Biokinetics of microbial consortia using biogenic sulfur as a novel electron donor for sustainable denitrification

A R T I C L E I N F O


Introduction
The increased depletion of resources, the rising water stress, the need of decreasing the carbon footprint and the stringent nutrient (WWTPs). However, heterotrophic denitrification and denitritation typically require the supply of organic matter, resulting in higher sludge production and operational costs (Sun and Nemati, 2012). To overcome these disadvantages, chemically synthesized elemental sulfur (S 0 ) has been used as a cheaper and effective electron donor for autotrophic denitrifying microorganisms treating wastewaters poor in organics . However, the low water solubility of chemically synthesized S 0 limits its availability to microorganisms and makes denitrification and denitritation kinetics slower than that achieved with more soluble electron donors (Kiskira et al., 2017a;Mora et al., 2015;Park and Yoo, 2009;Zou et al., 2016). Additionally, elevated NO 2 concentrations when using chemically synthesized S 0 during autotrophic denitrification can decrease the overall process efficiency (Christianson et al., 2015;Kostrytsia et al., 2018). Specifically, the sulfur to nitrogen (S/N) molar ratio, the feed pH and the microbial community structure are known to be among the main factors controlling the NO 2 accumulation. Therefore, it is crucial to investigate the potential of alternative electron donors for both NO 3 and NO 2 removal, such as biogenic S 0 . Biogenic S 0 (or biosulfur) is a biological product obtained from the incomplete oxidation of sulfide in gaseous streams under oxygen-limiting conditions by S-oxidizing microorganisms (Florentino et al., 2015). The Thiopaq® technology (Paques BV, the Netherlands) is, for instance, a well-established process aimed at the biological gas desulfurization that integrates hydrogen sulfide (H 2 S) removal and biogenic S 0 recovery, with more than 200 installations worldwide ('THIOPAQ Biogas desulphurization,' 2018). Biogenic S 0 globules, generated by different strains of bacteria, are considered to be hydrophilic, with a structure made of orthorhombic S 0 crystals surrounded by a hydrated layer of long-chain polymers or polythionates (Kamyshny et al., 2009;Kleinjan et al., 2003). The chemical composition of biogenic S 0 globules and their small particle size (2-40 μm) affect the S 0 (bio)chemical reactivity and make it more bioavailable for microorganisms (Findlay et al., 2014).
These exclusive properties of biogenic S 0 have promoted its application as a fertilizer (FERTIPAQ, the Netherlands) and in metal recovery technologies (Florentino et al., 2015). Only in the last two years, biogenic S 0 has been suggested for denitrification applications (Di Capua et al., 2016) due to its bioavailable nature and the possibility to offer a more affordable and sustainable nutrient removal solution. In this line, a possible integrated solution combining desulfurization of biogas with nitrogen removal from wastewaters can become applicable in the future, enabling to upgrade the current wastewater treatment configurations on a novel water resource recovery facility, in agreement with the EU Action Plan for the Circular Economy. To do so, more research on the chemistry and microbiology behind the use of biogenic S 0 for NO 3 and NO 2 removal is required.
The present research aims to investigate the fundamental aspects of autotrophic denitrification with biogenic S 0 (ADBIOS) using highstrength NO 3 and NO 2 synthetic waters. The main objectives of this study were to: (i) perform a physico-chemical and elemental characterization of the biogenic S 0 used; (ii) enrich a microbial consortium capable of NO 3 and NO 2 reduction and concomitant oxidation of biogenic S 0 in batch; (iii) use the enriched microbial community to evaluate the kinetics of biogenic S 0 -based autotrophic denitrification, denitritation, and simultaneous denitrification-denitritation in batch bioassays; and (iv) investigate the structure of the bacterial communities in the presence of NO 3 -, NO 2 or both. The impact of this study for the design and scale-up of biogenic S 0 -driven denitrification and denitritation systems is discussed.

