Rapid Solar Photoinactivation of Influenza Virus and Phi6 in Colored Surface Water

Abstract

Indirect transmission through aquatic environments is critical to interspecies transmission of influenza, but knowledge of sunlight inactivation of the virus, or any enveloped virus, in water is lacking. This study characterizes the photoinactivation of two enveloped viruses (Phi6 and influenza A virus, IAV), and a nonenveloped virus (MS2) in clear and colored surface waters using simulated sunlight. We measured light-screening corrected decay rate constants ( $\mathrm{k̂}$ ) using infectivity assays. $\mathrm{k̂}$ values were greater, especially for IAV, in colored surface water versus clear water. $\mathrm{k̂}$ values were generally greatest for IAV, followed by Phi6, then MS2, suggesting greater susceptibility of enveloped viruses to sunlight. Most $\mathrm{k̂}$ values for IAV and Phi6 did not differ with UV irradiance variations in colored surface water, indicating dominance of indirect, photochemically produced reactive intermediate–mediated inactivation pathways. Additional experiments with Phi6 as a representative enveloped virus suggest the importance of hydroxyl radicals, indicate adsorption of NOM promotes photoinactivation, and sublethal damage of the virus may occur. For IAV, modeled time for 99% inactivation in a well-mixed 1 m deep wetland water column is 1.2 h, compared to 1.8 h in clear water. These findings, particularly those for IAV, can inform human and animal health protection strategies.

Introduction

Enveloped viruses represent important emerging pathogens, and many have pandemic potential. For some enveloped viruses, like influenza, water may serve an important role in their transmission. For example, previous work documents the importance of environmental waters as a transmission pathway of avian influenza virus among wild and domestic bird populations. − ^, Understanding the fate of avian influenza virus and similar viruses in the environment is key for pandemic risk assessment and control, but critical knowledge gaps exist regarding the photoinactivation, or light-mediated inactivation, of enveloped viruses in natural waters. ,

Sunlight has been studied extensively for its ability to photoinactivate nonenveloped viruses in water, but photoinactivation processes are complex and not well understood. − Unlike UVC light (200–280 nm), which is largely absent from natural sunlight and contributes mainly direct photodamage, sunlight can contribute to both direct and indirect photoinactivation of viruses. UVB light (280–320 nm) can directly damage nucleic acids, lipids, and proteins, while UVB, UVA (320–400 nm), and visible light (400–700 nm) can contribute to indirect damage of nucleic acids, lipids, and proteins by generating photochemically produced reactive intermediates (PPRIs), including reactive oxygen species (ROS), which can oxidize viral components. , Key PPRIs which may contribute to viral photoinactivation include singlet oxygen, hydroxyl radicals, triplet state natural organic matter (NOM), and hydrogen peroxide. ,, In environmental waters, PPRIs may be produced via reactions between photons and constituents of the microorganism (endogenously) or through interactions between NOM (exogenously). , Additionally, particulates and dissolved NOM in water may absorb light, sorb to viruses, or shield viruses, affecting their susceptibility to photoinactivation. ,

Most prior work on viral photoinactivation has focused on nonenveloped viruses or UVC-based disinfection systems, leaving critical gaps in our understanding of enveloped virus photoinactivation by natural sunlight. − , Recent systematic reviews identified very limited data on enveloped virus decay in sunlit waters, , with only four published studies investigating sunlight-induced photoinactivation of enveloped viruses in liquid matrices akin to natural waters. Garver et al. considered solar photoinactivation of hematopoietic necrosis virus (an enveloped fish virus) suspended in seawater, Sagripanti et al. exposed vaccinia virus in phosphate-buffered saline (PBS) to simulated sunlight, and Gomes et al. exposed bacteriophage Phi6 to simulated sunlight in PBS. None of the three studies reported explicitly considered the effect of NOM on inactivation or investigated other photoinactivation mechanisms. The fourth study, Anderson and Boehm, provided photoinactivation rate constants for Phi6 and murine hepatitis virus in PBS with simulated sunlight, but did not explore photoinactivation mechanisms or the impact of photosensitizers.

Genome and protein composition of viruses may influence their susceptibility to solar photoinactivation, but the key determinants remain unclear, especially for enveloped viruses. In previous studies with both sunlight and UVC, viruses with longer genomes were found to be more susceptible to photoinactivation because they present an increased number of photoreactive pyrimidine nucleotides (uracil/thymine (U/T) and cytosine (C)) and photoreactive adjacent nucleotides (CC, CU, UC, UU, CT, TC, and TT). ,,,− Studies investigating both nonenveloped and enveloped virus photoinactivation by UVC light found that single-stranded genomes can be more susceptible to photoinactivation versus double-stranded genomes, and RNA genomes are less susceptible than DNA genomes, possibly due to different dimer product formation. , While viral genome damage is likely important for the photoinactivation of enveloped viruses by sunlight, most of these insights come from studies that used nonenveloped viruses or focus solely on UVC exposure. Similarly, there are little data on the damage to proteins and lipids of enveloped viruses from sunlight. Available evidence from studies with sunlight inactivation of nonenveloped viruses suggests that damage to viral proteins may cause photoinactivation in addition to genome damage, especially for liquid matrices containing PPRIs from NOM-sensitized photoprocesses. ,,−

