DNA methylation-driven gene FAM3D promotes colorectal cancer growth via the ATF4-SESN2-mTORC1 pathway

Methylation, expression, and clinical relevance of FAM3D in CRC patients

We have reported that the 3-CpG methylation prognostic model consisting of 3 methylation-driven genes (FAM3D, DAPP1, and PIGR) is an independent prognostic biomarker for CRC [13]. Based on this, our study further explored the prognostic significance of FAM3D in CRC as well as its impact on CRC cell functions and potential molecular mechanisms. The relevant workflow is shown in Supplementary Figure 1.

First, we examined the connection between the mRNA level and FAM3D promoter methylation using TCGA and GEO. The findings suggested an inverse relationship between FAM3D expression and methylation of its promoter region (rTCGA = -0.497 and rGSE106582+GSE101764 = -0.266; rGSE131013+GSE44076 = -0.405) (Figure 1A, 1B and Supplementary Figure 2A). Prognostic analysis found that hypermethylation of FAM3D not only resulted in decreased OS but also contributed to reduced DFS (Figure 1C, 1D). FAM3D methylation also acted as an independent predictor of the prognosis of CRC patients (HRTCGA (95% CI) = 2.02 (1.23-3.34), P = 0.006) (Supplementary Table 1). Next, we treated CRC cell lines (CW2 and LS513) with a series of concentrations of 5-Aza (Macklin) and examined FAM3D expression. Results showed elevated levels of FAM3D mRNA and protein after demethylation (Figure 1E, 1F).

Figure 1. Methylation, expression and prognosis of FAM3D. (A, B) Correlations between FAM3D mRNA expression and methylation of the FAM3D promoter in the (A) TCGA and (B) GEO databases (GSE106582 + GSE101764). (C, D) Kaplan-Meier curves of (C) OS and (D) DFS based on the methylation of the FAM3D promoter in tumor tissues from the TCGA cohort. (E, F) Relative FAM3D mRNA and protein expression after treatment with a series of concentrations of 5-Aza in (E) CW2 and (F) LS513 cells (differences compared to the DMSO group). (G) The mRNA expression of FAM3D in normal tissues and CRC tissues in the TCGA cohort. (H) The mRNA expression of FAM3D in normal cells and CRC cells in GSE132465. (I) The protein expression of FAM3D in normal tissues and COAD tissues in the CPTAC cohort. (J) Kaplan-Meier curve of OS based on FAM3D mRNA expression in tumor tissues from the TCGA cohort. *P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001.

We further explored FAM3D expression in CRC tissues in TCGA and GEO, which found that FAM3D mRNA was significantly reduced in CRC tissues (Figure 1G and Supplementary Figure 2B, 2C). In addition, consistently, FAM3D mRNA in CRC cells was significantly downregulated (Figure 1H), and it was mainly expressed in epithelial cells (Supplementary Figure 2D). Furthermore, according to the CPTAC database, FAM3D protein level was significantly reduced in colon cancer (COAD) tissues (Figure 1I). Although the expression of the FAM3D mRNA did not obviously change (Supplementary Figure 2E), FAM3D protein was significantly reduced in advanced-stage patients (Supplementary Figure 2F). Subsequent investigations showed that increased FAM3D mRNA expression was positively correlated with better OS and DFS (Figure 1J and Table 1 and Supplementary Figure 2G–2J). FAM3D mRNA was also discovered to be an independent prognostic factor for CRC patients (Supplementary Tables 2–4).

FAM3D promotes the malignant phenotypes of CRC cells in vitro

Immunofluorescence testing was used to determine the localization of the FAM3D protein in CRC cells (Figure 2A). Then, we knocked out the FAM3D gene in LoVo and HT29 cells. Sanger sequencing and Western blotting confirmed the KO efficiency of the construct (Supplementary Figure 3 and Figure 2B, 2C). Cell function tests were used to explore FAM3D’s impact on CRC progression. Colony formation and CCK-8 assays showed that FAM3D deficiency strongly inhibited the growth of LoVo and HT29 cells in vitro (Figure 2D–2F). In vitro migration and invasion tests showed that FAM3D KO cells had reduced motility (Figures 2G, 2H, 3A, 3B). To better understand the function of FAM3D, we established LoVo cells that overexpress FAM3D and confirmed the overexpression efficiency of the construct (Figure 3C, 3D). Conversely, increased expression of FAM3D promoted the malignant phenotypes of CRC cells (Figure 3E–3H). The results demonstrated that FAM3D promotes CRC progression in vitro.

