Final Report Summary - MITIN (INTEGRATION OF THE SYSTEM MODELS OF INSULIN SIGNALLING AND OF MITOCHONDRIAL FUNCTION AND ITS APPLICATION IN THE STUDY OF COMPLEX DISEASES)
Executive Summary:
The main goal of MITIN was to elucidate the interplay between the insulin signaling pathway and the mitochondrial function through the combination of computational tools and molecular omics technologies in a multidisciplinary, integrated system biological approach. MITIN scientific objectives are based on research evidence on the cross-link between insulin signaling and mitochondrial function.
The specific main objectives of the project were the following:
a. To develop a computational framework that integrates, as an interaction network, the already available, and the newly generated or predicted data on all components (gene expression, proteins or metabolites) of the Insulin signalling pathway and the Mitochondrial processes, and the relationships within and between both systems. This framework will mainly consist in a complex network that will be designed focusing on several important features, from which we want to emphasize: [1] the possibility of handling different types and levels of relationships (links) that connect components within and between both systems (from direct protein-protein interaction to common cellular location); and [2] the possibility of querying the network for potential changes at all levels when specific components or links are deliberately perturbed.
b. To enrich the quality of the computational framework by generating information related to specific perturbations known to selectively modify insulin signalling and/or mitochondria function.
c. To determine the predictability of the computational framework by validating specific hypothesis emerging from the developing model to be tested using appropriate in vitro and in vivo model.
d. To evaluate the translatability of the new knowledge to human diabetes or related diseases.
Project Context and Objectives:
PROJECT CONTEXT.
A common trait to disorders such as type 2 diabetes or obesity is the development of insulin resistance. The link between insulin signalling and these diseases is unclear but there is growing evidence indicating the existence of a bidirectional cross-talk between insulin signalling and mitochondrial function that may be relevant for the pathogenesis of these diseases. Based on this evidence, MITIN main goal was to identify novel mitochondrial-dependent mechanisms responsible for the development of insulin resistance. This was done by the use of technologies of Systems Biology and the generation of computer-based tools that will permit the study of complex biological systems that integrate different regulatory networks.
For the global visualization and interpretation of the “insulin signalling/mitochondria” complex system, our intention was to develop a computational framework that will store and integrate all possible data for each of the subsystems, both currently available data and data generated within the project by transcriptomics and lipidomics analysis. This would allow the prediction of functional associations and interactions between both processes, which were tested under specific hypothesis-driven studies in mammalian cells, mice and Drosophila. Furthermore, with this network, we simulate different complex metabolic and regulatory scenarios for normal conditions and diseases, and identify and predict possible changes when other parts of the system are deliberately perturbed.
The integrated model of “insulin signalling/mitochondria” were progressively refined as the new data generated within the project and integrated into the computational framework contributed to its enrichment. Thus we analyzed experimentally the effect of specific perturbations in the insulin-signalling pathway on mitochondrial function. The opposite paradigm were also investigated, so that the impact of primary defects in mitochondria function on insulin signalling were characterized. We took advantage of cross-species information by investigating these paradigms in mammalian cells and mice but also in Drosophila. By using a Systems Biology approach and investigating how different perturbations modulate cell transcriptomes, and lipidomes, we aim to identify integrated homeostatic responses involving the insulin signalling network and the mitochondria networks.
Validation of some of the strongest associations detected by virtue of the integrated systems biology of insulin signalling/mitochondria will generate high value targets of use in therapy against Complex Diseases such as diabetes, obesity, or pathophysiological traits of the Metabolic Syndrome. Besides, computer-based tools generated in the project will be applicable to the analysis of the mechanisms that trigger other Complex Diseases.
The development of new profiling technologies has generated an unprecedented opportunity to generate huge amount of heterogeneous data. This opened great expectations about the possibility of tackling Complex Diseases as well has generated new technological challenges such as integrating these databases and networks into meaningful models. For these, new computational tools need to be developed and experience gained by applying them to specific biomedical problems.
In this project we have explored how/if a Systems Biological approach is a good strategy to unravel a complex biomedical problem. The problem selected is insulin resistant diabetes and its interaction with mitochondrial function. This is an important medical problem given its epidemic proportions, its future prospects of resulting in a subsequent epidemic of cardiovascular events, its obvious financial implications and, more importantly, because it is cause of human suffering and death.
Defects in insulin action are commonly observed in a variety of different Complex Diseases or syndromes such as type 2 diabetes, obesity, hypertension or cardiovascular disease. In this regard, there is growing evidence indicating the existence of a bidirectional cross-talk between insulin signalling and mitochondrial function. Thus, alterations in mitochondrial function may trigger the development of insulin resistance in tissues and, in addition, insulin signalling has a direct impact on mitochondrial function.
Based on these observations and the level of their complexity, we have developed a Systems Biology approach to determine whether this research strategy is useful to provide a global insight on the molecular mechanisms linking both biological systems, i.e. insulin signalling and mitochondrial function. Specifically we used a multidisciplinary approach that will, on one hand generate and identify a wide range of available quantitative biological data and, on the other hand, develop computational tools to integrate this information into integrated models reflecting the nodes of interaction between insulin signaling and mitochondrial function.
To enhance the power of this approach we took advantage of data obtained from different levels of organization (cultured cells and in the in vivo animal) and in two different organisms (Drosophila and mice). Transcriptomics and lipidomics analyses will permit to identify homeostatic responses affecting insulin signalling and mitochondrial systems. The reasons by which we focus in this proposal on two omics (lipidomics and transcriptomics) are straightforward. Lipid metabolism is markedly modified by insulin or by changes in mitochondrial function, and, in turn, lipids are known to impact on insulin sensitivity. In addition, both insulin and changes in mitochondrial function modify the gene expression profile in cells or tissues, and in fact, insulin modifies the expression of genes encoding for mitochondrial proteins and conditions of insulin resistance show a deficient expression of genes encoding for subunit of the OXPHOS system, which parallel deficient mitochondrial activity.
Validation of some of the strongest associations detected by virtue of the integrated systems “insulin signaling/mitochondria” will generate high value targets that may be used for future development of pharmacological intervention strategies. Besides, generation of tools for the integration of Systems Biology will be of application to better understanding other Complex Diseases.
OBJECTIVES.
The main goal of MITIN was to elucidate the interplay between the insulin signaling pathway and the mitochondrial function through the combination of computational tools and molecular omics technologies in a multidisciplinary, integrated system biological approach.
MITIN scientific objectives were based on research evidence on the cross-link between insulin signaling and mitochondrial function (in which partners of the consortium have taken part). The project aimed at capitalising this knowledge and will pave the way for the translation of research into relevant applications.
The specific main objectives of the project were the following:
a) To develop a computational framework that integrates, as an interaction network, the already available, and the newly generated or predicted data on all components (gene expression, proteins or metabolites) of the Insulin signalling pathway and the Mitochondrial processes, and the relationships within and between both systems. This framework mainly consisted in a complex network that was designed focusing on several important features, from which we want to emphasize: [1] the possibility of handling different types and levels of relationships (links) that connect components within and between both systems (from direct protein-protein interaction to common cellular location); and [2] the possibility of querying the network for potential changes at all levels when specific components or links are deliberately perturbed.
b) To enrich the quality of the computational framework by generating information related to specific perturbations known to selectively modify insulin signalling and/or mitochondria function.
c) To determine the predictability of the computational framework by validating specific hypothesis emerging from the developing model to be tested using appropriate in vitro and in vivo model.
d) To evaluate the translatability of the new knowledge to human diabetes or related diseases.
Project Results:
The major work performed during the operation of MITIN has been the following:
a) The development and validation of the computational framework through retrieval of available information and the integration of data into a network.
b) The development and application of analytical protocols for metabolic profiling of mitochondria.
c) The analysis of the impact that mitochondrial dysfunction has on insulin sensitivity.
d) The analysis of the effect of alterations of insulin sensitivity on mitochondrial function.
a) The development and validation of the computational framework through retrieval of available information and the integration of data into a network.
