Optical and Photochemical Properties of Suwannee River Natural Organic Matter Fractions Generated by Preparative Free Flow Electrophoresis Garrett McKay, Kai Cheng, András Gáspár, Mourad Harir, Philippe Schmitt-Kopplin, Norbert Hertkorn, Ronald Benner, Karl Kaiser Abstract Fractionation of dissolved organic matter (DOM) has been used for decades to simplify the challenging task of bulk sample characterization and to gain insight into the structural features governing DOM reactivity. Here, we used preparative free-flow electrophoresis (PFFE), which separates molecules based on their charge-to-size ratio, to separate Suwannee River natural organic matter (SRNOM) into fractions with varying electrophoretic mobility. Optical and photochemical data of the fractions were paired with extensive analytical characterization. The least electrophoretically mobile fraction had the lowest number-average molecular weight (M n), aromaticity, and specific ultraviolet absorbance (SUVA254), while showing the largest quantum yields of fluorescence as well as singlet oxygen and hydroxyl radical formation. Structure–activity relationships for the fractions, which exhibited substantial differences in mobility, aromaticity, and M n did not follow the conventional trends observed in size-based DOM fractionation studies (e.g., SUVA254 ∝ aromaticity). These findings highlight the role of charge and structural differences in driving optical and photochemical properties apart from bulk aromaticity and M n. The distribution of absorbance, the sum of vanillyl and syringyl phenols, and DIOHBA (3,5-dihydroxybenzoic acid) indicates that the most electrophoretically mobile fractions with the highest absorbance are dominated by DIOHBA, a phenol monomer derived from tannins and flavonoids, rather than lignin. DIOHBA was inversely related to the fluorescence index and β/α. Overall, these results provide insight into SRNOM photophysics and highlight the utility of PFFE as a tool for resolving DOM structure–reactivity relationships. [es6c07200_0009.webp] [es6c07200_0007.webp] Introduction Dissolved organic matter (DOM) governs the mechanisms and rates of key processes in aquatic systems, such as elemental cycles, − attenuation of contaminants, − and productivity of the Earth’s oceans. , Many of these processes are initiated by the absorption of sunlight by chromophoric DOM, which comprises up to ∼50% of the total DOM pool in terrestrial environments. DOM is an exceedingly complex mixture of molecules, rich in oxygenated functional groups, and sp2 hybridized carbon with varying molecular size. − DOM molecular size and composition are altered during environmental processes, such as adsorption, oxidation, − and coagulation. DOM structure impacts its reactivity in natural and engineered systems. One approach to resolving these structure–reactivity relationships is to fractionate DOM by size or polarity. It has been extensively demonstrated that molecular size is an important determinant of DOM optical properties , and photoreactivity , using size-based methods such as gel permeation chromatography, , field-flow fractionation, , and ultrafiltration. ,, Although molecular weight depends only on the molecular structure, molecular size is a complex function of structure and environmental conditions such as pH and ionic strength. , For example, ultrafiltration separates molecules based on size (weight is calibrated with globular proteins), but the resulting fractions are nonetheless often described as high- and low-molecular-weight fractions. It has been shown that lower molecular weight DOM generated from size-based separations has lower specific absorbance, , larger spectral slope, higher fluorescence quantum yields (Φ f ), blue-shifted emission, and higher quantum yields for the photoproduction of reactive intermediates. ,, These lower molecular weight fractions also have a lower proportion of sp2-hybridized carbon than higher molecular weight fractions. Evidence for this comes both from nuclear magnetic resonance (NMR), which demonstrates a higher abundance of sp2 carbon, and Fourier transform ion cyclotron resonance mass spectrometry, which shows lower H/C ratios, higher double bond equivalence, and more formulas with an aromaticity index >0.5 for high relative to low molecular weight fractions. , Several recent efforts have also focused on resolving DOM optical spectra via polarity-based separations with commercial solid-phase extraction cartridges. − In contrast to separation by molecular size and polarity, DOM separation by charge density is less common. Capillary electrophoresis with optical or mass spectrometric detection demonstrated that electropherograms could act as a DOM fingerprint because the electrophoretic mobility is directly related to the charge and hydrodynamic friction of molecules in complex organic mixtures. In addition, Trubetskoj and coworkers investigated the optical and photochemical properties of DOM fractions of different electrophoretic mobility, which were generated by size exclusion chromatography (SEC) in the presence of 7 M urea. ,, The predominant acidic functional groups in DOM are carboxylic acids and phenols, and the protonation state of these functional groups has been shown to alter DOM optical properties and photoreactivity. , Although there are some