Figure 2. FAM3D promotes CRC cell proliferation, migration, and invasion in vitro. (A) The localization of FAM3D in LoVo and HT29 cells. (B) Schematic diagram of sgRNA targeting the human FAM3D gene locus. Two sgRNA sequences, sgRNA1 and sgRNA2, are marked in red, and the protospacer adjacent motif (PAM) sequences are presented in blue. (C) Western blot analysis of FAM3D in the supernatant (secreted) and cell lysate (lysate) of LoVo and HT29 cells after FAM3D KO. (D–F) Cell proliferation was detected by (D) colony formation assay and (E, F) CCK-8 assay in LoVo and HT29 cells. (G, H) Cell migration was detected by a wound healing assay in (G) LoVo and (H) HT29 cells. **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 3. FAM3D promotes CRC cell proliferation, migration, and invasion in vitro. (A, B) Migration and invasion were detected by transwell assays in (A) LoVo and (B) HT29 cells. (C, D) Overexpression of FAM3D was confirmed by (C) RT-qPCR and (D) Western blot in LoVo cells. (E, F) Proliferation was detected after FAM3D overexpression. (G, H) Migration and invasion were detected after FAM3D overexpression. **P < 0.01, ***P < 0.001, ****P < 0.0001.

FAM3D promotes the growth of CRC cells in vivo

To assess FAM3D’s effect on cell proliferation in vivo, we generated LoVo xenograft nude mouse models (Figure 4A). The results found that FAM3D deletion resulted in a reduction in tumor size and weight, while high expression of FAM3D led to an increase in both parameters (Figure 4B–4D). Moreover, the naked mice’s body weights did not vary much (Figure 4E). The data demonstrated that FAM3D enhances CRC tumor growth in vivo.

Figure 4. FAM3D promotes the growth of CRC cells in vivo. (A) Photographs of nude mice after the administration of FAM3D KO or FAM3D-overexpressing LoVo cells for 19 days. (B) Excised tumors on Day 19. (C) The weights of excised tumors on Day 19. (D) The tumor volumes were calculated as length × width² × 0.5 every 2–3 days. (E) The body weights of the nude mice were recorded every 2–3 days. *P < 0.05; **P < 0.01.

FAM3D expression is inversely associated with SESN2 expression

Differentially expressed genes (DEGs) between FAM3D-knockout cells and negative control (NC) cells were detected using RNA sequencing, which found that there were 1167 upregulated genes and 1919 downregulated genes after FAM3D KO (Figure 5A, 5B). A focus was placed on the upregulated genes for subsequent analyses. GO analysis revealed the roles of these genes in the cell response, apoptosis, differentiation, and protein folding (Figure 5C), and KEGG analysis demonstrated these genes’ participation in pathways relevant to tumors, such as the p53 and mTOR signaling pathways (Figure 5D). Parallel analyses of CRC cells with high and low FAM3D expression yielded consistent results (Supplementary Figure 4A, 4B). Next, we selected six key genes associated with the p53 and mTOR pathways and further verified their expression levels, which showed that SESN2 expression increased most significantly in FAM3D KO cells and decreased most significantly in FAM3D overexpressing cells (Figure 5E, 5F). Further analysis found that SESN2 mRNA was downregulated in tissues of CRC (Figure 5G, 5H). Although high SESN2 expression was associated with good CRC prognosis in GSE161158 (P = 0.046) (Figure 5I), the association between SESN2 expression and CRC prognosis was only marginally significant in TCGA (P = 0.068) (Figure 5J). Taken together, these findings suggested that FAM3D may regulate SESN2, thereby promoting CRC progression.

Figure 5. FAM3D expression is inversely associated with SESN2 expression. (A) Volcano plot and (B) hierarchical clustering heatmap showing DEGs between FAM3D KO and NC LoVo cells. (C) GO functional analysis showed the top 30 biological processes of significantly upregulated genes after FAM3D KO. (D) KEGG enrichment analysis of genes significantly upregulated after FAM3D KO. (E, F) RT-qPCR was used to detect changes in the expression of representative genes in the p53 and mTOR pathways in FAM3D (E) KO or (F) overexpressing LoVo cells. (G, H) The mRNA expression of SESN2 in normal tissues and CRC tissues in (G) GSE18105 and (H) GSE106582. (I, J) Kaplan-Meier curves of DFS in (I) GSE161158 and (J) TCGA. *P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001.

