M o l e c u l a r C l a s s i f i c a t i o n of

B re a s t Ca n c e r
Elena Provenzano, MBBS, PhD, FRCPatha,b, Gary A. Ulaner, MD, PhDc,d,
Suet-Feung Chin, BSc, PhDe,*

KEYWORDS
 Breast cancer  Molecular classification  Genes  Mutations  Copy number alterations

KEY POINTS
 Breast cancer is a heterogeneous disease.
 It can be classified based on its molecular profiles.
 These molecular subtypes have different prognostic indices and may require different clinical
management.

INTRODUCTION of genes in a single experiment. Recent advances,

for example, next-generation sequencing allowed
Breast cancer is not a single entity but a heteroge- detection of mutations, chromosomal rearrange-
neous group of diseases with highly variable clin- ments, and copy number alterations across the
ical behavior1. Pathologists have long recognized entire genome, including those only present within
this diversity at the morphologic level, and it is re- minor subclones of tumor cells.3,4 These high-
flected in the various special histologic types of throughput (HT) technologies have changed how
breast cancer with their distinct microscopic ap- we conceptualize and classify breast cancer as
pearances and associated clinical outcomes. well as provide a new level of insight into the
However, 70% to 80% of breast cancers fall into complexity of genetic changes and the existence
the ductal/no-special-type category (invasive of intratumoural heterogeneity.5–7
ductal carcinoma [IDC]), which, rather than repre- Historically, patient management decisions
senting a unique entity, show marked heterogene- have been based on traditional histologic features,
ity with respect to tumor morphology, underlying including tumor size, histologic grade, lymph node
molecular biology, and prognosis.2 status, and hormone and human epidermal growth
Cancer is driven by DNA alterations, including factor receptor 2 (HER2) receptor status in
chromosomal rearrangements, mutations, and conjunction with patient characteristics, such as
epigenetic changes, such as promoter hyper- age.8,9 These variables show a strong association
methylation resulting in activation of growth- with survival outcomes but, even when combined
promoting genes (oncogenes) or suppression of in algorithms, such as Nottingham Prognostic In-
growth-inhibiting genes (tumor suppressor genes). dex, Predict, or Adjuvant! Online, are a crude mea-
The advent of array-based technologies enabled sure of risk in individual patients.10–15 Accurate
quantification of DNA copy number changes and prediction of tumor behavior is key to oncological
global expression profiling of tens of thousands decision-making, avoiding overtreatment with

a
Cambridge Experimental Cancer Medicine Centre (ECMR), NIHR Cambridge Biomedical Research Centre,
Cambridge University Hospitals NHS Foundation Trust, Cambridge CB2 0QQ, UK; b Department of Histopathol-
ogy, Addenbrookes Hospital, Box 235, Hills Road, Cambridge CB2 0QQ, UK; c Department of Radiology, Me-
morial Sloan Kettering Cancer Center, 1275 York Avenue, Box 77, New York, NY 10065, USA; d Department
pet.theclinics.com

of Radiology, Weill Cornell Medical School, New York, NY 10065, USA; e Cancer Research UK Cambridge Insti-
tute, University of Cambridge, Li Ka Shing Centre, Robinson Way, Cambridge CB2 0RE, UK
* Corresponding author.
E-mail address: [email protected]

PET Clin 13 (2018) 325–338

https://doi.org/10.1016/j.cpet.2018.02.004
1556-8598/18/Ó 2018 Elsevier Inc. All rights reserved.
326 Provenzano et al

harmful drugs in patients with a good prognosis, been dropped, as it is thought to represent
and more aggressive intervention with first-line contamination by normal glands. Classification of
chemotherapy in patients with a poor prog- cohorts of breast cancers into the intrinsic sub-
nosis.16–18 However, use of algorithms based on types seems robust across studies; however,
these histologic variables result in a significant assignment of individual tumors to a subgroup
number of patients being overtreated, with as shows only moderate reproducibility depending
many as 85% of patients deriving no benefit from on the array platform used, composition of the
chemotherapy. At the other extreme, 20% of pa- entire tumor population, and setting of gene
tients still die despite receiving maximum expression thresholds.24–26 Identification of the
therapy.19 basal-like group is most reproducible, with the
The ultimate goal of modern oncological man- luminal B and HER2 groups the most poorly repro-
agement is personalized medicine, with a more ducible.26,27 The commercially available Predic-
precise determination of patient prognosis based tion Analysis of Microarray 50(PAM50) classifier
on tumor biology and the opportunity for targeted (Prosigna),28,29 based on expression of 50 genes
treatment directed at the underlying molecular that can separate tumors into the intrinsic sub-
aberrations driving individual tumor growth. HT types, has been shown to be an independent
molecular techniques offer the potential to revolu- marker of prognosis.30–33 Attempts to replicate
tionize patient management in this way. But these these groups using IHC-based panels, including
techniques are currently expensive compared with ER, PR, HER2, Ki67, and basal cytokeratins,
standard methods, such as immunohistochem- have produced modest concordance between
istry (IHC); the vast amounts of data generated gene expression and IHC-defined intrinsic sub-
require complex bioinformatic analyses limiting types at best.22,34
their clinical use currently.20
Luminal Breast Cancer
Intrinsic Subtypes
Luminal breast cancers are enriched for ER-
The mainstays of breast cancer characterization positive tumors and include special type cancers,
are still histologic subtype, tumor grade, and such as tubular, cribriform, lobular, and mucinous
stage, which provide a basic reflection of the de- carcinomas. Luminal cancers form a continuous
gree of tumor differentiation (tubule formation) spectrum that can be arbitrarily divided into 2
and growth rate (size and mitotic count). The sem- subgroups based on the expression of
inal article that led to the identification of 5 intrinsic proliferation-related genes. Luminal A tumors are
subtypes was published by Perou and colleagues5 typically low grade with an excellent prognosis,
in 2000. The investigators took a series of 38 inva- ER/PR positive and HER2 negative, with high
sive breast cancers (36 ductal and 2 lobular), 1 expression of ER-related genes and low expres-
case of ductal carcinoma in situ (DCIS), and 4 sion of proliferation-related genes.23,35 In contrast,
benign samples and undertook complementary luminal B tumors are higher grade with worse
DNA microarray gene expression analysis fol- prognosis and may be PR negative and/or HER2
lowed by hierarchical clustering of differentially positive with high expression of proliferation-
expressed genes and identified 5 subtypes pri- related genes.36,37 Clinically, the luminal A group
marily separated by estrogen receptor (ER) is likely to benefit from hormonal therapy alone,
expression; 2 ER-positive luminal subtypes, and whereas luminal B tumors with their increased pro-
3 ER-negative subtypes (HER2 enriched, basal liferation may be candidates for chemotherapy.
and normal-like). A follow-up study showed that Molecular signatures that separate ER-positive
these subtypes were associated with differences tumors into good and poor prognosis subgroups
in survival.21 form the basis for many of the multi-gene assays
These 5 intrinsic subtypes have been validated that are currently available for clinical use,
in other series and have changed how we think such as Oncotype Dx, Mammaprint, and
about the taxonomy of breast cancer.22,23 The EndoPredict.38–40 Although there is little overlap
separation into good and poor prognosis in the specific genes that make up these signa-
ER-positive, HER2-positive, and triple-negative tures, they all include genes involved in prolifera-
(ER, progesterone receptor [PR], and HER2 nega- tion and ER signaling.27,41 In studies whereby
tive: triple-negative breast cancer [TNBC]) groups multiple signatures are applied to the same patient
is highly clinically relevant, given that current ther- cohort, they all identify low- and high-risk groups
apeutic regimens are centered on antiestrogen with a significantly different prognosis; however,
therapy, chemotherapy, and HER2-targeted there is disagreement at the individual patient level
agents. The normal-like subtype has subsequently in many cases.42–46
 Molecular Classification of Breast Cancer 327

