AUSTRALIAN FRONTIERS OF SCIENCE, 2008

The Shine Dome, Canberra, 21-22 February

Punching a hole in the eukaryotic cell membrane: Family secrets of bacterial toxins
by Dr Galina Polekhina

Galina Polekhina has a PhD from Aarhus University in Denmark on elongation factor Tu (EF-Tu), one of the essential proteins involved in protein biosynthesis. The structures of the inactive EF-Tu and active EF-Tu in complex with tRNA were landmark in the field and highlight the importance of conformational flexibility at the molecular level. Since 1997 she has worked at St Vincent's Institute of Medical Research in Melbourne, first as a postdoctoral scientist and then as an independent researcher determining structures of many medically important proteins, including ubiquitin ligase Siah, glycogen binding domain of AMP-activated protein kinase, galectins and maleylacetoacetate isomerase. She is currently working on several drug targets in diabetes and cancer in collaboration with local and international scientists.

The subject of my talk is cholesterol dependent cytolysin, which James Whisstock has referred to in his talk, but for simplicity I will just say that my talk is on pore-forming toxins.

(Click on image for a larger version)

Pore-forming toxins are produced by a wide variety of bacteria. Several of them, such as Listeria, streptococci and pneumococci, cause invasive disease. Pore-forming toxins contribute to ill effects of bacterial infections that can cause conditions such as necrosis, gas gangrene, meningitis and pneumonia.

Several pathogenic and non-pathogenic Gram-positive bacteria from at least five different genera secrete water-soluble toxins. These toxins can oligeromerise on the target cell membrane and form pores, or holes. This leads to cell lysis and consequently cell death all of these toxins have 'lysin' in their name, reflecting their ability to lyse cells, or spill cells' guts.

The pores are formed by as many as 50 toxins and the largest known can reach a diameter of up to 30 nanometres. Even at sublytic concentrations, these toxins can interfere with the host immune system.

Two toxins will feature in my talk today: perfringolysin O (PFO) is a prototypical member of the family, and the most unusual of them all is intermedilysin.

(Click on image for a larger version)

The toxins are all secreted as water-soluble single chain polypeptides, with a molecular weight of 4760 kilodaltons, or, in other words, composed of about 500600 amino acid residues.

They share about 3070 per cent of pairwise sequence identity, and therefore are expected to share a three-dimensional structure and also to follow the same mechanism of action.

They display lytic activity towards many eukaryotic cells from various animal species.

The presence of cholesterol in the membrane is absolutely essential for function of the toxins. Free cholesterol can also inhibit lytic activity, suggesting that the toxins possess a cholesterol binding site.

There is a highly conserved tryptophan-rich region of 11 amino acids, near the C-terminal of the toxins, that is crucial for function. Nearly any amino acid substitution in it would render the toxin unable to lyse the cells, especially a replacement of any of three conserved tryptophans. However, there are two members of the family that do not comply with this sequence.

Most families have an 'odd' member, including the pore-forming toxin family. The odd one in this family is intermedilysin.

(Click on image for a larger version)

Intermedilysin is secreted by Streptococcus intermedius, and is a factor in deep-seated infections that can lead to abscess formation in the brain and the liver. Brain abscess formation may give rise to meningitis.

Intermedilysin, in contrast to all other members of the family, displays exclusive specificity to human cells. Indeed, it interacts with a protein receptor on the target cells, namely, a small glycoprotein of complement system CD59, which James Whisstock has already mentioned. Ironically, CD59 is an inhibitor of the complement membrane attack complex that protects a host from its own defence system, namely, its complement part. (Streptococcus intermedius is actually a normal constituent of our human mouth bacterial flora.)

Intermedilysin also, as I have already mentioned, has an unusual tryptophan-rich region, in that even though it is still inhibited by cholesterol, this occurs at concentrations much higher than for other toxins.

(Click on image for a larger version)

The approach I use in my research is X-ray crystallography. It allows us to determine the three-dimensional structure or capture a protein in one of sometimes many possible states. This information, combined with many biochemical data, helps us to understand how the protein works, and armed with this knowledge we can either look for ways to disrupt its function, if desirable, or even harness it for our own purposes.

To obtain the crystal structure: the protein of interest is crystallised, the crystals are exposed to X-rays, and diffraction images are collected. Normally there are several hundreds of them. A formula relates what is in the crystal to the diffraction picture. Therefore, using intensities recorded during the diffraction experiment and estimates of the information that was lost during the experiment, we can reconstruct what is in the crystal.

(Click on image for a larger version)

The first structure of a pore-forming toxin that was determined was the structure of perfringolysin O, which James Whisstock has already shown. It was done by Jamie Rossjohn and Susanne Feil at St Vincent's Institute. It became clear that it is a soluble form of the toxin. The molecule was roughly subdivided into four parts, Domain 1, Domain 2, Domain 3 and Domain 4.

