Ki67 expression at Kasai portoenterostomy as a prognostic factor in patients with biliary atresia

Background Biliary atresia is a rare paediatric biliary obliteration disease with unknown aetiology, and is the most common indication for paediatric liver transplantation (LT). However, no consensus for predicting Kasai portoenterostomy (KP) outcomes using liver histological findings exists. Ki67 is a popular biomarker for measuring and monitoring cellular proliferation. Methods Ki67 (clone, MIB‐1) liver parenchyma expression was measured by immunohistochemical staining of samples from living donors and patients with biliary atresia to assess its value in predicting outcomes after KP. Results Of 35 children with biliary atresia, 13 were native liver survivors (NLS), 17 were non‐NLS, and five had primary LT. The median proportion of Ki67 immunostained areas in donors and patients with biliary atresia at KP was 0·06 and 0·99 per cent respectively. Univariable analysis identified a high proportion of Ki67 areas, high Ki67 cell numbers and high Ki67‐positive/leucocyte common antigen‐positive cell numbers at KP as significant predictors of poor native liver survival after KP (hazard ratio 9·29, 3·37 and 12·17 respectively). The proportion of Ki67 areas in the non‐NLS group was significantly higher than that in the NLS group (1·29 versus 0·72 per cent respectively; P = 0·001), and then decreased at LT (0·32 per cent versus 1·29 per cent at KP; P < 0·001). Conclusion This study has demonstrated the clinical data and time course of Ki67 expression in patients with biliary atresia. High Ki67 expression at KP may be an important predictor of native liver survival following the procedure.


Introduction
Biliary atresia is a rare paediatric disease characterized by an obliterative cholangiopathy of unknown pathogenesis 1 . Kasai portoenterostomy (KP) is usually performed during early infancy, but subsequent liver transplantation (LT) is necessary in failed cases. Biliary atresia is the most common indication for paediatric LT, and accounts for up to 50 per cent of paediatric LTs globally 1 . Although studies have tested histological variables to predict outcomes following KP, the results have been contradictory and no consensus exists on the most appropriate histological finding, such as the number or size of biliary ductules in the remnant bile duct [2][3][4][5] , the presence of giant cell transformation [6][7][8] , the degree of bile duct inflammation [8][9][10][11] , or the amount of liver fibrosis and ductular reaction 6,[12][13][14][15] . Prediction for patients with biliary atresia. Thus, finding a good prognostic biomarker is important.
Ki67 has been a popular biomarker for measuring and monitoring tumour proliferation for many years, especially in breast cancer 16 . Expression of Ki67 varies throughout the cell cycle, reaching a peak during mitosis and being absent during the G0 stage of the cell cycle 16 . The present authors hypothesized that advanced liver damage at KP might result in a poor outcome. In view of the regeneration potential of the liver, cellular turnover is expected to be faster than normal when the liver is damaged [17][18][19][20] . Thus, Ki67 may be an objective biomarker with which to predict prognosis after KP. Only one report 21 has been published on the association between Ki67 expression in bile duct cells and clinical outcome in patients with biliary atresia, and there appear to be no reports on the relationship between Ki67 expression in other liver cells and clinical outcome in these patients. There is also a lack of studies reporting on the time course of Ki67 expression according to liver functioning. The aim of this study was to assess Ki67 expression as a histological, prognostic and liver functional marker after KP in patients with biliary atresia.

Methods
This retrospective study used data from patients with biliary atresia who had undergone KP and/or LT in Kumamoto University Hospital, Kumamoto, Japan, and from living LT donors between January 2005 and December 2016. The living LT donors are described as 'donors' and those who had primary LT are described as 'primary patients' in this article.
All clinical data were obtained from electronic medical charts. Liver specimens collected from the periphery of native livers by wedge resection during surgery were used for liver histological examination in clinical practice; in the present study, the remaining specimens were used for research purposes. The fibrosis score of liver specimens was determined using the new Inuyama classification 22 : F0, no portal fibrosis; F1, fibrous portal expansion; F2, bridging fibrosis; F3, bridging fibrosis with lobular distortion; and F4, cirrhosis.
The institutional review board at Kumamoto University Hospital approved this retrospective study (number 371). Written informed consent was obtained from each patient (if they could understand the study aim) and their parents, or only from parents (when patients were too young to understand the aim).

