leak for 1 hr. Media from the lower chamber was then collected and HRP concentration was assessed by addition of 0.5mM guaiacol and 0.6mM H₂O₂. Spectrophotometric analysis of absorbance at 490nm was used to assess permeability. For assays in which FITC dextran was used as a reporter, the same protocol was followed with two exceptions: the assay was carried out in EGM2-MV and fluorometric analysis of media in the lower chamber was performed using a microplate fluorometer to assess fluorophore concentration. Data are presented as mean ± SEM of at least three independent experiments and at least 12 individual replicates.

ARF6 activation assay

Primary HMVEC-D cells were treated with agonists and inhibitors as described in figure legends. Cells were then lysed and immunoprecipated with GGA3-conjugated agarose beads (Cell Biolabs, STA419). Immunoprecipitates (GTP-bound ARF6) and whole cell lysates (Total ARF6) were then analyzed via western blot with an anti-ARF6 antibody (1:1000) from Millipore (05-1149).

VE-Cadherin internalization

Internalization assays were adapted from previous protocols (15, 17). HMVEC-D cells were seeded on 8-well, glass-bottomed chamber slides coated with human fibronectin. Cells were incubated in dialyzed VE-Cadherin BV6 antibody solution from Millipore (MABT134) that was diluted in labeling buffer (EBM-2 (Lonza) + 3% fatty acid-free BSA + 20mM HEPES) at a concentration of 5ug/mL for 1h at 4°C. Media was then replaced with EGM-2 MV with 100ng/mL LPS or vehicle to stimulate endocytosis and 0.6mM Primaquine (Sigma) to block recycling of vesicles back to the plasma membrane. Cells were acid-washed 3 times in HBSS (Invitrogen)+ Ca + Mg + 3% fatty acid-free BSA + 50 mM glycine, pH 2.7 to remove non-internalized antibody and then fixed in 4% PFA for 10 minutes at room temperature. Fixative was washed away with HBSS and cells were permeabilized with 0.3 % triton in HBSS for 5 min followed by blocking in 5% non-fat milk + 10% normal donkey serum in HBSS for 1hr. Detection of VE-Cadherin was performed with by Alexa Fluor ® 488-conjugaed anti-mouse IgG (Invitrogen) and cells were mounted in antifade medium + 2 μg/mL DAPI. Five random fields/well were imaged on an Olympus FV1000 confocal microscope at 600X magnification with 3 individual Z-slices, 0.3μm apart. To score the image, the total number of cells with greater than 10 488nm positive vesicles divided by the total number of cells (DAPI-stained nuclei) per field of a collapsed Z-projection were counted. Statistics were calculated using a two-way ANOVA in PRISM.

Colocalization with EEA1 was performed in the same manner, but the rabbit EEA1 primary antibody (ABCAM) was added immediately after the blocking step and incubated for 1 hr at room temperature.

Immunoprecipitation

The coding sequence of each functional domain or full-length of MYD88 and ARNO were amplified from IMAGE cDNA clone by PCR and ligated into pcDNA3.1 vector after enzyme digestion. The ARNO and functional domain constructs contained a Myc and His epitope, while MYD88 constructs contained an HA epitope. 293T cells were transfected with indicated MYD88 or ARNO constructs. Cells were lysed in ice cold 50 mM tris (pH 7.5) 750 mM MgCl₂, 1% NP-40, and 10% glycerol supplemented with protease and phosphatase inhibitors (1861282, Thermo) and incubated with immobilized Anti-HA high affinity rat monoclonal antibody (clone 3F10) agarose beads (Roche 11 815 016 001) for 60 min at 4°C. Complexes were washed three times with lysis buffer and analyzed via western blot

Cytokine analysis

Human endothelial cell supernatants were treated as described in the figure legends. Cytokine expression was measured with the Millipore Milliplex magnetic bead assays as per manufacturer’s instructions. For murine plasma TNF-α concentrations, 3-4 month-old mice that had been treated I.P. with 25mg/kg LPS and I.V. with 40mmol/kg peptide treatment or vehicle were anesthetized and blood was collected via terminal cardiac puncture in the presence of EDTA. Whole blood was centrifuged and the plasma was obtained and stored at -80°C until analysis. TNF-α concentrations were analyzed using the Mouse TNF-alpha DuoSet ELISA (RnD systems DY410).