Viral envelope lipid composition also may contribute to viral photoinactivation by sunlight, but the topic is underexplored. Direct observations of lipid damage of enveloped viruses from sunlight exposure are not available, but available evidence obtained using lipids, and viral lipids exposed to UVC light can inform this work. In previous studies with UVC light, the decay of enveloped viruses was similar to that of nonenveloped viruses, indicating direct lipid damage may not drive photoinactivation. , In contrast, indirect damage from oxidation processes has been repeatedly found to be an important damaging factor to lipids. ,− Indirect damage may be the result of lipid hydroperoxides product formation or oxidation that gives rise to isomeric fatty acids; damages can change membrane properties like fluidity and polarity. The specific lipid makeup of envelope membranes, such as an increased proportion of unsaturated fatty acids in the lipid membrane or increased number of plasmalogens (a class of vinyl ether-containing phospholipids), has been found to increase virus vulnerability to oxidative damage. ,,, Thus, viral envelope composition, which depends on the virus host cell, may influence environmental persistence. −

The focus of the present work is to address fundamental gaps in our understanding of the mechanisms and time scales associated with enveloped virus photoinactivation from sunlight in natural waters. Specifically, we measured solar photoinactivation rate constants of influenza A virus (IAV), Phi6, and MS2 and assessed the relative importance of direct photodamage and PPRI-mediated processes through experiments with clear water and colored surface water. By comparing photoinactivation rate constants of three viruses (enveloped viruses IAV and Phi6, and nonenveloped MS2 as a control) with differing intrinsic characteristics we also aim to identify key determinants of solar photoinactivation susceptibility. We provide novel data on solar photoinactivation of Phi6 in colored surface water and solar photoinactivation of IAV in any water matrix. Filling these knowledge gaps is essential to improving our understanding of how long viruses persist in water and how they contribute to transmission risks. The findings will inform preventive measures such as water safety guidelines and public health strategies, and support predictive modeling for emerging viral threats.

Methods

Overview

Experiments were performed using a solar simulator to investigate the photoinactivation of three viruses: IAV, Phi6, and MS2 (Table provides virus properties). The first set of experiments quantified virus photoinactivation rate constants in phosphate-buffered saline (PBS) and natural colored surface water, using 50% long-pass cutoff filters at 280, 305, and 320 nm. Those experiments provide novel rate constants and document the susceptibility of viruses to UVA and UVB photons. The second set of experiments aimed to identify the mechanisms of enveloped virus solar photoinactivation in colored surface water; these experiments mainly used Phi6 as a model enveloped virus.

1. Viral Structural and Genomic Properties .

property	IAV	Phi6	MS2
host	mammalian cell	bacterial cell	bacterial cell
viral structure	enveloped	enveloped	nonenveloped
viral size	80–120 nm	85 nm	26 nm
genome structure	8 segments	3 segments	1 segment
genome type	ssRNA	dsRNA	ssRNA
genome length	13,158 nucleotides	13,385 nucleotides	3569 nucleotides
count of adjacent pyrimidine bases (CC, CU, UC, and UU)	2,370	6,158	855

^aIAV is influenza A virus.

Virus Propagation and Quantification

IAV (A/California/07/2009­(H1N1), kindly provided by Dr. Jeffrey Glenn (Stanford University, Stanford; identification confirmed by sequencing IAV genome, see Supporting Information, SI Text S1) was propagated in Madin Darby canine kidney (MDCK) cells (CCL-34, ATCC, Manassas, VA) and quantified through a 50% Tissue Culture Infectious Dose (TCID₅₀) assay as described previously. Phi6 (NBRC 105899, identity confirmed via whole genome sequencing, see Text S1) was propagated in Pseudomonas syringae (P. syringae, ATCC 21781) and was quantified through a double-layered plaque assay. P. syringae was cultured overnight at 30 °C in nutrient broth prior to use in experiments. MS2 (DMS No. 13767) was propagated in Escherichia coli (E. coli, ATCC 700891) and was quantified through a double-layered plaque assay. E. coli was cultured in tryptic soy both (Becton, Dickinson and Company; Franklin Lakes, NJ) supplemented with ampicillin sodium salt (Sigma; St. Louis, MO) and streptomycin sulfate (Sigma; St. Louis, MO) for approximately 3 h the day of the experiments at 37 °C, until the culture reached log-phase growth. Virus propagation, purification, and quantification methods are detailed in the SI, Text S2.

Colored Surface Water Collection

Colored surface water was collected in April 2024, from a marsh in Coyote Hills Regional Park in Fremont, California (lat 37.55419, long −122.08834). Roughly 20 L of colored surface water were collected in a carboy, sterilized prior to sample collection with 10% bleach and triple-rinsed with deionized (DI) water. Upon collection, the carboy was triple-rinsed with natural water. After collection, the water was passed through a 0.125 mm metal mesh sieve (W.S. Tyler, Inc.; OH) to remove large debris, then was passed through a 0.2 μm pore-sized PES filter (VWR International, LLC; Radnor, PA) to remove particles and microorganisms. After filtration, the water was stored in dark conditions in opaque containers at 4 °C until use in experiments between April 2024 and June 2025. Additional characteristics of the natural water were measured including the absorbance (measured May 2024, July 2024, and April 2025; averaged measurements from May 2024 and July 2024 were used in fluence calculations), pH (measured Sept. 2025 and Jan. 2026), salinity (measured April 2024), dissolved organic carbon (measured July 2024 and Jan 2026), nitrate (measured July 2024 and Jan 2026), metals content (elemental analysis, measured March 2025). Steady-state concentrations of hydroxyl radicals were measured under experimental sunlight exposure conditions (measured June 2024). Key parameters are summarized in the results; methods are included in the Supporting Information (SI, Text S3 and Figure S1). While storage of the water for an extended time is not ideal, it is necessary to complete the complex experiments described herein. Absorbance of the water measured in May 2024, July 2024, and April 2025 did not change (Figure S2), and other parameters were stable over time suggesting limited changes in properties of the water during storage. Additionally, experiments with MS2 and phi6 throughout the study duration (May 2024–April 2025) showed identical decay kinetics (Figure S2), suggesting limited changes in water characteristics relevant to the study. As the water was filter-sterilized, and stored at 4 °C, there was likely limited microbial activity to affect water chemistry, but it cannot be ruled out.

