Effect-Directed Toxicity Assessment of the Surface Water of Chao Phraya River by Integrating a Panel of Bioluminescent E. coli with ICP-OES and HPLC-DAD

Abstract

This study presents an effect-directed analysis (EDA) combining Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES), High-Performance Liquid Chromatography-Diode Array Detection (HPLC-DAD), and five bioluminescent Escherichia coli bioreporters to map surface water toxicity along a 372-km gradient of the Chao Phraya River, Thailand. Evaluating cytotoxicity, genotoxicity, metabolic disturbance, oxidative stress, and quorum sensing via a new Toxic Index (TI) and Dynamic Time Warping (DTW) kinetic alignment, the panel unmasked complex cellular stress gradients. Although ICP-OES revealed localized mercury spikes (119 μg/L), metals contributed minimally to overall toxicity. Instead, upstream freshwater exhibited severe, multi-pathway stress (Total Toxicity Index, TTI ∼ 2.48) statistically associated with trace bioactive organic fractions isolated via solid-phase extraction. In complex urban matrices, DTW effectively exposed non-monotonic toxicity masking. Conversely, the downstream estuarine TTI peak (3.06) was identified as an isolated, non-chemical osmotic shock response driven by extreme salinity (18,710 ppm total dissolved solids). This integrated framework demonstrates that relying on bulk chemical quantification inherently underestimates latent ecotoxicological hazards, proving the necessity of mechanistic biological diagnostics to accurately evaluate and protect complex aquatic ecosystems.

1. Introduction

River ecosystems, such as Chao Phraya River in Thailand, − face increasing ecological pressure from complex anthropogenic mixtures of heavy metals and organic micropollutants. − Effective monitoring of these dynamic systems is essential, yet traditional chemical techniques (such as ICP-OES and targeted chromatography) primarily provide high-precision mass–concentration data, without accounting for pollutant bioavailability or synergistic mixture interactions. − Furthermore, conventional “lights-off” microbial bioassays (such as Microtox) are widely used for rapid toxicity screening, − often yield only a general toxicity score without explaining the underlying mechanisms of action. − These traditional tools offer no mechanistic insights into how environmental contaminants are actively harming living systems and remain highly susceptible to false-negative reporting in the presence of complex environmental matrix interferences. ,

To overcome this analytical limitation, engineered whole-cell bioluminescent bioreporters have emerged as a highly sensitive diagnostic alternative. ,,− By fusing specific, stress-responsive bacterial promoters to the luxCDABE reporter operon, these genetically engineered microorganisms generate a dynamic “lights-on” bioluminescent signal directly proportional to the activation of distinct intracellular damage pathways. While these bioreporters offer exceptional diagnostic speed and mechanistic clarity, their large-scale, in-field use faces persistent challenges, including strict regulatory frameworks controlling the environmental release of genetically modified organisms (GMOs) and potential signal attenuation caused by environmental matrix extremes (such as high turbidity). Thus, these bioassays are mostly done in the laboratory settings.

Therefore, the primary objective of this study was to establish a comprehensive, effect-directed evaluation framework designed to isolate, map, and diagnose the underlying inducers of surface water toxicity across a 372 km longitudinal profile of the Chao Phraya River. By integrating multi-element ICP-OES and Solid-Phase Extraction (SPE) coupled with HPLC-DAD profiling alongside a high-resolution panel of five engineered Escherichia coli bioreporter strains targeting cytotoxicity (grpE), genotoxicity (recA), membrane disruption (fabA), oxidative stress (katG), and quorum-sensing communication (lasI), we aimed to offer a new solution for integrated river water toxicity assessment. Additionally, pairing targeted chemical fractionation with a novel absolute Toxic Index (TI) scoring system and Dynamic Time Warping (DTW) kinetic sequence alignment, we aim to unravel the nature of environmental contaminants by identifying the specific stress pathways activated, thereby inferring the dominant modes of action in complex water samples.

2. Materials and Methods

2.1. Sampling Design

Water samples were collected across the entire length of the Chao Phraya River, starting from its formation point (upstream, sample 1) down to its spillage point (downstream, sample 9) making it approximately 372 km (Figure ). In total, nine samples were collected in 50 mL polyethylene tubes (in duplicate) from nine provinces (Supporting Information, Table S1). Samples were collected from a depth of approximately 20–30 cm and a distance of 1–2 meters from the shoreline at temples and community areas, reflecting the close link between Thai rivers and daily life and cultural activities. While this near-shore sampling strategy effectively captures localized human exposure at community access points, it must be noted that shoreline waters are inherently more susceptible to local discharge outfalls, bank scouring, and sediment resuspension than the central river mainstream. During transit, the samples were stored in closed containers with ice and refrigerated at 4 °C upon return to the laboratory. All samples were collected on the same day of the study.

1.

Map of sampling sites along the Chao Phraya River and surrounding communities and temples.

2.2. Analytical Techniques

2.2.1. Chemical Analysis

2.2.1.1. ICP-OES

The determination of heavy metals in water samples was conducted using Inductively Coupled Plasma with Optical Emission Spectroscopy (ICP-OES), specifically with a Shimadzu Model ICPE-9820 Series. This technique is recognized for its high sensitivity, wide dynamic range, and efficiency in simultaneous multi-element analyses, ensuring high reproducibility and precision. , Prior to analysis, the samples were filtered through a 0.45-micron filter to remove particle interference. Calibration standards were prepared from known heavy metal stock solutions (arsenic, cadmium, chromium, copper, mercury, and nickel), and concentrations in the filtered water samples were subsequently determined via their emission spectra against established calibration curves (R ² > 0.998).

2.2.1.2. Physicochemical Parameters

Electrical conductivity (EC, μS/cm), total dissolved solids (TDS, ppm), and pH were measured for all water samples using a standardized multiparameter meter (Eutech PC 450, ThermoScientific-Eutech, Thailand). The instrument calculated TDS dynamically from the baseline EC values utilizing an internal nonlinear NaCl conversion algorithm. The measurements of EC and TDS provide a direct representation of the sample ionic strength, establishing a physicochemical context for interpreting potential bioassay influence.

2.2.1.3. HPLC-DAD

Chromatographic separation was performed using a Shimadzu Nexera LC-40 system equipped with a C18 reverse-phase column. A Diode Array Detector (SPD-M40) was used to detect a broad spectrum of organic classes, including aliphatic, aromatic, and phenolic compounds. , The total area under the curve (AUC) and specific peak areas were integrated to serve as indices of the total organic package for subsequent correlation with the bioreporter responses.

