Abstract
Although pluripotent stem cell (PSC) properties, such as
differentiation and infinite proliferation, have been well documented
within the frameworks of transcription factor networks, epigenomes, and
signal transduction, they remain unclear and fragmented. Directing
attention toward translational regulation as a bridge between these
events can yield additional insights into previously unexplained
mechanisms. Our functional CRISPR interference screen–based approach
revealed that EIF3D, a translation initiation factor, is crucial for
maintaining primed pluripotency. Loss of EIF3D disrupted the balance of
pluripotency-associated signaling pathways, thereby compromising primed
pluripotency. Moreover, EIF3D ensured robust proliferation by
controlling the translation of various p53 regulators, which maintain
low p53 activity in the undifferentiated state. In this way,
EIF3D-mediated translation contributes to tuning the homeostasis of the
primed pluripotency networks, ensuring the maintenance of an
undifferentiated state with high proliferative potential. This study
provides further insights into the translation network in maintaining
pluripotency.
__________________________________________________________________
Translation controls the homeostasis of key signaling pathways that
define primed pluripotent stem cell identity.
INTRODUCTION
Pluripotent stem cells (PSCs) have the capacity for self-renewal under
appropriate conditions while maintaining their distinct attributes,
including differentiation potential and unlimited proliferation
([38]1–[39]5). Pluripotency is categorized into two types: naïve,
resembling the preimplantation inner cell mass, and primed, akin to the
postimplantation epiblast ([40]6). These categories differ in their
specific requirements for self-renewal, differentiation capacity, and
epigenetic status. Previous studies have shown that inhibiting multiple
kinases helps sustain naïve pluripotency in both rodents and humans,
suggesting a conserved, kinase-independent strategy across species
([41]7–[42]10).
Conversely, shifting from kinase inhibition to specific growth factor
stimuli enables naïve PSCs to transition into primed pluripotency, a
state poised for differentiation into various somatic lineages ([43]7,
[44]11). Unlike the naïve state, the fate of primed pluripotency
depends on a range of signaling inputs, including fibroblast growth
factor (FGF), insulin-like growth factor, and transforming growth
factor–β (TGFβ) ([45]12–[46]14). Thus, kinase signaling dynamics are
crucial for the transition between these states and their ongoing
maintenance. Paradoxically, the same growth factors that support primed
pluripotency also initiate lineage-specific differentiation programs
([47]15, [48]16). Maintaining a delicate balance between strong and
weak kinase signaling is key to preserving the equilibrium between
self-renewal and differentiation induction ([49]17, [50]18). Although
primed pluripotency, maintained by finely tuned signaling, is an
important research area in stem cell and developmental biology, the
complex mechanisms governing this balance remain elusive.
Translation, the process by which RNA is converted to protein, is
critical for cellular homeostasis and the study of primed pluripotency.
In this context, it is notable that differentiation cues actively
enhance protein synthesis ([51]19). This shift in translation dynamics
highlights the importance of translational control in dictating
pluripotency and differentiation. Gene analysis showing discrepancies
between mRNA and protein levels has unveiled the critical role of
context-dependent posttranscriptional regulation in maintaining primed
pluripotency ([52]20). This suggests that translational modulation
substantially influences the primed pluripotent state, independent of
transcriptional regulation. Several translational factors associated
with primed pluripotency have been identified ([53]21–[54]24). For
example, one of them, eukaryotic translation initiation factor gamma 2
(EIF4G2) (also known as NAT1 and DAP5) has been reported to interact
with the eIF3 complex and regulate the translation of specific mRNAs in
a noncanonical manner; however, a comprehensive understanding of the
role of translational regulation in stem cell homeostasis and fate
determination remains elusive ([55]25, [56]26).
Here, we show that translational regulation by EIF3 subunit D (EIF3D)
is critical for maintaining the identity of human primed PSCs. Loss of
EIF3D resulted in impaired self-renewal of primed PSCs and their
differentiation potential into three germ layer lineages. As a
mechanism to explain these phenotypes, we revealed that EIF3D regulates
the activities of key signaling pathways for the self-renewal of primed
PSCs. In addition, EIF3D also controls the translation of a set of p53
regulators that render p53 activity low, resulting in robust
proliferation of undifferentiated PSCs. Collectively, we conclude that
the homeostasis of primed pluripotency is maintained through
translational regulation by EIF3D.
RESULTS
EIF3D is essential for the maintenance of human primed pluripotency
To identify the genes essential for maintaining primed pluripotency, we
conducted genome-wide functional screening using the CRISPR
interference (CRISPRi) platform ([57]27, [58]28). A single guide RNA
(sgRNA) library targeting genome-wide mRNAs was delivered into a human
PSC line carrying a doxycycline (Dox)–inducible KRAB-dCas9 transgene
([59]28). Self-renewing cells positive for stage-specific embryonic
antigen 5 (SSEA-5) ([60]29), a surface marker specifically expressed on
undifferentiated primed PSCs, were purified before (day 0) and 8 and 16
days after the induction of knockdown (KD).
The essential genes for the self-renewal of primed PSCs were identified
as targets of significantly depleted sgRNAs in purified SSEA-5–positive
cells. MAGeCK, the method for model-based normalization and robust
ranking aggregation, identified 1255 genes on day 8 and 1686 genes on
day 16 that positively influence the self-renewal of primed human PSCs
(data S1 and S2) ([61]30, [62]31). A comparison with the previously
reported CRISPR screens in human primed PSCs confirmed the reliability
of our screening results (fig. S1, A and B) ([63]32–[64]34). Gene
ontology (GO) analysis of these genes on day 16, critical for
maintaining primed PSC identity, highlighted translation-related terms
as significantly enriched (fig. S1C).
We focused on the eIF3 family, which interacts with EIF4G2, a
translational regulator that is critical for maintaining primed
pluripotency ([65]22). Our CRISPRi screen revealed that EIF4G2 was
essential for the self-renewal of primed human PSCs on days 8 and 16
(fig. S1D and data S1 and S2). The screening results suggested that
EIF3D, a cap-binding protein ([66]35), had a top-ranked notable role in
both early (day 8) and late (day 16) phases of KD among the eIF3 family
genes ([67]Fig. 1A and fig. S1E). Therefore, we focused on the role of
EIF3D in primed pluripotency. EIF3D is abundantly expressed in
undifferentiated induced PSCs (iPSCs) compared with that in human
dermal fibroblasts (HDFs), and its levels markedly decreased upon
differentiation ([68]Fig. 1B). This expression pattern is similar to
pluripotency factors, such as Octamer-binding transcription factor 3/4
(OCT3/4), SRY-box transcription factor 2 (SOX2), and Nanog homeobox
(NANOG), suggesting the potential role of EIF3D in primed PSCs.
Fig. 1. EIF3D is essential for maintaining primed pluripotency.
[69]Fig. 1.
[70]Open in a new tab
(A) Rank plot from CRISPRi screening. Red and blue dots represent genes
significantly increased or decreased (±1 SD) following 8 and 16 days of
KD, respectively. n = 3. The full list is provided in data S1 and S2.
(B) Protein expression in undifferentiated and differentiating PSCs (10
and 20 days post-FGF withdrawal) and HDFs. Vinculin (VCL) was used as a
loading control. (C) Representative images of control and EIF3D KD
iPSCs, 5 days post-KD induction. Scale bars, 100 μm. (D) Cell counts on
days 3 (P = 4.70 × 10^−4) and 5 (P = 3.53 × 10^−7) post-KD induction
(mean ± SD, n = 6). P values determined via unpaired t test. (E) Cell
cycle phase distribution (mean ± SD, n = 3). G[1]: P = 2.14 × 10^−3; S:
P = 4.43 × 10^−7; G[2]/M: P = 1.99 × 10^−4, calculated by unpaired t
test. (F) RNA expression of pluripotency and p53-related genes during
EIF3D KD. n = 3. (G) Expression of pluripotency and p53-associated
proteins during EIF3D KD. d, days; FC, fold change.
To explore its function, we generated Dox-inducible EIF3D KD iPSC lines
using a CRISPRi platform (fig. S2A). Control cell lines expressing a
sgRNA with nontarget spacer sequence were used for comparison in all
subsequent analyses. Following EIF3D KD, we observed morphological
changes characterized by flattened colonies with indistinct edges and
notably enlarged nuclei, evoking differentiation ([71]Fig. 1C and fig.
S2, B and C). In addition, EIF3D KD resulted in a marked reduction in
cell numbers between days 3 and 5 postinduction ([72]Fig. 1D).
Investigating this phenotype, we measured DNA synthesis via
5-ethynyl-2′-deoxyuridine (EdU) incorporation, revealing a significant
decrease in the S phase cell population and a corresponding increase in
cells in the G[1] and G[2]/M phases following EIF3D KD ([73]Fig. 1E and
fig. S2, D to F). These results collectively suggest that EIF3D KD
imposes growth arrest on primed human PSCs.
