Rhodococcus enclensis PDD-23b-28 was grown in 25 mL of R2A medium for 48 h at 17 ∘C, 130 rpm (Reasoner and Geldreich, 1985). Then cultures were centrifuged at 4000 rpm for 15 min at 4 ∘C. Bacteria pellets were rinsed first with 5 mL of NaCl 0.8 % and after with Volvic^® mineral water (pH =7.0), previously sterilized by filtration under sterile conditions using a 0.22 µm PES filter. The bacterial cell concentration was estimated by optical density at 600 nm using a UV3100 spectrophotometer to obtain a concentration close to 109 cell mL-1. Finally, the concentration of cells was precisely determined by counting the colonies on R2A Petri dishes.

Biotransformation. Rhodococcus enclensis PDD-23b-28 cells were re-suspended in 5 mL of 0.1 mM phenol (Fluka >99 %) solution, prepared in Volvic^® mineral water, and incubated at 17 ∘C, 130 rpm agitation for 48 h in the dark; 0.5 mL of this culture was incubated in 25 mL of the same medium and under the same conditions. In order to determine the concentration, the optical density for each strain was measured at 600 nm during the experiment. The strain concentration was ∼109 cells mL-1. The concentration ratio of bacterial cells to phenol was kept similar to that as measured in cloud water (Lallement et al., 2018b). We showed in the past that in repeated experiments identical cell / substrate ratios lead to the same biodegradation rates (Vaïtilingom et al., 2010).

A control experiment was performed by incubating phenol without bacteria; phenol concentration remained stable over time (0.1 mM of phenol was obtained at the end of the experiment). For phenol quantification over time in the incubation experiments, 600 µL samples were centrifuged at 12 500 rpm for 3 min and the supernatants were kept frozen until HPLC analysis. Complementary experiments were also performed consisting of incubation of the cells and 0.1 mM phenol in the presence of light without Fe(EDDS).

Phototransformation. A 0.1 mM phenol solution (Fluka >99 %), prepared in Volvic^® mineral water, was incubated at 17 ∘C, 130 rpm agitation for 48 h in photo-bioreactors designed by Vaïtilingom et al. (2011). OH radicals were generated by photolysis adding 0.5 mM Fe(EDDS) complex solution. The Fe(EDDS) solution (iron complex with 1 : 1 stoichiometry) was prepared from iron(III) chloride hexahydrate (FeCl3, 6H2O; Sigma-Aldrich) and (S,S)-ethylenediamine-N,N'-disuccinic acid trisodium salt (EDDS, 35 % in water). A complementary experiment was also performed consisting of incubation of a 0.1 mM phenol solution in the presence of light without an Fe(EDDS) complex.

The experimental conditions of the irradiation experiments (Sylvania Reptistar lamps; 15 W; 6500 K) are described by Wirgot et al. (2017). They mimic the solar light measured under cloudy conditions at the Puy de Dôme station (Fig. S1). The mechanism of the ⚫OH radical production under light irradiation is as follows (Brigante and Mailhot, 2015). R1Fe(III)-EDDS⟶hν[Fe(III)-EDDS]∗⟶Fe(II)+EDDS⚫R2EDDS⚫+O2⟶O2⚫-+EDDSoxR3HO2⚫⇆O2⚫-+H+R4HO2⚫+O2⚫-⟶H+H2O2+O2+-OHR5HO2⚫+HO2⚫⟶H+H2O2+O2R6Fe(III)+O2⚫-⟶Fe(II)+O2R7Fe(III)+HO2⚫⟶Fe(II)+O2+H+R8Fe(II)+H2O2⟶Fe(III)+⚫OH+-OH Using the specifications of the lamp, an overall rate constant of the photolysis of the Fe(III)-EDDS complex jR9=1.4×10-3 s-1 was calculated (Sect. S2). R9Fe(III)-EDDS⟶hν⚫OH+products Assuming steady-state conditions for ⚫OH at the beginning of the experiments (i.e., equal ⚫OH production and loss rates), an ⚫OH concentration of 8.3×10-13 M can be calculated. This concentration is at the upper limit of ⚫OH concentrations as derived from various measurements and model studies (Arakaki et al., 2013; Lallement et al., 2018a).

Photo-biotransformation. The protocols for biotransformation and photo-transformation of phenol in the presence of Fe(EDDS) as described above were combined.

