Strains used in this study and their origins are listed in Supplementary Table 1. Staphylococcus aureus MFP03, Staphylococcus epidermidis MFP04, Staphylococcus capitis MFP08, and Pseudomonas fluorescens MFP05 were isolated from the skin of healthy human volunteers (Hillion et al., 2013). S. aureus, S. epidermidis, and S. capitis were used as representative members of the Firmicutes phylum and P. fluorescens as a model of cutaneous proteobacterium. Corynebacterium tuberculostearicum CIP102622, originating from Pasteur institute collection, was included in this study as a member of Actinobacteria, one of the three major phyla of skin microbiota (Grice and Segre, 2011). Bacteria were grown overnight at 28°C for P. fluorescens and at 37°C for other species with shaking (180 rpm). Luria-Bertani (LB) medium was used as culture media for all bacteria and for most experiments, except for C. tuberculostearicum which was cultivated in modified LB medium supplemented with 1% Tween 80 (LBT80 medium) (Becton Dickinson, Heidelberg, Germany). From overnight cultures, fresh medium was inoculated (OD₅₈₀ = 0.08) to obtain cultures at mid-exponential phase (OD₅₈₀ ∼ 0.6). A defined volume of bacterial cultures was spread on each cellulose acetate membrane filter (pore size 0.2 μm, diameter 47 mm, Sartorius Biolab Products, Göttingen, Germany) impregnated with LB or LBT80 and pre-disposed on LB or LBT80 agar. Filters were incubated at optimum growth temperature of each testing bacteria during 4.1 generation time, resulting in a single layer bacterial population. This optimal incubation time was determined by preliminary studies. After incubation, filters with bacteria were placed on agar in one-well dishes (127.8 × 85.5 mm, Thermo Scientific Nunc, Roshester, NY, United States) and transferred into the gas exposure device (Figure 1A).

The gas exposure device (Ghaffari et al., 2005; Asimakopoulou et al., 2011) consisted of two sterile cylindrical Teflon chambers (Figure 1B), one for gNO₂ exposure and another for air exposure. These exposure chambers were used in parallel in a static incubator and maintained at the optimal growth temperature of the bacterial species. In order to mimic the interaction of gas with bacteria on the skin surface, bacteria on cellulose acetate filter were exposed to gNO₂ or air under a continuous flow during 2 h. Experiments were realized using gNO₂ 10 or 80 ppm. NO₂, N₂, and O₂ were provided by different gas bottles (Air Liquide, Mitry-Mory, France) and air by a medical grade air compressor. Gas flow rates were controlled by mass flow regulators (Alicat Scientific, Inc., Tucson, AZ, United States) in order to obtain the desired concentration of gNO₂, and 20% of O₂, with a global flow rate of 2 L/min. Temperature and relative humidity data were monitored inside the exposure chambers using a data logger (Lascar Electronics, Inc., Erie, PA, United States). After gas exposure, bacteria were resuspended by gentle agitation for 20 s per filter into physiological water and used for the further analysis.

Bacterial numeration was performed by the viable count technique after decimal dilution in physiological water and spreading onto LB or LBT80 agar with incubation at the optimal growth temperature. Results are expressed as CFU/Filter of suspension obtained after exposure. The cultivability was assessed by calculating the base 10 logarithm of the ratio of the count after exposure to air against that after exposure to gNO₂ as shown below:

Ninety-six-well Bioscreen microplates (Thermo Fisher Scientific, Roshester, NY, United States) were inoculated with bacterial suspensions in LB or LBT80 at an initial OD₅₈₀ of 0.08. Bacterial suspensions were directly obtained from bacteria grown on cellulose acetate filters exposed to air or gNO₂ and recovered as previously described. Growth kinetics were measured at OD₅₈₀ under 15 min interval using a Bioscreen C automated microplate reader (Labsystems Oy, Helsinki, Finland) allowing determination of generation time and lag phase length.

