A simplified qPCR method revealing tRNAome remodeling upon infection by genotype 3 hepatitis E virus

The landscape of tRNA–viral codons regulates viral adaption at the translational level, presumably through adapting to host codon usage or modulating the host tRNAome. We found that the major zoonotic genotype of hepatitis E virus (HEV) has not adapted to host codon usage, prompting exploration of the effects of HEV infection on the host tRNAome. However, tRNAome quantification is largely impeded by the extremely short sequences of tRNAs and redundancy of tRNA genes. Here, we present a length‐extension and stepwise simplified qPCR method that utilizes a universal DNA/RNA hybrid tRNA adaptor and degenerate primers. Using this novel methodology, we observe that HEV infection dramatically reprograms the hepatic tRNAome, which is likely to facilitate translation of viral RNAs. This tRNAome quantification method bears broad implications for future tRNA research and possibly tRNA‐based diagnostics.

The landscape of tRNA-viral codons regulates viral adaption at the translational level, presumably through adapting to host codon usage or modulating the host tRNAome. We found that the major zoonotic genotype of hepatitis E virus (HEV) has not adapted to host codon usage, prompting exploration of the effects of HEV infection on the host tRNAome. However, tRNAome quantification is largely impeded by the extremely short sequences of tRNAs and redundancy of tRNA genes. Here, we present a length-extension and stepwise simplified qPCR method that utilizes a universal DNA/RNA hybrid tRNA adaptor and degenerate primers. Using this novel methodology, we observe that HEV infection dramatically reprograms the hepatic tRNAome, which is likely to facilitate translation of viral RNAs. This tRNAome quantification method bears broad implications for future tRNA research and possibly tRNA-based diagnostics.
Keywords: hepatitis E virus; length-extension and simplified qPCR method; tRNAome Viral adaption is shaped by a variety of effectors, including the need to hijack the cellular protein-synthesizing machinery for efficient translation of viral genes [1]. Rate-limiting steps in the production of viral proteins differ depending on host species and the virus involved. But in many cell types, the efficiency of the translational machinery is critically dependent on tRNA availability for decoding both cellular and viral codons [2,3]. Accordingly, many viruses show adaptation with respect to codon usage toward their host [4][5][6][7].
Conversely, it has been reported that interferons, the potent antiviral cytokines [8], can alter the availability of tRNAs to facilitate the production of antiviral proteins and to attenuate decoding of viral codons [9,10]. Thus, studying the interplay between viral codons and cellular tRNAome is essential for understanding the infection biology.
tRNAs are the recognition modules that decode the mRNA in the ribosome and are covalently charged by amino acids [11]. Following recognition, the corresponding intramolecular bond is hydrolyzed and the released amino acid is covalently linked to the nascent peptide chain. The tRNAs encoded in genomic DNA are typically 70-90 base pairs in length often with multiple genomic copies of the same tRNA [12]. The human tRNAome is encoded by more than 600 tRNA gene loci [12]. Following transcription, tRNAs are subjected to post-transcriptional modification, and a common CCA ribonucleotide sequence is added to the 3 0 end of tRNA [13]. Accurate detection of the mature tRNAome is technically challenging and especially hampered by the short-length and redundant tRNA genes [14,15]. As the number of different tRNA sequences is relatively limited, we believe that optimized qPCR-based techniques hold possibility for quantifying tRNAs.
Hepatitis E virus (HEV) is a ssRNA(+) virus and is the leading cause of acute viral hepatitis [16]. Globally, HEV causes annually 20 million infections with over 56 000 lethal cases and especially in pregnant women [17]. Among the major genotypes, genotype 1 and 2 HEVs only infect humans, whereas genotypes 3 and 4 are zoonotic. Genotype 3 is highly prevalent in Western countries with a broad host spectrum [18,19]. In this study, we aim to first develop a simplified qPCR method for quantifying mature tRNAs, and then to investigate how codon usage of the zoonotic HEV genotype relates to the human tRNAome composition.

