Recently, the team of Gu Weili at the First Affiliated Hospital of Guangzhou Medical University published an article titled "Metabolic signatures of lymphangioleio-myomatosis in biofluids: nuclear magnetic resonance (NMR)-based metabonomics of blood plasma: a case-control study" in the journal Annals of Translational Medicine. Based on nuclear magnetic resonance (NMR) spectroscopy, the article explored and compared the characteristics and differences of blood metabolites between patients with lymphangioleiomyomatosis (LAM) and healthy controls, in order to find biomarkers that could be used for the diagnosis and classification of LAM. As a new non-invasive diagnostic method, NMR made a beneficial contribution to finding potential markers for LAM patients. The Tianjin Key Laboratory of Clinical Multi-omics, whose supporting institution ProteinT has reached a long-term strategic cooperation with Bruker on the promotion of NMR spectroscopy in China, jointly completed the project in cooperation with Guangzhou Medical University. Based on the advantages of NMR spectroscopy—non-destructive detection of trace samples, high-throughput full-process automation, unparalleled reproducibility, and an exclusive database platform—ProteinT was responsible for the NMR detection and data analysis parts of the project and was listed as a co-author of the article.
Research Background
Lymphangioleiomyomatosis (LAM) is a rare, low-grade malignant tumor that almost exclusively affects women. There are two types: sporadic LAM (S-LAM) without a genetic background and LAM associated with the genetic disease tuberous sclerosis complex (TSC-LAM). Their average incidence is about 4.9 per million, and the incidence of TSC in the population is ≈1/20,000. 80% of 40-year-old female TSC patients develop pulmonary cystic changes. Common symptoms include dyspnea, cough, chylous effusion, or pneumothorax, and as the disease progresses, lung function deteriorates. TSC-LAM also has other multi-system clinical features of TSC, including effects on the nervous system and skin. At present, pulmonary or extrapulmonary pathological diagnosis is the gold standard for diagnosing LAM, but obtaining pathological specimens is traumatic for patients and carries the risk of adverse reactions such as pneumothorax and bleeding.
Serum VEGF-D level can serve as a non-invasive biomarker to distinguish LAM from other lung diseases. However, the limitation of VEGF-D is that it increases only in patients with severe lymphatic involvement, and there are still many LAM patients whose VEGF-D does not increase. Therefore, it is necessary to find new biomarkers to further improve the diagnostic performance of LAM and hopefully to distinguish disease types.
Nuclear magnetic resonance (NMR) spectroscopy is an advanced measurement method that can provide a large amount of metabolomics data, including lipid indicators, metabolites, and amino acids. In terms of lipids, NMR can not only group different lipoproteins according to particle size (very-low-density lipoprotein (VLDL-1~5), intermediate-density lipoprotein (IDL), low-density lipoprotein (LDL-1~6), and high-density lipoprotein (HDL-1~4)), but can also quantitatively determine the content of lipoproteins (cholesteryl ester (CE), free cholesterol (FC), triglycerides (TG), and phospholipids (PL)) based on the different responses of substances in a magnetic field. In this study, we combined NMR spectroscopy and laboratory testing to compare plasma lipoproteins and metabolites between LAM patients and healthy subjects and between different types of LAM patients, looking for differences, so as to provide a reference for the non-invasive diagnosis and classification of LAM.
Research Methods
1. Population
From January 2020 to January 2022, 69 LAM patients from the outpatient clinic of the First Affiliated Hospital of Guangzhou Medical University were consecutively enrolled. All subjects met the criteria of the American Thoracic Society/Japanese Respiratory Society. Diagnosis was mainly based on clinical history, high-resolution computed tomography (CT) of the lungs (diffuse thin-walled cystic lesions), and pathology (LAM cells). Exclusion criteria were: (I) presence of other types of malignancy; (II) presence of rheumatic or endocrine diseases; (III) incomplete data. Of the 69 LAM patients, 2 were excluded due to other comorbidities (1 case of breast cancer and 1 case of hyperthyroidism), and 67 patients entered the next phase of the study, of whom 4 were excluded due to incomplete data and 2 patients refused blood collection. Finally, there were 61 LAM patients in total.
2. Grouping
51 sporadic LAM (S-LAM) patients, 10 tuberous sclerosis complex LAM (TSC-LAM) patients, and 30 healthy controls.
