Introduction – Iron Biogeochemistry in the Ocean and the Importance of Iron Speciation

Approximately 30% of surface waters in the open ocean are known as high nutrient low chlorophyll (HNLC) regions (Boyd et al., 2007). These areas are replete in the macronutrients nitrate and phosphate, but present lower phytoplankton biomass, in terms of chlorophyll concentrations, than expected from residual macronutrient concentrations (Figure 1). The restriction of phytoplankton growth in these regions is now acknowledged to be the result of iron (Fe) limitation (Martin and Fitzwater, 1988; Boyd et al., 2007). Fe is a micronutrient required for proteins involved in fundamental cellular processes, including both photosynthesis and respiration (Raven et al., 1999). Despite being the fourth most abundant element in the Earth’s crust (Liu and Millero, 2002), dissolved Fe (dFe: <0.2 or 0.45 μm) concentrations in open ocean surface waters typically fall below 0.2 nM, (De Baar and De Jong, 2001; Boyd and Ellwood, 2010) and dFe generally exhibits a nutrient-type depth profile in the ocean, depicting removal in surface waters by biological uptake, and increased concentrations at depth from remineralization processes occurring through the water column (Boyd and Ellwood, 2010). Although low, dFe concentrations in the ocean can be much higher than might be predicted given that the solubility of ferric hydroxide in seawater at pH 8.1 and 25°C has been determined to be as low as 0.01 nM (Liu and Millero, 2002). The presence of dFe at concentrations beyond the inorganic solubility of Fe is thought to be facilitated by organic complexation of Fe with stabilizing ligands, which buffer dFe in seawater against hydrolysis and ensuing precipitation (Hunter and Boyd, 2007; Boyd and Ellwood, 2010). However, the overall physico-chemical speciation of dFe, which encompasses all of its possible physical and chemical forms in seawater, is more complex than implied by the consideration of organic complexation alone. The different physico-chemical fractions of dFe include Fe(II), colloidal, truly soluble, and inorganic iron in addition to organically complexed iron. These different fractions have different environmental and biological mobility (Kuma et al., 1996; Maldonado et al., 2005; Hunter and Boyd, 2007; Kitayama et al., 2009; Boyd and Ellwood, 2010; Hassler et al., 2011b). The motivation for understanding the physico-chemical speciation of Fe, therefore, results from a desire to understand how these different fractions influence the overall biogeochemistry of Fe in the oceans.

It is important to note that dissolved iron (dFe) is operationally defined by filtration, with early studies employing 0.45 μm or, more recently, 0.2 μm membrane filters (De Baar and De Jong, 2001; Cutter et al., 2010). However, it has been shown that a significant proportion of dFe is colloidal (Fe_colloidal; Wu and Luther, 1994; Cullen et al., 2006; Bergquist et al., 2007; Kondo et al., 2008; Schlosser and Croot, 2008; Boye et al., 2010). Colloidal Fe is characterized as the difference between the Fe concentration determined in the <0.2 μm fraction (dFe) and the >1 kDa or >0.02 μm fraction, depending on whether cross flow filtration or membrane filtration techniques are used for the determination (Schlosser and Croot, 2008). The colloidal fraction is not measured directly, but inferred from the difference between dissolved (<0.2 μm) and soluble (<1 kDa or <0.02 μm) fractions.

The mass balances for Fe when considering its physical distribution can be described as

while the mass balance from a chemical perspective might be described as

where Fe′ represents labile inorganic Fe complexes, FeL represents Fe organic ligand complexes exchangeable within a time scale of <1 day, and Fe_inert represents the Fe fraction bound up in matrices that are essentially non-labile. As our analytical methods for the determination of the physico-chemical speciation of Fe tend to focus on either the physical (e.g., Schlosser and Croot, 2008; Baalousha et al., 2011) or the chemical (e.g., Gledhill and van den Berg, 1994; Rue and Bruland, 1995; van den Berg, 1995; Wu and Luther, 1995; Laglera et al., 2007; Mawji et al., 2008a; Velasquez et al., 2011) perspective, reconciling these two approaches remains a major challenge to Fe biogeochemists.

In recent years there has been a concerted effort to understand more about both the physical partitioning of Fe in the marine environment, and the chemical nature of the Fe ligand pool. The application of filtration with trace metal clean 0.02 μm pore size membrane filtration, ultrafiltration (10 kDa cut offs), and flow field flow fractionation (FFFF) coupled to ultra-violet (UV) and inductively coupled plasma-mass spectrometry (ICP-MS) detection techniques have considerably improved our knowledge of the physical partitioning of Fe in marine waters (Schlosser and Croot, 2008; Baalousha et al., 2011). Characterization of the FeL pool has been tackled through the utilization of high performance liquid chromatography–electrospray ionization-mass spectrometry (HPLC–ESI-MS) and development of novel electroanalytical techniques (McCormack et al., 2003; Laglera et al., 2007; Velasquez et al., 2011). In parallel to these advances a concerted effort is being made to improve our understanding of the robustness of competitive ligand exchange–adsorptive cathodic stripping voltammetry (CLE–ACSV), the technique most commonly used to determine Fe complexation in seawater (Buck et al., under review;Laglera et al., 2011). These advances have indicated that although the absolute physical partitioning determined varies from study to study as a result of the different techniques and filter cut offs, the colloidal Fe pool makes up between 30 and 91% of the dFe pool (Wu and Luther, 1994; Nishioka et al., 2001; Cullen et al., 2006; Bergquist et al., 2007; Hurst and Bruland, 2008; Kondo et al., 2008; Schlosser and Croot, 2008; Boye et al., 2010). The presence of Fe-binding ligands has been inferred in colloidal and measured in soluble fractions (Wu et al., 2001; Boye et al., 2010; Thuroczy et al., 2010), but both FFFF and ultrafiltration studies indicate that not all of the colloidal Fe (organic or inorganic) is exchangeable or under saturated with respect to Fe (Wu et al., 2001; Boye et al., 2010; Stolpe and Hassellov, 2010; Stolpe et al., 2010; Thuroczy et al., 2010). The existence of an inert colloidal fraction has broad implications for our understanding of Fe biogeochemistry and its significance in the ocean has yet to be properly assessed. Thus, while Fe_soluble might reasonably be expected to include Fe′, FeL, or Fe_inert are unlikely to be discreet to Fe_soluble, Fe_colloidal, or Fe_particulate.

