Functional omics analyses reveal only minor effects of microRNAs on human somatic stem cell differentiation

The contribution of microRNA-mediated posttranscriptional regulation on the final proteome in differentiating cells remains elusive. Here, we evaluated the impact of microRNAs (miRNAs) on the proteome of human umbilical cord blood-derived unrestricted somatic stem cells (USSC) during retinoic acid (RA) differentiation by a systemic approach using next generation sequencing analysing mRNA and miRNA expression and quantitative mass spectrometry-based proteome analyses. Interestingly, regulation of mRNAs and their dedicated proteins highly correlated during RA-incubation. Additionally, RA-induced USSC demonstrated a clear separation from native USSC thereby shifting from a proliferating to a metabolic phenotype. Bioinformatic integration of up- and downregulated miRNAs and proteins initially implied a strong impact of the miRNome on the XXL-USSC proteome. However, quantitative proteome analysis of the miRNA contribution on the final proteome after ectopic overexpression of downregulated miR-27a-5p and miR-221-5p or inhibition of upregulated miR-34a-5p, respectively, followed by RA-induction revealed only minor proportions of differentially abundant proteins. In addition, only small overlaps of these regulated proteins with inversely abundant proteins in non-transfected RA-treated USSC were observed. Hence, mRNA transcription rather than miRNA-mediated regulation is the driving force for protein regulation upon RA-incubation, strongly suggesting that miRNAs are fine-tuning regulators rather than active primary switches during RA-induction of USSC.


4
Legends for Supplementary Tables   Supplementary Table S1-S2-S3. Quantified mRNAs in USSC and XXL-USSC and cluster-specific gene ontology (GO) term analysis. Table S1: Transcriptome of native and XXL-USSC. Single CPM values are included for each biological replicate (columns C-N). Table S2: Quantitative transcriptome data showed an overlap of 1797 proteins which have been quantified by label-free mass spectrometry approach (accession numbers, column D). Mean CPM (columns E-H), log2FC, p-values and q-values are listed. Native USSC are compared to XXL incubated USSC for each analysed time point (columns I-Q)). Table S3: Cluster-specific gene ontology (GO) term analysis. Two clusters derived from cluster analysis (see Fig. 1C) were analysed including 746 and 857 mRNAs, respectively (column A and B). Overview of all biological processes (column C) found to be enriched by DAVID database with FDR corrected p-value (q < 0.05). In addition, numbers of transcripts (column D) and percentage of transcripts from total transcripts included in the depicted biological process (column E) are listed.
Supplementary Table S4-S5-S6. Quantified proteins in USSC and XXL-USSC and cluster-specific gene ontology (GO) term analysis. Table S4: Quantitative comparison of native, 3d XXL and 7d XXL incubated USSC by label-free mass spectrometry approach. Only proteins identified with at least 2 unique peptides (column E) and a score of at least 20 (column F) are included (1864 proteins).
Label-free quantification (LFQ) intensities of all replicates are listed (columns G-W). Table S5: Overview of mean LFQ intensities from biological replicates for each analysed time point (columns G-I). Fold changes (log2FC, columns K, M, O), q-value (column P) of < 0.05 and p-value (columns Q-S) of < 0.01 after pairwise comparison were considered for significantly changed abundances. Table S6: Cluster-specific gene ontology (GO) term analysis. Two clusters derived from cluster analysis (see Fig. 2B) were analysed including 391 and 381 proteins, respectively (column A and B). Overview of all biological processes (column C) found 5 to be enriched by DAVID database with FDR corrected p-value (q < 0.05). In addition, numbers of proteins (column D) and percentage of proteins from total proteins included in the depicted biological process (column E) are listed.
Supplementary Tables S9 and S10. Bioinformatic target gene predictions of significantly regulated miRNAs.
Herein, predictions from each algorithm are given separately (columns F-Q), together with the sum of algorithms predicting a given target (column R). Of note, the column "Gene name" (official gene symbols) contains redundancies due to several RefSeq IDs often matching a single gene (i.e. transcription variants). Column S in both files summarises all predicted genes from column C after removal of redundancies.
The sheets "Crossmatch..." in both files present crosstables of regulated miRNAs and inversely regulated predicted target proteins. Regulated proteins in XXL-USSC (3d and/or 7d) are listed in columns A (3d XXL) and/or B (7d XXL) separately, sorted by 6 the number of predicting miRNAs. Predictions from all 12 algorithms of miRWalk 2.0 are used. The subsets of predictions from at least 5 algorithms were also used for network constructions (Fig. 5).
Supplementary Table S11-S12-S13-S14. Quantitative proteome analysis after miRNA mimic or hairpin inhibitor transfections. Table S11: Label-free quantitative proteome analysis of control (n.t. siRNA) and hsa-miR27a-5p mimic transfected USSC and subsequent XXL incubation for 3 days. Only proteins identified with at least 2 unique peptides (column E) and a score of at least 20 (column F) are included (1,550 proteins). Proteins which are predicted targets for hsa-miR-27a-5p (1,469 proteins) as well as the number of prediction algorithms are listed (column G and H). In addition, label-free quantification intensities of all replicates are shown (columns I-Q). Fold changes (FC >1.5) and a q-value of < 0.05 are considered for significantly changed abundances.
Here, 1,577 proteins are included (column F); 1,444 were predicted hsa-miR-34a-5p targets (columns G and H). Table S14: Summary the predicted target proteins significantly (q < 0.05, FC > 1.5) lower or higher abundant upon miRNA mimic or hairpin inhibitor transfection (see Fig.   7) together with the number of algorithms (out of 12) predicting the particular protein.
Of note, all one-algorithm predictions stem from algorithm RNAhybrid (bold numbers).
Proteins in bold green are inversely regulated in XXL-USSC vs. USSC (see Fig. 7).