Skip to main content

Thank you for visiting nature.com. You are using a browser version with limited support for CSS. To obtain the best experience, we recommend you use a more up to date browser (or turn off compatibility mode in Internet Explorer). In the meantime, to ensure continued support, we are displaying the site without styles and JavaScript.

Age-correlated protein and transcript expression in breast cancer and normal breast tissues is dominated by host endocrine effects

Abstract

The magnitude and scope of intrinsic age-correlated and host endocrine age-correlated gene expression in breast cancer is not well understood. From age-correlated gene expression in 3,071 breast cancer transcriptomes and epithelial protein expression of 42 markers in 5,001 breast cancers and 537 normal breast tissues, we identified a majority of age-correlated genes as putatively regulated by age-dependent estrogen signaling. Surprisingly, these included genes encoding the chromatin modifier EZH2 (which had a negative age correlation) and associated H3K27me3 (which had an inverse, positive age correlation). Among The Cancer Genome Atlas lung, thyroid, kidney and prostate transcriptomes, the largest overlap with breast cancer in age-correlated transcripts was lung cancer, for which about one-third of overlapping age-correlated transcripts appeared to be estrogen regulated. Age-quartile-stratified outcomes analysis of 3,500 breast cancers using EZH2, H3K27me3, FOXA1 and BCL2 proteins revealed distinct age-related prognostic significance. Age correlation in gene expression may thus be an important factor in ER, EZH2, H3K27me3 and other biomarker assessment and treatment strategies.

Access options

Rent or Buy article

Get time limited or full article access on ReadCube.

from$8.99

All prices are NET prices.

Fig. 1: Data sources and age distributions.
Fig. 2: Age-dependent transcript volcano plots and expression patterns.
Fig. 3: Age-dependent transcript volcano plots.
Fig. 4: Age-associated transcripts and their ER binding site status.
Fig. 5: Age-correlated IHC measures.
Fig. 6: Kaplan–Meier survival curves.

Data availability

TCGA data (-BRCA, -PRAD, -KIRC, -LUAD and -THCA) are available at https://portal.gdc.cancer.gov/projects/. METABRIC data are available at https://ega-archive.org/studies/EGAS00000000083. The raw data and the Vancouver Big Series 3,991-case TMA and normal breast 537-case TMA data used to generate all of the results are available from the GitHub (https://github.com/BCCRCMO/BrCa_AgeAssociations) and Zenodo repositories (https://doi.org/10.5281/zenodo.3715548)62. The raw images from IHC that were applied to TMAs are available on the BC Cancer Research Institute servers described in the supplementary information63,64,65,66,67,68,69. All of the ChIP-Seq data generated in the Zwart et al.42 study and the RNA-Seq data are available from the Gene Expression Omnibus repository (accession codes GSE104399 and GSE104730, respectively). Methodology information is available in the associated publication25. Expression data fold-change estimates and age-associated P values for the 18,950 protein-coding genes for the 60 years age group, as well as all of the cases of the METABRIC and TCGA cancer datasets, are available as Supplementary Data 1.

Code availability

All code and algorithms used to generate the figures and results are available from the GitHub (https://github.com/BCCRCMO/BrCa_AgeAssociations) and Zenodo repositories (https://doi.org/10.5281/zenodo.3715548)62.

References

  1. 1.

    Benz, C. C. Impact of aging on the biology of breast cancer. Crit. Rev. Oncol. Hemat. 66, 65–74 (2008).

    Google Scholar 

  2. 2.

    Anderson, W. F., Jatoi, I. & Devesa, S. S. Distinct breast cancer incidence and prognostic patterns in the NCI’s SEER program: suggesting a possible link between etiology and outcome. Breast Cancer Res. Treat. 90, 127–137 (2005).

    PubMed  Google Scholar 

  3. 3.

    Nixon, A. J. et al. Relationship of patient age to pathologic features of the tumor and prognosis for patients with stage I or II breast cancer. J. Clin. Oncol. 12, 888–894 (1994).

    CAS  PubMed  Google Scholar 

  4. 4.

    Jenkins, E. O. et al. Age-specific changes in intrinsic breast cancer subtypes: a focus on older women. Oncologist 19, 1076–1083 (2014).

