BACTERIAL ADSORPTION OF AQUEOUS HEAVY METALS: MOLECULAR
SIMULATIONS AND SURFACE COMPLEXATION MODELS
A Dissertation
Submitted to the Graduate School
of the University of Notre Dame
in Partial Fulfillment of the Requirements
for the Degree of
Doctor of Philosophy
by
Kelly J. Johnson
____________________________________
Jeremy B. Fein, Director
Graduate Program in Civil Engineering and Geological Sciences
Notre Dame, Indiana
July 2006
BACTERIAL ADSORPTION OF AQUEOUS HEAVY METALS: MOLECULAR
SIMULATIONS AND SURFACE COMPLEXATION MODELS
Abstract
by
Kelly J. Johnson
Bacterial cell walls can adsorb a wide range of metal cations, potentially altering the
mobility of the metals in geologic systems. To constrain and mitigate contaminant
transport it is essential that geochemical models be developed to measure and quantify
adsorption of heavy metals onto bacteria. This dissertation presents the work of three
studies that apply molecular simulations (Ch. 2) and surface complexation modeling (Ch.
3 & 4) to improve our understanding of metal-bacterial adsorption reactions. Simulation
models (Ch. 2) enabled us to estimate the most stable configuration for bacterial surface
complexes and to compare binding affinities and interatomic distances with experimental
values to validate and predict metal adsorption behavior. We found that mechanics-based
simulations adequately describe the interactions of Cd with the cell wall, defining metal
ion coordinations and binding distances. However, this approach does not accurately
describe Pb-cell wall interactions, possibly due to limitations in the simulation
parameters, the propensity for Pb to form hydroxides at circumneutral pH, or other
adsorption mechanisms. We studied the effect of bacterial metabolism (Ch. 3) on the
Kelly J. Johnson
extent of Cd adsorption to Gram-positive and Gram-negative bacteria. We found that
while metabolically-active Gram-positive cells adsorb significantly less Cd than nonmetabolizing cells, Gram-negative cells show little difference in Cd adsorption. The
metabolic effect on adsorption for Gram-positive cells is likely due to the proton motive
force. The lack of effect in Gram-negative cells suggests that Cd adsorption occurs in a
region of the cell wall not affected by proton motive force. We use a thermodynamic
modeling approach to estimate that the effect of the proton motive force lowers the pH at
the cell wall from 7.0 to 5.7. We applied potentiometric titrations and metal adsorption
experiments (Ch. 4) and found that changes in bacterial diversity do not impact proton
and metal uptake of consortia grown from three locations and sampled throughout a year,
strongly suggesting universal adsorption behavior for the species present. Applying an
averaged-site surface complexation model we found a single set of averaged acidity
constants, site concentrations, and stability constants for metal-bacterial surface
complexes that can be used to model the adsorption behavior.
CONTENTS
FIGURES
………………..…………………...……………………………………...iv
TABLES
…….…………………………………………………………...……….....v
ACKNOWLEDGMENTS
……………………………………………………...….....vi
CHAPTER 1: INTRODUCTION
……………………………………………………..1
CHAPTER 2: MOLECULAR SIMULATIONS OF METAL ADSORPTION TO
BACTERIAL SURFACES …………...……….……………………………..…7
2.1
Introduction ……………………………………………………………..7
2.2
Cell Wall Characteristics
……………………………………………10
2.3
Methods and Model Development ………………….………………...19
2.3.1 Simulation Methods ……………………………………………20
2.3.2 Model Development ……………………………………………25
2.3.3 Metal Interactions
……………………………………………26
2.4
Results and Discussion
……………………………………………28
2.4.1 Ligand Model Development and Structural Optimization……….28
2.4.2 Ligand-M2+ Energy Minimization ……………………………28
2.4.2.1 Binding Energies
……………………………………29
2.4.2.2 Metal-Oxygen Distances
……………………………32
2.4.3 Molecular Dynamics of Periodic Hydrated Systems ………..…..32
2.4.3.1 Cd2+ Simulations
……………………………………35
2+
……………………………………39
2.4.3.2 Pb Simulations
2.5
Conclusions ……………………………………………………………43
CHAPTER 3: THE IMPACT OF METABOLIC STATE ON CD ADSORPTION ONTO
BACTERIAL CELLS
……………………………....…………………....45
3.1 Introduction
…………………………………………………………....45
3.2 Methods
……………………………………………………………49
3.2.1 Bacterial Strains and Culture Conditions
……………………49
3.2.2 Metabolic Treatment and Cadmium Binding ……………………49
3.2.3 Metabolic Activity Measurements ……………………………52
3.3
Results ……………………………………………………………………53
3.4
Discussion
……………………………………………………………59
3.5
Conclusions ……………………………………………………………64
CHAPTER 4: PROTON AND METAL ADSORPTION ONTO BACTERIAL
CONSORTIA: SIMILARITIES IN ADSORPTION BEHAVIORS AND
ESTIMATION OF GENERALIZED STABILITY CONSTANTS FOR METALBACTERIAL SURFACE COMPLEXES ……………………………………66
ii
4.1
4.2
4.3
4.4
Introduction ……………………………………………………………66
Methods
……………………………………………………………69
4.2.1 Sampling and Growth of Bacteria
……………………………69
4.2.2 Potentiometric Titrations and Metal Adsorption Experiments
……………………………………………………………………71
4.2.3 Gram Staining and DGGE Analysis ……………………………72
Results and Discussion
……………………………………………74
4.3.1 Bacterial Diversity ……………………………………………74
4.3.2 Potentiometric Titrations
……………………………………75
4.3.3 Metal Adsorption Experiments
……………………………76
4.3.4 Surface Complexation Modeling
……………………………77
Conclusions ……………………………………………………………95
CHAPTER 5: CONCLUSIONS
REFERENCES
…………………………………………………..100
…………………………………………………………………..103
iii
FIGURES
2.1
Molecular simulation models ……………………….…….………....….21
2.2
Radial distribution functions from molecular dynamics simulations of M2+
interaction with the carboxylate ligand of the peptidoglycan fragment ……......36
2.3
Radial distribution functions from molecular dynamics simulations of M2+
interaction with the phosphoryl ligand of the peptidoglycan fragment ……......41
3.1
Representative oxygen consumption by treated and non-treated bacterial cells
……………………………………………………………..…….............54
3.2
Comparison of the amount of Cd adsorbed onto metabolizing and nonmetabolizing bacterial cells for experiments containing 3, 10, and 20 ppm Cd
……………………………………………………………………….…..56
4.1
DGGE analysis gel
4.2
Potentiometric titration results (a) for all titrations conducted for river, soybean
crop and forest sites (b) Example experimental potentiometric titration curve
showing model fits ……………………………....…………….……………..78
4.3
Cd adsorption onto the bacterial consortia from the river, soybean crop, and forest sites
…………………………………………..…………………………………..80
4.4
Metal adsorption data for Ca, Cu, Pb, Sr, and Zn
4.5
Correlation plots showing metal adsorption constants calculated from natural
consortia and corresponding metal-acetate stability constants from Shock and
Koretsky (1993)
……………………………….…………………………..96
……………………………………………………….…..75
iv
…………………………..81
TABLES
2.1
PARTIAL CHARGES OF METAL AND LIGAND SPECIES USED FOR
MOLECULAR SIMULATIONS…...……………………………………………21
2.2
POTENTIAL ENERGY VALUES FOR CD2+ AND PB2+ FROM THE
MOLECULAR DYNAMICS SIMULATIONS FOR THE PERIODIC
SOLVATED METAL-LIGAND STRUCTURES………………………………23
2.3
BINDING ENERGIES (KCAL/MOL) FOR CD2+ AND PB2+ FOR THE GAS
PHASE SIMULATIONS OF METAL ADSORPTION TO THE
PEPTIDOGLYCAN LIGAND LINKED TO THE TEICHOIC ACID (PEP-TA
……………………………………………………………………………31
2.4
THE COORDINATION AND BINDING DISTANCES (Å) OF CATIONS
WITH 1:1 AND 1:2 METAL-LIGAND STOICHIOMETRIES
……………37
4.1
CALCULATED PROTON BINDING CONSTANTS (PKA) AND SURFACE
SITE CONCENTRATIONS ……………………………………………………89
4.2
CD BINDING CONSTANTS (LOG K) FOR BEST-FIT ADSORPTION
MODELS
……………………………………………………………………91
4.3
METAL BINDING CONSTANTS (LOG K) FOR BEST-FIT ADSORPTION
MODELS
……………………………………………………………………94
v
ACKNOWLEDGMENTS
I must first thank my advisor, Dr. Jeremy B. Fein, for sharing his vast knowledge
and infinite patience with me. I appreciate all the unique opportunities he has generously
provided me and I truly thank him for all he has taught me and all he has done for me.
Thanks for always answering when I knocked and being accessible all day, every day,
whether here or on some other continent. It truly makes a difference. I would also like to
extend a thank you to Dr. Randall T. Cygan for giving me the opportunity to work with
him at Sandia National Laboratories in the wonderful Southwest. Thank you for your
endless patience and understanding. I would like to thank the professors on my
dissertation committee, Drs. Peter C. Burns and Patricia A. Maurice, for their time and
help in my research efforts. My thanks go out to Dr. Louise Criscenti for helpful
discussions regarding molecular simulations development, as well as Jennifer E. S.
Szymanowski for the endless help on projects in the Environmental Molecular Science
Institute and my job search. I would like to extend my thanks to the laboratory assistants
Jennifer Forsythe (EMSI), and also to Dennis Birdsell for his help with instrumentation at
Notre Dame’s Center for Environmental Science and Technology (CEST). I would also
like very much to thank all the members of the Fein group who have been very helpful
throughout my years here at Notre Dame.
This dissertation would not be complete without expressing acknowledgement
and gratitude to my family and friends. I have to thank my parents for their top-notch
vi
advice, long telephone calls, airline tickets, for providing me with the opportunities that
have let me get this far, and for wanting me to be happy. I could not have asked for
better parents. I thank my siblings for being supportive, and a special thanks to Chris for
being both a great brother and outstanding friend. To my friends spread around at
various graduate schools, ski resorts, and international projects thanks for being cool and
inspiring me to do something different, I can’t wait to see you all again. To all my
friends from Notre Dame, I extend my deepest gratitude. Special thanks go to Sara
Nicholl for opening my eyes to not only great coffee and fitness, but for her unwavering
support and understanding. To Brad Weldon, thanks for being a good friend, making me
laugh, and letting me stay with you, and finally to my former officemate Dr. Katie C.
Young (future J.D.) for making me laugh so hard and introducing me to such great things
as Page 2 and Hoops &Yoyo. I would not have finished this without the support,
patience, and encouragement from the special friends I made at Notre Dame. Thank you.
Research funding was provided by a National Science Foundation Environmental
Molecular Science Institute grant (EAR02-21966) and the U. S. Department of Energy,
Office of Basic Energy Sciences. Sandia is a multiprogram laboratory operated by
Sandia Corporation, a Lockheed Martin Company for the United States Department of
Energy’s National Nuclear Security Administration under contract DE-AC0494AL85000. Fellowship funding was provided by EMSI, DOE, and a CEST (Bayer).
vii
CHAPTER 1
INTRODUCTION
Many years of mining, industrial activities, weapons and nuclear energy
production, and other processes have lead to widespread metal contamination in the
environment. In many cases, the mobility of heavy metals in the sub-surface is dependent
on the adsorptive properties of the geologic surfaces they come into contact with.
(Beveridge and Murray, 1976; Sposito, 1984; McCarthy and Zachara, 1989; Davis and
Kent, 1990; Dzombak and Morel, 1990; Fein et al., 1997). Therefore, it is essential to
develop accurate conceptual and quantitative models to assess metal adsorption processes
and contaminant migration.
Bacteria are present in most environments and their surfaces can adsorb a wide
range of aqueous metals (e.g., Beveridge and Murray, 1976; Beveridge and Koval, 1981;
Mullen et al, 1989; Konhauser et al., 1993), thereby impacting their mobility in many
water-rock systems. Metal binds to bacterial surfaces over a large pH range because of
the low isoelectric point of most bacterial surfaces. Bacterial surfaces contain carboxyl,
phosphoryl, hydroxyl, and amino functional groups (Beveridge and Murray, 1976; 1980)
and with increasing pH, these functional groups deprotonate, resulting in negatively
charged functional group sites capable of metal adsorption (Beveridge and Murray; 1980;
1
Beveridge, 1989). The high surface area to volume ratio of bacteria allows them to
accumulate metals in amounts greater than their own weight (Beveridge, 1989).
Various modeling techniques have been applied to quantify bacteria-metal
interactions. Empirically-based bulk partitioning models, such as those that employ
Freundlich and Langmuir isotherms, have been used to develop partitioning coefficients
to quantify the extent of metal adsorption to bacterial cell walls (e.g., Harvey and Leckie,
1985; Mullen et al., 1989; Ledin et al., 1996, Warren and Ferris, 1998). These models can
successfully describe adsorption processes for many conditions of interest, however, the
models do not account for surface and aqueous speciation, and are only accurate if
chemical conditions are relatively constant, i.e. the pH is stable and the surface properties
of the sorbent are relatively homogeneous (Bethke and Brady, 2000; Koretsky, 2000).
Surface complexation modeling, a thermodynamics-based modeling technique,
applies mass action and mass balance equations to describe proton and metal adsorption
to the bacterial surface (Fein et al., 1997). Surface complexation models (SCMs) use an
equilibrium constant to quantify the stability of the adsorbed metal-bacteria complex. The
mass action and mass balance equations are solved to determine the equilibrium constant
for the adsorption reactions and/or concentrations of the individual chemical species. The
stability constants developed from SCMs are independent with respect to most of the
parameters that affect partitioning-approach based models (i.e. pH, bacteria:metal ratio,
and ionic strength). However, SCM models are more complicated because the sorbent
surface must be characterized more precisely by determining the acidity constants and
site concentrations for the functional groups responsible for contaminant adsorption as
well as the stability constants for the important surface complexes. This dissertation
2
includes three related studies that apply modeling techniques to describe the mechanisms
and quantify the adsorption of heavy metals to bacterial cell walls.
In Chapter 2 we apply molecular simulations techniques to model heavy metal
adsorption to the bacterial cell wall. In recent years, the adsorption of heavy metal cations
onto bacterial surfaces has been extensively studied using both laboratory and field
techniques. Metal adsorption has generally been modeled as a bulk partitioning process,
ignoring the specific site interactions and only determining the quantity of metal
adsorbed. Bulk adsorption measurements, involving both protons and aqueous metal
cations, conducted as a function of pH and/or metal-bacteria concentration ratio, can be
used to indirectly constrain important adsorption reactions and to determine the
equilibrium constants for those reactions (e.g., Fein et al., 1997; Cox et al., 1999;
Martinez et al., 2001; Ngwenya et al., 2003; Borrok et al., 2004). Additional direct
constraints on the metal-bacterial cell wall binding environment have been offered by Xray absorption fine-structure spectroscopy (XAFS) investigations (e.g., Sarret et al.,
1998; Kelly et al., 2001; Boyanov et al. 2003a; Templeton et al., 2003; Francis et al.
2004). X-ray absorption spectroscopy can provide excellent constraints on the first and
second nearest neighbors to a metal of interest on the cell wall. However, this approach
only yields an averaged view of the binding environment consisting of numerous ligands
and binding orientations. Molecular simulation methods have the potential to be a
complementary third approach for studying metal-bacteria adsorption reactions,
providing a more detailed and atomistic understanding of how metal cations interact with
specific functional group types within the bacterial cell wall. Chapter 2 describes a study
in which we applied energy minimizations and molecular dynamics based simulations to
3
model Cd and Pb adsorption onto carboxyl and phosphoryl functional groups. This
approach yields information regarding the stability of a variety of metal:bacteria
complexes and allows comparison of binding affinities and interatomic distances. This
work has been submitted for publication to Geochimica et Cosmochimica Acta. Dr.
Randall T. Cygan (Sandia National Laboratories, Albuquerque, NM) and I performed the
mechanics based simulations. Dr. Cygan performed the quantum-based electronic
structure calculations. I wrote the manuscript and Drs. Cygan and Fein provided
intellectual insight and editing.
Chapter 3 includes a study of the impact of metabolic process on Cd adsorption
onto bacterial cells. Bacterial surfaces can adsorb a wide range of aqueous metals (e.g.,
Beveridge and Murray, 1976; Beveridge and Koval, 1981; Mullen et al, 1989). Many
experimental studies of bacteria-metal interactions have employed non-metabolizing
bacteria cells because they were focused on the passive adsorption of metal cations to
bacterial surface functional groups (e.g., Mullen et a.,. 1989; Fein et al., 1997; Haas et al.,
2001; Ngwenya et al., 2003). Metal cations bind to bacteria as a function of pH. The
extent of cation adsorption increases markedly with increasing pH due to the
deprotonation and increasing negative charge of the bacterial surface. Bacterial metabolic
activity can create an electric potential across the plasma membrane called a proton
motive force. During aerobic metabolism, protons are pumped across the plasma
membrane toward the outside of the cell (Ehrlich, 1996). The proton motive force moves
protons back into the cell through plasma membrane ATPases, enabling electrical
potential energy to be captured as chemical potential energy in ATP. When protons are
pumped out of the cytoplasm faster than protons can diffuse back through and move
4
away from the plasma membrane, the cell wall region influenced by the proton
accumulation possesses H+ activities that are elevated relative to that in the bulk solution
and other areas of the cell wall. The lower pH associated with the cell wall may impact
the extent of metal adsorption onto cell wall functional groups (Urrutia Mera et al.,
1992); therefore the effect of metabolism on metal cation desorption must be determined
in order to effectively model bacterial adsorption in metabolically active systems.
Chapter 3 describes a study in which we compared the adsorption of Cd in metabolizing
and non-metabolizing Gram-positive and Gram-negative bacteria. Consequently, these
results can be used to determine the impact of bacterial metabolism on the adsorption of
cations to the bacterial cell wall. This paper has been submitted to Geobiology. I wrote
the manuscript with Dr. David A. Ams, Adrianne N. Wedel (Department of biological
Sciences; Wichita State University), Jennifer E. S. Szymanowski, Dustin L. Weber
(Department of biological Sciences; Wichita State University), and Drs. Mark A
Schneegurt (Department of biological Sciences; Wichita State University), and Jeremy B.
Fein as co-authors. The adsorption experiments and analysis were performed by Jennifer
E. S. Szymanowski, Dr. David A. Ams, and me. The respiration measurements were
performed by Adrianne N. Wedel, Dustin L. Weber, and Dr. Schneegurt. Drs. Fein and
Schneegurt provided project advisement and manuscript assistance.
In Chapter 4 we examine if changing bacterial species diversity affects proton and
metal uptake of a natural consortia. To better constrain contaminant transport and design
more effective remediation strategies in the environment, it is important to develop
models that determine the influence of bacteria on the speciation and distribution of
heavy metals in the sub-surface. Site-specific SCMs, originally developed to quantify
5
cation adsorption to mineral surfaces, have been successfully used to account for proton
and metal adsorption to bacterial surfaces (e.g., Plette et al., 1996; Fein et al., 1997; Haas
et al., 2001; Martinez et al., 2002). However, if each bacterial species displays unique
metal adsorption behavior, determining the acidity constants, surface site concentrations,
and metal stability constants required to model metal adsorption onto bacteria in the
environment would be very difficult. Previous studies have observed similarities in
adsorption behavior of single bacterial species (e.g. Small et al., 1999: Daughney et al.,
1998; Kulczycki et al., 2002; Ngwenya et al., 2003). A single location in a natural system
can contain many bacterial species and bacterial diversity can change from one location
to another. Therefore, if SCMs are to be applied to real systems, it is important to
determine if proton and metal adsorption behavior is species-specific or if commonalities
exist among bacterial species. Chapter 4 describes a study in which we used adsorption
experiments onto bacterial consortia grown from samples collected over the course of a
year in various environments to extensively test if commonalities exist in the adsorption
behavior of protons and metals. The results of this study describe the proton and metal
uptake of natural bacterial consortia with consideration for the changes in species
diversity that occurs in natural environments. This work will be submitted to Chemical
Geology for publication. The sample collection and laboratory experiments were
performed by Jennifer E. S. Szymanowski and Melissa Baranay. I performed the
modeling and wrote the manuscript. Dr. Jeremy B. Fein provided intellectual insight and
manuscript assistance.
