Metal–organic complexation in the marine environment

We discuss the voltammetric methods that are used to assess metal–organic complexation in seawater. These consist of titration methods using anodic stripping voltammetry (ASV) and cathodic stripping voltammetry competitive ligand experiments (CSV-CLE). These approaches and a kinetic approach using CSV-CLE give similar information on the amount of excess ligand to metal in a sample and the conditional metal ligand stability constant for the excess ligand bound to the metal. CSV-CLE data using different ligands to measure Fe(III) organic complexes are similar. All these methods give conditional stability constants for which the side reaction coefficient for the metal can be corrected but not that for the ligand. Another approach, pseudovoltammetry, provides information on the actual metal–ligand complex(es) in a sample by doing ASV experiments where the deposition potential is varied more negatively in order to destroy the metal–ligand complex. This latter approach gives concentration information on each actual ligand bound to the metal as well as the thermodynamic stability constant of each complex in solution when compared to known metal–ligand complexes. In this case the side reaction coefficients for the metal and ligand are corrected. Thus, this method may not give identical information to the titration methods because the excess ligand in the sample may not be identical to some of the actual ligands binding the metal in the sample.


Introduction
In the last two decades, our knowledge of trace metal speciation has grown tremendously. With the advent of trace metal clean sampling techniques 1 and sensitive voltammetric techniques, 2-4 the marine community now recognizes that metal speciation in seawater and estuarine waters is dominated by complexation with organic compounds of unknown composition and origin. [5][6][7][8][9][10][11][12] Recent culture work [13][14][15][16][17][18] has shown that microorganisms produce a variety of low molecular weight organic compounds that complex metals with high stability constants. These compounds have a variety of functional groups that include phosphate, carboxylic acids, amines, thiol and hydroxy groups. Specific functional groups such as hydroxamate, catecholate and b-hydroxyaspartate are bidentate groups and organisms make molecules with three bidentate groups in a molecule. 14,[19][20][21] In addition, plant degradation products [22][23][24][25][26][27][28][29][30] such as porphyrins are significant organic ligands that bind metals through four N atoms in a square planar arrangement. These latter multidentate molecules have very high stability constants with metals and are also kinetically inert to metal-ligand dissociation processes. [31][32][33][34] For this reason, organisms generally uptake the free metal ion rather than a metal-ligand form. 35,36 Thus, an understanding of metal-organism interactions requires an understanding of the amount of dissolved free ion present relative to the total dissolved metal concentration as well as the metal acquisition methods that an organism can use. [35][36][37] In this paper we review and compare the principal voltammetric methods, which provide evidence for metalorganic complexes. Most voltammetric work is performed with the hanging mercury drop electrode (HMDE) or the rotating disk electrode (RDE) with a thin mercury film (TMF) because these permit the measurement of metal-organic complexation at (sub)nanomolar levels directly in the solution of interest. The actual experimental methods can be broken into two broad categories and are based on the electrochemical behavior of the metal bound to an organic ligand.
The first method consists of titration experiments that measure the amount of ligand in excess to the metal in the solution [38][39][40][41][42][43][44] and the conditional stability constant, K cond M'L , for the excess ligand with the metal. The K cond M'L is generally assumed to be a 1 : 1 metal-ligand complex and is given by The titration experiments include (1) anodic stripping voltammetry 2 (ASV), which is useful for metals that react at the electrode directly (Cu 2z , Zn 2z , Cd 2z , Pb 2z ), and (2) cathodic stripping voltammetry/competitive ligand exchange 3,8,9 (CSV-CLE) which is useful for metals that do not react at the electrode directly but have a metal-ligand complex that does (Fe 3z , Co 2z ). The CSV-CLE method depends on the measurement of a known metal-ligand complex (the competing ligand), that adsorbs to the mercury electrode. In addition, a kinetic CSV-CLE approach 10-12 for excess ligand binding a metal has been used to measure the metal organic formation rate constant (k f ), dissociation rate constant (k d ), the half-life or residence time (t 1/2 ) of the complex and K cond M'L (from k f /k d ). The second type of voltammetry method involves the breakdown of the actual complex in situ and is termed pseudovoltammetry, [45][46][47][48] which is useful for metals that react at the electrode directly. This method gives information on the amount of ligand binding to a specific complex with a thermodynamic constant, K therm , that differs from K cond ML . K cond ML is corrected for the side reaction coefficient of the metal but not the ligand whereas K therm is corrected for the side reaction coefficients of the metal and ligand via comparison to metal-ligand complexes of known K therm (chelate scale).
