ABO-compatible Liver Allograft Antibody-mediated Rejection: an update

Insights from the “Early Years”

Relatively crude cell-based donors-specific antibody (DSA) detection assays began the study of liver allograft antibody-mediated rejection (AMR): 8 – 15% recipients showed positive results (>50% lysis)[1-3]. It was quickly recognized that human liver allografts: a) were less sensitive than kidney allografts to acute adverse consequences of pre-formed DSA[4]; and b) could protect subsequent kidney and heart allografts from the same donor from AMR in most, but not all, sensitized experimental animals[5,6] and humans[7,8]: acute kidney allograft AMR was not seen in all sensitized recipients, but lower patient and graft survival (~40%) did not influence practice.

Most adequately powered studies revealed increased acute AMR and T cell-mediated rejection (TCMR) risk with or without graft failure when livers were transplanted into crossmatch-positive recipients[3,9-12]. AMR with or without TCMR was often treatable[1] and short-term penalties for ignoring crossmatch results were not felt to be clinically significant enough to institute routine HLA typing or DSA testing at most centers. Liver allografts, however, were not exempt from acute AMR: relative hepatic resistance could be overcome, in experimental animals[13,14] and humans[1,10,15], in very highly sensitized recipients.

Cataloging liver HLA class I and II antigen expression in normal liver was accomplished by immunostaining in peripheral liver tissue specimens[16-28]: all reported diffuse and strong class I HLA expression on all cell types, except hepatocytes where expression was weaker. HLA class II expression was largely restricted to portal, perivenular, and subcapsular dendritic cells, and Kupffer cells in normal livers with DQ demonstrating the weakest expression. Portal vein branch endothelia class II expression was consistently negative, but portal capillary, sinusoidal and central vein endothelia varied from negative to focally positive[16-28]. Since few studies specifically addressed class II expression in portal capillary/peribiliary plexus, lymphatic capillaries, inlet venules, and the peribiliary plexus of large extra-hepatic bile ducts; more work is needed on this topic.

All solid organ allografts, especially livers with co-existent pathology, show dynamic HLA expression, especially class II, which is regulated by class II transactivator [29,30]. Upregulation occurs following exposure to environmental/inflammatory stimuli (esp. γ-interferon) resulting in strong DR>DP>DQ expression in all endothelial cell compartments, biliary epithelium, and hepatocytes with co-existent liver pathology, especially TCMR[16-28].

Practical consequences of dynamic HLA expression include: 1) variable liver class II target antigen distribution versus comparatively constant heart and kidney microvascular endothelial expression (Figure 1A-D); 2) spectral downstream immunological effector mechanisms likely depend on endothelial cell antigen-antibody complex density[31-33]; and 3) precipitation and/or accentuation of potential adverse DSA consequences by co-existent pathology[34]. All these consideration likely affect underlying pathophysiological mechanisms and histopathology. Therefore, 4) a potential exists to improve outcomes with non-antibody-directed therapeutic interventions that down-regulate class II expression [35,36].

Extremely sensitized (>1: 540 CDC) recipients can experience rapid allograft failure caused by frank acute AMR: microvascular injury, thrombosis, and hemorrhagic necrosis[37]. A less fulminant acute AMR presentation is more commonly seen as graft dysfunction, thrombocytopenia, DSA persistence after post-transplant (more common with class II), appearance of circulating immune complexes[1,2,38], and histopathological changes described below. Lower-level sensitization usually resulted in rapid DSA disappearance and either no injury or transient antibody-mediated damage often misrepresented as “preservation injury”[1,2].

Understanding Why Liver Allografts Differed from Kidney and Heart Allografts

Early observations prompted studies that identified liver allograft AMR resistance mechanisms, all of which are likely contributory: 1) Kupffer cells could clear antibodies, activated complement, platelet aggregates, and immune complexes (formed by soluble donor class I HLA antigens that bound class I DSA)[5,6,39-41]. Supporting evidence includes: a) increased susceptibility to AMR and decreased protection of sequentially-placed extra-hepatic allografts in recipients of Kupffer cell-depleted liver allografts[5,6,40]; and b) delay or prevention of acute heart allograft AMR in sensitized recipients by gene therapy that delivers soluble donor class I antigens, similar to liver allografts[42]. 2) Variable hepatic [43] versus strong and constitutive kidney[44] and heart[45] microvascular class II expression provide less class II DSA targets. 3) Large liver size facilitates antigen-antibody complex dilution across the vast endothelial cell surface; potentially explaining increased AMR susceptibility in reduced-size allografts[46,47]. 4) Liver sinusoidal endothelial cells express Fc receptors[48] and lack a typical basement membrane; they are also normally lined by macrophages (Kupffer cells). All of these factors potentially influence antibody-endothelial interactions. 6) The liver’s regenerative capacity and ability to heal either without fibrosis or reverse fibrosis[49].

