Estrogen controls mTOR signaling and mitochondrial function via WNT4 in lobular carcinoma

Invasive lobular carcinoma of the breast (ILC) is strongly estrogen-driven, and represents a unique context for estrogen receptor (ER) signaling. In ILC, ER controls the expression of the Wnt ligand WNT4, which is critical for endocrine response and anti-estrogen resistance, yet signaling mediated by WNT4 is poorly understood. We utilized reverse phase protein array (RPPA) to characterize ER and WNT4-driven signaling in ILC cells, and identified WNT4 as a mediator of downstream mTOR signaling via p70-S6K. Independent of mTOR/p70-S6K, ER and WNT4 control levels of MCL-1, which is associated with mitochondrial function. In this context, knockdown of WNT4 caused accumulation of reactive oxygen species and decreased ATP production that precede cell death. WNT4 regulation of both mTOR signaling and MCL-1 levels was also observed in anti-estrogen resistant models of ILC. Further, we identified that high WNT4 expression is associated with similar mTOR pathway activation in serous ovarian cancer tumors, suggesting that this WNT4 pathway is important in multiple tumor types. The identified downstream pathways represent potential targets to inhibit WNT4 signaling in ovarian cancer and overcome anti-estrogen resistance for patients with ILC.


Introduction
Invasive lobular carcinoma of the breast (ILC), the second most common histologic subtype of breast cancer [2][3][4], appears to be exquisitely sensitive to and dependent on the steroid hormone estrogen.
Estrogen-based hormone replacement therapy (with or without progestin) more strongly increases the incidence of ILC versus the more common invasive ductal carcinoma (IDC)(reviewed in [3]), and nearly all ILC tumors express ER [4][5][6]. Further, ILC biomarkers are consistent with the hormone-dependent 'Luminal A' molecular subtype [4], supporting the current paradigm that patients with ILC are ideal candidates for treatment with anti-estrogen therapies. However, retrospective studies suggest patients with ILC may not receive similar benefit from anti-estrogens as patients with IDC, ie. poorer outcomes with adjuvant tamoxifen [7,8] and poorer long-term outcomes >5-10 years post-diagnosis [9][10][11][12].
Consistent with clinical observations, we identified tamoxifen-resistance in ER-positive ILC models, including de novo ER partial agonism by anti-estrogens [13,14]. These data raise the question of whether ER function and signaling is unique in the context of ILC.
Studies using laboratory models of ILC and tumor profiling through the Cancer Genome Atlas support that some aspects of ER signaling are distinct in ILC versus IDC cells. TCGA analyses [4] specifically compared Luminal A ILC to Luminal A IDC, and reverse-phase protein array (RPPA) data identified differential activity of PTEN and downstream Akt signaling. Additionally, each of the three transcriptional subtypes of ILC (Reactive, Immune, and Proliferative) showed distinct signaling features in RPPA analyses (eg. high c-kit; high STAT5; high DNA repair protein signature; respectively).
Similarly, we reported that ER regulates unique target genes in ILC cells via distinct ER DNA binding patterns [13]. These unique targets mediate ILC-specific signaling pathways driven by ER that are critical for endocrine response and resistance, eg. estrogen-driven atypical Wnt signaling via WNT4 [15,16]. However, our understanding of ER-driven signaling at the protein level in ILC cells remains limited, as studies to date either cannot define dynamic changes caused by ER activation (ie. are from static samples as in TCGA), or are focused on the ER-driven transcriptome. Proteomic studies in ILC with estrogen or anti-estrogen treatment are needed to better understand ER-driven signaling in ILC.
We identified WNT4 as a critical signaling molecule induced by ER specifically in ILC, but the mechanism by which WNT4 engages downstream signaling is unknown. WNT4 is unique among Wnt proteins in its diverse roles in either activating or suppressing canonical or non-canonical Wnt signaling in a tissue-specific manner (discussed in [16]). In the normal mammary gland, WNT4 is induced by progesterone in progesterone receptor (PR)-positive luminal epithelial cells, then secreted to act in a paracrine manner to activate canonical β-catenin-dependent signaling in neighboring myoepithelial cells [17][18][19][20]. In contrast, WNT4 is under direct control of ER, not PR, in ILC cell lines [13,15].
Canonical Wnt signaling is also dysfunctional in ILC cells; hallmark genetic loss of E-cadherin (CDH1) causes catenin protein dysfunction including destabilization and loss of β-catenin protein in both cell lines and tumors [4,15]. Further, we recently reported that WNT4 secreted from ILC cells (and other cancer models) is dysfunctional for paracrine activity, but instead activates signaling by a cellautonomous, intracellular mechanism (ie. "atypical" Wnt signaling [16]). Taken together, defining ERdriven signaling and novel WNT4-driven signaling in ILC is necessary to both understand ILC biology and identify new treatment targets for ILC.
To address these gaps in our understanding of ER function in ILC, we used RPPA analyses of ILC models to characterize ER-driven signaling in ILC, and to determine the role of WNT4 in mediating ERdriven signaling. These studies identified that estrogen activates distinct protein signaling pathways in ILC cells, including a distinct component of PI3K/Akt/mTOR signaling in ILC versus IDC cells. A subset of ER-driven signaling in ILC cells required WNT4, most notably mTOR activity and regulation of mitochondrial function. These observations led us to further examine the mechanisms by which ERdriven WNT4 signaling mediates cell proliferation and survival, and endocrine response and resistance, in ILC cells.

