Graphical abstract

   graphic file with name fx1.jpg
   [57]Open in a new tab

Highlights

     * •
       Oncofetal protein IGF2BP1 is a key stabilizer of IQGAP3 mRNA
     * •
       IGF2BP1’s targets SRF and FOXM1 act as additional regulators of
       IQGAP3 expression
     * •
       IGF2BP1-SRF-FOXM1-IQGAP3 network contributes to cancer stemness
     __________________________________________________________________

   Cell biology; Stem cells research; Cancer

Introduction

   The study of stem cell regulators is necessary to understand how
   proliferation and multipotency are maintained in stem cells and how
   they are hijacked in cancer. Our recent work identified cytoskeletal
   protein IQGAP3 (IQ motif containing GTPase-activating protein 3) as a
   highly specific marker of proliferating stem cells in the stomach
   corpus ([58]Matsuo et al., 2021). We further found rapid upregulation
   of IQGAP3 expression, which was accompanied by massive
   dedifferentiation of mature lineages, upon tissue damage ([59]Matsuo
   et al., 2021). Our observation that the depletion of IQGAP3 led to
   differentiation indicated a strong requirement for IQGAP3 in the
   maintenance of stem cell potential and proliferation ([60]Matsuo
   et al., 2021).

   IQGAP3 is a scaffold protein with multiple protein binding modules,
   such as the calponin homology domain (CHD) for F-actin-binding, the WW
   domain that interacts with ERK ½, calmodulin-binding IQ motifs, Ras
   GTPase-activating protein-related domain (GRD) that interacts with
   Rac1/Cdc42, and the RasGAP binding domain at the C-terminus (RGCT)
   ([61]Hedman et al., 2015; [62]Nojima et al., 2008). These domains allow
   IQGAP3 to integrate diverse biological pathways for important roles in
   cell adhesion, motility, proliferation, and signal transduction (eg.
   RAS-ERK cascade and Rho-Rac1) ([63]Nojima et al., 2008; [64]Wang
   et al., 2007). The Bgee database (Swiss Institute of Bioinformatics)
   revealed that IQGAP3 expression is highest during mammalian
   embryonic/fetal development when compared to the adult. Moreover,
   IQGAP3 is essential for cell proliferation and motility—through
   FGFR1-Ras signaling—during zebrafish embryonic development ([65]Fang
   et al., 2015), thereby reinforcing the notion that IQGAP3 is necessary
   for development across species.

   IQGAP3 is overexpressed in several types of human malignant tumors
   including gastric, colorectal, liver, pancreatic, lung, ovarian, and
   breast cancers ([66]Dongol et al., 2020; [67]Hu et al., 2016; [68]Oue
   et al., 2018; [69]Skawran et al., 2008; [70]Wu et al., 2019; [71]Xu
   et al., 2016; [72]Yang et al., 2014). Importantly, high expression
   levels of IQGAP3 are associated with poor prognosis in different cancer
   types ([73]Koon et al., 2004; [74]Kumar et al., 2017; [75]Skawran
   et al., 2008; [76]Wu et al., 2015). Our finding that IQGAP3 is involved
   in the maintenance of stem cell properties in human embryonic and
   undifferentiated gastric cancer cells ([77]Matsuo et al., 2021)
   suggests that IQGAP3 could be a potential therapeutic target for cancer
   treatment.

   The mechanisms underlying the tight association of IQGAP3 expression
   with proliferating stem cells and cancer cells are unclear. Our study
   aims to delineate the regulation of IQGAP3 expression in the
   pluripotent embryonic cancer cell line model NTERA-2 and the
   undifferentiated human gastric cancer cell line, HGC-27. We found that
   IQGAP3 RNA stability is enhanced by the oncofetal protein IGF2BP1
   (insulin-like growth factor 2 mRNA-binding protein 1) in an
   N^6-methyladenosine (m^6A) RNA methyltransferase METTL3/14-dependent
   manner. IGF2BP1 has earlier been shown to bind to m^6A-modified mRNAs,
   thereby stabilizing the RNA of cancer-associated genes such as MYC,
   E2F1, and Serum Response Factor (SRF) ([78]Degrauwe et al., 2016;
   [79]Huang et al., 2018; [80]Muller et al., 2019, [81]2020). SRF
   regulates gene expression through cooperation with ternary complex
   factors (TCFs) and with myocardin-related transcription factors (MRTFs)
   ([82]Clark and Graves, 2014). SRF-TCF complexes modulate the
   transcription of genes involved proliferation and response to growth
   factors while SRF-MRTF drives the transcription of genes related to
   cytoskeleton regulation, cell adhesion, and migration ([83]Clark and
   Graves, 2014; [84]Esnault et al., 2014; [85]Leitner et al., 2011). More
   recently, SRF has been shown to destabilize cellular identity through
   its regulation of cytoskeletal genes ([86]Ikeda et al., 2018). We found
   that SRF strongly upregulates IQGAP3 expression in NTERA-2 and HGC-27.
   We also identified FOXM1 (forkhead box protein M1) RNA as a novel
   direct target of IGF2BP1 and as an upstream regulator of IQGAP3.
   Together, our study suggests that the aberrant upregulation of IQGAP3
   in cancer may be owing to the re-expression of IGF2BP1 in malignant
   tissues and its interplay with SRF and FOXM1. IGF2BP1, SRF, FOXM1, and
   IQGAP3 may therefore form a critical axis for tumorigenesis and cancer
   stem cell viability of undifferentiated cancer cells, such as HGC-27.

Results

IGF2BP1 binds and stabilizes m^6A-modified IQGAP3 RNA in NTERA-2 cells

   To identify the regulatory factors of IQGAP3 gene expression, we first
   queried the public database The Encyclopedia of RNA Interactomes
   (ENCORI) ([87]http://starbase.sysu.edu.cn/) ([88]Li et al., 2014). This
   interface provides access to the enhanced crosslinking and
   immunoprecipitation (eCLIP) datasets, which are used for the
   identification of target sites of RNA binding proteins ([89]Van
   Nostrand et al., 2016). Distinct enrichment of IGF2BP1 at the 3′
   untranslated region (3′UTR) of IQGAP3 RNA from the human liver
   hepatocellular carcinoma cell line HepG2 and the multipotential,
   hematopoietic malignant cells K562 cell line were observed
   ([90]Figure S1A).

