Introduction
Breast cancer (BC) is the most frequently diagnosed cancer in the USA and the second most common cause of cancer-related death in women [
1], although there have been significant advances in treatment strategies over the past few decades. In addition to cytotoxic chemotherapy, endocrine therapy, and radiation therapy, in the molecular era, BC patients often receive targeted treatments based on their unique cancer biology in both adjuvant and neoadjuvant settings. For instance, anti-Human Epidermal Growth Factor Receptor 2 (HER2) therapy combined with chemotherapy has become the standard of care in HER2-positive (HER2+) BCs [
2,
3]. Hormone receptor-positive (HR+)/HER2- tumors have historically received endocrine therapy and, more recently, CDK4/6 inhibitors as targeted treatments [
4,
5]. Some molecularly targeted therapies have also emerged for triple-negative breast cancer (TNBC), such as poly-ADP ribose polymerase inhibitors for
BRCA-mutated tumors [
6,
7] and checkpoint inhibitors for PD-L1+ tumors [
9,
10]. Despite these advances, most patients with metastatic BC develop drug resistance or rapidly progress shortly after initiating guideline-informed systemic therapy. Increasingly, genomic testing is used to expand treatment options to overcome de novo or acquired resistance. Patients with HR+/HER2- tumors harboring a
PIK3CA mutation, for example, experience a prolonged progression-free survival when treated with a phosphoinositide-3-kinase (PI3K) inhibitor in combination with endocrine therapy [
8]. However, there is still much room for improvement in endocrine-resistant HR+/HER2- BC and TNBC survival due to their aggressive natures and relatively limited treatment options [
9,
10].
TNBC is much less common than the HR+/HER2- subtype in the US population [
11], but disproportionately affects patients of African ancestry (AA) compared to those of European ancestry (EA). Moreover, AA patients experience a 40% higher mortality rate than EA patients across all subtypes [
11], despite lower incidence rates of BC overall. The complex biologic and social drivers causing these disparities are beginning to be revealed through rigorous studies. While socioeconomic factors like inadequate access to quality care partially contribute to higher mortality, outcome disparities remain even after adjusting for these factors [
12]. Several studies point to biological differences in the molecular drivers and evolutionary trajectory of BC based on ancestry. For example, population-based studies have observed higher rates of germline
BRCA1 mutations in AA compared to EA patients [
13]. In addition to germline mutations, somatic mutational differences have also been reported. An analysis of tumor sequencing data in the Cancer Genome Atlas conducted by our group revealed a higher prevalence of
TP53 mutations in AA patients and a lower prevalence of alterations in
PIK3CA [
14]. In this same dataset,
TP53 mutations were found to be a positive predictor for recurrence. These data and similar reports demonstrate a significant contribution of genomic differences to the mortality gap between AA and EA patients with BC and point to differential drivers of disease that may be potential therapeutic targets [
15].
There is a paucity of molecular and clinical data from underserved and understudied populations, and the few studies using molecular assessments have been underpowered to demonstrate prognostic value in AA patients with BC in the USA [
16,
17]. Analyses of comprehensive, large-scale oncology databases are needed to close the knowledge gap and address the unmet clinical need of diverse patient populations. Here, we performed a retrospective analysis of genomic and transcriptomic breast tumor sequencing data from Tempus’ large database and compared results between AA and EA patients. Furthermore, we estimated genetic ancestry employing ancestry-informative markers (AIMs) tailored specifically to our genomic data [
18], which can fill gaps in race/ethnicity metadata from electronic health records (EHRs), provide more meaningful biological insights, and increase data availability in under-researched ancestral populations. Combining ancestry estimations, EHR metadata, genomics, and transcriptomics in this real-world cohort may guide the future development of treatment strategies by providing data for biomarker-informed research and precision cancer care.
