The present study was approved by the Institutional Review Board of Tungs' Taichung MetroHarbor Hospital, Taichung, Taiwan. (approval no. 113003). All experimental procedures followed the relevant ethical guidelines and informed consent was obtained from all patients with TC included in the present study. Thyroid tissue samples were collected from a cohort of these patients.

The patient samples were collected from Tungs' Taichung MetroHarbor Hospital, Taichung Taiwan. Samples were collected between 2024/05/01 and 2024/05/31. A total of 11 patients participated in the study, five samples for each patient, and a total of 55 samples were collected for analysis. Samples were preserved and processed for immunohistochemical (IHC) analysis to evaluate the expression of specific proteins. Clinical data on the patients' age, sex and cancer staging were obtained from the hospital's electronic medical records (Table I). All samples were anonymized to ensure patient confidentiality in accordance with Institutional Review Board (IRB) regulations.

The expression of proteins was assessed using IHC on formalin-fixed, paraffin-embedded tissue sections. The sections were cut into 4 µm thick slices and mounted on positively charged glass slides. The slides were deparaffinized in xylene and rehydrated using a decreasing series of graded alcohol solutions, followed by a rinse in distilled water. For antigen retrieval, heat-induced epitope retrieval using a citrate buffer (pH 6.0) was performed for 20 min in a pressure cooker. To block endogenous peroxidase activity, sections were incubated with 3% hydrogen peroxide (H₂O₂) in methanol for 10 min at room temperature, followed by washing in PBS. After retrieval, the slides were allowed to cool to room temperature and then washed in PBS. To block non-specific binding, the slides were incubated with 5% normal goat serum (Vector Laboratories, Inc. cat. no. ZH083) with blocking solution for 30 min at room temperature. The tissue sections were incubated overnight at 4°C with the following primary antibodies: Mouse monoclonal anti-CDK5 (cat. no. sc-249; Santa Cruz Biotechnology, Inc.; 1:20) or rabbit polyclonal anti-p21 (cat. no. 2947; Cell Signaling Technology, Inc.; 1:50). After washing the slides in PBS, they were incubated with appropriate biotinylated secondary antibodies for 30 min at room temperature (Goat anti-mouse IgG, cat. no. BA-9200 or Goat anti-rabbit IgG, cat. no. BA-1000; Vector Laboratories, Inc.; dilution 1:200), followed by a wash in PBS. Signals were visualized using a horseradish peroxidase (HRP)-conjugated streptavidin detection system (Vectastain ABC kit, Vector Laboratories, Inc.). The slides were developed using 3,3′-diaminobenzidine (DAB; Dako) as the chromogen for 5 min at room temperature and counterstained with hematoxylin to visualize nuclei for 2 min at room temperature. Finally, the sections were dehydrated, cleared in xylene and a coverslip was placed over them using a permanent mounting medium. Positive control slides were included for each staining run. Negative controls were performed by omitting the primary antibodies to assess non-specific binding of the secondary antibody.

Stained slides were scanned using a high-resolution digital microscope (Zeiss Imager. Z2; Carl Zeiss AG) and images were captured for analysis. Protein expression was semi-quantitatively evaluated based on staining intensity and the percentage of positive cells using TissueFAXS viewer version 7.1.6. supplied by TissueGnostics GmbH. The results were classified into low, moderate, or high expression levels. Statistical analysis was performed to evaluate the correlation between protein expression and clinical outcomes in patients with TC.