We have constructed a consensus insulin pathway from several public resources, including (Biocarta; www.biocarta.com Kegg; www.genome.jp/kegg/ Reactome; www.reactome.org and PID; http://pid.nci.nih.gov/) and a commercial resource (Biobase; www.biobase.de). This database was manually curated and refined by the participation of different partners of the consortium.
In order to select the parts lists that compose mitochondrial proteins or genes, we selected a total set of 900 proteins from mitoP2 database (www.mitop2.de/) and organized them according the biological process, mitochondrial compartment and molecular function. Like done for the insulin pathway, the set was manually curated by the participation of the expert groups in the consortium.
Besides human tissues and cells, the consortium aimed to investigate with both mice and drosophila melanogaster. We identified each mouse and fly orthologous gene/protein for all involved proteins. Orthologous genes were selected by the best one to one bidirectional blast hit between the three organisms.
One of the first steps towards the main objectives of this workpackage was to build a complete and manually curated list of the proteins involved in the insulin pathway and the mitochondrial function, and connect them in order to better understand how these two systems are functionally related. In that sense, while the data from perturbation models within the consortium is being generated, we have taken advantage of the publicly available data from several databases and "omics" experiments to construct a first version of these systems network.
The first version of the network contains both protein-protein interaction data, and gene co-expression relationships.
To identify genes that participate in the crosstalk between insulin and mitochondria, we have looked for protein-protein interactions. With that objective in mind, we have used a non-redundant set of 23 protein interaction datasets. These interactions had been mapped to expression data available, in order to estimate whether each interaction is likely to occur in each given tissue of interest (Bossi and Lehner 2009). According to these data, we have studied interactions possible to occur in liver, cardiac myocites, skeletal muscle, smooth muscle, and adipose tissue.
The first level of evidence of connection between the two systems are those proteins that appear both in the insulin signalling pathway and that are described to form part of the biology of the mitochondria. These proteins are encoded by the following genes: BAD, PRKAR2B, MAPK9, ARAF, GCK, YWHAE, RAF1, SHC1, YWHAZ, PPP1CC.
When exploring the direct interactions between mitochondria and insulin genes, we found seven insulin proteins that interact at least with two mitochondrial proteins (each of these interactions reported by two independent publications), and eight mitochondrial proteins that interact at least with two insulin proteins (each of these interactions reported by two independent publications).
We also explored the indirect interactions between mitochondria and insulin using one protein as intermediary (internode). That way, we identified 34 proteins that interact at least 3 times in total with mitochondria and/or insulin. After excluding those internodes that could have occurred by chance, we ended up with a total of 29 high confidence internodes.
Taking all these interactions into account, we could construct a preliminary MITIN general network, based on manually curated insulin and mitochondrial systems and highly stringent physical protein-protein interactions information. Taking into account the tissue expression of each of the nodes (genes) that participate in each of the interactions, a network for each of the tissues of interest including liver, fetal liver, cardiac myocytes, adipocytes, smooth and skeletal muscle, has been constructed.
Together with that, we have developed a method to calculate a signal-to-noise ratio of each observed interaction, using random permutation tests, assigning a p-value for each of the links observed.
Among all these high confidence interactions, some of them make the results promising, as the network allowed us to identify previously known relationships, such as the relationship between EGFR and insulin signalling (Borisov et al. 2009), or the connection between BCL2, MAPK1 and MAPK2 genes and insulin signalling. Regarding the internodes, ANXA2, RPS6KA1, and APP genes, among many others, have previously shown to have a functional connection between both MITIN systems, as reviewed by several members of partner 1 (IRB Barcelona). A manual inspection an literature search performed by the investigators of partner 1 (IRB Barcelona) showed that there was a functional known connection for around 75% of the internodes. Some of these predicted internodes have been selected to be experimentally validated at partner 1 (IRB Barcelona) laboratory. Although for the remaining 25 % of candidates there is no previous known evidence of connection between insulin and mitochondria, these candidates should better be considered as potential novel candidates rather than false positives.
As a second source of links that interact between insulin and mitochondrial function, we analyzed gene co-expression networks (GCN). GCN can be calculated from microarray datasets from which a sufficient amount of samples exists. Genes that are correlated across different samples (termed co-expressed), can be classified as pertaining to a common module and are likely to perform a similar function in the studied tissue. For the gene co-expression network, we have used the dataset of Schadt et. Al (Schadt et al 2008), which consists of expression data of 400 healthy human liver samples. We used WGCNA R package (Langfelder and Horvath 2008) to calculate the weighed gene co-expression network as well as identifying modules of co-expressed genes. The identification of modules that contain a significant enrichment of both insulin signaling and mitochondrial function genes will allow us to identify new candidate players between both systems. In addition, the co-expression interactions that overlap with protein-protein interaction findings will be used to positively score those relations.
All the computational predictions obtained during this activity and all the data that will be generated by the different members of the consortium should be integrated into a common, easily accessible, and flexible database. For that purpose we have integrated all the available data into a mySQL relational database. The MITIN database is structured in three main compartments. The main compartment is centered in Homo Sapiens data and contains most of the information. Within this compartment, there is data about mitochondria, insulin, and internodes as obtained by computational analysis. In addition several gene nomenclature and synonyms databases will allow performing queries coming from different sources. As one of the sources of data that will be needed to be integrated will come from microarray data, several probe-to-gene annotation tables have been integrated into the database. Finally, as one of the final goals of building the MITIN network is to identify new participants in the aetiology of Type 2 Diabetes and its related diseases, data on genes involved in mendelian diseases related to metabolism
(OMIM, http://www.ncbi.nlm.nih.gov/omim) and genetic variation associated to Type 2 Diabetes and related complex diseases (GWAS database, Hindorff et al 2009) has been included into the MITIN mySQL database. Therefore, the relational database allows us to query whether the MITIN network genes are related to both mendelian and complex diseases related to Type 2 Diabetes. This structured data will allow the integration of omics data coming from different partners, collaborators and clinicians.
The main human compartment of the mySQL database is connected through orthology to mouse and drosophila data. Both mouse and drosophila data contain several gene nomenclature and synonyms database to allow data coming from different resources to be integrated. This database allows us to query a set of genes that are transferred through orthology in the case of data coming from mouse or drosophila, to interrogate whether these genes have been previously implicated in both mendelian and complex diseases, as well as to interrogate functional genetic variation that might be contained within each of these genes.
Manually curated review of the candidates and selection of candidates.
We have performed a thorough review of the mitochondrial and insulin signalling proteins that functionally interact in mammalian cells. As a result, we have an updated view on the mitochondrial and insulin signaling proteins already known to physically and functionally interact in mammalian cells.
Among all the high confidence interactions detected, some of them make the results promising, as the network allowed us to identify previously known relationships, such as the relationship between EGFR and insulin signalling, or the connection between BCL2, MAPK1, MAPK2 and MAPK8 genes and insulin signaling (Figure 1). Regarding the internodes, ANXA2, RPS6KA1, and APP genes, among many others, have previously shown to have a functional connection between both MITIN systems, as reviewed by several members of partner 1 (IRB Barcelona). A manual inspection of a literature search performed by the investigators of partner 1 (IRB Barcelona) showed that there was a functional known connection for around 75% of the internodes.
Some of these predicted internodes have been selected to be experimentally validated at partner 1 (IRB Barcelona) laboratory. Although for the remaining 25 % of candidates there is no previous known evidence of connection between insulin and mitochondria, these candidates should better be considered as potential novel candidates rather than false positives. The signal-to-ratio analysis propose BRAF, ANXA2, MAPK8IP3, RPS6KA1, ANXA2P2, PPP2CA, CRK, TSC2, and IGF1R, as promising candidate internodal proteins to be evaluated experimentally. The disadvantage of these options is that most of them are not novel as candidates linking mitochondria and insulin signaling.
Construction of the MITIN network (second version of the computational framework after operation at a lower stringency).
