Unlabelled 2,4-D (chemical purity > 99%, δ¹³C = −29.1‰) was purchased from Sigma-Aldrich Co., Ltd. and ring-U-labeled ¹³C-2,4-D (99% of chemical purity, isotopic enrichment > 98%) was obtained from Dislab'system (France). Before any experiments, the ¹³C-2,4-D level of enrichment was checked by GC-IRMS (see description below).

All cultures were grown in a minimum medium (MM) containing K₂HPO₄ (1.5 g.L⁻¹), KH₂PO₄ (0.5 g L⁻¹), (NH₄)₂SO₄ (1 g.L⁻¹), MgSO₄ (H₂O)₇ (207 mg.L⁻¹), ZnSO₄ (H₂O)₇ (200 μg.L⁻¹), MgCl₂ (H₂O)₄ (10 μg.L⁻¹), H₃BO₃ (5 μg.L⁻¹), CoCl₂ (H₂O)₆ (25 μg.L⁻¹), CuSO₄ (100 μg.L⁻¹), NiCl₂ (H₂O)₆ (5 μg.L⁻¹), FeSO₄ (H₂O)₇ (250 μg.L⁻¹), and EDTA (ED4S) (125 μg.L⁻¹). 2,4-D (250 mg.L⁻¹) was added as the sole carbon source (the amount of C from EDTA was considered negligible compared to that in 2,4-D, 2,000 times higher). Two types of incubations were performed. Incubations with unlabelled 2,4-D were performed in Erlenmeyer's containing 200 mL of MM with or without (Control) 200 g of fine sterilized sand (sable de Fontainebleau, 100–200 μm, Prolabo). Incubations with ¹³C-labeled 2,4-D (δ¹³C = 134‰) were performed in serum bottles (120 mL) with Teflon® rubber stoppers crimped on with aluminum seals containing 50 mL of MM with or without (Control) 50 g of sand. There were 4 replicates per treatment in the unlabelled and ¹³C-labeled 2,4-D experiments. The sand was completely saturated and there was approximately 80 mL of free water phase. So, it was possible to shake and sample the liquid medium in the same way as for the control samples (without sand).

The inoculation of the pure C. necator JMP134 strain (courtesy of F. Martin-Laurent, INRA Dijon, France), collected at the end of the exponential growth phase, was done at a level of 10⁵ CFU.mL⁻¹. Pre-cultures were performed in a TY liquid medium: 5 g.L⁻¹ of tryptone (δ¹³C = −23.1‰), 3 g.L⁻¹ of yeast extract (δ¹³C = −26.1‰) and 50 mg.L⁻¹ of 2,4-D. According to the C content of each substrate, the average isotopic enrichment of the preculture was −23.9‰. The inoculum was rinsed once with phosphate buffer to limit the addition of C from the TY medium in the MM. All batch cultures were performed at 25°C in the dark on a rotary shaker at 150 rpm.

For all experiments (unlabelled and ¹³C-labeled 2,4-D experiments) we estimated 2,4-D (% of initial amount) and bacterial concentrations [Log(CFU).mL⁻¹] after 1, 2, 3, 5, and 10 days of incubation by measuring the optical density in the liquid medium using a Lambda 5 (Perkin-Elmer) spectrophotometer (λ_2,4−D = 282 nm and λ_Bacteria = 600 nm). At each sampling date, 4 replicates of each treatment were destructively sampled for microscopy and carbohydrates measurements (unlabeled 2,4-D experiments) or for Fatty Acid Methyl Ester (FAME) profiling (¹³C-labeled 2,4-D experiments).

Scanning Electron Microscopy (SEM) of sand surface was performed on the unlabelled 2,4-D experiments at each sampling date. In order to preserve the original structure of the sample we used a SEM equipped with a low temperature system (Chenu and Tessier, 1995). Sand samples of about 1 mm³ were mounted on the metal stubs of a specimen holder (Hexland cryostrans system CT 1,500) and were cryofixed by immersion in a nitrogen slush at −200°C, transferred to the microscope preservation chamber at −180°C where they were fractured with a cold blade. Ice was partially sublimated under vacuum at about −80°C for 20 min, and the disappearance of ice in the pores of the samples was controlled by observation under the microscope. The surface of the sample was coated at −180°C with gold, and the samples were then introduced in the refrigerated columns of the microscope (Philips SM 525 SEM) to be observed at a temperature < −160°C. For each case date of sampling three replicates were observed.

