You are on page 1of 32

CHAPTER SEVEN

Crosslinking glutamate receptor


ion channels
Andrew J.R. Plesteda,b,c,* and Mette H. Poulsend
a
Institute of Biology, Cellular Biophysics, Humboldt Universit€at zu Berlin, Berlin, Germany
b
Leibniz Forschungsinstitut f€ur Molekulare Pharmakologie (FMP), Berlin, Germany
c
NeuroCure Cluster of Excellence, Charite–Universit€atsmedizin Berlin, Corporate Member of Freie
Universit€at Berlin and Humboldt-Universit€at zu Berlin, and Berlin Institute of Health, Berlin, Germany
d
Department of Drug Design and Pharmacology, University of Copenhagen, Copenhagen, Denmark
*Corresponding author: e-mail address: andrew.plested@hu-berlin.de

Contents
1. Introduction 162
2. Receptor crosslinking with introduced disulfide bonds 164
3. Photocrosslinking with genetically-encoded unnatural amino acids 165
3.1 Control experiments 168
3.2 Using light-activated UAAs with electrophysiology 171
4. Fast perfusion coupled to chemical modification and state-dependent
crosslinking 172
4.1 A four-parallel-bore perfusion tool for fast solution application and chemical
modification 172
4.2 Materials 172
4.3 Preparing the fused silica capillaries 174
4.4 Pulling the 4-bore glass 174
4.5 Bending the tip 176
4.6 Cutting the nose piece off 177
4.7 Etching the tip 177
4.8 Assembling the tool 178
4.9 General advice 179
5. Biochemistry: Photoactivated Crosslinkers and purification of GluA2 180
5.1 Materials 181
5.2 Buffers 181
5.3 Day 1: Cell culture 183
5.4 Day 2: Transfection and UAA labeling 183
5.5 Day 5: UV treatment, harvesting and lysis of cells 183
5.6 Day 6: FLAG-TAG purification and Western blot 184
6. Modeling of the effects of crosslinking 185
6.1 Disulfide bonds and bifunctional crosslinkers 185
6.2 Photocrosslinking 187

Methods in Enzymology, Volume 652 Copyright # 2021 Elsevier Inc. 161


ISSN 0076-6879 All rights reserved.
https://doi.org/10.1016/bs.mie.2021.03.005
162 Andrew J. R. Plested and Mette H. Poulsen

7. General limitations 187


8. Summary and perspectives 188
Acknowledgments 189
References 189

Abstract
Combining crosslinking strategies with electrophysiology, biochemistry, and structural
in silico analysis is a powerful tool to study transient movements of ion channels during
gating. This chapter describes crosslinking in living cells using cysteine and photoactive
unnatural amino acids (UAAs) that we have used on glutamate receptor ion channels.
Here, we share the protocol for building a perfusion tool to enable rapid chemical mod-
ification of glutamate-gated AMPA receptors, optimized for their fast activation. This sys-
tem can be used to perform state-dependent crosslinking in receptors modified by
cysteines or UAA incorporation on the millisecond timescale. Introducing UAAs results
in receptors with lower expression levels relative to the introduction of cysteine resi-
dues. Reduced expression is principally a challenge for biochemical studies, and we
share here our approach to capture the light driven oligomerization of AMPA receptors
containing UAA crosslinkers. Finally, we describe strategies for computational analysis to
make sense of the crosslinking results in terms of structure and function.

1. Introduction
Ion channels and receptors are dynamic molecules that exist in a vari-
ety of conformations. The physiological role of conformational change is
evident for nearly half a century from the open-closed nature of single
channel recordings (Neher & Sakmann, 1976). However, the manner of
the structural changes is not. The cartography of ion channel activation
reveals their operating principles, and potentially opens new sites for exploi-
tation by drug-like molecules. Movements supporting characteristic transi-
tions between ion channel states (for example, isomerization between open
and desensitized states) span a range of distance scales, from side chain flips
(Cordero-Morales et al., 2006) to major translations and rotations of domains
(Meyerson et al., 2016). Since the advent of site-directed mutagenesis, cys-
teine reactive chemistry has been used to probe these motions (Karlin
& Akabas, 1998). More recently, co-opting transfer-RNAs (tRNAs) and
RNA synthetases from other trees of life has enabled the genetic encoding
of “foreign” amino acids with a wide range of chemical groups (Chin,
Martin, King, Wang, & Schultz, 2002), including photoactive crosslinkers
that may be introduced to proteins in mammalian cells (Hino et al.,
2005). Since its introduction (Wang et al., 2007), this technique has gained
traction among ion channel investigators.
Crosslinking glutamate receptor ion channels 163

When used as a complement to the rapidly-expanding catalog of ion


channel and receptor structures, such crosslinks, or trapping bridges, allow
hypothesis-driven testing for functional motions. The principle of state-
dependent crosslinking is to define functionally-relevant conformations
by catching receptors in the act (Fig. 1). In particular, dynamics on the

Fig. 1 Disulfide bonds and photo-crosslinking with unnatural amino acids (UAAs).
(A) Cysteine residues for forming disulfide bonds can be accessed on the extracellular
(Continued)
164 Andrew J. R. Plested and Mette H. Poulsen

millisecond scale can be detected (Ding & Horn, 2001; Salazar, Mischke, &
Plested, 2020), similar to the limit of what can be captured by electron
microscopy (Unwin, 1995). In principle, unstable or otherwise hidden states
can be isolated, as has been achieved for transporters (Reyes, Ginter, &
Boudker, 2009), although it is not clear that crosslinks always trap physio-
logical states. Trapping bridges are (perhaps surprisingly) quite selective in
their action, showing little evidence for non-selective action between intact
membrane bound receptors, for example (Lau et al., 2013). An important
related application permitted by all crosslinking techniques, but particularly
convenient with photoactive unnatural amino acids (UAAs) is peptide
photo affinity labeling (Seidel, Zarzycka, Zaidi, Katritch, & Coin, 2017;
Borg et al., 2020), which can, in principle, be performed symmetrically
(i.e., crosslinking group on either host receptor or peptide ligand).
In this chapter, we discuss two methods that we have employed in glu-
tamate receptor ion channels, crosslinking using cysteines, and crosslinking
using photoactive UAAs. Each approach has advantages and disadvantages—
in combination, they cover most needs. We also describe some computa-
tional techniques to assist in interpretation of the results.

2. Receptor crosslinking with introduced


disulfide bonds
Cysteine crosslinks have been used in the context of electrophysiol-
ogy, biochemistry, and crystallographic studies of glutamate receptors.
Disulfide bonds typically add considerable stability to proteins, but can also
be unstable depending on local chemistry (Heras, Edeling, Schirra, Raina, &
Martin, 2004). In particular, unstable disulfides show a longer thiol-thiol

Fig. 1—Cont’d side of glutamate receptors—that is, the same side as solution access
to activate the receptor with glutamate. Pairs of sites must be screened. (B) UAA
crosslinkers activated by light can be installed at single sites in any part of the receptor,
except the C-terminus, because in this region, a premature stop codon does not pre-
vent functional expression. (C) Following a change in conformation, redox chemistry
can reversibly trap the receptor with a disulfide bridge. (D) Following conformational
change, UV light exposure can crosslink or perturb the structure. (E) Photochemistry
of azido-phenylalanine (AzF) and benzoylphenylalanine (BzF) protein crosslinking.
The aryl azide undergoes irreversible conversion to nitrene following photon absorption
and feeds the major crosslinking pathway of ring expansion and crosslink formation to a
primary amine on the substrate, R. Several minor pathways exist including direct linking
to amines. For BzF, reversible formation of a ketyl triplet radical (lifetime 1 ms) leads
to hydrogen abstraction from carbon in good context and recombination with the
substrate, R.
Crosslinking glutamate receptor ion channels 165

distance on average than stable ones (2.18 Å vs 2.05 Å (Sun et al., 2017).
Residues on twofold axes between subunits are ideal targets for cysteine
crosslinking, because a rapid scan of single mutants can yield interesting
functional phenotypes that may also be amenable to structural studies
(Armstrong, Jasti, Beich-Frandsen, & Gouaux, 2006; Lau et al., 2013).
Bifunctional crosslinkers that span two cysteines can be employed if the
residues do not come into close contact (Armstrong et al., 2006; Tajima
et al., 2016). With some care, paired cysteines that form disulfides after a
conformational change can also be targeted (Baranovic & Plested, 2018).
Such crosslinks based on structural homology can be used to investigate
domain organization for receptors before structure was available (Das,
Kumar, Mayer, & Plested, 2010). Metal bridges formed from groups of his-
tidine residues are more flexible to use and better tolerated than cysteine
bridges (Baranovic et al., 2016; Lau et al., 2013). However, these bridges have
a low success rate, even when designed from known structure, and metal
binding is often much less avid than cysteine bridge formation (Harding,
Nowicki, & Walkinshaw, 2010). These bridges are less appropriate for bio-
chemistry or structural analysis as a result. Cysteines in membrane domains
can create metal ion binding sites, but only in solvent-accessible regions,
which tends to limit their scope to pore lining residues (Sobolevsky,
Yelshansky, & Wollmuth, 2004) or voltage sensors (Yang & Horn, 1995).