To establish the molecular links between the insulin and the mitochondrial systems, we used the Global Network that we generated to identify different types of relationships between the insulin and mitochondrial systems: direct interactions -genes from mitochondria or the insulin signaling pathway whose products either physically or functionally interact directly with at least one protein of the other system-, and indirect interactions, that is, allowing one intermediate gene, not belonging to the mitochondria nor to the insulin pathway but connected to at least one insulin and one mitochondria gene. These genes that functionally bridge these two systems will be termed from now on internodes or MITIN linker genes. The MITIN network consisted of a total 886 genes (nodes), 65 direct interactions and 1194 indirect interactions triggered by 286 linker genes. All these linker proteins were connected to a mean number of 2.1 Insulin genes and 1.7 mitochondria. When taking into account the number of functional evidences (regardless the number of nodes) between both systems the linker genes showed a mean of 2.6 functional evidences linking insulin pathway while a mean of 2.0 evidences linking the mitochondria genes. Figure 2 shows a high confident subset of linker genes that show at least three evidences linking insulin signaling and mitochondria genes simultaneously.
The same network also allowed us to define which mitochondrial genes are more connected to insulin, and vice-versa, and either directly or indirectly. The top five insulin genes most connected to mitochondria insulin genes were nucleolar and coiled-body phosphoprotein 1 (NOLC1), ribosomal protein S6 (RPS6), inhibitor of kappa light polypeptide gene enhancer in B-cells, kinase beta (IKBKB), pyruvate kinase, liver and RBC (PKLR), and v-src sarcoma (Schmidt-Ruppin A-2) viral oncogene homolog (avian) (SRC), with a total of 99, 40, 31, 28 and 22 indirect connections with mitochondria, respectively. Regarding mitochondria, the five most connected gens with insulin were Tu translation elongation factor, mitochondrial (TUFM), tumor protein 53 (TP53), solute carrier family 25 (mitochondrial carrier; adenine nucleotide translocator), member 5 (SLC25A5), polymerase (DNA directed), gamma (POLG), and estrogen receptor 1 (ESR1), with a total of 93, 36, 29, 25, and 19 indirect connections with insulin, respectively.
While a majority of the internodes are considered novel, as they have not previously been described as having a bridging role connecting both systems, some of them have already been shown to interact with both systems. For example, TRAF2 is connected to four insulin genes and two mitochondria genes, according to our analysis. Indeed, TRAF2 has been shown to be connected to MAP3K1 (MEKK1) and CAV1 (caveolin-1) insulin genes, and to MAP3K5 (ASK1) and CASP8 (caspase-8) mitochondrial genes. A possible connection to MTOR (mTOR) has also to be considered. Another example is NFKB1, showing interactions with four insulin signaling genes and three mitochondria proteins, according to our analyses.
We have validated the relevance of one specific gene, i.e. MAPK8 (JNK) as a linker between mitochondrial function and insulin signaling. In this regard, the following studies were performed: a) Mfn2 deficiency activates JNK in muscle cells; b) Mfn2 deficiency causes activation of JNK in conjunction with mitochondrial dysfunction, enhanced ROS production and ER stress in liver and muscle; c) Chronic treatment with TUDCA or N-acetylcysteine normalize JNK activity, insulin signaling and glucose tolerance in liver-specific Mfn2 knockout mice.
b) The development and application of analytical protocols for metabolic profiling of mitochondria.
The methods for primary metabolites (LC/MS/MS), global lipidomics (UPLC/MS), and global metabolomics (GCxGC-TOF/MS), and branched chain amino acids and their keto acids (HPLC) are available for analysis.
Additionally, we have used time under this task for improving the experimental protocols for tissues and cells, as well as improving the identification of lipids (using internal MS/MS libraries). Simultaneously, we have been developing the methods for analyzing the acylcarnitines and acyl-CoA:s (UPLC/MS/MS), and Cardiolipins (by UPLC/MS/MS).
c) The analysis of the impact that mitochondrial dysfunction has on insulin sensitivity.
Summary of progress
The specific objectives of this workpackage were:
[1] To study the effect of disruption of mitochondrial dynamics on insulin signalling in cultured cells, in mice and in Drosophila melanogaster;
[2] To analyse the mechanisms that control insulin sensitivity induced by selective defects in either PGC-1alpha or PGC-1beta.
As part of the first objective, partner I (IRB) has developed a battery of different loss and gain of function in vitro models to assess the metabolic impact of Mfn2, a protein involved in mitochondrial dynamics in insulin action in muscle cell model as L6E9 or C2C12 myotubes. In addition, to further study the effects of the aforementioned proteins involved in mitochondrial fusion in a more complex organism, partner 1 (IRB) also generated both skeletal and liver tissue specific KO mouse model for Mfn2. Metabolic profiling of both models have been extensively analysed. Lipidomics of specific key models have been generated and analysed by Partner 3 (VTT).
Partner 6 (DKFZ) has generated KO flies for marf (Mfn2) and fzo (Mfn1) to study the impact of impaired mitochondrial function on insulin signalling cascade under different nutritional stressors (palmitic acid). In addition to this, they have identified C69009 as a mitochondrial gene with homology to PGC1a that is regulated by insulin. Lipidome analyses have been performed by Partner 3 (VTT).
The second objective involved the global physiological characterization of rodent models As part of the second objective, aiming to establish the role of PGC1alpha/ PGC1beta dysfunction on the mechanisms controlling insulin sensitivity, partner 5 (UCAM) has developed and performed a battery of experimental models aimed to challenge energy homeostasis with particular interest in carbohydrate metabolism.
The experimental approach used involved selected nutritional, environmental and genetic interventions aimed to challenge the mitochondrial function resulting from total ablation or heterozygosis of PGC1s. Briefly these challenges involved chronic nutritional interventions with high fat diet (45% kCal from saturated fat) as well as high protein diet (40% kCal from protein) for some of the rodent models described below. As part of the genetic challenges Partner 5 (UCAM) generated the Pgc1beta x ob/ob model as a paradigm model of mitochondrial dysfunction in a context of massive energy surplus elicited by ob/ob. Moreover, we also created the double heterozygote PGC1beta x PGC1 alpha. In addition, the PGC1beta transgenic model to perform some specific comparison respect to the other models.
Partner 5 have also performed an extensive phenotypical characterization as a part pf our integrative physiology programme that includes analysis of energy expenditure, carbohydrate metabolism. We have also performed as part of the ex vivo profiling, gene and protein expression analysis, serum biochemistry measurements and histological analysis of key organs that includes adipose tissue, liver, pancreas. In addition, we have also produced percoll purified mitochondrial fractions for specific rodent models in liver and skeletal muscle to characterize the effects of PGC1s ablation on the mitochondrial lipidome (task performed by partner 3-VTT). Finally we have also produced gene expression data that has been used for the model generated by partner 2-BSC-CNS.
Significant results
As part of the in vitro studies in myotubes and myoblasts, Partner 1 showed that deletion of proteins involved in mitochondrial fusion and fission impaired myogenic processes affecting mitochondrial morphology and function (e.g. Mfn1, Mfn2, OPA1 knowdown, Drp1 knock-in and Drp1 overexpression) linked in some specific cases with impairment of insulin signalling cascade at different nodes. Microarray analysis of Mfn2 deletion as well as overexpression of drp1 (native and dominant negative form) revealed a list of relevant candidate genes to explore further. Moreover partner 1 also showed that the Mfn2 ablation induces endoplasmic reticulum vacuolization and unfolded protein response in cells treated with ER stress inducers. Microarray analysis of Mfn2 loss of function and drp gain and loss of function models revealed substantial differences in the identity of pathways dysregulated as well as the severity of the impact.