reports employing absorbance and fluorescence detectors during capillary electrophoresis, complete spectral and photochemical characterization of DOM separated by charge-based methods is largely absent in the literature. Here, we report the optical and photochemical properties of Suwannee River natural organic matter (SRNOM) fractions generated by preparative free-flow electrophoresis (PFFE) in a previous study. We characterized PFFE fractions and bulk SRNOM by optical measurements (absorbance and fluorescence), spectral pH titrations, fluorescence quenching, and measurements of apparent quantum yields for the photochemical production of singlet oxygen (1O2, ΦΔ) and hydroxyl radical (•OH, ΦOH). Correlations were examined between photochemical parameters and a wealth of compositional data, including NMR-derived functional group composition, molecular weight, amino acid content, and phenol monomers produced after copper oxide oxidation. A primary goal of this study is to assess the utility of PFFE as a DOM fractionation technique with respect to the molecular distinctness of its fractions. In addition, we expect that these data will provide new insight into the structural and compositional underpinnings of SRNOM optical and photochemical properties. Materials and Methods Preparative Free-Flow Electrophoresis PFFE of SRNOM (1R101N) was performed in Munich using an Octopus PZE device manufactured by Dr. Weber GmbH (Kirchheim, Germany), as described in earlier studies. − The device has a vertical separation chamber with dimensions of 500 mm (length) × 100 mm (width) × 0.5 mm (thickness), resulting in a chamber volume of 25 mL. The separation was performed at a temperature of 15 °C and a separation voltage of 500 V. The separation medium consisted of 5 mM ammonium carbonate (pH 9.2), while the electrode buffer consisted of 500 mM ammonium carbonate buffer at the same pH. The sample (20 mg mL–1 dry weight) was continuously pumped (2 mL h–1) into the bottom of the chamber and carried by the separation medium (965 mL h–1) in a laminar flow profile. At the top of the chamber, the separation medium and sample were collected in a tray containing 96 wells. Solutions from wells were binned into 11 fractions, and fractions were lyophilized to obtain solid for further analysis and to remove ammonium carbonate. The appearance of a dry, fluffy powder after lyophilization is strong evidence of the removal of ammonium carbonate. Although 11 fractions were collected during PFFE, only five were analyzed in detail here, as they are the main contributors to the absorbance detected in the electrophoretogram. Analytical Methods The SRNOM fractions and their analytical characterization were originally described by Gáspár, which included NMR and amino acid analyses. A brief description of all analytical methods is provided here (see complete details in the Supporting Information). Optical measurements, SEC, and photochemical characterization were performed in 2020 and are original to this work. Optical measurements, SEC, and photochemical characterization were completed at Texas A&M. Briefly, absorbance and fluorescence spectra were collected with an Aqualog (Horiba). Standard protocols for data treatment were implemented, such as blank subtraction, inner-filter corrections, Raman normalization, and Rayleigh masking. , Optical surrogates were calculated according to Korak and McKay and included the specific ultraviolet absorbance at 254 nm (SUVA254), E2:E3 ratio, spectral slope between 275 and 290 (S 275–290) and 350–400 (S 350–400), the fluorescence index (FI), and the freshness index (β/α). SEC was completed with an Ultimate 3000 UHPLC (Thermo Fisher) and Waters Protein-Pak (15 × 300 mm) column, with a mobile phase consisting of 0.1 M potassium chloride, 2 mM potassium dihydrogen phosphate, and 2 mM potassium hydrogen phosphate at pH 6.8. The flow rate was 1.0 mL min–1. Retention time was calibrated to molecular weight using polystyrenesulfonates (PSS) ranging from 246 to 18 000 Da. SEC is one of the most common techniques used in the DOM community for obtaining apparent molecular weights. ,, Although PSS do not perfectly represent DOM size, number-average molecular weights (M n) measured by SEC with PSS calibration agree well with values from vapor pressure osmometry. , Total organic carbon was measured with a Shimadzu V-SCH organic carbon analyzer on 0.45 μm-filtered samples using the non-purgeable organic carbon method. NMR analysis was performed in Munich, Germany, using a Bruker DMX spectrometer (B 0 = 11.7 T). We focus here on 13C data, which were collected for PFFE fractions (5–35 mg) redissolved in up to 135 μL CD3OD (99.96% 3H, Merck, Darmstadt, Germany) at 283 K. Shift regions were integrated to obtain quantitative estimates of functional group distributions (see Text S1). Amino acid analysis was performed in Columbia, South Carolina, while copper oxide oxidation and phenol monomer analysis were performed in Galveston, TX. Briefly, amino acid content was measured on hydrolyzed samples (6 M HCl, 20 h, 110 °C) by HPLC after in-line derivatization with o-phthaldialdehyde/N-isobutyryl-l-cysteine or o-phthaldialdehyde/N-isobutyryl-d-cysteine. Quantification of phenol monomers after copper oxide oxidation was performed by GC/MS following cleanup with solid-phase extraction (Oasis HLB cartridges). p-Hydroxyphenyl (P), vanillyl (V), syringyl (S), and cinnamyl (C) phenols were quantified, including p-hydroxybenzaldehyde (PAL), p-hydroxyacetophenone (PON), p-hydroxybenzoic acid (PAD), vanillin (VAL), acetovanillone (VON), vanillic acid (VAD), syringaldehyde (SAL), acetosyringone (SON), syringic acid (SAD), p-coumaric acid (CAD), and ferulic acid (FAD). The sum of V and S phenols (Σ6 = VAL + VON + VAD + SAL + SON + SAD) is a strong indicator of lignin content. 