Experimental models
For HEV infectious model, a full-length genotype 3 HEV genome (Kernow-C1 p6 clone; GenBank accession number: JQ679013) was used [20]. The human hepatoma Huh7 cell line was used to harbor the HEV genome and produce infectious virus particles for secondary infection of na€ ıve Huh7 cells [21]. The HEV subgenomic model was based on Huh7 cells containing the subgenomic genotype 3 HEV sequence (Kernow-C1 p6/luc) coupled to a Gaussia luciferase reporter gene [21,22]. The cell line was regularly checked for identity using STR verification as provided by the pathology department of the Erasmus MC. Key regents, viral strains, and software used in this study are listed in Table S1.

The qPCR protocol for quantifying mature tRNAs
Preparation of samples and reagents -Timing 1 day 1 Prepare primary cells, cell lines, or tissues to yield 1-4 µg total RNA for each sample. This amounts to roughly 10 5 -10 6 mammalian cells. These values are rough guidelines for experimental design, but yields will vary according to the cell types and amounts.
Notice: Because of the large number of tRNA sets, it is recommended to prepare tRNA-X-R and tRNA-X-F primer mixture to minimize loading variation. For example, to prepare 80 times of tRNA-R/F mix, we recommend that 20 µL (10 pm) of tRNA-X-R and tRNA-X-F was diluted by 200 µL ddH 2 O. Therefore, the finally used volume of tRNA primers mix is 3 µL. 17 Perform qPCR with the designed cycling condition according to Table S4.

Quantification
The 2 ÀDDCt was used to calculate the relative expression of each tRNA species, and the DC t values were determined by subtracting the average C t values of the endogenous control gene GAPDH from the average C t values of each tRNA type.

Statistical analysis
Linear correlations between the relative synonymous codon usage (RSCU) of human and animal HEV strains were estimated by the Pearson coefficients. Statistical comparisons of tRNAome data were performed with the Mann-Whitney U-test for nonpaired independent samples. *P ≤ 0.05 was indicated as significant. The statistical analysis was performed using the SPSS 19.0 software (IBM Corporation, Armonk, NY, USA).

