In our prior study, we demonstrated distinct expression patterns in liver sinusoidal endothelial cells (LSECs) when compared to other endothelial cells (Jamil et al., 2020). However, microRNAs were not previously analyzed in human LSECs. In our current investigation, we identified 72 out of 411 microRNAs that displayed significant differential expression (p < 0.05) between LSECs and other endothelial cells. Among these, 35 were found to be overexpressed, while 37 were under-expressed in LSECs compared to other endothelial cells (Figure 1A, Supplementary Table S1).

Upon comparing LSECs with individual endothelial cell types (HCMEC, HPAEC, and HPMEC), we identified several differentially expressed microRNAs. Notably, 11 microRNAs were found to be commonly differentially expressed in all three comparisons (LSECs-HPAEC, LSECs-HPMEC, LSECs-HCMEC) (Figure 1B Right, Supplementary Table S2). Furthermore, seven of these microRNAs were observed to target 427 out of 2,547 differentially expressed genes (DEGs) identified in our previous study comparing LSECs and other endothelial cells at p < 0.05 (Supplementary Table S3) (Jamil et al., 2020). The correlation analysis of these 11 microRNAs with F8 expression revealed a strong negative correlation for “miR-30c-2-3p” and a significantly positive correlation with “miR-122-5p” (Figure 1B Left). Pathway analysis of the 427-targeted genes (over/under expressed) indicated decreased activity in pathways related to “Induction of Apoptosis by HIV1” and “Osteoarthritis,” whereas the “AMPK Signaling” pathway displayed significant activation (Figure 1C).

Hepatocytes constitute 55%–65% of liver mass. Comparing hepatocytes (HH) with LSECs can offer insights into the specific microRNAs that define LSEC identity. We identified 193 differentially expressed microRNAs (134 upregulated and 59 downregulated in LSECs) between LSECs and HH at p < 0.05 (Figure 2A, Supplementary Table S4). Among these, 44 microRNAs (39 upregulated and 5 downregulated in LSECs compared to other endothelial cells) were common with differentially expressed microRNAs found when comparing LSECs with hepatocytes and other endothelial cells (Figure 2B). Overexpressed microRNAs (39 microRNAs) exhibited the highest expression in LSECs when compared to both hepatocytes and other endothelial cells (Figure 2C Right), while the mean expression of under-expressed microRNAs (5 microRNAs) in LSECs fell between that of hepatocytes and other endothelial cells (Figure 2C Left). The top 10 differentially expressed microRNAs in LSECs compared to hepatocytes and other endothelial cells included miR-122-5p, miR-192-5p (liver-specific microRNAs) (Chang et al., 2004; Liang et al., 2007; Coll et al., 2015), miR-19b-3p, miR-30c-2-3p (F8/HA associated microRNAs) (Jankowska et al., 2020a; Atreya et al., 2020), and several other microRNAs (miR-429, miR-200b-3p, miR-1185-5p, miR-200a-3p, miR-30a-5p, and miR-216b-5p) (Figure 2D).

The top 100 microRNAs expressed in LSECs consist of 54 upregulated and 46 downregulated microRNAs when compared to other endothelial cells (ECs) (Figure 3A, Supplementary Table S5). Intriguingly, 42 out of these top 100 microRNAs were also identified among the top 100 expressed microRNAs in a study by Oda et al. in 2017, which focused on rat LSECs samples (GSE97229) (Oda et al., 2018) (Supplementary Tables S5, S6).

Moreover, we observed that 75 out of these 100 top microRNAs target a substantial number of differentially expressed genes (DEGs), specifically 1,293 out of 2,547, as identified in our prior study comparing LSECs and other ECs (Jamil et al., 2020). Remarkably, the expression of these microRNAs exhibited a significantly negative correlation with the expression of only 252 genes (Figure 3B). Pathway analysis of these negative correlated-targeted genes highlighted the enrichment of specific signaling pathways, notably “SAPK/JNK Signaling” and the “T-Cell Exhaustion Signaling Pathway” (Figure 3C). Additionally, other significant pathways identified included “Induction of Apoptosis by HIV1” and “IL-15 Production.” These pathways are linked to an immune response which is known functionality for LSECs (Jamil et al., 2020). Similarly, stress or apoptosis-related pathways are linked to chronic liver diseases which are related to irregular FVIII expression (Pradhan-Sundd et al., 2021).

