Final Report Summary - HUMAN (Health and the Understanding of Metabolism, Aging and Nutrition)
Executive Summary:
The prevalence of metabolic diseases has dramatically risen owing to the great increase of life expectancy and cultural transitions in lifestyle and nutrition, with a predicted massive economic burden for the European health systems. In the last 15 years, great research efforts have been devoted to identify the genetic basis of age related and metabolic diseases, mostly by means of large genome-wide association studies (GWAS). GWAS studies have reported associations of thousands of single nucleotide polymorphisms (SNPs) for hundreds of traits/diseases. Despite the highly significant association with the trait of interest, the functional role of all these genetic variants need to be elucidated especially in term of genotype/phenotype correlation. Otherwise, the potential of the identified SNPs for clinical and pharmaceutical application for i.e. the treatment and prevention of metabolic diseases is dramatically reduced. To overcome these limitations one of the main goals of the HUMAN consortium was the generation of mouse models highly repopulated with human hepatocytes or carrying pancreatic beta-cells from either primary cells (hepatocytes) or induced pluripotent stem cells (iPSCs). This innovative approach offers the unique possibility of studying the physiological and pathophysiological processes associated to metabolic diseases in human-derived organs, but placed in an integrated living system (i.e the mouse body). A unique strength of the project is that humanised mice are a model for a translational research that is translatable to the human condition. iPSCs used to generate hepatocytes and beta-cells derives from patients affected by severe metabolic diseases such as type 2 diabetes (T2D) with or without complications, or subjects selected for nearly the complete opposite phenotype, for their nearly complete lack of disease and exceptional longevity. In the end of the HUMAN consortium, we have focused on the most relevant genetic variant (FCF7L2) amongst those that emerged from GWAS studies for its association with T2D and metabolic syndrome. Mice with humanised liver and/or pancreatic beta-cells have been generated, and the phenotype that results from the risk or protective alleles characterised. Since complex diseases such as T2D and metabolic syndrome are polygenetic and partially dependent on the genes selected here, the creation of the risk and protective cell lines and mouse models provides a platform technology for discovery of genes and metabolic pathways that cooperate with these known risk alleles to manifest the full spectrum of the disease. Furthermore, studies with drugs targeting nuclear receptors (i.e. Liver x Receptor; LXR) demonstrated the validity of the liver humanised mouse model to predict the possible negative outcomes observed in phase II human studies and thus allowing an early evaluation of the translatability to human of pre-clinical drug development. The main goal of the project is to offer European academia and and industries exclusive tools to tackle the challenge of functional validation of metabolic disease-associated genetic variants by offering innovative and robust humanised animal and cellular models and a portfolio of new and validated therapeutic targets for better understanding of metabolic diseases and healthy aging.
Project Context and Objectives:
WP1
The WP1 objective is to assess the role and the importance of selected genes/alleles/phenotypes in metabolism and longevity, by (i) selection of human donors for the generation of humanised mouse models, (ii) assessment of the role of selected genes/alleles/phenotypes in metabolism and longevity in human studies and (iii) comparison of results from human studies and those obtained from humanised mouse models.
The strategy for selection comprised the presence of risk and/or protective alleles for SNPs of interest in the TCF7L2, APOE, and FTO genes and in combination with metabolic phenotypes (related to glucose and lipid metabolism) and/or clinical phenotypes (covering the spectrum from healthy longevity and absence of disease towards cardiometabolic disease with complications). The activity in this work package were instrumental for WP2 and WP3 by delivering fibroblasts to be used for production of iPSC or by delivering hepatocytes to directly produce liver humanised mice.
WP2
The objective of WP2 is the generation and differentiation of induced pluripotent stem cells (iPSC) to hepatocytes and pancreatic β-cells. WP2 is closely linked to WP3 “generation of humanised liver and pancreas mouse models”.
In this work package, two separate groups worked to improve the ability to make liver cells (hepatocytes) from stem cells (iPS cells). The work has resulted in many scientific publications and more will appear. Compared to 5 years ago when the grant was written, the liver cells made today from stem cells are much closer to authentic liver cells, and they provide many mature liver functions. In addition to making better liver cells, we have provided a method and a database by which we and any other laboratory can compare their stem cell-derived liver product to authentic human fetal or adult liver samples. In this manner, one can easily determine the state of differentiation of the experimental group to the values found in real fetal or adult human liver. This database and method will be made public for general use. Many laboratories do not have access to authentic human liver samples like we do, so this public data base will be of great help to researchers all over the world.
WP3
The objective of this WP is the generation of humanised liver and/or pancreatic mouse models. Mice will be made from: 1) Transplanted hepatocytes from normal donors; 2) Transplanted hepatocytes from metabolic liver diseases; 3) Transplanted hepatocytes from specifically selected genotypes; 4) pancreas humanised mice where cell encapsulation device technology is used to transplant pancreatic progenitors into immune-compromised rodents and allow maturation towards fully differentiated β-cells in vivo; 5) mice with both liver and pancreas humanised mice.
In this work package, we created interesting and useful models of serious human diseases. Many mouse models do not faithfully reflect the human disease. To make the models more human relevant we have generated mice that are highly repopulated with human liver cells (hepatocytes). The liver cells in these liver-humanised mice perform their normal human liver functions and are human-relevant models for blood lipids such as cholesterol, the production of bile acids, the regulation of glucose levels, drug and ammonia metabolism. Liver–humanised mice were also created to model serious defects in liver function such as urea cycle deficiencies, which inhibit the patients ability to eat normal protein and excrete the excess nitrogen that is generated in that process. Liver humanised mice were created for two life-threatening liver-based diseases that affected the urea cycle. In both cases, the mouse model produced the symptoms characteristic of the human disease. These mouse models have proven useful to examine new drug, cell and gene editing therapies and will provide data that can be extrapolated to human patients since the target cells for the disease and the therapy are actual human liver cells.
WP4
The aim of this WP is to fully characterise the metabolic and behavioural phenotype of humanised mice by using a standardized approach and with attention to how the genetic variations affect metabolism towards a greater susceptibility to metabolic disorders.
The research groups working in this working package were in close contact with the research groups acting in WP1, WP2, and WP3. The characterization of the animals generated included analysis of the feeding habits, of how energy was produced
and whether more or less sugars were burned in comparison to fat. In addition, the characterization of this animal included the analysis of the changes in lipid in circulation, in the liver and in other tissue. Further, the liver humanised mice were also utilized for study aiming to demonstrate that this animals model gives result that are translatable to human condition when new drugs are developed. The different experiment performed also generated biological material that have been further analyzed in the WP5, WP7 and WP8.
WP5
The WP5 objective is to understand how the brain-liver–brain axes contribute to the phenotypic traits observed in WP4.
The hypothalamus integrates information on metabolism from hormonal and vagal nerve inputs. In turn, the brain provides adaptive responses both at the behavioural and metabolic levels, through endocrine and autonomic controls of peripheral organ activity notably liver, pancreas, muscle, adipose tissue and brown fat. Thyroid signalling is a prototypical endocrine candidate for controlling metabolism by liver /brain communication. The liver is a primary source of production of T3. Changes in thyroid status will in turn modify a number of metabolic pathways both directly in the periphery and indirectly through actions in the hypothalamus and the autonomic nervous system, e.g. central levels of T3 modulate glucose production by the liver through the sympathetic NS. Available wild type and mutant mouse models were used as tools to get more insights on the metabolomics and transcriptomic hypothalamic networks underlying central responses to either increased circulating T3 or increased intra-hypothalamic T3.
WP6
The overall task was to genetically convert different SNP/mutations in induced pluripotent stem cells (iPSCs) derived from fibroblasts using the CRISPR technology and screen the corrected clones for off target modifications. Additionally, the corrected clones with low off target frequency was selected for differentiation into hepatocytes and further transplanted into FRGN mice. Finally, the humanised mice were screened for normalised phenotype by the genetic correction.
In WP6 we greatly exceeded the studies initially proposed in the application. Building on the liver-humanised mice studies in WP3 we were able to create mice with serious human liver diseases that affect the metabolism of proteins and ammonia, called urea cycle defects. These diseases are so serious that the liver cells used for these studies were isolated from the liver of pediatric patients that had to receive a liver transplant to save their life. The diseases cells were isolated from the patient liver after it was removed for the transplant. As indicated in WP3, when we repopulated the mouse liver to high levels with the patient’s mutant cells, the mice displayed the same symptoms of the urea cycle defect as the donor patient. In separate studies, we were able to use state-of-the–art gene editing procedures to correct the mutation in the patients liver cells. When those “corrected” cells were transplanted into the mice, the mice no longer displayed the symptoms of the disease. Thus, we were able to demonstrate that it is possible genetically correct mutation in a patient own liver cells and if those cells are transplanted, the symptoms of the disease are corrected. Through WP 3 and WP6 we were able to firmly demonstrate that the genotype of the hepatocytes transplanted into the mice determines the phenotype of the mouse that is generated. This new data is highly relevant information to help move the gene editing therapy to the clinic. Given the abundance of research into gene editing technology and the importance of the human genome sequence and the human DNA replication and repair pathways to these processes, it would be advantageous to perform preclinical studies of gene editing technologies with human hepatocytes, and when possible, with the actual disease-affected human hepatocytes. The value of the data obtained from such studies will be greatly enhanced if those studies could be conducted, in vivo. Additionally, studies with live animals that can be followed after gene editing or other therapies for long periods of time will enable the assessment of long-term outcomes such as phenotypic correction of the disease and safety studies. The model reported here will be expected to be useful for investigations of modified RNA, gene, cellular and small molecule therapies for genetic-based liver diseases. Perhaps where these liver-humanised models will be most informative will be in the investigations of the safety and efficacy of gene editing technologies and their translation to the clinic.
WP7
The WP7 objective is to use metabolomics as tool for a comprehensive characterization of the biological specimens derived by the different models generated within the HUMAN Consortium. The major goal is to deliver systematic molecular profiles and to identify underlying biochemical pathways altered in metabolic diseases, correlated with disease-linked genotypes, or connected to longevity in order to improve understanding of healthy aging, disease aetiology and for suggestion of new treatment strategies.
WP7 has worked closely with WP1, WP2 and WP3 so that quality all through the different stages of the research path flow chain from donors, via cell cultivation to the humanised mice, was assured by analysis of the biochemical pathways. WP7 results provided WP8 with candidate proteins for targeted proteomics in order to validate the suggested biochemical pathways. The biochemical pathway analysis has also been instrumental to provide suggestions for possible new therapeutic strategies. The work in WP7 has been conducted in close collaboration with WP9 for data integration and interpretation. The working hypothesis of WP7 was that complex gene expression networks coordinate the metabolic fingerprint of the disease and that their function can be observed in the endogenous metabolic profile.
WP8
The WP8 objective is to characterize the molecular characteristics of mouse with humanised livers by a comprehensive analysis of the modification of DNA, the expression of the genes and of proteins (i.e. epigenomic, transcriptomic and proteomic) using advanced state-of-the-art technologies (i.e. high-throughput ‘Omics’ technologies).
The activity of WP8 were fundamental for the characterization at molecular level of the humanised mouse models by linking molecular alterations to the physiological phenotypes observed in the experiment conducted in WP4 and to the metabolic profiles obtained in WP7. As for WP7, the quality controlled data obtained in WP8 have been transferred to WP9 for integrative analysis and identification of affected pathways, drug targets and biomarkers.
WP9
The WP9 objective was the provision of a central database infrastructure, which enabled structured management and secure access to the raw/processed multi-omics data and associated phenotypic data generated within the HUMAN consortium, and the generation of high-quality processed data.
