Infection-induced 5′-half molecules of tRNAHisGUG activate Toll-like receptor 7

Toll-like receptors (TLRs) play a crucial role in the innate immune response. Although endosomal TLR7 recognizes single-stranded RNAs, their endogenous RNA ligands have not been fully explored. Here, we report 5′-tRNA half molecules as abundant activators of TLR7. Mycobacterial infection and accompanying surface TLR activation up-regulate the expression of 5′-tRNA half molecules in human monocyte-derived macrophages (HMDMs). The abundant accumulation of 5′-tRNA halves also occur in HMDM-secreted extracellular vehicles (EVs); the abundance of EV-5′-tRNAHisGUG half molecules is >200-fold higher than that of the most abundant EV-microRNA (miRNA). Sequence identification of the 5′-tRNA halves using cP-RNA-seq revealed abundant and selective packaging of specific 5′-tRNA half species into EVs. The EV-5′-tRNAHisGUG half was experimentally demonstrated to be delivered into endosomes in recipient cells and to activate endosomal TLR7. Up-regulation of the 5′-tRNA half molecules was also observed in the plasma of patients infected with Mycobacterium tuberculosis. These results unveil a novel tRNA-engaged pathway in the innate immune response and assign the role of “immune activators” to 5′-tRNA half molecules.


Introduction
There are many pathogenic microbes that induce a wide range of symptoms and diseases, including Mycobacterium tuberculosis (Mtb), one of the greatest threats to humans, causing more than 1.2 million deaths annually [1]. When a host is infected with pathogenic microbes, it has 2 essential arms of defense to eliminate them: the innate immune system and the adaptive immune system [2]. In the innate immune system, Toll-like receptors (TLRs) and other pathogen recognition receptors detect pathogen-associated molecular patterns (PAMPs) and initiate protective responses [3,4]. Among the 10 TLRs characterized in humans, TLR1, TLR2, TLR4, TLR5, TLR6, and TLR10 localize to the cell surface (surface TLRs), while TLR3, TLR7, TLR8, and TLR9 localize to intracellular compartments such as endosomes (endosomal TLRs). When TLRs recognize PAMPs, they recruit adaptor proteins, such as MyD88 and TRIF, to initiate signal transduction pathways that culminate in the activation of transcription factors such as NF-κB and AP-1, leading to the production of cytokines and chemokines for host defense [5,6]. occur not only in cell culture systems but also in actual pathological situations. Our study unveils a novel tRNA-engaged pathway in the innate immune response and newly assigned the role of immune activators to 5 0 -tRNA halves.

BCG infection and surface TLR activation induce the expression of 5 0 -tRNA halves in HMDMs
HMDMs express both surface and endosomal TLRs and have been used to study TLR pathways [35,36], while BCG has been used as a model bacterium for tuberculosis infection [37]. In the present study, THP-1-derived HMDMs were infected with viable or heat-killed (HK) BCG, and two 5 0 -tRNA halves (5 0 -tRNA HisGUG half and 5 0 -tRNA GluCUC half; previously abundantly detected in human breast cancer cells [28]) were quantified by tRNA half-specific Taq-Man quantitative reverse transcription PCR (RT-qPCR) [28,30], in which a 3 0 -adapter was ligated to the 5 0 -tRNA half, and then the ligation products were quantified using a TaqMan probe targeting boundary of the adapter and the tRNA half. As shown in Fig 1A, BCG infection enhanced the expression of both of the 5 0 -tRNA halves. The induction of 5 0 -tRNA half expression was independent of the viability of BCG (Fig 1A), suggesting that the induction could result from the pathway of surface TLRs, which recognize BCG PAMPs, or from the process of endocytosis. To examine the involvement of surface TLRs in 5 0 -tRNA half expression, we stimulated TLR4 and TLR2 by treating HMDMs with lipopolysaccharide (LPS) or peptidoglycan (PGN), respectively [38,39]. Successful stimulations of the TLRs were confirmed by upregulation of tumor necrosis factor α (TNFα) and the macrophage inflammatory factors, MIP-1α and MIP-1β (Fig 1B). Upon stimulation of the surface TLRs, the expression of 5 0 -tRNA halves was observed to be up-regulated by TaqMan RT-qPCR (Fig 1C) and northern blot ( Fig  1D). Notably, the expression levels of corresponding mature tRNAs were unchanged by surface TLR stimulation (Fig 1D). As described in previous studies [26,28,30,40], the production of 5 0 -tRNA halves did not influence the levels of mature tRNAs which are steadily maintained by an unknown mechanism. We further analyzed primary human monocyte-derived macrophages (PHMDMs) differentiated from CD14+ monocytes. As in the case of HMDMs, treatment of PHMDMs with LPS or PGN caused surface TLR activation (Fig 1E) and upregulation of 5 0 -tRNA half expression (Fig 1F), confirming the surface TLR-induced expression of 5 0 -tRNA halves in the primary cells of the human body.

