Clinicopathologic and Genomic Landscape of Breast Carcinoma Brain Metastases

Abstract

Background

Among patients with breast carcinoma who have metastatic disease, 15%–30% will eventually develop brain metastases. We examined the genomic landscape of a large cohort of patients with breast carcinoma brain metastases (BCBMs) and compared it with a cohort of patients with primary breast carcinomas (BCs).

Material and Methods

We retrospectively analyzed 733 BCBMs tested with comprehensive genomic profiling (CGP) and compared them with 10,772 primary breast carcinomas (not‐paired) specimens. For a subset of 16 triple‐negative breast carcinoma (TNBC)–brain metastasis samples, programmed death‐ligand 1 (PD‐L1) immunohistochemistry (IHC) was performed concurrently.

Results

A total of 733 consecutive BCBMs were analyzed. Compared with primary BCs, BCBMs were enriched for genomic alterations in TP53 (72.0%, 528/733), ERBB2 (25.6%, 188/733), RAD21 (14.1%, 103/733), NF1 (9.0%, 66/733), BRCA1 (7.8%, 57/733), and ESR1 (6.3%,46/733) (p < .05 for all comparisons). Immune checkpoint inhibitor biomarkers such as high tumor mutational burden (TMB‐high; 16.2%, 119/733); high microsatellite instability (1.9%, 14/733); CD274 amplification (3.6%, 27/733); and apolipoprotein B mRNA editing enzyme, catalytic polypeptide‐like mutational signature (5.9%, 43/733) were significantly higher in the BCBM cohort compared with the primary BC cohort (p < .05 for all comparisons). When using both CGP and PD‐L1 IHC, 37.5% (6/16) of patients with TNBC brain metastasis were eligible for atezolizumab based on PD‐L1 IHC, and 18.8% (3/16) were eligible for pembrolizumab based on TMB‐high status.

Conclusion

We found a high prevalence of clinically relevant genomic alterations in patients with BCBM, suggesting that tissue acquisition (surgery) and/or cerebrospinal fluid for CGP in addition to CGP of the primary tumor may be clinically warranted.

Implications for Practice

This study found a high prevalence of clinically relevant genomic alterations in patients with breast carcinoma brain metastasis (BCBM), suggesting that tissue acquisition (surgery) and/or cerebrospinal fluid for comprehensive genomic profiling (CGP) in addition to CGP of the primary tumor may be clinically warranted. In addition, this study identified higher positive rates for FDA‐approved immunotherapy biomarkers detected by CGP in patients with BCBM, opening a possibility of new on‐label treatments. Last, this study noted limited correlation between tumor mutational burden and PD‐L1 immunohistochemistry (IHC), which shows the importance of testing patients with triple‐negative BCBM for immune checkpoint inhibitor eligibility with both PD‐L1 IHC and CGP.

Short abstract

This article examines the genomic landscape of breast cancer brain metastasis samples with comprehensive genomic profiling in a large cohort of patients to define the potential applicability of recent therapeutic advances.

Introduction

Breast cancer remains a leading cause of morbidity and mortality for women globally. Despite therapeutic advances, 30% of women with early disease will relapse with incurable metastatic breast cancer. A growing clinical problem in patients with metastatic disease is the development of brain metastasis, and breast carcinoma brain metastasis (BCBM) occurs in 15%–30% of patients [1]. In particular, triple‐negative breast carcinoma (TNBC) and human epidermal growth factor receptor 2 (HER2)–positive breast cancer have a propensity for metastasizing to the central nervous system [2, 3].

Sequential advances in targeted systemic therapies for metastatic breast cancer have largely been predicated on assessing control of extracranial disease, with predictive biomarkers typically assessed on primary tumor or an extracranial metastasis. Recent advances include the U.S. Food and Drug Administration (FDA) approval of alpelisib plus fulvestrant for PIK3CA‐mutated disease in estrogen receptor–positive, HER2‐negative breast cancer based on the SOLAR‐1 clinical trial [4]. Although olaparib and talazoparib are now available for patients with metastatic breast cancer and germline BRCA1/2 mutation, the TBCRC‐048 trial also demonstrated benefit from olaparib in somatic BRCA1/2‐mutant breast cancer [5, 6, 7].

With regard to immunotherapy, the FDA has recently approved two immune checkpoint inhibitors (ICPIs) for patients with metastatic TNBC [8, 9]. First, atezolizumab plus nab‐paclitaxel was approved by the FDA in 2019 for TNBC with programmed death‐ligand 1 (PD‐L1) positivity (immune cell score ≥ 1) as defined by the Ventana PD‐L1 SP142 immunohistochemistry (IHC) assay (Ventana Medical Systems, Oro Valley, AZ) [10]. The second ICPI, pembrolizumab, was approved for all solid tumor types in patients with high tumor mutational burden or high microsatellite instability [11, 12]. In addition, based on the KEYNOTE‐355 trial, patients with TNBC with PD‐L1 positivity using the Dako 22C3 IHC assay (Dako, Carpinteria, CA) with combined positive score ≥ 10 may benefit from pembrolizumab in combination with chemotherapy [13]. Furthermore, the apolipoprotein B mRNA editing enzyme, catalytic polypeptide‐like (APOBEC) mutational signature has been associated with ICPI response in tumor types such as non‐small cell lung cancer and breast carcinoma [14, 15].

