tRNA‐Derived Small RNAs in Digestive Cancers: From Translational Regulation to Immune and Extracellular Communication

ABSTRACT

Transfer RNA‐derived small RNAs (tsRNAs), comprising tRNA‐derived fragments (tRFs) and stress‐induced tRNA halves (tiRNAs), have increasingly been recognized as an important regulatory class in gastric cancer (GC), colorectal cancer (CRC), hepatocellular carcinoma (HCC), and pancreatic cancer/pancreatic ductal adenocarcinoma (PC/PDAC). Research in this field has expanded from expression profiling and liquid biopsy to non‐canonical translational control, metabolic adaptation, therapy resistance, immune‐associated remodeling, and extracellular‐vesicle (EV)‐related communication. The most intensively studied and mechanistically developed area currently lies at the intracellular level. In digestive system tumors, tsRNAs can act through EIF4 displacement, AGO2/RISC‐dependent silencing, direct target repression, and ribosome‐associated interactions, with some of these mechanisms validated in animal models. Beyond direct regulation of gene expression, tsRNAs can also influence tumor metabolic state and thereby contribute to chemo‐ and radio‐resistance. By contrast, studies directly examining the effects of tsRNAs on immune‐cell populations remain relatively limited, and work on EV‐mediated systemic propagation still largely focuses on vesicle association and biomarker value. Functional delivery and recipient‐cell effects require further clarification. Among the currently summarized studies, pancreatic‐derived signaling that conditions the hepatic niche represents one of the few examples approaching a cross‐organ functional model. This review discusses the major functional layers of tsRNAs in digestive system tumors, beginning with the relatively mature intracellular mechanisms and then extending to emerging immune, extracellular/systemic, and host–microbe research. We also identify key unresolved problems, including nomenclature standardization, modification‐aware sequencing, criteria for EV functional delivery, causal validation of microbiota‐derived tsRNAs, and prospective biomarker validation against benign and inflammatory disease controls.

tRNA‐derived small RNAs (tsRNAs) function in digestive cancers across three scales: intracellular translational and metabolic control, extracellular vesicle–mediated immune remodeling, and systemic and host–microbe communication. This review organizes these mechanisms by evidence strength, from well‐established intracellular roles to emerging communication axes.

Abbreviations

AFP

alpha‐fetoprotein

AGO2

Argonaute 2

AUC

area under the receiver operating characteristic curve

CA19‐9

carbohydrate antigen 19–9

CA72‐4

carbohydrate antigen 72–4

CEA

carcinoembryonic antigen

CRC

colorectal cancer

EIF4A1

eukaryotic initiation factor 4A‐I

EMT

epithelial‐mesenchymal transition

EV

extracellular vesicle

GC

gastric cancer

HCC

hepatocellular carcinoma

HSC

hepatic stellate cell

ICI

immune checkpoint inhibitor

MDSC

myeloid‐derived suppressor cell

MICA

MHC class I chain‐related protein A

MISEV

Minimal Information for Studies of Extracellular Vesicles

ncRNA

non‐coding RNA

NK

natural killer

OS

overall survival

PC

pancreatic cancer

PDAC

pancreatic ductal adenocarcinoma

PDX

patient‐derived xenograft

PIVKA‐II

protein induced by vitamin K absence or antagonist‐II

PMN

pre‐metastatic niche

PSC

pancreatic stellate cell

RBP

RNA‐binding protein

RISC

RNA‐induced silencing complex

ROC

receiver operating characteristic

TAM

tumor‐associated macrophage

tDR

tRNA‐derived RNA

tiRNA

stress‐induced tRNA half

TME

tumor microenvironment

tRF

tRNA‐derived fragment

tsRNA

tRNA‐derived small RNA

1. Introduction

Once regarded as random degradation products, tRNA‐derived small RNAs (tsRNAs) are now recognized as regulated small RNAs with defined biogenesis, subclass diversity and biological function. Their significance cannot be fully explained by canonical small‐RNA silencing alone: tsRNAs can interact with translation‐initiation factors, ribosomal subunits, RNA‐binding proteins and AGO complexes through molecular modes that do not always have a clear miRNA counterpart [1, 2, 3, 4, 5, 6]. In addition, parental tRNA modifications shape tsRNA production and function, adding another layer of context dependence to fragment interpretation [4, 7, 8]. These features have moved the field from fragment cataloging toward mechanistic investigation.

Digestive system tumors provide a particularly suitable context for studying this layer of regulation. The gastric, colonic, hepatic, and pancreatic compartments are chronically exposed to metabolic fluctuation, hypoxia, oxidative stress, inflammatory remodeling, microbiota‐derived signals, and bile acid‐related stress; these cellular environments overlap with the stress and injury contexts in which tRNA cleavage and tsRNA generation are frequently observed [9, 10, 11, 12]. This biological fit is especially relevant because digestive cancers are not only driven by tumor‐cell‐intrinsic signaling but also shaped by metabolic state, inflammatory remodeling, immune surveillance, and intercellular or inter‐organ communication.

Research on tsRNAs in digestive cancers has therefore expanded from expression profiling and circulating biomarker discovery to mechanistic studies of translation control, RBP‐mediated regulation, AGO2/RISC‐dependent target repression, ribosome‐associated regulation, metabolic adaptation, therapy resistance, immune‐associated remodeling, and EV‐related communication [1, 2, 3, 5, 6, 7, 8, 13, 14, 15, 16, 17, 18, 19]. However, the strength of evidence is uneven across these layers. Intracellular mechanisms currently have the strongest experimental support, whereas immune remodeling, EV‐mediated functional transfer, systemic propagation, and microbiota‐derived host–microbe communication remain more exploratory and require stricter causal validation [13, 15, 16, 17, 18, 19].

Existing disease‐ or pathway‐oriented reviews provide important background for the field [1, 2, 3, 20, 21, 22, 23, 24, 25, 26, 27]. Here, we first introduce tRNA biogenesis and tsRNA subclass classification, and then place digestive‐tumor tsRNA studies into a mechanism‐oriented evidence framework. We emphasize primary mechanistic studies and functional validation evidence, while distinguishing expression association, circulating detectability, EV cargo enrichment, and causal activity in recipient cells throughout the discussion.

