As the others have said the P and N channels go can on at once, so this is a bad circuit.
As this is only running on 5V, you can probably connect M3-Gate to IN1. (I assume cmos 555 with 0-5V signals) Then it is effectively two cmos invertors chained as long as Rload is not too high
It is however very interesting that it worked on your breadboard, and using a scope on your breadboard setup might be very interesting.
[Don't know why I thought M1,M2 meant motor - oh well]
An actual DC motor for Rload has discontinuous currents as the commutator goes around, and flyback and back emf voltages. Protoboards may turn the current pulses into voltages (e.g. at the sources) that effect positive feedback. Or stray capacitance my be interacting with voltage spikes.
Try putting a 100n capacitor across the motor (on your protoboard circuit) and see if it still works. Try a 1000uF cap across the 5V. Does it still work? (I see your pcb has 10000uF. Did the protoboard?)
Another thing that happens with fets, is they can oscillate in the MHz when you get them anywhere near linear operation, especially with any sort of inductive load. That is why you see them with a 47ohm series resistor right at the gate. Breadboards exacerbate that sort of thing, however it might be making you circuit work by adding positive feedback.
If you have discovered the self-level shifting bridge drive, your name may yet join other hallowed minimalist giants: the inventor of the Regen radio, the teledyne opamp, and the discoverers of cold fusion.