Analyzing the effects of benzodiazepines on the virulence and biofilm formation of Pseudomonas aeruginosa

Abstract

Sedation with benzodiazepines (BZs) has eventual side-effects increasing the risk for ventilator-associated pneumonia (VAP) (e.g. immunity alterations and nervous/mechanical responses), but there are some knowledge gaps on the topic. For instance, whether BZs could cause a modulation of bacterial virulence, and/or influence the host-pathogen interaction in neglected contexts to facilitate VAP. Consequently, we analyzed relevant in vitro and in vivo infection-related parameters to decipher whether they could be affected by BZs to increase the success for infection of the top VAP-causing pathogen Pseudomonas aeruginosa. While most variables were unaltered, an attenuated pathogenic impact on lung A549 cells (invasion, cytotoxicity and inflammation reduced up to ≈ 50%) appeared upon BZs exposure at high therapeutic concentrations, potentially because of effects mostly on the cultured cells. These facts could entail a BZs-associated stealth pathogen-like behavior of P. aeruginosa consisting of a weak immune activation proportional to the mild damage caused, perhaps favoring VAP onset. BZs also triggered a significantly increased biofilm formation (up to ≈ 2-fold > controls) on plastic plates and endotracheal tubes (supported by the upregulation of biofilm-related genes/KEGG pathways and increased c-di-GMP accumulation), suggesting the BZ-dependent boosted formation of these sessile reservoirs which could potentially increase bacterial release to low airways and thus VAP progression.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-025-32848-4.

Introduction

Pseudomonas aeruginosa is one of the most relevant human opportunistic pathogens, characterized by its high adaptability mediated by a wide array of virulence factors and mechanisms for antibiotic resistance. It is capable of causing a great variety of infections, from chronic pulmonary in patients with underlying respiratory diseases to acute nosocomial of the most diverse nature^1–3. Among them the Ventilator-Associated Pneumonia (VAP) stands out, given that it is the most common infection in ICUs, entailing serious clinical and economic burdens. VAP incidence is ≈ 10 cases/1000 days of ventilation, and it can affect up to 40% of patients subjected to this procedure. Moreover, individuals undergoing mechanical ventilation for > 48 h have up to 10 times more risk of developing pneumonia than non-intubated ones. Furthermore, VAP increases the length of hospital stay by ≈ 7 days and the cost of medical care by ≈$40,000/patient^4–6. P. aeruginosa is the one of the top species causing VAP (≈ 20% of cases), with a mortality around 13.5%. In the case of VAP due to multidrug-resistant P. aeruginosa strains, mortality reaches ≈ 40%, demonstrating the great magnitude of the problem posed by this species as cause of VAP^4–6.

Individuals subjected to mechanical ventilation are generally anesthetized and sedated to improve comfort and tolerance of the endotracheal tube (ETT) and to prevent patient-ventilator desynchrony, goals for which different drugs are available. Among sedatives, benzodiazepines (BZs) are widely administered, but several publications called into question their real convenience because of undesired side effects. In this regard, different studies analyzing large cohorts of patients proposed the treatment with BZs over non-BZ drugs as a potential added risk factor for VAP^7–9. This is likely consequential to the secondary effects of BZs on the patient: delirium, agitation, increased need of re-intubation and consequent incremented aspiration facilitating the entry of opportunistic pathogens colonizing the upper respiratory tract into deep airways^7–9. Additionally, alteration of the immune response has been suggested as an additional cause underlying this higher risk of infection^10–12: there is evidence to support that BZs partially suppress phagocytic activity and inflammatory responses through the activation of γ-aminobutyric acid (GABA) A receptors (central BZ receptors) and peripheral BZ receptors (TspO proteins located in mitochondrial membranes) and their respective signaling pathways, finally impairing immune effectiveness^13–19. However, since BZs receptors are widely present in our different tissues’ cells, these drugs could have other neglected effects beyond immune-suppression, e.g. on certain features of the respiratory epithelium, somehow altering the interaction with the colonizing bacteria (in our case P. aeruginosa) to favor VAP development, a topic that has not been specifically investigated.

When a side effect of a drug is found to influence the outcome of an infection, it is usually attributed to impacts on the host rather than on the pathogen^20,21. Therefore, when this secondary effect in the patient is deciphered, no further research on the bacterium is undertaken. Thus, the possibility of significant concentrations of BZs reaching P. aeruginosa populations colonizing the airways and ultimately influencing its pathogenic performance to increase success for VAP has never been approached. However, this hypothesis could be not far-fetched because of previous clues: some studies in related species such as Escherichia coli or P. fluorescens proposed the existence of bacterial receptors able to recognize BZs and modulate virulence in response^22–24. Moreover, the existence of specific receptors for a compound to drive pathogenicity-modulating effects seems not always needed, as demonstrated for steroids causing envelope stress and derived virulence-related changes in P. aeruginosa^25,26. Therefore, without known specific receptors for BZs yet described for P. aeruginosa, our approach is timely to decipher whether unknown effects of these drugs impacting its pathogenic performance could be contributing to the documented higher incidence of VAP in BZ-treated patients^7–9.

Finally, despite the abovementioned evidence regarding undesired neurological, mechanical and immuno-suppressive effects associated to BZs, it has never been specifically approached whether these facts are reproduced in an ad hoc animal model of VAP by P. aeruginosa. Thus, it has not been established in vivo an unequivocal cause-effect relationship between BZs treatment and a significantly increased incidence/severity of VAP in comparison with other sedatives. Trying to fill this knowledge gap was therefore a key objective of this study. Additionally, in view of all the abovementioned questions still to be resolved, here we explore whether BZs at concentrations attainable in patients serum could modulate P. aeruginosa infection-related behavior, by analyzing different relevant parameters not only regarding the bacterium itself, but also the host-pathogen interplay. This study provides novel knowledge predominantly from the microbiological perspective that complements the array of factors supporting why BZs sedation may pose an added risk for VAP by P. aeruginosa.

Methods

Bacterial strains, reagents and growth conditions

Pseudomonas aeruginosa PAO1 and PA14 reference strains were used as specified in each assay. The previously constructed mutants PAOMS (PAO1 defective in mutS²⁷ and PAOMA (PAO1 defective in mucA²⁸ were used in the corresponding experiments as controls for hypermutation and robust biofilm formation respectively. For biofilm-related confocal microscopy assays, PAO1 was tagged with miniTn7 containing the cyan fluorescent protein gene (cfp) by electroporation (PAO1-cfp)²⁶. The reporter strain PAO1 PcdrA::luxCDABE was used for indirect quantification of c-di-GMP^26,29.

