Multiplexed screening reveals how cancer-specific alternative polyadenylation shapes tumor growth in vivo

Gabel, Austin M.; Belleville, Andrea E.; Thomas, James D.; McKellar, Siegen A.; Nicholas, Taylor R.; Banjo, Toshihiro; Crosse, Edie I.; Bradley, Robert K.

doi:10.1038/s41467-024-44931-x

Download PDF

Article
Open access
Published: 01 February 2024

Multiplexed screening reveals how cancer-specific alternative polyadenylation shapes tumor growth in vivo

Nature Communications volume 15, Article number: 959 (2024) Cite this article

1974 Accesses
1 Citations
8 Altmetric
Metrics details

Subjects

Abstract

Alternative polyadenylation (APA) is strikingly dysregulated in many cancers. Although global APA dysregulation is frequently associated with poor prognosis, the importance of most individual APA events is controversial simply because few have been functionally studied. Here, we address this gap by developing a CRISPR-Cas9-based screen to manipulate endogenous polyadenylation and systematically quantify how APA events contribute to tumor growth in vivo. Our screen reveals individual APA events that control mouse melanoma growth in an immunocompetent host, with concordant associations in clinical human cancer. For example, forced Atg7 3′ UTR lengthening in mouse melanoma suppresses ATG7 protein levels, slows tumor growth, and improves host survival; similarly, in clinical human melanoma, a long ATG7 3′ UTR is associated with significantly prolonged patient survival. Overall, our study provides an easily adaptable means to functionally dissect APA in physiological systems and directly quantifies the contributions of recurrent APA events to tumorigenic phenotypes.

Cleavage and polyadenylation machinery as a novel targetable vulnerability for human cancer

Article 17 April 2024

The transcriptional terminator XRN2 and the RNA-binding protein Sam68 link alternative polyadenylation to cell cycle progression in prostate cancer

Article 07 November 2022

Multimodal perturbation analyses of cyclin-dependent kinases reveal a network of synthetic lethalities associated with cell-cycle regulation and transcriptional regulation

Article Open access 11 May 2023

Introduction

Alternative cleavage and polyadenylation (APA)—the process whereby a pre-mRNA can be cleaved and have a poly(A) tail added at multiple distinct locations, leading to expression of mRNAs with different 3′ untranslated regions (UTRs)—affects most human genes¹. Although the biological roles of most APA events are unknown, detailed studies have revealed that APA can affect mRNA levels, localization, and translation, among other molecular phenotypes^2,3,4. Poly(A) site selection is coordinated by differential interactions between multiple factors, including several RNA-binding proteins and the U1 small nuclear RNA (snRNA). RNA-binding proteins and the U1 snRNA bind RNAs in a sequence-specific manner near the proximal poly(A) site and can either promote or inhibit assembly of the 3′ end processing complex, although this process is still not entirely understood^5,6,7,8. Differential polyadenylation site usage is commonly observed in many biological processes, frequently in a biased manner indicative of coordinated regulatory changes^9,10,11. For example, rapidly dividing cells frequently utilize gene-proximal over gene-distal poly(A) sites, and thus express mRNAs with correspondingly shorter 3′ UTRs, compared to non-dividing, terminally differentiated cells. mRNAs are differentially polyadenylated throughout development, where terminally differentiated cells tend to utilize more distal poly(A) sites³; immune cell subsets including monocytes, T cells, and B cells undergo global 3′ UTR shortening when stimulated by their respective chemokines to begin dividing¹²; many cancers express mRNAs with markedly shorter 3′ UTRs than do peritumoral, healthy tissues^13,14,15.

It is well-established that APA is recurrently dysregulated in virtually all cancer types, yet it remains unknown whether most APA dysregulation directly promotes cancer progression or instead simply a downstream consequence of rapid cell division^{16,17,18,19,20}. Consistent with a functional role for APA dysregulation in cancers, several studies have reported that transcriptome-wide APA dysregulation is associated with patient prognosis in a subset of tumor types and that overexpression of a specific mRNA with a short 3′ UTR, but not a long 3′ UTR, can promote enhanced cancer cell growth^21,22. Nonetheless, the potential functional relevance and roles of cancer-associated APA remains unclear. The vast majority of cancer-associated APA events have never been functionally studied. Most functional studies of APA have relied on transgenic expression, rather than manipulation of APA in an endogenous context, and no study has yet manipulated endogenous, cancer-associated APA events in the physiological context of tumorigenesis in vivo.

Here, we systematically identify cancer-associated APA dysregulation that plays a causative, rather than simply correlative, role in tumorigenesis by functionally studying endogenous APA site selection in the physiologically relevant setting of tumorigenesis in vivo. To do so, we developed a multiplexed, CRISPR-Cas9-based screen that permitted us to perform parallelized interrogation of individual APA events during tumorigenesis in vivo. This high-throughput functional screen revealed APA events that control key tumor phenotypes in an immunocompetent mouse host and have concordant associations in clinical human cancer.

Results

Global 3′ UTR lengthening predicts poor patient prognosis in human melanoma

To assess whether cancer-associated changes in APA are relevant to clinical phenotypes, we tested whether global APA dysregulation was significantly associated with patient outcomes. Several prior studies have analyzed RNA sequencing data from The Cancer Genome Atlas (TCGA) and observed significant associations between global APA dysregulation and patient survival; however, those studies limited analyses to the 17 tumor types for which there were sufficient patient-matched, peritumoral normal tissue samples available^13,15. In order to extend such analyses to all tumor types, including those for which peritumoral samples are not available, we instead took a stratification-based approach that relied on data from tumor samples alone. For each cancer subtype, we stratified patients into terciles representing whether their tumor transcriptomes preferentially expressed short, medium, or long 3′ UTRs by computing a median 3′ UTR length for each tumor across 7513 genes that are subject to alternative polyadenylation (Fig. 1A–H; Supplementary Fig. 1A–C). This measure of global 3′ UTR length was significantly positively correlated with PABPN1 expression in 28 of 30 cancer subtypes and significantly negatively correlated with CSTF2 expression in 21 of 30 cancer subtypes, consistent with PABPN1 and CSTF2’s known roles as a repressor or activator of proximal poly(A) site usage, respectively (Supplementary Fig. 1D–I)^5,6,7.

**Fig. 1: Global 3′ UTR length correlates with clinical outcomes across tumor types.**

We then performed Kaplan–Meier survival analysis by comparing patients with short or long median 3′ UTRs and found significant survival differences that varied substantially by cancer type (Fig. 1I; Supplementary Fig. 2A–D; Supplementary Data File 1). In ovarian carcinoma, renal cell carcinoma, breast carcinoma, colon adenocarcinoma, and lung adenocarcinoma, patients whose cancers expressed shorter 3′ UTRs exhibited significantly worse overall survival (Fig. 1J; Supplementary Fig. 2D). Several cancer subtypes displayed the opposite trend, where patients with globally lengthened 3′ UTRs exhibited significantly worse overall survival, including head and neck squamous cell carcinoma, low grade glioma, and cutaneous melanoma (Fig. 1I, J; Supplementary Fig. 2D). These variable associations illustrate the potentially clinically relevant roles of APA and highlight the likely complexity of the relationship between 3′ UTR length trends and prognosis.