3,5-Dihydroxybenzoic acid (DIOHBA) was also measured. Solution Preparation Solutions for SEC, optical, and photochemical analyses were prepared using ultrapure water (>18.0 MΩ cm) from a Barnstead water system (Thermo Fisher). PFFE fractions and bulk SRNOM were dissolved in ultrapure water. Sample pH was adjusted to 7.0 with 0.1 M NaOH, and solutions were stirred overnight and filtered with pre-rinsed 0.45 μm polyether sulfone syringe filters. These stock solutions were stored at 4 °C until dilution for subsequent experiments. Dilutions were prepared in either 5 mM potassium phosphate buffer at pH 7.2 (for optical and photochemical measurements) or, for fluorescence quenching (see Text S1), in acetic acid (pH 4) or sodium carbonate (pH 9). Photochemical Measurements A more detailed description of the probe compound methods is provided in Text S2. A Rayonet merry-go-round photoreactor with 12 365 nm bulbs, equipped with a fan, was employed for all irradiations, which were conducted in borosilicate glass tubes (effective path length of 0.84 cm). Briefly, 25 μM furfuryl alcohol and 2 mM benzoate were used for quantifying the formation rates of 1O2 and •OH, respectively. The rate of •OH formation was derived from the rate of salicylate formation using the yield of 0.204 mol salicylate per mol •OH. Apparent quantum yields were calculated using the measured 1O2 or •OH formation rate and the rate of light absorption, the latter calculated using the photon irradiance (measured with p-nitroanisole/pyridine), solution absorbance, and path length in the quartz tube. Formation of DOM triplet states was not measured in this study (e.g., using trimethylphenol) as 1O2 is considered an indicator of triplet reactivity. Although commonly used within the aquatic photochemistry literature, the 365 nm light source in our study likely results in higher ΦΔ values compared to a solar simulator due to ΦΔ’s wavelength dependence. However, our objective was to investigate fundamental structural features governing SRNOM optical properties and photochemistry, not to accurately represent environmental polychromatic conditions. Statistical Methods Correlation analysis was performed using the Statistics and Machine Learning Toolbox in MATLAB 2024b. Pearson correlation coefficients and their associated P values were calculated. Correlations with a P-value of 0.01 (indicated with an *) were classified as significant due to the small number of points in each correlation (n = 6). Correlations with 0.01 ≤ P < 0.05 are considered suggestive trends, and the ρP values are still shown. Results Preparative Free-Flow Electrophoresis Produces SRNOM Fractions with Varying Optical and Photochemical Properties PFFE separates molecules based on size and charge, with a molecule’s effective electrophoretic mobility being proportional to its charge-to-size ratio. Higher charge density translates into a higher velocity of the ions in solution in the electric field and thus effective electrophoretic mobility (field-strength-normalized ion velocity). The electropherogram demonstrates that the majority of SRNOM molecules have a narrow range of electrophoretic mobilities, between −0.03 and −0.005 cm2 V min –1 (). Although only five fractions are discussed below (Fr6–Fr10), the mass of organic carbon (OC) in all 11 fractions (Fr1–Fr11) was measured to assess the mass balance (Table S2). Approximately 80% of the total SRNOM mass was recovered, while slightly less of the OC was recovered (73%), which could be explained by the incomplete removal of ammonium carbonate. Relative to the mass quantified in the 11 fractions, fractions 6 through 10 comprise more than 95% of the SRNOM and OC mass as well as most of the absorbance at 254 nm in the electrophoretogram ( and S1). The unrecovered mass may include high-mobility molecules such as small, aliphatic carboxylic acids. Although we did not examine the organic carbon concentration in each tube after PFFE, an early report by Leenheer and Malcolm shows a substantial organic carbon peak that is weakly UV-absorbing with high electrophoretic mobility. Given the probable chemical nature of the missing mass, we do not expect that its absence will impact the optical and photochemical properties observed here. PFFE of SRNOM produces fractions with distinct chemical, optical, and photochemical properties ( and ). The least mobile Fr10 has the lowest SUVA254 (0.98 L mgC –1 m–1), highest FI (1.58) and β/α (0.831), lowest M n (344 Da), and lowest abundance of aromatic plus phenolic carbon (herein referred to as aromaticity) (24.8%). Although there is a general trend of decreasing M n and aromaticity with decreasing mobility (Fr6 → Fr10), the first three fractions are largely invariant in these properties. Absorbance-based surrogates (SUVA254 and E2:E3) and Φ f do not show a