Subgroup Leader : Gabriel MALOUF
Regulation of gene expression in cancer
Regulation of gene expression in cancer
For the past 30 years, we have been studying the molecular basis of regulation of gene expression. The group worked for over 25 years on the structure and function of the mammalian general transcription factor TFIID, its subunit TBP (TATA-binding protein TBP) as well as male germ cell expressed paralogs of TFIID component such as TBP-related factor 2 (TRF2) or TAF7L, publishing more than 50 papers on these subjects.
In 2008, the major focus of the group moved from the study of the general transcription factor TFIID to investigating regulation of gene expression and epigenetics in malignant melanoma. We focussed our attention initially on the role of transcription factors MITF and SOX10 in melanoma cell proliferation, identifying their target genes including long non-coding RNAs (lncRNAs) that are now a major focus of attention, and using mass-spectrometry to identify their potential co-factors. Amongst the MITF and SOX10 co-factors that we identified were the chromatin remodelling complexes SWI/SNF, NuRF and NuRD. We defined the role of BRG1, the ATPase subunit of the SWI/SNF PBAF complex, in human melanoma cells in culture showing how it is recruited by MITF and/or SOX10 to cis-regulatory elements driving expression of genes involved in melanoma cell proliferation. We defined the role of the BPTF scaffolding subunit of NuRF in human melanoma cells in culture and in the mouse melanocyte lineage in vivo where we identified its critical role in reactivation of the melanocyte gene expression program in adult melanocyte stem cells during anagen phase. We further defined the critical roles of BRG1 and BPTF as MITF and SOX10 co-factors in mouse melanoma in vivo. Investigation of the role of the CHD4 ATPase subunit of the NuRD complex led us in to the field of cancer cell metabolism. We defined a novel mechanism by which citrullination of the key glycolytic enzyme PKM2 regulates glycolysis and cell proliferation in a variety of cancer cells.
A full list of publications can be found here: https://orcid.org/0000-0001-5533-1171
1). The TFIID subunit TAF4.
Proper regulation of gene expression requires a complex and dynamic dialogue between transcription factors binding cis-regulatory elements, chromatin remodelling and modification complexes and the basal transcription machinery. The multi-subunit basal transcription complex TFIID plays a critical role in this communication. TFIID comprises the TATA-box binding protein (TBP) and 13-14 TBP-associated factors (TAFs) and plays a unique role in pre-initiation complex (PIC) formation. Recent electron microscopy studies showed that TFIID is organised in three lobes A-C that undergo topological reorganisation during the initial steps of PIC formation as it binds promoter DNA along with TFIIA and TFIIB. The histone-like TAF4-TAF12 heterodimer is crucial for the structural integrity of lobes A and B. In lobe B, a region of the conserved TAF4 C-terminal domain makes contacts with promoter DNA and both TAF4 and TAF12 contact the TFIIA-TBP module suggesting that they promote TBP DNA binding and fix the distance between the TBP binding site and the transcription start site.
Genetic knockout of Taf4 in mouse, as is the case with several other TAFs, leads to highly specific effects on PIC formation and gene expression that may in part be explained by redundancy with its paralog TAF4B that also heterodimerizes with TAF12 and integrates into TFIID, thus maintaining TFIID integrity in absence of TAF4. We used genetic inactivation to address the function of murine Taf4 in embryonic and adult murine epidermis, neonatal hepatocytes and adult pancreatic beta cells.
We found that TAF4 interacts with the transcription factor HNF4A acting as an essential co-factor for activation of post-natal metabolism genes in neo-natal hepatocytes. TAF4 inactivation in neonatal hepatocytes not only blocked PIC formation at the promoters of these genes, but also led to loss of HNF4A binding to its cognate sites in the promoters of these genes. TAF4 thus coordinates HNF4A genome occupancy to PIC formation and activation of gene expression in neonatal hepatocytes.
Alpern et al., doi: 10.7554/eLife.03613. https://elifesciences.org/articles/03613
As HNF4A also plays an important role in pancreatic beta cells, we selectively inactivated TAF4 in adult pancreatic beta cells. Rather than finding a role as a HNF4A we found an essential role for TAF4 in beta cell function and the maintenance of their identity. TAF4 inactivation led to increased glycaemia due to defective insulin signalling and secretion, a consequence of an immediate and major impact on gene expression. Single cell (sc)RNA-seq showed that 1 week after TAF4 inactivation cells with mixed beta and alpha identities were observed in a process of stress-induced beta cell trans-differentiation into glucagon-expressing alpha-like cells. Computational analysis of the scRNA-seq proposed a model for beta cell trans-differentiation into glucagon expressing alpha-like cells involving a progressive loss of beta cell identity associated with gain of activity for a series of aberrantly expressed transcription factors and a key role for the NuRF chromatin remodelling complex.
