Staufen1 Represses the FOXA1-Regulated Transcriptome by Destabilizing FOXA1 mRNA in Colorectal Cancer Cells

Abstract Transcription factors play key roles in development and disease by controlling gene expression. Forkhead box A1 (FOXA1), is a pioneer transcription factor essential for mouse development and functions as an oncogene in prostate and breast cancer. In colorectal cancer (CRC), FOXA1 is significantly downregulated and high FOXA1 expression is associated with better prognosis, suggesting potential tumor suppressive functions. We therefore investigated the regulation of FOXA1 expression in CRC, focusing on well-differentiated CRC cells, where FOXA1 is robustly expressed. Genome-wide RNA stability assays identified FOXA1 as an unstable mRNA in CRC cells. We validated FOXA1 mRNA instability in multiple CRC cell lines and in patient-derived CRC organoids, and found that the FOXA1 3′UTR confers instability to the FOXA1 transcript. RNA pulldowns and mass spectrometry identified Staufen1 (STAU1) as a potential regulator of FOXA1 mRNA. Indeed, STAU1 knockdown resulted in increased FOXA1 mRNA and protein expression due to increased FOXA1 mRNA stability. Consistent with these data, RNA-seq following STAU1 knockdown in CRC cells revealed that FOXA1 targets were upregulated upon STAU1 knockdown. Collectively, this study uncovers a molecular mechanism by which FOXA1 is regulated in CRC cells and provides insights into our understanding of the complex mechanisms of gene regulation in cancer.


Introduction
Transcription factors (TFs) are broadly responsible for controlling transcription by directly binding to DNA in a sequencespecific manner.TFs exert control over diverse cellular processes, including, but not limited to, cellular differentiation, developmental patterning, cell cycle control, and apoptosis. 1,2ue to their importance in regulating the transcriptome, TFs are just as tightly regulated as their targets.Mutations in or misregulation of TFs can lead to many diseases, including cancer, diabetes, and neurological disorders. 1,3iven the significance of TFs in development and disease, understanding the processes that govern their regulation is of great importance.Some TFs, such as p53, the "guardian of the genome," are regulated post-transcriptionally at the protein level via protein stability, ubiquitination, or phosphorylation. 4he most well-established mechanism of p53 regulation is through a negative feedback loop with MDM2. 2,5Other TFs are regulated post-transcriptionally, at the mRNA level via mRNA decay and splicing. 6Post-transcriptional gene regulation can play an important role in regulating many cellular processes and contribute to several cancer hallmarks, including cell proliferation, differentiation, and survival. 7A common mechanism of post-transcriptional regulation is at the level of mRNA decay, in which the 3 0 untranslated region (3 0 UTR) plays a key role. 83 0 UTRs are often regulated by microRNAs (miRNAs) and RNA-binding proteins (RBPs) via direct binding in a sequence-or structure-specific manner, leading to transcript degradation and/or translation inhibition. 7,8Importantly, some RBPs have been found to be dysregulated in various diseases, including colorectal cancer (CRC), and contribute to CRC pathogenesis. 9,10A recent example of this is RBP-J, whose expression is significantly upregulated in CRC tissues and CRC cells, promoting proliferation, migration, and invasion. 11nderstanding the complex interactions between TFs, their targets, and the elements that regulate them is essential to our understanding of the processes driving human health and disease.
The pioneer transcription factor forkhead box A1 (FOXA1) is one such TF vital for proper development, organogenesis, and differentiation of many tissues, including those of the gastrointestinal tract. 12FOXA1 -/-mice do not survive longer than 14 days following their birth and experience abnormalities such as severe growth retardation and hypoglycemia in that short time span. 13,14FOXA1 is a pioneer TF, meaning it can bind to both highly condensed chromatin (heterochromatin) and less condensed chromatin (euchromatin) to activate transcription of its targets, 15 which include both coding and noncoding genes, such as CDH1, EPHB3, LINC00675 (also called FORCP or TMEM238L), and CEACAM5, all of which have been studied for their roles in cancer pathogenesis. 16,17OXA1's functions continue into adulthood, where it-and other members of the FOX family-help maintain homeostasis in the body. 129][20] In prostate cancer, FOXA1 directly interacts with androgen receptor, driving the growth and survival of normal and tumorous prostate cells, yet it has also been found to have tumor suppressive roles in late-stage PCa by regulating gene expression and the epithelialmesenchymal transition (EMT) independent of androgen receptor. 21,22In breast cancer, FOXA1 enhances the binding of estrogen receptor-a to its target genes in luminal type A. 23,24 Despite the wealth of research on FOXA1 in these cancer types, its role in CRC is not fully understood.Some reports suggest that FOXA1 functions as an oncogene, promoting cell proliferation and inhibiting apoptosis, 25 while others report that FOXA1 acts as a tumor suppressor, inhibiting EMT and preventing metastasis. 26Multiple studies have reported significant decrease in FOXA1 expression in CRC tumor tissues as compared to normal human colon tissues. 16,17Contrary to this, Ma et al. 25 previously reported a significant increase in FOXA1 mRNA levels in CRC tissues as compared to matched tumor-adjacent normal tissues.We previously showed that FOXA1 expression is significantly decreased in colon adenocarcinoma (COAD) cohorts compared to normal human colon tissues. 16,17Moreover, we found that FOXA1 expression is significantly lower in poorly differentiated CRC cell lines as compared to well-differentiated CRC lines. 17This latter point is notable because poorly differentiated CRC tumors are typically more aggressive and more likely to metastasize than their well-differentiated counterparts.
Building upon our previous studies, here we report the molecular mechanism of regulation of FOXA1 expression in CRC.We found that FOXA1 is a target of mRNA degradation in well-differentiated CRC cell lines, and this instability is mediated via the RBP STAU1, which targets the FOXA1 3 0 UTR.Furthermore, transcriptome analysis after STAU1 knockdown suggested that genes repressed by STAU1 are enriched for FOXA1 target genes, indicating that STAU1 regulates its targets through both direct binding and the regulation of the key transcription factor FOXA1. Collectively, this study identifies the molecular mechanism by which FOXA1 is regulated in CRC cells, providing insights into the complex mechanisms of gene regulation in cancer.