3. Photocrosslinking with genetically-encoded


unnatural amino acids
Bearing in mind the constraint of solution access to accessible
cysteines, genetic encoding of commercially available photocrosslinkers
(Ye et al., 2008; Ye, Huber, Vogel, & Sakmar, 2009) is an inexpensive
and technically straightforward approach that requires little expertise in
chemistry. Light-activated crosslinkers are in principle applicable in any
environment, and available chemical groups are demonstrated to offer a bet-
ter time resolution (Ding & Horn, 2001) than cysteine bridging. Most
importantly, when performing scanning approaches, because all crosslinking
potential is contained within a single site, many more sites can be examined.
Compared to disulfide crosslinking, where search must proceed across pairs
of sites (N2), single site crosslinkers require only the square root of effort
(N, see Fig. 1). Cysteine mutants are introduced directly, by mutagenesis,
but the process of UAA incorporation is more involved. Various schemes
exist, usually based on repurposing Stop codons. The Amber Stop codon
166 Andrew J. R. Plested and Mette H. Poulsen

Fig. 2 Unnatural amino acid incorporation. (A) Endogenous translation components


(green) and the equivalent components for genetically-encoded unnatural amino acid
(UAA) mutagenesis (purple), from another tree of life. Transcription of the gene of inter-
est with an early introduced amber stop codon (red) yields mRNA with an early UAG
codon. (B) The early UAG in the mRNA leads to a truncated protein in normal conditions.
In the case of perfectly orthogonal translation systems, the exogenous synthetase char-
ges an exogenous amber-suppressor tRNA, “rescuing” the Amber stop codon and thus
expression of folded, labeled protein (cylinder). (C) 4 modes of non-orthogonal activity
by the exogenous translation machinery that reduce or abolish labeling. Incorrect amino
acylation involving mischarging of exogenous or endogenous tRNAs by their cognate
synthetases, or misrecognition of tRNAs by the non-cognate synthetases (less common)
are shown. The folded protein has uncertain incorporation of UAA and possibly no label-
ing at all. (D) Readthrough of amber codons by endogenous suppression elements is
variable by site and probably depends on local sequence, and can reduce labeling.
Readthrough can occur without any exogenous elements, or in their presence.

(TAG) is usually chosen because it has the lowest incidence, representing less
than 1 in 4 stop codons in humans (Liu, Brock, Chen, Chen, & Schultz,
2007) and less than 10% of those in Bacteria. Briefly, the site of interest is
replaced with an Amber codon by site-directed mutagenesis (Fig. 2A).
For expression in mammalian cells, the plasmid with the gene of interest,
Crosslinking glutamate receptor ion channels 167

now with an introduced Amber codon, is co-transfected with vectors


carrying genes for an orthogonal tRNA/synthetase pair, for example from
bacteria, because crossaminoacylation between translation systems from
other trees of life is low. Next, the transfected cells are incubated with
the UAA in the growth medium, which will be taken up by the endogenous
amino acid transport system. The tRNA-synthetase recognizes and attaches
the UAA to the orthogonal tRNA, and the UAA is then incorporated to
repress the Amber codon on the mRNA (Chin & Schultz, 2002; Huber &
Sakmar, 2014). This process is similar to the endogenous Amber suppression
that is reported for some animals and bacterial strains (Brown, Liu, & Deiters,
2018). If the gene of interest has a native Amber stop codon (like Rat GluA2),
this should be replaced, for example with the Opal stop codon.
To date, only a few UAA photocrosslinkers that are amenable to genetic
encoding are commercially available. Among the most widely applied are
p-azido-L-phenylalanine (AzF (Schwyzer & Caviezel, 1971) and p-benzoyl-
L-phenylalanine (BzF, also often abbreviated Bpa) (Kauer, Erickson-
Viitanen, Wolfe, & DeGrado, 1986). These probes were used in several
studies of glutamate receptors (Durham et al., 2020; Klippenstein, Ghisi,
Wietstruk, & Plested, 2014; Poulsen, Poshtiban, Klippenstein, Ghisi, &
Plested, 2019; Zhu et al., 2014). The photochemistry of AzF and BzF is well
studied (see Preston & Wilson, 2013 for comparison to photochemistry of
other crosslinkers). Following absorption of an UV photon (350–365 nm),
the keto group of BzF is excited to a triplet biradical that can crosslink to
CdH bonds within 3.1 Å. Critically, if recombination fails, the BzF relaxes
thermally to its original form, allowing crosslinking to accumulate over illu-
mination cycles (Fig. 1E). Photoactivation of AzF occurs at 254–400 nm
and generates an unstable nitrene that can crosslink to various chemical
groups, including primary amines (Fig. 1E). The activation of AzF is
irreversible, and in the absence of a crosslinking partner, the nitrene is
rearranged into a ketenimine, which reacts to generate unspecific crosslinks.
What factors should be considered when choosing to deploy AzF or
BzF? In spite of its more volatile behavior relative to BzF, AzF has been
extensively used as a photo-active UAA and is more readily incorporated
in buried sites, for example within the plasma membrane. Our biochemical
analysis showed robust incorporation of AzF and BzF at sites within the
AMPA receptor membrane domain (Poulsen et al., 2019). However, most
of the tested membrane-embedded sites harboring BzF gave non-functional
channels (Poulsen et al., 2019), whereas most AzF-labeled receptors had nor-
mal function. This observation contrasts to results in potassium and sodium
168 Andrew J. R. Plested and Mette H. Poulsen

channels, which were readily labeled with benzophenone or BzF, but may
have fortuitously exploited membrane-embedded regions with better solvent
access (Horn, Ding, & Gruber, 2000; Murray et al., 2016). In general, incor-
poration of AzF and BzF is better tolerated at sites of aromatic residues, which
to some extent biases their utility.
UAA photocrosslinkers of non-aromatic character have recently been
developed, such as 30 -azibutyl-N-carbamoyl-lysine (AbK) (Ai, Shen, Sagi,
Chen, & Schultz, 2011), (3-(3-methyl-3H-diazirine-3-yl)-propaminocar-
bonyl-Nε-L-lysine (DiZPK) (Zhang et al., 2011) and the optimized version
of DiZPK, Se-(N-(3-(3-methyl-3H-diazirin-3-yl)propyl)propanamide)-
3-yl-homoselenocysteine (DiZHSeC). The latter can undergo oxidation-
mediated cleavage to produce a mass spec identifiable label (Yang et al.,
2016). Thus, this growing palette of aromatic and non-aromatic UAA
photocrosslinkers enables introduction of covalent linkages to capture tran-
sient movements, protein-protein interactions and mapping of unknown
protein binding sites.

3.1 Control experiments


In contrast to the introduction of cysteines, to exclude some false negative
and false positive outcomes, a somewhat laborious panel of control exper-
iments are unavoidable when using UAA mutagenesis. These controls have
the aim of establishing a good orthogonality of the UAA incorporation,
parallel to and separate from the endogenous translation machinery.
These checks should be performed for each system, and in some cases, each
mutant. Ideally, controls should be performed for each type of experiment
(electrophysiology, biochemistry). Five outcomes that reduce the efficiency
of in vivo (as opposed to isolated in vitro reactions) incorporation must be
considered (Fig. 2).
1. Nonspecific UAA incorporation occurs if endogenous synthetases accept
exogenous UAAs and charge native tRNAs with them (Fig. 2C). This
possibility is rare because of the lack of compatibility and general increased
bulk of UAAs.
Control: Incubating cells overexpressing the wild-type form of
the protein of interest with UAA alone (without adding the tRNA/
Synthetase plasmids) should not lead to a photoactivated effect. UV illu-
mination should have no effect on cells or the protein of interest. This
experiment can also serve to confirm the expected overall stability of the
responses.
Crosslinking glutamate receptor ion channels 169

