Survey
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project
Insulin-like Growth Factor Binding Protein 5 (IGFBP5) as a putative biomarker for breast cancer risk 20th February 2015 Olivia Fletcher Group leader, Genetic Epidemiology Making the discoveries that defeat cancer Structure of my talk • Background to the 2q35 breast cancer risk locus • Identifying target(s) • Tissue specificity of IGFBP5 expression • Developing IGFBP5 assays in (i) FFPE tumour microarrays • (ii) Plasma • Some preliminary results • Statistical issues • What next? 2 3 rs13387042 (Chr2:217,905,832) is associated with a 20% increased risk of ER+ breast cancer, P=1.3 x 10-13 Replicated in subsequent GWAS and by the Breast Cancer Association Consortium Stacey, S et al; Nature Genetics 2007 Milne, R et al; J Natl Cancer Inst 2009 2q35 risk locus – a gene dessert 4 Ghoussaini, M et al; Nature Communications 2014 Identifying target(s) Expression Quantitative Trait Locus (eQTL) analysis Physical interactions between regulatory elements and the genes they influence (chromosome conformation capture methods) Both methods IGFBP5 (protein coding gene) DIRC3 (long non-coding RNA) 5 IGFBP5 expression IGFBP5 is expressed and secreted by breast epithelial cells. eQTL and Chromosome Conformation Capture are reflecting local expression. IGFBP5 is also expressed by the liver (endocrine) Liver expression responsible for (circulating) plasma IGFBP5 levels Your average epidemiological study has – • buffy coat, • plasma/serum, • possibly urine • but not breast epithelial cells (at least in controls) 6 The plan! Measure IGFBP5 in (pre-diagnostic) plasma samples from incident cases and matched controls Measure IGFBP5 in tumour material from the cases Qus (i) Are levels of IGFBP5 in tumours correlated with levels in plasma (in cases)? (ii) Are plasma levels of IGFBP5 associated with subsequent risk of breast cancer? (iii) Are plasma levels of IGFBP5 associated with established risk factors for breast cancer? 7 Breakthrough Generations Study Population based cohort study Over 100,000 female volunteers age 16 or over UK- based Recruited between 2003 and 2013 Questionnaire data Blood samples – entry and after 6 years (subset) ~ 5,000 prevalent cases ~ 1,250 incident cases (excluding DCIS) PIs Swerdlow/Ashworth 8 Measuring IGFBP5 in tumours Immunohistochemistry (IHC) staining of tissue microarray sections from one ER-positive and one ER-negative tumour 9 10 Two commercially available IGFBP5 assays…….. A DELFIA standard curve (R&D systems ELISA kit) B 10 80000 OD = 450nM Europium counts 100000 ELISA standard curve (Raybiotech ELISA kit) 60000 40000 1 0.10 20000 1 10 100 IGFBP5 (ng/ml) 1000 0.01 0.1 1 10 IGFBP5 (ng/ml) 100 1000 Our IGFBP5 standard curve 11 Pilot data (1) Assayed 224 cases and 223 controls frequency matched on Age, Year of entry Ethnicity Number of days the blood sample was in the post Assayed samples in triplicate, and took the mean of the three readings. Re-assayed 5 samples with IGFBP5>900ng/ml 1 sample with undetectable levels 21 samples with CV>20% 14 random samples Intra-assay CV=6.2%, Inter-assay CV=12.8% 12 13 Pilot data (2) IGFBP5 ng/ml Comparing cases and controls (1) Mean plasma IGFBP5 in Controls – 243.0 ng/ml (206.3-279.8) Cases – 262.9 ng/ml (226.1 – 299.7) P=0.08 But excluding the two outlier cases Controls – 249.1 ng/ml (223.0-275.1) Cases – 256.9 ng/ml (230.8 – 282.9) P=0.34 14 15 Comparing cases with controls (2) Controls N (%) Cases N (%) OR (95% Ci) ORadj (95% Ci) Q1 (63.3 – 174.2) 55 (24.8) 47 (21.2) 1.00 1.00 Q2 (174.3 – 239.3) 56 (25.2) 66 (29.7) 1.38 (0.81 – 2.34) 1.38 (0.77 – 2.48) Q3 (239.4 – 307.7) 55 (24.8) 49 (21.1) 1.04 (0.60 – 1.80) 1.11 (0.58 – 2.12) Q4 (307.8 – 986.4) 56 (25.2) 60 (27.0) 1.25 (0.74 – 2.14) 1.35 (0.69 – 2.66) Ptrend=0.67 Padj=0.55 *three outliers (women whose plasma IGFBP5 levels were >5 standard deviations from the mean), were excluded 16 Association with age at FFTP Number of women* Mean IGFBP5 (95% CI) <20 17 222.3 (171.7 – 273.0) 20-24 132 248.8 (230.6 – 267.0) 25-29 163 249.5 (233.1 – 265.8) 30-34 53 259.8 (231.2 – 288.5) ≥35 16 248.6 (196.4 – 300.8) P value Age at FFTP (years) Ptrend=0.40 (Padj=0.50) * three outliers (women whose plasma IGFBP5 levels were >5 standard deviations from the mean), one woman for whom IGFBP5 levels were undetectable and 12 women for whom the duration of their only pregnancy/pregnancies was less than 26 weeks are excluded from these analyses. 17 Association with breast feeding Number of women** Mean IGFBP5 (95% CI) None 70 247.6 (222.4 – 272.8) <6 138 254.3 (236.5 – 272.1) 6 -11 60 250.6 (223.6 – 277.6) 12 - 24 82 249.9 (226.5 – 273.4) ≥24 29 233.7 (194.2 – 273.3) P value Duration of breast feeding (months) **data on breastfeeding was missing for six women Ptrend=0.64 (Padj=0.58) 18 Association with parity Number of women Mean IGFBP5 (95% CI) Non-parous 51 259.8 (214.0 – 305.7) Parous 381 249.4 (239.0 – 259.8) 1 child 40 291.1 (258.5 – 323.8) 2 children 213 250.3 (236.2 – 264.5) 3 children 97 235.5 (214.6 – 256.5) ≥4 children 31 233.0 (195.8 – 270.1) P value Parity P=0.50 (Padj=0.85) Ptrend=0.01 (Padj=0.02) Statistical issues Outliers - to re-assay or not - remove from quantitative analyses? - include in analyses of quartiles? Power – 1,250 cases and 1,250 controls, 70% power at 5% significance for OR=1.35 (comparing top and bottom quartiles) • Applied for funding to assay an additional 1,026 cases and 1,026 controls • Temporal variation – assay plasma samples taken from the same woman five years apart (subset of 62 women) 19 Acknowledgments ICR - Gene Function Laura Broome Nichola Johnson Richard Elliott Chris Lord Alan Ashworth Bioscale Colin Merrifield Breakthrough Generations Study Tony Swerdlow Montse Garcia-Closas University of Bristol Nick Orr Jeff Holly Minouk Schoemaker Claire Perks Michael Jones Caroline Jarrret Katarzyna Tomczyk