Online High-Performance Size Exclusion Chromatography−Nuclear

Dec 23, 2009 - ... parameters, but the end result is unparalleled detection capabilities with only a ..... dissolved organic matter in the ocean Natur...
0 downloads 0 Views 1MB Size
Environ. Sci. Technol. 2010, 44, 624–630

Online High-Performance Size Exclusion Chromatography-Nuclear Magnetic Resonance for the Characterization of Dissolved Organic Matter GWEN C. WOODS,† MYRNA J. SIMPSON,† BRIAN P. KELLEHER,‡ MARGARET MCCAUL,‡ WILLIAM L. KINGERY,§ AND ´ J . S I M P S O N * ,† ANDRE Department of Chemistry, University of Toronto, Scarborough Campus, Toronto, Ontario, Canada M1C 1A, School of Chemical Sciences, Dublin City University, Dublin 9, Ireland, and Department of Plant and Soil Sciences, Mississippi State University, Mississippi 39762

Received October 11, 2009. Revised manuscript received November 25, 2009. Accepted December 5, 2009.

The substantial heterogeneity of dissolved organic matter (DOM) inhibits detailed chromatographic analysis with conventional detectors as little structural information can be obtained in the presence of extensive coelution. Here we examine the direct hyphenation of high-performance size exclusion chromatography (HPSEC) with nuclear magnetic resonance (NMR) spectroscopy to determine how size-distinguished fractions differ in composition. The results support the applicability of using HPSEC to generate more homogeneous fractions of DOM prior to NMR analysis and demonstrate that structure is significantly altered with size. The largest fractions are enriched in carbohydrate- and aromatic-type structures. The midsized material is substantial and is representative of carboxylrich alicyclic molecules (CRAMs). The smallest material has strongsignaturesofmaterialderivedfromlinearterpenoids(MDLT). Both CRAMs and MDLT have been recently hypothesized as major components of DOM, and detection by HPSEC-NMR confirms their existence as unique and separable entities. This preliminary work focuses on NMR hyphenation to HPSEC due to widespread use of HPSEC to characterize DOM. Online hyphenation is useful not only for time-efficient analysis of DOM but also for that of other highly complex samples such as those found in many environmental analyses.

Introduction Dissolved organic matter (DOM) constitutes one of the largest reservoirs of actively cycling organic carbon on Earth (1). As CO2 is a primary product of DOM mineralization, an intimate link exists between this dissolved pool of carbon and the atmosphere. It has been suggested that enhanced oxidation of marine DOM brought the planet out of a series of severe glaciations and into the Cambrian explosion of life (2). * Corresponding author phone: (416) 287-7547; fax: (416) 2877279; e-mail: [email protected]. † University of Toronto. ‡ Dublin City University. § Mississippi State University. 624

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010

Understanding DOM mineralization processes is thus critical to assessing environmental issues such as climate change, but despite considerable research designated to the characterization of DOM, the individual molecular structures, and to some extent the major structural components, are still not well understood. Access to such molecular-level information would provide insight into origins of water masses, clarity of DOM interactions with environmental contaminants, and a better understanding of DOM significance in the global carbon cycle. Through the use of multidimensional NMR, predictions, and simulations, our research group has been able to describe the major categories of components present in freshwater DOM (3) but have concluded that new analytical techniques will be required to elucidate exact structures. Originating at least in part from the degradation of biomolecules, DOM is comprised of a myriad of degraded and reworked products and is subsequently one of the most complex natural mixtures. In laboratory analyses, excessive signal overlap with most detectors limits the amount of detailed information that can be obtained. In attempting to decipher information, researchers frequently employ multiple techniques and instruments, either hyphenated into a single analytical run or conducted offline. HPLC is frequently hyphenated to mass spectrometry (MS) (4-6). Ultra-highresolution Fourier transform ion cyclotron resonance MS (FT-ICR-MS) can be used to calculate the molecular formula of thousands of constituents in DOM and is complemented by offline NMR analyses (7). Spectral information from UV-vis and fluorescence is frequently used to characterize sources of DOM, and these techniques have been used in combination with high-performance size exclusion chromatography (HPSEC) (8). Despite such multifaceted approaches, detailed structures remain indiscernible. Even when using advanced multidimensional NMR experiments that have successfully unraveled components in soil organic matter (9-13), the lack of appropriate standards for DOM and extensive spectral overlap result in only basic information of major structural components (3). Individual structures themselves remain elusive and warrant further research. NMR is a powerful detector for unknown constituents in mixtures, but hyphenated NMR is a fitting solution when faced with the unprecedented complexity of DOM. If suitable separation can be obtained with an instrument online with NMR, more detailed information can be obtained than with offline analyses. Online HPLC-NMR finds application in such areas as the analysis of pharmaceuticals, food products, and contaminants, with the intent of resolving mixtures of metabolites and degradation products (14-17). Sample mixtures are simplified chromatographically and directly detected by NMR without extensive laboratory procedures. The greatest advantage is simply that samples are analyzed quickly and frequent data acquisition is possible (i.e., “slices” can be taken more often than in comparable offline techniques, translating into less coelution per fraction). The drawbacks to the online system include sample “dilution” during HPLC separation, leading to sensitivity issues with NMR detection as well as problems with solvent signal suppression of nondeuterated HPLC solvents. The consequence of these drawbacks is that compromises must be made when designing experimental parameters, but the end result is unparalleled detection capabilities with only a single chromatographic run. Online HPLC-NMR has found widespread use in pharmaceutical screening but rarely in environmental applications (17), with only a preliminary online analysis of DOM 10.1021/es903042s