Genome-wide RNA stability assays identify FOXA1 as a target for mRNA degradation
To identify unstable RNAs in CRC cells, we performed genome-wide RNA stability assays from two CRC cell lines.Similar to CRC tumors, CRC cell lines can be either well-or poorly differentiated, and as a result may have differences in gene expression and gene regulation.In biological triplicates, we treated a well-differentiated CRC cell line (LS180) and a poorly differentiated CRC cell line (HCT116) for 0, 2 and 4 h with actinomycin D (ActD) to inhibit transcription and subsequently performed mRNA-seq (Figure 1A).In both cell lines, treatment with ActD led to robust changes in gene expression, with thousands of genes significantly downregulated after 2 and 4 h of ActD treatment (log2FC < -1; adj.p < 0.05) (Supplementary Table S1 and Table S2).These unstable RNA transcripts belonged to three categories: (1) more unstable in LS180 than HCT116, (2) more unstable in HCT116 than in LS180, and (3) unstable in both.1,591 of the downregulated RNAs, including, but not limited to, the transcription factor and proto-oncogene MYC, were unstable in both cell lines, while 1,445 RNAs such as CDKN1A (p21) were significantly more unstable only in LS180 cells, and 494 RNAs, including the WNT signaling pathway inhibitor DKK1, were significantly more unstable only in HCT116 cells (Supplementary Figure S1A to F).
In our RNA-seq data following ActD treatment, we found that FOXA1 mRNA is unstable in both poorly differentiated and well-differentiated CRC cells with a half-life of �2 h (Figure 1B and C).We have previously observed significantly decreased FOXA1 expression in CRC tissues compared to normal human colon tissues. 16,17We found that high FOXA1 expression was significantly correlated with better patient prognosis in the TCGA COAD data (Supplementary Figure S2), suggesting that FOXA1 functions as a potential tumor suppressor in CRC.We therefore sought to determine how FOXA1 mRNA is post-transcriptionally regulated in CRC cells.To validate our RNA-seq data we performed RT-qPCR from multiple CRC cell lines using both ActD and a second transcription inhibitor, 5,6-dichlorobenzimidazole 1-b-d-ribofuranoside (DRB).While ActD inhibits all RNA polymerases, DRB inhibits the activity of only RNA polymerase II.The half-life of FOXA1 mRNA was �2 to 3 h, confirming that it is indeed an unstable mRNA (Figure 1D, Supplementary Figure S1G to I).
We next examined the stability of FOXA1 mRNA in physiologically relevant CRC organoids derived from two patients.RNA stability assays using ActD indicated that FOXA1 mRNA is unstable with a half-life of �2 h in both CRC organoids (Figure 1E to F); MYC mRNA served as a positive control for unstable mRNA.These data demonstrate that FOXA1 mRNA is unstable not only in CRC cells, but also in patient-derived CRC organoids.Moving forward, we chose to use welldifferentiated CRC cell lines for our experiments because of more robust FOXA1 expression compared to poorly differentiated CRC lines.

FOXA1 mRNA instability is mediated via its 3 0 UTR but not by miRNAs
Due to the role of 3 0 UTRs in mediating mRNA stability 8 and the unstable nature of FOXA1 mRNA, we hypothesized that the instability of FOXA1 mRNA is regulated via its 3 0 UTR.FOXA1 has a relatively long 3 0 UTR of �1.4 kb that is highly conserved between human and mouse (Supplementary Figure S3A and data not shown).There is an established correlation between 3 0 UTR length and mRNA instability, further supporting our hypothesis that the 3 0 UTR could mediate this process.To investigate the effect of the FOXA1 3 0 UTR on FOXA1 mRNA stability, we used two approaches.First, we performed RNA stability assays using ActD in HCT116 cells that stably overexpressed either the FOXA1 open reading frame (ORF) (FOXA1-ORF-Flag) or the FOXA1-ORF-Flag þ 3 0 UTR (FOXA1-ORF-Flag þ 3 0 UTR).The flag-tag was inserted in frame before the stop codon.We achieved this overexpression using the pLVX-puro lentiviral-based system and chose HCT116 cells since they have low endogenous FOXA1 expression.RT-qPCR for the flag-tagged FOXA1 constructs in these cells showed that the half-life of FOXA1 mRNA decreased from �5 h to �2 h in the presence of the FOXA1 3 0 UTR (Figure 1G) suggesting that the 3 0 UTR could mediate the instability of the FOXA1 mRNA.
In the second approach, we performed dual luciferase reporter assays using the psiCHECK-2 vector.In this assay, we inserted either the human or mouse FOXA1 3 0 UTR into the 3 0 UTR of the Renilla luciferase gene of psiCHECK2; the Firefly luciferase gene of psiCHECK2 served as an internal control.The presence of the FOXA1 3 0 UTR led to a significant reduction in luciferase activity when compared to the empty vector control indicating that the FOXA1 3 0 UTR may have a repressive effect on gene expression (Figure 1H).The psiCHECK-2-FORCP-3 0 UTR served as a positive control. 27ince miRNAs can function to promote RNA degradation through interaction with target 3 0 UTRs, 28 we next sought to determine if the miRNA pathway regulates FOXA1 expression post-transcriptionally.To do this, we performed siRNA knockdowns of DICER, a protein essential for miRNA biogenesis. 29ollowing 48 h of transfection in LS180 cells, we found a significant reduction in DICER mRNA expression but no significant change in FOXA1 mRNA (Supplementary Figure S3B).As positive controls for miRNA targets, the MYC and ZEB1 mRNAs were modestly but significantly upregulated.Consistent with this data, at the protein level, upon knockdown of DICER1 in LS180 cells for 72 h, we did not observe a change in FOXA1 expression (Fig S3C).These results suggest that miRNAs likely do not play a major role in the regulation of FOXA1 expression.