Exo-enzymatic activities were tested by culture on specific media. The lipolytic activity (lipase production) was investigated on tributyrin agar plates for Staphylococci (Xie, 2012) and rhodamine B agar plates for P. fluorescens MFP05 and C. tuberculostearicum CIP102622 (Kouker and Jaeger, 1987). Esterase production was assayed on trypcase soy agar (TSA) medium (Sigma-Aldrich, St Quentin Fallavier, France) complemented with 1% Tween 80, 1% CaCl₂, and 1% phenol red (Plou et al., 1998). Caseinase (protease) activity was tested on TSA with 10% semi-skimmed milk (Riffel and Brandelli, 2006). Determination of the lecithinase (phospholipid hydrolase) activity was performed on TSA with 5% egg yolk agar (McClung and Toabe, 1947).

Pseudomonas fluorescens MFP05 was the unique motile species in the series of selected cutaneous bacteria. This bacterial species can move by swimming, swarming, and twitching depending on the surface. These different modes of motility involve flagella, type IV pili, and surfactant secretion (Henrichsen, 1972; Déziel et al., 2001). P. fluorescens swimming was tested by surface inoculation of LB medium-agar 0.3% (wt/vol), swarming by using LB medium-agar 0.6% (wt/vol) and twitching by inoculation in depth LB medium-agar 1% (wt/vol) plates, according to Déziel et al. (2001). The diameter of the migration halo was measured after 24 h of incubation at 28°C.

Biofilm formation was investigated in brain heart infusion (BHI) medium enriched with 2% (wt/vol) glucose for Staphylococci, LB medium for P. fluorescens MFP05 or LBT80 medium for C. tuberculostearicum CIP102622. Bacterial pellets obtained after gas exposure were diluted into 1 ml of fresh medium at OD₅₈₀ = 0.8 and grown in 24-well glass-bottomed microplates (SensoPlate^TM, Greiner Bio-One, Courtaboeuf, France) for 24 h at their respective optimal growth temperatures. After incubation, biofilms were rinsed twice with sterile physiological water in order to eliminate planktonic bacteria and traces of growth media. Biofilms were stained by adding 5 μM of SYTO 9 green fluorescent nucleic acid stain (Invitrogen, Molecular Probes, Carlsbad, CA, United States) prepared in sterile physiological water and incubated at room temperature for 15 min in the dark and then washed again with physiological water. Biofilm images were acquired with a Zeiss LSM710 confocal laser scanning microscopy (CLSM) (Carl Zeiss, Oberkochen, Germany) using laser excitation at 488 nm, emission band pass 500–550 nm, and an ×63 oil immersion objective. At least five image stacks (image taken every micrometer throughout the biofilm depth) from five independent experiments with two replicates for each experimental condition were used. 3D images were generated using the Zen 2.1 software.¹ Biofilm average thickness, biovolume, and roughness (Ra) were determined using the COMSTAT2 software² (Heydorn et al., 2000; Vorregaard, 2008).

Antimicrobial susceptibility testing was done on Mueller-Hinton II agar (MH-II agar) (Becton Dickinson, Heidelberg, Germany) for all strains except for C. tuberculostearicum CIP102622 which was grown on a modified MH-II agar completed with 1% Tween 80. The disk diffusion technique was used according to the EUCAST (Société Française de Microbiologie, 2019). All tested antibiotic disks (Oxoid, Basingstoke, Hampshire, United Kingdom) are listed in Supplementary Table 2. Antibiotics susceptibility profiles were interpreted on the basis of the EUCAST guide (Société Française de Microbiologie, 2019).

3-nitrotyrosine, a nitrosative stress biomarker, was quantified with the nitrotyrosine competitive ELISA kit (Abcam, Cambridge, United Kingdom), usually used for eukaryotic cells. The protocol was adapted for bacteria. After air or gNO₂ treatments, bacteria cells resuspended from the filter were centrifuged at 12,000 × g for 10 min and rinsed thrice in a PBS solution. At the third rinsing step, pellets were recovered in 1 ml of 1× sample extraction and transferred into 1 ml vial with 0.1 mm bead glass (Bertin Instruments, Montigny-le-Bretonneux, France). Bacterial lysis was performed at 4°C using a Precellys24 Bead Beater (6,800 rpm, 3 × 30 s, 120 s break) equipped with a Cryolys cooling system (Bertin Instruments, Montigny-le-Bretonneux, France). Samples were centrifugated and the supernatant was used to determine the 3-nitrotyrosine (3-NT) level according to the manufacturer’s instructions.

