Background
Invasive lobular carcinoma (ILC) represents the second most common subtype of breast cancer, and overall it is the sixth most common cancer diagnosis in women in the United States [
1]. ILC cases typically present with favorable biomarkers, as >90 % of ILCs are estrogen receptor (ER)-positive and progesterone receptor (PR)-positive, <10 % are human epidermal growth factor receptor 2 (HER2)-positive, and the majority are low Ki-67-positive [
1‐
5]. On the basis of these biomarkers, ILCs are an archetype of the luminal A molecular subtype, and patients with ILC may be expected to have favorable outcomes when treated with adjuvant endocrine therapy. However, recent retrospective analyses of the BIG 1-98 trial [
6] and ABCSG-8 trial [
7] suggest that, compared with patients with similar invasive ductal carcinomas (IDCs), a subset of patients with ILC may in fact have poorer outcomes with endocrine therapy. Improved understanding of ER signaling, endocrine response, and the development of endocrine resistance in ILC is critical to improving patient outcomes.
We previously reported a study of unique ER-mediated gene expression and signaling in ILC model systems using gene expression microarrays coupled with ER chromatin immunoprecipitation sequencing (ChIP-seq) [
8]. In this study, the most strongly induced ILC-specific ER target gene was the Wnt ligand
WNT4. Additionally, ChIP-seq identified an ILC-specific estrogen receptor binding site (ERBS) at the
WNT4 locus, approximately 1.5 kb downstream from the
WNT4 transcription start site, an evolutionarily conserved region [
9] that contains two predicted estrogen response elements (EREs) (diagrammed in Additional file
1: Figure S1). These observations suggest that direct ER binding at this site may be responsible for estrogen-induced
WNT4 expression. Importantly, ILC cells may be co-opting
WNT4 regulation by placing it under ER control, as Wnt4 is a transcriptional target and downstream effector of PR signaling in the murine adult mammary gland [
10‐
14]. In this context, Wnt4 is critical to maintaining a mammary progenitor cell population (reviewed by Brisken et al. [
15]). Decreased progenitor cell potential during parity (and subsequent parity-induced breast cancer protection) is linked to downregulation of
Wnt4 [
11], but progenitor cell proliferation is rescued by
Wnt4 induction [
16] or exogenous WNT4 [
11]. On the basis of these observations, we hypothesized that WNT4 may play a critical role in estrogen-regulated phenotypes in ILC.
To test this hypothesis, we assessed regulation and expression of WNT4, WNT4 signaling, and WNT4-mediated phenotypes in ILC- and IDC-derived breast cancer cell lines. In addition, we established a series of long-term estrogen-deprived (LTED) endocrine-resistant variants of the ILC cell lines MDA-MB-134-VI (MM134) and SUM44PE (44PE), and examined the role of WNT4 in endocrine resistance in these models. Our findings suggest that WNT4 signaling is a putative target to modulate endocrine response and combat endocrine resistance for ILC.
Methods
Cell culture
MCF-7 and T47D (American Type Culture Collection [ATCC], Manassas, VA, USA) cells were maintained as described elsewhere [
17]. MM134 (ATCC) and 44PE (Asterand Bioscience, Detroit, MI, USA) cells were maintained as described previously [
8]. MDA-MB-330 cells (MM330; ATCC) were maintained as described for MM134. HCC1428 and HT1080 cells (ATCC) were maintained in DMEM (11965; Life Technologies, Carlsbad, CA, USA) + 10 % FBS (26140; Life Technologies). BCK4 (University of Colorado Anschutz) was maintained as described elsewhere [
18]. All lines were incubated at 37 °C in 5 % CO
2. Cell lines were authenticated annually by polymerase chain reaction (PCR)-restriction fragment length polymorphism analyses at the University of Pittsburgh Cell Culture and Cytogenetics Facility and confirmed to be mycoplasma-negative. Authenticated cells were in continuous culture for <6 months. Cells were hormone-deprived using charcoal-stripped FBS (CSS) (12676, lot 1176965; Life Technologies), as described previously [
17], in phenol red-free improved minimum essential medium (IMEM) + 10 % CSS (2 % CSS for SUM44PE only). This single lot of CSS was used for all experiments and was confirmed to have complete hormone deprivation [
19].
