Supervised learning reveals circulating biomarker levels diagnostic of hepatocellular carcinoma in a clinically relevant model of non-alcoholic steatohepatitis; An OAD to NASH

Abstract

Although cirrhosis is a key risk factor for the development of hepatocellular carcinoma (HCC), mounting evidence indicates that in a subset of patients presenting with non-alcoholic steatohepatitis (NASH) HCC manifests in the absence of cirrhosis. Given the sheer size of the ongoing non-alcoholic fatty liver disease (NAFLD) epidemic and the dismal prognosis associated with late-stage primary liver cancer there is an urgent need for HCC surveillance in the NASH population. Using serum levels of HCC biomarkers as vectors and biopsy-proven HCC or no HCC as outputs / binary classifier, a supervised learning campaign was undertaken to develop a minimally invasive technique for making a diagnosis of HCC in a clinically relevant model of NASH. Adult mice randomized to control diet or a fast food diet (FFD) were followed for up to 14 mo and serum level of a panel of HCC-relevant biomarkers was compared with liver biopsies at 3 and 14 mo. Both NAFLD Activity Score (NAS) and hepatic hydroxyproline content were elevated at 3 and 14 mo on FFD. Picrosirius red staining of liver sections revealed a filigree pattern of fibrillar collagen deposition with no cirrhosis at 14 mo on FFD. Nevertheless, 46% of animals bore one or more tumors on their livers confirmed as HCC in hematoxylin-eosin-stained liver sections. In this training set, receiver operating characteristic (ROC) curves analysis for serum levels of the HCC biomarkers osteopontin (OPN), alpha-fetoprotein (AFP) and Dickkopf-1 (DKK1) returned concordance-statistic/area under ROC curve of ≥ 0.89. Serum levels of OPN (threshold, 218 ng/mL; sensitivity, 82%; specificity, 86%), AFP (136 ng/mL; 91%; 97%) and DKK1 (2.4 ng/mL; 82%; 81%) diagnostic for HCC were confirmed in a test set comprising mice on control diet or FFD and mice subjected to hepatic ischemia-reperfusion injury. These data suggest that levels of circulating OPN, AFP and DKK1 can be used to make a diagnosis of HCC in a clinically relevant model of NASH.

Fig 1. Fatty liver disease continuum.

Accumulation of lipid droplets in the liver or steatosis can lead to steatosis+inflammation or NASH which can progress to scarring of the liver or NASH with fibrosis. The next stage cirrhosis is accompanied by significant scarring of the entire liver and presents enhanced risk for HCC. However in the setting of fatty liver disease, HCC can also occur prior to cirrhosis (large red arrow).

Fig 2. Hepatic phenotype on FFD.

Compared to livers from animals on a control diet for 3 mo, livers from FFD (3 mo) animals had larger mass and liver-to-body mass ratio (#, p < 0.01 vs. control).

Fig 3. FFD-induced HCC.

Of the 76 animals entered into the training set, 65 animals bore no tumors on the liver. Of the 24 animals on FFD for 14 mo, 11 exhibited liver tumors which manifested as one or more polyp-like structures on the liver. Animals on FFD for 14 mo had larger liver and liver to body mass ratios compared to the control diet cohort. Animals with liver tumors (labeled HCC FFD) had the highest values for these endpoints (#, p < 0.01 vs. control; *, p < 0.01 vs. FFD).

Fig 4. ROC curves for liver mass and liver-to-body mass.

ROC curves showing the performance of liver mass as a diagnostic for HCC (AUROC 0.83) and liver-to-body mass ratio for HCC (AUROC 0.83). Data from the entire training set of 76 animals were used for plotting these curves.

Fig 5. Liver histopathology.

(Top left) At 3 mo on the diets, NAS was elevated in the FFD cohort vs. the control diet cohort (#, p < 0.01 vs. control). (Bottom left) At 14 mo on the diets, NAS was elevated in the FFD cohort vs. the control diet cohort (#, p < 0.01 vs. control). Animals exhibiting liver tumors (HCC FFD) exhibited highest NAS (*, p < 0.01 vs. FFD). (Top right) Representative H&E stained liver sections (10 X) from the 14 mo control, FFD and HCC FFD cohorts are shown. Steatosis was prominent in the FFD cohort. tTe HCC FFD cohort showed a trabecular growth pattern of atypical hepatocytes and clusters of multinucleated hepatocytes with a distinct margin (arrows) between the cancerous and non-cancerous parenchyma. (Bottom right) Representative H&E stained liver sections (20 X) from HCC FFD livers showing aspects of NASH (steatosis and inflammation) and HCC (arrows).

Fig 6. Liver hydroxyproline.

(Top left) Animals randomized to FFD for 3 mo exhibited ~ 50% higher liver hydroxyproline content compared with animals randomized to a control diet for an equivalent time period (#, p < 0.01 vs. control). (Bottom Left) Animals on FFD for 14 mo but free of HCC had >4-fold higher liver hydroxyproline content than animals on a control diet for the same time interval (#, p < 0.01 vs. control). Liver hydroxyproline content was highest in the 14 mo HCC FFD cohort (*, p < 0.01 vs. FFD). (Right) ROC curve for liver hydroxyproline as a diagnostic for HCC yielded an AUROC of 0.9.

Fig 7. FFD model of NASH with fibrosis.

(Top) Representative H&E (left) and Picrosirius red (right) stained liver sections (10 X) from an animal on a control diet for 14 mo. (Middle) Representative H&E (left) and Picrosirius red (right) stained liver sections (10 X) from an HCC FFD animal that was on FFD for 14 mo. (Bottom) 20 X section from that animal (left, H&E; right Picrosirius red). While a filigree pattern of Picrosirius red staining was observed in the HCC FFD cohort, there was no histological evidence of cirrhosis.

Fig 8. ROC curves for HCC biomarkers.

Performance of the biomarkers listed in Table 1 as diagnostics for HCC are shown as ROC curves. Only OPN, AFP and DKK1 had AUROCs ≥ 0.8.

Fig 9. Diagnostic performance of HCC biomarkers in test set.

(Top) Representative H&E-stained livers sections (20 X) from a Kcnn4/ KCa3.1 null mice mouse on a control diet (left), a C57BL/6 mouse on FFD for 7 mo (middle) exhibiting steatosis and CD-1 mouse submitted to hepatic ischemia and 24 hr reperfusion (right) exhibiting necroinflammatory injury. There was no evidence of HCC in the test set. (Bottom) Diagnostic accuracy of OPN, AFP and DKK1 in the test set.