Kleiber et al., doi: 10.1038/s41419-021-04067-y. https://www.nature.com/articles/s41419-021-04067-y
We also showed that Taf4 is dispensable for preimplantation development in vivo and embryonic stem cell (ESC) viability. However, Taf4 plays an essential role in ESC differentiation as neurons or cardiomyocytes. Taf4 was required for de novo PIC formation at promoters of genes expressed late in the neurogenic program leading to an arrest of differentiation. Interestingly, Pou5f1-expressing PGCs were formed in Taf4-/- embryos. Generation of PGCs involves complex transcriptional and epigenetic reprogramming events including reactivation of pluripotency genes as well as changes in histone and DNA methylation. Taf4 is therefore not required for this naturally induced reprogramming, an observation that contrasts with a reported role of Taf4 in maintaining ESC pluripotency and promoting induced pluripotent stem cell (iPS) reprogramming in vitro.
Langer et al., doi: 10.1038/ncomms11063. https://www.nature.com/articles/ncomms11063
2). Gene regulation, epigenetics and metabolism in melanoma.
Melanoma is the most malignant skin cancer. Despite the introduction of MAP kinase inhibitor (MAPKi) and immune checkpoint (ICI) therapies, a significant fraction of patients is still considered as non-responders or develop resistance.
Heterogeneity in melanoma is a major obstacle to efficient therapy. Much effort has been devoted by the field to understand the molecular mechanisms, both genetic and epigenetic, driving resistance to MAPKi and ICI therapy. Melanoma tumours comprise a variety of cell types characterized by specific gene expression signatures, proliferative, invasive and drug resistant properties.
While, naïve tumours comprise a large majority of melanocytic type melanoma cells, drug and ICI treatment lead to the appearance of other cell states notably the neural crest stem cell like-state (NCSC) and the undifferentiated/mesenchymal state. Both of these cell states are involved in minimal residual disease and acquisition of MAPKi resistance. Each cell state is driven by a unique set of transcription factors, with MITF and SOX10 playing pivotal roles in the melanocytic state, SOX10 in the neural crest like-state, while in the undifferentiated/mesenchymal state, several factors have been shown to be important including; SOX9, TEAD/AP1, ZEB1 and GATA6.
Melanoma Heterogeneity and tumour evoluation.
We used single cell RT-qPCR to characterize melanoma cells grown under different culture conditions and grown as CDX in vivo with a collection of 160 genes that defined the melanocytic and undifferentiated cell states. This approach highlighted how cell heterogeneity was modulated by culture conditions and the CDX microenvironment in vivo.
Ennen et al., doi: 10.1038/onc.2014.262. https://www.nature.com/articles/onc2014262
Molecular analysis of primary human melanoma is limited by the necessity to preserve intact biopsies for histopathology analysis. As single-cell gene expression analyses require only small numbers of cells, these can be isolated as microbiopsies without compromising subsequent histologic evaluation. Using single cell RT-qPCR coupled with RNAscope in situ hybridization we provided the first in depth analyses of gene expression and inter and intra-tumoral heterogeneity of primary melanoma lesions. We identified not only melanocytic and mesenchymal state cells but also a new population expressing markers of both cell states. RNAscope also illustrated how cells in these different states could be organized either as distinct regions or intermingled amongst each other.
