Nontargeted and targeted metabolomics approaches reveal the key amino acid alterations involved in multiple myeloma

Study design and patient recruitment

In this study, a total of 70 participants in two cohorts were recruited for metabolomic detection. Patients with MM were collected from Lanzhou University Second Hospital, and NC subjects were recruited through conventional physical examination in the hospital. MM patients were diagnosed according to the criteria of International Myeloma Working Group (IMWG) (Rajkumar, 2016). The International Staging System (ISS) was used to evaluate disease progression (Loehrer, 2006). The patients with MM in the study were all newly diagnosed cases. Subjects with immunodeficiency disease, medication use, smoking habits, or underlying chronic disease were excluded from the study at the preliminary screening stage. All MM patients in this study were not treated with radiotherapy or chemotherapy. Finally, in the cohort 1, 40 participants including 20 MM patients and 20 NC subjects, were selected for nontargeted- and targeted-metabolomics. We also collected the information about clinical variables, including the biochemical detection of serum albumin (ALB), interleukin 6 (IL-6), lactic dehydrogenase (LDH), β2-microglobulin (B2-MG), globulin (GLB), and serum creatinine (SCR) using the Beckman automatic biochemistry analyzer (Beckman Chemistry Analyzer AU5800, Beckman Coulter, Inc., Brea, CA, USA) and Roche electrochemiluminescence system (Cobas e 801 Analyzer, Roche Diagnostics GmbH, Mannheim, Germany), for all the 40 participants in cohort 1. The study of cohort 1 mainly contained two parts: (1) all 40 participants underwent nontargeted metabolomics analysis, and (2) 30 participants (including 15 MM patients and 15 NC subjects) underwent targeted metabolomics analysis. In the cohort 2, another 30 cases, including 15 MM patients and 15 NC subjects, were recruited as the validation group of the targeted metabolomics analysis. Figure 1 summarized the study design, and detailed information on the participants could be found in Table 1 (cohort 1) and Table S1 (cohort 2). This study was approved by the Medical Ethics Committee of Lanzhou University Second Hospital (2018a-037). Written informed consent was obtained from all participants.

Nontargeted metabolomics analysis using ultrahigh-performance liquid chromatography (UHPLC) combined with quadrupole time-of-flight mass spectrometry (Q-TOF-MS)

UHPLC-Q-TOF-MS analysis

UHPLC analysis was performed with an Agilent 1290 Infinity LC UHPLC system. Chromatographic separation was carried out at 25 °C on a hydrophilic interaction chromatography (HILIC) column with a flow rate of 0.3 mL/min. The mobile phase consisted of solvent A (water + 25 mM ammonium acetate + 25 mM ammonia) and solvent B (acetonitrile). Gradient elution was performed as follows: 95% solvent B with 0–0.5 min; the linear change of solvent B was from 95% to 65% within 0.5–7 min, from 65% to 40% with 7–8 min, and maintained at 40% for 8–9 min; then, the linear change of solvent B was from 40% to 95% within 9–9.1 min and maintained at 95% for 9.1–12 min. The samples were placed in an automatic sampler at 4 °C throughout the whole analysis. To avoid the influence of signal fluctuation caused by instrument detection, continuous analysis of samples was carried out in random order. Moreover, pooled quality control (QC) samples (generated by taking an equal aliquot of all the samples included in the experiment) were run at the beginning of the sample queue for column conditioning and every ten injections thereafter to assess inconsistencies that are particularly evident in large batch acquisitions in terms of retention time drifts and variation in ion intensity over time.

