Initial population included 204 paired sera from 102 patients presenting anaphylaxis and recruited from three Spanish hospitals (Fundación Jiménez Díaz University Hospital, Cruz Roja Central Hospital and Niño Jesús University Children’s Hospital). However, 35 of them were discarded due to the hemolysis of their samples.

Anaphylaxis diagnosis was confirmed by an allergist in agreement with the definition of anaphylaxis established by the “National Institute of Allergy and Infectious Disease and Food Allergy and Anaphylaxis Network” (19). In addition, from each patient were recorder their gender, age, the trigger of the reaction and if they were treated before or after obtaining the acute sample. Furthermore, their clinical signs and symptoms were also collected and, based on these, the severity of the reaction was determined according to the criteria established by Brown (20).

The study was approved by the Ethics Committee (CEIm FJD, PIC057-19), authors adhered to the declaration of Helsinki and all patients were included after giving informed consent by the donors or their relatives. Inclusion criteria were acceptance to participate in the study and an objective diagnosis of the reaction. Exclusion criteria were the presence of a blood-borne disease and/or any psychic or psychological diseases that would prevent acceptance for the study.

Serum samples were collected from each patient under two conditions: during acute phase (anaphylaxis) and at basal phase (control), at least 14 days after the reaction. Considering the heterogeneity of participants, the acute sample data was normalized by calculating the increase or decrease ratio upon baseline individual sample values (ratio acute/basal).

All samples were obtained from patients who developed anaphylaxis accidentally. If the allergic reaction occurred accidentally, acute phases were collected in emergency departments. On the other hand, if the hypersensitivity reaction was provoked by a controlled challenge test, sera were obtained in allergy units.

For serum processing, tubes with a separator gel (BD Vacutainer) were used. Once the peripheral blood was obtained, it was centrifuged at 1,200 g for 10 minutes at 4°C. Afterwards, the serum was aliquoted in 0.5 ml and stored at -80°C until its use in each of the experiments.

The serum circulating miRNAs profile was determined by Next Generation Sequencing (NGS) at Qiagen Genomic Services, as previously described by Nuñez-Borque et al. (17, 21). For the analysis, acute and baseline samples from 5 adult patients with drug anaphylaxis were tested in the same batch.

RNA was extracted from 200 μl of serum using the miRNeasy Serum/Plasma Kit (Qiagen). Moreover, 52 exogenous sequences were included to confirm the correct isolation of miRNAs. In turn, libraries were performed with the QIAseq miRNA Library Kit (Qiagen). For this purpose, RNA was retrotranscribed to copy DNA (cDNA), amplified by PCR (22 cycles) and purified from samples. Libraries were carried out by a gel-free system through the addition of Unique Molecular Identifiers (UMIs) and were specific for small RNA species ranging from 15-40 nucleotides. Quality controls were determined using the Bioanalyzer 2100 or the TapeStation 4200 (Agilent Technologies). In turn, miRNAs data were normalized to equimolar ratios depending on the quality of the library, the quality of the inserts and the concentration measurements. Finally, libraries were quantified by quantitative PCR (qPCR) and sequenced on an Illumina NextSeq500 instrument. Raw data files (FASTQ) were generated for each sample using the bcl2fastq program (Illumina Inc.) and Cutadapt (1.11) software was utilized to remove adapter sequences and collapse the reads by UMIs. Readings were mapped against the miRbase_20 database using the Bowtie2 program (2.2.2) and no errors or more than 1 mismatch were allowed.

To the validation step, paired samples from each patient were always analyzed simultaneously. Serum miRNAs were extracted using the miRNeasy Serum/Plasma Advanced kit (Qiagen). For each sample, 300 μl of serum were used and added 90 μl of lysis buffer and 1 μl of UniSp2/4/5 mix (miRCURY LNA RNA Spike-in kit, Qiagen), synthetic miRNAs that serve as a quality control of the technique. Following this, samples were incubated for 3 minutes at room temperature, 30 μl of precipitation buffer were added, and a new 3-minute incubation at room temperature was performed. Tubes were centrifuged at 12,000 g for 3 minutes, supernatants were mixed with an equal volume of isopropanol and transferred to a specific column where RNA was retained. Subsequently, three washes were performed and 15 μl of RNAase-free water was added to the column, which was incubated for 1 minute at room temperature and centrifuged for 1 minute to elute the isolated RNA.

