Higher TP53 mRNA expression is associated with worse clinical outcomes of paediatric neuroblastoma
We collected public paediatric neuroblastoma cohorts and designed a study process to determine the prognostic effects of TP53 and TP53-associated genes in paediatric neuroblastoma (Supplementary Fig. 1). First, the prognostic value of TP53 expression, the TP53 pathway and TP53-associated genes was determined using the TARGET, EMBL-EBI and GEO datasets. The independent prognostic factors in paediatric neuroblastoma were also determined. Finally, we also constructed a nomogram model to predict the overall survival of paediatric neuroblastoma based on age and TP53-associated genes.
In total, 152 paediatric neuroblastoma patients from the TARGET dataset along with 2325 paediatric neuroblastoma patients from the EMBL-EBI and GEO datasets, including the E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets, were collected (Fig. 1a). The number of paediatric neuroblastoma patients in each dataset and the gene microarray platform of each dataset are shown in Fig. 1a.
First, the prognostic effects of TP53 mRNA expression in paediatric neuroblastoma were determined. Univariate Cox regression analysis suggested that TP53 expression was associated with the event-free survival of paediatric neuroblastoma in the E-MTAB-161, E-MTAB-1781, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets (Fig. 1a). However, in the TARGET and E-MTAB-8248 datasets, TP53 mRNA expression was not significantly associated with the event-free survival of paediatric neuroblastoma (Fig. 1a). Moreover, TP53 expression was associated with the overall survival of paediatric neuroblastoma in all eight independent datasets: TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 (Fig. 1a).
Furthermore, paediatric neuroblastoma patients in each dataset were divided into TP53 higher and lower subgroups based on the mRNA expression levels of TP53. The different clinical outcomes of TP53 higher and lower subgroups were determined by Kaplan–Meier survival analysis. Lower TP53 expression was associated with prolonged event-free survival and overall survival of paediatric neuroblastoma in the TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets (Fig. 1b and Supplementary Fig. 2). These results suggested that TP53 mRNA expression was a critical prognostic factor of paediatric neuroblastoma.
A higher enrichment score of the TP53 signalling pathway is associated with worse clinical outcomes of paediatric neuroblastoma
Using the ssGSEA, we determined the enrichment score of the TP53 signalling pathway in each paediatric neuroblastoma patient. We found that the enrichment score of the TP53 signalling pathway was associated with the event-free survival and overall survival of paediatric neuroblastoma in the TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets (Fig. 2a). Moreover, paediatric neuroblastoma patients with lower enrichment scores of the TP53 signalling pathway had prolonged event-free survival and overall survival in the eight independent datasets (Fig. 2b and Supplementary Fig. 3).
The TP53-associated genes CCNE1, CDK2, CHEK2 and SESN1 are correlated with the clinical outcomes of paediatric neuroblastoma
In the GSEA dataset, the TP53 signalling pathway lists 68 TP53-associated genes. We then determined the prognosis of the TP53-associated genes in the TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets. The prognosis of these 68 genes is shown in the supplementary table. Four TP53-associated genes, CCNE1, CDK2, CHEK2 and SESN1, were all detected and were associated with the overall survival of paediatric neuroblastoma in the eight independent datasets (Fig. 3a). However, the prognostic effects of CCNE1, CDK2, CHEK2 and SESN1 were different in paediatric neuroblastoma. Higher expression of CCNE1, CDK2 or CHEK2 was an unfavourable prognostic factor, while higher expression of SESN1 was a favourable prognostic factor in paediatric neuroblastoma (Fig. 3a).
The Kaplan–Meier survival analysis further showed that overall survival was decreased in paediatric neuroblastoma patients with high CCNE1, CDK2 or CHEK2 expression (Fig. 3b and Supplementary Fig. 4). However, overall survival was increased in paediatric neuroblastoma patients with high SESN1 expression in the TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets (Fig. 3b and Supplementary Fig. 4).
