Mouse long-chain acyl-CoA synthetase 1 is active as a monomer

Article abstract of DOI:10.1016/j.abb.2021.108773

Fatty acids are essential cellular building blocks and a major energy source. Regardless of their metabolic fate, fatty acids first need to be activated by forming a thioester with a coenzyme A group. This reaction is carried out by acyl-CoA synthetases (ACSs), of which ACSL1 (long-chain acyl-CoA synthetase 1) is an important member. Two bacterial homologues of ACSL1 crystal structures have been solved previously. One is a soluble dimeric protein, and the other is a monomeric peripheral membrane protein. The mammalian ACSL1 is a membrane protein with an N-terminal transmembrane helix. To characterize the mammalian ACSL1, we purified the full-length mouse ACSL1 and reconstituted it into lipid nanodiscs. Using enzymatic assays, mutational analysis, and cryo-electron microscopy, we show that mouse ACSL1 is active as a monomer.

Full text of DOI:10.1016/j.abb.2021.108773

Archives of Biochemistry and Biophysics 700 (2021) 108773
Contents lists available at ScienceDirect
Archives of Biochemistry and Biophysics
journal homepage: www.elsevier.com/locate/yabbi
Mouse long-chain acyl-CoA synthetase 1 is active as a monomer
Holly Dykstra, Chelsea Fisk, Cassi LaRose, Althea Waldhart, Xing Meng, Gongpu Zhao,
Ning Wu*
Van Andel Institute, Grand Rapids, MI, 49503, USA
A R T I C L E I N F O
A B S T R A C T
Keywords:
Fatty acids are essential cellular building blocks and a major energy source. Regardless of their metabolic fate,
fatty acids first need to be activated by forming a thioester with a coenzyme A group. This reaction is carried out
by acyl-CoA synthetases (ACSs), of which ACSL1 (long-chain acyl-CoA synthetase 1) is an important member.
Two bacterial homologues of ACSL1 crystal structures have been solved previously. One is a soluble dimeric
protein, and the other is a monomeric peripheral membrane protein. The mammalian ACSL1 is a membrane
protein with an N-terminal transmembrane helix. To characterize the mammalian ACSL1, we purified the full-
length mouse ACSL1 and reconstituted it into lipid nanodiscs. Using enzymatic assays, mutational analysis,
and cryo-electron microscopy, we show that mouse ACSL1 is active as a monomer.
Acyl-CoA synthetase
Fatty acid metabolism
Structure-function
Electron microscopy
Nanodisc
Enzyme kinetics
1. Introduction
interact with various protein networks to orchestrate FA metabolism
[3].
Fatty acids (FAs)1 are essential to cellular function. They need to be
activated by conjugating to coenzyme A (CoA) in order to be used for
β-oxidation, incorporation into phospholipids, or neutral lipid synthesis.
Acyl-CoA synthetases (ACSs) are classified into families according to the
chain length of their substrates: those families are named short-chain
ACS (ACSS), medium-chain ACS (ACSM), long-chain ACS (ACSL), very
long-chain ACS (ACSVL), Bubblegum ACS (ACSBG), and ACS family
(ACSF) [1]. The ACSL family is particularly important because those
enzymes esterify free FAs of 12–20 carbons, which are prevalent in the
human diet. There are five ACSL isoforms, and they differ in their tissue
distribution, subcellular localization, and substrate preference [2].
ACSL1 is expressed in many tissues, and especially highly in liver and
adipose tissue. It is found in the plasma membrane, outer membrane of
mitochondria, and the ER, depending on the cell type [3]. The first of the
two classic roles of ACSL1 is facilitating FA transport across the plasma
membrane where CoA attachment prevents free FAs from diffusing back
out of the cell [4]. The second role is channeling FAs toward β-oxidation
[5]. On the mitochondrial outer membrane, ACSL1 delivers acyl-CoA to
CPT1 proteins for conversion into acyl-carnitine, which then enters the
mitochondria for FA oxidation [3,6]. ACSL1 has been reported to
The current structural understanding of mammalian ACSL comes
largely from bacterial crystal structures. The crystal structures of a Thermus
thermophilus ACSL homologue, ttLC-FACS [7], and a Mycobacterium tuber-
culosis ACSVL homologue, FadD13 [8], have been reported. ttLC-FACS has
substrate length ranging from 12 to 18 carbons. It is a soluble enzyme that
works as a dimer with two independent active sites. FadD13 is a
membrane-bound peripheral enzyme that functions as a monomer with
substrate carbon length up to C26. Both share a two-domain structure,
having a large N-terminal domain and a small C-terminal domain. During
the reaction, the C-terminal domain moves towards the N-terminal domain
to complete the active site [9]. The general catalytic mechanism for these
enzymes is thought to involve two steps. First, the free FA condenses with
ATP to form an acyl-AMP intermediate with a release of pyrophosphate.
Second, the intermediate binds CoA, which replaces the AMP. In the
FadD13 structure, membrane attachment could allow the reaction to occur
without completely extracting the FA from the membrane. Unlike either of
the bacterial proteins, mammalian ACSL1 is a membrane protein with a
single N-terminal transmembrane helix predicted by sequence analysis [3].
For the mouse protein, it is amino acids 25LPTNTLMGFGA-
FAALTTFWYA45, which is 100% conserved between human, mouse, and
* Corresponding author.