Because of the present high cost of these com- 7% basal-like, and 18% normal-like. Of note, the
mercial multi-gene assays, attempts have been HER2-enriched tumors had a significantly higher
made to recapitulate the luminal A and B subtypes pathologic complete response (pCR) rate versus
using IHC markers of proliferation, such as Ki67. HER2-luminal tumors (53% vs 29% respectively)
Cheang and colleagues37 looked at a series of 357 and had a larger improvement in event-free sur-
cancers with known gene expression profiling and vival with the addition of trastuzumab.50
identified an optimum Ki67 cut point of 13.25% for
distinguishing luminal A from B tumors, with a sensi-
Basal-Like/Triple-Negative Breast Cancers
tivity of 72% and a specificity of 77%. This cut point
was rounded to 14% for clinical use. Hence, the sur- The basal-like breast cancers are typically high-
rogate IHC definition of luminal A tumors was ER grade TNBC that are characterized by upregula-
positive and HER2 negative with a Ki67 index less tion of genes expressed by basal/myoepithelial
than 14% and of luminal B tumors was ER positive cells, including high-molecular-weight cytokera-
and HER2 negative with a Ki67 index of greater tins (CK5 and 14), P-cadherin, and epidermal
than 14% or HER2 positive. In 2 large clinical trial us- growth factor receptor.23 The basal-like group
ing this IHC surrogate definition, 81% to 85% of includes diverse histologic types of breast cancer;
luminal A tumors were correctly classified, whereas although most are high-grade IDC, they also
35% to 52% of luminal B tumors were misclassified include medullarylike cancers with a prominent
as IHC luminal A.35 The IHC definition of luminal A lymphocytic infiltrate; metaplastic cancers, which
tumors was subsequently modified to include PR may show squamous or spindle cell differentiation;
expression of greater than 20% to better identify a and rare special type cancers like adenoid
good prognosis subgroup akin to the luminal A tu- cystic carcinoma (AdCC), which carry a good
mors defined by gene expression profiling. Despite prognosis.51–55
the initial enthusiasm, Ki67 is not currently accepted Breast cancers arising in BRCA1 mutation car-
as a routine prognostic marker in breast cancer by riers are typically basal-like; BRCA1 dysfunction
the American Society of Clinical Oncology because mediated by alternative mechanisms, such as
of disagreement as to the best scoring method and methylation, has been identified in non-BRCA1
suboptimal reproducibility in scoring.47 mutated basal-like tumors.51,56,57 The terms
basal-like and TNBC have been used interchange-
ably; however, not all TNBC are of the basal type.
Human Epidermal Growth Factor Receptor
On gene expression profiling, TNBCs have been
2–Positive Breast Cancer
subdivided into 6 subgroups with different molec-
ER-negative cancers comprise biologically distinct ular drivers and variable clinical outcomes and
entities with different drivers that can be divided response to neoadjuvant chemotherapy, although
into 2 main groups: HER2 enriched and basal- this has subsequently been revised to 4
like/TNBC. The HER2-enriched group is driven groups.58–60
by overexpression of HER2 and genes associated An important subgroup is the luminal androgen
with related pathways or with the HER2 amplicon receptor (LAR) group, characterized by high
on chromosome 17q12. Before the introduction expression of androgen receptor and hormonally
of HER2-targeted therapies, they had the worst regulated pathways with similarities to the molec-
prognosis of all breast cancer subtypes. Although ular apocrine group of breast cancers.61 They have
most tumors within this subgroup (>80%) show a relatively good prognosis and show a lower pCR
HER2 gene amplification or HER2 protein over- rate following neoadjuvant chemotherapy (10%)
expression on IHC, not all clinically defined more akin to ER-positive tumors.60 As well as
HER2-positive tumors fall into this subgroup; expressing androgen receptor, LAR tumors
many ER-positive/HER2-positive tumors fall into frequently have PIK3CA mutations and may be
the luminal B group mentioned earlier.23,48 In one candidates for antiandrogenic agents and/or
series looking at clinically HER2-positive breast phosphatidylinositide 3-kinase (P13K) pathway
cancers, 47% were of the HER2-enriched sub- inhibitors.62
type, whereas conversely only 65% of HER2- There are 2 basal-like subgroups, BL1 and BL2,
enriched subtype cancers were clinically HER2 enriched for genes involved in proliferation, DNA
positive.49 In a retrospective analysis of the damage response (BL1), and growth factor recep-
NeoAdjuvant Herceptin (NOAH) clinical trial look- tor signaling pathways (BL2). The BL1 group
ing at neoadjuvant therapy with or without trastu- shows high pCR rates following neoadjuvant
zumab for clinically HER2-positive breast cancer, chemotherapy (52%), whereas the BL2 group
on PAM50 testing only 55% of the cancers were shows poor response.60 The immune modulatory
classed as HER2 enriched; 21% were luminal, group is enriched for genes involved in immune
328 Provenzano et al