In drosophila, partner 6 has demonstrated that Marf KO (Mfn2) development and life is severely compromised with clear evidence of mitochondrial defects. The heterozygote showed alterations in mitochondrial morphology and impairment in mitochondrial function and number as well as a reduction in oxygen consumption that is consistent with increased TG levels and impairment in insulin signalling (in presence of palmitic acid as equivalent to HFD in rodents) and resembles the data generated in rodent models as explained in detail below) surprisingly, these heterozygotes also show a significative increase in the life expand respect to wt. Microarray analysis suggest alterations in carbohydrate metabolism. Moreover, the heterozyote for fzo (Mfn1) showed a substantial increase in fat content (elevated Tg levels) in combination with hypoglycaemia.
In vivo experiments using the rodent models aforementioned Partner 1 (IRB) has demonstrated that:
a)Ablation of Mfn2 causes marked mitochondrial dysfunction at multiple levels and insulin resistance in skeletal muscle. Interestingly, conditional KO mice for mfn2 in skeletal muscle showed impaired glucose tolerance and hyperinsulinemia when fed with HFD. Moreover, ablation of Mfn2 in muscle caused a severe impairment in the insulin signalling cascade in line with the data observed in vitro.
b) Specific ablation of Mfn2 in liver causes marked hepatic glucose production, glucose intolerance in both chow and HFD fed mice and severe insulin resistance after being challenged by HFD. At molecular level, the livers of these mice showed a severe impairment in insulin signalling cascade as well as evidences of endoplasmic reticulum stress and altered ER morphology and calcium homeostasis. Lipidomics analysis of these mice showed profound alterations in the lipidome with marked increased in triglycerides and decrease in phospholipid species consistent with the presence of hepatic steatosis.
Partner 5 (UCAM) highlights the following results regarding specific PGC1 ablated models used:
a) the ablation of PGC1beta has a positive impact on carbohydrate metabolism (increased glucose tolerance) in aged mice, profile accompanied by reduction in fat percentage and liver mass despite confirmed mitochondrial dysfunction and decreased global energy expenditure in aged mice (over 15 months old).
b) The ablation of PGC1beta in the context of massive energy surplus depicted in the leptin deficient mouse ob/ob cause a significant amelioration of the hyperglycemia normally observed in the ob/ob mice consistent with a mild increase in fasting insulin levels. Nevertheless, this picture is accompanied by increased markers of micro and macrosteatosis that may indicate liver failure. Paradoxically, the increase in hepatic fat content may act as a buffer protecting other organs such pancreas. Decreased energy expenditure in double ko respect ob/ob littermates specially during day phase is consistent with alterations in carbohydrate metabolism after o/n fasting.
c) The reduction in the allelic expression of both PGC1beta and PGC1 alpha (DHET) produced a more insulin sensitive mouse with higher energy expenditure suggestive of a hypermetabolic state consistent with the activation of thermogenic markers in brown adipose tissue and the leaner phenotype observed. DHET also protected against fatty liver imposed by HFD.
d) Preliminary analysis of global lipidomics, acylcarnitines and cardiolipins (performed by partner 3) reveals changes in the lipidome of key metabolic organs as well as isolated mitochondria in PGC1 deficient mice.
d) The analysis of the effect of alterations of insulin sensitivity on mitochondrial function.
Using a lentivirus shRNA-mediated approach we successfully generated myocyte cell lines with deficient CAP expression, in a cardiac (H9C2) and a muscular (C2C12) cells. Immunoblotting of lysates obtained from puromycin selected individual clonal cells showed that the inhibition of CAP protein in these cells was greater than 95%. Under these conditions, the cells were able to grow and differentiate into myotubes as well as control cells. The H9C2 CAP deficient cells exhibited a reduction of basal mitochondrial membrane potential compared to control cells, a reduced glucose uptake in response to insulin stimulation while there were no changes in the expression of insulin signaling proteins, such as insulin receptor, p85 regulatory subunit of PI3Kinase, AKT or ERK1/2. A slight decrease of Cbl expression was also observed in the CAP deficient cells. Insulin signaling through PI 3Kinase and AKT phosphorylation in response to insulin was normal. We observed however an elevated basal activation of MAPK pathway (p42/p44) in the CAP deficient H9C2 cells compared to controls. We have also examined the phenotype of CAP depletion in the muscle cell line C2C12. Similarly to the results found in H9C2 cells, depletion of CAP did not impair differentiation of CC12 cells into myotubes. Insulin mediated phosphorylation of AKT was normal in these cells as well as activation of the MAPK cascasde (p42/p44). Unlike the H9C2 we did not detected an enhanced basal activation of the MAPK pathway in these cells. Further characterization of the CAP gene depletion was performed by examining the gene expression profiles of these cells by quantitative real time PCR. We found changes in the mRNA levels of the gene coding for the coactivator PGC1-alpha, but not PGC1-beta, or the transcription factors NRF1 or TFAM. No statistically significant changes were observed in the level of expression of the mRNA coding for mitochondrial proteins, COX2, COX4 or ATP synthase, however there was a trend for COX2 to be reduced in the CAP depleted cells. We found a statistically significant reduction of the gene coding for UCP2 in the CAP depleted C2C12 cells compared to controls. We did not detect any changes in the protein content of the mitochondrial proteins Cox1 or Porin. The amount of Mitofusin2 detected in these cells was also similar to controls.
We have also achieved an efficient knockdown of c-Cbl. Analysis of these cells revealed no major changes in the Phosphatydilinositol 3 kinase/AKT or MAPK pathways by insulin. Activation of the stress kinase JNK was also comparable to that observed in control cells. The mitochondrial membrane potential of the Cbl Knocked down cells wass elevated compared to control cells. Cbl depleted C2C12 cells show increased mitochondrial membrane potential compared to control and reduced ROS production. ATP content was similar to that of control cells, whereas citrate synthase activity was increased in both Cbl and CAP depleted cells compared to control cells. Cbl depleted but not CAP depleted cells also display increased activation of AMPK and ACC.
We also generated a cell line with impaired signaling throught the PI 3 Kinase pathway following incubation of the cells with insulin for 48 hrs. This strategy sought to recreate a model described in the literature that has been previously utilized to downregulate the expression and signaling through the insulin receptor. As expected, we observed that the insulin receptor expression was reduced approximately 50% in the insulin treated cells. Insulin mediated activation of AKT (via PI3 Kinase) was significantly reduced in these cells, whereas activation of the ERK1/2 enzymes (MAPK cascade) was only decreased by approximately 10%. Under these conditions, we observed that the mitochondrial membrane potential decreased about 50% in respect of control cells (no insulin treated). However measurements of ATP were not significantly different than control cells. Consistent with this, gene expression profiling using quantitative real time PCR indicated that the mRNA levels for subunit of the ATP synthase were no statistically different than those found in control cells. However we detected a tendency for a decrease in the amount of mRNA for PGC-1alpha, COX2 and COX4 in the chronically treated cells. There was a marked reduction in the amount of Uncoupling protein (UCP2) whereas we did not detect changes for the mRNA for PGC-1 beta, NRF1 or Tfam gene expression. The protein content of the mitochondrial proteins Cox1, Porin and Mitofusin 2 were also unaffected. In addition, ROS production over time was measured as quantitative DCF-DA fluorescence and was slightly lower in these cells respective to controls. A complete gene profile analysis using Affymetrix Microarray technology has been carried out to determine the molecular mechanisms by which insulin carries out these effects.
Potential Impact:
During the operation of the MITIN consortium we have interacted with clinical scientists in order to analyze the different pathways to validate the data generated by MITIN. The general conclusion we have reached is that at this point a realistic possibility is to undertake an extensive collaboration with clinical scientists to demonstrate the implication of the genes identified in the MITIN project in the acquisition of clinically-relevant traits within the scope of the Metabolic Syndrome.
Discussions maintained with different clinical investigators with expertise in the Metabolic Syndrome permit to conclude that an excellent approach is to genotype multiple informative SNPs in all the genes identified by the MITIN consortium, and to analyze whether the genotypes associate to relevant parameters measured in the cohorts. This approach should be done in parallel with researchers holding different well phenotyped cohorts. We will select collaboration with clinical scientists having access to cohorts showing a number of properties, namely, that they are extremely well phenotyped and for which they have collected plasma, DNA and subcutaneous and visceral adipose tissue and they have measured the expression of a substantial number of genes, and plasma biochemistry.
