Fusion-Associated Carcinomas of the Breast: Diagnostic, Prognostic, and Therapeutic Significance

Abstract

Recurrent gene fusions comprise a class of viable genetic targets in solid tumors that have culminated several recent break-through cancer therapies. Their role in breast cancer, however, remains largely underappreciated due to the complexity of genomic rearrangements in breast malignancy. Just recently, we and others have identified several recurrent gene fusions in breast cancer with important clinical and biological implications. Examples of the most significant recurrent gene fusions to date include 1) ESR1-CCDC170 gene fusions in luminal B and endocrine resistant breast cancer that exert oncogenic function via modulating the HER2/HER3/SRC complex, 2) ESR1 exon 6 fusions in metastatic disease that drive estrogen-independent ER transcriptional activity, 3) BCL2L14-ETV6 fusions in a more aggressive form of the triple negative subtype that prime epithelial-mesenchymal transition and endow paclitaxel resistance, 4) the ETV6-NTRK3 fusion in secretory breast carcinoma that constitutively activates NTRK3 kinase, 5) the oncogenic MYB-NFIB fusion as a genetic driver underpinning adenoid cystic carcinomas of the breast that activates MYB pathway, and 6) the NOTCH/MAST kinase gene fusions that activate NOTCH and MAST signaling. Importantly, these fusions are enriched in more aggressive and lethal breast cancer presentations and appear to confer therapeutic resistance. Thus, these gene fusions could be utilized as genetic biomarkers to identify patients that require more intensive treatment and surveillance. In addition, kinase fusions are currently being evaluated in breast cancer clinical trials and on-going mechanistic investigation is exposing therapeutic vulnerabilities in patients with fusion positive disease.

Introduction

Gene fusions are a subtype of chromosomal structural rearrangements that coalesce the genetic material from two genes. These genomic aberrations can result in enhanced expression of the 3’ fused gene partner, formation of in-frame fusion transcripts that encode chimeric proteins with altered cellular function¹, as well as development of inactivating out-of-frame fusion transcripts^1,2. Previous studies have cited that genetic gene fusions function as tumorigenic events in 16.5% of cancers and appear to be druggable in 6% of cases³. In addition to genomic rearrangements, recent studies suggest the existence of non-genetic gene fusions arising from intergenic cis- or trans-splicing due to aberrant RNA splicing. Cis-splicing between adjacent genes (typically within 30kb) is a splicing event that occurs within a single pre-mRNA resulting from transcriptional read-through event between the two genes⁴. In contrast to cis-splicing, aberrant products from trans-splicing arise from two separate mRNAs of different genes^5–7. Although trans-spliced fusion events are less abundant, these fusions can be locked into place by genomic rearrangements that lead to malignancy⁵. The more abundant cis-spliced fusions have previously been assumed to be non-pathological, however, our recent study suggests that cancer-testis specific (testis overexpressed) cis-spliced fusions could contribute to breast cancer pathobiology and may confer potential therapeutic vulnerabilities⁸.

From the clinical perspective, recurrent gene fusions represent a critical class of oncogenic aberrations that have helped shape the precision medicine landscape. In particular, the recent discoveries of recurrent gene fusions in epithelial tumors have generated tremendous clinical impacts⁹. The EML4-ALK and ROS1 fusions found in 4–5% and 1–2% of lung cancer cases respectively have been targeted with effective novel therapies resulting in subsequent durable disease control^10,11. Recently, Larotrectinib has received FDA approval as the first targeted therapy with tissue-agnostic indications against Neurotrophic Tyrosine Receptor Kinase (NTRK) fusions found in ~1% of solid tumors¹². Although infrequent in proportion, these driver gene fusions can help discern genetic subtyping of solid tumors that can be managed through individualized fusion-targeted therapies. In addition to serving as actionable drug targets, recurrent gene fusions may also contribute as diagnostic, predictive, and prognostic markers in specific cancer entities. This is best represented by the ETV6-RUNX1 fusion that is used clinically as a predictive biomarker for tailored chemotherapies in acute lymphoblastic lymphoma¹³. In addition, the TMPRSS2-ERG gene fusion is currently being evaluated as a diagnostic and prognostic biomarker in prostate cancer clinical trials¹⁴.

Breast cancer is the most common malignancy diagnosed in women and harbors the highest number of gene fusions among cancer types¹⁵. Breast cancer has four major molecular subtypes: the hormone-receptor (HR) positive luminal A, the HR positive luminal B, the HER2-enriched (HR negative and HER2 receptor positive) subtype, and the triple-negative (HR and HER2 negative; TNBC) subtype. Among these molecular classifications, luminal B and TNBC are considered as the more aggressive tumor forms that lack effective interventions. The luminal B subtype accounts for 15–20% of all breast cancers¹⁶ and is the most common subtype in young women¹⁷. Luminal B disease can only be approximately defined based upon ER (+), HER2 (−), PR (<20%), and Ki67 (>14%), or based upon a PAM50 multi-gene signature^18–20. There is a lack of definitive genetic biomarkers and treatment options are limited²¹. These tumors are characterized by more aggressive clinical behavior, early relapse following endocrine therapy, and high risk of metastatic dissemination. TNBC accounts for 10–20% of breast cancer and bears the highest mortality rate²². Chemotherapy remains the cornerstone for TNBC treatment due to a lack of well-defined molecular targets. The pathomolecular events underlying these more aggressive breast cancer forms remains ill understood and recent genome sequencing studies reveal a paucity of actionable drivers specific to luminal B or TNBCs^23,24, hindering the efforts to devise effective targeted therapies.

