Quantification of serum metabolites in early colorectal adenomas using isobaric labeling mass spectrometry

Abstract

A major challenge in reducing the death rate of colorectal cancer is to screen patients using low-invasive testing. Blood test shows a high compliance rate with reduced invasiveness. In this work, a multiplex isobaric tag labeling strategy coupled with mass spectrometry is adopted to relatively quantify primary and secondary amine-containing metabolites in serum for the discovery of metabolite level changes of colorectal cancer. Serum samples from patients at different risk statuses and colorectal cancer growth statuses are studied. Metabolite identification is based on accurate mass matching and/or retention time of labeled metabolite standards. We quantify 40 metabolites across all the serum samples, including 18 metabolites validated with standards. We find significantly decreased levels of threonine and asparagine in the patients with growing adenomas or high-risk adenomas (p < 0.05). Glutamine levels decrease in patients with adenomas of unknown growth status or high-risk adenomas. In contrast, arginine levels are elevated in patients with low-risk adenoma. Receiver operating characteristic analysis shows high sensitivity and specificity of these metabolites for detecting growing adenomas. Based on these results, we conclude that a few metabolites identified here might contribute to distinguishing colorectal patients with growing adenomas from normal individuals and patients with unknown growth status of adenomas.

Graphical Abstract

INTRODUCTION

Colorectal cancer is the 3^rd most diagnosed cancer with an incidence of 10.0% and it is ranked 2^nd in mortality (9.4%) in cancer-caused death.¹ Early diagnosis and treatment are critical for reducing the mortality of colorectal cancer. Currently, optical colonoscopy (OC) is the gold standard for the detection of colonic lesions, and computed tomographic colonoscopy (CTC) is also considered an acceptable alternative.² However, these two methods are inaccessible to isolated populations and can be expensive for screening large populations. The identification of low-invasiveness biomarkers for early colorectal cancer detection is challenging. To improve compliance and reduce the invasiveness of detection, stool-based tests have been developed.³ Although they perform well for frank colorectal cancer, these methods can only detect 11% to 42% of the advanced colonic adenoma.³

For colorectal polyps, polypectomy in the early stage of this disease is efficient for long-term prevention.⁴ Based on longitudinal analysis of colorectal polyps in patients by CTC, it is found that growing adenomas are prone to develop into high-risk adenomas and then become colorectal cancer.^5–7 However, only about one-fourth of (6–9 mm) polyps grow and most colorectal polyps stop growing or regress.^{6, 8} Therefore, overdiagnosis should be avoided to achieve improvements in the prevention of colorectal cancer. Taken together, the longitudinally monitored CTC cohort provides support to determine the degree of correlation of changed levels of serum components with the growing adenoma. In addition, samples from colonoscopy/cancer-free cases classified as screening normal were used as control. Previous studies reported significant changes in the level of several serum proteins in patients carrying premalignant colonic adenomas.^{9, 10} Meanwhile, efforts on discovering metabolite biomarkers have been made and a few potential metabolite biomarkers have been reported in blood samples.^11–13 However, metabolite biomarkers associated with early growing colonic adenomas are rarely studied.

Mass spectrometry (MS)-based quantification is useful in metabolomics investigations. Typically, by spiking an isotopic analog of the analyte as an internal standard, absolute quantification can be achieved.¹⁴ However, suitable isotopic internal standards for some targeted metabolites may not be available and the cost can be high if many targeted analytes need to be quantified. Alternatively, relative quantification can be achieved by label-free quantification and chemical derivatization approaches. Compared to label-free quantification, stable isotope labeling is more accurate and offers high throughput capability for relative quantification. Stable isotope labeling strategy for quantification can be classified as MS1 level-based (mass difference or mass defect) quantification and tandem mass spectrometry (MS/MS) reporter ion-based quantification. Mass difference labeling like mass differential tags for relative and absolute quantification (mTRAQ),¹⁵ formaldehyde dimethylation,¹⁶ and dansylation¹⁷ could increase mass spectrometry complexity. Strategies like hyperplexing isobaric tags and mass defect-based stable isotope labeling can give more clean spectra for quantification but require high-resolution mass spectrometers.^{18, 19} In contrast, isobaric labeling strategy can incorporate isotopic tags to analytes with nearly the same precursor ion mass shifts, and MS2 fragmentation can produce reporter ions for relative quantification. Isobaric labeling methods, such as iTRAQ,²⁰ TMT,²¹ and isobaric N,N-dimethyl leucine (DiLeu),²² provide clean MS1 spectra and only require relatively low-resolution mass spectrometers for quantification. Relative quantification of metabolites using an isobaric labeling strategy has been used in many studies and has demonstrated good performance in discovering metabolite biomarkers.^{19, 22–24} Therefore, this study employs 4-plex DiLeu tags for the relative quantification of secondary and primary amine-containing metabolites in serum samples. To control the intrinsic variance among patients, serum samples were taken during prepolypectomy and postpolypectomy. Meanwhile, serum samples from screening normal individuals were also analyzed as control.

