Unsupervised Multi-Scale Analysis for the Identification of Placental Subtypes of Human Preeclampsia and Fetal Growth Restriction

Preeclampsia (PE) and fetal growth restriction (FGR) are two of the most common pathologies of pregnancy, both thought to be primarily driven by placental dysfunction. Despite decades of research into the underlying etiologies of these disorders, as well as potential biomarkers and treatments, no single discovery has been found to be applicable to the entire clinical spectrum of PE or FGR patients, likely due to the existence of multiple disease subtypes. Therefore, the main goal of this thesis was to investigate if the application of unsupervised clustering techniques to placental gene expression data could elucidate transcriptional subtypes of PE and FGR with increased clinical and histopathological homogeneity. Clustering of three overlapping large microarray datasets revealed 3-5 molecular clusters, depending on the study. However, three subtypes of PE placentas were consistently identified within clusters 1-3, and, eventually, each co-clustered with a group of placentas from normotensive suspected FGR pregnancies. Within cluster 1, PE and suspected FGR samples demonstrated less severe clinical outcomes, molecular similarity to healthy term controls, and either no placental lesions or mild histopathology, suggesting a dominant non-placental source of the disease. Cluster 2 PE and FGR placentas revealed overwhelming evidence of “canonical” maternal vascular malperfusion features and increased placental expression of hypoxia and hormone activity genes. Cluster 3 PE and FGR samples displayed signs of an “immunological” pathology, with a transcriptional signature of immune response, apoptosis, and cytokine activity, and histological lesions affiliated with allograft rejection, such as massive perivillous fibrin deposition. In the largest microarray dataset (N=330), two additional clusters were discovered. Cluster 4 samples were preterm controls with histological chorioamnionitis, while cluster 5 was associated with confined placental mosaicism, but no clinical or histological cohesion. Furthermore, specific differences in the expression of three genes by qPCR were found to be sufficient for separating placentas into transcriptional clusters 1-4, which will allow future studies to focus on the identification of subtype-specific biomarkers and therapeutics for PE/FGR without having to first cluster microarray data. Matched maternal samples will also be necessary to comprehend the development of hypertension in some patients but not others with similar placental profiles.

Department dc:contributor.department

Physiology

Year dc:date.issued

2018

Author dc:creator

• Leavey, Katherine Michelle

Advisor dc:contributor.advisor

• Cox, Brian J

• fetal growth restriction
• gene expression
• histopathology
• placenta
• preeclampsia
• subtypes

Handle dc:identifier.uri

http://hdl.handle.net/1807/97890

OAI identifier oai:identifier

oai:utoronto.scholaris.ca:1807/97890