In this installment, I will discuss why it is difficult to discover, design and develop a drug, in view of our current knowledge of physiology.
With numerous, intertwined reactions happening, our body is a complex clockwork of biomachinery gears. What do you do, then, if some gears fail—that is, if you got sick? On one hand, it is a consolation that many gears are what biologists call 'redundant', which means that it's alright that a certain gear fails, because there are other gears that can take over its function. On the other hand, due to the intricacy of the gears, it is hard to pinpoint which gear is the problem, let alone fixing it. And the sheer number of gears: ICD-10 classifies tens of thousands diagnoses — tens of thousands ways the gears can fail — and those are only the ones we know; how about those we don't? Granted, some are not caused by our own gears failing, but by interferences of other, pesky gear systems: viruses, bacteria, misfolded proteins, errant microbiome, etc.; but the sense of magnitude is there.
So we take drugs. In the past, the way we administer drugs is the equivalent of throwing various types of wrenches to the clockwork and then observe whether the gears are working again. Today we know more about the gears and the various shapes and teeth so that we are able to design a more sophisticated and targeted wrench, but we still don’t know enough.
Our current thinking is that the machinery gears, mostly proteins, are of distinct shapes, or perhaps more fittingly, different tooth shapes. A drug has gotta fit into these various shapes. We think that if we can find 'keys' that fit to these 'locks' we can modulate that particular gear's activity: turn it up or down, switch it on or off. The key would fit into the lock, the lock would be induced to change shape / dissociate / do other curious stuff; which triggers happenings in the next gear in line. Like the dominoes falling off in line. Only that ‘the curious stuff’ may be more than a domino keeling over or a gear turning, but something that is more wackily messy, something like a Rube Goldberg contraption than a precise-looking clockwork, in this respect.
How should we shape the serrated key? The first problem is scale. These gears are small — not microscopically, but nanoscopically so: To design the keys, we need the moulds, and so these very gears are the moulds. Structural biology to the rescue — X-rays crystallography/multidimensional NMR/cryo-electron microscopy can characterise the gears with varying resolutions.
However, structural determination techniques face a big challenge: the interconnectedness of the system. You can't take out a gear out of context of the surrounding gears, examine it, hoping that your examinations will be valid and/or useful. Well, you sort of can. Sure, systems biology is important to gain a bird's-eye view of the whole shebang, but structural biologists routinely single out a gear and determine its structure to make it crystal-clear (heh heh) what its possible activation mechanism is, how it interacts with its activator/inhibitor, how it transduces signal to the next gear, and so on.
The structural data, then, has to be taken with a grain of sodium chloride because essentially it is performed—in the physics equivalent of—in vacuo. That said, the gleaned information can be incredibly useful. Case in point: enfuvirtide. This anti-HIV-1 drug mimics a region—C-heptad repeats (CHR)—of the viral envelope glycoprotein, gp41, which is a crucial part of viral:cell membrane fusion machinery [1]. CHR is supposed to fold back to NHR like a hairpin (they are connected) and then Stuff happens, with gory details I will spare you from (Oh, fine: a channel is opened between HIV-1 membrane and that of your soon-to-be-infected T-cell; viral particles are pumped through and soon hijack your T-cell to become baby-virus-producing zombie until it bursts releasing said baby virions. Happy?) Now, the 'folding back' part is a mechanism that was uncovered through cleverly-devised experiments founded in structural studies. Thanks to this, we can deduce that if we somehow have a fake CHR, this 'folding back' can be circumvented. And that's exactly what enfuvirtide is: a dummy gear that connects to NHR gear, but not connected to the gears down the road, thus—hip hip hurray—no zombie outbreak.
This is kinda cheating though. Enfuvirtide is essentially free-roaming CHR region of gp41 and no rational design work was done. For example, peptidomimetic strategy could have been used to find something similar to enfuvirtide but is able to survive the gut—since a peptidic drugs like enfuvirtide won’t, so they have to be taken intravenously. But of course, HIV-1-infected individuals don’t have the luxury of time to wait for further work on enfuvirtide optimisation.
Back to the 3D protein mould model built by structural determination techniques. Ideally, we can start building up the drug à la Lego bricks, fitting the chimaera into the protein-shape mould, right, right? Couldn't be more wrong. Enter multidimensional fitting. You see, a protein ain't like your Mom's muffin pan. Besides topology, there are other dimensions to fit—electrostatic charges, hydrophilicity/hydrophobicity, to name a few. The topology is not necessarily fixedly rigid either — playing with those poppin’ stick-and-ball molecular models may give us the illusion that proteins are rigid, but protein electron clouds are more like wobbly pudding (Mmmm, pudding...); plus, different environments (pH, oxidising level) may give rise to largely different topologies (e.g. due to different protonation states, broken disulphide bridge(s), etc.).