Bacterial membrane integrity was investigated by flow cytometry (CytoFlex S flow cytometer, as previously described) after staining of the bacteria with a combination of fluorescent dyes [SYTO 9 and propidium iodide (PI)] (LIVE/DEAD^TM BacLight^TM Bacterial Viability kit, Invitrogen, Molecular Probes, Carlsbad, CA, United States). Both dyes bind specifically to nucleic acids. Bacterial membranes are permeable to SYTO 9 (green fluorescence) that stains all bacteria, whereas PI (Red fluorescence) only penetrates bacteria with damaged membranes and quenches the SYTO 9 signal. Three populations can be distinguished corresponding to intact green bacteria, permeabilized red and orange bacteria, or unstained cells debris (Léonard et al., 2016). This procedure was impossible to apply with C. tuberculostearicum CIP102622 which, as all Corynebacteria, is known to be difficult to stain with both dyes, probably because of the specific composition of its cell wall (Neumeyer et al., 2013). Then, C. tuberculostearicum CIP102622 was only stained with PI. For each analysis, a positive control was realized using an aliquot of bacterial suspension treated with 100% ethanol for 1 h. SYTO 9 and PI were excited with a 22 mW blue laser (488 nm) and their fluorescence emission were respectively detected at 525/40 nm (green channel) and 690/50 nm (red channel). A total of 200,000 events were recorded at a flow rate of 10 μL.min^–1. Data were analyzed using the CytExpert Software.

Direct analysis by flow cytometry of bacteria without staining provides information regarding the bacterial morphology and structure. The forward scattering (FSC) fraction of light and the fraction of light scattered laterally, known as side scatter (SSC) provide respectively an estimation of the cell size and an analysis of the surface heterogeneity, called granularity (Léonard et al., 2016). Measures were performed on 200,000 events per sample recorded with the CytoFlex S flow cytometer (Beckman coulter Life science, Indianapolis, IN, United States) and the data were analyzed with the internal CytExpert Software.

Atomic Force Microscopy (AFM) imaging was performed by using a Nanoscope III Multimode microscope (Veeco instrument, Santa Barbara, CA, United States) for S. capitis MFP08 or a Nanoscope 8 Multimode microscope (Bruker Nano Surfaces, Santa Barbara, CA, United States) for S. aureus MFP03, S. epidermidis MFP04, P. fluorescens MFP05, and C. tuberculostearicum CIP102622. In all cases, a 100 μm piezoelectric scanner was employed. Imaging was achieved in air using the contact mode for all bacteria except S. aureus MFP03 for which the PeakForce^® mode was used. In the contact mode, the cantilevers, characterized by a low spring constant of 0.06 N/m, were equipped with a classical pyramidal NiSi tip. In this condition, all measurements were performed with the feedback loop on (constant force from 10^–9 to 10^–8 N). In the PeakForce^® mode, a tapping cantilever with a spring constant of 40 N/m and a classical Si tip was used. Images were obtained with a peak force tapping frequency of 2 kHz with the auto-amplitude on. All images are presented in the height mode and are top view. Flatten and three points leveling operations were realized using the Gwyddion AFM software.³ Images were also processed using a local contrast filter (kernel size 2 px, blending depth 2, and weight 1) in order to visualize the bacteria surface. Then, the local contrast image was overlayed to the topographic image using the GIMP software⁴ in order to get more meaningful image.

All experiments were carried out at least three times. For analysis and graphical presentation, GraphPad Prism^® Software (V8.3.1, San Diego, CA, United States) was used. Shapiro–Wilk normality test was used to verify normality of the data and a ratio paired t-test or a paired t-test was applied to compare air and gNO₂ conditions. To compare the results obtained with the different strains, one-way ANOVA and Tukey’s multiple comparison were performed. To evaluate the correlation between different parameters, a Pearson correlation test was used. A significant difference was considered as (^∗) for P < 0.05, (^∗∗) for P < 0.01, and (^∗∗∗) for P < 0.001.

