Genetic modifiers of CHEK2-associated and familial breast cancer

Doctoral Programme in Biomedicine (DPBM)

Genetic modifiers of CHEK2-associated and familial breast cancer

Taru A. Muranen

Department of Obstetrics and Gynecology Helsinki University Hospital

Faculty of Medicine University of Helsinki

Helsinki, Finland

Academic Dissertation

To be discussed, with permission of the Faculty of Medicine, University of Helsinki, in Biomedicum 1, Lecture Hall 2, Haartmaninkatu 8, Helsinki

on 2 November 2018, at 12 noon.

Helsinki 2018

Supervised by:

Adjunct Professor Heli Nevanlinna, PhD Department of Obstetrics and Gynecology

Helsinki University Hospital and University of Helsinki, Finland Associate Professor Dario Greco, PhD

Faculty of Medicine and Life Sciences

Institute of Biosciences and Medical Technology University of Tampere, Finland

Reviewed by:

Adjunct Professor Minna Tanner, MD, PhD Faculty of Medicine and Life Sciences University of Tampere, Finland Professor Matti Nykter, PhD

Faculty of Medicine and Life Sciences University of Tampere, Finland Official Opponent:

Associate Professor Ingrid Hedenfalk, PhD Division of Oncology and Pathology Department of Clinical Sciences Lund University, Sweden

Cover image: Three versions of the same pedigree overlaid: one colored by disease status (on the bottom), one colored by genotype of a moderate penetrance mutation (middle), and one colored by polygenic risk score (on the top).

Dissertationes Scholae Doctoralis Ad Sanitatem Investigandam Universitatis Helsinkiensis ISBN 978-951-51-4503-1 (Paperback)

ISBN 978-951-51-4504-8 (PDF) ISSN 2342-3161 (print)

ISSN 2342-317X (online)

Unigrafia

3

Itseoppinut on ainoa oppinut. Muut ovat opetettuja.

Erno Paasilinna

Ursalle, Elselle, Eerolle ja Urholle

Table of Contents

Table of Contents ... 4

Abstract ... 7

List of Original Publications ... 9

Abbreviations ... 10

Gene and protein names ... 11

1 Introduction ... 12

2 Review of the Literature ... 13

2.1 General cancer characteristics ... 13

2.1.1 Cancer progression ... 13

2.1.2 Cancer genes ... 15

2.2 Breast cancer ... 16

2.2.1 Mammary gland ... 17

2.2.2 Breast cancer risk factors ... 20

2.2.3 Breast cancer subtypes ... 20

2.2.4 Origin of breast cancer... 22

2.3 Breast cancer treatment ... 23

2.3.1 Adjuvant endocrine therapy ... 23

2.3.2 Other targeted biological therapies ... 23

2.3.3 Adjuvant chemotherapy ... 24

2.4 Genetic predisposition to breast cancer ... 25

2.4.1 Breast cancer heritability ... 25

2.4.2 High-risk genes ... 28

2.4.3 Moderate-risk genes ... 29

2.4.4 The Breast cancer pathway ... 29

2.4.5 Common predisposing variants ... 31

2.5 CHEK2... 31

2.5.1 CHEK2 protein function ... 31

2.5.2

2.5.3

2.5.4 CHEK2 in breast tumors ... 34

3 Aims of the Study ... 35

5

4.1 Study subjects and data sources ... 36

4.1.1 Breast tumors (I, II) ... 36

4.1.2 Study subjects from the Breast Cancer Association Consortium (II, III) ... 36

4.1.3 Study subjects of the Helsinki breast cancer study (IV) ... 37

4.2 Methods ... 37

4.2.1 Microarray data processing and analyses (I, II) ... 37

4.2.2 Permutation analysis (I: unpublished data) ... 39

4.2.3 Survival analyses (II) ... 40

4.2.4 Tumor pathology analyses (II) ... 40

4.2.5 The Polygenic risk score (III, IV) ... 40

4.2.6 Risk association analyses (III, IV) ... 40

4.2.7 Feature selection (III: unpublished data) ... 40

4.2.8

4.3 Ethics statement ... 41

5 Results ... 42

5.1 c.1100delC and p.(I157T) carrier tumors (I, II) ... 42

5.1.1 c.1100delC-associated copy number aberrations (I) ... 42

5.1.2 c.1100delC-associated differences in gene expression (I and unpublished data) ... 42

5.1.3 Combined analysis of aCGH and GEX data (I and unpublished data) ... 43

5.1.4 p.(I157T)-associated gene expression (II) ... 45

5.1.5 Clinico-pathological characteristics (II) ... 45

5.2 p.(I157T) or c.1100delC carrier survival (II) ... 45

5.3 Genetic modifiers of c.1100delC-associated breast cancer risk (III)... 45

5.3.1 Synergistic risk effect of common variants for c.1100delC carriers (III)... 45

5.3.2 The sparse model (III: unpublished data) ... 46

5.3.3

5.4 Risk modifiers in breast cancer families (IV) ... 47

6 Discussion ... 48

6.1

6.1.1 Germline

6.1.2 Genomic profiling elucidates the steps of

6.1.3 1p22 loss might complement

6.1.4 Elevated expression of olfactory receptors in c.1100delC carrier tumors (I) ... 51

6.1.5 WNT pathway deregulation – typical for c.1100delC breast cancers (I,III)? ... 52

6.1.6 Hypothetical model for c.1100delC-associated breast cancer progression (I, III) ... 55

6.1.7 Is p.(I157T) ‘the first hit’ for germline mutation carriers (II)? ... 55

6.2 Survival of breast cancer patients carrying germline

6.2.1 Increased mortality associated with c.1100delC ... 57

6.2.2

6.2.3 p.(I157T), lobular carcinoma, and patient survival warrant further research ... 59

6.3 Common genetic variants in breast cancer risk prediction (III, IV) ... 59

6.3.1 PRS could be used in risk stratification of c.1100delC carriers (III) ... 59

6.3.2 No epistatic interaction exists between c.1100delC and the common variants (III) ... 59

6.3.3 PRS explains part of the increased familial risk (IV) ... 60

7 Summary and Conclusions ... 62

8 Acknowledgments... 64

References ... 66

Appendix

7

Abstract

Aims

CHEK2 (checkpoint kinase 2) is a moderate-risk breast cancer susceptibility gene. By definition, the CHEK2 susceptibility mutations do not segregate consistently with breast cancer within pedigrees, and other genetic factors have been proposed to modify the penetrance of the CHEK2 mutations. The primary purpose of this study was to identify the risk-modifying genetic factors using risk association analyses. Furthermore, genomic profiling of mutation carrier tumors could suggest candidate loci for further risk analyses, but more importantly shed light on the events that have led to tumor development, complementing the CHEK2 deficiency.

CHEK2 c.1100delC has been suggested to be associated with poor prognosis after breast cancer diagnosis. We tested whether the same effect would be shared by p.(I157T), another recurrent CHEK2 mutation in the Finnish population. Additionally, breast cancer phenotypic features associated with the two mutations were examined in terms of pathological characteristics and differential gene expression.

By now, collaborative international studies have identified a vast number of common variants associated with a modest increase in the risk of breast cancer. However, combining multiple variants into a polygenic risk score (PRS) has been assumed to have potential in breast cancer risk stratification. We assessed the applicability of the PRS in risk prediction for women at elevated baseline risk, namely carriers of c.1100delC and women with a positive family history of breast cancer.