Discussion
The results from our large-scale study provide innovative insight by presenting genomic and transcriptomic differences between BC tumors from AA and EA patients, with findings stratified by clinical features in a real-world cohort. The findings indicate the utility of assessing molecular landscapes when conducting both basic research and clinical trials according to the specific ancestries, as well as reveal biologically different mechanisms between AA and EA patients. Importantly, the mechanisms identified carry implications for use of molecularly targeted therapy to broaden early access to clinical trials, including the use of combination therapies in patients with residual disease after neoadjuvant therapy to prevent metastases and late recurrences.
In this cohort, the frequency of
TP53 mutation was higher in AA than EA, which is consistent with previous reports, but there were no significant differences between ancestries in the stratified subtypes [
14,
42,
43]. While
TP53 mutations were much more frequent in AA patients with TNBC, our results indicate that the higher frequency of
TP53 mutation in AA tumors from the entire cohort is due to the higher incidence of TNBC in AA. We also observed a significantly lower mutation rate of
PIK3CA in AA tumors, particularly in patients with HR+/HER2- tumors. Luminal-type tumors are more likely to have
PIK3CA mutations, providing opportunities for molecularly targeted therapy to improve survival [
44]. Several compounds targeting
PIK3CA have been studied and the US Food and Drug Administration has approved the first PI3K inhibitor alpelisib for advanced luminal BC harboring the
PIK3CA mutation [
8]. While the frequency of
PIK3CA mutation was lower, any AA patient with identifiable mutations would be expected to benefit from alpelisib, but lack of access to molecular diagnosis could continue to drive racial disparities in outcomes for these patients. This underscores the need to consider each patient as an individual and utilize NGS testing early to capture patients with unique tumor biology and druggable pathways who would benefit from molecularly targeted therapy and/or immunotherapy.
We observed trends suggesting differences in the rates of somatic
BRCA1 and
BRCA2 mutations. Previous studies have reported elevated frequency of
BRCA1 and
BRCA2 germline mutations among AA patients [
45‐
47]; however, the somatic mutation differences by race have not been well documented up until now. While the differences were not significant, we observed a potential racial difference of somatic
BRCA1 and
BRCA2 mutations in both HR+/HER2- and TNBC subtypes. AA patients were more likely to have
BRCA2 mutations within the HR+/HER2- subtype, whereas EA patients were more likely to have
BRCA1 mutations within the TNBC subtype. These findings and previous studies of germline alterations might also have implications for DNA repair-targeted therapies and/or immunotherapies [
48].
KMT2C, a member of the myeloid/lymphoid or mixed-lineage leukemia family that encodes a histone methyltransferase, is one of the most frequently mutated genes in HR+ BC [
49]. The deletion of
KMT2C has been reported to be associated with resistance to endocrine therapy and worse prognosis [
50]. We observed that the frequency of
KMT2C mutation was significantly higher in AA than EA among TNBC patients (23% vs. 12%) as well as patients with HR+/HER2- tumors (24% vs. 15%), suggesting that the loss of
KMT2C function might disproportionately affect the survival outcome of AA patients in both subtypes. Previous reports focusing on mutational differences between races also indicated that
GATA3, which has a critical role in the development of luminal type BC, is mutated in around 10% of both AA and EA patients [
42,
43]. In contrast, 22% of AA patients in our cohort had tumors with mutated
GATA3 in HR+/HER2- subtypes, which might contribute to the racial disparity of luminal BC prognosis.
From the full-transcriptome RNA-sequencing data, we identified significantly different expression of over 8000 independent genes between the two ancestral groups. Various genes were exclusively expressed according to subtype (HR+/HER2- or TNBC) or disease stage (stage I-III or stage IV). Here, it is worth mentioning that
RPL10, which is responsible for DNA replication stress and promoting proliferation and oncogenesis [
40], was expressed higher in AA patients throughout almost every subtype and stage. Whereas the higher expression of
RPL10 is reported to be associated with poor prognosis mainly in hematologic malignancy [
51,
52], its relation to prognosis in breast malignancy has not been fully investigated. This transcriptomic change in AA patients regardless of subtype or stage might explain the prognostic disparity of AA patients and could be a potential therapeutic target in BC. On the other hand,
HSPA1A, a member of the heat shock protein 70 family, was significantly lower in AA patients in both HR+/HER2- BC and TNBC. The clinical significance of
HSPA1A status is unknown in malignant tumors, including BC [
53].