These results are necessarily based upon computational inference of 3′ UTR length using poly(A)-selected RNA-seq data, rather than assays like 3′-seq that provide nucleotide-level resolution of 3′ UTR length, due to the nature of most patient transcriptomic data. We therefore confirmed that our results were agnostic to the specific computational algorithm used to infer 3′ UTR usage by measuring 3′ UTR length using a distinct method, APAlyzer²³. We repeated our analyses for the cutaneous melanoma cohort, which demonstrated the most significant associations between 3′ UTR length and patient survival among TCGA cohorts, and found concordant and similarly strong associations between 3′ UTR length and survival (Supplementary Fig. 2E, F; Supplementary Data File 2). Together, these analyses highlight a subset of cancer types where global APA is strongly correlated with patient survival, suggesting possible functional relationships between APA and tumor growth.

A mouse model of melanoma exhibits cancer-associated aberrant alternative polyadenylation

Motivated by these strong associations between APA and patient prognosis, we sought to experimentally test the hypothesis that APA dysregulation affecting specific genes functionally contributes to tumorigenesis. As clinical melanoma exhibited the strongest correlations between 3′ UTR length and patient survival (Fig. 1I, J), we turned to syngeneic mouse models of untransformed melanocytes and melanoma to functionally study APA in biologically relevant and experimentally tractable systems. B16-F10 cells are mouse melanoma cells that readily engraft and form aggressive tumors in immunocompetent C57BL/6 mice, while Melan-A cells are immortalized but non-tumorigenic melanocytes derived from the same C57BL/6 background (Fig. 2A)^24,25. These two models therefore provided an opportunity to identify melanoma-associated APA by comparing immortalized melanocytes and melanoma cells in a consistent genetic background context.

**Fig. 2: Identification of differentially polyadenylated RNAs in a murine model of melanoma.**

We profiled global gene expression and poly(A) site selection in each of the two syngeneic cell lines with high-replicate RNA-seq and Poly(A)-seq (Fig. 2B). We identified significant differences in APA site selection between the two cell lines (n = 6 replicates per cell line) with APAlyzer²³. This analysis revealed 654 significantly differentially polyadenylated transcripts for which differential APA site selection was supported by both Poly(A)-seq and RNA-seq (Fig. 2C–F; Supplementary Data File 3). 204 genes displayed strong correlations between APA and differential gene expression, but the majority did not, consistent with the idea that many APA events have gene- or protein-specific effects rather than simply modulating transcript levels in an easily predictable manner (Fig. 2G)⁴.

We next compared APA in mouse and human melanoma. For each gene, we calculated the relative distal poly(A) site use in the B16-F10 mouse melanoma model system and for the human gene ortholog in the 424 TCGA melanoma samples and then computed the Pearson correlation for those gene-level values. This analysis revealed a modest but significant positive correlation between APA in mouse and human melanoma (R = 0.24, p < 2.2 × 10⁻¹⁶; Supplementary Fig. 3A–C). Overall, these data demonstrate that widespread alterations in APA characterize the B16-F10 model of mouse melanoma, suggesting that it may be a useful system for dissecting the functional contributions of individual APA events to tumorigenesis.

CRISPR-Cas9 paired-guide RNAs enable functional manipulation of alternative polyadenylation

In order to identify possible functional roles for the APA events that we identified in melanoma, we sought to manipulate endogenous poly(A) site selection in order to test how each APA event influenced tumor growth. We utilized a well-established approach for endogenous poly(A) site manipulation, wherein CRISPR-Cas9 paired-guide RNAs (pgRNAs) are used to precisely delete an individual poly(A) site. This approach has been previously utilized by other groups to force usage of specific poly(A) sites in Trp53, Calm1, and other genes^26,27,28,29.

We tested the efficacy of this approach in our melanoma model system using Sap30l, which has two distinct poly(A) sites separated by several kilobases. B16-F10 cells expressed SAP30L mRNA overwhelmingly utilizing the proximal poly(A) site, whereas Melan-A cells utilize the distal poly(A) site significantly more (Fig. 2C, F). We generated constitutive Cas9-expressing B16-F10 cells that we then transduced with two distinct pgRNAs designed to delete the proximal polyadenylation signal sequence within the Sap30l 3′ UTR (Fig. 3A, B). Cas9-expressing B16-F10 cells treated with either pgRNA designed to knock out the proximal polyadenylation signal sequence of Sap30l, referred to as proximal knockouts (pKO), exhibited appreciable levels of genomic DNA excision at the expected site (Fig. 3C; Supplementary Fig. 4A, B). Poly(A)-seq of polyclonal cells treated with either of two distinct Sap30l pKO pgRNAs caused specific increases in use of the Sap30l distal poly(A) site (Fig. 3D; Supplementary Fig. 4C). We generated seven monoclonal lines with heterozygous excision of the proximal Sap30l polyadenylation signal, five monoclonal lines with homozygous excision, and six monoclonal negative control lines transduced with a non-targeting control (NTC) pgRNA (Supplementary Fig. 4D). Treatment with either of the two distinct Sap30l pKO pgRNAs significantly increased use of the distal polyadenylation signal sequence in both the polyclonal and monoclonal settings as measured by RT-PCR, where cells with homozygous deletion of the proximal poly(A) signal displayed the highest use of the distal poly(A) site (Fig. 3E; Supplementary Fig. 4E). Using qRT-PCR to measure SAP30L mRNA levels, we found that homozygous excision of the proximal poly(A) signal led to a median reduction in total SAP30L mRNA levels of 47.9% (Fig. 3F; p = 0.017 computed across the monoclonal isolates). Consistent with this, we observed a strong negative correlation between distal polyadenylation site usage and total mRNA abundance for Sap30l (Fig. 3G; p = 0.0084). We concluded that targeted deletion of proximal poly(A) sites using a CRISPR-Cas9 pgRNA approach can effectively force distal poly(A) site use in both polyclonal and monoclonal populations of B16-F10 cells.

**Fig. 3: A CRISPR-Cas9 paired-guide RNA strategy can force targeted distal poly(A) site usage.**

High-throughput functional screening identifies alternative polyadenylation that regulates murine melanoma growth in vitro and in vivo

To test if APA is functionally relevant in melanoma, we utilized the CRISPR-Cas9-based approach to manipulating endogenous poly(A) site selection to create a multiplexed genetic screen. We designed a custom pgRNA library targeting 143 proximal polyadenylation signal sequences that we identified as cancer-associated from our genomic analyses (Supplementary Fig. 5A; Supplementary Data File 4). Approximately 25% of these APA events were dysregulated in both human and mouse melanoma (36 targets), while 75% were specific to mouse melanoma (107 targets) (Supplementary Fig. 5B). We created downloadable BED files to visualize pgRNA library design alongside the Poly(A)-seq data from B16-F10 and Melan-A cells that we used to select targets (Supplementary Fig. 6A, B). The final library comprises 8–10 unique targeting pgRNAs per proximal poly(A) signal, 150 positive control pgRNAs designed to knock out 15 distinct genes including core essential genes and tumor suppressor genes as negative and positive controls, respectively, and 150 negative control pgRNAs targeting proximal polyadenylation signals in genes that are not expressed in B16-F10 cells. We included ample negative control pgRNAs in order to allow for accurate determination of empirical false discovery rates (FDRs) and robust statistical significance testing (Supplementary Fig. 5B)³⁰. The resulting targeting, positive control, and negative control pgRNAs each have similar distributions of on- and off-target scores relative to pgRNA libraries that we and others have previously used for functional interrogation of tumorigenesis (Supplementary Fig. 5C, D)^30,31. We then cloned this library of 1718 unique pgRNAs via oligo array cloning as previously described³² and confirmed excellent library diversity with next-generation sequencing of the final plasmid pool (Supplementary Fig. 5E, F).

