Computational Identification and Characterization of New microRNAs in Human Platelets Stored in a Blood Bank

Platelet concentrate (PC) transfusions are widely used to save the lives of patients who experience acute blood loss. MicroRNAs (miRNAs) comprise a class of molecules with a biological role which is relevant to the understanding of storage lesions in blood banks. We used a new approach to identify miRNAs in normal human platelet sRNA-Seq data from the GSE61856 repository. We identified a comprehensive miRNA expression profile, where we detected 20 of these transcripts potentially expressed in PCs stored for seven days, which had their expression levels analyzed with simulations of computational biology. Our results identified a new collection of miRNAs (miR-486-5p, miR-92a-3p, miR-103a-3p, miR-151a-3p, miR-181a-5p, and miR-221-3p) that showed a sensitivity expression pattern due to biological platelet changes during storage, confirmed by additional quantitative real-time polymerase chain reaction (qPCR) validation on 100 PC units from 500 healthy donors. We also identified that these miRNAs could transfer regulatory information on platelets, such as members of the let-7 family, by regulating the YOD1 gene, which is a deubiquitinating enzyme highly expressed in platelet hyperactivity. Our results also showed that the target genes of these miRNAs play important roles in signaling pathways, cell cycle, stress response, platelet activation and cancer. In summary, the miRNAs described in this study, have a promising application in transfusion medicine as potential biomarkers to also measure the quality and viability of the PC during storage in blood banks.


Introduction
Platelet concentrate (PC) transfusions are widely used to save the lives of patients suffering from acute blood loss and are more often used in supportive prophylactic therapy for patients with various hematological diseases [1].
This blood component requires special storage in blood banks, normally being stored up to a maximum of five days at a temperature of 22 ± 2 • C, with gentle and continuous agitation, because even under ideal storage conditions, the PC can undergo modifications or degradations known as platelet storage lesion (PSL), a term introduced by Murphy et al., in 1971 [2], for describing the multifactorial mechanisms of this problem, which include the methods of collection, processing, storage, handling before or after collection, and expiration date [3][4][5]. In Brazil, the validity of the PC is three to seven We performed the reading mapping pipeline in genome and library mode, in which both modes used a common preprocessing step, mapping, expression profile, and miRNA prediction [28]. For this, the 3' adapter sequence was trimmed, and the sequence length distribution was analyzed. Sequences with a reading length between 15 and 27 nucleotides (nt) were aligned with the miRNA precursor sequences (pre-miRNA) of human miRBase, version 22 [29].
To this analysis, we applied the standard sRNAbench parameters as follows: (i) minimum length of adapter that needs to be detected, 10; (ii) alignment type, Bowtie seed alignment (GRCh38_p10_mp), seed length for alignment 20, minimum read count 2, minimum read length 15, allowed number of mismatches 1, the maximum number of multiple mappings 10; (iii) MEAN quality filtering was used, 20; and (iv) the annotation used miRNAs for species hsa (Homo sapiens).

MicroRNA Expression and Quality Analysis between Different PCs
We assessed the level of miRNA expression by counting per million reads (CPM) using the following formula: CPM = (reads number of one miRNA)/(total mapped reads to all annotated miRNAs) × 10 6 . We defined a miRNA expressed in a PC with a CPM > 1 in more than 50% of the six PC samples. The PC-specific miRNA was defined as a miRNA expressed exclusively in PC on the first day (PC-1), was used in this study as the platelet quality control.
Because PCs are stored in a blood bank for a maximum period that varies between three and five days, depending on the plasticizer in the conservation bag [6], our analyses considered the first day as the high-quality control of platelets and the seventh day corresponded to the low-quality control of platelets [20].
In our computational results obtained with sRNA-Seq, we analyzed miRNAs expressions according to the methodology described by Pontes et al. [20] to check which PC bags were in good condition for transfusion. This method compared the relative expression between miR-127 and miR-320a [30]. When miR-127 presented a lower expression (<80%) as compared with miR-320a, storage lesions in this blood component were considered, suggesting the blockage of the PC bag for transfusion.