Human TC cell lines BCPAP and TPC-1 were purchased from the Food Industry Research and Development Institute in Taiwan. The two TC cell lines were provided by collaborators Dr Chen-Kai Chou and Dr Hong-Yo Kang in Kaohsiung Chang Gung Hospital (Taiwan) (46). Cells were maintained in RPMI-1640 culture medium (Gibco; Thermo Fisher Scientific, Inc.), supplemented with 10% FBS (Gibco; Thermo Fisher Scientific, Inc.), penicillin-streptomycin at concentrations of 100 IU/ml and 100 µg/ml, respectively (MilliporeSigma), along with 2 mM L-glutamine, 1.5 g/l sodium bicarbonate (NaHCO₃) (MilliporeSigma), 10 mM HEPES (MilliporeSigma) and 1 mM sodium pyruvate (MilliporeSigma). For cell transfection, cells were transfected with either short hairpin (sh)RNA targeting CDK5 (shCDK5) or a scrambled shRNA control (shCtrl) using Lipofectamine® 2000 (Invitrogen; Thermo Fisher Scientific, Inc.) as the transfection reagent for 6 h at 37°C. Overexpression plasmids were transfected in the same manner. Transfections were performed in serum-free medium according to the manufacturer's protocol. After 6 h of incubation, the transfection medium was replaced with fresh complete medium and cells were cultured for an additional 24–48 h before proceeding with downstream analyses. Plasmids used for transfection were obtained from the National RNAi Core Facility (Academia Sinica). Specifically, the pLKO.1 plasmid harboring shRNA sequence targeting CDK5 or scrambled control (shCtrl) were used at a concentration of 2 µg/ml. The target sequences of shGFP and shCDK5 were: TRCN0000072192 (shGFP): 5′-GAACGGCATCAAGGTGAACTT-3′, TRCN0000021466 (shcdk5 #1): 5′-TGTCCAGCGTATCTCAGCAGA-3′, TRCN0000199652 (shcdk5 #2): 5′-GTGAACGTCGTGCCCAAACTC-3′, TRCN0000194974 (shcdk5 #3): 5′-CCTGAGATTGTAAAGTCATTC-3′.

Total RNA was extracted from cultured cells using TRIzol reagent (Invitrogen; Thermo Fisher Scientific, Inc.) following the manufacturer's instructions. RNA quality and concentration were assessed by spectrophotometry. Reverse transcription was performed using 1 µg of total RNA with the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems; Thermo Fisher Scientific, Inc.) according to the manufacturer's instructions. qPCR was performed using SYBR Green (Roche Applied Science) on a real-time PCR system by using Applied Biosystems StepOnePlus (Applied Biosystems; Thermo Fisher Scientific, Inc.) as described previously (22). Gene expression levels were normalized to the housekeeping gene, and relative quantification was calculated using the 2^−ΔΔCq method as described previously by Livak and Schmittgen (47). The thermocycling conditions were as follows: initial denaturation at 95°C for 10 min, followed by 40 cycles of denaturation at 95°C for 15 sec, annealing and extension at 60°C for 1 min. Each experiment was performed in triplicate and independently replicated at least three times. The following primers were used to amplify the cDNA: CDKN1A (5′-ACCCTAGTTCTACCTCAGGC-3′ and 5′-AAGATCTACTCCCCCATCAT-3′), CDK5 (5′-TCTTTTTCCCGGCAATGAT-3′ and 5′-TCTGGCAGCTTGGTCATAGA-3′), and actin (5′-TTGCCGACAGGATGCAGAA-3′ and 5′-GCCGATCCACACGGAGTACT-3′).