We used this library to identify APA that influenced tumor growth by performing paired in vitro and in vivo screens in B16-F10 cells (Fig. 4A). In brief, we infected Cas9-expressing B16-F10 cells with the lentiviral library at an MOI of 0.2. 24 h later, we added 0.5 µg/mL puromycin, which we left for 2 days, and then cultured for one more day in puromycin-free media. We then split the cells into 8 replicates. For each replicate, we collected a day 0 fraction, plated 500,000 cells for culturing in vitro, or subcutaneously injected 500,000 cells into a mouse flank for tumor growth in vivo. After 20 days of monitored growth, cells and tumors were harvested, and total genomic DNA (gDNA) was extracted using established methods³³. gDNA from the plasmid pool, day 0 time points, and day 20 in vitro and in vivo time points was then PCR amplified and subjected to next-generation sequencing with ~2500-fold coverage per pgRNA.

**Fig. 4: A high-throughput functional CRISPR-Cas9 screen reveals APA events that influence melanoma growth.**

We performed enrichment analysis and FDR computation using our previously described statistical framework³⁰. In brief, we normalized the fold-changes relative to day 0 for a given replicate such that the median of all pgRNAs targeting poly(A) sites in unexpressed genes was equal to 1 and pooled computed fold-changes across all replicates for a given condition to maximize statistical power (Supplementary Data File 5). We then calculated a p value for significant enrichment/depletion associated with each poly(A) site target by comparing the distribution of fold-changes for all pgRNAs targeting that poly(A) site to the distribution of pgRNAs targeting poly(A) sites in unexpressed genes. We computed empirical FDRs for each targeted poly(A) site using a subsampling procedure across negative control pgRNAs targeting unexpressed genes (n = 10,000 samples).

Both positive and negative control pgRNAs performed as expected in vitro and in vivo (Fig. 4B–D; Supplementary Fig. 7A–C). pgRNAs designed to knock out genes that are essential for tumor growth exhibited marked depletion, while pgRNAs designed to delete proximal polyadenylation sites of unexpressed genes exhibited negligible enrichment or depletion, with 96.0% or 96.7% of pgRNAs falling within two standard deviations of the median fold-change for our in vitro and in vivo experiments, respectively (Fig. 4C, D; Supplementary Fig. 7A–E). In contrast, numerous pgRNAs targeting proximal polyadenylation sites of expressed genes exhibited significant enrichment or depletion, with deletion of some proximal polyadenylation sites associated with alterations in B16-F10 cell growth in vitro and in vivo of similar magnitudes to positive controls (Fig. 4C, D).

We next validated two particularly striking hits that emerged from our screen. pgRNAs targeting the proximal polyadenylation sites of Atg7 and Egln1 were strongly depleted and enriched, respectively, both in vitro and in vivo. We introduced individual pgRNAs targeting these polyadenylation sites into Cas9-expressing B16-F10 cells and systematically profiled polyadenylation site selection, protein levels, and effects on cell growth. All experiments were performed in comparison to Cas9-expressing B16-F10 cells transduced with a pgRNA targeting a poly(A) site in the unexpressed gene Crabp1, which was selected as a negative control because its normalized fold-change in the screen was close to the median of all control pgRNAs targeting poly(A) sites within unexpressed genes. These experiments confirmed that poly(A) site knock out (pKO) pgRNAs targeting Atg7 and Egln1 induced significant increases in use of the distal polyadenylation sites, as intended (Fig. 4E, H). pKO pgRNA treatment significantly reduced ATG7 protein levels but was associated with only modest alterations in EGLN1 protein levels (Fig. 4F, I). Introduction of pKO pgRNAs to force use of distal polyadenylation sites in Atg7 and Egln1 recapitulated the in vitro growth phenotypes observed in the screen format (Fig. 4G, J). Overall, these data demonstrate that multiplexed pgRNA screening effectively nominates specific APA events that influence melanoma cell growth both in vitro and in vivo.

A long Atg7 3′ UTR suppresses murine melanoma growth in vitro and in vivo

We selected Atg7 for additional studies due to its significant depletion both in vitro and in vivo. We performed a rescue experiment to confirm that reduced levels of the short Atg7 isoform were responsible for the reduced cellular growth that we observed in our Atg7 pKO B16-F10 cells (Fig. 4G) by expressing transgenic Atg7 mRNA in the Atg7 pKO background and measuring cell growth. We generated a cDNA construct harboring the Atg7 coding sequence and the endogenous short Atg7 3′ UTR sequence as well as 200 nucleotides downstream of the poly(A) signal sequence (Supplementary Fig. 8A). We additionally generated a construct with the Atg7 coding sequence and an SV40 poly(A) signal. Stable integration and expression of either construct in the Atg7 pKO background restored ATG7 protein levels and rescued the cellular growth phenotype observed in Atg7 pKO cells (Supplementary Fig. 8B, C). We observed rescue of cell growth for both constructs (with the endogenous short Atg7 3′ UTR and with the SV40 poly(A) signal), suggesting that decreased ATG7 protein levels, rather than low levels of the short Atg7 3′ UTR-containing mRNA itself, underlies the reduced growth characteristic of Atg7 pKO cells.

To further characterize differences between the Atg7 short and long 3′ UTR isoforms, we completed an Actinomycin D time course to assess mRNA stability of the respective mRNAs. We found that the percent of all Atg7 mRNA utilizing the long isoform decreased relative to the fraction of total Atg7 mRNA across the Actinomycin D treatment time course (Fig. 5A, B). These data demonstrate that the Atg7 mRNA utilizing the long 3′ UTR is less stable than the Atg7 mRNA utilizing the short 3′ UTR, likely explaining at least in part the reduced protein levels we observed in B16-F10 Atg7 pKO cells (Fig. 4F).

**Fig. 5: *Atg7* alternative poly(A) site selection alters melanoma cell growth in vitro and in vivo.**

We then sought to replicate the in vivo growth phenotype observed in the CRISPR screen for Atg7 pKO cells in focused assays. Cas9-expressing B16-F10 cells were treated with either a pKO pgRNA targeting Atg7 or a control pgRNA targeting a poly(A) site in an unexpressed gene, and then engrafted into C57BL/6 mice. Consistent with results from the multiplexed screen, Atg7 pKO tumors exhibited reduced growth in vivo, leading to significantly prolonged host survival using Kaplan–Meier survival analysis (p = 0.0053) (Fig. 5C; Supplementary Fig. 9A, B).

We next sought to understand the molecular consequences of Atg7 APA. Autophagy related protein 7, ATG7, is considered a critical protein for autophagosome formation and function, and loss of ATG7 protein inhibits autophagy³⁴. We therefore investigated the consequences of Atg7 APA for autophagy. Complete Atg7 knock out and consequent loss of ATG7 protein caused a significant increase in p62 protein levels, as expected (Supplementary Fig. 10A). In contrast, treating cells with an Atg7 pKO pgRNA to force distal polyadenylation site usage did not cause detectable deficits in autophagy at baseline or following exposure to serum and amino acid starvation, measured by both p62 accumulation and a flow cytometry-based LC3-GFP-mCherry reporter (Supplementary Fig. 10A–E)³⁵. These results are distinct from a previous study reporting that ATG7 3′ UTR lengthening was associated with autophagy inhibition in pro-B cells, a difference that may be attributable to cell type-specific effects or that study’s reliance on KD of all ATG7 isoforms for functional studies rather than specific manipulation of poly(A) site usage³⁶. Although forcing distal polyadenylation site usage reduced ATG7 protein levels by ~60% (Fig. 4F; Supplementary Fig. 10A), we detected none of the autophagy impairment that was readily apparent upon Atg7 KO, highlighting the utility of directly manipulating poly(A) site usage for functional inference.