Total RNA was retrotranscribed using the miRCURY LNA RT kit (Qiagen). For this purpose, 2 μl of the reaction-specific buffer, 4.5 μl of RNAase-free water, 1 μl of the enzyme mixture, 2 μl of the RNA previously obtained in the extraction and 0.5 μl of UniSp6, a synthetic miRNA used as a quality control of the reverse transcription, were included for each sample. Once the mix was done, samples were placed in the PTC-100 Thermal Cycler (MJ research, Inc) where retrotranscription was carried out.

Finally, miRNAs quantification (miRCURY LNA SYBR Green PCR kit, Qiagen) was performed by quantitative PCR (qPCR) with specific primers (Qiagen) in the LightCycle96 Real Time PCR System (Roche Life Science). Samples from each patient were always measured at the same time and in the same batch. UniSp2, UniSp4, UniSp5 and UniSp6 were used as exogenous controls of the extraction and reverse transcription, while miR-451a and miR-23a-3p were measured as endogenous controls of sample hemolysis (22, 23). Moreover, miR-30e-5p was chosen as endogenous control for the normalization of the target miRNAs, as suggested by the NGS according to the levels of all molecules identified. Data obtained were analyzed using the 2^-ΔΔCT method (24).

The purification of EVs from serum was performed using the miRCURY Exosome isolation Kit – Serum and Plasma (Qiagen) according to Colletti et al. (25). For this purpose, 500 μl of serum were centrifuged at 12,000 g for 20 minutes at 10°C. Subsequently, 200 μl of the precipitate buffer were added to the supernatant and the mix was incubated for 3 hours at 4°C. After this time, samples were centrifuged at 1,500 g for 30 minutes and supernatants were removed. Finally, EVs were collected adding 270 μl of the resuspension buffer.

Suitable isolation and characterization of EVs was determined by Western blot (WB), nanoparticle tracking analysis (NTA) and electron microscopy (EM) as previously described (26) and following MISEV2018 guidelines (27). Subsequently, the extraction, retrotranscription and quantification of miRNAs contained in EVs were carried out as previously described in the section “Serum miRNAs extraction, retrotranscription and quantification”. Paired samples from each patient were always analyzed simultaneously.

Three bona fide markers (CD9, CD63 and TSG101) of EVs were characterized by WB. Proteins were separated by electrophoresis on sodium dodecyl sulfate polyacrylamide gels (SDS-PAGE) at 12.5% using the PowerPac Basic Power Supply section (BioRad). Subsequently, proteins were transferred to a nitrocellulose membrane (BioRad) using the Trans-Blot Turbo Transfer system (BioRad). Non-specific membrane junctions were blocked by applying a solution of phosphate buffered saline (PBS) with 0.1% Tween (Sigma) and 5% milk. After this, the membrane was incubated overnight with the corresponding primary antibodies diluted 1:500 in all cases: anti-CD9 (Invitrogen), anti-CD63 (Invitrogen) and anti-TSG101 (Abcam). For signal detection, the membrane was incubated for 1 hour with a rabbit anti-mouse (RAM) peroxidase-conjugated secondary antibody (Jackson Laboratory) diluted at 1:5,000 in 0.05% PBS-Tween solution with 3% milk. Lastly, proteins were detected in the Amersham Imager 600 system (GE Healthcare) using the chemiluminescent ECL Prime Western Blotting Detection Reagent (Thermo Scientific).

For the EVs visualization by EM, 5 μl of the previously purified particles were fixed for 24 hours at 4°C using 50 μl of PBS with 2% paraformaldehyde (PFA, Sigma). After this, EVs were adsorbed on copper/carbon grids (Fedelco) for 3 minutes, washed with water for 1 minute and counterstained with 1% uranyl acetate for 30 seconds. Then, samples were observed using a JEOL 1010 transmission EM (JEOL) and employing a voltage of 100 kV. The analysis and image processing were carried out with the Soft Imaging Viewer software.

NTA was performed at the Spanish National Cancer Research Center (CNIO, Madrid) with the NS500 nanoparticle characterization system (NanoSight) equipped with a blue laser (405 nm) and diluting samples 1:500.