Essential methods

Genomic copy number aberrations (CNA) associated with c.1100delC were analyzed using data from 26 c.1100delC carrier and 76 non-carrier tumors. Analyses were performed in R environment for statistical computing using Bioconductor packages CGHcall, CGHregions, and WECCA.

Associations between CNA regions and c.1100delC were tested with Wilcoxon rank-sum test.

C.1100delC-associated differential gene expression was examined using data from 13 c.1100delC carrier and 65 non-carrier tumors and p.(I157T)-associated gene expression with data from 10 p.(I157T) carrier and 162 non-carrier tumors. Analyses were performed with Bioconductor package limma. Functional enrichment of the differentially expressed genes was analyzed with DAVID functional annotation tool and Gene Set Enrichment Analysis (GSEA) using gene libraries available at mSigDB.

Survival and tumor pathologic characteristics of breast cancer patients carrying CHEK2 mutations were studied in collaboration with the Breast Cancer Association Consortium (BCAC) in a dataset consisting of 25940 non-carriers, 590 p.(I157T) carriers, and 271 c.1100delC carriers. Survival analyses were performed with Cox regression and pathology analyses with Cochran-Mantel- Haenszel test.

Risk effect associated with about 75 common variants was studied in a BCAC dataset of 78 354

non-carriers and 848 c.1100delC carriers as well as in a Finnish dataset consisting of 1 303

unselected cases, 378 additional familial index cases, 1 272 population controls, and 429 women

from 52 breast cancer families. A polygenic risk was calculated as a product of per-variant log

odds ratios, and standardized according to healthy population controls. Risk association analyses were performed with logistic regression. Nested regression models were compared with likelihood-ratio test.

Results

We identified seven chromosomal locations, whose copy number aberrations were more frequent in c.1100delC carrier tumors than in non-carrier tumors. Functional in silico analysis of CNA regions and differentially expressed genes suggested that loss of GBP genes, elevated activity of olfactory receptors and deregulation of the WNT pathway could be recurrent driver events in c.1100delC carrier cancers.

Gene expression analysis suggested that CDH1 inactivation is a frequent event in p.(I157T) carrier breast cancers, possibly accounting for most of the observed differences between p.(I157T) carrier and non-carrier breast cancers.

Germline CHEK2 mutations c.1100delC and p.(I157T) differ in terms of their association with patient prognosis, c.1100delC being a marker for poor prognosis. Both mutations predispose to estrogen receptor (ER)-positive breast cancer. P.(I157T) is associated with lobular breast cancer, whereas c.1100delC is not associated with any specific breast cancer histological subtype.

The breast cancer risk associated with the PRS was similar for c.1100delC carriers, women from breast cancer families, and unselected women. When accounting for the elevated background risk associated with c.1100delC, about 20% of mutation carriers with the highest PRS values were estimated to have higher than 30% lifetime risk. Furthermore, even though PRS explained part of the excess familial risk, the high PRS values retained predictive value in risk stratification of women with a positive family history of breast cancer.

Conclusions

The genomic analyses of CHEK2 mutation carrier tumors could lay a foundation for a model of c.1100delC-associated tumorigenesis. Furthermore, c.1100delC and p.(I157T) might have different roles in the origin and development of breast cancer. The findings from these hypothesis- generating studies could benefit future functional in vitro and in vivo studies on breast cancer etiology.

The poor survival of breast cancer patients carrying c.1100delC warrants further examination. The data presented in this work indicate that the survival association is not shared by p.(I157T), emphasizing that findings based on a certain mutation cannot always be generalized to other mutations of the same gene.

On a general population level, the usability of the current PRS is limited due to the very low proportion of unselected women stratified into the high-risk category by PRS alone. However, for women at elevated background risk as a consequence of an inherited moderate penetrance mutation, like c.1100delC, or positive family history, the PRS of about 75 variants could provide significant clinical benefit in identifying women at high lifetime risk.

9

List of Original Publications

This thesis is based on the following original publications, referred to in the text by their Roman numerals. In addition, unpublished data exploring Studies I and III further are included.

I. Muranen TA, Greco D, Fagerholm R, Kilpivaara O, Kämpjärvi K, Aittomäki K, Blomqvist C, Heikkilä P, Borg Å, Nevanlinna H. Breast tumors from CHEK2 1100delC- mutation carriers: genomic landscape and clinical implications. Breast Cancer Res.

2011;13:R90.

II. Muranen TA, Blomqvist C, Dörk T, Jakubowska A, Bojesen SE, Fagerholm R, Greco D, Aittomäki K, Shah M, Dunning AM, Rhenius V, Hall P, Czene K, Brand JS, Darabi H, Chang-Claude J, Rudolph A, Nordestgaard BG, Couch FJ, Hallberg E, Figueroa J, García-Closas M, Fasching PA, Beckmann MW, Li J, Liu J, Andrulis IL, Knight JA, Winqvist R, Pylkäs K, Mannermaa A, Kataja V, Lindblom A, Margolin S, Lubinski J, Dubrowinskaja N, Bolla MK, Dennis J, Michailidou K, Wang Q, Easton DF, Pharoah PDP, Schmidt MK, Nevanlinna H. Patient survival and tumor characteristics associated with CHEK2 I157T: findings from the Breast Cancer Association Consortium. Breast Cancer Res. 2016 Oct 3;18(1):98.

III. Muranen TA, Greco D, Blomqvist C, Aittomäki K, Khan S, Hogervorst F, Verhoef S, Pharoah PDP, Dunning AM, Shah M, Luben R, Bojesen SE, Nordestgaard BG, Schoemaker M, Swerdlow A, García-Closas M, Figueroa J, Dörk T, Bogdanova NV, Hall P, Li J, Khusnutdinova E, Bermisheva M, Kristensen V, Borresen-Dale AL, Investigators N, Peto J, Dos Santos Silva I, Couch FJ, Olson JE, Hillemans P, Park-Simon TW, Brauch H, Hamann U, Burwinkel B, Marme F, Meindl A, Schmutzler RK, Cox A, Cross SS, Sawyer EJ, Tomlinson I, Lambrechts D, Moisse M, Lindblom A, Margolin S, Hollestelle A, Martens JWM, Fasching PA, Beckmann MW, Andrulis IL, Knight JA, Investigators K, Anton-Culver H, Ziogas A, Giles GG, Milne RL, Brenner H, Arndt V, Mannermaa A, Kosma VM, Chang-Claude J, Rudolph A, Devilee P, Seynaeve C, Hopper JL, Southey MC, John EM, Whittemore AS, Bolla MK, Wang Q, Michailidou K, Dennis J, Easton DF, Schmidt MK, Nevanlinna H. Genetic modifiers of CHEK2 c.1100delC- associated breast cancer risk. Genet Med. 2017 May;19(5):599-603.

IV. Muranen TA, Mavaddat N, Khan S, Fagerholm R, Pelttari L, Blomqvist C, Aittomäki K, Easton DF, Nevanlinna H. Polygenic risk score is associated with increased disease risk in 52 Finnish breast cancer families. Breast Cancer Res Treat. 2016 Aug;158(3):463-9.

These publications are reprinted with the permission of their copyright holders.

