Peptide‑based therapeutics targeting the SLC39A14‑PIWIL2 fusion in hepatocellular carcinoma.

Introduction
Molecular and genetic complexity of hepatocellular carcinoma (HCC) impedes early diagnosis and limits treatment efficacy, thereby contributing to poor survival outcomes [1]. Advances in high-throughput genomic sequencing have unraveled the intricate molecular prerequisites of HCC, revealing its molecular heterogeneity with aberrations across various pathways, including inflammation, cell cycle control, invasion, metabolism, and metastasis [2–4]. In addition, chromosomal rearrangements and fusion genes stand out as critical oncogenic drivers across multiple cancer types [5–8]. These fusion genes serve as diagnostic biomarkers, such as ERG and ETV1, while others, such as ALK, RET, BRAF, and FGFR1–4, represent therapeutically targetable genes [9, 10]. In HCC, a limited number of fusion genes have been now identified with potential implications in tumor progression and prognosis. The DNAJB1-PRKACA fusion serves as a driver mutation in fibrolamellar carcinoma, a subtype of HCC [11]. The LINE1-MET fusion has also been highlighted in HCC development [12]. Furthermore, fusions such as SLC45A2-AMACR, ITCH-ASIP, and RNF138-RNF125 have been associated with better HCC prognoses, while MAN2A1-FER, CCNH-C5orf30, and SLC45A2-AMACR, detectable in serum, show potential for HCC diagnosis [13].
In this study, we identified a novel SLC39A14-PIWIL2 fusion gene in HCC patients’ samples. This fusion was formed by chromosomal inversion, joining SLC39A14 exon 1 with PIWIL2 exons 7–23, resulting in the upregulation of PIWIL2 and promoting HCC progression. Additionally, structural modeling of PIWIL2 and SLC39A14-PIWIL2, which encodes a truncated form (tPIWIL2), and their interactions with downstream proteins, alongside the design of PIWIL2 inhibitory peptides, has further highlighted its potential as a therapeutic target in HCC, opening avenues for the development of PIWIL2-focused treatments.

Materials and methods

Identification of fusion transcripts from RNA-seq data of HCC patient samples
The presence of fusion transcripts was identified from our previously published RNA-seq dataset (68 HCC and 10 non-tumor tissues; GSE113617) [14]. By applying three different methods of SOAPfuse [15], ChimeraScan [16], and TopHat-Fusion [17] with default parameters and filtering out transcripts with fewer than 30 reads at the breakpoints, we could identify 14 fusion transcripts which were detected by at least two of the methods.

Human cancer cell lines, anticancer drug, and plasmids
The SNU398 (Seoul, Republic of Korea; Cat# 00398_SNU-398, RRID: CVCL_0077), SNU449 (KCLB Cat# 00449_SNU-449, RRID: CVCL_0454), HepG2 (KCLB Cat# 88065_HepG2, RRID: CVCL_0027), and Huh7 (KCLB Cat# 60104_Huh7, RRID: CVCL_0336) cell lines were purchased from Korean cell line bank and maintained in DMEM, MEM, or RPMI-1640 (Gibco BRL, Grand Island, NY) supplemented with 10% FBS and 1% antibiotics, at 37 °C in a 5% CO₂ incubator. PIWIL2, PIWIL2 without exon 1 (X1-PIWIL2) and SLC39A14-PIWIL2 constructs (tPIWIL2) were cloned into PCDNA3.1 (RRID: Addgene_70219) or C-terminal 3xFLAG-tagged PCDNA3.1 (RRID: Addgene_208616) using the In-Fusion cloning method (Clontech, Mountain View, CA). PCR products were amplified with CloneAmp HiFi PCR Premix (TAKARA, Tokyo, Japan) and specific primers (Table S1), followed by insertion into PCDNA3.1 plasmid using HindIII and ApaI. Constructs were confirmed by Sanger sequencing (Macrogen, Seoul, South Korea). The plasmids were transfected into the liver cancer cell lines (2 × 106 cells per 60-mm dish) using 6 µg of Lipofectamine 3000 (Invitrogen, Thermo Fisher Scientific, Inc.), and incubated for 48 h at 37 °C in a CO₂ incubator. The cells expressing PIWIL2 were treated with various concentrations of 5-FU and peptides for 4 days in media supplemented with 10% FBS.