2. tRNA Biogenesis and tsRNA Classification

Canonical tRNAs are transcribed by RNA polymerase III as precursor tRNAs that contain a 5′ leader sequence, a 3′ trailer sequence, and, in some cases, intronic regions. During maturation, the 5′ leader is removed by RNase P, the 3′ trailer is processed by RNase Z/ELAC2, introns are spliced when present, and a CCA tail is added to the 3′ end. Mature tRNAs then acquire extensive chemical modifications that stabilize their structure, maintain decoding fidelity, and influence their susceptibility to ribonuclease‐mediated cleavage. This maturation process is important for understanding tsRNA biology because different tsRNA subclasses originate from different regions of precursor or mature tRNAs, rather than representing a single homogeneous small‐RNA population [1, 2, 3, 4].

According to their position within the parental tRNA, tsRNAs are commonly divided into several major subclasses. tRF‐1 fragments are generated from the 3′ trailer sequence of precursor tRNAs after RNase Z/ELAC2‐mediated processing. tRF‐5 fragments arise from the 5′ end of mature tRNAs and can be further subdivided according to length and cleavage position. tRF‐3 fragments are derived from the 3′ end of mature tRNAs and often retain the post‐transcriptionally added CCA tail. Internal tRFs originate from internal regions of mature tRNAs and are less easily classified by terminal origin alone. In contrast, tiRNAs, also termed tRNA halves, are longer fragments of approximately 30–40 nucleotides that are usually generated by cleavage around the anticodon loop, particularly under cellular stress. These subclasses differ not only in length and origin but also in their potential loading into AGO complexes, binding to RNA‐binding proteins, interaction with translational machinery and incorporation into extracellular vesicles [1, 2, 3, 4, 28].

The generation of tsRNAs is regulated by both enzymatic cleavage and the modification status of parental tRNAs. Angiogenin/RNase 5 is a well‐established stress‐responsive nuclease that cleaves mature tRNAs at or near the anticodon loop to generate tiRNAs, whereas Dicer and other ribonucleases have been implicated in the production of shorter tRFs. tRNA modifications can either protect tRNAs from cleavage or promote the production of specific fragments, depending on the modification type, the modifying enzyme, and the cellular context. Therefore, the abundance of a given tsRNA reflects not only the expression level of its parental tRNA but also the combined effects of tRNA maturation, chemical modification, stress exposure, and nuclease activity [4, 7, 11, 12, 29, 30].

This classification is useful for interpreting digestive cancer studies because different subclasses may support different functional claims. For example, AGO2‐associated tRFs can act through miRNA‐like target repression, tiRNAs can regulate translation initiation and stress‐granule formation, and selected 3′ tsRNAs or internal fragments may engage ribosomal or RNA‐binding protein‐associated mechanisms. Thus, throughout this review, tsRNA function is interpreted in relation to fragment origin, molecular interaction mode, and level of functional validation rather than by expression change alone [5, 6, 7].

3. Intracellular Effector Mechanisms

Stress‐induced tRNA cleavage is not merely a tumor‐cell‐intrinsic regulatory event, but is also linked to digestive injury, tissue homeostasis and tumor‐initiation contexts. Selitsky et al. found that 5′ tRNA halves were increased in non‐malignant liver tissue from patients and chimpanzee models with chronic HBV or HCV infection and exceeded microRNA abundance in most infected tissues. In matched HCC tissues, however, 5′ tRH abundance was reduced and the relative abundance of individual 5′ tRHs was altered; in HBV‐associated HCC, 5′ tRH abundance also correlated with angiogenin expression [31]. In the intestine, cytoplasmic RNase 5 restrains global translation by generating tiRNAs, thereby suppressing crypt hyperproliferation and tumor initiation [32]. Together, these observations support the idea that stress‐responsive tRNA cleavage can function as a tissue‐protective brake in digestive contexts, although this role is fragment‐ and tissue‐specific and should not be generalized beyond the relevant biological setting.

3.1. Translation‐Initiation Control

The core significance of translation‐initiation factor displacement is that tsRNAs can rapidly reshape cellular stress responses, proliferation, and tumor progression by suppressing the initiation of protein synthesis [5, 33]. Ivanov et al. found that selected tiRNAs can displace eIF4G/eIF4A from mRNAs and release eIF4F from the m7G cap, thereby inhibiting both cap‐dependent and cap‐independent protein synthesis; this process depends on a terminal oligoguanine motif and can induce phospho‐eIF2α‐independent stress‐granule assembly [5]. YB‐1 further participates in tiRNA‐induced stress‐granule assembly as part of the ANG‐mediated stress‐response program [34]. This biochemical mechanism provides a mechanistic basis for related examples in digestive system tumors: in HCC, Wu et al. found that downregulated 5′‐tiRNA‐Gln can fold into an intramolecular G‐quadruplex and bind EIF4A1, thereby reducing EIF4A1–mRNA engagement, partially repressing cap‐dependent translation and ultimately constraining tumor progression [35]. Therefore, the G‐quadruplex/EIF4A1 axis mechanistically parallels the broader oligoguanine/eIF4F‐displacement principle and provides a relatively clear example in digestive system tumors: tsRNAs can act at the level of translation initiation, rather than only regulating target genes at the mRNA‐message level.

Xiong et al. showed, using patient cohorts, in vitro experiments and in vivo models, that tiRNA‐Val‐CAC‐2 binds the single‐stranded nucleic‐acid‐binding protein FUBP1 in pancreatic ductal adenocarcinoma (PDAC). This interaction activates c‐MYC transcription and promotes PDAC cell metastasis [36]. Cui et al. reported that, in gastric cancer (GC), tRF‐Tyr acts through hnRNPD to suppress tumor progression [37], whereas tRF‐Val acts through EEF1A1 to promote tumor‐cell proliferation and inhibit apoptosis [38]. In pancreatic cancer (PC), tRF‐Gly‐CCC‐012 regulates the HNRNPC/PHGDH axis, linking tsRNA–RBP interactions to serine synthesis and metabolic reprogramming [39].