cell processes, such as B- and T-cell receptor these special-type cancers are described in
signaling, cytokine signaling, and antigen presen- more detail later.
tation, and is now thought to reflect infiltration by
immune cells rather than represent a true TNBC
Invasive Lobular Carcinoma
subtype.59 The mesenchymal (M) group shows
enrichment for genes involved in cell motility, dif- The most common special-type cancer is invasive
ferentiation, growth signaling pathways, and lobular carcinoma (ILC), accounting for 5% to 15%
extracellular matrix interactions, with low expres- of breast cancers (Fig. 1).2 ILC is composed of
sion of proliferation-related genes and high small, monotonous cells with round nuclei that
expression of genes associated with stem cells. lack cohesion, infiltrating as cords and single files
The last group was associated with worse 5-year around existing normal breast structures. There is
distant recurrence-free survival consistent with often minimal stromal reaction; this absence of
upregulation of pathways involved in motility and disruption to background breast architecture is
metastasis; the M group also showed the poorest what results in the difficulties in radiological detec-
disease-free survival and overall survival following tion of these tumors, with up to 43% being occult
neoadjuvant chemotherapy.60 on standard mammography.69 ILCs are typically
TNBCs demonstrate particularly elevated up- ER positive (95%–99%) and HER2 negative
take of fluorodeoxyglucose (FDG) on PET.63,64 (>95%), and most fall into luminal subtypes with
They are also known to result in early metastatic 30% to 85% being luminal A and 20% to 64%
disease and have a propensity for extraskeletal luminal B.70 A small subset show more aggressive
metastases,65 increasing the importance of imag- histologic features, such as a solid growth pattern
ing for systemic staging. and/or severe nuclear pleomorphism. The latter
group are referred to as pleomorphic ILC and
Integrative Clusters can have a more aggressive biological phenotype
with ER negativity in up to 25% and HER2 gene
Breast cancer is a copy number disease.66
amplification in 15% to 35%.71,72
Recently, a classifier based on integrated analysis
The characteristic molecular feature of ILC is
of both genomic and transcriptomic data from
downregulation of the CDH1 gene on 16q that en-
2000 primary breast tumors MolEcular TAxonomy
codes the E-cadherin protein. E-cadherin is a cell
of BReast cancer International Consortium
adhesion molecule that forms part of the
(METABRIC) was produced, which revealed 10
cadherin-catenin complex and is responsible for
groups termed integrative clusters (IntClusts) with
the glandular architecture of the breast.73 How-
a distinct clinical outcome67 further refining the
ever, 12% to 16% of classic ILC retain expression
existing expression-based subgroups (Table 1).
of the E-cadherin protein on IHC and 25% to 50%
IntClust 5 is composed of HER2-positive tumors;
of IDCs may show reduced or absent E-cadherin
TNBCs predominantly fall in IntClust 10 with
expression, so ILC is defined by morphology and
some in IntClust 4, which is characterized by im-
not by absent E-cadherin staining alone.74–76
mune infiltration. The remaining 7 IntClusts are
Clinically, ILC have a similar prognosis to stage-
predominantly composed of ER-positive/HER2-
matched IDC in most series, although long-term
negative cancers highlighting the heterogeneity
prognosis may be poorer with higher rates of late
present within this subtype includes an ER-
relapse after 10 years.77 Grade is a significant
positive group (IntClust2) with a prognosis worse
prognostic factor, with grade 3 tumors showing a
than ER-negative disease characterized by
worse outcome. ILC show a poor response to neo-
CCND1 amplification.
adjuvant chemotherapy, with low pCR rates and a
lower rate of tumor downsizing and conversion to
Special Histologic Subtypes
breast-conservation therapy.78,79 On Oncotype
There are more than 20 special histologic sub- Dx testing, most fall into the low- (32%–77%)
types of breast cancer, each associated with a and intermediate-risk (23%–57%) groups, with
distinct histologic appearance and clinical only 1% to 10% having a high-risk recurrence
behavior.2 New genomic techniques have led to score, consistent with the limited benefit of
considerable interest in interrogating the molecu- chemotherapy in these patients.80
lar profiles of these special-type tumors to help un- Historically, ILC and IDC were thought to
derstand them.68 Although some special-type develop via separate pathways, with ILC origi-
tumors have been associated with characteristic nating in the lobules and IDC originating from
genetic changes, there is also considerable ge- ducts; it is now known that most breast cancers
netic heterogeneity within many special-type can- arise in the terminal duct lobular units, which is
cers despite their distinct morphology. Some of also the location of the stem cell niche.81 ILC and
 Molecular Classification of Breast Cancer 329

Table 1
Description of the 10 integrative cluster subtypes

Distinguishing Histologic ER/HER2 (Intrinsic

IntClust Frequency (%) Prognosis Molecular Features Features Subtype)
1 7 Intermediate 17q23 amplification High grade, NST 85% ER1
GATA3 mutations 15% HER21
High genomic instability (Lum B)
2 4 Poor 11q13–14 amplification — 94% ER1
(CCND1) 8% HER21
High genomic instability (Lum B)
3 15 Good Low genomic instability Low grade, 94% ER1
with few copy tubular, lobular, 0% HER21
number changes mixed NST/ (Lum A)
PIK3CA and CDH1 special types
mutations
4 17 Good Low genomic instability Non–grade 3, 70% ER1
Upregulation immune lymphocytic 8% HER2 1
response genes infiltrate
5 10 Poor ERBB2 amplification Grade 3 51% ER1
85% HER21
(HER2E)
6 4 Intermediate 8p12 amplification Non–low grade 98% ER1
(ZNF703) 4% HER21
High genomic instability (Lum B)
7 10 Good 16p gain, 16q loss, 8q Non–grade 3 97% ER1
amplification 0.5% HER21
MAP3K1 mutations (Lum A)
8 15 Good 1q gain, 16q loss Low grade 96% ER1
PIK3CA and GATA3 1% HER21
mutations (Lum A)
9 7 Intermediate 8q gain, 20q Grade 3 87% ER1
amplification 12% HER21
High genomic instability (Lum B)
TP53 mutations
10 11 Poor 5q loss, 8q gain, 10p Grade 3, 14% ER1
gain, 12p gain medullarylike 3% HER21
Impaired DNA (basal)
checkpoint regulation
TP53 mutations
Abbreviations: Lum, luminal; NST, no special type.

low-grade ER-positive IDC share many genetic Gene expression profiling of ILC has identified 2
similarities, including gain on chromosome 1q subtypes: one is hormone related with upregula-
and 16p and loss of 16q, and are now thought to tion of ESR1 and cell cycle and ER target genes
share common precursor lesions, such as atypical and the other is immune related with activation of
lobular hyperplasia, atypical ductal hyperplasia, immune signaling and cytokine pathways and
LCIS, and low-grade DCIS, that together form associated with an increased lymphocytic infiltrate
the low-grade ER-positive breast neoplasia on histology.84 In the METABRIC study, ILC fell
pathway.82 In a comparison with grade-matched across several intrinsic clusters, with approxi-
ER-positive IDC, ILC showed differential ex- mately one-third in IntClust 3 associated with
pression of CDH1 and other genes involved in CDH1 and PIK3CA mutations85,86; in fact, it was
cell assembly and cell-cell interactions as well as the increased number of ILC in IntClust 3 that
reduced expression of cell cycle genes.83 largely accounted for the CDH1 mutations87. The
Other genes commonly mutated in ILC include second largest group fell in IntClust 4, with low
PIK3CA (35%–48%), PTEN (14%), and ERBB2 genome instability and increased expression of
(4%–18%).70 immune signatures. Given the generally poor
330 Provenzano et al

Fig. 1. Invasive lobular carcinoma (ILC). (A) Screening mammogram from a 69-year-old woman showing a 20-mm
area of architectural distortion in the upper outer quadrant. (B) On MR imaging, the tumor was much larger than
apparent on the mammogram. (C) Core biopsy revealed ILC, with the characteristic single-file growth pattern and
minimal stromal reaction (H&E, original magnification x 10). (D) Higher power showing cells with regular round
nuclei and occasional mucin vacuoles, which demonstrates the potential limitations, including undermeasure-
ment of the malignancy, when imaging ILC by mammography (H&E, original magnification x 20).