 2010 American Chemical Society

Published on Web 12/23/2009

to date (18). The development of HPLC-NMR techniques is significant not only for DOM elucidation but also for analysis of other complex environmental samples such as organic matter in soils, sediments, and the atmosphere. Here we examine the structural components of DOM with NMR hyphenated to HPSEC (a commonly used chromatographic technique with DOM). As such this work describes the successful hyphenation of HPSEC with NMR to examine the size-distinguished structural variability in DOM.

Materials and Methods Sample Collection and Preparation. Three DOM samples were used for this study. Nordic Reservoir natural organic matter (NRNOM) and Suwannee River natural organic matter (SRNOM) were purchased from IHSS. Both samples were isolated via reverse osmosis following 0.4 µm filtration (details on the IHSS Web site, http://ihss.gatech.edu/ihss2). The third sample was collected from a wetland in the Lynde Shores Conservation Area, Whitby, Ontario, in August of 2007. The Lynde Shores sample (LSDOM) was collected via pressure filtration with 0.45 µm PVDF membranes and isolated on DEAE cellulose. Samples were extracted under N2 with 0.1 M NaOH, ion exchanged to remove Na+, and lyophilized (18, 19). All samples were prepared in the HPSEC mobile phase, adjusted to pH 12 with NaOH, and syringe filtered (0.45 µm). HPSEC Separation. HPSEC separation and HPSEC-NMR analyses were conducted on an Agilent HP1100 HPLC system, equipped with a column heater, diode array detector (DAD), and fraction collector (Foxy Jr., ISCO) and controlled with Hystar software (Bruker), version 3.0. Readers should note that the DAD signal was in most cases swamped due to the large and concentrated injections required for NMR spectroscopy. As such the DAD data are not considered here. Future studies designed to use both the DAD and NMR as parallel detectors would need a software-controlled splitter and dilutor prior to DAD. Two columns, Ultrahydrogel 250 and 120 (Waters, rated pH 2-12), were used in series. Column performance was assessed daily using a mixture of poly(styrenesulfonic acid) standards, and no qualitative or quantitative changes were noted to occur over the duration of this study. Additionally, quantitative recovery of DOM was assessed by collecting all eluate after injections both with and without the columns, and 100.1 ( 2.3% recovery was noted on the basis of multiple injections. Although the column was calibrated with PSS standards, molecular weight estimates of DOM are not presented. The hydrodynamic radii as well as interactions with the stationary phase will differ between common commercial standards and humic substances as debated heavily in the literature (20). Appropriate calibration standards for DOM are difficult to assess pending better insight into molecular structures and aggregation of this complex material (21). HPSEC is used here as a means to size-separate DOM, and molecular weights are not presented due to concerns as to the accuracy of such data. An isocratic solvent system at 40 °C was used for HPSEC separations with an aqueous buffer comprised of 0.1 M NaCl and 0.03 M NH4Cl, adjusted to pH 11 with NH4OH. A similar buffer has been cited elsewhere (22), but a lower NaCl concentration was used here for purposes of NMR compatibility (which is particularly crucial for fraction collection where salts are concentrated during drying). Separation was not significantly affected by this reduction in salts. For directly coupled HPLC-NMR experiments, a 90:10 H2O/D2O version of the buffer was used. Fraction Collection. HPSEC-eluted fractions were acquired for comparison to online techniques. Fractions of the DOM were taken in 2 min intervals over the duration that DOM eluted from the HPSEC column. This process was repeated 30 times, and combined fractions were lyophilized