Q-TOF-MS analysis was conducted on an AB Triple TOF 6600 mass spectrometer (AB SCIEX, Framingham, MA, USA). After HILIC chromatographic separation, the electrospray ionization (ESI) source conditions were as follows: ion Source Gas1 (Gas1): 60 psi, Ion Source Gas2 (Gas2): 60, Curtain gas (CUR): 30, source temperature at 600 °C, ionSapary Voltage Floating (ISVF): ±5,500 V (positive and negative model); TOF MS scan m/z range: 60–1,000 Da, product ion scan m/z range: 25–1,000 Da, TOF MS scan accumulation time: 0.20 s/spectra, and product ion scan accumulation time: 0.05 s/spectra. Information-dependent acquisition (IDA) was used to obtain the secondary MS, as well as high sensitivity mode, declustering potential (DP): ±60 V (positive and negative modes), and collision energy: 35 ± 15 eV. IDA was set as exclude isotopes within 4 Da and candidate ions to monitor per cycle: 6.

Data processing and analysis

The raw data were converted into. mzXML format files using the ProteoWizard converter tool and then processed using XCMS software for peak alignment, retention time correction, and peak area extraction. Then, the structures of metabolites were identified by comparing the accuracy of m/z values (<25 ppm) and matching second-stage spectra with the laboratory’s self-built database (Applied Protein Technology Co., Ltd., Shanghai, China).

After data preprocessing by Pareto scaling, multidimensional data analyses were carried out, including unsupervised principal component analysis (PCA), supervised partial least squares discrimination analysis (PLS-DA), and orthogonal partial least squares discriminant analysis (OPLS-DA). Single-dimensional statistical analyses, including Student’s t-test and fold change (FC) analysis, were conducted. The variable importance in the projection (VIP) value (>1) in the PLS-DA model and P value analyzed by Student’s t-test were combined to confirm the significance of differential metabolites. Significant differential metabolites were obtained when VIP > 1 and P value < 0.05, and differential metabolites were selected when VIP > 1 and 0.05 < P value < 0.1. Moreover, Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway (http://www.genome.jp/kegg/) pathway enrichment analysis for significant differential metabolites was conducted by the MetaboAnalyst online tool (http://www.metaboanalyst.ca/MetaboAnalyst/) to investigate the metabolomics pathways.

Targeted metabolomics analysis of amino acids using multiple reaction monitoring-mass spectrometry (MRM-MS)

MRM-MS analysis

The analysis was performed with an Agilent 1290 Infinity LC UHPLC system. The mobile phases were as follows: solvent A was 25 mM ammonium formate + 0.08% formic acid (FA) in water, and solvent B was 0.1% FA-acetonitrile. The sample was placed in an automatic sampler at 4 °C, the column temperature was 40 °C, the flow rate was 250 µL/min, and the injection volume was 1 µL. Gradient elution was performed as follows: the linear change of solvent B was from 90% to 70% within 0–12 min; from 70% to 50% with 12–18 min; from 50% to 40% within 18–25 min, from 40% to 90% with 30–30.1 min; and maintained at 90% for 30.1–37 min. In the sample queue, a QC sample was also set in a certain number of experimental samples at every interval to test and evaluate the stability and repeatability of the system. MS analysis was carried out in positive ionization mode on a 5500 QTRAP mass spectrometer (AB SCIEX, Framingham, MA, USA). The conditions of the 5500 QTRAP ESI source were as follows: ion source temperature, 500 °C; Gas1, 40 psi; Gas2, 40 psi; CUR, 30 psi; and ISVF, 5500 V. MRM mode was then used to detect the ion pairs.

Data processing and analysis

The chromatographic peak area and analyte retention time were extracted using Multiquant software. Based on the standards of amino acids and their derivatives, the analyte retention time was corrected, followed by identification of the metabolites. PLS-DA was also conducted to explore the amino acids profiling and differential amino acid metabolites between groups. Score scatter plots and loading plots were generated to visualize the separation of samples and metabolites. MedCalc software (v19.0.4) was utilized for receiver operating characteristic curve (ROC) analysis to evaluate the diagnostic accuracy of amino acid metabolites. Moreover, Spearman correlation analysis was performed with the statistical platform R package (v3.2.4). KEGG pathway enrichment analysis for significant differential metabolites was also conducted by the MetaboAnalyst online tool.

