Introduction
Neuroblastoma, an embryonic tumor originating from neural crest cells, represents a significant concern in pediatric oncology, being responsible for over 15% of childhood cancer-related fatalities. In the United States alone, around 650 new cases of neuroblastoma are diagnosed annually, underscoring the urgent need for continued research, early detection methods, and improved treatment options for this challenging disease. However, due to its distinct heterogeneity in genetic, clinical, and morphological aspects, the efficacy of current treatment modalities for neuroblastoma remains limited.1 Given this complexity, it is imperative to explore safer and more effective treatment approaches grounded in the genetic variations and molecular signatures of neuroblastoma.
5-methylcytosine (m5C) is a prevalent RNA modification found in many RNA species, including ribosomal RNAs, transfer RNAs (hereinafter referred to as tRNAs), messenger RNAs (mRNAs), enhancer RNAs, and noncoding RNAs, playing a pivotal role in numerous aspects of both nuclear gene and mitochondrial expression. Its functions encompass RNA stability, export, ribosome assembly, and translation regulation. Remarkable advancements have been made in elucidating the molecular interactions between m5C writers, demethylases, readers, and RNAs. Among these, DNA methyltransferases (DNMTs) and NOP2/Sun domain family members (NSUNs) function as m5C writers, catalyzing the methylation of mRNA, while ten-eleven translocation proteins (TETs) serve as m5C demethylases, primarily responsible for removing m5C modifications from mRNA. Additionally, proteins such as alpha-ketoglutarate-dependent dioxygenase ALKB homolog 1 (ALKBH1), RNA-binding protein ALY/REF export factor (ALYREF), and Y-Box binding protein 1 (YBX1) are capable of recognizing m5C modifications, acting as m5C-binding proteins.2 Notably, m5C RNA methylation has emerged as a significant player in the progression of various cancers, spanning a range of tumor types, including breast cancer,3 glioma,4 prostate cancer, and neuroblastoma.5,6 Given the positive association between m5C and cancer migration and metastasis, targeting m5C for cancer therapy has emerged as a promising avenue of research.
NSUN6 functions as an mRNA methyltransferase, specifically catalyzing the methylation of 5-methylcytidine (m5C). Its role in mRNA methyltransferase activity and translation termination within the context of m5C modification is crucial. The activity of mRNA and translation processes is enhanced upon NSUN6 engagement with m5C-containing mRNAs. Previous research has indicated low expression of NSUN6 in human tumors, suggesting that high NSUN6 expression may serve as an indicator of better outcomes in specific tumors.7 In a study by Yang et al.,8 the association between pancreatic cancer and m5C-related genes was analyzed using quantitative polymerase chain reaction (PCR) and immunohistochemistry. Notably, NSUN6 demonstrated promising performance in evaluating patient survival and tumor recurrence in pancreatic cancer cases.8
To identify novel susceptibility factors associated with m5C modification—with a specific focus on single-nucleotide polymorphisms (SNPs) of the key m5C modification-related gene NSUN6 and its neighboring gene calcium voltage-gated channel auxiliary subunit beta 2 (CACNB2), both linked to neuroblastoma—we conducted a comprehensive case-control study. Our investigation unveiled significant associations between NSUN6 and its neighboring gene CACNB2 SNPs and the risk of neuroblastoma among children in Jiangsu province.
Materials and methods
Study subjects
This study was a case-control study, conducted at the Children’s Hospital of Nanjing Medical University (Nanjing, Jiangsu, China) with ethical approval (No. 202112141-1) and adherence to the Declaration of Helsinki (as revised in 2024), from 2021 to 2023. Guardians provided written informed consent. Cases (n = 402): Children with histologically confirmed neuroblastoma (2021–2023), staged via the International Neuroblastoma Staging System, with complete clinical data (age, sex, tumor origin) and no prior chemotherapy/radiotherapy or recurrent disease. Controls (n = 473): Age- and sex-matched cancer-free children undergoing routine health checks (same period), with no genetic syndromes or chronic diseases. Exclusion criteria: Incomplete clinical/genetic data, prior immunosuppressive therapy, or refusal to provide samples. Our experimental flow chart is shown in Figure 1.