E-mail address: Ning.wu@vai.org (N. Wu).
1
The abbreviations used are: FA, fatty acid; CoA, coenzyme A; ACS, acyl-CoA synthetase; ACSL1, long-chain ACS 1; ACSS, short-chain ACS; ACSM, medium-chain
ACS; ACSVL, very long-chain ACS; ACSBG, Bubblegum ACS; ACSF, ACS family; ER, endoplasmic reticulum; CPT, carnitine palmitoyl transferase; ttLC-FACS, long-
chain fatty ACS from T. thermophilus HB8; FadD13, bacterial fatty ACS; MSP, membrane scaffold protein; SMA, styrene-maleic acid copolymer; cryo-EM, cryo-electron
microscopy; EM, electron microscopy; 2D, two-dimensional; AppNp, adenosine 5’-(β, γ-imido) triphosphate; HPLC, high performance liquid chromatography.
https://doi.org/10.1016/j.abb.2021.108773
Received 31 October 2020; Received in revised form 15 January 2021; Accepted 16 January 2021
Available online 21 January 2021
0003-9861/© 2021 Elsevier Inc. All rights reserved.

H. Dykstra et al.
Archives of Biochemistry and Biophysics 700 (2021) 108773
rat ACSL1. It is unclear whether this N-terminal transmembrane helix of
ACSL1 is important for catalysis. It is also unclear whether the mammalian
ACSL1 is a monomer or a dimer.
developed to enclose membrane proteins and the lipids together, using
either lipid-binding proteins or polymers to generate soluble protein–lipid
particles called nanodiscs [10–12]. These methods use engineered de-
rivatives of apolipoprotein A1 scaffold proteins (called membrane scaffold
protein, MSP), saposin A, and styrene–maleic acid copolymer (SMA). These
have proven to be useful in enzyme activity assays without detergent and
One of the difficulties of working with membrane proteins is the need
for detergent to keep the proteins in solution. Detergents can adversely
affect protein stability and function. Recently, methods have been
Fig. 1. Full-length ACSL1 is a monomer by EM. (A). ACSL1 elution fractions 1 to 6 from Ni-NTA. The last lane is the PD10 buffer-exchanged protein after the Ni-NTA
column. (B). Superose 6 FPLC elution profile of ACSL1/soy polar lipids/MSPΔH5 nanodiscs. Standard molecular weight markers are indicated by dashed lines. (C). A
representative negative-stain EM micrograph of ACSL1 nanodiscs. (D). 2D class averages from negative-stain images, showing multiple ways of ACLS1 insertion into
the nanodisc. Yellow arrowheads point to ACSL1 and blue arrowheads to the nanodisc. Scale bar represents 10 nm. (E). A schematic model of the various methods of
ACSL1 insertion into nanodisc. (F). A representative cryo-EM micrograph at 30,000× magnification. (G). Representative cryo-EM 2D class averages of ACSL1 alone (I
and II) or in complex with the nanodisc (III–V). ACSL1 in complex with nanodisc had poorer resolution. Yellow arrowheads point to ACSL1 and blue arrowheads to
the nanodisc. Orange arrowheads point to the C-terminal domain in the open position. Scale bar represents 5 nm.
2

H. Dykstra et al.
Archives of Biochemistry and Biophysics 700 (2021) 108773
in high-resolution structural determinations by cryo-electron microscopy
(cryo-EM). Among the different types of protein–lipid particles, MSP is
relatively easy to use, and MSPs of different lengths have been developed
to generate nanodiscs of different sizes.
2.2. Full-length ACSL1 is active in nanodiscs
Because detergents used in membrane protein isolation and stabili-
zation can directly affect ACSL1 activity, we hypothesized that purifi-
cation and reconstitution into nanodiscs could provide a more native
environment for assessing enzymatic kinetics. Using the full-length
ACSL1 in nanodiscs, we performed ACSL1 activity assays using oleate
In this study, we successfully purified the full-length mouse ACSL1
and reconstituted it into nanodiscs using MSPΔH5, eliminating the need
to use detergents during enzyme assays. Using cryo-EM single-particle
analysis and an N-terminal-truncated version of ACSL1, we show that
mouse ACSL1 is active as a monomer.
(C18:1) as a substrate at concentrations ranging from 10 μM to 500 μM
and with saturating amounts of CoA and ATP. The product (AMP) was
monitored by HPLC (Fig. 2A and B). No detergent was present during the
assay. We found that the full-length protein was active and had a relative
2. Results
Km of 58 ± 4
μM for C18:1, which is on the lower end of the normal
plasma oleic acid concentration of 30
μ
M to 3.2 mM [13].
2.1. ACSL1 incorporates into lipid nanodiscs as a monomer
Due to the sequence homology between the T. thermophilus ttLC-
FACS and the mammalian ACSL [7], we hypothesized that the
conserved T277 and E463 in the mouse ACSL1 are important active site
residues, and that mutation of these two residues into alanine should
render the enzyme inactive. We therefore made a double alanine (AA)
mutant (T277A/E463A) and inserted it into the MSPΔH5 nanodiscs. The
mutant protein behaved similarly to the wild type during purification
(Fig. 2C), and just as expected, it was enzymatically inactive (Fig. 2D).