and stored for later analysis. Lyophilized samples were reconstituted with D2O and analyzed via 1H NMR. Solution 1H NMR. HPSEC-NMR analyses were achieved on a Bruker Avance 500 MHz spectrometer at 298 K. A dualtuned 1H-19F flow probe (120 µL) fitted with an actively shielded z-gradient was used for NMR detection. For continuous-flow analyses run at 0.5 mL/min, 16 scans were acquired, while at 0.05 mL/min 88 scans were acquired. Stopped-flow experiments were performed using 96 scans. Fraction-collected and whole DOM samples were further analyzed in D2O (with 10 µL of NaOD added to whole samples to ensure solubility) with a 5 mm, triple-resonance broadbanded inverse (TBI) probe using 128 scans. For all NMR experiments 16 384 time domain points were used with a recycle delay of 2 s. NOESY presaturation (pulse program NOESYPR1D) was used with a 400 µs mixing time to suppress the signal from the mobile phase (water at 4.7 ppm). During Fourier transform the residual water signal was reduced using a Gaussian function centered at the water frequency (∼4.7 ppm) and corresponding to a bandwidth in the transform spectrum of 0.7 ppm (23). Reduced contributions from water allow the profiles and projections from the DOM signal to be more reliably discerned. The disadvantage is that the region around the water appears artificially “smoothed” and is seen most clearly in samples with low S/N ratio (e.g., Figure 2, row C). Readers should recognize this as an artifact from the processing employed. Spectra were apodized through multiplication with an exponential decay corresponding to 1 Hz and processed using a zero filling factor of 2. All 1D spectra generated from online experiments were compiled into pseudo-2D NMR chromatograms (1D 1H NMR spectra along the x-axis and HPSEC elution volume on the y-axis) using Bruker Topspin (Bruker), version 2.1.

Results and Discussion Online Techniques. Hyphenated HPLC-NMR may be accomplished in either continuous-flow (also called on-flow) or stopped-flow mode, with advantages and disadvantages to both. Both methods include directly flowing eluate from the HPLC to the NMR flow cell. Continuous-flow is accomplished with the HPLC pump and NMR running simultaneously, whereas stopped-flow involves stopping the pump in intervals during which time NMR experiments are run on the static sample (controlled here via Hystar software and stopped for ∼10 min per fraction). Continuous-flow is limited in the number of scans per spectrum (translating into reduced sensitivity) due to limited time in the NMR cell. The process of continuously flowing also causes inhomogeneities within the magnetic field which can lead to reduced spectral resolution. In continuous-flow, however, chromatographic separation is not interrupted and the influence of diffusion is much less compared to that in stopped-flow. Thus, despite drawbacks, continuous-flow is useful for fast screening of samples and/or comparison to stopped-flow. In contrast, stopped-flow allows the sample to be shimmed, generating a more homogeneous field inside the NMR cell, as well as the possibility to collect more scans per slice, hence improving detection limits (15). Due to the potential for complementary information and/or for comparison, we examined online analyses in both stopped- and continuous-flow. Stoppedflow experiments were run in time-slice mode, which essentially stops flow when the detection cell is refreshed with new material, ensuring that all material is detected. With continuous-flow, a limited number of scans can be collected, and thus, we examined the effect of a reduced flow by dropping the rate by 10-fold (i.e., run at both 0.5 and 0.05 mL/min). Readers should keep in mind throughout the remainder of the text that compromised chromatography is a necessity of successful online HPLC-NMR. The reduced VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