Measurements of the amount and the ¹³C enrichment of the CO₂ evolved were performed during the ¹³C-labeled 2,4-D experiments. At each sampling date, 3 gas samples were taken from the serum bottle headspace and the CO₂ was directly quantified with a micro-GC (Agilent 3,000A) equipped with a HP-Plot-U column with helium as the carrier gas. The gas sample was introduced into the micro GC via a built-in vacuum pump. The column and the detector were maintained at 80°C. The isotopic content of the CO₂ was measured with a GC-IRMS (Isochrom Optima, Micromass) on 3 other gas samples injected with a syringe into the GC. The GC was equipped with a Haysep Q column (Valco Instruments Co. Inc.) and helium as the carrier gas. The injector and detector were maintained at 180°C. The column temperature was programmed at 80°C then ramped at 30°C per min to 160°C. The δ¹³C standard deviations were 0.1‰or better. At the beginning of the experiment and after measuring CO₂, all the flasks were flushed with reconstituted air (19% O₂, 81% N₂).

Carbohydrate concentrations of each sample were measured as hexoses by the phenol-sulfuric method as previously described by Dubois et al. (1956). Briefly, 2.5 μl of phenol (80%), and 250 μl of concentrated sulfuric acid were added to 100 μl of each cell wall fraction. Following incubation at 30°C for 30 min, color was detected by an automated plate reader at 490 nm. The carbohydrate concentration was determined using glucose controls (37.5, 75, 150, and 300 g.mL⁻¹). Three analytical replicates were performed for each experimental replicate and each date of sampling.

At each sampling date, cells from control samples were harvested by centrifugation (3,500 g for 20 min at 4°C) using a Beckman-Coulter Adventi J20 XP1 and then lyophilised before lipid extraction (see protocols described below). Samples containing sand were first filtered (GF/F, < 0.7 μm, Whatmann) and then lyophilised with the filter. For ¹³C-labeled 2,4-D experiments, the 4 replicates samples were put together to obtain a sufficient amount of biomass. The total lipid fraction was preferred to the PLFA method because it may also contain information about 2,4-D and its metabolites, as these are extractable in the FAME extractants (Lerch et al., 2007). Briefly, samples were extracted with a mixture of dichloromethane and methanol by using the modified Bligh/Dyer method as described previously by White et al. (1979). The total lipids obtained were transesterified with BF₃/Methanol to recover the FAME (Morrison and Smith, 1964) and 2,4-D as methyl-2,4-D (Cochrane et al., 1981) in hexane/ether.

The FAME were separated and quantified by GC with flame ionization detection (GC-FID), identified by mass spectrometry (GC-MS) and their individual isotopic analysis was done by isotopic ratio mass spectrometry (GC-IRMS) as follows. The fatty acid methyl esters were analyzed by capillary GC-FID on a Hewlett-Packard model 5890 series II chromatograph with a 60 m SGE BPX5 column (0.32 mm inside diameter and 0.25 μm film thickness) with helium as the carrier gas. The injector and detector were maintained at 280°C. The column temperature was programmed at 50°C for 1 min, then ramped at 5°C per min to 300°C, followed by an isothermal period of 10 min. Samples were injected in splitless mode. The preliminary peak identification was done by comparison of retention times with commercial standards for FAME (BAME Mix, Supelco) and some molecules, such as 2,4-D, 2,4-DCP (2,4-dichlorophenol) and 3,5-CC (3,5-dichlorocatechol). Results were expressed as equivalent peak responses to these standards. Peak identification was completed by GC-MS of selected samples with a Hewlett-Packard model 6890N gas chromatograph interfaced to a HP 5373 quadrupole mass spectrometer by using the same column and temperature program as previously described. The electron impact ion source was maintained at 250°C, and the quadrupole temperature was 150°C. Mass spectra were obtained by electron impact at 70 eV.

For ¹³C-labeled 2,4-D experiments, the individual δ¹³C measurement of each FAME was performed in triplicate on an Isochrom III Isotopic mass spectrometer (Micromass-GVI Optima) coupled to a GC HP5890 with the same column and conditions as those previously described for GC-MS analysis. The column effluent was combusted on-line in an oxidation oven (copper-nickel-platinum catalyst at 980°C) and passed through a reactor with elemental copper (600°C) to reducing NOx and remove surplus O₂. Combustion gases were purified in a cryogenic trap containing liquid N₂ (−100°C).