EXPERIMENTAL

Human subject protocol

Detailed information on how the CT method for colorectal cancer screening, derivation of in vivo colorectal polyp volume, and how blood specimens were obtained as described in previous publications.^{6, 9} The Institutional Review Board at the University of Wisconsin–Madison approved this study.

Chemicals and reagents

Optima LC/MS grade acetonitrile and water, anhydrous N,N-dimethylformamide (DMF), and 4-(4,6-dimethoxy-1,3,5-triazin-2-yl)-4-methylmorpholinium tetrafluoroborate (DMTMM) were purchased from Fisher Scientific (Fair Lawn, NJ). N-Methylmorpholine (NMM) was purchased from TCI America (Tokyo, Japan). Twenty-four metabolite standards (glycine, L-alanine, L-cysteine, L-serine, L-proline, L-valine, L-threonine, L-leucine, L-isoleucine, L-aspartic acid, L-asparagine, L-glutamic acid, L-glutamine, L-methionine, L-dopamine, L-lysine, L-histidine, L-phenylalanine, L-arginine, L-tyrosine, L-tryptophan, serotonin, γ-aminobutyric acid, dopamine, kynurenine) and triethylammonium bicarbonate (TEAB) was purchased from Sigma (St. Louis, MO). Formic acid (FA) was purchased from Fluka (Buchs, Switzerland). SCX Ziptips were purchased from Millipore.

DiLeu reagent synthesis and activation

The synthesis of 4-plex isobaric DiLeu tags was previously described by Frost et al.²⁵ DiLeu reagents were activated to DiLeu triazine ester immediately before metabolite labeling. Each 1 mg dried DiLeu tag was activated by mixing with 100 μL activation solution (1.41 mg DMTMM and 0.47 μL NMM in anhydrous DMF) and then vortexed at room temperature for 45 minutes. The mixture was centrifuged at 10 000xg for 1 min, and the supernatant was used immediately for metabolite labeling.

Labeling metabolite standards

Stock solution containing 1 mM of each metabolite standard was prepared and stored at −20 °C before use. A diluted stock solution containing different amounts of each metabolite was dried with SpeedVac concentrator (Thermo Electron Corporation, West Palm Beach, FL), and redissolved in 10 μL pH8 0.5 M TEAB solution to constitute metabolites solution with different concentrations in a ratio of 1:1:1:1 (20 μM each) and a ratio of 1:2:4:8 (5 μM, 10 μM, 20 μM and 40 μM). Each aliquot was mixed with 40 μL activated DiLeu tag 115, 116, 117, or 118, respectively. For each labeling reaction, activated DiLeu reagents and metabolite mixture reacted at a molar ratio higher than 10: 1 to ensure efficient and complete labeling under room temperature for 2 h with shaking. Ten microliter 0.5 M TEAB without metabolites were also mixed with 4-plex DiLeu and combined as the negative control. The reaction was then quenched by adding hydroxylamine to 0.25% (v/v). Subsequently, labeled samples were each dried in vacuo and combined at equal ratios. Millipore SCX Ziptips (10 μL) were used to remove residual labeling chemicals, with 0.1% FA in water used as the reconstitution and washing solutions, and 5% NH₃·H₂O in 30% MeOH as elution solution. The eluate was dried down and stored at −20 °C until analysis.