Despite the substantial number of genomic rearrangements observed in breast cancer genomes, the number of recurrent gene fusions only represents a tiny proportion of these structural variants, and even less are expected to be of pathological significance.^2,25,26 In the past decade, we and others have identified several recurrent gene fusions in breast cancer with important pathobiological functions. Fascinatingly, most of these recurrent gene fusions are preferentially detected in the more aggressive breast cancer forms or in metastatic disease. In addition to canonical simple gene fusions, recently functional double-hop fusion transcripts transcribed from three distinct genomic regions as a result of complex genomic rearrangements have been described in multiple breast cancer cases²⁷. Such fusion transcripts are difficult to identify with short-read sequencing methods and may represent unique contribution of complex structural variants to breast cancer pathobiology. Herein, we focus our review on fusion-associated carcinomas of the breast and discuss the diagnostic, prognostic, and therapeutic significance of these recently identified pathological recurrent gene fusions.

Clinical and therapeutic relevance of ESR1-CCDC170 associated luminal B and endocrine resistant breast cancers

Prior to our studies, estrogen-receptor (ER) positive luminal breast cancer had not been associated with recurrent gene fusions. In 2014, we reported the first ESR1 fusions generated by localized rearrangements between the ESR1 gene encoding ERα and its immediate centromeric neighbor gene coiled-coil domain containing 170 (CCDC170)²⁸. ESR1-CCDC170 fusion events are detected in ~8% of luminal B breast cancers, a majority of which are tandem duplications. The fusion between ESR1 and CCDC170 represents a type of intrachromosomal fusion involving genes within a 500kb range which we coined as an adjacent gene rearrangement (AGR), which could be generated by tandem duplications²⁸, interstitial deletions²⁹, or cryptic balanced rearrangements. We found that these unique AGRs may appear more frequently than previously noted in breast cancer. In fact, our data shows that about half of the somatic translocations in breast tumors are within a 500kb span, suggesting a measurable prevalence of such cryptic events.

Distinct from other ESR1 fusions that preserve exons 1–6 and thus the ER transcriptional activation domain³⁰, most ESR1-CCDC170 fusions join the 5’ untranslated region (UTR) of ESR1 to the coding region of CCDC170 resulting in the expression of an amino(N)-terminally truncated CCDC170 (ΔCCDC170) driven by a constitutively active ESR1 promoter. Wildtype CCDC170 belongs to the structural maintenance of chromosome (SMC) protein family that maintains chromosome conformation through SMC-domain dependent looping and microtubule stabilization during interphase and mitosis^31,32. Genetic variants of CCDC170 have been reported to be prominent breast cancer risk factors^33,34. SMC proteins are formed by two long coiled-coil domains connected by a non-helical sequence, the ‘hinge’, which presumably corresponds to the low complexity region of CCDC170. SMC proteins contain a highly conserved ATP-binding cassette (ABC) that drive dimerization and subsequent SMC folding at the hinge region through antiparallel coiled-coil interactions³⁵. While a wide-variety of ESR1-CCDC170 variants have been observed, only five distinct forms of ΔCCDC170 proteins are encoded due to the five silent ATG start codons present within the CCDC170 open-reading frame (ORF) (Fig. 1a). Although these fusion variants result in varying degrees of SMC domain deletion, each retains a putative high-affinity ATP-binding pocket at the C-terminus.

Figure 1. Schematics of ESR1-CCDC170 fusion protein structure and the underlying genetic mechanism.

(a) Schematic of ESR1-CCDC170 fusion variants and the encoded proteins. CC, Coiled-coil; ABC, ATP-binding cassette; LCC, Low compositional complexity. The specific fusion variants are indicated on the left (i.e., E1/2-E9/10 indicates four fusion variants resulting from fusion of ESR1 exon 1 or 2 to CCDC170 exon 9 or 10). (b) Schematic showing the potential mechanism engaged by ESR1-CCDC170 to endow breast cancer cell survival under endocrine stress and the druggable hypothesis.

Consistent with the behavior of luminal B tumors, our recent data suggests that ESR1-CCDC170 fusions endow ligand-independent growth factor signaling leading to increased cell motility, invasion, anchorage-independent growth and reduced endocrine sensitivity in vitro, as well as enhanced tumor formation in vivo²⁸. Of note, all the ESR1-CCDC170 variants V2–5 share the ability to promote some level of reduced endocrine sensitivity, with the smaller-sized variants (E2-E8 and E2-E10) conferring the most potent effects. Importantly, our most recent study suggests that ESR1-CCDC170 fusions may endow breast cancer cell survival under endocrine stress via physical binding of the fusion to the HER2/HER3/SRC complex and activation of downstream signaling pathways³⁶. HER2/HER3 forms a heterodimer when bound to the HER3 ligand NRG1/2, which phosphorylates the HER3 C-terminal tail thus activating its kinase activity³⁷. HER3 plays a central role in HER2 positive breast cancer and is the most potent activator of AKT/SRC^37,38. SRC is broadly overexpressed in luminal breast cancers³⁹ and can crosstalk with HER2 when facilitated by other molecules, leading to phosphorylation of HER2 at Y877^40–42. Our results suggest that ESR1-CCDC170 may engender a unique exploitable vulnerability in these lethal tumors that can be potentially targeted by HER2/SRC inhibitors.