Current DSA Testing Era

Solid phase DSA testing defined the current era, during which prior observations were validated and extended [50]: pre-transplant CDC+-causing antibodies are encountered in ~10-15% of recipients with a female and autoimmune predilection[51-56]. Data linking the two eras show the ~96% of cell-based CDC- recipients also lacked DSA; however, >50% of isolated class I or II DSA+ patients were CDC-[51].

When DSA+ (defined as MFI≥5000) recipients underwent orthotopic liver transplantation (OLTx), the vast majority of lower MFI class I DSA (<10,000 MFI) disappeared without short-term overt liver allograft damage, but C4d deposits were detected in some highly sensitized recipients early after OLTx and long term consequences, if any, are unknown[51,56]. Regardless, preformed DSA did not adversely influence short-term survival in the vast majority of low to moderately sensitized recipients[51-56]. In contrast to class I, 1/3 of patients with high-MFI class II DSA (≥10,000) experienced persistence[51] with an increased risk of early TCMR, and perhaps, mixed TCMR and acute AMR[51]. A tiny fraction (<5%) of highly sensitized (DSA+) recipients have sufficient DSA (usually multiple class I and II usually in high MFI/titers)to cause clinically and histopathologically significant liver injury[50,55,57], which also depends on the baseline immunosuppressive regimen[55,56,58].

Precise DSA characterization helped clarify mechanisms underlying the “partial protection” liver allografts afford sequentially-placed kidney allografts from the same donor: low-level class I DSA rarely causes problems, but class II DSA resulted in kidney, and less-likely liver, allograft acute AMR[59,60]. It is tempting to speculate that less efficient class II DSA clearing is attributable to lower density class II expression and secretion.

De novo DSA develops in ~8 - 15% of liver allograft recipients[61,62], the vast majority directed at HLA class II, preferentially DQ[61,62]. Risk factors for de novo DSA include cyclosporine versus tacrolimus use, low levels of immunosuppression, young age, low Model for End-Stage Liver Disease (MELD) score[61], and previous transplants[62]. Multivariate analyses show that de novo DSA is associated with decreased patient and allograft survival[61] and fibrosis development[62]. IgG3 subclass testing may help facilitate identification of preformed and de novo DSAs associated with the highest risk of allograft damage[63,64]. Late onset acute AMR has been reported in suboptimally immunosuppressed de novo DSA+ individuals[62].

Plasma cell hepatitis (PCH) [a.k.a. de novo autoimmune hepatitis (AIH)], is an uncommon (~3-5% of recipients) cause of late (usually >1 year) graft dysfunction that resembles native liver AIH. More prevalent and severe bile duct damage, IgG4+ plasma cell-bias, and aggressive central perivenulitis in PCH, compared to standard AIH, suggests that this entity represents an overlap between auto- and alloimmunity[65]. TCMR and steroid-dependence are PCH risk factors in pediatric recipients[66]. Plasma cell-rich infiltrates and aggressive perivenular necro-inflammatory activity correlate with serological evidence of autoimmunity[67].

Atypical liver/kidney microsomal autoantibodies directed against the cytosolic enzyme glutathione-S-transferase T1 (GSST1 have been associated with PCH[68-70] in null GSTT1 genotype recipients of GSTT1+ donor livers; others have not seen this association[68-70]. A variety of other autoantibodies detected in the setting of PCH include cytokeratin 8/18 auto-antibodies[71] and atypical LKM antibodies directed at isoforms of carbonic anhydrase III, subunit β1 of proteasome, and members of different glutathione S-transferase (GST) families[72]. Angiotensin II Type-1 receptor DSA impairs renal allograft outcomes[73], and possibly contributes to fibrosis in combination with HLA DSA in liver allografts[74].