consistent trend with decreasing mobility (B and D). In contrast, FI and β/α (fluorescence-based surrogates) generally increase with decreasing mobility (C). Fr10 has the highest Φ f , ΦΔ, and ΦOH, Fr7 has the lowest values, and values for Fr6 and Fr8 are slightly larger than those for Fr7 (D–F). This pattern matches the trend of E2:E3. These results contrast with prior reports employing ultrafiltration, which typically show continual changes in optical surrogates and photochemistry with varying nominal molecular-weight cutoff of the ultrafiltration membrane. ,, For example, Fr7 has the second highest M n and aromaticity but the second lowest SUVA254 (3.2 L mg C–1 m–1), while Fr6 has the highest SUVA254 and highest M n (938 Da). Fr7 and Fr9 have different aromaticity and M n but similar SUVA254 (A and B). Although chemical composition and molecular weight are not directly measured in all ultrafiltration studies, most of those that measure these parameters (e.g., by mass spectrometry for aromaticity or SEC for molecular weight) report that permeate DOM has lower aromaticity and molecular weight. A conservative mixing model was applied to the data in which the signal for bulk SRNOM (Sbulkcalc) was calculated based on the fractional mass (fi ) and each fraction’s signal (S i ) such that Sbulkcalc=Σi=15fiSi (). Additional details are shown in Text S3. When fi is based on total mass, Sbulkcalc underestimated measured signals by between 14 and 53% ([Sbulkcalc−Sbulkmeas]/Sbulkmeas×100) . In contrast, model calculations are closer to measured signals when the total mass of OC is used to calculate fi , with percent differences between −41 and +5%. When using OC mass for fi , differences between modeled and calculated FI, M n, aromaticity, and λpH4 are greater than −10% while SUVA254 has a −13% difference. Φ f , ΦΔ, and ΦOH exhibit the poorest agreement between model and measured signals, with model values underestimating the measured signals by 30% or more. The disagreement between the conservative mixing model and bulk sample measurements suggests nonadditive interactions in bulk SRNOM that are not retained in PFFE fractions. The incomplete mass balance obtained during PFFE should be noted as a caveat warranting future examination. Nevertheless, bulk values are within the range of PFFE fractions, indicating that fractionation captures the full extent of compositional variability. Fluorescence Quenching and Spectral pH Titration Reveal Charge-Density Differences of Chromophores in PFFE Fractions PFFE does not discriminate between chromophoric and nonchromophoric molecules. Thus, one would expect that the chromophores present in Fr6 would have greater charge-to-size ratios than those of later fractions. Fr6 also has the largest measured molecular weight (A); thus, it should have a far greater number of charged sites than other fractions to have the largest charge density and electrophoretic mobility. pH titrations and fluorescence quenching were used to evaluate the charge density of chromophores in the PFFE fractions. Fluorescence quenching with tempamine (TA) and tempol (TO) provides an assessment of the surface charge of fluorophores. , The ratio of Stern–Volmer constants for each quencher (K TA/K TO), defined as λ, is related to the apparent attractive electrostatic potential (Ψ′) via λ = exp­(−Ψ′/k B T), where k B is the Boltzmann constant and T is the temperature. Another way of conceptualizing λ is as the ratio of local to bulk quencher, which is unity for neutral tempol and >1 for cationic tempamine. The measured value of λ is lower than the actual value due to dynamic quenching processes. Nonetheless, λ is a useful measure for assessing Ψ′′. λ values for all PFFE fractions are smaller at pH 4 than at pH 9 (), consistent with the larger expected charge density at pH 9 due to a greater fraction of deprotonated carboxylic acids and phenols. λ at pH 4 and 9 decrease with decreasing electrophoretic mobility, consistent with Fr6 having the highest charge-to-size ratio and therefore the largest attractive electrostatic potential. The difference in λ between pH 4 and 9 for each fraction is largest for Fr6 and bulk SRNOM, whereas λ are more similar at each pH for Fr8 through Fr10. Overall, the quenching behavior of SRNOM fractions corroborates the effective separation of fluorophores by PFFE into molecules with varying charge density. Spectral pH titrations provide an additional assessment of the chromophore charge density. The response of absorbance to an increase in pH has been linked to the roles of carboxyl- and phenol-containing chromophores. ,, The ratio of absorbance at 450 nm as a function of pH normalized to pH 2 (A pHX/A pH2) increases more rapidly with increasing pH for Fr6 relative to Fr9 across the pH 2 to 6 (carboxyl) and pH 8 to 12 (phenol) ranges (B). Thus, the pH-dependent absorbance behavior of PFFE fractions further corroborates the effective separation of chromophores within SRNOM. Relationships between Optical Surrogates and Compositions Relationships between compositional data and photochemical properties were assessed for bulk SRNOM and PFFE fractions using Pearson correlation coefficients (ρP ). The compositional data included NMR-derived functional group composition, SEC estimates of molecular weight, and the abundance of phenol