We describe the use of these methods for unknown ligands in seawater as well as with model ligands in UV irradiated seawater for the metals Cu(II), Zn(II) and Fe(III). In the case of CSV-CLE, we show for known Fe(III)-organic complexes that the use of different ligands [1-nitroso-2-napthol, or 1N2N, and salicylaldoxime, or SAL) gives comparable K and ligand concentration data.

Experimental
The details of the experimental procedures for ASV and pseudovoltammetry work have been previously described by our group. 45,46 Total Zn and Cu concentrations were performed using the method of Bruland et al. 1 CSV-CLE and kinetic Fe(III) measurements with 1N2N were performed as we have outlined previously. [10][11][12] CSV-CLE experiments with SAL were performed as described by Rue and Bruland. 6,7 Examples of model ligands commonly used in experiments are given in Appendices 1 and 2. Appendix 2 shows types of strong ligands (functional groups are circled) that bind to Fe(III) and which may bind to other metals.

Metal-ligand complexes
Voltammetry can provide information on a ligand actually binding a metal because many metal-ligand complexes give a discrete peak or half-wave potential. In a sample, these peaks can be compared to known ligand-metal complexes in the form of a metal-chelate scale (see pseudovoltammetry below). Fig. 1A shows the voltammetric reduction of inorganic Zn(II) in UV irradiated seawater (E p~2 1.05 V) and Fig. 1B shows the reduction when Zn(II) is bound to two nitrilotriacetic acid molecules (NTA; E p~2 1.52 V). The reduction is more negative for the Zn complex with NTA than for inorganic Zn(II) because the ligand donates electrons more strongly than simple monodentate ligands such as chloride and hydroxide. In addition, two to four atoms in one NTA molecule can bind to Zn(II) and the displacement of monodentate inorganic ligands by multidentate ligands gives rise to higher stability constants via the ''chelate'' effect which is an entropy driven reaction; i.e., there are more product molecules than reactant molecules for the reaction 49 (generalized eqn. (1) and (2); charges omitted for simplicity).
where L-L indicates a bidentate ligand Every ligand that reacts with a metal can have a unique reduction potential that can be used for analysis and this is the basis for both the CSV-CLE and pseudovoltammetry approaches.

ASV titrations
We first discuss the titration approach for the measurement of metal-organic ligand complexes for metals that react directly at the Hg electrode (ASV experiment). In titration experiments, metal is added to an unknown sample and the inorganic form of the metal (e.g., Fig 1A for inorganic Zn indicates that the deposition potential should be more negative than 21.1 V) is analyzed via deposition experiments for possible reaction at the Hg electrode. More than 95-99% of the metal is normally bound to an unknown organic compound(s), which is in excess to the metal in the sample. Fig. 2A shows that the inorganic Zn reduction peak from a Delaware Bay sample is suppressed until the excess ligand has been titrated by the addition of inorganic Zn. Linearization of the titration data is typically performed by use of the Langmuir or Ruzic transformation [38][39][40][41] first transition series in seawater is usually v0.2 log K units). 50,51 In Fig. 2B, the linearization plot for data in Fig. 2A shows that there is a single straight line showing only one complex with a C L~3 6.1 nM and a log K cond ML~9 .03. The Scatchard transformation is given by eqn. (4) and shown in Fig. 2C for the data in Fig. 2A.
gives a slope which is K cond ML and [C L ] is the x-intercept for the regression line. In this linearization, two separate slopes are noted with a total ligand content of 37.1 nM. These data suggest that two ligands or ligand classes may be present in the sample. By convention, L 1 is the stronger ligand with a higher log K cond ML of 9.13 (concentration~33.7 nM) and L 2 is the weaker ligand with a smaller log K cond ML of 8.89 (concentration is 37.1 2 33.7~3.4 nM). The Scatchard transformation is usually the better of the linearization methods for determining separate ligand classes especially when the log K cond ML data are similar. More recently, non-linear methods 42 have been gaining popularity.