Our results imply a potential mechanism whereby the N-terminal truncations of CCDC170 delete the CC1 and CC2 domains and expose the CC4 coiled-coil domain via reduction in antiparallel coiled-coil interactions and thus ultimately alters protein interaction and localization properties. Through ATP-driven autonomous homodimerization, ESR1-CCDC170 proteins may facilitate interactions between HER2, HER3, or SRC, leading to uncontrolled activation of their downstream signaling targets (Fig. 1b). This mechanistic model can help explain the following data: 1) Fusion variants V2–5 with CC1/2 domain removal have been associated with worse patient survival and reduced endocrine sensitivity; 2) Increased exposure of the CC4 domain in fusion variants V4–5 confers more pronounced endocrine resistance; 3) Ectopic expression of ESR1-CCDC170 leads to activation of SRC/AKT, which are repressed when endogenous ESR1-CCDC170 is silenced; And 4) HER2/HER3 dimerization prevents receptor internalization and lysosomal degradation⁴³, which may explain the decreased HER2/HER3 protein levels following fusion knockdown⁴⁴. While unregulated activity of growth factor pathways has been linked to breast cancer endocrine resistance, the underlying genetic cause(s) remains ill understood. Here our studies shed light on a potential new mechanism of genetic aberration induced growth factor activity.

To date ESR1-CCDC170 remains the most frequent pathological gene fusion detected in luminal breast cancer and its recurrence has been subsequently supported by several recent studies^26,45–48. In addition to breast cancer, a recent publication reported ESR1-CCDC170 as a recurrent event in ovarian cancer associated with exceptional short-term survival⁴⁹. Even more interesting, an independent report in Science Translational Medicine assessed ESR1 fusions in surgical samples of stage I-III ER positive and HER2 negative breast cancer cases following neoadjuvant letrozole treatment⁴⁷. This study detected ESR1-CCDC170 in three out of 27 endocrine resistant or intermediate resistant tumors (11%) among which two tumors harbored E2-E6 variants and one tumor harbored the E2-E8 variant. The authors also detected an ESR1-CCDC170 fusion in one out of 29 endocrine sensitive tumors. The fusion, however, was an E4-E5 variant encoding a C-terminally truncated ESR1 protein with a disrupted DNA binding domain and thus was likely nonfunctional. This suggests that fusion variants V2 and V4 are associated with lack of response to neoadjuvant letrozole treatment and that truncation of CCDC170 that removes CC1 and CC2 domains is sufficient to confer reduced endocrine sensitivity in patients. Furthermore, another clinical study detected ESR1-CCDC170 in the primary tumors of luminal breast cancer patients that experienced metastatic relapse⁵⁰. The most relevant fusion variants involved ESR1 E2 and CCDC170 E4–10 (V2–5) which were detected in 28 out of 307 patients (9.1%) and resulted in a significantly worse disease-free survival following initial surgery. As all patients included in the study were endocrine naïve, first-line endocrine therapies were administered upon distant relapse. The average progression-free survival for ESR1-CCDC170 positive and negative patients were 9.1 vs 12.0 months following tamoxifen treatment and 10.5 vs 13.8 months following aromatase inhibitor treatment, respectively⁵⁰. These differences, however, did not reach statistical significance due to overall poor response of metastatic tumors to endocrine therapy (83.7% patients progressed within 3 years). This study emphasizes the need for future investigation of fusions in primary breast cancer patients treated with adjuvant endocrine therapy and the necessity for long-term follow-up of these patients to verify the correlation of ESR1-CCDC170 with endocrine response.

The clinical and therapeutic implications of endocrine resistant and metastatic breast cancers harboring ESR1 exon 6 gene fusions

In addition to ESR1-CCDC170 fusions, two recent studies identified pathological recurrent translocation events at ESR1 exon 6 in metastatic breast cancers that result in translatable fusion protein products⁴⁶. These recurrent ESR1 fusions arise through an ESR1 intron 6 intra- or inter-chromosomal translocation breakpoint that is in-frame with a translocated protein partner. The in-frame translocation event enables transcription and subsequent translation to produce a stable ER fusion product. Intron six breakpoints result in a truncated ESR1 transcript (exons 1–6; ESR1ΔCTD) with loss of ESR1 exons 7–9. The fusion protein thus retains the N-terminal domain, the Activation Function-1 (AF1) domain, the nuclear localization sequence and the DNA Binding Domain (DBD). The DBD is necessary for ER interaction with palindromic Estrogen Response Element (ERE) sequences on the genome and regulation of the ER transcriptome. Critically, the truncated ER protein loses its ligand binding domain (LBD) which is responsible for binding estradiol (E2) and subsequent activation of the ER classical pathway. Contrary to the ESR1-CCDC170 fusion variants, ESR1 (exon 6) fusion proteins have a diverse number of 3’ binding protein partners (Fig. 2) each with corresponding translocated and potentially functional domains. The first described in-frame and stable ESR1 (exon 6) gene fusion was discovered in a patient derived xenograft model harboring an ESR1-YAP1 fused transcript (Yes associated protein 1)⁵¹. The ESR1-YAP1 transcript retains YAP1’s transcriptional activation and WW domains. More recently, sequencing of therapy-refractory advanced breast cancer samples identified a similar breakpoint in ESR1 intron 6 to the disable-2 protein (DAB2)⁴⁶ (Fig. 2). Seven additional ESR1 (exon 6) fusions were identified through hybrid-capture-based target sequencing with four fusions discovered in solid tumors (ESR1-SOX9, ESR1-MTHFD1L, ESR1-PLKHG1 and ESR1-TFG) and three identified in circulating tumor DNA (ctDNA) (ESR1-NKAIN2, ESR1-AKAP12 and ESR1-CDK13). In addition, ESR1-GYG1 was identified in a bone metastatic sample⁴⁶. RNAseq analysis of metastatic breast cancer samples discovered two additional stable ESR1 (exon 6) fusion products, ESR1-PCDH11X and ESR1-NOP2. The previously published ESR1-YAP1 fusion was also discovered again in this RNAseq cohort⁴⁸. Discerning the functional input from the 5’ ESR1 and 3’ diverse partners in these fusion proteins is critical to determining ESR1 (exon 6) fusions phenotypes and transcriptomic/cistomic signatures. Unfortunately, no comprehensive study has detected ESR1 (exon 6) fusion proteins. In a cohort of advanced breast cancer patients, a 4.5% rate of ESR1-AKAP12 fusions was reported utilizing RT-PCR⁵⁰, whereas in an evaluation of ER positive breast cancer patients with targeted sequencing, a 1.6% rate was determined for ESR1 fusions with breakpoints at ESR1 intron 6⁵². Although prevalence is likely underestimated due to limitations in detecting ESR1 (exon 6) fusions resulting from diverse 3’ binding partners, new targeted sequencing techniques are being explored taking advantage of the consistent presence of ESR1 exon and intron 6 in these fusions³⁰. Emergence of liquid biopsy methods utilizing ctDNA and tumor-derived exosomes has also provided a non-invasive and feasible mechanism to uncover and track these fusion events longitudinally.