Systems Biology Analysis (SBA) of miR-375-3p was performed in silico using all its target genes with more than 50 target score (269 genes) in the miRDB database (http://mirdb.org/) (28). All genes were mapped against their corresponding gene ontology (GO) in the Ingenuity Pathway Knowledge Base using the Ingenuity Pathway Analysis (IPA) software (Qiagen). This program provided different functional categories: “ingenuity canonical pathways”, “upstream regulators”, “disease and disorders”, “network” and “tox lists”. In addition, it evaluated the significance using a one-tailed Fisher’s test. Among the different functional categories obtained, “disease and disorders” and “ingenuity canonical pathways” were the only one considered.

Serum cytokine levels were determined using the multiplex assay RayPlex Human Cytokine Storm Array 1 Kit (RayBiotech). For this purpose, acute and basal samples from 40 patients with anaphylaxis were studied. Dilution of the samples and standards was performed according to the manufacture indications. A panel of 25 mediators involved in the inflammatory process was tested: basic Fibroblast growth factor (bFGF), Eotaxin, Granulocyte colony-stimulating factor (G-CSF), Granulocyte macrophage colony-stimulating factor (GM-CSF), Interferon-γ (IFNγ), Interleucin-10 (IL-10), IL-12p70, IL-13, IL-15, IL-17A, IL-1β, IL-1RA, IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, Monocyte chemoattractant protein-1 (MCP-1), Macrophage inflammatory protein-1α (MIP-1α), MIP-1β, Platelet-derived growth factor-BB (PDGF-BB), RANTES, Tumor necrosis factor-α (TNFα), Vascular endothelial growth factor (VEGF). Mean fluorescent intensities of the samples and standards were determined in a paired manner using a BD FACSCanto II flow cytometer (Becton Dickinson).

Human dermal microvascular ECs (HMVEC-D) were acquired from Lonza (CC-2543) and maintained in EGM-2MV BulletKit medium, as previously described (29). First, HMVEC-D were seeded and allowed to grow for 24 hours to reach 70-80% confluence. Next day, transfection was conducted incubating cells with Opti-MEM medium (Thermo Scientific) and using Transit-X2 reagent and mimics (50 mM) of the miR-375-3p (Qiagen). In addition, fluorescently labeled mimic scrambles (50 mM) were applied as controls of the technique (Qiagen). Subsequently, after 72 hours, molecular and functional studies were performed in HMVEC-D. In turn, the correct development of the transfection was confirmed by qPCR and confocal microscopy.

Extraction of miRNAs from HMVEC-D was carried out using the MasterPure Complete DNA & RNA Purification Kit (Lucigen). Cells were lysed by adding 300 μl of lysis buffer and 1 μl of Proteinase K. The lysate obtained was collected and incubated at 65°C for 15 minutes. Then, samples were incubated at 4°C for 5 minutes and 150 μl of precipitation buffer were added. Subsequently, each sample was centrifuged at 10,000 g for 10 minutes and supernatants were collected. An equal volume of isopropanol was added, and mixtures were centrifuged again. Next, to remove the remaining DNA, the precipitate was resuspended in 200 μl of the DNA nuclease buffer and 5 μl of DNA nuclease I. Consecutively, samples were incubated at 37°C for 30 minutes and 200 μl of the lysis and precipitation buffers were added. Then, samples were incubated at 4°C for 5 minutes and centrifuged at 10,000 g for 10 minutes at 4°C. After this, an equal volume of isopropanol was added, and samples were centrifuged again. Finally, the RNA extracted was resuspended in 20 μl of trisaminomethane buffer (TE) and 1 μl of RNAse inhibitor. Finally, reverse transcription and miRNAs quantification were performed as previously described in the section “Serum miRNAs extraction, retrotranscription and quantification”.