Biotransformation. As for phenol, Rhodococcus enclensis PDD-23b-28 cells were re-suspended in 5 mL of 0.1 mM catechol (Fluka >99 %) solution, prepared in Volvic^® mineral water, and incubated at 17 ∘C, 130 rpm agitation for 48 h in the dark. Four experiments were carried out with different cell concentrations (109, 108, 107 and 106 cell mL-1). For catechol quantification over time in the incubation experiments, 600 µL samples were centrifuged at 12500 rpm for 3 min and the supernatants were kept frozen until LC-HRMS analysis.

Before analysis, all samples were filtered on an H-PTFE filter (pore size at 0.2 µm and diameter of 13 mm from Macherey-Nagel, Germany). Phenol detection was done on an HPLC VWR Hitachi Chromaster apparatus fitted with a DAD detector and driven by Chromaster software. Isocratic mode was used with a reverse-phase end-capped column (LiChrospher^® RP-18, 150 mm × 4.6 mm, 5 µm, 100 Å). The mobile phase was composed of acetonitrile and filtered water (Durapore^® membrane filters, 0.45 µm HVLP type, Ireland) in 25/75 ratio with a flow rate at 1.2 mL min-1. Sample injection volume was 50 µL, spectra were recorded at 272 nm and the runtime was 10 min.

LC-HRMS analyses of catechol were performed using an RSLCnano UltiMate™ 3000 (Thermo Scientific™) UHPLC equipped with an Q-Exactive™ Plus Hybrid Quadrupole-Orbitrap™ Mass Spectrometer (Thermo Scientific™) ionization chamber. The same conditions were used for analyzing EDDS. Chromatographic separation of the analytes was performed on a Kinetex^® EVO C18 (1.7 µm, 100 mm × 2.1 mm, Phenomenex) column with a column temperature of 30 ∘C. The mobile phases consisted of 0.1 % formic acid and water (A) and 0.1 % formic acid and acetonitrile (B). A three-step linear gradient of 95 % A and 5 % B in 7.5 min, 1 % A and 99 % B in 1 min, and 95 % A and 5 % B in 2.5 min was used throughout the analysis. This device was associated with a Thermo Scientific™ Dionex™ UltiMate™ DAD 3000 detector (200–400 nm).

The Q-Exactive ion source was equipped with an electrospray ionization (ESI) and the Q-Orbitrap™. The Q-Exactive was operated in either full MS-SIM; the full MS scan range was set from m/z 80 to 1200. The mass resolution was set to 70 000 fwhm, and the instrument was tuned for maximum ion throughput. The AGC (automatic gain control) target or the number of ions to fill the C-Trap was set to 106 with a maximum injection time (IT) of 50 ms. The C-Trap is used to store ions and then transfer them to the Orbitrap mass analyzer. Other Q-Exactive generic parameters were gas (N2) flow rate set at 10 a.u. (arbitrary units), sheath gas (N2) flow rate set at 50 a.u., sweep gas flow rate set at 60 a.u., spray voltage at 3.2 kV in positive mode and 3 kV in negative mode, capillary temperature at 320 ∘C, and heater temperature at 400 ∘C. Analysis and visualization of the data set were performed using Xcalibur™ 2.2 software from Thermo Scientific™.

The degradation rates of phenol and catechol were calculated after normalization based on the ratio of the concentration at time t (C) and the concentration at time t=0 (C0). The pseudo-first-order rate constants (kphenol and kcatechol) were determined using Eq. (1): 1ln⁡C/C0=f(t)=-kphenol(or  kcatechol)t.

We use a multiphase box model to compare the loss rates of phenol and catechol in the gas and aqueous phases by radicals (⚫OH, NO3⚫) in both phases and bacteria only in the aqueous phase over a processing time of 15 min to simulate chemical and biological processing in a single cloud cycle. For each set of processes (⚫OH/NO3⚫, phenol/catechol), the three terms in the following equation are calculated and the relative importance of each process is determined: 2dAromaticdtmoleccmgas3s=-kchem,gasRadical(gas)Aromatic(gas)︸loss by gas-phase chemistry-kchem,aqRadical(aq)Aromatic(aq)︸loss by aqueous-phase chemistry+kbact,aqCellAromatic(aq)︸loss by microbial processesin the aqueous phaseLWCNA0.001, where [Aromatic] denotes the phenol or catechol concentration, [Radical] the ⚫OH or NO3⚫ concentration in the gas or aqueous phase, respectively, and kchem,gas, kchem,aq and kbact are the rate constants as listed in Table S1 in the Supplement. The units of the aqueous-phase processes are converted into the same units as the gas-phase processes (molec cm-3 s-1), with LWC (liquid water content =9.7×10-7 L(aq) L(gas)-1), NA=6.022×1023 molecules mol-1 (Avogadro constant) and 0.001 to convert from L to cm3.

