Sobia Raza

A gene expression atlas of the domestic pig

BackgroundThis work describes the first genome-wide analysis of the transcriptional landscape of the pig. A new porcine Affymetrix expression array was designed in order to provide comprehensive coverage of the known pig transcriptome. The new array was used to generate a genome-wide expression atlas of pig tissues derived from 62 tissue/cell types. These data were subjected to network correlation analysis and clustering.ResultsThe analysis presented here provides a detailed functional clustering of the pig transcriptome where transcripts are grouped according to their expression pattern, so one can infer the function of an uncharacterized gene from the company it keeps and the locations in which it is expressed. We describe the overall transcriptional signatures present in the tissue atlas, where possible assigning those signatures to specific cell populations or pathways. In particular, we discuss the expression signatures associated with the gastrointestinal tract, an organ that was sampled at 15 sites along its length and whose biology in the pig is similar to human. We identify sets of genes that define specialized cellular compartments and region-specific digestive functions. Finally, we performed a network analysis of the transcription factors expressed in the gastrointestinal tract and demonstrate how they sub-divide into functional groups that may control cellular gastrointestinal development.ConclusionsAs an important livestock animal with a physiology that is more similar than mouse to man, we provide a major new resource for understanding gene expression with respect to the known physiology of mammalian tissues and cells. The data and analyses are available on the websites http://biogps.org and http://www.macrophages.com/pig-atlas.

Can DCs be distinguished from macrophages by molecular signatures

187 eages. Sets of coregulated genes can be associated with the transcription factors that regulate them2,3. We downloaded the data from the ImmGen Project from the Gene Expression Omnibus database of the National Center for Biotechnology Information, included all the monocyte and macrophage sets, and reclustered those data with the network tool BioLayout Express3D (ref. 2). The resulting data are available as a network graph (http://www.macrophages.com/ HumeNI2013), along with the rederived clusters. Consistent with our analysis of the BioGPS data2, we were unable to identify any DC-associated cluster. None of the genes proposed by Miller et al. as a core DC signature or as having significantly higher expression in DCs1 were expressed in all of the DC populations or were restricted to lymphoid tissue DCs only. In every case, their expression was all shared with macrophages in some location, state of differentiation or activation, or with myeloid progenitor cells, mast cells, T cells or B cells. They also did not form a distinct cluster of coexpression, which suggested they do not share transcriptional regulation2,3. The extent to which the choice of prototypical macrophages determines the identification of DC-specific genes is evident from our analysis (Supplementary Fig. 1) showing all of the genes selected as being specific to DCs (the core DC set) or with significant upregulation in DCs for the full set of samples from the ImmGen Project. A sample-to-sample correlation matrix graph of these data grouped all of the cells (Fig. 1) and indicated their relatedness as members of the mononuclear phagocyte family4–6. The spleen and lymph node DCs (clusters 2 and 3) segregate from all other tissue DCs (clusters 1 and 4). Cluster 5 contains samples of MHC class II–positive cells from the peritoneal cavity subclassified on the basis of differences in the expression of F4/80, CD11c and CD11b but that are all very similar. The plasmacytoid DCs and each of the tissue macrophages also cluster independently of the myeloid DCs and of each other. Those