Moreover, EIF3D KD reduced the expression of transcripts encoding core
pluripotency transcription factors, which are indicative markers of
PSCs ([74]Fig. 1F). The decrease in NANOG expression was more rapid and
pronounced than that of POU5F1 (encoding OCT3/4) and SOX2. Protein
expression analyses exhibited a similar trend ([75]Fig. 1G and fig.
S2G). Notably, transcriptional targets of p53, such as CDKN1A and MDM2
mRNAs, and their translation products were significantly up-regulated
in EIF3D KD cells, akin to the changes seen in differentiated iPSCs
induced by FGF withdrawal ([76]Fig. 1, B, F, and G). Other hallmarks of
senescence (INK4A and ARF) and cell death (FAS) markers were also
elevated ([77]Fig. 1F). Similar to iPSC differentiation, the absence of
EIF3D led to increased p53 protein expression, while TP53 mRNA showed
only modest changes ([78]Fig. 1, B, F, and G, and fig. S2G). The p53
protein up-regulated by EIF3D KD increased in its size over time. This
observation can be explained by the fact that the short form lacking
the N terminus (known as dN40) is expressed in undifferentiated PSCs,
whereas the p53 protein induced by EIF3D KD is the full-length form
(fig. S2H) ([79]36). In correlation with the increase in p53 protein,
the levels of p21 and MDM2 proteins increased in EIF3D KD PSCs
([80]Fig. 1, B and G, and fig. S2G). In addition, we detected
remarkable elevation of the phosphorylated Ser^51 of EIF2α on day 5 of
EIF3D KD, suggesting the increase of cellular stress ([81]Fig. 1G).
Overall, the KD phenotypes suggested that EIF3D loss diminishes the
primed PSC identity along with several intriguing changes.
On the basis of the EIF3D KD phenotype in primed PSCs, we investigated
the effects of EIF3D KD on cell morphology and proliferation in other
cell types. Short hairpin RNA (shRNA) against EIF3D efficiently
suppressed EIF3D expression in 293T/17 cells, HepG2 cells, and HDFs
(fig. S3A). EIF3D KD did not alter the morphology of these three cell
types (fig. S3B). One of the two shRNAs against EIF3D caused a
significant decrease in the proliferation of 293T/17 cells. At the same
time, there was no significant change in the proliferative capacity of
HepG2 cells or HDFs (fig. S3C). These data suggest that the effects of
EIF3D KD on cell morphology and proliferation are limited in the cell
types tested, and the phenotype of primed PSC is remarkable.
To confirm that the phenotypes observed in the primed PSCs were due to
EIF3D KD, we introduced EIF3D and its mutants into the EIF3D KD iPSCs
to perform the rescue experiments. Although the mCherry transgene
introduced as a control could not prevent the phenotypes observed in
the EIF3D KD iPSCs, the EIF3D transgene prevented the emergence of the
phenotypes, including morphological changes, reduction of PSC marker
expression, activation of the p53 pathway, and proliferation defect
(fig. S4, A to D). Furthermore, either phosphomimetic (S528D/S529D) or
phosphoinhibitory (S528N/S529N) mutation ([82]37) in EIF3D, known to
inhibit or promote Cap binding, respectively, did not affect the
prevention of the EIF3D KD phenotypes by EIF3D transgene (fig. S4, A to
D). These data demonstrate that the EIF3D KD-induced disruption of
self-renewal in human primed PSCs shown above is independent of the
phosphorylation status of EIF3D.
Given these results, we explored the potential roles of increased p53
levels and reduced NANOG expression in the phenotypes arising from
EIF3D KD. Introducing exogenous NANOG did not restore pluripotency
marker expression or resolve the proliferative impairment (fig. S5, A
to D). In addition, the concurrent KD of TP53 with EIF3D did not fully
restore pluripotency marker expression (fig. S5, E to H). However,
these experiments showed partial recovery of cellular proliferation and
a decrease in elevated p21 expression (fig. S5, G and H). These
findings suggest that analyzing hallmark genes associated with
undifferentiated or differentiated states alone is insufficient to
fully understand the role of EIF3D in pluripotency. Nonetheless, the
data indicate the involvement of the p53-p21 pathway in regulating the
proliferation of primed PSCs, although other pathways likely contribute
to EIF3D-mediated self-renewal. In summary, the collective findings
highlight the multifaceted role of EIF3D in maintaining primed
pluripotency through complex molecular interactions.
Loss of EIF3D diminishes primed pluripotency with limited impact on three
germ layer specifications
We then conducted a comprehensive genome-wide transcriptome analysis to
further investigate the underlying mechanisms from a broader
perspective. To elucidate the cell fate changes in primed PSCs induced
by EIF3D KD, we compared the global gene expression profiles of EIF3D
KD cells with those of iPSCs differentiated through suppression of core
transcription factors (fig. S6, A to C), and iPSC-derived cells
directed toward endoderm (EN), mesoderm (ME), and neuroectoderm (NE)
lineages (fig. S6D).
EIF3D KD resulted in an incremental increase in differentially
expressed genes (DEGs) over successive days compared with that in the
controls ([83]Fig. 2A and fig. S6E). Notably, despite notable changes
in gene expression, the EIF3D KD profile showed less similarity to the
corresponding comparatives such as trilineage differentiated cells and
core transcription factor KD cells ([84]Fig. 2A and fig. S6F). Instead,
it seemed to enter an independent state, marked by a lack of clear
lineage commitment to any of the three germ layers. This was
accompanied by down-regulating key pluripotency and primed PSC markers,
including ZIC2, CD24, and SFRP2 ([85]Fig. 2, A and B, and table S1)
([86]32, [87]38–[88]42). By contrast, the increase or decrease in
transcription factors and housekeeping genes ([89]43–[90]45) due to
EIF3D KD was limited (fig. S6G). GO analysis revealed that EIF3D KD
up-regulated genes associated with inconsistent differentiation terms,
while genes related to cell cycle and division were down-regulated
(fig. S6H). These results are consistent with the observed EIF3D KD
phenotypes, including growth retardation and loss of pluripotency, with
minimal contribution to specific lineages. The evidence indicates that
EIF3D maintains primed pluripotency through mechanisms distinct from
those of core transcription factors or the inhibition of specific
lineage commitments.
Fig. 2. Loss of EIF3D impairs trilineage specifications.
[91]Fig. 2.
[92]Open in a new tab
(A) Principal components (PC) analysis of RNA-seq data. Each dot
represents the average value of replicates. n = 3. (B) Volcano plots
displaying DEGs between control and EIF3D KD iPSCs, 3 and 5 days after
Dox addition. Understated-colored dots represent DEGs.
Highlighted-colored dots denote key pluripotency markers. Red and blue
dots indicate genes significantly up-regulated and down-regulated,
respectively [|log[2] fold change (FC)| > 1, adjusted P < 0.05]. n = 3.
See full gene list and FC in table S1. See also fig. S6. (C) The scheme
of trilineage differentiation. Three days before differentiation
induction, Dox was added to the control and sgEIF3D cell lines. On day
3, the cells were exposed to the specific culture conditions of each
lineage and simultaneously maintained under the PSC condition for
comparison. Dox continued to be added until day 8. (D) Phase contrast
and immunocytochemistry images of control (top) and sgEIF3D (bottom)
cells after differentiation, showing specified proteins (red). Nuclei
visualized with Hoechst 33342 (blue). Scale bars, 100 μm. (E) RNA
expression of pluripotency and lineage marker genes after
differentiation (mean ± SD, n = 9). *P < 0.0001 for control versus
sgEIF3D in the same condition determined by one-way analysis of
variance (ANOVA).
Loss of EIF3D impairs trilineage differentiation
The impairment of self-renewal by EIF3D KD, including the
down-regulation of pluripotency genes, suggests that EIF3D contributes
to the capacity of primed PSCs. Therefore, we next tested whether EIF3D
KD primed PSCs could differentiate into three germ layers. After 3 days
of KD induction, a time when EIF3D proteins have almost disappeared,
the cells were exposed to culture conditions for differentiation into
each differentiated lineage ([93]Fig. 2C). Following a 5-day period of
differentiation, the expression of distinct lineage markers, including
SOX17 and FOXA2 for EN, HAND1 and FOXF1 for ME, and PAX6 and SOX1 for
NE, was assessed. It was shown that the control cells exhibited
successful differentiation, while the EIF3D KD cells displayed limited
expression of these specific markers, indicating a failure to
differentiate ([94]Fig. 2, D and E). The expression of the pluripotency
marker NANOG was completely silenced in well-differentiated control
cells, whereas EIF3D KD cells retained its partial expression ([95]Fig.
2E). EIF3D KD cells not only lacked lineage marker activation but also
exhibited a morphology characterized by enlarged nuclei in all culture
conditions, including the PSC condition ([96]Fig. 2D). In addition,
cells could survive even after a total of 8 days of EIF3D KD regardless
of culture conditions. These results indicate that primed PSCs lacking
EIF3D not only fail to maintain self-renewal but also exhibit
deficiencies in their ability to differentiate into the trilineage, a
hallmark of pluripotency.
