From my experience of job interviews at Microsoft I can tell that I never got just a direct question about graphs, but many questions just implied the knowledge of graph algorithms.
The most common topic here are trees as data structures and algorithms on trees: breadth-first and depth-first searches, building the minimal spanning trees.
For directed and/or weighted graphs finding the shortest path between two nodes (the known algorithm of Dijkstra) is a very common problem and I saw it in many collections of sample interview questions.
There are many data structures that are based upon trees, for example heaps, that are quite common in interviews and you should be at least aware of why they are organized this way.
Less common are the algorithms that include the maximal flow calculation, coloring and other specific problems.
The question can be formulated in quite an absract way. During my interview for an intern developer position at Microsoft, for example, I was asked to implement a short algorithm to generate all possible letter combinations for a given phone number and I used tree traversion to derive a recursive algorithm, to discuss its pros and contras and to turn it finally into an iterative one.