Extracellular polymeric substances from soil-grown bacteria delay evaporative drying

When soils dry, water flow and nutrient diffusion cease as the hydraulic microenvironments vital for soil life become fragmented. To delay soil drying locally and related adverse effects


Introduction
Microbial life in soil faces numerous challenges, including soil water content fluctuations that fundamentally affect the hydraulic microenvironment and thus bacterial growth and reproduction (Schimel, 2018;Tecon and Or, 2017). During soil drying, the diminished water connectivity of the pore space limits nutrient diffusion and water flow. Central to these processes are the high surface tension and the low viscosity of water, which govern the dynamics of liquid connectivity in bacterial aqueous habitats. Bacteria and plants evolved a common strategy to counteract eventual disruptions in transport and hydration conditions. The resulting soil water dynamics of the rhizosphere and biological soil crusts are thus different from bulk soil conditions (Carminati et al., 2010;Rossi et al., 2018;Verrecchia et al., 1995). These modifications are commonly ascribed to extracellular polymeric substances (EPS) or mucilage released by bacteria and plant roots, respectively. Bacterial and plant-derived EPS are polymeric blends that share main chemical and physical properties , despite their high compositional diversity (Flemming and Wingender, 2010;Naveed et al., 2019). The interconnected network formed by EPS retains large quantities of water (Flemming and Wingender, 2001;McCully and Boyer, 1997;Read et al., 1999;Roberson and Firestone, 1992;Segura-Campos et al., 2014) and reduces soil hydraulic conductivity Kroener et al., 2018;Or et al., 2007;Rosenzweig et al., 2013;Zheng et al., 2018). Dissolved polymers also increase the viscosity of the soil solution Wingender, 2001, 2010;Naveed et al., 2017;Stoodley et al., 2002a) and reduce the surface tension of liquids (Cooper et al., 1981;Raaijmakers et al., 2010;Read et al., 2003). All these features decelerate soil drying and reduce fluctuations in soil transport properties.
During evaporative soil drying, water drawn from the wet soil profile is lost to the unsaturated atmosphere at the soil surface. In wet soils, the evaporative demand of the atmosphere is met by transport of water to the soil surface via capillary action (termed stage one evaporation). Once the evaporative demand at the soil surface can no longer be supplied by capillary flow, the vaporization plane recedes into the soil profile and water loss is governed by diffusion of water vapor through the air-filled pore space; so called stage two evaporation (Lehmann et al., 2008;Shokri and Or, 2011). This second stage is accompanied by substantial reduction of evaporation rates and a deceleration of soil drying that leads to hydraulic decoupling. The onset of stage two evaporation is a function of capillary forces, soil hydraulic conductivity, and the evaporative demand of the atmosphere. A mismatch between the evaporative demand and the soil water supply arises when capillary forces or soil hydraulic conductivity are too low to sustain the evaporative fluxes. As EPS were shown to lower surface tension, hence, capillary forces, and reduce soil hydraulic conductivity, they promote hydraulic decoupling by accelerating the transition to stage two evaporation. This might explain the reduced water loss during evaporation from soils inoculated with EPS-producing bacteria (Volk et al., 2016;Zheng et al., 2018), soils amended with EPS , and inoculated soil micromodels (Deng et al., 2015). Although, microbially-induced reduction of evaporative water loss is ascribed an adaptation to the challenging soil environment, the time needed for adaptation, and water related stress are mostly disregarded. The same holds true for local soil water dynamics, which are likely heterogeneous in space but are rarely quantified.
Here, we quantify the potential of bacteria to engineer their porous environment when they are exposed to repeated wetting and drying cycles. We hypothesize that hydraulic decoupling between the soil surface and the evaporation front caused by bacterial growth and EPS production shelters soil water from evaporation. To test our hypothesis, we used sand microcosms inoculated with either Bacillus subtilis NCIB 3610, which is known for complex biofilm formation or its domesticated counterpart Bacillus subtilis 168 trp + (Gallegos- Monterrosa et al., 2016;McLoon et al., 2011). Sand microcosms were subject to daily wetting-drying cycles for one week. At the end of this adaptation period, local water loss was monitored during evaporative drying with time-series neutron radiography to spatially map the distribution of water losses. Additionally, we verified the observed drying dynamics in microcosms inoculated with strain NCIB 3610 by estimating the thickness of the sub-viscous atmospheric boundary layer and the depth of the evaporation plane, and confirmed the role of polymeric substances in hydraulic decoupling by conducting an evaporation experiment using sand amended with xanthan as an EPS analogue.

