The fate of BRCA1-related germline mutations in triple-negative breast tumors

Abstract

The preservation of pathogenic BRCA1/2 germline mutations in tumor tissues is usually not questioned, while it remains unknown whether these interact with somatic genotypes for patient outcome. Herein we compared germline and tumor genotypes in operable triple-negative breast cancer (TNBC) and evaluated their combined effects on prognosis. We analyzed baseline germline and primary tumor genotype data obtained by Sanger and Next Generation Sequencing in 194 TNBC patients. We also performed multiple tests interrogating the preservation of germline mutations in matched tumors and breast tissue from carriers with available material. Patients had been treated within clinical trials with adjuvant anthracyclines-taxanes based chemotherapy. We identified 50 (26%) germline mutation carriers (78% in BRCA1) and 136 (71%) tumors with somatic mutations (83% in TP53). Tumor mutation patterns differed between carriers and non-carriers (P<0.001); PIK3CA mutations were exclusively present in non-carriers (P=0.007). Germline BRCA1/2 mutations were not detected in matched tumors and breast tissues from 14 out of 33 (42%) evaluable carriers. Microsatellite markers revealed tumor loss of the germline mutant allele in one case only. Tumors that had lost the germline mutation demonstrated a higher incidence of somatic TP53 mutations as compared to tumors with preserved germline mutations (P=0.036). Germline mutation status significantly interacted with tumor TP53 mutations for patient disease-free survival (interaction P=0.026): In non-carriers, tumor TP53 mutations did not affect outcome; In carriers, those with mutated TP53 tumors experienced more relapses compared to those with wild-type TP53 tumors (36% vs. 9% relapse rate, respectively). In conclusion, we show that loss of germline BRCA1/2 mutations is not a rare event in TNBC. This finding, the observed differences in tumor genotypes with respect to germline status and the prognostic interaction between germline BRCA1-related and tumor TP53 mutation status prompt for combined germline and tumor genotyping for the classification of TNBC, particularly in the context of clinical trials evaluating synthetic lethality drugs.

Introduction

Triple-Negative Breast Cancers (TNBC) are characterized by extensive biological heterogeneity and are frequently deficient in the homologous recombination repair pathway [1]. In the largest multi-centered study to date, 14% of unselected TNBC patients carried loss-of-function germline mutations in homologous recombination genes [2]. Depending on the age at diagnosis, family history and ethnic descent, 8-16% of TNBC cases can be attributed to germline BRCA1 mutations [2-4].

The standard perception of how BRCA1 mutations contribute to tumor formation is that of classic tumor suppressor genes, i.e., preservation of the mutant and loss of the wild-type allele in tissue stem cells from which tumors originate [5]; Notably though, in this early and in most following studies, loss-of-heterozygosity (LOH) was only present in a fraction of tumors tested, while the lost allele (wild-type or mutant) was not clearly presented. Interestingly, loss of the wild-type allele may not be a prerequisite for BRCA-related tumorigenesis [6]. LOH was demonstrated in normal and precancerous breast tissue as well [7,8], while loss of the mutant inherited BRCA1 allele was occasionally observed in breast lesions [8]. However, LOH testing is traditionally performed for germline variant classification into pathogenic, while the preservation of BRCA1 germline mutations in matched tumor tissues is practically never questioned.

At the tissue level, TNBC are hallmarked by mutations in TP53, while mutations in other genes are infrequent and non-recurrent [9-11]. Tumors in BRCA1/2 carriers frequently have TP53 mutations [12] and they are also characterized by specific patterns of genomic instability [13]. Germline BRCA1/2 status and the associated genomic instability may be particularly important for treatment selection; Such carriers seem to benefit from platinum-based treatment and not from the conventionally used taxanes, a difference not observed in non-carriers [14,15]. It remains unknown whether clinical benefit from conventional or investigative treatments also depends on specific tumor genotypes, in addition to germline mutation status.

In order to address the above open questions, herein we investigated whether baseline germline mutations were preserved in primary tumors from carriers. We also compared germline and tumor genotypes for obtaining specific patterns associated with prognosis upon anthracyclines-taxanes-based adjuvant chemotherapy, which is still used for TNBC patients.

Patients, materials and methods

We retrospectively studied available germline and primary tumor genotype data from 196 TNBC patients who had been treated during 1997-2012 in adjuvant clinical trials by the Hellenic Cooperative Oncology Group (HeCOG), i.e., HE 10/97, HE 10/00, HE 10/04A, HE 10/04B, HE 10/05, HE 10/08, HE 10/10, or had been treated with taxanes-based adjuvant chemotherapy in HeCOG affiliated clinical centers, as previously described [16] (Figure 1). Clinical material (formalin-fixed, paraffin-embedded [FFPE] tissue blocks and peripheral blood) had been collected prior to initiation of adjuvant chemotherapy. Germline and tumor genotype data had been obtained in separate series in two different laboratories. Patients had provided written consent for the use of their biologic material for research purposes and the study was approved by the Bioethics Committee of the Aristotle University of Thessaloniki School of Medicine (#77/10 June 2014) and by the Institutional Review Board of Papageorgiou Hospital of Thessaloniki (#725/10 May 2013). Germline status [2,3] and tumor genotype data [16] were previously published for extended cohorts including these patients. Peripheral blood samples had been submitted for germline testing based on TNBC phenotyping (ER/PgR/HER2 negative with immunohistochemistry [IHC]) upon diagnosis in local pathology laboratories. Patient characteristics are shown in Table 1.

