Transfer RNA-derived fragments (tRFs) are dysregulated in mTLE

To assess brain tRF signatures in mTLE, we performed sncRNA-seq of hippocampal and adjacent neocortical tissue in control and mTLE non-HS samples (Fig. 1a). For data analysis, we first validated our bioinformatics pipeline using a publicly available dataset. Our analysis pipeline robustly validated previous results identifying the most differentially expressed (DE) candidate tRFs of the study (Gly-GCC (Supplementary Fig. 1a-c), Gly-CCC (Supplementary Fig. 1d-f) and Glu-CTC (Supplementary Fig. 1 g-i))(Supplementary Table S6; ([29]). Detailed comparative analysis of sample groups (Supplementary Fig. 2a-e) or of fresh frozen and postmortem-treated cortical tissue (Supplementary Fig. 3a-e) did not reveal differences in alignment percentage, match type to tRNA, sequence origin or tRF length. Sample separation was best for mTLE non-HS and hippocampal tissues (cytoplasm: Fig. 1b; nucleus: Supplementary Fig. 4a-f). Therefore, further analyses were performed using hippocampal samples.

Fig. 1

Transfer RNA-derived fragment (tRFs) expression is deregulated in the cortex and hippocampus of mTLE patients. a Schematic overview of the tissue fractionation performed on human cortical and hippocampal mTLE tissue. Created in BioRender. Pasterkamp, J. (2025) (https://BioRender.com/1gha316). b Principal component analysis (PCA) of the indicated groups based on cytoplasm tRF transcriptome changes in hippocampal tissue. c Heatmap representation of DE tRFs between controls and mTLE non-HS patients in the cytoplasm fraction of hippocampal samples. Of the detected tRFs 3% (120/3913) were DE in mTLE non-HS patients as compared to controls. d 5'tRF-His-GTG, 5'half-His-GTG and 5'half-Lys-CTT relative expression in controls and mTLE non-HS patient hippocampus, normalized to 5S. Each dot represents one individual and bars show relative mean expression ± SD (Mann–Whitney test; *P ≤ 0.05; N = 7 individuals per group). e Regulation profile of 5'tRF-His-GTG and 5'half-His-GTG in mTLE non-HS and mTLE + HS patients, normalized to 5S. Individual symbols depict relative mean fold-changes ± SD (Two-way ANOVA with Uncorrected Fisher’s LSD; ****P < 0.0001; N = 7–8 individuals per group)

Unsupervised hierarchical clustering of cytoplasmic (Fig. 1c) and nuclear (Supplementary Fig. 4b) samples revealed a large set of differentially expressed (DE) tRFs. Altogether, 134 DE (FDR < 0.05) tRFs were detected in the nucleus and 120 (FDR < 0.05) in the cytoplasm. To select candidates for further analysis, the following criteria were used: 1) an expression threshold of > 1000 counts, 2) FC ≥ 2 with FDR < 0.05, and 3) 100% alignment to the mature or pre-tRNA sequence. Since sample clustering was more robust for cytoplasmic (Fig. 1c) as compared with nuclear samples (Supplementary Fig. 4b), candidate tRFs were selected from cytoplasm using the abovementioned criteria. This revealed that tRF fragments cleaved from tRNA-His-GTG were most strongly deregulated. By visualizing tRF alignment onto precursor mature tRNA (n = 1 for control and mTLE non-HS), tRNA-His-GTG fragments (downregulated) were detected that originated from the 5’end of the tRNA (Supplementary Fig. 5a-b). An independent set of hippocampal patient samples was used to validate the downregulation of 5'tRF-/5'half-His-GTG with RT-qPCR. This confirmed decreased expression of 5'tRF-His-GTG (0.18-fold, U7,7 = 6.5, p = 0.0192) and 5'half-His-GTG (0.21-fold, U7,7 = 7, p = 0.0245) in mTLE non-HS hippocampus (Fig. 1d). To further validate this result, the analysis was replicated with Taqman RT-qPCR chemistry which showed downregulation of 5’half-His-GTG (Spearman r = 0.94, p < 0.0001; Supplementary Fig. 5c). Further, a tRF that did not show changes in the DESeq analysis, 5’half-Lys-CTT, was selected for qPCR analysis to further test our data-analysis pipeline. Expression of this fragment was not altered in our validation samples (0.65-fold, U7,7 = 18, p = 0.4347) (Fig. 1d). Interestingly, when mTLE subtypes were compared, a more robust downregulation of 5'tRF-His-GTG (0.05-fold, t7,8 = 39.02; DF = 26, p < 0.0001) and 5'half-His-GTG (0.04-fold, t7,8 = 37.04; DF = 26, p < 0.0001) was detected in mTLE + HS as compared to mTLE non-HS patient samples (Fig. 1e). This strong decrease observed in mTLE + HS tissue could be attributed to the significant neuronal cell loss that characterizes the hippocampus of human mTLE + HS patients [7, 22]. In conclusion, our bioinformatics pipeline identified previously unexplored tRF expression changes in mTLE brain tissue, including a downregulation of 5’tRNA-His-GTG fragments.

