Molecular drivers of lobular carcinoma in situ

In contrast to the role of LCIS as a risk factor for the development of breast cancer,
a role for direct progression of LCIS into invasive cancer is less well accepted,
and the molecular basis is currently poorly understood. LCIS must traverse myoepithelial
cells and the basement membrane in order to invade locally. Schematically shown in
Fig. 2 are four proposed mechanisms that could influence this process, and that are likely
not mutually exclusive. Cells may acquire genetic (and/or epigenetic) changes in critical
pathways that allow migration into the stroma. Alternatively, or in concert, cells
may receive signals that cause progression from the stroma, including fibroblasts,
adipocytes, and immune cells. Another possibility is that tumor suppressive myoepithelial
cells may become compromised, allowing LCIS cells to break through and gain access
to the stroma. Finally, enlarging LCIS cellular density may cause physical strain
on the myoepithelial cells and basement membrane such that the cells may physically
rupture a lobule. It is most likely that the progression of LCIS to IBC occurs through
a combination of these events.

Fig. 2. Proposed mechanisms of lobular carcinoma in situ progression to invasive breast cancer. Progression of lobular carcinoma in situ (LCIS) to invasive breast cancer may be influenced by multiple factors, including
cell-intrinsic changes, such as mutations, extrinsic factors from interaction with
the microenvironment, changes within the myoepithelial cells (MEPs), and physical
strain on the basement membrane components, exerted by LCIS within the lobule, causing
cells to rupture the lobule. ILC, invasive lobular carcinoma

To better identify LCIS lesions that may progress versus those that remain dormant
requires a thorough knowledge of the mechanisms that drive progression to invasive
disease. Current research on LCIS has focused on four main areas: (1) prognostic markers,
(2) genomic changes, (3) factors related to epithelial to mesenchymal transition (EMT),
and (4) signaling pathways. We summarize and discuss those areas of research below.

Prognostic markers

The expression of nuclear receptors – especially ER – can be used to predict clinical
outcome of tumors 41], 42]; 80 to 100 % of LCIS cases express ER?, most of which show moderate to strong immunoreactivity
by immunohistochemistry 43]–46] (Table 1). Similarly, ILC is also a highly ER-positive disease, with greater than 90 % ER
positivity 47], 48]. Some aggressive variants of LCIS are more likely to be ER-negative. For example,
80 % of apocrine PLCIS lesions are ER-negative 49]. This suggests that ER-negativity in LCIS may be a potential marker of progression
of more aggressive lesions. However, because most CLCIS and many PLCIS lesions are
ER-positive, additional prognostic markers are clearly needed to better differentiate
ER-positive lesions that will progress versus those that will not.

Table 1. Common prognostic markers in classical lobular carcinoma in situ

Expression of PR is regulated by ER and is considered a prognostic marker of IBC 50]. Loss of PR expression is associated with endocrine resistance 51], and luminal B tumors are more often PR-negative/low compared to less aggressive
luminal A tumors. Approximately 47 to 90 % of LCIS lesions express PR. However, the
expression of PR is lower in LCIS lesions associated with IBC 43]–46], and in apocrine PLCIS 31], implying that PR-low lesions are more likely to progress to invasive cancer 52]. Intriguingly, a recent study showed an inverse relationship between ER/PR status
and Ki67 proliferation rate in ductal cancer but not in lobular cancer, such that
ER-negative status did not correlate with high Ki67 in invasive lobular cancers whereas
it did with invasive ductal cancers 52]. However, this has not yet been studied in detail in LCIS.

In contrast to ER?, the role of ER?1 and the spliced variant ER?2 in breast cancer
is less well understood 53]. Some hypothesize that tamoxifen is an agonist of ER? 54], thus suggesting that ER? could be a marker of poor prognosis, due to its ability
to oppose the anti-proliferative effects of tamoxifen-binding ER?. Recently, Huang
and colleagues 55] measured ER? expression in DCIS, invasive ductal cancer (IDC), and ILC. They concluded
that while ER? expression is high in normal mammary epithelial cells, ER? expression
is low in DCIS and IDC. In contrast, ILC tumors express higher levels of ER?, with
a reduction in expression in late stage ILC. LCIS samples were not included in this
study. Results from an earlier study showed that LCIS has higher ER?2 expression compared
with normal epithelium, but that ER?1 expression is not different 43]. Huang and colleagues concluded that the spliced variant ER?2 is an indicator of
hypoxia, not malignancy, which may explain the increased ER?2 spliced variant in LCIS
43], 55]. In contrast to this observation, Nonni and colleagues 56] showed that ER? expression in LN is significantly lower than in normal epithelium,
although this study had a smaller sample size (n?=?30).