Figure 1.

REMARK diagram. Tissue material was formalin-fixed paraffin-embedded (FFPE). *: Patients were treated in the context of clinical trials or in HeCOG affiliated centers, as indicated; ^: Hot-spots testing for the Greek founder alleles; ^^: The entire coding sequence and splice junctions of the two genes was sequenced; **: Testing targeted the germline mutation in available matched FFPE tissues; Positive: Germline mutation carriers; MS ID: Microsatellite testing for validating blood and tissue sample origin from the same patient; GT: Cases with available excess peripheral blood and tumor tissue DNA for interrogating the presence of the identified germline mutation in tissue material following matched MS ID test results.

Table 1.

Patient demographic, clinicopathological, germline and tumor genotype data

Variable	N (%) for categorical variables
Age (N: 194)
Mean ± SD	51.4±12.6
Median	51
Min-max	21-77
Menopausal status
Postmenopausal	99 (51.0%)
Premenopausal	95 (49.0%)
Tumor size
≤2 cm	74 (38.1%)
>2 cm	118 (60.9%)
Missing/not applicable	2 (1.0%)
Positive nodes
≥4	42 (21.7%)
0-3	149 (76.8%)
Missing/not applicable	3 (1.5%)
Adjuvant hormonotherapy
NO	165 (85.0%)
YES	29 (15.0%)
Adjuvant radiotherapy
NO	54 (27.8%)
YES	140 (72.2%)
Histological grade
I or II	28 (14.4%)
III	166 (85.6%)
Histological type
Lobular	6 (3.1%)
Medullary	11 (5.7%)
Metaplastic	5 (2.5%)
Ductal, no special type	159 (82.0%)
Other	13 (6.7%)
TNBC upon central testing
NO	27 (13.9%)
YES	167 (86.1%)
Ki67 (%; N: 177)
Mean ± SD	53.9±31.7
Median	56
Min-max	0-100
Basal^
YES	160 (82.5%)
NO	26 (13.4%)
Missing	8 (4.1%)
TILs density % (N: 188)
Mean ± SD	22.2±21.0
VMedian	15
Min-max	0-85
Disease-free survival data
N patients	194
Median follow up (months)	59
Relapse, N (%)	54 (27.8%)
Event free at 3 years, N (%)	144 (74.2%)

^{^}EGFR and/or CK5 IHC positive.

Tumor processing and genotyping

Central tumor assessment had been carried out at the Laboratory of Molecular Oncology (MOL; Hellenic Foundation for Cancer Research/HeCOG/Aristotle University of Thessaloniki, Thessaloniki, Greece) and included histology review; ER/PgR/HER2 IHC and HER2 FISH where needed; Ki67, CK5 and EGFR IHC [17]; And, stromal tumor infiltrating lymphocytes (TILs) density as previously described [18]. IHC, FISH and DNA extraction had been applied on in-house low-density tissue microarrays (TMA) containing two 1.5 mm cores per tumor. FFPE DNA samples were processed for targeted next generation sequencing (NGS) in a Proton sequencer (Life Technologies/Ion Torrent; Paisley, UK) with a previously published tissue panel [19] that had been designed to target mutations in 43 TNBC related genes based on published comprehensive genomic data [9,11]; The BRCA1-related targets within this panel covered areas <500 nt per gene [19]. Samples were accepted for further evaluation if 90% of amplicons had been read >100 times. Variants obtained from Ion Reporter v.4 were filtered out if non-annotated, if indels with GC-stretches (reading artifacts with semiconductor sequencing); If position coverage <100 and if variant coverage <40. The present samples had median mean depth 875 (mean: 1338; Min-max: 433-9744); Median uniformity 81 (mean: 80; Min-max: 74-91); Median number of variants 12 (mean: 14; Min-max: 5-133). Variants were called mutations if these were amino acid or splice site changing, and if minor allele frequency (MAF) in dbSNP was <0.1%. Clonal mutations were called for variant allele frequency >25% [20]. Informative NGS data from FFPE tumors were initially available for 193 patients (Figure 1).

Germline genotyping

Germline genotyping, using DNA extracted from peripheral blood, was implemented at the Molecular Diagnostics Laboratory (MDL) at NCSR Demokritos, Athens, Greece. All participants signed informed consent prior to genetic testing. The study was approved by the Bioethics committee of NCSR “Demokritos” (240/EHΔ/11.3), updated on February 2014. Screening initially involved testing for founder and recurrent BRCA1 mutations, all of which cluster in the BRCA1 C-terminal (BRCT) domain, as previously published [21,22]. All individuals found negative at these loci, were Sanger sequenced for the entire BRCA1 and BRCA2 coding sequence and splice junctions genes or tested with NGS by a multi gene panel (Figure 1), as previously described [2]. All germline mutations that were identified with NGS were orthogonally validated with Sanger sequencing.