TP53, CCNE1, CDK2, CHEK2 and SESN1 expression is correlated with MYCN amplification in neuroblastoma
Previous results showed that TP53 was a direct target of MYCN in paediatric neuroblastoma [21]. Therefore, we determined the associations of TP53 expression with MYCN amplification in paediatric neuroblastoma cohorts. We found that TP53 mRNA levels were upregulated in MYCN-amplified paediatric neuroblastoma patients in the TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets (Fig. 4a). Moreover, the enrichment score of the TP53 signalling pathway was also associated with MYCN amplification in paediatric neuroblastoma patients (Fig. 4b).
The mRNA expression levels of CCNE1, CDK2, CHEK2 and SESN1 in paediatric neuroblastoma patients with or without MYCN amplification were also investigated. Similar to TP53, CCNE1, CDK2 and CHEK2 expression levels were all upregulated in paediatric neuroblastoma patients with MYCN amplification (Fig. 4c). In contrast, SESN1 expression was downregulated in paediatric neuroblastoma patients with MYCN amplification in the TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 independent paediatric neuroblastoma cohorts (Fig. 4c). The results were consistent with overexpression of CCNE1, CDK2 or CHEK2 being worse prognostic factors, while increased regulation of SESN1 was a better prognostic factor in paediatric neuroblastoma.
CCNE1 and SESN1 are independent prognostic markers of neuroblastoma
Age at diagnosis and MYCN amplification are critical determinants of the clinical outcomes of paediatric neuroblastoma [5]. We then assessed the associations of age at diagnosis, MYCN amplification, TP53, CCNE1, CDK2, CHEK2 and SESN1 in the prediction of the overall survival of neuroblastoma using a multivariate Cox regression assay. We found that age at diagnosis was an independent prognostic factor of paediatric neuroblastoma in the E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets (Fig. 5). MYCN amplification was also an independent prognostic factor in the E-MTAB-161, E-MTAB-1781, GSE49710 and GSE85047 datasets (Fig. 5). However, TP53 was not an independent prognostic factor of paediatric neuroblastoma in any of the eight independent paediatric neuroblastoma cohorts (Fig. 5).
Moreover, CCNE1 was a prognostic factor of paediatric neuroblastoma in the E-MTAB-161, E-MTAB-1781, E-MTAB-8248, GSE16476 and GSE49710 datasets, independent of MYCN amplification and age at diagnosis (Fig. 5). CDK2 was also an independent prognostic factor of paediatric neuroblastoma in the E-MTAB-161, E-MTAB-1781 and GSE85047 datasets (Fig. 5). CHEK2 was an independent prognostic factor of paediatric neuroblastoma in the TARGET, E-TABM-38 and GSE49710 datasets (Fig. 5). SESN1 was an independent prognostic factor of paediatric neuroblastoma in the TARGET, E-TABM-38, GSE49710 and GSE85047 datasets (Fig. 5). Overall, age at diagnosis, MYCN amplification, CCNE1 and SESN1 were independent prognostic factors in at least four independent paediatric neuroblastoma cohorts.
Synergistic prognostic effects of CCNE1 with age at diagnosis in neuroblastoma
Since CCNE1 was a prognostic maker of neuroblastoma independent of age at diagnosis, the combinations of CCNE1 with age at diagnosis could achieve better prognostic effects in patients with paediatric neuroblastoma. First, in the TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE49710 and GSE85047 datasets, the expression levels of CCNE1 were higher in paediatric neuroblastoma patients with an age at diagnosis ≥ 18 months than in paediatric neuroblastoma patients with an age at diagnosis < 18 months (Fig. 6a).
