In this study we set out to explore a number of questions. Firstly, our interest in pathways underpinning macrophage activation [21
] motivated us to want to analyse in parallel the contribution of a number of known factors to the IFNγ response in these cells. Secondly, previous RNAi studies in our laboratory had identified a number of ISGs as contributing to the enhanced antiviral state of IFNγ-primed BMDMs to mCMV infection. We reasoned that we may further our understanding of their mechanism of action by analysing the affect of their knockdown at the transcriptional level. Finally, we wished to explore the potential of using a combination of the recently available exon level microarrays and improved RNAi targeting capabilities to gain insights into the interferon signalling pathways. In order to address these questions, 11 genes were targeted with siRNAs followed by IFNγ treatment and microarray analysis on the Affymetrix mouse exon 1.0 ST array platform. This study provides one of the few reports investigating the utility of combining RNA interference with global transcript profiling in macrophages. Our results highlight the potential of the approach, as well as some of the associated difficulties in performing this work.
In carrying out this investigation, we had to contend with a number of technical issues. The induction of an IFN response by double stranded RNA has been shown to be an issue in a number of different cell types [23
]. This problem is potentially exacerbated in dendritic cells and macrophages due to their expression of TLRs, RNA helicases and other pattern recognition receptors involved in the sensing of pathogen-associated molecular patterns (PAMPs). The immuno-stimulatory properties of Lipofectamine2000 and other cationic lipid-based reagents have also been documented previously [32
]. In order to minimise these known effects we therefore used final lipid and siRNA concentrations lower than recommended by the supplier (Thermo Fisher). However as we observed even these 'mild' transfection conditions still induced a significant type I IFN response in BMDMs. This response was characterized by the up-regulation of pro-inflammatory cytokines, transcription factors and other IFN-induced genes between 5–24 hours post-transfection of control siRNA and represented a significant shift in the transcriptional activity of these cells. The use of of 1,2-dioleoyl-3-trimethylammonium-propane (DOTAP) lipid formations to transfect siRNA in mouse cells has also been shown to induce the type I interferon response [33
] and the immuno-stimulatory properties of lipid-based plasmid DNA transfections are well documented (reviewed in [35
]). Furthermore, we have detected up-regulation of IFN-induced transcripts in response to Dharmafect 1 (Thermo Fisher) 48 hours post treatment of mouse fibroblasts (NIH-3T3) (unpublished data). These studies support the notion that IFN stimulation by siRNA and transfection reagents may be a widespread effect occurring in a number of different cell types [36
A second technical issue of our study was the relatively low and variable knockdown efficiencies achieved when performing transfections in BMDMs, as measured by qPCR and array analysis. This was due in part to the low concentrations of reagents used, but also to the generally low efficiency of DNA/RNA delivery by transfection to primary cells such as macrophages. Primary macrophages are known to be considerably more difficult to transfect than cultured cell lines [37
], making efficient gene knockdowns difficult to achieve in this study.
Despite these technical issues, we generated high quality microarray expression data from targeted transfection studies which was analysed using a combination of conventional statistical and network-based approaches [26
]. At the exon level we were unable to observe any convincing evidence for alternative splicing events between the comparisons and therefore all further examination of the data was restricted to gene level analyses. Using network analysis it was possible to visualise relationships between differentially regulated transcripts and cluster them into distinct groups based on the similarity of their expression profiles across samples.
Network analysis of the data identified five major groups, or clusters, of co-expressed genes that were regulated by siRNA treatment. Genes within each cluster were found to be biologically related according to functional annotation and transcription factor binding site analysis, and co-regulated by IFNγ and/or siRNA treatment. The median expression profiles between clusters were markedly different, representing five distinct transcriptional networks. However, across all clusters, we observed a strong influence from the activity of six siRNAs targeted to the Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 genes. These siRNAs induced a global change in the macrophage transcriptome altering the expression of several hundred downstream genes. This effect was not observed in response to treatment of the cells with the RISC-free control siRNA or the other five siRNAs used in this study (targeting the Casp4, Ifi47, Lyn, Sod2 and Traf1 transcripts). The analyses presented here suggest that Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 all contribute to the control of genes regulated by both classes of interferon.