FAM3D functions in CRC cells by targeting SESN2 and activating the mTORC1 pathway

To verify the regulatory relationship between FAM3D and SESN2, rescue tests were performed through knocking down SESN2 in FAM3D KO cells and overexpressing SESN2 in FAM3D overexpression cells. Figure 6A, 6B showed that SESN2 was successfully downregulated or upregulated, respectively. Functional experiments revealed that reducing SESN2 expression strongly enhanced, while increasing SESN2 expression strongly suppressed LoVo cell growth and motility; reduction of SESN2 reversed the tumor suppressive effects of FAM3D KO, while upregulation of SESN2 counteracted the oncogenic effects of high FAM3D expression (Figures 6C–6H, 7A, 7B). Numerous studies have revealed that SESN2 inhibits the mTORC1 pathway to have a tumor suppressor impact on many cancer types, including CRC [28, 29], endometrial cancer [30], non-small cell lung cancer [31], and neuroblastoma [32]. Thus, we performed Western blotting to explore whether FAM3D influences the mTORC1 pathway by modulating SESN2, which found that FAM3D KO reduced the phosphorylation of p70 S6K, a key marker of mTORC1 activation. This decrease was reversed by SESN2 knockdown. Conversely, FAM3D overexpression increased the phosphorylation of p70 S6K, which was suppressed by SESN2 overexpression (Figure 7C, 7D). These findings suggested that FAM3D promotes CRC progression by downregulating SESN2 and activating the mTORC1 pathway.

Figure 6. FAM3D functions in CRC cells by targeting SESN2 and activating the mTORC1 pathway. (A) The mRNA and protein levels of SESN2 in FAM3D-knockout LoVo cells with SESN2 knockdown. (B) The mRNA and protein levels of SESN2 in FAM3D-overexpressing LoVo cells with SESN2 overexpression. (C) Colony formation assay of FAM3D-knockout LoVo cells with SESN2 knockdown. (D) Colony formation assay of FAM3D-overexpressing LoVo cells with SESN2 overexpression. (E) CCK-8 assay of FAM3D-knockout LoVo cells with SESN2 knockdown. (F) CCK-8 assay of FAM3D-overexpressing LoVo cells with SESN2 overexpression. (G) A wound healing assay of FAM3D-knockout LoVo cells with SESN2 knockdown. (H) A wound healing assay of FAM3D-overexpressing LoVo cells with SESN2 overexpression. *P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001.

Figure 7. FAM3D functions in CRC cells by targeting SESN2 and activating the mTORC1 pathway. (A) The migration and invasion of FAM3D-knockout LoVo cells with SESN2 knockdown. (B) The migration and invasion of FAM3D-overexpressing LoVo cells with SESN2 overexpression. (C) Western blot analysis of FAM3D-knockout LoVo cells with SESN2 knockdown. (D) Western blot analysis of FAM3D-overexpressing LoVo cells with SESN2 overexpression. ****P < 0.0001.

FAM3D activates the mTORC1 pathway via ATF4-mediated downregulation of SESN2

Next, we explored how FAM3D suppresses the expression of SESN2. Previous studies have reported that both p53 and activating transcription factor 4 (ATF4) are transcription factors of SESN2 and can lead to its transcriptional activation in CRC [28, 33]. Thus, we analyzed ATF4 expression in our RNA-seq and GEO data. The results showed that ATF4 expression was significantly upregulated in FAM3D KO cells (Figure 8A) and was negatively correlated with FAM3D expression (rGSE39582 = -0.293 and rGSE106582 = -0.257) (Figure 8B, 8C), which was further confirmed by detecting ATF4 mRNA and protein (Figure 8D, 8E). Thus, we hypothesized that FAM3D reduces ATF4 expression by inhibiting its transcription. We performed a dual-luciferase reporter test. It was found that FAM3D caused a decrease in luciferase reporter activity regulated by the ATF4 promoter (Figure 8F, 8G), supporting our initial hypothesis. Further, Western blotting found that FAM3D KO significantly upregulated ATF4 protein level, consequently increasing SESN2 transcription and inhibiting mTORC1 (Figure 8H), while FAM3D overexpression played the opposite role (Figure 8I). These findings suggested that FAM3D promotes CRC progression through the ATF4-SESN2-mTORC1 pathway.