Immunoblotting (western blotting)
Cells were harvested and lysed using lysis buffer (REF87787, Thermo Fisher Scientific), then centrifuged at ~ 13,000 × g for 10 min at 4 °C. Protein concentrations were determined with a Bradford protein assay kit (#5,000,006, Bio-Rad, Hercules, CA, USA). Equal amounts (30 µg) of protein were separated using 10% SDS-PAGE (Bio-Rad) and transferred to nitrocellulose membranes (#1,620,115, Bio-Rad) for immunoblotting. Membranes were washed three times with PBS (Welgene, Gyeongsangbuk-do, Republic of Korea) containing 0.1% Tween 20 (PBST; Sigma-Aldrich), blocked with PBST containing 1% bovine serum albumin (BSA, Bovogen, Melbourne, Australia) for 1 h at room temperature, and incubated with primary antibodies in PBST with 1% BSA overnight at 4 °C. After washing, membranes were incubated with secondary antibodies (1:1000 dilution) against goat anti-rabbit IgG-HRP (Cell Signaling Technology Cat# 7074S, RRID: AB_2099233) or anti-mouse IgG-HRP (Cell Signaling Technology Cat# 7076S, RRID: AB_330924) for 1 h at room temperature and washed again. Membranes were developed with ECL Buffer (REF34580, Thermo Fisher Scientific) and images captured using an iBright 1500 imaging system (REF34580, Thermo Fisher Scientific). The following antibodies were used: FLAG (1:1000, Sigma-Aldrich, Cat# F1804, RRID: AB_262044), CTNNB1 (1:1000, Santa Cruz Biotechnology, Cat# sc-7963, RRID: AB_626807), c-myc (1:1000, Cell Signaling Technology, Cat# 9402, RRID: AB_2151827), p-Akt (1:1000, Cell Signaling Technology Cat# 4060S, RRID:AB_2315049), p-GSK-3β (1:1000, Cell Signaling Technology, Cat# 9336S, RRID: AB_331405), p-STAT3 (1:1000, Cell Signaling Technology, Cat# 9145S, RRID: AB_2491009), β-actin (1:2000, Santa Cruz Biotechnology, Cat# sc-47778, RRID:AB_626632), and GAPDH (1:5000, Abcam Cat# ab8245, RRID: AB_2107448).

Immunocytochemistry analysis
To examine c-myc localization, PIWIL2-overexpressing liver cancer cells were incubated with or without NEP1 peptide, fixed with 4% paraformaldehyde, and permeabilized with 0.25% Triton X-100. After washing, cells were blocked with 1% BSA and 0.1% Tween 20 for 1 h, then treated with primary antibodies against c-myc (1:100, Cell Signaling Technology, Cat# 9402, RRID: AB_2151827) and PIWIL2 (1:100, Abnova, Cat# MAB0843, RRID: AB_1204794) at 4 °C for 24 h. Following primary antibody incubation, cells were incubated with Alexa Fluor 594-conjugated donkey anti-rabbit IgG (1:200, Molecular Probes Cat# A-21207, RRID: AB_141637) and Alexa Fluor 488-conjugated donkey anti-mouse IgG (1:200, Molecular Probes Cat# A-21202, RRID: AB_141607) secondary antibodies for 2 h at room temperature. Nuclei were stained with DAPI-containing mounting solution, and cells were visualized under an Axiovert 200 fluorescence microscope (Carl Zeiss).

RT-quantitative PCR analysis
Total RNA was isolated using RNeay (Qiagen, Hilden, Germany) for RT-qPCR analysis of target genes. RT-qPCR was performed using the iQ SYBR Green supermix (Bio-Rad, CA, USA) and the CFX96™ Real-Time system (Bio-Rad, Singapore). Reverse transcription was performed using TOPscript™ RT DryMIX (Enzynomics, Daejeon, Republic of Korea). The relative amounts of target genes were normalized to those of glyceraldehyde-3-phosphate dehydrogenase (GAPDH). The primer sets used are listed in table S1 in supplementary data. The 2−ΔΔCq method was adopted to determine the fold changes (control vs. sample).

