Water-driven microbial nitrogen transformations in biological soil crusts causing atmospheric nitrous acid and nitric oxide emissions

Biological soil crusts (biocrusts) release the reactive nitrogen gases (Nr) nitrous acid (HONO) and nitric oxide (NO) into the atmosphere, but the underlying microbial process controls have not yet been resolved. In this study, we analyzed the activity of microbial consortia relevant in Nr emissions during desiccation using transcriptome and proteome profiling and fluorescence in situ hybridization. We observed that < 30 min after wetting, genes encoding for all relevant nitrogen (N) cycling processes were expressed. The most abundant transcriptionally active N-transforming microorganisms in the investigated biocrusts were affiliated with Rhodobacteraceae, Enterobacteriaceae, and Pseudomonadaceae within the Alpha- and Gammaproteobacteria. Upon desiccation, the nitrite (NO2−) content of the biocrusts increased significantly, which was not the case when microbial activity was inhibited. Our results confirm that NO2− is the key precursor for biocrust emissions of HONO and NO. This NO2− accumulation likely involves two processes related to the transition from oxygen-limited to oxic conditions in the course of desiccation: (i) a differential regulation of the expression of denitrification genes; and (ii) a physiological response of ammonia-oxidizing organisms to changing oxygen conditions. Thus, our findings suggest that the activity of N-cycling microorganisms determines the process rates and overall quantity of Nr emissions.


Materials and Methods
Sampling Cyanobacteria-dominated biocrusts were collected in small petri dishes (55 mm diameter and 15 mm height). For sampling, the bottom of the petri dish was placed upside down on the biocrust surface, pressed into the substrate, and with the help of a trowel pushed below, the biocrust part was lifted together with the underlying soil. In order to minimize metabolic activity, the samples were collected in an air-dried state and subsequently transported (darkened in a plastic box) by car and airplane to Germany. Before measurements, the samples were stored at the Max Planck Institute for Chemistry (MPIC, Mainz, Germany) for two and a half months (from the end of April to the beginning of July) at room temperature (25°C) in the dark.
Overall experimental setup Dynamic chamber measurements were performed according to the study of Weber et al. [1] following the work of Su et al. [2] and Oswald et al. [3].
Nine measurements with three measurements/replicates for each stage of desiccation were carried out.
The first sample set was removed from the chamber close to full water holding capacity (T1, early wetting; mean ~99% WHC; 20-30 minutes after wetting; Table S1), the second one at increasing emissions (at rising HONO and NO emissions before maximum was reached; T2, intermediate drying; mean ~30% WHC; 3.4 -5 hours after wetting; Table S1), and the third one close to the end of desiccation cycle (T3, late drying; mean ~4% WHC; 6.3 -8 hours after wetting; Table S1). Once the samples were removed from the chamber they were analyzed with different techniques (as described below) and therefore the whole material was used up for each measurement/replicate. Based on preliminary measurements, were full desiccation cycles were performed for the related samples (same biocrust type, study site and, collection date) and based on experiences, the second and third stage of desiccation (T2, T3) was estimated.

