If you don't remember your password, you can reset it by entering your email address and clicking the Reset Password button. You will then receive an email that contains a secure link for resetting your password
If the address matches a valid account an email will be sent to __email__ with instructions for resetting your password
Division of Biomedical Sciences, School of Medicine, University of California, Riverside, CA, USAMolecular Medicine Program, Division of Urology, Department of Surgery, University of Utah School of Medicine, Salt Lake City, UT, USA
Correspondence: Changcheng Zhou, PhD, FAHA, Division of Biomedical Sciences, School of Medicine, 900 University Avenue, University of California, Riverside, CA 92521. Phone: 951-827-9139; Fax: 951-579-4118;
Small non-coding RNAs (sncRNAs) play diverse roles in numerous biological processes. While the widely used RNA sequencing (RNA-seq) method has advanced sncRNA discovery, RNA modifications can interfere with the cDNA library construction process, preventing the discovery of highly modified sncRNAs including transfer RNA-derived small RNAs (tsRNAs) and ribosomal RNA-derived small RNAs (rsRNAs) that may have important functions in disease development. To address this technical obstacle, we recently developed a novel PANDORA-seq method to overcome RNA modification-elicited sequence interferences. To identify novel sncRNAs associated with atherosclerosis development, LDL Receptor-deficient (LDLR-/-) mice were fed a low-cholesterol diet (LCD) or high-cholesterol diet (HCD) for 9 weeks. Total RNAs isolated from the intima were subjected to PANDORA-seq and traditional RNA-seq. By overcoming RNA modification-elicited limitations, PANDORA-seq unveiled a rsRNA/tsRNA-enriched sncRNA landscape in the atherosclerotic intima of LDLR-/- mice which was strikingly different from that detected by traditional RNA-seq. While microRNAs (miRNAs) were the dominant sncRNAs detected by traditional RNA-seq (55.9% of total sncRNAs), PANDORA-seq substantially increased the reads of rsRNAs and tsRNAs which account for 77.4% and 5% of total sncRNAs, respectively. PANDORA-seq also detected 1,383 differentially expressed sncRNAs induced by HCD feeding, including 1,160 rsRNAs and 195 tsRNAs. One of HCD-induced intimal tsRNAs, tsRNA-Arg-CCG may contribute to atherosclerosis development by regulating the pro-atherogenic gene expression in endothelial cells. Overall, PANDORA-seq revealed a hidden rsRNA and tsRNA population associated with atherosclerosis development. These understudied tsRNAs and rsRNAs, which are much more abundant than miRNAs in the atherosclerotic intima of LDLR-/- mice, warrant further investigations.
Atherosclerosis is a chronic inflammatory disease characterized by the accumulation of cholesterol, immune cells, and fibrous elements in the sub-intimal layer of the artery leading to plaque formation and cardiovascular events (
). At the cellular level, the initial stages of atherosclerosis development involve endothelial dysfunction followed by macrophage infiltration, lipid accumulation, and vascular smooth muscle cell (SMC) migration in the intimal layer of the vasculature (
). In addition to well-defined risk factors, numerous noncoding RNAs, such as long non-coding RNAs (lncRNAs) and small non-coding RNAs (sncRNAs), have been shown to play important roles in regulating atherosclerosis development (
Noncoding RNAs in Cardiovascular Disease: Current Knowledge, Tools and Technologies for Investigation, and Future Directions: A Scientific Statement From the American Heart Association.
Noncoding RNAs in Cardiovascular Disease: Current Knowledge, Tools and Technologies for Investigation, and Future Directions: A Scientific Statement From the American Heart Association.
Noncoding RNAs in Cardiovascular Disease: Current Knowledge, Tools and Technologies for Investigation, and Future Directions: A Scientific Statement From the American Heart Association.
). In addition to miRNAs, there are many other sncRNA categories including transfer RNA-derived small RNAs (tsRNAs), ribosomal RNA-derived small RNA (rsRNAs), YRNA-derived small RNAs (ysRNAs), and piwi-interacting RNA (piRNAs) (
Deep sequencing of RNA from immune cell-derived vesicles uncovers the selective incorporation of small non-coding RNA biotypes with potential regulatory functions.
). However, the functions of these understudied sncRNAs in atherosclerosis and cardiovascular disease (CVD) are mostly unknown. As compared to miRNAs, many of these sncRNAs including tsRNAs and rsRNAs are highly modified since their precursors (e.g., tRNAs and rRNAs) bear various RNA modifications such as RNA methylations (e.g., m1G, m1A and m3C) (
Akiyama, Y., S. M. Lyons, M. M. Fay, T. Abe, P. Anderson, and P. Ivanov. 2019. Multiple ribonuclease A family members cleave transfer RNAs in response to stress. biorxiv.
). These modifications can interfere with the reverse transcription and adaptor ligation processes in the widely used cDNA library construction protocol for standard RNA sequencing (RNA-seq) analysis (
). To address this issue, we recently developed an innovative RNA-seq protocol, PANDORA-seq (Panoramic RNA Display by Overcoming RNA Modification Aborted Sequencing) which employs a combinatorial enzymatic treatment protocol to remove key RNA modifications that block adapter ligation and reverse transcription processes during cDNA library construction (
). PANDORA-seq has enabled us to detect the highly modified sncRNAs including tsRNAs and rsRNAs, resulting in the discovery of higher abundant levels of tsRNAs/rsRNAs in various tissues and cell types that were undetectable by traditional RNA-seq (
In the current study, we used PANDORA-seq to identify novel sncRNAs that are associated with atherosclerosis development in LDL receptor-deficient (LDLR-/-) mice. By overcoming RNA modification-elicited limitations, PANDORA-seq unveiled a rsRNA- and tsRNA-enriched sncRNA landscape in the atherosclerotic intima of LDLR-/- mice, which was strikingly different from that detected by traditional RNA-seq. We also found that one of the tsRNAs, tsRNA-Arg-CCG was upregulated in the atherosclerotic intima of LDLR-/- mice. Interestingly, tsRNA-Arg-CCG can affect the expression of pro-atherogenic genes in human endothelial cells, which may contribute to atherosclerosis development.