In contrast to these advances for the control of extracranial disease, only tucatinib has been specifically licensed for the treatment of BCBM and only in the context of HER2‐positive disease, as patients with active BCBM have been excluded from the vast majority of registration trials in breast cancer [16]. However, limited evidence of potential central nervous system activity of targeted agents, including alpelisib and olaparib, is available from case reports [17, 18]. Additionally, in the NALA trial, neratinib in combination with capecitabine showed good efficacy in patients with HER2‐positive BCBM [19]. Studies have previously examined the genomic landscape of BCBM and reported genomically distinct features from primary breast cancer as well as extracranial metastasis [20]. Given the limited systemic options for the treatment of BCBM, identifying the prevalence of biomarkers that may select for patients who may benefit from targeted therapies warrants further study. Here, we examine the genomic landscape of BCBM samples with comprehensive genomic profiling (CGP) in a large cohort of patients to define the potential applicability of recent therapeutic advances.

Materials and Methods

Patient Cohort

This study was approved by the Western Institutional Review Board under protocol no. 20152817. We performed a retrospective analysis of samples from 733 consecutive patients with breast carcinoma that metastasized to the brain and were tested with FoundationOne/FoundationOne CDx (Foundation Medicine, Inc., Cambridge, MA) between August 2014 and June 2020 as part of routine clinical care. In addition, 10,772 primary breast carcinoma specimens were used as a comparison group. Age, sex, and site of specimen of patient were extracted from accompanying pathology reports. TNBC status was confirmed in a subset of patients who received PD‐L1 IHC testing by reviewing the accompanying pathology report for hormone receptor status and ERBB2 (HER2) status from CGP.

Comprehensive Genomic Profiling of Breast Carcinoma Samples

CGP was performed using FoundationOne/FoundationOne CDx in a Clinical Laboratory Improvement Amendments (CLIA)–certified and College of American Pathologists (CAP)–accredited laboratory (Foundation Medicine, Cambridge, MA) using previously described methods [21]. FoundationOne/FoundationOne CDx uses a next‐generation sequencing platform and a hybrid capture methodology that detects base substitutions, insertions and deletions, and copy number alterations in up to 324 genes and select gene rearrangements, as well as tumor mutational burden (TMB) and microsatellite instability (MSI). H&E‐stained slides from each sample were reviewed by a board‐certified pathologist for presence of adequate tumor (≥20% of nucleated cells are tumor cells) before sequencing. ERBB2 amplification was determined by CGP and defined as greater than or equal to ploidy +3 (copy number 5 in a diploid sample) in accordance with the FDA‐approved companion diagnostics (CDx) claim for the FoundationOne CDx assay. TMB was determined on up to 1.14 megabases of sequenced DNA, and TMB ≥10 mutations per megabase was considered high TMB (TMB‐high) per the CDx approval [12, 22]. MSI analysis was performed from DNA sequencing across 114 loci, and high MSI (MSI‐high) was considered positive [11, 23]. APOBEC mutational signatures were called as described by Zehir et al. [24]. Genomic ancestry of patients was determined using a principal component analysis of genomic single nucleotide polymorphisms trained on data from the 1000 Genomes Project, and each patient was classified as belonging to one of the following super‐populations: African, Central and South American, East Asian, European, and South Asian [25, 26]. Somatic/germline status for BRCA1/2 short variant mutations was computationally predicted using previously described methods [27].

Ventana PD‐L1 SP142 Immunohistochemistry Testing

For a subset of cases, PD‐L1 IHC was performed using the Ventana PD‐L1 SP142 assay per manufacturer's instructions in a CLIA‐certified and CAP‐accredited reference laboratory (Foundation Medicine, Morrisville, NC) [28]. PD‐L1 IHC slides were interpreted by board‐certified pathologists using the tumor‐infiltrating immune cell (IC) scoring method, where the IC score is the proportion of tumor area that is occupied by PD‐L1 staining ICs of any intensity per the interpretation guide [29]. The CDx cutoff for atezolizumab plus nab‐paclitaxel for TNBC is an IC score of ≥1% [8].

Statistical Analysis

Clinicopathologic differences between different breast cancer cohorts were analyzed using analysis of variance, chi‐squared contingency test, or Fisher's exact test. To examine the differences in the genomic landscape of the different cohorts, we identified the top 25 genes that have genomic alterations (GAs) and compared these same genes with Fisher's exact test. Values of p were adjusted for multiple comparisons using the Bonferroni method, and adjusted values of p < .05 were considered significant [30].

