Hi,
First, in general MOS process, the source and drain are isolated with PN junctions. In order to ensure proper operation of MOS device, the PN junctions have to be reverse biased.
Secondly, as you know, NMOS has a P-type substrate with N-type source and drain, while PMOS has a N-type substrate with P-type source and drain. In order to reverse bias the PN junction, you have to connect the substrate of NMOS to GND and the substrate of PMOS to VDD.
You might probably argue that one does not have to do that to achieve a reverse bias. Yes, as long as you maintain the voltage of N side higher than the P side, you can achieve a reverse bias. But the problem is in a representative N well process, the P type substrate is the universal substrate of all NMOS devices on a single die. You simply don't know at what voltage will the PN junction safely biased, except the lowest available voltage GND. The same applies in N type substrate.
Hope this helps
regards
ceyjey