    PubMed  PubMed Central  Google Scholar 

  5. 5.

    Quong, J. et al. Age-dependent changes in breast cancer hormone receptors and oxidant stress markers. Breast Cancer Res. Treat. 76, 221–236 (2002).

    CAS  PubMed  Google Scholar 

  6. 6.

    Sørlie, T. et al. Repeated observation of breast tumor subtypes in independent gene expression data sets. Proc. Natl Acad. Sci. USA 100, 8418–8423 (2003).

    PubMed  Google Scholar 

  7. 7.

    Harley, C. B., Futcher, A. B. & Greider, C. W. Telomeres shorten during ageing of human fibroblasts. Nature 345, 458–460 (1990).

    CAS  PubMed  Google Scholar 

  8. 8.

    Nakamura, K.-I. et al. Comparative analysis of telomere lengths and erosion with age in human epidermis and lingual epithelium. J. Invest. Dermatol. 119, 1014–1019 (2002).

    CAS  PubMed  Google Scholar 

  9. 9.

    De Lange, T. Telomere-related genome instability in cancer. Cold Spring Harb Symp Quant Biol. 70, 197–204 (2005).

    CAS  PubMed  Google Scholar 

  10. 10.

    Bailey, S. M. & Murnane, J. P. Telomeres, chromosome instability and cancer. Nucleic Acids Res. 34, 2408–2417 (2006).

    CAS  PubMed  PubMed Central  Google Scholar 

  11. 11.

    Gisselsson, D. et al. Telomere dysfunction triggers extensive DNA fragmentation and evolution of complex chromosome abnormalities in human malignant tumors. Proc. Natl Acad. Sci. USA 98, 12683–12688 (2001).

    CAS  PubMed  Google Scholar 

  12. 12.

    Hastie, N. D. et al. Telomere reduction in human colorectal carcinoma and with ageing. Nature 346, 866–868 (1990).

    CAS  PubMed  Google Scholar 

  13. 13.

    Kurabayashi, R. et al. Luminal and cancer cells in the breast show more rapid telomere shortening than myoepithelial cells and fibroblasts. Hum. Pathol. 39, 1647–1655 (2008).

    CAS  PubMed  Google Scholar 

  14. 14.

    Lee, J. K. et al. Age and the means of bypassing stasis influence the intrinsic subtype of immortalized human mammary epithelial cells. Front. Cell Dev. Biol. 3, 13 (2015).

    PubMed  PubMed Central  Google Scholar 

  15. 15.

    Miyano, M. et al. Age-related gene expression in luminal epithelial cells is driven by a microenvironment made from myoepithelial cells. Aging 9, 2026–2051 (2017).

    CAS  PubMed  PubMed Central  Google Scholar 

  16. 16.

    Zane, L., Sharma, V. & Misteli, T. Common features of chromatin in aging and cancer: cause or coincidence? Trends Cell Biol. 24, 686–694 (2014).

    CAS  PubMed  PubMed Central  Google Scholar 

  17. 17.

    Yau, C. et al. Aging impacts transcriptomes but not genomes of hormone-dependent breast cancers. Breast Cancer Res. 9, R59 (2007).

    PubMed  PubMed Central  Google Scholar 

  18. 18.

    Cui, J., Shen, Y. & Li, R. Estrogen synthesis and signaling pathways during aging: from periphery to brain. Trends Mol. Med. 19, 197–209 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

  19. 19.

    Curtis, C. et al. The genomic and transcriptomic architecture of 2,000 breast tumours reveals novel subgroups. Nature 486, 346–352 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  20. 20.

    Weinstein, J. N. et al. The Cancer Genome Atlas Pan-Cancer Analysis Project. Nat. Genet. 45, 1113–1120 (2013).

    PubMed  PubMed Central  Google Scholar 

  21. 21.

    Benjamini, Y. & Hochberg, Y. On the adaptive control of the false discovery rate in multiple testing with independent statistics. J. Educ. Behav. Stat. 25, 60–83 (2000).

    Google Scholar 

  22. 22.

    Phipps, A. I. et al. Defining menopausal status in epidemiologic studies: a comparison of multiple approaches and their effects on breast cancer rates. Maturitas 67, 60–66 (2010).