Cell viability, spheroid formation, migration and invasion assay
For phenotypic changes, cells were seeded in 96-well plates (2 × 103 cells/well) and incubated overnight at 37 °C with 5% CO₂. After transfection, 5 mg/mL MTT solution was added and incubated for 2 h. The blue precipitate was dissolved in 150 μl DMSO, and absorbance (550 nm) was measured using a microplate reader. All experiments were done in triplicate. For colony formation, cells were transfected for 48 h, then seeded at 500 cells/well in 6-well plates and incubated for 14 days. Colonies were washed, fixed with 3.7% paraformaldehyde, and stained with 1% crystal violet. Cell viability was assessed using the Cell Titer-Blue kit, and fluorescence intensity (555–585 nm, gain: 57) was measured using a SynergyHTX Fluorescent Microplate Fluorometer. For spheroids, cells were suspended in complete medium, seeded in ultra-low attachment plates, and incubated for 4 days. The spheroid viability was tested using the Cell Titer-Blue assay after trypsinization. Migration and invasion assays were performed in Transwells with 8-μm-pore filters, uncoated for migration or coated with matrigel for invasion. After incubation, non-migrated or non-invaded cells were removed, and the migrated or invaded cells were fixed, stained with crystal violet, and counted under microscope.

Structural modeling, peptides design, and synthesis
While this study was ongoing, the cryo-EM structure of human PIWIL2 (Hili, PDB: 7YFX) was published [18]. However, we used the AlphaFold-predicted model in this study, as the two structures showed an overall root mean square deviation of 1.45 Å (Fig. S1A). Since both wild-type and tPIWIl2 retains MID and PIWI domains that bind HDAC3 and NME2, we used the wild-type model for docking analysis.
For NME2, we utilized its crystal structure (PDB: 7KPF) to analyze its binding with PIWIL2. This model contains the N-terminal hotspots that have been reported to bind PIWIL2 [19]. For HDAC3, the crystal structure (PDB: 4A69) lacks the C-terminal domain (aa 376–428) [20], which has been reported to exclusively bind the PIWI domain of PIWIL2 [21]. To add this motif into HDAC3, we retrieved the full-length HDAC3 model from the AlphaFold database for docking analysis [22, 23].
Protein–protein docking was performed using MOE, ClusPro and AlphaFold 3. AlphaFold generated high-confidence models using five random seeds and the interface were compared with that of previously reported hotspots [19, 21]. Models deviating from experimental results were excluded and the rest were ranked by AlphaFold’s predicted Local Distance Difference Test (pLDDT) and Predicted Aligned Error (PAE) scores. The final docking model was selected based on three criteria: (1) Lowest interface PAE (< 5 Å) and highest mean pLDDT (> 85); (2) Consistency of interface residues with previously reported binding hotspots [20, 31]; and (3) Thermodynamic stability confirmed by 100 ns molecular dynamics simulations (MDS). Models showing poor interface stability (fluctuation > 2 Å RMSF at the hotspot residues) or inconsistent interaction patterns were excluded. The final model used for analysis originated from the AlphaFold 3 were consistent in both interface accuracy and post-simulation stability. Final models underwent molecular dynamics simulations using GROMACS and binding free energy (BFE) calculations using MMPBSA methods [24]. We also employed MOE and ClusPro protein–protein docking platforms using AlphaFold-predicted peptide structures; however, the resulting docking poses were relatively inconsistent compared to those generated by AlphaFold 3.
Critical binding motifs within HDAC3 and NME2 contributing to BFE and interface stability were identified and subjected to in silico alanine mutagenesis [25]. These motifs served as templates to design synthetic decoy peptides (NEP1, NEP2, and HDEP1) targeting PIWIL2's interaction with HDAC3 or NME2, as described previously [25]. Peptide candidates were docked against PIWIL2 using alphaFold3, as described above.
All peptides were linked with CPP by their N-terminal and synthesized by GeneCust (Boynes, France) at a purity of over > 90%, as determined by reversed-phase high-performance liquid chromatography (HPLC; Shimadzu Prominence), described previously [25]. The HPLC reports and related information about peptide synthesis are provided in supplementary data (see Supplementary data).