Mutations introducing a premature termination codon were classified as loss-of-function. The missense variants detected were interpreted based on the American College of Medical Genetics and Genomics guidelines [23]; Three BRCA1 missense alleles were classified as definitely pathogenic (p.C61G, p.G1738R and p.V1833M).

Validation of germline mutations in tumor tissues

The panel used for tissue genotyping did not systematically target the entire coding region of cancer predisposing genes and only incidentally revealed such mutations in tumors. Therefore, it was necessary to individually validate whether the identified germline mutations were preserved in the corresponding tumors. For this purpose, germline mutations identified at MDL were interrogated in matched tissue DNA (tumor and/or normal) with Sanger sequencing at MOL. Re-sequencing was applied for matched blood and tumor samples using the same primers targeting the germline mutation. This validation step was possible for 36 patients with adequate matched tissue DNA (Figure 1).

But validation of the detected germline mutations failed in tumor and normal tissue samples from 17 patients and therefore it appeared mandatory to perform identity testing on peripheral blood and tissue DNA. The common origin of the samples was assessed by microsatellite identity testing (MS-ID). Markers for BRCA1 and BRCA2 were used as indicated in Table 2.

Table 2.

Microsatellite markers and ID test results in the 16 cases with discordant germline/tumor mutation status

Case	DNA sample	BRCA1 locus (17q21.31)			BRCA2 locus (13q13.1)
		D17S323*	D17S1322*	D17S801**	D13S260**	D13S1698**	D13S267**
1	Germline	151/157	116/119	228/239
	Tumor	151/157	116/119	228/239
2	Germline	148/150	113/116	232/232
	Tumor	148/150	113/116	232/232
	Normal	148/150	113/116	232/232
3	Germline	148/151	113/116	232/234
	Tumor	148/151	113/116	232/234
4	Germline	148/151	119/122	238/240
	Tumor	148/151	119/122	238/240
5	Germline	148/150	116/116	232/232
	Tumor	148/150	116/116	232/232
6	Germline	148/150	116/116	232/238
	Tumor	148/150	116/116	232/238
7	Germline	148/150	113/116	232/232
	Tumor	148/150	113/116	232/232
	Normal	148/150	113/116	232/232
8	Germline	146/146	116/116	230/232
	Tumor	146/146	116/116	230/232
9	Germline	148/150	116/116	232/232
	Tumor	148/150	116/116	232/232
10	Germline	148/150	116/116	230/232
	Tumor	148/150	116/116	230/232
11	Germline	148/150	116/116	232/238
	Tumor	148/150	116/116	232/238
12	Germline	148/150	116/116	230/232
	Tumor	148/150	116/116	230/232
13	Germline	Not available
	Germline#	150/160	116/119	228/236	159/163	153/155	147/154
	Tumor	148/150	116/119	232/236	159/160	151/155	148/154
14	Germline	148/150	113/116	232/242	155/165	155/161	141/141
	Tumor	148/150	116/116^	232/232^	165/165	150/155	143/143
15##	Germline	151/157	116/125	228/232	155/167	159/171	149/156
	Tumor	151/153	122/122	240/240	159/160	164/151	148/154
	Tumor (2nd)				159/160	164/151	148/154
16##	Germline	150/159	116/119	240/240
	Tumor	148/150	116/116	232/232

^*Intragenic markers;

^**Extragenic markers;

^{^}Germline allele lost in tumor (case TN021);

^#Genomic DNA was unavailable from the proband, so her daughther’s DNA was used instead;

^##Mismatched cases, excluded from the analysis.

Two cases were excluded from further analysis since there was a mismatch for blood/tissue samples, emphasizing the necessity of the undertaken testing for performing reliable germline/tumor comparisons. Therefore, germline mutations were analyzed in 194 patients, out of whom 191 had tumors with informative tissue panel NGS genotypes; germline/tumor mutation status could be compared in 33 patients (Figure 1); Out of these patients, 31 had informative tumor NGS genotypes.

Statistics

Categorical data are presented as frequencies and corresponding percentages; While continuous data are presented as mean, standard deviation, median and range values. The Fisher’s exact or Pearson chi-square tests were used for group comparison of categorical data, while for continuous data the non-parametric Mann-Whitney test were used where appropriate.

Disease-free survival (DFS) was measured from the date of diagnosis until verified disease relapse or death, whichever occurred first, or loss from follow-up. Patients who survived without relapse or were lost from follow-up without previous relapse were censored at the date of their last contact. Survival curves were estimated using the Kaplan-Meier method and compared with the log-rank test. The associations between factors of interest and relapse rate were evaluated with hazard ratios (HR) estimated with univariate and multivariate Cox proportional hazards models. HR were estimated for the following factors: Age (≥50 vs. <50), menopausal status (post- vs. premenopausal), tumor size (>2 vs. ≤2 cm), number of positive nodes (≥4 vs. 0-3), adjuvant radiotherapy (yes vs. no), mutation (any vs. no), germline (positive vs. negative), TP53 mutation (mut vs. wt), TILs continuous in 10% increments. Evidence for effect modification between the above factors in the multivariate analyses was evaluated by fitting additional interaction terms in the models with the main effects only and comparing the resulting nested models with the likelihood ratio test.