    PubMed  PubMed Central  Google Scholar 

  23. 23.

    Tam, C. Y. et al. Risk factors for breast cancer in postmenopausal Caucasian and Chinese–Canadian women. Breast Cancer Res. 12, R2 (2010).

    PubMed  PubMed Central  Google Scholar 

  24. 24.

    Collaborative Group on Hormonal Factors in Breast Cancer. Menarche, menopause, and breast cancer risk: individual participant meta-analysis, including 118 964 women with breast cancer from 117 epidemiological studies. Lancet Oncol. 13, 1141–1151 (2012).

    PubMed Central  Google Scholar 

  25. 25.

    Severson, T. M. et al. Characterizing steroid hormone receptor chromatin binding landscapes in male and female breast cancer. Nat. Commun. 9, 482 (2018).

    PubMed  PubMed Central  Google Scholar 

  26. 26.

    Parker, J. S. et al. Supervised risk predictor of breast cancer based on intrinsic subtypes. J. Clin. Oncol. 27, 1160–1167 (2009).

    PubMed  PubMed Central  Google Scholar 

  27. 27.

    Efron, B. & Tibshirani, R. J. An Introduction to the Bootstrap (Monographs on Statistics and Applied Probability No. 57) (Chapman & Hall/CRC, 1993).

  28. 28.

    Cheang, M. C. et al. Immunohistochemical detection using the new rabbit monoclonal antibody SP1 of estrogen receptor in breast cancer is superior to mouse monoclonal antibody 1D5 in predicting survival. J. Clin. Oncol. 24, 5637–5644 (2006).

    CAS  PubMed  Google Scholar 

  29. 29.

    Welch, W. J. Construction of permutation tests. J. Am. Stat. Assoc. 85, 693–698 (1990).

    Google Scholar 

  30. 30.

    Liao, S. et al. The molecular landscape of premenopausal breast cancer. Breast Cancer Res. 17, 104 (2015).

    PubMed  PubMed Central  Google Scholar 

  31. 31.

    Bernhardt, S. M. et al. Hormonal modulation of breast cancer gene expression: implications for intrinsic subtyping in premenopausal women. Front. Oncol. 6, 241 (2016).

    PubMed  PubMed Central  Google Scholar 

  32. 32.

    Pirone, J. R. et al. Age-associated gene expression in normal breast tissue mirrors qualitative age-at-incidence patterns for breast cancer. Cancer Epidemiol. Biomarkers Prev. 21, 1735–1744 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  33. 33.

    Dai, C., Heemers, H. & Sharifi, N. Androgen signaling in prostate cancer. Cold Spring Harb. Perspect. Med. 7, a030452 (2017).

    PubMed  PubMed Central  Google Scholar 

  34. 34.

    Handelsman, D. J., Sikaris, K. & Ly, L. P. Estimating age-specific trends in circulating testosterone and sex hormone-binding globulin in males and females across the lifespan. Ann. Clin. Biochem. 53, 377–384 (2016).

    CAS  PubMed  Google Scholar 

  35. 35.

    Chakraborty, S., Ganti, A. K., Marr, A. & Batra, S. K. Lung cancer in women: role of estrogens. Expert Rev. Resp. Med. 4, 509–518 (2010).

    CAS  Google Scholar 

  36. 36.

    Derwahl, M. & Nicula, D. Estrogen and its role in thyroid cancer. Endocr. Relat. Cancer 21, T273–T283 (2014).

    CAS  PubMed  Google Scholar 

  37. 37.

    Schveigert, D., Krasauskas, A., Didziapetriene, J., Kalibatiene, D. & Cicenas, S. Smoking, hormonal factors and molecular markers in female lung cancer. Neoplasma 63, 504–509 (2016).

    CAS  PubMed  Google Scholar 

  38. 38.

    Parkin, D. M., Bray, F., Ferlay, J. & Pisani, P. Global cancer statistics, 2002. CA Cancer J. Clin. 55, 74–108 (2005).

    PubMed  Google Scholar 

  39. 39.

    Carroll, J. S. et al. Genome-wide analysis of estrogen receptor binding sites. Nat. Genet. 38, 1289–1297 (2006).