Statistics and reproducibility
All the statistical analysis was performed using SigmaPlot v12.5 software (Systat Software, Inc., San Jose, CA, USA), R packages (www.r-project.org), and GraphPad Prism (version 7). All experiments were performed in triplicate, and the data are represented as the mean ± standard deviation. *P < 0.05, **P < 0.01, and ***P < 0.001 compared to siNC or vector group. Significant difference was determined using the two-tailed Student’s t-test or One Way ANOVA Tukey test. P < 0.05 was considered statistically significant. Data were analyzed using SigmaPlot software (Systat Software, Inc.) to evaluate the two parameters (logistic three and quadratic) and determine the IC50 of the peptide or drug. The isobologram analysis [26] evaluates the nature of interaction of two drugs, i.e., drug A and drug B, at a given effect level.
Combination index (CI) is calculated as below:
A CI of less than, equal to, and more than 1 indicates synergy, additivity, and antagonism, respectively.

Results

Identification of SLC39A14-PIWIL2 fusion transcript in HCC
We used three methods—SOAPfuse, ChimeraScan, and TopHat-Fusion—to identify tumor-specific fusion transcripts with over 30 chimeric reads at the breakpoint involving protein-coding genes. Fourteen potential fusion transcripts were consistently detected by all methods (Fig. 1A). While no recurrent fusions were observed, we hypothesized that functional fusions might be expressed at higher levels than native transcripts. Among these, the SLC39A14-PIWIL2 fusion exhibited the highest expression (4.5-fold) relative to its native form (Fig. 1B). This fusion, caused by a chromosomal inversion on chromosome 8, joins exon 1 of SLC39A14 with exons 7–23 of PIWIL2 (Fig. 1C). Expasy Translate predicted a ~ 82.48 kDa PIWIL2 protein with exon 1 of SLC39A14 not contributing to the functional ORF (Fig. 1D, Fig. S2A). AlphaFold modeling revealed that the ~ 82.48 kDa PIWIL2 product lacks the intrinsically disordered (ID) region and the L0 motif in the N-terminal domain required for RNA binding but retains the PAZ, MID, and PIWI domains (Fig. 1D). We designated this product as truncated PIWIl2 “tPIWIL2”.
Sanger sequencing confirmed SLC39A14-PIWIL2 expression at the RNA level in the HCC sample (AJHCC007) but not at the genomic DNA level (Fig. 1E). Interestingly, SLC39A14 expression in HCC samples, including AJHCC007, was lower than in non-tumor samples, while PIWIL2 expression was markedly elevated in AJHCC007. However, SLC39A14 levels remained higher than PIWIL2 in other HCC and control samples, suggesting liver-specific induction (Fig. 1F, Fig. S2B) [27].
Using The Cancer Genome Atlas (TCGA), we identified SLC39A14-PIWIL2 fusions in stomach adenocarcinoma (STAD) and lung squamous cell carcinoma (LUSC), with breakpoints differing from those in AJHCC007 (Fig. S2C). In STAD, the fusion encodes full-length PIWIL2, while in LUSC, it produces a shorter PIWIL2 (530 aa, 60.7 kDa) similar to PL2L60, which promotes tumorigenesis via NF-κB [28]. Despite varying lengths, all forms are associated with tumorigenesis when expressed outside the testis.