The main goal of MITIN was to elucidate the interplay between the insulin signaling pathway and the mitochondrial function through the combination of computational tools and molecular omics technologies in a multidisciplinary, integrated system biological approach. MITIN scientific objectives are based on research evidence on the cross-link between insulin signaling and mitochondrial function.
The specific main objectives of the project were the following:
a. To develop a computational framework that integrates, as an interaction network, the already available, and the newly generated or predicted data on all components (gene expression, proteins or metabolites) of the Insulin signalling pathway and the Mitochondrial processes, and the relationships within and between both systems. This framework will mainly consist in a complex network that will be designed focusing on several important features, from which we want to emphasize: [1] the possibility of handling different types and levels of relationships (links) that connect components within and between both systems (from direct protein-protein interaction to common cellular location); and [2] the possibility of querying the network for potential changes at all levels when specific components or links are deliberately perturbed.
b. To enrich the quality of the computational framework by generating information related to specific perturbations known to selectively modify insulin signalling and/or mitochondria function.
c. To determine the predictability of the computational framework by validating specific hypothesis emerging from the developing model to be tested using appropriate in vitro and in vivo model.
d. To evaluate the translatability of the new knowledge to human diabetes or related diseases.
Project Context and Objectives:
PROJECT CONTEXT.
A common trait to disorders such as type 2 diabetes or obesity is the development of insulin resistance. The link between insulin signalling and these diseases is unclear but there is growing evidence indicating the existence of a bidirectional cross-talk between insulin signalling and mitochondrial function that may be relevant for the pathogenesis of these diseases. Based on this evidence, MITIN main goal was to identify novel mitochondrial-dependent mechanisms responsible for the development of insulin resistance. This was done by the use of technologies of Systems Biology and the generation of computer-based tools that will permit the study of complex biological systems that integrate different regulatory networks.
For the global visualization and interpretation of the “insulin signalling/mitochondria” complex system, our intention was to develop a computational framework that will store and integrate all possible data for each of the subsystems, both currently available data and data generated within the project by transcriptomics and lipidomics analysis. This would allow the prediction of functional associations and interactions between both processes, which were tested under specific hypothesis-driven studies in mammalian cells, mice and Drosophila. Furthermore, with this network, we simulate different complex metabolic and regulatory scenarios for normal conditions and diseases, and identify and predict possible changes when other parts of the system are deliberately perturbed.
The integrated model of “insulin signalling/mitochondria” were progressively refined as the new data generated within the project and integrated into the computational framework contributed to its enrichment. Thus we analyzed experimentally the effect of specific perturbations in the insulin-signalling pathway on mitochondrial function. The opposite paradigm were also investigated, so that the impact of primary defects in mitochondria function on insulin signalling were characterized. We took advantage of cross-species information by investigating these paradigms in mammalian cells and mice but also in Drosophila. By using a Systems Biology approach and investigating how different perturbations modulate cell transcriptomes, and lipidomes, we aim to identify integrated homeostatic responses involving the insulin signalling network and the mitochondria networks.
Validation of some of the strongest associations detected by virtue of the integrated systems biology of insulin signalling/mitochondria will generate high value targets of use in therapy against Complex Diseases such as diabetes, obesity, or pathophysiological traits of the Metabolic Syndrome. Besides, computer-based tools generated in the project will be applicable to the analysis of the mechanisms that trigger other Complex Diseases.
The development of new profiling technologies has generated an unprecedented opportunity to generate huge amount of heterogeneous data. This opened great expectations about the possibility of tackling Complex Diseases as well has generated new technological challenges such as integrating these databases and networks into meaningful models. For these, new computational tools need to be developed and experience gained by applying them to specific biomedical problems.
In this project we have explored how/if a Systems Biological approach is a good strategy to unravel a complex biomedical problem. The problem selected is insulin resistant diabetes and its interaction with mitochondrial function. This is an important medical problem given its epidemic proportions, its future prospects of resulting in a subsequent epidemic of cardiovascular events, its obvious financial implications and, more importantly, because it is cause of human suffering and death.
Defects in insulin action are commonly observed in a variety of different Complex Diseases or syndromes such as type 2 diabetes, obesity, hypertension or cardiovascular disease. In this regard, there is growing evidence indicating the existence of a bidirectional cross-talk between insulin signalling and mitochondrial function. Thus, alterations in mitochondrial function may trigger the development of insulin resistance in tissues and, in addition, insulin signalling has a direct impact on mitochondrial function.
Based on these observations and the level of their complexity, we have developed a Systems Biology approach to determine whether this research strategy is useful to provide a global insight on the molecular mechanisms linking both biological systems, i.e. insulin signalling and mitochondrial function. Specifically we used a multidisciplinary approach that will, on one hand generate and identify a wide range of available quantitative biological data and, on the other hand, develop computational tools to integrate this information into integrated models reflecting the nodes of interaction between insulin signaling and mitochondrial function.
To enhance the power of this approach we took advantage of data obtained from different levels of organization (cultured cells and in the in vivo animal) and in two different organisms (Drosophila and mice). Transcriptomics and lipidomics analyses will permit to identify homeostatic responses affecting insulin signalling and mitochondrial systems. The reasons by which we focus in this proposal on two omics (lipidomics and transcriptomics) are straightforward. Lipid metabolism is markedly modified by insulin or by changes in mitochondrial function, and, in turn, lipids are known to impact on insulin sensitivity. In addition, both insulin and changes in mitochondrial function modify the gene expression profile in cells or tissues, and in fact, insulin modifies the expression of genes encoding for mitochondrial proteins and conditions of insulin resistance show a deficient expression of genes encoding for subunit of the OXPHOS system, which parallel deficient mitochondrial activity.
Validation of some of the strongest associations detected by virtue of the integrated systems “insulin signaling/mitochondria” will generate high value targets that may be used for future development of pharmacological intervention strategies. Besides, generation of tools for the integration of Systems Biology will be of application to better understanding other Complex Diseases.
OBJECTIVES.
The main goal of MITIN was to elucidate the interplay between the insulin signaling pathway and the mitochondrial function through the combination of computational tools and molecular omics technologies in a multidisciplinary, integrated system biological approach.
MITIN scientific objectives were based on research evidence on the cross-link between insulin signaling and mitochondrial function (in which partners of the consortium have taken part). The project aimed at capitalising this knowledge and will pave the way for the translation of research into relevant applications.
The specific main objectives of the project were the following:
a) To develop a computational framework that integrates, as an interaction network, the already available, and the newly generated or predicted data on all components (gene expression, proteins or metabolites) of the Insulin signalling pathway and the Mitochondrial processes, and the relationships within and between both systems. This framework mainly consisted in a complex network that was designed focusing on several important features, from which we want to emphasize: [1] the possibility of handling different types and levels of relationships (links) that connect components within and between both systems (from direct protein-protein interaction to common cellular location); and [2] the possibility of querying the network for potential changes at all levels when specific components or links are deliberately perturbed.
b) To enrich the quality of the computational framework by generating information related to specific perturbations known to selectively modify insulin signalling and/or mitochondria function.
c) To determine the predictability of the computational framework by validating specific hypothesis emerging from the developing model to be tested using appropriate in vitro and in vivo model.
d) To evaluate the translatability of the new knowledge to human diabetes or related diseases.
Project Results:
The major work performed during the operation of MITIN has been the following:
a) The development and validation of the computational framework through retrieval of available information and the integration of data into a network.
b) The development and application of analytical protocols for metabolic profiling of mitochondria.
c) The analysis of the impact that mitochondrial dysfunction has on insulin sensitivity.
d) The analysis of the effect of alterations of insulin sensitivity on mitochondrial function.
a) The development and validation of the computational framework through retrieval of available information and the integration of data into a network.