monomers released during copper oxide oxidation. Amino acid yield is also included as it is often related to fluorescent DOM to infer information about source and level of processing.62 13C-NMR-derived functional group assignments, phenol monomer content, and amino acid yield are normalized to the mass of organic carbon in each fraction. Pearson was chosen over Spearman rank coefficients because the most interesting and relevant scatter plots exhibited linearity. The choice of Pearson vs. Spearman does result in differences in the correlation coefficient, but ρP and ρS exhibit a roughly one-to-one relationship (Figure S2). Differences in P-values between the two analyses can impact interpretations. For example, the FI-aromaticity relationship exhibits good linearity (R 2 = 0.816) and a significant Pearson correlation coefficient (ρ P = −0.91 and P < 0.01) but the calculated Spearman rank coefficient (ρ S = −0.83) has P > 0.01. A shows a heatmap of correlation coefficients, while B–U displays selected scatter plots. ρP with P-values <0.05 are printed, ρP with P-values <0.01 are marked with an ∗, and those with P > 0.05 are not displayed. Considering structural characteristics from each category, aromatic and carbonyl carbon, amino acid yield, M n, and DIOHBA exhibited the highest frequency of correlations with optical surrogates with P < 0.05. FI had eight correlations with P < 0.05 and five with P < 0.01, which was the most of any of the optical surrogates or quantum yields. Examining the correlation coefficients in A might suggest that the relationships between composition and optical or photochemical properties are consistent with original studies linking DOM composition to optical surrogates. For example, SUVA254 is positively correlated to aromaticity and negatively correlated to M n (B and C), consistent with Weishaar et al. and Chin et al. (for diverse hydrophobic organic acid isolates) and ultrafiltration studies; , spectral slope (S 275–295) and E2:E3 are negatively correlated to aromaticity and M n, consistent with past work on fractionated DOM (A, G, and H); , and ΦΔ is negatively correlated to M n, consistent with ultrafiltration studies. , Despite the appearance of significance based on ρP values, scrutiny of individual scatter plots is necessary to decipher which relationships are meaningful. Selected scatter plots with a linear regression fit and R 2 values are shown in B–U. By far, the tightest linear relationships are for FI, with R 2 > 0.9 for M n, amino acid yield, and DIOHBA content. DIOHBA shows the highest frequency of correlations with P < 0.05 (four have P < 0.01) among the phenol monomers, being negatively correlated to β/α (Figure S3) and ΦΔ (T). S 275–295 is not significantly correlated to any phenol monomer (P > 0.01) but is significantly related to aromaticity (P = 0.0388) and M n (P = 0.0269) (G and H). This contrasts with the results of Fichot and co-workers, which show a strong exponential relationship between S 275–295 and the sum of p-hydroxy, vanillyl, and syringyl phenols in whole water samples from the northern Gulf of Mexico and coastal areas worldwide. This is most likely due to the more limited range of S 275–295 values in our study (10–18 μm–1) compared to Fichot (10–40 μm–1) which included photobleached samples with higher S 275–295. S 350–400 is well correlated to p-hydroxyphenyl (P = 0.0132) and cinnamyl (P = 0.0361) phenols (ρP < −0.9) but not aromaticity or M n (P > 0.2). None of the relationships between phenol monomers and E2:E3 or SUVA254 are significant, although we should note that the P-value for the SUVA254-DIOHBA correlation is 0.070. DIOHBA has been interpreted as an indicator of tannins and flavonoids, distinct from the syringyl and vanillyl phenols traditionally considered surrogates for lignin. Interestingly, amino acid yield shows a strong relationship to FI and β/α (P ≤ 0.001 ρP > 0.97), which are surrogates for DOM source and age. This finding complements the results of Yamashita et al., who reported a positive correlation between a tryptophan-like parallel factor analysis component and amino acid yield in the northern Gulf of Mexico. Our results for PFFE fractions suggest that this correlation occurs for regions of the excitation–emission matrix that are less prone to low fluorescence intensities. Relationships between 1O2 and • OH Quantum Yields and DOM Structure ΦΔ varies by over two-fold between the fractions (0.0145 to 0.0321), whereas ΦOH values are largely similar except for Fr10 (1.2 × 10–4) which is four-fold larger than the lowest value, Fr7 (2.9 × 10–5). ΦΔ is correlated to aromaticity (P < 0.01), carbonyl content (P = 0.0336), and M n (P = 0.0363). In contrast, ΦOH is not significantly correlated to any of the compositional parameters (P> 0.01). Scatter plots of these data with a linear regression model are shown in Figure S5. The negative dependence of ΦΔ on both aromaticity and M n is consistent with previous studies employing ultrafiltration, , while the negative relationship to carbonyl content has, to our knowledge, not been reported previously. Discussion PFFE Produces Fractions with Distinct Structures but Similar Optical and Photochemical Properties A primary finding of this research is that the molecular structure, composition, and photoactivity of PFFE fractions appear