Sustained organic loading perturbation favors nitrite accumulation in bioreactors with variable resistance and resilience of nitrification and nitrifiers

Sustained perturbations are relevant for environmental biotechnology as they can lead systems to alternative stable states that may not be reversible. Studies assessing these concerns are scarce as robust replication is required. Here, we tested the effect of sustained organic loading variations (food-to-biomass ratio, F:M; carbon-to-nitrogen ratio, C:N) on both structure and function of activated sludge bacterial communities, focusing on nitrification and nitrifiers. Two sets of replicate 5-liter sequencing batch reactors were operated at two different, low (n=4) and high (n=3), F:M (0.19-0.36 mgCOD/mgTSS/d) and C:N (3.5-6.3 mgCOD/mgTKN) conditions for a period of 74 days, following 53 days of sludge acclimation. The inoculum was taken from a full-scale treatment plant. Resilience was tested during the last 14 days by operating all seven reactors at low F:M-C:N. Samples were analyzed using metagenomics, 16S rRNA gene amplicon sequencing, and effluent characterization. High F:M-C:N reactors exhibited different ecosystem functions and nitrifier abundance compared to the ones at low F:M-C:N. Perturbed high F:M-C:N reactors displayed quantifiable and initially variable functional resistance. Stable nitrite accumulation (77%) was achieved through high F:M-C:N loading with concurrent suppression of Nitrospira, revealing a new partial nitrification strategy for nitritation-denitritation systems. Subsequently, only two of the three reactors experiencing a switch back from high to low F:M-C:N recovered the nitrification function, with an increase in Nitrobacter (r-strategist) abundance as the predominant NOB replacing the niche initially occupied by Nitrospira (K-strategist). Overall, the AOB community was more diverse and resilient than the NOB community. We showed that functional resistance and resilience can vary across replicate reactors in a closed system, and that nitrification resilience need not coincide with a return to the initial nitrifying community structure.


37
Improving stability and optimizing performance of wastewater treatment processes are central 38 tenets of environmental engineering and biotechnology, to help achieve the sustainable development 39 goal of guaranteeing availability and sustainable management of water and sanitation for everyone 40 [1]. In ecology, perturbations or disturbances are believed to have direct effects on the stability of 41 ecosystems by altering community structure and function (Table 1) [2]. For engineered systems like 42 activated sludge bioreactors, it is important to identify the effect of different disturbances on the 43 microbial community structure so as to relate them to changes in process performance [3]. However, 44 disturbance is also deemed to affect underlying mechanisms of community assembly [4] which, if 45 predominantly stochastic, could drive microbial communities to divergent trajectories in terms of 46 composition and function [5]. Therefore, robust replication is required to properly assess the effect of 47 perturbations in the stability of sludge bioreactors. 48 Sustained perturbations or press disturbances that impose a long-term continuous change of 49 species densities through an alteration of the environment [6,7], are relevant since they can lead a 50 system to alternative stable states that may or may not be reversible in terms of both community 51 composition and function [8]. Press disturbance studies on wastewater treatment systems often 52 employ harsh perturbations like toxic aromatic pollutants [9] or washout of organisms through sludge 53 discharge [10,11]. However, alterations in the environment that are not directly detrimental for 54 organisms but still provide opportunities for low abundance members within the community are also 55 considered a disturbance [12]. In bioreactors, a switch in the substrate feeding scheme employed 56 could then elicit changes in community structure and function. Such perturbations can occur in the 57 form of organic shocks within activated sludge [13][14][15] and anaerobic reactor systems [16,17]. In 58 wastewater treatment, the food-to-biomass ratio (F:M) or sludge loading rate is an important 59 parameter as it determines the growth type and settleability of sludge microorganisms [18]. Moreover, 60 the F:M is also related to the carbon-to-nitrogen ratio (C:N) as both depend on the amount of organic 61 carbon in the feed. However, the few studies describing the effect of alterations in F:M on sludge 62 microbial communities often neglect uncontrolled covariations in C: N [19, 20], despite the fact that 63 nitrite and nitrate) in the liquid phase using colorimetric tests and ion chromatography. Nitrite 116 accumulation percentage in the effluent was calculated as the ratio of nitrite concentration and the 117 sum of nitrate and nitrite concentrations. Total organic carbon and total Kjeldahl nitrogen (TKN) were 118 also measured in the influent. To control the F:M, sludge biomass was measured as total (TSS) and 119 volatile suspended solids (VSS) twice a week, after which sludge wastage was done to target 1500 120 mg/L of TSS. Sludge volume index (SVI) was calculated from the liquid and sludge volumes 121 measured in the reactors after 30 min settling and the TSS values obtained in the same cycle. 122 Microbial community function was also investigated in the form of intensive sampling (every 30 to 60 123  Sequences were truncated after 280 and 255 nucleotides for forward  142   and reverse reads, respectively, length at which average quality dropped below a Phred score of 20.  143 After truncation, reads with expected error rates higher than 3 and 5 for forward and reverse reads 144 were removed. After filtering, error rate learning, ASV inference and denoising, reads were merged 145 with a minimum overlap of 20 bp. Chimeric sequences (0.18% on average) were identified and 146 removed. For a total of 104 samples, 19679 reads were kept on average per sample after processing, 147 representing 49.2% of the average input reads. Taxonomy was assigned using the SILVA database 148 (v.132) [40]. Adequacy of sequencing depth after reads processing was corroborated with rarefaction 149 curves at the ASV level (Fig. S2). 150