Materials and Methods
Cell culture MDA MB 134VI (MM134) and SUM44PE (44PE) were maintained as described [13]. MCF-7 and HCC1428 were maintained in DMEM/F12 (Corning Life Sciences, cat#10092CV) supplemented with 10% fetal bovine serum (FBS; Nucleus Biologics, cat#FBS1824). Hormone-deprivation was performed as described [21] in IMEM (Gibco/ThermoFisher, cat#A10488) supplemented with 10% charcoalstripped fetal bovine serum (CSS; prepared as described [21] with the same FBS as above). WNT4overexpressing models were previously described [16], and cultured in the same conditions as parental cell lines. Long-term estrogen deprived (LTED) model establishment and culture conditions were previously described [15]. All lines were incubated at 37°C in 5% CO 2 . Cell lines are authenticated annually via the University of Arizona Genetics Core cell line authentication service and confirmed to be mycoplasma negative every four months. Authenticated cells were in continuous culture <6 months.

Reverse phase protein array (RPPA)
Cells were hormone-deprived prior to reverse transfection with 10nM siRNA as described above. 24 hours post-transfection, cells were treated with vehicle (0.01% EtOH) or 100pM E2 and harvested 24 hours later (48h post-transfection; 24h post-treatment). Cells were lysed according to core facility instructions (see below, [22] Multiple testing correction was applied using the Benjamini-Hochberg method. Raw/normalized RPPA data and treatment comparisons will be deposited to an OSF page upon formal publication [1], or will be provided upon request.

Immunoblotting
Whole-cell lysates were obtained by incubating cells in RPPA lysis buffer (above) for 20' on ice. Cells were centrifuged at ~16,000xg for 10m at 4°C and the resulting supernatant was collected for analysis.
Protein concentration was measured using the Pierce BCA Protein Assay Kit (#23225). Standard methods were used to perform SDS-PAGE. Proteins were transferred onto PVDF membranes.

Cell proliferation and cell cycle analyses
Total double-stranded DNA was measured as a proxy for total cell number by hypotonic lysis of cells in ultra-pure H 2 O, followed by addition of Hoechst 33258 (ThermoFisher Scientific, #62249) at 1μg/mL in Tris-NaCl buffer (10mM Tris, 2M NaCl; pH 7.4) at equivalent volume to lysate. Fluorescence (360nm ex / 460nm em) was measured on a Bio-Tek Synergy 2 microplate reader. For cell cycle analyses, cells were fixed in 70% EtOH overnight at 4°C. After fixation, cells were treated with RNase A, and stained with 50μg/mL propidium iodide + 0.1% Triton X-100 in PBS overnight at 4°C. PI intensity was assessed on a Gallios Flow Cytometer (Beckman Coulter) at the U. Colorado Cancer Center Flow Cytometry Core Facility; at least 10,000 events were counted per sample.

Quantitative PCR analyses
RNA extractions were performed using the RNeasy Mini kit (Qiagen); mRNA was converted to cDNA on an Eppendorf Mastercycler Pro (Eppendorf) and using Promega reagents: Oligo (dT) 15  Expression data were normalized to RPLP0. Primer sequences were published previously [13,15].

