Alternative poly-adenylation modulates α1-antitrypsin expression in chronic obstructive pulmonary disease

Abstract

α1-anti-trypsin (A1AT), encoded by SERPINA1, is a neutrophil elastase inhibitor that controls the inflammatory response in the lung. Severe A1AT deficiency increases risk for Chronic Obstructive Pulmonary Disease (COPD), however, the role of A1AT in COPD in non-deficient individuals is not well known. We identify a 2.1-fold increase (p = 2.5x10-6) in the use of a distal poly-adenylation site in primary lung tissue RNA-seq in 82 COPD cases when compared to 64 controls and replicate this in an independent study of 376 COPD and 267 controls. This alternative polyadenylation event involves two sites, a proximal and distal site, 61 and 1683 nucleotides downstream of the A1AT stop codon. To characterize this event, we measured the distal ratio in human primary tissue short read RNA-seq data and corroborated our results with long read RNA-seq data. Integrating these results with 3' end RNA-seq and nanoluciferase reporter assay experiments we show that use of the distal site yields mRNA transcripts with over 50-fold decreased translation efficiency and A1AT expression. We identified seven RNA binding proteins using enhanced CrossLinking and ImmunoPrecipitation precipitation (eCLIP) with one or more binding sites in the SERPINA1 3' UTR. We combined these data with measurements of the distal ratio in shRNA knockdown experiments, nuclear and cytoplasmic fractionation, and chemical RNA structure probing. We identify Quaking Homolog (QKI) as a modulator of SERPINA1 mRNA translation and confirm the role of QKI in SERPINA1 translation with luciferase reporter assays. Analysis of single-cell RNA-seq showed differences in the distribution of the SERPINA1 distal ratio among hepatocytes, macrophages, αβ-Tcells and plasma cells in the liver. Alveolar Type 1,2, dendritic cells and macrophages also vary in their distal ratio in the lung. Our work reveals a complex post-transcriptional mechanism that regulates alternative polyadenylation and A1AT expression in COPD.

Fig 1. Tissue expression and alternative polyadenylation of the SERPINA1 mRNA.

A) Tissue expression in Transcripts per Million (TPM) on a log scale for SERPINA1 mRNA for 17382 samples sequenced by the GTEx RNA-seq consortia [10,11,62]. The six tissues with median SERPINA1 expression above average (~1000 TPM, indicated with dashed line) were further analyzed. B) Variation (as measured by the tissue specific standard deviation) in tissues. We observe particularly high levels of variation among individuals within blood samples. C) Read coverage of GTEx RNA-seq data for the six tissues highly expressing SERPINA1 mRNA are shown in pink. Read coverage for HepG2 and HepG2 3′ RNA-seq experiments shown in turquoise. All these RNA-seq data sets confirm the exon structure of SERPINA1 as shown on top in black and grey. We observe tissue specific differential exon usage in the 5’ UTR (exons 1–3) consistent with previous work characterizing differential isoform usage in this region of SERPINA1 [4]. D) Zoom of the 3′ UTR region for primary liver data mean depth (top) and HepG2 3′end seq experiment identifying the proximal and distal APA sites. E) PacBio long read sequencing of primary human liver tissue data confirming isoforms using proximal (blue) and distal (yellow) APA sites.

Fig 2. Distal ratio across tissues and in COPD.

The distal ratio is measured as the relative depth of SERPINA1 3′ UTR distal vs. proximal reads in RNA-seq data. A) Distal ratio measured in the six tissues expressing SERPINA1 mRNA above 1000 TPM in Gtex RNA-seq consortium data [10,11]. We observe the highest distal ratio in liver, but important variation in the lung. B) The distal ratio measured in a publicly available [26] short-read lung tissue RNA-seq data set from n = 82 (red) COPD subjects and n = 64 (blue) controls. C) Distal ratio measured from lung tissue RNA-seq data in the Lung Tissue Research Consortium for n = 376 COPD (red) subjects and n = 267 (blue) normal participants. In both studies we observe a significant increase of the distal ratio in COPD subjects. D) The characteristic drop at the proximal APA site indicative of alternative polyadenylation for mean COPD (red) and Normal (blue) lung tissue for n = 376 COPD subjects and n = 267 normal patients. E) Distal ratio analysis of LTRC subjects broken down by SERPINA1 M (normal), S (mild disease) and Z (severe disease) alleles showing that the S and Z alleles exacerbate use of the distal alternative poly-adenylation site in the lungs of individuals with COPD.

Fig 3. Luciferase reporter and mRNA stability assays to measure the effect of long vs. short 3′ UTRs.