Loss of EIF3D decreases naïve pluripotency slower than in primed PSC
To further investigate the effects of EIF3D KD in primed PSCs, beyond
the typical three germ layers derived from these cells, our study
expanded to naïve PSCs which resemble an earlier developmental fate.
Therefore, we examined the differentiation of naïve PSCs into a primed
state to further evaluate the role of EIF3D in maintaining primed
pluripotency ([97]Fig. 3A). Before the transition to the primed state,
we induced EIF3D KD in naïve PSCs for 3 days, which did not result in
any observable abnormalities ([98]Fig. 3B). However, when we changed
the culture conditions from the kinase-inhibiting naïve state to the
growth factor-rich primed state, the EIF3D KD naïve PSCs showed an
inability to differentiate into the primed state. This was marked by
massive cell death within 4 days ([99]Fig. 3B). Given this evidence of
the inability to self-renew primed PSCs following EIF3D KD, we conclude
that EIF3D is essential for the maintenance of primed pluripotency.
Fig. 3. The comparison of EIF3D’s role in primed and naïve PSCs.
[100]Fig. 3.
[101]Open in a new tab
(A) Scheme of differentiation of naïve PSCs into primed state. (B)
Differentiation of naïve PSCs to primed state. Induction of
differentiation from naïve to primed PSCs, 3 days post-Dox addition, by
altering culture conditions. Representative images of specified cell
lines and days are shown. Scale bars, 100 μm. (C) Scheme of the
comparison between naïve and primed PSCs with EIF3D KD. (D)
Representative images of control and sgEIF3D naïve PSCs, 5 days
post-Dox addition. Scale bars, 100 μm. (E) Protein expression in
control and EIF3D KD naïve (N) and primed (P) PSCs, 5 days post-Dox
addition. (F) Hierarchical clustering of the averaged values of RNA-seq
data. (G) Scatterplots of log[2]FC (control and EIF3D KD) from the Wald
tests of primed 5 and 3 days (left) and primed 5 days and naïve 5 days
(right) post-Dox addition. Colors show up (red) and down (blue) from
the result of primed control and EIF3D KD 5 days after Dox addition.
Pearson’s correlation values and P values of t test were shown on top.
(H) The DEG numbers between control versus EIF3D KD. (I) RNA expression
of EIF3D and pluripotency marker genes after 7 days of Dox addition
(mean ± SD, n = 3). *P < 0.05 determined by unpaired t test.
To assess the EIF3D KD phenotype in naïve PSCs, we compared primed and
naïve PSCs after 5 days of EIF3D KD ([102]Fig. 3C). EIF3D KD attenuated
naïve pluripotent signatures, as evidenced by decreased expression of a
naïve PSC marker KLF17, although without noticeable morphological
changes in naïve PSC colonies ([103]Fig. 3, D and E). In addition, the
comparison with the samples shown in [104]Fig. 2A such as control and
EIF3D KD primed PSCs, trilineage differentiated cells, and core
transcription factor KD cells revealed that the impact of EIF3D KD on
gene expression changes in naïve PSCs was less pronounced than those in
the primed state ([105]Fig. 3F). The variance of fold changes between
control versus EIF3D KD and the number of DEGs also supported the
significant alterations of gene expression by EIF3D KD day 5 post-KD
induction in primed PSCs rather than naïve state ([106]Fig. 3, G and
H). However, when the culture of EIF3D KD naïve PSCs reached day 7, it
could not be maintained due to a notable decrease in cell number. The
quantitative reverse transcription polymerase chain reaction (qRT-PCR)
results showed that the expression of PSC markers was reduced by EIF3D
KD ([107]Fig. 3I). Although the phenotype emergence of naïve PSCs was
slower than that of primed PSCs, EIF3D was required for self-renewal of
naïve PSCs. While EIF3D plays critical roles in naïve and primed
pluripotency, we further focused on its role in primed PSCs.
EIF3D orchestrates translation of key signaling pathways in primed
pluripotency
On the basis of global transcriptome data, we observed a lesser degree
of change induced by EIF3D KD on day 3 compared to day 5 ([108]Fig. 2,
A and B, and fig. S6, E and F). Therefore, to understand the initial
response to EIF3D loss, we analyzed translation status on day 3 post-KD
induction. Puromycin incorporation showed that EIF3D KD reduced de novo
protein synthesis to 45% relative to control cells ([109]Fig. 4A and
fig. S7). Polysome profile analysis indicated a significant
accumulation of the 80S ribosomal subunit with EIF3D KD, suggesting
decreased translation initiation ([110]Fig. 4, B and C). These results
indicated that the absence of EIF3D in primed PSCs led to noticeable
alterations in translation.
Fig. 4. EIF3D-mediated translation regulates multiple signaling pathways.
[111]Fig. 4.
[112]Open in a new tab
(A) Quantification of de novo protein synthesis by detecting
incorporated puromycin. n = 3. P = 6.75 × 10^−3 determined by unpaired
t test. a.u., arbitrary units. (B) Representative polysome profiles of
sgEIF3D primed PSCs compared to control lines on day 3 post-Dox
addition. (C) Area under the curve (AUC) quantification for specified
ribosomal fractions (mean ± SD, n = 3). Ratios 60S/40S: P = 0.011;
80S/40S: P = 0.012; polysomes (PS)/40S: P = 0.094, calculated using
unpaired t test. (D) Categorization of ORFs with varying translation
efficiency (TE) during EIF3D KD. Displayed are all ORFs translated in
human iPSCs (all) and those down-regulated (down) or up-regulated (up)
by EIF3D KD over 3 days. (E) Volcano plot showing up-regulated DTEGs
(uDTEG; red) and down-regulated DTEGs (dDTEGs; blue) (|log[2]FC| > 1,
adjusted P < 0.05). See full gene list, FC, and adjusted P value in
data S3 and S4. (F) Pathway enrichment analysis of dDTEGs using
WikiPathways. (G) Volcano plots displaying log[2] FC of transcripts in
specified pathways according to WikiPathways (WP437, WP399, WP481,
WP382, and WP366 for EGF, WNT, insulin, MAPK, and TGFβ pathways,
respectively). Understated-colored dots represent DTEGs.
Highlighted-colored dots denote transcripts in specified pathways. Red
and blue dots indicate genes whose TEs are significantly up-regulated
and down-regulated, respectively (|log[2] FC| > 1, adjusted P < 0.05).
n = 3. See full gene list, FC, and adjusted P value in data S5 to S9.
(H) Phosphorylation status of key proteins in the signaling pathways
across the timeline of EIF3D KD.
To determine the translational changes due to EIF3D KD, we performed
ribosome profiling (fig. S8, A to C). To consider translation in
noncanonical manner, we identified 31,263 open reading frames (ORFs)
undergoing translation in human primed iPSCs, classified as follows:
45% as annotated coding sequences (CDS), 42% as variant CDS, 2% as
unidentified ORFs, and 8% as upstream ORFs (uORFs) ([113]Fig. 4D).
EIF3D KD increased translation efficiency (TE) in 284 ORFs and
decreased it in 1340 ORFs. The increased TE group featured a higher
percentage of uORFs (35.07%), whereas the decreased TE group showed no
significant preference in ORF classification. uORF is a cis-regulatory
element that regulates TE of downstream ORF ([114]25). Therefore, we
investigated the changes in TE of downstream ORFs of the up- and
down-regulated uORFs and the nonsignificant uORFs, following EIF3D KD.
The TE of ORFs located downstream of up-regulated uORFs showed minimal
differences compared to that of ORFs located downstream of
down-regulated uORFs (fig. S8D), indicating that EIF3D did not
contribute to cis-regulatory changes between ORFs but rather influenced
the TE of all ORFs across the transcripts. These results suggest that
EIF3D is involved in the regulation of ribosome binding to uORFs in
primed PSCs and also plays a role in the regulation of TE across the
entire transcript, including uORFs and annotated ORFs such as CDS.
We subsequently examined the transcript-wise translation status
following EIF3D KD. Ribosome profiling revealed that EIF3D KD
significantly altered the TE of 1321 genes (increased in 402; decreased
in 919) ([115]Fig. 4E and data S3 and S4), which we term “differential
translation efficiency genes” (DTEGs). Compared with a previous study
that performed photoactivatable ribonucleoside-enhanced cross-linking
and immunoprecipitation and identified JUN as a target of EIF3A, EIF3B,
EIF3D, and EIF3G in 293T cells ([116]35), in this study, there are a
small number of overlapped genes with the ribosome profiling data of
EIF3D KD primed PSCs (fig. S8E). In addition, we confirmed that TE of
JUN was not significantly changed in EIF3D KD primed PSCs (fig. S8F),
suggesting a cellular context-dependent manner of translational
regulation by EIF3D and supporting the importance of the cap-binding
independent role of EIF3D in primed PSCs (Fig. S4). Pathway analysis
indicated that down-regulated DTEGs (dDTEGs) were linked to several
signaling pathways, including epidermal growth factor (EGF), WNT,
insulin, mitogen-activated protein kinase (MAPK), and TGFβ, all of
which are critical for the maintenance of primed pluripotency
([117]Fig. 4F) ([118]46). In addition, dDTEGs were also associated with
several terms, including “pluripotency.” By contrast, up-regulated
DTEGs (uDTEGs) did not show significantly enriched terms. Given the
EIF3D KD phenotype, which includes the loss of primed pluripotency, it
is plausible that these signaling pathways involving dDTEGs are
implicated.