Surface TLR-activated NF-κB up-regulates the expression of ANG mRNA
In mammalian cells, ANG cleaves the anticodon loops of tRNAs to produce tRNA halves [25,26,28]. To confirm the involvement of ANG in the tRNA half production in LPS-treated HMDMs, we performed siRNA-mediated knockdown (KD) of ANG expression, which reduced the ANG mRNA levels to around 35% (S1A Fig). The ANG KD decreased the expression of 5 0 -tRNA halves (S1B Fig), suggesting that tRNA halves are generated by ANG-mediated cleavage of tRNAs in LPS-treated HMDMs. Because the expression levels of ANG mRNA were up-regulated upon LPS or PGN treatment in HMDMs (S1C Fig) and PHMDMs (S1D Fig), we reasoned that the transcription factors downstream of surface TLR signal transduction pathways, such as NF-κB, could induce the expression of ANG mRNA. Indeed, direct binding of NF-κB to the region upstream of the ANG gene was suggested by chromatin immunoprecipitation and sequencing (ChIP-seq) data for NF-κB in lymphoblastoid B cells [41] (S1E Fig). The potential involvement of NF-κB in ANG mRNA expression was examined by treating HMDMs with JSH-23, an inhibitor of NF-κB [42,43], which reduced the immune To explore whether EVs secreted from HMDMs are carriers of tRNA halves, we isolated the EVs from the culture medium of LPS-treated HMDMs by an ultracentrifugation-based method. Western blots for the isolated EVs confirmed the presence of CD63 and Alix, proteins known for EV accumulation [44,45], and the absence of calnexin and cytochrome c, which are non-EV proteins [46,47] (Fig 2A). Nanoparticle tracking analysis (NTA) showed the abundant presence of EVs from 80 to 120 nm at a high concentration (approximately 2.0 × 10 7 particles/ ml EV solution) (Fig 2B, S1 and S2 Movies). The isolated EVs were further observed by transmission electron microscopy (Fig 2C), the results of which collectively confirmed the successful isolation of HMDM EVs. The isolated EVs were subjected to TaqMan RT-qPCR for two 5 0 -tRNA halves, 5 0 -tRNA HisGUG half, and 5 0 -tRNA GluCUC half, as well as to stem-loop RT-qPCR for 2 miRNAs, miR-21 and miR-150, which are known to abundantly accumulate in HMDM EVs [44]. We obtained clear amplification signals from all of the 4 examined RNAs. While the EVs treated with RNase alone yielded similar amplification signals to untreated EVs, the EVs treated with both RNase and detergent yielded drastically reduced amplification signals (Fig  2D), confirming that the detected 5 0 -tRNA halves and miRNAs were present inside the isolated EVs and were not captured as non-EV contaminants. We further explored the absolute amounts of the 5 0 -tRNA HisGUG half and miR-150 in LPS-treated HMDMs and their EVs. The calculation of the amounts was based on the standard curve from synthetic RNAs, which showed excellent linearity between input amounts and amplification signals (S2 and S3 Figs).
The determined abundances of the 2 RNAs per μg of total HMDM RNA or per μl of EV fraction are shown in Fig 2E. Although miR-150 was reported as the most abundant miRNA species expressed in HMDMs and their EVs [44], the abundance of the 5 0 -tRNA HisGUG half was much more pronounced than that of miR-150: 136-fold and 215-fold higher in HMDMs and EVs, respectively. 5 0 -tRNA halves are produced from specific tRNA species in HMDMs and are selectively packaged into EVs Given the abundant accumulation of 5 0 -tRNA halves in HMDMs and their EVs, we next identified the expression profiles of the 5 0 -tRNA halves. Although short RNA-seq was previously performed for HMDMs and their EVs [48,49], standard RNA-seq cannot accurately capture 5 0 -tRNA halves because they possess a cP at their 3 0 -end that hinders adapter ligation [28]. Instead, we employed "cP-RNA-seq," which can selectively amplify and sequence cP-RNAs, namely 5 0 -tRNA halves [28,32]. The cP-RNA-seq procedure was first applied to gel-purified short RNA fractions of HMDMs, which successfully amplified approximately 140-to 160-bp bands (considering adapters' lengths, inserted RNA sequences were estimated to be approximately 22 to 42 nucleotides [nt] in length) (Fig 3A). Consistent with the up-regulation of HMDM tRNA half expression by LPS treatment (Fig 1), cP-RNA-seq amplified more abundant cDNAs from the LPS-treated HMDMs than from the untreated cells (Fig 3A). In contrast, we failed to amplify clear cDNA bands from the RNAs of HMDM EVs by cP-RNA-seq, possibly due to the limited amounts of EV-RNAs present. The cP-RNA-seq procedure includes a periodate oxidation step, which might be harsh enough to damage whole RNAs if the initial RNA amounts are limited. Therefore, for EV-RNAs, we decided to capture all short RNA species containing not only a cP but also a phosphate (P) or a hydroxyl group (OH) at the 3 0 -end. For this, EV-RNAs were first treated with T4 polynucleotide kinase (T4 PNK), which can remove cP and P from the 3 0 -end of RNAs, and then were subjected to the short RNA-seq procedure. This yielded abundant approximately 140-to 160-bp cDNA bands (Fig 3B), similar to the bands obtained from cP-RNA-seq of HMDMs (Fig 3A). Interestingly, RNAs treated with a mutant T4 PNK, which lacks 3 0 -dephosphorylation activity [50], yielded only faint cDNA bands, suggesting that the majority of short RNA species in EVs contain a 3 0 -terminal cP or P and RNAs containing a 3 0 -OH end, such as miRNAs, are the minor species in EVs; this is consistent with the experimental results shown in Fig 2E. Illumina sequencing of the gel-purified approximately 140-to 160-bp cDNAs from HMDMs and their EVs yielded approximately 35 to 44 million raw reads, of which >82% to 95% were extracted as reads with a length of 25 to 50 nt (S1 Table). tRNA-mapped reads were enriched in EV libraries (Fig 3C); among them, the 5 0 -tRNA halves were the most major species, as expected (Fig 3D). While 5 0 -tRNA halves comprised approximately 57% of tRNAderived reads in HMDMs, they accounted for over 93% of tRNA-derived reads in EVs, suggesting that 5 0 -tRNA halves could be selectively packaged into EVs to a greater extent than other tRNA-derived RNAs. Considering that the human genome encodes 55 cytoplasmic (cyto) tRNA isoacceptors with different anticodon sequences [51], the identified 5 0 -tRNA halves were derived from a rather focused subset of tRNAs, such as cyto tRNA ValCAC , tRNA Va-lAAC , tRNA GlyGCC , tRNA HisGUG , and tRNA GluCUC , which are in aggregates the sources of 88% to 90% of the identified 5 0 -tRNA halves in EVs (Fig 3E). Among the 5 major 5 0 -tRNA halves, the relative abundance of the 5 0 -tRNA HisGUG half in EVs was considerably greater than that in HMDMs, while the other four 5 0 -tRNA halves were similarly abundant in both libraries ( Fig  3E and 3F), implying preferential incorporation of the 5 0 -tRNA HisGUG half into EVs. tRNA His-GUG contains an additional nucleotide at nucleotide position (np; according to the nucleotide numbering system of tRNAs [52]) -1 of its 5 0 -end. Our recent analyses of BT-474 human breast cancer cells showed that the majority (approximately 60%) of cyto tRNA HisGUG contains G -1 , but a significant proportion contains U -1 or lacks the -1 nucleotide (contains G 1 as a 5 0terminal nucleotide) [53]. As shown in Fig 3G, while the 5 0 -tRNA HisGUG half containing G -1 was the major species in HMDMs, the majority of the 5 0 -tRNA HisGUG halves in EVs lacked the -1 nucleotide (G 1 ). Similarly, while the major 3 0 -terminal nucleotide was U 33 for HMDM 5 0 -tRNA HisGUG halves, the majority of the EV-5 0 -tRNA HisGUG halves contained G 34 as the 3 0 -end. The 5 0 -tRNA HisGUG half from G 1 to G 34 comprised approximately 80% of EV-5 0 -tRNA HisGUG halves but only 5% of HMDM 5 0 -tRNA HisGUG halves. Inconsistency of the identified species between HMDMs and EVs was also observed in some other major 5 0 -tRNA half species (S4 Fig), implying that the efficiency of EV loading may not be equal for all 5 0 -tRNA halves and specific species could be preferentially packaged into EVs.

