Context.—Molecular diagnostics play a role in the management of many cancers, including breast cancer.
Objective.—To provide an update on molecular testing in current clinical practice, targeted at practicing pathologists who are not breast cancer specialists.
Data Sources.—This study is a narrative literature review.
Conclusions.—In addition to routine hormone (estrogen and progesterone) receptor testing, new and recurrent tumors are tested for HER2 amplification by in situ hybridization or overexpression by immunohistochemistry. Intrinsic subtyping of tumors represents a fundamental advance in our understanding of breast cancer biology, but currently it has an indirect role in patient management. Clinical next-generation sequencing (tumor profiling) is increasingly used to identify potentially actionable mutations in tumor tissue. Multianalyte assays with algorithmic analysis, including MammaPrint, Oncotype DX, and Prosigna, play a larger role in breast cancer than in many other malignancies. Given that a proportion of breast cancers are familial, testing of nontumor tissue for cancer predisposition mutations also plays a role in breast cancer care.
In this review, we survey the landscape of breast cancer through the lens of molecular techniques used for diagnosis and treatment. By any measure, this disease is both common and lethal. It is the most commonly diagnosed cancer in women, excluding nonmelanoma skin cancer, and is also a common cause of cancer death, second only to lung cancer.1 It is estimated that 231 840 new invasive breast cancer cases and 40 290 deaths occurred in the United States in 2015.1 In the United States, 1 woman in 8 can expect to receive a diagnosis of breast cancer. Because this disease is commonly seen in most surgical pathology laboratories, it is essential for the pathologist to have a working knowledge of molecular techniques used in routine practice. It is equally valuable to have some knowledge of emerging techniques that, while promising, have not yet entered common use.
This review is focused on invasive mammary carcinoma, which represents most breast cancer. Not covered are benign tumors, carcinoma in situ, sarcomas of the breast, and other miscellaneous entities.
HORMONE RECEPTOR TESTING (RECEPTORS FOR ESTROGEN AND PROGESTERONE)
Basis of Test
Although the diagnosis of breast cancer is primarily made histologically, ancillary testing supports diagnosis, classification, prognosis, and prediction of response to therapy. Several well-established nonmolecular tests are part of the standard of care at first diagnosis of breast cancer.2 Hormone receptor status is determined by the tumor cells' expression of nuclear receptors for estrogen (ER) and progesterone (PR; Table). Tumor cells expressing these receptors may be stimulated by circulating estrogen or progesterone. Inhibiting the stimulatory effect of steroid hormones is an important component of therapy for hormone receptor–positive (HR+) breast cancer.
Biochemical ligand-binding assays were initially used to detect ER and PR,3 but they were cumbersome and could not be performed on routine formalin-fixed, paraffin-embedded tissue. Immunohistochemical assays have become routine for this purpose.4–6 In the clinically validated Allred scoring method, the proportion (on a scale from 0 to 5) and intensity (from 0 to 3) of staining are summed to give an overall score. Cutoffs for ER and PR positivity (proportion + intensity >2) were determined from clinical data recorded in patients receiving endocrine therapy.7,8
Hormone receptor status predicts response to endocrine therapy including tamoxifen9 and aromatase inhibitors.10 National Comprehensive Cancer Network (NCCN) guidelines recommend that endocrine therapy be considered for any patient with HR+ breast cancer. For this reason, all new and recurrent breast cancers must be tested for ER and PR expression.5
ERBB2 (HER2) TESTING
Basis of Test
The erb-b2 receptor tyrosine kinase 2 (ERBB2, also known as HER2) gene is overexpressed in 20% to 30% of breast cancer cases,11,12 most often because of copy-number variation (amplification). The HER2 protein is analogous to other receptor tyrosine kinases: it is expressed at the cell surface; it exists there in homodimeric form, as well as in heterodimers with other ErbB family members; and it possesses intracellular domains capable of transphosphorylation and interaction with downstream effectors.13 It has no known ligand, and therefore is hypothesized to serve a regulatory role through its heterodimerization with other ErbB family members.14
