Re: pervasiveness of surrogate (also called synthetic) keys

i.e for an Invoice, we have at least 2 tables (more in practice...):

Invoice Header

    -Invoice Number

    -Date

    -CustomerID

    -Currency

Invoice Line

    -ItemID

    -qty

    -Price

This is not real design for tables, just basic real info stored to represent 1 Invoice Document.

Now, let say for Invoice Header, "Natural" PK is Invoice Number, so we will add that column to Invoice Line table as FK....

What should be "Natural" PK for Invoice Line table?

Also, if we now add additional tables i.e. Serial Numbers, related to Invoice Line, what should be "Natural" key in it?

Thanks,

Misa