Leukocyte Rolling and Adherence assay

The protocol was adapted from previous studies (18, 19). HMVEC-D were seeded into each well of a human fibronectin-coated parallel plate flow chamber (U-plate 0.4 VI Fibronectin, Ibidi GMBH) at a density of 3 × 10⁴ cells/chamber. Media was changed daily until the cells had grown to confluence.

Human polymorphonuclear leukocytes (PMNs) were isolated using previously described protocols (20, 21). Human peripheral venous blood was collected from healthy, medication-free adult subjects and drawn into acid-citrate-dextrose (14% final concentration) through standard venipuncture technique and used immediately upon collection. All subjects provided informed consent and Institutional Review board approval was received for blood collection at the University of Utah. Platelet-rich-plasma was removed upon centrifugation of whole blood at 150 × g for 20 min at room temperature. The remaining red/white blood cell mixture was resuspended back to the original volume with 0.9% saline solution. Six percent dextran 70 was added to the cell mixture and left for 1 hr. The leukocyte-rich supernatant was removed and centrifuged at 400 × g for 5 min at 4°C. The supernatant was discarded and the pellet was resuspended in 0.2% NaCl immediately followed with an equal part 1.6% NaCl. Cells were then centrifuged at 400 × g for five minutes at 4°C. The pellet was resuspended in Hank’s Balanced Salt Solution supplemented with 1% human serum albumin (HBSS/A) and layered over an equal volume of Ficoll-Paque Plus. The suspension was centrifuged for 30 min at 400 × g at 4°C. The mononuclear leukocyte layer was removed, followed by the RBC layer, HBSS/A layer, and Ficoll-Paque layer. The remaining pellet containing at least 95% PMNs was washed with HBSS/A and resuspended to 10⁶ cells per mL in warm Ultrasaline.

Endothelial cells previously seeded into flow chambers were treated with LPS (100 ng/mL) and/or SC-514 for 3 hr in complete media. A syringe pump (Harvard Apparatus) was used to propel isolated and unstimulated primary human leukocytes through the parallel plate flow chambers (at 1 dynes × cm^-1 (typical venous shear stress (22)) in Ultrasaline. Differential Interference contrast microscopy (DIC) images were taken 1X per second for 30-seconds using an Olympus inverted microscope and a 10X objective, which is sufficient for visualizing a large number of endothelial cells. The imaged area was in the center (in both length and width) of each well, a point confirmed by the manufacturer using computational fluid dynamics to demonstrate the most uniform flow characteristics. The total number of PMNs rolling and adhered in all 30 images was quantified using MetaMorph software. Data presented is a quantification of the cumulative number of cells observed to be adhered or rolling on the endothelial monolayer in each image (total leukocytes adhered/30 seconds) ± SEM and represents data from one large field of view in the center of the well from at least 4 independent replicates. Statistics were calculated using ANOVA and Dunnett’s multiple comparison post-hoc test.

VE-Cadherin area quantification

HMVEC-D were seeded on BD-Falcon clear-bottom 96-well tissue culture plates (catalogue number 353219) at a density of 10,000 cells per well. Cells were incubated with denoted treatments for 5 hr and then fixed with 4% paraformaldehyde in PBS for 10 min. Cells were then permeabilized with 0.5% triton-X 100 for 2 min, blocked with Odyssey™ Blocking Buffer (Licor Biosciences) for 30 min at room temperature, and incubated with a mouse monoclonal VE-Cadherin antibody (BD, 555661) over night at 4°C. Cells were then rinsed with PBS and incubated with Alexa Fluor® 488 donkey anti-mouse secondary and Hoechst 33258 dye for 4 hr at room temperature. Cells were rinsed and the Molecular Devices ImageXpress Micro XLS collected nine non-overlapping images at the center of each well with a 20X objective. VE-cadherin area of each image was quantified with ImageJ by batch thresholding all images from an individual experiment and quantifying the number of pixels that were VE-Cadherin positive within each image. VE-cadherin area was then divided by the number of nuclei per image and then normalized to the mock-treatments. Data are compiled from three independent experiments with three replicate wells per condition in each experiment and composed of nine images per well for a total of 81 images per condition.