Dynamic chamber measurements
The dynamic chamber measurements are continuous measurements in the course of desiccation using a constant flow rate. During the experiments the chamber was kept in darkness in order to avoid photochemical reactions. The emissions and mixing ratios of NO, HONO, NO 2 (nitrogen dioxide), and H 2 O were measured at the outlet of a Teflon (PFE) chamber (volume 0.047 m 3 ), which was purged with purified and dried air (PAG 003, Ecophysics, Duernten, Switzerland) at a flow rate of 1x10 -4 m 3 s -1 .
In the course of desiccation, HONO was determined every 30 minutes, whereas NO was analyzed at 10-seconds interval. Fluxes F N (ng m -2 s -1 ) were calculated using the following formula [3]: For a time span of at least 20 minutes, before and after each measurement, zero air (empty chamber) was determined as a reference.
The water content of the samples was determined in the following way: In a first step, the total amount of water that could be held by the samples without dripping (= 100% WHC) was determined using a weight balance. Subsequently, the amount of H 2 O (ppth) evapotranspirated from the sample was measured in 10-second intervals over the course of desiccation, using an infrared gas analyzer (LI-7000, LI-COR Biosciences GmbH, Germany). At the end of the measurement, the amount of water in the samples was checked gravimetrically to ensure that the calculations over the course of the measurements were correct (which was always the case).
Soil water contents and WHC were calculated using the following formulae: m FC = mass of water at field capacity [g]. Field capacity is the amount of water held in a soil after gravitational water drainage stops.
In fully saturated samples, the mass of water in soil equals the field capacity [3].
NO and NO 2 were analyzed with a gas chemiluminescence detector, which was equipped with a blue light converter (Model 42C, Thermo Electron Corporation, USA; limit of detection (LOD) NO ≈ 120 ppt and LOD NO2 ≈ 300 ppt) [3]. HONO was detected spectrophotometrically using a long path absorption photometer (LOPAP, QUMA Elektronik & Analytik, Wuppertal, Germany; total accuracy 10% and detection limit ~3-6 ppt) [2]. The retention and response time accounted to 20 and 13 minutes, respectively. With this method, an acidic solution of sulphanilamide is used to sample HONO with a stripping coil directly connected to the chamber [1][2][3][4][5][6]. Upon reaction, HONO is immediately transformed into a diazonium salt, which serves as a precursor of a diazotization, which causes the formation of an azo dye. The concentration of the azo dye, which is equivalent to the concentration of HONO in the sampled air, is determined by means of VIS-spectrophotometry. A detailed description of the instrument was presented by Heland et al. [7] and Kleffmann et al. [8].

Microsensor measurements
Oxygen saturation was analyzed using oxygen microsensors (OX-100, Unisense A/S, Aarhus, Denmark) with a tip diameter of 100 µm. The measurement principle is based on diffusion of oxygen through a silicone membrane to an oxygen reducing cathode, which is polarized against an internal Ag/AgCl anode. An additional guard electrode ensures signal stability by removal of all oxygen diffusing towards the sensor from the internal electrolyte (Revsbech, 1989; Revsbech and Jørgensen, 1986).
We analyzed the oxygen saturation within the photoautotrophic layer (surface to 400 µm depth) and the heterotrophic layer (> 400 µm to 3000 µm) at T1 (~99% WHC) and T2 (~30% WHC). This was accomplished by vertical profiles at 200 µm steps. The water content (WHC) was determined with the help of a weight balance.

Nitrite and nitrate analyses
The NO 3 − and NO 2 − content of the biocrust samples was measured before (Pre) and after (Post) a desiccation cycle. In order to identify biological processes causing a potential accumulation of mineral N, also samples treated with methyl iodide (CH 3 I) to suppress microbial activity were investigated.
Methyl iodide treatment of samples was performed according to a procedure established for soil samples by Oswald et al. [3]. For this, biocrust samples were placed into a vacuum desiccator and 1 ml CH 3 I (99% Reagent Plus, SIGMA-ALDRICH Chemie GmbH, Germany) was added in a separate beaker. Subsequently, a vacuum was generated and the samples were exposed to a high partial pressure of CH 3 I for 24 hours [3]. NO  Subsequently, the samples were centrifuged at 10,000 x g and 4°C for 5 min, and were then washed two times in 1 x PBS buffer (1.5 mL), with each step being followed by a centrifugation with the same characteristics. Afterwards, the samples were stored in 1. EtOH and allowed to dry. After this step, filters were cut into 12 pieces and were stored at -20 °C until in situ hybridization and quantification of cells [9,10].
In situ hybridization: A hybridization buffer set to a probe specific formamide concentration (see Quantification: An eyepiece reticle (net micrometer covering an area of 10 x 10 mm, Nikon, Tokyo, Japan) was used for cell counting. For 10 microscopic fields of sight, that were randomly chosen on the filter sections, the cells in 100 squares were counted (10 areas of 10 x 10 mm; 1000 squares in total/filter section). Taking into account the dilution of the soil samples in the course of the fixation, sonication and filtration procedure, these cell numbers were extrapolated to cells per gram of soil (dry weight).