response to traditional chemotherapy, these ge- The lobular subtype has important implications
netic profiles may offer new treatment opportu- for imaging studies (Fig. 1). Primary ILC lesions
nities, such as PI3K pathway and immune of the breast are more difficult to detect than
checkpoint inhibitors. ductal/IDC on all current imaging modalities,
 Molecular Classification of Breast Cancer 331

including mammography, ultrasound, magnetic increase in expression of ER and PI3K signaling

resonance, and FDG PET.88–93 On FDG PET, ILC pathways.103
demonstrates lower FDG avidity than comparable Within the METABRIC study, tubular cancers fell
IDC malignancies in both primary88,89,91,92 and across several good prognosis subtypes, indicating
metastatic lesions.94,95 In addition, ILC differs in that despite their distinct morphology and clinical
its patterns of metastatic spread, including a pro- behavior they have diverse molecular drivers. The
pensity for infiltrative growth along the peritoneum, largest group fell into IntClust 3 and harbored
retroperitoneum, gastrointestinal and genitouri- PIK3CA mutations, although as expected they
nary viscera, and the leptomeninges.96–99 This lacked the CDH1 mutations also seen in this cluster,
infiltrative growth often makes detection of meta- whereas others fell into IntClusts 4, 7, and 8.87
static lesions more difficult on imaging95 (Figs. 2
and 3). Unique molecular alterations in ILC may Mucinous Carcinoma
make this breast cancer subtype more amenable
Invasive mucinous carcinomas are composed of
to imaging with novel PET radiotracers targeting
islands of tumor cells floating in pools of extracel-
specific metabolic pathways.100,101
lular mucin.2 Mucinous carcinomas can be difficult
to diagnose, as the lack of an associated stromal
Tubular Carcinoma
response can give an indeterminate appearance
Invasive tubular carcinoma is an excellent prog- on mammography and ultrasound104 (Fig. 4). Clin-
nosis special-type cancer that is often screen ically, pure mucinous carcinomas tend to occur in
detected.102 As the name suggests, it is character- older patients and have a favorable prognosis.
ized by prominent tubule formation (>90%) and is They are almost always ER positive and HER2
almost universally ER positive and HER2 negative negative and fall into the luminal subtypes with
of the luminal A subtype. They typically have the most being luminal A.105,106 At the genomic level,
1q gain and 16q loss seen in members of the mucinous carcinomas are distinct from IDC, with
low-grade neoplasia family. On gene expression low genomic instability and absence of the 1q
profiling, tubular cancers cluster together with gain and 16q loss seen in other low-grade ER-pos-
matched low-grade ER-positive IDC; however, itive tumors and a relative paucity of PIK3CA
on hierarchical analysis, they show a relative mutations.106,107

Fig. 2. (A) FDG PET maximum-intensity-projection image of a patient with newly diagnosed lobular breast cancer
demonstrates physiologic FDG avidity, without abnormal foci. (B) Sagittal computed tomography (CT) on bone
window demonstrates widespread sclerotic osseous metastases (arrows). Biopsy-confirmed osseous metastases.
(C) Fused FDG PET/CT demonstrates the widespread osseous metastases are not appreciably FDG avid. This finding
demonstrates the potential for substantial ILC malignancy with little or no FDG avidity on PET.
332 Provenzano et al

Fig. 3. (A) FDG PET maximum-intensity-projection image of a patient with newly diagnosed lobular breast cancer
demonstrates physiologic FDG avidity, without abnormal foci. (B) Axial PET. (C) Axial CT. (D) Axial fused FDG
PET/CT demonstrates gastric wall thickening (arrow), peritoneal free fluid (arrowhead), and hydronephrosis
(curved arrow), better appreciated on CT than FDG PET. Endoscopy and biopsy confirmed ILC metastases to
the stomach. Hydronephrosis is probably from ILC metastases to the retroperitoneum obstructing the ureters.
This image is an example of widespread, multi-organ system ILC metastases with little or no FDG avidity.

Within the METABRIC cohort, mucinous carci- grade variants of spindle cell metaplastic carci-
nomas fell within the ER-positive IntClusts but noma (fibromatosislike metaplastic carcinoma),
were relatively evenly distributed across clusters low-grade adenosquamous carcinoma, and rare
and did not associate with any one cluster; hence, salivary glandlike tumors that occur in the breast,
despite their unique morphology, as a group including AdCC and secretory carcinoma, both
mucinous carcinomas show considerable hetero- characterized by the presence of specific fusion
geneity in terms of underlying genetic drivers.87 genes.53,55,108,109
Basal-like breast cancers are typically associ-
ated with aggressive clinical behavior and a poor
Adenoid Cystic Carcinoma
outcome, but there is a subset of special-type can-
cers that fall within the basal group but has an AdCCs are composed of a dual population of
excellent prognosis. This subset includes low- luminal and basaloid cells that form islands with

Fig. 4. A 49-year-old woman presented with a breast mass. (A) Mammogram showed dense breast tissue with no
apparent focal lesions. (B) Ultrasound showed a circumscribed lobulated lesion with heterogeneous echogenicity.
(C) MR imaging showed a 15-mm ring-enhancing lesion that did not reach threshold, with increased signal on T2
and low signal on T1. (D) Core biopsy showed an invasive mucinous carcinoma, with islands of tumor cells floating
in background mucin. (E) The tumor cells stained strongly positive for estrogen receptor but were negative for
HER2 (F). This finding demonstrates the potential limitations, including nonvisualization of the malignancy,
when imaging mucinous tumors by mammography.
 Molecular Classification of Breast Cancer 333

a cribriform or sievelike architecture with spaces appear deceptively innocuous on imaging, partic-
filled with bluish mucin and pink basement mem- ularly in young women.53,104
branelike material, although solid and tubular Classic medullary carcinoma is defined by strict
patterns can also be seen.53,110 They were origi- histologic criteria that are poorly reproducible, and
nally described in the salivary gland but can it is now thought that the relatively favorable prog-
occur in the breast or lung. In the breast, AdCCs nosis compared with other grade-3 TNBCs is due
usually occur in older women and are low-grade to the associated immune response, so medullary,
with an excellent prognosis with adequate local atypical medullary, and medullarylike IDC are
therapy; lymph node and distant metastases grouped together in the most recent World Health
are rare.2 Regardless of site, AdCCs harbor a Organization classification as carcinoma with
chromosomal rearrangement involving the MYB medullary features.2 These tumors show high
gene, typically resulting in a t(6;9)(q22–23;p23– levels of genomic instability with complex copy
24) translocation with the formation of a MYB- number aberrations and structural gene rear-
NFIB fusion gene. This translocation has been rangements. Structural gene rearrangements
detected in 23% to 100% of breast AdCCs.53 resulting in the creation of neo-antigens, which
Apart from this translocation, the genome is sta- then activate the host immune response.114,115
ble and they lack copy number changes, such as The cancers that arise in BRCA1 mutation carriers
8q gain and 5q loss or mutations in TP53 and often have medullary features with a very high
PIK3CA that are commonly seen in other mitotic rate, although conversely only 13% of
TNBCs.111 Rare cases with progression to poorly women diagnosed with medullarylike cancers
differentiated basaloid carcinoma accompanied have BRCA1 mutations.116,117 Alternative mecha-
by accumulation of additional genetic alterations nisms, such as promoter hypermethylation, may
have been described.112 be responsible for BRCA1 downregulation in
cases lacking a germline mutation.118 Medullary-
like carcinomas also frequently contain TP53
Secretory Carcinoma
mutations; this was the only high-frequency muta-
Secretory carcinoma is a very rare tumor that was tion seen in medullary carcinomas within the
originally described in children and adolescents METABRIC cohort, present in more than 80% of
but can occur in patients of all ages and in cases.86
men.2,113 It is composed of cells with abundant In conclusion, new HT genetic technologies have
granular eosinophilic to foamy cytoplasm forming enhanced our understanding of breast cancer
microcystic, tubular and solid patterns, with peri- biology and changed the way we conceptualize
odic acid–Schiff–positive intracellular and extra- breast cancer classification. However, the tradi-
cellular secretory material that gives it its name. tional histologic subtypes of breast cancer retain
Despite being triple negative, they are low grade their value because of their clinical and radiological
and have an excellent prognosis. Secretory carci- correlations and provide some clues as to the un-
noma has a characteristic t(12;15) translocation derlying molecular biology. Hence, the two are
that produces a ETV6-NTRK3 fusion gene; this complementary; both play an important role at pre-
translocation is unique to secretory carcinoma in sent in determining patient management. The his-
the breast but has been described in other tumors, tologic and molecular subtypes of breast cancer
such as infantile fibrosarcoma and congenital have important implications for how tumors are
mesoblastic nephroma.53,109 Similar to AdCC, visualized on imaging studies. Several subtypes
the genome is otherwise stable with few copy of breast cancer are more difficult to image on
number alterations. Interestingly, a salivary gland mammography. Basal-like/TNBCs tend to be high-
counterpart to secretory carcinoma has now ly FDG avid on PET, whereas the lobular histologic
been recognized, in part because it harbors the subtype may be less avid and less perceptible on
same t(12;15) translocation. FDG PET imaging.