The data transferred to WP9 have been assessed for quality by using standardised and validated high-performance workflows. Data pre-processing and quality control, including report generation, have been done according to validated workflows that have been developed by a research partner of the Human consortium. The results have been generated to ensure an efficient high throughput and standardized analysis and to meet the project’s specific requirements for the latest ‘omics technologies. Recommendations for re-processing or selective exclusion of raw data from further analyses has been provided to project partners when necessary. WP9 has performed an integrated, multivariate analysis of all ‘omics data types to establish statistically significant associations and their relevance determined by elucidation of cross-platform evidence for regulatory and metabolic networks.
Project Results:
For the generation of humanised mice, human cells were from two different sources, namely (i) primary human hepatocytes for direct transplantation in mice, and (ii) human fibroblasts for pluripotent stem cells (iPSC) generation and subsequent differentiation into mature hepatocytes and beta cells prior to transplantation in mice.
The HUMAN consortium had access to several large collections of human primary hepatocytes that upon genotyping were shown to exhibit a wide distribution of protective and risk alleles for TCF7L2, APOE, and FTO. Hepatocytes from donors with protective alleles for all selected SNPs in TCF7L2, APOE, MC4R, and FTO as well as hepatocytes from donors with risk alleles in TCF7L2, APOE, and/or FTO were selected for later transplantation in mice.
The researchers working in the WP1 had also access to unique collections of cryopreserved fibroblasts from cohorts of semi-supercentenarians and the Leiden Longevity Study. After Ethical approval, selected fibroblasts were further characterized genetically and epigenetically and used for the generation of iPSC lines. Genetic characterization revealed that the mTOR coding sequence is extremely conserved, indicating that this gene is likely involved in vital physiological functions and barely tolerates amino acidic variability. In a novel study for which ethical approval was obtained in first year of life of the HUMAN Consortium, fibroblasts were collected and cryopreserved from biopsies from 24 newly recruited patients with type 2 diabetes mellitus, and selected based on presence of risk alleles and clinical diabetic complications.
In first years of HUMAN, the main focus was on the selection of human fibroblasts based on a the presence of risk and/or protective alleles for selected SNPs in TCF7L2, APOE, and FTO in combination with metabolic and/or clinical phenotypes. Due to technical challenges related to derivation of differentiated human hepatocyte by iPSC of fibroblasts, our focus later changed to selection of primary hepatocytes, based on the distribution of protective and risk alleles for TCF7L2, FTO or APOE risk alleles for transplantation into mice and the characterization of the mice thus obtained.
To further elucidate the role of selected genes/alleles/phenotypes in human metabolism and longevity, an important focus of WP1 has been the integrated analysis of genotype/phenotype correlations in different human donors/cohorts. These analyses were performed around three main themes: i) genotype-phenotype associations; ii) sleep and cardiometabolic disease; and iii) causal inference.
Genotype-phenotype associations.
Regarding APOE, we estimated in the Leiden Longevity Study that the contribution of the APOE isoforms to the previously identified phenotypic and metabolic differences between offspring and controls (e.g. diabetes mellitus prevalence and HDL cholesterol levels) was minimal. Based on literature, we put forward the hypothesis that increased physical activity and increased intake of omega-3 fatty acids are potential working mechanisms to diminish the detrimental effects of APOE ε4 on the risk of developing age-related diseases. Regarding TCF7L2, we observed in the Active and Healthy Aging Study, and the Leiden Longevity Study that carriers of the risk allele had higher nocturnal glucose levels, similar levels of glucose during the day, and that they displayed a comparable degree of variability in glucose levels. However, part of this association could be explained by adiposity, since adiposity was associated with higher nocturnal and diurnal glycaemia, but not with variability of glucose levels.
In the Netherlands Epidemiology of Obesity Study, we observed that genetic variation in TCF7L2 was mainly associated with insulin levels and with type 2 diabetes mellitus treated with insulin analogues, which suggests that TCF7L2 associated with a type 2 diabetes mellitus phenotype which is more critically dependent on insulin secretion rather than insulin resistance. Regarding FTO, we found no evidence for an association of the FTO risk allele rs1421085 with measures of energy metabolism in the Netherlands Epidemiology of Obesity Study.
Sleep and cardiometabolic disease
In the Netherlands Epidemiology of Obesity Study, we found that short as well as bad habitual sleep were associated with an increased body mass index. We did not find evidence that short or bad habitual sleep were specifically associated with a higher amount of visceral fat, as measured using magnetic resonance imaging. In the same study population, we additionally found that short as well as bad habitual sleep was associated with an increased insulin resistance and higher concentrations of triglycerides in blood and liver. In collaboration with the international CHARGE consortium, we initiated a sleep-SNP interaction study on serum lipid levels with the purpose to study the biological background of sleep-associated dyslipidemia. Using data from over 40 cohorts of different ancestries (N>125,000), multiple genetic variants mapped to genes not earlier described in relation were identified and replicated in an independent analysis.
Causal inference
In HUMAN, we conducted several Mendelian Randomization studies to infer causal associations of phenotypes of interest to the consortium. We were also part of an international collaboration to examine the causal association between (central) adiposity and age-related disease outcomes, which found that central adiposity and overall adiposity have a distinct cardiometabolic risk profile. In addition, we found, using publically available data from the DIAGRAM consortium, that high concentrations of the liver enzyme gamma-glutamyltransferase was not causally associated with an increased risk of type 2 diabetes mellitus, although observational research suggested such relationship. Furthermore, increased concentrations of thyroid stimulating hormone (TSH) have been associated with increased risk of developing type 2 diabetes mellitus. Since, this association was causal it is evident that common risk facors and life styles can lead to type 2 diabetes mellitus and disturbances of the thyroid gland. Increased low-grade systemic inflammation, however, was found to associate with an increased risk of type 2 diabetes mellitus using Mendelian Randomization, in line with observational studies. However, low-grade systemic inflammation is complicated and is dependent on several factors. Based on our results, is could be inferred that other inflammatory factors are causally associated with the risk of type 2 diabetes mellitus. Based on a newly performed genome-wide association study on circulating cholesterol ester transfer protein (CETP) , we identified 3 independent genetic variants mapped to the CETP gene that could be used in Mendelian Randomization studies. However, we did not observe a causal association between genetically-determined CETP concentrations and the risk of cardiovascular disease and found that CETP was causally associated with the concentrations of VLDL particles rather than LDL particles was previously thought.
In WP2 we proposed to make liver cells from stem cells. Now with the much-improved protocols we established, it is possible to make billions of human liver cells from stem cells with significant levels of liver function. One of the groups involved in WP2 was the group that conducted the first liver cell transplants in patients to correct liver disease, and has the largest group of patients treated by liver cell transplantation. Once issues of scalability of the processes and safety are met, the ability to generate liver cells from a seemingly endless source of stem cells, will enable us to expand liver cell therapy to treat liver disease to many more patients than are currently possible. This could have a huge impact on the health, welfare and quality of life for thousands of patients, and should help correct liver diseases that are now only possible to treat by liver transplants. Since iPS could be made from the patients themselves, or universal donor cell banks could be generated, the future liver cell transplants could be performed without suppressing the patient’s immune system with drugs. This program funding allowed us, for the first time, to begin gene-editing technology. Cells from many normal patients, centenarians and those with severe liver diseases were reprogrammed to iPSc. These stem cells were studied by themselves were coaxed, in culture to become liver cells and were also used as models for the different human diseases from which they originated. The stem cells were the first targets of the gene editing programs because they were abundantly available which allowed us sufficient cells for our initial trial and error experiments to improve gene-editing technology. These studies proved crucial to the later steps where actual human liver cells isolated from the patient’s own liver after it was removed during a liver transplant procedure were edited to correct the mutation in the patient’s liver cells.
It is clear that animal models of many diseases that affect human patients do not accurately reflect the disease observed in the patient. Of the over 20,000 genes in the human genome, more than 1/3 of them are expressed in the human liver. Mutations in many of these genes result in serious human live diseases. We started from the hypothesis was that the best model for a human metabolic liver disease that affects hepatocytes (liver cells), is one generated by actual disease-affected human hepatocytes. If that model could be investigated in vivo, the value of the model would be greatly enhanced. We created models where the liver of mice can become nearly completely replaced by human liver cells and showed that when normal liver cells are transplanted a normal mouse is generated, but when diseased human live cells are transplanted, the human disease is now recreated in the mouse. In WP 3, these liver humanised mice were generated and studies in great detail. The results proved that the physiology and functions of the human liver can be studied in these liver-humanised mice. This is a huge breakthrough for researchers because the modes accurately reflect the human disease and the studies will reduce the number of animals needed for research studies since the humanised models will provide data that is more relevant to human patients and many studies will be able to be conducted on live animals.
As all well-planned research programs should, success in one part feeds the research in multiple other parts of the project. Studies in WP2 reprogramming cells to stem cells allowed us to use the stem cells to begin our gene editing projects in WP6 and verify that the gene editing also changed the biology and the properties and functions of the liver cells made from the stem cells. Likewise, progress with gene editing of stem cells in WP6 and the generation of the liver-humanised mice conducted in WP3 fed the progress made in gene editing with actual liver cells as reported in WP3 and WP6.
For several genetic traits, we firmly established that the genes that are expressed (the genotype) in the human liver cells that are transplanted into the mice determines the characteristics (phenotype) observed in the liver-humanised animals. These observations are of critical importance for studying normal physiology and biology as well as human liver disease. Mice made with specific genetic defects in liver function, recreated the symptoms of the disease observed in human patients. Thus, we have verified that the liver-humanised mouse model is a faithful model and surrogate with which to study and correct these serious human diseases. Also, the diseased hepatocytes used for these studies were derived from the liver of patients that had to receive a liver transplant for life-threatening diseases. Thus, we have created models of serious human liver diseases in liver-humanised mice from the actual diseased liver cells removed from the patient. With these newer technologies like gene editing, the target of the editing procedure is critical to the outcome. Editing of a specific gene in mouse cells or mouse liver will tell us virtually nothing about the safety or efficiency of that procedure in human cells or human liver. Critical information for gene editing procedures are the DNA sequence being edited and the DNA replication and repair pathways in the cells themselves. Since these are much different between mouse and human cells, the mouse is not a sufficient model to test the safety or efficiency of the gene editing procedures. One cannot extrapolate gene-editing data from mice to human patients. It would be advantageous to perform preclinical studies of gene editing technologies with human hepatocytes, and when possible, with the actual disease-affected human hepatocytes. The value of the data obtained from such studies will be greatly enhanced if those studies could be conducted, in vivo. All of those aims have been met with the liver-humanised mice generated here. Additionally, studies with live animals that can be followed after gene-editing or other therapies for long periods of time will enable the assessment of long-term outcomes such as phenotypic correction of the disease and safety studies. The model reported here will be expected to be useful for investigations of modified RNA, gene, cellular and small molecule therapies for genetic-based liver diseases. Perhaps where these liver-humanised models will be most informative will be in the investigations of the safety and efficacy of gene editing technologies and their translation to the clinic to correct liver disease in patients.