   IGF2BP1 is expressed mainly during embryonic development, with likely
   roles in cell migration, metabolism, and stem cells ([91]Degrauwe
   et al., 2016). After birth, IGF2BP1 expression drops to very low or
   negligible levels in most tissues ([92]Degrauwe et al., 2016).
   Importantly, it is re-expressed in a broad range of cancer types and
   typically associated with poor prognosis ([93]Degrauwe et al., 2016).
   We compared the RNA expression levels of Iqgap3 and Igf2bp1 in the
   mouse blastocyst stage and various adult tissues and found stronger
   co-expression of both in the blastocyst compared to adult tissues
   ([94]Figure 1A). It would seem that Iqgap3 and Igf2bp1 play prominent
   roles in embryonic development. Interestingly, Iqgap3 expression is
   higher in the adult stomach than in tissues such as adult brain, heart,
   kidney, and liver, whereas Igf2bp1 levels are negligible in all adult
   tissues ([95]Figure 1A). We had earlier detected high IQGAP3 expression
   specifically in rapidly proliferating isthmus stem cells in the adult
   stomach corpus ([96]Matsuo et al., 2021). It is possible that, while
   markedly reduced in the adult, IQGAP3 expression is still maintained in
   proliferative tissues with high regeneration capacity. By contrast,
   IGF2BP1 expression and function appear to be mainly fetal. As IGF2BP1
   is expressed in pluripotent and various cancer cell lines, we selected
   the pluripotent human testicular embryonal carcinoma cell NTERA-2 to
   further explore the oncofetal association between IQGAP3 and IGF2BP1.

Figure 1.

   [97]Figure 1
   [98]Open in a new tab

   IGF2BP1 binds to 3′UTR of IQGAP3 in an m^6A-dependent manner in NTERA-2
   cells

   (A) Comparison of Igf2bp1 and Iqgap3 mRNA expression in mouse
   blastocyst and various adult mouse tissues. Statistical significance
   was determined by the one-way ANOVA method.

   (B) Motif enrichment in IGF2BP1 binding sites in RIP-seq.

   (C) GO pathway enrichment analysis of IGF2BP1 binding target genes
   identified by RIP-seq.

   (D) Distribution of IGF2BP1 binding sites across MYC and SRF as
   identified by RIP-seq in NTERA-2 overexpressed with FLAG-IGF2BP1.

   (E) IGF2BP1 binding to IQGAP3 3’UTR as identified by RIP-seq.

   (F) RIP-qPCR validation using IGF2BP1 antibody.

   (G) RIP-qPCR using m^6A antibody in NTERA-2 cells reveals the presence
   of m^6A modification in IQGAP3, MYC, and SRF mRNAs.

   Data bars represent mean values with standard deviation (n=3). Data in
   (G and H) were statistically analyzed using two-tailed Student’s t-test
   (∗∗∗∗ for P-value of <0.0001, ∗∗∗ for P-value of <0.001 and NS for not
   significant). See also [99]Figures S1 and [100]S2.

   We introduced FLAG-tagged IGF2BP1 into NTERA-2 cells and used
   FLAG-affinity beads to immunoprecipitate FLAG-IGF2BP1-ribonucleoprotein
   complexes for RNA immunoprecipitation sequencing (RIP-seq). Motif
   analysis revealed the enrichment of various motifs resembling the known
   IGF2BP1 consensus binding motif 5’-GGAC-3′at the IGF2BP1 peaks
   ([101]Figure 1B). Gene ontology (GO) analysis of the IGF2BP1 binding
   sites indicated significant enrichment for genes involved in Rho GTPase
   signaling, insulin receptor signaling, organelle organization,
   developmental growth, positive regulation of epithelial cell
   proliferation, and negative regulation of differentiation
   ([102]Figure 1C). As expected, RIP-seq showed enriched IGF2BP1 binding
   at the previously identified targets MYC and SRF mRNA, thereby
   indicating the specificity of our screen ([103]Figure 1D). Importantly,
   the IGF2BP1 peaks observed throughout the IQGAP3 RNA—with strong
   enrichment at the 3′ UTR—confirmed IQGAP3 as a high-confidence RNA
   target of IGF2BP1 ([104]Figure 1E). It is interesting that of the three
   IQGAP family members, IQGAP3 mRNA showed the strongest binding by
   IGF2BP1: IGF2BP1 binding of IQGAP1 RNA was 2-fold less than that in
   IQGAP3, while the binding of IGF2BP1 to IQGAP2 RNA was negligible
   ([105]Figure S1B). Intriguingly, we also detected the interaction
   between IGF2BP1 and FOXM1 RNA ([106]Figure S2A). FOXM1 is a
   transcription factor that has been strongly linked to cell
   proliferation as well as stem cell self-renewal ([107]Liao et al.,
   2018). In the same vein, we also identified RNA of pluripotency factors
   NANOG and KLF4 as IGF2BP1 binding targets ([108]Figure S1C).

   To validate the RIP-seq data, we performed RIP of endogenous IGF2BP1 in
   NTERA-2 cells. RIP-quantitative PCR (RIP-qPCR) confirmed strong
   enrichment of IQGAP3 RNA in IGF2BP1 immunoprecipitates, relative to the
   HIST2H3A antibody which served as negative control ([109]Figure 1F).
   Similarly, MYC and SRF RNA were also bound by IGF2BP1, which served as
   positive controls for the RIP-qPCR ([110]Figure 1F). IGF2BP1
   preferentially binds m^6A-modified RNA. Therefore, we performed
   methylated RNA immunoprecipitation with m^6A antibody followed by qPCR.
   We confirmed that IQGAP3 RNA was modified by m^6A methylation at levels
   comparable to that of SRF and MYC RNA [111]Figure 1G).

   The heterodimer comprising methyltransferases METTL3 and METTL14
   catalyzes m^6A methylation of RNA ([112]Liu et al., 2014). We found
   that knockdown (KD) of METTL3/L14 abolished the ability of IGF2BP1 to
   interact with IQGAP3 RNA, thus indicating that m^6A modification of
   IQGAP3 RNA is necessary for binding to IGF2BP1 ([113]Figure 2A).
   Likewise, the interactions of IGF2BP1 with its known targets MYC and
   SRF RNA were strongly impaired after METTL3/L14 knockdown
   ([114]Figure 2A). Of note, IQGAP3, MYC, and SRF protein levels were
   depleted after METTL3/L14 knockdown, suggesting that m^6A may be
   involved in regulating the protein abundance of its target genes
   through the modulation of RNA levels ([115]Figure 2B).

Figure 2.

   [116]Figure 2
   [117]Open in a new tab

   IGF2BP1 stabilizes IQGAP3 RNA in a METTL3/14 dependent manner in
   NTERA-2 cells

   (A) Binding of IGF2BP1 to IQGAP3, SRF, and MYC mRNAs after the
   depletion of m^6A via METTL3/14 knockdown (KD). Data bars represent
   mean values with standard deviation (n = 3). Statistical analysis was
   conducted using two-tailed Student’s t-test (∗∗∗∗ for P-value of
   <0.0001, ∗∗∗ for P-value of <0.001 and NS for not significant).

   (B) Immunoblot showing IQGAP3, SRF, and MYC protein expression upon
   METTL3/14 KD.

   (C–E) Half-life analyses of IQGAP3, (D) MYC and (E) SRF mRNAs after
   IGF2BP1 (left) or METTL3/14 (right) KD.

   Each data point represents mean values with a standard deviation of 3
   independent experiments. Linear regression model was used to fit the
   data points. See also [118]Figure S3.

