Quantitative Histopathology of Fibrosis in an Animal Model of NASH

The hepatic fibrosis has been the most challenging aspect in the animal models for nonalcoholic steatohepatitis (NASH). Although it could induced by nutritional (high-fat died with high cholesterol, methionine and choline deficient diet, etc), chemical (carbon tetrachloride, thioacetamide, α-naphthylisothiocyanate etc.), and surgical (bile duct ligation) means, the clinical relevance of the resulting fibrosis remains an open question. The Center of Excellence for NASH at the WuXi AppTec (Shanghai) has established these fibrosis models and tested compounds in the NASH clinical trials for pharmacological validation. From these studies, we concluded that the mouse model based on high-fat diet (HFD) induced steatosis and carbon tetrachloride (CCL4) induced fibrosis displays histopathological features that bear resemblance to those observed in human NASH livers.
Figure 1. Fibrosis in healthy control animals (fed on regular diet), those fed on HFD, treated with CCL4, and in the HFD+CCL4 NASH model without and with OCA treatment.
In the HFD+CCL4 model for NASH, the CCL4 component is largely responsible for the induction of fibrosis. The combination of CCL4 with HFD induces an extra level of fibrosis (ΔF). It has been observed that obeticholic acid (OCA) improves the fibrosis in this model to the extent that the remaining fibrosis is at the similar level as that induced by CCL4 alone (Figure 1). Given that OCA has no significant efficacy on the fibrosis induced by CCL4 alone, we suspected that ΔF is specific to the NASH model, hence alleviated by the OCA treatment. In our latest effort to understanding the hepatic fibrosis in the CCL4 and the HFD+CCL4 models in mouse, we collaborated with HistoIndex/Choutu and conducted a retrospective study of artificial intelligence (AI)-based quantitative histopathology of hepatic fibrosis.
HistoIndex has successfully developed an AI-based approach of two-photon excitation (PTE) imaging coupled with second harmonic generation (SHG) imaging. The automated SHG/TPE microscope identifies collagen (key component of fibrosis) and the surrounding structures, in the SHG and TPE channels respectively, directly from unstained tissue slides (formalin-fixed and paraffin-embedded). In the case of liver (biopsy) samples, a proprietary AI algorithm recognizes and segregates the liver sections into different sub-anatomical regions. The fibrosis features are then identified and quantified from the SHG channel and, at the same time, other histopathological features in different regions are gathered from the TPE channel. The coupled SHG/TPE imaging, aided by AI-based analysis, affords a much less biased assessment of histopathology that contains quantitative and sub-anatomical information.
We selected four animal groups from several independent experiments for the retrospective study, including animals fed on HFD and treated with CCL4 for induction of fibrosis, and NASH model animals without and with OCA treatment (represented in Figure 1, 20 animals for each group). The SHG scanning results are shown in Figure 2. A basal level of total fibrosis (regardless of structure or location) is detected in the livers of animal fed on HFD, which is similar to that observed in the health control animals (fed on regular diet) (data not shown). CCL4 induces the total fibrosis that is significantly higher than the HFD group (~1.6% vs. ~0.6% SHG of total liver area). In the NASH model, the total fibrosis level is higher than CCL4 induction, by ~0.5% SHG. Consistent with the result shown in Figure 1, OCA reduces the total fibrosis in the model to the level which is comparable with that in CCL4 induction (p >0.05). The induction of the extra fibrosis in the NASH model is statistically significant when compared with the CCL4 induction alone, whereas the improvement of fibrosis by OCA is also significant (Figure 2).
Figure 2. Quantitative histopathology by SHG/TPE. SHG quantification of the total fibrosis in HFD, CCL4, NASH and NASH + OCA groups, and their distributions in three sub-anatomical regions, PS, perisinusoidal; PT, portal tract; and CV, central vein. The SUM group is the sum of PS, PT and CV. Statistical significance, p value, *<0.05; **<0.01 and ***<0.001.