Ennen et al. doi: 10.1158/1078-0432.CCR-17-0010. https://clincancerres.aacrjournals.org/content/23/22/7097.long
In a related study, we took advantage of a unique atypical melanoma case history to define how the genetic make-up of melanoma tumours can evolve over time. We analysed the genetics of multiple melanoma tumours occurring in an individual at Strasbourg University Hospital dermatology clinic taken over a 6–7-year period. The individual developed around 82 surgically removed epidermotropic melanoma lesions without development of extra-cutaneous metastases. However, after almost 8 years, the individual developed metastasis in the left supraclavicular lymph nodes and finally in the brain that were un-responsive immune checkpoint therapy. Whole exome sequencing analyses of two early and two late-stage tumours allowed us to reconstruct the nature history of the disease and the polyclonal relationships amongst the tumours along with its genetic evaluation. Contrary to expectations, tumour evolution over a prolonged period was associated with selection of subclones with fewer mutations and reduced neoantigen load and we found no BRAFV600E or NRAS mutations or other mutations commonly found in melanoma. Later tumours therefore displayed lower neoantigen burden compared to early tumours suggesting that clonal evolution was at least in part driven by counter selection of subclones with high neoantigen burdens. Our analysis indicated that the disease course may in large part be explained by elimination of subclones with high mutational and neoantigen burden favouring emergence of subclones with low mutational and neoantigen burden. Negative selection of subclones with high neoantigen burden is further consistent with the reduction in tumour infiltrating immune cells in the latter lesions. We identified neoantigen mediated clonal selection as the driver of the natural disease history of this individual why the individual poorly responded to immune checkpoint therapy when subclones with low neoantigen burden eventually gave rise to lymph node and visceral metastases. Moreover, we also identified a rather unique combination of driver mutations in these tumours involving an AKAP9-BRAF fusion and a MAP2K1F53Y amino acid substitution. These 2 mutations act cooperatively to activate MAP kinase signalling and hence drive tumour growth.
Davidson et al., doi: 10.1016/j.jid.2019.01.027.
Chromatin remodelling and epigenetics.
Role of the SWI/SNF and NuRF chromatin remodelling complexes.
Using tandem affinity purification of tagged MITF or SOX10 from human melanoma cells coupled with mass-spectrometry, we identified were the chromatin remodelling complexes SWI/SNF, NuRF and NuRD as common protein cofactors. We defined the role of BRG1, the ATPase subunit of the SWI/SNF PBAF complex, in human melanoma cells in culture showing how it is recruited by MITF and/or SOX10 to cis-regulatory elements driving expression of genes involved in melanoma cell proliferation.
Laurette etal., doi: 10.7554/eLife.06857. https://elifesciences.org/articles/06857
We defined the role of the BPTF scaffolding subunit of NuRF in human melanoma cells in culture and in the mouse melanocyte lineage in vivo where we identified its critical role in reactivation of the melanocyte gene expression program in adult melanocyte stem cells during anagen phase.
Koludrovic et al., doi: 10.1371/journal.pgen.1005555. https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1005555
We investigated the functions of BRG1 and BPTF in a mouse model of melanoma in vivo. We established mice where expression of oncogenic BRAFV600E and inactivation of PTEN can be induced in melanocytes by Tamoxifen treatment leading to the rapid development of melanoma. We found that as melanoma progresses in these animals and the cells invade deeply into the dermis, they change morphology and lose pigmentation. By ChIP-seq we defined that developing melanoma cells undergo epigenetic reprogramming leading to the silencing of melanocyte identity genes such as MITF and its targets, whereas SOX10 remains strongly expressed and displays a ‘super-enhancer’ signature. These and other analyses we performed show that in this model progressing melanoma cells lose their melanocyte identity and adopt a neural-crest like identity. This de-differentiation is likely induced by the stroma as culture of cells from late-stage tumours in vitro leads to rapid reactivation MITF expression along with its melanocyte identity target genes. Remarkably, while the melanocyte identity genes are marked by H3K27me3 and loss of H3K27ac and RNA polymerase II in late-stage tumours, the MITF locus retains H3K27ac but not Pol II and does not accumulate H3K27me3. MITF therefore remains primed to be reactivated explain its rapid re-expression in the ex vivo cultured cells.
By breeding these mice with those where the genes encoding Brg1 or Bptf are floxed, we could coordinate tumour initiation with their inactivation. Loss of these factors strongly inhibited tumour formation. Some tumours did however grow from cells that had escaped recombination. When these cells were grown in vitro and treated with 4-hydroxy-tamoxifen the residual floxed Brg1 or Bptf alleles were recombined leading to apoptosis. In the ex vivo cultured cells, Mitf together with Sox10 co-regulated a large number of genes essential for their growth. RNA-seq after siRNA-mediated Mitf or Sox10 silencing or 4-hydroxtamoxifen Brg1 and Bptf inactivation revealed how these chromatin remodellers acts as cofactors for the Mitf and Sox10 regulated gene expression programs. Together these transcription factors and chromatin remodelling complexes orchestrate essential gene expression programs in mouse melanoma cells.
Laurette et al., doi: 10.1038/s41418-019-0333-6. https://www.nature.com/articles/s41418-019-0333-6
Regulation of cancer cell glycolysis.