The pH value of cloud water is assumed to be constant (pH =4) to represent conditions of a continental, moderately polluted cloud. It should be pointed out that the choice of the pH value in the simulations does not affect the results for a wide range of pH values (3<pH<6), as this is typical of clouds influenced by marine and continental air masses (Deguillaume et al., 2014). None of the parameters in Eq. (2) is pH dependent within the range relevant for cloud water (cf. Sect. S3.3). In addition to the data for Rhodococcus obtained in the current study, we also include literature data on the biodegradation of phenol and catechol by Pseudonomas putida and Pseudonomas aeruginosa (Sect. 3.2), which are usually more abundant in the atmosphere than Rhodococcus.

The processes considered in the gas and aqueous phases are summarized in Table S1 and Fig. 1. In both phases, the reaction of phenol with ⚫OH is assumed to yield 50 % catechol; other products of these reactions are not further tracked in the model. The reaction of phenol with NO3⚫ results in nitrophenols (Bolzacchini et al., 2001; Harrison et al., 2005); the loss of these products is not explicitly included in the model either as we solely focus on the comparison of the degradation rates. Recently, it was suggested that the reactions with ozone and HO2⚫/O2⚫- might represent major sinks (∼50 % and ∼20 %, respectively) of catechol in the aqueous phase (Hoffmann et al., 2018). However, the only available rate constant for the ozone reaction was derived at pH =1.5 by Gurol and Nekouinaini (1984), who postulate that at higher pH (∼5–6), the reaction with ⚫OH likely dominates the overall loss. Therefore, in our base case simulations, we limit the reactions of phenol and catechol to the reactions with ⚫OH and NO3⚫ radicals. Sensitivity studies including the HO2⚫/O2⚫- and O3 reactions are discussed in the Supplement (Sect. S4).

Microbial activity in the aqueous phase by Rhodococcus and Pseudonomas is usually expressed as rates (mol per cell per hour) (Vaïtilingom et al., 2013). We converted these experimentally derived rates into “rate constants” (liters per cell per hour) in order to adjust them to the substrate and cell concentrations as assumed in the aqueous phase in the model (Sect. S3.2), equivalent to the treatment of chemical processes. In order to account for the numerous additional loss processes of ⚫OH(aq) and NO3⚫(aq) in clouds, sinks for both radicals have been added: a general rate constant of ⚫OH with total water-soluble organic carbon (WSOC) (kOH,WSOC=3.8×108 M-1 s-1) lumps the main loss processes of ⚫OH in cloud water (Arakaki et al., 2013); assuming an average WSOC concentration of 5 mM results in a first-order loss process of kOH=2×106 s-1. The main losses of NO3⚫(aq) are likely reactions with halides (Herrmann et al., 2000); as a proxy, we assume here a first-order loss process (kNO3=105 s-1), reflecting the sum of the major NO3⚫(aq) sinks. These lumped sink processes lead to aqueous-phase radical concentrations of [⚫OH(aq)]day∼10-15 M and [NO3⚫(aq)]night∼10-14 M, respectively, in agreement with predictions from previous model studies (Ervens et al., 2003). Kinetic-phase transfer processes between the two phases are described for the radicals and aromatics based on the resistance model by Schwartz (1986); all phase transfer parameters (Henry's law constants KH, mass accommodation coefficients α and gas-phase diffusion coefficients Dg) are summarized in Table S1.

Initial concentrations of 4 ppt catechol and phenol are assumed in the gas phase that partition between both phases and are chemically consumed over the course of the simulation (15 min). These initial mixing ratios correspond to equivalent aerosol mass concentrations on the order of several 10s ng m-3, in agreement with measurements of phenol compounds in aerosol samples (Bahadur et al., 2010; Delhomme et al., 2010) and nanomolar concentrations in cloud water (Lebedev et al., 2018). It should be noted that the assumption about the initial aromatic concentrations does not affect any conclusions of our model studies, as we compare the loss fluxes of all processes in a relative sense. Two simulations are performed for each set of conditions to simulate day- or night-time conditions, respectively, that only differ by the radical concentrations ([⚫OH]day=5×106 cm-3; [NO3⚫]night=5×108 cm-3) that are constant throughout the simulations. Two types of bacteria are assumed (Rhodococcus and Pseudomonas). They have been found to contribute 3.6 % and 19.5 % to the total number concentration of bacteria cells isolated from cloud waters and present in our lab collection. Using a typical cell concentration in cloud water of 6.8×107 cell L-1 (Amato et al., 2017), the assumed bacteria cell concentrations in the model are 2.7×106 cell L-1 and 1.3×107 cell L-1 for Rhodococcus and Pseudomonas, respectively. The simulations are performed for the conditions for monodisperse droplets with a diameter of 20 µm. The drop number concentration of 220 cm-3 results in a total liquid water content of 0.9 g m-3. These parameters do not change over the course of the simulation.

