Role of endothelial microRNA 155 on capillary leakage in systemic inflammation

All chemical and reagents, unless otherwise specified, were purchased from Sigma-Aldrich. Antibodies against Claudin-1, ZO-1, GAPDH and Alexa Fluor 546 Phalloidin, 4′,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich, St.Louis, MO) and Aqua-Poly/Mount were used.

RNA sequencing library was generated from 0.1 to 1 µg of total RNA using TruSeq® Small RNA Library Prep Kits v2 (Illumina) according to manufacturer’s protocols. The libraries were sequenced on Illumina HiSeq2500 using TruSeq SBS Kit v3-HS (50 cycles, single ended run) with an average of 10 × 106 reads per RNA sample. FASTQ files were trimmed with fastq-mcf (ea-utils, https://​expressionanalys​is.​github.​io/​ea-utils/​) removing Illumina RNA adapter sequences (TGGAATTCTCGGGTGCCAAGG) and nucleotides with phred scores below 20. [7‐10]

Human umbilical vein endothelial cells (HUVECs) were isolated from human umbilical veins (donor approval, Hannover Medical School Nr. 1303-2012). Specific miRCURY LNA miRNA mimic (No. 339173) (Qiagen, Hilden, Germany) was used to overexpress the microRNA 155. To inhibit the microRNA 155, the miRCURY LNA miRNA Inhibitor 5 nM was used (No. 339121) (Qiagen, Hilden, Germany).

All mouse experiments were approved by the local authorities at Hannover Medical School and conducted in accordance with institutional and governmental guidelines (LAVES Lower Saxony, Ref. No. 18/2817). Male B6.Cg-Mir155tm1.1Rsky/J were used as MIR155 knockout mouse. The mice, 10–12 weeks of age, were either injected intraperitoneally (i.p.) with 17.5 mg/kg bodyweight lipopolysaccharide (LPS) or a cecal ligature puncture (CLP) as described elsewhere [11] was conducted. Briefly, the LPS originated from Escherichia coli, serotype O111:B4 (Sigma-Aldrich, St. Louis, MO) dissolved in 10 mL/kg sterile NaCl. The injected mice were monitored and scored according to Additional file 1: Table S4. After 16 h, they were killed and organs harvested. For the performance of the CLP experiment, the mice were anaesthetized with isoflurane (1–2% in medical air) and a midline laparotomy was placed by a single-blinded operator. The anti-mesenteric border was ligated, and through one single 20G needle puncture 1 mm of stool was extracted. As a sham surgery, laparotomy with cecal mobilization was performed. Followed by the two-layering closure, mice were given 200 µL of NaCl fluids s.c. and for analgesia once 10 mg/mL Butorphanol (Zoetis Manufacturing & Research, Spain). Until the organ harvest 24 h later, the mice were monitored and scored by a single-blinded investigator. The Evans Blue permeability assay (EB) was conducted as described [11]. In brief, the mice were injected with the 100 µL of 2% wt/vol EB in the tail vein 12 h after LPS injection. For lung cuffing, mice were challenged with LPS (17.5 mg/kg BW) and killed 16 h later and organs were harvested. After the staining process, a lung cuffing score and a percentage of cuffing (yes / no) were raised by a single-blinded person (see Additional file 1: figure S1 and table S2). For Kaplan–Meier survival studies, both models, i.e., LPS and CLP were used. All experiments were performed and analyzed by a blinded investigator.

The Western Blot Analysis was conducted as in [12] described using the SuperSignal™ West Pico Chemiluminescent Substrate (Life Technologies) and Versa Doc Imaging System Model 3000 (BioRad, Hercules, CA) was used to visualize the bands.

HUVECs were grown to confluency on Collagen (Sigma-Aldrich, St.Louis, MO) covered coverslips. Antibodies and reagents were used as described above.

Paraffin slices were deparaffined and blocked with 10% donkey serum (Jackson Immuno Research Inc., West Grove, PA). GR-1 was used as primary antibody and pictured with the secondary antibody (goat anti-rat IgG (Alexa flour 555)) (Invitrogen, CA). To evaluate the amount of lung cuffing through scores, mice lungs were stained with Periodic Acid Schiff (PAS) Staining and scored according to Additional file 1: Table S2. The images were taken with the Leica DMI 6000B microscope under the same gain and offset conditions.

For the isolation, the miRNeasy Mini Kit was used (Qiagen, Hilden, Germany) to extract both miRNA and RNA from HUVECs and organ tissue followed by Prime Script RT Reagent Kit (TaKaRa Bio Europe SAS, St Germain-en-Laye, France) after the manufacturer’s instructions.

HUVECs were plated at 1.8 × 105 cells/well in 6 well-plate overnight and transfected with 200 pM biotinylated control miRNA mimics or 5′-Biotinylated MIR-155 mimics (No. 339178) (Qiagen, Hilden, Germany) using HiPerfect (Qiagen, Hilden, Germany) in accordance with the manufacturer’s instructions. After 48 h of transfection, cells were harvested in 700 µL lysis buffer supplemented with Protease Inhibitor Cocktail (Roche, South San Francisco, CA) and RNaseOUT (Invitrogen) and incubated on ice for 20 min. Lysates were centrifuged at 10,000 × g for 15 min at 4 °C after which 50 µl of the cytoplasmic lysate (input) was transferred into a new tube for RNA extraction. The remaining supernatant was incubated with activated Streptavidin-Dynabeads (Dynabeads M-280 Streptavidin, Invitrogen) for 4 h at 4 °C. After several washing steps and centrifugation, the supernatant was taken for RNA extraction. RNA was subjected to qPCR using Claudin-1 specific primers. The analysis was done as follows: MIRNA pull-down/control pull-down (‘A’), miRNA input/control input (‘B’); fold enrichment = A/B.

Female Tg(flk1:mCherry) zebrafish were mated with male Tg(l-fabp:eGFP-DPB) to generate Tg(flk1:mCherry/l-fabp:eGFP-DPB) zebrafish offspring. Eggs were microinjected with a MIR155 mimic or a scrambled MIR (mirVana, life technologies) at the one to four cell stage at a concentration of 25 μM [13]. At 96 h post-fertilization, the vascular integrity of the transgenic larvae based on flk1 expression in ECs was determined. In parallel, the plasma protein loss of eGFP-DBP was analyzed by measurement of fluorescence intensity in the retinal vessel plexus [13]. The zebrafish animal studies were performed according to the National Institutes of Health Guideline for the Care and Use of Laboratory Animals. The Mount Desert Island Biological Laboratory (Bar Harbor, ME) animal care committee approved the animal protocol (IACUC protocol #1703). The analysis was performed using ImageJ (Version 1.60, National Institutes of Health, Bethesda, MD) and reported in arbitrary units.

TER was measured using an electric cell-substrate impedance sensing system (ECIS) (Applied BioPhysics Inc.). The continuous values were pooled at discrete time points and plotted versus time. Each conditions’ end point resistance was divided by its starting resistance to give the normalized TER [14].

Kidney biopsies were obtained directly post-mortem from patients aged 18 years or older, who died of sepsis. Kidneys from patients diagnosed with kidney cancer who underwent a total nephrectomy, served as controls as described in Aslan et al. Crit. Care. (2014) [15]. The postmortem biopsies were waived by the Medical Ethical Committee of the UMCG, Groningen, The Netherlands (METc 2011/372) [15].

MIR155 expression was measured in serum and BALf samples collected within 24 h of disease onset in patients with ARDS (n = 16), with clinical details provided in Table 1 and in healthy controls (n = 5). The collection of the BALf and serum samples was approved according to the ethics committee of Hannover Medical School (MHH, EK 8146_BO_K_2018). Total RNA was isolated from human BALf and serum samples using miRNeasy Serum/Plasma Advanced Kit (Qiagen, Cat. No. 217204) following manufacturer’s instructions. For the BALf analysis, we established a novel method base on the above kit used for the serum. Cel_MIR-39 miRNA mimic (Qiagen, Cat. No. 219610) at a concentration of 1.6 × 108 copies/µL was used as a spike-in miRNA control. The reverse transcription for miRNAs was done with equal volume of starting total RNA for each sample and specific TaqMan probes (Applied Biosystems; for MIR-155: Assay ID 002623 and for Cel_MIR-39: Assay ID 000200) using TaqMan MicroRNA Reverse Transcription Kit (Applied Biosystems, Cat. No. 4366597) as per the manufacturer’s guidelines. ViiA7 system (Applied Biosystems) was used to perform miRNAs’ quantification PCR (qPCR) with MIR-155 or Cel_MIR-39 specific TaqMan assays (Applied Biosystems) and Absolute Blue qPCR Mix (Abgene, Cat. No. AB-4136/B). The collection of the BALf and serum samples was approved according to the ethics committee of Hannover Medical School (MHH, EK 8146_BO_K_2018).

Statistical significance was evaluated using Mann–Whitney test or one-way ANOVA unless otherwise noted. All experimental results are presented as mean ± SD, and a two-tailed p value of less than 0.05 was considered to indicate statistical significance. GraphPad Prism 6.0 (La Jolla, CA).

Mice were challenged intraperitoneally with a high dosage of LPS (17.5 mg/kg bodyweight, n = 3) or vehicle (n = 3) to induce a systemic inflammatory response that regularly leads to severe pulmonary capillary leakage. Mice were killed after 24 h, and lungs were harvested for endothelial cell (EC) separation with a magnetic CD146 antibody strategy followed by isolation of MIRs. SmallRNA sequencing was performed together with the Helmholtz Institute (Brunswick, Germany). A heatmap of regulated MIRs of the 6 included animals is shown in Fig. 1a and Additional file 1: Table S1. In summary, out of 318 investigated MIRs, 5 were down- and 31 up-regulated (Fig. 1b and Additional file 1: Table S2). Of note, MIR155—that had not been studied in the septic endothelium so far—showed the strongest signal in our analysis and was therefore followed-up on in the next experiments.

To confirm these earlier findings of the smallRNA seq, MIR155 was first quantified by RT-PCR from whole lung lysates in endotoxemic and control mice at different time-points. Already after 4 h, we found a 6 × fold increase in the LPS-treated mice (Fig. 1c, p < 0.01). Given that MIR155 is also expressed by other cells types (e.g., epithelial and immune cells), we selected specifically CD146⁺ pulmonary ECs, which showed an even larger upregulation (i.e., 8 × fold MIR155 increase, Fig. 1d, p < 0.01). To confirm these results in a cellular context and to set a system for further mechanistic in vitro studies, confluent HUVEC monolayers were stimulated with 50 ng/mL TNFα (a key cytokine of systemic inflammation) and harvested in a time-dependent fashion. Indeed, TNFα challenge of ECs was sufficient to induce a significant upregulation of the MIR155 already after 2 h (p < 0.01) with a maximum response exceeding 15 × fold after 24 h (p < 0.001, Fig. 1e).

Given the lack of knowledge on the functional role of MIR155 in the endothelium, we investigated it in a transgenic Tg(flk1:mCherry)/Tg(l-fabp:eGFP-DBP) zebrafish line. This zebrafish model allowed us (1) to rapidly overexpress MIR155 by microinjection of a MIR155 mimic and (2) to analyze vascular phenotypes. Simplified, this transgenic fish expresses a red cherry protein in the vasculature and has an enhanced green fluorescent protein bound to Vitamin D binding protein (eGFP-DBP) (corresponds to the intravascular compartment, i.e., the plasma). Both loss of GFP intensity in the retinal vessels and the development of pericardial effusion can be used as a surrogate of vascular leak [13, 16]. Indeed, overexpression of MIR155 led to a visually detectable loss of GFP in the red vascular tree and fluid accumulation in the pericardium (Fig. 2a–c). Visualization and quantification of the retinal GFP expression showed a decrease in fluorescence intensity (p < 0.0001) supporting the hypothesis that experimental overexpression of MIR155 is sufficient to induce vascular permeability (even in the absence of systemic inflammation) (Fig. 2d–f).

To assess the role of the MIR155 in mammals, knockout mice and littermate wildtypes were challenged with LPS or vehicle and permeability was assessed in vivo by the Evans Blue (EB) method. Septic wildtypes showed a 2.2 × fold increase in lung EB extravasation quantified by spectroscopic analysis of the grounded tissue. Interestingly, homozygote knockouts (MIR155^−/−) were not protected, but the heterozygote MIR155 mice (MIR155^+/−) showed barrier properties that were indistinguishable from healthy controls (Fig. 3a). Bronchiolar cuffing is a histological method to visualize pulmonary edema (Additional file 1: Fig. S1). Again, septic wildtype mice showed a severe cuffing phenotype (Fig. 3b, second panel) that was less severe in the heterozygote context (third panel) but not different in the MIR155^−/− mice (fourth panel). This finding was semi-quantified by a blinded investigator both with regard to the percentage of cuffing positive bronchi (Fig. 3c) and the severity of cuffing (Fig. 3d). In addition to these observations with regards to permeability, we found additional evidence that the endothelial inflammatory response might be positively influenced if MIR155 is experimentally reduced (Additional file 1: Fig. S2). Together these findings support that MIR155 upregulation in systemic inflammation might be an injurious contributor to vascular leakage across different species.