Results

Cohort of Patients with Breast Carcinoma Brain Metastasis

A total of 733 BCBM samples were included in this study. Median patient age was significantly higher in the primary breast carcinoma (BC) cohort compared with the BCBM cohort (55 and 53, respectively, p < .001; Table 1). Predominant genetic ancestry of patients was not significantly different between brain metastasis and primary breast carcinoma samples (p = .22; Table 1).

Table 1.

Characteristics of patients with primary breast carcinoma and breast carcinoma brain metastases

Patient characteristics	Primary breast carcinoma (n = 10,772)	Breast carcinoma brain metastases (n = 733)	p valuea
Age, yearsb			<.001
Median	55	53
Mean	55	53
Predominant ancestry,c n (%)			.22
African	1,564 (14.8)	94 (13.0)
Central and South American	1,250 (11.9)	103 (14.3)
East Asian	442 (4.2)	27 (3.7)
European	7,094 (67.3)	481 (66.7)
South Asian	192 (1.8)	16 (2.2)
Immunotherapy biomarkers, n (%)
TMB‐high (≥10 mutations/Mb)	586 (5.4)	119 (16.2)	<.001
MSI‐high	42 (0.4)	14 (1.9)	<.001
CD274 amplification	186 (1.7)	27 (3.6)	<.001
APOBEC mutational signature	384 (3.6)	43 (5.9)	.003

^a All p values from Fisher's exact test except for age, which is from analysis of variance, and predominant ancestry, which is from chi‐squared contingency test.

^b Age was not available for three breast carcinoma brain metastases and 28 primary breast carcinoma samples.

^c Predominant genetic ancestry could not be determined for a few cases (n = 721 for total breast carcinoma brain metastases and n = 10,542 for total primary breast carcinoma).

Abbreviations: APOBEC, apolipoprotein B mRNA editing enzyme, catalytic polypeptide‐like; Mb, megabase; MSI‐high, high microsatellite instability; TMB‐high, high tumor mutational burden.

In a subset of samples with PD‐L1 IHC testing (n = 188), 16 TNBC brain metastasis (BM) samples and 172 primary TNBC samples had CGP performed concurrently with the PD‐L1 IHC testing. In this cohort, age and predominant genetic ancestry were not significantly different between the TNBC‐BM and primary TNBC disease subsets (p = .45 and p = .81, respectively; Table 2).

Table 2.

Characteristics of patients with primary TNBC and TNBC brain metastases

Patient characteristics	Primary TNBC (n = 172)	TNBC brain metastases (n = 16)	p value
Age, years			.454a
Median	54	58.5
Mean	54.1	56.9
Predominant ancestry, n (%)			.805b
African	43 (25.0)	3 (18.8)
Central and South American	21 (12.2)	3 (18.8)
East Asian	5 (2.9)	0 (0)
European	99 (57.6)	10 (62.5)
South Asian	4 (2.3)	0 (0)
PD‐L1 positivity rate (IC ≥1), n (%)	110 (64.0)	6 (37.5)	.057c
Number of GAs per sample			<.001a
Median	5	8
Mean	5.7	8.3
Immunotherapy biomarkers, n (%)
TMB‐high (≥10 mutations/Mb)	4 (2.3)	3 (18.8)	.014c
MSI‐high	0 (0)	0 (0)	1.000c
CD274 amplification	5 (2.9)	3 (18.8)	.022c

^aFisher's exact test.

^b Chi‐squared contingency test.

^c Analysis of variance.

Abbreviations: GA, genomic alteration; Mb, megabase; MSI‐high, high microsatellite instability; PD‐L1, programmed death‐ligand 1; TMB‐high, high tumor mutational burden; TNBC, triple‐negative breast carcinoma.

Genomic Landscape of Breast Carcinoma Brain Metastases

CGP analysis of 733 BCBMs revealed the following top 10 most altered genes: TP53 (72.0%, 528/733), PIK3CA (28.7%, 218/733), ERBB2 (25.6%, 188/733), MYC (25.5%, 187/733), PTEN (16.8%, 123/733), CCND1 (14.9%, 109/733), RAD21 (14.1%, 103/733), FGF3 (13.5%, 99/733), FGF19 (13.5%, 99/733), and FGF4 (13.4%, 98/733) (Fig. 1A). BCBMs were enriched for GAs in TP53 (72.0% [528/733] vs. 59.7% [6,405/10,772], p < .001), ERBB2 (25.6% [188/733] vs. 11.8% [1,268/10,772], p < .001), RAD21 (14.1% [103/733] vs. 10.3% [1,108/10,772], p = .046), NF1 (9.0% [66/733] vs. 5.7% [615/10,772], p = .016), BRCA1 (7.8% [57/733] vs. 4.5% [486/10,772], p = .005), and ESR1 (6.5% [46/733] vs. 3.7% [397/10,772], p = .024) compared with the primary BC cohort (Fig. 1B; supplemental online Table 1). In addition, we identified 2.2% (16/733) of patients with BCBM with a fusion, including three patients with a NOTCH1‐SEC16A fusion and one patient with a RFX6‐ROS1 fusion (supplemental online Table 2).

Figure 1.