   We next ascertained whether IGF2BP1 regulates IQGAP3 mRNA turnover in
   an m^6A-dependent manner. NTERA-2 cells were treated with control
   siRNA, siRNA targeting IGF2BP1 or METTL3/L14, followed by actinomycin D
   to inhibit the transcription. Cells were harvested at various time
   durations and assayed for IQGAP3 mRNA by RT-qPCR. We found that IGF2BP1
   knockdown was associated with a significant reduction of IQGAP3 mRNA
   half-life ([119]Figure 2C). Similarly, METTL3/L14 knockdown also
   resulted in reduced IQGAP3 mRNA half-life. We also found that the
   half-lives of known IGF2BP1 targets MYC and SRF mRNAs were reduced upon
   the knockdown of IGF2BP1 or METTL3/14 ([120]Figures 2D and 2E).
   Therefore, our results indicate that IGF2BP1 bound and stabilized m^6A
   modified IQGAP3 mRNA ([121]Figure 2C).

   We further examined whether IGF2BP1 controls the stability of IQGAP3,
   MYC, and SRF mRNAs in another cell type, HEK293T. Similar to NTERA-2,
   knockdown of IGF2BP1 in HEK293T led to a dramatic reduction in the
   half-life of IQGAP3 mRNA ([122]Figure S3A). The half-lives of MYC and
   SRF mRNAs were also reduced after the knockdown of IGF2BP1 in HEK293T
   cells ([123]Figures S3B and S3C). Of note, the mRNA decay process in
   HEK293T cells follows the one-phase decay phenomenon (non-linear) while
   the linear regression analysis was conducted to generate the line that
   best fits the data of NTERA-2 cells. Moreover, the half-lives of
   IQGAP3, MYC, and SRF mRNAs in HEK293T cells were shorter compared to
   that of NTERA-2 cells.

The IGF2BP1-IQGAP3-SRF-FOXM1 network regulates the pluripotent transcriptome

   We recently reported the specific expression of IQGAP3 in rapidly
   proliferating isthmus stem cells of the stomach corpus ([124]Matsuo
   et al., 2021). Moreover, we showed that IQGAP3 knockdown in NTERA-2
   cells was associated with the reduction of the mRNA of Yamanaka factors
   NANOG, KLF4, SOX2, and MYC ([125]Matsuo et al., 2021). Therefore, in
   addition to its critical role in proliferation ([126]Nojima et al.,
   2008), we hypothesize that IQGAP3 is necessary for the maintenance of
   pluripotency and/or stem cell properties. Indeed, the induction of
   differentiation in NTERA-2 cells through treatment with retinoic acid
   (RA) led to the dramatic reduction of mRNA and protein levels of both
   IQGAP3 and IGF2BP1, with concomitant decrease of pluripotency factors
   ([127]Figure 3A). The expression of neuronal differentiation marker
   GFAP was strongly induced, thereby confirming differentiation onset
   ([128]Figure 3A). We next knocked down IGF2BP1 in NTERA-2 cells. qPCR
   and Western blot analysis revealed drastic reductions of the
   pluripotency/stem cell markers (ie. OCT4, CD44v9, SOX2, NANOG, KLF4),
   reminiscent of RA-induced differentiated cells ([129]Figure 3B).
   Importantly, IGF2BP1 knockdown in NTERA-2, similar to IQGAP3 knockdown,
   was associated with morphological changes resembling differentiation
   ([130]Figure 3E). Of note, FOXM1 and SRF expression levels were reduced
   following IGF2BP1 depletion and onset of differentiation.

Figure 3.

   [131]Figure 3
   [132]Open in a new tab

   The IGF2BP1-IQGAP3-SRF-FOXM1 network regulates the pluripotent
   transcriptome

   (A) NTERA-2 cells were treated with retinoic acid (RA). qPCR and
   immunoblots show the mRNA and protein expression of IQGAP3, IGF2BP1,
   SRF, FOXM1, and other stem cell factors, respectively. C and D
   represent control and differentiated cells, respectively. Data bars
   represent mean values with SD(n = 3).

   (B–D) mRNA and protein expression levels of IQGAP3 and stem cell
   markers after IGF2BP1, (C) SRF or (D) FOXM1 KD in NTERA-2 cells.
   Differentiation status was assessed by GFAP expression. Data bars
   represent mean values with SD(n = 3).

   (E) Morphological changes in NTERA-2 cells after KD of IQGAP3, IGF2BP1,
   SRF, or FOXM1 compared to RA-induced differentiated cells. Scale bars
   represent 100 μm.

   (F) GSEA of differentially expressed genes in NTERA-2 cells after KD of
   IGF2BP1, IQGAP3, SRF, or FOXM1.

   (G) Venn’s diagram and H) GO pathway enrichment analysis of commonly
   downregulated genes after KD of IGF2BP1, IQGAP3, SRF, or FOXM1. Pathway
   enrichment was conducted using MetaScape software.

   See also [133]Figure S4 and [134]Table S3.

   Our RIP-seq data revealed IGF2BP1 binding sites within the FOXM1 RNA,
   with the highest peaks at the 3′ UTR. We further validated the RIP-seq
   data by performing IGF2BP1 RIP-qPCR assay to show strong enrichment of
   IGF2BP1 on FOXM1 RNA ([135]Figure S2B). The FOXM1 RNA bound by IGF2BP1
   was enriched with m^6A, as shown by m^6A-IP ([136]Figure S2C).
   Depletion of m^6A by METTL3/14 knockdown resulted in reduced binding of
   IGF2BP1 to FOXM1 RNA ([137]Figure S2D). It was reported that FOXM1
   expression increases at the start of the S-phase and that FOXM1 induces
   cell proliferation ([138]Laoukili et al., 2005). Moreover, FOXM1 has
   been described as an oncofetal protein that regulates stem cell renewal
   ([139]Bella et al., 2014). All these traits are reminiscent of IQGAP3,
   which is induced by serum and associated with stem cell potential
   ([140]Matsuo et al., 2021; [141]Nojima et al., 2008). Moreover, the
   gene expression analysis of The Cancer Genome Atlas (TCGA) dataset
   which comprises 407 stomach adenocarcinoma samples revealed a strong
   positive correlation of IQGAP3 mRNA expression with that of FOXM1
   ([142]Figure S2E). The UCSC genome browser ChIP-seq public database
   showed small but clear FOXM1 binding in proximity to the H3K4me3 mark
   at the IQGAP3 promoter (data not shown).

   We and others have shown that SRF is a direct target of IGF2BP1 (this
   work and [143]Muller et al., 2020). SRF is a well-established
   transcription regulator of mitogen-activated genes (eg. cytoskeletal
   genes) ([144]Posern and Treisman, 2006). More recently, SRF was
   reported to repress cell-type-specific genes and promote pluripotency
   ([145]Ikeda et al., 2018). We queried the publicly available UCSC
   genome browser ChIP-seq database using the IGV tool ([146]Kent et al.,
   2002; [147]Robinson et al., 2011) and observed the occupancy of SRF,
   which overlapped with its consensus binding sequence, at the IQGAP3
   gene promoter in HCT-116 and MCF-7 cells ([148]Figures S4A and S4B).
   The SRF binding site was enriched with the H3K4me3 epigenetic mark,
   which indicates active transcription. We, therefore, performed
   ChIP-qPCR using SRF antibody and confirmed the SRF binding site between
   200 and 450 base pairs upstream of IQGAP3 promoter in NTERA-2 cells
   ([149]Figures S4C and S4D). Unfortunately, our ChIP-seq experiment did
   not show SRF binding to IQGAP3, suggesting highly dynamic or transient
   interaction (data not shown).

