Canagliflozin protects against sepsis capillary leak syndrome by activating endothelial α1AMPK

Sepsis capillary leak syndrome (SCLS) is an independent prognostic factor for poor sepsis outcome. We previously demonstrated that α1AMP-activated protein kinase (α1AMPK) prevents sepsis-induced vascular hyperpermeability by mechanisms involving VE-cadherin (VE-Cad) stabilization and activation of p38 mitogen activated protein kinase/heat shock protein of 27 kDa (p38MAPK/HSP27) pathway. Canagliflozin, a sodium-glucose co-transporter 2 inhibitor, has recently been proven to activate AMPK in endothelial cells. Therefore, we hypothesized that canagliflozin could be of therapeutic potential in patients suffering from SCLS. We herein report that canagliflozin, used at clinically relevant concentrations, counteracts lipopolysaccharide-induced vascular hyperpermeability and albumin leakage in wild-type, but not in endothelial-specific α1AMPK-knockout mice. In vitro, canagliflozin was demonstrated to activate α1AMPK/p38MAPK/HSP27 pathway and to preserve VE-Cad’s integrity in human endothelial cells exposed to human septic plasma. In conclusion, our data demonstrate that canagliflozin protects against SCLS via an α1AMPK-dependent pathway, and lead us to consider novel therapeutic perspectives for this drug in SCLS.


Abbreviations
www.nature.com/scientificreports/ in e-AMPK KO mice demonstrated that canagliflozin protection was abrogated in the absence of endothelial α1AMPK. Indeed, endotoxemia-induced myocardial edema and albumin leakage persisted despite canagliflozin treatment in e-AMPK KO animals (Fig. 2c, e and g). Taken all results together, these data demonstrate that canagliflozin is indeed able to protect septic mice against LPS-induced vascular leakage, based on an endothelial α1AMPK-dependent mechanism.
Canagliflozin activates α1AMPK/p38 MAPK/HSP27 pathway in HMECs. To better understand the molecular mechanisms underlying canagliflozin protection of endothelial barrier function, we assessed the impact of canagliflozin treatment on AMPK activation and its downstream p38MAPK/HSP27 pathway in human endothelial cells. The p38MAPK/HSP27 pathway is known to mediate AMPK-dependent stabilization of inter-endothelial junctions by reorganizing the actin cytoskeleton 15,25,34 . Indeed, HSP27 is an actin-capping protein that inhibits actin polymerization by binding the microfilaments positive ends. HSP27 phosphorylation downstream of the p38MAPK releases HSP27 from actin, thereby enabling further filament polymerization in order to reinforce IEJ anchorage. Since SCLS primarily occurs at the level of capillaries and post-capillary venules, experiments were performed on human endothelial cells derived from the microcirculation (HMECs). Results indicate that canagliflozin dose-dependently activates AMPK, as assessed via phosphorylation of AMPK (Thr172) and its bona fide substrate acetyl-CoA carboxylase (ACC) (Ser79) (Fig. 3a, b). Furthermore, canagliflozin significantly increases phosphorylation of both p38 MAPK (Thr180/Tyr182) and HSP27 (Ser82), and this in a dose-dependent manner (Fig. 3c, d). Supplementary Fig. 1 shows that incubation of HMECs with 3 µM canagliflozin for increasing time periods also resulted in a significant and sustained AMPK activation, as represented by both AMPK (Thr172) and ACC (Ser79) phosphorylation. Taken together, these results demonstrate that clinically relevant canagliflozin concentrations do indeed activate the α1AMPK/p38MAPK/HSP27 pathway in HMECs, thereby potentially reinforcing inter-endothelial junctions by modulating actin cytoskeleton organization (Fig. 3e).
Canagliflozin-induced AMPK activation protects VE-Cad organization and endothelial barrier function against LPS injury. Next, we investigated the impact of canagliflozin on VE-Cad, i.e., the junctional protein known to be the gatekeeper of endothelial barrier function 12 . Therefore, immunostainings were performed on HMECs, either depleted or not in α1AMPK 25 , before being treated with canagliflozin (3 µM) and LPS (50 µg/mL) (Fig. 4a). Figure 4b illustrates that the continuous peripheral staining of VE-Cad under basal conditions appears to be disorganized in response to LPS treatment. This is associated with the formation of intercellular gaps. In contrast, canagliflozin likely strengthens VE-Cad anchorage within the plasma membrane, preserving its organization and preventing the formation of intercellular gaps in response to LPS. The response to canagliflozin treatment is abrogated in α1AMPK-deficient cells, as confirmed by VE-Cad signal quantifications ( Fig. 4b and c). Finally, the impact of canagliflozin-induced AMPK activation on the endothelial barrier function was evaluated in vitro by measuring the clearance of HRP-coupled streptavidin through the HMECs monolayer (Fig. 4d). Because cellular transfection affects by itself the barrier integrity, we employed the pan-AMPK inhibitor SBI0206965 to abrogate AMPK activation. As expected, LPS treatment was revealed to increase endothelial permeability, whereas canagliflozin was demonstrated to protect against LPS-induced endothelial    Data are expressed as mean ± SEM (3 biological replicates for each condition). # p < 0.05 is relative to respective non-treated HMECs, $ p < 0.05 is relative to LPS-only treated HMECs, and *p < 0.05 is relative to cells treated with DMSO. The data underwent two-way ANOVA. www.nature.com/scientificreports/ barrier disruption. The SBI0206965 compound, when given alone, was shown to significantly impair endothelial barrier function, whereas this agent completely abolished canagliflozin-induced protection. Of interest, these results are remarkably supported by our previous data demonstrating that α1AMPK is essential in both maintaining expression and architecture of IEJs under basal conditions, and mediating the protective effect of 991 compound, its best pharmacological activator, against IEJs disruption caused by LPS insult 25 .