We next performed immunohistochemistry (IHC) for the Ki67 proliferation marker on control and Atg7 pKO tumors. Although Atg7 pKO tumors grew slower (Fig. 5C; Supplementary Fig. 9A, B), they exhibited a significantly higher fraction of nuclei with moderate and strong Ki67 staining relative to control tumors (Fig. 5D, E; Supplementary Fig. 9C, D). The levels of Ki67 fluctuate significantly throughout the cell cycle, with RNA and protein levels rising through S phase and peaking in G2/M phase³⁷. We therefore performed cell cycle analysis with propidium iodide staining and found that Atg7 pKO increased the fraction of B16-F10 cells in S phase and G2/M phase. These altered cell cycle kinetics are consistent with the increased Ki67 staining intensity observed by tumor IHC and concordant with a previous study demonstrating that ATG7 KD by siRNA in human bladder cancer cells increased the fraction of cells in S phase and G2/M phase (Fig. 5F; Supplementary Fig. 9E, F)³⁸.

Atg7 undergoes APA in both mouse and human cells, raising the possibility that ATG7 3′ UTR length could influence human melanoma growth just as it does mouse melanoma. We first assessed mutation status of ATG7 across the 424 patients in TCGA’s melanoma cohort. No nonsense, frame-disrupting, or recurrent missense mutations were observed across this cohort, confirming that ATG7 is not a common mutational target in clinical melanoma (Fig. 5G). We then performed Kaplan–Meier survival analysis after stratifying these 424 cutaneous melanoma patients by ATG7 3′ UTR length. We observed that patients whose tumors expressed a longer ATG7 3′ UTR exhibited significantly better progression-free survival, consistent with our finding that forcing long Atg7 3′ UTR usage slowed mouse melanoma growth (Fig. 5H). Given this strong association between ATG7 APA and survival that we observed in both mouse and clinical human cancer, it is interesting to speculate that ATG7 APA provides a post-transcriptional mechanism by which cancers may regulate ATG7 protein levels to influence their growth even in the absence of ATG7 mutations.

Discussion

Dysregulated alternative polyadenylation is a pervasive feature of most cancers, and thousands of cancer-associated APA events have been identified. However, very few of these APA events have been functionally studied or linked to cancer phenotypes. Here, we describe a high-throughput, in vivo screening platform to interrogate the functional roles of individual APA events in cancer and identify APA isoforms which enhance or reduce tumor growth in an immunocompetent mouse model. Our study systematically reveals individual APA events that play a causative, rather than simply a correlative, role in influencing key cancer phenotypes.

We primarily focused on APA events whose modulation was associated with notable effects on tumor growth. However, it is interesting to note that most APA events queried in our screen were not associated with detectable alterations in tumor cell growth either in vitro or in vivo. Although we cannot rule out the possibility that such negative results arise from ineffective pgRNA targeting and APA modulation, this seems unlikely given the relatively high efficiency of APA modulation that we observed even in polyclonal settings for all APA events that we studied individually. These data therefore suggest that many APA events have modest or no effects on tumor growth kinetics in our system and therefore either alter other facets of tumorigenesis or are bystander events of low functional relevance. This dichotomy, where some APA events play critical functional roles and others do not, may help to resolve the ongoing controversy regarding the likely functional relevance of most APA^12,14. In either case, our data highlight the utility of high-throughput screens for identifying functionally relevant events for detailed study.

It is important to note that our in vivo studies focused on determining of how APA modulation altered tumor cell growth, which is only one of many important cancer phenotypes that influence patient prognosis. For example, overall survival is frequently strongly associated with metastasis, which our assays did not measure. A natural and important extension of the current study would be to determine how cancer-specific APA influenced tumor cell dissemination.

We utilized the CRISPR-Cas9 pgRNA system for endogenous APA modulation because it has been extensively tested and validated^26,27,28,29. Future studies of APA may take advantage of the diverse other genome editing technologies that are now amenable to high-throughput screens. For example, dead Cas9 (dCas9) targeted downstream of the poly(A) signal has been used to enhance proximal poly(A) site selection³⁹, providing a useful counterpart to the suppression of poly(A) site use via pgRNAs that we relied upon. A fusion of the RNA-targeting dead Cas13b with adenosine deaminase 2 (ADAR2) can specifically modify adenosine residues in RNA⁴⁰ and as such could potentially be used to disrupt the A-rich canonical polyadenylation signal sequences AATAAA and ATTAAA, thereby allowing for functional manipulation of poly(A) sites at the level of RNA, rather than DNA.

Although we focused on cancer in the current study, we speculate that high-throughput functional interrogation of APA will prove equally fruitful in other biological contexts. For example, embryonic development and tissue differentiation^41,42 display coordinated, temporal regulation of poly(A) site selection, while poly(A) site selection is globally dysregulated in many genetic diseases, including Oculopharyngeal Muscular Dystrophy and Duchenne Muscular Dystrophy^43,44. Targeted sequencing has cataloged thousands of context-specific APA alterations in these and other systems, the vast majority of which have never been functionally studied. High-throughput manipulation of endogenous poly(A) site selection offers an efficient way to bridge the gap between APA identification and functional understanding.

Methods

Animal use

All animal work and procedures were completed in accordance with the Guidelines for the Care and Use of Laboratory Animals and approved by the Fred Hutchinson Cancer Center Institutional Animal Care and Use Committee. Six-week-old C57BL/6 male mice were obtained from the Jackson Laboratory for all animal experiments listed here. Animals were housed for 1 week prior to initiation of any experiments in line with IACUC guidance.

Cell lines and culture

B16-F10 cells were obtained from ATCC (CRL-6475) and cultured per the manufacturer’s instructions. Cell lines tested negative for mycoplasma contamination. Cell lines were authenticated via RNA-seq expression profiling. B16-F10 cells were also spot checked as they express melanin and cell pellets are black (no other cell lines in use are black when pelleted). The Melan-A cell line was obtained from the Wellcome Trust Functional Genomics Cell Bank and cultured per their instructions²⁴. B16-F10 cells were infected at 0.3 MOI with Cas9 lentivirus (Addgene 52962-LV) and selected with 2 ug/mL Blasticidin for 7 days. Polyclonal cells were then single cell sorted, and individual clones were assessed for on-target editing efficiency, all subsequent CRISPR assays were completed with this B16-F10 Cas9 clone with the highest editing rate.

Poly(A)-Seq library preparation

RNA was isolated from cell pellets with the Direct-zol RNA MiniPrep kit (Zymo Research). Protocol was adapted from Derti et al.¹. In brief, 5 µg of RNA was heated at 65° for 2 min and then placed on ice. Poly(A) selected mRNA was isolated using DynaBeads (ThermoFisher 61006) per the manufacturer’s instructions. 1st strand cDNA synthesis using SuperScript IV Reverse Transcriptase (ThermoFisher 18090010) and the primer RKB4087 (a modified oligo d(TVN) primer) to generate cDNA reads which begin synthesis just before the poly(A) tail. This mixture was then purified using a 2x SPRI bead clean up and then followed by 2nd strand synthesis and then a 1x SPRI bead clean up. This cDNA library was then subject to PCR (NEBNext High Fidelity 2x Master Mix) and primers RKB4089 (universal primer) and RKB4090-RKB4101 (indexed primers). Specific primer sequences are listed in Supplementary Data 6. The final libraries generate a broad range of fragment sizes, as such we gel extracted libraries from a 2% agarose gel corresponding to 300–500 bp size. Following extraction and purification, library size was analyzed with a 4200 TapeStation System before sequencing.