RNA-binding protein Staufen1 binds to FOXA1 mRNA
RBPs are known to bind to their target mRNAs via a specific sequence or RNA secondary structure and regulate processes including RNA splicing, localization, stability, and degradation. 9,10To determine a potential role of RBPs in regulating FOXA1 mRNA expression by binding to the FOXA1 3 0 UTR in CRC cells, we performed in vitro streptavidin pulldowns (PD) following incubation of a biotinylated-FOXA1 3 0 UTR or biotinylated-luciferase RNA with LS180 whole cell lysates (Figure 2A).Mass spectrometry from biological triplicates of these PDs identified 79 proteins bound to the FOXA1 3 0 UTR (Supplementary Table S3).To determine if a subset of these RBPs could regulate FOXA1 expression in CRC, we intersected the pulldown data with the genes in the TCGA COAD cohort that have a significant negative correlation with FOXA1 expression (Spearman Correlation < -0.3, adj.p < 0.05).1][32] The average peptidespectrum matches (PSMs) for STAU1 in the negative control luciferase PDs was 9, while in the FOXA1 3 0 UTR PDs the average PSMs was 27 (Supplementary Figure S3D).We confirmed this PD-mass spectrometry data by immunoblotting and observed higher levels of STAU1 protein in the FOXA1 3 0 UTR PD as compared to the luciferase negative control PD (Figure 2C).The RBP RBM47 served as a negative control.
To further validate our mass spectrometry results we utilized a recently published method of RNA pulldown from cell lysates that utilizes biotinylated antisense oligos (ASOs). 33We designed four ASOs spanning the FOXA1 mRNA and performed RT-qPCR after FOXA1 mRNA PD from LS180 cell lysates.FOXA1 mRNA was enriched �13-to 56-fold in all four pulldowns compared to the input (Figure 2D and E).There was no enrichment of the negative control ACTB mRNA in these PDs, confirming the specificity.Immunoblotting from these pulldowns showed that STAU1 was enriched in PDs using ASOs 1, 3 and 4 and undetectable in the beads only negative control (Figure 2F).5][36] Although we PD STAU1 using three out of four ASOs targeting FOXA1 mRNA, it is unclear why we did not PD STAU1 in the FOXA1 ASO2 PDs.As a negative control, RBM47 was only detected in the input, confirming the specificity of the PDs (Figure 2F).
To further confirm the interaction between STAU1 and FOXA1 mRNA, we next performed native STAU1 ribonucleoprotein immunoprecipitations (RNP IPs).Immunoblotting confirmed enrichment of STAU1 in the IP (Figure 2G).By RT-qPCR, we observed �5-fold enrichment of FOXA1 mRNA in the STAU1 IPs, while there was no enrichment in the RBM47 IPs (Figure 2H).ARF1 and SERPINE1 mRNAs are known STAU1 targets 31,36,37 and were enriched �7-and �40-fold in the STAU1 IPs (Figure 2H).There was no enrichment of MALAT1 in the STAU1 IPs, further confirming the specificity (Figure 2H).These data suggest that the STAU1 protein interacts with FOXA1 mRNA.

STAU1 regulates FOXA1 mRNA stability via the FOXA1 3 0 UTR
We next examined the effect of STAU1 knockdown on FOXA1 expression at the RNA and protein level.We transfected three CRC cell lines with a SmartPool of four siRNAs against STAU1 for 72 h and determined the effect on FOXA1 expression by RT-qPCR and immunoblotting.STAU1 knockdown resulted in increased FOXA1 expression at both the mRNA and protein levels (Figure 3A to D and Supplementary Figure S3E to G).Specifically, we observed �1.5-and 3-fold increase in FOXA1 mRNA expression in LS180 and SW1222 cells (Figure 3A and  C).At the protein level, we observed a robust increase of FOXA1 in LS180 cells, but modest increase in DLD1 and SW1222 cells (Figure 3B and D and Supplementary Figure S3E to G).
To determine if the observed increase in FOXA1 expression upon STAU1 knockdown was due to an increase in FOXA1 mRNA stability, we performed RNA stability assays.
Depletion of STAU1 resulted in increased FOXA1 mRNA stability, with the half-life reproducibly increasing from �1 to 3 h in LS180 cells (Figure 3E).Of note, although the levels of FOXA1 mRNA were approximately the same in siCTRL and siSTAU1 samples after 6 h of ActD treatment, we observed a change in the rapid clearance of the FOXA1 mRNA upon STAU1 knockdown (Figure 3E).In dual luciferase assays, we observed a significant rescue of luciferase activity of the human FOXA1 3 0 UTR reporter upon STAU1 knockdown (Figure 3F).Moreover, STAU1 knockdown in HCT116 cells transduced with a lentivirus expressing the FLAG-tagged ORF of FOXA1 but lacking the FOXA1 3 0 UTR, did not alter FOXA1-FLAG protein expression, suggesting that the FOXA1 ORF is not regulated by STAU1 (Supplementary Figure S3H).These data suggest that STAU1 destabilizes the FOXA1 mRNA via the FOXA1 3 0 UTR.
In line with these data, we observed a significant negative correlation between FOXA1 and STAU1 expression in the TCGA COAD cohort (Figure 3G, Supplementary Table S4).Interestingly, there was no correlation between FOXA1 and STAU1 mRNA in normal human colon tissues from the TCGA COAD cohort (Figure 3H, Supplementary Table S5).This raises the possibility that the post-transcriptional repression of FOXA1 expression by STAU1 may be restricted to CRC tumors; further experiments are required to investigate this.