Cells were lysed at approximately 80–90% confluency using lysis buffer [50 mM Tris-HCl (pH 7.4), 150 mM NaCl, 1% NP-40, 0.5% sodium deoxycholate, 0.1% SDS] supplemented with a protease inhibitor cocktail (MilliporeSigma). Protein concentrations were measured using a BCA Protein Assay Kit (Pierce; Thermo Fisher Scientific, Inc.). Equal amounts of protein (30 µg per lane) were separated by electrophoresis on 10% SDS-polyacrylamide gels (SDS-PAGE) and transferred onto polyvinylidene fluoride (PVDF) membranes (MilliporeSigma). Membranes were blocked in 5% non-fat dry milk (MilliporeSigma) in TBS containing 0.1% Tween-20 (TBST) for 1 h at room temperature. After blocking, membranes were incubated overnight at 4°C with primary antibodies diluted in blocking buffer: anti-CDK5 (cat. no. ab40773; Abcam; 1:1,000) and anti-p21 (cat. no. 2947; Cell Signaling Technology, Inc.; 1:1,000). After incubation, membranes were washed three times (5 min each) in TBST (Tris-buffered saline containing 0.1% Tween-20). Subsequently, membranes were incubated with HRP-conjugated secondary antibodies (goat anti-rabbit IgG-HRP, cat. no. 7074, 1:5,000; goat anti-mouse IgG-HRP, cat. no. 7076, 1:5,000; Cell Signaling Technology, Inc.) at room temperature for 1 h. Following incubation, membranes were washed three times (5 min each) in TBST. Proteins were visualized using enhanced chemiluminescence (ECL) detection reagent (Pierce ECL Western Blotting Substrate; Thermo Fisher Scientific, Inc.) according to the manufacturer's protocol. Signal intensities were captured using a chemiluminescent imaging system (Bio-Rad Laboratories, Inc.). β-actin (cat. no. 4970; Cell Signaling Technology, Inc.; dilution 1:2,000) was used as the loading control. Each experiment was independently repeated at least three times.

IP was performed as previously described (48). Briefly, BCPAP and TPC-1 cells were seeded into 10-cm culture dishes at a density of ~1×10⁶ cells per dish and cultured until reaching 80–90% confluency. BCPAP and TPC-1 cells were lysed in IP lysis buffer (50 mM Tris-HCl pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% NP-40 and protease inhibitors). Protein concentration was determined using a BCA Protein Assay Kit (Pierce; Thermo Fisher Scientific, Inc.). Equal amounts of protein lysate (500 µg) were incubated with 2 µg anti-CDK5 antibody (Abcam) overnight at 4°C with gentle rotation. Protein A/G agarose beads (Thermo Fisher Scientific, Inc.) were added and incubated for an additional 2 h at 4°C. Input controls (30 µg of total protein lysate without immunoprecipitation) were included for western blot analysis to confirm the presence of target proteins. Beads were washed three times with ice-cold washing buffer (20 mM Tris-HCl pH 7.4, 150 mM NaCl, 0.1% NP-40) and proteins were eluted with SDS sample buffer by boiling for 5 min. Eluates were analyzed using 10% SDS-PAGE followed by western blotting using an anti-p21 antibody (cat. no. 2947; Cell Signaling Technology, Inc.) as aforementioned.

Protein stability assays were performed as previously described (49). To assess p21 protein stability, BCPAP and TPC-1 cells were treated with cycloheximide (CHX; MilliporeSigma) at a final concentration of 50 µg/ml to inhibit protein synthesis. Cells were collected at 0, 2, 4 and 6 h post-CHX treatment. For proteasome inhibition, cells were treated with 10 µM MG132 (Calbiochem; Merck KGaA) for 6 h prior to lysis. Cell lysates were prepared and protein levels of p21 and CDK5 were determined by western blotting as aforementioned. β-Actin (Cell Signaling Technology, Inc.) was used as the loading control.