Even though we have not fully characterized either colloidal or organic Fe associations, efforts to understand the overall effects of Fe speciation on Fe biogeochemistry have been made (Archer and Johnson, 2000; Moore et al., 2004; Parekh et al., 2005; Weber et al., 2005; Fan, 2008; Moore and Braucher, 2008; Tagliabue et al., 2009; Fan and Dunne, 2011; Tagliabue and Voelker, 2011). Modelers have made considerable progress toward capturing the complexity of iron speciation, incorporating inorganic iron scavenging and up to two ligand classes, in order to investigate the large scale implications of iron speciation on ocean productivity and the potential effects of a changing climate (Tagliabue et al., 2009; Ye et al., 2009; Tagliabue and Voelker, 2011). Fe biogeochemical models incorporating Fe speciation tend to highlight the importance of photoreduction processes in determining the dissolved concentrations of Fe in surface waters, while indicating that organic ligands are likely to strongly influence the thermodynamic solubility of Fe (Weber et al., 2005; Fan, 2008; Tagliabue et al., 2009; Tagliabue and Voelker, 2011). Furthermore, variations in organic ligand concentration and conditional stability constants have also been shown to influence the residence time and potential bioavailability of Fe in models (Tagliabue et al., 2009). Other studies have shown that different organic ligands have different chemical labilities and different susceptibilities to photoreduction (Hutchins et al., 1999; Barbeau et al., 2001, 2003; Maldonado et al., 2005; Hassler et al., 2011b). Overall, such studies highlight the need to characterize more fully the physico-chemical speciation of dFe.

In this paper, we will review current knowledge of the organic complexation of Fe, the characterization and distributions of specific FeL complexes, and the characterization of colloidal Fe fractions as a first attempt at reconciling the different approaches to understanding the physico-chemical speciation of dFe in seawater. We will not consider the reduced Fe pool, Fe(II), in this review as our understanding of the interactions between Fe(II), Fe(III), and organic matter (e.g., Toner et al., 2009; Bligh and Waite, 2010) require further investigation. However, it should be born in mind that Fe(II) is likely important to Fe biogeochemistry as it is thought to be highly biologically available and can make up 50–60% of the dFe pool in illuminated surface waters or near sediments, hydrothermal vent systems, or oxygen minimum zones (Hong and Kester, 1986; Kuma et al., 1992; Johnson et al., 1994; Gledhill and van den Berg, 1995; Croot et al., 2005; Ussher et al., 2007; Hansard et al., 2009; Sarthou et al., 2011).

Iron Complexation as Determined by Competitive Equilibrium Experiments

The organic complexation of dFe, including organic Fe-binding ligand concentrations ([L_i]) and their associated conditional stability constants is primarily measured in seawater using the electrochemical technique CLE–ACSV. The ambient Fe-binding ligands determined by this technique are typically described as ligand “classes,” which are operationally defined by the associated conditional stability constant measured. Ligand classes are denoted as L_i, where i = 1 for stronger ligand classes and i = 2, 3, etc., for progressively weaker ligand classes. Conditional stability constants for these ligand classes are expressed as either where and α_Fe′ is the inorganic side reaction coefficient for Fe, with α_Fe′ = [Fe′]/[Fe³⁺]. α_Fe′ varies with pH (Byrne et al., 1988), but a value of 10¹⁰ is commonly used for α_Fe′ in pH 8 seawater (Hudson et al., 1992; Sunda and Huntsman, 2003), although other values have been used (Gledhill and van den Berg, 1994; Rue and Bruland, 1997; Gledhill et al., 1998; Nolting et al., 1998). In CLE–ACSV, samples are buffered prior to analyses, typically to pH 8, and mostly equilibrated and measured at room temperature, so the use of an α_Fe′ = 10¹⁰ is reasonable for these cases, regardless of original sample pH or temperature. To better facilitate comparisons across studies, we encourage analysts to consider adjusting their using the appropriate α_Fe′ for the analytical pH employed. Similarly, those who report are encouraged to specify the α_Fe′ used.

Competitive Ligand Exchange–Adsorptive Cathodic Stripping Voltammetry

Aliquots of a filtered, but otherwise unaltered, seawater sample are first buffered to maintain pH near 8, and then titrated with increasing additions of dissolved Fe (dFe) from +0 to typically 10×–20× the ambient dFe concentration. It is recommended to have two +0 Fe additions in the titration, with at least eight Fe amendments, for a total of 10 or more titration points to provide the best data structure for interpretation (Garnier et al., 2004; Sander et al., 2011). DFe additions are left to equilibrate with ambient Fe-binding ligands, often for several hours, before each aliquot is amended with a given concentration of a well-characterized added ligand (AL) and allowed an additional equilibration time of minutes to hours. Once equilibrated with both added dFe and AL, each sample aliquot of the titration is analyzed sequentially by ACSV on a hanging mercury drop electrode (HMDE).

Over a deposition time predetermined to allow sufficient preconcentration of Fe(AL)_x in the sample, the Fe(AL)_x complex formed in the titrations is adsorbed at a set initial potential to the surface of the HMDE while stirring the solution. As the deposition time employed is applied to all aliquots in a given titration, it is essential that the length of time chosen is long enough to distinguish the small increases in the added dFe between vials early in the titration, but not so long that the larger amendments of dFe at the end of the titration saturates the HMDE surface. After the deposition time concludes, there is often a few seconds of “quiet time,” when stirring is stopped, before the voltage applied to the HMDE is ramped in the negative direction (cathodic stripping) following the optimal excitation signal wave form (differential pulse, linear sweep, Osteryoung square wave) for the Fe(AL)_x chosen.

At the reduction potential of Fe, roughly −0.5 V (depending on AL chosen), Fe is reduced from the Fe(AL)_x complex, or, in some cases, the Fe(AL)_x itself may reduce, generating a peak in current measured at the HMDE. The height of this peak in current, typically expressed in absolute nA units, is recorded for each aliquot as a titration point and plotted against the added dFe to generate a titration curve. This raw titration data is then interpreted using either linear (Scatchard, 1949; Ružic, 1982; Van Den Berg, 1982) or non-linear (Gerringa et al., 1995) transformations to determine ambient Fe-binding ligand concentrations and their thermodynamic conditional stability constants.

The concentration and choice of AL employed, along with the thermodynamic characteristics of the Fe(AL)_x complex formed, defines the analytical window of the analyses. This analytical window is described as where The analytical window applied constrains the measured ligand concentrations and conditional stability constants such that the end result is best described as averages of the different ligand classes identified within an analytical window. These parameters produce an where In most cases, only one ligand class is identified in Fe speciation titrations, but in the case of more than one ligand class,

While the use of multiple analytical windows has not yet been addressed for Fe speciation, there is some indication from recent GEOTRACES inter calibration efforts that the analytical window applied to Fe speciation may impact calculated results for [L_i] and (Buck et al., under review). This effect is much less pronounced, however, than has been established previously for dissolved copper (Cu) speciation (Bruland et al., 2000), likely reflecting the very different inorganic side reactions of Cu and Fe, as well as the relative breadth of the ligand pools for these elements.

