Temporal Coding in Spiking Neural Networks With Alpha Synaptic Function: Learning With Backpropagation.

The timing of individual neuronal spikes is essential for biological brains to make fast responses to sensory stimuli. However, conventional artificial neural networks lack the intrinsic temporal coding ability present in biological networks. We propose a spiking neural network model that encodes information in the relative timing of individual spikes. In classification tasks, the output of the network is indicated by the first neuron to spike in the output layer. This temporal coding scheme allows the supervised training of the network with backpropagation, using locally exact derivatives of the postsynaptic spike times with respect to presynaptic spike times. The network operates using a biologically plausible synaptic transfer function. In addition, we use trainable pulses that provide bias, add flexibility during training, and exploit the decayed part of the synaptic function. We show that such networks can be successfully trained on multiple data sets encoded in time, including MNIST. Our model outperforms comparable spiking models on MNIST and achieves similar quality to fully connected conventional networks with the same architecture. The spiking network spontaneously discovers two operating modes, mirroring the accuracy-speed tradeoff observed in human decision-making: a highly accurate but slow regime, and a fast but slightly lower accuracy regime. These results demonstrate the computational power of spiking networks with biological characteristics that encode information in the timing of individual neurons. By studying temporal coding in spiking networks, we aim to create building blocks toward energy-efficient, state-based biologically inspired neural architectures. We provide open-source code for the model.

Publisher Correction: c-Maf controls immune responses by regulating disease-specific gene networks and repressing IL-2 in CD4+ T cells.

In the version of this article initially published, the Supplementary Data file was an incorrect version. The correct version is now provided. The error has been corrected in the HTML and PDF version of the article.

c-Maf controls immune responses by regulating disease-specific gene networks and repressing IL-2 in CD4+ T cells.

The transcription factor c-Maf induces the anti-inflammatory cytokine IL-10 in CD4+ T cells in vitro. However, the global effects of c-Maf on diverse immune responses in vivo are unknown. Here we found that c-Maf regulated IL-10 production in CD4+ T cells in disease models involving the TH1 subset of helper T cells (malaria), TH2 cells (allergy) and TH17 cells (autoimmunity) in vivo. Although mice with c-Maf deficiency targeted to T cells showed greater pathology in TH1 and TH2 responses, TH17 cell-mediated pathology was reduced in this context, with an accompanying decrease in TH17 cells and increase in Foxp3+ regulatory T cells. Bivariate genomic footprinting elucidated the c-Maf transcription-factor network, including enhanced activity of NFAT; this led to the identification and validation of c-Maf as a negative regulator of IL-2. The decreased expression of the gene encoding the transcription factor RORγt (Rorc) that resulted from c-Maf deficiency was dependent on IL-2, which explained the in vivo observations. Thus, c-Maf is a positive and negative regulator of the expression of cytokine-encoding genes, with context-specific effects that allow each immune response to occur in a controlled yet effective manner.

Analysis of Transcriptional Signatures in Response to Listeria monocytogenes Infection Reveals Temporal Changes That Result from Type I Interferon Signaling.

Analysis of the mouse transcriptional response to Listeria monocytogenes infection reveals that a large set of genes are perturbed in both blood and tissue and that these transcriptional responses are enriched for pathways of the immune response. Further we identified enrichment for both type I and type II interferon (IFN) signaling molecules in the blood and tissues upon infection. Since type I IFN signaling has been reported widely to impair bacterial clearance we examined gene expression from blood and tissues of wild type (WT) and type I IFNαß receptor-deficient (Ifnar1-/-) mice at the basal level and upon infection with L. monocytogenes. Measurement of the fold change response upon infection in the absence of type I IFN signaling demonstrated an upregulation of specific genes at day 1 post infection. A less marked reduction of the global gene expression signature in blood or tissues from infected Ifnar1-/- as compared to WT mice was observed at days 2 and 3 after infection, with marked reduction in key genes such as Oasg1 and Stat2. Moreover, on in depth analysis, changes in gene expression in uninfected mice of key IFN regulatory genes including Irf9, Irf7, Stat1 and others were identified, and although induced by an equivalent degree upon infection this resulted in significantly lower final gene expression levels upon infection of Ifnar1-/- mice. These data highlight how dysregulation of this network in the steady state and temporally upon infection may determine the outcome of this bacterial infection and how basal levels of type I IFN-inducible genes may perturb an optimal host immune response to control intracellular bacterial infections such as L. monocytogenes.

The Blood Transcriptome of Experimental Melioidosis Reflects Disease Severity and Shows Considerable Similarity with the Human Disease.

Melioidosis, a severe human disease caused by the bacterium Burkholderia pseudomallei, has a wide spectrum of clinical manifestations ranging from acute septicemia to chronic localized illness or latent infection. Murine models have been widely used to study the pathogenesis of infection and to evaluate novel therapies or vaccines, but how faithfully they recapitulate the biology of human melioidosis at a molecular level is not known. In this study, mice were intranasally infected with either high or low doses of B. pseudomallei to generate either acute, chronic, or latent infection and host blood and tissue transcriptional profiles were generated. Acute infection was accompanied by a homogeneous signature associated with induction of multiple innate immune response pathways, such as IL-10, TREM1, and IFN signaling, largely found in both blood and tissue. The transcriptional profile in blood reflected the heterogeneity of chronic infection and quantitatively reflected the severity of disease. Genes associated with fibrosis and tissue remodeling, including matrix metalloproteases and collagen, were upregulated in chronically infected mice with severe disease. Transcriptional signatures of both acute and chronic melioidosis revealed upregulation of iNOS in tissue, consistent with the expression of IFN-γ, but also Arginase-1, a functional antagonist of the iNOS pathway, and was confirmed by immunohistochemistry. Comparison of these mouse blood datasets by pathway and modular analysis with the blood transcriptional signature of patients with melioidosis showed that many genes were similarly perturbed, including Arginase-1, IL-10, TREM1, and IFN signaling, revealing the common immune response occurring in both mice and humans.