Altered (pre)tRNA expression in the mTLE hippocampus

The differences in tRF expression in mTLE hippocampus prompted us to assess whether levels of corresponding mature tRNAs were also altered. The overall level of each tRNA was determined based on the total number of aligned reads, i.e. all detected tRFs. Our bioinformatics analysis revealed downregulation of tRNA-His-GTG (Fig. 2a) and RT-qPCR on independent samples confirmed this result (0.36-fold, U7,7 = 8, p = 0.0379; Fig. 2b). As a control tRNA-Gln-GTC, unchanged in the RNA-seq data, was included and was found to be unaltered (0.47-fold, U7,7 = 15.5, p = 0.2721; Fig. 2b). We also assessed the expression and distribution of pre-tRNAs by calculating a nucleus/cytoplasm ratio using a mTLE total RNA-seq dataset. This showed compartment-specific enrichment for select pre-tRNAs. For example, pre-tRNA-His was predicted to be more abundant in the cytoplasm of mTLE non-HS patients as compared to controls (Fig. 2c). Count data from the pseudoalignment analysis confirmed this result by showing increased pre-tRNA-His levels in cytoplasmic samples (2.09-fold, U5,5 = 0, p = 0.0079; Fig. 2d). Based on highest expression level and largest difference between groups in the pseudoalignment analysis, three pre-tRNA-His-GTG genes were selected for RT-qPCR. All three genes (pre-His-GTG-1, pre-His-GTG-6, pre-His-GTG-8) were found to be upregulated (6.33-fold, U7,7 = 7.5, p = 0.0256; 7.83-fold, U7,7 = 5.5, p = 0.0105; 21.18-fold, U7,7 = 0, p = 0.0006; respectively) in mTLE non-HS samples (Fig. 2e). Comparison of mTLE + HS and mTLE non-HS data did not reveal differences for pre-tRNA-His-GTG (0.80-fold, t7,8 = 0.08; DF = 26, p = 0.9338) and tRNA-His-GTG (0.48-fold, t7,8 = 0.28; DF = 26, p = 0.7817) (Fig. 2f). Together these data show altered brain expression of tRNAs and pre-tRNAs in mTLE. Interestingly, while levels of mature tRNA-His-GTG were decreased, expression of pre-His-GTG, originating from three different genomic loci, was increased.

Fig. 2

Dysregulation of mature tRNA and pre-tRNA in the hippocampus of mTLE patients. a A heatmap representing cumulative values of small-RNA reads aligned to each transfer RNA (tRNA) in hippocampus samples. b tRNA-His-GTG relative expression in controls and mTLE non-HS patient samples, normalized to 5S. tRNA-Gln-GTC was used as a negative control. Each dot represents one individual and bars show mean expression ± SD (Mann–Whitney test; *P ≤ 0.05; ns; N = 7 individuals per group). c Shifts in pre-tRNA nucleus/cytoplasm ratios as well as changes in pre-tRNA quantities were observed when controls (black) were compared to mTLE non-HS samples (purple). d Heatmap representation of pre-tRNA fold-changes in the nucleus and cytoplasm samples from hippocampal tissue of mTLE non-HS patients (Mann–Whitney test; *P ≤ 0.05; N = 5). e Pre-tRNA-His-GTG relative expression in controls and mTLE non-HS patients, normalized to 5S. Custom primers were designed against the three different genes (pre-His-GTG-1-1; -1-6 and -1-8). Each dot represents one individual and bars show relative mean expression ± SD (UMann-Whitney test; **P ≤ 0.01; *P ≤ 0.05; N = 7 individuals per group). f Regulation profile of pre-tRNA-His-GTG and tRNA-His-GTG in mTLE non-HS and mTLE + HS patient samples, normalized to 5S. Individual symbols depict relative mean fold-changes ± SD (N = 7–8 individuals per group)

Because these data hinted at a disruption of tRNA-His-GTG biogenesis (Supplementary Fig. 6a), expression of genes with reported roles in tRNA biogenesis was analyzed in total RNA-seq data. This revealed no negative or positive enrichment of genes related to tRNA processing or modification (Supplementary Fig. 6b). Five canonical enzymes involved in tRNA biogenesis were assessed in more detail but no changes were observed in their expression (Drosha Ribonuclease III (DROSHA; 0.96-fold, U7,7 = 23, p = 0.9015), DICER (0.96-fold, U7,7 = 23, p = 0.8747), Ribonuclease/Angiogenin Inhibitor 1 (RNH1; 1.05-fold, U7,7 = 22, p = 0.8048), Ribonuclease Z (RNase Z; 1.21-fold, U7,7 = 11, p = 0.0973) and ANG (0.88-fold, U7,7 = 10, p = 0.0728)) (Supplementary Fig. 6c). Also at the protein level, no changes were observed for DROSHA (0.95-fold, U6,6 = 18, p > 0.9999), DICER (1.50-fold, U6,6 = 6, p = 0.0649) or ANG (1.49-fold, U6,6 = 14, p = 0.5887) (Supplementary Fig. 6d). This suggests that in the chronic epileptic hippocampus, disrupted tRNA-His-GTG biogenesis may not be caused by changes in the expression of canonical enzymes.

