Dysregulation of microRNA Modulatory Network in Abdominal Aortic Aneurysm

Abdominal artery aneurysm (AAA) refers to abdominal aortic dilatation of 3 cm or greater. AAA is frequently underdiagnosed due to often asymptomatic character of the disease, leading to elevated mortality due to aneurysm rupture. MiRNA constitute a pool of small RNAs controlling gene expression and is involved in many pathologic conditions in human. Targeted panel detecting altered expression of miRNA and genes involved in AAA would improve early diagnosis of this disease. In the presented study, we selected and analyzed miRNA and gene expression signatures in AAA patients. Next, generation sequencing was applied to obtain miRNA and gene-wide expression profiles from peripheral blood mononuclear cells in individuals with AAA and healthy controls. Differential expression analysis was performed using DESeq2 and uninformative variable elimination by partial least squares (UVE-PLS) methods. A total of 31 miRNAs and 51 genes were selected as the most promising biomarkers of AAA. Receiver operating characteristics (ROC) analysis showed good diagnostic ability of proposed biomarkers. Genes regulated by selected miRNAs were determined in silico and associated with functional terms closely related to cardiovascular and neurological diseases. Proposed biomarkers may be used for new diagnostic and therapeutic approaches in management of AAA. The findings will also contribute to the pool of knowledge about miRNA-dependent regulatory mechanisms involved in pathology of that disease.


Introduction
Abdominal aortic aneurysms (AAA) are segmental dilatations of the abdominal aorta measuring 50% greater than the proximal normal segment, or >3 cm in maximum diameter [1,2]. Screening programs launched in different populations reported the prevalence of AAA between 4% and 8% in general population of men aged 65-80 years [3] and lower (0.45%) in Asians [4]. AAA rupture is responsible for 0.3-0.4% of all death cases and approximately 1% of deaths among men above 65 years globally, causing 130,000 to 180,000 fatalities per year [5]. Although mortality of AAA is decreasing in the 21st Century in many countries including United States and United Kingdom (mostly due to introduction of more advanced endovascular and open surgery repair techniques and better risk factor management), in other countries (Hungary, Romania), AAA mortality is still increasing [6][7][8][9].
The specific mechanism initiating and leading to progression of AAA has not yet been elucidated; however, AAA development has been associated with a variety of infections like brucellosis, salmonellosis and tuberculosis, trauma and connective tissue disorders, Takayasu disease and Marfan syndrome. Identified risk factors for aneurysm development include older age, male gender, cigarette smoking, obesity, dysregulation of lipid levels, hypertension [2,[10][11][12] and genetic predisposition [13][14][15][16].
Patients with AAA may report nonspecific symptoms like abdominal and back pain; however, in many cases disease progress is asymptomatic. The prolonged course of asymptomatic phase of AAA provides a relatively long diagnostic window before rupture [17]. Despite many studies targeting circulatory biomarkers of AAA, there is still lack of robust molecular methods able to classify affected individuals with satisfactory precision [18][19][20].
MiRNA-dependent regulation of gene expression emerged as a new tool providing novel opportunities in diagnosis of AAA. MiRNAs are approximately 18-25 nucleotides long, non-coding and single-stranded RNAs, which exhibit gene expression regulating effect by binding to mRNA. The pairing effect of miRNA-mRNA interactions predominantly inhibits gene expression by repression of translation, destabilization and cleavage of mRNA [21,22]. Currently miRNAs are a particularly intensively studied with promising preliminary results opening door to novel diagnostic and treatment approaches [23].
Large studies comparing miRNA and gene expression patterns in patients with AAA and healthy individuals may provide novel biomarkers with good discriminative value improving our diagnostic capability of detecting aneurysm, its rates of progression and complications; however, their introduction to clinical practice requires further investigations [24][25][26][27].
Although differential expression of miRNAs in human AAA cases was reported in abdominal aortic tissue, whole blood, serum and plasma samples, deregulation of miRNA expression in PBMCs (peripheral blood mononuclear cells) was not extensively studied.
In the present study, we applied next generation sequencing to analyze miRNA and gene expression in PBMCs of AAA patients and healthy volunteers with a goal to find the most capable miRNA and gene expression biomarkers of AAA and to research a potential role of identified biomarkers in pathogenesis of AAA.
The study design, methodology, article structure and language have been inspired by our previous studies regarding deregulation of miRNA regulatory network in lower extremities arterial disease [28] and chronic venous disease [29].

Study Participants Characteristics
The study was performed in accordance with the Declaration of Helsinki. The study design was approved by the Ethics Committee of Medical University of Lublin ( 1 Mean ± SD, 2 range. Statistical significance (p) of differences between AAA and control group in age and body mass index were determined using two-sided Mann-Whitney U test, and in sex and smoking habits were determined using two-sided Fisher's exact test. AAA-Abdominal Aortic Aneurysm, "NA"-not applicable.

