I use a planes on the center of the wheel mated a distance from the top of the V angle and a second mate to keep it along the center of the angle.

Two tangent mates also work, from either side of the angle to either side V-bearing.  Distance and coincident mates seem to be more stable to me and that is why I use them.

I have tried this both ways as described by Wayne, and I have to say that the model will likely be more stable and run quicker if you create some equivalent reference geometry to use for the actual mating. It is my preferred method.

However, that practice does have the down side that you may need to periodically manually check the fit between the wheel and the rail against unintended changes to the interface due to changes to other components - your reference geometry may not keep up with other changes in your assembly unless it is based on bulletproof context-based geometry. Unintended changes when you are using tangent mates are more likely to throw up mate errors and often are more obvious to the eye.

One way to do it is set up a width mate between the sides of the bearing and the sides of the rail.  Then do a tangent mate between one of the V-edges and the corresponding V-edge of the rail.