Materials and methods
2.1. Source of biogenic S 0 and development of the biogenic S 0 -oxidizing microbial consortium The biogenic S 0 (Fertipaq BV, the Netherlands, purity > 99%, 11% moisture content) from the Thiopaq process (Paques BV, the Netherlands) was used as electron donor in the batch bioassays aimed at denitrification and denitritation. An activated sludge collected from the denitrifying tank of a municipal wastewater treatment plant (Cassino, Italy) was used as inoculum (10% v/v) in the batch bioassays. The biogenic S 0 -based denitrifying bacterial cultures were enriched for 3.5 months in 125 ml serum bottles with a working volume of 100 ml. The bottles were fed with the basal medium and trace elements as reported by Kostrytsia et al. (2018). NO 3 or NO 2 were individually added to the serum bottles with an initial concentration of 240 mg N/l. To ensure the presence of an adequate concentration of biogenic S 0 for complete denitrification or denitritation, an excess S 0 was used to maintain an S:N (g/g) ratio of 3.76 (1.5 times higher than the stoichiometric value). NaHCO 3 was added as buffer and carbon source with a concentration of 2.0 g/l. Each bottle was purged with helium gas for 3 min to exclude oxygen, prior to sealing the bottle with a rubber stopper and an aluminum crimp. Subsequently, the bottles were placed on a gyratory shaker at 300 rpm and temperature was maintained at 30 ( ± 2)°C by means of a water bath. The enrichment was subcultured when NO 3 --N or NO 2 --N concentrations approached zero. An enrichment was treated as 'stable' when the achieved denitrification or denitritation rates of the subcultures alternated by less than 5%.

Kinetic experiments
To knowledge of the authors, for the first time biogenic S 0 -oxidizing microbial consortia capable of reducing NO 3 or NO 2 were developed.
Three sets of batch bioassays were set up using the enriched biomass to investigate the kinetics of ADBIOS, i.e. denitrification (A: NO 3 − and S 0 ), denitritation (B: NO 2 and S 0 ) and simultaneous denitrification-denitritation (C: NO 2 -, NO 3 and S 0 ). The initial NO 3 or NO 2 concentrations were similar (i.e. 5% difference) to those used in a previous batch study on autotrophic denitrification with chemically synthesized S 0 . In the first experiment (A), batch bioassays were conducted to investigate the denitrification characteristics (NO 3 --N reduction rate, NO 2 --N accumulation and NO 2 --N reduction rates). During the second set of the experiments (B), NO 2 --N was used as the sole electron acceptor in order to evaluate the denitritation kinetics (NO 2 --N reduction rate) and assess the impact of biomass acclimation to NO 2 --N degradation. The simultaneous denitrification-denitritation experiment (C) was performed to study the effect of NO 3 --N on NO 2 --N degradation. At the beginning of each experiment, the required amount of NO 3 --N and NO 2 --N from stock solutions was added into the serum bottles to achieve the desired initial concentration as reported in Table 1. Biogenic S 0 , NaHCO 3 , the basal medium and trace elements solution were supplied at the same concentrations as in the enrichment phase. Each serum bottle was inoculated with an enriched culture with an amount of approximately 217.5 ( ± 2.5) mg/l of volatile suspended solids (VSS). Abiotic controls were used to monitor the possible chemical reactions involving the electron donor and electron acceptor. Controls without electron donor (biogenic S 0 ) or electron acceptor (NO 3 or NO 2 -) were carried out to estimate their possible degradation not associated with S 0 -driven denitrification or denitritation. In each experiment, the denitrification and denitritation rates were calculated from the slope of the curve describing NO 3 --N and NO 2 --N degradations, respectively, versus time and expressed as mg NO x − -N/l·d. The biomass specific denitrification and denitritation activities (mg NO x − -N/g VSS·d) were calculated by normalizing the denitrification and denitritation rate data with the initial biomass concentration (g VSS/l).