Choice of bacterial strains
Two Bacillus subtilis strains were selected based on their contrasting biofilm forming potential (Gallegos-Monterrosa et al., 2016;McLoon et al., 2011) and were used to inoculate sand microcosms. Both strains were stored at -80 • C in glycerol and lysogeny broth (LB) media. The "wildtype" strain B. subtilis NCIB 3610 (Stanley et al., 2014) and the "domesticated" strain B. subtilis 168 trp + (Tanous et al., 2008) differ in their ability to develop robust biofilms in liquid media and on solid agar due to several genes associated with extracellular matrix production that are altered in the domesticated strain (McLoon et al., 2011). The domestication of B. subtilis with respect to biofilm formation could be traced to five genetic alterations between strain 168 and NCIB 3610: point mutations in coding sequences sfp, epsC, and swrA; a promotor mutation in degQ; and the absence of a plasmid-borne gene rapP (McLoon et al., 2011). Although the strain 168 trp + (a tryptophan-prototrophic derivative of 168) was also found to develop architecturally complex colonies under certain conditions, it carries the point mutation epsC (Gallegos-Monterrosa et al., 2016). Furthermore, the biofilm formation is repressed under high-salinity conditions (Rath et al., 2020) that likely occurred in our microcosm experiment during evaporative drying (that increases the salt concentration over time). Hence, we expect the wildtype to reduce evaporation from microcosms due to its biofilm forming potential and the production of EPS, which is a gel-like matrix consisting of macromolecules, like polysaccharides, proteins, and nucleic acids (Branda et al., 2005;Stoodley et al., 2002b).

Pre-culture and growth rates in liquid culture
Strains were transferred from a frozen -80 • C stock to a 5 ml TSB (tryptic soy broth; VWR International) liquid culture and incubated overnight in a shaking incubator at 28 • C and 280 RPM. The stationarystage cultures were diluted (1:32) and re-grown for two hours before plating on TSB agar (1.5% w/v; European bacteriological agar, Lab Logistics Group GmbH) for further use. A monoclonal colony of each strain was inoculated in liquid overnight cultures used in the microcosm experiment. Agar plates were stored at room temperature due to the sensitivity of B. subtilis to low temperatures. The strains displayed contrasting phenotypes when spotted on fresh TSB agar plates with high liquid content. The dense colonies of the wildtype were able to take up liquid from the plates that resulted in extruded structures (SI Fig. 1b). Overall, the wildtype displayed the expected intricate biofilm architecture that was absent in the domesticated strain, as previously reported (McLoon et al., 2011).
To compare the growth potential of the two strains, we determined their maximal growth rates. Triplicate liquid cultures containing 15 ml TSB in 50 ml Erlenmeyer flasks were inoculated with 100 μl of an overnight culture. Optical density (OD) dynamics at 600 nm (OD 600 , Ultrospec 10 Cell density meter, Amerham Biosciences) were recorded by extracting 1 ml from the cultures at different timepoints. To account for non-linearity of OD 600 measurement, readings were corrected using a relative density (RD) approach for calibration of OD 600 (Lin et al., 2010). To estimate growth rates during exponential phase, we used linear regression on the logarithm of RD timeseries normalized to initial conditions at inoculation (SI Fig. 1a).