With respect to ligand classes, it should be noted that analysts commonly apply the notation of “L₁” to the strongest (or only) ligand class detected in their sample, regardless of the associated . Thus, ligand class definitions may vary widely between analysts and sample data sets, complicating comparisons across studies. For example, the range of reported in the literature, for either the single ligand class identified or the designated L₁ in two ligand systems, covers more than four orders of magnitude (from ∼9 to ∼13.5; Table 1). Even in field studies distinguishing two ligand classes, the reported for the stronger ligand class range between 11.1 and 13.9, overlapping with the weaker ligand class values of 9.7–11.95 reported (Table 1; Rue and Bruland, 1995, 1997; Cullen et al., 2006; Buck and Bruland, 2007; Buck et al., 2007; Ibisanmi et al., 2011). Correspondingly, it would be useful to agree on definitions for ligand classes based on the range of reported in natural systems. Along these lines, we would propose using “L₁” for “L₂” for log and “L₃” for Regardless of whether such new definitions are employed, when comparing results between studies we implore analysts to consider making comparisons between instead of designated ligand classes.

Limitations of Interpretation

Competitive ligand exchange–ACSV is an indirect approach to determining the organic complexation of dFe. As such, there are some fundamental assumptions and limitations inherent to CLE–ACSV data interpretation, which are worth reviewing in brief before attempting to contextualize published field data. First, raw titration is interpreted assuming that ambient ligands are coordinated with dFe in a 1:1 ratio. In reality, the wide range of dissolved organic matter present in seawater likely includes ligands that may bind dFe in ratios greater than or less than 1:1. Thus, the reported ligand concentrations measured from CLE–ACSV represent the binding capacity for organic ligand classes present in the sample, not necessarily an absolute concentration of ligand molecules. As a consequence, some analysts report their measured ligand concentrations in terms of nanoequivalents (nEq; e.g., Nolting et al., 1998; Gerringa et al., 2008) or nanomolar Fe equivalents (e.g., Croot et al., 2004) instead of simply nanomolar concentrations to highlight this distinction.

A second, and more problematic, assumption in the interpretation of CLE–ACSV data is that all of the dFe in a given sample is in an exchangeable form with respect to the added competitive ligand (AL). Arising from this assumption, any dFe that does not bind to the AL to form Fe(AL)_x is then assumed in the data interpretation to be complexed by an organic ligand class with a larger than the applied. This is clearly problematic in natural samples, where much of the dFe, especially in high dFe environments, may exist as stable inorganic or organically associated colloids, or as other phases of dFe not exchangeable with AL (e.g., Fe_inert). It has been noted previously that inclusion of Fe_inert in calculations of ligand concentrations results in an overestimation of [L_i] and a shift in the toward higher stabilities (Bruland and Rue, 2001). This becomes particularly critical if ligand titrations are carried out in unfiltered samples (Gledhill et al., 1998; Thuroczy et al., 2010), but given the evidence for an inert colloidal fraction (see below), is also likely to affect determination of ligands in dissolved samples. Determinations of [L_i] and in the ocean are, thus, potentially subject to systematic errors when using the total dFe concentration in the interpretation of CLE–ACSV titration data.

To further explore the implications of overestimating exchangeable Fe, we consider the data in Thuroczy et al. (2010) and data on titrations in unfiltered samples from Gledhill et al. (1998), recalculated using the reactive Fe concentration (Fe_R) determined in the unfiltered samples (Gledhill et al., 1998). Although the datasets used were both obtained for unfiltered samples, the same systematic errors with respect to Fe_inert and exchangeable Fe will occur within the dFe pool, albeit to a lesser degree.

Plots of [L_i] and determined using Fe_T (unfiltered), against [L_i] and determined using either dissolved or reactive Fe concentration (dFe or Fe_R), are presented in Figures 2A,B. In these studies, Fe_T was determined in unfiltered samples either by flow injection analysis (FIA) after acidification (pH 1.8, samples analyzed after 1 year; Thuroczy et al., 2010) or by ACSV after acidification (pH 2) and UV irradiation (Gledhill et al., 1998). Dissolved Fe was determined by FIA after filtration (<0.2 μm) and acidification (pH 1.8, samples analyzed after 12 h, Thuroczy et al., 2010) and Fe_R was determined in unfiltered samples by ACSV and defined as the amount of Fe freely complexed by 20 μM 1-nitroso-2-napthol at pH 6.9 (Gledhill et al., 1998).

From these plots, it is clear that both [L_i] and log are overestimated if the Fe concentration used in data transformation includes Fe_inert (Figures 2A,B). An overestimation of [L_i] and is not desirable as this will result in an overestimation of the degree to which L can stabilize Fe in solution, and can thus underestimate the potential for competing biogeochemical processes such as colloidal aggregation, scavenging, and uptake. Calculation of the excess ligand (eL_i) for each case confirms results shown previously (Thuroczy et al., 2010), that this parameter is more or less independent of the Fe concentration used in the transformation, at least at low excess ligand concentrations (Figure 2C).

Figure 2D shows that overestimation of is likely to be important only in areas where the Fe_inert concentration is high. Such areas will include coastal regions as observed in Gledhill et al. (1998), hydrothermal vents, and the edge of the ice shelf (Bennett et al., 2008; van der Merwe et al., 2009; Batchelli et al., 2010; Stolpe and Hassellov, 2010; Stolpe et al., 2010). In these regions a good distinction between the dissolved Fe_inert and exchangeable Fe could be very informative, and may strongly influence our interpretation of the role of ligands in stabilizing Fe in solution (e.g., Sander and Koschinsky, 2011). Further work is therefore required in order to identify a practical method which can distinguish between Fe_inert and dFe. One approach may simply be the standard addition of dFe with over-competition with a higher concentration of AL, although this may be complicated by limits in the concentration of AL that may be used before saturation of the HMDE becomes problematic.

Clearly, the fractionation of Fe, particularly within the dissolved phase, between Fe_inert and exchangeable Fe deserves further attention. In the meantime, analysts may qualify ligand concentrations by reporting excess L ([eL] = [L] − [dFe]) as a proxy for ligand under saturation (Wu and Luther, 1995; Witter and Luther, 1998; Witter et al., 2000; Boye et al., 2001; Tian et al., 2006; Rijkenberg et al., 2008b; Thuroczy et al., 2010; Ibisanmi et al., 2011).

Distributions of Fe-Binding Ligands in the Marine Environment

The organic complexation of Fe, with typically >99.9% of dFe complexed by ligands in the marine environment, is a nearly ubiquitous feature of dFe speciation in seawater. While these Fe-binding ligands are present seemingly everywhere, from surface to deep waters of the coastal and open ocean, there are some distinguishable trends in their distributions and thermodynamic characteristics. Previous reviews have highlighted the predominant feature of higher ligand concentrations in surface waters of depth profiles, often with stronger stability constants, compared to deep waters (Bruland and Rue, 2001; Hunter and Boyd, 2007). Here we review trends of organic complexation of Fe in terms of excess ligand concentrations ([eL_i] = [L_i] − [dFe]), since total ligand concentrations ([L_i]) determined by CLE–ACSV will include stable inorganic colloidal Fe or other Fe_inert components (see above).

