Science - 06.12.2019

SCIENCE sciencemag.org

GRAPHIC: V. ALTOUNIAN/

SCIENCE

spectrum of cargo proteins and cargo

recognition is profoundly affected by

the length of the cargo’s TM domains

rather than by a specific amino acid

sequence motif (or motifs) ( 4 ).

In this context, it came as a sur-

prise that in mammals, two of the

four cornichon homologs, CNIH2

and CNIH3, are specialized toward

AMPARs, the most abundant ex-

citatory neurotransmitter receptor

in the brain, by recognizing the

pore-forming GluA1–4 subunits

as exclusive “cargo” or interaction

partners ( 5 ). However, the CNIH2-

GluA and CNIH3-GluA interac-

tions differ from the CNIH1-cargo

interaction in two characteris-

tic ways. One is that CNIH2 and

CNIH3 can only bind to their cargo

in the ER after GluAs have been as-

sembled into tetramers, which, in

neurons, requires an ER-based as-

sembly machinery ( 6 ). Another is

that CNIH2 and CNIH3 do not re-

lease their cargo, but rather remain

tightly associated and travel with

the AMPARs to the plasma mem-

brane, where they have been found

in both synapses and extrasynaptic

sites ( 7 – 9 ) (see the figure).

Thus, in addition to their role as

cargo exporters, CNIH2 and CNIH3

are auxiliary subunits of AMPARs that have

a profound impact on the gating properties

of the receptor channels: Both CNIH2 and

CNIH3 stabilize the AMPAR channels in

the open or conducting state(s) by slowing

the deactivation (channel closing after ag-

onist removal) and desensitization (chan-

nel closing with agonist bound) processes

( 5 ). In excitatory synapses, such prolonged

opening enhances the influx of Na+ (and

Ca2+) into the postsynapse, which, through

increased and longer-lasting depolariza-

tion of the postsynaptic neuron, leads to

more reliable propagation of the electrical

signal. This was directly observed in hip-

pocampal neurons ( 7 ).

For all distinct cornichon functions, the

characterization of CNIH3-GluA2 com-

plexes by Nakagawa provides interesting

insights. The most intriguing and unpre-

dicted features of this cryo-EM structure

are the topology and overall folding of

CNIH3: The protein displays four TM do-

mains with amino and carboxyl termini

facing the ER lumen (which becomes the

extracellular side of the plasma mem-

brane). Most parts of the protein are bur-

ied in the membrane; only the linker be-

tween TM domains 3 and 4, and parts of

TM domains 1 and 2, extend into the cy-

toplasm. This unexpected topology may

be initiated by a 12–amino acid stretch

at the amino terminus of CNIH1–3 that is

highly conserved among mammals, flies,

and worms and operates as a signal pep-

tide that is not cleaved off [therefore called

“uncleavable membrane inserting peptide”

(UMIP) by the author]. The amino-termi-

nal half of the UMIP contains three con-

served phenylalanine residues that make

close contact with the TM domains of the

GluA subunits.

Projection of the primary sequence(s)

onto the CNIH3 structure highlights a 4.5-

turn a-helical extension at the cytoplasmic

side of TM2 as the major difference be-

tween the GluA-specific CNIH2 and CNIH3

and the cargo receptors CNIH1 and CNIH4.

This extension, together with the helical

domain of TM1 and the (structurally unre-

solved) linker, represents the hallmark fea-

ture that distinguishes the CNIHs from the

other inner core constituents of AMPARs,

the TARPs (transmembrane AMPAR regula-

tory proteins), and GSG1L (germ cell–spe-

cific gene 1–like protein) ( 10 , 11 ). TARPs and

GSG1L also modulate channel gating and

comprise four TM segments, but their “ex-

tensions” are directed toward the extracel-

lular side of the membrane where they can

contact the ligand-binding domains (LBDs)

of the GluA subunits during the conforma-

tional changes that drive channel

gating after agonist binding ( 11 ).