Diazepam (≥ 98%) and Midazolam (United States Pharmacopeia Reference Standard, Sigma-Aldrich) were used to prepare 2 mg/ml stocks with the typical solvent solution of commercial diazepam preparations (e.g. Hospira, DASH pharmaceuticals): 40% propylene glycol, 10% alcohol, 4.9% sodium benzoate and 0.1% benzoic acid. In accordance with literature, 10 µg/ml was chosen as an archetypical top concentration of BZ that can be attained in patients’ serum^19,30,31 and therefore, generally used in all the experiments unless otherwise specified. Parallel assays with the same amount of solvent (final concentration: 0.5%) were always performed to discard the possibility of some effect being due to this instead of BZs. As general rule, BZs or solvent were added in the media in which the assays were going to be performed, whereas only in the animal model (see below) these additives were also incorporated into the overnight LB broth cultures prepared before the assays (37 °C, 180 rpm agitation).

Bacterial growth assays

In vitro exponential growth rate assays were performed according to procedures previously described by our group³². Briefly, 1 mL samples taken from liquid LB overnight cultures were diluted in 50 mL of fresh LB broth (containing or not diazepam 10 µg/ml, midazolam 10 µg/ml or solvent 0.5%) in 250-mL flasks and incubated at 37 °C with agitation at 180 rpm. The doubling times of exponentially growing cells were determined by plating serial dilutions onto LB agar plates at 60 min intervals, using standard formulas for the calculation of specific growth rate constant, µ = .

[N standing for bacterial counts and t for time in minutes, at initial (0) and final points] and duplication time, g = (minutes). At least three independent experiments were performed for each of the strains/conditions.

Bacterial motility

Swimming, swarming and twitching motilities were determined as described previously in Petri dishes containing the appropriate medium (complemented or not with diazepam 10 µg/ml, midazolam 10 µg/ml or solvent 0.5%, always added after autoclave sterilization)³². Petri dishes with the mentioned semi-solid media were incubated at 37 °C for 16 h sealing them with plastic wrap to avoid evaporation, and the diameter of the motility zones measured. Eight measurements for each condition and motility type were recorded.

Antibiotic susceptibility testing

Minimum inhibitory concentrations (MICs) of the antipseudomonal drugs ceftazidime, cefepime, ceftolozane, piperacillin, aztreonam, imipenem, meropenem, ciprofloxacin, gentamicin, and colistin were determined by microdilution according to EUCAST guidelines (www.eucast.org), using Mueller-Hinton (MH) broth alone or supplemented with the corresponding additive. Three independent assays for each MIC determination were performed for each strain/condition, and the median values were recorded to display in the Results section. MICs of diazepam, midazolam and solvent were determined following the same procedure.

Estimation of spontaneous mutants frequency

The frequencies of emergence of spontaneous single-step rifampicin-resistant mutants were determined according to previously described protocols³³, supplementing or not the Mueller-Hinton broth cultures with the corresponding additives. After the colony counts in Mueller-Hinton agar versus Mueller-Hinton agar plus rifampicin (300 µg/ml), mutant frequencies were calculated by dividing the median numbers of mutants by the median numbers of total cells in the five independent cultures of each media (Mueller-Hinton broth alone versus with the corresponding additive). At least 5 independent mutation frequency experiments (each one with the mentioned five independent cultures) were performed for all the conditions.

Cell culture infection experiments: invasion, cytotoxicity and inflammation

The day before the assays, the human lung alveolar A549 cells (source: human lung carcinoma; catalogue number: 86012804-1VL; supplier: Sigma-Aldrich; RRID: CVCL_0023), always used between passages 3 and 30, were seeded at ≈ 0.5 × 10⁵ cells per well in 24-well plates, using RPMI-1640 without phenol red supplemented with 10% of heat-inactivated fetal bovine serum, 10mM HEPES, 2 mM L-glutamine and 1X antibiotic-antimycotic solution (Biowest) as maintenance medium. The day after, the cells were ≈ 90% confluent (≈ 1 × 10⁵ cells/well) and were infected at a multiplicity of infection (MOI) of 100, according to procedures previously described by our group^32,34. Briefly, bacteria from overnight liquid LB cultures were washed with PBS and diluted in RPMI-1640 without serum/antibiotic-anti-mycotic solution [henceforth called infection medium (IM), to which the different BZ/solvent additives were added when indicated]: ≈1 × 10⁷ PAO1 or PA14 cells/500µl of IM were used to infect the A549 monolayer in each well after discarding the initial cell culture medium and washing with PBS. After a period of infection of 3 h, the IM was collected and stored at −80 °C prior to additional analysis (inflammation and cytotoxicity, see below).

To determine the invasion capacity, PAO1 was used in aminoglycoside exclusion assays previously described by our group^32,34. After the 3 h infection, the IM was replaced with 1 mL/well of fresh RPMI-1640 containing 0.4 mg/ml of amikacin for 1.5 h to kill extracellular bacteria. The medium was removed and the cells washed twice with PBS. 0.5 mL of PBS containing 0.1% of Triton X-100 (Sigma-Aldrich) were added to each well, incubated for 10 min to lyse the cells and release the intracellular bacteria, which were quantified after plating in LB agar. The abovementioned supernatants from the PAO1 3 h-infections were used for IL-8 release quantification as inflammatory marker, using a Human IL-8 Instant ELISA kit (Invitrogen) following the manufacturer’s instructions. Cytotoxicity was assessed in the PA14 3 h-infection supernatants as previously described by our group^32,34 using a Cytotoxicity Detection Kit PLUS (Roche), following the manufacturer’s instructions for LDH release quantification. Samples for all the cell culture infection experiments were obtained from at least 9 wells (3 from 3 independent plates) per strain/condition.

LPS purification and stimulation of cell cultures

LPS from PAO1 strain was purified following previously described protocols with minor modifications³⁵. Briefly, bacterial cells from overnight liquid cultures were harvested by centrifugation. The resulting pellets were washed and resuspended in a final volume of 8 mL of ultrapure water and incubated for 10 min at 68 °C. Then, an equal volume of phenol solution was added, mixed, incubated for 30 min at 68 °C and cooled to 4 °C. The mixtures were centrifuged at 1,000 x g for 20 min at 4 °C to separate the aqueous and phenol phases. The upper aqueous phase was recovered and dialyzed against ultrapure water at 4 °C for 7 days. The phenol-free dialysates were centrifuged at 8000 x g for 15 min. The resulting supernatants were sequentially treated with RNase, DNase and Proteinase K, followed by ultracentrifugation. The final pellets were resuspended in ultrapure water and lyophilized to quantify the amount of purified LPS to further adjust the concentration as needed. LPS quality was analyzed by 15% polyacrylamide gels (SDS-PAGE) and silver staining^34,35. Once resuspended in sterile double distilled water, the purified LPS was used to stimulate A549 monolayers at a final concentration of 20 µg/ml in regular maintenance medium (with or without the different BZ or solvent additives) for 24 h, following published procedures^36,37. After this period, cell supernatants were used to determine the amount of released IL-8 and LDH (cytotoxicity) using the abovementioned kits.

Rat model of VAP

Housing of animals and experiments were conducted in compliance with the Spanish legislation (RD53/2013) (and therefore European Community legislation) and approved by the local Ethic Committees [Animal Experimentation Ethics Committee of IdISBa and of UIB (references CEEA-005–2023 and CEEA 223-09-23)], and Authorities [Conselleria d’Agricultura, Pesca i Medi Natural (reference SSBA 03/2024 AEXP)], Balearic Islands, Spain. Accordingly, the animal model was conducted in accordance with ARRIVE guidelines (https://arriveguidelines.org) and thus with all relevant guidelines and regulations as mentioned above.

