Design logic of a cannabinoid receptor signaling network that triggers neurite outgrowth

Abstract

Cannabinoid receptor 1 (CB1R) regulates neuronal differentiation. To understand the logic underlying decision-making in the signaling network controlling CB1R-induced neurite outgrowth, we profiled the activation of several hundred transcription factors after cell stimulation. We assembled an in silico signaling network by connecting CB1R to 23 activated transcription factors. Statistical analyses of this network predicted a role for the breast cancer 1 protein BRCA1 in neuronal differentiation and a new pathway from CB1R through phosphoinositol 3-kinase to the transcription factor paired box 6 (PAX6). Both predictions were experimentally confirmed. Results of transcription factor activation experiments that used pharmacological inhibitors of kinases revealed a network organization of partial OR gates regulating kinases stacked above AND gates that control transcription factors, which together allow for distributed decision-making in CB1R-induced neurite outgrowth.

Fig. 1

Identification of positive and negative regulators of CB1R-induced neurite outgrowth. (A) Arrays of transcription factor activation in Neuro2A cells treated with DMSO as a control or 2 μM HU-210 (CB1R agonist) for 20 min. The right panel highlights several of the activated transcription factors. The colors in the panel correspond to the circled spots in the arrays. TCF1, T cell factor 1. (B) Effects of transcription factor inhibition on neurite outgrowth. Neuro2A cells were transfected with the indicated siRNAs or transfected with DN Stat3 constructs [Y→F and DNA-binding domain (DBD)], DN CREB, wild-type c-Myb, or pcDNA3 (see fig. S11 for construct expression) and then stimulated with HU-210 to induce neurite outgrowth. Error bars, mean ± SEM (n = 3 independent experiments); *, P < 0.05 (statistically significant difference by Student's t test) versus the control Luc siRNA; **, P < 0.05 (Student's t test) versus control pcDNA3. The figure is a composite of multiple experiments. The siRNA transfections were performed as two experimental sets. Set 1: Luc, AP-2, PAX6, c-Myb, and USF1. Set 2: Luc, NR3C1, Smad3, RARα, CEBPα, NFYA, and SPI1. Transfections of each DN construct were first done independently and then repeated as one experimental set. Depletion of transcription factor expression was confirmed by quantitative real-time reverse transcription polymerase chain reaction (RT-PCR) or immunoblot (fig. S12).

Fig. 2

Construction of networks and identification of BRCA1 and a PI3K-AKT-PAX6 pathway as regulators of CB1R-induced neurite outgrowth. (A) Eight mammalian protein-protein interaction databases and one signaling network were consolidated into a single network made of 67,379 human protein-protein and protein-ligand interactions (I). This network was filtered by removing interactions from research articles that reported more than three interactions. The lists of activated and nonactivated transcription factors (TFs) at 20 min were used as input nodes to find direct and neighboring interactions and to identify paths from the CB1R receptor to the transcription factors (II), enabling us to identify and rank regulators within the network (III). (B) A subnetwork created by finding the shortest paths of a maximum of seven steps from the HU-210 node (HU) to the 23 activated transcription factors (orange nodes) at 20 min. (I) Paths were found for 17 out of the 23 factors. BRCA1, PI3K, and AKT1 are highlighted (green nodes). HU and CB1R nodes are highlighted in blue. (II) Pathway connecting CB1R to PAX6 through PI3K and AKT1 (edited manually after literature review). (III) Table showing the ranking of components in pathways detected in a control subnetwork (fig. S4 and table S3) versus the activated subnetwork using the ranking method described in the supporting online material (SOM) (7). I, II, and III in (B) correspond to I, II, and III in (A), respectively. CREBBP, CREB-binding protein; PIK3CA, PIK3 catalytic, alpha; PIP2, phosphatidylinositol 4,5-bisphosphate; PIP3, phosphatidylinositol 3,4,5-trisphosphate. (C) Subnetwork created by finding the shortest paths of a maximum of two steps between the 23 activated transcription factors. Nineteen of the factors were connected using this method (orange nodes). A binomial proportions test was used to prune out most of the less important intermediates. BRCA1 is highlighted in green.

Fig. 3

Regulation of CB1R-induced neurite outgrowth by BRCA1. (A) Effect of BRCA1 siRNA on cannabinoid-induced neurite outgrowth. Neuro2A cells were transfected with Luc siRNA or BRCA1 siRNA and stimulated with 2 μM HU-210 to induce neurite outgrowth or with DMSO as a control. Amounts of neurite outgrowth in cells exposed to Luc siRNA and HU-210 were normalized to 1, and baseline amounts of neurite outgrowth were normalized to 0 (7). Treatment of cells with an SC siRNA (SC siRNA HU) resulted in similar amounts of cannabinoid-induced neurite outgrowth as Luc siRNA (Luc siRNA HU). Error bars, mean ± SEM (n = 3 independent experiments), *, P < 0.01 (student's t test) versus Luc siRNA HU-210 control. (B) Regulation of Stat3 localization. Neuro2A cells were transfected with Luc siRNA or BRCA1 siRNA and treated with DMSO or HU-210 for 20 min. Cells were fixed, permeablized, and stained with Stat3 antibodies. Purple and yellow arrows indicate cytosolic and nuclear Stat3 localization, respectively. Scale bars indicate distance in micrometers. Nuclei were visualized with Hoescht stain. (C) Decreased BRCA1 expression in response to CB1R stimulation. Neuro2A cells were stimulated with HU-210 and RNA was isolated at the indicated times. Quantitative real-time RT-PCR was performed as described in the SOM (7). Error bars, mean ± SEM (n = 4 independent experiments); *, P < 0.01 (Student's t test) versus 0 min control.