The WP3 successfully created liver-humanised mice with specific genotypes, and the phenotype of the mice reflects the genotype of the human hepatocyte donor. Importantly, we can gene edit the human hepatocytes and the phenotype of the liver-humanised animals reflects the genotype of the donor human hepatocytes originally transplanted. As was shown in other work packages with the lipoprotein profiles, the bile acids, the blood glucose levels and here with the urea cycle activity, the phenotype of the liver-humanised mice faithfully reflects the genotype of the human hepatocyte donor. Concerning gene editing, HUMAN has demonstrated for the first time that hepatocytes can be gene edited, ex vivo, and that the gene edited cells have a similar engraftment and growth rate when compared to unedited hepatocytes. This indicates that the edited cells are healthy and function normally. Despite the publication of high profile papers that suggest that gene editing will produce so many genetic alterations that the cells will not function properly, we demonstrate here that the cells not only function, but they are able to regenerate an organ and form a complete humanised liver, and that the regenerated organ functions properly.
In WP4, the HUMAN Consortium was able to characterise the metabolic and behavioural phenotype of humanised mice being produced within the consortium. In particular, we have analysed our model in regard to the species-related differences between humans and mice, and to the metabolic differences originating from the xenograft with hepatocytes from different human donors. In both cases, we have acquired an amount of compelling data enriching our knowledge on hepatic physiology and metabolism. New insight on human lipoprotein and bile acid metabolism were gained. In the last part of HUMAN, our attention shifted to characterization of intrahepatic lipid metabolism in lever humanised mice. We characterized in details the APOB-LDLR-PCSK9 axis in liver-humanised mice, and some insights into their lipid metabolism. HUMAN has also assessed the lipoprotein phenotype of liver-humanised mice transplanted with protective or risk alleles for TCF7L2, detecting differences in plasma cholesterol levels.
We have also highlighted the role of glucocorticoid signalling in the glucose phenotype displayed by humanised animals. Our results demonstrate that humanization of hepatocyte leads to change in circadian period of the animal which translate into advanced feeding, respiratory quotient, activity with a 2hr advanced maximum compared to murinised controls. We also demonstrate that human hepatocyte impose their circadian rhythm to the clock gene of the neighbouring rodent hepatocyte in within the humanised liver. We have further confirmed that liver humanization leads to change in circadian behaviour in mice, but additionally we have attempted to modulate the shift induced by humanization through direct force-feeding manipulation. We have uncovered that liver humanization translate in altered rhythmic expression in the clock gene in muscle
HUMAN has also been able to validate liver-humanised mice as a more reliable in vivo platform to study human LXR pharmacodynamics and to predict liver-related negative outcomes of the clinical trials that have been conducted in human.
In WP5 several experiments using in vivo microdialysis in control FRG-NOD (FRG KO) and liver-humanized FRG-NOD (Hu-FRG KO) mice were performed. We wanted to evaluate whether the extracellular and tissue levels of monoaminergic neurotransmitters (i.e. DA, NA and 5-HT) in the medial prefrontal cortex (mPFC) of liver-humanized mice could be dysregulated, particularly when considering the limited supply of L-Tyr and its restricted metabolism in FRG mice. HUMAN was able to demonstrate that repopulation of FRG mice with human hepatocytes did not cause any major differences in basal and pharmacologically stimulated monoamine levels in the mPFC. The only exception was NA. Humanization of liver in FRG KO mice lead to a higher sensitivity to the effect of duloxetine resulting in higher NA levels. This interesting finding suggests that the crass-talk between liver and brain may be not equal for all the different pathways affecting the levels of neurotransmitters.
It has been demonstrated that a lack of coordination between nutrient intake and biological clock contribute to the development of the metabolic syndrome and that the central biological clock can be reset by liver inputs. Alteration of feeding schedule or hyperphagia directed to palatable food could reflect stress/anxiety phenotype associated with a change in liver activity.
In HUMAN, behaviour and locomotion patterns of liver humanized, of liver murinised mice, and of wild type C57BL mice were studied. The control wild type C57BL mice preferred the dark box during a 120-second trial. This behaviour was attenuated in the FRG KO mice and completely abolished in the mice injected with the neurotoxin MPTP and the Hu-FRG KO mice. The FRG KO and the Hu-FRG KO mice showed an analogous pattern of locomotor activity as did the naïve C57BL mice, the total distance of forward movement over the 10-min period was not different between these groups. Metabolic and behavioural analysis revealed a striking phenotype in liver humanized animals. Feeding behaviour and locomotor activity displayed a prolongation of 2 hours, which in addition to the 22hr30min (rodent circadian period) result in a circadian period similar to the human one (24 hours). This ground-breaking observation suggests that the liver has the ability to influence the central clock and that human hepatocytes drive a 24hr-humaized circadian rhythm. Furthermore, changes in feeding pattern were also associated to altered expression of hypothalamic orexigenic neuropeptide. Exposure to dark further confirmed that liver-born inputs were sufficient to drive endogenous clock. These data reveals an important new mechanism by which liver-born inputs, presumably direct to the brain could act as a dominant signals in the control of feeding behaviour and metabolism.
HUMAN activities also focused on the WSB/EiJ mouse model because it is long-lived, resistant to obesity induced by dietary fats, and depict low circulating levels of thyroid hormones. Since transcriptomic analysis suggests that hypothalamic T3 content of these mice could be increased, we used the WSB/EiJ mouse as model to set up proteomics studies. Different studies using mice at different age (i.e. 3 month and 18 month old) challenged with 3 different dietary regimens (i.e. control diet, high fat diet for 3 days, high fat diet for 8 weeks) were performed . We could confirm that differences in metabolic adaptability and in longevity observed in mice with different T3 status could be the results of differential regulations of proteins involved in different processes, especially those which are known to be regulated by T3. These results have identified several critical pathways that are potentially involved in the aging process, as telomere regulation, mitochondria and lipid metabolism. Moreover, our results on the link between central D2 expression/activity and BAT thermogenesis suggest that central sensitivity to thyroid hormone could be crucial for an effective peripheral energy expenditure, and thus, for the homeostasis of the energy balance. Transcriptomic and proteomics analyses showed that WSB/EiJ mice (hypermetabolic phenotype despite lower circulating TH, higher longevity) have a hypothalamic gene expression pattern different from the one of C57Bl/6J mice (sensitive to HFD induced obesity), especially in the nuclei involved in the central control of food intake and energy expenditure. The pathways implicated in this specific pattern included the mitochondria and obesity pathways. These results confirm that body energy balance is centrally controlled by specific gene networks, which interact to drive energy expenditure and fat storage. Also, proteomics and transcriptomics data analyses from mouse model of increased central sensitivity to TH emphasise the importance of mitochondrial pathways in the central and peripheral control of metabolism, favouring healthy aging. In particular, long-lived mice present less detoxification enzymes and molecules controlling oxidative stress than normal lived mice, which can suggest that high need of detoxification can exhaust the organism and be deleterious for lifespan, as is a highly challenged metabolism.
A fundamental pillar of HUMAN was the development of robust methods and standard operating procedures (SOPs) for metabolic profiling of multiple biological matrices including plasma, tissues (brain, muscle, liver, kidney, pancreas and gut), cell media, cells (fibroblasts and iPSCs), cerebrospinal fluid, brain microdialysates, faeces and urine. The developed methods were also applied for metabolic profiling of changes in cell culture media that take place during differentiation of human iPSCs into hepatocyte-like cells. Consumption of nutrients from the cell media, as well as secretion of metabolites in to the cell media were monitored in order to identify the biochemical pathways important for cell growth. This resulted in the optimization of the protocols for iPSC cultivation and differentiation.
The effect of liver humanization on the metabolic profiles of mice brain and plasma was investigated by comparing metabolic profiles from brain tissues (cerebellum, cortex, hippocampus, hypothalamus and striatum) of humanized mice and non-humanized mice. All types of investigated tissues from humanised mice showed distinct differences in the metabolic profiles when compared to those observed in the control mice, which proved that humanisation of the liver had a profound effect on the overall metabolic profile of brain and plasma. Several metabolic pathways showing restoration of the liver function in humanized mice after cell transplantation were also detected, confirming successful humanization and hence validating both the applied humanization procedure and metabolomics as a reliable approach for analysis of samples from humanized mice.
In order to understand the cross-talk between human hepatocytes and other organs in the humanized mouse model, a metabolomics approach was applied to investigate differences in metabolic profiles of tissues from humanized and murinised mice (kidney, pancreas, liver and plasma). In all tissues clear differences between both types of mice were observed, mainly connected to the changes in lipid metabolism. Effect of the genotype of cells used for humanization (with protective or risk allele for type 2 diabetes) was seen only in liver samples and was characterized by lower levels of sugar phosphates and higher levels of the majority of fatty acids and cholesterol in the livers humanized with hepatocytes with the risk allele. This finding has to be put further into the biochemical perspective. Integration of metabolomics data from studies on humanized mice with the data produced by other partners of the HUMAN Consortium, will enhance our understanding of the humanized mice as a model for studying human diseases and its application in development of novel treatments for these.
The metabolomics approach was also applied to pinpoint metabolic pathways changed in age-related diseases, pathways connected to healthy aging and influence of genes on these. Human samples from two cohorts were analysed – from the Leiden Longevity Study and the Bologna Healthy Aging and Longevity Study. It was possible to pinpoint metabolic pathways changed in such diseases as type 2 diabetes, metabolic syndrome, cirrhosis and those changed in centenarians and semi-supercentenarians as compared to control groups. Metabolites correlated with early insulin resistance in the non-diabetic population were also detected, adding additional knowledge to our understanding of the pathogenesis of type 2 diabetes and being potential supportive biomarkers for early detection of insulin resistance. Metabolic profiles connected to a TCF7L2 polymorphism in both studied cohorts pinpointed several metabolites that connected the TCF7L2 mutation with inflammation and with action of gut microbiota. In summary, the majority of the identified changes that were correlated with age-related diseases, healthy aging and influence of genes on these were connected to inflammation, energy and lipid metabolism as well as with action of microbiota. Based on these results lipid-related and inflammation-related pathways, together with modulation of microbiota activity, were identified as potential targets for treatment or support of treatment of age-related diseases. Obtained results suggested also that treatment effect could be at least partly achieved by changes in the dietary habits of the sick individuals.
HUMAN extensively characterized humanized mice with a multi-omic approach forcing us to optimize the methods to study the epigenomic (including DNA methylation and histone marks enrichment profiles), transcriptomic and proteomic profile in a condition in which human and mouse molecules co-exists and should be specifically detected. Hence, innovative workflows were proposed for sample handling and human-murine data disambiguation in order to achieve human-specific, multi-omic data analysis and integration. HUMAN has thus developed a novel method for rapid assessment of degree of humanization in mice. The method relies on the ability to distinguish and quantitated human and murine proteins (namely Albumin, Transferrin and Apolipoprotein-E) in blood thus giving a quantitative humanisation percentage of the mouse.
In addition, proteomic and post-translational modification (PTM) analysis in long-lived and short-lived mouse, showed that the main differences was a lower PTM levels (specifically carbonylation, methylation and dimethylation) in the long-living mouse livers compared to the short-living. Remarkably, high fat diet did not produce any significant changes.
Innovative NGS-based methodologies has been applied for the concomitant analysis of epigenomic and transcriptomic profiles of humanised mouse livers carrying TCF7L2 genetic variants. All library-preparation protocols (either RNA-seq, ChIP-seq or RRBS) relied on a LIMS-supported platform and were automatized in order to reduce manual error and increase reproducibility. Data analysis was performed with optimised and fully automated workflows comprising five major nodes: i) raw data processing; ii) mapping and disambiguation; iii) quantification; iv) quality control; and v) statistical analysis. NGS data determined the percentage of liver humanisation. Since humanisation levels proved to be highly conserved (with an estimate deviation from true value of 3%) among different NGS data types, HUMAN was thus able to generate robust protocols of the high quality.
