Plants produce specialized metabolites for their defence. However, specialist herbivores adapt to these compounds and use them for their own benefit. Plants attacked predominantly by specialists may be under selection to reduce or eliminate production of co‐opted chemicals: the defence de‐escalation hypothesis.
We studied the evolution of pyrrolizidine alkaloids (
PA
s) in Apocynaceae, larval host plants for
PA
‐adapted butterflies (Danainae, milkweed and clearwing butterflies), to test if the evolutionary pattern is consistent with de‐escalation.
We used the first
PA
biosynthesis specific enzyme (homospermidine synthase,
HSS
) as tool for reconstructing
PA
evolution. We found
hss
orthologues in diverse Apocynaceae species, not all of them known to produce
PA
s. The phylogenetic analysis showed a monophyletic origin of the putative
hss
sequences early in the evolution of one Apocynaceae lineage (the
APSA
clade). We found an
hss
pseudogene in
Asclepias syriaca
, a species known to produce cardiac glycosides but no
PA
s, and four losses of an
HSS
amino acid motif.
APSA
clade species are significantly more likely to be Danainae larval host plants than expected if all Apocynaceae species were equally likely to be exploited.
Our findings are consistent with
PA
de‐escalation as an adaptive response to specialist attack.