Soluble Axl is an accurate biomarker of cirrhosis and hepatocellular carcinoma development: results from a large scale multicenter analysis

Abstract

Patients with chronic liver disease (CLD) and cirrhosis are at high risk for hepatocellular carcinoma (HCC). Current diagnostic tools for HCC detection include imaging techniques and serum biomarkers such as α-fetoprotein (AFP). Yet, these methods are limited in sensitivity and specificity to accurately detect early HCC. Here we focused on the potential of soluble Axl (sAxl) as a biomarker in CLD patients by analyzing serum samples of 1067 patients and healthy controls from centers in Europe and Asia. We show that serum concentrations of sAxl were significantly increased at early (82.57 ng/mL) and later stages of HCC (114.50 ng/mL) as compared to healthy controls (40.15 ng/mL). Notably, no elevated sAxl levels were detected in patients with CLD including chronic viral hepatitis, autoimmune hepatitis, cholestatic liver disease, or non-alcoholic fatty liver disease versus healthy controls. Furthermore, sAxl did not rise in liver adenomas or cholangiocarcinoma (CCA). Yet, patients with advanced fibrosis (F3) or cirrhosis (F4) showed enhanced sAxl concentrations (F3: 54.67 ng/mL; F4: 94.74 ng/mL). Hepatic myofibroblasts exhibited an increased release of sAxl, suggesting that elevated sAxl levels arise from these cells during fibrosis. Receiver operating characteristic curve analysis of sAxl displayed a strongly increased sensitivity and specificity to detect both cirrhosis (80.8%/92.0%) and HCC (83.3%/86.7%) with an area under the curve of 0.935/0.903 as compared to AFP. In conclusion, sAxl shows high diagnostic accuracy at early stage HCC as well as cirrhosis, thereby outperforming AFP. Importantly, sAxl remains normal in most common CLDs, liver adenomas and CCA.

Keywords: biomarker; cirrhosis; fibrosis; hepatocellular carcinoma; soluble Axl.

Figure 1. sAxl levels in HCC patients

(A) sAxl serum concentrations in healthy controls (n = 75) compared to all HCC patients (n = 347) and in HCC patients either with cirrhosis (n = 272) or without cirrhosis (n = 35). (B) Analysis of sAxl serum concentrations in healthy controls (n = 75), early HCC (BCLC A: n = 45), intermediate HCC (BCLC B: n = 45), advanced HCC (BCLC C: n = 67) and end-stage HCC (BCLC D: n = 12). (C) Statistical comparison between HCC patients with cirrhosis and HCC patients without cirrhosis. Serum samples were diluted with LowCross-buffer^® (Candor, Germany) 1:200 and analyzed for sAxl levels by ELISA. Horizontal bars indicate median levels with interquartile ranges. Statistical significances of the differences between groups were evaluated with Mann-Whitney U test. Ns: not significant. *** p ≤ 0.001.

Figure 2. sAxl levels in fibrosis and CLDs

(A) sAxl serum concentrations in healthy controls (n = 75) compared to patients without fibrosis (F0: n = 54), with mild fibrosis (F1: n = 75), moderate fibrosis (F2: n = 80), advanced fibrosis (F3: n = 36) and cirrhosis (F4: n = 155). (B) sAxl serum levels of all HCC patients (n = 347) versus fibrosis and cirrhosis patients. (C) sAxl serum concentrations in healthy controls (n = 75) compared to patients with viral hepatitis (n = 26), and autoimmune hepatitis (AIH: n = 18). (D) sAxl serum levels of all HCC patients versus viral hepatitis and AIH patients. (E) Analysis of sAxl serum concentrations in healthy controls (n = 75), cholestatic liver disease (n = 30) and patients with non-alcoholic fatty liver disease (NAFLD; n = 116). (F) Statistical comparison between all HCC patients (n = 347) to cholestatic liver disease patients and NAFLD. Serum samples were diluted with LowCross-buffer^® (Candor, Germany) 1:200 and analyzed for sAxl levels by ELISA. Horizontal bars indicate median levels with interquartile ranges. Statistical significances of the differences between groups were evaluated with Mann-Whitney U test. Ns: not significant. ** p ≤ 0.01, *** p ≤ 0.001.

Figure 3. sAxl levels in CCA and liver adenoma patients

(A) Analysis of sAxl serum concentrations in healthy controls (n = 75), CCA patients (n = 44) and patients with benign liver adenoma (n = 11). (B) Statistical comparison between all HCC patients to CCA and adenoma patients. Serum samples were diluted with LowCross-buffer^® (Candor, Germany) 1:200 and analyzed for sAxl levels by ELISA. Horizontal bars indicate median levels with interquartile ranges. Statistical significances of the differences between groups were evaluated with Mann-Whitney U test. Ns: not significant. *** p ≤ 0.001.

Figure 4. Detection of HCC and cirrhosis by sAxl

(A) ROC curve of sAxl in healthy controls (n = 75) versus HCC (n = 347). (B) ROC curve of sAxl in healthy controls (n = 75) versus cirrhosis (F4, cirrhosis w/o HCC; n = 155). (C) ROC curve of sAxl, AFP and a combination of both in healthy controls (n = 28) versus HCC patients (n = 274). (D) ROC curve of sAxl, AFP and a combination of both in healthy controls (n = 28) versus cirrhosis patients (F4, cirrhosis w/o HCC; n = 24). Numbers in parentheses represent area under the curve.

Figure 5. sAxl release of liver myofibroblasts, mesenchymal and epithelial hepatoma cell lines and CCA cell lines

Evaluation of total sAxl release of seven liver myofibroblast cell lines (white), two mesenchymal hepatoma cell lines (dark gray), two epithelial hepatoma cell lines (gray) and three cholangiocellular carcinoma cell lines (black) by ELISA. Data are expressed as mean ± s.d.