The multi-omic strategy highlighted variations in molecular profiles in humanised mouse models associated with humanisation, donor genetic background, diet and hormone treatment.
The epigenetic and transcriptomic modifications observed in mouse with humanized livers versus mouse with livers containing only mouse hepatocytes showed significant enrichments in genes encoding proteins belonging to metabolic pathways (e.g. bile acid metabolism, xenobiotic metabolism, amino acid metabolism, lipid and lipoprotein metabolism).
Multi-omic modifications were assessed also between humanised mouse models either carrying TCF7L2 rs7903146 protective or risk allele. In particular, RNA-seq analysis highlighted changes in the expression of genes associated with cell cycle (e.g. regulation of cell replication and chromosomal segregation); extracellular matrix organization, metabolic pathways (e.g. collagen and steroid hormone biosynthesis, biotransformation of xenobiotics) and immunological mechanisms (e.g. interferon signalling, cytokine secretion, ROS response). Interestingly, changes in TCF7L2 expression were not highlighted. RRBS analysis revealed enrichment of differentially-methylated regions (DMRs) on genomic regions associated with cell death mechanisms (e.g. death receptor and TNFR1-induced signalling as well as regulation of caspase 8 and c-FLIP activity).
ChIP-seq analysis identified differential, H3K27/H3K4 enrichment. Genotype-associated changes in H3K27/H3K4 enrichment were found at loci linked to metabolic processes (e.g. biotransformation of xenobiotics, metabolism of lipids, bile acid, fatty acids, etc.). In particular, changes in H3K4me1 enrichment were found associated with vesicle trafficking and immunologic process mediated by cytokines and chemokines (e.g. inflammatory response); while changes in H3K27ac enrichment were mainly found in association with cellular organisation and gene regulation loci (e.g. cytoskeleton organisation, meiosis and chromatin organisation). Interestingly, no significant enrichment at TCF7L2 nor TCF4 binding sites was found. Cross-omic comparison of NGS data indicated several changes in biologic processes including extracellular matrix organisation, vesicles trafficking, xenobiotics metabolism and steroid hormone biosynthesis. Among the most prominent genes affected by TCF7L2 genotype both at transcriptomic and epigenomic levels are TRIM31, ADGRE5 and PEG which are associated with metabolic diseases and longevity. Another set of differentially regulated genes comprise NAAA, EPB41L4A, GATA3 and BIRC5 which are implicated in Wnt signalling pathway (TCF7L2 belongs to Wnt signalling pathway).
Human was able to elaborate the concept that TCF7L2 variants ectopically affect epigenomic and transcriptomic profiles of humanised mouse livers. Indeed, whereas no changes in TCF7L2 expression nor TCF7L2 promoter histone enrichment were found, changes at multiple omic levels were reported for a variety of molecular pathways including Wnt signalling. The interpretation of these data has to include the possibility that the expression of TCF7L2 mRNA and protein in humanized livers is likely to be dynamic in living mouse. Thus, the analysed mRNA levels at the endpoint of the study (upon sacrificing the mouse) does not necessarily reflect the expression dynamics of TCF7L2 protein in vivo.
In order to manage all the volume of data generated and the complexity of the project, HUMAN developed a central infrastructure for data storage, management, and sharing. The quality of the raw ‘omics data has been assessed using standardised and validated high-performance workflows. Data pre-processing and quality control, including report generation, has been performed for array- and sequencing-based as well as mass spectrometry-based data. Phenotypic data generated from the diverse models has been integrated and analysed together with the ‘omics data, to identify: i) molecular events associated with phenotypic and disease endpoints, and ii) phenotypic readouts associated with disease endpoints. New bioinformatics, biostatistics, and visualisation tools have been developed. Methods have been implemented specifically to address the challenges of analysing data from humanised mouse models, e.g. to discriminate NGS reads originating from human vs. mouse cells.
Infrastructure
A central, structured FTP server with access control and user management has been established, and storage capacity has been expanded to 6 TB by the end of the project.
To enable data harmonization necessary for integrated analysis, a sample naming and annotation convention has been established, and sample naming templates following this convention have been provided to all project partners.
Bioinformatics pipelines
Several enhancements and new functionalities needed for the HUMAN project have been implemented into the Genedata Profiler® software. For the analysis of ChIP-Seq data, QC metrics, algorithms and statistical tools have been developed and an automatable data processing workflow has been set up. For the analysis of RRBS data from xenograft samples new algorithms have been adopted and an automatable SOP data processing workflow for the detection of differentially methylated regions has been implemented. Specific workflows for the processing and statistical analysis of transcriptomics and epigenomics data created with the platforms NanoString nCounter and Illumina Infinium MethylationEPIC BeadChip have been developed. New algorithms and statistical tools for the analysis of RNA-Seq data have been incorporated into the Genedata Profiler® software and two automatable data processing workflows have been configured for the detection of differential expression and differential splicing, respectively.
Data Analysis
Partner UMIL has been supported in the establishment of an optimized ChIP-Seq protocol by processing and comprehensive quality assessment of ChIP-Seq pilot data. Proteomics data from partners CNRS and UCL obtained from seven organs of two mouse strains under different dietary conditions has been quality controlled, processed and analysed. Data from the UPD RRBS study has been analysed using a refined SOP workflow for read mapping, disambiguation and assessment of humanisation levels of mouse liver samples and integrated with sample-matched metabolic data. A workflow for processing, humanisation assessment and statistical analysis of data from the KI3 RNA-seq study has been set up and executed, which has been reported as part of deliverable D8.3 (Transcriptome signatures of humanised livers). Data from the UMIL ChIP-seq study has been processed and analysed, including quality control and humanisation assessment of the grafted mouse liver samples as well as the systematic validation and functional annotation of the detected H3K27ac peaks. In a collaboration with partner NIHS, machine learning-based models for the prediction of iPSC clone differentiation efficiency based on molecular signatures detected from sample-matched, multi-omics data have been constructed and validated. Partner CNRS has been supported with statistical and network analysis of qPCR data from mouse brain samples to characterise the gene expression signatures associated with strain, age or diet effects in relevant hypothalamic nuclei. Humanised mouse liver models have been characterised by genome-wide histone mark profiles based on H3K4me3/H3K27ac ChIP-seq data, and systematic validation, cross-study comparison, functional annotation and motif analysis of differential peaks has been performed. Extensive cross-omics analyses have been performed through statistical comparison and pathway-based, functional characterization of disease-derived vs. control humanised mouse liver models by integrated analysis of gene expression (RNA-seq), histone mark (ChIP-seq) and DNA methylation (RRBS) profiles.
Potential Impact:
The research activities and the results generated by the HUMAN consortium have a great impact on different levels in the study of longevity, development of metabolic syndrome, on the physiological regulation of metabolisms, and on the cross talk between the brain and gut.
Regarding the SNPs studied, HUMAN was able to confirm the importance of the genetic variances of TCF7L2 and APOE confirm the role of these genes in cardiometabolic health and disease. HUMAN put forward the hypothesis that omega-3 fatty acid intake and physical activity may modify the impact of ApoE ɛ4 on disease risk. This hypothesis may impact a public health perspective.
The collaboration of HUMAN with the international CHARGE consortium, resulted in multiple novel hypotheses and candidate mechanisms linking sleep habits to SNPs to be further explored in future studies. Further, we identified novel mechanisms that could potentially be of help in the development of novel interventions for the prevention of cardiometabolic disease in patients with sleep disorders.
Epidemiological research has often an observational nature which limits the potential of causal inference. Although a randomized controlled trial is considered the golden standard to infer causality of observational associations between a certain exposure and outcome, it is often not feasible. Mendelian Randomization is an increasingly used method to infer causality using genetic variants for the exposure as instrumental variables. In HUMAN, we conducted several Mendelian Randomization studies to infer causal associations of phenotypes of interest to the consortium, and the results of these are od potential clinical importance. The inhibition of Cholesteryl Transfer Protein (CETP) has been long thought to be a therapeutic target for the prevention of cardiovascular disease. Our finding that genetically-determined CETP concentrations are not causally associated with risk of cardiovascular disease, are pivotal to understand the failed trails using CETP inhibitor drugs.
The activity of HUMAN on the generation of iPSCs, their genetic modification with gene-editing technologies and the differentiation towards mature hepatocytes have a tremendous direct and future innovative impact. HUMAN can and has provided the international community with robust and validated protocols and working pathways on how select cell to be differentiated in iPSC, how to grows the latter and genetically modify them with success. Further HUMAN has defined an advanced and accurate way how evaluate the differentiation of iPCS cells into mature human hepatocytes. The successful gene-editing protocols established by HUMAN which has led to the recreation of model of human disease, and of their correction in mice are milestone for other for the development of similar model targeting other pathologies.
All the animal model created, with humanised liver or humanised pancreas, has been turned out to be an example for the development of personalized models that allow the study of the individual subject or the test of new therapeutic approaches, either based on gene editing or based on small molecules. Hence, the research generated by HUMAN is an example of science that is “translatable” by to human condition, and not only translational to it. A significant contribution to the need for generation of animal models that recreated the human condition is the prove that the humanised liver mice, were able to reproduce the negative outcome of the first clinical trials in human in which a nuclear receptor was targeted in the liver.
The work of HUMAN may thus positively shorten the time between discovery and implementation into clinical care, by showing a way how already in a pre-clinical stage predict the outcomes of phase 2 clinical trials.
The discovery of HUMAN on lipoprotein and bile acid metabolism and how human liver affects the macrobiota may positively influence the future development of strategy aiming to prevent and cure cardiometabolic diseases.
The identification, by the integration of proteomics, trancriptomics and metabolomics data sets from long-lived, obesity resistant mice, of several pathways which could contribute to the healthy aging phenotype. These pathways, when confronted with proteomics analysis from “supercentenarian” human samples, were found to be similar in some extents to pathways identified in these human samples. Thus, a potential impact of the work carried out by HUMAN could be the targeting of these pathways in a therapeutic perspective, to achieve healthy aging in human. This important perspective will be refined by an integrated analysis of all the “omics” datasets obtained in mice, and confronted to human datasets, in order to provide strong results showing the most important pathways to target. The result of this big data integrative analysis, comparing mice and human data, will be made available to the international scientific community.
The applied metabolomics approach of HUMAN improved the understanding of biochemical processes disturbed in pancreatic and metabolic diseases. Metabolic pathways that were found to be affected in the studied diseases are potential targets for novel treatments which, together with proper dietary interventions, would in long run contribute to healthier aging of European citizens.
HUMAN was able to identify multi-omic molecular signatures for a thorough characterisation of the humanised mouse model. The set-up and optimisation of a series of high-throughput, NGS-based techniques able to deal with the difficulties associated with the use of animal models in which human and murine genomes are present contextually have a great impact of the future reaserch in which chimeric models are used. Multi-omic data (either proteomic, transcriptomic or epigenomic) generated by the HUMAN Consortium allow for highly-reproducible and robust assessment of liver humanisation in the humanised mouse models and the definition of workflows enabling automated processing of gene expression, DNA methylation and histone-mark analyses reduce can be utilized in the studies of many other diseases. Furthermore, the automated analysis pipeline allows for rapid GO and pathways annotation of hits identified in different NGS-data. Lastly, the proposed NGS data analysis pipeline include cross-omic comparison thereby allowing identification of biological processes affected at a multiple omic level.
The results achieved from the analysis of humanised mouse livers highlighted human genotype-associated changes in the function of the human hepatocytes with particular involvement of metabolic and cellular organisation processes. These results represent the initial step towards the discovery of potentially new regulatory pathways in physiology and in disease onset and progression.
