You are on page 1of 50

Mechanical Behaviour of Bacterial Cell Walls

JOHN J . THWAITES" and NEIL H . MENDELSONb

"Department of Engineering. University of Cambridge. Cambridge CB2 I PZ. UK.


and bDepartment of Molecular and Cellular Biology. University of Arizona.
Tucson. AZ 85721. USA

1. Introduction . . . . . . . . . . . . . . . 174
A . The cell surface . . . . . . . . . . . . . 174
B . The mechanical role of the cell wall . . . . . . . . . 174
I1 . Thestructureofcell-wallmaterial . . . . . . . . . . 177
A . Peptidoglycan . . . . . . . . . . . . . 177
B . Peptides and cross-linking . . . . . . . . . . . 179
C . Other polymers . . . . . . . . . . . . . 181
111. Thephysicalstateofthecellwall . . . . . . . . . . . 182
A . Chargedcell-wallpolymers . . . . . . . . . . 182
B . The dynamic structure of the cell wall . . . . . . . . 183
C. Order in cell-wall material . . . . . . . . . . . 185
D . Thesignificanceofmacrofibres . . . . . . . . . . 186
IV . Mechanical properties . . . . . . . . . . . . . 189
A . Early observations . . . . . . . . . . . . 189
B. Direct measurements by means of bacterial thread . . . . . 189
C . Measuredproperties . . . . . . . . . . . . 192
D . Environmental effects . . . . . . . . . . . . 196
E . Visco-elasticity . . . . . . . . . . . . . 200
F. Molecular arrangement in the cell wall . . . . . . . . 202
V . Cell-wall models . . . . . . . . . . . . . . 202
A . Theaimsofmechanicalmodelling . . . . . . . . . 202
B . Geometrical models . . . . . . . . . . . . 203
C . Models involving surface tension-like stress . . . . . . . 205
D . A model involving anisotropic cell-wall material . . . . . . 207
E. A cell-wall growth model . . . . . . . . . . . 214
VI . Conclusions . . . . . . . . . . . . . . . 218
References . . . . . . . . . . . . . . . . 220

ADVANCES IN MICROBIAL PHYSIOLOGY. VOL .32 CopynghtO 1991. by Academic Press Limited
ISBN C-124277324 All rights of reproduction in any form reserved
174 J. J. THWAITES A N D N. H. MENDELSON

I. Introduction

A. THE CELL SURFACE

The external boundaries of living cells provide the interface between two
very different environments, one highly regulated by homeostatic bio-
chemical mechanisms within cells, the other highly variable to extremes in
the chemical and physical nature of the environment. The functions of the
boundary are complex: it plays a structural role, it accommodates the
selective movement of materials through itself, it undergoes changes made
necessary by growth within or in response to forces from without and it
transfers information about the environment into the cell. The materials
that serve as cell boundaries consist of: (i) lipids organized into bilayer
membranes, (ii) sugar polymers built into exoskeletal girdles (cell walls) and
(iii) proteins (Rogers et al., 1980). The properties of these materials dictate
the kinds of function each is suited to perform. We will not deal with
membranes here. Sugar polymers, such as cellulose and peptidoglycan that
comprise the cell-wall materials of plant and bacterial cells, respectively, are
made of covalently linked, sometimes multilayered, sheets of material that
form porous networks of considerable strength. They derive their integrity
from the chemical nature of the linkages that bind the monomers together
into polymers, or that link polymers to one another, as well as from
hydrogen-bonding, electrostatic and hydrophobic interactions. Thus, the
strength of cell-wall materials, a most significant feature, originates in the
strength of the bonds between the components and the structural organiza-
tion of the components in the wall (Marquis, 1988). The properties of
proteins also have bearing on bacterial cell-wall mechanics. The folded
three-dimensional shape of a protein polymer involves helical and sheet
regions held in position by a combination of cross-bridges, charge interac-
tions and hydrodynamic factors (Cantor and Schimmel, 1980). The final
configuration is highly inflexible but not totally unable to change shape in
response to diverse stimuli. Many of the mechanical properties of peptido-
glycan described in this article reflect the behaviour of the peptide portion of
the polymer. It is a considerable challenge, therefore, to understand the
behaviour of bacterial cell walls as a material. The effort is warranted in view
of the significant role cell walls play in cell growth and survival.

.
B THE MECHANICAL ROLE OF THE CELL WALL

I . Mechanical Requirements
When dealing with bacterial cellular processes that involve the cell surface it
is often difficult to separate the contributions of the cell membrane and the
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 175
cell wall. When it comes to the mechanical role of surface components,
however, distinctions become much clearer owing to the physical nature of
the materials involved. For example, stiffness in the cell surface, which is
required to maintain shapes other than spherical, can only be a property of
the wall. This is easily shown by removing the wall from a rod-shaped cell
under osmotically stabilized conditions. The resulting wall-less form is
always spherical ,indicating that surface-tension forces predominate and the
cell membrane assumes a minimum surface area (Gilpin and Nagy, 1976).
Similarly, if cell walls from rod-shaped cells are purified under conditions
where they remain intact, the collapsed empty structures clearly resemble
the shape of the original cylindrical cells (Beveridge, 1981). Wall stiffness
must therefore maintain shape in living cells even under conditions where
cells grow, which raises the question of whether stiffness and shape might
influence the growth process itself.
Bacterial cells of all shapes have a kind of structural polarity that becomes
evident during growth and division when their structures gradually change in
an ordered way from that of the initial shape to elongated versions of the
same form. As new cell wall is assembled a continuity of shape is preserved.
The cell wall is stiff enough to maintain that shape and, at the same time,
ductile enough to permit the expansion made necessary by synthesis of
cellular materials and consequent growth of the cell within. Growth polarity
appears to be an essential aspect of an ordered cell cycle, possibly because
the cell surface plays a role in segregating DNA to progeny cells. The ability
to maintain shape while expanding in one dimension and at the same time
retain flexibility suggests that the structural organization of the cell-wall
polymers cannot be highly crystalline. Instead the wall acts mechanically
more like an elastic gel. Its flexibility governs its response to transient
deformations and its elasticity is shown by its ability to recover from such
temporary shape changes. Unlike an inert gel, however, a cell wall can also
undergo changes linked to wall metabolism or growth that persist for
generations because the force imposed by growth continues and the material
itself becomes altered biochemically. The ability to undergo adaptive change
is the hallmark of many biological materials, including bacterial cell walls.
Perhaps the most obvious requirement in the cell wall is strength. It must
provide protection for the cell membrane from outside forces. It must also,
like the wall of a pressure vessel, be strong enough to withstand the turgor
pressure within the cell. Turgor is essentially a hydrostatic pressure,
maintained osmotically, which forces the membrane outwards against the
inside of the cell wall (Csonka, 1989). It is almost certain that there are no
internal cytoplasmic structures that can bear any of the osmotic force, and
the membrane itself is relatively weak. Finally, the wall must withstand the
electrostatic repulsion between its parts. For it is highly charged and, under
176 1. 1. THWAITES AND N. H. MENDELSON

normal growth conditions, the charges are not entirely neutralized. There
are some complicating factors, however. It is not absolutely clear how the
cell membrane engages the wall and thus how the turgor pressure is
transferred to the wall polymers. Also, when the external osmolality is
varied experimentally to change turgor pressure, cells respond by homeo-
static mechanisms that keep the turgor pressure within a limited range,
suggesting an important role for turgor in cell physiology. These difficulties
notwithstanding, turgor-pressure measurements reveal that bacterial cell-
wall material is very strong. And, as might be expected, the more material
that is in a wall, the greater is the turgor-pressure load it can bear.

2. Other Functions
The cell wall has other functions for which its mechanical properties are
important and which are of significance in cell physiology. Three of these
are: (i) permeability through the wall, (ii) anchorage of structures in the cell
wall or their protrusion through it and (iii) the possible role cell walls might
play in sensing and/or transferring information into cells. Everything that
passes into or out of bacterial cells has got to traverse the wall as well as the
cell membrane (or membranes). Normally, walls are thought to act as sieves
that cannot pass materials larger than a molecular weight of 1200, or that
have a radiusgreater than 1.1nm (Scherrer and Gerhardt, 1971). There are,
however, exceptions and its appears likely that the charge structure of the
wall and related features of its physical state, such as its degree of
contraction, almost certainly play a role in controlling what can and cannot
pass through it. Flagella and pili are protein-containing structures that
penetrate the cell wall yet do not jeopardize its structural integrity. They
somehow effectively plug the opening through which they penetrate, thus
preventing turgor pressure from extruding the membrane through it. The
means by which they d o so is not clear. Transfer of information into bacteria
through the cell wall is a subject of new interest and great potential
significance in that it provides a means by which the physical nature of the
environment can be linked directly to regulation of cell processes. The
precise role of the wall is not yet understood, however, as can be seen by
considering some recent examples.
A number of systems have been described in which bacteria regulate
expression of their genes in response to mechanical rather than chemical
stimuli. The cell surface is involved in all of these, so the cell wall may play an
important role in transfer of information into cells. Stress in, or deformation
of, the wall structure may be part of the process. When the marine bacterium
Vibrio parahuemolyticus swims in solutions of the same viscosity as its
normal ocean habitat, it uses a single polar flagellum for propulsion. If the
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 177
viscosity of the solution is increased or the cells made to grow on a surface
rather than in solution, they respond by turning on genes otherwise not
being expressed, at least one of which leads to production of lateral flagella.
Other phenotypic changes also arise in addition to the new means of
propulsion (Belas et al., 1986). Evidence that the large polar flagellum may
act as a dynamometer comes from studies of mutants with defects in its
structure or function in which lateral-flagella genes were always expressed
regardless of medium viscosity (McCarter el al., 1988). It appears, there-
fore, that the normal functioning of the polar flagellum is needed to keep
lateral-flagella genes turned off. The signal leading to gene regulation could
be the surface drag engendered by moving through the solution, the torque
applied to the cell body as a result of polar flagellar rotation, or the power
required by the polar flagellum to drive the cell through the solution.
In another system, gene expression in Escherichia coli is affected by
changes in osmotic pressure. The cell wall obviously participates, but
whether as a sensor, transducer or just scaffolding for them is not known
(Csonka, 1989). Both turning on and turning off of genes in response to
osmotic changes have been found. The results suggest that the state of
turgidity in a bacterial cell provides information in some mechanical way for
gene regulation. Changes in turgidity set into play a cascade of biochemical
reactions leading eventually to changes in gene expression. A mechanical
component is needed to initiate the cascade. A third bacterial system in
which physical stimuli have been found to initiate changes in gene
expression involves a marine organism that inhabits deep-sea environments
(Bartlett et al., 1989). These cells respond to barometric pressure. The
amount of gene expression observed is positively correlated with the
barometric pressure under which the cells are grown. These three examples,
and also the newly discovered “touch-sensitive genes” in plants (Braam and
Davis, 1990), show that forces acting on the cell surface can evoke responses
at the level of gene expression. The mechanical properties of the cell wall
must, therefore, be involved in transfer of information. To understand how
cell walls participate in the processing of such information requires
knowledge of their material properties as well as their structure and
composition.

11. The Structure of Cell-Wall Material

A. PEF’TIDOGLYCAN

A considerable amount is known about the structure, composition and


synthesis of walls from both Gram-positive and Gram-negative bacteria. As
178 I. I. THWAITES AND N. H. MENDELSON