17β-Estradiol (E2) and 4-hydroxytamoxifen (4-OHT) were obtained from Sigma-Aldrich (St. Louis, MO, USA); other compounds were obtained from Tocris Biosciences (Bristol, UK). E2, 4-OHT, ICI 182,780 (ICI), and progesterone (P4) were dissolved in ethanol; RU486 (RU, mifepristone), BMS-345541 (BMS), staurosporine (STS), endo-IWR1 (IWR), and JW 67 (JW) were dissolved in dimethyl sulfoxide.
Proliferation and viability assays
For cellular proliferation assays, we used the FluoReporter double-stranded DNA quantitation kit (F2692; Life Technologies) according to the manufacturer’s instructions. Cell death was assessed using CellTox Green (G8741; Promega, Madison, WI, USA) according to the manufacturer’s instructions. For each assay, cells were plated in 96-well plates and allowed to attach overnight prior to the indicated treatments. Fluorescence was assessed using a VICTOR X4 plate reader (PerkinElmer, Waltham, MA, USA). For assays, points and/or bars represent the mean of five or six biological replicates ± SD.
RNA interference
Small interfering RNAs (siRNAs) were reverse-transfected using Lipofectamine RNAiMAX reagent (Life Technologies) according to the manufacturer’s instructions. A list of constructs used in this study is available in Additional file
2. Notably, the efficacy of
WNT4 knockdown varied across commercially available constructs. The extent of knockdown correlated with effects on growth (Additional file
3: Figure S2). The reagent indicated (Additional file
2) outperformed other reagents tested (additional details available on request).
Gene expression analyses
For RNA extractions, we used the illustra RNAspin Mini Kit (GE Healthcare Life Sciences, Little Chalfont, UK) or the RNeasy Mini Kit (QIAGEN, Hilden, Germany). For complementary DNA conversion, we used iScript master mix (Bio-Rad Laboratories, Hercules, CA, USA), and for quantitative PCR (qPCR) reactions, we used SsoAdvanced SYBR Green Master Mix (Bio-Rad Laboratories) on a CFX384 thermocycler (Bio-Rad Laboratories), according to the manufacturer’s instructions. Expression data were normalized to
RPLP0. Primer sequences are available in Additional file
2.
Chromatin immunoprecipitation
Cells were hormone-deprived as described above prior to treatment with 0.1 % EtOH, 1 nM E2, 1 μM ICI, 100 nM P4, or 1 μM RU486 for 45 minutes. ChIP experiments were performed as described previously [
20] with minor modifications:
1.
Nuclei were extracted prior to sonication by resuspending the fixed cell pellet in nuclei preparation buffer (5 mM 1,4-piperazinediethanesulfonic acid (PIPES), 85 mM KCl [pH 8.0] + 0.5 % Nonidet P-40 + protease inhibitor) with rotation at 4 °C for 30 minutes. Nuclei were then pelleted and lysed and/or sonicated as described.
2.
SDS was omitted from buffers Tris-sucrose-ethylenediaminetetraacetic acid I (TSEI) and TSEII.
3.
Carrier molecules were added during immunoprecipitation [
21].
In immunoprecipitation experiments, we used ERα (HC-20) and rabbit immunoglobulin G (sc2027) antibodies (Santa Cruz Biotechnologies, Dallas, TX, USA). PCR was performed as described above, normalized to percentage input. Primer sequences are available in Additional file
2.
Long-term estrogen deprivation
Endocrine-resistant variants of MM134 and 44PE were generated by maintaining cells in hormone-deprived conditions using IMEM + 10 % CSS. As 44PE cells are only modestly hormone-responsive and their basal medium has minimal hormone content, we first subderived cells in fully hormone-replete conditions by maintaining SUM44PE as described for MM134 (DMEM/L-15 + 10 % FBS) for 3 months. The resulting variant, termed SUM44/F, has an increased proliferative response to E2 (Additional file
4: Figure S3). To generate ILC-LTED lines, MM134 and SUM44/F were hormone-deprived as described above and subsequently plated in a 6-well plate. Each well was maintained independently over 6–12 months until cells could be passaged routinely; this generated four LTED MM134 and two LTED SUM44/F lines.