List of Websites:
www.fp7human.eu
The prevalence of metabolic diseases has dramatically risen owing to the great increase of life expectancy and cultural transitions in lifestyle and nutrition, with a predicted massive economic burden for the European health systems. In the last 15 years, great research efforts have been devoted to identify the genetic basis of age related and metabolic diseases, mostly by means of large genome-wide association studies (GWAS). GWAS studies have reported associations of thousands of single nucleotide polymorphisms (SNPs) for hundreds of traits/diseases. Despite the highly significant association with the trait of interest, the functional role of all these genetic variants need to be elucidated especially in term of genotype/phenotype correlation. Otherwise, the potential of the identified SNPs for clinical and pharmaceutical application for i.e. the treatment and prevention of metabolic diseases is dramatically reduced. To overcome these limitations one of the main goals of the HUMAN consortium was the generation of mouse models highly repopulated with human hepatocytes or carrying pancreatic beta-cells from either primary cells (hepatocytes) or induced pluripotent stem cells (iPSCs). This innovative approach offers the unique possibility of studying the physiological and pathophysiological processes associated to metabolic diseases in human-derived organs, but placed in an integrated living system (i.e the mouse body). A unique strength of the project is that humanised mice are a model for a translational research that is translatable to the human condition. iPSCs used to generate hepatocytes and beta-cells derives from patients affected by severe metabolic diseases such as type 2 diabetes (T2D) with or without complications, or subjects selected for nearly the complete opposite phenotype, for their nearly complete lack of disease and exceptional longevity. In the end of the HUMAN consortium, we have focused on the most relevant genetic variant (FCF7L2) amongst those that emerged from GWAS studies for its association with T2D and metabolic syndrome. Mice with humanised liver and/or pancreatic beta-cells have been generated, and the phenotype that results from the risk or protective alleles characterised. Since complex diseases such as T2D and metabolic syndrome are polygenetic and partially dependent on the genes selected here, the creation of the risk and protective cell lines and mouse models provides a platform technology for discovery of genes and metabolic pathways that cooperate with these known risk alleles to manifest the full spectrum of the disease. Furthermore, studies with drugs targeting nuclear receptors (i.e. Liver x Receptor; LXR) demonstrated the validity of the liver humanised mouse model to predict the possible negative outcomes observed in phase II human studies and thus allowing an early evaluation of the translatability to human of pre-clinical drug development. The main goal of the project is to offer European academia and and industries exclusive tools to tackle the challenge of functional validation of metabolic disease-associated genetic variants by offering innovative and robust humanised animal and cellular models and a portfolio of new and validated therapeutic targets for better understanding of metabolic diseases and healthy aging.
Project Context and Objectives:
WP1
The WP1 objective is to assess the role and the importance of selected genes/alleles/phenotypes in metabolism and longevity, by (i) selection of human donors for the generation of humanised mouse models, (ii) assessment of the role of selected genes/alleles/phenotypes in metabolism and longevity in human studies and (iii) comparison of results from human studies and those obtained from humanised mouse models.
The strategy for selection comprised the presence of risk and/or protective alleles for SNPs of interest in the TCF7L2, APOE, and FTO genes and in combination with metabolic phenotypes (related to glucose and lipid metabolism) and/or clinical phenotypes (covering the spectrum from healthy longevity and absence of disease towards cardiometabolic disease with complications). The activity in this work package were instrumental for WP2 and WP3 by delivering fibroblasts to be used for production of iPSC or by delivering hepatocytes to directly produce liver humanised mice.
WP2
The objective of WP2 is the generation and differentiation of induced pluripotent stem cells (iPSC) to hepatocytes and pancreatic β-cells. WP2 is closely linked to WP3 “generation of humanised liver and pancreas mouse models”.
In this work package, two separate groups worked to improve the ability to make liver cells (hepatocytes) from stem cells (iPS cells). The work has resulted in many scientific publications and more will appear. Compared to 5 years ago when the grant was written, the liver cells made today from stem cells are much closer to authentic liver cells, and they provide many mature liver functions. In addition to making better liver cells, we have provided a method and a database by which we and any other laboratory can compare their stem cell-derived liver product to authentic human fetal or adult liver samples. In this manner, one can easily determine the state of differentiation of the experimental group to the values found in real fetal or adult human liver. This database and method will be made public for general use. Many laboratories do not have access to authentic human liver samples like we do, so this public data base will be of great help to researchers all over the world.
WP3
The objective of this WP is the generation of humanised liver and/or pancreatic mouse models. Mice will be made from: 1) Transplanted hepatocytes from normal donors; 2) Transplanted hepatocytes from metabolic liver diseases; 3) Transplanted hepatocytes from specifically selected genotypes; 4) pancreas humanised mice where cell encapsulation device technology is used to transplant pancreatic progenitors into immune-compromised rodents and allow maturation towards fully differentiated β-cells in vivo; 5) mice with both liver and pancreas humanised mice.
In this work package, we created interesting and useful models of serious human diseases. Many mouse models do not faithfully reflect the human disease. To make the models more human relevant we have generated mice that are highly repopulated with human liver cells (hepatocytes). The liver cells in these liver-humanised mice perform their normal human liver functions and are human-relevant models for blood lipids such as cholesterol, the production of bile acids, the regulation of glucose levels, drug and ammonia metabolism. Liver–humanised mice were also created to model serious defects in liver function such as urea cycle deficiencies, which inhibit the patients ability to eat normal protein and excrete the excess nitrogen that is generated in that process. Liver humanised mice were created for two life-threatening liver-based diseases that affected the urea cycle. In both cases, the mouse model produced the symptoms characteristic of the human disease. These mouse models have proven useful to examine new drug, cell and gene editing therapies and will provide data that can be extrapolated to human patients since the target cells for the disease and the therapy are actual human liver cells.
WP4
The aim of this WP is to fully characterise the metabolic and behavioural phenotype of humanised mice by using a standardized approach and with attention to how the genetic variations affect metabolism towards a greater susceptibility to metabolic disorders.
The research groups working in this working package were in close contact with the research groups acting in WP1, WP2, and WP3. The characterization of the animals generated included analysis of the feeding habits, of how energy was produced
and whether more or less sugars were burned in comparison to fat. In addition, the characterization of this animal included the analysis of the changes in lipid in circulation, in the liver and in other tissue. Further, the liver humanised mice were also utilized for study aiming to demonstrate that this animals model gives result that are translatable to human condition when new drugs are developed. The different experiment performed also generated biological material that have been further analyzed in the WP5, WP7 and WP8.
WP5
The WP5 objective is to understand how the brain-liver–brain axes contribute to the phenotypic traits observed in WP4.
The hypothalamus integrates information on metabolism from hormonal and vagal nerve inputs. In turn, the brain provides adaptive responses both at the behavioural and metabolic levels, through endocrine and autonomic controls of peripheral organ activity notably liver, pancreas, muscle, adipose tissue and brown fat. Thyroid signalling is a prototypical endocrine candidate for controlling metabolism by liver /brain communication. The liver is a primary source of production of T3. Changes in thyroid status will in turn modify a number of metabolic pathways both directly in the periphery and indirectly through actions in the hypothalamus and the autonomic nervous system, e.g. central levels of T3 modulate glucose production by the liver through the sympathetic NS. Available wild type and mutant mouse models were used as tools to get more insights on the metabolomics and transcriptomic hypothalamic networks underlying central responses to either increased circulating T3 or increased intra-hypothalamic T3.
WP6
The overall task was to genetically convert different SNP/mutations in induced pluripotent stem cells (iPSCs) derived from fibroblasts using the CRISPR technology and screen the corrected clones for off target modifications. Additionally, the corrected clones with low off target frequency was selected for differentiation into hepatocytes and further transplanted into FRGN mice. Finally, the humanised mice were screened for normalised phenotype by the genetic correction.
In WP6 we greatly exceeded the studies initially proposed in the application. Building on the liver-humanised mice studies in WP3 we were able to create mice with serious human liver diseases that affect the metabolism of proteins and ammonia, called urea cycle defects. These diseases are so serious that the liver cells used for these studies were isolated from the liver of pediatric patients that had to receive a liver transplant to save their life. The diseases cells were isolated from the patient liver after it was removed for the transplant. As indicated in WP3, when we repopulated the mouse liver to high levels with the patient’s mutant cells, the mice displayed the same symptoms of the urea cycle defect as the donor patient. In separate studies, we were able to use state-of-the–art gene editing procedures to correct the mutation in the patients liver cells. When those “corrected” cells were transplanted into the mice, the mice no longer displayed the symptoms of the disease. Thus, we were able to demonstrate that it is possible genetically correct mutation in a patient own liver cells and if those cells are transplanted, the symptoms of the disease are corrected. Through WP 3 and WP6 we were able to firmly demonstrate that the genotype of the hepatocytes transplanted into the mice determines the phenotype of the mouse that is generated. This new data is highly relevant information to help move the gene editing therapy to the clinic. Given the abundance of research into gene editing technology and the importance of the human genome sequence and the human DNA replication and repair pathways to these processes, it would be advantageous to perform preclinical studies of gene editing technologies with human hepatocytes, and when possible, with the actual disease-affected human hepatocytes. The value of the data obtained from such studies will be greatly enhanced if those studies could be conducted, in vivo. Additionally, studies with live animals that can be followed after gene editing or other therapies for long periods of time will enable the assessment of long-term outcomes such as phenotypic correction of the disease and safety studies. The model reported here will be expected to be useful for investigations of modified RNA, gene, cellular and small molecule therapies for genetic-based liver diseases. Perhaps where these liver-humanised models will be most informative will be in the investigations of the safety and efficacy of gene editing technologies and their translation to the clinic.
WP7
The WP7 objective is to use metabolomics as tool for a comprehensive characterization of the biological specimens derived by the different models generated within the HUMAN Consortium. The major goal is to deliver systematic molecular profiles and to identify underlying biochemical pathways altered in metabolic diseases, correlated with disease-linked genotypes, or connected to longevity in order to improve understanding of healthy aging, disease aetiology and for suggestion of new treatment strategies.
WP7 has worked closely with WP1, WP2 and WP3 so that quality all through the different stages of the research path flow chain from donors, via cell cultivation to the humanised mice, was assured by analysis of the biochemical pathways. WP7 results provided WP8 with candidate proteins for targeted proteomics in order to validate the suggested biochemical pathways. The biochemical pathway analysis has also been instrumental to provide suggestions for possible new therapeutic strategies. The work in WP7 has been conducted in close collaboration with WP9 for data integration and interpretation. The working hypothesis of WP7 was that complex gene expression networks coordinate the metabolic fingerprint of the disease and that their function can be observed in the endogenous metabolic profile.
WP8
The WP8 objective is to characterize the molecular characteristics of mouse with humanised livers by a comprehensive analysis of the modification of DNA, the expression of the genes and of proteins (i.e. epigenomic, transcriptomic and proteomic) using advanced state-of-the-art technologies (i.e. high-throughput ‘Omics’ technologies).
The activity of WP8 were fundamental for the characterization at molecular level of the humanised mouse models by linking molecular alterations to the physiological phenotypes observed in the experiment conducted in WP4 and to the metabolic profiles obtained in WP7. As for WP7, the quality controlled data obtained in WP8 have been transferred to WP9 for integrative analysis and identification of affected pathways, drug targets and biomarkers.
WP9
The WP9 objective was the provision of a central database infrastructure, which enabled structured management and secure access to the raw/processed multi-omics data and associated phenotypic data generated within the HUMAN consortium, and the generation of high-quality processed data.