In most waters, Fe-binding ligand concentrations exceed dFe (Table 1). Excess ligand concentrations are commonly highest and most variable in the upper water column (Rue and Bruland, 1995; van den Berg, 1995, 2006; Witter and Luther, 1998; Boye et al., 2001, 2006, 2010; Croot et al., 2004; Gerringa et al., 2006; Hopkinson and Barbeau, 2007; Kondo et al., 2007; Rijkenberg et al., 2008b; Thuroczy et al., 2010; Ibisanmi et al., 2011; Mohamed et al., 2011), and relatively static at depth within individual profiles (Rue and Bruland, 1995; van den Berg, 1995; Nolting et al., 1998; Boye et al., 2001, 2006, 2010; Ibisanmi et al., 2011). The highest excess ligand concentrations are often associated with the fluorescence or chlorophyll biomass maxima (Rue and Bruland, 1995; van den Berg, 1995, 2006; Boye et al., 2001, 2006; Croot et al., 2004; Gerringa et al., 2006, 2008; Tian et al., 2006; Buck and Bruland, 2007; Wagener et al., 2008; Ibisanmi et al., 2011). The uptake of dFe associated with fluorescence or biomass maxima results in elevated and variable excess ligand concentrations at low dFe (e.g., Buck and Bruland, 2007).

An atmospheric source has been hypothesized for surface waters of the NE Atlantic with low chlorophyll (Gerringa et al., 2006), though a reduction in excess ligand after dust deposition was observed by Rijkenberg et al. (2008b). Excess ligands in low chlorophyll surface waters may alternatively be the remnants of a previous bloom, as excess ligands have been shown in incubation experiments to increase in proportion to chlorophyll consumption by grazers (Sato et al., 2007). In some cases, a minimum in excess ligands within the upper water column is attributed to photochemical ligand destruction at the surface (Boye et al., 2001; Croot et al., 2004), although the photochemical lability of ambient Fe-binding ligands has also been shown to be rather unpredictable (Powell and Wilson-Finelli, 2003b; Rijkenberg et al., 2006b).

Anomalously high excess ligand concentrations have been reported in shelf and bottom boundary layers (Croot and Johansson, 2000; Gobler et al., 2002; Boye et al., 2003; Buck et al., 2007; Kondo et al., 2007; Gerringa et al., 2008; Batchelli et al., 2010), as well as in river plumes (Powell and Wilson-Finelli, 2003a; Buck et al., 2007; Kondo et al., 2007). The excess Fe-binding ligands measured in these coastal waters may be humic substances (HS), which have been suggested to be an important component of the ligand pool for dFe in margin environments (Laglera et al., 2007; Laglera and van den Berg, 2009). Regardless of chemical nature, river plumes and sediment resuspension on shelves appears to be a source of both dFe and Fe-binding ligands to the marine water column (Croot and Johansson, 2000; Powell and Wilson-Finelli, 2003a; Buck et al., 2007; Gerringa et al., 2008; Batchelli et al., 2010). As with dFe, total dissolved ligand concentrations also typically decrease from shore to sea (Boye et al., 2003; Buck and Bruland, 2007).

When two ligand classes are detected, a particularly strong ligand class is reported primarily in the upper water column, and the weaker ligand class exists throughout the water column (Rue and Bruland, 1995, 1997; Cullen et al., 2006; Ibisanmi et al., 2011). When only one ligand class is measured, no trend in values (∼11–12) is reported with depth in profiles, with similar log values through the entire water column (van den Berg, 1995, 2006; Witter and Luther, 1998; Boye et al., 2001, 2006, 2010; Croot et al., 2004; Gerringa et al., 2006, 2008; Thuroczy et al., 2010). values reported for coastal and open ocean samples are within the same range (∼9–14; Table 1), as are values from river plumes (10.4–13.9; Powell and Wilson-Finelli, 2003b; Buck et al., 2007; Table 1). Transects conducted between the shore and sea found no significant trends in values with distance from shore (Boye et al., 2003; Buck and Bruland, 2007; Hopkinson and Barbeau, 2007).

Recent size fractionation studies of the dissolved ligand pool have shown that the excess ligand is observed predominantly in the soluble size fraction throughout the water column (Cullen et al., 2006; Boye et al., 2010; Thuroczy et al., 2010), but particularly in upper waters (Cullen et al., 2006; Kondo et al., 2008; Boye et al., 2010; Thuroczy et al., 2010). Colloidal ligands, on the other hand, approach Fe saturation (Gobler et al., 2002; Boye et al., 2005, 2010; Cullen et al., 2006; Gerringa et al., 2007; Kondo et al., 2008; Batchelli et al., 2010; Thuroczy et al., 2010), with the colloidal fraction of Fe, and presumably L, increasing with depth (Thuroczy et al., 2010).

The conditional stability constants of Fe-binding ligands in the soluble (<1–1000 kDa) size fraction are largely within 1 SD of the values reported in the total dissolved (<0.2–0.4 μm) size fraction (Boye et al., 2005, 2010; Cullen et al., 2006; Gerringa et al., 2007; Kondo et al., 2008; Thuroczy et al., 2010; Table 1). When two ligand classes were detected within the soluble and dissolved size fractions (Cullen et al., 2006), most of the ligands of both classes were present in the soluble (<0.02 μm) fraction, and excess ligand concentrations of both L₁ and L₂ were higher in the soluble fraction. The conditional stability constants of these ligand classes were similar in both size fractions (Cullen et al., 2006; Table 1). In mesoscale Fe fertilization experiments, Fe additions were observed primarily in the colloidal fraction (Boye et al., 2005; Kondo et al., 2008), and both Fe and ligand concentrations increased in response to Fe-enrichment (Rue and Bruland, 1997; Boye et al., 2005; Kondo et al., 2008). Excess ligand concentrations in these experiments, however, typically decreased as Fe additions saturated ligands, and some of the increase in [L_i] may reflect an artifact in interpretation. At the conclusion of the SEEDS II experiment in the NW Pacific, excess ligand concentrations increased, predominantly in the soluble fraction, as the stimulated bloom declined (Kondo et al., 2008). Shipboard microzooplankton grazing experiments conducted during the SEEDS II experiment demonstrated an increase in ligand concentrations proportional to the chlorophyll consumed by added grazers (Sato et al., 2007).

In environments high in Fe, dFe concentrations may meet or exceed measured ligand concentrations. Examples of this, where [dFe] > [L_T], have been reported in hydrothermal vent plumes (Bennett et al., 2008), artificial Fe-enrichment experiments (Rue and Bruland, 1997; Boye et al., 2005; Kondo et al., 2008), and Fe-rich coastal shelf environments (Croot and Johansson, 2000; Gobler et al., 2002; Powell and Wilson-Finelli, 2003a; Buck and Bruland, 2007; Buck et al., 2007; Kondo et al., 2007; Batchelli et al., 2010). In the Eastern Tropical North Pacific suboxic zone, dFe concentrations also approach dissolved ligand concentrations, resulting in diminished excess ligand concentrations compared to surrounding oxygenated waters (Hopkinson and Barbeau, 2007). The conditional stability constant of the ligand measured in the suboxic zone was also slightly stronger than the ligands measured in the oxic waters above (Hopkinson and Barbeau, 2007). In the suboxic zone of the Arabian sea, on the other hand, excess ligand concentrations were much greater than in the oxic waters, with a similar or slightly lower conditional stability constant of these suboxic zone ligands (Witter et al., 2000).