EV-5 0 -tRNA halves are delivered into endosomes in recipient HMDMs
Because EV-miRNAs have been shown to be ligands for endosomal TLRs [10,11], we examined whether the abundantly identified EV-tRNA halves are delivered into endosomes in recipient cells. We chemically tagged synthetic 5 0 -tRNA HisGUG half or 5 0 -tRNA GluCUC half with fluorescein-5-thiosemicarbazide (FTSC) [54] and transfected it into HMDMs, as shown in S5A and S5B Fig. We then isolated the EVs containing the labeled 5 0 -tRNA halves from the transfected cells and subsequently applied them to recipient HMDMs. As a result, we observed the incorporation of the labeled EV-5 0 -tRNA halves into recipient cells. Clear overlap between the signals of the 5 0 -tRNA halves and Rab7 (Fig 4A and 4B), an endosome marker [55], and TLR7 (S5C and S5D Fig) confirmed the delivery of EV-tRNA halves into the endosomes of the recipient HMDMs. These results experimentally proved that tRNA halves in HMDMs are packaged into EVs and secreted outside of the cells, which are then delivered into the endosomes of recipient cells.

0 -tRNA HisGUG half activates endosomal TLR7
Given the abundant accumulation and endosome-targeted delivery of 5 0 -tRNA halves in HMDM EVs, we further assessed the activity of the 5 0 -tRNA halves in stimulating ssRNA-sensing endosomal TLRs (i.e., TLR7 and TLR8). As described in earlier studies [11,56], HMDMs were primed with interferon γ and then transfected with 5 0 -tRNA HisGUG half or 5 0 -tRNA GluCUC half using the cationic liposome 1,2-dioleoyloxy-3-trimethylammonium-propane (DOTAP) which mimics EVs. As controls, a 20-nt HIV-1-derived ssRNA termed ssRNA40 (S2 Fig), known to strongly activate endosomal TLRs [7], and its inactive mutant (ssRNA40-M), in which U is replaced with A, were also transfected. As shown in Fig 5A, transfections of the 5 0 -tRNA HisGUG half and ssRNA40, a positive control, increased the production of TNFα, interleukin (IL)-1β, and IL-12p40 mRNAs, whereas transfections of the 5 0 -tRNA GluCUC half and ssRNA40-M, a negative control, did not. Induction of the secretion of TNFα and IL-1β into culture medium upon the transfection of the 5 0 -tRNA HisGUG half, as well as ssRNA40, was further confirmed by ELISA ( Fig  5B). Transfection of the 5 0 -tRNA HisGUG half using Lipofectamine reagents did not show such inductions (S6 Fig), confirming that the delivery of 5 0 -tRNA HisGUG half to endosomes, not to the cytoplasm, is necessary for the inductions. The strong activation of endosomal TLR by the DOTAP-fused 5 0 -tRNA HisGUG half was further observed in PHMDMs. Upon transfection of 5 0 -tRNA HisGUG half into PHMDMs, increased production of TNFα, IL-1β, and IL-12p40 mRNAs (Fig 5C) and enhanced secretion of TNFα and IL-1β (Fig 5D) were observed. While the calculation of Fig