3. Sample Collection and Transport
After all subjects had fasted for 10 hours, venous blood samples were drawn into vacuum tubes, kept at room temperature for 30 min, and centrifuged (1200×g, 10 min). Plasma samples were collected in vacuum tubes and transported on dry ice at −80°C.
4. Statistical Analysis
Data were analyzed using ProteinT's Umbrella program. Comparisons between the two groups were performed using the Wilcoxon test, independent-sample t-test, or χ2 test. Using P value (<0.05) and fold change (FC) (<0.83 or >1.2) as screening criteria, differential indicators were identified and logistic regression modeling was performed.
Research Approach
Among S-LAM, TSC-LAM, and healthy individuals, the Wilcoxon test, independent-sample t-test, or χ2 test, and logistic modeling correlation analysis were used to determine differential indicators, so as to provide a reference for the non-invasive diagnosis and classification of LAM.
Research Results
1. Basic Demographic Information and Clinical Characteristics
According to the clinical practice guidelines of the American Thoracic Society/Japanese Respiratory Society, 61 stable female patients met the criteria (51 S-LAM, 10 TSC-LAM). All patients completed CT scans of the lungs before enrollment. Lung CT showed diffuse distribution of bilateral pulmonary cysts, consistent with the features of LAM. Blood was collected from patients within 1 month after diagnosis, and they were not receiving any treatment at the time of blood collection. Patient age and lung function indicators were normally distributed data and are expressed as mean ± standard deviation. Most clinical indicators showed no significant difference between the S-LAM group and the TSC-LAM group. All LAM patients and controls were Asian, with a mean age of 40.70±8.59 years (S-LAM patients 40.90±8.68 years; TSC-LAM patients 39.70±8.49 years). None had a smoking history, and 3 cases (4.92%) had a family history of LAM or TSC. Among the 61 patients, 44 (72.13%) underwent biopsy, 28 (45.90%) had pneumothorax, 5 (8.20%) had chylothorax, and 39 (63.93%) had lung involvement. Among the involved extrapulmonary organs, renal AML accounted for the largest proportion (23 cases, 37.70%), followed by the liver, retroperitoneal AML, uterine fibroids, and chylous ascites. The rate of lung involvement in the TSC-LAM group was significantly higher than in the S-LAM group, especially the renal involvement. The authors also tested lung function at baseline. Pulmonary ventilation function indicators included forced expiratory volume in the first second (FEV1) predicted % (70.26%±27.50%), forced vital capacity (FVC) predicted % (90.63%±16.03%), and FEV1/FVC% (66.93%±23.54%). The single-breath % predicted diffusing capacity of the lung for carbon monoxide (DLCO-SB%) was 57.28%±21.63%.
2. Differential Indicators Between LAM Patients and Healthy Controls
There were 15 different lipoprotein indicators between LAM patients and healthy controls, mainly the fifth and sixth components of low-density lipoprotein (LDL), and 18 differential metabolites (all P<0.05 and FC >1.2 or <0.83, Table 2).
3. Differential Indicators Between S-LAM Patients and Healthy Controls
There were 10 different lipoprotein indicators, mainly the sixth component of LDL, and 18 differential metabolites (all P<0.05 and FC >1.2 or <0.83, Table 2).
4. Differential Indicators Between TSC-LAM Patients and Healthy Controls
There were 14 different lipoprotein indicators, mainly the sixth component of LDL, and 8 different metabolites (all P<0.05 and FC >1.2 or <0.83, Table 2).
5. Differential Indicators Between S-LAM and TSC-LAM
The metabolites with significant differences between S-LAM and TSC-LAM were creatinine and acetone (all P<0.05 and FC<0.83 or >1.2, Table 2). There was no significant difference in lipoproteins between the two groups.
6. Logistic Regression Analysis of Indicators Selected to Distinguish LAM Patients from Healthy Controls
The 15 differential lipoprotein indices and 18 differential metabolites obtained by comparing LAM patients with healthy controls were modeled by logistic regression. In the validation cohort, the AUC of the median model was 0.9737, and the sensitivity and specificity of the model were 100% and 89%, respectively. In all subjects of this group, ROC curves were plotted separately for the high-frequency indicators or indicators appearing in the median model, to distinguish patients from healthy individuals. Among them, the indicators with higher diagnostic performance were methionine (AUC=0.929, sensitivity=73.8%, specificity=100%, cutoff=0.011 mmol/L) and acetic acid (AUC=0.966, sensitivity=95.1%, specificity=90%, cutoff=0.006 mmol/L).