2.2.1.4. Sample Preparation

Surface water samples (30 mL) were acidified to pH 3.0 using a 0.1% formic acid solution to stabilize organic analytes and optimize the recovery of acidic compounds. Extraction was performed via SPE using Oasis HLB cartridges (150 mg, 6 cc, Waters Corp.), which contained a universal polymeric reverse-phase sorbent. The cartridges were preconditioned with 5 mL of HPLC-grade methanol and equilibrated with 5 mL of acidified ultrapure water (pH 3). The samples were filtered through a 0.45 μm syringe filter and loaded at a controlled flow rate of 2 mL/min to ensure efficient mass transfer. Following a wash step with 5% methanol in ultrapure water to remove inorganic salts, the cartridges were dried under vacuum for 10 min. The analytes were then eluted with 1 mL of methanol, resulting in a 30-fold concentration of the original river water matrix, which is a critical step to enhance the detection sensitivity for trace organic contaminants potentially present at low environmental concentrations. All extracts were filtered through 0.45 μm PTFE membranes before analysis.

2.2.1.5. Instrument Parameters

The separation of organic compounds was performed using a C18 reverse-phase analytical column (150 mm × 4.6 mm, 5 μm particle size) maintained at 35 °C. The mobile phase consisted of (A) ultrapure water with 0.1% formic acid and (B) HPLC-grade acetonitrile with 0.1% formic acid. Linear gradient elution was selected at a flow rate of 1.0 mL/min, increasing from 5% B to 95% B over 25 min, followed by a 5 min column wash and equilibration period (Supporting Information, Table S2). Organic compounds were detected using an SPD-M40 Diode Array Detector (DAD) with five distinct extracted ion chromatogram channels monitored at 200, 210, 230, 254, and 270 nm. Data acquisition and peak integration were performed using the Shimadzu LabSolutions software (version 5.99). The total area under the curve (AUC) for each channel was calculated as an index of the cumulative organic package for further correlation with the biological effects of the bioassays.

2.2.2. Biological Assay

2.2.2.1. Bacterial Strains

This study utilized a panel of five genetically modified E. coli strains, each harboring a plasmid-borne fusion of a specific stress-responsive promoter to the luxCDABE operon. This design enables these bacteria to produce light as a reporter of distinct metabolic or cellular stress pathways. The bioluminescence mechanism involves luciferase-catalyzed oxidation of reduced flavin mononucleotide (FMNH₂) and a long-chain fatty aldehyde, yielding FMN, water, a fatty acid, and light emission (490 nm). The luxCDABE operon encodes the necessary luciferase subunits (lux A, B) and fatty acid reductase complex (lux C, D, E) for light production. The specific promoter-reporter gene fusions for each strain are listed in Table . The bacterial strains were provided by Prof. Shimshon Belkin (Hebrew University, Jerusalem, Israel).

1. Bioluminescent E. coli Reporter Strains and Targeted Pathways.

Strain	Plasmid/host	Promoter	Strain sensitivity	Inducer of bioluminescence	Ref.
TV1061	pGrpELux/RFM443	grpE	Heat shock (general stress-cytotoxicity)	Ethanol
DPD2794	pRecALux/RFM443	recA	SOS (genotoxicity)	Doxorubicin
DPD2544	pFabALux/W3110	fabA	Fatty acid availabilty (general stress)SOS	Phenol, ethanol, hydrogen peroxide
DPD2511	pKatGLux/RFM443	katG	Protein damage	Hydrogen peroxide
K802NR	Psb1075	Iasl	Quorum sensing	Nalidixic acid

The panel comprised strains targeting the main cellular stress responses. Metabolic changes and general stress (cytotoxicity) were assessed using the strain TV1061, which incorporates the grpE promoter. − This strain is sensitive to various protein-damaging compounds and features a tolC mutation that enhances the detection of hydrophobic molecules. Strain DPD2794, containing a recA promoter fusion, was used to detect genotoxicity, specifically activating the DNA repair SOS system in response to DNA damage. ,, Doxorubicin was used as a positive control for genotoxicity. Fatty acid biosynthesis disruption, indicative of metabolic imbalance, was monitored by strain DPD2544, which utilizes the fabA promoter. Oxidative stress was detected by strain DPD2511, containing the katG promoter. This strain responds with increased luminescence to peroxides, hydrogen peroxide, and redox-cycling agents. Lastly, strain K802NR was used to evaluate quorum sensing activity. This strain carries a lasI promoter fusion, which responds to long-chain acyl-homoserine lactones (AHLs), which are key signaling molecules in Gram-negative bacterial communication. ,− Their sensitivity to QS-modulating chemicals can be affected by their influence on bacterial metabolism. The positive controls included ethanol for grpE, phenol for fabA, hydrogen peroxide for katG, and nalidixic acid for lasI. − Double-distilled water (DDW) was used as a negative control.

2.2.2.2. Growth Protocol and Measurements of Bioluminescence Kinetics

Briefly, the five E. coli bioreporter strains were maintained through long-term storage at −80°C in 50% glycerol and propagated on antibiotic-supplemented Petri dishes (kanamycin at 50 μg/mL for K802NR; ampicillin at 100 μg/mL for TV1061, DPD2794, DPD2544, and DPD2511). Prior to bioassays, bacterial cultures were grown in Luria Bertani broth with appropriate antibiotics for 12–24 h, followed by a 2–3 h secondary culture without antibiotics, ensuring cultures reached exponential growth phase (OD600 between 0.185–0.215). Bioluminescence assays were conducted in white opaque 96-well microtiter plates, where 100 μL of bacterial culture was exposed to 10 μL of undiluted (raw sample) or serially diluted (10-fold and 100-fold) water samples, with all exposures performed in triplicate to ensure the reliability of the data. Bioluminescence curves were created using a Varioskan Flash luminometer (Thermo Fisher Scientific). The recording of bioluminescence kinetics was done every 5 min for a total of 6h for each experiment.

2.3. Statistical Analysis

The primary response metric was defined as the Response Ratio (RR), calculated as the ratio between the maximum sample bioluminescence and double distilled water control (DDW).

$$\mathrm{R}\mathrm{R}=\frac{\max \mathrm{b}\mathrm{i}\mathrm{o}\mathrm{l}\mathrm{u}\min \mathrm{e}\mathrm{s}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}\mathrm{s}\mathrm{a}\mathrm{m}\mathrm{p}\mathrm{l}\mathrm{e}}{\max \mathrm{b}\mathrm{i}\mathrm{o}\mathrm{l}\mathrm{u}\min \mathrm{e}\mathrm{s}\mathrm{c}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}\mathrm{c}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}}$$

A significant result was considered when an RR exceeded 1.5 (light induction) or when RR dropped below 0.5 (light inhibition). To mathematically integrate both induction and inhibition without signal loss into a single metric, RR values were converted into an absolute Toxic Index (TI) score. This score measures the absolute deviation from the baseline (RR = 1), normalized to the 0.5 significance interval

$$\mathrm{T}\mathrm{I}=\frac{|\mathrm{R}\mathrm{R}-1|}{0.5}$$

Under this transformation, both a 50% signal increase and a 50% signal decrease yield a TI of 1.0, ensuring that acute toxicity and cellular stress are weighed equally. To account for toxicity masking in complex matrices, the maximum TI score across all tested dilutions (undiluted, 1:10, 1:100) was selected for each strain. By doing this, any score ≥1.0 immediately flags biological significance, regardless of whether it was caused by stress induction or inhibition.