Sampling and analytical techniques
The liquid samples were taken twice a day and stored at −20°C prior to analysis. NO 3 -, NO 2 and sulfate (SO 4 2− ) concentrations were determined by ion chromatography, as reported elsewhere (Kiskira et al., 2017b). Elemental S 0 was determined by reversed-phase chromatography as originally described by Kamyshny et al. (2009). In this study, a highperformance liquid chromatography (HPLC) system (Prominence LC-20A Series, Shimadzu, Japan) equipped with a Kinetex LC column (C18, 5100 Å) and a UV/Vis detector (SPD-20A, Shimadzu, Japan) at 230 nm was used to quantify elemental S 0 . Prior to and at the end of the batch kinetic experiments, total suspended solids (TSS) and VSS of the liquid samples were determined according to the Standard Methods (APHA, 2011).
Laser size particle analysis (LSPA) was performed to determine the particle size distribution (PSD) of the raw and freeze-dried biogenic S 0 in a deionized water by a Mastersizer 2000 (Malvern Instruments, UK) laser diffraction particle sizer equipped with a HydroG sample dispersion wet unit. The measurement range of the instrument was from 0.02 to 2000 µm. Size parameters of the diameters d 0.1 , d 0.5 and d 0.9 were presented with 10%, 50% and 90%, respectively, of the volume of the particles below the given number.
To investigate the chemical and structural origin of biogenic S 0 , Raman spectra were obtained at random positions on the biogenic S 0 material using a Horiba LabRAM II Raman spectrometer (Horiba Jobin-Yvon, France). The instrument was equipped with a 600 groove·mm −1 diffraction grating, a confocal optical system, a Peltier-cooled CCD detector and an Olympus BX41 microscope arranged in 180°backscatter geometry. The measurements were performed using a 532 nm laser channeled through a Leica L100X/0.75 objective, providing a laser spot diameter of ∼1.5 μm.

Microbial community analysis
2.4.1. DNA extraction and high-throughput sequencing The total genomic DNA was extracted from the inoculum (Section 2.1) and the biomass at the beginning and at the end of the experiments (Table 1) in triplicate, following the protocol described by Griffiths et al. (2000). A high-throughput sequencing of partial 16S rRNA gene on DNA samples was conducted by the Illumina MiSeq sequencing service (FISABIO, Spain). The primers 515F and 806R were applied to target the 16S rRNA gene. The raw sequence files supporting the results of this article are available in the European Nucleotide Archive under the project accession number PRJEB27906.

Bioinformatics
The abundance table was generated by constructing operational taxonomic units (OTUs). Initially, the paired-end reads were preprocessed as described by Schirmer et al. (2015). Briefly, the paired-end reads were trimmed and filtered using Sickle v1.200. Then, PANDAseq v2.4 was used to assemble the forward and reverse reads into a single sequence spanning the entire V4 region. This resulted in consensus sequences for each sample on which VSEARCH v2.3.4 was used for OTU construction. The preprocessed reads from each sample were pooled together while barcodes were added. The reads were then dereplicated and sorted in order of decreasing abundance (Schirmer et al., 2015). Subsequently, the reads were clustered based on 97% similarity, followed by a removal of clusters (vsearch). Finally, the OTU table was generated by matching the original barcoded reads against clean OTUs (a total of 1104 OTUs for n = 19 samples) at 97% similarity. The representative OTUs were taxonomically classified against the SILVA SSU Ref NR v123 database. Multisequence aligned the OTUs and used them with FastTree v2.1.7 to generate the phylogenetic tree in NEWICK format. The biom file for the OTUs was then generated by combining the abundance table with the taxonomy information using Qiime workflow.

Statistical analysis
Statistical analyses were performed in R v3.4.4 using the combined data generated from the bioinformatics as well as meta data associated with the study. The vegan package was used for alpha and beta diversity analyses. For alpha diversity measures, the following indexes were calculated: rarefied richnessthe estimated number of species after rarefying the abundance table to minimum library size; Shannon entropy a commonly used index to measure balance within a community. The ordination of the OTU table in a reduced space was done using Principal Coordinate Analysis (PCoA) plots of OTUs using two different distance measures: Bray-Curtis distance metric which considered OTU abundance counts and; Weighted Unifrac distance that combined the phylogenetic distances weighted with relative abundances. Phylogenetic distances within each sample were further characterized by calculating the nearest taxa index (NTI) and net relatedness index (NRI) (Kembel et al., 2010). This analysis helped to determine whether the community structure was stochastic (driven by competition among taxa) or deterministic (driven by environmental pressure).
Sparse Projection to Latent Structure -Discriminant Analysis (sPLS-DA) was performed with the R's mixOmics package (Rohart et al., 2017). The procedure constructed artificial latent components of the predicted variables (OTU table collated at genus level) and the response variables by factorizing these matrices into scores and loading vectors in a new space such that the covariance between the scores of these two matrices in this space was maximized. The loading vector was constructed with the coefficients indicating the importance of each variable to define the component, i.e. non-zero coefficients in the loading vectors indicated the genera that vary significantly between the categories and were thus deemed as discriminants (Rohart et al., 2017). Fine tuning of the algorithm was applied by splitting the data into training and testing sets and then finding the classification error rates, employing two metrics, i.e. overall error rates and balanced error rates (BER).