SNP selection and genotyping
SNP selection
SNPs were chosen based on: (1) localization in functional regions (e.g., introns with transcription factor binding sites); (2) minor allele frequency > 5% in Chinese Han populations (1000 Genomes Project); (3) linkage disequilibrium R2< 0.8 (Haploview 4.2).
Ultimately, three SNPs met these criteria (rs3740102 A>C, rs12780826 T>A, and rs61842187 G>C). The rs3740102 A>C, rs12780826 T>A, and rs61842187 G>C are located in the intron region of the NSUN6 gene. The presumed function of these SNPs involves binding with transcription factors.
DNA extraction & genotyping
DNA extraction: Peripheral blood genomic DNA was isolated using the TIANamp Genomic/Blood DNA Kit (TianGen, Beijing, China) per the manufacturer’s protocol. TaqMan assay: Genotyping was performed via TaqMan Genotyping PCR on an Applied Biosystems 7500 Fast Real-Time PCR System. Reaction system (10 µL): 5 µL 2× Master Mix, 0.25 µL 20× SNP Assay, 2 µL DNA (50 ng/µL), 2.75 µL nuclease-free water. Cycling conditions: 95°C for 10 m; 40 cycles of 95°C for 15 s, 60°C for 1 m. We employed the TaqMan SNP genotyping method to analyze these three SNPs within both the case and control groups, with 10% of samples randomly chosen for method repetition. Secondary genotyping confirmed 100% consistency.
Expression quantitative trait locus (eQTL) analysis (the Genotype-Tissue Expression (GTEx) database)
eQTL analysis used the GTEx Portal (v9). Relevant tissues (adrenal gland, tibial nerve, whole blood) were selected (linked to neuroblastoma pathogenesis). Associations between SNPs and target gene expression were extracted, including nominal P-values and expression fold changes.
Statistical analysis
The cases and controls were divided into different subgroups based on their age, sex, tumor origin, and International Neuroblastoma Staging System stage. We utilized the chi-square test to analyze the discrete variables and the t-test to analyze the continuous variables like age range. The goodness-of-fit chi-square test was employed to assess the Hardy-Weinberg equilibrium in controls. Multiple logistic regression analysis was conducted to calculate the odds ratio (OR) value and 95% confidence interval (CI) value after adjusting for age and sex. All of these analyses were conducted using SAS 9.4, with a significance level set at P < 0.05. We investigated the effect of rs3740102, rs12780826, and rs61842187 on NSUN6 or nearby genes by querying the GTEx online website. Furthermore, the expression levels of NSUN6 and CACNB2 were analyzed by the R2: Genomics Analysis and Visualization Platform (http://r2.amc.nl ). The relationship between NSUN6/CACNB2 and overall survival (OS) and event-free survival (EFS) of neuroblastoma patients was investigated using the Sanger-box platform. Moreover, non-parametric testing was utilized to analyze differences in gene expression between two groups: those classified as high risk or without high risk of neuroblastoma, and those with or without MYCN proto-oncogene, BHLH transcription factor (MYCN) amplification.
Results
Association study
This study recruited 402 cases and 473 controls, all of whom were successfully genotyped. Complete demographic information for all participants is provided in Table S1. Relevant information can also be found in our previous research.9,10 All of these gene polymorphisms (rs3740102 A>C, rs12780826 T>A, and rs61842187 G>C) conformed to Hardy-Weinberg equilibrium (P > 0.05) (Table 1). No significant association was found between neuroblastoma and the individual genotypes (rs3740102 AA/AC, rs12780826 TT/TA, and rs61842187 CC). However, compared to subjects with zero to one risk genotype, those with two to three risk genotypes were observed to have an association that may be related to lower neuroblastoma risk (2–3 vs. 0–1: adjusted OR = 0.41, 95% CI = 0.23–0.73, P = 0.002).