2. Nonspecific endogenous AA incorporation (Fig. 2C) occurs if the exogenous


synthetase charges exogenous tRNAs with endogenous AAs such as
Tyrosine. Amber codons are rescued but the protein is not labeled
with UAA.
Control: Activity of the exogenous synthetase to incorporate native
AAs at the TAG site is visible from current rescue in absence of added
UAA. The existence of this problem is clear from differences between
exogenous synthetases. For example, the BzF synthetase from E. coli is
more specific for BzF (less “leaky”) over endogenous AAs than the
AzF synthetase (Ye et al., 2008, 2009) is for AzF. Rescue in the absence
of UAA is less common with the BzF synthetase.
3. Nonspecific charging of exogenous tRNA with endogenous AA by
endogenous synthetases (Fig. 2C).
Control: Crossinteraction between tRNAs and synthetases from dif-
ferent trees of life is weak but can be identified if the exogenous tRNA
alone (without adding the synthetase plasmid or UAA) rescues expres-
sion of a gene of interest with an introduced early TAG stop codon.
4. Nonspecific charging of endogenous-tRNAs with UAA by the exoge-
nous synthetase (Fig. 2C).
Control: The same no-UAA control as for case #2 covers this possi-
bility, which is again rare because the crossreactivity between mamma-
lian tRNAs and bacterial synthetases is weak.
In the ideal case of perfect orthogonality, all these four confounding
interactions would be absent, but a fifth possibility exists:
5. Non-specific readthrough based on skipping the TAG codon (Fig. 2D).
This process exists in nature, with the translation of key viral proteins
depending on the readthrough of UAG stop codons. This readthrough
is driven by their neighboring nucleotide sequences (Skuzeski, Nichols,
Gesteland, & Atkins, 1991). In mammalian cells, the presence of an
adenine at the “–1” position is critical (Cassan & Rousset, 2001) for
readthrough. Recent work fully characterised Amber suppression as a
function of local sequence, and could predict about half of the effect
(Bartoschek et al., 2021). Although in principle, this knowledge could
be exploited to reduce readthrough, it may not be possible to change
the flanking coding sequence with synonymous codons.
Control: The same no-UAA control as for case #2 (and #4) covers
this possibility. It is essential to measure the enrichment of abundance
from including UAA (over control conditions) using Western blot anal-
ysis. Severe cases of readthrough can be assessed from solo expression of
TAG mutants of genes of interest.
170 Andrew J. R. Plested and Mette H. Poulsen

The above controls usually indicate the worst-case scenario because even
in case #2, endogenous AA incorporation is probably much slower than that
of exogenous UAAs for the pre-optimized synthetases. This means that
labeling with the exogenous AA can still in principle dominate even if there
is a competition between endogenous and exogenous AAs. Therefore, most
readthrough seems to come from #5. With premature stop codons, a methi-
onine later in the sequence can result in ectopic late start codon (Kalstrup &
Blunck, 2015). Thus, removal of the responsible Methionine residues would
avoid the expression of a truncated, but functional, protein. However, this
strategy is not always possible. Removing adventitious Kozak sequences
(Kozak, 1986) from around them is not reliable. These effects place limits
on fluorescence microscopy of GFP-fusion proteins as a readout of successful
rescue. Certain truncated constructs will give a nice cellular GFP signal
(if not at the membrane) even if no functional receptor rescue has been
achieved.
If rescue fails, a useful way to deduce if the problem is genetic (that is,
bad constructs) or chemical (poor UAA uptake) is to try to rescue Amber stop
codons with the exogenous wild-type E. coli tyrosyl synthetase and engineered
bacterial tRNA, because these exploit endogenous tyrosine and will rescue
expression selectively. If the protein expression is rescued using the exogenous
tyrosyl synthestase and tRNA, then the problem is related to poor take-up of
the UAA, and not failure to implement Stop codon rescue per se.
To enable electrophysiological recordings from cells which have effec-
tively rescued expression of AMPA receptors with early TAG stop codons
with the UAA of interest, we included TAG on the GFP as well (at residue
37). This dual reporter was essential for regularly finding cells with cur-
rents. When using separate plasmids, we found many green (transfected)
cells, but we only rarely found current responses to glutamate in
outside-out patches (Klippenstein, 2015). While these observations are
qualitative, a more sensitive test is to compare the biophysical properties
of currents in readthrough conditions at a previously characterized site.
Probing the desensitization behavior of L483TAG mutants expressed
without UAAs (Tyr at this site blocks desensitization, other aromatics
do not; Stern-Bach, Russo, Neuman, & Rosenmund, 1998) produced
an intermediate phenotype, consistent with no clear bias for native AA
incorporation (Klippenstein, 2015). Distinct desensitization from recep-
tors produced by receptors under conditions of AzF or BzF incorporation
further suggested that the rescue by UAAs is specific.
Crosslinking glutamate receptor ion channels 171

3.2 Using light-activated UAAs with electrophysiology


For photocrosslinking in outside-out patches from HEK cells expressing
UAA containing AMPA receptors, we adapted a standard patch clamp setup.
Our setup is based on an inverted research-grade microscope, and we direct
UV light to patches via the epifluorescence illumination path with a exci-
tation filter with 90% transmission (e.g., Semrock FF01–377/50–25)
and dichroic mirror with >95% reflectivity up to 400 nm (e.g., Semrock
FF409-Di03-25  36). For further details of the setup see (Klippenstein
et al., 2014).
Flash lamps (such as Rapp Optoelectronic JML-C2 or SP-20) can pro-
duce very brief (>1 ms) pulses of UV, and this approach was effective on
benzophenone conjugated to voltage-gated channels (Ding & Horn,
2001). But in the case of BzF, dependent on the local stereochemistry at
the site of incorporation, such illumination can fail to trigger recomb-
ination with measurable efficiency, and worked poorly in our hands.
The low repetition rate of such flash lamps limits their applicability.
The UVICO source from Rapp Optoelectronic has an inbuilt shutter
(allowing 50 ms exposures) and gives a reliable, intense UV light through
a liquid light guide: around 30 mW through the objective. Over the past
decade, brighter UV LEDs have become available and are also appropriate
for UAA photocrosslinking (Rook, Williamson, Lueck, Musgaard, &
Maclean, 2020). Such high power LEDs (like ThorLabs Solis type or
the Rapp Optoelectronic KSL2–365) represent a cost saving over metal
halide systems, and approach their brightness, with the advantage of a bet-
ter temporal control of illumination (better than 1 ms). Using shutters or
light-sources that are under computer control, the illumination can be
time locked to occur at a specific stage of ligand application, either before,
during or after a pulse of glutamate.
With regard to the choice of objective, we find that a quartz objective
was not necessary. The poor optical quality outweighs the usefulness of
better transmission. We use an Olympus LUCPLFLN 20 objective which
has 80% transmission at 365 nm and a steep cut-off below this point to <10%
at 320 nm. These optical properties are ideal for filtering out short wave-
length UV, and pre-filtering the illumination light with, for example,
Schott glass (Gurney, 1994) was not necessary. UV transmission through
the objective is easily detected with fluorescent Euro banknotes, available
everywhere in Europe or in any Bureau de Change.
172 Andrew J. R. Plested and Mette H. Poulsen

4. Fast perfusion coupled to chemical modification


and state-dependent crosslinking
Most early work with Cysteine chemistry targeted voltage-gated
channels, or slow ligand gated channels with substantial steady-state activity
(Boileau, Evers, Davis, & Czajkowski, 1999). For fast activating receptors,
fast perfusion is necessary to capture the true activation profile (Colquhoun,
Jonas, & Sakmann, 1992; Lester, Clements, Westbrook, & Jahr, 1990). To
catch motions on a 10 ms timescale, the chemical modification (DTT/
CuPhen) can also be delivered through a rapid perfusion system. The same
system can be used to capture the fast gating of AMPA receptors in a
state-dependent manner using photoactive UAA crosslinkers. In principle,
both approaches could be combined.

4.1 A four-parallel-bore perfusion tool for fast solution


application and chemical modification
This design for perfusion tools was originally developed by Dr. Mark Mayer
at NIH. Custom-made square or rectangular profile glass tubing (see Fig. 3
for profiles) is preferred over round profile theta-tubing because the barrels
are easily lined up and more barrels are available for doing control
and test jumps. A rectangular section tool allows the flowing interface
between solutions to be reliably vertical (perpendicular to the microscope
focal plane), which simplifies setup greatly. The same techniques work well
for four-square barrel tools to make fast binary exchanges with a piezo stack
(e.g., Phyisk Instrument P-840.40: 60 μm travel) and coarse manipulator to
switch between the “upper” and “lower” pair of barrels. In the following,
we outline the protocol for making 4 parallel barrel perfusion tools.

4.2 Materials
- Deactivated fused silica capillaries (for example, Agilent 160-2455-10).
- Borosilicate Glass multi-bore tubing (see Fig. 3 for designs, Vitrocom;
supplied in Europe by www.cmscientific.com).
- Fine silicon rubber tubing.
- Araldite Rapid or similar fast-setting epoxy adhesive.
- Hydrofluoric Acid (695,068 FLUKA).
- Small Bunsen burner.
- 1 mm 80/20 Nichrome wire coil (1.385 Ohm/m) mounted in heatproof
holder (e.g., from an old electric cooker).
Crosslinking glutamate receptor ion channels 173

Fig. 3 Multibarrel glass profiles. Original drawings for glass supplied by Vitrocom
(distributed by CM Scientific) in 30 cm lengths. The four barrel profile (C) tends to
bow out slightly.

- Dissecting microscope (e.g., Olympus SZ61) or digital microscope.


- Coarse Manipulators (e.g., Narishige MMN-1, Narishige oil-hydraulic
like MMO-203).
- 20A laboratory power supply (e.g., https://www.heidenpower.com/
produkt/zplus/).
174 Andrew J. R. Plested and Mette H. Poulsen

- Pristine diamond knife (Ted Pella).


