Comprehensive genomic profiling to identify actionable alterations for breast cancer brain metastases in the Chinese population

Abstract

Background

Breast cancer brain metastasis (BCBM) is a crucial issue in the treatment of breast cancer and is associated with poor prognosis. Therefore, novel therapeutic targets are urgently needed in clinical practice. In this study, we aimed to identify potential actionable targets in brain metastases (BMs) utilising the FoundationOne® CDx (F1CDx).

Patients and methods

Formalin-fixed paraffin-embedded archived specimens including 16 primary breast tumours (PTs), 49 BCBMs and 7 extracranial metastases (ECMs) from 54 patients who underwent surgery for BCBM were tested using F1CDx. Tumour-infiltrated lymphocytes (TILs) of BMs were also tested using haematoxylin–eosin staining.

Results

The median tumour mutational burden (TMB) and TILs in BMs were 5.0 (range 0-29) mut/Mb and 1.0% (range 0%-5.0%), respectively. High TMB (≥10 mut/Mb) was detected in four cases (8%). Genomic alterations (GAs) were detected in all samples. The top-ranked somatic mutations in BMs were TP53 (82%), PIK3CA (35%), MLL2 (22%), BRCA2 (14%) and ATM (14%) and the most prevalent copy number alterations were ERBB2 (64%), RAD21 (36%), CCND1 (32%), FGF19 (30%) and FGF3 (30%). The most prevalent GAs were relatively consistent between paired PTs and BMs. Actionable GAs were detected in 94% of all BMs. Consistent rate in actionable GAs was 38% (6/16) between paired PTs/ECMs and BMs. Compared to matched PTs/ECMs, additional actionable GAs (BRAF, FGFR1, PTEN, KIT and CCND1) were discovered in 31% (5/16) of the BMs.

Conclusions

TMB and TILs were relatively low in BCBMs. Comparable consistency in actionable GAs was identified between BCBMs and matched PTs/ECMs. It was, therefore, logical to carry out genomic testing for BCBMs to identify potential new therapeutic targets when BCBM specimens were available, as ∼31% of samples carried additional actionable GAs.

Highlights

• •BCBM genomic profiling is similar to that of the PTs: both have high mutation frequencies in TP53 and PIK3CA.
• •TMB and TILs were relatively low in BCBMs and ∼8% of BCBMs had high TMB.
• •Actionable GAs and additional actionable GAs undetected in the matched samples were detected in 94% and 31% of BCBMs.

Introduction

Breast cancer is a major global health issue, accounting for 30% of cancers in females and is the leading cause of cancer-related deaths among women according to the American Cancer Society.¹ Breast cancer brain metastasis (BCBM) is a critical issue in breast cancer research as it is a common cause of mortality among patients with metastatic breast cancer.^2,3 Previous studies have found that the average survival time after brain metastasis (BM) was 2-16 months.⁴ Patients with BMs are treated with multimodal therapies, including surgery, radiotherapy, chemotherapy, immunotherapy and targeted therapy.⁵ However, the prognosis of patients with BCBM remains poor. This is because surgery and radiotherapy are often unlikely to be carried out repeatedly due to invasive approaches or dosage limitations, especially when disease progression occurs and most of the agents cannot fully cross the blood–brain barrier to be effective.⁶ Therefore, novel therapeutic targets are still needed in clinical practice.

Several studies have examined the concordance and differences between BMs and primary breast tumours (PTs), from clinicopathological features to gene levels. Most of the metastatic lesions have greater genomic heterogeneity than PTs.^7,8 A next-generation sequencing (NGS) study that included 20 matched cases has found that ∼20% of patients with human epidermal growth factor receptor 2 (HER2)-negative primary lesions had HER2 status changes (including amplification and mutation) in BMs.⁹ These results suggest that the therapeutic targets of BMs may have differences comparable to those of PTs. Thus, in this study, we aimed to describe the genomic profiling of BCBM and, more importantly, to reveal the consistency of actionable genomic alterations (GAs) between BMs and the matched PTs.

Patients and methods

Patient inclusion and tissue sample acquisition

This was a retrospective study of 54 patients with BCBM who underwent surgery for BMs at Sun Yat-sun University Cancer Center between 1995 and 2018. Formalin-fixed paraffin-embedded (FFPE) biopsy specimens from 54 BCBM patients, including 16 PTs, 49 BMs, and 7 extracranial metastases (ECMs), were included in the study. This study was approved by the Institutional Review Board of Sun Yat-sen University Cancer Center (No. B2019-200-Y01). A flow chart of the study sample screening is shown in Supplementary Figure S1, available at https://doi.org/10.1016/j.esmoop.2024.102389.

