Environ. Sci. Technol. 2010, 44, 650–656
Discrete Site Surface Complexation Constants for Lanthanide Adsorption to Bacteria As Determined by Experiments and Linear Free Energy Relationships B R Y N E T . N G W E N Y A , * ,† MARISA MAGENNIS,† VALERIE OLIVE,‡ J. FRED W. MOSSELMANS,§ AND ROBERT M. ELLAM‡ Microbial Geochemistry Laboratory, School of GeoSciences, University of Edinburgh, West Mains Road, Edinburgh EH9 3JW, Scottish Universities Environmental Research Centre, Rankine Avenue, East Kilbride, G75 0QF, and Diamond Light Source Ltd., Diamond House, Chilton, Didcot, Oxfordshire OX11 0DE
Received May 13, 2009. Revised manuscript received November 13, 2009. Accepted November 23, 2009.
Bacteria are abundant in many natural and engineered environments where they are thought to exert important controls on the cycling, mobility, bioavailability, and toxicity of metal contaminants. In order to probe their role in moderating the behavior of lanthanides, pH-dependent adsorption edges of 13 individual lanthanides and yttrium to the Gram-negative bacterium Pantoea agglomerans were used to generate discrete site surface complexation constants. The calculated surface complexation constants were compared with stability constants estimated using linear free energy relationships based on a number of hydroxyl-containing ligands. The experimental data suggests that lanthanide adsorption edges below pH 6.5 are consistent with adsorption to phosphate groups for the light and some of the middle lanthanides (La to Gd), whereas some of the middle and heavy lanthanides appear to favor carboxyl co-ordination (Tb to Yb), although exceptions occur in each grouping. The experimentally derived surface complexation constants for carboxyl coordination were of similar magnitude to stability constants estimated from linear free energy correlations using fulvic acid stability constants. The implication is that the adsorption of lanthanides to bacterial surfaces could be modeled reasonably well using lanthanide stability constants for natural organic matter, except perhaps at low pH where phosphate binding dominates.
1. Introduction A large body of experimental work has shown that the adsorption of major and transition metals to bacterial surfaces is significant and likely plays a major role in the cycling, mobility, bioavailability, and toxicity of such metals in subsurface and aquatic systems (1-8). By contrast, relatively few experiments have been conducted on the adsorption of * Corresponding author: E-mail:
[email protected]; tel: +44 131 650 8507/8505; fax: +44 131 668 3184. † University of Edinburgh. ‡ Scottish Universities Environmental Research Centre. § Diamond Light Source Ltd. 650
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010
lanthanides to bacterial cells. These studies generally indicate that cellular uptake of lanthanides by bacteria is minimal (9) but that surface adsorption dominates (10), with strong surface complexation constants (11). Thus bacteria are likely to exert important controls on the mobility, cycling, and fractionation of lanthanides in the environment. Unfortunately, the large number of elements involved has meant that to date, no single study has investigated the adsorption of all the lanthanides to determine their thermodynamic surface complexation constants, although constant pH distribution coefficients for the simultaneous adsorption of all 15 lanthanides have been reported (12). Given their importance in (i) interpreting geochemical processes (13), (ii) pollution source tracing (14, 15), and (iii) biological activity (16), the availability of such data is critical to maximize their utility in modeling these processes. The enhanced mobility of lanthanides (and by implication of actinides for which lanthanides act as surrogate tracers) upon complexation with natural organic matter (17), combined with recent assessments that bacterially derived biomass constitutes more than 50% of soil organic matter (18), provides additional motivation for acquiring these data. In the absence of such data, we measured adsorption edges for 13 lanthanides plus Y and used surface complexation modeling to generate a suite of stability constants with the Gram-negative bacterium Pantoea agglomerans. The calculated stability constants were compared, where possible, with values estimated using linear free energy relationships (19). Although linear free energy relationships have been developed for metal-bacteria discrete site stability constants (6, 20), our study is the first to use the resulting equations in a predictive capacity.