Medullarylike Carcinoma ACKNOWLEDGMENTS

In contrast, medullarylike carcinomas are TNBCs Dr S. Chin acknowledges Cancer Research UK.
characterized by marked nuclear pleomorphism, Dr G.A. Ulaner acknowledges the Department of
a high mitotic rate, syncytial growth pattern, and Defense Breast Cancer Research Program Break-
pushing margins with a prominent lymphocytic in- through Award BC132676 and the MSKCC Radio-
flammatory infiltrate. Despite the aggressive histo- chemistry and Molecular Imaging Probe Core
logic features, the presence of pushing margins (NIH/NCI Cancer Center Support Grant P30
means that they are well circumscribed and can CA008748).
334 Provenzano et al

REFERENCES 16. Brenton JD, Carey LA, Ahmed AA, et al. Molecular
classification and molecular forecasting of breast
1. Ali HR, Rueda OM, Chin SF, et al. Genome-driven in- cancer: ready for clinical application? J Clin Oncol
tegrated classification of breast cancer validated in 2005;23(29):7350–60.
over 7,500 samples. Genome Biol 2014;15(8):431. 17. Sinn P, Aulmann S, Wirtz R, et al. Multigene assays
2. Lakhani SR, Ellis IO, Schnitt SJ, et al. WHO classi- for classification, prognosis, and prediction in
fication of tumours of the breast. 4 edition. Lyon breast cancer: a critical review on the background
(France): International Agency for Research on and clinical utility. Geburtshilfe Frauenheilkd 2013;
Cancer (IARC) Press; 2012. 73(9):932–40.
3. Cavallaro S, Paratore S, de Snoo F, et al. Genomic 18. Sonnenblick A, Fumagalli D, Azim HA Jr, et al. New
analysis: toward a new approach in breast cancer strategies in breast cancer: the significance of mo-
management. Crit Rev Oncol Hematol 2012;81(3): lecular subtypes in systemic adjuvant treatment for
207–23. small T1a,bN0M0 tumors. Clin Cancer Res 2014;
4. Rizzo JM, Buck MJ. Key principles and clinical ap- 20(24):6242–6.
plications of "next-generation" DNA sequencing. 19. Early Breast Cancer Trialists’ Collaborative Group
Cancer Prev Res (Phila) 2012;5(7):887–900. (EBCTCG), Peto R, Davies C, Godwin J, et al.
5. Perou CM, Sorlie T, Eisen MB, et al. Molecular por- Comparisons between different polychemotherapy
traits of human breast tumours. Nature 2000; regimens for early breast cancer: meta-analyses of
406(6797):747–52. long-term outcome among 100,000 women in 123
6. Russnes HG, Navin N, Hicks J, et al. Insight into the randomised trials. Lancet 2012;379(9814):432–44.
heterogeneity of breast cancer through next- 20. Maia AT, Sammut SJ, Jacinta-Fernandes A, et al.
generation sequencing. J Clin Invest 2011; Big data in cancer genomics. Curr Opin Syst Biol
121(10):3810–8. 2017;4:78–84.
7. Sotiriou C, Neo SY, McShane LM, et al. Breast can- 21. Sorlie T, Perou CM, Tibshirani R, et al. Gene
cer classification and prognosis based on gene expression patterns of breast carcinomas distin-
expression profiles from a population-based guish tumor subclasses with clinical implications.
study. Proc Natl Acad Sci U S A 2003;100(18): Proc Natl Acad Sci U S A 2001;98(19):10869–74.
10393–8. 22. Prat A, Pineda E, Adamo B, et al. Clinical implica-
8. Rakha EA, Reis-Filho JS, Baehner F, et al. Breast tions of the intrinsic molecular subtypes of breast
cancer prognostic classification in the molecular cancer. Breast 2015;24(Suppl 2):S26–35.
era: the role of histological grade. Breast Cancer 23. Weigelt B, Baehner FL, Reis-Filho JS. The contribu-
Res 2010;12(4):207. tion of gene expression profiling to breast cancer
9. Schnitt SJ. Classification and prognosis of invasive classification, prognostication and prediction: a
breast cancer: from morphology to molecular tax- retrospective of the last decade. J Pathol 2010;
onomy. Mod Pathol 2010;23(Suppl 2):S60–4. 220(2):263–80.
10. Haybittle JL, Blamey RW, Elston CW, et al. 24. Mackay A, Weigelt B, Grigoriadis A, et al. Microar-
A prognostic index in primary breast cancer. Br J ray-based class discovery for molecular classifica-
Cancer 1982;45(3):361–6. tion of breast cancer: analysis of interobserver
11. Olivotto IA, Bajdik CD, Ravdin PM, et al. Popula- agreement. J Natl Cancer Inst 2011;103(8):662–73.
tion-based validation of the prognostic model 25. Pusztai L, Mazouni C, Anderson K, et al. Molecular
ADJUVANT! for early breast cancer. J Clin Oncol classification of breast cancer: limitations and po-
2005;23(12):2716–25. tential. Oncologist 2006;11(8):868–77.
12. Ravdin PM. A computer program to assist in mak- 26. Weigelt B, Mackay A, A’Hern R, et al. Breast cancer
ing breast cancer adjuvant therapy decisions. molecular profiling with single sample predictors: a
Semin Oncol 1996;23(1 Suppl 2):43–50. retrospective analysis. Lancet Oncol 2010;11(4):
13. Wishart GC, Azzato EM, Greenberg DC, et al. 339–49.
PREDICT: a new UK prognostic model that predicts 27. Zhao X, Rodland EA, Sorlie T, et al. Systematic
survival following surgery for invasive breast can- assessment of prognostic gene signatures for
cer. Breast Cancer Res 2010;12(1):R1. breast cancer shows distinct influence of time
14. Wishart GC, Bajdik CD, Dicks E, et al. PREDICT and ER status. BMC Cancer 2014;14:211.
Plus: development and validation of a prognostic 28. Parker JS, Mullins M, Cheang MC, et al. Super-
model for early breast cancer that includes HER2. vised risk predictor of breast cancer based on
Br J Cancer 2012;107(5):800–7. intrinsic subtypes. J Clin Oncol 2009;27(8):1160–7.
15. Wishart GC, Rakha E, Green A, et al. Inclusion of 29. Wallden B, Storhoff J, Nielsen T, et al. Development
KI67 significantly improves performance of the and verification of the PAM50-based Prosigna
PREDICT prognostication and prediction model breast cancer gene signature assay. BMC Med
for early breast cancer. BMC Cancer 2014;14:908. Genomics 2015;8:54.
 Molecular Classification of Breast Cancer 335

