Potential conflict of interest: Nothing to report.
See Editorial on Page 1888
Evidence suggests that most neonatal hemochromatosis (NH) is the phenotypic expression of gestational alloimmune fetal liver injury. Gestational alloimmune diseases are induced by the placental passage of specific reactive immunoglobulin G and often involve the activation of fetal complement by the classical pathway leading to the formation of membrane attack complex (MAC) as the effector of cell injury. We examined liver specimens from cases of NH, from cases of non-NH liver disease, and from infants without liver disease to determine if they would provide evidence that MAC is involved in hepatocyte injury. Sections were immunostained with anti-human C5b-9 complex, the terminal complement cascade (TCC) neoantigen formed in the assembly of MAC. Fetal liver injury in NH cases is associated with a severe loss of hepatocytes. In all NH cases examined, most remaining hepatocytes showed intense staining for TCC neoantigen, whereas hepatocytes in non-NH liver disease cases showed variable light staining. The percentage of hepatocytes containing TCC neoantigen in NH was much greater than that in non-NH liver disease, and there was no overlap between the groups. Findings in both groups suggest that hepatocytes have mechanisms to protect against MAC, including a biliary pathway for its excretion. Conclusion: The finding that all cases of proven NH contained TCC neoantigen far in excess of cases of other neonatal liver diseases suggests that a single process, namely congenital alloimmune hepatitis, is the principal cause of NH. MAC-mediated alloimmune injury in congenital alloimmune hepatitis is a novel mechanism of liver injury that results from an interplay of maternal adaptive immunity and fetal innate immunity. HEPATOLOGY 2010
The etiology and pathophysiology of neonatal hemochromatosis (NH) remain uncertain. The disease has been defined by the concomitant findings of pathological siderosis of various extrahepatic tissues in newborns with liver disease and has been thought to possibly result from a genetically determined error of iron metabolism.1-3 However, more recently it has become evident that iron overload is not the primary disease mechanism involved; instead, siderosis seems to be the phenotypic result of fetal liver injury.4, 5 We have proposed that gestational alloimmunity causes the fetal liver injury in most or all cases in a process that we call congenital alloimmune hepatitis.6
Congenital (gestational) alloimmune disease requires interplay between the maternal adaptive immune system and the fetus, which is an allograft with respect to the mother.7 The mother becomes sensitized to an alloantigen expressed by her fetus but not herself and responds by forming specific reactive antibodies. As only immunoglobulin G (IgG) among the immunoglobulin classes can be transported across the placenta to gain access to the fetal circulation, fetal injury due to maternal alloimmunity is mediated through IgG.8 No cells or other effectors of the maternal adaptive immune system are involved because they have no access to the fetus. The major pathology of NH is restricted to the liver, and most cases show extensive parenchymal injury. Putting the theoretical alloimmune mechanism together with the observed liver pathology suggests that maternal IgG targets an antigen on fetal hepatocytes and that the interaction of specific alloimmune IgG with the fetal liver results in hepatocyte injury and death. Because the simple binding of IgG to a cell should not cause its injury, some effector components of the fetal immune system must be involved.9
The innate immune system of the fetus acts in concert with maternally derived IgG to ward off infections.10 Complement is synthesized by the fetal liver, and this begins relatively early in gestation. IgG subclasses 1 and 3 have the characteristics of being able to engage the placental immunoglobulin receptor, which acts as an IgG chaperone during transport, and to activate complement.8 The binding of IgG1 and/or IgG3 in sufficient density to a microbe or fetal cell can be expected to activate the classical complement pathway.9 Once activated on the surface of the microbe or cell, the complement cascade will culminate in the production of membrane attack complex (MAC).11 MAC is formed in the terminal complement cascade (TCC) by the sequential interaction of C5b with C6, C7, and C8. The complex thus formed attaches to the lipid membrane of the cell or microbe and attracts C9, which polymerizes with the C5b-8 complex to form a MAC unit that penetrates the membrane. In the case of NH, we propose that the binding of a specific maternal alloantibody to fetal hepatocytes activates fetal complement to produce MAC and liver injury.
