PCC is used to distribute traffic among WANs. @anav’s suggestion prefers ISP 1 for LAN 1 and LAN 2, and prefers ISP 2 for LAN 3 and LAN 4. No traffic distribution is used.
You can simplify it by using a single common routing-mark for traffic of both LAN 3 and LAN 4.
You can improve it by using recursive next-hop search to monitor the transparency of the uplink all the way to the internet rather than monitoring just the availability of the ISP’s gateway router, using @Chupaka’s suggestion.
For the “common routing” (routing table called main, consisting of routes with no routing-mark), the route via ISP 1 is preferred because it has lower value of distance, so the route via ISP 2 only becomes active if the ISP 1’s gateway becomes unresponsive.
For the route via ISP 2 with the routing-mark, the failover to ISP 1 consists in a fallback to routing table main if no route with the required routing-mark is available.
So the route rules could look as follows (just an example to be adjusted to your reality):
/ip route rule
add interface=br-lan3 action=lookup table=prefer-isp2
add interface=br-lan4 action=lookup table=prefer-isp2