PLOS BIOLOGY
nucleotides, Q34 has been reported to block ANG-mediated anticodon cleavage [62] and thus would be absent in the 5 0 -tRNA HisGUG half. The synthetic 5 0 -tRNA HisGUG half containing the other 4 modified nucleotides (S2A Fig) activated endosomal TLRs as strongly as unmodified RNA (Fig 5E), suggesting that the endogenous, modified 5 0 -tRNA HisGUG half would have the activity. Although mature tRNAs have been reported to be incorporated in EVs [19,20], interestingly, the full-length tRNA HisGUG was incapable of stimulating endosomal TLR (Fig 5F) possibly due to its rigid secondary and tertiary structures. These results suggest that shortening mature tRNA HisGUG into less-rigid 5 0 -half molecules by anticodon cleavage is necessary to activate endosomal TLR.
We next examined whether the 5 0 -tRNA HisGUG half activates endosomal TLR7 and/or TLR8. siRNA-mediated KD of TLR7 alone or simultaneous KD of TLR7 and TLR8 in HMDMs abolished the up-regulation of TNFα, IL-1β, and IL-12p40 by DOTAP transfection of the 5 0 -tRNA HisGUG half, whereas TLR8 KD alone did not (Fig 6A, S8A-S8C Fig). Relative abundance half stimulates endosomal TLR7 as strongly as ssRNA40, but not TLR8. To further confirm the involvement of TLR7 in the activity of the 5 0 -tRNA HisGUG half, by using CRISPR/Cas9 approach, we generated TLR7 knockout (KO) THP-1 cell lines in which TLR7 expression is completely abolished (Fig 6B). The 5 0 -tRNA HisGUG half did not show the activity to stimulate endosomal TLR in TLR7 KO cells (Fig 6C), confirming that the 5 0 -tRNA HisGUG half activates endosomal TLR7.
To test whether the EV-5 0 -tRNA HisGUG half activates TLR7, we transfected the 5 0 -tRNA HisGUG half, 5 0 -tRNA GluCUC half, and ssRNA40-M (negative control) into HMDMs, and the EVs isolated from the cells were applied to recipient HMDMs. As shown in Fig 7A, EVs isolated from HMDMs that transiently expressed the 5 0 -tRNA HisGUG half were able to induce immune response. To further confirm the activity of endogenous EV-5 0 -tRNA HisGUG halves, we utilized antisense oligonucleotides of the 5 0 -tRNA HisGUG half and control oligonucleotides with scrambled sequences. In a DOTAP transfection experiment, both oligonucleotides did not show activity for endosomal TLR by themselves (Fig 7B). When mixed with an equal amount of 5 0 -tRNA HisGUG half, the antisense oligonucleotides impaired TLR7 activation by the 5 0 -tRNA HisGUG half, but the control oligonucleotides did not (Fig 7B), confirming the antisense oligonucleotides' activity to block the 5 0 -tRNA HisGUG half. In the experiment using the EVs isolated from HMDMs, strikingly, the antisense oligonucleotides of the 5 0 -tRNA HisGUG half reduced the EVinduced up-regulation of TNFα and IL-1β by 40% to 60% (Fig 7C). Taken together, these results confirmed that endogenous 5 0 -tRNA HisGUG halves, which are transferred from EVs to recipient cells, have activity to promote cytokine productions by stimulating endosomal TLR7.   S9C Fig), the detected 5 0 -tRNA halves in plasma samples were expected to be mostly present inside plasma EVs. We then quantified the 5 0 -tRNA haves in the plasma samples from healthy individuals or Mtb-infected patients. Because the expression of tRNA halves can be affected by sex hormones [28] and aging [34], we limited the examined individuals to males aged 30 to 35 years. During RNA extraction, a synthetic mouse piRNA was added as a spike-in control, and its abundance was used for normalization. As shown in Fig 8B, the expression levels of 2 examined 5 0 -tRNA halves were markedly enhanced in Mtb-infected patients compared to healthy individuals. The 5 0 -tRNA HisGUG half in particular was highly elevated at approximately 10-fold higher in Mtb-infected patients than in control individuals. These results suggest that the up-regulation and secretion of 5 0 -tRNA halves upon infection are not limited to cell culture settings but also occur in actual pathological situations in pathogenic microbe-infected patients.

Discussion
Here, we identified a novel role of 5 0 -tRNA halves as activators of TLR7. Both BCG infection and PAMP-mediated surface TLR activation induced the expression of 5 0 -tRNA halves in HMDMs. Considering the results of earlier studies on the function of 5 0 -tRNA halves in the stress response, translation, and cell proliferation [28,[63][64][65], infection-induced 5 0 -tRNA halves could function in various biological processes inside macrophages. In the present study, we focused on the secretion of 5 0 -tRNA halves into EVs and their role as stimulators of endosomal TLRs in recipient cells. Strikingly, our analyses revealed the abundant accumulation of 5 0 -tRNA halves in HMDM-secreted EVs and their delivery to endosomes in recipient cells for the activation of TLR7. We propose that infection-induced 5 0 -tRNA halves function as "immune activators" by being delivered to endosomes in surrounding cells via EV-mediated cell-cell communication and by activating TLR7 (Fig 9).

