Does natural selection explain the universal response of metazoans to volatile anesthetics?

Abstract

Volatile anesthetics disrupt behavior in every metazoan thus far tested. The question that has been put before us is why? In particular, might natural selection underlie the relatively uniform response of a wide range of organisms to volatile anesthetics? To address this question, I will consider two alternative mechanisms to natural selection for anesthetic sensitivity that might explain the wide ranging response to anesthetics. Organisms ranging from yeast to human have been shown to be affected by volatile anesthetics. This fact is obviously consistent with natural selection for volatile anesthetic sensitivity but is clearly not sufficient to conclude such. To be clear on what I mean here by natural selection for anesthetic sensitivity, the hypothesis is that response to anesthetics confers some sort of advantage for the organism in terms of survival and propagation of progeny. In its simplest form, the hypothesis requires all organisms for which there is such a selection to be exposed to anesthetics at some time during their life. Although a large number of compounds can produce general anesthesia, such an exposure is clearly not the case at least for most multicellular organisms. A variant of this natural selection hypothesis is that the selection is for an anesthetic mimetic as proposed by Sonner in the accompanying article.1 This, of course, cannot be excluded and is clearly the case for response to opiates in mammals. However, alternative explanations seem more plausible. One such explanation is that volatile anesthetics bind to, and functionally alter, a large number of biologically active molecules. If so, probability alone might lead to an observable anesthetic effect in every organism. This probability becomes even higher if one does not limit consideration to anesthetic effects at clinical concentrations against neurons. The probability argument has some experimental support for and against it. In favor, a number of proteins or peptides that clearly cannot mediate general anesthesia have been shown to bind volatile anesthetics even at clinical concentrations.2–6 The most famous example of this is firefly luciferase,2,7–10 binding of volatile anesthetics to which fits every pharmacological criteria for a relevant anesthetic target except, of course, that the protein is not neuronal and does not mediate behavior. If such seemingly random proteins have a sufficient affinity for anesthetics, it is not hard to imagine that a substantial number of neuronal proteins in every organism also bind and have their function disrupted by anesthetics. Indeed, Xi et al. has shown that at least 90 proteins are photo-affinity labeled by clinical concentrations of halothane.11 Of the 34 proteins that were identified among these 90, most are highly evolutionarily conserved. Thus, anesthetic binding to and action against some of them, thereby producing an observable behavioral effect in every metazoan, seems a plausible hypothesis. However, several genetic results argue against the probabilistic argument, at least one where there are a large number of relevant anesthetic targets in every organism. In the nematode C. elegans, mutations in a single gene, encoding a presynaptic protein, can produce a 30-fold variation in isoflurane sensitivity against coordinated locomotion, and one particular mutation can produce an animal that is fully resistant to human clinical concentrations of volatile anesthetic.12 Obviously, this result is completely inconsistent with a highly distributed anesthetic mechanism and argues against the probabilistic model. Likewise, mouse knockout mutants of a specific neuronal potassium channel produced a 50% increase in halothane minimum alveolar anesthetic concentration.13 This result, of course, is not inconsistent with a probabilistic model but does argue against a large number of relevant anesthetic targets, each contributing only a small portion to the total anesthetic mechanism. For IV anesthetics, specific mouse mutations of the γ-aminobutyric acid type A (GABAA) receptor fully block specific behavioral effects of etomidate and propofol.14 Obviously, volatile anesthetics are structurally distinct from IV anesthetics, but this genetic result clearly suggests that anesthetic action against single targets is capable of producing anesthetic end-points. Perhaps a more plausible hypothesis than the probabilistic model is that one or more anesthetic targets is highly conserved from lower to higher eukaryotes. This model requires not only that the target be conserved but that the anesthetic binding site and anesthetic action on the target be conserved across species. A clear distinction should be made here between natural selection for anesthetic action on this conserved target, as proposed by Sonner, and natural selection for the target itself. In the latter case, the selection is on the target itself independent of any anesthetic or anesthetic mimetic. A nice example of this is serum albumin, to