A committee convened by the American Society of Clinical Oncology (ASCO) and the College of American Pathologists (CAP) released in 2007 a consensus statement, subsequently updated in 2011 and 2013, to provide guidance on HER2 testing in breast cancer.15,16 Key aspects of the guideline include a recommendation that all primary, recurrent, and metastatic breast cancers be tested for HER2, either by immunohistochemistry (IHC; to detect overexpression) or fluorescence in situ hybridization (FISH; to detect amplification), using a validated test (Table). US Food and Drug Administration (FDA)–approved assays are available for both of these modalities, and both are accepted as first-line testing. Both IHC and FISH have the potential to return a positive, negative, or equivocal result. Equivocal cases are recommended to undergo reflex testing by the other modality (eg, equivocal IHC cases are reflexed to FISH). In a laboratory using FISH for first-line testing, reflex HER2 IHC provided a clear positive or negative result in only 5 of 17 FISH-equivocal cases (29%).17 In contrast, in a laboratory using first-line IHC, reflex FISH for 2+ IHC yielded a clear amplified or nonamplified result in 43 of 50 cases (86%).18 It should be recognized that both IHC and FISH represent an attempt to convert a continuous biologic variable into a dichotomous category, an exercise that is clearly dependent on reproducible application of a validated cutoff. In particular, the precise definition of an equivocal result has a pronounced effect upon the use and operating characteristics of reflex testing.19
HER2 testing by FISH uses probes directed at the HER2 gene, generally reported as a ratio normalized to chromosome 17 centromeric probes (HER2:CEP17 ratio). A ratio less than 2.0 with average HER2 copy number of fewer than 4.0 signals per cell indicates no HER2 amplification. A ratio of 2.0 or greater regardless of HER2 copy number, or a ratio of less than 2.0 with 6.0 or more HER2 signals per cell, indicates amplification. Cases with a ratio less than 2.0 and with 4.0 to 6.0 signals per cell are reported as equivocal.15 Polysomy 17, or at least an increase in CEP17 count, is present in some cases and may result in a nonamplified ratio despite presence of an increased HER2 gene dosage. The significance of this latter finding has not been well established.20
HER2 testing by IHC is reported as 0, 1+, 2+, or 3+. The result is 3+ (positive) as defined by ASCO/CAP if more than 30% of invasive tumor cells show uniform, intense, thick circumferential membrane staining. The result is 1+ in the presence of weak, incomplete membrane staining in any fraction of tumor cells, or weak, complete membrane staining in less than 10% of cells. The 2+ (equivocal) cases are those that fall between 1+ and 3+. As a memory aid, these may be thought of as failing to reach 3+ on any single criterion (ie, staining is not thick, not intense, not circumferential, or not present in more than 30% of cells, provided that it is present in at least 10% of cells). Given that these categories are very clearly defined by ASCO/CAP, it is recommended that pathologists practice with explicit reference to the guideline.15
HER2 amplification was shown to be an independent risk factor for recurrence and death in studies conducted before the advent of targeted therapy.12,21 At that time, HER2 status was investigated as a predictor of response to various nontargeted adjuvant chemotherapy regimens.22,23 Today, it has gained particular importance because of the availability of trastuzumab, a humanized monoclonal antibody against HER2.24–26 Trastuzumab improves progression-free survival and overall survival, as part of adjuvant treatment for both early high-risk disease27,28 and metastatic disease.29 Pertuzumab, a second-generation monoclonal antibody, and lapatinib, a small-molecule kinase inhibitor, are also FDA approved as HER2-targeted therapies for breast cancer.
What of tumors that are negative for ER expression, PR expression, and HER2 overexpression? These triple-negative cancers cannot be treated by currently approved targeted therapies—although targets of an experimental nature may be present—and cytotoxic chemotherapy is usually pursued. Triple-negative tumors have a less favorable prognosis than other subtypes, attributable in part to lack of molecular targets.