All analyses were performed with the SAS software (SAS for Windows, version 9.3, SAS Institute Inc., Cary, NC, USA). Statistical significance was 2-sided P=0.05.

Results

Tumor genotypes

With the tissue panel we recorded 208 mutations distributed in 136 of the 191 tissue NGS informative tumors (71.2%); Mutation distribution, mutation type and patterns are shown in Figure 2A. TP53 was by far the most frequently mutated gene (Figure 2A and 2B). PIK3CA mutations were in the known hotspot codons Glu542, Glu545 and His1047; CDH1 mutations were mostly loss-of-function (Figure 2A). Among the 136 mutated tumors, 27 were co-mutated in TP53 and other genes (19.9%), 23 were mutated in non-TP53 genes (16.9%) and 86 (63.2%) in TP53 only. Clonal mutations comprised 69% of all mutations and were found in 113 tumors; 90% of these were in TP53.

Figure 2.

Mutation distribution in tumors and in the germline of TNBC patients. A: Mutation distribution per tumor, frequency of mutations per gene, and histogram of the number of mutations per tumor. Tumors had mean ± SD 1.1±1.2 mutations, affecting up to 7 different genes with the panel used. B: Genes mutated in at least 2 tumors are shown; percentages correspond to the number of tumors with a specific mutated gene among the 136 tumors that were found positive for at least one mutation among the 191 tumors with NGS informative genotypes. C: Incidence of germline mutation carriers among 194 TNBC patients; percentages correspond to the number of carriers among all patients in the cohort. Among carriers, 78% of the mutations were found in BRCA1 and 8% in BRCA2. Mutations in other genes (RAD51C, BARD1, RAD50, NBN, MRE11A and BRIP1) mostly presented as single genetic events.

As compared to tumors without mutations in the genes targeted with the tissue panel, tumors with mutations had larger size (P=0.028) and higher Ki67 labeling (P=0.025), while exhibiting a trend for higher histological grade (0.075) and the basal-like phenotype (P=0.074). TP53 mutations were significantly associated with the basal-like phenotype (P=0.039). PIK3CA mutations were associated with older age (P=0.006) and lower Ki67 labeling (P=0.009). Overall, the present tumor mutation patterns and their characteristics are in line with previous reports in basal-like tumors [9] and TNBC [10,11].

Germline mutations

We identified 50 germline loss-of-function mutations in eight cancer predisposing genes among 194 TNBC patients (Figure 2C). The observed prevalence (26%) appears higher than the 14.6% previously reported for TNBC patients [2] and should be attributed to the strong founder effects described for Greek patients. Germline mutations are described in detail in Table 3.

Table 3.

Germline mutations in the 50 carriers, along with germline/tumor (G/T) status

Gene	Mutation (DNA)	Mutation (protein)	Location	N carriers	N G/T conc	N G/T disc	N G/T pair not tested
BRCA1	c.68_69delAG	p.E23fsX17	Ex 2	1	1	0	0
BRCA1	c.181T>G	p.C61G	Ex 5	2	1	0	1
BRCA1	c.1299_1302dupCAGT	p.D435fsX2	Ex 11	2	1	1	0
BRCA1	c.1687C>T	p.Q563X	Ex 11	1	0	0	1
BRCA1	c.1953_1956delGAAA	p.K651fsX49	Ex 11	1	1	0	0
BRCA1	c.3375_3376delTC	p.P1125fsX6	Ex 11	1	0	1	0
BRCA1	c.3624delA	p.K1208fsX2	Ex 11	1	0	1	0
BRCA1	c.3700_3704delGTAAA	p.V1234fsX8	Ex 11	2	0	2	0
BRCA1	c.4065_4068delTCAA	p.N1355fsX10	Ex 11	1	1	0	0
BRCA1	c.4391_4393delCTAinsTT	p.P1464fsX2	Ex 14	1	1	0	0
BRCA1	c.5212G>A	p.G1738R	Ex 20	8	5	1	2
BRCA1	c.5251C>T	p.R1751X	Ex 20	3	0	0	3
BRCA1	c.5266dupC	p.Q1756fsX74	Ex 20	6	1	5	0
BRCA1	c.5431C>T	p.Q1811X	Ex 23	1	1	0	0
BRCA1	c.5406+664_*8273del11052	p.0	Ex 23, 24	3	2	0	1
BRCA1	c.5467G>A	p.G1803_A1823del21	Ex 23	1	1	0	0
BRCA1	c.5468-285_5592+4019del4429_insCACAG	p.0	Ex 24	3	0	0	3
BRCA1	c.5497G>A	p.V1833M	Ex 24	1	0	1	0
BRCA2	c.2339C>G	p.S780X	Ex 11	1	0	1	0
BRCA2	c.3847_3848delGT	p.Val1283fsX2	Ex 11	1	0	0	1
BRCA2	c.5946delT	p.S1982fsX22	Ex 11	1	0	1	0
BRCA2	c.9097dupA	p.T3033fsX11	Ex 23	1	1	0	0
RAD51C	c.706-2A>G	p.V236_A279del44	Int 4	2	2	0	0
BARD1	c.896_897insG	p.P300fsX2	Ex 4	1	0	0	1
BRIP1	c.2334_2337dupTGTC	p.I779fsX17	Ex 12	1	0	0	1
MRE11	c.1995-2A>G	p.?	Int 17	1	0	0	1
RAD50	c.326_329delCAGA	p.T109fsX2	Ex 3	1	0	0	1
NBN	c.265C>T	p.R89X	Ex 3	1	0	0	1

Notes: Ex: Exon; Int: Intron; N: Number; G/T conc: Germline mutation preserved in tumor; G/T disc: Germline mutation not present in tumor (highlighted); Pair not tested: Due to unavailable material; ?: Protein change not yet described.