Genomic landscape of breast carcinoma brain metastases (BCBMs). (A): Co‐mutation plot of the 733 BCBM samples. The top five genes with genomic alterations (GAs) were TP53, PIK3CA, ERBB2, MYC, and PTEN. (B): Long‐tail plot of genes with GAs and comparison between the BCBM and primary breast carcinoma (BC) cohorts. The genomics of the BCBM samples was different from that of the primary BC samples. Enrichment of GAs in TP53, ERBB2, RAD21, NF1, BRCA1, and ESR1 was present in the BCBM cohort when compared with the primary BC cohort (p < .05).

In this BCBM cohort, 51.7% (379/733) of the patients were positive for at least one CDx biomarker as determined by CGP (Fig. 2). In terms of specific FDA‐approved therapies with an associated biomarker, 24.3% (178/733) had ERBB2 amplifications, 26.7% (196/733) had PIK3CA mutations, 0% (0/733) had NTRK1/2/3 fusions, and 11.9% (87/733) had BRCA1/2 mutations. Based on the somatic/germline status bioinformatic predictions, 28.6% (22/87) were germline mutations, 27.3% (21/87) were somatic mutations, and the algorithm was not able to make a prediction in 57.1% (44/87) of the mutations. In the BCBM cohort, 26.2% (192/733) had at least one GA in one of the 14 genes involved in the homologous recombination deficiency pathway (7.8% [57/733] BRCA1, 5.7% [42/733] BRCA2, and 3.1% [23/733] ATM; supplemental online Table 3). ESR1 mutations almost always occur following endocrine therapy in patients with estrogen receptor–positive breast cancer and have been shown to be a biomarker of endocrine resistance [31]. In this cohort of BCBM, 4.9% (36/733) of the patients harbored ESR1 mutation(s), which could help inform decision‐making for these patients with BCBM as well as highlighting the importance of including such patients in any relevant clinical trials (supplemental online Table 4).

Figure 2.

Eligibility of patients with breast carcinoma brain metastasis (BCBM) for therapy based on comprehensive genomic profiling (CGP) companion diagnostics data set. In this BCBM cohort, more than half of the patients (51.7%, 379/733) were positive for at least one companion diagnostic biomarker as determined by CGP. Abbreviations: amp, amplification; mut, mutation; TMB‐H, high tumor mutational burden.

Table 3.

Genomics of 11 paired primary breast carcinoma and breast carcinoma brain metastasis samples

