It depends on how the questionnaire was created. Many questionnaires are just thrown together and you can do whatever you like to them - as many people who have used SUS have shown. SUS is like a raft: it won't sink but the quality of the ride is terrible.
QUIS is slightly different in that it tries to target different aspects of the system being evaluated. I think you'll be safe enough if you take sections out of QUIS and use them as chunks. I'd be wary of editing or leaving out questions from each section though.
Please note that a LIKERT scale means a collection of items which has been shown to exhibit a fairly rigorously defined set of statistical attributes - it's not just a bunch of "1 to 5" questions! Likert scales are most probably the best measurement tools we have for quantifying user experience and in such scales it is utterly foolhardy to edit or otherwise tamper with the items. A good Likert scale is like a finely-constructed piece of clockwork: tamper at your peril.