Endocrine Abstracts

Background: GH deficiency (GHD) has a spectrum of severity as characterised by GH stimulation tests; the cut-off level for GHD has been a long standing contentious issue. An independent biological correlate of severity would be valuable.

Objectives: To identify patterns of gene expression (GE) that correlate with severity in GHD.

Methods: Pre-pubertal children with GHD (n=72) were enrolled from the PREDICT study (NCT00256126). GHD severity was defined using two standard GH stimulation tests with peak GH levels (pGH) <10 μg/l (range 0.2–9.3 μg/l). Whole blood GE was determined prior to treatment using Affymetrix U133v2 microarrays. GE was correlated with pGH using rank regression (gender, ethnicity, age, and BMI as co-variates]. Models of GHD severity based on network clusters were generated, which were then used to investigate mechanistic relationships (Moduland algorithm and causal network analysis (CNA) (Ingenuity Pathway Analysis software]). Associated biological functions were determined with the hypergeometric test.

Results: Rank regression identified 1631 genes that were correlated with pGH: 619 positively (R>+0.28) and 1012 negatively related (R<−0.28) (P<0.05). These genes were causally associated with pathways related to ‘failure of growth’ (P<2.4×10⁻³). Sub-clusters of similar gene expression patterns could be recognised in GHD children with pGH >7 μg/l (borderline GHD) and pGH <2 μg/l (very severe GHD) and to a lesser extent for pGH ≥2–≤5 μg/l (severe GHD) and >5–≤7 μg/l (moderate GHD). In the model of negatively correlated GE, functional clusters were causally related to the cyclin binding gene, CUL3 and in the model of positively correlated GE to the cyclin dependent kinase gene, CDK9 with both clusters causally associated with genes related to body size (P<2.8×10⁻³).

Conclusions: This study has demonstrated a relationship between gene expression patterns and pGH; this could provide biological correlates and mechanisms for the range of severity of childhood GHD.