6
CHAPTER 2
MOLECULAR SIMULATIONS OF METAL ADSORPTION TO
BACTERIAL SURFACES
2.1
Introduction
Bacterial surfaces can adsorb a wide range of aqueous metals (e.g., Beveridge and
Murray, 1976; Beveridge and Koval, 1981; Mullen et al, 1989), thereby impacting the
mobility of mass in many water-rock systems. In recent years, the adsorption of aqueous
metal cations onto bacterial surfaces has been extensively studied using both laboratory
and field techniques. However, most adsorption reactions have been modeled as bulkpartitioning processes, with the major concern being the amount of metal adsorbed to the
adsorbent and not the specific site of adsorption or the mechanism of adsorption.
Molecular simulation techniques can be used to better constrain the binding mechanisms
involved in bacteria-metal interactions, thereby creating more powerful, flexible, and
quantitative models to examine the effects of adsorption on mass transport.
A number of experimental approaches have been used recently to elucidate the
molecular-scale controls for metal binding onto bacterial cell walls. Bulk adsorption
measurements, involving both protons and aqueous metal cations, conducted as a
function of pH and/or metal-bacteria concentration ratio, can be used to indirectly
constrain the important adsorption reactions and to determine the equilibrium constants
7
for those reactions (e.g., Fein et al., 1997; Cox et al., 1999; Martinez et al., 2001;
Ngwenya et al., 2003). More direct constraints on the metal-bacterial cell wall binding
environment have been offered by X-ray absorption fine-structure spectroscopy (XAFS)
investigations (e.g., Sarret et al., 1998; Kelly et al., 2001; Boyanov et al. 2003a;
Templeton et al., 2003; Francis et al. 2004). X-ray absorption spectroscopy can provide
excellent constraints on the first and second nearest neighbors to a metal of interest on the
cell wall. However, this approach only yields an averaged view of what may be a
complex binding environment consisting of numerous ligands and binding orientations.
Molecular simulation methods have the potential to be a complementary third approach
for studying metal-bacteria adsorption reactions, providing a more detailed and atomistic
understanding of how metal cations interact with specific functional group types within
the bacterial cell wall.
Molecular simulations have previously been applied to model components of the
bacteria cell. However, these studies focused on the lipid bilayer of the cell wall, not the
metal-binding macromolecules within the cell wall (Bandyopadhyay et al., 2001; Shelley
et al., 2001a,b). For example, Shelley et al. (2001b) simulated the self-assembly of the
lipid bilayer starting with a random configuration, providing insight into two different
phospholipid phases. The lipid bilayer research utilized coarse-grained simulation models
to demonstrate the properties of the layer, grouping individual atoms with similar
functionality into one entity. This scaling process enables modeling of relatively large
systems and has reasonably low computational cost. However, because of the
simplifications, the approach describes only bulk processes as opposed to atomic level
interactions.
8
Molecular modeling methods can be used to calculate the total potential energy of
a molecular cluster or of a periodic system through either molecular mechanics or
quantum mechanics. Molecular mechanics evaluates the interactions of individual atoms
or molecules while quantum methods extend the simulation tools to the electron level,
evaluating the electronic structure of the system. Molecular mechanics methods require
analytical expressions to describe the potential energy as a function of atomic geometry
(Cygan, 2001). The energy expressions are typically parameterized by experimental
observation or quantum calculations. Through molecular mechanics methods, such
techniques as energy minimization, conformational analysis, and molecular dynamics can
be applied to a system of interest, for example those involving many hundreds and
thousands of atoms of macromolecules representing a bacterial surface.
Molecular modeling studies of bacterial surfaces (e.g. lipopolysaccharide
structures associated with the cell membranes of Gram-negative bacteria) have been
completed over the past decade (Kastowsky et al., 1992; Wang and Hollingsworth, 1996;
Obst et al., 1997; Kotra et al., 1999; Lins and Straatsma, 2001; Shroll and Straatsma,
2003). These studies have used various levels of atomic abstractions and classical
molecular mechanics to evaluate the structure and dynamics of these complex surfaces.
Previous research includes Shroll and Straatsma (2003) employing classical molecular
simulation techniques to model the adhesion of Pseudomonas aeruginosa to the mineral
goethite, and Obst et al. (1997) examining the impact of Ca2+ on the lipopolysaccharide
structure of Escherichia coli.
The objective of the present study was to use a classical molecular mechanics
approach to identify the binding mechanisms involved in Cd and Pb adsorption onto two
9
cell wall macromolecules that are thought to be the foci of metal binding in a number of
Gram-positive bacterial species. The cations Cd2+ and Pb2+ were chosen because previous
laboratory and XAFS research of these ions characterizes the interaction of these specific
metals with bacteria surfaces and their relevant functional groups (Fein et al., 1997;
Boyanov et al., 2003a; Boyanov et al., 2003b; Templeton et al., 2003; Borrok and Fein,
2005). Also, Cd tends to form more stable complexes at circumneutral pH, while Pb is
more complex due to its 6s2 outer shell electronic configuration. The lone pair electrons
are often stereochemically active and induce a strong deformation of divalent lead
polyhedra (Galy et al., 1975). This allowed us to test the applicability of molecular
simulations to describe increasingly complex metal-ligand interactions.
We used energy minimization methods to derive binding energies of metal-ligand
complexes and we applied molecular dynamics (MD) simulations to analyze equilibrium
structures, coordination, bond distances of metal-ligand complexes, and to derive radial
distribution functions for correlation to XAFS observations. We also used molecular
dynamics to study the solvation of metal-ligand complexes in water molecules and to
compare the resulting structures to gas phase simulations of metal-cell wall complexes.
2.2
Cell Wall Characteristics
Our molecular simulations are focused on the metal-binding cell wall constituents
of Bacillus subtilis (a common Gram-positive soil bacterium) because both the
biochemistry and the surface chemistry have been well characterized (Beveridge and
Murray, 1980). However, titration experiments, XAFS, and attenuated total reflectance
Fourier transform infrared (ATR-FTIR) spectroscopy show that most Gram-positive and
10
Gram-negative cell walls contain similar metal binding functional groups (Beveridge and
Murray, 1980; Fein et al., 1997, Yee and Fein, 2001; Borrok et al., 2004; Jiang et al.,
2004). Therefore, the results of our study are likely to be widely applicable for
understanding metal-binding onto a range of similar bacterial species.
The primary components of the Gram-positive cell wall are peptidoglycan,
teichoic acid, and teichuronic acid (Elwood and Tempest, 1969; Beveridge and Murray,
1980; Beveridge, 1999). All three constituents contain functional groups that, when
deprotonated, can effectively bind metal cations. Peptidoglycan contains carboxyl,
hydroxyl, and amine functional groups, teichoic acid includes phosphoryl groups, and
teichuronic acid is similar to teichoic acid, but contains carboxyl functional groups rather
than the phosphoryl groups of teichoic acid. Gram-negative cell walls include a lesser
amount of peptidoglycan than Gram-positive cells and have a complex outer membrane,
but they do not include teichoic and teichuronic acid constituents (Beveridge, 1999). The
outer membrane of Gram-negative bacteria contains phospholipids, lipoproteins,
lipopolysaccharides, and various proteins. The phospholipids have phosphoryl groups in
the same local coordination environment as the phosphoryl groups in teichoic acid.
The peptidoglycan structure consists of two sugars, N-acetylglucosamine and Nacetylmuramic acid (NAG, NAM), with a side peptide chain attached to the NAM. The
peptide chain includes four amino acid groups with the D-glutamic acid and the mesodiaminopimelic acid (DAP) containing the two carboxyl groups of interest for metal
cation adsorption. Peptidoglycan constitutes up to 50 % of the cell wall by weight
(Beveridge and Murray, 1980; Graham and Beveridge, 1994). Teichoic acids comprise
the other major portion of the Gram-positive cell wall. Teichoic acid is a polymer of
11
glycerol linked by phosphoryl groups, which are the active adsorption sites (Figure 2.1);
d-alanine may also be present in teichoic acid, but is not shown in the figure. There are
generally 20-30 residues present in a chain and teichoic acid can represent up to 70% of
the dry weight of the cell wall (Elwood and Tempest, 1969). Teichoic acid is linked
covalently to the peptidoglycan sugars by a linkage unit containing two sugars and a
phosphoryl group. The phosphoryl group in the linkage unit may also be active in
adsorption of cations (Araki and Ito, 1989).
The cell walls of Gram-positive bacteria can exhibit a negative charge due to the
deprotonation of the carboxyl, phosphoryl, and hydroxyl functional groups (Beveridge
and Murray, 1980). At low pH, the functional groups located on the cell wall are mostly
protonated, and, therefore, little to no metal adsorption occurs. As pH increases, the
surface functional groups deprotonate successively, resulting in the overall negative
charge on the cell wall and an increasing number of sites available for metal adsorption.
Potentiometric titration experiments (e.g., Fein et al., 1997; Cox et al., 1999; Ngwenya et
al., 2003; Fein et al., 2005) have shown these surface sites can be represented by discrete
sites on the cell wall, each of which undergoes deprotonation according to the following
reaction:
R − AH 0 ⇔ R − A − + H +
(1)
where R represents the bacterial cell wall macromolecule to which each functional group
type, A, is attached. The pKa values for the carboxylic and phosphoryl sites are 4.8 and
6.8 respectively, leading them to be deprotonated at circumneutral pH (Fein et al., 2005).
Surface complexation modeling can be used to model bulk metal adsorption
12
Figure 2.1. The molecular simulation models developed from the schematic (a) of a
peptidoglycan chain attached to a teichoic acid dimer. Peptidoglycan and teichoic acid
are the major metal binding constituents of the B. subtilis cell wall at neutral pH. The
optimized molecular model (b) is a representation of (a). For all simulations: grey atoms
are carbon, white are hydrogen, red are oxygen, purple are phosphorous, blue are
nitrogen, green are cadmium, and pink are lead. Four peptidoglycan dimers were
linked/bridged together to form more of a "fabric" representation of the cell wall (c).
These structures were studied in gas phase simulations only due to their prohibitive size.
Periodic simulation cells were used to study the effects of solvation on the interaction of
the metal with the ligand. Initially the metal ion was placed in a periodic cell containing
only water molecules (d). We also studied ligand-water associations (not shown) before
placing both the metal and the ligand in the cell. Both (e) and (f) represent 1:2
metal:ligand stoichiometries. (e) Cd2+ associated with the peptidoglycan fragment and (f)
Pb2+ with the teichoic acid. Both of these simulation cells contain more than 500 water
molecules that have been removed to improve viewing of the metal-ligand complex.
13
Figure 2.1 (continued)
(a)
Teichoic Acid
Linkage Unit
Peptidoglycan
O
CH2OH
O
O
H2CO
R-O
P
HO
OCH2
OCH2
CH2OH
H2CO
P
O
O
O
O
O
OH
NAc
P
O
CH2OH
CH2
O
O
O
AcN
O
O
OH
O
HC
NAc
CH3
CO
NH
L-alanine
HC CH3
CO
NH
D-glutamic acid
HC
(CH2)2COOH
CO
NH
meso-diaminopimelic acid
NH2
HC (CH2)2CHCOOH
CO
NH
D-alanine
HC
CH4
14
COOH
OH
OH
NAc
Figure 2.1 (continued)
(b)
15
Figure 2.1 (continued)
(c)
16
Figure 2.1 (continued)
(d)
(e)
17
Figure 2.1 (continued.)
(f)
18
measurements assuming interaction between the deprotonated functional groups and the
aqueous metal cations.
M m + + R − COO − ⇔ R − COO( M ) ( m −1) +
(2)
M m + + R − PO − ⇔ R − PO( M ) ( m −1)+
(3)
The equlibirum constants derived for reactions in this form can account for the observed
adsorption behavior as a function of pH and bacteria-metal concentration ratio (see Fein,
2000 for a review of these approaches). In this molecular modeling study, we consider
the interactions between aqueous Cd2+ and Pb2+ and deprotonated carboxylate and
phosphoryl functional groups of the bacterial cell wall. We assume the valence electrons
of the deprotonated functional groups to be delocalized between the two oxygen atoms.
2.3
Methods and Model Development
A series of molecular models of metal adsorption to the bacterial surface were
developed using molecular simulation methods. Initial gas phase models of the
macromolecules were created from published structures of peptidoglycan and teichoic
acid (Beveridge and Murray, 1980; Araki and Ito, 1989; Navarre and Schneewind, 1999).
These models were examined in different configurations to determine the optimal energy
minimized structures. Orientations of molecular residues were systematically varied and
then fully relaxed to obtain the global minimized configuration. Next, a cation was placed
proximate to the ligand and the new configuration was minimized by again allowing
relaxation of all atoms. MD-based simulations were then conducted on the optimized gas
phase metal-ligand models. Finally, solvation cells with periodic boundaries were
19
developed to study the effect of full water solvation on metal-ligand interaction using
MD simulations.
2.3.1 Simulation Methods
Initial gas phase molecular simulations were used to graphically develop threedimensional models of the peptidoglycan and teichoic acid molecules (Beveridge and
Murray, 1980). The Constant Valence Force Field (CVFF) was applied to evaluate the
interatomic potentials among the various atoms of the system. Through this force field,
each atom has an assigned partial charge (Table 2.1) and a set of parameterized analytical
functions to describe the potential energy of bonded and non-bonded interactions. All
atomic positions were allowed to freely translate during each simulation; no constraints
were imposed on the models. The CVFF force field was originally parameterized for
applications involving peptide and protein structures by Dauber-Osguthorpe et al. (1988).
The non-bonded parameters needed to describe the metal cations-ligand interactions are
discussed below.
To model the critical intra-molecular interactions of the constituents of the cell
wall, the potential energy of the system must be defined. The summation of the following
energy components provides the total potential energy for the simulation:
ETotal = E Coul + EVDW + E BondStretch + ETorsion + E AngleBend
(4)
The Coulombic and van der Waals energies represent the non-bonded terms, and
the bond stretch, torsion, and angle bend correspond to the bonded interactions. The nonbonded terms control the binding and adsorption of the metal cation to the organic
20
TABLE 2.1
PARTIAL CHARGES OF METAL AND LIGAND SPECIES USED FOR
MOLECULAR SIMULATIONS
Metal
Cd
2.0
Pb
2.0
Water
H
0.41
OW
-0.82
Carboxylate
C
0.14
OL
-0.57
Phosphoryl
P
1.4
OL
-0.85
molecules, whereas the bonded terms generally describe the atomic configuration within
the organic molecules. The ECoul term accounts for the long-range electrostatic
interactions and is represented by:
ECoul = K ∑
i≠ j
qi q j
rij
(5)
The partial charges qi and qj are typically obtained from quantum mechanics calculations,
K is a constant, and rij is the distance between the two atoms of the summation. The van
der Waals energy, EVDW, represents the short-range interactions that prevent the overlap
of atomic electronic clouds. It is represented by a Lennard-Jones function:
EVDW
6
 R 12
 Ro  
o
= ∑ Do   − 2  
 rij  
 rij 
i≠ j
  

(6)
where Do and Ro are empirical parameters derived from the fitting of the potential energy
model to observed structural and physical property data.
21
Values for the Lennard-Jones parameters for Cd and Pb interacting with oxygen
(Eqn. 6) were previously unknown. We therefore chose to derive these potentials using
an appropriate analog such as Ba2+ from the parameters of Åqvist (1990). Through a
comparison of Åqvist Ba2+, Åqvist Sr2+, and Palmer Sr2+ (Åqvist, 1990; Palmer et al.,
1996) Lennard-Jones parameters, we determined that Cd and Pb potentials derived from
the Åqvist Ba2+ value were validated by consistent coordination numbers, solvation
energies, and metal-ligand distances for both metal cations when comparing to
experimentally determined values (Franks, 1973; Baes and Mesmer, 1976; Ohtaki et al.
1993). The Lennard-Jones parameters (Eqn. 6) for Cd are D0 = 0.0470 kcal/mol and R0 =
3.1011 Ǻ; and for Pb are D0 = 0.0470 kcal/mol and R0 = 3.8364 Ǻ. Solvation energies for
Cd and Pb derived from MD simulations using periodic water boxes (cation with 216
water molecules), as seen in Table 2.2, are -373.4 and -325.3 kcal/mol respectively.
These values are within 15% of the experimental solvation energy of Franks (1973),
respectively, -436.9 kcal/mol and -359.0 kcal/mol.
Molecule models were optimized by first completing a series of energy
minimizations (also referred to as geometry optimizations) to test various initial
configurations and to obtain the most stable configuration for the molecules. Our initial
modeling emphasized the simulation of isolated molecular clusters, or gas phase
representations, of the cell wall components. Energy minimizations involve the repeated
sampling of the potential energy surface until the potential energy minimum is obtained
corresponding to a configuration where the forces on all atoms are zero (Cygan, 2001).
Multiple initial structures were tested to ensure the true global energy minimum has been
obtained and avoid any configuration corresponding to a local energy minimum.
22
TABLE 2.2
POTENTIAL ENERGY VALUES FOR Cd2+ AND Pb2+ FROM THE
MOLECULAR DYNAMICS SIMULATIONS FOR THE PERIODIC SOLVATED
METAL-LIGAND STRUCTURES
Total PE
(kcal/mol)
σ
(kcal/mol)
Number
of
Waters
20.5
22.8
22.5
System
Model
Water
Cd-Water
216 H2O
Cd-H2O
Pb-Water
Pb-H2O
-1965.0
-2338.4
-2290.3
Peptidoglycan
Fragment
Carb frag
Cd-1Carb
Cd-2Carb
Cd-Dis
Pb-1Carb
Pb-2Carb
Pb-Dis
-4752.5
-5134.5
-5179.7
-5123.1
-5071.8
-5118.6
-5125.7
Teichoic Acid
Fragment
Phos Frag
Cd-1Phos
Cd-2Phos
Cd-Dis
Pb-1Phos
Pb-2Phos
Pb-Dis
-4709.6
-5086.7
-5096.4
-5084.9
-5037.0
-5050.7
-5052.3
PE/MLPE
(kcal/mol)
σ MLPE
(kcal/mol)
er (%)
216
216
216
-9.1
-373.4
-325.3
MLPE
n/a
n/a
n/a
n/a
n/a
n/a
35.2
38.4
40.3
39.5
41.8
39.4
36.4
508
508
512
508
508
512
512
-131.1
-513.1
-522.0
-501.7
-450.4
-460.8
-467.9
47.2
49.7
51.2
50.5
52.3
50.5
48.2
36.0
9.7
9.8
10.1
11.6
11.0
10.3
35.0
30.2
32.5
39.3
37.8
38.8
35.8
508
508
508
508
508
508
508
-88.2
-465.3
-475.0
-463.5
-415.6
-429.3
-431.0
47.1
43.6
45.2
50.4
49.2
49.9
47.7
53.4
9.4
9.5
10.9
11.8
11.6
11.1
The potential energies (PE) and metal-ligand potential energies (MLPE) are obtained by
subtracting the potential energy of the water (the number of waters multiplied the self interaction
energy of water) from the total potential energy of the simulation cell. σ MLPE denotes the
standard deviation of the calculated MLPE and finally er (%) represents the percent of relative
error. Cd-dis and Pb-dis denote simulations in which the cation was not adsorbed or associated
directly with the ligand. The PE values for the Cd-Water and Pb-Water simulations are equivalent
to the hydration enthalpy for the cation.
23
Minimizations are an important tool for examining energies as well as determining metalligand bond distances and coordination.
MD simulations were also utilized in this work to examine the significance of
thermal processes on the energy-optimized molecular configuration. The MD method is a
deterministic technique that allows the molecular system to evolve in response to a
distribution of atomic motions and velocities dictated by the force field (Cygan, 2001). In
dynamics simulations, Newton's equations of motion are iteratively solved for typically
femtosecond time steps. MD simulations overcome some of the limitations associated
with energy minimization by allowing the kinetic energy of the system to assist atoms in
an improved sampling of the potential energy surface and leading to a thermally
equilibrated configuration. From these dynamics simulations we can better assess
equilibrium structures, coordinations, bond distances of metal-ligand complexes, and
derive radial distribution functions for comparison to XAFS data. We can also examine
the explicit solvation of metal-ligand complexes in water using periodic simulation cells.
To create a periodic cell a peptidoglycan or teichoic acid sub-unit, a cation, and over 500
water molecules are placed in a simulation cell of appropriate size for the density of
interest. Surface effects are eliminated by the three-dimensional periodic boundary
conditions and the minimum image convention; the simulation cell is effectively
surrounded in all directions by translated copies of itself. MD simulations were
performed on a gas phase peptidoglycan monomer linked to a teichoic acid dimer (PepTA) and on solvated periodic cell structures of metal adsorption to either peptidoglycan
or teichoic acid sub-units. The gas phase MD simulations were completed to ensure
equilibrium was reached and to determine the average distances for cations adsorbed to
24
the macromolecule. Solvation boxes containing both the metal and ligand were examined
to obtain adsorption energies, metal coordination number, and ion-water and ion-organic
binding distances.