one would expect on evolutionary grounds, diverse materials and organiza-


tions are utilized by different species (Murray et al., 1965), but the common
aspects of them all are not difficult to identify. The key material is a
heteropolymer, murein or peptidoglycan, consisting of a polysaccharide
backbone to which short peptides are covalently linked. This is the strength-
bearing polymer of virtually all eubacteria. Its molecular composition
consists of a glycan backbone made of disaccharide (p-( 1+4)-linked
alternating residues of N-acetylglucosamine and N-acetylmuramic acid).
The glycans are cross-linked by covalent connections between short
peptides that emanate from some o r all N-acetylmuramic acid residues along
the backbone. The meshwork or sacculus so formed is the material which
governs the mechanical nature of the wall.
The organization of peptidoglycan in the bacterial wall was revealed using
the electron microscope. In Gram-negative cells, only a thin 2-3 nm layer is
present lying between the cell membrane and an outer membrane (Murray
et al., 1965). Gram-positive cells, in contrast, lack the outer membrane but
have much more peptidoglycan, on average a layer about 30 nm thick
(Shockman and Barrett, 1983). Calculations based upon the amount of
peptidoglycan in each cell, ideas about the molecular structure of peptido-
glycan, its density and the surface area that must be covered, led initially to
the idea that the thin layer in Gram-negative cells was barely sufficient to
form a monomolecular sheet (Braun et al., 1973). Re-examination of the
molecular models later suggested that perhaps three layers could be formed
(Schwarz and Glauner, 1988). Recent findings about turnover and re-
utilization of peptidoglycan, and also details of its cross-linking, support the
idea that at least two layers are present in Gram-negative cells (Goodell and
Schwarz, 1985). New microscopical techniques have also contributed to a
revision of ideas about the way in which the Gram-negative cell wall may be
organized. The full volume between inner and outer membranes might be
filled with a gel-like peptidoglycan material (Hobot et al., 1984). Whether in
the form of a sheet or a gel, the glycan backbones must lie predominantly in
the plane of the surface, for they are too long to stick straight out radially
(Rogers etal., 1980). The orientation of the glycans on the surface, however,
has not yet been definitively determined. Partially fragmented and digested
purified walls, when examined by electron microscopy, reveal a meshwork
pattern suggesting an alignment primarily perpendicular to the cylinder axis
(Verwer et al., 1980). But the overall organization in a circumferential
pattern appears not to be highly regular. There is considerable evidence
against any crystalline order of the polymers within bacterial walls
(Labischinski et al., 1983). Other polymers, however, are known to lack
crystallinity yet still, on average, be highly ordered. It is reasonable to
suppose that peptidoglycan may fall into this class of material.
The walls of Gram-positive bacteria (such as Bacillus subtilis) maintain
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 179
their thickness by a turnover process that requires constant synthesis and
shedding (Pooley, 1976 a,b). The structure of the wall is therefore in a
constant dynamic state. The organization of glycans within the wall is, again,
not certain. The length of chains, however, precludes their sticking straight
out in a radial fashion unless folded, A more likely arrangement is as in
Gram-negative cells; the glycans lie parallel to the cell surface. Chemical
analysis of glycan lengths in B. subtih (which, if anything, is likely to
underestimate, owing to cleavage by autolytic enzymes) indicates a range of
30-600 nm (30-590 disaccharide repeats) (Shockman and Barrett, 1983).
Even the longest strands cannot, therefore, stretch the entire length (about
3-4 pm) or circumference (about 2.5 pm) of a cell. A linked network is
needed, therefore, to cover the entire surface. However, there is anything
but agreement as to how the glycans are arranged in the linked cell-wall
network. Everything from a totally random net, resembling a non-woven
material (Koch 1989), to a highly precise crystalline organization (Burman
and Park, 1984), has been proposed. Recent molecular models depict the
glycan backbone as a right-handed helix with a repeat distance of four
disaccharide units per turn. The twisted nature of the backbone enables the
peptides to point in any direction. And, since cross-linking of one glycan to
another can only arise when the two peptides that join are within a
prescribed distance from each other, the backbone geometry limits the kind
of regular three-dimensional structure that can be formed. Even with this
limitation, however, there remain at least seven different possible arrange-
ments of the glycans relative to one another, known as topotypes. Each
topotype has its own characteristic level of possible cross-linking, its own
number of layers of glycans and its own predicted extensibility(Labischinski
et al., 1985). The latter is based on the idea that the backbone glycans are
rigid and unable to be stretched or compressed along their length very much
before rupturing bonds, whereas the peptides are thought to lie in a ring-like
arrangement flat on the backbone with the ability to assume a straightened
configuration, thereby extending the distance between neighbouring
glycans by a factor of about three. Other degrees of freedom which might
contribute to extension have not been considered in these topotypes.

B. PEPTIDES AND CROSS-LINKING

Although the chemical composition of the glycan portion of peptidoglycan is


essentially the same for all organisms from which the polymer has been
isolated," the composition of the peptide moiety, the types of linkage that

Occasional variationsinvolve acetylation or phosphorylationof muramic acid residuesor loss


of N-acetyl groups in the endospore cortex conversion of a muramic acid residue to a lactam,
and in some mycobacteriaand corynebacteriareplacement of an N-acetyl group on a muramic
acid residue by an N-glycol group (Heymer et ul., 1985).
180 J . J. THWAlTES AND N. H. MENDELSON

join peptides, and the presence or absence of interpeptide bridge structures,


differ greatly among various bacterial species (Ghuysen, 1968). Regardless
of precisely which residues are present in the stem peptides, they always
appear to consist of alternating L- and D-amino acids with an L form linked to
the glycan. For example, in B. subtilis, the stem pentapeptide consists of
residues of L-alanine, D-glutamic acid, rneso-diaminopimelic acid and D-
alanyl-malanine. It is not certain what function this pattern serves, but it
may confer resistance to protease hydrolysis which would otherwise lead to
solubilization of the wall network (Schleifer and Kandler, 1983). At least
three variations on linkage structure between peptides have been found. A
common pattern involves linkage between the third-position amino-acid
residue from one stem peptide to the fourth-position amino-acid residue
from another, as for example the linkage of a diaminopimelic acid residue
from stem A to a D-alanine residue in stem B in B. subtilis and E. coli. A
second linkage pattern involves bond formation between the second-
position amino-acid residue from one stem peptide and the fourth-position
amino-acid residue from another. The third variety, recently found in E.
coli, involves a direct connection of third-position amino-acid residues to
one another (Schwarz and Glauner, 1988). These differences and others
involving the joining of stem peptides through interpeptide bridge material
(Schleifer and Stackebrandt, 1983) are very likely to influence the mechani-
cal behaviour of peptidoglycan and thus the material properties of the walls
from different species.
Two very important structural parameters in polymers are chain length
and degree of cross-linking; these parameters for peptidoglycan are
significant in terms of cell-wall biomechanics. Predictions of the cross-
linking index for peptidoglycans have been made from models which take
account of the restraints imposed by the geometry of the glycan backbones.
Values obtained for different topotypes range from 24% in a type that could
only form monolayers to 72% in a type that could form a multilayer network
similar to the polymer in the wall of a typical Gram-positive bacterium.
Measurements of the cross-linkingindex give values ranging from 20 to 93%
in different bacterial species. Even the lowest values for peptidoglycan
cross-linking are high in comparison with other polymers. Peptidoglycan
chain length, however, is much shorter. A relationship between chain length
and degree of cross-linkinghas recently been noted from studies of polymers
in E. coli mutants. It appears that, as chain length decreases, the degree of
cross-linking goes up (Schwarz and Glauner, 1988). It has also been
suggested that such variations may exist in different parts of the same cell,
such as in rod-shaped cells, the cylinder and the poles. If so, one might
expect the corresponding regions to behave quite differently mechanically.
There are other reasons for believing that the poles and cylindrical portions
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 181
of rod-shaped cells may differ mechanically. The walls in these two regions
are assembled in a different manner in Gram-positive cells; they have vastly
different rates of turnover, and they also bear the turgor pressure in
different ways.

C. OTHER POLYMERS

The proportion of cell wall that is peptidoglycan can range from 40% (by
weight) to as high as 95% (Marquis, 1988). Most of the remainder of the wall
in Gram-positive cells is a negatively charged accessory polymer, namely
teichoic acid or teichuronic acid (Rogers et al., 1980). The presence of these
anionic polymers has a profound effect on cell-wall structure. If, for
example, in B. subtilis, the glycerol teichoic acid polymer normally
covalently attached to C-6 of the N-acetylmuramic acid residue in the
peptidoglycan is lost by mutation, the remaining wall becomes highly
disorganized, thickened, and extremely roughened on its outer surface. Cell
shape changes under these conditions from rod to sphere (Cole et al., 1970).
Similar phenotypes have been observed in other situations where cell walls
become deficient in negatively charged polymers (Rogers, 1979). The
mechanical properties of walls must also be influenced by the electrostatic
nature of these polymers. Accessory polymers become inserted into the wall
attached to new peptidoglycan, and are shed into the medium with it during
turnover (Mauck and Glaser, 1972; Boylan et al., 1972). Their spatial
arrangement within the wall is not certain. Those on the surface project
radially from it (Birdsell et al., 1975), while those in the peptidoglycan
network probably lie parallel to the surface (Archibald et al., 1973). If so,
the chains within the wall would be in a position to interact with one another
as well as with charged groups on the peptidoglycan peptide.
Very little is known about how other polymers that are found in, or
associated with, cell walls might contribute to their mechanical behaviour.
In walls of B. subtilis, a dozen or so proteins ranging in mass from 14to over
200 kDa have been found (Studer and Karamata, 1988; Doyle et al., 1977).
Neither their function nor their location is well enough understood to permit
an assessment of their role, if any, in wall mechanics. The same can be said
for the cell membrane, which, in Gram-negative cells, is known to penetrate
in regions through the peptidoglycan (Bayer’s adhesions; Bayer, 1979).
Wall-associated polysaccharides, such as exocellular capsules and slimes,
could be significant in terms of mechanics, but virtually nothing is yet known
about their mechanical properties (Isaac, 1985; Costerton et al., 1981).
182 J. J. THWAITES AND N. H. MENDELSON

111. The Physical State of the Cell Wall

A. CHARGED CELL-WALL POLYMERS

Structural studies have not yet provided a clear picture of the physical state
of the cell wall, but other experimental approaches have given an outline of
the situation. Key factors are: (i) the charged nature of wall polymers, (ii)
the dynamic nature of wall structure, and (iii) the degree of order in it. The
wall appears to be a somewhat open porous network, fully hydrated, in
which counter-ions maintain neutrality (Marquis, 1988). In vivo, the
network is under pressure from within the cell. It is stretched, but it can
expand or contract and, as described later, generally behaves as a visco-
elastic material.
The electronegative nature of the bacterial wall is due to dissociation of
charged groups, which can be identified by titration. Using the electrophore-
tic mobility of whole cells and of purified walls as functions of pH value, in
order to measure ionization, carboxyl groups in E. coli and both carboxyl
and phosphate groups in B. subtilis were shown to be the source of
electronegativity (Neihof and Echols, 1968). Carboxyl groups in peptido-
glycans are located on residues of D-glumatic acid, meso-diaminopimelic
acid and terminal D-alanine residues of the peptide. Phosphate groups are
present along the backbone of teichoic acids. At neutral pH values, cell
surfaces are highly negatively charged because of these groups. Electrostatic
repulsion between negative forces in the wall represents a significant factor
in terms of wall mechanics. If the charges were unneutralized, the repulsion
they generate would far exceed the strength of the wall, essentially blowing
it apart. Two sources of cations are available to neutralize negative charges,
namely amino groups in the matrix of peptidoglycan, and counter-ions from
the environment. In walls of some Gram-positive bacteria there are so many
anionic groups to be neutralized that the capacity of the walls to bind cations
is on a par with that of commercial ion-exchange resins (Marquis et al.,
1976). Walls from other species are not so electronegative, however. Factors
such as the degree of cross-linking (each cross-link eliminates two charged
groups), amidation of glutamic acid residues, which converts the carboxyl to
an uncharged amide, or alanyl substitutions on teichoic acids, which
neutralize neighbouring phosphate groups, all serve to decrease negative
charge. Even so, in all cases, peptidoglycan appears to behave like a flexible
polyelectrolyte gel that swells or shrinks in response to electrostatic factors
and, thus, the density and porosity of the cell wall must vary accordingly
(Marquis, 1968).
The compactness of the cell wall varies from organism to organism. The
range of measured dextran-impermeable volumes is between 2 and 13 ml
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 183
(g dry weight)-'. These values can be used in conjunction with the known
density of hydrated walls to determine the fraction of wall volume occupied
by polymer and by water. The wet density of walls has been measured and
found to be very low, less than 1.1g ml-' even in highly compact walls (Ou
and Marquis, 1972). Thus, the cell wall appears to consist of about 74%
water by weight when the walls are contracted and about 93% water when
expanded (Marquis, 1988). If these values are an accurate estimate of the
state in vivo, there does not appear to be any chance for wall polymers to be
organized in a crystalline fashion. Instead, a flexible, open, sponge-like
material is indicated. One would not expect hydrogen bonding or hydro-
phobic interactions to be significant under these circumstances since the
molecular chains within the network are not close enough to one another.
The effects of temperature and urea on cell-wall compactness do, however,
show that weak interactions are present though not extensive (Ou and
Marquis, 1972). The bacterial cell wall is quite different, therefore, from
those of plants and fungi that consist of cellulose and chitin. The latter both
form mechanically rigid structures containing crystallites rather than flexible
gels. The peptide side chains in the bacterial wall, short as they may be, are
long enough to allow formation of a very open flexible network.