This finding lends strength to our purification methods.
ACSL1 is predicted to have a single transmembrane segment at the N-
terminus. Since the lipid environment may influence the function of the
enzyme, we decided to express the full-length protein in insect cells to
ensure proper incorporation of the protein into the membrane. Mouse
ACSL1 with a C-terminal His10 tag was cloned into the pFastBac system
and expressed in High Five cells. The protein expressed and purified well
using Ni-NTA affinity resin (Fig. 1A). The eluted protein was quickly
buffer-exchanged to eliminate the high concentration of imidazole using
a PD10 column and then was mixed with soy polar lipids that had
already been sonicated into small unilamellar liposomes. Then MSPΔH5
was added to encircle the protein within the lipid/detergent mixture; the
detergent was eliminated using biobeads. A typical size distribution of
ACSL1/MSPΔH5 nanodiscs on an analytical Superose 6 column is shown
in Fig. 1B. The center fraction of the peak at about 15 mL was used for
single-particle analysis by electron microscopy (EM).
2.3. ACSL1 Δ58 is catalytically active as a soluble enzyme
Cryo-EM data (Fig. 1G) showed that during grid freezing, some ACSL1
protein broke away from the nanodiscs. In addition, sequence alignment of
FadD13 and mouse ACSL1 showed none of the arginine residues important
for FadD13 association with the membrane are conserved in mouse ACSL1
(Supplementary Fig. 1). We wondered whether we could get higher-
resolution structures by just analyzing the soluble catalytic portion of
mouse ACSL1. We deleted amino acids 1–57 to eliminate the trans-
membrane region and some of the flexible linker region. The Δ58 truncated
protein was expressed in High Five cells to be consistent. It was soluble, so
we purified it from the cell lysate supernatant fraction instead of the
membrane fraction. No detergent was used during the purification.
Compared with the full-length protein in nanodisc, Δ58 was less stable and
tended to precipitate out upon storage at 4 ◦C. For the cryo-EM work, we
added AppNp and C18:1 to stabilize the protein, and for the enzymatic work
we used freshly prepared protein to minimize protein inactivation. As shown
in Fig. 3A, Δ58 complexed with AppNp and C18:1 ran as a single peak on
Superdex 200. We used SDS-PAGE and BSA standards to estimate the ACSL1
protein concentration (Fig. 3B). Using the same amount of protein as for full-
length WT ACSL1, we found the Δ58 still had activity, although the Km was
larger than the full-length protein (Fig. 3C). Nevertheless, the results support
the idea that mouse ACSL1 is active as a monomer.
First, we used negative-stain EM to analyze the oligomeric state of
ACSL1. The micrographs showed good particle distribution and density
and allowed visualization of the complex (Fig. 1C). The resulting
reference-free, two-dimensional (2D) classes showed that ACSL1 has
different ways to insert into the nanodisc (Fig. 1D and E). If the mouse
ACSL1 works as a dimer, then the MSPΔH5 would most likely enclose a
dimer together in one nanodisc. However, our negative-stain data
indicated that while multiple ACSL1 monomers were often incorporated
into a single nanodisc, they did not appear to be interacting. Indeed,
some 2D classes showed ACSL1 monomers inserting on opposite sides of
the nanodisc, while other classes showed two monomers on the same
side but with no apparent interaction. This observation led us to spec-
ulate that ACSL1 is a monomer.
To investigate ACSL1 structure at higher resolution, we examined
our sample via cryo-EM. ACSL1 is about 75 kDa, which is small for high-
resolution, single-particle analysis by cryo-EM. The bacterial enzyme
structures suggest that ACSL1 has two domains with the enzyme active
site formed at their interface [7,8]. The smaller C-terminal domain can
move away from the larger N-terminal domain to allow substrate
binding and product release. This variable conformational positioning
between the two domains increases the difficulty of particle alignment
in 2D classification. Therefore, we tested various substrates and prod-
ucts homologues in search of the most stable complex for cryo-EM study.
We found that the combination best for enzyme stability and cryo-EM
was AppNp (a nonhydrolyzable ATP analogue) and C18:1 (oleic acid).
Cryo-EM images showed that the particles were monodisperse (Fig. 1F).
Particles were autopicked from collected micrographs and then were
subjected to iterative, reference-free, 2D classifications. The resulting
classes showed ACSL1 complexed with the disc-like density of the nano-
disc in some classes, while in others, ACSL1 appears to have broken away
from the nanodiscs (Fig. 1G). Despite changing the grid-freezing param-
eters, we continued to see the same results. The resolution of 2D classes of
ACSL1 alone was superior to the resolution of 2D classes of ACSL1 in
complex with nanodiscs. There are many reasons for this, including 1) the
nanodisc orientation relative to ACSL1 is not fixed, and 2) the ACSL1
soluble domain is relatively small compared with the nanodisc. However,
it is clear from the 2D classes that ACSL1 exists as a monomer.
2.4. The Cryo-EM structure of Δ58 shows folding similar to that of
bacterial homologues
Next, we attempted to determine the structure of Δ58 ACSL1 via
cryo-EM to confirm it is a monomer. We assessed sample density and
distribution (Fig. 4A) and collected a modest-sized data set on Krios
(about 3400 micrographs). After picking particles and eliminating poor-
quality particles through iterative 2D classifications, the two-domain
structure can be seen in some 2D classes (Fig. 4B).