to differ from size-based fractionation techniques such as ultrafiltration. Studies employing ultrafiltration typically show increasing E2:E3 with decreasing molecular weight, as defined by the nominal molecular weight cutoffs of the ultrafilters. For example, Maizel and Remucal demonstrate that ultrafiltration permeates of Pony Lake and Suwannee River fulvic acid have larger E2:E3 values than both the bulk sample and retentate, and show a concomitant increase in the abundance of formulas with an aromaticity index >0.5. Likewise, Mostafa et al. show that ultrafiltration permeates of SRNOM and a wastewater derived DOM have longer SEC retention times and larger E2:E3 values than their respective bulk and retentate fractions. In contrast, PFFE fractions of SRNOM show minimal change in E2:E3 from Fr6 through Fr9, which vary in molecular weight by nearly two-fold and aromaticity by 1.5-fold. There is no consistent trend in SUVA254 from Fr6–Fr9; however, Fr10 has the lowest M n, aromaticity, and highest E2:E3 of all fractions. In general, absorbance-based optical surrogates that characterize spectral slope do not vary considerably until Fr10. Surprisingly, Φ f displays a pattern similar to E2:E3. This contrasts with FI and β/α which generally increase with decreasing electrophoretic mobility and exhibit tight correlations to aromaticity and M n. Trubetskoj and coworkers investigated the optical properties and photoreactivity of DOM fractions with varying electrophoretic mobility, ,, and some of their observations align with those from PFFE. In Trubetskoj’s studies, DOM was separated by low-pressure SEC in the presence of 7 M urea and the electrophoretic mobility of the fractions was measured post hoc with gel electrophoresis. They observed that the most electrophoretically mobile fractions (C + D) had higher SUVA280 than the least mobile fractions (A and B). , They also showed that the more mobile fractions had similar E2:E3, similar to Fr6–Fr9 in our study. One discrepancy is that, for PFFE separation, fractions with higher electrophoretic mobility also have the largest molecular size, whereas Trubetskoj et al. found that the most mobile fraction had the smallest molecular size. In addition, Trubetskoj et al. observed that the most electrophoretically mobile fraction had higher photosensitizing efficiency, , which is opposite to the result seen here that Fr10 has the largest ΦΔ and ΦOH values. Although speculative, these discrepancies could be due to disruption of intermolecular interactions (e.g., H-bonding) by 7 M urea in the most electrophoretically mobile fractions that have a high abundance of H-bond donors and acceptors. Such disruptions could decrease the apparent size of the DOM, as measured by SEC, leading to higher photosensitization efficiency of the high-mobility fraction. Trubetskoj et al. dialyzed samples prior to photochemical measurements; thus, any disaggregation resulting from 7 M urea would need to be preserved post-dialysis for this explanation to be valid. In addition, because Trubetskoj and coworkers characterized electrophoretic mobility after SEC, separation by charge density (i.e., pulling apart molecules with similar size but different charge) would not have occurred during SEC. Clearly, more work is needed to evaluate these hypotheses. Regardless, the agreement between our study and Trubetskoj et al. that high-mobility fractions may have similar absorbance-based properties further supports the differences between charge-based separation and ultrafiltration. We propose two possible explanations for the lack of systematic relationships between aromaticity, M n, and absorbance-based surrogates among PFFE fractions. First, aromaticity and M n are coarse descriptors that do not fully capture the nuances of DOM structure. Separation by charge-density results in high-mobility fractions having a greater abundance of ionizable functional groups (Table S3). Spectra of fractions with hhigher charge density will be impacted by these additional functional groups in ways not easily captured with bulk measures of aromaticity and molecular weight. For example, phenol and acetovanillone have the same number of aromatic carbon atoms but quite different absorbance and fluorescence spectra due to the electronic nature of their substituents. The more mobile fractions exhibit red-shifted spectra relative to lower-mobility fractions (Table S4, Figure S6). This observation is consistent with the positive correlation between fluorescence asymmetry (a measure of red-shifted emission) and the relative abundance of aromatic and highly oxygenated formulas reported by Nikolaev and coworkers. Nikolaev and coworkers further hypothesized that this relationship is driven by the proportion of oxidized aromatic compounds with low pK a values, which are most likely to be carboxylic acids. These same structures govern separation of DOM molecules by PFFE and are more abundant in the high-mobility fractions (Table S3). A second explanation is related to the PFFE separation mechanism. Ultrafiltration separates DOM primarily on the basis of size, whereas PFFE separates DOM on the basis of size and charge. Thus, molecules with the same molecular weight but different charge densities will be separated in PFFE but not by ultrafiltration. This explains why Fr6 through