Immunoblotting
SDS-PAGE was performed using standard methods. Proteins were transferred to polyvinylidene difluoride (PVDF) membranes for Western blot analysis using chemiluminescence detection. Antibodies were used according to the manufacturers’ recommendations: ERα (clone 6 F11; Leica Biosystems, Buffalo Grove, IL, USA), p65 (8242; Cell Signaling Technology, Danvers, MA, USA), phospho-p65 (serine 536, CS 3033; Cell Signaling Technology), RelB (CS 4922; Cell Signaling Technology), c-Rel (CS 4727; Cell Signaling Technology), NFKB1 (p105/p50, CS 12540; Cell Signaling Technology), p21 (CS 2946; Cell Signaling Technology), DVL2 (CS 3216; Cell Signaling Technology), DVL3 (CS 3218; Cell Signaling Technology), and tubulin (T9026; Sigma-Aldrich).
Transcription factor response element reporter assays
A targeted screen for transcription factor activity was performed using the Cignal 45-Pathway Reporter Array System (QIAGEN). Plasmids were reverse-transfected using Attractene (QIAGEN). The following day, cells were treated with 0.01 % EtOH or 100 nM ICI. All conditions were performed in biological triplicate. Cells were assayed for reporter activity 42 h posttreatment using the Dual-Luciferase Reporter Assay System (Promega).
For canonical Wnt signaling reporter assays, we used TOP and Renilla plasmids, a kind gift from the Monga laboratory (University of Pittsburgh). Wnt expression plasmids were obtained from the Open Source Wnt Project (Addgene, Cambridge, MA, USA). The plasmid kit was a gift from Marian Waterman, David Virshup, and Xi He (kit 1000000022; Addgene). Plasmids were cotransfected using Lipofectamine LTX and PLUS reagent (Life Technologies). Cells were assayed for reporter activity 24 h posttransfection using the Promega Dual-Luciferase Reporter Assay System.
Statistical analyses
Curve-fitting and statistical analyses for in vitro studies were performed using Prism version 5.04 software (GraphPad Software, La Jolla, CA, USA). For in silico analyses, we used expression values derived from The Cancer Genome Atlas (TCGA) [
22] breast cancer cases and normal tissue (in units of transcripts per million) downloaded from the Gene Expression Omnibus database [GEO:GSE62944] [
23]. PAM50 subtypes for the TCGA tumors were defined using the “genefu” package in R (version 2.4.2). Briefly, 50:50 distributions of ER
+/ER
− tumors were sampled for the median centering step, and subtypes were assigned to all tumors. This process was repeated 100 times, and the consensus subtype for each tumor was taken. Molecular Taxonomy of Breast Cancer International Consortium (METABRIC) [
24] data were downloaded from the Synapse software platform (syn1688369; Sage Bionetworks, Seattle, WA, USA). Microarray probes were selected for individual genes based on the probe set with the highest interquartile range.
Discussion
ILCs typically present with clinical biomarkers consistent with endocrine responsiveness, and nearly all patients with ILC are treated with adjuvant endocrine therapy [
1,
36]. However, recent retrospective clinical trial data suggest that at least a subset of patients with ILC may have poor outcomes with endocrine therapy [
6,
7], suggesting that, compared with IDC, ER biology and signaling may be unique in ILC cells. Consistent with this, our report on endocrine response in ILC model systems demonstrated that ER signals via unique transcriptional targets in ILC cells [
8], which we hypothesized would mediate endocrine response in ILC. We identified the Wnt ligand WNT4 as a putative novel effector of ER signaling specifically in ILC cells, and we demonstrate in the present study that WNT4 is a driver of endocrine response and resistance in ILC (Fig.
6d).