The STAU1-repressed transcriptome is enriched for FOXA1 target genes in CRC cells
Since FOXA1 is a transcription factor that regulates the expression of many genes, 15,38,39 and its expression increases upon STAU1 knockdown in CRC cells, we next determined if the expression of FOXA1 target genes increases upon STAU1 knockdown.We therefore performed RNA-seq from LS180 cells transfected with siCTRL or siSTAU1 and compared it with our recent RNA-seq data upon FOXA1 knockdown and FOXA1 ChIP-seq from LS180 cells. 17As expected, there was a significant reduction in STAU1 mRNA following STAU1 knockdown (Figure 4A).Thirty-six genes were significantly upregulated upon STAU1 knockdown (log2FC > 0.6, adj.p value < 0.05), significantly downregulated upon FOXA1 knockdown (log2FC < −0.6, adj.p value < 0.05), and had FOXA1 ChIPseq peaks in their promoters or enhancer regions (Figure 4B, Supplementary Table S6), including the canonical FOXA1 target gene TFF1.The binding sites for FOXA1 in these 36 genes vary from the promoter region, to upstream or downstream of the gene, to within the gene body.We observed increased reads for each of the FOXA1 target genes, including ISX, TFF1, and CCN2 (aka CTGF), upon STAU1 knockdown in LS180 cells and each gene had a FOXA1 ChIP-seq peak in the promoter (Figure 4C and D).These findings indicate that the STAU1-repressed transcriptome is enriched for a subset of FOXA1 target genes in CRC cells.While it is unclear why only 36 FOXA1 target genes were upregulated upon STAU1 knockdown, it may be the case that these are high-affinity FOXA1 targets.
Since these RNA-seq and ChIP-seq experiments were performed in LS180 cells, we sought to validate these data in other CRC cell lines.In each of the cell lines LS180, DLD1, and SW1222, STAU1 knockdown resulted in increased mRNA expression of the FOXA1 targets ISX, TFF1, and CCN2 by �1.5-3.5-fold(Figure 4E).Intriguingly, although we observed an increase in CCN2 expression upon STAU1 knockdown in the LS180 RNA-seq data, we could not validate this by RT-qPCR following STAU1 knockdown in LS180 cells (Figure 4E).Finally, gene ontology analysis (GO: biological process) for the 36 genes transcriptionally regulated by FOXA1 and posttranscriptionally repressed by STAU1 suggested a role of STAU1 in regulation of genes important for differentiation and development (Supplementary Figure S4).
In summary, we propose a model according to which the RBP STAU1 binds to the FOXA1 3 0 UTR, leading to degradation of the FOXA1 mRNA (Figure 5).Depletion of STAU1 reverses this effect, and results in an increase in both FOXA1 expression and the expression of specific FOXA1 target genes.Collectively, these data demonstrate that the STAU1repressed transcriptome consists not only of mRNAs directly degraded by STAU1, but also mRNAs that are indirectly regulated via the transcription factor FOXA1.