Figure 8. FAM3D activates the mTORC1 pathway via ATF4-mediated downregulation of SESN2 and reduces the chemosensitivity of CRC cells. (A) The mRNA expression of ATF4 in FAM3D KO and NC cells. (B, C) Correlations between ATF4 mRNA expression and FAM3D mRNA expression in (B) GSE39582 and (C) GSE106582. (D, E) The mRNA and protein levels of ATF4 in LoVo cells after FAM3D (D) KO or (E) overexpression. (F) Schematic description of the luciferase reporter. (G) Relative luciferase activity of the ATF4 wild type (WT) promoter in LoVo cells with knockout or overexpression of FAM3D. (H, I) Western blot analysis of LoVo cells with (H) knockout or (I) overexpression of FAM3D. (J) Estimated IC₅₀s of oxaliplatin, cisplatin and 5-fluorouracil in FAM3D-knockout and NC cells. (K) CCK-8 assay was used to detect the viability of FAM3D-knockout or FAM3D-overexpressing LoVo cells after treatment with a series of concentrations of oxaliplatin, cisplatin, and 5-fluorouracil. *P < 0.05; **P < 0.01, ***P < 0.001, ****P < 0.0001.

FAM3D reduces the chemosensitivity of CRC cells

Finally, we tested whether FAM3D has an impact on the chemosensitivity of CRC cells. Using our RNA-seq results, we calculated the predicted IC₅₀ concentrations of drugs or chemicals using the “oncoPredict” package (Supplementary Table 5). The results showed that FAM3D KO cells had lower IC₅₀ values for oxaliplatin, cisplatin, and 5-fluorouracil (Figure 8J). Consistently, the CCK-8 assay revealed that FAM3D KO increased, whereas FAM3D overexpression decreased CRC cell sensitivity to three chemotherapeutic drugs (Figure 8K).

Data acquisition

For the TCGA database, we obtained publicly available CRC data on DNA methylation, RNA expression, and clinical information from the UCSC (https://xena.ucsc.edu/). For the CPTAC database, we downloaded COAD protein data from cBioPortal (http://cbioportal.org/). In addition, we downloaded DNA methylation (GSE131013 and GSE101764) and RNA expression (GSE18105, GSE106582, GSE17536, GSE17537, GSE39582, GSE161158, GSE44076, and GSE132465) data of CRC patients from the GEO database (https://www.ncbi.nlm.nih.gov/geo/).

Cell culture

LoVo and LS513 cells were bought from the Cell Bank of Chinese Academy of Sciences. HT29, CW2, and HEK293T cells were acquired from Pricella Life Science & Technology Co., Ltd. LoVo, LS513, HT29, CW2, and HEK293T cells (used for generating lentivirus) were grown in Ham’s F-12K, RPMI-1640, McCoy’s 5A, DMEM, and DMEM containing 10% FBS (Procell). We cultured all cells in an incubator at 37° C and 5% CO₂.

RT-qPCR

After extraction with a Total RNA Kit I (Omega, R6834), Total RNA was quantified and cDNA was synthesized by ReverTra Ace® kit (Toyobo, FSQ201). The mRNA expression level was measured using SYBR green mix (Toyobo, QKD-201). Transcript levels were normalized and determined utilizing GAPDH and 2^-ΔΔCt methodology, correspondingly. The RT-qPCR primers used are listed in Supplementary Table 6.