B . THE DYNAMIC STRUCTURE OF THE CELL WALL

The cell wall is in a dynamic state. The most obvious manifestation of this is
that the cell grows. In rod-shaped cells, for example, the length increases
rapidly while the diameter remains remarkably constant (Trueba and
Woldringh, 1980). At any given time, the wall material must be stretched.
Evidence that the wall is stretched in living cells comes from measurements
of cell shrinkage caused by eliminating turgor pressure (Koch, 1984; Koch et
al., 1987). The stretched surface is apparently between 20 and 46% greater
in area than it is without turgor. The stretching effect of turgor pressure may
play a major role in expansion of the wall during growth, possibly even in
regulation of growth itself (Koch, 1983). The magnitude of turgor pressure
in bacterial cells has been estimated from: (i) measurement of the total
weight of internal solutes in crude cell extracts (from which water activity
could be calculated), (ii) using vapour-pressure diffusion equilibrium
methods, (iii) quantitative determination of the osmolarity that induces
plasmolysis, and (iv) using collapse of intracellular gas vacuoles as a function
of applied hydrostatic pressure (for a survey see Csonka, 1989). All of the
methods used have been criticized for one reason or another, so the values
reported are probably not sufficiently accurate to warrant critical predic-
tions of how strong the cell wall must be in order to bear the turgor pressure.
Estimated values of turgor pressure in Gram-positive cells under normal
184 J. J. THWAITES A N D N. H. MENDELSON

growth conditions are in the range of 1.5-2.0 MPa. The values for Gram-
negative cells are lower (about 0.1-0.5 MPa). The upper value of about 20
atm is remarkably high to be contained by an open-structure wall that is
mainly water!
Many of the other aspects of the dynamic nature of the wall have been
explored in Gram-positive cells. The basic pattern involves a constant loss
and replacement of wall polymers during growth. In B. subtilis, newly
inserted peptidoglycan starts to be shed approximately one cell-generation
time after it has been inserted into the wall. The rate of turnover (shedding)
is proportional to growth rate, the major factor apparently being the time it
takes for polymer in the cylindrical wall to be pushed up to the surface by
new material inserted below it (Pooley, 1976a). During steady-state growth
of wild-type cells of B. subtilis 168, about 50% of the cylindrical wall
material is shed in each generation. For reasons not yet clear, very little wall
material is turned over in the cell poles; 15-20 generations are required to
turn over pole material compared to between three and four for cylindrical
wall (Doyle and Koch, 1987; Doyle et al., 1988). Pole and cylindrical wall
appears to be controlled differently and, perhaps, the enzymes involved are
less effective on polar wall. Prevalent as it is, cell-wall turnover is not
required at the rate usually found in wild-type cells. Bacillus subtilis
mutants, such as the FJ7 strain that carry a lyt2 defect which results in a
decrease in autolytic enzyme activity to between 5 and 10% of that of the
wild-type, grow perfectly well, but turn over their walls at less than 20% of
the wild-type rate (Mobley et al., 1984). The only obvious phenotypic
defects noted in such strains are an inability to separate cells following
septation, which leads to formation of chains of cells and a lack of flagella,
presumably because there is insufficient autolysin to open the wall in order
to permit their penetration (Fein, 1979). There is now evidence that
peptidogylcan in walls of Gram-negative bacteria undergoes constant
turnover. About 50% of the peptides in mature peptidoglycan in E. coli is
released during each cell generation (Goodell and Schwarz, 1985). The
outer membrane, however, blocks it so that only 8% of the peptidoglycan
breakdown products are shed into the environment. The remainder is
trapped in the periplasm and is re-utilized. Surprisingly, some of it,
consisting of tetra- and tripeptides, appears to be recycled directly into new
peptidoglycan instead of being processed first to individual amino acids. The
relationship of synthesis to turnover in walls of Gram-negative bacteria must
be very precisely regulated in view of the very small amount of material
present and the seemingly indispensable role it plays in maintenance of cell
structure and in cell growth.
Breakdown of peptidoglycan either in the periplasm of Gram-negative
cells or at the surface of the wall on Gram-positive bacteria requires cleavage
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 185
of covalent bonds. Stresses in the wall due to hydrostatic pressure and
expansion due to growth may contribute, but the primary mechanism of
peptidoglycan breakdown involves cleavage by specific enzymes. Such
enzymes are potentially lethal to cells; their activities must be well
regulated. Four means to do so have been proposed. One involves inhibition
by lipoteichoic acids, a second relies upon the association of autolysin and an
activator protein into a functional complex, the third is based on the idea
that an energized cell membrane inhibits nearby autolysins, and, lastly,
extracellular proteases may inactivate or activate autolysins at the cell
surface (Shockman and Barrett, 1983). In B. subtilis, two autolysins have
been identified. One cleaves the glycan backbone, the other severs the
connection of stem peptides to the backbone. Both these activities are
decreased in lyt mutants. Strains totally devoid of them have never been
found, however, which suggests that some peptidoglycan remodelling must
be required during growth. In E. coli, both kinds of peptidoglycan
hydrolases have also been found. Both activities are needed to generate the
tetra- and tripeptide breakdown products already described. There is much
similarity, therefore, between the dynamic behaviour of cell-wall polymers
in Gram-positive and Gram-negative bacteria.

C. ORDER IN CELL-WALL MATERIAL

One of the most compelling indications that there must be some order in the
arrangement of peptidoglycan in the cell wall is the fact that bacterial cells
are usually not spherical. An isotropic wall would respond to turgor pressure
during growth by expanding to a sphere just as protoplasts do. The tendency
to produce a spherical shape must, therefore, be resisted by an anisotropic
material in order that shapes such as quasi-spherical, cylindrical, helical or
differentiated, as in the stalked cells of Caulobacter spp., can be produced
and maintained. From a mechanical perspective, there is no alternative
short of eliminating turgor pressure, which is apparently the case in the
square bacterium Halobacterium sp. (Walsby, 1980), but not in others.
Another is the observation that rod-shaped cells twist as they elongate.
This has been inferred from the behaviour of long multicellular filaments
and of twisted multifilament structures (macrofibres, see Section 1II.D)
which form as the result of filament plying (Fein, 1980; Mendelson, 1982a).
In these, the growth process is itself anisotropic in a controlled way. The
obvious conclusion is that cell walls which maintain their structural integrity
are also anisotropic, either because of the mode of insertion of new polymer
or perhaps because of the twisted nature of the peptidoglycan molecule
itself. Measurement of twist in multicellular structures indicates that, for B.
subtilis, the helix angle of the cell-wall twist can be as large as about 8" right
186 J J THWAITES AND N H. MENDELSON

hand and 6" left hand (Mendelson et af.,1984) and, depending on a number
of a factors, anywhere between. All of the factors appear to operate by
controlling some aspect of the insertion of new polymer into the cell wall
(Mendelson and Thwaites, 1988). The result is that a given cell in a given
environment grows with a particular geometry, one of many that it is capable
of producing. From these observations we infer that the bacterial cell wall
must be able to assume a range of structural states, each perfectly stable and
able to pass from parent to progeny so long as the environment remains
constant. Studies of twist in macrofibres provide evidence that peptido-
glycan metabolism, the integrity of the glycan backbone and electrostatic
interactions in the cell wall of individual cells are involved in the establish-
ment and maintenance of cell-wall twist (Mendelson and Thwaites, 1988;
Favre et al., 1986; Mendelson et al., 1985; see also Section 1II.D). None of
these, however, requires that the molecular architecture of the wall be
crystalline or otherwise highly regular. Amorphous polymers are well
known to possess order and to have anisotropic properties (Ferry, 1980).

D. THE SIGNIFICANCE OF MACROFIBRES

Macrofibres are self-assembling twisted structures formed in various cul-


tures of the same cell separation-suppressed mutants of B. subtilis as those
used to make bacterial thread (Mendelson, 1978, 1982a). During growth,
the ends of individual cellular filaments consisting of long chains of cells have
for many years been observed to rotate in opposite directions as the
filaments elongate. This phenomenon has been observed in Mendelson's
laboratory since 1977 in time-lapse films. As the filaments twist, they also
bend, so producing a writhing motion which persists until a filament touches
itself, whereupon it plies into a two-fold twisted structure (Fein, 1980). The
two-fold structure behaves in exactly the same way as a single filament so
that four-fold, and eventually manifold, twisted structures are formed.
Although the twist has not yet been measured for a single filament, it can be
calculated from measurements of surface-helix angle and diameter in
micrographs for macrofibres of all orders from about four-fold upwards. For
a given environment, no matter what the order, the twist (turns per unit
length) is the same. Twist can also be obtained dynamically, i.e. by
measuring the relative rotation rate of the ends of a macrofibre fragment and
relating this to the elongation rate. Such measurements made from stop-
action microcinematograph films agree with measurements from still images
(Mendelson et al., 1984). Since the behaviour of individual filaments is
qualitatively the same as that of the multifilament structures of all orders,
the obvious conclusion is that the filament twist has the same value as
macrofibre twist.
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 187
The simplest explanation for formation and development of macrofibres
is that, when a filament (or a macrofibre) touches itself, its twisting-with-
elongation is blocked (Mendelson, 1976). This requires an axial torque of
opposite handedness to the twist, which is provided in some way by the
contact. A long filament under torque but under no tension (indeed, since it
is growing through a fluid, under slight thrust) is not in a stable state, even
less so when in a loop. The mechanical torsion is relieved by plying in just the
same way as textile yarns ply together. The ply twist is in the opposite
direction to the torque, i.e. of the same handedness as the blocked twisting
of the filament. Thus, for a given environment, the twist of macrofibres of all
orders is the same. Other possible explanations have been advanced
involving much less plausible assumptions. Koch (1988), for example,
finding difficulty with the idea that cells at the centre of a macrofibre can
metabolize properly, suggests that the outer filaments, growing faster, might
be restrained by the inner and so buckle in a helical fashion. This would not
explain the plying of single filaments or of macrofibres of low order, where
there are no central cells. Nor can it explain the range of twist, both left- and
right-hand, that can be engendered in macrofibres by changes in environ-
ment (see, for example, Mendelson et al., 1985; Mendelson and Thwaites,
1988a). In addition, buckling due only to thrust never takes a helical or
twisting form, whereas torsion is relieved in this way. Centrally placed cells
in large macrofibres probably do metabolize at a slower rate but, since each
filament migrates due to the twist in a macrofibre, from inside to outside, all
filaments behave in essentially the same way.
Although the origin of the twisting of cellular filaments is reasonably well
understood (see Section V), filament bending during growth is less so. It
may in part be due to random differential variations in growth rate in
different segments of the cell wall round the circumference. However,
observations of filament behaviour following an imposed disturbance
suggest that it may be a dynamic material response. A major cause must be
the thrust developed by filament growth through the culture medium.
Although the viscosity is approximately that of water, the ends of a long
filament are moving apart at a speed of the order of one cell diameter per
second. The thrust developed is not sufficient to cause buckling in the
classical sense, but it does not take much thrust to cause writhing, as can be
observed in the motion of the hosepipes that are used for cleaning (by their
motion) the bottoms of swimming pools. The thrust in this case is due to the
water efflux at the pipe tip. However, when the viscosity of the culture
medium is artificially increased lOW-fold, macrofibres develop by classical
buckling of an almost straight structure without contact of any kind. It can be
shown that the thrust due to growth is all that is required for such a
phenomenon (Mendelson and Thwaites, 1990).
188 J. J. THWAITES AND N. H. MENDELSON

Macrofibres can grow with a wide range of twist. The corresponding cell-
wall helix angle can lie anywhere in a spectrum extending from about 8”right
hand to about 6” left hand (Mendelson and Thwaites, 1988). The cell surface
must, therefore, assume a corresponding range of structural states. Since
cells growing with twist appear to be completely normal in growth and
heritability, their walls must be perfectly functional. Once a twist state has
been established, it passes from parent to progeny cells as long as
environmental conditions remain the same. This suggests that the cell
surface is assembled with a particular structural or mechanical anisotropy
that governs the characteristic pattern of twisting with growth. Evidence
that peptidoglycan is involved in both the establishment and maintenance of
twist comes from: (i) the fact that twist establishment is blocked by low
concentrations of penicillin G (Zaritsky and Mendelson, 1984), (ii) the
discovery that twist establishment is influenced by the concentration of D-
alanine, or its antagonist, D-CyClOSerine, in the growth medium (Mendelson,
1988) and (iii) the finding that lysozyme digestion of live cells causes rapid
changes in twist before liberation of sphaeroplasts (Favre et al., 1986).
Peptidoglycan cross-linking may, therefore, play a role in twist establish-
ment. The integrity of the glycan backbone is necessary for twist maintenance.
Wall charge must also play a significant role, for changes in pH value and ions
such as bromide also induce rapid twist changes (Mendelson et al., 1985).
The significance of macrofibre experiments in relation to the mechanical
behaviour of cell walls is that they magnify certain cell-wall behavioural
phenomena such as twisting-with-elongation.Not only this, like the
bacterial thread system, they offer a means of quantitatively describing this
behaviour, for example, by being able to measure twist. They are also very
sensitive indicators of change in the cell wall (see, for example, Favre et al.,
1985,1986). So they provide information about cell shape and wall assembly
that relates to single cells, and which is not obtainable by other means. The
information obtained does not just relate to mutants. There is no reason to
suppose that cell walls of B. subtilis strain FJ7, for example, are significantly
different from those of the standard 168 strain of B. subtilis which will, if
grown in very low-density cultures, form filaments (Zaritsky and MacNab,
1981). Mutants are used because their filaments are much more stable.
Macrofibre formation has been observed in at least six other species, namely
B. cereus (Roberts, 1938), B. licheniformis (Robson and Baddiley, 1977),an
unnamed “new caldoactive filamentous bacterium’’ (Hudson et al., 1984),
Mastigocladus laminosus (Hernandez-Muniz and Stevens, 1988),
Clostridium autobutylicum (M. Young, private communication, 1986) and
B. sphaericus (Y. Hoti, private communication, 1987). There is every
reason to believe that the cell walls of these organisms can be successfully
investigated by means of macrofibre experiments.
MECHANICAL BEHAVIOUR OF BACXERIAL CELL WALLS 189
IV. Mechanical Properties

A. EARLY OBSERVATIONS

Until very recently little was known about the mechanical properties of
bacterial cell walls. Manipulation of cells (Wamoscher, 1930) had shown
walls to be extensible, flexible and, in some sense, elastic. By mounting on
films which were then stretched they were shown to be highly ductile (Knaysi
et al., 1950). The first measurements of deformation in relation to its cause
showed that walls could shrink by up to 50% in volume in salt solutions
depending on the bacterial species, due to electrostatic forces (Marquis,
1968; Ou and Marquis, 1970). Later, attempts were made to relate
deformation to turgor (Koch, 1984; Koch et al., 1987). Contractions in
length in the range 12-17% were inferred for E. coli when turgor was
suppressed, but there were complicating factors in these experiments which
make the figures unreliable as estimates of the extent of deformation due to
stress in walls alone. All of these observed deformations were much less than
the 100% that could be imposed (slowly) on the plasma membrane (Corner
and Marquis, 1969) and the theoretical figure of about 200%, based on
models of peptide configuration in peptidoglycan (Labischinskiet al., 1983).
The only knowledge of cell-wall strength was that it was large enough to
bear turgor pressure (Mitchell and Moyle, 1956) but, since the magnitude of
this was a matter for some dispute and accurate measures of the cell-wall
cross-section were not initially available, there was little point in calculating
a value for cell-wall stress. Because of the small size of bacteria the only
methods of approach to mechanical properties in normal cultures were
indirect, often with tenuous lines of inference. Direct measurements
became possible with bacterial thread, which is a fibrillar fibre made from
cultures of a cell separation-suppressed mutant that can be investigated by
standard fibre-testing techniques. The existence of filament-forming
mutants has so far restricted measurements to B. subtilis but, in principle,
the technique could be used on any filament-forming micro-organism.