Metagenomics sequencing and reads processing 151
Libraries were sequenced in one lane on an Illumina HiSeq2500 sequencer in rapid mode at a 152 final concentration of 11pM and a read-length of 250 bp paired-end. In total, around 325 million 153 paired-end reads were generated, with 3.4 ± 0.4 million paired-end reads on average per sample (total 154 48 samples). Illumina adaptors, short reads, low quality reads or reads containing any ambiguous 155 bases were removed using cutadapt [41]. High quality reads (91.0 ± 1.4% of the raw reads) were 156 randomly subsampled to an even depth of 4,678,535 for each sample prior to further analysis. 157 Taxonomic assignment of metagenomics reads was done following the method described by Ilott 158 [42]. High quality reads were aligned against the NCBI non-redundant (NR) protein database (March 159 2016) using DIAMOND v.0.7.10.59 [43] with default parameters. The lowest common ancestor 160 approach implemented in MEGAN Community Edition v.6.5.5 [44] was used to assign taxonomy to 161 the NCBI-NR aligned reads with the following parameters (maxMatches=25, minScore=50, 162 minSupport=20, paired=true). On average, 36.8% of the high-quality reads were assigned to cellular 163 organisms, of which 98.4% were assigned to the bacterial domain. Adequacy of sequencing depth was 164 corroborated with rarefaction curves at the genus taxonomic level (Fig. S2). 165

Dynamics in bioreactor performance 167
During the acclimation phase, the F:M-C:N values were maintained at 0.21 (mg COD/mg 168 TSS/d) and 3.5 (mg COD/mg TKN), respectively ( Table 2). Ammonium concentrations in the effluent 169 decreased gradually while nitrate concentrations increased (Fig. 1). Sludge related parameters like 170 settleability (SVI) and biomass fraction (VSS:TSS) varied during this acclimation period (Fig. S3). 171 Most of these variations decreased after 30 d and trends were stable after 45 d. 172 During the perturbation phase of the study (d54 onwards), sludge was wasted more often to 173 better control the TSS and thus the F:M ( To ensure that the partial nitrification was due to different F:M-C:N values and not a lack of available 182 dissolved oxygen, we increased the aeration rate from 1 to 4 L min -1 from d97 onwards without 183 observing significant changes in effluent compounds. The last two weeks of the study involved 184 shifting operational parameters in the high F:M-C:N reactors to match those of the low F:M-C:N ones 185 (Table 2). During this period a transition towards recovery of the nitrification function was observed, 186 with high variability of effluent values for NO 2 --N, NO 3 --N and COD across reactors (Fig. 1). 187

