Fig. 1

High-content screening and data analysis workflow. a E18 embryonic mouse neurons were seeded in 96-well cell culture plates and infected with lentiviral particles containing shRNA constructs against the 41 genes of interest (4–5 shRNAs per gene). Neurons were only cultured in the inner 60 wells of a 96-well plate and infected on day 1 in vitro (DIV1). On each plate, at least three wells were infected with a scrambled control shRNA, three wells with a positive control shRNA (against NLGN1), and at least three wells were left uninfected. The remaining wells were infected with ~ 50 different experimental shRNAs. Controls and experimental shRNAs were always in a randomized order as to minimize plate position effects. Cultures were fixed at DIV7, DIV14, or DIV21. For each time point, shRNA replicates (n = 3) were divided over three different culture plates as to minimize plate effects. b Neurons were stained with a nuclear marker (Hoechst), a dendritic marker (anti-MAP2), a presynaptic marker (anti-synapsin) and a postsynaptic marker (anti-PSD-95) and imaged using automated confocal high-content microscopy. Neurons were first imaged at × 10 magnification (panel 1) to determine neuron numbers (panel 2; MAP2-positive neurons in green; MAP2-negative cells in red), and total dendrite length and numbers of primary dendrites and branch points (panel 3; selected neurons in red, traced dendrites in green). Neurons were subsequently imaged at × 40 magnification (panel 4) to quantify presynaptic puncta (panel 5; selected puncta in red), postsynaptic puncta (panel 6; selected puncta in green), and colocalized pre-and postsynaptic puncta (panel 7; colocalized puncta in yellow). Images are representative examples of DIV14 neurons. Examples of neurons at all DIV are included in Supplementary Fig. S2. Scale bars: 100 μm (panels 1–3), 20 μm (panels 4–7). c Based on these primary measurements, 10 core parameters were derived that measure relevant aspects of neuronal viability and survival, neuronal network formation, and synaptic connectivity. d Data were checked for batch, plate, and edge effects and outliers were removed. After normalization, multi-level exploratory data analysis was performed to determine the relative contribution of both experimental and technical sources of variation to overall variance in the data. Statistical analysis was then performed to detect significant gene-level effects in the data without prior removal of individual shRNAs. In parallel, robust biological effects were determined by filtering out individual shRNAs that produce inconsistent phenotypes that might represent off-target effects. Based on these two selection criteria candidate genes were selected for proteomics analysis and subsequent cellular pathway analysis