On the basis of the pathway analysis results, we confirmed the TEs of
EIF3D targets in each enriched pathway. The volcano plots revealed
significant TE alterations in the transcripts of the EGF, WNT, insulin,
MAPK, and TGFβ pathways ([119]Fig. 4G, fig. S8G, and data S5 to S9)
([120]47). Next, we examined the phosphorylation status of key proteins
in these signaling pathways potentially regulated by EIF3D. EIF3D KD
led to the hyperactivation of the MAPK, insulin (AKT, mTOR, and
p70S6K), and WNT pathways while simultaneously suppressing the TGFβ
pathway through SMAD2 ([121]Fig. 4H). These results confirm that EIF3D
KD disrupts the balance of multiple signaling activities in primed PSCs
through translational regulation.
EIF3D inhibits p53 protein expression by modulating translation of p53
regulators
Besides the dysregulation of multiple kinase pathways, there is a
notable increase in p53 protein and subsequent activation of the p53
pathway due to EIF3D KD ([122]Fig. 1G and fig. S2G). However, ribosome
profiling data revealed no significant change in p53 TE following EIF3D
KD (likelihood ratio test, fold change = 1.01, adjusted P = 0.92). This
led us to hypothesize about an indirect regulatory mechanism. To deepen
our understanding, we performed a comparative analysis between DTEGs
and a compilation of posttranscriptional regulators of p53 protein
expression ([123]48). Setting a TE threshold of 1.5 fold change, EIF3D
KD resulted in significant translation dysregulation of 207 of 818
genes, representing 25.3% of the list ([124]Fig. 5A and data S10). In
addition, we found 54 genes under the condition of a TE threshold of
twofold change between control and EIF3D KD ([125]Fig. 5A and data
S10). The protein analysis confirmed the decreased expression of p53
regulator proteins in EIF3D KD primed PSCs ([126]Fig. 5B), suggesting
that EIF3D indirectly affects p53 protein expression by regulating the
translation of its regulators.
Fig. 5. Indirect regulation of p53 protein through EIF3D-mediated
translational regulation.
[127]Fig. 5.
[128]Open in a new tab
(A) Venn diagrams display the overlap between DTEGs with significant TE
(|log[2] FC| > 0.58) and p53 regulators. Genes with higher TE (|log[2]
FC| > 1) are highlighted in red. See full gene list, FC, and adjusted P
value in data S10. (B) Protein expression of p53 regulators
translationally controlled by EIF3D. High TE (|log[2] FC| > 1) includes
RBBP6, SSU72, and TCP1; moderate TE (|log[2] FC| > 0.58) includes
KDM5C, WDR5, WDR82, and REST. (C) Luciferase reporter assays showing
the effects of RBBP6 (P = 2.89 × 10^−9) and GAPDH (P = 0.85) 5′UTRs on
translation in control and EIF3D KD primed PSCs (mean ± SD, n = 6). P
values determined via unpaired t test. (D) Representative images of
specified cells, 5 days post-KD induction. Scale bars, 100 μm. (E) Cell
counts of the cells depicted in [129]Fig. 5D (mean ± SD, n = 6). (F)
Relative gene expression in cells from [130]Fig. 5D. Values normalized
to GAPDH and compared to control without Dox. n = 3. (G) Expression of
specified proteins in cells from [131]Fig. 5D.
For example, we identified RB binding protein 6 (RBBP6), a negative
regulator of p53 protein stability in the dDTEG list (data S3)
([132]49). Luciferase reporter assays revealed that the 5′ untranslated
region (5′UTR) of RBBP6 was responsible for EIF3D-mediated
translational regulation in primed PSCs ([133]Fig. 5C). By contrast,
the absence of EIF3D did not affect the translation of the GADPH
5′UTR-containing reporter ([134]Fig. 5C). These data together with the
ribosome profiling results indicate that RBBP6 is a target of EIF3D in
primed PSCs. As expected, RBBP6 KD mimicked EIF3D KD phenotypes,
including cell growth defects with morphological changes, decreased
expression of pluripotency markers, increased p53 protein, and elevated
expression of CDKN1A/p21 and MDM2 ([135]Fig. 5, D to G). These findings
indicate that EIF3D plays a role in modulating p53 protein expression
by controlling the translation of its regulators.
Artificial dysregulation of signaling balance recapitulates loss of
pluripotency
Last, we investigated whether deliberately disrupting the balance of
pluripotency-associated signaling pathways and activating the p53
pathway could recapitulate the loss of primed pluripotency. We used
chemical compounds and a growth factor to induce changes observed in
primed EIF3D KD PSCs, including the activation of EGF/MAPK, mTOR, AKT,
WNT, and p53, and the inhibition of TGFβ ([136]Fig. 6A). Treatment with
five factors (5F consisted of EGF, AKT activator, mTOR activator, WNT
activator, and TGFβ inhibitor) decreased cell number without obvious
morphological changes compared to the dimethyl sulfoxide (DMSO)–treated
control cells ([137]Fig. 6, B and C). Treatment solely with Nutlin-3a
(N3a), a p53 activator, resulted in marked morphological changes with
clear cell borders and irregular colony edges, yet it did not affect
the expression of representative PSC markers ([138]Fig. 6, B and D).
Notably, PSCs treated with N3a and EIF3D KD for 3 days exhibited
clustering ([139]Fig. 6D). Furthermore, the conditions containing N3a
(6F, N3a, N3a + 5F, and N3a + 6F) induced massive cell death, with a
much greater reduction in cell numbers than 5F ([140]Fig. 6, B and C).
The p53 overexpression and MDM2 KD also resulted in the death of
undifferentiated primed PSCs, highlighting the critical role of p53
suppression in primed PSCs for survival and proliferation, rather than
maintaining undifferentiation (fig. S9).
Fig. 6. Disruption of primed pluripotency through manipulation of EIF3D’s
target pathways.
[141]Fig. 6.
[142]Open in a new tab
(A) Overview of chemical compound and growth factor treatments. Control
cells were treated with DMSO. (B) Representative images of iPSCs
treated with specified factors. Scale bars, 100 μm. (C) Cell counts on
day 5 (mean ± SD, n = 9). *P < 0.0001 compared to the control;
†P < 0.05 versus 5F determined by one-way ANOVA. (D) Sample clustering
from (A), alongside control and sgEIF3D PSC core transcription factor
(TF) KD iPSCs, as well as trilineage differentiated cells, highlighting
selected PSC marker genes. n = 3. (E) Schematic representation of the
model describing how EIF3D-mediated translational regulation balances
homeostasis of critical signaling pathways in primed pluripotency.
The dysregulation of multiple signaling pathways by the 5F and 6F
treatment for 5 days substantially decreased the expression of
pluripotency-associated genes, and they were clustered with PSCs after
5 days of EIF3D KD ([143]Fig. 6D). These results suggest that the
disruption of multiple signaling pathways induced by EIF3D KD is
responsible for the loss of primed pluripotency. Collectively, this
study demonstrates that EIF3D plays a key role in safeguarding
self-renewal and robust proliferation in primed pluripotency by
coordinating essential signaling pathways and inhibiting the p53
pathway ([144]Fig. 6E).
DISCUSSION
This study demonstrates that EIF3D is critical for maintaining primed
pluripotency through translational regulation that finely balances
kinase signaling and suppresses the p53 pathway. The absence of EIF3D
decreased global translation in primed PSCs by approximately 50%.
However, previous research indicated that reduced global translation
did not necessarily result in the loss of pluripotency, making it
unlikely that the EIF3D KD phenotype is a consequence of stress due to
a lowering translation ([145]24). On day 3 post-EIF3D KD, attenuated
pluripotency, reduced translation, and changes in the activity of the
p53 pathway were detected, whereas an increase in the phosphorylation
of EIF2α, a sign of cellular stress, was not observed until day 5. The
disparity in time indicates that the EIF3D KD phenotype in primed PSCs
is not simply due to stress. Meanwhile, since translation factors such
as EIF2B5, which are essential for self-renewal of primed PSCs, are
targets of EIF3D, loss of EIF3D may affect the primed PSC identity by
altering the translation machinery. Further research into comprehensive
translational control mechanisms provides further insights.
One of the remarkable phenotypes of EIF3D KD in primed PSCs shown in
this study is the rapid loss of proliferation. We have also shown that
the reduction in proliferative capacity caused by EIF3D KD also occurs
in 293T/17 cells as non-PSCs. Previous studies have also shown that
EIF3D promotes proliferation in some types of cancer cell lines
([146]50–[147]53). This is not unexpected given that EIF3D is expressed
in many cell types and is responsible for regulating translation
initiation. In this study, we did not analyze which signaling pathways
are regulated by EIF3D KD in non-PSCs. Further analysis will provide
more insight into the function of EIF3D on cellular proliferation. On
the other hand, we have shown that the p53 pathway is activated by
EIF3D KD in primed PSCs, leading to cell proliferation defects.