Although subunits of the NuRD complex are found in both the MITF and SOX10 interactomes, we could find no evidence that NuRD acted as a cofactor for MITF or SOX10 in melanoma cells. Nevertheless, siRNA CHD4 silencing led to strongly reduced cell proliferation. RNA-seq showed that CHD4 silencing strongly induced the expression of PADI1 (Protein Arginine Deiminase 1) and PADI3 encoding enzymes that convert peptidyl-arginine to citrulline. Immunoprecipitation with a pan-citrulline antibody and mass-spectrometry revealed increased citrullination of many glycolytic enzymes including PKM2 siCHD4 silenced cells. PKM2 is a highly regulated enzyme playing a central role in integrating cellular metabolic status and cell cycle with control of glycolysis. We confirmed increased PKM2 citrullination by immunoblot and identified several citrullinated arginine residues including R106. Moreover, siCHD4 silencing or ectopic PAD1/PAD3 expression increased basal and maximal glycolytic rates in a series of melanoma cell lines as well as cervical, renal and breast cancer lines and stimulated PKM2 enzymatic activity. As a result of this excess glycolysis, reduced pyruvate is available for mitochondrial metabolism that is a much more prolific source of ATP. Consequently, ATP levels are strongly depleted resulting in reduced cell proliferation.
PKM2 activity is positively regulated by serine (Ser), fructose 1,6-biphosphate (FBP) and succinylaminoimidazole-carboxamide riboside (SAICAR) and negatively regulated by tryptophan (Trp), alanine (Ala) and phenylalanine (Phe), thus coupling glycolytic flux to the level of critical anabolic intermediate metabolites. The free amino acids compete for binding to a pocket in PKM2 with R016 located at the entrance of the pocket. In the apo state, R106 mostly faces the solvent, but upon free amino acid binding, it rotates towards the pocket where its guanidino group interacts with the carboxylate group of the bound amino acid. Ser forms an extensive hydrogen bond network within the pocket stabilizing the active conformation, whereas upon Trp, Ala, or Phe binding, their hydrophobic side chain causes an allosteric transition to the inactive state. We hypothesized that loss of R106 positive side chain charge upon citrullination will weaken its interaction with free amino acids, but that due to its extended network of hydrogen bonds within the pocket Ser binding would be less affected than the hydrophobic amino acids that induce important structural changes. Consequently, R106 citrullination would weaken the inhibitory effect of Trp, Ala and Phe thereby shifting the equilibrium towards activation by Ser. We could fully substantiate this idea by evaluating the effects of altering Ser, Phe and Trp concentrations on live cell glycolysis and PKM2 enzymatic activity. PKM2 citrullination therefore represents a physiological mechanism to regulate glycolysis and cell proliferation adding another layer of complexity to the control of PKM2 activity.
Expression of the PAD1 and PAD3 enzymes is tightly regulated and they are only weakly expressed under most conditions. However, their expression is strongly induced by hypoxia. By analyses of patient data in TCGA, we found a strong correlation of PAD1 and PAD3 expression with hypoxic score in several cancer types such as pancreatic or lung carcinoma or clear cell renal carcinoma. We demonstrated experimentally that they were induced in cells cultured in hypoxic and pseudo-hypoxic conditions leading to enhanced R106 citrullination and increased glycolysis.
We therefore identified a novel pathway regulating cancer cell proliferation where CHD4 or hypoxia regulates PADI1 and PADI3 expression and their potential to citrullinate key arginines in PKM2 involved in its allosteric regulation to modulate glycolysis and cell proliferation. PAD1 and PAD3-mediated PKM2 citrullination also contributes to the increased glycolysis seen under hypoxic conditions, a hallmark of many cancers and RA and may be active in other pathological contexts associated with increased glycolysis.
Coassolo et al., doi: 10.1038/s41467-021-21960-4. https://www.nature.com/articles/s41467-021-21960-4
GATA6 a new actor in melanoma drug resistance.
