I fail to see the difference and I work in wet neuroscience.
Let's translate to rough computational terminology. Your serotonin signaling (depression is actually more complicated and the 5HT hypothesis is almost surely wrong, but you stated it) was compromised and the computation was returning results outside the norm for your physiological optimal.
Now you, as a creature perceiving inside-the-loop and inside the algorithm, seem to believe because this is being computed using SERT and DAT and the plethora of 5HTx receptors on both sides the synapse, this is in someway different than a computer having an arbitrary "5HT signal" which weighs out state. I can assure you it's not and your just being a silly human.
EDIT: In the not so distant future, we will shed these legacy ideals of [evil] consciousness and sentience and start looking at computation as the physical and universal process that it is and and realize that we already pause and stop and slow down our computation when we die and sleep and go under anesthesia or coma. We will routinely be running super complicated routines that are subcomponents of us on alternative substrates that are out-of-order, asynchronous, non-local and such. We might spin up a subroutine to do some work on a project in Australia or Mars, then shut it down. Now, what to do with the original copy that likes feeling alive...