Dynamics in nitrification and nitrifiers 188
Nitrite accumulation was found in the effluent of high F:M-C:N reactors, together with a higher 189 residual COD (Fig. 1). The acclimation phase (d1-d53) displayed negligible nitrite accumulation at 190 0.3% (± 1.5%). Low F:M-C:N reactors had zero percent nitrite accumulation after d61 and only 1.1% 191 (± 2.0%) during the first week (d54-d60). Conversely, high F:M-C:N reactors showed an initial 192 transient nitrite accumulation of 18% (± 21%) on d54-d60, which subsequently increased and 193 stabilized at 77% (± 6.0%) during the d61-d113 period. Finally, after shifting from high to low F: M-194 C:N conditions, nitrite accumulation decreased to 55% (± 29%) in the first week (d114-d120), and all 195 the way to zero in the second week (d121-d127) for two of the three reactors ( Fig. 1). Following the shift from high to low F:M-C:N it was Nitrobacter that rose to be the dominant 204 NOB instead of Nitrospira, but only in two of three reactors (Fig. 2). Variations in performance 205 among replicate reactors were also evident from cycle study profiles before (d110) and after (d124) 206 the shift in feeding regime for the high F:M-C:N reactors (Fig. 3). Two weeks after the change, only 207 two high to low F:M-C:N reactors displayed functional profiles similar to those of the low F:M-C:N 208 reactors. The reactors which recovered functionality were the same as those that registered around 1% 209 of Nitrobacter abundance (Fig. 2). 210 The higher resolution of 16S rRNA gene amplicon sequencing allowed us to taxonomically 211 identify four ASVs as Nitrospira, twenty-one as Nitrosomonas, and one as Nitrobacter (Fig. 4). From 212 these, only two ASVs were identified at the species level. The genus Nitrospira was dominated by the 213 N. defluvii species, with the other three ASVs detected only at low abundances across the low F:M-214 C:N reactors after d97, the day the aeration rate was increased. The dominant Nitrosomonas ASVs 215 during the perturbation phase were different from the initial ones. Furthermore, N. europaea and 216 Nitrosomonas ASV-70 saw their relative abundances increasing with time across high F:M-C:N 217 reactors (Fig. 4). 218

Function stabilization and nitrifier dynamics during acclimation phase 220
Sludge acclimation served to stabilize important functions like nitrification, organic carbon 221 removal, and sludge settling capacity across reactors (Fig. 1, Fig. S3). The most abundant nitrifying 222 genus was Nitrospira, which increased around 50% after the acclimation stage. The next most 223 abundant nitrifier, Nitrosomonas, increased sixfold ( Fig. 2) and the changes in abundance of different 224 Nitrosomonas ASVs suggested a succession of organisms within this genus (Fig. 4). Additionally, 225 through the use of exact sequence variants [39] we could observe that Nitrosomonas was the most 226 diverse nitrifying genus with twenty-one different ASVs being detected (Fig. 4). 227

Disturbance leads to stable partial nitrification unveiling system's resistance 228
Ecosystem function in terms of COD removal, ammonia removal, and complete nitrification 229 was optimal and stable for the low F:M-C:N reactors, particularly towards the end of the study (