Image processing and cell counting
To quantify the proportion of immunostained areas per field (described in this study as the 'area proportion'), using ImageJ 1.46 (National Institutes of Health, Bethesda, Maryland, USA), three non-overlapping, randomly selected areas (except portal areas) were viewed for each patient; for CK19-immunostained sections, the portal areas were included. Immunostained sections were photographed with a microscope (BX51; Olympus, Tokyo, Japan) at 40× magnification for CK19 and 200× magnification for the others. Images of the immunostained sections were converted into 8-bit greyscale images that had only colour intensity. Threshold values for positive signals of the selected images were adjusted (from 0 to 255), and were set at 100 for Ki67 and CK19, and 120 for CD163. This was because compensated images were more reflective of original images. The area proportion was then calculated. To quantify immunostaining by cell counting, six non-overlapping areas (except portal areas) were selected randomly for the analyses. Immunostained sections were photographed at 400× magnification. Two pathologists, blinded to patient information, evaluated all sections.

Statistical analysis
All data are presented as median (range) values. The Mann-Whitney U test was used for comparisons between groups, and the Wilcoxon test for comparisons between two paired groups. To estimate the predictors of native liver survival after KP, receiver operating characteristic (ROC) curves were generated for each clinical parameter generally used in clinical practice and for Ki67 values, and the areas under the ROC curves (AUROCs) were compared. Cut-off values were determined by the Youden index based on the ROC curves, and patients with biliary atresia were divided into two groups on the basis of those values. Kaplan-Meier curves were generated for native liver survival rates, and groups were compared with the log rank test. Spearman's rank test was used to study associations between two variables. All tests were two-sided, and P < 0⋅050 was considered statistically significant. Statistical analyses were performed with GraphPad Prism ® 7 (GraphPad Software, San Diego, California, USA).

Results
At October 2018, of 35 children with biliary atresia, 13 were native liver survivors (NLS) and 17 were non-NLS; these 17 patients underwent both KP and the subsequent LT at the authors' institution. The remaining five patients were primary patients. Table 1 presents basic clinical and biochemical data for patients with biliary atresia, excluding primary patients. Nine patients were boys and 21 were girls. Their median gestational age, birthweight and age at KP were 39 (31-41) weeks, 2877 (940-4250) g and 69 (27-143) days respectively. Twenty-nine patients had type III biliary obstruction 1 , and one patient had type I. The jaundice clearance rate was 53 per cent (16 of 30). Table 2 presents basic clinical and biochemical data for patients with biliary atresia in the NLS, non-NLS (at KP), primary patient and non-NLS (at LT) groups.
Univariable analysis was performed to explore the risk factors associated with native liver survival after KP. High Ki67 area proportion, high Ki67 cell number, high Ki67+/LCA+ cell number, low platelet count, high PT-INR and low duration of surgery were significant predictors of poor native liver survival ( Table 4; Fig. S1d-f , supporting information). No association was found between survival and other clinical parameters, Ki67+/HepPar1+ cell number or Ki67+/αSMA+ cell number. Livers with high Ki67 area proportions had a significantly worse native liver survival rate than those with low Ki67 area proportions (hazard ratio (HR) 9⋅29, 95 per cent c.i. 3⋅47 to 24⋅91; P = 0⋅008) (Fig. S1d, supporting information). Livers with a high proportion of Ki67+/LCA+ cells had a significantly worse native liver survival rate than those with a low proportion (HR 12⋅17, 3⋅92 to 37⋅78; P = 0⋅002) (Fig. S1f , supporting  information).

