The Wilson Current mirror doesn't provide much advantage when using MOS. It's biggest strength is eliminating base current error when using BJT's.
However, here's how it functions in your configuration:
Assuming initially all devices are off, the following sequence of events will happen during start-up:
1) I4 will start increasing voltage vz.
2) When vz > VT, M5 will start increasing voltage vy.
3) When vy > VT, M3 will start to conduct current. It will keep increasing until it sinks all of I4.
The net voltage will be:
vy = vgs of M3 necessary to pass all of I4 current.
vz = vy + vgs of M5 necessary to pass all of ratioed current in M4. So if M4 = M3, it will be the vgs necessary to pass I4 current.