Shedding Light on tRNA-Derived Fragment Repertoire in Facs Blood Cells

1. Introduction

tRNA-derived fragments (tRFs) are a relatively recently discovered class of small RNA molecules that are shown to be involved in the pathogenesis of various diseases, particularly cancer and viral infections. Processed from transfer RNAs, tRFs regulate gene expression through various mechanisms like miRNA-like silencing of target mRNAs [1,2,3,4].

There are several types of classified tRFs based on the cleavage site within the tRNA molecule: tRF-1: Cleaved from the 3' trailer sequence of pre-tRNAs. tRF-2: Cleaved from the D-loop of mature tRNAs. tRF-3: Cleaved from the 3' end of mature tRNAs. tRF-5: Cleaved from the 5' end of mature tRNAs. i-tRF: Internal tRFs derived from the internal region of mature tRNAs [5].

Different tissues and organs possess a different set of tRNA fragments [2]. Nevertheless, it has not been studied how the pattern of these molecules differs in different blood cells. Supposably, the repertoire of these molecules changes in correspondence with changes in the immunological state of patients. In particular, it is not clear how these molecules will change in blood cells during the cytokine storm induced by COVID-19.

It has been previously reported by Wu et. al. that tRFs were the most significantly affected small non-coding RNAs in nasopharyngeal swabs of COVID-19 patients, they had also observed that SARS-CoV-2-infected airway epithelial cells exhibit the same tendency [4]. In an attempt to additionally shed light on tRNA-derived fragments and their involvement in host-virus interactions it seems relevant to study tRFs in different pathophysiological circumstances. In the case of this study our aim was to reveal the differential expression of tRFs in fluorescence activated sorted cells of healthy control donors and SARS-CoV-2 infected patients, as well as to observe and characterise tRFs during viral infections.

2. Materials and Methods

2.1. Patients and Data Collection

Six healthy donors and five RT-PCR confirmed SARS-CoV-2 positive patients made up the two groups of research participants. Three patients were hospitalised to the critical care unit and two patients were moved to the infectious disease unit. Severe patients satisfied the following criteria for admission to the intensive care unit (ICU): body temperature ≥ 39 °C, respiration rate ≥ 30/min, and oxygen saturation (SpO2) ≤ 93%.

2.2. Cell Sorting

Сell sorting was performed on a MoFlow Astrios EQ device (Beckman Coulter). Erythrocytes were isolated using antibodies against CD235 protein (Beckman Coulter , IM2212U). To isolate leukocytes, erythrocytes were first lysed with VersaLyse lysing solution (Beckman Coulter). Granulocytes (neutrophils, basophils and eosinophils) were isolated using duraclone IM granulocyte antibody panel (B88651, Beckman Coulter). Lymphocytes and monocytes were isolated using antibodies CD45 (A79392 Beckman Coulter), CD16 (6607118 Beckman Coulter), CD14 (A70204 Beckman Coulter, USA). Cell sample purity for all cell populations was >95% according to flow cytometry data. A detailed description of cell sorting protocols is provided in the following article [6].

2.3. RNA Separation, Library Preparation and Next Generation Sequencing

RNA was isolated\extracted using ExtractRNA reagent (Evrogen) according to the provided protocol. After extraction, RNA was dissolved in 10 µl RNAse free water. The quality of the obtained RNA was checked on a TapeStation (Agilent). Only samples with RIN>5 were taken for sequencing. Short RNA libraries were prepared using the Small RNA Library Prep Kit (BGI, 1000006383). Sequencing was performed on a DNBSEQ-G400 instrument (BGI)

2.4. Bioinformatics

A general design of the bioinformatic processing pipeline is presented in Figure 1.

Read Quality Control

For quality control we applied FastQC [7]. All datasets qualified quality control standards. Each FastQ file contained between 17 and 27 million reads.

Adapter Trimming

Adapter trimming was executed using Trimmomatic [8]. Long clipping sequence was set as AGTCGGAGGCCAAGCGGTCTTAGGAAGACAA. We opted to use optimal run settings for SE (single-end) reads (TruSeq3-SE:2:30:10 LEADING:3 TRAILING:3 SLIDINGWINDOW:4:15 MINLEN:20) which allowed us to minimise execution time with no data loss. As an input, trimmomatic takes the adapter.fa file, in which our adapter sequence is registered. For each FastQ file from 50 to 68% of the input reads were marked as dropped after running Trimmomatic.

Read Mapping, Count and Normalisation

Full genome alignment and transriptome-based pseudoalignment methods are not applicable for small-RNA-seq identification due to tRNA cleavage events, which led us to utilising "BLAST"(Blast Local Alignment Search Tool) [9].

