My thought is that if you turn the 2D layer into a 3D layer you don't need any expressions other than the one that links the point of interest to the null. As long as you keep the POI null on the 2D plane, not moving it along the z axis, it should work nicely that way.
Agreed with Kallehikki, 2D should only be used for some necessary effects, and the 3D layer should be clamped down to your map, and scripts should be pulling the 3D data to the 2D null and the camera. You should not be pushing 2D data back to the 3D null.
Since there was too many steps/scripts to try and instruct, I just build a demo version for how I would do this.