Additionally, a series of gas phase electronic structure calculations was performed
on a set of peptidoglycan and techoic acid fragments (sub-units), identical to those used
in the MD study of the hydrated periodic systems. The quantum simulations provide a
critical independent check on the validity of the force field parameters, and provide a
molecular orbital basis for describing the metal-organic interactions. Optimized
configurations of the fragments with and without the metal cations were obtained using
the all-electron density functional code Dmol3 (Delley, 1990; 2000). Nonlocal gradientcorrected electron correlation (generalized gradient approximation) with double
numerical plus polarization functionals was implemented (Perdew et al. 1992). A selfconsistent field solution was obtained through iteration of the wave equations and an
energy tolerance of 0.0063 kcal/mol. Geometry optimization of each system was obtained
through a series of steepest descent, conjugate gradient, and Newton Raphson methods
with full atomic relaxation and an energy convergence of 0.013 kcal/mol.
2.3.2 Model Development
The Cerius2 graphical-based molecular simulation software package (Accelrys,
Inc., San Diego) was employed for the development of all molecular models. Energy,
energy optimization, and molecular dynamics calculations were performed with the OFF
energy software available within the modeling package. The CVFF force field was
applied to the simple monomer representations of peptidoglycan and teichoic acid. The
25
potential energy for each model was evaluated with a spline cutoff distance of 8.5 Å for
the non-bonded van der Waals interactions and an Ewald summation for the periodic
cells was used for the Coulombic interactions to ensure proper energy convergence (Tosi,
1964; Allen and Tildesley, 1987). As the result of charged systems in the periodic models
(due to deprotonated functional groups and/or the presence of metal cations), a
background screening correction was used to compensate excess charge and provide a
neutral simulation cell.
Energy minimizations were performed on gas phase models to obtain the energy
optimized configuration for each structure. Once the peptidoglycan and teichoic acid
monomers were developed, they were linked to create dimers, peptidoglycan-teichoic
acid structures (Figure 1b), and a larger peptidoglycan strand (Figure 2.1c). The
optimized potential energies from these various structures were recorded and used to
evaluate the metal-organic interactions based on the stability of the metal-ligand
complexes.
2.3.3 Metal Interactions
After obtaining energy-optimized models of the peptidoglycan and teichoic acid
structures, the carboxylic and phosphoryl functional groups of interest were deprotonated
to represent a circumneutral pH; amino groups were subsequently protonated to reflect
the pH conditions. The structures were further energy optimized and examined to ensure
that a global energy minimum was attained. Once fully optimized, a Cd or Pb cation was
placed at an arbitrary distance from each functional group of interest. The system was
again minimized, resulting in a metal ion coordinated or adsorbed to a deprotonated
26
functional group. By varying the initial metal position, we ensured an optimum final
configuration that was confirmed by comparing the potential energy values. The binding
energies for the metal-cell wall association were then derived by comparison of the
potential energy of the cell wall macromolecule models with those models containing the
macromolecule and its associated cation.
Dynamics simulations were used to evaluate the solvated interactions of metal
with peptidoglycan and teichoic acid abstract models and water. Due to computational
cost, the largest periodic box contained 512 water molecules, requiring a smaller organic
model than the full peptidoglycan or teichoic acid macromolecule structures used for the
gas phase calculations. To create the smaller abstract molecule, the ligands were
terminated beyond the carbon group that followed the functional groups of interest.
MD simulations were performed by placing the organic ligands in a cubic
simulation cell with a volume of approximately 5900 Å3 (during molecular dynamics
simulation the box length of approximately 18.1Å changed by no more than 0.2 Å in any
one dimension) with periodic boundary conditions allowing all atoms to have complete
freedom to translate and cross cell boundaries if necessary (Figures 2.1d-2.1f). NPT
canonical ensemble MD simulations were performed at 1 bar and 300K using NoseHoover (Hoover, 1985) and Parrinello-Rahman (Parrinello and Rahman, 1981) methods
to control temperature and pressure, respectively, of the simulation. The MD time step
was 1 fs. Initially, the simulation cells are not at thermodynamic equilibrium, causing the
temperature of the cell to significantly fluctuate during the first few picoseconds of the
simulations. To avoid these thermal excursions and to obtain an equilibrated molecular
configuration, a 30 ps equilibration run was first conducted, followed by a 50 ps
27
production MD run. We observe this combination of dynamics simulations to be
sufficient in allowing full system equilibration; potential and kinetic energies and
temperature attained steady state values within this period. Dynamics trajectories
representative of the equilibrated system were stored for only the last 30 ps of the total 80
ps simulation time. Radial distribution functions (RDF) can be derived from the atomic
trajectories saved from the MD simulations. The RDF represents the distribution of
distances between coordinating atoms during the simulation and can be compared
directly with similar distributions derived from XAFS experimental data.
2.4
Results and Discussion
2.4.1 Ligand Model Development and Structural Optimization
Energy-optimized models were obtained for a peptidoglycan monomer, teichoic
acid monomer, dimers of both of these structures, a peptidoglycan monomer linked to a
teichoic acid dimer (Pep-TA), and, finally, a larger strand of four linked peptidoglycan
dimers. Each peptidoglycan monomer contains two carboxylate groups and each teichoic
acid monomer contains two phosphoryl groups. During energy minimization of the PepTA, the carboxylate and phosphoryl groups did not interact with one another.
2.4.2. Ligand-M2+ Energy Minimization
Though the various models allow for full atomic and molecular flexibility, the
structures of the peptidoglycan and teichoic acid remain relatively stable with little
configurational change along the molecular chains when a cation is associated with the
primary ligand. Most conformational change occurs in the orientation of the atoms within
28
or near the deprotonated functional group to obtain the most favorable metal-ligand
complex configuration. Because all atoms in the molecules possess a partial charge,
ligand atoms located close to the cation respond by “moving” away (same charge) or
closer to the cation (opposing charges) during the energy minimization. Deprotonation of
the carboxyl and phosphoryl functional groups is pH related, therefore, these atomistic
models allow for a better understanding of the response of the cell wall to both pH
changes and cation interaction. Due to the static nature of the molecular models and the
limitations of the non-reactive CVFF force field, the protonation state of the functional
groups is assigned during model development, and therefore is fixed and does not change
during the simulations.
2.4.2.1 Binding Energies
Energy minimizations were conducted on peptidoglycan and teichoic acid
structures in the presence of Cd and Pb. Initial calculations assumed the models to exist
as isolated gas phase molecules, without incorporation of the effects of solvating water
molecules. Studying the individual energies of the Pep-TA structure with and without the
metal cations, we were able to compare the binding energies of the individual functional
groups (Table 2.3). We use the term binding energy to represent the association energy of
the metal complex reactions as described by Eqns. (2) and (3), where the negative sign
indicates the stable formation of the complex. The derived binding energies should only
be compared in a relative sense because of the limitations associated with any empirical
force field like CVFF, and the introduction of specialized Lennard-Jones parameters for
the metal cations. When compared to experimental values, the theoretical energies are
29
typically an order of magnitude greater. These greater theoretical values are not
surprising owing to the lack of any solvating water molecules coordinating to the metalligand association. Note that binding energies derived from electronic structure
calculations of the small-sized proxy metal-organic complexes are similar to those
derived using the classical force field approach and those above for the larger cell wall
models. The binding energies can be used to determine which cation is preferentially
bound to the ligand of interest. The complexation of metal with two ligands provides a
more negative binding energy for both Cd and Pb than complexing with one ligand,
suggesting a 1:2 metal:ligand coordination is more stable as expected. Cd interaction with
the two phosphoryl groups of the teichoic acid displays the most negative binding energy.
For 1:1 metal-organic pairings, the glutamic acid carboxylate group exhibits the most
negative binding energy for Cd, while the Pb binds more tightly to the phosphoryl group.
The Cd modeling result is concurrent with the findings of Beveridge and Murray (1980)
that the glutamic acid site on the peptidoglycan is the most apparent site for metal
complexation on B. subtilis.
Cd displays a more negative binding energy than Pb for all adsorption sites,
suggesting Cd is bound more tightly to the ligand than Pb. These results are inconsistent
with trends observed in bulk adsorption laboratory studies, in which bacterial cell walls
are observed to adsorb significantly more Pb than Cd under identical experimental
conditions (e.g., Fein et al., 1997). The difference between our simulation models and the
results of Fein et al. (1997) are likely due to hydration effects or the presence of covalent
bonding between the metal cation and the organic ligand.
30
TABLE 2.3
BINDING ENERGIES (KCAL/MOL) FOR CD2+ AND PB2+ FOR THE GAS PHASE
SIMULATIONS OF METAL ADSORPTION TO THE PEPTIDOGLYCAN LIGAND
LINKED TO THE TEICHOIC ACID (PEP-TA).
Ligand
Carb-D
Carb-G
Carb-DG
Phos-L
1Phos
2Phos
Cd2+
-381.3
-453.4
-482.1
-422.1
-430.8
-493.4
Pb2+
-329.3
-320.3
-398.8
-359.2
-346.9
-463.2
The energies are obtained by subtracting the potential energy of the energy-minimized Pep-Ta
structure and the free metal cation (0 kcal/mol) from the total potential energy when the ligand is
associated with the cation. Carb-D, Carb-G, and Carb-DG denote which carboxylate ligand
(meso-diaminopimelic acid and/or D-glutamic acid) the metal is associated with. Phos-L indicates
the phosphoryl group that links the peptidoglycan and teichoic acid.
Results for electronic structure optimizations of the fragment representations of
peptidoglycan and techoic acid support those obtained from the classical simulations.
Force field-based simulations of the identical fragment systems were used to compare the
two different theoretical methods. The quantum simulations provide binding energies for
the Cd-ligand and Pb-ligand complexes that are 25 to 80 kcal/mol stronger than the
values obtained by the force field method; the mean relative difference between the two
methods is approximately 14%. As observed for the large cell wall models, Cd binds
more strongly than Pb with the peptidoglycan complexes with carboxylate ligands more
favored than those involving the phosphoryl groups of techoic acid. These results are
consistent with either quantum or classical method. Comparison of the optimized metal
complex structures is quite good with metal-ligand distances in agreement by less than
5% difference. All optimized structures exhibit the metal ions coordinated by four
31
oxygen ligands. This coordination is most enhanced for the Cd-peptidoglycan complex
where the two carboxylate groups form a more tightly bound and relatively planar
coordination about the smaller Cd ion. Conformations of the organic backbone for the
optimized structures derived using the two methods are in very good agreement with only
subtle differences observed.
2.4.2.2 Metal-Oxygen Distances
The metal cation-oxygen distances from the non-solvated gas phase MD models
are dependent on the type of metal, the functional group, and the number of sites
involved in the metal binding. In gas phase simulations Cd exhibits a shorter carboxylate
binding distance than Pb, 2.19 Å versus 2.46 Å, respectively, correlating reasonably with
the 2.3 Å and 2.5 Å respectively of Franks (1973). The binding distances for the fully
optimized structures are less than the observed values for both metals likely due to the
gas phase simulations not addressing the effect solvation has on experimental systems. In
addition, the metal-oxygen distances for metals complexing with two ligands are less
than those for single ligand complexes due to increased electrostatic attractions.
2.4.3 Molecular Dynamics of Periodic Hydrated Systems
Five types of periodic systems were examined using MD to determine the binding
energies of Cd and Pb to the peptidoglycan and teichoic acid fragments in hydrated
periodic systems: metal only, ligand only, metal bound to two functional groups, metal
bound to one functional group, and a cell containing the metal dissociated from the
ligand. The metal and ligand only simulations were developed in order to differentiate
32
their energetics from those simulations containing metal cations. Additionally, the matrix
of simulations provides an opportunity to reduce the binding reaction to the fundamental
components and energies. Each type of ligand-metal MD simulation was performed for
Cd and Pb. The potential energy from the molecular simulations takes into account all
atoms in the solvation box: the organic ligand, the metal ion, and water molecules. The
energies (PE) reported in Table 2.2 for the fragment simulations have significant relative
error, and caution must be taken when trying to compare them to thermodynamic
enthalpies.
In Table 2.2, metal-ligand potential energies (MLPE) were defined by subtracting
the energy of the water molecules (the number of waters multiplied by the self interaction
energy of water) from the total potential energy of the solvation box. The systems in
which the metal was bound to two functional groups (either phosphoryl or carboxyl)
resulted in lower potential energies for both Cd and Pb. For example, Pb interacting with
two ligands (-460.8 kcal/mol) has a PE 10.4 kcal/mol lower than when it is interacting
with one ligand (-450.4 kcal/mol). This energy difference is less than the standard
deviation in the total PE for the MD simulations; nonetheless, this trend can be seen for
both metals with both ligands.
The energy difference between the 1:1 and 1:2 metal-ligand stoichiometries
reflects the greater stability achieved when the metal is coordinated to both functional
groups. Boyanov et al. (2003a), using EXAFS analysis, were unable to determine if the
metal:ligand stoichiometry is 1:1 or 1:2 for Cd-cell wall interactions due to overlapping
error bars in their analysis, and therefore they based their structural models on a 1:1
stoichiometry. Fein et al. (1997) obtained better fits for their bulk Cd and Pb adsorption
33
data using 1:1 metal-ligand stoichiometry and Boyanov et al. (2003b) observed a 1:1 Pbcarboxylate stoichiometry for their study of Pb adsorbed to a monolayer.
Similar to the gas phase simulations, the stabilization energies suggest that both
peptidoglycan and teichoic acid components of the cell wall have a greater binding
strength for Cd cations than for Pb. However, bulk adsorption studies have documented
that the cell wall has a greater affinity for Pb than for Cd in both individual and
competitive adsorption experiments (Fein et al., 1997; Fowle and Fein, 1999; Borrok and
Fein, 2005). For example, Borrok and Fein (2005) conducted separate adsorption
experiments in which 10 ppm of either Pb or Cd were reacted with 3 g/L Pseudomonas
mendocina, a Gram-negative bacterium. At pH 6.5, only half of the Cd was adsorbed
onto the cell wall, while nearly all of the Pb was adsorbed under identical experimental
conditions. The simulations presented here, therefore, must not fully describe the aspect
of the binding mechanisms that account for the differences between Pb and Cd
adsorption. The models also portray Pb binding the most strongly in systems where the
cation is completely dissociated from the critical ligands, although these comparisons are
associated with large uncertainty overlap. Our models account for the strength of metal
adsorption through van der Waals and long-range electrostatic forces, but not covalent
effects or possible metal hydroxide complexes that may influence the affinity and
amounts of metal binding in aqueous systems.
XAFS techniques have been used in various ways to investigate the adsorption of
cations to the bacterial surface. Here, we compare XAFS results to the results of MD
simulations of Cd and Pb adsorption onto peptidoglycan and teichoic acid components of
the bacterial cell to validate the molecular simulation models. In general, the radial
34
distribution functions (RDFs) for cation-oxygen, cation-carbon and cation-phosphorous
are similar for both Pb and Cd, with the overall peak shape and distribution being
comparable for both metal cations (data not shown). As anticipated, the RDFs exhibit
differences in mean distances and overall shape (distribution) of the curve. RDFs were
calculated using the force field type for each of the atoms of interest, which allows us to
discriminate the metal-oxygen distance for the metal-ligand complexes from that of the
metal-water.
2.4.3.1 Cd2+ Simulations
The first shell interactions of Cd-carboxylate complexation with both the oxygen
of the peptidoglycan ligand and water oxygen in the solvated periodic systems can be
compared in the RDF presented in Figure 2.2. The highest peak shows the Cd-O distance
for a 1:1 Cd-carboxylate complex is calculated to be 2.27Å. When Cd is coordinated with
two carboxylate groups the first shell is slightly expanded and the first peak maximum is
at 2.33 Å. These metal cation-OT (all oxygens; ligand and water) distances are both
comparable to the 2.3 Å Cd-oxygen XAFS distances measured by Boyanov et al (2003a).
Analogous calculated Cd-oxygen distances for the 1:1 and 1:2 metal-ligand coordinations
onto teichoic acid are 2.27 Å and 2.31 Å, respectively (Table 2.4). The XAFS
measurements of Boyanov et al, (2003a) placed the first shell for Cd interacting with
solution and the phosphoryl group oxygen of teichoic acid at 2.27 Å. Therefore, the
models for the interaction of Cd with the metal-binding macromolecules of the cell wall
are consistent with the XAFS results.
35
25
Cd-O 2.33 Ǻ
Pb-O 2.57 Ǻ
20
Pb - 1 Carb
Pb - 2 Carb
Cd -1 Carb
Cd -2 Carb
RDF
15
10
5
0
1.5
2.0
2.5
3.0
3.5
r (Å)
Figure 2.2. Radial distribution functions from molecular dynamics simulations of
M2+ interaction with the carboxylate ligand of the peptidoglycan fragment. The fine
lines denote the RDFs for the 1:1 metal-ligand coordination and the thick lines are
for the 2:1 metal-ligand coordination. The arrows indicate the simulated average
metal-OT distance for 1:2 coordination.
36
TABLE 2.4
THE COORDINATION AND BINDING DISTANCES (Å) OF CATIONS WITH 1:1
AND 1:2 METAL-LIGAND STOICHIOMETRIES
Peptidoglycan
Run
Shell
Cd-2Carb
OL
OW
C
OT
Cd-1Carb
OL
OW
C
OT
Pb-2Carb
OL
OW
C
OT
Pb-1Carb
OL
OW
C
OT
Avg CN
R
σ
4
4
2
8
2.33
2.33
2.67
2.33
0.102
0.087
0.074
0.096
2.2
5
1
7
2.27
2.29
2.63
2.27
0.086
0.076
0.061
0.080
4
5
2
8.8
2.51
2.59
2.93
2.59
0.195
0.108
0.113
0.153
2
6.6
1
8
2.55
2.61
2.97
2.61
0.121
0.104
0.091
0.111
Techoic Acid
Run
Shell
Cd-2Phos
OL
OW
P
OT
Cd-1Phos
OL
OW
P
OT
Pb-2Phos
OL
OW
P
OT
Pb-1Phos
OL
OW
P
OT
Avg CN
R
σ
3
4
2
7
2.13
2.31
2.95
2.31
0.339
0.080
0.317
0.115
2
5
1
7
2.19
2.33
2.99
2.27
0.093
0.075
0.056
0.086
3
4.8
2
7.8
2.45
2.59
3.19
2.57
0.117
0.101
0.338
0.125
1
7
1
8
2.47
2.59
3.79
2.59
0.065
0.126
0.154
0.140
The coordination and binding distances (Å) of cations with 1:1 and 1:2 metal-ligand
stoichiometries derived from equilibrated NPT-ensemble molecular dynamics simulations of the
hydrated peptidoglycan and techoic acid fragments. OL = carboxylate or phosphoryl oxygen, OW
= water oxygen, OT = total oxygens, C = carboxylate carbon, and P = phosphoryl phosphorous.
Analysis of the molecular simulation results help to differentiate between the
ligand oxygen (OL) and water oxygen (OW) coordinated with the cation of interest in
contrast to XAFS techniques where no chemical distinction can be made. This is helpful
when attempting to differentiate among the various ligands coordinating to a metal either
in solution or on a surface. Table 2.4 shows the average metal cation-oxygen distances
for the different oxygen types and, similar to the RDFs, these results represent the
average of the first coordination shell. While Cd is coordinated with one and two
37
carboxylate ligands, the binding distances to the specific ligand oxygen are 2.27 Å and
2.33 Å respectively. These are very close to the average metal cation-OT distances, where
OT represents both ligand and water (total) oxygens. However, when binding with the
phosphoryl groups of the teichoic acid, the 1:1 and 1:2 Cd-O distances are 2.19 Å and
2.13 Å respectively. These distances are both smaller than the average metal cation-OT
distances for the Cd-phosphoryl group(s) interaction, which are 2.27 Å and 2.31 Å
respectively. Both the carboxylate and phosphoryl ligand sites have similar
configurations, where an electron is delocalized between the two oxygen of the functional
group. The Cd may be bound closer to the phosphoryl oxygen relative to the OT due to
the higher partial charge of the phosphoryl group oxygen (Table 2.1).
The Cd-peptidoglycan second coordination shell contains carbon at a distance of
2.63 Å from the metal for 1:1 stoichiometry and 2.67 Å for the 1:2 complex. For Cd
complexation onto phosphoryl sites on teichoic acid, phosphorous is the second nearest
neighbor at 2.99 Å and 2.95 Å (Table 2.4), respectively, for complexation with 1 and 2
phosphoryl groups. Boyanov et al. (2003a) fit their carboxylate data with a carbon shell at
2.7 Å and their teichoic acid with a phosphorus shell at 3.43 Å. Due to overlapping error
bars XAFS could not be used to determine the Cd-carboxylate stoichiometry, as noted
previously. The C-shell distance matches the XAFS results (within 2%), however, there
is a sizeable shortening (approximately 13%) for the simulated Cd-P-shell distance with
the one determined by XAFS.
