Rethinking the Philosophy of Antimicrobal Drug Design

May 1, 2002

Rethinking the Philosophy of Antimicrobal Drug Design

By Glen D. Armstrong, PhD

To cause infections, microbial pathogens must first tightly adhere to host
cell surfaces. This allows pathogenic microbes to oppose cleansing action --
such as the flushing of urine or tears, peristaltic movement of the gut or
mucociliary function of respiratory epithelial cells -- that would otherwise
quickly eliminate them from their host's body. Many pathogenic bacteria express
specialized surface proteins, called adhesins, which mediate their binding to
host cells. Most, but not all, microbial adhesins are long filamentous
structures microbiologists call pili or fimbriae. On the host cell, the complex
carbohydrate sequences of glycoproteins or glycolipids represent receptors to
which many microbial pili are known to bind. It is this binding of pili to
complex carbohydrate receptors that allows many microbes to successfully
colonize host cell surfaces. With few exceptions, microbial exotoxins bind to
host cell glycolipids, thus mediating exotoxin access to their host cell

In theory, therefore, soluble carbohydrate receptor analogs might
competitively antagonize the pathophysiological effects of exotoxins or inhibit
bacterial colonization of host cell surfaces.

The continuing rise in microbial resistance to antibiotic drugs is a major
concern to all healthcare practitioners. It has stimulated infection control
professionals and medical microbiologists to seek new ways of preventing or
treating classical and emerging infectious diseases.

The merits of education and the enforcement of infection control measures are
undisputed. In many cases, the simplest and most economical of measures,
frequent handwashing for example, have been shown to have profound positive
effects.1 Recent advances in vaccine development have contributed
significantly to infection control. The application of immunization to a
particular problem is only practical in cases where a subject's immune system is
functioning normally. In many cases, it is simply not practical to develop
vaccines for infectious diseases where the annual incidence would be less than
the anticipated rate of adverse effects in a mass immunization program.
Effective drugs will remain a last resort in cases where immunization does not
represent a viable option or when other control measures fail.

The main problem associated with current antimicrobial drugs is the huge
capacity of microorganisms to become resistant to these drugs, thereby
eventually rendering them ineffective. In the presence of such drugs, microbes
which are unaffected by them soon become the predominant members of the
population. The acquisition of microbial resistance mechanisms requires the
pharmaceutical industry to search for new derivatives to which microbial
resistance has not developed. In most cases, this enterprise involves chemically
altering the structures of known drugs or creating completely synthetic agents
with suitable antimicrobial properties. Nonetheless, the principles of drug
discovery have not changed and continue to involve the perpetual quest for
agents designed to prevent the in-vivo replication of pathogenic microbes.
Perhaps it is time to rethink the philosophy of antimicrobial drug design and
investigate the feasibility of producing agents that will not create
environments favoring the proliferation of genotypically resistant microbes.

Recent advances in the fields of bacterial genomics, proteomics and molecular
biology have allowed us the benefit of a much more thorough understanding of
pathogenic mechanisms and the intricate interplay between microbes and their
hosts.2 These studies reveal potential methods of interfering with
the mechanisms of the infection process in ways that may not promote the
proliferation of microbes insensitive to the desired inhibitory effects of these
agents. Such novel agents should probably be referred to as anti-infective
rather than antimicrobial since they will be designed to inhibit the infection
process rather than the microbes themselves.

In virtually every example, the attachment of pathogenic microorganisms or
their toxic exoproducts to host cells represents the first step in the process
leading to a clinically apparent condition. My research is based on the
observation that many pathogenic microbes or their exotoxins attach to complex
carbohydrate sequences, also referred to as glycans, displayed on host cell
surfaces.3 These attachment processes are highly specific, usually
involving a microbial protein which recognizes a specific host cell complex
carbohydrate sequence or sequences.4 It is conceivable that soluble
carbohydrate receptor-based drugs designed to competitively inhibit these
microbial attachment processes may be therapeutically beneficial in cases where
conventional antibiotics are, for one reason or another, ineffective.5
The rest of this article will elaborate on how we and others have been testing
this concept.

Enterohemorrhagic E. coli (EHEC) cause a condition called hemorrhagic
colitis (HC) in humans.6 Although several EHEC serotypes have been
identified, E. coli O157:H7 serotype is isolated from the majority of HC
cases in North America, Europe and Japan.7 Patients who develop EHEC-mediated
HC are also at risk of developing a potentially life-threatening post-infection
complication known as the hemolytic-uremic syndrome (HUS). This serious
complication of HC is directly linked to the production of potent EHEC
cytotoxins.8 These represent a family of cytotoxins which are related
to the Shiga toxin (Stx) produced by Shigella dysenteriae Type 1. One of
the EHEC cytotoxins, Stx1, is virtually identical to the Stx expressed by Shigella
9-10 and was formerly called Shiga-like toxin I (SLT-I).
This cytotoxin was also called Verotoxin 1 (VT 1) because its cytotoxic activity
was initially described in African green monkey (Vero) kidney cells.11
The other EHEC cytotoxin, formerly called SLT-II or VT 2 but now referred to as
Stx2, is more distantly related to Stx1. In addition, several variant forms of
Stx2 have been described in EHEC isolates. Regardless of their relationships,
though, all the EHEC Shiga toxins share structural and functional similarities.8