The data transferred to WP9 have been assessed for quality by using standardised and validated high-performance workflows. Data pre-processing and quality control, including report generation, have been done according to validated workflows that have been developed by a research partner of the Human consortium. The results have been generated to ensure an efficient high throughput and standardized analysis and to meet the project’s specific requirements for the latest ‘omics technologies. Recommendations for re-processing or selective exclusion of raw data from further analyses has been provided to project partners when necessary. WP9 has performed an integrated, multivariate analysis of all ‘omics data types to establish statistically significant associations and their relevance determined by elucidation of cross-platform evidence for regulatory and metabolic networks.
Project Results:
For the generation of humanised mice, human cells were from two different sources, namely (i) primary human hepatocytes for direct transplantation in mice, and (ii) human fibroblasts for pluripotent stem cells (iPSC) generation and subsequent differentiation into mature hepatocytes and beta cells prior to transplantation in mice.
The HUMAN consortium had access to several large collections of human primary hepatocytes that upon genotyping were shown to exhibit a wide distribution of protective and risk alleles for TCF7L2, APOE, and FTO. Hepatocytes from donors with protective alleles for all selected SNPs in TCF7L2, APOE, MC4R, and FTO as well as hepatocytes from donors with risk alleles in TCF7L2, APOE, and/or FTO were selected for later transplantation in mice.
The researchers working in the WP1 had also access to unique collections of cryopreserved fibroblasts from cohorts of semi-supercentenarians and the Leiden Longevity Study. After Ethical approval, selected fibroblasts were further characterized genetically and epigenetically and used for the generation of iPSC lines. Genetic characterization revealed that the mTOR coding sequence is extremely conserved, indicating that this gene is likely involved in vital physiological functions and barely tolerates amino acidic variability. In a novel study for which ethical approval was obtained in first year of life of the HUMAN Consortium, fibroblasts were collected and cryopreserved from biopsies from 24 newly recruited patients with type 2 diabetes mellitus, and selected based on presence of risk alleles and clinical diabetic complications.
In first years of HUMAN, the main focus was on the selection of human fibroblasts based on a the presence of risk and/or protective alleles for selected SNPs in TCF7L2, APOE, and FTO in combination with metabolic and/or clinical phenotypes. Due to technical challenges related to derivation of differentiated human hepatocyte by iPSC of fibroblasts, our focus later changed to selection of primary hepatocytes, based on the distribution of protective and risk alleles for TCF7L2, FTO or APOE risk alleles for transplantation into mice and the characterization of the mice thus obtained.
To further elucidate the role of selected genes/alleles/phenotypes in human metabolism and longevity, an important focus of WP1 has been the integrated analysis of genotype/phenotype correlations in different human donors/cohorts. These analyses were performed around three main themes: i) genotype-phenotype associations; ii) sleep and cardiometabolic disease; and iii) causal inference.
Genotype-phenotype associations.
Regarding APOE, we estimated in the Leiden Longevity Study that the contribution of the APOE isoforms to the previously identified phenotypic and metabolic differences between offspring and controls (e.g. diabetes mellitus prevalence and HDL cholesterol levels) was minimal. Based on literature, we put forward the hypothesis that increased physical activity and increased intake of omega-3 fatty acids are potential working mechanisms to diminish the detrimental effects of APOE ε4 on the risk of developing age-related diseases. Regarding TCF7L2, we observed in the Active and Healthy Aging Study, and the Leiden Longevity Study that carriers of the risk allele had higher nocturnal glucose levels, similar levels of glucose during the day, and that they displayed a comparable degree of variability in glucose levels. However, part of this association could be explained by adiposity, since adiposity was associated with higher nocturnal and diurnal glycaemia, but not with variability of glucose levels.
In the Netherlands Epidemiology of Obesity Study, we observed that genetic variation in TCF7L2 was mainly associated with insulin levels and with type 2 diabetes mellitus treated with insulin analogues, which suggests that TCF7L2 associated with a type 2 diabetes mellitus phenotype which is more critically dependent on insulin secretion rather than insulin resistance. Regarding FTO, we found no evidence for an association of the FTO risk allele rs1421085 with measures of energy metabolism in the Netherlands Epidemiology of Obesity Study.
Sleep and cardiometabolic disease
In the Netherlands Epidemiology of Obesity Study, we found that short as well as bad habitual sleep were associated with an increased body mass index. We did not find evidence that short or bad habitual sleep were specifically associated with a higher amount of visceral fat, as measured using magnetic resonance imaging. In the same study population, we additionally found that short as well as bad habitual sleep was associated with an increased insulin resistance and higher concentrations of triglycerides in blood and liver. In collaboration with the international CHARGE consortium, we initiated a sleep-SNP interaction study on serum lipid levels with the purpose to study the biological background of sleep-associated dyslipidemia. Using data from over 40 cohorts of different ancestries (N>125,000), multiple genetic variants mapped to genes not earlier described in relation were identified and replicated in an independent analysis.
Causal inference
In HUMAN, we conducted several Mendelian Randomization studies to infer causal associations of phenotypes of interest to the consortium. We were also part of an international collaboration to examine the causal association between (central) adiposity and age-related disease outcomes, which found that central adiposity and overall adiposity have a distinct cardiometabolic risk profile. In addition, we found, using publically available data from the DIAGRAM consortium, that high concentrations of the liver enzyme gamma-glutamyltransferase was not causally associated with an increased risk of type 2 diabetes mellitus, although observational research suggested such relationship. Furthermore, increased concentrations of thyroid stimulating hormone (TSH) have been associated with increased risk of developing type 2 diabetes mellitus. Since, this association was causal it is evident that common risk facors and life styles can lead to type 2 diabetes mellitus and disturbances of the thyroid gland. Increased low-grade systemic inflammation, however, was found to associate with an increased risk of type 2 diabetes mellitus using Mendelian Randomization, in line with observational studies. However, low-grade systemic inflammation is complicated and is dependent on several factors. Based on our results, is could be inferred that other inflammatory factors are causally associated with the risk of type 2 diabetes mellitus. Based on a newly performed genome-wide association study on circulating cholesterol ester transfer protein (CETP) , we identified 3 independent genetic variants mapped to the CETP gene that could be used in Mendelian Randomization studies. However, we did not observe a causal association between genetically-determined CETP concentrations and the risk of cardiovascular disease and found that CETP was causally associated with the concentrations of VLDL particles rather than LDL particles was previously thought.
In WP2 we proposed to make liver cells from stem cells. Now with the much-improved protocols we established, it is possible to make billions of human liver cells from stem cells with significant levels of liver function. One of the groups involved in WP2 was the group that conducted the first liver cell transplants in patients to correct liver disease, and has the largest group of patients treated by liver cell transplantation. Once issues of scalability of the processes and safety are met, the ability to generate liver cells from a seemingly endless source of stem cells, will enable us to expand liver cell therapy to treat liver disease to many more patients than are currently possible. This could have a huge impact on the health, welfare and quality of life for thousands of patients, and should help correct liver diseases that are now only possible to treat by liver transplants. Since iPS could be made from the patients themselves, or universal donor cell banks could be generated, the future liver cell transplants could be performed without suppressing the patient’s immune system with drugs. This program funding allowed us, for the first time, to begin gene-editing technology. Cells from many normal patients, centenarians and those with severe liver diseases were reprogrammed to iPSc. These stem cells were studied by themselves were coaxed, in culture to become liver cells and were also used as models for the different human diseases from which they originated. The stem cells were the first targets of the gene editing programs because they were abundantly available which allowed us sufficient cells for our initial trial and error experiments to improve gene-editing technology. These studies proved crucial to the later steps where actual human liver cells isolated from the patient’s own liver after it was removed during a liver transplant procedure were edited to correct the mutation in the patient’s liver cells.
It is clear that animal models of many diseases that affect human patients do not accurately reflect the disease observed in the patient. Of the over 20,000 genes in the human genome, more than 1/3 of them are expressed in the human liver. Mutations in many of these genes result in serious human live diseases. We started from the hypothesis was that the best model for a human metabolic liver disease that affects hepatocytes (liver cells), is one generated by actual disease-affected human hepatocytes. If that model could be investigated in vivo, the value of the model would be greatly enhanced. We created models where the liver of mice can become nearly completely replaced by human liver cells and showed that when normal liver cells are transplanted a normal mouse is generated, but when diseased human live cells are transplanted, the human disease is now recreated in the mouse. In WP 3, these liver humanised mice were generated and studies in great detail. The results proved that the physiology and functions of the human liver can be studied in these liver-humanised mice. This is a huge breakthrough for researchers because the modes accurately reflect the human disease and the studies will reduce the number of animals needed for research studies since the humanised models will provide data that is more relevant to human patients and many studies will be able to be conducted on live animals.
As all well-planned research programs should, success in one part feeds the research in multiple other parts of the project. Studies in WP2 reprogramming cells to stem cells allowed us to use the stem cells to begin our gene editing projects in WP6 and verify that the gene editing also changed the biology and the properties and functions of the liver cells made from the stem cells. Likewise, progress with gene editing of stem cells in WP6 and the generation of the liver-humanised mice conducted in WP3 fed the progress made in gene editing with actual liver cells as reported in WP3 and WP6.
For several genetic traits, we firmly established that the genes that are expressed (the genotype) in the human liver cells that are transplanted into the mice determines the characteristics (phenotype) observed in the liver-humanised animals. These observations are of critical importance for studying normal physiology and biology as well as human liver disease. Mice made with specific genetic defects in liver function, recreated the symptoms of the disease observed in human patients. Thus, we have verified that the liver-humanised mouse model is a faithful model and surrogate with which to study and correct these serious human diseases. Also, the diseased hepatocytes used for these studies were derived from the liver of patients that had to receive a liver transplant for life-threatening diseases. Thus, we have created models of serious human liver diseases in liver-humanised mice from the actual diseased liver cells removed from the patient. With these newer technologies like gene editing, the target of the editing procedure is critical to the outcome. Editing of a specific gene in mouse cells or mouse liver will tell us virtually nothing about the safety or efficiency of that procedure in human cells or human liver. Critical information for gene editing procedures are the DNA sequence being edited and the DNA replication and repair pathways in the cells themselves. Since these are much different between mouse and human cells, the mouse is not a sufficient model to test the safety or efficiency of the gene editing procedures. One cannot extrapolate gene-editing data from mice to human patients. It would be advantageous to perform preclinical studies of gene editing technologies with human hepatocytes, and when possible, with the actual disease-affected human hepatocytes. The value of the data obtained from such studies will be greatly enhanced if those studies could be conducted, in vivo. All of those aims have been met with the liver-humanised mice generated here. Additionally, studies with live animals that can be followed after gene-editing or other therapies for long periods of time will enable the assessment of long-term outcomes such as phenotypic correction of the disease and safety studies. The model reported here will be expected to be useful for investigations of modified RNA, gene, cellular and small molecule therapies for genetic-based liver diseases. Perhaps where these liver-humanised models will be most informative will be in the investigations of the safety and efficacy of gene editing technologies and their translation to the clinic to correct liver disease in patients.
The WP3 successfully created liver-humanised mice with specific genotypes, and the phenotype of the mice reflects the genotype of the human hepatocyte donor. Importantly, we can gene edit the human hepatocytes and the phenotype of the liver-humanised animals reflects the genotype of the donor human hepatocytes originally transplanted. As was shown in other work packages with the lipoprotein profiles, the bile acids, the blood glucose levels and here with the urea cycle activity, the phenotype of the liver-humanised mice faithfully reflects the genotype of the human hepatocyte donor. Concerning gene editing, HUMAN has demonstrated for the first time that hepatocytes can be gene edited, ex vivo, and that the gene edited cells have a similar engraftment and growth rate when compared to unedited hepatocytes. This indicates that the edited cells are healthy and function normally. Despite the publication of high profile papers that suggest that gene editing will produce so many genetic alterations that the cells will not function properly, we demonstrate here that the cells not only function, but they are able to regenerate an organ and form a complete humanised liver, and that the regenerated organ functions properly.