Maintaining low p53 activity is essential in undifferentiated PSCs,
although the exact regulatory mechanism remained unclear ([148]54). Our
study sheds light on the indirect inhibition of p53 by EIF3D through
targeted translation modulation of p53 regulators. Moreover, the marked
increase in p53 protein in EIF3D KD, combined with the inverse
correlation between increased p53 protein and decreased EIF3D
expression during PSC differentiation, strongly suggests the pivotal
role of EIF3D in suppressing p53. On the other hand, primed PSCs
expressed a short isoform of p53 translated from an AUG start codon
downstream of the canonical translation start site. Such a mechanism
may also be involved in the tight regulation of p53 protein expression.
The role of EIF3D, particularly its involvement in selective
translation, has been recognized in the context of oncogenic and stress
responses. On the other hand, we explicitly state that this study does
not assert the importance of selective translation by EIF3D in primed
PSCs. Recent studies have demonstrated the influence of EIF3D on the
translational regulation of the MAPK pathway in commonly used cell
lines, such as 293T and HeLa cells ([149]37, [150]55). These previous
findings not only bolster our current results but also suggest a
potential overlap of EIF3D targets in different cell types. The diverse
abnormalities resulting from EIF3D depletion in different cell types
may be due to the cellular context-dependent role of a particular
pathway as a target of EIF3D. Together, the data from this and previous
studies suggest that the regulation of the MAPK and p53 pathways, which
are critical for proliferation in many cell types, is a universal
rather than a selective role of EIF3D. The EIF3D target data in primed
PSC provided in this work may serve as a resource for future research
aimed at elucidating the selectivity of translational regulation by
EIF3D.
Recently, it has become possible to stop the proliferation of PSCs by
inhibiting mTOR ([151]56, [152]57). This state, called diapause-like,
slows proliferation while maintaining the proportion of cells in S
phase. On the other hand, the decrease in proliferation caused by EIF3D
KD in primed PSCs results in the loss of cells in S phase, which is
different from the diapause-like state. In addition, while PSCs in the
diapause-like state retained their differentiation potential, EIF3D
depletion in primed PSCs resulted in failure to differentiate into the
three germ layers. The differentiation failure of EIF3D KD PSCs may be
due to proliferation arrest rather than its critical role in lineage
commitment; thus, further studies are warranted to determine whether
EIF3D directly contributes to differentiation potential. On the other
hand, since EIF3D contributes to the regulation of protein expressions
and their phosphorylation status in multiple pathways, including mTOR
signaling, further research on the regulation of its activities (e.g.,
phosphatase) may reveal a comprehensive mechanism of proliferation
control in PSC.
Although EIF3D and EIF4G2 are known to interact ([153]26), the
phenotype of EIF3D KD shown in this study differed from the previously
reported phenotype of EIF4G2 KD in primed PSCs using the same
experimental condition ([154]22). First, the time required for EIF3D KD
to lose self-renewal was rapid (3 to 5 days), whereas that for EIF4G2
KD was relatively slow (6 to 12 days). Furthermore, while EIF4G2 was
required for differentiation into NE but not EN and ME, EIF3D KD
inhibited differentiation into all three germ layers. These results
suggest that the simple interaction of EIF3D and EIF4G2 is not
sufficient to maintain primed pluripotency. This study showed that both
phosphorylation site mutants of EIF3D, which promote or inhibit Cap
binding, can avoid the EIF3D KD phenotype in primed PSCs. Further
research is warranted to clarify the detailed mechanism; this study
provides the possible role of EIF3D in safeguarding pluripotency
independent of Cap binding.
Research on transcription factor networks and signaling pathways has
advanced our understanding of pluripotency. This study highlights the
importance of translational regulation as a link between these two
aspects in pluripotency. Our CRISPRi screen indicated that, besides
EIF3D, other translational regulators might play a role in maintaining
primed pluripotency; deeper exploration into translational control will
offer further insights into the fundamental nature of pluripotency.
MATERIALS AND METHODS
Experimental design
To investigate the importance of translational regulation in primed
PSCs, we performed functional analyses of EIF3D, which was identified
through a functional screening process. We generated each of three
independent lines of control and EIF3D KD iPSCs and used them for all
comparative assays. In certain assays, the parental 1B4 iPSC line
([155]27) was also used. All experiments were performed with a minimum
of two to three replicates with appropriate statistical analyses.
Cell lines
Human iPSC lines WTB6 and 1B4 (CRISPRi Gen 1 clone 4 derived from WTB6)
were gifts from B. R. Conklin of Gladstone Institutes ([156]27,
[157]58). HDFs derived from fetal (HDF1419) and adult (TIG-120) donors
were acquired from Cell Applications Inc. and K. Kaji, respectively.
The 293T/17 and HepG2 cells were purchased from American Type Culture
Collection and Japanese Collection of Research Bioresources,
respectively.
Cell culture
Human iPSC lines were maintained on tissue culture plates coated by
iMatrix 511 silk (Matrixome) using StemFit AK02N media (Ajinomoto), as
previously described ([158]59). For passaging, cells were washed once
with Dulbecco’s phosphate-buffered saline (D-PBS; Nacalai Tesque) and
incubated in TrypLE Express (Thermo Fisher Scientific) for 10 min at
37°C. Subsequently, cells were dissociated into single cells and washed
in Dulbecco’s modified Eagle’s medium/Ham’s F-12 (DMEM/F-12; WAKO)
containing 0.1% bovine serum albumin (BSA; WAKO). After cell counting
and centrifugation, the cells were resuspended in StemFit AK02N media
supplemented with iMatrix-511 silk (1.67 μg/ml) and 10 μM Y-27632
(Nacalai Tesque) ([159]60). G-banding tests conducted by Nihon Gene
Laboratories confirmed that all PSC lines used in this study showed no
apparent karyotypic abnormalities. HDFs, HepG2, and 293T/17 cells were
maintained in DMEM (Nacalai Tesque) with 10% fetal bovine serum (Cosmo
Bio). Routine testing confirmed the absence of mycoplasma infection.
Transposon-mediated gene transfer
We transfected 1 μg of a plasmid containing the inverted terminal
repeats of either PiggyBac (PB) or Sleeping Beauty (SB) transposons,
together with 0.5 μg of a plasmid encoding a hyperactive PB transposase
(hyPBase) or SB transposase (SB100X), into 5 × 10^5 human PSCs. This
was accomplished using the P3 Primary Cell 4D-Nucleofector X Kit S
(Lonza) and Program CA-137 on the 4D Nucleofector device (Lonza). Two
days posttransfection, the transfectants were selected with the
appropriate drug until nontransfected cells were completely eradicated.
Subsequently, single-cell–derived colonies that uniformly expressed the
transduced fluorescent protein were isolated and expanded.
CRISPR interference
To generate inducible CRISPRi iPSC lines targeting a specific gene, we
introduced a vector containing U6 promoter–driven sgRNA along with CAG
promoter–driven fluorescence protein and a drug resistance marker into
the 1B4 human iPSC line (passages 23 to 28) using the PB-mediated gene
transfer method previously described ([160]27). Starting on day 2
posttransfection, drug selection was initiated and continued until
nontransfected cells were eliminated. Subsequently, single-cell–derived
colonies uniformly expressing the fluorescent protein were isolated and
expanded. To induce KD, we administered Dox (1 μg/ml; WAKO) for the
specified duration. KD clones within 20 passages postsubcloning were
used for the study. The spacer sequences are listed in table S2.
EN differentiation
EN differentiation was conducted as previously described, with minor
modifications ([161]61, [162]62). Primed PSCs were seeded at a density
of 1 × 10^6 cells per well in iMatrix 511-coated six-well plates using
StemFit AK02N media, supplemented with 10 μM Y-27632. The following
day, cells were washed once with DMEM/F-12, and the media was replaced
with differentiation media 1 (DM1) consisted of DMEM/F-12 (Thermo
Fisher Scientific), 2% B27 supplement (Thermo Fisher Scientific), 1%
MEM nonessential amino acids (NEAA; Thermo Fisher Scientific), and 0.1
mM 2-mercaptoethanol (2-ME; Thermo Fisher Scientific), supplemented
with activin A (100 ng/ml; Nacalai Tesque), 3 μM CHIR99021 (Nacalai
Tesque), basic fibroblast growth factor (bFGF; 20 ng/ml; Nacalai
Tesque), and 50 nM PI-103 (Cayman Chemical). After 24 hours, the cells
were washed with DMEM/F-12, and the medium was replaced with DM1
supplemented with activin A (100 ng/ml) and 250 nM LDN193189
(Stemgent). Two days later, following another wash with DMEM/F-12, the
cells were cultured in DM1 with activin A (100 ng/ml) for an additional
48 hours.
