| Background:Pheochromocytomas and paragangliomas(PPGL)are a potential cause of endocrine hypertension,and the uncontrolled secretion of catecholamine hormones by PPGL tumours can lead to severe cardiovascular manifestations,including multiple endorgan damage and hypertensive crisis.Furthermore,although surgery is the treatment of choice for PPGL,there is no effective treatment for patients with PPGL who develop systemic metastases or for whom surgery is more difficult.Therefore,exploring appropriate laboratory tests and screening strategies to accurately identify and diagnose PPGL,and to uncover potential mechanisms for the treatment of PPGL,requires further investigation.The current focus of biochemical screening for PPGL includes three catecholamines(CAs):epinephrine(E),norepinephrine(NE)and dopamine(DA)and their corresponding catecholamine metabolites(Metanephrines(MNs):metanephrine(MN),methoxynorepinephrine(NMN)and methoxytyramine(3-MT).However,there are no Chinese standards for reference intervals(RI)for CAs and MNs,and the choice of optimal diagnostic cut-offs and combinations of CAs and MNs in the diagnosis of PPGL remains controversial in different studies.Also,the levels of MNs may provide additional information about the nature of these tumours in addition to reflecting the presence or absence of PPGL,yet there are few supporting studies.Furthermore,tumour stem cell characteristics have been widely shown to be associated with the development and progression of a variety of tumours,while their study in PPGL is still lacking.Therefore,this study firstly proposes to establish RI and diagnostic models for CAs and MNs by mining patient data from the Chinese population.Secondly,by analysing the relationship between MNs and tumour tissue characteristics and genetic features in PPGL,we will investigate whether MNs can be used to reflect key tumour characteristics and guide clinical optimisation of screening procedures.Finally,the tumour stem cell characteristics of PPGL were analysed to explore their associated survival and treatment prognosis related markers.Methods:This study was based on a dataset of apparently healthy subjects,using three different direct method algorithms:transformed parametric,non-parametric and robust methods,to establish direct method RIs for plasma MNs,24-hour urine MNs and CAs,respectively.Two different indirect algorithms were used:the refine R algorithm and the Hoffman algorithm,which are based on the data set of "real-world mixed samples" and"real-world screened samples" derived from the clinical data management system without PPGL diagnosis and exclusion.The gender differences in the upper limit of the RI for each indicator were determined by gender bias ratio analysis of the RI.Next,a receiver operating characteristic curve(ROC)analysis.grid search and various machine learning algorithms were used in the training dataset of the PPGL Diagnostic Study consisting of patients with a definite PPGL diagnosis and those with a definite PPGL diagnosis excluded.The optimal diagnostic cut-off values for each indicator of the ROC source and the multi-indicator association diagnostic model were developed using receiver operating characteristic(ROC)analysis,grid search models and various machine learning algorithms.Then,in the validation dataset of the PPGL Diagnostic Study,the diagnostic performance of the upper RI limit for each indicator,the optimal diagnostic threshold for each indicator from the ROC source,the diagnostic threshold for each indicator used by the Mayo laboratory and the multiple indicator model were compared,and the optimal diagnostic models for plasma MNs,24-hour urine MNs and CAs were identified.The best diagnostic model for CAs was identified.In this study,the relationship between the secretion characteristics of MNs and the synthesis characteristics of CAs in PPGL tumours was analysed by in situ mass spectrometry,and the best indicators and models for predicting the location and size of PPGL tumours were established and validated by ROC analysis and KNN algorithm.The patients were also screened for mutations in PPGL susceptibility genes carried by 237 PPGL patients based on second-generation sequencing technology,and classified into two catecholaminesecreting phenotypes(noradrenergic and adrenergic)based on the proportion of elevated blood MN,and then assessed the predictive effect of different catecholamine-secreting phenotypes on mutations in PPGL susceptibility genes.Finally,based on the transcriptome data of 186 PPGLs