B. DIRECT MEASUREMENTS BY MEANS OF BACTERIAL THREAD

The mutant used is FJ7 of the 168 strain of B. subtilis (Mendelson and
Karamata, 1982; Mendelson et al., 1984). Under the right growth conditions
(Mendelson and Thwaites, 1989b) the bacteria grow in long cellular
filaments, containing on average about 300 cells, in a structure like a textile-
fibre web. From this a “thread” can be drawn into which the individual
filaments are drawn radially and compressed together into a multifilament
fibre with a circular cross-section. A standard thread is produced by
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 191

FIG. 1. Scanning electron micrographs of bacterial threads. (i) A standard thread,


i.e. with culture medium removed by washing in water. (ii) An unwashed thread, i.e.
as drawn from the culture. (iii) A lysozyme-treatedthread, i.e. washed a second time
in lysozyme (15 minutes at 50 pg 1-' concentraton).Bars are equal to 5 pm for (i) and
(ii) and 10 pm for (iii). Micrographs (i) and (ii) were orignially published in
Mendelson and Thwaites (1989b).

resuspension in a large volume of de-ionized water in order to wash off


residual culture medium. Although the filaments separate upon immersion,
they retain enough cohesion for a thread to be redrawn in which they are
closely packed and highly aligned parallel to its axis (Fig. 1). They also
adhere strongly to one another so that, when a thread is extended, it behaves
integrally in the same way that naturally occurring fibrillar fibres, such as
wool, do. Threads produced in this way can be up to 0.5 m in length and 100
pm in diameter, i.e. they contain about SO00 filaments and more than lo9
cells. These are similar dimensions to those that can be obtained without
washing (Thwaites and Mendelson, 1989).
Tensile tests are made on threads using equipment of a standard kind
placed in an atmospheric enclosure in which the humidity is maintained
constant. (Thwaites and Mendelson, 1989) Specimens which are stored in
192 J J THWAITES AND N H MENDELSON

the laboratory atmosphere are equilibrated in the enclosure before testing.


It has been established that such a procedure produces the same results as
when specimens are drawn and maintained at the relative humidity of the
test. A standard test consists of extending a specimen to break, at a constant
speed of about half the elongation rate of cells in vivo, while recording the
tension as a function of the extension. Using an estimate of the wall cross-
sectional area (Thwaites and Mendelson, 1989), the tension is converted to
wall stress and extension is expressed as a proportion of the initial length
by use of the term strain. The reasons why the properties being measured
are those of the cell wall and are not assemblage-related are given by the
authors (Thwaites and Mendelson, 1989). The main ones are as follows.
There are no signs of interfilament slippage in any stresdstrain curve.
Almost complete recovery can be otained from extensions that are
substantial fractions of the breaking extension, even at high relative
humidities. In fact, the observed stress/strain behaviour is just like that of
other polymeric materials. The cytoplasm can be shown to exert negligible
effect and the individual filaments break, not by the pulling apart of septa,
but in the cylinder wall.

C. MEASURED PROPERTIES

Much information about material behaviour can be obtained from observa-


tion of the way in which stress depends on strain. Typical stressktrain curves
for a cell wall are shown in Fig. 2 (J. J. Thwaites and U. C. Surana,
unpublished observations). At lower relative humidities (less than about
50%) the curves are approximately straight lines with much steeper slopes,
and breaks occur at slightly more than 1%strain. This behaviour is typical of
any glassy polymer but, at high levels of humidity, the shape of the stress/
strain curve, the ductile nature of the wall and its low strength are all
characteristic of a polymer above its glass transition, i.e. when it has become
rubbery. Representative parameters used to describe behaviour are the
extensibility and the strength (i.e. the strain and the stress at break, the
stress being calculated from the area of the cross-section at break) and also
the initial (Young’s) modulus, calculated from the slope of the stresdstrain
curve at zero strain. Experimental results for these are shown in Figs 3 and 4,
in the relative humidity range 2Q-95%. Cell walls are strong (about 300
MPa) but brittle at low relative humidities. They are also stiff (modulus
about 15 GPa). These properties are maintained up to about 50% relative
humidity but, as humidity is further increased, they become ductile
(extensibility rises to the range 60430%) and strength and modulus fall
dramatically, to values extrapolated to 100% humidity of 13 and 30 MPa,
respectively.
i
x 65.1

I I I I I I
0 10 20 30 40 50 60
Strain (O/O)

FIG. 2. Nominal stresdstrain curves for cell walls of Bacillus subtilis FJ7 from
standard threads at four representative values of relative humidity, indicated by
values on the graph. Corresponding curves for unwashed walls (from threads drawn
directly from the culture) at approximately 18% lower relative humidity are
indicated by t. The crosses indicate break points. Nominal stress is based on the
average diameter before testing. The rate of strain was 3.3% per minute.
Unpublished observations of J. J. Thwaites and U. C. Surana.
194 J. 1. THWAITES AND N. H. MENDELSON

FIG. 3 A plot of the tensile strength of cell walls in threads of Bacillussubtilis FJ7 as
a function of relative humidity. The response of a standard (i.e. washed in water)
wall is shown by A and that of an unwashed wall upshifted by 18% relative humidity
by 0.The tensile strength is the breaking load divided by the cell wall cross-sectional
area at break point.

These properties are, in part, similar to those of other biopolymers, but


differ in significant ways. When dry, they are, except for extensibility,
comparable with those of cellulose (Meredith, 1959). When wet, cell-wall
strength is about the same as that of chitin (Thor and Henderson, 1940), and
the modulus is about ten times that of elastin (Gotte et al., 1968). However,
although the polymer backbones of chitin, and to a lesser extent of cellulose,
are similar to that of peptidoglycan, both are crystalline to some degree and
are much higher polymers. Cellulose, for example, has about 100 times as
many residues in each chain. Its properties change with humidity, but by
much smaller factors than do those of a bacterial cell wall. On the other
hand, although the variation in modulus of elastin with humidity is similar to
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 195
""

*so
A
20-

A
AA
10

O80r3, ~

5-
OOQ
-a" 8.
-W
2-

-
3
Y)

n
: 1

-i
0
$
- 05- ?&i$ 0
A
?$

02- e "bta A
ao6bDA
00 n
01
*%
0.05
- h o 1(
8
0.02 I I I I I I I , Q
10 20 30 40 50 60 70 80 90 '0

FIG. 4. Initial (Young's) modulus of cell walls in threads of Bacillus subtilis FJ7 as a
function of relative humidity. The response of a standard (i.e. washed in water) wall
is shown by A and that of unwashed wall upshifted by 18% relative humidity by 0.
The modulus is derived from the tangent to the stresdstrain curve at the origin.

that of a cell wall, elastin is an insoluble protein rubber. Because


peptidoglycan and its accessory polymers have many ionizable and other
hydrophilic groups, there is ample opportunity for water to associate with
the cell wall, competing for the hydrogen-bond sites which provide links
between peptides when it is dry. This could explain why a stiff polymer
network becomes increasinglymore flexible as humidity increases. However, it
is possible that increasing ordering of water molecules, as occurs in elastin
and other proteins, may also contribute (Scheraga, 1980).
196 1. 1. THWAITES AND N . H. MENDELSON

If the extrapolated values for strength can be taken as appropriate for fully
hydrated walls, the cells in vivo should be able to withstand a turgor pressure
of about 2.6 MPa, which is comparable with values deduced for other Gram-
positive bacteria (Mitchell and Moyle, 1956; Marquis and Carstensen,
1973). That is, provided the wall is stronger in the circumferential (hoop)
direction, the hoop stress in a cylinder wall due to internal pressure is twice
the longitudinal stress. This does not seem unreasonable. The wall is clearly
stiffer in this direction or else the diameter could not be maintained so
constant during growth. It should be noted that the conclusion about turgor
pressure does not depend on the fact that the wall cross-sectional area is an
estimate; the same factor (four times the wall thickness divided by the cell
diameter) is involved in calculating wall stress from turgor pressure as it is
from thread tension. There are, however, other caveats. (i) The stress is,
because of inside-to-outside wall growth, unlikely to be uniformly distributed
through the wall so that, for a maximum stress of 13 MPa on the outside, the
average wall stress and, therefore, the internal pressure would be less. (ii)
The material is visco-elastic, and its properties depend upon the speed of
deformation (see Section 1V.E); it may therefore be stronger in vivo. (iii) It
is by no means certain that extrapolation to 100% relative humidity
represents the fully hydrated state. (iv) Turgor pressure may not be the only
source of stress in cell walls. Electrostatic forces have been shown to exert a
strong influence on cell-wall deformation (Marquis, 1968). Their influence
during growth is unknown. Some of these difficultiescan be addressed by the
use of mathematical models of the cell wall (see Section V). Meantime, it is
worth noting that Gram-negative bacteria have both thinner walls and lower
turgor pressure than do Gram-positive bacteria. Taking the latter to be
0.3 MPa (Stock et al., 1977), and assuming that the periplasm has no osmotic
function, the stress in the cell wall of E. coli should be about the same as that
in B. subtilis.

D. ENVIRONMENTAL EFFECTS

The chief effect is, of course, that due to water; the variation in properties
with the degree of hydration has already been described. It is to be expected
that the ionic environment might also have a profound effect. At first, this
was thought to be the reason for differences in cell-wall properties measured
using standard threads and unwashed threads, to which residual culture
medium adheres (Fig. 1). The first property measurements were made on
such threads (Thwaites and Mendelson, 1985, 1989; Mendelson and
Thwaites, 1989b) and shows a similar variation with humidity to those
derived from standard threads. With fill data sets available for both types, it
is clear that all of the properties of unwashed preparations are the same as
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 197
the standard properties at about 18% higher relative humidity. Figure 2
shows that this relates not just to the representative parameters but to the
whole stresshtrain curve. Although the ions in the residual culture medium
are a possible cause, the closeness of the agreement suggests two more likely
causes. Firstly, changes in mechanical properties may be related to water
content rather than water activity as given by relative humidity. This would
require two quite different sorption isotherms, which seems unlikely, but
measurements of this kind have not been made. Secondly, and more likely,
since the dried culture medium is very hygroscopic, the cell walls, which are
surrounded by it within the thread, experience an atmosphere which is more
humid than the bulk atmosphere. Neither of these hypotheses explains the
fact that properties obtained from the differently treated threads, when
extrapolated to 100%relative humidity, differ by substantial factors (ten for
the moduli). If 100% humidity were the equivalent of full hydration, this
would not be so. It may not be equivalent for measurement of mechanical
properties. It certainly cannot be for diffusion, and it is only by diffusion
that culture medium is removed in order to make standard threads. A
further possibility which does not have this difficulty, but which other-
wise is somewhat unlikely because the strain is fyt-, is that there is a small
amount of autolysin present in the residual growth medium and that the
water activity at high relative humidity is sufficient for some enzyme activity.
This might explain why properties for the differently treated threads are not
the same at 100% relative humidity. However, another mechanism would
be needed to explain the differences in the middle range of relative
humidity.
The effects of ions on mechanical properties have been demonstrated by
immersing standard threads a second time, firstly in water of different pH
values. There is an increase in cell-wall ductility in the pH range from 3.3 to
9.0, but no significant changes in either strength or initial modulus (J. J.
Thwaites, U. C. Surano and A. M. Jones, unpublished observations). The
increase in ductility, however, leads to a significant decrease in average
modulus as the pH value is increased. The effects of inorganic ions have
been shown by a second immersion in solutions of known concentration and
the drawing out of salt-washed threads. Their behaviour, like that of
standard threads, varies with the humidity of the atmosphere. Apart from
the response at relatively low humidity levels, the general effect is to make
walls more ductile, less strong and less stiff (Fig. 5 ) . However, the effects
cannot be described by a simple shift in relative humidity, which makes it
unlikely that the residual culture-medium effect already discussed is due to
ions. Ions probably influence mechanical properties through changes in
polymer conformation due to ion binding. It is known to diminish the
electrostatic forces between charged groups (Marquis, 1968) and it is
198 J. J. THWAITES AND N. H. MENDELSON

Salt concentration (MI

FIG. 5. Initial (Young's) modulus of cell walls in threads of Bacillus subtilis FJ7 as a
function of salt concentration at three values of relative humidity (mean k SD).
Treatment was with sodium chloride, shown by V ,or ammonium sulphate, shown by
V. Controls are shown by A . Values on the graph indicate percentage relative
humidity.

thought that, for concentrations greater than about 0.2 M, polysaccharides


behave like uncharged polymers. The effects of salt on mechanical
properties increase in magnitude as salt Concentration increases, with some
indication, certainly, in the strength and extensibility changes, of saturation
at about 0.5 M.
An attempt has been made to change peptidoglycan structure in a more
obvious way by treatment with lysozyme both in the culture prior to drawing
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 199
threads and by washing standard thread in lysozyme solutions. There is no
difference in any cell-wall property, including relaxation behaviour (see
below), between treated and standard threads. This is so, even though the
lysozyme concentration was only just below the value at which it became
impossible to draw threads. The threads are certainly changed in appearance
(Fig. 1).There are distinct signs of preferential attack near the cross walls in
the cellular filaments and even of cell separation. It is clear that adhesion of
filaments to each other is very strong and is, even with some cell separation,
enough to maintain threads as integral fibres during testing.
Although the properties of cell walls vary with relative humidity, with
salt concentration and because of the presence of residual culture medium,
there is one constant feature. Their strength is related to their extensibility
always in the same way; Fig. 6 shows this. Allowing for the spread, the points
lie on the same “failure locus”, whatever their treatment. This indicates

FIG. 6. Tensile strength of walls in threads of Bacillus subtilis FJ7 as a function of


extensibility (breaking strain). The response of a standard wall is shown by A ,
unwashed by 0, salt-treated by V and lysozyme treated by A.
200 J. J . THWAITES AND N. H. MENDELSON

that, despite the amount of molecular flexibilty that must exist in the wall
(see Section IV.F), the average amount of disentanglement required in
order that a given number of bonds be broken is the same whatever the
molecular conformation might be.