Clusters 1 and 2 in the data set (see Table and Figure ) represent genes directly induced or repressed by IFNγ treatment respectively. Genes within these two clusters were also influenced by the activity of the Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 siRNAs (to a varying extent). Our analysis of transcripts regulated by IFNγ stimulation was consistent with our previous time-course experiments (ArrayExpress: E-MEXP-1490) and with previous profiling studies in this area [15
]. Genes up-regulated by IFNγ stimulation (Cluster 1) reflected a broad range of immunomodulatory function, including an up-regulation of class II antigen presentation capabilities through the co-activator of MHC class 2 genes (Ciita), and histocompatibility class II antigens. Up-regulation of chemokines, complement components (C3, C4), caspases (Casp1, Casp7), interleukins (Il15, Il18) and interferon-induced proteins was also observed (for complete list see Additional file 8
). In contrast, Cluster 2 contains many genes that are known to be suppressed by IFNγ stimulation and whose basal and IFNγ-stimulated expression level increased following treatment with the six siRNAs relative to the controls and other siRNA treatments.
Cluster 3 was made up of a group of genes that did not respond significantly to IFNγ stimulation, but were markedly down-regulated by the activity of the six siRNAs mentioned above. Importantly, Cluster 3 contained many well known type I IFN anti-microbial effector genes encoding interferon inducible proteins and chemokines suggesting regulation by IFNβ. Furthermore, statistically significant over-representation of ISRE promoter sequences in the 5' flanking regions of these genes, again suggests a dependency on type I IFN regulation and the IFN-induced transcription factor complex, ISGF3. Transcripts within Cluster 3 therefore appear to have a strong transcriptional dependency on type I IFNs and were the most markedly down-regulated by siRNA treatment in the dataset. In contrast, genes with a dependency on type II IFN i.e. genes involved in MHC class II antigen presentation [9
], were not affected in this same manner by the six siRNAs. Genes with a co-dependency on both type I and type II IFN, we believe, are those in the data set being regulated by both IFNγ and the Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 siRNAs.
Cluster 4 in the data set is made up almost exclusively of genes whose function can be associated with cell cycle progression. Many transcripts within Cluster 4 (69/86) were found to be down-regulated in our mock transfection experiments during immune stimulation by Lipofectamine and RISC-Free siRNA. This suggests that transcripts within this cluster, most of which are associated with cell cycle control, are down regulated or suppressed during macrophage activation. This was supported by very low expression levels for these transcripts observed in RISC-Free control samples 24 hours post IFNγ treatment in our second series of experiments. The expression of genes within this cluster was markedly induced (or de-repressed) in response to Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 siRNAs (see Figure ). This suggests a link between the six genes targeted and the control of the cell cycle, which we believe may be a secondary effect of disrupting the IFN pathway. Cluster 5 in the data set consists of a group of 44 genes, many associated with the NF-kB signalling system supporting a link between this system and the IFN pathway in BMDMs [40
]. Details of this involvement are however ill-defined.
In trying to explain these observations regarding the Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 knockdown phenotypes, our hypothesis is that suppression of Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 using siRNA all result in a perturbation of the type I IFN response in BMDMs. We believe this occurs either by a direct perturbation of IFNβ induction following activation of pathogen detecting systems (as seen with Irf3, Irf5, Ifnb1 and Nfkb2 siRNAs) or by perturbation of signalling downstream of the type I receptor complex (as seen with Stat1 and Stat2 siRNAs). Perturbation at either of these levels in the pathway, we believe, is what accounts for the common downstream alteration of several hundred interferon-regulated transcripts as observed in this study. We also believe the perturbation has also influenced NF-kB signalling and resulted in a modulation of the cell cycle. The common phenotype induced by Ifnb1, Irf3, Irf5, Stat1, Stat2 and Nfkb2 siRNAs observed in our study suggests each of the genes targeted are operating at a similar level or hierarchy within the interferon pathway, and that suppression of these genes has a similar effect on the macrophage transcriptome.