The PC bag was considered to be suitable for transfusion when it had one of the following possibilities: (i) expression of miR-127 > miR-320a, (ii) equal expression between miR-127 and miR-320a, (iii) and a difference less than 20% in expression between these two miRNAs. In this study, the results generated for the expression levels of miRNAs were analyzed using computational biology simulations, employing unsupervised grouping and principal component analysis (PCA).

Validation of microRNAs by qPCR on 100 PC Units
Each PC which was used contained platelets from five healthy donors. Therefore, validation was performed in 500 donors (250 men and 250 women in the age group that comprises young adults, between 18 and 40 years old). All biosafety policy guidelines were applied in the involved laboratories, with the approval of the Ethics Committee (#194, 196, approval 17 October 2012).
Seven days of storage were analyzed, and the first day was used as the control for the following days, totaling five comparisons of expression levels for each of the six miRNAs identified, accounting for 30 analyses. As the validation occurred in 100 PC units, we totaled this validation in 3000 analyses. Mir-191 was selected as an internal control for miRNA input and reverse transcription efficiency because the miRNA was most highly expressed on seven different days of storage [20]. All real-time quantitative PCR (qPCR) reactions were performed in triplicate for both miRNAs. Expression levels of miR-486-5p, miR-92a-3p, miR-103a-3p, miR-151a-3p, miR-181a-5p, and miR-221-3p were isolated from 100 PC units with mirVana TM miRNA Isolation Kit (Thermo Fisher Scientific, Waltham, MA, USA).

IsomiR Annotation Analysis
In this analysis, we detected the miRNAs sequence variants called isomiR [31]. To detect isomiRs, we applied the following steps for the sRNAbench pipeline: (i) mapping the reads with the pre-microRNA genome or sequence using the Bowtie seed option [32], (ii) determining the coordinates of the mature microRNA, (iii) clustering of all reads that mapped within a window of the canonical sequence of the mature microRNA (miRBase), and (iv) applying a hierarchical classification of the variants [27].
Then, a subsequent analysis was performed to detect multiple NTA sequence variants which involved non-templated additions (enzymatically addition of a nucleotide to the 3 end, i.e., adenylation and uridylation) that included NTA(A), number of reads with a non-templated A (adenine) addition; NTA(U), number of reads with a non-templated U (uracil) and (T) thymine addition; NTA(C), number of reads with a non-templated C (cytosine) addition; and NTA(G), number of reads with a non-templated G (guanine) addition.
The second class of length variants included 5 and 3 trimming and extension in the following forms: lv3pE, number of reads with 3 length extension (longer than the canonical sequence); lv3pT, number of reads with 3 length trimming (shorter than the canonical sequence); lv5pE, number of reads with 5' length extension (longer than the canonical sequence); lv5pT, number of reads with 5 length extension (shorter than the canonical sequence); and mv, number of reads classified as multiple length variants [28]. We organized the results of this analysis by the ranking of isomiRs identified in PCs and mature miRNA most expressed according to individual reading counts (CPM expression).
For this analysis, we applied an information retrieval feature known as target mining in the miRWalk predictor, version 3.0 (http://mirwalk.umm.uni-heidelberg.de/), which hosts the three aforementioned predictors, to obtain the information from miRNA-gene interactions that were organized in a table with various predictive metrics, which were considered for binding probability (p > 0.95) [36], sites preferably conserved within the 3' UTR (untranslated region) and validated interactions [37].

Construction of the microRNA-Gene Interaction Network
The subset files containing the predicted and validated interactions, identified in the previous analysis, were used in the construction of miRNA-gene interaction networks. For that, we filtered only the predicted and validated interactions of interest to remove those annotated symbol genes for more than one Refseq identifier. Then, we simulated the construction of two miRNA-gene interaction networks for the two predicted and validated interaction files (1) and (2), respectively, which were loaded and viewed in Cytoscape, version 3.8.0 (https://cytoscape.org/). The two networks were merged according to the tutorial (http://manual.cytoscape.org/en/stable/ Merge.html) with an intersection operator to obtain a single miRNA-gene interaction network common to the two previous networks, formed exclusively by interaction data predicted and validated. The central region of the network with the densest connections was detected with CytoHubba [38]. We applied the maximal click centrality (MCC) method to identify miRNA-gene interaction clusters.