Comparison of Ki67 expression in native liver survivors and non-survivors
The Ki67 area proportion, identified as a significant predictor of poor native liver survival in univariable analysis, was compared in NLS and non-NLS groups. The median Ki67 area proportion was significantly higher in the non-NLS group (1⋅29 per cent versus 0⋅72 per cent in the NLS group; P = 0⋅001) (Fig. 2a). The median Ki67 cell number was also significantly higher in the non-NLS group (489 versus 374 cells/mm 2 respectively; P = 0⋅036) (Fig. 2b).   Numbers of Ki67+/LCA+ cells, also identified in univariable analysis as a significant prognostic predictor in livers with biliary atresia, were then compared. Numbers of Ki67+/LCA+ cells were higher in the non-NLS than in the NLS group, but the difference was not significant (P = 0⋅130) (Fig. 2c). Numbers of Ki67+/HepPar1+ and Ki67+/αSMA+ cells were also higher in the non-NLS group, but again the differences were not significant (P = 0⋅307 and P = 0⋅404 respectively) (Fig. 2d,e).

Change in Ki67 expression between Kasai portoenterostomy and liver transplantation
To explore the association between Ki67 expression and the time course after KP, Ki67 expression was examined in donor livers (as normal livers) and livers with biliary atresia (Fig. 3a), obtained at KP (30 livers), primary LT (5) or LT (17). LT specimens were obtained at the time of LT from patients with biliary atresia who had undergone KP in the authors' institution. Median patient age at LT was 209 (123-622) days ( Table 2). The median Ki67 area proportion in the KP group was higher than that in the donor group (Fig. 3a). The median Ki67 area proportion in the primary group was significantly lower than that in the KP group (0⋅12 versus 0⋅99 per cent respectively; P < 0⋅001). The Ki67 area proportion in the LT group was significantly lower than that in the KP group (0⋅32 versus 0⋅99 per cent respectively; P < 0⋅001) (Fig. 3a).
To ascertain the time course of Ki67 expression in detail, Ki67 expression in livers of patients in the non-NLS group was compared between the times of KP and LT. Almost all of the Ki67 area proportions decreased with time, and median Ki67 expression at LT was significantly lower than that at KP (0⋅32 versus 1⋅29 per cent respectively; P < 0⋅001, Wilcoxon test) (Fig. 3b). Fig. 3c,d shows representative images that demonstrate the time course of Ki67 expression at KP and LT in the same patient. To investigate the association between Ki67 expression and liver function, a dot plot was generated to demonstrate Ki67 expression and the fibrosis score for all patients with biliary atresia. Patients in the NLS group appeared mainly in the left lower part of the plot, those in the non-NLS group in the upper part, and the primary patients in the right lower part (Fig. 4a).