In the murine mammary gland, Wnt4 is critical in pregnancy-induced ductal elongation and branching [
10], as well as in maintenance of the progenitor cell niche [
11‐
13]. In these contexts,
WNT4 serves as an effector of PR signaling and is directly regulated by P4/PR. We observed PR regulation of
WNT4 in the PR-positive BCCL T47D (Fig.
2a), but elucidating the mechanism that places
WNT4 under ER control in ILC is an important future direction for research. Interestingly, the EREs at the WNT4 ERBS are canonical half-EREs but deviate from the consensus full-ERE sequence (Additional file
1: Figure S1); thus, a specific transcription factor context may be required to access and/or use this site in ILC. Interestingly, potential ILC-specific modifiers of ER function were reported in recent large-scale genomic studies in which researchers identified differential expression and mutation of
FOXA1 vs
GATA3 [
3] and amplification of
ESR1 [
5]. However, we have not detected these aberrations in MM134 or 44PE (T. Du, K. Levine, M. J. Sikora, et al., unpublished manuscript; also see [
22]), and thus
WNT4 regulation is likely mediated by other factors. Putative factors from murine tissues where
Wnt4 is hormone-regulated include Foxo1 [
37], Foxc2 [
38], Wt1 [
39,
40], Mta3 [
41], MED1 [
42], and Egr1 [
43]. Another putative cofactor may be YAP and/or TAZ, which cross-talk with both canonical and noncanonical Wnt signaling [
44,
45]; nuclear (active) YAP is elevated in ILC tumors [
46]. Finally, the activation of Oct-4 observed in ILC-LTED (Fig.
5a) paralleled
WNT4 expression in ILC-LTED cells (i.e., ICI sensitivity in 44:LTED), and thus Oct-4 may play a role in maintaining ER regulation of WNT4 in 44:LTED. Oct-4 may also connect differentiation or progenitor state to
WNT4 regulation. Perhaps consistent with this,
Wnt4 in the pubertal murine mammary gland is in fact modestly E2-induced, but it is strictly P4-induced in the adult gland [
14]. Though these observations are intriguing, the Oct-4 response element used is likely promiscuous across Oct/POU family members, and thus Oct signaling in ILC-LTED may be tied to a number of related transcription factors, though Oct-4 itself has recently been implicated in endocrine resistance [
47].
Noncanonical Wnt4 signaling pathways have been examined in murine tissues but are greatly context-dependent. β-Catenin-independent WNT4 signaling has not been extensively characterized in the breast or in breast cancer. In ILC cells, we identified
CDKN1A/p21 as a novel WNT4 signaling target and demonstrated that
CDKN1A/p21 regulation is a critical component of WNT4-mediated growth (Fig.
6). p21 has previously been shown as a direct transcriptional repressor of
Wnt4 [
48], but we observed the converse, that WNT4 is an upstream regulator of
CDKN1A. Though our transcription factor screen did not identify a putative effector of WNT4 to suppress
CDKN1A, a number of factors and/or pathways have been reported downstream of Wnt4 in murine tissues and thus may be functioning in ILC, including p38/Jnk [
49], SF-1(NR5A1) [
50,
51], EAF1 and EAF2 [
52,
53], Runx-1 [
54], and Fst [
55]. Yu et al. also demonstrated that noncanonical Wnt4 signaling could block ovariectomy-induced osteoporosis via inhibition of receptor activator of nuclear factor kB ligand-induced NF-kB signaling [
56] (though we did not observe changes in NF-kB signaling upon siWNT4 in ILC-LTED; Fig.
5d and e). Additionally, Wnt4 regulates steroidogenesis in ovarian and adrenal models [
57‐
59], which may have significant implications should this also be true in ILC. However, WNT4 signaling is clearly multifaceted, as siCDKN1A only partially rescued the growth suppression by siWNT4 (Fig.
6c). Identification of additional WNT4 target genes in ILC, as well as the WNT4 receptor and downstream signaling components, is a critical future direction for research.
