Genome-Wide Polyadenylation Maps Reveal Dynamic mRNA 3′-End Formation in the Failing Human HeartNovelty and Significance
Rationale: Alternative cleavage and polyadenylation (APA) of mRNA represents a layer of gene regulation that to date has remained unexplored in the heart. This phenomenon may be relevant, as the positioning of the poly(A) tail in mRNAs influences the length of the 3′-untranslated region (UTR), a critical determinant of gene expression.
Objective: To investigate whether the 3′UTR length is regulated by APA in the human heart and whether this changes in the failing heart.
Methods and Results: We used 3′end RNA sequencing (e3′-Seq) to directly measure global patterns of APA in healthy and failing human heart specimens. By monitoring polyadenylation profiles in these hearts, we identified disease-specific APA signatures in numerous genes. Interestingly, many of the genes with shortened 3′UTRs in heart failure were enriched for functional groups such as RNA binding, whereas genes with longer 3′UTRs were enriched for cytoskeletal organization and actin binding. RNA sequencing in a larger series of human hearts revealed that these APA candidates are often differentially expressed in failing hearts, with an inverse correlation between 3′UTR length and the level of gene expression. Protein levels of the APA regulator, poly(A)-binding protein nuclear-1 were substantially downregulated in failing hearts.
Conclusions: We provide genome-wide, high-resolution polyadenylation maps of the human heart and show that the 3′end formation of mRNA is dynamic in heart failure, suggesting that APA-mediated 3′UTR length modulation represents an additional layer of gene regulation in failing hearts.
The poly(A)tail is found at the 3′-end of fully processed eukaryotic mRNAs, and it is crucial for mRNA stability and transport of mRNA from the nucleus to the cytoplasm. This poly(A)tail of ≈200 adenosines is added by a tightly coupled 2-step process of mRNA cleavage and polyadenylation, and it is carried out by a surprisingly large multiprotein complex, which in humans constitutes of ≈85 proteins.1 It was recently discovered that about half of the human genes generate alternative mRNA isoforms that differ in length of their 3′-untranslated region (UTR) because of a process called alternative cleavage and polyadenylation (APA).2 Given the role of the 3′UTR in mRNA stability, shortening of the 3′UTR by APA may have a profound impact on gene expression. In fact, it has been shown that shorter mRNA isoforms of specific genes can produce 10-fold more protein, partly by escaping miRNA-mediated repression.3 Although APA has already been described 30 years ago, the implementation of deep-sequencing techniques allowed only recently to better appreciate the extent of APA. The first transcriptome-wide studies on APA reported that proliferative cells (eg, cancer cells) globally shift toward shorter mRNA isoforms, whereas differentiating cells generally lengthen their 3′UTR.3,4 The biological consequence of specific APA events was reported in cancer cells, where expression of the shorter isoform of the proto-oncogene IGF2BP1/IMP-1 caused oncogene transformation.3 Altogether, these recent discoveries highlight APA as a widespread regulatory mechanism to control gene expression.
Here, we performed e3′-Seq to map and quantify 3′end-cleavage and poly(A)sites at the nucleotide resolution in healthy and failing human hearts. These polyadenylation profiles allowed us to identify disease-specific APA signatures in dozens of genes, indicating that APA-mediated 3′UTR length modulation represents an additional layer of gene regulation in the failing heart.
Detailed methods are provided in the Online Data Supplement.
Global Analysis of APA in the Heart
To reliably monitor genome-wide APA events, we developed an enhanced 3′-Seq protocol (e3′-seq) that improves 3′-mRNA cleavage site (CS) position mapping and quantification5 (Methods section in the Online Data Supplement). A flow chart of the e3′-Seq procedure and an example of 3′read mapping and CS identification is shown in Figures 1A and 1B and Online Figure I. We measured CSs in 5 control and 5 dilated cardiomyopathy (DCM) hearts. In total, we obtained information on the CS in 12.317 3′UTRs, of which ≈50% contained ≥2 CS (Online Figure II). e3′-Seq data can be visualized directly in the UCSC (University of California Santa Cruz) browser; α-cardiac actin (ACTC1) is shown as an example in Figure 1C. ACTC1 has 1 CS, and the read distribution is similar between control and DCM samples. Strikingly, a CS is used that is located shortly after the stop codon, and not, as expected at the end of the 3′UTR. This may be highly relevant when one investigates the expression regulation of ACTC1 by miRNAs or RNA-binding proteins.
We determined shortening versus lengthening of 3′UTRs based on CS_J scores, a metric which determines the center of mass of all reads mapping to CSs within a 3′UTR (Methods section in the Online Data Supplement). On the basis of these scores, we examined the differences in APA in control and DCM hearts and identified 1370 transcripts with a shift in CS usage (Online Table II, χ2 test, P<0.001 after Bonferroni correction for the number of 3′UTRs with >2 CS). In ≈50% of these transcripts, the shift is toward the distal CS. Equal proportions shifting toward the distal and the proximal CS indicates that there is no global shortening or lengthening of 3′UTRs in DCM, and this is confirmed when calculating the proximal usage index distribution per sample, which also revealed no significant difference in the global usage of proximal CS (Figure 1D; Methods section in the Online Data Supplement).