The hepatic pathology of NH is complex, yet the consistent findings of hepatocyte necrosis, loss of hepatocyte mass, or both suggest that the hepatocyte is the primary target of injury. The formation of MAC on fetal hepatocytes could cause such injury. In this study, we show that liver specimens from patients with NH uniquely express large quantities of MAC in most hepatocytes and that liver specimens from children of similar ages with and without liver disease express much less MAC in fewer hepatocytes. These findings suggest that MAC-mediated hepatocyte injury is a unique and defining feature of congenital alloimmune hepatitis that leads to most cases of NH.
Tissue samples from cases of NH were obtained from postmortem examinations or at the time of liver transplantation. Cases were proven to have NH by the standard criteria of demonstrated liver disease and extrahepatic siderosis. To identify cases of non-NH neonatal liver disease, the surgical pathology records of Children's Memorial Hospital for the years 2000-2009 were subjected to a computer search for cases for which the diagnosis was known and biopsy had been performed at less than 6 months of age. The tissue archives were searched to determine which of these cases had sufficient remaining tissue to permit examination. Autopsy records of Children's Memorial Hospital and Prentice Women's Hospital were searched to identify newborns for which the autopsy report showed no evidence of liver disease, and this resulted in the selection of normal newborn specimens with a range of gestational ages similar to those of the NH cases. The collection and study of these samples were approved by the institutional review board of Children's Memorial Hospital by exemption and with an approved Health Insurance Portability and Accountability Act waiver.
Histology and Immunohistochemistry.
Paraffin-embedded tissue specimens were used, and from these, 5-μm sections were obtained for histological examination (hematoxylin and eosin and trichrome stains) and immunohistochemistry. The immunohistochemistry techniques were previously reported.12, 13 Briefly, after paraffin was removed with xylene, sections were hydrated by passage through a graded series of alcohol solutions. Endogenous peroxidase activity was quenched by the incubation of the sections in 0.3% hydrogen peroxide in methanol for 5 minutes. After being washed, the sections were transferred into a prewarmed antigen retrieval solution (Dako Corp., Carpinteria, CA) and maintained at 95°C for 30 minutes. The sections were then incubated with monoclonal antibody to human SC5b-9 neoantigen (TCC neoantigen; Quidel Corp., San Diego, CA) overnight at 4°C, washed, and treated with the appropriate biotinylated antibody (Vector Laboratories, Burlingame, CA); this was followed by development with Vectastain ABC reagent (Vector Laboratories) according to the manufacturer's instructions. The sections were counterstained with hematoxylin. Because cholestasis and/or siderosis of the specimens might have given false-positive results via endogenous peroxidation, we performed control stains without the primary antibody.
TCC Neoantigen Expression in the Liver.
Liver sections stained for TCC neoantigen were photographed to determine the fraction of hepatocytes displaying MAC. Three random, nonoverlapping images (×400 magnification) were obtained of the hepatic parenchyma from each liver section stained for TCC neoantigen. The photographs excluded regenerative nodules if they were present in the specimen. Digital images were overlaid with a grid of 13 lines × 17 lines (this provided 221 intersections per image), and the position of each intersection was visually determined, whether it was within the confines of a hepatocyte or not. The tally of points in three images (663 points counted in all) landing within a hepatocyte containing clear-cut TCC neoantigen was divided by the tally of points falling within any hepatocyte to compute the percentage of TCC neoantigen-positive hepatocytes.
The findings in the three groups—NH cases, non-NH neonatal liver disease cases, and asphyxiated newborns—were compared for significance with the Student t test for unpaired samples. Group data are presented as means and standard deviations.
Thirty-three cases of NH were studied. Autopsy materials were used in the study of 27 cases, and hepatic explants were used in 6 cases. No sibling cases were included. Four cases were stillborn at 20, 23, 29, and 38 weeks of gestation. The gestational age at birth of the live-born cases was 36 ± 3 weeks (range = 29-39 weeks), and tissue was obtained at 26 ± 30 days of age (range = 0-120 days). With the inclusion of the four stillborn cases, the tissue was obtained at 38 ± 7 weeks (range = 20-48 weeks) after conception. All cases met the conventional criteria for having NH: extrahepatic siderosis in association with severe liver disease. None of the children in this study survived. Extrahepatic siderosis was demonstrated at autopsy in 29 cases (2 after liver transplantation) and by magnetic resonance imaging and/or oral mucosal biopsy in 4 cases that did not undergo autopsy.