Only PAO1 as infecting strain and midazolam as BZ were used, in order to reduce the number of animals. The model was developed based on previous publications in which bacterial inoculation is followed by ventilation to ensure the development of infection, with minor modifications^37–39. Anesthetic/sedative dosages were also taken from previously published studies⁴⁰. Briefly, ≈ 10–12 weeks Wistar rats (4 males + 4 females/group) weighing ≈ 200–250 g, were treated with intraperitoneal ketamine [Anesketin^® (Dechra)]/dexmedetomidine [Sedadex^® (Dechra)] at 75/0.25 mg/kg (as anesthetic/sedative agents) or ketamine/dexmedetomidine/midazolam (Normon) 75/0.125/5 mg/kg, and ventilated for 1 h using a SAR-1000 device (CWE Inc.), with standard parameters (tidal volume 1.5–1.9 ml; respiratory rate 60–70 bpm; peak-inspiratory pressure 10 cm H₂O; flow rate 0.180–0.225 ml/min; inspiration time 0.43–0.50 s; inspiratory/expiratory ratio 1:1). Intubation was conducted using the Rat Endotracheal Intubation Kit (RWD Life Science), applying an Introcan Safety G16 catheter (Braun) as endotracheal tube (ETT), to which the ventilator cannulas were connected. Temperature, heart rate and oxygen saturation were monitored using an IMEC8 Vet device (Mindray). Atipamezole hydrochloride [Revazol^® (Dechra)] subcutaneous at a 0.5 mg/kg dose was used for sedation reversal after the ventilation period finished.

Besides non-infected control animals (only anesthetized/sedated with the abovementioned treatments and ventilated without bacterial inoculation), infection groups were intra-tracheally inoculated through the abovementioned catheter with 100 µl of saline solution containing ≈ 5E⁷ PAO1 cells (proceeding from overnight LB cultures exposed or not to midazolam 10 µg/ml) before ventilation. 24 h after ventilation the animals were sacrificed through CO2 exposure (following standard procedures in compliance with European legislation), and their lungs excised and homogenized in PBS for bacterial plating and colony counts and IL-8 ELISA (SunLong Biotech) as inflammatory marker following manufacturer’s instructions.

Assessment of biofilm formation on plastic plates

Biomass quantification by crystal Violet (CV)

the CV assay was performed following previously described protocols³³. Briefly, ≈2E⁶ stationary phase cells of the indicated P. aeruginosa strains/conditions (untreated, treated with 10–100 µg/mL of diazepam or midazolam, or solvent at 0.5 or 5%), were inoculated into each of eight wells/strain in 96-well round bottom microtiter plates (Nunc), to attain a final volume of 100 µl of fresh LB broth. After incubations of 24 h at 37 °C, the plates were carefully rinsed with water to remove the planktonic cells. Afterwards, the plates were air-dried and stained for 10 min with 125 µl of 0.1% CV solution/well. The plates were rinsed with water and air-dried again, and the CV dye retained by the biofilm matrix was then solubilized with 200 µl of 30% acetic acid (30 min, room temperature). Once mixed by pipetting, 125 µl of each well’s contents were individually transferred to clear flat-bottom microtiter plates to read the absorbance at 590 nm using a Synergy H1 plate reader (BioTek). The final expressed values are means of eight wells from each of 5 independent plate replicates.

Biofilm quantification by confocal microscopy²⁶: 5 ml culture of PAO1-cfp (tagged in the chromosome) was grown overnight in LB broth at 37 °C and 180 rpm, washed with 0.9% saline solution and normalized to an OD600nm of 0.1 in RPMI-1640 medium untreated, treated with 10–100 µg/mL of diazepam or midazolam or solvent at 0.5 or 5%, as previously explained. A total of 200 µL/well were added to 8 well micro-slide Ibitreat chambers (Ibidi GmbH), inoculating 2 wells per condition. The slides were incubated statically at 37 °C for 24 h or 48 h in a humid chamber. After incubation, the medium was carefully removed and the eDNA stained with 1 µM of DiYO-1 (AAT Bioquest) for 10 min, at room temperature and in the dark. After the stain was removed, 100 µL of 0.9% saline solution were added to prevent desiccation prior to laser scanning confocal microscopy (CLSM). Biofilms were imaged using the enhanced cyan fluorescent protein (ECFP) configuration (Emission 405 nm/Excitation 454 nm), while the fluorescein configuration, as recommended by the manufacturer (Emission 488 nm/Excitation 579 nm) was used for eDNA imaging, using an LSM 710 device (Carl Zeiss). At least three independent replicates were performed and a total of 4 Z-stack images per well (8 per condition) were taken for each repeat. Biomass of the biofilms (for both bacterial cells and eDNA) were quantified using COMSTAT2 software following previously described protocols⁴¹, applying automatic thresholding (Otsu’s method) without connected volume filtering.

Quantification of viable cells within biofilms

the PAO1 initial inoculum was prepared as explained for microscopy assays, and 200 µl of the OD600nm = 0.1 normalised culture in RPMI-1640 alone, with 10 µg/ml of diazepam, midazolam or 0.5% solvent, were added in triplicate in a clear 96 well plate (Nunc, Thermofisher) being afterwards covered with a 96-peg lid (Nunc Inmuno TSP, Thermofisher), as previously described⁴². After 24 h, pegs were rinsed for 10 min in dH₂O and the lid was transferred to a opaque white 96-well plate containing 100 µl of cation-adjusted Mueller-Hinton and centrifuged for 20 min at 800 x g to recover the biofilm biomass, as described elsewhere⁴². Afterwards, 100 µl of BacTiterGlo reagent (Promega) was added to each well and incubated for 5 min before reading the luminescence in an automated plate reader (Synergy H1, BioTek), following manufacturer’s instructions and previously described protocols⁴³.

Assessment of biofilm formation on endotracheal tubes (ETTs)

In order to test the ability of BZs to potentially induce an increment of P. aeruginosa biofilm in ETTs at longer periods, the following experimental set ups (extracted and combined from Pericolini et al. and Latorre et al.^44,45were followed. Briefly, 0.5 cm pieces of an ETT (7.5 mm internal diameter, OXYGEM) were cut under sterile conditions and further UV sterilized for 60 min in a biosafety hood. Three pieces per condition were placed into a 24-well plate and inoculated with 500 µL of an overnight culture of PAO1 normalized to an OD600nm of 0.1 in TSB + 1% glucose untreated, treated with 10 µg/mL of diazepam or midazolam or solvent at 0.5%. After the first 24 h and during the following 4 days, the ETT pieces were daily moved to new 24-well plates containing fresh medium and the corresponding additives. The experiment was performed in duplicate for CV and CLSM analysis and three biological repeats were performed. The last day, the pieces were gently washed with 0.9% saline solution to remove planktonic cells and further analyzed.