SLC39A14-PIWIL2 (tPIWIL2) promotes HCC progression
SLC39A14 maintains metal ion homeostasis in the liver and pancreas, resulting in higher expression in these organs compared to others, as indicated by the Human Protein Atlas dataset (Fig. S3A) [27]. Conversely, PIWIL2, a member of the PIWI subfamily of Argonaute proteins, is crucial for genome integrity during germ cell development and is predominantly expressed in the testis and duodenum, with minimal expression in liver tissues (Fig. S3B) [29]. These observations suggest that the rearrangement of SLC39A14 exon 1 into the 5' exon 7 of PIWIL2 likely enhances PIWIL2 expression in HCC. We hypothesized that tPIWIL2 expression in HCC is driven by the SLC39A14 promoter. RT-PCR analysis of the predicted promoter region (1000 bp) in AJHCC007 confirmed transcriptional activity, suggesting that the SLC39A14 promoter and exon 1 are rearranged into the 5′ region of PIWIL2 exon 7, inducing aberrant PIWIL2 expression (Fig. S3C).
Since abnormal transcripts or proteins are often degraded via mechanisms like nonsense-mediated mRNA decay or the ubiquitin–proteasome system [30], we sought to determine whether SLC39A14-PIWIL2 produces a functional protein. We cloned and expressed WT PIWIL2, tPIWIL2, and X1-PIWIL2 (without exon 1) in Huh7 and HepG2 cells. RT-PCR and western blot analyses confirmed the expression of tPIWIL2 and X1-PIWIL2 at similar molecular weights (~ 80 kDa), consistent with exon 1 of SLC39A14 not contributing to the functional ORF (Fig. 2A, Fig. S3D). These results validate that SLC39A14-PIWIL2 encodes a functional PIWIL2 protein (tPIWIL2).
Aberrant PIWIL2 expression has been implicated in promoting proliferation in HCC and other cancers [19, 31], with its tumorigenic functions mediated through interactions with HDAC3, NME2, β-catenin (CTNNB1), and others via the PIWI and MID domains [19, 21, 28, 32]. In agreement, we demonstrated that tPIWIL2, like WT PIWIL2, enhances proliferation, invasion, and migration of liver cancer cells (HepG2 and Huh7, Fig. 2B). These findings confirm that the SLC39A14-PIWIL2 fusion transcript expresses an oncogenic tPIWIL2 protein, retaining its functional PIWI and MID domains.

Structural insights into the PIWIL2 binding proteins, NME2 and HDAC3
To explore the downstream pathways contributing to the tumor-promoting functions of the PIWIL2 fusion product, we focused on two known PIWIL2-binding proteins, NME2 and HDAC3, both of which are implicated in oncogenic signaling—NME2 through its role in metastasis, proliferation, and nucleoside diphosphate kinase activity, and HDAC3 through epigenetic repression of tumor suppressor genes and promotion of cell survival. To gain structural insights into their interactions with PIWIL2, both molecules were docked with PIWIL2 (Fig. 2C). It is reported that NME2, but not its homolog NME1, binds PIWIL2 [19]. To understand this specificity, we aligned the sequences of NME1 and NME2, identifying critical amino acid differences that influence PIWIL2 binding (Fig. 2C, left). Structural analysis revealed that NME2 interacts with both the MID and PIWI domains of PIWIL2 via residues unique to NME2 (Fig. 2D, left). Meanwhile, HDAC3 binds exclusively to the PIWI domain of PIWIL2, utilizing a C-terminal helix motif consistent with previous excremental findings (Fig. 2D, right) [21].
To evaluate the stability of these interactions, MDS were performed on the PIWIL2-NME2 and PIWIL2-HDAC3 complexes. BFE calculations and hydrogen bond density analyses demonstrated that both complexes maintained approximately 7.5 hydrogen bonds on average (Fig. 3A). Interestingly, BFE analyses revealed that PIWIL2 binds HDAC3 with greater affinity than NME2 (Fig. 3B), likely due to the higher number of electrostatic bonds observed in the PIWIL2-HDAC3 complex (Fig. 2D, right). Root mean square fluctuation analysis indicated that the N and PAZ domains of PIWIL2 exhibit higher flexibility compared to the MID and PIWI domains, with the latter being further stabilized upon binding to NME2 and HDAC3 (Fig. 3C). These findings underscore the importance of HDAC3 and NME2 in mediating PIWIL2's oncogenic functions, with interactions occurring primarily through the MID and PIWI domains, independent of the N-terminal and PAZ domains.

