Determination of molecular formulas of natural organic matter molecules by (ultra-) high-resolution mass spectrometry

Journal of Chromatography A, 1216 (2009) 3687–3701

Contents lists available at ScienceDirect

Journal of Chromatography A journal homepage: www.elsevier.com/locate/chroma

Review

Determination of molecular formulas of natural organic matter molecules by (ultra-) high-resolution mass spectrometry Status and needs Thorsten Reemtsma Technical University of Berlin, Department of Water Quality Control, Sekr KF 4, Strasse des 17 Juni 135, 10623 Berlin, Germany

a r t i c l e

i n f o

Article history: Received 19 September 2008 Received in revised form 24 January 2009 Accepted 12 February 2009 Available online 21 February 2009 Keywords: Natural organic matter Fulvic acids Humic acids Mass spectrometry Extraction Liquid chromatography Ozonation

a b s t r a c t Electrospray ionization (ESI) combined with ultra-high-resolution mass spectrometry on a Fourier transform ion cyclotron resonance mass spectrometer has been shown to be a very powerful tool for the analysis of fulvic and humic acids and of natural organic matter (NOM) at the molecular level. With this technique thousands of ions can be separated from each other and their m/z ratio determined with sufficient accuracy to allow molecular formula calculation. Organic biogeochemistry, water chemistry, and atmospheric chemistry greatly benefit from this technique. Methodical aspects concerning the application of Fourier transform ion cyclotron resonance mass spectrometry (FTICR-MS) to NOM isolated from surface water, groundwater, marine waters, and soils as well as from secondary organic aerosol in the atmospheric are reviewed. Enrichment of NOM and its chromatographic separation as well as possible influences of the ionization process on the appearance of the mass spectra are discussed. These steps of the analytical process require more systematic investigations. A basic drawback, however, is the lack of well defined single reference compounds of NOM or fulvic acids. Approaches of molecular formula calculation from the mass spectrometric data are reviewed and available graphical presentation methods are summarized. Finally, unsolved issues that limit the quality of data generated by FTICR-MS analysis of NOM are elaborated. It is concluded that further development in NOM enrichment and chromatographic separation is required and that tools for data analysis, data comparison and data visualization ought to be improved to make full use of FTICR-MS in NOM analysis. © 2009 Elsevier B.V. All rights reserved.

Contents 1. 2.

3. 4.

5.

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.1. Ionization and modifiers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2.2. Mass analyzers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Molecular formula calculation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Graphical presentation and exploitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.1. Reconstructed spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.2. Van Krevelen diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.3. Kendrick diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.4. C versus mass diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.5. Other diagrams . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.1. Biogeochemistry . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.2. Investigation of transformation reactions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.3. Nitrogenous molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.4. Sulfur containing molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5.5. Phosphorus containing molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

E-mail address: [email protected]. 0021-9673/$ – see front matter © 2009 Elsevier B.V. All rights reserved. doi:10.1016/j.chroma.2009.02.033

3688 3689 3689 3689 3690 3691 3691 3692 3692 3692 3693 3694 3694 3694 3694 3697 3697

3688

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

6.

Unsolved issues . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1. Isolation procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2. LC–MS, the benefits of chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.1. Size-exclusion chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.2.2. Reversed-phase chromatography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3. Electrospray ionization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.1. In-source fragmentation and disaggregation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.2. Adduct formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.3. Selective ionization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4. Data exploitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.1. Determination of average molecular weights . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4.2. Database problem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

7.

1. Introduction Complex, poorly defined mixtures of natural organic matter (NOM) can be found in virtually all compartments of the earth, as particulate organic matter in soils and sediments, as suspended in water, and in dissolved form (DOM) in pore water of soils and sediments, in surface water, in groundwater, in the worlds oceans and even in aerosol of the atmosphere. NOM consists of an incredibly large number of different molecules (several thousands can be easily determined in one sample) usually in low concentration. These NOM molecules do not only comprise the majority of molecules in environmental samples, but also represent a major portion of the organic carbon content in water, soil and air. Therefore, NOM is of importance for carbon transport and cycling and also for other biogeochemical processes. In the atmosphere NOM alters the physical and chemical properties of aerosol, and in soil it interacts with the cycling and transport of inorganic compounds such as nutrients and metals. Besides that NOM is linked with the fate of environmental contaminants as it may, inter alia, act as a source of new contaminants, for example as so-called disinfection byproducts (DBP) during chlorination processes of drinking water, as a carrier or cosolvent for organic compounds of limited polarity, and as a sink by providing sites for chemical reactions (bound residue formation). These are some of the reasons why investigations into the formation, transformation, decay, and function of NOM are an important issue in soil science, in aquatic chemistry, and water treatment, in biogeochemistry as well as in atmospheric sciences. The analysis of NOM in soil and water has a long tradition. In the last decades NMR spectroscopy has been used extensively and has provided most important information on structural properties of NOM thus far. The role of mass spectrometry, however, remained limited for most of this time as it required the destruction and derivatization of NOM molecules to make them amenable to gas chromatography. Except for certain compound classes, whose presence had been confirmed by target analysis such as sugars, amino acids, fatty acids etc., the identity of molecules of NOM remained unknown because the molecules could not be made visible by any analytical technique. Indeed, the term NOM in this review is being used to denote just those molecules whose chemical nature had remained unknown for such a long time. Two inventions in mass spectrometry have changed this situation dramatically and have given mass spectrometry the prime position in present NOM research. The first was electrospray ionization (ESI), which allowed the infusion of aqueous solutions into the mass spectrometer and the coupling of mass spectrometry with liquid chromatography. This circumvented the need to destruct and derivatize NOM molecules and allowed their direct analysis by mass spectrometry. The first paper providing an impression of what ESIMS could contribute to NOM research was published by McIntyre

3697 3697 3697 3698 3698 3698 3698 3699 3699 3699 3699 3700 3700 3701

et al. in 1997 [1] (Fig. 1). The authors already recognized important mass spectral characteristics of fulvic acids. The second invention was ultra-high-resolution mass spectrometry in the form of Fourier transform ion cyclotron resonance (FTICR-) MS (e.g. [2]). First applications of ESI-FTICR-MS to NOM research that showed the general feasibility of this approach began in the late 1990s [3–5]. The real potential of this technique became visible from work in the early 2000s (e.g. [6]). Stenson et al. [7] already published thousands of molecular formulas from one fulvic acid sample. To date, ultra-highresolution mass spectrometry is the only technique that is able to separate molecular species from these complex NOM mixtures and to allow the analysis of NOM at the molecular level without any fractionation. With ESI-FTICR-MS instruments soil scientists, aquatic chemists, biogeochemists as well as atmospheric chemists have received an analytical instrument of remarkable analytical power. This tool has opened a perspective on a largely unknown kingdom of organic molecules, that had long seemed a bit mysterious. Meanwhile, thousands of molecules have been encountered in NOM isolates in all environmental compartments yet investigated and a remarkable similarity between these molecular series has been recognized, from dissolved organic matter in the deep ocean and in freshwater, over soil organic matter to secondary organic aerosol in the atmosphere. The present state of the analytical methodology developed for NOM analysis by ESI-MS and the experiences that have been made in the past 10 years are reviewed in this paper. Critical methodical aspects are outlined and discussed and issues that deserve our attention and our future efforts are highlighted. The analysis of molecular structure of these NOM molecules by means of collision induced dissociation in hyphenation mass spectrometry, with triple-quadrupole-MS or quadrupole time-of-flight MS (Q-

Fig. 1. First ESI(−)-MS spectrum of NOM (in this case: groundwater organic acids) generated with a quadrupole MS. Reproduced with permission from [1].

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

3689

Table 1 Minimum mass differences to the next C, H, O-only molecule of the same integer mass in NOM isolates. Add

Remove

C, H, O-only molecules O CH4 C4 O3 Heteroatomic molecules S O2 N2 CO SH4 C3 PH O2 13 N CH Adducts NaH a b c

C2

Mass difference (mDa)a

R requiredb

Found in

Reported by

Each integer mass in all isolates Overlap of “islands”

See Fig. 2 See Fig. 2

36.39 15.26

14000 33000

−17.76 11.23 3.37 −8.24 −8.15

28000 45000 150000 61000 61000

Marine NOM Atmospheric aerosol Calculatedc Calculatedc

−2.37

210000

Positive ion mode

[11] [8,9] [11]

[8]

Larger differences are obtained by adding increments of 36.39 mDa. Mass resolution (m/m50% ) at FWHM required for an ion of m/z 499 and two signals of comparable intensity. Calculated, not explicitly mentioned in literature.