Abbreviations

aCGH Array comparative genomic hybridization BAC Bacterial artificial chromosome BCAC Breast Cancer Association Consortium

CI Confidence interval

CMF Cyclophosphamide

–

–

CNA Copy number aberration

COGS Collaborative Oncological Gene-Environment Study

ER Estrogen receptor

FFPE Formalin-fixed paraffin-embedded

FHA Fork head-associated

FWER Family-wise error rate

G1/G2 Gap 1/2

GEO Gene Expression Omnibus

GEX Gene expression

GSEA Gene set enrichment analysis GWAS Genome-wide association study

HR Hazard ratio

IC-NST Invasive carcinoma of no special type ILC Invasive lobular carcinoma

KD Kinase domain

M Metastasis

M Mitosis

N Status of adjacent lymph nodes

OR Odds ratio

PgR Progesterone receptor

PRS Polygenic risk score

Q Glutamine

S; Ser Serine

S Synthesis

SCD SQ/TQ cluster domain

SNP Single-nucleotide polymorphism

T; Thr Threonine

T Tumor size

TCGA The Cancer Genome Atlas

TEB Terminal end bud

UTR Untranslated region

11

Gene and protein names

ACIII Adenylate cyclase III AKT AKT serine/threonine kinase 1 ALDH Aldehyde dehydrogenase

ALG14 ALG14, UDP-N-

acetylglucosaminyltransferase subunit

ANKLE1 Ankyrin repeat and LEM domain containing 1

APC APC, WNT signaling pathway regulator

ATE1 Arginyltransferase 1 ATM ATM serine/threonine kinase AURKA Aurora kinase A

BARD1 BRCA1 associated RING domain 1 BRCA1/2 BRCA1/2, DNA repair associated CALCOCO1 Calcium binding and coiled-coil

domain 1

CDC25A Cell division cycle 25 A

CDH1 Cadherin 1

CDK1/2/4/6 Cyclin dependent kinase 1/2/4/6 CHEK2 Checkpoint kinase 2

CLCA1 Chloride channel accessory 1 CSAD Cysteine sulfinic acid decarboxylase EGFR Epidermal growth factor receptor ELL Elongation factor for RNA

polymerase II

FANCM Fanconi anemia complementation group M

FGF Fibroblast growth factor FGFR2 Fibroblast growth factor receptor 2 FTO FTO, alpha-ketoglutarate dependent

dioxygenase

FZD1 Frizzled class receptor 1 GBP1-7 Guanylate binding protein 1-7 GNAL G protein subunit alpha L HER2 Human epidermal growth factor

receptor 2 IFN-Ȗ Interferon gamma IL-1ȕ Interleukin 1 beta

JAK Janus kinase

KRT5 Keratin 5

LEF1 Lymphoid enhancer binding factor 1

LHRH Luteinizing hormone releasing hormone

LRP1 LDL receptor related protein 1 LRRC8D Leucine rich repeat containing 8

VRAC subunit D

MLH1 mutL homolog 1

MMP Matrix metalloproteinase MRE11 MRE11 homolog, double strand

break repair nuclease

MSH2 mutS homolog 2

MTOR Mechanistic target of rapamycin kinase

MYC MYC proto-oncogene, bHLH transcription factor

NBN Nibrin

NF1 Neurofibromin 1

NQO1 NAD(P)H:quinone oxidoreductase

OR Olfactory receptor

OR6C3 Olfactory receptor family 6 subfamily C member 3 PALB2 Partner and localizer of BRCA2 PARP Poly(ADP-ribose) polymerase PIK3CB Phosphatidylinositol-4,5-

bisphosphate 3-kinase catalytic subunit beta

PRKDC Protein kinase, DNA-activated, catalytic polypeptide

PTEN Phosphatase and tensin homolog

PVT1 Pvt1 oncogene

RAD50 RAD50 double strand break repair protein

RAD51 RAD51 recombinase RasGEF Ras-type guanine nucleotide

exchange factors

RB1 RB transcriptional corepressor 1 STAT Signal transducer and activator of

transcription

STK11 Serine/threonine kinase 11 TMED5 Transmembrane p24 trafficking

protein 5

TOP2A Topoisomerase II alpha TP53 Tumor protein p53

1 Introduction

CHEK2 has been established as a moderate-penetrance breast cancer susceptibility gene. Two CHEK2 mutations, protein truncating c.1100delC and missense p.(I157T), are relatively common in the Finnish population, having carrier frequencies of 1.4% and 5.3%, respectively.

The relative risk associated with c.1100delC and other truncating CHEK2 mutations is two- to threefold, whereas the risk effect of p.(I157T) is considerably lower.

C.1100delC predisposes to familial breast cancer. However, it does not segregate consistently with the disease within breast cancer families. Furthermore, the moderate-risk level associated with c.1100delC (odds ratio (OR):

2.26; [95% confidence interval (CI) 1.90-2.69]) has rendered its applicability in genetic counseling limited.

About 30% of breast cancer incidence has been estimated to be caused by genetic factors.

High- and moderate-risk mutations account for about one-fifth of the heritability.

However, they do not operate alone. The best genetic model explaining both familial clustering and population-level incidence of breast cancer consists of rare high-penetrance mutations and common variants with low effect sizes contributing together in a multiplicative fashion to increase the risk.

The discovery of multiple risk-modifying variants in genome-wide association studies (GWAS) studies has paved the way for investigation of genetic variants modifying the risk associated with CHEK2 c.1100delC.

In addition to risk-modifying effects, common genetic variation has been predicted to contribute to familial clustering of breast cancer.

Furthermore, the multiplicative model suggests that the nominal risk effects associated with single variants could be combined in order to estimate risk of individual women. Recently, a polygenic risk score (PRS) was introduced for risk prediction on a population level,

and we have investigated its applicability in breast cancer families.

Hereditary cancer is typically caused by an inherited loss-of-function mutation in a tumor

suppressor gene.

The loss of the intact allele is assumed to be often the initiating event of a

multistep path to cancer.

The later steps of tumorigenesis arise as a result of random somatic

events. However, only those changes that endow a growth advantage in combination with the

earlier events are selected during the course of tumor evolution. The final cancer phenotype

reflects the accumulation of novel features associated with the driver events.

Since the driver

changes are specific for the cancer relative to adjacent healthy tissue, and since tumor growth is

dependent on them, they represent an appealing target for cancer therapy. Genomic analyses of

copy number aberrations and gene expression changes in tumor tissue can be used for

characterization of the driver events leading to cancer in specific cancer subgroups.

13

2 Review of the Literature

2.1 General cancer characteristics 2.1.1 Cancer progression

Cancer is a progressive disease in which the cancerous cells gradually lose their tissue-typical morphology, proliferate in an uncontrolled manner, invade the surrounding tissue, and eventually spread via lymphatic and blood vasculature to give rise to metastatic growths.

Cancer progression is driven and accompanied by mutations, which can be considered as stochastic events whose probability is increased by two types of factors: those increasing the number of cell divisions and those causing DNA damage (Figure 1).

Most of the neoplastic events, i.e.

emergence of driver mutations, take place in progenitor/transit-amplifying cells, which have the capacity to dedifferentiate into stem cell state, but which also proliferate at a frequency high enough for accumulation of a sufficient number of malignant mutations.