Functional Enrichment Analysis
The target genes of the miRNA-gene interaction network from the previous analysis were used for functional enrichment with the tool the Database for Annotation, Visualization and Integrated Discovery (DAVID, version v6.8) [39]. These target genes were organized in a list containing only the Refseq identifier of the species (Homo sapiens).
To avoid redundancies in terms, high-stringency classifiers with a similarity threshold and multiple linkage threshold equal to 0.50 were applied. The most significant terms for each gene were obtained from functional annotation clusters GO (Gene Ontology) [40] and from KEGG pathways (Kyoto Encyclopedia of Genes and Genomes) [41] with significant values noted with Log 10 , p-value < 0.05.

Statistical Analysis
All statistics were performed using R (https://www.r-project.org). The averages obtained with the non-parametric tests were calculated using the compare_means function. Principal component analysis (PCA) with the FactoMineR package [42] was used for exploratory investigation of multivariate data from miRNA using the prcomp function. The unsupervised grouping and heatmaps were built with the heatmap function.2. The correlation coefficients of the isomiRs were calculated with the following functions: cor and rcorr. Correlations with p-value > 0.01 were considered not significant. The ANOVA test was used to test the means among the miRNA variants.

Data Preprocessing and Abundance of microRNAs in Platelet Concentrate
The results of sRNA-Seq analysis from PCs of the GSE61856 repository, made with sRNAbench, showed that after a preprocessing of the data, more than 95% of the reads were recovered, which mapped between 81 and 85% of unique regions of the genome, with coverage genomics comprised of 95%, ( Table 1). We observed that the amount of read counts (RC) detected for miRBase hairpins showed a decrease from the fourth to the fifth day, increasing only on the last day. The same pattern was observed for mature miRNAs that presented more than 35% of miRNA expressed on the last day of storage (PC-7) ( Table 1). The abundance of miRNAs in stored PC bags described in Table 1, shows an increase in the number of reads until the third day of storage (PC-3), followed by a decrease in PCs on the fourth (p = 0.023) and the fifth (p = 8.9E-08) days, with an increase in the median expression, exceeding the average of normalized expression of 4.8 CPM ( Figure 1A). As a direct consequence of two more days of storage, the PC on the seventh day (PC-7) presented the largest number of reads and 939 miRNAs, which represented an increase of 2.5% observed after seven days of blood collection, confirmed by the decrease in the median. These results confirm that storage for more than five days in a blood bank causes a decrease in the levels of miRNA expression in the PC.   Boxplot is designed from the 75th to the 25th percentile. The vertical lines above and below the box define the maximum and minimum values and the dots indicate outliers, the horizontal line inside the box represents the median value. Kruskal-Wallis test (p-value <0.001) was applied to compare the means between the six groups (** p < 0.01, **** p < 0.0001); (B) The heatmap shows an expression profile defined by the most abundant miRNA families in all PCs. Z-score was the metric applied to infer the best clustering between miRNA families. Gradients with a red tendency represent families of miRNAs with a lower Z-score and gradients with a blue tendency with a higher Z-score; (C-H) Donut chart shows the ranking of miRNA family positions on all PCs. PC, platelet concentrate.          Items in the main columns of the table: gene family, coordinate string, pre-microRNA. UR, number of unique reads; RC, read count; RPM (lib), the read per million normalized by the total number of reads mapped to the library to a known microRNA; RPM (total), the reads per million normalized by the total number of genome mapped reads (genome mode) or the total number of reads in the analysis (sequence library mode); mature microRNAs sequence, miR 5p-/3p-arms, arm sequence as defined by the miRBase annotation. The order of classification of miRNAs is based on the RC.