PLOS BIOLOGY
Previous studies have shown that stress stimuli and sex hormone signaling pathways induce ANG-catalyzed cleavage of the anticodon loop of tRNAs, leading to the expression of tRNA halves termed tRNA-derived stress-induced RNAs (tiRNAs) and sex hormone-dependent tRNA-derived RNAs (SHOT-RNAs), respectively [25,26,28]. In tiRNA biogenesis, tRNA cleavage is triggered by decreased levels of RNH1, an ANG inhibitor, which increase ANG availability for tRNA cleavage [66]. Although the mechanism of SHOT-RNA biogenesis is unknown, estrogen or androgen receptors, functioning as transcription factors, might regulate the expression of ANG and/or RNH1. In the case of infection-induced tRNA halves, our analyses revealed that TLR-activated NF-κB up-regulates the expression levels of ANG mRNA, potentially leading to enhanced levels of ANG protein available for tRNA cleavage. If this is the mechanism behind tRNA half production, because not only surface TLR pathways but also the TLR7 pathway culminates in NF-κB activation, there could be a feed-forward loop in which TLR7 activation by 5 0 -tRNA halves induces the expression of 5 0 -tRNA halves for further activation of TLR7. In addition, because dysregulation of NF-κB is linked to various diseases, such as cancers and inflammatory and autoimmune diseases [67][68][69], the potential regulation of tRNA half production by NF-κB suggests the involvement of tRNA halves in such diseases.
By using cP-RNA-seq, we identified the complete expression repertories of 5 0 -tRNA halves in HMDMs and their secreted EVs, revealing that only specific tRNA species serve as major substrates for infection-induced tRNA half expression. The molecular mechanism underlying the anticodon loop cleavage of specific tRNA species remains unknown. Because major substrate tRNAs such as cyto tRNA ValCAC , tRNA ValAAC , tRNA GlyGCC , tRNA HisGUG , and tRNA GluCUC were also identified as major sources of SHOT-RNAs in human breast cancer cells [28], those tRNAs may be universally susceptible to ANG cleavage, or the molecular factors determining the susceptibility of tRNAs to anticodon cleavage, such as tRNA modifications, may be regulated similarly between the biogenesis of sex hormone-and infectioninduced tRNA halves. The difference in the expression profiles of 5 0 -tRNA halves between HMDMs and their secreted EVs suggests selective packaging of 5 0 -tRNA halves into EVs. Selective packaging of the 5 0 -tRNA HisGUG half into EVs is intriguing as this half is highly active in TLR7 stimulation. Although the mechanism of EV RNA content selection is unknown, biased EV incorporation has been also shown for miRNAs [70][71][72] and tRFs [73,74]. Because Y-box protein 1 (YBX1) has been reported to interact with 5 0 -tRNA halves [64] and has been implicated in the sorting of miRNAs for packaging into EVs [72], such RNA-binding proteins could be involved in the selective packaging of 5 0 -tRNA halves. Among the cellular 5 0 -tRNA His-GUG half species, only a specific 5 0 -tRNA HisGUG half, from G 1 to G 34 , is preferentially packaged into EVs. Specific sequences and/or secondary/tertiary structures may contribute to preferential binding to RNA-binding proteins responsible for EV packaging. Indeed, in the case of miRNAs, specific 3 0 -terminal sequences are required to interact with heterogeneous nuclear ribonucleoprotein A2/B1 for preferential incorporation into EVs [70].
One of the most remarkable characteristics of 5 0 -tRNA halves is their abundance. Although miR-150 was identified as one of the most abundant miRNAs in HMDMs and their EVs [44], the present quantification revealed the abundance of the 5 0 -tRNA HisGUG half in HMDMs and EVs to be over 130-fold and 210-fold higher, respectively. Although miRNAs have been shown to function as ligands for TLR7, considering ligand-receptor interactions, 5 0 -tRNA halves with much more abundance could be more efficient, superior TLR ligands than miRNAs. Given that T4 PNK treatment greatly enhanced amounts of EV-cDNAs during our sequencing procedure, it is predicted that EV-short ncRNA species are mostly 3 0 -P-or cP-containing RNAs, such as 5 0 -tRNA halves, and that 3 0 -OH-containing RNAs, such as miRNAs, are minor species. While studies on EVs have established the role of EV-RNAs as cell-cell communication agents [75], most current studies rely on standard RNA-seq, which cannot capture the 3 0 -P or cP-containing RNAs that account for the majority of short RNA species in EVs. Our results suggest the necessity of shedding light on these previously unrecognized RNAs by pretreating EV-RNA fractions with T4 PNK in sequencing studies. Giraldez and colleagues revealed previously unexplored mRNA and lncRNA fragments by phosphor-RNA-seq whose procedure includes T4 PNK treatment [76].
Another striking feature of the 5 0 -tRNA HisGUG half is its ability to strongly activate TLR7, but not TLR8. This selective activity for TLR7 might result from the high sensitivity of TLR7 to GU-rich ssRNAs, such as the 5 0 -tRNA HisGUG half, while TLR8 senses AU-rich ssRNAs [77]. The activation of TLR7 by the 5 0 -tRNA HisGUG half is as high as that by SSRNA40, suggesting the role of the 5 0 -tRNA HisGUG half as an endogenous ligand for TLR7 with the full capacity to produce an immune response. On the other hand, the 5 0 -tRNA GluCUC half did not activate TLR7. Because the 5 0 -tRNA GluCUC half and the 5 0 -tRNA HisGUG half were similarly delivered to recipient endosomes in our delivery experiments, the inactivity of the 5 0 -tRNA GluCUC half is probably due to its inefficient binding to TLR7. The lack of 3 0 -terminal GU-rich sequences may be one of the reasons for the inefficient activity of 5 0 -tRNA GluCUC toward TLR7 as previous study showed significance of 3 0 -terminal GU sequences in let-7 miRNA for TLR7 activation [10]. Intriguingly, unlike the 5 0 -tRNA HisGUG half, the full-length tRNA HisGUG is incapable of activating TLR7, suggesting the cruciality of tRNA cleavage and production of tRNA half molecules to yield active ligands for TLR7.
Finally, we showed the elevation of 5 0 -tRNA half levels in the plasma of Mtb-infected patients, demonstrating the expressional induction and secretion of 5 0 -tRNA halves in actual pathological situations. Because up-regulation of 5 0 -tRNA half expression has been reported upon infection with respiratory syncytial virus [78,79], Rickettsia [80], and hepatitis B and C viruses [81], induction of 5 0 -tRNA halves could be a universal phenomenon among infectious diseases. Considering the expressional differences and the demonstrated roles of 5 0 -tRNA halves in the innate immune response, further characterization of 5 0 -tRNA halves may lead to the use of 5 0 -tRNA halves as potential target candidates for future therapeutic applications and/or circulating biomarkers for noninvasive testing to estimate the severity of infectious diseases and the status of the immune response.