Solar Simulator Experimental Conditions

Experiments were conducted using a solar simulator (Atlas Suntest CPS+; Linsengericht-Altenhaßlau, Germany) equipped with a xenon lamp and quartz filter to create a collimated light spectrum that mimics solar radiation. The solar simulator intensity was set to 250 W/m² (integrated from 280 to 700 nm) and additional 50% long-pass cutoff filters (referred to in the text as “cutoff filters”) were used (280, 305, and 320 nm, N-WG-280, Schott N-WG-305, and Schott N-WG-320 filters, respectively; Edmund Optics; Barrington, NJ) to modulate the intensity of UVB and UVA wavelengths. Solar simulator irradiance is similar to that in Palo Alto, California (latitude of 37.43°N), in mornings (08:00–10:00) and evenings (17:00–19:00) in the spring and fall. Light intensity inside the solar simulator was measured five times after startup for each 50% long-pass cutoff filter type using a SpectriLight ILT 950 fiber optic spectroradiometer (International Light; Peabody, MA) and averaged, while natural sunlight was measured outside a single time. Inside the solar simulator, we ran a continuous circulating water cooler set to 15 °C for all experiments except for anoxic experiments, when the water cooler was set to 10 °C to achieve similar water temperatures given the different experimental setups.

Experimental Setup

Autoclaved 50 mL beakers with cylindrical geometry (radius of 19 mm) were used in all experiments. Each beaker was equipped with a sterilized plastic-coated magnetic stir bar and its bottom and sides were covered in black electrical tape to reduce light reflection into the beaker. Detailed experimental times, trials, and number of data points (n) are shown in Table and are described in the methods section. Varying experimental times, trials, and number of data points were a result of logistical constraints. At minimum, two trials per experimental condition were conducted. Experimental setup images are shown in Figure S3. Mean starting concentrations of viruses in all experiments were 2 × 10⁴ TCID₅₀/mL for IAV, 3 × 10⁷ PFU/mL for Phi6, and 8 × 10⁷ PFU/mL for MS2. In PBS, experiments with MS2 and Phi6 were carried out at the same time. There were negligible interaction effects between MS2 and Phi6 in PBS. IAV experiments in PBS were separate. In colored surface water, experiments were separate for all viruses.

2. Sampling Information for All Experimental Conditions Tested .

^a n = number of data points and trials = number of experimental trials.

PBS and Colored Surface Water Experiments

Depending on the experiment, 20 mL of either PBS or colored surface water, each containing viruses at the concentrations previously described, were added to a beaker, a 50% long-pass cutoff filter was placed atop the beaker, and the beaker was exposed to light in the solar simulator. The total sunlight exposure time was up to 8 h, and samples were withdrawn at time 0 and 8 times after, for a total of 9 samples. Experiments were repeated between two and five times for each virus, matrix, and 50% long-pass cutoff filter combination. Experimental data for Phi6 and MS2 in PBS were previously published, and all other experimental data (IAV in PBS and colored surface water, MS2 and Phi6 in colored surface water) are novel.

Mechanistic Experiments

PPRI Experiments

Experimental treatments investigated potential solar photoinactivation pathways involving various PPRIs using Phi6 as a model enveloped virus. A model enveloped virus was used as it was not feasible to complete all mechanistic experiments with IAV due to the expense and time associated with the viral infectivity assays. The 305 nm 50% long-pass cutoff filter was used to reduce the impact of UVB on virus solar photoinactivation during the experiments. We measured solar photoinactivation rate constants for the following solutions: (1) 50% colored surface water and 50% D₂O (Fisher Scientific; Fair Lawn, NJ) compared to 50% colored surface water and 50%H₂O, (2) 100% colored surface water with a final concentration of 250 mM isopropyl alcohol (MP Biomedicals, LLC; Solon, OH), (3) 100% colored surface water with a final concentration of 10 μM 2,4,6 trimethylphenol (Fisher Scientific; Fair Lawn, NJ), and (4) 100% colored surface water with no dissolved oxygen (mean dissolved oxygen concentrations = 0.82 mg/L, SD = 1.4 mg/L; details in Text S3). Conditions (1), (2), (3), and (4) were intended to allow insight into the roles of singlet oxygen, hydroxyl radical, triplet state NOM, and oxygen-mediated formation and quenching of PPRIs in the bulk phase, respectively, on enveloped virus solar photoinactivation rate constants. Preliminary experiments were conducted to determine isopropyl alcohol and 2,4,6 trimethylphenol concentrations that would not cause viral decay in dark controls (see Text S4). Experimental treatments were exposed to up to 8 h of sunlight in the solar simulator, with samples taken at 4 time points. Experiments were repeated between two to four times for each experimental treatment.

Impact of Dissolved Natural Organic Carbon Concentration

We evaluated solar photoinactivation first-order decay rate constants corrected for light screening in solutions containing different concentrations of NOM. Experiments were carried out using IAV and Phi6 using the 305 nm 50% long-pass cutoff filter to reduce direct photoinactivation. We added Phi6 and IAV to three different solutions of colored surface water, diluted with with H₂O (molecular-grade DI water; Gibco; Grand Island, NY): (1) 50% colored surface water and 50%H₂O, (2) 75% colored surface water and 25%H₂O, and (3) 90% colored surface water and 10%H₂O. The resultant solutions had 12.74, 19.11, and 22.93 mg/L dissolved organic carbon (DOC), respectively, based on DOC concentration of 25.48 mg/L DOC in the collected water. Experiments were repeated between three and five times for each virus and matrix combination.