$${\mathrm{T}\mathrm{I}}_{\max }=\max (\mathrm{T}{\mathrm{I}}_{\mathrm{u}\mathrm{n}\mathrm{d}\mathrm{i}\mathrm{l}\mathrm{u}\mathrm{t}\mathrm{e}\mathrm{d}},\mathrm{T}{\mathrm{I}}_{1:10},\mathrm{T}{\mathrm{I}}_{1:100})$$

Finally, the Total Toxicity Index (TTI) for each site was calculated using the Root Mean Square (RMS) of the five TI_max scores. This quadratic mean prevents the mathematical loss of highly specific pathway activations, ensuring that severe localized stress or cytotoxicity is properly represented in the overall site assessment.

$$\mathrm{T}\mathrm{T}\mathrm{I}=\sqrt{\frac{\sum _{i=1}^5{(\mathrm{T}{\mathrm{l}}_i)}^2}{5}}$$

Similarity analysis was conducted between the environmental samples and double-distilled water (DDW) using dynamic time warping (DTW).

$$\mathrm{D}\mathrm{T}\mathrm{W}\mathrm{d}\mathrm{i}\mathrm{s}\tan \mathrm{c}\mathrm{e}=\sum ^{\mathrm{s}\mathrm{e}\mathrm{q}\mathrm{u}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}}\mathrm{e}\mathrm{r}\mathrm{r}\mathrm{o}\mathrm{r}$$

The DTW algorithm was used for pattern recognition in the bioluminescence time-series data generated by bacteria exposed to environmental samples, with the objective of understanding the overall shape and structure of the toxicity fingerprint rather than a strict, element-by-element effect.

Pearson Correlation Analysis (R) was performed between the organic fraction concentration (HPLC peak height) and the TI_max of the bioreporters. An R value of <0.3 signifies a weak or no correlation, 0.3 < R < 0.7 signifies a moderate correlation, and R > 0.7 signifies a strong correlation.

$$R=\frac{\sum _{i=1}^n(x_i-\mathrm{x̅})(y_i-\mathrm{y̅})}{\sqrt{\sum _{i=1}^n{(x_i-\mathrm{x̅})}^2}\sqrt{\sum _{i=1}^n{(y_i-\mathrm{y̅})}^2}}$$

where n is the total number of data points (sample size), x _i is the individual values of the x variable, y _i is the individual values of the y variable, x̅ is the sample mean (average) of all x values, and y̅ is the sample mean (average) of all y values.

3. Results and Discussion

3.1. Sampling Locations

The Chao Phraya River represents a heterogeneous aquatic environment influenced by urban discharge, industrial activities, and downstream estuarine mixing. The geographic distribution of the sampling sites is illustrated in Figure , and the detailed site positions are labeled as samples 1–9 along the longitudinal gradient.

Samples 1 and 2 represent the freshwater source headwaters characterized by direct exposure to agricultural runoff, municipal wastewater outfalls, and localized upstream industrial inputs. As the river cuts through the midstream transitional zones (Samples 3–6), it integrates continuous domestic and industrial loading alongside natural hydrological mixing processes that alters contaminant transportation, speciation, and environmental outcomes. − Reaching the lower metropolitan and tidal sections (Samples 7–8), the river flows through dense urban centers, receiving heavy stormwater runoff and intense river vessel transport inputs. Finally, Sample 9 sits directly at the river mouth, capturing the dynamic brackish mixing zone where freshwater meets seawater. This downstream estuarine interface significantly influences contaminant dynamics through dilution, particle aggregation, flocculation, and sediment complexation, while simultaneously introducing a distinct chemical matrix transition. −

3.2. Physicochemical Characteristics

The baseline physicochemical parameters provide essential environmental context for interpreting both target chemical concentrations and whole-cell biological stress kinetics. Among the measured variables, electrical conductivity (EC) and total dissolved solids (TDS) exhibited the most pronounced variations along the longitudinal gradient (Figure , and Supporting Information, Table S3).

2.

pH and TDS across the nine sampling sites along the Chao Phraya River. To support inferences regarding osmotic stress, TDS values were dynamically derived from direct instrumental measurements of EC.

The upstream and midstream sites (samples 1–6) consistently maintained low TDS values ranging from 99.25 to 221.9 ppm, alongside corresponding EC levels (188.6–422 μS/cm), characteristic of tropical freshwater systems. , The lower urban transition zone (samples 7 and 8) exhibited a moderate rise in conductivity (869.8–3230 μS/cm, TDS 457.6–1660 ppm), indicating cumulative urban discharges before reaching the marine mixing zone. Across the entire gradient, pH remained moderately stable (pH range: 7.5–7.8), providing a compatible baseline for basic microbial viability and standard contaminant partitioning (Supporting Information, Table S3).

The downstream estuarine site (Sample 9) displayed an intense TDS value of 18710 ppm. While TDS alone can represent both ionic and non-ionic dissolved matter, this measurement was dynamically derived from the EC reading of 36,300 μS/cm. This corresponding EC data provides the necessary quantitative confirmation that the elevated TDS at Sample 9 is determined by an intense influx of conductive marine salts, directly establishing the presence of severe ionic strength and high osmotic pressure at the river mouth.

3.3. ICP-OES Metal Distribution

ICP-OES analysis revealed variability in dissolved metal concentrations across the nine sampling locations, indicating different contamination along the river (Figure and Supporting Information, Table S4). Mercury (Hg), nickel (Ni), and arsenic (As) were the principal metals detected, while cadmium, chromium, and copper were either not present or below the detection limits at most sites.

3.

Heavy metal concentrations were measured using ICP-OES (SD < 1%). Left axis concentration represents Hg, right axis concentration represents the other metals.

Arsenic and Nickel were present, however, their dissolved concentrations remained at trace levels across the river. Ni fluctuated between 1.77 μg/L (sample 3) and 3.43 μg/L (sample 6). Trace levels of As were found only in two sites, registering 1.7 μg/L at sample 2 and 0.137 μg/L at sample 5.

In contrast to the trace baseline of the other heavy metals, Hg appeared to be the principal heavy metal contaminant. Highly elevated concentrations of dissolved mercury were detected in the lower urbanized sites, peaking at 119 μg/L at sample 7 (Wat Yai Sawang Arom) and 114 μg/L at sample 8 (Wat Kalayanamit Pier). Additionally, a concentration of 75.8 μg/L was recorded directly at the upstream source (sample 1). Such dangerous, localized spikes in an open river channel are highly unusual. They strongly suggest intense, site-specific anthropogenic inputs, such as agricultural chemical runoff upstream, and concentrated urban drainage, heavy river transport activities, and localized industrial point-source discharges within the densely populated metropolitan sections downstream.