Study Material Preparation and Sequencing
The procedure of study material preparation and sequencing was conducted as previously described in [28].
Peripheral blood mononuclear cells (PBMCs) were isolated from whole blood specimens using density gradient centrifugation with Gradisol L reagent (Aqua-Med, Łódź, Poland). Proportions of white blood cells subpopulations in AAA group were obtained from venous blood morphology analysis results and were presented in Figure S1.
Small RNA fractions (for miRNA expression analysis) were isolated from PBMCs specimens of twenty eight AAA patients and nineteen control subjects using MirVana microRNA Isolation Kit (Ambion, Austin, TX, USA).
Total RNA specimens (for transcriptome analysis) were isolated from PBMCs samples of seven randomly selected AAA patients and seven randomly selected controls using TRI Reagent Solution (Applied Biosystems, Foster City, CA, USA).
Small RNA and transcriptome libraries were prepared using Ion Total RNA-Seq Kit v2, Magnetic Bead Cleanup Module kit, Ion Xpress RNA-Seq Barcode 01-16 Kit and sequenced on Ion 540 chips (all Life Technologies, Carlsbad, CA, USA) using Ion S5 XL System (Thermo Fisher Scientific, Waltham, MA, USA). Raw sequences of small RNA and transcriptomic libraries were aligned to 2792 human miRNAs from miRBase v21 (http://www.mirbase.org) and to 55,765 genes of hg19 human genome, respectively.

Statistical and Bioinformatical Analysis
Detailed description of methodology applied to statistical and bioinformatical analysis was provided in our previous study [28].
The differences of AAA and control groups in age and BMI were evaluated using two-sided Mann-Whitney U test (wilcox.test function in R), and in sex and smoking using Fisher's exact test (fisher.test function in R).
Statistical analysis of miRNA expression data (resulted from sequencing of small RNA libraries) and gene expression data (resulted from sequencing of transcriptome libraries) was performed using R environment (version 3.5.2, https://www.r-project.org). Analysis was conducted on biological replicates. Differential expression analysis was performed using DESeq2 and UVE-PLS (uninformative variable elimination by partial least squares) [30] methods implemented in DESeq2 1.18.1 (https: //bioconductor.org/packages/release/bioc/html/DESeq2.html) [31] and plsVarSel 0.9.3 (https://cran. r-project.org/web/packages/plsVarSel/index.html) [32] packages, respectively. MiRNA and gene transcripts found by DESeq2 method with p value < 0.05 after adjustment by Benjamini-Hochberg false discovery rate were considered as statistically significant. UVE-PLS analysis was performed for miRNA and gene expression data using 3 and 2 PLS components, respectively. UVE-PLS analysis was executed with 1,000 iterations and default cut-off threshold.
Functional analysis was performed for genes included in the network using DAVID (Database for Annotation, Visualization and Integrated Discovery) 6.8 database (https://david.ncifcrf.gov/) [40,41] and it's supporting resources: KEGG (Kyoto Encyclopedia of Genes and Genomes) pathway maps, the Reactome Database of Signaling Pathways, GAD (Genetic Association Database) and GO (Gene Ontology). As a background, default whole Homo sapiens genome was applied. All associated terms of KEGG, Reactome and GAD categories were harvested as well as associated GO terms with Expression Analysis Systematic Explorer (EASE) score < 0.05.

Results
The summary of research process is presented on Figure 1. J. Clin. Med. 2020, 9, x FOR PEER REVIEW 6 of 33

Results
The summary of research process is presented on Figure 1.

Primary Results
Results of libraries assessments, Ion Sphere Particles enrichment quality control and results of sequencing data primary analysis are shown in Tables S1 and S2. To assess sequencing data quality, MA plot, boxplot of Cook's distances across samples and histogram of p values frequency were performed for miRNA ( Figure S3) and transcriptome ( Figure S4) sequencing results.

Differential Expression Analysis of miRNA
Differential expression analysis of miRNA between 28 AAA patients and 19 non-AAA controls was performed using DESeq2 and UVE-PLS methods.
DESeq2 comparative analysis of the miRNA expression signatures revealed 1107 differentially expressed miRNA transcripts in AAA group. Altered expression of 187 miRNA transcripts was characterized by statistical significance (p < 0.05) after adjustment by the Benjamini-Hochberg false discovery rate. To limit false positive results, for further comparison with UVE-PLS results we selected 36 differentially expressed miRNA transcripts (for 32 mature miRNAs) with adjusted p < 0.0001 (Table S3).
UVE-PLS analysis has returned 75 informative miRNA transcripts (Table S4). The arrangement of prediction error and PLS components as well as cross-validated predictions versus measured values were presented on Figure S5.
In the next step, the set of 36 differentially expressed miRNA transcripts (p < 0.0001) identified by DESeq2 method and the set of 75 differentially expressed miRNA transcripts identified by UVE-PLS method as informative were compared on Venn diagram (Figure 2a). The comparison disclosed 33 miRNA transcripts selected by both methods (Figure 2a). Differential expression of these 33

Primary Results
Results of libraries assessments, Ion Sphere Particles enrichment quality control and results of sequencing data primary analysis are shown in Tables S1 and S2. To assess sequencing data quality, MA plot, boxplot of Cook's distances across samples and histogram of p values frequency were performed for miRNA ( Figure S3) and transcriptome ( Figure S4) sequencing results.