2. Experimental Methods and Data Sources 2.1. Lanthanide Adsorption Experiments. We conducted sorption experiments with all the lanthanides (except Pm and Tm) and Y as a function of pH using lyophilized cells of the Gram-negative bacterium Pantoea agglomerans. All experiments used a single batch of lyophilized cells in order to avoid interculture variability reported in studies that use fresh cells (21). Viability tests using LIVE/DEAD BacLight molecular probes have shown that most of the cells (>90%) are viable after this treatment (22). For each element, at least one adsorption edge with a nominal 0.2 g/L biomass and 4 ppm lanthanide concentration was performed. In order to ensure that the modeled values were robust, a second set of experiments with 2 ppm lanthanide starting concentration and 0.2 g/L biomass was also carried out for selected elements (La, Nd, Sm, and Yb). Experimental suspensions were made from a stock suspension of lyophilized cells initially rehydrated for 1 h in 0.01 M NaClO4. Cells were then rinsed in the electrolyte three times, each followed by centrifugation at 23,420g for 10 min. After the final rinse, enough electrolyte was added to dilute the cell suspension to the desired concentration, followed by addition of metal from 1000 mg/L stock solutions in 1% HNO3. The pH of this stock suspension was then adjusted upward in 0.25 pH steps using 1 M NaOH, and at each pH, 20 mL was transferred into acid-cleaned, 50 mL polycarbonate centrifuge tubes and equilibrated for 3 h on a rotating carousel. This equilibration time was determined from time course experiments which showed that constant, reversible adsorption was attained between 1 and 3 h (22). Prior to addition of metal to the stock suspension, two 5 mL subsamples were transferred into preweighed glass vials and evaporated to constant weight in order to determine the exact biomass concentration, after correcting for a 5 mL electrolyte 10.1021/es9014234
2010 American Chemical Society
Published on Web 12/15/2009
blank. Previous studies with P. agglomerans in our laboratory have shown that this strain produces soluble organics around circum-neutral pH values (23), which can decrease sorption density. Thus, by starting with low pH, the impact of these soluble organics was restricted potentially to circumneutral pH suspensions. Nevertheless, adsorption reversibility experiments have shown that this is not a problem for pH values below 6.5, and further tests showed that there was no difference in adsorption between fresh and freeze-dried cells (22). Suspension pH was measured at room temperature (23 ( 1 °C) using a glass combination electrode after a threepoint calibration using Merck buffers of 4.00, 7.00, and 9.22. The response of the pH probe was checked against a pH 4.00 buffer made in 0.01 M NaClO4 and recorded a pH of 3.99 ( 0.01. Experiments were limited to the acidic pH range (e6.5) where speciation calculations using hydrolysis constants from Klungness and Byrne (24) showed that between 97% (Yb) and 99% (La) of the lanthanide was in the form of the hydrated trivalent ion and the rest as LnOH2+. Sampling involved pelleting the cells and filtering ∼10 mL of the supernatant through a 0.2 µm syringe filter into an acid-cleaned bottle. These solutions were acidified to 2% v/v HNO3 and stored at 4 °C before metal analysis by ICP-MS. Solutions were analyzed using a VG Elemental PlasmaQuad II+ Quadrupole mass spectrometer after appropriate dilutions in 5% HNO3, with metal concentration calculated by reference to a three-standard (normally 0.5, 1, and 2 ppb plus blank) calibration line (details in Supporting Information, S1). Indium, rhenium, and ruthenium were selected as internal standards to monitor the condition of the VG PQII+ within each session. The accuracy of the procedure was checked by including an international environmental reference material BCR-1 (25), which although not representative of the sample matrix, is enriched in light REE and so is ideal for assessing the stability of the ICP. Precision of sample preparation using three duplicate pH values showed differences smaller than 10%. 