Clinical review

Clinical and pathological information, such as age, date of diagnosis of breast cancer, date of diagnosis of BCBM, date of operations, TNM (tumour–node–metastasis) classification, hormonal receptor (HR) status, HER2 status, the molecular subtypes of BMs, number of BMs, treatment regimens, BM-free survival (BMFS) and overall survival (OS), were included. BMFS was defined as the time from the operation of the brain lesions to relapse or death of the brain lesions. OS was defined as the time from the date of diagnosis of BCBM to death from any cause.

Genetic alteration assessment

FFPE samples from 54 patients with BCBM were tested utilising FoundationOne® CDx (F1CDx, Foundation Medicine, Inc., Cambridge, MA), an NGS-based diagnostic platform detecting alteration of 324 genes.¹⁰ Sequencing was carried out with the Illumina HiSeq® 4000 (Illumina, San Diego, CA) to identify all classes of GAs, including base substitutions, indels, copy number alterations (CNAs) including amplification and homozygous gene deletions, and selected genomic rearrangements such as gene fusions. Microsatellite instability and tumour mutational burden (TMB) were also reported. Actionable GAs were defined as alterations that target either the GAs or the pathway activated by the alterations, with efficacy reported in humans.

Assessment of tumour-infiltrating lymphocytes

Tumour-infiltrating lymphocytes (TILs) were evaluated according to a five-step standardized scoring system developed by the International Immuno-Oncology Biomarker Working Group.^11,12 As previously described,^13,14 haematoxylin–eosin-stained whole-slide images of 55 BM samples were reviewed by a pathologist, TILs encompass all mononuclear cells (including lymphocytes and plasma cells), and polymorphonuclear leukocytes (neutrophils) were excluded. The percentage of stromal TILs in the entire examined area (stromal TILs%) was assessed within the invasive tumour borders of BCBM samples, which was considered a semiquantitative continuous parameter indicating how much of the demarcated stromal area exhibited dense mononuclear infiltrates.

Statistical analysis

The BMFS and OS with 95% confidence intervals (CIs) were assessed using Kaplan–Meier estimates and the log-rank tests. Statistical analysis of all genes was based on a dichotomy (i.e. presence/absence of any alteration). Differences in frequency between groups were determined using the chi-square or Fisher’s exact test. All statistical analyses were carried out using SPSS version 22.0 and R version 3.4.3 software. Venn diagrams were constructed using EVenn.¹⁵ Two-sided statistical tests were used in this study and P values <0.05 were considered statistically significant.

Results

Clinicopathological characteristics and prognosis of patients with BCBM

The median age at breast cancer diagnosis was 43 (range 26-62) years. The median age at surgery for BMs was 48 (range 27-65) years. The median time from breast cancer diagnosis to BM was 38 (range 0-229.5) months. Half of the patients had only BMs during surgery. Most patients underwent surgery for BMs <3 months or >12 months after diagnosis of BM. Before surgery, only five patients (9%) received radiotherapy, and over 50% of them received systemic therapy. At the time of surgery, 50% of the patients had no evidence of ECMs, and only 8 patients had disease progression of the ECMs. The positivity rate of HR in BMs was similar to the negativity rate (48% versus 52%), whereas the positivity rate of HER2 was higher (57% versus 37%). For molecular subtypes according to immunohistochemical staining, 24% of BMs were luminal B HER2 negative, and only seven BMs (13%) were triple negative (Table 1).

Table 1.

Patients’ characteristics

Variables	n	%
Age at diagnosis
Median, range (years)	43 (26-62)
≥45	20	37
<45	34	63
TNM staging at diagnosis
I	3	6
II	8	15
III	20	37
IV	9	17
Unknown	14	26
Hormonal receptor (BM)
Positive	26	48
Negative	28	52
HER2 status (BM)
Positive	31	57
Negative	20	37
Unknown	3	6
DFS
De novo stage IV	9	17
<12 months	7	13
≥12 months	38	70
First metastatic site
Brain only	27	50
Extracranial only	22	41
Brain and extracranial	5	9
Time from diagnosis of BM to brain surgery
≤3 months	25	46
3-6 months	0	0
6-12 months	6	11
>12 months	23	43
Age at brain surgery
Median, range (years)	48 (27-65)
≥45	22	41
<45	32	59
Number of BMs
1	44	81
2	2	4
3	4	7
>3	4	7
RT
None	25	46
Pre-operation	5	9
Post-operation	24	44
Systemic therapy prior surgery
Yes	24	44
No	30	56
BM only at surgery
Yes	25	46
No	29	54
Extracranial disease status at surgery
Complete response/no extracranial disease	27	50
Partial response	1	2
Stable disease	11	20
Progressive disease	8	15
Unknown	7	13
Molecular subtype (BM IHC)a
Luminal B HER2 negative	13	24
Luminal B HER2 positive	15	28
HER2	16	30
TNBC	7	13
Unknown	3	6

BM, brain metastasis; HER2, human epidermal growth factor receptor 2; IHC, immunohistochemistry; TNBC, triple-negative breast cancer; TNM, tumour–node–metastasis.