Discussion
The transcription factor FOXA1 is essential for mouse development, is involved in organogenesis and tissue differentiation, and has target genes ranging from coding to noncoding. 12,13,16,17,38,40As such, FOXA1, and many of its target genes, have been studied in the context of cancer pathogenesis. 16,17,26,27,41Despite the wealth of literature on FOXA1 regulation of its targets, little is known about the regulation of FOXA1 expression itself, especially in the context of CRC.To date, there is limited evidence of transcriptional and posttranslational regulation of FOXA1 expression in cancer.In androgen-dependent PCa, deletion or repression of a set of six cis-regulatory elements in the FOXA1 regulatory plexus leads to significant decreases in FOXA1 expression and PCa growth, 42 while in estrogen receptor negative (ER-) breast cancer, FOXA1 is transcriptionally regulated via multiple factors in the ErbB2-ERK pathway-including ERK, AP2a, and the TFs CREB1 and c-Fos. 43In CRC, SNAIL1 directly represses FOXA1 by binding to its promoter, ultimately resulting in the downregulation of all FOXA factors and an increase in mesenchymal gene expression along with changes in morphology. 26More recently, FOXA1 has been found to be posttranslationally regulated in PCa in a mechanism involving EZH2-mediated methylation at lysine-295, whereby there is reduced FOXA1 ubiquitination and increased FOXA1 protein stability. 44In CRC cell lines, NEDD4 similarly acts as an E3 ubiquitin ligase targeting FOXA1 protein for degradation. 45espite evidence of FOXA1 expression regulation at the transcriptional and post-translational levels, there is an apparent lack of understanding how FOXA1 expression is regulated post-transcriptionally, and our current study seeks to fill this gap.
We report here that FOXA1 mRNA is highly unstable in CRC cells, and this instability is mediated via the FOXA1 3 0 UTR, consistent with the known functions of the 3 0 UTR. 8e observed no significant change in FOXA1 mRNA expression following inhibition of the miRNA processing pathway through DICER1 knockdown, suggesting that miRNAs do not play a role in this process.Contrary to this, FOXA1 has previously been shown to be a direct target of both miR-3064-5p and miR-212 in hepatocellular carcinoma (HCC), 46,47 and a target of miR-760 in CRC. 48Our siRNA knockdowns of DICER1 suggest that FOXA1 mRNA is not regulated by miRNAs even though known miRNA targets such as MYC and ZEB1 were found to be upregulated.Alternatively, the CRC cell lines utilized in Cong et al. were poorly differentiated, 48 while our knockdowns of DICER1 were performed in well-differentiated CRC cells.Also, it may be the case that miRNAs upregulated in CRC may repress FOXA1 expression which can be explored in future studies.
The only protein enriched in our FOXA1 3 0 UTR RNA pulldown and mass spectrometry data that has a significant negative correlation with FOXA1 expression in TCGA COAD patient data was STAU1, an ortholog of Drosophila.STAU1 is an established RBP with four double-stranded RNA-binding domains (dsRBDs) that has been well-studied for its role in regulating mRNA decay in the pathway Staufen1-mediated mRNA decay (SMD). 30,31,3636]49 The FOXA1 3 0 UTR does not contain Alu elements.While there are predicted RNA secondary structures that could serve as SBSs according to mRNA structure prediction software and there is STAU1 hiCLIP data, 34 we did not find binding of STAU1 to the FOXA1 mRNA in hiCLIP.This could be because we found that the cells from which STAU1 hiCLIP was performed have very low FOXA1 expression (data not shown).Due to these limitations, the specific sequence in the FOXA1 3 0 UTR where STAU1 binds in CRC cells remains unclear.In addition, in our study we may have missed RBPs that stabilize the FOXA1 mRNA by intersecting the mass spectrometry data from the FOXA1 3 0 UTR PDs with the genes positively correlated with FOXA1 expression in TCGA data.This analysis may discover FOXA1 stabilizing RBPs that are downregulated in CRC and target the FOXA1 3 0 UTR.
As there is approximately the same level of FOXA1 mRNA at 6 h ActD treatment upon STAU1 depletion as compared to samples receiving control siRNAs, it is possible there is some compensatory mechanism.The second ortholog of Staufen, STAU2, has structural similarities to STAU1 and may have similar functions, yet there is only 51% sequence identify between the two. 50In microarrays for a subset of the different STAU1 and STAU2 isoforms (STAU1, 48 STAU2, 51 and STAU2 52 ), there is a subset of cellular mRNAs that are present in all ribonucleoprotein complexes for each isoform. 50It has previously been suggested that there is some redundancy between the two paralogs, and STAU2 can compensate for STAU1 loss or even heterodimerize with STAU1 and bind to proteins involved in SMD in HeLa cells. 50,53Contrary to this, STAU2 expression is not affected by disruption of the STAU1 locus in Stau1 mutant mice, arguing against redundant mechanisms between the two. 51In general, STAU2 is primarily expressed in the brain, and has no significant correlation with FOXA1 expression in CRC (data not shown), suggesting that STAU2 is not compensating for a loss of STAU1 in CRC, and there is no redundancy between the two in the context of CRC.
Our study further demonstrates that STAU1 regulates the transcriptome not only through individual targets, but also through the regulation of TFs like FOXA1, which themselves regulate the expression of hundreds of genes.We find that upon STAU1 knockdown, there are significant increases in expression of large number of genes, a subset of which are FOXA1 targets.Many of these targets have been previously studied for their roles in promoting or preventing cancer pathogenesis.Furthermore, there are significant increases in other TFs upon STAU1 knockdown according to our RNA-seq data, including KLF4, CRIP12, and FOXP2, another member of the FOX family (data not shown).Each of these TFs has been studied in cancer and they inhibit cell proliferation, angiogenesis, or cancer stem cell fates, respectively. 52,54,55Other RBPs have been found to similarly regulate the transcriptome in both healthy and disease states via binding to and regulating the expression of TFs.The RBP HuD, which is implicated in neuronal development and disease, [56][57][58] promotes normal neuronal differentiation of neural stem cells by stabilizing the mRNA of the TF STAB1. 59In the gastrointestinal tract, the RBP RBMS3 regulates the expression of TF Prx1, which plays critical roles in the activation of hepatic stellate cells (HSCs), the mesenchymal cells of the liver, and consequently in the development of liver fibrosis. 60These RBPs serve as examples of those that can enhance or repress the expression of TFs, and ultimately indirectly enhance or repress the regulation of subsets of the transcriptome, with consequences on human health and disease progression.
In summary, the establishment of the post-transcriptional mechanism by which FOXA1 expression is regulated in CRC, together with other studies, stresses the importance of understanding the complex mechanisms governing the expression of genes with critical biological functions.Furthermore, we establish that the STAU1-regulated transcriptome in CRC is not limited to its direct binding targets, but STAU1 also regulates the transcriptome through the direct regulation of TF expression.Elucidating the interactions between TFs, their targets, and the elements that regulate them is essential to our understanding of the processes driving human health and disease.Future studies can use the foundation of data here to determine the physiological effects of STAU1 regulation on the FOXA1-regulated transcriptome in CRC and reveal further STAU1 or FOXA1 regulatory networks that target genes with cancer-related functions.