    CAS  PubMed  Google Scholar 

  40. 40.

    Fullwood, M. J. et al. An oestrogen-receptor-α-bound human chromatin interactome. Nature 462, 58–64 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  41. 41.

    Ross-Innes, C. S. et al. Differential oestrogen receptor binding is associated with clinical outcome in breast cancer. Nature 481, 389–393 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  42. 42.

    Zwart, W. et al. Oestrogen receptor–co-factor–chromatin specificity in the transcriptional regulation of breast cancer. EMBO J. 30, 4764–4776 (2011).

    CAS  PubMed  PubMed Central  Google Scholar 

  43. 43.

    Garbe, J. C. et al. Accumulation of multipotent progenitors with a basal differentiation bias during aging of human mammary epithelia. Cancer Res. 72, 3687–3701 (2012).

    CAS  PubMed  PubMed Central  Google Scholar 

  44. 44.

    McCready, J., Arendt, L. M., Rudnick, J. A. & Kuperwasser, C. The contribution of dynamic stromal remodeling during mammary development to breast carcinogenesis. Breast Cancer Res. 12, 205 (2010).

    PubMed  PubMed Central  Google Scholar 

  45. 45.

    Yamaguchi, Y. et al. Tumor–stromal interaction through the estrogen-signaling pathway in human breast cancer. Cancer Res. 65, 4653–4662 (2005).

    CAS  PubMed  Google Scholar 

  46. 46.

    Cao, R. & Zhang, Y. The functions of E(Z)/EZH2-mediated methylation of lysine 27 in histone H3. Curr. Opin. Genet. Dev. 14, 155–164 (2004).

    CAS  PubMed  Google Scholar 

  47. 47.

    Kleer, C. G. et al. EZH2 is a marker of aggressive breast cancer and promotes neoplastic transformation of breast epithelial cells. Proc. Natl Acad. Sci. USA 100, 11606–11611 (2003).

    CAS  PubMed  Google Scholar 

  48. 48.

    Bachmann, I. M. et al. EZH2 expression is associated with high proliferation rate and aggressive tumor subgroups in cutaneous melanoma and cancers of the endometrium, prostate, and breast. J. Clin. Oncol. 24, 268–273 (2006).

    CAS  PubMed  Google Scholar 

  49. 49.

    Bhan, A. et al. Histone methyltransferase EZH2 is transcriptionally induced by estradiol as well as estrogenic endocrine disruptors bisphenol-A and diethylstilbestrol. J. Mol. Biol. 426, 3426–3441 (2014).

    CAS  PubMed  Google Scholar 

  50. 50.

    Bae, W. K. et al. The methyltransferase EZH2 is not required for mammary cancer development, although high EZH2 and low H3K27me3 correlate with poor prognosis of ER-positive breast cancers. Mol. Carcinogen. 54, 1172–1180 (2015).

    CAS  Google Scholar 

  51. 51.

    Wei, Y. et al. Loss of trimethylation at lysine 27 of histone H3 is a predictor of poor outcome in breast, ovarian, and pancreatic cancers. Mol. Carcinogen. 47, 701–706 (2008).

    CAS  Google Scholar 

  52. 52.

    Bate-Eya, L. T. et al. Enhancer of zeste homologue 2 plays an important role in neuroblastoma cell survival independent of its histone methyltransferase activity. Eur. J. Cancer 75, 63–72 (2017).

    CAS  PubMed  Google Scholar 

  53. 53.

    Kim, K. H. et al. SWI/SNF-mutant cancers depend on catalytic and non-catalytic activity of EZH2. Nat. Med. 21, 1491–1496 (2015).

    CAS  PubMed  PubMed Central  Google Scholar 

  54. 54.

    Kim, J. et al. Polycomb- and methylation-independent roles of EZH2 as a transcription activator. Cell Rep. 25, 2808–2820.e4 (2018).

    CAS  PubMed  PubMed Central  Google Scholar 

  55. 55.

    Hammond, M. E. H. et al. American Society of Clinical Oncology/College of American Pathologists guideline recommendations for immunohistochemical testing of estrogen and progesterone receptors in breast cancer (unabridged version). Arch. Pathol. Lab. Med. 134, e48–e72 (2010).

