ESPE Abstracts (2019) 92 FC12.5

Integrated Analysis of Baseline Blood Transcriptome and Genome Identifies Clusters of Turner Syndrome Patients with Different Responses to Recombinant Human Growth Hormone

Robert Sellers1, Amina Amin1, Kajal Patel1, Terence Garner1, Andrew Whatmore1, Ekaterina Koledova2, Philip Murray1, Pierre Chatelain3, Peter Clayton1, Adam Stevens1

1University of Manchester, Manchester, United Kingdom. 2Merck Healthcare KGaA, Darmstadt, Germany. 3Université Claude Bernard, Lyon, France

Responsiveness to recombinant human growth hormone (rhGH) treatment in Turner syndrome (TS) is highly variable. Previous research has characterised genetic variants associated with rhGH response but these only have a minor impact. The relationship of these genetic variants to the blood transcriptome is unknown. The aim of this analysis was to relate unsupervised baseline blood transcriptome and genetic data from TS patients to their phenotype, karyotype and responsiveness to rhGH.

Data on 91 TS patients from the PREDICT study were analysed. Patients were assessed for change in clinical biomarkers over the first month of treatment including IGF-I, IGFBP-3, triglyceride levels and insulin resistance (measured by homeostatic model assessment). Annual change in height was measured for five years after the commencement of rhGH treatment. Pharmacogenomic analysis was performed using 11 single nucleotide polymorphisms (SNPs) previously associated with rhGH response, along with baseline transcriptome (mRNA) from blood. Unsupervised analysis of the transcriptome was conducted using Principal Component Analysis (PCA). Patients were clustered by expression profiles using similarity network fusion (SNF). Genetic and transcriptomic data were linked by defining expression quantitative trait loci (eQTL).

Using 9963 transcriptomic probes, 91 TS patients were clustered by PCA. Three patient clusters (Clusters 1, 2 and 3; n=31, 32, 28) were identified using SNF. A profile of 7851 differential expressed genes (DEGs) defined these patient clusters (0.05>q>1.8x10-20; q=false-discovery modified p-value). For the first year of rhGH treatment, Cluster 3 displayed greater height velocity (HV1) in comparison to other clusters (Clusters 1, 2 and 3; mean HV1 (cm/yr ± standard deviation[SD])=7.1±0.9, 7.2±1.2, 8.6±1.6; P<0.009), as well as a greater decrease in fasting triglyceride over the first month of treatment (P=0.014). Total growth over the five year study was significantly elevated for Cluster 3 (Clusters 1, 2 and 3; mean growth (cm ± SD)=25.7±0.8, 28.9±3.3, 31.1±4.3; P=0.04). Cryptic Y material, not detected by karyotyping, was defined by transcriptomic analysis. eQTL mapping using the 11 SNPs, identified 849 transcriptomic associations (0.02>p>5.7x10-8); 409 (48%) of which were associated with DEGs in Cluster 3 (2.7 fold-enrichment, P<1x10-6).

A baseline blood transcriptome profile can be used to identify TS patients with 'good' growth and metabolic response to rhGH. In addition genetic variants known to associate with response to rhGH were linked to levels of gene expression in this group. Integrated transcriptomic and genomic analysis represents a novel approach to the personalisation of rhGH treatment for TS.