ATRX, which has a critical role in chromatin remodeling, was also expressed lower in AA patients in both HR+/HER2- BC and TNBC. Previous studies revealed that
ATRX loss is associated with an increase in cancer aggressiveness [
54]. Among a large number of genes with little genomic and clinical annotation,
NUTM2F was significantly overexpressed in AA patients throughout subtypes and stages. Our comprehensive birds-eye view analysis has identified many potential genes to consider in future basic and clinical research.
When breaking down the enrichment analysis of hallmark gene sets into the subtypes and stages, 125 exhibited significantly different enrichment between AA and EA patients across the entire cohort. No significant differences were observed when comparing gene set expression between ancestral groups within the TNBC subpopulation, even when stratifying by stage. Although the gene sets evaluated here are informative markers of oncogenic pathways, there are other collections of curated gene sets available for analysis, such as the KEGG pathways, and future studies could include these to provide a more thorough investigation of differences between AA and EA patients. Nevertheless, notable differences in many mutations and individual gene expressions between the two ancestral groups in TNBC were demonstrated as an initial assessment.
Meanwhile, among the stage IV HR+/HER2- group, 10 differentially expressed gene sets were identified. As a translation from research to clinic, we revealed four differentially expressed gene sets relevant to BC treatment: ERBB2_UP.V1_UP, LTE2_UP.V1_UP, HALLMARK_FATTY_ACID_METABOLISM, and HALLMARK_ANDROGEN_RESPONSE [
28,
41]. In the stage IV HR+/HER2- group, these pathways could contribute to worse prognosis of AA patients and, accordingly, would be worth investigating in further prognostic studies consisting of AA and EA patients.
Our findings could have implications for prognosis, response to therapy, and enrollment of diverse populations in clinical trials and precision oncology studies. If genomic and transcriptomic differences are not considered, applying the results of clinical trials on the studied population (e.g., EA) to another population (e.g., AA) could lead to suboptimal treatment decisions. The molecular distinctions identified here indicate that the efficacy of novel targeted therapy combinations may differ by patient ancestry, and thus highlight an opportunity to optimize neoadjuvant and adjuvant clinical trial design. Additionally, the availability of this data on a population level and for each patient should ultimately result in more clinical trial opportunities including early access to trials for more patients.
Some potential limitations of this study are the broad analyses and heterogeneity of the cohort, as the patients included in the study were highly selected and derived from various institutions and spanned multiple subtypes and stages. Another is the lack of treatment data, as there is likely a mix of treatment-naïve and treatment-refractory patients included in this study, and outcomes data. Nevertheless, in previous studies from our group, we examined NGS data using the Tempus xT assay in patients undergoing neoadjuvant chemotherapy from the Chicago Multiethnic Breast Cancer Cohort and found similar patterns [
55]. Furthermore, the patient characteristics were not entirely balanced between AA and EA, considering there were higher-grade tumors and more TNBCs in the AA patient group. Although our large sample size and stratifications by stage and subtype may have ameliorated the impact of differences in patients’ background between the two groups, our analyses did not directly address socioeconomic differences that could play roles in treatment outcomes. The incorporation of those additional factors is beyond the scope of this study and could be included in future studies.
Overall, these data show important differences in BC mutational spectrums, gene expression, and relevant transcriptional pathways between patients with genetically determined African and European ancestries, particularly within the HR+/HER2- BC and TNBC subtypes. To serve diverse populations of patients diagnosed with BC, promote equitable access to clinical trials, and accelerate the development of clinical decision tools for precision care, future studies should focus on geography and genetic ancestry when conducting biomarker-informed, early-phase clinical trials.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.