aka “Antibiotic Uptake in Gram Positive and Gram Negative Bacteria”

This is a little different from the usual fare, but let’s talk biology.


Membrane Diffusion in Drug Discovery

Permeability is one of the key properties of drug candidates during the development process (the others being efficacy, toxicity, stability and solubility), as drugs with good permeability can not only be formulated as pills, but will be well distributed throughout the body.  It does no good for a lead compound to inhibit its target enzyme in the low nanomolar range if the target sits inside a cell while the lead is stuck outside.

The membrane bilayer of mammalian cells is fairly well understood, and for our purposes can be considered a simple amphiphilic bilayer [1].  Polar lipid head groups line the outside of the membrane, interacting with water molecules, while hydrophobic lipid tails are paired end to end within the bilayer, preventing ready transfer of dissolved drugs into or out of the cell.  A number of proteins stud the membrane, giving the cell its shape, allowing it to communicate with its environment, and facilitating passage of hydrophilic nutrients like glucose through the bilayer.

Essential Cell Biology. 2e. Garland Science. 2004.

Passage through the mammalian membrane is primarily via passive diffusion, and is linked to physiochemical properties, not to any specific functional groups on the compound of interest.  In the heady days of combinatorial chemistry (~1990) this made pharmacokinetics a low priority during the synthesis of high throughput screening compounds.  Though it was before my time, the thinking appears to have been that permeability could be instilled into HTS hits later in the development cycle, after optimizing for activity.  This strategy worked poorly.

When it became apparent that projects were being dropped due to poor permeability, several sets of ‘rules’ were used to cull these bricks from the high-throughput screening libraries.  Lipinski’s rules are the most famous, and they helped reduce failure during PK/Bioavailability optimization about fourfold.

Lipinski’s Rules

Orally available drugs generally:

  1. Have a molecular wight elss than 500 g/mol.
  2. Have less than five hydrogen bond donors (NH and OH).
  3. Have less than ten hydrogen bond acceptors (=O and =N).
  4. Have an log[octanol]/[water] (logP) ratio of less than five.

The rules were derived from empirical observations, with molecular weight only correlated with poor permeability.  As compounds increase in mass they naturally grow more complex, with a greater number of rotatable bonds and a larger polar surface area.  Both of these attributes limit permeability, if for different reasons.  Lipids in the bilayer generally rest at full extension. This creates a constricted space that is perturbed by flexible compounds, imposing an entropic penalty to passage through the membrane.  At around seven rotatable bonds the penalty is high enough to prevent the passage of a large fraction of compounds, regardless of their molecular weight.

A compounds polar surface area is of course linked to the number of hydrogen bond donors and acceptors, and few compounds with greater than 110 Ǻ2 PSA are able to cross the lipid bilayer.  All dissolved compounds form a network of hydrogen bonds to water molecules, creating a network of coordinated hydration spheres.   These spheres provide a large enthalpic boost to solubility, but must be shed for passage through the membrane.  110 Ǻ2 PSA appears to mark the border region, when the hydration sphere is bound too tight for removal.

Conversely, the greater the logP the weaker the interactions with water.  A measurement of compound distribution through equal volumes of octanol and water, compounds with high logP values generally prefer the hydrophobic membrane, and may readily diffuse across.  However, if they have little or no polarity they will have poor solubility, and will make poor drug candidates.  The sweet spot generally lies in the 2-5 range.

Exceptions to Lipinski’s Rules

Most antibiotics violate one or more of Lipinski’s rules [2].  Conversely compounds which follow Lipinski’s rules tend to fare very poorly against both bacterial enzymes and whole cell bacteria.  Unlike mammalian cells, bacteria are surrounded by a hostile environment, filled with toxins and chemical weapons.  As a result their enzymes tend to have more solvent accessible binding pockets, necessitating large, polar inhibitors.  This runs counter to what many medicinal chemists would prefer, and has to date mostly limited antibiotic scaffolds to naturally produced secondary metabolites.

Natural products have made a splash working against mammalian enzymes too.  One of my favourite rule-breakers is Cyclosporine A: a large, cyclic peptide that has found use as an immunosuppressant.  With a molecular mass of 1202 Da, a cLogP of 14.4, and a whopping 279 Ǻ2 PSA, one would expect this drug to have incredibly poor permeability.  However, approximately 30% of every oral dose makes its way through the intestinal wall into our bloodstream (though the solubility is very low).  In non-polar environments cyclosporine is able to fold in on itself, forming a number of intramolecular hydrogen bonds that offset much of the enthalpic penalty inherent with shedding the large hydration sphere.  This behaviour can be modeled with molecular dynamics simulations, and has been experimentally verified in CNS drug discovery with a cute little NK1 antagonist.