Figure 2. Schematics of ESR1 exon 6 fusion protein structures with diverse 3’ binding partner proteins.

Wildtype ER is shown at the top. Position of the two major hotspot missense mutations in ESR1 Y537 and D538 are shown in red. Example of five ESR1 fusions all occurring in intron 6 and resulting in truncation of ER at exon 6 and fusion to genes each with different protein domains. Activation function (AF), DNA binding domain (DBD), Ligand binding domain (LBD), Phosphotyrosine interaction domain (PID), SRC homology 3 binding domain (SH3), Rsp5 or WWP domain (WW), Nuclear export sequence (NES), LIM zinc-binding domain, transcriptional activation domain (TAD), Proline-, glutamine- and alanine-rich (PQA), Proline-, glutamine- and serine-rich (PQS), Substrate binding region (Sub), Glycogen Synthase 1 interaction region (GYS1).

A biophysical understanding of these ESR1 (exon 6) fusions lacks as their three-dimensional spatial arrangements are yet to be studied. Despite the reoccurring involvement of the AF1 domain in these fusions (Fig. 2), this AF1 region is known to be intrinsically disordered, and the structural knowledge available is somewhat limited. Circular dichroism and two-dimensional ¹H-¹⁵N spectral analyses demonstrated AF1’s overall lack of secondary structures⁵³ (including a shorter version with amino acid truncation⁵⁴). Small-angle x-ray scattering analyses reaffirmed this structural disorder and reported that the AF1 domain is more compact than expected in the three-dimensional space⁵⁵, when compared to other disordered proteins with low-complexity amino acid sequences. Biophysical studies provided a glimpse into AF1’s dynamic ensemble of structures (Fig. 3a) and revealed a previously uncharacterized feature of long-range amino acid interactions (Fig. 3b). In addition, this AF1 domain is known to interact with a general transcription factor TBP (TATA-binding protein)⁵³ or Pin1 (a peptidyl-prolyl cis/trans isomerase; via an AF1 mutant, not the wildtype)⁵⁴. Interestingly, the AF1 remains disordered upon Pin1 binding⁵⁴ or TBP binding⁵⁵. However, the molecular nature and mechanism of the AF1 remains elusive in the context of these endocrine resistant ESR1 fusions, and its interplay with various fusion partners is critical for better understanding of the resistance driven function at the structural level.

Figure 3. Biophysical studies of ESR1-AF1 structures.

(a) The ensemble of AF1 structures determined by integrating the biophysical data of small-angle X-ray scattering (SAXS) and hydroxyl radical protein footprinting (HRPF) with atomic-level molecular dynamics (MD) simulations. A set of 8,491 clusters of structures were determined via K-means clustering with a Ca-RMSD cutoff of 5 Å from a total of 35,240 MD-simulated structures and further optimized simultaneously against experimental SAXS and HRPF data. Shown are the best-fit ensemble structures, represented by a set of top 10 clusters. (b) Interactions of AF1 amino acids determined from the AF1 ensemble structures. Of note, long-range interactions between two segments centering around Ile33 and Ser118 (indicated by arrow) were observed and subsequently validated. Modified from Ref.⁵⁵.

LBD missense mutations occur at ~20–30% frequency in advanced breast cancer and have been reported to drive constitutive ER activity and endocrine resistance⁵⁶. Similarly, ESR1 (exon 6) fusion proteins null of LBD display endocrine therapy resistance and enhanced metastatic properties⁴⁸. ESR1 (exon 6) fusion proteins demonstrated enhanced ER activity through ERE reporter assays in absence of E2 stimulation⁴⁶. Additionally, Hartmaier et al. observed ER hyperactivity in the presence of endocrine therapy in ESR1-DAB2 and ESR1-SOX9 fusions transiently expressed in HEK293T cell lines. Nanostring analysis found upregulation of classical ER responsive genes, GREB1, PGR and TFF1, in ESR1-DAB2 tumor samples⁴⁶. Lei et al. observed similar ligand independent activation of the classical ER transcriptome in ESR1-YAP1 and ESR1-PCDH11X stable T47D cell line models as well as enhanced proliferation phenotypes in the fusion positive cell lines compared to controls⁴⁸. There was also an enhanced epithelial-mesenchymal transition (EMT) transcriptome in these cell line models via RNAseq analysis with supporting findings in in vitro motility assays and in vivo metastatic xenograft models⁴⁸. Together, these data provide strong support of ESR1 (exon 6) pro-cancer functionality through 1) their enrichment in metastatic disease, 2) fusion E2 independent and endocrine resistant growth phenotypes 3) fusion E2-independent transcriptomic regulation and 4) EMT signature upregulation providing a transcriptomic framework for cancer progression.