CRISPR/Cas9-mediated gene knockout and lentivirus-mediated overexpression

To generate FAM3D gene KO CRC cells, a lentivirus-based CRISPR/Cas9 system was used. We used the CRISPick (https://portals.broadinstitute.org/gppx/crispick/public) to design two single guide RNA (sgRNA) sequences targeting the human FAM3D gene. Then, we synthesized double-stranded sgRNA through annealing, which was further cloned into the LentiCRISPR V2 vector (Miaolingbio, plasmid no. 52961), a third-generation lentiviral backbone that coexpresses CRISPR-associated protein 9 (Cas9) and sgRNA. The recombinant LentiCRISPR V2 vector, pMD2.G (Miaolingbio, plasmid no. 12259), and pSPAX2 (Miaolingbio, plasmid no. 12260) were co-transfected into HEK239T cells with Lipofectamine™ 2000 (Invitrogen) using the recommended protocol. After continuing to culture for 48 hours, the supernatant was filtered to collect lentivirus. Subsequently, LoVo and HT29 cells were infected with the collected lentivirus. Seventy-two hours post-infection, cells were cultured in complete medium containing puromycin (LoVo: 4 μg/ml; HT29: 2 μg/ml) for selection of stable FAM3D KO cells. About 14 days later, to generate monoclonal cells, CRC cells with stable knockout of FAM3D were plated into a 96-well plate at a concentration of one cell per well. Approximately 4 weeks later, single cell–derived clones were digested and transferred to a 24-well plate for further expansion to establish pooled populations of stable FAM3D KO cells. Finally, the efficiency of FAM3D KO in each clone was assessed using Sanger sequencing and Western blotting.

In addition, we established FAM3D-overexpressing LoVo cells via a lentiviral expression system. FAM3D mRNA and protein levels were detected to verify FAM3D overexpression efficiency.

Gene silencing

The negative control (sc-37007) and SESN2 (sc-106544) siRNAs were obtained from Santa Cruz Biotechnology. Lipofectamine™ 2000 was used to transfect siRNA at 25nM into LoVo cells. After incubating for an additional 48 hours, the knockdown efficiency of SESN2 was evaluated by detecting SESN2 mRNA and protein levels.

Western blotting

To extract total protein, cell lysis was performed using RIPA buffer (Beyotime, P0013B) containing phosphatase inhibitor (Roche, 4906837001) and 1% PMSF (Beyotime, ST506). After quantification, 30 μg of denatured total protein was transferred to the SDS-PAGE loading well and electrophoresed at 120V for about 70 minutes. Then transfer the proteins to the PVDF membrane at a current of 200mA for about 60 minutes. After blocking, the membrane was incubated successively with primary and secondary antibodies. Finally, the proteins were developed with ECL reagent and visualized with a Tanon 5200. Differences in protein expression levels were assessed using β-actin expression as a reference. The antibodies used for Western blotting analysis included FAM3D (Proteintech, 12336-1-AP, 1:6000), SESN2 (Proteintech, 10795-1-AP, 1:6000), total S6K (Proteintech, 14485-1-AP, 1:4000), phospho-S6K-Thr389 (Proteintech, 28735-1-AP, 1:6000), ATF4 (Proteintech, 10835-1-AP, 1:1000), and β-actin (OriGene, TA811000, 1:2000).

Detection of FAM3D protein in the cell culture supernatant

1 million cells were plated in Petri dishes. After culturing overnight, serum was withdrawn and starvation culture continued for 36 hours. The supernatant was collected and centrifuged to eliminate cells and fragments, after which the proteins were precipitated using 10% trichloroacetic acid (TCA) for 24 hours on ice. After centrifugation, the protein precipitates were washed for 3×5 minutes with acetone and dissolved in 80 μl 1×SDS-PAGE protein loading buffer. Protein levels were detected using Western blotting [53].

Cell viability assays

Cell growth and response to chemotherapy were evaluated using CCK-8 (Dojindo). To test proliferation, LoVo and HT29 cells in logarithmic growth phase were digested and counted, then diluted to 1.5 × 10⁵ cells per ml of complete medium. Mix thoroughly and inoculate into a 96-well plate (100 μl/well), place in an incubator, and continue culturing for 4 hours. Remove the 96-well plate from the incubator and detect cell proliferation, and then detect it every 24 hours for a total of 5 times; for the chemosensitivity test, after 24 hours of pre-incubation, LoVo cells were treated with different doses of cisplatin, oxaliplatin, and 5-fluorouracil for 48 h, after which cell survival was assessed. At specified intervals, after incubation with 10% CCK-8 reagent (LoVo: 2 hours; HT29: 1 hour), a microplate reader was used to detect the optical density of the reaction solution.