The initial compound, 1, had good activity in vitro (1.05 nM), but poor solubility (2 ug/mL).  Adding on a tertiary amine to form a thick hydration sphere increased solubility about three-hundred fold (and doubled activity in vitro), but would normally have completely sunk passage through the blood brain barrier.  However, the tertiary amine was able to form an intramolecular hydrogen bond with a pre-existing polar amide, resulting in a roughly four-fold increase in accumulation in brain homogenate and brain supernatant [3].

The Bacterial Cell

However, cyclosporine A has it easy.  Bacterial cell envelopes are built to reduce such passive diffusion, and have proven largely successful.  Gram negative bacteria have two membranes, with a thin peptidoglycan layer spanning the periplasm and a lipopolysaccharide chains embedded on the outer face.  This creates alternating hydrophobic and hydrophilic bands, and the physiochemical properties optimal for passage through one band will lead to poor passage through the next.  Even Gram positive bacteria, with their single membrane and thick peptidoglycan coat (which has all the stopping power of tissue paper) are less permeable than mammalian cells.  Most antibiotics must bypass one or more of these barriers to enter the cell.

Microbiology. 6e. McGraw Hill. 2005.

Like mammalian cells, bacteria use integral membrane proteins to take in nutrients and export waste.  Of clinical relevance are the porins, which span the outer membrane of Gram negative bacteria.  Large, water filled channels about 0.8-1.1 nm across, the porins allow most polar molecules about 600 Da or smaller access to the periplasm, including beta-lactams.  Mutations in porin encoding genes have been linked to variations in antibiotic susceptibility, and help to explain the resilience of Pseudomonas aeruginosa.  A few antibiotics use substrate specific carrier proteins to enter the cell, but this route is generally of limited use as the drugs must closely mimic the proteins’ natural substrates.

A nice example of this specificity is fosfomycin, which inactivates MurA by acting as a phosphoenolpyruvate analogue (a suicidal inhibitor, in this case).  Fosfomycin rides not just one, but two carrier proteins into the cell, taking advantage of its structural similarity to both glycerophosphate and glucose-6-phosphate.  If the cells are pretreated with G6P to induce expression of the G6P transporter the intracellular concentration of fosfomycin can actually reach double the extracellular concentration.  Any significant alteration in the structure of fosfomycin could be expected to reduce its affinity to not only MurA, but these two carrier proteins, creating a narrow path to successful analogue development.

Self-Promoted Uptake

Co-opting a bacterial protein works well for several important classes of antibiotics, but is a susceptible to resistance development via mutations in those transporters.  Aminoglycosides and the cationic antimicrobial peptides (CAMPs) have decided to take a more… direct approach.

The lipopolysaccharide strands coating the outer membrane of Gram negative bacteria are bridged by divalent cations like magnesium and calcium.  With five to six positive charges the aminoglycosides form much better bridging agents, and are able to pull several LPS strands to a relatively small area.  This alters the organization of phospholipids in the outer membrane, which in effect forms “cracks” through which other aminoglycosides molecules can flow.  CAMPs work similarly in vitro (they are attenuated by hydrophobic serum proteins in vivo), with the added bonus of being able to directly insert into the bacterial membrane.  This insertion forms transient pores, scrambling lipids between the inner and outer leaflets and depolarizing the cell.  Like mitochondria, bacteria rely on a pH gradient for ATP production, and cell depolarization is rarely beneficial to the organism.  These pores also allow passage of the peptides into the cytosol, where they are free to interact with negatively charged macromolecules such the ribosome or hydrophobic structures like chaperone proteins.  The result is death from a thousand cuts, as the bacteria’s internal processes begin to break down.  At exceedingly high concentrations the peptides may even dissolve the cell, acting similar to detergents.

Resistance to such an assault is difficult, as any large-scale alteration in the cell envelope will interfere with the delicate balance of membrane proteins.  Strains with these alterations demonstrate resistance to CAMPs, but are weaker than their non-resistant brethren in less hostile environments.  Mutations in internal processes are a bit of a non-starter, as CAMPs target based on physiochemical properties, and have dozens or hundreds of targets.  This selectivity cuts both ways, and the antibacterial activity of CAMPs is almost entirely driven by their physiochemical properties.  Even some FDA approved drugs have been found to have clinically relevant antimicrobial activity, especially those with CNS targets [4].