The decrease of bacterial cultivability after exposure to gNO₂ 10 and 80 ppm in comparison to air is presented in Figure 2 and Supplementary Figure 1. A 2 h exposure to gNO₂ 80 ppm resulted in a mean cultivability loss of 1.36 LOG units for the five strains, with a maximum loss of 1.54 and 1.50 LOG units for C. tuberculostearicum and S. capitis, respectively, and a minimum of 1.11 LOG units for S. aureus. The cultivability decrease varied significantly between strains at 10 ppm (P = 0.0055) and 80 ppm (P = 0.0003). A lower LOG reduction (<1 LOG) was observed for 10 ppm, with an average cultivability loss of 0.47 LOG units and a maximum loss of 0.68 and 0.65 LOG units for C. tuberculostearicum and S. capitis, respectively. Surprisingly, after a 10 ppm exposure, P. fluorescens showed no significant loss of cultivability, unlike the other four bacterial species.

Exposure to gNO₂ altered the growth kinetics of S. aureus, S. epidermidis, S. capitis, P. fluorescens, and C. tuberculostearicum (Figures 3A–E, respectively). A 80 ppm exposure increased the lag phase of the five strains, from a minimum of +2.16 h for S. capitis to a maximum of +13 h for S. epidermidis, showing a strain-related effect of gNO₂ (P = 0.0165). At 10 ppm, the lag phase increase was less pronounced and even absent in P. fluorescens, but always variable between the strains (P = 0.0098) (Figure 3F). Except for P. fluorescens, the generation time generally tended to increase or remain unchanged, depending on the strain, for both gNO₂ tested concentrations (P < 0.0001) (Figure 3G). S. epidermidis, unlike the other Staphylococci, is difficult to grow in LB culture medium. This may explain the high variability of the obtained results as well as the significant increase in generation time and latency phase observed after exposure to gNO₂, compared to other species.

Pseudomonas fluorescens was the unique motile microorganism selected in the present study. This bacterium is known to be able to use different types of motility according to the culture conditions, namely swimming, swarming, and twitching. Bacteria exposed to gNO₂ 80 ppm remained capable to migrate using the three types of movement. The diffusion distance and aspect of halo were unchanged (Supplementary Table 3). However, the migration was delayed by 24 h compared to bacteria exposed air.

The lipase, phospholipid hydrolase (lecithinase), protease (caseinase), and esterase exo-enzymatic activities of the five selected cutaneous bacterial strains were tested in the absence or presence of gNO₂ 80 ppm. These activities were unchanged between control and gNO₂ exposed microorganisms although a 24 to 48 h delay was observed for appearance of the phenotype in comparison to bacteria in contact with air (Supplementary Table 4).

The effect of gNO₂ 80 ppm on biofilm formation was studied by CLSM in order to visualize the different morphological alterations (Figure 4A). Image analysis was leading to calculate the mean biovolume, roughness, and average thickness of the biofilm (Figure 4B). gNO₂ had limited effects on S. aureus biofilms formation and only the mean roughness was significantly decreased. Conversely, in the case of S. epidermidis the roughness was not modified by gNO₂ but the biovolume and average thickness of the biofilm were increased. S. capitis showed another reaction to gNO₂ with a decrease of biofilm biovolume and thickness and an important rise of the roughness. For P. fluorescens no significant effect of gNO₂ on the biofilm formation activity was observed. C. tuberculostearicum was also marginally affected by gNO₂ with just a decrease of the average thickness.

The sensitivity of our five bacterial strains to a wide range of antibiotics covering the principal active families was tested as recommended by the EUCAST. No effect of gNO₂ 80 ppm pretreatment was observed on the antibiotic susceptibility. All antibiotics susceptibility profiles are available in Supplementary Table 2.