Links between Biological Activity and Ligand Concentrations

Previous reviews have emphasized the important interplay between biological activity and Fe-binding ligand cycling (Bruland and Rue, 2001; Hirose, 2006, 2007; Hunter and Boyd, 2007; Boyd and Ellwood, 2010). There are multiple lines of evidence in support of a biological source of Fe-binding ligands in the marine environment. As mentioned above, field studies have commonly found highest excess ligand concentrations within and around the biomass maximum in the water column (Rue and Bruland, 1995; van den Berg, 1995, 2006; Boye et al., 2001, 2006; Croot et al., 2004; Gerringa et al., 2006, 2008; Tian et al., 2006; Buck and Bruland, 2007; Wagener et al., 2008; Ibisanmi et al., 2011; Mohamed et al., 2011). Excess ligand concentrations have been observed to show an annual cycle (Wagener et al., 2008), with increases in excess ligand concentrations observed during the most productive periods. Surface transects also show elevated excess ligand concentrations with high productivity in Fe-depleted waters (Boye et al., 2003; Gerringa et al., 2006; Tian et al., 2006; Buck and Bruland, 2007). The particularly strong Fe-binding ligand class, detected only in the upper water column (Rue and Bruland, 1995, 1997; Cullen et al., 2006; Ibisanmi et al., 2011), presents the same class of log K values (log K > 12) as siderophore-type ligands measured by CLE–ACSV (Rue and Bruland, 1995; Witter et al., 2000; Buck et al., 2010; Poorvin et al., 2011).

Recent incubations of natural surface seawater have shown active production of Fe-binding ligands concomitant with diatom growth under Fe-limiting conditions (Buck et al., 2010; King et al., 2012). In these incubations, ligand production was only observed in the unamended bottles, Fe-amended bottles presented much higher growth but no ligand production, suggesting that the Fe-limitation status of the diatoms in the unamended bottles was related to the ligand production observed (Buck et al., 2010; King et al., 2012). In the field, Gerringa et al. (2006) found that 63% of the variability in ligand concentrations in profiles from the Canary Basin was explained by changes in phytoplankton biomass and silicic acid concentrations, indicating a correlation between diatom growth and Fe-binding ligand concentrations. Recent incubation studies have furthermore shown an increase in Fe bioavailability due to redox speciation changes specifically in diatom cultures (Rijkenberg et al., 2008a). In combination, these studies suggest a connection between Fe-stressed diatom communities and Fe-binding ligand concentrations that deserves further attention.

The passive production of Fe-binding ligands from grazing and bacterial remineralization of organic matter is another important biological source of ligands. HS, degradation products of terrestrial and marine organic matter, may be a substantial component of the Fe-binding ligand pool in some coastal and deep ocean waters (Laglera and van den Berg, 2009). A recent study of coastal estuarine waters found that Fe was strongly but reversibly complexed with HS, predominantly in the colloidal fraction (Batchelli et al., 2010).

Increased excess ligand concentrations, largely in the soluble size fraction, were reported during the bloom decline of the SEEDS II experiment (Kondo et al., 2008). Shipboard grazer experiments, also conducted during SEEDS II, showed that dFe and ligand concentrations increased as chlorophyll biomass was consumed (Sato et al., 2007). The highest increases in ligand concentrations, normalized to biomass consumption, were measured in the copepod grazing experiments (Sato et al., 2007). The ligands produced in these grazer experiments were similar to the L₁ type ligands and were found to be more bioavailable to diatom species than to picoplankton in the incubations (Sato et al., 2007).

The ability of ligands to solubilize natural Fe sources may also be linked to biological productivity. An annual cycle in the dissolution of Fe from Saharan dust has been observed in water sampled at different times of the year from the Mediterranean (Wagener et al., 2008), with dust derived Fe being less soluble in water sampled during the winter period (Dec–Feb). This was linked to a change in excess ligand concentration, and a potential change in ligand characteristics (although stability constants were not reported in the study), as Fe dissolution could not be determined in the winter, despite the presence of excess ligands (Wagener et al., 2008).

Bacterial remineralization of sinking biogenic particles, on the other hand, produces weaker Fe-binding ligands ( or lower) concomitant with dFe release (Boyd et al., 2010). Viral lysis of cells in grazing experiments has also recently been shown to produce similarly weak Fe-binding ligands (Poorvin et al., 2011). The weakly complexed Fe released by these passive grazing processes (Boyd et al., 2010; Poorvin et al., 2011) may be more biologically available to the phytoplankton community than the strongly complexed siderophore-bound Fe (Poorvin et al., 2011).

The Chemical Characterization of the Dissolved Iron Ligand Pool

Since the first evidence for Fe complexation by natural organic ligands in seawater was presented (Gledhill and van den Berg, 1994; Rue and Bruland, 1995; van den Berg, 1995; Wu and Luther, 1995), the question of the identity, chemical structure, and source of Fe-binding organic ligands in the oceans has fueled research across disciplines. While we have learned a great deal about Fe-binding ligands, it is still not possible to give definitive answers to these questions some 17 years later. Initially, the strength and concentration of the ligands measured by competitive equilibrium techniques indicated that ambient ligands had a high affinity for Fe (Gledhill and van den Berg, 1994; Rue and Bruland, 1995; van den Berg, 1995; Wu and Luther, 1995), while, as had been highlighted above, many profiles of ligand distributions in the ocean point to an autochthonous biological source. Determination of the conditional stability constants of potential ligand types produced by phytoplankton and bacteria indicated that some of these ligand types had very similar binding strengths in seawater to the detected natural ligands (Rue and Bruland, 1995; Witter et al., 2000; Macrellis et al., 2001). Hence these biologically produced ligand types, which may include compounds like siderophores and porphyrins, were hypothesized to make up a major part of the ligand pool. However, characterization of Fe ligands in seawater is a considerable challenge because the ligands are of unknown composition, are likely complex in chemical nature and are present at very low concentrations in a matrix of high ionic strength. The recent progress made in this field has been greatly facilitated by improvements in the sensitivity, mass accuracy and robustness of both inorganic and organic mass spectrometers. Such improvements have resulted in the detection of specific iron organic complexes such as siderophores by HPLC–ESI-MS (McCormack et al., 2003; Gledhill et al., 2004; Mawji et al., 2008a, 2011; Velasquez et al., 2011) and the detection of Fe associated with more complex organic fractions by FFFF–ICP-MS (Stolpe and Hassellov, 2010; Stolpe et al., 2010; Baalousha et al., 2011).