30. Dowsett M, Sestak I, Lopez-Knowles E, et al. Com- 42. Ebbert MT, Bastien RR, Boucher KM, et al. Charac-
parison of PAM50 risk of recurrence score with on- terization of uncertainty in the classification of multi-
cotype DX and IHC4 for predicting risk of distant variate assays: application to PAM50 centroid-
recurrence after endocrine therapy. J Clin Oncol based genomic predictors for breast cancer treat-
2013;31(22):2783–90. ment plans. J Clin Bioinforma 2011;1:37.
31. Filipits M, Nielsen TO, Rudas M, et al. The PAM50 43. Fan C, Oh DS, Wessels L, et al. Concordance
risk-of-recurrence score predicts risk for late among gene-expression-based predictors for
distant recurrence after endocrine therapy in post- breast cancer. N Engl J Med 2006;355(6):560–9.
menopausal women with endocrine-responsive 44. Kelly CM, Bernard PS, Krishnamurthy S, et al.
early breast cancer. Clin Cancer Res 2014;20(5): Agreement in risk prediction between the
1298–305. 21-gene recurrence score assay (Oncotype
32. Gnant M, Filipits M, Greil R, et al. Predicting distant DX(R)) and the PAM50 breast cancer intrinsic clas-
recurrence in receptor-positive breast cancer pa- sifier in early-stage estrogen receptor-positive
tients with limited clinicopathological risk: using breast cancer. Oncologist 2012;17(4):492–8.
the PAM50 Risk of Recurrence score in 1478 post- 45. Prat A, Parker JS, Fan C, et al. Concordance
menopausal patients of the ABCSG-8 trial treated among gene expression-based predictors for
with adjuvant endocrine therapy alone. Ann Oncol ER-positive breast cancer treated with adjuvant
2014;25(2):339–45. tamoxifen. Ann Oncol 2012;23(11):2866–73.
33. Gnant M, Sestak I, Filipits M, et al. Identifying 46. Varga Z, Sinn P, Fritzsche F, et al. Comparison of
clinically relevant prognostic subgroups of post- EndoPredict and Oncotype DX test results in hor-
menopausal women with node-positive hormone mone receptor positive invasive breast cancer.
receptor-positive early-stage breast cancer treated PLoS One 2013;8(3):e58483.
with endocrine therapy: a combined analysis of 47. Harris LN, Ismaila N, McShane LM, et al. Use of
ABCSG-8 and ATAC using the PAM50 risk of recur- biomarkers to guide decisions on adjuvant sys-
rence score and intrinsic subtype. Ann Oncol 2015; temic therapy for women with early-stage invasive
26(8):1685–91. breast cancer: American Society of Clinical
34. Prat A, Perou CM. Deconstructing the molecular por- Oncology clinical practice guideline. J Clin Oncol
traits of breast cancer. Mol Oncol 2011;5(1):5–23. 2016;34(10):1134–50.
35. Prat A, Cheang MC, Martin M, et al. Prognostic 48. Rouzier R, Perou CM, Symmans WF, et al. Breast
significance of progesterone receptor-positive tu- cancer molecular subtypes respond differently to
mor cells within immunohistochemically defined preoperative chemotherapy. Clin Cancer Res
luminal A breast cancer. J Clin Oncol 2013;31(2): 2005;11(16):5678–85.
203–9. 49. Prat A, Carey LA, Adamo B, et al. Molecular fea-
36. Ades F, Zardavas D, Bozovic-Spasojevic I, et al. tures and survival outcomes of the intrinsic sub-
Luminal B breast cancer: molecular characteriza- types within HER2-positive breast cancer. J Natl
tion, clinical management, and future perspectives. Cancer Inst 2014;106(8) [pii:dju152].
J Clin Oncol 2014;32(25):2794–803. 50. Prat A, Bianchini G, Thomas M, et al. Research-
37. Cheang MC, Chia SK, Voduc D, et al. Ki67 index, based PAM50 subtype predictor identifies higher
HER2 status, and prognosis of patients with luminal responses and improved survival outcomes in
B breast cancer. J Natl Cancer Inst 2009;101(10): HER2-positive breast cancer in the NOAH study.
736–50. Clin Cancer Res 2014;20(2):511–21.
38. Filipits M, Rudas M, Jakesz R, et al. A new molec- 51. Badve S, Dabbs DJ, Schnitt SJ, et al. Basal-like and
ular predictor of distant recurrence in ER-positive, triple-negative breast cancers: a critical review with
HER2-negative breast cancer adds independent an emphasis on the implications for pathologists
information to conventional clinical risk factors. and oncologists. Mod Pathol 2011;24(2):157–67.
Clin Cancer Res 2011;17(18):6012–20. 52. Jacquemier J, Padovani L, Rabayrol L, et al.
39. Paik S, Shak S, Tang G, et al. A multigene assay to Typical medullary breast carcinomas have a
predict recurrence of tamoxifen-treated, node- basal/myoepithelial phenotype. J Pathol 2005;
negative breast cancer. N Engl J Med 2004; 207(3):260–8.
351(27):2817–26. 53. Pareja F, Geyer FC, Marchio C, et al. Triple-nega-
40. van de Vijver MJ, He YD, van’t Veer LJ, et al. tive breast cancer: the importance of molecular
A gene-expression signature as a predictor of sur- and histologic subtyping, and recognition of low-
vival in breast cancer. N Engl J Med 2002;347(25): grade variants. NPJ Breast Cancer 2016;2:16036.
1999–2009. 54. Weigelt B, Kreike B, Reis-Filho JS. Metaplastic
41. Kittaneh M, Montero AJ, Gluck S. Molecular breast carcinomas are basal-like breast cancers:
profiling for breast cancer: a comprehensive re- a genomic profiling analysis. Breast Cancer Res
view. Biomark Cancer 2013;5:61–70. Treat 2009;117(2):273–80.
336 Provenzano et al