Physico-chemical and elemental characterization of biogenic S 0
The PSD of biogenic S 0 is shown in Fig. 1. The raw biogenic S 0 sample consisted of particles with a median size of 241.16 µm and the 10% (6.88 µm) and 90% (508.89 µm) as measures of variability (Fig. 1). A surface weighted mean of 24.78 µm was quantified, which estimated the average diameter based on the surface area. In contrast, in previous studies on chemically synthesized S 0 -based denitrification, the S 0 particle size was between 500 and 16000 µm (Christianson and Summerfelt, 2014;Sahinkaya et al., 2014;Sahinkaya and Kilic, 2014), which is approximately two orders of magnitude higher than that of the S 0 used in the current study.
To evaluate the degree of agglomeration of the biogenic S 0 and its behavior in suspension, the size distribution of the S 0 particles was also determined after mixing them in water (5 min at 300 rpm and 30°C). The measured particles were with the median size of 4.69 µm and variability of the 10% (1.37 µm) and 90% (12.8 µm). This shows that the particle size of the biogenic S 0 was of the same order of magnitude of the S 0 particles (up to 1 µm) produced by sulfide-oxidizing bacteria in aqueous environment (Findlay et al., 2014). The contact of biogenic S 0 with water under mixing (i.e. 300 rpm) was likely to break sulfur agglomerations (Fig. 1). This could be explained by the hydrophilic surface of biogenic S 0 , which hinders the particle-aggregation in water (Kleinjan et al., 2003). In contrast, chemically produced S 0 is more hydrophobic and aggregates quickly in aqueous solutions (Findlay et al., 2014).
The specific surface area (SSA) of elemental S 0 particles is a main driver for its biooxidation rate, including oxidation coupled to denitrification and denitritation . In previous studies, the higher reactivity of biogenic S 0 was attributed not only to its unique surface characterization, but also to a higher SSA associated with a smaller particle size compared to that of the chemically produced S 0 (Kleinjan et al., 2003). In this study, the small biogenic S 0 grain size of 4.69 µm obtained after mixing provided a high SSA of 3.38 m 2 /g, compared to that of raw biogenic sulfur that had a 0.242 m 2 /g SSA corresponding to a 241.16 µm grain size. In addition, the results of LSPA (Fig. 1) and Raman spectroscopy (data not shown) confirmed that the biogenic S 0 used in this study is an elemental microcrystalline orthorhombic sulfur. These findings are in line with a previously proposed biogenic model of a microcrystalline solid elemental sulfur covered by biopolymers (Janssen et al., 1999). Therefore, the microcrystallinity of biogenic S 0 particles results in its higher reactivity and solubility, as suggested by Pasteris et al. (2001).
During denitrification, the reactions of NO 3 or NO 2 reduction to nitrogenous oxides are catalyzed by metalloenzymes (Shao et al., 2010). Among the quantifiable trace metals detected in biogenic S 0 (Table 2), copper (Cu), molybdenum (Mo) and iron (Fe) are co-factors of metalloenzymes (Shao et al., 2010). Nitrite reductase (NiR) and nitrous oxide reductase (N 2 OR) enzymes contain Cu. Mo is covalently attached to the protein in nitrate reductase (NaR), and Fe is cofactor for nitric oxide reductase (NOR). Additionally, Fe-S proteins, so-called ferredoxins, mediate electron transfer during NO 3 and nitric oxide (NO) reduction (Shao et al., 2010). Thus, the supply of trace metals is essential for the high performance of denitrification and denitritation, and biogenic S 0 can effectively provide the necessary trace metals during these processes ( Table 2). The possible inhibitory effect of heavy metals released by biosulfur on the activity of denitrifying biomass can be taken into consideration in a future study.