Cells were seeded and transfected in P6 plates (Corning) as previously described in the “Transfection of miRNA” section. 30 µM cyclic AMP (cAMP) was used to stabilize the endothelial barrier during 30 minutes. Once the incubation time was over, proteins were extracted by adding 100 μl of lysis buffer supplemented with 5 mM protease inhibitor (Sigma), 1 mM phenylmethylsulfonyl fluoride (PMSF, Sigma) and 5 mM dithiothreitol (DTT, BioRad). Cells were detached using a scraper and the lysates obtained were kept in agitation for 15 minutes at 4°C. After this, they were centrifuged for 15 minutes at 12,000 rpm at 4°C and the supernatants, which contained the proteins, were stored at -20°C until use. Rac1-CDC42 levels were analyzed by WB following the protocol previously described in the “Western Blot” section. Membranes were incubated overnight with primary antibody anti-Rac1-Cdc42 (Cell Signaling) diluted 1:500. For signal detection, the membrane was incubated for one hour with a goat anti-rabbit (GAR) peroxidase-conjugated secondary antibody (Jackson Laboratory) diluted at 1:5,000 in 0.05% PBS-Tween soluction with 3% milk. Likewise, an anti-β-actin antibody (Santa Cruz) was used to detect β-actin as a housekeeping protein diluted at 1:500 while the RAM secondary antibody was diluted 1:5,000. Bands intensity was quantified using the ImageJ software and normalization to the amount of total protein loaded was calculated using the ratio of the Rac1-Cdc42/β-actin signal.

The endothelial barrier integrity was assessed by measuring the trans-endothelial electrical resistance (TEER) value of the monolayer through an EndOhm chamber (WPI) using 24-well Transwell (TWs) cell culture inserts (26, 29). For determining the effect of miR-375-3p in a cAMP-stabilized endothelial system, HMVEC-Ds were transfected on TW. Four different HMVEC-D systems were registered and analyzed: non transfected, incubated with Transit-X2 alone, transfected with the scramble, as control, and transfected with the miR-375-3p mimic (50 mM). All these systems were evaluated in presence of cAMP and TEER measures were obtained at different times (10, 30, 45, 60, 90, 120 and 180 minutes).

For confocal microscopy HMVEC-D were seeded in P8 chambers (IBIDI). Cells were fixed with 4% PFA for 15 minutes and permeabilized with 0.1% Triton (Sigma) for 5 minutes at 4°C. After this, non-specific junctions were blocked by adding a PBS solution with 2% BSA for 20 minutes. Cells were incubated overnight with an anti-VE-cadherin antibody (Cell Signaling) diluted 1:400 in a solution of PBS with 2% BSA. The next day, cells were incubated for 1 hour with a donkey anti-rabbit Alexa Fluor 647-conjugated secondary antibody (Invitrogen). Then, actin filaments were red stained by incubating HMVEC-D for 30 minutes with Texas-Red-X phalloidin reagent (Invitrogen) at a concentration of 1:40 in a solution of PBS with 2% BSA. Subsequently, a second 10-minutes incubation was carried out with a solution of DAPI (Sigma) at a concentration of 1:1,000 in distilled water to stain the nuclei blue. Lastly, the slides were assembled using the FlourSave Reagent (Calbiochem) and visualized on the LSM 700 inverted confocal microscope (Carl Zeiss). Transfection was confirmed by visualization of the fluorescent scramble (green) inside ECs.

Data obtained from NGS were normalized and analyzed with the statistic software R 3.5.3. The principal component analysis (PCA) and the heat map were carried out on the ClustVis website (https://biit.cs.ut.ee/clustvis/) (30). Graphical representation and statistical evaluation were performed by using the Graph Pad Prism 8 software (La Jolla, CA, USA).

Qualitative variables were described as frequency and percentage. However, in quantitative variables, data distribution was always first assessed by normality tests. When data presented a normal distribution, parametric analyses were used, and the graphs were represented by the mean ± standard error of the mean (SEM). On the other hand, when the data did not follow a normal distribution, non-parametric tests were used, and the graphs showed the median ± interquartile range (IQR). Differences were considered significant at level of p<.05.

NGS analysis were performed with the Prostar package (http://live.prostar-proteomics.org/) distributed by Bioconductor and implemented in R (R Core Team, 2019). Abundance data were transformed (log₂) using R 3.5.3 software to obtain a symmetrical distribution. Moreover, all miRNAs not detected in at least two patients were discarded. The matrices were normalized with the Cyclic Loess method to reduce systemic variances (31). Finally, Student’s paired t-test was used to determine the significant differences between both conditions (acute vs basal) (32).