Organ permeability

This assay was adapted from previously published protocols (1, 23, 24). Male C57Bl/6 mice between 8 and 12 weeks old were purchased from Jackson Labs were anesthetized with ketamine and given 40 mmol/kg peptide followed by 40 mg/kg Evan’s Blue dye in saline delivered intravenously through the retro-orbital sinus injection. Anesthesia was reversed with 50 μL of 500 μg/mL antisedan administered subcutaneously. Six hr later, mice were then euthanized via CO₂ asphyxiation. The mass of each organ was measured. Organs were placed in 1 mL formamide for 72 hr at 60°C to elute the dye from each organ. Organs were centrifuged and the Evan’s blue accumulation was measured at 620nm with 740nm correction. Values were compared to a standard curve to calculate total amount of dye per gram (wet-weight) in each organ.

Endotoxemiasurvival

Male C57Bl/6 mice 8 to 12 weeks old were purchased from Jackson Labs. Mice were given 40 mmol/kg peptide I.V. through retro-orbital sinus injection and 25 mg/kg of LPS I.P. and monitored at least four times per day until all living mice resumed normal behaviors. To reduce death as an endpoint, all mice observed to be unresponsive to touch were euthanized immediately by CO₂ asphyxiation.

Statistics

Unless otherwise denoted, statistics and P values were calculated using one-way ANOVA with appropriate post-hoc test in PRISM software version 5.0f. * denotes P<0.05, ** denotes P< 0.01 and *** denotes P<0.001

All animal experiments were approved by our institutional animal care and use committee under protocol number 12-10002.

Results

TLRs induce vascular leak by activation of ARF6

To determine whether ARF6 is involved in TLR-mediated endothelial permeability, we utilized a well-characterized endothelial cell, primary human dermal microvascular endothelial cells (HMVEC-D), which is known to respond to TLR stimulation and is also widely utilized in endothelial cell-permeability assays (25-27). We treated primary HMVEC-D with LPS, the well-characterized agonist of TLR4 (28), and found that LPS was sufficient to induce ARF6 activation as measured by the increased presence of ARF6-GTP (Figure 1a-b) to biologically relevant levels consistent with other documented ARF6 functions (29-31). We also found that siRNA directed against ARF6 was sufficient to reduce permeability increased by LPS in a transwell system (Figure 1c). During the IL-1β signaling cascade in endothelial cells, the guanine nucleotide exchange factor ARNO interacts with the inflammatory adapter protein MYD88 to activate ARF6(15). We hypothesized that the LPS-mediated increase of permeability was controlled by this MYD88-ARNO interaction. Indeed, siRNAs directed against either MYD88 or ARNO were sufficient to prevent LPS-induced endothelial permeability (Figure 1d-e).

Figure 1. LPS utilizes MYD88-ARNO-ARF6 cascade to destabilize endothelial barriers.

(a) HMVEC-D were exposed to 100 ng/mL LPS and active, GTP-bound ARF6 was assessed by GGA3 immunoprecipitation assay and compared to total ARF6 in the lysate. (b) Quantification of (a) showing the geometric mean and 95% confidence interval; asterisks denote statistically significant results in which the lower bound of the 95% confidence interval for the normalized ratio does not cross 1 (the null value). (c) HMVEC-D treated with siRNAs directed toward ARF6 or a non-targeting siCTL RNA were assessed for their ability to reduce the leak of an HRP-reporter after treatment with 100 ng/mL LPS. HMVEC-D were transfected with siRNAs targeting ARNO (d) or MYD88 (e) and assessed for the flow of HRP across an endothelial monolayer after exposure to 100 ng/mL LPS. One-way ANOVA with Dunnett’s multiple comparisons test were used for statistical analysis. (f) Domain map of MYD88 and ARNO constructs. (g) Lysates from HEK 293T cells expressing ARNO-Myc and the indicated MYD88-HA constructs were immunoprecipitated with anti-Myc antibodies and immunoblotted with anti-HA antibodies. (h) Lysates from HEK 293T cells expressing MYD88-HA and the indicated ARNO-Myc constructs were immunoprecipitated with anti-Myc antibodies and immunoblotted with anti-HA antibodies.

