It is my honor and pleasure to present this year's Thomas Starzl Transplant Surgery State of the Art Lecturer for AASLD 2021. Dr. Anthony Demetrius is the Director of Liver and Transplantation Pathology and the Thomas Starzl Professor of Pathology at the University of Pittsburgh Medical Center. Dr. Demetrius is a practicing surgical pathologist and coordinator of the liver subsection of the Banff Foundation for Allograft Pathology. His research interests and laboratory focus primarily on the immune microenvironment in human livers and allografts and using liver biopsies from clinical trials. His laboratory specializes in maximizing the diagnostic value and pathophysiological insights from human liver biopsies through the integration of multiple data platforms using handcrafted and automated data analysis tools, including high-resolution digital imaging, multiplex labeling, image analysis, serology, mRNA expression, and mutational profiling. Dr. Demetrius has co-authored over 600 manuscripts of which nearly half have been with Dr. Starzl who was his academic mentor. The title of his presentation for the State of the Art Lecture is cross-platform analysis of human biopsies provide insights into the liver-based immune responses. Dr. Demetrius. Today, I'm going to discuss how traditional pathology is evolving and how we can use these advancements to understand antibody and T-cell mediated immune responses in human livers using formalin-fixed, paraffin-embedded tissue specimens. First, however, I would like to express my deep gratitude for the honor of being chosen for this presentation. Even though I am a pathologist, Dr. Starzl was my academic mentor. He attracted the best and brightest to Pittsburgh in the 1980s and 1990s when I had the tremendous advantage of working with the most brilliant surgical colleagues in the world. My hope is to reasonably represent that era and those people. As Dr. Starzl always reminded us, liver transplantation is an unfinished product that will only be finished when recipients can experience a normal morbidity-free life expectancy. Our current approach to native or allograft liver biopsy analysis is illustrated in this slide. Multiple inputs that include the clinical history and results of imaging studies, traditional histopathology, single histochemical, and immunohistochemical findings, and serological results such as liver injury test, donor-specific antibodies, autoantibody assays are integrated by the pathologist to determine the most appropriate diagnosis and incorporate that diagnosis into a patient management algorithm. Traditional liver pathology is undergoing an evolution from glass to digital, from single biomarkers to multiplexing, with semi-quantitative to dynamic range quantitative assessments, and from simply establishing a diagnosis to enabling and guiding therapy. In this presentation, I will illustrate how using next-generation pathology has led to a better understanding of antibody and cell-mediated immune reactions in human livers. I will focus on allografts, but the same principles apply to native livers. Before illustrating the value of next-generation pathology, there's one take-home message that I want to emphasize. Specifically, spatial context is important and should not be ignored or destroyed. In addition, next-generation pathology currently provides the highest resolution images. For example, a standard MRI has a resolution on a scale of several cubic millimeters. In contrast, a slide scanned at 40x magnification yields images with a submicron resolution scale. This results in 10,000 times more resolving power in the digital slides which enables examination of intercellular relationships and segregation of molecules on the surface of a single cell, as will be illustrated later during the presentation. For example, the top-right image shows an immune synapse formed between a Kupfer cell and a lymphocyte. The bottom image illustrates CD45-positive lymphocytes within the lumen of a CD34-positive portal capillary. This is referred to as capillaritis, an important feature of antibody-mediated liver injury. Each of these cellular interactions were automatically identified by image analysis software and are easily quantified over the entire biopsy specimen. The first part of the presentation will focus on antibody-mediated rejection. Acute AMR is the most straightforward to recognize, but is relatively uncommon. The contribution of antibodies to chronic injury has been more challenging to unravel. Traditionally, whether for allografts or native livers, circulating antibodies are thought to be biomarkers of immune activation and nothing more. However, we now know that antibodies can directly cause liver allograft damage. This was first recognized with ABO isoagglutinins. Currently, treatment regimens are used to decrease circulating isoagglutinin titers to enable transplantation across ABO barriers. Several years later, it was recognized that alloantibodies can also directly cause liver allograft damage, although less so in the case of native livers. Although less so than the isoagglutinins. Second, few experimental animal models exist in the literature. The reason for this is twofold. First, there is a clinical bias that antibodies are not that important in liver transplantation. And second, rodents generally show much lower levels or lack microvascular endothelial cell, MHC class 2 expression when compared to humans. And finally, interactions among antibodies, complement, and T cells with respect to liver injury is ripe for further investigation. This slide illustrates the baseline expression of major histocompatibility complex antigens in normal human liver tissue. The left panel shows the expression pattern of class 1 MHC antigens. They're expressed on