TOF-MS), or by excitation inside an FTICR-MS is not covered in this review. This second and equally exciting field of mass spectrometric research on NOM is left for a separate review. 2. Analysis 2.1. Ionization and modifiers It is well known for ESI-MS that the composition of the solution in which an analyte is introduced into the ion source can have significant influence on the ionization procedure. The solutions pH as well as the presence of inorganic or organic modifiers influence the ionization procedure and, thus, the appearance of the mass spectrum generated by ESI-MS. For NOM components, most of which are expected to be of anionic nature at neutral pH, the negative ion mode would be the “natural” choice. Indeed, the first publications [1] on ESI-MS of natural organic matter and most that followed have used the negative ion mode. In some cases the solution was made slightly basic by ammonium bicarbonate or ammonium hydroxide. Astonishingly, the first studies that involved FTICR-MS preferred the positive ion mode. This may have been due to the fact that FTICRMS had previously been used primarily for protein analysis, which is also performed in positive ion mode. One disadvantage of using the positive ion mode is the occurrence of sodium adducts with a mass difference of only 2.4 mDa to other non-sodiated species [8] (Table 1). More recent studies using FTICR-MS do now also use the negative ion mode. Positive ionization could be favorable in case that nitrogenous and basic NOM compounds are of interest. Both modes have been compared in two studies: while one study found no advantage of the positive ion mode even for nitrogenous molecules [9], the other one recorded clear differences in the mass spectra obtained in either mode [10]. The use of both ionization modes is attractive for structure elucidation of NOM by hyphenation mass spectrometry because the site of ionization and the processes of fragmentation may well differ in either mode. Therefore the information gathered in both modes for one parent molecule can be complementary [11]. Fulvic acids and most other NOM isolates from aqueous solution are well water soluble and do not require the use of an organic modifier for solubilization. Such modifiers, however, can stabilize the spray formation and may reduce the droplet size in the electrospray process, which may improve the formation of ions and support their transition into the gas phase. Therefore, organic modifiers may be used even though they are not required to solubilize the analytes of interest. In the, yet rare, cases that on-line chromatography has been used the choice of inorganic and organic modifiers may be dictated by the needs of the chromatography (see Section 6.2).

It has been reported that fulvic acids stored in MeOH over prolonged periods of time showed an increasing content of methyl esters [12]. The use of MeOH in eluents or dilutions for infusion, however, should not pose any problem because the exposure time is comparatively short. No systematic study on the influence of the solution composition on the appearance of mass spectra of NOM has yet been published. Often, not even the pH of the infusion solutions has been controlled. This is especially critical when spectra of different isolates are compared, because differences may then be due to differences in the modifier composition rather than to differences in the NOM isolates themselves. 2.2. Mass analyzers The term “resolution” in mass spectrometry is often implicitly used as a synonym for mass accuracy and precision. A high resolving power, however, does not guarantee accurate m/z values (an adequate calibration is needed for this purpose) and precision may be low even with a high-resolution instrument if its calibration is not stable. In NOM analysis, however, resolution is really needed in its original sense, to resolve one molecular species from the other. The widely available quadrupole mass spectrometers and ion trap mass spectrometers were clearly insufficient [13]. While TOF-MS was a step forward [14,15] especially when used in the Wmode [16,17] it was, still, not sufficient for a full resolution. This is obtained, only, by ultra-high-resolution mass spectrometry with FTICR-MS. How much resolution is required to separate molecular species in NOM isolates? Obviously, this depends upon the complexity of the isolates. If solely C, H, O-only molecules occur that appear at odd m/z values, the resolution needed is quite clear. An omnipresent theme at the level of isobaric ions of NOM is series of molecules that differ by the formal replacement of O by CH4 , resulting in a mass difference of 36.4 mDa [18] (Table 1, Fig. 2, left and right). This pattern has been found at all odd m/z ratios in all isolates investigated. For such isobaric ions with a mass difference of 36.4 mDa, the number of carbons and oxygens per molecule is constant. In Fig. 2 this sum is 26 at m/z 367 (Fig. 2 left) and 27 at m/z 373 (Fig. 2, right). But between m/z 367 and m/z 373 there is a number of odd m/z values ions where both series overlap (Fig. 2, middle). In that region the minimum mass difference between isobaric ions is reduced to 15.26 mDa, corresponding to (+O3 –C4 ) (Table 1). 13 C2 compounds are not relevant in this mass range. Thus, in case that solely C, H, Oonly molecules are present and ESI is used in the negative ion mode, an Orbitrap-MS (R of 60 000) should provide sufficient resolving power [19]. If only major ions are of interest even a recent TOF-MS would do (resolution up to 15 000).

3690

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

Fig. 2. Ultra-high-resolution (FTICR-MS) spectrum for m/z 367, m/z 371 and m/z 373 of a fulvic acid from river water (SRFA). Numbers at the peak tops denote the numbers of C/H/O. Molecules with a sum of carbon and oxygens of 26 marked with a circle; molecules with a sum of carbon and oxygen of 27 marked with a triangle.

Obviously the inclusion of heteroatoms increases the requirements in terms of mass resolving power much further (Table 1). In that case molecular formula determination requires the use of FTICR-MS, which provides a mass resolution of 200 000 and higher. In some studies performed at Marshalls group at the National High Magnetic Field Laboratory (NHMFL) mass resolutions ranging from 400 000 to 500 000 [20] have been used. In a recent study with a commercial 12 T instrument a mass resolution up to 580 000 [21] has been reported. This appears sufficient to solve all resolution problems related to NOM. It was shown recently that a mass resolution and accuracy of 0.1 mDa would be required to resolve all theoretically possible isobaric ions in a mass range up to 500 Da and to unambiguously detect their elemental composition [22]. For an ion at m/z 500 this would correspond to a resolving power of 5 Mio. This is (only) about one order of magnitude above the resolving power provided by the most recent commercial FTICR-MS instruments. Usually a large number of FTICR-MS spectra are accumulated to improve the signal to noise ratio and in this way good precision in mass determination is also achieved. But besides that accuracy in the mass determination is an inevitable prerequisite for molecular formula determination. This is best achieved by internal calibration of the spectra. Omnipresent fatty acids may be used for that purpose [21], while in another study with a large number of N and S-containing molecules, the well known pattern of the C, H, O-only ions was used for internal calibration [11]. 3. Molecular formula calculation With the mass resolution offered by FTICR-MS several thousands of ions may be detected in a comparatively narrow mass range up to m/z 1000 in one NOM isolate [7]. Formulas are proposed by the MS software based on preselected elements (e.g. C, H, O only, or also N, S or even P) and double bond equivalents and allowing for a certain mass error. One such programme is also available from the NHMFL [19,21]. But even with a mass error window as narrow as ±1 ppm (0.5 mDa for m/z 500) about 10 molecular formulas will be proposed for m/z 500 [8]. The number of possible formulas may be reduced by considering isotope patterns [23] but this requires sufficient intensity of the signal of the far less frequent heavier isotope [8], which is often not the case, and could only be done on an ion by ion basis. Therefore, one needs to select the most likely formula from those proposed by the software. The easiest approach would be to select the one with the lowest mass error (e.g. up to 4 ppm as in [24]) but the closest match is not necessarily the best selection. If sufficient time is available additional criteria for selecting the appropriate formula for each signal may be used [25]. Generally, all additional criteria that may be used for “manual” as well as automated formula selection are based on the finding that in NOM isolates series of

molecules are found [7,16,25], often with a Gaussian intensity distribution [26], whereas isolated formulas are far less frequent and not really interesting (they would represent the “weed” in the molecular “grass” of NOM molecules). These series have characteristic exact mass differences (e.g. 14.0157 Da for alkyl chain homologues, 2.0157 Da for hydrogen homologues, 15.9949 Da for oxygen homologues. Additional characteristic differences arise from the formal exchange of elements, 1.0034 Da for 13 C against 12 C, 0.9952 Da for NH against CH2 and 0.0364 Da for CH4 against O [7] (Table 1). These differences can be used to decide on which of the proposed molecular formulas is the most likely for one ion by considering the mass distance to lower or higher homologues of putative series or to neighbouring ions with the same integer mass. A powerful means for molecular formula determination (and selection) is the calculation of the Kendrick mass defects (KMDs) [20]. This calculation method was first proposed by Kendrick in 1963 for computing high-resolution data of alkyl chain homologues in petrochemical analysis [18]. For the case of molecular formula determination the now-called Kendrick mass (KM) of a mass spectrometric signal is calculated from the detected mass according to Eq. (1): KM = massdetect