ME differentiation
Directed differentiation to ME was carried out with minor modifications
from previously described methods ([163]62, [164]63). A day before
differentiation, primed PSCs were seeded at a density of 1 × 10^6 cells
per well in iMatrix 511-coated six-well plates using StemFit AK02N
media supplemented with 10 μM Y-27632. The following day, cells were
washed once with DMEM/F-12 and then cultured in DM1 medium containing
activin A (30 ng/ml), bone morphogenetic protein 4 (BMP4; 40 ng/ml;
PeproTech), 6 μM CHIR99021, bFGF (20 ng/ml), and 100 nM PIK-90
(MedChemExpress) for 24 hours. Subsequently, after a wash with
DMEM/F-12, the medium was replaced with DM1 supplemented with BMP4 (40
ng/ml), 1 μM A83-01, and 4 μM CHIR99021, and cells were maintained for
an additional 48 hours. Then, the cells were washed once more with
DMEM/F-12 and then cultured in DM1 medium supplemented with BMP4 (40
ng/ml) for another 48 hours.
Ectoderm differentiation
NE differentiation was conducted as previously described ([165]64,
[166]65). A day before induction, primed PSCs were seeded at a density
of 1 × 10^6 cells per well in iMatrix 511-coated six-well plates, using
StemFit AK02N medium supplemented with 10 μM Y-27632. The following
day, the cells were washed once with DMEM/F-12 and then cultured in
Glasgow’s MEM (WAKO) containing 8% knockout serum replacement (KSR;
Thermo Fisher Scientific), 1 mM sodium pyruvate (Sigma-Aldrich), 1% MEM
NEAA, 0.1 mM 2-ME, 1 μM A83-01, and 250 nM LDN193189. This was
maintained for 5 days, with daily media changes.
Generation and maintenance of naïve PSCs
Primed PSCs were converted to a naïve pluripotent state as previously
described ([167]11). Before conversion, we maintained primed PSCs on
γ-ray irradiated primary mouse embryonic fibroblasts (MEFs) in DFK20
media, composed of DMEM/F-12, 20% KSR, 1% NEAA, 0.1 mM 2-ME, and bFGF
(4 ng/ml). For harvesting, cells were treated with CTK solution
(ReproCELL) and dissociated into single cells. We then seeded
1.5 × 10^5 primed PSCs onto inactivated MEFs in a well of a six-well
plate using DFK20 media supplemented with 10 μM Y-27632. The cells were
incubated at 37°C in a hypoxic environment (5% O[2]). The following
day, the media was replaced with NDiff227 (Takara) supplemented with 1
μM [168]PD325901 (Stemgent), leukemia inhibitory factor (LIF; 10 ng/ml;
EMD Millipore), and 1 mM Valproic acid (WAKO). After 3 days, we
switched the media to PXGL, consisting of NDiff227 supplemented with 1
μM [169]PD325901, 2 μM XAV939 (WAKO), 2 μM Gö6983 (Sigma-Aldrich), and
LIF (10 ng/ml). Upon the emergence of round-shaped colonies, the cells
were dissociated using a 1:1 mixture of TrypLE Express and 0.5 mM EDTA
and then plated onto fresh inactivated MEF feeders in PXGL media
containing 10 μM Y-27632. We replaced the media daily and passaged the
cells every 3 to 5 days. The cells were used for assays after a minimum
of 30 days postconversion.
Differentiation of naïve PSCs to the primed state
Before differentiating naïve PSCs into a primed state, we treated the
cells, which were grown in PXGL media on iMatrix 511-coated plates,
with Dox for 3 days. Subsequently, the media was replaced with StemFit
AK02N, also supplemented with Dox, and the cells were incubated under
normoxic conditions (20% O[2]). The media was changed daily.
Treatment with chemical compounds and growth factors
Before treatment, we seeded 1B4 iPSCs at a density of 2 × 10^5 cells
per well in a six-well plate coated with iMatrix 511, using StemFit
AK02N media supplemented with 10 μM Y-27632. On the following day, we
replaced the media with StemFit AK02N containing various combinations
of 10 μM SC79 (Sigma-Aldrich), 10 μM MHY1485 (MedChemExpress), 1 μM
A83-01, 2 μM CHIR99021, EGF (50 ng/ml; WAKO), and 1 μM N3a
(Sigma-Aldrich). As a control, cells were treated with DMSO at a 1/1000
volume ratio. The treatment was conducted for either 3 or 5 days, with
daily media changes.
RNA isolation and reverse transcription polymerase chain reaction
Cells were washed once with D-PBS and lysed using QIAzol reagent
(QIAGEN). Total RNA was extracted using the Direct-zol RNA Miniprep kit
(Zymo Research), including on-column genomic DNA digestion as per the
provided instructions. For reverse transcription, 1 μg of RNA was used,
using the ReverTra Ace qPCR RT Master Mix (TOYOBO). qRT-PCR was
conducted with gene-specific primers (refer to table S2) using either
THUNDERBIRD Next SYBR qPCR Mix (TOYOBO) or TaqMan assays (Thermo Fisher
Scientific) with TaqMan Universal Master Mix II, no UNG (Thermo Fisher
Scientific) on a QuantoStudio 5 Real-Time PCR System (Applied
Biosystems). Raw threshold cycle (Ct) values were normalized against
ACTB or GAPDH expression using the delta-delta Ct method. Relative
expression was then calculated as fold change relative to the control.
Size-based protein analysis
Cells were washed once with D-PBS and lysed using
radioimmunoprecipitation assay (RIPA) buffer (Sigma-Aldrich)
supplemented with a protease inhibitor cocktail (Sigma-Aldrich). The
crude lysates were centrifuged at 15,300g for 15 min at 4°C, and the
cleared supernatant was transferred to a new tube. The concentration of
the cleared lysate was measured using a Pierce BCA Protein Assay Kit
(Thermo Fisher Scientific) and an EnVision 2104 plate reader
(PerkinElmer), following previously described methods. For quantitative
and specific detection of target proteins, we used either a Wes or Jess
automated capillary electrophoresis platform (ProteinSimple) with 12-
to 230-kDa or 60- to 440-kDa Separation Modules (ProteinSimple). We
loaded 2 μg of cell lysate per detection, along with the following
antibodies (see also table S3): mouse monoclonal anti-OCT3/4 (1:500,
Santa Cruz Biotechnology), goat polyclonal anti-SOX2 (1:40, R&D
Systems), goat polyclonal anti-NANOG (1:40, R&D Systems), mouse
monoclonal anti-p53 (DO-7) (1:200, Novus Biologicals), goat polyclonal
anti-p53 (1:100, R&D Systems), rabbit monoclonal anti-p21 (1:50, Cell
Signaling Technology), rabbit monoclonal anti-MDM2 (1:50, Cell
Signaling Technology), rabbit polyclonal anti-EIF3D (1:250,
Proteintech), rabbit polyclonal anti-eIF2α (1:50, Proteintech), rabbit
polyclonal anti-phospho-eIF2α (Ser^51) (1:20, Cell Signaling
Technology), rabbit polyclonal anti-SOX11 (1:500, Proteintech), rabbit
polyclonal anti-KLF17 (1:200, Sigma-Aldrich), rabbit monoclonal
anti-ERK1/2 (1:50, Cell Signaling Technology), rabbit monoclonal
anti–phospho-ERK1/2 (Thr^202/Tyr^204) (1:50, Cell Signaling
Technology), rabbit monoclonal anti-SMAD2 (1:50, Cell Signaling
Technology), rabbit polyclonal anti–phospho-SMAD2 (Ser^245/250/255)
(1:50, Cell Signaling Technology), rabbit monoclonal anti-mTOR (1:50,
Cell Signaling Technology), rabbit monoclonal anti–phospho-mTOR
(Ser2448) (1:50, Cell Signaling Technology), rabbit monoclonal anti-AKT
(1:50, Cell Signaling Technology), rabbit monoclonal anti–phospho-AKT
(Ser^473) (1:50, Cell Signaling Technology), rabbit polyclonal
anti-p70S6K (1:50, Cell Signaling Technology), rabbit polyclonal
anti–phospho-p70S6K (Thr^389) (1:50, Cell Signaling Technology), rabbit
polyclonal anti–phospho-p70S6K (Thr^421/Ser424) (1:50, Cell Signaling
Technology), rabbit monoclonal anti–β-catenin (1:50, Cell Signaling
Technology), rabbit monoclonal anti–phospho-β-catenin (Ser^552) (1:50,
Cell Signaling Technology), mouse monoclonal anti-puromycin (1:20,
Developmental Studies Hybridoma Bank), rabbit polyclonal anti-RBBP6
antibody (1:50, Sigma-Aldrich), rabbit polyclonal anti-TCP1 (1:200,
Proteintech), rabbit polyclonal anti-SSU72 (1:100, Proteintech), rabbit
polyclonal anti-REST (1:100, Proteintech), rabbit polyclonal anti-KDM5C
(1:250, Proteintech), rabbit polyclonal anti-WDR5 (1:100, Proteintech),
rabbit polyclonal anti-WDR82 (1:50, Proteintech), rabbit monoclonal
anti-VINCULIN (1:250, Cell Signaling Technology), and rabbit polyclonal
anti–α tubulin (1:200, Proteintech). Data visualization and analysis
were conducted using Compass for SW6.0 software (ProteinSimple).