all cells, but are strongest on endothelium and bile duct epithelium and weakest on hepatocytes. The right panel illustrates MHC class 2 expression that is stained red and localized primarily to macrophages. However, the second panel overlay shows CD34 staining, which highlights portal capillaries. Some of these capillaries also show low-level MHC 2 expression. Lastly, the cyan stained CD163 are cells belonging to the macrophage lineage. You can appreciate the extensive overlap with MHC 2. This slide is from Jackie O'Leary's publication regarding the fate of preformed donor-specific antibodies in the peripheral circulation of presensitized human liver allograft recipients. Dr. O'Leary's seminal publications were largely responsible for the resurgence of interest regarding the influence of allo antibodies in liver transplantation. As can be seen on the left side of this figure, most anti-class 1 antibodies are depleted from the serum within days of transplantation. However, as shown on the right side of the diagram, many anti-class 2 antibodies persist, and the difference between class 1 and class 2 is significant. This slide illustrates the distribution of allo antibody binding in target tissue within the first several weeks after transplantation in a presensitized liver kidney allograft recipient from the same donor. The kidney biopsy shown in the top row was obtained on day 15. The liver biopsy shown on the bottom row was obtained on day 13. The left column shows the H and E findings. The right column shows the C4D staining pink, the capillary deposition of C4D. Note the widespread distribution of C4D staining in the kidney with its restricted distribution in the liver. Both organs showed sludging of leukocytes within the C4D positive capillaries. In the kidney, this is immediately recognizable as acute antibody-mediated rejection. In the liver, without C4D staining, these same findings are usually suspected as illustrating a biliary tract issue. Note that the C4D deposits primarily in CD34-positive portal capillaries, but much less so in sinusoids. This slide illustrates two important points. First, the type of patients at risk for acute antibody-mediated rejection and the typical clinical course. And second, routine histopathological and immunohistochemical findings. The left panel shows that patients at risk are typically highly sensitized individual with previous, often multiple, exposures to foreign MHC antigens. They also often suffer from autoimmune disorders and harbor multiple anti-donor, anti-HLA antibodies before transplantation. Immediately after transplantation, they usually experience otherwise unexplained refractory thrombocytopenia. As can be seen from this representative recipient, the left column on the right side of the slide shows diffuse C4D deposition in the portal vein branches and portal capillaries, which is stained red. The right column shows marked endothelial cell hypertrophy and leukocyte sludging within C4D positive vessels, sometimes leading to complete inflammatory cell occlusion. The top panel of this slide shows the gastroenterology visual abstract from the multicenter IWF Immunosuppression Withdrawal Study publication led by Drs. Sandy Fang and John Bucavales. In my opinion, Dr. Fang rivals Dr. Starzl in her ability to successfully carry out translational research in the clinical arena. Dr. Alberto Sanchez-Fueo conducted the mRNA expression studies and our laboratory served as the central histopathology, multiplexing, and imaging core. The upper left corner shows that these were highly selected patients. From nearly 3,000 patients from 12 centers, only 157 recipients who were greater than four years after transplantation were eligible and consented for immunosuppression withdrawal. These are the best of the best recipients with normal liver injury tests. At entry, all participants underwent a liver allograft biopsy to determine eligibility according to BAMF criteria. Histopathology parameter clustering, but not mRNA expression profiling, was able to separate the participants into three groups or clusters. Biopsies from Cluster 1 participants showed interface hepatitis. Biopsies from Cluster 2 recipients showed increased fibrosis with or without inflammation. Biopsies from Cluster 3 were near normal. However, once the biopsies had been separated into distinct clusters by histopathology, two important observations were made. First, Cluster 1 patients with interface hepatitis showed an mRNA expression profile typical of T-cell-mediated rejection. Second, the same Cluster 1 patients also showed the highest prevalence of donor-specific antibodies. Those with normal biopsies showed no mRNA evidence of T-cell-mediated rejection and the lowest prevalence of donor-specific antibodies. The purpose of showing this slide is that it foreshadows our cross-platform approach shown in the next slide. This slide represents our current approach to triage of data derived from clinical trial specimens. The left side shows the multiple data inputs used in our analyses. Included are traditional histopathology, multiplex staining and quantitative morphometry, clinical information and imaging study results, serology and genomic data. Data from each platform is color-coded with arrows on the left and amalgamated using a bi-clustering data mining approach on the right. Instead of manually and sequentially filtering through data harvested from different platforms, as we did for the gastroenterology publication shown in the previous slide, we are now employing bi-clustering data mining approaches. The latter enables amalgamation of data harvested from multiple platforms. Bi-clustering algorithms have been available for years and are freely available in R. We are using this approach to better understand