 14.00000  14.01565

(1)

The difference between the Kendrick mass and the closest integer mass is called the KMD which is calculated by Eq. (2): KMD = massinteger − KM

(2)

Molecules that differ only in the number of alkyl groups exhibit the same KMD value, while molecules with different number of double bond equivalents, nitrogens, oxygens or sulfur atoms differ in their KMD values. For example adding one O to a molecule increases its KMD by 0.0229, while adding H2 decreases the KMD by 0.0134 (Fig. 3). This greatly helps to extend the range, in which molecular formulas can be derived from the measured masses, from the low mass range, where formulas may be unambiguous, into the moderate mass range (>300 Da) or even up to a higher mass range of m/z 900 [27], where a decision on which of the proposed formulas is most likely would otherwise become very difficult and prone to errors. The KMD calculation and plotting in a Kendrick diagram can also compensate for a drift in the mass axis (0.15 mDa from m/z 200 to m/z 450 in the data shown in Fig. 3). Calculation of KMDs appears to be widely used for molecular formula selection in crude oil analysis [20]. This KMD concept may be extended to any other functional group found in NOM, such as carboxylate groups [24,28] or oxygens [29]. In these cases the Kendrick mass would be calculated by normalizing the experimentally determined mass by either (44.0000/43.9898) or (16.0000/15.9949) and then all molecules differing only in the number of carboxylate groups or in the number

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

3691

Fig. 3. Kendrick plot (CH2 -normalized) for organosulfur compounds determined in secondary organic aerosol, mostly organosulfates. Molecular formulas denote the start and the end member of the respective CH2 homolog series. Data from [11].

of oxygens would exhibit the same KMD. But as alkyl chain homologues are most common in NOM the KMD calculated on the basis of CH2 groups is the most useful one and the one most frequently used. Obviously, manual approaches of molecular formula selection, even when supported by KMD calculation, are very time consuming and may sometimes be influenced not only by the experience of the analyst but also by his expectation. To improve productivity mathematical approaches for molecular formula selection have been developed that make use of the concepts outlined above in one way or another [8,30]. 4. Graphical presentation and exploitation Considering the large number of molecules detected by FTICRMS in NOM samples it is, obviously, important to find adequate ways to visualize these datasets. This should be done in a way that either shows the systematic pattern in one NOM isolate or the most characteristic differences between two isolates. FTICR-MS datasets of NOM are not only large but also multidimensional. For one molecule the number of several elements is known (C, H, O as a minimum, but also of N, S and/or P), together with the molecular mass of the molecule, and its signal intensity. If chromatographic separation was used the retention time would be an additional figure. Thus, any graphical representation in three dimensions requires data reduction, by normalizing data and/or by selecting the most appropriate data for a certain aspect. 4.1. Reconstructed spectra The most straightforward presentation of FTICR-MS data is in the form of reconstructed spectra. Reconstructed spectra can be generated from a subset of the identified ions by plotting their signal intensity against their mass. In this way the complexity of a full

Fig. 4. Reconstructed spectra for NOM molecules determined in secondary organic aerosol: (a) C, H, O-only molecules; (b) C, H, O, S molecules; (c) C, H, O, N molecules. Redrawn from [11].

mass spectrum is reduced and periodicities occurring in one subset of molecules as well as differences compared to another subgroup are more clearly visible. This is exemplarily shown for C, H, O-only molecules and those bearing one S or one N in Fig. 4; all three series show the characteristic periodicity in the signal intensity of 14 Da, but the maxima occur at different masses. Such printed mass spectra do not provide sufficient graphical resolution to keep ions with the same integer mass distinguishable.

3692

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

Fig. 5. Van Krevelen diagram for a set of NOM isolated from a swamp water. Typical elemental ratios of different classes of organic molecules are marked by the eliptical areas. Lines mark systematic changes in the elemental ratio of molecules by: (A) methylation/demethylation, (B) hydrogenation/dehydrogenation; (C) condensation/hydrolysis; (D) oxidation/reduction. Reproduced with permission from [34].

4.2. Van Krevelen diagram A so-called van Krevelen diagram was derived for the classification of coals and its precursor materials based on elemental analysis data [31]. For this purpose their H/C ratios are plotted against the respective O/C ratios. The same diagram has later been used to compare elemental analysis data of humic substances [32]. When FTICR-MS was applied to fulvic acids and other NOM isolates van Krevelen diagrams were also employed to display compositional characteristics of large sets of molecules from NOM isolates (Fig. 5) [27,33]. To date, it appears to be the most widely used mode of presenting molecular formula data of larger sets of NOM molecules. For molecules with heteroatoms a three dimensional variant may be used where the carbon normalized content of other heteroatoms is shown in the third dimension [27,34]. Alternatively, the third dimension may be used for plotting the ion intensities [17,33]. Initially van Krevelen developed his diagram to visualize the differences in the elemental composition of a limited number of molecules or in elemental analysis data of a few samples [31]. The use of this kind of diagram for plotting H/C and O/C ratios of molecular formulas of thousands of molecules brings about some disadvantages. As it normalizes to the carbon number, a van Krevelen diagram sacrifices a lot of information. For example it cannot illustrate changes with regard to the molecular mass of the molecules. Moreover, molecules with different molecular formulas plot at the same point in the diagram, if their H/C and O/C ratios coincide. For larger datasets of hundreds or thousands of formulas this is regularly the case. The problem of normalization can be weakened a bit by constructing separate plots for selected mass ranges of one dataset. A second weakness is the fact that the two dimensions of the diagram are not truly orthogonal, as H/C and O/C ratios are not fully independent from each other. Moreover, literature shows that the use of van Krevelen diagrams for interpreting FTICR-MS data of NOM is prone to some misunderstandings: (a) It has been recognized that different classes of biogenic compounds tend to plot in certain areas of a van Krevelen diagram (Fig. 5). But simply because the elemental ratios O/C and H/C of one molecule fall into the region of one of these compound classes, this molecule does not necessarily belong to that compound class. A very simple example: the presence of one N and 2O atoms in a molecule may indicate an ␣-aminocarboxylic

acid; but only because a molecule carries NO2 it need not be an amino acid. Obviously the elemental composition and, even more so, the elemental ratio is an insufficient criterion to ascribe a molecule to a certain class of compounds. Also the molecular structure needs to coincide. Although this may seem obvious, literature shows several examples in which van Krevelen diagrams have been erroneously and repeatedly used to suggest an interrelationship between NOM molecules, for which elemental compositions have been determined by FTICR-MS, and biogenic compound classes [21,29,35]. (b) A similar misunderstanding is related to the lines of oxidation, hydrogenation and methylation that can be plotted in a van Krevelen diagram (lines D, B and A in Fig. 5). If NOM molecules plot along such lines this is sometimes interpreted as an indication for an educt–product interrelationship between those molecules. However, these lines indicate a formal interrelationship in the H/C and O/C ratios, only. One has to be aware that neighbours in such a diagram may exist completely independent from each other without any meaningful interrelationship at all. As a matter of fact the ions represented by neighbouring points in a van Krevelen diagram may represent molecules that are separated by hundreds of mass units. To conclude, van Krevelen diagrams may be suited to provide a broad overview on the average properties of NOM samples and as they are impressive they are being used extensively, but their diagnostic value is limited. 4.3. Kendrick diagram As mentioned previously the calculation of the KMDs was first proposed by Kendrick in 1963 for computing high-resolution data of alkyl chain homologues in petrochemical analysis [18]. First, the Kendrick mass is calculated from the elemental composition of an ion or molecule according to Eq. (3): KM = massIUPAC

 14.00000  14.01565

(3)

Then the KMD can be calculated, usually normalized for alkyl chain homologues according to Eq. (2). If the Kendrick mass defects of molecules are plotted against their integer mass it rapidly becomes obvious that series of alkyl homologs arrange along horizontal lines with a nominal mass spacing of 14 Da (Fig. 3) [20]. All series of ions with other differences in the elemental composition have other KMDs and are, thus, vertically separated from each other. For example molecules that differ by the number of hydrogens or oxygens have KMDs of 0.0134 and 0.0229 (Fig. 3). To keep their clarity Kendrick plots can only be generated for a limited number of molecular series, but over the whole mass range investigated. The so-called nominal mass series z* that can be calculated for the molecular formula of each ion [36] can be used to presort ions for KMD diagrams [7]. Such Kendrick diagrams are diagnostically useful and support molecular formula selection from measured masses (see Section 3), but they are not very illustrative and hardly understandable by intuition. 4.4. C versus mass diagrams Another approach that appears to be related to the Kendrick plot is a C versus mass diagram. In this diagram the number of carbon atoms of each of the molecules is plotted against its molecular mass (Fig. 6b) [11,37,38]. The molecules in NOM isolates appear to arrange themselves in island-shaped arrangements and all members of such an “island” share the same sum of C and O atoms per molecule. From one island to the next this number and, thus, the sum of carbons