3.2. AGO2/RISC‐Mediated Target Silencing

Canonical RISC loading represents a second relatively well‐defined mode of action. Zhang et al. analyzed 454 tissue samples spanning normal gastric mucosa, gastritis and early gastric carcinogenesis, and found that tRF‐33‐P4R8YP9LON4VDP was progressively lost during gastric cancer (GC) development. Gain‐of‐function experiments showed that tRF‐33 suppresses cell proliferation and metastatic capacity while inducing apoptosis; in nude mouse models, tRF‐33 also reduced tumor‐forming ability. RNA immunoprecipitation, in situ hybridization and dual‐luciferase assays further demonstrated that tRF‐33 binds AGO2 and silences STAT3 through its 3′‐UTR, thereby reducing STAT3/p‐STAT3 and the downstream effectors MMP‐9 and Bcl‐2 [14]. Similarly, Ma et al. found that tRF‐Ser‐TGA‐011 also binds AGO2 and attenuates JNK signaling and Cyclin D1 expression by targeting MAP3K13, thereby inducing cell‐cycle arrest in GC [40]. Based on personalized transcriptome and HITS‐CLIP analyses across cancer types, Telonis et al. further showed that AGO loading of tRFs is cell‐type specific [41]. Therefore, AGO2‐dependent function should not be inferred from AGO2 co‐immunoprecipitation alone; it should be supported by target‐site mutagenesis, dual‐luciferase reporter assays and functional rescue experiments to establish both target specificity and phenotypic causality.

3.3. Direct Target Repression

Direct repression of a validated target represents a third mode of action, especially when causality is supported by phenocopy and rescue experiments. Liu et al. found that, in hepatocellular carcinoma (HCC), 5'tRF‐Gly is highly expressed and associated with tumor size and metastasis. In vitro functional experiments showed that 5'tRF‐Gly promotes tumor‐cell growth and metastasis, and this effect was further validated in nude mouse models. CEACAM1 was confirmed as a direct target of 5'tRF‐Gly; CEACAM1 knockdown reproduced the phenotypic changes induced by 5'tRF‐Gly, whereas restoration of CEACAM1 expression reversed these changes [42]. Zhu et al. reported that tRF‐Pro‐CGG is downregulated in pancreatic cancer (PC). Restoration of tRF‐Pro‐CGG expression inhibited cell proliferation, migration, and invasion, while promoting apoptosis. Dual‐luciferase assays confirmed CSF1 as a direct target of tRF‐Pro‐CGG, and this fragment also modulated the PI3K‐AKT signaling pathway. However, this study was mainly confined to cultured tumor cells and did not further validate in vivo effects or CSF1‐related immune‐cell functions [43]. Other direct‐targeting axes include the tRF‐34‐P4R8YP9LON4VHM/DAB2IP → MEK/ERK → VEGFA axis in HCC [44], the tRF‐23‐Z87HFK8SDZ/IRS1 axis in PC [45], and the tumor‐suppressive HCETSR/β‐catenin‐complex [46], tRF‐24‐6VR8K09LE9 [47] and tRF‐34‐86J8WPMN1E8Y2Q/LRAT [48] axes in GC.

3.4. Ribosome‐Associated Mechanisms

Direct involvement of tsRNAs in the protein‐translation apparatus, especially in ribosome‐related processes, represents a fourth mode of action. A defining foundational example is the 3′ fragment of Leu‐CAG tRNA, namely LeuCAG 3′‐tsRNA. This fragment binds at least two ribosomal‐protein mRNAs, RPS28 and RPS15, and enhances their translation. Loss of this tsRNA reduces RPS28 protein levels, blocks pre‐18S rRNA processing and decreases the number of 40S ribosomal subunits; inhibition of this tsRNA also induces apoptosis in rapidly proliferating cells and in a patient‐derived orthotopic HCC mouse model [6]. Follow‐up work further localized this regulation to a post‐initiation step and showed that its CDS target site is evolutionarily conserved in vertebrates [49]. Chen et al. further showed that ALKBH3 removes m1A/m3C modifications from tRNAs, making them more susceptible to ANG‐mediated cleavage and thereby promoting the generation of anticodon‐region tDRs. These tDRs enhance ribosome assembly and suppress cytochrome‐c‐triggered apoptosis in cancer cells [7]. Together, these studies show that the functional scope of tsRNAs is not limited to RBP‐ or RISC‐mediated regulatory modes. They also suggest that tsRNAs may help explain some tumor‐associated phenotypes by directly influencing ribosome biogenesis and protein‐translation processes. The mechanistic studies underlying this section, together with related foundational models, are summarized in Table 1. TsRNA biogenesis, subclass diversity and the molecular mechanisms through which tsRNAs act in digestive cancers are summarized in Figure 1.

TABLE 1.

Representative mechanistic tsRNA studies in digestive system tumors and related foundational models.