2.2. Data Analysis. Metal adsorption was calculated by mass balance from the difference between initial concentration and the concentration in solution after equilibration. The calculated adsorption edges were modeled using the FITEQL 4 code (26) to determine intrinsic metal-site stability constants, using variance between the experimental data and the model to select the best fitting model. These values could then be compared to values predicted using linear free energy relationships. A constant capacitance electric field model with activity correction using the Davies equation was adopted, with the same surface area (140 m2/g), capacitance (8F/m2), deprotonation constants and surface site densities as in Ngwenya et al. (3). These are pKa ) 4.3 and 5.5 × 10-4 mol/g sites for the carboxyl ligand, and pKa ) 6.9 and 2.2 × 10-4 mol/g sites for the phosphate ligand, based on dry weight of bacteria. A constant capacitance model was adopted to account for the high lanthanide valence (27) but also because attempts with nonelectrostatic models did not always produce consistent results between different lanthanide to biomass ratios. Acid-base equilibria for water were included in the equilibrium problem, together with lanthanide hydrolysis stability constants from Klungness and Byrne (24). 2.3. Data Sources for Linear Free Energy Relationships and Implementation. Because we lack literature data with which to compare experimentally determined surface complexation constants, we have used linear free energy relationships in an attempt to predict surface complexation constants for lanthanides. Linear free energy relationships require a suite of different metals for which surface complexation constants to specific sites on the bacterial surface are known. These complexation constants can then be regressed against their stability constants with well-known ligands, preferably those containing functional groups sus-
pected to be present on bacterial cell surfaces. The resulting equations are then applied inversely to predict surface complexation constants for the lanthanides, using published stability constants of the lanthanides with each chosen ligand. This approach has been used successfully to estimate complexation constants for the lanthanides with humic substances (28, 29), which were then used to model speciation of lanthanides in natural fluids (28) and experimental data (29) with reasonable success. Metal adsorption data for regression equations was sourced from several studies, including one study on Pantoea agglomerans (3), and several on Bacillus subtilis (1, 20, 30). Although B. subtilis is Gram-positive, the stability constants determined by these studies for Cu, Pb, and Zn were shown to be statistically similar to those determined for Pantoea agglomerans (3) and other Gram-positive bacteria (6). Moreover, all four studies used the constant capacitance model, leading to data that is internally consistent. Although some of the experimental data is consistent with adsorption to phosphate sites as the best fitting model (section 3.1), we will focus on estimating stability constants for carboxyl complexation only, because relatively few measurements exist for metal complexation to phosphate sites on bacterial cell surfaces for generating meaningful regression equations, and also because the phosphoryl surface complexation constants may not be as well constrained (see below and Supporting Information S2). The data for surface complexation to carboxyl sites are summarized in Supporting Information (Table S2). With respect to ligands, we restricted our analysis to hydroxyl, acetate, citrate, humic, and fulvic acids. Reasons behind such choices are detailed in Supporting Information (S3). They include (i) the fact that the majority of surface active functional groups contain hydroxyl moieties and thus may provide a close structural model for linear free energy correlations (19), particularly for mineral oxides (31), (ii) spectroscopic (32) and acid-base titrations (33) that suggest that carboxyl functional groups form important metal binding sites (1, 3, 34) and which have previously been correlated with acetate stability constants (1, 6, 20), (iii) the possibility of multidentate binding represented by citrate, and (iv) recent studies suggesting that the acid-base chemistry of natural organic matter (NOM) can be modeled using deprotonation constants measured for bacterial surfaces (35) with the pKa of the acidic site being close to the widely accepted carboxyl site pKa of 3.3 for fulvic acid (36). For humic and fulvic acids, the stability constants for the linear free energy correlation plots were taken from the model V database, whereas for the inverse calculation, we compared the performance of the pKMHA and pKMFA values from Tang and Johannesson (28) and Sonke (29). Humic acids were also included in order to probe the efficacy of using natural organic matter as a surrogate for modeling lanthanide adsorption by bacteria, given recent findings that microbial biomass constitutes more than 50% of the organic matter in soils (18).