RNA extraction and RT-qPCR
Total RNA was extracted with TRIzol reagent (Invitrogen) according to the manufacturer's instructions.Five hundred nanograms of extracted RNA was used for reverse transcription using iScript TM Reverse Transcription Supermix (Bio-Rad).All quantitative RT-qPCR reactions were carried out on StepOnePlus real-time PCR System (Applied Biosystems) using FastStart SYBR green Master Mix (Millipore Sigma).Expression was normalized using GAPDH mRNA, and relative RNA expression was calculated using the 2

siRNA transfections
All gene knockdown experiments were conducted using a SMARTpool of four siRNAs.STAU1 SMARTpool siRNAs, FOXA1 SMARTpool siRNAs, and DICER1 SMARTpool siRNAs were purchased from Horizon Discovery; AllStars Negative Control siRNA was purchased from Qiagen.Cells were reverse transfected with a final siRNA concentration of 20 nM using Lipofectamine RNAiMax (Invitrogen) in Opti-MEM I Reduced Serum Medium (Gibco) according to the manufacturer's protocol.After 48 or 72 h transfection, cells were harvested for RNA extraction and RT-qPCR was performed as described above, or for whole cell lysates were prepared for immunoblotting.

Dual luciferase reporter assays
Human and mouse FOXA1 3 0 UTR inserts were generated with gBlocks purchased from IDT.The inserts were cloned into the 3 0 UTR of Renilla luciferase in the psiCHECK-2 dual luciferase vector (Promega) using the restriction enzymes XhoI and NotI (New England Biolabs).Ligation was performed with T4 DNA Ligase (New England Biolabs) and ligation products were transformed into DH5a competent E. coli (Thermo Fisher Scientific) then incubated overnight at 37 � C on LB agar plates containing 100 mg/mL ampicillin.Colonies were inoculated overnight at 37 � C in liquid cultures containing 100 mg/mL ampicillin, then plasmid DNA was isolated using the MonarchV R Plasmid Miniprep Kit (New England Biolabs).For luciferase assays, cells were forward transfected with 100 ng plasmid DNA using Lipofectamine 2000 (Life Technologies Invitrogen) in Opti-MEM I Reduced Serum Medium (Gibco) according to the manufacturer's protocol.At 48 h post-transfection, luciferase activity was measured using the dual luciferase system (Promega).psiCHECK-2 empty vector served as a negative control, and psiCHECK-2-FORCP-3 0 UTR served as a positive control as previously described. 27

Generation of stable cell lines
FOXA1 ORF stable expressing HCT116 cells were previously generated. 17For FOXA1 ORF þ 3 0 UTR stable expression HCT116 cells, the FOXA1 3 0 UTR was PCR amplified from the psiCHECK-2-FOXA1 3 0 UTR vector using the following primers: ATATATCTAGACTCCCGGGACTG and ATATATCTAGATTTTGTT AACATTTGATTT.PCR products and the pLVX-FOXA1-ORF vector were digested using the restriction enzyme XbaI (New England Biolabs).Digested FOXA1 3 0 UTR insert was ligated to digested pLVX-FOXA1-ORF vector using T4 DNA ligase, then transformed, inoculated, and isolated from bacterial culture as described above.Then 293 T cells were forward transfected with 1200 ng pLVX-FOXA1-ORF along with lentiviral packaging vectors using Lipofectamine 2000 (Life Technologies Invitrogen), according to manufacturer instructions.Medium containing packaged viral particles was collected and replenished at 48, 56, and 72 h post-transfection and stored at -80 � C until further use.Virus titer was determined by serial dilution method, and a multiplicity of infection (MOI) of �0.5 was used to generate stable cell lines.Selection was performed using 1 mg/mL puromycin.HCT116 cells stably expressing FOXA-FLAG (for Supplementary Figure S3H) were generated as previously described. 17

Biotinylated RNA pulldowns and mass spectrometry analysis
For the biotinylated RNA pulldowns, the FOXA1 3 0 UTR was amplified from psiCHECK-2-FOXA1-3 0 UTR using primers containing the T7 promoter sequence: TAATACGACTCACT ATAGGGCTCCCGGGACTGGGGGGTTT and TTGGACACAACTTA ATTCTA at the 5 0 end.Control luciferase plasmid was linearized prior to in vitro transcription using XbaI restriction enzyme (New England Biolabs).The PCR amplified or linearized DNA was used as a template for the MEGAscript TM T7/SP6 Transcription Kit (Thermo Fisher Scientific) using Biotin-16-UTP (Roche).In vitro transcribed RNA was treated with Dnase I for 15 min at 37 � C then purified using Thermo Scientific Spin Columns.RNA concentration was measured using a Nanodrop, and RNA quality was checked on a Bioanalyzer.Approximately 30-50 million cells were harvested and snap-frozen at -80 � C, then resuspended in RIPA buffer (Thermo Fisher Scientific) with 1� protease inhibitor cocktail (Roche) and 0.1 U/mL RNaseOUT TM Recombinant Ribonuclease Inhibitor (Invitrogen) added.To prepare lysate, resuspended cells were sonicated on ice and centrifuged at full speed at 4 � C for 15 min; supernatant was transferred to fresh Eppendorf microcentrifuge tubes.Dynabeads TM M-280 Streptavidin (Thermo Fisher Scientific) were conjugated with 10-30 pmol purified biotinylated RNA and pulled down using a magnetic separation stand (Promega).Lysates were pre-cleared by rotation in pre-washed Dynabeads TM M-280 Streptavidin, then split in half and added to the RNAconjugated Dynabeads TM , rotating overnight at 4 � C. Proteins were eluted by boiling the washed beads in Laemmli Sample Buffer (Bio-Rad) at 95 � C for 5 min.Ten percent of eluate was saved for immunoblotting, the rest was subjected to mass spectrometry analysis as previously described. 63ulldowns of endogenous RNA via biotinylated antisense oligos (ASOs) were performed using the protocol outlined in Yu et al. 33 with the following modifications: (1) ASOs were diluted to a stock concentration of 200 mM, and 1 mL was used per reaction, (2) immunoblotting was performed from input lysate and eluate.