E. VISCO-ELASTICITY

The properties of cell walls depend upon the speed of deformation; at higher
extension rates they appear to be stiffer and more brittle, and vice versa.
This is a consequence of their visco-elastic nature, which is common to most
other polymers. Visco-elastic parameters can be measured in several ways,
the simplest being by observing either (a) creep, i.e. continuing extension at
a constant stress, towards, after very long times, an asymptotic final
extension (Thwaites and Mendelson, 1985) or (b) stress relaxation, i.e.
reduction in stress at a constant extension towards an asymptotic relaxed
stress. (Thwaites and Mendelson, 1989). The decay in stress with time
explains why cell walls appear less stiff when extended more slowly. A
typical stress-relaxation curve for a cell wall resembles, but only superficially,
an exponential decay. However, there is no single time constant; it is as if the
time constant were getting longer with time! An initial value of this
characteristic time can be estimated provided a good estimate of the fully
relaxed stress can be obtained. For cell walls an estimate has been obtained
after 30 minutes. The initial time constant lies in the range 21-84 seconds,
depending on humidity (J. J. Thwaites and U. C. Surana, unpublished
observations). The relaxed modulus, derived from the relaxed stress and the
imposed (constant) strain, also varies with humidity, with a minimum value
of about 0.2 times the initial modulus, at about 75% relative humidity, which
corresponds to the minimum characteristic time. The full range of relaxed
modulus as a function of initial modulus is shown in Fig. 7. The corresponding
data for cell walls in threads with residual culture medium also lie essentially
on the same locus, i.e. within the scatter.
If the stress-relaxation curves are plotted against the logarithm of time, it
becomes clear that 30 minutes is not long enough to achieve a true relaxed
modulus. (Thwaites and Mendelson, 1989). The data suggest a very wide
distribution of relaxation time constants, a strictly phenomenological
concept (Ferry, 1980, Chapter 3). For many polymers, there is a temperature-
time equivalence which allows a master relaxation curve to be drawn for the
whole range of times (and temperatures). For some, e.g. nylon, there is a
humidity-time equivalence. Using this idea and a plausible shift factor, data
for cell walls can tentatively be plotted on a master curve extending over 22
decades of time (Thwaites and Mendelson, 1990). The form of the curve
suggests that a fully relaxed modulus might only exist in the state of full
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 201

1 I I I I
0.02 0.05 0.1 0.2 0.5

Initial modulus (GPO)

FIG. 7. Relaxed modulus of cell walls in threads of Bacillus subtib FJ7 after 30
minutes as a function of initial modulus. The response of a standardwall is shown by
A and that of an unwashed wall by 0. (J. J. Thwaites and U. C. Surana,unpublished
observations).

hydration. The visco-elasticbehaviour of some polymers can be described in


terms of very sharp transitions as temperature, or humidity, varies. The so-
called glass-transition temperature is often used (Ferry, 1980, Chapter l l).
Although the behaviour of cell wall changes from glass-like when dry to
rubber-like when wet, the transition is not a sharp one and it is unlikely that a
glass-transition temperature could be posited. This is a common feature of
other highly cross-linked (amorphous) polymers and, although peptidoglycan
in B. subtilis is less highly cross-linked than that of many other bacteria, it is
highly cross-linked by the standards of man-made polymers.
202 1.1. THWAITES A N D N. H. MENDELSON

F. MOLECULAR ARRANGEMENT IN THE CELL WALL

Despite the lack of measurements other than in the axial direction, there is
much evidence to indicate that the mechanical properties of the cell wall are
anisotropic. The way in which rod-shaped bacteria maintain a constant
diameter during growth suggests that the cell wall is much stiffer in the
hoop direction. The twisting-with-elongation growth pattern of B.
subtilis, established by observations of macrofibres (Mendelson et al.,
1984), suggests that the anisotropy is helical. This implies some order
in the structural polymer peptidoglycan, that is the backbones are on
average in a preferred direction. But this does not indicate a regular
structure, as has been proposed for some bacteria (Burman and Park, 1984).
Order is a feature of all amorphous polymers. The mechanical behaviour of
cell walls is in all respects just like that of amorphous polymers and suggests
no other arrangement of peptidoglycan than an entanglement network.
Where peptidoglycans differ from more well-known polymers is in being
polyelectrolytes, i.e. having a lot of interaction between charged groups.
Marquis (1968) has shown this, and it is also most likely that this is
why mechanical properties are so much influenced by ions. The conforma-
tional changes required in order to effect this influence require a lot of
molecular flexibility which would not be possible with a regular molecular
arrangement.

V. Cell-Wall Models

A. THE AIMS OF MECHANICAL MODELLING

The general purpose behind wall modelling is to investigate states of wall


deformation and stress with a view to explaining how cell shape is
determined and maintained in terms of surface assembly and mechanics. A
subsidiary purpose is to understand how molecular structure leads to the
mechanical properties required for such behaviour. A long-term goal is to
try to determine whether, and if so how, forces generated during growth
govern other vital cell processes.
A cell-wall model is, or should be, the quantitative embodiment of some
hypothesis. Its usefulness lies in predicting phenomena that can be
measured, so that the validity of the hypothesis can be tested and ideas
modified or discarded. It should also suggest which experiments are likely to
prove most fruitful in terms of further understanding. The models described
in Sections V.D and E may, because they are couched in the language of
applied mechanics, strike microbiologists as unusual, but they result in
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 203
several predictions that can be checked experimentally as well as explaining
already observed phenomena. They also open up the possibility of
investigating the dynamics of disturbed growth and of the control processes
at work.
Some of the questions that should be addressed first are as follows:

How does a rod-shaped bacterium combine \ongitudina\ ductility


with such high circumferential stiffness?
Why do the cells of B. subtilis and, by implication, several other rod-
shaped bacteria (see Section 1II.D) twist as they grow?
Why is the rod shape so stable?
Indeed, how do different organisms maintain their characteristic
shape not only during growth but also during division?
Is growth a semi-autonomous process in which the cell wall simply
responds to the stresses within it or is it necessary for it to be under
close control by some other process?
If the latter, what kind of control signals are required? What
varying parameters might be sensed by cells in order to exercise
control?

In the past, some attention has been given to questions (a), (b), (e)
and even (d). The models described later bear on questions (a), (b), (c)
and (e).

B. GEOMETRICAL MODELS

Models which take no account of stress due to turgor, and possibly other
body forces, should not be decried; understanding geometry is a prerequisite
for good mechanical understanding. For example a rod-growth model in
which new material is added to the cylindrical surface in a helical pattern
explains in an elementary way how cell-twisting with elongation might
come about (Mendelson, 1976). It also shows that, if the rate of growth
varies circumferentially round the rod, the result is not merely that the rod
becomes curved, as it would if material were added parallel to the cylinder
axis, but that it takes up a helical shape (Fig. 8), as is observed not only in the
filaments af macrofibres but also in individual filaments (Tilby, 1977). The
argument is that, if a certain treatment produces differential growth, the
helical wall build would automatically establish a helical shape.
A variant of this model shows how cells could relate temporal order to
structure (Mendelson, 1982b). If the material is added in strings side-by-side
so that the helical pitch must change in order to accommodate them, the rate
of twisting in the rod must decrease eventually to zero. This change in speed
204 J . J . THWAITES AND N. n. MENDELSON

FIG.8. Elongation and bending of a cylinder due to the addition of new material, in
parallel alignment (above) or helically (below). Bending results from non-uniform
growth rate round the circumference. Helical addition results in relative rotation of
the ends and (with non-uniform growth rate) helical shape. From Mendelson and
Thwaites (1989a).

or even the string geometry could be sensed by the cell and used to trigger
other processes. The general idea behind such models is that, although the
rules may be established genetically and can be changed by genetic
manipulation, the actual process could thereafter be autonomous. We shall
see that this is also a possibility for more sophisticated models. Geometrical
models of this and similar kinds (see, for example, Burman and Park, 1984)
suffer not only from lack of consideration of the forces involved but also in
that they predict a very regular molecular structure in the wall. This is, as
remarked in Section 111, not detectable by ultrastructural investigation.
Furthermore, all of the evidence about mechanical properties (Section IV)
indicates the wall to be an amorphous polymer, with some order, but not this
kind of regularity.
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 205
C. MODELS INVOLVING SURFACE TENSION-LIKE STRESS

Although the existence of turgor pressure had been established for many
years (Mitchell and Moyle, 1956), the first serious attempt to relate it to
bacterial shape was made only in 1981 by Koch and his colleagues. In their
model, pressure is linked to surface curvature and a surface tension-like
stress and, as for a soap bubble, the mechanics of the situation is described
by what they call an energy-conservation equation. It is really a statement
about mechanical work, of a kind often used in applied mechanics. During
growth the situation is a changing one and, therefore, for each incremental
change, the work done by the turgor pressure in expanding the cell volume
(p dV) is equal to the work done against the surface tension in extending the
surface area (T dA). Unfortunately, the problem attempted is an explana-
tion of the shape of a coccal pole, forming from a splitting septum. It is a very
difficult one of its kind, the geometry of which almost certainly implies
continuous variation in stress over the whole surface, whereas surface
tension would be uniform. In order to make headway, artificial constraints
are introduced, such as that the material becomes “rigid” immediately it is
externalized from the septum. Not surprisingly, the fitting of the predicted
geometry to observed shape is not good. Nor is the rationalization of the
differences very convincing.
Nonetheless the basic idea is a good one. As subsequently stated by Koch
(1983), it is that growth and division are driven, not in some mysterious way
by the molecular architecture of wall insertion, but by the stress in the cell
wall due to turgor. This is of course another version of the “autonomous
process following rules” idea already referred to. Development of the idea
has, unfortunately, despite the number of papers devoted to it, not been
thorough, largely because the state of stress in walls has not been properly
analysed. The fact that the stress level normally varies over the cell surface
has been dealt with by introducing a variable surface tension-like stress
(variable T) related to localized differences in the polymerization process
(Koch, 1983). Clearly, all manner of soap bubble-like shapes can (and have
been) produced by this convenient device. The organism must of course
know what shape it needs to be and act accordingly! No doubt this is just
possible but it defies one of the basic tenets of the so-called “surface-
stress theory”, and relies on the (forsworn) mysterious wall-insertion
process.
Furthermore, it is obvious, and was well known when the ramifications of
the basic idea were first worked on (Koch 1983), that the two stress
components in the bacterial cell wall are not, in general, equal. For example,
in the cylindrical part of a rod-shaped bacterium subject to turgor pressure,
the hoop stress is twice the longitudinal stress, wheras surface tension is the
206 J. J. THWAlTES AND N. H. MENDELSON