Materials and Methods
Animals
3-week-old male C57BL/6 LDLR-/- mice (Jackson Laboratories) were fed ad libitum on a semisynthetic low-fat (4.2% fat) AIN76 diet containing either low cholesterol (LCD; 0.02% cholesterol; Research Diets) or high cholesterol (HCD; 0.5% cholesterol; Research Diets) for 9 weeks until they were euthanized at 12 weeks of age (
Induction of atherosclerosis by low-fat, semisynthetic diets in LDL receptor-deficient C57BL/6J and FVB/NJ mice: comparison of lesions of the aortic root, brachiocephalic artery, and whole aorta (en face measurement).
). All animals were housed in pathogen-free microisolator cages in a temperature-controlled environment with a 12-h light-dark cycle. All experimental mice used in this study were male. However, studying a single sex has limitations since sex differences have been widely reported in mouse atherosclerosis studies (
). On the day of sacrifice, mice were fasted for 6 h following the dark cycle (feeding cycle) and blood and major organs were collected as previously described (
). All animal studies were performed in compliance with the approved protocols by the Institutional Animal Care and Use Committee of the University of California, Riverside.
Blood analysis
Serum total cholesterol and total triglyceride concentrations were measured using the Wako Cholesterol E enzymatic colorimetric assay (Wako, 999-02601) and the Wako L-type TG M assay (Wako, 994-02891) according to the manufacturer’s instructions (FUJIFILM Medical Systems U.S.A., Inc., Richmond, VA, USA). The lipoprotein fractions were isolated in a Beckman Coulter XPN100-IVD ultracentrifuge as previously described (
). Lipoprotein fractions were isolated by centrifuging 60 μL of serum at 40,000 RPM for 5 h at 4°C in a Beckman Coulter Type 42.2 Ti Rotor at its own density (1.006g/mL). The infranatant was then adjusted to a density of 1.063g/mL with solid potassium bromide to harvest the VLDL (<1.006g/mL), LDL (1.006 ≤ d ≤ 1.063g/mL), and HDL (d>1.063g/mL) fractions by spinning at 40000 RPM for 24 h at 4°C. The cholesterol content of the lipoprotein fractions was then measured enzymatically (Wako, cholesterol 999-02601).
Atherosclerotic lesion analyses
The atherosclerotic lesion sizes were quantified as previously described (
). To quantify the plaque area at the aortic root, Optimal Cutting Temperature (OCT)-compound-embedded hearts were sectioned at a 12-μm thickness keeping all the three valves of the aortic root in the same plane, and stained with oil red O. To quantify atherosclerotic plaque area at the brachiocephalic artery (BCA), the OCT-embedded brachiocephalic arteries were sectioned from distal to proximal at a thickness of 10μm. BCA atherosclerotic lesions from the lumenal to the internal elastic lamina were quantified in three equidistant oil red O-stained sections at 200, 400, and 600μm proximal from the branching point of the brachiocephalic artery into the carotid and subclavian arteries. Images were taken and plaque size was quantified using a Nikon microscope (Nikon, Melville, NY, USA).
Immunohistochemistry
The aortic root of the mice freshly embedded in OCT and sectioned were first fixed in 4% PFA for 15 minutes and permeabilized with 0.1% Triton X-100 in PBS (PBST) for 15 min. The slides were then blocked with PBST containing 5% bovine serum albumin (BSA) (Sigma-Aldrich, A9647) for 1 h at room temperature. For immunostaining, slides were incubated with antibodies against CD68 (Bio-Rad AbD Serotec, MCA1957) and αSMA (Abcam, ab5694) at 4°C for 12 to 15 hours (
). The sections were rinsed with PBS and incubated with fluorescein-labeled secondary antibodies (Life Technologies). The nuclei were stained by mounting the slides with DAPI medium (Vector Laboratories). Samples were imaged and analyzed with a Nikon fluorescence microscope. Images were taken, and the positively labeled area was quantified using a Nikon microscope. For trichrome staining, the aortic root freshly embedded in OCT and sectioned were stained with a trichrome stain (Masson) kit (Sigma-Aldrich, HT15) according to the manufacturer’s instructions with some modifications to accommodate for frozen tissue (
). Briefly, sections were fixed in 4% PFA for 30 min then stained with Biebrich Scarlet-Acid Fuchsin Solution for 1 min to stain the fibers red. The samples were rinsed with ddH2O and placed in Working Phosphotungstic/Phosphomolybdic Acid Solution for 1 hr followed directly with Aniline Blue staining for 1 min. Lastly, samples were placed in 1% Glacial Acetic Acid for 2 min and mounted with Permount. Images were taken and collagen content was quantified using a Nikon microscope (Nikon, Melville, NY, USA).
Cell culture and transfection
Human endothelial cell line HMEC-1 was purchased from AATC (CRL-3243) (
). HMEC-1 cells passaged less than 20 times were used for transfection. Synthetic control Oligo and tsRNS-Arg-CCG Oligo (Sigma-Aldrich) was transfected into the cells with Lipofectamine RNAiMAX (Thermo Fisher, 13778100) (
). Total RNAs isolated from HMEC-1 cells were separated by a 10% urea-PAGE gel followed by SYBR gold staining of nucleic acids (Thermo Fischer, S11494) and immediately imaged. The RNA was then transferred to a positively charged nylon membrane (Roche, 11417240001) and ultraviolet crosslinked with 0.12J of energy. Membranes were hybridized with PerfectHyb Plus Hybridization Buffer (Sigma-Aldrich, H7033) for 1 hour at 42˚C. To detect tsRNA-Arg-CCG, membranes were incubated overnight in 42˚C with a DIG-labeled oligonucleotide probe (5’DIG- CGAACCCACAATCCCCAGCT-3’).
Intimal RNA isolation
The aortas of LDLR-/- mice were isolated and flushed with PBS followed by intimal peeling using TRIzol reagent (Sigma-Aldrich, T9424) (
). A total of ∼300-400 μL of Trizol was flushed into the aorta for 10 seconds (∼100μL) followed by a 10-second pause and repeated 3 times, as previously described (
). Total RNAs from the remaining aorta (media and adventitia) were also isolated using TRIzol for gene expression analyses. Intimal RNA integrity and concentrations were confirmed using an Agilent 2100 Bioanalyzer and Aligent RNA 6000 Nano Kit (Agilent Technologies Inc, 5067-1511) (
Quantitative Real-Time PCR (QPCR) were performed by using gene-specific primers and SYBR Green PCR kit (Bio-Rad Laboratories) on a Bio-Rad CFX Real-Time-PCR Machine as previously described (
). Here we provided detailed information for PANDORA-seq of intimal small RNAs. A schematic of PANDORA-seq workflow is also included in Figure 1.