Adsorption Experiments

To investigate the impact of NOM sorption to Phi6, we pretreated the virus prior to experiments. Phi6 was added to colored surface water and incubated for 1 h and 40 min in darkness or in sunlight with the 305 nm 50% long-pass cutoff filter to promote adsorption of NOM to the virus (dark incubation) and adsorption and sublethal photodamage (sunlit incubation). After the incubation period, Phi6 was isolated and purified from the colored surface water solution using an Amicon Ultra centrifugal filter unit (100,000 Da cutoff; Merck Millipore Ltd.; Carrigtwohill Co. Cork, Ireland) at 14,000 g for 5 min in batches until all of the matrix was processed. Phi6 from the filter retentate was then used, without propagation or modification, to inoculate beakers containing PBS. The absorbance of the PBS + filter retentate was measured to confirm no NOM was transferred to the PBS (Figure S4). The resultant solution was then exposed to sunlight in the solar simulator with either a 280 or a 305 nm 50% long-pass cutoff filter placed atop the beaker. Solutions were exposed to 4 h of sunlight, with samples taken every hour. The precise experimental conditions are presented in Table . In total, adsorption experiments were carried under the following 4 conditions: (1) dark incubation of Phi6 in colored surface water, then Phi6 isolated and added to PBS and exposed to sunlight with the 280 nm 50% long-pass cutoff filter, (2) sunlight incubation of Phi6 in colored surface water with the 305 nm 50% long-pass cutoff filter in colored surface water then PBS with the 280 nm 50% long-pass cutoff filter, (3) dark incubation in colored surface water then PBS with the 305 nm 50% long-pass cutoff filter, or (4) sunlight incubation with the 305 nm 50% long-pass cutoff filter in colored surface water then PBS with the 305 nm 50% long-pass cutoff filter. Experiments were repeated three times for each of the four conditions.

3. Adsorption Experiment Conditions for Phi6 .

	Step 1			Step 2
condition number	matrix	sunlight or dark	50% longpass cutoff Filter	matrix	sunlight or dark	50% longpass cutoff filter
1	colored surface water	dark	none	PBS	sunlight	280 nm
2	colored surface water	dark	none	PBS	sunlight	305 nm
3	colored surface water	sunlight	305 nm	PBS	sunlight	280 nm
4	colored surface water	sunlight	305 nm	PBS	sunlight	305 nm

^aStep 1 represents the incubation of virus in colored surface water. Step 2 represents the light exposure of Phi6 in PBS after Step 1.

Dark Controls

In addition to positive and negative quantification assay controls, three dark control experiments were performed for each experimental condition. In these experiments, the setup was equivalent to those previously described with the addition of a tinfoil cap on the beaker to block light. For anoxic experiments, the tinfoil cap was not able to be added within the sealed system, so the dark experiments were performed in the same system setup as the sunlight experiments, but without turning on the solar simulator.

Data Analysis

Data were analyzed using R (R: A Language for Statistical Computing, version 4.2.3; R Foundation for Statistical Computing, Vienna, Austria).

For IAV, all scored wells without restriction were input into a TCID₅₀ calculator, which used the Spearman-Kärber method to estimate viral concentration. Phi6 and MS2 plates outside of the countable range of plaques (10–300 PFU) were excluded from the data analysis.

To enable comparison of the decay of viral infectivity across experiments run in matrices with different absorbances and under different sunlight conditions, the first-order decay rate constant was corrected for light screening. Depth-averaged UVA and UVB fluence (MJ/m²) was calculated by multiplying the depth-averaged UVA + UVB light intensity, ⟨I⟩ _z (MW/m²), by time (seconds), which is appropriate for a well-mixed reactor like ours. Depth-averaged light intensity is equal to

$${⟨I⟩}_z=\sum _{\lambda }{⟨I_{\lambda }⟩}_z=\sum _{\lambda }{I}_{\lambda ,0}{S}_{\lambda }=\sum _{\lambda }{I}_{\lambda ,0}\frac{1-{10}^{-{\alpha }_{\lambda }z}}{2.303{\alpha }_{\lambda }z}$$ 1

where ${⟨I_{\lambda }⟩}_z$ (MW/m²) is the depth-averaged intensity at a given wavelength λ (nm) for an experimental solution of depth z (m), I _λ,0 is the light intensity at the surface of the experimental solution at a given wavelength (MW/m²), and S _λ (unitless) is the light-screening factor. S _λ is calculated using the absorbance coefficient at a given wavelength (α_λ, m^–1). Depth changed with each time point as samples (900 μL per time point) were withdrawn from the reactor and was calculated considering the geometry of the beaker. Wavelengths 280–400 nm (UVB and UVA light screening correction) were summed in eq to calculate the depth-averaged fluence.

Data were fit to a first-order decay rate model with respect to fluence

$$\ln \left(\frac{C_t}{C_0}\right)={\beta }_0-\mathrm{k̂}\mathrm{F}\mathrm{l}\mathrm{u}\mathrm{e}\mathrm{n}\mathrm{c}\mathrm{e}$$ 2

where C ₀ is the initial viral concentration, C _t is the viral concentration at each time point, β₀ is the intercept, $\mathrm{k̂}$ is the light-screening corrected first-order decay rate constant (m²/MJ), and Fluence has units of MJ/m². C ₀ and C _t units are PFU/mL (bacteriophages) or TCID₅₀/mL (IAV).