Table 2.4 includes the coordination of the different atoms surrounding the Cd ion.
To determine an average coordination value, the number of atoms surrounding the cation
was counted and averaged every ten ps of the trajectory. X-ray scattering studies of
38
solvated Cd have identified an octahedral hydration shell around the aqueous Cd2+ ion
(Ohtaki et al., 1974; Caminiti et al., 1984; Marcus, 1988; Ohtaki and Radnai, 1993). In
the molecular simulations Cd bound to two ligands was solvated by four water molecules
and for single ligand coordination, the hydration sphere contained five water molecules.
This inner sphere complex was seen by Boyanov et al. (2003a) in their XAFS models for
Cd-B. subtilis experiments and also in X-ray scattering works on similar reference
solutions (Caminiti and Johansson, 1981; Caminiti, 1982; Caminiti et al., 1984).
Although XAFS methods cannot be used to differentiate between ligand and
water oxygen, the agreement of the molecular simulations and XAFS results for the
coordination numbers and distances of Cd with the cell wall sites is quite good. The
correlation between these simulations, XAFS experiments, and laboratory experiments
provide validation, at least to some extent, that the molecular simulations offer a
reasonably accurate view of the adsorption of Cd and similarly behaving divalent cations
onto the reactive cell wall components of a wide range of Gram-positive bacteria. These
results, along with the previous electronic structure validation, suggest that the force field
and the Cd Lennard-Jones parameters derived from the Åqvist (1990) data set are
sufficient for modeling these systems.
2.4.3.2 Pb2+ Simulations
In the case of the Pb2+ ion, there are no XAFS data for direct comparison of Pb
binding to the cell wall of Gram-positive bacteria. However, there are other studies of Pb
adsorption to Gram-negative bacteria, fungal cells, and Langmuir monolayers, all
containing carboxylate and phosphoryl functional groups (Sarret et al., 1998; Boyanov et
39
al., 2003b; Templeton et al., 2003). Sarret et al. (1998) examined Pb binding to fungal
cell walls, comparing carboxylate and phosphoryl complexes. Boyanov et al. (2003b)
studied Pb adsorption to a fatty acid Langmuir monolayer that contained carboxylate
head groups, and Templeton et al. (2003) applied XAFS to study adsorption and
biomineralization within biofilms of the Gram-negative bacteria Burkholderia cepacia.
Although these various substrates lack the full-scale peptidoglycan and teichoic acid
macromolecules, the binding of Pb to the carboxylate and phosphoryl functional groups
can be compared to the molecular simulations of this study.
Our determinations for the average distances for Pb-oxygen in the first
coordination shell are inconsistent with the experimental XAFS results. The calculated
molecular simulation first shell Pb-OT distances for 1:1 and 1:2 metal-ligand
stoichiometries for carboxylate and phosphoryl group interactions are between 2.57 Å
and 2.61 Å (Figures 2.2 and 2.3). Although 2.59 Å is an average Pb-O distance for a
hydrated Pb2+ cation (Franks, 1973), Templeton et al. (2003) found two distinct Pb-O
distances of 2.30 ± 0.02 Å and 2.51 ± 0.02 Å with these distances being similar to Pb-O
distances in lead organic model compounds determined by XAFS spectroscopy (Xia et
al., 1997; Boyanov et al., 2003b). The shorter Pb-O distance and second neighbor Pb-(C,
P) distance that Templeton et al. (2003) measured suggests the ligand forms an innersphere complex with the cation, while the longer Pb-O distance measured by the same
group is consistent with outer-sphere aqueous Pb2+ complexes. However, in comparing
our modeling results to spectroscopic data, it should be noted that obtaining correct firstshell interactions (Pb-O) is much more important than if second shell interactions (Pb-P,
Pb-C) are similar, especially with the complex adsorption behavior of Pb.
40
25
Cd-O 2.31 Ǻ
Pb-O 2.57 Ǻ
20
Pb - 1 Phos
Pb - 2 Phos
Cd - 1 Phos
Cd - 2 Phos
RDF
15
10
5
0
1.5
2.0
2.5
3.0
3.5
r (Å)
Figure 2.3. Radial distribution functions from molecular dynamics simulations of M2+
interaction with the phosphoryl ligand of the teichoic acid fragment. The fine lines
denote the RDFs for the 1:1 metal-ligand coordination and the thick lines are for the
2:1 metal-ligand coordination. The arrows indicate the simulated average metal-OT
distance for 1:2 coordination.
The calculated second shell interatomic distances for 1:1 and 1:2 Pb-C
coordinations are 2.97 Å and 2.93 Å, respectively, for the peptidoglycan macromolecule.
The calculated Pb-P distances for Pb bound to one and two phosphoryl groups of teichoic
acid are 3.79 Å and 3.19 Å, respectively. The calculated Pb-C distances are the same as
Boyanov et al. (2003b) found when looking at the interaction of Pb with the carboxylate
groups on a Langmuir monolayer, and the Pb-P distance is similar to the 3.24 ± 0.04 Å
Pb-P distance Templeton et al. (2003) measured for B. cepacia.
41
When coordinated with the phosphoryl group(s) of teichoic acid, the Pb is
preferentially bound in a monodentate structure with oxygen, and when bound to both
ligands the Pb interacted with three of the four oxygen ligands. This is a different result
than observed by simulation for the Cd interaction when it was coordinated with one
phosphoryl functional group where both ligand oxygens coordinated with the metal in a
bidentate structure. Both Boyanov et al. (2003b) and Templeton et al. (2003) observed
1:1 stoichiometry for Pb-carboxylate and Pb-phosphoryl binding.
The simulations of Pb coordinated to one or two functional groups show the
cation first coordination shell containing eight total oxygen atoms, with five to seven
from the coordinating water molecules. The gas phase electronic structure simulations
performed on the metal-ligand fragments support these force field results. The
coordination of Pb with both oxygen and the respective C or P from the carboxylate or
phosphoryl functional groups is difficult to discern using XAFS techniques due to the
presence of the lone pair of electrons associated with the Pb2+ cation; this situation
creates large disorder in the local coordination environment, particularly with organic
complexes (Sarret et al., 1998). Both Boyanov et al. (2003b) and Templeton et al. (2003)
observed Pb coordinated by four oxygen atoms. Our electronic structure calculations on
the fragment models indicate substantially more transfer of electrons from the Pb
(including the 6s2 lone pair) to the molecular bonding orbitals than observed for Cd;
however, steric and conformational effects associated with the organic backbone while
coordinating to the smaller ion contributed to a more stable Cd complex.
The molecular simulations for the adsorption of Pb onto the carboxylate and
phosphoryl ligands of the peptidoglycan and teichoic acid molecules are consistent with
42
some aspects of XAFS and surface complex modeling findings. Our calculated Pboxygen distances are in good agreement with experimental Pb-oxygen distances for a
solvated Pb2+ cation, and the cation-P or cation-C distances are consistent. However, our
simulations do not support the shorter inner-sphere Pb-oxygen bond distance measured
by Templeton et al. (2003), or the Pb-O coordination observed by both Boyanov et al.
(2003b) and Templeton et al. (2003). These discrepancies are perhaps the result of
limitations in the force field approach or the existence of a different mechanism of
adsorption for Pb to the cell wall. The classical-based models do not account for the
transfer of electrons and the formation of covalent bonds associated with the Pb cation, or
the ability of Pb to form hydroxide phases at circumneutral pH.
2.5
Conclusions
The results of the molecular simulations of this study can be used as a
complement to surface complexation modeling and X-ray absorption spectroscopy for
providing constraints on the nature of the binding mechanisms involved in cation
adsorption onto bacterial surfaces. Using energy minimization and molecular dynamics
simulations in this study, we modeled Cd2+ and Pb2+ adsorption onto the carboxylate and
phosphoryl groups of peptidoglycan and teichoic acid that are present within the cell wall
macromolecules of Gram-positive bacteria. The force field-based models enable us to
estimate the most stable complex configuration and compare binding affinities and
interatomic distances with experimentally-determined values to validate and predict
metal cation adsorption behavior. MD simulations were incorporated to extend the
molecular configurations derived from energy minimizations, and to model the influence
43
of explicit water solvation of the organic components in the presence of solvated Cd and
Pb cations.
The results of our molecular simulations of Cd-cell wall interactions indicate that
molecular mechanics simulation techniques can adequately describe the interaction of Cd
with the cell wall when comparing simulations with XAFS techniques and laboratory
experiments. The molecular dynamics periodic cell simulations described both atom
coordinations and binding distances that correlate very well with spectroscopic data.
While simulations of Pb-ligand interactions do not agree with XAFS results as well as
those obtained for the Cd models, their inconsistency can be construed as a need to refine
force field parameters or to develop an alternative mechanism for Pb adsorption onto the
cell wall.
The application of force field-based simulation methods allows us to examine
relatively large and complex molecular systems such as the linked peptidoglycan dimers
shown in Figure 1c. These theoretical approaches are useful for studying the adsorption
of a cation among multiple ligand sites, the rigidity of the major cell wall constituents,
and adsorption strength, binding distance, and coordination number of various metal
cations without the computational cost and limited system size required to use electronic
structure methods.
44
CHAPTER 3
THE IMPACT OF METABOLIC STATE ON CD ADSORPTION ONTO
BACTERIAL CELLS
3.1
Introduction
Bacterial cell walls can adsorb a wide range of aqueous metal cations, potentially
altering the mobility of the metals in geologic systems (e.g., Beveridge & Murray, 1976,
1980; Beveridge & Koval, 1981; Crist et al., 1981; Harvey & Leckie, 1985; Goncalves et
al., 1987). Most experimental studies of bacterial surface adsorption have involved
bacterial cells that were not metabolically active during the period of adsorption, focusing
on the passive binding that occurs between bacterial surface functional groups and
aqueous metal cations (e.g., Mullen et al,. 1989; Fein et al., 1997; Haas et al., 2001;
Ngwenya et al., 2003). Metal cations appear to bind predominantly to deprotonated sites
within the bacterial cell wall. The extent of metal adsorption onto bacterial surface
functional groups decreases markedly with decreasing pH due to protonation reactions,
and, hence, neutralization of negatively charged surface functional groups at lower pHs.
The decrease in adsorption can be viewed as competitive adsorption of H+ and aqueous
metal cations on available surface sites.
Bacterial metabolic activity can create an electric potential across the plasma
membrane, called a proton motive force. During aerobic metabolism, protons are pumped
45
across the plasma membrane toward the outside of the cell into the periplasmic space,
and electrons or negatively charged species such as OH- concentrate inside the cell
(Ehrlich, 1996). The proton motive force is an essential component of bacterial
metabolism, for the movement of protons back into the cell, down the concentration
gradient through plasma membrane ATPases, enables electrical potential energy to be
captured as chemical potential energy in ATP. If protons are pumped out of the
cytoplasm faster than protons can diffuse back through and move away from the plasma
membrane, then the cell wall region influenced by the proton accumulation by
metabolizing bacterial cells should possess H+ activities that are elevated relative to that
in the bulk solution and other areas of the cell wall. It has been postulated that the lower
pH associated with the cell wall environment of metabolizing cells could diminish the
extent of metal cation adsorption onto cell wall functional groups (Urrutia Mera et al.,
1992). Clearly, in order to model bacterial adsorption in systems where a significant
fraction of the bacteria are undergoing active metabolism, the effect of metabolism on
metal cation adsorption must be determined.
Urrutia Mera et al. (1992) conducted a set of experiments aimed at determining
the effect of the proton motive force on the adsorption of metal cations onto cell wall
functional groups. In these experiments, Gram-positive Bacillus subtilis cells were
inactivated by exposing them to either 1 mM NaN3 or 40 µM carbonyl cyanide mchlorophenylhydrazone (CCCP) in solution, or by exposing them to approximately
35,000 rads/h of gamma radiation for 1.5 h. In control experiments, untreated
metabolically active B. subtilis cells were suspended in distilled water. The final pH for
all control and treated experiments was between 6.40 and 6.53. Each bacterial sample
46
was suspended in a 1 mM metal-bearing solution (either uranyl acetate or scandium
chloride), allowed to react for 10 minutes, washed, and then the amount of metal
associated with the cell walls was determined by bulk elemental ICP-OES analysis. In
general, Urrutia Mera et al. (1992) observed elevated uranyl and scandium concentrations
on the inactivated cells, and they concluded that the proton motive force caused the
diminished adsorption observed with control cells due to enhanced H+ competition.
These results are consistent with similar experiments by Kemper et al. (1993), who also
observed enhanced cation adsorption by inactivated B. subtilis cell walls.
Although the work of Urrutia Mera et al. (1992) suggests that the proton motive
force exerts a significant effect on adsorption, a number of their experimental procedures
may have affected the results. The metal concentrations used by Urrutia Mera et al.
(1992) were relatively high and mineral precipitation may account for some of the
observed metal loss from solution. For example, the solubility of schoepite under the
experimental conditions of a solution open to the atmosphere at pH 6.4 is less than 1 µM.
Even considering aqueous uranyl-acetate complexation, the 1 mM experimental solutions
were significantly oversaturated with respect to schoepite, and precipitation may have
occurred in some of the experiments. Additionally, the ‘active’ bacterial controls
consisted of cell suspensions in distilled water and the level of metabolic activity may
have been quite low.
Claessens et al. (2006) studied the effects of bacterial metabolism on cell wall site
protonation and surface charge of a Gram-negative (Shewanella putrefaciens) species,
and they examined the role of cell wall structure by comparing the protonation behavior
of metabolically-active S. putrefaciens to that of metabolically-active B. subtilis. Live S.
47
putrefaciens cells exhibited rapid initial consumption of acid under all pH conditions
studied, likely reflecting the initial protonation behavior of cell wall functional groups. At
pH 4, proton uptake by suspensions of live cells stopped after 50 min, likely due to loss
of viability. However, at pH 8 and 10, Claessens et al. (2006) observed that
deprotonation continued at a slower rate for the entire 5-h duration of the experiment,
likely due to active respiration and proton motive force generation by the cells.
Inactivation of S. putrefaciens cells caused no effect on initial acid or base consumption
by the cells, however the inactivation caused long-term protonation/deprotonation to
cease or proceed at a very slow rate. Active B. subtilis cells exhibited a greater extent of
initial proton consumption than did active S. putrefaciens cells, indicating that the B.
subtilis cells have more functional groups present on the cell wall. In contrast, the S.
putrefaciens cells exhibited much higher extents of long-term deprotonation at pH 8 and
10 than did the B. subtilis cells, suggesting that the Gram-negative species had a larger
proton motive force effect on protonation of the cell wall functional groups.
The objective of our study is to further investigate the effect of metabolic activity
on the ability of bacterial cell wall functional groups to adsorb aqueous metal cations.
These experiments are similar to those conducted by Urrutia Mera et al. (1992) in that we
compare metal uptake onto inactivated cells to that observed for metabolically-active
control suspensions. However, in this study the adsorption of a range of concentrations of
aqueous Cd2+ is studied under clearly undersaturated conditions, metal-cell suspensions
are allowed to equilibrate for 2.5 h, and all experiments are conducted in a nutrient
growth medium to insure substantial metabolic activity in our untreated control cultures.
Varying the Cd2+ concentration allows us to observe the effects of metabolism on Cd
48
adsorption at different metal-bacteria ratios and to demonstrate that effects on Cd binding
are not due to metal toxicity, but rather to changes in the metabolic state of the cells.
Furthermore, we examine a set of both Gram-positive (B. subtilis and Bacillus cereus)
and Gram-negative (Pseudomonas fluorescens and Shewanella oneidensis) bacterial
species in order to determine if cell wall structure influences the metabolic effects on
adsorption.
3.2
Methods
3.2.1 Bacterial Strains and Culture Conditions
Cultures of P. fluorescens str. ATCC 11764, B. cereus str. ATCC 6462, S.
oneidensis str. MR1, and B. subtilis str. ATCC 6051 were maintained as liquid 100-ml
shake-flask cultures (150 rpm; 1-in stroke dia) at room temperature or at 32 °C in LB
broth (Sambrook et al., 1989). Fresh liquid cultures were inoculated every two weeks
from clonal colonies on LB agar plates. Actively growing log-phase subcultures were
used for growth and binding experiments.
3.2.2 Metabolic Treatment and Cadmium Binding
Although Urrutia Mera et al. (1992) and Kemper et al. (1993) used a range of
metabolic inhibitors, the current study used formalin (a 37% by weight formaldehyde
solution in water) treatments of cell suspensions to inhibit the metabolism of both Grampositive and Gram-negative bacterial cells. Our measurements (see below) demonstrate
that formalin is faster and more effective than sodium azide at inhibiting cellular
metabolism, and is unlikely to have the potentially damaging effects that high doses of
49
radiation may have on cell wall structures. However, in order to facilitate direct
comparison of these results to those of Urrutia Mera et al. (1992) and Kemper et al.
(1993) separate experiments were conducted using only B. subtilis in which formalin was
replaced with sodium azide as the metabolic inhibitor.
Two 100-mL aliquots of bacterial inoculum from a 500-ml stock culture were
distributed into 250-mL polypropylene bottles. One aliquot was treated with 2.5% (w/v)
formalin, and one aliquot was left as an untreated control. Both the formalin–treated
culture and the untreated control culture were incubated under identical conditions (at 25
°C) for 30 min. The B. subtilis-sodium azide experiments were conducted in the same
manner, however one aliquot was treated with 0.5 % (w/v) sodium azide and the
experimental cultures were incubated for 6 h. Preliminary respirometry experiments
demonstrated that bacteria require a longer exposure period to sodium azide to inactivate
cells. Note that in the untreated control experiments, cell reproduction occurred and
culture density increased during the incubation period, while this did not occur to a
significant extent in the treated suspensions. To account for this unavoidable difference in
cell concentration, Cd adsorption concentration results were normalized using cell mass
present at the end of the incubation. Since the formalin treatment requires a shorter
incubation time to be effective, there was less of a difference between the cell
concentrations in the treated and untreated systems, and this was another reason why the
formalin approach was favored over the sodium azide treatment to inactivate the bacterial
cells. After the incubation period, the bacterial cultures were harvested by centrifugation
for 10 min at 6000 g. Bacterial cell pellets were then resuspended in 100 mL of assay
medium (in g/L: KNO3, 0.55; NaNO3, 0.47; HEPES, 2.37; with glycerol used as the
50
carbon and energy source at 0.5% (w/v)), and supplemented with the appropriate Cd
concentration (3, 10, or 20 ppm from a 1000 ppm Cd nitrate reference solution). The
solution pH was adjusted to ~7.0 with HCl or NaOH and maintained by the HEPES
buffer. Cd concentrations were well below saturation values. The assay medium was
designed to limit competing ions such as phosphate and chloride, although the ionic
strength was similar to typical basal salts media for bacteria. All adsorption experiments
were performed with three or more replicates. All metal incubations were performed
while rotating the reaction vessel end over end on a carousel for 2.5 h. Preliminary
experiments showed that this duration was more than sufficient to ensure binding
equilibrium.
After the 2.5 h reaction period, the pH of each assay culture was measured and
cell density determined by measuring absorbance at 600 nm (Cary 300
spectrophotometer). Experiments were performed with each bacterial species to
determine a conversion factor between absorbance and wet weight, with 1 OD unit at 600
nm corresponding to 2.613, 1.558, 1.743, and 2.920 g L-1 wet weight for B. subtilis, B.
cereus, P. fluorescens, and S. oneidensis, respectively. These values were used to
normalize the Cd adsorption data, and adsorption results are reported in terms of g of Cd
bound per mg wet weight of bacteria. After cell densities were determined for each
experiment, the supernatant from each assay was collected by centrifugation at 10,000 g
for 3 min and filtered with a 0.45 µm filter (Osmonics Cameo 30N). The filtered
solutions were then acidified with a small aliquot of concentrated HNO3 and stored at 4
ºC for no longer than 1 wk. The final dissolved Cd concentration in each of the assay
supernatants was determined by inductively coupled plasma – optical emission
51
spectroscopy (ICP-OES), with matrix–matched standards for calibration. The amount of
Cd that was adsorbed onto the bacteria during each assay was determined as the
difference between the amount of Cd in solution at the end of the assay and the initial Cd
concentration. Control experiments without bacteria were performed simultaneously with
the cell binding assays to determine the amount of nonspecific Cd adsorption onto the
experimental apparatus. The average adsorption of three controls was 0.2 ppm Cd loss
from an initial 10 ppm Cd solution, and this value was subtracted from each experimental
value to account for loss of Cd.
