constitutive and acute depletion of LIS1 pheno-
copied dynein-dynactin inhibition and led to
spindle pole defocusing in aMTOC-free mouse
oocytes (fig. S9, F to M). To determine if dynein,
dynactin, and LIS1 function together, we in-
terfered with individual components, acutely
perturbed their poleward transport during
metaphase I with nocodazole, and examined
their accumulation at kinetochores ( 87 ). P150-
CC1–mediated inhibition and LIS1 depletion,
but not NUMA depletion, each disrupted the
localization of dynein, dynactin, and LIS1 in
aMTOC-free mouse oocytes (fig. S10). Thus,
stably associated NUMA, together with a
dynein-dynactin-LIS1 complex, clusters mi-
crotubule minus ends at aMTOC-free spin-
dle poles.
Dynein-dynactin focuses the spindle poles in
human oocytes
To test whether our findings in aMTOC-free
mouse oocytes translate to human oocytes, we
first analyzed the localization of dynactin at
different stages of meiosis in human oocytes.
Dynactin was not recruited during early spin-
dle assembly but localized predominantly
to kinetochores and spindle microtubules
before the metaphase spindle had assembled
(Fig. 3A). At metaphase I and II, dynactin
was most prominently detected at the spin-
dle poles (Fig. 3, A and B). Dynactin was in
proximity to LIS1 and NUMA at the spindle
poles (Fig. 3B), in line with their localization
in aMTOC-free mouse oocytes (figs. S7, A and
B, and S9, A and B).
We then inhibited dynein-dynactin in hu-
man oocytes using P150-CC1. NUMA remained
bound to microtubule minus ends and spindle
poles became defocused in dynein-inhibited
human oocytes (Fig. 3, C to E), recapitulating
our observations in aMTOC-free mouse oocytes
(fig. S8, B to D). Thus, dynein-dynactin and
possibly LIS1 are required for pole focusing in
human oocytes.
Human oocytes are deficient in the molecular
motor KIFC1/HSET
Next, we asked whether spindle instability is
a general feature of mammalian oocytes that
lack aMTOCs and use NUMA for spindle pole
organization. To this end, we performed live
imaging of spindle assembly in bovine, por-
cine, and aMTOC-free mouse oocytes. These
oocytes progressed through similar stages of
spindle assembly as human oocytes (figs. S4,
H and J, and S11, A and B; and movies S11
and S12). However, unstable spindle poles
were observed in only 6% of bovine oocytes,
4.4% of porcine oocytes, and in none of the
aMTOC-free mouse oocytes as compared to in
82% of human oocytes ( 2 ) (Fig. 3F and fig. S11,
C and D). Thus, additional mechanism(s) must
stabilize the spindles in nonhuman mamma-
lian oocytes.
When we examined the defocused spindles
in human and aMTOC-free mouse oocytes,
we noticed that around 40% of the defocused
spindles in human oocytes failed to align mi-
crotubules within the central region, whereas
none of the defocused spindles in aMTOC-free
mouse oocytes showed this misalignment
(Fig. 3, G and H). Given that multipolar spin-
dles were observed in a similar fraction of
untreated human oocytes, but in none of the
aMTOC-free mouse oocytes (Fig. 3, F and I),
we hypothesized that human oocytes lack a
spindle-stabilizing protein that is present in
mouse, bovine, and porcine oocytes. Deficiency
of this protein would thus lead to spindle
instability and misalignment of microtubules
within the central region of the human oocyte
spindle.
We designed an RNAi screen to identify the
protein that protects aMTOC-free mouse oo-
cytes from misalignment of microtubules within
the central region of the spindle. Specifically, we
codepleted NUMA with one of 20 candidate
proteins in aMTOC-free mouse oocytes using
follicle RNAi. The candidate proteins included
microtubule cross-linking proteins [ASPM,
DLG5/HURP (discs large MAGUK scaffold
protein 5), and TPX2] ( 42 , 88 ), proteins re-
lated to spindle bipolarization [HAUS6 (HAUS
augmin-like complex subunit 6), KIF11/EG5,
KIF15/HKLP2 (kinesin superfamily protein 15),
and KIFC1/HSET (kinesin superfamily protein
C1)] ( 42 , 89 , 90 ), proteins related to micro-
tubule dynamics [KIF2A (kinesin superfamily
protein 2A), KIF18A (kinesin superfamily pro-
tein 18A), CENPE (centromere-associated pro-
tein E), CLASP1 (CLIP-associated protein 1),
and CLASP2 (CLIP-associated protein 2)]
( 91 , 92 ), and proteins related to bridging fibers
and the central spindle [RACGAP1/CYK4 (Rac
GTPase-activating protein 1), PRC1 (protein
regulator of cytokinesis 1), KIF4A (kinesin
superfamily protein 4A), KIF12 (kinesin super-
family protein 12), KIF14 (kinesin superfamily
protein 14), KIF20A/MKLP2 (kinesin super-
family protein 20A), KIF20B/MPP1 (kinesin
superfamily protein 20B), and KIF23/MKLP1
(kinesin superfamily protein 23)] ( 93 , 94 ).