PIWIL2 inhibiting peptides design and in vitro validation
Given the critical role of the PIWIL2-HDAC3 and PIWIL2-NME2 binding in initiating oncogenic pathways, we used in silico alanine mutagenesis and identified two critical motifs in NME2 (amino acids 34–56 and 115–140) and a helical motif in HDAC3 (amino acids 351–370) that significantly contribute to the binding energies of their respective complexes (Fig. 4A). Based on these insights, we designed decoy peptides that could selectively disrupt these interactions, using decoy-based peptide design strategy (Fig. 4B). We have previously utilized this strategy in designing TLRs and SARS-CoV-2 inhibiting peptides [25, 33]. We found that all three peptides, NEP1, NEP2, and HDEP1, compete HDAC3 and NME2 for PIWIL2 binding, as suggested by the peptides clustering around MID and PIWI domains of PIWIL2 (Fig. 4C). This competitive binding is expected to attenuate the PIWIL2-mediated stabilization of HDAC3, the c-Myc regulatory function of NME2 and downstream oncogenic effects, such as cell proliferation, and modulation of critical signaling pathways like Wnt and Src/STAT3.
Using liver cancer cell lines (Huh7, HepG2, SNU449, SNU398) overexpressing PIWIL2 (Fig. S4A), we assessed peptide efficacy through cell viability and sphere formation assays. NEP1 consistently demonstrated the highest potency, achieving IC50 values of 8 μM in SNU398 and 11 μM in SNU449 for cell viability assays, and 9.6 μM (Huh7) and 8.6 μM (SNU398) in sphere formation assays. HDEP1 showed moderate activity, with IC50 values of 10 μM (SNU398) and 28 μM (SNU449) for cell viability, and 10 μM (Huh7) and 9.9 μM (SNU398) for sphere formation (Fig. 4D, E).

NEP1 suppresses liver cancer oncogenesis by targeting the PIWIL2 pathway
To investigate NEP1's effects on the PIWIL2 pathway, we analyzed RNA levels of PIWIL2-associated proteins in SNU398 and SNU449 cells. PIWIL2 overexpression significantly upregulated CTNNB1, c-myc, K8, and NME2 transcripts (Fig. 5A), while treatment with 10 μM NEP1 suppressed these changes, indicating NEP1's regulatory impact. Immunoblotting confirmed elevated p-AKT, GSK3β, and c-myc levels in PIWIL2-overexpressing cells, which NEP1 treatment restored to baseline (Fig. 5B). This aligns with prior findings that PIWIL2 knockdown reduces AKT and GSK3β phosphorylation [34]. Discrepancies in CTNNB1 mRNA and protein levels may result from p-GSK3β-mediated CTNNB1 stabilization, which increases protein accumulation without altering transcription [35–37]. Thus, RT-qPCR reflects CTNNB1 transcription, while immunoblotting captures post-translational regulation. Next, immunocytochemistry was performed to assess c-myc subcellular localization under PIWIL2 overexpression and NEP1 treatment. PIWIL2 caused significant nuclear accumulation of c-myc in both cell lines, which NEP1 effectively reversed (Fig. 5C). This aligns with previous findings that PIWIL2 interacts with NME2 to promote c-myc-driven proliferation in HeLa and HepG2 cells [19]. NEP1, in contrast, suppresses c-myc expression, inhibiting cell proliferation.
5-Fluorouracil (5-FU) is a pyrimidine analog that disrupts DNA and RNA synthesis [38] and is used to treat various cancers, including colorectal, breast, gastric, pancreatic, and head and neck malignancies [39]. We examined the combined effects of NEP1 and 5-FU on liver cancer cell line SNU449 (Fig. 5D) and SNU398 (Fig. 5E). PIWIL2 overexpression increased the IC50 of 5-FU but decreased the IC50 of NEP1. Co-administration of NEP1 and 5-FU at equal concentrations (0, 2, 5, 10, 25, and 50 µM) reduced the IC50 of the combination under PIWIL2 overexpression. CI analysis showed antagonism at IC30 but synergy at IC50, with CI values of 0.693 and 0.842 in SNU398 and SNU449, respectively. These findings suggest that NEP1 reverses PIWIL2-driven oncogenic changes, including c-myc nuclear accumulation and chemoresistance, and synergistically enhances 5-FU efficacy, supporting its potential as a targeted cancer therapy. This approach also highlights opportunities for personalized treatment strategies targeting specific molecular drivers.