The standard notation for expressing high-precision gas IRMS results in δ is defined as follows:

where R_Sample and R_VPDB are the ¹³C/¹²C isotope ratios corresponding respectively to the sample and to the international internal standard Vienna PeeDee Belemnite (R_VPDB = 0.0112372 ± 0.0000090).

The derivatization of the fatty acids introduces one additional carbon which is not present in the parent compound and which alters the original isotope ratio of the fatty acids. However, the derivatization process introduces a distinct reproducible fraction that is constant for each fatty acid (Goodman and Brenna, 1992). The measured isotope ratios of the FAME were corrected for the isotope ratio of the methyl moiety to obtain the isotope ratios of the fatty acids. This was done by using the Formula (2):

where δ¹³C_FA is the δ¹³C of the fatty acid, C_n is the number of carbons in the fatty acid, δ¹³C_FAME is the δ¹³C of the FAME, and δ¹³C_MeOH is the δ¹³C of the methanol used for the methylating reaction (−63.2‰) to calculate the isotope ratios of the fatty acids.

Considering the 2,4-D as the sole carbon source in the medium, we used the mass conservation Equation (3) and the mixing isotopic mass Equation (4) in order to calculate the amount of ring-C and chain-C in a considered fraction:

where C_Fraction is the amount of total C, C_Ring is the amount of C coming from the 2,4-D benzenic ring (ring-C) and C_Chain is the amount of C coming from the 2,4-D acetate chain (chain-C). By simplifying equations (3) and (4), the equations giving C_Ring and C_Chain in a given fraction are the following:

where δ¹³C_Fraction is the measured value, δ¹³C_Chain (−32.1‰) and δ¹³C_Ring (+177‰) are the isotopic signatures of the acetate chain and benzene ring previously determined by Lerch et al. (2007). Equations 5 and 6 were used for every FAME analyzed and the CO₂ evolved. It is important to notice that the C of the inoculum was neglected since it was at least 100 fold lower than the amount of biomass C after 1 day of incubation.

Significant differences among treatments (with or without sand) were tested by ANOVA followed by Tukey's HSD test. Principal component analysis (PCA) was performed both on total FAME profiles and on the percentage of C_Ring incorporated into each FAME. All multivariate analyses were performed using the “ade4TkGUI” package in R while graphical representations were performed with SigmaPlot 11.0. All statistical analyses were performed with R software (version 2.12.0 R Development Core Team, 2008).

Electron microscopy observations revealed that C. necator JMP134 was able to form a biofilm in a minimum medium containing only 2,4-D as C source and sand as a solid phase. After just 1 day of incubation, bacteria (about 1 μm diameter) and exopolymers (thin white filament) were detected on the surface (Figure S1). After 3 days, the density of sessile bacteria increased slightly as well as the amount of polymers between sand grains. After 5 days, sand grains were totally recovered by a film of polymers of approximately 1 μm thick. When analyzing sand grains at a larger scale (Figure 1), we observed that the porous media became totally clogged by biopolymers after 5 days of incubation.

The concentration of 2,4-D in solution was always (P < 0.001) lower in samples containing sand than in control samples (Figure 2A). After 1 day of incubation, the amount of 2,4-D in samples with sand was already significantly lower than for control samples. For both curves, a minimum was reached after 3 days of incubation. At the end of the experiment, the 2,4-D concentrations in solution were 5 ± 3% and 30 ± 3% of the initial amount in samples with or without sand, respectively. The amount of 2,4-D mineralized throughout the experiment was always higher (P < 0.001) in samples containing sand (Figure 2B). After 10 days of incubation, the cumulative amounts of CO₂ evolved were 68 ± 3% and 60 ± 2% of the initial amount of C in the medium in samples with or without sand, respectively.

It should be mentioned that whatever the sampling date, the lipid fraction of control samples did not contain any residues of 2,4-D or its metabolites (data not shown). In contrast, the analysis of the lipid fraction extracted from samples containing sand revealed the presence of 2,4-D (Figure S2). Other 2,4-D metabolites were not detected. The quantity of 2,4-D in this fraction corresponded to 9 ± 3% of the initial amount on the first day and then remained stable between 13 ± 4 and 15 ± 4% of the initial amount from day 2 onwards.