- Forceps (older, coarser ones are fine).
- Drying oven to cure glue and store tools between experiments.
To perform experiments, you will also need a piezo lever, mounting hard-
ware, a coarse manipulator for positioning on the microscope, low-voltage
piezo amplifier (e.g., E610 Phyisk Instrument) and filter to smooth the
amplifier input signal.

4.3 Preparing the fused silica capillaries


Use a razor blade to cut the silica capillaries to the appropriate length
(about 12 cm). If you make a half-cut in the polyimide coating, it should snap
clean. The ends can be smoothed off the with superfine (>1000 grit)
abrasive paper.
Bath sonicate the capillaries in a 50 mL Falcon tube using 70%
non-denatured EtOH for 15 min (to clean for dust and glass splinters).
Rinse capillaries 2–3 times in MilliQ water and air dry (or dry in the oven
at 40–50 °C)).
Mix epoxy glue 1:1 and add a small drop, 0.5 cm from end of each cap-
illary. Creating a cardboard jig with holes appropriate in size to hold the
capillary can be helpful. Rotate the capillary positioned in the cardboard
while adding the glue so the capillary is in the middle of the glue ball.
This glue ball will act as a bush for holding the fine silicon tubing as a
force-fit later. If the glue ball is too big, there is risk to break the fused silica
capillary when attaching the rubber tubing.
Dry the silica capillaries, still positioned in the cardboard holder, in the
oven (40–50 °C) for minimum 2 h.

4.4 Pulling the 4-bore glass


First, cut the raw glass tubing (Vitrocom—see Fig. 3 for designs) into 10 cm
lengths by scoring with a single stroke of the diamond knife on 1 face
(Fig. 4A and B). Break glass by pulling apart with your fingers (latex gloves
can help grip), not by bending. These 10 cm long glass capillaries will each
result in two perfusion tools.
The four parallel bore tubing is 2.6 mm wide and does not easily fit in a
pipette puller. To make the narrow tip of the perfusion tool, we draw the
glass out over a small Bunsen flame (Fig. 4D). For smaller profile glass (e.g.,
4-square, Fig. 3), we would use a Sutter P1000 with 3 mm box filament.
Crosslinking glutamate receptor ion channels 175

Fig. 4 Perfusion tool preparation. (A) Cut glass to 10 cm lengths by scoring with
a diamond knife. (B) Pull glass apart to achieve a clean break. (C) Clean-broken tip.
(Continued)
176 Andrew J. R. Plested and Mette H. Poulsen

The pulled glass might have an hourglass shape—it must be long enough
so that after cutting the glass into two pieces, the tip can be bent (see below).
It is very hard to pull the 4-parallel bore glass straight over the flame, and you
will usually pull the glass into two pieces (Fig. 4E) that you cut back sepa-
rately. Fortunately, extra bends are largely without consequence. If not
already in two parts, divide the capillary into two pieces by cutting in the
middle of the taper.
Cutting the tip is best done by mounting the 4-bore tubing in manipu-
lator under dissecting microscope (Fig. 4F–H) and mounting a diamond
knife in a second manipulator (a Narishige hydraulic or manual types work
nicely). In the middle of the hourglass-shaped taper, or near the tip, repeat-
edly score 1 surface using up/down movements of diamond knife. Use
enough pressure to cause the glass to deflect slightly. Keep scoring, and per-
haps push a little, until glass breaks spontaneously. You can also use your
finger to break the glass after scoring. If two manipulators are not available,
the glass tip can be nicked with a steady hand. The break should be as clean
and square as possible. The resulting tip has to be thinner/smaller than
500 μm so that activating the piezo lever can move the tool across all the
barrels. You can store the cut perfusion tubes indefinitely on wax or
Blu-tac strips in a petri dish.

4.5 Bending the tip


If necessary, use a nichrome wire heater to bend the tip of the tool. Thus,
the tool can be mounted later at an angle that enables the outflow to be par-
allel to the focal plane (see Fig. 5D). Mount the cut glass capillary on a
heat-resistant holder (Perspex can work) and place it under microscope.
Slide in between nichrome wire bent double with little clearance, heat
the wire (10 A) and bend it the glass to about a 45° angle while heating
by pressing with a pair of forceps (see Fig. 4I and J). You need a surprising
amount of heat to melt the glass. We use a 20 A lab power supply to the
filament.
Fig. 4—Cont’d (D) A 10 cm length is pulled by hand over a Bunsen flame. (E) Two raw
pulled tubes pulled over the Bunsen flame. (F) Under the dissecting microscope, mount
the tool end on to cut across the barrels and obtain the tip (photos G and H). (I) Closeup
of glass mounted in Perspex holder for bending in the nichrome heater. (J) The holder
slides in a slot in a metal jig to allow the glass to be bent in a nichrome filament heater.
(K) With the tip cut back, bent and the nose piece pulled off (in the same way as in
panels A–C), the polyimide tubing can be introduced into the shaft. (L) Positioning is
checked by sliding the tip back on and reforming the join. (M) Raw tip before etching
in HF, and after two rounds of etching. This tool, albeit having a rather wide tip, can now
be glued together.
Crosslinking glutamate receptor ion channels 177

Fig. 5 Final perfusion tool. (A) Tool made from three parallel barrel glass. (B) Small balls
of dry epoxy glue to hold silicon tubing in place by interference fit. (C) Sucking back a
few milliliters of MilliQ water through the tip of a finished tool to clear out any debris.
(D) Tool in use mounted to piezo lever in Siskiyou mount on an inverted microscope.
Lever is in turn mounted on a manual manipulator for positioning over the microscope
objective.

4.6 Cutting the nose piece off


Carefully score across the face of the perfusion tool as close to the bended tip
as possible (approx. 10 mm from tip). This reduces the dead volume during
the recordings. Carefully snap off and keep the nose cone. This can be done
using forceps or your hands. If using forceps, put rubber around the metal so
that the glass does not break. Silicone tubing is ideal for this. If using your
hands, you will find that gloves help to provide grip on the glass. Label the
two pieces with ink dots to enable correct orientation when you reassemble
the tool.

4.7 Etching the tip


Etching the septa with hydrofluoric acid (HF) is a straightforward way to
narrow the interface between solutions, giving a faster rise time following
a solution jump (Robert & Howe, 2003). There is a danger of etching
too much and simply destroying the tip, however. Fig. 4M shows the thin-
ning of a three-barrel tip before and after etching.
All procedures must be done with care in a fume hood with double pairs of gloves
and eye-protection goggles –HF is very corrosive! We advise to wear two lab coats as
well.
Pipette a 100 μL droplet of HF (48%) into a 35 mm plastic petri dish.
Hold the nose cone using a pair of forceps. Dip the bended tip of the nose
piece into the HF and keep it for a couple of seconds until HF has been
178 Andrew J. R. Plested and Mette H. Poulsen

sucked into the barrels by capillary action. Put the tip aside in a dish with HF
still in the barrels for 15–20 min (never exceed 25 min!) inside the fume
hood. The incubation time depends on the size of the tip, larger tips have
thicker glass and can be etched for longer. Add some drops of dilute NaOH
in MilliQ water onto the tips to stop the etching.
Carefully wash the tips with MilliQ water. Add 500 mL MilliQ water in a
beaker, draw water into the 50 mL syringe and now assemble with 0.22 μm
filter (Millipore) and silicon rubber tubing. Insert the nose piece in the tub-
ing with the tip pointing away from the 50 mL syringe (still wearing double
gloves and glasses!) and push the water through the tip. This step is critical,
and it is easy to destroy the tip in the process. Ensure a good grip on the nose
piece before pushing MilliQ water through. Work over cling film draped
across the sink in case you end up squirting off the precious tip. Repeat this
washing step 3 times. Dispose the HF in appropriate container in the fume
hood. Check the tip under the dissecting microscope—you might need to
etch twice if the appearance of the tip walls is unchanged. For the second
etching, time is more critical (to avoid breaking the septum) and varies
for different tubes, from 2 to 10 min. After a second etching, repeat the
washing protocol.

4.8 Assembling the tool


To assemble the perfusion tool, collect all the parts; the prepared silica tubes,
the nose piece that was cut off and etched (see above), the remaining part
of the quartz glass capillary and silicon tubing. Push the silica tubes all the
way through the barrels in the quartz glass capillary, with the glue ball
now pointing away from where the nose piece was cut off. Now smear
epoxy glue over the silica tubes where they protrude from the shaft and push
back and forth to work glue into the barrels of the glass capillary. Most impor-
tant is to seal the shaft. Assemble the perfusion tube by connecting the nose
piece, using the ink marks for correct orientation. Slide each silica capillary
individually into each barrel in the nose piece and as far as possible into the
tip. Be careful not to block the silica capillary with glue. Slide the nose piece
until it butts against the cut surface of the glass capillary. Remove excess glue
and wiggle to persuade epoxy glue into the gaps. Splay out the protruding
silica capillaries at the back end of the tool using a cut-off pipette tip with glue.
Cure the glue in the oven (>40 °C) for minimum 2 h. To enable connection
Crosslinking glutamate receptor ion channels 179

of the perfusion tool to the perfusion system fit 1 cm segments of silicone


rubber tubing over the glue bushes on the capillaries.