In vitro, naïve or MIR155 modulated HUVECs were challenged with thrombin (a mediator of systemic inflammation and an established inductor of EC permeability in vitro) to recapitulate the above described in vivo findings. Thrombin stimulation primed the formation of F-actin polymerization and gap formation between adjacent ECs leading to an increase in permeability (white arrows) (Fig. 4a, lower left). ECs that have been challenged with a MIR mimetic (w/o thrombin) partly phenocopied these permeability patterns (Fig. 4a, upper right). Interestingly, from a therapeutic point of view, ECs that have been co-stimulated with thrombin + anti-MIR155 peptide were protected from the development of the before described endothelial morphological changes (Fig. 4a, lower middle panel). To reliably quantify these qualitative changes in morphology, we measured TER in real-time with Electric Cell-substrate Impedance Sensing (ECIS). MIR155 overexpression was not sufficient to induce leak in this assay by itself, but MIR155 overexpression critically interfered with the usual recovery after stimulation with thrombin (Fig. 4b). More important from a translational aspect, inhibition of MIR155 (anti-MIR155) was sufficient to reduce the deleterious effect of thrombin in this highly sensitive assay (Fig. 4c, p < 0.05). Together these findings support our hypothesis that upregulation of MIR155 in the septic endothelium is indeed injurious and that its inhibition might represent a therapeutic target.

Using an in silico analysis tool (http://​www.​targetscan.​org), we identified 556 potential transcripts with preserved sites among various species that might be influenced by MIR155. Given the permeability phenotype in our in vivo and in vitro models, we specifically looked at proteins involved in the junctional apparatus. The analysis predicted that Claudin-1 could be a possible target that additionally showed conserved binding sites among different species (e.g., human, cow, rat, mouse, dog). First, we checked if Claudin-1 expression is indeed altered in LPS-treated mice and found a significant reduction in their lungs (Fig. 4d). Experimental up-regulation of MIR155 in ECs in vitro to a similar extent that we earlier detected in systemic inflamed mice (Additional file 1: Fig. S3) induced a significant downregulation of Claudin-1 on mRNA and protein levels (Fig. 4e, f). Using an immunoprecipitation strategy from ECs stimulated with an MIR155 mimetic, we could confirm the physical binding of MIR155 with claudin-1 (Fig. 4g). However, inhibition of MIR155 was not sufficient to increase Claudin-1 expression spontaneously (data not shown).

Next, we tested if the previously observed attenuation of permeability in the MIR155 heterozygote mice might be sufficient to protect from organ dysfunction and death. Analysis of functional organ parameters from serum samples 24 h after LPS administration in all genotypes showed trends towards improved organ function and survival in a 100% lethal model (Additional file 1: Table S3 and Fig. S3.) in the heterozygous MIR155^+/− mice (median survival MIR155^+/− 25.5 h vs MIR155^+/+ 36 h). The 100% lethality of our endotoxemia model combined with increasing doubts regarding the clinical relevance of this model in the sepsis community [17] inspired us to test outcome in a clinically more meaningful polymicrobial sepsis model, i.e., CLP. Upon CLP surgery, mice were regularly scored for severity of disease by a single-blinded investigator using a standardized activity score (Additional file 1: Table S4). Consistent with the permeability data, MIR155^+/− mice had a better physical performance indicating a less severe disease manifestation than all other groups over a 96 h observation period (Fig. 5a) and indeed showed a clear survival benefit by approximately 60% compared to the MIR155^+/+ mice. (Mantel–Cox Test, p < 0.05, Fig. 5b).

Given the discovery of both cross-species regulation and relevance of MIR155 in increased endothelial permeability, we set out to investigate its regulation in a human organism. To do this, we followed two strategies. First, we analyzed BALf and serum from critically ill patients suffering from ARDS—a syndrome that is pathophysiologically based on increased vascular permeability. Patients characteristics are summarized in Table 1. To our surprise, circulating MIR155 levels were not different between controls and ARDS patients (Fig. 6a). Given that we identified MIR155 with the RNA seq specifically in the pulmonary endothelium, we analyzed local MIR155 abundance in the alveoli (i.e., BALf) of healthy controls and ARDS patients. The healthy volunteers (n = 5) had a median age of 26.9 years [IQR 22.5–33.1] and were all male. Consistent with our previous findings, we found a massive MIR155 upregulation in ARDS patients in this local compartment, close to the pulmonary endothelium (Fig. 6b). To test MIR155 regulation in other organs frequently affected during sepsis, we analyzed immediate post-mortem kidney biopsies from patients with septic acute kidney injury compared to healthy parts of tumor nephrectomy samples. Again, we could confirm that MIR155 was significantly upregulated in human septic kidneys (Fig. 6c). Together, these findings highlight the translational relevance of our results.