Abiotic degradation. In the presence of light and Fe(EDDS), phenol concentration decreases with time in the first 2 h of the experiments and then remains rather stable (Fig. 2). In parallel, catechol, the first intermediate of phenol transformation, is formed (Fig. S2a) and accumulates over time. Catechol concentration is quite low because it is further oxidized over time to yield CO2. Phenol degradation slows down after 2 h due to the lack of OH radical production resulting from the destruction of the EDDS ligand with time (Fig. S2b). Phenol is not directly photolyzed in the presence of light, while it is oxidized in the presence of Fe(EDDS) complex (Figs. 2 and S2).

Biotic degradation. In the dark, phenol is biotransformed by Rhodococcus enclensis cells (Fig. 2) and completely degraded after 5.5 h. A lag time of about 2.5 h is observed, during which phenol is degraded extremely slowly. This is a well-known phenomenon under lab conditions corresponding to the induction period of the gene expression (Al-Khalid and El-Naas, 2012). Catechol is slowly formed in parallel until t=3.5 h and is further biodegraded when bacteria have started to be more active (Fig. S2a).

Abiotic and biotic combined transformation. When light (in the absence of Fe(EDDS)) is present, no major change is observed for the biodegradation of phenol by Rhodococcus enclensis (Fig. 2); the lag time is still observed. When light and Fe(EDDS) are present, the lag time is no longer observed and the degradation of phenol is completed within 2.5 h instead of 5.5 h when the bacteria are in the dark. The microbial activity compensates for the limitation of radical processes due to the destruction of the Fe(EDDS) complex (after 2 h). In parallel, the production of catechol is increased compared to biotic or abiotic conditions alone (Fig. S2a). Catechol accumulates over approximately 3 h, after which it decreases. As observed previously, this decrease is likely a result of the bacterial activity.

Comparison of the rates of phenol transformation under the different conditions. If we consider the numerous uncertainties, the rates of transformation under abiotic, biotic and combined conditions are within the same order of magnitude, namely ∼10-5 mol L-1 h-1 (Table 1). Biotic and combined conditions can be further compared in more detail by normalizing the transformation rates with the exact number of cells present in the different incubations (three biological replicates for each condition). Note that the number of cells varied from 4×108 to 8×109 cell mL-1. After normalisation to the cell concentration used in the individual experiments, it is evident that the rates of phenol transformation are very close to each other and in the range of 10-16 mol per cell per hour (Table 2).

As catechol is an intermediate of phenol transformation, we measured its biotransformation rate by Rhodococcus enclensis under dark conditions. When the cell concentration was 108 or 109 cell mL-1, the catechol biodegradation was too fast to be detected within the time resolution of the experiments (Fig. 3). We performed various experiments with reduced cell concentrations, from 107 to 106 cell mL-1 (Fig. 3). Finally, we used the results corresponding to 107 cell mL-1 to derive the initial rate of catechol biotransformation. It was estimated as (15±0.5)×10-16 mol per cell per hour. This value is 8.5 times higher than the biodegradation rate of phenol and was used in the model (Sect. 3.2).

Straube (1987) showed that the activity of the catechol-1, 2-dioxygenase of Rhodococcus sp. P1 was higher than that of its phenol hydroxylase. This trend is in agreement with our results, as we know from the genome sequencing of our Rhodococcus enclensis strain that a catechol-1,2-dioxygenase is involved (and not a catechol-2,3-dioxygenase) (Lallement et al., 2017). As opposed to the results for phenol in Fig. 2, it can be seen in Fig. 3 that no lag time is observed for catechol biodegradation. This suggests that the first step of oxidation of phenol to catechol by a phenol hydroxylase might be a limiting step as it needs to be induced, while the second step – corresponding to the opening of the ring cycle by a catechol-dioxygenase – is not induced and, thus, is faster.