To measure effect on translation efficiency and mRNA stability of the long and short 3′ UTR sequences we performed a series of luciferase reporter assays. A) Schematic of the luciferase reporter assay, combining a nanoluciferase reporter, PEST domain and the SERPINA1 exon 7 coding sequence upstream of the long and short 3′ UTRs. The proximal APA site is mutated to inhibit use of this site (indicated with an x on the long construct). These constructs are co-transfected in HepG2 and A549 cell lines with a control firefly reporter and the ratio of nanoluciferase protein to firefly protein measured. B) Log normalized luminescence, which indicates gene expression, measured for short (blue) and long (tan) SERPINA1 isoforms. The expression is significantly higher for the short isoform by close to two orders of magnitude in both lung derived A540 cells and liver derived HepG2 cells. C) Pulse chase experiments in A549 cells using ethylene uridine (EU) and click-it chemistry for labeling with biotin-azide to measure relative mRNA stability by qRT-PCR [63]. We confirmed that GAPDH was consistently stable over the time period and similar declines in both the long and short SERPINA1 RNAs, indicating similar stability for both long and short 3′ UTRs. D) Deletion construct design to identify regions controlling gene expression in SERPINA1 3′ UTR. Six constructs were designed to selectively delete regions 1–4. E) Relative luminescence which indicates expression for the six deletion constructs compared to short (blue) and long (tan) SERPINA1 3′ UTRs. F) Predicted vs. measured expression of deletion constructs using regression model described by Eq 1. This model yields the translation coefficients of the four regions reported in Table 1.

Fig 4. Role of RNA binding proteins and structure in modulating translation efficiency.

Mapping of eCLIP (enhanced CrossLinking and ImmunoPrecipitation) from recent ENCODE (Encyclopedia of DNA Elements) experiments on RNA binding proteins carried out in HepG2 cell lines [–36] onto the long 3′ UTR isoform of SERPINA1 mRNA. Each rectangle indicates a protein binding site and sites are colored by RBP. B) SHAPE (Selective 2’ Hydroxyl Acylation by Primer Extension) structure probing long SERPINA1 3′ UTR. SHAPE data identifies flexible (unpaired) nucleotides in the RNA structure revealing regions of high accessibility for protein binding. Red indicates highly reactive nucleotides, yellow intermediate and black low. C) Using SHAPE data, we compute the entropy of the SERPINA1 mRNA structure. Low entropy regions adopt single, well-defined structures, while high entropy regions are more disordered [64,65]. D) SHAPE derived secondary structure model for short (top) and long SERPINA1 3′ UTR indicated as an arc diagram. Only highly probable base-pairs are shown (green). As can be seen QKI (Quaking Homolog) binds to a low-entropy region in the 3′ UTR with extensive local base-pairing. E) SHAPE reactivity for the nucleotides for the eCLIP QKI binding site including the putative binding motif of the RNA binding protein 5’-NACUAAY-N(1,20)-UAAY-3′ [46]. F) We found the QKI binding site to impact gene expression from the SERPINA1 3’UTR using our nanoluciferase assay. Mutating the QKI binding site increases expression from the long 3’UTR isoform (QKImutant) while adding the QKI binding site and flanking nucleotides decreases expression of the short 3’UTR isoform (Short+QKI) in A549 cells. G) Secondary structure model derived from experimental SHAPE data for QKI binding region, both regions of the motif are accessible for binding. H) QKI is known to retain mRNAs in the nucleus. We measured the SERPINA1 distal ratio in Hepatocellular carcinoma (●) and HepG2 (Δ) nuclear and cytoplasmic RNA-seq fractions and find the highest ratio in the nucleus. This suggests QKI binds and retains the long SERPINA1 isoform in the nucleus thereby inhibiting translation.

Fig 5. Role of RNA binding proteins and relative single-cell populations in modulating SERPINA1 APA.

A) Two-dimensional volcano plot of endogenous shRNA distal ratio in SERPINA1 (y-axis) as a function of log₂ differential expression fold change (log₂(DE_FC)) in LTRC primary lung tissue from 376 COPD cases and 267 controls for the 224 corresponding RNA binding proteins. The horizontal line represents the mean distal ratio in the corresponding empty vector shRNA controls, while the vertical line is centered on zero. Blue points (bottom right quadrant and top left quadrant) will decrease the distal ratio in COPD, while orange points (top left and bottom right quadrant) will increase the distal ratio. Filled circles indicate RNA binding proteins for which the distal ratio changes from shRNA control, and log₂(DE_FC) are both significant in both data sets with p_adj<0.05. B.) log₂ computed Euclidean distance from center of two-dimensional volcano plot for significant (p_adj<0.05) RNA binding proteins affecting COPD. Again, those RNA binding proteins expected to lower the distal ratio are colored blue and negative, while those expected to increase the distal ratio are orange and positive.