Endocrine therapies such as selective estrogen receptor modulators (SERMs), selective estrogen receptor degraders (SERDs), and aromatase inhibitors (AIs) target ER’s dependence on E2 for transcriptional activation, cell proliferation and growth. Although these hormonal therapies are successful in a majority of ER positive cases, 20–30% of advanced breast cancer cases will develop resistance to treatment. ER deletion events and LBD missense mutations have served as the primary source of ER endocrine resistant research until recently with the discovery of novel ESR1 (exon 6) fusion proteins^46,48. ESR1 (exon 6) fusions lack of response to endocrine therapies, specifically the SERM tamoxifen and the SERD fulvestrant, suggest that early fusion detection could redirect the therapeutic regimen to better combat disease. Cyclin-dependent kinase (CDK) 4/6 inhibitors, such as Palbociclib, work to inhibit G1-S phase progression. Palbociclib was FDA approved in postmenopausal ER positive HER2 negative breast cancer in combination with an AI, letrozole (PALOMA-1)⁵⁷, or with fulvestrant (PALOMA-3)⁵⁸. Treatment with CDK4/6 inhibitors early in fusion harboring tumors may provide an essential and necessary cancer targeting therapeutic. Monitoring ER positive patients periodically through non-invasive methodologies such as liquid biopsies are also critical for early detection of ER mechanisms of endocrine resistance and, in doing so, will better allow clinicians to efficiently prescribe patient specific treatment regimens.

Clinical implications of luminal breast cancers associated with the RAD51AP1-DYRK4 non-traditional fusion transcript

In addition to ESR1 fusions, our most recent large-scale analysis of TCGA RNAseq data discovered a neoplastic fusion transcript (RAD51AP1-DYRK4) preferentially overexpressed in luminal B breast cancers (7–17.5%), the expression of which tends to be mutually exclusive with ESR1-CCDC170 fusions⁸. The fusion partner genes RAD51AP1 and DYRK4 are co-linearly placed neighboring genes located approximately 2kb apart on the same strand of chromosome 12. No genomic rearrangements are detected at RAD51AP1-DYRK4 loci in the fusion positive tumors. Further, our data suggests that under normal conditions this fusion transcript is exclusively expressed in the testis, whereas under pathological conditions this fusion transcript is ectopically activated in luminal B and metastatic breast cancers⁸. Thus, this fusion is likely a nontraditional fusion transcript generated by cancer-testis specific intergenic splicing. RAD51AP1 is a RAD51-interacting protein specific to vertebrates. Besides its known function in homologous recombination repair^59,60, RAD51AP1 has been found to promote lung cancer metastasis⁶¹ and the growth of cholangiocarcinoma⁶². DYRK4 belongs to a conserved family of serine/threonine kinases⁶³. RAD51AP1-DYRK4 encodes a C-terminally truncated RAD51AP1 protein fused to a frame-shift peptide derived from DYRK4 and resulting in a cytoplasmic-localized chimeric protein that retains the intrinsic disordered (ID) regions but lacks the RAD51-interacting domain (Fig. 4). ID regions are polypeptide segments that lack sufficient hydrophobic amino acids for protein folding and are characterized by structural plasticity⁶⁴. ID regions perform vital functions in cell signaling via facilitating specific protein interactions^65,66, and play key roles in the development of cancer⁶⁶.

Figure 4. Schematic of RAD51AP1-DYRK4 major fusion variants and their encoded proteins identified in breast cancer cell lines and tissues.

Numbers represent different exons in the corresponding wildtype genes or fusion genes. E9-E2 and E8-E2 represent different fusion variants resulted from fusion between exon 9 or exon 2 of RAD51AP1 and exon 2 of DYRK4, respectively. E8s-E2 is the fusion variant resulted from an alternative splicing donor site in the exon 8 of RAD51AP1 spliced to the exon 2 of DYRK4. Open reading frames (ORFs) are depicted in dark shades, with the intrinsic disordered regions highlighted on each of the variant proteins.

Our data suggest that overexpression of RAD51AP1-DYRK4, but not wildtype RAD51AP1, activates MEK/ERK signaling and endows increased aggressiveness in breast cancer cells. MEK/ERK signaling plays key roles in migratory and invasive phenotypes⁶⁷, epithelial-mesenchymal transition⁶⁸, anoiksis resistance⁶⁹, dormancy escape, and distant colonization⁷⁰. RAD51AP1-DYRK4 appears to rewire key oncogenic signaling and induce cancer cell addiction to MEK/ERK signaling through attenuating PI3K/AKT cascades and thus could constitute as an Achilles heel to attack these deadly breast cancers. More importantly, RAD51AP1-DYRK4 may confer a potential therapeutic vulnerability in these lethal tumors by conferring increased sensitivity to the MEK inhibitor trametinib (Mekinist). Trametinib is an FDA approved MEK inhibitor used for treating melanoma with BRAF mutations. Trametinib monotherapy has recently emerged as a new treatment option for recurrent low-grade serous ovarian cancer based upon a phase III study⁷¹. In breast cancer, trametinib has been tested in 33 triple-negative patients albeit with limited efficacy⁷². This phase II trial, however, did not test luminal breast cancer and lacks biomarker guidance to identify responsive patients. Future preclinical studies will be required to provide evidence for personalized trametinib treatment based on RAD51AP1-DYRK positivity. The discovery of RAD51AP1-DYRK reveals a new class of pathological molecular events accountable for the increased aggressiveness and metastasis-prone behavior of lethal breast cancer forms. Such non-traditional fusion events that are silent in human somatic tissues but are overexpressed in certain tumor entities have not been previously identified in breast cancer. We envision that tumor-specific intergenic splicing could play underappreciated roles in the pathobiology of lethal breast cancer subtypes which could provide novel pathological insights and illuminate new therapeutic strategies.