Germline positive tumors were more frequent in premenopausal and in younger women (both P<0.001); Were of higher grade (P=0.015); And had higher TILs density (P=0.007). Importantly though, when distinguishing BRCA1 and non-BRCA1 carriers, the latter were found in older, menopausal patients, with higher nodal burden (Table 4), in accordance to previous reports [24]. Tumors from the 11 non-BRCA1 carriers appeared more frequently discordant upon phenotyping in local and central pathology laboratories (P=0.008), which is a novel finding needing validation in larger series.

Table 4.

Clinicopathological associations based on germline mutation status

194 patients informative for germline mutations
	Germline negative, N=144	Germline positive, N=50	P value
Menopausal status			<0.001*
Post	86 (59.31%)	13 (26.00%)
Pre	59 (40.69%)	37 (74.00%)
Histological grade			0.015*
I or II	26 (17.93%)	2 (4.00%)
III	119 (82.07%)	48 (96.00%)
Age (yrs)			<0.001^
N patients	144	50
Mean ± SD	53.7±12.5	44.6±10.3
Median	54.5	43.5
TILs density			0.007^
N patients	140	48
Mean ± SD	20.5±20.7	27.8±21.2
Median	10	24
50 patients with germline mutations
	BRCA1	Non-BRCA1
TNBC central			0.017*
NO	5 (12.82%)	5 (45.45%)
YES	34 (87.18%)	6 (54.55%)
Menopausal status			0.014*
Post	7 (17.95%)	6 (54.55%)
Pre	32 (82.05%)	5 (45.45%)
Positive nodes			<0.001*
≥4	4 (10.26%)	7 (63.64%)
0 to 3	35 (89.74%)	4 (36.36%)
Age (yrs)			0.020^
N patients	39	11
Mean ± SD	42.3±8.4	52.5±12.5
Median	42	56

^*Pearson’s chi sq;

^{^}Non-parametric.

Comparison between germline and tumor genotypes

The pattern of mutated genes in tumors was significantly different between germline carriers and non-carriers (Table 5). Due to the small number of tumors with mutations in specific genes, comparisons were only possible for TP53 and PIK3CA. Strikingly, PIK3CA mutations were exclusively present in tumors from non-carriers (P=0.007). The rate of TP53 mutations and of TP53-mutated tumors with or without clonal mutations (Table 5) did not differ between germline carriers and non-carriers. Although the overall distribution of TP53 domain mutations (Figure 3A) did not differ in tumors from germline carriers as compared to non-carriers, we observed a higher number of clonal mutations in the TP53 tetramerization domain (P=0.029). The distribution of mutations with respect to functional BRCA1 domains did not differ in tumors with and without TP53 mutations (Figure 3B).

Table 5.

Tumor mutation characteristics based on germline mutation status

191 patients with informative germline and tumor genotypes
	Germline negative	Germline positive	chr sq p (Pearson’s)
Mutations per gene$			<0.001
TP53	103	31
PIK3CA	20	0
CDH1	13	1
MAP3K1	4	1
AKT1	2	0
EGFR	2	0
GATA3	2	1
NCOR1	2	0
PTEN	2	2
BRCA1	1*	2**,^
CCND1	1	1
ESR1	1	0
MAP2K4	1	2
MET	1	0
NOTCH4	1	0
PALB2	1#	0
VEGFA	1	1
ARID1B	0	5
TBX3	0	3
PIK3CA mut status$$			0.007
PIK3CA wt	125 (86.80%)	47 (100%)
PIK3CA mut	19 (13.19%)	0 (0%)
TP53 mut status$$			0.337
TP53 wt	56 (38.89%)	22 (46.81%)
TP53 mut	88 (61.11%)	25 (53.19%)
TP53 clonal mut$$			0.364
TP53 wt	39 (27.08%)	16 (23.88%)
TP53 non clonal	31 (21.53%)	6 (12.77%)
TP53 clonal	74 (51.39%)	25 (53.19%)
31 patients tested for germline/tumor mutation status
	Discordant, N=12	Concordant, N=19
Mutations per gene			0.034
TP53	9	7
BRCA1	1	1
CCND1	0	1
GATA3	0	1
MAP2K4	0	1
MAP3K1	0	1
PTEN	0	1
TP53 mut status			0.038
TP53 wt	3 (25%)	12 (63.16%)
TP53 mut	9 (75%)	7 (36.84)
TP53 clonal mut			0.036
TP53 wt	3 (25%)	12 (63.16%)
TP53 non clonal	0 (0%)	1 (5.26%)
TP53 clonal	9 (75%)	6 (31.58%)

^$Number of mutations among genes is compared;

^$$Number of tumors is compared;

^*BRCA1 p.E33K present in the tumor only at variant allele frequency (VAF) of 14%;

^**BRCA1 p.Q1811ter in the tumor at VAF 94% instead of the p.Q1756fsX74 in the germline;

^{^}BRCA1 p.A1823T germline mutation preserved in the tumor at VAF 92%;

^#PALB2 p.E860K present in the tumor only at very low VAF (6%);

Wt: Wild-type; Non clonal indicates TP53 mutation with VAF <25%.

