Laser Desorption Ionization Mass Spectrometry On Silicon Nano-substrate And Its Application In Clinical Metabolomics

Metabolites are closely related to cell signal release,neurotransmission and hormone transportation and other life processes in organisms.They are intermediates or final products of metabolic processes,reflecting biological events that have occurred and closely related to pathological and physiological conditions.Metabolite is a key link between internal regulation and phenotype.By monitoring the changes of endogenous metabolites,metabolic biomarkers can be mined to characterize the physiological conditions of specific disease phenotypes,which can provide new ideas for IVD developments.At present,the main techniques for metabolomics include nuclear magnetic resonance(NMR)and mass spectrometry(MS).In contrast with other methods,MS exhibits unique advantages in sensitivity,specificity,accuracy,detection speed and detection flux.However,traditional MS approaches rely on tedious sample pretreatment due to the low molecular abundance and high sample complexity in biofluids,which limited the application of metabolomics in potential biomarkers discovery,clinical diagnosis and disease mechanism investigation.In this paper,we aimed to develop an analytical technique for complex biofluids based on vertical silicon nanowires(SiNWs)chips to achieve low background noise,high sensitivity,high reproducibility,and high stability of metabolite detection.Suitable metabolite extraction methods were proposed for different types of biofluids to shorten the sample pretreatment time.Besides,a high-throughput metabolomics platform for clinical disease diagnosis was constructed by combining laser desorption ionization-mass spectrometry(LDI-MS)platform with artificial intelligence.In detail,we designed a novel vertical SiNWs chip with low background noise for rapid detection of metabolites in complex biofluids.Noninvasive diagnosis of early-stage tumors was further realized by combining vertical SiNWs-based LDI-MS technique with saliva metabolomics.In addition,a tip-contact extraction method was proposed for extracting endogenous metabolites and exogenous substance in salty urine samples.The acquired urinary metabolic fingerprinting could mirror the occurrence and development of various tumors and chronic diseases,which was utilized for accurate disease identification with the assistance of machine learning.Furthermore,we established a pretreatment-free LDI MS platform based on porous silicon particles to decode the evolutionary metabolic landscape from chronic hepatitis B to hepatocellular carcinoma and identify different stages of liver disease.The main contents of this paper are as follows:In Chapter 1,we firstly summarized the major sample preprocessing methods,main mass spectrometry techniques and untargeted metabolomics strategies.Besides,we reviewed the correlation of metabolic disorder and various diseases,and their significance in clinical diagnosis,drug and therapy development.Finally,we introduced the application of nanomaterials in metabolite extraction,MS signal enhancement,high-abundance protein and salt removal as well as the main mechanism involved in LDI-MS.Also,the basis of topic selection and the overall framework are proposed at the end of this part.In Chapter 2,we constructed a fluorinated ethylene propylene modified vertical silicon nanowires(FEP@SiNWs)chip and investigated its feasibility for rapid metabolite detection in complex biological samples.On the one hand,the FEP initiator coating on the surface of vertical SiNWs chip produced few dissociation fragment peaks,and inhibited the generation of silicon ion clusters under laser irradiation,which resulted in ultra-low background noise in the low molecule-weight region.On the other hand,the phase transition of initiator promoted the desorption of analytes.Benzylpyridinium salt[BP]~+and tetraphenylboron salt[TB]~-verified the enhanced desorption mechanism on the FEP@SiNWs surface.The feasibility of this ultra-low background noise substrate in real biofluid profiling was confirmed by analyzing saliva samples from patients with type 2 diabetes mellitus(T2DM)and healthy controls.Based on the differential salivary metabolites sorted out through multivariate statistical analysis,T2DM patients and controls were successfully separated.In Chapter 3,we performed high-throughput salivary metabolite profiling on the ultra-low noise tip-enhanced LDI-MS platform for non-invasive diagnosis of early-stage lung cancer.Salivary secretion is influenced by multiple external factors such as water consumption.To minimize the impact of non-disease factors on downstream metabolic analysis,we evaluated six kinds of data-driven normalization methods in terms of four aspects,including variation of replicated measurements,dilution correction effects,variation of metabolites in urine samples from the same group,and discrimination