RNA-seq library preparation

RNA was isolated from cell pellets with the Direct-zol RNA MiniPrep kit (Zymo Research). Poly(A)-selected, unstranded Illumina libraries were prepared following the TruSeq protocol per the manufacturer’s instructions. Library size and distribution was analyzed with a 4200 TapeStation System before sequencing on an Illumina HiSeq as 2 × 50 bp to obtain ~40 million reads per sample.

RNA-seq and Poly(A)-seq data analysis

RNA-seq was analyzed as previously described⁴⁵. RNA-seq reads were mapped to an annotated transcriptome created using Ensembl 71^45,46, UCSC knownGene⁴⁷ and Misov2.0⁴⁸ annotations using RSEM version 1.2.4⁴⁹ (modified to call Bowtie⁵⁰ with option ‘-v 2’). Unaligned reads were then mapped to the corresponding genome (hg19/GRCh37 assembly, mm10/GRCmc38 assembly) and a database containing all possible pairings of 5’ and 3′ splice sites per gene in our merged transcriptome annotation using TopHat version 20.8b⁵¹. Mapped reads were then merged and input into MISO v2.0. For TCGA studies, we analyzed 9045 available samples across 29 cancer types.

For Poly(A)-seq, data was mapped similarly but in a stranded fashion. BAM files were input into the APAlyzer package in R²³ and the gene level log₂(distal reads/proximal reads) was computed per sample for each respective strand.

Survival analyses

Cancer type abbreviations are the same as TCGA standards (https://gdc.cancer.gov/resources-tcga-users/tcga-code-tables/tcga-study-abbreviations). Survival analyses were completed with the Kaplan–Meier estimator and statistical test were performed with a logrank test (R package survival). Stratification per cancer subtype was completed for Fig. 1b, Supplementary Figs. 1A–C and 2A–F by computing the median 3′ UTR length per sample and dividing each cancer subtype into terciles (short, medium and long) and comparing the short (≤33%) vs long (≥66%) bins. Gene level 3′ UTR measurements were downloaded from a previously published database⁵² or computed using the APAlyzer package in R²³.

For Fig. 5F, patients from the TCGA SKCM cohort were stratified based on the ATG7 3 UTR length per sample, again into terciles and downstream analyses were performed identically as described above.

pgRNA library construction and cloning

Target genes selected from the Poly(A) sequencing and TCGA SKCM analysis were largely selected based on manual inspection of BAM coverage plot quality. For each gene, the most utilized APA sites within the 3′ UTR were identified from PolyA-DB, filtered based off number of tissues where a particular APA site is utilized and the fraction of samples it has been detected in²³. Of those Poly(A) sites, the most proximal site was selected for targeting. The 100 base-pairs up and downstream of the given poly(A) site were selected as the genomic window wherein all possible gRNAs were generated using Guidescan Version 1.0⁵³. The downloaded gRNAs were then identified as up or downstream of the poly(A) signal (PAS), and then the guides were combined in a pairwise fashion to generate all possible pgRNAs which disrupt the poly(A) signal sequence. For each target we excluded gRNAs with more than two two-nucleotide off-target sites, more than thirty three-nucleotide off-target sites or an on-target editing efficiency of less than thirty according to Rule Set 2 scoring⁵⁴. All input target sites were then filtered to only include targets with at least eight pgRNAs. For sites with more than ten possible pgRNAs, pgRNAs were ranked based first on targeting efficiency and then specificity to select the ten pgRNAs to be utilized in the library. This same process was used to generate all possible pgRNAs targeting highly utilized PAS sites of genes that are not expressed in B16-F10 or Melan-A cells. In this control library, 150 pgRNAs were selected randomly as controls to be included in the library, of note this random selection process was run several times until the distribution of control pgRNAs matched the true library in terms of efficiency and specificity. An additional 15 genes were included as growth controls based on previous literature, either core essential genes identified from DepMap or genes which promote more rapid B16-F10 cell growth in vivo identified by mining data generated from previous genome wide gene knock out CRISPR screens in B16-F10 cells⁵⁵.

gDNA PCR and on-target editing verification

gDNA was extracted using the DNeasy Blood and Tissue kit (Qiagen) per the manufacturer’s protocol. A window around the Sap30l proximal poly(A) site was selected to allow for simple detection of deletion by band separation on gel electrophoresis. PCR amplicons were then purified and submitted for AmpliconEZ sequencing (Azenta/Genewiz). Reads were trimmed, mapped and the fraction of reads with an intact or disrupted poly(A) signal sequence were identified using the CRISPresso2 software^56,57. Specific primer sequences are listed in Supplementary Data 6.

pgRNA screen

9 × 10⁷ Cas9 expressing B16-F10cells were infected at an MOI of 0.2. Twenty-four hours post infection, cells were then selected with 1 µg/mL Puromycin for 48 h. Remaining cells were then pooled and divided into eight replicates. 500,000 cells were isolated from each replicated and frozen as a time point 0. Of the remaining cells, 500,000 cells were then plated into 10 cm plates and another 500,000 cells were injected subcutaneously into C57BL/6 mice. Cells grown in culture were passed every 3–4 days, and tumors were monitored and then harvested when reaching 1.5 cm in any single dimension or after 20 days, whichever condition was met first. On the day a given tumor was harvested, digested and frozen, the corresponding cells growing in vitro were also frozen. gDNA was then extracted from all samples using previously published methods³³, and quantified via Nanodrop. 1.5 µg of each sample was then utilized as input for PCR to amplify out integrated pgRNA constructs as previously described^30,31. Final, purified libraries were then submitted for Next-Generation Sequencing using a custom approach performed as previously described.

pgRNA sequencing and data analysis

Analyses were performed identically as previously described³⁰. In brief, each read includes independent reads for gRNA1 and gRNA2 as well as an index read. Each gRNA was separately mapped to a database of pgRNAs using BowTie, and correct pairings were identified by ensuring gRNA1 and gRNA2 come from the same pgRNA, incorrect pairings were discarded. We generated a per pgRNA pseudocount which was added to the raw counts and served to regularize the fold-change computations so that they are proportional to the relative representation of each individual pgRNA in the library.

Fold-changes were then computed by comparing each time point to the counts at the day 0 time point. These fold-changes were then normalized so that the median of all pgRNAs targeting poly(A) sites in unexpressed genes (150 pgRNAs in total) was equal to 1. For each poly(A) site target or gene KO control target a p value was computed by comparing the distribution of normalized log₂ fold-changes per pgRNA (8–10 per target) to the distribution of all 150 pgRNAs targeting poly(A) sites in unexpressed genes using a two-sided Wilcoxon rank-sum test. FDRs were computed by generating a distribution of fake target p values by randomly subsampling 10 pgRNAs from our control pgRNAs 10,000 times to estimate the distribution of expected p values. We then estimated FDR per real poly(A) site or gene KO target via the cumulative distribution function of the distribution of the fake target p values.

In vitro validation studies

Cas9 expressing B16-F10 cells were grown in standard conditions and then transduced with lentivirus containing indicated pgRNAs and then selected in 1 µg/mL Puromycin for 72 h. For cell growth assays, 100,000 cells were plated into each well of a 24 well plate and 20% of the well was passaged every 2 days into a new 24 well plate. 100 µL of the remaining cell suspension was used as input for a CellTiter-Glo (ProMega Catalog Number G9242) assay performed per the manufacturer’s instructions.

