(corresponding to segmentsa37 toa47 in the
structure) of Nup160 were missing in a patient
with steroid-resistant nephrotic syndrome ( 34 ).
Our EM maps identify two Nup93 molecules
within each CR subunit, expanding the list of
ACE1-containing nucleoporins (ACE1 proteins)
within the CR. In the updated model of the
CR subunit, four of sevena-helical domain–
containing nucleoporins have the ACE1 fold
( 20 ), which exhibit a similar conformation in
both inner and outer copies (fig. S17). Cross-
talk among the ACE1 domains of Nup93-O,
Nup107-O, and Nup96-I, which form an ACE1
core at the stem region of the Y complexes, is
mediated through interfaces that differ from
the canonical crown-crown association ( 35 )
(Fig. 4B). Our recent analysis of the IR subunit
also revealed a homodimerization interface of
Nup93 through its crown and tail ( 22 ). These
observations corroborate a key role of the
multifaceted ACE1 proteins within the NPC
assembly ( 20 ). Published EM maps ( 23 , 36 , 37 )
indicate that the Nup93-ACE1-O is also pre-
sent in the NR subunit from a vertebrate NPC.
Together with four molecules of Nup93 in the
IR subunit ( 12 , 22 , 25 , 30 ), there are at least
seven Nup93 molecules within each NPC spoke.
Both Nup205 and Nup188 adopt a super-
helical fold that is structurally reminiscent of
the karyopherin family ( 24 , 38 ). Our EM map
of the core region reveals that two Nup205
molecules, rather than Nup188 molecules, are
present within each CR subunit. The IR sub-
unit of theX. laevisNPC also comprises two
Nup205 molecules ( 22 ), whereas only one
Nup205-O molecule is found in the EM maps
of the NR subunit fromX. laevis( 23 , 36 )and
Homo sapiens( 37 ). These data suggest that at
least five Nup205 molecules are present with-
in each NPC spoke. The potentially differen-
tial placement of Nup205 in the CR and NR
mark one point of structural asymmetry be-
tween these two rings.
Among the eighta-helical nucleoporins in
the CR subunit, Nup85, Nup96, Nup93, Nup107,
and Nup205 each contain a single helical
solenoid that confers their structural elasticity
(fig. S17). Nup133 and Nup160 each contain an
elongateda-helical domain. Unlike the others,
Nup358 contains multiple tetratricopeptide re-
peat (TPR)–like repeats ( 39 ) that are clustered
to two distinct superhelical solenoids. The
unique shape and structural features of Nup358
further support our assignment of Nup358 to
the five clamp-shaped EM densities. Relative
movements of the two solenoids confer ad-
ditional conformational elasticity on Nup358
(fig. S13B). Identification of the fifth copy of
Nup358 expands our understanding of the
role of Nup358 and highlights its complex
modality in the assembly of CR. Both Nup358
clamp 1 and Nup93-ACE1-O engage in direct
interfaces that connect the inner and outer
Y complexes (Figs. 4 and 5 and figs. S11 and
S14). Such interweaving may strengthen the
association between the two concentric rings
of the vertebrate CR ( 10 ). In total, our current
model accounts for <30% of the full-length
Nup358. The invisible functional domains of
Nup358 may project into the cytosol for events
such as Ran binding ( 7 ). Because of residual
anisotropy and structural flexibility, the cur-
rent EM density falls short of sequence assign-
ment for the peripheral regions of the CR
subunit, including five Nup358 clamps and
the two Nup93 ACE1 domains. Future improve-
ment for these regions of the CR subunit may
confer a more conclusive assignment.
In summary, our final model reported in
this study, with >200 nucleoporins assigned, is
estimated to account for ~90% of the molec-
ular mass for the ordered region of a verte-
brateCR.AssemblyoftheCRscaffoldrelieson
thea-helical nucleoporins; the CR scaffold is
structurally solidified by 12b-propellers in
each CR subunit (fig. S18).Materials and Methods
Cryo-EM sample preparation
The NE fromX. laevisoocytes was prepared
largely as previously described ( 9 , 14 )butwith
the following improvements. First, to reduce
conformational heterogeneity of the NPC caused
by mechanical distortion, the opened NE was
spread onto the grids as gently as possible,
with minimal force applied. This practice is
in contrast to methods used in previous studies
( 9 , 14 ), in which mechanical force was applied
to ensure flat placement of the NE on the grids.
