Big thanks to Vic Jefferson, Nick Catanzaro, Ralph Baric and Michael Letko @fviromics.bsky.social and big congrats to
@jarelelgin.bsky.social
, our talented PhD student who led the work!! 🥳🥳

The bat viruses with the closest spike or loop2 to that of nvHKU5r-CoV were also sampled in provinces with primary mink fur processing and wholesale hubs, further implicating fur farming as an important risk factor for virus spillover.

Lastly we did comprehensive recombination and molecular dating analysis to explore the origins of this nvHKU5r-CoV. Our findings support that the virus transmitted from bats to mink within the last decade, consistent with an expansion of mink fur farming in China.

AlphaFold3 co-structure predictions were also consistent with entry when this single mutation was introduced, revealing the likely molecular interaction that enables hACE2 compatibility.

Using a safe, replication competent VSV system expressing the nvHKU5r-CoV spike we did passaging in mixed cells expressing both Pabr and hACE2. To our surprise, the spikes acquired a single mutation on the loop2, R548S, which enabled hACE2 usage!

Then we asked, since ACE2 ortholog usage seems to shift quite easily in these viruses, could nvHKU5r-CoV potentially use human ACE2?

We further confirmed our pseudovirus results with luciferase reporter live virus, showing that nvHKU5r-CoV specifically replicated well in both PabrACE2 and nvACE2 expressing cells.

Impressively, AlphaFold3 predictions of RBD-ACE2 pairs were very consistent with pseudovirus results! RBDs with ACE2 entry had consistent binding footprints while RBDs without entry were predicted to bind inaccessible parts of the ACE2s.

...the only two long loop2 RBDs that can use nvACE2 being nvHKU5r-CoV and Q265, but not WZ2! This shows that loop2 is a key determinant for enabling the use of non-bat ACE2s.

Taking these RBDs in the lab, we show with pseudovirus assays that all HKU5r-CoVs can use their native P abramus ACE2, but only few can use mink ACE2, especially ones with long loop2s...

This part of the RBD corresponds to a loop in the RBD structure called loop2, known to have different lengths between HKU5r-CoVs. By examining the evolution of the loop2 we show a single deletion to a shorter genotype which has in turn recombined into many different RBDs.

However, just like all CoVs merbecoviruses are notorious for recombining parts of their genome, especially the spike. Looking more closely, we detected one small part of the RBD where nvHKU5r-CoVs were curiously closer to another bat virus, Q265!

The mink HKU5 virus (nvHKU5r-CoV) was sampled in Shendong and seems to be overall closest to different bat HKU5s sampled in P. abramus bats across south and east China. 🦇
When focusing on the spike and RBD, one virus, WZ2, sampled in Zhejiang, is the single closest relative.

This was a large team effort. Big thank you to our structural collaborator, James Rini at University of Toronto, Craig Wilen at Yale, the Baric lab at UNC, our Italian collaborators for discovering these viruses and sharing cool sequence data, and of course, Vic, the talented postdoc in my lab.(6/6)

To expand on the RBD clade classifications from our last paper:
CLADE 1 = MERS-CoV, HKU4 = DPP4 users
CLADE 2 = HKU5 complex, HKU25, H.Savii = ACE2 users
CLADE 3 = PDF2180, NeoCoV = ACE2 users
CLADE 4 = ErinCoVs (all of them) = APN users

Such receptor diversity for one subgenus! 🤩
(5/6)

We also provide not one, but TWO cryo-EM co-structures of ErinCoV RBDs bound to hedgehog APN. Both German and Italian hedgehog coronaviruses interact with hedgehog APN identically. Binding assays show that ErinCoVs have a lower affinity for APN than other CoVs have for their receptors. (4/6)

Aminopeptidase N- the classic alpha-CoV receptor - facilitates ErinCoV entry. And we don't show this for just one virus, we show it for all of them. Every merbecovirus in hedgehogs across Europe and Asia- from ErinCoV-12-19 to HKU31 - all use APN. This is a first for beta-CoVs! (2/6)