KI-67 PROLIFERATION INDEX
The Ki-67 proliferation index has been investigated as a breast cancer predictive and/or prognostic factor in various settings.30 The Ki-67 antigen is expressed in the nuclei of cells that are in cycle (ie, not in G0 phase), and therefore reflects cell proliferation. Although Ki-67 can be measured by immunohistochemistry, its analytic validity has not been well established, and formal interlaboratory standardization is not in place.30 The NCCN Breast Cancer Guidelines do not currently recommend Ki-67 in routine clinical workup.2
INTRINSIC SUBTYPES AND THEIR SURROGATES
Basis of Test
Classic studies in the late 1990s revealed that breast cancers can be classified by gene expression profiling into at least 4 “intrinsic subtypes” (luminal A, luminal B, HER2-enriched, and basal-like).31,32 It has been suggested that these subtypes form a spectrum that recapitulates the differentiation of breast ductal cells, from progenitors (basal-like) to an intermediate state (HER2-enriched) to mature luminal cells (luminal A and B).33 A “normal breast–like” subtype has been identified, but its existence has been questioned on the grounds that many of the normal breast–like tumors have been paucicellular and therefore have consisted mainly of nontumor tissue,33 which of course, upon expression profiling, resembled normal breast. Rare intrinsic subtypes include claudin-low, with low expression of cell adhesion genes,34 and molecular apocrine-like,35 which often expresses androgen receptors.36
Although the discovery of the intrinsic subtypes has increased our understanding of breast cancer, subtyping by expression profiling was clinically unavailable until the advent of the Prosigna assay (NanoString Technologies, Seattle, Washington), described later in this review. However, HR and HER2 status can be used as a surrogate for intrinsic subtyping by molecular methods. Tumors that express either ER or PR (so-called hormone receptor–positive cases) are likely to be luminal subtype. HR+/HER2− status correlates closely with luminal A type, whereas HR+/HER2+ correlates with luminal B. HR−/HER2+ cases correspond to the HER2-enriched intrinsic type. The triple-negative HR−/HER2− cases are predominantly basal-like type, but transcriptomic studies have revealed that some of these tumors have a luminal gene expression profile, whereas others represent uncommon intrinsic subtypes.37
These subtypes can be conceptualized as different disease entities with discrete epidemiologic and clinical correlations. The subtype identity of a given tumor persists through treatment31 and correlates with clinical outcome.38 Luminal subtype is associated with benefit from tamoxifen treatment.39 Nonluminal (HER2-enriched and basal-like) subtypes were associated with pathologic complete response to anthracycline and taxane chemotherapy40 in a retrospective series. Subtype can also be correlated with genetic background: germ line breast cancer 1 (BRCA1)–mutated tumors are generally of basal-like subtype,41 and sporadic basal-like tumors may have loss of BRCA1 function due to mutation, epigenetic events (promoter methylation), or other regulatory mechanisms.42,43
NEXT-GENERATION SEQUENCING OF BREAST CANCER TISSUE
Basis of Test
High-throughput “next-generation” sequencing (NGS) has made it possible to sequence large numbers of genes to identify actionable mutations in the tumor (Table).44 Clinically actionable mutations are those that are either prognostic (correlated with behavior of the neoplasm irrespective of any specific therapy) or predictive (correlated with response to a therapeutic agent, whether targeted or nontargeted). Of note, somatic mutations may be pathogenic (“driver” mutations) without being specifically actionable at the present time.
Tests to simultaneously sequence multiple genes by NGS (“gene panel testing”) are offered by many laboratories. Notable advantages of the NGS panel sequencing approach, compared with single-gene testing, include the ability to cover a large number of genes on a single platform, without multiple procurements of tissue from the paraffin block. Next-generation sequencing has the potential to detect, within the reportable range of a single test, all 4 canonic classes of genetic variation: single-nucleotide variants, insertions/deletions, copy-number variants, and structural variants (rearrangements, translocations).45,46 As the cost of performing NGS does not increase in direct proportion to the number of genes sequenced, NGS can be both cost- and time-effective compared with single-gene testing.
The diagnostic yield of breast NGS has been studied but is difficult to define and measure. Sequencing of 46 cancer-related genes in 415 breast cancer samples (including some primary-metastatic pairs) revealed somatic nonsynonymous mutations in 220 of 354 patients (62.1%). A total of 13 of 61 pairs (21%) demonstrated additional mutations in the metastasis, suggesting a possible incremental utility for testing metastases.47 These authors did not comment on the actionability of the identified mutations. In a series of metaplastic breast cancers, sequencing a panel of 236 cancer-related genes revealed alterations that were considered potentially targetable in 19 of 20 cases (95%),48 but this included many alterations, such as tumor protein p53 (TP53) mutations, that would be difficult to target in clinical practice at the present time. A systematic multinational study of NGS profiling in breast cancer is underway.49 A survival benefit associated with undergoing and acting on the testing has not yet been shown in breast cancer specifically. If done (as it has been for non–small cell lung cancer50 and a cohort of mixed tumor types51), this would strongly support clinical use of NGS tumor profiling.