Amplification of c-erbB-2 (HER2) is a marker of poor prognosis in patients with IBC.
Fortunately, anti-HER2 antibodies have been effective drugs for HER2-positive tumors
57]. Understanding expression levels of HER2 in LCIS may shed light on its malignant
nature. In LCIS, 0 to 11 % of tumors have HER2 amplification (Table 1). More aggressive LCIS subtypes are more likely to have amplified HER2 44], 45]; 18 % of FLCIS and 31 % of apocrine PLCIS show HER2 amplification 33], 49].

Ki-67 expression is a marker of the proliferation rate of a tumor, and higher proliferative
rates correlate with poor clinical outcomes 58]. In many LCIS lesions, Ki-67 expression is very low, corresponding to a 0 to 2 %
proliferation rate in some studies 44], 45]. Other studies have shown that some LCIS lesions express a higher than 10 % proliferation
rate 59]. Patients who have LCIS with higher proliferation rates may have a higher probability
of relapse after surgery 59]. Currently, however, Ki67 is not used clinically to guide management decisions for
LCIS.

The tumor suppressor gene encoding p53 is often dysregulated in human cancers 60]. In LCIS, p53 overexpression (reflecting protein stabilization as result of mutation)
has been shown to be relatively low, ranging from 0 to 19 % using immunohistochemistry
44], 46]. Although the mutation rate of the p53 gene has not been assessed for LCIS, loss
of heterozygosity has been observed for chromosome 17p, which is the location of the
gene that encodes p53.

Recently, in a study by Andrade and colleagues, 23 patient-matched samples of normal
breast tissue, LCIS, and ILC were subjected to microarray analysis to determine which
genes might be involved in the progression of LCIS 61]. They identified 169 candidate genes involved in LCIS progression. The same study
also showed that 40 CLCIS patient samples clustered in two groups, suggesting heterogeneity
between CLCIS lesions at the transcriptomic level, even if they may otherwise appear
homogenous.

The prognostic markers mentioned above do not reliably and accurately predict the
potential of LCIS lesions to progress to invasive disease. Therefore, there is a critical
need to identify better markers of progression, which might be used clinically to
guide management.

Genomic changes

Much of what is known about LCIS has been generated from studies utilizing aCGH. These
studies, and others, suggest that LCIS and ILC are genetically similar and clonally
related 20], 22], 23], 34], 62]. aCGH studies have also revealed similarities between lobular lesions and other low-grade
lesions, including flat epithelial atypia, atypical ductal hyperplasia, low-grade
DCIS and low grade IDC 34], 63], 64]. In light of these data, some have proposed that a broadly defined low-grade family
of breast neoplasia exists, which has similar molecular drivers during disease progression
65], 66]. Characterization of breast cancer subtypes using gene expression profiling and DNA
copy number variation has led to depiction of HER2-positive and ‘triple negative’
breast cancers as part of a ‘high-grade pathway’ and certain low-grade ER/PR-positive
breast cancers as part of the ‘low-grade pathway’ 67]. Recently, this ‘low-grade precursor hypothesis’ has been challenged, with evidence
that LCIS can progress into both low-grade and high-grade tumors 22], 67] and that LCIS can be a precursor to both ILC and IDC 62].