Sand microcosms and treatments
To evaluate the impact of bacterial growth and EPS release on soil drying dynamics, a sterile control and two treatments incubated with the contrasting strains of B. subtilis were prepared. Acid washed and autoclaved quartz sand (0.1-0.3 mm; Carlo Bernasconi AG, Switzerland) was filled to 2.5 mm below the upper edge of aluminum containers (16 mm x 6 mm x 30 mm). To test whether aluminum impaired the growth of the two strains, containers were placed on TSB agar plates and inoculated with 100 μl of each strain's liquid culture at OD 600 of 0.1. No inhibition of growth was observed near containers even after prolonged exposure of several weeks. Microcosms were closed from the bottom with a membrane (20 µm mesh-size) to allow drainage of excess liquid. Prepared containers were buried so that 2/3 or the lower 20 mm of each container was immersed in glass beads (1 mm diameter) to allow drainage and to limit evaporation from the bottom. All containers were placed in a laminar flow cabinet (EF/S4, Clean Air Techniek B. V.) to ensure constant evaporation by air flow. The cabinet was sterilized by UV light exposure (> 15 min) prior to inoculation. On day one, 0.8 ml of sterile TSB (tryptic soy broth; VWR International) solution was applied at the surface of all control treatment replicates. This negative control without cell inoculum was used to account for potential effects of TSB on evaporation dynamics and to test for contamination. Microcosms of the other two treatments were inoculated with 0.8 ml of diluted (1:32) overnight cultures that were re-grown in TSB for four hours to ensure cells were in exponential growth phase before diluting to an OD 600 of 0.1. The microcosms of each treatment were placed on balances to monitor cumulative changes in weight during the adaptation period. Every 24 h, 0.8 ml of sterile dilute (0.1) TSB solution was applied to the surface of each microcosm.
A second set of larger glass containers (76 mm x 10 mm x 200 mm) were prepared with a coarser sand from the same manufacturer (0.3-0.9 mm diameter; Carlo Bernasconi AG, Switzerland), saturated with a xanthan solution (E-415, Solegraells) at a content of 2 mg xanthan per gram of sand and a control column saturated with deionized water. The evaporation dynamics of this sand were previously analyzed (Lehmann et al., 2008) and this material was chosen because the capillary flow paths are relatively short (in the range of 100 mm). Thus, disruption of capillary flow occurred after a few days of evaporation. For the finer sand used in inoculated microcosms, the experiment would last several weeks until the flow paths break. Containers were placed on scales and evaporation dynamics of both treatments were derived from recorded mass loss over time. The profile of gravimetric water content was obtained after 12.3 days by oven drying individual layers of wet sand removed from containers at 10 mm increments. This experiment was conducted to study the effect of evaporation suppression due to an EPS analogue and to support the observations made for microcosms inoculated with B. subtilis NCIB 3610.

Dynamics of cell counts and biomass
Every 24 h, a replicate of each treatment was destructively sampled to determine the number of colony forming units (CFU) and the liquidextractable biomass. Sand was removed from the containers and cells were extracted by suspension in 10 ml phosphate buffered saline (PBS, Phosphate Buffered Saline pH 7.4, gibco) followed by 10 s of vortexing at maximum speed and 2 min in an ultrasonic bath. The OD 600 of the extraction was recorded with reference to pure PBS. To determine CFU counts, the suspension was serially diluted (1:10 3 , 1:10 4 , and 1:10 5 ) and 100 μl were plated on TSB agar (1.5% w/v) at the highest dilution (1:10 5 ). Also, undiluted suspensions were spotted on agar plates to serve as controls (negative control for treatment without cells and positive control for others). The weight of the dry sand was recorded prior to cell extraction. In addition, 5 ml of the extracted PBS solution were filtered using Isopore™ Millipore with 0.4 μm pore size. The filters were weighed after airdrying for 24 h to estimate the amount of extractable biomass. Extractable biomass and CFU counts were measured between day 1 and 6 of the experiment. On day 7, the microcosms were sealed with parafilm (Parafilm "M" Laboratory Film, BEMIS) to prevent evaporation prior to the monitored dry-down experiment on day 8 by means of neutron radiography.
Dynamics of cell counts and biomass are summarized in the supplementary materials (SI Fig. 2).

Time-series neutron radiography
The loss of water from the upper sand surface of each microcosm was monitored with time-series neutron radiography at intervals of about 10 minutes. Neutron radiography is highly sensitive to H-rich liquids, like water or TSB in porous media due to their strong attenuation of neutrons (Moradi et al., 2009). This allows for non-invasive quantification of liquid volume distribution (Carminati et al., 2007) by analysis of the differences between signals of dry and wet samples (the direct computation based on the neutron cross-section of the constituting materials is not possible due to the heterogeneity of the sand material).