Figure 1Schematic of PNADORA-seq workflow. Step 1: Small RNAs from intimal total RNAs were excised from PAGE gel. Step 2: Purified small RNAs were subjected to T4PNK treatment. Step 3: Purified small RNAs were subjected to Alkb treatment. Step 4: NEBNext Small RNA Library Prep Set for Illumina was used for cDNA library construction. Step 5: PCR products were purified with PAGE gel. Step 6: Final product was sequenced and annotated using the SPORTS1.1 pipeline. A detailed protocol of PANDORA-seq is included in the Materials and Methods section. Image was created by using BioRender.com.
The RNA sample from the intima was mixed with an equal volume of 2× RNA loading dye (New England Biolabs, B0363S) and incubated at 75 °C for 5 min. The mixture was loaded into 15% (wt/vol) urea polyacrylamide gel (10 ml mixture containing 7 M urea (Invitrogen, AM9902), 3.75 ml Acrylamide/Bis 19:1, 40% (Ambion, AM9022), 1 ml 10× TBE (Invitrogen, AM9863), 1 g l−1 ammonium persulfate (Sigma–Aldrich, A3678-25G) and 1 ml l−1 TEMED (Thermo Fisher Scientific, BP150-100) and run in a 1× TBE running buffer at 200 V until the bromophenol blue reached ∼1cm from the bottom of the gel. Small RNA of 15-50 nucleotides was visualized with SYBR Gold solution (Invitrogen, S11494) and excised based on a small RNA ladder (New England Biolabs, N0364S) (
). The RNA was eluted in 0.3 M sodium acetate (Invitrogen, AM9740) and 100 U ml−1 RNase inhibitor (New England Biolabs, M0314L) overnight at 4 °C. The samples were then centrifuged for 10 min at 12,000x g at 4°C. The aqueous phase was mixed with pure ethanol, 3 M sodium acetate and linear acrylamide (Invitrogen, AM9520) at a ratio of 3:9:0.3:0.01 and incubated at -20 °C for 2 hrs and centrifuged for 25 min at 12,000 x g at 4°C (
). After removing the supernatant, the precipitate was resuspended in nuclease-free water.
Treatment of RNA with AlkB
The intimal RNA was incubated in a 50 μL reaction mixture containing 50 mM HEPES (Fisher Scientific, 15630080), 75 μM ferrous ammonium sulfate, 1 mM α-ketoglutaric acid (Sigma–Aldrich; K1128), 2 mM sodium ascorbate, 50 mg l-1 bovine serum albumin (Sigma-Aldrich, A9647), 4 μg mL-1 AlkB, 2,000 U ml-1 RNase inhibitor and RNA at 37°C for 30 min (
). Then, the mixture was added into 500 μL TRIzol reagent for RNA isolation.
Treatment of RNA with T4PNK
The intimal RNA was incubated in 50μL reaction mixture containing 5μL 10× PNK buffer (New England Biolabs, B0201S), 1 mM ATP (New England Biolabs, P0756S), 10 U T4PNK (New England Biolabs, M0201L) and RNA at 37°C for 20 min (
). Then, the mixture was added into 500μL TRIzol reagent for RNA isolation.
Small RNA library construction and deep sequencing
The adapters were obtained from the NEBNext Small RNA Library Prep Set for Illumina (New England Biolabs, E7330S) and ligated sequentially. First, we added a 3’ adapter system under the following reaction conditions: 70°C for 2 min and 16°C for 18 h. Second, we added a reverse transcription primer under the following reaction conditions: 75°C for 5 min, 37°C for 15 min and 15°C for 15 min. Third, we added a 5’ adapter mix system under the following reaction conditions: 70°C for 2 min and 25°C for 1 h. First-strand cDNA synthesis was performed under the following reaction conditions: 70°C for 2 min and 50°C for 1 h. PCR amplification with PCR Primer Cocktail and PCR Master Mix was performed to enrich the cDNA fragments under the following conditions: 94°C for 30 s; 17 cycles of 94°C for 15 s, 62°C for 30 s and 70°C for 15 s; 70°C for 5 min; and hold at 4°C. Then, the PCR product was purified from PAGE gel.
Small RNA annotation and analyses for PANDORA-seq data
Small RNA sequences were annotated using the software SPORTS1.1 with one mismatch tolerance (SPORTS1.1 parameter setting: -M 1). Reads were mapped to the following individual non-coding RNA databases sequentially: (
) the non-coding RNAs defined by Ensembl and Rfam 12.3. The tsRNAs were annotated based on both pre-tRNA and mature tRNA sequences. Mature tRNA sequences were derived from the GtRNAdb and mitotRNAdb sequences using the following procedures: (
) a G nucleotide was added to the 5′ ends of histidine tRNAs. The tsRNAs were categorized into four types based on the origin of the tRNA loci: 5′ tsRNA (derived from the 5′ end of pre-/mature tRNA); 3′ tsRNA (derived from the 3′ end of pre-tRNA); 3′ tsRNA-CCA end (derived from the 3′ end of mature tRNA); and internal tsRNAs (not derived from 3′ or 5′ loci of tRNA). For the rsRNA annotation, we mapped the small RNAs to the parent rRNAs in an ascending order of rRNA sequence length to ensure a unique annotation of each rsRNA (for example, the rsRNAs mapped to 5.8S rRNA would not be further mapped to the genomic region overlapped by 5.8S and 45S rRNAs). We then employed the edgeR algorithm (
) to perform the comparison of sncRNA expression patterns between groups. We applied the TMM algorithm to perform reads count normalization and effective library size estimation and the likelihood ratio test to identify the differentially expressed sncRNA species. The sncRNA species with a false discovery rate (FDR)< 0.1 and fold change (FC) > 2 were deemed differentially expressed.