Differential Expression Analysis of miRNA
Differential expression analysis of miRNA between 28 AAA patients and 19 non-AAA controls was performed using DESeq2 and UVE-PLS methods.
DESeq2 comparative analysis of the miRNA expression signatures revealed 1107 differentially expressed miRNA transcripts in AAA group. Altered expression of 187 miRNA transcripts was characterized by statistical significance (p < 0.05) after adjustment by the Benjamini-Hochberg false discovery rate. To limit false positive results, for further comparison with UVE-PLS results we selected 36 differentially expressed miRNA transcripts (for 32 mature miRNAs) with adjusted p < 0.0001 (Table S3).
UVE-PLS analysis has returned 75 informative miRNA transcripts (Table S4). The arrangement of prediction error and PLS components as well as cross-validated predictions versus measured values were presented on Figure S5.
In the next step, the set of 36 differentially expressed miRNA transcripts (p < 0.0001) identified by DESeq2 method and the set of 75 differentially expressed miRNA transcripts identified by UVE-PLS method as informative were compared on Venn diagram (Figure 2a). The comparison disclosed 33 miRNA transcripts selected by both methods (Figure 2a). Differential expression of these 33 miRNA transcripts is visualized on heat-map with Euclidean clustering and PCA plot (Figure 2b,c, respectively). 1 According to miRBase 22 (http://www.mirbase.org/). Presented 33 miRNA transcripts give 31 mature miRNAs (miRNA IDs). The table presents p (after Benjamini-Hochberg false discovery rate correction) and fold change values resulted from DESeq2 analysis, PLS (partial least squares) coefficients resulted from UVE-PLS (uninformative variable elimination by partial least squares) analysis and areas under ROC (receiver operating characteristics) curves (ROC-area under curves (AUC)) resulted from ROC analysis. MiRNA transcripts were divided into upregulated and downregulated groups and ordered according to increasing p value.  Discriminative value of altered expression in selected 33 miRNA transcripts was assessed using ROC analysis. Calculated areas under the curve ranged between 0.981 and 0.795, indicating good ability of AAA classification (Table 2 and Table S5, Figure S6). Therefore, a set of 31 mature miRNAs (16 upregulated and 15 downregulated) encoded by selected 33 miRNA transcripts was proposed as the most promising miRNA biomarkers of AAA (Table 2).

Differential Expression Analysis of Genes
Transcriptomic analysis was performed for randomly selected 7 AAA patients and 7 non-AAA controls. Differential expression analysis of genes was performed using DESeq2 and UVE-PLS methods.
DESeq2 analysis revealed 26,816 differentially expressed genes in AAA group when compared to controls. Altered expression of 2238 genes resulted with statistical significance p < 0.05, after adjustment by Benjamini-Hochberg false discovery rate. To limit false positive results, a set of 155 differentially expressed genes with adjusted p < 0.0001 was chosen for further comparison with UVE-PLS results (Table S6).
UVE-PLS analysis disclosed 91 informative genes, which expression differentiated AAA and control groups (Table S7). Figure S7 presents the arrangement of prediction error and PLS components and also cross-validated predictions versus measured values.
The comparison between the set of 155 differentially expressed genes revealed by DESeq2 (p < 0.0001) and the set of 91 informative genes selected by UVE-PLS disclosed 51 genes common for both methods (Figure 3a). A potential of these 51 genes to differentiate AAA and control groups was evaluated by PCA analysis and Canberra clustering (Figure 3b,c, respectively).

Correlation Analysis
Demographical characteristics (age, BMI), clinical parameters (maximum aneurysm diameter, thrombus volume and aneurysm neck length) and expression data of 34 selected miRNA transcripts and 51 selected genes of AAA group were included to the correlation analysis (Table 4, broaden results are provided in Table S9 and S10). Among demographical and clinical characteristics, ROC analysis showed strong discriminative value of changed expression of these 51 genes with an area under the ROC curve varying from 0.939 to 1 (Table 3 and Table S8, Figure S8). Therefore, these 51 genes was considered as a panel of transcriptomic biomarkers of AAA (Table 3). Table 3. Set of 51 differentially expressed genes, which significance of differential expression was confirmed by DESeq2 analysis with p < 0.0001 and by uninformative variable elimination by partial least squares (UVE-PLS) analysis in patients with abdominal aortic aneurysm in comparison to controls.