Ethical approval
The Office of Human Research (OHR) of Thomas Jefferson University (TJU) approved our use of patient samples without private information in accordance with all federal, institutional, and ethical guidelines (#OHR-19: Expressions of noncoding RNAs in human plasma and serum samples). We obtained the plasma samples from a company BioIVT (Westbury, New York, United States of America) without receiving patients' information.

EV isolation
EVs were isolated from the culture medium of LPS-treated HMDMs according to an ultracentrifugation-based method described previously [44]. In brief, dead cells and cell debris in the culture medium were removed by successive centrifugation at 300 g for 10 min, 2,000 g for 10 min, and 10,000 g for 30 min. The supernatant was then ultracentrifuged using Sorvall WX + Ultracentrifuge Series (Thermo Fisher Scientific) at 110,000 g for 2 h. The pellet was washed with PBS and ultracentrifuged again at 110,000 g for 2 h to eliminate contaminant proteins. The final pellet was collected as the EV fraction. The data regarding EV isolation and characterization are available in EV-TRACK database (EV-TRACK ID: EV190062) [85]. To confirm the presence of EV-RNAs, the isolated EVs were incubated with PureLink RNase A (200 ng/μl, Thermo Fisher Scientific) with or without 0.1% Triton X-100 at 37˚C for 30 min.

NTA and transmission electron microscopy
Size distributions of the isolated EVs were analyzed by NTA using NanoSight NS300 (Malvern Analytical, Malvern, United Kingdom), as described previously [86], at the Flow Cytometry Facility of the Sidney Kimmel Cancer Center at TJU. The isolated EVs were further visualized by transmission electron microscopy (JEOL, Akishima, Tokyo, Japan) at the Centralized Research Facilities at Drexel University.

Quantification of RNAs by TaqMan RT-qPCR, stem-loop RT-qPCR, and standard RT-qPCR
Total RNA from the cells and EVs was isolated using TRIsure (Bioline, Swedesboro, New Jersey, USA). TaqMan RT-qPCR for specific quantification of 5 0 -tRNA halves was performed according to our previously described tRNA half quantification method [28]. Briefly, to remove cP from 5 0 -tRNA halves, total RNA was treated with T4 PNK, followed by ligation to a 3 0 -RNA adapter by T4 RNA ligase. Ligated RNA was then subjected to TaqMan RT-qPCR using the One Step PrimeScript RT-PCR Kit (Takara Bio, Kusatsu, Shiga, Japan), 200 nM of a TaqMan probe targeting the boundary of the target RNA and the 3 0 -adapter, and forward and reverse primers. The TaqMan probe and primer sequences are shown in S2 Table. Stem-loop RT-qPCR for quantification of miRNAs and piRNAs was performed as previously described [87,88]. In brief, total RNA was treated with DNase I (Promega, Madison, Wisconsin, USA) and subjected to reverse transcription using SuperScript III reverse transcriptase (Thermo Fisher Scientific) and a stem-loop reverse primer. The synthesized cDNAs were then subjected to PCR using SsoFast Evagreen Supermix (Bio-Rad, Hercules, California, USA) and forward and reverse primers. Sequences of the primers used are shown in S3 Table. Standard RT-qPCR was used for quantification of mRNAs. Briefly, DNase I-treated total RNA was subjected to reverse transcription using RevertAid Reverse Transcriptase (Thermo Fisher Scientific) and a reverse primer. The synthesized cDNAs were then subjected to PCR using 2×qPCR Master Mix (Bioland Scientific, Paramount, California, USA) and forward and reverse primers. Sequences of the primers used are shown in S4 Table. Northern blot Northern blot was performed as previously described [28] with the following antisense probes: 5 0 -tRNA HisGUG half, 5 0 -CAGAGTACTAACCACTATACGATCACGGC-3 0 ; 5 0 -tRNA GluCUC half, 5 0 -GCGCCGAATCCTAACCACT-3 0 ; and miR-16, 5 0 -GCCAATATTTACGTGCTGC TA-3 0 .