We provided a panoramic view of the distribution of these miRNA families in all PCs ( Figure 1C-H), where we could see the predominance of the mir-486, let-7, mir-25, and mir-191 families, except for on PC-3 which presented the most abundant mir-423. From this information, we also identified the top-20 miRNAs most expressed. Table 2 shows the detailed information on the precursor (pre-microRNA) and the mature miRNAs identified in this study from the annotation files. It shows a decrease in the level of expression of these miRNAs until PC-5 and an increase after two days on PC-7. In Supplementary Materials File S1, we present detailed information on several canonical miRNAs, with specimens that had their functions elucidated in platelets, in addition to a wide variety of miRNAs that have not been described in the literature.
In all PCs, high levels of expression of miR-486-5p and miR-191-5p were identified, let-7i-5p being the third most expressed miRNA until PC-3, followed by miR-92a-3p, miR-181a-5p, and the other miRNAs that also occupied prominent positions listed in Table 2. We show the ranking of miRNAs with the normalized expression on the Log 2 CPM scale in two separate groups in the graphs depicted in Figure 2A,B.
These two groups of miRNAs had their levels of expression analyzed by the analytical method which compares the relative expression of miRNAs identified by Pontes et al. In a PC bag we can test the following: i.
Our results show that miR-486-5p, miR-92a-3p, miR-103a-3p, miR-151a-3p, miR-181a-5p, and miR-221-3p decrease from the fourth to the fifth day of PC storage ( Figure 3). We compared the average expression of these miRNAs in 100 units of PC on the fourth day (PC-4), confirming that the storage time caused the decrease of these miRNAs, in a much more accentuated way for miR-486-5p more expressed in the sRNA-Seq data ( Table 2). All six miRNAs increased their levels of expression on PC-7 ( Figure 3).
To understand how the increased of storage time caused changes in miRNAs profiles, we applied computational biology simulation to sRNA-Seq and qPCR data. The results showed miRNAs grouped with similar expression profiles, both for increasing and decreasing expression ( Figure 4A,C), and these miRNAs, most likely, could or could not have the same function in PCs with activated platelets. The PCA analysis ( Figure 4B,D) was able to identified MiRNAs groups that suffered expression variation caused by the increased storage time, as we observed miR-486-5p, miR-92a-3p, and miR-151a-3p at extreme points of the ellipses and miR-103a-3p, miR-181a-5p, and miR-221-3p (overlapping ellipses). The first principal components (PC1) explained 64.1% of the total miRNA expression variations in the qPCR experiment ( Figure 4D).
The hierarchical grouping ( Figure 4A) shows an expression pattern of 20 miRNAs in six PCs identified with sRNA-Seq. Z-score was the metric applied to infer miRNAs with similar levels of expression. Gradients with a red tendency represent miRNAs with a lower Z-score and gradients with a blue tendency with a higher Z-score. The PCA analysis graph ( Figure 4B) shows the grouping of miRNAs with expression level <80% (in red, according to the miR-127-3p expression reference) and miRNAs with expression level 80% on the PC (in blue, according to the miR-320a-3p expression reference). The hierarchical cluster ( Figure 4C) and PCA ( Figure 4D) identify the associations of miRNAs, miR-486-5p, miR-92a-3p, miR-103a-3p, miR-151a-3p, miR-181a-5p, and miR-221-3p validated with qPCR related to storage lesions. In the PCA analysis graph, ellipses were predicted with a probability of 0.95. The X-and Y-axes show principal component 1 and principal component 2. The first principal component (PC1), explained 94.5% and 64.1% of the total miRNA expression variations in the two experiments, respectively. In both PCAs, the divergences in the first two main components reflect the differences in miRNA profiles with a particularly distinct division between groups.