Immunocytochemistry was performed as previously described (3). Briefly, cells were fixed for 5 min in 4% paraformaldehyde and 2% sucrose in PBS at room temperature. Subsequently, the cells were permeabilized and blocked in PBS buffer containing 0.1% Triton X-100 and 5% bovine serum albumin (Cyrus Bioscience, Lot no. 1260601) for 15 min at room temperature. The primary antibodies CDK5 (Santa Cruz Biotechnology, Inc.; cat. no. sc-249; dilution 1:100) and p21 (Cell Signaling Technology, Inc.; cat. no. 2947; dilution 1:200) were diluted in 5% BSA in PBS and added to the cells for overnight incubation at 4°C. Then, the cells were incubated with Alexa Fluor 488-conjugated secondary antibodies (Thermo Fisher Scientific, Inc.; cat. no. A-11008; dilution 1:1,000) and Alexa Fluor 546-conjugated secondary antibodies (Thermo Fisher Scientific, Inc.; cat. no. A-11003; dilution 1:1,000) for 1 h at room temperature. The cells were washed again with PBS and counterstained with DAPI (Thermo Fisher Scientific, Inc.) for 5 min at room temperature. Cells were subsequently washed with PBS, mounted using Fluoromount-G (Southern Biotech; cat. no. 0100-01) and analyzed using a fluorescence microscope (Olympus BX-51; Olympus Corporation) and a Zeiss LSM510 confocal microscope (Carl Zeiss AG).

The 3D structures of proteins were visualized as previously described (3). The interaction between CDK5 and p21 was predicted using AlphaFold 3.0, an AI-based protein structure prediction tool (50). The protein structures were visualized and analyzed for interaction sites using UCSF Chimera software (Resource for Biocomputing, Visualization, and Informatics, University of California, San Francisco, CA, USA). Chimera was used to map potential interaction interfaces and validate the predicted binding sites.

TCGA data analysis was performed as previously reported (5). Expression data and clinical information for patients with TC were downloaded from TCGA using the UCSC Xena browser (https://xenabrowser.net/). Pearson correlation analysis was performed to evaluate the relationship between CDK5 and p21 mRNA expression levels. Kaplan-Meier survival analysis was conducted using the ‘survival’ package in R to compare overall survival between high and low-expression groups for CDK5 and p21. Kaplan-Meier survival analysis was performed using the ‘survival’ package in R. Differential expression analysis between tumor and normal tissues was assessed using the Wilcoxon rank-sum test. Violin plots were generated to visualize the stage-specific expression of CDK5 and p21 across different stages of TC. Differential expression analysis was conducted to compare CDK5 and p21 expression in tumor compared with normal tissues, with statistical significance assessed using the Wilcoxon rank-sum test. Statistical analyses and visualizations, including Pearson correlation, Kaplan-Meier survival analysis, violin plots, and differential expression analysis, were conducted using R software (R Core Team, 2023). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria (http://www.R-project.org/).

All statistical analyses were conducted using GraphPad Prism version 8.0.1 (Dotmatics). Data are presented as the mean ± SD of at least three independent experiments. Comparisons between two groups were performed using an unpaired Student's t-test. Pathological stage plots were obtained from GEPIA (http://gepia.cancer-pku.cn/), and differential gene expression was analyzed by one-way ANOVA followed by Tukey's post hoc test. P<0.05 was considered to indicate a statistically significant difference.

The interaction between CDK5 and p21^CIP1 is pivotal for understanding the regulatory mechanisms underlying cell cycle progression and tumor suppression in TC. To investigate this interaction, AlphaFold 3.0, a cutting-edge AI-based protein structure prediction tool, was used to generate a model of the CDK5-p21 protein-protein interaction. The accuracy of AlphaFold in predicting protein structures based on amino acid sequences provided a reliable foundation for the analysis (50). Using data generated from AlphaFold, Chimera was used to visualize and analyze the interaction sites between CDK5 and p21. The analysis revealed that CDK5 directly interacted with specific regions of the p21 protein, indicating potential sites for regulation through phosphorylation and other post-translational modifications (Fig. 1A-D). This interaction is significant as it implies modulation of the stability of p21 and activity in TC cells, potentially contributing to enhanced cell proliferation and malignancy. To experimentally validate the CDK5-p21 interaction, IP assays were performed using the TC cell line TPC-1. Additionally, cells were treated with MG132, a proteasome inhibitor, to prevent proteasomal degradation and ensure the detection of any ubiquitin-dependent degradation products. The IP analysis confirmed that CDK5 biochemically interacted with p21 protein in TPC-1 cells, reinforcing the computational predictions and suggesting a direct regulatory relationship between CDK5 and p21 in TC (Fig. 1D). These findings provide a comprehensive understanding of the CDK5-p21 interaction, demonstrating its potential role in modulating cell cycle regulation and promoting the malignancy of TC.