3.2.3 Metabolic Activity Measurements
The effectiveness of the inactivation treatments was determined by viability
staining, INT staining, and O2 respiration measurements. Sodium azide affects cells in
such a way that for some time after treatment the cells appear alive with viability staining
and INT staining. Therefore, respiration was also measured in treated cultures using a
Clark-type oxygen electrode (YSI 5000) calibrated with air-saturated water. INT staining
was performed following the method of Bovill et al. (1994) where cells treated with a
tetrazolium dye were harvested, washed, resuspended in buffer, and examined
microscopically for refractile spots diagnostic of metabolic activity. Viability staining
used as a measure of membrane integrity was performed following the manufacturer's
instructions (BacLight Kit, Molecular Probes, Eugene, OR).
52
3.3
Results
A variety of techniques were used to measure the effectiveness of the formalin
and sodium azide treatments in inhibiting cell metabolism, and to confirm that metabolic
activity occurred in the untreated systems. Figure 3.1 shows respirometer measurements
for suspensions of both treated and untreated B. subtilis cells, with and without Cd
present in the system. These results are representative of all of the bacterial species used
in the current study, but for clarity only the B. subtilis results are shown. The
respirometer measurements show that untreated cells, with or without Cd present, remove
90-95% of the oxygen in the system within the first minute of the assay. Cells treated
with either formalin or sodium azide remove a maximum of 10-15% of the oxygen in the
system over the entire length of the experiment. Thus, the respirometry experiments
provide evidence that untreated bacteria maintain active metabolism under the conditions
of the binding assays, even in the presence of Cd, and that cells treated with either
formalin or sodium azide metabolize at a greatly diminished rate. The results of INT
staining and viability staining of formalin-treated cells were consistent with the
respiration results, demonstrating that the cells were no longer metabolically active. For
the azide-treated cells, the plasma membrane apparently remained intact and INT staining
is not an appropriate measure of electron transport activity, as azide blocks the system
beyond the point detected by the INT reaction.
The design of these experiments allows for comparison of the amount of Cd
bound to metabolically active cells relative to that bound to inactivated cells. Preliminary
experiments showed a measurable increase in bacterial culture density over the course of
53
Oxygen Concentration
(% atmospheric)
100
90
80
70
60
50
40
30
20
10
0
0
1
2
3
4
5
6
Time (min)
Untreated Cells (w/out Cd)
Formalin-treated Cells
Untreated Cells (w/Cd)
Azide-treated Cells
Figure 3.1. Representative oxygen consumption by treated and non-treated bacterial cells.
the experiments with untreated cells, another indication that bacterial metabolism was
active in the assay medium. Conversely, there was no difference between the initial and
final bacterial culture densities in experiments with treated cells. Since the final bacterial
culture densities differed in each experiment due to normal variations in growth, the
concentrations of adsorbed Cd given in Figures 3.2-3.5 were normalized to the final
bacterial culture density. Error bars in Figures 3.2-3.5 represent the standard error of all
replicate experiments. The uncertainties for these experiments are potentially higher than
those typically seen for metal adsorption experiments due to the treatment of the bacterial
cells prior to the metal adsorption experiments. Also, the cells were not acid washed or
rinsed with an electrolyte to remove cations that may have adsorbed onto the cells from
the growth medium.
54
In all of the assays with Gram-positive cells, higher Cd adsorption was observed
with inactivated cells than in the systems that contained metabolically active cells.
Depending on the experimental Cd concentration, Cd adsorption onto metabolically
active B. subtilis was between 24 and 48% less than that exhibited by the inactive cells
that were treated with formalin; and between 22 and 32% lower than cells that were
treated with sodium azide (Figure 3.2a). Cd adsorption onto untreated metabolizing B.
cereus cells was between 30 and 53% lower than formalin-treated cells (Figure 3.2b).
Although the inactivation treatment had a consistent effect on the extent of Cd adsorption
onto the two Gram-positive species, there was no distinguishable trend in the effect as a
function of Cd concentration for either species. The results of the B. subtilis experiments
suggest that sodium azide had a consistently smaller impact on cell metabolism, as cells
inhibited with formalin exhibited higher metal adsorption than those treated with sodium
azide (Figure 3.2a).
The results for B. subtilis adsorption are in agreement with those of Urrutia Mera
et al. (1992) in that an increase in cation adsorption was observed in the experiments that
involved inactivated cells. However, the effect observed by Urrutia Mera et al. (1992)
was larger than that reported here, possibly due to differences in wash procedures
between the two studies or differences in the metals used in each study. Urrutia Mera et
al. (1992) also observed a greater change in bulk solution pH during their experiments.
The initial pH of 7.0 in the Urrutia Mera et al. (1992) experiments decreased to 6.5 for
their azide-treated cells and to 6.4 for their control cells, whereas the pH in the current
experiments decreased from 7.0 to pH 6.8.
55
Figure 3.2. Comparison of the amount of Cd adsorbed onto metabolizing and nonmetabolizing bacterial cells for experiments containing 3, 10, and 20 ppm Cd. F =
formalin–treatment and SA = sodium azide–treatment (for B. subtilis only), where Active
denotes untreated cells incubated for identical times to Inactive cells treated with
formalin or sodium azide, respectively: a. B. subtilis; b. B. cereus; c. P. fluorescens; and
d. S. oneidensis. The error bars represent the standard error.
56
Figure 3.2 (continued)
(a)
Cd (ug) adsorbed per mg bacteria
B. subtilis
4
Active
3.5
Inactive
3
2.5
2
1.5
1
0.5
0
F3
F10
F20
SA3
(b)
Cd (ug) adsorbed per mg bacteria
B. cereus
4
3.5
Active
Inactive
3
2.5
2
1.5
1
0.5
0
F3
F10
F20
57
SA10
SA20
Figure 3.2 (continued)
(c)
Cd (ug) adsorbed per mg bacteria
P. fluorescens
10
9
8
Active
Inactive
7
6
5
4
3
2
1
0
F3
F10
F20
(d)
Cd (ug) adsorbed per mg bacteria
S. oneidensis
5
Active
Inactive
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
F3
F10
F20
58
In contrast to what was observed in the experiments with Gram-positive bacterial
cells, the results from experiments with Gram-negative bacterial cells show no systematic
effect of metabolism on the extent of Cd adsorption onto the cells. The inactivated cells
exhibited approximately the same extent of Cd adsorption as the actively metabolizing
cells. For P. fluorescens, the difference between the observed extent of adsorption by
active and inactive cells for each Cd concentration studied was less than 12% (Figure
3.2c). Similar behavior for S. oneidensis to that seen for P. fluorescens is depicted in
Figure 3.2d: No consistent difference was observed in the extent of Cd adsorption onto
active cells relative to the inactive cells, and this is true for all Cd concentrations studied.
The 10-ppm experiments with P. fluorescens and S. oneidensis exhibited a slight increase
in the extent of adsorption onto the treated cells, but the opposite was observed in the 20ppm experiments. The observation of a larger metabolic effect in Gram-positive species
relative to that observed for Gram-negative species is inconsistent with the potentiometric
titrations conducted by Claessens et al. (2006), who documented greater long-term proton
release in suspensions of metabolically-active Gram-negative cells compared with Grampositive cells.
3.4 Discussion
The results of this study suggest that adsorption of Cd onto Gram-positive
bacteria is significantly diminished by the presence of a proton motive force, while the
effect on Gram-negative bacteria is small to negligible. As suggested by Urrutia Mera et
al. (1992), who conducted similar experiments to these using only the Gram-positive
species B. subtilis, the decrease in adsorption associated with active metabolism observed
59
for Gram-positive bacteria is likely due to increased competition between Cd2+ and the
H+ ions effluxed into the periplasmic space, adjacent to negatively charged functional
groups on the cell walls of metabolizing bacteria. Urrutia Mera et al. (1992)
hypothesized that the proton motive force exerts a similar effect on both Gram-positive
and Gram-negative bacterial species. However, the differences in adsorption behavior
between Gram-positive and Gram-negative species observed here suggest that metal
binding to Gram-negative and Gram-positive bacteria is affected by the proton motive
force in substantially different ways.
Gram-positive bacterial cell walls consist of a thick layer of peptidoglycan
overlying the plasma membrane (Beveridge & Murray, 1980). For Gram-positive
bacteria, the proton motive force is generated by protons effluxed into the periplasmic
space between the plasma membrane and the peptidoglycan layers, thus, acidifying to
some extent the cell wall region. Gram-negative bacterial cell walls consist of a plasma
membrane overlain with a relatively thin peptidoglycan layer, which in turn is overlain by
a lipopolysaccharide outer membrane. The proton motive force in Gram-negative species,
like that in Gram-positive species, should at least partially neutralize the electronegativity
of the peptidoglycan layer. The observation that metabolizing and non-metabolizing
Gram-negative bacteria exhibit similar extents of Cd adsorption can be explained by the
following hypotheses: either 1) the extent of the influx of protons into the cell wall is
smaller for Gram-negative bacterial species than it is for Gram-positive species; or 2) the
primary sites of binding of Cd in Gram-negative cells under these experimental
conditions is exterior to the peptidoglycan layer, and the proton-motive force does not
extend out to these primary sites of metal binding within the Gram-negative cell wall
60
structure. It is possible that the flux of protons during Gram-negative metabolism
neutralizes the local charge within the peptidoglycan layer, but does not affect the more
external surface of the outer membrane and associated structures. This interpretation
suggests that a significant component of the metal adsorbed onto Gram-negative cells is
bound to the phosphoryl groups of the outer membrane and associated structures, likely
involving the functional groups within the outer membrane phospholipids and
lipopolysaccharides, and does not involve the peptidoglycan layer as much as is
suggested for Gram-positive species (Beveridge, 1989).
In this study, B. subtilis cells treated with formalin adsorbed more Cd than those
treated with sodium azide (Figure 3.2a); this suggests that formalin is more effective at
retarding metabolism in these bacteria. The lower oxygen consumption exhibited by the
formalin-treated cells relative to those treated with sodium azide (Figure 3.1) also
supports this premise, however it should be noted that the two are likely within
experimental error. In an analysis of the effectiveness of various treatments in retarding
microbial activity in sediment trap material, Lee et al. (1992) found that formalin and
chloroform treatments reduced microbial activity to undetectable levels, but that the
activity of azide-treated samples were reduced only to 10-20% of their initial value. The
difference in Cd adsorption between the formalin- and the sodium azide-treated cells
(Figure 3.2a) is likely due to the manner in which the treatments impact the bacterial cell.
Sodium azide is a proton-conducting uncoupler; it inhibits oxidative phosphorylation by
binding to the terminal cytochromes of most respiring bacteria, preventing electron
transfer (Harold, 1972). Cytochromes are generally membrane-bound proteins that carry
out electron transport or catalyze reductive/oxidative reactions. Conversely, formalin
61
inactivates proteins by forming covalent crosslinks with several functional groups; this
dehydrates the cells and replaces the normal fluid with a gel-like rigid complex.
Surface complexation modeling can be used to quantify the proton flux
responsible for the observed decrease in Cd adsorption capacity in metabolically active
Gram-positive cells. As discussed above, the decreased adsorption associated with the
metabolizing Gram-positive cells is likely due to a lower local pH in the near-surface
region, inhibiting adsorption onto these cells. Since surface complexation modeling can
be used to account for the pH dependence of adsorption (e.g., Plette et al., 1995, 1996;
Fein et al., 1997; Cox et al., 1999), it is possible that this modeling approach could be
used to model the effects of metabolism on adsorption if the extent of the surface pH
change was known. Conversely, a thermodynamic modeling approach can be used to
estimate the pH of the cell wall region of metabolizing Gram-positive bacteria. Based on
the decrease in adsorption that accompanies bacterial metabolism, surface complexation
modeling was used to determine the effective pH of the region where Cd binding occurs
on the B. subtilis cell wall. This calculation can only be conducted for the B. subtilis
experiments, because acidity constants of the cell wall functional group sites and
associated Cd binding constants have not been determined for the other bacterial species
used in this study. The surface complexation model accounts for the protonation
behavior of the cell wall using the deprotonation constants and surface site concentrations
of discrete cell wall functional groups in the 4-site model developed by Fein et al. (2005).
The deprotonation reactions are represented by the following equation:
−
R − Ln H 0 ⇔ R − Ln + H +
62
(1)
where R is the bacterium to which each functional group type Ln is attached. The mass
action equation for the above deprotonation reaction is
−
[ R − Ln ]a H +
Ka =
(2)
[ R − Ln H 0 ]
where Ka represents the acidity constant, a is the activity of the subscripted species, and
brackets denote the concentration of surface sites in mol/L of solution. Although Fein et
al. (2005) model the bacterial cell wall acidity using 4 sites, under near neutral pH
conditions, Sites 2 and 3, with log Ka values of 4.8 and 6.8, respectively, are the metal
adsorption sites of interest. The two-site model of Borrok et al. (2004b) was used to
account for Cd adsorption onto B. subtilis:
Cd 2+ + R − Ln
( −1)
⇔ R − Ln − Cd +
(3)
The corresponding mass action equation relates the Cd binding constant, Kads, to
components of reaction (3):
K a ds =
[ R − Ln − Cd + ]
−
a Cd 2 + [ R − Ln ]
(4)
The log Kads values for Cd adsorption onto cell wall Sites 2 and 3 are 3.4 and 4.6,
respectively. The above mass action equations (equations 2 and 4), along with mass
action equations for aqueous Cd hydrolysis, and mass balance constraints on total Cd and
bacterial site concentrations were used to determine pH conditions at the binding sites of
the actively metabolizing bacteria. The initial bulk solution pH of both metabolizing and
non-metabolizing experiments was 7.0, and it was assumed that the observed decrease in
adsorption was due exclusively to a decrease in pH at the binding sites. The 3.0 ppm Cd
experiments involving formalin-treated B. subtilis cells exhibited a change in adsorbed
63
Cd from 1.3 ppm to 0.7 ppm associated with metabolic activity. Calculations show that
this decrease in adsorption would be caused by a decrease in pH at the binding sites from
7.0 to 5.7. Similarly, the calculations for the 10- and 20-ppm experiments suggest a pH
decrease that is associated with bacterial metabolism from 7.0 to 6.3 and from 7.0 to 6.2,
respectively, yielding an average calculated effective pH at the binding sites of actively
metabolizing cells of 6.1 ± 0.3. The experiments involving sodium azide-treated B.
subtilis cells yield pH changes (from the bulk solution pH of 7.0) associated with
metabolism to values of 6.1, 6.4, and 6.0 for the 3-, 10-, and 20-ppm Cd experiments,
respectively, with an average calculated effective pH at the Cd binding sites on the
actively metabolizing cells of 6.2 ± 0.2.
3.5 Conclusions
This study demonstrates that the metabolic activity of Gram-positive bacteria has
a significant impact on Cd adsorption onto cell wall functional groups. Metabolizing
Gram-positive cells adsorbed significantly less Cd than did non-metabolizing Grampositive cells. Conversely, metabolizing and non-metabolizing Gram-negative cells
exhibited roughly similar extents of Cd adsorption. The effect of bacterial metabolism on
Cd adsorption onto Gram-positive cells is likely due to a local decrease in pH in the cell
wall region where Cd is bound. The lack of an observable effect of metabolism on Cd
adsorption onto Gram-negative cells suggests that Cd binding occurs at a greater distance
from the inner plasma membrane than occurs within the cell wall of Gram-positive cells.
A surface complexation modeling approach was used to estimate that the effect of the
metabolic proton motive force on the Gram-positive cell wall was to decrease the pH of
64
this region by approximately one pH unit. This study demonstrates that bacterial
metabolic state can influence the extent of passive metal adsorption onto cell wall
functional groups, at least for bacterial species similar to the Gram-positive species
studied here. Adsorption onto actively metabolizing bacterial cells has been modeled
using a surface complexation approach and the results suggest that the decrease in pH in
Gram-positive cell walls due to metabolism was approximately one pH unit. The results
also suggest that biosorption remediation strategies that involve Gram-positive bacterial
species may be more efficient at metals removal from solution if bacterial metabolism is
inhibited during the sorption process.
65
CHAPTER 4
PROTON AND METAL ADSORPTION ONTO BACTERIAL CONSORTIA:
SIMILARITIES IN ADSORPTION BEHAVIORS AND ESTIMATION OF
GENERALIZED STABILITY CONSTANTS FOR METAL-BACTERIAL SURFACE
COMPLEXES
4.1
Introduction
Bacterial surfaces can adsorb a wide range of heavy metal contaminants (e.g.,
Beveridge and Murray, 1976; Beveridge and Koval, 1981; Mullen et al, 1989). To better
constrain and mitigate contaminant transport in the environment, it is important to
develop models that determine the influence of bacteria on the speciation and distribution
of heavy metals in the sub-surface. Site-specific surface complexation models, originally
developed to quantify cation adsorption to mineral surfaces, have been successfully used
to account for proton and metal adsorption to bacterial surfaces (e.g., Plette et al., 1996;
Fein et al., 1997; Haas et al., 2001; Martinez et al., 2002). However, if each bacterial
species exhibits unique adsorption characteristics as mineral surfaces do, then it would be
an overwhelming task to determine the stability constants, site concentrations and acidity
constants necessary for modeling metal adsorption onto all of the bacteria of
environmental and geologic interest. A single location in a natural system can contain
many bacterial species, and bacterial diversity can change from one location to another.
66
Consequently, if surface complexation models are to be applied to realistic systems, it is
important to determine if proton and metal adsorption behavior is species-specific or if
commonalities exist among bacterial species.
A number of studies have noted similar adsorption behavior among individual
bacterial species (e.g., Daughney et al., 1998; Small et al., 1999; Kulczycki et al., 2002;
Ngwenya et al., 2003) and among artificial mixtures of pure strains of bacteria (Yee and
Fein, 2003). Yee and Fein (2001) hypothesized that similarities in adsorption
mechanisms exist for a wide range of bacterial species, and they conducted
potentiometric titrations and Cd-bacteria adsorption experiments using a range of Grampositive and Gram-negative species. Yee and Fein (2001) observed similar adsorption
behavior for the variety of bacteria studied, suggesting that the structures that give rise to
metal and proton adsorption are common over a wide range of bacterial species.
The hypothesis of universal bacterial adsorption behavior has been supported by a
number of subsequent experimental studies. For example, Jiang et al. (2004)
demonstrated that the attenuated total reflectance Fourier-transform infrared spectra of
both Gram-positive and Gram-negative bacteria are similar and exhibit similar variations
as a function pH. These similarities suggest a similarity in binding environments for
metals between species, supporting a universal adsorption behavior that arises from
similar cell wall functional group chemistries. Borrok et al. (2004a) measured H+ and Cd
adsorption onto bacterial consortia from a range of natural environments, demonstrating
that the consortia exhibit similar proton and Cd adsorption behaviors, and that the
adsorption onto all of the consortia can be modeled using a single set of stability
constants. In addition, Borrok et al. (2005) compiled all currently available
67
potentiometric titration datasets for individual bacterial species and bacterial consortia,
noting general similarities in the proton adsorption behaviors and presenting an
internally-consistent averaged set of ‘universal’ thermodynamic proton binding and site
density parameters for modeling bacterial adsorption reactions in geologic systems.
Although a large number of bacterial species appear to exhibit broadly similar adsorption
behavior, some studies suggest that at least some bacteria have significantly different
adsorptive properties. For example, Borrok et al. (2004b) showed that some bacteria that
thrive in hydrocarbon-contaminated environments exhibit significantly enhanced
adsorptive behavior compared to those from uncontaminated systems.
In this study, we expand the study of natural consortia of Borrok et al. (2004a) to
test whether we observe similarities in binding environments for a much wider range of
bacterial species than was tested by Borrok et al. (2004a), and we measure the extent of
adsorption of other metals onto bacterial consortia as well. We obtain our range of
bacterial diversity by growing consortia from samples taken from three natural settings
and sampling those settings over the course of a year. We conduct potentiometric
titrations using these bacterial consortia, and we measure the extents of Ca, Cd, Cu, Pb,
Sr, and Zn adsorption onto the consortia as well. The results suggest strong similarities in
binding environments on the bacterial cell walls, and we use the measurements to
determine average stability constants for the important metal-bacterial surface complexes.
We use the stability constants to constrain relationships between these values and metalacetate stability constants so that our results can be extrapolated to other metals of
environmental interest.
68
4.2
Methods
4.2.1 Sampling and Growth of Bacteria
Sample locations were in northern Indiana and included a river, a forest, and a
soybean field site. Samples were collected 7 times from all three sites over the course of a
year (October 2004 through September 2005) for potentiometric titration and Cd
adsorption experiments. River water samples were also collected from October 2005
through January 2006 for follow-up experiments involving other metal cations. Bottles
and scoops used to collect samples were sterilized and sealed in plastic bags before use.