We have constructed a consensus insulin pathway from several public resources, including (Biocarta; www.biocarta.com Kegg; www.genome.jp/kegg/ Reactome; www.reactome.org and PID; http://pid.nci.nih.gov/) and a commercial resource (Biobase; www.biobase.de). This database was manually curated and refined by the participation of different partners of the consortium.
In order to select the parts lists that compose mitochondrial proteins or genes, we selected a total set of 900 proteins from mitoP2 database (www.mitop2.de/) and organized them according the biological process, mitochondrial compartment and molecular function. Like done for the insulin pathway, the set was manually curated by the participation of the expert groups in the consortium.
Besides human tissues and cells, the consortium aimed to investigate with both mice and drosophila melanogaster. We identified each mouse and fly orthologous gene/protein for all involved proteins. Orthologous genes were selected by the best one to one bidirectional blast hit between the three organisms.
One of the first steps towards the main objectives of this workpackage was to build a complete and manually curated list of the proteins involved in the insulin pathway and the mitochondrial function, and connect them in order to better understand how these two systems are functionally related. In that sense, while the data from perturbation models within the consortium is being generated, we have taken advantage of the publicly available data from several databases and "omics" experiments to construct a first version of these systems network.
The first version of the network contains both protein-protein interaction data, and gene co-expression relationships.
To identify genes that participate in the crosstalk between insulin and mitochondria, we have looked for protein-protein interactions. With that objective in mind, we have used a non-redundant set of 23 protein interaction datasets. These interactions had been mapped to expression data available, in order to estimate whether each interaction is likely to occur in each given tissue of interest (Bossi and Lehner 2009). According to these data, we have studied interactions possible to occur in liver, cardiac myocites, skeletal muscle, smooth muscle, and adipose tissue.
The first level of evidence of connection between the two systems are those proteins that appear both in the insulin signalling pathway and that are described to form part of the biology of the mitochondria. These proteins are encoded by the following genes: BAD, PRKAR2B, MAPK9, ARAF, GCK, YWHAE, RAF1, SHC1, YWHAZ, PPP1CC.
When exploring the direct interactions between mitochondria and insulin genes, we found seven insulin proteins that interact at least with two mitochondrial proteins (each of these interactions reported by two independent publications), and eight mitochondrial proteins that interact at least with two insulin proteins (each of these interactions reported by two independent publications).
We also explored the indirect interactions between mitochondria and insulin using one protein as intermediary (internode). That way, we identified 34 proteins that interact at least 3 times in total with mitochondria and/or insulin. After excluding those internodes that could have occurred by chance, we ended up with a total of 29 high confidence internodes.
Taking all these interactions into account, we could construct a preliminary MITIN general network, based on manually curated insulin and mitochondrial systems and highly stringent physical protein-protein interactions information. Taking into account the tissue expression of each of the nodes (genes) that participate in each of the interactions, a network for each of the tissues of interest including liver, fetal liver, cardiac myocytes, adipocytes, smooth and skeletal muscle, has been constructed.
Together with that, we have developed a method to calculate a signal-to-noise ratio of each observed interaction, using random permutation tests, assigning a p-value for each of the links observed.
Among all these high confidence interactions, some of them make the results promising, as the network allowed us to identify previously known relationships, such as the relationship between EGFR and insulin signalling (Borisov et al. 2009), or the connection between BCL2, MAPK1 and MAPK2 genes and insulin signalling. Regarding the internodes, ANXA2, RPS6KA1, and APP genes, among many others, have previously shown to have a functional connection between both MITIN systems, as reviewed by several members of partner 1 (IRB Barcelona). A manual inspection an literature search performed by the investigators of partner 1 (IRB Barcelona) showed that there was a functional known connection for around 75% of the internodes. Some of these predicted internodes have been selected to be experimentally validated at partner 1 (IRB Barcelona) laboratory. Although for the remaining 25 % of candidates there is no previous known evidence of connection between insulin and mitochondria, these candidates should better be considered as potential novel candidates rather than false positives.
As a second source of links that interact between insulin and mitochondrial function, we analyzed gene co-expression networks (GCN). GCN can be calculated from microarray datasets from which a sufficient amount of samples exists. Genes that are correlated across different samples (termed co-expressed), can be classified as pertaining to a common module and are likely to perform a similar function in the studied tissue. For the gene co-expression network, we have used the dataset of Schadt et. Al (Schadt et al 2008), which consists of expression data of 400 healthy human liver samples. We used WGCNA R package (Langfelder and Horvath 2008) to calculate the weighed gene co-expression network as well as identifying modules of co-expressed genes. The identification of modules that contain a significant enrichment of both insulin signaling and mitochondrial function genes will allow us to identify new candidate players between both systems. In addition, the co-expression interactions that overlap with protein-protein interaction findings will be used to positively score those relations.
All the computational predictions obtained during this activity and all the data that will be generated by the different members of the consortium should be integrated into a common, easily accessible, and flexible database. For that purpose we have integrated all the available data into a mySQL relational database. The MITIN database is structured in three main compartments. The main compartment is centered in Homo Sapiens data and contains most of the information. Within this compartment, there is data about mitochondria, insulin, and internodes as obtained by computational analysis. In addition several gene nomenclature and synonyms databases will allow performing queries coming from different sources. As one of the sources of data that will be needed to be integrated will come from microarray data, several probe-to-gene annotation tables have been integrated into the database. Finally, as one of the final goals of building the MITIN network is to identify new participants in the aetiology of Type 2 Diabetes and its related diseases, data on genes involved in mendelian diseases related to metabolism
(OMIM, http://www.ncbi.nlm.nih.gov/omim) and genetic variation associated to Type 2 Diabetes and related complex diseases (GWAS database, Hindorff et al 2009) has been included into the MITIN mySQL database. Therefore, the relational database allows us to query whether the MITIN network genes are related to both mendelian and complex diseases related to Type 2 Diabetes. This structured data will allow the integration of omics data coming from different partners, collaborators and clinicians.
The main human compartment of the mySQL database is connected through orthology to mouse and drosophila data. Both mouse and drosophila data contain several gene nomenclature and synonyms database to allow data coming from different resources to be integrated. This database allows us to query a set of genes that are transferred through orthology in the case of data coming from mouse or drosophila, to interrogate whether these genes have been previously implicated in both mendelian and complex diseases, as well as to interrogate functional genetic variation that might be contained within each of these genes.
Manually curated review of the candidates and selection of candidates.
We have performed a thorough review of the mitochondrial and insulin signalling proteins that functionally interact in mammalian cells. As a result, we have an updated view on the mitochondrial and insulin signaling proteins already known to physically and functionally interact in mammalian cells.
Among all the high confidence interactions detected, some of them make the results promising, as the network allowed us to identify previously known relationships, such as the relationship between EGFR and insulin signalling, or the connection between BCL2, MAPK1, MAPK2 and MAPK8 genes and insulin signaling (Figure 1). Regarding the internodes, ANXA2, RPS6KA1, and APP genes, among many others, have previously shown to have a functional connection between both MITIN systems, as reviewed by several members of partner 1 (IRB Barcelona). A manual inspection of a literature search performed by the investigators of partner 1 (IRB Barcelona) showed that there was a functional known connection for around 75% of the internodes.
Some of these predicted internodes have been selected to be experimentally validated at partner 1 (IRB Barcelona) laboratory. Although for the remaining 25 % of candidates there is no previous known evidence of connection between insulin and mitochondria, these candidates should better be considered as potential novel candidates rather than false positives. The signal-to-ratio analysis propose BRAF, ANXA2, MAPK8IP3, RPS6KA1, ANXA2P2, PPP2CA, CRK, TSC2, and IGF1R, as promising candidate internodal proteins to be evaluated experimentally. The disadvantage of these options is that most of them are not novel as candidates linking mitochondria and insulin signaling.
Construction of the MITIN network (second version of the computational framework after operation at a lower stringency).