Whereas oocytes codepleted of NUMA and
DLG5, TPX2, or HAUS6 did not assemble a
spindle, oocytes codepleted of NUMA and all
other proteins except one assembled defocused
spindles with aligned microtubules within
the central region (Fig. 4A). Only oocytes co-
depleted of NUMA and KIFC1 failed to align
microtubules within the central region of the
spindle (Fig. 4A). This was further confirmed
by an automated analysis of microtubule di-
rectionality, which uses pseudocolor to rep-
resent variations in the orientation of spindle
microtubules (Fig. 4A).
To determine whether KIFC1 is deficient in
human oocytes, we examined data from pre-
vious proteomics studies on mouse and humanoocytes ( 95 , 96 ). We noticed that KIFC1 could
only be detected in the mouse dataset. How-
ever, owing to different depths of proteome
coverage for these two studies, we also ana-
lyzedKIFC1expression using data from pre-
vious RNA sequencing (RNA-seq) studies of
mammalian oocytes and embryos ( 97 – 102 ).
Mouse, bovine, and porcine oocytes had a prom-
inent pool of maternalKIFC1messenger RNA
(mRNA), which was depleted upon fertilization,
and embryonicKIFC1mRNA was expressed from
the two- to four-cell stage onward (fig. S12A).
By contrast,KIFC1mRNA was barely detect-
able in human oocytes and zygotes but was
readily expressed from the two- or four-cell
stage onward in embryos (fig. S12, A and B).
Such discrepancies in gene expression between
different mammalian species were not observed
forNUMAand for the conserved zona pellucida
proteinZP3(fig. S12C).
We subsequently examined KIFC1 protein
levels in oocytes and in asynchronized HeLa
cells as a positive control for immunoblot
(Fig. 4, B to D). To ensure comparable load-
ing of oocyte lysate from different species, we
performed on-blot total protein normaliza-
tion, which outperformed the sensitivity and
linearity of all canonical housekeeping pro-
teins (Fig. 4B and fig. S13, A to G). Although we
could readily detect KIFC1 in HeLa cell, mouse
oocyte, bovine oocyte, and porcine oocyte
lysates (Fig. 4C and fig. S13H), we could not
detect KIFC1 in comparable amounts of hu-
man oocyte lysate (Fig. 4C). Only after over-
exposing the entire blot could we detect a
very faint smear in the human oocyte lane
(Fig. 4C). Thus, human oocytes are deficient
in KIFC1.Depletion of KIFC1 in aMTOC-free mouse
oocytes and in bovine oocytes recapitulates the
spindle instability of human oocytes
To mimic the deficiency of KIFC1 in human
oocytes, we depleted KIFC1 in aMTOC-free
mouse oocytes and in bovine oocytes using
follicle RNAi and Trim-Away, respectively.
Around 35% of KIFC1-depleted aMTOC-free
mouse oocytes assembled multipolar spindles
(Fig. 5, A to C), as confirmed by the significant
increase in the number of NUMA clusters per
oocyte (Fig. 5D). Moreover, another 30% of
KIFC1-depleted aMTOC-free mouse oocytes
assembled round spindles with broad poles
(Fig. 5, B and C). These spindles closely re-
sembled the“apolar”spindles that have been
observed in live human oocytes ( 2 ). Quantifi-
cation of the total volume of NUMA clusters
revealed no significant difference between
control and KIFC1-depleted oocytes (Fig. 5E),
suggesting that these pole defects are caused
by the failure of NUMA clusters to coalesce
rather than by the de novo assembly of ad-
ditional NUMA clusters. Similarly, around
60 and 25% of KIFC1-depleted bovine oocytesSoet al.,Science 375 , eabj3944 (2022) 11 February 2022 5 of 19
RESEARCH | RESEARCH ARTICLE