55. Wetterskog D, Lopez-Garcia MA, Lambros MB, reveals novel subgroups. Nature 2012;486(7403):
et al. Adenoid cystic carcinomas constitute a ge- 346–52.
nomically distinct subgroup of triple-negative and 68. Horlings HM, Weigelt B, Anderson EM, et al.
basal-like breast cancers. J Pathol 2012;226(1): Genomic profiling of histological special types of
84–96. breast cancer. Breast Cancer Res Treat 2013;
56. Foulkes WD, Stefansson IM, Chappuis PO, et al. 142(2):257–69.
Germline BRCA1 mutations and a basal epithelial 69. Johnson K, Sarma D, Hwang ES. Lobular breast
phenotype in breast cancer. J Natl Cancer Inst cancer series: imaging. Breast Cancer Res 2015;
2003;95(19):1482–5. 17:94.
57. Severson TM, Peeters J, Majewski I, et al. BRCA1- 70. Ciriello G, Gatza ML, Beck AH, et al. Comprehen-
like signature in triple negative breast cancer: mo- sive molecular portraits of invasive lobular breast
lecular and clinical characterization reveals sub- cancer. Cell 2015;163(2):506–19.
groups with therapeutic potential. Mol Oncol 71. Monhollen L, Morrison C, Ademuyiwa FO, et al.
2015;9(8):1528–38. Pleomorphic lobular carcinoma: a distinctive clin-
58. Lehmann BD, Bauer JA, Chen X, et al. Identifica- ical and molecular breast cancer type. Histopathol-
tion of human triple-negative breast cancer sub- ogy 2012;61(3):365–77.
types and preclinical models for selection of 72. Simpson PT, Reis-Filho JS, Lambros MB, et al. Mo-
targeted therapies. J Clin Invest 2011;121(7): lecular profiling pleomorphic lobular carcinomas of
2750–67. the breast: evidence for a common molecular ge-
59. Lehmann BD, Jovanovic B, Chen X, et al. Refine- netic pathway with classic lobular carcinomas.
ment of triple-negative breast cancer molecular J Pathol 2008;215(3):231–44.
subtypes: implications for neoadjuvant chemo- 73. Cowin P, Rowlands TM, Hatsell SJ. Cadherins and
therapy selection. PLoS One 2016;11(6):e0157368. catenins in breast cancer. Curr Opin Cell Biol 2005;
60. Masuda H, Baggerly KA, Wang Y, et al. Differential 17(5):499–508.
response to neoadjuvant chemotherapy among 7 74. Da Silva L, Parry S, Reid L, et al. Aberrant expres-
triple-negative breast cancer molecular subtypes. sion of E-cadherin in lobular carcinomas of the
Clin Cancer Res 2013;19(19):5533–40. breast. Am J Surg Pathol 2008;32(5):773–83.
61. Farmer P, Bonnefoi H, Becette V, et al. Identification 75. Doyle S, Evans AJ, Rakha EA, et al. Influence of
of molecular apocrine breast tumours by microar- E-cadherin expression on the mammographic
ray analysis. Oncogene 2005;24(29):4660–71. appearance of invasive nonlobular breast carci-
62. Lehmann BD, Bauer JA, Schafer JM, et al. PIK3CA noma detected at screening. Radiology 2009;
mutations in androgen receptor-positive triple 253(1):51–5.
negative breast cancer confer sensitivity to the 76. Rakha EA, Patel A, Powe DG, et al. Clinical and
combination of PI3K and androgen receptor inhib- biological significance of E-cadherin protein
itors. Breast Cancer Res 2014;16(4):406. expression in invasive lobular carcinoma of the
63. Basu S, Chen W, Tchou J, et al. Comparison of breast. Am J Surg Pathol 2010;34(10):1472–9.
triple-negative and estrogen receptor-positive/ 77. Guiu S, Wolfer A, Jacot W, et al. Invasive lobular
progesterone receptor-positive/HER2-negative breast cancer and its variants: how special are
breast carcinoma using quantitative fluorine-18 flu- they for systemic therapy decisions? Crit Rev On-
orodeoxyglucose/positron emission tomography col Hematol 2014;92(3):235–57.
imaging parameters: a potentially useful method 78. Balmativola D, Marchio C, Maule M, et al. Patho-
for disease characterization. Cancer 2008;112(5): logical non-response to chemotherapy in a neoad-
995–1000. juvant setting of breast cancer: an inter-institutional
64. Groheux D, Giacchetti S, Moretti JL, et al. Correla- study. Breast Cancer Res Treat 2014;148(3):
tion of high 18F-FDG uptake to clinical, patholog- 511–23.
ical and biological prognostic factors in breast 79. Petruolo OA, Pilewskie M, Patil S, et al. Standard
cancer. Eur J Nucl Med Mol Imaging 2011;38(3): pathologic features can be used to identify a sub-
426–35. set of estrogen receptor-positive, HER2 negative
65. Groheux D, Hindie E, Delord M, et al. Prognostic patients likely to benefit from neoadjuvant chemo-
impact of (18)FDG-PET-CT findings in clinical stage therapy. Ann Surg Oncol 2017;24(9):2556–62.
III and IIB breast cancer. J Natl Cancer Inst 2012; 80. Tsai ML, Lillemoe TJ, Finkelstein MJ, et al. Utility of
104(24):1879–87. oncotype DX risk assessment in patients with inva-
66. Ciriello G, Miller ML, Aksoy BA, et al. Emerging sive lobular carcinoma. Clin Breast Cancer 2016;
landscape of oncogenic signatures across human 16(1):45–50.
cancers. Nat Genet 2013;45(10):1127–33. 81. Weigelt B, Geyer FC, Reis-Filho JS. Histological
67. Curtis C, Shah SP, Chin SF, et al. The genomic and types of breast cancer: how special are they? Mol
transcriptomic architecture of 2,000 breast tumours Oncol 2010;4(3):192–208.
 Molecular Classification of Breast Cancer 337