Intimal transcriptome analysis
The creation of cDNA libraries and sequencing were performed using the Illumina standard operation pipeline as previously described (
). For data analysis, we applied the Salmon tool to quantify the mRNA expression from the raw sequencing data with the default setting, based on the Ensembl mouse cDNA annotation (GRCm38). We then employed the edgeR algorithm (
) to perform the comparison in transcriptomic pattern between groups, using the TMM algorithm to perform read count normalization and effective library size estimation and the likelihood ratio test to identify the DEGs. The genes with FDR <0.1 and FC >2 were deemed differentially expressed. We further performed pathway analysis upon the differentially expressed genes using the definition from Kyoto Encyclopedia of Genes and Genomes (KEGG) project. For each KEGG pathway, we computed a geneset score, using the Functional Analysis of Individual Microarray Expression (FAIME) algorithm (
). A higher FAIME score suggests a higher overall expression of a given pathway. All the RNA-seq datasets have been deposited in the Gene Expression Omnibus (GSE213305).
Statistical Analysis
All data except the high-throughput sequencing data are presented as the mean ± SEM. Individual pairwise comparisons were analyzed by two-sample, two-tailed Student’s t-test unless otherwise noted, with P<0.05 was regarded as significant using GraphPad Prism. One-way analysis of variance (ANOVA) with uncorrected Fisher’s LSD test was used for analyzing different origins of the tsRNA/miRNA expression ratio under different treatments. N numbers are listed in figure legends.
Results
Feeding LDLR-/- mice a low-fat, high-cholesterol diet promotes hypercholesterolemia without affecting adiposity and metabolic phenotypes
To investigate sncRNAs associated with atherosclerosis development, male LDLR-/- mice were fed a low-fat (4.3% fat) AIN76 diet containing 0.02% or 0.5% cholesterol for 9 weeks (
Induction of atherosclerosis by low-fat, semisynthetic diets in LDL receptor-deficient C57BL/6J and FVB/NJ mice: comparison of lesions of the aortic root, brachiocephalic artery, and whole aorta (en face measurement).
). We and others have successfully used this diet to induce atherosclerosis development without eliciting obesity and associated metabolic disorders in LDLR-/- or ApoE-/- mice (
Induction of atherosclerosis by low-fat, semisynthetic diets in LDL receptor-deficient C57BL/6J and FVB/NJ mice: comparison of lesions of the aortic root, brachiocephalic artery, and whole aorta (en face measurement).
). Consistently, we found that LDLR-/- mice fed the relatively high-cholesterol diet (HCD, 0.5% cholesterol) had similar body weight and growth curve as mice fed the low-cholesterol diet (LCD, 0.02% cholesterol) (Figure 2A). HCD- and LCD-fed mice also had similar body composition, including fat mass and lean mass (Figure 2B). Glucose tolerance tests also demonstrated that exposure to these diets did not alter glucose tolerance in LDLR-/- mice (Figure 2C). Consistent with previous results (
Induction of atherosclerosis by low-fat, semisynthetic diets in LDL receptor-deficient C57BL/6J and FVB/NJ mice: comparison of lesions of the aortic root, brachiocephalic artery, and whole aorta (en face measurement).
), HCD feeding led to elevated serum total cholesterol levels without affecting triglyceride levels (Figure 2D). Lipoprotein fraction analysis was then performed, and mice fed HCD had significantly higher atherogenic LDL and VLDL cholesterol levels, but similar HDL cholesterol levels as compared with LCD-fed mice (Figure 2E). Thus, the low-fat HCD can promote hypercholesterolemia without inducing obesity or metabolic disorders in LDLR-/- mice.
Figure 2A low-fat, high-cholesterol diet effectively promotes hypercholesterolemia and atherosclerosis without inducing obesity and insulin resistance in LDL receptor-deficient mice. Three-week-old male C57BL/6 LDLR-/- mice were fed a low-cholesterol diet (LCD, 0.02% cholesterol) or high-cholesterol diet (HCD, 0.5% cholesterol) for 9 weeks until euthanasia at 12 weeks of age. (A-C) Growth curves (A), body composition (B) and intraperitoneal glucose tolerance test (C) of LCD- and HCD-fed LDLR-/- mice. (D) Serum total cholesterol and triglyceride levels were measured. (E) Lipoprotein fractions (VLDL-C, LDL-C, and HDL-C) were isolated, and the cholesterol levels of each fraction were measured. (F) Quantitative analysis of the lesion area in the aortic root. Representative oil red O-stained aortic root sections displayed below the quantification data (scale bar = 200μm). (G) Quantitative analysis of the lesion area in the brachiocephalic artery (BCA). Representative oil red O-stained BCA sections displayed below quantification data (scale bar = 200μm). LCD, low-cholesterol diet; HCD, high-cholesterol diet; VLDL-C, very low-density lipoprotein cholesterol; LDL-C, low density lipoprotein cholesterol; HDL-C, high density lipoprotein cholesterol. All data are plotted as means ± S.E.M. (n=5-8 each group; *P<0.05; **P<0.01; ***P<0.001).
High-cholesterol diet feeding effectively induces atherosclerosis development in lean LDLR-/- mice
Atherosclerotic lesion areas were then analyzed in the aortic root and brachiocephalic artery (BCA) of LCD or HCD-fed LDLR-/- mice. We found that HCD feeding significantly increased atherosclerotic lesion areas in the aortic root of LDLR-/- mice as compared with LCD-fed mice (183,487.7 ± 48,193.4 μm2 vs. 8,378.9 ± 4,142.7 μm2) (Figure 2F). Short-term LCD feeding did not induce observable atherosclerosis development in BCA of LDLR-/- mice (Figure 2G). However, exposure to HCD significantly increased the atherosclerotic lesion areas in the BCA of LDLR-/- mice (8,559.8 ± 3,652.6 μm2 vs. 0 ± 0 μm2) (Figure 2G).
In addition to lesion size, we characterized multiple factors that are associated with atherosclerosis development, including macrophage infiltration, SMC migration, and collagen production. As expected, immunostaining for macrophage and SMC markers showed increased macrophage contents and SMC migration in the atherosclerotic plaque of HCD-fed LDLR-/- mice (Supplemental Figure 1A-1C). Furthermore, Masson’s Trichrome staining also demonstrated that HCD feeding led to increased collagen content in the atherosclerotic lesions of LDLR-/- mice as compared to LCD-fed mice (Supplemental Figure 1D). Collectively, these results demonstrate that the low-fat HCD feeding can effectively induce atherosclerosis development in lean LDLR-/- mice.