625

FIGURE 1. Comparison of HPSEC-NMR techniques (SRNOM, 100 mg/mL, 100 µL injection). Column I (rows A-C): pseudo-2D chromatograms from online analyses, elution profiles along the y-axes. Column I (row D): sum profiles from selected NMR regions from column I, row A (indicated by dashed lines and colored regions): red, aromatics, 6.5-7.8 ppm; green, carbohydrates, 3.2-4.5 ppm; blue, carboxyl-rich alicyclic molecules (CRAMs), 1.6-3.2 ppm; purple, material derived from linear terpenoids (MDLT), 0.6-1.6 ppm (as highlighted in column I, row A). These profiles are discussed later in the paper but included here so that the reader can visualize how the sum profiles are created from 2D HPLC-NMR data sets. Rows A, B, and C: NMR spectra from stopped-flow, slow continuous-flow, and fast continuous-flow, respectively. Row D (columns II-IV): NMR spectra from offline fraction collection. Columns II, III, and IV: NMR spectra of material eluted at 50%, 75%, and 95% of the total sample elution volume. Readers should be aware that under continuous-flow conditions (rows B and C) the sample is only in the NMR cell for a finite amount of time; hence, the number of scans cannot be increased. However, in the case of stopped-flow (row A) and fraction collection (row D) the number of scans can be increased substantially, permitting the detection of components at much lower concentrations. The asterisk indicates carbohydrates; methoxyl from lignin also contributes to this region. sensitivity of NMR compared to traditional HPLC detectors requires that conditions such as large sample load and slower flow are used to achieve sufficient material within the NMR flow cell. HPLC is essentially a means of separation prior to NMR analysis, and optimal separation, although desirable, is not necessary to obtain structural information from NMR (17). Figure 1 illustrates the chromatographic variability of conducting the above-mentioned modes on SRNOM. Rows A, B, and C depict stopped-flow, slow continuous-flow, and fast continuous-flow, respectively. For these rows, column I illustrates 2D NMR chromatograms. The y-axes of these 2D chromatograms contain profiles of SRNOM elution with mobile-phase volume and are constructed from the sum of NMR signals (excluding solvent signal from water). Examination of these y-axes illustrates that continuous-flow eluted material over larger volumes than stopped-flow trials and that continuous-flow resulted in later apex elution (see also Table 1). Theoretically it would be expected that peaks in stopped-flow would be broader due to the substantial diffusion that is permitted to occur from both motionless conditions and considerably longer run times (e.g., here ∼17 h for stopped-flow vs ∼4 h for slow and ∼1 h for fast continuous-flow). Diffusion is also likely to promote DOM disaggregation, and thus, stopped-flow should result in smaller material at larger retention volumes. Stopped-flow instead generated a sharper peak and earlier apex than continuous-flow, contrary to what was expected. A plausible explanation for these elution differences is the effects of sample viscosity. As noted to occur in preparative HPLC, “viscous fingering” occurs with large sample injections, 626

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010

TABLE 1. Parameters and Apex Retention Volume for HPSEC-NMR Analyses method

flow rate (mL/min)

sample

stopped-flow continuous-flow continuous-flow fraction collection stopped-flow stopped-flow stopped-flow stopped-flow

0.5 0.05 0.5 0.5 0.5 0.5 0.5 0.5

SRNOM SRNOM SRNOM SRNOM SRNOM SRNOM NRNOM LSNOM

concn apex retention (mg/mL) vol (mL) 100 100 100 100 20 5 100 100

14.25 16.83 16.08 16.14 13.88 13.63 14.25 14.63

causes a delay in retention volume, and is known to generate false banding after apex material has passed (24, 25). In simple terms, viscous fingering occurs when a large plug of sample is injected into a less viscous mobile phase and is essentially “run over” by the mobile phase, resulting in poor separation, broadened peaks, and false “bands” at the tail (which was noted to occur with samples run in continuous-flow mode). Stopped-flow, permitting abundant diffusion, would be expected to generate ample time for the sample edges to mix with the mobile phase and therefore generate fewer viscous fingering effects. The early elution and tighter sample bands produced by stopped-flow indeed suggest that viscous fingering did not occur nearly as much as with continuousflow conditions. To further test the role viscosity might play, much less concentrated/viscous samples were also run and found to elute earlier than more concentrated samples (Table