Biofilm quantification by CV

the pieces of the ETT were stained with 500 µL of a 0.1% CV solution for 10 min. The ETT portions were rinsed with water and air-dried. The CV dye retained by the biofilm matrix was then solubilized with 500 µl of 30% acetic acid (30 min, room temperature). A 1/10 dilution of the well content was added in triplicate in a 96 well plate and the absorbance at 590 nm was read using the Synergy H1 plate reader (BioTek).

Biofilm quantification by CLSM

the biofilm was inactivated by freezing the ETT segments at 80 °C for at least 72 h. After letting the segments thaw during 30 min, the pieces were stained with 500 µL of the Syto9 stain [1.5 µL Syto9 per 1 mL of 1x PBS, pH 7.4 according to manufacturer’s instructions (Live/Dead Baclight kit, Thermofiser)] during 10 min in the dark. The ETT segments were mounted on coverslips and imaged using the CLSM, using the Syto9 configuration. The images (six per segment) were processed as previously explained.

c-di-GMP reporter assay

Following published protocols^26,29, production of the second messenger c-di-GMP was measured indirectly by quantifying the expression of a cdrA::lux transcriptional fusion in PAO1^26,29. Bacteria with or without BZs or solvent were cultured statically in black bottom 96-well microtiter plates at 37 °C. OD600nm and luminescence were quantified in an automated plate reader (Synergy H1, BioTek) recorded every 30 min over 24 h.

Transcriptomic analysis

RNA extraction: 10 mL of LB containing the different treatments (control or midazolam 10 µg/ml as representative BZ) were inoculated with 500 µL of a previous overnight PAO1 liquid culture, and grown until OD600nm reached ≈ 0.5 at 37 °C and 180 rpm. One milliliter of the culture was mixed with 2 ml of RNA protect (QIAgen) and the RNA extraction was performed using the RNeasy kit (QIAgen). Contaminant DNA was removed using 2 U of the TURBO DNAse kit (Invitrogen) and incubated for 30 min at 37 °C. Bacterial RNA was extracted from three independent replicates for each experimental condition.Library Construction, Quality Control and Sequencing (Novogene Corporation Inc.): ribosomal RNA was removed from total RNA using the Illumina Ribo-Zero Plus rRNA Depletion Kit followed by ethanol precipitation. For the library preparation, the Novogene NGS Stranded RNA Library Prep Set (PT044) was used. After fragmentation, the first strand cDNA was synthesized using random hexamer primers. During the second strand cDNA synthesis, dUTPs were replaced with dTTPs in the reaction buffer. The directional library was ready after end repair, A-tailing, adapter ligation, size selection, USER enzyme digestion, amplification, and purification. The library was checked with Qubit and real-time PCR for quantification and bioanalyzer for size distribution detection. Quantified libraries were pooled and sequenced on an Illumina NovaSeq 6000 using the PE150 strategy, according to effective library concentration and data amount required. Data were unequally shared with 3G raw data (20 million paired reads) for each sample.

RNA-seq bioinformatics analysis: raw RNA-seq reads were quality-checked with FastQC v0.11.9. Adapter trimming and removal of low-quality (< Q30) bases were performed using Trimmomatic v0.39⁴⁶. Ribosomal RNA sequences were filtered out with SortMeRNA v4.3.6⁴⁷ using default rRNA databases in paired-end mode. Filtered reads were aligned to the P. aeruginosa PAO1 reference genome (NCBI RefSeq: NC_002516.2, assembly GCF_000006765.1) using STAR 2.7.10a⁴⁸, generating sorted BAM files and gene-level counts. The reference genome and annotation files were downloaded from NCBI (https://www.ncbi.nlm.nih.gov/nuccore/NC_002516). Read summarization was also performed with featureCounts v2.0.3⁴⁹, using strand-specific settings and gene annotations from the corresponding GFF file. Quality reports from all steps were aggregated using MultiQC v1.26. Differential gene expression analysis was performed using DESeq2 v1.44.0⁵⁰. Differentially expressed genes (DEGs) were identified using the Benjamini–Hochberg method for controlling the false discovery rate (FDR), considering genes significant if FDR ≤ 0.05 and absolute log₂ fold change ≥ log₂(1.5) as threshold. Functional enrichment, i.e. Kyoto Encyclopedia of Genes and Genomes (KEGG)^51,52 biological pathways/functions (www.kegg.jp/kegg/kegg1.html) over-representation analysis was carried out using the clusterProfiler v4.12.6⁵³ R package. Pathways with an adjusted P value ≤ 0.05 were considered significantly enriched.​.

Quantification of gene expression through real-time RT-PCR. To confirm RNA-seq results, the mRNA of selected genes (Table S1) of PAO1 strain [exponential phase, grown in regular LB (control) or supplemented with midazolam at 1, 5, or 10 µg/ml] was quantified through real-time reverse transcription PCR (RT-PCR) and specific primers (Table S1), following described protocols³². Briefly, an amount of 50 ng of purified RNA extracted as explained above was used for real-time RT-PCR using the QuantiTect SYBR Green RT-PCR Kit (Qiagen) in a CFX Connect device (Bio-Rad). The rpsL housekeeping gene was used to normalize mRNA levels using previously described primers³², and the results were referred to the values of PAO1 grown in LB alone. All RT-PCRs were performed in duplicate, and mean expression levels from three independent RNA extractions were considered.

Data analysis

Quantitative variables were analyzed using GraphPad Prism 7 software, through one-way ANOVA (with Tukey’s post hoc multiple comparisons test), Student’s t test (two tailed, unpaired), or Mann-Whitney U test (two-tailed) as appropriate. A P value < 0.05 was considered statistically significant.

Results

Exposure to benzodiazepines does not affect growth, motility, antibiotic susceptibility, or spontaneous mutation frequency of P. aeruginosa

Exposure to BZs concentrations around the maximum values attainable in patients’ serum (10 µg/ml^19,30,31 did not affect P. aeruginosa exponential growth rate in liquid medium (expressed in terms of duplication time), motility patterns (swimming, swarming or twitching), or antibiotic susceptibility profile against relevant antipseudomonal drugs (Supplementary Figures S1 and S2, and Table 1). In the case of the frequency of spontaneous one-step rifampicin-resistant mutants (Figure S3), although slightly increased values compared to PAO1 did appear, these never reached the statistical significance threshold, at a clear distance from the hyper-mutator control. These increments appeared also in samples exposed to BZs solvent, which suggests that in any case, BZs would not be their trigger. On the other hand, both BZs caused the inhibition of P. aeruginosa growth at 250 µg/ml (Table 1), whereas the minimum percentage of solvent in the Mueller-Hinton broth that had this effect was 12.5%, an expected result given its concentration and composition with different preservatives (see Methods). These inhibitory concentrations are much higher than those used in our subsequent experiments (BZs at 10 ug/ml and solvent at 0.5%), therefore it is unlikely that these latter had any influence on the remainder of our results.

Table 1.