Case	Days b/w specimen collections	Pt age at pBC	Ancestry	Histologic subtype	HR/HER2 status	Primary breast carcinomaa	BCBMa	Addition of genomic findings	Loss of genomic findings
1	720	56	European	Invasive ductal carcinoma	HR+/HER2−	MYC amp, MDM2 amp, ZNF703 amp, ESR1 D538G, TMB‐high	MYC amp, MDM2 amp, ZNF703 amp, FRS2 amp, ARFRP1 amp, PIK3CA E545K, ESR1 D538G	PIK3CA E545K, FRS2 amp, ARFRP1 amp	TMB‐high
2	2,937	58	European	Invasive ductal carcinoma	HR−/HER2− (TNBC)	EGFR amp, TP53 R342fs*3, PIK3CA V105_R108del	EGFR amp, EMSY amp, FGF19 amp, CCND1 amp, FGF3 amp, FGF4 amp, TP53 R342fs*3, IKZF1 G141*, PIK3CA V105_R108del	EMSY amp, FGF19 amp, CCND1 amp, FGF3 amp, FGF4 amp, IKZF1 G141*	none
3	343	33	East Asian	Invasive ductal carcinoma	HR(unk)/ HER2−	FGFR2 amp, MYC amp, TP53 A159P, PIK3R1 N673fs*19, STK11 V337fs*22	FGFR2 amp, TP53 A159P, CDH1 duplication, STK11 V337fs*22	CDH1 duplication	MYC amp, PIK3R1 N673fs*19
4	608	41	European	Invasive ductal carcinoma	HR+/HER2−	CCND1 amp, FGFR2 amp, PIK3CA H1047R, TP53 R280K	FGFR2 amp, CCND1 amp, EMSY amp, PIK3CA H1047R, TP53 R280K	EMSY amp	none
5	172	31	European	Invasive ductal carcinoma	HR−/HER2− (TNBC)	REL amp, PDGFRA amp, KIT amp, KDR amp, TP53 M237I	REL amp, KIT amp, KDR amp, TP53 M237I	none	PDGFRA amp
6	639	27	European	Invasive ductal carcinoma	HR−/HER2+	ERBB2 amp, MYC amp, TP53 R175H	ERBB2 amp, MYC amp, TP53 R175H, RB1 splice site 1390‐16_1421 + 29del77	RB1 splice site 1390‐16_1421 + 29del77	none
7	610	65	European	Invasive lobular carcinoma	HR(unk)/HER2+	ERBB2 amp, MLH1 del, TP53 R267G, ERBB2 V777L, CDK12 truncation, BCORL1 P1681fs*20, CDH1 N315fs*6, CHD4 Q1596*, NOTCH3 N1961fs*5	ERBB2 amp, MLH1 del, ERBB2 V777L, TP53 R267G, ABL1 P310fs*9, CDK12 truncation, NOTCH3 N1961fs*5, MSH3 K383fs*32, BCORL1 P1681fs*20, CDH1 N315fs*6, SMO P694fs*82, TMB‐high, MSI‐H	ABL1 P310fs*9, MSH3 K383fs*32, SMO P694fs*82, TMB‐high, MSI‐H	CHD4 Q1596*
8	1,008	61	African	Invasive ductal carcinoma	HR−/HER2+	TOP2A amp, ERBB2 amp, SMAD2 del, PIK3CA G1049R, TBX3 E275fs*7, TP53 D281fs*24	ERBB2 amp, SMAD2 del, ZNF703 amp, FGFR1 amp, TOP2A amp, PIK3CA G1049R, TBX3 E275fs*7, TP53 D281fs*24	ZNF703 amp, FGFR1 amp	none
9	618	39	African	Invasive ductal carcinoma	HR−/HER2− (TNBC)	MYC amp, LYN amp, MYST3 amp, BCL2L2 amp, TP53 E204*, EP300 truncation	LYN amp, MYST3 amp, MYC amp, BCL2L2 amp, NCOR1 G150R, TP53 E204*	NCOR1 G150R	EP300 truncation
10	784	51	European	Breast carcinoma (nos)	HR+/HER2−	ZNF217 amp, GNAS amp, AURKA amp, TP53 R196Q, XRCC2 R91Q, PIK3CA E545K, ARAF E568*, NF1 splice site 4577 + 1G > A, IRF2 E30*, GATA3 S408*, APOBEC mutational signature, TMB‐high	AURKA amp, ZNF217 amp, GNAS amp, PIK3CA E726K, PIK3CA E545K, XRCC2 R91Q, NF1 S1954*, CDKN1B S138*, APOBEC mutational signature	PIK3CA E726K, NF1 S1954*, CDKN1B S138*	TP53 R196Q, ARAF E568*, NF1 splice site 4577 + 1G > A, IRF2 E30*, GATA3 S408*, TMB‐high
11	473	41	Central and South American	Breast carcinoma (nos)	HR−/HER2− (TNBC)	EGFR amp, PRKCI amp, TERC amp, FGF12 amp, TP53 P300fs*6, TMB‐high	RAD21 amp, SOX2 amp, MYC amp, CCNE1 amp, MYCL1 amp, RPTOR amp, CD274 amp, PRKCI amp, EGFR amp, VEGFA amp, PIK3CA amp, TERC amp, PDCD1LG2 amp, FGF12 amp, NOTCH3 amp, JAK2 amp, CCND3 amp, MYB‐AHI1 rearrangement, TP53 P300fs*6	RAD21 amp, SOX2 amp, MYC amp, CCNE1 amp, MYCL1 amp, RPTOR amp, CD274 amp, VEGFA amp, PIK3CA amp, PDCD1LG2 amp, NOTCH3 amp, JAK2 amp, CCND3 amp, MYB‐AHI1 rearrangement	TMB‐high

^a Immune checkpoint inhibitor biomarkers of TMB‐high, MSI‐H, CD274 amplification, and APOBEC mutational signature are also included when present.

Abbreviations: amp, amplification; APOBEC, apolipoprotein B mRNA editing enzyme, catalytic polypeptide‐like; b/w, between; BCBM, breast carcinoma brain metastasis; del, deletion; HER2, human epidermal growth factor receptor; HER2−, HER2 negative; HER2+, HER2 positive; HR, hormone receptor; HR−, HR negative; HR+, HR positive; HR(unk), unknown HR status; MSI‐H, high microsatellite instability; nos, not otherwise specified; pt, patient; pBC, primary breast carcinoma; TMB‐high, high tumor mutational burden; TNBC, triple‐negative breast carcinoma.

In the CGP‐defined ERBB2‐amplified BCBM cohort (n = 178), there were fewer genomic alterations in PTEN (1.7% [3/178] vs. 21.6% [120/555], p < .001), RB1 (2.2% [4/178] vs. 12.8% [71/555], p < .001), and BRCA1 (1.1% [2/178] vs. 9.9% [55/555], p < .001) compared with the ERBB2‐nonamplified BCBM cohort (supplemental online Table 5). This contrasts with the CGP‐defined TMB‐high BCBM cohort (n = 119), where there was enrichment for alterations in PIK3CA (47.4% [55/119] vs. 26.5% [163/614], p = .001) and ARID1A (16.4% [19/119] vs. 5.7% [35/614], p = .009) compared with the non–TMB‐high BCBM cohort (supplemental online Table 6).

In the confirmed TNBC‐BM cohort, the top five genes with GAs were TP53 (87.5%, 14/16), RAD21 (56.3%, 9/16), PTEN (37.5%, 6/16), MYC (31.3%, 5/16), and VEGFA (18.8%, 3/16) (Fig. 3A). A significantly higher number of GAs were present in the TNBC‐BM cohort when compared with the primary TNBC cohort, although significant differences were not found in individual genes (p < .001, p ≥ .05, respectively; Table 2; Fig. 3B; supplemental online Table 7).