results.Salivary metabolic fingerprinting of healthy controls,early lung cancer and advanced lung cancer patients may provide valuable information for mapping the metabolic disorder in lung cancer.Dysregulation of amino acid and nucleotide metabolism found from the salivary metabolomics experiment was supported by transcriptomic data from online TCGA repository.An artificial neural network model was established based on the verified differential metabolites and achieved a good discrimination of early lung cancer.We believed that this non-invasive diagnostic model may open up a new possibility for large-scale screening of lung cancer in high-risk groups such as heavy smokers.In Chapter 4,we developed a sensitive and reproducible urine detection platform based on the“tip-contact extraction”(TCE)method coupled with negative LDI-MS.Through a simple TCE procedure,endogenous metabolites and exogenous substance can be effectively extracted onto the surface of FEP@SiNWs for subsequent LDI-MS analysis.Compared with the traditional"drop-dry"method,TCE technique displayed great advantages in signal enhancement and high salt tolerance.Besides,stable and reproducible MS spectra was obtained by TCE technique in the presence of extra salts or during the dilution process.Considering that urinary diseases directly affected the levels of metabolites in urine samples,we applied TCE technique in nontargeted metabolomic analysis of urine samples from healthy volunteers and patients diagnosed with bladder cancer,in order to reveal metabolic disorders associated with bladder cancer and achieve non-invasive disease diagnosis.In Chapter 5,we constructed a high-throughput multi-omics platform and integrated urine metabolomics and peptidomics to enhance the diabetic kidney disease(DKD)diagnosis,especially for early-stage DKD.Urine samples were collected from healthy controls,type 2diabetes mellitus,and DKD patients at various stages.In this study,non-targeted metabolic profiles of urine samples were acquired by TCE method based on FEP@SiNWs chip coupled with LDI-MS detection,while peptide profiles in urine samples was uncovered by MALDI-TOF MS after capturing urine peptides by porous silicon microparticles.After multivariate statistical analysis,six peptides and ten metabolites were uncovered to be stepwise regulated in different DKD stages.Bioinformatics analysis revealed the disturbed metabolic pathways and biological processes in DKD progression,which were further verified by renal transcriptomic data.Compared with single omics,integrative omics model constructed with machine learning algorithm significantly increase the diagnostic outcomes for T2DM and all stages of DKD.Finally,a stepwise discriminant model was constructed for early DKD diagnosis and DKD status discrimination,which was further verified by an independent external validation cohort.In the future,this integrative approach may provide a new choice for diabetes health management.In Chapter 6,we proposed a pretreatment-free LDI MS technology based on porous silicon particles for reproducible detection and storage of serum metabolic information.(3-aminopropyl)-triethoxysilane modified porous silicon(APTES-pSi)particles exhibit excellent ion desorption efficiency,size exclusion effect,salt tolerance and protein tolerance,so that metabolites in in native serum samples(only 100 n L)can be rapidly and sensitively detected without any pretreatment.To minimize the influence of external factors on downstream serum metabolic analysis,we evaluated and screened six kinds of data-driven normalization methods.In addition,we demonstrated that APTES-pSi particles can effectively store metabolic information in serum samples.Compared with natural serum without any pretreatment,the pre-treated serum sample displayed better stability under different storage conditions and freeze-thaw cycles.In view of the good metabolic information storage ability of porous silicon particles,it is expected to be a good medium for long-distance transport of biological samples,and promote clinical multi-center collaboration and large cohort omics studies.In Chapter 7,we performed nontargeted metabolomic analysis on the pretreatment-free LDI MS platform for accurate diagnosis of different stages of liver disease.In this work,a total of1192 serum samples were collected from healthy controls,chronic hepatitis B,liver cirrhosis,and HCC volunteers.Multivariate statistical analysis revealed the progressive metabolic disorders associated with HCC development.The alteration of metabolic pathways was concentrated into amino acid metabolism and nucleotide metabolism,which was consistent with online transcriptomic dataset.With the assistance of machine learning,a stepwise diagnostic model was constructed to characterize different types of liver diseases in real-time.Compared with clinical AFP indexes and imaging results,this model greatly improves the diagnostic sensitivity of HCC patients.