Discussion
Fusion-driven oncogenesis often involves promoter activation of oncogenes by tissue-specific or highly expressed genes. For example, TMPRSS2-ERG and SLC45A3-BRAF fusions in prostate cancer similarly amplify oncogene expression via highly active promoters [40]. The SLC39A14-PIWIL2 fusion identified in this study follows this mechanism, with the SLC39A14 promoter driving PIWIL2 overexpression, amplifying its tumor-promoting potential in HCC. The rearrangement activates the SLC39A14 promoter, which drives the expression of PIWIL2, an oncogene typically restricted to gonads [41], leading to low molecular weight PIWIL2 (tPIWIL2) overexpression. This promotes HCC progression by activating oncogenic pathways.
PIWIL2 promotes tumorigenesis through interactions with key proteins and pathways. It stabilizes HDAC3 via its PIWI domain by preventing degradation and enhancing phosphorylation by CK2α, promoting cell proliferation and suppressing apoptosis [21]. It interacts with NME2, supporting c-Myc-mediated oncogenesis [19], and binds CTNNB1 via its PAZ domain, implicating the Wnt signaling pathway [32]. Additionally, PIWIL2 inhibits apoptosis by forming a PIWIL2/K8/p38 complex, stabilizing K8, reducing Fas, and repressing p53 phosphorylation [42]. These multifaceted interactions make PIWIL2 a critical driver of oncogenesis and a promising therapeutic target.
The protein product of the SLC39A14-PIWIL2 fusion retains the oncogenic MID and PIWI domains of PIWIL2, driving proliferation, invasion, and migration in Huh7 and HepG2 cells despite losing the intrinsically disordered region and L0 motif. Interestingly, similar fusion transcripts were found in STAD and LUSC, with distinct breakpoints and functions. In LUSC, the truncated PIWIL2 isoform resembles PL2L60, associated with NF-κB activation and tumorigenesis [28], underscoring its potential as a universal therapeutic target. Our structural modeling revealed that tPIWIL2 interacts with HDAC3 and NME2 via its MID and PIWI domains, thereby potentially driving its aberrant oncogenicity in liver tissues. Therefore, key residues essential for these interactions were identified through molecular dynamics simulations and alanine mutagenesis, enabling the design of decoy peptides (NEP1 and HDEP1) to competitively disrupt these interactions.
Current HCC treatments, such as immune checkpoint inhibitors and multi-kinase inhibitors, face limitations due to tumor heterogeneity and resistance [43]. Targeting PIWIL2 through small molecules, RNA-based approaches, or peptide therapeutics offers a novel strategy for addressing these challenges. The peptides developed in this study effectively attenuate PIWIL2-mediated oncogenic signaling, demonstrating therapeutic potential. Moreover, the SLC39A14-PIWIL2 fusion could serve as a biomarker for early diagnosis and disease stratification. Detecting fusion transcripts in serum, as shown with other fusions like SLC45A2-AMACR [44], could enable minimally invasive diagnostic assays tailored to HCC.

Limitations and future directions
We identified a rare case of SLC39A14-PIWIL2 expression in HCC, with evidence of its occurrence in other cancer types. Although the expression frequency of the fusion transcript is low, targeted therapies against this fusion may provide promising approach for managing patients harboring this alteration. In addition, this study establishes the oncogenic role of SLC39A14-PIWIL2 in HCC; however, further investigation is needed to elucidate its role in the tumor microenvironment and its interplay with immune evasion mechanisms. Future studies should also focus on optimizing the delivery and stability of PIWIL2-inhibiting peptides for in vivo applications. Although we partly evaluated (5-FU), combining these targeted therapies with existing treatment modalities, such as immune checkpoint inhibitors or kinase inhibitors, could provide synergistic benefits and overcome resistance mechanisms in advanced HCC.

Supplementary Information