    CAS  PubMed  Google Scholar 

  56. 56.

    Turashvili, G. et al. P-cadherin expression as a prognostic biomarker in a 3992 case tissue microarray series of breast cancer. Mod. Pathol. 24, 64–81 (2011).

    CAS  PubMed  Google Scholar 

  57. 57.

    Rajput, A. B. et al. Stromal mast cells in invasive breast cancer are a marker of favourable prognosis: a study of 4,444 cases. Breast Cancer Res. Treat. 107, 249–257 (2008).

    PubMed  Google Scholar 

  58. 58.

    Friedman, J. H. A Variable Span Scatterplot Smoother Technical Report 5 (Laboratory for Computational Statistics, Stanford University, 1984); https://www.slac.stanford.edu/pubs/slacpubs/3250/slac-pub-3477.pdf

  59. 59.

    Cleveland, W. S., Grosse, E. & Shyu, W. M. Local Regression Models (Wadsworth & Brooks/Cole, 1992).

  60. 60.

    Van Belle, G., Fisher, L., Heagerty, P. J. & Lumley, T. Biostatistics: a Methodology for the Health Sciences 2nd edn (John Wiley & Sons, 2004).

  61. 61.

    Fiteni, F. et al. Endpoints in cancer clinical trials. J. Visc. Surg. 151, 17–22 (2014).

    CAS  PubMed  Google Scholar 

  62. 62.

    McKinney, S. BCCRCMO/BrCa_AgeAssociations: BCCRC Molecular Oncology breast cancer age associated biomarker study. Zenodo https://zenodo.org/record/3715548#.XpbqmdNKjeQ (2020).

  63. 63.

    Cheang, M. C. et al. Basal-like breast cancer defined by five biomarkers has superior prognostic value than triple-negative phenotype. Clin. Cancer Res. 14, 1368–1376 (2008).

    CAS  PubMed  Google Scholar 

  64. 64.

    Chia, S. et al. Human epidermal growth factor receptor 2 overexpression as a prognostic factor in a large tissue microarray series of node-negative breast cancers. J. Clin. Oncol. 26, 5697–5704 (2008).

    CAS  PubMed  Google Scholar 

  65. 65.

    Habibi, G. et al. Redefining prognostic factors for breast cancer: YB-1 is a stronger predictor of relapse and disease-specific survival than estrogen receptor or HER-2 across all tumor subtypes. Breast Cancer Res. 10, R86 (2008).

    PubMed  PubMed Central  Google Scholar 

  66. 66.

    Jensen, K. C. et al. New cutpoints to identify increased HER2 copy number: analysis of a large, population-based cohort with long-term follow-up. Breast Cancer Res. Treat. 112, 453–459 (2008).

    CAS  PubMed  Google Scholar 

  67. 67.

    Cheang, M. C. et al. Ki67 index, HER2 status, and prognosis of patients with luminal B breast cancer. J. Natl Cancer Inst. 101, 736–750 (2009).

    CAS  PubMed  PubMed Central  Google Scholar 

  68. 68.

    Turashvili, G. et al. Inter-observer reproducibility of HER2 immunohistochemical assessment and concordance with fluorescent in situ hybridization (FISH): pathologist assessment compared to quantitative image analysis. BMC Cancer 9, 165 (2009).

    PubMed  PubMed Central  Google Scholar 

  69. 69.

    Liu, S. et al. Progesterone receptor is a significant factor associated with clinical outcomes and effect of adjuvant tamoxifen therapy in breast cancer patients. Breast Cancer Res. Treat. 119, 53–61 (2010).

    CAS  PubMed  Google Scholar 

Download references

Acknowledgements

S.A. and C.J.E. are supported by the BC Cancer Foundation, Canadian Cancer Research Society (S.A. 705617), Terry Fox Research Institute (S.A. 1082) and Canadian Institutes for Health Research (S.A. 148429). S.A. is also supported by Canada Research Chairs and Breast Cancer Resesarch Foundation (BCRF-19-180). T.O. is supported by a Molecular Oncologic Pathology Fellowship from the Canadian Institute of Health Research and the Terry Fox Foundation, and by awards from the Sumitomo Life Welfare and Culture Foundation, Mochida Memorial Foundation for Medical and Pharmaceutical Research and Takashi Tsuruo Memorial Fund. H.L. is supported by the Severance Research Initiative Scholarship from Yonsei University.