Understanding the mechanism by which WNT4 activates its signaling cascade (i.e., in an autocrine vs paracrine mechanism) is an important future direction for research. A key observation from
Wnt4-transgenic mice mammary gland studies was that though
Wnt4 knockout ablated ductal elongation and branching [
10], the converse was not true;
Wnt4 overexpression did not induce hyperplasia or tumorigenesis [
29]. However, researchers in an earlier study did observe that
Wnt4 overexpression induced mammary hyperplasia [
60], and, taken together, these studies highlight the potential importance of the specific cell population expressing Wnt4 (discussed in [
29]). Interestingly, studies that have demonstrated the role of Wnt4 in maintenance of the progenitor cell niche [
11‐
13] clearly showed a paracrine role for Wnt4, wherein PR-positive, Wnt4-expressing cells secrete Wnt4 to activate signaling in neighboring hormone receptor-negative cells. Additionally, regulation of
Wnt4 may be modified during pregnancy, wherein E2 and P4 may cooperate to induce paracrine Wnt4 signaling [
61]. It is unclear whether similar mechanisms may be maintained in ILC cells, whether ILC-derived WNT4 can signal in a paracrine mechanism with the tumor microenvironment, or whether WNT4 operates in a cell-autonomous vs nonautonomous manner to drive proliferation of tumor cells.
Studies examining endocrine resistance mechanisms specifically in ILC are in their infancy [
1,
62]. Beyond our LTED models, only one other ILC acquired endocrine resistance model has been characterized: SUM44/LCCTam, which is a tamoxifen-resistant variant of 44PE [
63]. Intriguingly,
WNT4 is among the top 50 differentially expressed genes between LCCTam and 44PE (upregulated more than twofold in LCCTam vs 44PE [GEO:GSE12708]), suggesting that WNT4 may be a common mechanism of acquired endocrine resistance in ILC cells. We further characterized WNT4-mediated endocrine resistance in ILC-LTED and identified that WNT4 signaling is maintained via activation of NF-kB signaling in 134:LTED. Activation of NF-kB is a driver of antiestrogen resistance in MCF-7 models of acquired [
64], Akt-driven [
65], and HER2-driven [
66] resistance. In these contexts, ER is a repressor of NF-kB activity [
67], and loss of canonical ER activity (and parallel loss of chicken ovalbumin upstream promoter transcription factor II [
68]) leads to reactivation of NF-kB. This inverse correlation between ER and NF-kB activity has also been observed in patient tumor samples (reviewed by Sas et al. [
69]). However, it does not appear that NF-kB activity is downstream of ER in ILC cells, as ICI treatment had minimal or no effect on reporter output in ILC parental or LTED cells (Fig.
5a). Thus, though NF-kB may be a shared endocrine resistance mechanism in breast cancers, the mechanism of activation and potentially its signaling may differ in IDC vs ILC. The presence of NF-kB/Rel binding sites at the WNT4 ERBS (Additional file
10: Figure S9b) suggests that
WNT4 may be a direct target of NF-kB signaling in ILC, and future studies will elucidate the context required for NF-kB to regulate
WNT4.
The increased expression of
WNT4 in ER-positive breast tumors is consistent with the role of WNT4 in mediating hormone response in the normal mammary gland, and with our observations regarding the role of WNT4 in endocrine response in ILC cells. Though we observed ER regulation of WNT4 specifically in ILC cells, the association of
WNT4 expression with PR status across ER-positive tumors suggests that PR may regulate
WNT4 in IDC, as we observed in T47D cells, or that
WNT4 may be a marker of functional ER signaling. Importantly, the expression data currently available represent static, pretreatment measurement of
WNT4 expression in breast tumors; regulation of
WNT4 expression following endocrine therapy may be a superior biomarker for ILC biology. Though gene expression data from ILC tumors following neoadjuvant letrozole therapy have been reported [
70], the expression of many ILC-specific ER target genes [
8], including
WNT4, were excluded from the analyses due to issues related to the use of multiple expression platforms. Future analyses of
WNT4 regulation may be possible on the basis of ongoing studies such as POETIC [
71] or our neoadjuvant trial for patients with ILC (ClinicalTrials.gov identifier NCT02206984 [
72]).