Time-series neutron radiography was conducted at the ICON (Imaging with Cold Neutrons) beamline of the Paul Scherrer Institute, Villigen, Switzerland (Kaestner et al., 2011). A CCD camera with a field of view of 15 × 15 cm was used to capture the mean local decrease in liquid volume across microcosms i.e., across 0.6 cm thickness of porous sand at a pixel size of 85 µm over a period of 52 h. The attenuation coefficient corresponding to the signal of a given thickness of liquid was derived by use of step wedges of specific thickness (in direction of the neutron beam) filled with the respective liquid. Resulting neutron attenuation coefficient of water and undiluted TSB solution were nearly identical (SI Fig. 3). Acquired images were also corrected for neutron scattering (Boillat et al., 2018;Carminati et al., 2019), normalized, spot cleaned and filtered.

Estimation of diffusive fluxes from evaporation plane
To validate if hydraulic decoupling is a plausible mechanism in the microcosms inoculated with B. subtilis NCIB 3610, we estimated the thickness of the sub-viscous atmospheric boundary layer and the depth of the evaporation plane at stage two (i.e., the thickness of the dry sand layer) based on the observed total water losses from the time-series neutron radiography measurements.
At stage one evaporation, the evaporation rate Q [m s − 1 ] is determined by diffusion across the sub-viscous atmospheric boundary layer L air [m]. For a given cross-sectional area and a density of water (1000 kg m − 3 ), we fitted L air to match the observed evaporation rate. Then, the diffusion length through the dry sand layer, L sand was fitted to match observed rate Q at the beginning of stage two evaporation assuming diffusion across the sand and air boundary layers arranged in series. Parameters were derived using MATLAB (Version 2020a; The Math-Works inc.) by minimizing the difference between simulated and observed Q at the two time points. Q is calculated according to with the diffusion coefficient of water vapor in air, D air = 2.42 × 10 − 5 m 2 s − 1 at 20 • C. The diffusion coefficient in sand, D sand [m 2 s − 1 ] is calculated according to Millington and Quirk (1961) by assuming a porosity ϕ of 0.5 and a volumetric air content ε of 0.48 for the dry sand: The difference in vapor concentration Δc is derived according to the ideal gas law defined as with the saturation vapor pressure p sat of 2.334 kPa (Buck, 1981), relative humidity RH of 0.8 [Pa Pa − 1 ], molar mass of water vapor M mol of 0.018 kg mol − 1 , universal gas constant R of 8.314 J K − 1 mol − 1 , and temperature T of 293 K (20 • C).

Results
The difference in maximal exponential growth rates in liquid culture between the two B. subtilis strains was smaller than the uncertainty of the triplicate measurements (0.75 ± 0.02 h − 1 and 0.77 ± 0.02 h − 1 for 168 trp + and NCIB 3610, respectively). Values were consistent with previous reports of 0.75 h − 1 for 168 trp + (Jules et al., 2009) and correspond to a minimum generation time of approximately one hour. Shaking of liquid cultures of strain 168 trp + resulted in quickly dispersing foam while for liquid cultures of NCIB 3610 stable foam was created that persisted for several hours. The biomass extracted from the sand microcosms and the obtained CFU counts fluctuated over time but did not display systematic differences between the two strains. Thus, we assumed similar cell densities and fitness for both strains throughout the microcosm experiment (SI Fig. 2b). No cross contamination was detected in undiluted suspensions extracted and spotted on agar plates.
The following qualitative observations were made for which no quantities were derived. During the growth period prior to neutron imaging, microcosms inoculated with either strain of B. subtilis developed a robust crust-like sand layer of a few mm in thickness at the surface after two days into the experiment. This robust sand layer might have been the reason for the reduced infiltrability resulting in temporal ponding of applied nutrient solution observed thereafter. No alteration of infiltrability nor a change in sand cohesion was observed in the control treatment with rapid infiltration of the nutrient solution after each application. Also, the porosity in microcosms inoculated with strain NCIB 3610 increased over time which led to a lift of the sand surface of about 2.5 mm.
Although both strains appeared to alter infiltration to a similar degree, evaporation dynamics were distinctly different between microcosms inoculated with either strain. Monitored fluctuations in weight between liquid applications for each treatment showed a continuous adaptation of evaporation dynamics for microcosms inoculated with strain NCIB 3610 (SI Fig. 4a. 5). Over the growth period, the early evaporation rate increased for microcosms inoculated with either strain. Near the end of the incubation period on day six, the initially high evaporation rate was followed by a similar characteristic decrease for the NCIB 3610 treatment, as observed subsequently with time-series neutron radiography. Note, that the captured cumulative weights for each treatment during the growth period might include weight differences caused by evaporation of liquid which drained from the microcosms into the coarse glass bead packing below.
