Prolonged Stagnation Reduces Treated Wastewater Biostability by Altering Microbial Community: Insights From Metaproteomics

ABSTRACT

Reclaimed wastewater is increasingly reused for irrigation and other non‐potable applications; however, inadequately treated effluent has raised concerns regarding environmental and public health impacts. Water quality in reclaimed distribution systems is shaped by multiple factors, particularly hydraulic stagnation in pipes and storage reservoirs. Stagnation can alter microbial community stability and facilitate persistence of pathogenic taxa. To investigate how prolonged stagnation affects microbial community structure and function, we integrated metagenomics and metaproteomics analyses of biofilms under flow and stagnant conditions over 3, 5 and 7 months. Prolonged stagnation caused pronounced compositional shifts, including strong reductions in nitrogen‐removing taxa such as Nitrospira and Nitrosomonas. Correspondingly, key nitrification and denitrification proteins were depleted ≥ twofold under stagnation, indicating impaired nitrogen conversion processes. Stagnation also enriched motility‐ and transport‐related functions and promoted Acidovorax persistence, a genus including phytopathogenic species. In contrast, flow conditions sustained nitrogen‐cycling activity, contaminant‐degrading enzymes, and quorum‐quenching proteins, supporting greater biostability. Overall, our findings show that prolonged stagnation disrupts microbial community balance, suppresses essential nitrogen‐cycling and detoxification pathways, and reduces the functional robustness of treated wastewater. Maintaining hydraulic flow within reclaimed water systems is therefore critical for preserving microbial functionality and ensuring safe and reliable reuse in irrigation and other non‐potable applications.

Water quality in reclaimed distribution systems is shaped by hydraulic stagnation. Prolonged stagnation suppresses essential nitrogen‐cycling and detoxification pathways and reduces the functional robustness of treated wastewater. Maintaining hydraulic flow within reclaimed water systems is critical for ensuring safe and reliable reuse in irrigation and other non‐potable applications.

1. Introduction

Treated wastewater is increasingly used for agricultural and landscape irrigation, and agriculture alone accounts for 69% of global water withdrawals. The demand for wastewater reuse continues to rise, particularly, in water‐scarce regions where reclaimed effluents have become an essential alternative water source. To safeguard food and environmental health, the World Health Organization (WHO) and the Food and Agriculture Organization (FAO) have developed sanitation‐based guidelines that define health‐based targets, outline agricultural water‐quality management requirements, identify key hazards and provide protective measures and risk‐assessment frameworks for the safe use of reclaimed water (World Health Organization ²⁰⁰⁶; Drechsel et al. ²⁰²³; United Nations Environment Programme ²⁰²³). Despite these guidelines, the use of inadequately treated wastewater remains a documented concern for food safety (Helmecke et al. 2020). This underscores the need to understand how treated wastewater quality is shaped not only by treatment efficiency but also by the conditions encountered during storage and distribution.

The quality of treated wastewater during distribution and storage can be altered by environmental factors such as temperature, pH and the length and age of distribution pipes, all of which influence its physical, chemical and biological characteristics. These fluctuations may result in reduced water quality at the point of use and can negatively affect agricultural systems, including the emergence of phytopathogens capable of damaging soils and crops (Dion 2010; Mohammed et al. ²⁰²¹; Xu et al. ²⁰²¹; Kokina et al. ²⁰²²). In addition to these environmental factors, hydraulic conditions play a critical role in altering treated wastewater quality, shaping the microbial composition and the functional activities (Douterelo et al. 2013). Continuous flow dynamics, maintained through regular flushing, improve disinfectant residual concentrations and limit contaminant accumulation in distribution systems (Vedrin et al. 2024; Waegenaar et al. ²⁰²⁵). In contrast, extended residence times (stagnation), particularly over months with minimal hydraulic movement as may occur during pandemic‐related shutdowns, maintenance periods or between crop cycles, can lead to water quality deterioration and increased microbial risks (Nisar et al. 2020; Bhaskar and Singh ²⁰²⁵). Studies have demonstrated that as few as 6 days of stagnation can significantly alter microbial community composition and increase the abundance of opportunistic pathogens in drinking water, including Legionella pneumophila , Mycobacterium avium and Pseudomonas aeruginosa (Ling et al. 2018). Stagnation has further been identified as a major driver of Legionella spp. proliferation (Nisar et al. 2020; Montagnino et al. ²⁰²²). However, while stagnation is well studied in drinking water systems, its impacts on treated wastewater remain poorly understood, particularly, with respect to microbial structure and functional responses that underpin the system's biostability. In this study, biostability is operationally assessed through three metaproteomic indicators directly measured in the dataset: (i) the diversity of expressed proteins, quantified as Shannon diversity of the metaproteomic profile; (ii) the relative preservation of nitrogen‐cycling enzyme expression and (iii) the maintenance of proteins associated with xenobiotic transformation.

To comprehensively assess these impacts, appropriate methodological tools are essential. The study of microbial dynamics depends on the analytical approach used. While metagenomics (MGs) provides insights into microbial diversity and functional potential, it does not directly reflect microbial activity (Van Den Bossche et al. 2021). In contrast, metaproteomics has emerged as a powerful tool for assessing actual functional responses in complex environments (Van Den Bossche et al. 2025). Integrating both approaches can improve understanding of how prolonged stagnation alters microbial community structure and function in treated wastewater systems, potentially supporting the persistence of pathogens (Salvato et al. 2021). Understanding these functional traits involved could inform strategies to mitigate risks associated with microbial survival in reclaimed water distribution systems. Metaproteomics is gaining momentum across multiple fields, including medical and environmental studies (Armengaud 2023). Yet, the combination of MGs and metaproteomics has seldom been used to investigate the complex bacterial diversity in biofilms, especially for treated wastewater systems.