Minimum inhibitory concentrations (MICs) of representative antipseudomonal antibiotics, BZs and the solvent used, determined in PAO1 strain.

Conditionsa	MICsb
	SOL	DIA	MID	CAZ	FEP	TOL	PIP	AZT	IMP	MER	CIP	GEN	COL
CONTROL	12.5	250	250	1	1	0.5	2	2	1	0.5	0.12	1	0.5
SOL 0.5%	-	-	-	1	1	0.5	4	2	2	1	0.12	1	0.5
DIA 10 µg/ml	-	-	-	1	1	0.5	4	2	1	1	0.12	1	0.5
MID 10 µg/ml	-	-	-	1	1	0.5	2	2	2	1	0.12	1	0.5

^aMICs were determined in regular Mueller-Hinton broth (control), or supplemented with each additive displayed on the left column at the concentration shown. ^bMIC of solvent is expressed in percentage (vol./vol.), whereas those of BZs and antibiotics are in µg/ml. Abbreviations: SOL: solvent (40% propylene glycol, 10% alcohol, 4.9% sodium benzoate and 0.1% benzoic acid); DIA: diazepam; MID: midazolam; CAZ: ceftazidime; FEP: cefepime; TOL: ceftolozane; PIP: piperacillin; AZT: aztreonam; IMP: imipenem; MER: meropenem; CIP: ciprofloxacin; GEN: gentamicin; COL: colistin.

Benzodiazepines cause a significant decrease of the values of P. aeruginosa infection-related parameters in co-cultures with human pulmonary A549 cells

Figure 1 shows the typical cell culture infection-derived parameters used to assess the pathogenic behavior of bacteria in a model utilizing human pulmonary A549 cells^32,54,55: invasion capacity, elicited inflammatory response, and cytotoxicity. As can be seen in Fig. 1A, the value of invasive PAO1 cells (those surviving in the aminoglycoside exclusion assays through penetration into A549 cells’ cytosol, see Methods) was significantly decreased upon BZs exposure, especially in the case of midazolam, reaching a reduction of ca. 50%. Additionally, quantification of the IL-8 released during the same 3 h infection with PAO1 revealed that incubation with BZs entailed a remarkably lower release of this inflammatory marker, to almost a third of the control values (Fig. 1B). On the other hand, the cellular death caused by PA14 (cytotoxicity) was slightly but significantly reduced upon BZs exposure, up to 15% compared to infection conducted in medium supplemented with the solvent. Solvent per se already had an appreciable effect reducing LDH release compared to control (Fig. 1C), but this effect was significantly boosted in the presence of BZs, indicating their specific impact in the cytotoxicity attenuation.

Fig. 1.

Human pulmonary A549 cell culture infection-related parameters obtained with the different conditions and P. aeruginosa strains indicated. (A) PAO1 invasion assay. The number of invasive colony forming units (CFUs) per well of A549 cells (multiplicity of infection [MOI] of 100) after 3 h is shown. (B) Quantification of IL-8 released to supernatants during the mentioned PAO1 infection. (C) Cytotoxicity results (LDH released as percentage of that from a completely lysed A549 cells well) after infection with PA14 (strain chosen because of its cytotoxic profile; PAO1 causes virtually no cytotoxicity effects in 3 h)³⁵. (D) Quantification of IL-8 secretion by A549 cells after stimulation with 20 µg/ml of LPS (PAO1 strain) for 24 h. Results are grouped with horizontal lines when there was no statistical difference between them (n.s.: not significant). Values displayed are the means (columns) plus SDs (error bars) obtained in assays from three wells from each of three independent plates. *P < 0.05 in the one-way ANOVA plus Tukey’s post hoc test.

To obtain more clues regarding the attenuated inflammatory response elicited by P. aeruginosa upon BZs exposure, we stimulated A549 cells with purified PAO1-derived lipopolysaccharide (LPS, 20 µg/ml, 24 h) in cell culture medium with or without the different additives. As can be seen in Fig. 1D, although not achieving statistical significance due to high variance (P > 0.05), BZs on average induced IL-8 release from A549 cells at 50% reduction relative to BZ-negative groups when exposed to LPS. This suggests that the decreased IL-8 release upon PAO1 infection is likely due, at least partially, to impacts of BZs on the A549 cells behavior rather than solely on bacterial features. To discard a potential influence of an eventual LPS-linked cytotoxicity on these IL-8 results, LDH was also quantified in the supernatants of LPS-stimulated cells, but cellular death was below 5% in all cases (data not shown).

Short term sedation with benzodiazepines in a rat model of VAP does not significantly change the P. aeruginosa infection output

Results in the previous section suggest a short-term attenuated pathogenic behavior of P. aeruginosa over cultured A549 cells in response to BZs. Consequently, we decided to conduct a VAP model in rat to ascertain whether sedation with versus without BZs could have similar acute impacts for the infection output in vivo, a topic never investigated before in an ad hoc animal model. Nevertheless, as can be seen in Fig. 2, none of the measured parameters (bacterial load and inflammatory response in lungs) showed statistically significant differences comparing the animals sedated with versus without midazolam. Therefore, our results suggest that sedation of rats with one single dose of midazolam and for a short period of ventilation (1 h) has not significant acute impacts for the VAP output.

Fig. 2.

Parameters derived from the rat model of VAP by P. aeruginosa PAO1 strain, with the animals treated as specified in the X axis. Filled symbols represent male whereas empty ones correspond to female rats. (A) Quantification of IL-8 in the lungs homogenates of control and infected animals. Horizontal bars correspond to the mean values calculated from the 8 animals of each group. P values obtained for the two-tailed Student’s t test were > 0.05 for the comparisons between the animals treated with midazolam versus those not treated with this BZ. (B) CFUs per lungs excised from each animal. Horizontal bars represent the median value of bacterial load for each group of rats. Two-tailed Mann Whitney U test P values were > 0.05 for all the comparisons entailing presence versus absence of midazolam. Abbreviations: KET: ketamine; DEX: dexmedetomidine.

Exposure to benzodiazepines enhances biofilm formation by P. aeruginosa

Biofilm growth is one of the P. aeruginosa hallmarks in the chronic niche (e.g. cystic fibrosis airways infection), but these sessile communities also pose essential bacterial reservoirs in ETTs and other invasive devices enabling typical nosocomial acute infections such as VAP^56,57. Thus, to complement our A549 cells- and rat model-derived results, we decided to investigate whether exposure to BZs could influence P. aeruginosa biofilm formation in vitro. First, we conducted classical experiments to measure biofilms formed on 96-well plastic microtiter plates (24 h and 48 h)³³, to provide a general idea of the potential impact of BZs in the biofilm formation capacity of P. aeruginosa. Additionaly, not as a comparison but as a complement to these assays, and to provide a more clinically impactful perspective simulating conditions of critical patients who are intubated and ventiladed for longer periods, we quantified biofilms formed on sections of ETTs after 5 days^45,57–59.