cP-RNA-seq and bioinformatics
For cP-RNA-seq, 25-50-nt RNAs were gel-purified from the total RNA of LPS-treated HMDMs and subjected to the cP-RNA-seq procedure as previously described [28,30,[32][33][34]. For EV-5 0 -tRNA half sequencing, EV-RNA was first treated with T4 PNK to remove cP from the 5 0 -tRNA halves, followed by adapter ligation and cDNA amplification using the TruSeq Small RNA Sample Prep Kit (Illumina, San Diego, California, USA). The amplified cDNAs were gel-purified and sequenced using the Illumina NextSeq 500 system at the MetaOmics Core Facility of the Sidney Kimmel Cancer Center at TJU. The sequence libraries contain approximately 35 to 44 million raw reads (S1 Table) and are publicly available from the NCBI Sequence Read Archive (accession no. SRR8430192, SRR8430191, and SRR8430193). Bioinformatic analyses were performed as described previously [33,34]. Reads were mapped to 471 mature cyto tRNAs obtained from GtRNAdb [51], and then to mature rRNAs, to mRNAs of RefSeq with NM-staring accession numbers (NM is an accession prefix of known RefSeq), to the mitochondrial genome (GenBank: CM001971.1 sequence plus 22 mitochondrial tRNA sequences), and to the whole genome (GRCh37/hg19).

In vitro RNA synthesis
The synthetic RNAs used in this study are shown in S5 Table. While antisense oligonucleotides, miRNAs, and a piRNA (spike-in) were synthesized by Integrated DNA Technologies, 5 0 -tRNA halves, FL-tRNA HisGUG , and ssRNA40 were synthesized by an in vitro reaction as described previously [53]. dsDNA templates were synthesized using PrimeSTAR GXL DNA Polymerase (Takara Bio) and the primers shown in S6 Table. The templates were then subjected to an in vitro transcription reaction with T7 RNA polymerase (New England Biolabs, Ipswich, Massachusetts, USA) at 37˚C for 4 h. For 5 0 -tRNA GluCUC half production, the in vitro synthesized RNA contained the ribozyme sequence to generate a mature 5 0 -end as described previously [89], so the reaction mixture was further incubated for 3 cycles at 90˚C for 2.5 min and 37˚C for 15 min, allowing the ribozyme reaction. The synthesized RNAs were then gelpurified using denaturing PAGE with single-nucleotide resolution, and the quality of the gel-

Fluorescent labeling of 5 0 -tRNA halves and their EV-mediated delivery to cells
The synthetic 5 0 -tRNA HisGUG half and 5 0 -tRNA GluCUC half were fluorescent-labeled at their 3 0end based on a previously described method [54]. In brief, synthetic RNAs were incubated in 100 mM NaOAc (pH 5.2) and 100 μM NaIO 4 at room temperature for 90 min, followed by ethanol precipitation. Then the pellet was dissolved in a solution containing 1.5 mM FTSC (Cayman Chemical, Ann Arbor, Michigan, USA) and 100 mM NaOAc (pH 5.2), followed by overnight incubation at 4˚C. After ethanol precipitation, the labeled RNAs were subjected to Centri-Spin 10 (Princeton Separations, Adelphia, New Jersey, USA) purification to remove unreacted FTSC. Then, 80 pmol of the labeled RNA was transfected into HMDMs using RNAiMAX (Thermo Fisher Scientific). After 24 h, the cells were washed with PBS and further incubated for 12 h with LPS, and the cell culture medium was subjected to EV isolation as described above. The isolated EV fraction was then added to HMDMs, followed by incubation for 6 h, and visualization of the labeled 5 0 -tRNA halves with Rab7 and TLR7 by confocal microscopy as described below.

DOTAP-mediated RNA delivery to endosomes
To deliver RNAs to endosomes, we used DOTAP liposomal transfection reagent (Sigma-Aldrich) as previously described [11,56]. In brief, 230 pmol or other various amounts of synthetic RNAs were mixed with 60 μl of HBS buffer and 15 μl of DOTAP reagent and incubated for 15 min. The RNA-DOTAP solution was then added to 1 ml HMDM or PHMDM medium, followed by incubation of the cells for 16 h.