Statistics:
To test for significant differences in cell counts among measurements, we applied a Bayesian hierarchical model, fitted via the Stan Hamiltonian Monte Carlo sampler [12], implemented using the "brms" package (version 2.13.5; [13]) in the R programming language (version 4.0.2; [14]).
We fit separate models for EUB, Archaea, and NOB, with soil layer and stop time coded as categorical variables. Differences among treatments were quantified based on the posterior distribution of contrasts between these categorical variables, e.g. as the difference between estimated cell counts for UL (T1) vs. UL (T2). Pairs of treatment for which this difference did not include zero within the onetailed 95% credibility interval were interpreted as being "significanty" different from one another. For all models, we ran 4 chains for 2000 iterations, and discarded the first 1000 iterations as a burn-in period. To account for pseudoreplication, we included a "random intercept term" (i.e. fitted categorical variables constrained by a shared Gaussian prior) for sample ID nested within dish ID, of the form GeoChip functional gene microarray The functional gene microarray (FGA) GeoChip allows to study the genes involved in the biogeochemical cycling of nitrogen, carbon, sulfur, and phosphorus [15,16]. The GeoChip 5.0 version, manufactured by Agilent Technologies (Santa Clara, CA, USA), has 57,000 oligonucleotide probes (60K). Only genes for proteins containing catalytic subunits or active sites are included [16].
The gene markers used for the detection of nitrification included amoA and hao, whereas narG, nirK, nirS, norB and nosZ indicated denitrification processes.

RNA extraction, cDNA synthesis, labeling
Before extraction, the samples were placed on liquid N 2 and homogenized using a RNase/DNase free polystyrene spatula. From 2 g of soil, taken from several areas and over the entire height (15 mm) of the petri dish, RNA was extracted using the RNeasy PowerSoil® Total RNA kit (Qiagen, Venlo, Netherlands), following manufacturer´s instructions. We also included a negative control (without soil).
The purity and concentration of the extracted RNA was determined using the Multiskan GO UV/VIS cDNA extracts were stored at -20˚C until use. cDNA was used for functional gene array hybridization.
Sample labeling, hybridization, scanning, and image processing were performed at the company Glomics Inc. (Norman, OK, USA). Labeling of DNA for microarray hybridization was performed with fluorescent cyanine dyes [16].

Data Processing and Analysis
Prior to data analysis, normalization was performed to take into account unequal quantities of starting RNA, differences in labeling or detection efficiencies between the fluorescent dyes, and systematic differences across datasets [17]. For microarray data normalization and comparison of gene array data across different samples and stages during desiccation, the common oligonucleotide reference standard (CORS) method was applied [16,18]. The CORS is an artificial sequence probe, which is co-spotted with each gene probe. The complementary CORS target is labeled with a fluorescent dye different from that of the sample and is spiked into each sample before hybridization [16,18]. For data normalization, the average signal intensity of CORS was calculated for each array and the maximum average value was used to normalize the signal intensity of samples in each array. Secondly, for each array the sum of the signal intensity of samples was calculated. The maximum sum value was applied to normalize the signal intensity of all spots in an array, generating a normalized value for each spot.
Spots were considered as positive if the signal-to-noise ratio was ≥2.0 [SNR = (signal meanbackground mean)/background standard deviation], and the signal intensity was at least 1.3 times the background. Furthermore, spots with signal intensities less than ~200 were removed [16]. Of 24 983 positive spots, 5 275 (21.1%) were excluded because they were detected in only one of the three replicates. In the current study, only probes involved in the biogeochemical cycling of nitrogen have been analyzed. The Geochip data analysis was performed with a software pipeline developed by the company Glomics. The pipeline includes tools to perform descriptive statistics, such as relative abundances of genes/gene categories/subcategories, and richness, alpha, and beta diversity indices of functional genes [15,16]. A logarithmic transformation was carried out to improve the characteristics of the data distribution and variables were scaled before measuring inter-observation dissimilarities.
Class comparison/differential expression: The microarray experiment was designed as class comparison experiment [19,20]. We looked for genes with statistically significant differences between the three stages during desiccation. For differential expression analysis, a global test, linear models for microarray (limma method), was applied, which analyzes each probe separately [19,20]. To assess the differential expression of each probe by means of limma, the moderated t-statistic was applied, where the standard errors have been moderated across genes using a Bayesian model [21,22]. As testing thousands of genes is likely to produce hundreds of false positives, multiple testing correction is necessary. The chosen approach was to control the false discovery rate (FDR; e.g. Benjamini-Hochberg; BH(p) = significant if ≤ α). The FDR is a less restrictive approach than the family wise error rate procedure [19].

n i f H n a r G n i r K n i r S n o r B n o s Z g d h u r e C a m o A h a o n x r A n a p A n r f A h z o h z s A n a r B n a s A n i r A n i r B
Upper layer Lower layer