   It would seem that IGF2BP1 regulates SRF and FOXM1 mRNA, which in turn
   influences IQGAP3 gene activity. Therefore, we performed siRNA
   knockdown of SRF and FOXM1 in NTERA-2 cells. Similar to NTERA-2 singly
   treated with IQGAP3 and IGF2BP1 siRNA, SRF and FOXM1 siRNA led to
   reductions in the expression of the core pluripotency markers NANOG,
   KLF4, SOX2, and MYC, whereas GFAP levels were increased
   ([150]Figures 3C and 3D). Moreover, the morphology of cells after
   IQGAP3, IGF2BP1, SRF, or FOXM1 knockdown resembled that of
   differentiated NTERA-2 cells, suggesting that the NTERA-2 cells had
   undergone differentiation after IQGAP3, IGF2BP1, SRF, or FOXM1
   knockdown ([151]Figure 3E).

   Taken together, our results are indicative of an
   IGF2BP1-SRF-FOXM1-IQGAP3 cooperative in maintaining pluripotency in
   NTERA-2 cells. To further investigate the biological significance of
   this network, NTERA-2 cells treated with control siRNA (siC), siRNA
   targeting IQGAP3 (siIQGAP3), IGF2BP1 (siIGF2BP1), SRF (siSRF), or FOXM1
   (siFOXM1) were subjected to RNA sequencing (RNAseq)-based transcriptome
   profiling. Gene set enrichment analysis (GSEA) of the RNAseq data
   revealed blastocyst formation among the top-ranked enriched gene
   signatures which were downregulated in siIGF2BP1-treated cells
   ([152]Figure 3F). IGF2BP1 knockdown, therefore, promoted a more
   differentiated and less proliferative state in NTERA-2 cells.
   Similarly, GSEA of siSRF cells showed that genes related to blastocyst
   formation were downregulated ([153]Figure 3F). Given that SRF
   destabilizes cell identity ([154]Ikeda et al., 2018), and that SRF is a
   downstream target of IGF2BP1, it is not surprising that SRF knockdown
   is associated with a shift to the differentiation state. The GSEA of
   siFOXM1 also revealed the downregulation of blastocyst formation,
   suggesting the shutdown of embryogenesis upon FOXM1 downregulation
   ([155]Figure 3F). Analysis of siIQGAP3 revealed the downregulation of
   genes associated with self-renewal ([156]Figure 3F). These gene
   signatures indicate that IQGAP3 plays regulatory roles in the
   determination of cell fates and stemness.

   The GSEA of siIGF2BP1, siSRF, siFOXM1, and siIQGAP3 indicated the
   induction of the differentiated state, in agreement with the
   observations obtained from changes in cell morphology and RT-qPCR
   analysis after siRNA treatment. We next queried the shared gene targets
   of IGF2BP1, SRF, FOXM1, and IQGAP3. Of the 2755 genes that were
   downregulated in siIGF2BP1, siSRF, siFOXM1, and siIQGAP3 cells, 74
   genes were common targets ([157]Figure 3G and [158]Table S3). Gene
   ontology analysis (GO) of these 74 genes revealed signaling pathways
   regulating pluripotency of stem cells, cellular response to retinoic
   acid, and let-7 inhibition of ES cell reprogramming as the top-ranked
   program, reflecting the transition of NTERA-2 cells toward the
   differentiated state upon knockdown of IGF2BP1, SRF, FOXM1, and IQGAP3
   ([159]Figure 3H). Moreover, pathways that have been highly associated
   with IQGAP3 and SRF such as the positive regulation of MAPK cascade,
   regulation of cell-cell adhesion, and regulation of actin cytoskeleton
   are among the top significant pathways, highlighting the important
   roles of SRF and IQGAP3 in this network ([160]Figure 3H).
   Interestingly, the significant associations of these shared
   differentially expressed genes with tissue regeneration and degradation
   of extracellular matrix is reminiscent of the notion that “cancer is a
   wound that never heals” ([161]Ge et al., 2017). We queried the TCGA
   dataset and observed that IQGAP3 expression is significantly increased
   in tumor lesions, compared to adjacent normal, in most of the tissues
   ([162]Figures S5A and S5B). Moreover, IQGAP3 overexpression is
   associated with poor survival across all TCGA cancer datasets, and a
   similar pattern is observed for IGF2BP1 ([163]Figure S5C).

Inhibition of the IGF2BP1-SRF-FOXM1-IQGAP3 signaling axis promotes
differentiation in undifferentiated HGC-27 gastric cancer cells

   We, therefore, examined the interplay of IGF2BP1, SRF, FOXM1, and
   IQGAP3 in regulating stem cell potential in cancer and the possibility
   of inhibiting this interplay for cancer treatment. Undifferentiated
   gastric cancer cell line HGC-27 was treated with siRNA targeting
   IGF2BP1, SRF, FOXM1, and IQGAP3. Similar to NTERA-2 cells, knockdown of
   IGF2BP1, SRF, FOXM1, or IQGAP3 resulted in reduced stemness and a more
   differentiated state. qPCR and Western blot analysis revealed drastic
   reductions in stem cell markers, such as OCT4, SOX2, NANOG, and CD44v9
   ([164]Figures 4A–4C). In gastric cells, KLF4 was reported to reduce
   self-renewal and promote differentiation ([165]Miao et al., 2020).
   Accordingly, KLF4 levels remained unchanged in the differentiated
   cells. Moreover, differentiated cell marker Pepsinogen C (PGC) was
   induced in siIGF2BP1, siSRF, siFOXM1, and siIQGAP3, compared to control
   (siC) cells. To ascertain the proliferative capacity of the stem cells
   in control (siC) and IGF2BP1, SRF, FOXM1, and IQGAP3 siRNA-treated
   HGC-27 cells, we next performed the tumor spheroid assay. The treated
   cells were assessed for growth in serum-free and non-adherent
   conditions as only stem cells can proliferate in this environment
   ([166]Johnson et al., 2013). In agreement with the gene expression data
   reported here ([167]Figures 4A–4C) and previously reported data
   ([168]Matsuo et al., 2021), siIGF2BP1, siSRF, siFOXM1, and siIQGAP3
   treated cells showed a significant reduction in size and viability,
   relative to control ([169]Figure 4D). Together, the siRNA treatments
   showed that the individual depletion of IGF2BP1, SRF, FOXM1, and IQGAP3
   was sufficient to reduce stemness and promote the differentiation of
   HGC-27.

Figure 4.

