Antiviral difficulty

Why are drugs that kill bacteria so much easier to find than drugs that kill viruses?

Within about two decades of penicillin, medicine had several unrelated classes of antibiotic, most of them pulled out of soil organisms by brute screening. Viruses had nothing comparable for far longer, and even now the list we can reliably treat is short.

The usual explanation is that a virus is simpler, and simple things are harder to break. That should sound suspicious: fewer moving parts normally means easier to jam, not harder. So what is it about a virus that resists?

Start by asking what a drug actually has to achieve. Not killing the pathogen — bleach kills every virus and every bacterium on a surface. The requirement is to damage the pathogen far more than the patient. Selectivity, not potency, is the scarce good, and selectivity has to be bought with some difference between them.

So ask what differs. A bacterium is a whole second cell living beside yours, and it is built differently almost everywhere you look. It has a peptidoglycan cell wall, a structure with no human counterpart at all. Its ribosome is the 70S kind, distinguishable from our 80S. It runs its own DNA gyrase. It has to synthesise folate from scratch, where we simply absorb ours from food. Four differences — and penicillins, tetracyclines and macrolides, quinolones, and sulfonamides map onto exactly those four. The differences were sitting there before anyone went looking; screening merely found molecules that exploited them.

Now do the same audit on a virus inside an infected cell. What is making the new virions? Our ribosomes are translating the viral proteins. Our transfer RNAs, our nucleotide pool, our membranes, our energy. Almost every chemical step between infection and a released virus particle is our chemistry, running normally, on instructions that happen to be foreign. Point a drug at any of it and you have pointed it at the patient.

What is left to aim at is only what the virus brings itself, and that list is genuinely short. HIV's genome is about 9.7 kilobases encoding roughly fifteen proteins; a bacterium like E. coli carries something on the order of four thousand genes (both recalled figures — the orders of magnitude are what the argument needs, not the digits). The target list is not hard to search. It is short.

That gives a test with a visible answer. If the binding constraint is how much machinery the virus owns, then viruses carrying more of their own enzymes should be the easiest to drug, regardless of how nasty they are. Check it. Herpesviruses have large genomes — on the order of 150 kilobases (recalled) — and carry their own thymidine kinase and DNA polymerase; we have aciclovir, ganciclovir and foscarnet. HIV brings a reverse transcriptase, a protease and an integrase; we have a drug class for each. Hepatitis C brings its own protease and polymerase, and direct-acting antivirals now cure well over ninety per cent of treated patients (recalled). Against rhinoviruses and noroviruses, which borrow almost everything, there is still nothing routine. The refuting observation would be the opposite ordering — the most parasitic viruses turning out to be the well-drugged ones — and that is not what we see.

Aciclovir compresses the whole argument into one molecule. It arrives inert, and the phosphorylation that activates it is performed efficiently only by the virus's own thymidine kinase, so it is switched on essentially only inside infected cells — after which it inhibits the viral polymerase far more strongly than ours. Two virus-owned proteins, stacked, to buy the selectivity a bacterium hands over for free.

"Viruses are simpler, so they are harder" turns out to be right for the wrong reason. Simplicity is not the obstacle; the ratio of borrowed to owned machinery is. A virus with a large genome is not harder to treat than a small one — it is easier, which is the reverse of what the folk version predicts.

Two things weaken my own account. First, host targets are not forbidden: maraviroc blocks CCR5, a human receptor, and works precisely because that receptor is dispensable. So the real rule is not "only viral proteins" but "any protein the patient can spare" — the virus's own proteins are simply the largest class of those. Second, selectivity is not the only constraint. Many viral illnesses have already peaked by the time symptoms bring someone to a doctor, which is why influenza antivirals shorten illness only modestly and only if started early; and RNA viruses replicate without proofreading, so single-drug therapy selects escape mutants quickly.

There is also a live dispute about whether the gap is scientific at all. Some argue the binding constraint was commercial — a short course for a self-limiting illness repays development poorly. That is a real force, but it does not explain why chronic viral diseases with obvious markets waited for structure-guided chemistry rather than yielding to the soil screening that produced the antibiotics.

• Why antiviral resistance behaves differently from antibiotic resistance, given that one arises within a single patient and the other spreads between people.
• How broad-spectrum antivirals are attempted by targeting host pathways a patient can temporarily spare.

Every term the collection defines is gathered in the glossary.