3.4. Individual Reporter Responses

To isolate the biological impacts of these water samples, the five engineered E. coli strains were exposed to the raw water samples, and their 10-fold (1:10) and 100-fold (1:100) dilutions. By tracking the maximum RR relative to DDW water control (Figure , and Supporting information, Figures S1–S5) the panel mapped specific cellular damage pathways along the gradient (Figure ). Additionally, the TI_max index was used to calculate a linear regression (R ²) across the river to determine the toxicity gradient, excluding sample 9 because of its extreme estuarine salinity, which induces an osmotic physical shock that separates the biological response from the underlying chemical toxicity gradient.

4.

Integrated mechanistic toxicity fingerprints from multi-bioreporter responses.

3.4.1. Heat Shock Bioreporter TV1061, Promoter grpE

Strain TV1061, bearing the grpE heat-shock promoter, serves as a sensor for protein damage and general cytotoxicity. Previous characterization of TV1061 has shown it to be responsive to a broad range of heavy metals, including Cd, Cu, and Hg, as well as organic compounds. The maximum Toxic Index (TI_max) analysis revealed a strong, statistically significant longitudinal gradient across the freshwater samples (Samples 1–8, R ² = 0.917, p < 0.001). However, the cytotoxic load did not increase toward the urban center; rather, it peaked heavily at the upstream source (Sample 1, TI = 2.71) and steadily decreased as the river flowed downstream toward the lower urban zones (Sample 8, TI = 1.80). This distinct downward trend indicates that the river receives its most highly concentrated load of sublethal protein-damaging agents at its upstream formation points, which is subsequently diluted by increasing river volume and hydrological mixing as it travels toward the estuary.

Serial dilution of the samples unmasked classic toxicity masking effects, particularly in the lower urban reaches (Samples 6–8). In these matrices, 1:10 and 1:100 dilutions frequently yielded higher bioluminescent signals than the raw, undiluted water. This non-monotonic behavior indicates that concentrated matrix interferences or acute toxicants initially suppressed bacterial metabolism, a physiological limitation that was only relieved by dilution. Conversely, at the estuarine mouth (Sample 9), the cytotoxic signal was entirely suppressed (TI = 1.11). Rather than indicating a pristine environment, this sudden signal drop could also be attributed to severe osmotic shock induced by the high-salinity marine matrix (36,300 μS/cm), which completely quenches the bioluminescent signal of the freshwater E. coli.

3.4.2. Genotoxicity Bioreporter DPD2794, Promoter recA

The recA promoter in strain DPD2794 responds to DNA damage (SOS response). Following the TI_max regression model, the genotoxic profile exhibited an exceptionally strict linear gradient across the river (Samples 1–8, R ² = 0.975, p < 0.0001). Similar to the cytotoxic response, the highest genotoxic hazard was registered at the upstream source (Sample 1, TI = 2.39) and gradually attenuated downstream (Sample 8, TI = 1.50). Critically, this upstream genotoxic activation remained highly persistent and did not decrease across serial dilutions (1:10 and 1:100), demonstrating the presence of highly potent genotoxins active at trace levels. The magnitude of this upstream activation far exceeds the toxic potential predicted by the low heavy metal concentrations measured in Samples 1 and 2 via ICP-OES. This difference supports the hypothesis that the upstream samples are heavily impacted by unmeasured organic genotoxins, such as agricultural pesticides, herbicides, or pharmaceutical residues, that undergo natural attenuation or dilution as the river approaches the marine mixing zone.

In the highly impacted upstream samples (Samples 1 and 2), the genotoxic signal remained remarkably persistent across serial dilutions (1:10 and 1:100). This lack of dilution-driven signal decay demonstrates the presence of highly potent, hydrophilic genotoxins that remain biologically active and capable of triggering DNA repair mechanisms even at trace environmental concentrations. In contrast, the estuarine station (Sample 9) completely quenched the genotoxic response. The intense ionic strength of the seawater mixing zone physically stopped the biological translation required for the recA SOS repair mechanism to produce light, highlighting the limitations of applying freshwater bioreporters directly to marine matrices without accounting for osmotic stress.

3.4.3. Fatty Acid Biosynthesis Inhibition Bioreporter DPD2544, Promoter fabA

Strain DPD2544 detects disruptions in membrane fatty acid biosynthesis. Biological tracking revealed a sustained membrane stress response that closely reflected the genotoxic gradient, demonstrating a statistically significant linear decrease from the upstream samples to the urban sites (Samples 1–8, R ² = 0.779, p < 0.01). The activation peaked upstream (Sample 1, TI = 1.95) and steadily declined (Sample 8, TI = 1.51). The longitudinal correlation between fabA and recA activations indicates the co-occurrence of complex agricultural or domestic discharges capable of simultaneously compromising bacterial membrane integrity and damaging DNA structures. ,

The membrane stress pathway exhibited distinct non-monotonic dilution behaviors in the lower urban sections of the river. Raw, undiluted samples regularly presented low signals that increased upon 1:10 dilution, unmasking the presence of concentrated lipophilic organic pollutants, such as commercial surfactants or industrial solvents, whose initial concentrations severely compromised bacterial membrane integrity and suppressed baseline metabolism. Upon reaching the estuary (Sample 9), the fabA response dropped entirely to baseline (TI = 1). The saline matrix effectively incapacitated the specific membrane lipid biosynthesis pathways required for signal induction, confirming that the high-salt environment exerts a dominant physiological control on the fabA reporter system.

3.4.4. Oxidative Stress Bioreporter DPD2511, Promoter katG

The oxidative stress bioreporter DPD2511 displayed no trend profile compared to that of the other strains (Figures and S4). While metals such as Cr, Cu, and Hg are known to generate Reactive Oxygen Species (ROS) via Fenton-like reactions, , the low katG activation suggests that oxidative stress was not the dominant mechanism of toxicity in these samples. Alternatively, the complex organic matrix of the river water may have scavenged free radicals, reducing the ROS signal detected by the bacteria. This supports other studies showing that environmental matrices can alter the bioavailability and oxidative potential of metals. ,

While the freshwater reaches showed minimal oxidative stress across all dilutions, the estuarine site (Sample 9) triggered a substantial, isolated biological event, with the katG TI increase rapidly to 6.53. This extreme activation was not induced by chemical oxidants, but rather by the severe osmotic shock of the 36,300 μS/cm matrix, which forces the bacteria into rapid osmoregulation, generating a massive intracellular burst of ROS. When the sample was subjected to serial dilution, the katG signal decreased proportionally with the declining salinity but still maintained an activation level greater than 50% above the control. This dilution behavior confirms that osmotic stress, rather than trace chemical pollutants, was the primary inducer of this specific oxidative burst.

