Serum metabolomic analysis identified serum biomarkers predicting tumour recurrence after Bacillus Calmette-Guérin therapy in patients with non-muscle invasive bladder cancer

Abstract

Background: Metabolomic research and metabolomics-based biomarkers predicting treatment outcomes in bladder cancer remain limited.

Objective: We explored the serum metabolites potentially associated with the risk of recurrence after intravesical Bacillus Calmette-Guérin (BCG) therapy.

Methods: Two independent cohorts, a discovery cohort (n = 23) and a validation cohort (n = 40), were included in this study. Blood was collected before the induction of BCG therapy (pre-BCG blood; both discovery and validation cohorts) and after six doses of BCG (post-BCG blood; only discovery cohort). Metabolome analysis of serum samples was conducted using capillary electrophoresis time-of-flight mass spectrometry. The endpoint was intravesical recurrence-free survival, which was analysed using Kaplan-Meier estimates, the log-rank test, and the Cox proportional hazard model.

Results: Of the 353 metabolites quantified, nine (2.5%) and four (1.1%) were significantly upregulated and downregulated, respectively. The heatmap of hierarchical clustering analysis and principal coordinate analysis for the fold changes and in serum metabolites differentiated 10 recurrent cases and 13 non-recurrent cases in the discovery cohort. A metabolome response-based scoring model using 16 metabolites, including threonine and N6,N6,N6-trimethyl-lysine effectively stratified the risk of post-BCG recurrence. Additionally, pre-BCG metabolome-based score models using six metabolites, octanoylcarnitine, S-methylcysteine-S-oxide, theobromine, carnitine, indole-3-acetic acid, and valeric acid, were developed from the discovery cohort. Univariate and multivariate analyses confirmed a high predictive accuracy in the validation and combination cohorts.

Conclusions: We demonstrated that numerous types of serum metabolites were altered in response to intravesical BCG and developed high-performance score models which might effectively differentiated the risk of post-BCG tumour recurrence.

Keywords: BCG; Bacillus Calmette-Guerin; metabolites; metabolomics; mycobacterium bovis; prognosis; recurrence; serum; urinary bladder neoplasms.

Figure 1.

Study overview of the patient cohorts and analysis. This study consisted of three phases: enrolment, serum metabolomic analysis, and the development of prognostic models. Two cohorts, the discovery cohort and the validation cohorts were enrolled in this study. NMIBC, non-muscle invasive bladder cancer; BCG, Bacillus Calmette–Guérin; MetRes score, Metabolite response score; MetPreBCG score, Metabolite Pre-BCG score.

Figure 2.

BCG-induced change of serum metabolites and risk of post-BCG recurrence. A, The upper heatmap visualisation of hierarchical clustering analysis of changes in serum metabolites before and after intravesical BCG therapy. Lower heatmap visualisation of serum metabolites in which FC was associated with outcome after induction of BCG therapy. Sixteen of the 353 metabolites were identified as potential prognostic factors. In the upper 14 metabolites, a low FC was associated with a higher risk of recurrence. Among the two lowest metabolites, high FC was associated with a higher risk of recurrence. The red and blue blocks indicate upregulation and downregulation in the post-BCG group compared to the pre-BCG group, respectively. B, Score plot of principal coordinate analysis (PCoA) for response in serum metabolomic profiles of the non-recurrent (open circles on a red ellipse) and recurrent (black circles on a blue ellipse) groups. C, Metabolite Response score (MetRes-score; full score = 100) was calculated as the sum of the risk scores, as shown in Table 3. The scatter plot shows the MetRes scores of 23 cases, consisting of 13 nonrecurrent cases (white circles) and 10 recurrent cases (black circles) in the discovery cohort. The median age was 43 years, and the interquartile range was 23–66. D, Recurrence-free survival curves according to two-group classification into patients with low MetRes score ≤ 43 or with high MetRes score > 43 and according to three-group classification into patients with low MetRes score ≤ 23, intermediate MetRes score, or high MetRes score > 66.

Figure 3.

External validation of the developed MetPreBCG-score models. A, MetPreBCG-score model 1 used serum levels of six metabolites: octanoylcarnitine, S-methylcysteine-S-oxide, theobromine, carnitine, indole-3-acetic acid, and valeric acid. The MetPreBCG score model 1 (full score = 26) was calculated as the sum of the risk scores, as shown in Table 3. The recurrence-free survival curves of the two-group classification (upper) and three-group classification (lower) are shown separately. The left panels show the data of the discovery cohort, and the right panels show those of the validation cohort. B, MetPreBCG-score model 2 used serum levels of carnitine, indole-3-acetic acid, and valeric acid. The MetPreBCG score model 2 (full score = 13) was calculated as the sum of the risk scores, as shown in Table 3. The recurrence-free survival curves of two-group classification (upper) and three-group classification (lower) are shown separately. The left panels show the data of the discovery cohort, and the right panels show those of the validation cohort.