In vivo validation studies

Cas9 expressing B16-F10 cells were grown in standard conditions and then transduced with lentivirus containing indicated pgRNAs and then selected in 1 µg/mL Puromycin for 72 h. 5 × 10⁶ cells were then injected subcutaneously into each flank of adult male C57BL/6 mice, and monitored using calipers. Animals were euthanized when a tumor reached 1.5 cm in any dimension in accordance with IACUC guidelines, and no tumors ever exceeded this threshold in any dimension. Tumor material was isolated either for an archived flash frozen sample, fixed in 10% formalin at room temperature for histology studies or placed in TRIzol reagent (ThermoFisher 15596018) for subsequent RNA isolation.

Immunohistochemistry

Tissues from tumors were processed, embedded and stained through the Fred Hutch Experimental Histopathology core. Mouse Ki-67 (CST Clone 12202 1:2000 dilution) staining was performed using a rabbit monoclonal antibody. Staining was performed with a BOND RX autostainer (Leica Biosystems) and images were then acquired with an Aperio ImageScope at 40x magnification (Leica Biosystems). Image analysis was completed using HALO Image Analysis software.

RT-PCR analyses

RNA was isolated using the Direct-zol RNA MiniPrep (Zymo Research). SuperScript IV reverse Transcriptase was used to synthesize cDNA per the manufacturer’s instructions, but using a specific oligo d(TVN) primer to amplify the DNA directly upstream of the poly(A) tail (ThermoFisher Scientific). Nested RT-PCR was performed with a universal poly(A) tail primer and two gene specific primers to amplify the 3′ UTR. The final product visualized using agarose gel electrophoresis and band intensity quantification was performed with FIJI/ImageJ and reported as the percent distal isoform (pdi) defined as the percent of total signal arising from bands corresponding to distal poly(A) site use. For the Actinomycin D treatment time course, cells were plated and treated with 1 µg/mL Actinomycin D (Sigma-Aldrich A1410-2MG) for the indicated time frame, whereafter cells were harvested and RNA was isolated as described above. Specific primer sequences are listed in Supplementary Data 6.

Western blotting

Total protein lysates were isolated in 1x RIPA buffer and quantified with the Pierce 660 nm Protein Assay Reagent. Total protein lysates were electrophoretically separated and then transferred onto a nitrocellulose membrane using the NuPAGE system (ThermoFisher Scientific). Each membrane was blocked for 1 h at room temperature and then probed with primary antibody diluted in a blocking buffer overnight at 4 °C. Atg7 (AbCam Ab133528, 1:1000), Egln1 (Cell-Signaling Technology D31E11, 1:1000), p62/Sqstm1 (AbCam Ab109012, 1:1000), and Alpha-tubulin (Sigma-Aldrich Clone DM1, 1:2000) primary antibodies were used. Anti-mouse or anti-rabbit IRDye (LI-COR Biosciences) secondary antibodies and the Odyssey CLx Imager (LI-COR Biosciences) were utilized for detection and imaging.

Fluorescent LC3 reporter assay

The FUW mCherry-GFP-LC3 reporter plasmid was obtained from Addgene (Plasmid #110060) which was utilized to generate lentivirus in 293 T cells. B16-F10 Cas9-expressing cells were then infected and grown for 48 h. Cells were then plated and treated with standard DMEM with 10% FBS or Hanks Balanced Salt Solution for the indicated time. Following the time course all cells were washed 3x with PBS and then resuspended in 100 uL of PBS and stained with LIVE/DEAD Fixable Violet (ThermoFisher L34955) stain per the manufacturer’s instructions.

Cells were then passed through a 40-micron filter to generate a single cell suspension and then run on a BD FACSCelesta. Single cells were gated for live cells and then the individual GFP and mCherry signal were measured in the FITC and CF594 channels, respectively. Data were analyzed in FlowJo v10.

Statistics and reproducibility

Statistical analyses were performed in the R with Bioconductor, and tables and plots were generated using dplyr⁵⁸ and ggplot2⁵⁹ packages. All cancer type abbreviations follow TCGA standards (https://gdc.cancer.gov/resources-tcga-users/tcga-code-tables/tcga-study-abbreviations). All box plots follow the same formatting, where the middle line indicates the median, the hinges represent the 25th and 75th percentile, notches indicate the 95% confidence interval, and whiskers represent the most extreme data point within 1.5x the interquartile range from hinge Sample sizes for in vivo validation experiments were based off power analysis to reliably detect a 25% reduction in tumor growth with greater than 90% confidence incorporating measured standard error rates from previous in vivo experiments³⁰.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Data availability

Our proximal-Poly(A) KO (pKO) library has been deposited with Addgene (Pooled Library #81543 [https://www.addgene.org/Robert_Bradley/]). RNA-seq data generated as part of this study has been deposited in the Gene Expression Omnibus (accession number GSE212278). Gene-level 3′ UTR measurements for all TCGA samples were downloaded from http://tc3a.org⁵¹. Source data are provided with this paper. Data for Figs. 1–2, 4–5 and Supplementary Figs. 1–3, 5, 7–8 are included in Supplementary Data 1–5 and the Source Data file. All remaining data can be found in the Article, Supplementary and Source Data files. Source data are provided with this paper.