In WP4, the HUMAN Consortium was able to characterise the metabolic and behavioural phenotype of humanised mice being produced within the consortium. In particular, we have analysed our model in regard to the species-related differences between humans and mice, and to the metabolic differences originating from the xenograft with hepatocytes from different human donors. In both cases, we have acquired an amount of compelling data enriching our knowledge on hepatic physiology and metabolism. New insight on human lipoprotein and bile acid metabolism were gained. In the last part of HUMAN, our attention shifted to characterization of intrahepatic lipid metabolism in lever humanised mice. We characterized in details the APOB-LDLR-PCSK9 axis in liver-humanised mice, and some insights into their lipid metabolism. HUMAN has also assessed the lipoprotein phenotype of liver-humanised mice transplanted with protective or risk alleles for TCF7L2, detecting differences in plasma cholesterol levels.
We have also highlighted the role of glucocorticoid signalling in the glucose phenotype displayed by humanised animals. Our results demonstrate that humanization of hepatocyte leads to change in circadian period of the animal which translate into advanced feeding, respiratory quotient, activity with a 2hr advanced maximum compared to murinised controls. We also demonstrate that human hepatocyte impose their circadian rhythm to the clock gene of the neighbouring rodent hepatocyte in within the humanised liver. We have further confirmed that liver humanization leads to change in circadian behaviour in mice, but additionally we have attempted to modulate the shift induced by humanization through direct force-feeding manipulation. We have uncovered that liver humanization translate in altered rhythmic expression in the clock gene in muscle
HUMAN has also been able to validate liver-humanised mice as a more reliable in vivo platform to study human LXR pharmacodynamics and to predict liver-related negative outcomes of the clinical trials that have been conducted in human.
In WP5 several experiments using in vivo microdialysis in control FRG-NOD (FRG KO) and liver-humanized FRG-NOD (Hu-FRG KO) mice were performed. We wanted to evaluate whether the extracellular and tissue levels of monoaminergic neurotransmitters (i.e. DA, NA and 5-HT) in the medial prefrontal cortex (mPFC) of liver-humanized mice could be dysregulated, particularly when considering the limited supply of L-Tyr and its restricted metabolism in FRG mice. HUMAN was able to demonstrate that repopulation of FRG mice with human hepatocytes did not cause any major differences in basal and pharmacologically stimulated monoamine levels in the mPFC. The only exception was NA. Humanization of liver in FRG KO mice lead to a higher sensitivity to the effect of duloxetine resulting in higher NA levels. This interesting finding suggests that the crass-talk between liver and brain may be not equal for all the different pathways affecting the levels of neurotransmitters.
It has been demonstrated that a lack of coordination between nutrient intake and biological clock contribute to the development of the metabolic syndrome and that the central biological clock can be reset by liver inputs. Alteration of feeding schedule or hyperphagia directed to palatable food could reflect stress/anxiety phenotype associated with a change in liver activity.
In HUMAN, behaviour and locomotion patterns of liver humanized, of liver murinised mice, and of wild type C57BL mice were studied. The control wild type C57BL mice preferred the dark box during a 120-second trial. This behaviour was attenuated in the FRG KO mice and completely abolished in the mice injected with the neurotoxin MPTP and the Hu-FRG KO mice. The FRG KO and the Hu-FRG KO mice showed an analogous pattern of locomotor activity as did the naïve C57BL mice, the total distance of forward movement over the 10-min period was not different between these groups. Metabolic and behavioural analysis revealed a striking phenotype in liver humanized animals. Feeding behaviour and locomotor activity displayed a prolongation of 2 hours, which in addition to the 22hr30min (rodent circadian period) result in a circadian period similar to the human one (24 hours). This ground-breaking observation suggests that the liver has the ability to influence the central clock and that human hepatocytes drive a 24hr-humaized circadian rhythm. Furthermore, changes in feeding pattern were also associated to altered expression of hypothalamic orexigenic neuropeptide. Exposure to dark further confirmed that liver-born inputs were sufficient to drive endogenous clock. These data reveals an important new mechanism by which liver-born inputs, presumably direct to the brain could act as a dominant signals in the control of feeding behaviour and metabolism.
HUMAN activities also focused on the WSB/EiJ mouse model because it is long-lived, resistant to obesity induced by dietary fats, and depict low circulating levels of thyroid hormones. Since transcriptomic analysis suggests that hypothalamic T3 content of these mice could be increased, we used the WSB/EiJ mouse as model to set up proteomics studies. Different studies using mice at different age (i.e. 3 month and 18 month old) challenged with 3 different dietary regimens (i.e. control diet, high fat diet for 3 days, high fat diet for 8 weeks) were performed . We could confirm that differences in metabolic adaptability and in longevity observed in mice with different T3 status could be the results of differential regulations of proteins involved in different processes, especially those which are known to be regulated by T3. These results have identified several critical pathways that are potentially involved in the aging process, as telomere regulation, mitochondria and lipid metabolism. Moreover, our results on the link between central D2 expression/activity and BAT thermogenesis suggest that central sensitivity to thyroid hormone could be crucial for an effective peripheral energy expenditure, and thus, for the homeostasis of the energy balance. Transcriptomic and proteomics analyses showed that WSB/EiJ mice (hypermetabolic phenotype despite lower circulating TH, higher longevity) have a hypothalamic gene expression pattern different from the one of C57Bl/6J mice (sensitive to HFD induced obesity), especially in the nuclei involved in the central control of food intake and energy expenditure. The pathways implicated in this specific pattern included the mitochondria and obesity pathways. These results confirm that body energy balance is centrally controlled by specific gene networks, which interact to drive energy expenditure and fat storage. Also, proteomics and transcriptomics data analyses from mouse model of increased central sensitivity to TH emphasise the importance of mitochondrial pathways in the central and peripheral control of metabolism, favouring healthy aging. In particular, long-lived mice present less detoxification enzymes and molecules controlling oxidative stress than normal lived mice, which can suggest that high need of detoxification can exhaust the organism and be deleterious for lifespan, as is a highly challenged metabolism.
A fundamental pillar of HUMAN was the development of robust methods and standard operating procedures (SOPs) for metabolic profiling of multiple biological matrices including plasma, tissues (brain, muscle, liver, kidney, pancreas and gut), cell media, cells (fibroblasts and iPSCs), cerebrospinal fluid, brain microdialysates, faeces and urine. The developed methods were also applied for metabolic profiling of changes in cell culture media that take place during differentiation of human iPSCs into hepatocyte-like cells. Consumption of nutrients from the cell media, as well as secretion of metabolites in to the cell media were monitored in order to identify the biochemical pathways important for cell growth. This resulted in the optimization of the protocols for iPSC cultivation and differentiation.
The effect of liver humanization on the metabolic profiles of mice brain and plasma was investigated by comparing metabolic profiles from brain tissues (cerebellum, cortex, hippocampus, hypothalamus and striatum) of humanized mice and non-humanized mice. All types of investigated tissues from humanised mice showed distinct differences in the metabolic profiles when compared to those observed in the control mice, which proved that humanisation of the liver had a profound effect on the overall metabolic profile of brain and plasma. Several metabolic pathways showing restoration of the liver function in humanized mice after cell transplantation were also detected, confirming successful humanization and hence validating both the applied humanization procedure and metabolomics as a reliable approach for analysis of samples from humanized mice.
In order to understand the cross-talk between human hepatocytes and other organs in the humanized mouse model, a metabolomics approach was applied to investigate differences in metabolic profiles of tissues from humanized and murinised mice (kidney, pancreas, liver and plasma). In all tissues clear differences between both types of mice were observed, mainly connected to the changes in lipid metabolism. Effect of the genotype of cells used for humanization (with protective or risk allele for type 2 diabetes) was seen only in liver samples and was characterized by lower levels of sugar phosphates and higher levels of the majority of fatty acids and cholesterol in the livers humanized with hepatocytes with the risk allele. This finding has to be put further into the biochemical perspective. Integration of metabolomics data from studies on humanized mice with the data produced by other partners of the HUMAN Consortium, will enhance our understanding of the humanized mice as a model for studying human diseases and its application in development of novel treatments for these.
The metabolomics approach was also applied to pinpoint metabolic pathways changed in age-related diseases, pathways connected to healthy aging and influence of genes on these. Human samples from two cohorts were analysed – from the Leiden Longevity Study and the Bologna Healthy Aging and Longevity Study. It was possible to pinpoint metabolic pathways changed in such diseases as type 2 diabetes, metabolic syndrome, cirrhosis and those changed in centenarians and semi-supercentenarians as compared to control groups. Metabolites correlated with early insulin resistance in the non-diabetic population were also detected, adding additional knowledge to our understanding of the pathogenesis of type 2 diabetes and being potential supportive biomarkers for early detection of insulin resistance. Metabolic profiles connected to a TCF7L2 polymorphism in both studied cohorts pinpointed several metabolites that connected the TCF7L2 mutation with inflammation and with action of gut microbiota. In summary, the majority of the identified changes that were correlated with age-related diseases, healthy aging and influence of genes on these were connected to inflammation, energy and lipid metabolism as well as with action of microbiota. Based on these results lipid-related and inflammation-related pathways, together with modulation of microbiota activity, were identified as potential targets for treatment or support of treatment of age-related diseases. Obtained results suggested also that treatment effect could be at least partly achieved by changes in the dietary habits of the sick individuals.
HUMAN extensively characterized humanized mice with a multi-omic approach forcing us to optimize the methods to study the epigenomic (including DNA methylation and histone marks enrichment profiles), transcriptomic and proteomic profile in a condition in which human and mouse molecules co-exists and should be specifically detected. Hence, innovative workflows were proposed for sample handling and human-murine data disambiguation in order to achieve human-specific, multi-omic data analysis and integration. HUMAN has thus developed a novel method for rapid assessment of degree of humanization in mice. The method relies on the ability to distinguish and quantitated human and murine proteins (namely Albumin, Transferrin and Apolipoprotein-E) in blood thus giving a quantitative humanisation percentage of the mouse.
In addition, proteomic and post-translational modification (PTM) analysis in long-lived and short-lived mouse, showed that the main differences was a lower PTM levels (specifically carbonylation, methylation and dimethylation) in the long-living mouse livers compared to the short-living. Remarkably, high fat diet did not produce any significant changes.
Innovative NGS-based methodologies has been applied for the concomitant analysis of epigenomic and transcriptomic profiles of humanised mouse livers carrying TCF7L2 genetic variants. All library-preparation protocols (either RNA-seq, ChIP-seq or RRBS) relied on a LIMS-supported platform and were automatized in order to reduce manual error and increase reproducibility. Data analysis was performed with optimised and fully automated workflows comprising five major nodes: i) raw data processing; ii) mapping and disambiguation; iii) quantification; iv) quality control; and v) statistical analysis. NGS data determined the percentage of liver humanisation. Since humanisation levels proved to be highly conserved (with an estimate deviation from true value of 3%) among different NGS data types, HUMAN was thus able to generate robust protocols of the high quality.
The multi-omic strategy highlighted variations in molecular profiles in humanised mouse models associated with humanisation, donor genetic background, diet and hormone treatment.