3.4.5. Quorum Sensing Bioreporter K802NR, Promoter Iasl

Strain K802NR monitors inter-cellular communication (quorum sensing). Across all sampling sites and serial dilutions, the quorum-sensing reporter maintained a uniformly low baseline activation. The serial dilutions did not unmask any latent endocrine-disrupting signals, confirming that specific autoinducer-mimicking compounds were either absent from the river or remained well below the biological detection threshold. At the estuarine mouth (Sample 9), the extreme salinity further constrained any potential density-dependent signaling. Because quorum sensing relies heavily on optimal bacterial growth and membrane stability to detect signaling molecules, the osmotic stress at Sample 9 resulted in slight baseline inhibition rather than induction, confirming an absence of communication disruptors even in the transition zone.

3.5. DTW Kinetic Fingerprints

Dynamic Time Warping (DTW) analysis was applied to the time-resolved bioluminescence responses to generate kinetic fingerprints describing the mode of action of contaminants across sampling sites. Unlike single end point measurements, DTW captures differences in response timing, signal intensity, and recovery dynamics, enabling the discrimination between stress pathways and mixture effects (Figure ). This time-series alignment approach has been widely used for pattern recognition in environmental bioassays and improves the interpretation of complex biological signals. The DTW-derived fingerprints revealed clear differences between the sites. As seen with the strain DPD2794 (genotoxicity), most samples exhibited kinetic patterns characterized by delayed peak emission and prolonged recovery, consistent with sustained cellular stress and exposure to reactive contaminants in the cytoplasm. In contrast, the estuarine sample suppressed kinetic responses across most reporters, reflecting the dilution of bioavailable toxicants and physiological inhibition of freshwater-adapted bacteria under high-salinity conditions. These results are similar to those of previous studies demonstrating that bioreporter kinetics are sensitive indicators of environmental stress and contaminant exposure. ,

Cluster analysis of the DTW further demonstrated that samples with similar mechanistic profiles were grouped together regardless of environmental proximity, indicating that contaminant composition, rather than location alone, induced biological responses. Sites showing concurrent activation of recA and fabA formed a distinct cluster associated with genotoxic and membrane-active compounds, while samples with minimal pathway activation clustered separately, suggesting lower bioavailable toxicity or matrix induced suppression. Effect-based monitoring studies have similarly shown that kinetic bioassay patterns can differentiate between contaminant classes and mixture interactions. , The DTW approach also enabled differentiation between temporary and persistent toxic effects. Rapid peak recovery kinetics suggested short-term metabolic disruption (represented by a low DTW value), whereas prolonged luminescence suppression indicated sustained exposure to the inhibitory mixtures (high DTW value).

The DTW analysis separated distinct modes of action. Genotoxic exposures (recA) consistently mapped to a unique cluster characterized by high DTW values (>80) paired with moderate RR values. This kinetic fingerprint represents a delayed, slow-rising luminescent emission that stops recovering within the 6-hour window, showing the slow, continuous accumulation of DNA damage and the sustained activation of the bacterial SOS mechanism. In contrast, standard cytotoxic responses (grpE) and membrane stress (fabA) mapped to tightly bound, low-DTW clusters (30–50), reflecting rapid, sharp luminescent spikes that quickly clear as the cells adapt or stabilize. This time-resolved alignment confirms that whole-cell kinetic tracking provides excellent diagnostic resolution, distinguishing between persistent structural damage and temporary metabolic stress within complex environmental matrices.

3.6. Toxicity Gradient

To evaluate the overall biological hazard across the sampling sites and combining all pathway-specific responses, the integrated Total Toxicity Index (TTI) was calculated using the Root Mean Square of the maximum Toxic Index (TI_max) scores. This approach unmasked a complex spatial gradient of mechanistic toxicity along the Chao Phraya River. The upstream freshwater source locations (Samples 1 and 2) exhibited a highly concerning, multi-pathway toxic package, with elevated TTI scores (2.47 and 2.48, respectively). These sites demonstrated widespread biological impacts, induced by severe concurrent spikes in the katG (oxidative stress, TI ∼3.6), grpE (cytotoxicity, TI ∼2.8), and recA (genotoxicity, TI ∼2.4) pathways. This broad activation profile indicates exposure to a highly reactive, complex mixture of anthropogenic organic pollutants capable of causing true multi-mechanistic cellular damage at the formation points of the river. As the river transitions through the midstream and urban zones (Samples 3–8), the TTI scores moderated (ranging from 1.67 to 2.11) but maintained sustained biological significance across multiple pathways, reflecting continuous, non-lethal organic loading. , When interpreting this spatial toxicity profile, particularly the pronounced, multi-pathway toxicity observed at upstream Samples 1 and 2, the physical sampling strategy must be considered. Because samples were collected close to the shoreline, they are highly sensitive to localized non-point discharges, bank scouring, and shallow sediment resuspension. Therefore, while these elevated upstream TTI scores accurately represent the intense toxicological burden present in the shoreline zones where community interaction is highest, they may not accurately represent the heavily diluted water quality of the river’s central mainstream. This highlights a limitation in generalizability across the full hydrodynamic cross-section of the river.

Notably, the downstream estuarine site at the river mouth (Sample 9) registered the highest overall TTI (3.06). However, the mechanistic resolution of the bioreporter panel revealed that this peak was almost exclusively induced by an extreme, singular spike in the katG oxidative stress pathway (TI = 6.53), while the other stress pathways remained near baseline levels (TI 0.85–1.13, Figure ). This specific biological fingerprint directly confirms the physicochemical data. Rather than indicating increased chemical pollution, this isolated oxidative burst is characteristic of severe osmotic shock. The extremely high EC and salinity at the estuarine mixing zone impose intense physiological stress on the freshwater-adapted E. coli bioreporters, triggering high ROS defense mechanisms. ,, Furthermore, across all sampling sites, the IasI quorum sensing pathway was the only bioreporter to consistently remain below the significance threshold (TI < 1.0). This indicates that while chemical and osmotic stressors were dominant, specific endocrine-disruptor-like signals were either absent or present at concentrations too low to impair bacterial communication networks during the broader toxic package. Ultimately, the RMS-based TTI demonstrates that relying on arithmetic averages or single-end point assays fundamentally underestimates the highest ecological hazards present in complex river matrices.

5.

Total toxicity index gradient and normalized biological response across sites.

3.7. HPLC Organic Compounds Analysis

To determine whether the total organic concentration predicts toxicity, the correlation between the Total Area Under the Curve (AUC), representing all organic matter across the 30 min run, and the five stress panels were evaluated. Statistical analysis demonstrated that the Total AUC is an inadequate predictor of cellular stress. Of the 20 tested correlations, only two reached statistical significance, both occurring at 230 nm. Notably, the 210 nm Total AUC, which contains the vast majority of organic mass, showed no correlation with any biological response. This lack of correlation supports the approach of focusing on specific chromatographic peaks rather than bulk concentration, as the observed biological effects strongly correlate with minor, specific chemical constituents hidden within the overall mixture.

By fractionating the samples into specific peaks, the class of compounds responsible for cellular damage could be isolated (Table ). The polarity and molecular structure were assumed based on chromatographic retention time and UV absorption profiles. In this reverse-phase setup, early eluting peaks represent highly polar compounds, while late-eluting peaks indicate lipophilic constituents.