Enriched functional groups among the 1370 genes displaying altered APA were examined using the online tool PANTHER (Protein Analysis Through Evolutionary Relationships). Genes displaying 3′UTR shortening were mostly enriched for categories related to RNA binding, whereas genes that displayed 3′UTR lengthening were found in categories of actin binding and structural constituent of cytoskeleton (Table).
Candidate Genes Displaying Altered APA in DCM
Six genes displaying a clear shift in APA and in DCM hearts are shown in Figure 2; Online Figures IV and X. These candidates were selected from Online Table III, based on: (1) their previously described functions in the heart (eg, regulator of calcineurin-16 and CDC42 effector protein-3)7 or (2) their absolute difference of CS usage between control and DCM hearts (eg, small EDRK-rich factor 2 [SERF2] and phosphatidylinositol glycan K [PIGK], WEE1 G2-checkpoint kinase [Wee1], and fibrosin-like 1 [FBRSL1]). CDC42 effector protein-3, PIGK, regulator of calcineurin-1, and Wee1 show significant 3′UTR shortening in DCM hearts, manifested by a prominent increase in the proximal CS usage relative to the distal one. FBRSL1 and SERF2 are 2 examples of transcripts with 3′UTR lengthening; 3′end qRT-PCR (quantitative reverse transcriptase-polymerase chain reaction) was used as an independent method to confirm the results of the e3′-Seq. As shown in Online Figure V, 3′end qRT-PCR confirmed the APA shifts in DCM hearts for CDC42 effector protein-3, PIGK, FBRSL1, SERF2, and Wee1, but the latter 3 did not reach statistical significance in this assay.
Effect of APA Changes on Gene Expression
To explore whether the observed APA changes may affect mRNA abundance of the corresponding genes, we used an extensive RNA-seq database of independent control hearts and DCM hearts (Methods section in the Online Data Supplement). As shown in Figure 3A, we found significant upregulation of the genes with a shortened 3′UTR (ie, PIGK and Wee1) and a downregulation of those with a longer 3′UTR (ie, FBRSL1 and SERF2). Western blotting for FBRSL1 and SERF2 revealed that these 2 proteins are also reduced at the protein level in DCM hearts (Online Figure VI). The mRNA expression of regulator of calcineurin-1 and CDC42 effector protein-3 were not different. For one of our main candidates, PIGK, we performed luciferase assays after subcloning the short and long 3′UTR fragment downstream of a Renilla cassette (Online Figure VII). Transfection of these constructs into H10 cells revealed that the 3′UTR shortening of PIGK enhanced Renilla luciferase, which is in line with the enhanced PIGK mRNA expression in DCM hearts where 3′UTR shortening is observed. Finally, when calculating the expression levels of the hundred most significantly changed APA genes, we show that almost 50% of the candidates with a shortened 3′UTR are upregulated in DCM. Vice versa, genes that display 3′UTR lengthening are more often downregulated than the ones with 3′UTR shortening (Online Figure VIII).
APA Changes May Be Driven By Disturbed Expression of Cleavage and Polyadenylation Factors
Mechanisms that may underlie the altered APA in DCM hearts could be related to altered expression of cleavage and polyadenylation factors.1 Therefore, we analyzed the expression levels of several of these factors and found that the expression of poly(A)-binding protein nuclear-1 (PABPN1) was downregulated not only on the mRNA level but also on the protein level (Figure 3B and 3C). In addition, we also found that the mRNA expression of cleavage and polyadenylation-specific factor-4 was downregulated, whereas cleavage and polyadenylation factor subunit-11 was upregulated in DCM hearts (Online Figure IX).
In this study, we provide genome-wide, high-resolution polyadenylation maps of the human heart and show that in subsets of genes, 3′end formation of mRNA changes in failing hearts. For the vast majority of genes, the importance of APA remains unknown, but this work and that of others,8 indicate that APA shifts toward a proximal CS, resulting in shorter 3′UTRs, may contribute to an increased expression. Along the same lines, an APA shift toward a distal CS contributes to downregulation of the mRNA. We did not find global shifts in 3′UTR length in failing hearts, but identified groups of genes where the 3′UTR ratio changed. Genes displaying 3′UTR shortening were mostly enriched for categories related to RNA binding, whereas the genes that displayed 3′UTR lengthening were found in categories of actin binding and structural constituent of cytoskeleton. Strikingly, we did not find APA shifts in genes important for contractile functions, suggesting that APA does not play an important role in these genes. Assuming that UTR length is inversely correlated with expression, the pathway analysis suggests that changes in APA patterns in DCM stimulate (post)transcriptional regulation, whereas repressing genes involved in the structural cytoskeleton organization of the cell. As cardiac remodeling involves alterations in both processes, it is conceivable that many of the APA changes described in this study contribute to these aspects of heart failure.
