A two-gene blood test for methylated DNA sensitive for colorectal cancer

Abstract

Background: Specific genes are methylated with high frequency in colorectal neoplasia, and may leak into blood. Detection of multiple methylated DNA biomarkers in blood may improve assay sensitivity for colorectal cancer (CRC) relative to a single marker. We undertook a case-control study evaluating the presence of two methylation DNA markers, BCAT1 and IKZF1, in circulation to determine if they were complementary for detection of CRC.

Methods: Methylation-specific PCR assays were developed to measure the level of methylated BCAT1 and IKZF1 in DNA extracted from plasma obtained from colonoscopy-confirmed 144 healthy controls and 74 CRC cases.

Results: DNA yields ranged from 2 to 730 ng/mL plasma (mean 18.6ng/mL; 95% CI 11-26 ng/mL) and did not correlate with gender, age or CRC status. Methylated BCAT1 and IKZF1 DNA were detected in respectively 48 (65%) and 50 (68%) of the 74 cancers. In contrast, only 5 (4%) and 7 (5%) controls were positive for BCAT1 and IKZF1 DNA methylation, respectively. A two-gene classifier model ("either or" rule) improved segregation of CRC from controls, with 57 of 74 cancers (77%) compared to only 11 of 144 (7.6%) controls being positive for BCAT1 and/or IKZF1 DNA methylation. Increasing levels of methylated DNA were observed as CRC stage progressed.

Conclusions: Detection of methylated BCAT1 and/or IKZF1 DNA in plasma may have clinical application as a novel blood test for CRC. Combining the results from the two methylation-specific PCR assays improved CRC detection with minimal change in specificity. Further validation of this two-gene blood test with a view to application in screening is now indicated.

Fig 1. Methylation signal versus recovered DNA and subject age.

Correlation plots of average Ct signals measured in the BCAT1 (black triangles) and IKZF1 (white triangles) methylation assays versus (A) mass of recovered bisulphite converted DNA per mL plasma or (B) age of subjects at the time of blood draw. Assay replicates with ‘no signal’ were assigned the arbitrary Ct value of ‘50’ for depiction purposes.

Fig 2. Detection of methylated BCAT1 and IKZF1 DNA in plasma specimens.

BCAT1 (top panel, black triangles) and IKZF1 (bottom panel, white triangles) methylation specific assays were used to assess the appearance of circulating methylated BCAT1 and IKZF1 DNA in 218 plasma specimens including 144 healthy controls and 74 cancers. Mann-Whitney t-tests were performed on calculated positivity rates for each phenotypic class to derive the P-values. The calculated 95% confidence intervals (95% CI) are indicated.

Fig 3. Discriminatory performance of one-gene versus two-gene testing.

The sensitivity, specificity and discrimination accuracy values for the BCAT1, IKZF1 and combined two-gene assays were calculated using the true and false positivity rates measured in 144 healthy control and 74 cancer cases (A) and subsequently plotted on Receiver Operator Characteristic (ROC) space plots (B). Single-gene assays: A plasma specimen was qualitatively called ‘positive’ if at least one PCR replicate resulted in a Ct signal. Two-gene assay: A specimen was qualitatively called ‘positive’ if at least one replicate in either the BCAT1 or IKZF1 assays produced a Ct value.

Fig 4. Mass of methylated BCAT1 and IKZF1 DNA in circulation.

(A) Correlation plot of average Ct values (log(Ct)) measured by the BCAT1 or IKZF1 assays from 144 healthy controls (black circles) and 74 CRC (white circles). PCR replicates with no signal were assigned an arbitrary Ct value of 50 for depiction purposes. Square quadrants (a) to (d) represent: (a) 15 BCAT1 negative but IKZF1 positive cases; (b) 42 BCAT1 and IKZF1 positive cases including 41 cancer and 1 control; (c) 11 BCAT1 positive but IKZF1 negative cases; (d) 150 BCAT1 and IKZF1 negative cases. A linear regression analysis was performed on 38 of the 41 BCAT1 and IKZF1 positive CRC cases (quadrant b), omitting three CRC outliers. Broken diagonal lines indicate the 95% CI range. The calculated R square value is shown. The mass (ng) of methylated BCAT1 (B) and IKZF1 (C) DNA was calculated and plotted as mass of methylation per mL plasma drawn from 144 control cases, 4 Stage I, 28 Stage II, 23 Stage III, 8 Stage IV (the 8 CRC cases with unknown staging omitted). Average and median mass levels in the five phenotypic classes are indicated below the graphs. Mann-Whitney t-test performed on median values, ***: p value < 0.05, ns: non-significant, as compared with the 144 healthy control cases.

Fig 5. Estimation of class probabilities.

The estimated average mass of (A) BCAT1 and (B) IKZF1 DNA in each of the triplicate assays containing DNA isolated from the equivalent of 0.5mL plasma from healthy controls (black lines, BCAT1: n = 5, IKZF1: n = 2*), early stage cancer (blue lines, Stage I+II, BCAT1: n = 19, IKZF1: n = 17) and late stage cancer (red lines, Stage III+IV, BCAT1: n = 22, IKZF1: n = 26) were used to calculate the empirical probability density plots, omitting ‘no signal’ results (broken lines). Solid lines: Population means (μ) and standard deviations (σ) were calculated assuming a normal distribution for each of the three phenotypic classifications. *Five of the seven positive IKZF1 control cases had Ct signals within the last 5 cycles of PCR, therefore no mass estimation. (C) Cumulative probability of a plasma sample with BCAT1 or IKZF1 DNA methylation levels equal to or greater than 20-, 50- or 100pg methylated DNA coming from a patient with the indicated phenotypic classification; the mass thresholds were determined as the area under the curve (solid lines in panels A and B) for ln(pg) values greater than 3, 3.9 and 4.6, respectively (see dotted vertical lines and arrows) in Panels A and B. Likelihood ratios relative to healthy controls are shown in brackets.