In this study, we integrated metagenome sequencing and metaproteomics to obtain comprehensive taxonomic and functional profiles of microbial communities, investigating the impact of stagnation compared with continuous flow dynamics. To achieve this, we employed laboratory‐scale reactors fed with membrane bioreactor (MBR) effluent. The objectives were to evaluate: (i) how stagnation affects the diversity and taxonomic composition of microbial genes and proteins; (ii) how prolonged stagnation over 3, 5 and 7 months drives temporal changes in community and functional dynamics; (iii) which functions are enriched or depleted under stagnation and (iv) how these taxonomic and functional shifts relate to the ecological quality and biostability of reclaimed water. By understanding not only what microbes are present but also what they are doing through metaproteomics, we can assess how microbial activities impact water quality. This approach enables evaluation of treated wastewater quality both post‐treatment and during distribution and storage in irrigation systems.

2. Materials and Methods

2.1. Reactors Setup—Operation and Sampling

The experimental setup was designed to test flow and stagnant conditions using laboratory‐scale aerobic Center for Disease Control (CDC) reactors (Biosurface Technologies, USA) (see detailed workflow in Figure S1, figure was created by BioRender.com). The reactors were fed with treated wastewater effluent from an MBR‐based wastewater treatment plant in KAUST, with operating conditions described previously (Jumat et al. 2017). Fresh MBR effluent used to feed the reactors contained free chlorine concentrations ranging from 0.2 to 0.5 mg/L, measured using a Hach CN‐70 colorimetric test kit (Hach, USA; Cat. No. 1454200) prior to reactor feeding, operated at 30°C ± 2°C in the dark. The control (continuous flow) samples were maintained at a continuous flow rate of 1.3 mL/min and a stirring rate of 200 rpm to impose a controlled shear environment typical for CDC biofilm cultivation. Previous fluid dynamic characterisations of CDC reactors indicate that coupon surface shear stresses under rotation occur on the order of tenths of a Pascal, consistent with low to moderate shear conditions relevant to drinking water distribution system hydraulics (Johnson et al. 2021; Silva et al. ²⁰²⁵). The hydraulic retention time (HRT) was calculated to be approximately 5 h.

Stagnant conditions were implemented using a fill‐and‐draw protocol in which 350 mL of fresh MBR‐chlorinated effluent was exchanged weekly while biofilm coupons remained attached and undisturbed. Water replacement was performed every 7 days throughout the experimental period, corresponding to a 7‐day water exchange interval under stagnation. Between exchange events, reactors were maintained under completely stagnant conditions without flow or mixing. The reactor working volume of 350 mL was selected to match the wet volume used under flow conditions. Extended stagnation periods and intermittent usage cycles are characteristic of low‐use zones, dead ends in distribution systems and premise plumbing during prolonged reductions in water demand, including building closures during the COVID‐19 pandemic, where chemical disinfectant residuals decline and microbial communities evolve during multiple days without flow (Zlatanović et al. 2017). Biofilms were grown for three time periods (3, 5 and 7 months) to capture long‐term biofilm development under consistently applied hydraulic conditions. Biofilms were harvested using 1X phosphate‐buffered saline (PBS) solution. The experiment was performed in two biological replicates (two independent setups and operation times).

2.2. DNA Extraction and MGs Analysis

The harvested biofilms were pelleted, and DNA was extracted using DNeasy PowerLyzer PowerSoil Kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol. DNA concentration was quantified using Qubit dsDNA HS (High Sensitive) Assay Kit (Thermo Scientific, Invitrogen, Carlsbad, CA, USA). Two samples were carried out as negative control with only 1X PBS and showed that DNA concentration was below the detection limit (0.5 ng/mL).

The MG libraries were prepared using Ovation Ultralow library systems V2 (NuGEN, San Carlos, CA, USA). Briefly, genomic DNA for each sample was normalised to 8 ng and sheared to an average size of 500 bp using M220 ultrasonicator (Covaris, Woburn, MA, USA). Sheared DNA samples were subjected to paired‐end indexed library construction and purified using AMPure XP beads (0.8×; Beckmann Coulter Genomics, Brea, CA, USA). The concentrations of the purified libraries were measured using Qubit dsDNA HS Assay Kit and the quality was assessed using Bioanalyzer (Agilent Technologies, Santa Clara, USA). The indexed libraries were pooled in equimolar concentration for shotgun MG sequencing (Bidirectional sequencing of 150 bp), carried out on an Illumina NovaSeq 6000 platform and submitted to the Bioscience Core Lab facility at King Abdullah University of Science and Technology, Saudi Arabia.

Raw MG reads were quality‐trimmed using Trimmomatic (Bolger et al. 2014) and assembled de novo with MEGAHIT (Li et al. 2015). Contigs were filtered using DeepMicroClass (Hou et al. 2024) to retain prokaryotic‐associated sequences. MG reads were aligned to these contigs using BWA and depth of coverage was computed with appropriate tools. Binning was performed with five tools: (i) Maxbin2 (Wu et al. 2016), (ii) Metabat2 (Kang et al. 2019), (iii) CONCOCT (Alneberg et al. 2014), (iv) VAMB (Nissen et al. 2021), Semibin2 (Pan et al. 2023) in (iv) single‐ and (v) multi‐sample mode, followed by refinement with MAGscot (Rühlemann et al. 2022) and dereplication using dRep (Olm et al. 2017), yielding a non‐redundant set of metagenome‐assembled genomes (MAGs). MAGs were taxonomically classified with CAT/BAT (Von Meijenfeldt et al. 2019; Meijenfeldt et al. ²⁰²⁴) and The Genome Taxonomy Database (GTDB) (Chaumeil et al. 2022), then functionally annotated using Bakta (Schwengers et al. 2021). CAT/BAT and GTBD integration was performed to enhance the annotation and taxonomic profiling (Hauptfeld et al. 2024). MAG‐ and gene‐level quantification was performed using Salmon (Patro et al. 2017), and counts were summarised with R Statistical Environment. All Bakta‐predicted proteins from MAGs were compiled into custom FASTA databases with taxonomic metadata and decoy/contaminant sequences for compatibility with FragPipe (https://github.com/Nesvilab/FragPipe) workflows.