   [170]Figure 4
   [171]Open in a new tab

   The IGF2BP1, SRF, FOXM1, and IQGAP3 cooperation maintains stemness in
   HGC-27 cancer cells

   (A–C) mRNA and protein expression levels of IQGAP3, stem cell markers,
   and stomach differentiation markers PGC and KLF4 after IGF2BP1, (B) SRF
   or (C) FOXM1 KD in HGC-27 cells. Data bars represent mean values with
   SD(n = 3).

   (D) Tumor spheroid sizes and viability after KD of IGF2BP1, SRF, FOXM1,
   or IQGAP3.

   Data bars represent mean values with SD (n = 3). Statistical analysis
   was conducted using two-tailed Student’s t test (∗∗∗∗ for p-value of
   <0.0001 and ∗∗∗ for p-value of <0.001). Scale bars represent 100 μm.
   See also [172]Figure S5.

Discussion

   IQGAP3 expression is elevated in most, if not all, cancer cells
   ([173]Dongol et al., 2020; [174]Jinawath et al., 2020; [175]Kumar
   et al., 2017; [176]Matsuo et al., 2021; [177]Monteleon et al., 2015;
   [178]Shi et al., 2017; [179]Wu et al., 2015, [180]2019; [181]Xu et al.,
   2016). This indicates the importance of IQGAP3 as a cancer driver gene
   and its involvement in the fundamental aspect of carcinogenesis. That
   IQGAP3 is essential for cell proliferation and maintenance of stemness
   suggests that it is mainly associated with proliferating cancer stem
   cells. Depletion of IQGAP3 reduced the proliferation of breast and
   gastric cancer cell lines ([182]Hu et al., 2016; [183]Oue et al.,
   2018). Moreover, IQGAP3 has been shown to regulate metastasis and EMT
   ([184]Jinawath et al., 2020; [185]Shi et al., 2017). IQGAP3 is
   therefore a likely target for the treatment of aggressively
   proliferating cells as well as metastasis in most cancers. However,
   what causes the specific increase of IQGAP3 expression in cancer is
   unclear.

   Cell cycle regulator E2F1 was shown to transactivate the IQGAP3
   promoter ([186]Lin et al., 2019). More recently, YAP was reported to
   promote the binding of B-MYB to the IQGAP3 promoter and enhancer,
   thereby upregulating IQGAP3 transcription ([187]Leone et al., 2021).
   Here, we identified regulators of IQGAP3 expression, which may explain
   its specific expression in proliferating stem cells and cancer cells.
   We report the importance of IGF2BP1-SRF-FOXM1-IQGAP3 signaling axis in
   maintaining stem cell properties in a pluripotent stem cell NTERA-2 and
   an undifferentiated gastric cancer cell line HGC-27. IGF2BP1 has long
   been implicated in cell metabolism and stem cell maintenance during
   early development ([188]Degrauwe et al., 2016). We show that IGF2BP1
   and IQGAP3 expression levels are highest in the blastocyst stage and
   that expression of both showed corresponding decreases during
   adulthood. The re-expression of IGF2BP1 during malignant transformation
   is likely to increase IQGAP3 mRNA stability. On the other hand, the
   differences between the Iqgap3 and Igf2bp1 expression levels in adult
   mouse stomach tissues suggest that there are mechanisms other than
   IGF2BP1, which regulate the IQGAP3 expression, especially in rapidly
   proliferating tissues similar to stomach. E2F1 or YAP, previously
   reported to have a role in the transcriptional regulation of IQGAP3,
   may be accountable for these mechanisms.

   Moreover, IGF2BP1-mediated increase of SRF and FOXM1 mRNA levels afford
   a further layer of IQGAP3 regulation, namely a link to mitogenic
   stimulation. SRF is a major regulator of actin cytoskeleton genes
   ([189]Miano et al., 2007). It is, therefore, not surprising that
   IQGAP3, with its activities in the reorganization of cytoskeletal
   architecture and cell proliferation ([190]Nojima et al., 2008;
   [191]Wang et al., 2007), is an SRF downstream gene. Although the public
   dataset and our ChIP-qPCR showed SRF binding at or near its consensus
   sites at the IQGAP3 promoter, our inability to use ChIP-seq to show SRF
   binding to the IQGAP3 gene indicates that its binding is transient.
   FOXM1 is important for proliferation, self-renewal, and tumorigenesis
   ([192]Liao et al., 2018). Similar to IQGAP3, FOXM1 is found mainly in
   highly proliferative cells, such as stem/progenitor cells, regenerating
   tissues, and cancer cells ([193]Liao et al., 2018; [194]Matsuo et al.,
   2021).

   The similarities between robust proliferation during tissue damage
   repair/regeneration and cancer have long been observed. IGF2BP1 plays
   important role in the tissue repair and homeostasis of the colonic
   epithelium ([195]Chatterji et al., 2021; [196]Hamilton et al., 2015;
   [197]Manieri et al., 2012; [198]Singh et al., 2020). Although our
   earlier study on stomach tissues did not show the upregulation of
   IGF2BP1 after tamoxifen-induced tissues damage, FOXM1 and SRF levels
   were increased, suggesting that both may be contributing to the strong
   IQGAP3 upregulation in the stomach during tissue repair.

   Our findings, which linked SRF and FOXM1 to the upregulation of IQGAP3
   gene expression, raise many interesting questions: are they critical
   for IQGAP3’s ability to maintain self-renewal and drive proliferation
   in cancer cells? FOXM1 and SRF have been linked to chemoresistance
   ([199]Okada et al., 2013; [200]Whitson et al., 2018). Do they, in
   response to external stimuli, activate IQGAP3 to stimulate
   proliferation in rare quiescent cancer stem cells and thus confer
   chemotherapy resistance? Clearly, more research is required to
   determine the mechanistic basis of the associations between IQGAP3,
   SRF, and FOXM1.

   Undifferentiated cancer cells are regarded to be more malignant than
   well-differentiated cancers, which possess limited tumorigenic
   potential. Undifferentiated cancer cells, with stem-like properties,
   are responsible for aggressive metastasis. Drugs that target this
   IGF2BP1-SRF-FOXM1-IQGAP3 axis will be therapeutically attractive as
   inducers of terminal differentiation in cancer stem cells. Given the
   commonality of elevated IQGAP3 in many cancer types, our work suggests
   that, in addition to identifying drug inhibitors of IQGAP3, inhibitors
   of IGF2BP1 binding activity ([201]Mahapatra et al., 2017), SRF/Rho/MRTF
   activity ([202]Leal et al., 2019) and FOXM1 transcription
   ([203]Petrovic et al., 2010; [204]Sleiman et al., 2011)—some of which
   are in clinical trials—may potentially be used to deplete IQGAP3 and
   improve cancer treatment.