Ribonucleoprotein immunoprecipitation
For RNP-IPs, 10 mg of anti-STAU1 antibody (Abcam), or anti-RBM47 antibody (Millipore Sigma-Aldrich) was incubated with 30 mL of Pierce TM Protein A/G magnetic beads (Thermo Fisher Scientific) overnight at 4 � C, then washed twice with NT2 buffer (50 mM Tris pH 7.4, 150 mM NaCl, 1 mM MgCl 2 , 0.05% NP-40).For cell lysate preparation, 10 cm dishes of LS180 cells at �70% confluency were harvested via trypsinization, then moved to 1.5 mL Eppendorf microcentrifuge tubes.The cells were spun down and supernatant was removed.Cell pellets were resuspended in Polysome Extraction Buffer (10 mM Tris-HCL pH 7.5, 100 mM KCl, 5 mM MgCl 2 , 0.5% NP-40, and 1� protease inhibitor cocktail) with 40 U/mL RNaseOUT TM Recombinant Ribonuclease Inhibitor (Thermo Fisher Scientific) and 1 mM DTT added, then incubated on ice for 10 min.Lysates were spun down for 10 min at 4 � C, and 5% of the supernatant was kept for protein input while 5% was kept for RNA input.Remaining lysates were split between the samples and added to antibody-conjugated beads, rotating for 4 h at 4 � C, then washed four times with NT2 buffer.Beads were incubated at 37 � C for 10-15 min with Dnase I (Thermo Fisher Scientific), then washed with NT2 buffer.Ten percent of the sample was removed to check protein efficiency, then the beads were incubated for 15 min with 10% SDS and Proteinase K at 55 � C with shaking.RNA was isolated from the supernatant of these samples as well as from the RNA input sample using Phenol:Chloroform (Ambion), with overnight precipitation in 2.5 volumes EtOH and 0.1 volumes of 3 M Sodium Acetate.RT-qPCR and immunoblotting were performed on these samples as described above.

RNA-seq and ChIP-seq analysis
For genome-wide stability assays, LS180 and HCT116 cells were treated with ActD in biological triplicate for 0, 2, and 4 h.Total RNA was isolated using the Rneasy Plus Kit (Qiagen).Samples were sent to the CCR Sequencing Facility (Frederick, MD) for poly (A) RNA sequencing on a NovaSeq S1 sequencer.Reads were mapped to the hg19 genome with the Gencode v19 annotation using the STAR aligner v2.7.0f, 64 followed by quantification of raw read counts using RSEM v1.3.1. 65Differential expression analysis was performed by the Developmental Therapeutics Branch, National Cancer Institute (NCI) using the DESeq2 package in R. 66 For the STAU1 knockdown RNA-seq samples in LS180, transient transfection with siRNAs against STAU1 or control siRNAs was performed in biological duplicate as described above.Cells were harvested and total RNA was isolated using the Rneasy Plus Kit (Qiagen).Samples were sent to the CCR Sequencing Facility (Frederick, MD) for poly (A) RNA sequencing on a NextSeq 2000 P2 sequencer.Sample reads were aligned with the hg38 reference genome with Gencode v30 annotation using the STAR aligner v2.7.0f, 64 followed by quantification of raw read counts using RSEM v1.3.1. 65ifferential gene expression analysis was performed by the Genetics Branch Omics Bioinformatics Facility, NCI using DESeq2. 66or the FOXA1 knockdown RNA-seq data and FOXA1 ChIP-seq data, we utilized our previously analyzed and published data. 17

TCGA gene expression data
TCGA gene expression data for correlation analysis in Figure 2 was accessed via the publicly available cBioPortal database. 67FOXA1 was queried for mRNA expression z-scores relative to all samples in TCGA PanCancer Atlas COAD dataset.TCGA gene expression and pathology data for survival analysis in Supplementary Figure S2 was accessed via the publicly available Human Protein Atlas database.

Quantitation and statistical analysis
All data were plotted in GraphPad Prism (v8), unless otherwise noted.Error bars represent standard deviation, and statistical analysis was performed on biological triplicates using the Student's t-test.