Fr8 can have varying charge densities but similar M n. The additional separation dimension helps to resolve compositional heterogeneity among molecules with overlapping molecular weights. In contrast to absorbance-based surrogates, which do not vary systematically with electrophoretic mobility and exhibit weaker correlations to aromaticity and M n, FI and β/α show both a consistent increase with increasing electrophoretic mobility and have tight correlations to M n and aromaticity. The strong inverse correlation of these fluorescence-based surrogates to M n and aromaticity is consistent with the genesis studies of these optical surrogates as well as continued inquiry using ultrafiltered samples. ,− It has been well-documented that FI and β/α are inversely related to the maximum emission wavelength (λemmax) at the excitation wavelength used to determine these surrogates (see Figure S4). , Because λemmax is inversely proportional to the energy gap between the first excited and ground electronic states, this result suggests that fluorophores in more mobile fractions emit from lower energy states. The Importance of DIOHBA Another explanation for the lack of systematic relationship between structure and absorbance-based surrogates for SRNOM fractions is derived from the observation that Σ6 and DIOHBA have different distributions in the PFFE fractions (). DIOHBA is most abundant (i.e., μmol mgOC –1) in Fr6 and decreases monotonically with decreasing mobility, whereas Σ6 increases from Fr6 to Fr8 and decreases thereafter. Importantly, the same mobility pattern holds when absolute DIOHBA and Σ6 amounts (i.e., μmol) are compared to the electropherogram (Figure S9). We hypothesize that differences in the distribution of Σ6 and DIOHBA impact the structure of the aromatic chromophores in each fraction. Lignin phenol concentrations, which do not include DIOHBA, are often considered the main predictor of chromophoric DOM in natural waters ,, and are used as a tracer of terrigenous DOM. In contrast, our results suggest that material with the highest absorbance in the electropherogram (Fr6) is dominated by DIOHBA. In addition, DIOHBA abundance has a strong inverse correlation to FI and β/α. This study is the first to our knowledge to report significant correlations between DOM optical surrogates and DIOHBA. In contrast, studies of terrestrial and coastal systems ,, have reported both intrinsic , and extrinsic ,, correlations to lignin phenol composition. For the SRNOM fractions, Σ6 was correlated only to S 350–400. Combined, these observations strongly indicate that DIOHBA precursors are a dominant force in shaping the fluorophores governing FI and β/α, while lignin phenols play a role in S 350–400. DIOHBA is often interpreted as derived from tannins and flavonoids as opposed to lignin. Alkali-extracted lignin has Φ f values much lower and λemmax values higher than DOM isolates. Tannic acid fluorescence is complicated by self-oxidation in water, and model compounds for tannin and lignin tend to have higher Φ f than the bulk materials. , Insight into compositional differences of DIOHBA-dominated Fr6 and lignin-dominated Fr9 comes from examining the intensity-weighted emission wavelength (λemavg) of these fractions and bulk SRNOM as a function of pH (Figure S10). The pH-dependent λemavg of bulk SRNOM more closely matches the behavior of Fr9 at λ ex shorter than ∼290 nm wherein a slight red shift is observed across the carboxyl pH range (∼2 to 6). In contrast, Fr6 λemavg values are comparatively constant across all pH at these excitation wavelengths. At λex of 300 and 310 nm, there is little change in λemavg for bulk SRNOM and Fr6, while Fr9 emission red shifts from pH 2 to 6. At UVC excitation wavelengths, both bulk SRNOM and Fr6 emission spectra blue shift with increasing pH, while average emission for Fr9 remains relatively constant. Thus, the pH-dependent behavior of bulk SRNOM more closely resembles DIOHBA-dominant Fr6 at λex longer than ∼300 nm and more closely resembles Σ6-dominant Fr9 at λ ex shorter than ∼300 nm. These observations show how different structures in Fr6 and Fr9 contribute to SRNOM fluorescence behavior at different excitation wavelengths. Given the strong relationships observed here between PFFE fractions and DIOHBA, we suggest that future studies DIOHBA quantification (post cupric oxidation) with DOM optical measurements. With additional validation, variation in spectral shape or in emission maxima under varying conditions such as pH could serve as a proxy for DOM fluorophores linked to DIOHBA precursors. Structural and Compositional Controls on 1O2 and • OH Formation Correlations of 1O2 and •OH formation with optical surrogates indicate that SRNOM fractions behave more consistently with the broader DOM literature, including ultrafiltration studies, than the above discussion would predict. For example, although E2:E3 and Φ f do not vary systematically with electrophoretic mobility, ΦΔ and ΦOH are both well correlated to these parameters, consistent with a pattern frequently observed for ultrafiltered DOM, ,,, whole-water samples, , and ozone-treated DOM (). This suggests an overlap between the structural features governing E2:E3 and the excited states that are precursors to Φ f , ΦΔ, and ΦOH. Dalrymple et al. suggested that E2:E3 is