Using a total of 83,000 particles, the 3D reconstruction of Δ58 ACSL1
revealed two possible conformations, open (26% of particles) and closed
(52% of particles) (Fig. 4C). The remaining particles fell into classes with
poorly defined features (Supplementary Fig. 2). Further refinement of
each conformation was low-pass-filtered to a resolution of 10 Å with an
adjusted threshold of 0.0013 for the open conformation and with an
adjusted threshold of 0.0027 for the closed conformation. Comparing
the size of these two dominant classes with the crystal structure of
FadD13, they matched quite closely (Fig. 4C and D). This clearly sup-
ports that Δ58 ACSL1 is a monomer.
3

H. Dykstra et al.
Archives of Biochemistry and Biophysics 700 (2021) 108773
Fig. 2. Enzyme kinetics of ACSL1 nanodisc with oleic acid. (A). Initial velocity with various concentrations of oleic acid (C18:1). (B). Michaelis–Menten kinetic
enzyme activity curves from purified, full-length ACSL1 assayed with oleic acid. (C). Protein gel of ACSL1 full-length wild type (WT) and AA (double alanine) mutant
used to standardize the amounts of protein used for activity comparison. L = ladder; 1–5 = BSA standards (1 = 250 μg/mL, 2 = 125 μg/mL, 3 = 50 μg/mL, 4 = 25 μg/
mL, 5 = 5 μg/mL); 6–7 = WT samples in duplicate; 8–9 = WT sample 1:2 dilution in duplicate; 10–11 = AA mutant in duplicate; 12–13 = AA mutant 1:2 dilution in
duplicate. (D). Initial velocities comparison of WT and AA mutant at 500
than the symbols.
μ
M C18:1. Results are mean ± SD. When no error bars appear, the SD values were smaller
3. Discussion
that individual ACSL1 proteins do not interact with each other. The Δ58
truncation later confirmed that ACSL1 is active as a monomer. However,
We report the expression, purification, and lipid nanodisc insertion
of the mouse ACSL1 protein along with a low-resolution structure of the
mouse ACSL1 catalytic domain by cryo-EM. The major conclusion is
mouse ACSL1 has the same two-domain architecture as ttLC-FACS and
FadD13. The AA mutant data and the sequence homology [7] supports
that mouse ACSL1 works as a monomer and follows the same catalytic
steps as that reported for ttLC-FACS.
it is important to point out that Δ58 (especially the apo protein) was
much less stable than the full-length protein in nanodiscs during storage
at 4 ◦C and was less active than the full-length protein. Protein stability
may contribute to the lower Vmax of Δ58. The higher Km value for Δ58
may indicate that the membrane surface may be important for maximum
catalytic efficiency of ACSL. There are about 40 amino acid linkers
predicted between the transmembrane helix and the rest of the folded
structure. From the EM images, we did not detect any significant asso-
ciation of the soluble folded domains with the nanodisc lipid surface. In
fact, the ACSL1 and the nanodisc seem to be able to rotate freely relative
to each other. Therefore, we do not know how the transmembrane helix
or the close proximity to the lipid surface contribute to catalysis.
From our cryo-EM studies on the full-length ACSL1 in nanodiscs, we
found the conditions used to freeze EM grids were harsh enough to cause
some protein to come out of its lipid environment. This was quite sur-
prising to us. We do not know if this property is generally true for single
transmembrane proteins or is specific to ACSL1. However, it has been
reported that the air-water interface is detrimental to low affinity protein
complexes and can even cause protein denaturation [15,16]. This freezing
effect may possibly affect the ligand occupancy in our Δ58 structure. We
saturated the protein solution with ligands before freezing the grids, so we
should only see one closed conformation. However, the final classification
showed two major conformations, not just one. The open conformation
implies that the ligands must have dissociated from the protein.
The difficulties of studying the kinetics of lipid-modifying enzymes
include the extraction and purification of membrane-bound proteins
from their hydrophobic environment while keeping them stable and
active. The detergents used for protein purification can introduce arti-
facts because the enzyme may need the native lipid to function properly.
We used fos-choline 12 to extract ACSL1, quickly replaced the detergent
with soy polar lipids, and then successfully enclosed ACSL1 and lipids in
MSPΔH5 nanodiscs. This stabilized ACSL1 and we could carry out ki-
netic assays without any detergent. Compared to the published values
[14], our Km of the full-length ACSL1 was larger (58 ± 4
μ
M vs. 6.7 ±
M) and the Vmax was smaller (454 ± 8 nmol/min/mg vs. 2089 ±
μmol/min/mg), likely due to these modifications. In addition,
1.4
μ
108
cyclodextrin was used in the reaction to keep the oleic acid in solution
because no detergent was present. This may affect the accessibility and
kinetics of the ACSL1 and oleic acid interaction. As to the real specific
activity in vivo, there are more parameters to be considered, such as the
local availability/concentration of FAs and whether the substrate needs
to be delivered to the enzyme via a FA–binding protein.