Transcriptome analysis reveals altered atherosclerosis-related gene expression in the intima of HCD-fed LDLR-/- mice
We next isolated intimal RNAs for regular RNA sequencing (RNA-seq) analysis to understand the transcriptomic changes in the intima of LDLR-/- mice. Total RNAs isolated from the intima and remaining aortic fraction were first characterized for endothelial marker (e.g., VE-Cadherin and Tie2) and SMC marker (Myh11 and α-smooth muscle actin) expression. Consistent with previous studies (
), our results confirmed the enrichment of endothelial markers and the low expression levels of SMC markers in the intima as compare with that of media/adventitia (Supplemental Figure 2). By contrast, the media/adventitial fraction has the high expression levels of SMC markers but low expression levels of endothelial cell markers (Supplemental Figure 2).
RNA-seq analysis of the intima uncovered 1,313 differentially expressed genes (DEGs) in the intima of HCD-fed LDLR-/- mice as compared with LCD-fed mice with a false discovery rate (FDR) of < 0.1 and fold change (FC) > 2 as the cut-off threshold (Figure 3A and Supplemental Table 1). Kyoto Encyclopedia of Genes and Genomes (KEGG) analysis then revealed that many upregulated DEGs were enriched in several biological processes that contribute to atherosclerosis development including “cytokine-cytokine receptor interaction”, “chemokine signaling pathway”, and “cell adhesion molecules” (Figure 3B). In addition, some DEGs were also enriched in lysosome and phagosome associated pathways that are also important for atherosclerosis development (Figure 3B). By using the Functional Analysis of Individual Microarray Expression (FAIME) algorithm, we verified the geneset scores of these pathways were significantly increased in HCD-fed LDLR-/- mice as compared with LCD-fed mice (Figure 3C). The genes associated with these pathways were also upregulated in HCD-fed LDLR-/- mice (Figure 3D). Taken together, intimal RNA-seq analysis confirmed the altered atherosclerosis-related genes and pathways that are associated with the increased atherosclerosis in HCD-fed LDLR-/- mice.
Figure 3High-cholesterol diet feeding affects many atherosclerosis-related gene expression in the intima of LDL receptor-deficient mice. Three-week-old male LDLR-/- mice were fed an LCD or HCD for 9 weeks. Total RNA was isolated from the intima and used for RNA-seq analysis. (A) Volcano plot of differential expressed genes (DEGs) in the intima of HCD-fed LDLR-/- mice as compared with LCD-fed mice. Colored dots represent the enriched (red dots) or depleted (blue dots) DEGs with a false discovery rate (FDR) of < 0.1 and fold change (FC) > 2 as a cut-off threshold. (B) KEGG pathways significantly associated with upregulated DEGs in intimal of HCD-fed LDLR-/- mice. The P-values were computed by Fisher’s exact test. The vertical dash line indicates the significance level of α = 0.05. The y-axis displays the KEGG pathways while the x-axis displays the adjusted P-values. (C) Geneset scores of the prioritized KEGG pathways. The geneset score was calculated using the FAIME algorithm. (D) Heatmap representation of DEGs involved in the pathways of “Cytokine-cytokine receptor interaction”, “Chemokine signaling pathway”, and “Cell adhesion molecules” “Lysosome”, “Phagosome”, “Hematopoietic cell lineage”, “B cell receptor signaling pathway”, and “Antigen processing and presentation”. Each column shows one individual gene and each row shows a biological replicate of mouse. Red represents relatively increased gene expression while blue denotes downregulation (n=5 each group).
). To overcome this obstacle, we recently developed a novel small RNA sequencing method, PANDORA-seq, to eliminate the RNA modification-elicited sequence interferences (Figure 1) (
). To determine whether PANDORA-seq can detect novel sncRNAs associated with atherosclerosis development, intimal RNA was collected and processed through both traditional small RNA-seq and PANDORA-seq followed by SPORTS1.1 bioinformatics analysis (
). Indeed, traditional-seq detected a miRNA-enriched sncRNA landscape in the intima of both LCD (47.9%) and HCD (55.9%)-fed LDLR-/- mice (Figure 4A). However, PANDORA-seq revealed a totally different sncRNA landscape in which rsRNAs and tsRNAs account for 83.1% (LCD) and 82.4% (HCD) of total detected sncRNAs in the intima of LDLR-/- mice (Figure 4A).
Figure 4Read summaries and length distributions of different sncRNA categories in the intima of LDL receptor-deficient mice under traditional RNA-seq and PANDORA-seq. Three-week-old male LDLR-/- mice were fed an LCD or HCD for 9 weeks. Total RNA was isolated from the intima and used for PANDORA-seq or traditional small RNA sequencing. (A) Dynamic landscapes of intimal sncRNAs including miRNAs, tsRNAs, rsRNAs, and piRNAs detected by traditional-seq and PANDORA-seq protocols. Zoomed panels of miRNAs and tsRNAs detected by PANDORA-seq are shown on the right. (B) Relative tsRNA/miRNA ratios under traditional-seq and PANDORA-seq protocols. (C) tsRNA responses to traditional-seq and PANDORA-seq in regard to different origins (5′ tsRNA, 3′ tsRNA, 3′ tsRNA-CCA end, and internal tsRNAs). The y axes represent the relative expression levels compared with total reads of miRNA. Different letters above the bars indicate statistically significant differences (P< 0.05). Same letters indicate P> 0.05. Statistical significance was determined by two-sided one-way ANOVA with uncorrected Fisher’s LSD test. All data are plotted as means ± S.E.M. (n=4-5 in each group).
