AR isoform 2 is widely expressed in human tissues and acts as a decoy receptor to regulate AR transcriptional activity

Together with the Estrogen, Progesterone, and Glucocorticoid Receptors, the Androgen Receptor (AR) is a member of the type I nuclear receptor superfamily, a class of transcription factors that are activated by steroid hormones1. Nuclear receptors share a common three-domain structure, comprising a highly variable amino-terminal domain (NTD), with potent transcriptional activator capability, a central two zinc finger motifs DNA-binding domain (DBD), and a well-conserved carboxy-terminal ligand-binding domain (LBD) responsible for interacting with the transcriptional machinery1. Given their modular structure, expression of isoforms resulting from alternative splicing or starting and termination events have profound effects on nuclear receptor activity and function. Several AR variants (AR-Vs) have been identified from prostate cancer clinical specimens, arising from multiple mechanisms, including genomic rearrangements2 and upregulation of splicing factors3. This seminal discovery has provided the molecular basis for a deeper understanding of AR-related diseases, with potential to be translated in identification of novel therapeutic targets and promising prognostic markers4–6. By deeply sequencing the transcriptome of human brains, here we identified AR-2 as the most expressed AR isoform. Furthermore, we found that AR-2 is widely expressed in other human tissues, forms heterodimers with the full length AR and acts as a repressor of its transcriptional activity. Delivery of this isoform via adeno-associated virus (AAV9) resulted in significant amelioration of the disease phenotype in a mouse model of spinal and bulbar muscular atrophy (SBMA), a neuromuscular disease with unmet clinical need caused by a toxic gain of function of polyglutamine-expanded (polyQ) AR7. Collectively, these results establish a physiological role of AR-2 in fine-tuning full length AR activity and demonstrate the feasibility and therapeutic benefit of modulating this variant for treating SBMA and other AR-related diseases. RNA sequencing was performed on a total of 26 human samples. Of these samples, three represented three different brain regions for one SBMA patient, five represented three different brain regions for two neurological controls, six represented three different brain regions for two non-neurological controls, six represented iPSC-derived motor neurons from three different SBMA patients cultured under EtOH or DHT condition, and six represented iPSC-derived motor neurons from three controls cultured under EtOH or DHT condition. Neurological controls = brain samples from patient(s) that have a diagnosed neurological disease/disorder other than SBMA NonNeurological_Controls = brain samples from patients that do not have a diagnosed neurological disease/disorder, but rather represent brain with non-disease/non-disorder neurological deficits associated with age.