It is important to reiterate that the K cond ML data cannot be corrected for the side reaction coefficient of the unknown ligand in samples. Bruland 2 showed that the Zn-EDTA complex has a log K cond ML~7 .9 in UV seawater but log K cond ML w 11 in 0.1 M KCl of the same pH. The difference in these constants is due to the interaction of Ca and Mg in seawater with the carboxylic acid functional groups of EDTA. However, the actual thermodynamic constant for Zn-EDTA is log K therm~1 6.3. The fact that a log K cond ML w 11 is calculated indicates that there is a titration window for these types of ASV titration experiments. The titration window depends on the concentration of the unknown ligand and the metal. In general, there is a window of about six log K units for these types of titrations.

CSV-CLE titrations
Any known metal-organic complex, which gives a voltammetry signal, can be used to study the interactions of that metal with an unknown ligand(s) in a sample. In this example, the known ligand is a competitive ligand, one competing for the metal in a sample. This approach must be used for metals such as Fe(III) 3,6,12 and Co(II) 8,9 that do not react directly at the mercury electrode. Several studies have also used this approach for metals 40,41,44,46 such as Zn and Cu, which can be measured at the electrode. Comparison of the ASV and CSV-CLE methods 44 for these metals shows similar [C L ] and K cond ML data.
In the CSV-CLE case, metal in increasing concentration (from zero added metal) is added to a series of electrochemical cells containing the sample with the same amount of a competitive ligand. After analyzing each electrochemical cell, a plot similar to Fig. 2A results. Linearization of the data is given in eqn. (5), which is identical to eqn. (4) except for a ML , which is the side reaction coefficient for the metal with the competitive ligand. Much work has been performed to understand Fe(III) speciation in seawater. For Fe(III), the K cond Fe(III)L of a Fe-natural ligand complex and total natural ligand concentration [ (10 8.4 at pH~7; 10 10.0 at pH~8) 34 and a ML is the side reaction coefficient for Fe(III)L competitive ligand complexes. For Fe(III) with 1N2N, 3,12,52 the a Fe1N2N is about 10 13.04 at pH~7 and 8. For salicylaldoxime, 6 the a FeSal is about 10 2 at pH~8. The window for determination of K cond Fe(III)L is smaller that the ASV method (about two log units) but can be varied by changing the ligand concentration. The low a FeSal for salicylaldoxime indicates that the K cond Fe(III)L calculated is dependent on the accuracy of a Fe' used. Byrne et al. 53 have estimated a value of a Fe' of 10 11.5 so K cond Fe(III)L can vary 1.5 log units based on the a Fe' used. Fig. 3 shows the log K cond Fe(III)L data for several model ligands in UV seawater determined by CSV-CLE titrations with the two competitive ligands (1N2N and SAL). In these calculations 54 an a Fe' of 10 10.0 at pH~8 was used. The data show that the log K cond Fe(III)L data are similar-usually within one log K unit-when using either competitive ligand. The use of 1N2N at pH~7, where the Fe1N2N voltammetric peak is most sensitive, does not compromise the data. The main reason for this is the high a Fe1N2N when compared to the a Fe' of Fe(III) at these pH values. The vertical lines in Fig. 3A and 3B show the range of reported Fe(III)L log K values from the world's oceans. Fig. 3 also shows that the model ligands binding Fe(III) give similar log K cond Fe(III)L data regardless of their structure. This will be discussed below.