However, the actual origin of a cancer could be any cell that retains proliferative capacity, ranging from stem cells to their more differentiated descendants, depending on tissue hierarchy and cell half-lives.

After the initiating event, the progeny of any pre-neoplastic cell may normally take their place and function in the tissue hierarchy or, alternatively, form a benign growth or even be erased by innate mechanisms controlling tissue homeostasis until further mutations endowing a growth advantage emerge. Thus, tumor progression is a cellular-level combination of stochastic events and Darwinian evolution.

Figure 1. Summary of factors increasing the probability of cancer progression.

The branching evolution characteristic for cancer development can be seen in genomic profiles of excised tumors and metastatic growths. The benign clonal cell populations co-exist with their more aggressive descendants. The different subpopulations can be distinguished by their mutation and gene expression profiles, highlighting the intrinsic heterogeneity of any single tumor.

Healthy tissue

Inherited mutations

replicationDNA

errors Carcinogenic

exposure Increasing

number of stem cell

divisions Intrinsic

oxidative stress

Age

Inflammation, hormonal exposure etc.

Increasing probability of tumor driver

mutations

Cancer

Figure 2.

Cancer Hallmarks introduced by Hanahan and Weinberg can be categorized as changes taking place primarily in the neoplastic cell lineage (inner circle) and changes affecting the interactions between the cancer cells and their tissue environment (outer circle).

Owing to the progressive nature of the disease, each cancer is unique. Hanahan and Weinberg summarized features shared by most cancers into six ‘Cancer Hallmarks’ and later refined the model by addition of two novel hallmarks and two enabling characteristics (Figure 2).

First of all, alterations in cellular and tissue level programs regulating cell proliferation and survival are divided into three hallmarks: sustaining proliferative signaling, evading the growth suppressors, and escaping the intrinsic apoptotic programs. The number of replicative cycles is restricted by telomere erosion in human cells. The rampantly dividing cells face a telomere crisis and need to reactivate the telomerase enzyme to gain replicative immortality, the fourth hallmark. Another restrictive mechanism for tumor growth is the shortage of oxygen and nutrients. Two further hallmarks enable cancer to overcome this challenge: inducing neo-vasculature and reshaping energy metabolism. The ultimate hallmark transforming cancer into a systemic disease is tissue invasion and metastasis, requiring changes in cell phenotype and adaptation to a foreign cellular environment. Finally, all the way through the tumor progression, the cancer cells must escape the surveillance of the immune system in order to triumph. In addition to these eight hallmarks, Hanahan and Weinberg named two features as specific cancer-enabling characteristics: genomic instability and chronic inflammation (Figure 2).

The concept of cancer hallmarks is a simplified framework; all hallmarks cannot be considered to apply to all cancer cells at all times, not to all cancer stem cells, and not even to all cancers. Floor

Sustaining proliferative

signaling Resisting

cell death

Deregulating cellular energetics Genome

instability and mutation

Enabling replicative immortality

Activating invasion and metastasis

15

(Figure 3). In this model, the ‘Cancer Hallmarks’ form the highest hierarchical level, the general cancer characteristics. The hallmarks result from changes in cellular pathways arising from genetic, epigenetic, or lysogenic oncogenic events, which on the bottom level are caused by various intrinsic and environmental factors (Figure 1). Importantly, the hierarchical levels of the model are connected by complex one-to-many and many-to-one relations.

This model emphasizes the need for a molecular-level understanding of cancer. A good start for this effort would be harvesting the oncogenic events, i.e. the acquired somatic and predisposing germline mutations in genetic studies. Elucidating how these changes affect the complex network of cellular pathways and differentiation programs giving rise to the cancer hallmarks will be the future goal of cancer research.

Figure 3.

Hierarchical model of carcinogenesis etiology with potential simple and complex interactions at all levels of hierarchy. Adapted from Floor

2.1.2 Cancer genes

Cancer-associated genes are categorized into two groups based on how their aberrations accelerate tumorigenesis: activated oncogenes and silenced tumor suppressor genes; both promote cancer progression.

By definition, oncogenes are genes encoding proteins involved in cellular growth and differentiation programs that have lost important gene- or protein-level regulatory constraints.

Cellular oncogenes have often been identified at characteristic chromosomal translocations due to sequence similarity to viral oncogenes.

Oncogene activation is usually a somatic event, and mutations in proto-oncogenes are rarely associated with hereditary cancer.

The majority of tumor suppressor genes have been discovered in studies of cancer families.

Typically, inactivation of both alleles is required for cancer initiation, as suggested in Knudson’s

‘two-hit hypothesis’.

If one hit, i.e. a loss-of-function mutation of a tumor suppressor gene, is

inherited, cancer probability is higher than if both hits were to be acquired as somatic changes. In

addition to this simplistic model, haplo-insufficiency has been recognized as another mechanism

for tumor suppressor-related cancer initiation; under certain environmental or intrinsic stress,

expression of only one intact allele is not sufficient to protect the cell from additional neoplastic

changes.

Tumor suppressor genes are further divided into three functional categories: gatekeepers, caretakers, and landscapers. Gatekeepers refer to the original idea of tumor suppressors as anti- oncogenes. Those are genes encoding proteins that limit cell cycle progression and entry into mitosis such as APC (APC, WNT signaling pathway regulator) in colorectal cancer and RB1 (RB transcriptional corepressor 1) in retinoblastoma. Caretakers include DNA repair genes, like BRCA1 and BRCA2 (BRCA1/2, DNA repair associated), whose mutations predispose to breast and ovarian cancer, or MSH2 (mutS homolog 2) and MLH1 (mutL homolog 1), which are associated with colorectal cancer.

Landscaper genes are defined to encode proteins involved in regulating the tumor micro- environment. The tumor-initiating mutation of a landscaper gene might even take place in a stromal cell instead of the cancer cell lineage. Juvenile polyposis is a characteristic syndrome for a germline landscaper gene deficiency. It is manifested by multiple hamartomatous polyps of the colon at a young age. The abnormal growth of the epithelium has been concluded to be induced by a mutation in the surrounding stromal cells, causing sustained proliferative signaling. The elevated number of cell divisions in epithelial cells raises the probability of somatic neoplastic events, thus increasing the risk of carcinoma development.

Functional classification of tumor suppressor genes identifies the key features of cancer-associated genes. However, it is a rough simplification. One gene or protein may serve in different roles depending on the context. For example, TP53 (tumor protein p53) serves as both a gatekeeper and caretaker, connecting the surveillance of genomic integrity to apoptosis,

and BRCA1 is involved in DNA repair, control of cell division, and regulation of differentiation via gene expression.

Furthermore, since even oncogene activation induces apoptotic programs, it is justified to shift the focus from single genes to the pathway level and consider cancer as a consequence of disturbed cellular programs.

Molecular and cellular cancer research has focused in a reductionist fashion on characterization of cancer cell lineages like the transformed epithelial cells in carcinomas. However, cancer is not a cell-autonomous disease. Instead, the interactions with the microenvironment, extracellular matrix, stromal cells, and immune system contribute to cancer development. There is some experimental evidence implying that aneuploid cancerous cells could be normalized under regulation of a healthy cellular environment.

Furthermore, the risk of tumor spread has been suggested to depend more on immune response than on the features of the cancer cells themselves.

2.2 Breast cancer

Breast cancer is the most common cancer and the leading cause of cancer-related death in women in developed countries.