Table 1Association of NSUN6 gene polymorphisms with neuroblastoma risk in children from Jiangsu province
| Genotype | Cases (n = 402) | Controls (n = 473) | Pa | Crude OR (95% CI) | P | Adjusted OR (95% CI)b | Pb |
|---|
| rs3740102 A>C (HWE = 0.559) |
| AA | 216 (53.73) | 242 (51.16) | | 1.00 | | 1.00 | |
| AC | 141 (35.07) | 189 (39.96) | | 0.84 (0.63–1.11) | 0.218 | 0.84 (0.63–1.11) | 0.218 |
| CC | 45 (11.19) | 42 (8.88) | | 1.20 (0.76–1.90) | 0.435 | 1.20 (0.76–1.90) | 0.436 |
| Additive | | | 0.955 | 0.99 (0.81–1.21) | 0.955 | 0.99 (0.81–1.22) | 0.955 |
| Dominant | 186 (46.27) | 231 (48.84) | 0.448 | 0.90 (0.69–1.18) | 0.449 | 0.90 (0.69–1.18) | 0.447 |
| AA/AC | 357 (88.81) | 431 (91.12) | | 1.00 | | 1.00 | |
| CC | 45 (11.19) | 42 (8.88) | 0.254 | 1.29 (0.83–2.02) | 0.255 | 1.30 (0.83–2.02) | 0.254 |
| rs12780826 T>A (HWE = 0.327) |
| TT | 316 (78.61) | 357 (75.48) | | 1.00 | | 1.00 | |
| TA | 71 (17.66) | 105 (22.20) | | 0.76 (0.55–1.07) | 0.117 | 0.76 (0.55–1.07) | 0.118 |
| AA | 15 (3.73) | 11 (2.33) | | 1.54 (0.70–3.40) | 0.285 | 1.54 (0.70–3.41) | 0.285 |
| Additive | | | 0.613 | 0.93 (0.72–1.22) | 0.614 | 0.93 (0.72–1.22) | 0.614 |
| Dominant | 86 (21.39) | 116 (24.52) | 0.273 | 0.84 (0.61–1.15) | 0.274 | 0.84 (0.61–1.15) | 0.274 |
| TT/TA | 387 (96.27) | 462 (97.67) | | 1.00 | | 1.00 | |
| AA | 15 (3.73) | 11 (2.33) | 0.222 | 1.63 (0.74–3.59) | 0.226 | 1.63 (0.74–3.60) | 0.226 |
| rs61842187 G>C (HWE = 0.157) |
| GG | 225 (55.97) | 257 (54.33) | | 1.00 | | 1.00 | |
| GC | 153 (38.06) | 175 (37.00) | | 1.00 (0.75–1.32) | 0.992 | 1.00 (0.75–1.32) | 0.988 |
| CC | 24 (5.97) | 41 (8.67) | | 0.67 (0.39–1.14) | 0.140 | 0.67 (0.39–1.14) | 0.139 |
| Additive | | | 0.312 | 0.90 (0.73–1.11) | 0.312 | 0.90 (0.72–1.11) | 0.309 |
| Dominant | 177 (44.03) | 216 (45.67) | 0.628 | 0.94 (0.72–1.22) | 0.628 | 0.94 (0.72–1.22) | 0.626 |
| GG/GC | 378 (94.03) | 432 (91.33) | | 1.00 | | 1.00 | |
| CC | 24 (5.97) | 41 (8.67) | 0.129 | 0.67 (0.40–1.13) | 0.132 | 0.67 (0.40–1.13) | 0.131 |
| Combine protective genotypesc |
| 0–1 | 37 (9.20) | 19 (4.02) | | 1.00 | | 1.00 | |
| 2–3 | 365 (90.80) | 454 (95.98) | 0.002 | 0.41 (0.23–0.73) | 0.002 | 0.41 (0.23–0.73) | 0.002 |
Stratified analysis
In the context of further analyzing the influence of NSUN6 gene polymorphisms on neuroblastoma risk, we stratified our variables by age, sex, site of origin, and clinical stage for stratification (Table 2). All comparisons below refer to the contrast between two to three protective genotypes and zero to one protective genotypes. Significant risk reductions were observed in the following subgroups: ≤ 18-month subgroup (adjusted OR = 0.11, 95% CI = 0.03–0.49, P = 0.004), the male subgroup (adjusted OR = 0.25, 95% CI = 0.09–0.70, P = 0.008), the III + IV clinical stages (adjusted OR = 0.26, 95% CI = 0.14–0.49, P < 0.0001), the adrenal gland origin subgroup (adjusted OR = 0.39, 95% CI = 0.17–0.90, P = 0.027), the retroperitoneal subgroup (adjusted OR = 0.39, 95% CI = 0.20–0.79, P = 0.008), and others (adjusted OR = 0.20, 95% CI = 0.05–0.77, P = 0.019).