Various scenarios arising in the interpretation of cancer genomic data may challenge the clinical genomicist as well as the treating physician. Mutations at a rarely mutated amino acid position may not be as well studied as more common mutations in the same gene. Genotype-phenotype correlations established in one disease may not necessarily be able to be directly extrapolated to other diseases.52 The medical literature may hold tantalizing preclinical data suggesting actionability of a given mutation, without the evidence being sufficient to guide clinical practice. When only tumor DNA is sequenced, it is usually impossible (except by inference) to determine whether a given variant is somatic or present in the germ line, which may give rise to ethical questions surrounding return of results. Sequencing paired nontumor DNA resolves the question but increases the cost of performing the assay. Given the complex issues involved in interpreting these data sets, some institutions have adopted a “molecular tumor board” approach to facilitate multidisciplinary decision-making.53–55
Although tumor profiling may reveal intriguing molecular targets for specific breast cancers and for breast cancer in general, current standards of care are not predicated on the results of NGS assays.2 In particular, NGS has a limited role in the management of HR+ and/or HER2-amplified cancers (together representing 80% of breast cancer cases56), because the expression of ER, PR, or HER2 drives the selection of adjuvant therapy. Next-generation sequencing is more accurately viewed as a means for matching tumors with novel therapies and/or clinical trials, particularly in triple-negative tumors, which are not amenable to hormonal or anti-HER2 therapy, have a less favorable prognosis,57 and are a genetically heterogeneous group.58
Novel and investigational therapies that may be considered based on NGS results include poly(ADP-ribose) polymerase (PARP) inhibitors for breast tumors with BRCA mutations. These mutations impair DNA repair by homologous recombination. PARP inhibition is hypothesized to act by inducing single-strand breaks that, if unrepaired, lead to double-strand breaks, which the cell is unable to repair.59–61 This treatment strategy is FDA approved for advanced ovarian cancer in the setting of germ line BRCA mutation.62 Other sporadic mutations that may also indicate “BRCAness” include the Fanconi anemia genes (RAD51 recombinase [RAD51], MRE11 homolog A double-strand break repair nuclease [MRE11], nibrin [NBN], partner and localizer of BRCA2 [PALB2], RAD50 double-strand break repair protein [RAD50], BRCA1-associated RING domain 1 [BARD1], BRCA1-interacting protein C-terminal helicase 1 [BRIP1], Fanconi anemia complementation group D2 [FANCD2], cyclin-dependent kinase 12 [CDK12]).63 Genetic alterations implicated in BRCAness include not only point mutations in the above-mentioned entities, but also copy-number variation (amplification) of a BRCA2-interacting transcriptional repressor (EMSY).64 Epigenetic alterations also play a role: another Fanconi gene, Fanconi anemia complementation group F (FANCF), can occasionally be inactivated by somatic methylation.65 Likewise, the level of expression of an unmutated gene may play a role: BRCAness has been associated with increased expression of the BRCA1 repressor, inhibitor of DNA-binding 4, HLH protein (ID4).43
The PI3K/AKT/mTOR pathway is activated in as many as 10% of basal-like breast cancers, chiefly due to mutations in phosphoinositide-3-kinase regulatory subunit 1 (PIK3R1), phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit alpha (PIK3CA), phosphatase and tensin homolog (PTEN), and v-akt murine thymoma viral oncogene homolog 1 (AKT1).66 The fraction of luminal and HER2-enriched tumors that activate this pathway is even higher. Agents targeting this pathway, for example by inhibiting the pro-survival serine/threonine kinase AKT, have been investigated.58
In triple-negative breast cancers with amplification of the fibroblast growth factor receptor genes FGFR1 or FGFR2,66,67 FGFR inhibition has been investigated, including by multikinase inhibitors and by more selective agents.58
Triple-negative breast cancer can harbor mutations and amplifications in notch (NOTCH1, NOTCH2, and NOTCH3). Mutations have been noted specifically within the proline–glutamic acid–serine–threonine (PEST) domain, which functions as a negative regulator.68 Similar mutations are implicated in leukemogenesis.69 Pharmacologic inhibition of gamma-secretase, which cleaves Notch to release an intracellular transcriptional activator domain, is under investigation.58
An emerging variation on the concept of HER2 testing in breast cancer is the fact that some tumors lacking HER2 amplification have been found to harbor activating HER2 mutations, which can be identified by NGS. In vitro, many of these mutations are activating and are amenable to inhibition by the small molecule lapatinib or by the irreversible kinase inhibitor neratinib.70 An in vivo benefit of neratinib has been reported in a patient with HER2-mutated breast cancer.71
Tumor profiling at present is normally done on tumor tissue alone, usually with some enrichment of the sample for tumor cells but without any comparison to a paired nontumor sample.44 Without paired nontumor sequencing, it is typically not possible to exclude that any given tumor mutation is in fact present in the germ line. Testing of nontumor tissue may be considered in appropriate cases (eg, in patients with a history suspicious for inherited cancer). However, inherited cancer predisposition is rare, even in a cancer population. It would be unwieldy (not to mention distressing to patients) to insist on reflex germ line testing for every somatic mutation. Mutations identified in tumor DNA should not be taken as prima facie evidence of a germ line mutation, and the risk that this misinterpretation will be made is one reason why testing should be done in association with genetic counseling. In the future, paired nontumor tissue sequencing may become more common and may help in addressing this issue.