Small, Defined Organic Ligands

Siderophores are compounds produced by bacteria in order to sequester Fe from their environment (Hider and Kong, 2010). Siderophores are known to be produced by a wide variety of bacteria, including marine bacteria (Amin et al., 2009; Cabaj and Kosakowska, 2009; Vraspir and Butler, 2009). Bacteria appear to be able to produce families of different but related siderophores (Martinez et al., 2000, 2003; Ito and Butler, 2005; Martinez and Butler, 2007; Homann et al., 2009a,b), but uptake of siderophores by bacteria is not necessarily specific (Stintzi et al., 2000), so that a bacterial species may be able to acquire siderophores produced by other species. There has been much progress in the characterization of specific siderophores produced by marine bacteria (Amin et al., 2009; Vraspir and Butler, 2009), in particular by the A. Butler research group at the University of California in Santa Barbara.

Relatively large quantities of isolated siderophore are currently required for a complete description of chemical structure so that it is only possible to fully characterize siderophores in bacteria that can be cultured in the laboratory. In addition, the time and effort required also imposes restrictions on the number of siderophores that can be fully characterized. Characterized siderophores are, thus, likely to represent only a fraction of the potential siderophore pool. Marine siderophores produced in laboratory cultures have been found to contain all the major siderophore chelating groups of hydroxamate, catecholate, and carboxylate functional groups (Vraspir and Butler, 2009). To date, the majority of marine siderophores characterized contain mixed ligand carboxylate groups, and many of these characterized marine siderophores have fatty acid tails attached to the chelating head (Martinez et al., 2000, 2003; Xu et al., 2002; Owen et al., 2005; Martin et al., 2006; Martinez and Butler, 2007; Homann et al., 2009a,b). Such fatty acid tails are likely to strongly influence siderophore biogeochemistry, affecting siderophore partitioning between dissolved and particulate phases and perhaps mitigating against diffusive losses of siderophores from the bacterial cell (Xu et al., 2002; Martinez and Butler, 2007).

Progress has also been made in the detection and characterization of siderophores in natural seawater itself (Gledhill et al., 2004; Mawji et al., 2008a, 2011; Velasquez et al., 2011; Table 2; Figure 3). The first reports of dissolved siderophores in seawater were made by Kosakowska et al. (Kosakowska et al., 1999; Mucha et al., 1999) using capillary electrophoresis to detect hydroxamate siderophores in the Baltic Sea. Macrellis et al. (2001) used siderophore assays to report the presence of siderophore like functional groups in coastal NE Pacific upwelling waters. More recently, chromatographic techniques have been coupled to ESI or ICP-MS in order to either characterize or quantify specific siderophores in seawater (Mawji et al., 2008a; Velasquez et al., 2011; Table 2). Separation techniques coupled to mass spectrometry show much promise for both the identification and quantification of specific metal complexes like with siderophores, largely as a result of the high sensitivity of these techniques, which are capable of detecting in the picomolar–nanomolar range. Mass spectrometry has been used to detect unknown and known siderophores in the Southern Ocean and in the Atlantic Ocean (Mawji et al., 2008a; Velasquez et al., 2011). In the Southern Ocean, unidentified siderophore-type complexes were detected with the molecular masses varying from station to station (Velasquez et al., 2011). In the Atlantic Ocean, ferrioxamine type siderophores were detected ubiquitously and were found to be present at concentrations of upto 20 pM, and to make up between 0.5 and 5% of the total dissolved Fe pool (Mawji et al., 2008a). The somewhat sparse data so far obtained indicates that siderophore distributions may be linked to bacterial abundance, but otherwise gives little clue as to how siderophore concentrations might vary with, for example, depth. It is clear, however, that the presence of marine siderophores is not limited to low dFe waters.

It is notable that, while the available data is limited, the siderophores detected in seawater to date are all hydroxamates, with no reports of any identifiable siderophores containing carboxylate or catecholate functional groups being recovered from the dissolved phase in seawater. The lack of detected carboxylate or catecholate type siderophores indicates that either these siderophore types are not present in the dissolved phase, or more likely, that the analytical techniques used to detect these compounds are biased toward the detection of hydroxamate siderophores. Current methods to detect siderophores in seawater rely on a preconcentration step, necessary in order to both remove interfering matrices, and to increase siderophore concentrations to detectable ranges (Mucha et al., 1999; Mawji et al., 2008a; Velasquez et al., 2011). Such preconcentration techniques introduce bias and probably restrict the types of siderophores that can be detected. For example, at the natural pH of seawater (∼8), both carboxylate and catecholate type Fe–siderophore complexes will be deprotonated and negatively charged (Harris et al., 1979a; Loomis and Raymond, 1991). Thus, the preconcentration efficiency of catecholate and carboxylate siderophore types onto the commonly used hydrophobic resins (e.g., C18, polystyrene divinyl benzene, XAD) will be reduced. Acidification of the sample prior to preconcentration (Mucha et al., 1999; Velasquez et al., 2011), in order to neutralize negatively charged complexed siderophores and thus make them more hydrophobic may also result in siderophore hydrolysis or precipitation (Harris et al., 1979b; Loomis and Raymond, 1991). Changes in sample pH during analysis are also common, with chromatographic separations performed at low pH (McCormack et al., 2003; Velasquez et al., 2011). However, catecholate type siderophore–Fe complexes undergo hydrolysis at low pH and are not detectable when either the sample pH is low or the chromatographic conditions employ low pH eluants (Loomis and Raymond, 1991; Gledhill, unpublished data). Consequently, further work is necessary in order to develop robust preconcentration and analysis techniques applicable to a wider variety of siderophore types.

Identification of siderophore complexes in mass spectra takes place via the utilization of distinctive isotopic ratios endowed upon the complex by the metal ion (McCormack et al., 2003; Velasquez et al., 2011). Two strategies are currently employed; the first involves complexation of the siderophore with gallium, which has a very identifiable isotopic ratio for ⁶⁹Ga:⁷¹Ga of 3:2 and allows for several mechanisms of checking against false positives, increasing the robustness of the analysis (McCormack et al., 2003; Mawji et al., 2011). Such checks include analysis of the original sample for the presence of the Fe-complexed or apo- (metal free) siderophore and analysis by ICP-MS in order to check that gallium is indeed present at the expected relative retention time (Mawji et al., 2011). The gallium exchange method suffers from disadvantages in that it involves reduction of the sample pH in order to keep gallium in solution and, thus, is likely to result in losses of chemically unstable siderophores (see above).

A second promising technique recently applied to the identification of siderophores in seawater is based on the distinctive isotopic ratio endowed on a molecule by Fe (Velasquez et al., 2011). The natural abundance of the ⁵⁴Fe isotope is approximately 5.6% of the ⁵⁶Fe isotope. When incorporated into an organic complex, the combination of the Fe and carbon isotopic abundances increases the abundance of the lighter molecular ion so that ferrioxamine B will have expected isotopic ratios of 6.4:100:32.4 for m/z (M + H⁺) 612, 614, and 617. A key component of this technique is the ability to detect the putative ⁵⁴Fe containing molecular ion and then confirm it otherwise has the same structure as the ⁵⁶Fe containing molecular ion. This has been shown to be possible with a nano-HPLC coupled to a high resolution mass spectrometer (Velasquez et al., 2011). Unfortunately, the background noise in lower resolution instruments working with higher flow rates may make it difficult to identify the ⁵⁴Fe isotope, as it is present at quite a low relative abundance in MS spectra. However, even in lower resolution instruments fragment ions obtained on collision induced dissociation of pseudo molecular ions can be used to provide evidence for Fe complexes (Velasquez et al., 2011) as Fe is strongly retained in fragment ions (Mawji et al., 2008b; Velasquez et al., 2011). In theory, Fe could also be determined in parallel by ICP-MS in analogy to the gallium technique in order to confirm the presence of the metal at the expected retention time, although more care would have to be taken with respect to contamination. Furthermore the determination of Fe by ICP-MS can be more difficult due to isobaric interferences (Mawji et al., 2008a, 2011).