2. Chemical Family Characterization of Major HPLC Peaks.

Peak and RT window	Polarity	Predicted chemical families	Observed bioreporter effect	Ref.
Peak A (2.7–2.8 min)	Highly Polar	Organic acids, sugars, and inorganic ions (nitrates).	No toxicity or stress induction
Peak B (4.6–4.9 min)	Mid-Polar	Small phenolics, chlorophenols, and polar solvents.	Inhibition of grpE, fabA, recA
Peak C (15.4 min)	Mid-Low Polarity	Quinones, oxy-PAHs, substituted benzenes.	Induction of katG
Peak D (26.8 min)	Highly Lipophilic	PAHs, POPs (PCBs), and heavy fatty acids.	Induction of IasI

The compound class assignments listed in Table are strictly presumptive and tentative, derived solely from reverse-phase hydrophobic retention dynamics and characteristic UV chromophore absorption bands. In complex environmental surface waters, multi-component co-elution is common, and highly toxic organic structures lacking distinct UV chromophores would remain completely undetected by a DAD setup. Therefore, these chemical family classifications serve as exploratory guides and require future validation via High-Resolution Mass Spectrometry coupled with non-target screening workflows to absolutely identify the molecular structures. To evaluate the relationship between these specific organic fractions and the whole-cell responses, Pearson correlation coefficients (R) were calculated between individual peak heights and the bioreporters TI_max index across three selective channels (230, 254, and 270 nm), as illustrated in the heatmap (Figure ).

6.

Correlation heatmap illustrating the relationship between targeted HPLC organic fractions (Peaks A–D) and bacterial bioreporter TI_max. The color gradient represents the Pearson correlation coefficient (R), with dark red indicating strong positive correlation, while dark blue indicating strong negative correlation. *denote statistical significance levels: * p < 0.05, ** p < 0.01, *** p < 0.001.

To isolate and identify the specific organic inducers of biological stress within the complex river matrix, an Effect-Directed Analysis (EDA) framework was applied. Pearson correlation coefficients (R) were calculated between the specific chemical mass of fractionated organics, quantified via the Area Under the Curve (AUC) of isolated HPLC peaks (A–D), and the corresponding whole-cell stress responses. To successfully account for the non-monotonic toxicity masking observed in the raw matrices, the correlation model utilized the maximum Toxic Index (TI_max) scores for each bioreporter strain. This mathematical integration ensures that the true, uninhibited biological effect is accurately plotted against the targeted chemical dose.

The correlation matrix revealed highly distinct, fraction-specific toxicological mechanisms. Most notably, the mid-polar organic fraction eluting as Peak B (quantified at 270 nm) exhibited intense, statistically significant negative correlations with genotoxicity (recA, R = −0.94), cytotoxicity (grpE, R = −0.93), and membrane disruption (fabA, R = −0.88). Within the context of bioluminescent bioreporters, this severe negative correlation does not indicate an absence of toxicity. Rather, it mathematically confirms acute cellular inhibition. As the concentration of Peak B increases, the biological signal is severely decreased, indicating that this specific mid-polar fraction (likely comprising concentrated industrial phenolics or aromatic solvents) rapidly penetrates the cellular membrane and induces widespread translational arrest before the bacteria can produce a luminescent stress response.

Conversely, strong positive correlations successfully unmasked the specific chemical fractions actively driving sublethal cellular damage. The mid-polar fraction eluting as Peak C (254–270 nm) exhibited a strong statistical correlation with oxidative stress in the freshwater samples (katG pathway, R = 0.86 at 270 nm). This suggests that the compounds within Peak C actively undergo intracellular redox cycling, generating a heavy load of ROS. Furthermore, the late-eluting lipophilic fraction (Peak D) demonstrated broad-spectrum positive correlations, acting as a secondary correlated factor of oxidative stress (R = 0.61 at 230 nm) while strongly influencing quorum-sensing pathways (lasI, R = 0.72 at 270 nm). The lipophilic nature of Peak D suggests the presence of heavier, highly permeable organic pollutants capable of prolonged intracellular residence times and membrane accumulation. Overall, this correlation framework demonstrates that the severe, multi-pathway toxicity observed at the upstream source points of the Chao Phraya River is not statistically linked to the trace dissolved heavy metals detected via ICP-OES, but rather correlates strongly with highly potent, complex mixtures of mid-polar and lipophilic organic pollutants.

Critically, statistical correlation does not establish direct biological causation. While the Pearson matrix alignments reveal strong associations between isolated chromatographic peaks (A–D) and pathway-specific bioreporter responses, these attributions remain strictly tentative. In complex environmental surface waters, multi-component co-elution and highly toxic, non-chromophoric organic compounds can easily confound UV-DAD signals. Definitive causal verification of these specific biological drivers requires future non-target high-resolution mass spectrometry (such as LC-MS/MS or GC-MS) paired with bioassay validation using authentic analytical standards.

3.8. Matrix Effects and Toxicity Masking

Serial dilution experiments revealed significant matrix-driven influences on the biological responses observed. These effects are commonly associated with complex contaminant mixtures, elevated salinity, or high organic content, which constrain microbial metabolism and obscure pathway-specific responses. , These dilution-dependent patterns are shown in Figure by their similarity of mechanistic toxicity. Each sample was classified across reporter strains using two features, the maximum TI (TI_max), which represents the response magnitude, and the dynamic time warping distance (DTW), which represents the deviation of the response pattern from the water control. For each sample–strain pair, the responses were categorized as dose-dependent, toxicity masking, negligible response and osmotic effect.

7.

Integrated Toxicity Profile mapping Effect Magnitude, Dilution Behavior, and Kinetic Deviation across the Chao Phraya River. The heatmap visually synthesizes three distinct dimensions of environmental toxicity across the longitudinal river gradient (Station 1: upstream source to Station 9: estuarine boundary). Color intensity indicates the Effect Magnitude, quantified by the Maximum Toxic Impact (TI_max) achieved across a three-point serial dilution gradient (raw, 1:10, 1:100). Categorical letters define the kinetic Dilution Behavior: D (Dose-Dependent) indicates standard unmasked toxicity; M (Masking Effect) indicates that acute matrix interferences suppressed the raw signal, requiring serial dilution to unmask the real biological effect; N indicates a negligible biological response; and O+/O– denote high osmotic induction or complete quenching driven by the saline matrix rather than trace chemicals. Finally, statistical asterisks mark Kinetic Deviation based on Dynamic Time Warping (DTW) scores against a pure water control. A single asterisk (*) indicates a significant kinetic deviation (DTW ≥ 60), while a double asterisk (**) highlights extreme kinetic disruption (DTW ≥ 100), such as acute translational arrest or sudden osmoregulatory shock.