the relationships among results obtained from serology such as DSA, multiplex staining and quantitative image analysis, and mRNA expression profiles from the same eye-width patient population. Numerous studies show a strong association between DSA presence and liver allograft inflammation and damage. Whether this association is causally related or not is an unsettled question that we have been addressing. This slide illustrates the two-hit hypothesis we published with Jackie O'Leary. We are currently testing this hypothesis. The upper left panel shows the normal circumstance where very little MHC2 is expressed on the allograft microvasculature. However, as shown in the upper right panel, after injury or when the graft is inflamed, there is upregulation of MHC class 2 on the portal microvasculature, which serves as a trigger for sensitization and a target for the DSA. Increased MHC class 2 expression facilitates DSA binding to target antigens that can result in either direct endothelial damage through complement-mediated endothelial cell injury, as shown in the lower right panel, or facilitation of microvascular injury through recruitment of macrophages and NK cells via ADCC mechanisms and microvascular inflammation, as shown in the lower left panel. Using our approach, we can directly test the two-hit hypothesis using pre-existing data harvested from image analysis software. The left panel shows that patients with neither DSA nor C4D binding showed the least capillaritis or margination of leukocytes within the portal microvasculature. In contrast, participants with two hits or a positive DSA and C4 deposition showed significantly more margination. The right panel, Y-axis, shows increasing MHC class 2 expression on the portal microvasculature, and DSA MFI sum is shown on the X-axis. The extent of capillaritis or margination is illustrated by the colored dots. Green dots represent low scores, with a gradual transition to red dots representing high scores. As can be seen, participants without DSA and low MHC class 2 expression show the lowest level of capillaritis or margination. The right panel, Y-axis, for expression of MHC class 2 on the allograft microvasculature, is both a persistent stimulus for sensitization and a target of the antibodies form. Using a biclustering approach, if we then seed mRNA expression profiles with increasing microvascular inflammation, and low MHC class 2 expression on their microvasculature, and a strong positive serum DSA, pathway enrichment analysis yields genes associated with both T-cell mediated rejection and antibody mediated rejection, or possibly a mixed pattern of rejection. Included are positive regulation of leukocyte and T-cell activation, regulation of leukocyte adhesion, cell-to-cell adhesion, and complement activation via the classical pathway. The important take-home message is that the iterations among data derived from various platforms can provide new insights into tissue biology. It can also be used as a discovery and validation tool by re-examining or re-staining tissue specimens for positively regulated networks. Perspectives on interactions of T-cells with liver tissues. First, T-cells are thought to be primary mediators of liver injury and most human chronic necro-inflammatory liver diseases. Experimental animal models suggest that, under normal circumstances, T-cells default toward Treg cell differentiation and ineffective effector T-cell maturation. However, under a pro-inflammatory liver immune microenvironment, liver priming promotes effector T-cell maturation. This slide nicely illustrates the power of multiplex staining in image analysis. These H&E stains can be used to compare pre-weaning biopsies from two separate eye-width patients. Note that, by routine histopathology, they are similar in appearance and without inflammation and fibrosis. However, serial slices of the same biopsy stained with CD34, CD45, and MHC class II shows dramatic differences that were not apparent on the H&E sections. The top panel was from a participant that was deemed tolerant after lowering immunosuppression. The bottom panel came from a recipient that experienced rejection during weaning. Note the dramatic differences between the two, especially the number of lymphocytes located immediately adjacent to antigen-presenting cells within the lobules. The APCs are most likely Kupfer cells. This is the first time we realized a link between basic science animal models and humans. Importantly, the immune microenvironment of the liver is important in the quality of T-cell priming that occurs. This slide illustrates the approach used to verify that we were indeed identifying true immune synapses in formalin-fixed, paraffin-embedded tissue sections. It combined both morphometric and immunohistochemical findings. Numerous basic science studies show that when lymphocytes engage in true immunologic synapses, they become closely adherent to antigen-presenting cells and undergo flattening. Therefore, we defined a synapse as follows. The edge of a lymphocyte nucleus was within 3 microns of the antigen-presenting cell nucleus. The lymphocyte nucleus was flattened, and the pair showed nearly parallel long axes alignments. Pairs identified by morphometry were then examined for supramolecular activation complex formation that indicates stable binding and true immune synapse formation. As shown in the top panel on the right, the supramolecular activation complex is identified at high-resolution microscopy by exclusion of CD45 and concentration of CD3 staining. Transient binding and unbound cells can also be recognized by their staining patterns as shown in the middle and bottom panels on the right. The bottom panel on the left shows that lymphocytes engaged in pairings with APCs showed