My institution subscribes to Nature, and using my library's proxy to access the Nature website, I can use the "Share/bookmark" menu to generate links like http://rdcu.be/bKk4, http://rdcu.be/bKlc, http://rdcu.be/bKld, and http://rdcu.be/bKli, which can be viewed in the browser (or maybe only because I also just installed the ReadCube app?).

The articles linked to above span several months, but it's generating serial links, so I can only assume that it's able to track visits back to the subscriber and/or my university account.

The ReadCube HTML5 reader looks nice, but does not work with JavaScript disabled (no surprise there). It uses JavaScript to override text selection (disabling copy&paste), but after a little meddling with the developer tools and element inspector, you can find a decently near ancestor to the text and copy the DOM as html. Stick that into a new file and you can select (and copy) the text without too much further hassle.

The DOM is awkward and split up kind of like a PDF (selecting a range of text goes haywire in unpredictable cases), but in comparing the HTML DOM hierarchy to the text object structure in the original PDF (which, as a subscriber, I can download), I found no obvious similarities, so I'm guessing they aren't translating the PDF to HTML directly.