Immunocytochemistry
The cells were washed once with D-PBS and fixed with 4%
paraformaldehyde (Nacalai Tesque) for 15 min at room temperature. They
were then blocked in D-PBS containing 1% BSA, 2% normal donkey serum
(Sigma-Aldrich), and 0.2% Triton X-100 (Teknova) for 45 min at room
temperature. Subsequently, the fixed cells were incubated overnight at
4°C with primary antibodies diluted in D-PBS containing 1% BSA.
Following this, the cells were washed with D-PBS and incubated for 45
min at room temperature in 1% BSA containing fluorescence-conjugated
secondary antibodies and Hoechst 33342 (1 μg/ml; Thermo Fisher
Scientific). After a final wash in D-PBS, fluorescence was detected
using a BZ-X810 imaging system (KEYENCE). Merged images were generated
using a BZ-X Analyzer (KEYENCE). Nuclear size was quantified by
analyzing Hoechst images with a Hybrid Cell Count Module (KEYENCE). The
antibodies and their dilutions were as follows: mouse monoclonal
anti-OCT3/4 (1:200), goat polyclonal anti-SOX2 (1:100), goat polyclonal
anti-NANOG (1:100), goat polyclonal anti-p53 (1:200), rabbit monoclonal
anti-p21 (1:400), goat polyclonal anti-SOX17 (1:100, R&D Systems), goat
polyclonal anti-HAND1 (1:50, R&D Systems), rabbit polyclonal anti-PAX6
(1:100, BioLegend), Alexa 647 Plus anti-mouse immunoglobulin G (IgG;
1:500, Thermo Fisher Scientific), Alexa 647 Plus anti-rabbit IgG
(1:500, Thermo Fisher Scientific), and Alexa 647 Plus anti-goat IgG
(1:500, Thermo Fisher Scientific).
Puromycin incorporation
After washing the cells grown in three wells twice with prewarmed
D-PBS, we added StemFit AK02N media containing cycloheximide (CHX; 100
μg/ml; Sigma-Aldrich) to one well and StemFit AK02N media alone to the
other two wells. Following a 10-min incubation at 37°C, we added 1 μM
puromycin (Thermo Fisher Scientific) to one well containing CHX-treated
cells and to one of the two nontreated wells and then continued
incubation for 30 min at 37°C. Postincubation, cells were washed with
ice-cold D-PBS and lysed using RIPA buffer, supplemented with a
protease inhibitor cocktail. Subsequently, the samples underwent
size-based protein analysis as described previously.
Cell cycle analysis
As previously described ([170]66), we conducted cell cycle analysis
using the Click-iT EdU Alexa Fluor 647 Flow Cytometry Assay Kit (Thermo
Fisher Scientific). Cells seeded at a density of 5 × 10^5 cells per
well in a six-well plate were cultured for 5 days in StemFit AK02N
media supplemented with Dox. Subsequently, the cells were incubated in
media containing 10 μM EdU for 135 min at 37°C. The cells were then
harvested and washed with 1% BSA. Following centrifugation, the cells
were resuspended in Click-iT fixative and incubated for 15 min at room
temperature. After washing the fixed cells with 1% BSA, they were
permeabilized with 1× Click-iT Perm/Wash reagent for 15 min at room
temperature. For EdU detection, we added D-PBS containing Copper
Protectant, Alexa Fluor 647 picolyl azide, and 1× Click-iT EdU buffer
additive to the cell suspension. The samples were then washed with 1×
Click-iT Perm/Wash reagent and stained with FxCycle Violet (1 μg/ml;
Thermo Fisher Scientific) for 30 min at room temperature. We analyzed
1 × 10^4 cells using a FACS (fluorescence-activated cell sorting) Aria
II (BD Biosciences) and BD FACSDiva software (BD Biosciences). EdU
(detected with Alexa 647) and DNA (detected with FxCycle Violet) were
analyzed using APC (650/660 nm) and Pacific Blue (405/455 nm) filters,
respectively. Data analysis was conducted using FlowJo software (FlowJo
LLC).
Genome-wide CRISPRi screens
Ten micrograms of the genome-wide CRISPRi library hCRISPRi-v2 (courtesy
of J. Weissman; Addgene, #83969) which contains five sgRNA for each of
18,905 mRNAs along with 3.75 μg of psPAX2 (courtesy of D. Trono;
Addgene, #12260) and 1.25 μg of pMD2.G (courtesy of D. Trono; Addgene,
#12259) were transfected into 293T/17 cells (passage 27) using
TransIT-Lenti Transfection Reagent (Mirus). Cells, plated at 5 million
per 100-mm collagen I–coated dish, were transfected the day before. Two
days posttransfection, the virus-containing supernatant was filtered
through a 0.45-μm pore size polyvinylidene difluoride filter
(Millipore), and lentiviral particles were concentrated using the
Lenti-X Concentrator (Takara) and titrated using a Lenti-X qRT-PCR
Titration Kit (Takara) as per the instructions. The lentivirus was then
infected into 1B4 iPSCs (passage 23) at a multiplicity of infection of
<0.4 (as determined by TagBFP fluorescence in the lentiviral vector) to
achieve a coverage of >1000×. Three days postinfection, cells were
selected with puromycin (1.5 μg/ml) until all noninfected cells
perished ([171]28). Subsequently, the cells were plated at 10 million
per 150-mm dish in StemFiT AK02N containing 10 μM Y-27632 and
iMatrix-511 silk. The following day, the media was replaced with
StemFiT AK02N supplemented with Dox. Cells were split every 2 to 3
days, maintaining a minimum of 100 million cells, corresponding to a
1000× coverage. Cells maintained without Dox (day 0) and those with Dox
for 8 and 16 days were harvested, and SSEA-5 (+) cells were collected
using an autoMACS Pro Separator (Miltenyi Biotec). We conducted the
screening three times independently. Genomic DNA was purified from at
least 100 million cells of each sample using NucleoSpin Blood XL
(Takara) or QIAamp DNA Blood Midi Kit (QIAGEN). The purified DNA was
digested overnight with SbfI-HF (New England Biolabs) and separated on
a 0.8% Tris-acetate-EDTA (TAE) agarose gel. Postelectrophoresis, DNA
fragments ranging from 350 to 700 bp were excised from the gel and
purified using a QIAGEN gel extraction kit (QIAGEN). PCR and library
preparation were conducted as previously described ([172]28). The
libraries were sequenced using a NextSeq 500/550 High Output v2 Kit
(Illumina) with custom primers, following the manufacturer’s protocol.
Reads were aligned to the hCRISPRi-v2 sequences, counted, analyzed
using MAGeCK (version 0.5.9.5), and then visualized using the MAGeCK
flute (version 1.12.0) package in R (version 4.1.1) ([173]30, [174]31).
GO analysis was conducted using clusterProfiler (version 4.2.2)
([175]67, [176]68).
Polysome profiling
The method used for polysome fractionation was based on a previously
described method with minor modifications ([177]69). A single
semiconfluent well of a 10-cm dish containing either control or sgEIF3D
iPSCs was placed on a CoolBox XT Workstation (Biocision) to maintain a
temperature of 4°C. This was followed by one gentle wash with 5 ml of
ice-cold DPBS. The cells were then gently scraped and dissociated in
0.6 ml of ice-cold lysis buffer, consisting of 20 mM tris-HCl (pH 7.5),
150 mM NaCl, 5 mM MgCl[2] (Nacalai Tesque), 1 mM dithiothreitol (DTT;
WAKO), a protease inhibitor cocktail, CHX (100 μg/ml), chloramphenicol
(100 μg/ml), and 1% Triton X-100. The cell suspension was collected
into a prechilled 1.5-ml DNA LoBind tube (Eppendorf). The lysate was
incubated for 15 min on ice with TURBO deoxyribonuclease (DNase; 25
U/ml; Thermo Fisher Scientific) before centrifugation at 20,000g for 10
min at 4°C. The cleared supernatant was then transferred to a fresh
1.5-ml tube. The samples were rapidly frozen using liquid nitrogen and
stored at −80°C.
A continuous sucrose gradient ranging from 10 to 45% was prepared using
10 and 45% sucrose solutions (Sigma-Aldrich) in a 14 × 95 mm polyclear
tube (Seton). The gradient was created in the presence of CHX (100
μg/ml) and 1 mM DTT in polysome buffer [25 mM tris-HCl, (pH 7.5), 150
mM NaCl, and 15 mM MgCl[2]] using the Biocomp Gradient Master program
(Biocomp). Thawed cell lysates were measured for RNA concentration
using the Qubit RNA BR Assay Kit (Thermo Fisher Scientific). A
consistent volume of cell lysate containing 40 μg of RNA from each
sample (300 μl) was layered onto the continuous sucrose gradient. The
polysomes were separated by centrifugation in a Himac ultracentrifuge
using a P40ST rotor (Himac) at 36,000 rpm for 2.5 hours at 4°C. The
relative RNA abundance in ribosomal subunits, monosomes, and polysomes
was detected using a 254-nm ultraviolet light with the Biocomp Piston
Gradient Fractionator (Biocomp). Area under the curve was calculated
using GraphPad Prism 8.0.2.