The EHEC Shiga toxins are classical hexameric "A/B5" exotoxins
consisting of one enzymatically active A subunit and 5 identical B subunits.12-13
Although the cytotoxic action of EHEC Shiga toxins is expressed by their A
subunits in the cytoplasm of host cells, absent the B pentamer that binds to
glycan receptors on the surface of the host cell,3 the A subunit is
atoxic. The Stx B pentamer is critical to the intoxication process because it
acts as a mediator for A subunit binding to and translocation into the host cell
cytoplasm. Given its critical binding function, it seemed feasible that soluble
B pentamer glycan receptor analogs might competitively inhibit Stx binding to
and intoxication of host cells. The therapeutic benefits of such Stx receptor
analogs might be to prevent HUS developing in subjects suffering from HC.

Typically, protein-carbohydrate interactions are of the low affinity variety.14
Unlike enzymes, which contain a deep cleft or tunnel into which the
substrate tightly binds, the carbohydrate ligand binding domains of proteins
such as the EHEC Shiga toxins usually occur as shallow depressions on the
surface of the protein. Much of the binding energy in carbohydrate-protein
interactions is derived from the displacement of disordered water molecules from
the depression in the surface of the protein by the ordered hydroxyl groups
displayed by the glycan sequence.15 Nature compensates for the weak
binding by simply multiplying the number of equivalent ligand binding domains in
carbohydrate-binding proteins. This allows these proteins to achieve tight
binding by simultaneously engaging multiple receptor sequences.

We have used the X-ray crystal structure of the Stx1 B pentamer, complexed
with soluble glycan receptor analogs,16 as a template for creating a
symmetrical soluble inhibitor, we call Starfish, capable of simultaneously
embracing the multiple domains found on the receptor binding face of this
protein.17 The resulting Starfish analog displayed a solid phase
binding inhibition constant that was superior, by 10 million times, to that of a
monovalent inhibitor capable of engaging only one of the multiple Stx1 B
pentamer receptor binding domains. In further co-incubation studies,17
the polyvalent carbohydrate-based soluble inhibitor protected Vero cells, at
concentrations in the ?M range, from a lethal challenge dose of both Stx1 and

After we disclosed our multivalent inhibitor in the literature, a group of
Japanese investigators presented data demonstrating that ?M doses of a
polyvalent inhibitor they called "Super Twig" protected mice from
lethal challenge doses of both Stx1 and Stx2 [(Nishikawa, K., et al. Development
of a new type of Shiga toxin neutralizer, carbosilane dendrimer, which
completely rescues the lethality of Shiga toxin 2 in mice. 2000. Abstract 415,
4th International Symposium and Workshop on Shiga toxin [Verocytotoxin],
producing Escherichia coli infections, Kyoto, Japan).] In this report,
the inhibitor, like the toxins, was administered intravenously and appeared to
be well tolerated by the mice. Since the Japanese report, we have obtained
compelling evidence demonstrating the efficacy of symmetrical polyvalent
inhibitors in protecting mice from the lethal effects of Stx118 and
now Stx2 (in preparation). The dosages used in our in vivo experiments were
equivalent to a few milligrams of inhibitor per Kg of body weight and so far,
we, like the Japanese, have not observed any toxic effects of these
carbohydrate-based inhibitors in mice.

Others19 have used an approach similar to ours to develop
effective inhibitors for two additional bacterial exotoxins in the A/B5
classification group; Cholera toxin and the Heat-labile (LT) toxins expressed by
enterotoxigenic E. coli (ETEC), for example. It would seem that, at least
for bacterial exotoxins that display symmetrical arrays of multiple ligand-binding
sites, it is generally feasible to produce carbohydrate-based soluble inhibitors
that have exceedingly beneficial effects at pharmacological dosages. These
recent findings bode well for potentially commercializing polyvalent
carbohydrate-based receptor analogs for treating diseases caused by EHEC, ETEC,
and, possibly, Vibrio cholerae.