Fig. 4

Regulation of neuronal differentiation by BRCA1. (A) BRCA1 regulates neurite outgrowth in rat primary hippocampal neuron cultures. Hippocampal cultures were transfected with Luc siRNA or BRCA1 siRNA after plating and adhesion. Cells were fixed 30 hours after transfection. The mean number of processes per cell in each field was analyzed morphometrically. The mean number of processes for neurons treated with Luc siRNA was normalized to 1. Error bars, mean ± SEM (n = 10 wells of four fields per well for each experiment set); *, P < 0.05 (Student's t test) versus Luc siRNA control. (B) BRCA1 regulates synaptic density in hippocampal neuron cultures. Neurons were transfected with Luc siRNA or BRCA1 siRNA using NeuroPORTER after 3 days in cultures. Four days later, cells were fixed and stained with synaptophysin (red) and β-tubulin (blue) antibodies. Yellow arrows denote synaptophysin puncta. Quantification is shown in fig. S7B.

Fig. 5

Effects of PI3K-Akt signaling to PAX6 on CB1R-induced neurite outgrowth. (A) Phosphorylation of Akt. Neuro2A cells were stimulated with 2 μM HU-210 for the indicated times in the absence (−LY) or presence of (+LY) the PI3K inhibitor LY. Cells were lysed and immunoblot analysis was performed with antibodies to phospho Akt (pAkt) or total Akt. pAkt levels were normalized to total Akt. Error bars, mean ± SEM (n = 3 independent experiments); *, P < 0.05 (Student's t test) versus 0 min control. (B) Effects of pharmacological inhibitors on neurite outgrowth. Neuro2A cells were either treated with the indicated inhibitors or transfected with DN CREB and then stimulated with HU-210 to induce neurite outgrowth. Error bars, mean ± SEM (n = 4 independent experiments for PD, n = 3 for all others); *, P < 0.02 (Student's t test) versus DMSO control; **, P < 0.01 (Student's t test) versus pcDNA3 control. The figure is a composite of multiple experiments. DN CREB + LY was performed with the other DNA transfections in Fig. 1B. (C) Gel shift assay of PAX6 binding. Neuro2A cells were stimulated with 2 μM HU-210 for the indicated times in the absence (−LY) or presence (+LY) of LY. Nuclear extracts were prepared and gel shift assays were performed with oligonucleotides containing consensus-binding sites for PAX6. (D) PAX6 phosphorylation in rat primary hippocampal neuron cultures. Hippocampal cultures were stimulated with 1 μM HU-210 for the indicated times in the absence (−LY) or presence (+LY) of 10 μM LY 294002. Cells were lysed, and immunoprecipitations were performed with rabbit antibodies to PAX6 or rabbit immunoglobin G as a control. Immunoblot analysis was performed with mouse antibodies directed against phospho-Thr. Phospho-Thr levels were normalized to PAX6 after stripping and re-probing the blots with PAX6 antibodies. Values were averaged from two independent experiments. (E) Simple schematic of signal flow through Src, MAPK, and PI3K during neurite outgrowth. CB1R stimulation by HU-210 (HU) activates the alpha subunits of G_i and G_o (α_i/o) and leads to activation of Stat3 through the kinase Src. BRCA1 is depicted in blue. The putative interaction between BRCA1 and Akt is shaded gray (38).

Fig. 6

Integration of activated transcription factors with the upstream signaling network during CB1R-induced neurite outgrowth. (A) Pharmacological inhibition of transcription factor activation during CB1R stimulation. The Venn diagram (data from fig. S2B) shows the inhibition of transcription factor activation by the PI3K inhibitor LY (blue), the MAPK pathway inhibitor PD (yellow), and the Src inhibitor PP2 (magenta). (B) Proposed decision logic for cell-state change during CB1R-stimulated neurite outgrowth. A set of three pOR gates connects the alpha subunits of G_i and G_o (Gα_i/o for clarity) and the Gβγ subunit to the kinases PI3K, Src, and MAPK. This study's results and the experimental literature (table S6) suggest that these gates are not true OR gates and are thus represented as pOR gates (see fig. S13 for details). Stacked below are three AND gates that connect the kinases to the transcription factors. The components and connections are in black. The gray arrows and gate symbols are in gray to denote information flow and the abstract nature of the pOR and AND gates.