Data were also fit to a first-order decay rate model with respect to time

$$\ln \left(\frac{C_t}{C_0}\right)={\beta }_0-k_{\mathrm{o}\mathrm{b}\mathrm{s}}t$$ 3

where C ₀ is the initial viral concentration, C _t is the viral concentration at each time point, β₀ is the intercept, k _obs is the first-order decay rate constant (hour^–1), and t is time (hour). C ₀ and C _t units are the same as eq . Equation yields a raw k _obs, uncorrected for light screening with inverse time units.

To convert the fluence-based solar photoinactivation rate ( $\mathrm{k̂}$ ) to units of inverse time, Fluence was divided by the integrated irradiance from 280 to 400 nm of the standard intensity spectrum across the United States (NREL AM1.5G, which is considered to be a reasonable average for the 48 contiguous U.S. states over a period of one year). The resulting time values were then used in eq to obtain ${\mathrm{k̂}}_{\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}$ , which we define as a light-screening corrected first-order decay rate constant with units of inverse time (hour^–1).

For tables and plotting, k _obs, $\mathrm{k̂}$ , and ${\mathrm{k̂}}_{\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}$ were calculated using all trials within each condition. To obtain sufficient degrees of freedom for statistical tests, $\mathrm{k̂}$ were calculated for each trial. Comparisons of $\mathrm{k̂}$ between two experimental conditions was made with a Student’s t-test, while comparisons of $\mathrm{k̂}$ between more than two experimental conditions were made using an analysis of variance (ANOVA) test followed by a Tukey’s posthoc test. Mean $\mathrm{k̂}$ across all trials is reported for statistical comparisons. An alpha of 0.05 was chosen as the threshold for statistical significance, but marginal significance (alpha = 0.1) is considered. Arithmetic means and standard deviations (SD) of various experimental parameters, like virus stock concentration and temperature, are reported.

$\mathrm{k̂}$ calculated from the experiments with different NOM concentrations were fit to DOC concentrations with a linear regression and a nonlinear regression (log–linear regression model). The results were fit using the Akaike Information Criterion (AIC). The AIC is a measure of prediction error that balances model fit with complexity to avoid overfitting; a lower AIC is better. We choose a log–linear regression model as an estimate of a nonlinear relationship with a plateau. These two models were chosen as they have precedent in the literature. An exponential model was also considered, and although the data fit the model well, the model allows for the rate constant to increase unbounded as DOC increases and we therefore did not pursue it further, and retained the more parsimonious linear and log–linear models. Time to 99% inactivation calculations are described in the SI, Text S5.

Results

Experiment Properties

Solar simulator irradiance has a similar spectral shape to the standard intensity spectrum across the United States (NREL AM1.5G) and a spring day at Stanford, although at a lower integrated intensity, which made collecting experimental observations more feasible for rapidly decaying viruses (Figure a). Incident UVB (Figure b) is greatest with the 280 nm 50% long-pass cutoff filter, followed by the 305 nm 50% long-pass cutoff filter, and the 320 nm 50% long-pass cutoff filter had the least incident UVB. Absorbance (Figure c) is greatest for the undiluted colored surface water and least for PBS, with colored surface water dilutions falling in between those limits. The addition of different viruses and stabilizers/quenchers to the matrices had negligible impact on absorbances (Figure S4).

1.

Solar simulator (set to an intensity of 250 W/m² from 280 to 700 nm, 24.6 W/m² from 280 to 400 nm) and water matrix properties. (A) Irradiance spectra of the solar simulator, NREL AM1.5G, and natural sunlight on a spring day (Mar-14-2025, 3:15 pm) at Stanford, CA from 280 to 700 nm. (B) Irradiance spectra of the solar simulator from 280 to 340 nm with no filter, and the 280, 305, and 320 nm 50% long-pass cutoff filters used compared to natural sunlight. (C) Absorbance spectra of colored surface water dilutions at 50% colored surface water and 50%H₂O, 75% colored surface water and 25%H₂O, 90% colored surface water and 10%H₂O, and 100% colored surface water compared to PBS.

The colored surface water has the following properties: pH of 7.9 ± 0.1 (n = 2), 1.7 ppt salinity, 23.4 ± 2.9 mg/L DOC (n = 2), and 0.6 μM NO₃ ^– (n = 1). A nitrate measurement obtained 8 months after the conclusion of the experiments was higher at 70 μM (Text S6). Of metals measured in elemental analysis (Table S1), Mn and Sr concentrations were notably high, at 27.3 μM and 12.6 μM, respectively. Under simulated sunlight exposure with the 280 nm cutoff filter and the open-atmosphere experiment setup, the steady state hydroxyl radical was ∼8 × 10^–15 M. For all open-atmosphere experiments, temperature within the beaker averaged 24.4 °C, SD = 0.5 °C while the solar simulator was running and mean DO was 233 μM, SD = 9.4 μM (Figure S5). For anoxic experiments, temperature averaged 23.8 °C, SD = 1.9 °C, and mean DO was 0.82 μM, SD = 1.4 μM (Figure S5).

All negative virus controls for plaque assays and TCID₅₀ assays had zero plaques and no positive wells, respectively, indicating no virus contamination. All positive controls for plaque assays and TCID₅₀ assays had expected virus concentrations, indicating working quantification assays. For all dark experimental controls, first-order rate constants with respect to time (k _obs) were less than 0.063 h^–1 or had p-values were <0.05, indicating little to no viral decay (Figure S6, Table S2).