No.
Gene  Deconvolution procedure revealed estimated proportions of 11 cell subpopulations in PBMCs specimens subjected to transcriptome analysis. The "quanTIseq" method enables to obtain comparisons between cell types and specimens ( Figure S9) and "MCPcounter" method provides further information in differences between samples ( Figure S10). Certain differences in proportions of 11 cell subpopulations between samples were noticed; however, further statistics suggests that cell subpopulations composition in PBMCs specimens had no significant influence on the study outcome.

Correlation Analysis
Demographical characteristics (age, BMI), clinical parameters (maximum aneurysm diameter, thrombus volume and aneurysm neck length) and expression data of 34 selected miRNA transcripts and 51 selected genes of AAA group were included to the correlation analysis (Table 4, broaden results are provided in Table S9 and S10). Among demographical and clinical characteristics, statistically significant and weak positive correlation was indicated between age and maximum aneurysm diameter (R = 0.42, p = 0.025), as well as between BMI and thrombus volume (R = 0.38, p = 0.045) (Table S9). Two upregulated miRNAs: hsa-miR-34a-5p and hsa-miR-574-5p were positively correlated with maximum aneurysm diameter and thrombus volume, respectively, what make them potential targets for AAA prognosis. Hsa-miR-769-5p and hsa-miR-7847-3p are associated with age, pointing them as possible age-associated risk factors of AAA. Statistically significant correlations between genes and age (AC092620.2, PDCD4, SNHG5, SUFU, ZRANB2) as well as BMI (GIT2, RP1-102E24.1, RPL3P9) were also revealed ( Table 4). Relatively low number of samples does not allow to make categorical conclusions, therefore further studies with larger populations should be performed to confirm these results. To evaluate effects of smoking, coronary artery disease, diabetes mellitus type 2 and hypertension presence in AAA group on miRNA and gene expression, DESeq2 method was applied to find differentially expressed miRNAs and genes in AAA subjects with these conditions in comparison to AAA subjects without them and any of analyzed miRNAs and genes was statistically significantly differentially expressed. This result suggests, that miRNAs and genes identified as potential biomarkers of AAA do not depend on smoking, coronary artery disease, diabetes mellitus type 2 and hypertension; however, this finding should be confirmed in further studies. Table 4. Correlation analysis between maximum aneurysm diameter, thrombus volume, aneurysm neck length, age, body mass index (BMI) and expression of 33 selected miRNA transcripts and 51 selected genes identified as potential abdominal aortic aneurysm signatures. MiRNA transcripts and genes with at least one statistically significant correlation (p < 0.05) were presented. All correlations results are provided in Table S9 and S10 in Supplementary File.

In Silico Identification of miRNA:Gene Interactions
Identification of validated and predicted miRNA:gene interactions between 31 miRNAs and 51 genes revealed as potential biomarkers of AAA was processed by multiMiR package. In the analysis, five validated miRNA:gene pairs (Table S11) and 60 top 10% predicted miRNA:gene pairs (Table S12) were returned. Received interactions were visualized on regulatory network generated using Cytoscape 3.5.1 software (Figure 4).

No information
Gene Ontology terms associated with EASE score <0.05

GO Biological Process
Regulation of transcription, DNA-templated, regulation of nucleic acid-templated transcription, regulation of RNA biosynthetic process, regulation of RNA metabolic process, nucleic acid-templated transcription, RNA biosynthetic process, regulation of cellular macromolecule biosynthetic process, regulation of nucleobase-containing compound metabolic process, regulation of macromolecule biosynthetic process, regulation of cellular biosynthetic process, regulation of biosynthetic process, regulation of gene expression, nucleobase-containing compound biosynthetic process, regulation of nitrogen compound metabolic process, heterocycle biosynthetic process, aromatic compound biosynthetic process, organic cyclic compound biosynthetic process, RNA metabolic process, cellular nitrogen compound biosynthetic process, cellular macromolecule biosynthetic process, nucleic acid metabolic process, macromolecule biosynthetic process, gene expression GO enrichment analysis assigned upregulated genes (ANKRD13D, CPT1A, GGT1, GIT2, HTT, NBEAL2, SUFU, THOC5, UBE4B, UPF1, ZSWIM8, and ZZEF1) to positive regulation of cellular catabolic process, mRNA transport and developmental processes, while downregulated genes (KIAA1549L, PDCD4, PRDM13, SNORA60, SNORD94, and ZRANB2) were associated with RNA biosynthesis and gene expression (Table 5).