Activity-dependent and pro-inflammatory changes in 5’tRNA-His-GTG-derived fragments

To further study the role of tRFs in mTLE, we focused on tRNA-His-GTG and its 5’ fragments. These RNA molecules showed a robustly dysregulated expression profile and detectable expression levels, in the absence of alignment errors of sequencing reads. To begin to dissect the neuronal function of tRNA-His-GTG, expression of tRNA-/5'tRF-/5'half-His-GTG was examined during mouse hippocampal development (E18-P365). Hippocampal expression of tRNA/5'tRF-/5'half-His-GTG was found from E18 onwards (Fig. 3a). In primary hippocampal (PHN) cultures expression increased as cultures matured (DIV1-28) (Fig. 3b). As the hippocampus is composed of different cell types in addition to neurons, Magnetic-Activated Cell Sorting (MACS) was used to obtain astrocytes and microglia from mouse hippocampus tissue. Expression of tRNA-/5'tRF-/5'half-His-GTG was found in astrocytes and microglia (Fig. 3c).

Fig. 3

5'tRF-/5'half-His-GTG expression analysis. a Relative expression (log10 transformed) during hippocampal development, normalized to 5S. Bar plots represent mean ± SD, normalized to E18. N = 3 animals/time-point; E, embryonic day; P, postnatal day. b Relative expression (log10 transformed) in vitro in murine hippocampal neurons (PHN), normalized to 5S. Bar plots represent mean ± SD, normalized to day in vitro (DIV)1. Each dot depicts one experiment (N = 1–3 independent cultures). c Relative expression in MACS-isolated astrocytes and microglia from adult mouse hippocampus, normalized to 5S. Bar plots represent mean ± SD. N = 3 independent experiments. d Schematic overview showing the culturing method and protocol used for different activity-based assays. Created in BioRender. Pasterkamp, J. (2025) (https://BioRender.com/5k5x3r4). e tRNA-/5'tRF-/5'half-His-GTG and 5'half-Lys-CTT fold-change upon neuronal depolarization of mouse PHN with KCl, normalized to 5S. Bar plots represent mean ± SD, normalized to control condition (saline) (One-way ANOVA with Dunnett’s multiple comparison test; **P ≤ 0.01; *P ≤ 0.05; ns, P > 0.05; N = 5 independent experiments). f tRNA-/5'tRF-/5'half-His-GTG intracellular fold-change upon Mg2+ depletion treatment of mouse PHN, normalized to 5S. Bar plots represent mean ± SD, normalized to control condition (Mg2+) (Unpaired t-test; **P ≤ 0.01; *, P ≤ 0.05; ns, P > 0.05; N = 3 independent experiments). g Schematic overview of the glial cell culture preparation method and LPS treatment. Created in BioRender. Pasterkamp, J. (2025) (https://BioRender.com/f7ge5zo). h 5'tRF-/5'half-His-GTG fold-change in MACS-isolated astrocytes and microglia upon LPS-treatment. Bar plots represent mean ± SD, normalized to control condition (Unpaired t test; *P ≤ 0.05; ns, P > 0.05; N = 3 independent experiments)

As epilepsy is characterized by increased neuronal activity, next the effect of enhanced neuronal activity on the tRNA-His-GTG axis was tested. For this, PHN cultures were used given their robust expression of tRNA-His-GTG and 5'tRF-/5'half-His-GTG (Fig. 3b, d). Increased levels of 5'tRF-His-GTG (2.18-fold, q5,5 = 3.138, DF = 12, p.adj = 0.0158) and 5'half-His-GTG (2.50-fold, q5,5 = 3.962, DF = 12, p.adj = 0.0035), but not of tRNA-His-GTG (0.94-fold, q5,5 = 0.2120, DF = 12, p.adj = 0.9677), were observed at 4 h following KCl treatment (Fig. 3e). Levels of 5’half-Lys-CTT were also elevated at 4 h after KCl treatment (2.49-fold, q5,5 = 3.265, DF = 12, p.adj = 0.0125). At 24 h, the increase had declined and only 5’half-Lys-CTT expression was significantly increased (2.21-fold, q5,5 = 2.637, DF = 12, p.adj = 0.0393; Fig. 3e). This suggests that tRF changes following neuronal stimulation may be quick and transient. In line with the effect of KCl, increased 5'tRF-His-GTG (1.58-fold, t3,3 = 3.168, df = 4, p = 0.0339) and a trend towards upregulation of 5'half-His-GTG (1.39-fold, t3,3 = 1.791, df = 4, p = 0.1478) expression were detected in PHNs exposed to Mg2+-depleted media (mg-) (Fig. 3f). Mature tRNA-His-GTG expression was unchanged (0.68-fold, t3,3 = 0.8251, df = 4, p = 0.4557). No changes in ANG or DICER were detected following KCl and Mg2+-depletion treatments (Supplementary Fig. 6e). Finally, given their expression in astrocytes and microglia, the influence of LPS on 5’tRF/5’half-His-GTG expression was tested. LPS induced a mild decrease in 5’half-His-GTG expression in mouse microglia, but not astrocytes (0.73-fold, t3,3 = 3.159, df = 4, p = 0.0342; Fig. 3g, h).

