I haven't looked closely to figure out the exact nature of their sins, but when I fire up my own copy of WWW::Mechanize, I see that that there is only one link on the page /nature/current/index.html.

It is therefore a trivial matter to fetch the link path, prefix it with the protocol and host, and fetch the resulting address:

$mech->get('http://www.nature.com/nature/current/index.html');
my $link = 'http://www.nature.com' . ($mech->links)[0];
$mech->get($link);
[download]