4.9 General advice


Important! Before first use, wash the perfusion tool. Connect a syringe to the
perfusion tool via the silicon tubing and dip the tip into MilliQ water. Draw
MilliQ water back through the tip to clear any debris (Fig. 5C). If you try to
eject solution through the tip, you can easily block the tip of the tool. Wash
with 20% EtOH and water extensively after use, air dry with filtered com-
pressed air as well as possible and store the perfusion tool in a drying oven
>40 °C. With luck, a good perfusion tool can last for the experiments for an
entire paper.
For work on AMPA receptors, we exchanged the usual fast perfusion
system (using a piezo stack) for a piezo lever with a much longer travel
(400 μm, P601.4OL Physik Instrument), on which we mounted custom
glass with four parallel barrels. The mount for the glass tool can be made from
Perspex or aluminum (commercially available as Siskiyou MXPZT-300 system,
see Fig. 5D). 3D printed parts should also work well. The priority is to minimize
weight. Although the low natural frequency of these devices restricts their
motion to the 1–10 ms range, the solution exchange happens in a narrow mixed
region between two, or more, laminar flowing solutions (see Fig. 6) and is
complete on a faster timescale ( Jonas, 1995). For a theoretical and empirical
treatment (see Sachs, 1999). The rise time for the open tip response (junction
potential between two solutions with different ionic strength) should be
100–200 μs if mounted on a fast piezo. For a piezo lever the exchange is not
so fast. If uncontrollable oscillation of the tip occurs, reduce the filter frequency
on the piezo amplifier input (100–300 Hz) and/or adjust the mounting
position of the tool to change the natural frequency of the vibrating parts.
To record the fast gating of the AMPA receptor, we excised outside-out
patches from HEK cells expressing the receptor. The perfusion tool is then
moved so that the pipette with the outside-out patch sits equidistant from
the flanking interfaces. This reduces the tendency of mechanical oscillations
of the solution flow that follow switches to reach the patch (see Fig. 6). In
order to rapidly apply oxidizing conditions in the millisecond range, we
swept the tool rapidly from one barrel across a second to a third (Salazar
et al., 2020), varying the exposure time by changing the slew rate of the
perfusion tool.
180 Andrew J. R. Plested and Mette H. Poulsen

Fig. 6 Perfusion protocol using a tube with four parallel barrels. Tool is mounted on a
piezo lever. During the experiment, the tool is moved between four positions (virtual
patch pipette positions marked in photo). GluA2 A665C mutant in a membrane patch
is exposed to glutamate in reducing conditions (DTT, green) to establish the baseline.
Cyclothiazide was present throughout to block desensitization. For crosslinking, perfu-
sion tool is rapidly moved so that the red barrel is flowing over the patch. The tool com-
mand voltage is shown, with the colors indicating the solution that is flowing over the
patch. Interfaces are visible due to sucrose in alternating solutions (e.g., red and gray).

5. Biochemistry: Photoactivated Crosslinkers


and purification of GluA2
We chose to do crosslinking in live cells, to maintain the proper
oligomeric state of the channel. Most ion channels contain cysteines. The
AMPA receptor subunit GluA2 has 11, but 4 are in conserved disul-
fide bonds. Therefore, there are up to seven free cysteines. Although in
native folding, these are not available to make adventitious crosslinks, during
the extraction and purification of receptors for biochemical analysis, the
receptor unfolds and these cysteines readily form non-physiological covalent
crosslinks which confound the appearance of any crosslinks from UV acti-
vation. The GluA2-S729TAG described below contains 3xCys deletions to
Crosslinking glutamate receptor ion channels 181

reduce the tendency of the receptor to form non-physiological higher order


oligomers during UV crosslinking. However, removing Cysteines in GluA2
also reduces expression dramatically (Lau et al., 2013). Therefore, we also
include 20–40 mM N-ethyl-maleimide (NEM, see Fig. 7). The construct
is purified with a C-terminal FLAG-tag. Therefore, only receptors for
which the early Amber stop codon was replaced by UAA should be purified,
and not any truncated forms. In the following, we outline the protocol for
biochemical detection of UAA crosslinking between subunits. Within sub-
unit crosslinking can be detected by inserting a TEV protease site between
the UAA and its target, and assessing the ability of crosslinking to protect
monomeric forms (Poulsen et al., 2019; Xu, Ramu, Shin, Yamakaze, &
Lu, 2013).
We use HEK 293 T cells (DSMZ no.: ACC 635) obtained from the
German Collection of Microorganisms and Cell cultures (www.dsmz.de)
and culture them according to standard techniques using 90% Dulbecco’s
minimal essential media (DMEM) with 10% Fetal Bovine Serum, following
the supplier’s instructions.

5.1 Materials
– GE Healthcare Microspin Columns (Cat # 27-3565-01)
– Sigma-Aldrich FLAG peptide (Cat # F3290)
– Sigma-Aldrich anti-FLAG M2 affinity gel (Cat # A2220)
– Agilent Stratagene anti-flag M2 antibody (Cat # 200472-21)
– Invitrogen pre-cast 4–12% Tris gels (Cat# NP0336)

5.2 Buffers
Make up Lysis buffer (LB) in MilliQ water. This buffer is used to wash the
column too—NEM is always present:
– 300 mM NaCl.
– 50 mM Tris-HCl (pH 8.0).
– 1 protease inhibitor mixture (Roche).
– 1% Dodecylmaltoside (DDM).
– 40 mM N-ethyl-maleimide (NEM).
– Loading buffer SDS 4 .
– 250 mM Tris-HCl.
– 8% SDS.
– 40% Glycerin.
– 0.4% Bromophenol Blue.
Fig. 7 Biochemical preparation for UV crosslinking of AMPA receptors with
UAAs. (A) 10 cm dishes with Duroplan covers to filter out short wave UV, on ice.
(B) Crosslinking procedure shows dishes in the oven (door open) with UV meter visible
at the back. (C) Step by step protocol, showing Western blot against wild-type GluA2
and S729TAG constructs. Note lack of monomer band (no “rescue”) in the absence of
BzF, but wild-type shows very strong exposure-dependent crosslinking, through unwanted
disulfide formation during denaturation. (D) Blocking cysteines with NEM during experi-
mental steps and including BME in the gel loading buffer reduces wild-type crosslinking
and reveals clear BzF dependent crosslink in the S729BzF mutant. Note much less protein
loaded in WT lanes (to keep monomer amount roughly constant). The RNA-synthetase pro-
tein also has a FLAG-tag and this can usually be seen quite well. In the wild-type GluA2
lanes, the RNA-synthetase-Flag (RS-FLAG) band is present but very faint.
Crosslinking glutamate receptor ion channels 183

5.3 Day 1: Cell culture


Split HEK-293 T cells (with the SV40 large T-Antigen, for better transfec-
tion) and plate them in 10 cm dishes. You want cells to be 75% confluent
for transfection. Non-UV-treated HEK-293 T cells should be used as con-
trols so prepare extra dishes accordingly. One dish per lane of the Western
blot should be enough. Cells are cultured in DMEM with 5–10% serum
(DMEM complete).

5.4 Day 2: Transfection and UAA labeling


Transfect the HEK-293 T cells with desired plasmids using polyethyleneimine
(PEI, 1 mg/mL, linear 25kDa Polysciences, Inc., Eppelheim, Germany) in the
ratio 1:3 of DNA:PEI. We use PEI because it is inexpensive and has a good
transfection yield.
To an Eppendorf tube add 900 μL of OptiMEM and 63 μL PEI.
Then add a total amount of 21 μg DNA of GluA2-S729TAG-FLAG:
tRNA:RS in a mass ratio of 4:1:1. Let the mixture stand in room temp.
For 20–30 min before adding the mixture dropwise on the cells. The
amount of DNA needed will vary depending on the protein of interest.
After 6 h aspirate the cell medium and add 10 mL of freshly-prepared
DMEM complete containing the UAA 1 mM BzF (MW: 269.3; Bachem
Holding AG, Bubendorf, Switzerland) or 0.5 mM AzF (MW: 206.1;
Chem-Impex International, Inc., Wood Dale, IL, United States). Although
for electrophysiology, we often use methyl ester forms (available commer-
cially from AsisChem) for good uptake, these are too expensive for biochem-
istry scale preps.
BzF is dissolved in 1 M HCl and AzF in 1 M NaOH. The dissolved
UAAs should immediately be added to prewarmed (37 °C) DMEM com-
plete to a final concentration of 1 mM for BzF and 0.5 mM for AzF. The
DMEM-complete containing BzF or AzF should be adjusted to pH 7.3
and sterilized using a 0.22 μm PVDF filter (Millex®-GV; Millipore) (Hino,
Hayashi, Sakamoto, & Yokoyama, 2006; Ye et al., 2009). The medium
should be supplemented with 1 mM kynurenic acid, a low affinity, cheap
AMPA receptor blocker to prevent improve cell health for cells expressing
AMPA receptors.

