So Eisen is correct, it's the same thing. A non-open access publisher could make a non-open access equivalent of PLOS ONE if they wanted to (and maybe somebody has). It's not a question of open access.
But it is worth noting that a move to APCs dramatically changes the incentive structure for a publisher, and there are built-in mechanisms in such a system to produce new strange dynamics that we haven't really seen much of before (ie the tension between needing to accept a certain volume to stay in business vs try to keep the reputation of the journal high).
Consider that PLOS ONE, because it doesn't review for importance, has a much higher acceptance rate. Only accepted papers pay processing charges, so each accepted article has to subsidize the review and rejection of all the rejected papers. In PLOS Biology or PLOS Medicine, with a low acceptance rate, each accepted article has to subsidize review of numerous other articles, driving the cost per published article way up.
"The costs are largely the same between open-access journals and non-open-access journals," said Michael Eisen, a co-founder of the open-access project PLOS. "We’re doing essentially the same thing as traditional publishers."
The type of peer review that megajournals (like PLOS One) do is categorically different than a lot of traditional journals. A megajournal will typically review for scientific soundness only. If the methodology is sound then the paper should be accepted regardless of whether it breaks new ground in the field or is seen as important.
This is not the same kind of peer review that you will find when submitting to Nature, where they are trying to curate for importance in addition to vet for soundness. I won't argue one way or the other about whether one type of peer review is better than the other, there are arguments to be made on all sides. But they are not the same, and they do not cost the same amount.
Methodological peer review is cheaper. That's why the megajournals like PLOS One, et al do it that way (well, it's one of the reasons, the other main one being a philosophical reason). You could not run the big OA operations at their price points with traditional peer review. Their revenue is purely driven by acceptance volume, and so therefore anything that slows down peer review or rejects too many papers has a serious material impact on revenue. It's a balancing act between ensuring you don't accept quack science accidentally but don't reject too much than you have to.