The clinical and pathological implications of BCL2L14-ETV6 associated more aggressive TNBC tumors

Despite the distinctive receptor negative status of TNBC, recent genomic sequencing studies have revealed a paucity of TNBC-specific base pair mutations. In our most recent study, we performed a landscape analysis of recurrent gene fusions in breast cancer based on whole-genome sequencing (WGS) data from the International Cancer Genome Consortium (ICGC) which cataloged a total of 99 recurrent fusions. Interestingly, we found that TNBC tumors appear to harbor more AGRs than other breast tumors and further, we discovered several recurrent fusions that are preferentially detected in TNBC. This suggests that AGR serves as a special type of genomic rearrangement that may be far more frequent than realized in TNBC and has been largely overlooked in previous studies. Such enigmatic rearrangements cannot be detected by conventional cytogenetic approaches or the earlier generation of genomic profiling such as SNP arrays that have limited resolution in detecting exon aberrations and lack detection of balanced rearrangements. Therefore, the resulting aberrant transcripts are usually buried within numerous cis-, trans-, and back-spliced chimeras detected by RNAseq. Thus, AGRs could be a class of genetic alterations in solid tumors that have largely been understudied and could be an overshadowed area of the TNBC-specific genetic landscape that plays a crucial role in TNBC pathobiology.

Intriguing, this analysis revealed a recurrent gene fusion involving the prototype cancer gene ETV6 and its immediate telomeric neighbor gene BCL2L14, which is exclusively detected in 6–12% of TNBC according to WGS data. BCL2L14 encodes a Bcl-2 family protein and functions as a pro-apoptotic factor⁷³. ETV6 encodes a ubiquitously expressed transcriptional repressor that serves as a tumor suppressor unless it forms oncogenic fusions⁷⁴. ETV6 fusions have been intensively studied in leukemia, most notably ETV6-RUNX1, which led to substantially improved patient survival through tailored chemotherapies¹³. Our subsequent clinical sample studies of 134 TNBC tumors and 200 ER positive tumors suggested BCL2L14-ETV6 as a TNBC-specific pathological AGR which is detected in about 4.5% of TNBC tumors. It is notable that the RT-PCR assay we used in this study could underestimate the incidence of BCL2L14-ETV6 in TNBC as the primer set used cannot detect the fusion variants involving exon 1 of variant 1 or exons 1–3 of a longer transcript variant 2 from a distal promoter. This emphasizes the difficulty to amplify BCL2L14-ETV6 variants using primers from distant 5’ exons.

Clinically BCL2L14-ETV6 associated TNBC breast tumors tend to show necrosis areas, a higher tubule formation score, and higher nuclear pleomorphism. Amid TNBC subtypes, BCL2L14-ETV6 is positive in ~19% of these tumors and most frequently detected in the mesenchymal TNBC subtype. Subsequent experimental studies suggest that BCL2L14-ETV6 fusions enhance cell motility and invasiveness, prime EMT, and endow paclitaxel resistance of TNBC⁷⁵. While the fusion variants encode diverse open-reading frames that encode either chimerical proteins or truncated BCL2L14 proteins, these variants appear to induce similar gene expression and phenotypic changes distinctive from wildtype ETV6. This is the first report of a TNBC-specific recurrent gene fusion and the first report of a 3’ ETV6 gene fusion in solid tumors. Future studies of BCL2L14-ETV6 induced EMT and taxane resistance will help establish its role as a genomic determinant of taxane sensitivity and reveal optimized chemotherapy regimens as well as new therapeutic vulnerability of BCL2L14-ETV6 positive tumors.

Fusions associated rare low-grade TNBC entities and the clinical modality for ETV6-NTRK3 fusion associated secretory breast carcinoma

In addition to the recently discovered recurrent gene fusions in major breast cancer subtypes, a subset of low-grade TNBCs are defined by recurrent gene fusions and are pathologically distinctive⁷⁶. Adenoid cystic carcinomas (ACC) of the breast constitute a rare subtype of TNBC characterized by a distinctive histology of abnormal nests encompass or infiltrate breast glandular structures and are relatively clinically indolent⁷⁷. Genetically a vast majority of ACCs (83%) are defined by the MYB-NFIB fusion⁷⁸, and the rest ACC tumors are associated with MYBL1 rearrangements or MYB amplification, all of which lead to activation of MYB or MYBL1⁷⁹. Assays detecting MYB-NFIB fusion such as MYB labeling by immunohistochemistry may serve as powerful diagnostic tool for ACC⁸⁰ as in salivary gland neoplasms⁸¹.

Besides the MYB-NFIB fusion, the ETV6-NTRK4 gene fusion first reported in 2002 is now established as a hallmark pathognomonic genetic aberration that defines a rare breast cancer subtype called secretory breast carcinoma (SBC)^82,83. SBC is a form of infiltrating ductal carcinoma (IDC) accounting for about 0.15% of all breast cancer^84,85. SBC is known to affect both children and adults of both sexes and is usually attributed to the basal subtype of breast cancer⁸⁵. SBC presents with distinct histological and molecular features characterized by the presence of secretory material within or outside of the tumor cells⁸⁶. ETV6-NTRK3 t(12;15) is the result of a balanced chromosome rearrangement. This fusion generates an in-frame chimeric oncoprotein, connecting the Helix-Loop-Helix (HLH) dimerization domain of the ETV6 protein to the PTK domain of the NTRK3 protein^83,87. The HLH dimerization domain is thought to drive ligand-independent dimerization of the fusion protein and subsequent constitutive activation of the PTK domain⁸⁸. This fusion protein demonstrates potent oncogenic transforming activities through dysregulating the MAPK and PI3K-AKT signaling pathways, both of which are major effector pathways of wildtype NTRK3 that play important roles in the proliferation and survival of breast cells^88,89.