This is the first study that directly measures global APA in the human heart. Park et al9 analyzed mRNA isoforms in the hypertrophied mouse heart by microarrays and found global shortening of 3′UTRs, something that we did not observe in human hearts. Nevertheless, microarray-based analysis of APA has serious limitations: it is restricted by the design of the array probes, which never cover the complete mRNA, and quantification of APA isoforms with >2 poly(A)sites is unreliable.
An important question about APA in heart failure relates to the mechanism that underlies the altered usage of poly(A)sites. We found disturbed mRNA expression in failing hearts of several genes that are involved in the 3′end-processing machinery. A main finding is the robust decrease of PABPN1 protein because this gene has been identified as a suppressor of APA.5 This suggests that the observed loss of PABPN1 in heart failure may have contributed to 3′UTR shortening of at least some candidates. Gene-targeting studies that directly examine the role of PABPN1 in the mouse heart would be important to further delineate its role in heart failure. Moreover, other factors, such as RNA-binding proteins, splicing and transcriptional mechanisms, and even histone modifications could all affect APA choices. The mechanism underlying 3′UTR lengthening in the failing heart remains elusive.
In conclusion, we show that the 3′-end formation of numerous mRNAs is altered in the failing heart, which is paralleled by a reduced expression of PABPN1. Interestingly, many of the genes with shortened 3′UTRs in DCM are involved in RNA binding, whereas the genes with 3′UTR lengthening seem involved in cytoskeletal organization.
This suggests that specific APA events fulfill a (patho-)biological function. Future studies that address the pathophysiological consequences of these APA changes are required to evaluate their role in the pathogenesis of heart failure and whether manipulation of APA can be considered a therapeutic option for heart failure.
We thank Cris dos Remedios (Sydney Heart Bank), Nanette Bishopric, Alfred George, and Andras Varro for providing human heart samples.
Sources of Funding
This work was supported by grants from the Netherlands Organization for Scientific Research (NWO-836.12.002 and NWO-821.02.021) and the Netherlands Cardiovascular Research Initiative (grants CVON-ARENA-2011-11 and PREDICT-2012-10). A.P. Ugalde was supported by the Human Frontier Science Program (LT000640/2013).
In November 2015, the average time from submission to first decision for all original research papers submitted to Circulation Research was 15.52 days.
3′-Seq bam-files are available at NCBI BioProject accession PRJNA288418.
The online-only Data Supplement is available with this article at http://circres.ahajournals.org/lookup/suppl/doi:10.1161/CIRCRESAHA.115.307082/-/DC1.
- Nonstandard Abbreviations and Acronyms
- 3′-untranslated region
- alternative polyadenylation
- cleavage site
- dilated cardiomyopathy
- enhanced 3′end RNA sequencing
- fibrosin-like 1
- poly(A)-binding protein nuclear-1
- polyadenylation signal
- phosphatidylinositol glycan K
- small EDRK-rich factor 2
- Received July 2, 2015.
- Revision received December 11, 2015.
- Accepted December 15, 2015.
- © 2015 American Heart Association, Inc.
- Tian B,
- Hu J,
- Zhang H,
- Lutz CS.
- Sandberg R,
- Neilson JR,
- Sarma A,
- Sharp PA,
- Burge CB.
- Jenal M,
- Elkon R,
- Loayza-Puch F,
- van Haaften G,
- Kühn U,
- Menzies FM,
- Oude Vrielink JA,
- Bos AJ,
- Drost J,
- Rooijers K,
- Rubinsztein DC,
- Agami R.
- Rothermel BA,
- McKinsey TA,
- Vega RB,
- Nicol RL,
- Mammen P,
- Yang J,
- Antos CL,
- Shelton JM,
- Bassel-Duby R,
- Olson EN,
- Williams RS.
- Colgan DF,
- Manley JL.
- Park JY,
- Li W,
- Zheng D,
- Zhai P,
- Zhao Y,
- Matsuda T,
- Vatner SF,
- Sadoshima J,
- Tian B.
Novelty and Significance
What Is Known?
The 3′-untranslated region is a well-known hotspot for microRNAs and RNA-binding protein interactions, and as such, is important for gene expression regulation.
The polyadenylation machinery determines the length of the 3′-untranslated region by precise positioning of the poly(A) tail.
About half of the human genes generate alternative mRNA isoforms that differ in length at the 3′end by a process called alternative cleavage and polyadenylation (APA). However, in the human heart, APA is uncharted.
What New Information Does This Article Contribute?
We performed 3′-end sequencing and provide genome-wide polyadenylation maps of healthy and failing hearts.
Heart failure–specific APA signatures are present in numerous genes.
We identified robust downregulation of one of the main APA regulators, poly(A)-binding protein nuclear-1, in failing hearts.
APA is emerging as an important layer of gene regulation. It was recently discovered that about half of the human genes generate alternative mRNA isoforms that differ in length of their 3′-untranslated region because of this APA process. In this study, we provide genome-wide polyadenylation maps of the human heart and show that in subsets of genes, 3′end formation of mRNA alters in failing hearts. We found that changes in length of the 3′-untranslated region inversely correlated with expression level of the affected genes in heart failure. We identified a robust downregulation of poly(A)-binding protein nuclear-1 in failing hearts as a putative mechanism underlying these APA changes.