Melanoma heterogeneity is a source of therapeutic resistance. Exposure of melanoma cells in vitro or melanoma PDX in vivo to MAP kinase inhibitors (MAPKi) leads to phenotype switching of melanocytic cells towards drug tolerant/resistant NCSC and/or mesenchymal states. We investigated how melanoma cells become resistant to THZ1 an inhibitor of the TFIIH-associated kinase CDK7. We found that TFIIH/CDK7 played a critical role in driving MITF expression via the super-enhancers that characterize this locus in melanoma cells. Consequently, exposure to THZ1 induces melanocytic cells to lose MITF expression and hence undergo a switch to a mesenchymal state resistant to both THZ1 and MAPKi. THZ1 treatment of partially mesenchymal MM047 cells also led them to switch to a more fully mesenchymal state. A comparison of the gene expression programs of the THZ1-treated melanocytic and MM047 cells identified a group of genes whose expression was not generally related to the mesenchymal switch, but more specifically to acquisition of the drug resistant state. Amongst these genes is the ABCG2 transporter that exports MAPKi and THZ1 from cells and in thus an important contributor to the resistant state.
Mining of the drug resistant signature identified transcription factor GATA6 as a potential regulator. Expression of MITF and GATA6 is inversely correlated in patient melanoma and in MAPKi treated PDX where in addition GATA6 expression is upregulated in the IFNg-active drug resistant state. We found that MITF bound to a negative intronic regulatory element in GATA6 to repress its expression. Thus, upon THZ1 inhibition or other stimuli that induce de-differentiation, such as IFNg or TNFa, GATA6 expression is upregulated. Together, these data identify GATA6 as a novel transcription factor activated in the mesenchymal state and important in regulating a gene expression program involved in drug resistance.
Berico P, et al., doi: 10.15252/embr.202051683.
TEAD transcription factors in myogenic differentiation.
In 1988, we purified the TEAD1 transcription factor from Hela cell nuclear extracts and subsequently cloned genes for the entire family.
Davidson et al., doi: 10.1016/0092-8674(88)90108-0. https://www.sciencedirect.com/science/article/pii/0092867488901080?via%3Dihub
Xiao et al., doi: 10.1016/0092-8674(91)90088-g. https://www.sciencedirect.com/science/article/pii/009286749190088G?via%3Dihub
Jacquemin et al., doi: 10.1074/jbc.271.36.21775. https://www.sciencedirect.com/science/article/pii/S0021925819618582?via%3Dihub
More than 20 years later we returned to the study of TEAD factors, intrigued by the fact that TEAD factors act as mediators of the Hippo signalling pathway interacting with the YAP and TAZ transcriptional co-activators to regulate proliferation, oncogenesis, stem cell maintenance and differentiation and control of organ size, but that they also played an important role in skeletal, cardiac, and smooth muscle differentiation and physiology both in vitro and in vivo. In a first study, we showed that stable shRNA-mediated Tead4 knockdown led to formation of shortened C2C12 myotubes and ChIP-chip experiments in C2C12 cells revealed that Tead4 occupied 867 promoters including muscle structural and regulatory proteins.
However, defining TEAD transcription factor function in myogenesis has proved elusive due to overlapping expression of family members and their functional redundancy. We subsequently showed that silencing of either Tead1, Tead2 or Tead4 did not effect primary myoblast (PM) differentiation, that was strongly impaired only by their simultaneous knockdown. In contrast, Tead1 or Tead4 silencing impaired C2C12 differentiation showing their different contributions in PMs and C2C12 cells. ChIP-seq identified enhancers associated with myogenic genes bound by combinations of Tead4, Myod1 or Myog. Tead4 regulated distinct gene sets in C2C12 cells and PMs involving both activation of the myogenic program and repression of growth and signaling pathways. ChIP-seq from mature mouse muscle fibres in vivo identified a set of highly transcribed muscle cell-identity genes and sites bound by Tead1 and Tead4. Nevertheless, Tead4 inactivation in mature muscle fibres had no appreciable phenotype. In contrast, Tead4 inactivation delayed notexin-induced muscle regeneration revealing its important role in myogenic differentiation in vivo.
By combining knockdown in cell models in vitro with Tead4 inactivation in muscle in vivo, we provided the first comprehensive description of the specific and redundant roles of Tead factors in myogenic differentiation and muscle regeneration. We also pioneered a new protocol for generation of ChIP-grade chromatin from mature murine muscle fibres. This allowed us to identify super-enhancer associated genes involved in muscle identity as well as TEAD-factor binding sites in muscle fibres.
Joshi et al., doi: 10.3791/56504. https://www.jove.com/t/56504/improved-protocol-for-chromatin-immunoprecipitation-from-mouse
Renal medullary carcinoma is the third most common kidney cancer in young adults. Resistant to therapies used against other types of kidney cancer,…