Discussion
This study found that Ki67 expression in liver parenchyma of patients with biliary atresia might be used as a biomarker for predicting native liver survival after KP, and that high Ki67 expression and Ki67+/LCA+ cell numbers at the time of KP were associated with poor clinical prognosis. In addition, Ki67 expression in livers with biliary atresia decreased as liver function deteriorated.
Immunohistochemical detection of the Ki67 antigen has been used for many years to assess cancer proliferation 16 . However, few reports have focused on Ki67 as a biomarker in non-tumour diseases such as biliary atresia. Funaki and colleagues 24 demonstrated that the Ki67 labelling index in the biliary ducts of livers with biliary atresia was higher than that in normal control livers, but the differences were not statistically significant. Kinugasa and co-workers 21 found that the MIB-1 index in biliary cells did not differ between patients with biliary atresia and poor bile drainage and those with good bile drainage. These results suggested that assessment of Ki67 expression in bile ducts could not be used for prognosis in biliary atresia. However, Hossain et al. 7 reported that the expression in hepatocytes of proliferating cell nuclear antigen, a marker of cellular proliferation, was closely related to the prognosis of patients with biliary atresia. Therefore, measuring Ki67 alone in liver parenchyma may help obtain an accurate prognosis for patients with biliary atresia. This result is reasonable, because an increase in Ki67 cell numbers occurs as a result of active cellular reactions that may be the response to liver injury 17,20 . Patients with a high Ki67 area proportion may have more severe liver damage than those with a low Ki67 area proportion. IHC assessment of the Ki67 labelling index is the method used most widely to determine cell proliferation; however, standardization of Ki67 staining and cell counting is problematic, and interlaboratory reproducibility is variable 25 .
Few reports have been published on the preoperative identification of patients with biliary atresia and a poor prognosis following KP. Azarow and colleagues 10 found that the presence of syncytial giant cells, lobular inflammation, focal necrosis, bridging necrosis and cholangitis in preoperative liver biopsy samples was associated with failure of KP, consistent with the present authors' findings that a high number of Ki67+/LCA+ cells at KP was associated with poor prognosis. LCA is a marker of inflammatory cells, including Kupffer cells. Therefore, a high accumulation of inflammatory cells in livers with biliary atresia may be an important factor for assessing prognosis after KP. Liver fibrosis has also been associated with the prognosis of patients with biliary atresia after KP 12,15 . High perisinusoidal deposition of type I collagen at KP was associated with poor progression after KP 26 . A high level of αSMA expression was also significantly associated with poor KP outcomes [27][28][29] . Although the number of Ki67+/αSMA+ cells in the non-NLS group was higher than that in the NLS group in the present study, this did not appear to have any association with native liver survival. CD163 is a specific marker of activated macrophages associated with liver fibrosis 30,31 ; however, little evidence was found in the present study for an association with Ki67 (Fig. S2,  supporting information). This may indicate that the activity of inflammatory cells may be more important for the prognosis of patients with biliary atresia than that of liver stellate cells. One report 32 demonstrated that pharmacological inhibition and antibody neutralization of serum matrix metalloproteinase (MMP) 7 suppressed the experimental biliary atresia phenotype in neonatal mice infected by rotavirus. MMP acts on proinflammatory cytokines, chemokines and other proteins to regulate inflammation 33 . Thus, inhibition of inflammation may contribute to improving the prognosis of patients with biliary atresia before or after KP as a future therapeutic possibility.
The Ki67 expression changes between KP and LT were an intriguing finding. Ki67 expression increased at KP and decreased at LT to the same level as that in donors, and Ki67 expression decreased from KP to that at LT among the same patients in the non-NLS group. Ki67 expression in the primary patients was significantly lower than that in patients with biliary atresia in the KP group. These results suggest that the change in Ki67 expression was due to liver function changes and/or liver damage in patients with biliary atresia.
The combination of Ki67 expression and fibrosis score may be helpful to separate patients with biliary atresia into four types (groups A-D) that predict different prognoses (Fig. 4b), and initial treatment (KP or primary LT) may be selected according to this grouping. At present, no consensus exists on primary LT indications. Thus, if liver biopsy is used before intraoperative cholangiography to determine Ki67 expression and the fibrosis score, the values of these two histological parameters may be used as an indicator for primary LT (for example, extremely high Ki67 expression or extremely low Ki67 expression with a high fibrosis score).
A limitation of this study is the small number of patients included. In addition, the age of control patients was different to that of patients with biliary atresia, and there was no age-matching.
As KP procedures have improved, the clinical outcome of patients with biliary atresia has also improved; however, the jaundice clearance rate is only around 60 per cent 1,34 . Therefore, many patients already have irreversible liver changes at the time of KP. High-quality biomarkers that can assess success before surgery are needed. This study compared liver histological findings at the time of surgery, and the findings need to be verified using preoperative biopsy samples. The use of biopsy samples to assess outcomes before surgery would enable the selection of patients for primary LT, although for patients with low expression of Ki67 it would be difficult to decide on the indication for KP or primary LT. This approach should be validated in future studies with more patients.
Cell Pathology, Graduate School of Medical Sciences, Kumamoto University, for their assistance with IHC techniques and useful advice. Enago (www.enago.jp) is thanked for the English-language review. This work was supported by JSPS KAKENHI (number 16 K11349) and Japan Agency for Medical Research and Development  (18ek0109258h0002 and 19ek0109258h0003). Disclosure: The authors declare no conflict of interest.