PBS and Colored Surface Water Experiments

Corrected first-order rate constants with respect to fluence ( $\mathrm{k̂}$ , eq , Table S2 and Figure ) for IAV in PBS are 82.1, 29.5, and 13.2 m²/MJ and 1125, 714, and 867 m²/MJ in colored surface water for the 280, 305, and 320 nm 50% long-pass cutoff filters, respectively. For Phi6, $\mathrm{k̂}$ are 16.2, 2.95, and 2.20 m²/MJ in PBS and 46.6, 36.0, and 39.8 m²/MJ in colored surface water for the 280, 305, and 320 nm 50% long-pass cutoff filters, respectively. For MS2, $\mathrm{k̂}$ are 4.62, 0.934, and −0.699 m²/MJ in PBS and 9.49, 6.52, and 6.39 m²/MJ in colored surface water for the 280, 305, and 320 nm 50% long-pass cutoff filters, respectively. The differences in $\mathrm{k̂}$ between PBS and colored surface water are significant for all viruses and 50% long-pass cutoff filters used (p < 0.005 for all experimental conditions).

2.

Photoinactivation of viruses as a function of UVA and UVB fluence. (A) Measured natural-log transformed normalized concentrations of IAV, Phi6, and MS2 as a function of fluence. For IAV, median ln­(C _t /C ₀) are plotted, as the median is provided from the TCID₅₀ calculator results; for Phi6 and MS2, mean ln­(C _t /C ₀) are shown based on plaque assay results. Each point represents a trial and time point for each experimental condition. Error bars indicate ± the standard deviation from quantification assay replicates. (B) The bar plot shows the calculated $\mathrm{k̂}$ (eq ) across trials for each condition. Error bars indicate ± the standard error of the linear regression slope across all trials. Symbol *** indicates p < 0.001, ** indicates p < 0.01, * indicates p < 0.05; p ≥ 0.05 are not plotted.

When examining the effect of incident UVA and UVB in PBS, $\mathrm{k̂}$ is similar between the 305 and 320 nm 50% long-pass cutoff filter conditions (Figure ). In PBS, $\mathrm{k̂}$ is different between the 280 and 320 nm cutoff filter conditions for all viruses (p < 0.001 for IAV, Phi6, and MS2; mean difference of 69.0 m²/MJ for IAV, for 14.0 m²/MJ Phi6, and 5.3 m²/MJ for MS2) and is also different between the 280 and 305 nm cutoff filter conditions for all viruses (p < 0.001 for IAV and Phi6, and p = 0.004 for MS2; mean difference of 52.6 m²/MJ for IAV, for 13.3 m²/MJ Phi6, and 3.7 m²/MJ for MS2). In colored surface water, $\mathrm{k̂}$ is significantly different between the 280 and 305 nm cutoff filters for IAV (p = 0.04; mean difference of 411 m²/MJ) and MS2 (p = 0.03; mean difference of 3.0 m²/MJ). It is also significantly different between the 280 and 320 nm cutoff filters for MS2 (p = 0.02; mean difference of 3.1 m²/MJ).

Raw, uncorrected first-order decay rate constants (k _obs, eq ) are shown in Table S2 and Figure S7. For IAV in PBS k _obs is 6.85, 2.46, and 1.04 h^–1 for the 280, 305, and 320 nm 50% long-pass cutoff filters, respectively. In colored surface water k _obs are 80.4, 51.1, and 60.4 h^–1 for the 280, 305, and 320 nm 50% long-pass cutoff filters, respectively. For Phi6, k _obs are 1.34, 0.242, and 0.173 h^–1 in PBS and 3.03, 2.33, and 2.51 h^–1 in colored surface water for the 280, 305, and 320 nm 50% long-pass cutoff filters, respectively. For MS2, k _obs are 0.381, 0.077, and −0.055 h^–1 in PBS and 0.626, 0.428, and 0.404 h^–1 in colored surface water for the 280, 305, and 320 nm 50% long-pass cutoff filters, respectively. Corrected first-order rate constants with respect to time ( ${\mathrm{k̂}}_{\mathrm{t}\mathrm{i}\mathrm{m}\mathrm{e}}$ , eq ) are described in detail in Table S2 and Figure S8.

Mechanistic Experiments

In these experiments, we first investigated the impact of various PPRIs on Phi6 solar photoinactivation with quenchers/stabilizers using the 305 nm 50% long-pass cutoff filter (Figure ). The 305 nm 50% cutoff filter was chosen over the 280 nm 50% cutoff filter to reduce direct inactivation pathways. Addition of D₂O to the system, intended to promote singlet oxygen-mediated damage, increased $\mathrm{k̂}$ but the difference between treatment and control was only marginally statistically significant ( $\mathrm{k̂}$ = 22.8 m²/MJ with H₂O versus 19.4 m²/MJ with D₂O, p = 0.09); given the effect size, singlet oxygen can account for 18% of $\mathrm{k̂}$ observed for the control (see Text S7). Addition of isopropyl alcohol to the experimental system, known to quench hydroxyl radicals, reduced $\mathrm{k̂}$ ( $\mathrm{k̂}$ = 20.5 m²/MJ with isopropyl alcohol versus 36.0 m²/MJ without, p = 0.003). Addition of 2,4,6-trimethylphenol to the system, intended to quench triplet state NOM, did not affect $\mathrm{k̂}$ ( $\mathrm{k̂}$ = 41.0 m²/MJ with 2,4,6-trimethylphenol versus 36.0 m²/MJ without, p = 0.29). Finally, when oxygen was removed from the experimental system, $\mathrm{k̂}$ decreased by more than half ( $\mathrm{k̂}$ = 14.0 m²/MJ under anoxic versus 36.0 m²/MJ under oxic conditions, p < 0.001), suggesting removal of oxygen reduces production of PPRI involved in photoinactivation.

3.