Kinetic approach
This approach has been used to assess the rate constants for formation and dissociation of Fe(III)L complexes. The approach is briefly described but is detailed elsewhere. [10][11][12]55 Excess Fe' is added to a sample without any competitive ligand so that the excess Fe' can bind to the organic ligand(s) in seawater (eqn. (6)). Aliquots of this solution are measured over time at the pH of the sample after addition of a competitive ligand to the aliquot. The k f (rate of formation of FeL is determined from this experiment) for the excess ligand binding to Fe' is determined by kinetic analysis of the time course.
The k d and t 1/2 are determined by recovering Fe' in FeL by adding a competitive ligand such as 1N2N to an equilibrated sample (eqn. (7)). This is monitored over time.
Eqn. (7) can be broken into two eqn. (8) and (9) FeL A Fe' z L The k d is evaluated using the steady state approximation for Fe' which simplifies the kinetic expression to ln[FeL]~k d t. The K cond Fe'L~kf /k d and K cond Fe(III)L~Kcond Fe'L (a Fe' ) where a Fe'~1 0 10 . Fig. 4 shows the log K cond Fe(III)L data obtained from the kinetic approach at pH~8 and the CSV-CLE approach for model ligands bound to Fe(III) in UV seawater. The agreement is excellent indicating that both methods give comparable results. To date the window for K cond Fe(III)L using this method is log K 18-23. In addition to the stability constant data, the kinetic data for Fe'L (Table 1) reflect the fast reaction rates via k f and slow dissociation rates via k d . The t 1/2 and residence times for Fe(III)L complexes come directly from the k d data (t 1/2 6 k d~0 .693) and correlate with other estimates of iron residence times in the ocean. 56,57 These CSV-CLE and kinetic data as well as solubility data 58,59 indicate that Fe(III) is primarily complexed by natural organic compounds in seawater.

Pseudovoltammograms and chelate scales
Metal reduced to an amalgam; e.g. ZnL z 2e 2 A Zn(Hg) z L. When a metal-ligand complex is reduced to a metal amalgam, the half-wave potential of a metal complex, E 1/2 ', or the peak potential, E p , can be directly related to the thermodynamic stability constant, K therm , 45-48 by eqn. (10): where E 1/2 is the reduction potential of the free metal ion and n is the number of electrons involved in the process and K therm~Kox~   simplicity). K therm is corrected for ionic strength effects, the side reaction coefficients of the metal and the ligand in the solution of interest and is a pH independent constant. This particular form of the Lingane equation assumes: (a) no dependence on the reduced metal since it is an amalgam; thus, the complex is destroyed which is a measure of the bond strength and K therm ; (b) E 1/2 ' is independent of ligand concentration, which can be checked by titrating the metal with ligand until no further change in E 1/2 ' is observed. At trace concentrations for metal-ligand complexes, pseudopolarograms or pseudovoltammograms are recorded by performing a stripping experiment. Deposition experiments are performed over a range of potentials and the current recorded at each potential. The range of potentials should provide current values where the analyte is and is not electroactive. Plots of i vs. deposition potential (E dep ) give an ''s''-shape Nernstian curve from which E 1/2 can be evaluated. For complexes which give a discrete E 1/2 based on the pseudopolarogram, E' 1/2 for the complex is directly related to the decomposition of the metal-ligand complex via K therm (eqn. (11)) E' 1/2~E1/2 z [RT ln K therm ]/nF (11) where E 1/2 is the potential of the analyte in the absence of complexation by any organic ligand in the matrix of interest. A plot of E' 1/2 vs. ln K therm for a series of metal-ligand complexes can be constructed from the literature or from experiment to derive information on K therm from unknown ligands in natural samples. Fig. 5A shows SWV peaks for 100 micromolar Zn(II) with NTA and Fig. 5B shows a pseudovoltammogram for 10 nanomolar Zn(II) with NTA. The data are similar for the Zn(NTA) 2 