The mean cumulative water losses across six microcosms of each treatment were monitored with time-series neutron radiography (Fig. 1a). In treatments inoculated with strain NCIB 3610, water loss was reduced by about 58% after 52 h when compared to the control treatment. Water loss from the NCIB 3610 treatment was reduced by about 51% in comparison to microcosms inoculated with strain 168 trp + . The water loss from NCIB 3610 treatments was initially accelerated and ceased after ca. 15 h (Fig. 1b) from 5.4 mm day − 1 to a stable rate of ca. 0.8 mm day − 1 . Evaporation from control treatments and microcosms inoculated with B. subtilis 168 trp + decreased gradually from 3.5 mm day − 1 to about 2.3 mm day − 1 and 1.8 mm day − 1 after 52 h, respectively.

Fig. 1. Derived water losses and evaporation rates from sand microcosms monitored
with time-series neutron radiography with 0.8 mm day − 1 evaporation rate indicated for treatment inoculated with B. subtilis NCIB 3610 (red dashed line). a) Mean and 95% confidence interval of mean total water loss of sand microcosms observed from time-series neutron radiography (n = 6). Control treatment (black dots), treatment inoculated with B. subtilis 168 trp + (blank dots), and treatment inoculated with B. subtilis NCIB 3610 (grey dots) are shown. b) Mean (dots) and standard error (error bars) of evaporation rate of inoculated and control sand microcosms observed from timeseries neutron radiography.
For the treatment with B. subtilis NCIB 3610, an immediate decrease in evaporation rate i.e., a transition from stage one to stage two evaporation was observed. Initial water loss by evaporation from the soil surface was followed by a detachment of the evaporation plane from the surface and diffusion of water vapor through a dry soil layer (stage two, Lehmann et al., 2008). Note that the evaporation front did not gradually recede, but abruptly "jumped" to form a new evaporation plane at deeper layers (Shokri and Or, 2011).
The water loss dynamics over depth show a similar average profile of water depletion for the control and the 168 trp + treatments (Fig. 2a). Initial water losses occurred at a comparable depth and magnitude with a gradual increase in depletion depth over time. For the NCIB 3610 treatment high initial water losses were observed within the first few millimeters of the profile and ceased after 15 h followed by minor depletion at a reduced mean evaporation rate (i.e., Fig. 2b). Estimates of the atmospheric boundary layer thickness and the depth of the dry sand layer were derived for the NCIB 3610 treatment, which displayed a clear separation of two soil drying stages. By considering an initial mean evaporation rate of 5.4 mm day − 1 and a reduced evaporation rate of 0.8 mm day − 1 , which was reached after ca. 15 h, at the onset of stage two (see Fig. 1), the estimated atmospheric boundary layer thickness was 1.3 mm and soil dry depth was 1.9 mm. Derived values are consistent with the observed rapid water loss near the sand surface in NCIB 3610 treatments (Fig. 2a).
The average profile of water losses shows increased depletion near the surface and up to ca. 10 mm depth, and slightly reduced loss below this depth for the 168 trp + treatment (Fig. 2b). Similarly, high water loss occurred near the surface of microcosms inoculated with B. subtilis NCIB 3610 while the cumulative water loss below this region was markedly reduced. The results indicate that the evaporation from columns inoculated with NCIB 3610 was limited after a thin surface layer of a few millimeters had dried. (1-5) are indicated by grey boxes next to treatment's plotted profiles. a) Mean water loss profile over time (20 h) for each treatment monitored with time-series neutron radiography (n = 6). b) Mean cumulative water loss across microcosm depth (solid line) with standard deviation (grey band; average of six samples) after 5 h and 52 h. The water loss below 10 mm depth is substantially reduced for the treatment with NCIB 3610.