2.3. Protein Extraction and Metaproteomics Analysis

In a previous study, different protein extraction methods were evaluated for their efficiency in extracting proteins from biofilms (Almulhim and Hong 2023). Therefore, proteins in this study were extracted from the best evaluated method using the iST kit (PreOmics GmbH, Martinsried, Germany) in three technical replicates (independent samples processing) with minor modifications to the manufacturer's protocol. In detail, biofilm pellets were lyophilised overnight, a maximum of 0.22 mg dry weight was resuspended in 50–100 μL lysis buffer (the volume was adjusted according to starting material) and incubated at 95°C for 10 min. To reduce and alkylate cysteine residues, 25 mM Tris(2‐carboxyethyl)phosphine (TCEP) was added, and samples were incubated at 37°C while shaking at 1000 rpm for 30 min. To ensure further protein denaturation stability, 56 mM iodoacetamide (IAA) was added, and samples were incubated at room temperature, shaking at 1000 rpm for 30 min. Protein concentrations were determined using the Pierce Micro BCA assay kit (Thermo Fisher Scientific Inc.), and volumes containing a maximum of 100 μg of protein were subjected to digestion by adding 40:1 Trypsin/Lys‐C mix and incubating at 37°C, shaking at 1000 rpm overnight.

Samples were loaded onto the cartridge filters for purification and spun down at 2200 rcf for 2 min or until complete flow‐through. Wash 0 was prepared by mixing 100% hexane and 0.2% trifluoroacetic acid (TFA) and loaded 200 μL into the samples and spun down at 2200 RCF for 2 min. This step was repeated three times. Wash X from the PreOmics Phoenix kit for peptide clean‐up was used to carry out the purification step by adding 200 μL to the samples and spinning down at 2200 RCF for 2 min. This step was repeated three times. Then, proceeded with wash 1 by adding 200 μL to the samples and spinning down at 2200 RCF for 2 min; this step was repeated twice. The final washing step was adding 200 μL of wash 2 once. Purified peptides were recovered by adding 100 μL elute solution and spinning down at 2200 RCF for 2 min. The elution step was repeated twice and placed in a SpeedVac (45°C) for approximately 2 h. Peptides were resuspended in 50–100 μL LC‐LOAD solution (the volume was adjusted according to the final peptide concentrations). Peptide concentrations were measured using Pierce Quantitative Colorimetric Peptide Assay (Thermo Fisher Scientific Inc.).

2.4. Liquid Chromatography–Tandem Mass Spectrometry Analysis

LC–MS/MS measurement was performed using an Orbitrap Exploris 480 mass spectrometer (Thermo Fisher Scientific) coupled to a Vanquish Neo nano‐HPLC system (Thermo Scientific). Peptide samples were first loaded onto a PepMap NEO C18 trap column (300 μm × 5 mm, 5 μm particle size; Thermo Scientific), and subsequently separated on a PepMap RSLC C18 analytical column (2 μm particle size, 100 Å pore size, 75 μm × 50 cm; Thermo Scientific). Chromatographic separation was conducted at a flow rate of 250 nL/min over an 87‐min run time. The mobile phases consisted of water with 0.1% formic acid (solvent A) and 95% acetonitrile with 0.1% formic acid (solvent B). Peptides were eluted using a multistep linear gradient of solvent B as follows: 2%–55% over 75 min, followed by a ramp to 99% B over 2 min. The column was then held at 99% B for an additional 10 min to ensure complete peptide elution before re‐equilibration with 2% B.

Electrospray ionisation was initiated by applying 1900 V to the EASY‐Spray emitter. Data acquisition was performed using Xcalibur software in data‐dependent acquisition (DDA) mode, employing a top speed method with a cycle duration of 3 s. The Orbitrap mass spectrometer was operated in positive ion mode with FAIMS (Field Asymmetric Ion Mobility Spectrometry) enabled at a compensation voltage (CV) of −45 V. Full MS survey scans were acquired in profile mode over an m/z range of 350–1400 Th at a resolution of 120,000. The automatic gain control (AGC) target was set to a custom value corresponding to 300% normalised AGC, with the maximum injection time set to auto. The most intense precursor ions with charge states from 2+ to 5+ were selected for fragmentation via higher‐energy collisional dissociation (HCD) using a fixed collision energy mode. The normalised collision energy (NCE) was set to 26%, and precursor ions were isolated with a 1.6 Th isolation window. Fragment ion spectra were acquired at a resolution of 15,000, with AGC set to standard and the injection time set to auto.

Protein identification and label‐free quantification (LFQ) were performed using FragPipe v 23.0 (Kong et al. 2017). All Bakta‐predicted proteins from MAGs were compiled into custom FASTA databases with taxonomic metadata. Database searches were conducted against a custom protein database generated from Bakta‐predicted proteins from MAGs in this study, comprising 1,025,600 sequences. Carbamidomethylation of cysteine residues was set as a fixed modification, while oxidation of methionine, deamidation of asparagine and glutamine and N‐terminal acetylation of proteins were specified as variable modifications. Up to two missed cleavages were allowed for tryptic digestion. The precursor and fragment mass tolerances were set to 20 ppm and 0.02 Da, respectively. A false discovery rate (FDR) of 1% was applied at both the peptide‐spectrum match (PSM) and protein levels. Only proteins identified with a minimum of two distinct peptides (unique + razor) were considered for further analysis. Quality‐control metrics for all LC–MS/MS runs, including injected peptide load, total ion current (TIC), MS/MS spectra, PSMs, identified peptides and identified proteins, are provided in Table S1 and summarised in Figure S2.

2.5. Statistical Analysis

For MGs, diversity and richness metrics were calculated from two biological replicates per time point. For metaproteomics, three technical replicates were acquired per biological replicate and aggregated before analysis. To assess overall differences between flow and stagnant conditions with adequate statistical power, comparisons of Shannon diversity, genus richness and protein richness (Figure 1) were performed by pooling all timepoints (3, 5 and 7 months) within each condition (total n = 6 per condition). This pooling approach enables statistical assessment of overall condition‐level differences while preserving temporal information for exploratory visualisation in subsequent figures. These pooled comparisons were assessed using permutation tests and Wilcoxon rank‐sum tests, and statistical significance in boxplots was visualised using bracket annotations generated with the ggpubr package (Kassambara 2025). Subsequent figures display data separated by timepoint to illustrate temporal patterns; however, the analysis was intended to evaluate overall flow‐versus‐stagnant effects rather than timepoint‐specific differences. Temporal patterns presented in subsequent figures are therefore exploratory and descriptive.