Dynamics of nitrifiers during the perturbation phase 245
Understanding community and activity dynamics of nitrifying bacteria is essential for 246 improving design and operation of wastewater treatment biological processes [47]. The suppression of 247 Nitrospira in the high F:M-C:N reactors (Fig. 2) was likely due to competition for DO with 248 heterotrophs and Nitrosomonas that possess a higher affinity for DO [48,49]. The same competition 249 for DO occurs between Nitrosomonas and heterotrophs, which explains the observed lower abundance 250 of this genus in high F:M-C:N reactors. It is known that AOB have a higher oxygen affinity than NOB 251 [50]. Still, the increase in aeration rates from d97 onward did not prevent nitrite accumulation, 252 indicating that a low DO in the system was not the main reason behind our observations. DO 253 concentrations higher than 1 mg/L are enough to achieve optimal nitrification performance [51], 254 which was the case for the reactors in this study (Fig. 3). However, nitrifying communities grow in 255 stratified biofilms where AOB are located closer to the water interface and NOB are in the interior 256 zone [52, 53]. The concentration of oxygen deep inside a biofilm or floc is lower than in the mixed 257 liquor. Moreover, stratification in AOB biofilms due to an increase in C:N has been reported [48], 258 highlighting that increases in biofilm thickness due to heterotrophic growth further reduce oxygen 259 diffusion inside, which is detrimental to the growth of NOB. In our study, the period during the 260 aerobic phase of a cycle when almost all ammonia had been removed and COD concentrations were 261 either low or remained constant (around 400-500 min, Fig. 3) implies that heterotrophs and AOB were 262 not competing with NOB for oxygen anymore. Although this should have provided sufficient oxygen 263 to NOB to be active during the remainder of the aerobic phase, an increase in nitrate production was 264 not observed. A possible reason for this could be nitrite accumulation, which was reported to be toxic 265 to NOB at high concentrations [54]. In summary, the observed NOB suppression at high F:M-C:N 266 could have been due to a combination of competition for DO with heterotrophs and Nitrosomonas, 267 reduced oxygen diffusion into the nitrifier biofilm due to heterotrophic growth, and nitrite 268 accumulation due to AOB activity. 269 AOB growth rates are normally higher than those of NOB at 30 °C, which implies that the 270 SRT can be reduced to achieve partial nitrification [55]. In our study, increasing F:M while keeping 271 TSS constant implied an SRT reduction of 35% in the high F:M-C:N reactors compared to the low 272 F:M-C:N reactors. It is conceivable that part of the observed reduction in nitrifiers was due to washout 273 given their low growth rates. It was suggested based on mathematical modelling that a reduction in 274 SRT has a stronger effect on NOB than on AOB [56]. However, the SRT of 8. Our reactors constituted a closed system, which implies that the Nitrobacter colonizers came 351 from low-abundance seed-bank populations (Table 1) Here we showed that functional resilience greatly differed from nitrifying community 358 resilience. Reactors with recovered function after returning to low F:M-C:N conditions (Fig. 3) 359 remained distinct from the control reactors in terms of NOB composition (Fig. 2)    suggest that AOB populations were more diverse than NOB populations. 657 Table 1. Glossary of key microbial ecology concepts 658

Concept Definition
Community A group of populations of two or more different microbial taxa (e.g. nitrifiers) occupying the same geographical space (e.g. bioreactor) at a particular time [2].
Community structure The composition and relative abundance of different microbial taxa within a community at a given time [2].

Disturbance
An event that significantly alters the environment of a community, creating opportunities for individuals to grow and reproduce [12] (e.g. a change in feeding strategy). Also referred to as perturbation.

Ecosystem
A community of living organisms in conjunction with the abiotic components of their environment, interacting as a system [2] (e.g. a bioreactor).
Ecosystem function Outputs of functional processes at a whole ecosystem level [83] (e.g. ammonia removal within a bioreactor).

Immigration
Large-scale movement of members of a species to a different environment [84] (e.g. bacteria within the feed entering a bioreactor).

Niche
The abiotic and biotic conditions that a microbial taxa need to grow, survive and reproduce [85].
Press disturbance A long-term or continuous disturbance [7] (e.g. a sustained change in organic loading).
r-and K-strategists Classification given to organisms under the r-vs. K-strategies ecological framework. Among nitrifiers, Nitrosomonas and Nitrobacter are rstrategists able to grow fast under high substrate concentration conditions, while Nitrosospira and Nitrospira are K-strategists that grow slowly but are adapted to low substrate concentrations [70, 72].

Resilience
The capacity of a community and/or function to return to a pre-disturbance condition [86].

Resistance
The degree to which a community and/or function is insensitive to a disturbance [86].
Seed-bank population Refers to the members within a community that are present in low abundances, but are able to grow when given the appropriate conditions [87].

Stability
The response of a community to disturbance [12]. It is comprised of resistance and resilience, which are two quantifiable metrics useful for