To confirm that evaporation losses in the microcosms inoculated with NCIB 3610 could be caused by EPS, we conducted additional experiments using sand columns amended with xanthan as an EPS analogue. The drying dynamics measured for the sand-xanthan mixture were similar as for the microcosms modified by B. subtilis NCIB 3610. An initially high evaporation rate was followed by a relatively constant low rate of water loss (Fig. 3a). The period of stage one evaporation for the initially water saturated control lasted for six days before the transition to stage two evaporation occurred. The stage two evaporation rate of the control was higher than for the xanthan treatment after eight days. This difference could have been caused by dry xanthan structures deposited in the pores above the evaporation plane limiting water vapor diffusion. Reduced gas diffusion was reported for coarse sand amended with dried seed mucilage (Haupenthal et al., 2021).
The distribution of soil liquid content over depth after 12.3 days showed that the top 20 mm were completely dry for the xanthan column, indicating that the evaporation plane was at the depth of about 20 mm at the end of the experiment. Below 50 mm, the column was nearly saturated (Fig. 3b). In the control column, the evaporation plane was at depth of about 10 mm (completely dry top layer of 10 mm thickness). Below this dry layer, the water content increased gradually with depth and was similar to the spatial profile of water loss observed with neutron radiography in the control treatment and the microcosms inoculated with B. subtilis 168 trp + (Fig. 2a).

Discussion
The comparison of evaporative water loss illuminates the potential impact of complex microbiome processes on soil water dynamics. The substantial decrease in evaporative water loss observed after a growth period of one week for microcosms inoculated with B. subtilis NCIB 3610 is explained by EPS production and likely its accumulation in the pore space. The modification of soil properties requires time, indicated by the change in evaporation dynamics during the growth period (SI Fig. 4 a. 5). The observed deceleration of evaporative soil drying resulted in a substantial reduction of total water loss. Differences between the treatments are caused by an accelerated shift from stage one to stage two evaporation that effectively decouples the evaporation front from the soil surface allowing only the slow diffusion of water vapor from the retreating evaporation plane across the dry soil layer. Decoupling of the evaporation front was followed by an overall reduced water loss at greater depth Fig. 2a. Also, the depth-time profiles of local water loss were much smoother and less scattered when compared to the other treatments. The overall reduction in water loss is caused by a shift from capillary transport of water towards the soil surface to diffusion of water from the profile towards the surface. This transition occurred immediately after initiation of evaporative soil drying. Typically, at stage two the evaporation rate gradually declines as the evaporation plane moves downwards. In our case, the evaporation rate remained fairly constant after 15 h (0.8 mm day − 1 ), indicating no considerable retreat of the evaporation plane. The estimated diffusion length at stage two was in good agreement with the thickness of the dry region obtained from neutron imaging. This suggests that hydraulic decoupling is a plausible mechanism for the observed dynamics of evaporative water loss.
We observed equivalent dynamics with the formation of a retreating evaporation plane and a dry layer near the surface for xanthan treated sand. This suggests that EPS could limit the capillary flow to the evaporating plane. The reduced water loss near ca. 1 cm in depth in treatments with strain NCIB 3610 could be the result of decreased water retention i.e., water content in this region caused by locally increased pore sizes.
Neither the control treatment nor the microcosms inoculated with B. subtilis strain 168 trp + exhibited evaporation dynamics or spatially differentiated water loss comparable to changes induced by strain NCIB 3610. Nevertheless, the 168 trp + treatment showed a tendency for increased water loss near the soil surface and a slight decrease in water loss at greater depth. This alteration might have been caused by the accumulation of bacterial cells near the surface which could potentially induce a reduction in soil hydraulic conductivity due to pore clogging which would also explain the decrease of liquid infiltration observed.
Despite the substantial fluctuations in water content, both strains appeared to thrive within the porous environment, reflected by the CFUs obtained (SI Fig. 2b). The decrease in extracted biomass over time (with exception of the extract at day 4) could be related to the accumulation of EPS in the pore space that might have impaired the extraction process.
We conclude that structurally complex polymeric substances and an adaptive period are required to induce the observed modifications to the evaporation dynamics. The observed early shift to diffusion-limited water loss could be caused by multiple soil physical alterations. However, potential modifications caused by strain 168 trp + were comparably ineffective in this context.