Figure 1 .
Figure 1.FOXA1 mRNA is unstable in CRC cells and its instability is mediated via its 3 0 UTR.(A) Schematic showing the experimental procedure for identifying unstable RNAs.(B and C) Snapshots from Integrated Genome Viewer (IGV) showing RNA-seq read coverage of the FOXA1 locus in HCT116 (B) or LS180 (C) cells following treatment with ActD for 0, 2 and 4 h.(D-F) RNA stability assays were performed for FOXA1 mRNA by measuring its levels by RT-qPCR (normalized to GAPDH mRNA) following ActD treatment at the indicated time points in LS180 cells (D), CRC patient-derived organoid 1 (E), and CRC patient-derived organoid 2 (F).The half-life of FOXA1 mRNA is indicated as t 1/2 .MYC mRNA serves as a positive control for unstable mRNA.(G) RNA stability assays were performed for exogenous FOXA1 open reading frame (ORF) or FOXA1 ORF þ 3 0 UTR following stable overexpression in HCT116 cells.Half-life of FOXA1 ORF is �5 h whereas that of FOXA1 ORF þ 3 0 UTR is �2 h.The RT-qPCR was normalized to GAPDH mRNA.(H) The human or mouse FOXA1 3 0 UTR was cloned downstream of the Renilla luciferase gene in psiCHECK2, and dual luciferase assays were performed following transfections for 48 h in LS180 cells.psiCHECK-2 empty vector served as a negative control, and psiCHECK-2-FORCP-3 0 UTR served as a positive control.��� p < 0.001, ���� p < 0.0001.

Figure 2 .
Figure 2. STAU1 interacts with FOXA1 mRNA.(A) Schematic showing the experimental procedure for in vitro biotinylated RNA pulldowns followed by mass spectrometry.IVT refers to in vitro transcription.(B) Venn diagram showing the intersection of proteins bound to the FOXA1 3 0 UTR in three replicates of biotinylated RNA pulldowns and RBPs that have a significant negative correlation with FOXA1 expression in TCGA COAD data (Spearman correlation < −0.3, adj.p < 0.05).(C)Immunoblotting was performed to verify STAU1 binding to the biotinylated FOXA1-3 0 UTR following RNA pulldowns.Biotinylated luciferase RNA and RBM47 served as a negative controls.(D-F) Endogenous FOXA1 mRNA was pulled down from LS180 whole cell lysates using four biotinylated antisense oligos (ASOs) spanning the FOXA1 mRNA (D), and successful pulldown of the FOXA1 mRNA was confirmed by RT-qPCR (E).ACTB mRNA was used as a negative control.(F) Immunoblot shows that STAU1 is pulled down using three out of four ASOs but absent in the beads only control.RBM47 served as a negative control.(G) Immunoblotting was performed following IP from LS180 cell lysates using an IgG or anti-STAU1 Ab. (H) Fold enrichment of specific mRNAs in STAU1 RNP IPs was determined by RT-qPCR from RNA IPs from LS180 cell lysates.RBM47 IPs served as a negative control.ARF1 and SERPINE1 served as positive controls for STAU1 targets whereas MALAT1 served as a negative control.

Figure 3 .
Figure 3. STAU1 knockdown increases FOXA1 mRNA stability and inhibits FOXA1 3 0 UTR-mediated repression.(A-D) STAU1 knockdowns were performed by transiently transfecting LS180 (A, B), SW1222 (C, D), and DLD1 cells (D) with siRNAs against STAU1 or control siRNA for 72 hr.(A, C) RT-qPCR was performed for STAU1 and FOXA1 mRNAs after siRNA knockdown of STAU1.SDHA mRNA served as a negative control in panel A. (B, D) Immunoblotting from whole cell lysates is shown for FOXA1 and STAU1 after siRNA-mediated knockdown of STAU1.GAPDH served as loading control.(E) RNA stability assays were performed for FOXA1 mRNA (normalized to GAPDH mRNA) following transient transfection of LS180 cells with siRNAs against STAU1 or control siRNA for 72 h.The half-life of FOXA1 mRNA in siCTRL samples was �1 h and �3 h in siSTAU1 samples.(F) FOXA1 3 0 UTR was cloned downstream of the luciferase gene in the psiCHECK2, and dual luciferase assays were performed following cotransfection of the cells with psiCHECK2 empty vector (EV) or psiCHECK2-FOXA1 3 0 UTR and siRNAs against STAU1 or control siRNA for 72 h.(G, H) Correlation between FOXA1 and STAU1 mRNA expression in CRC patient tumor tissues (G) or normal human colon tissues (H) from the TCGA COAD cohort.� p < 0.05, �� p < 0.01, ��� p < 0.005, ���� p < 0.0001.

Figure 4 .
Figure 4.The STAU1-repressed transcriptome is enriched for a subset of FOXA1 targets.mRNA-seq was performed from LS180 cells transfected with siRNAs against STAU1 or control siRNA for 72 h.(A) IGV snapshot showing read coverage at the STAU1 locus following siRNA knockdown.(B) Venn diagram showing the intersection between genes significantly upregulated (log2FC > 0.6, adj.p < 0.05) in mRNA-seq from STAU1 knockdown LS180 cells, genes significantly downregulated (log2FC < −0.6, adj.p < 0.05) in mRNA-seq from FOXA1 knockdown LS180 cells, and genes bound by FOXA1 in the FOXA1 ChIP-seq from LS180 cells.(C, D) IGV snapshots for the three FOXA1 target genes ISX, TFF1, and CCN2 (that were in the intersection of the dataset shown in panel B) from the STAU1 knockdown mRNAseq (C) or FOXA1 ChIP-seq (D).(E) Increased expression of the three FOXA1 target genes identified in (B-D) was validated by RT-qPCR in LS180, DLD1 and SW1222 cells.

Figure 5 .
Figure 5. STAU1 represses the FOXA1-regulated transcriptome by binding to the FOXA1 3 0 UTR.Model showing that STAU1 binds to FOXA1 mRNA at the 3 0 UTR to inhibit FOXA1 expression, which in turn downregulates FOXA1-target gene expression.
All RNA-seq data are under the Data Availability Statement.