In summary, tumor profiling has become clinically available for detection of somatic mutations in breast cancer. These assays use NGS to detect potentially actionable variants in a set of cancer-related genes, and they may be of greatest use in triple-negative tumors. At present, many of the known actionable variants are of an investigational nature. It is clearly possible to validate NGS assays for clinical use.72–76 Unresolved questions continue to surround patient selection, time point (new diagnosis versus recurrence), tissue selection (archival primary tumor versus metastasis), reporting, and, crucially for the development of the field, clinical utility and reimbursement.77
MULTIANALYTE ASSAYS WITH ALGORITHMIC ANALYSIS: MAMMAPRINT
Basis of Test
More than other areas of oncology, breast cancer management has begun to employ multianalyte assays with algorithmic analysis (MAAAs) in routine clinical care.78,79 An MAAA is an in vitro diagnostic test that combines measurements of multiple genes or other analytes to derive predictive or prognostic information. The measurements are integrated using a specific, closed-form, often proprietary algorithm to yield a clinically validated result. Multianalyte assays with algorithmic analysis are also referred to in the literature as gene expression profiles because they are largely based on measuring mRNA levels for selected genes. Several of these tests have been developed for use in breast cancer,80 including MammaPrint (Agendia, Amsterdam, the Netherlands, and Irvine, California), Oncotype DX (Genomic Health, Redwood City, California), and Prosigna (Table).
The MammaPrint assay was cleared by the FDA in 2007 and was initially limited by its requirement for fresh tissue, but it is now validated and cleared for formalin-fixed, paraffin-embedded tissue. MammaPrint measures the expression of 70 genes using a microarray platform, and reports a binary result (low risk or high risk) for recurrence without adjuvant chemotherapy at 10 years. This information is intended to spare patients at low risk of recurrence from receiving adjuvant chemotherapy, with its attendant morbidity. It is not intended to predict the response, per se, to chemotherapy; rather, via its prognostic significance, it helps to select patients who are likely to benefit from chemotherapy. MammaPrint is not intended to affect the decision for adjuvant endocrine therapy.
The 70 genes included in the assay were selected by supervised classification of genes whose expression differed in good-prognosis and poor-prognosis tumors, among a group of 78 sporadic tumors.81 This process was agnostic to the function of the genes. To perform the assay, a test tumor's gene expression profile is compared to the average good-prognosis profile, and a correlation coefficient is calculated. This coefficient is compared to a validated threshold to place the tumor into the good- or poor-prognosis group. This threshold was optimized to minimize misclassifications, and in particular to minimize misclassification of poor-prognosis tumors into the good-prognosis group.81
The assay has now been validated in both lymph node–positive and lymph node–negative patients, and in both ER+ and ER− tumors.82–86 The prospective RASTER trial showed that MammaPrint testing could spare from chemotherapy 94 of 295 patients (32%) for whom chemotherapy would have been recommended by a standard (non–gene-based) decision tool.86 Another analysis showed that MammaPrint low-risk versus high-risk status predicted distant metastasis-free survival at 10 years for T1a–b cancer, with a hazard ratio of 3.45 (95% confidence interval [CI], 1.04–11.50).85 In these and in other studies, the assay has been clinically validated. Further study is likely to determine whether the effect size is sufficient to justify routine adoption.
MULTIANALYTE ASSAYS WITH ALGORITHMIC ANALYSIS: ONCOTYPE DX BREAST CANCER ASSAY
Basis of Test
The Oncotype DX breast cancer assay is a MAAA for prognostication based on reverse transcriptase–polymerase chain reaction analysis of 21 genes. Initially validated for ER+, HER2−, node-negative invasive cancer,87 the validation has now been extended to node-positive disease (1–3 nodes) as well.88,89 In contrast to MammaPrint, which has received 510(k) clearance from the FDA, Oncotype DX has been marketed under the form of a laboratory-developed test (“homebrew assay”) for which FDA clearance is not required.