1). This finding further suggests that viscosity acts to delay and broaden eluting material in the more concentrated samples. An alternative theory might be that late elution is the result of a reduction in conformational size with increasing concentration, but NMR research has shown that not to be the case with SRNOM (21), and thus, it is not likely to be the case here. Further comparison of the two continuous-flow experiments reveals that slow flow resulted in broader and later elution than fast flow trials (Figure 1 (column I, row C, and column I, row B); see also Table 1), suggesting that both diffusion and viscosity affect samples run under slow continuous-flow conditions. Along with differences in chromatographic profiles, the NMR data provide information as to the separation of structural entities in the fast, slow, and stopped-flow experiments. In all cases the NMR spectrum for the major fraction eluting at the apex is dominated by carboxyl-rich alicyclic molecules (CRAMs), signified by the dominant signal from ∼1.6 to 3.2 ppm (Figure 1, column II, rows A-C). By the latest fractions (95% elution), however, CRAMs are depleted in the stopped-flow and fast continuous-flow experiments but are still prominent in the slow continuous-flow. The reduced variability apparent in the slow trial compared to the others suggests that slow continuous-flow provides the least effective separation. Offline fraction collection of SRNOM is illustrated in Figure 1 (row D, columns II-IV). These spectra are similar to the online spectra, but differences exist that are likely due to the 8× greater sample volume collected for the offline fractions (1.00 mL vs 0.125 mL elution volume). These offline spectra represent 240 times more material (8× larger fractions and 30 HPLC runs), 240 times higher salt content, longer NMR experiments, and material freeze-dried following fraction collection. Excess repetitions of HPLC runs were conducted to acquire more sample material from the most dilute front and tail fractions in hopes of obtaining information unobtainable from online analyses. No structural information was gained that was not discernible in online experiments. Fraction collection is, however, more readily available to researchers, and depending on the type of separation or research goals, fraction collection may in many instances be more practical than online HPLC-NMR. The goal of research presented here is to develop online techniques that are in many ways complementary to offline techniques as both have advantages and disadvantages. For example, the online experiments result in a degree of carryover within the NMR flow cell (dependent upon the volume and design of the flow cell and connections). Fraction collection, in turn, is limited in the sense that increasing the numbers of fractions becomes an issue of resources and labor. From our research experience, 100 fractions could be accomplished within ∼1 h (continuousflow) and ∼17 h (stopped-flow) while online. The comparable offline technique took weeks to obtain and run 25 fractions. Improved HPLC separations in the future (with potentially hundreds to thousands of peaks) would make fraction collection extremely laborious to achieve the sort of resolution (i.e., minute fractions) capable with online HPLC-NMR. Additionally, salts from buffers are concentrated while fractions are collected, which can make accurate matching and tuning difficult for NMR. Of the online methods compared in this study, stoppedflow provided the best separation (due to reduced viscous fingering) and has the potential for enhanced sensitivity over other methods (as one can increase the number of scans). Though diffusion could be problematic, evidence suggests that more time for equilibration between the sample and solvent system in stopped-flow reduces undesirable viscosity effects which at least in this study outweigh negative effects from diffusion. Environmental samples are frequently very heterogeneous and due to the vast numbers of constituents

present in very small quantities require large HPLC injections. The effects of viscosity are therefore an important consideration of online HPLC-NMR with environmental applications. Effects of Concentration. Concentrations as low as 5 mg/ mL are known to result in the aggregation of DOM into conformationally larger material (21). Thus, it is important to investigate whether increased concentration leads to reduced separation, either from sample aggregation or from column overload, which in turn leads to a loss of information via NMR detection. Figure 2 illustrates strong similarities in the eluting material between the three tested concentrations (5, 20, and 100 mg/mL). At 5 mg/mL (Figure 2, row C), the early- and late-eluting material produced spectra with low S/N ratio and sample peaks are overshadowed by large artifacts resulting from water suppression during data processing (see the Materials and Methods). At such low concentrations, distortions from the Gaussian filter used to reduce the water signal are emphasized and weak sample signals are difficult to discern. This is best exemplified by the region at ∼3.7 ppm adjacent to the water and highlighted with an arrow in Figure 2 (column I, row A). This signal arises from a combination of lignin methoxyl and carbohydrates (21), is clearly seen in the more concentrated sample, is likely present in the 20 mg/mL separation, and is not discernible in the 5 mg/mL separation. Figure 2 demonstrates that eluting material appears similar at all tested concentrations but also illustrates the importance of using large sample quantities to obtain sufficient NMR signal for more detailed and robust identifications. Structural Information. Recent work on the structural identification of DOM has identified material thought to be derived from cyclic terpenoids known as CRAMs (3, 26) and material derived from linear terpenoids (MDLT) (3) as major components of DOM. Recent 1H and 13C NMR research on DOM and humic acids provides evidence of aromatics and carbohydrates as major constituents of conformationally larger DOM, while CRAMs and MDLT generally constitute smaller material (21, 27, 28). For the analyses of spectra described in this study, constituents present in 1D 1H NMR spectra were assigned the following chemical shifts: aromatics (6.5-7.8 ppm), carbohydrates (3.2-4.5 ppm), CRAMs (1.6-3.2 ppm), and MDLT (0.6-1.6 ppm). The 1D slices from all experiments reveal that the bulk of carbohydrate- and aromatic-type material eluted in the early slices (Figures 2, column I, and 3, column II). The apex material was in turn characterized by a strong CRAM signature (Figures 2, column II, and 3, column III), and the late-eluting material was characterized by strong signals from the MDLT region as well as sharp peaks in the carbohydrate region suggestive of small sugars (Figure 1, column IV). To analyze the elution order of major components present in each sample, 1D profiles were generated from selected regions of the pseudo-2D NMR chromatograms using Advanced Chemistry Development (ACD laboratories) Spectrum Manager (version 11.0). As indicated by the colored regions in Figure 1 (column I, row A), profiles of the aromatics, carbohydrates, CRAMs, and MDLT were generated by compiling the sums of these regions. An example of the resulting profiles is illustrated in the bottom left corner of Figure 1 (column I, row D). The insets in Figure 3 illustrate the component order of elution for three DOM samples (SRNOM, NRNOM, and LSNOM) and are discussed in detail below. The maxima of the four material types eluted independently and with the general elution volumes of aromatics < carbohydrates < CRAMs < MDLT. This trend was typical for all experiments and samples with the exception of LSNOM. LSNOM did not separate well as indicated by an elution order of aromatics ) carbohydrates ) CRAMs < MDLT (further discussion below). VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