Figure 3.

Spectra of tumor TP53 and germline BRCA1 mutations, with respect to germline mutation preservation or loss in tumors. A and B: Lolliplots show all detected mutations for TP53 in tumors and for BRCA1 in the germline. The colors are used to show preservation of the germline mutation in the tumor (green); Loss of the germline mutation in the tumor (red); And, presence of TP53 mutations in tumors from non-carriers (grey). Orange dots correspond to germline mutations that could not be tested in the matched tumors. The majority of TP53 mutations were found mostly in the DNA binding domain (65.7%) and BRCA1 mutations in the BRCA1 C-terminal domain (64.1%). C: Preservation and loss of the mutant germline allele in tumors is shown in four example cases. Two germline mutations differently preserved in tumors have been selected. The same germline mutation in individual carriers may be preserved (red letters for tumor genotypes) or not preserved (black: Wild-type) in matched tumors. Tumor TP53 mutation status is also shown. Note that tumors with preserved germline mutations had a more pronounced stromal component and appeared less anaplastic compared to those with discordant germline/tumor status.

Unexpectedly, in the 33 patients with evaluable germline mutation status in the tumor (Figure 1), germline loss-of-function mutations were preserved in 19 (concordant germline/tumor status [GTS], 58%) but were absent in 14 (discordant GTS; 42%) tumors. Sample mismatch was excluded for these samples based on MS-ID testing (Table 2). Technical issues that might have caused mutation artifacts, as described when shearing DNA for capture NGS [25] did not interfere with the results, since the majority of germline mutations were detected by dd-sequencing (Figure 1), including 10/14 of discordant GTS mutations. With respect to FFPE, dd-sequencing yielded acceptable results for the wild-type region for the same mutation and for the same type of mutation (Figure 3C); Hence, inability to detect the expected mutations could not be attributed to technical underperformance of these samples.

The status of normal tissue matched that of the corresponding tumor: In the case with concordant GTS, the germline mutation was also present in the available normal sample; In 6 discordant cases the germline mutation was absent in both the tumor and matched normal tissue. GTS was not BRCA1/2 mutation type specific (Figure 3B). Of note, the MS-ID results in Table 2 also indicated the absence of LOH in 13/14 matched samples with germline mutation loss in the tumor. In one case with discordant GTS, the germline BRCA1 p.Q1756fsX74 (c.5266dupC) was lost in the tumor, where a somatic BRCA1 mutation, p.Q1811* was present (Figure 3C); In this case, loss of the germline mutant allele was demonstrated upon MS-ID analysis (Table 2) and the new mutant allele prevailed with 94% VAF in the tumor. Although not statistically significant, tumors in patients with concordant GTS more frequently exhibited heterogeneous differentiation with basophilic extracellular matrix production, as compared to those with discordant status, which were highly anaplastic and solid (Figure 2C).

With the tissue panel, mutations were present in 9/12 tumors with discordant and in 11/19 with concordant GTS. All mutant discordant tumors (9/9) carried clonal TP53 mutations. In comparison, TP53 mutations were less frequent and clonal in concordant tumors (P=0.038 and P=0.036, respectively) (Table 5). No further difference was observed between concordant and discordant GTS cases with respect to tumor mutation and clinicopathological parameters.

Tumor TP53 mutations interacted with germline mutation status affecting patient outcome

Follow-up and disease-free survival (DFS) data are shown in Table 1. Germline status per se, or distinguished in BRCA1 and non-BRCA1 carriers, was not related to patient outcome (Table 6). Germline status interacted with tumor TP53 mutations with regards to DFS. Non-carriers with or without TP53 mutations had similar DFS (Figure 4A). In comparison, in carriers, relapses were noticed in 9/25 (36%) patients with TP53 mutated tumors but in only 2/22 (9%) patients with TP53 wild-type tumors (Figure 4B). This interaction was statistically significant (P=0.0266) (Table 6).

Table 6.

