For sea slugs, via RNA transfer. At Discover:
There’s a couple of reasons why I don’t think this is evidence of “memory transfer”.
Firstly, what was transferred here was hardly a memory in the usual sense of the word. It is simply an increase in the sensitivity of a set of neurons, a single reflex pathway. This ‘memory’ is not specific to any particular stimulus. The training consisted of shocking the animals, which makes them more likely to withdraw in response to touch – not to shock, but any touch. It’s just “turning up the dial” on that reflex. It is hard to see how this relates to the far more complex types of memory in humans. More.
See also: Neuroscience: RNA can transfer memory in sea slugs Science Mag: “‘It’s a very controversial idea,’ admits study coauthor David Glanzman, a neuroscientist at UCLA.” Replication studies are needed here, if they are allowed.
and
Neuroskeptic: Research casts doubt on the idea that the brain is a machine with parts