FIGURE 1.

Alpha and beta diversity of microbial genes and proteins under flow and stagnant conditions. (A) Shannon diversity of metagenomics communities (Wilcoxon rank‐sum test; p = 0.11). (B) Shannon diversity of metaproteomic profiles (Wilcoxon rank‐sum test; p = 0.002). (C) Principal coordinates analysis (PCoA) of metagenomics communities based on Bray–Curtis dissimilarity (PERMANOVA: Pseudo‐F _1,10 = 7.83, R ² = 0.44, p = 0.0069). Middle cross represents the centre. (D) PCoA of metaproteomic functional profiles based on Bray–Curtis dissimilarity (PERMANOVA: Pseudo‐F _1,10 = 4.02, R ² = 0.29, p = 0.0109). Boxplots were performed on six biological samples per condition (Flow n = 6; Stagnant n = 6), pooled across the 3, 5 and 7 month timepoints to visualise the overall trend of flow and stagnation. Timepoints are shown as distinct point shapes for visualisation purposes only; temporal patterns are exploratory and descriptive.

Protein quantification and differential abundance analyses were carried out in FragPipe‐Analyst (Hsiao et al. 2024). A target‐decoy strategy was implemented by appending reversed decoy sequences (rev_) to the custom database prior to peptide‐spectrum matching. Peptide and protein identifications were filtered at a 1% FDR within the FragPipe/IonQuant workflow and only proteins identified with a minimum of two unique peptides were retained for downstream analysis. A total of 25,022 protein groups were identified prior to filtering, of which 14,088 were retained for downstream statistical analysis. Protein intensities were processed using variance‐stabilising normalisation (VSN). Missing values were subsequently imputed using the default Perseus‐like left‐shift approach implemented in FragPipe‐Analyst, where missing values were replaced by random draws from a Gaussian distribution with a width of 0.3 and a down‐shift of 1.8 standard deviations (SDs) relative to the mean of each sample. Differential expression testing was performed using protein‐wise linear models with empirical Bayes statistics via the limma Bioconductor package. Significance was determined using Benjamini–Hochberg FDR correction, and proteins were considered differentially abundant at adjusted p < 0.05 and |log₂ fold‐change| ≥ 1. All exported normalised matrices and DE tables were subsequently processed and visualised in R (v4.4.1) using dplyr (Wickham et al. 2022), tidyverse (Wickham et al. 2019) and ggplot2 (Wickham 2009).

Taxonomic identity of metaproteomic proteins was derived from the MAG of origin. MAGs were taxonomically classified using the integrated CAT/BAT and GTDB‐Tk workflows, and taxonomic annotations were incorporated into the custom proteomics database during database construction, allowing each identified protein to inherit the taxonomy of its source MAG. Shared peptides were handled during protein inference in FragPipe using a parsimony‐based protein grouping approach (Kong et al. 2017; Hauptfeld et al. ²⁰²⁴; Van Den Bossche et al. ²⁰²⁵).

All processed metaproteomic data used in this study, including protein identifications, taxonomy assignments, normalised MaxLFQ intensities, differential abundance statistics (log₂ fold‐changes and adjusted p‐values) and functional annotations (COG, KEGG, GO), are provided in Supporting Information Dataset S2. A curated list of all statistically significant proteins across timepoints, including protein, taxonomy, enrichment direction, fold‐changes, adjusted p‐values and functional annotations, is provided in Table S3.

3. Results

3.1. Diversity Patterns of Microbial Genes and Proteins Under Flow and Stagnation

To address our first objective of understanding how stagnation shapes microbial community composition and protein diversity, we first assessed alpha and beta diversity metrics from both MG and metaproteomic datasets. Although Shannon diversity from MGs did not differ significantly between flow and stagnant (Wilcoxon rank‐sum test; p > 0.1), stagnant samples exhibited a modest but non‐significant 13% higher mean diversity of 2.27 compared to flow (2.01), corresponding to a 1.13‐fold increase (Figure 1A). In contrast, metaproteomic profiles showed the opposite pattern, with flow displaying significantly higher Shannon diversity than stagnant (8.17 vs. 7.10; Wilcoxon rank‐sum test; p = 0.002), representing a 15% increase (1.15‐fold) (Figure 1B).

Beta diversity, evaluated using principal coordinates analysis (PCoA), revealed clear compositional differences between hydraulic conditions in both datasets. Significant separation was detected in MG microbial genes (pseudo‐F _1,10 = 7.83, R ² = 0.44, p = 0.0069; Figure 1C) and metaproteomic functional profiles (pseudo‐F_1,10 = 4.02, R² = 0.29, p = 0.0109; Figure 1D). Timepoints (3, 5 and 7 months) were shown as different point shapes in the PCoA for visualisation purposes, but were not analysed as independent factors due to limited replication per timepoint (n = 2); therefore, statistical comparisons focused on overall condition effects rather than temporal dynamics. While MG data did not exhibit distinct temporal patterns within each condition, metaproteomic data showed tighter clustering of protein expression profiles across all months under flow, whereas stagnant samples formed more separated clusters by timepoint.

3.2. Taxonomic Composition and Temporal Dynamics Reveal Condition‐Specific Community Shifts

Building on the diversity patterns, we examined taxonomic composition at the genus level and evaluated how community structure changed over 3, 5 and 7 months. The number of genera detected at both the gene and protein levels followed a similar pattern, with flow conditions exhibiting slightly higher richness than stagnant conditions, although the enrichment under flow was not statistically significant (Wilcoxon rank‐sum test; p > 0.05) (Figure 2A). In contrast, the genus‐level composition derived from the metaproteomic profile revealed clear shifts in community structure shaped by hydraulic conditions (Figure 2B). Across all samples, 54 genera were detected: 29 were enriched under flow and 25 under stagnant conditions.

FIGURE 2.