Photoinactivation of Phi6 with PPRI quenches/stabilizers. All plots are shown as a function of fluence with points showing the mean log reduction per trial and error bars indicating ± the standard deviation from quantification assay replicates. (A) ln­(C _t /C ₀) of Phi6 with 50% D₂O, compared to 50%H₂O and PBS. (B) ln­(C _t /C ₀) of Phi6 with isopropyl alcohol and colored surface water, compared to pure colored surface water and PBS. (C) ln­(C _t /C ₀) of Phi6 with 2,4,6-trimethylphenol and colored surface water, compared to pure colored surface water and PBS. (D) ln­(C _t /C ₀) of Phi6 in anoxic conditions with colored surface water, compared to colored surface water open to the atmosphere and PBS.

For both IAV and Phi6, there appears to be a linear trend between $\mathrm{k̂}$ and DOC concentration with a slope of 24.5 (m² L/(MJ mg)) for IAV and 1.26 (m² L/(MJ mg)) for Phi6 (Figures and S9). The AIC for IAV is 64.0 for a linear regression versus 71.0 for the nonlinear log–linear regression and the AIC for Phi6 is 27.1 for a linear regression versus 33.3 for the nonlinear log–linear regression, indicating the linear regression fits the data better than the nonlinear log–linear regression. If there is no NOM adsorption to the virus, after accounting for light screening impacting fluence, we would expect a linear trend in corrected first-order decay rate constants as NOM increased (i.e., the colored surface water was less diluted).

4.

Photoinactivation of viruses in various dilutions of colored surface water with respect to dissolved oxygen concentration (DOC). Points show the calculated $\mathrm{k̂}$ across trials for each condition. Error bars indicate ± the standard error of the linear regression across all trials. Linear regressions (solid line) of $\mathrm{k̂}$ data are plotted. Plots are separated by virus species.

We incubated the virus in colored surface water in the dark, then isolated and purified Phi6 from the colored surface water, added the incubated virus to PBS, and exposed the virus to sunlight with the 280 and 305 nm 50% long-pass cutoff filters. Experiments conducted using the 280 nm long-pass cutoff filter showed $\mathrm{k̂}$ was not significantly different for the incubated Phi6 versus the unincubated Phi6 ( $\mathrm{k̂}$ = 20.0 m²/MJ in treatment versus 16.2 m²/MJ in control, p = 0.95) (Figures and S10). Experiments conducted using the 305 nm long-pass cutoff filter showed $\mathrm{k̂}$ was greater for incubated Phi6 versus the unincubated Phi6 in PBS ( $\mathrm{k̂}$ = 9.3 m²/MJ in treatment versus 2.9 m²/MJ in control, p < 0.001).

5.

Photoinactivation rate constants of Phi6 in PBS with 280 and 305 nm 50% long-pass cutoff filters after incubation in colored surface water in darkness and with sunlight exposure with the 305 nm 50% long-pass cutoff filter. Bar plots show the calculated $\mathrm{k̂}$ across trials for each condition, with error bars representing ± the standard error of the linear regression across all trials. Comparison $\mathrm{k̂}$ values are shown for nonincubated Phi6 in PBS and colored surface water.

In addition to incubating Phi6 in colored surface water in darkness, we also exposed it to sunlight, then isolated and purified the virus from the surface water, transferred it to PBS, and re-exposed it to sunlight. $\mathrm{k̂}$ was greater for Phi6 after incubation in colored surface water and exposure to sunlight than after incubation in darkness (for the 280 nm cutoff filter $\mathrm{k̂}$ = 30.7 m²/MJ in treatment versus 20.0 m²/MJ in control, p = 0.02; for the 305 nm cutoff filter $\mathrm{k̂}$ = 19.5 m²/MJ in treatment versus 9.3 m²/MJ in control, p = 0.005) (Figure , log reduction graphs in Figure S10).

Discussion

Photoinactivation rate constants for IAV, Phi6, and MS2 are larger in colored surface water than in clear water with sunlight exposure, even without controlling for light screening. This is especially evident for the enveloped viruses. Solar photoinactivation rate constants for IAV reported here are roughly 2 orders of magnitude greater than those observed for the nonenveloped MS2, indicating rapid decay. Direct comparisons of decay rate constants measured here and in previous work are challenging due to differences in NOM concentrations, artificial or natural sensitizer use, and irradiance conditions. Generally, however, prior studies reporting decay rate constants of influenza virus in PBS exposed to polychromatic UVA and UVB light (not sunlight) are of a similar order of magnitude to those observed here.

Reducing incident UVB wavelengths did not significantly affect most decay rate constants of the enveloped viruses studied in colored water. The exception was a difference in $\mathrm{k̂}$ between the 280 and 305 nm cutoff filters for IAV. This general trend indicates the dominance of indirect, PPRI-mediated solar photoinactivation pathways for enveloped viruses, which contrasts with our observations in clear water. Our nonenveloped virus results agree with observations by others conducted using bacteria and nonenveloped viruses in clear water, ,,,,, where rate constants decrease with decreasing UVB, illustrating the importance of direct solar photoinactivation.

Experiments exploring the importance of different PPRIs in the indirect solar photoinactivation pathway suggest hydroxyl radicals generated in the bulk colored water matrix may play a more important role than singlet oxygen or triplet state NOM. Prior work ,,, with nonenveloped MS2 that found singlet oxygen was critical in promoting indirect solar photoinactivation in water containing commercially available NOM, yet our results suggest a potentially smaller role (Text S7). Studies using bacteria and nonenveloped rotavirus , found hydroxyl radicals, as well as hydrogen peroxide, played important roles in solar photoinactivation. Reduced solar photoinactivation rate constants observed under anoxic conditions further support the importance of hydroxyl radicals, which are formed photochemically at lower rates in anoxic, NOM-containing waters compared to oxic waters. ,,, It is important to note that PPRI formation and loss mechanisms can be complex and the experimental treatments we used intending to limit formation or loss of different PPRIs are imperfect. For example, D₂O can both enhance formation rates and reduce loss rates of singlet oxygen in the presence of different NOM molecules. , Additionally, all triplet state NOM may not have been quenched in our experiments, due to possible presence of oxidative and energy transfer types, , and owing to the lower than ideal concentrations of quencher we had to use to avoid negative impacts of the quencher on virus stability. Therefore, triplet state NOM contributions to photoinactivation may be underestimated. We also did not explicitly consider the role of other PPRIs including H₂O₂. Additional work is warranted to better understand the role of PPRIs generated in the bulk solution.