same in all directions. Apart from this, it means that the simple work
equation, basic to the “surface-stress theory” is not applicable; more
complicated and (to some at least) well-known relations between stress and
strain must be used. This has not been done. Instead, artificial constraints
have usually been devised in order to surmount difficulties. For example, it
is alleged that cylindrical extension in the Gram-positive rod is stable
provided only that turgor pressure and surface-tension stress are constant
and that the poles are rigid (Koch, 1983). This is in fact not true; a mass
balance is also required and the rigidity of poles leads, in response to the
slightest disturbance, to “barrelling” of the cylinder; but the artificiality of
the requirement of pole rigidity is astounding. For, in other attempted
explanations of shape, the septum must split and stretch in order to take up
the required pole shape (Burdett and Koch, 1984). This shape, which is of
course not a hemisphere, nonetheless maximizes the enclosed volume (for a
given surface area) under the constraint that there is no cirumferential
expansion where the pole joins the cylinder (Koch and Burdett, 1986);
because of the rigidity of the cylinder, presumably. But one cannot argue
that cylinder rigidity maintains pole shape and at the same time that pole
rigidity is responsible for cylinder stability.
This is not to say that attempts to develop the basic idea have been
worthless. Consideration of stress levels has given a unique explanation
of why there is much less turnover in the poles of Gram-positive rods than
in the cylinder wall. The argument is based on the notion that bonds
subjected to greater stress are more readily cleaved. The wall, in changing
from a flat septum to a curved pole, is subject to stress which varies little
through its thickness whereas, because of upwelling during growth, the
stress on the outside of the cylinder is much greater than the average stress
and, therefore, much greater than the pole stress. The “split and stretch”
idea is also attractive, but it applies to somewhat intractable problems.
Nascent poles are, in general, not the same shape as completed poles. It is
often difficult to solve the equations that result from analysis of deformation
from known geometry, let alone ones in which the geometry is itself an
unknown.
Recently, attention has turned to possible cell-wall material properties,
not by the experimental method of Section IV, but again in terms of the
relation of stress to the breaking of bonds, (Koch, 1988) and to the cracking
of cell surfaces. The fact that the exterior surface of the walls of Gram-
positive bacteria is extremely rough presents the modeller with a problem,
but there is little point in attempting to analyse a cracked cell surface if the
corresponding whole surface cannot be properly dealt with. Our inference,
on the basis of much experimental work on the twisting of macrofibres and
individual filaments (see, for example, Mendelson 1976, 1982a; Mendelson
MECHANICAL BEHAWOUR OF BACERIAL CELL WALLS 207
et al., 1984), that B. subtilis cells twist as they elongate appears finally to
have been agreed. It is natural to try to explain this by analysis of the stresses
in the cell wall. We do so below. Koch (1989) makes the attempt by arguing
that the high levels of stress on the outside of the cylinder wall must lead to
cracks. There is no experimental evidence for this but, since the stress must
be non-uniform, it could be so. He then proposes, without analysis, that a
subtle combination of hoop stress and high longitudinal strain somehow
produces helical cracks. This proposition has no foundation in mechanics
whatsoever. The helical cracks are further supposed to propagate and, in so
doing, in an unexplained manner, to produce rotation of one end of the cell
relative to the other. This is a poor hypothesis; it is apparently aimed at
explaining one already known phenomenon. It leads to no prediction
concerning, for example, the amount of twist, nor even its direction. It
cannot be applied to explain the wide spectrum of twist, both left and right
hand, that has been observed. This would still be true even if there were any
convincing stress analysis relating to the alleged cracks.
In fact, the major shortcomings of attempts to date to deal with cell-wall
stress are that the problems are not properly posed mechanically, and that
there has been no proper stress analysis. Some of the ideas are good and
their limitations can be explored by a relatively simple analysis of stress. The
same approach would soon lead to the discarding of others. This, we believe,
is “what the microbiologist can learn from the textile engineer” (Koch,
1988) or, indeed, any mechanical engineer. It is what we attempt to do in the
following sections. We choose as a first example one of the less difficult
geometrical problems.

D. A MODEL INVOLVING ANISOTROPIC CELL-WALL MATERIAL

1. Analysis of Stress
The simplest geometrical shape, apart from a sphere, in which stress can be
analysed is a circular cylinder. Since rod-shaped bacteria are essentially
cylinders with polar caps this is a good starting point. It is well known that
the hoop stress, o h , due to internal pressure in the wall of a closed cylinder is
twice the longitudinal stress q,i.e. the stress on the faces of axial planes in
the wall material is twice that on the faces of radial planes (Fig. 9). This
follows only from consideration of the equilibrium of forces. It does not
mean that the hoopwise deformation is twice as great as the lengthwise
deformation, even for an isotropic elastic material. This is because the
elastic strain, i.e. extension divided by length, in each direction depends not
only on the stress component in that direction but also on a fraction
(Poisson’s ratio) of the component in the perpendiculardirection. Nonetheless,
208 J. J. THWAITES AND N. H. MENDELSON

less, for an isotropic material, the two deformations are of the same order.
This is clearly not the case for rod-shaped bacteria in which the cylinder
extends by growth while the diameter remains remarkably constant. In
order to explain this, indeed to investigate theoretically Koch’s proposition
that growth is a semi-autonomous process in which the cell wall simply
extends in response to stress, the material must be highly anisotropic, being
much stiffer in the hoop direction than in the longitudinal direction. It is
likely that the stiffer direction represents the average orientation of the
peptidoglycan backbone.

FI i . 9. Diagram illustrating longitudinal stress 0, and hoop stress ohin the wall of I
circular cylinder under internal pressure p.

This is not all, however. Twisting with growth has been observed in
filaments of B. subtilis and other species, and measurements of twisting rate
have been deduced from experiments on macrofibres (Mendelson et al.,
1984; see also Section 1II.D). This too may be an essentially mechanical
response of the cell wall to stress. When a helically reinforced pipe is
subjected to internal pressure, it twists, so that it is sensible to consider a
model in which the stiff direction in the wall material is inclined slightly to
the cylinder hoop direction. This has little effect on hoopwise stiffness, but
does allow us to investigate whether the proposition about twisting with
growth is theoreticaly possible. It is certainly a more plausible proposition
than either of those already described.
Finally, the state of stress in the wall need not be due solely to turgor
pressure. The wall is, in general, quite highly charged (Marquis, 1968).
Much of the charge lies on teichoic acids and, in B. subtilis, teichoic acid-
deficient mutants are frequently misshapen (Cole etal., 1970).This suggests
that electrostatic forces are significant mechanically. Even if only a tiny
fraction of wall charge remains unneutralized, the resulting electrostatic
repulsive force could engender stresses of the same order as those due to
turgor (see below). The effect would not be the same as that of an internal
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 209
pressure nor would it necessarily be the same at all points in the cylinder
wall, even if the net charge were to be uniformly distributed, which, for B.
subtilis and most likely for other bacteria, may not be so (Sonnenfeld et al.,
1985). The effect would be to add substantially more to the longitudinal
component of wall stress than to the hoop component, so that the combined
effect of turgor and electrostatic repulsion could produce a longitudinal
stress greater than the hoop stress (J. J. Thwaites, unpublished observation).
Thus, an appropriate “thin-shell’’ model of the cylinder wall is one in
which the material has helical anisotropy with the major axis (the stiffer
direction) inclined at a small angle a to the hoop direction, and where the
cylinder is subjected to internal pressure, p l , together with a longitudinal
tension, which it is convenient to think of as a pressure,p2, acting only on the
closed ends of the cylinder. The hoop and longitudinal stress components
are then given, respectively, by

and

where h is the ratio of wall thickness to cylinder radius. In order to obtain the
resulting components of deformation, the stress components on planes
perpendicular to the material axes must be found, and the resulting
deformations transformed back in terms of cylinder axes (Thwaites, 1977).
For a small angle, a, this is a simple matter. The wall-material properties are
both non-linear and time-dependent (see Section IV), but the present state
of knowledge would not justify the effort required in order fully to model
these features. Instead, the material is here assumed to be either a linear
elastic solid, i.e. one obeying a generalized Hooke’s law, or a linear viscous
gel of high viscosity. The property values chosen are averages of those
measured or, where not measured, plausible estimates. True cell-wall
behaviour should lie somewhere between those predicted by these two
models.
For a linear elastic material, the strains in the hoop and longitudinal
directions, &h and differ negligibly from what they would be if the material
axes co-incided with the cylinder axes. They are given by

and

where E is the elastic modulus in the less stiff principal direction for the
material, i.e. approximately along the axial direction, e is the ratio of E to
the elastic modulus along the stiff material axis, and v is Poisson’s ratio.
210 1. 1. THWAITES AND N. H. MENDELSON

There is also a shear strain which results in a relative rotation of the


cylinder ends through an angle 8 given by (J. J. Thwaites, unpublished
calculation)

where a and 1 are the cylinder radius and length, respectively, and G is the
shear modulus for the material. There are similar relations for the viscous-
gel material, giving strain rates and rotation rate in terms of stress and
corresponding viscous-material parameters.
The measurements of the mechanical properties of walls of B. subtilis
described in Section IV show that the average modulus, E, must be less than
the initial modulus. For model purposes it is taken to be 20 MPa; p 1 is taken
to be 1 MPa. This is less than generally accepted for turgor pressure but,
since the longitudinal stress must not exceed the observed breaking stress, it
is a reasonable value. Thus, for a thicknesshadius ratio of 0.1, q,is 10 MPa.
No measurements of transverse modulus or Poisson's ratio have been made
but, for other highly anisotropic polymers, v is small, of the order of 0.1.
Also, there is good reason to believe that, like many other polymers,
cell walls are approximately incompressible (see Section V.E), in which case
the material properties are not independent and e = 2v (Thwaites, 1977).
The transverse modulus is therefore taken to be 100 MPa. Using these
values

and

where x = q h h = 0.5(l+p2/pl). For equal pressures the stress components


are equal, &h = 0.05 and = 0.45;forpz = 1.4p1, i.e. o1= 1.20h, the strains
are 0.04 and 0.55, respectively. The longitudinal strains are, not sur-
prisingly,of the order of those measured (Section IV). The hoop strains of a
few per cent indicate why the cell diameter can remain approximately
constant.
Corresponding results giving strain rates for a viscous material are
obtained by taking a viscosity (instead of E) of 20 MPa h. This is a reasonable
value representing the stress-relaxation behaviour as measured. The
longitudinal strain rate El is then 0.55 h-', which implies a length doubling
time of about 75 minutes, i.e. of the same order as B. subtilis growing at
20°C. Strictly, these figures relate only to a cylinder consisting of a given
amount of material and take no account of addition of new wall or of
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 21 1
turnover. Nonetheless, they are substantially correct when the model is
modified in order to take account of these effects.
In the above analysis, the choice of pressure ratios (p2/p1)of 1 and 1.4,
respectively, was somewhat arbitrary, but the size of p2 implied is not
unreasonable. The extra longitudinal pressure for a cylinder with uniform
surface-charge density, q , depends on the cylinder lengthhadius ratio; it is of
the order of q2/2c0 where c0 is the dielectric permittivity. This is 1.4 MPa
when q = 0.45 C m -2, i.e. for walls of B. subtilis about 11 mEq 1-’. Thus,
taking Marquis’ value for charge (derived from Doyle et al., 1980),
only about 5% of wall charge needs to be unneutralized to achieve
comparability with turgor. This does not appear to be unreasonable, but the
question of charge neutralization in relation to cell-wall stress needs further
research.
The inclusion of electrostatic repulsion provides an explanation of a
hitherto puzzling finding, namely that, when turgor pressure is removed,
cells (strictly cellular filaments) shrink in length much less than one would
expect (Koch, 1984; Koch et al., 1987). Using the property values already
referred to and taking p2 to be 1.4 MPa, removal of turgor (pl = 0) lowers
the value for cl from 0.55 to 0.30, i.e. a shrinkage (based on the original
length) of 16.1%. If p2 is taken to be 1 MPa, the value would be 17.2%.
These values are close to the shrinkage of 17% observed. It is true that the
observations were made on filaments of E. coli but, as remarked in Section
IV.B, it is not unreasonable to suppose that the stress in its wall is of
comparable magnitude to that in the wall of B. subtilis. A shrinkage of the
same order should be observable if p 2 were to be removed, for example by
changing the pH value. Shrinkage of roughly 10% was observed during the
experiments determining the effects of neutral salts (at 50 mM) on
mechanical properties (Section 1V.B.3), but no proper measurements were
made. Small changes in turgor or electrostatic repulsion could not be
detected in this way because their effect would be difficult to measure with
any confidence. Changes in twist, however, are much more readily
observable.

2. Cell-Wall Twist
The models predict (equation (3)) that, even in the absence of external
forces, the cell should twist under the influence of p1 andp2. The unknown
factor is the shear modulus, G, which has yet to be measured for B. subtilis.
For very open networks such as the peptidoglycan network in cell walls
(Marquis, 1968; Ou and Marquis, 1970), it must be much less than Young’s
moduli. This is true of materials in sheet form and also of many textile
structures (including non-wovens) which are good analogues of open-
212 J. J. THWAITES AND N. n. MENDELSON
network sheets. For the purposes of calculation, G is taken to be 2 MPa. A
value of 0.1E for G is somewhat arbitrary, but many of the following
conclusions do not depend on this particular value. If the shear modulus
were larger, relatively small changes in the ratiop2/pl would lead to the same
conclusions. Using this and values already used, the twist is given by
aell = ( 3 . 9 ~- 4.7)~
which is zero forp, = 1.4pl. Moreover, for changes in p 2 of only +lo%,ae-1
takes values of f d4,i.e. the twist changes sign without the need for a to do
so. It seems distinctly possible that the positive and negative (right- and left-
hand) twists in cells, inferred from observations of macrofibres, are a direct
result of helical anisotropy in the cell wall, and that the handedness is
controlled in part by quite small changes in the net unneutralized charge, but
not necessarily by changes in the handedness of the material anisotropy.
Rapid changes in twist, i.e. non-growth-related changes, can be induced by
lysozyme attack, by changing the pH value and by neutral salts (Mendelson et
al., 1985). Lysozyme achieves its effect by cleaving certain glycosidic bonds of
the peptidoglycan backbone, in model terms by decreasing the transverse
modulus, i.e. increasing e. Corresponding changes in 8 follow from
d9 2ohla
_ -
- -
de Ea
so that, provided a is positive, i.e. the helix of material anisotropy is right
handed, twisting towards the right hand is predicted whatever the existing
twist. This is as observed (Favre et al., 1986).
Rapid changes in twist ought also to result from removal of turgor
pressure. They were not reported by Koch (1984), but this may be only
because it is difficult to observe such twisting in relatively short single
filaments. However, changes in turgor should produce observable twisting
in macrofibres: e.g. for the property values used above, a 20% decrease inpl
should give an increase in aW1 of about 0.53~.Attempts to observe such
changes by using sucrose to change medium osmolarity have been inconclu-
sive. A decrease in electrostatic repulsion due to charge neutralization, by
changing the pH value for example, should produce positive twisting of the
same order as that produced by reducing turgor. Unfortunately, lowering
the pH value not only decreases the unneutralized charge but, since the cell
wall becomes less ductile, it increases the average modulus. This, by
equation (3), has a contrary effect so that it is not possible to make a good
prediction. However macrofibres, and by implication cell walls, do twist
rapidly when the pH value is decreased (Mendelson et al., 1985). They also
twist when treated with certain ions. The theoretical situation is similar in
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 213
this case. Mechanical properties are changed (Section IV.D), perhaps by a
different mechanism involving water ordering, and, presumably, charge
neutralization takes place also. Until more is known about the relative
strength of these contrary effects, there is no point in speculating about the
predictive value of the model in these situations.
The model does, however, show clearly that it is not necessary to change
the helix handedness of material anisotropy in order to achieve a reversal in
cell-wall twist. Such a mechanism, though quite possible during growth and,
therefore, slow, would be an unlikely one for the rapid changes observed. In
the same way, the continuous twisting observed during growth need not be
related to different degrees (or even helix handedness) of anisotropy. The
twist could be controlled by factors which affect the material properties and
the unneutralized charge only. Strictly, continuous twisting can only be
described by a model which takes account of new wall insertion, but this
effect can be thought of as follows. Each new cylinder element (of length l)
extends by l q and one end moves circumferentially, in the cylinder surface,
relative to the other by a distance ue. The helix angle of continuous twisting
is therefore uO/lq. For a viscous material model, even without new wall
insertion and certainly with it, this helix angle is u6/ltl,where the dots denote
rates of change. The helix angle y~of twisting is, therefore, given by the same
expression, derived from equations ( 2 ) and (3):