3. Results and Discussion 3.1. Experimentally Determined Discrete Site Stability Constants. We demonstrate the analysis using data acquired with nominally 0.2 g/L biomass and 4 mg/L lanthanide concentrations (for comparison of model fits at different metal to bacteria ratios, some graphs include 2 mg/L data). For convenience, we show in Figure 1 adsorption edges for six elements, representing two from each of the light (La and Ce), middle (Sm and Gd), and heavy (Yb and Lu) lanthanide groupings. As expected, all the elements show adsorption edges with cationic behavior, increasing in adsorption with increasing pH. Stability constants based on surface complexation models using FITEQL 4 optimization routine (26) are summarized VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
651
FIGURE 1. Adsorption edges for representative lanthanides with starting lanthanide concentrations of 2 ppm or 4 ppm. Curves represent model fits to carboxyl (e.g., La4C) or phosphoryl (e.g., La4P) sites. Most experiments were conducted using approximately 0.2 g/L biomass. The panels represent (a) lanthanum, (b) cerium, (c) samarium, (d) gadolinium, (e) ytterbium, and (f) lutetium. in Table 1. In all cases, the best fitting model was consistent with monodentate adsorption involving proton exchange to either a carboxyl or a phosphate group (section S1 in Supporting Information). Some of the light lanthanides were best fit assuming adsorption to phosphate sites, while carboxyl complexation appeared to fit the heavy lanthanides and some of middle lanthanides. We believe that these outcomes reflect the inability of surface complexation models to unambiguously distinguish which adsorption sites are involved, particularly given the limited pH range over which experiments were conducted, and also because site concentrations are 1 order of magnitude higher than total lanthanide concentration. However, data beyond this pH range does not yield additional information because adsorption remains constant at 100% above pH 6 (22). The ambiguity may also reflect the fact that both carboxyl and phosphate sites are involved in binding, with phosphate dominating at low pH, as demonstrated spectroscopically in our previous study (22). Mean stability constants calculated from the experimental data (Table 1) have been plotted against atomic number in 652
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010
Figure 2. They display a weak, monotonic increase with atomic number, particularly between Sm and Yb. However, there is evidence for the existence of convex up tetrad effects (37) in the first (La-Nd), third (Gd-Ho), and fourth (Er-Lu) tetrads in both carboxyl and phosphate stability constants. Tetrad effects have been reported in other studies involving adsorption of lanthanides to bacteria (12) and inorganic particles, including MnO2 (38) and zeolites (39), but these studies calculated distribution coefficients based on adsorption of a mixture of the lanthanides rather than single element stability constants. The existence of these tetrad effects has produced a trend that globally appears flat. Furthermore, Eu and Lu deviate from the general trend, plotting below the global pattern. Additionally, calculated phosphoryl stability constants covary with carboxyl stability constants, suggesting that they may not be as well constrained. The lack of literature data with which these stability constants can be compared was partly the motivation for this study. Fein et al. (20) calculated a log K of 5.1 ( 0.2 for monodentate Nd adsorption to carboxyl sites on Bacillus subtilis cells, close to our experimentally determined value
TABLE 1. Experimentally Determined (modeled) Stability Constants for Adsorption to Carboxyl (R-COO-Ln2+) and Phosphoryl (R-POO-Ln2+) Sites element
R-COO-Ln2+
R-POO-Ln2+
La Ce Pr Nd Pm Sm Eu Gd Tb Dy Ho Er Tm Yb Lu Y
4.87 ( 0.01 5.00 ( 0.26 4.97 ( 0.08 5.06 ( 0.40 ND 5.13 ( 0.04 4.83 ( 0.13 5.14 ( 0.01 5.37 ( 0.04 5.18 ( 0.28 5.22 ( 0.30 5.29 ( 0.13 ND 5.33 ( 0.23 4.95 ( 0.04 4.90 ( 0.33
8.13 ( 0.01 8.39 ( 0.24 8.20 ( 0.51 8.50 ( 0.05 ND 8.35 ( 0.03 7.97 ( 0.02 8.40 ( 0.08 8.59 ( 0.04 8.43 ( 0.27 8.43 ( 0.27 8.52 ( 0.13 ND 8.50 ( 0.23 8.20 ( 0.05 8.10 ( 0.24
of 5.06 ( 0.4. However, these authors did not report phosphate stability constant. Conversely, Markai et al. (11) calculated a log K value of 7.13 ( 0.4 for Eu-carboxyl complexing on Bacillus subtilis cells from pH-dependent adsorption edges. This value was different from that required to fit adsorption isotherm data at a fixed pH of 5, which was 5.97 ( 0.44, still nearly 1 order of magnitude higher than our calculated value of 4.83 for this element. Clearly, more studies are required to constrain these macroscopic adsorption parameters, although it is also encouraging that our stability constant for Eu binding to phosphate sites (7.97) is close to that of Markai et al. (11), which was 8.14 ( 1.0. Nevertheless, because of the suspected poor constraint of the phosphoryl stability constants, the rest of the paper focuses on carboxyl stability constants and how these compare with linear free energy predictions. 3.2. Linear Free Energy Correlations and Estimated Stability Constants. The correlation equations produced using the different ligands, together with statistical analysis of the regression slopes, are summarized in the Supporting Information S3 (Table S3). Some of these have been published previously (20, 28) so only a brief description is given here
FIGURE 2. Variation of experimentally determined stability constants for adsorption to cell surface sites (log K(R-S-Ln)2+), where S is either carboxyl (R-COO-Ln2+) or phosphoryl (R-POO-Ln2+) sites on Pantoea agglomerans. Error bars are 1× standard deviation for experiments with duplicate biomass and/ or starting lanthanide concentration; otherwise, errors in the FITEQL output are quoted.
