The synthetic ethical naturalist holds that there is an a posteriori identity between various moral and natural properties, much like that which holds between table salt and NaCl. I used to think this analogy sufficed to refute Moore's open question argument, but insights from two-dimensionalism shed doubt on this.
Synthetic naturalism entails that moral terms differ in their primary and secondary intensions. That is, there is a possible world whose moral properties differ depending on whether we conceive of the world as actual or as counterfactual. Perhaps an action there fits the "good-making role", but fails to qualify as truly "good" because its underlying natural properties differ from those natural properties which play the actual good-making role. (Compare: XYZ plays the watery role in Twin Earth, but fails to count as truly "water" because the latter term rigidly designates whatever actually plays the water role, namely, H2O.)
But that just seems transparently absurd. Moral properties aren't to be held hostage to actual world contingencies. Ethics isn't a matter of natural kinds and rigid designators. If something satisfies the "good-making role" in a world, then that is good; it doesn't matter if its underlying properties differ from those of its functional analogue in our world. The primary and secondary intensions of 'good' ought to coincide. So if some moral truths are necessary truths, they must also be a priori.
Now, moral facts clearly supervene on the natural, if they exist at all, and hence give rise to necessary conditional truths. So (as we've seen) these truths must also be a priori. Hence, the only adequate realist naturalist meta-ethics is analytic ethical naturalism. The challenge remains to provide the conceptual analysis in question, or at least indicate how a reduction of the normative to the natural might go.
(See also the relevant section of my old essay on the fact/value gap, and the reference to related arguments in Michael Smith's excellent article, 'Moral Realism'. He points out that synthetic naturalists still need to provide an a priori analysis of the "good-making role", prior to rigidification under this role. He also reinforces the crucial point that Moore's open question argument merely shows that no reductive analysis of moral terms is obvious, not that none is possible.)