Taxonomic community distribution at the genus level. (A) Genus richness follows a similar trend in metagenomics and metaproteomics with no significant difference (Wilcoxon rank‐sum test; p > 0.05) (Flow n = 6; Stagnant n = 6), pooled across the 3, 5 and 7 months timepoints. Timepoints are shown as distinct point shapes for visualisation purposes only; temporal patterns are exploratory and descriptive. (B) Mean log₂ fold‐change (stagnant vs. flow) plotted against mean relative abundance for the most abundant genera, selected based on their total relative abundance across all samples. Bubble size represents the number of proteins annotated from each genus, (C) the top genera highly expressed in stagnant (Acidovorax) while depleted in flow and the top genera highly expressed in flow (Nitrospira) while depleted in stagnant, the shaded area represents the standard deviation ± SD (n = 2). This panel presents an exploratory temporal visualisation based on n = 2 biological replicates per timepoint; temporal patterns are descriptive and not statistically tested.

Among the dominant taxa, Nitrospira was strongly enriched in flow, accounting for ~25% of the community compared to < 1% in stagnant, representing a ~49‐fold difference. Vineibacter was also more abundant in flow (17%) than in stagnant (6%), corresponding to a ~3‐fold increase. In stagnant conditions, distinct dominant genera emerged, led by Acidovorax (13% in stagnant vs. < 1% in flow; ~37‐fold increase) and Sphingopyxis (11% vs. 2%; ~3‐fold difference). Several genera showed comparable abundances between conditions, including Aestuariivirga (~7%) and Bradyrhizobium (~5%).

The Top 15 genera driving community structure across timepoints were visualised in a stacked bar plot of genus‐level relative abundances (Figure S3). This revealed greater variability among replicates at 3 months, whereas community composition appeared more consistent at 5 and 7 months under both flow and stagnant conditions. This compositional shift from Nitrospira‐dominated profiles under flow to Acidovorax‐ and Sphingopyxis‐dominated profiles under stagnation represents one of the clearest compositional contrasts between the two hydraulic conditions (Figure S3). Temporal shifts also influenced the persistence of key taxa: Nitrospira, the most abundant genus in flow, increased from 3 to 5 months before declining at 7 months. Conversely, Acidovorax, the dominant genus in stagnant conditions, decreased from 3 to 5 months before stabilising at 7 months (Figure 2C). Individual temporal profiles for the Top 15 genera are provided in Figure S4. These profiles show that most flow‐associated genera maintained relatively stable temporal abundance patterns over the experimental period, whereas several stagnant‐associated genera displayed more variable temporal trajectories under stagnant conditions.

3.3. Functional Enrichment and Depletion Under Flow and Stagnation

To identify which functions are enriched or depleted under stagnation, we examined protein expression patterns and performed differential abundance analysis. The protein expression profiles showed clear differences between flow and stagnant conditions. Stagnant biofilms exhibited significantly fewer detected proteins compared to flow (Wilcoxon rank‐sum test, p = 0.002; Figure 3A). Across both conditions, a total of 10,668 unique proteins were identified. Flow biofilms contained 7081 proteins (66.4%) that were unique to flow, whereas stagnant biofilms contained 1201 proteins (11.3%) unique to stagnant. A set of 2386 proteins (22.4%) was detected in both conditions (Figure 3B).

FIGURE 3.

Protein distributions among different water dynamics. (A) The number of proteins identified showing a significantly lower number of proteins in stagnant compared to flow (Wilcoxon rank‐sum test; p = 0.002). (B) Upset plot showing the total proteins identified in flow and stagnant, as well as the unique and shared proteins among the conditions. (C) Volcano plots visualising differentially abundant proteins between stagnant and flow conditions at 3, 5 and 7 months. Proteins enriched in flow (negative log₂ fold change) are shown in blue, while proteins enriched in stagnant conditions (positive log₂ fold change) are shown in red (adjusted p‐value < 0.05). Selected labels indicate representative differentially abundant proteins associated with key functional processes across conditions.

Differential protein abundance analysis revealed distinct temporal patterns across the three sampling months (Figure 3C). In flow biofilms, the number of significantly enriched proteins increased over time, with 210 proteins at 3 months, 728 at 5 months and 1138 at 7 months. In contrast, stagnant biofilms showed limited protein enrichment at 3 months (41 proteins), followed by a peak at 5 months (374 proteins) and a slight decline at 7 months (275 proteins). Overall, flow biofilms exhibited approximately fivefold higher numbers of enriched proteins than stagnant biofilms at 3 months, twofold at 5 months and sevenfold at 7 months. The complete metaproteomic dataset, including protein intensities, annotations and differential abundance statistics, is provided in Supporting Information S2.

The uniquely enriched proteins showed distinct functional patterns shaped by water dynamics (Figure 4). Proteins favoured under flow conditions were predominantly associated with nitrogen removal and contaminant degradation. Key nitrification and denitrification enzymes, including ammonia monooxygenase (AMO), which catalyses the conversion of ammonia to hydroxylamine (NH₃ → NH₂OH) (Liu et al. 2015), nitrate reductase (NAR) and nitrite reductase (NIR), responsible for the first two steps of denitrification (NO₃ ⁻ → NO₂ ⁻ → NO) (Robertson and Groffman 2007), were consistently expressed across all three timepoints. Additional flow‐enriched proteins involved in contaminant transformation, such as LLM class flavin‐dependent oxidoreductase (Pandey et al. 2025), flavin‐dependent trigonelline monooxygenase (Perchat et al. 2018) and carboxymethylenebutenolidase (He et al. 2022), were consistently detected. Proteins associated with adhesion, including fimbrial proteins and Type VI secretion system protein Hcp (Liu et al. 2015; Jønsson et al. ²⁰²³), were also prevalent under continuous flow. Notably, the biofilm‐regulating quorum‐quenching lactonase YtnP (Annapoorani et al. 2012; Liu et al. ²⁰²⁰) became enriched under flow beginning at 5 months and remained stable through 7 months.

FIGURE 4.