Specific chromosomal alterations are found frequently and consistently in LCIS. The
chromosomal changes most commonly associated with LCIS are loss of 16q and gain of
1q 34]. Chromosome 16q contains several tumor suppressor genes, including E-cadherin (CDH1),
a member of the calcium-dependent adhesion family of transmembrane proteins. Loss
of other genes on 16q, including those encoding dipeptidase 1 (DPEP1) and CCCTC-binding
factor (CTCF), have also been implicated in ILC 34], 68]. Loss of chromosome 16q, combined with mutations often resulting in premature stop
codons and thus truncated proteins, transcriptional repression, and possibly gene
promoter methylation, can lead to biallelic inactivation of CDH1. In addition to the
16q- and 1q?+?signature, many LCIS (both classical and pleomorphic) lesions demonstrate
loss of 17p, which maps the gene encoding p53 33]. Loss or amplification of 11q (containing the cyclin D1 gene) and loss of 8p are
seen with a higher incidence in PLCIS compared with CLCIS. Furthermore, some FLCIS
harbor amplification of 17q (spanning the gene encoding HER2), a finding seen less
commonly in CLCIS 33]. Losses of 16p, and gains of 6q are also sometimes observed in LCIS 9]. Amplification of 16p and losses of 3q, 11q and 13q have also been described 49]. Results of aCGH experiments have shown that while most chromosomal changes in LCIS
are not consistent, those that are most consistent (namely, 16q loss and 1q amplification)
are found early in the progression to invasive disease. Although this information
can be helpful for determining the relatedness of different lesions, it is less likely
to be helpful clinically in distinguishing LCIS lesions that will progress from those
that will not. Employing modern genomic techniques, such as next generation sequencing,
will be critical in expanding our understanding of the genetic changes involved in
the progression of LCIS.

LCIS is often multicentric, sometimes arising from 10 or more foci 69], and bilateral LCIS is also common. Furthermore, according to one study, about 23 %
of women who develop LCIS have at least one first-degree relative with IBC 70]. Consistent genomic changes in LCIS may shed light on the genetic inheritance of
the disease. There is evidence that germline polymorphisms in the CDH1 gene (E-cadherin) predispose women to LCIS 71], and LCIS was also found in some patients with CDH1-related hereditary diffuse gastric
cancer syndrome 72].

Recently, Sawyer and colleagues 73] analyzed SNPs in a total of 6539 lobular cancers, including 436 cases of pure LCIS,
to identify those which specifically predisposed women to lobular disease. This study,
which is part of GLACIER, a UK study of lobular breast cancer, utilized the iCOGS
chip, a custom SNP array that comprises 211,155 SNPs enriched at predisposition loci
for breast and other cancers 74]. Six SNPs were found that were strongly associated with ILC and LCIS, but not with
IDC, with rs11977670 (7q34) showing the strongest association. Preliminary data in
this study suggest that this SNP may influence levels and/or activity of JHDM1D, or
SLC37A3, proteins with histone demethylase and sugar-phosphate exchanger functions,
respectively. It is also possible that this SNP interacts with expression or function
of the nearby BRAF gene, or that it controls expression of other non-coding genes. ENCODE data show
overlap of the SNP with an area of H3K27 acetylation, supporting a role of this region
in gene regulation. A SNP in LGR6 (rs6678914) showed specific associations with LCIS,
and not with ILC. Similarly, other variants had stronger effect sizes in LCIS compared
with ILC – for example, SNPs at TOX3, ZNF365 and MLLT10 loci. There were also SNPs that were more strongly associated with ILC compared with
LCIS, including variants in the FGFR2 and MAP3K1 genes. Intriguingly, none of the 56 CDH1 SNPs present on the iCOGS chip showed significant association with lobular cancer.
This study provided an outstanding starting point for further functional studies of
the identified pathways, especially to decipher their roles in development and progression
of LCIS.

Epithelial to mesenchymal transition markers in lobular carcinoma in situ

EMT is a process by which epithelial cells gain characteristics of mesenchymal cells,
thereby promoting motility through tissue stroma 75], 76]. It has been proposed to be an essential step in breast cancer progression and metastasis.
A critical component of EMT is the reduced function of cell-cell junctions, and it
is feasible that EMT could also play a role in the development of LCIS, which is characterized
by decreased cohesiveness within the lobule. Decreased expression of E-cadherin and
dissociation of the cadherin-catenin complex is both a necessary step of EMT and a
hallmark of lobular disease. E-cadherin loss and accumulation of cytosolic p120 catenin
are frequently used diagnostically to differentiate between lobular and ductal lesions
11], 77]. A combination of mechanisms has been shown to contribute to the loss of E-cadherin,
including somatic mutations, chromosomal loss, epigenetic silencing, and transcriptional
repression (Table 2). The tight junction protein claudin 4, which plays a role in loss of cellular adhesion
during EMT, was also shown to be downregulated in LCIS compared with normal tissue
78], and might thus also contribute to the decrease in cellular adhesion in lobular disease.