Because CNIH2 and CNIH3 lack

contacts with the LBDs, Nakagawa

suggests that structural rearrange-

ments induced by CNIH3 in the

“filter” region of the receptor chan-

nel likely explain the pronounced

effects on channel gating. The fil-

ter contains a particular site that

controls Ca2+ permeability and may

influence the open-state configura-

tion of the receptor channel ( 12 ).

Whether these rearrangements are

relevant for AMPAR gating remains

unclear, but they are strongly remi-

niscent of the “selectivity-filter gate”

that opens (and closes) various

types of K+ channels in response to

distinct ligands and pharmacologi-

cal agents ( 13 ).

The CNIH3-GluA2 complexes are

rarely (if at all) observed in the ro-

dent brain. Most CNIH-containing

AMPARs (a total of ~80% of all

AMPARs) have both CNIH and

TARP proteins co-assembled in a

2:2 stoichiometry, as estimated by

quantitative proteomics ( 14 ) and

confirmed by the cryo-EM analysis

of native AMPARs ( 15 ). Most of the

successfully resolved structures of

native AMPARs ( 15 ) have contained

two TARP subunits and two subunits with

four densities in the membrane plane; the

CNIH3-GluA2 structure from Nakagawa

suggests that these may be CNIH2 and/or

CNIH3 proteins.

The unexpected structure of CNIH3 and

its distinction from the TARPs will provide

a blueprint for coming analyses of its role in

gating and trafficking of AMPARs, but also

in studying the ER export processes medi-

ated by the CNIH family. Finally, it should

not be forgotten that predictions from data-

bases and algorithms are good, but rigorous

experimental data are better. j

REFERENCES AND NOTES

1. T. Nakagawa, Science 366 , 1259 (2019).


2. S. Roth et al., Cell 81 , 967 (1995).


3. C. Bökel et al., Development 133 , 459 (2006).


4. Y. H e r z i g et al., PLOS Biol. 10 , e1001329 (2012).


5. J. Schwenk et al., Science 323 , 1313 (2009).


6. J. Schwenk et al., Neuron 104 , 680 (2019).


7. S. Boudkkazi et al., Neuron 82 , 848 (2014).


8. B. E. Herring et al., Neuron 77 , 1083 (2013).


9. A. S. Kato et al., Neuron 68 , 1082 (2010).


10. B. Herguedas et al., Science 364 , 353 (2019).


11. E. C. Twomey et al., Nature 549 , 60 (2017).


12. I. D. Coombs et al., Nat. Commun. 10 , 4312 (2019).


13. M. Schewe et al., Science 363 , 875 (2019).


14. J. Schwenk et al., Neuron 74 , 621 (2012).


15. Y. Zhao et al., Science 364 , 355 (2019).

ACKNOWLEDGMENTS

We thank members of the Fakler lab for discussions.

10.1126/science.aaz8642

At the ER

CNIH2 and CNIH3

specifcally bind to

AMPARs through

interactions with

the amino terminus

and the peptide

extensions of the

cytoplasm-facing

region of these CNIHs.

At the synapse

CNIH2 and CNIH3

are auxiliary

subunits of AMPARs

that stabilize the

channels in the

open state and

thereby enhance

signal transduction.

GluA

subunit

CNIH2

or CNIH3

Coat protein complex II

CNIH1

or CNIH4

Secreted

cargo

Endoplasmic reticulum

Postsynaptic

neuron

Presynaptic

neuron

Transmembrane

AMPAR

regulatory

protein

Functions

of cornichon

proteins

Cornichon homolog 2

(CNIH2) and CNIH3

control endoplasmic

reticulum (ER) export

and gating of AMPA-type

glutamate receptors

(AMPARs, comprising

GluA subunits). CNIH1 and

CNIH4 are nonspecific

cargo receptors.

6 DECEMBER 2019 • VOL 366 ISSUE 6470 1195

Published by AAAS

on December 12, 2019^

http://science.sciencemag.org/

Downloaded from