Results
The codon usage of genotype 3 HEV has not adapted to its human host As in many biological systems, tRNA availability [23] and codon context [24] constitute the rate-limiting step for protein synthesis. Viruses are expected to adapt codon usage to fit host codon usage patterns. With respect to the zoonotic genotype 3 HEV, however, this has not yet been analyzed. Hence, we compared the RSCU of three open reading frames (ORFs) of genotype 3 HEV isolated from human to the counterpart of the human host. Similar analysis was also performed for HEV strains isolated from animals, including swine, deer, mongoose, and wild boar (Table 1). Strikingly, we did not find evidence that the codon usage of genotype 3 HEV isolated from human and the animal host has adapted to the human codon usage pattern for any of the three ORFs investigated (Fig. 1), because linear correlations between the RSCU of human and animal HEV strains are very poor (Pearson coefficients less than 0.40). Moreover, the RSCU analysis of HEV between the nonhuman primate (e.g., monkey) and humans also indicated that the usage of major codons is similar except serine codon (UCU) (Fig. S1). Besides RSCU, the CAI has also been suggested to predict the efficiency of translation elongation [25]. Our results indicated that the CAI of HEV was similar among different host species-derived strains, and much lower than that of human housekeeping gene (e.g., GAPDH) (Fig. 1). We conclude that there has been very little evolutionary pressure on genotype 3 HEV to adapt to the codon usage of its human host. Thus, we decided to investigate how HEV infection regulates the host tRNAome.
A simplified qPCR method for characterizing the human mature tRNAome To develop a novel qPCR method for easy quantification of mature tRNAs, we first designed a DNA/ RNA hybrid adaptor (universal adaptor, U-adaptor).
It is compatible with all types of mature tRNAs with a common 3 0 CCA acceptor (Fig. 2). Following binding of the adaptor to the tRNA molecule, the adaptor can be ligated to both the 5 0 end and 3 0 end of the mature tRNAs. Of note, the mature tRNAs are frequently charged by amino acids that hamper the above ligation reaction. Thus, a deaminoacylation reaction for the charged tRNA is needed during RNA isolation (see Materials and methods). Importantly, the adaptor is designed to form a loop and hybridize on tRNA itself. This hybridization is compatible with any type of base pairs adjacent to 3 0 CCA tail of tRNAs, as the pairing base of the U-adaptor involved is degenerately designed. The ligation reaction specifically links the 5 0 end and 3 0 end of the adaptor to the corresponding ends of tRNA molecule that is catalyzed by T4 RNA ligase (Fig. 2). Finally, a hybrid pseudocircular molecule, containing the deaminoacylated original mature tRNA molecule and the Uadaptor, is formed. This strategy serves two important ends: It extends the length of tRNA for subsequent qPCR while simultaneously eliminating unmature tRNAs. Afterward, tRNAs can be determined by conventional qPCR (see the experimental strategy and Fig. 3).
To this end, a mixture of tRNA-specific reverse primers was used, with the primers designed to bind to the region between the anticodon loop and the D loop. A complication with regard to designing reverse primers for characterizing the tRNAome is that many iso-decoder tRNAs (i.e., tRNA with the same anticodon but vary in backbone) are responsible for decoding of the identical genetic codon [12]. We thus retrieved all 57-human genomic tRNA genes from the GtRNA 2.0 database and accordingly designed degenerate primers (Table S2). Using RNA obtained from Huh7 human liver cells, all primer pairs successfully amplified their iso-decoder tRNA molecules in qPCR (Figs S2 and S3). We conclude that our approach employing the U-adaptor and degenerate primers allows easy quantification of cellular tRNA profiles.

Normalization of tRNAome by reference gene
In order to normalize tRNA level, tRNA-specific reverse primers, and random and oligo (dT) primers were concurrently used for reverse transcription reaction. GAPDH, a commonly accepted reference gene, is selected for tRNAome normalization [26]. tRNA-His-GUG is the only tRNA species capable of decoding histidine codons (CAC and CAU), and thus, we expect that its expression would be more stable. Indeed, we observed good correlation between levels of GAPDH mRNA and tRNA-His-GUG (Fig. S4). Comparing the U-adaptor and tRNA-histidine adaptor for ligation reaction, we found that the U-adaptor did not compromise the ligation efficiency as it caused only 0.11 variation calculated by DC t (Table S1). Efforts were made for further optimization of the protocol, for instance, by combining steps including annealing, but these efforts failed (Fig. S5). The final protocol (refer to Materials and methods) is organized in seventeen discrete steps and allows quantitative assessment of mature tRNAome.

HEV infection provokes remodeling of the mature tRNAome in host cells
Although viruses are considered to adapt their codon usage to that of their host [3,27], we observe that HEV is not subject to such adaptation. Hence, we decided to compare tRNAome composition in the presence and absence of HEV infection. To this end, Huh7 cells were infected with full-length HEV or exposed to vehicle control (Fig. S4A), and the levels of different tRNAs were determined (Fig. S6A). We observed a major effect on tRNAome remodeling by HEV infection. A total of 29 of 57 tRNA species showed significant upregulation. The most marked example is tRNA-Pro-GGG that shows 162 times higher expression in HEV-infected cells as compared to noninfected cells. tRNA-Gly-CCC and tRNA-Pro-UGG were approximately 50 times and about 35 times upregulated in HEV-infected cells, respectively. Conversely, Fig. 1. Lack of adaptation of genotype 3 HEV codon usage to host codon usage. Codon usage of human HEV was correlated with overall human codon usage. The codon usage of swine, wild deer, mongoose, and wild boar HEV was also related to human codon usage. For each possible codon and 64 in total, its RSCU in the human genome is plotted against the y-axis and the corresponding RSCU in the relevant viral ORF on the x-axis. Results were shown for all three ORFs in the upper panels. In the lower panel, CAI of genotype 3 HEV isolated from different host species was compared for ORF1-3. The CAI of GAPDH from deer, wild boar, and humans is also indicated. The CAI calculation was performed by CODONW software with Saccharomyces cerevisiae as reference set.