Water samples were collected by dipping the sample jar directly into the river. Soil
samples were collected by removing the top 5 to 10 cm of topsoil and debris, and then
directly scooping the soil specimen using the glass sample jar. Lids were placed loosely
over the jars to allow for aerobic conditions and to prevent contamination.
Approximately 10 mL of water from the river samples were used to inoculate 2 L
of LB broth (Sambrook et al., 1989). Ten g of soil from the forest or soybean samples
were used to inoculate 75 mL of LB broth. To dilute the solid fraction present in the soil
samples, approximately 10 mL of the initial bacteria-broth suspension were used to
inoculate larger quantities of identical broth solutions. Samples were shaken gently at
room temperature for a total of 7 days before they were harvested for experiments.
Many bacterial species are unculturable, and experiments have shown that within
a natural consortium many species cannot survive repeated inoculations (Kaeberlein et
al., 2002). Therefore, the bacteria grown from our experiments are likely a subset of the
total bacterial population at each sample site. For example, because all growth conditions
69
were aerobic, all anaerobes were eliminated through the growth procedures. However,
our experimental approach employing a single re-inoculation insured that the bacterial
consortia produced from each sampling had a range of the bacteria that was present in
each environment, but also had sufficient biomass to conduct the experiments. All
experiments were conducted within seven days of sampling.
To prepare bacteria for potentiometric titrations and metal adsorption
experiments, following the initial 7 day growth period, bacteria were harvested by
centrifugation for 10 min at 6000 rpm (2220 g) and then rinsed 5 times with a 0.1 M
NaClO4 solution. NaClO4 was chosen as the electrolyte because perchlorate does not bind
protons or the metals of interest to an appreciable extent under the experimental
conditions. After each wash, bacteria were suspended in clean electrolyte in a test tube
using a vortex machine and stir rod. Bacteria were then centrifuged for 3 min at 7200 g to
form a bacterial pellet, and the supernatant was decanted. Following the final wash,
bacteria were resuspended in test tubes and centrifuged (7200 g at 25˚C) for 1 h, stopping
2 times to decant the supernatant. Following centrifugation, the wet weight of the bacteria
was determined and used for calculation of bacterial concentrations for experiments. For
a discussion on wet vs. dry weight, see Borrok et al. (2004a). The bacteria were
immediately used in titrations or metal adsorption experiments. The bacterial cells remain
viable after the washing treatment; however, they are not likely to be undergoing active
metabolism due to the thorough wash, the lack of nutrients and electron donors in the
experimental solutions, and the relatively short (<3 hours) experimental durations.
70
4.2.2 Potentiometric Titrations and Metal Adsorption Experiments
Prior to titrating bacteria, 0.1 M NaClO4 was purged of CO2 by N2 bubbling for
60 min. Following this step, the harvested bacteria (0.41 ± 0.02 g) were suspended in
12.13 ± 0.13 mL of the electrolyte, and titrations were conducted in an N2 atmosphere
with an automatic burette assembly. When conducting titrations, the acid or base was
added in minute amounts when the stability of the suspension attained a change of 0.1
mV sec-1 or less. The suspensions were titrated with 1.001N HNO3 to pH ≈ 2.3, and then
they were titrated with 1.037N NaOH to pH ≈ 10. We chose to titrate to the lower pH
because previous research shows proton adsorption occurs down to at least pH 2.3 (Fein
et al., 2005). Potentiometric titrations were performed in triplicate for all river, soybean
field, and forest sample consortia grown throughout the year.
Metal adsorption experiments were performed as a function of pH using Ca, Cd,
Cu, Pb, Sr, and Zn. Cd adsorption experiments were completed for all river, soybean
field, and forest sample consortia grown throughout the year. All other metal adsorption
experiments were conducted using the October, 2005 through January, 2006 river
samples. All experiments were conducted with freshly sampled/grown bacteria. Cd
adsorption kinetics experiments were conducted for each of the types of samples.
Experiments were conducted at a pH between 6.0 and 6.5, and full equilibrium was
reached within 1.5 h for all consortia tested.
For all Cd adsorption experiments, approximately 10 gL-1 of a bacterial
consortium was suspended in a pH-neutralized stock solution of 0.1M NaClO4 and
approximately 10 ppm Cd. In all adsorption experiments, the exact concentrations of both
the bacteria and the metal of interest were determined gravimetrically. The respective
71
bacterial concentrations for experiments involving Ca, Cu, Pb, Sr, and Zn, are: 11.4, 5.72,
3.3, 20.05, and 11.34 g L-1, respectively, and their respective metal concentrations are:
5.24, 5.25, 10.13, 3.19, and 10.31 ppm. Following a 10 min initial equilibration period,
the metal/bacteria stock solution was divided into individual polypropylene reaction
vessels. The pH of the suspensions in these vessels was then adjusted to the desired pH
by adding minute aliquots of 0.1 to 1 M HNO3 or NaOH. The vessels were then rotated
slowly end over end on a rotating rack for 2 h, and the final pH was measured. The
reaction vessel was then centrifuged at 4500 g for 3 min, and the supernatant was
collected after being filtered through a 0.45 µm filter (Osmonics Cameo 30N). The
filtered solutions were acidified with a small aliquot of concentrated HNO3 and stored at
4 ºC for no longer than 7 days prior to analysis for dissolved metal concentration. The
final dissolved metal concentration in each of the sample supernatants was determined by
inductively coupled plasma – optical emission spectroscopy (ICP-OES), with matrix–
matched standards for calibration. The amount of metal that was adsorbed onto the
bacteria during each experiment was determined as the difference between the measured
concentration of metal in solution at the end of the experiment and the known initial
metal concentration. Control experiments without bacteria were performed
simultaneously to determine the amount of nonspecific metal adsorption onto the
experimental apparatus.
4.2.3 Gram Staining and DGGE Analysis
Gram staining was performed periodically throughout the year on consortia grown
from each of the 3 sample sites by heat-fixing the cells to a glass slide and then staining
72
the cells using a PROTOCOL Gram stain kit from Fisher Scientific. Denaturing gradient
gel electrophoresis (DGGE) analysis was performed for all consortia collected between
October 2004 and September 2005. DGGE employs a linear denaturing gradient to
separate fragments of DNA by size; therefore, we are able to better constrain community
speciation and qualitatively assess the change in bacterial diversity throughout the year.
A MoBio Laboratories, Inc. Ultraclean soil DNA kit was used to extract DNA, which
was frozen at -20 ˚C prior to amplification. To prepare the DNA for DGGE analysis, a
polymerase chain reaction (PCR) process using a custom-made universal bacterial primer
set (EUB 341 and EUB 534, 200 base pairs in length with a GC-clamp) was used to
amplify a specific 16s rDNA sequence (Muyzer et al., 1993). DGGE was performed
using a Dcode universal mutation detection system (Bio-Rad). The PCR product DNA
was loaded into a gel with a 30% to 60% gradient composed of urea and formamide (a
chemical denaturant), and an electrical potential was applied to force the DNA to travel
through the gel. Gels were run at 60 ˚C and a potential of 60 V for 14 h. Ethidium
bromide was used to stain the gel, it was then photographed using a Kodak EDAS 290
photographic system. Bacterial species have different DNA base pair sequences,
therefore, the DNA for each species forms a characteristic band in the gel because it
denatures at a specific point along the gel. Because the intensity of the band is directly
related to the concentration of DNA in the sample, analysis of the band positions and
intensities can determine the minimum number of bacterial species present and their
relative abundances. The DGGE analysis sufficiently demonstrates the change in
bacterial populations with sample times and locations, but additional sequencing of the
73
bands would be required to verify that each band represents only one species and to
determine the identity of the species.
74
4.3
Results and Discussion
4.3.1 Bacterial Diversity
The Gram staining results indicate that the bacterial populations of each
consortium contained both Gram-negative and Gram-positive bacterial species. There
was not a distinguishable trend in the relative abundances throughout the year. In each
sample, there was generally a mix of rod-shaped, cocci, and spirilla bacteria, and the
Gram-negative bacteria were generally 2 to 10 times smaller than the Gram-positive
bacteria. The consortia grown from the river water samples were generally dominated by
Gram-negative bacteria, but there were two months (January and May) when the bacteria
had equal amounts of both, and one month (September) when Gram-positives were more
abundant. The bacteria harvested from the forest and soy field sample sites typically
contained a similar ratio of Gram-positive and Gram-negative, with a larger number of
Gram-negative bacteria during the winter months (January) and more Gram-positive
species toward the end of summer (September).
The consortia used in this study displayed between 6 and 14 bands (Figure 4.1) in
the DGGE analysis. The greatest population diversity occurred in the March consortia for
all sample sites (the river and forest site consortia displayed 14 bands each, and the soy
field site consortium displayed 13 bands). The lowest number of bands generally
occurred during the months of November and December. Bacterial species populations
typically vary from site to site and as a function of time; therefore our sampling approach
was successful in creating bacterial consortia with a wide range of diversity, thereby
75
River
Soybean Crop
Forest
Figure 4.1. DGGE gel with labeled site locations. Each section contains a lane for each
sampling month, Oct., Nov/Dec., Jan., March, May, June, and September. Each band in a
lane represents a different bacterial species.
enabling a rigorous test of adsorption behavior as a function of changing bacterial
diversity.
4.3.2 Potentiometric Titrations
All data are plotted in terms of mmoles of deprotonated sites per mass of bacteria
(mmol/g),
Net Molality Protons Added = (Ca – Cb – [H+] + [OH-])/mb
(1)
where Ca and Cb are the concentrations of acid and base added at each step of a titration,
brackets represent mmolal species concentrations, and mb is the bacterial wet weight
suspension concentration (g L-1). Figure 4.2a depicts the results from the titration for the
76
21 consortia grown in this study, and which are virtually identical to each other. Although
there are slight differences between individual titration curves, there are no significant
trends in those differences between the titration curves for the consortia from one site
relative to the others, nor as a function of sampling time during the year. All consortia
displayed a significant buffering capacity over the entire pH range of this study (2.5-9.5).
The curves exhibited a similar shape to those exhibited by natural consortia and by a
wide range of individual bacterial species (e.g., Plette et al., 1995; Fein et al., 1997; Haas
et al., 2001; Yee and Fein, 2001; Martinez et al., 2002; Ngwenya et al., 2003; Borrok et
al., 2004a; Fein et al., 2005). On a per gram basis, the natural consortia exhibited an
average buffering capacity of 4.58 x 10-4 mol g-1 over the experimental pH range, a
similar buffering capacity to the 3.0 x 10-4 mol g-1 buffering capacity exhibited by
Bacillus subtilis over a pH range of 3-10 (Fein et al., 2005).
4.3.3 Metal Adsorption Experiments
In general, the extent of Cd adsorption to natural consortia was similar for the
river, soy field, and forest sites (Figure 4.3). While the river consortia in general adsorbed
slightly less Cd than the others, the consortia grown from the soy field and forest
environments demonstrated indistinguishable adsorption capacities. There was not a
noticeable trend in Cd adsorption as a function of time of sampling. These results indicate
that the large changes in species diversity that is depicted in the DGGE results from
Figure 4.4.1 do not have a large impact on the metal adsorption behaviors. The extent of
Cd adsorption in this study is higher than that observed in the similar study conducted by
Borrok et al. (2004a); both studies were conducted at the same ionic strength and
77
bacterial and Cd concentrations, however Borrok et al. (2004a) used a different growth
medium and they froze some consortia samples prior to experimentation. At pH 7, we
observed 75 to 92% Cd adsorption and Borrok et al. (2004a) observed 52-70% metal
adsorption for their natural consortia experiments; however, in their study of Cd
adsorption to consortia grown from contaminated environments, Borrok et al. (2004b)
reported 75-95% Cd adsorption at pH 7. In single species Cd adsorption experiments,
Fein et al. (1997) observed a similar extent of Cd adsorption to the results in this study. In
general, the observed extents of Ca, Cu, Pb, Sr, and Zn adsorption onto the river water
consortia were similar to those observed for individual bacterial species (Fein et al.,
1997; Fowle and Fein, 1999; Fein et al., 2001; Borrok et al., 2005) under similar
experimental conditions.
4.3.4 Surface Complexation Modeling
The diversity of the bacterial species represented in the natural consortia of this
study did not have a large impact on the proton or metal uptake capabilities of the
consortia. This suggests that the adsorption behavior of all of the consortia can be
modeled using a thermodynamic modeling approach with a single set of averaged
thermodynamic parameters. Although a range of types of models can be used to account
for the proton and metal adsorption behavior, we choose to apply a discrete site, nonelectrostatic surface complexation model (SCM), similar to that applied by Borrok et al.
(2004a) and by Fein et al. (2005). When applying this approach to model the acidity of
the surfaces of the consortia, we assume that adsorption is due to proton and metal cation
78
interaction with negatively charged organic acid functional groups on the bacterial cell
walls. This approach implicitly treats the numerous cell wall types in each consortium
79
Figure 4.2. (a) Potentiometric titration results for all titrations conducted for (+) river, (∆)
soybean crop, and (O) forest sites. (b) Example experimental potentiometric titration data
(O). Model curve calculated using the averaged pKa’s and surface site concentrations.
Dashed curve represents titration specific model, solid curve represents averaged pKa’s
model.
80
(a)
0.8
Net Molality Protons Added (mM/g)
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0.0
-0.1
-0.2
2.00
4.00
6.00
8.00
10.00
pH
(b)
Net Molality Protons Added (mM/g)
0.50
0.40
0.30
0.20
0.10
0.00
-0.10
1
2
3
4
5
6
pH
81
7
8
9
10
11
100
4-Site Model
90
% Cd Adsorbed
80
70
3-Site Model
60
50
40
30
20
10
0
2
3
4
5
6
7
8
9
10
11
pH
Figure 4.3. Cd adsorption onto the bacterial consortia from the (+) river, (∆) soybean crop, and
(O) forest sites. Cd experiments were conducted with 10.5 ± 0.1 g bacteria/L (wet weight) and
10.31 ± 0.02 ppm Cd in a 0.1 M NaClO4 solution. Solid and dashed curves represent the 4- and
3-site models, respectively. The curves were developed from the average of the individual SCM
defined metal adsorption equilibrium constants.
82
Figure 4.4 Metal adsorption data for Ca, Cu, Pb, Sr, and Zn. (◊) = experimental data, and
solid curve depicts the best fit curve.
(a)
25
5.24 ppm Ca; 11.4 g/L biomass
% Ca adsorbed
20
15
10
5
0
2
3
4
5
6
7
pH
(b)
100
5.25 ppm Cu; 5.72 g/L biomass
90
% Cu Adsorbed
80
70
60
50
40
30
20
10
0
2
3
4
5
pH
83
6
7
Figure 4.4 (continued)
(c)
100
10.1 ppm Pb, 3.30 g/L biomass
90
% Pb Adsorbed
80
70
60
50
40
30
20
10
0
2
3
4
5
6
7
pH
(d)
50
3.19 ppm Sr; 20.1 g/L biomass
% Sr Adsorbed
40
30
20
10
0
2
3
4
5
pH
84
6
7
Figure 4.4 (continued)
(e)
70
10.3 ppm Zn; 11.3 g/L biomass
% Zn adsorbed
60
50
40
30
20
10
0
2
3
4
5
6
7
pH
studied as an average cell wall with a limited number of types of functional groups.
Clearly, if individual cell walls exhibit unique adsorption characteristics, it would be
impossible to successfully apply this type of simplified approach. Therefore, a major
objective of this modeling exercise is to determine if such an approach can successfully
account for the observed adsorption behaviors.
In the discrete site modeling approach, the deprotonation of each functional group
type is represented as a single deprotonation reaction (e.g., Borrok et al., 2004a; Fein et
al., 2005; Fein et al., 1997; Ngwenya et al., 2003). This SCM is likely a simplification of
the mechanisms involved and should not be taken as an exact representation of what is a
complex and heterogeneous chemical system. In our modeling approach, the surface
charge of the bacterial cell walls is due to deprotonation reactions according to the
following stoichiometry:
85
−
R − AiH0 ⇔ R − Ai + H+
(2)
where R represents the bacterial cell wall to which each functional group type, Ai, is
attached. Mass balance equations are used to quantify the distribution of protonated and
deprotonated functonal groups, in terms of the acidity constant for Site Ai, Ka:
−
Ka =
[R − A i ]a H +
[R − A i H 0 ]
(3)
where [R – Ai-] and [R – AiH0] represent the concentration of deprotonated and
protonated sites, respectively, in mol/L, and aH+ represents the activity of protons in the
bulk solution. The data are modeled using a non-electrostatic approach because all
experiments were conducted at the same ionic strength. Therefore, potential ionic
strength effects on the surface electric field could not be determined. Also, in order to
apply an electrostatic model to this system, the surface area of the bacteria of interest
would have to be determined. However, each consortium contained a variety of bacterial
species that changed throughout the sampling period. Therefore, the overall surface area
can not be calculated easily or with any certainty. Because the solvent contained protons,
the same species as the one that is reacting with the surface of interest, we define a zero
proton condition for the cell wall to account for the change in the proton concentrations
relative to that condition (Westall et al., 1995). The same approach was used by Fein et
al. (2005). The fully protonated cell wall was chosen to represent our zero proton
condition, using FITEQL 2.0 (Westall et al., 1982) to solve for the initial protonation
state of the cell walls in each titration.
The titration data were used not only to determine how many types of functional
groups must be invoked in order to account for the observed buffering capacities, but also
86
to constrain the functional group site concentrations and their associated proton binding
constants (Ka). Similar to what Borrok et al. (2004a) observed for titrations of bacterial
consortia, in each case a model that includes four different functional group types yields a
better fit to the titration data than do models with fewer site types. In each case, a fivesite model does not converge, indicating that parameters for five sites can not be
constrained by the available data, and that the system is over-determined with five sites.
The goodness of fit is determined by comparing the FITEQL calculated variance function
values, V(Y), for each data set, and the model that yields a V(Y) value closest to 1.0
represents the best fit. Figure 4.2b shows an example of a model developed using the four
site model and a fit to titration data (Sept. soy field sample consortia) that is
representative of the general shape of other titrations.
All titration modeling results are compiled in Table 4.1, the table shows averages
for all of the river water, forest soil, and soybean field soil samples separately. Calculated
values show no significant trend of pKa values (all values are consistent, within
experimental uncertainty) or site concentrations as a function of sampling site or as a
function of time of year – so clearly bacterial diversity does not affect proton uptake
significantly and a single set of averaged pKa values and site concentrations can be used
to account for the observed proton uptake by all of the consortia in this study. These
values that are averaged over entire dataset are also compiled in Table 4.1. The
functional group sites with pKa values of 3.15, 4.77, 6.54, and 9.18 are hereafter referred
to as sites A1 through A4, respectively (Table 4.1). Figure 4.2b shows the fit of the
averaged pKa values and surface site concentrations to a representative titration. The
averaged pKa values are similar to those calculated by Borrok et al. (2004a; 2004b) for
87
both the natural consortia from contaminated and uncontaminated environments. Site
densities for each of the four discrete sites are similar for all consortia and the average
total site concentrations (the sum of the average concentrations of binding sites for each
of the four surface sites) range from 3.7 × 10-4 to 3.9 × 10-4 mol of sites per gram of
consortium. Titration data alone cannot be used to determine the identity of the individual
functional group types present on the cell walls. However, it is interesting to note that the
buffering capacities of the large number of complex mixtures of bacterial species studied
here can be modeled well using the same set of averaged acidity constants and site
concentrations. It is possible that the commonality in these parameters represents a
commonality in binding site types among these bacterial species, but a rigorous testing of
this hypothesis must await further spectroscopic study.
The uncertainties determined for the averaged acidity constants reported in Table
4.2 represent a range of 1 σ standard deviation., and vary from 0.07 to 0.76 (on a log
scale). These values are of similar magnitude to those reported by previous studies for
single species of bacteria (e.g. Fein et al., 2005; Fein et al., 1997; Haas et al., 2001;
Martinez et al., 2002; Ngwenya et al., 2003). The uncertainties associated with the
averaged values from this study suggest that the degree of observed variability for the
range of consortia studied arises more from experimental uncertainties than from real
differences in the buffering capacities of the different consortia. The calculated surface
site concentrations from this study (Table 4.1) are significantly higher than those
determined by Borrok et al. (2004a) for similar bacterial consortia. These differences may
be due to a number of influences: Borrok et al. (2004a) froze some of their samples prior
to experimentation, their DGGE analyses demonstrated that on average their consortia
88
contained fewer species than were present in our consortia, and they cultured their
bacteria in different broth than was employed in this study.