Author information

Affiliations

Authors

Contributions

T.O. and S.A. designed the research. T.O.N., T.O., G.T. and S.M. designed the TMAs and reviewed the antibody staining. S.M. and D.C. analyzed the data. D.W., H.L., T.O., S.M. and S.A. wrote the paper. S.J. and W.Z. contributed ER binding data, performed the analyses and edited the manuscript. T.O.N., J.T.E., C.C. and C.J.E. contributed data and edited the manuscript. S.A. supervised the research.

Corresponding author

Correspondence to Samuel Aparicio.

Ethics declarations

Competing interests

S.A. is a founder and consultant for Contextual Genomics. S.A. advises Sangamo Therapeutics and Repare Therapeutics.

Additional information

Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Extended data

Extended Data Fig. 1 Age-correlated transcripts among different cancer types and biphasic age-correlated transcripts.

a, Venn sets bar plot of intersection sizes (upper panel) and histogram of age-correlated gene counts by cancer site (left panel). Vertical connectors below the Intersection Size bar plot show the corresponding set intersections. The breast cancer cohorts showed the most age-correlated gene counts (left three bars, thousands of age-associated genes). Lung and thyroid cohorts also showed hundreds of age-correlated genes, with considerable overlap with breast cancer cohorts. Kidney and prostate cohorts showed very few age-correlated genes. b, Biphasic age-correlated transcripts from TCGA Breast expression dataset (N = 1079 cases per graph). c, Biphasic age-correlated transcripts from the METABRIC expression dataset (N = 1992 cases per graph). The red curves are loess smoother curves indicating the mean expression for any given age, yielding a semi-parametric age-correlated expression pattern for the indicated biomarker. The tan coloured interval represents a 99% confidence interval for the smoother curve. The Benjamini-Hochberg multiple comparison adjusted p-values were obtained from linear model ANOVA (2-sided F-test) of cases with age < 60 (METABRIC N = 991 cases, TCGA N = 595 cases), age > 60 (METABRIC N = 1001 cases, TCGA N = 484 cases), and across all ages (METABRIC N = 1992 cases, TCGA N = 1079 cases). Genes with absolute fold change > 1.25 and adjusted p-value < 0.01 for cases age < 60 or cases age > 60 or all cases were declared as age-associated.

Extended Data Fig. 2 Age dependent transcript volcano plots for estrogen receptor (ER) and human epidermal growth factor receptor (HER2) subgroups for the TCGA dataset.

Horizontal axis displays the maximum fold change measured from the three regression line fits to expression data from the TCGA Breast expression dataset (N = 1079 cases) for cases < 60 years of age, cases > 60, and all cases. Fold changes to the right (Up) indicate genes exhibiting an increase in expression levels from younger to older cases. Fold changes to the left (Down) indicate genes exhibiting a decrease in expression levels from younger to older cases. Genes exhibiting a large absolute fold change and strong statistical evidence of association (small p-values) appear in the upper left and right areas of the volcano plots. P-values from linear model ANOVA (2-sided F-test) based on the sample sizes shown above each plot. Genes with absolute fold change > 1.25 and adjusted p-value < 0.01 for cases age < 60 or cases age > 60 or all cases are coloured according to the legend categories for ER binding.

Extended Data Fig. 3 Age-associated protein expression in breast cancer.

Immunohistochemistry (IHC) of tissue microarrays (TMAs) from a, Vancouver Big series, and b, METABRIC series. Even though pathologists score TMA staining patterns with a variety of schemas, e.g. percent of tumour cells positively stained, staining intensity, or a combination thereof, biphasic and other age-associated patterns are still revealed via supersmoother fits to the data. The age at which biphasic inflection occurs varies by biomarker, e.g. near 60 years of age for EZH2 and p16, near 70 years for ER. The red curves are supersmoother curves fitted to the IHC data. The tan coloured intervals are randomization test confidence bands under the null hypothesis of no age association (see Methods) at one of the four confidence levels indicated by the associated p-value bound in each graph. Red curves that extend beyond the (1-α) confidence bounds at any location indicate age-association at the given confidence level with a two-sided p-value p < α (95%, 99%, 99.9% and 99.99% confidence levels assessed via randomization tests yielding p < 0.05, 0.01, 0.001 and 1e-04 respectively, exact p-values not available, see Methods).