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

3693

molecules differ by the formal exchange of one oxygen by (C + 4H), leaving the sum of C + O unchanged. This replacement reflects the previously recognized mass difference of 0.0364 Da between isobaric ions which appears to be present in all NOM isolates yet investigated (Fig. 2) (e.g. [7,16,25]). Interestingly such patterns appear to be present in all sets of C, H, O-only molecules in all NOM isolates yet investigated. To keep the clarity of the C versus mass diagrams separate graphs should be drawn for molecules with heteroatoms [9,11]. Contrary to other plots the island diagram does not normalize anything. Thus, all information on a molecule, its elemental composition and its mass, is preserved in this graph. One may also add relative signal intensity as a third dimension, for example by drawing a contour plot where the colors code for the relative signal intensity [37–39] (Fig. 6c). If one compares the molecular composition of samples in an island plot, those with a higher contribution of oxygenated molecules plot lower in vertical direction (more oxygen and less carbon at the same m/z value), while those with more aromatic molecules are shifted in horizontal direction to the left (lower mass for the same number of oxygens and carbons per molecule). This provides a very clear evidence for compositional differences of samples for the selected mass range. The regularity and homogeneity in the elemental composition of the NOM molecules isolates from very different environment is one of the most striking findings in FTICR-MS analysis of NOM thus far. It is not clear, yet, to which extent this homogeneity is an artefact of the analytical process (see Section 6) or a reflection of the source material and the transformation processes of NOM [37]. 4.5. Other diagrams The extent of unsaturation of a molecule is an important structural feature and a measure of its potential reactivity in the presence of sunlight as well as towards electrophils. The unsaturation may be expressed by the so-called double bond equivalents (DBEs), i.e. the number of double bonds + rings systems. This parameter can be calculated according to Eq. (4) [40]: DBE = −

Fig. 6. FTICR-MS data of a river water fulvic acid (m/z 260–400) (SRFA); (a) spectrum, (b) island plot for 329 fulvic acid molecules; (c) contour plot of (b) with the relative signal intensity indicated by the colored areas. Redrawn from [37]. Numbers in the diagram denote the sum of carbon and oxygen atoms for all members of one island.

and oxygens per molecule increases by one. In the mass range m/z 260–400 in Fig. 6 10 of these islands are found but this pattern continues from lower and towards higher masses. Also within each island the molecules elemental composition changes systematically, with clear consequences also for the structure of these molecules: Horizontally (Fig. 6) within one island the saturation increases stepwise (+H2 = +2.0157 Da). Vertically (Fig. 6)

3A5 A3 A1 + + A4 + + 2A6 + 1 2 2 2

(4)

where A1 –A6 denotes the number of atoms with the valency 1–6. The DBE of a molecule can be easily calculated from its molecular formula for C, H, O-only molecules. Unfortunately, DBE calculation for a molecule on the basis of its molecular formula is not possible when S or P is present, because these elements exhibit different valencies, depending on their oxidation state. Only if a sample originates from strictly reducing environment (petrochemistry) one can assume S being bivalent as in sulfides or thiols and P trivalent as in phosphines. Then a DBE calculation solely on the basis of the molecular formula may be reasonable. Most NOM samples yet investigated stem from oxic environment, however, and then sulfur may also occur in tetravalent (sulfoxid) or hexavalent (sulfone, sulfonic acid, sulfate ester) form, while P may be pentavalent (phosphonates, phosphate ester). Thus, in the presence of S and P the DBE calculation requires information on the oxidation state of these elements. If that is not available, no DBE value can be calculated. Unfortunately, this has not been always considered [21,34,41]. Also in the case of nitrogen one may question, whether the DBE calculation with a fixed valency of 3 for nitrogen is always adequate. When nitrogen occurs in its highest oxidation state (+III) as in nitrogroups or nitrate esters it could, at least formally, be considered a pentavalent state even though the fifth bond is not present but replaced by the positive charge on the nitrogen and the negative charge on one of the oxygens.

3694

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

Fig. 7. Frequency distribution of carbon normalized double bond equivalents (DBE/C) for a soil charcoal isolate and a soil porewater isolate. Reproduced with permission from [42].

For C, H, O-only molecules, however, and for nitrogen in most of its forms, the following simplified formula for DBE calculation from the molecular formula Cx Hy Nz On can be used [40]: DBE = x −

y z + +1 2 2

(5)

NOM samples have been compared for the relative abundance of double bonds by drawing frequency distributions of the carbon normalized DBE values (DBE/C) of all identified molecules (Fig. 7) [42] or the DBE-O values versus mass [29]. Besides such visual ways of comparing the NOM composition of different samples mathematical tools need to be used to compare these large datasets. These measures should be able to express the degree of similarity between two datasets in a quantitative way and to detect and describe the most specific differences between samples. Provided that analytical conditions are reproducible and identical for all samples of a set the relative signal intensity should also be included into such a comparison. Likely, such tools exist already for data processing in proteomics or metabolomics. 5. Applications 5.1. Biogeochemistry Electrospray ionization–high-resolution-mass spectrometry using FTICR-MS has been successfully used to determine elemental compositions of natural organic molecules in very diverse environments (Table 2). Among them were fulvic acids from freshwater [6,7,25], NOM in marine environment, in coastal waters [21,29,43] as well as in the open and deep ocean [9,24], NOM in freshwater environment [35], black carbon derived molecules in soil environment [28,42], NOM deposited in an Arctic ice core [44] and organic matter formed in the atmosphere as so-called secondary organic aerosol (SOA) [11]. In these investigations various scientific issues have been touched, such as the putative sources of fulvic acids [7,17,37] and of organic matter in the coastal environments [21,29,43], structures and regularities in NOM molecules [17,24,37]. The changes in DOM composition from coastal regions into the open ocean have been investigated [21,29]. With respect to soil environment FTICR-MS has been utilized to study transformation of black carbon [28,42]. Only single studies have been published on the quality of organic matter deposited in the Arctic region [44] and the identity of SOA [11]. 5.2. Investigation of transformation reactions One of the most striking improvements provided by ESI-FTICRMS is the possibility to observe chemical reactions at the molecular level in these complex NOM mixtures. The transformation of

aquatic NOM by biological means [35,45], by photolysis [45] and by chemical oxidation (ozonation) [39] has been studied. FTICR-MS data of the reaction of NOM molecules with ozone are shown here to illustrate this potential. Extensions of FTICRMS spectra of Suwannee River fulvic acid (SRFA) before and after ozonation are shown in Fig. 8. Of those ions found at m/z 267 before ozonation (Fig. 8a, top) the least oxidized species (C15 H24 O4 ) reacted fastest and was quantitatively removed by an ozone dose of 2.5 mg/mg DOC (Fig. 8a, bottom). The more oxidized fulvic acid molecules of this set of isobaric ions (e.g. C12 H12 O7 ) reacted slower and less complete. The reactivity towards ozone was much higher for all m/z 345 ions (Fig. 8b). These were completely removed by the same ozone dosage. Instead highly oxidized ozonation products now appeared at m/z 345 that carry much more oxygen and less carbon (O/C ratio 0.7–1.1) (Fig. 8b, bottom). These changes upon ozonation can be summarized for all ions in the mass range 310–370 Da in a carbon number versus mass diagram, where color codes for the relative signal intensity of the ions (Fig. 9). Compared to the non-ozonated isolate (Fig. 9a) three changes are well visible in the contour plot of the ozonated sample (Fig. 9b): (i) all islands are shifted to very low carbon numbers, reflecting the inclusion of several moles of oxygen into the molecules; (ii) For each island the intensity maximum is shifted towards higher masses (for island #24 from 338 Da to 346 Da), indicating that more saturated molecules remain after ozonation because the unsaturated ones are more reactive towards electrophils; (iii) the maximum in ion intensity of the whole mass range is shifted to lower masses (from island #26, 366 Da to island #22, 316 Da), illustrating that heavier (and larger) molecules are preferentially removed and smaller molecules are generated by ozonation (which is, indeed, the case [39]). 5.3. Nitrogenous molecules The carbon to nitrogen (C/N) ratio of dissolved organic matter ranges from 10 to 12 in marine environment and from 15 to 50 for many freshwater systems [46,47]. Assuming that the nitrogen was evenly distributed within the DOM one nitrogen would be bound to every tenth carbon in marine DOM. On a molecular basis each molecule visible by FTICR-MS should carry at least one nitrogen atom. However, the overwhelming majority of published molecular formulas of DOM do not include heteroatoms but is formed by C, H, O, only (Table 2). This is, at least partly, due to the isolation procedures (see Section 6.1). Molecular formulas carrying up to five nitrogens were recently reported from a freshwater isolate [35] and detected by FTICR-MS in the broth of a bacterial culture [45]. In the context of soil analysis graphs that include nitrogenous molecules have been shown [34] but no molecular formulas were reported. The same is true for an Arctic ice core [44]. In crude