7. Logistic Regression Analysis of Indicators Selected with Significant Differences Between S-LAM Patients and Healthy Controls
The 10 differential lipoprotein indicators and 18 differential metabolites of S-LAM patients and healthy individuals were modeled by logistic regression. In the validation cohort, the AUC of the median model was 0.9852, and the sensitivity and specificity of the model were 100% and 93%, respectively. In all subjects of this group, ROC curves were plotted separately for the high-frequency indicators or indicators appearing in the median model, to distinguish patients from healthy populations. Among them, the indicators with higher diagnostic performance were methionine (AUC=0.924, sensitivity=72.5%, specificity=100%, cutoff=0.011 mmol/L) and acetic acid (AUC=0.963, sensitivity=94.1%, specificity=90%, cutoff=0.006 mmol/L).
8. Logistic Regression Analysis of Indicators Selected with Significant Differences Between TSC-LAM Patients and Healthy Controls
The 14 differential lipoprotein indices and 8 differential metabolites obtained by comparing TSC-LAM patients with healthy controls were modeled by logistic regression. In the validation cohort, the AUC of the median model was 1, and the sensitivity and specificity of the model were 100% and 100%, respectively. In all subjects of this group, ROC curves were plotted separately for the high-frequency indicators or indicators appearing in the median model, to distinguish patients from healthy individuals. Among them, the indicators with higher diagnostic performance were methionine (AUC=0.948, sensitivity=80%, specificity=100%, cutoff=0.011 mmol/L), acetic acid (AUC=0.975, sensitivity=100%, specificity=90%, cutoff=0.007 mmol/L), and creatinine (AUC=0.917, sensitivity=80%, specificity=95%, cutoff=0.059 mmol/L).
9. Correlation Analysis Between Differential Metabolites and Clinical Indicators
The clinical indicators of 61 LAM patients were collected, including age, lung function (FEV1%, FVC%, FEV1/FVC, DLCO-SB%), pneumothorax, chylothorax, renal AML, liver lesions, chylous ascites, uterine fibroids, and retroperitoneal AML. The correlation between clinical indicators and the differential indicators with high diagnostic efficiency (alanine, methionine, L6TG, acetic acid, L6PL, creatinine, histidine, 2-aminobutyric acid, acetone, sarcosine) was analyzed. Methionine was significantly correlated with the occurrence of pneumothorax (P<0.05), and creatinine was significantly correlated with uterine fibroids (P<0.05) (Figure 5).
Discussion
This is the first time that the NMR method has been used to conduct metabolomic analysis of LAM patients in an Asian population. This study aimed to find blood biomarkers for LAM and to evaluate the effectiveness of this non-invasive examination. The study found that, compared with healthy controls, S-LAM and TSC-LAM patients jointly showed upregulated expression of some sixth components of LDL. The authors also found metabolic abnormalities of some amino acids, such as downregulation of methionine and alanine, and upregulation of acetic acid and creatinine. These indicators were high-frequency indicators or indicators that appeared in all three models in the median model. In addition, the researchers found significant differences in creatinine and acetone between the S-LAM and TSC-LAM subgroups. Methionine levels were significantly lower in LAM patients than in healthy individuals. Methionine is an amino acid required to maintain cell growth and protein translation. Abnormal methionine metabolism (downregulation) has been observed in many tumors, such as glioma and lung cancer. In LAM patients and the two LAM subtypes (S-LAM and TSC-LAM), methionine was significantly decreased, which suggests that pulmonary cystic degeneration may be associated with the downregulation of methionine. Moreover, methionine could effectively distinguish LAM from the healthy population, with an AUC of 0.929 and a cutoff of 0.011 mmol/L. Methionine was significantly correlated with the occurrence of pneumothorax, suggesting that lower methionine levels in patients' blood may be an indicator for early prediction of pneumothorax risk. In the future, in-depth research is still needed to confirm this possibility. Another indicator with higher diagnostic performance is acetic acid. Compared with healthy controls, its expression was significantly upregulated in LAM patients. The AUC was 0.966 (sensitivity=95.1%, specificity=90%), with a cutoff of 0.006 mmol/L. A previous study found that, in their metabolomic analysis of breast cancer patients, acetic acid levels were elevated compared with healthy controls, making it a potential biomarker. Acetic acid is significantly elevated in lung cancer and liver cancer, confirming that acetic acid and its metabolites may promote tumor growth.