We implemented BLAST through “tRNAExplorer" [10], a Python-based pipeline, optimised for tRF-profile analysis. In order to utilise tRNAExplorer, we needed a *.bed genome annotation file as a reference, which we would BLAST our reads *.fa files against. While there is an option to generate a custom annotation file, we used an already compiled grch.38 database which came within the tRNAExplorer package. To run the pipeline, we created project directories with sample lists individually for each type of cells and custom configuration file, with data directories paths and launch options. While tRNAExplorer supports trimming and QC, we omitted these steps due to previous data processing. As an input each run of the pipeline takes all the *.fa files with read data from the specified directory, list of samples, and path to database *.bed file. We automated these runs with another custom Python script. As a final result of data processing we obtained *.csv and *.tsv text files, containing BLAST run results for every sample, cleavage sites data and a collective read count table for all provided samples with annotations.

This data was already sufficient for differential expression analysis and visualisation, but we applied another extra step, running a custom Python script which organises data and performs statistical tests.

2.5. Statistics

Wilcoxon–Mann–Whitney test was used to perform statistical analysis on the data. P-values less than 0.05 were considered statistically significant. The p-values for the RNAseq volcano graphics were log converted to log10 (1.31) for data visualisation.

3. Results

3.1. tRF Length Distribution in Main Cell Types

Analysing tRF length distribution in erythrocytes, monocytes, lymphocytes, neutrophils, basophils and eosinophils we observed a remarkable similarity between control and severe COVID-19 patient’s cells (Figure 2). This data is additionally supported by our heatmap, which demonstrated a distinguishable pattern of tRF expression depending on cell type. Erythrocytes are presented as the most distinguishable group of cells in terms of tRF length and tRF expression patterns, which may be due to the fact that the erythrocyte is an inactive cell in terms of synthetic processes.

3.2. tRF Type Composition in Main Cell Types

Analysing the main tRF types in sorted lymphocytes of control donors and severe COVID-19 patients we observed the prevalence of tRF-5 in 7 out of 8 samples. 1 severe COVID-19 patient exhibited dominating above all presence of i-tRF and tRF-3 types (Figure 3). It is worth noting that this patient had extremely high levels of IL-6 - 2 398 (pg/ml) on the same day that this sample was taken.

3.3. tRF Expression in Control Donors and Severe COVID-19 Patients

To compare tRF expression in different cell types we calculated the 150 most abundant tRFs based on RAW read counts, log normalised for data presentation. The heatmap demonstrates a relatively vivid frontier for erythrocytes (Figure 4).

3.4. tRF Differential Expression in Erythrocytes

Analysing the expression of tRFs in sorted erythrocytes we observed 12 upregulated and 21 downregulated fragments (Figure 5, Table 1). tRF#33-QJ3KYUYRR6RBD2 stood out with significant downregulation, however, there is no data on the involvement of this fragment in any known biological processes.

3.5. tRF Differential Expression in Lymphocytes

tRFs in lymphocytes demonstrated significant upregulation with 49 upregulated fragments, and zero downregulated, which vividly sets lymphocytes apart from other cells (Table 2). We applied log4 transformation for data presentation (Figure 6).

3.6. tRF Differential Expression in Monocytes

Monocytes were presented with the largest overall amount of significant hits. 38 upregulated tRFs and 53 downregulated (Figure 7, Table 3).

4. Discussion

This pilot study aimed at demonstrating a distinguishable tRF pattern for different sorted cell types. Given the fact that tRFs have been discovered relatively recently, research aimed at understanding their origin can help elucidate their biological purposes.

Analysing tRF length distribution in erythrocytes, monocytes, lymphocytes, neutrophils, basophils and eosinophils we observed a remarkable similarity between control and severe COVID-19 patient’s cells (Figure 2). This data is additionally supported by our heatmap (Figure 4), which demonstrated a distinguishable pattern of tRF expression depending on cell type. Erythrocytes are presented as the most distinguishable group of cells in terms of tRF length and tRF expression patterns, which may be due to the fact that the erythrocyte is an inactive cell in terms of synthetic processes.

Regarding tRF type composition, It has been previously noted that 3-tRFs are produced in response to various cellular stresses like oxidative stress, hypoxia, and viral infection [11]. Some 3-tRFs can inhibit viral replication by interfering with viral gene expression or packaging. They can act as signalling molecules to mediate stress responses [12,13].

Given the fact that IL-6 is a key mediator of the "cytokine storm" that leads to acute respiratory distress syndrome (ARDS) and multi-organ dysfunction in severe COVID-19 [14], it is assumable that such distribution of tRF types may be due the condition the patient underwent. Other cell types did not exhibit any significant alterations in tRF composition.

