Histopathology of livers in patients with congenital portosystemic shunts (Abernethy malformation): a case series of 22 patients

Congenital portosystemic shunt (CPSS) are congenital anomalies resulting in partial or complete diversion of the portal blood into the systemic circulation. First described by John Abernethy [1], they are thought to be related to defective portal vein embryogenesis, as discussed by Wanless et al. [2].

CPSS are increasingly being diagnosed in the neonatal period when an ultrasound is being performed for neonatal jaundice. Another large proportion is picked up incidentally when having an ultrasound for unrelated reasons. The remainder are usually identified due to the associated complications of neurocognitive dysfunction, developmental delay, hepatocellular nodules and/or hepatopulmonary syndrome. If gone unnoticed until adult life, presentation may be due to symptoms related to a hepatic nodule, signs of low-grade encephalopathy, pulmonary hypertension, hypoglycaemia, chronic abdominal pain and/or elevated ammonia levels.

The literature on the histological changes in livers of patients with CPSS is limited. Wanless et al. [2] were the first to describe the histological findings and in particular portal vein hypoplasia and large regenerative nodules in the liver of a patient with persistent ductus venosus. In addition to other individual case reports [3‐7], the case series by Lisovsky and colleagues [8] on five CPSS patients, to our knowledge, is the most comprehensive and detailed histological description to date. We describe, therefore, our findings in a larger series of 22 patients with CPSS who were managed clinically in our institution over a period of 15 years, and from whom background liver tissue was available for histological review. The emphasis in this paper is on the hepatic histological changes related to the effect of the underlying anomaly rather than the more general clinical aspects or details of the liver tumours identified in some of these patients.

Our case series consists of 22 of 46 consecutive patients diagnosed and managed in our institution for CPSS between 2001 and 2016 for which liver specimens were available (core needle biopsy, wedge biopsy or liver explant).

The type of CPSS was subclassified into four types:

We defined PDV according to the criteria by Blanc et al. [12]: ‘intrahepatic shunt running from the proximal part of the left portal branch to the terminal part of the left hepatic vein and located in the depth of the Arantius sulcus between the left and caudate lobes of the liver’. We considered as intrahepatic shunts any intrahepatic shunts that did not fulfil Blanc’s criteria for PDV.

The histology review was carried out by AQ and CDV and the following features were assessed: Fibrosis was assessed in a descriptive semiquantitative fashion noting its location (portal or centrilobular/perisinusoidal based), and whether it was associated with septa and parenchymal nodular transformation. This assessment was carried out with the aid of the Gordon and Sweet reticulin and Sirius Red stains carried out routinely in our laboratory. Particular attention was paid to changes to portal vascular structures. The term ‘hypoplasia’ was used for portal vein branches showing a thickened wall and/or abnormally small calibre relative to the size of the nearby bile ducts and/or arteries, in line with the terminology used by Wanless et al. [2] and Lisovsky et al. [8]. Portal tracts containing one or more arterial profiles and at least one interlobular or septal bile duct and without an obvious portal vein branch, or with one or more slit-like spaces which needed immunohistochemistry to be characterised further were recorded as ‘dyads’. Portal tracts containing a bile duct, and an artery of matching size and a larger normal-sized portal vein branch with a thin wall was recorded as a ‘triad’ [14]. Individual portal structures were recorded as monads and the type specified. Unless individual arteries had a significant cuff of stroma around them they would be recorded separately as interlobular individual arteries rather than portal arterial monads. Portal tracts containing prominent, sometimes dilated, thin-walled channels (PTWCs), which needed immunohistochemistry (podoplanin, CD34) to be characterised further, were not recorded as dyads but counted separately. Dilated thin-walled channels protruding through the limiting plate were recorded as dilated inlet venules. Portal arterial branches were considered abnormal if appeared increased in number against the reference standard described by Crawford et al. [14]. Small thick-walled capillaries were considered to be minute arterial branches. In line with the findings by Lisovsky et al. [8], the presence of abnormal arteries, much larger than the nearby accompanying bile duct was also recorded. The presence of isolated intralobular arterioles and increased sinusoidal stain for CD34 was also recorded when applicable. Additional features assessed were changes to bile ducts (bile duct loss, bile duct attenuation and peribiliary lymphocytosis) and hepatocytes (normal hepatic plates, hepatic plate disarray, nodular regenerative hyperplasia, aberrant stain for CK7, periportal accumulation of copper-binding protein, periportal hepatocyte nuclear vacuolation and steatosis).