Discussion
Searching for precise and robust biomarkers of early AAA is crucial due to very subtle symptoms of the disease and high mortality of ruptured aneurysm. Examining deregulations in miRNA network and consequential effects on gene expression appears as an interesting research tactics for finding novel biomarkers of AAA [25][26][27].
In the presented study, we performed integrated analysis of miRNAome and transcriptome expression in PBMC specimens obtained from patients with AAA and healthy controls. The PBMCs pool is involved in inflammation, an important element of AAA pathology, and therefore, should provide an abundance of information about condition and disorders of vascular system. Moreover, high accessibility of PBMCs facilitates translation of obtained results into clinical practice.
Application of next generation sequencing and multi-stage statistical methodology allowed to select 31 miRNAs (Table 2) and 51 genes (Table 3) as the most favorable candidates for detection of AAA (Figure 1). Selection of proposed biomarkers was preceded by control of potential false positive results through adopting the higher threshold of statistical significance (p < 0.0001, adjusted by Benjamini-Hochberg false discovery rate) and eliminating uninformative variables using UVE-PLS. High diagnostic value of proposed biomarkers was confirmed in ROC analysis (Tables 2 and 3, Tables  S5 and S8, Figures S6 and S8). Such stringent criteria applied for biomarkers selection were introduced due to the inability to predict in advance the number of miRNA/genes that should be validated by qPCR, thus allowing us not to design proper experiment accordingly.
There is a limited number of studies regarding miRNA expression investigations by next generation sequencing in PBMC samples in AAA individuals [26]. The comparison of our results and findings obtained in similar studies is presented in Table 6. Relatively poor overlap with literature data and our miRNAs list could be explained by differences in methodology applied and biological material subjected to experiments. In depth discussion regarding functions of DEMs (differentially expressed miRNAs) of this research vastly exceeds capacity of the current paper. Table A1 in Appendix B focuses on most important information regarding possible mechanisms of presented miRNAs actions in AAA. The set of revealed miRNAs clearly points to deregulation of numerous signaling pathways like mTOR, PI3K/AKT, TGF-β, NOTCH, MAPK, and NF-κB. Those affect general processes like cell adhesion, proliferation and motility as well as more detailed ones like wound healing and vascular growth. Taken together, this may point to processes engaged in AAA onset and development like vascular wall remodeling, hypoxia, subsequent revascularization and hemorrhage. On the other hand it is worth to draw attention to selected miRNAs, due to their engagement in regulation of genes, occurring in our sequencing data. The table presents studies on miRNAs in AAA. Differentially expressed miRNAs are revealed from AAA subjects and control groups. MiRNAs overlapping with biomarkers proposed in the current study are shown. For more comprehensive review of this topic please refer to [26].
Expression levels of miR-21 may vary between types of biological material. In AAA aortic tissue is significantly upregulated [46] whereas in plasma of AAA patients it may be downregulated [47]. We observed exhibited upregulation of miR-21, suggesting that PBMCs may better reflect processes ongoing in affected aortic tissue. Higher levels of miR-21 were observed in low inflammatory state abdominal aneurysms compared to high inflammation phase aneurysm [48]. On the other hand, elevated level of miR-21 in macrophages promote apoptosis and vascular inflammation in atherogenesis [49] and was shown as a biomarker of low extremities arterial disease [28]. Those facts suggests more detailed assessment of miR-21 function in inflammation, atherogenesis and AAA in the future.
Overexpression of miR-21 leads to downregulation of PDCD4 and PTEN inducing cell proliferation, decreasing apoptosis in the aortic wall, alleviating aneurysm expansion and protect against cell injury caused by hydrogen peroxide exposure [50][51][52]. PDCD4 is involved in atherosclerosis pathology probably through enhancing levels of IL-6 and IL-8 and promoting apoptosis of VSMC in animal models of coronary atherosclerosis [53]. In our study, both upregulation of miR-21-5p and downregulation of PDCD4 were observed in PBMCs samples of AAA subjects, suggesting pro-inflammatory and antiapoptotic effects in AAA.
We confirmed findings of Lenk et al. [54] demonstrating upregulation of GGT1 in patients with AAA. This phenomenon was also reported as a signature of low extremities arterial disease (LEAD) [28], suggesting non-specific character of GGT1 deregulation.
There was no overlap between our findings and gene expression biomarkers of AAA found in some other studies [55][56][57]. The differences in results may stem from dissimilarities in study material, criteria for participants inclusion and methodological approaches.
Integration of miRNA and gene expression analysis enabled identification of miRNA:gene regulatory pairs in AAA, presented on regulatory network (Figure 4). Elevated expression of UBE4B may be caused by many miRNAs selected as biomarkers of AAA ( Figure 4) and lead to aneurysm expansion through inhibition of endothelial growth factor receptor (EGFR)-mediated proliferation of vascular cells by enhancing EGFR degradation [58]. Upregulation of ANKRD13D may affect endocytic trafficking of EGFR by inhibition of its ubiquitinated form from the cell surface, attenuating pro-proliferative signaling of internalized EGFR [59], potentially aggravating AAA. EGFR signaling might be compromised also by upregulation of miR-424-5p, which is a negative regulator of EGFR expression, as observed in tumor cells [60]. According to bioinformatic analysis, this miRNA might be a regulator of both UBE4B and ANKRD13D, affecting also EGFR expression [60]. This net of reciprocal regulations may be considered as a part of a mechanism decreasing cell proliferation in AAA through modulation of EGFR signaling.
The preliminary functional analysis of genes regulated by miRNAs revealed terms closely related to vascular pathology, including lipid metabolism, inflammation, atherosclerosis and aging (Table 5). Interestingly, there were seven genes associated with neurological disorders, including Alzheimer's disease (CPT1A, SUFU, ZSWIM8, PDCD4), schizophrenia (HTT, NBEAL2) and Parkinson's disease (PRDM13). For further comment on neurological relationships of our findings please refer to Appendix C.
The highly enriched terms associated with genes regulated by biomarker miRNAs like RNA biosynthesis and transport, positive regulation of catabolic processes and developmental processes suggest more general mechanisms also involved in control of gene expression in AAA.
We are aware of limitations of our study design. It was not established, whether alterations in expression of miRNAs and genes proposed as biomarkers were predictive or responsive to AAA development. Although proposed biomarkers were characterized by high predictive value, supported by high level of statistical significance, multistage selection and ROC confirmation, the clinical application requires confirmation in studies with larger cohorts. PBMCs consists of lymphocytes and monocytes subpopulations, varying in miRNA and gene expression patterns. Nuances in proportions of these subpopulations may affect diagnostic value of indicated biomarkers, thus this effect should be further investigated.
Co-existing diseases may bias evaluation of PBMCs expression profiles on a systemic scale. For this reason many conditions were established as exclusion criteria, as mentioned in experimental section. Such strict evaluation helped us to find expression patterns potentially reflecting the local changes in AAA; however, it entailed statistically significant differences in demographic characteristics between AAA and control groups (Table 1). Differences in gender, age and smoking habits might have potentially influenced the study outcome.
The AAA group in our study has men overrepresentation (89.3%, 25 patients) whereas control group is more sex-balanced and include 47% of healthy men (9 subjects) ( Table 1). Those characteristics may potentially introduce a gender-associated bias into our data. Some of proposed miRNA biomarkers of AAA have already been connected to sex differences in humans (Table A2 in Appendix D), suggesting that deregulation of these miRNAs may reflect gender differences occurring between AAA and control group. In the case of gene expression, comparison of Deegan et al. paper [61] draw only five overlapping genes, suggesting minor gender bias, while from [62] there were no such ones (Table A3 in Appendix E). Interestingly, Cui et al. discovered that blood is poor material for distinguishing sex associated transcriptome patterns due to relative invariability between genders [63].