Nontargeted metabolomics analysis using UHPLC-Q-TOF-MS

Multivariate statistical analysis

To analyze the metabolic changes between MM patients and NC, nontargeted metabolomics analysis was carried out using UHPLC-Q-TOF-MS. As a result, the PCA score plots showed a clear separation between MM patients and NC in both positive-ion and negative-ion modes (Fig. 2A). To further identify ion peaks that could possibly be used to differentiate the metabolic profiles of MM patients and NC, the supervised PLS-DA model was conducted. The results showed that MM patients were also separated from NC by the PLS-DA score plots in both modes (Fig. 2B). Consistent results were also obtained by the OPLS-DA score plots in both modes (Fig. 2C). Further permutation tests consisting of 200 permutations demonstrated that the model was not overfitted (positive-ion model: R2 = (0.0, 0.707), Q2 = (0.0, −0.3673); negative-ion model: R2 = (0.0, 0.8331), Q2 = (0.0, −0.4174); Fig. 2D). These data indicated that there was a significant metabolic change between MM patients and NC.

Analysis of differential metabolites between MM patients and NC

To compare differences in metabolites between MM patients and NC, differential metabolites were screened out between MM patients and NC in both positive-ion and negative-ion modes based on VIP >1 in the PLS-DA model and P < 0.1 analyzed by Student’s t-test. The results showed that a total of 62 differential metabolites were identified between MM patients and NC in positive-ion mode, of which 56 were significant (P < 0.05). Similarly, 45 differential metabolites were identified between MM patients and NC in negative-ion mode, of which 38 were significant (P < 0.05) (Table S2). These data confirmed that there were significant differences in metabolites between MM patients and NC.

Metabolic pathway analysis

To investigate the metabolomic pathways involved in MM development, the differential abundant metabolites were enriched for the related metabolic pathway analysis using the MetaboAnalyst online tool. As shown in Fig. 3, differential metabolites were significantly enriched in multiple metabolic pathways, such as protein digestion and absorption (including L-alanine, L-glutamine, L-tryptophan, L-histidine, L-asparagine, L-valine, and L-isoleucine), ABC transporters (including L-alanine, L-glutamine, L-histidine, L-valine, and L-isoleucine), and biosynthesis of amino acids (L-alanine, L-glutamine, L-tryptophan, L-histidine, L-asparagine, L-valine, L-citrulline, and L-isoleucine). Notably, these pathways were associated with amino acid metabolism.

Targeted metabolomics analysis of amino acids using MRM-MS

Serum amino acid metabolites of MM patients and NC showed distinguished profiles

Previously, the results indicated that the altered metabolic profile of MM was mainly involved in amino acid metabolomics features. We further conducted a target amino acid metabolomics analysis to gain more insight into MM metabolism. Generally, in the cases of cohort 1, the PLS-DA two-dimensional (2D) score plot showed clearly distinguished profiles between the MM and NC groups (Fig. 4A). Moreover, scatter loading plot analysis was carried out to evaluate whether these differential amino acid metabolites accounted for the distinguished profiles. MM patients exhibited higher levels of choline, creatinine, glutamate, and asparagine as well as lower levels of alanine/sarcosine, valine, tryptophan, and cystine than NC (Fig. 4B). These data suggested that altered amino acid profiles could discriminate MM patients from NC.