Univariate Cox analysis results for DFS

Variable	Total N	N patients	N events	HR	95% CI low	95% CI high	P value
Age ≥50 vs. <50)	194	105 vs. 89	27 vs. 27	0.777	0.456	1.326	0.355
Menopausal (post vs. pre)	194	99 vs. 95	24 vs. 30	0.740	0.432	1.265	0.271
Tumor size (>2 vs. ≤2)	192	116 vs. 76	37 vs. 16	1.833	1.018	3.298	0.043
Number of positive nodes (≥4 vs. 0-3)	190	44 vs. 146	21 vs. 31	2.735	1.570	4.764	<0.001
Adjuvant Radiotherapy (yes vs. no)	194	140 vs. 54	35 vs. 19	0.657	0.376	1.150	0.141
TILs (continuous, 10% increments)	188			0.805	0.678	0.956	0.013
Germline (positive vs. negative)	194	50 vs. 144	11 vs. 43	0.802	0.413	1.557	0.515
Tumor mutation (any vs. none)	191	136 vs. 55	43 vs. 11	1.722	0.888	3.340	0.108
TP53 (MUT vs. nonMUT)	191	113 vs. 78	33 vs. 21	1.133	0.655	1.958	0.656
Germilne* TP53 MUT status#	191						0.027^
Germline positive, TP53 MUT vs. TP53 WT	47	25 vs. 22	9 vs. 2	4.121	0.882	19.250
Germline negative, TP53 MUT vs. TP53 WT	144	88 vs. 56	24 vs. 19	0.827	0.453	1.510

Notes: HR: Hazard ratio;

^#Interaction between germline status and tumor TP53 mutations;

^{^}Likelihood ratio test for interaction;

^*Interaction (statistical);

WT: Wild-type; Significant results are shown in bold.

Figure 4.

Tumor TP53 mutations diversely affect DFS according to germline mutation status. A: Non-carriers; B: Carriers.

Upon multivariate analysis involving patients with informative data for this interaction and adjustments for menopausal status, tumor size, nodal status, radiotherapy and TILs at 10% increments, germline mutations and tumor TP53 mutations were not independently significant, but the above interaction was retained with marginal significance in the model (Table 7). Classic parameters independently predicting for favorable DFS were low nodal burden, higher TILs and adjuvant radiotherapy.

Table 7.

Multivariate analysis results for DFS

Multivariate Model (all pts: 180)	N patients	N events	HR	95% CI low	95% CI high	p-value
Germline (positive vs. negative)	45 vs. 135	11 vs. 37	0.514	0.115	2.301	0.384
TP53 (MUT vs. nonMUT)	106 vs. 74	29 vs. 19	0.821	0.426	1.581	0.555
Germline* TP53 MUT status#						0.053^
Germline positive; TP53 MUT vs. WT	24 vs. 21	9 vs. 2	3.34	0.615	18.125
Germline negative; TP53 MUT vs. WT	82 vs. 53	20 vs. 17	0.847	0.439	1.636
Menopausal status (post vs. pre)	89 vs. 91	20 vs. 28	0.694	0.379	1.272	0.238
Tumor size (>2 vs. ≤2)	107 vs. 73	34 vs. 14	1.465	0.756	2.839	0.257
Number of positive nodes (≥4 vs. 0-3)	41 vs. 139	20 vs. 28	3.78	1.999	7.145	<0.001
Adjuvant Radiotherapy (yes vs. no)	134 vs. 46	31 vs. 17	0.486	0.257	0.918	0.026
TILs (continuous, 10% increments)	n.a.	n.a.	0.769	0.624	0.948	0.014

^#Interaction between germline status and tumor TP53 mutations;

^{^}Likelihood ratio test for interaction;

^*Interaction (statistical);

HR: Hazard ratio; N.a.: Not applicable for continuous variables; Significant results are shown in bold.

The group of 14 patients with discordant GTS experienced more relapses as compared to the 19 patients with concordant GTS (29% vs. 21%). Due to the small subgroup sample size, however, this observation did not reach statistical significance; for the same reason, further subgroup analyses with respect to TP53 mutations were not possible.

Discussion

This study provides novel aspects on the impact of germline mutations on tumor biology and on the outcome of TNBC patients treated with adjuvant chemotherapy. We show (A) that the germline mutation may be lost in adjacent non-cancerous and tumor tissues from BRCA1/2 carriers, and (B) that tumor genotypes, particularly tumor TP53 mutations, seem to interact with germline status for patient outcome.

We observed loss of the pathogenic germline mutation in roughly 40% of matched peripheral blood and tumor samples. This contradicts the classical perception of how tumors develop in individuals with mutations in tumor suppressor genes, for example BRCA1 [5]. As described, this observation could not be attributed to sample mismatch and further technical issues. Although still a taboo in genetics, the usually overlooked loss of the pathogenic germline mutation in tissues and tumors has been reported for known cancer predisposition genes, e.g., in breast for BRCA1/2 [6,8] and ATM [26]; In endometrial carcinomas for MLH1 [27]; In a murine pancreatic carcinoma model for Brca2 [28]. Of note, the observation here concerns the loss of the germline mutation and not necessarily of the entire mutated BRCA1/2 allele in tumors. Based on recent knowledge, cells may not opt for the germline mutation in an attempt to counteract haploinsufficiency, which, as shown for BRCA1, results in genomic instability and premature senescence in human breast epithelial cells [29]. In a different context but again in order to avoid excessive genomic instability induced by synthetic lethality and platinum treatment, cells and tumors may become resistant to such treatments by losing the inherited mutant BRCA2 mutation and by developing a more survival-compatible mutation, a process called mutation reversion [30]. Although still unproven, it may well be that life-long haploinsufficiency occasionally results in the above described reversion of the inherited mutation in tumor related tissues for conferring survival advantage. Our discordant germline/tumor data are consistent with this model.