Thirty-seven cases of non-NH neonatal liver disease were studied. They included the following: 13 cases of biliary atresia (acquired type; proven by operative cholangiography and the surgical pathology of the biliary remnant); 6 cases of progressive familial intrahepatic cholestasis (3 cases of familial intrahepatic cholestasis type 1 and 3 cases of bile salt export pump disease; proven by mutation analysis); 3 cases of bile acid synthetic defect (BASD; all with Δ4-oxosteroid reductase deficiency; proven by mass spectrometry of urinary bile acids); 3 cases of Alagille syndrome (clinical diagnosis by standard criteria); 3 cases of total parenteral nutrition-associated cholestasis after bowel resection; 2 cases of alpha-1-antitrypsin deficiency (PiZZ genotype); 2 cases of tyrosinemia type 1 (urine succinyl acetone-positive); and 1 case each of abetalipoproteinemia (plasma apolipoprotein B analysis), glycogen storage disease type 1 (liver glucose-6-phosphtase activity), inspissated bile syndrome after cardiac surgery, tricho-hepato-enteric syndrome (hair morphology), and herpes simplex 1 acute liver failure (tissue polymerase chain reaction). Liver biopsy was the source of the tissue in 31 cases, none of which met the standard criteria for acute liver failure.14 Six cases met the criteria for acute liver failure: hepatic explant specimens were studied in three cases (one case of BASD and two cases of tyrosinemia type 1), and autopsy was performed in three cases (two cases of BASD and one case of herpes simplex 1). The gestational ages of these subjects were undiscoverable. Tissue was obtained at 2.0 ± 1.2 months of age (range = 0.5-5.0 months).
Eleven newborns without liver disease were studied. All materials were obtained from autopsies of live-born babies, the cause of death in all being perinatal asphyxia. The postconception ages ranged from 24 to 39 weeks.
Hepatic Histopathology in NH.
As shown in Fig. 1, most of the cases examined in this study showed histopathology that has been well described and is most consistent with subacute or chronic injury.1, 15 In these cases, there appeared to be a substantial loss of hepatocyte mass, as has been reported.12, 16 The morphology of the remaining hepatocytes was profoundly abnormal in most cases (see Figs. 1 and 2 for examples). Giant cells and pseudoacini were common in many cases. The nonhepatocyte parenchyma in these cases consisted mainly of fibrosis and collapsed reticulum. Regenerative nodules were observed in 12 cases. A few samples showed evidence of panlobular hepatocyte necrosis with little fibrosis. This pathological appearance has been reported rarely17 and may represent a more acute onset of liver injury than is typical of NH. The overall hepatocyte morphology in these cases was less disturbed than that in the other cases (see Fig. 3 for examples).
Demonstration of TCC Neoantigen by Immunohistochemistry.
The finding of immunoreactive TCC neoantigen on or in a cell is considered to be irrefutable evidence that MAC has been assembled on its plasma membrane.18 Because the presence of bile and/or iron might cause a false-positive immunohistochemistry result, we performed controls with only the secondary horseradish peroxidase-labeled antibody in several NH and non-NH liver disease cases, none of which showed positive staining in the absence of the specific antibody (Fig. 4).
TCC Neoantigen Expression in the Liver.