inversely related to DOM’s ability to produce charge-transfer states. These states would serve as a radiationless deactivation pathway for 1DOM*, explaining the decrease in Φ f , ΦΔ, and ΦOH with decreasing E2:E3. Additional research such as measuring spectra in varying solvent polarities or sodium borohydride reduction could further delineate the role of charge transfer in PFFE fractions. Regardless of the mechanism, these results indicate that the photophysical processes funneling excited states away from fluorescence are also funneling excited states away from triplet DOM (1O2 precursor) and •OH precursors. Environmental Implications In this work, we have shown that separation of SRNOM by PFFE results in structurally distinct fractions with optical and photochemical properties different from those of apparent molecular weight fractions generated by ultrafiltration. Moreover, the relationships between absorbance-based optical surrogates and compositional parameters (i.e., molecular weight and aromaticity) in some cases go against classical interpretations in the literature. The molecular size and aromaticity of Fr6 through Fr9 vary substantially, yet the E2:E3 values of these fractions are highly similar and their SUVA254 values do not vary systematically. Practically, this suggests that caution should be used when interpreting changes in optical properties in relation to structure for natural or engineered processes that fractionate DOM on the basis of charge density. At a fundamental level, these results provide evidence that additional structural descriptors (beyond aromaticity and molecular size) are necessary to fully understand E2:E3 and perhaps SUVA254. We have also shown that DIOHBA precursors are strongly correlated with DOM optical properties, especially fluorescence. DIOHBA is believed to be linked to tannins and flavonoids and not lignin. Although tannins have been hypothesized to be a source of chromophoric DOM, − there is, to our knowledge, very little direct evidence for their role, in stark contrast to the multitude of studies showing correlations between absorbance and lignin phenol concentrations. , Future studies should employ cupric oxide oxidation and DIOHBA quantification in environmental field studies as well as mechanistic investigations to further understand the role of DIOHBA precursors in DOM optical properties. Despite the atypical behavior of absorbance-based surrogates for PFFE fractions, correlations between photochemical quantum yields and optical surrogates (i.e., E2:E3 and Φ f ) hold across the fractions in a manner typical of ultrafiltration-generated fractions. This points to E2:E3 being a predictor of reactivity that cannot be ascribed to something as simple as molecular weight alone (e.g., the abundance and nature of interchromophore interactions). On the basis of its unique results compared to ultrafiltration, PFFE has shown utility in separating DOM, which remains an outstanding analytical challenge to characterize. The in-depth analysis of PFFE-generated fractions presented herein is an exemplary top-down approach for tackling this challenge. , PFFE offers the fundamental advantage of avoiding interactions with a solid-phase matrix. In addition, electrophoretic separation is based not only on size but also on charge density, which more directly relates the separation process to DOM structure and, therefore, the mechanistic role of DOM in natural and engineered aquatic systems. This is an advantage over ultrafiltration methods, which function primarily based on size. Drawbacks of PFFE are that the separation device itself is an instrument with higher cost than the materials typically used for DOM fractionation, such as ultrafiltration cells and membranes or solid-phase extraction cartridges. Other disadvantages are the need to separate the resulting fractions from the aqueous buffer employed in the separation medium and the comparatively low throughput compared with solid-phase extraction or ultrafiltration. It is worth considering PFFE for the separation of additional DOM isolates readily available in high quantities (e.g., IHSS isolates). For example, it would be of interest to see if the trends observed for SRNOM are different for fulvic acid extracts (which have higher carboxyl group abundance than SRNOM). Furthermore, investigating the behavior of PFFE fractions in mineral sorption, microbial processing, organic pollutant and metal interactions, and photomineralization is another productive research avenue, given that DOM size and charge are known to influence each of these processes. For example, one could quantify the metal-binding constants of PFFE fractions and compare them to the bulk sample with a conservative mixing model. Finally, DOM is one of many complex carbon-based mixtures for which structure–reactivity relationships remain elusive, suggesting that PFFE may also prove to be a powerful separation tool beyond the DOM field. The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acs.est.6c07200. - Experimental and analytical methods, details regarding the analysis of data, and additional data and figures (PDF) The authors declare no competing financial interest. Published as part of Environmental Science & Technology special issue “60th Anniversary of Environmental Science and Technology”.