significantly more supramolecular activation complex formation and true synapse formation. This slide shows an updated prediction cube of the one we originally published in Hepatology earlier this year. The axes include lymphocyte-APC pairs or synapses, recent monocyte immigrants expressing MAC387 or calprotectin, and lobular CD8-positive cells. Those inside the cube are predicted to be tolerant and represented by solid circles. Green circles are those actually tolerant, and red were ultimately rejecters. The lobular lymphocyte-APC pairs exert the strongest prediction influence. The updates we used included an improved nuclear segmentation algorithm and a focus only on lobular CD8-positive cells instead of total CD8-positive cells. This approach was taken for several reasons. First, previous studies decades ago from Cambridge showed lobular CD8-positive cells were more important than the total. Second, we improved and validated our immune synapse detection. And third, experimental animal models with in vivo microscopy showed that priming by lobular APCs played a major role in the quality of T-cell priming within the liver. As can be seen for the IWF patient population, with the updated approach, we can correctly predict all tolerant recipients and misidentified only 9 of 35 or 25% of the non-tolerant participants. This showed that a preexisting inflammatory liver microenvironment promotes rejection after weaning. We are currently testing whether this approach can be validated in independent patient populations. When mRNA expression profiles were seeded with increasing lobular immune synapses and pre-weaning biopsies from IWF patients, the most correlated gene list and enrichment analysis yielded the interesting results shown here. Specifically, bioclustering analysis showed upregulation of pathways associated with complement activation and humoral immunity. As expected, it also showed MHC protein complex binding and positive regulation of T-cell activation pathways, which we observed after immunosuppression weaning. These results point toward links between complement antibodies and T-cell priming in the liver, perhaps by opsonizing targets. Also, it is well known that local production of complement by T-cells and antigen-presenting cells enhances alloreactivity. However, to verify the validity of this approach, we then circle back to the tissue specimens to determine the spatial context of protein expression of these coordinately regulated molecules. Regardless, the roles of antibody and complement are an understudied area of liver immunobiology. We are currently using the approach described today in preclinical translational research and clinical settings. We have found it useful for transplant experimental animal models, studying the cancer immune microenvironment, immunosuppression weaning studies, drug development, and extracorporeal perfusion. In essence, next-generation pathology enables robust, reliable, and quantifiable spatial histopathomics output that can be integrated with data from multiple other platforms. It leads to more reproducible and quantifiable results. In some cases, it completely eliminates subjectivity and can lead to artificial intelligence-enabled pathology and automated grading for certain parameters, as was shown for immune synapse formation and microvascular inflammation. Before closing, I would like to acknowledge all of those who made this presentation possible. Although the list is too long to mention all by name, some highlights are provided. At the top of the list is Dr. Thomas Starzl, who put his trust in me starting as a resident. That trust continued throughout many productive years. I would also like to acknowledge all of my friends and surgical colleagues from that golden era in Pittsburgh. We accomplished a lot and had a lot of fun in doing so. My pathology colleagues in Pittsburgh, who have been up with me for decades, especially Dr. Zivi, who is my alloantibody mentor. The IWITH leadership team and site PIs, who organized the IWITH trial and provided invaluable material to study. The Immune Tolerance Network, transmedics, Novartis, and CTOTC, who trust in our abilities as the central core laboratory for their pathology studies. The transplant program at Baylor Medical Center in Dallas, some of whom have retired or moved on, especially Goran Klintman and Jackie O'Leary, but also the new leadership team. The BAMF Foundation on Allograft Pathology that provides a respected platform for discussing, testing, and publishing consensus criteria on various aspects of liver allograft pathology. The current surgical and hepatology transplant teams in Pittsburgh, they have indeed created a new era in Pittsburgh with all of their hard work. And last but not least are the folks in my lab who do much of the work that was presented today. Drew Lesniak, who has unparalleled expertise in digital imaging and image analysis. Michelle Wood-Trageser, who keeps the ship in order. Kayla Golnowski, our multiplex panel developer. And Ben Popp, a digital slide scanner and software coder. In conclusion, traditional pathology is evolving towards digital imaging, multiplexing, and automated image analysis, which adds substantial value for translational research and patient management. Spatial context at a submicron resolution scale provides critical data that can be used to uncover underlying pathophysiologic mechanisms of allograft injury and acceptance. And finally, cross-platform analyses using large data sets and by clustering data mining algorithms enable translational research to be conducted using the same approaches we use for patient management.

Dr. Anthony Demetrius

University of Pittsburgh Medical Center

liver pathology

transplantation pathology

immune microenvironment

digital imaging