To investigate the role of CDK5 in the regulation of p21^CIP1 in TC cells, several experiments were performed to analyze protein interactions, degradation and cellular localization. TC cells transfected with either an empty vector (EV) or CDK5 were treated with CHX to inhibit protein synthesis. Western blot analysis showed that p21 levels decreased more rapidly in cells overexpressing CDK5 compared with the EV control. This suggested that CDK5 accelerated p21 degradation (Fig. 2A and B). Next, cells were treated with MG132, a proteasome inhibitor, to block protein degradation. In mock-treated cells, overexpression of CDK5 resulted in lower p21 levels compared with the EV control. In MG132-treated cells, p21 levels did not differ markedly between CDK5-overexpressing cells and EV controls. In addition, IHC was used to further confirm the results. In control cells (vector), p21 was normally expressed, while in CDK5-overexpressing cells, p21 levels were visibly reduced. Upon MG132 treatment, p21 levels were restored in CDK5-overexpressing cells, confirming the role of proteasome-mediated degradation (Fig. 2C-E). The data collectively demonstrate that CDK5 downregulates p21 in TC cells by promoting its ubiquitin-dependent degradation via the proteasome pathway. This regulatory mechanism suggests a significant role for CDK5 in modulating cell cycle progression and tumor development in TC by targeting p21 for degradation. To further elucidate the mechanism by which CDK5 regulated p21, Co-IP experiments were performed to analyze the interaction between CDK5 and p21. Wild-type (WT) p21 and its phosphorylation-deficient mutant (MT; S130A) were compared. Western blot analysis revealed that CDK5 effectively interacted with WT, as shown by the co-precipitation of CDK5 and p21 proteins (Fig. 2F). However, the intensity of the interaction signal was lower for MT p21, suggesting that the S130 site influenced the strength of the CDK5-p21 interaction (Fig. 2F). To assess the functional impact of CDK5-mediated p21 regulation on cell proliferation, TC cells were analyzed for their proliferation rates in the presence of WT and MT p21. The results showed that CDK5 overexpression markedly enhanced cell proliferation when WT p21 was present. By contrast, cells expressing MT p21 (S130A) displayed suppressed CDK5-mediated proliferation (Fig. 2G), suggesting that the phosphorylation of p21 at the S130 site is essential for CDK5 to promote cell proliferation. These findings indicated that the phosphorylation-deficient mutant p21 (S130A) blocks the ability of CDK5 to enhance cell proliferation, thereby acting as a dominant-negative form to counteract CDK5-mediated oncogenic effects. The data collectively suggested that CDK5 promoted cell proliferation in TC by targeting p21 for phosphorylation-dependent degradation and that MT p21 can serve as a potential therapeutic tool to inhibit CDK5-driven tumor progression.

To further elucidate the regulatory relationship between CDK5 and p21^CIP1 in TC cells, several experiments examining mRNA and protein expression levels, as well as the effects of CDK5 knockdown were performed. RT-qPCR was used to measure the mRNA expression levels of CDK5 in BCPAP and TPC-1 cells. The results showed markedly higher CDK5 mRNA expression in BCPAP cells compared with TPC-1 cells (Fig. 3A). In addition, TPC-1 cells had markedly higher p21 mRNA expression levels than BCPAP cells (Fig. 3B). Western blot analysis was performed to examine the protein expression levels of CDK5 and p21 in BCPAP and TPC-1 cells, both with and without MG132 treatment. In the absence of MG132, TPC-1 cells showed higher levels of CDK5 and lower levels of p21 expression compared with BCPAP cells. MG132 treatment increased p21 levels in both cell lines, but the increase was more pronounced in TPC-1 cells, indicating that CDK5 promoted p21 degradation through the proteasome pathway (Fig. 3C). In addition, TPC-1 and BCPAP cells were transfected with shRNAs targeting CDK5 (shCDK5 #1, shCDK5 #2 and shCDK5 #3) or a control shRNA (shGFP). Following CDK5 knockdown, p21 protein expression levels were measured. Cells with CDK5 expression knocked down exhibited a significant increase in p21 levels compared with the shGFP control (Fig. 3D and E). This indicated that CDK5 negatively regulates p21 expression, as knockdown of CDK5 led to an increase in p21 protein expression levels.