In general, the similarity in titration curves among all of the consortia studied
here, and the associated low degree of uncertainty in the averaged acidity constants and
site concentrations, demonstrates that one set of averaged acidity constants and surface
site concentrations is adequate to reasonably describe the proton adsorption behavior of
all of the natural consortia in this study. These similarities imply a commonality of
binding sites among the various bacterial cell wall functional groups present in all of the
consortia in this study, and suggest that the averaged acidity constants and site
concentrations may be used to provide a reasonable model for proton adsorption,
buffering capacity, and surface charge for bacterial consortia in a wide range of natural
systems.
Following the approach of Borrok et al. (2004a), we use FITEQL 2.0, along with
the average acidity constant and site concentration values determined from the titration
models, to account for cation (M2+) adsorption onto the cell wall according to the
reaction:
M +2 + R − A ix -1 ⇔ R − A i H x (M) x +1
(4)
where R – AiHx(M)x+1 represents the metal-bacterial surface complex. The mass balance
equation for reaction 4 is:
K ads =
[R − AH x (M) x +1 ]
a M + 2 [R − AH xx -1 ]
(5)
where the brackets represent concentration in moles of sites L-1, a represents the activity
of the subscripted species, and Kads is the thermodynamic equilibrium constant for
89
TABLE 4.1
CALCULATED PROTON BINDING CONSTANTS (PKA) AND
SURFACE SITE CONCENTRATIONS
Proton Binding Constants (-log Ka)
Consortia
River
A2
A3
A4
A1
A2
A3
A4
Oct.
3.5
4.8
6.6
9.2
7.6
10.8
6.8
9.9
Nov.
3.8
5.1
6.8
9.2
6.7
8.2
6.1
11.0
3.7
5.1
7.0
9.2
6.7
8.1
7.0
11.2
3.7
5.1
7.2
9.5
6.2
7.6
10.0
14.2
Jan.
March
May
June
Sept.
Forest
A1
Site Concentrations
(x10-5 mol/g wet weight)
Oct.
Nov.
Jan.
March
May
June
Sept.
3.6
5.1
6.9
9.4
9.2
10.9
5.5
9.8
3.6
4.9
6.6
9.2
8.7
10.6
6.4
9.7
3.7
5.1
6.9
9.3
9.3
10.6
5.3
9.9
3.0
4.7
6.6
9.3
12.3
14.5
6.6
9.6
2.8
4.7
6.5
9.2
17.1
21.0
9.3
14.4
3.5
4.9
6.6
9.1
8.4
10.6
4.9
7.9
3.4
4.8
6.5
8.9
7.3
10.1
4.8
8.2
3.5
4.8
6.9
9.2
7.7
11.0
4.6
9.4
2.5
4.6
6.1
9.0
18.1
13.1
7.1
7.7
2.2
4.3
5.8
8.8
22.0
11.2
8.1
6.6
3.1
4.7
6.4
9.0
11.4
12.6
7.0
7.6
2.7
4.4
6.1
8.9
12.7
12.1
8.1
7.0
3.1
4.8
6.4
9.1
11.5
12.1
6.7
7.6
3.6
4.9
6.7
9.2
9.9
10.5
5.5
10.1
3.4
4.9
6.6
9.2
11.4
11.8
6.7
9.2
3.6
4.9
6.7
9.3
11.1
11.6
6.0
11.7
3.2
4.6
6.5
9.2
9.7
13.5
7.4
12.3
3.7
5.1
7.1
9.5
10.1
13.3
5.1
14.4
3.5
5.0
6.9
9.3
8.2
12.0
4.8
11.7
3.5
4.8
6.6
9.2
7.7
13.3
5.7
13.7
3.4
4.8
6.6
9.3
12.2
14.0
6.3
11.0
3.2
4.7
6.5
9.2
10.0
14.8
6.3
11.0
3.3
4.7
6.5
9.1
8.8
12.3
5.7
8.2
3.5
4.8
6.7
9.2
8.1
11.0
5.2
9.1
3.4
4.8
6.1
9.2
8.0
1.0
5.0
8.0
3.1
4.7
6.6
9.2
11.5
13.2
6.4
8.6
3.1
4.7
6.6
9.2
11.3
12.8
6.2
8.4
3.1
4.7
6.5
9.2
11.9
14.0
6.4
9.3
3.1
4.6
6.2
9.1
10.2
15.2
6.7
1.1
3.2
4.7
6.6
9.2
11.4
18.8
5.1
1.3
90
TABLE 4.1 (continued)
Soy
Oct.
Nov.
Jan.
March
May
June
Sept.
3.5
4.9
6.8
9.3
8.8
13.4
4.7
11.2
3.4
4.9
6.7
9.1
9.6
10.8
6.2
8.4
3.6
5.0
6.8
9.2
8.6
9.3
5.5
8.4
3.5
5.0
6.7
9.3
8.5
9.4
5.8
9.3
3.4
4.9
6.8
9.3
10.2
13.0
5.6
8.3
3.5
4.9
6.7
9.2
10.3
13.9
6.1
9.2
3.6
5.0
6.7
9.3
9.7
12.1
5.5
11.6
3.2
4.8
6.7
9.3
11.3
13.5
5.7
7.7
3.3
4.8
6.7
9.3
10.7
13.9
5.4
9.5
3.4
4.8
6.7
9.3
9.7
13.2
5.4
9.9
3.4
4.9
6.7
9.2
10.1
13.6
5.5
9.3
3.4
4.8
6.7
9.1
8.8
12.8
5.2
8.1
3.5
4.9
6.7
9.3
9.0
11.8
5.4
9.4
2.9
4.7
6.5
9.3
13.2
13.3
6.7
8.0
2.7
4.6
6.3
8.9
13.7
12.4
6.9
6.3
2.9
4.7
6.4
9.1
11.7
11.5
6.2
6.0
3.2
4.8
6.6
9.2
11.1
12.0
5.5
8.0
3.2
4.7
6.6
9.3
10.2
13.2
5.8
9.8
3.3
4.8
6.6
9.3
9.8
12.4
6.0
11.8
2.97
3.3
3.23
3.15
4.73
4.77
4.81
4.77
6.43
6.59
6.64
6.54
9.18
9.21
9.22
9.18
10.8
10.0
10.3
10.4
11.5
13.1
12.4
12.4
6.7
5.9
5.7
6.1
9.5
10.6
9.0
9.7
+0.2/-0.38
+0.16/-0.27
+0.26/-0.76
+0.07/-0.25
± 2.78
± 2.28
± 1.07
± 2.04
Averages a
River
Forest
Crop
Overall b
Stdev
c,d
a
The binding constants reported are an average calculated from the individual (non-log) equilibrium
constants for the specific site for the consortia of interest.
b
Overall averages are an average calculated from the individual (non-log) equilibrium constants. These
values are utilized for metal adsorption models.
c
Because of outlier values, the standard deviation is larger than the average and can not be reported in log
form. Therefore, we neglect the K values greater than 1.5 × 10-3 in calculating the standard deviation for
A1.
d
Reported uncertainties were calculated using the following formulas: + uncertainty = log (Ave + stdev) log (Ave), - uncertainty = log (Ave) - log (Ave - stdev) where Ave represents the average, site-specific
equilibrium constant calculated from all experimental data. Stdev represents the standard deviation of the
average.
91
TABLE 4.2
CD BINDING CONSTANTS (LOG K) FOR BEST-FIT
ADSORPTION MODELS
Consortia
River
Forest
Soybean
Averages
Stdev
b
Cd Binding Constants (Log K)
1
2
3
A
A
A
4
A
Oct.
Nov.
3.1
3.0
3.4
2.9
2.8
6.0
Jan.
March
May
June
Sept.
Oct.
Nov.
Jan.
March
May
June
Sept.
Oct.
Nov.
Jan.
March
May
June
Sept.
3.4
4.2
4.3
4.2
4.0
4.1
3.9
4.3
4.5
4.2
4.2
4.5
3.6
4.7
4.1
3.5
4.3
4.3
4.6
4.2
5.8
2.5
2.9
2.9
3.3
3.2
3.4
2.7
3.1
3.0
2.9
3.5
3.1
2.9
3.0
3.1
3.2
2.8
3.0
3.6
3.0
2.9
3.0
3.4
3.4
3.4
3.3
3.5
3.5
3.6
3.3
3.2
3.6
3.4
3.4
3.3
3.6
3
3.1
3.2
3.1
3.2
3.4
3.5
3.4
4.1
4.4
4.3
4.3
5.8
5.6
5.9
5.8
+0.20/-0.38
+0.18/-0.29
+0.23/-0.50
+0.28/-0.94
5.0
6.3
5.5
5.5
5.8
5.6
5.8
4.7
5.6
a
Water
Corn
Forest
Overall
a
The metal stability constants reported represent an average calculated from the individual (non-log)
equilibrium constants for the specific site and consortia of interest.
b
Uncertainties were calculated using the following formulas: + uncertainty = log (Ave + stdev) - log (Ave),
- uncertainty = log (Ave) - log (Ave - stdev) where Ave represents the average, site-specific metal stability
constant calculated from all experimental data. Stdev represents the standard deviation of the average.
reaction 5. A 1:1 metal:surface site stoichiometry was used in all models. These
calculations account for aqueous metal hydrolysis reactions using equilibrium constants
from Baes and Mesmer (1976), and they use the water dissociation constant from Wolery
92
(1992). Metal adsorption measurements conducted as a function of pH were used to
determine the minimum number of binding sights and the best fit metal-bacterial
adsorption stability constants (Table 4.2).
The number of sites required to adequately describe cation adsorption varies,
depending mostly on the pH range of the data and the extent of metal adsorption
observed for a specific metal. For the Cd experiments, we only considered models in
which the metal adsorbs onto sites with sequential pKa values (i.e., adsorption to sites A1
and A2 or onto A1, A2, and A3). In cases in which the pH range of the data was limited,
only a 2 or 3-site model was required to fit the experimental data. In all cases, when
adsorption data were collected at pH values higher than 8, a fourth site was required to
account for the adsorption at high pH. The best-fit model for each consortium and the
average stability constant value for each site are tabulated in Table 4.2. The metal
stability constant for A4 could be better constrained with more experiments at high pH
with varying metal:bacteria ratios. Other than for the A4 site, the uncertainties for the
averaged metal-bacterial stability constants are of similar magnitude to those reported by
previous studies for single species of bacteria (e.g., Fowle and Fein, 1999; Haas et al.,
2001; Yee and Fein, 2001; Ngwenya et al., 2003), providing evidence that the variability
in metal adsorption behavior is largely due to experimental uncertainty rather than to real
differences in the adsorption capacities of individual bacterial species.
An averaged four-site adsorption model, constructed using the averaged acidity
constants, site concentrations, and the averaged Cd-consortia stability constants given in
Tables 4.1 and 4.2, yields a reasonable fit to the observed Cd adsorption behavior (Figure
4.3). The model fit lies within 20% of the observed adsorption percentages in all cases,
93
and for much of the pH range the fit is within 10% of the extremes in adsorption that we
observed. At high pH (8.5 and above), the averaged model over-predicts the extent of
adsorption by up to 5-10%, suggesting that the Cd-A4 binding environment is not welldescribed by the model. The best-fitting three-site model (also depicted in Figure 4.3)
yields a bigger misfit to the observed adsorption under high pH conditions, clear evidence
that a fourth binding site must be involved in the uptake of Cd at these pH values. There
is considerably more uncertainty associated with the model fit to the Cd adsorption data
for the consortia studied here than is typically associated with model fits to metal
adsorption onto single bacterial species. However, the level of uncertainty is acceptably
small for field applications of this modeling approach where it is likely that the
uncertainties associated with using these averaged parameters to model Cd adsorption are
relatively small compared to the uncertainties in determining bacterial concentrations in
realistic settings.
A similar modeling approach was applied to account for the adsorption of the
other metals studied here onto the river water consortia, and the results are compiled in
Table 4.3 and depicted in Figure 4.4. The number of surface sites required to account for
the adsorption of Ca, Cu, Pb, Sr, and Zn depends on the pH range of the adsorption data
as well as on the extent of adsorption for each metal. In general, the best-fit models
account for the data well, with the Pb and Zn models displaying the best fits. While the
stability constants obtained for Ca, Pb, Sr, and Zn were similar to those obtained
previously for single species of bacteria (Fowle and Fein, 1999; Fein et al., 2001;
Ngwenya et al., 2003; Yee and Fein, 2003; Borrok and Fein, 2005), we were unable to
find similar comparisons for Cu.
94
TABLE 4.3
METAL BINDING CONSTANTS (LOG K) FOR BEST-FIT
ADSORPTION MODELS
Metal
Ca
Cu
Pb
Sr
Zn
Metal Binding Constants (Log K)
A1
A2
A3
1.8
2.3
2.9
3.8
3.9
5.5
7.1
2.5
1.9
3.1
2.6
2.8
3.5
A4
5.7
The Cd experiments rigorously tested the hypothesis of universal binding
mechanisms using a wide range of bacterial consortia. Given the similarities in binding
that we observed, the primary purpose for conducting the Ca, Cu, Pb, Sr, and Zn
experiments was not to repeat tests on a diverse collection of consortia, but to calibrate a
linear free-energy approach (Langmuir, 1979) for estimating metal-consortia stability
constants for metals not studied here. In this approach, we relate the stability constants
for the metal-bacterial surface complexes that are calculated in this study to the stability
constants for acetate complexes involving these same metals. This approach yields good
correlations for single bacterial species, such as Bacillus subtilis (Fein et al., 1997; Fein
et al., 2001). Other than the Pb data which were fit with a Site A3-only model, all model
fits included metal binding onto the A2 site (4.77 pKa) and the A3 site (6.54 pKa). We
relate the metal-A2 and metal-A3 stability constants to the corresponding metal-acetate
stability constants from Shock and Koretsky (1993), with the results depicted in Figure
4.5. There is a reasonable relationship in each case, with metal-bacterial stability
constants increasing in general with increasing metal-acetate stability constant value. We
95
fit the data with linear relationships, with correlation coefficients of 0.95 and 0.92,
respectively, for the Site A2 and Site A3 relationships, respectively. These correlations
likely result from similarities in binding environments between acetate and the A2 and A3
sites on the bacterial consortia. The equations of the best-fitting lines shown in Figure
4.5a and 4.5b are
Y = (1.37) X + 0.70
(5)
Y = (2.53) X + 0.05
(6)
where X and Y represent the logarithm of the stability constant values for the metalacetate and metal-A2 and metal-A3 complexes respectively. These relationships can be
used to estimate metal-A2 and metal-A3 stability constants for bacterial consortia, if the
stability constant for the corresponding metal-acetate complex is known. Therefore, these
types of relationships are important for extrapolating these experimental and modeling
results to estimate the extent of adsorption that would occur between bacterial consortia
and metals not studied here.
4.4
Conclusions
In this study, we used adsorption experiments onto bacterial consortia grown from
a range of environments collected over the course of a year to extensively test if
commonalities exist in the adsorption behavior of protons and Cd. We observed nearly
identical proton adsorption behavior and similar Cd adsorption behavior for all of the
consortia tested, and we quantify the acidity constants, site concentrations, and stability
constants for the important metal-bacterial surface complexes. The uncertainties
associated with the averaged acidity constants, site concentrations, and stability constant
96
Figure 4.5 Correlation plots showing metal adsorption constants calculated from natural
consortia and corresponding metal-acetate stability constants from Shock and Koretsky
(1993). Linear correlation coefficients are shown for each data set.
97
Site 2 Consortia Log K Values
(a)
4.5
cc = 0.95
4
2+
Cu
3.5
3
Ca
2+
Cd
2.5
2+
2+
Zn
2
1.5
2+
Sr
1
0.5
0
0.5
1
1.5
2
2.5
Acetate Log K Values
Site 3 Consortia Log K Values
(b)
8
0.92
7
Pb
2+
6
5
Ca
4
2+
2+
Sr
2+
Cu
Cd
3
2+
2+
Zn
2
1
0
0.5
1
1.5
2
Acetate Log K Values
98
2.5
3
values from all of the consortia datasets are similar in magnitude to those associated with
acidity constants, site concentrations, and stability constant values from adsorption
experiments that involved single bacterial species. Therefore, the variability that is seen
in the adsorption behaviors in this study likely results primarily from experimental
uncertainty and not from real trends in binding behavior as a function of bacterial
diversity. The consortia that we used in the adsorption experiments represent a wide
range in bacterial diversity. Therefore, although there clearly are some bacteria in nature
that exhibit different adsorption behaviors (e.g., Borrok et al., 2004b), it is likely that the
common adsorption behavior documented here holds for an even wider range of bacteria
than was tested in this study.
The Cd experiments demonstrate the universality of metal adsorption behavior
onto a wide range of bacterial consortia. Based on these results, we measured the
adsorption of Ca, Cu, Pb, Sr, and Zn onto individual consortia in order to calibrate a
linear free energy approach that enables predictions of the adsorption behavior of other
metals onto bacterial consortia. Therefore, this study yields a set of averaged acidity
constants, surface site concentrations, and metal stability constants that can be used to
estimate metal speciation and adsorption in realistic bacteria-bearing systems. The
usefulness of this averaged model depends on the particular application of interest. In
engineered systems, it may be useful and possible to determine the extent of bacterial
metal adsorption with greater precision than this averaged model provides. However, in
complex geologic systems, it is impossible to study the adsorption properties of each
metal and each bacterial species of interest. Therefore, this generalized model, with its
99
slightly higher level of uncertainty, offers a useful approach for quantifying the fate and
transport of metals.
100
CHAPTER 6
CONCLUSIONS
The research presented in this dissertation employs laboratory experiments and
modeling techniques to provide insight into the metal adsorption behavior of bacteria. To
better understand metal adsorption to the cell wall at the atomic level, a molecular
simulations approach was employed to examine metal adsorption to the major metal
binding components of the cell wall structure. Our results indicate that simulations
techniques can successfully describe Cd adsorption to the cell wall as the metal
coordinations and binding distances that were modeled are very similar to those
determined by X-ray adsorption spectroscopy. The inability of simulations to produce a
Pb-ligand geometry similar to XAFS results is possibly due to limitations in the classical
mechanics-based models to simulate the covalent binding behavior of Pb, or more likely
the ability of Pb to form hydroxide phases at circumneutral pH. Future research
associated with molecular simulations of metal-bacteria interactions includes the
development and refinement of force field parameters, development of larger
representative cell wall models, analysis of multiple metal adsorption, and competitive
adsorption processes.
Many studies of metal-bacteria interaction have used non-metabolizing bacteria to
focus on passive binding. Our results indicate that bacterial metabolic processes have a
101
significant impact on Cd adsorption to the cell wall functional groups. Similar to Urrutia
Mera et al., (1992), we found that metabolizing Gram-positive cells adsorbed
significantly less Cd than non-metabolizing Gram-positive cells. However, we did not
observe any difference in metal uptake for metabolizing and non-metabolizing Gramnegative cells. Our results suggest that the differences in metal adsorption for Grampositive cells is likely due to a decrease in pH in the local cell wall region because of the
proton motive force of metabolically-active cells. This study also applied surface
complexation modeling to determine the change in pH around in the local cell
environment that would result in the magnitude of decrease in metal adsorption on the
Gram-positive cells. Further research to investigate the localized changes in near surface
pH of bacteria would be useful. Also, studies involving other bacterial species with
varying cell wall structures and varying the parameters (i.e. bacteria:metal ratio, pH, etc.)
would help constrain the impact of metabolic processes on metal adsorption to the
bacterial cell wall.
The results of adsorption experiments onto bacteria consortia, collected repeatedly
over the course of a year to ensure changing bacterial diversity, suggest that
commonalities exist in the proton and Cd adsorption behavior. The extensive titration and
adsorption data were modeled to determine acidity constants, surface site concentrations,
and metal stability constants. Because the uncertainties associated with these values were
comparable to those calculated for metal adsorption to single bacterial species, we were
able to ascertain that the variability is likely due to experimental uncertainty and not the
binding behavior of individual species in the consortia. Therefore these values can be
used to estimate the metal-bacterial adsorption component of surface complexation
102
models of natural systems. A future direction for this research is to perform similar
experiments with different cations, altering the metal:bacteria ratios to better constrain
the metal stability constants.
The research described in this dissertation elucidates the mechanisms of metal
adsorption onto bacteria – examining the specifics and looking for generalities between
species. Furthermore, a better understanding of the influence of bacterial processes, such
as metabolism, on bacterial metal adsorption is important in order to developing accurate
and quantitative geochemical models that increase the effectiveness of remediation
strategies. Awareness of the mechanisms of bacterial adsorptive processes is an important
step towards being able to quantify metal transport in the environment.