82. Lopez-Garcia MA, Geyer FC, Lacroix-Triki M, et al. 95. Hogan MP, Goldman DA, Dashevsky B, et al. Com-
Breast cancer precursors revisited: molecular fea- parison of 18F-FDG PET/CT for systemic staging of
tures and progression pathways. Histopathology newly diagnosed invasive lobular carcinoma
2010;57(2):171–92. versus invasive ductal carcinoma. J Nucl Med
83. Weigelt B, Geyer FC, Natrajan R, et al. The molecular 2015;56(11):1674–80.
underpinning of lobular histological growth pattern: a 96. Borst MJ, Ingold JA. Metastatic patterns of invasive
genome-wide transcriptomic analysis of invasive lobular versus invasive ductal carcinoma of the
lobular carcinomas and grade- and molecular breast. Surgery 1993;114(4):637–41 [discussion:
subtype-matched invasive ductal carcinomas of no 641–2].
special type. J Pathol 2010;220(1):45–57. 97. He H, Gonzalez A, Robinson E, et al. Distant meta-
84. Michaut M, Chin SF, Majewski I, et al. Integration of static disease manifestations in infiltrating lobular
genomic, transcriptomic and proteomic data iden- carcinoma of the breast. AJR Am J Roentgenol
tifies two biologically distinct subtypes of invasive 2014;202(5):1140–8.
lobular breast cancer. Sci Rep 2016;6:18517. 98. Kane AJ, Wang ZJ, Qayyum A, et al. Frequency
85. Dawson SJ, Rueda OM, Aparicio S, et al. A new and etiology of unexplained bilateral hydroneph-
genome-driven integrated classification of breast rosis in patients with breast cancer: results of a
cancer and its implications. EMBO J 2013;32(5): longitudinal CT study. Clin Imaging 2012;36(4):
617–28. 263–6.
86. Pereira B, Chin SF, Rueda OM, et al. The somatic 99. Lamovec J, Bracko M. Metastatic pattern of infil-
mutation profiles of 2,433 breast cancers refines trating lobular carcinoma of the breast: an autopsy
their genomic and transcriptomic landscapes. Nat study. J Surg Oncol 1991;48(1):28–33.
Commun 2016;7:11479. 100. Ulaner GA, Goldman DA, Corben A, et al. Prospec-
87. Abhik Mukherjee, Roslin Russell, Suet-Feung Chin, tive clinical trial of (18)F-fluciclovine PET/CT for
et al. Associations between genomic stratification determining the response to neoadjuvant therapy
of breast cancer and centrally reviewed tumour pa- in invasive ductal and invasive lobular breast can-
thology in the METABRIC cohort. British Journal of cers. J Nucl Med 2017;58(7):1037–42.
Cancer (in press). 101. Ulaner GA, Goldman DA, Gonen M, et al. Initial re-
88. Avril N, Menzel M, Dose J, et al. Glucose meta- sults of a prospective clinical trial of 18F-fluciclo-
bolism of breast cancer assessed by 18F-FDG vine PET/CT in newly diagnosed invasive ductal
PET: histologic and immunohistochemical tissue and invasive lobular breast cancers. J Nucl Med
analysis. J Nucl Med 2001;42(1):9–16. 2016;57(9):1350–6.
89. Avril N, Rose CA, Schelling M, et al. Breast imaging 102. Rakha EA, Lee AH, Evans AJ, et al. Tubular carcinoma
with positron emission tomography and fluorine-18 of the breast: further evidence to support its excellent
fluorodeoxyglucose: use and limitations. J Clin On- prognosis. J Clin Oncol 2010;28(1):99–104.
col 2000;18(20):3495–502. 103. Lopez-Garcia MA, Geyer FC, Natrajan R, et al.
90. Berg WA, Gutierrez L, NessAiver MS, et al. Diag- Transcriptomic analysis of tubular carcinomas of
nostic accuracy of mammography, clinical exami- the breast reveals similarities and differences with
nation, US, and MR imaging in preoperative molecular subtype-matched ductal and lobular
assessment of breast cancer. Radiology 2004; carcinomas. J Pathol 2010;222(1):64–75.
233(3):830–49. 104. Larribe M, Thomassin-Piana J, Jalaguier-
91. Bos R, van Der Hoeven JJ, van Der Wall E, et al. Bio- Coudray A. Breast cancers with round lumps: cor-
logic correlates of (18)fluorodeoxyglucose uptake in relations between imaging and anatomopathology.
human breast cancer measured by positron emis- Diagn Interv Imaging 2014;95(1):37–46.
sion tomography. J Clin Oncol 2002;20(2):379–87. 105. Caldarella A, Buzzoni C, Crocetti E, et al. Invasive
92. Buck A, Schirrmeister H, Kuhn T, et al. FDG uptake breast cancer: a significant correlation between
in breast cancer: correlation with biological and histological types and molecular subgroups.
clinical prognostic parameters. Eur J Nucl Med J Cancer Res Clin Oncol 2013;139(4):617–23.
Mol Imaging 2002;29(10):1317–23. 106. Lacroix-Triki M, Suarez PH, MacKay A, et al.
93. Lopez JK, Bassett LW. Invasive lobular carcinoma Mucinous carcinoma of the breast is genomically
of the breast: spectrum of mammographic, US, distinct from invasive ductal carcinomas of no spe-
and MR imaging findings. Radiographics 2009; cial type. J Pathol 2010;222(3):282–98.
29(1):165–76. 107. Buttitta F, Felicioni L, Barassi F, et al. PIK3CA muta-
94. Dashevsky BZ, Goldman DA, Parsons M, et al. tion and histological type in breast carcinoma: high
Appearance of untreated bone metastases from frequency of mutations in lobular carcinoma.
breast cancer on FDG PET/CT: importance of histo- J Pathol 2006;208(3):350–5.
logic subtype. Eur J Nucl Med Mol Imaging 2015; 108. Geyer FC, Lambros MB, Natrajan R, et al. Gen-
42(11):1666–73. omic and immunohistochemical analysis of
338 Provenzano et al

adenosquamous carcinoma of the breast. Mod 114. McGranahan N, Furness AJ, Rosenthal R, et al.
Pathol 2010;23(7):951–60. Clonal neoantigens elicit T cell immunoreactivity
109. Krings G, Joseph NM, Bean GR, et al. Genomic and sensitivity to immune checkpoint blockade.
profiling of breast secretory carcinomas reveals Science 2016;351(6280):1463–9.
distinct genetics from other breast cancers and 115. Turajlic S, Litchfield K, Xu H, et al. Insertion-and-
similarity to mammary analog secretory carci- deletion-derived tumour-specific neoantigens and
nomas. Mod Pathol 2017;30(8):1086–99. the immunogenic phenotype: a pan-cancer anal-
110. Foschini MP, Morandi L, Asioli S, et al. The morpho- ysis. Lancet Oncol 2017;18(8):1009–21.
logical spectrum of salivary gland type tumours of 116. Armes JE, Egan AJ, Southey MC, et al. The histologic
the breast. Pathology 2017;49(2):215–27. phenotypes of breast carcinoma occurring before
111. Martelotto LG, De Filippo MR, Ng CK, et al. age 40 years in women with and without BRCA1 or
Genomic landscape of adenoid cystic carcinoma BRCA2 germline mutations: a population-based
of the breast. J Pathol 2015;237(2):179–89. study. Cancer 1998;83(11):2335–45.
112. Fusco N, Geyer FC, De Filippo MR, et al. Genetic 117. Lakhani SR, Jacquemier J, Sloane JP, et al. Multi-
events in the progression of adenoid cystic carci- factorial analysis of differences between sporadic
noma of the breast to high-grade triple-negative breast cancers and cancers involving BRCA1
breast cancer. Mod Pathol 2016;29(11):1292–305. and BRCA2 mutations. J Natl Cancer Inst 1998;
113. Li D, Xiao X, Yang W, et al. Secretory breast carci- 90(15):1138–45.
noma: a clinicopathological and immunopheno- 118. Osin P, Lu YJ, Stone J, et al. Distinct genetic and
typic study of 15 cases with a review of the epigenetic changes in medullary breast cancer.
literature. Mod Pathol 2012;25(4):567–75. Int J Surg Pathol 2003;11(3):153–8.