As rsRNA reads were dominant in PANDORA-seq results which consumed the relative reads of other sncRNAs (Figure 4A), we filtered out rsRNAs from the total sncRNA reads and found that PANDORA-seq but not traditional RNA-seq detected an increased tsRNA:miRNA ratios in both LCD- and HCD-fed mice (Figure 4B). The origins, from which the tsRNAs are derived (5’tsRNAs, 3’tsRNAs, 3’tsRNAs with a CCA end, and internal tsRNAs), were also analyzed. PANDORA-seq revealed an increased relative expression of specific tsRNA origins as compared to traditional RNA-seq (Figure 4C). Consistent with previous studies, a majority of tsRNAs are derived from the 5’ end of mature tRNAs (
). Specific rRNA loci, where rsRNA are derived, were also evaluated, and the data demonstrate increased rsRNA detection by PADNORA-seq (Supplemental Figure 3). These results suggest that PANDORA-seq can detect highly modified sncRNAs such as tsRNAs and rsRNAs in the atherosclerotic intima of LDLR-/- mice that were otherwise undetectable by the traditional RNA-seq method.
PANDORA-seq detects more differentially expressed sncRNAs associated with atherosclerosis development in the intima of HCD-fed LDLR-/- mice
We further analyzed both PANDORA-seq and traditional RNA-seq results to determine how HCD feeding alters the expression levels of sncRNAs in the intima of LDLR-/- mice. While traditional RNA-seq only detected a small number of differentially regulated sncRNAs, including only 16 rsRNAs and tsRNAs, PANDORA-seq detected a total of 1,383 differentially regulated sncRNAs (FC >2 and FDR <0.1), including 1160 rsRNAs and 195 tsRNAs (Figure 5A and Supplemental Table 2-7). PANDORA-seq also detected 28 differentially regulated miRNAs in the intima of HCD-fed LDLR-/- mice (Figure 5B and Supplemental Table 4). Many of those miRNAs are consistent with previous reports such as miR-146b, miR-155, and miR-125a, though the functions of these miRNAs on atherogenesis have not been completely understood (
Figure 5Identification of significantly altered intimal sncRNAs associated with atherosclerosis development in LDL receptor-deficient mice by PANDORA-seq and traditional RNA-seq. Three-week-old male LDLR-/- mice were fed an LCD or HCD for 9 weeks. Total RNA was isolated from the intima and used for PANDORA-seq or traditional small RNA sequencing. (A) Volcano plot of differentially expressed intimal sncRNAs identified by Traditional (left) and PANDORA-seq methods (right). Colored dots represent the differentially expressed tsRNAs (red dots), rsRNAs (pink dots), or miRNAs (blue dots) with a false discovery rate (FDR) < 0.1 and fold change (FC) > 2 as the cut-off threshold. (B) Heatmap representation of differentially expressed intimal miRNAs detected by PANDORA-seq. (C) Comparison of rsRNA-generating loci by rsRNA mapping data on 5S rRNA, 12S rRNA, and 28S rRNA detected by PANDORA-seq. (D) Heatmap representation of differentially expressed intimal tsRNAs detected by PANDORA-seq. (E) Dynamic responses to LCD or HCD of representative individual tsRNAs (pictured right). Biological replicates are represented in each row (B) or column (D). Red represents relatively increased expression while blue represents decreased expression with an FDR < 0.1 and FC >2 as the cut-off threshold (n=4-5 each group).
Next, we compared rsRNA expression patterns between LCD and HCD-fed LDLR-/- mice from individual ribosomal RNA (e.g., 5S, 12S, and 28S) by analyzing the specific loci where rsRNAs were derived (Figure 5C and Supplemental Figure 3). The mapping of rsRNAs from 5S, 12S, and 28S, used as examples, showed dynamic expression patterns in response to HCD feeding detected by PANDORA-seq (Figure 5C). We also generated a heatmap to compare the 195 significantly altered intimal tsRNAs between HCD- and LCD-fed mice detected by PANDORA-seq (Figure 5D and Supplemental Table 2). Further, mapping of tsRNA expression patterns on individual tRNA length scales (tRNA-Trp, tRNA-Arg, tRNA-Ser, tRNA-Leu, and tRNA-Asp used as examples) revealed that those tsRNAs also contain distinct dynamic responses to HCD feeding (Figure 5E). The functions of those tsRNAs are mostly unknown and more studies are required to investigate the role of those tsRNAs in atherogenesis.
High-cholesterol diet-induced tsRNA-Arg-CCG affects pro-atherogenic gene expression in endothelial cells in vitro
Endothelial cells play an essential role in regulating vascular inflammation and the initiation and progression of atherosclerosis (
). To investigate the potential function of tsRNAs identified by PANDORA-seq in atherosclerosis, we selected one of the HCD-induced tsRNAs, tsRNA-Arg-CCG for in vitro analysis. tsRNA precursors, tRNAs, are a highly conserved RNA species (
), and human and murine tRNA-Arg-CCG also shared the same sequence (Figure 6A). We transfected synthetic tsRNA-Arg-CCG oligonucleotides into human endothelial cells, HMEC-1 cells (
). Norther blot was then performed to confirm the successful overexpression of tsRNA-Arg-CCG in HMEC-1 cells (Figure 6B). Interestingly, we found that overexpression of synthetic tsRNA-Arg-CCG led to increased expression of several pro-atherogenic genes including IL-6, IL-1α, ICAM-1, VCAM-1, and MCP-1 in HMEC-1 cells (Figure 6C). Therefore, tsRNA-Arg-CCG may have pro-atherogenic properties in vivo, which warrants further investigation.
Figure 6tsRNA-Arg-CCG affects pro-atherogenic gene expression in human endothelial cells in vitro. (A) Sequences of human and murine tRNA-Arg-CCG and tsRNA-Arg-CCG (top), and representative figure of tRNA-Arg and tsRNA-Arg-CCG indicated with a red line (bottom; black arrowheads indicating cleavage site). (B and C) Human endothelia cells, HMEC-1 cells were transfected with synthetic tsRNA-Arg-CCG oligo. The expression levels of tsRNA-Arg-CCG after transfection was assessed with Northern blot (B). Expression levels of indicated genes were analyzed by QPCR (C) (n=4; **P<0.01; ***P<0.001).