(Click on image for a larger version)

The tryptophan-rich region turned out to be at the tip of Domain 4, with its hydrophobic tryptophans folded against the body of the domain, hiding them from the hydrophilic environment. The tryptophan-rich region has been known to insert into the membrane; however, it came as somewhat of a surprise that this region in the molecule can also traverse the membrane in the pore. Moreover, amino acid residues in this region were alternating between being in an aqueous and a non-aqueous environment when in the pore, thus suggesting that it will adopt an extended rather than helical conformation.

(Click on image for a larger version)

No obvious cholesterol-binding site could be found on the toxin. It has, however, been suggested that the tryptophan-rich region will unfold, insert into the membrane, and present its tryptophans for a stacking interaction with a cholesterol molecule in the membrane.

(Click on image for a larger version)

The membrane-inserted form would look somewhat as shown on this slide.

(Click on images for larger versions)

Clearly, a quite dramatic conformational change would be needed. It is believed that the tryptophan-rich region inserts into the membrane first. This sends a signal along the molecule and the contacts between Domain 3 and Domain 2 are lost, followed by some sort of collapse of the molecule and unwinding of the short helical regions. The transformation from alpha-helical conformation to beta-strand is not that unusual it has been seen before in several other proteins.

We and others have tried hard, but so far we have been unable to capture by X-ray crystallography a cholesterol-bound form of the toxin. To do so for a membrane-inserting form will be even harder, because the pores formed by the toxin are non-homogeneous, composed of different numbers of toxins, and therefore are not suitable for crystallisation. But what about other members of the family? And, especially, what about the most unusual of them all? This and the medical relevance of it prompted us to determine the structure of intermedilysin, which I did some time ago.

(Click on image for a larger version)

As expected, the overall structure is preserved. But there are some unexpected differences in the important parts of the molecule.

(Click on image for a larger version)

The tryptophan-rich region is unfolded, and is more like the proposed cholesterol-bound form, rather than the crystal structure of PFO. Intermedilysin has lost the second invariant tryptophan altogether, and the conformation of the third one is not optimal for a stacking interaction with cholesterol. Intermedilysin also is less sensitive to inhibition by cholesterol.

(Click on image for a larger version)

The structure of PFO has been captured in two different crystal forms. One of them was obtained at lower than physiological pH. The difference between the two crystal structures is a slight domain movement in the bottom part of the molecule, somewhat on the scale of 58. That is accompanied by the loosening of contacts between Domain 3 and Domain 2.

Intermedilysin has a much larger swing of Domain 4, thus leaving the transmembrane helices free to unfold, with no need for a large collapse of Domain 2. It is possible that this swing is necessary to accommodate binding to the receptor. The binding site of CD59 has been mapped on intermedilysin, and it is located in Domain 4, immediately underneath the transmembrane helices.

(Click on image for a larger version)

So is intermedilysin poised to insert? It is tempting to speculate that the form of the toxin that intermedilysin presented us with is the one that is immediately prior to the insertion of transmembrane helices into the membrane, with the tryptophan-rich region already tethered into the membrane. Intermedilysin activity is known to be stimulated by specific ions, and indeed it was essential to use sodium sulfate in crystallisation, so it is not that surprising that we saw two sulfate ions in the crystal structure, near Domain 2, one of them acting as a lock for this peculiar conformation of Domain 4 to the rest of the molecule.

(Click on image for a larger version)

Of course, if it is the form that is immediately prior to insertion, there isn't just one toxin there, there are 50 of them and they are assembled in a huge pore, all of them tethered to the membrane by the tryptophan-rich region. You may remember that I said intermedilysin displays exclusive specificity to human cells, thus overruling the role of cholesterol as a receptor.

(Click on images for larger versions)

So now, when the toxins are assembled and ready to insert, really we need a trigger, because we don't want conformational change to happen prematurely it would be disadvantageous. This trigger turned out to be cholesterol. Once the transmembrane helices are unwound, a huge barrel is formed through the cell membrane.

(Click on image for a larger version)

The model of the pore constructed this way is quite consistent with various descriptions and dimensions derived from mainly electron microscopy work done on different members of the family. Basically, it describes two concentric circles, the inner concentric circle corresponding to Domain 1, the outer one corresponding to Domain 4. And a thinner circle separating the two corresponds to Domain 2.

We are currently pursuing the structure of intermedilysin in complex with its receptor CD59. Because cholesterol has been ruled out as a receptor for other members of the family, it is quite possible that they use a still unknown protein receptor on the target cells.

(Click on image for a larger version)

Last but not least, I want to acknowledge people who have done the work: Rod Tweten, from the University of Oklahoma, who has been a long-time collaborator with Professor Michael Parker and does all the functional work on the toxins; Harry Tong for help with the synchrotron; Professor Parker for giving me the opportunity to work in his lab; NHMRC for supporting me; and the Australian Synchrotron Research Program (ASRP) for sponsoring the trips to the synchrotron.