Figure 3.

Genomic landscape of triple‐negative breast carcinoma brain metastasis (TNBC‐BM). (A): Co‐mutation plot of the TNBC‐BM cohort. In the confirmed TNBC‐BM cohort, the top five genes with genomic alterations (GAs) were TP53, RAD21, PTEN, MYC, and VEGFA. (B): Long‐tail plot of genes with GAs and comparison between the TNBC‐BM and primary TNBC cohorts. A significantly higher number of GAs were present in the TNBC‐BM cohort compared with the primary TNBC cohort, although significant differences were not found in individual genes with GAs (p < .001, p ≥ .05). Abbreviation: TNBC, triple‐negative breast carcinoma.

ICPI Biomarkers

ICPI biomarkers of TMB‐high (16.2% [119/733] vs. 5.4% [584/10,772], p < .001), MSI‐high (1.9% [14/733] vs. 0.4% [42/10,772], p < .001), CD274 (which encodes for PD‐L1 protein) amplification (3.6% [27/733] vs. 1.7% [186/10,772], p < .001), and APOBEC mutational signature (5.9% [43/733] vs. 3.6% [384/10,772], p = .003) were significantly higher in the BCBM cohort when compared with the primary BC cohort (Table 1). In addition, we also examined the ICPI biomarker prevalence in disease subsets positive and negative for ERBB2 amplification, PIK3CA mutations, BRCA1/2 mutations, and ESR1 mutation (supplemental online Table 8). In the PIK3CA mutation–positive cohort, prevalence of TMB‐high (26.0% [51/196] vs. 12.7% [68/527], p < .001) and APOBEC mutational signature (14.8% [29/196], 2.6% [14/537], p < .001) was significantly higher when compared with the PIK3CA mutation–negative cohort. No significant difference was found in the other comparisons.

The frequency of PD‐L1 positivity was lower in the TNBC‐BM cohort than in the primary TNBC cohort (37.5% [6/16] vs. 64.0% [110/173], p = .057). However, when examining the other ICPI biomarkers (TMB‐high: 18.8% [3/16] vs. 2.3% [4/172], p = .014; CD274 amplification: 18.8% [3/16] vs. 2.9% [5/172], p = .022), we saw a significantly higher prevalence in the TNBC brain metastatic cohort when compared with the primary TNBC cohort. No patients with MSI‐high status were identified in the primary TNBC and BM cohorts. When using both CGP and PD‐L1 IHC, 37.5% (6/16) of patients with TNBC were eligible for atezolizumab based on PD‐L1 IHC, 18.8% (3/16) of patients were eligible for pembrolizumab based on TMB‐high status, and 12.5% (2/16) of patients were eligible for both atezolizumab and pembrolizumab based on PD‐L1 IHC and TMB‐high status (supplemental online Fig. 1).

Paired Primary BC and BCBM Samples

In our cohort, 11 paired primary BC and BCBM samples were identified (Table 3). The time between the collection date of the primary BC sample and that of the BCBM sample ranged from 5.7 months to 8 years. Overall, there were an additional 23 amplifications, 9 mutations, and 1 fusion detected in the paired BCBM samples when compared with the paired primary BC samples. In addition, of the 11 paired cases, 90.9% (10/11) had at least one additional GA discovered in the BCBM sample when compared with the primary BC sample. Also, 45.5% (5/11) of the BCBM samples did not have at least one GA that was found on the primary BC sample. Of importance, case 1 (hormone receptor positive and HER2 negative) had gained a PIK3CA E545K mutation in the BCBM sample and did not have any PIK3CA mutations on the primary BC sample, and case 10 (hormone receptor positive and HER2 negative) had gained a PIK3CA E726K mutation in the BCBM sample in addition to the PIK3CA E545K mutation in the original primary BC sample (Table 3). For ICPI biomarkers, case 11 showed a gain of a CD274 amplification and case 7 had changed to a TMB‐high and MSI‐high status in their paired BCBM samples, and three cases had lost TMB‐high status in their paired BCBM samples (Table 3).

We next compared the genomic profiles of the 11 primary breast cancer cases that eventually metastasized to the brain with the overall primary BC cohort. Here, we saw higher prevalence of GAs in TP53 (90.9% [10/11] vs. 59.4% [6,395/10,761], p = .860), PIK3CA (36.4% [4/11] vs. 31.4% [3,375/10,761], p = 1.000), ERBB2 (27.3% [3/11] vs. 11.8% [1,265/10,761], p = 1.000), and MYC (36.4% [4/11] vs. 21.1% [2,272/10,761], p = 1.000) in the primary BC that eventually metastasized to the brain, although no significance was found because of the limited number of these samples (supplemental online Table 9).

