You can either point the first proxy to the second proxy, or point it to the backends directly. Depends if you have firewalls in the way that stop the VPS proxy reaching your backends directly; or if that internal nginx instance is dong anything clever like handling auth, adding headers etc. etc.
In your instance I'd more likely have the VPS locked down and unable to access my internal resources and just open up its access to my internal nginx instance. Therefore chaining proxies would be my approach but there's no right or wrong.