5.5 Day 5: UV treatment, harvesting and lysis of cells


Old hybridization systems (e.g., the Stratalinker made by Stratagene) can be
used, but these are no longer commercially available and the UV bulbs
should be changed from the usual 254 nm for nucleic acids. We have also
184 Andrew J. R. Plested and Mette H. Poulsen

used UV curing LEDs but the best results with BzF were obtained with a
UV oven (Luzchem, LZC-1; Ottawa, Canada) which has a heat extraction
fan, with UV spectrum peaking in the 300–400 nm band. In general, the
intensity of UV illumination that can be achieved in these experiments is
at least an order of magnitude less than what is possible with the focusing
illumination with a microscope objective. Therefore, illumination intervals
must be extended.
Keep everything on ice (see Fig. 7). Wear UV-blocking goggles
and cover your arms—UV light is a mutagen. Turn on the UV incubator
on full intensity. The incubator has to warm up for 15–20 min prior incu-
bation of cells, and the UV intensity should be checked using an integrated
UV sensor. The intensity at the position where the dishes are (see Fig. 7)
should be around 330–440 mW/m2. Before UV treatment, check the cells
for green fluorescence at the microscope to evaluate efficiency of transfec-
tion. You want at least 10% efficiency, indicating 10% of cells took up all
three plasmids and incorporate UAA to the receptor.
Prepare a solution of 40 mM N-ethyl-maleimide (NEM, to quench
Cysteines, see Fig. 7) in PBS, stir the solution and pH adjust to between
7 and 7.2. Keep the solution on ice. Aspirate the cell medium and wash
the cells carefully with cold (4 °C) PBS, aspirate the PBS and add 10 ml
of the prepared cold 40 mM NEM-PBS (keep on ice). Now, put the cell
dish on ice in the UV oven(without any plastic lid, it blocks UV, but option-
ally with Duraplan (Pyrex) lids, see Fig. 7) for 30 min. Immediately after UV
treatment harvest cells in 1 mL cold PBS and centrifuge at 1000 g for 5 min at
4 °C.
Discard supernatant and re-suspend pellet in 500 μL of Lysis Buffer (LB).
Sonicate each sample 20 times in the cold room at 4 °C using a standard
immersion tip. Our sonicator was manufactured by Dr. Hielscher, model
UP 200S (0.5 cycles and 50% amplitude).
Rotate or nutate the samples at 10 rpm for 1 h at 4 °C. Centrifuge at
20,000 g for 20 min at 4 °C to separate pellet from Clear Lysate (CL).
Re-suspend the CL with 80 μL of beads anti-FLAG M2 affinity gel (previ-
ously washed 3  at 800 rpm for 3 min with LB). Rotate the beads with the
clear lysate at 10 rpm overnight at 4 °C.

5.6 Day 6: FLAG-TAG purification and Western blot


Collect the samples from the cold room and centrifuge them at
55  g (800 rpm in Eppendorf 5418 microcentrifuge) for 3 min. Prepare
microspin columns by removing their caps and placing them in 2 mL
Crosslinking glutamate receptor ion channels 185

receiving tubes. Re-suspend the beads in enough loading buffer to load the
entire contents of each tube in a microspin column (Your FLAG protein
should be bound to the beads). The liquid part will flow through the col-
umn, while the beads will be retained. Following the supplier’s protocol,
wash each column 3 times with 800 μL LB. Discard the flow through each
time. After the last wash spin the columns for 30 s at 800 rpm. Prepare the
Elution Buffer (EB), calculating for each sample:
7 μL Flag competitor Protein +13 μL Lysis Buffer ¼ 20 μL (Elution 1—E1)
7 μL Flag competitor Protein +13 μL Lysis Buffer ¼ 20 μL (Elution 2—E2)
Add 20 μL of EB to each column. Normally there is very little
FLAG-tagged protein eluting with E1, but collect the sample in a new
1.5 mL tube anyway (just in case the E2 does not show any FLAG-tagged
protein) by centrifuging for 3 min at 55  g. Store the E1 samples at
20 °C. Now collect the Elution 2 (E2). This sample should contain the
FLAG-tagged protein of interest.
Place the columns (with the cap on) in new 1.5 mL tubes. Add 20 μL of
EB and put them to shake at 300 rpm for 15 min. This step improves the
yield. Remove the caps from the columns and centrifuge at 55  g for
3 min. Store the E2 samples at 20 °C. Prior to running a gel with the
E2 samples check that the volume of each sample is 20 μL (if not adjust with
H2O) and add 5 μL of Loading Buffer (SDS 4) with 500 μM
β-mercaptoethanol added. Heat the samples for 5 min at 80 °C before load-
ing the precast 4–12% Bis Tris gel.
The Western blot is run according to the machine you are using. We
use the Invitrogen system and run the gel for about 1 h at 200 V at 4 °C.
We transfer at 4 °C either for 2 h at 25 V or overnight at 15 V. We otherwise
do everything as written in their manual. Primary antibody mouse anti-
FLAG (1:1000) from Agilent (cat# 200472-21) in 5% milk O.N. at 4 °C.
If doing chemiluminescence detection, incubate with a secondary anti
mouse-HRP (1:10000) 1 h RT. For superior linear detection, we use IR
secondary antibodies and imager from LiCor.

6. Modeling of the effects of crosslinking


6.1 Disulfide bonds and bifunctional crosslinkers
Structures of crosslinked receptors from crystallography (Lau et al., 2013)
and CryoEM (Herguedas et al., 2016) can be useful for relating the geometry
of the crosslinked form to a functional state. However, these are not always
available. Disulfide crosslinks can be used to detect close contacts, but
bifunctional crosslinkers tend to measure ideal distances, and between
186 Andrew J. R. Plested and Mette H. Poulsen

two sites a favored length of crosslinker may be observed (Baranovic &


Plested, 2018). To address the new geometries, we used rigid body docking
of binding domains to define the conformational space of possible crosslinks.
This approach enabled us to define forbidden and allowed regions of
conformational space and relate them to known structures. This docking
defined a compact conformation of the LBD layer of the AMPA receptor
subunit GluA2 that in this specific case could be attained in complex with
Stargazin, an auxiliary subunit of AMPA receptors (Chen et al., 2000). The
procedure (using the burymyCys.py script from https://github.com/aplested/
cystance) is outlined in Fig. 8. As described in Langtangen’s magisterial book
(Langtangen, 2013), PYTHON is used to “glue” together command line
and scriptable tools such as molecular graphics software (PyMol; Schr€odinger,
2015) and utilities from CCP4 (AREAIMOL and NCONT) (Winn et al.,
2011). These compiled utilities do the heavy lifting and PYTHON script can
semi-interactively manipulate domains to find plausible geometries where

Fig. 8 Calculating a conformation where cysteine crosslinking cannot occur. An exam-


ple of structural modeling to interpret crosslinking data. What is the geometry of a new
state where crosslinking is absent? The procedure is repeated to generate 50–100
models and thereby map conformational space. See Baranovic and Plested (2018) for
usage of this script to determine a plausible compact conformation.
Crosslinking glutamate receptor ion channels 187

cysteines are inaccessible due to steric constraints. This method uses close
atom contacts (<2 Å) as a proxy for energetically-unfavorable conformations.

6.2 Photocrosslinking
At this stage, we must acknowledge that sampling of sufficient crosslinking
sites to integrate or interpret structural dynamics is difficult. Certain sites do
not work for incorporation and certain sites have no effect. After these are
excluded only 50% of sides tried will give an effect. To achieve a meaningful
discrimination (clustering) of effects, groups need to have multiple members.
Therefore, at a minimum, a back of the envelope calculation suggests
screening 20 + sites. We screened about 40 for our recent study, with about
20 giving reproducible results (Poulsen et al., 2019). We discarded some sites
because truncation did not produce a non-functional AMPA receptor (from
the middle of the M4 helix onward). Finally, the lack of clear intersubunit
crosslinking results from biochemistry limited our interpretation to struc-
tural perturbations, rather than distance constraints.
In our hands, mapping crosslinking effects during electrophysio-
logy experiments to known structure did not give meaningful patterns.
Nonetheless, we noticed that certain crosslinking sites provoked different
changes in the ion channel activity. To understand the consequences of
crosslinking activation of UAA we employed a kinetic modeling technique
(using the Aligator PYTHON scripts, https://github.com/aplested/aligator)
whereby we progressively change and transition energies between states
(resting, open, desensitized) in a toy model of AMPA receptor activation
(Poulsen et al., 2019). The rationale for this approach is that crosslinking
or changing structure might have an impact on some transitions more than
others. The results suggested that some sites affected desensitization more
than activation, whereas other sites were much stronger on activation gating.
Critically, one site in the selectivity filter was an outlier, having a complex
mixture of effects, which we could test with further experiments (Poulsen
et al., 2019).