References

Derti, A. et al. A quantitative atlas of polyadenylation in five mammals. Genome Res. 22, 1173–1183 (2012).
Article CAS PubMed PubMed Central Google Scholar
Berkovits, B. D. & Mayr, C. Alternative 3′ UTRs act as scaffolds to regulate membrane protein localization. Nature 522, 363–367 (2015).
Article ADS CAS PubMed PubMed Central Google Scholar
Ji, Z., Lee, J. Y., Pan, Z., Jiang, B. & Tian, B. Progressive lengthening of 3’ untranslated regions of mRNAs by alternative polyadenylation during mouse embryonic development. Proc. Natl. Acad. Sci. USA 106, 7028–7033 (2009).
Article ADS CAS PubMed PubMed Central Google Scholar
Spies, N., Burge, C. B. & Bartel, D. P. 3′ UTR-Isoform choice has limited influence on the stability and translational efficiency of most mRNAs in mouse fibroblasts. Genome Res. 23, 2078–2090 (2013).
Article CAS PubMed PubMed Central Google Scholar
Jenal, M. et al. The poly(A)-binding protein nuclear 1 suppresses alternative cleavage and polyadenylation sites. Cell 149, 538–553 (2012).
Article CAS PubMed Google Scholar
Li, W. et al. Systematic profiling of Poly(A)+ transcripts modulated by Core 3’ end processing and splicing factors reveals regulatory rules of alternative cleavage and polyadenylation. PLoS Genet. 11. https://doi.org/10.1371/journal.pgen.1005166 (2015).
Yao, C. et al. Transcriptome-wide analyses of CstF64-RNA interactions in global regulation of mRNA alternative polyadenylation. Proc. Natl. Acad. Sci. USA 109, 18773–18778 (2012).
Article ADS CAS PubMed PubMed Central Google Scholar
Berg, M. G. et al. U1 snRNP determines mRNA length and regulates isoform expression. Cell 150, 53–64 (2012).
Article CAS PubMed PubMed Central Google Scholar
Lianoglou, S., Garg, V., Yang, J. L., Leslie, C. S. & Mayr, C. Ubiquitously transcribed genes use alternative polyadenylation to achieve tissue-specific expression. Genes Dev. 27, 2380–2396 (2013).
Article CAS PubMed PubMed Central Google Scholar
Zhang, H., Lee, J. Y. & Tian, B. Biased alternative polyadenylation in human tissues. Genome Biol. 6, 1–13 (2005).
Google Scholar
Brumbaugh, J. et al. Nudt21 controls cell fate by connecting alternative polyadenylation to chromatin signaling. Cell 172, 106–120.e21 (2018).
Article CAS PubMed Google Scholar
Sandberg, R., Neilson, J. R., Sarma, A, Sharp, P. A. & Burge, C.B. Proliferating cells express mRNAs with shortened 3’ untranslated regions and fewer microRNA target sites. Science 320, 1643–1647 (2008).
Goering, R. et al. LABRAT reveals association of alternative polyadenylation with transcript localization, RNA binding protein expression, transcription speed, and cancer survival. BMC Genom. 22, 476 (2021).
Mayr, C. & Bartel, D. P. Widespread shortening of 3′UTRs by alternative cleavage and polyadenylation activates oncogenes in cancer cells. Cell 138, 673–684 (2009).
Article CAS PubMed PubMed Central Google Scholar
Xia, Z. et al. Dynamic analyses of alternative polyadenylation from RNA-seq reveal a 3’2-UTR landscape across seven tumour types. Nat. Commun. 5. https://doi.org/10.1038/ncomms6274 (2014).
Gruber, A. J. & Zavolan, M. Alternative cleavage and polyadenylation in health and disease. Nat. Rev. Genet. 20, 599–614 (2019).
Article CAS PubMed Google Scholar
Hu, C. et al. Construction of Two Alternative Polyadenylation Signatures to Predict the Prognosis of Sarcoma Patients. Front. Cell Dev. Biol. 9. https://doi.org/10.3389/fcell.2021.595331 (2021).
Xu, C. & Zhang, J. Alternative polyadenylation of mammalian transcripts is generally deleterious, not adaptive. Cell Systems 6, 734–742.e4 (2018).
Article CAS PubMed PubMed Central Google Scholar
Zhang, Y., Wang, Y., Li, C., & Jiang, T. Systemic analysis of the prognosis-associated alternative polyadenylation events in breast cancer. Front. Genet. 11. https://doi.org/10.3389/fgene.2020.590770 (2020).
Zhang, Y., Xu, Y., & Wang, Y. Alternative polyadenylation associated with prognosis and therapy in colorectal cancer. Sci. Rep. 12. https://doi.org/10.1038/s41598-022-11089-9 (2022).
Andres, S. F. et al. IMP1 3′ UTR shortening enhances metastatic burden in colorectal cancer. Carcinogenesis 40, 569–579 (2019).
Article CAS PubMed Google Scholar
Tan, S. et al. CPSF6 links alternative polyadenylation to metabolism adaption in hepatocellular carcinoma progression. J. Exp. Clin. Cancer Res. 40, https://doi.org/10.1186/s13046-021-01884-z (2021).
Wang, R. & Tian, B. APAlyzer: a bioinformatics package for analysis of alternative polyadenylation isoforms. Bioinformatics 36, 3907–3909 (2020).
Article CAS PubMed PubMed Central Google Scholar
Bennett, D. C., Cooper, P. J. & Hart, I. R. A line of non-tumorigenic mouse melanocytes, syngeneic with the B16 melanoma and requiring a tumour promoter for growth. Int. J. Cancer 39, 414–418 (1987).
Article CAS PubMed Google Scholar
Overwijk, W. W. & Restifo, N. P. B16 as a melanoma model for human. Curr. Protoc. Immunol. https://doi.org/10.1002/0471142735.im2001s39.B16 (2001).
Bae, B. et al. Elimination of Calm1 long 3′-UTR mRNA isoform by CRISPR-Cas9 gene editing impairs dorsal root ganglion development and hippocampal neuron activation in mice. https://doi.org/10.1261/rna (2020).
Mitschka, S. & Mayr, C. Endogenous p53 expression in human and mouse is not regulated by its 30utr. ELife, 10. https://doi.org/10.7554/eLife.65700 (2021).
Pereira-Castro, I. et al. MCL1 alternative polyadenylation is essential for cell survival and mitochondria morphology. Cell. Mol. Life Sci. 79. https://doi.org/10.1007/s00018-022-04172-x (2022).
Wang, Q. et al. Cell cycle regulation by alternative polyadenylation of CCND1. Sci. Rep. 8. https://doi.org/10.1038/s41598-018-25141-0 (2018).
Thomas, J. D. et al. RNA isoform screens uncover the essentiality and tumor-suppressor activity of ultraconserved poison exons. Nat. Genet. 52, 84–94 (2020).
Article CAS PubMed PubMed Central Google Scholar
Parrish, P. C. R. et al. Discovery of synthetic lethal and tumor suppressor paralog pairs in the human genome. Cell Rep. 36. https://doi.org/10.1016/j.celrep.2021.109597 (2021).
Gasperini, M. et al. CRISPR/Cas9-mediated scanning for regulatory elements required for HPRT1 expression via thousands of large, programmed genomic deletions. Am. J. Human Genet. 101, 192–205 (2017).
Article CAS Google Scholar
Chen, S. et al. Genome-wide CRISPR screen in a mouse model of tumor growth and metastasis. Cell 160, 1246–1260 (2015).
Article CAS PubMed PubMed Central Google Scholar
Komatsu, M. et al. Impairment of starvation-induced and constitutive autophagy in Atg7-deficient mice. J. Cell Biol. 169, 425–434 (2005).
Article CAS PubMed PubMed Central Google Scholar
Leeman, D. S. et al. Lysosome activation clears aggregates and enhances quiescent neural stem cell activation during aging. Science 359, 1277–1283 (2018).
Article ADS CAS PubMed PubMed Central Google Scholar
Park, S. M. et al. U2AF35(S34F) promotes transformation by directing aberrant ATG7 Pre-mRNA 3’ end formation. Mol. Cell 62, 479–490 (2016).
Article CAS PubMed PubMed Central Google Scholar
Sobecki, M. et al. Cell-cycle regulation accounts for variability in Ki-67 expression levels. Cancer Res. 77, 2722–2734 (2017).
Article CAS PubMed Google Scholar
Zhu, J. et al. ATG7 overexpression is crucial for tumorigenic growth of bladder cancer in vitro and in vivo by targeting the ETS2/miRNA196b/FOXO1/p27 Axis. Mol. Ther. - Nucleic Acids 7, 299–313 (2017).
Article CAS PubMed PubMed Central Google Scholar
Shin, J. et al. CRISPRpas: programmable regulation of alternative polyadenylation by dCas9. Nucleic Acids Res. 50, E25 (2022).
Article CAS PubMed Google Scholar
Cox, D. B. T. et al. RNA editing with CRISPR-Cas13. https://www.science.org (2017).
Agarwal, V., Lopez-Darwin, S., Kelley, D. R. & Shendure, J. The landscape of alternative polyadenylation in single cells of the developing mouse embryo. Nat. Commun. 12. https://doi.org/10.1038/s41467-021-25388-8 (2021).
Cheng, L. C. et al. Widespread transcript shortening through alternative polyadenylation in secretory cell differentiation. Nat. Commun. 11, 1–14 (2020).
Article CAS Google Scholar
Batra, R. et al. Loss of MBNL leads to disruption of developmentally regulated alternative polyadenylation in RNA-mediated disease. Mol. Cell 56, 311–322 (2014).
Article CAS PubMed PubMed Central Google Scholar
Raz, V., Dickson, G. & ’t Hoen, P. A. C. Dysfunctional transcripts are formed by alternative polyadenylation in OPMD (8, 43). www.impactjournals.com/oncotarget (2017).
Dvinge, H. et al. Sample processing obscures cancer-specific alterations in leukemic transcriptomes. Proc. Natl. Acad. Sci. USA 111, 16802–16807 (2014).
Article ADS CAS PubMed PubMed Central Google Scholar
Flicek, P. et al. Ensembl 2013. Nucleic Acids Res. 41. https://doi.org/10.1093/nar/gks1236 (2013).
Meyer, L. R. et al. The UCSC Genome Browser database: extensions and updates 2013. Nucleic Acids Res. 41. https://doi.org/10.1093/nar/gks1048 (2013).
Katz, Y., Wang, E. T., Airoldi, E. M. & Burge, C. B. Analysis and design of RNA sequencing experiments for identifying isoform regulation. Nat. Methods, 7, 1009–1015 (2010).
Li, B. & Dewey, C. N. RSEM: accurate transcript quantification from RNA-Seq data with or without a reference genome. BMC Bioinform. 12. https://doi.org/10.1186/1471-2105-12-323 (2011).
Langmead, B., Trapnell, C., Pop, M. & Salzberg, S. L. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome. Genome Biol. 10. https://doi.org/10.1186/gb-2009-10-3-r25 (2009).
Trapnell, C., Pachter, L. & Salzberg, S. L. TopHat: discovering splice junctions with RNA-Seq. Bioinformatics 25, 1105–1111 (2009).
Article CAS PubMed PubMed Central Google Scholar
Feng, X., Li, L., Wagner, E. J. & Li, W. TC3A: the cancer 3′ UTR Atlas. Nucleic Acids Res. 46, D1027–D1030 (2018).
Article CAS PubMed Google Scholar
Perez, A. R. et al. GuideScan software for improved single and paired CRISPR guide RNA design. Nat. Biotechnol. 35, 347–349 (2017).
Article CAS PubMed PubMed Central Google Scholar
Doench, J. G. et al. Optimized sgRNA design to maximize activity and minimize off-target effects of CRISPR-Cas9. Nat. Biotechno. 34, 184–191 (2016).
Article CAS Google Scholar
Manguso, R. T. et al. In vivo CRISPR screening identifies Ptpn2 as a cancer immunotherapy target. Nature 547, 413–418 (2017).
Article CAS PubMed PubMed Central Google Scholar
Clement, K. et al. CRISPResso2 provides accurate and rapid genome editing sequence analysis. Nat. Biotechnol. 37, 224–226 (2019).
Huber, W. et al. Orchestrating high-throughput genomic analysis with Bioconductor. Nat. Methods 12, 115–121 (2015).
Article CAS PubMed PubMed Central Google Scholar
Wickham, H., François, R., Henry, L. & Müller, K. dplyr: a grammar of data manipulation. R package version 0.7.6. (2018).
Wickham, H. ggplot2: Elegant Graphics for Data Analysis (Springer, 2009).