The distribution of fibrosis in three sub-anatomical regions (perisinusoidal, PS; portal tract, PT; and central vein, CV) is then analyzed (Figure 2). In the HFD animals, the basal levels of total fibrosis are observed in the three regions, with quantitative difference. CCL4 induces similar levels of total fibrosis in all regions. However, in the NASH model, the PS region contains the highest level of fibrosis, which is significant when compared with the CCL4 induction in the same region. OCA alleviates significantly fibrosis in all three regions, and PT region being most affected, as the fiber level drops to a level lower than that induced by CCL4, with statistical significance (Figure 2).
Figure 3. Quantitative histopathology by SHG/TPE. The quantification of the fiber strings in HFD, CCL4, NASH and NASH + OCA groups, and their distributions in three sub-anatomical regions, PS, perisinusoidal; PT, portal tract; and CV, central vein. The SUM group is the sum of PS, PT and CV. Statistical significance, p value, *<0.05; **<0.01 and ***<0.001.
The SHG/TPE quantitative histopathology allows identification of fiber strings in the three regions (Figure 3). The number of fiber string is the highest in the PS region in all animals groups (Figure 3), in contrast to the SHG fibrosis, which is distributed equally in the three regions (Figure 2), indicating that the fibrosis in the perisinusoidal region is better structured than two other regions; i.e., a large portion of the fibrosis in the PT and CV regions are not structured fiber strings.
Compared with CCL4 induction alone, the fiber strings in the NASH model are significantly higher in number, which are significantly reduced by the OCA treatment. This pattern is largely preserved in the PS region. However, the fiber strings in the PT and CV regions are not statistically different in the CCL4 and the NASH animals. OCA reduces significantly the fibers strings in the PT but not the CV regions (Figure 3).
Collectively, the results of quantitative histopathology indicate that the fibrosis level in the NASH model is significantly higher than that induced by CCL4 alone, and that OCA alleviates the overall fibrosis in the PS, PT and CV regions of the livers. However, the fiber strings in the PS region are the primary target for the anti-fibrotic effect of OCA, and those in PT region to a less extent. 
Figure 4. Histopathology of hepatic fibrosis in the HFD+CCL4 model alone (A) and with OCA treatment (B). The blue arrows indicate bridging fibrosis between the portal triads, and the green in the portal tract regions (A), both of which are suppressed by the OCA treatment (B).
In the HFD+CCL4 model for NASH, a significant portion of the hepatic fibrosis is perilobular (mainly in zone 1); i.e., occurring around hepatic lobules, and bridging between portal triads is pronounced (Figure 4-A). However, fibrotic septa is not common in this model. When this model is used for efficacy test, compounds are usually administered at the same time CCL4 is injected intraperitoneally. Administered in this fashion, the OCA treatment prevents the bridging of perilobular fibers (Figure 4-B). Further the fibrosis in the portal tract regions are also reduced (Figure 4-B). These observations are largely in agreement with the quantitative histopathology by the SHG/TPE imaging and AI analysis, supporting the notion that the HFD+CCL4 model displays histopathological features of clinical relevance to NASH. This is further supported by the observation of “chicken-wire” fibrosis in the PS region of the HFD+CCL4 model, but not the CCL4 model (data not shown) – “chicken-wire” in the PS region is the characteristic pattern of fibrosis in NASH patients.
The hepatic fibrosis in the HFD+CCL4 model (i.e., NASH model) is different from that induced by CCL4 alone. In particular, the fibrosis in the perisinusoidal region of the NASH model is significantly elevated when compared with the same fibrosis in the CCL4 model. Further, perilobular bridging between portal triads is significantly more pronounced in the NASH model than in the CCL4 model (data not shown). These differences may account for the ΔF described above (Figure 1), and are the primary targets of the anti-fibrotic efficacy of OCA.
← Return to Resources