collected in The Cancer Genome Atlas(TCGA)database,a stemness index(mRNA expression-based stemness index)reflecting the strength of stemness characteristics of tumour cells was calculated using the OCLR machine learning algorithm-(mRNAsi).The relationship between mRNAsi and PPGL susceptibility gene mutation status,clinical and pathological characteristics was assessed,while survival analysis and multiple immune assessment algorithms were used to compare the differences in survival outcomes and immune microenvironment characteristics between high and low mRNAsi groups.This was followed by screening for differential genes between the high and low mRNAsi groups and using weighted gene co-expression network analysis to identify a pivotal gene cluster that best reflects the stemness profile of PPGL.Pathway enrichment analysis and target drug prediction were performed for this key gene cluster.In addition,based on this key gene cluster,a stemness risk model was developed to predict the prognosis of PPGL survival by LASSO-COX regression analysis,and univariate Cox analysis and ROC curves were used to assess the accuracy of the prediction.Finally,differences in immune cell infiltration,immune checkpoint expression,immune scores,and responsiveness to immunotherapy were compared between high and low stemness risk groups.Results:There were gender differences in the upper RI limits for plasma MNs,24-hour urine MNs and CAs,and the upper RI limits were significantly higher in men than in women.In addition,the upper RI limits for the Hoffman method were higher than those for the other methods,except for 24-hour urine 3-MT,which was based on the "indirect real-world screening sample" dataset.For 24-hour urine 3-MT,the upper RI limit was highest for the non-parametric method.The RI source achieved the best diagnostic performance for each indicator compared to the ROC source and the Mayo laboratory source.The 24-hour urine NMN(diagnostic threshold:55.36 mcg/24h)was the best way to diagnose PPGL(sensitivity:89.5%;specificity:99.7%;diagnostic agreement:98.0%)relative to each single-indicator or multiple-indicator combination model.In situ mass spectrometry imaging results showed that the secretion pattern of MNs in peripheral blood of PPGL was consistent with the secretion profile of CAs within its tumour tissue.Tumours were more likely to be located in the adrenal glands when the relative elevation of plasma MNs exceeded 2.3%(sensitivity:44.8%;specificity:87.3%)or 6.1%(sensitivity:44.7;specificity:86.5%)in PPGL patients.In addition,the catecholamine secretion phenotype could be used to identify mutations in PPGL susceptibility genes,with an accuracy of 70.4%in identifying mutations in RET,NF1 and TMEM127,and 70.4%in identifying mutations in SDH.Analysis of PPGL tumour stemness profiles revealed that higher mRNAsi may be associated with tumour metastasis in SDHB wildtype PPGL patients,as well as with lower immune and stromal scores and a suppressive tumour immune microenvironment.Correspondingly the pathway enrichment results likewise suggest that the gene cluster associated with the tumour stemness profile of PPGL is mainly involved in the regulation of immune cell chemotaxis in the tumour microenvironment.In addition,the tyrosine kinase inhibitor 4.5.dianilinophthalimide was identified as a compound that may inhibit PPGL progression by interfering with the action of tumour stemness-associated hub genes.The stemness risk scoring system can be used for prognostic prediction in patients with PPGL and has high predictive power(AUC=0.908).Patients with a high stemness risk score have lower expression levels of multiple immune checkpoints and a lower proportion of immune cells that exert antitumour immune activity such as M1-like macrophages and NK cells in the tumour microenvironment compared to PPGL patients with a low stemness risk score.In the TCGA-PPGL patient cohort as well as in the real-world patient cohort receiving immunotherapy,patients with lower stemness-related risk scores showed better responsiveness to immunotherapy.Conclusion:MNs are not only efficient in diagnosing PPGL,but also in predicting tumour location,size and mutational genetic information in PPGL patients.the high tumour stemness characteristics of PPGL patients’ tissues correlate with their malignant behaviour and suppressive immune microenvironment characteristics.A stemness-related risk scoring system,which can be applied to predict survival prognostic outcomes and guide the choice of treatment strategies in PPGL. |