The level of 3-NT is usually employed as marker of protein damage induced by RNS in eukaryotic cells and results from the action of the peroxynitrite ion (ONOO^–) generally generated from nitric oxide (NO) (Pacher et al., 2007; Teixeira et al., 2016). This method was successfully applied to the global evaluation of the effect of gNO₂ on bacterial proteins. The level of 3-NT increased significantly for all the bacterial species tested after exposure to gNO₂ 10 or 80 ppm (Figure 5). For the two concentrations, the effect of gNO₂ was species-depended (P = 0.0208 for 10 ppm and 0.0028 for 80 ppm). At gNO₂ 80 ppm, S. capitis showed a significantly greater increase (5.08 LOG units) compared with S. aureus (P = 0.0232), P. fluorescens (P = 0.0109), S. epidermidis (P = 0.0097), and C. tuberculostearicum (P = 0.0039). At 10 ppm, the least affected strain was P. fluorescens (2.24 LOG units).

Membrane permeability was assessed by flow cytometry and labeling with SYTO 9 / propidium iodide (PI) or PI alone. The membrane permeability of all strains was strongly increased after exposure to gNO₂ 10 or 80 ppm (Figure 6). The percentage of bacteria permeable to PI following exposure to air varied between species (P < 0.0001 for 10 ppm and P = 0.0331 for 80 ppm). The increase in permeability caused by gNO₂ was therefore evaluated by considering the difference of percentages measured in gNO₂ and air exposed bacteria (Table 1). This increase ranged from +61% for S. aureus to +90.1% for S. capitis. In coherence with the loss of cultivability observed at gNO₂ 80 ppm, a lower increase of permeability was observed for all bacteria exposed to this concentration of gNO₂, with a marginal effect on permeability for S. aureus (+1.8%) and P. fluorescens (+0.99%) (Table 1).

As mentioned above, for all the parameters tested, different effect of gNO₂ were observed according to bacterial species considered and the pollutant concentration. Figure 7 summarizes the P-values obtained by the multiple comparison of the strains using the Tukey’s test for all the parameters (except biofilm formation not tested at 10 ppm) and both concentrations. More significant differences appeared at 10 ppm rather that at 80 ppm, probably because 80 ppm induced a saturation effect, masking inter-species variations.

Direct analysis by flow cytometer without staining of bacteria after gas exposure was performed for the two gNO₂ concentrations: 10 and 80 ppm. Only the 80 ppm concentration induced a modification of the FSC/SSC dot plot clouds (Figures 8A–C). Changes in size and granularity appear differently depending on the species and three different behaviors were observed. In the case of S. aureus (Figure 8A) and the other studied Staphylococci (Supplementary Figures 2A,B), an increase in the mean cell size and complexity was observed. Concerning P. fluorescens, exposure to gNO₂ 80 ppm induced an increase in mean cell size and a decrease in complexity (Figure 8B). C. tuberculostearicum showed another behavior, characterized by a minor size increase and an absence of cell complexity variation (Figure 8C).

In order to complete the flow cytometry analysis and visualize possible alterations of the bacterial envelope, an AFM analysis was performed. In topographic images, depression areas were observed on the surface of many S. aureus cells exposed to 80 ppm gNO₂ (Figure 8D). Similar alterations were found in higher occurrence in S. epidermidis and S. capitis (Supplementary Figures 3A,B). Another type of alteration, characterized by appearance of pore-like structures, was observed on the surface of P. fluorescens exposed to gNO₂ 80 ppm. In addition, whereas control bacteria showed wavelet structures, this pattern was totally absent after exposure to gNO₂ 80 ppm (Figure 8E). Interestingly, no visible surface alteration was observed in gNO₂ treated C. tuberculostearicum (Figure 8F). Measurement of frictional forces revealed no difference in S. aureus (Figure 8G) and P. fluorescens (Figure 8H) after gNO₂ exposure. In contrast, chemically different surface domains observed in control C. tuberculostearicum disappeared in bacteria exposed to gNO₂ (Figure 8I), suggesting that gNO₂ induced a rearrangement of the cell wall.