Together, these data show tRNA/tRF- and cell type-specific regulation by acute seizure-relevant neuronal stimuli and pro-inflammatory signals in vitro.

mTLE-associated changes in tRNA-His-GTG affect neuronal morphology and gene expression

To unravel the functional role of tRNA-His-GTG in neurons, antisense LNA gapmeR technology was used. The His-GTG gapmeR caused knockdown of tRNA-/5'tRF-/5'half-His-GTG in HEK293T and Neuro-2a cell lines, in comparison to control (NC) (Supplementary Fig. 7a-c). As a positive control, MALAT1 knockdown was performed in both cell lines (Supplementary Fig. 7a, 7c). Analogous to hippocampal mTLE tissue (Fig. 2e), increased pre-tRNA-His-GTG expression (pre-His-GTG-1, pre-His-GTG-6, pre-His-GTG-8) was observed in HEK293T cells (Supplementary Fig. 7b). Thus, the gapmeR induced expression changes in the tRNA-His-GTG axis in vitro that resembled those observed in mTLE hippocampus in vivo. Next, His-GTG gapmeR and a pCK-GFP vector were co-transfected in DIV1 PHN. At DIV4, PHN transfected with His-GTG gapmeR showed increased cell body size (245.0 ± 111.5 as compared to 215.6 ± 90.3 in NC; U113,106 = 4954, p = 0.0272) and decreased neurite length (78.4 ± 51.7 as compared to 93.1 ± 61.4 in NC; U113,105 = 5015, p = 0.0486) in comparison to NC (Supplementary Fig. 7d, e). Together, these data for the first time show that downregulation of tRNA-His-GTG and its 5’ fragments induce modest but significant changes in neuronal morphology.

To investigate the molecular consequences of knockdown of tRNA-His-GTG and its fragments, total RNA-seq was performed on gapmeR-transfected Neuro-2a cells (Fig. 4a). For this study, a knockdown threshold of > 35% was applied during sample selection. PCA showed clear separation between NC and His-GTG gapmeR samples (Fig. 4b). RNA-seq revealed 441 DE transcripts (Fig. 4c, Supplementary Table S7), of which many were upregulated (251 versus 190 downregulated transcripts). The most representative GO clusters were related to “RNA processing”, “DNA and chromatin” and “Translation and protein” processes (Fig. 4d, Supplementary Fig. 8a, b; Supplementary Table S8, S9). In addition, using GSEA analysis, we observed that the transcripts encoding for Histidine repeat proteins were enriched towards downregulation following His-GTG gapmeR treatment (ES = -0.30, FDR = 0.0290; Supplementary Fig. 8c).

Fig. 4

tRNA/5’tRF/5’half-His-GTG depletion regulates neuronal gene expression. a Schematic overview of the workflow used for knockdown and RNA sequencing. Created in BioRender. Pasterkamp, J. (2025) (https://BioRender.com/c28fahe). b PCA of the two experimental groups (His-GTG and negative control (NC) gapmeR) based on the transcriptome changes in Neuro-2a (N2A; N = 4 samples/group). c Heatmap representation of the 441 DE transcripts found between the His-GTG and NC gapmeR groups. d Enrichment map representation of GO biological processes associated with the transcriptome changes between the His-GTG and NC gapmeR groups. Different clusters represent a group of related GO biological processes. The size of each node is proportional to the number of genes enriched in a certain process (FDR < 0.05)

Overall, these results show that regulation of the tRNA-His-GTG axis as observed in human mTLE influences gene and protein expression regulation in vitro.

