Tissue Metabonomic Phenotyping for Diagnosis and Prognosis of Human Colorectal Cancer

Abstract

Colorectal cancer (CRC) is one of the leading causes of cancer-related death worldwide and prognosis based on the conventional histological grading method for CRC remains poor. To better the situation, we analyzed the metabonomic signatures of 50 human CRC tissues and their adjacent non-involved tissues (ANIT) using high-resolution magic-angle spinning (HRMAS) (1)H NMR spectroscopy together with the fatty acid compositions of these tissues using GC-FID/MS. We showed that tissue metabolic phenotypes not only discriminated CRC tissues from ANIT, but also distinguished low-grade tumor tissues (stages I-II) from the high-grade ones (stages III-IV) with high sensitivity and specificity in both cases. Metabonomic phenotypes of CRC tissues differed significantly from that of ANIT in energy metabolism, membrane biosynthesis and degradations, osmotic regulations together with the metabolism of proteins and nucleotides. Amongst all CRC tissues, the stage I tumors exhibited largest differentiations from ANIT. The combination of the differentiating metabolites showed outstanding collective power for differentiating cancer from ANIT and for distinguishing CRC tissues at different stages. These findings revealed details in the typical metabonomic phenotypes associated with CRC tissues nondestructively and demonstrated tissue metabonomic phenotyping as an important molecular pathology tool for diagnosis and prognosis of cancerous solid tumors.

Figure 1

Average 600 MHz ¹H HRMAS NMR spectra of ANIT (A), stage I CRC tumor (B) and stage IV CRC tumor (C). The region of δ 5.7–8.5 was vertically expanded 16 times compared with δ 0.8–4.2. Metabolite keys: 1, isoleucine; 2, leucine; 3, valine; 4, lactate; 5, threonine; 6, alanine; 7, lysine; 8, arginine; 9, proline; 10, glutamate; 11, methionine; 12, glutamine; 13, creatine; 14, choline; 15, glycine; 16, tyrosine; 17, phenylalanine; 18, scyllo-inositol; 19, lipid; 20, aspartate; 21, asparagine; 22, glutathione; 23, cysteine; 24, phosphorylcholine/glycerophosphocholine; 25, taurine; 26, myo-inositol; 27, phosphoethanolamine; 28, uracil; 29, cytosine; 30, isocytosine; 31, acetate; 32, fumarate; 33, inosine; 34, formate.

Figure 2

PCA scores plots obtained from NMR data of CRC tumor tissues at different stages (I–IV) with (A) or without (B) ANIT. ANIT (), stage I (), stage II (), stage III (), and stage IV ().

Figure 3. ROC curves determined using the cross-validated predicted Y-values of the ¹H NMR OPLS-DA models from CRC tumor and ANIT.

(A) ANIT vs CRC tumor, (B) stages I-II tumor vs stages III-IV tumor, (C) stage I tumor vs ANIT, (D) stage II tumor vs ANIT.

Figure 4

OPLS-DA scores (left) and coefficient-coded loadings plots (right) showing the discrimination between (A) ANIT () and CRC tumor () (n = 50, |r| > 0.29) and (B) stages I-II tumor () and stages III–IV tumor () (n = 22, |r| > 0.41). Metabolite keys are given in Fig. 1 and Table S1.

Figure 5. The ratios of metabolite changes for CRC tumor tissues at different stages (I-IV) against ANIT.

Figure 6. Fatty acid levels in ANIT and CRC tumor tissues.

*p < 0.05 when compared to the ANIT, Δ p < 0.05 when compared to low-grade (stages I-II) tumor tissues.