Phenotypic rescue via mTOR inhibition in neuron-specific Pten knockout mice reveals AKT and mTORC1-site specific changes

Phosphatase and Tensin Homolog (PTEN) is a dual-specific protein and lipid phosphatase that regulates AKT and downstream signaling of the mechanistic target of rapamycin (mTOR). PTEN functions as a tumor suppressor gene whose mutations result in PTEN Hamartoma Tumor Syndrome (PHTS) characterized by increased cancer risk and neurodevelopmental comorbidity. Here, we generated a novel neuron-specific Pten knock-out mouse model (Syn-Cre/Pten HOM) to test the ability of pharmacologic mTOR inhibition to rescue Pten mutation-associated disease phenotypes in vivo and in vitro. We found that treatment with the mTOR inhibitor, everolimus, increased the survival of Syn-Cre/Pten HOM mice while some neurologic phenotypes persisted. Transcriptomic analyses revealed that in contrast to mice harboring a neuron-specific deletion of the Tuberous Sclerosis Complex 2 gene (Syn-Cre/Tsc2 KO), genes that are under AKT regulation were significantly increased in the Syn-Cre/Pten HOM mice. In addition, genes associated with synapse, extracellular matrix, and myelination were broadly increased in Syn-Cre/Pten HOM mouse neocortex. These findings were confirmed by immunostaining of cortical sections in vivo, which revealed excessive immunoreactivity of myelin basic protein and perineuronal nets (PNN), the specialized extracellular matrix surrounding fast-spiking parvalbumin (PV) interneurons. We also detected increased expression of Synapsin I/PSD95 positive synapses and network hyperactivity phenotypes in Syn-Cre/Pten HOM mice neurons compared to wild-type (WT) neurons in vitro. Strikingly, everolimus treatment rescued the number of synapses and network hyperactivity in the Syn-Cre/Pten HOM mice cortical neuron cultures. Taken together, our results revealed in vivo and in vitro molecular and neuronal network mechanisms underlying neurological phenotypes of PHTS. Notably, pharmacologic mTOR inhibition by everolimus led to successful downstream signaling rescue, including mTOR complex 1 (mTORC1) site-specific suppression of S6 phosphorylation, correlating with phenotypic rescue found in our novel neuron-specific Syn-Cre/Pten HOM mice. Pten homozygous and heterozygous mutant mice of the indicated genotypes were sacrificed at P45 (n=4 HET, 4 HOM), and the cortex was collected. RNA extraction was performed using TRIzol (Invitrogen). Total RNA was used as an input for cDNA library preparation using the NEBNext Ultra II RNA Library Prep Kit by Illumina, as per manufacturer’s protocols. Sequencing was performed on an Illumina analyzer, and 150bp paired-end reads were obtained. Reads passing quality filters were mapped using Hisat2 (http://daehwankimlab.github.io/hisat2/; v2.0.5) onto the mouse genome. For differential expression, raw gene counts from protein coding genes from heterozygous and homozygous animals were used as input for EdgeR (https://bioconductor.org/packages/release/bioc/html/edgeR.html). Functional annotation was performed using DAVID (https://david.ncifcrf.gov/summary.jsp), and redundant or non-specific categories were manually removed. Comparison with ASD candidate genes was performed by comparing differentially expressed genes to SFARI genes (category 1-3). For RNA sequencing of Tsc2 model (previously described by Yuan and colleagues29), mice from the indicated genotypes were sacrificed at P30 and frontal cortex was dissected. Total RNA was isolated using a Qiagen RNAeasy kit, and cDNA sequencing libraries were constructed with the Kapa Hyperprep library kit using polyA selection (Roche). Sequencing was performed on an Illumina analyzer, and 75bp single-end reads were obtained. Reads passing quality filters were mapped onto the mouse genome using Hisat2. For WGCNA, the WGCNA package from CRAN was used, and gene counts from protein coding genes from both Pten mutant and Tsc2 mutant mice were normalized using log2 counts per million separately. Co-expression networks were created from genes with a coefficient of variation > 0.005 using b=6, and clusters were isolated using the cutreeDynamic package. Module overlap was determined by calculating the Pearson correlation of the kME between two modules.