Considering that hepatocytes are recognized as non-FVIII-producing cells (Shahani et al., 2014), this section exclusively focuses on the analysis of LSECs and endothelial cells in relation to F8-targeting microRNAs. We incorporated all three LSEC samples, notably including two samples (samples B and C) exhibiting relatively high F8 expression and one sample (sample A) displaying lower F8 expression (Figures 4A, B).

Our analysis, utilizing the IPA database, identified 106 microRNAs that target F8. Out of these, only twenty-five microRNAs were expressed in our microRNA-sequencing data (Figures 4B, C, Supplementary Tables S7, S8), and these microRNAs were capable of distinguishing LSECs from other endothelial cells. A 3D-PCA and heatmap analysis revealed that among the three LSEC samples, one (Sample A) exhibited overexpression of a majority of microRNAs targeting F8 (Figure 4B Left and Middle, Supplementary Table S7). Notably, this LSEC sample demonstrated significantly lower levels of F8 expression, even lower than that observed in other ECs (Figure 4A).

Furthermore, we observed a significant positive correlation (R ² = 0.68) between the first principal component analysis (PCA) of microRNAs targeting F8 and the expression of F8 itself (Figure 4B Right panel). Conducting a differential expression analysis of microRNAs between LSECs and other ECs unveiled four potential microRNAs (miR-204-3p, miR-2682-5p, miR-122-5p, and miR-214-3p) that not only distinctively separated LSECs from other ECs but most probably post-transcriptionally regulated F8 (Figures 4C, D). Additionally, it was noted that these microRNAs also regulated other transcription factors (TFs), such as ETS1, MAX, ARNT2, and FOXP1, known to bind to the F8 promoter (Figure 4E).

We selected two microRNAs, miR-122-5p (positively correlated) and miR-30-c (negatively correlated) with F8 expression levels, to assess their impact on primary LSECs. Figure 5 illustrates that both of these selected microRNAs have an effect on F8 expression, albeit to varying degrees. To investigate this, we chose one mimic and one inhibitory molecule for each of the two selected microRNAs.

As anticipated, the miR-122-5p inhibitory molecule significantly reduced F8 expression in LSECs compared to LSECs-WT (p < 0.01, logFC = −1.67). However, we observed a statistically non-significant overexpression of F8 for the mir-30-c inhibitory molecule. Furthermore, there was a slight, non-significant under-expression observed for both the miR-30-c mimic and miR-122-5p mimic.

After 48 h following transfection, we evaluated the functional impact of the miRNA on primary LSECs by conducting quantitative real-time PCR (QRT-PCR) to assess the expression levels of targeted genes. To perform this analysis, we extracted total RNA from the transfected cells using the Invitrogen PureLink RNA Mini Kit (Product Code: 10359103). Subsequently, we used 10 ng of total RNA for a one-step RT-PCR assay to measure F8 expression using an ABI 7500 real-time PCR system. Specific primers for F8 (Hs00252034_m1, Thermo Fisher, Catalog No. 4331182) were utilized, along with β-actin primers as an internal control. The amplifications were carried out utilizing 25x RT-PCR Master Mix (AgPath-ID™ One-Step RT-PCR Reagents, Applied Biosystems by Thermo Fisher) and 0.6 pmol of each primer. The PCR protocol involved an initial step at 95 °C for 5 min, followed by 40 cycles of denaturation at 95 °C for 30 s, annealing at 60 °C for 30 s, extension at 72 °C for 45 s, and a final extension step at 72 °C for 7 min. To ensure the reliability and reproducibility of our findings, we conducted all experiments in triplicate, with biological triplicates being included. Ct-values from the RT-PCR were transformed into fold changes using the 2^−ΔΔCT method. Subsequently, the data underwent statistical analysis using appropriate tests, and the results were presented as the mean ± confidence interval (CI).