Limitations of the study

   A limitation was that our study was performed using only two cell
   types, namely the pluripotent embryonic cancer cell line model NTERA-2
   and the undifferentiated human gastric cancer cell line HGC-27. To
   assess the commonality of the IGF2BP1-SRF-FOXM1-IQGAP3 network, more
   cell types including adult stem cells as well as undifferentiated
   cancer cell lines from diverse tissues should be studied. Moreover, the
   comparison of protein expression levels between IGF2BP1, SRF, FOXM1,
   and IQGAP3 in tumor tissue microarrays may be needed. Nevertheless,
   this proof-of-concept study reveals the hitherto unknown relationship
   of IGF2BP1, SRF, FOXM1, and IQGAP3 and suggests avenues for future
   mechanistic studies on the regulation and pharmacological inhibition of
   IQGAP3 in cancer stem cells.

STAR★Methods

Key resources table

   REAGENT or RESOURCE SOURCE IDENTIFIER
   Antibodies
     __________________________________________________________________

   SRF Cell Signaling Technology 5147
   FOXM1 Cell Signaling Technology 20459
   IQGAP3 Proteintech 25930-1-AP
   IGF2BP1 Cell Signaling Technology 8482
   IGF2BP1 MBL RN007P
   NANOG Cell Signaling Technology 4903
   OCT4 Abcam ab19857
   KLF4 Abcam ab106629
   CD44v9 Cosmo Bio LKG-M001
   GFAP Proteintech 16825-1-AP
   PGC Abcam ab93886
   GAPDH Cell Signaling Technology 2118
   SOX2 Proteintech 11064-1-AP
   cMYC Cell Signaling Technology 5605
   Rabbit IgG Cell Signaling Technology 2729
   m6A Sigma-Aldrich MABE1006
     __________________________________________________________________

   Chemicals, peptides, and recombinant proteins
     __________________________________________________________________

   All-Trans Retinoic Acid Sigma-Aldrich R2625
   Actinomycin D Sigma-Aldrich A9415
   Formaldehyde Sigma-Aldrich 252549
     __________________________________________________________________

   Critical commercial assays
     __________________________________________________________________

   RNeasy kit Qiagen 74104
   iScript cDNA synthesis kit Bio-Rad 1708890
   iTaq Universal SYBR green kit Bio-Rad 1725150
   QIAquick PCR purification kit Qiagen 28106
   CellTiter-GLO Promega G9681
   Pierce BCA Protein Assay kit Thermo Fisher Scientific 23225
     __________________________________________________________________

   Deposited data
     __________________________________________________________________

   RNA seq and RNA-IP seq data This paper [205]GSE196406
     __________________________________________________________________

   Experimental models: Cell lines
     __________________________________________________________________

   NTERA-2 ATCC CRL-1973
   HGC-27 CellBank Australia, the European
   Collection of Cell Cultures 94042256
   HEK293T ATCC CRL-3216
   HEK293FT ATCC CRL-11268
     __________________________________________________________________

   Oligonucleotides
     __________________________________________________________________

   siRNAs targeting IGF2BP1, SRF, FOXM1, and IQGAP3 Horizon Discovery
   Refer to [206]Table S1
   siRNAs targeting METTL3 and METTL14 IDT Refer to [207]Table S1
   Primers for qPCR of human target genes in knockdown experiments IDT
   Refer to [208]Table S2
   Primers for qPCR of mouse genes IDT Refer to [209]Table S2
   Primers for qPCR of target genes in RNA IP and m6A-IP qPCR assays IDT
   Refer to [210]Table S2
   Primers for ChIP-qPCR assays in IQGAP3 promoter regions IDT Refer to
   [211]Table S2
     __________________________________________________________________

   Software and algorithms
     __________________________________________________________________

   Prism 7.03 software GraphPad
   [212]https://www.graphpd.com/scientific-software/prism/
   QuantStudio3 Real-Time PCR systems Thermo Fisher Scientific
   [213]https://downloads.thermofisher.com/quantstudio3_5/v1.5.1/quantstud
   io3_5-1.5.1.update
   ImageJ software Fiji [214]https://imagej.nih.gov/ij/download.html
   [215]Open in a new tab

Resource availability

Lead contact

   Further information and requests for resources and reagents should be
   directed to and will be fulfilled by the lead contact, Yoshiaki Ito
   ([216]yoshi_ito@nus.edu.sg).

Materials availability

   This study did not generate new unique reagents.

Experimental model and subject details

Cell culture

   NTERA-2, human malignant pluripotent embryonal carcinoma cell line,
   obtained from ATCC (NTERA-2 cl.D1, Cat#CRL-1973, Lot#70004008) were
   cultured in DMEM (Dulbecco’s Modified Medium, Nacalai Tesque) with 10%
   FBS (Fetal bovine serum) at 37°C in a humidified 5% CO2 incubator
   according to the manufacturer’s instructions. Cells were maintained at
   high density and the medium was changed every 2–3 days. On confluency,
   cells were passaged at 1:4 ratio.

   For differentiation of NTERA-2, cells were treated with 10 μM of
   all-trans retinoic acid (Sigma-Aldrich, Cat#R2625) every 3–4 days for
   28 days. HGC-27, human gastric cancer cell line, obtained from CellBank
   Australia, the European Collection of Cell Cultures (Cat#94042256), and
   HEK293T and HEK293FT cells obtained from ATCC were cultured
   respectively in RPMI 1640 (Nacalai Tesque) or DMEM (Dulbecco’s Modified
   Medium, Nacalai Tesque) with 10% FBS at 37°C in a humidified 5% CO2
   incubator according to the manufacturer’s instructions. Cells were
   passaged at 1:10 ratio when the confluency reached 80–90%.

Method details

siRNA-based knockdown of target genes

   ON-TARGET plus SMARTpool siRNAs targeting human FOXM1, IGF2BP1, IQGAP3,
   SRF, METTL3/14 and a non-targeting siRNA pool (Dharmacon,
   Cat#D001810-10-50) as the control siRNA were used for the knockdown
   experiments. At 24 h after seeding in a 6-well culture plate, cells
   were transfected with 35 nM of siRNAs using jetPRIME (Polypus
   Transfection) according to the manufacturer’s protocol. At 48 h after
   transfection, cells were reseeded into a 6-well plate and a 6 cm dish
   which were then transfected in a similar manner at 6-8 h after seeding,
   and cells were harvested at 72 h later for RNA and protein
   quantification. The details of siRNA sequences are as described in
   [217]Table S1.

RNA extraction and qPCR

   Total RNA from the cell pellets was extracted using RNeasy Kit (Qiagen)
   with the removal of genomic DNA using the RNase-free DNase Set (Qiagen)
   according to the manufacturer’s instructions. Complementary DNA (cDNA)
   was prepared from 1 μg of total RNA extracted using iScript reverse
   transcription supermix for RT-qPCR (Bio-Rad). qPCR was performed using
   iTaq Universal SYBR green kit (Bio-Rad) and QuantStudio3 Real-Time PCR
   systems (Thermo Fisher Scientific), and the primers used are listed in
   [218]Table S2. The fold-change in the expression levels of target genes
   in treated cells compared to control cells was calculated using the
   2^-ΔCT method where GAPDH was used as a housekeeping gene.