Membrane Disruption in Drug Discovery

There are quite a few antibiotics whose spectrum of activity isn’t limited by which bacteria have susceptible enzymes, but by which bacteria have sufficiently porous membranes.  Rifampin is a good example, as it targets the DNA-dependent RNA polymerase found in most clinically relevant bacteria.  Rifampin’s therapeutic use is limited to Gram positives and mycobacteria, as it is unable to cross the Gram negative membrane.  However, when administered (in vitro) in tandem with CAMPs Rifampin’s MIC improves about a hundred-fold, as the cell envelope no longer represents a significant barrier.

Interestingly, chemically linking CAMPs to drugs can impart both cell and organelle specific targeting.  The Kelley lab at the University of Toronto has shown this quite well, linking FDA approved drugs to a mitochondria targeting peptide they developed.  They especially favour toxic agents, with a recent publication exploring the dihydrofolate reductase (DHFR) inhibitor, methotrexate.  DHFR catalyzes an essential step in folate biosynthesis in both prokaryotic and mammalian metabolism, limiting the drug’s use to immunosuppression and chemotherapy.

When linked to the mitochondria targeting peptide methotrexate is sequestered in an organelle lacking DHFR, which eliminates much of the eukaryotic toxicity.  However, the same pull to negatively charged membranes directs the conjugate to the interior of bacterial cells, increasing the antibacterial effect about forty-fold.  The end result is roughly a 180, 000-fold increase in the therapeutic window [5].

The next step in the grand scheme of things would be to combine a hydrophobic and a cationic drug, giving the hybrid organelle targeting features while keeping the molecular weight low (not to mention retaining the synergy of two drugs).  This is something that I’ve done a bit of work in, if you’re interested in a self-link.  Regardless, moving on.

Drug Efflux Pumps

From the bacteria’s perspective antibiotics are the barbarians at the gates, and so it’s not surprising that they’ve developed countermeasures.  Mutations in LPS and membrane structure have been found, and give a 2-4 fold decrease in the apparent MIC against aminoglycosides and other permeabilizers.  However, the greater threat is a more proactive response, drug efflux pumps.

Simple active transporters with poor substrate specificity, efflux pumps pull antibiotics out of the cell, preventing drug action.  The transporters’ poor specificity is an asset to the bacterium, as a single efflux pump can target an entire antibiotic class and provide broad spectrum resistance.  Pump inhibitors have been developed, and are quite effective in a lab setting, but must combat both dosing concerns and the expression of multiple pump types.    In the CNS and cancer world altering the structure of the lead compound can reduce efflux by pumps like P-glycoprotein, but this approach is again limited in a bacterial setting by the large number of efflux pump types.  The most effective tactic may be to saturate the pumps, by increasing compound permeability or by improving self-promoted uptake.  As in many things, the best defense is a good offense.


Far from the world of Lipinski’s rules and passive diffusion, the majority of antibiotics appear to enter bacterial cells by either hijacking a bacterial transporter or by disrupting the integrity of the bacterial membrane.  This is due to alternating polar and hydrophobic barriers in Gram negative bacteria, which significantly reduces passive diffusion, and shift towards larger, more polar enzyme inhibitors.  Several classes of antibiotics promote their own uptake by disrupting the bacterial membrane.  Incorporating such mechanisms into to drug research has the potential to increase permeability in large, polar compounds that may have otherwise been removed from high throughput screening libraries.  These compounds are more likely to be active against not only bacterial enzymes, but also challenging drug targets like protein-protein interactions.


[1] An idea that is not without controversy.

[2] Those that I work with tend to break all five of Lipinski’s rules.  Who doesn’t like a challenge?

[3] It’s pretty rare to improve efficacy, solubility and permeability in a single stroke.  If you’re interested in learning more about intramolecular hydrogen bonds, these two papers are a good place to start.

[4] Positive charges help association with the negative phospholipids in the blood brain barrier, while CNS agents are generally more hydrophobic than the norm, to improve permeability.

[5] The targeting peptide will likely be susceptible to serum proteases, so this targeting could break down in vivo.  But, with a little work it should be entirely possible to replicate these results with more metabolically stable compounds.