TABLE 2. Discrete Site Stability Constants Estimated Using Linear Free Energy Equationsa element hydrolysis acetate citrate La Ce Pr Nd Sm Eu Gd Tb Dy Ho Er Tm Yb Lu
3.91 3.97 4.03 4.06 4.08 4.11 4.06 4.08 4.06 4.06 4.08 4.14 4.14 4.17
4.62 4.83 4.81 4.76 4.97 4.92 4.74 4.74 4.74 4.72 4.72 4.72 4.63 4.74
4.47 4.57 4.57 4.60 4.56 4.53 4.51 4.52 4.53 4.54 4.58 4.62 4.62
TJHA
TJFA
4.43 4.53 4.58 4.59 4.71 4.67 4.62 4.58 4.55 4.55 4.55 4.58 4.60 4.63
4.67 4.80 4.84 4.86 4.99 4.96 4.91 4.88 4.84 4.86 4.87 4.88 4.91 4.96
SONKE- SONKEHA FA 4.63 4.68 4.73 4.78 4.89 4.94 4.99 5.05 5.09 5.14 5.19 5.24 5.30 5.35
4.80 4.86 4.90 4.94 5.03 5.07 5.12 5.16 5.20 5.24 5.29 5.33 5.37 5.42
a Estimates based on humic and fulvic acids are split in two based metal-stability constants from Tang and Johannesson (TJ-HA and TJ-FA, ref 28) and those from Sonke (SONKE-HA and SONKE-FA, ref 29).
to highlight important aspects. Regression with acetate yields the highest correlation coefficient (0.93). This value is slightly lower than that found by Fein et al. (20), which was 0.97, partly because the data used by Fein et al. did not include stability constants for Pantoea agglomerans and partly because in our treatment we excluded Nd3+, as its stability constant also needs to be estimated. Fein et al. (20) suggested adopting acetate to estimate stability constants for adsorption of metals to carboxyl sites, because of the higher correlation coefficient. However, other studies have found that acetate complexes do not yield the highest linearity and have argued against the universal application of acetate for different bacteria (6). Surface complexation constants estimated using the chosen ligands are tabulated in Table 2. They vary widely between ligands, by about 1 order of magnitude for the light lanthanides. The difference increases to 1.3 orders of magnitude for the heavy lanthanides. As shown in Figure 3a, the predicted stability constants produce different trends when plotted against atomic number, each trend reflecting that of the original regressed data. Thus, stability constants estimated from acetate solution complexing are convex with a maximum around the middle lanthanides, mimicking lanthanide-acetate stability constants from Shock and Koretsky (40). By contrast, values estimated from humic and fulvic acid stability constants from Sonke (29) increase with increasing atomic number, also reflecting their recommended values. These differ, however from those of Tang and Johannesson (28) for the same ligands, which like acetate complexes, are convex around the middle lanthanides. Comparison of the measured stability constants for the 13 lanthanides with estimated stability constants indicates clearly that using hydrolysis constants can be ruled out because they are too low, by at least 1 order of magnitude in most cases. Burnett et al. (6) also dismissed hydrolysis constants for predicting stability constants for adsorption of selected transition metals to carboxyl sites, but this was based on a lower regression coefficient rather than predicted values. Similarly, values estimated using acetate, citrate, and humic and fulvic acid values from Tang and Johannesson (28) are somewhat low relative to experimental data from this study, particularly for the light and heavy lanthanides. A better match between measured and estimated values is apparent using pKMFA values from Sonke (29) for fulvic acids. Note, however that significant discrepancies occur for Ce, Pr, Eu, and Lu even for fulvic acid. Two of these (Ce and Eu) can VOL. 44, NO. 2, 2010 / ENVIRONMENTAL SCIENCE & TECHNOLOGY
9
653
FIGURE 3. (a) Comparison of experimentally determined carboxyl stability constants with estimates (curves) based on linear free energy relationships with selected ligands, plotted against atomic number. The ligands are hydroxyl (OH), acetate, citrate, humic acid, and fulvic acid. Estimates based on the latter two are split in in two based on metal-stability constants from Tang and Johannesson (TJ-HA and TJ-FA, ref 28) and those from Sonke (SONKE-HA and SONKE-FA, ref 29). Note the close correspondence between experimental values and those predicted using fulvic acid (SONKE-FA) and humic acid (SONKE-HA) values from Sonke (ref 29). (b) A linear plot of the predicted versus experimental values shows that most data plot close to the 1:1 line (solid line) and within the error envelope (thin lines) of (0.3 log-units of the experimental data. Elements showing large deviations are highlighted. potentially occur in oxidation states other than trivalent, but it is unclear