Immunoblot analysis

   Cell pellets were first lysed by resuspending of the cell pellets with
   the lysis buffer (50 mM Tris pH 8, 150 mM NaCl, 0.1% SDS, 0.5% Sodium
   Deoxycholate and 1% NP-40) supplemented with Halt phosphatase inhibitor
   (Thermo Fisher Scientific), Complete protease inhibitor (Merck) and
   PMSF (Sigma-Aldrich) and incubating at 4°C for 45 min. After
   centrifugation of the lysates at 13500 rpm for 15 min at 4°C, the
   supernatant of each lysate was collected and the protein concentrations
   were measured using the Pierce BCA Protein Assay kit (Thermo Fisher
   Scientific, Cat # 23225) according to the manufacturer’s protocol.
   40 μg of protein was loaded into each well of 8% SDS/PAGE. The
   antibodies used listed in [219]key resources table.

Tumour spheroid assay

   For tumour spheroid assay, 1000 cells per well were seeded in an
   ultra-low attachment round bottom 96-well plate (Thermo Fisher
   Scientific, Cat # 174929), and cultured in normal growth medium. Cells
   were treated with siRNAs targeting genes of interest using reverse
   transfection method where suspension of cells were treated with 35 nM
   of siRNAs prior to seeding and jetPRIME (Polypus Transfection) was used
   according to the manufacturer’s protocol. Spheroid growth viability was
   determined on 5 days after seeding using CellTiter-GLO (Promega, Cat
   #G9681). Spheroid size was measured by using ImageJ software.

Mouse tissue extraction for gene expression analysis

   Blastocysts and adult mouse tissues samples were derived from C57BL/6
   mouse. For adult mouse tissue samples, 16-week old mice were sacrificed
   and tissues (brain, heart, kidney, liver and stomach) were extracted.
   Tissues samples were washed in ice-cold PBS for 3 times, and processed
   for RNA extraction using Trizol reagent (Invitrogen). After adding
   Trizol, aqueous phase was collected, and RNA was precipitated at −20°C
   for overnight using isopropanol and glycogen. Precipitated RNA was then
   cleaned up using ethanol. cDNA synthesis and qPCR analysis were done as
   described above.

RNA Immunoprecipitation

   RNA immunoprecipitation (RIP) was performed with some modifications
   ([220]Kwok et al., 2021). Briefly, cells seeded in a 10-cm plate were
   cross-linked by UV at 375J/cm^2 and harvested by scraping. Total cell
   extracts were prepared using polysome lysis buffer (100 mM KCl, 5 mM
   MgCl[2] , 10 mM Hepes (pH 7) and 0.5% NP-40) supplemented with 1×
   protease inhibitor, 200 units/ml Rnase out (Invitrogen, Cat #10777019)
   and 200 units/ml SUPERase IN (Ambion, Cat #AM2694), and 10% of the
   lysate was kept as an input. IGF2BP1 (MBL, Cat# RN007P) or rabbit IgG
   antibody (CST, Cat#2729) were incubated with Protein A/G Dynabeads®
   (Life Technologies, Cat #10015D) in 5% BSA (Bovine serum albumin) in
   PBS for 2 h at 4°C followed by an overnight incubation with 400 μg of
   cellular extracts at 4°C. Antibodies were used at the dilution of 1:75
   for one RIP reaction. Reaction mixture was then washed for 5 times with
   NT2 buffer (50 mM Tris-HCL (pH 7.0), 150 mM NaCl, 10 mM MgCl2, 0.05%
   (v/v) NP-40) and RNA was eluted by incubating beads with 200ul of
   1%SDS-TE buffer supplemented with 20ug of Proteinase K at 55°C for
   30mins. RNA was extracted using Trizol reagent (Invitrogen),
   precipitated with ethanol and cleaned up using
   Phenol:Chloroform:Isoamyl Alcohol. For RIP-qPCR assay, cDNA was
   generated from input and IP samples using iScript cDNA synthesis kit
   (BioRad) and the transcript enrichment of known target genes and genes
   of interest were examined by doing real-time qPCR with iTaq Universal
   SYBR green kit (Bio-Rad) and QuantStudio3 Real-Time PCR systems (Thermo
   Fisher Scientific). Primers used in the experiments were listed in
   [221]Table S2.

   For RIP-sequencing, NTERA-2 cells overexpressed with FLAG-IGF2BP1 and
   FLAG-affinity beads (Sigma Aldrich) were used, and RIP was performed as
   described above. RNA samples were sent to Beijing Genomics Institute
   (BGI) for cDNA synthesis, PCR amplification, transcriptome library
   preparation and sequencing.

Half-life analysis of RNA stability

   Cells were treated with 5 μg/ml actinomycin D (Sigma, Cat# A9415) and
   harvested at 0 (no treatment), 30, 60, 120, 240 and 360 min after
   treatment. The extraction of total RNA from the cell lysate, cDNA
   synthesis, and the measurement of mRNA levels of target genes and
   housekeeping genes were done as described previously. The expression
   levels of target genes were calculated using the 2^-ΔCT method where
   GAPDH was used as a housekeeping gene. The mRNA levels of target genes
   at different time points were normalized to the amount of mRNA at time
   0. Using the GraphPad Prism 7.03 software, the linear regression
   analysis was done to generate the line that best fits the data from
   NTERA-2 cells, while the non-linear curve (one-phase decay) was used
   for the data from HEK293T cells, and the half-life of mRNA was
   calculated. For statistical analysis, area under curve (AUC) was
   calculated and compared between the lines that best fits the data using
   GraphPad Prism 7.03.

m^6A-IP

   m^6A-IP was done with some modifications from previously reported
   method ([222]Dominissini et al., 2012). Total RNA was extracted using
   TRIzol (Invitrogen) and DNase I (QIAGEN) treatment was done to
   eliminate genomic DNA contamination. 150 ng of total RNA as an input,
   and 3 μg of RNA was used to incubate with 10 μg of m^6A antibody
   (Sigma-Aldrich, Cat# MABE1006) or rabbit IgG antibody (CST, Cat#2729)
   conjugated to Protein A/G Dynabeads® (Life Technologies, Cat #10015D)
   in IP buffer (50 mM Tris-HCL (pH 7.4), 750 mM NaCl and 0.5% NP-40)
   supplemented with 200 units/ml Rnase out (Invitrogen, Cat #10777019)
   and 200 units/ml SUPERase IN (Ambion, Cat #AM2694). After overnight
   incubation at 4°C, RNA-antibody complexes were pelleted and washed with
   IP buffer for 3 times followed by the isolation of RNA with Trizol
   reagent (Invitrogen) and the RNA precipitation with ethanol. RNA was
   then cleaned up using Phenol:Chloro-form:Isoamyl Alcohol method. cDNA
   was generated from input and IP samples using iScript cDNA synthesis
   kit (BioRad) and the transcript enrichment of known target genes and
   genes of interest were examined by doing real-time qPCR with iTaq
   Universal SYBR green kit (Bio-Rad) and QuantStudio3 Real-Time PCR
   systems (Thermo Fisher Scientific). Primers used in the experiments
   were listed in [223]Table S2.

