Analysing the relationship between immune infiltration and tissue architecture in high grade serous ovarian cancer

The aim of this project was to understand the distribution and compartmental effects of immune populations and the tissue architecture that underlies them in high grade serous ovarian cancer (HGSOC). The quantitative distribution of three key immune infiltrates was initially assessed in the SEARCH cohort (n=332); CD8+ cytotoxic T-cells, CD45RO+ memory T-cells and CD68+ macrophages. The stromal infiltration of CD68+ (HR = 0.44, p = 3×10−4 ) and CD45RO+ (HR = 0.76, p = 7×10−4 ) cells was positively associated with survival and the positive prognostic value of CD8 + cells increased when the density was averaged across both stroma and tumour compartments(HR = 0.79, p = 6 × 10 −4 ). Multiple correlated immune infiltrates were combined to view patterns across samples. This analysis was limited to the subset of patients with both tumour and stroma present(n=152). A multi-dimensional view of these immune populations (an ‘immunospace’) was obtained and the principal components of this space were analysed and used to build complex survival models. The largest variation between patient samples was the strength of a coordinated multi-cell immune response that was associated with improved survival (HR = 0.88, p = 0.016). Tumour architecture was investigated by deriving metrics from the positions of epithelial cells in samples in the OV04 cohort (n=101). Both multiplexed pancytokeratin(CK), FOXP3 and CD8 staining and H&E serial sections were used for the localisation of epithelial cells. The nearest neighbour distance between epithelial cells was significantly correlated across serial sections (R = 0.65, p = 2.7 × 10 −5 ), between sections from different cores from the same tissue block (R = 0.58, p = 0.00064) and sections from differing tumour sites (R = 0.78, p = 7.8 × 10 −5 ). This measure of epithelial cell packing is therefore tumour-intrinsic. Correlations were also examined between these structural metrics and the FOXP3 and CD8 immune infiltration. Epithelial CD8 + density was strongly correlated with epithelial nearest neighbour distance (R = 0.91, p = 1 × 10 −14 ) and moderately with border cell percentage and epithelial fraction of the tumour. Regulatory T-cell (FOXP3 + ) infiltration was positively correlated with the number of epithelial clusters across a core. The conservation of structural measures across tumours and the link between structure and immune infiltrate supported the physical border hypothesis of immune exclusion. It was hypothesised that the link between these two features was collagen deposition. Collagen structure was assessed via dual IF/SHG imaging. Structural metrics were correlated with collagen Energy, Correlation and Coherency features. There was no correlation between these collagen features and immune infiltration. Imaging Mass Cytometry was carried out on the ICON7 cohort (n=313) to produce a novel data set. Preliminary analyses of these images were carried out to further investigate links between collagen, infiltrates and structure. The thesis presents novel data and numerous novel conclusions regarding the prognostic value of specific immune infiltrate variables and quantitative links between tissue structure and immune response in HGSOC.

Name dc:type.qualificationname

Doctor of Philosophy (PhD)

Level dc:type.qualificationlevel

Doctoral

Grantor dc:publisher.institution

University of Cambridge

Year dc:date.issued

2021

Author dc:creator

• Barker-Clarke, Rowan

Advisor dc:contributor.advisor

• Brenton, James

• ovarian cancer
• tumour microenvironment
• immune oncology
• cancer imaging
• machine learning

Licence

• https://apollo8-f-pro.lib.cam.ac.uk/bitstreams/897e8286-5bbd-48a1-8932-f5415a2078d1/download
• https://www.rioxx.net/licenses/all-rights-reserved/

Language dc:language

eng

Author Identifier

0000-0002-5738-6683

OAI identifier oai:identifier

oai:www.repository.cam.ac.uk:1810/310110