IsomiR Quantification
We identified a dominant pattern of expression of mature miRNAs in 5p-arm and 3p-arm, which were investigated systematically. Our results show that non-activated platelets (PC-1), present more than 88.73% of specific miRNAs in 5p-arm and only 11.27% in 3p-arm ( Figure 5A). We extended this count across all PCs in an attempt to find significant differences in 5p-arm and 3p-arm expression dominance.
The data indicate that the percentage decrease in miRNA, caused by storage, was 85.75% in 5p-arm and 14.25% in 3p-arm (PC-5) as compared with previous PCs ( Figure 5A). We measured the density of miRNA expression by the log 2 ratio (5p-arm/3p-arm). A significant difference was found in the level of expression, tens, or hundreds of times greater in the 5p-arm than in the 3p-arm ( Figure 5B). A large number of miRNA different sequences variants were identified in all PCs in a reproducible way (Table 3). In Table 3(1) we identified the total of isomiR expressed in count reads in the form of NTA and length variants. In Table 3(2) we account for the mean and standard deviation of these variants in count reads, detected in the canonical sequence of the 20 miRNAs. The averages were tested with ANOVA (p < 0.001) and were significant. On the one hand, the amount of the NTA(U) variant increased from PC-1 to PC-7 to 67.31%, on the other hand, the NTA(A) variant decreased from PC-1 to PC-7 to 30.42%. The variants NTA(C) and NTA(G) showed less relevant variations in increase and decrease.  (C) Correlogram calculated to highlight the miRNAs that are most correlated with the NTA variants, with an emphasis on miR-486-5p, miR-92a-3p, miR-320a-3p, miR-127-3p, let-7i -5p, let-7b-5p, miR-103a-3p, miR-151a-3p, miR-221-3p, miR-26a-5p, miR-423-3p, and miR-423-5p; (D) Correlogram calculated correlation to highlight the miRNAs that most correlate with the length variants, highlighting miR-486-5p, miR-127-3p, let-7i-5p, miR-103a-3p, miR-423-3p, miR-181a-5p, miR-22-3p, let-7d-5p, miR-423-5p, and let-7g-5p. MiRNAs that have a positive correlation for both variants at the same time were miR-486-5p, miR-127-3p, let-7i-5p, miR-103a-3p, miR-423-3p, and miR-423-5p. In the correlograms, the scales with blue gradient are positively correlated, while the gradients in red color are negatively correlated. Blank scale are miRNAs correlations with p-value >0.01 are considered no significant.
The most abundant length variants were lv3pT which decreased from PC-1 to PC-7 to 63.64%, whereas lv5pE increased from PC-1 to PC-7 to 30.11%. The lv5pE, lv5pT, and mv variants were less abundant. Sequence variants were correlated to miRNAs by calculating the correlation matrix ( Figure 5C,D), which highlighted the following miRNAs: miR-486-5p, miR-127-3p, let-7i-5p, miR-103a-3p, miR-423-3p, and miR-423-5p (Table 3). These showed a positive correlation for both variants at the same time. We gathered more detailed information on these variants which is available in Supplementary Materials File S2.

Functional microRNA-gene Interaction on the PC
The results of the miRNAs target prediction genes showed that the subsets of predicted and validated interactions share 275 (23.3%) of accounted interactions ( Figure 6A). The first network of miRNA-gene interaction constructed with file (1), presented 339 nodes and 1070 edges, while the second network constructed with file (2) presented 127 nodes and 560 edges. The final network merged from the first two networks presents 108 nodes and 220 edges ( Figure 6B). The network topology presents a central region formed by denser connections with miRNA-gene interactions, ordered by the MCC mainly for let-7d-5p, let-7a-5p, let-7i-5p, let-7b-5p, let-7f-5p, and let-7g-5p, in addition to the YOD1 gene and compositions formed by miR-92a-3p, miR-423-5p and miR-103a-3p. Most of the genes in the network in Figure 6B were enriched in GO, mainly in the categories of molecular function for protein binding (GO: 0005515), cellular component for nucleoplasm (GO: 0005654), and in a biological process for negative regulation of transcription, DNA-templated (GO: 0045892) ( Figure 6C). These genes were also enriched with DAVID [39], generating a wide pathway panel with an important functional repertoire for the P53 signaling pathway, described in signs of oxidative stress, including DNA damage, activation of oncogenes, cell cycle arrest, senescence, and apoptosis. Pathways with an impact on cancer, cell cycle and platelet activation, and other pathways directly associated with cellular stimuli, signal transduction, cell signaling, and stress response were predicted ( Figure 6D). The results of these analyses are in Supplementary Material file S3.