HEV infection-triggered tRNAome remodeling is largely dependent on ORF2
The capsid protein of HEV encoded by ORF2 features many ß-sheets in its secondary structure [28] and consequently requires incorporation of a multitude of proline and glycine residues during synthesis. By far, the most prominent effects of HEV infection on the tRNAome are the upregulation of tRNAs supporting proline and glycine decoding (tRNA-Pro-GGG, tRNA-Pro-UGG, and tRNA-Gly-CCC). Thus, we investigated the role of ORF2 in HEV-induced tRNA remodeling. We quantified tRNAome in cells carrying subgenomic HEV replicon with ORF2 deletion. This subgenomic HEV replicon contains a luciferase reporter, and viral replication can be quantified by luciferase activity (Fig. 4A). Intriguingly, although effects on tRNAome composition were still observed in the subgenomic replicon (Fig. S6B), they are much moderate as compared to those in the full-length HEV model (Fig. S6C). Effects on tRNA-Pro-GGG, tRNA-Pro-UGG, and tRNA-Gly-CCC levels were not statistically significant in the subgenomic replicon. Further insight came from unsupervised clustering analysis of the tRNAome profiles [29]. Three biologically independent experiments were performed with na€ ıve Huh7 cells, Huh7 cells infected by HEV with the full-length genome, and cells carrying the subgenomic replicon. The experimental conditions segregated the profiles generated. As expected, the effects on tRNAome provoked by infection of HEV with fulllength genome were much more pronounced as compared to those evoked by the subgenomic replicon (Fig. 4B). We, thus, conclude that ORF2 is essential for HEV-triggered tRNA remodeling.
HEV-induced tRNA remodeling may counteract decoding of interferon-stimulated genes Innate antiviral immunity, especially that mediated by the so-called ISGs, is important for controlling HEV infection [30]. Interestingly, tRNAome remodeling by HEV mainly supported better decoding of ORF2 when Fig. 2. A U-adaptor that interacts specifically with all mature tRNAs. (A) More than 600 tRNA-encoding genes can be transcribed by RNA polymerase III. After post-transcriptional processing, splicing and adding of ACC acceptor to these new tRNA transcripts are performed. The resulting mature tRNAome is available for all cellular mRNA translation and also for viral RNA decoding. The final step in tRNA maturation is adding the CCA sequence to which the acylating amino acid is coupled. The CCA tail is thus common and specific to all mature tRNAs. (B) A universal (U) DNA/RNA hybrid adaptor was designed. The À2 to À4 of 3 0 -terminal bases (TGG) of the U-adaptor are specifically designed to pair CCA acceptor. To further improve the compatibility of U-adaptor to all mature tRNAs, the last 3 0 -terminal nucleotide was degenerately designed. The interface of U-adaptor and mature tRNA is detailed at the right of the panel. compared to ORF1 and ORF3 (Fig. 4C,D). Hence, we analyzed the effects of tRNAome remodeling triggered by full-length infectious HEV on the decoding of essential ISGs. We selected a panel of ISGs known to inhibit HEV infection. We found that HEV-provoked tRNA remodeling is not correlated with the RSCU of these ISGs (Fig. 4C and Table S5), whereas viral ORF2 translation per se is related to the same changes (Fig. 4C,D). We, thus, conclude that HEV infection reprograms the cellular tRNAome, which is likely to facilitate viral translation but hamper cellular antiviral immunity.