Several types of NTRK inhibitors (NTRKi) have shown impressive efficacy in treating ETV6-NTRK3 positive SBC. Among these, Larotrectinib is an orally available highly selective NTRKi that acts against NTRK1, NTRK2 and NTRK3 proteins⁹⁰. Larotrectinib acts by antagonizing the activities of NTRK tyrosine kinases thereby causing cancer cell death and growth inhibition⁹¹. Larotrectinib was granted accelerated approval by the FDA in 2018 as the first tumor-agnostic therapy for pediatric and adult tumors harboring NTRK gene fusions. Larotrectinib has shown striking clinical outcomes in the treatment of SBC patients harboring ETV6-NTRK3 fusions with a response rate of 80%^91–94. Remarkably, patients were reported to respond within the first two cycles of treatment with Larotrectinib, with rapid resolution of cancer-related symptoms and tolerable adverse events⁹⁴. Entrectinib, another NTRKi which is potent but less selective than Larotrectinib targets the fusion proteins involving NTRK, c-ros oncogene 1 (ROS1) or anaplastic lymphoma kinase (ALK). The results from three global phase I/II Entrectinib trials on breast cancer revealed that the response rate for NTRK fusion-positive breast cancer was 83.3%. Overall, Entrectinib is well-tolerated with durable response in NTRK fusion-positive breast cancer⁹⁵, which suggests the potential use of this drug in managing SBC patients with the ETV6-NTRK3 fusion protein in the future.

In addition to SBC, a study that involved 4,854 patients with breast cancer identified a total of 4 patients with NTRK1 fusions-positive metastatic disease. This includes the LMNA-NTRK1 fusion, ZBTB7B-NTRK1 fusion and SCP2-NTRK1 fusion⁹⁶. Patients with LMNA-NTRK1 fusions demonstrated positive clinical response upon treatment with Larotrectinib. Treatment with Larotrectinib diminished the growth advantage and eliminated the endocrine therapy resistance observed in breast cancer cells, proving the oncogenic potential of LMNA-NTRK1 fusion gene⁹⁶. Considering the effectiveness of Larotrectinib and Entrectinib in treating NTRK-positive breast cancer, identification of more NTRK fusion-driven breast cancer will be helpful in improving the prognosis of such patient groups.

The therapeutic implication of NOTCH/MAST kinase gene fusion associated breast carcinomas

Gene fusions involving the NOTCH or microtubule-associated serine-threonine (MAST) family members (NOTCH/MAST) were also reported to be recurrent events in breast cancer. In a study by Dan Robinson and colleagues, the presence of fusions affecting MAST family members, typically MAST1 and MAST2 genes, were found in a subset of around 3–5% of breast cancer while fusions involving NOTCH1 or NOTCH2 gene were found to be exclusive in ER negative breast cancer. The fusion partners of the NOTCH/MAST family kinase genes are promiscuous, involving the fusion of different exons of NOTCH/MAST genes with different partner genes, specifically at the 5’ end for the MAST1 or MAST2 gene and either at the 5’ or 3’ end for the NOTCH1 or NOTCH2 gene⁹⁷. Overexpression of the MAST1 or MAST2 fusion gene with the MAST genes retaining intact PDZ and 3’ kinase-like domains confer growth and proliferation advantage in benign breast epithelial cells, whereas knockdown of the ARID1A-MAST2 fusion in MDA-MB-468 cells leads to reduction of growth in both the cell lines and mouse xenograft models⁹⁷. The ER negative cell lines that harbor endogenous NOTCH fusion genes (SEC16A-NOTCH1, SEC22B-NOTCH2 and an intragenic NOTCH1 fusion) demonstrated increased NOTCH responsive transcriptional activity relative to other cell lines and increased expression of NOTCH target genes⁹⁷. Inhibition of the NOTCH pathway using DAPT, a γ-secretase inhibitor, repressed NOTCH reporter activity and proliferation of HCC2218 and HCC1599 cells, respectively. The HCC1599 cells xenograft mouse model also showed reduction in tumor volume following treatment with DAPT⁹⁷. Recently, a pan-NOTCH γ-secretase inhibitor (AL101) was reported to show potent therapeutic effect on TNBC patient derived xenografts (PDX) models that harbor NOTCH gene fusions⁹⁸. This highlights the potential of developing AL101 into a form of targeted therapy for TNBC positive NOTCH fusions. These data suggest the ability of NOTCH fusion genes in activating NOTCH-responsive genes and the potential therapeutic value of NOTCH inhibitors in fusion positive breast cancer patients.

The therapeutic significance of FGFR and other actionable kinase gene fusions in breast cancer

Several recent reports also presented findings of kinase fusions in breast cancer which are drawing increasing attention due to the availability of actionable drugs. A recent study on kinase gene fusions in a cohort of 4,854 breast cancer patients revealed enrichment of actionable kinase fusions in metastatic breast cancers negative for ESR1 mutations (0.6%)⁹⁶. A total of 27 patients were detected to harbor actionable kinase fusions, including FGFR1, FGFR2, FGFR3, BRAF, NTRK1, RET, ROS1, ALK, ERBB2, and MET. FGFR family kinases include four highly conserved receptor tyrosine kinases, with FGFR4 being the least involved in gene fusions^96,99–101. The FGFR fusions detected in breast cancer have highly promiscuous fusion partners, but the majority involve FGFR proteins as the 5’ partner and retain most of their functional domains^96,101. It is possible that some of these fusions may enable ligand-independent dimerization of FGFR as reported in other cancers^102–104. One such example is the FGFR3-TACC3 fusion identified in various types of cancer in which the dimerization of FGFR was made possible through the coiled-coil domain of TACC3¹⁰⁵. The FGFR inhibitors being investigated in breast cancer that harbor FGFR amplifications, mutations, or gene fusions include Erdafitinib¹⁰⁴, AZD4547¹⁰⁶, PD173074^101,107 and Infigratinib^108,109.