sncRNAs are a major family of non-coding RNAs that play critical roles in numerous biological processes. Certain small RNA populations such as miRNAs and piRNAs have been extensively studied and have also been implicated in atherosclerosis (
). In addition to those well characterized sncRNAs, recent studies have revealed the wide existence and unexpected functions of new classes of sncRNAs including tsRNAs and rsRNAs (
). However, the functions of those non-canonical sncRNAs in atherosclerosis or CVD are mostly unknown. A major obstacle in discovering and studying these sncRNA is that the currently widely used RNA-seq protocol generates biased sequencing results and often fails to detect these species (
). Many sncRNAs including tsRNAs and rsRNAs bear RNA modifications that interfere with adapter ligation and reverse transcription processes during cDNA library construction process, leading to unsuccessful detection of tsRNAs and rsRNAs in various tissues and cells (
). For example, tsRNAs contain N1-methyladenosine (m1A), N3-methylcytidine (m3C), and N1-methylguanosine (m1G) modifications that hinder reverse transcription process during traditional cDNA construction (
). Additionally, some sncRNA species, including tsRNAs and rsRNAs, contain 3’ terminal modifications such as 5’-phosphate and 2’3’-cyclic phosphate that block adaptor ligation and escape traditional library construction (
). Therefore, the modified tsRNAs, rsRNAs, and other sncRNAs may escape library construction and remain undetected by traditional RNA-seq. To overcome this obstacle, we recently developed PANDORA-seq based on a combination of enzymatic treatments (e.g., T4PNK and AlkB treatments) with optimized protocols that improv both RNA 3’ and 5’ adapter ligation and reverse transcription during cDNA library construction (
). PANDORA-seq has major advantages over previous methods designed to target either adapter ligation or reverse transcription processes alone, leading to the identification of abundant modified tsRNAs and rsRNAs in various tissues and cells (
). In the current study, PANDORA-seq unveiled a rsRNA- and tsRNA-enriched sncRNA landscape in the atherosclerotic intima of LDLR-/- mice which was strikingly different from that detected by traditional RNA-seq. While miRNAs were the dominant sncRNAs detected by traditional RNA-seq (e.g., ∼56% of total sncRNAs in HCD-fed mice), PANDORA-seq detected a substantially different sncRNA population with rsRNAs and tsRNAs accounting for >82% of total sncRNAs. In addition, traditional RNA-seq only detected a few differentially expressed rsRNAs and tsRNAs induced by HCD feeding but PANDORA-seq detected more than 1,300 differentially expressed rsRNAs and tsRNAs. Thus, by overcoming RNA modification-elicited limitations, PANDORA-seq revealed the hidden rsRNAs and tsRNAs population associated with atherosclerosis development undetected by traditional RNA-seq.
To identify novel sncRNAs associated with atherosclerosis development, we fed LDLR-/- mice a low-fat LCD or HCD for 9 weeks. As expected, HCD feeding resulted in severe hypercholesterolemia and large atherosclerotic lesions in LDLR-/- mice. The atherosclerotic lesions of LCD-fed mice were quite small, partially due to the relatively short feeding period (9 weeks) and young age of the mice (12 weeks old) at the time of euthanasia. Previous studies using the same LCD but longer feeding period resulted in larger atherosclerotic lesions (
Induction of atherosclerosis by low-fat, semisynthetic diets in LDL receptor-deficient C57BL/6J and FVB/NJ mice: comparison of lesions of the aortic root, brachiocephalic artery, and whole aorta (en face measurement).
). However, the relatively big difference of atherosclerotic lesion sizes between LCD and HCD-fed mice is ideal for the current study. Aiming to identify genes and sncRNAs associated with atherosclerosis development, we also focused on the intima of hypercholesterolemic LDLR-/- mice and performed transcriptome and sncRNA analyses via high-throughput RNA-seq and PANDORA-seq. Previous studies have used whole aortic tissues for traditional RNA-seq (
Aortic Gene Expression Profiles Show How ApoA-I Levels Modulate Inflammation, Lysosomal Activity, and Sphingolipid Metabolism in Murine Atherosclerosis.
). Since intima is the main site for atherosclerotic plaque initiation and development, we chose to isolate RNAs from the intimal for further analyses. As expected, transcriptomic analysis revealed enriched pathways for inflammatory and immune responses, which is consistent with previous studies on whole atherosclerotic arteries and the well-established role of the immune response in atherosclerosis (
Aortic Gene Expression Profiles Show How ApoA-I Levels Modulate Inflammation, Lysosomal Activity, and Sphingolipid Metabolism in Murine Atherosclerosis.
). For example, the C-C motif chemokine ligand family (CCL2, CCL3, CCL4 and CCL5) associated with increased inflammatory responses in atherosclerosis (
), were upregulated in the intima of HCD-fed LDLR-/- mice. Our transcriptomic results are consistent with known functions of the genes and pathways in atherogenesis based on the previous human and rodent studies (
Aortic Gene Expression Profiles Show How ApoA-I Levels Modulate Inflammation, Lysosomal Activity, and Sphingolipid Metabolism in Murine Atherosclerosis.
Transcriptome analysis of genes regulated by cholesterol loading in two strains of mouse macrophages associates lysosome pathway and ER stress response with atherosclerosis susceptibility.
Noncoding RNAs in Cardiovascular Disease: Current Knowledge, Tools and Technologies for Investigation, and Future Directions: A Scientific Statement From the American Heart Association.
). Our PANDORA-seq results also revealed differentially expressed miRNAs that are consistent with traditional-seq results and previous reports. For example, miRNA-146, which targets the 3’UTR region of TRAF6 to regulate NF-κB activation, has been shown to be associated with atherosclerosis development or CVD in humans (
Expression of miR-146a/b is associated with the Toll-like receptor 4 signal in coronary artery disease: effect of renin-angiotensin system blockade and statins on miRNA-146a/b and Toll-like receptor 4 levels.
). Consistently, we found miRNA-146b was upregulated in the intima of HCD-fed LDLR-/- mice as compared with LCD-fed mice. In addition, we also found that miRNA-31 was upregulated in atherosclerotic intima of HCD-fed LDLR-/- mice. miRNA-31 is expressed in both endothelial cells (
Cutting edge: TNF-induced microRNAs regulate TNF-induced expression of E-selectin and intercellular adhesion molecule-1 on human endothelial cells: feedback control of inflammation.
Cutting edge: TNF-induced microRNAs regulate TNF-induced expression of E-selectin and intercellular adhesion molecule-1 on human endothelial cells: feedback control of inflammation.