7. General limitations
Why can gating changes due to UAA activation by light be detected
with electrophysiology, but differences between open and closed states can-
not be rationalized from available structures, and intersubunit crosslinking
in the membrane domain is minimal? These dichotomies points toward
a limited power of the method to resolve conformational changes. Two
188 Andrew J. R. Plested and Mette H. Poulsen

possible explanations are that (a) the structures determined so far do not cap-
ture all details of the dynamic receptor and, more likely, (b) the energetic
changes due to isomerization are somewhat opaque. The energies related
to gating motions vary strongly over small distances, and cannot be read
off directly from structures. Hydrogen bonds and van der Waals interactions
are important for determining ion conduction and the energies involved
change twofold with distances in the Angstrom range, similar to the size
of the atoms themselves. Therefore, structural noise, and the difficulty to
summate all energetic contributions might wash out small effects. On the
other hand, the high signal to noise of the patch clamp can detect 10%
changes in conductance trivially. In future, it may be possible to combine
molecular dynamics simulations of AMPA receptors (Biedermann,
Braunbeck, Plested, & Sun, 2021) with crosslinking data to understand per-
meation and gating of ion channels. However, more work is required to
understand the disruption to structure, and thus to gating, that is made by
activation of the UAA crosslinkers by light. In this respect, the tendency
of AzF to follow multiple reaction pathways following photon absorption
probably contributes to its greater effectiveness in perturbing gating, even
if the structural nature of the perturbations is unclear.

8. Summary and perspectives


In this chapter, we have described two complementary strategies for
studying transient movements occurring during AMPA receptor gating.
Studying these movements is crucial for understanding not only receptor
activation and functional states, but also the potential druggable landscape
of individual receptor subtypes. Both cysteine and UAA mediated cross-
linking discussed have their pros and cons. Cysteine crosslinking is straight-
forward to initiate but solution exchange is more challenging to implement
and more sites must be screened. Using light to activate UAA crosslinkers
simplifies experiments and provides the opportunity to crosslink over many
sites independent of environment. However, the size of the UAA might
interrupt gating, as seen for BzF in the membrane domain of the AMPA
receptor. Further, several additional controls are required to ensure the effi-
cacy and specificity of the UAA mutagenesis. The toolbox of UAA photo-
crosslinkers continues to expand, allowing insertion of UAAs of different
size and photochemistry. In principle, this powerful tool to study receptor
mechanisms could be used in parallel with cysteine bridges to provide a full
depiction of receptor movements. The protocols outlined in this chapter
Crosslinking glutamate receptor ion channels 189

have been optimized to study crosslinking during electrophysiological


recordings and biochemical analysis of fast gating receptors with low expres-
sion and should be applicable to other ion channels and membrane proteins
in general. In the case of AMPA receptors, a panel of structures is available
that can be used to model structural transitions between states captured
by crosslinking, and more abstract methods can be used to model receptor
gating energetically, with chemical kinetic schemes. The principal advan-
tage of such modeling is to identify new aspects of receptor function, exem-
plified in our approach as the novel observation, confirmed by experiments
that the AMPA receptor selectivity filter can have a minor conductance in
the desensitized state.

Acknowledgments
We thank Jelena Baranovic, Hector Salazar, Valentina Ghisi, Viktoria Klippenstein
and Anahita Poshtiban for their refinements to the protocols. We thank Thomas Sakmar
(Rockefeller) for the gift of optimised tRNAs and RNA-synthetase constructs. We thank
Hector Salazar and Eva Meyer-Keller for some of the photographs of the glass tools in
Figs. 4 and 6. This work was funded by DFG (PL 619/7-1, Heisenberg Professor and
Cluster of Excellence NeuroCure EXC2049) and ERC (CoG “GluActive” 647895). We
thank Sam Goodchild and Chris Ahern for their advice on UV crosslinking and Mark
Mayer for comments on the perfusion tool protocol.

References
Ai, H.-W., Shen, W., Sagi, A., Chen, P. R., & Schultz, P. G. (2011). Probing
protein-protein interactions with a genetically encoded photo-crosslinking amino acid.
ChemBioChem, 12, 1854–1857.
Armstrong, N., Jasti, J., Beich-Frandsen, M., & Gouaux, E. (2006). Measurement of confor-
mational changes accompanying desensitization in an ionotropic glutamate receptor.
Cell, 127, 85–97.
Baranovic, J., & Plested, A. J. (2018). Auxiliary subunits keep AMPA receptors compact dur-
ing activation and desensitization. eLife, 7.
Baranovic, J., Chebli, M., Salazar, H., Carbone, A. L., Faelber, K., Lau, A. Y., et al. (2016).
Dynamics of the ligand binding domain layer during AMPA receptor activation.
Biophysical Journal, 110, 896–911.
Bartoschek, MD, Ugur, Enes, Nguyen, T, Rodschinka, G, Wierer, M, Lang, K, et al. (2021).
Identification of permissive amber suppression sites for efficient non-canonical amino
acid incorporation in mammalian cells. Nucleic Acids Research. https://doi.org/10.
1093/nar/gkab132.
Biedermann, J., Braunbeck, S., Plested, A. J. R., & Sun, H. (2021). Nonselective cation
permeation in an AMPA-type glutamate receptor. Proceedings of the National Academy
of Sciences of the United States of America, 118.
Boileau, A. J., Evers, A. R., Davis, A. F., & Czajkowski, C. (1999). Mapping the agonist
binding site of the GABAA receptor: Evidence for a beta-strand. The Journal of
Neuroscience, 19, 4847–4854.
190 Andrew J. R. Plested and Mette H. Poulsen

Borg, C. B., Braun, N., Heusser, S. A., Bay, Y., Weis, D., Galleano, I., et al. (2020).
Mechanism and site of action of big dynorphin on ASIC1a. Proceedings of the National
Academy of Sciences of the United States of America, 117, 7447–7454.
Brown, W., Liu, J., & Deiters, A. (2018). Genetic code expansion in animals. ACS Chemical
Biology, 13, 2375–2386.
Cassan, M., & Rousset, J. P. (2001). UAG readthrough in mammalian cells: Effect of
upstream and downstream stop codon contexts reveal different signals. BMC Molecular
Biology, 2, 3.
Chen, L., Chetkovich, D. M., Petralia, R. S., Sweeney, N. T., Kawasaki, Y.,
Wenthold, R. J., et al. (2000). Stargazin regulates synaptic targeting of AMPA receptors
by two distinct mechanisms. Nature, 408, 936–943.
Chin, J. W., & Schultz, P. G. (2002). In vivo photocrosslinking with unnatural amino acid
mutagenesis. ChemBioChem, 3, 1135–1137.
Chin, J. W., Martin, A. B., King, D. S., Wang, L., & Schultz, P. G. (2002). Addition of a
photocrosslinking amino acid to the genetic code of Escherichia coli. Proceedings of the
National Academy of Sciences of the United States of America, 99, 11020–11024.
Colquhoun, D., Jonas, P., & Sakmann, B. (1992). Action of brief pulses of glutamate on
AMPA/kainate receptors in patches from different neurones of rat hippocampal slices.
The Journal of Physiology, 458, 261–287.
Cordero-Morales, J. F., Cuello, L. G., Zhao, Y., Jogini, V., Cortes, D. M., Roux, B., et al.
(2006). Molecular determinants of gating at the potassium-channel selectivity filter.
Nature Structural & Molecular Biology, 13, 311–318.
Das, U., Kumar, J., Mayer, M. L., & Plested, A. J. R. (2010). Domain organization
and function in GluK2 subtype kainate receptors. Proceedings of the National Academy of
Sciences of the United States of America, 107, 8463–8468.
Ding, S., & Horn, R. (2001). Slow photo-cross-linking kinetics of benzophenone-labeled
voltage sensors of ion channels. Biochemistry, 40, 10707–10716.
Durham, R. J., Paudyal, N., Carrillo, E., Bhatia, N. K., Maclean, D. M., Berka, V., et al.
(2020). Conformational spread and dynamics in allostery of NMDA receptors.
Proceedings of the National Academy of Sciences of the United States of America, 117,
3839–3847.
Gurney, A. M. (1994). Flash photolysis of caged compounds. Microelectrode Techniques
(pp. 389–406).
Harding, M. M., Nowicki, M. W., & Walkinshaw, M. D. (2010). Metals in protein
structures: A review of their principal features. Crystallography Reviews, 16, 247–302.
Heras, B., Edeling, M. A., Schirra, H. J., Raina, S., & Martin, J. L. (2004). Crystal structures
of the DsbG disulfide isomerase reveal an unstable disulfide. Proceedings of the National
Academy of Sciences of the United States of America, 101, 8876–8881.
Herguedas, B., Garcı́a-Nafrı́a, J., Cais, O., Fernández-Leiro, R., Krieger, J., Ho, H., et al.
(2016). Structure and organization of heteromeric AMPA-type glutamate receptors.
Science, 352, aad3873.
Hino, N., Okazaki, Y., Kobayashi, T., Hayashi, A., Sakamoto, K., & Yokoyama, S. (2005).
Protein photo-cross-linking in mammalian cells by site-specific incorporation of a
photoreactive amino acid. Nature Methods, 2, 201–206.
Hino, N., Hayashi, A., Sakamoto, K., & Yokoyama, S. (2006). Site-specific incorporation of
non-natural amino acids into proteins in mammalian cells with an expanded genetic
code. Nature Protocols, 1, 2957–2962.
Horn, R., Ding, S., & Gruber, H. J. (2000). Immobilizing the moving parts of voltage-gated
ion channels. The Journal of General Physiology, 116, 461–476.
Huber, T., & Sakmar, T. P. (2014). Chemical biology methods for investigating G
protein-coupled receptor signaling. Chemistry & Biology, 21, 1224–1237.
Crosslinking glutamate receptor ion channels 191