In Finland, the cumulative disease risk by the age of 75 years has increased steadily since the 1950s, reaching 9.9% in 2014.

The high rank of breast cancer as a cause of mortality is partially misleading; it is mostly due to the high frequency of the disease itself. In general, breast cancer is a manageable disease, with a five-year survival rate of about 88%.

Male breast cancer is a rare disease, with an average of 25 diagnosed cases per year in Finland.

36

17

Breast cancer covers a range of phenotypically different neoplastic diseases. However, here, the term ‘breast cancer’ is used to strictly refer to carcinomas originating from the epithelial cells of the mammary gland, distinct from connective tissue, lymphoid, or skin neoplasias occurring in the chest area.

2.2.1

Mammary gland

The human mammary gland is a tree-like structure, with primary, secondary, and tertiary ducts forming the stem and branches, and lobules forming the leaves. The hormone-independent early development of the mammary gland starts during embryogenesis, with formation of a bilateral mammary ridge, followed by development of placodes and rudimentary ductal trees. After birth, the gland remains in a quiescent stage until puberty, when the branching morphogenesis continues in response to estrogen stimulus.

At birth, the rudimentary ductal tree consists of short primitive ducts ending in terminal end buds (TEBs). Thereafter, the primary ducts are formed by bifurcation of the TEBs. While the primary ducts elongate, the secondary and tertiary ducts emerge as lateral appendages, with novel TEBs at each branch end. Eventually, the TEBs give rise to small ductules, which first develop into virginal lobules, i.e. terminal ductal lobular units, and later differentiate into milk secreting alveoli.

The tertiary ducts and the lobules are sensitive to oscillating exposure to ovarian and pituitary hormones, which induce a growth and differentiation pulse followed by a period of regression during each menstrual cycle. However, until the age of 35 years the regression never returns to the starting point, the net effect being cumulative growth and differentiation.

While the lobules develop, the number of ductules per lobule increases and the size of the ductules decreases.

The full maturation of the ductal tree takes place during the first half of the first pregnancy, when the epithelial luminal cells differentiate in anticipation of lactation. Pregnancy, lactation, and weaning are followed by partial gland involution. Russo et al.

classify the lobules into four types based on the number of ductules per lobule. The predominant lobule type for nulliparous women is ‘Lob1’ with about 11 ductules, whereas for parous women it is ‘Lob3’ with about 81 ductules representing the most abundant type. ‘Lob2’ is an intermediate structure and ‘Lob4’ refers to fully maturated structure during the latter half of pregnancy and lactation.

In addition to structural changes, pregnancy induces epithelial cell differentiation by altering the balance of the activated cellular pathways, and by reducing the proliferative activity of the epithelial cells.

Menopause is associated with major structural involution of the ductal tree, irrespective of parity. Thus, the mammary gland of menopausal parous and nulliparous women is structurally similar and occupied mainly by ‘Lob1’ lobules. However, on a cellular level there are significant differences in chromatin condensation and proliferative activity.

The ducts are lined with two layers of epithelial cells: the luminal layer and the basal myoepithelial

layer. Alveolar, milk-producing cells differentiate from dedicated alveolar precursors scattered

around in the luminal layer.

All of these epithelial cell lineages originate from the same

precursors – the mammary stem cells, which dwell in the basal layer.

The stem cell activity is

restricted mainly to the period of embryonic development, and in the adult tissue cell regeneration

is maintained by multiplication of luminal and basal lineage-specific precursors.

Basal lamina

separates the epithelial cell layers from fibroblast-rich stroma, which surrounds the ductal tree

amid the adipocytes of the mammary fat pad.

Table 1. Regulators of mammary gland development.^{39, 46} Place of

expression/

action

Gene/Protein Agonists

and ligands

Antagonists Targets Effects

Systemic regulators Ovaries E1/E2

(Estrogens)

ESR1 pd: epithelial cell proliferation preg: maintenance of alveolar cells

Ovaries P

(Progesterone)

PGR preg: tertiary branching and alveologenesis Pituitary gland GH

(Growth hormone)

IGF1, ESR1 pd: epithelial cell proliferation

Pituitary gland PRL

(Prolactin) PRLS preg: progesterone expression,

lobulo-alveolar development

Liver IGF1

(Insulin-like growth factor) pd: epithelial cell proliferation

Liver PLG

(Plasminogen)

KLK1 ECM iv: ECM breakdown, disruption of

cell-cell contacts

Local regulators

Epithelium WNT3, WNT6, WNT10B Wnt family members

TBX3, FGFR2B

LEF1/TCF ed: epithelial cell proliferation

Mesenchyme WNT5A, WNT11

Wnt family members TGFB1 LEF1/TCF ed: epithelial cell proliferation

pd: inhibition of ductal elongation Epithelium WNT4

Wnt family member

P LEF1/TCF preg: tertiary branching

Epithelium and mesenchyme

LEF1/TCF

lymphoid enhancer binding factor 1 / TCF family transcription factors

WNT-

ligands Multiple

transcriptional level targets

mammary gland development

Mesenchyme TBX3 T-box 3

BMP4 WNT10B,

BMP4

ed: localization of developing gland

Epithelium BMP4

bone morphogenetic protein 4 TBX3 BMPR1A,

TBX3 ed: localization of developing gland Mesenchyme BMPR1A

bone morphogenetic protein receptor type 1A

PTH1R WNT

signaling

ed: localization of developing gland, nipple formation

Epithelium PTHLH

parathyroid hormone like hormone PTH1R ed: branching and niplle formation

Mesenchyme PTH1R

parathyroid hormone 1 receptor

PTHLH BMPR1A ed: branching and niplle formation

Mammary line Epithelium

FGFR2

fibroblast growth factor receptor 2

FGFs SPRY2 ed: placode placement

pd: epithelial cell proliferation Somites

Stroma

FGF10(and other FGF ligands)

fibroblast growth factor 10 GLI3 FGFRs ed: placode placement

pd: epithelial cell proliferation

Somites GLI3

GLI family zinc finger 3 FGF10 ed: placode placement

Mesenchyme NRG3 neuregulin 3

Integrins, ECM, ERBB4

ed: regulation of cell-cell interactions

Epithelium ERBB4

erb-b2 receptor tyrosine kinase 4

NRG3 ed: regulation of cell-cell interactions

Epithelium ERBB2

erb-b2 receptor tyrosine kinase 2

EGFR- coreceptor

pd: ductal morphogenesis

Stroma STX2

syntaxin 2 / epimorphin

Metallo- enzymes

CEBPB, MMP2, MMP3

pd: GH dependent branching

Epithelium CEBPB

CCAAT/enhancer binding protein beta

STX2 Multiple

transcriptional level targets

pd: GH dependent branching

Stroma IGF1

insulin like growth factor 1

GH IGFBP5 IGF1R pd: epithelial cell proliferation

19

Place of expression/

action

Gene/Protein Agonists

and ligands

Antagonists Targets Effects

Luminal cells AREG amphiregulin

ESR1, ADAM17

HSPGs EGFR pd: proliferation of ER-negative cells

Luminal cells ADAM17

a disintegrin and metalloproteinase 17 PPCs TIMP3 AREG pd: epithelial cell proliferation