Differentially expressed proteins enriched under flow (negative log₂ fold change) or stagnant (positive log₂ fold change) conditions across the 3, 5 and 7 months timepoints. Displayed proteins were selected from statistically significant differentially abundant proteins (adjusted p < 0.05, |log₂FC| ≥ 1) and assigned to broad functional categories using KEGG annotations together with manual curation based on protein descriptions and known functional roles in wastewater biofilm systems: Adhesion, motility and chemotaxis, nitrogen removal, quorum quenching, transport and nutrient scavenging and xenobiotic degradation and detoxification. Point size reflects −log₁₀(adjusted p‐value); colour indicates the condition of enrichment.

In contrast, stagnant conditions favoured proteins linked to motility, transport and nutrient scavenging. These included flagellar motor protein MotB (Haiko and Westerlund‐Wikström ²⁰¹³) and multiple transporters such as the Vitamin B12 transporter BtuB, critical for microbial viability (Pieńko and Trylska 2020). Additional enriched proteins supported solute acquisition, including MOMP‐like family proteins, which mediate uptake of extracellular compounds (Atanu et al. 2013). Although many adhesion‐related proteins were depleted under stagnant conditions, the OmpA‐like domain‐containing protein, a factor associated with adhesion, virulence and antibiotic resistance (Scribano et al. 2024), was uniquely enriched. All significantly enriched or depleted proteins, with their annotations and timepoint‐specific statistics, are listed in Table S3.

KEGG orthologue (KO)–based functional annotation was performed to summarise the major metabolic pathways represented by the identified proteins (Kanehisa and Goto 2000). Six functional categories were detected across conditions and are summarised in Figure S5: nitrogen removal and metabolism, oxidative stress defence, motility and chemotaxis, transport and nutrient scavenging, contaminant degradation and carbon utilisation and energy metabolism. This KEGG‐based summary shows that nitrogen metabolism‐ and xenobiotic degradation‐associated proteins were more represented among flow‐enriched proteins, whereas transport‐ and secretion‐related functions dominated the stagnant‐enriched protein set. Proteins in three categories, nitrogen removal and metabolism, contaminant degradation and carbon utilisation and energy metabolism, showed higher abundance under flow relative to stagnation. Conversely, proteins associated with motility, chemotaxis, transport and nutrient scavenging exhibited higher abundance under stagnation. Oxidative stress defence proteins were detected in both conditions with variable temporal patterns.

3.4. Taxonomic‐Functional Linkages Reveal Ecosystem Implications

The taxonomic origin and temporal dynamics of significantly expressed proteins revealed clear functional differentiation between the two water‐flow regimes (Figure 5; Table S4). Under flow, the microbial community displayed broad functional capacity, with pronounced enrichment of nitrogen‐cycle enzymes. Multiple nitrification proteins were consistently expressed, including AMO and nitrite oxidoreductase (NXR), both dominated by Nitrospira, in agreement with its role as a canonical nitrifier (Daims et al. 2015). AMO was produced by two taxa: Nitrospira, uniquely detected in flow and members of the class Betaproteobacteria, detected in both conditions but showing twofold increase in flow. NXR β‐subunit was largely specific to Nitrospira and uniquely observed under flow. Hydroxylamine oxidoreductase (HAO), a key enzyme that oxidises hydroxylamine to nitrite (Simon and Klotz 2013), was expressed by Nitrospira and Nitrosomonas at 5‐ and 7‐month timepoints in both conditions, though differences between flow and stagnant were not significant (Table S2).

FIGURE 5.

Taxonomic–functional mapping of significantly enriched proteins under the two water‐flow regimes. The heatmap shows the top taxa and their associated proteins across the three developmental timepoints (3, 5 and 7 months) for flow‐enriched proteins (top; n = 20) and stagnant‐enriched proteins (bottom; n = 20), grouped by broad functional categories. Proteins were selected from statistically significant differentially abundant proteins (adjusted p < 0.05, |log₂FC| ≥ 1) and ranked by absolute log₂ fold‐change magnitude, with the Top 20 per condition displayed. Functional categories were assigned using KEGG annotations together with manual curation based on protein descriptions and known functional roles relevant to wastewater biofilm processes. Each row represents a specific taxon–protein pair and columns represent sampling timepoints. Colours indicate scaled protein abundance (z‐score), where higher‐than‐average expression is shown in purple and lower‐than‐average expression in orange, relative to each protein's mean expression across timepoints. Taxonomic assignments were derived from the MAG‐based database using CAT/BAT and GTDB‐Tk classifications. This visualisation illustrates the distinct taxonomic contributors and functional roles favoured under flow versus stagnant conditions.

Consistent with downstream steps of nitrogen removal, several denitrification enzymes were also enriched under flow. NAR, responsible for the first step of denitrification (NO₃ ⁻ → NO₂ ⁻) (Richardson et al. 2001), was primarily expressed by Nitrospira and members of the family Acetobacteraceae, showing a threefold increase at 5 months. NIR, which converts nitrite to nitric oxide, was expressed by Nitrospira, unique to flow and Nitrosomonas, which increased threefold at 5 and 7 months. Nitric oxide reductase (NOR) also appeared exclusively at 5 months and increased further by 7 months. The terminal denitrification enzyme, nitrous oxide reductase (NosZ), mainly derived from Dechloromonas and Bradyrhizobium, was detected in both conditions but showed a fourfold reduction under stagnation.

Nitrogen assimilation proteins exhibited broad taxonomic representation. Glutamine synthetase and glutamate–ammonia ligase, essential for assimilatory ammonium incorporation (Reitzer 2003), were uniquely detected in flow and expressed by the family Parvularculaceae and the genus Mycobacteroides. Flow conditions also favoured diverse xenobiotic‐transformation proteins, including 4‐nitrophenol 2‐monooxygenase from Acidimicrobiales (flow‐specific), carboxymethylenebutenolidase from Ideonella, LLM‐class oxidoreductases from Vineibacter and dehydrogenase/reductase proteins from Ferrovibrio. Most showed ≥ 20‐fold higher abundance under flow, highlighting enhanced contaminant‐degradation capacity. Additional flow‐specific functions included gluconolactonase from Caldilinea and the quorum‐quenching lactonase YtnP from Bradyrhizobium.