Table 2 Application of (ultra-) high-resolution MS to NOM analysis. NOM origina

Enrichment

Introduction; modifiersb

Ionization; MSc

Heteroatoms

Remarks

Reference

Groundwater River water; soil elutriate River water (SRFA) River water (SRFA)

Unknown RO, UF; NaOH XAD XAD

Infusion; none Infusion; MeOH, NH4 OH Infusion; MeOH, acetic acid/propanol, NH4 OH Infusion; MeOH, acetic acid/propanol, NH4 OH

Pos/neg; Q-TOF Pos; FTICR (9.4 T) Pos/neg; FTICR (9.4 T) Pos/neg; FTICR (9.4 T)

None None None None

+MS/MS spectra

[14] [6] [25] [7]

River water (SRFA)

XAD

Infusion; MeOH

Neg; FTICR (9.4 T)

N

Soil ash elutriate River water (SRFA) River water, groundwater, peat elutriate Arctic marine water (Weddell Sea) + mangrove River water (SRFA)

Freeze drying XAD XAD C-18 XAD

Infusion; MeOH, NH4 OH SEC; MeOH, NH4 HCO3 SEC; MeOH, NH4 HCO3 Infusion; MeOH; formic acid SEC; MeOH, NH4 HCO3

Neg; FTICR (9.4 T) Neg; Q-TOF (W-mode) Neg; Q-TOF (W-mode) Pos; FTICR (7 T) Neg; Q-TOF (W-mode)

None None None None None

Arctic ice core Soil pore water Atmospheric aerosol Surface water, peat elutriate Freshwater

C-18 C-18 C-18 XAD C-18

Infusion Infusion; MeOH, NH4 OH Infusion; MeOH, NH4 HCO3 SEC; MeOH, NH4 HCO3 Infusion; MeOH,

Pos; FTICR (9.4 T) Neg; FTICR (9.4 T) Neg; FTICR (6 T) Neg; FTICR (6 T) Neg; FTICR (9.4 T)

N, S None N, S None N

Marine surface and deep water (Pacific)

UF; freeze drying

Infusion; formic acid

Pos; FTICR (7 T)

None

Mangrove swamp + estuary River water (SRFA) Deep marine water

C-18 XAD XAD/freeze-drying and desalting

Infusion; propanol, NH4 OH On-line SEC; MeOH, NH4 HCO3 Infusion; MeOH, NH4 HCO3

Neg; FTICR (9.4 T) Neg; FTICR (6 T) Neg; FTICR (6 T)

None None N

River to ocean

C-18

Infusion; MeOH, NH4 OH

Neg; FTICR (12 T)

N, S, P

River water (SRFA)

XAD

Infusion; AcCN, NH4 OH

Neg; FTICR (9.4 T)

None

Arctic marine water (Weddell Sea)

Polymer SPE

Infusion; MeOH

Neg; FTICR (9.4 T)

N, S

Aerosol from biomass burning River water (SRFA)

XAD

Infusion; Infusion; MeOH, formic acid/MeOH

Pos; Orbitrap Pos/neg; FTICR (9.4 T)

None None

c

+MS/MS spectra +MS/MS spectra Before and after ozonation; MS/MS spectra Soil, black carbon +MS/MS spectra Comparison before and after biodegradation of DOM Ultrafiltrated DOM; formulas available > 420 Da

14

C-age of several thousand years; highly hydrophilic fraction isolated; low resolution MS/MS spectra Comparison of C18 and direct injection for one sample Off-line RPLC fractionation Off-line RPLC fractionation Comparison of ESI, APCI and APPI

[45]

[28] [16] [17] [43] [39]

[44] [42] [11] [37] [35]

[24]

[29] [38] [9]

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

a b

About 4000 formulas available Alteration by photolysis and biodegradation

[21]

[55] [50] [19] [10]

SRFA denotes Suwannee River fulvic acid standard of the International Humic Substances Society. Introduction into the ionization source by infusion or size-exclusion chromatography (SEC); modifiers: inorganic and organic additives to the sample solution or to the chromatographic eluent. Ionization by ESI in either positive or negative mode and kind of mass spectrometer used.

3695

3696

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

Fig. 8. Extended FTICR-MS spectra for m/z 267 and m/z 345 of SRFA. top: before ozonation; bottom: after ozonation with 2.5 mg O3 /mg DOC. Numbers at the peak tops denote the number of C/H/O of the respective molecules.

oils and coal, however, molecules with heteroatoms, among them nitrogen, appear to be quite frequent [20,48]. At least for aqueous samples the lack of reports on nitrogenous molecules may partly be due to the very polar and possibly zwitterionic character of DON species. This has, likely, prevented their isolation/enrichment from aqueous samples on hydrophobic stationary phases and made the nitrogenous compounds escape from the enrichment, and, thus, from analysis (see Section 6.1). Only recently highly hydrophilic marine DOM was isolated for the first time in a complex multistep procedure [49]. In that

hydrophilic fraction of marine NOM hundreds of nitrogen containing molecules could be identified and molecular series with one, two and three nitrogens per molecule have been identified with high mass accuracy [9]. Ions carrying one nitrogen atom are rapidly visible as increased signal intensity at even m/z values. The minimum distance to 13 C -isotopic signals of C, H, O-only molecules is 8.15 mDa. A char1 acteristic mass difference of 11.23 mDa was found in many odd m/z ion envelopes which is due to the formal replacement of CO by 2N between molecules consisting only of C, H and O and those with

Fig. 9. Contour island plot for fulvic acids (mass range 310–370 Da). (a, left): before ozonation; (b, right): after ozonation with 1 mg O3 /mg DOC.

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

2N (Table 1). The same distance between pairs of signals was also found for even m/z ions, between ions containing 1 and 3N [9]. Also in atmospheric aerosol nitrogenous compounds have been identified by FTICR-MS [11], which could be identified as organonitrates by their product ion spectra. 5.4. Sulfur containing molecules While sulfur-organic molecules appear to be quite frequent in crude oils and related products [20,48] their occurrence in NOM has yet been recorded only seldomly (Table 2). Sulfur organic compounds, presumably organosulfates, have been identified in organic matter of an Arctic ice core [44], and in large number in secondary organic aerosol of the atmosphere [11]. In this case (low resolution) product ion spectra proved that truly organosulfates were present. The presence of sulfur containing molecules in marine waters has been reported recently [21,50]. However, the relevancy of organosulfur NOM on the earth’s surface cannot be assessed on the basis of the data presently available. From a mass spectrometric point of view the occurrence of sulfur-organic molecules is quite challenging because the mass difference to non-sulfur ions can be as low as 3.37 mDa (Table 1) and high mass resolution and accuracy is, thus, required for their identification. This may be one of the reasons why organosulfur components have only rarely been detected, yet. 5.5. Phosphorus containing molecules Dissolved organophosphorus compounds in NOM have hardly been studied by FTICR-MS, yet. After a selective enrichment procedure from lakewater molecular formulas of about 20 compounds have been identified in a comparatively early application of FTICRMS [51]. Recently, the occurrence of P-containing NOM molecules has been reported for a swamp water with a high DOC content exceeding 100 mg/L [21]. The lack of previous detections of Pcontaining NOM molecules appears to be due to two reasons: (a) a lack of sensitivity for P compounds, which are much more diluted as compared to C, H, O-only molecules or to nitrogenous molecules, and (b) the comparatively low mass difference of 8.24 mDa to the next C, H, O-only molecule (Table 1), which hampers its detection in case of a weak signal even when the mass resolution is nominally sufficient. While considerable progress has been made to determine NOM molecules containing heteroatoms like N, S and P, information on the functional groups in which these heteroatoms occur is very limited. Clear information could only be gathered for atmospheric aerosol where characteristic fragmentations occurred in MS/MS experiments [11] while this was not possible for marine nitrogenous material [9].