The above model of germline mutation loss as an escape from the anti-survival effects of haploinsufficiency in normal breast cells has been linked to BRCA1 mutations in the RING and DNA-binding domain [29]. Our carriers with discordant germline/tumor status had such mutations but they also had mutations in the BRCT domain, including the most common BRCA1 c.5266dupC [31]. Notably, opting for germline mutation loss does not necessarily mean that cells opt for BRCA1/2 proficiency. As we show in one case with LOH of the germline BRCA1 c.5266dupC allele, this truncating germline frameshift mutation was replaced in the tumor by a nonsense somatic BRCA1 substitution in the same functional domain (p.Q1811*). This might possibly have taken place because this type of mutation is less energy demanding [32] for the establishment of deficient BRCA1 function.

In the present study we used data from a larger project that was not specifically designed for interrogating the status of BRCA1/2 and additional genes that were targeted with germline panels at the tissue level. Because the study is retrospective, there was no additional tissue material available for targeted testing of the detected germline mutation in the tumors of all 50 identified carriers nor to examine the entire coding BRCA1/2 regions in tumor tissues; hence, at present we do not know whether additional, potentially multiple BRCA1/2 alterations were present in the examined tumors. For the same reason, additional normal (e.g., buccal) DNA and full pedigree information could not be obtained from the above patients; Hence, we do not know whether germline mutation loss concerned the tumor and tissue-of-origin only or other tissues as well. This would be important since rare cases of BRCA1 mosaicism that may be established during embryonic development cannot be completely excluded [33]. Despite the above shortcomings, our data clearly show that BRCA1/2 germline mutations may be lost in tumors at far higher rates than previously thought. The impact of germline mutation loss on tumor biology and clinical behaviour, which could not be demonstrated in the present small patient subgroups, needs to be addressed in larger studies.

With respect to tumor TP53 mutations in the present series, the incidence of such mutations did not differ between BRCA1-related germline carriers and non-carriers. This finding is not directly comparable to studies reporting on associations between TP53 mutations and tumor BRCA-ness, assessed with different methods [34,35]. Although we haven’t tested forBRCA-ness, we observed a higher incidence of clonal TP53 mutations in the TP53 tetramerization domain among tumors from germline carriers, which is in accordance to old reports where TP53 mutations preferably occur outside the exon 5-8 hotspots in BRCA1/2 carriers [12,36]. This implies different functional impact of TP53 mutations in the presence or absence of BRCA1-related germline mutations in TNBC, which is further supported by our finding on a higher prevalence of TP53 mutations in tumors with discordant, as compared to those with concordant BRCA1/2 germline/tumor mutation status. Clearly, cause-and-effect for this novel report cannot be demonstrated here and the finding needs functional validation.

A functional interference between tumor TP53 mutations and BRCA1-related germline status is also suggested by the herein presented prognostic interaction between these two features on TNBC patient outcome upon standardanthracyclines-taxanes-based adjuvant chemotherapy. Demonstrating a clear prognostic or predictive impact of TP53 mutations in breast cancer still poses one of the greatest challenges in Oncology [37]. The present finding may be one of the first reports within the TNBC subtype; It seems that standard adjuvant treatment benefits germline carriers with TP53 wild-type tumors, while germline carriers with TP53 mutated tumors may need a different type of treatment. Although the number of our patient subgroups was small and our findings should be considered as hypothesis generating needing validation in larger studies, the adverse prognostic role of TP53 mutations in germline carriers is worth considering, since this effect was completely absent in non-carriers.

In tumors from carriers, we observed absence of PIK3CA mutations, as previously published [35], and absence of AKT1, EGFR and MET mutations. Such patients constituted approximately 17% of non-carriers in the present series; whether they would benefit from anti-PI3K treatments needs to be investigated.

Life-long haploinsufficiency in BRCA1 carriers appears to confer systemic effects including the immune system, e.g., reduced BRCA1 expression in blood leykocytes [38] and impaired immune function [39]. These effects, in combination with the genomic instability induced by BRCA-deficiency, may explain the herein observed increased rates of tumor infiltrating lymphocytes in TNBC carriers and supports the proposed immunotherapeutic interventions for this disease [40].

In conclusion, we have shown that germline mutation loss in tumors is not a rare phenomenon in TNBC and we have discussed that this condition may develop in the context of haploinsufficiency with a mechanism previously proposed for the acquisition of resistance to synthetic lethality drugs. We have also shown that preservation or loss of the germline mutation in tumors is linked to somatic TP53 mutations and that the latter interact with germline mutation status for patient outcome. Overall, these findings prompt for combined germline/tumor genotyping for the classification of TNBC, particularly in the context of clinical trials. The direct implication of this finding on genetic association studies is that LOH, as currently tested, cannot predict cause-and-effect for germline mutations in a particular gene. Pathogenic germline mutation loss in tumors may appear as dogma-challenging but it certainly merits further investigation for its impact on tumor biology and clinical behaviour.

Disclosure of conflict of interest

None.