Histopathology

Post-reperfusion biopsies from patients with high titer/high MFI despite serial dilutions DSA show platelet aggregates in portal and/or central veins [1,97]. Diffuse portal microvasculature C4d staining[50,56], portal microvascular endothelial hypertrophy and cytoplasmic eosinophilia, occasionally resulting in “hobnailing” appear within days after OLTx in those who develop acute AMR[50,57]. Accompanying features include “micro-vasculitis” involving portal vein branches, inlet venules, portal capillaries, peribiliary plexus capillaries, and occasionally central veins. Endothelial inflammation is often mediated by eosinophils, macrophages, and neutrophils[50,57] and accompanied by a ductular reaction, spotty acidophilic necrosis of hepatocytes, centrilobular hepatocellular swelling and hepatocanalicular cholestasis; focal bile duct necrosis; and arterial changes strongly suggestive of arterial vasospasm[1]. Superimposed TCMR is almost universally present[1].

Inflammatory/necrotizing arteritis is rare, but is diagnostic of acute AMR when seen in conjunction with diffuse C4d deposits and DSA. Some histopathological changes resemble preservation/reperfusion injury and obstructive cholangiopathy, but marked portal microvascular endothelial cell hypertrophy and cytoplasmic eosinophilia and “microvasculitis”, especially when involving central veins, distinguish acute AMR. Regardless, stringent diagnostic criteria are needed to establish an AMR diagnosis with certainty: 1) histopathological changes consistent with AMR; 2) exclusion of other insults causing a similar injury pattern; 3) serum DSA; and 4) strong and diffuse complement (C4d) deposition[50,55,57,95,98,99], defined as strong portal vein and capillary and usually periportal sinusoidal endothelial staining involving a majority of portal tracts.

The “signature” histopathological lesion of acute AMR in kidney and heart allografts, “capillaritis”, is defined by endothelial cell hypertrophy, margination of leukocytes [macrophages, neutrophils, and lymphocytes (e.g. NK cells)], and dilatation with or without capillary disruption, depending on injury severity[77]. Capillaritis is usually accompanied by C4d staining and observed regardless of the time after heart or kidney transplantation. Similar changes are seen in liver acute AMR, but capillaritis can be difficult to recognize more than several weeks post-OLTx.

There are likely several key reasons for these inter-organ differences: 1) Distinguishing among typical portal capillaries, lymphatic capillaries, and inlet venules is extremely difficult[100]; made even more difficult in inflamed livers because small portal vessels are: a) obscured by co-existent inflammation; b) intermixed among normal elastic fibers that non-specifically stain for C4d, especially on frozen tissue (Figure 1E-F); and c) are gradually destroyed during chronic AMR[101-103]. Endothelial stains (e.g. CD34, CD31, D2-40) can help facilitate identification. 2) Lower class II expression might affect downstream effector mechanisms and instead of “capillaritis”, DSA might trigger endothelial cell phenotype changes and/or activation of pathways that promote or retard cellular inflammation, coagulation, apoptosis and complement deposition[31,32,75-78,80-83]. 3) Direct activation of stellate cells is possible given the strong empiric association between DSA and non-inflammatory fibrosis[62,104,105].

Interpreting of Liver Allograft Immune and C4d Staining

Classic immune deposits (e.g. IgG, C3, C4) are ephemeral in acute AMR, even in frozen tissue[1,95,106]. In severe cases, deposits of IgG, and/or IgM, C3, C4, and C4d can be diffusely detected in frozen sections along the sinusoids and in perihilar arteries, portal veins, and peribiliary plexus if tissue samples are obtained early in acute AMR development [1,95,106]. Recognition that C4d can: a) persist for several days; b) be detected in formalin-fixed paraffin-embedded (FFPE) tissues; and c) correlate with circulating DSA improved the diagnostic accuracy of liver allograft AMR[99,107].

If liver AMR is suspected, saving frozen tissue for immunofluorescence (IF) C4d staining may improve diagnostic accuracy. No studies directly compared frozen and immunoperoxidase staining on formalin-fixed, paraffin-embedded C4d staining for liver allografts. The first author’s experience can be summarized as: IF is more sensitive than immunoperoxidase, but more difficult to interpret because of high background/non-specific staining and difficulty in discerning the underlying architecture, especially the portal microvasculature. Sinusoidal C4d labeling is more common in frozen samples. Immunoperoxidase staining technique affects sensitivity: pressure cooker and high pH (pH = 9.0 vs. 6.0) antigen retrieval yield more sensitive results, but background staining can emerge leading to interpretational problems [107-115].