5’tRNA-His-GTG fragments regulate gene expression

tRFs have been proposed to regulate gene-expression in a miRNA-like fashion[42]. Therefore, to select specific candidates that may act downstream of 5’tRNA-His-GTG fragments in mTLE, miRNA-like targets were predicted for these molecules (Fig. 5a, b). 5'tRF-His-GTG targets were positively enriched when GSEA was used with our His-GTG gapmeR RNA-seq dataset (ES = 0.27, FDR = 0.0010; Fig. 5a). However, no significant enrichment was found for the predicted targets of 5’half-His-GTG in these data (ES = 0.25, FDR = 0.6070; Fig. 5b). To assess possible miRNA-like functions of tRFs, bioinformatics analysis of sncRNA sequenced AGO-IP samples was performed (HEK293 cells; Fig. 5c). Total RNA samples had a relatively high number of detected tRFs as compared to AGO-IPs with two different antibodies. However, 151 tRFs were also present in immunoprecipitated samples. Importantly, both 5’tRF-His-GTG and 5’half-His-GTG were detected in AGO-IP samples (Fig. 5d). To functionally validate these datasets, AGO2 immunoprecipitation (IP) was performed from Neuro-2a cell lysates. A mild enrichment of 5’tRF-His-GTG was found in AGO2-IP in comparison to IgG-IP, confirming the interaction of 5’tRF-His-GTG with AGO2 (1.74-fold; Fig. 5e). To assess if this functional enrichment of 5’tRF-His-GTG and 5’half-His-GTG on Ago2 is present in vivo, we analyzed publicly available Ago2-seq smRNAseq data for tRFs, from controls and intra-amygdala kainate mouse model at different timepoints post-kainate injections [80]. Similar to the enrichment on Ago2 in vitro (Fig. 5d), a moderate enrichment of 5’tRF-His-GTG and 5’half-His-GTG was also observed in vivo in adult hippocampal samples of control or kainate-treated mice different timepoints. This further confirms the association of these fragments with the Ago2-associated RNA-inducing silencing complex (RISC) (Fig. 5f). Next, qPCR was used to assess selected top predicted 5’tRF-His-GTG targets based on rank score[58]. Several of these targets were enriched (4/6) after gapmeR treatment. These included Slc12a8 (1.64-fold, t7,7 = 4.943, DF = 84, p.adj < 0.0001), Stau2 (1.52-fold, t7,7 = 3.947, DF = 84, p = 0.0011), Ankrd17 (1.48-fold, t7,7 = 3.483, DF = 84, p = 0.0055) and Ksr2 (1.64-fold, t7,7 = 4.920, DF = 84, p < 0.0001) (Fig. 5g). These data indicate that 5’tRF-His-GTG is present in a complex with AGO2 and some of its predicted targets are upregulated after gapmeR treatment.

Fig. 5

Identification of 5'tRF/5’half-His-GTG targets. a Gene Set Enrichment Analysis (GSEA) for predicted positive enrichment of 5’tRF-His-GTG targets (ES = 0.27; FDR < 0.01) after His-GTG gapmeR treatment as compared to negative control. b GSEA for predicted 5’half-His-GTG targets. No enrichment was observed. c A venn diagram showing the overlap between tRFs detected by total RNA sequencing and in AGO immunoprecipitation (AGO-IP) samples with two different antibodies in HEK293T. From all tRFs present in HEK293T cells, 7% (151/2101) were present in samples immunoprecipitated with both AGO antibodies. d 5'tRF-/5'half-His-GTG enrichment in Abcam and Diagenode AGO-IP samples (CPM, counts per million). e 5'tRF-/5'half-His-GTG enrichment in AGO2-IP in Neuro-2a (N2A). Enrichment was calculated as the fold-change between AGO2-IP and IgG-IP, after normalization with 5S. Bar plots represent fold-change relative to control condition (IgG) ± SD. N = 3 independent experiments. f 5'tRF-/5'half-His-GTG enrichment on Ago2-IP samples in hippocampal samples of control and intra-amygdala kainite mice. Each dot represents the detected average total read counts at different timepoints. Bar plots represent mean ± SD. g Bar plots represent the fold changes between 5'tRF-His-GTG targets in gapmeR- and negative control (NC)-transfected N2A cells. Bar plots represent mean ± SD, Two-way ANOVA with Fisher’s LSD test. ****P ≤ 0.0001; ***P ≤ 0.001; **P ≤ 0.01; N = 7 independent transfections