Discussion
Aging is characterized by a functional decline in many physiological systems that can be triggered by environmental and endogenous stress, including the wear on telomeres, genomic instability, epigenetic changes, and loss of proteostasis, favoring cell damage and the progression of physiological aging [43].
Platelets are small anucleated cells that essentially originate from the fragmentation of pseudopods from the megakaryocyte cytoplasmic membrane in the bone marrow [44]. In a healthy person, platelets circulate in the blood for about seven to ten days, and then are removed from circulation and destroyed in the spleen [45,46]. Similarly, this aging also occurs in vitro, wherein blood banks in most countries, blood components widely used and routinely supplied in the form of PCs are discarded after five days of storage [47].
Platelet storage in blood banks causes a decrease in the abundance of miRNAs due to shear stress and platelet activation, being the two main factors responsible for the release of microparticles (MPs) rich in miRNAs [11][12][13]. Studies of this nature have shown a relationship between miRNA profiles with subsequent platelet reactivity, suggesting an important role in post-transcriptional regulation during storage [7,24,48].
In this study, we identified the 20 most expressed miRNAs in PCs stored for seven days with a genomic coverage of 95%. Specifically, in the PC with high-quality platelets (PC-1), we identified a total of 916 miRNAs (Table 1). However, we confirmed a 22.4% decrease in miRNA levels from the first to the fifth day, much more accentuated than the value found in our previous study [20], increasing the number of miRNAs only after seven days of blood collection with a rate of 2.5, which is, most likely, a response to inhibit the translation of proteins induced by stress caused by aging for more than seven days of storage [20]. On the basis of these results, we emphasize that the use of different bioinformatics pipelines to analyze the same sequencing data, generates results that can differ substantially. Whereas, some studies have highlighted the urgent need to ensure that the bioinformatics pipelines used for next-generation sequencing (NGS) analysis, undergo better validation, especially for applications in translational genomic medicine [49].
In our data, we observed the influence of storage on the unequal distribution of the abundance of miRNA families in all PCs (Figure 1). For example, the largest mir-486 family showed declines in expression, losing the most abundant position on PC-3 to the mir-423 family. The mir-191 family was replaced by members of the let-7 family on PC-2, which are very common in platelets [50]. Probably, post-transcriptional changes influenced the biogenesis and stability of miRNA during storage, as has been shown in studies that used molecular changes in DICER1 to reduce the number of miRNAs that strongly regulated platelet reactivity [51,52].
The measurement of miRNAs expression in the PCs is a variable that we have demonstrated to be associated with the quality of these blood components. When we evaluated the expression levels of miRNA in PCs using computational methodologies, we concluded that, in clinical practice, miRNAs are a very useful tool for testing PC bags that are close to expiration date during storage in a blood bank.
In this study, we found a large number of miRNAs that are candidates as storage damage biomarkers that can replace miR-127, miR-191, and miR-320a miRNAs in clinical trials, which were found in our first study, especially miR-191 which has been used as an internal control for qPCR validation analysis [20].
In addition, we selected six new miRNAs, of the most expressed RNAs, for further validation by qPCR. Relative quantification indicates decreased expression levels of miR-486-5p, miR-92a-3p, miR-103a-3p, miR-151a-3p, miR-181a-5p, and miR-221-3p from the fourth to the fifth day ( Figure 3). We also highlighted that this decrease occurred more accentuated for miR-486-5p, which was more expressed in the sRNA-Seq data (Table 2). Additionally, all six miRNAs increased their levels of expression on PC-7 ( Figure 3).