Stroma EGFR

epidermal growth factor receptor

AREG, NRG3 and other EGF ligands

pd: ductal elongation

Stroma GHR

growth hormone receptor GH IGF pd: epithelial cell proliferation

Epithelium

Stroma TGFB1

transforming growth factor beta 1

TGFBR2 WNT

signaling

pd: negative regulator of branching and duct elongation

Epithelium CSF1

colony stimulating factor 1

CSF1R pd: macrophage recruitment

Stroma CCL11

C-C motif chemokine ligand 11

CCR3 pd: eosinophil recruitment

Luminal cells PGR

progesterone receptor P WNT4,

TNFSF11 preg: tertiary branching and alveologenesis Luminal cells TNFSF11/RANKL

TNF superfamily member 11 / RANK ligand

PGR, JAK2 /

STAT5 TNFRSF11A preg: tertiary branching

and alveologenesis

Luminal cells TNFRSF11A/RANK

TNF receptor superfamily member 11a TNSF11 CCND1,

NFKB1 preg: tertiary branching and alveologenesis Luminal cells PRLR

prolactin receptor

PRL TNFS11,

JAK2/STAT5

preg: tertiary branching and alveologenesis

Luminal cells Integrins ECM JAK2/STAT5 preg: tertiary branching

and alveologenesis Luminal cells SIRPA

signal regulatory protein Į

ECM JAK2/STAT5 preg: tertiary branching

and alveologenesis Luminal cells JAK2 / STAT5

janus kinase 2 /

signal transducer and activator of transcription 5

PRLR, Integrins, SIRPA

SOCS SOCS,

TNSF11 preg: tertiary branching and alveologenesis

Luminal cells SOCS family

suppressors of cytokine signaling

JAK2 / STAT5

JAK2 / STAT5

preg: tertiary branching and alveologenesis Epithelium LIF

leukemia inhibitory factor

Milk stasis STAT3 iv: involution inducing signal

Luminal cells STAT3

signal transducer and activator of transcription 3

LIF PI3 kinase,

IGFBP5 apoptotic programs

iv: inhibition of proliferative signals, apoptosis

Luminal cells IGFBP5

insulin-like growth factor binding protein-5

STAT3 IGF iv: inhibition of proliferative signals

Epithelium KLK1 kallikrein 1

PLG iv: ECM breakdown, disruption of cell-cell contacts

Stroma MMP2, MMP3 and MMP14 matrix metalloproteinases

EGFR, TGFB, ESR1, STX2

TIMPs 1-4, TGFB

ECM, basal lamina

pd: duct elongation, secondary branching

iv: disruption of basal lamina Stroma TIMPs 1-4

tissue inhibitor of metalloproteinases

MMPs pd and iv: MMP regulation Abbreviations: ed – embryonic development, pd – pubertal development, preg – pregnancy, iv - involution

The embryonic development of the mammary gland is independent of sex hormones and does not

differ between women and men. It is regulated by reciprocal interactions between the epithelium

and the underlying mesenchyme. Pubertal development is orchestrated by growth hormone

secreted from the pituitary gland and ovarian estrogen, which together induce global and local

downstream effects leading to mammary gland expansion (Table 1). Growth cycles associated

with the menstrual period rely mainly on ovarian progesterone stimulus, and the full maturation of the mammary gland taking place during pregnancy is driven by progesterone together with pi- tuitary gland prolactin. The key regulators in different developmental phases are listed in Table 1.

In brief, WNT and FGF (fibroblast growth factor) pathways have an essential role in ductal growth and branching from embryogenesis until full maturation. The JAK/STAT (janus kinase/signal transducer and activator of transcription) pathway together with matrix metalloproteinases (MMPs) are involved in pubertal and antenatal development as well as in involution. Overall, the mammary gland development is a complex interplay between the epithelial cell layers, extracellular matrix, stromal fibroblasts, and recruited white blood cells and requires a strict control of cell proliferation and differentiation as well as matrix remodeling.

2.2.2 Breast cancer risk factors

Breast cancer risk is increased by multiple factors related to life-style or to reproductive and medical history (Table 2).

Pregnancy is associated with a transient increase in the risk of estrogen receptor (ER)-negative breast cancer. In the long run, pregnancy and breast feeding have a protective effect. However, if the time between the first menstrual period and the first full-term pregnancy exceeds 15 years, the protective effect fades away. The pregnancy-associated fluctuation in breast cancer risk has been proposed to reflect the intrinsic mammary gland biology.

The periodically proliferating epithelial cells of a virgin gland are vulnerable to carcinogenic attack, whereas the full maturation of the gland entails protective changes in the form of increased chromatin condensation and lowered proliferative activity. However, gland expansion, accompanied by epithelial cell division and matrix remodeling, transiently raises the risk of malignant development during pregnancy.

Table 2. Breast cancer risk factors.

Life-style Reproductive history Medical history

Tobacco smoking Nulliparity* Oral contraceptives*

Alcohol consumption* High age at first full term pregnancy* Hormonal replacement therapy*

Overweight* Early menarche* Chest area X-rays

Lack of physical activity Late menopause* Mammographic density*

Benign breast disease

* Factors associated with increased exposure to estrogen.

Many of the risk factors are directly related to increased exposure to estrogen (Table 2).

Estrogen contributes to breast cancer risk both by increasing the number of cell divisions and by inducing genotoxic stress; estrogen metabolite estradiol generates free oxygen radicals and forms DNA adducts, leading to depurination and increased risk of error-prone DNA repair.

Oophorectomy or several years’ administration of antiestrogens halves the breast cancer risk of women at high familial risk.

2.2.3 Breast cancer subtypes

Breast cancers are categorized in many ways in order to assist in prognosis and in the choice of

treatment. Grading is based on tumor cell nuclei morphology, proliferation, and the extent to

21

according to tumor size (T), spread to adjacent lymph nodes (N), and distant metastases (M).

Histopathologic classification relies on multicellular structures and proportion of infiltrating cells.

The most common histological type is ‘Invasive carcinoma of no special type’ (IC-NST), previously called ‘Invasive ductal carcinoma’, and the second most common is ‘Invasive lobular carcinoma’ (ILC). Additionally, expression of four marker proteins, ER, PgR (progesterone receptor), HER2 (human epidermal growth factor receptor), and ki67 is often assessed in clinics.

ER/PgR or HER2 positivity is a direct indicator for endocrine or HER2-targeted therapies, respectively, whereas ki67 expression indicates high proliferation and suggests that the patient may benefit from adjuvant chemotherapy.

The diversity in external breast cancer features has evoked an attempt to classify tumors by their inherent cellular biology, leading to development of the intrinsic tumor subtypes and corresponding gene expression signatures.

The current consensus signature consist of 50 genes, whose expression levels divide breast tumors into luminal A, luminal B, basal-like, and HER2+ enriched subtypes.

The proportions of different subtypes have varied in the published literature, depending on cohort composition. However, in unselected patient series, luminal A is the most common subtype, assigned typically to about half of all cases. Luminal B is more frequent than basal-like and HER2+ enriched subtypes, which are equally common.

The classification has gained popularity because of the added value it gives to prognostic estimation and treatment choices. In brief, the good prognosis luminal A tumors could be spared from chemotherapy, whereas for the other three subgroups chemotherapy would be justified.

If gene expression data are unavailable, the intrinsic breast cancer subtypes can also be estimated using surrogate histopathological markers; ER expression makes the major division between luminal and basal branches, and HER2 expression divides the ER-positive luminal and ER-negative basal branches further. PgR, ki67, and grade can be used for subdividing the luminal branch (Table 3).

However, it is not uncommon that the surrogate marker-based prediction deviates from the gene expression-based classification. Especially, the differentiation of luminal A from luminal B tumors or HER2+ enriched from basal-like tumors remains challenging.

Table 3. Surrogate intrinsic subtypes defined on the basis of immunohistochemical measurement of marker

proteins as suggested in St Gallen 2013.

Luminal A-like Luminal B-like HER2 positive Triple negative / Basal like

ER

positive positive negative negative

PR

positive Either PR-negative, negative negative

Ki-67

low Ki-67 high, or any any

HER2

negative HER2-positive positive negative

Alongside tumor subtypes, gene expression signatures have been developed either for classifying

breast cancer patients into good and poor prognosis groups,

for predicting benefit from a

certain treatment specimen,

or for estimating specific tumor characteristics.

Even though

there is little overlap in the individual genes included in these signatures, they have not been shown

to differ in their ability to predict patient survival.

A critical meta-analysis showed that the

common feature for these signatures is their ability to detect tumor proliferation, which has a direct

association with patient survival.

In fact, gene expression in general is confounded by cell

proliferation status and the breast cancer-associated signatures have not been able to outperform random signatures in predicting patient survival.

Furthermore, different classification methods give opposite predictions on an individual patient level, and the traditional approach based on ER, HER2, grade, and TNM is still widely used for prognostic estimation.

Recently, IntClust classification has challenged the intrinsic subtypes as the best method for molecular taxonomy of breast cancer. IntClust is based on identification of recurrent chromosomal aberrations driving the tumorigenesis and affecting gene expression in cis.

The method was developed using copy number and gene expression data in parallel, but later a surrogate signature using only gene expression was developed.

The IntClust classification has 10 categories.

Intrinsic basal tumors cluster almost exclusively to a single IntClust category, HER2+ enriched cancers to another category, but luminal tumors are dispersed across multiple IntClust categories.

In the original study, IntClust outperformed the intrinsic subtyping especially in identifying a subgroup of chemo-insensitive ER-positive/luminal A breast cancers with poor prognosis, as well as in defining a signature indicative of a high number of infiltrating T-cells associated with good prognosis irrespective of the intrinsic subtype.

However, perhaps the most important contribution of the IntClust classification to breast cancer taxonomy is the shift in focus from primarily prognostic estimation to identification of the molecular events leading to breast cancer and characterization of specific druggable aberrations.

2.2.4 Origin of breast cancer

Development of the intrinsic molecular subtypes encouraged scientists to hypothesize that luminal tumors originate from luminal layer cells and basal tumors from myoepithelial cells.

However, multiple ensuing studies have disputed this and indicated that luminal and basal breast cancers share a common origin, the luminal progenitor cells. Only later steps in tumor progression determine the resulting tumor phenotype.

BRCA1-deficient cancer has served as an archetypic model for basal breast cancer, and the luminal origin of basal tumors was first discovered in a BRCA1 knock-down experiment. Loss of BRCA1 function in organoid and mouse models altered gene expression in luminal progenitor cells and diverted them from their predestined differentiation program.

Furthermore, it has been shown that even though the basal branch tumors typically do not express estrogen receptor, estrogen exposure plays an important role in the initiation of basal tumors.

Interestingly, transformation of the basal progenitors leads to development of a metaplastic carcinoma, a rare breast cancer subtype, characterized by low expression of the Claudin genes.

Another model for breast cancer etiology has been proposed by mathematical modeling of age- related breast cancer incidence. The key observations were that the age distribution of breast cancer risk had a bimodal shape and that the risk of ER-negative breast cancer decreased with age.

According to the suggested model, the two etiologic breast cancer subtypes would include ER-

negative early-onset breast cancer and ER-positive breast cancer with a linearly increasing

cumulative lifetime risk. The division was suggested to be caused by differences in the steps of

tumor progression arising from biological differences between pre- and postmenopausal

mammary glands.

23 2.3 Breast cancer treatment

The primary treatment for breast cancer is surgical removal of the tumor mass with sufficient margin or excision of the entire mammary gland. Radiation is commonly recommended to lower the risk of local recurrence. Furthermore, the treatment regimen generally includes a combination of hormonal, cytostatic, and biological adjuvant therapies to reduce the risk of death due to metastatic disease. Neoadjuvant therapy, i.e. therapy preceding the surgery, can be used for increasing the operability of an inflammatory or locally advanced breast cancer, or for reducing the tumor size for better success of a breast-conserving operation.

2.3.1 Adjuvant endocrine therapy

Estrogen receptor expression in tumor cells is an indication of benefit from adjuvant endocrine therapy.

The rationale behind endocrine therapy is that the proliferation of tumor cells depends on uninterrupted supply of ovarian hormones.

Tamoxifen, the first antiestrogen drug used in an adjuvant setting, is still routinely used in treatment of premenopausal women with ER-positive breast cancer.

Tamoxifen competes with endogenous estrogens in binding to estrogen receptor. The tamoxifen-receptor complex is able to dimerize and bind to estrogen-responsive elements in DNA, but does not induce transcription of estrogen target genes in breast tissue.

Tamoxifen therapy in premenopausal women may be combined with drugs suppressing ovarian function, e.g. luteinizing hormone releasing hormone (LHRH) agonists, which overstimulate LHRH receptors in the pituitary gland, thus reducing LHRH levels and estrogen production in the ovaries.

In postmenopausal women, the primary estrogen source is androgen metabolism in peripheral tissues. Aromatase inhibitors (AIs) bind either covalently or reversibly to aromatase enzyme, blocking androgen conversion, and outperform antiestrogens in efficiency in treating post- menopausal women with ER-positive breast cancer.

The most common mechanism for acquisition of resistance to endocrine therapy involves activation of the PIK3CB-AKT-MTOR pathway (phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit beta; AKT serine/threonine kinase 1; mechanistic target of rapamycin kinase).

A specific MTOR inhibitor, Everolimus, has recently been approved for treatment of advanced ER-positive breast cancer.

Furthermore, a specific PIK3CB inhibitor has recently been reported to be effective in clinical trials, but it has also been associated with severe adverse effects, preventing its wider use.

There is a continuously ongoing research effort to further improve endocrine regimens. The most recent advances include cyclin dependent kinase (CDK4/6) inhibitors and histone deacetylase inhibitors as well as refinement of AI therapy by addition of adjuvant bisphosphonates to protect against fractures and bone metastasis.

2.3.2 Other targeted biological therapies

Biological cancer therapy is based on agents that specifically target the drivers of tumorigenesis.

Since the driver events often arise as a consequence of somatic mutations or by re-activation of

embryonic pathways, they should not be present in the healthy adult system,

and the systemic

adverse effects of the treatment should be minimal. On the other hand, the treatment benefit is

restricted to the subgroup of patients whose tumors carry these specific aberrations.

Besides the estrogen receptor, the most important target in breast cancer therapy is the HER2 receptor, which is overexpressed or amplified in about 15% of breast cancers. Trastuzumab, a monoclonal antibody against HER2, was first introduced in a clinical trial in 1998. After the release of impressive and consistent results of two large trials combining trastuzumab with adjuvant chemotherapy in 2005, trastuzumab was widely adopted for adjuvant treatment of HER2- positive breast cancer.