To establish the molecular links between the insulin and the mitochondrial systems, we used the Global Network that we generated to identify different types of relationships between the insulin and mitochondrial systems: direct interactions -genes from mitochondria or the insulin signaling pathway whose products either physically or functionally interact directly with at least one protein of the other system-, and indirect interactions, that is, allowing one intermediate gene, not belonging to the mitochondria nor to the insulin pathway but connected to at least one insulin and one mitochondria gene. These genes that functionally bridge these two systems will be termed from now on internodes or MITIN linker genes. The MITIN network consisted of a total 886 genes (nodes), 65 direct interactions and 1194 indirect interactions triggered by 286 linker genes. All these linker proteins were connected to a mean number of 2.1 Insulin genes and 1.7 mitochondria. When taking into account the number of functional evidences (regardless the number of nodes) between both systems the linker genes showed a mean of 2.6 functional evidences linking insulin pathway while a mean of 2.0 evidences linking the mitochondria genes. Figure 2 shows a high confident subset of linker genes that show at least three evidences linking insulin signaling and mitochondria genes simultaneously.
The same network also allowed us to define which mitochondrial genes are more connected to insulin, and vice-versa, and either directly or indirectly. The top five insulin genes most connected to mitochondria insulin genes were nucleolar and coiled-body phosphoprotein 1 (NOLC1), ribosomal protein S6 (RPS6), inhibitor of kappa light polypeptide gene enhancer in B-cells, kinase beta (IKBKB), pyruvate kinase, liver and RBC (PKLR), and v-src sarcoma (Schmidt-Ruppin A-2) viral oncogene homolog (avian) (SRC), with a total of 99, 40, 31, 28 and 22 indirect connections with mitochondria, respectively. Regarding mitochondria, the five most connected gens with insulin were Tu translation elongation factor, mitochondrial (TUFM), tumor protein 53 (TP53), solute carrier family 25 (mitochondrial carrier; adenine nucleotide translocator), member 5 (SLC25A5), polymerase (DNA directed), gamma (POLG), and estrogen receptor 1 (ESR1), with a total of 93, 36, 29, 25, and 19 indirect connections with insulin, respectively.
While a majority of the internodes are considered novel, as they have not previously been described as having a bridging role connecting both systems, some of them have already been shown to interact with both systems. For example, TRAF2 is connected to four insulin genes and two mitochondria genes, according to our analysis. Indeed, TRAF2 has been shown to be connected to MAP3K1 (MEKK1) and CAV1 (caveolin-1) insulin genes, and to MAP3K5 (ASK1) and CASP8 (caspase-8) mitochondrial genes. A possible connection to MTOR (mTOR) has also to be considered. Another example is NFKB1, showing interactions with four insulin signaling genes and three mitochondria proteins, according to our analyses.
We have validated the relevance of one specific gene, i.e. MAPK8 (JNK) as a linker between mitochondrial function and insulin signaling. In this regard, the following studies were performed: a) Mfn2 deficiency activates JNK in muscle cells; b) Mfn2 deficiency causes activation of JNK in conjunction with mitochondrial dysfunction, enhanced ROS production and ER stress in liver and muscle; c) Chronic treatment with TUDCA or N-acetylcysteine normalize JNK activity, insulin signaling and glucose tolerance in liver-specific Mfn2 knockout mice.
b) The development and application of analytical protocols for metabolic profiling of mitochondria.
The methods for primary metabolites (LC/MS/MS), global lipidomics (UPLC/MS), and global metabolomics (GCxGC-TOF/MS), and branched chain amino acids and their keto acids (HPLC) are available for analysis.
Additionally, we have used time under this task for improving the experimental protocols for tissues and cells, as well as improving the identification of lipids (using internal MS/MS libraries). Simultaneously, we have been developing the methods for analyzing the acylcarnitines and acyl-CoA:s (UPLC/MS/MS), and Cardiolipins (by UPLC/MS/MS).
c) The analysis of the impact that mitochondrial dysfunction has on insulin sensitivity.
Summary of progress
The specific objectives of this workpackage were:
[1] To study the effect of disruption of mitochondrial dynamics on insulin signalling in cultured cells, in mice and in Drosophila melanogaster;
[2] To analyse the mechanisms that control insulin sensitivity induced by selective defects in either PGC-1alpha or PGC-1beta.
As part of the first objective, partner I (IRB) has developed a battery of different loss and gain of function in vitro models to assess the metabolic impact of Mfn2, a protein involved in mitochondrial dynamics in insulin action in muscle cell model as L6E9 or C2C12 myotubes. In addition, to further study the effects of the aforementioned proteins involved in mitochondrial fusion in a more complex organism, partner 1 (IRB) also generated both skeletal and liver tissue specific KO mouse model for Mfn2. Metabolic profiling of both models have been extensively analysed. Lipidomics of specific key models have been generated and analysed by Partner 3 (VTT).
Partner 6 (DKFZ) has generated KO flies for marf (Mfn2) and fzo (Mfn1) to study the impact of impaired mitochondrial function on insulin signalling cascade under different nutritional stressors (palmitic acid). In addition to this, they have identified C69009 as a mitochondrial gene with homology to PGC1a that is regulated by insulin. Lipidome analyses have been performed by Partner 3 (VTT).
The second objective involved the global physiological characterization of rodent models As part of the second objective, aiming to establish the role of PGC1alpha/ PGC1beta dysfunction on the mechanisms controlling insulin sensitivity, partner 5 (UCAM) has developed and performed a battery of experimental models aimed to challenge energy homeostasis with particular interest in carbohydrate metabolism.
The experimental approach used involved selected nutritional, environmental and genetic interventions aimed to challenge the mitochondrial function resulting from total ablation or heterozygosis of PGC1s. Briefly these challenges involved chronic nutritional interventions with high fat diet (45% kCal from saturated fat) as well as high protein diet (40% kCal from protein) for some of the rodent models described below. As part of the genetic challenges Partner 5 (UCAM) generated the Pgc1beta x ob/ob model as a paradigm model of mitochondrial dysfunction in a context of massive energy surplus elicited by ob/ob. Moreover, we also created the double heterozygote PGC1beta x PGC1 alpha. In addition, the PGC1beta transgenic model to perform some specific comparison respect to the other models.
Partner 5 have also performed an extensive phenotypical characterization as a part pf our integrative physiology programme that includes analysis of energy expenditure, carbohydrate metabolism. We have also performed as part of the ex vivo profiling, gene and protein expression analysis, serum biochemistry measurements and histological analysis of key organs that includes adipose tissue, liver, pancreas. In addition, we have also produced percoll purified mitochondrial fractions for specific rodent models in liver and skeletal muscle to characterize the effects of PGC1s ablation on the mitochondrial lipidome (task performed by partner 3-VTT). Finally we have also produced gene expression data that has been used for the model generated by partner 2-BSC-CNS.
Significant results
As part of the in vitro studies in myotubes and myoblasts, Partner 1 showed that deletion of proteins involved in mitochondrial fusion and fission impaired myogenic processes affecting mitochondrial morphology and function (e.g. Mfn1, Mfn2, OPA1 knowdown, Drp1 knock-in and Drp1 overexpression) linked in some specific cases with impairment of insulin signalling cascade at different nodes. Microarray analysis of Mfn2 deletion as well as overexpression of drp1 (native and dominant negative form) revealed a list of relevant candidate genes to explore further. Moreover partner 1 also showed that the Mfn2 ablation induces endoplasmic reticulum vacuolization and unfolded protein response in cells treated with ER stress inducers. Microarray analysis of Mfn2 loss of function and drp gain and loss of function models revealed substantial differences in the identity of pathways dysregulated as well as the severity of the impact.
In drosophila, partner 6 has demonstrated that Marf KO (Mfn2) development and life is severely compromised with clear evidence of mitochondrial defects. The heterozygote showed alterations in mitochondrial morphology and impairment in mitochondrial function and number as well as a reduction in oxygen consumption that is consistent with increased TG levels and impairment in insulin signalling (in presence of palmitic acid as equivalent to HFD in rodents) and resembles the data generated in rodent models as explained in detail below) surprisingly, these heterozygotes also show a significative increase in the life expand respect to wt. Microarray analysis suggest alterations in carbohydrate metabolism. Moreover, the heterozyote for fzo (Mfn1) showed a substantial increase in fat content (elevated Tg levels) in combination with hypoglycaemia.