), miRNAs can then be readily detected by traditional-seq and becomes the major small RNA species in traditional-seq results. However, these results are misleading as traditional-seq failed to detect sncRNAs with extensive terminal/internal modifications (
), and demonstrated the advantages of PANDORA-seq enabled by overcoming RNA modifications which generate sequencing bias in the traditional-seq method. In addition to miRNAs, PANDORA-seq can detect other highly modified rsRNAs and tsRNAs which were not detected by traditional-seq (
). Our PANDORA-seq results showed that percentage of miRNA reads is very small (∼0.4-0.6%) but rsRNAs and tsRNAs are much more abundant as compared to miRNA reads, which is consistent with our recent studies (
). However, these results do not mean that PANDORA-seq failed to detect miRNAs, but rather more objectively represent the true compositions of sncRNA population in the intima.
As compared to miRNAs, little is known about the function of tsRNAs and rsRNAs in CVD or atherosclerosis (
Noncoding RNAs in Cardiovascular Disease: Current Knowledge, Tools and Technologies for Investigation, and Future Directions: A Scientific Statement From the American Heart Association.
). However, the role of the tsRNAs in atherosclerosis or CVD are mostly unknown. In the current study, PANDORA-seq but not traditional-seq identified many tsRNAs affected by HCD feeding in the intima of LDLR-/- mice. We found that one of HCD-induced intimal tsRNAs, tsRNA-Arg-CCG may play a role in regulating pro-atherogenic gene expression in endothelial cells in vitro. It is also important to note that tsRNA modifications can affect their secondary structure and function (
). Although we demonstrated the pro-atherogenic properties of synthetic unmodified tsRNA-Arg-CCG in vitro, endogenous tsRNA-Arg-CCG with modifications may exert stronger or different phenotypes in vivo. Future studies are required to understand the functions of those tsRNAs in endothelial cell function and atherosclerosis and CVD.
As a low-input and high-throughput method, PANDORA-seq allows us to detect a vast amount of small RNAs from a low amount of RNA input from the intima. Nevertheless, a potential limitation of PANDORA-seq is the loss of RNA products during multiple RNA purification steps, which make it difficult to reach the level of single-cell library preparation. To address this limitation in the future, one potential solution is to improve the enzymes used in PANDORA-seq, which allows the library construction process to be performed in a one-pot reaction, thereby reducing the number of RNA purification steps and minimizing RNA loss. Combining the improved PANDORA-seq method with fluorescence-activated cell sorting would enable investigators to perform single-cell analysis for detecting cell type-specific highly modified sncRNAs. Future studies using single-cell PANDORA-seq would increase our understanding of the cell-type specific role of understudied sncRNAs in atherosclerosis.
In summary, we used a novel PANDORA-seq to identify sncRNAs associated with atherosclerosis development in LDLR-/- mice. By overcoming RNA modification-elicited limitations, PANDORA-seq substantially increased the reads of rsRNAs and tsRNAs, and revealed a rsRNA- and tsRNA-enriched sncRNA landscape in the atherosclerotic intima of LDLR-/- mice. We also found that one of HCD-induced intimal tsRNA-Arg-CCG has pro-atherogenic properties and may contribute to endothelial cell dysfunction and atherosclerosis development. To our knowledge, this is the first study to utilize PANDORA-seq to investigate sncRNAs in atherosclerosis or CVD. Our results suggest that the understudied rsRNAs and tsRNAs are much more abundant than miRNAs in atherosclerosis-prone tissues. Findings from our studies will hopefully stimulate further investigations of the functions of these previously underexplored sncRNAs in atherosclerosis or CVD.
Data availability
The RNA-Seq datasets have been deposited in the Gene Expression Omnibus (GSE213305). Data supporting the plots within this paper and other findings of this study are available from the corresponding author upon request.
Acknowledgements
The authors thank all lab members for their technical assistance and Dr. Wenxin Zhao for preparing recombinant AlkB. This work was supported in part by National Institutes of Health (NIH) grants (R01HL167206, R01HL131925, R01ES023470, R01ES032024, and R35GM128854) and an American Heart Association (AHA) grant (19TPA34890065). R.H. was supported by an NIH National Research Service Award T32 training grant (T32ES018827) and an AHA predoctoral fellowship (23PRE1018751). The content is solely the responsibility of the authors and does not necessarily represent the official views of NIH and AHA.
Noncoding RNAs in Cardiovascular Disease: Current Knowledge, Tools and Technologies for Investigation, and Future Directions: A Scientific Statement From the American Heart Association.
Deep sequencing of RNA from immune cell-derived vesicles uncovers the selective incorporation of small non-coding RNA biotypes with potential regulatory functions.
Akiyama, Y., S. M. Lyons, M. M. Fay, T. Abe, P. Anderson, and P. Ivanov. 2019. Multiple ribonuclease A family members cleave transfer RNAs in response to stress. biorxiv.
Induction of atherosclerosis by low-fat, semisynthetic diets in LDL receptor-deficient C57BL/6J and FVB/NJ mice: comparison of lesions of the aortic root, brachiocephalic artery, and whole aorta (en face measurement).
Aortic Gene Expression Profiles Show How ApoA-I Levels Modulate Inflammation, Lysosomal Activity, and Sphingolipid Metabolism in Murine Atherosclerosis.
Transcriptome analysis of genes regulated by cholesterol loading in two strains of mouse macrophages associates lysosome pathway and ER stress response with atherosclerosis susceptibility.
Expression of miR-146a/b is associated with the Toll-like receptor 4 signal in coronary artery disease: effect of renin-angiotensin system blockade and statins on miRNA-146a/b and Toll-like receptor 4 levels.
Cutting edge: TNF-induced microRNAs regulate TNF-induced expression of E-selectin and intercellular adhesion molecule-1 on human endothelial cells: feedback control of inflammation.
00025-1&id=gr1.jpg){kind=link}
00025-1&id=gr2.jpg){kind=link}
00025-1&id=gr3.jpg){kind=link}
00025-1&id=gr4.jpg){kind=link}
00025-1&id=gr5.jpg){kind=link}
00025-1&id=gr6.jpg){kind=link}