Table 2. Mechanisms of E-cadherin loss

EMT has been shown to be driven by intrinsic transcription factors, such as SNAIL,
SLUG, TWIST, and ZEB1, and by paracrine signaling molecules, including TGF-? and Wnt
76], 77], 79], 80]. In a subset of LCIS, some EMT genes, such as TWIST, are expressed 77], 81]. There is some evidence that in normal epithelial tissues, TWIST is epigenetically
silenced through hypermethylation of its promoter region and its overexpression in
LCIS is at least in part a result of hypomethylation 81]. TWIST expression is increased even more so in ILC 77], and high expression of ZEB1 was reported in poorly differentiated ILC 79], 80]. Thus, TWIST and ZEB1 may play a role in the development of ILC by promoting EMT
through two major steps: dissociation of cell junctions with loss of polarity, and
cytoskeletal changes that promote motility 76]. Another aspect of EMT involves cytoskeletal changes and increased motility 76]. Rho-GTPases control actin remodeling and are regulated by p120 catenin 82]. With the accumulation of cytosolic p120 catenin in lobular cancer, it is not surprising
that p120 appears to be a major driver of the lobular phenotype 83]. LCIS cells demonstrate an affinity to interact with extracellular matrix components
by increasing mesenchymal surface molecules like N-cadherin 84] and laminin receptor 1 85]. Matrix metalloproteinase 9, well known to cause degradation of extracellular matrix
to promote migration into stroma, was shown to be highly expressed in LCIS compared
with normal mammary epithelium 78]. Collectively these data suggest the early LCIS lesions are poised for invasion;
however, most will not progress to invasive disease.

Activation of other signaling pathways in lobular carcinoma in situ

Several signaling pathways are commonly altered in lobular cancer. Perhaps most frequently,
PIK3CA activating point mutations, long implicated in tumorigenesis, are found in
both in situ and invasive lobular 86], 87]. In fact, in one study, 44 % (7 of 16 cases) of lobular neoplasias harbored activating
PIK3CA mutations. Such mutations are also found in ductal cancers, and are not unique
to breast carcinoma. As a comparison, these point mutations were found in 10 out of
21 (48 %) cases of DCIS and 13 out of 37 (35 %) invasive carcinomas 86].

Similarly to a variety of cancers, c-Src was found to be activated in both LCIS and
ILC. Interestingly, some c-Src downstream targets such as Fak and Stat-3 were only
active in ILC, but not in pre-invasive lobular neoplasia 84], 88]. Such activation thus represents a possible switch to allow LCIS cells to invade.
In addition to Stat3, there is also some evidence for Stat5a playing a potential role
in LCIS development and progression 89]. Stat5 is an important signaling molecule in the development of normal milk-producing
mammary cells, and provides survival signals to mammary epithelial cells during lactation
90]. There is also evidence that increased Stat5 levels prevent apoptosis normally initiated
by oncoproteins and involution 91]. Bratthauer and colleagues 89] reported strong staining for STAT5a in normal mammary epithelial cells, but loss
in DCIS and IDC. Intriguingly, LCIS and ILC lesions retained STAT5 expression in 32 %
and 17 % of the samples, respectively 89]. Amplification of prolactin receptor – an upstream activator of STAT5a signaling
in breast tissue – is also observed in LCIS and ILC lesions, but not in DCIS lesions
92], 93]. These data suggest that STAT5a might provide survival signals to neoplastic cells
in LCIS.

And finally, there is a report showing that cyclooxygenase-2 (COX-2) localizes within
calveolae-like structures in the membrane, especially in more low-grade lesions 94], 95]. COX-2 expression has been implicated in the development of cancers by promoting
an inflammatory environment conducive to tumor development 96] and, despite limitations, COX-2 inhibition may hold promise for cancer therapy and
prevention. Further studies are necessary to understand the role of COX2, and in more
general the role of the immune environment on development and progression of LCIS.