3697

Because inorganic salts, cations in the positive ion mode and oxoanions in the negative ion mode, interfere with the ionization in the electrospray process, freeze drying alone is not suitable for enrichment of NOM for mass spectrometric investigation. As far as humic and fulvic acids are concerned, which are major constituents of natural organic matter, enrichment from water by XAD resins is well established [7,16]. Even though this procedure is cumbersome, it is inevitable in humic and fulvic acid research, because both compound classes are only operationally defined through this XAD procedure. For NOM research enrichment by solid-phase extraction on C-18 stationary phases under strongly acidic conditions (pH 2) is much easier and has become quite common (Table 2). Both enrichment procedures are incomplete and extraction efficacies in the range of 20–40% of the DOC have been reported [9,21,43]. In other words, up to 80% of DOM are not enriched and could not be considered in FTICR-MS analyses of DOM. Obviously, our present view on NOM is, still, quite selective. It has been recognized that nitrogenous compounds are strongly depleted in this enrichment [24,43]. Recently, the molecular composition of the C-18 extract of a swamp water has been compared to the directly injected sample with a DOC content of 136 mg/L [21]. It was shown that highly oxygenated C, H, O-only molecules as well as nitrogenous molecules were lost in the C-18 enrichment procedure. On this basis we have to consider that the striking similarity that has been found between fulvic acid as well as NOM isolates of so many different environments is partly due to the loss of more unique constituents in these XAD or C-18 enrichment procedures. To improve this situation Leenheer has developed a multistep procedure that allowed to recover hydrophilic organic matter (HPI) that had not be retained on XAD-8 resin [49,52]. With this procedure it was possible for the first time to determine molecular formulas for hundreds of nitrogenous molecules, even from deep marine water with its very low organic matter but high salt content (Fig. 10) [9]. Probably the success in detecting the nitrogenous compounds in this case was due to two effects. Firstly, the isolation procedure was suitable for enrichment of the polar nitrogenous compounds. And, secondly, the separation of the nitrogenous matter from the bulk organic matter by this multistep procedure supported the ionization of the nitrogen-containing molecules in the electrospray process. The sensitivity of their detection was quite weak and they, likely, would not have shown up at all in the presence of compounds that are more easily ionized. It seems quite obvious that enrichment, isolation and fractionation of NOM isolates prior to ESI-MS analysis have a severe effect on the appearance of the final mass spectra. Nevertheless, systematic studies comparing the influence of enrichment by different sorbents and under different conditions on the FTICR-MS data of NOM have not been published, yet. Therefore we have to be careful in comparing data from different studies or different kinds of water, even when seemingly the same enrichment process has been used.

6. Unsolved issues 6.2. LC–MS, the benefits of chromatography 6.1. Isolation procedures To date enrichment of NOM from water is required prior to its analysis by FTICR-MS. This is even the case when samples exhibit DOC concentrations of several mg/L. Generally, ultra-highresolution MS requires higher concentrations in terms of DOC than low resolution ESI-MS, because increasing resolution splits larger signals into an increasing number of signals of weaker intensity. Thus, the initial total concentration is distributed to several thousands of ions of different exact mass. Moreover this total concentration is not evenly distributed but certain signals are quite prominent while others occur in amounts that are orders of magnitude lower.

Most FTICR-MS investigations of NOM published so far are based on infusion of the whole complex mixture into the ion source (Table 2). This is an easy and fast, but not a very convincing procedure. During infusing of such complex and highly concentrated mixtures into the ESI source the processes leading to ionization are difficult to control and may not be reproducible (see Section 6.3). A chromatographic separation prior to FTICR-MS analysis should be beneficial as it could reduce the complexity of the NOM mixture entering the mass spectrometer at a given time. The mixtures are, however, far too complex to allow complete chromatographic separation of molecules according to their hydrophobicity or along other molecular criteria.

3698

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

ents promised more success. A reversed-phase separation of NOM would reduce the number of isobaric ions entering the MS at a given time, because those differ in their oxygen content (O versus CH4 ) with a clear effect on polarity. Moreover, it would separate inorganic matrix components from the NOM molecules and, thus, reduce the risk of adduct formation. First successful attempts of RPLC-MS coupling for NOM analysis have been reported in the past years, e.g. for aerosol organic matter [54]. Coupling of RPLC with FTICR-MS is not straightforward as the time required for recording one spectrum by FTICR-MS is comparatively long and more than hundred spectra may be summed to improve the signal to noise ratio. Moreover, column overloading may occur due to the high DOC concentrations required for ultra-high-resolution MS (see Section 6.1). These problems can be circumvented by repeated off-line fractionation by RPLC followed by infusion FTICR-MS [50,55]. One of these studies provided indications that RPLC could separate isomers (with the same exact mass and molecular formula) [55], which is impossible by mass spectrometric means. In this respect RPLC and ESI-MS can be truly complementary. An off-line combination of RPLC and MS is, obviously, not a very productive strategy. But if one considers that one such analysis delivers data that require days and weeks for processing and exploitation, such an off-line ‘coupling’ may be acceptable. 6.3. Electrospray ionization

Fig. 10. Mass spectra of NOM fractions isolated from the deep ocean. (a) Hydrophobic fraction retained on XAD-8 and (b) hydrophilic fraction (HPI) that was isolated from the XAD-filtrate. Extended sections show signals at m/z 397. Redrawn from [9].

6.2.1. Size-exclusion chromatography The separation method traditionally used widely in NOM research is size exclusion chromatography (SEC), that separates according to the molecular volume of the molecules with the larger molecules eluting before the smaller ones. MS coupling has been developed using a completely volatile buffer system and a low resolution quadrupole MS [26,53], followed by high-resolution Q-TOF-MS [16] and, recently, also with ultra-high-resolution FTICRMS [37,38] (Table 2). By the parallel use of DOC- and UV-detection the SEC–MS system could be used to investigate operational influences on the ionization efficacy of the ESI process and on in-source fragmentation [53]. In the case of coupling to FTICR-MS the limited chromatographic resolution of SEC was advantageous, because it provided enough measuring time to generate spectra of the high, moderate and low mass fraction [38]. This system showed that high molecular weight fulvic acids ions follow the same system of elemental composition as the low molecular weight ions but are more aromatic and enriched in carboxylate moieties [38]. 6.2.2. Reversed-phase chromatography Owing to the generally high polarity of NOM dissolved in water previous attempts to use reversed-phase HPLC have been of limited success. With ESI-MS coupling the interest to use RPLC for NOM separation has recovered and novel stationary phases that keep their retention capacity also with almost pure aqueous elu-

The FTICR-MS analysis provides access to molecular formulas of thousands of ions in one NOM isolate. However, one needs to be aware that electrospray ionization is a complex physico-chemical process [56] that may affect polar and less polar, acidic and basic, smaller and larger members of a NOM mixture in different ways. Thus, the mass spectrum of an NOM isolate, whether generated with high or low mass resolution, is not the complete and undistorted representation of all the molecules in solution. A basic methodical problem at this stage of mass spectrometric analysis of NOM is our limited knowledge of the exact structure of NOM molecules. As long as we neither know the exact molecular structure of defined NOM molecules nor have a selection of these compounds in pure form in hand we cannot investigate the differences between the composition of a NOM solution and its mass spectrum. The three major issues of concern in the context of electrospray ionization of NOM are in-source fragmentation, adduct formation and selective ionization. Multiple charging, which was initially considered problematic, has repeatedly been shown not to be an issue. 6.3.1. In-source fragmentation and disaggregation Although ESI is a soft ionization technique weak covalent bonds may fragment and molecular aggregates of NOM polyelectrolytes may disaggregate during ionization. Thus, a part of the ions detected by mass spectrometry may originate from fragments rather than from intact molecules or from molecules rather than from aggregates. When analyzing molecules of unknown elemental composition and structure such fragmentations or disaggregations are, unfortunately, hardly recognizable. Disaggregation of non-covalently bound arrangements of molecules may be induced by changes in solution chemistry while the LC eluent enters the ESI interface. The electrophoretic field invoked by the high voltage electrical field present at the capillary tip forces charge separation. During the evaporation of spray droplets an increase in electrolyte and analyte concentration per unit volume occurs, while the repeated formation of offspring droplets and the drastic decrease in droplet volume may finally isolate one molecule from the next. During solvent evaporation the pH of the liquid phase and its dielectric constant may change drastically [56].

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

3699

Fig. 11. ESI-MS spectra of the HMW fraction of SRFA at a cone voltage of (a) 20 eV and (b) 140 eV. The inserts show extended mass spectra of m/z 220–244. Reprinted with permission from [53]. Copyright (2003) American Chemical Society.