In contrast, stagnation displayed a different functional profile, with broader taxonomic diversity contributing chiefly to transport, motility and stress‐response functions. Transport processes were distributed across at least seven taxa, including Sphingopyxis (BtuB vitamin B₁₂ transporter; twofold increase), Sediminibacterium (TonB‐dependent receptor P3; threefold increase), Dongiaceae (oligopeptide and arginine ABC transporters; twofold increase), Vineibacter (ABC permease; unique), Bradyrhizobium (amino acid/amide transporter; threefold increase), Chitinophagaceae (ExbD biopolymer transporter; unique) and Parvularcula (heme transporter BhuA; threefold increase). Motility and chemotaxis proteins also exhibited stagnation‐specific signatures, including flagellar motor proteins from Novosphingobium (unique) and flagellin from Plastoroseomonas (threefold increase). Stress‐response functions were expressed by multiple taxa, such as acid‐stress chaperone HdeA from Rhodospirillaceae (twofold increase), bacteriocins from Planctomycetia (unique) and CsbD‐like proteins from Hyphomicrobium (threefold increase).

Temporal patterns further highlighted these regime‐specific functional traits. Most flow‐associated proteins showed stable enrichment from 3 to 7 months, whereas many stagnant‐associated proteins peaked at 5 months. Together, this taxonomic–functional mapping demonstrates that flow conditions favour functional specialists dominated by Nitrospira‐driven nitrogen cycling, while stagnation promotes functional generalists, with multiple taxa sharing roles in transport, motility and stress adaptation.

4. Discussion

This study investigated how prolonged stagnation influences microbial adaptability and functionality in treated wastewater, compared with continuous flow. Previous stagnation research has primarily focused on the persistence of emerging pathogens (Yi et al. 2019; Chen et al. ²⁰²⁰; Jalili et al. ²⁰²²; Ye et al. ²⁰²²; Lippai et al. ²⁰²⁴). However, we expanded the scope to explore microorganisms that contribute to reclaimed water stability during distribution and overall ecosystem function. Our findings reveal how altered hydraulic conditions reshape microbial structure and function, with stagnation undermining biostability and ecosystem resilience.

4.1. Community Restructuring: Diversity Loss and Taxonomic Shifts Under Stagnation

Our analysis of diversity patterns and taxonomic composition revealed that stagnation fundamentally altered both the structural and functional diversity of microbial communities. At the genomic level, the MGs showed a slight diversity increase in stagnant conditions, albeit with no significant difference compared to the flow conditions; however, the metaproteomic analysis showed the opposite pattern with significantly lower protein profiles in stagnation. This contrast indicates that the genes' presence exemplified by means of DNA does not necessarily lead to protein expression. Specifically, stagnation may maintain taxonomic diversity but reduce functional diversity. This difference reflects the well‐established distinction between genetic potential and expressed function: MGs captures what a community could do, while metaproteomics reveals what it is actively doing under different environmental conditions. This distinction is well documented in environmental microbiology (Van Den Bossche et al. 2021) and metaproteomics therefore provides more direct insight into active microbial function under different hydraulic conditions.

Key taxa encoding nitrogen‐cycling proteins were significantly depleted in stagnation, including Nitrospira, Nitrosomonas and Vineibacter. The enrichment of nitrifying taxa under flow conditions is likely attributed to the continuous delivery of ammonium and dissolved oxygen from the MBR effluent, which supports the growth requirements of obligate aerobic nitrifiers such as Nitrospira and Nitrosomonas (Daims et al. 2015). Continuous flow likely maintains more consistent substrate and oxygen availability, favouring these slow‐growing, specialised taxa. Conversely, stagnant conditions create a closed batch environment in which ammonium and dissolved oxygen may become limited over the seven‐day exchange interval, potentially restricting nitrifier growth and activity. Together, these conditions may contribute to the depletion of Nitrospira and Nitrosomonas observed under stagnation. Under stagnation, proteins associated with motility, transport and nutrient scavenging were also enriched. These transport proteins include systems for uptake of extracellular compounds such as oligopeptides, amino acids and heme (via MOMP‐like proteins and specific ABC transporters), reflecting a metabolic shift from biosynthesis to scavenging of pre‐formed compounds under nutrient‐limited conditions (Atanu et al. 2013). This energy‐conserving strategy enables microbial persistence when resources are scarce under stagnant conditions. Indeed, among the genera enriched during stagnation, Acidovorax was particularly dominant and persistent across time points. Several Acidovorax species are known phytopathogens responsible for bacterial fruit blotch of cucurbits, a globally distributed disease causing major agricultural losses (Webb and Goth 1965; Bahar et al. ²⁰⁰⁹; Adhikari et al. ²⁰¹⁷; Song et al. ²⁰²⁰). The expression of cell surface‐associated proteins by Acidovorax observed in this study aligns with a previous report describing the ability of Acidovorax to undergo phenotypic variation in response to environmental stressors, promoting persistence and adaptation (Shrestha et al. 2013). The consistent detection of Acidovorax in stagnant water underscores the potential risk of phytopathogen dissemination through irrigation systems supplied with stagnant reclaimed water.

4.2. Temporal Progression Reveals Functional Instability Under Prolonged Stagnation

Beyond initial compositional shifts, our temporal analysis across 3, 5 and 7 months revealed different temporal patterns between conditions, although these observations remain exploratory given the limited replication per timepoint (n = 2). Metaproteomic diversity, as visualised by PCoA, showed flow samples clustering more tightly across all timepoints, whereas stagnant samples showed greater temporal separation, suggesting greater temporal divergence under stagnation. At the individual genus level, temporal trends also varied among genera. For example, Nitrospira increased from 3 to 5 months, then returned by 7 months to a similar abundance observed at 3 months. Members of Nitrospira spp. are known to participate in nitrite oxidation and some comammox Nitrospira are additionally capable of complete ammonium oxidation, suggesting that fluctuations in ammonium and nitrite availability may influence their abundance and succession dynamics (Spasov et al. 2020). However, ammonium and nitrite concentrations were not monitored within the reactors during operation, which represents a key limitation of this study and prevents direct attribution of Nitrospira temporal dynamics to nitrogen substrate availability specifically. The drivers of this pattern therefore remain uncertain and substrate availability, community succession and biofilm maturation are all plausible contributing factors. Under stagnant conditions, Acidovorax decreased from 3 to 5 months and then stabilised through 7 months, which may reflect adaptation to low‐oxygen and nutrient‐limited conditions.