This heatmap reveals a highly dynamic, nonlinear toxicological landscape that fundamentally challenges standard water quality assumptions. By combining the severity of the hazard (color), the matrix behavior (letters), and the kinetic disruption of the cells (asterisks), a clear distinction can be made. The upstream (samples 1–4) represent the most severe and bioavailable chemical hazard zone. This is explicitly demonstrated by the intense genotoxic (DPD2794) and oxidative (DPD2511) stress pathways, which not only exhibit high TI_max scores but are heavily flagged with extreme kinetic deviations. This indicates that upstream pollutants are not merely triggering standard stress responses, they are strongly altering the timing and duration of the bacteria’s DNA repair and redox kinetics. As the river transitions into the densely populated urban midstream (Stations 5–8), the toxicity profile shifts significantly from standard dose-dependent dynamics to widespread matrix masking (M). In this urban basin, the raw environmental water contains complex mixtures of organic pollutants, domestic discharge, and surfactants that are acutely toxic. Instead of inducing a bright luminescent response, these concentrated mixtures induce rapid metabolic suppression and translational arrest, artificially quenching the light output. The true magnitude of this urban hazard is only captured because the TI_max metric isolates the highest biological deviation observed after dilution relieves this physiological limitation. Finally, the estuarine boundary (Station 9) presents as a complete biological difference. The dark red katG cell combined with an (O+) designation confirms that the sudden intrusion of seawater physically overrides all trace chemical toxicity. This extreme salinity physically shocks the freshwater-adapted E. coli, induce a high, isolated osmotic burst that completely dissociates the biological response from the underlying chemical pollution gradient.

The significant takeaway from this visualization is that environmental matrix interference actively conceals severe ecological hazards in urbanized rivers. If regulatory bodies rely exclusively on standard, undiluted ecotoxicity assays, the highly contaminated urban basin will mathematically present as safe due to widespread metabolic quenching (toxicity masking). This heatmap shows that utilizing serial dilution to calculate the TI_max, combined with kinetic deviation tracking, is an absolute necessity to prevent the major underestimation of biological risk in complex aquatic systems.

3.9. Contribution of Metals and Organic Contaminants to Observed Toxicity

While ICP-OES confirmed the presence of dissolved metals along the Chao Phraya River, determining their direct contribution to the observed biological effects is complex. A weak correlation was observed between the combined metal load and the biological response (RR), indicating that metals alone did not explain the mechanistic toxicity patterns. One major factor is the bioavailability of metals. Chemical analyses quantify total dissolved concentrations, whereas bioluminescent bioreporters primarily respond to the bioavailable fraction that can enter bacterial cells and interact with intracellular targets. Complexation with dissolved organic matter and other ligands likely reduces metal uptake, rendering part of the measured metal load biologically inactive under assay conditions. , Mixture interactions also play an important role in this regard. Metal combinations often exhibit non-additive behavior, producing synergistic or antagonistic effects depending on their composition. For example, copper–cobalt mixtures can enhance toxicity, whereas copper–zinc interactions may suppress toxicity. In the present study, the complex mixture of metals may have resulted in antagonistic interactions that diminished the oxidative stress responses typically associated with individual metal exposure. In contrast, the strong activation of genotoxic (recA) and membrane/metabolic stress (fabA) pathways suggests the influence of unmeasured organic toxicants. These responses are commonly associated with pharmaceuticals, pesticides, and other anthropogenic micropollutants frequently detected in wastewater-impacted river systems but not captured by ICP-OES metal analyses. , Together, these findings indicate that although metals were present, they were unlikely to be the primary drivers of the observed toxicity. Instead, the observed biological effects were more consistent with complex contaminant mixtures dictated by organic stressors, supporting the integration of chemical analysis with mechanistic bioassays to identify the dominant toxicity sources.

3.10. Estuarine Influence on Physicochemical Conditions and Bioreporter Responses

The downstream estuarine site (Sample 9) registered a distinct biological signature, showing a complete suppression of cytotoxicity (grpE), genotoxicity (recA), and membrane stress (fabA), contrasted against an intense spike in the katG oxidative stress pathway (TI = 6.53). As established by the physicochemical baseline, Sample 9 has a high total dissolved solids concentration (18,710 ppm TDS) and high electrical conductivity, confirming an intense marine mixing zone. Placing freshwater-adapted E. coli bioreporters directly into this high-salinity matrix imposes a severe osmotic shock. , To survive, the bacterial cells are forced to execute rapid osmoregulatory adjustments, generating a severe burst of intracellular ROS as a side product of osmotic stress. − Therefore, the high activation of the katG promoter at Sample 9 does not reflect chemical oxidation from riverine pollutants. Instead, it represents an isolated, non-chemical osmotic shock response directly caused by the high ionic strength of the estuarine matrix. At the same time, this intense salinity exerts a strong physiological suppression on the other four pathways, quenching their light output and creating a false interpretation of low chemical toxicity. This dual effect emphasizes that comparisons between freshwater and estuarine zones must be made with extreme caution, such that high salinity can simultaneously alter chemical bioavailability, dilute riverine pollutants, and introduce major non-chemical biological stress pathways.

3.11. Proposed Mechanistic Evaluation Framework

To contextualize the toxicological dynamics observed across the Chao Phraya River, Table merges established literature regarding the specific cellular targets of the bioreporter panel with the empirical kinetic and chromatographic data obtained in this study. In complex environmental matrices, attributing whole-cell stress to singular compounds is difficult without advanced mass spectrometry. However, by associating the specific promoter pathways with known chemical categories and directly mapping them against the correlated HPLC organic fractions, we can establish a strong, effect-directed profile of the water matrix. This approach avoids definitive chemical identification while categorizing the supposed modes of action induced by the isolated polar, mid-polar, and lipophilic fractions.