Human serum sample preparation

Serum samples were thawed on ice and centrifuged at 2000 xg for 10 min to remove particulates and debris. Molecular weight cut-off filters (MWCO, 3 kDa, Millipore Amicon Ultra, Burlington, MA) were prerinsed 3 times with optima water at 14000 xg for 20 min. Twenty microliter serum supernatant from each sample was diluted in 380 μl water and added to the filter then centrifuged for 20 min at 14000 xg. Flowthrough was collected. Then 400 μL water was added to the filter to rinse the sample. An additional rinse step was applied. All the flowthrough (about 1.2 mL) was combined and dried down in Speedvac and stored at −20 °C until labeling. For normalization, a pooled serum sample from screening normal individuals was also prepared. For labeling of serum samples, 40 μL activated DiLeu reagents were mixed with serum flowthrough dissolved in 10 μL 0.5 M TEAB and shaken for 2 h. After the reaction was quenched, 2.5 μL of each reaction mixture with different DiLeu tags were combined at 1: 1: 1: 1 ratio and mixed well. To compare the relative abundance of each set of 4-plex DiLeu labeled metabolites, one channel from each 4-plex combination was selected and combined with 118-DiLeu tag labeled pooled serum samples. Ten μL of the mixture was taken for drying down and desalted using SCX Ziptips. Samples were dissolved in 15 μL 0.1% FA and 3 μL was injected into LC-MS for data acquisition with 3 technical replicates.

Liquid chromatography coupled with MS/MS data collection

Four-plex pooled samples were reconstituted in 0.1% formic acid before injection. The HPLC-MS/MS analysis was conducted using a Waters nanoACQUITY UPLC coupled with Thermo Q Exactive Orbitrap MS. The separation column was in-house made with an emitter tip and dimensions of 75 μm inner diameter × 15 cm length. The column was packed with 1.7 μm, 150 Å, ethylene-bridged-hybrid (BEH) C18 material (Waters, Milford, MA). Mobile phase A was water containing 0.1% formic acid, and mobile phase B was acetonitrile containing 0.1% formic acid. The flow rate was set as 0.3 μL/min, and the LC gradient was 55 min and set as follows: 0−10 min, 3%−30% solvent B; 10−30 min, 30−80% B; 30–30.5 min, 80%−95% B; 30.5− 40.5 min, 95% B; 40.5−41 min, 95%−3% B; 41–55 min, 3% B. Positive ionization mode was used and full MS scans were acquired from m/z 180 to 800 at a resolution of 60 k, automatic gain control (AGC) was set as 5× 10^5, and a maximum injection time as set as 30 ms. The top 20 precursors were selected for normalized collision energy (NCE) dissociation (NCE = 30) with an isolation window of m/z 1, fixed first m/z 110, dynamic exclusion of 5 seconds, charge exclusion of >2, and a resolution of 35 k.

Data Analysis

Raw data files were converted into mgf. format files via the msConvert.²⁶ Metandem was used to process three technical replicates of each 4-plex DiLeu labeled metabolite for quantification.²⁷ The average precursor mass shift due to labeling was 145.1273 Da. Data analysis parameters were optimized automatically in Metandem software. Output files with reporter ion information were merged. Among all the features detected in DiLeu labeled pooled serum (1:1:1:1 mixed after labeling), only the ratios of reporter ion intensity (116/115 and 117/115 and 118/115) between 0.67 and 1.5 were kept for further analysis. Then selected features detected in the serum mixture were compared with DiLeu labeled blanks and features that appeared in the blank along with the same retention time were excluded. The remaining features were used for calculating the monoisotopic molecular weight based on mass shift caused by labeling and then searched against the Human Metabolome Database (HMDB). Only putative metabolites with primary and secondary amine were kept, while putative metabolites without the amine group were excluded. To compare the relative abundance of each set of 4-plex DiLeu labeled metabolites, all the intensities of reporter ions were normalized based on the reporter ion intensities of the shared channel for comparison. Mann Whitney U test was performed to compare metabolite levels in different groups. ROC analysis in MedCalc (version 20.118) was performed to evaluate the sensitivity and specificity of biomarkers for detecting growing adenomas compared with adenomas of unknown growth status or compared with normal controls. Leave-one-out cross-validation was performed for different groups (control group vs. adenoma carrier group, control group vs. growing adenoma group, and growing adenoma group vs. unknown growth group) via R studio (R version 4.2.2).