Adsorption of NOM to viruses has been shown to facilitate indirect solar photoinactivation by possibly enhancing local exposure to ROS generated at the virion interface. We found that allowing NOM to sorb to Phi6 enhanced its solar photoinactivation with the 305 nm cutoff filter in clear water, suggesting this could be occurring in our system when direct UVB-driven inactivation is limited. If this is the case, addition of substances to the bulk solution to modify steady-state PPRI concentrations, as we did in some of our experiments with the 305 nm filter, would not necessarily affect PPRIs generated at the interface. This leaves open the possibility that, when UVB is limited, PPRIs generated at the virion interface enhance indirect solar photoinactivation pathways for Phi6.

Interestingly, we found that allowing NOM to sorb to the virus and exposing it to light prior to the start of the experiments further enhanced its solar photoinactivation in clear water. This suggests that sublethal photodamage of the virus, caused by indirect pathways, further enhances its susceptibility to solar photoinactivation. Recent work by Shin et al. described this phenomena for adenovirus 2 where sublethal damage by singlet oxygen to a viral enzyme involved in transcription/replication made the virus more susceptible to DNA damage upon UVB exposure. Phi6 is a dsRNA virus, so the results from Shin et al. do not apply; however sublethal damage could occur to viral components critical to replication such as proteins in the virion, like nucleocapsid proteins or RNA polymerase proteins.

If PPRIs generated at the virion interface by NOM controlled solar photoinactivation, then we would expect that as bulk phase NOM concentration increases, that the virus surface might become saturated with NOM molecules and that rate constants would remain constant as NOM concentrations increase. This was previously observed by Kohn et al. who studied MS2 solar photoinactivation in an aqueous matrix containing commercially available NOM. However, we did not observe that to occur in our system, even with higher DOC concentrations than those used in Kohn et al. Instead, Phi6 and IAV rate constants increased linearly with NOM concentration. This could suggest that even at the highest concentrations of NOM (100% colored surface water matrix), the Phi6 and IAV surfaces did not become saturated with NOM, and/or bulk phase PPRIs, whose steady-state concentrations scale with NOM, remain important even though interfacial PPRI generation is occurring. We hypothesize that both interfacial and bulk-phase generation of PPRIs are important in the indirect solar photoinactivation pathways of Phi6 and IAV, as evidenced by the higher rate constants observed in the bulk colored surface water matrix compared to the presorption experiments.

The biological targets of PPRIs could be viral nucleic acids, proteins, or lipids. Experiments conducted herein were not designed to identify the biological targets and therefore future work should do so. To date, previous studies examining biological damage caused by light are limited to UVC irradiation and/or analysis of nonenveloped viruses, so there is a clear knowledge gap to be addressed. ,,− ,,− We are currently examining the potential roles of lipid oxidation in the inactivation of enveloped viruses in sunlight.

Environmental Implications

The rapid decay of IAV in colored surface water has important implications for viral persistence in the environment. The values for $\mathrm{k̂}$ reported here can be used to estimate the time to reach 99% solar photoinactivation in an actual water body (T ₉₉). Even though $\mathrm{k̂}$ values are much larger in colored water than clear water for the viruses, T ₉₉ will depend on the available UVB and UVA photons as well as $\mathrm{k̂}$ . In darkness, T ₉₉ would be roughly 2 weeks for Phi6 and IAV, based on mean values from a systematic review. We modeled T ₉₉ for the viruses as a function of depth for a well-mixed water column with standard intensity spectrum across the United States (NREL AM1.5G) incident at the surface (Figure S11). For a 1 m depth well-mixed water column with constant, collimated flux, T ₉₉ for IAV is approximately 1.2 h, whereas T ₉₉ for Phi6 and MS2 are 28 h and 5.7 days, respectively. In clear water, the time to reach 99% solar photoinactivation for IAV, Phi6, and MS2 would be 1.8, 8.9, and 31 h, respectively, in a 1 m well-mixed water column.

A complete data set for this paper is published in the Stanford Digital Repository, available at https://purl.stanford.edu/qs728mp7384 and 10.25740/qs728mp7384.

The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.5c13495.

• Text S1: viral sequencing information, Text S2: virus propagation methods, Text S3: Metadata collection methods for the collected surface water, Text S4: preliminary quenching experiments, Text S5: time to inactivation calculation, Text S6: nitrate concentratons, Text S7: singlet oxygen contribution calculation, Table S1: elemental analysis, Table S2: corrected first-order rate constants with respect to time, Figure S1: sampling location and water image, Figure S2: absorbance over time and decay of viruses at different time points of water storage, Figure S3: experimental setup, Figure S4: additional absorbance measurements, Figure S5: temperature and DO, Figure S6: dark control results, Figure S7: log reduction versus time, uncorrected, Figure S8: log reduction versus time, corrected, Figure S9: log reduction for dissolved organic carbon concentrations, Figure S10: log reduction for adsorption experiments, and Figure S11: components of T99 as a function of depth (PDF)

The authors declare no competing financial interest.