Figure 3A shows the absorbance values derived from CV assays performed on 96-well plastic microtiter plates (24 h). Although exposure to BZs at 10 µg/ml apparently caused no difference on PAO1 biofilm biomass, when the CV assays were repeated with BZs at 100 µg/ml to simulate an accumulative effect (i.e. BZs could be additively retained within the biofilm after several sedative doses and days of intubation-ventilation, reaching concentrations higher than those in serum), an increase superior to 2-fold was seen compared to controls (especially in the case of midazolam). Moreover, as can be observed in Fig. 3B, when quantifying 24 h-old biofilms formed on 96-well plastic microtiter plates through confocal microscopy, exposure to BZs at 10 µg/ml demonstrated statistically significant increases (up to ≈ 50% compared to controls) in the values of biomass (PAO1-Cfp cells) and extracellular DNA (key component of the biofilm matrix). These results were likely due to the greater sensitivity of the microscopy techniques compared to CV assays, in which a need for exaggerated BZs concentrations to reveal measurable differences seemingly existed. To discard that CV results were due to sequestration of the dye by the BZs at 100 µg/ml (which would be on their turn adsorbed on the walls of wells), negative controls free of P. aeruginosa were performed, following the same CV assay procedure. Absorbances from wells with versus without additives were virtually the same, ruling out this artifact possibility (Supplementary Figure S4). Moreover, to confirm our microscopy-derived data for biofilm quantification, we determined the cell viability in the 24 h-old biofilms formed on plastic microtiter plates through a luminescence commercial assay as previously described^42,43. As can be observed in the Fig. 3C, exposure to BZs at 10 µg/ml triggered a significantly increased amount of viable P. aeruginosa cells compared to controls, posing a clear parallelism with the abovementioned microscopy-derived parameters. Finally, the Supplementary Figure S5 displays the same previously mentioned microscopy-derived parameters in 48 h-old biofilms treated with BZs at 10 µg/ml, revealing the conservation of trends mentioned for 24 h.

Fig. 3.

24 h-old biofilm-derived parameters obtained under the experimental conditions indicated in Material and Methods for P. aeruginosa PAO1 grown on plastic plates. (A) Absorbance of samples after CV assays: columns represent the mean values of eight wells from each of four independent plate replicates, with error bars corresponding to SDs. (B) biomass data were obtained by analyzing cyan fluorescence (PAO1-Cfp, filled columns) while eDNA values by staining with Di-YO1 (excitation 491 nm-emission 508 nm, stripe pattern). Columns represent the mean values from at least three independent experimental replicates while the error bars correspond to SDs. (C) Quantification of PAO1 cells viability in 24 h-old biofilms untreated, treated with BZs or solvent, measured through the BacTiterGlo assay. The negative control column corresponds to the media without bacteria. Columns represent the mean values of the experimental replicates, with error bars corresponding to SDs. RLU: relative luminiscence units. (D) Production of c-di-GMP by PAO1 in response to BZs: light output was quantified as a function of growth (RLU/OD600) for PAO1 PcdrA::luxCDABE. Representation of results was performed by calculating the area under the curve (AUC) of RLU/OD600 values for the 24 h period, and displayed through columns and error bars (SDs), from three independent experimental replicates. Statistical significance was determined by one-way ANOVA plus Tukey’s post-hoc test.*P < 0.05. n.s.: not significant.

Transition from planktonic to biofilm lifestyle, and also from an acute (virulent) to a chronic infection (attenuated) profile, has been linked to the accumulation of the signaling molecule cyclic diguanylate (c-di-GMP) in P. aeruginosa⁶⁰. Therefore, to check that the observed boosted biofilm formation was supported by this intracellular acute-chronic switch and thus was not an artifact, we quantified the accumulation of c-di-GMP through a previously described indirect method^26,29. As shown in Fig. 3D, the accumulation of c-di-GMP was significantly increased during the 24 h of assay (up to ≈ 20% of increment for midazolam compared to control) thus following a parallel trend to that of the abovementioned microscopy-derived parameters for 24 h biofilm quantification.

In accordance with plastic microtiter plates data, our results from ETTs revealed statistically significant increases (up to ca. 30% compared to controls) in absorbance after CV staining (Fig. 4A) appearing in response to BZs exposure. A greater difference was seen for total biomass quantification in this case through double stranded DNA Syto9 staining (thus binding both to intracellular and extracellular DNA) revealing proportional increases as expectable (up to ca. two-fold compared to controls, Fig. 4B). Interestingly, these assays demonstrate an apparent trend of midazolam causing more robust effects than diazepam. Representative 3D pictures of 5-day-old biofilms formed on ETTs sections (obtained through the Zen black-ZEISS software) clearly supporting the data mentioned in this section, are displayed in the Supplementary Figure S6.

Fig. 4.

5 days-old biofilm-derived parameters obtained under the indicated experimental conditions for P. aeruginosa PAO1 grown on endotracheal tubes (ETTs) sections. (A) Absorbance of samples after CV assays: columns represent the mean values of 3 sections from each of 3 independent replicates, with error bars corresponding to SDs. (B) biomass data were obtained by Syto9 staining and confocal microscopy analysis. Columns represent the mean values from at least three independent experimental replicates while the error bars correspond to SDs. Statistical significance was determined by one-way ANOVA plus Tukey’s post-hoc test.*P < 0.05. n.s.: not significant.

Transcriptomic analysis of BZs exposure effects on P. aeruginosa

In view of our results indicating that BZs exposure triggered a significantly increased capacity for biofilm formation and perhaps a certain intrinsic virulence attenuation of P. aeruginosa on cell culture, we sought to investigate which basis could support these outputs. Besides the abovementioned quantification of the quintessential messenger involved in biofilm and virulence regulation (c-di-GMP), transcriptomic analysis to delve into additional pathways explaining our results was the next objective. In this vein, we performed RNA-seq of PAO1 strain grown in planktonic culture (with versus without the addition of midazolam as representative BZ), to ascertain which transcriptomic scaffold could support the transition to a profile of increased biofilm formation (and perhaps, of attenuated virulence). As can be seen in the DataSet S1, midazolam exposure even for such a short period (only ca. 2 h, until the refreshed overnight liquid culture reached an OD600nm of 0.5, see Methods) exerted very important changes on PAO1 transcriptome, which supports the idea of BZs capacity to readily modulate bacterial behavior. More specifically, from a total of 570 differentially-transcribed genes [based on P values (FDR) below 0.05], a total of 66 were considered as significantly upregulated in the midazolam-treated samples (at least 1.5-fold compared to PAO1 without BZ), whereas 26 were down-regulated (at least 1.5-fold compared to PAO1 without BZ).

Interestingly, several of the genes significantly upregulated in the midazolam-exposed cells were previously related to adhesion and biofilm formation/modification. For instance, that of the two-component system sensor pprA (involved in fimbriae assembly, adhesion and thus transition between planktonic to community lifestyles); Type IVb pilin (flp); LecA and LecB lectins; or the alginate o-acetyltransferase AlgF^61–65 (see “notes” column in DataSet S1). Additionally, most of the genes significantly downregulated upon BZs exposure are involved in iron uptake (synthesis of pyochelin, receptors for pyoverdin, etc., for a total of ten genes). Finally, as can be seen in Fig. 5, KEGG functional enrichment analyses (www.kegg.jp/kegg/kegg1.html) revealed that biofilm formation was the pathway with higher number of differentially expressed genes (e.g. 29 genes belonging to the biofilm formation pathway, KEGG ID pae02025), from a total of seven significantly enriched pathways, including exopolysaccharyde synthesis or chemotaxis, obviously related to biofilm formation, among others.