The limited number of reports on siderophore concentrations in the dissolved phase in the ocean indicates that the concentrations of individual siderophores are likely to be quite low, in the picomolar range (Mawji et al., 2008a). However, as has been pointed out above, while the pool of potential siderophores is apparently quite large (e.g., Amin et al., 2009; Cabaj and Kosakowska, 2009; Vraspir and Butler, 2009), it is likely that only a limited number of siderophores in seawater are detectable using the currently applied methods. Thus, although it is now clear that siderophores can be present in the open ocean, it is not possible at this stage to say what proportion of the dissolved ligand pool is made up of siderophores.

Porphyrins (molecular weight 600–1000 Da) are another biologically produced class of compounds that have been suggested as potential Fe ligands (Witter et al., 2000; Hunter and Boyd, 2007). Porphyrins function as prosthetic groups in proteins and are useful for their ability to absorb light, transfer electrons, and bind oxygen (Mochizuki et al., 2010). Porphyrins, which include chlorophylls, chlorophyll breakdown products like phaeophytin, hemes, and vitamin B12, are produced by almost all living organisms via a well conserved tetrapyrrole biosynthesis pathway (Mochizuki et al., 2010). The biological production of these compounds is tightly controlled, as they cause oxidative stress if present in excess of their proteins (Mochizuki et al., 2010). Hemes have been detected at picomolar concentrations in particulate material (Gledhill et al., under review) and heme like compounds have been detected in the dissolved fraction at nanomolar concentrations in river and estuarine waters (Vong et al., 2007). However, given that porphyrins have a very low solubility in water at pH 8 and more recent work has found little evidence to show that Fe(III) spontaneously complexes with porphyrins in solution (Rijkenberg et al., 2006a; Schlosser and Croot, 2008) it remains questionable as to whether porphyrins form a part of the dissolved ligand pool.

Domoic acid (DA), an algal toxin produced by Pseudo-nitzschia spp., has also been shown to complex Fe (Rue and Bruland, 2001), although with a weak stability constant that is unlikely to make it an effective competitor in the ocean. However, there is evidence that DA, through its ability to complex dissolved Cu, may facilitate the high affinity Fe uptake from strongly complexed Fe by diatoms in Fe-limited environments (Wells et al., 2005), regions where DA levels can be elevated (Silver et al., 2010). Thus, while possibly not a significant component of the Fe-binding ligand pool, DA may nonetheless play an important role in the biogeochemical cycling of Fe, particularly in Fe-limited environments.

Larger, Less Well-Defined Organic Ligands

Recent evidence has indicated that more diffuse, less well-defined organic compounds such as HS and exopolysaccharides (EPS) may contribute to the Fe ligand pool (Batchelli et al., 2009, 2010; Laglera and van den Berg, 2009; Stolpe and Hassellov, 2010; Stolpe et al., 2010; Hassler et al., 2011a,b). Laglera and van den Berg recently reported an electrochemical technique that allows for the direct detection of humic like substances in seawater by ACSV (Laglera et al., 2007; Laglera and van den Berg, 2009). The technique determines the catalytically enhanced reduction current produced by Fe bound to HS, and is standardized using the Suwannee River fulvic acid (SRFA) standard (Laglera et al., 2007). Determination of the conditional stability constants and binding capacities of these putative humic like substances by competition with ethylene-diaminetetraacetic acid (EDTA) has shown that these ligands have similar stability constants to the natural organic ligands measured by CLE–ACSV (Laglera and van den Berg, 2009). Apparent binding capacities for these humic like substances in seawater are also similar to those determined by CLE–ACSV (Laglera and van den Berg, 2009), and it has been suggested that HS can account for the majority of the total ligand pool in coastal and deep waters (Batchelli et al., 2009, 2010; Laglera and van den Berg, 2009). The coincidence of similar reduction potentials and binding capacities for the detected ambient Fe ligand, SRFA, and Suwannee River humic acid (SRHA) make it tempting to identify the ambient ligand detected using the technique as a humic like substance (Laglera et al., 2007). However, the technique is not specific to HS, as EPS will also become electrochemically active upon addition of Fe and are, thus, indistinguishable from HS by ACSV (Hassler et al., 2011b). Further support for identification of a ligand fraction as HS is, however, provided by FFFF. Studies in coastal waters using FFF–ICP-MS and FFFF coupled to UV and fluorescence detectors have indicated that dFe in coastal waters is associated with fluorescent, colored, dissolved humic like organic matter (Stolpe and Hassellov, 2010; Stolpe et al., 2010), and that, furthermore, freshly added Fe readily associates with this fraction (Stolpe et al., 2010).

By their very nature, both HS and EPS are challenging to chemically characterize. Both HS and EPS are oxygen rich – the SRFA used in Laglera and van den Berg (2009) consists of 44% oxygen (% w/w, http://www.ihss.gatech.edu/elements.html), while EPS are predominantly composed of neutral sugars and contain a significant fraction of acidic polysaccharides like uronic acids (Mancuso Nichols et al., 2004). Uronic acids are known to complex Fe (Gyurcsik and Nagy, 2000), with complexes being more stable at high pH (Gyurcsik and Nagy, 2000), as observed for Fe–HS complexes (Laglera et al., 2007). Classical pH-metric determinations of stability constants for simple acidic polysaccharide-Fe complexes show that they are low (log β 3–4; Gyurcsik and Nagy, 2000) and Fe complexation of model polysaccharides carrageenan, laminarin, and alginic acid in seawater has been found to be undetectable by CLE–ACSV (Strmecki et al., 2010), although this may have been an artifact of the selected AL (Laglera et al., 2011). In other studies, EPS from bacteria have been observed to be associated with Fe (Hassler et al., 2011a,b) and EPS has been shown to reduce the lability of Fe to other Fe ligands while enhancing Fe availability to phytoplankton when compared to Fe–siderophore complexes (Hassler et al., 2011b).

In addition to an Fe–HS fraction, Stolpe and Hassellov (2010) also observed an association of Fe with a higher molecular weight organic fraction described as nanofibrils, particularly at a time of enhanced productivity (Stolpe and Hassellov, 2010). The nanofibrils were consistent in size and appearance to transparent exopolymer (TEP) like compounds (Stolpe and Hassellov, 2010). However, although Fe present in the samples was found to co-elute with the nanofibril fraction, freshly added Fe did not associate with the nanofibrils. Stolpe and Hassellov (2010) suggested that this behavior could be explained as an inorganic Fe-nanoparticle association (Stolpe and Hassellov, 2010), although it could also be a reflection of under saturation of binding sites in the HS fraction, coupled to a weaker affinity between Fe and the larger nanofibrils.