RESULTS AND DISCUSSIONS

Patient demographics

We studied 43 patients in total here. Twenty patients screened normal by OC were set as control. Meanwhile, 23 patients with adenomas discovered by CTC got their blood drawn before and after polypectomy (Figure 1). Among the 23 patients carrying adenomas, 15 patients were classified histologically as high-risk and the remaining 8 patients were classified as low-risk cases. In the high-risk cases, 10 were classified as growing by longitudinal CTC analysis, one as static, and 4 as unknown growth status.^{6, 9} In these low-risk cases, 4 were classified as growing, 2 as static, and 2 as unknown status (Figure 1). More details about the patients and screening normal controls can be found in Table S1 and Table S2 or previous studies.^{9, 10}

Figure 1.

A summary of patient cases enrolled into this study. For patients carrying polyps, the polyps were examined and classified by standard histopathologic criteria as low-risk adenomas or high-risk (advanced) adenomas. Blood draw was performed before and after polypectomy for patients carrying tumor in this study. When available, tumors were classified as growing and static independently based on the longitudinal size profiles. The level of a biomarker of interest was compared between prepolypectomy and postpolypectomy sera.

Quantitative performance of 4-plex labeling strategy

The structure of the DiLeu isobaric tag is composed of a reporter group, a balance group, and an amine-reactive group (Figure S1A). The tag can react with the amine group of metabolites (Figure S1B) and introduce a mass shift of 145.1263 Da or 145.1283 Da (Figure S1C). Though 12-plex and even 21-plex DiLeu tags have been developed for relative quantification of peptides and/or amine-containing metabolites,^{18, 23} we only selected four tags that had the smallest mass value differences (only 0.002 Da). This could facilitate accurate mass matching and enhance confidence in identifying features of those unavailable metabolite standards. It is noted that the metabolite identifications mainly relied on metabolite standards and accurate mass matching. To evaluate the accuracy and reproducibility of relative quantitation of metabolite standards using the 4-plex DiLeu tags, two sets of metabolite standard mixtures with different molar ratios (1: 1: 1: 1 and 1: 2: 4: 8) were labeled by 4-plex DiLeu and combined. Combined labeled metabolites were purified and analyzed by nanoLC ESI-MS/MS analysis. Relative quantification of the metabolite was achieved by calculating the intensity ratios of DiLeu reporter ions (m/z 115.1, 116.1, 117.1, 118.1) (Figure S1C) generated in MS2 fragmentation. The results were summarized in the box plots in Figures 2A,B. To further evaluate the accuracy and reproducibility of the relative quantitation of metabolites in serum samples, two sets of serum flowthrough with different volumes of starting material (20: 20: 20: 20 μL and 5: 10: 20: 40 μL) were labeled by 4-plex DiLeu. The results were summarized in the box plots in Figures 2C,D. The median ratios measured among 21 metabolite standards were 1: 1.08: 0.98: 1.03 with a theoretical ratio of 1: 1: 1: 1, and 1: 2.52: 4.93: 7.56 with a theoretical ratio of 1: 2: 4: 8, respectively. The median ratio measured among 40 metabolites in pooled serum were 1: 1.08: 1.07: 1.09 with a theoretical ratio of 1: 1: 1: 1, and 1: 2.06: 4.01: 6.90 with a theoretical ratio of 1: 2: 4: 8, respectively. Both groups for labeling of serum metabolites showed satisfactory accuracy (within 14% error) for relative quantification. This result suggested that the amount of tags used here was suitable for labeling serum samples with starting volumes ranging from 5 μL to 40 μL. Therefore, the accuracy for relative quantification of metabolites in 20 μL serum was confirmed.

Figure 2.

Box plots of reporter ion ratios of DiLeu labeled metabolite standards at a theoretical 1: 1: 1: 1 ratio (A) and 1 : 2 : 4 : 8 ratios (B), and box plots of reporter ion ratios of DiLeu labeled serum metabolites at a theoretical 1: 1: 1: 1 ratio (C) and 1 : 2 : 4 : 8 ratios (D). Each box contains 21 data points (average of 3 replicates) from 21 metabolite standards for metabolite standards, and 40 data points for serum metabolites, obtained from LC-ESI-MS/MS analysis. Box denotes 25^th and 75^th percentiles; line within box denotes 50^th percentile; whiskers denote standard deviation.