Transwell migration and invasion assays

Transwell chamber (Corning, 3422) was used to detect the in vitro motility of CRC cells. We resuspended 1 × 10⁵ cells in basal medium and transferred them to the top chamber, with 480 μl basal medium and 120 μl FBS in the bottom chamber. Following a 72-hour incubation, fixation was performed with 600 μl of 4% paraformaldehyde for 15 minutes. After rinsing, cells were stained with 600 μl of 0.1% crystal violet for an additional 15 minutes. Next, cells that did not pass through the holes were carefully wiped away, while the migrating cells were randomly photographed by fluorescence microscopy (LoVo: 160×magnification; HT29: 100×magnification) and counted using NIH ImageJ analysis software. For cell invasion assays, the membrane was precoated with Matrigel (BD Biosciences, 356234).

Colony formation assay

LoVo and HT29 cells were digested, quantified, and placed in a 6-well plate (2000 cells/well). After continuing to culture for 2-3 weeks, fix cells with 4% paraformaldehyde for 15 minutes, rinse twice, stain them with crystal violet for another 15 minutes, and finally rinse them with PBS 5-6 times. The plates were allowed to dry overnight. Clones that had at least 50 cells were counted.

Wound healing assay

Colorectal cancer cells were digested and counted, then inoculated into the 6-well plate at the same concentration and placed in an incubator to continue culturing. When the cells are all fused, use a 200-μl pipette tip to quickly make scratches. After rinsing three times, the serum was removed and starvation culture was performed. At 0, 24, and 48 hours after injury, images of the scratches were collected using fluorescence microscopy (40×magnification).

RNA-seq and data analysis

RNA was isolated from FAM3D KO and NC LoVo cells. Following an assessment of the RNA quality and concentration, cDNA libraries were synthesized and sequenced on an Illumina NovaSeq 6000 by Shanghai Bioprofile Technology Company Ltd. HISAT2 was used to map the clean data to Homo sapiens GRCh38, and transcript expression was normalized by the FPKM method. DEGs were defined as follows: 1, |log₂ fold change| ≥ 1; 2, corrected P-value < 0.05. The DEGs were then functionally annotated using GO, and pathway enrichment analysis was performed using KEGG.

In vivo tumorigenesis assay

Beijing Vital River Laboratory Animal Technology Co., Ltd. supplied sixteen BALB/c male nude mice, aged four weeks. 3×10⁶ LoVo cells were subcutaneously inserted into the center and back regions of the right axilla of naked mice for experimental xenograft research. Subsequently, we kept feeding the mice with caution and recorded the growth of the tumors and the body weight of the naked mice every 2-3 days. On the 19th day after the tumor was implanted, all of the naked mice were killed, and the tumor tissue was taken out and weighed. The formula for calculating tumor volume was 0.5 × width² × length.

Immunofluorescence

We seeded LoVo and HT29 cells on coverslips in a 6-well plate and cultured in an incubator for approximately 12 hours. And we discarded the supernatant and fixed cells with paraformaldehyde for 15 minutes. Discarding the paraformaldehyde, cell permeabilization was performed using 0.1% Triton X-100 for 10 minutes. After blocking on a shaker, we incubated cells with FAM3D primary antibody (Affinity, DF4832, 1:300) overnight at 4° C. Incubating cells with a CoraLite 594-conjugated secondary antibody (Proteintech, SA00013-4, 1:300) was performed for 60 minutes on a shaker in the dark. Then, cells on coverslips were stained using DAPI (Beyotime, C1005). Finally, we used fluorescence microscopy to capture blue and red fluorescence in the same field of view (160×magnification).

Dual-luciferase reporter assay

ATF4 promoter was subcloned and inserted a luciferase reporter plasmid (pGL3). After being cultivated for 12 hours, LoVo cells were co-transfected with the recombinant plasmid pGL3 and pGL4.75. After 48 hours, the experiments were carried out by a dual-luciferase reporter assay kit (Promega, E1910). Finally, we used the microplate reader to detect the signal intensity of luciferin.

Statistical analysis

Unless otherwise stated, three independent replicates of the experiment were performed to generate data. All analyzes were carried out using R 4.2.2. Variances among groups were estimated using one-way ANOVAs, Student’s t tests, Pearson and Spearman correlation tests, Wilcoxon rank sum tests, and χ² tests. Survival probabilities were determined using the Kaplan–Meier curve alongside the log-rank examination. Prognostically relevant variables were identified by Cox proportional hazards models. P < 0.05 was considered to be statistically different.