In vivo experiments using the rodent models aforementioned Partner 1 (IRB) has demonstrated that:
a)Ablation of Mfn2 causes marked mitochondrial dysfunction at multiple levels and insulin resistance in skeletal muscle. Interestingly, conditional KO mice for mfn2 in skeletal muscle showed impaired glucose tolerance and hyperinsulinemia when fed with HFD. Moreover, ablation of Mfn2 in muscle caused a severe impairment in the insulin signalling cascade in line with the data observed in vitro.
b) Specific ablation of Mfn2 in liver causes marked hepatic glucose production, glucose intolerance in both chow and HFD fed mice and severe insulin resistance after being challenged by HFD. At molecular level, the livers of these mice showed a severe impairment in insulin signalling cascade as well as evidences of endoplasmic reticulum stress and altered ER morphology and calcium homeostasis. Lipidomics analysis of these mice showed profound alterations in the lipidome with marked increased in triglycerides and decrease in phospholipid species consistent with the presence of hepatic steatosis.
Partner 5 (UCAM) highlights the following results regarding specific PGC1 ablated models used:
a) the ablation of PGC1beta has a positive impact on carbohydrate metabolism (increased glucose tolerance) in aged mice, profile accompanied by reduction in fat percentage and liver mass despite confirmed mitochondrial dysfunction and decreased global energy expenditure in aged mice (over 15 months old).
b) The ablation of PGC1beta in the context of massive energy surplus depicted in the leptin deficient mouse ob/ob cause a significant amelioration of the hyperglycemia normally observed in the ob/ob mice consistent with a mild increase in fasting insulin levels. Nevertheless, this picture is accompanied by increased markers of micro and macrosteatosis that may indicate liver failure. Paradoxically, the increase in hepatic fat content may act as a buffer protecting other organs such pancreas. Decreased energy expenditure in double ko respect ob/ob littermates specially during day phase is consistent with alterations in carbohydrate metabolism after o/n fasting.
c) The reduction in the allelic expression of both PGC1beta and PGC1 alpha (DHET) produced a more insulin sensitive mouse with higher energy expenditure suggestive of a hypermetabolic state consistent with the activation of thermogenic markers in brown adipose tissue and the leaner phenotype observed. DHET also protected against fatty liver imposed by HFD.
d) Preliminary analysis of global lipidomics, acylcarnitines and cardiolipins (performed by partner 3) reveals changes in the lipidome of key metabolic organs as well as isolated mitochondria in PGC1 deficient mice.
d) The analysis of the effect of alterations of insulin sensitivity on mitochondrial function.
Using a lentivirus shRNA-mediated approach we successfully generated myocyte cell lines with deficient CAP expression, in a cardiac (H9C2) and a muscular (C2C12) cells. Immunoblotting of lysates obtained from puromycin selected individual clonal cells showed that the inhibition of CAP protein in these cells was greater than 95%. Under these conditions, the cells were able to grow and differentiate into myotubes as well as control cells. The H9C2 CAP deficient cells exhibited a reduction of basal mitochondrial membrane potential compared to control cells, a reduced glucose uptake in response to insulin stimulation while there were no changes in the expression of insulin signaling proteins, such as insulin receptor, p85 regulatory subunit of PI3Kinase, AKT or ERK1/2. A slight decrease of Cbl expression was also observed in the CAP deficient cells. Insulin signaling through PI 3Kinase and AKT phosphorylation in response to insulin was normal. We observed however an elevated basal activation of MAPK pathway (p42/p44) in the CAP deficient H9C2 cells compared to controls. We have also examined the phenotype of CAP depletion in the muscle cell line C2C12. Similarly to the results found in H9C2 cells, depletion of CAP did not impair differentiation of CC12 cells into myotubes. Insulin mediated phosphorylation of AKT was normal in these cells as well as activation of the MAPK cascasde (p42/p44). Unlike the H9C2 we did not detected an enhanced basal activation of the MAPK pathway in these cells. Further characterization of the CAP gene depletion was performed by examining the gene expression profiles of these cells by quantitative real time PCR. We found changes in the mRNA levels of the gene coding for the coactivator PGC1-alpha, but not PGC1-beta, or the transcription factors NRF1 or TFAM. No statistically significant changes were observed in the level of expression of the mRNA coding for mitochondrial proteins, COX2, COX4 or ATP synthase, however there was a trend for COX2 to be reduced in the CAP depleted cells. We found a statistically significant reduction of the gene coding for UCP2 in the CAP depleted C2C12 cells compared to controls. We did not detect any changes in the protein content of the mitochondrial proteins Cox1 or Porin. The amount of Mitofusin2 detected in these cells was also similar to controls.
We have also achieved an efficient knockdown of c-Cbl. Analysis of these cells revealed no major changes in the Phosphatydilinositol 3 kinase/AKT or MAPK pathways by insulin. Activation of the stress kinase JNK was also comparable to that observed in control cells. The mitochondrial membrane potential of the Cbl Knocked down cells wass elevated compared to control cells. Cbl depleted C2C12 cells show increased mitochondrial membrane potential compared to control and reduced ROS production. ATP content was similar to that of control cells, whereas citrate synthase activity was increased in both Cbl and CAP depleted cells compared to control cells. Cbl depleted but not CAP depleted cells also display increased activation of AMPK and ACC.
We also generated a cell line with impaired signaling throught the PI 3 Kinase pathway following incubation of the cells with insulin for 48 hrs. This strategy sought to recreate a model described in the literature that has been previously utilized to downregulate the expression and signaling through the insulin receptor. As expected, we observed that the insulin receptor expression was reduced approximately 50% in the insulin treated cells. Insulin mediated activation of AKT (via PI3 Kinase) was significantly reduced in these cells, whereas activation of the ERK1/2 enzymes (MAPK cascade) was only decreased by approximately 10%. Under these conditions, we observed that the mitochondrial membrane potential decreased about 50% in respect of control cells (no insulin treated). However measurements of ATP were not significantly different than control cells. Consistent with this, gene expression profiling using quantitative real time PCR indicated that the mRNA levels for subunit of the ATP synthase were no statistically different than those found in control cells. However we detected a tendency for a decrease in the amount of mRNA for PGC-1alpha, COX2 and COX4 in the chronically treated cells. There was a marked reduction in the amount of Uncoupling protein (UCP2) whereas we did not detect changes for the mRNA for PGC-1 beta, NRF1 or Tfam gene expression. The protein content of the mitochondrial proteins Cox1, Porin and Mitofusin 2 were also unaffected. In addition, ROS production over time was measured as quantitative DCF-DA fluorescence and was slightly lower in these cells respective to controls. A complete gene profile analysis using Affymetrix Microarray technology has been carried out to determine the molecular mechanisms by which insulin carries out these effects.
Potential Impact:
During the operation of the MITIN consortium we have interacted with clinical scientists in order to analyze the different pathways to validate the data generated by MITIN. The general conclusion we have reached is that at this point a realistic possibility is to undertake an extensive collaboration with clinical scientists to demonstrate the implication of the genes identified in the MITIN project in the acquisition of clinically-relevant traits within the scope of the Metabolic Syndrome.
Discussions maintained with different clinical investigators with expertise in the Metabolic Syndrome permit to conclude that an excellent approach is to genotype multiple informative SNPs in all the genes identified by the MITIN consortium, and to analyze whether the genotypes associate to relevant parameters measured in the cohorts. This approach should be done in parallel with researchers holding different well phenotyped cohorts. We will select collaboration with clinical scientists having access to cohorts showing a number of properties, namely, that they are extremely well phenotyped and for which they have collected plasma, DNA and subcutaneous and visceral adipose tissue and they have measured the expression of a substantial number of genes, and plasma biochemistry.