627

FIGURE 2. Comparison at varying concentrations (stopped-flow, SRNOM, 100 µL injection). Rows A, B, and C: NMR spectra of 100, 20, and 5 mg/mL, respectively. Columns I, II, and III: material eluted at 25%, 50%, and 75% of the total sample elution volume. The arrow is discussed in the text.

FIGURE 3. Comparison of DOM samples (stopped-flow, 100 mg/mL, 100 µL injection). Rows A, B, and C: SRNOM, NRNOM, and LSNOM, respectively. Column I: whole samples run in a traditional 5 mm NMR tube. Columns II, III, and IV: material eluted at 25%, 50%, and 75% of the total sample elution volume (100 mg/mL, stopped-flow). The insets in column I (rows A-C) illustrate elution profiles from HPSEC-NMR; the axes are the retention volume (mL). (Refer to the Figure 1 caption for color and chemical shift assignments.) The asterisk indicates carbohydrates; methoxyl from lignin also contributes to this region. Variability of Environmental Samples. To assess environmental variability, DOM samples isolated from three very different bodies of water were compared. Figure 3 illustrates whole DOM samples (column I) and HPSEC-eluted material at 25%, 50%, and 75% of the total sample elution volume (columns II-IV). The whole sample NMR spectra of all three samples consist of similar broad NMR profiles (characteristic of DOM). The profiles of size-distinguished fractions, how628

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010

ever, reveal clear molecular differences between samples. This variability is key to elucidating and understanding the origins of the organic material as well as the physical and biological processes that act to degrade and mineralize DOM. The fractions displayed in Figure 3 are merely representative of the types of material eluted prior to, during, and after the chromatographic apex of each sample; these fractions do not add up to make the whole of the bulk samples but do

illustrate the variability in composition between samples. With the largest material, the NRNOM sample has the strongest carbohydrate contribution (methoxyl from lignin may also resonate here) compared to the other samples. With the midsized fraction, NRNOM has the weakest contribution of CRAMs, while the smallest fraction has the strongest MDLT component. The SRNOM and LSNOM samples appear most similar to each other by HPSEC-NMR, but differences not apparent from the total NMR spectra are apparent. For example, the SRNOM contains more MDLT in the smaller fraction than the LSNOM, despite the fact that LSNOM contains more MDLT as a whole. Further analysis of apex retention volumes (Table 1) reveals that the LSNOM eluted later than the other samples, indicative of smaller material (or greater viscosity). The SRNOM and NRNOM samples, in turn, eluted earlier than the LSNOM sample but with retention volumes equal to one another. Finally, the profiles of aromatics, carbohydrates, CRAMs, and MDLT for all three samples are illustrated in the insets of Figure 3, column I. From these insets it is apparent that the LSNOM sample separated the least, with nearly all component maxima coeluting while the SRNOM and NRNOM samples had distinct elution volumes for the four major groups. The inset in column I (row B) further illustrates that the NRNOM sample had the most successful chromatographic separation with a second major peak after apex elution. This second peak is rich in MDLT-type material and is structurally very distinct from the earlier material (hence the higher contributions from the purple trace in the second peak). The variation in HPSEC elution could be the result of a number of factors. The samples originated from diverse water sources and were collected in different years with varying methods of isolation. The SRNOM and NRNOM samples originated from a blackwater river and drinking reservoir, respectively, and were both concentrated via reverse osmosis (in 1999 and 1997). The LSNOM sample was collected from a productive marshland in 2007 and concentrated on DEAE cellulose followed by alkaline extraction. All samples were stored in an identical fashion (freeze-dried, sealed, and in the dark) but factors such as origin, parent vegetation, microbial inputs, degradation processes, mentioned isolation techniques, and age could contribute to the variability in HPSEC separations. The LSNOM sample was notably the most difficult to separate into chemically distinct fractions, while the NRNOM sample was the easiest and even produced a second major chromatographic peak (refer to the inset in Figure 3, column I, row B). More research is necessary to determine the reasons for elution variability between samples. These differences do, however, demonstrate that while all samples are similar in terms of their 1D NMR (column I), HPSEC-NMR provides additional information on the subcomponents within DOM which may prove useful in understanding how components vary temporally and spatially and their role in the formation and persistence of DOM in the environment. Despite the compromises necessary for online HPSECNMR, analyses of slices eluted at regular intervals indicate variations in material from the largest to the smallest sized fractions, and material generally eluted in the order of (1) aromatics, (2) carbohydrates, (3) CRAMs, and (4) MDLT. A noteworthy phenomenon is the strong NMR signal of CRAMtype material during apex elution (i.e., the biggest fraction of the sample). This supports the existence of CRAMs as a separable and unique entity in DOM that has only recently been proposed in the literature (26). Similarly, MDLT dominates the smaller size fraction, supporting the hypothesis that MDLT also contributes significantly to DOM and is a unique and separable entity. Comparison of whole sample spectra to size-separated fractions suggests that DOM separates into structurally different material that varies