We used a chi-squared test to investigate whether there was an association between the histological changes and the type of shunting, dividing our patients in two groups, one with type 1 shunting (complete absence of portal flow) and the other group with type 2, intrahepatic shunts of persistent ductus venosus, and whether there was an association between the histological changes and the presence or absence of liver tumours.

Our cohort consisted of 22 patients (ten females). Most were children at presentation (age range = 2 months–18 years; median 5.5 years). The only patient presenting in adulthood was a 49-year-old gentleman who was investigated for symptoms related to a large hepatocellular carcinoma. Radiological signs of an underlying CPSS were discovered during his work-up.

In terms of shunt type, six patients had a type 1, 13 had a type 2, two had an intrahepatic type and one had a persistent ductus venosus. Of the two intrahepatic shunts, one (patient 14) run from the left portal vein to the middle hepatic vein, whilst the other (patient 16) run from the left portal vein through the parenchyma of the liver to a peripheral subcapsular vessel away (to the left) from the Arantius sulcus, and subsequently fed into the left hepatic vein. The shunt type was characterised by cross-sectional imaging and portal venogram when present and then correlated with surgical findings.

Thirty-two samples in total were available for review. In three patients, the samples examined were obtained from whole livers removed at transplantation. In the other 19 patients, the samples reviewed were 16 wedge biopsies obtained at laparotomy for shunt closure and ten percutaneous core needle biopsies. Of the ten percutaneous liver biopsies, four were obtained from four children at their first clinical presentation before CPSS was discovered. One patient was a 2-month-old baby girl who presented with conjugated hyperbilirubinaemia, and the other three were older children with abnormal liver biochemistry or hepatosplenomegaly. The other six core needle biopsies were taken during surgery or to complement the clinical work-up of known CPSS. The details are shown in Table S1.

The histological findings in the three explanted livers are summarised in Table 1. All three explant specimens had multiple nodular lesions with a histological appearance ranging from focal nodular hyperplasia (FNH) to hepatocellular adenoma (HCA), well-differentiated hepatocellular neoplasm (WDHN) [15‐17] or hepatocellular carcinoma (HCC) in two, and a single large HCC with marked atrophy of the remaining parenchyma in one.

The samples of background liver were limited as they were perilesional in two specimens (patients 1 and 2), and we cannot exclude that at least some of the changes noted were the result of tumour compression. Particularly in patient 2, there was marked parenchymal atrophy, severe fibrosis and parenchymal micronodular transformation, making this case not comparable to the other ones. Of note, the portal vein was present in the sections from the porta hepatis in all three specimens, but there were no representative systematic samples from the large portal tracts to comment further on the status of the large intrahepatic portal vein branches.

The amount of fibrosis in patient 1 was variable ranging from severe bridging fibrosis in the vicinity of the HCC to minimal portal fibrosis in the vicinity of the lesions with features of focal nodular hyperplasia. The small portal tracts in the less fibrotic areas showed some hypoplastic portal vein branches (Fig. 1) and some PTWCs in places. Some hepatic artery branches were hyperplastic or enlarged. Bile ducts were intact and cholangitis was present in one patient. There was some sinusoidal dilatation and isolated hepatic arteries were identified in the lobules.