Discussion
The development of tRNA quantification methods has been greatly facilitated by the tRNAscan-SE program that allows accurate identification of genomic tRNA sequences. Earlier methods to detect tRNAs usually involve thin-layer chromatography [31], liquid chromatography-mass spectrometry [32], DNA arraying [33], and single tRNA-based qPCR [15]. Recently, quantification of the tRNAome at transcriptional level has become more efficient through high-throughput sequencing [14,34], but is dependent on programming and very specific reagents. In this study, we established a simplified qPCR method for rapid quantification of the mature tRNAome. This strategy is characterized by a length-extension step, a universally compatible adaptor and degenerate primers applicable for the entire mature tRNAome. The present study was only demonstrated in a human cell model, but we envision that it is applicable to other organisms by adjusting primer design. As a proof of principle, this qPCR methodology was used for characterizing the effects of HEV Fig. 3. Workflow of quantifying tRNAome. First, total RNA was extracted from the experimental systems involved according to a routine column-based RNA isolation protocol. Before elution, the column is incubated with diacylation buffer at 37°C for 40 min to remove acylating amino acids at tRNA CCA tail followed by elution with TE buffer. Subsequently, the tRNA and U-adaptor are linked by T4 RNA ligase followed by annealing with reverse primer mix. Finally, the cDNA of total mRNA and U-adaptor-linked tRNAs are concurrently synthesized for qPCR amplification. The relative expression of tRNAs is calculated by 2 ÀDDCt method that can be normalized by housekeeping genes, such as GAPDH. The stepwise procedures are detailed by seventeen discrete steps in Materials and methods.  infection on host tRNAome. We observed substantial remodeling of tRNA composition following HEV infection. For example, the level of tRNA-Pro-GGG was increased by 162 times by HEV infection. Such regulations appear in a fashion that facilitates translation of viral proteins while simultaneously hampering the decoding of antiviral ISGs. Interestingly, these observed effects relate especially to HEV ORF2, which is the dominant ORF of this virus with respect to demands on the host translational machinery [35]. Our data suggest that counteracting HEV-dependent tRNA remodeling may constitute a novel avenue for supporting host defense in combating viral infection, but this will require further understanding of the molecular pathways involved. It has been reported that codon usage bias of the zoonotic genotypes 3 and 4 of HEV is much weaker than that of the nonzoonotic genotype 1 [19]. This indicates that less bias of HEV codon usage may be involved in zoonotic infection. Consistently, we found that genotype 3 HEV has not adapted its codon usage to that of its human host. This suggests that an alternative pathway may be used by HEV to facilitate cross-species infection, such as remodeling of tRNAome. It is well known that viral infection can hijack host cell physiology for weakening host defense and facilitating viral protein synthesis, which is dependent on the cellular mature tRNAome.
Recent studies have indicated that tRNAome disturbance has broad implications in many diseases, especially in cancer development and metastasis [23,36]. Further understanding of the role of tRNAs in biology and pathogenesis requires easy techniques to detect and quantify these molecules. In this respect, the U-adapter strategy explored in the present study may provide new impetus and may form the basis for novel diagnostic possibilities. Hence, we envision that a convenient method, as the one presented in the current study, may prove instrumental for many aspects of contemporary biomedical research and practice with respect to tRNA biology.

Supporting information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Fig. S1. Comparative analysis of codon usage bias of HEV isolated from monkey and human. Fig. S2. Usefulness of U-adaptor to detect 57 types of tRNA sets in human genome.  Fig. S4. Specific and quantifiable expression of tRNAs that can be related to HEV infection levels. Fig. S5. Failed efforts aimed at protocol optimization. Fig. S6. Relative transcription levels of 57 tRNA types following infection with either full length HEV or an ORF2-lacking subgenomic replicon. Table S1. Key materials used in this study. Table S2. Primers used in mature tRNAome detection. Table S3. Preparation of qPCR mix. Table S4. Specifics of the qPCR analysis. Table S5. Interferon stimulated genes (ISGs) used in this study.