The AAA and control groups were age-unmatched (66.39 ± 4.52 years and 36.58 ± 9.97 years, respectively) ( Table 1). Literature analysis revealed that some of proposed miRNA biomarkers of AAA may also be considered as senescence indicators (Table A2 in Appendix D). On the other hand, gene expression in AAA has no or little overlap with senescence/aging biomarkers. One of the proposed AAA biomarkers, SNORA33, has been previously associated with normal human aging [64]. In a comprehensive transcriptomic analysis of eight senescence in vitro models [65], 20 out of 51 genes reported in our study were also differentially expressed in senescent cells (Table A4 in Appendix F). It could be then possible that senescence characteristics present in our data may be the outcome of general stress(es) due to disease process itself, not only the age.
Smoking is AAA risk factor affecting miRNA expression [66] and in the case of our studies might be another bias-introducing factor, since AAA group includes current smokers (9 persons, 32.1%), while control group is devoid of them (Table 1). Only a small number of miRNAs indicated potential bias after comparison with literature (Table A2 in Appendix D). Other research regarding signatures of smoking did not provide us with any transcriptomic patterns recurring in our data (Table A3 in Appendix E). This suggests an absence of significant smoking-associated bias, probably due to relatively low number of current smokers in AAA group.
Despite of in-depth literature analysis we were unable to exclude unambiguously any miRNAs or gene transcripts from AAA biomarker panels. MiRNAs represent a group of RNAs with pleiotropic regulatory behavior. This does not exclude particular miRNA to act in various cellular processes in different spatiotemporal context. After detailed analysis, we noticed great variability of applied methodologies across literature (refer to literature in Tables A3 and A4). It is possible that existing discrepancies may reflect methodological rather than sex/age/smoking bias.
Taking all of this into account, it should be considered that presented data shows either real AAA biomarkers or biomarkers associated with both AAA and AAA predisposing factors.
Due to technical (data storage server capacity) and financial limitations, gene expression analysis was performed for 14 out of 47 study participants. It could be a potential source of bias affecting investigations on miRNA:gene regulatory network; however, we confirmed some previously validated interactions (Table S11) and predictive interactions with high probability (Table S12). More research with larger populations is needed to confirm our findings and to validate predictive targets.
The results obtained in our study confirm the important role of miRNA in the pathogenesis of AAA, opening a door to deeper understanding of miRNA functions and regulatory network. AAA biomarkers proposed in this research, after further validation in studies with larger and demographically matched cohorts, can be prospectively applied into clinics for differentiation, diagnosis and therapy of AAA.
Supplementary Materials: The following are available online at http://www.mdpi.com/2077-0383/9/6/1974/s1. Figure S1: proportions in white blood cells subpopulations in AAA patients resulted from blood morphology analysis. Figure S2: evaluation of differences between AAA group (n = 28) and the control group (n = 19) for age (A), BMI (B), sex (C) and smoking (D). Statistical significance (p values) were calculated using two-sided Mann-Whitney U test (for age and BMI) and a two-sided Fisher's exact test (for sex and smoking). Distributions of age and BMI were presented on boxplots (A) and (B), respectively, with whiskers defining region between minimum and maximum values, boxes covering values between 25% and 75% quantile and horizontal lines inside boxes marking median value. Distribution of sex (black-female, gray-male) and smoking (black-no-smokers, grey-smokers) were presented on spine plots (C) and (D), respectively. Figure S3: quality control of data obtained from sequencing of small RNA libraries and results of differential expression analysis performed by DESeq2 package in 28 AAA individuals and 19 healthy controls. (A) Boxplot presents Cook's distances of miRNAs across samples. Whiskers define range between minimum and maximum value of Cook's distance, boxes range between 25% and 75% quartile, horizontal lines inside boxes mark median value. (B) Histogram presenting distribution of DESeq2 p values. (C) MA plot showing relation between log2 of fold changes of differentially expressed miRNAs and averages of normalized counts. MiRNAs with p value < 0.1 were marked as red points. Figure S4: quality control of data obtained from sequencing of transcriptome libraries and results of differential expression analysis performed by DESeq2 package between group of 7 AAA individuals and a group of 7 healthy controls. (A) Boxplot presenting Cook's distances of miRNAs across samples. Whiskers define range between minimum and maximum value of Cook's distance, boxes range between 25% and 75% quartile, horizontal lines inside boxes mark median value. (B) Histogram presenting distribution of DESeq2 p values. (C) MA plot showing relation between log2 of fold changes of differentially expressed miRNAs and averages of normalized counts. MiRNAs with p value < 0.1 were marked as red points. Figure S5: UVE-PLS differential expression analysis of miRNA expression data of 28 AAA patients compared to 19 healthy controls. Plots present the arrangement of prediction error and PLS components (A) and cross-validated predictions versus measured values (B). Figure S6: results of receiver operating characteristics (ROC) analysis performed for selected 33 miRNA transcripts differentially expressed in AAA. Values of area under curves (AUC) with 95% confidence interval (in brackets) were included in each plot. Figure S7: UVE-PLS differential expression analysis of transcriptomic expression data of 7 AAA individuals compared to 7 healthy controls. Plots present the arrangement of prediction error and PLS components (A) and cross-validated predictions versus measured values (B). Figure S8: results of receiver operating characteristics (ROC) analysis of 51 genes selected as signatures of AAA. Values of area under curves (AUC) with 95% confidence interval (in brackets) were included in each plot. Figure S9: results of deconvolution procedure performed on gene expression datasets of 7 AAA patients and 7 controls using "quanTIseq" method implemented to immunedeconv 2.0.0 package. Figure S10: results of deconvolution procedure performed on gene expression datasets of 7 AAA patients (AAA) and 7 controls (control) using "MCPcounter" method implemented to immunedeconv 2.0.0 package. Two-sided Mann-Whitney test (wilcox.test function in R) was used to calculate statistical significance of differences in score values between AAA and the control group. Table S1: assessment of small RNA samples, small RNA libraries and results of primary analysis of small RNA sequencing data carried out with Ion Torrent small RNA Plugin v5.0.5r3. Table S2: assessment of transcriptome libraries and results of primary analysis of transcriptome sequencing data carried out with Ion Torrent RNASeqAnalysis plugin v.5.0.3.0. Table S3: the set of 36 differentially expressed miRNA transcripts in the group of 28 AAA individuals compared to 19 healthy controls, resulted from DESeq2 analysis with p < 0.0001. MiRNA transcripts were ordered according to increasing p value. Table S4: the set of 75 differentially expressed miRNA transcripts in the group of 28 AAA individuals compared to 19 healthy controls, resulted from UVE-PLS analysis. MiRNA transcripts were ordered according to decreasing PLS coefficients. Table S5: results of ROC analysis for 34 miRNA transcripts selected as signatures for AAA. Table S6: results of DESeq2 differential analysis of genes in the group of 7 AAA individuals compared to the group of 7 healthy controls. The table presents 155 differentially expressed genes with p < 0.0001, ordered accordingly to increasing p value. Gene names were searched using HUGO multi-symbol checker and gene symbols not assigned to names are termed as "unmatched". Synonyms or previous gene symbols are placed in brackets. Table S7: results of UVE-PLS analysis of miRNA in the group of 7 AAA individuals compared to the group of 7 healthy controls. The table presents 91 differentially expressed genes recognized as informative, ordered according to decreasing PLS coefficients. Gene names were searched using the HUGO Multi-Symbol Checker and gene symbols not assigned to names are termed as "unmatched". Synonyms or previous gene symbols are placed in brackets. Table S8: results of ROC analysis for 51 genes selected as signatures of AAA. Table S9: correlation analysis between maximum aneurysm diameter, thrombus volume, aneurysm neck length, age, BMI and expression of 33 selected miRNA transcripts identified as potential AAA signatures. Table S10: correlation analysis between maximum aneurysm diameter, thrombus volume, aneurysm neck length, age, BMI and expression of 51 selected genes identified as potential AAA signatures. Table S11: experimentally validated miRNA:gene pairs found in silico among 31 miRNAs and 51 genes indicative for AAA. Table S12:  Data Availability: All datasets generated for this study can be found in the FigShare repository (link will be provided upon acceptance).