Discussion

This retrospective cohort study of 733 BCBM and 10,772 primary BC specimens revealed that brain metastases were more likely to exhibit TP53, ERBB2, RAD21, NF1, BRCA1, and ESR1 GAs. These genes have important clinical implications. For example, TP53 mutations has been shown to have distinct prognostic relevance, ERBB2 amplifications are an indication for the use of HER2 inhibitors, RAD21 expression confers resistance to chemotherapy, NF1 GAs are associated with contralateral breast cancer and poor survival, BRCA1/2 mutations are an indication for the use for poly (ADP‐ribose) polymerase (PARP) inhibitors, and ESR1 mutations are resistance biomarkers for aromatase inhibitor therapy [5, 6, 31, 32, 33, 34, 35]. In the confirmed TNBC cohort, the number of GAs per sample was also increased in the brain metastasis cohort, and in many cases it was different from the primary BCs. Because of these differences, for patients with breast carcinoma that has metastasized to the brain, the metastatic tissue in the brain or cerebrospinal fluid specimen, when safely available, should be considered as specimens for CGP testing [36, 37].

In the 11 paired primary BC and BCBM samples, we found that there was at least one additional GA in 90.9% (10/11) of BCBM (post‐treatment) samples when compared with the primary BC samples and that 45.5% (5/11) of the BCBM samples lost at least one GA when compared with the primary BC samples. One case had gained a PIK3CA mutation on the BCBM sample with no PIK3CA mutation detected in the primary BC sample, thus making the BM targetable with PIK3CA inhibitors. Another BM case gained an additional PIK3CA mutation on top of the original PIK3CA mutation, which likely portends a higher sensitivity to PIK3CA inhibitors when compared with a single PIK3CA mutation [38, 39]. Last, one case gained a CD274 amplification in the BCBM sample not previously detected in the primary BC sample, which confers sensitivity to ICPI [40]. Previously, it was shown that HER2‐positive and TNBC breast carcinoma have an increased risk of developing brain metastasis [20]. In our 11 patients with primary BC who eventually developed BM, we discovered a higher prevalence of GAs in TP53, PIK3CA, ERBB2, and MYC, which suggests that GAs in these genes could play a role in the metastasis of BC to the brain. It is important to note that our cohort is small, and so conclusions cannot be made on these data alone, but these findings do highlight important trends that should be expanded upon in a larger data set.

In the overall breast carcinoma cohort, the ICPI biomarkers of TMB‐high, MSI‐high, CD274 amplification, and APOBEC mutational signature were all enriched in the BCBM samples when compared with the primary BC samples. Importantly, several biomarkers have documented efficacy for immune checkpoint therapy in brain metastases (mostly from lung, melanoma, and renal cell cancers) [41]. These results further suggest that a subset of BCBMs are positive for ICPI biomarkers and could be considered for treatment with ICPI. We detected enrichment of TMB‐high and CD274 amplification in the TNBC‐BM samples; however, we also found a lower PD‐L1 positivity rate among the TNBC‐BM cohort compared with the primary TNBC cohort, similar to what was previously described [42]. One explanation for this observed difference in PD‐L1 positivity rate could be the small sample size. However, it is more likely that preanalytic factors in processing brain specimens or biologic reasons caused a lower PD‐L1 positivity rate in TNBC‐BM specimens when compared with primary TNBC specimens, and this difference could be considered when choosing a sample for PD‐L1 IHC testing for treatment purposes or enrollment into a clinical trial.

Last, in the TNBC‐BM confirmed cohort with PD‐L1 testing (n = 16), we saw that 37.5% (6/16) of patients with TNBC were eligible for atezolizumab based on PD‐L1 IHC, 18.8% (3/16) of patients were eligible for pembrolizumab based on TMB‐high status, and 12.5% (2/16) of patients were eligible for both atezolizumab and pembrolizumab based on PD‐L1 IHC and TMB‐high status. Although there is a subset of patients eligible for both atezolizumab and pembrolizumab based on TMB and PD‐L1 IHC, there is a distinct subset of patients only eligible for atezolizumab or pembrolizumab, exemplifying the importance of testing with both CGP and PD‐L1 IHC in these patients. Given the recent approval of the Dako 22C3 PD‐L1 IHC assay in TNBC for pembrolizumab, it would be important to determine if the positivity for this assay overlaps with TMB and Ventana PD‐L1 SP142 PD‐L1 IHC assays.

The major strength of this study is the large number of BCBM samples all undergoing centralized CGP using a single assay. However, a primary limitation of this study is the limited clinical information available with the patient samples. It is likely that some patients with primary BC in our study could also have had concurrent BCBM; however, in the absence of clinical histories, we do not know the extent of their disease. However, in general, most samples received at our institution are from patients with advanced disease at time of testing. Also, the rates of patients with BCBM who undergo surgery and obtain a surgical specimen are low (Sperduto et al. [21.1%, 521/2,473] and Lin et al. [15.8%, 46/291]), and this study only represents the patients who had a tissue specimen for CGP testing [43, 44]. In addition, whereas we have the HER2 status of the patients in this study based on CGP testing, we do not have the estrogen or progesterone receptor status of most of the patients to further stratify the patients based on this status. Another limitation is that the only FDA‐approved therapy specifically for BCBM is tucatinib. Although the actionability of the biomarkers described in this study has been associated with breast carcinoma, they have not been approved by the FDA for patients with BCBM, and so further clinical studies are needed to formally assess the actionability of these biomarkers in patients with BCBM.