_y~ -- (x-l)g+2(e-vx) - 2 (4)


a (x-v)
where g and e are corresponding ratios of either moduli or viscosities. For
the parameter values already used, w changes from approximately - d 2 to
+ d 2 for a +lo% change inpz about the value 1 . 4 ~Since ~ . the twist angles
inferred from macrofibre measurements on B. subtilis mutant FJ7 are in the
range -5" to +6" (Mendelson et uf., 1984), the helix angle of anisotropy
would not need to be more than about 10" if this were the cause of twist
changes. But the variety of influences on macrofibre twist, and by
implication cell-wall twist, suggests that changes in material properties, and
even in a, must also be involved. (Mendelson and Thwaites, 1988).
A final quantitative prediction concerns the forces needed to suppress
twisting during growth and bears on the argument (see Section 1II.D) that
sufficient torque is engendered by contact between filaments for them to ply
and eventually form macrofibres. For a twist 0/1, the torque required is
approximately 21tGu~he.l.With values already used and a equal to 10" ,this
torque could be produced by a tangential force at the cell surface of about
0.8 X lo-' N, i.e. a little under 1 pg weight. Whether or not this is a
reasonable force in bacterial terms must await suitable experiments on
inhibition of twist.
214 J . J . THWAITES AND N. H. MENDELSON

In summary, a “thin-shell’’ model of the cylinder wall of rod-shaped


bacteria, on the assumption of helical anisotropy and using known material
properties or plausible estimates, can be used to predict quantitatively, or
explain, changes in length and diameter and also rapid twisting due to the
following: (i) changes in turgor pressure; (ii) changes in effective wall charge
engendered by variation in p H value or by ions; (iii) changes in material
properties, due, for example, to enzyme attack.
Several of these phenomena have yet to be measured. In addition, it can
be adapted to predict the rate of twisting-with-growth and the way in which it
depends on (i) to (iii). In particular, it explains why it could be possible to
have changes in twist handedness without needing a change in the material
anisotropy. Strictly, this adaption needs justifying by what follows.

E. A CELL-WALL GROWTH MODEL

Although a “thin-shell” model can be used to describe rapid changes in


response to various stimuli and might be adapted, with much greater
geometrical sophistication, in order to describe aspects of growth in which
material is inserted in given bands, it cannot describe fully what might be
happening in walls in Gram-positive cells. It has been well established for
many years that growth is accompanied by upwelling of wall material from
the plasma membrane and by turnover of material at or near the outer
surface (Archibald, 1976; Archibald and Coapes, 1976; Pooley, 1976 a,b).
This involves increasing deformation during upwelling, with consequent
changes in the state of stress. One deals with this by means of a so-called
“thick-shell’’ model, which is not to imply that the wall thicknesskell radius
ratio is any greater, but simply that variations through the wall are taken into
account.
The basis for such a model is the observation that wall material in B.
subtilis is inserted uniformly over the inner surface of the cylindrical portion
(though not also at the poles) and that it wells up uniformly too (Pooley,
1976a; Merad et al., 1989), so that material at a given radius in the wall is all
of the same age. While this may not be precisely true, there is sufficient
averaging over the cylinder surface for it to be a good basis. The model is
made tractable by the further assumption that the material is incompressible,
i.e. that it does not change in volume due to application of stress. As
remarked earlier, this is true of many polymers, and it is argued that it must
be approximately so for cell walls, for they contain a large proportion of
water, the bulk modulus of which (about 2 GPa) is so very much greater than
the measured tensile modulus (and by implication the transverse modulus).
Volumetric strains due to the levels of stress inferred (and those permissible
in measurements such as those described in Section IV) would be minute.
MECHANICAL BEHAWOUR OF BAClERIAL CELL WALLS 215

FIG. 10. Diagram of the cross-section of a cylindrical cell wall showing mass flow.
Material initially inserted at the membrane (radius a ) is, after time t , at radius r.
Insertion rate: h per unit volume; shedding rate (radius b): p per unit volume.

So we consider the material that was initially inserted at the membrane


(radius a in Fig. 10) to be, at a time t later, at radius r. The volume occupied
by a length 1 of material inserted during that time (shaded in Fig. 10) is
increasing at a rate R d(? - a ')f/dt. This must equal the rate of insertion of
new material which is taken to be h per unit wall volume. The resulting
differential equation is
2rt - 2ad + (3 - a2)q = (b2 - a2)h (5)
where q is the longitudinal strain rate ill and dots denote rates of change.
Note that q is the same for all values of r, but is not necessarily constant in time.
There is a similar equation for the whole thicknessof the wall, which shows that
the proportional rate of increase of the wall cross-sectional area, x(b'-a'), is
equal to h-p-q, where pis the rate of wall turnover per unit volume. There is
an obvious balance when the rate of insertion minus the rate of loss due to
turnover is equal to the rate of volume increase due to extension.
Equation ( 5 ) allows expression of the hoop strain rate, f / r ,as a function
of r. The hoop strain, In (du), can be calculated directly; it increases from zero
at the membrane to approximately h (the thicknesdradius ratio) at the outer
surface. The longitudinal strain is also zero at the membrane, but increases
to In (l+q/p) at the outer surface. Since, particularly in lyt- mutants, p must
216 I. J. THWAITES AND N. H. MENDELSON

be much smaller than q, this implies much larger strain than can be borne by
the cell wall in experiments (Section IV).
The basis for stress analysisfor this model, the details of which will appear
elsewhere (J. J. Thwaites, unpublished observation), is that there are
equations of the same form as equations (2) and (3) relating strains and twist
(or their rates of change) to stress at a general radius r. Average values of
longitudinal and hoop stress, which are identical with those used in the thin-
wall model, are found by integration over the whole wall cross-section, i.e.
with respect to varying r. The variation in stress with radius is then given by
equations analogous to equations (2). It is substantial for the elastic-material
model. At the membrane, all stress components are compressive and equal
to the pressure pl. This state is called “hydrostatic”. Both the hoop and
longitudinal stresses become tensile and increase rapidly with radius,
reaching values at the outer surface which are several times the average
values. This is illustrated, for the case p2 = 1.4pl, in Fig. 11. The viscous

to” I

Rod 0 Axis Rod b

FIG. 11. The state of stress at the membrane (radius a ) and the outer surface of the
wall (radius b ) , according to wall-growth models for (i) elastic material and (ii)
viscous-gel material. The figures are the ratios of the stresses to the average hoop
stress o h .
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 217
model produces a quite different stress distribution with almost no variation
in longitudinal stress with radius, and hoop stress slowly decreasing from
inside to outside (Fig. 11). A point to note is that wall material, although it
starts without strain, is not stress-free at any time during its journey through
the wall. Given the measured properties for walls of B. subtilis, the
predicted level of stress at the outer surface, for an elastic material, is greater
than can be borne in our experiments. Some of the material would be
fractured, so that average values of moduli would effectively be lower than
those used in the calculations. It is not at all clear, despite assertions to the
contrary, that this kind of fracture does occur (Koch, 1989). If it does, the
mode of fracture would certainly not be as Koch suggests. In any case, the
material is visco-elastic; the levels of stress must lie between those predicted
for elastic and viscous materials. A visco-elastic model is the subject of
current research.
Other conclusions to be drawn from analysis of wall-growth models are,
firstly, that equations (2) holds provided the stresses are the average stresses
given by equations (1) and the strains are average strains (or strain rates).
For the viscous-material model, this gives a direct expression for the
longitudinal strain rate, i.e. the rate of growth, in terms of average
stresses, which is the same as predicted by thin-shell modelling, i.e.
q = (0, - v o h ) / E rwhere
, E' is the viscosity. The average longitudinal strain
for the elastic material is, however, because of the relation between radius
and time, a more complicated expression which yields quite a different
answer for the overall longitudinal strain rate. It is interesting to observe
that a purely elastic-material model, because of continuous material
insertion and turnover, can predict a strain rate, but a full visco-elastic
treatment is required in order to obtain a better expression for q. Secondly,
provided the helix angle of anisotropy does not vary much from the inside to
the outside of the wall, the helix angle of twisting-with-growth is given by
equation (4)with, since the material has been taken to be incompressible,
e = 2v, so that conclusions relating to twisting can all be obtained by
essentially thin-shell arguments. There is little experimental evidence of any
kind concerning material anisotropy in the wall, so further analysis does not
seem to be appropriate. It should be stated, however, that polymeric
materials, when subject to the levels of strain envisaged in the model,
usually become more highly oriented. For example, a high degree of
anisotropy is achieved in fibres by drawing.
A further conclusion follows from the inclusion in the upwelling equation
(5) of a term for variation in overall cylinder radius with time and derivation
of an expression for the rate of change of wall cross-sectional area. This is
that the wall thicknesdradius ratio, h, tends with time, after a disturbance,
towards an equilibrium value 2v(20h-o,)lEr(h+p). This, of course, alters
218 J. J. THWAITES AND N. H. MENDELSON

with changes in the ratio p 2 / p I ,but the changes need not be large. Taking p
= 0.2h (Pooley, 1976a), and giving h its equilibrium value, a +lo% change
in p 2 of about 1 . 4 produces
~ ~ approximately f13% change in h about the
value 0.1. The ratio h tends towards its equilibrium value with a time
constant l/(h+p). If the system is in balance in terms of material insertion, this
is of the order of the length-doubling time which is long in terms of maintaining
control over the wall. Nonetheless, h tends towards its equilibrium value
automatically without the need for a control signal. Consequently, the
cylinder radius is under control provided the wall area is constant, i.e.
provided h - p = q. Maintaining this material balance is a matter requiring a
control signal. One might speculate that the cell senses stresses in its wall, or
strain rates, and controls the rate of insertion of wall accordingly.
To summarize, a wall-growth model in which the stress analysis given in
Section V.D is applied to the wall of Gram-positive bacteria, taking account
of upwelling and turnover, first of all confirms the twisting-with-growth
conclusions obtained with a thin-shell model. It also can be used to predict
the dynamics of change during growth, in length, diameter, wall thickness
and twist, due to the following: (i) changes in turgor; (ii) changes in effective
charge (variation in p H value; ions); (iii) changes in material properties
(e.g. enzyme attack); (iv) changes in material insertion and turnover rates.
Again, many of these phenomena have yet to be measured. In addition, it
predicts parameters that cannot be measured, but are nonetheless useful
conceptually, such as stress distribution through the wall. Possible future
developments include: (a) better modelling of the wall material, e.g.
including compressibility and variations in properties through the wall; and
(b) quantitative investigation of control mechanisms for maintaining
diameter and wall thickness. The more difficult problem of non-cylindrical
shape, in particular its establishment during division, should also be
attacked.

VI. Conclusions

In the preceding sections, we have attempted to show the value of


investigating the mechanical behaviour of bacterial cell walls and also how it
can be done both experimentally and theoretically. We now suggest the
directions that future research should take. Concerning material properties,
the anisotropy of walls needs to be examined. Shear behaviour will be
measured in torsional experiments on bacterial threads, but the more
difficult measurement of transverse modulus should also be attempted. A
fuller investigation of visco-elasticity in walls is needed, including the effects
of temperature. The effects of variations in wall polymers should be
MECHANICAL REHAVIOUR OF BACTERIAL CELL WALLS 219
examined by using treatments that modify them in known ways structurally
and possibly by using walls of different polymeric composition as a result of
physiological or genetic manipulation. Fully hydrated material needs to be
tested; this will require more sensitive instrumentation, for example to
experiment on the mechanical properties of macrofibres, possibly in vivo.
Single-cell measurements will not be possible in the near future.
Mechanical modelling suggests experiments involving dynamic changes in
the observable geometry of live cells, such as wall thickness, cell length and
diameter, and twisting rate in some cases. How these change in response to
stimuli that affect the rate of insertion of new wall material, for example, or
that change the mechanical properties in known ways, should be measured.
These experiments include the possibility of observing the effects of
upwelling of new material, following an abrupt change, of different
properties, as with reversal of twist handedness in macrofibres following a
rapid change in environment. Other experiments involve growth under
constraint. The aim should be to determine whether an imposed deforma-
tion could be set, i.e. that it would persist when the constraint was removed.
The implication would be that the growth pattern had changed owing to the
imposition of stress and that the cell-wall material had acquired a memory of
form even though this had been externally imposed.
Theoretical modelling iself, after improvements suggested in the last
section, should be aimed at the determination of shape. The difficulties of
doing so should, however, not be underestimated. The problems of working
with an unknown geometry, unknown that is in the sense that it is what is to
be determined, are mathematically difficult, but not intractable. Models
involving possible feedback links, using mechanical signals such as wall
stress to control, via plausible biochemical means, such parameters as the
insertion rate of new material and even, to some extent, its properties (e.g.
by variations in cross-linking) should also be investigated. Models of a
somewhat different kind should aim at the molecular organization of the
cell-wall polymers and the way in which the wall is assembled.
Finally, it should be emphasized that all of the tools of genetics,
biochemistry and structural biology can be applied in support of mechanical
investigations, by providing controlled variations against which changes in
mechanical behaviour can be evaluated. Once cell-wall mechanical behavi-
our is understood, its role in the regulation of cellular processes can be
examined. Micro-organisms successfully integrate such processes and
mechanics. We should try to do likewise.
220 J. J. THWAITES AND N. H. MENDELSON

REFERENCES

Archibald, A. R. (1976). Journal of Bacteriology 127, 956.