For example, bacteria can reduce the surface tension of the gas-liquid interface (Raaijmakers et al., 2010). This modification causes a decrease in capillary forces responsible for transporting water to the soil surface and reduce water flow i.e., by facilitating quick depletion of soil water near the surface. In presence of strain NCIB 3610 an increase in porosity Fig. 3. Evaporation dynamics and water distribution in water saturated sand (control) and sand saturated with hydrated xanthan at a content of 2 mg g − 1 (i.e., mg dry xanthan per g dry sand). a) Evaporation rate derived from evaporative mass loss over time. b) Profile of gravimetric water content (g of water per g of dry sand) measured after 12.3 days derived from mass difference of wet and dry sand after oven drying. of the sand in microcosms was observed which led to a rise of the soil profile by about 2.5 mm. This increase might have been caused by surface active substances which lead to the formation of stable gas bubbles when water infiltrated from the sand surface. Stable foam (Razafindralambo et al., 1996) in liquid cultures of strain NCIB 3610 is associated with the production of surfactin that is absent in strain 168 (Julkowska et al., 2005). Surfactin is a bacterial lipopeptide surfactant that lowers the surface tension of water and is produced extensively in the presence of solid carriers (Yeh et al., 2008). The skimming of such foam was used by Cooper et al. (1981) to quantify the production of surfactin, which was reported to lower the surface tension to a minimum of 27 mN m − 1 . Mixtures of surfactant and polymers can enhance foam stability (Pugh, 2016), which could explain the observed increase in porosity. Additionally, bacterial EPS decrease the soil hydraulic conductivity (Rosenzweig et al., 2013), thus limiting the capillary flow to the soil surface. Therefore, the reduced capillary forces and higher friction i.e., reduced soil hydraulic conductivity can simultaneously contribute to the observed hydraulic decoupling.
For both bacterial treatments, reduced infiltration resulting in the short-term ponding of the daily applied nutrient solution occurred after two days. This reduction was accompanied by the formation of a stable crust at the sand surface of a few millimeters. These observations emphasize the effect of bacterial activity on soil hydraulic conductivity. Several studies showed a marked decrease of soil hydraulic conductivity in presence of polymeric substances Colica et al., 2014;Rosenzweig et al., 2013;Volk et al., 2016;Zheng et al., 2018). Such reduction would also explain the rapid hydraulic decoupling of capillary flow from the surface observed in the NCIB 3610 treatment as high EPS content is usually associated with the formation of a stable crust at the soil surface (Rossi et al., 2018). However, similar ponding of liquid and the formation of a stable crust was also observed in the 168 trp + treatment. Since the distinct hydraulic decoupling during evaporative drying was exclusively observed in sand inoculated with strain NCIB 3610, both strains appear to have modified infiltration dynamics by different means.

Conclusions
Hydraulic decoupling of the evaporation plane from the surface during evaporative soil drying reduces water losses and can be induced by bacteria. The process is explained by the EPS mediated reduction of soil hydraulic conductivity and capillary forces that can effectively reduce water loss from the soil surface by sheltering water below a layer of dry soil as validated by use of xanthan amended sand. These modifications of soil physical properties occurred during a period of fluctuations in soil water content and appeared to be part of a continuous adaptive process. The comparison of the control treatment and treatments with B. subtilis 168 trp + and B. subtilis NCIB 3610 highlights the impact of differences in soil bacterial (genetic) traits beyond growth rates and population sizes. Although infiltration was similarly decreased in the presence of both strains, the cumulative and local water losses during evaporation were distinctly different and hydraulic decoupling was solely observed in microcosms modified by the complex biofilm producer NCIB 3610. This "wild type" strain has the potential to drastically engineer the architecture of the soil pore space that shapes the rates and spatial distribution of soil water losses. Our study demonstrates how bacterial induced modifications of soil physical properties can affect evaporative soil drying. Beyond what we observed, EPS in soil and its effect on local soil water distribution and soil water dynamics remains surprisingly unexplored. Especially in regions like the rhizosphere, where microbial growth is stimulated by exudates released from plant roots, the processes underlying microbial modifications of soil water dynamics remain schematic and hypothetical.
Understanding the potential of bacteria to divert soil water fluxes during dry periods is critical for unraveling the role of specific genetic traits in the adaptation to environmental changes. To draw conclusions for more complex environments such as the rhizosphere or biological soil crusts, the interplay of distinct microbial traits, taxa, and associated dynamics in resource availability need to be considered.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability
Data will be made available on request.