Fragmentation, i.e. rupture of covalent bonds is more likely to occur after ionization of molecules and their transfer into the gas phase by collision with residual gas molecules. The risk of fragmentation is higher for heavier ions as those require a higher cone voltage to direct them into the aperture of the mass spectrometer [57]. By using SEC–MS [53] it was found that the total ion intensity of the HMW fraction could be increased by increasing the mass spectrometers cone voltage, i.e. the potential used to ionize the analytes and to accelerate them towards the mass spectrometers aperture. But then the m/z ratio of the ions underlying this chromatographic signal decreased as visible from the MS spectra (Fig. 11) [53]. Several authors have recognized that ions exceeding m/z 800 are hardly found in ESI-MS spectra of NOM and that intensity maxima occur in the range of m/z 350–500 (see Figs. 1, 4 and 6) [1,7,26,33,58]. To this point it is not clear, however, whether larger units present in solution, if those have existed, were (a) molecules and fragmentation of weak covalent bonds had taken place in the interface or whether (b) the initial units were aggregates the constituents of which became isolated one from another during the electrospray ionization process. Again the lack of reference material hampers significant progress in this respect. It should be noted, however, that ESI is a soft ionization technique that has been shown to be able to ionize many fragile molecules without a significant extent of fragmentation. Moreover, it may also be inevitable that the shape of a large and complex molecular arrangement changes during its transformation from a polyelectrolyte in solution to a singly charged ion in the gas phase. Much heavier ions than those determined by ESI-MS have recently been reported in humic acid by using matrix assisted laser desorption (MALDI)-TOF MS [59]. Intensity maxima in the range of m/z 40 000–50 000 were determined. From their experiments the authors could not decide on whether these ions were aggregates generated in the MALDI process or true molecular ions.

6.3.2. Adduct formation Comparatively high concentrations of NOM in the grams per litre range are infused into the mass spectrometer in most FTICR-MS studies [7,37] and adduct formation may occur under these circumstances, either with (polyvalent) cations in the positive ion mode or with oxoanions in the negative ion mode (see Section 2.1). The use of on-line chromatography instead of infusion could reduce this problem as it should separate the inorganic ions from the organic analyte molecules [54]. Adduct formation with MeOH during ESI-MS analysis in negative ion mode has been observed for fresh NOM prepared by ozonolysis of monoterpenes, when hemiacetals were formed from aldehydes and ketones [60]. The authors recommended to use

acetonitrile instead of methanol. This problem may be less critical with aged and less reactive NOM isolated from environmental samples. Also with DMSO may adducts be formed [55]. The aspect of adduct formation is a real vicious circle of NOM analysis by ESI-MS. If no adducts are detected, one may argue that they have been present in solution and were destroyed during ionization (see above), but if adducts would be detected one may call them analytical artefacts. 6.3.3. Selective ionization It is known from target analysis that not all classes of organic molecules are equally well ionized by ESI, whether in the positive or in the negative ion mode. This is one reason why not all constituents of a chemically diverse DOM mixture are equally represented in its ESI-MS spectrum. Some hints on the potential selectivity of ESI-MS were obtained by SEC separation of humic and fulvic acids and using an organic carbon detector (SEC-DOC) parallel to a mass spectrometer [53]. With this approach it was shown that high molecular weight components are detected less sensitively by ESI-MS than those of low molecular weight: while the organic carbon trace of the SEC chromatogram was dominated by the early eluting material of high molecular weight the corresponding MS signal was comparatively weak (Fig. 12a) [53]. Unfortunately, neither the extent of fragmentation/disaggregation, nor the extent of adduct formation nor the extent of selectivity of ESI-MS towards chemically differing DOM constituents can be experimentally determined or compensated for without exact knowledge on the identity of larger numbers of molecules in such isolates. Very recently, atmospheric pressure chemical ionization (APCI) and atmospheric pressure photoionization (APPI) have been used for fulvic acid analysis parallel to ESI [10]. Significant differences in the mass spectra were noted, as APCI and APPI tend to ionize less polar and smaller molecules than ESI [10]. As APCI requires more rigid ionization conditions than ESI, one could expect stronger fragmentation than at ESI conditions. 6.4. Data exploitation 6.4.1. Determination of average molecular weights The determination of the average molecular mass or weight of NOM isolates has been an important topic of NOM research and various physico-chemical methods have been used to obtain such information, among them ultracentrifugation, SEC or field flow fractionation. A mass spectrum, which is a graphical representation of the mass distribution of the ions detected in one sample and of their relative intensity, seemed to be an ideal basis for such a cal-

3700

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

their results. The rare exceptions from this rule are Stenson et al. [7] for SRFA, Hertkorn et al. [24] for marine DOM and Hatcher and coworkers [42] for organic matter in soil. These authors have provided printed lists of the identified molecular series. Printed lists are, however, not the most productive way of providing access to thousands of molecular formulas. And relative signal intensities have not been published in any of these cases. But the intensity distributions can be more characteristic for an isolate than the presence or absence of certain formulas. Presently, NOM identification by means of mass spectrometry and the use of such data is severely compromised by this lack of available data, as new data can hardly be compared with existing data. However, comparing one’s own data with those published previously is one of the very basic requirements in science. This requirement cannot be met as long as the data are not electronically available. An open access database should be built, that makes molecular formula data from samples analyzed by FTICR-MS electronically available. 7. Conclusion

Fig. 12. SEC chromatograms of SRFA with OC detection and ESI-MS detection. (b) Influence of the cone voltage of the ESI interface on the number average molecular weight (Mn ) of SRFA and SRHA. Redrawn from [53].

culation. Thus, attempts have been undertaken to calculate average molecular weights on the basis of NOM mass spectra [61,62]. Considering the putative variation in ionization efficacy for molecules of different elemental composition and structure, the fragmentation or disaggregation of larger NOM units in the electrospray source and the unknown risk of adduct formation, it becomes obvious that the relative signal intensities in NOM mass spectra must not be translated into relative concentrations. Moreover, the intensity distribution in ESI mass spectra is heavily influenced by instrumental conditions. If the electrostatic potential between the capillary outlet and the mass spectrometers aperture (cone voltage) is increased the intensity distribution is shifted towards lower m/z values as visible from the average molecular weight calculated from the data (Fig. 12 b) [53]. Besides that the frequency of the octapole or quadrupole ion storage and transfer optics that is used to deliver the ions to the FT-ICR MS has been reported to influence the appearance of the spectra [25]. Hence, a calculation of average molecular weights on the basis of ESI mass spectra does not seem to be justified. It has been argued that spectra of different samples generated under the same instrumental conditions may be compared one another [38]. While this may be true, comparison with data of other studies where analyses were performed under different conditions is to be avoided. 6.4.2. Database problem It is equally remarkable and curious that hardly any molecular formulas of NOM are publicly available, although more than 20 studies reported results of FTICR-MS analysis of NOM from various environments and report determination of molecular formulas for hundreds or thousands of molecules, each (Table 2). In most of these studies various graphs are presented (see Section 4). But the original data (molecular formulas and their relative intensity) have never been published, so that the respective authors, only, know

With FTICR-MS an analytical instrument of remarkable analytical power has been made available that has opened a perspective on a largely unknown kingdom of organic molecules in NOM. The elemental composition of thousands of NOM molecules has already been determined using this technique. While the high resolving power of FTICR-MS appears sufficient for the present needs in molecular formula determination, this review showed that many methodical questions in the analytical process before and after the FTICR-MS measurement itself need to be addressed in more detail. The major tasks that are located before the MS analysis are: 1 Enrichment procedures for NOM should become much broader and less selective. In any case clarity about their limits needs to be gathered. 2 Chromatography requires improvement so that stable reversedphase systems for NOM are available and the on-line coupling to FTICR-MS can be considered. 3 Electrospray ionization requires a closer inspection to obtain clarity on its efficacy and selectivity and on the alterations occurring during ionization (disaggregation, fragmentation, adduct formation). Only by solving these issues we can become sure about what we really see when applying FTICR-MS to NOM and what this instrumentation is unable to show us. It should be noted that reference compounds, rather than model compounds, are needed for these issues, which are not yet available. The major tasks following the generation of data by FTICR-MS are: 4 New graphical presentation methods for the large sets of molecular formula data would be useful. 5 Statistical methods for data comparison need to be developed. 6 Documentation of molecular formula including relative signal intensity need to be improved and these data made electronically available. All these issues merit our intensified effort. If they would be solved datasets generated by the application of FTICR-MS to NOM analysis would be more reliable and our interpretation would be more profound. In this way the benefit of using FTICR-MS in the field of NOM analysis could be greatly enhanced. But even when these issues were solved we would, still, not know the structure of these thousands of molecules in NOM. To pro-