Jonas, P. (1995). Fast application of agonists to isolated membrane patches. In Single-channel


recording (pp. 231–243). Boston, MA: Springer US.
Kalstrup, T., & Blunck, R. (2015). Reinitiation at non-canonical start codons leads to leak
expression when incorporating unnatural amino acids. Scientific Reports, 5, 11866.
Karlin, A., & Akabas, M. H. (1998). Substituted-cysteine accessibility method. Methods in
Enzymology, 293, 123–145.
Kauer, J. C., Erickson-Viitanen, S., Wolfe, H. R., & DeGrado, W. F. (1986). P-benzoyl-
L-phenylalanine, a new photoreactive amino acid. Photolabeling of calmodulin with
a synthetic calmodulin-binding peptide. The Journal of Biological Chemistry, 261,
10695–10700.
Klippenstein, V. (2015). Photoinactivation of glutamate receptors by genetically encoded unnatural
amino acids. Dissertation.
Klippenstein, V., Ghisi, V., Wietstruk, M., & Plested, A. J. R. (2014). Photoinactivation of
glutamate receptors by genetically encoded unnatural amino acids. The Journal of
Neuroscience, 34, 980–991.
Kozak, M. (1986). Point mutations define a sequence flanking the AUG initiator codon that
modulates translation by eukaryotic ribosomes. Cell, 44, 283–292.
Langtangen, H. P. (2013). Python scripting for computational science. Springer Science & Business
Media.
Lau, A. Y., Salazar, H., Blachowicz, L., Ghisi, V., Plested, A. J. R., & Roux, B. (2013). A
conformational intermediate in glutamate receptor activation. Neuron, 79, 492–503.
Lester, R. A., Clements, J. D., Westbrook, G. L., & Jahr, C. E. (1990). Channel kinetics
determine the time course of NMDA receptor-mediated synaptic currents. Nature,
346, 565–567.
Liu, W., Brock, A., Chen, S., Chen, S., & Schultz, P. G. (2007). Genetic incorporation of
unnatural amino acids into proteins in mammalian cells. Nature Methods, 4, 239–244.
Meyerson, J. R., Chittori, S., Merk, A., Rao, P., Han, T. H., Serpe, M., et al. (2016).
Structural basis of kainate subtype glutamate receptor desensitization. Nature, 537,
567–571.
Murray, C. I., Westhoff, M., Eldstrom, J., Thompson, E., Emes, R., & Fedida, D. (2016).
Unnatural amino acid photo-crosslinking of the IKs channel complex demonstrates a
KCNE1:KCNQ1 stoichiometry of up to 4:4. eLife, 5.
NEHER, E., & SAKMANN, B. (1976). Single-channel currents recorded from membrane
of denervated frog muscle fibres. Nature, 260, 799–802.
Poulsen, M. H., Poshtiban, A., Klippenstein, V., Ghisi, V., & Plested, A. J. R. (2019). Gating
modules of the AMPA receptor pore domain revealed by unnatural amino acid muta-
genesis. Proceedings of the National Academy of Sciences of the United States of America,
116, 13358–13367.
Preston, G. W., & Wilson, A. J. (2013). Photo-induced covalent cross-linking for the analysis
of biomolecular interactions. Chemical Society Reviews, 42, 3289–3301.
Reyes, N., Ginter, C., & Boudker, O. (2009). Transport mechanism of a bacterial homo-
logue of glutamate transporters. Nature, 462, 880–885.
Robert, A., & Howe, J. R. (2003). How AMPA receptor desensitization depends on recep-
tor occupancy. The Journal of Neuroscience, 23, 847–858.
Rook, M. L., Williamson, A., Lueck, J. D., Musgaard, M., & Maclean, D. M. (2020). β11-12
linker isomerization governs acid-sensing ion channel desensitization and recovery.
eLife, 9.
Sachs, F. (1999). Practical limits on the maximal speed of solution exchange for patch clamp
experiments. Biophysical Journal, 77, 682–690.
Salazar, H., Mischke, S., & Plested, A. J. R. (2020). Measurements of the timescale and con-
formational space of AMPA receptor desensitization. Biophysical Journal, 119, 206–218.
192 Andrew J. R. Plested and Mette H. Poulsen

odinger, L. L. C. (2015). The PyMOL molecular graphics system, version 2.0.


Schr€
Schwyzer, R., & Caviezel, M. (1971). P-azido-L-phenylalanine: A photo-affinity ‘probe’
related to tyrosine. Helvetica Chimica Acta, 54, 1395–1400.
Seidel, L., Zarzycka, B., Zaidi, S. A., Katritch, V., & Coin, I. (2017). Structural insight into
the activation of a class B G-protein-coupled receptor by peptide hormones in live
human cells. eLife, 6. https://doi.org/10.7554/eLife.27711.
Skuzeski, J. M., Nichols, L. M., Gesteland, R. F., & Atkins, J. F. (1991). The signal for a leaky
UAG stop codon in several plant viruses includes the two downstream codons. Journal
of Molecular Biology, 218, 365–373.
Sobolevsky, A. I., Yelshansky, M. V., & Wollmuth, L. P. (2004). The outer pore of the
glutamate receptor channel has 2-fold rotational symmetry. Neuron, 41, 367–378.
Stern-Bach, Y., Russo, S., Neuman, M., & Rosenmund, C. (1998). A point mutation in the
glutamate binding site blocks desensitization of AMPA receptors. Neuron, 21, 907–918.
Sun, M. A., Wang, Y., Zhang, Q., Xia, Y., Ge, W., & Guo, D. (2017). Prediction of revers-
ible disulfide based on features from local structural signatures. BMC Genomics, 18, 279.
Tajima, N., Karakas, E., Grant, T., Simorowski, N., Diaz-Avalos, R., Grigorieff, N., et al.
(2016). Activation of NMDA receptors and the mechanism of inhibition by ifenprodil.
Nature, 534, 63–68.
Unwin, N. (1995). Acetylcholine receptor channel imaged in the open state. Nature, 373,
37–43.
Wang, W., Takimoto, J. K., Louie, G. V., Baiga, T. J., Noel, J. P., Lee, K.-F., et al. (2007).
Genetically encoding unnatural amino acids for cellular and neuronal studies. Nature
Neuroscience, 10, 1063–1072.
Winn, M. D., Ballard, C. C., Cowtan, K. D., Dodson, E. J., Emsley, P., Evans, P. R., et al.
(2011). Overview of the CCP4 suite and current developments. Acta Crystallographica.
Section D, Biological Crystallography, 67, 235–242.
Xu, Y., Ramu, Y., Shin, H.-G., Yamakaze, J., & Lu, Z. (2013). Energetic role of the paddle
motif in voltage gating of shaker K(+) channels. Nature Structural & Molecular Biology, 20,
574–581.
Yang, N., & Horn, R. (1995). Evidence for voltage-dependent S4 movement in sodium
channels. Neuron, 15, 213–218.
Yang, Y., Song, H., He, D., Zhang, S., Dai, S., Lin, S., et al. (2016). Genetically encoded
protein photocrosslinker with a transferable mass spectrometry-identifiable label. Nature
Communications, 7.
Ye, S., K€ ohrer, C., Huber, T., Kazmi, M., Sachdev, P., Yan, E. C. Y., et al. (2008).
Site-specific incorporation of keto amino acids into functional G protein-coupled recep-
tors using unnatural amino acid mutagenesis. The Journal of Biological Chemistry, 283,
1525–1533.
Ye, S., Huber, T., Vogel, R., & Sakmar, T. P. (2009). FTIR analysis of GPCR activation
using azido probes. Nature Chemical Biology, 5, 397–399.
Zhang, M., Lin, S., Song, X., Liu, J., Fu, Y., Ge, X., et al. (2011). A genetically incorporated
crosslinker reveals chaperone cooperation in acid resistance. Nature Chemical Biology, 7,
671–677.
Zhu, S., Riou, M., Yao, C. A., Carvalho, S., Rodriguez, P. C., Bensaude, O., et al. (2014).
Genetically encoding a light switch in an ionotropic glutamate receptor reveals subunit-
specific interfaces. Proceedings of the National Academy of Sciences of the United States of
America, 111, 6081–6086.

You might also like