Synaptic Subgroup Identification - a high content method for synaptic disease modeling and its future applications

Abstract

Cortical interneurons establish a broad, distributed network of local inhibitory circuits. However, because interneuron synapses are so numerous and widespread, new tools needed to be developed that provide high-content information for each synapse beyond standard metrics (size, shape, etc.), while also assessing overall interneuron synaptic coverage. To meet both of these requirements, we recently developed a novel image-based machine learning method to analyze cortical interneuron synapses at scale, an approach we term Synaptic Subtype IDentification (SSID). Using SSID, we extract hundreds of spatial and intensity features from each synapse imaged, thereby generating a high content, quantitative, deep phenotyping read-out of interneuron synaptic state across conditions. This rich dataset allows us to 1) Identify previously unrecognized synaptic subgroups; 2) Quantify each synaptic subgroup’s relative representation in the whole population; 3) Determine the spatial distribution of synaptic subgroups within large regions of interest (eg. within a cortical column). In our preliminary data, we show that applying SSID to a FXS mouse model identifies deficits in both the relative representation of synaptic subgroups, as well as the spatial arrangement of interneuron synapses.  However, we expect that FXS will manifest differently between mice and humans. Therefore, we have developed a strategy to use SSID to directly compare synaptic effects of FXS across cognate mouse and human interneurons. To model FXS in interneurons, we developed a rapid and highly efficient direct reprogramming system to generate human induced neurons (iiNs) from iPSCs using transcription factors (TFs). This approach, which stands in contrast to traditional extrinsic factor-driven directed differentiation, offers several key advantages. Our collaborators  established high-efficiency, inducible iPSC-iiN cell lines where the reprogramming TFs are inserted into the genome via piggyBac (PB) transposons. These constructs are finely controlled, with each TF driven by a doxycycline-inducible promoter (TRE), alongside the necessary doxycycline-dependent TF (rTTA). This precise control is enables us to generate iiNs of defined cell lineages with up to 95% efficiency, with level of maturity comparable to those of cultured endogenous mouse cortical neurons.