Next, we wanted to confirm the upregulation of the selected targets using a different knockdown strategy (Fig. 6). Inhibitors were designed that directly target 5’tRNA-His-GTG fragments in the cytoplasm, in contrast to the His-GTG gapmeR that targets both mature tRNA and fragments in the nucleus. The inhibitor induced strong downregulation of 5'tRF-His-GTG in human HEK (0.12-fold, t6,6 = 3.902, df = 10, q = 0.0030) and SH-SY5Y cells (0.53-fold, t6,6 = 4.893, df = 10, q = 0.0019), and mouse Neuro-2a cells (0.43-fold, t6,6 = 3.958, df = 10, q = 0.0030) (Fig. 6a). Similarly, 5’half-His-GTG was downregulated in human and mouse cell lines (HEK 0.19-fold, t6,6 = 3.378, df = 10, q = 0.0071; SH-SY5Y cells 0.58-fold, t6,6 =  = 3.572, df = 10, q = 0.0071; N2a 0.50-fold, t6,6 =  = 3.687, df = 10, q = 0.0071) (Fig. 6a). Mature tRNA-His-GTG was either strongly upregulated (HEK 2.83-fold, t6,6 = 3.723, df = 10, q = 0.0002; N2a 2.67-fold, t6,6 = 5.735, df = 10, q = 0.0002) or unaltered (SH-SY5Y cells onefold, t6,6 = 0.039, df = 10, q = 0.3265) (Fig. 6a). Expression of pre-tRNA-His-GTG-1 (7.14-fold, t6,6 = 3.314, df = 10, q = 0.0079) and pre-tRNA-His-GTG-8 (34.12-fold, t6,6 = 4.664, df = 10, q = 0.0018) was increased (HEK 3.14-fold, t6,6 = 3.611, df = 10, q = 0.0024; SH-SY5Y 32.73-fold, t6,6 =  = 14.700, df = 10, q = 0.0001) (Fig. 6b). Thus, the inhibitor successfully targeted 5’tRNA-His-GTG fragments and induced upregulation of mature and pre-tRNA-His-GTG.

Fig. 6

Targeting of 5'tRNA-His-GTG fragments in the cytoplasm using LNA inhibitors causes gene expression changes. a Bar plots represent the fold changes of 5'tRF-His-GTG, 5'half-His-GTG and tRNA-His-GTG in inhibitor- and negative control-transfected Neuro-2a (N2A), SH-SY5Y and HEK293T cells. Bar plots represent mean ± SD. Multiple t tests. *P < 0.05. N = 6 independent transfections. b Bar plots represent the fold changes of the different pre-His-GTG genes in inhibitor- and negative control-transfected SH-SY5Y and HEK293T cells. Bar plots represent mean ± SD. Multiple t tests. *P < 0.05. N = 6 independent transfections. c Bar plots represent the fold changes of 5'tRF-His-GTG targets in inhibitor- and negative control-transfected N2A cells, SH-SY5Y and HEK293T cells. Bar plots represent mean ± SD. Multiple t tests. *P < 0.05. N = 6 independent transfections

Analysis of the selected targets showed upregulation of 6/7 candidates following addition of the inhibitor to Neuro-2a cells (Cnr1 (1.46-fold, t6,6 = 6.281, df = 10, q = 0.0002), Parva (1.30-fold, t6,6 = 2.645, df = 10, q = 0.0237), Slc12a8 (2.18-fold, t6,6 = 3.253, df = 10, q = 0.01212), Ankrd17 (6.84-fold, t6,6 = 11.590, df = 10, q = 0.0001), Cmtm4 (1.71-fold, t6,6 = 3.100, df = 10, q = 0.01412), and Ksr2 (1.20-fold, t6,6 = 2.564, df = 10, q = 0.0237)) (Fig. 6c). Five of these genes were validated in human SH-SY5Y cells (CNR1 (2.22-fold, t6,6 = 2.215, df = 10,q = 0.0073), SLC12A8 (4.94-fold, t6,6 = 12.600, df = 10, q = 0.0001), ANKRD17 (1.31-fold, t6,6 = 1.31, df = 10, q = 0.00005), CMTM4 (1.57-fold, t6,6 = 5.150, df = 10, q = 0.0001), and KSR2 (0.83-fold, t6,6 = 3.474, df = 10, q = 0.0015)) (Fig. 6c). Finally, two targets were confirmed in HEK293T cells (CNR1 (3.04-fold, t6,6 = 3.783, df = 10, q = 0.0036) and KSR2 (0.90-fold, t6,6 = 0.660, df = 10, q = 0.2648)) (Fig. 6c).

Together, these data show that knockdown of 5’tRNA-His-GTG fragments leads to an upregulation of mRNAs with predicted seed regions for 5’tRF-His-GTG in their 3’UTR’s.