Mechanism category	Tumor/model	tsRNA	Main mechanism/target	Evidence
Translation initiation & RBP control	HCC	5′‐tiRNA‐Gln	G‐quadruplex → EIF4A1, translation repression [35]	Target validation; in vivo
Translation initiation & RBP control	PDAC	tiRNA‐Val‐CAC‐2	FUBP1 → c‐MYC → metastasis [36]	Rescue; in vivo
AGO2/RISC silencing	GC	tRF‐33	AGO2 → STAT3 3′‐UTR → MMP‐9/Bcl‐2 (14)	Reporter/RIP; in vivo
AGO2/RISC silencing	GC	tRF‐Ser‐TGA‐011	AGO2 → MAP3K13 → JNK/Cyclin D1 [40]	Reporter/RIP; in vivo
Direct target repression	HCC	5'tRF‐Gly	CEACAM1 direct target [42]	Phenocopy/rescue; in vivo
Ribosome‐engaging mechanism	HCC PDX (foundational)	LeuCAG3'tsRNA	Binds RPS28/RPS15 mRNAs → ribosome biogenesis [6, 49]	Target validation; PDX
Metabolism/therapy resistance	CRC	tsRNA‐08614	ALDH1A3 → lactylation → oxaliplatin sensitivity [50]	Rescue; in vivo
Modification‐driven resistance	CRC	5′‐tiRNA‐Gly‐GCC	METTL1 m7G → JAK1/STAT6, SPIB → 5‐FU resistance [8]	In vivo; nanoparticle co‐delivery
Immune‐associated remodeling	HCC	tsr_019759	TNFSF15↓/JAK2‐STAT3 → M2 macrophages [15]	In vivo; delivery reversal
Immune escape	CRC	3′‐tRF‐Ala‐CGC	MICA cleavage → NK‐cell escape [17]	In vivo
EV/cross‐organ	PDAC	exo‐tRF‐GluCTC‐0005	WDR1/YAP → HSC → MDSC → hepatic niche [19]	Cross‐organ in vivo chain

Note: Reference numbers correspond to the manuscript reference list.

Abbreviations: AGO2, Argonaute 2; CEACAM1, carcinoembryonic antigen‐related cell adhesion molecule 1; CRC, colorectal cancer; EIF4A1, eukaryotic initiation factor 4A‐I; FUBP1, far upstream element‐binding protein 1; GC, gastric cancer; HCC, hepatocellular carcinoma; HSC, hepatic stellate cell; MDSC, myeloid‐derived suppressor cell; MICA, MHC class I chain‐related protein A; PC, pancreatic cancer; PDAC, pancreatic ductal adenocarcinoma; PDX, patient‐derived xenograft; RBP, RNA‐binding protein; RIP, RNA immunoprecipitation; RISC, RNA‐induced silencing complex; tiRNA, stress‐induced tRNA half; tRF, tRNA‐derived fragment; tsRNA, tRNA‐derived small RNA.

FIGURE 1.

tsRNA biogenesis, subclass diversity and molecular mechanisms in digestive cancers. Three zones are shown. Left: The cloverleaf structure of a tRNA, indicating the 5′ leader and 3′ trailer of the precursor (removed by RNase P and RNase Z/ELAC2, respectively) and the mature 3′‐CCA end, together with where representative tsRNA subclasses arise—tRF‐5 (5′ portion of mature tRNA), tRF‐3 (3′ portion, retaining CCA), tRF‐1 (pre‐tRNA 3′ trailer, generated by RNase Z/ELAC2), internal tRFs (i‐tRF), and tiRNAs/tRNA halves (anticodon‐loop cleavage, mediated by angiogenin under stress); tRFs are processed by Dicer or other RNases in selected contexts. Center: Representative molecular mechanisms and their targets, drawn from a common tsRNA pool—5′‐tiRNA‐Gln (G‐quadruplex‐forming) reduces engagement of the eIF4F helicase EIF4A1 with mRNA, lowering cap‐dependent translation; tRF‐33 (and tRF‐Ser‐TGA‐011) acts through AGO2/RISC on the 3′UTR of STAT3/MAP3K13 to guide silencing; tRF/tiRNA species either directly repress targets (CEACAM1, DAB2IP, CSF1) or act via RNA‐binding proteins (FUBP1 → c‐MYC; HNRNPC → PHGDH); and LeuCAG 3′‐tsRNA stabilizes RPS28/RPS15 mRNAs to enhance 40S ribosome biogenesis. Right: Disease‐relevant functional domains to which these mechanisms contribute (tumor growth/invasion, metabolic adaptation/therapy resistance, immune remodeling, and extracellular communication); these are context‐dependent groupings rather than one‐to‐one mechanism–phenotype links. Solid arrows denote biogenesis or a defined molecular mechanism, T‐bars denote repression or factor displacement, and braces denote context‐dependent functional grouping. The amber, dashed domain (extracellular communication) is an emerging area mapped separately, with its evidence boundaries detailed in Figure 2.

4. Metabolic Reprogramming and Therapy Resistance

Stress‐responsive tsRNA fragments may influence therapeutic response by resetting metabolic states, which provides an important logic linking them to drug and radiation tolerance in digestive system tumors. Because chemotherapy and radiotherapy impose energetic, oxidative, and DNA‐damage stress on tumor cells, tsRNA‐mediated metabolic reprogramming may affect not only basal tumor growth but also adaptive survival under therapeutic pressure [51, 52].

Several axes converge on glycolytic and mitochondrial control. In PC, 5′‐tRF‐19‐Q1Q89PJZ restrains hexokinase‐1‐mediated glycolysis [53]. In CRC, tsRNA‐08614 inhibits glycolysis and histone lactylation via ALDH1A3, suppressing EFHD2 transcriptional activity and conferring oxaliplatin sensitivity [50]. In HCC, hypoxia‐induced tRF‐3Thr‐CGT remodels mitochondrial metabolism through mtDNA translation [54], and in GC the tRF‐Ser/CNBP/HSPA8 axis constrains energy metabolism [55]; the HNRNPC/PHGDH axis couples an RBP interaction to serine‐synthesis flux in PC [39]. This metabolic regulation is not limited to glycolysis and mitochondrial function, but also extends to lipid metabolism and ferroptosis‐related processes. tRF‐23‐Q99P9P9NDD binds the 3′‐UTR of ACADSB and is proposed to reshape lipid metabolism and ferroptosis in gastric cancer [56], whereas the tumor‐suppressive tRF‐E enhances ferroptosis sensitivity in hepatocellular carcinoma, with low expression predicting poor prognosis [57].