Table 2Stratification analysis for the association between NSUN6 genotypes and neuroblastoma susceptibility in Jiangsu children
| Variables | rs61842187 (cases/controls)
| Adjusted ORa (95% CI) | Pa | Protective genotypes (cases/controls)
| Adjusted ORa (95% CI) | Pa |
|---|
| GG/GC | CC | 0–1 | 2–3 |
|---|
| Age, month |
| ≤18 | 134/127 | 5/12 | 0.40 (0.14–1.15) | 0.090 | 16/2 | 123/137 | 0.11 (0.03–0.49) | 0.004 |
| >18 | 244/305 | 19/29 | 0.82 (0.45–1.50) | 0.515 | 21/17 | 242/317 | 0.62 (0.32–1.20) | 0.154 |
| Gender |
| Females | 178/205 | 13/20 | 0.75 (0.36–1.55) | 0.436 | 21/14 | 170/211 | 0.54 (0.27–1.09) | 0.084 |
| Males | 200/227 | 11/21 | 0.59 (0.28–1.26) | 0.176 | 16/5 | 195/243 | 0.25 (0.09–0.70) | 0.008 |
| Sites of origin |
| Adrenal gland | 87/432 | 6/41 | 0.72 (0.30–1.75) | 0.467 | 9/19 | 84/454 | 0.39 (0.17–0.90) | 0.027 |
| Retroperitoneal | 154/432 | 13/41 | 0.89 (0.46–1.71) | 0.726 | 16/19 | 151/454 | 0.39 (0.20–0.79) | 0.008 |
| Mediastinum | 115/432 | 5/41 | 0.46 (0.18–1.19) | 0.108 | 7/19 | 113/454 | 0.67 (0.27–1.64) | 0.378 |
| Others | 18/432 | 0/41 | / | / | 3/19 | 15/454 | 0.20 (0.05–0.77) | 0.019 |
| Clinical stages |
| I + II + 4s | 164/432 | 9/41 | 0.57 (0.27–1.20) | 0.137 | 9/19 | 164/454 | 0.83 (0.37–1.89) | 0.660 |
| III + IV | 149/432 | 14/41 | 0.99 (0.52–1.87) | 0.969 | 22/19 | 141/454 | 0.26 (0.14–0.49) | <0.0001 |
Functional effects of rs3740102 AA/AC, rs12780826 TT/TA, and rs61842187 CC genotypes on nearby genes
Given the significant impact of these protective genotypes on neuroblastoma risk, we conducted eQTL analyses using the GTEx database for the variants rs3740102 AA/AC, rs12780826 TT/TA, and rs61842187 CC (Fig. 2). The rs3740102 C>A variant demonstrates an upregulation of the NSUN6 gene expression in cultured fibroblast cells (Fig. 2a), adrenal gland (Fig. 2b), and whole blood (Fig. 2c). Additionally, the A allele of rs3740102 is associated with enhanced expression of the CACNB2 gene in whole blood (Fig. 2d), as well as increased expression of genes such as RP11-499P20.2 (Fig. 2e and f) and ARL5B (Fig. 2g and h). The rs61842187 G>C variant upregulates NSUN6 gene expression in cultured fibroblast cells (Fig. 2i), while downregulating ARL5B gene expression in whole blood (Fig. 2j). Furthermore, the rs61842187 G>C variant is correlated with decreased expression of RP11-499P20.2 (Fig. 2k) and enhanced expression of the CACNB2 gene in the tibial artery (Fig. 2l). The rs12780826 A allele is significantly associated with increased expression of the NSUN6 gene in cultured fibroblast cells (Fig. 2m), and elevated expression of the CACNB2 gene in the tibial artery (Fig. 2n). Conversely, the rs12780826 A allele is linked to decreased expression of the ARL5B (Fig. 2o) and RP11-499P20.2 genes in whole blood (Fig. 2p).