Canagliflozin-induced AMPK activation protects VE-Cad integrity in HMECs challenged with human septic plasma.
Finally, in order to reinforce the translational perspectives of our work, the effects of canagliflozin were evaluated on HMECs incubated with human plasma collected from either control healthy volunteers (HV) or septic shock patients (SS) ( Fig. 5a and b). Supplementary Fig. 2 shows experiments performed with supplemental donors. Clinical characteristics of healthy donors and septic shock patients are summarized in Supplementary Tables 1 and 2. Immunostainings show that both HV and SS plasma affect VE-Cad architecture, with SS plasma inducing higher VE-Cad disruption, as represented by discontinuous jagged signals and intercellular gaps formation. Of major interest, canagliflozin importantly preserved VE-Cad integrity and linear organization, while slightly enhancing its membrane expression in HMECs exposed to both HV and SS plasma. On the other hand, α1AMPK depletion was associated with reduced, disrupted VE-Cad signal, and drastic decrease of canagliflozin protective effects. These, however, also seem to involve AMPK-independent mechanisms, since their abrogation appears inconstant in AMPK depleted cells.

Discussion
Our work highlights canagliflozin's protective effects on endotoxemia-induced vascular hyperpermeability and demonstrates that endothelial α1AMPK and its downstream p38MAPK/HSP27/VE-Cad regulatory pathway are involved in this protection. During the past decade, canagliflozin, along with other SGLT2i, have emerged as antidiabetic drugs that exhibit remarkable cardiovascular protection 30,31,35-37 , which is not fully explained by their blood glucose-lowering properties [38][39][40] . Extensive clinical studies are currently conducted to further characterize this protective action. Emerging hypotheses notably suggest that glucosuria and natriuresis, decreased inflammation, or reduced oxidative stress may all contribute to improve cardiovascular function 40 . Interestingly, the diuretic effects of SGLT2i have been shown to selectively reduce interstitial edema with minimal depletion of circulating blood volume [41][42][43] . Beyond postulating that endothelial barrier integrity possibly plays a significant role in this particular SGLT2i feature, such integrity appears particularly relevant in the SCLS setting.
Here, we have demonstrated that canagliflozin-induced microvascular protection depends, at least to some extent, on endothelial α1AMPK activation. While the SGLT2-induced AMPK activation is attracting growing interest, the hypothesis that this kinase mediates SGLT2's cardiovascular protective effect is still incompletely explored. Tampering inflammation [44][45][46] , reducing oxidative stress [47][48][49] , regulating nitric oxide production 45,50,51 , or preventing energy depletion 45,52,53 are overlapping cardiovascular protective mechanisms of SGLT2i and AMPK. Our data combined with recent findings supporting an empagliflozin-induced AMPK-dependent protection of microvascular barrier function 53 enable us to postulate that the endothelial barrier regulation as induced by AMPK activation also represents a key mechanism contributing to the SGLT2i-related cardiovascular protection.
One limitation of our study is that in our model, canagliflozin treatment was administered before LPS challenge. This protocol does, thus, not reflect the clinical reality of sepsis. The canagliflozin impact should further be evaluated and figured as a treatment of declared sepsis. A recent study demonstrating improved survival of septic mice subsequently treated with SGLT2i supports that promising results may reasonably be expected 54 . This perspective, however, raises several issues and questions. It should, first, be determined whether other SGLT2is could exert a similar protective mechanism. Indeed, although canagliflozin was initially described to activate AMPK more robustly compared to other SGLT2i 29 , both empagliflozin and dapagliflozin were subsequently reported to activate AMPK in vivo in both mice total heart samples 45,46 and cardiac fibroblasts 46 . Moreover, empagliflozin was proven particularly beneficial for microvascular barrier function 53 and against sepsis injury 45,54 . Therefore, we believe that vascular barrier protection would not be restricted to canagliflozin. Second, SGLT2i administration could be further optimized in order to avoid per os formulations for intensive care unit (ICU) settings. In this respect, it is worth mentioning that the feasibility of intravenous canagliflozin administration has been demonstrated recently 50 . Finally, owing to the heterogeneity of clinical sepsis presentations and based on the increasing relevance attached to genetic variants concerning host septic responses, it is unlikely that all septic patients would benefit to the same extent from receiving SGLT2i inhibitors. Dynamic protocols reflecting the integrity of the microcirculation-i.e., orthogonal polarization spectral imaging 55 -could be useful for early identifying patients that are most likely to respond to this new therapeutic approach.
Given the urgent need for therapies targeting SCLS 18 , along with the emerging cardiovascular protective role of SGLT2i, we strongly believe that these aforementioned findings will likely help better link these two research fields and ultimately provide a promising therapeutic approach for SCLS. It must additionally be mentioned that SGLT2i have been recently approved in other indications than diabetes such as heart failure with reduced ejection fraction, extending their clinical applications and daily uses.

Conclusion
This study highlights endothelial barrier protection by the SGLT2 inhibitor canagliflozin during sepsis, along with α1AMPK/p38MAPK/HSP27/VE-Cad pathway to play a key role in this effect. Canagliflozin could be considered a new therapeutic option in sepsis-induced capillary leak syndrome. www.nature.com/scientificreports/ Mice and breeding. All animals were housed with a 12-h/12-h light/dark cycle, with the dark cycle occurring from 6.00 p.m. to 6.00 a.m. Mice were observed daily with free access to water and standard chow. C57BL/6J males (age 8-12wk) were purchased from the Janvier labs (Le Genest Saint Isle, France). C57BL/6J Cdh5-iCre-ERT2 mice 56 were kindly provided by Ralf Adams, and crossed with mice carrying a floxed allele of PRKAA1 gene (PRKAA1 fl / fl , #014141, the Jackson Laboratory). Cdh5-iCreERT2+ /− PRKAA1 fl / fl mice were administered Tamoxifen (500 μg, intraperitoneally) for five consecutive days at 8 weeks, and used for experiment three weeks after the last Tamoxifen injection. The animals were maintained under a 12:12-h light-dark cycle with free access to food and water.

Lung endothelial cell isolation and model validation.
Mouse lungs were harvested, rinsed and incubated in Dulbecco's modified Eagle's medium containing 2 mg/mL collagenase I for 45 min at 37 °C. The cells were then centrifuged at 1000 g for 5 min at 4 °C, resuspended in buffer 1 (0.1% bovine serum albumin, 2 mM EDTA, in PBS), and incubated with anti-rat immunoglobulin G-coated magnetic beads precoupled with rat anti-mouse PECAM-1 antibody for 30 min at 4 °C in an overhead shaker. Beads were separated from the solution with a magnetic particle concentrator (Dynal MPC-S). The supernatant was kept and the beads were washed five times with buffer 1. Cells-to-CT 1-Step TaqMan kit was used for both the supernatant and the purified endothelial cells, before performing Taqman PCR technology for α1AMPK expression quantification. Data were analyzed with the 2 (-Delta Delta C(T)) method 57 , and expressed as fold of controls.
In vivo model of endotoxemia, cardiac permeability assessment, and plasmatic measurements. Canagliflozin was suspended in saline solution containing 0.5% carboxymethylcellulose and 0,025% Tween-20 and administered by oral gavage (100 mg/Kg, 10µL/g), as described previously 29 . Endotoxemia was induced by intraperitoneal (IP) injections of either LPS (10 mg/Kg) or saline vehicle. For myocardial permeability studies, Evans Blue Dye (EBD) was administered by IP injections (20 mg/Kg) and used to quantify albumin extravasation, as described previously 5 . For heart sampling, animals were euthanized with IP injections of pentobarbital (300 mg/Kg) following 24 h. Vascular leakage, corresponding to the dye amount within the extravascular compartment, was quantified using image J software (Wayne Rasband, National Institutes of Health, Bethesda, MD), as the relative fluorescence (594 nm) surface on frozen Sects. (6-um thick). For blood collection, mice were bled under ketamine and xylazine anesthesia from the retro-orbital plexus. Plasma was obtained by centrifugation at 3000 g for 15 min, followed by 14800 g for 3 min. Albumin was measured by colorimetric method, using FUJI Dry-Chem NX500 biochemical system. For plasma quantification, canagliflozin was analyzed by HPLC-MS/MS system consisting in a Xevo TQ-S mass spectrometer (Waters) coupled to an Acquity UPLC Class H system (Waters). Dapagliflozin was employed as internal standard. The chromatographic separation was performed using a Kinetex C18 HPLC column. Multiple reaction monitoring analysis was performed following electrospray ionization in positive mode.
Human plasma sampling. Patients with septic shock admitted at Cliniques universitaires Saint-Luc, Brussels, were included in the analysis. Septic shock was defined as a sepsis with vasopressor therapy needed to elevate mean arterial blood pressure (MAP) ≥ 65 mmHg, and lactate > 2 mmol/L, despite adequate fluid resuscitation of 30 mL/kg of intravenous crystalloid within 6 h. Patients on therapeutic oral or parenteral anticoagulation therapy (including heparins, fondaparinux, vitamin K antagonist, or novel oral anticoagulants), with previous history of thrombocytopenia (< 100,000 platelets/mm3), recent (less than 1 month) chemotherapy, cirrhosis (Child Pugh > A), or recent (less than 48 h) major surgery, and those patients with active inflammatory disease, hemophilia, or other coagulopathy were excluded from the analysis. The control group comprised healthy volunteers. For the experimental group, blood samples were obtained in the ICU using the routinely inserted central venous catheter, within 48 h of septic shock diagnosis. For the control group, blood samples were collected by venous puncture. Platelet-rich-plasma (PRP) was obtained after centrifugation at 800 g for 5 s, followed by centrifugation at 100 g for 5 min. Next, platelets were pelleted by centrifugation at 400 g for 10 min. Apyrase and Integrilin were added to limit platelet activation during the preparation. In vitro transwell assay. For the endothelial permeability assay, HMECs (10 5 cells/well) were seeded on gelatin-coated Transwell inserts of 24-well plates, in 250μL complete with Endothelial Cells Growth Medium MV. They were then incubated for 72 h at 37 °C and with 5% CO2. The cells were incubated in free M200 medium for two hours before stimulation. The cells were then incubated with the different compounds, as indicated in the figure legends. After treatment, the upper chamber medium was replaced by 300μL of M200, containing HRP-coupled streptavidin. The lower chamber medium was collected after 10-min incubation at 37 °C, and every condition was aliquoted in triplicate. The TMB substrate was added for 10 min, and 2 N H 2 SO 4 was applied to stop the reaction before acquiring 450 nm absorption in an Elisa reader. Resultant absorption intensity values were normalized over the vehicle control condition. Each experiment was repeated three times.
In vitro immunofluorescence staining and image analysis. HMECs were seeded on non-coated glass coverslips at a density of 20 × 10 3 cells/cm 2 , 72 h before treatment. After treatment, cells were fixed in 4% paraformaldehyde, permeabilized with 0.3% triton X-100 for 10 min, and then blocked with 10% BSA for 45 min. Cells were then stained as previously described 5 , using VE-Cad primary antibodies (1:25) and Alexa Fluor-coupled secondary antibodies (1:1000). Nuclei were stained using 4' ,6-diamidino-2-phenylindole (DAPI). Stainings were visualized under a Zeiss Imager Z1 microscope that was equipped with an ApoTome device. Pictures were acquired using an ×20 objective. Each experiment was repeated three times. Quantitative image analysis was performed on uncompressed images (native format: zvi) with Fiji 1.52n on MacOS (10.14.5). One image was analyzed per condition and for each experiment. Intercellular junctions, evidenced by VE-Cad staining, were automatically delimited using a fixed-value threshold method. The stained area was quantified, and the mean signal intensity was calculated with this section. Stained membrane segments were subsequently detected using the Analyze Particles ad Skeletonize tools, and automatically counted. For normalization purposes, all images' nuclei were automatically counted using a threshold method and the analyze particles tool.

Ethics approval and consent to participate. The study was approved in 2018 by the Ethical Review
Board of Cliniques universitaires Saint-Luc/UCLouvain (V1 04/12/2018). All methods were carried out in accordance with relevant guidelines and regulations. All participants provided written informed consent.
Animal handling and experimental procedures were approved by local authorities at UCLouvain (Comité d' éthique facultaire pour l' expérimentation animale, 2016/UCL/MD/027) and performed in accordance with the Guide for the Care and Use of Laboratory Animals, published by the US National Institutes of Health (NIH Publication, revised 2011). All the authors complied with the ARRIVE guidelines.
Statistical analyses. The sample size was not pre-determined based on statistical analysis, and it was chosen according to previous publications. Statistical analyses were conducted using SPSS v.25 Software (IBM Corp., Armonk, NY, USA), and graphs were build using GraphPad Prism 7.0 (GraphPad Software, La Jolla, CA, USA). All tests were two-sided, with statistical significance set at the 0.05 probability level. Data were expressed as mean ± standard deviation. Means were compared using unpaired Student's t-test or a one-way or two-way analysis of variance, as appropriate. The Bonferroni correction was applied for multiple comparisons.

Data availability
The datasets used during the current study are available from the corresponding author on reasonable request.