The link between tsRNA‐mediated metabolic reprogramming and therapy resistance is one of the most translationally relevant findings in this field. Xu et al. showed that METTL1‐dependent 5′‐tiRNA‐Gly‐GCC promotes 5‐FU resistance in colorectal cancer (CRC) through SPIB and JAK1/STAT6 signaling. In the same study, poly(β‐amino ester) nanoparticles co‐delivering 5‐FU and a tsRNA‐GlyGCC inhibitor enhanced 5‐FU sensitivity in subcutaneous tumor models [8]. In CRC, tRF‐17‐877S6V2 modulates radioresistance through ENO1 [58], whereas tRF‐5a regulates radioresistance through MKNK1 [59]. Related studies further suggest that tsRNA‐mediated metabolic regulation can extend to tumor stemness and signaling states. For example, the NDFIP2/AKT axis enhances liver cancer stem cell‐like properties in HCC [60] whereas lipid and vitamin D3 metabolic reprogramming influences CRC progression through Wnt signaling [61]. The link between tsRNAs and ferroptosis is supported by some experimental evidence, but it is not uniformly validated across studies; in several cases, ferroptosis involvement is inferred mainly from pathway changes rather than from direct endpoints such as lipid peroxidation or rescue with ferroptosis inhibitors. Therefore, tsRNA‐mediated metabolic changes may influence not only therapy resistance but also the immune microenvironment, providing a transition to the following discussion of immune‐associated remodeling.

5. Immune‐Associated Remodeling of the Tumor Microenvironment

In the gastrointestinal tumor microenvironment, tumor‐derived metabolic and stress signals can reshape the functional states of innate immune cells, including macrophages, myeloid populations, and NK cells, thereby influencing immune surveillance and therapy response [62]. Therefore, tsRNAs that regulate translational and metabolic states in tumor cells may also indirectly affect macrophage polarization, NK‐cell function, and the recruitment of suppressive myeloid cells.

Zhou et al. found in HCC that tsr_019759 is upregulated, suppresses TNFSF15, and activates JAK2/STAT3 signaling, thereby promoting tumorigenesis through crosstalk with M2 tumor‐associated macrophages. The same study also showed that cell‐membrane‐modified polymer nanoparticles targeting this fragment could reverse this effect [15]. Also in HCC, radiotherapy combined with a self‐gelling powder encapsulating a tRF5‐GlyGCC inhibitor enhanced NK‐cell immunity and prevented post‐resection recurrence [16]. In CRC, Zhang et al. found that 3′‐tRF‐Ala‐CGC cleaves the membrane protein MICA, thereby weakening NK‐cell‐mediated cytotoxicity; inhibition of this fragment in vivo reduced tumorigenicity [17]. Overall, current direct immune evidence is mainly concentrated in macrophage‐ and NK‐cell‐related models, whereas the effects of tsRNAs on other immune‐cell populations remain less directly validated.

6. Extracellular tsRNAs and EV‐Associated Communication

The extracellular layer remains less mature than the intracellular and metabolic layers. Evidence for extracellular export of tsRNAs and for intercellular or inter‐organ communication is still limited. Methodologically, the MISEV guidelines issued by ISEV provide basic standards for EV research. These guidelines emphasize that a specific function should not be attributed to EVs solely on the basis of biological activity detected in crude vesicle preparations; relevant studies should also report EV isolation and characterization methods and validate recipient‐cell effects through functional experiments [18]. Therefore, moving from “detection of tsRNAs in EVs” to “demonstration of their function in recipient cells” still requires several key steps, including selective cargo loading, EV heterogeneity, recipient‐cell uptake and validation of downstream functional effects [13].

Among current studies on tsRNAs in digestive system tumors, the PDAC–liver pre‐metastatic niche axis represents the clearest cross‐organ functional example. PDAC‐derived exosomes carrying tRF‐GluCTC‐0005 activate hepatic stellate cells by upregulating WDR1/YAP signaling, promote MDSC infiltration and liver pre‐metastatic niche formation, and ultimately drive early liver metastasis [19]. This study links EV cargo, recipient hepatic stellate cells, downstream immune events, and an in vivo metastatic outcome, and therefore comes closer to functional communication evidence than simple EV cargo detection.

Local EV‐mediated communication has also been reported within the pancreatic tumor niche. Cao et al. found that pancreatic stellate cell‐derived exosomal tRF‐19‐PNR8YPJZ promotes pancreatic cancer (PC) cell proliferation and motility through AXIN2 [63]. Other reported intra‐niche communication routes involve immune‐cell remodeling: CRC plasma‐exosomal tRF‐3022b modulates M2 macrophage polarization by binding cytokines [64] whereas HCC migrasome‐borne tsRNA‐10,105 has also been associated with M2 polarization, although this route is currently supported by fewer confirmations [65].

By contrast, most remaining EV studies mainly show that tsRNAs are vesicle‐associated and disease‐discriminating. Examples include EV‐associated tRF‐3004a [66], tsRNA‐Gly‐5‐0007 [67], GC EV‐sncRNA signatures [68] and a canine HCC‐versus‐adenoma tRNA‐Val study [69]. These studies have biomarker value, but they do not yet demonstrate dose‐resolved transfer, recipient‐cell uptake, or phenotypic causation. They are therefore better interpreted as circulating EV‐associated biomarkers than as evidence of functional systemic signaling. Because EV isolation methods strongly influence the recovered small‐RNA profile [18, 70], EV‐tsRNA inventories are also intrinsically method‐dependent.

Beyond tumor‐ and stromal‐cell‐derived EVs, microbiota‐derived extracellular small RNAs may represent an additional communication layer of potential relevance to digestive system tumors. Bacteria also generate tRNA‐derived fragments, and microbial EVs can package these fragments and deliver them to eukaryotic host cells, where they may regulate host gene expression and immune responses [71]. Recent evidence further showed that gut commensal Klebsiella pneumoniae ‐derived EVs (KpEVs) are enriched in bacterial tsRNAs that are markedly elevated in the serum of patients with HCC and suppress macrophage‐derived nitric oxide production; the KpEVs themselves promote an M2‐like macrophage phenotype and facilitate gut‐to‐liver bacterial translocation [72]. Although these findings do not yet establish a direct causal role for bacterial tsRNAs in digestive tumor initiation or progression, they extend the conceptual framework of tsRNA‐mediated communication from tumor–stroma exchange to host–microbe and gut–liver signaling.

Taken together, the current extracellular tsRNA literature in digestive system tumors can be broadly organized into three evidence tiers: functional transfer models, EV‐associated biomarker associations, and an emerging host–microbe communication axis. The PDAC–liver pre‐metastatic niche axis remains the clearest tumor‐derived cross‐organ functional example, whereas most EV‐associated tsRNA studies still require recipient‐cell functional validation. Microbiota‐derived tsRNAs further extend this framework toward gut–liver signaling, but their cancer‐specific causal roles remain to be established.

The mechanisms and biological significance of tsRNA‐mediated intercellular communication, together with the boundary between functionally supported transfer and EV‐associated biomarker signals, are summarized in Figure 2, and the individual reported communication events are catalogued in Table 2.

FIGURE 2.

Mechanisms and biological significance of tsRNA‐mediated intercellular communication in digestive cancers. A generic communication route (left)—donor‐cell packaging of tsRNAs into extracellular vesicles (EVs) or migrasomes, transit, recipient‐cell uptake, reprogramming, and biological outcome—is instantiated by three representative routes. (i) Immune remodeling: Tumor‐derived EV/migrasome tsRNAs (CRC EV tRF‐3022b; HCC migrasome tsRNA‐10,105) drive macrophage M2 polarization and an immunosuppressive tumor microenvironment. (ii) Cross‐organ metastasis: PDAC‐derived EV tRF‐GluCTC‐0005 acts on hepatic stellate cells via WDR1/YAP to promote MDSC recruitment and a hepatic pre‐metastatic niche. (iii) Stromal‐to‐tumor communication: Pancreatic stellate‐cell EV tRF‐19‐PNR8YPJZ acts on cancer cells via AXIN2/Wnt to promote tumor progression. The bottom box separates functionally supported transfer (donor cargo → recipient uptake or response → recipient phenotype) from cargo/readout‐only EV‐tsRNA associations (e.g., circulating tRF‐3004a and tsRNA‐Gly‐5‐0007, for which recipient function has not been demonstrated) and the emerging host–microbe axis ( Klebsiella pneumoniae EV‐borne bacterial tsRNAs), for which a causal role in cancer remains unresolved.

TABLE 2.

TsRNA‐mediated intercellular communication in digestive cancers.

Cancer/context	Vesicle and tsRNA	Donor → recipient	Mechanism in recipient	Phenotype/biological significance	Evidence tier	Ref.
PDAC (→ liver)	Exosomal tRF‐GluCTC‐0005	PDAC cell → hepatic stellate cell	↑ WDR1/YAP signaling → MDSC infiltration	Liver pre‐metastatic niche; early liver metastasis	Functional transfer	[19]
Pancreatic cancer	Exosomal tRF‐19‐PNR8YPJZ	Pancreatic stellate cell → PC cell	AXIN2 (Wnt) signaling	↑ proliferation & motility	Functional transfer	[63]
Colorectal cancer	Plasma‐exosomal tRF‐3022b	CRC cell → macrophage	Binds cytokines → M2 polarization; modulates apoptosis	Immunosuppressive microenvironment	Functional transfer	[64]
Hepatocellular carcinoma	Migrasome‐borne tsRNA‐10,105	HCC cell → macrophage	M2 polarization (fewer confirmations)	Immunosuppressive microenvironment	Functional transfer (limited)	[65]
CRC, GC, HCC (multiple GI)	tRF‐3004a · tsRNA‐Gly‐5‐0007 GC EV‐sncRNA signatures tRNA‐Val (canine HCC)	Not established (circulating)	Vesicle‐associated; no demonstrated transfer, uptake or phenotype	Diagnostic/prognostic biomarker	Cargo/biomarker only	[66, 67, 68, 69]
HCC (gut–liver)	Bacterial EV (KpEV) tsRNAs	Gut K. pneumoniae → macrophage (gut → liver)	↓ macrophage NO; ↑ M2 phenotype; gut‐to‐liver translocation	Elevated in HCC serum; hepatic immune modulation/gut–liver crosstalk; cancer causality unresolved	Host–microbe (exploratory)	[71, 72]

Note: Reference numbers correspond to the manuscript reference list. Evidence tiers: Functional transfer, donor cargo linked to recipient‐cell uptake/response and a recipient phenotype; Cargo/biomarker only, vesicle‐associated, disease‐discriminating tsRNAs without demonstrated transfer, uptake or phenotype; Host–microbe (exploratory), microbial EV‐borne bacterial tsRNAs with an as‐yet unresolved causal role in cancer.

Abbreviations: CRC, colorectal cancer; EV, extracellular vesicle; GC, gastric cancer; HCC, hepatocellular carcinoma; KpEV, Klebsiella pneumoniae ‐derived extracellular vesicle; MDSC, myeloid‐derived suppressor cell; NO, nitric oxide; PC, pancreatic cancer; PDAC, pancreatic ductal adenocarcinoma.

7. Clinical Translation: Liquid Biopsy, Methodology and the Road to Validation

The clinical value of tsRNAs is mainly supported by their high abundance and stability in serum, plasma, and exosomes, as well as the technical feasibility of qRT‐PCR‐based detection. These features make them particularly suitable for early, non‐invasive detection of digestive system cancers. Previous studies have reported multiple single tsRNA markers with diagnostic or prognostic relevance in gastric cancer, colorectal cancer, and hepatocellular carcinoma [73, 74, 75, 76, 77, 78, 79, 80, 81]. However, from the perspective of clinical translation, the more informative direction is not the repeated discovery of individual markers but the development of multi‐analyte models, with study designs that explicitly include direct comparisons with established clinical markers and benign‐disease controls.

7.1. Clinical Benchmarking Against Benign Disease and Established Markers

Recent tsRNA biomarker studies with greater clinical informativeness have moved beyond simple “cancer versus healthy control” comparisons in three complementary directions: inclusion of benign or inflammatory disease controls, such as chronic gastritis in gastric cancer and hepatitis or cirrhosis in HCC; assessment of pre‐ and post‐operative dynamics; and evaluation of the incremental value of tsRNA markers over established clinical markers, including CEA, CA19‐9, CA72‐4, AFP, and PIVKA‐II.

In gastric cancer, Gu et al. found that serum tRF‐28‐P4R8YP9LOND5 outperformed CEA and CA19‐9 in distinguishing patients with gastric cancer from both healthy individuals and chronic gastritis controls. When combined with other markers, its discriminative performance for gastric cancer in the setting of chronic gastritis was further improved, and its serum level declined significantly after curative resection [82]. Serum hsa_tsr016141 showed a similar profile, with diagnostic performance superior to traditional markers and a marked post‐operative decline in paired samples [83]. Yuan et al. constructed a three‐tsRNA panel and found that, when combined with CEA, CA19‐9, and CA72‐4, this panel achieved high overall discriminative performance in TNM stage I/II tumors; by contrast, the AUCs of traditional markers at this stage were close to chance level [84]. However, because this study included only healthy controls, it mainly supports the early diagnostic gain of the tsRNA panel and does not fully establish its specificity against benign disease.

In HCC, Jin et al. reported that serum tsRNA‐Thr‐5‐0015 had an AUC of 0.731 when used alone for diagnosis; when combined with AFP and PIVKA‐II, the overall diagnostic performance was substantially improved. This improvement was more pronounced in comparisons with healthy controls and was smaller but still consistent in comparisons with hepatitis controls. The post‐operative level of tsRNA‐Thr‐5‐0015 declined toward that of healthy controls, and this fragment was specifically overexpressed in HCC rather than broadly elevated across other digestive system tumors [85]. Two additional studies further define the clinical evidence boundaries of tsRNA biomarkers from different angles. In HBV‐related HCC, the tRF‐3a‐Pro study incorporated liver cirrhosis controls, thereby better reflecting the clinical need to distinguish HCC from benign liver disease. This marker achieved an AUC of 0.863 when used alone and could be integrated with AFP through a logistic‐regression model to form a combined predictor; however, its molecular target remains experimentally unvalidated [86]. In addition, TGF‐β‐induced CRC EMT‐related tsRNAs, including tRF‐Phe‐GAA‐031 and tRF‐VAL‐TCA‐002, were validated in 68 CRC tumor samples, with ROC AUCs of approximately 0.75 and associations with patient survival. Unlike the preceding studies that incorporated benign disease controls, this study mainly supports the diagnostic and prognostic association of EMT‐related tsRNAs, but does not yet establish their specificity against benign intestinal disease; moreover, their potential targets remain computationally predicted [87]. These clinically oriented studies are summarized in Table 3.

TABLE 3.

Representative circulating or EV‐associated tsRNAs with clinical relevance in digestive system tumors.

Tumor	Candidate tsRNA/panel	Comparator and design	Main clinical signal
GC	hsa_tsr016141 (serum) [83]	GC vs. healthy and gastritis; pre/post‐op	AUC 0.814 vs. healthy; head‐to‐head superior to CEA, CA19‐9; significant post‐op decline
GC	tRF‐28‐P4R8YP9LOND5 (serum) [82]	GC vs. healthy and gastritis; combined with CEA/CA19‐9/CA72‐4	Multi‐marker panel reaches AUC 0.821 vs. healthy and 0.883 vs. gastritis
GC	Three‐tsRNA serum panel [84]	GC vs. healthy only; early‐stage analysis	Six‐marker combination AUC 0.956; outperforms CEA, CA19‐9, CA72‐4 in early TNM I/II
HCC	tsRNA‐Thr‐5‐0015 (serum) [85]	HCC vs. hepatitis and healthy; AFP/PIVKA‐II combination; pre/post‐op	Three‐marker combination AUC 0.913 vs. healthy; post‐op decline; cross‐tumor specific
HCC	tRF‐3a‐Pro (serum) [86]	HCC vs. liver cirrhosis (LC) and healthy; AFP integration	AUC 0.863; logistic‐regression model integrates with AFP
CRC	tRF‐Phe‐GAA‐031/tRF‐VAL‐TCA‐002 (tissue) [87]	CRC tumor vs. adjacent non‐tumor	ROC AUC ~0.75; correlates with metastasis, stage and OS
GC, CRC, HCC	EV‐associated tsRNAs [66, 67, 68, 69]	Cancer vs. control plasma EV	EV cargo disease‐discriminating

Note: Reference numbers correspond to the manuscript reference list.

Abbreviations: AFP, alpha‐fetoprotein; AUC, area under the receiver operating characteristic curve; CA19‐9, carbohydrate antigen 19–9; CA72‐4, carbohydrate antigen 72–4; CEA, carcinoembryonic antigen; CRC, colorectal cancer; EV, extracellular vesicle; GC, gastric cancer; HBV, hepatitis B virus; HCC, hepatocellular carcinoma; LC, liver cirrhosis; OS, overall survival; PIVKA‐II, protein induced by vitamin K absence or antagonist‐II; ROC, receiver operating characteristic; TNM, tumor‐node‐metastasis; tRF, tRNA‐derived fragment; tsRNA, tRNA‐derived small RNA.

A separate methodological point should also be emphasized: abnormal tsRNA expression does not necessarily represent a tumor‐specific signal, because chronic inflammation and non‐malignant liver disease can themselves reshape the background abundance of tsRNAs. In chronic HBV/HCV‐associated liver disease, 5′ tRNA halves are increased in non‐malignant liver tissue and can exceed microRNAs in abundance; in HBV‐associated HCC, their abundance also correlates with angiogenin [31]. Therefore, HCC tsRNA biomarker studies need to include disease controls such as hepatitis and cirrhosis, rather than relying only on healthy controls.

7.2. The Meta‐Analytic Signal and Its Limits

A GC systematic review and meta‐analysis reported that upregulated tsRNAs predicted poor prognosis (HR ≈2.48) and downregulated tsRNAs a favorable one (HR ≈0.55), with diagnostic AUCs of 0.81 versus healthy and 0.74 versus inflammation [88]. Because the analysis pools differently named, oppositely directed fragments as single categories, it is best read as evidence that the direction of association is reproducible across heterogeneous markers rather than as a precise pooled effect size; “upregulated tsRNAs” is not, in this context, a single biological quantity. The signal nevertheless aligns with the directional pattern recovered by single‐marker and panel studies above.

7.3. Barriers to Clinical Translation

Five substantive barriers still limit the clinical translation of tsRNAs. First, nomenclature remains unstable, which compromises the portability of assay design and limits molecule‐level meta‐analysis. Although resources such as MINTbase and MINTmap have made fragment‐level cataloging increasingly tractable, a stable molecule‐level naming convention has not yet been established [4, 89, 90]. Second, sequencing strategies that insufficiently account for RNA modifications can systematically affect the expression differences used to nominate biomarkers. This is not a marginal technical issue: AlkB‐ or PANDORA‐type pretreatments can substantially reshape the recovered tsRNA fragment landscape, whereas low‐bias library strategies such as randomized splint ligation can further reduce the systematic under‐representation of 2′‐O‐methylated fragments [30, 91, 92, 93]. Third, existing clinical cohorts are usually small, single‐center, and retrospective. Future studies should compare tsRNA markers against benign and inflammatory digestive diseases, including chronic gastritis, hepatitis, cirrhosis, and chronic pancreatitis, and should further test whether tsRNA panels provide quantifiable incremental value beyond established markers such as AFP, CEA, and CA19‐9 across disease stages and pre‐/post‐operative time points. Fourth, EV isolation is strongly dependent on pre‐analytical procedures, making EV‐related findings difficult to compare directly across studies; therefore, adherence to MISEV‐type guidelines should be a basic prerequisite for developing vesicle‐targeted clinical assays [13, 18, 70]. Finally, all current therapeutic strategies remain at the preclinical proof‐of‐concept stage, including engineered‐vehicle approaches in HCC [15, 16] and the poly(β‐amino ester) co‐delivery system in CRC [8]. Their clinical translation will require not only safety validation but also pharmacokinetic and biodistribution data appropriate for systemic delivery.

8. Discussion and Perspectives

The digestive‐tumor tsRNA literature now presents an evidence landscape that is uneven in strength but increasingly coherent in logic. Intracellular effector mechanisms currently have the strongest support, including RBP sequestration and translation‐initiation control, AGO2/RISC‐dependent silencing, direct target repression, and ribosome‐associated interactions; several of these mechanisms have been validated in animal models. These findings constitute the firmest evidential ground of the field and represent a key contribution that distinguishes tsRNA biology from conventional small‐RNA research because they include not only miRNA‐like silencing mechanisms but also non‐canonical modes such as EIF4F displacement, FUBP1‐mediated transcriptional activation, and direct engagement of ribosomal‐protein mRNAs.

The link between tsRNAs, metabolic state and therapy resistance is also well supported and carries clear translational potential. At least one combination strategy, inhibition of tsRNA‐GlyGCC together with 5‐FU, has advanced to a deliverable preclinical formulation [8]. By contrast, evidence that tsRNAs directly remodel immune cells has begun to emerge but remains limited to a small number of studies, whereas evidence for EV‐mediated systemic propagation is still largely associative, with only one example approaching a complete cross‐organ functional chain. A pan‐cancer analysis further suggests that tRF expression can be regulated by heritable genetic variation and may contribute to cancer risk, adding a germline dimension to tsRNA research; however, this direction remains markedly underexplored in digestive system tumors specifically [94].

To secure the outer and less firmly established parts of this evidence landscape, three priorities are likely to be most decisive. First, modification‐aware sequencing and a stable molecule‐level nomenclature should move from specialist options to field‐wide defaults [4, 30, 89, 90, 92]. Second, functional claims about EV‐mediated delivery and immune regulation should meet recipient‐cell‐level evidential criteria, including dose‐resolved transfer, recipient‐cell uptake and loss‐of‐function validation in the recipient cell or recipient compartment, rather than being inferred from EV cargo characterization alone [13, 18]. Third, biomarker studies should be designed against benign and inflammatory disease comparators and should test the incremental value of tsRNA markers over established markers such as AFP, CEA, CA19‐9 and PIVKA‐II, rather than relying only on comparisons with healthy controls [31, 82, 83, 84]. Beyond these priorities, whether microbiota‐derived tsRNAs act causally in digestive tumors—rather than merely marking gut–liver inflammatory states—remains an open question that will require direct perturbation in tumor models rather than serum association alone.

For the field, credibility will depend less on adding further correlative studies than on maintaining several key distinctions: circulating detectability is not equivalent to functional signaling; immune association is not equivalent to immune causality; and the presence of a cargo in EVs is not equivalent to effective delivery and functional activity in recipient cells.

In summary, tsRNA research in digestive system tumors has established a relatively stable intracellular mechanistic core, together with a well‐supported metabolic and therapy‐resistance dimension. Immune‐associated effects are beginning to emerge but remain at an early evidential stage, whereas cross‐organ propagation still depends largely on a small number of studies approaching a complete functional chain. Future progress will depend less on cataloging additional dysregulated fragments than on applying the methodological and validation standards needed to make the more ambitious functional claims genuinely testable. The graphical table of contents further summarizes the overall scope of this review, spanning intracellular regulation, immune‐associated remodeling, EV/migrasome‐mediated communication and systemic host–microbe or gut–liver signaling (Figure 3).

FIGURE 3.

Graphical table of contents. The graphical table of contents summarizes how tRNA‐derived small RNAs in digestive cancers connect intracellular regulation, immune remodeling, extracellular‐vesicle/migrasome‐mediated transfer and systemic host–microbe or gut–liver communication.

Author Contributions

Wang Xitan: writing – original draft, investigation. Li Han: conceptualization, writing – review and editing.

Funding

The authors have nothing to report.

Ethics Statement

The authors have nothing to report.

Consent

The authors have nothing to report.

Conflicts of Interest

The authors declare no conflicts of interest.

Data Availability Statement

The authors have nothing to report.