Abdominal Aorta Aneurysm and Control Groups Construction
All 28 patients from analyzed abdominal aorta aneurysm (AAA) group were hospitalized in Department of Vascular Surgery of the Independent Public Clinical Hospital No. 1 in Lublin for AAA as the main diagnosis. The lack of comorbidities was a crucial condition to qualify the patient to the study to create as homogenous group as possible. During the recruitment processes, the medical history and comorbidities were analyzed three times by three different physicians: vascular surgeon, anesthetist (two weeks before admission) and then, one more time, at the day of admission by vascular surgeon in order to conclude on current health state before surgery. In this study, we screened 232 AAA patients during first stage of recruitment. A total of 198 patients was excluded due to chronic venous disease (C2-C6 stage), inflammatory aneurysm, false aneurysm, thoracic aorta aneurysm, isolated popliteal or iliac artery aneurysm, aortic and/or arterial dissection, stroke, transient ischemic attack, myocardial infarction, diabetes mellitus type I, symptomatic peripheral arterial disease (ABI < 0.8), connective tissue disorders including rheumatoid disease, impaired hepatic or renal function, corticoid therapy, infection within previous six weeks, deep venous thrombosis (less than one year), pulmonary embolism, inflammatory and/or infectious disease and cancer. In fact, at the age of 60-70 it is quite impossible to find AAA patients without any kind of comorbidities.
To construct the control group, we aimed to include healthy (confirmed by vascular surgeon) and non-smoking subjects. AAA mainly affects elderly men and often is accompanied with other diseases, what is the reason of presence of differences in age, BMI, gender and smoking between AAA patients and controls. Therefore, it is difficult to find age-and gender-matched control group including healthy subjects and to avoid differences in those features during the study participants' inclusion. We cope with this biological bias by extension of standard statistical methodology by noise reduction using advanced statistical methods (UVE-PLS), and strict significance thresholds (p < 0.0001 for miRNAs and genes). This particular group of subjects was used also in our both previous research as the control group [28,29].
Appendix B Table A1. Selected most prominent targets and processes affected by differentially expressed miRNAs from presented study, drawn from literature analysis.
hsa-miR-24 Downregulated in plasma of AAA patients and murine AAA models [67].
hsa-miR-122 Role in Alzheimer's disease through regulation of genes involved in lipid metabolism [68].