To probe the interaction of MYD88 and ARNO, we first wanted to confirm the relationship by defining the interacting domains of MYD88 and ARNO. MYD88 is composed of three domains: the TLR- and IL-1R-related domain (TIR), the intermediate domain (ID), and the death domain (DD) (Figure 1f). The TIR domain interacts with both the cytoplasmic tail of TLRs and IL-1 receptor (IL-1R); the DD binds interleukin receptor associated kinase (IRAK); the binding partners of the ID are ill-defined (32). Interestingly, ARNO was immunoprecipitated with both the DD and TIR domains of MYD88 only when the ID was present. Furthermore, the coiled-coil (CC) and Pleckstrin homology (PH) domains of ARNO were found to be sufficient for ARNO’s interaction with MYD88 (Figure 1g-h).

A MYD88-ARNO-ARF6-controlled permeability hypothesis suggests that, like IL-1β, modulation of ARF6 may be independent of LPS-induced changes in gene expression and that those changes may be dispensable for LPS-induced permeability. Indeed, siRNAs targeting ARF6 were unable to block LPS-induced TNF-α expression in endothelial cells (Figure 2a). Additionally, blocking LPS-induced transcription factor NF-κB- or IRF3 activation using SC-514 or BX-795, respectively, failed to inhibit LPS-induced permeability but blocked other LPS-induced responses including leukocyte rolling, TNF-α, and IL-6 expression (Figure 2b-f)(33, 34).

Figure 2. LPS-induced permeability is independent of canonical TLR4 pathways.

(a) HMVEC-D were transfected with siRNAs targeting ARF6, exposed to 100 ng/mL LPS and assessed for TNF-α levels in the supernatants after 24 hr. (b-c) Inhibitors of IKK-β (50 μM SC-514) for NF-κB or IKK-ε (10 μM BX795) for IRF3 were administered to HMVEC-D and were assessed for their ability to reduce leak of an HRP-reporter across an endothelial monolayer post LPS exposure.(d), leukocyte rolling (e) TNF-α or (f) IL-6 expression at 24 hr post LPS-exposure, respectively. Statistics were performed using one-way ANOVAs and Bonferroni (a) or Dunnett’s (b-f) multiple comparisons tests.

LPS induces permeability through VE-Cadherin disruption

ARF6 has previously been implicated in the regulation of the internalization of the adherens junction protein VE-Cadherin, which suggests that the LPS-mediated increase in endothelial permeability occurs in a paracellular manner. To test this, we assessed leak of 2MDaFITC-dextran, a reporter too large for transcellular transport across endothelial monolayers (35). We observed enhanced movement of the 2MDa dextran across the monolayer upon LPS stimulation, suggesting that LPS enhances paracellular leak across the endothelium (Figure 3a). LPS also induced VE-Cadherin internalization as shown by enhanced presence of VE-Cadherin into endocytic vesicles containing early endosome antigen1 (EEA1) (Figure 3b-d). This internalization was abolished upon treatment of endothelial cells with siRNA directed against ARF6 mRNA further supporting the hypothesis that LPS mediates permeability through ARF6-mediated adherens junction internalization (Figure 3, e-f).

Figure 3. LPS induces permeability though VE-Cadherin disruption.

(a) HMVEC-D monolayers were incubated with either a 40 kDa FITC-dextran or a paracellular transport-restricted 2 MDa FITC dextran after LPS-exposure. (b, d) HMVEC-D were exposed to antibodies directed toward the extracellular domain of VE-cadherin and treated with 100 ng/mL LPS. At the indicated times, the monolayers were acid-washed to remove uninternalized antibodies then assessed by immunofluorescence for the percentage of cells with at least one VE-Cadherin (green) positive vesicle. (c) Immunofluorescence of VE-cadherin (green) overlaid with EEA1 (red) in HMVEC-D after 60min100 ng/mL LPS treatment. (e, f) HMVEC-D were transfected with siRNAs directed against ARF6 and assessed for their ability to internalize VE-Cadherin at the denoted times after initial LPS exposure. Scale bars are 30 μm.

Inhibition of ARF6 with its N-terminus reduces vascular leak

Having established that activation of ARF6 plays a critical role in the enhancement of endothelial permeability, we wanted to determine if deactivation of ARF6 would enhance endothelial barriers and reduce leak. A property of the ARF-family GTPases is that peptides constructed of their N-termini are proposed to inhibit ARF GDP-GTP exchange (36, 37). We synthesized a myristoylated peptide composed of amino acids 2-13 of ARF6 (MyrARF6 2-13). The ARF6 N-terminal peptide, but not scrambled (MyrSCR 2-13) or non-myristoylated (ARF6 2-13) controls, was sufficient to reduce ARF6 activation in primary human dermal endothelial cells (Figure 4a-b). Our model predicts that reduced ARF6 activation would result in enhanced endothelial barrier function. The ARF6 N-terminus, but not non-myristoylated or scrambled controls, reduced permeability across an endothelial monolayer (Figure 4c). This reduction in permeability corresponded to an increase of VE-Cadherin at cell junctions (Figure 4d-e). This ARF6 inhibitory function was also active in vivo and prevented the leak of Evan’s blue dye from the blood into both the lungs and the kidneys (Figure 4f-g). These data suggest that the synthetic ARF6 N-terminus may be useful as a novel therapeutic to prevent pathologies resulting from excess inflammatory vascular leak.

Figure 4. An N-terminus ARF6 peptide inhibits ARF6 activation and reduces vascular permeability.

(a-b) HMVEC-D were exposed to 25μM treatments of the myristoylated ARF6 2-13 peptide, non-myristoylated ARF6 2-13 peptide, scrambled myristoylated ARF6 2-13 peptide, or vehicle and lysates were assessed for the ratio of ARF6-GTP to total ARF6 in the cell lysate. Data are presented as the geometric mean and 95% confidence interval; asterisk denotes a statistically significant result in which the upper bound of the 95% confidence interval for the normalized ratio does not cross 1 (the null value). HMVEC-D were exposed to the denoted treatments and assessed for their ability to inhibit leak ofHRP(c) or increase VE-cadherin surface area (d-e). Mice were administered the 40mmol/kg of denoted peptides I.V. with Evan’s blue. Kidneys (f) as well as lungs (g) were removed 6 hr post injection. The concentration of dye in each tissue is reported. One-way ANOVA and Tukey’s multiple-comparison test were used for statistical analysis. Representative panels in d were selected after enhancement of contrast to 0.4% saturation of pixels and the scale bar is 30 μm.

The synthetic N-terminus of ARF6 enhances survival during endotoxic shock

Sepsis is a condition characterized by immense vascular leak that leads to tissue dysfunction, shock, and eventually death. We sought to determine if the stabilizing activity of the ARF6 N-terminus could enhance survival of mice in the endotoxemia model of sepsis. Lethal doses of LPS were administered I.P. immediately following I.V. injection of peptides. Mice were monitored at least four times per day until they resumed normal grooming behaviors for at least 24 hours. After 48 hours only 20% of the mice treated with LPS plus vehicle or a non-myristoylated peptide were alive. However, 85% of the mice treated with LPS plus ARF6 N-terminal peptide were alive at hour 48, and over 70% of the mice completely recovered (Figure 5a). Interestingly, peptide treatment enhanced survival, but did not significantly affect serum cytokine levels (Figure 5b). The observation that peptide treatment does not affect cytokine levels provides further evidence that ARF6 pathway inhibition can protect mice from the deleterious effects of inflammation while allowing canonical inflammatory cascades to continue.

Figure 5. The synthetic ARF6 N-terminal peptide enhances survival during endotoxemia.

Mice were anesthetized with ketamine and I.P. administered a lethal dose (25 mg/kg) of LPS concurrently with 40 mmol/kg-denoted peptide I.V. and monitored for survival (a) and plasma TNF-α concentration after 5 h, n=4 mice per group (b). The Mantel-Cox test (a) and one-way ANOVA with Dunnett’s multiple comparisons tests (b) were used to assess statistical significance.

Discussion

We have defined a signaling cascade in endothelial cells by which lipopolysaccharides induce vascular leak. We have also described a peptide inhibitor of this pathway, MyrARF6 2-13, which stabilized the vasculature in both in vitro and in vivoassays and enhanced survival in a mouse model of endotoxic shock. The endothelial cell monolayer is the critical barrier between blood and the peripheral tissues. It regulates the migration of fluid, nutrients, and cells from blood into the interstitial space. The integrity of this barrier is maintained by a balance of opposing cues that regulate endothelial cell transcytosis and cell-cell junctions (38). For example, homophilic interactions of VE-Cadherin at the cell-cell junction are enhanced upon stimulation by TIE2-dependent angiopoietin signaling (39) and ROBO4-dependent SLIT signaling (1) but are disrupted after stimulation with destabilizing cues such as VEGF (17), IL-1β(15), or LPS as described in this manuscript. The ability of the stabilizing receptor ROBO4 to prevent leak and pathologic angiogenesis is dependent on its ability to recruit the GTPase activating protein, GIT1, to inactivate ARF6(40). Conversely, the ability of the inflammatory mediators LPS and IL-1β to destabilize the vasculature is dependent on their ability to activate ARF6(15). These observations, combined with the data in this manuscript, place ARF6 at the nexus of multiple pathways that disrupt the endothelium. ARF6 activation leads to an unstable, leaky vasculature, while ARF6 inhibition leads to a stable, less permeable vasculature. This tightly regulated balance of endothelial stability is disrupted in conditions such as sepsis or pulmonary infections. During these conditions the inflammatory response induces intractable vascular leak, hypovolemic shock and/or acute respiratory distress syndrome, which can all precipitate death. By blocking ARF6 with its N-terminal peptide and reducing vascular leak, we are able to enhance survival of mice in the endotoxemia model of sepsis.

Our data in vitro and in vivo demonstrates that ARF6 regulates permeability independent from downstream inflammatory gene expression. Both knockdown of ARF6 expression in endothelial cells using specific siRNAs and the systemic in vivo administration of the ARF6 N-terminal peptide do not block LPS-induced TNF-α secretion. Thus far, all of the evidence in animal models, including the present study, suggests that ARF6, while necessary for disease pathology, does not regulate cytokine release (15). In contrast, it has been proposed that inhibition of ARF6 in fibroblasts, macrophages, and dendritic cells decreases LPS-induced cytokine production by preventing the association of MYD88 with TIRAP and TLR4(41, 42). However, LTA-induced cytokine expression, which also requires the association of MYD88 to TIRAP, was not affected by ARF6 inhibition suggesting a greater complexity in these TLR-signaling pathways (42). Furthermore, ARF6 inhibition has been demonstrated to have no effect on LPS-induced cytokine expression in dendritic cells (43). Therefore, the preponderance of in vitro evidence as well as all in vivo studies published to date suggests that ARF6 inhibition does not block cytokine expression or release.

Aside from enhancing vascular stability, inhibition of a highly conserved molecule such as ARF6 may have a number of developmental effects. For example, genetic ablation of ARF6 leads to the failure of proper hepatic cord formation and is embryonically lethal (44). However, the specific ablation of ARF6 in adults has never been described. In adult mice, inhibition of the ARF-family of GTPases prevented signal transduction of the insulin-receptor complex leading to transient insulin resistance in mice (45). Additionally, inhibition for two weeks showed no obvious toxicity but lowered inflammatory pathologies in mouse models of arthritis (15). ARF inhibition has also reduced invasion and metastasis of melanoma xenografts (46). Furthermore, ARF6 inhibition blocked proper VEGF receptor signaling and angiogenesis (40, 47) as well altered platelet activation (36, 48). Determining the effects of ARF6 inhibition on other tissues, especially in the context of hyperinflammation and sepsis is an important and ongoing area of research in our lab and others.

This manuscript identifies a novel host-pathogen interaction in which bacterial lipopolysaccharides activate aMYD88-ARNO-ARF6 cascade that induces endothelial leak independent of cytokine expression after stimulation (figure 6). We characterize an ARF6 inhibitor that reduces vascular leak and increases survival during lethal LPS-induced endotoxemia. MYD88 is a central effector of innate immunity and blocking the MYD88-ARNO-ARF6 cascade can potentially serve to stabilize the vasculature and prevent pathologies induced by multiple diverse inflammatory events regardless of the stimulus. We expect this study will open new avenues for the development of therapeutics that could be used for not only for the treatment of inflammatory processes related to infection, but also any inflammatory process associated with vascular instability, including trauma, epilepsy, and autoimmune diseases such as multiple sclerosis (49-51).

Figure 6. LPS utilizes a MYD88-ARNO-ARF6 cascade to induce VE-Cadherin internalization and vascular permeability independent from cytokine expression.

LPS stimulates TLR4 which recruits the adapter proteins MYD88 and TRIF to facilitate downstream transcriptional regulation such as cytokine expression. LPS also stimulates an independent and divergent pathway through the MYD88-ARNO-ARF6 signaling axis that dissociates VE-Cadherin from the cell-cell junctions allowing for vascular leak to occur. MyrARF6 2-13 blocks this divergent pathway by deactivating ARF6 and reducing vascular leak.