between samples. It has been previously proposed that while biomolecules undergo similar biogeochemical processes, it seems likely that there are similarities between samples that have undergone significant degradation (26). Whether the differences found in the separations here are caused more by source material or degradation warrants further research. With this initial HPSEC-NMR hyphenation, we have shown that online HPLC-NMR can be used for the structural analyses of DOM. This technique is useful and perhaps advantageous over other analytical techniques in that samples can be analyzed very quickly and that large quantities of data and detailed information are made available in a short period of time. Of particular interest for further research is to examine stationary phases better suited for the separation of DOM constituents. Improved chromatography online with NMR has the potential to provide unsurpassed structural characterization of DOM and is a technique applicable to analyses of other complex natural samples.

Acknowledgments We thank the Natural Science and Engineering Research Council of Canada (Discovery Grant, A.J.S.), the International Polar Year (IPY), the Science Foundation of Ireland (Grant GEOF509), and the Irish Environmental Protection Agency (STRIVE program) for providing funding. We also thank the government of Ontario for providing an Early Researcher Award (A.J.S.).

Literature Cited (1) Amon, R. M. W.; Benner, R. Rapid-cycling of high-molecularweight dissolved organic matter in the ocean. Nature 1994, 369, 549–552. (2) Peltier, W. R.; Liu, Y. G.; Crowley, J. W. Snowball Earth prevention by dissolved organic carbon remineralization. Nature 2007, 450, 813–U811. (3) Lam, B.; Baer, A.; Alaee, M.; Lefebvre, B.; Moser, A.; Williams, A.; Simpson, A. J. Major structural components in freshwater dissolved organic matter. Environ. Sci. Technol. 2007, 41, 8240– 8247. (4) Dittmar, T.; Whitehead, K.; Minor, E. C.; Koch, B. P. Tracing terrigenous dissolved organic matter and its photochemical decay in the ocean by using liquid chromatography/mass spectrometry. Mar. Chem. 2007, 107, 378–387. (5) Peuravuori, J.; Bursakova, P.; Pihlaja, K. ESI-MS analyses of lake dissolved organic matter in light of supramolecular assembly. Anal. Bioanal. Chem. 2007, 389, 1559–1568. (6) Reemtsma, T. The use of liquid chromatography-atmospheric pressure ionization-mass spectrometry in water analysissPart I: Achievements. TrAC, Trends Anal. Chem. 2001, 20, 500–517. (7) Dittmar, T.; Paeng, J. A heat-induced molecular signature in marine dissolved organic matter. Nat. Geosci. 2009, 2, 175–179. (8) Wu, F. C.; Evans, R. D.; Dillon, P. J. Separation and characterization of NOM by high-performance liquid chromatography and on-line three-dimensional excitation emission matrix fluorescence detection. Environ. Sci. Technol. 2003, 37, 3687– 3693. (9) Simpson, A. J.; Simpson, M. J.; Smith, E.; Kelleher, B. P. Microbially derived inputs to soil organic matter: Are current estimates too low? Environ. Sci. Technol. 2007, 41, 8070–8076. (10) Simpson, A. J.; Song, G. X.; Smith, E.; Lam, B.; Novotny, E. H.; Hayes, M. H. B. Unraveling the structural components of soil humin by use of solution-state nuclear magnetic resonance spectroscopy. Environ. Sci. Technol. 2007, 41, 876–883. (11) Kelleher, B. P.; Simpson, A. J. Humic substances in soils: Are they really chemically distinct? Environ. Sci. Technol. 2006, 40, 4605–4611. (12) Kelleher, B. P.; Simpson, M. J.; Simpson, A. J. Assessing the fate and transformation of plant residues in the terrestrial environment using HR-MAS NMR spectroscopy. Geochim. Cosmochim. Acta 2006, 70, 4080–4094. (13) Simpson, A. J.; Kingery, W. L.; Hatcher, P. G. The identification of plant derived structures in humic materials using threedimensional NMR spectroscopy. Environ. Sci. Technol. 2003, 37, 337–342. (14) Kitayama, T.; Ute, K. On-line SEC-NMR. In Modern Magnetic Resonance; Webb, G. A., Ed.; Springer: Dordrecht, The Netherlands, 2008; pp 399-405. VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