MiRNAs reported in the present study as downregulated in AAA
hsa-let-7g Increases viability of lung cancer and osteosarcoma cells via downregulation of HOXB1 and activation of NF-kB pathway [79,80].
hsa-miR-31 Knockdown of this miRNA inhibits expression of Collagen I and III and Fibronectin in hypertrophic scar formation [81], regulator of senescence in cancer cells [82].

Neurological Associations of Genes and miRNAs Involved in Abdominal Aortic Aneurysm (AAA) Pathology
Although our research was focused on AAA, some additional associations with nervous system diseases were discovered. Interestingly, there were seven genes associated with neurological disorders, like Alzheimer disease (CPT1A, SUFU, ZSWIM8, PDCD4), schizophrenia (HTT, NBEAL2) and Parkinson's disease (PRDM13) ( Table 5). SUFU and PDCD4 genes were linked to Alzheimer's disease and were shown as predictive targets of miR-34a-5p. Upregulation of miR-34a-5p was earlier reported in blood mononuclear cells from individuals with Alzheimer's disease [103]. Both SUFU and PDCD4 were also predictive targets of miR-122-5p involved in development of Alzheimer's disease through regulation of genes involved in lipid metabolism [68]. Another target for miR-34a-5p was PRDM13, which was found to be associated with susceptibility to Parkinson's disease. Upregulation of miR-34a was reported in in vitro model of Parkinson's disease [104]. The regulatory network including miR-34a-5p, miR-122-5p, SUFU, PDCD4, and PRDM13 may be involved in development of aneurysm-related neurodegenerative diseases. The link between nervous system and cardiovascular diseases is well known and was also confirmed in our earlier research [28]. hsa-miR-24 [106,107] hsa-miR-34a [108] hsa-miR-122 [106] hsa-miR-424 [63,105] [106]
Appendix F   Table A4. Associations of genes proposed in our study as AAA biomarkers with aging, found in the most relevant literature.

No. Gene Symbol
Associations with Aging: