Copy number alterations associated with clinical features in an underrepresented population with breast cancer.

BACKGROUND: As the most incident tumor among women worldwide, breast cancer is a heterogeneous disease. Tremendous efforts have been made to understand how tumor characteristics as histological type, molecular subtype, and tumor microenvironment collectively influence disease diagnosis to treatment, which impact outcomes. Differences between populations and environmental and cultural factors have impacts on the origin and evolution of the disease, as well as the therapeutic challenges that arise due to these factors. We, then, compared copy number variations (CNVs) in mucinous and nonmucinous luminal breast tumors from a Brazilian cohort to investigate major CNV imbalances in mucinous tumors versus non-mucinous luminal tumors, taking into account their clinical and pathological features.
METHODS: 48 breast tumor samples and 48 matched control blood samples from Brazilian women were assessed for CNVs by chromosome microarray. Logistic regression and random forest models were used in order to assess CNVs in chromosomal regions from tumors.
RESULTS: CNVs that were identified in chromosomes 1, 5, 8, 17, 19, and 21 classify tumors according to their histological type, ethnicity, disease stage, and familial history.
CONCLUSION: Copy number alterations described in this study provide a better understanding of the landscape of genomic aberrations in mucinous breast cancers that are associated with clinical features.

INTRODUCTION

As the most incident tumor among women worldwide, breast cancer also causes the highest number of deaths in the female population, especially in developing countries where the diagnosis of late‐stage disease is made in most cases (World Health Organization, 2018). Breast cancer is also a heterogeneous disease, where the individual's genetics in combination with the influence of tumor histological type, molecular subtype, and tumor microenvironment contribute to disease progression. A better understanding of these factors in relation to early diagnosis and disease treatment impacting overall survival is critical (Cecilio et al., 2015). In addition, differences between populations and also environmental and cultural factors significantly affect the origin and evolution of the disease, and therefore bring additional therapeutic challenges (IARC, 2014).

Ductal carcinomas account for more than 70% of breast tumors and include all histological types that cannot be classified into defined types. Their prognosis depends mainly on the molecular subtype and other features such as stage that includes tumor size, affected lymph nodes, and the presence of metastasis (IARC, 2014). Among the histological types of breast tumors, mucinous carcinomas of the breast are rare and comprise 1%–6% of all breast tumor cases, especially in women over 75 years of age (Ha, Deleon, & Deleon, 2013). Genomic studies involving this type of tumor are understudied, in part because of its low incidence. A portion of the cases that did not respond well to standard‐of‐care treatments were characterized as presenting positivity for ERBB2 and P53, with a higher probability of metastasis. Cases that present the mucinous histological type in less than 90% of the tumor or, in association with invasive ductal tumors, also tend to be more aggressive (Lacroix‐Triki et al., 2010). In addition, chromosome analysis in pure mucinous tumors in conjunction with other histological types showed gains in 1q and 16p arms and losses in the 16q and 22q arms, despite lower genetic instability compared to invasive ductal tumors. Studies have shown that a number of genes such as ERBB2, FGFR1, CCND1, FGF3, FGF4, FGF19, PIK3CA, BRCA1, TSC2, STK11, AKT3, and ESR1, among others, present changes in tumors of this type (Lei, Yu, Chen, Chen, & Wang, 2016; Ross et al., 2016). Hence, a better understanding is needed of altered genomic landscape in aggressive, treatment‐refractory mucinous breast tumors.

Majority of defined breast cancer molecular subtypes were derived from ductal invasive breast tumors, and largely lacked profiling from other histological types of breast tumors (Dieci, Orvieto, Dominici, Conte, & Guarneri, 2014; Perou et al., 2000; The Cancer Genome Atlas [TCGA], 2012). Few studies have described how molecular features from different histological types may influence treatment response (Caldarella et al., 2013; Weigelt et al., 2008). Mucinous tumors are often described as Luminal A, and recent studies have shown that this subtype tended to have worse responses to cytotoxic agents and develop resistance to chemotherapy compared earlier to other histological subtypes (Araki & Miyoshi, 2018; Martelotto, Ng, Piscuoglio, Weigelt, & Reis‐Filho, 2014).

Although breast cancer comes in many histological forms, the mucinous histological type remains understudied, in part due to its low incidence. In addition, the Brazilian population of breast cancer patients is understudied regardless of the tumor phenotype. Current demographic data shows that the Brazilian population is composed of mixed ethnicities (Instituto Brasileiro de Geografia e Estatística [IBGE], 2018). Since Brazil is a genetically underrepresented population, studies that include Brazilian cohorts may uncover previously unknown genetic drivers of therapeutic resistance and lead to the discovery of new biomarkers. The genetic composition of tumors in the Brazilian population is also dissimilar from that of populations living in other regions of the globe, even in neighboring Latin America countries, since the patterns of colonization and intrinsic miscegenation between colonizers and the native populations vary markedly across these countries (Giolo et al., 2012; Popejoy & Fullerton, 2016).

In this study, we compare the genomic features in terms of copy number variations (CNVs) in mucinous and nonmucinous luminal breast tumors of a Brazilian cohort. With this methodological approach, we were able to describe major CNV imbalances in mucinous tumors versus ordinary luminal A/B tumors in association with clinical and pathological features.

SUBJECTS AND METHODS

The procedures for obtaining the samples used in this study, as well as the informed consent form signed by all the women participating in this study, followed the recommendations of the Declaration of Helsinki and were approved by the Research Committee of CAISM—Women's Hospital/UNICAMP (approved project n.° 082/2013) on 12/12/2013 and by the Research Ethics Committee of UNICAMP and CONEP—National Research Committee (approved project n.° 1.166.843) on 7/30/2015. Tumor and blood samples of women who agreed to participate in the study and signed the consent form for this purpose were collected by the Division of Gynecological Oncology and Breast Pathology of CAISM—Women's Hospital/UNICAMP. Medical records were reviewed to obtain women clinical and epidemiological data. For this study, only ductal and mucinous tumors with or without other minor components were selected after the histopathological characterization of the biopsy. A skilled pathologist selected tumor and normal areas for microdissection. Tumor areas were used to obtain 10μm fragments from which DNA extraction using phenol/chloroform protocol was performed. A similar protocol was used for DNA retrieval from blood samples.

DNA was verified in agarose gel and considered adequate only when hosting >80% of integrity. DNA was then diluted at concentrations between 40 and 60 ng/μl, which were verified by the Epoch spectrophotometer (Biotek®, Winooski, VT). These concentrations are suitable for use with Affymetrix® Cytoscan™ HD Array assay kits (Thermo Fisher Scientific Inc., Santa Clara, CA). The protocol was performed as per manufacturer recommendations, comprising the steps of preparing the genomic DNA, digestion, ligation, PCR, purification, quantification, fragmentation, labeling, hybridization, washing, staining, and chip scanning. After scanning, data was processed by Affymetrix Molecular Diagnostic Software (AMDS) and quality control was generated by ChAS analysis software (Chromosome Analysis Suite, Affymetrix®). 48 chips were hybridized for the tumor samples and 48 chips for the blood samples of the same woman, the latter being used as control of constitutive CNVs.

For CNV analysis, data were normalized via the ASCRMA and raw copy algorithms. Then, the normalized data was segmented using the Parent‐specific circular binding segmentation (Olshen et al., 2011), copynumber, GADA, and CBS protocols. Only alterations contemplating at least 25 microarray probes for deletions or 50 probes for amplifications were considered, along with fragments of 100kb and with low‐rank representation (LRR) ≤ −0.3 for deletions and LRR ≥0.3 for amplifications. The data were also evaluated by the intersection of methods performed and described above: only samples with CNVs present in three or more of the methods were considered as altered for the variation found. Afterward, two statistical tests were applied to rank the most relevant CNVs by comparing between ductal and mucinous samples and also to evaluate the most relevant CNVs in relation to the clinical and pathological characteristics. Functional pathways associated with these CNVs were searched using DAVID 6.8 (The Database for Annotation, Visualization and Integrated Discovery, p‐value ≤0.05) (Huang, Sherman, & Lempicki, 2009) and UCSC Table Browser was used to retrieve information on variants already described that are in association with the verified CNVs.

RESULTS

Table 1 shows the clinical and epidemiological features of the women included in the study, per tumor histological type. The majority of the women were above 45 years of age and were postmenopausal. Disease stage was predominantly I or II. About 81% of the women were Caucasian versus 19% Afro‐descendants. Fourteen women reported one or more cases of breast cancer in their families. Majority of the cases (n = 35) were classified as Luminal A, 11 Luminal B and 2 Luminal B/HER2 enriched.

Table 1

Description of the clinical and epidemiological features of the women included in the study

	Mucinous Samples		Ductal Samples
	n	%	n	%
Age at diagnosis
35–45	0	0	3	6
>45	10	21	35	73
Ethnicity
Caucasian	9	19	30	62
Afro‐descendant	1	2	8	17
Menopausal status
Post	9	19	28	58
Pre	1	2	10	21
Disease stage
I/II	9	19	27	57
III	1	2	11	22
Familial history—breast cancer
Yes	1	2	13	27
No	9	19	25	52
Molecular subtype
Luminal A	7	15	28	58
Luminal B	1	2	10	21
Luminal B/HER2	2	4	0	0

The frequencies of CNVs, by chromosome, in relation to clinical/pathological data are shown in Table 2. Interestingly, the described CNVs related to later disease stage and presence of family history were found on the same chromosomes (chr 5, 19 and 21), although more CNVs in chromosome 19 (46%) were associated with late stage and more CNVs in chromosome 21 (49%) were associated with family history. For histological type, when comparing ductal to mucinous breast carcinomas, CNVs in chromosomes 1 and 8 accounted for almost 49% of all alterations found in the mucinous tumors analyzed. Similarly, CNVs in chromosome 19 summed to 46.27% of alterations related to later disease stage, alterations in chromosome 21 summed to 49.42% for familial history presence and chromosomes 1 and 17 summed to 31.08% for ethnicity (Caucasian).

Table 2

Percentage of CNVs found, by chromosome, including association with clinical and pathological features

	Chromosome	Percentage (%)
Histological type	chr8	27.81
	chr1	21.16
	chr15	0.88
	chr16	7.00
	chr14	6.67
	chr12	4.82
	chr11	4.80
	chr18	4.32
	chr17	4.09
	chr19	3.16
	chr6	2.50
	chr13	2.50
	chr20	1.77
	chr22	1.54
	chr3	1.47
	chr9	1.37
	chr21	1.23
	chr7	0.95
	chr4	0.90
	chr2	0.62
	chr10	0.35
	chr5	0.10
Ethnicity	chr1	17.85
	chr17	13.23
	chr10	12.15
	chr19	12.12
	chr8	8.71
	chr16	7.88
	chr11	5.83
	chr14	5.51
	chr20	2.61
	chr13	2.59
	chr6	2.11
	chr12	1.82
	chr21	1.54
	chr3	1.51
	chr5	1.14
	chr22	0.87
	chr2	0.71
	chr4	0.70
	chr7	0.52
	chr9	0.29
	chr18	0.21
Disease stage	chr19	46.27
	chr21	35.07
	chr5	18.66
Familial history	chr21	49.42
	chr19	28.79
	chr5	21.79

Table 3 describes the genes related to CNVs in each chromosome, according to the features they were most associated with. Logistic Regressions and Random Forests models were used to assess these regions, comparing the genomic profiles of the samples, in which a power of discrimination (AUC) of 73% was obtained. The CNVs ranking data distinguishing between histological types and other clinical/pathological tumors’ characteristics were assessed to evaluate how these alterations contributed to the separation between considered groups.

Table 3

Description of chromosomes containing most alterations per feature and related genes, verified by logistic regression/random forests

	Histological type			Familial history	Disease stage					Ethnicity
Chromosome	1		8	21	19					1	17
Percentage	21.16%		27.81%	49.42%	46.27%					17.85%	13.23%
Genes	FNDC5	NBPF13P	TATDN1	SAMSN1	FKRP	C19orf70	C19orf24	SYNGR4	CENPBD1P1	RORC	MEIS3P2
	PHBP12	CC2D1B	RNF139‐AS1	SLC19A1	ZNF257	CRX	SF3A2	BSPH1	CTBP2P7	EMBP1	DDX42
	STXBP3	ZC3H11A	LYN	RAD23BLP	ZNF573	ZNF439	UHRF1	GCDH	ZNF155	SLC2A1‐AS1	CRLF3
	DIRAS3	PGBD2	DLGAP2	PRDM15	ZNF468	ZNF700	PLEKHA4	ZNF814	SEMA6B	C1orf131	NXN
	MYCL	POLR3C	DMTN	PCBP3	ZSCAN5A	KLK8	SNORD23	TMEM143	DOT1L	RFWD2	APOH
	FKSG48	TRIM58	ZNF250	ERG	MIR3940	KLK15	OR7A10	CYP4F2	LMTK3	HHAT	ATAD5
	LMO4	HHIPL2	ANK1	LINC00320	ZNF725P	RNU6‐902P	DBP	PCGF7P	RPL23AP2	RGL1	CCDC144CP
	PPIE	CFL1P2	RNF139	RPL31P1	SLC27A5	SAFB	KCNN1	RN7SL513P	TTYH1	RNU2‐12P	PSMD7P1
	VAV3	BMP8B	LINC01109	RPL34P3	ZFP30	ZNF606	SPPL2B	SSC5D	ZNF793	GPATCH2	CTNS
	ZZZ3	EIF4G3	COL22A1	ITGB2	ZNF222	ZIM2	ZNF780B	EHD2	SIGLEC7	FMO4	MYO18A
	LRRC7	SYDE2	PXDNL	DIP2A	ZNF121	SNRPEP4	NTF4	LENG8	ZNF432	ZBTB41	TRIM16L
	AK5	AK2	TTPA	LINC00159	CHMP2A	OSCAR	MIR3189	BIRC8	LAIR1	CYCSP4	CCDC47
	BRINP2	PBX1	FAM66A	RPL23AP4	ZNF69	DHDH	CYTH2	GYS1	SIGLEC6	LAMB3	RAI1‐AS1
	PIGK	SF3B4	IKBKB	SNX19P1	UQCR11	MAN2B1	RNA5SP468	ZNF285	ZNF135	PLD5	SREBF1
	C1orf185	MACF1	CSMD1	C21orf91	ZNF737	TPM3P6	RFX2	CALR	SAFB2	ARID4B	MAP2K3
	GCLM	S100PBP	LRRC69	OR4K11P	CTU1	ZNF813	SPHK2	SIPA1L3	MEGF8	RGS7	SCPEP1
	COL24A1	TFAP2E	NPM1P6	BACH1	SIGLEC5	ZNF446	RASIP1	ZNF571‐AS1	ZNF433	C1orf100	ATP2A3
	S100A16	SMYD3	PKHD1L1	TTC3	SLC5A5	ZNF675	CLPP	ZNF780A	RPL23AP80	ESRRG	SNORD3C
	ABL2	DISP1	CLVS1	PRMT2	CACNG8	RNU6‐1337P	ZNF324	ZNF254	RFXANK	OPTC	SMCR5
	RNF115	ZMPSTE24	LYPLA1	SLC37A1	GALP	MIR517C	ZNF473	ZNF726	PTPRS	WDR64	ZSWIM5P2
	MRPS6P2	MIR4423	PRKDC	TTC3‐AS1	ZNF221	WDR87	ZNF878	SYDE1	CD70	DENND1B	RAI1
	THBS3	CFHR2	MFHAS1	PKNOX1	POU2F2	MIR520H	KLKP1	KIR3DP1	RNU6‐751P	CD1A	MIR33B
	GNG12‐AS1	RPS15AP6	FAM66B	CHODL‐AS1	C5AR2	MIR521‐1	KCNJ14	MYO9B	NPAS1	PPP1R12B	CDRT15L2
	OXCT2	SMAP2	RAB2A	HSF2BP	MEF2B	FUT2	RNA5SP465	KHSRP	GRIK5	ZP4	BRIP1
	TUBB8P6	USP33	ASPH	LSS	ELSPBP1	MIR522	RNA5SP464	CCDC130	FPR1	GNPAT
	HENMT1	TRIT1	MYOM2	SNORD74	KDELR1	ZNF571	LSM4	CYP4F3		FCGR1A
	STRIP1	CCNT2P1	CYCSP22	TIAM1	POLR2E	FUT1	MED25	CACNA1A		HNRNPA1P59
	YARS	ASTN1	TRMT12	RNU4‐45P	ZNF628	ZNF582	MAU2	TMEM145		FLVCR1
	PLD5	RLF	ELP3	GRIK1‐AS2	MZF1	MIR518A2	ZNF235	CA11		IBA57
	ROR1	MTX1	SGCZ	YBEY	ZNF43	FTL	ZNF611	GDF15		COLGALT2
	CHRM3	MUC1	SNORD112	GTF2IP2	C19orf18	CSNK1G2‐AS1	LINC00662	PRR19		SLC35F3
	SPATA17	GIPC2	EBAG9	LINC00314	ZNF45	FAM90A28P	TUBB4A	SYT3		PBX1
	DAB1	TMEM56	THAP1	C21orf58	TNFSF9	OR1AB1P	MIR4321	SULT2A1		MDM4
	RGS7	FUBP1	FUT10	SCAF4	SNRNP70	ZNF420	TMEM161A	ZNF283		TTC13
	RN7SL854P	MIR1256	TRPS1	LINC00315	MIR1227	MIR519A2	KLK5	SULT2B1		LHX4
	EEF1A1P14	CFHR1	FER1L6‐AS2	TIMM9P2	IGSF23	MIR516A2	TMEM160	LYPD5		CDC73
	RNU6‐877P	RNF19B	FGFR1	KCNE2	TNFAIP8L1	MIR7‐3	BAX	GTF2F1		CDC42BPA
	AKT3	HPCAL4	FBXO32	RPS26P5	SIGLEC9	MIR516A1	MIR4323	ALKBH7		COL11A1
	ST6GALNAC5	MIR4421	SNTG1	DSTNP1	KIR3DL3	MIR527	LGALS14	PRR12		CFHR2
	RN7SL370P	LPGAT1	LPL	H2AFZP1	ZNF578	MIR519A1	KCNA7	TPRX1		HMGN1P5
	RNF11	FMN2	NRG1	DYRK1A	ZNF350	BTBD2	OR7C2	EPN1		HIST2H3D
	GJA5	FNDC7	LINC01111	HMGN1P2	CABP5	THEG	OR7A5	ZNF490		SLC2A1
	CLCA4	DNAJB4	POLB	MIRLET7C	JSRP1	SAE1	JUND	RNU6‐982P		TRMT1L
	ARL5AP3	KIF14	SLC10A5	CYP4F29P	URI1	ZNF566	BCL3	MAMSTR		GNG4
	DTL	TRAF5	FER1L6	FAM207A	SPACA4	RPS9	FAM83E	IZUMO1		PTPN14
	NSRP1P1	RCOR3	GSR	LIPI	ZNF28	ASF1B	KIAA1683	RPL39P38		CFHR1
	ACOT11	LIN9	ENPP2	UBASH3A	ZNF709	SUGP2	RN7SL693P	PRRG2		KLHL12
	PLA2G12AP1	DDX59	C8orf22	APP	LRRC25	ZNF702P	CYP4F8	RN7SL121P		CSRP1
	TAF1A	KIAA1522	ZNF705G	LINC00205	PSPN	TRIM28	SLC25A42	KLK9		MROH9
	HMCN1	TRIM46	IMPA1	ANKRD30BP1	ERCC1	SEC1P	RPL7P51	NAT14		CD46
	NME7	RNA5SP52	MIR4662A	MIR3156‐3	AP3D1	CEP89	NOSIP	SIGLECL1		HIST2H2BF
	SLC35F3	TMEM54	DPY19L4	PDE9A	ZNF470	ZNF546	RN7SL708P	ZNF841		KDM5B
	MIR92B	FAM129A	RNF170	SAMSN1‐AS1	ZNF761	PPP1R14A	SIGLEC18P	ACSBG2		BRINP3
	CDC42BPA	C8B	WISP1	MCM3AP‐AS1	CRB3	DENND1C	PGPEP1	SPINT2		CRB1
	CDKN2C	ZNF672	DECR1	NCAM2	PPP1R13L	ZBTB45	ZNF676	KLK6		PDE4B
	EVI5	GBAP1	RPS3AP30	NRIP1	ZNF808	NDUFA3P1	GLTSCR1	KLK10
	FAM102B	RNA5SP44	FABP5	TRAPPC10	ZNF233	ZNF835	GLTSCR2	KLK7
	SRGAP2	TPR	RPL5P23	COL18A1	RUVBL2	RPL28	ZSCAN5C	ZNF763
	MRPS21	NOTCH2	PCM1	POTED	CEACAM22P	ATP1A3	LONP1	LRG1
	OSBPL9	TTC39A	XKR6	CRYAA	CEACAM16	PLIN3	SLC25A41	VN1R85P
	WDR63	USP25	RNU6‐756P	MRPL51P2	KLK12	ZNF443	MIR7‐3HG	VN1R84P
	HSPE1P25	CSMD2	DEFB109P1B	MCM3AP	KLK14	DYRK1B	TCF3	LRRC4B
	HPCA		SYBU	BRWD1‐IT1	NR2C2AP	NTN5	ISOC2	HOOK2
	LINC01057		CHD7	WDR4	UBE2S	ZNF415	TPRX2P	ZFP82
	NFASC		MCM4	MIR125B2	LGALS17A	CGB5	ZNF100	CGB7
	RAVER2			RNU1‐98P	FBL	ARHGEF1	LGALS16	DPP9
	DDAH1			PTTG1IP	RDH13	ZNF563	PAFAH1B3	KLK13
	OR2W3			SIK1	CIRBP	SLC25A23	ZNF665	ZFP28
	GLMN			ERLEC1P1	HDGFRP2	SMIM7	RPL18	CCDC9
	RN7SKP98			CBS	MIDN	RNU6‐1028P	TTC9B	ZNF225
	NEXN‐AS1			DSCAM	TM6SF2	KLK11	RPL18A	ZNF226
	NEXN			TMPRSS15	PRKACA	RNU6‐1041P	SEPT7P8	ZNF320
	DPYD			RNU6‐1326P	HAPLN4	PLIN4	ZNF816	ZNF227
	JAK1			LINC00308	CYP4F22	MLLT1	ZNF540	ZNF112
	PSMB2			PPIAP1	FEM1A	NCAN	ZNF92P2	RABAC1
	PIAS3			PCNT	CLC	MAP3K10	SNORD112	ZNF805
	TMCO2			MIR99A	ZNF568	TPM3P9	ZNF92P3	ZNF230
	RN7SKP12			RSPH1	MAST3	MIER2	HKR1	KLK4
	PDE4B			PCP4	RNU6‐165P	NLRP8	LENG8‐AS1	HMGN1P32
	RNA5SP21			PSMG1	KDM4B	GRWD1	TICAM1	RPL36
	TMEM56‐RWDD3			MIR1283‐2	TPM4	ZNF83	TINCR	ZNF564
	RNA5SP20			LINC00322	LSM7	ABCA7	BRD4	CD33
	WLS			MIR5692B	ZNF234	CSNK1G2	ARRDC5	ZNF223
	SV2A			LINC00307	HSD11B1L	ONECUT3	SNORA68	ZNF224
	CACHD1			RUNX1	ZNF321P	RPL32P34	ZNF574	MIR3188
	RNU7‐121P			ADARB1	RPSAP58	KIR2DP1	INSR	PLEKHJ1
	GPATCH2			GRIK1	FCGBP	ZSCAN5B	CIRBP‐AS1	CGB8
	FAF1			ERVH48‐1	ARMC6	KIR3DL1	RNA5‐8SP4	ZNF208
	ORC1			C2CD2	CATSPERD	CGB1	CEACAM19	ATP8B3
	PRPF38B			POFUT2	CYP4F23P	MIR5684	ADAT3	FAM90A27P
	ENAH			TPT1P1	CLEC11A	SUGP1	SCAMP4	BBC3
	ST7L			RNU6‐286P	SYCE2	KLK2	GATAD2A	SIGLEC14
	NUDT17			CLDN14	ZNF799	CCDC124	LEUTX	ARRDC2
	EPS15			VDAC2P1	GRIN2D	C19orf81	ZNRF4	SHISA7
	ZFYVE9			NDUFV3	ACER1	PLIN5	ZNF284	ZC3H4
	NFIA			BRWD1	AMH	ATP5D	ZNF404	RCN3
	PDE4DIP			U2AF1	PGK1P2	KIR2DL1	C5AR1	DEDD2
	IRF6			RNU6‐113P	UBE2M	KIR2DL3	SHANK1	RAD23A
				KRTAP10‐3		KIR2DL4	ZNF585A	AKT2
							IL12RB1	HIPK4

Table 4 summarizes the annotation findings in terms of functional pathways closely associated to the CNV‐related genes found in the most altered chromosomes, depending on the analyzed trait. Pathways involved with alternative splicing and polymorphisms were mainly associated with most of the altered regions.

Table 4

Functional annotation of genes and enriched pathways associated with CNVs described (DAVID 6.8 Database)

Feature	Chr.	Pathway/function	Gene count	%	p‐value
Histological type	1	Alternative splicing	37	57.8	0.0017
		Splice variant	29	45.3	0.0096
		Cytoplasm	20	31.2	0.015
		Mutagenesis site	11	17.2	0.041
	8	Polymorphism	87	55.1	0.021
		Alternative splicing	81	51.3	0.0056
		Phosphoprotein	69	43.7	0.0014
		Splice variant	67	42.4	0.0012
		Cytoplasm	44	27.8	0.0047
Familial history	21	Alternative splicing	40	40.8	0.0043
		Phosphoprotein	35	35.7	0.0014
		Protein binding	32	32.7	0.038
		Nucleus	28	28.6	0.0003
		Cytosol	18	18.4	0.0055
Disease stage	19	Polymorphism	224	53.8	0.013
		Nucleus	149	35.8	1E‐12
		Transcription	118	28.4	4E‐28
		Metal binding	117	28.1	6E‐13
		DNA binding	106	25.5	1E‐26
Ethinicity	1	Alternative splicing	33	60.0	0.024
		Splice variant	28	50.9	0.01
		Ubl conjugation	10	18.2	0.016
	17	Splice variant	12	54.5	0.0028

Supplementary Table S1 shows the variants already described associated with the CNVs found in this study. The information of cancer‐related phenotypes, genes, and clinical status were assessed in order to better describe variants and their clinical interpretation. It is worth noting that all variants have been previously linked to breast or other forms of human neoplasms and roughly 60% of the CNVs found are of uncertain significance or have conflict of interpretation. Our observations add up to this data to be part of a more accurate interpretation in the future.

DISCUSSION

The results shown describe altered chromosome regions that better classify tumors according to their histological type, ethnicity, disease stage, and familial history. For this set of tumors, almost half of CNVs were present in chromosomes 1 and 8 in mucinous tumors. Similarly, approximately half of the CNVs were found in chromosome 19 when considering cases with later disease stage, and in chromosome 21 when considering cases with presence of family history. Virtually 1/3 of the CNVs were found on chromosomes 1 and 17 when considering cases classified by ethnicity. Also, genes found in CNV regions described in this study were significantly enriched in gene sets related to alternative splicing, polymorphisms, DNA‐binding, transcriptional regulation, phosphoproteins, and mutagenic sites, among others.

Polymorphisms of single nucleotides or of larger DNA fragments and all the other above mentioned pathways are widely associated with the development of cancer in general. Aberrant activation of these pathways in breast cancer is part of the oncogenic mechanisms contributing to disease progression and is the focus of many current studies, since the disruption of mechanisms affected by these pathways may lead to pathogenic events (Mocellin, Valpione, Rossri, & Pooley, 2018; Nicolini, 2017.; Ziv et al., 2017). The description of these changes is very relevant from the point of view of genetic susceptibility.

Alternative splicing has been extensively linked to activation of many tumor processes, because RNA processing is vital for the production of variant proteins that are involved in steps such as angiogenesis, invasion, and antiapoptosis. These processes are also influenced not only by genetic but also environmental factors, for example, chemical and immune responses, heat stress, and DNA damage (Anczuków & Krainer, 2016; Pai & Luca, 2018). Copy number alterations were described as having a particular association to alternative splicing, especially large ones, as seen in our study (Sebestyén et al., 2016; Singh & Eyras, 2017). Also, hereditary breast cancer was reported as enriched for splicing mutations, what often leads to loss of functions in cancer (Rhine et al., 2018).

Thus, in relation to clinical features, namely histological type, ethnicity, disease stage, and familial history, there are particularities worth pointing out. As previously stated, CNVs in chromosomes 1, 8, 17, 19, and 21 explain around half of the alterations found in these samples when associated with one of these clinical characteristics. Alterations in chromosome 1 have been described in 50%–60% of breast tumors and are associated with disease initiation, presence of amplification sites, and a large number of copy number alterations, especially in the 1q arm, which harbor many oncogenes as MYCL1, JUN, NRAS, SHC1, and NCSTN, for example, all verified in samples from our current study (Goh et al., 2017; Orsetti et al., 2006; Silva et al., 2015). Chromosome 8p arm CNVs are widely linked to poor prognosis and metabolic disruptions in breast cancer; moreover, recent studies showed that loss of multiple genes in this region may create greater genomic instability, leading to different effects from loss of a single gene (Cai et al., 2016; Lebok et al., 2015). These two chromosomes are mainly associated with differentiation of ductal and mucinous types, which explain why they were found linked to histological type alterations (Afghahi et al., 2015; Lacroix‐Triki et al., 2010).

Ethnicity was found to be associated with CNVs on chromosomes 1 and 17. A recent study suggests that genes near BRCA1 in 17q are correlated with breast cancer in African Americans (Ochs‐Balcom et al., 2015). However, there is a lack of studies that confirm this association, although genes related to heredity could also contribute to this finding. Interestingly, familial history presence correlated mainly to CNVs in chromosome 21. The gene NRIP1 localized at 21q21 was described to be a susceptibility locus (Ghoussaini et al., 2012) and this region was among our identified CNVs. Also, other chromosome 21 regions were identified, containing genes as SAMSN1, associated with several cancer types such as multiple myeloma, lung cancer, glioblastoma, and RUNX1, implicated as an oncogene and tumor suppressor in breast cancer (Browne et al., 2015; Mercado‐Matos, Matthew‐Onabanjo, & Shaw, 2017; Noll et al., 2014; Yamada et al., 2008; Yan et al., 2013). Late disease stage was correlated to chromosome 19 copy number alterations. These regions have been described in association with high‐grade breast cancers for other studies (Yu, Kanaan, Bae, Baed, & Gabrielson, 2009) and are characterized by aggressiveness and poor prognosis tumors.

Since this study focused on a Brazilian cohort, it is worth mentioning that the genetic composition of the Brazilian population is sharply mixed and is genomically underrepresented in studies that consider variants and tumor markers (Popejoy & Fullerton, 2016). There might be considerable genetic differences underlying tumor biology in these cases, so it is critical to consider understudied populations to better understand breast cancer worldwide. Despite the restricted sample size, this is the first study to evaluate breast cancer CNVs in this specific population, associating them to tumor clinical features. CNV regions identified from these samples and their correlated genes could potentially be different from non‐Brazilian cohorts. In a previous study comparing Brazilian and TCGA (The Cancer Genome Atlas) data (data not shown), we found striking differences between these two cohorts, which were related to genes involved in different carcinogenic pathways, since pathways related to FGF and Wnt were most commonly affected in the Brazilian samples, whereas those associated with cholecystokinin receptor (CCKR) signaling and inflammation mediated by chemokine and cytokine signaling pathways were most commonly affected in the TCGA samples.

We conclude that the copy number alterations described in this study provide an overview of the chromosomal regions affected by CNVs and their association with clinical and pathological features. New molecular targets can be inferred from this study and these CNV regions should be investigated in more detail, potentially driving more dedicated studies focusing on breast tumors from Brazilian cohorts.

CONFLICT OF INTEREST

The authors declare no potential conflicts of interest.

Click here for additional data file.