Serum Metabolite Identification and Quantification.

A schematic illustration of the workflow used in this study is shown in Figure 3. Briefly, serum metabolites from patients or controls were collected after serum was passed through a 3 kDa MWCO filter. And 4-plex DiLeu tags were used for labeling secondary and primary amine-containing metabolites from each individual and then combined at a ratio of 1: 1: 1: 1. And relative abundance of each metabolite can be determined by comparing the relative intensity of reporter ions.

Figure 3.

Workflow for relative quantification of serum metabolites using DiLeu-labeling strategy.

Among all the features detected in DiLeu labeled pooled serum sample (1: 1: 1: 1 labeled and combined), we applied the criteria to filter: 1) the m/z values were not shown in DiLeu labeled blank samples or features had different retention time (>1 min) or had the same retention time, but the peak intensity was 100 fold lower than that in the serum samples; 2) the MS2 spectra of the features contained all the four reporter ions; 3) the ratio of reporter ions (116/115, 117/115, 118/115) ranged from 0.67 to 1.5; 4) the calculated mass of the features can match a secondary or primary metabolite in Human Metabolome Database (HMDB) (mass differences < 10 ppm). When we calculated the masses of unlabeled metabolites, we subtracted precursor masses by 145.1273 Da (for one-tag-labeled) or 290.2546 (two-tag-labeled), then the masses were searched against a human metabolite database (HMDB). Among hundreds of features detected in the serum samples, a total of 59 passed the filtering criteria. Furthermore, 40 features were detected in all the serum samples, including 18 features validated using metabolite standards (Table S3).

The 40 metabolites were assessed for statistical significance in detecting colorectal cancer from cancer-free cases. The identified metabolites were compared quantitatively among adenoma-free people (“screening normal”), people carrying adenomas of unknown growth, people carrying growing adenoma, and people carrying adenoma (including both unknown growth and growing adenoma). Among the validated metabolites, threonine, asparagine, and glutamine in patients with high-risk adenomas showed significantly decreased serum levels compared with screening normal cases (p-value <0.05 with Benjaminin-Hochberg to control the false discovery rate q at <0.05) (Figure 4). Consistently, threonine and glutamine also showed decreased serum levels in growing adenoma compared with screening normal. Glutamine showed a significant decrease in sera of patients with unknown growth adenoma compared with normal controls (Figure 4). Significant level changes of asparagine, glutamine and arginine were observed in adenoma carriers compared to screening normal controls (Figure 4). In contrast, arginine showed significantly enhanced levels in the sera of patients with low-risk adenoma compared with normal controls (Figure 4). Cysteine in the blood is not stable and can be oxidized to form cystine and levels of cystine may be affected during sample preparation. Therefore, cystine was excluded from discussion here. We also looked into the comparison of the ratios of post-/pre-polypectomy between different groups, including a comparison between unknown growth and growing adenoma, and a comparison between low-risk and high-risk adenoma, but we did not find significant differences between them.

Figure 4.

Relative ratio-to-pooled serum mixture for six comparison groups in four metabolites that showed statistically significant tumor-associated level changes. Blood draw was performed pre-polypectomy. Mann Whitney U test was performed. Asterisks represent the significance level (*P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001) across the different adenoma groups with Benjaminin-Hochberg to control the false discovery rate q at <0.05.

We then explore whether the levels of these altered metabolites reflected the total adenoma volume in a patient. Previous studies showed that levels of serum protein biomarkers F5, ITIH4, LRG1, and VTN in the Apc^Pirc/+ rat positively correlated with the number of colonic adenomas, while no significant correlation between the total volume of adenomas and the prepolypectomy levels of the four serum markers.^{9, 28} In this study, we found no significant correlation between the four metabolites (threonine, asparagine, glutamine, arginine) and adenoma volume (Table S4) by Spearman test. Furthermore, we asked if there were any correlations between these metabolites and protein biomarkers identified in previous work (Table 2 in ref 8). We generated a correlation heatmap for metabolites identified in this study and previously identified protein biomarkers (Figure 5A). Interestingly, we found that there was a strong positive correlation between ITIH4 and F5 (ρ >0.7) (Figure 5B). Moderate positive correlation was found between CRP and VTN (ρ >0.4) (Figure 5C). In contrast, there was a strong negative correlation between LRG1 and asparagine (-ρ <−0.7) (Figure 5D).