In addition to FGFR fusions, a limited number of studies reported the presence of BRAF gene fusions in metastatic breast cancer patients, with KIAA1549 as the most common fusion partner at the 5’ end^96,110. Furthermore, gene fusions involving the RET gene were found in less than 1% of the patients in a cohort of 9,693 breast cancer cases, with RET kinase domain-coding exons remaining intact in all the cases. Among the RET fusions genes, CCDC6-RET, NCOA4-RET, RASGEF1A-RET were found to be oncogenic, with the latter two fusions driving oncogenicity through MAPK and PI3K pathways¹¹¹. The use of Cabozantinib, a nonselective multi-kinase inhibitor on a breast cancer patient with a tumor harboring a NCOA4-RET fusion revealed rapid radiographic and clinical response. This suggests the potential use of this drug in managing breast cancer patients with RET fusions¹¹¹. Other kinase fusions that involve ROS1, ALK, ERBB2 and MET are very rare in breast cancer. Of note, functional studies of the EML4-ALK fusion detected in a ESR1 wildtype metastatic breast cancer patient showed that this fusion gene confers hormone-independent growth advantage and resistance to endocrine therapy in T47D breast cancer cells. Combination treatment of Ceritinib/Fulvestrant diminished the endocrine therapy resistance observed in the cancer cells⁹⁶, suggesting their potential to treat breast cancer patients harboring ALK fusions. The recurrent gene fusions discussed above and selected private actionable gene fusions identified in breast cancer are summarized in Table 1.

Table 1: Summary table of selected fusion genes in breast cancer.

BrCa, breast cancer; LumB, Luminal B; NOS, Not otherwise specified.

Gene fusion	5’ gene	3’ gene	BrCa Subtype	Frequency	Clinical Implication	References
ESR1-CCDC170	ESR1	CCDC170	LumB, Endocrine resistant BrCa	6% – 8% LumB, 9% of luminal BrCa experiencing distant relapse	Endows endocrine resistance, confers sensitivity to HER2/SRC inhibitors	28,47,50,112
ESR1 exon 6 fusions	ESR1 exon 6	Promiscuous	Metastatic breast cancer	4.5%	Potentially manageable with CDK4/6 inhibitors	30,46,48,57,58
RAD51AP1-DYRK4	RAD51AP1	DYRK4	LumB	7% – 17.5% of LumB	Confers sensitivity to the MEK inhibitor trametinib	8
BCL2L14-ETV6	BCL2L14	ETV6	More aggressive TNBC tumors with necrotic and mesenchymal features	4.4% – 12.2% of TNBC	Primes EMT and endows resistance to paclitaxel	75
ETV6-NTRK3	ETV6	NTRK3	Secretory breast carcinoma (SBC)	92% of SBC	80% and 83.3% response rate to Larotrectinib and Entrectinib, respectively	83,92–95
MAST/NOTCH family fusions	Promiscuous	MAST1 or MAST2		3% – 5%		97,98
	NOCTH1 / NOTCH2 / promiscuous genes	NOCTH1 / NOTCH2 / promiscuous genes	ER negative	3 in 11 TNBC tumors	Potentially treatable with pan-Notch gamma secretase inhibitor AL101
FGFR Family fusions	FGFR2	Promiscuous	Metastatic breast cancer without ESR1 mutations	0.2% – 2.5%	Potentially manageable with pan-FGFR inhibitor or combination therapy that includes pan-FGFR inhibitor, CDK6/4 inhibitor and ER down-regulator	96,101,106,108,109,113
	FGFR3	TACC or TACC3
	Promiscuous	FGFR1
BRAF fusions	Promiscuous	BRAF	Metastatic breast cancer, IDC, NOS	<1%	Potentially treatable with MEK inhibitor Trametinib	110,114–117
MYB-NFIB	MYB	NFIB	Adenoid cystic carcinoma	83%	Defines a pathological distinct rare subtype of TNBC	78,80,81

Conclusion and perspectives

With the advancement of next generation sequencing (NGS) technology, an increasing number of pathological recurrent and kinase gene fusions are being identified in breast cancer. The identification of these gene fusions as a mutation that contributes to the aggressiveness and therapy-resistant nature of breast cancer could illuminate new therapeutic and/or patient management strategies. From a diagnostic perspective, these recurrent gene fusions could serve as genetic biomarkers to distinguish more aggressive and lethal breast cancers that require intensive individualized treatment with robust clinical follow-up. Although the therapeutic avenue for breast cancer patients positive for recurrent gene fusions remains limited, emerging evidence has suggested potential therapeutic vulnerabilities of a subset of the fusion-associated breast carcinomas, as exemplified by the carcinomas associated with actionable kinase gene fusions or associated with ESR1-CCDC170 and RAD51AP1-DYRK4 fusions. In addition, future studies of chemosensitivity of fusion associated breast carcinomas could highlight tailored chemotherapy regimens as in the case of ETV6-RUNX1 associated acute lymphoblastic leukemia¹³. To translate our fundamental knowledge about fusion-associated breast carcinomas into clinical applications, larger-scale clinical studies will be required to understand fusion pathological characteristics, clinical behaviors, predictive and prognostic associations, and chemosensitivities. In addition, future mechanistic studies will be required to pinpoint the pathobiological mechanisms driven by these gene fusions and to generate druggable hypotheses, in the hope to devise new clinical management and therapeutic strategies.