Back to the question: what does the Ames test tell us? One thing it does is something that all toxicological tests do - that, as Paracelsus put it, "the dose makes the poison." There's hardly a more important tox principle than that. You can get a lot of things to show positive for mutagenicity if you're willing to load up on them.
Beyond that key point, Ames himself has made the argument that synthetic compounds and naturally-occuring ones have the same hit rate in these assays. Plants have evolved a variety of pesticides and antifeedant compounds, many of which are reactive and toxic at some level - therefore, most (as in 99.99%, according to his estimate) of the pesticides in the human diet are those found in the plants themselves. The cruciferous vegetables (broccoli, cabbage, mustard and so on) are particularly rich in compounds that will light up an Ames test. A fine article of his from 1990 (Ang. Chem. Int. Ed.,29, 1197) states that ". . .it is probably true that almost every plant product in the supermarket contains natural carcinogens."
And that's before cooking. Many of these reactive compounds are destroyed by heating, but many others are formed, especially in browning or charring of proteinaceous foods. There are two ways to react to news like this: either you can panic at the thought that every meal you take is full of mutagens, or you can decide that (since people aren't dropping all around you) that we've apparently got some method of dealing with them.
That we do: the digestive processes, gut and liver especially, the same things that are the bane of medicinal chemists for tearing up our carefully-designed wonder drugs. They give the same treatment to most everything you eat. In most cases, they're successful at detoxifying whatever compounds might be present, even if they were at harmful concentrations. But there's a limit - if you chow down on plants containing cyanogenic glycosides (raw cassava root, apricot pits, etc.,) nasty amino acids (some kinds of Lathyruspeas,) or fluoroacetic acid (some South African weeds,) then not much is going to help you.
Ames's point is that the mental division many people have between "artificial" or "synthetic" chemicals (bad) and "natural" ones (good) is nonsense. The same number of toxic compounds are found in each category, and we're exposed to far more of the latter. Instead of worrying about parts-per-billion of pesticide residues, we should worry about greater public health risks like smoking, alcohol, etc. Going crazy about the minute amounts of synthetic compounds that we can now detect not only diverts time and money from more useful concerns - it can lead to decisions that end up doing more harm than the compound residues ever could. Ame's article is a fierce broadside against this sort of thinking.
If you're going to sound the alarm about chemicals, he suggests, look at high-dose occupational exposures. Here we get into toxicity that has less to do with a compound's mutagenic potential. At very high doses, you're basically causing cell death, irritation, and tissue injury. That leads to increased rates of cell division, leading to an increased chance of carcinogenesis.
We're back to "the dose makes the poison." The principle applies not only to people who are exposed to huge doses of chemicals, but to unlucky lab rats as well. Ames has forcefully made the point that testing compounds in animals at or near their maximum tolerated dose (MTD) is a poor measure of their cancer-causing potential. About half the compounds so tested show up as carcinogens, but the dose-response curves aren't linear. It's a complete mistake to assume that half of all chemicals cause cancer, unless you're soaking your feet in solvent while doing ice-cold shots of fungicide.
The implications for the toxicity testing of pharmaceuticals? We don't usually test our drugs at such high levels in chronic studies. Instead of working down from the MTD, as an environmental toxicologist might, we work up from the MED, the minimum efficacious dose. If a compound makes it through OK at some multiple (10x, 50x, 100x) of the MED, then we feel safe enough to go on.
As for Ames testing of pharmaceuticals, since we also don't go to such high levels, we don't see that many true positives. There are some known pitfalls: antibiotics can be tricky to assay, as you'd guess, since the procedure uses bacteria. Some of the drugs that target DNA-manipulating enzymes (like the fluoroquinolones) will give you a false positive because of the way the bacteria have been crippled for the test.
Since a real Ames-positive is uncommon in the drug industry, we pay attention when we get one. Would these compounds really be mutagens in humans? And if they were, would they be carcinogenic? Maybe not! But for the most part, no one knows, and no one's going to find out, either. It takes a lot of nerve to continue developing such a compound, and there really aren't enough data points to draw a conclusion.
It's the same way with animal tests. If something serious happens with your whole-animal tox (especially if it happens across species,) you usually pack it in and cut your losses. No doubt some of these compounds could have gone on safely, but we'll never know. At least not until we're a lot better at this than we are now. . .