The amount of fibrosis in patient 3 was minimal and portal-based in two samples from the right and left lobe, which had been taken as far away as possible from any of the nodular lesions. There was a patchy parenchymal atrophy with irregular spacing of the portal tracts including crowding and convergence [8] in places. Small intrahepatic portal vein branches were absent or hypoplastic in most portal tracts. Dilated inlet venules were not identified. PTWCs and arterial branches increased in size or in number were present in places. There was a hepatic plate disarray, and areas of nodular regenerative hyperplasia were also noted. Sinusoids were dilated in some of the samples, and their endothelium stained focally for CD34 often as an enhancement of the normal periportal stain as observed in the biopsy specimens. Individual lobular arteries were also observed in places.

Cytokeratin 7 positive hepatocytes and periportal deposits of copper-binding protein were present in two patients. Of note, periportal hepatocytes with vacuolated nuclei in the periportal regions were not identified in any of these three patients.

Twenty-nine biopsy samples from 19 patients were available for review. Sixteen samples were from wedge biopsies and the other 13 from core needle biopsies. Three core needle biopsy specimens from patients 13, 17 and 22, in addition to those described in Table S1 were excluded from histological review because they were too small.

Ten patients underwent a single biopsy of background liver (five core needle biopsies and five wedge biopsies). Nine patients underwent multiple background liver biopsies. Three patients (4, 7 and 9) underwent wedge liver biopsies from both right and left lobes at the time of the shunt closure. The other six patients underwent two or more background liver biopsies at different time points. The details are described in Table 2.

The ten core needle biopsies, which were reviewed further, ranged in size from 7 mm to 35 mm (average 14.1 mm) in length and included a range of 2 to 11 complete portal tracts (median = 3.5). In contrast, the wedge biopsy specimens included a range of 5 to 71 portal tracts (median = 17). An overall summary of the histological findings in all biopsies is provided in Table S1.

Our findings are summarised in Table S1 (individual patients) and Table 2 (comparison between wedge and core needle biopsies).

We did not find any significant statistical difference (not shown) in histological changes when comparing the wedge biopsies of two patients with type 1 shunt with those of ten patients with type 2 shunt, or the wedge biopsies of two patients with tumours with those of ten patients without. We did not include the core needle biopsies and the explanted livers because they were not comparable to the wedge biopsies, due their much smaller size (core needle biopsies), and relatively limited amount of non-lesional tissue (explanted livers), as mentioned earlier.

The literature on the spectrum of histological changes in livers with patients with CPSS is limited to case reports [2‐7] and the single case series by Lisovsky and colleagues [8]. Our larger cohort has allowed us to confirm broadly on previous observations and add some new findings. Compared to biopsies, surgical specimens should have the advantage of providing more tissue for histological examination. In our cases, however, all resection specimens contained tumours, and the amount of non-neoplastic tissue available for this retrospective review was limited and mainly perilesional. We cannot therefore exclude that the changes present in these specimens were due, at least in part, to the effect of tumour compression.

In contrast to Lisovski et al. [8], we could identify the portal vein in the porta hepatis of all three liver explants. In patient 1, the shunt consisted of a persistent ductus venosus suggesting that it formed in the context of a developed intrahepatic portal vein tree. Patients 2 and 3 had a type one shunt characterised by an end to side connection between the extrahepatic portal vasculature and the inferior vena cava with complete lack of intrahepatic portal flow. The close and complex developmental relationships between the prerenal section of the inferior vena cava and the plexiform subdiaphragmatic anastomosis and remodelling of the vitelline veins from which the portal vein derive [18], could explain the paradoxical coexistence of a complete shunt and the hilar portal vein.