Figure 5.

Correlation between tumor volume and differential marker level. Correlation heatmap for tumor volume, protein biomarkers and metabolite markers (A). Correlation between ITIH4 and F5 (B). Correlation between CRP and VTN (C). Correlation between LRG1 and asparagine (D). A Spearman test was carried out for correlation between the monotonic rank orders of tumor volume, protein biomarker level and metabolite biomarker. The Spearman ρ values and their P values indicated the strength of the correlation between different biomarkers. The corresponding levels of each of the biomarkers were determined, relative to the pooled serum samples from normal control samples of each metabolite.

We assessed quantitatively the sensitivity and specificity for using these serum metabolic markers to detect growing adenomas compared with normal controls. Receiver operating characteristic (ROC) analysis showed that the nominal area under curve (AUC) values for these individual markers ranged from 0.829 (glutamine and threonine) to 0.871 (glutamine, threonine, asparagine, and arginine) (Figure 6A). When patients with growing adenomas were compared with patients with unknown growth of adenomas, the AUC values of the markers ranged from 0.905 to 0.925 (Figure 6B).

Figure 6.

ROC analysis showing sensitivity and specificity of panels of potential metabolite biomarkers for detecting growing adenomas compared with normal controls (A) or compared with adenomas of unknown growth status (B).

The identification of low-invasiveness biomarkers for early colorectal cancer detection is challenging. Although minimally invasive tests for occult fecal blood and tumor-derived DNA have been widely used for the detection of colorectal cancer, performance for detecting advanced colonic adenoma needs to be further improved.³ To increase the sensitivity and specificity in predicting colorectal cancer and improve the understanding of the pathological mechanism of colorectal cancer development, efforts have been made to discover protein and metabolite biomarkers. Recent studies have reported that potential metabolite biomarkers show dysregulated levels in the blood of colorectal cancer patients compared with healthy controls. However, most of these studies only focus on identifying biomarkers between colorectal cancer and healthy controls.¹¹ One of the studies grouped the patients based on the risk of colorectal cancer precursors (including conventional adenomas and serrated polyps) and found that lipid metabolism and metabolites may be involved in the early stage of serrated pathway- related colorectal carcinogenesis.¹³ In this study, we focused on investigating potential serum metabolite biomarkers that can distinguish healthy controls, and adenoma carriers with different growth rates or risks. This study is a follow-up of a previous work aiming at discovering serum protein biomarkers related to growing early colorectal adenomas.⁹ A more comprehensive profiling of serum content (including proteins and metabolites and even post-translational modifications of proteins) may enable enhanced detection of colorectal cancer and offer deeper insights into the pathological mechanisms of colorectal cancer. It would be interesting to investigate what power can emerge from using two or more modalities (metabolomics and proteomics) and then from combining different analytes from the same modality or from different modalities. Figure 6 reported cases of enhanced ROC scores by combining metabolite markers that were positively correlated with each other. A combination of orthogonal markers that were each correlated with a tumor phenotye but not with each other (with an absolute value of Spearman ρ less than 0.4) also showed enhanced ROC scores compared with each individual marker (Figure S2). In the ROC analysis performed in MedCalc, we used all the data set to generate ROC curves. We also evaluated cross-validated area under the ROC curve (AUC) using leave-one-out validation in R studio (Table S5).³⁰ Compared with unvalidated AUCs conducted using MedCalc, we saw a general decrease of the value in the cross-validated AUCs. We could find that some metabolites still showed decent performance in distinguishing different groups of patients in this case. For example, a combination of threonine and asparagine got a cross-validated AUC of 0.8 for control group and growing adenomas group. We also noticed that ROC analysis for growing adenoma group and unknown growth did not work well (Table S5). This result could be caused by the relatively small number in these two groups, 14 samples and 6 samples, respectively. Furthermore, paired prepolypectomy and postpolypectomy serum samples from colorectal cancer patients enabled us to investigate the metabolite profile changes before and after removing polyps.