findings reinforce evidence of the known plasticity of mononuclear phagocytes4–6 but do not support the delineation of non– lymphoid tissue DCs from macrophages. The BioGPS data2 and the extended data from the ImmGen Project indicate that the expression of Flt3 and Kit is higher in DCs relative to their expression in macrophages. Flt3 expression is low or absent from the CD11b+ populations grouped with the DCs (Supplementary Fig. 1). The expression of neither gene is higher in DCs than in myeloid progenitors, and Kit expression is much higher in mast cells (http://www. biogps.org/#goto=genereport&id=3815). The expression of Flt3l, which encodes a molecule (Flt3L) that controls the differentiation of lymphoid tissue DCs and a subset of tissue DCs7, is much higher in naive T cells than in other hematopoietic cells3, which could provide one explanation for the enrichment for Flt3-expressing DCs in T cell areas. To the Editor: An article from the Immunological Genome (ImmGen) Consortium published in Nature Immunology focuses on the differentiation of dendritic cells (DCs) in the mouse and efforts to distinguish those cells from macrophages1. In that article, Miller et al. report a small set of markers whose expression is restricted to or shows higher expression in DCs relative to their expression in four prototypical tissue macrophage populations1. Those findings contrast to our own observations obtained with large microarray data sets of macrophages and DCs2,3, and we wish to provide a different perspective. The authors exclude from the set of prototypical macrophages used for comparison most of the macrophage-related samples available in the ImmGen Project data set, including monocytes, subcapsular sinus and medullary macrophages, elicited peritoneal macrophages and salmonella-infected macrophages. Many of the genes shown to be DC specific (such as Traf1, Slamf7, Gpr68, Gpr132 and Hmgn3) actually have high expression in the macrophage samples excluded from that analysis. Those and other genes inferred to be DC markers (such as Jak2 and Pstpip1) are clearly also expressed by macrophages, according to the data in the BioGPS gene-annotation portal, which includes large data sets from highly purified mouse splenic DC subpopulations, bone marrow–derived macrophages, thioglycollate-elicited peritoneal macrophages and microglia, as well as other leukocytes and progenitors. The DC-specific marker set reported by Miller et al.1 does not include any of the genes encoding surface markers, such as Itgax (CD11c), Itgae (CD103), Ly75 (DEC205) or Cd209 (DC-SIGN), considered before to be DC markers but that are clearly not DC restricted3, even though CD11c and CD103 are used by Miller et al. as markers to purify the tissue DCs examined1. The molecules encoded by the DC-specific set of genes do not obviously show enrichment for any biological function, nor does that set include genes encoding molecules involved in antigen presentation. Conversely, molecules encoded by the genes whose expression is higher specifically in DCs, as identified by Miller et al., include the transcriptional transactivator CIITA, and its targets, the major histocompatibility complex (MHC) class II molecules1. Those have been clustered together before but have not been limited to lymphoid tissue DCs2,3. In the study by Miller et al.1, DCs seem to have higher relative expression of those genes they because the prototypical macrophages chosen for comparison were actually selected based in part on their lack of expression of MHC class II. Expression of MHC class II is induced on classically activated macrophages, and MHC class II is expressed on most tissue macrophages, especially those associated intimately with epithelia4. So in this way, the chosen macrophages are not prototypical. Coexpression clustering gives a different perspective on cell linCan DCs be distinguished from macrophages by molecular signatures? C o r r e s p o n D e n C e