One important function of ACSL1 is to provide acyl-CoA to CPT1 for
transport into the mitochondria for oxidation. Two known ways that FA
oxidation can be regulated via CPT1A is its expression increase induced
by high fat content in diet [17] and by the inhibition of CPT1 activity by
From our negative-stain EM studies, one can see that multiple ACSL1
proteins can be enclosed within one MSPΔH5 nanodisc, sometimes on
the same side of the nanodisc and sometimes on opposite sides. It is clear
4

H. Dykstra et al.
Archives of Biochemistry and Biophysics 700 (2021) 108773
Fig. 3. Truncation mutant Δ58 purification and activity. (A). Superdex 200 FPLC elution of Δ58 ACSL1 at about 11 mL. The second peak is AppNp added to stabilize
the protein. Standard molecular weight markers are indicated by dashed lines. (B). Protein gel of Δ58 and full-length (FL) ACSL1 showing the relative amount of
protein used for activity assay. L = ladder; 1–5 = BSA standards (1 = 250 μg/mL, 2 = 125 μg/mL, 3 = 50 μg/mL, 4 = 25 μg/mL, 5 = 5 μg/mL); 6–7 = Δ58 in
duplicate; 8–9 = Δ58 1:2 dilution in duplicate; 10–11 = FL in duplicate; 12–13 = FL 1:2 dilution in duplicate. (C). Michaelis–Menten kinetic enzyme activity curves
from purified Δ58 ACSL1 assayed with oleate (C18:1). Results are mean ± SD. When no error bars appear, the SD values were smaller than the symbols.
malonyl-CoA [18], product of FA synthesis. In addition, CPT1A has been
reported to be an oligomer [19] that shifts between hexameric and
trimeric states [20,21]. It is conceivable that ACSL1 interaction with
CPT1A is also regulated by metabolic demands. Knowing that ACSL1 is a
monomer can help clarify the structure-function of ACSL1 and its rela-
tionship with associated proteins.
The frozen cells were thawed and mechanically homogenized with a
Teflon tissue homogenizer. Cell membranes were collected by centrifu-
gation and solubilized with resuspension buffer (25 mM Tris 7.5, 150 mM
NaCl, 10 mM MgSO4, 5% glycerol, 0.2% fos-choline 12). For purification
of the recombinant protein, the supernatant was batch bound with Ni-NTA
beads (Qiagen) for 1 h and was packed into an empty PD10 column and
washed with 25 mM Tris 7.5, 150 mM NaCl, 30 mM imidazole, 10 mM
MgSO4, 5% glycerol, 0.00125% soy PC (95%), and 0.05% fos-choline 12.
Protein was eluted with 300 mM imidazole in the wash buffer. The most
concentrated fractions were buffer-exchanged into 30 mM Tris 7.5, 50 mM
NaCl, 0.5 mM MgSO4, 0.00125% soy PC (95%), and 0.05% fos-choline 12
using a PD10 column. Soy polar lipids were added to the buffer-exchanged
protein. For stabilization, 0.5 mM AppNp and 0.15 mM oleic acid were
added, chosen specifically for their compatibility with future structural
analysis. Following a 2 h incubation period, MSPΔH5 was added for 1 h.
BioBeads were added overnight to adsorb fos-choline 12. The resulting
nanodiscs were concentrated using a 30-kDa cut-off concentrator and
loaded onto an analytical Superose 6 SEC column, equilibrated with 30
mM Tris, pH 7.5, 50 mM NaCl, and 0.25 mM TCEP. The correct fractions
were collected and stored on ice for further analysis.
4. Methods and materials
4.1. Cloning
Mouse ACSL1 was cloned into a homemade pFastBac1-His10 vector
at BamHI and NotI sites, with His10 tag at the C-terminus. Baculovirus
was generated from SF9 cells according to the Bac-to-Bac manufac-
turer’s instructions (Invitrogen). High Five cells were infected with the
baculovirus for protein production.
4.2. Protein purification
Soy polar lipids (Avanti 541602) in chloroform were dried with ni-
trogen gas and rehydrated with buffer (30 mM Tris pH 7.5, 50 mM NaCl,
0.1 mM MgSO4) and probe-sonicated to prepare unilamellar, uniform
liposomes. Soy PC (95%) (Avanti 441601) was similarly rehydrated in
resuspension buffer (25 mM Tris 7.5, 150 mM NaCl, 10 mM MgSO4, 5%
glycerol, 0.2% fos-choline 12) and probe-sonicated.
4.3. Enzyme activity assay
Enzyme activity was measured by AMP formation as analyzed on
an HPLC (Dionex Ultimate 3000 UHPLC+) at 259 nm. The reaction
5

H. Dykstra et al.
Archives of Biochemistry and Biophysics 700 (2021) 108773
Fig. 4. Single-particle cryo-EM analysis of Δ58 complexed with AppNp and C18:1. (A). A representative cryo-EM micrograph at 215,000× magnification. (B).