The concentration of suspended cells in solution was always (P < 0.001) lower in samples containing sand than in control samples (Figure 3A). The planktonic bacterial concentration reached a plateau after 3 days for control (approximately 9 × 10⁷ cells.mL⁻¹) and 2 days for sand samples (2 × 10⁷ cells.mL⁻¹). It should be noted that no significant differences were observed between unlabelled and ¹³C-labeled experiments, indicating that both incubations were similar. The total amount of FAME (detailed below) was similar in both treatments throughout the incubation (Figure 3B), indicating similar microbial biomass in planktonic and biofilm cultures. The maximum amount of FAME was reached after 5 days of incubation with 243 ± 14 μg and 241 ± 17 μg per replicate for sample with or without sand, respectively. The amount of total carbohydrates (Figure 3C) was also significantly (P < 0.001) higher in the biofilm culture than in the suspended cell cultures from 3 to 10 days of incubation (approximately 47 to 63% more).

The 10 most concentrated FAME identified by GC analyses are expressed in percentage of their relative amount (Figure S3). Five other FAME (iC16:0, iC17:0, C17:0, C18:0, C18:1ω9t and C19:0) were present at trace levels (relative amount < 0.5%) and were not represented. The ANOVA revealed significant differences among sampling dates, but, no significant differences between sessile bacteria (control) and sessile + planktonic bacteria (+ sand), except at day 10. At this sampling date, the relative amount of C16:0 was significantly lower and that of the cycC17:0 was significantly higher for samples containing sand grains. In both treatments, four major FAME (C16:1ω9c, C16:0, cycC17:0, and C18:1ω9c) accounted for approximately 90% of total FAME. Six minor fatty acids (C14:0, iC5:0, aC15:0, C15:0, C18:0, and cycC19:0) were found at a level of approximately 10%. The proportion of three of the major FAME (C16:1ω9c, cycC17:0, and C18:1ω9c) clearly evolved with time (P < 0.001). The evolution of C16:0 proportion was less significant (P < 0.05). It should be mentioned that the cyclopropyl/precursor ratio (cycC17:0+cycC19:0/C16:1ω9c + C18:1ω9c) increased significantly (P < 0.001) for both treatments (Table S1). This ratio was significantly higher (P < 0.01) for sessile bacteria, especially from day 3 onwards. PCA performed on FAME profiles (Figure 4) showed change with time (F1 = 39%) and a distinction between planktonic and sessile cells (F2 = 37%). A redundancy analysis showed that 95% of the variability was explained by these two factors.

The isotopic analysis of the CO₂ did not show significant differences in the origin of C respired between samples with or without sand (Figure S4). In both treatments, the proportion of C originating from the benzenic ring (C_Ring) was slightly lower in the CO₂ (≈ 73%) than in the 2,4-D (75%), indicating that C. necator JMP134 preferentially used the C originating from the acetic chain (C_Chain) for energy.

The isotopic analysis of individual FAME did not reveal significant differences between samples containing sand and control samples. As a result, the proportion of C originating from the benzenic ring (C_Ring) and the acetic chain (C_Chain) were similarly incorporated in sessile and planktonic bacteria. For both treatments, the proportion of C_Ring in FAME increased with time (Figure S5). PCA performed on the proportion of C_Ring in FAME (Figure 5) showed a small change with time (F1 = 81%) and a little distinction between planktonic and sessile cells during the first 2 days of incubation (F2 = 7%).

The similarity of the total FAME amounts in both treatments (with or without sand) suggested that there was no difference in the total microbial biomass. Spectrophotometric measurements of the bacterial density in the liquid phase showed that C. necator JMP134 cultures in the presence of 2,4-D presented similar growth as reported previously (Lerch et al., 2007, 2011). For both treatments, the exponential phase reached its maximum after 3 days followed by a stationary phase that lasted for the next 2 days. Thereafter, there was a similar stage of cellular death with decreasing OD. While both treatments showed the same kinetics, the concentration of suspended bacteria in samples containing sand was approximately 25% that of control samples. In parallel, electron microscopy observations showed that bacterial cells were already attached to the sand grain surfaces after 1 day of incubation. Thus, we conclude that from the early growth stage, most of the cells were attached to the sand grains than being in suspension. By using mass conversion of cell density and with comparison to optical density measurements, we estimated that approximately 80% of the cells in samples containing sand were fixed to the mineral phase from 2 days onwards. Assuming sand grain were perfect spheres (specific area of 0.02 m².g⁻¹ and density of 2.7 g.cm⁻³), the surface density of sessile bacteria was estimated to be 0.02 to 0.1 cell.μm⁻², which is consistent with SEM analyses.

