I know how arp poisoning works in the LAN environment and I know that routers also contains mac-address tables with the neighbouring router's mac-addresses, that means routers exchange arp frames. So is there any possibility to send the arp poisoning attack to ISP level? Till now is there any attacking techniques exists regarding arp poisoning in WAN. If not how the attackers are doing man-in-the-middle attacks & what is the security mechanism the routers use to stop these attacks?
Based on your question, I don't think you understand ARP poisoning as well as you think you do.
In most cases, WAN links are point to point, so there's nowhere to redirect traffic to.
If you're not within the layer 2 broadcast domain, you won't be able to send or receive ARP messages, legitimate or otherwise.
If the WAN is a shared ethernet, maybe. That tends to only happen in cheap/lazy data centers. Common (residential) WAN protocols like DSL and DOCSIS are point-to-point; while they may be ethernet and appear to be a shared medium, all traffic flows through the ISP headend (DSLAM/CMTS), CPE's cannot directly communicate.
Beyond that, ISP technology intended to be used in such a manor has various security mechanisms to prevent this sort of thing.