In addition, the authors also found differential expression of alanine. Compared with healthy controls, plasma alanine levels in cancer patients were significantly reduced, very likely because it serves as a major gluconeogenic precursor to meet the high glucose uptake and demand of tumor cells. Pancreatic cancer cells depend on extracellular alanine as a carbon source to promote the tricarboxylic acid cycle in a glutamate-pyruvate transaminase 2 (GPT2)-dependent manner. A recent study re-established the role of alanine in T cell activation. GPT2 is responsible for alanine catabolism, converting alanine into pyruvate, which is the main substrate of mitochondrial metabolism. Pyruvate can compensate for glutamine consumption by replenishing the tricarboxylic acid cycle and maintain the anabolic processes of cancer by utilizing the activities of different enzymes (such as the pyruvate dehydrogenase complex or pyruvate carboxylase). In this study, acetone was significantly higher in S-LAM patients than in healthy controls and was able to distinguish S-LAM from TSC-LAM. A previous study found that lung cancer cells produce higher levels of acetone, and blood acetone levels in small cell lung cancer patients are significantly elevated. Ketone bodies are mainly responsible for delivering energy to cells and participate in regulating various cellular processes. The increased ketogenesis in cancer cells may be due to alterations in lipogenesis, gluconeogenesis, and cholesterogenesis, all of which are closely related to ketone body regulation, or due to the high competition between cancerous and healthy cells for primary energy compounds. This study also found upregulation of creatinine. 48-80% of TSC patients have kidney disease, accompanied by AML, cysts, cancer, and/or progression to renal insufficiency.
In this study, the AML involvement rate in the TSC-LAM group was as high as 90%, consistent with the result of elevated creatinine. Previous studies have shown that 34-80% of TSC-LAM patients have AMLs, cortical cysts, malignant lesions, and/or chronic renal insufficiency. Compared with healthy controls, creatinine was significantly higher in TSC-LAM patients, reflecting the possibility of renal involvement in TSC-LAM patients. In addition, this indicator showed differential expression between S-LAM and TSC-LAM, showing promise as an indicator to distinguish the two subtypes. The authors analyzed the plasma lipid metabolism profile of LAM patients and compared it with healthy controls. In S-LAM patients, plasma LDL-6 triglycerides were significantly elevated, while in TSC-LAM patients, LDL-6 phospholipids were significantly elevated. Previous studies have shown that breast cancer is associated with lipid disorders, and elevated blood LDL levels may be related to tumor lipid reprogramming metabolism, consistent with the results of this study. Using the NMR platform, the elevated lipoprotein subgroups were classified, and the sixth component of LDL was found to be elevated, which is usually overlooked in routine clinical lipid testing. Many studies have found that LDL can promote the proliferation, metastasis, and angiogenesis of breast cancer cells. The association between these metabolites and their potential biological roles in cell survival and signal transduction suggests that lipids may be both disease-related biomarkers and potential therapeutic targets for LAM. In summary, the experimental data are consistent with previous reports on other types of tumors, indicating that LDL is significantly elevated in LAM, but the AUC of LDL-6 triglycerides and LDL-6 phospholipids is not high. The exact mechanism between lipoprotein abnormalities and the occurrence of LAM is still unclear and awaits further exploration in more basic research in the future.
Because LAM is a rare disease, the number of study subjects enrolled in this study was limited; in particular, TSC-LAM is a rare disease within a rare disease, so the number of cases is very small. The authors hope to overcome this limitation in the future by expanding the number of patients and conducting multicenter research. In addition, although some lipoproteins and metabolites differ significantly between LAM patients and healthy controls, the underlying mechanisms remain unclear. In the future, research on animal models or cell models can further confirm the relevant mechanisms, such as their metabolic pathways, and achieve breakthroughs in therapeutic drugs.
Summary
In this study, the blood metabolic characteristics of LAM patients were explored in depth using NMR technology for the first time. NMR has good discriminatory ability for blood samples, thereby opening up new possibilities for a non-invasive diagnostic method for LAM. Metabolic disorder is one of the manifestations of LAM. Methionine and acetic acid levels in the plasma of LAM patients can serve as biomarkers for disease diagnosis. Methionine was also found to be associated with pneumothorax in LAM patients. Acetone and creatinine are very promising metabolic markers for distinguishing S-LAM from TSC-LAM.