complex. Inorganic Zn(II) varies because of the much higher concentration in Fig. 5A than Fig. 5B. These data show that inorganic ions in seawater do not bind high concentrations of Zn(II) effectively. Based on these types of data, a chelate scale (E 1/2 vs. log K therm ) can be constructed for Zn with a variety of ligands. Fig. 6 shows a scale for seven known ligands 45 covering the range of log K therm 4 to 18. These data indicate that the window for estimation of log K therm data is much larger for the chelate scale approach than the ASV titration approach. The upper limit for log K therm for Zn as well as other metals is controlled by sodium ion reduction which begins near 21.75. Fig. 7A and Fig. 7B show pseudovoltammograms obtained from rainwater (September 5, 1992) and seawater (June 26, 1992) from the mouth of Delaware Bay with the Atlantic Ocean. The rainwater sample shows no organic complexation for Zn(II) whereas the seawater sample shows that there are two moderate-strength ZnL complexes at -1.24 V (log K therm7 .77) and 21.40 V (log K therm~1 1.45, respectively. A possible  weak third-ligand complex at 21.082 V (log K therm4 .14 M 2 1) is due to inorganic ligands and/or weak acids such as oxalate. The Zn concentration bound to each of these complexes in increasing negative potential is 1.7, 0.90 and 3.5 nM (5.7 nM combined based on the Zn peak sensitivity) whereas the total Zn concentration in the sample is 24.7 nM. Thus, 19 nM of complexed Zn compounds are still unaccounted for. This could be due to strong organic complexes (log K therm w 18) or multinuclear sulfide complexes 60 which have been found in natural waters that have log K therm w 40. These Zn-ligand complexes cannot be determined by the pseudovoltammetry approach because of sodium ion reduction, which permits an upper limit of log K therm~1 8 for Zn.
These data are now compared with the ASV titration approach shown in Fig. 2. The latter method indicates that one complex (perhaps a second) with ligand in excess to the metal is present with a value for the conditional log K cond ML~9 .03.
The conditional K cond ML and K therm data are not readily comparable for Zn(II) because K therm data are due to the actual ligand complexes in the sample and K cond ML data are for the ligands in excess to the metal in seawater. The actual ligands binding Zn may or may not be the same as the excess ligands to total Zn in the sample. The log K therm data that are less than 9.03 are weak complexes that are not detected by both Langmuir and Scatchard linear transformations. The complex with log K therm~1 1.45 (close to the ZnEDDA complex) 45 may not be related to the log K cond ML data of 9.03 either because the actual ligand concentration binding Zn in this case via the pseudovoltammograms is smaller than the ASV titration calculation of 36.1 nM. Thus, the two methods appear to be giving information on different Zn complexes.
A similar approach has been used for Cu(II) 46 as shown in Fig. 8. In that study, 17 known organic ligands were used to develop a chelate scale with a log K therm range of 12-26.5. The upper limit for this scale based on the sodium reduction wave is log K therm~4 9. Interestingly, the largest log K therm value for a CuL ligand is smaller than the estimated CuL data from field and culture samples (E' 1/2 is more negative for the field samples) demonstrating that very strong CuL complexes can be formed. The strong CuL complex found in Martha's Vineyard waters was matched by a ligand produced by a strain of Synechoccous. The moderately strong CuL complex in Eel's Pond and in Martha's Vineyard waters did not match the ligands from other cultures. The three cultures tested showed a great variability of ligands that can be produced by different organisms. The log K cond Cu(II)L for these complexes as determined by ASV titration ranged from 10.8 to 14.3. These conditional constants indicate that the side reaction coefficients for the ligand(s) are high and similar to what has been observed for ligands that form Fe(III) complexes, which are discussed below.