Furthermore, based on the expression levels of CCNE1 and age at diagnosis, paediatric neuroblastoma patients in each dataset were divided into four subgroups. Paediatric neuroblastoma patients with an age at diagnosis < 18 months and with CCNE1 lower expression levels had significantly better prognosis and mostly survived in the following timeframe (Fig. 6b). In contrast, paediatric neuroblastoma patients with an age at diagnosis ≥ 18 months and with higher CCNE1 expression levels had a significantly worse prognosis (Fig. 6b). Paediatric neuroblastoma patients with an age at diagnosis ≥ 18 months and with CCNE1 lower expression levels had medium overall survival risks (Fig. 6b). Paediatric neuroblastoma patients with an age at diagnosis < 18 months and with CCNE1 higher expression levels had the most diverse prognosis than other subgroups (Fig. 6b).
Synergistic prognostic effects of SESN1 with age at diagnosis in neuroblastoma
Consistent with the previous results that overexpression of SESN1 was a favourable prognostic factor of paediatric neuroblastoma, the expression levels of SESN1 were lower in paediatric neuroblastoma patients with an age at diagnosis ≥ 18 months in the TARGET, E-MTAB-161, E-MTAB-1781, E-MTAB-8248, E-TABM-38, GSE16476, GSE49710 and GSE85047 datasets (Fig. 7a).
We also observed superior prognostic effects in the combinations of SESN1 with age at diagnosis in paediatric neuroblastoma. Paediatric neuroblastoma patients with an age at diagnosis < 18 months and SESN1 higher expression levels had a significantly better prognosis (Fig. 7b). In contrast, paediatric neuroblastoma patients with an age at diagnosis ≥ 18 months and SESN1 lower expression levels had a significantly worse prognosis (Fig. 7b). Paediatric neuroblastoma patients with an age at diagnosis ≥ 18 months and SESN1 higher expression levels had medium overall survival risks (Fig. 7b). Paediatric neuroblastoma patients with an age at diagnosis < 18 months and SESN1 lower expression levels had the most diverse prognosis than other subgroups (Fig. 7b).
Construction of a nomogram model to predict the overall survival of paediatric neuroblastoma based on age and TP53, CCNE1, CDK2, CHEK2 and SESN1 expression
Our results suggested that TP53 and its associated genes CCNE1, CDK2, CHEK2 and SESN1 were all associated with the overall survival of paediatric neuroblastoma. We then constructed a nomogram model based on age at diagnosis and TP53, CCNE1, CDK2, CHEK2 and SESN1 expression features to predict the clinical overall survival of paediatric neuroblastoma (Fig. 8a). The risk point of each paediatric neuroblastoma patient in the E-MTAB-161, E-MTAB-8248, E-TABM-38, GSE16476 and GSE49710 datasets was obtained in the nomogram model. Paediatric neuroblastoma in the lower risk subgroup had significantly longer overall survival (Fig. 8b). Moreover, the ROC analysis in the E-MTAB-161, E-MTAB-8248, E-TABM-38, GSE16476 and GSE49710 datasets indicated that the nomogram model could predict the three-year, five-year or ten-year overall survival of paediatric neuroblastoma with high specificity and sensitivity (Fig. 8c).
The nomogram model could predict the overall survival of MYCN nonamplified paediatric neuroblastoma
We previously showed that the expression levels of TP53, CCNE1, CDK2, CHEK2 and SESN1 were correlated with MYCN amplification in neuroblastoma patients. We further tested whether the TP53-based nomogram model could discriminate paediatric neuroblastoma cases with unfavourable outcomes among MYCN nonamplified patients. MYCN nonamplified patients from the TARGET, E-MTAB-161, E-MTAB-8248, E-TABM-38, GSE16476 and GSE49710 datasets were selected for investigation. MYCN nonamplified paediatric neuroblastoma patients with higher risk points had significantly shortened overall survival (Supplementary Fig. 5). Moreover, the ROC analysis in the TARGET, E-MTAB-161, E-MTAB-8248, E-TABM-38, GSE16476 and GSE49710 datasets showed that the nomogram model could predict the overall survival of MYCN nonamplified paediatric neuroblastoma (Supplementary Fig. 5).