Functional annotation of NSUN6 and CACNB2 genes in neuroblastoma
Since the polymorphisms rs3740102 C>A, rs12780826 T>A, and rs61842187 G>C have significant effects on NSUN6 mRNA expression, we further investigated the relationship between NSUN6 mRNA levels and neuroblastoma prognosis using the R2: Genomics Analysis and Visualization Platform. Kaplan–Meier survival analysis was employed to predict the OS and EFS in neuroblastoma cases from the GSE62564 dataset (Fig. 3). Our analysis revealed that higher NSUN6 mRNA expression (n = 249) is associated with better OS (n = 249; P = 2.54e-03; Fig. 3a) and EFS (n = 249; P = 7.90e-04; Fig. 3b) compared to neuroblastoma patients with lower NSUN6 mRNA expression. Similarly, elevated CACNB2 mRNA expression is associated with better prognosis, as evidenced by better OS (n = 249; P = 6.35e-06; Fig. 3c) and EFS (n = 249; P = 4.64e-06; Fig. 3d). Box plots were used to illustrate differences in NSUN6 and CACNB2 expression between high-risk and low-risk groups, as well as between patients with and without MYCN amplification. The results indicated that increased NSUN6 expression is associated with a potential lower risk of neuroblastoma (Fig. 4a) and the absence of MYCN amplification (Fig. 4b). Likewise, a higher level of CACNB2 is associated with a potential lower risk of neuroblastoma (Fig. 4c) and the absence of MYCN amplification (Fig. 4d).
Discussion
It is challenging to devise a universal therapy for high-risk neuroblastoma patients due to the specific heterogeneity of their biological, morphological, clinical, and genetic characteristics.11 With advancements in understanding the genomic aberrations and disrupted pathways associated with neuroblastoma, genome-wide association studies, high-throughput genome analysis, transcriptional profiling, and genome sequencing offer new therapeutic targets to enhance patient survival rates. Novel molecular treatments may be developed to target these genomic aberrations and disrupted pathways directly.1 However, conducting genome-wide association studies using Bonferroni correction may result in underpowered heritability estimates.12 In our previous studies, we elucidated the relationship between neuroblastoma and m5C methyltransferase genes, including TET2,10TET1,9ALKBH1 rs2267755,13 and NSUN2 rs13181449.14 As a type of m5C enzyme, NSUN6 exhibits strong substrate specificity, enabling it to install m5C72 by targeting tRNACys/Thr.7,15 However, the biological functions and mechanisms of NSUN6 for m5C modification are still unknown.16 Accordingly, we carried out this study to elucidate the correlation between neuroblastoma and the NSUN6 gene.
Our study demonstrated that NSUN6 rs3740102 AA/AC, rs12780826 TT/TA, and rs61842187 CC genotypes were associated with protective effects. Compared to individuals with zero to one protective genotype, those with a combination of two to three protective genotypes were observed to have an association that may be related to lower neuroblastoma risk. Even after adjusting for age, gender, tumor origin sites, and clinical stages, the combination of two to three NSUN6 protective genotypes remained associated with lower neuroblastoma susceptibility. Additionally, stratified analysis showed that this association was notably more significant across all subgroups, and this result was supported by Bonferroni correction. To verify the effect of NSUN6 rs3740102, rs12780826, and rs61842187 genotypes, we conducted eQTL analyses in the GTEx database to investigate the functional effect of rs3740102 AA/AC, rs12780826 TT/TA, and rs61842187 CC genotypes on nearby genes. Additionally, we utilized Kaplan-Meier survival analysis to explore the functional annotation of NSUN6 and CACNB2 genes in neuroblastoma. Ultimately, the results indicated that higher expression of NSUN6 and CACNB2 genes was associated with better prognosis, low risk, and non-MYCN amplification.