The trend of the differentially expressed proteins from the volcano plots showed a continuous increase in flow. This increased expression was observed in proteins like LLM class flavin‐dependent oxidoreductase and carboxymethylenebutenolidase, which are involved in pollutant degradation. The increased expression of these proteins over time may indicate an adaptive response that helps limit the accumulation of contaminants during prolonged operation. In addition, some proteins in flow were not detected in 3 months; they only started to emerge at 5 months and continued, for example, quorum‐quenching enzyme—lactonase YtnP, which might have started to express in response to the biofilm progression over time. In contrast, protein expression under stagnation reached a peak at 5 months before declining by 7 months. While the drivers of this pattern cannot be determined from the current dataset, several explanations are plausible, including community succession, nutrient limitation over the extended stagnation period or a shift towards more dormant physiological states as biofilm communities adapt to resource‐limited conditions. The declining expression of transport proteins such as TonB‐dependent receptor P3 and the vitamin B₁₂ transporter BtuB by 7 months may reflect reduced metabolic investment under prolonged nutrient scarcity, though this remains speculative without direct metabolic or chemical measurements. These observations suggest that prolonged stagnation may progressively alter microbial functional stability, with potential implications for reclaimed water quality during extended storage periods such as irrigation pond retention or prolonged system shutdowns.

4.3. Biostability Implications for Water Quality and the Safe Reuse of Reclaimed Water

These taxonomic and functional shifts have direct implications for the biostability of reclaimed water and its safe use in irrigation systems. In this study, continuous flow preserved all three indicators associated with biostability, protein diversity, nitrogen‐cycling enzyme expression and xenobiotic transformation‐associated proteins, whereas stagnation reduced these across the experimental period. Continuous flow exhibited high biostability, supported by active nitrogen cycling, detoxification capacity and biofilm regulation. Under stagnation, however, loss of biostability was evident as functional microbial communities shifted towards survival‐adapted niches under low‐oxygen and nutrient‐limited conditions. This functional imbalance may have practical implications for irrigation systems, potentially leading to crop disease, yield losses and food safety concerns.

The loss of biostability described above reflects functional destabilisation of the microbial community rather than a direct pathogenic risk assessment. Within the studied bacterial community, concern was primarily limited to the genus‐level enrichment of Acidovorax, which contains phytopathogenic species; however, species‐level identification and virulence characterisation would be required to substantiate this risk further. Nonetheless, our results demonstrate that continuous flow more effectively preserves beneficial microbial functions, including nitrogen‐cycling‐associated protein expression and proteins linked to contaminant transformation. Consequently, maintaining hydraulic flow within reclaimed water systems may help sustain microbial functional integrity and reduce the accumulation of undesirable compounds during storage and distribution.

Overall, the metaproteomic patterns indicate that prolonged stagnation disrupts microbial community balance, suppresses key functional pathways and reduces the biostability of treated wastewater. Maintaining hydraulic movement within reclaimed water systems is therefore critical to preserve microbial functionality and ensure safe reuse for irrigation.

It is important to note that these observations were derived from lab‐scale reactors that only partially mimic real distribution systems. While our controlled system allowed isolation of hydraulic effects, several real‐world variables were not captured. Regarding pipe material, the CDC reactor coupons used in this study were composed of PVC, which is consistent with materials commonly used in water distribution systems; however, other materials such as iron or copper, which promote corrosion‐associated microbial communities, may support different biofilm communities under stagnation (Zhang et al. 2017; Lee et al. ²⁰²¹). Regarding hydrodynamics, shear stresses in the CDC reactors were characterised at low‐to‐moderate levels consistent with drinking water distribution system hydraulics, though larger‐scale turbulent flow regimes and pressure fluctuations in actual networks may alter biofilm dynamics beyond what was captured here (Johnson et al. 2021; Silva et al. ²⁰²⁵). Reactors were operated in the dark to mimic enclosed pipe conditions; however, systems exposed to sunlight, where UV exposure and temperature fluctuations may further influence microbial community dynamics, were not addressed in this study. Additionally, limited biological replication per timepoint (n = 2) necessitated pooling across timepoints for statistical comparisons, which enabled assessment of overall condition effects but precluded detailed temporal statistical analysis. Future pilot‐scale studies incorporating these variables, with greater temporal and seasonal replication and more realistic operational conditions, will be essential for validating and extending these results.

Author Contributions

Shaman Narayanasamy: data curation, formal analysis, software. Pei‐Ying Hong: conceptualization, supervision, funding acquisition, project administration, resources, writing – review and editing. Ma'an Amad: methodology, investigation. Fatimah Almulhim: methodology, conceptualization, investigation, formal analysis, visualization, data curation, writing – original draft, writing – review and editing. Papita Mandal: methodology, investigation. Dalila Bensaddek: methodology, investigation. Changzhi Wang: software, formal analysis, data curation.

Funding

This work was supported by King Abdullah University of Science and Technology (BAS/1/1033‐01‐01).

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: Overview of the experimental workflow.

Figure S2: Quality‐control metrics for LC–MS/MS analyses of flow and stagnant biofilm samples collected after 3, 5 and 7 months of operation.

Figure S3: Relative abundance of the Top 15 microbial genera detected in biofilms grown under flow and stagnant conditions across the three developmental timepoints (3, 5 and 7 months).

Figure S4: Temporal trajectories of selected dominant and condition‐associated microbial genera across biofilm development at 3, 5 and 7 months under flow and stagnant conditions.

Figure S5: KEGG‐based functional profiling of the metaproteome showing condition‐specific metabolic pathway dynamics across biofilm development (3, 5 and 7 months).

Table S1: Quality‐control metrics for all LC–MS/MS runs.

Table S2: Full metaproteomics dataset.

Table S3: Statistically significant proteins across timepoints.

Table S4: The taxonomic‐functional linkage.

Data Availability Statement

Metaproteomics mass spectrometry data have been deposited in the PRIDE repository via the ProteomeXchange Consortium under dataset identifier PXD072096.