The central problem with evolving a wheel is that wheels pretty much need to be totally separate elements from the animal at large.
Think about how a wheel rotates. If it just spun and spun, there could be no connections to the animal at large. No nerves, no blood supply, nothing, because such connections would get twisted off after only one full rotation.
Are there any workarounds? Yes.
1. Wheels are dead and/or inorganic. The animal in question places a wheel stub into a muscular orifice which then rotates the wheel - and replaces them as they wear out. It's hard to imagine how this could evolve. Maybe an animal, for some reason, begins running on spherical seed pods like someone balancing on a ball?
2. Wheels evolve as a symbiosis between two animals. this is a little bit easier. Let's say a limbless animal (snail, snake, etc) develops a relationship with another animal which already has a habit of rolling quickly downhill to pursue or avoid prey. At first, the snake-thing just rides along the top, surfing along with them, but with time it can better grip the other animals without slowing down the speed of rotation. Give it a few million years and you might have effective "roller" wheels. How optimized the wheels could develop probably depends upon degenerate the wheel-symbionts could get. I can't see vertebrates - even as symbionts - having their mid-body developing quite into an axle.