There was a proposal for this in the 90s called Hashcash. Senders would have to brute force some random data to append onto email headers until they found some such that the first n bits of the hashed header were 0s. Sending emails would be computationally expensive, but not to a point that it would impact non-bulk users.

The proof-of-work email use case never took really took off, but if you're acquainted with cryptocurrencies some parts of the implementation might sound familiar :)