Fig. 5.

KEGG enrichment analysis of DEGs in P. aeruginosa PAO1 strain after treatment with midazolam 10 µg/ml, based on RNA-seq results. The IDs of each of the seven P. aeruginosa PAO1 enriched pathways are shown (www.kegg.jp/kegg/kegg1.html).

To confirm our RNA-seq data, three of the up-regulated (flp, glpK and arcA) and two of the down-regulated genes (fpvA and pchF) were chosen as representatives to quantify their transcription levels through real-time RT-PCR. As can be seen in the Supplementary Figure S7, real-time RT-PCR results are in line with the outputs of our RNA-seq analysis, confirming significant up-/down-regulations of the selected genes upon exposure to midazolam 10 µg/ml.

Discussion

Despite data supporting that sedation of critically ill patients with BZs has undesired secondary effects with potential severe consequences, their use is still widespread^66,67. BZs sedation is associated with delayed awakening and extubation, higher risks of delirium and cognitive dysfunction, longer hospital stays, and increased mortality^8,68,69. Moreover, BZs have been linked to increased risks specifically for VAP^7–9, because of immunity alterations and a boosted penetration of opportunistic pathogens to low airways^12–19. The lower cost and more versatile administration of BZs compared to propofol for instance, the reluctance to change year-long established routine protocols, and the fact that the mentioned evidence regarding their undesired side effects is not unequivocal yet, explains the lack of a consensus for BZs substitution^66,67. In view of these facts, we sought to complete knowledge on the topic from a new perspective, i.e. the potential existence of secondary effects of BZs on P. aeruginosa biology and/or host-bacteria interplay, that may favor VAP. A similar study claiming that catecholamine inotropes (medication often administered to ventilated patients) act as contributory factor for VAP by P. aeruginosa, was published more than a decade ago⁷⁰, supporting the pertinence of our approach.

We first analyzed whether exposure to BZs could influence some intrinsic P. aeruginosa parameters with evident impact on infection output: growth rate and motility, as markers for proliferation and dissemination capacity, and antibiotic susceptibility and the frequency of appearance of spontaneous mutants, which are intimately related to development of resistance. In accordance with previous data in E. coli^71,72, our hypothesis was that BZs may impact P. aeruginosa features to increase its antibiotic resistance in an inducible fashion and/or to act as pro-mutagen agents driving to a boosted appearance of mutants. However, we did not see any change on these parameters that may aggravate VAP output. In fact, in accordance with data showing that BZs at high doses display antimicrobial activity^73,74, they showed inhibitory effects on P. aeruginosa growth in our assays, although at concentrations more than 20–fold higher than those usually attained in patients’ serum^19,30,31.

On the other hand, our results with A549 cultures indicate a significantly reduced pathogenic impact of P. aeruginosa infection upon BZs exposure. Regarding the release of IL-8, it has been reported that BZs and related ligands can reduce the inflammatory response both in isolated leukocytes ^12–19,^76,77) and animal models^{12–14,77–79}. Upon BZs binding, central and peripheral receptors of cells turn on different downstream signaling networks (NF-kappaB- or p38-dependent, among others^76,77,79 which are responsible for the mild inflammatory responses. Since A549 cells have peripheral BZs receptors and likely express at least some of the subunits of the central receptors^80,81, it was perhaps anticipated that exposure to these drugs also decreased the inflammatory activation in these cells, although these results have not been demonstrated. In this vein, our results suggest that this lower inflammatory response elicited by P. aeruginosa and also, albeit not statistically significantly, by purified LPS, seems likely explained at least partially, by the impact of BZs on A549 cells. However, we cannot rule out that BZs could also have some influence transiently reducing the intrinsic pro-inflammatory features of P. aeruginosa. Similar effects have been reported for P. aeruginosa when exposed to chlorhexidine⁸², supporting the pertinence of the idea that a drug may show side effects causing the modulation of the inflammatory behavior of a bacterium.

Similar to the IL-8 data, the decreased values of invasion and cytotoxicity could be at least partially due to BZs effects on the A549 cells. In this sense, it is known that some anti-inflammatory drugs contribute to maintain the physical integrity of the epithelial barriers, which boosts their resistance against bacteria^83,84. More specifically, it has been shown that midazolam increases the expression of protein zonula occludens-1 (ZO-1), essential for the epithelial intercellular tight junctions⁸⁵. Additionally, it is well-known that P. aeruginosa actively disturbs this kind of paracellular seals through different mechanisms to increase invasiveness and cytotoxicity^56,86–88. Therefore, it is plausible that our reduced values in these parameters were due to the mentioned increase in ZO-1 expression enabled by BZs, which would curb the alteration of tight junctions by P. aeruginosa, and thus invasion and cytotoxicity⁵⁶. On the other hand, among the P. aeruginosa genes downregulated upon BZs exposure revealed by our RNA-seq data, several actors linked to iron uptake did appear (e.g. related to pyochelin synthesis and pyoverdin uptake, DataSet S1). Since iron is a key element to enable bacterial virulence⁸⁹, one could argue that its decreased incorporation could contribute to the reduced damage caused on A549 cells.

Although our initial hypothesis was that BZs could act as inducers for P. aeruginosa virulence (as described for steroids for instance²⁵ and consequently promote and/or aggravate VAP, our results with A549 cells seem to contradict this idea. However, regardless of the actor (P. aeruginosa versus A549 cells) predominantly hosting the effects of BZs entailing a mild damage/response on the cultured cells, the possibility of this circumstance being a paradoxical trigger for VAP must be considered. Since P. aeruginosa and other so-called stealth pathogens such as Klebsiella pneumoniae exploit different strategies to dampen initial host sensing and, consequently, alter the immune/inflammatory response in order to thrive within tissues, the inflammation brake posed by BZs could be a facilitator for infection^90–92. Moreover, it has been proposed that in the beginning of an infection, cellular invasion could be more detrimental than beneficial for bacteria since internalization promotes a greater immune activation and neutrophil recruitment, which triggers desquamation of infected cells and thus physical removal of bacteria⁵⁶. Therefore, an impaired P. aeruginosa invasion as that in our results, could indirectly favor VAP by giving rise to weak alarm-immune responses and impaired cell detachment-mediated bacterial clearance. In this regard, the reported short-termmutation-driven evolution of P. aeruginosa leading to attenuated virulence in strains causing VAP⁹³ would be in line with our results and the mentioned stealth behavior.