The intent of this analysis was to determine if MAC was involved in the injury of surviving hepatocytes, which presumably were still being injured at the time of the sampling. Figure 2 shows examples of TCC neoantigen expression in the livers of typical NH cases with subacute or chronic injury, Fig. 3 shows its expression in NH cases with histology suggesting acute injury, and Fig. 5 shows its expression in examples from the two comparison groups. The percentage of hepatocytes displaying TCC neoantigen was determined in each sample as a measure of the intensity of complement-mediated cell injury. The results shown in Fig. 6 demonstrate that hepatocytes in NH cases more frequently expressed TCC neoantigen than hepatocytes in non-NH liver disease cases. In NH cases, 93.0% ± 6.5% (range = 76%-100%) of hepatocytes expressed TCC neoantigen, whereas in non-NH liver disease cases, 10.8% ± 12.5% (range = 0%-45%) did (P < 0.001). In addition, substantial qualitative differences were observed between NH cases and non-NH liver disease cases. As demonstrated in Figs. 2 and 3, hepatocytes in many NH cases showed diffuse staining throughout the hepatocyte cytoplasm, whereas in others, they showed highly consolidated staining in what appeared to be vesicles, perhaps autophagosomes or lysosomes. Hepatocytes in non-NH liver disease cases, in contrast, showed at most faint sparse speckling with TCC neoantigen (Fig. 5). Several NH and non-NH liver disease cases with hepatocellular cholestasis, such as cases of BASD, showed intense TCC neoantigen staining of bile in canaliculi and duct structures (Fig. 4). This novel finding suggests that bile may be an excretory pathway for MAC excised from the plasma membrane. Newborn control cases expressed virtually no TCC neoantigen in hepatocytes: 0.9% ± 1.3% (range = 0%-3%) of hepatocytes stained positively (P < 0.001 versus NH and non-NH cases).
NH is evidently the phenotypic result of fetal liver injury,19, 20 and a prevailing theory regarding pathogenesis is that NH results from severe fetal liver injury due to any of many potential causes.1 However, important questions have remained regarding what disease process could produce fetal liver injury severe enough to cause NH and whether there is one predominant process involved.21 The current work involves the study of a large collection of liver specimens from cases of NH proven by standard anatomic criteria to determine if congenital alloimmune hepatitis is the predominant cause. We reasoned that the most likely mechanism of immune injury that could result in the extensive loss of hepatocytes seen in NH would involve maternal IgG-mediated activation of fetal complement by the classical pathway and that MAC formation would be proximate to hepatocyte injury. Thus, we expected to find MAC associated with residual injured hepatocytes in cases of congenital alloimmune hepatitis. Our finding of extensive MAC expression in most remaining hepatocytes in all NH liver specimens examined suggests that congenital alloimmune hepatitis is the likely cause of most cases of NH. These results also demonstrate a novel mechanism of liver injury in congenital alloimmune hepatitis. To our knowledge, no other liver disease has been shown to result directly from humoral immune injury of hepatocytes. Fetal tissue injury in gestational alloimmune disease is driven by the maternal immune system via passage of IgG to the fetus, and the fetal immune system is engaged to produce tissue injury.7 In congenital alloimmune hepatitis, gestational alloimmune mechanisms directly injure fetal hepatocytes and produce severe fetal liver disease leading to phenotypic NH.
NH has been defined in its earliest descriptions by siderosis of extrahepatic tissues and has been widely considered to result from an inborn error of metabolism.22 It remains classified as part of the hereditary hemochromatosis group of disorders (OMIM 231100). The liver is virtually always iron-laden in NH,16 but that finding is not diagnostic of the condition because several other liver diseases of neonates, including tyrosinemia type 1 and inborn errors of bile acid synthesis, also show hepatic siderosis. The mechanism of hepatic iron deposition in these various forms of neonatal liver disease, including NH, is unknown, but it is certainly the result of liver injury and not the cause. The mechanism of iron overload of extrahepatic tissues in NH is unknown. In contrast to the liver, the extrahepatic tissues showing siderosis do not appear injured, and this suggests an indirect mechanism of iron overload. Severe fetal liver injury could result in a failure to regulate the flux of iron across the placenta. Hepcidin is the hepatic signal molecule that regulates placental iron flux by its interaction with ferroportin.5, 23 As it does in postnatal life, the fetal liver senses iron sufficiency and increases hepcidin expression to limit the influx of iron.24 A severely injured liver might fail to exert this negative feedback on placental iron flux, and this could result in iron overload.5 NH is also associated with decreased serum transferrin levels (probably the result of hepatic synthetic failure) and increased iron saturation.25 Increased non-transferrin-bound iron results, and as in hereditary hemochromatosis, this iron is deposited as hemosiderin in a specific tissue distribution. Why this occurs uniquely in congenital alloimmune hepatitis among neonatal liver diseases is probably related to the severity of the fetal liver injury, which may be unmatched by any other disease, and/or the early onset of liver injury, which does not occur in infectious or metabolic diseases.21 It is probable that some cases of congenital alloimmune hepatitis will produce fetal liver injury of lesser severity or later onset and as a result will not be associated with extrahepatic siderosis. Finding such cases would support the notion that extrahepatic siderosis is the result of fetal liver injury and not a manifestation of a primary defect in iron metabolism.