MYCN amplification accounts for 20–25% of neuroblastoma patients, often resulting in poor prognosis and a significant proportion of cancer-related deaths.17 Our research found that higher expression of NSUN6 always accompanied lower MYCN amplification, which is characteristic of low risk and good prognosis. Recently, NSUN6 has shown survival benefits among glioblastoma patients and in other cancers.7,8,18,19 Our study also confirmed that higher expression of the NSUN6 gene can improve the prognosis of neuroblastoma. Selmi T and his colleagues revealed that in certain cancer types, high expression of NSUN6 in mice indicated a better patient survival rate.7
In lung cancer, Lu et al.20 found that NM23-H1 was expressed at low levels while NSUN6 was overexpressed. Further testing confirmed that high levels of NSUN6 can regulate NM23-H1 in an m5C-dependent manner to inhibit tumor proliferation, migration, and epithelial-mesenchymal transition.20 Temozolomide is a type of therapy for glioblastoma cancer. Awah et al.18 presented evidence that NSUN6 can influence the response to Temozolomide therapy through NELFB and RPS6BK2, which are regulated by m5C in glioblastoma cancer. NSUN6 rs12780826 T>A may cause the variant of intron and upstream transcript; NSUN6 rs3740102 C>A, rs61842187 G>C may lead to variants in the 5’ UTR, intron, and upstream transcript. However, there are no reports detailing the clinical consequences of these three SNPs.21 In this study, we found that the combination of NSUN6 rs3740102 AA/AC, rs12780826 TT/TA, and rs61842187 CC genotypes is associated with a potential lower neuroblastoma risk. Further analyses indicated that higher NSUN6 gene expression is correlated with a potential lower neuroblastoma risk and the absence of MYCN amplification. Overall, the potential association of these genotypes with neuroblastoma prognosis may be related to the following: the A allele of rs3740102 A>C, A allele of rs12780826 T>A, and C allele of rs61842187 C>G may be linked to the potential upregulation of NSUN6 mRNA expression and a lower likelihood of MYCN amplification, and this expression pattern (higher NSUN6, lower MYCN amplification) is in turn associated with better neuroblastoma prognosis. As a member of the voltage-gated superfamily characteristic of calcium channels, CACNB2 can encode a subunit of channel protein in voltage-dependent calcium channels. Recent studies revealed that CACNB2 plays an important role in many cancers, including breast cancer,22 lung cancer, and gastric cancer.23,24 Masuelli et al.22 demonstrated that in left-side breast tumors, CACNB2 can participate in a 6-ion channel-gene signature. This signature, comprising six ion channel genes, was inversely correlated with DNA methylation writers and cancer progression markers such as stemness and proliferation. Our research suggests that NSUN6 rs3740102 C>A, rs61842187 G>C, and rs12780826 T>A may be associated with the potential upregulation of CACNB2 expression; this potential upregulation of CACNB2 expression is in turn associated with a potential lower neuroblastoma risk, the absence of MYCN amplification, and better patient prognosis.
This is the inaugural revelation of the association between NSUN6 and the risk of neuroblastoma, and we have also clarified the potential mechanism. Our survey encompassed a large sample from Jiangsu, China, enhancing the credibility of our study. However, there are some limitations. Firstly, our subjects were recruited only from Nanjing, China, which may introduce sampling error and reduce the credibility of extending our conclusions. Secondly, being a case-control study, further animal trials will be necessary to validate and elucidate the potential mechanism.25
In future studies, further functional experiments are needed to clarify the regulatory mechanisms of NSUN6 SNPs on CACNB2 expression and MYCN amplification. Multi-center, multi-ethnic population validations should be conducted to confirm the generalizability of our findings. Additionally, exploring the potential of NSUN6/CACNB2-related markers in clinical risk stratification and personalized treatment may provide new insights for neuroblastoma management.
Conclusions
Our study found NSUN6 rs3740102 AA/AC, rs12780826 TT/TA, and rs61842187 CC genotypes can improve the prognosis of neuroblastoma by increasing the expression NSUN6 gene, while decreasing the amplification of MYCN. This conclusion has implications for prognosis prediction and suggests avenues for further mechanistic studies in neuroblastoma.
Supporting information
Supplementary material for this article is available at https://doi.org/10.14218/CSP.2025.00019 .
Table S1
Demographic characteristics of neuroblastoma patients and cancer-free controls from Jiangsu province.
(DOCX)
Declarations
Ethical statement
This study was conducted in accordance with the Declaration of Helsinki (as revised in 2024). Approval was obtained from the institutional review board of Guangzhou Women and Children’s Medical Center (Ethical Approval No: 202016601) and the Children’s Hospital of Nanjing Medical University (Approval No: 202412006-1). Guardians provided written informed consent.
Data sharing statement
All the data were available upon request from the corresponding authors.
Funding
This study received support from grants provided by the National Natural Science Foundation of China (Grant No. 81973063).
Conflict of interest
The authors have declared that they have no competing interests.
Authors’ contributions
Conceptualization (ZZ), methodology (SJ, YS, CZ), formal analysis and investigation (SJ, YH, CZ), writing - original draft preparation (SJ), writing - review and editing (JH), funding acquisition (ZZ), resources (HW, WZ), and supervision (ZZ, CZ). All authors have approved the final version and publication of the manuscript.