Not all bacterial carbohydrate-binding proteins display the same five-fold
rotational symmetry as the A/B5 family of exotoxins. Some bacterial exotoxins,
those expressed by Clostridium difficile, the organisms responsible for
antibiotic-associated colitis in hospitalized patients or residents of
extended-care facilities, for example, contain a linear array of multiple
carbohydrate-binding domains.20 Filamentous bacterial adhesions may
also display linear arrays of multiple carbohydrate-binding domains.21
Alternately, the pili or fimbriae expressed by some bacterial pathogens may
display carbohydrate-binding domains only at their distal tips. In these
"tip domain" situations, polyvalency may be achieved not only by the
symmetrical arrangement of multiple carbohydrate binding sites at the tips of
the individual pili but also by expressing multiple pili per organism.21

The situation presents an even greater challenge to developing effective
inhibitor analogs because the flexibility of these filaments contributes to a
relative lack of symmetry in these systems. Once an organism has attached itself
to a host cell, it may be impossible for a soluble inhibitor to reverse this
process. We22, however, have demonstrated that a synthetic
glycoconjugate consisting of bovine serum albumin (BSA) containing covalently
bound N-acetyllacosaminyl glycoside sequences inhibited bundle-forming
pili (BFP) -- mediated attachment of enteropathogenic E. coli (EPEC) to
cultured human epithelial cells. This inhibitory activity appeared to depend on
the average number, i.e., potential valency, of N-acetyllacosaminyl
glycosides attached to each BSA molecule and was specific for this particular
carbohydrate sequence. In EPEC, at least, the relative absence of a symmetrical
arrangement of the BFP-localized carbohydrate-binding domains did not hinder the
inhibitory activity of a BSA glycoconjugate containing multiple glycan ligands.

Further investigation into the EPEC host cell binding inhibition mechanism of
BSA glycoconjugates revealed the process was more complex than we originally
believed. These studies demonstrated that inhibition likely involved two
mechanisms, one involving simple competitive inhibition, and the other involving
a time-dependent glycoconjugate-mediated loss of BFP from the surface of the
bacteria. At a glance, this would appear to be counterintuitive to the role of
BFP in EPEC attachment to host cells. However, EPEC colonize host cells by a
complex multi-step process. BFP-mediated attachment represents only the first
step in EPEC colonization of host cells. Once this first step has successfully
occurred, the continuing presence of BFP on the surface of EPEC may interfere
with the subsequent colonization steps. We are speculating, therefore, that EPEC
BFP binding to specific carbohydrate receptors may produce an as yet unknown
signal informing the bacterium that it has made initial contact with an
accommodating host cell and may now discard its BFP in preparation for the next
phase of colonization.

Once EPEC have colonized a specific site in the host, it may be necessary for
some organisms to be released from this primary colonization site. In the
absence of any host response, these released organisms would then be free to
colonize additional, perhaps downstream host sites, thereby perpetuating the
infection process. Once they have been released from the influence of their host
cell glycan receptors, these planktonic organisms may once again produce BFP
that would enable them to colonize the new sites in the host. It is during this
secondary phase of the infection process, the colonization of additional host
sites, that the administration of a multivalent carbohydrate receptor analog may
be of benefit in a subject suffering from EPEC-mediated diarrhea.

Others have also noted that EPEC alter the expression of their adhesins in
response to contacting host cell surfaces and there is some speculation that
glycan receptors may be generally involved in these sensing reactions.23
Many pathogenic microbes, including EPEC, possess the ability to assemble a
complex multimeric organelle, called a Type III secretory apparatus, when they
contact host cell surfaces.6 This Type III secretory apparatus is
necessary for injecting virulence factors into the cytoplasm of host cells.
These factors cause host cells to modify their physiological behavior conducive
to the pathogenic process. In light of what is known about the possible role of
glycan receptors in regulating adhesin expression in EPEC, it is possible that
specific host cell surface carbohydrate sequences may also influence the
function of microbial Type III secretory systems. Should this assumption prove
to be valid, soluble carbohydrate receptor analogs may also eventually be useful
for disrupting the function of Type III secretory systems. Such analogs might,
for example, trigger Type III-mediated secretion of virulence factors from
planktonic bacteria, thereby causing these factors to disperse harmlessly to the
exterior of the host cells.

It is unlikely microorganisms will rapidly develop resistance to
carbohydrate-based anti-infective drugs because this would involve altering
their ability to bind such agents. Organisms that alter their ability to bind to
soluble host cell carbohydrate receptor analogs would not gain an advantage over
their wild-type siblings because these mutants would also lose their ability to
bind or respond properly to host cells. Alternatively, microorganisms may
acquire the capacity to express adhesins with a specificity for different host
cell receptors. However, although it is quite possible for microbes to acquire
the genetic capacity to express complex multi-protein virulence factors such as
pili or Type III secretory systems, en block6, this may
require a sustained environment which highly favors organisms displaying the new
phenotype. It is unlikely the use of carbohydrate-based anti-infective drugs
only during the acute phase of an infection would create such a sustained
environment. To be of any use to the microbes, the newly acquired adhesins would
also have to display a specificity for other unique carbohydrate sequences
already present on host cells.

Glen D. Armstrong is a professor in the Department of Medical Microbiology
and Immunology at the University of Alberta, Edmonton, AB, Canada.