whether this contributes to the deviation. Figure 3b presents in linear form the log Kcarboxyl values predicted using fulvic acid stability constants from Sonke (29) against experimental values. Most predicted values are close to the 1:1 line and fall within 0.3 log-units (i.e., two times error envelope, this being the maximum error bound of the experimental data) except Lu. Larger deviations are also apparent for Eu and Tb, but these are still within the error envelope. These graphical observations are supported, to some extent, by the regression statistics summarized in Table S4 of the Supporting Information (S4), where regression slopes (t-ratio > 2) are significant only for the SONKE-FA predictions and root-mean-square errors are also lower. In particular, the statistics improve significantly when the Lu outlier (SONKE-FA*) is excluded. This results in a regression slope close to unity (0.77), an improved (normalized) R2 value (0.60) and p ) 0.002 and the intercept is closer to zero, with a positive offset most likely due to the higher log Kcarboxyl for the light lanthanides. The model bias for SONKE-FA* was also lower at 0.035 compared with the TJ-FA model with a bias of 0.114. 3.3. Application to Adsorption Modeling. From the preceding discussion, we can reasonably conclude that fulvic acid provides more reasonable estimates of the stability 654
9
ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 44, NO. 2, 2010
FIGURE 4. Comparison of the predicted percentage adsorption for representative lanthanides using experimentally determined carboxyl stability constants for 4 ppm initial lanthanide suspensions (e.g., LA4C) and estimated stability constants for fulvic acid based on metal stability constants from Sonke (SONKE-FA, ref 29) and Tang and Johannesson (TJ-FA, ref 28). Note the poor fit at high pH for La with all three stability constants, due to the fact that La adsorption is best modeled by adsorption to phosphate rather than carboxyl sites.
constants for carboxyl complexation on bacterial surfaces. It also emerges that a high regression coefficient does not necessarily produce the best estimates. This is a counterintuitive result, although it might reflect differences in structural configuration of the metals between the soluble ligand and bacterial surfaces. In general, differences between estimated and calculated stability constants using humic and fulvic acid values from Sonke (29) are relatively small and fall within experimental errors often found for metal-site stability constants in metal-bacteria systems (1). As a result, the estimated values perform reasonably well in predicting the measured adsorption, as demonstrated with selected elements (Figure 4), where the solid curves represent predicted adsorption using average stability constants calculated from the experimental data. Thus for lanthanum, predicted adsorption using estimated stability constants perform as well as the experimentally derived stability constants, all three underestimating adsorption at pH greater than 5. This is consistent with experimental data where optimal fits to the data were consistent with adsorption to phosphate sites (Figure 1a).
Equally, both fulvic acid estimates predict adsorption values identical to the predicted experimental adsorption curve for Sm (optimally fit with carboxyl sites: see Supporting Information), while the prediction is slightly better using fulvic acid constants based on Sonke (29) for Yb. In general, estimates based on pKMFA values from Sonke (29) perform slightly better than those based on values from Tang and Johannesson (28) in predicting lanthanide adsorption to carboxyl sites on bacterial surfaces. The exception is Lu where values recommended by Tang and Johannesson (28) are marginally superior. 3.4. Wider Implications. Stability constants calculated for adsorption of the lanthanides to carboxyl and phosphate sites on bacterial surfaces are high relative to other ligands. Thus, bacteria are likely to exert important controls on the mobility, cycling, and fractionation of lanthanides in the environment (1, 3, 4, 6, 41). What remains to be resolved is to what extent bacteria contribute to the total organic matter content in the subsurface and other environmental settings, and whether such a contribution warrants separate formulation in models of fate and cycling of these metals. In a recent study by Simpson et al. (18), it has been suggested that bacterially derived biomass can constitute more than 50% of the total soil organic matter, even though live bacterial biomass is generally low (