Effect of electron acceptor on ADBIOS kinetics
The accumulation was most likely ascribed to a higher enzyme activity of NaR compared to NiR, as also reported elsewhere (Du et al., 2016;Sun and Nemati, 2012). The high NO 2 --N build-up was followed by a drop of the NO 3 --N degradation rate to 7.8 mg/l·d from day 5 onwards. This was likely due to the inhibition of a NO 2 --N concentration above 60 mg/l on the activity of the denitrifying biomass (Guerrero et al., 2016).
In the denitritation experiments, the potential of the biogenic S 0oxidizing biomass enriched on NO 2 to reduce high NO 2 concentrations was investigated (Fig. 2b and e). The bacteria were capable of completing NO 2
SO 4 2− -S was the only sulfur product of the biogenic S 0 oxidation ( Fig. 2 and f). This observation was confirmed by the stochiometric consumption of the biogenic S 0 with NO 3 --N or NO 2 --N (Sun and Nemati, 2012). No denitrification and denitritation were observed in the abiotic and electron donor-free controls (data not shown). In this study, specific denitrification and denitritation activities of 223.0 mg NO 3 --N/g VSS·d and 339.5 mg NO 2 --N/g VSS·d, respectively, were achieved by the biogenic S 0 -oxidizing microbial consortium (Table 3). The high solubility of biogenic S 0 , which was likely attributed to the hydrophilic properties and the lower particle size of the biogenic S 0 particles (Section 3.1), induced a significantly higher NO 2 --N degradation (Fig. 2a). The kinetics of AD-BIOS (including both denitrification and denitritation) was characterized by 10-time higher rates compared to those obtained with chemically synthesized S 0 , with both studies being performed at similar initial NO 3 and NO 2 concentrations.