Reciprocal regulation of 5’tRNA-His-GTG fragments and targets during experimental epileptogenesis

Given the strong downregulation of 5’tRF-His-GTG in human mTLE patients, we next examined the expression of this tRF during epileptogenesis using a clinically relevant animal model. Analysis of an intra-cortical kainic acid (ICK) model[22] revealed upregulation of 5’tRF-His-GTG at 24h post-SE (Fig. 7a; 7.02-fold, t3,3 = 2.919, df = 4, p = 0.0433). While this is apparent contrast to the observed downregulation in human TLE tissue, it is in line with the upregulation observed following enhanced neuronal activation in vitro (Fig. 3e, f). This state of enhanced neuronal activity mimics the high level of seizure activity at this experimental timepoint. A trend towards upregulation was detected for 5’half-His-GTG (6.02-fold, t3,3 = 2.721, df = 4, p = 0.0529; Fig. 7b), while mature tRNA levels were unchanged (1.83-fold at 24h post-SE as compared to control, t3,3 = 2.058, df = 4, p = 0.1087; Fig. 7b). No regulation of 5’tRNA-His-GTG fragments was detected at later timepoints (Fig. 7b). In the same samples, downregulation of the 5’tRF-His-GTG predicted targets Cnr1 (0.38-fold, t3,3 = 5.546, DF = 20, p = 0.0001) and Ksr2 (0.53-fold, t3,3 = 4.262, DF = 20, p = 0.0019) was observed (Fig. 7c). Similarly, analysis of 5’tRF-His-GTG predicted targets in hippocampal mTLE non-HS samples, revealed a significant upregulation of CNR1 (1.35-fold, p = 0.0029) in cytoplasmic cell fractions (Fig. 7d). Further, a trend towards upregulation of CB1 receptor protein (encoded by CNR1) was found in mTLE non-HS samples as compared to controls (1.56-fold, U6,6 = 12, p = 0.3939) (Fig. 7e, f). Further, functional annotation of CNR1 co-expressed genes in the mTLE dataset revealed regulation of proteins involved in epilepsy (category: UP_KW_disease; p = 0.0001) and intellectual disability (category: UP_KW_disease; p = 0.0023), and proteins located in synapse (category: UP_KW_cellular_component; p = 0.0100), integral component of membrane (category: GOTERM_CC; p = 0.0110), glutamatergic synapse (category: GOTERM_CC; p = 0.0120) and neuron projection (category: GOTERM_CC; p = 0.0120) (Supplemental Fig. 9a).

Fig. 7

Analysis of 5'tRF-His-GTG targets in experimental TLE in vivo. a Schematic overview of the intra-cortical (ICK) mouse model and the different timelines and stages of the development of epilepsy. SE, status epilepticus. b Graphs showing the Log2 fold change of tRNA-/5'tRF-/5'half-His-GTG expression at different timepoints after SE induction. Unpaired t-test. *P ≤ 0.05 N = 3 animals/group. c Bar graphs showing the relative expression changes of 5’tRF-His-GTG targets. Bar plots represent mean ± SD. Two-way ANOVA with Fisher’s LSD test. *P ≤ 0.05; N = 3 animals/group. d Bar plot representing the fold changes of normalized counts from RNA-seq data for KSR2 and CNR1 transcripts in mTLE non-HS hippocampal tissue compared to post-mortem controls. DESeq2; **P ≤ 0.01; N = 5 mTLE non-HS, N = 5 post-mortem controls. e CB1 receptor (CNR1) protein levels in mTLE non-HS and post-mortem controls. Representative images of Western blots (upper panel) and quantification (lower panel) are shown. The number of controls (C) and Epilepsy (E) samples refer to the corresponding control and mTLE non-HS samples (Supplemental Table S1) used in the study. f Bar plots represent mean ± SEM, normalized to β-actin (Mann Whitney test; ns, P > 0.05; N = 6 individuals per group) (C, Control; E, Epilepsy; kDa, kilodalton)

Together, our data reveals opposite expression changes between 5’tRF-His-GTG and select targets in both human mTLE brain tissue and at acute stages of an experimental TLE model. Interestingly, CNR1 showed robust and consistent reciprocal regulation and is closely connected to several epilepsy-related and synaptic proteins in mTLE.

Inhibition of 5’tRNA-His-GTG fragments causes aberrant brain activity and pathology