^aPercentage exceeds 100% because of rounding.

After the median follow-up of 34.3 months, the median BMFS and OS of all patients were 24.7 months (95% CI 14.6-34.7 months) and 62.5 months (95% CI 23.1-101.8 months), respectively (Supplementary Figure S2, available at https://doi.org/10.1016/j.esmoop.2024.102389). The HER2 status of the BMs was related to prognosis after surgery for BMs. Patients with HER2-postive status achieved longer BMFS than those with HER2-negative status. However, we did not observe any difference in OS. Other factors, including estrogen receptor status, number of BMs, disease status of the extracranial disease and whether radiotherapy or systemic therapy was administered before surgery, did not influence BMFS or OS.

GAs of BCBM

We carried out comprehensive genomic profiling analysis of ∼72 samples. GAs were detected in all the samples. A heatmap of the top 30 mutated genes is shown in Figure 1. TP53 (85%), PIK3CA (37%), MLL2 (24%), BRCA2 (17%), CDH1 (14%), ATM (13%), BCOR (13%), CREFBBP (13%), FGFR4 (13%) and MSH6 (13%) were the top 10 most frequent somatic mutations in all samples (including all PTs, BMs and ECMs). The top-ranked somatic mutations in BMs were TP53 (82%), PIK3CA (35%), MLL2 (22%), BRCA2 (14%) and ATM (14%). When focusing on CNAs of the BMs, we found that most of the alterations were amplifications and only a few CNA deletions were detected. Common CNAs, including ERBB2 (64%), RAD21 (36%), CCND1 (32%), FGF19 (30%) and FGF3 (30%), were detected in our study (Figure 2).

Figure 1.

Somatic mutational profiling of 71 samples (top 30 frequent mutations, 1 breast cancer brain metastasis sample with only KEL mutation not shown).

Figure 2.

Copy number alterations of all brain metastases.

The heatmap of 11 matched samples of the PTs and BMs showed that the consistency of the top 10 alterations was high, both in somatic alteration genes and CNAs (Figure 3A and B). However, the total consistency between PTs and BMs was low. When comparing the frequency of the commonly observed GAs between all PTs and BMs, no significant alterations were found, although alterations in some genes were more likely to be observed in BMs (Figure 3C). ATM mutation was detected in seven BCBMs but not in PTs, although the difference was not statistically significant.

Figure 3.

Comparison of GAs between BMs and PTs. (A) Somatic mutational profiling of 11 matched PTs and BMs. (B) Copy number alterations of 11 matched PTs and BMs. (C) Forest plot of somatic mutations with greatest difference in frequency in all BMs versus all PTs. BMs, brain metastases; GA, genomic alteration; NS, not significant; PTs, primary breast tumours.

TMB, TILs and actionable GAs of BCBM

The microsatellite status of all samples (72) was stable. We analysed TILs for all BM samples and found that TILs were relatively low, with an average of 1.2% (standard deviation ±1.3, range 0-5). The median TMB of all BMs was 5 mut/Mb (range 0-29 mut/Mb). There were four cases with high TMB (≥10 mut/Mb; Figure 4A and B). Spearman’s correlation analysis showed no correlation between TMB and TILs. Among four cases with high TMB, one case had TILs of 5% and the other three cases had TILs of 1%. One case with TMB of 29 mut/Mb also carried amplification of ERBB2, CCND1 and AKT3. All of the four cases were HER2 positive and none of them received immunotherapy until the last updated follow-up.

Figure 4.

TILs, actionable GAs and TMB of all BMs. (A) Boxplots of TILs, actionable GAs and TMB of all BMs. (B) Frequency of all actionable GAs in all BMs. BMs, brain metastases; GA, genomic alteration; TILs, tumour-infiltrating lymphocytes; TMB, tumour mutational burden.