Pathway enrichment analysis of target genes

   For the pathway enrichment analysis of IGF2BP1’s target genes
   identified by RNA-IP sequencing, the target genes with the most
   statistically significant enrichment scores were selected and analysed
   using Metascape analysis tool ([224]Zhou et al., 2019). The details of
   the analysis could be found on the webpage
   ([225]https://metascape.org). Briefly, for each given gene list,
   pathway and process enrichment analysis has been carried out using the
   different ontology sources including KEGG Pathway, GO Biological
   Processes, Reactome Gene Sets, Canonical Pathways and WikiPathways.
   Gene ontology terms with p value < 0.01, a minimum count of 3, and an
   enrichment factor >1.5 (the enrichment factor is defined as the ratio
   between the observed counts and the counts expected by chance) are
   grouped into clusters. The term with the most statistically significant
   value within a cluster is taken to represent the cluster. "Log10(P)"
   means the p value in log base 10, and "Log10(q)" means the multi-test
   adjusted p value in log base 10.

   For the pathway enrichment analysis of commonly downregulated genes in
   all 4 different knockdown gene sets (downregulated genes in NTERA-2
   cells after IGF2BP1, IQGAP3, SRF or FOXM1 knockdown), commonly
   downregulated genes were firstly defined using the publicly available
   tool called Draw Venn Diagram (ugent.be). The 74 commonly downregulated
   genes were then analyzed using Metascape analysis tool as described
   above.

Chromatin IP (ChIP)

   Chromatin IP was performed with some modifications of previously
   published method ([226]Lee et al., 2006). Briefly, cells were fixed
   with 1% formaldehyde for 10 min at room temperature (RT) and quenched
   by glycine (final concentration 125 mM) for 5 min at RT. Cells were
   washed 3 times by ice-cold PBS supplemented with 1 mM PMSF, collected
   by scrapping and then spinning down at 1500 rpm for 5 min. Nuclear
   extraction was done and nuclear extracts were resuspended in shearing
   buffer (1 mM EDTA, 10 mM Tris-HCl pH 7.6, 0.1% SDS) followed by
   sonication with ME220 Focused-ultrasonicator (Covaris) to achieve DNA
   fragments of sizes between 200 and 500 bp. Sonicated samples were then
   centrifuged at 13,000 rpm for 10 min to remove the cell debris and 1 mL
   of supernatant was incubated with 3 μg of desired antibody conjugated
   with Protein A/G Dynabeads (Life Technologies, Cat #10015D) for
   overnight at 4°C. Antibodies were listed in [227]key resources table.
   Reaction mixture was then washed with ice-cold low salt wash buffer,
   high salt wash buffer, LiCl wash buffer and TE buffer supplemented with
   proteinase inhibitor. Bound DNA was then eluted by incubating with
   elution buffer (50 mM Tris-HCl, pH 8.0, 10 mM EDTA, 1% SDS) at 65°C for
   30 min where tubes were quickly vortexed every 7–8 min. The eluted
   DNA-protein complexes were then treated with RNase A and proteinase K
   followed by incubating the reaction for overnight at 65°C for reverse
   cross-linking. DNAs were then purified by using QIAquick PCR
   purification kit (Qiagen). qPCR was done to measure the enrichment of
   the genes of interest bound by the target proteins.

TCGA and GTEX dataset analysis

   For the analysis of the IQGAP3 mRNA expression in different human
   cancer types, IQGAP3 mRNA expression data available in TCGA Pan-cancer
   Atlas dataset was downloaded via UCSC Xena analysis tool
   ([228]http://xena.ucsc.edu) ([229]Goldman et al., 2020). The expression
   data for three subtypes which are adjacent normal, primary tumor, and
   metastatic tumor tissue samples in 33 different cancer types (n =
   10,833) were then analyzed using GraphPad Prism 7.03 software.

   To compare IQGAP3 mRNA expression levels between normal and tumor
   samples in 18 different human tissues, IQGAP3 mRNA expression data
   available in TCGA and GTEX datasets were downloaded from UCSC Xena
   analysis tool ([230]http://xena.ucsc.edu) ([231]Goldman et al., 2020).
   The expression data derived from the normal and tumor tissue samples
   (n = 11493) were analysed using GraphPad Prism 7.03 software.

   GEPIA ([232]http://gepia.cancer-pku.cn/) was used to analyse the
   association between the expression levels of IQGAP3 and the overall
   survival in months across 33 different human cancer types available on
   TCGA datasets([233]Tang et al., 2017). GEPIA uses the Mantel-Cox test
   to calculate the hazards-ratio, and high-expression and low-expression
   cohorts were defined with the cut-off threshold set at 50%. Percent
   survival was calculated as the percentage of patient in the defined
   cohort who remained alive for a given period in months. Hazards ration
   of greater than 1 was defined as high and p value of less than 0.05 was
   defined as statistical significance.

RNA-seq analysis and GSEA

   RNA-seq reads were mapped by STAR to the human reference genome GRCh37
   with reference gene annotation GENCODE 26 ([234]Dobin et al., 2013;
   [235]Harrow et al., 2012). Reads with low mapping quality <20 were not
   used for further analysis. Next, PCR duplicates were removed in the
   paired-end alignments using SAMtools. Gene expression levels of
   individual genes/transcripts were generated using featureCounts
   ([236]Liao et al., 2014). Cross-sample normalization of reads counts
   was based on total mappable counts using RPKM. Normalized read counts
   were then applied to perform GSEA analysis to identify significantly
   enriched pathways/gene sets between the wildtypes and their
   corresponding knockdowns. The full version of MSigDB v7.2 was employed
   as GSEA dataset database.

RIP-seq analysis

   Reads from the RIP-Seq samples and the corresponding controls were
   mapped against human reference genome GRCh37 by STAR with GENCODE 26
   for transcriptome annotation. First, peak calling was performed by
   MACS2 to identify RIP binding enriched regions ([237]Zhang et al.,
   2008). For reads mapped to multiple exons or peaks presented in
   neighboring exons, the peaks were merged together. Sequences covered by
   the reproducible peaks of the duplicated samples were extracted for
   motif finding. The upstream and downstream sequences of reproducible
   peaks were also extracted and served as the background sequences in the
   motif finding. Motif discovery is performed with MEME suite using
   differential enrichment mode ([238]Bailey et al., 2009; [239]Bailey and
   Elkan, 1994).

Quantification and statistical analysis

   Statistical analyses were done by using Student’s t-tests (two-tailed)
   and all experiments were conducted at least for three replicates as
   indicated in the figure legends.

   For the overall survival curves, log rank analysis method was used to
   determine statistical significance. Data were presented as mean with
   standard deviation and p-value of <0.05 was considered as a statistical
   significance. For RNA-seq, three sets of RNA samples were used, and for
   RIP-seq, two sets of RNA samples were used. All western blot and
   brightfield images were representative of three independent
   experiments.

Acknowledgments