In view of our results up to this point, we thought that if BZs had the capacity to alter P. aeruginosa behavior in the interaction with respiratory epithelium cells, this could have a measurable effect in an animal model of VAP. However, we did not appreciate any change associated to BZs in the parameters extracted from our rat VAP model, neither aggravating the outcome, nor making infection less virulent. The explanation to this is likely that the sedative treatment and ventilation periods used were too short to cause effects similar to those seen in A549 assays. Moreover, the concentrations of midazolam reaching the respiratory epithelium in rats could have been much lower than those in our cell culture experiments, and thus not enough to cause any measurable effect on P. aeruginosa. Therefore, our VAP model design was likely not ideal to reveal the altered behavior of P. aeruginosa observed in A549 assays, but neither the influence of factors potentially intervening in the longer term (days instead of hours: biofilm formation), a circumstance we did not foresee when setting up our rat model, conducted before biofilm experiments.

Moving on to biofilm-related assays, our results suggest that BZs exposure acts as a trigger for biofilm formation, in accordance with data demonstrating that a synthetic ligand of the BZs peripheral receptor orthologue present in P. fluorescens boosted biofilm production²⁴. Based on these results, one could argue that in BZs-treated individuals, the derived robust biofilms may act as improved bacterial reservoirs –out of reach for antibiotics and/or immunity elements– on the ETTs, thus favoring VAP in these patients^56–59. Biofilms growing on the inner and external surfaces of these tubes play a well-known role for VAP pathogenesis by releasing bacteria to the low airways^{45,56–59,94}. Thus, any factor promoting biofilm formation –e.g. BZs as suggested by our results– should be considered as a potential risk factor for VAP, worth to be deeply investigated. In this regard, for instance, models with superior mammals (e.g. pigs) allowing prolonged sedation/ventilation periods^95–97 would be necessary to assess a clear cause-effect relationship between BZs sedation and an increased biofilm formation on the ETT, and thus definitely prove that these sedatives pose an added risk for VAP by P. aeruginosa. Moreover, additional in vitro experiments to quantify the potential short-term effects of BZs on P. aeruginosa capacity to adhere to the ETTs surfaces, and to finely determine the dynamics of biofilm increase along different hour/days periods of BZs exposure could complement our study.

Regarding the molecular basis for our findings, a first point to take into account is the increased accumulation of c-di-GMP upon BZs exposure. It is well-known that high levels of this molecule drive the switch from planktonic to biofilm lifestyle in P. aeruginosa involving a parallel decrease in acute virulence features⁶⁰, which agrees with our biofilm- and A549 infection-related results. Interestingly, it has been demonstrated that c-di-GMP signaling does not only work through modulation of gene expression, but also impacting other variables such as protein function through post-translational mechanisms^60,98–100, a phenomenon that cannot be discarded as partial contributor to our results. In any case, a short BZs exposure exerted very significant impacts on transcriptome (more than 500 genes differentially expressed compared to controls), which supports the idea that these sedatives have a striking capacity to modulate P. aeruginosa performance. Among the upregulated genes detected from BZ-exposed cells (at least 1.5-fold compared to control), some of them have been related to adhesion and biofilm success. For instance: pprA⁶¹ and psrA⁶⁴ belonging to two-component systems linked to transition to sessile lifestyle; flp and tadC⁶⁵ involved in pili synthesis and assembly and thus in adhesion and biofilm initiation, as are lecB⁶² and lecA⁶³ lectins; algF required for alginate acetylation¹⁰¹; the enzymes of the arginine deiminase pathway (arcA-arcD), positively selected in response to the specific environment of cystic fibrosis¹⁰², etc. Accordingly, KEGG analysis of our transcriptomic data revealed that biofilm formation (pae02025 KEGG ID) was the pathway more altered upon midazolam exposure (29 genes with statistically significant variations compared to control). Other pathways related to biofilm formation^103–106 such as bacterial chemotaxis (19 genes), exopolysaccharide biosynthesis (6 genes) pyruvate metabolism (15 genes) and propanoate metabolism (14 genes), were also significantly enriched in the BZ-exposed samples, making sense of our biofilm-related results.

While all these transcriptomic changes are fairly related to biofilm formation, it is yet to be elucidated whether they could have more impact on the initial stages (adhesion to surfaces), and/or at longer terms, as our results support the hypothesis that higher concentrations of BZs (i.e., in a accumulation–type situation) may exert significant effects on P. aeruginosa promoting biofilm formation. Additionally, investigating whether these BZ-driven alterations are due to their interaction with specific receptors in P. aeruginosa as seemingly happens in other species^22–24 or due to facts such as the generation of stress in bacterial envelopes^25,26, and which could be the downstream signaling pathways involved, pose exciting research lines to be opened in future.

Besides the points stated in the last paragraphs, our study has some other limiting aspects that remain to be approached: (i) the conditions of our rat model were not ideal to assess changes in acute virulence or biofilm formation in the ETTs, and thus the influence of BZs in VAP output; (ii) our in vitro findings could not necessarily be reproduced in ventilated patients, for instance because BZs concentrations needed to modulate bacterial behavior may not be reached in the P. aeruginosa colonizer populations; (iii) the results could be variable depending on the species or even the P. aeruginosa strains (e.g. antibiotic susceptible versus resistant) and thus, our conclusions cannot be extended beyond PAO1 and PA14.

Concluding remarks

Altogether our results clearly demonstrate that BZs at clinically attainable concentrations significantly influence the virulence-related performance of P. aeruginosa in vitro, which could pose a contributory factor for VAP as previously suggested for other medications administered to ventilated individuals⁷⁰. First, our results suggest that BZs determine a stealth pathogen-like behavior of P. aeruginosa causing attenuated damages and inflammatory responses in infected cells, a circumstance that could be an advantage for VAP onset because of the weak alarm respose elicited. Our data also demonstrate that BZs boost P. aeruginosa biofilm formation in ETTs and plastic surfaces. If this was reproduced in intubated patients, these BZ-mediated robust biofilms could be an enhancer for VAP, since these communities are known to act as key reservoirs that release bacteria to low airways, promoting/aggravating the infection. Besides the accumulation of c-di-GMP as second messenger, we found significant alterations in the expression of certain genes/pathways compatible with a decreased acute virulence and increased biofilm formation, posing robust mechanistic bases for our results.

We believe that this proof of concept-type study provides data strong enough to be carefully considered as a warning sign for the generalized BZs use. Future studies will help to support the biological relevance of our findings and to ascertain whether BZs sedation definitely increases the risk for VAP by P. aeruginosa, which would be a clear point in favor of a cautious use of these sedatives.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

A.O. and C.J. designed the study and the experiments; I.M.B., E.J.-L., M.E.-S., J.S.-S., M.A.C.-M., M.A.E., A.G.-A., A.T. and M.T. conducted the experiments; G.T., S.A., A.O. and C.J. analyzed the results and wrote the initial draft of the manuscript. All authors reviewed the manuscript.

Data availability

Data that are not directly available in the manuscript will be made available upon request. Transcriptomic raw and processed data are available at the Gene Expression Omnibus (GEO) repository (NCBI), under the accession number GSE302777.

Declarations

Competing interests

The authors declare no competing interests.