Actionable GAs were detected in 46 BM samples (94%). A total of 96 actionable GAs were detected with a median number of 2 (range 0-5). The most prevalent alterations were ERBB2 (28/96, 29%), PIK3CA (15/96, 16%), CCND1 (14/96, 15%), FGFR1 (7/96, 7%) and NF1 (5/96, 5%). We also found three cases with BRCA2 alterations and one case with BRCA1 alteration in BM samples none of which harboured ERBB2 alterations (Figure 4B). Focusing on the consistency of actionable GAs in BMs and PTs or ECMs, 6/16 (38%) pairs were totally consistent between PTs/ECMs and BMs (Supplementary Figure S3, available at https://doi.org/10.1016/j.esmoop.2024.102389). There were five cases (31%) with additional actionable GAs (BRAF, FGFR1, PTEN, KIT and CCND1) in BMs, which were not detected in the matched PTs (two cases) or ECMs (three cases), whereas five cases (31%) had a reduced number of actionable GAs in BMs. The actionable GAs of all BM samples and matched PT/ECM samples and the corresponding potentially effective therapies are available in Supplementary Table S1, available at https://doi.org/10.1016/j.esmoop.2024.102389.

Discussion

To our knowledge, this is the largest study focusing on genomic profiling and actionable GAs and evaluating the TILs of BCBM in Chinese patients. The results of the 49 BMs in this study showed that the most commonly mutated genes were TP53 and PIK3CA, which is consistent with a recently reported systematic review.⁸ TP53 and PIK3CA were the most prevalent genes detected in extracranial tumours, as previously reported.^16,17 This revealed that mutation of these two genes was maintained throughout disease progression and metastasis. In addition, among the additional 20 genes commonly detected in BMs reported in this review, five genes (BRCA2, CDH1, ATM, BRCA1, ARID1A) were in the top 30 in our study. As most of the samples were HER2 positive, the most frequent CNA was ERBB2, which also proved that the HER2-positive status usually did not change in BMs.⁹ Amplifications of genes involved in the DNA repair pathways, the cell cycle and the fibroblast growth factor family were second most prevalent CNAs in our study. In addition to TP53, PIK3CA and ERBB2, discrepancies were observed for other genes. The relatively low consistency among the other genes may be due to patients’ selection and different gene panels.

We also aimed to identify alterations that were substantially different between BMs and the PTs/ECMs. However, no significant changes in frequencies were observed. The reason may be small sample sizes of the PT (11/54) and ECM (7/54) samples. In a systematic review including 11 906 PT, 5541 ECM and 1485 BM samples, the authors reported that ESR1, ERBB2, EGFR, PTEN, BRCA2 and NOTCH1 had a significantly higher mutation prevalence¹⁸; however, no data from matched samples were reported. We found that, in 11 matched pairs of PTs and BMs, the genomic profiling of the BMs was diverged from the matched PTs. This has been reported in several previous studies.^19,20 With regard to actionable GAs, only 5 of the 16 (31%) cases detected additional GAs that were not detected in the matched PTs or ECMs. Studies on concordance of actionable GAs between the BMs and PTs or ECMs vary differently.21, 22, 23 Discrepancies were often observed because the sample sizes and percentages of molecular subtypes varied. In addition, the brain microenvironment differs dramatically from that of the extracranial lesions, which further influences the evolution of the tumour cells in the brain and generates a heterogeneous state.²⁴

Four of the 49 BM samples were defined as having a high TMB which is an indication for immunotherapy.^25,26 BRAF, FGFR1, PTEN, KIT and CCND1 detected only in BM samples are targets of certain off-label drugs, which provided more therapeutic options.

TILs have also been explored as biomarkers for prognosis and response rate in breast cancer, with a high percentage of TILs associated with a better response rate to immunotherapy and prognosis in certain breast cancer subtypes.²⁷ In our study, we discovered that TILs in BMs were very low and failed to further sub-classify the different lymphocytes. A study²⁸ assessing TILs in 46 pairs of primary and brain metastatic samples has shown that the median TIL category of the BMs was 5%, which was significantly lower than that of the PTs. TMB was not correlated with TILs in our study, which is consistent with the results of a previous study on melanoma.²⁹

Our study had some limitations. Owing to the difficulty in obtaining BM samples, we did not set restrictions when screening the cases. Therefore, information regarding PTs was unavailable for some patients and we could not analyse the switching rate of molecular subtypes between the PT and BM samples. Secondary, although the BM sample size in our study was large, the corresponding matched PT or ECM sample sizes were not large enough, which may have restricted the discovery of significant genomic findings during disease progression and evolution.

Conclusion

TMB and TILs were relatively low in BMs. Most BCBMs harboured at least one actionable GA. Comparable consistency in actionable GAs was identified between PTs and BMs. Approximately 31% of BMs harboured additional actionable GAs; thus, it is reasonable to carry out genomic testing for novel therapeutic options when BCBM specimens are available.