Second, to stabilize the NPC conformation, the
NE on the grids was cross-linked for 30 min on
ice using 0.5% glutaraldehyde in a low-salt
buffer (10 mM Hepes, pH 7.5, 1 mM KCl, and
0.5 mM MgCl 2 ). Third, copper grids were re-
placed by gold ones in this study. Finally, to
improve sample uniformity, each batch of sam-
ples was prepared on the same day using
oocytes from the same frog. The gold EM grids
(R1.2/1.3, R2/1, and R2/2; Quantifoil, Jena,
Germany) were blotted for 8 s with a blot force
of 15 and vitrified by plunge-freezing into liquid
ethane using a Vitrobot Mark IV (Thermo
Fisher Scientific) at 8°C under 100% humidity.
Thequalityofthesamplewasexaminedusing
an FEI Glacios microscope (Thermo Fisher
Scientific) operating at 200 kV.Data acquisition of intact X. laevis NPC
Grids were transferred to a Titan Krios electron
microscope (FEI) operating at 300 kV and
equipped with a Gatan GIF Quantum energy
filter (slit width 20 eV). A total of 46,143
micrographs were recorded with the grids tilt-
ing at angles of 0°, 30°, 45°, and 55° ( 14 , 40 ). A
K3 detector (Gatan) was used in the super-
resolution mode with a nominal magnifica-
tion of 64,000×, resulting in a calibrated pixel
size of 0.6935 Å for the movie files (fig. S1 andtable S1). The movie images were then binned
twice during motion correction, arriving at
a pixel size of 1.387 Å for the final motion-
corrected images. The total dose followed a
cosinealphaschemeinwhichthedoseis
inversely proportional to the cosine value of
the tilting angle. Within each stack, the ex-
posure time for each frame and the dose rate
were kept the same. Detailed statistics of data
collection are reported in table S1. All frames
in each stack were first aligned and summed
using MotionCor2 ( 41 ). Dose weighting was
also performed using MotionCor2 ( 41 ). The av-
erage defocus values were set between–1.5 and- 3.0 mm and estimated using Gctf ( 42 ).
Initial model of the CR
We examined all 46,143 micrographs and man-
ually selected 33,747 for further processing.
A total of 800,825 particles were manually
selected from these micrographs (fig. S2A).
Initial defocus estimation was performed as
previously described ( 14 ) before all other data-
processing procedures.
We used the 18-Å map of the CR from our
previous study ( 14 ) as the initial reference, which
was resampled to a pixel spacing of 11.096 Å
and low-pass filtered to 40 Å to reduce po-
tential model bias. We then performed one run
of a global search (K = 1) three-dimensional
classification, with the solvent mask covering
onlytheCR.Thesearchhad40iterations.The
resulting star files from the last several itera-
tions were then subjected to local search three-
dimensional classifications using multiple
reference seeds for guidance. All good classes
were then merged and duplicated particles were
removed, generating a dataset of 660,302 NPC
particles. A final round of autorefinement with
a solvent mask on the CR resulted in a recon-
structionat22-Åresolution(fig.S2A).The
resulting reconstruction was nearly identical
to that in our previous study ( 14 )(fig.S2B).C8
symmetry was applied throughout this stage
of data processing.Data processing and reconstruction of the
CR subunit
We extracted the CR subunit particles on the
basis of the alignment parameters of the 22-Å
CR reconstruction. We updated the orienta-
tion, shift, and defocus parameters for each
subunit according to a published protocol ( 14 ).
In short, a cropping center for each subunit
was defined in the map, and the Euler angles
alongwithtwo-dimensionalshiftsforeachsub-
unit particle were deduced from the CR particle.
The recentering procedure was repeated eight
times, defining a three-dimensional geometric
relationship among eight subunits of the same
CR. A total of 5,148,474 particles of the CR sub-
unit were extracted using a box size of 256 and a
binned pixel size of 2.774 Å (fig. S3A). We then
performed three rounds of CTF refinement withZhuet al., Science 376 , eabl8280 (2022) 10 June 2022 6of10
RESEARCH | STRUCTURE OF THE NUCLEAR PORE