Primers:

Beta-Actin-F: ACC​TTC​TAC​AAT​GAG​CTG​CG,

Beta-Actin-R: CCT​GGA​TAG​CAA​CGT​ACA​TGG,

Probe Beta-Actin: JOE-ACCTGGGTCATCTTCTCGCGGTTG-BHQ1.

The microRNAs were prepared and sequenced using the “TruSeq Small RNA Library Preparation Kits | Sequence micro and small RNAs” kit from Illumina (Prep Kit Code: RS-200-0012). The quality of the sequence data was assessed using the “fastq” software. To ensure data integrity, we utilized Cutadapt for the trimming of adapter sequences and removal of low-quality reads. For mapping the sequences, we used the “miARMA-seq” pipeline on the GRCh38 reference genome. The mature microRNAs GFF file necessary for feature count analysis was obtained from Ensemble. To facilitate downstream analysis, the count table generated by miARMA-seq was converted into FPKM (Fragments Per Kilobase Million) using Qlucore Omics Explorer 3.5. In order to focus our analysis on meaningful microRNAs, those with very low read counts across all samples were filtered out prior to conducting differential microRNAs expression analysis. Finally, the assessment of differential microRNA expression was carried out using Qlucore Omics Explorer 3.5, employing a statistically significant threshold of p < 0.05. This rigorous approach ensured the robustness and reliability of our microRNAs expression analysis.

We used the microRNA target filter function within the Ingenuity Pathway Analysis (IPA) to identify potential microRNA targets. Since microRNAs are well-known negative regulators, we specifically focused on microRNAs-mRNA interactions with a negative correlation or showing negative regulation. The subsequent processing of the microRNAs-mRNA target data was carried out using the R platform, allowing for further in-depth analysis. To identify microRNAs that bind to the F8 gene, we utilized IPA. We then conducted a sequence match analysis by comparing these microRNAs with the 1808-base pair long 3′ mRNA sequence of F8, which we obtained from UCSC. For the microRNA sequences, we sourced the data from miRbase. Pairwise sequence alignment was performed using the EBI Emboss Matcher. This comprehensive approach, (Supplementary Table S8), allowed us to explore potential microRNAs interactions with the F8 gene and further our understanding of the regulatory mechanisms involved.

Pathway analysis was conducted using Ingenuity Pathway Analysis (IPA). To perform this analysis, we entered the differentially expressed genes into IPA, utilizing the mean difference for enrichment analysis and the calculation of activity Z-Score. This approach allowed us to gain valuable insights into the functional pathways and their regulatory activity associated with the observed gene expression changes.

Transcription factor was identified using TRANSFAC database. F8 promoter of 1,000 bp upstream and 200 bp downstream was entered into TRANSFAC match tool with a profile selected for vertebrate_non_redundant_minSum and data version 2020.2. Transcription factor and microRNAs interaction was identified using IPA analysis.

We identified potential transcription factors by utilizing the TRANSFAC database. Specifically, we analyzed the F8 promoter region, encompassing 1,000 base pairs upstream and 200 base pairs downstream of the F8 gene. This sequence was entered into the TRANSFAC match tool with a profile selected for vertebrate_non_redundant_minSum and using data from version 2020.2 of the database.

Furthermore, we conducted an analysis of the interactions between transcription factors and microRNAs using Ingenuity Pathway Analysis (IPA). This comprehensive approach allowed us to explore the regulatory networks involving transcription factors and microRNAs, providing insights into the intricate regulatory mechanisms associated with the F8 gene and related processes.

All statistical analyses were conducted within the statistical environment of R or GraphPad Prism 6. Specifically, for the generation of 3D-PCA (Principal Component Analysis) and heatmaps, we utilized Qlucore Omics Explorer 3.5. For other types of plots and data visualization, we used either R or GraphPad Prism 6.