For the validation stage and measurements of miRNAs within EVs, data did not follow a normal distribution, so the two-tailed Wilcoxon matched-pairs non-parametric test was used to determine the significant differences between acute and basal phases. In turn, for the characterization of EVs purified from serum, non-parametric tests were used and both size and concentration were analysed with the Wilcoxon matched-pairs test. On the other hand, the study of serum cytokines was carried out using the Wilcoxon matched-pairs test in all cases. In turn, comparison of their levels with those of miR-375-3p was performed using Spearman’s non-parametric correlation test. Moreover, data obtained from in vitro assays followed a normal distribution and, consequently, analyses were performed using the paired t-test or 2 way ANOVA.

The studied cohort included 67 patients with anaphylaxis recruited from different Spanish hospitals ( Supplemental Figure 1 ). Their clinical characteristics and those of their reactions are shown in Table 1 . This population aged between 4 and 76 years old and more than half were females. The most frequent symptoms were cutaneous and respiratory, followed by gastrointestinal, mucosal, cardiovascular and central nervous system manifestations. Overall, two thirds of the reactions were classified as moderate (Grade 2), while the remaining ones were categorized as severe (Grade 3). Outstandingly, the frequency of Grade 3 cases was lower in children with food-induced reactions compared to any of the two adult populations (11% vs 43% and 46%, respectively). Finally, even though patient health was always prioritized, a percentage of acute samples were obtained immediately before treatment administration. In these subjects, the most used medications were epinephrine, histamine receptor 1 antagonists and corticosteroids.

The profile of circulating miRNAs was determined by NGS in an exploratory study carried out in samples from patients with drug anaphylaxis. For this purpose, acute and baseline serum samples were used. The similarity among the biological replicates was verified by the PCA, which exhibited a clear separation between the two stages ( Figure 1A ). In turn, the correct performance of NGS was confirmed through the analysis of different quality controls ( Supplementary Figure 2 ).

A total of 334 miRNAs were identified ( Supplementary Table 1 ) and after statistical analysis only 21 of them showed significant differences between acute and basal phases ( Figure 1B ; Supplementary Table 2 ). Each sample´s signals dispersion of significant miRNAs is depicted in Figure 1C . Among them, 7 were increased during the acute phase, while the remaining 14 were decreased ( Figure 1D ). Interestingly, more than half of these identified miRNAs have been previously described in allergy, although they have never been associated with anaphylaxis ( Supplementary Table 3 ).

To evaluate miRNAs capacity as diagnostic biomarkers we extended analysis to 134 samples obtained from 67 patients with anaphylaxis (34 adults with drug-induced reactions and 33 [14 adults and 19 children] with food-induced reactions). 15 patients were discarded from the study because acute and/or basal serum sample presented hemolysis (≥7 cycles) ( Figure 2A ). Therefore, the final cohort included 52 patients with anaphylaxis: 23 adults with drug-induced reactions and 29 (10 adults and 19 children) with food-induced reactions. In all these cases, UniSp2, UniSp4 and UniSp5 were detected approximately 5-7 cycles apart, confirming the correct development of the extraction ( Supplementary Figure 3A ). Furthermore, UniSp6 was amplified around cycle 18 demonstrating that reverse transcription had been successfully carried out ( Supplementary Figure 3B ).

After sample quality control, we choose 6 NGS identified miRNAs (miR-211-5p, miR-139-5p, miR-133a-3p, miR-375-3p, miR-193b-5p, miR-885-3p) to validate their serum level. Statistically significant differences were found exclusively for miR-375-3p, which is decreased during the acute phase of anaphylaxis compared to baseline in adults with drug-induced reactions, as observed in NGS data (FC: -1,63) ( Figure 2B and Supplementary Figures 3C–E ). To enlarge its value as a plausible anaphylactic biomarker, we extended the study of miR-375-3p into two cohorts of adults and children food-induced reactions, showing a decrease which was consistent with the results observed in samples from drug-induced reactions ( Figures 2C, D ). Overall, the global analysis of circulating serum levels of miR-375-3p showed a significant reduction during acute phase when compared to baseline in anaphylaxis ( Figure 2E ).

As a great portion of circulating miRNAs is contained in EVs (13), we investigated vesicle-associated miR-375-3p levels in anaphylactic human serum samples. We characterized circulating EVs from acute and baseline stages by different techniques. First, we confirmed vesicles purification by CD9, CD63 and TSG101 immunodetection, all of them considered bona fide protein markers of EVs ( Supplementary Figure 4A ). Transmission em showed heterogeneous size particles consistent previous EVs reports (27) ( Supplementary Figure 4B ). Next, we performed NTA in 9 paired samples (3 adults with drug-induced reactions and 6 (3 adults and 3 children) with food-induced reactions. ( Supplementary Figure 4C ). We did not observe differences in size or concentration when acute and basal vesicles were compared ( Figures 3A, B ).