The Oncotype DX breast cancer assay was constructed by hand-selecting 250 candidate genes, then identifying those with prognostic significance in 3 independent studies.90 The 16 cancer-related genes included in the final assay fall into several groups: a proliferation group, an ER-related group, an invasion-related group, and several miscellaneous genes. Five housekeeping genes are used for normalization. A formula is used to calculate a recurrence score between 0 and 100, which is further trichotomized into low, medium, or high risk.90
The initial use for Oncotype DX was to predict risk of recurrence at 10 years in node-negative patients, in whom the estimated relative risk in high-risk versus low-risk cases was 4.49 (95% CI, 3.38–5.60).90 It has also been clinically validated to predict the benefit of adding chemotherapy with cyclophosphamide, doxorubicin, and fluorouracil88—or cyclophosphamide, methotrexate, and fluorouracil87,91—to tamoxifen. The NCCN guidelines explicitly recommend use of Oncotype DX in specific scenarios, ostensibly in recognition of the robust validation that has been performed.
In addition to the Oncotype DX assay for invasive cancer, a 12-gene subset of the test is offered to predict recurrence in patients who have had a local excision (partial mastectomy) for ductal carcinoma in situ. The ductal carcinoma in situ algorithm uses a separate set of coefficients to calculate a recurrence score that has been validated both in patients who did and did not receive adjuvant tamoxifen.92,93 The recurrence score is further segmented into low-, intermediate- and high-risk groups to facilitate clinical decision-making.
MULTIANALYTE ASSAYS WITH ALGORITHMIC ANALYSIS: PROSIGNA BREAST CANCER ASSAY
Basis of Test
The Prosigna breast cancer assay, like MammaPrint and Oncotype DX, is based on tumor gene expression, but differs in that it represents a clinical implementation of the intrinsic subtype concept. The intrinsic subtypes were identified by measuring the expression of thousands of genes by microarray and performing clustering analysis. Although well suited to discovery, this method was not readily amenable to implementation as a clinical test. Thus, a key step in the development of the Prosigna assay was to design a classifier able to reproducibly identify the intrinsic subtype of a tumor based on the expression of a smaller set of genes: this is the Prediction Analysis of Microarray-50 (PAM50) classifier.94 A continuous risk of recurrence score, calculated from the PAM50 output and from clinical data, provides prognostic information and predicts the effectiveness of chemotherapy.94,95
As developed for clinical use in the FDA-cleared Prosigna Breast Cancer Prognostic Gene Signature Assay, the expression of the fifty PAM50 genes and 8 housekeeping genes is measured on RNA extracted from formalin-fixed, paraffin-embedded tissue on the NanoString nCounter Analysis System.96 This “digital” nucleic acid profiling platform measures mRNA levels without an amplification step, thus reducing technical complexity and bias. The technology is based on use of 2 sequence-specific probes (“capture” and “reporter,” each 35–50 nucleotides in length) that jointly hybridize with each target molecule. The reporter probes are labeled with fluorophores whose linear order, determined by high-resolution imaging, creates a color code that reveals the identity of each captured target.97
In clinical validation studies, the Prosigna risk of recurrence score compared favorably with the Oncotype DX assay as well as with an IHC-based prognostic model and clinical predictor.98,99 Prosigna placed fewer patients into the intermediate-risk group than did Oncotype DX, potentially facilitating treatment decisions.98,100 The assay is currently validated for postmenopausal women with HR+, node-negative (stage I or II) or node-positive (stage II) disease.101 Unlike some other MAAAs, Prosigna is not performed in a single, centralized laboratory,95 but instead is performed in multiple reference laboratories, all using the same platform, probe library, and algorithms. This approach may have benefits from the standpoint of turnaround time, proficiency testing, and other regulatory concerns.
The regulatory and payment environment surrounding MAAAs is rapidly evolving. The assays challenge conventional definitions of laboratory testing. Because the result of the assay is a computed score, and the individual analytes may not be reported separately, the Centers for Medicare & Medicaid Services (CMS) initially took the position that a MAAA does not indicate the “presence or absence of a substance or organism in the body,” and therefore is not a clinical laboratory test.102 CMS has nonetheless shown some willingness to reimburse these assays, first under miscellaneous Current Procedural Terminology (CPT) codes,103 and more recently under discrete CPT codes,104 for which the reimbursement has variably been set by the gapfill process and by crosswalking. Gapfilling is a value-based process for setting reimbursement for novel procedures ab initio, whereas crosswalking attempts to set a fair reimbursement by comparison to other previously reimbursed codes. The details of the method used have a dramatic effect on the eventual reimbursement. Private payers have looked to CMS for guidance, lending particular importance to the outcome of CMS's process.
