For a one-time \$99 fee, anyone can publish a single paper a year for life (although the first dozen authors on the paper all have to sign up). \$259 buys any author a lifetime membership, with the ability to publish as many papers as they choose.
PeerJ has several interesting business twists that seem to make this feasible, but the one I find the most ingenious is this one: "For long term archiving, the publication will be placed at the National Institutes of Health's PubMed Central archive." The article isn't clear, but if this means that PeerJ isn't hosting their older articles on their own servers, they've brought a cool new twist to publishing.
Traditional journal publishers have two common critiques about open access, (which I am not really qualified to judge, but they do sound plausible a priori):
Charging authors one time up-front won't cover the ongoing future costs (which theoretically keep going forever) to hosting a journal article.
Mandating PubMed Central archiving removes long-term revenue by limiting how much publishers can make serving articles, which means they can't cover the long-term costs.
PeerJ would turn these business critiques on their head by essentially offloading long-term archiving costs onto the Federal government via the PubMed Central NIH mandate. They managed to kill two birds by throwing them at each other!
On second thought, since PeerJ does intend to host post-publication review, comments, and other epi-article materials, it doesn't seem like unloading the articles on PubMed Central will do much for their costs. Maintaining user-generated data like that will incur far more manpower and server costs than the papers, since papers are a fairly static medium aside from errata and retractions.