Most biotreatment experiments using biofilms are enhanced by previous encapsulation and/or support material with high surface area in conjunction with co-substrate to enhance the immobilization of the cells (e.g., Yanez-Ocampo et al., 2009). Our study showed that C. necator JMP134 is capable to rapidly attach to sand grains to reach a density of 0.1 cell per μm² and develop a thin biofilm using 2,4-D as a sole carbon source of a biofilm with the cell. The ability of this bacteria to form biofilm has been reported previously in the literature. Alcaligenes eutrophus AE1308, a transconjugant of A. eutrophus CH34 and A. eutrophus JMP134 (former name of C. necator JMP134) have been successfully used to form active biofilm into membrane reactor for wastewater treatment (Diels et al., 1995). Other studies carried out in different environments showed that strains similar to C. necator JMP134 were able to form biofilms: Ralstonia eutropha (Steinle et al., 1998; Simkus et al., 2004), Ralstonia solanacaerum (Tans-Kersten et al., 2001; Kang et al., 2002), A. eutrophus (Mergeay et al., 1985; VanRoy et al., 1997), Alcaligenes denitrificans (Elvers et al., 1998), Alcaligenes xylosoxydans (Meade et al., 2001). We did not measure exactly the surface area and thickness of the biofilm. However, electron microscopy images showed that after 3 days of incubation, the thickness of the biofilm was approximately that of the bacteria (c.a. 1 μm), while the surface area of sand grains was almost entirely covered with polymers. Karthikeyan et al. (1999) reported that the thickness/area ratio of a biofilm was correlated to physicochemical characteristics of the organic pollutants being degraded by the biofilm, such as the octanol/water partition coefficient (K_OW). It is hypothesized that a close positioning of cells is a disadvantage when cells compete for a labile carbon source. The more recalcitrant and hydrophobic is a molecule, the higher thickness/surface area ratio of the biofilm. This finding is in line with the observation we did with 2,4-D, a very soluble molecule: the biofilm surrounding cells was relatively thin but covered a large part of the grain sand surface. We also observed that after 5 days, all pores were clogged, which certainly modified the diffusion of substrates and metabolites (Chenu and Roberson, 1996) as well as exchanges between the liquid and the solid phases.

Cell immobilization has been employed for biological removal of pesticides because it allows maintaining a catalytic activity over long periods of time. Advantages over cultures using suspended cells include greater cellular content in the support, enhanced cellular viability (weeks or months) and greater tolerance to high concentrations of pollutants. For example, a membrane reactor composed of a microbial consortium showed a better removal of 2,4-D than activated sludge (Gonzalez et al., 2006). Our results suggest that even a single population of bacteria is able to remove more 2,4-D when organized into a biofilm. Some studies using bacterial strains similar to C. necator JMP134 showed that the cell immobilization enhanced the removal of monochlorophenol (Balfanz and Rehm, 1991; Menke and Rehm, 1992) or dichlorophenol (Steinle et al., 1998). Here, we showed that the greater 2,4-D decrease in the liquid medium was partly due to a higher mineralisation. It could be hypothesized that metabolic activity of the cell is higher when immobilized than in planktonic form (Kinniment and Wimpenny, 1992). In the present study, the total amount of biomass was not affected by the bacterial attachment onto the surface. However, SEM observations and carbohydrate measurements suggested that higher exopolymer amounts where produced when C. necator JMP134 was grown with sand. This EPS production likely required more energy, which could also explain the relative increase in 2,4-D mineralisation. The better efficiency of 2,4-D uptake may be related to the exchange of metabolites between cells and/or the nutrient recycling after cell death (Laspidou and Rittmann, 2002, 2004). These possible advantages of sessile bacteria over their planktonic counterparts may explain results obtained in this study and also those previously published (Wolfaardt et al., 1998; Karthikeyan et al., 1999).