A logic-based diagram of signalling pathways central to macrophage activation

BackgroundThe complex yet flexible cellular response to pathogens is orchestrated by the interaction of multiple signalling and metabolic pathways. The molecular regulation of this response has been studied in great detail but comprehensive and unambiguous diagrams describing these events are generally unavailable. Four key signalling cascades triggered early-on in the innate immune response are the toll-like receptor, interferon, NF-κB and apoptotic pathways, which co-operate to defend cells against a given pathogen. However, these pathways are commonly viewed as separate entities rather than an integrated network of molecular interactions.ResultsHere we describe the construction of a logically represented pathway diagram which attempts to integrate these four pathways central to innate immunity using a modified version of the Edinburgh Pathway Notation. The pathway map is available in a number of electronic formats and editing is supported by yEd graph editor software.ConclusionThe map presents a powerful visual aid for interpreting the available pathway interaction knowledge and underscores the valuable contribution well constructed pathway diagrams make to communicating large amounts of molecular interaction data. Furthermore, we discuss issues with the limitations and scalability of pathways presented in this fashion, explore options for automated layout of large pathway networks and demonstrate how such maps can aid the interpretation of functional studies.

Construction of a large scale integrated map of macrophage pathogen recognition and effector systems

BackgroundIn an effort to better understand the molecular networks that underpin macrophage activation we have been assembling a map of relevant pathways. Manual curation of the published literature was carried out in order to define the components of these pathways and the interactions between them. This information has been assembled into a large integrated directional network and represented graphically using the modified Edinburgh Pathway Notation (mEPN) scheme.ResultsThe diagram includes detailed views of the toll-like receptor (TLR) pathways, other pathogen recognition systems, NF-kappa-B, apoptosis, interferon signalling, MAP-kinase cascades, MHC antigen presentation and proteasome assembly, as well as selected views of the transcriptional networks they regulate. The integrated pathway includes a total of 496 unique proteins, the complexes formed between them and the processes in which they are involved. This produces a network of 2,170 nodes connected by 2,553 edges.ConclusionsThe pathway diagram is a navigable visual aid for displaying a consensus view of the pathway information available for these systems. It is also a valuable resource for computational modelling and aid in the interpretation of functional genomics data. We envisage that this work will be of value to those interested in macrophage biology and also contribute to the ongoing Systems Biology community effort to develop a standard notation scheme for the graphical representation of biological pathways.

Characterisation of a Novel Fc Conjugate of Macrophage Colony-stimulating Factor

We have produced an Fc conjugate of colony-stimulating factor (CSF) 1 with an improved circulating half-life. CSF1-Fc retained its macrophage growth-promoting activity, and did not induce proinflammatory cytokines in vitro. Treatment with CSF1-Fc did not produce adverse effects in mice or pigs. The impact of CSF1-Fc was examined using the Csf1r-enhanced green fluorescent protein (EGFP) reporter gene in MacGreen mice. Administration of CSF1-Fc to mice drove extensive infiltration of all tissues by Csf1r-EGFP positive macrophages. The main consequence was hepatosplenomegaly, associated with proliferation of hepatocytes. Expression profiles of the liver indicated that infiltrating macrophages produced candidate mediators of hepatocyte proliferation including urokinase, tumor necrosis factor, and interleukin 6. CSF1-Fc also promoted osteoclastogenesis and produced pleiotropic effects on other organ systems, notably the testis, where CSF1-dependent macrophages have been implicated in homeostasis. However, it did not affect other putative CSF1 targets, notably intestine, where Paneth cell numbers and villus architecture were unchanged. CSF1 has therapeutic potential in regenerative medicine in multiple organs. We suggest that the CSF1-Fc conjugate retains this potential, and may permit daily delivery by injection rather than continuous infusion required for the core molecule.

Pleiotropic effects of extended blockade of CSF1R signaling in adult mice.

We investigated the role of CSF1R signaling in adult mice using prolonged treatment with anti‐CSF1R antibody. Mutation of the CSF1 gene in the op/op mouse produces numerous developmental abnormalities. Mutation of the CSF1R has an even more penetrant phenotype, including perinatal lethality, because of the existence of a second ligand, IL‐34. These effects on development provide limited insight into functions of CSF1R signaling in adult homeostasis. The carcass weight and weight of several organs (spleen, kidney, and liver) were reduced in the treated mice, but overall body weight gain was increased. Despite the complete loss of Kupffer cells, there was no effect on liver gene expression. The treatment ablated OCL, increased bone density and trabecular volume, and prevented the decline in bone mass seen in female mice with age. The op/op mouse has a deficiency in pancreatic β cells and in Paneth cells in the gut wall. Only the latter was reproduced by the antibody treatment and was associated with increased goblet cell number but no change in villus architecture. Male op/op mice are infertile as a result of testosterone insufficiency. Anti‐CSF1R treatment ablated interstitial macrophages in the testis, but there was no sustained effect on testosterone or LH. The results indicate an ongoing requirement for CSF1R signaling in macrophage and OCL homeostasis but indicate that most effects of CSF1 and CSF1R mutations are due to effects on development.

Combined genome-wide expression profiling and targeted RNA interference in primary mouse macrophages reveals perturbation of transcriptional networks associated with interferon signalling