Hippocampal CA1 region shows differential regulation of gene expression in mice displaying extremes in behavioral sensitization to amphetamine: relevance for psychosis susceptibility?

RATIONALE: Psychosis susceptibility is mediated in part by the dopaminergic neurotransmitter system. In humans, individual differences in vulnerability for psychosis are reflected in differential sensitivity for psychostimulants such as amphetamine. We hypothesize that the same genes and pathways underlying behavioral sensitization in mice are also involved in the vulnerability to psychosis. OBJECTIVES: The aim of the current study was to investigate which genes and pathways may contribute to behavioral sensitization in different dopaminergic output areas in the mouse brain. METHODS: We took advantage of the naturally occurring difference in psychostimulant sensitivity in DBA/2 mice and selected animals displaying extremes in behavioral sensitization to amphetamine. Subsequently, the dopamine output areas, prefrontal cortex, nucleus accumbens, and cornu ammonis 1 (CA1) area of the hippocampus, were isolated by laser microdissection and subjected to DNA microarray analysis 1 h after a challenge dose of amphetamine. RESULTS: A large number of genes with differential expression between high and low responders were identified, with no overlap between brain regions. Validation of these gene expression changes with real-time quantitative polymerase chain reaction demonstrated that the most robust and reproducible effects on gene expression were in the CA1 region of the hippocampus. Interestingly, many of the validated genes in CA1 are members of the cAMP response element (CRE) family and targets of the glucocorticoid receptor (GR) and myocyte enhancer factor 2 (Mef2) transcription factors. CONCLUSION: We hypothesize that CRE, Mef2, and GR signaling form a transcription regulating network, which underlies differential amphetamine sensitivity, and therefore, may play an important role in susceptibility to psychosis.

NMR structure of the amino-terminal domain from the Tfb1 subunit of TFIIH and characterization of its phosphoinositide and VP16 binding sites.

General transcription factor IIH (TFIIH) is recruited to the preinitiation complex (PIC) through direct interactions between its p62 (Tfb1) subunit and the carboxyl-terminal domain of TFIIEalpha. TFIIH has also been shown to interact with a number of transcriptional activator proteins through interactions with the same p62 (Tfb1) subunit. We have determined the NMR solution structure of the amino-terminal domain from the Tfb1 subunit of yeast TFIIH (Tfb1(1-115)). Like the corresponding domain from the human p62 protein, Tfb1(1-115) contains a PH domain fold despite a low level of sequence identity between the two functionally homologous proteins. In addition, we have performed in vitro binding studies that demonstrate that the PH domains of Tfb1 and p62 specifically bind to monophosphorylated inositides [PtdIns(5)P and PtdIns(3)P]. NMR chemical shift mapping demonstrated that the PtdIns(5)P binding site on Tfb1 (p62) is located in the basic pocket formed by beta-strands beta5-beta7 of the PH domain fold. Interestingly, the structural composition of the PtdIns(5)P binding site is different from the composition of the binding sites for phosphoinositides on prototypic PH domains. We have also determined that the PH domains from Tfb1 and p62 are sufficient for binding to the activation domain of VP16. NMR chemical shift mapping demonstrated that the VP16 binding site within the PH domain of Tfb1 (p62) overlaps with the PtdIns(5)P binding site on Tfb1 (p62). These results provide new information about the recognition of phosphoinositides by PH domains, and point to a potential role for phosphoinositides in VP16 regulation.

NMR structure of a complex containing the TFIIF subunit RAP74 and the RNA polymerase II carboxyl-terminal domain phosphatase FCP1.

FCP1 [transcription factor IIF (TFIIF)-associated carboxyl-terminal domain (CTD) phosphatase] is the only identified phosphatase specific for the phosphorylated CTD of RNA polymerase II (RNAP II). The phosphatase activity of FCP1 is enhanced in the presence of the large subunit of TFIIF (RAP74 in humans). It has been demonstrated that the CTD of RAP74 (cterRAP74; residues 436-517) directly interacts with the highly acidic CTD of FCP1 (cterFCP; residues 879-961 in human). In this manuscript, we have determined a high-resolution solution structure of a cterRAP74cterFCP complex by NMR spectroscopy. Interestingly, the cterFCP protein is completely disordered in the unbound state, but forms an alpha-helix (H1'; E945-M961) in the complex. The cterRAP74cterFCP binding interface relies extensively on van der Waals contacts between hydrophobic residues from the H2 and H3 helices of cterRAP74 and hydrophobic residues from the H1' helix of cterFCP. The binding interface also contains two critical electrostatic interactions involving aspartic acid residues from H1' of cterFCP and lysine residues from both H2 and H3 of cterRAP74. There are also three additional polar interactions involving highly conserved acidic residues from the H1' helix. The cterRAP74cterFCP complex is the first high-resolution structure between an acidic residue-rich domain from a holoenzyme-associated regulatory protein and a general transcription factor. The structure defines a clear role for both hydrophobic and acidic residues in proteinprotein complexes involving acidic residue-rich domains in transcription regulatory proteins.