To investigate the contribution of 5’tRNA-His-GTG fragments during the acute phase of TLE pathogenesis, we targeted the increase in 5’tRF-His-GTG observed at 24 h post-SE in mice using inhibitors (Fig. 7b). First, knockdown was validated in wild-type mice. Upon targeting 5’tRNA-His-GTG fragments by intra-hippocampal injection of inhibitors, a decrease in the expression of 5’tRF-His-GTG (0.55-fold, t4,6 = 3.644, df = 8, p = 0.0066) and 5’half-His-GTG (0.56-fold, t4,6 = 3.091, df = 8, p = 0.0149) was detected. Expression of mature tRNA-His-GTG expression was unchanged (Supplemental Fig. 10a-c). Inhibitor treatment also caused an increase in Ankrd (2.83-fold, t3,3 = 4.100, df = 24, p = 0.0025) and Cnr1 (2.37-fold, t3,3 = 2.934, df = 24, p = 0.0428) expression in mouse hippocampus in vivo. These data further link changes in 5’tRF-His-GTG expression and dysregulation of Cnr1 and other miRNA-like targets (Supplemental Fig. 10d). In the ICK model mice, inhibitors or negative controls were injected at 24 h post-SE induction and telemetric EEG recordings were performed during the first two weeks and later in the chronic period (Fig. 8a). In all mice, the SE period was followed by a similar latent period (mean ± SD (NC: 2.50 ± 2.43); (Inhibitor: 3.50 ± 3.78)) (Fig. 8b). Further, no significant difference in the number of seizures was observed until day14 (D14; Fig. 8c). However, during the chronic stage a consistently higher number of spontaneous seizures were observed on the days of recording (D23, D25, D27 and D29) in mice treated with inhibitors. (Fig. 8d). A significant increase in the total number of spontaneous seizures among the four recording days was observed (difference in means = 7.667 ± 3.308, df = 10, t6,6 = 2.317, p = 0.0430) (Fig. 8e). In line with these data, total time spent in ictal activity was higher in mice injected with inhibitor when data from all four days were combined (difference in means: 302 ± 94.46, t6,6 = 3.197, df = 10, p = 0.0952) (Fig. 8f, G). To further understand the brain activity changes caused by reduced tRF levels, EEG spectral analysis was performed by estimating total EEG power during seizures. Power spectral analysis (PSD) was conducted and the power of major frequency bands delta (0,5 ~ 4 Hz), theta (4 ~ 8 Hz), alpha (8 ~ 12 Hz), beta (12 ~ 30 Hz), gamma (30 ~ 100 Hz) and high ripples that were larger than 100 Hz (100 ~ 249 Hz) were estimated. This revealed a reduction in theta and alpha EEG frequencies [(D23, theta; ANOVA: 0.0018, Cohen’s d: 2.43, alpha; ANOVA: 0.0284, Cohen’s d:1.47)(D25, theta; ANOVA: 0.0021, Cohen’s d: 2.37, alpha; ANOVA: 0.0306, Cohen’s d:1.45)] (Fig. 8h, i; Supplemental Fig. 10e-f), indicating an excitatory/inhibitory imbalance. Finally, cellular changes were evaluated by immunostaining for different markers. As expected, marked hippocampal neuronal loss, granule cell dispersion and glial changes (indicated by strong GFAP and Iba1 staining) were detected in the ipsilateral hippocampus of ICK mice (Supplemental Fig. 11a, 11b). Interestingly, mice injected with inhibitor displayed an increase in GFAP staining compared to controls (Fig. 8j) (5.081 ± 2.594, t9,9 = 1.959, df = 16, p = 0.0339; Fig. 8k), while Iba1 signals were unchanged (Supplemental Fig. 11c, 11d).

Fig. 8

Knockdown of 5’tRNA-His-GTG fragments exacerbates seizure progression and pathology in vivo. a Schematic showing the injection schedule and timepoints at which EEG and corresponding telemetric video recordings were performed. Created in BioRender. Pasterkamp, J. (2025) (https://BioRender.com/osmpiu1). b Bar graphs showing the difference in latency period after kainate (KA) injections. Bar plots represent mean ± SD. N = 6 negative control (NC)- and N = 8 inhibitor-injected animals. c Line graph showing the average number of seizures from day1 to day14 after (status epilepticus) SE induction. Error bars are SD, N = 6 NC and N = 8 inhibitor-injected animals. d Line graph showing the average number of SRS at day23, day25, day27 and day29 after (status epilepticus) SE induction. Error bars are SD, N = 6 animals per group. e Bar plots showing the average number of seizures on day23, day25, day27 and day29. Bar plots represent mean ± SD. Unpaired t test, *P < 0.05. N = 6 animals per group. f Bar plots show total time spent in ictal activity on day23, day25, day27 and day29. Bar plots represent mean ± SD Unpaired t test, **P < 0.01. N = 6 animals per group. g Representative EEG traces depict spontaneous recurrent seizures (SRS) in negative control (NC) and inhibitor injected animals on day 25. h Line plots of the EEG power spectral analysis showing the average PSD spectrum for a 24 h period between the NC- and inhibitor-injected mice at day 23. N = 6 animals per group. i Boxplots showing the individual bandwise PSD power between the NC and inhibitor injected mice at day 23. ANOVA **P < 0.01, *P < 0.05. N = 6 animals per group. j Representative images of GFAP signal in day30 kainate mice injected with inhibitor or negative control oligos. Scale bar, 20 µm. Kainic acid (KA), negative control (NC), non-injected (Noninj), injected (inj). k Barplot shows the quantification of signal intensity in kainate mice between inhibitor and negative control (NC) injected. Bar plots represent mean ± SD. Unpaired t test, *P < 0.05. N = 3 sections per mice from 3 NC and 3 inhibitor injected animals

Thus, inhibition of (enhanced) 5’tRNA-His-GTG fragments expression during experimental TLE exacerbates changes in seizure frequency, brain activity and gliosis.