Roughly speaking, you can't. There is no good way to do what you want. All of the possible approaches have major drawbacks.
You can turn on LUKS encryption of the partition. However, this has two significant disadvantages:
Someone will need to enter the password in at the console each time the machine boots. If the machine gets rebooted, it will sit there hanging, waiting for someone to enter the password manually, before it can start the email server and access the encrypted data. For most scenarios, that's not acceptable.
More significantly, the scheme provides limited or no security. Suppose the server gets compromised, and the hacker gains control of the root account (or the account that the mailer runs under). Then the hacker will have all the same access that the mailer program does. In particular, this means the hacker will be able to access the decrypted emails -- so the encryption doesn't actually do any good against this threat. The only threat it protects is the possibility that the hardware might be lost or stolen; if the thief powers down the equipment, then they won't be able to regain access to the data, since they won't know the LUKS password.
If you think about it, there is no good way to avoid these security limitations -- they are pretty much unavoidable. If a hacker compromises your mailer, they'll be able to see all your email, in cleartext. (See also my answer elsewhere for more.)
Therefore, I think you need to think through carefully what you are trying to achieve. Cryptography is not magic pixie dust that you can sprinkle on a system to make it secure. By asking "how do I use cryptography?", you have zeroed in on a particular mechanism and implicitly assumed that cryptography is going to be the solution to your problem; but in this situation, that may not be the case. Therefore, I suggest you avoid such implicit assumptions and see if you can articulate what problem, exactly, you are looking to solve.
The standard way to think through your goals carefully is to work out what is your threat model. If you don't have a clear threat model, you're going to have a hard time devising a useful scheme. A threat model should identify what kinds of threats/attacks/risks you are trying to defend against, and which ones are out of scope. Once you have a threat model, one can start to consider what is the best mechanism. If you want to revise/update your question with a description of what you are trying to achieve and the threat model you have in mind, we can try to provide you with more helpful advice.