GERM LINE TESTING IN BREAST CANCER
Basis of Test
A fraction of breast cancer cases are caused by inherited (germ line) mutations. Although breast cancer predisposition is in most cases a complex trait, attributable to variants in multiple genes,105 approximately 10% of breast cancer cases show Mendelian inheritance and are attributed to mutations in a single gene.106,107
Most familial breast cancer occurs in the setting of the hereditary breast-ovarian cancer syndrome (caused by mutation in BRCA1 or BRCA2), accounting for approximately 90% of kindreds for which a causative mutation is able to be identified. Hereditary breast-ovarian cancer syndrome is relatively common among cancer predisposition syndromes. In a study of unselected patients (ie, individuals without cancer), 120 of 5331 Ashkenazi Jews (2%) were carriers of a BRCA1/BRCA2 mutation.108 The carrier rate is lower in the general US population—1 of 345 individuals (0.3%)109—although this estimate should be interpreted with caution, given the prevailing mixed genetic background. The carrier rate is much higher in patients with breast cancer, and even higher in those with a suspicious family history.107
Other inherited syndromes in which breast cancer is a major manifestation include Li-Fraumeni syndrome (TP53 mutation), Cowden syndrome (PTEN mutation), Peutz-Jeghers syndrome (serine/threonine kinase 11 [STK11] mutation), hereditary diffuse gastric cancer (cadherin 1 [CDH1] mutation), and ataxia-telangiectasia (ATM serine/threonine kinase [ATM] mutation). A fraction of breast cancer kindreds carry germ line mutations in Fanconi anemia genes,110 and patients with biallelic germ line BRCA2 mutation present with Fanconi anemia.111 Cancer predisposition is typically inherited in an autosomal dominant manner with variable penetrance, with tumors showing somatic loss of the second allele.
Assessment for hereditary cancer predisposition is part of the clinical management of breast cancer (Table). Particular emphasis is placed on the evaluation of patients with certain risk factors: early age (≤50 years) at onset of breast cancer, multiple primaries, male breast cancer, clustering of multiple syndromically related tumor types within the pedigree, high-risk ethnicity, and family members of individuals with known breast cancer susceptibility mutations.112 In patients meeting defined criteria, early referral to a cancer genetics professional is recommended, with consideration for eventual genetic testing. Individuals with appropriate expertise for pretest and posttest counseling include genetic counselors, medical geneticists, oncologists, surgeons, oncology nurses, and others.112
The discovery of a germ line mutation may prompt increased screening or treatment of the proband or her relatives for breast cancer and for other conditions associated with the mutation. For example, a proband with breast cancer who tests positive for BRCA1/BRCA2 mutation may choose to undergo risk-reducing salpingo-oophorectomy or risk-reducing contralateral mastectomy.113,114 Decision analysis shows that this specific intervention results in increased life expectancy.115 Relatives of the proband may choose to undergo targeted testing for a known familial germline mutation. Of note, the impact of genetic testing can potentially propagate across the pedigree, with dramatic effects on reproductive or lifestyle decisions, and appropriate genetic counseling is therefore essential.