KD via RNA interference
To produce lentiviruses, pLKO.1-Blast (Addgene, #26655) encoding
scramble shRNA (Addgene, #26701), EIF3D shRNA-1 or shRNA-2 (table S2)
was cotransfected with psPAX2 and pMD2.G into 293T/17 cells as
described above ([178]70). The collected viral supernatant was
concentrated using a Lenti-X Concentrator and titrated using a Lenti-X
qRT-PCR Titration Kit. These viral stocks were exposed to 293T/17 cells
(passage 27), HDF1419 (passage 12), and HepG2 cells (passage 13) at
5000 viral particles per cell overnight at 37°C. Starting 2 days later,
cells were selected with blasticidin S (10 μg/ml; Wako) until
noninfected cells were completely dead.
mRNA synthesis and luciferase assay
For the luciferase reporter assay, a T7 promoter–driven NanoLuc
luciferase (NLuc) was constructed. The 5′UTR sequences of RBBP6 (1061
bp from [179]NM_006910.5) and GAPDH (76 bp from [180]NM_002046.7) were
inserted between the T7 promoter and NLuc using NEBuilder assembly
cloning (New England Biolabs) according to the instructions. To prepare
a control plasmid for normalization, we introduced a firefly luciferase
(FLuc) gene into the linearized vector of the Takara IVTpro mRNA
Synthesis System (Takara). Using these plasmids, mRNAs were synthesized
according to the manufacturer’s protocol. We used CleanCap Reagent AG
(Trilink BioTechnologies) as the cap analog and N1-methylpseudouridine
5′-triphosphate sodium salt (TCI Chemicals) instead of uridine
triphosphate and performed quality control using a bioanalyzer
(Agilent).
After 2 days of Dox addition, the control and sgEIF3D iPSCs were
replated at a density of 7.5 × 10^4 cells per well in iMatrix
511-coated 24-well plates with StemFit AK02N media, supplemented with
10 μM Y-27632 and Dox. The following day (day 3 post-Dox addition), we
transfected the mRNA mix (125 ng each of NLuc reporter and FLuc
control) into the cells using Lipofectamine RNAi Max reagent (Thermo
Fisher Scientific). The next day, cells were lysed in 1× passive lysis
buffer (Promega) and measured luciferase activities using a GloMax
Discover Microplate Reader (Promega) and a Nano-Glo Dual-Luciferase
Reporter Assay System (Promega).
RNA sequencing
Cells were lysed using QIAzol reagent, and total RNA was purified
according to the protocol mentioned earlier. RNA quality was assessed
using an Agilent RNA6000 Pico Kit on a Bioanalyzer 2100 (Agilent). The
library preparation and subsequent analysis were carried out following
methods outlined in previous studies ([181]71, [182]72). Briefly, 100
ng of DNase-treated total RNA was used for library preparation with the
Illumina Stranded Total RNA Prep Ligation with Ribo-Zero Plus kit
(Illumina). The libraries were evaluated using an Agilent
High-Sensitivity DNA Kit (Agilent) and then sequenced using either a
NextSeq 500/550 High Output v2 Kit (Illumina), NextSeq 1000/2000 P2
Reagents (100 cycles) v3 (Illumina), or HiSeq X (Illumina). The adapter
sequence was trimmed using cutadapt-1.12 ([183]73). Reads mapping to
ribosomal RNA (rRNA) were excluded using SAM tools (version 1.10)
([184]74) and Bowtie 2 (version 2.2.5) ([185]75). Reads were aligned to
the hg38 human genome using STAR Aligner (version 2.7.10b) ([186]76).
Quality checks were performed using RSeQC (version 4.0.0) ([187]77).
Reads were counted with HTSeq (version 0.13.5) ([188]78) using the
GENCODE annotation file (version 35) ([189]79). Counts were normalized
using DESeq2 (version 1.34.0) in R (version 4.1) ([190]80). The DESeq2
package was also used to perform Wald tests. Principal components
analysis and heatmaps were generated using prcomp and pheatmap,
respectively. GO analysis was conducted and visualized using
clusterProfiler ([191]67, [192]68). Hierarchical clustering was
performed by using Euclidean distance and complete linkage method in R.
Pearson’s correlation and P values of t test were calculated in R.
Ribosome profiling
Ribosome profiling was conducted as previously outlined ([193]81,
[194]82). Cells were lysed in a buffer containing 20 mM tris-HCl (pH
7.5), 150 mM NaCl, 5 mM MgCl[2], 1 mM DTT, 1% Triton X-100,
chloramphenicol (100 μg/ml), and CHX (100 μg/ml), followed by a 15-min
DNase treatment on ice. RNA concentrations in the lysate were measured
using the Qubit RNA BR assay kit (Thermo Fisher Scientific). We treated
10 μg of RNA with ribonuclease I (Epicentre) for 45 min at 25°C. The
ribosome footprint RNA was then concentrated via ultracentrifugation
using a sucrose cushion [20 mM tris-HCl (pH 7.5), 150 mM NaCl, 5 mM
MgCl[2], 1 mM DTT, SUPERase-In (20 U/ml; Thermo Fisher Scientific), 1 M
sucrose, chloramphenicol (100 μg/ml), and CHX (100 μg/ml)]. The
resulting pellets were resuspended in pellet buffer [20 mM tris-HCl (pH
7.5), 300 mM NaCl, 5 mM MgCl[2], 1 mM DTT, 1% Triton X-100, and
SUPERase-In (20 U/ml)] and purified using the Direct-zol RNA Microprep
kit (Zymo Research). The RNA samples were separated by electrophoresis,
and fragments ranging from 17 to 34 nt were excised and purified using
Dr. GenTLE Precipitation Carrier (Takara). These purified ribosome
footprint RNAs were ligated with linker oligonucleotides containing an
inner index sequence and a unique molecular identifier (UMI), followed
by rRNA depletion using riboPOOLs for Ribo-seq (siTOOLs Biotech). The
residual RNAs were reverse-transcribed using ProtoScript II (New
England Biolabs) and circularized with circLigase2 (Epicentre). The
cDNA templates were amplified using Phusion polymerase (New England
Biolabs) with index-sequenced primers.
To calculate translational efficiency, corresponding RNA sequencing
(RNA-seq) experiments were performed using RNA extracted from the lysis
buffer. We used TRIZOL LS reagent (Thermo Fisher Scientific) and the
Direct-zol RNA Microprep kit for RNA extraction. The RNA-seq libraries
were prepared as instructed by manufacturer’s protocol, except using
riboPOOLs for RNA-seq (siTOOLs Biotech) in the rRNA depletion step and
xGen UDI-UMI Adapters (Integrated DNA Technologies) in the adapter
ligation step. The cDNA libraries were sequenced following the RNA-seq
protocol. The reads were demultiplexed using the inner index, and
adapters were removed using fastp (version 0.22.0) and fastx-split
([195]81, [196]83). To filter out reads mapping to rRNA, Bowtie2 and
SAMtools were used. The remaining reads were aligned to the human
genome (hg38) using STAR (version 2.7.10b), and duplicates were removed
on the basis of UMI using bam-suppress-duplicates. Quality control
statistics were calculated using fp-framing. Read counting and
normalization were performed using fp-count and DESeq2, respectively
([197]81). TE and fold change values were calculated using the average
values of replicates as follows (i indicates a gene)
[MATH:
TEi=Ribosome profiling
averaged normalized
countiRNA
sequencing averaged normalized
counti :MATH]
[MATH: Fold
changei=TEi of sgEIF3DTEi of control
:MATH]
For identifying differentially translated transcripts, we used DESeq2,
using a likelihood ratio test (full model: experiment + target +
experiment:target; reduced model: experiment + target). In pathway
enrichment analysis, Enrichr was used with the WikiPathways_2021_Human
database. To analyze upstream, downstream, and newly identified ORFs,
we first obtained a bed file using ORF-RATER with ribosome profiling
data of the parental iPSC line WTB6, treated with CHX and harringtonine
([198]58, [199]84). Using this bed file, reads from control and sgEIF3D
(Dox + 3 days) samples were counted by fp-count. We then conducted a
statistical analysis to identify significantly different transcripts
between these conditions, using DESeq2 with a Wald test.
Statistics
The quantitative measurement results are presented as individual data
points, depicted by colored dots, with means indicated by bars. In some
instances, bar graphs with individual data points and error bars
representing SDs are used. Statistical analyses included unpaired
two-tailed t tests to calculate P values, assessing differences between
two groups. Furthermore, one-way analysis of variance (ANOVA) was used
for multiple comparisons. These analyses were performed using GraphPad
Prism version 8.0.2 (GraphPad) and Excel (Microsoft). Statistical
significance was determined by P values or adjusted P values less than
0.05, denoted by asterisks in the figures. The specific values are
detailed in the figure legends.
Supplementary Material
20240409-1
[200]sciadv.adq5484.v1.pdf^ (9.8MB, pdf)
Acknowledgments