103
REFERENCES
Allen M.P. and Tildesley D.J. (1987) Computer Simulation of Liquids. 385 p. Oxford
University Press, Oxford.
Åqvist J. (1990) Ion-water interaction potentials derived from free energy perturbation
simulations. Journal of Physical Chemistry 94, 8021-8024.
Araki Y. and Ito E. (1989) Linkage units in cell walls of gram-positive bacteria. Critical
Reviews in Microbiology 17, 121-135.
Baes C.F. and Mesmer R.E. (1976) The Hydrolysis of Cations. Wiley.
Bandyophadhyay S., Shelley J.C., and Klein M.L. (2001) Molecular dynamics study of
the effect of surfactant on a biomembrane. Journal of Physical Chemistry B 105,
5979-5986.
Bethke C.M. and Brady P.V. (2000) How the Kd approach undermines ground water
cleanup. Ground Water 38(3) 435-443.
Beveridge T.J. (1989) Role of cellular design in bacterial metal accumulation and
mineralization. Annual Reviews in Microbiology 43, 147-171.
Beveridge T.J. (1999) Structures of gram-negative cell walls and their derived membrane
vesicles. Journal of Bacteriology 181, 4725-4733.
Beveridge T.J., Fyfe W.S. (1985) Metal fixation by bacterial-cell walls. Canadian
Journal of Earth Science 22, 1893-1898.
Beveridge T.J. and Koval S.F. (1981) Binding of metals to cell envelopes of Escherichiacoli-K-12. Applied and Environmental Microbiology 42, 325-335.
Beveridge T.J. and Murray R.G.E. (1976) Uptake and retention of metals by cell walls of
Bacillus subtilis. Journal of Bacteriology 127, 1502-1518.
Beveridge T.J. and Murray R.G.E. (1980) Sites of metal deposition in the cell wall of
Bacillus subtilis. Journal of Bacteriology 141, 876-887.
104
Borrok D.M. and Fein J.B. (2005) The impact of ionic strength on the adsorption of
protons, Pb, Cd, and Sr onto the surfaces of Gram negative bacteria: testing nonelectrostatic, diffuse, and triple-layer models. Journal of Colloid and Interfacial
Science 286, 110-126.
Borrok D., Fein J.B., and Kulpa C.F. (2004a) Proton and Cd sorption onto natural
bacteria consortia: Testing universal adsorption behavior. Geochimica et
Cosmochimica Acta 68, 3231-3238.
Borrok, D., Fein, J.B. and Kulpa, C.F. (2004b) Cd and Proton Adsorption onto Bacterial
Consortia Grown from Industrial Wastes and Contaminated Geologic Settings.
Environmental Science and Technology 38, 5656-5664.
Borrok D.B., Fein J.B., Tischler M., O’Loughlin E., Meyer H., Liss M. and Kemner K.M.
(2004c) The effect of acidic solutions and growth conditions on the adsorptive
properties of bacterial surfaces. Chemical Geology 209, 107-119.
Bovill R.A., Shallcross J.A., Mackey B.M. (1994) Comparison of the fluorescent redox
dye 5-cyano-2,3-ditolyltetrazolium chloride with p-iodonitrotetrazolium violet to
detect metabolic activity in heat-stressed Listeria monocytogenes cells. Journal of
Applied Bacteriology 77, 353-358.
Boyanov M.I., Kelly S.D., Kemner K.M., Bunker B.A., Fein J.B., Fowle D.A. (2003a)
Adsorption of cadmium to Bacillus subtilis bacterial cell walls: A pH-dependent
X-ray adsorption fine structure spectroscopy study. Geochimica et Cosmochimica
Acta 67, 3299-3311.
Boyanov M.I., Kmetko J., Shibata T., Datta A., Dutta P., Bunker B.A. (2003b)
Mechanism of Pb adsorption to fatty acid Langmuir monolayers studied by X-ray
adsorption fine structure spectroscopy. Journal of Physical Chemistry B 107,
9780-9788.
Caminiti R. (1982) Nickel and cadmium phosphates in aqueous solution-cation-anion
complex formation and phosphate-H2O interaction. Journal of Chemical Physics
77, 5682-5686.
Caminiti R. and Johansson G. (1981) On the structures of cadmium sulfate complexes in
aqueous solutions. Acta Chemica Scandinavica Series A 35, 373-381.
Caminiti R., Cucca P. and Radnai T. (1984) Investigation on the structure of cadmium
nitrate aqueous solutions by X-ray diffraction and Raman Spectroscopy. Journal
of Physical Chemistry 88, 2382-2386.
Claessens J., van Lith Y., Laverman A.M. and Van Cappellen P. (2006) Acid-base
activity of live bacteria: Implications for quantifying cell wall charge.
Geochimica et Cosmochimica Acta 70, 267-276.
105
Cox J.S., Smith D.S., Warren L.A., and Ferris F.G. (1999) Characterizing heterogeneous
bacterial surface functional groups using discrete affinity spectra for proton
binding. Environmental Science and Technology 33, 4515-4521.
Crist R.H., Oberholser K., Shank N. and Nguyen M. (1981) Nature of bonding between
metallic ions and algal cell walls. Environmental Science and Technology 15,
1212-1217.
Cygan R.T. (2001) Molecular modeling in mineralogy and geochemistry. In R.T. Cygan,
and J.D. Kubicki, Eds. Reviews in Mineralogy and Geochemistry: Molecular
Modeling Theory: Applications in the Geosciences, p. 1-35. The Geochemical
Society, Washington D.C.
Dauber-Osguthorpe P., Roberts V.A., Osguthorpe D.J., Wolff J., Genest M., and Hagler
A.T. (1988) Structure and energetics of ligand-binding to proteins: Escherichiacoli dihydrofolate reductase trimethoprim, a drug-receptor system. Proteins 4, 3147.
Daughney C.J., Fein J.B. and Yee N. (1998) A comparison of the thermodynamics of
metal adsorption onto two common bacteria. Chemical Geology 144, 161-176.
Daughney, C.J., Fowle, D.A. and Fortin, D.E. (2001) The effect of growth phase on
proton and metal adsorption by Bacillus subtilis. Geochimica et Cosmochimica
Acta 65, 1025-1035.
Davis, J.A. and Ken, D.B. (1990) Surface complexation modeling in aqueous
geochemistry. Mineral-water interface geochemistry. In M.F. Hochella Jr., and
A.F. White, Eds. Reviews in Mineralogy: Mineral-water interface geochemistry,
23, p. 177-260. Mineralogical Society of America, Washington, D.C.
Delley B. (1990) An all-electron numerical method for solving the local density
functional for polyatomic molecules. Journal of Chemical Physics 92, 508-517.
Delley B. (2000) From molecules to solids with the DMol3 approach. Journal of
Chemical Physics 113, 7756-7764.
Ehrlich H.L. (1996) Geomicrobiology. Third Edition. Marcel Dekker, Inc. New York.
Elwood D.C. and Tempest D.W. (1969) Control of teichoic acid and teichuronic acid
biosynthesis in chemostat cultures of Bacillus subtilis var. niger. Biochemical
Journal 111, 1-5.
Fein J.B. (2000) Quantifying the effects of bacteria on adsorption reactions in water-rock
systems. Chemical Geology 169, 265-280.
106
Fein J.B., Boily J.F., Yee N., Gorman-Lewis D. and Turner B.F. (2005) Potentiometric
titrations of Bacillus subtilis cells to low pH and a comparison of modeling
approaches. Geochimica et Cosmochimica Acta 69, 1123-1132.
Fein J.B., Daughney C.J., Yee N. and Davis T.A. (1997) A chemical equilibrium model
for metal adsorption onto bacterial surfaces. Geochimica et Cosmochimica Acta
61, 3319-3328.
Fein, J.B., Martin, A.M. and Wightman, P.G. (2001) Metal adsorption onto bacterial
surfaces: development of a predictive approach. Geochimica et Cosmochimica
Acta 65, 4267-4273.
Fortin, D., Ferris, F.G. and Beveridge, T.J. (1997) Surface-mediated mineral
development by bacteria. In Banfield, J.F., Nealson, K.H. (Eds.),
Geomicrobiology: Interactions Between Microbes and Minerals, Reviews in
Mineralogy, 161-180. Mineralogy Society of America, Washington, D.C.
Fowle D.A. and Fein J.B. (1999) Competitive adsorption of metal cations onto two gram
positive bacteria: Testing the chemical equilibrium model. Geochimica et
Cosmochimica Acta 63, 3059-3067.
Francis A.J., Billow J.B., Dodge C.J., Harris R., Beveridge T.J., and Papenguth H.W.
(2004) Uranium association with halophilic and non-halophilic bacteria and
archaea. Radiochimica Acta 92, 481-488.
Franks F. (1973) Ed. Water: A Comprehensive Treatise. V.3. Aqueous Solutions of
simple electrolytes. Plenum Press, New York.
Galy J., Meunier G., Andersson S., and Astrom A. (1975) Stereochemistry of elements
which are unbound pairs – GE (II), AS (III), SE (IV), BR (V), SN (II), SB (III),
TE (IV), I (V), XE (VI), TL (I), PB (II), AND BI (III) (OXIDES, FLUORIDES
AND OXYFLUORIDES). Journal of Solid State Chemistry 13, 142-159.
Goncalves M.L.S., Sigg L., Reutlinger M. and Stumm W. (1987) Metal ion binding by
biological surfaces: Voltammetric assessment in the presence of bacteria. Science
of the Total Environment 60, 105-119.
Graham L.L., and Beveridge T.J. (1994) Structural differentiation of the Bacillus subtilis
168 cell wall. Journal of Bacteriology 176, 1413-1421.
Haas J.R., Dichristina T.J. and Wade Jr. R. (2001) Thermodynamics of U(VI) sorption
onto Shewanella putrefaciens. Chemical Geology 180, 33-54.
Harold F.M. (1972) Conservation and transformation of energy by bacterial membranes.
Bacteriological Reviews 36, 172-230.
107
Harvey R.W., Leckie J.O. (1985) Sorption of lead onto two Gram-negative marine
bacteria in seawater. Marine Chemistry 15, 333-344.
Hoover W.G. (1985) Canonical dynamics: Equilibrium phase space distributions.
Physical Review A 31 1695-1697.
Jiang W., Saxena A., Song B., Ward B.B., Beveridge T.J., and Myneni S.C.B. (2004)
Elucidations of functional groups on gram-positive and gram-negative bacterial
surfaces using infrared spectroscopy. Langmuir 20, 11433-11442.
Kaeberlein, T., Lewis, K. and Epstein, S.S. (2002). Isolating “uncultivable”
microorganisms in pure culture in a simulated natural environment. Science 296,
1127-1129.
Kastowsky M., Gutberlet T. and Bradaczek H. (1992) Molecular modeling of the threedimensional structure and conformational flexibility of bacterial
lipopolysaccharide. Journal of Bacteriology 174, 4798-4806
Kelly S.D., Boyanov M.I., Bunker B.A., Fein J.B., Fowle D.A., Yee N., and Kemner
K.M. (2001) XAFS determination of the bacterial cell wall functional groups
responsible for complexation of Cd and U as a function of pH. Journal of
Synchrotron Radiation 8, 946-948.
Kemper M.A., Urrutia M.M., Beveridge T.J., Koch A.L. and Doyle R.J. (1993) Proton
motive force may regulate cell wall-associated enzymes of Bacillus subtilis.
Journal of Bacteriology, 175, 5690-5696.
Konhauser K.O., Fyfe W.S., Ferris F.G. and Beveridge T.J. (1993) Metal sorption and
mineral precipitation by bacteria in two Amazonian river systems: Rio Solimoes
and Rio Negro, Brazil. Geology 21, 1103-1106.
Koretsky C. (2000) The significance of surface complexation reactions in hydrologic
systems: a geochemist's perspective. Journal of Hydrology 230(3-4), 121-171.
Kotra L.P., Golemi D., Amro N.A., Liu G.Y., and Mobashery S. (1999) Dynamics of the
lipopolysaccharide assembly on the surface of Escherichia coli. Journal of the
American Chemical Society 38, 8707-8711.
Kulczycki, E., Ferris, F.G. and Fortin, D. (2002) Impact of cell wall structure on the
behavior of bacterial cells as sorbents of cadmium and lead. Geomicrobiology
Journal 19, 553-565.
Langmuir D. (1979) Techniques of estimating thermodynamic properties for some
aqueous complexes of geochemical interest. In Jenne, E.A. (Ed.), Chemical
Modeling in Aqueous Systems, 357-387. American Chemical Society.
108
Ledin M., Krantz-Rülcker C. and Allard B. (1996) Zn, Cd and Hg accumulations by
microorganisms. Organic and inorganic soil components in multi-compartment
systems. Soil Biology and Biochemistry 28, 791-799.
Lee C., Hedges J.I., Wakeham S.G. and Zhu N. (1992) Effectiveness of various
treatments in retarding microbial activity in sediment trap material and their
effects on the collection of swimmers. Limnology and Oceanography, 37, 117130.
Lins R.D. and Straatsma T.P. (2001) Computer simulation of the rough
lipopolysaccharide membrane of Pseudomonas aeruginosa. Biophysical Journal
81, 1037-1046.
Marcus Y. (1988) Ionic radii in aqueous solutions. Chemical Reviews 88, 1475-1498.
Martinez R.E. and Ferris F.G. (2001) Chemical equilibrium modeling: Techniques for the
analysis of high-resolution bacterial metal sorption data. Journal of Colloid and
Interface Science 243, 73-80.
McCarthy J.F. and Zachara J.M. (1989) Subsurface transport of contaminants.
Environmental Science and Technology 23, 496-502.
Mullen M.D., Wolf D.C. Ferris F.G. Beveridge T.J. Flemming C.A. and Baily G.W.
(1989) Bacterial sorption of heavy metals. Applied and Environmental
Microbiology 55, 3143-3149.
Muyzer, G., De Waal, E.C. and Uitterlinden, A.G. (1993). Profiling of complex
microbial populations by denaturing gradient gel electrophoresis analysis of
polymerase chain reaction-amplified genes coding for 16S rRNA. Applied and
Environmental Microbiology 59, 695-700.
Navarre W.W. and Schneewind O. (1999) Surface proteins of gram-positive bacteria and
mechanisms of their targeting to the cell wall envelope. Microbiology and
Molecular Biology Reviews 63, 174-229.
Ngwenya B.T., Sutherland I.W., and Kennedy L. (2003) Comparison of the acid-base
behaviour and metal adsorption characteristics of a gram-negative bacterium and
other strains. Applied Geochemistry 18, 527-538.
Obst S., Kastowsky M., and Bradaczek H. (1997) Molecular dynamics simulations of six
different fully hydrated monomeric conformers of Escherichia coli relipopolysaccharide in the presence and absence of Ca2+. Biophysical Journal 72,
1031-1046.
109
Ohtaki H., Maeda M., and Ito S. (1974) X-ray diffraction studies of aqueous solutions of
cadmium perchorlate and sodium tetraiodocadmate. Bulletin of the Chemical
Society of Japan 47, 2217-2221.
Ohtaki H. and Radnai T. (1993) Structure and dynamics of hydrated ions. Chemical
Reviews 93, 1157-1204.
Pagnanelli F., Papini M.P., Toro L., Trifoni M.and Veglio F. (2000) Biosorption of metal
ions on Arthrobacter sp.: Biomass characterization and biosorption modeling.
Environmental Science and Technology 34, 2773-2778.
Palmer B.J., Pfund D.M., and Fulton J.L. (1996) Direct modeling of EXAFS spectra from
molecular dynamics simulations. Physical Chemistry 100, 13393-13398.
Parrinello M. and Rahman A. (1981) Polymorphic transitions in single crystals: A new
molecular dynamics method. Applied Physics 52, 7182-7190.
Perdew J.P. and Wang Y. (1992) Accurate and simple analytic representation of the
electron-gas correlation energy. Physical Review B: Condensed Matter. 45,
13244-13249.
Plette A.C.C., Benedetti M.F., Van Reimsdijk W.H. and Van der Wal A. (1995) pH
dependent charging behavior of isolated cell walls of a Gram-positive soil
bacterium. Journal of Colloid and Interface Science 171, 354-363.
Plette A.C.C., Benedetti M.F. and Van Reimsdijk W.H. (1996) Competitive binding of
protons, calcium, cadmium, and zinc to isolated cell walls of a Gram-positive soil
bacterium. Environmental Science and Technology 30, 1902-1910.
Sambrook J., Fritsch E.F., Maniatis T. (1989) Molecular cloning, a laboratory manual, 2nd
ed: New York, Cold Spring Harbor Laboratory Press, 3 vols.
Sarret G., Manceau A., Spadini L., Roux J.C., Hazemann J.L., Soldo Y., Eybert-Berard
L. and Menthonnex J.J. (1998) Structural determination of Zn and Pb binding
sites in Penicillium chrysogenum cell walls by EXAFS spectroscopy.
Environmental Science and Technology 32, 1648-1655.
Shelley J.C., Shelley M.Y., Reeder R.C., Bandyopadhyay S., and Klein, M.L. (2001a) A
coarse grain model for phospholipid simulations. Journal of Physical Chemistry
B. 105, 4464-4470.
Shelley J.C., Shelley M.Y., Reeder R.C., Bandyopadhyay S., Moore P.B. and Klein M.L.
(2001b) Simulations of phospholipids using a coarse grain model. Journal of
Physical Chemistry B 105, 9785-9792.
110
Shock, E.L. and Koretsky, C.M. (1993) Metal-organic complexes in geochemical
processes: Calculation of standard partial molal thermodynamic properties of
aqueous complexes at high pressures and temperatures. Geochimica et
Cosmochimica Acta 57, 4899-4922.
Shroll R.M. and Straatsma T.P. (2003) Molecular basis for microbial adhesion to
geochemical surfaces: computer simulation of Pseudomonas aeruginosa adhesion
to goethite. Biophysical Journal 84, 1765-1772.
Small, T.D., Warren, L.A., Roden, E.E. and Ferris, F.G. (1999) Sorption of strontium by
bacteria, Fe(III) oxide, and bacteria-Fe(II) oxide composites. Environmental
Science and Technology 33, 4465-4470.
Sposito G. (1984) The Surface Chemistry of Soils. Oxford University Press.
Templeton A.S., Trainor T.P., Spormann A.M., Newville M., Sutton S.R., Dohnalkova
A., Gorby, Y., and Brown G.E. (2003) Sorption versus biomineralization of
Pb(II) within Burkholderia cepacia biofilms. Environmental Science and
Technology 37, 300-307.
Tosi M.P. (1964) Cohesion of ionic solids in the Born model. Solid State Physics 131,
533-545.
Urrutia Mera M., Kemper M., Doyle R. and Beveridge T.J. (1992) The membraneinduced proton motive force influences the metal binding ability of Bacillus
subtilis cell walls. Applied and Environmental Microbiology 58, 3837-3844.
van der Waals J.H. and Platteeuw J.C. (1959) Clathrate solutions. Advance in Chemical
Physics 2, 1.
Wang Y. and Hollingsworth R.I. (1996) An NMR spectroscopy and molecular mechanics
study of the molecular basis for the supramolecular structure of
lipopolysaccharides. Biochemical Journal 35, 5647-5654.
Warren L.A. and Ferris F.G. (1998) Continuum between sorption and precipitation of
Fe(III) on microbial surfaces. Environmental Science and Technology 32, 23312337.
Westall, J.C., Jones, J.D., Turner, G.D. and Zachara, J.M. (1995). Models for association
of metal ions with heterogeneous environmental sorbents. 1. Complexation of
Co(II) by leonardite humic acid as a function of pH and NaClO4 concentration.
Environmental Science and Technology 29, 951-959.
Westall, J.C. (1982) FITEQL, a computer program for determination for chemical
equilibrium constants from experimental data. Version 2.0. Oregon St.
University, Corvallis, OR.
111
Wolery, T.J. (1992) EQ4NR, A computer program for geochemical aqueous speciationsolubility calculations: Theoretical manual, user’s guide, and related
documentation (Version 7.0). UCRL-MA-110662-PT-III. Lawrence Livermore
Natl. Lab., Livermore, CA, USA.
Xia K., Bleam W., and Hemke P.A. (1997) Studies of the nature of Cu2+ and Pb2+ binding
sites in soil humic substances using X-ray adsorption spectroscopy. Geochimica
et Cosmochimica Acta 61, 2211-2221.
Yee N. and Fein J.B. (2001) Cd adsorption onto bacterial surfaces: A universal
adsorption edge? Geochimica et Cosmochimica Acta 65, 2037,-2042.
Yee N., Fein J.B. and Daughney C.J. (2000) Experimental study of the pH, ionic strength,
and reversibility behavior of bacteria-mineral adsorption. Geochimica et
Cosmochimica Acta 64, 609-617.
112