Representative 2D class averages showing secondary structure elements of Δ58 ACSL1. Orange arrowheads point to the C-terminal domain in the open position. Scale
bar represents 5 nm. (C). The Δ58 ACSL1 single-particle reconstruction in the open (left) and closed (right) conformation. 3D reconstructions were low-pass-filtered
to 10 Å with a raised threshold of 0.0013 (open) and 0.0027 (closed). (D). Rainbow coloring from blue (N-terminal) to red (C-terminal) of the previously published
Mycobacterium tuberculosis FadD13 crystal structure (PDB ID: 3R44) with comparable dimensions and shape to the Δ58 ACSL1 3D reconstruction.
was carried out in 30 mM Tris 7.5, 50 mM NaCl, 0.5 mM MgSO4, 2.5
mg/mL cyclodextrin, with 200 M CoA, 1 mM ATP, and various
4.5. Cryo-EM sample preparation and data acquisition
μ
concentrations of oleic acid. The reaction was allowed to proceed for
between 1 and 7 min. Each time point was terminated with chloro-
form. Chloroform extraction was performed a second time to remove
all protein and lipids. The resulting aqueous sample was mixed with
an equal amount of 5% perchloric acid for AMP stability then run
isocratically with a mobile phase (100 mM sodium acetate, 75 mM
monosodium phosphate, pH 4.6) on a C18 column (Thermo fisher BDS
Hypersil). Michaelis–Menten kinetic enzyme activity curves were
drawn, and Vmax and Km values were calculated using GraphPad Prism
8.0.
Quantifoil R1.2/1.3 300-mesh Au grids (SPI) were used with all
samples. In our case, negative glow discharging of the grids led to a
strongly preferred orientation of our ACSL1 complexes. We also sus-
pected denaturation of our complex at the air–water interface. Orien-
tational bias makes three-dimensional analysis very challenging so we
experimented with alternate methods of positive glow-charging. We had
the best results when grids were glow-discharged in the presence of
amylamine for 20 s which yielded positively charged grids and improved
particle orientation. The protein sample (3 μL) was applied to the grids,
blotted for 4 s, and then plunge-frozen in liquid ethane using a Thermo
Fisher Scientific Vitrobot Mark IV at 100% RH and 4 ◦C.
For ACSL1 in lipid nanodiscs: Cryo-EM data was collected using the
Thermo Fisher Scientific Talos Arctica electron microscope operating at
200 kV. Images were acquired with a K2 Summit direct detection camera
(Gatan) operating in counting mode at a nominal 36,000× magnifica-
tion, corresponding to a calibrated physical pixel size of 1.156 Å with a
4.4. Negative-staining electron microscopy
To determine sample quality and incorporation into the nanodisc,
negative-stain grids were prepared and imaged. The protein sample was
applied to a freshly plasma-cleaned, carbon-coated, copper grid (300-
mesh, EMS) for 1 min. Grids were then blotted and stained with two 20
defocus range between ꢀ 1.5 and ꢀ 2.1 μm, using SerialEM, an auto-
mated acquisition program [23]. Each micrograph was exposed for 6 s
with 10 e–/A2/s dose rate (total accumulated dose, 60 e–/A2), and about
30 frames were captured per movie stack.
μ
L droplets of 2% uranyl formate. Images were collected at a nominal
30,000× magnification on a Thermo Fisher Scientific Tecnai Spirit G2
BioTWIN microscope at 120 kV with a Gatan Orius 830 CCD camera.
Image analysis, particle picking, and 2D class averaging were performed
using RELION 3.0 [22].
For Δ58 ACSL1: Cryo-EM data were collected using the Thermo
Fisher Scientific Titan Krios electron microscope operating at 300 kV,
6

H. Dykstra et al.
Archives of Biochemistry and Biophysics 700 (2021) 108773
aligned for parallel illumination. Images were acquired with a K2
Summit direct detection camera (Gatan) and a phase plate (FEI) oper-
ating in super-resolution counting mode at a nominal 215,000×
magnification, with a super-resolution pixel size of 0.629 Å. Over a
period of 3 d, 3398 images with defocus values ranging from ꢀ 0.4 to
ꢀ 0.8 Å were recorded with a dose rate of 8.6 e–/A2/s. Each micrograph
was exposed for 8 s, resulting in an accumulated dose of 68.8 e–/A2 and a
total of 40 frames per stack, using SerialEM.
Author contributions
Ning Wu: conceptualization, data curation, formal analysis, funding
acquisition, investigation, methodology, project administration, re-
sources, supervision, validation, visualization, writing – original draft
preparation, writing – review and editing. Holly Dykstra: data curation,
formal analysis, investigation, methodology, software, visualization,
writing – original draft preparation, writing – review and editing.
Chelsea Fisk: formal analysis, investigation. Cassi LaRose: investiga-
tion, methodology. Althea Waldhart: investigation. Xing Meng: formal
analysis, methodology, software, writing – review and editing. Gongpu
Zhao: methodology, resources, software.
4.6. Cryo-EM image processing and 3D reconstruction
Movies were motion-corrected with dose weighting using Motion-
Cor2 [24]. The remainder of the processing was performed in RELION
3.0. Contrast transfer function (CTF) was determined using CTFFIND4.1
(512 box size, 30 Å minimum resolution, 5 Å maximum resolution, 0.10
amplitude contrast) [25].
Declaration of competing interest
The authors declare that they have no competing interests.