Expression of differential amino acid metabolites and the KEGG functional analysis

According to the metabolomics features, with the purpose of finding potential diagnostic biomarkers or therapeutic targets for MM, we focused on the differential amino acid metabolites in MM patients compared to those in NC. To further investigate the detailed variety degree of each metabolite, we analyzed the intensity differences in a total of 28 amino acid metabolites. Compared to NC, MM patients exhibited differential metabolic profiles. Overall, in the cases of cohort 1, there was a total of 12 significantly differential abundant amino acid metabolites in MM patients relative to NC (P < 0.05), including lysine, leucine, isoleucine, histidine, valine, threonine, glutamine, tryptophan, choline, ornithine, creatinine, and alanine/sarcosine (Table S3). Furthermore, we validated these findings with the samples in cohort 2, and the results showed similar dysregulation trends with the findings of cohort 1, that we identified 14 significantly differential abundant amino acid metabolites in MM group (Table S4). Specifically, the high expression levels of creatinine (FC = 2.029, P < 0.01 in cohort 1; FC = 1.632, P < 0.05 in cohort 2) and choline (FC = 1.508, P < 0.001 in cohort 1; FC = 1.234, P = 0.125 in cohort 2) were found in MM patients in both cohort 1 and cohort 2. Besides, since glutamate signaling has been reported involved in malignant disorders (Willard & Koochekpour, 2013), we also examined glutamate levels and found slightly high levels in MM patients (FC = 1.186, P = 0.245 in cohort 1; FC = 1.292, P = 0.346 in cohort 2).

To determine the pathway impact of these differential amino acid metabolites, we performed pathway classifications provided by KEGG analysis. Results indicated that the 12 differential abundant metabolites in cohort 1 were significantly enriched in 22 metabolic pathways, such as aminoacyl-tRNA biosynthesis, valine/leucine/isoleucine biosynthesis, phenylalanine/tyrosine/tryptophan biosynthesis, D-glutamine/D-glutamate metabolism, and histidine metabolism (Figs. 4C, 4D).

Differential amino acid metabolites may discriminate MM patients from NC

We subsequently evaluated the performances of these differential abundant metabolites in distinguishing MM from NC group. ROC analysis showed that, the amino acid metabolites classifier could act as potential disease diagnostic biomarkers in both cohort 1 and cohort 2 (AUC > 0.7, P < 0.05, Table 2 and Fig. S1). In MM patients of cohort 1, the AUC values of two representative upregulated amino acid metabolites, choline and creatinine, were 0.822 and 0.822, respectively (Table 2). Similarly, the AUC of creatinine in the cohort 2 was 0.747 (Table 2). Meanwhile, the ROC analysis also showed good diagnostic values of the downregulated amino acid metabolites in MM, such as leucine (AUC = 0.871 in cohort 1, AUC = 0.920 in cohort 2) and tryptophan (AUC = 0.884 in cohort 1, AUC = 0.902 in cohort 2). Specifically, we observed extremely high AUC values of the metabolite valine in both cohort 1 (AUC = 0.964) and cohort 2 (AUC = 0.960). The results above suggested that these key differential amino acid metabolites might be regarded as potential biomarkers for the diagnosis or therapeutic targets of MM.

Differential amino acid metabolites were associated with MM phenotypes

To investigate the correlations between differential amino acid metabolites and MM-associated phenotypes, we analyzed the correlations among 6 clinical indices/variables and 12 amino acid metabolites that differed significantly in abundance between MM patients and NC in the cohort 1 (Fig. 5). The clinical variables included serum albumin (ALB), interleukin 6 (IL-6), lactic dehydrogenase (LDH), β2-microglobulin (B2-MG), globulin (GLB), and serum creatinine (SCR). Among 12 amino acid metabolites whose abundances differed in MM patients, two mainly correlation clusters were obtained. The cluster of upregulated components, creatinine and choline in MM patients, was positively correlated with the levels of LDH, SCR, and B2-MG. Moreover, in the downregulated amino acids cluster, 10 amino acid metabolites were negatively correlated with the levels of IL-6 and GLB. Among these metabolites, the relationship between GLB and valine/isoleucine/tryptophan/leucine/sarcosine exhibited high correlation coefficients (adjusted P < 0.01). These data suggested that the combination of clinical indices and amino acid metabolites alterations facilitate a better understanding of MM progression.