The epigenetic and transcriptomic modifications observed in mouse with humanized livers versus mouse with livers containing only mouse hepatocytes showed significant enrichments in genes encoding proteins belonging to metabolic pathways (e.g. bile acid metabolism, xenobiotic metabolism, amino acid metabolism, lipid and lipoprotein metabolism).
Multi-omic modifications were assessed also between humanised mouse models either carrying TCF7L2 rs7903146 protective or risk allele. In particular, RNA-seq analysis highlighted changes in the expression of genes associated with cell cycle (e.g. regulation of cell replication and chromosomal segregation); extracellular matrix organization, metabolic pathways (e.g. collagen and steroid hormone biosynthesis, biotransformation of xenobiotics) and immunological mechanisms (e.g. interferon signalling, cytokine secretion, ROS response). Interestingly, changes in TCF7L2 expression were not highlighted. RRBS analysis revealed enrichment of differentially-methylated regions (DMRs) on genomic regions associated with cell death mechanisms (e.g. death receptor and TNFR1-induced signalling as well as regulation of caspase 8 and c-FLIP activity).
ChIP-seq analysis identified differential, H3K27/H3K4 enrichment. Genotype-associated changes in H3K27/H3K4 enrichment were found at loci linked to metabolic processes (e.g. biotransformation of xenobiotics, metabolism of lipids, bile acid, fatty acids, etc.). In particular, changes in H3K4me1 enrichment were found associated with vesicle trafficking and immunologic process mediated by cytokines and chemokines (e.g. inflammatory response); while changes in H3K27ac enrichment were mainly found in association with cellular organisation and gene regulation loci (e.g. cytoskeleton organisation, meiosis and chromatin organisation). Interestingly, no significant enrichment at TCF7L2 nor TCF4 binding sites was found. Cross-omic comparison of NGS data indicated several changes in biologic processes including extracellular matrix organisation, vesicles trafficking, xenobiotics metabolism and steroid hormone biosynthesis. Among the most prominent genes affected by TCF7L2 genotype both at transcriptomic and epigenomic levels are TRIM31, ADGRE5 and PEG which are associated with metabolic diseases and longevity. Another set of differentially regulated genes comprise NAAA, EPB41L4A, GATA3 and BIRC5 which are implicated in Wnt signalling pathway (TCF7L2 belongs to Wnt signalling pathway).
Human was able to elaborate the concept that TCF7L2 variants ectopically affect epigenomic and transcriptomic profiles of humanised mouse livers. Indeed, whereas no changes in TCF7L2 expression nor TCF7L2 promoter histone enrichment were found, changes at multiple omic levels were reported for a variety of molecular pathways including Wnt signalling. The interpretation of these data has to include the possibility that the expression of TCF7L2 mRNA and protein in humanized livers is likely to be dynamic in living mouse. Thus, the analysed mRNA levels at the endpoint of the study (upon sacrificing the mouse) does not necessarily reflect the expression dynamics of TCF7L2 protein in vivo.
In order to manage all the volume of data generated and the complexity of the project, HUMAN developed a central infrastructure for data storage, management, and sharing. The quality of the raw ‘omics data has been assessed using standardised and validated high-performance workflows. Data pre-processing and quality control, including report generation, has been performed for array- and sequencing-based as well as mass spectrometry-based data. Phenotypic data generated from the diverse models has been integrated and analysed together with the ‘omics data, to identify: i) molecular events associated with phenotypic and disease endpoints, and ii) phenotypic readouts associated with disease endpoints. New bioinformatics, biostatistics, and visualisation tools have been developed. Methods have been implemented specifically to address the challenges of analysing data from humanised mouse models, e.g. to discriminate NGS reads originating from human vs. mouse cells.
Infrastructure
A central, structured FTP server with access control and user management has been established, and storage capacity has been expanded to 6 TB by the end of the project.
To enable data harmonization necessary for integrated analysis, a sample naming and annotation convention has been established, and sample naming templates following this convention have been provided to all project partners.
Bioinformatics pipelines
Several enhancements and new functionalities needed for the HUMAN project have been implemented into the Genedata Profiler® software. For the analysis of ChIP-Seq data, QC metrics, algorithms and statistical tools have been developed and an automatable data processing workflow has been set up. For the analysis of RRBS data from xenograft samples new algorithms have been adopted and an automatable SOP data processing workflow for the detection of differentially methylated regions has been implemented. Specific workflows for the processing and statistical analysis of transcriptomics and epigenomics data created with the platforms NanoString nCounter and Illumina Infinium MethylationEPIC BeadChip have been developed. New algorithms and statistical tools for the analysis of RNA-Seq data have been incorporated into the Genedata Profiler® software and two automatable data processing workflows have been configured for the detection of differential expression and differential splicing, respectively.
Data Analysis
Partner UMIL has been supported in the establishment of an optimized ChIP-Seq protocol by processing and comprehensive quality assessment of ChIP-Seq pilot data. Proteomics data from partners CNRS and UCL obtained from seven organs of two mouse strains under different dietary conditions has been quality controlled, processed and analysed. Data from the UPD RRBS study has been analysed using a refined SOP workflow for read mapping, disambiguation and assessment of humanisation levels of mouse liver samples and integrated with sample-matched metabolic data. A workflow for processing, humanisation assessment and statistical analysis of data from the KI3 RNA-seq study has been set up and executed, which has been reported as part of deliverable D8.3 (Transcriptome signatures of humanised livers). Data from the UMIL ChIP-seq study has been processed and analysed, including quality control and humanisation assessment of the grafted mouse liver samples as well as the systematic validation and functional annotation of the detected H3K27ac peaks. In a collaboration with partner NIHS, machine learning-based models for the prediction of iPSC clone differentiation efficiency based on molecular signatures detected from sample-matched, multi-omics data have been constructed and validated. Partner CNRS has been supported with statistical and network analysis of qPCR data from mouse brain samples to characterise the gene expression signatures associated with strain, age or diet effects in relevant hypothalamic nuclei. Humanised mouse liver models have been characterised by genome-wide histone mark profiles based on H3K4me3/H3K27ac ChIP-seq data, and systematic validation, cross-study comparison, functional annotation and motif analysis of differential peaks has been performed. Extensive cross-omics analyses have been performed through statistical comparison and pathway-based, functional characterization of disease-derived vs. control humanised mouse liver models by integrated analysis of gene expression (RNA-seq), histone mark (ChIP-seq) and DNA methylation (RRBS) profiles.
Potential Impact:
The research activities and the results generated by the HUMAN consortium have a great impact on different levels in the study of longevity, development of metabolic syndrome, on the physiological regulation of metabolisms, and on the cross talk between the brain and gut.
Regarding the SNPs studied, HUMAN was able to confirm the importance of the genetic variances of TCF7L2 and APOE confirm the role of these genes in cardiometabolic health and disease. HUMAN put forward the hypothesis that omega-3 fatty acid intake and physical activity may modify the impact of ApoE ɛ4 on disease risk. This hypothesis may impact a public health perspective.
The collaboration of HUMAN with the international CHARGE consortium, resulted in multiple novel hypotheses and candidate mechanisms linking sleep habits to SNPs to be further explored in future studies. Further, we identified novel mechanisms that could potentially be of help in the development of novel interventions for the prevention of cardiometabolic disease in patients with sleep disorders.
Epidemiological research has often an observational nature which limits the potential of causal inference. Although a randomized controlled trial is considered the golden standard to infer causality of observational associations between a certain exposure and outcome, it is often not feasible. Mendelian Randomization is an increasingly used method to infer causality using genetic variants for the exposure as instrumental variables. In HUMAN, we conducted several Mendelian Randomization studies to infer causal associations of phenotypes of interest to the consortium, and the results of these are od potential clinical importance. The inhibition of Cholesteryl Transfer Protein (CETP) has been long thought to be a therapeutic target for the prevention of cardiovascular disease. Our finding that genetically-determined CETP concentrations are not causally associated with risk of cardiovascular disease, are pivotal to understand the failed trails using CETP inhibitor drugs.
The activity of HUMAN on the generation of iPSCs, their genetic modification with gene-editing technologies and the differentiation towards mature hepatocytes have a tremendous direct and future innovative impact. HUMAN can and has provided the international community with robust and validated protocols and working pathways on how select cell to be differentiated in iPSC, how to grows the latter and genetically modify them with success. Further HUMAN has defined an advanced and accurate way how evaluate the differentiation of iPCS cells into mature human hepatocytes. The successful gene-editing protocols established by HUMAN which has led to the recreation of model of human disease, and of their correction in mice are milestone for other for the development of similar model targeting other pathologies.
All the animal model created, with humanised liver or humanised pancreas, has been turned out to be an example for the development of personalized models that allow the study of the individual subject or the test of new therapeutic approaches, either based on gene editing or based on small molecules. Hence, the research generated by HUMAN is an example of science that is “translatable” by to human condition, and not only translational to it. A significant contribution to the need for generation of animal models that recreated the human condition is the prove that the humanised liver mice, were able to reproduce the negative outcome of the first clinical trials in human in which a nuclear receptor was targeted in the liver.
The work of HUMAN may thus positively shorten the time between discovery and implementation into clinical care, by showing a way how already in a pre-clinical stage predict the outcomes of phase 2 clinical trials.
The discovery of HUMAN on lipoprotein and bile acid metabolism and how human liver affects the macrobiota may positively influence the future development of strategy aiming to prevent and cure cardiometabolic diseases.
The identification, by the integration of proteomics, trancriptomics and metabolomics data sets from long-lived, obesity resistant mice, of several pathways which could contribute to the healthy aging phenotype. These pathways, when confronted with proteomics analysis from “supercentenarian” human samples, were found to be similar in some extents to pathways identified in these human samples. Thus, a potential impact of the work carried out by HUMAN could be the targeting of these pathways in a therapeutic perspective, to achieve healthy aging in human. This important perspective will be refined by an integrated analysis of all the “omics” datasets obtained in mice, and confronted to human datasets, in order to provide strong results showing the most important pathways to target. The result of this big data integrative analysis, comparing mice and human data, will be made available to the international scientific community.
The applied metabolomics approach of HUMAN improved the understanding of biochemical processes disturbed in pancreatic and metabolic diseases. Metabolic pathways that were found to be affected in the studied diseases are potential targets for novel treatments which, together with proper dietary interventions, would in long run contribute to healthier aging of European citizens.
HUMAN was able to identify multi-omic molecular signatures for a thorough characterisation of the humanised mouse model. The set-up and optimisation of a series of high-throughput, NGS-based techniques able to deal with the difficulties associated with the use of animal models in which human and murine genomes are present contextually have a great impact of the future reaserch in which chimeric models are used. Multi-omic data (either proteomic, transcriptomic or epigenomic) generated by the HUMAN Consortium allow for highly-reproducible and robust assessment of liver humanisation in the humanised mouse models and the definition of workflows enabling automated processing of gene expression, DNA methylation and histone-mark analyses reduce can be utilized in the studies of many other diseases. Furthermore, the automated analysis pipeline allows for rapid GO and pathways annotation of hits identified in different NGS-data. Lastly, the proposed NGS data analysis pipeline include cross-omic comparison thereby allowing identification of biological processes affected at a multiple omic level.
The results achieved from the analysis of humanised mouse livers highlighted human genotype-associated changes in the function of the human hepatocytes with particular involvement of metabolic and cellular organisation processes. These results represent the initial step towards the discovery of potentially new regulatory pathways in physiology and in disease onset and progression.
List of Websites:
www.fp7human.eu