We have been modelling the IFN system based on findings reported in the literature [21
] and have used this model to help further interpret the findings of this study (a simplified version of the model is shown in Figure ). As the model indicates, macrophages possess many cell surface and intracellular receptors for the detection of a broad range of molecular species specifically found in pathogenic organisms. It is some of these receptors that are undoubtedly activated by the transfection reagents/siRNA. The most likely candidates are those with RNA binding function such as Ddx58 (RIG-I) and Ifih1 (Mda5) which detect cytoplasmic viral ssRNA and dsRNA [5
] and/or the endosomal TLR receptors namely Tlr3, Tlr7 and Tlr8, that are also activated by these molecules [44
]. It is possible that these receptors may be sensitive to synthetic siRNA in activating the IFN response. Other TLR receptors e.g. Tlr1/2 and Tlr2/6 that are known to sensitive to lipopeptides and peptidoglycans might additionally be activated by the transfection reagent. According to our model, activation of all of these receptors ultimately leads to the phosphorylation, dimerization and translocation of Irf3 and/or Irf7 to the nucleus where they activate Ifnb1 expression. This formation of Irf3:Irf3 and Irf3:Irf7 dimers is an important regulatory event during the induction of IFNβ [45
] and subsequent up-regulation of ISGs following pattern recognition receptor activation [46
]. Therefore the suppression Irf3 using siRNA would be expected to have a strong influence on IFN regulation and subsequent downstream ISG expression. Indeed this is what we observed. In a similar fashion, if the IFNβ (Ifnb1) transcript itself was targeted for suppression, it might also be predicted to have a direct effect on downstream expression of type I IFN-induced genes (as also observed in this study). From this perspective, the phenotypes observed in response to Irf3 and Ifnb1 siRNAs in this study are as expected.
Figure 5 Model of known components of the IFN signalling pathway and explanation of observed results. Transfection of siRNA using Lipofectamine2000 in mouse BMDMs inductes a type I IFN response. This probably occurrs through the activation of pattern recognition (more ...)
Stat1 and Stat2 are primary transcriptional regulators of the IFN response and are essential components of the JAK-STAT signalling pathway. Their phosphorylation by ligand-activated interferon receptors leads them to form the Stat1-Stat2-Irf9 (ISGF3 complex) which is crucial for the transcriptional regulation of the IFN response via ISRE elements [48
] following induction of IFNβ. Stat1 also homodimerizes following activation of the type II IFN receptor complex to form the Stat1:Stat1 AAF complex to induce further transcription via GAS sites [12
]. The suppression of either Stat1 or Stat2 function would be predicted to affect IFN signalling directly and alter ISG expression by preventing transcription factor binding to ISG promoter sites. Indeed, we have observed a phenotype consistent with this in our study following Stat1 and Stat2 suppression using siRNA.
The role of the transcription factor Irf5 in the type I IFN response is less well established, although recently Paun et al
. has demonstrated that murine Irf5 can be activated by both TBK1 and MyD88 to form homodimers which bind to and activate transcription of type I IFN and inflammatory cytokine genes [49
]. Perturbation of Irf5 through siRNA knockdown in this study suggests that Irf5 could influence type I IFN-induced transcriptional networks at a similar level to Irf3. Further studies however will be required to clarify the role Irf5 in this context. Likewise, the role of Nfkb2 (p52/p100 subunit) in type I IFN signalling is difficult to explain based on the current understanding of this protein in the regulation of innate immunity. Nfkb2 is known to form transcription factor complexes with RelB and/or Bcl3 as part of the 'alternative' NF-KB pathway, often associated with B-cell maturation and lymphoid development [50
]. Our study strongly suggests that Nfkb2 may play a central role in the regulation of the type I IFN response in mouse BMDMs, however this observation is only partially supported in the literature [41
]. The presence of NF-kB binding elements in the IFNβ promoter (enhanceosome) [45
] raises the possibility of a direct interaction of this protein in IFNβ regulation, however further studies will of course be necessary to support this hypothesis.