Extended Data Fig. 4 Age-correlated protein expression in normal breast epithelium and age-correlated distribution of estrogen receptor (ER) in breast cancer.

a, Age-correlated protein expression in a tissue microarray (TMA) from normal breast epithelium (Normal Breast 13 TMA, N = 537 cases). The red curves and tan confidence bands are described above in Extended Data Fig. 3. Since breast reduction mammoplasty is not performed as often for women over 65 years of age, trends in older cases are more variable, indicated by the wide confidence bands around age 70-80. b, Distribution of percent of tumour cells positively stained for ER proteins in immunohistochemistry tissue samples (Big Series TMA, N = 3991 cases). The youngest cases showed the lowest rates of percentages above 50%. The distribution of percent tumour cells positively stained was strongly associated with age (Pearson 𝜒2 (15) = 348.3, p = 4.8×10−65). c, Distribution of percent of tumour cells positively stained for ER proteins in immunohistochemistry tissue samples for ER positive cases only (from Big Series TMA, N = 3991 total cases, N = 2760 ER positive cases). Though all these cases were designated ER positive because they had 1% or more of tumour cells positively stained, the younger cases showed much fewer instances of 50% or more of tumour cells positively stained (Pearson 𝜒2 (10) =215.7, p = 8.3×10−41). Younger ER positive cases had a much higher rate of individual tumour cells that did not show ER staining, which presumably would not be responsive to ER targeting therapies such as tamoxifen.

Extended Data Fig. 5 Kaplan-Meier survival curves and unadjusted two-sided logrank test p-values for EZH2 and H3K27me3 scored by immunohistochemistry within age (rows) and ER (columns) strata.

The two left columns show survival curves for EZH2 and H3K27me3 for ER+ cases (Big Series TMA, N = 3991 total cases). The two right columns show survival curves for EZH2 and H3K27me3 for ER- cases. Age group and sample sizes for the solid and dashed line survival curves are shown above each graph.

Extended Data Fig. 6 Kaplan-Meier survival curves and unadjusted two-sided logrank test p-values for H3K27me3 within age and EZH2 strata, and EZH2 within age and H3K27me3 strata.

The two left columns show survival curves for H3K27me3 low (< 60% tumour cells positively stained) versus H3K27me3 high (60% or more tumour cells positively stained) within age (rows) and EZH2 (columns) strata (Big Series TMA, N = 3991 total cases). The cohort sizes N for H3K27me3 low, and H3K27me3 high respectively are shown above the curves. The two right columns show survival curves for EZH2 negative (< 10% tumour cells positively stained) versus EZH2 positive (10% or more tumour cells positively stained) within age (rows) and H3K27me3 (columns) strata. The cohort sizes N for EZH2 negative, and EZH2 positive respectively are shown above the curves. The youngest cases show a marked difference in survival between H3K27me3 low and high cohorts within the EZH2 negative strata, while the oldest cases show a marked difference in the EZH2 positive strata.

Extended Data Fig. 7 Kaplan-Meier survival curves and unadjusted two-sided logrank test p-values for H3K27me3 within age and EZH2 strata, and EZH2 within age and H3K27me3 strata, scored by immunohistochemistry, for ER+ cases.

The two left columns show ER positive cases’ survival curves for H3K27me3 low (< 60% tumour cells positively stained) versus H3K27me3 high (60% or more tumour cells positively stained) within age (rows) and EZH2 (columns) strata (Big Series TMA, N = 3991 total cases). The cohort sizes N for H3K27me3 low, and H3K27me3 high respectively are shown above the curves. The two right columns show ER positive cases’ survival curves for EZH2 negative (< 10% tumour cells positively stained) versus EZH2 positive (10% or more tumour cells positively stained) within age (rows) and H3K27me3 (columns) strata. The cohort sizes N for EZH2 negative, and EZH2 positive respectively are shown above the curves. The youngest cases show a marked difference in survival between H3K27me3 low and high cohorts within the EZH2 negative strata, while the oldest cases do not show a marked difference in the EZH2 positive strata. (See Extended Data Fig. 8 for ER- cases).