Conclusion

We found a higher prevalence of clinically relevant GAs in patients with BCBM, which suggests that metastatic tissue to the brain or cerebrospinal fluid specimen should be considered as specimens for CGP testing [36]. In addition, we saw a higher frequency of immunotherapy biomarker positivity in the BCBM cohort, suggesting that patients with breast carcinoma metastasized to the brain should be assessed for ICPI biomarkers and advanced to such treatment modalities when clinically appropriate. Last, we found only a weak relationship between TMB and PD‐L1 IHC, which exemplifies the importance of testing with both PD‐L1 IHC and CGP for ICPI eligibility in patients with TNBC who have brain metastases.

Author Contributions

Conception/design: Richard S.P. Huang, Shakti H. Ramkissoon

Provision of study material or patients: Richard S.P. Huang, Kimberly McGregor, Douglas A. Mata, Brennan Decker, Matthew C. Hiemenz, Mirna Lechpammer, Natalie Danziger, Kelsie Schiavone, Jeffrey S. Ross, Shakti H. Ramkissoon

Collection and/or assembly of data: Richard S.P. Huang, James Haberberger

Data analysis and interpretation: Richard S.P. Huang, James Haberberger, Kimberly McGregor, Douglas A. Mata, Brennan Decker, Matthew C. Hiemenz, Mirna Lechpammer, Natalie Danziger, Kelsie Schiavone, James Creeden, Ryon P. Graf, Roy Strowd, Glenn J. Lesser, Evangelia D. Razis, Rupert Bartsch, Athina Giannoudis, Talvinder Bhogal, Nancy U. Lin, Lajos Pusztai, Jeffrey S. Ross, Carlo Palmieri, Shakti H. Ramkissoon

Manuscript writing: Richard S.P. Huang, James Haberberger, Kimberly McGregor, Douglas A. Mata, Brennan Decker, Matthew C. Hiemenz, Mirna Lechpammer, Natalie Danziger, Kelsie Schiavone, James Creeden, Ryon P. Graf, Roy Strowd, Glenn J. Lesser, Evangelia D. Razis, Rupert Bartsch, Athina Giannoudis, Talvinder Bhogal, Nancy U. Lin, Lajos Pusztai, Jeffrey S. Ross, Carlo Palmieri, Shakti H. Ramkissoon

Final approval of manuscript: Richard S.P. Huang, James Haberberger, Kimberly McGregor, Douglas A. Mata, Brennan Decker, Matthew C. Hiemenz, Mirna Lechpammer, Natalie Danziger, Kelsie Schiavone, James Creeden, Ryon P. Graf, Roy Strowd, Glenn J. Lesser, Evangelia D. Razis, Rupert Bartsch, Athina Giannoudis, Talvinder Bhogal, Nancy U. Lin, Lajos Pusztai, Jeffrey S. Ross, Carlo Palmieri, Shakti H. Ramkissoon

Disclosures

Richard S.P. Huang: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); James Haberberger: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Kimberly McGregor: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Douglas A. Mata: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Brennan Decker: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Matthew C. Hiemenz: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Mirna Lechpammer: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Natalie Danziger: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Kelsie Schiavone: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); James Creeden: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Ryon P. Graf: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Nancy U. Lin: Genentech, Merck, Pfizer, Seattle Genetics (RF), Seattle Genetics, Daichii Sankyo, Astra Zeneca, Denali Therapeutics, California Institute for Regenerative Medicine, Prelude Therapeutics (C/A); Lajos Pusztai: Seagen, Pfizer, AstraZeneca, Merck, Novartis, Bristol‐Myers Squibb, Genentech, Eisai, Pieris, Immunomedics, Clovis, Syndax, H3Bio, Radius Health, Daiichi (C/A, H), Seagen, AstraZeneca, Merck, Pfizer, Bristol‐Myers Squibb (RF—institution); Jeffrey S. Ross: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI); Shakti H. Ramkissoon: Foundation Medicine, Inc. (a wholly owned subsidiary of Roche) (E, OI). The other authors indicated no financial relationships.

(C/A) Consulting/advisory relationship; (RF) Research funding; (E) Employment; (ET) Expert testimony; (H) Honoraria received; (OI) Ownership interests; (IP) Intellectual property rights/inventor/patent holder; (SAB) Scientific advisory board

Supporting information

See http://www.TheOncologist.com for supplemental material available online.

Supplemental Figure 1 Venn diagram of relationship between tumor mutational burden and PD‐L1 SP142 immunohistochemistry positivity in triple negative breast carcinoma brain metastases (TNBC‐BM) patients

Appendix S1. Supporting tables