EV-mediated RNA delivery to endosomes
Synthetic 5 0 -tRNA HisGUG half, 5 0 -tRNA GluCUC half, and ssRNA40-M (80 pmol) were transfected to HMDMs (9 × 10 6 cells) using RNAiMAX (Thermo Fisher Scientific). After 24 h, the cells were washed with PBS and further incubated for 12 h, and the cell culture medium was subjected to EV isolation as described above. The isolated EVs were then added to HMDMs (1 × 10 6 cells), followed by incubation for 12 h, RNA extraction, and RT-qPCR quantification of TNFα, IL-1β, and IL-12p40 mRNAs.
Regarding experiments using antisense oligonucleotides, control oligonucleotides with scrambled sequences or antisense oligonucleotides for the 5 0 -tRNA HisGUG half (S5 Table) were first infused with DOTAP as described above. EVs isolated from LPS-treated HMDMs were mixed with the DOTAP-oligonucleotides solution and then were applied to recipient HMDMs, followed by incubation for 16 h. To eliminate possible effects of potential endotoxin (LPS) contamination, EVs isolated from LPS-treated HMDMs were incubated with 10 mg/ml polymyxin B (PMB) (Sigma-Aldrich) at 4˚C for 1 h prior to mixing with the DOTAP-oligonucleotides solution.

ELISA
For ELISA experiment, RNA transfection using DOTAP was performed in Opti-MEM (Thermo Fisher Scientific), and the culture medium from 1 × 10 6 HMDMs or 1 × 10 5 PHMDMs was subjected to ELISA (R&D Systems, Minneapolis, Minnesota, USA) for quantification of TNFα and IL-1β. Their absolute amounts were calculated based on standard curves.

TLR7 KO THP-1 cell lines
TLR7 KO THP-1 cells were generated using the CRISPR/Cas9 system at Genome Editing Institute in ChristianaCare. Two different clones, KO #1 and KO #2, were generated using gRNA1 (5 0 -ACUUUCAGGUGUUUCCAAUG-3 0 ) and gRNA2 (5 0 -UAGGAAACCAUCUAGCCC CA-3 0 ), respectively. The KO cells were differentiated into HMDMs and used for transfection of DOTAP-fused RNAs as described above. Confirmation of TLR7 depletion in the KO cells was done using western blot analysis as described above.

Human plasma samples and RNA isolation
Human plasma samples were derived from healthy or Mtb-infected males aged 30 to 35 years and obtained from BioIVT. For RNA isolation, 500 μl of plasma was first centrifuged at 16,060 g for 5 min, then 400 μl of supernatant was mixed with synthetic mouse piR-3 spike-in control (20 fmol) and subjected to RNA extraction using TRIzol LS (Thermo Fisher Scientific). The extracted RNAs were further subjected to purification using the miRNeasy Mini Kit (Qiagen, Hilden, Germany). Based on the quantification of miR-451 and miR-23a-3p and calculation of "miR ratio" as described earlier [91], no hemolysis was observed in any of the plasma samples. The extracted RNA samples were subjected to quantification of 5 0 -tRNA halves, and quantification of piR-3 (spike-in) was used for normalization. EVs produced from host HMDMs containing the labeled 5 0 -tRNA HisGUG half or 5 0 -tRNA GluCUC half were isolated and applied to recipient HMDMs. Delivery of the labeled, EV-5 0 -tRNA HisGUG half (C) or EV-5 0 -tRNA GluCUC half (D) into endosomes was observed in green. Immunofluorescence staining of TLR7 is shown in red, and DNA was counterstained with DAPI in blue. Scale bar, 100 μm. Clear co-localization of the labeled 5 0 -tRNA halves and TLR7 was observed. EV, extracellular vehicle; HMDM, human monocyte-derived macrophage; TLR, Toll-like receptor; tRNA, transfer tRNA. The indicated amounts of the synthetic 5 0 -tRNA HisGUG half were transfected into HMDMs using DOTAP. Total RNAs from the cells were subjected to RT-qPCR for the indicated mRNAs. Averages of 3 experiments with SD values are shown. (B) After the RNA transfection, culture medium was subjected to ELISA for quantification of TNFα and IL-1β. DOTAP, 1,2-dioleoyloxy-3-trimethylammonium-propane; HMDM, human monocyte-derived macrophage; IL, interleukin; mRNA, messenger RNA; RT-qPCR, quantitative reverse transcription PCR; SD, standard deviation; TLR, Toll-like receptor; TNFα, tumor necrosis factor α; tRNA, transfer tRNA. (EPS) S8 Fig. siRNA-mediated KD of TLR7 and TLR8 in HMDMs. (A) HMDMs were transfected with control siRNA (siControl) or siRNA targeting TLR7 (siTLR7) or TLR8 (siTLR8). To confirm the reduction of the targeted mRNA, total RNAs from the cells were subjected to RT-qPCR for TLR7 and TLR8 mRNAs (RPLP0: control). (B) Double KDs of TLR7 and TLR8 were performed by simultaneously transfecting both siTLR7 and siTLR8, and reduction of the both mRNAs was confirmed by RT-qPCR. (C) In HMDMs, the expression of both TLR7 and TLR8 was silenced by siRNAs and then DOTAP-fused 5 0 -tRNA HisGUG half or ssRNA40-M was transfected. Total RNAs from the cells were subjected to RT-qPCR for the indicated mRNAs. DOTAP, 1,2-dioleoyloxy-3-trimethylammonium-propane; HMDM, human monocytederived macrophage; KD, knockdown; mRNA, messenger RNA; RT-qPCR, quantitative reverse transcription PCR; TLR, Toll-like receptor.  (TIF) S1 Data. Numerical data underlying Fig 1A-1C, 1E and 1F; Fig 2B and 2D; Fig 3C-3G; Fig  5A-5F; Fig 6A and 6C; Fig 7A-7C; Fig 8A and 8B; and S1A-S1D and S1F and S1G