Acknowledgments
For ACSL1 in lipid nanodiscs: RELION 3.0’s Laplacian-of-Gaussian
was used to automatically pick an initial set of particles that were
then subjected to reference-free 2D classification [26]. The best classes
that represented views of ACSL1 in lipid nanodiscs were then used for
template-based particle picking. A total of 176,167 particles were
extracted (1.156 Å/pixel, 150-pixel box size) from 798 dose-weighted
micrographs and subjected to reference-free 2D classification using a
256 Å mask diameter. The best 2D class averages showing a range of
views were selected and subjected to more rounds of reference-free 2D
classification. 2D classification results showed classes with ACSL1
broken off from the nanodisc as well as classes with ACSL1 in complex
with the nanodisc. Since classes with ACSL1 alone (no nanodisc) showed
higher resolution, we used a truncated version of ACSL1 that did not
require the nanodisc to examine the ACSL1 structure to greater
resolution.
This work was supported by Van Andel Institute funding to NW. We
also thank Ruda de Luna Almeida Santos (Van Andel Institute) for
technical assistance with cryo-EM sample preparation and image pro-
cessing and David Nadziejka (Van Andel Institute) for critical reading of
the manuscript.
Appendix A. Supplementary data
Supplementary data to this article can be found online at https://doi.
org/10.1016/j.abb.2021.108773.
References
For Δ58 ACSL1: Aligned micrographs with a CTF max resolution
above 8.0 were eliminated from further processing. Manual picking was
used to pick an initial subset (about 7500 particles) that were then
extracted (224-pixel box size) and subjected to reference-free 2D clas-
sification. The best seven classes that represented various views of
ACSL1 were then used for template-based particle picking. From 2708
dose-weighted micrographs, 682,548 particles were extracted and bin-
ned 4 × 4 (2.516 Å/pixel, 56-pixel box size) and subjected to reference-
free 2D classification using a 120 Å mask diameter. The best 2D class
averages showing top/bottom and side views were selected and sub-
jected to more rounds of reference-free 2D classification using a 110 Å
mask diameter. The best 2D class averages showing secondary-structure
and multiple views were selected (83,000 particles) for further pro-
cessing and were recentered and reextracted unbinned (0.629 Å/pixel,
224-pixel box size). An ab initio map generated in RELION without
symmetry imposed (C1) was low-pass-filtered to 60 Å and was used as an
initial reference map for 3D classification with C1 symmetry and four
classes. 3D classification revealed that most particles separated into one
of two classes, representing a possible closed (52%) and an open (26%)
conformation (as judged by the differing positions of the C-terminal
domain relative to the N-terminal domain). The stable classes showing
the potential closed conformation (44,125 particles) and open confor-
mation (21,620 particles) showed detailed secondary structure and were
each subjected to 3D refinement. Postprocessing (sharpening and
masking) resulted in a closed conformation map that was low-pass-
filtered to a resolution of 10 Å and adjusted to a threshold of 0.0013
and an open conformation map that was low-pass-filtered to a resolution
of 10 Å and a adjusted to a threshold of 0.0027 using UCSF Chimera
[27]. The previously solved Mycobacterium tuberculosis fatty acyl-CoA
synthetase crystal structure (PDB ID: 3R44) [8] was visualized and
measured in Chimera in order to compare the size and shape of the
monomer to our 3D reconstruction. Figures were prepared using UCSF
Chimera.
[1] M. Lopes-Marques, I. Cunha, M.A. Reis-Henriques, M.M. Santos, L.F. Castro,
Diversity and history of the long-chain acyl-CoA synthetase (Acsl) gene family in
vertebrates, BMC Evol. Biol. 13 (2013) 271.
[2] D.G. Mashek, L.O. Li, R.A. Coleman, Rat long-chain acyl-CoA synthetase mRNA,
protein, and activity vary in tissue distribution and in response to diet, J. Lipid Res.
47 (2006) 2004–2010.
[3] P.A. Young, C.E. Senkal, A.L. Suchanek, T.J. Grevengoed, D.D. Lin, L. Zhao, A.
E. Crunk, E.L. Klett, J. Fullekrug, L.M. Obeid, et al., Long-chain acyl-CoA
synthetase 1 interacts with key proteins that activate and direct fatty acids into
niche hepatic pathways, J. Biol. Chem. 293 (2018) 16724–16740.
[4] F. Gao, X.F. Yang, N. Fu, Y. Hu, Y. Ouyang, K. Qing, Roles of long-chain acyl
coenzyme A synthetase in absorption and transport of fatty acid, Chin. Med. Sci. J.
31 (2016) 62–64.
[5] J.M. Ellis, L.O. Li, P.C. Wu, T.R. Koves, O. Ilkayeva, R.D. Stevens, S.M. Watkins, D.
M. Muoio, R.A. Coleman, Adipose acyl-CoA synthetase-1 directs fatty acids toward
beta-oxidation and is required for cold thermogenesis, Cell Metabol. 12 (2010)
53–64.
[6] K. Lee, J. Kerner, C.L. Hoppel, Mitochondrial carnitine palmitoyltransferase 1a
(CPT1a) is part of an outer membrane fatty acid transfer complex, J. Biol. Chem.
286 (2011) 25655–25662.
[7] Y. Hisanaga, H. Ago, N. Nakagawa, K. Hamada, K. Ida, M. Yamamoto, T. Hori,
Y. Arii, M. Sugahara, S. Kuramitsu, et al., Structural basis of the substrate-specific
two-step catalysis of long chain fatty acyl-CoA synthetase dimer, J. Biol. Chem. 279
(2004) 31717–31726.