Reduction of a metal complex to a lower valency (no amalgam formation) Fe 3z L z e 2 < Fe 2z L. Similar chelate scale data can be obtained for metal complexes which do not decompose at the electrode to form metal amalgams. 47,48 In this case, E' 1/2 is proportional to the ratio of the thermodynamic stability constants of the reduced and oxidized complexes according to eqn. (12): (12) where K ox and K red are the stability constants of Fe 3z L and  Fe 2z L, respectively. If the K red values for all complexes are similar as shown for Fe(III) then the K red term can be incorporated into the intercept and the equation simplifies to eqn. (13): For this case the electrode processes are reversible (checked by CV or SWV) because the complex does not dissociate or become destroyed, and E' 1/2 is independent of the ligand concentration (check by titrating the metal with ligand until no further change in E' 1/2 is observed). A chelate scale for Fe(III) 47 has been developed using seven natural ligands (Table 2 and Fig. 9) which react with Fe(III) to form complexes spanning 20 log K therm units. For this example we discuss the binding properties with regard to eqn. (2). The first 3 complexes contain CDTA or NTA with or without a catechol. The low E p and log K reflect that carboxylic acids do not bind (lower DH) with Fe(III) like the other complexes which contain only catechol functional groups. Fe(cat) 2 2 is stronger than these but weaker that the tris-catechol complexes. The nitro-catechol binds more weakly in the tris complex to Fe(III) than catechol because the nitro group is an electron withdrawing group. The enterobactin complex with Fe(III) shows the stronger binding effect of one molecule with three catechol groups than three separate catechol ligands. This is related to entropic effects via the ''chelate'' effect. 49 For example, Fe(cat) 3 32 and Fe(ent) 32  The larger log K therm reflects that DG is controlled by DS in Fig. 8 A plot of E' 1/2 from pseudovoltammograms vs. log K therm for Cu(II)-organic complexes dissolved in seawater. Blue symbols are from cultures and green symbols are from natural waters as indicated.   Table 2. eqn.(14b) because the DH term for catechol functional groups is similar for eqn.(14a) and (14b). These data suggest that k d for the Fe(ent) 32 containing a tris-catechol structure is smaller than for three separate catechol groups in Fe(cat) 3 32 . Because K~k f /k d , K and DG increase with smaller k d . 49 Fig. 9 also shows data for catechol ligands produced by different organisms. 47,48 Mytilus edulis produces a 100 kDa foot protein (mefp1) which contains several catechol groups. It is not known how many catechol groups bind to Fe(III) in the protein but E p and log K therm are larger than the bis-catechol Fe(III) complexes. Tryptic digests of the foot protein produce decapeptides that react to form FeL 2 , bis-catechol complexes, as in Fig. 10. These bind more strongly with Fe(III) than two catechols in Fe(cat) 2 2 and this stronger binding appears related to interaction of the peptide chains with each other which helps to lower k d . Similar results 47 have been noted for the bis(catechol) complex of alterobactin-B from Alteromonas luteoviolacea.
These data show that the known ligands have a significant difference in log K therm . However, these ligands and other known Fe(III) binding ligands have remarkably similar log K cond Fe(III)L values ( Fig. 3 and 4) despite having different structures. This suggests that proton loss from the ligand and not Mg, Ca binding are important for the binding of Fe by the ligand. The side reaction coefficients for these ligands differ in such a way that when correcting for proton effects, the K therm data is different. The model ligands range from catecholate groups, which have 2 protons per functional group (6 total for enterobactin), to one proton per functional group for hydroxamates (3 total for desferrioxamine because there is no proton attached to the CLO group). The three proton difference for enterobactin (also alterobactin-A) vs. that for desferrioxamine leads to a different K therm . In addition, porphyrins have 2 protons per functional group and all four N atoms can bind Fe. The effect of losing 6 protons from Alterobactin-A, 3 protons from desferrioxamine and 2 protons from a porphyrin lead to similar log K cond Fe(III)L .

Conclusions
The ASV and CSV-CLE methods for the determination of organic-metal complexation give similar results for K cond ML and ligand concentration. These data relate to the ligand in excess to the metal in solution. ASV titrations have a larger K cond ML window than CSV-CLE, which can be varied by changing the competing ligand concentration. The excess ligand may or may not be the same as the actual ligand in the metal-ligand complex in the sample. The pseudovoltammogram method gives K therm and ligand concentration data on the actual complex(es) in solution within the window limited by the reduction of sodium ion. The data from the titration methods and the pseudovoltammogram data are not necessarily similar as shown for Zn. For Fe(III), the choice of competitive ligand for the CSV-CLE methods does not appear to affect the log K cond Fe(III)L data. The kinetic approach also agrees with the CSV-CLE methods. These similarities are due to measuring the same ligand types; i.e. excess ligand to the metal in the sample.