Approaching an Overview of Iron Complexation and Physico-Chemical Speciation in Seawater

Several types of Fe organic complexes and associations have now been shown to exist in seawater, ranging from high affinity siderophores present at low concentrations, to weaker but more abundant HS and EPS associations, and likely also including associations between inorganic Fe nanoparticles and larger TEP like organic macromolecules (Figure 4). These organic complexes and associations are currently best distinguished by size fractionation, as they generally fall along an increasing size spectrum from siderophores to macromolecules. Often coupled to this increasing size spectrum is a decreasing thermodynamic stability of the complexes as measured by CLE–ACSV, but increasing kinetic inertness, leading to an overall potential decrease in biological availability. It is, therefore, becoming possible to unravel the different biogeochemical behaviors of the many components of the dFe pool.

Siderophores are associated with bacterial productivity and may only be produced in significant quantities when readily available carbon is abundant (Mawji et al., 2011). Siderophores maybe important in regions where organic carbon concentrations are high enough to support significant bacterial productivity, but where dFe is not readily available. The photochemical lability of siderophores suggests that many siderophore types will not persist in the water column, although their breakdown products may be more stable and form weaker, more bioavailable, Fe complexes (Barbeau et al., 2001, 2003). Siderophores may play an important role in increasing the availability of Fe for particular types of bacteria, influencing bacterial diversity and productivity in the ocean. Siderophores may also play a role in the solubilization of particulate and colloidal Fe. Given their origin, function and reactivity, siderophores are anticipated to contribute to Fe-binding ligand pools primarily in the upper water column, and especially in the truly soluble size fraction of ligands, although siderophores may also be associated with colloidal and particulate fractions as well.

Field studies of Fe-binding ligand distributions consistently report higher excess ligand concentrations in the upper water column, and these maxima in eL_i are often, though not always, associated with elevated fluorescence and phytoplankton biomass. When size fractionation is assessed, these excess ligand concentrations also appear to be predominantly in the soluble size fraction. In combination, field studies support a siderophore-type origin of ambient Fe-binding ligands in the upper oceans. However, siderophores extracted from natural waters to date have been found in only the picomolar range, while ambient strong Fe-binding ligands are measured by CLE–ACSV in the nanomolar range. Recent incubation studies of natural surface seawater have shown strong Fe-binding ligand production, also measured by CLE–ACSV in the nanomolar range, under conditions of both Fe-limited phytoplankton growth and stimulated bloom decay (Sato et al., 2007; Buck et al., 2010). Furthermore, the thermodynamic stability constants (log K′) of some known Fe–siderophore complexes are too high to detect with usual CLE–ACSV approaches. Altogether, this would suggest that, in addition to siderophores, other strong Fe-binding ligand types are also likely to contribute to the Fe-binding ligand pool measured by CLE–ACSV in surface waters.

Humic substances and EPS may be either terrestrial or autochthonous in nature, and are typically found in the colloidal fraction. It appears that the HS fraction is quite refractory and persists into the deep ocean while EPS is likely to be produced in surface waters as it is, by definition, associated with phytoplankton productivity. Fe bound to model humic acids and Fe bound to EPS have been shown to be as available to phytoplankton as inorganic Fe (Chen and Wang, 2008; Hassler et al., 2011b). However, both EPS and HS are complex molecules that remain poorly defined; greater effort is required in order to characterize the association between Fe, HS, and EPS, and the contribution of HS and EPS to the ambient ligand pool in the open ocean. It is apparent that current techniques may not distinguish particularly well between HS and EPS, yet the biogeochemistry of these two organic fractions is potentially quite different.

The biological availability of colloidal Fe is rather hard to define as it is probably highly dependent on the chemical nature of the colloids (Kuma and Matsunaga, 1995; Yoshida et al., 2006), as well as the capabilities of organisms present to extract Fe from the colloids, whether by reduction (Maldonado and Price, 2001; Rijkenberg et al., 2008a) or by dissolution via a high affinity uptake mechanism (Vraspir and Butler, 2009). The biological availability of some colloids, like Fe nanoparticles, is poorly understood. Colloidal Fe biogeochemistry is also likely to be strongly influenced by temperature, another consideration that has yet to be investigated fully. However, it is becoming apparent that Fe biogeochemistry in the open ocean may be strongly influenced by the stability of the colloidal fraction. While it has been suggested that organic Fe colloids formed in surface waters are scavenged quite rapidly through the water column (Wu and Luther, 1994; Nishioka et al., 2001; Wu et al., 2001; Cullen et al., 2006; Bergquist et al., 2007; Hurst and Bruland, 2008; Kondo et al., 2008; Schlosser and Croot, 2008; Boye et al., 2010; Thuroczy et al., 2010), stable Fe colloids from continental shelves have also been implicated in the long range transport of Fe to the open ocean (Elrod et al., 2004; Lam and Bishop, 2008). More information is, thus, needed on the nature of Fe colloids in open ocean surface and continental shelf waters in order to understand more fully these potentially contrasting behaviors. Investigation of an organic Fe colloid-like fraction in future global ocean models incorporating Fe speciation would also allow the role of this fraction in the Fe biogeochemical cycle to be further examined.

Taken together, the evidence described above points toward the existence of a spectrum of Fe ligands in seawater, and we found that the different experimental approaches employed by Fe biogeochemists appear to be converging toward a consistent overview. Increases in the size and complexity of the Fe species appear to be accompanied by a weakening of the thermodynamic stability of the Fe complex. The weaker complexation between Fe′ and the larger sized organic macromolecules is, however, potentially mitigated by the existence of associations between inert Fe nanoparticles and nanofibrils like TEP, resulting in an apparently Fe saturated inert colloidal ligand pool. Substantial progress has been made in trace metal clean size fractionation, CLE–ACSV, and HPLC–ESI-MS techniques over the last decade. We anticipate that future studies will increasingly couple these techniques, along with incubations and other biological process studies, providing much needed insight on the nature and cycling of Fe-binding ligands in the oceans. Finally, we conclude that the biogeochemical role of each of the organic ligand fractions appears to be very different and deserves further investigation, as this spectrum of organic Fe-binding ligands overwhelmingly complexes dFe in the marine environment.

Conflict of Interest Statement

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Acknowledgments

Gledhill was supported by UK – NERC Advanced Fellowship (NE/E013546/1) and contributed to the manuscript whilst a visiting scientist at the Royal Netherlands Institute of Sea Research, Texel, The Netherlands. Buck was supported by institutional funding from the G. Unger Vetlesen Foundation at the Bermuda Institute of Ocean Sciences (BIOS). This manuscript constitutes BIOS contribution number 2017 for co-author Buck.

References