The finding of hepatocyte MAC accumulation in cases of non-NH neonatal liver disease was somewhat unexpected and raises the question of whether complement activation contributes to hepatocyte injury in these diseases. The neonatal diseases examined, with the exception of biliary atresia, were not thought to have an immune mechanism. Several of the non-NH liver disease cases showed an accumulation of TCC neoantigen both in their cytoplasm and in bile within canaliculi. This presents the intriguing possibility that hepatocytes may possess the capacity to defuse MAC as some other cell types do,26-28 and they may use transport into bile as a means of disposal. The mechanics of this operation deserve to be explored. The non-NH cases with the highest percentage of hepatocytes with MAC expression were surprisingly mainly genetic/metabolic diseases: the nine cases in the upper quartile of values included three cases of progressive familial intrahepatic cholestasis, one case of Alagille syndrome, one case of BASD, one case of tyrosinemia type 1, and tricho-hepato-enteric syndrome. Two cases of biliary atresia also fell into the upper quartile; this is not altogether unexpected because an immune mechanism is thought to be involved.29 Possible mechanisms in the others include failure to clear MAC by hepatocytes as a result of cholestasis. However, this seems unlikely to be the entire reason given that only 2 of 13 biliary atresia cases fell into the upper quartile and only 7 fell above the median. Two cases (BASD and tyrosinemia type 1) that had clinical liver failure fell into the top quartile for MAC expression. It is possible that retention of MAC is a function of global hepatocyte failure in such cases. No matter what the mechanism is for the acquisition of MAC in non-NH liver disease, quantitative immunohistochemistry provided complete separation of NH from other severe liver diseases in this age group.
Our findings have important ramifications. Assuming no selection bias led to the inclusion of only alloimmune NH cases in the study population, one can conclude with a high degree of certainty that cases of NH proven by conventional grounds have an alloimmune etiology. This may reduce the need to look for other potential causes of liver disease in children with the NH phenotype and eliminate unneeded testing, such as searching for infectious causes and testing for genetic/metabolic diseases including hereditary hemochromatosis. This in turn would permit an earlier focus on potentially helpful treatment of affected neonates.30 Furthermore, the finding that gestational alloimmunity is involved provides an indication for treating all women who have had a baby affected with NH during subsequent gestations to prevent recurrence.31, 32 In addition, the involvement of maternal IgG directed against a target that we propose is a unique antigen expressed by fetal hepatocytes leads to the possibility of a serological test for diagnosing NH in the future. Finally, the absence of overlap of this finding in NH cases with other neonatal onset liver diseases suggests that this could provide a useful approach to the diagnosis of NH in archival liver tissue and might permit diagnosis in cases in which extrahepatic siderosis could not be demonstrated.
The findings of this study provide evidence that congenital alloimmune hepatitis is the etiology for most or all cases of NH. Our attempts to identify the specific fetal liver target antigen of alloimmunity have not yet been successful. As a result, we cannot say if congenital alloimmune hepatitis represents a single disease, that is, one involving a single alloantigen. However, the finding that every one of this large set of unselected cases of proven NH had extensive hepatocyte-associated MAC suggests that a single mechanism is involved in most or all cases of NH. Why in some cases the immune injury appears to be more acute whereas most cases are more subacute or chronic remains to be determined. It is possible that different antigens expressed at different times in fetal liver development are targeted in the two forms of the disease. The current findings provide an impetus to remove NH from the list of primary hemochromatosis disorders. This is important because as long as it remains on that list it will be considered to be a genetic disease, and genetic counselors will give inappropriate advice with respect to risk and detection. NH appears to be a secondary form of hemochromatosis due to fetal liver injury that in most or all cases is caused by congenital alloimmune hepatitis.