Extended Data Fig. 8 Kaplan-Meier survival curves and unadjusted two-sided logrank test p-values for H3K27me3 within age and EZH2 strata (left two columns), and EZH2 within age and H3K27me3 strata (right two columns), for ER- cases.

The two left columns show ER negative cases’ survival curves for H3K27me3 low (< 60% tumour cells positively stained) versus H3K27me3 high (60% or more tumour cells positively stained) within age (rows) and EZH2 (columns) strata (Big Series TMA, N = 3991 total cases). The cohort sizes N for H3K27me3 low, and H3K27me3 high respectively are shown above the curves. The two right columns show ER negative cases’ survival curves for EZH2 negative (< 10% tumour cells positively stained) versus EZH2 positive (10% or more tumour cells positively stained) within age (rows) and H3K27me3 (columns) strata. The cohort sizes N for EZH2 negative, and EZH2 positive respectively are shown above the curves. The youngest cases do not show a marked difference in survival between H3K27me3 low and high cohorts within the EZH2 negative strata, while the oldest cases show a marked difference in the EZH2 positive strata. (See Extended Data figure 7 for ER+ cases).

Extended Data Fig. 9 Kaplan-Meier survival curves and unadjusted two-sided logrank test p-values for FOXA1 and BCL2 within age strata for ER positive cases (left two columns) and ER negative cases (right two columns).

Big Series TMA data (N = 3991 cases) survival patterns within ER and age group strata for FOXA1 and BCL2 protein staining levels. The cohort sizes N for low and high protein staining rates within each ER and age group strata are shown above the curves. For FOXA1, all survival differences occur in cases under age 55 years, specifically for 40-55 year old cases in the ER positive strata, and for 20-40 year old cases in the ER negative strata. For BCL2, survival differences are seen in the three youngest age strata for ER positive cases, but in the oldest three age strata for ER negative cases.

Extended Data Fig. 10 Bootstrap analysis for age-association assessment within IntClust subgroups.

A random sample with replacement of size N was taken from the METABRIC expression set (1992 cases in total) in each of the ten IntClust groups to yield equally sized IntClust groups, so as to avoid the issue of smaller p-values in larger IntClust groups as seen in Fig. 4 where IntClust groups 3, 4 and 8 appear to show more age-correlated gene targets. Age-association status was reassessed for all gene targets with the bootstrapped sample and the proportion of age-correlated targets within each IntClust group recalculated, as shown by the red points plotted. Proportion of age-associated genes from each of the 20 bootstrap analyses are shown along with their associated boxplot, suggesting that IntClust groups 1, 2, 6 and 9 also contain sizable proportions of gene targets with age-correlated expression. The bootstrap sample size N per IntClust group is shown at the top of each graph. Boxplot elements show median (centre line), first and third quartiles (box top and bottom “hinges”), and whiskers extending from the hinge to the largest value no further than 1.5 * IQR from the hinge (where IQR is the inter-quartile range, or distance between the first and third quartiles).

Supplementary information

Supplementary Information

Supplementary Discussion and Supplementary Tables 1–7.

Reporting Summary

Supplementary Data 1

Expression data fold-change estimates and age-associated P values for 18,950 protein-coding genes for the ages <60 years and >60 years and all cases combined, for the METABRIC and TCGA cancer datasets.

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Osako, T., Lee, H., Turashvili, G. et al. Age-correlated protein and transcript expression in breast cancer and normal breast tissues is dominated by host endocrine effects. Nat Cancer 1, 518–532 (2020). https://doi.org/10.1038/s43018-020-0060-4

Download citation

Further reading

Search

Quick links

Nature Briefing

Sign up for the Nature Briefing newsletter — what matters in science, free to your inbox daily.

Get the most important science stories of the day, free in your inbox. Sign up for Nature Briefing