While the additional 2,4-D removal from the liquid medium was approximately 30% of added substrate in samples containing sand, only 10% more of added 2,4-D was mineralized. 2,4-D directly from the immobilized cells represented approximately 15% of the added substrate after 2 days of incubation (data not shown), which could explain the difference between the 2 treatments. Since 2,4-D is not adsorbed on the sand grains (preliminary experiments, data not shown) and the accumulation of 2,4-D into the cells is unlikely to happen (Lerch et al., 2007), we hypothesize that the retention of the pesticide was mainly due to the exopolysaccharides (EPS) matrix. It seems unlikely that EPS surrounding directly the immobilized cells may play a major role with regard to the low thickness (1 μm) of the biofilm in this area. However, SEM images revealed much more EPS in the porosity of the sand grains. In these specific area, the 2,4-D diffusion coefficient in the EPS matrix could be lower than in water, as shown already for glucose (Chenu and Roberson, 1996). Many authors have considered this phenomenon as a microbial process that enables to trap nutrients in various environments (Decho, 1990). The EPS matrix may also adsorb 2,4-D in soils as it has been reported for other molecules, such as dissolved organic matter (Sherr, 1988; Kaiser and Guggenberger, 2003), and organic pollutants (Flemming, 1995; Holden et al., 1997; Spath et al., 1998; Holden et al., 2002). An accumulation of pesticide diclofop in the EPS matrix was reported by Wolfaardt et al. (1994) with no consequence on its mineralisation rate. In our study, the accumulation of 2,4-D in the biofilm was accompanied by a higher mineralisation rate and higher carbohydrate amounts that are likely involved in EPS production. Considering that retention within the EPS matrix can allow further chemical transformation that would not happen in suspension (Confer and Logan, 1998), the increase in carbohydrate production could be a storage strategy to improve microbial survival since the production of biopolymers changed the habitat locally and improve substrate immobilization. In soils, the retention of 2,4-D on such biopolymers might play an important role in the formation of non-extractable residues (Lerch et al., 2009a) but also on their mineralisation (Lerch et al., 2009b).

The ANOVA performed on each FAME did not reveal significant differences between sessile and planktonic bacteria. However, the second principal component (37% of the variability) of the PCA performed on the whole FAME profiles was partly linked to the microbial lifestyle. When comparing the cyclopropyl/precursor ratios, we observed that sessile bacteria synthetized more cyclopropyl fatty acid than planktonic bacteria after 3 days of incubation. For both treatments, important changes in FAME profile were related to the sampling date (39% of the variability). These results suggest that the C. necator JMP134 FAME composition was influenced by the growth stage as much as by its attachment to a solid surface. Such growth stage influence on FAME composition is important to consider when interpreting FAME (or PLFA) profiles in complex environments. Variations in the lipid biomarker composition are not only due to the modification of the microbial composition but also to physiological changes. For example, the sensitivity of FAME profiles to change in growth substrate has been reported (Lerch et al., 2011). According to the present study, the bacterial sessile mode of life also alters the signature of lipid biomarkers.

It should be noted that we did not analyse hydroxy-fatty acids in this study since a sylilation is a prerequisite for GC analyses and this derivatisation involves the incorporation of many exogenous C atoms with their own isotopic signature. These lipopolysaccharides (LPS) components have been reported in the FAME profile of R. eutropha B8562, a strain similar to C. necator JMP134 (Volova et al., 2006). The LPS amount is likely to increase with the formation of biofilm (Brandenburg et al., 2003; Keinänen et al., 2003). Thus, it could be hypothesized that the proportion of hydroxy-fatty acids also increases during this phenomenon. However, most of the studies using lipid biomarkers for in situ assessment of the microbial communities do not consider hydroxy-fatty acids.

The analysis of FAME isotopic signatures gives a good estimate of the whole biomass isotopic signature (Lerch et al., 2007, 2009a). Here, we did not report significant changes in the FAME ¹³C enrichment between sessile and planktonic bacteria, suggesting that the formation of a biofilm by C. necator JMP134 did not change its catabolic pathway. As observed for the FAME profiles, the main changes in the isotopic signature were due to the growth stage. Consequently, the formation of biofilms in complex environments may not alter the isotopic signature of a mono-specific population.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The handling Editor declared a shared affiliation, though no other collaboration, with the authors and states that the process nevertheless met the standards of a fair and objective review.