One aspect of nanoscopic scale that is somewhat entangled with multiparameter fitting is that at this scale, there is evidence that quantum effects play an important role. Several examples of such systems have been studied like photosystems and birds’ navigation, but of more interest to a medicinal chemist would be tunnelling effect in certain enzymatic reactions. Who knows if quantum effects are more routinely utilised? Our view of physiology is biasedly mechanical—even my clockwork allegory evokes the mechanistic, so a paradigm shift may well be in order as more is known about the inner workings of such systems. If you blindly design a drug that target such systems, well, you will get an entangled mess—hopefully not the quantum kind.
Moving on to pharmacokinetic restrictions: the human body imposes further restrictions from the non-negotiables (e.g. a drug cannot be too insoluble otherwise how can it dissolve in the bloodstream; a drug cannot have side effects outweighing its efficacy), to convenience (e.g. a drug is preferable to be ingested rather than injected). Our own physiology thus severely restricts the chemical space of entities that make up our drug candidate pool. As I mentioned earlier, enfuvirtide is a peptide, so it won’t survive stomach acidity and peptidases in the gut. Even with intravenous administration, it would have a short half-life due to blood proteases. All peptidic drugs—insulin is one—suffer from these problems. And for drugs targeting the central nervous system (CNS), they have to overcome another obstacle, the blood brain barrier.
There seems to be some sort of patterns to the drug chemical space. To wit, some have observed that certain chemical scaffolds occur more frequently than others; they are so-called privileged scaffolds (e.g. benzodiazepines) [2] and an experienced medical chemist would be able to take a look at a chemical structure and decide whether it's 'drug-like' (while an inexperienced chemist like me would only know that a drug-like molecule can't be too small and simplistic, possesses some heteroatoms, usually has an aromatic ring or two—that's about it). Problem being, at its current state, drug-likeness is an empirical measure. No one has formulated a set of rules or equations to produce a predictive model. Lipinski's Rule of Five, for example, surveys already-existing drugs and look at the prevalent drug-like characteristics. Useful as rule of thumb; hardly predictive.
Next is the issue of specificity. If you choose to take a drug topically, that’s fine and dandy because you can apply the drug locally to the area in need of treatment. But if you take a drug via oral/intravenous/other numerous administration routes, the drug is going to circulate in your bloodstream. How would you ensure that your wrench would reach, and affect only, the faulty gear? You can’t—not with certainty, at least—the wrench is going to wreck another gear, and that’s why you always have side effects. If we are dealing with pathogens, it’s mighty nice that they possess a different set of biomachinery which we can specifically target. Penicillin, for example, targets the bacterial cell wall, which we as eukaryotes don’t possess. A huge class of antibiotics also target the prokaryotic ribosome, which is markedly different from ours.
But what if we are dealing with our own cells? Well, we do have some sort of defense, the immune system. They are the elves who repair the clockwork once the watchmaker is tired and goes to sleep—well, not really, since they work ‘round the clock, but something like that. They are good in recognising foreign pathogens, but sometimes they are overzealous, giving rise to autoimmune diseases. They can also be compromised, though diseases that debilitate the immune system like AIDS are not common.
How about dealing with our own cells on vendetta?—yes, I’m talking about cancerous cells. From outside, they look the same to the immune as other cells, possessing the same receptors, but inside they have turned their allegiance to—gasp— the Dark Side. Bb-b-b-but there is something different about them, rr-r-right? Something we can target on? Sure, some cancerous tissue has been observed to have lower pH and enhanced retention and permeability but generally what is toxic to the cancer cell will still be toxic to the surrounding healthy cells — we haven’t come up with something that specific yet. Now, even if you look for phenotypic markers peculiar to cancer cells, it turns out that there is great genetic diversity among the cancer cells themselves [3]. You may have heard about tailored drug for individual, since, not surprisingly, a drug works slightly differently in different individuals Quite a number of people I know, for example, get more sleepy upon ingesting caffeine, while for me I need that liquid black magic right after waking up to undo my zombification. Even the details of this individual tailoring of drug treatment may not be to the structural level—not practical—but maybe different cocktail composition or the like. But bespoke drug treatment for each phenotypically distinct cancer cell? Kidding me, right?
In closing, I hope I have convinced you that this field is mega-complex. It’s biology—and also beyond. It’s synthesis of biology and chemistry and unicorn magic. It’s something worth pursuing and I’m proud to be part of this great undertaking.
Stay tuned for next installments as I discuss more about tackling drug-related challenges and Pharma.
[1] Eckert, D., & Kim, P. (2001). Mechanisms of Viral Membrane Fusion and Its Inhibition Annual Review of Biochemistry, 70 (1), 777-810 DOI: 10.1146/annurev.biochem.70.1.777
[2] Welsch, M., Snyder, S., & Stockwell, B. (2010). Privileged scaffolds for library design and drug discovery Current Opinion in Chemical Biology, 14 (3), 347-361 DOI: 10.1016/j.cbpa.2010.02.018
[3] Marusyk A, Almendro V, & Polyak K (2012). Intra-tumour heterogeneity: a looking glass for cancer? Nature reviews. Cancer, 12 (5), 323-34 PMID: 22513401
No comments:
Post a Comment