Germ line testing is performed by sequencing DNA from nontumor tissue, often peripheral blood. The sequencing can be carried out by Sanger sequencing, covering individual genomic regions of up to approximately 1 kb, or by NGS, covering multiple regions on a single platform by obtaining numerous overlapping short reads. The genes associated with hereditary breast cancer are predominantly classical tumor suppressors, in which mutations at numerous amino acid positions can cause the loss of protein function and a subsequent cancer predisposition phenotype. This scenario contrasts with the “hot spot” phenomenon seen with classical oncogenes, in which the activating mutations may cluster at a few known positions (eg, BRAF p.V600E). As a result, comprehensively sequencing the entire coding region of the gene of interest, rather than genotyping selected positions, is recommended to ensure sufficient clinical sensitivity of testing for germ line breast cancer predisposition. One exception to this principle is that patients from populations with a known founder effect, such as those of Ashkenazi, Icelandic, Swedish, Hungarian, and Dutch extraction, may be candidates for genotyping of founder-specific positions.112
In patients in whom the clinical presentation is highly suggestive of a single specific syndrome (eg, Cowden syndrome in a patient with breast cancer, macrocephaly, and thyroid cancer), testing can be focused on the single most likely causative gene (in this case, PTEN). In some cases the pathologic features of the tumor may suggest its genetic basis (eg, lobular histology in CDH1-related cancer116 ; medullary histology or triple-negative in BRCA1-related cancer117,118 ; ER+ for BRCA2-related cancer118). However, the syndromes that include breast cancer can have overlapping presentations. When, as is often the case, the clinical presentation does not meet criteria for any single syndrome, sequencing of multiple genes must be considered.112 One approach is to initially test for the most likely genes, with reflex testing for less likely genes in case of a negative result. Next-generation sequencing has, however, made practical a “panel” approach to testing in which multiple genes are sequenced on the same platform and as part of a single test. Panel testing has the advantage of a simplified workflow and has the potential to reduce bias and increase discovery (of unexpected genes and unexpected types of variation).119 As the cost of NGS testing scales less than linearly with the number of genes, the NGS approach becomes more cost-effective as the number of genes requiring testing increases. Because, however, most hereditary breast cancer cases are attributed to only 2 genes (BRCA1/BRCA2), the relative cost-effectiveness of NGS versus single-gene testing remains in flux.
Germ line testing reports enumerate the variants identified and estimate the likelihood that they are clinically significant. Recommendations for reporting of germ line testing have been issued by the American College of Medical Genetics and Genomics.120 Variants that have been documented to cause a disorder are reported as “pathogenic,” whereas those that have been documented to be neutral are reported as “benign.” Variants with lower degrees of evidence are reported as “likely pathogenic” and “likely benign.” Forms of evidence that can be adduced include population data, pedigree (segregation) analysis, evolutionary data, computational predictions, and functional observations from clinical or laboratory material,120 which may be incorporated into a multifactorial model.
When available evidence is insufficient to ascertain the pathogenicity of a variant, it is reported as a variant of unknown significance. They are individually rare but collectively numerous (2.1 per patient in one report of a 42-gene panel, with at least 1 identified in 154 of 175 patients [88%]121), and by definition, existing knowledge is insufficient to fully guide their interpretation. By their nature, they may be frustrating to patients and physicians,122 although most are likely to be benign. Patients overestimate the significance of variants of unknown significance123 and in some cases have elected surgical management out of proportion to the significance of the identified variant.124 Importantly, testing for a larger number of genes increases the patient's exposure to variants of unknown significance. Thus, germ line variants of unknown significance represent a major challenge in genetic counseling and clinical management,125 and limit the utility of multigene germ line testing.112
As more data are collected, pathologists' and clinicians' ability to interpret germ line variants in breast cancer–related genes is likely to improve. Interpretation of BRCA1 and BRCA2 mutations is somewhat limited by the unique history of BRCA testing in the United States. Until 2013, this test was offered in a single laboratory under patent protection.126 This laboratory amassed during the course of 20 years a rich, proprietary database of variants with correlative clinical data, making it possible to determine which variants were associated with disease.127 In Association for Molecular Pathology vs. Myriad Genetics, the Supreme Court of the United States ruled that naturally occurring DNA sequences cannot be patented.126,128 That decision opened the door to BRCA1/BRCA2 testing in numerous laboratories, but it did not place the variant database into the public domain. Collaborative platforms have now been established to aid in the interpretation of variants in BRCA1/BRCA2 and other genes associated with hereditary breast cancer,129,130 but even with data sharing, consistent interpretation remains a challenge.131
Ancillary workup of invasive breast cancer includes nonmolecular testing as well as tests based on the assessment of nucleic acids. Nonmolecular testing includes HRs and HER2 status, which can also be assessed at the DNA level via FISH. Next-generation sequencing to identify actionable mutations is an emerging area whose role has not yet been defined. Predictive testing via MAAA has become common and is recommended in specific scenarios to support treatment decisions. Germ line testing for inherited cancer predisposition holds scientific challenges as well as ethical, legal, and societal ones. Optimal breast cancer care requires that both pathologists and clinicians maintain a working knowledge of all of these rapidly evolving techniques.
The author has no relevant financial interest in the products or companies described in this article.
Presented at the 2nd Princeton Integrated Pathology Symposium: Breast Pathology; February 8, 2015; Plainsboro, New Jersey.