To understand the clinical relevance of CDK5 and p21^CIP1 expression in TC, data from TCGA were analyzed. The analysis included correlation, survival, stage-specific expression and differential expression between tumor and normal tissues. A scatter plot showing the correlation between CDK5 and p21 mRNA expression levels in TC samples is shown in Fig. 4. The correlation coefficient (R) was 0.42, with a P-value <0.03, indicating a significant correlation between CDK5 and p21 expression in TC tissues (Fig. 4A). In addition, patients with high CDK5 expression had worse overall survival compared with those with low CDK5 expression (Fig. 4B). Conversely, patients with high p21 expression tended to have improved overall survival (Fig. 4C), indicating a negative correlation between CDK5 and p21 expression. Violin plots were used to show the expression levels of CDK5 and p21 across different stages of TC. CDK5 expression increases with advancing cancer stage, showing a difference between early-stage (Stage I/II) and late-stage cancer (Stage III/IV) (Fig. 4D). However, p21 expression decreased with advancing cancer stage, with a difference between early and late stages (Fig. 4E). Analysis of data from TCGA revealed a significant negative correlation between CDK5 and p21 expression in TC, supporting the experimental findings that CDK5 negatively regulated p21. Survival analysis indicated that high CDK5 expression was associated with poorer overall survival, while high p21 expression tended to be linked with improved survival outcomes. Additionally, CDK5 expression increased and p21 expression decreased with advancing cancer stage, suggesting their roles in TC progression. Differential expression analysis confirmed that CDK5 was up-regulated and p21 was down-regulated in TC tissues compared with normal tissues. These findings underscore the clinical relevance of targeting the CDK5-p21 axis in TC treatment.

IHC staining was performed to assess CDK5 and p21 expression in TC specimens from 11 patients (Fig. 5A). The expression profiles were analyzed in relation to tumor stage, sex and malignancy features (TNM staging; Table I and Data S1). In all cases, CDK5 and p21 were localized to the nuclei and cytoplasm of tumor cells. A distinct pattern emerged where patients with more advanced TNM stages, lymph node involvement, or recurrent tumors exhibited higher CDK5 and lower p21 expression (Fig. 5 and Table I). Notably, high CDK5 expression, particularly in patients 2, 3 and 9, was associated with more advanced-stage cancer (II, III and IVA) and malignancy features such as recurrence and invasion into surrounding structures. By contrast, moderate CDK5 and p21 expression were predominantly observed in early-stage cancers (Stage I), such as in patients 1, 6, 7, 8 and 11, where tumors remained smaller and confined to the thyroid. Notably, high p21 expression, as seen in patient 5, was rare and may suggest the presence of a unique tumor suppression pathway (Fig. 5A, Table I and Data S1). A comparative analysis of protein expression revealed that patients with more advanced cancer had a predominance of high CDK5 and low p21 expression, particularly in those with nodal involvement or invasive features, while early-stage patients had higher expression levels of p21 relative to CDK5 (Fig. 5B and C). These observations suggest a potential role of CDK5 in promoting tumor progression and a tumor-suppressive function for p21, with differential CDK5 and p21 expression profiles (Fig. 5), serving as valuable indicators for assessing tumor progression and identifying high-risk patients.