Download references

Acknowledgements

We thank A. Rajan for feedback and advice on autophagy-related studies. We thank D. Bennett and the Wellcome Trust Functional Genomics Cell Bank for providing the Melan-A cells. A.M.G is an ARCS Foundation scholar. R.K.B. was supported in part by the NIH/NCI (R01 CA251138), NIH/NHLBI (R01 HL128239, R01 HL151651) and the Blood Cancer Discoveries Grant program through the Leukemia & Lymphoma Society, Mark Foundation for Cancer Research, and Paul G. Allen Frontiers Group (8023-20). R.K.B is a Scholar of The Leukemia & Lymphoma Society (1344-18) and holds the McIlwain Family Endowed Chair in Data Science. Computational studies were supported in part by FHCC’s Scientific Computing Infrastructure (ORIP S10 OD028685). Experimental studies were supported in part by the Experimental Histopathology, Flow Cytometry, and Genomics Shared Resources of the Fred Hutch/University of Washington Cancer Consortium (NIH/NCI P30 CA015704). The results in this publication are based in part on data from The Cancer Genome Atlas Research Network (http://cancergenome.nih.gov).

Author information

Authors and Affiliations

Computational Biology Program, Public Health Sciences Division, Fred Hutchinson Cancer Center, Seattle, WA, USA
Austin M. Gabel, Andrea E. Belleville, James D. Thomas, Siegen A. McKellar, Taylor R. Nicholas, Toshihiro Banjo, Edie I. Crosse & Robert K. Bradley
Basic Sciences Division, Fred Hutchinson Cancer Center, Seattle, WA, USA
Austin M. Gabel, Andrea E. Belleville, James D. Thomas, Siegen A. McKellar, Taylor R. Nicholas, Toshihiro Banjo, Edie I. Crosse & Robert K. Bradley
Department of Genome Sciences, University of Washington, Seattle, WA, USA
Austin M. Gabel & Robert K. Bradley
Medical Scientist Training Program, University of Washington, Seattle, WA, USA
Austin M. Gabel, Andrea E. Belleville & Siegen A. McKellar
Molecular and Cellular Biology Program, University of Washington, Seattle, WA, USA
Andrea E. Belleville & Siegen A. McKellar

Authors

Austin M. Gabel
View author publications
You can also search for this author in PubMed Google Scholar
Andrea E. Belleville
View author publications
You can also search for this author in PubMed Google Scholar
James D. Thomas
View author publications
You can also search for this author in PubMed Google Scholar
Siegen A. McKellar
View author publications
You can also search for this author in PubMed Google Scholar
Taylor R. Nicholas
View author publications
You can also search for this author in PubMed Google Scholar
Toshihiro Banjo
View author publications
You can also search for this author in PubMed Google Scholar
Edie I. Crosse
View author publications
You can also search for this author in PubMed Google Scholar
Robert K. Bradley
View author publications
You can also search for this author in PubMed Google Scholar

Contributions

AMG, JDT and RKB designed the study. AMG, JDT, AEB, SAM, TRN, TB and EC conducted experiments. AMG analyzed the experimental data and performed genomic analyses. AMG and RKB wrote the manuscript, with input from all authors.

Corresponding author

Correspondence to Robert K. Bradley.

Ethics declarations

Competing interests

R.K.B. is a founder and scientific advisor of Codify Therapeutics and Synthesize Bio and holds equity in both companies. R.K.B. has received research funding from Codify Therapeutics unrelated to the current work. The remaining authors declare no competing interests.

Peer review

Peer review information

Nature Communications thanks Ultan McDermott, and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary information

Supplementary Information

Supplementary Data 1

Supplementary Data 2

Supplementary Data 3

Supplementary Data 4

Supplementary Data 5

Supplementary Data 6

Reporting Summary

Peer Review File

Description of Additional Supplementary Files

Source data

Source Data

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions

About this article

Cite this article

Gabel, A.M., Belleville, A.E., Thomas, J.D. et al. Multiplexed screening reveals how cancer-specific alternative polyadenylation shapes tumor growth in vivo. Nat Commun 15, 959 (2024). https://doi.org/10.1038/s41467-024-44931-x

Download citation

Received: 21 December 2022
Accepted: 10 January 2024
Published: 01 February 2024
DOI: https://doi.org/10.1038/s41467-024-44931-x

This article is cited by

Cleavage and polyadenylation machinery as a novel targetable vulnerability for human cancer
- Giulia Pagani
- Paolo Gandellini
Cancer Gene Therapy (2024)

Comments

By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.