In this work, we applied our in-house synthesized isobaric tags for the relative quantification of primary and secondary amine-containing metabolites. Among the 40 quantified 40 metabolites, we found that threonine, asparagine, glutamine, and arginine exhibited significant differences between the healthy controls and at least one colorectal cancer group (Figure 4). ROC analysis combining these metabolic markers showed decent specificity and sensitivity in detecting colorectal cancer. When we compared the metabolic markers with previously identified potential protein biomarkers in the shared patients, we found that LRG1 showed a strong negative correlation with asparagine. It would be interesting to validate if a correlation exists using a bigger sample size and investigate if any biological pathways are involved in this correlation once validated. A combination of protein biomarkers with metabolite biomarkers should improve the sensitivity and specificity for detecting early growing adenomas of colorectal cancer.

Here, we putatively quantified 40 amine-containing metabolites in the serum, which is a little bit too low compared with some other labeling strategies like dansylation labeling.³¹ Some reasons for this situation can be that 1) we excluded hundreds of m/z values shown in the DiLeu-labeled blank tube, inevitably removing some potential metabolites; 2) This method was based on MS2 quantification, so that it required the presence of higher concentrations of analytes for quantification since relative quantitation is achieved through comparison of reporter ions in tandem MS compared to MS1 quantification; 3) Because labeled metabolites are mainly in relative low mass range (m/z 200 – 500), co-isolation of some labeled metabolites for tandem mass fragmentation affected quantification accuracy, then we excluded these features; 4) Desalting tips used here could hold only very limited amounts of metabolites, therefore the total injection amount of metabolites for LC-MS/MS analysis was small. Though it is not the focus of this study, there is still some room for improving the DiLeu-based relative quantification method for metabolite quantification in serum samples. Although the performance of the 4-plex DiLeu tags utilized in this study cannot compete with dansylation labeling in identified metabolite numbers, the DiLeu tags are superior in its high throughpout, especially when 12-plex or even 21-plex DiLeu tags are applied for targeted primary/secondary amine-containing metabolites where metabolite standards are available.^{18, 23} Additionally, the tags only label primary/secondary amines, there is no doubt that other metabolites without the amine group cannot be identified here. Therefore, a combination of other types of chemical tags targeting different classes of metabolites or detecting sera metabolites using a label-free quantification strategy could expand the coverage of metabolites identified/quantified.

Enhancing prevention of colorectal cancer must be balanced by minimizing overdiagnosis. As only about 5% of all screening adults contain large polyps and the lifetime cancer risk,⁵ identifying the subset of early adenomas that are growing would be a key step to attenuate overdiagnosis. It is worth mentioning that this is a small cohort study with limited controls, so the confidence to claim the outcome is compromised. Although ROC analysis shows that some metabolites quantified here can distinguish samples from screening normal and adenoma carriers, there is still a lack of confidence to claim that these metabolites can be potential biomarkers for colorectal cancer because the sample size is small. Therefore, this study would benefit from further validation using a larger cohort of colorectal cancer patients with a classification of growth status. For example, with an estimated diagnostic accuracy (AUC) of 0.85 and marginal errors of 0.05 with 95% confidence level, the calculated sample sizes for each group of control and adenoma cases are 184.³² Although there is a long way to go before we know whether these metabolites can be validated as biomarkers by independent validation studies, this study illustrates a small step toward the direction of developing early detection in colorectal cancer. Hopefully, this work and strategies could improve early diagnosis of colorectal cancer and finally lead to a reduced death rate of colorectal cancer.

Supplementary Material

Supporting Information

The Supporting Information is available free of charge.

Table S1. An overview of CTC patient data;

Table S2. Colonoscopy/Cancer free cases;

Table S3. Metabolites identified using LC-MS/MS platform;

Table S4. Test for correlation between tumor volume and differential biomarker level.

Figure S1. General information of 4-plex DiLeu isobaric tags;

Figure S2. ROC analysis of markers for detecting colorectal cancer.

Data availability

This study is available at the NIH Common Fund’s National Metabolomics Data Repository (NMDR) website, the Metabolomics Workbench, https://www.metabolomicsworkbench.org where it has been assigned Study ID ST002420. The data can be accessed directly via its Project DOI: http://dx.doi.org/10.21228/M89M65. The study is scheduled to be released on 2024-06-12. Participant’s raw files will be available in a repository as required. The data will be publicly available once the manuscript is published.