Subsequently, other anti-HER2 agents and regimen modifications have become a focus of intensive research.

Dual inhibition of the HER2 pathway with a combination of drugs targeting different components of the pathway has been the most promising approach.

Poly(ADP-ribose) polymerase (PARP) inhibitors represent the most promising emerging therapy for breast cancer. PARP silencing is lethal for cells devoid of BRCA1 or BRCA2 function. Thus, the PARP inhibitors are an ideal therapy with minimal side-effects for carriers of germline BRCA1 or BRCA2 mutations.

Furthermore, since most of the moderate-risk breast cancer susceptibility genes, including ATM, PALB2, FANCM, and CHEK2, are involved in the same pathway controlling DNA repair via homologous recombination, PARP inhibitors may have potential also in treatment of breast cancer patients with germline mutations in any of these other risk genes.

The potential of other biological therapies for breast cancer, including antiangiogenic agents and inhibitors of epidermal growth factor receptor (EGFR), has been studied intensively. To date, however, the success has been limited.

2.3.3 Adjuvant chemotherapy

Chemotherapy refers to a wide range of cytotoxic agents targeting actively proliferating cells on a systemic level. The rationale is that since active proliferation is what distinguishes cancer from normal tissue the agents would have selective toxicity for cancer cells. However, typical side- effects include immunosuppression and hair loss due to killing of actively dividing precursor cells.

Ovarian suppression or premature menopause resulting from death of germ cell precursors may contribute to chemotherapy efficacy, but is an unwanted side-effect for younger women.

Cytotoxic compounds have been used in combinations to treat breast cancer since the 1970s.

CMF (cyclophosphamide; methotrexate; 5-fluorouracil) was introduced in a clinical trial in 1973

and is still included in the recommended adjuvant regimens.

CMF combines one alkylating agent (cyclophosphamide) and two antimebolites (methotrexate and 5-fluorouracil/

capecitabine) administered at regular intervals separated by periods of recovery.

Cyclophosphamide and its metabolites (4-hydroxycyclo-phosphamide and aldophosphamide) pass through the circulation in chemically inactive forms. Cytotoxic phosphoramide mustard is generated from aldophosphamide only in target cells with a low concentration of aldehyde dehydrogenase (ALDH), which is able to metabolize aldo-phosphamide into inactive carboxyphosphamide. Many normal tissues have sufficiently high expression of ALDH to protect them from the toxic side-effects.

Phosphoramide mustard causes DNA cross-strand links at guanine nucleotides, leading to DNA damage and cell death. Additionally, cyclophosphamide has antiangiogenic and immunostimulatory effects, which may contribute to its efficacy as a chemotherapeutic agent.

Methotrexate and 5-fluorouracil specifically block two enzymes required for thymine metabolism, dihydrofolate reductase and thymidylate synthase, respectively,

123, 124

25

Anthracyclines (epirubicin, doxorubicin/adriamycin) stabilize topoisomerase II alpha (TOP2A) complex bound on cleaved DNA. This results in a mitotic catastrophe when the cell cycle proceeds from G2- (gap 2) to M-phase (mitosis).

TOP2A is expressed from late S-phase (synthesis) to M-phase and regulates DNA topology during DNA replication.

TOP2A copy number aberrations, HER2 gene amplification, high proliferation rate, and ki67 expression have been suggested as individual markers for benefit from anthracycline therapy.

On the other hand, NQO1 (NAD(P)H:quinone oxidoreductase) germline variant rs1800566 has been suggested as a counter indication for anthracycline use.

However, none of these markers have currently been included in the treatment guidelines.

Anthracyclines have proven superior to the older- generation CMF in prolonging overall and relapse-free survival.

The treatment efficacy has, however, come at the cost of an increasing amount of adverse side-effects. Anthracyclines cause neutropenia in up to 25% of patients, secondary leukemia in about 1% of patients, and increased short- and long-term risk of heart failure, depending on the cumulative dose.

Taxanes (paclitaxel and docetaxel) represent a newer generation of chemotherapeutic agents.

Their administration alternately with anthracyclines improves patient prognosis relative to anthracycline-based therapies alone.

Taxanes promote rapid assembly of overly stable microtubules, stalling cells in the M-phase and leading to apoptosis or immune-related cell lysis.

138

Adverse effects include fatigue, neutropenia, peripheral neuropathy, and decreased cognitive performance as a result of axon demyelination and direct neural cell toxicity.

Other microtubule-targeting drugs with similar toxicity profiles include eribulin and vinca alkaloids, which are primarily used in treatment of metastatic breast cancer.

Chemotherapy is recommended as adjuvant treatment for patients at intermediate or high risk of recurrence or progression.

Few biological markers indicating sensitivity or resistance to any specific agent have been identified thus far.

The choice of an optimal treatment combi- nation for an individual patient is made following general guidelines, and adverse effects are monitored and the agent or dose is changed if necessary.

As the most efficient agents tend to be increasingly toxic, further research is required to identify the patient groups for whom the benefit will outweigh the harm, and to develop better-tolerated means for drug administration.

2.4 Genetic predisposition to breast cancer 2.4.1 Breast cancer heritability

Genetic factors account for slightly less than one-third of breast cancer incidence; twin studies have estimated the heritability to be about 31%.

Having a first-degree relative with breast cancer is associated with a twofold increase in breast cancer risk, and the risk has been suggested to increase along with the number of affected relatives, so that the cumulative lifetime risk for a woman with three affected first-degree relatives would reach 30-40%.

The familial aggregation of breast cancer is attributable to shared environmental factors as well as genetic variants acting multiplicatively to increase the risk, with common low-penetrance variants modifying the penetrance of higher-risk mutations.

Typically for a polygenic disorder, the currently known genetic risk variants are distributed on a

diagonal, reaching from rare high-risk mutations to moderate-risk mutations and further to

common low-penetrance variants, covering an area in which the balance between effect size and

variant frequency is sufficient to make them biologically or clinically interesting (Figure 4).

The breast cancer-predisposing variants have been discovered mainly using three different methods, each best-suited to discovery of certain classes of variants on the risk-frequency axis (Figure 4, Table 4). Historically, the golden age of linkage studies preceded the burst of resequencing of candidate genes, which has given way to the currently dominating genome-wide studies.

Figure 4. Breast cancer-predisposing genes and variants. High- and moderate-risk genes are named in the

figure, the rest being low-penetrance variants with a relative risk below 1.5.

The most important single risk factors are truncating mutations of BRCA1 or BRCA2 genes,

explaining about 16% of breast cancer genetic background (Figure 5). These together with

mutations in other high- or moderate-risk genes add up to about 20%.

Common predisposing

variants with low effect sizes account for an additional 18%. Interestingly, a recent genome-wide

association study identifying multiple novel predisposing variants indicated that an equally large

fraction of breast cancer predisposition could be explained by as yet unidentified variants, which

were included on or imputable from the genotyping array, but which alone did not exceed the

genome-wide significance threshold (‘Unknown common variants’ in Figure 5).

Still, a large

proportion of breast cancer genetic background remains unexplained. The missing heritability has

been suggested to be attributable to rare or private variants, structural variants, imprinting, or

epistasis resulting from genetic interactions.