629

(15) Albert, K., Ed. On-Line LC-NMR and Related Techniques; John Wiley & Sons: New York, 2002. (16) Duarte, I. F.; Godejohann, M.; Braumann, U.; Spraul, M.; Gil, A. M. Application of NMR spectroscopy and LC-NMR/MS to the identification of carbohydrates in beer. J. Agric. Food Chem. 2003, 51, 4847–4852. (17) Levsen, K.; Preiss, A.; Codejohann, M. Application of highperformance liquid chromatography coupled to nuclear magnetic resonance and high-performance liquid chromatography coupled to mass spectrometry to complex environmental samples. TrAC, Trends Anal. Chem. 2000, 19, 27–48. (18) Simpson, A. J.; Tseng, L. H.; Simpson, M. J.; Spraul, M.; Braumann, U.; Kingery, W. L.; Kelleher, B. P.; Hayes, M. H. B. The application of LC-NMR and LC-SPE-NMR to compositional studies of natural organic matter. Analyst 2004, 129, 1216–1222. (19) Lam, B.; Simpson, A. J. Passive sampler for dissolved organic matter in freshwater environments. Anal. Chem. 2006, 78, 8194– 8199. (20) Perminova, I. V. Size exclusion chromatography of humic substances: Complexities of data interpretation attributable to non-size exclusion effects. Soil Sci. 1999, 164, 834–840. (21) Lam, B.; Simpson, A. J. Investigating aggregation in Suwannee River (USA) dissolved organic matter using diffusion-ordered nuclear magnetic resonance spectroscopy. Environ. Toxicol. Chem. 2009, 28, 931–939. (22) Samburova, V.; Zenobi, R.; Kalberer, M. Characterization of high molecular weight compounds in urban atmospheric particles. Atmos. Chem. Phys. 2005, 5, 2163–2170.

630

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010

(23) Marion, D.; Ikura, M.; Bax, A. Improved solvent suppression in one-dimensional and two-dimensional NMR-spectra by convolution of time-domain data. J. Magn. Reson. 1989, 84, 425– 430. (24) Cherrak, D.; Guernet, E.; Cardot, P.; Herrenknecht, C.; Czok, M. Viscous fingering: a systematic study of viscosity effects in methanol-isopropanol systems. Chromatographia 1997, 46, 647–654. (25) Mayfield, K. J.; Shalliker, R. A.; Catchpoole, H. J.; Sweeney, A. P.; Wong, V.; Guiochon, G. Viscous fingering induced flow instability in multidimensional liquid chromatography. J. Chromatogr., A 2005, 1080, 124–131. (26) Hertkorn, N.; Benner, R.; Frommberger, M.; Schmitt-Kopplin, P.; Witt, M.; Kaiser, K.; Kettrup, A.; Hedges, J. I. Characterization of a major refractory component of marine dissolved organic matter. Geochim. Cosmochim. Acta 2006, 70, 2990–3010. (27) Conte, P.; Spaccini, R.; Smejkalova, D.; Nebbioso, A.; Piccolo, A. Spectroscopic and conformational properties of size-fractions separated from a lignite humic acid. Chemosphere 2007, 69, 1032–1039. (28) Conte, P.; Spaccini, R.; Piccolo, A. Advanced CPMAS-13C NMR techniques for molecular characterization of size-separated fractions from a soil humic acid. Anal. Bioanal. Chem. 2006, 386, 382–390.

ES903042S