Next, the hemolysis level of paired samples of 53 patients with anaphylaxis (19 adults with drug-induced reactions and 34 [19 adults and 15 children] with food-induced reactions) was evaluated. The coefficient between miR-23a-3p and miR-451a excluded 23 samples from 18 subjects that were discarded ( Supplementary Figure 4D ). Consequently, the determination of the miRNAs cargo was performed in 35 patients with anaphylaxis (15 adults with drug-induced reactions and 20 [8 adults and 12 children] with food-induced reactions). The correct extraction and reverse transcription of the miRNAs contained in EVs was confirmed using exogenous controls ( Supplementary Figures 4E, F ). Afterwards, the determination of vesicular miR-375-3p levels revealed a decrease during the acute phase of anaphylaxis compared to baseline, similarly as it was observed in serum ( Figure 3C ). Specifically, the independent analysis accordingly to trigger and age showed significant differences in the different cohorts of patients, confirming the drop of this miRNA in the acute phase of anaphylaxis ( Figure 3D ).

miR-375-3p serum and EVs-associated reduced levels during anaphylactic reactions could have downstream consequences and participate in the underlying molecular mechanisms of anaphylaxis. Therefore, we conducted an in silico SBA on miR-375-3p 269 target genes. The analysis pointed to several diseases and disorders related to anaphylaxis pathophysiology, such as gastrointestinal and dermatological diseases. In addition, different signaling pathways regulated by miR-375-3p were found to be closely related to the pathological reaction as Phosphoinositide 3-kinases/Protein kinase B (PI3K/Akt), G-Protein coupled receptor (GPCR) and cAMP ( Table 2 ).

Precisely, various of these mechanisms were associated with the inflammatory process, thus we evaluated a panel of inflammatory cytokines to compare their levels with those of miR-375-3p. The analysis performed in serum of patients with anaphylaxis revealed an increase of several of these molecules during the acute phase of the reaction ( Figure 4A ). Furthermore, increased GM-CSF and MCP-1 proteins correlated with decreased serum levels of miR-375-3p, although no association was observed with the other cytokines studied ( Figure 4B ). In turn, no relation was found between miR-375-3p levels within EVs and the molecules assessed by the multiplex assay ( Figure 4C ).

Increased vascular permeability is a process tightly related to anaphylaxis that involves endothelial barrier rupture. Among the mediators and molecular mechanisms affecting barrier stability, cAMP is crucial for integrity maintenance (11). To evaluate miR-375-3p role in this canonical pathway, we performed in vitro endothelial assays. For this purpose, HMVEC-Ds were transfected with an unlabeled mimic to overexpress miR-375-3p. First, the correct performance of the technique was confirmed by measuring the intracellular levels of miR-375-3p and by immunofluorescence images. A high number of miR-375-3p copies was observed by qPCR in those transfected mimic cells ( Figure 5A ), and its corresponding fluorescent labeled scramble was correctly visualized inside cells ( Figure 5B ). Subsequently, we evaluated the functional role of this miRNA on the endothelial barrier integrity. First, the stabilizing effect of cAMP was confirmed by increased values of the cellular monolayer´s resistance in non-transfected ECs ( Figure 5C ). However, cells transfected with miR-375-3p failed to response at the strengthening role exerted by this mediator ( Figure 5D ). Precisely, significant differences in endothelial resistance levels at 30 and 45 minutes were observed in cells transfected with miR-375-3p when compared to control systems (medium, TransIT-X2 and scramble) ( Figures 5E, F ). Next, we tested whether the functional change induced by miR-375-3p was also detectable morphologically via staining of VE-cadherin together with actin filaments ( Figure 5G ). Both in non-transfected ECs and scramble cells, VE-cadherin was localized at cellular membranes and actin was adjacently distributed forming the typical cobblestones cell structure. Contrary, separated cell borders were visualized in miR-375-3p ECs showing VE-cadherin disruption. In addition, miR-375-3p ECs exhibited diminished Rac1-Cdc42 protein levels ( Figure 5H ).