[8] C.S. Andersson, C.A. Lundgren, A. Magnusdottir, C. Ge, A. Wieslander, D. Martinez
Molina, M. Hogbom, The Mycobacterium tuberculosis very-long-chain fatty acyl-
CoA synthetase: structural basis for housing lipid substrates longer than the
enzyme, Structure 20 (2012) 1062–1070.
[9] G. Kochan, E.S. Pilka, F. von Delft, U. Oppermann, W.W. Yue, Structural snapshots
for the conformation-dependent catalysis by human medium-chain acyl-coenzyme
A synthetase ACSM2A, J. Mol. Biol. 388 (2009) 997–1008.
[10] I.G. Denisov, S.G. Sligar, Nanodiscs for structural and functional studies of
membrane proteins, Nat. Struct. Mol. Biol. 23 (2016) 481–486.
[11] J.M. Dorr, S. Scheidelaar, M.C. Koorengevel, J.J. Dominguez, M. Schafer, C.A. van
Walree, J.A. Killian, The styrene-maleic acid copolymer: a versatile tool in
membrane research, Eur. Biophys. J. 45 (2016) 3–21.
[12] J. Frauenfeld, R. Loving, J.P. Armache, A.F. Sonnen, F. Guettou, P. Moberg, L. Zhu,
C. Jegerschold, A. Flayhan, J.A. Briggs, et al., A saposin-lipoprotein nanoparticle
system for membrane proteins, Nat. Methods 13 (2016) 345–351.
[13] S.A. Abdelmagid, S.E. Clarke, D.E. Nielsen, A. Badawi, A. El-Sohemy, D.M. Mutch,
D.W. Ma, Comprehensive profiling of plasma fatty acid concentrations in young
healthy Canadian adults, PLoS One 10 (2015), e0116195.
7

H. Dykstra et al.
Archives of Biochemistry and Biophysics 700 (2021) 108773
[14] E.L. Klett, S. Chen, A. Yechoor, F.B. Lih, R.A. Coleman, Long-chain acyl-CoA
synthetase isoforms differ in preferences for eicosanoid species and long-chain fatty
acids, J. Lipid Res. 58 (2017) 884–894.
[21] Z.A. Jenei, G.Z. Warren, M. Hasan, V.A. Zammit, A.M. Dixon, Packing of
transmembrane domain 2 of carnitine palmitoyltransferase-1A affects
oligomerization and malonyl-CoA sensitivity of the mitochondrial outer membrane
protein, Faseb. J. 25 (2011) 4522–4530.
[15] E. D’Imprima, D. Floris, M. Joppe, R. Sanchez, M. Grininger, W. Kuhlbrandt,
Protein denaturation at the air-water interface and how to prevent it, Elife 8
(2019).
[22] S.H. Scheres, RELION: implementation of a Bayesian approach to cryo-EM
structure determination, J. Struct. Biol. 180 (2012) 519–530.
[16] R.M. Glaeser, Proteins, interfaces, and cryo-em grids, Curr. Opin. Colloid Interface
Sci. 34 (2018) 1–8.
[23] D.N. Mastronarde, Automated electron microscope tomography using robust
prediction of specimen movements, J. Struct. Biol. 152 (2005) 36–51.
[24] S.Q. Zheng, E. Palovcak, J.P. Armache, K.A. Verba, Y. Cheng, D.A. Agard,
MotionCor2: anisotropic correction of beam-induced motion for improved cryo-
electron microscopy, Nat. Methods 14 (2017) 331–332.
[17] L. Moody, G.B. Xu, H. Chen, Y.X. Pan, Epigenetic regulation of carnitine
palmitoyltransferase 1 (Cpt1a) by high fat diet, Biochim Biophys Acta Gene Regul
Mech 1862 (2019) 141–152.
[18] D.W. Foster, Malonyl-CoA: the regulator of fatty acid synthesis and oxidation,
J. Clin. Invest. 122 (2012) 1958–1959.
[25] A. Rohou, N. Grigorieff, CTFFIND4: fast and accurate defocus estimation from
electron micrographs, J. Struct. Biol. 192 (2015) 216–221.
[19] A. Faye, C. Esnous, N.T. Price, M.A. Onfray, J. Girard, C. Prip-Buus, Rat liver
carnitine palmitoyltransferase 1 forms an oligomeric complex within the outer
mitochondrial membrane, J. Biol. Chem. 282 (2007) 26908–26916.
[20] Z.A. Jenei, K. Borthwick, V.A. Zammit, A.M. Dixon, Self-association of
transmembrane domain 2 (TM2), but not TM1, in carnitine palmitoyltransferase
1A: role of GXXXG(A) motifs, J. Biol. Chem. 284 (2009) 6988–6997.
[26] J. Zivanov, T. Nakane, B.O. Forsberg, D. Kimanius, W.J. Hagen, E. Lindahl, S.
H. Scheres, New tools for automated high-resolution cryo-EM structure
determination in RELION-3, Elife 7 (2018).
[27] E.F. Pettersen, T.D. Goddard, C.C. Huang, G.S. Couch, D.M. Greenblatt, E.C. Meng,
T.E. Ferrin, UCSF Chimera–a visualization system for exploratory research and
analysis, J. Comput. Chem. 25 (2004) 1605–1612.
8