3. Connection between Toxicant Classes, Cellular Targets, and Responses of Bioluminescent E. coli Reporter Strains.

Bioreporter strain (pathway)	Chemical category (specific toxicants)	Mechanism of action	Observed matrix behavior & HPLC correlation (current study)	Ref.
TV1061 (grpE - Heat Shock/Cytotoxicity)	Toxic Metals (As, Cd, Cu, Hg), Protein-damaging organics	Protein denaturation, intracellular folding stress, and ROS generation.	High dose-dependent activation upstream. Negatively correlates with mid-polar Peak B (R = −0.93***) and lipophilic Peak D (R = −0.69*), consistent with acute translational arrest caused by concentrated solvent or metal fractions.	,,
DPD2794 (recA - Genotoxicity)	Genotoxins/Organics (e.g., Quinolones, Mitomycin C, PAHs)	Structural DNA lesions, inhibition of DNA replication, induction of SOS repair.	Persistent dose-dependent activation at headwaters. Strongly suppressed by mid-polar Peak B (R = −0.94***), suggesting acute matrix toxicity masks the genotoxic signal in the urban reaches.	,
DPD2544 (fabA - Membrane Disruption)	Membrane Disruptors (e.g., Phenols, Chlorophenols, Surfactants)	Disruption of membrane lipid bilayer integrity, inhibition of FabA/FabB enzymes.	Elevated membrane stress upstream. Strong negative correlation with mid-polar Peak B (R = −0.88**), aligning with the mechanism of highly membrane-permeable organics that rapidly compromise cell envelope stability.	,,,
DPD2511 (katG - Oxidative Stress)	Oxidative Agents (e.g., Peroxides, Paraquat, redox-cycling organics)	Disturbance of cellular redox balance, generation of intracellular superoxide/hydrogen peroxide.	Low freshwater baseline. Strong positive correlation with mid-polar Peak C (R = 0.86**), suggesting the presence of organic fractions capable of intracellular redox cycling. Massive osmotic induction at the estuary.	,,,
K802NR (lasI - Quorum Sensing)	Quorum Sensing Inhibitors (e.g., Furanones, high conc. metals)	Interference with LuxI/LuxR autoinducer synthesis or receptor binding.	Flat baseline across the freshwater gradient. Positive correlation with late-eluting lipophilic Peak D (R = 0.72*), indicating putative interactions between heavy, hydrophobic fractions and membrane-mediated signaling.	,

This integrated table aims to bridge the gap between established toxicological mechanisms and the empirical data gathered from the Chao Phraya River. The most critical takeaway from this synthesis is that the toxicity of complex environmental matrices can be effectively decoded without immediately relying on exhaustive chemical quantification. By supporting the known cellular targets of the bioreporter panel with the isolated chromatographic fractions, the table reveals a highly specific, effect-directed profile of the river. For instance, the strong negative correlations associated with the mid-polar Peak B explicitly map to the widespread metabolic suppression observed in the genotoxic, cytotoxic, and membrane-stress pathways. Conversely, the oxidative stress pathway uniquely correlates with the mid-polar Peak C, while the quorum-sensing pathway highlights known interactions with late-eluting lipophilic compounds in Peak D.

The implications of utilizing this specific multi-pathway bioreporter panel are useful for modern environmental monitoring. Traditional whole-organism or single-end point bioassays come with a huge limitation, that is they indicate whether a water sample is toxic but fundamentally cannot explain the underlying mode of action. By adopting this carefully selected panel of genetically transformed E. coli, regulatory bodies can decouple overlapping environmental hazards. The panel successfully differentiates between physical matrix differences, such as the high osmotic shock observed at the estuarine boundary, and true trace chemical toxicity. Furthermore, mapping the phenomenon of toxicity masking within the urban reaches confirms that utilizing a comprehensive, promoter-specific biosensor array is essential to avoid underestimating the ecological risks posed by concentrated organic mixtures.

Moving forward, the next step is to transition from this effect-directed categorization to definitive chemical identification. While the current methodology successfully isolates the specific mid-polar and lipophilic fractions driving the cellular damage, the compound class assignments remain speculative based on UV absorption and retention time gradients. To confirm the exact molecular drivers of this toxicity, future research must subject the isolated Peak B, C, and D fractions to high-resolution mass spectrometry, such as LC-MS/MS or GC-MS. This advanced analytical confirmation will conclusively identify the individual environmental pollutants responsible for the widespread metabolic suppression and targeted cellular stress, thereby completing the transition from biological hazard detection to targeted chemical regulation.

4. Conclusion

This study established a comprehensive monitoring framework for the Chao Phraya River, revealing a highly dynamic and nonlinear toxicological landscape induced by specific organic fractions rather than bulk chemical loading. While chemical analysis confirmed the presence of heavy metals (such as As, Ni, and Hg) across the river, the mechanistic E. coli bioreporter panel demonstrated that the most severe, bioavailable genotoxic (recA) and membrane-disrupting (fabA) hazards actually originate at the upstream sites. As the river transitions into the densely urbanized midstream, concentrated anthropogenic mixtures induce severe toxicity masking and acute metabolic suppression, effectively hiding the true ecological risk. Finally, at the estuarine boundary, extreme salinity masks the trace chemical toxicity, inducing an isolated osmotic oxidative burst (katG). A fundamental finding of this assessment is the statistical disconnect between bulk organic mass and actual biological risk. The Total Area Under the Curve (AUC) at 210 nm, which represents the vast majority of the organic carbon present in the river, showed no correlation with any biological stress response. Instead, an Effect-Directed Analysis (EDA) framework demonstrated that toxicity was strongly associated with trace bioactive packages identified through chromatographic fractionation. Peak B (∼4.6 min) was putatively linked to acute cytotoxicity, correlating with the rapid metabolic inhibition and widespread toxicity masking observed in the urban reaches. Conversely, Peak C (∼15.4 min) and Peak D (∼26.8 min) acted as highly specific sublethal disruptors of redox balance and quorum sensing (lasI), respectively, suggesting that trace pollutants actively interfere with bacterial communication pathways. Ultimately, these findings prove that relying on single-end point chemical monitoring or undiluted ecotoxicity assays significantly underestimates environmental risk, as complex matrix interferences actively conceal severe hazards. To make a shift from a strictly chemical concentration-based approach to a biological effect-based paradigm, this study proposes a four-tiered mechanistic evaluation framework. This analytical depth was used to unmask latent toxicity utilizing multi-dilution kinetic profiling and Maximum Toxic Impact (TI_max) scoring. Future environmental monitoring must expand beyond surface sampling to include deep-water profiles and sediment samples, as complex bioactive organics frequently accumulates into these matrices. Finally, the specific bioactive fractions isolated in this study must be conclusively identified using high-resolution gas chromatography-mass spectrometry (GC-MS), and osmoregulatory models must be established to ensure the reliable application of freshwater bioreporters across complex estuarine zones.

All data can be found through the manuscript and Supporting Material.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsestwater.6c00629.

• Data analysis (XLSX)

• Sample collection points across the river; HPLC elution gradient; heavy metals content in water samples across the river; physicochemical parameters; grpE, recA, fabA, katG, lasI, responses; chromatogram of the HPLC for all water samples, at 254 nm (PDF)

C.T.: Conceptualization, methodology, software, validation, formal analysis, investigation, data curation, writingoriginal draft preparation, visualization; W.P.: Methodology, software, formal analysis, data curation; C.P.: Resources, writingreview and editing, supervision, project administration, funding acquisition; R.S.M.: Conceptualization, methodology, resources, writingreview and editing, visualization, supervision, project administration, funding acquisition; Y.P.: Conceptualization, methodology, validation, investigation, resources, writingoriginal draft preparation, writingreview and editing, visualization, supervision, project administration, funding acquisition.

This study was supported by Matching Fund between Thammasat University Research Fund and Ben-Gurion University, Contract No MF 8/2566; Chulabhorn International College of Medicine Contract No G 7/2567; “Innovative sediment management framework for a SUstainNable DANube black SEa system (SUNDANSE)” co-funded by the European Union, grant number 81203411.

The authors declare no competing financial interest.