In most of our wedge biopsy specimens, portal arterio-biliary dyads were observed frequently. They do not necessarily indicate a pathological status, as they are present in up to 38% of portal tracts in liver biopsies from adult normal liver [14]. In many of our cases, a slit-like space was present in the vicinity of a dyad but it was not possible to be certain whether it represented a portal vein remnant. We do not use routinely trichrome stains, which may be helpful in evaluating portal structures and scarred vein remnants [8], and, as mentioned earlier, some of these small structures tend not to be represented in sections at different levels. We believe, however, that a threshold of 38% of portal tracts with dyads as a physiological finding does not apply to paediatric livers with CPSS because of the presence of the other vascular abnormalities in most portal tracts. The presence therefore in a biopsy specimen of a combination of dyads, biliary monads, PTWCs, dilated inlet venules and increased number of portal arteries should suggest the presence of a CPSS. PTWCs in particular were a frequent finding and relatively easy to identify. They appeared to be composed of a combination of portal vein branches and lymphatics, in some instances, and portal vein branches only in others, but this should be explored in more details by using serial sections, double epitope immunohistochemistry and, ideally, 3D reconstruction. The application of cognitive algorithms could also be informative [19, 20]. Such an approach could also be helpful to clarify whether the presence of porto-lymphatic shunting suggested in one of the biopsies does really occur. As suggested by Kanazawa et al. [10], the configuration of the microscopic portal circulation could help in predicting intrahepatic portal flow after shunt occlusion or plan for a two-stage approach. This consists of initial shunt reduction to allow the portal network to adapt to the increase in portal inflow, followed by complete occlusion a few months later [21].

Vacuolation of the nuclei of periportal hepatocytes is routinely observed in livers from young patients [22]. The absence or marked reduction of this change in most of the young patients in our series is intriguing. Commonly observed in pathological conditions such as diabetes mellitus and Wilson disease, Aravinthan et al. [23] suggested that it may be an indication of hepatocyte senescence. Serial liver biopsies from depancreatised baboons [24] showed that periportal hepatocytes develop nuclear vacuolation within 4 days, and that by 8 days almost all hepatocytes in the lobule show vacuolated nuclei. This change is reduced, but not eliminated, by the administration of insulin. Starzl et al. [25] showed in an animal model of partial portocaval transposition that exposure of the liver parenchyma to caval blood or blood of small intestinal origin results in atrophy. In contrast, the splanchnic blood returning from the pancreas, proximal part of the duodenum, stomach and spleen leads to hypertrophy and hyperplasia. In terms of hepatotropic factors, Starzl et al. [25] concluded that insulin is the dominant anabolic hormone, counterbalanced to some extent by glucagon and epinephrine. The lack of periportal hepatocyte nuclear vacuolation in our paediatric patients is therefore a paradox as one would expect the lack or reduced portal circulation to be detrimental by reducing the delivery of insulin to hepatocytes. An alternative explanation resides in the concept that nuclear vacuolation could be an indication of hepatocyte senescence, and that senescence is part of liver remodelling, including during embryonal development [26‐30]. Nuclear vacuolation in children, therefore, could be an indirect sign of liver growth in the context of a preserved portal circulation or, in a broader sense, a balanced artero-portal perfusion. Unbalanced porto-arterial perfusion due to the presence of a CPSS could alter senescence and remodelling and ultimately the growth process. Whether perturbation of the senescence process influences the development of tumours in these patients is a possibility.

The presence of biliary changes and signs of chronic cholestasis in some of our patients is similar to the findings of Lisovski et al. [8] but their cause remains uncertain.

In conclusions, our series of patient with CPSS has shown a constellation of changes in the liver parenchyma, which are probably the combined effect of the underlying developmental abnormality and compensatory effects. In our experience, the most characteristic findings in the peripheral liver parenchyma include the presence of portal prominent thin-walled channels, arterial-biliary dyads, increased arterial profiles in the portal tracts and the lobule and frequent lack of the physiological periportal-vacuolated hepatocytes in children. These changes may not be fully represented in core needle biopsy samples due to sampling variation. Whether and how these changes provide the background to the development of hepatocellular tumours in these patients remain uncertain. Patients with CPSS should be considered to be at risk of developing hepatocellular tumours with the potential to progress to HCC and should be managed accordingly.