We applied computational biology simulations (hierarchical grouping and PCA analysis) to the data generated by sRNA-Seq and qPCR which revealed how the increase in platelet storage time caused changes in the miRNA profiles confirmed in the validation. These computational methodologies have resulted in more accurately identifying miRNAs located in different groups based on the days of storage [53]. In our previous study, we used these methodologies which identified changes in the expression profiles of 14 miRNAs that were associated with PSL [21]. In this current study, we confirmed that changes in the profiles of the new miRNAs correlated with the instability of the half-life of these transcripts on the fourth day, which coincided with the time of onset of PSL. For example, PC bags that are tested and confirm that the expression of miRNAs (miR-151a-3p, miR-103a-3p, and miR-221-3p) is <80% of the expression of (miR-486-5p, miR-92a-3p, and miR-181a-5p) means that there are storage lesions.
The miRNA's stability varies widely, with half-lives of~1.5 h, more than 13 h, and up to 48 h, in human biofluids [54,55]. Measuring the relative levels of miRNA in PC is subject to some challenges that need to be taken into account, because the relative stability of miRNAs has implications for their ability to transfer regulatory information, as they are very short and have highly divergent sequences, with a wide variation of the GC content that can favor the different hybridization properties among different miRNA sequences [55].
In a blood bank, the analytical method for testing PCs can be implemented quickly and with low cost to test PC bags stored for more than four days whihc still contain physiologically normal platelets. The durability of platelet physiology depends on individuals with ideal suboptimal health status (SHS), which is considered to be a subclinical and reversible stage of chronic disease. Individuals with SHS can have a progressive accumulation of senescent cells and a relative shortening of the telomeres that produces the early biological aging of platelets [56][57][58].
The increase of miRNA levels in the PCs suggests that after platelet activation they stabilize within the circulating MPs for their transport of action [7,13,59], undergoing changes in their profiles and using selective platelet packaging pathway for MPs [13,60]. For example, we justify expression changes on several PCs, based on the dominant expression pattern in 5p-arm about 3p-arm ( Figure 5B). The density measurement confirmed a significant difference in the level of expression, tens or hundreds of times greater in the 5p-arm than in the 3p-arm, as has been observed in other studies [61][62][63].
Our global estimates of miRNAs expression on all PCs pointed to a decline in this dominant pattern from PC-2 to PC-5, but only increased on PC-7 (Table 3). We found several non-model and length variants, positively correlated with recently emerged miRNAs (miR-486-5p, miR-92a-3p, miR-103a-3p, miR-151a-3p, miR-181a-5p, and miR-221-3p), which have not yet been reported in other platelet miRNomes (complete sequencing of miRNAs) and provide greater quality and innovation to the analytical test mentioned in that study.
Our analysis of miRNA-gene functional interaction, pointed to the existence of a molecular regulation mechanism in platelets, which was inferred based on the topology of the interaction network formed by recently emerged miRNAs (miR-103a-3p, miR-423-5p, and miR-92a-3p) and conserved miRNAs of the let-7 family interacting with the YOD1 gene, a desubiquitinating enzyme that is very expressed in platelet hyperactivity [56]. We also identified the functional roles of significant target genes in signaling pathways, cell cycle, stress response, platelet activation, and cancer.
In the future, our investigations should be repeated in a larger number of samples, studying the sixth day of storage (not carried out with the current data). Fresh platelets (PC-0) should also be studied to obtain a broader profile of expression variation with storage days. The option to extend the study after seven days of storage would also be an advantage.
In summary, our results have a promising application in transfusion medicine, because we describe a new collection of miRNAs (miR-486-5p, miR-92a-3p, miR-103a-3p, miR-151a-3p, miR-181a-5p, and miR-221-3p) that shows a sensitivity expression pattern due to biological platelet changes during storage. These miRNAs could be applied, in blood banks, as potential biomarkers to also measure the quality and viability of the PC during storage.