Normal liver allograft biopsies are usually negative for endothelial cell C4d staining, but background/nonspecific C4d labeling can be seen in: arterial elastic lamina; portal and perivenular elastic fibers; necrotic and steatotic hepatocytes, and areas of sinusoidal fibrosis (unpublished observation)(Figure 1E-F). Portal vein and capillary, sinusoidal, central vein, and arterial endothelium, lymphoid nodules, and periductal and portal stromal cell C4d staining has been described in native pediatric livers with hepatitis B (HBV) and C (HCV), and AIH[116] and in allografts when other insults are thought to be the primary cause of allograft dysfunction (e.g. biliary obstruction[107], recurrent HBV[108] or HCV[111], and plasma cell hepatitis (de novo AIH))[117]. Endothelial C4d deposits in rejection-unrelated allograft disorders, however, are reportedly less widespread and intense than in severe ACR or acute AMR[56,94,99,109,118,119].

Caveats aside, portal venous and capillary, arterial, and sinusoidal endothelial C4d staining has been significantly associated with CDC+ and DSA+ recipients more often than their negative controls[113]; in those with isolated AMR[95,115]; and associated with macrophage and plasma cell infiltrates[109], micro-vasculitis[1,50], and TCMR[55,98,107-115], which, in some studies, was directly proportional to Banff grade[107-115]. Portal microvascular and sinusoidal endothelial cell C4d staining appears to be most specific for acute AMR, but “portal C4d stromal” staining has also been described in ABO-incompatible AMR[110], ACR[113] and CR[98,120].

Since C4d deposition depends on the presence of target antigen and microvascular endothelial cell HLA class II expression can vary, the pattern of positive C4d staining can be contextual. For example, central perivenular form of TCMR can locally upregulate HLA class II leading to preferential C4d deposition in central vein and perivenular sinusoidal endothelium.

Peri-Transplant DSA

Since DSA pre-OLTx can impact early post-OLTx outcomes, more programs are testing; albeit in a minority of standard risk primary liver-only allograft recipients. High-risk patients (e.g. re-transplant candidates and dual organ recipients) should all have HLA testing, ideally before OLTx, to facilitate immunosuppression optimization and, in rare cases, unacceptable antigen listing. Sometimes flow crossmatch is the only information available, but ideally, centers further risk-stratify positive patients with single antigen bead testing. Highest risk recipients for acute AMR have preformed class I DSA in high-titer (or MFI >15,000 despite serial dilutions)[50], but risk is exacerbated by key factors including recipient illness (MELD score), and poor donor quality (Donor Risk Index (DRI)) (Figure 2). Lower level sensitization (MFI≥5000) in high MELD scores recipients or in recipients of suboptimal organs can increase the risk of allograft loss when class I DSA is present pre-OLTx.

When acute AMR is diagnosed, most patients have already failed steroid recycle and T cell-depleting-antibody therapy for steroid-resistant rejection and receive B-cell or plasma cell depletion with either Rituximab or a proteasome inhibitor, such as Bortezomib, with or without plasmapheresis and/or IVIG[134,135]. Since patients are already severely immunosuppressed, strict criteria are needed to establish an acute AMR diagnosis emphasizing specificity over sensitivity[50,55,57]. Early diagnosis is key to treatment success since delay usually results in irrevocable allograft damage and possible development of niche resident plasma cells that facilitate ongoing chronic AMR. Absent overt acute AMR, preformed DSA can also be associated with TCMR[1,38,51,58,98]. Fortunately, induction immunosuppression in patients with preformed class II DSA has been shown to markedly decrease this risk[51].

Managing Chronic DSA

Chronic DSA is almost always directed against class II, but DSA should never be interpreted in isolation: when discovered correlation with biopsy findings is needed. If pathology (inflammation and fibrosis) is present, the following should be considered: 1) optimize immunosuppression to facilitate/ensure compliance; 2) utilize a tacrolimus-based immunosuppression regimen, whenever possible; 3) increase the intensity of immunosuppression unless non-compliance is strongly suspected, and 4) treat/cure any concomitant disease that may be contributing to chronic allograft inflammation, such as chronic viral hepatitis B or C. Outside this clinical algorithm the best treatment for chronic AMR remains unknown. Clinical studies can only begin after diagnostic criteria for chronic AMR are created and endpoints for trials are agreed upon.