Effect of different electron acceptors on microbial communities performing ADBIOS
The efficiency of biological NO 3 and NO 2 reduction depends on the community composition of microorganisms (Shao et al., 2010). Thus, the genera prevailing under each experimental condition (Fig. 3a), i.e. denitrification (A), denitritation (B) and simultaneous denitritation-denitrification (C) at the beginning (T 0 ) and at the end (T F ) of the experiments, as well as the microbial community of the raw activated sludge (AS) used as inoculum, were analyzed in this study (Figs. 3 and 4). The enrichment in both denitrification (A) and simultaneous denitrification-denitritation (C) (i.e. NO 3 and NO 2 − both involved) led to similar microbial communities with samples A T F and C T F clustering closer to each other on the PCoA plots (Fig. 3d). In contrast, when denitritation was performed alone (B) (i.e. with the sole NO 2 involved), a distinct community was formed (B T F ) (Fig. 3d). Thus,  Table 3 The highest denitrification and denitritation rates obtained with different S 0 sources as electron donor by a 3.5-month enriched biomass at initial concentrations of 225.0 and 240.0 mg/l for NO 3 --N and NO 2 --N, respectively, and 220 and 215 mg VSS/l in denitrification and denitritation experiments with biogenic S 0 , respectively, and 1000 mg VSS/l in experiments with chemically synthesized S 0 . different key representative genera are selected when both (A and C) or only one electron acceptor (B) are used, with the former enriching for Petrimonas, Bacillus, Truepera, Ferruginibacter, Castellaniella, Aminobacter and the latter comprising of Comamonas, Truepera, Bacillus and Clostridium sensu stricto 13, based on taxa differential analysis (Fig. 3b). Some members of the genera Comamonas and Bacillus have been shown to be involved in NO 3 reduction (Park and Yoo, 2009;Zhang et al., 2015), while in this study these were also abundant in the denitritation (B) experiment. Additionally, in the top 25 most abundant genera Thiobacillus and Moheibacter were predominantly selected for the conditions with two electron acceptors (A and C) (Fig. 3b). In contrast, the condition with one electron acceptor (B) selected for different communities predominantly (∼75%) comprising of Thiobacillus and Thermomonas (Fig. 3b). Thiobacillus has been reported as ubiquitous in denitrification applications with reduced sulfur compounds, e.g. particulate chemically synthesized S 0 and soluble S 2 O 3 2− (Di Capua et al., 2016; Kostrytsia et al., 2018) and is capable of withstanding high NO 2 concentrations (Chen et al., 2018;Gao et al., 2017;Zhang et al., 2015), as also observed in this study with hydrophilic biosulfur (Fig. 2b)  Following the sPLS-DA algorithm, only 40 genera were varying between the conditions in the kinetic experiments (Fig. 4a-c). The communities, when two (A and C) or one electron acceptor (B) were used, mainly differed in terms of genera represented by Strenotrophomonas, Kaistia, Moheibacter, Brevundimonas, Thauera, Propionicicella, Seculamonas ecuadoriensis (block b1); and Bryobacter, Diaphorobacter, Actinotalea, Rhodanobacter, Microbacterium, Pseudaminobacter, Dokdonella, Intrasporangium, Halothiobacillius, Thermomonas, and Sphingopyxis (block b4). Block 1 was under expressed in denitritation (B), whereas block b4 was over expressed in denitritation (B), and vice versa for denitrification (A) and simultaneous denitrification-denitritation (C). Similarly, Strenotrophomonas, Thauera, Diaphorobacter and Halothiobacillius were also reported in denitrifying reactors with chemically synthesized S 0 (Xu et al., 2015;Zhang et al., 2015). Rhodanobacter, Dokdonella and Thermomonas genera within the Xanthomonadaceae family are capable of using organic products from cell lysis to fuel denitrification (Xu et al., 2015). Pseudaminobacter is capable to oxidize reduced sulfur compounds directly to SO 4 2− (Ghosh and Dam, 2009). The co-presence of the two electron acceptors mainly selected for (Fig. 4d): Shinella, Rhizobium, Pleomorphomonas, Simplicispira, Limnobacter (block b2); and Nakamurella, Truepera, Cellulomonas, Petrimonas, Clostridium sensu stricto 13, Bacillus, Ferruginibacter, Aminobacter, Castellaniella, Isosphaera (block b3). Shinella, Rhizobium, Simplicispia and Limnobacter were detected in reactors with reduced sulfur compounds treating  The Nakamurella, Truepera, Petrimonas, Clostridium sensu stricto 13, Bacillus, Ferruginibacter Aminobacter, Castellaniella, Isosphaera genera are known to comprise of some denitrifying bacteria (He et al., 2018;Zhao et al., 2016). In addition, looking at block b3, the right tree comprising Clostridium sensu scricto 13, Bacillus, Ferruginibacter, Aminobacter, Castellaniella, and Isosphaera is overexpressed for both denitritation (B) and denitrification (A), indicating that those genera can degrade NO 2 - (Park and Yoo, 2009;Spain and Krumholz, 2012;Xu et al., 2017).
3.4. Opportunities for ADBIOS as a part of a sustainable and integrated wastewater treatment system ADBIOS ( Fig. 5 [1]) provides a sustainable technological solution for biological nitrogen removal fueled by biogenic S 0 , as a by-product of biogas desulfurization (Fig. 5 [2]). The benefits of the process, such as a 10-time faster kinetics (Section 3.2) compared to that of autotrophic denitrification with chemically synthesized S 0 , make it technologically attractive and economically feasible. In addition, the high NO 2 degradation rate in the presence of a NO 2 --enriched biomass suggests that ADBIOS can also be applied for NO 2 removal from wastewaters ( Fig. 5 [1]). Generally, ADBIOS implements the reuse of a waste resource (S 0 ) into conventional nitrogen removal systems and creates a potential for an integrated process combining wastewater and flue gas treatment. Therefore, the scale-up of ADBIOS is of a great interest, and the current study can serve as the basis of the necessary fundamental information on the process. However, not each site may have readily-available biogenic S 0 supply, and biosulfur transportation might be required, which needs to be considered within an economic balance.

Conclusions
The biogenic S 0 -oxidizing microbial consortia capable of reducing NO 3 or NO 2 mostly included Thiobacillus, Moheibacter and Thermomonas. The biogenic S 0 showed an orthorhombic crystalline structure, having a 4.69 µm median particle size and a 3.38 m 2 /g SSA, which made it particularly reactive and bioavailable. The specific denitrification and denitritation activities as high as 223.0 mg NO 3 --N/g VSS·d and 339.5 mg NO 2 --N/g VSS·d, respectively, resulted in enhanced denitrification and denitritation rates compared to those of chemically synthesized S 0 . Moreover, the use of biogenic S 0 induced a lower accumulation of NO 2 -, alleviating the activity of the denitrifying consortia.