Overlooking tRF expression in control donors and severe COVID-19 patients, it is worth noting how the expression is increased in the majority of tRF in granulocytes (neutrophils, basophils, eosinophils). A similar tendency to increased expression is observed in monocytes. Granulocytes and monocytes both originate from CFU-GM (Colony Forming Unit–Granulocyte–Macrophage), also known as granulocyte–macrophage progenitor (GMP) [15].

The aforementioned suggests that tRFs may dominantly persist from the progenitor stages of cell development, and are only slightly modified by environmental or physiological factors.

Most knowledge today regarding tRFs is concentrated around cancer research [2,16,17,18].

In our study we encountered fragments which have been previously associated with several types of cancer.

Lymphocytic tRFs: It is noted that tRF#31-R29P4P9L5HJVE previously has been acknowledged as a marker for lung cancer prediction among smokers 10 years prior to being diagnosed [19].

tRF#19-VKS4I71Z has been mentioned as an abundant novel-trf in a 2016 study of RNA-seq data from human prostate tissue [20].

In a 2023 research dedicated to studying tRFs in cancer, a high level tRF#34-5QKDN6QQ1362HQ has been mentioned as a predictor of breast cancer improved survival [2].

Monocytic tRFs: tRF#22-WEK6S1852 in previous research was found to be significantly downregulated and associated with human malignant mesothelioma [16].

A 2019 colon cancer study revealed significantly differentially expressed tRFs between colon cancer tissues and peritumor tissues, whereas another upregulated tRF that we observed - tRF#22-9LON4VN11 was mentioned, demonstrating downregulation in colon cancer tissues with a log2 fold change of -1.26 [17].

Another fragment we observed - tRF#18-HSQSD2D2, downregulation of which has been previously associated with early-stage breast cancer [2].

In a study devoted to examining the dysregulation of different tRFs in chronic lymphocytic leukemia - tRF-20-RK9P4P9L was amongst the top 15 differentially expressed sRNAs in aggressive chronic lymphocytic leukemia vs. normal controls, with the linear fold change being −76.64 and −258.08, respectively. Samples were composed of CD5+/CD19+ B cells. It is worth noting that in our study, the same tRF was also significantly downregulated but only in CD14++ CD16− monocytes [18].

tRF#22-WEPSJR852 was found in peripheral blood of fibromyalgia patients in a dissertation dedicated to the search of morphological substrate to fibromyalgia [21].

Erythrocytes appeared to be the only cell type which showed differential expression of tRFs that have not previously been associated with any type of cancer or disease. The fact that the erythrocyte is an inactive cell in terms of synthetic processes may elucidate these findings, however, regarding different cell types, tRFs and their interrelations with the conditions mentioned earlier it is more likely to suppose a coincidence, rather than some significant finding.

Differences in the length, type, and composition of specific tRFs between different cells are supposedly due to the specificity of the tRNA cleavage system. The same can be said about changes that occur during cytokine storm. The regulation system of tRNA processing is still poorly understood.

Nonetheless, we can say that it occurs at several “levels”. First of all - the regulation of nucleotide modifications in certain molecules [22]. Another stage is dependent on the regulation of RNAses that cleave tRNAs at specific sites.

The main types of nucleases involved in tRF biogenesis are: Dicer - cleaves 5' ends of mature tRNAs. Angiogenin (ANG) - cleaves mature tRNAs at anticodon loops. RNase Z/ELAC2 - cleaves 3' trailer sequences of pre-tRNAs [11,23,24].

RNase P: excision of external transcribed spacer (ETS) and internal transcribed spacers (ITS) from pre-tRNA transcripts [25].

Regulation of these enzymes can occur at the level of transcription initiation, as well as post-translational modifications of the enzyme

It is also likely that tRNA cleavage is regulated by proteins that bind this molecule [26] and make certain sites inaccessible for the aforementioned process. Unfortunately, we cannot state by what mechanism the difference between tRFs in blood cells is regulated.

In summary, it should be noted that TRF profiles significantly differ in different types of blood cells and demonstrate dramatic differential expression (sometimes more than 500-fold) during cytokine storm. Such profound differences suggest a major role of tRNA-derived fragments in the functioning of blood cells.

5. Conclusions

The composition of tRFs in erythrocytes monocytes, lymphocytes, neutrophils, basophils and eosinophils was analysed on the basis of sRNA-SEQ. We demonstrate notable alterations in the length and types of these molecules in main blood cell populations. We additionally observed a significant change in the profile of tRFs in the erythrocytes, monocytes and lymphocytes of patients infected with SARS-CoV-2.