If you are interviewing for a Ph.D. position in Germany, the supervisor will probably already know what you will be working on, since he will have gotten a grant on some particular topic, and you will be the warm body to actually do the work. So, in addition to everything else already mentioned:
What is the specific project that will pay your salary? Can the supervisor send you a copy of the successful grant proposal? Read that proposal carefully, especially concerning any timetables and financial information.
How much leeway will you have in actually implementing the grant proposal?
Presumably, the grant proposal builds on previous work by the supervisor's working group. Can you talk to current Ph.D. students working on the topic? What are the specific challenges in this topic? Would it be possible to get any additional information, like manuscripts? Will you have to take over existing stuff, like experiments that are already running, or legacy code (which pretty much nobody wants)?
The goal would be to get a feeling for the topic you will be working on for the next couple of years. Will this be something you can get sufficiently excited about?
And of course there are additional questions. In no particular order:
What kind of supporting infrastructure is in place? Is there a full-time sysadmin, or will you keep your IT running by yourself? Is there a statistician to help you with data analysis? Somebody technical to keep the apparatuses running? Someone to feed the lab rats? Or will all this be your responsibility?
If you are interested in this kind of thing: is there a possibility for you to work in a different lab for a few months, perhaps abroad? This kind of thing can be enormously valuable for your network.
@aeismail already mentioned asking how often you will meet with your supervisor, and I think this is one of the most important questions. If your supervisor tells you that he meets with each and every Ph.D. student once a week for half an hour, great. If you only meet once a semester when you give a seminar talk... not so great.