Archibald, A. R. and Coapes, H. E. (1976). Journal of Bacteriology 125, 1195.
Archibald, A. R . , Baddiley, J. and Heckels, J. E. (1973). Nature New Biology 241, 29.
Bartlett, D., Wright, M., Yayanos, A. A. and Silverman, M. (1989). Nature 342, 572.
Bayer, M. E. (1979). In “Bacterial Outer-Membranes: Biogenesis and Function” (M.
Inouye, ed.) pp. 167-202. John Wiley, New York.
Belas, R., Simon, M. and Silverman, M. (1986). Journal of Bacteriology 167, 210.
Beveridge, T. J. (1981). International Reviews in Cytology 72, 229.
Birdsell, D. C., Doyle, R. 1. and Morgenstern, M. (1975). Journal of Bacteriology 121,726.
Boylan, R. J., Mendelson, N. H., Brooks, D. and Young, F. E. (1972). Journalof Bacteriology
110, 281.
Braam, J. and Davis, R. W. (1990). Cell 60, 357.
Braun, V., Gnirke, H., Henning, V. and Rehn, K. (1973). JournalofBacteriology 114,1264.
Burdett, I. D. J. and Koch, A. L. (1984). Journal of General Microbiology 130, 1711.
Burman, L. G. and Park, J. T. (1984). Proceedings of the National Academy of Sciences ofthe
United States of America 81, 1844.
Cantor, C. R. and Schimmel, P. R. (1980). “Biophysical Chemistry, Part I: The Conformation
of Biological Macromolecules”. Freeman, San Francisco.
Cole, R. M., Popkin, T. J., Boylan, R. J. and Mendelson, N. H. (1970). Journalof Bacteriology
103, 793.
Corner, T. R. and Marquis, R. E. (1969). Biochimica et Biophysica Acta 183, 544.
Costerton, J. W., Irvin, R. T. and Cheng, K.-I., (1981). Annual Review of Microbiology 35,
299.
Csonka, L. N. (1989). Microbiological Reviews 53, 121.
Doyle, R. J. and Koch, A. L. (1987). Critical Reviews in Microbiology 15, 169.
Doyle, R. J., Streips, U. N., Fan, V. S., Brown, W. C., Mobley, H. L. T. and Mansfield, J. M.
(1977). Journal of Bacteriology 129, 547.
Doyle, R. J., Matthews, T. H. and Streips, U. N. (1980). Journal of Bacteriology 143, 471.
Doyle, R. J., Chaloupa, J. and Vinter, V. (1988). Microbiological Reviews 52, 554.
Favre, D., Thwaites, J. J. and Mendelson, N. H. (1985). Journal of Bacteriology 164, 1136.
Favre, D., Mendelson, N. H., andThwaites, J. J. (1986). Journalof GeneralMicrobiology 132,
2377.
Fein, J. (1979). Journal of Bacteriology 137, 1141.
Fein, J. (1980). Canadian Journal of Microbiology 26, 330.
Ferry, J. D. (1980). “Visco-elastic Properties of Polymers”. John Wiley, New York.
Ghuysen, J.-M. (1968). Bacteriological Reviews 32, 425.
Gilpin, R . W. and Nagy, S. S. (1976). Journal of Bacteriology 127, 1018.
Goodell, E. W. and Schwarz, U. (1985). Journal of Bacteriology 162, 391.
Gotte, L., Mammi, M. and Pezzin, G. (1968). In “Symposium of Fibrous Proteins” (W. G.
Crewther, ed.) pp. 236245. Butterworths, London.
Hernandez-Muniz, W. and Stevens, S. E. (1988). Journal of Bacteriology 170, 1519.
Heymer, B., Seidl, P. H. and Schleifer, K. H. (1985). In “Immunology of the Bacterial Cell
Envelope” (D. E. S. Stewart-Tull and M. Davies, eds), pp. 11-46. John Wiley, New York.
Hobot, J. A., Carleman, E., Villigen, W. and Kellenberger, E. (1984). Journalof Bacteriology
141, 143.
Hudson, J. A., Morgan, H. W. and Daniel, R. M. (1984), FEMS Microbiology Letters 22,149.
Isaac, D. H. (1985). In “Polysaccharides: Topics in Structure and Morphology” (E. D. T.
Atkins, ed.), pp. 141-184. VCH, Deerfield Beach, FL.
Knaysi, G., Hillier, J. and Fabricant, C. (1950). Journal of Bacteriology 60, 423.
Koch, A. L. (1983). Advances in Microbial Physiology 24, 301.
Koch, A. L. (1984). Journal of Bacteriology 159, 914.
Koch, A. L. (1988). Microbiological Reviews 52, 337.
Koch, A. L. (1989). Journal of Theoretical Biology 141, 391.
Koch, A. L. and Burdett, I. D. J. (1986). Journal of General Microbiology 132, 3451.
MECHANICAL BEHAVIOUR OF BACTERIAL CELL WALLS 221
Koch, A. L., Higgins, M. L. and Doyle, R. J. (1981). Journalof GeneralMicrobiology 123,151.
Koch, A. L., Lane, S. L., Miller, J. and Nickens, D. (1987). Journal of Bacteriology 166,
1979.
Labischinski, H., Barnickel, G. and Naumann, D. (1983). In “The Target of Penicillin” (R.
Hackenbeck, J.-V. Holtje and H. Labischinski, eds), pp. 49-54. Walter de Gruyter, Berlin.
Labischinski, H., Barnickel, G. Naumann, D. and Keller, P. (1985). Annals of the Institute
PasteurlMicrobiology 136A, 45.
Marquis, R. E. (1968). Journal of Bacteriology 95, 775.
Marquis, R. E. (1988). In “Antibiotic Inhibition of Bacterial Cell Surface Assembly and
Function” (P. Actor, L. Daneo-Moore, M. L. Higgins, M. R. J. Salton and G. D. Shockman,
eds), pp. 21-32. American Society for Microbiology, Washington, DC.
Marquis, R. E. and Carstensen, E. L. (1973). Journal of Bacteriology 113, 1198.
Marquis, R. E., Mayzel, K. and Carstensen, E. L. (1976). Canadian Journal of Microbiology
22, 975.
Mauck, J. and Glaser, L. (1972). Journal of Biological Chemistry 247, 1180.
McCarter, L., Hilmen, M. and Silverman, M. (1988). Cell 54, 345.
Mendelson, N. H. (1976). Proceedings of the National Academy of Sciences of the United States
of America 73, 1740.
Mendelson, N. H. (1978). Proceedings of the National Academy ofSciences of the United States
of America 75, 2478.
Mendelson, N. H. (1982a). Journal of Bacteriology 151, 438.
Mendelson, N. H. (1982b). Journal of Theoretical Biology 94, 209.
Mendelson, N. H. (1988). Journal of Bacteriology 170, 2336.
Mendelson, N. H. and Karamata, D. (1982). Journal of Bacteriology 151, 450.
Mendelson, N. H. andThwaites, J. J. (1988). In “Antibiotichhibition of Bacterial Cell Surface
Assembly and Function” (P. Actor, L. Daneo-Moore, M. L.Higgins, M. R. J. Salton and G.
D. Shockman, eds), pp. 109-125. American Society for Microbiology, Washington DC.
Mendelson, N. H. and Thwaites, J. J. (1989a). Comments in Theoretical Biology 1, 217.
Mendelson, N. H. and Thwaites, J. J. (1989b). Journal of Bacteriology 171, 1055.
Mendelson, N. H. and Thwaites, J. J. (1990). Materials Research Society Symposia 174, 171.
Mendelson, N. H., Favre, D. and Thwaites, J. J. (1984). Proceedings of the National
Academy of Sciences of the United States of America 81, 3562.
Mendelson, N. H., Thwaites, J. J. Favre, D., Surana, U., Breihl, M. M. and Wolfe, A. J.
(1985). Annals of the Institute PasteurlMicrobiology 136, 99.
Merad, T., Archibald, A. R., Hancock, I. C., Harwood, C. R. andHobot, J. (1989). Journalof
General Microbiology 135, 645.
Meredith, R. (1959). “The Mechanical Properties of Textile Fibres”. North Holland,
Amsterdam.
Mitchell, P. and Moyle J. (1956). Symposia of the Societyfor General Microbiology 6 , 150.
Mobley, H. L. T., Koch, A. L., Doyle, R. J. and Strieps, U. N. (1984). Journalof Bacteriology
158, 169.
Murray, R. G. E., Steed, P. and Elson, H. E. (1965). Canadian Journalof Microbiology 11,547.
Neihof, R. A. and Echols, W. H. (1968). “Biophysical Studies of Microbial Cell Walls, Part 2,
Electrophoresis: Apparatus and Exploratory Experiments”, ONR Report 6795.
Washington, DC.
Ou, L.-T. and Marquis, R. E. (1970). Journal of Bacteriology 101,92.
Ou, L.-T. and Marquis, R. E. (1972). Canadian Journal of Microbiology 18, 623.
Pooley, H. M. (1976a). Journal of Bacteriology 125, 1127.
Pooley, H. M. (1976b). Journal of Bacteriology 125, 1139.
Roberts, J. L. (1938). Science 87, 260.
Robson, R. L. and Baddiley, J. (1977). Journal of Bacteriology 129, 1051.
Rogers, H. J. (1979). Advances in Microbial Physiology 19, 1.
Rogers, H. J., Perkins, H. R. and Ward, J. B. (1980). “Microbial Cell Walls and Membranes”.
Chapman and Hall, London.
Scheraga, H. A. (1980). In “Protein Folding” (R. Jeunicke, ed.), pp. 261-268, Elsevier,
Amsterdam.
222 1. I. THWAITES AND N. H. MENDELSON

Scherrer, R. and Gerhardt, P. (1971). Journal of Bacteriology 107, 718.


Schleifer, K. H. and Kandler, 0. (1983). In “TheTarget of Penicillin” (R. Hackenbeck, J.-V.
Holtje and H. Labischinski, eds), pp. 11-17. Walter de Gruyter, Berlin.
Schleifer, K. H. and Stackebrandt, E. (1983). Annual Review of Microbiology 37, 143.
Schwarz, U. and Glauner, B. (1988). In “Antibiotic Inhibition of Bacterial Cell Surface
Assembly and Function” (P. Actor, L. Daneo-Moore, M. L. Higgins, M. R. J. Salton and G.
D. Shockman, eds), pp, 33-40. American Society for Microbiology, Washingon, DC.
Shockman, G. D. and Barrett, J. F. (1983). Annual Review of Microbiology 37, 501.
Sonnenfeld, E. M., Beveridge, T. J., Koch, A. L. and Doyle, R. J. (1985). Journal of
Bacteriology 163, 1167.
Stock, J. B., Rauch, B. and Roseman, S. (1977). Journal of Biological Chemistry 252,7850.
Struder, R. E. and Karamata, D. (1988). In “Antibiotic Inhibition of Bacterial Cell Surface
Assembly and Function” (P. Actor, L. Daneo-Moore, M. L. Higgins, M. R. J. Salton and G.
D. Shockman, eds), pp 140-150, American Society for Microbiology, Washington, DC.
Thor, C. J. B. and Henderson, W. F. (1940). American Dyestuff Reporter 29, 461.
Thwaites, J. J. (1977). International Journal of Mechanical Sciences 19, 161.
Thwaites, J. J. and Mendelson, N. H. (1985). Proceedingsofthe National Academy of Sciences
of the United States of America 82, 2163.
Thwaites, J. J. and Mendelson, N. H. (1989). International Journal of Biological
Macromolecules 11, 201.
Thwaites, J. J. and Mendelson, N. H. (1990). Materials Research Society Symposia 174, 179.
Tilby, M. J. (1977). Nature 266, 450.
Trueba, F. and Woldringh, C. L. (1980). Journal of Bacteriology 142, 869.
Verwer, R. W. H., Beachey, E. H., Keck, W., Staub, A. M. and Poldermans, J. E. (1980).
Journal of Bacteriology 141, 327.
Walsby, A. E. (1980). Nature 283, 69.
Wamoscher, L. (1930). Zeitschrifr fiir Hygiene und Infektionskrankheiten 111, 422.
Zaritsky, A. and MacNab, R. M. (1981). Journal of Bacteriology 147, 1054.
Zaritsky. A. and Mendelson, N. H. (1984). Journal of Bacteriology 158, 1182.

You might also like