BackgroundInterferons (IFNs) are potent antiviral cytokines capable of reprogramming the macrophage phenotype through the induction of interferon-stimulated genes (ISGs). Here we have used targeted RNA interference to suppress the expression of a number of key genes associated with IFN signalling in murine macrophages prior to stimulation with interferon-gamma. Genome-wide changes in transcript abundance caused by siRNA activity were measured using exon-level microarrays in the presence or absence of IFNγ.ResultsTransfection of murine bone-marrow derived macrophages (BMDMs) with a non-targeting (control) siRNA and 11 sequence-specific siRNAs was performed using a cationic lipid transfection reagent (Lipofectamine2000) prior to stimulation with IFNγ. Total RNA was harvested from cells and gene expression measured on Affymetrix GeneChip Mouse Exon 1.0 ST Arrays. Network-based analysis of these data revealed six siRNAs to cause a marked shift in the macrophage transcriptome in the presence or absence IFNγ. These six siRNAs targeted the Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 transcripts. The perturbation of the transcriptome by the six siRNAs was highly similar in each case and affected the expression of over 600 downstream transcripts. Regulated transcripts were clustered based on co-expression into five major groups corresponding to transcriptional networks associated with the type I and II IFN response, cell cycle regulation, and NF-KB signalling. In addition we have observed a significant non-specific immune stimulation of cells transfected with siRNA using Lipofectamine2000, suggesting use of this reagent in BMDMs, even at low concentrations, is enough to induce a type I IFN response.ConclusionOur results provide evidence that the type I IFN response in murine BMDMs is dependent on Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2, and that siRNAs targeted to these genes results in perturbation of key transcriptional networks associated with type I and type II IFN signalling and a suppression of macrophage M1 polarization.

Co-expression of FBN1 with mesenchyme-specific genes in mouse cell lines: implications for phenotypic variability in Marfan syndrome

Mutations in the human FBN1 gene cause Marfan syndrome, a complex disease affecting connective tissues but with a highly variable phenotype. To identify genes that might participate in epistatic interactions with FBN1, and could therefore explain the observed phenotypic variability, we have looked for genes that are co-expressed with Fbn1 in the mouse. Microarray expression data derived from a range of primary mouse cells and cell lines were analysed using the network analysis tool BioLayout Express3D. A cluster of 205 genes, including Fbn1, were selectively expressed by mouse cell lines of different mesenchymal lineages and by mouse primary mesenchymal cells (preadipocytes, myoblasts, fibroblasts, osteoblasts). Promoter analysis of this gene set identified several candidate transcriptional regulators. Genes within this co-expressed cluster are candidate genetic modifiers for Marfan syndrome and for other connective tissue diseases.

The mEPN scheme: an intuitive and flexible graphical system for rendering biological pathways

BackgroundThere is general agreement amongst biologists about the need for good pathway diagrams and a need to formalize the way biological pathways are depicted. However, implementing and agreeing how best to do this is currently the subject of some debate.ResultsThe modified Edinburgh Pathway Notation (mEPN) scheme is founded on a notation system originally devised a number of years ago and through use has now been refined extensively. This process has been primarily driven by the authors attempts to produce process diagrams for a diverse range of biological pathways, particularly with respect to immune signaling in mammals. Here we provide a specification of the mEPN notation, its symbols, rules for its use and a comparison to the proposed Systems Biology Graphical Notation (SBGN) scheme.ConclusionsWe hope this work will contribute to the on-going community effort to develop a standard for depicting pathways and will provide a coherent guide to those planning to construct pathway diagrams of their biological systems of interest.

Analysis of the transcriptional networks underpinning the activation of murine macrophages by inflammatory mediators

Macrophages respond to the TLR4 agonist LPS with a sequential transcriptional cascade controlled by a complex regulatory network of signaling pathways and transcription factors. At least two distinct pathways are currently known to be engaged by TLR4 and are distinguished by their dependence on the adaptor molecule MyD88. We have used gene expression microarrays to define the effects of each of three variables—LPS dose, LPS versus IFN‐β and ‐γ, and genetic background—on the transcriptional response of mouse BMDMs. Analysis of correlation networks generated from the data has identified subnetworks or modules within the macrophage transcriptional network that are activated selectively by these variables. We have identified mouse strain‐specific signatures, including a module enriched for SLE susceptibility candidates. In the modules of genes unique to different treatments, we found a module of genes induced by type‐I IFN but not by LPS treatment, suggesting another layer of complexity in the LPS‐TLR4 signaling feedback control. We also observe that the activation of the complement system, in common with the known activation of MHC class 2 genes, is reliant on IFN‐γ signaling. Taken together, these data further highlight the exquisite nature of the regulatory systems that control macrophage activation, their likely relevance to disease resistance/susceptibility, and the appropriate response of these cells to proinflammatory stimuli.