T. Reemtsma / J. Chromatogr. A 1216 (2009) 3687–3701

ceed into that direction further development in mass spectrometric methods is mandatory to obtain ultra-high-resolution product ion spectra with sufficient sensitivity. References [1] C. McIntyre, B.D. Batts, D.R. Jardine, J. Mass Spectrom. 32 (1997) 328. [2] M.W. Senko, C.L. Hendrickson, L. Pasa-Tolic, J.A. Marto, F.M. White, S. Guant, A.G. Marshall, Rapid Commun. Mass Spectrom. 10 (1996) 1824. [3] A. Fievre, T. Solouki, A.G. Marshall, W.T. Cooper, Energy Fuel 11 (1997) 554. [4] T. Solouki, M.A. Freitas, A. Alomary, Anal. Chem. 71 (1999) 4719. [5] T.L. Brown, J.A. Rice, Anal. Chem. 72 (2000) 384. [6] E.B. Kujawinski, P.G. Hatcher, M.A. Freitas, Anal. Chem. 74 (2002) 413. [7] A.C. Stenson, A.G. Marshall, W.T. Cooper, Anal. Chem. 75 (2003) 1275. [8] B.P. Koch, T. Dittmar, M. Witt, G. Kattner, Anal. Chem. 79 (2007) 1758. [9] T. Reemtsma, A. These, M. Linscheid, J. Leenheer, A. Spitzy, Environ. Sci. Technol. 42 (2008) 1430. [10] N. Hertkorn, M. Frommberger, M. Witt, B.P. Koch, P. Schmitt-Kopplin, E.M. Perdue, Anal. Chem. 80 (2008) 8908. [11] T. Reemtsma, A. These, P. Venkatachari, X. Xia, P.K. Hopke, A. Springer, M. Linscheid, Anal. Chem. 78 (2006) 8299. [12] C. McIntyre, C. McRae, D. Jardine, B.D. Batts, Rapid Commun. Mass Spectrom. 16 (2002) 785. [13] C. McIntyre, D. Jardine, C. McRae, Rapid Commun. Mass Spectrom. 15 (2001) 1974. [14] G. Plancque, B. Amekraz, V. Moulin, P. Toulhoat, C. Moulin, Rapid Commun. Mass Spectrom. 15 (2001) 827. [15] V. Moulin, P. Reiller, B. Amekraz, C. Moulin, Rapid Commun. Mass Spectrom. 15 (2001) 2488. [16] A. These, M. Winkler, C. Thomas, T. Reemtsma, Rapid Commun. Mass Spectrom. 18 (2004) 1777. [17] T. Reemtsma, A. These, Environ. Sci. Technol. 39 (2005) 3507. [18] E. Kendrick, Anal. Chem. 35 (1963) 2146. [19] J.S. Smith, A. Laskin, J. Laskin, Anal. Chem. 81 (2009) 1512. [20] C.A. Hughey, C.L. Hendrickson, R.P. Rodgers, A. Marshall, K. Qian, Anal. Chem. 73 (2001) 4676. [21] R.L. Sleighter, P.G. Hatcher, Mar. Chem. 110 (2008) 140. [22] S. Kim, R.P. Rodgers, A. Marshall, Int. J. Mass Spectrom. 251 (2006) 260. [23] N. Stoll, E. Schmidt, K. Thurow, J. Am. Soc. Mass Spectrom. 17 (2006) 1692. [24] N. Hertkorn, R. Benner, M. Frommberger, P. Schmitt-Kopplin, M. Witt, K. Kaiser, A. Kettrup, J.I. Hedges, Geochim. Cosmochim. Acta 70 (2006) 2990. [25] A.C. Stenson, W.M. Landing, A.G. Marshall, W.T. Cooper, Anal. Chem. 74 (2002) 4397. [26] T. Reemtsma, A. These, Anal. Chem. 75 (2003) 1500. [27] Z. Wu, R.P. Rodgers, A.G. Marshall, Anal. Chem. 76 (2004) 2511. [28] R.W. Kramer, E.B. Kujawinski, P.G. Hatcher, Environ. Sci. Technol. 38 (2004) 3387.

3701

[29] L.B. Tremblay, T. Dittmar, A.G. Marshall, W.J. Cooper, W.T. Cooper, Mar. Chem. 105 (2007) 15. [30] E.B. Kujawinski, M.D. Behn, Anal. Chem. 78 (2006) 4363. [31] D.W. van Krevelen, Fuel 29 (1950) 269. [32] J.A. Rice, P. MacCarthy, Org. Geochem. 17 (1991) 635. [33] S. Kim, R.W. Kramer, P.G. Hatcher, Anal. Chem. 75 (2003) 5336. [34] R.L. Sleighter, P.G. Hatcher, J. Mass Spectrom. 42 (2007) 559. [35] S. Kim, L.A. Kaplan, P.G. Hatcher, Limnol. Oceanogr. 51 (2006) 1054. [36] C.S. Hsu, K. Qian, Y.C. Chen, Anal. Chim. Acta 264 (1992) 79. [37] T. Reemtsma, A. These, A. Springer, M. Linscheid, Environ. Sci. Technol. 40 (2006) 5839. [38] T. Reemtsma, A. These, A. Springer, M. Linscheid, Water Res. 42 (2008) 63. [39] A. These, T. Reemtsma, Environ. Sci. Technol. 39 (2005) 8382. [40] V. Pellegrin, J. Chem. Educ. 60 (1983) 626. [41] B.P. Koch, T. Dittmar, Rapid Commun. Mass Spectrom. 20 (2006) 926. [42] W.C. Hockaday, A.M. Grannas, S. Kim, P.G. Hatcher, Org. Geochem. 37 (2006) 501. [43] B.P. Koch, M. Witt, R. Engbrodt, T. Dittmar, G. Kattner, Geochim. Cosmochim. Acta 69 (2005) 3299. [44] A.M. Grannas, W.C. Hockaday, P.G. Hatcher, L.G. Thompson, E. MosleyThompson, J. Geophys. Res. 111 (2006) D04304. [45] E.B. Kujawinski, R. Del Vecchio, N.V. Blough, G.C.K.A.G. Marshall, Mar. Chem. 92 (2004) 23. [46] P. Kaehler, W. Koeve, Deep Sea Res. I 48 (2001) 49. [47] J.J. Alberts, M. Takacz, Org. Geochem. 30 (1999) 385. [48] A.G. Marshall, R.P. Rodgers, Acc. Chem. Res. 37 (2004) 53. [49] J. Leenheer, in preparation. [50] B.P. Koch, K.-U. Ludwichowski, G.K.T. Dittmar, M. Witt, Mar. Chem. 111 (2008) 233. [51] J.M. Llewelyn, W.M. Landing, A.G. Marshall, W.T. Cooper, Anal. Chem. 74 (2002) 600. [52] J.A. Leenheer, T.I. Noyes, C.E. Rostad, M. Lee Davisson, Biogeochemistry 69 (2004) 125. [53] A. These, T. Reemtsma, Anal. Chem. 75 (2003) 6275. [54] Y. Gomez-Gonzalez, J.D. Surratt, F. Cuyckens, R. Szmigielski, R. Vermeylen, M. Jaoui, M. Lewandowski, J.H. Offenberg, T.E. Kleindienst, E.O. Edney, F. Blockhuys, C.v. Alsenoy, W. Meanhaut, M. Claeys, J. Mass Spectrom. 43 (2008) 371. [55] A.C. Stenson, Environ. Sci. Technol. 42 (2008) 2060. [56] R.B. Cole, J. Mass Spectrom. 35 (2000) 763. [57] S.M. Hunt, M.M. Sheil, M.B. Belov, P.J. Derrick, Anal. Chem. 70 (1998) 1812. [58] J.A. Leenheer, C.E. Rostad, P.M. Gates, E.T. Furlong, I. Ferrer, Anal. Chem. 73 (2001) 1461. [59] G. Chilom, O. Chilom, J.A. Rice, Rapid Commun. Mass Spectrom. 22 (2008) 1528. [60] A.P. Bateman, M.L. Walser, Y. Desyaterik, J. Laskin, A. Laskin, S.A. Nizkorodov, Environ. Sci. Technol. 42 (2008) 7341. [61] T. Pfeifer, U. Klaus, R. Hoffmann, M. Spiteller, J. Chromatogr. A 926 (2001) 151. [62] C.E. Rostad, J.A. Leenheer, Anal. Chim. Acta 523 (2004) 269.