Metabolic analyses
Cellular reactive oxygen species (ROS) were measured by CM-H 2 DCFDA (ThermoFisher, cat#C6827) fluorescence. Cells were reverse transfected as described above, and 24h later loaded with 5μM dye per manufacturer's instructions. 24h later (ie. 48h post-transfection), fluorescence-positive cells were identified using an Incucyte Zoom live cell imaging platform (Essen Bioscience) and normalized to total cellular confluence. For analysis of ATP per cell, cells were reverse transfected as above, and at the indicated timepoint, parallel plates were assessed for total dsDNA and total ATP (Cell Titer-Glo, Promega). Cell number by each method was normalized to a standard curve, and analyzed as a ratio of (Cell Number by ATP)/(Cell Number by dsDNA).

TCGA data analyses
TCGA data were accessed via the cBio portal in March-June 2019; provisional datasets were used for all analyses described. Z-score cutoffs for WNT4 expression (RNAseq) were selected based on the minimum z-score needed to exclude any low WNT4-expressing tumors. All statistical analyses were derived from cBio analysis tools, ie. Enrichments>Protein>RPPA.
Of note, estrogen suppressed total histone H3 in ILC (and two H3 post-translational modifications, reflective of the total H3 change). This represents a decrease in the pool of soluble non-nucleosomal H3, consistent with chromatin remodeling [24][25][26][27], rather than a change in total cellular H3 (Supplemental Figure 1, also see RPPA lysis conditions in Materials and Methods).
Since ILC-specific E2-regulated targets were primarily related to PI3K-mTOR signaling, we examined all PI3K-related proteins on the RPPA that were changed by E2 treatment in any cell line ( Figure 1D) to identify components of the signaling cascade in ILC. Based on the observed estrogen-induced phosphorylation of mTOR, p70-S6K (RPS6KB1), and S6 in ILC cells (Figure 1), we expected to see an upstream increase in phosphorylated Akt [28]. However, estrogen-induced Akt phosphorylation was only observed in MCF-7 ( Figure 1D), and levels of phospho-and total Akt were extremely low in ILC cells ( Figure 1E). Additionally, Akt-mediated phosphorylation of PRAS40 was only induced by estrogen in MCF-7, not in ILC cells, suggesting mTOR activation is mediated by another pathway in ILC cells.
Since non-canonical Wnt signaling has previously been linked to Akt-independent mTOR activation [29], we further investigated the role of WNT4 in estrogen-regulated signaling and mTOR activity in ILC cells. We compared hormone-deprived ILC cells treated with estrogen (as above) transfected with nontargeting versus WNT4-targeting siRNA (siNT vs siWNT4, respectively). Estrogen-regulation of WNT4 and siWNT4 efficacy were confirmed by western blot (Figure 2A). Consistent with the critical role of WNT4 in ILC cell proliferation and survival [15], siWNT4 broadly dysregulated overall signaling (Supplemental Figure 2A-B). We compared siWNT4-mediated signaling changes to estrogenregulated signaling in ILC cells, and identified 9 targets for which estrogen-regulated changes were ablated by WNT4 knockdown (Figure 2B). WNT4 knockdown blocked estrogen-mediated FOXM1 induction, S6 phosphorylation, MCL1 depletion, and soluble Histone H3 depletion in ILC cells; siWNT4 also increased phospho-ER (consistent with relieved negative feedback from mTOR signaling [30]). siWNT4 did not have similar effects on these targets in IDC cell lines MCF-7 or HCC1428 ( Figure 2C, Supplemental Figure 2C). HCC1428 expresses high WNT4 mRNA similar to that observed in ILC cell lines but does not depend on WNT4 for proliferation or survival [15], supporting that the identified signaling targets represent ILC-specific WNT4 signaling. Based on these data, we hypothesized that WNT4 mediates ILC cell proliferation and survival via regulation of FOXM1 and mTOR.

FOXM1 is not a direct target of WNT4 signaling
FOXM1 is a critical mediator of estrogen-driven cell cycle progression [31], has been linked to control of ER-driven transcription (in MCF-7 cells), and is a potential driver of anti-estrogen resistance [31][32][33].
This led us to examine whether FOXM1 is a critical WNT4 signaling target. Though FOXM1 siRNA suppressed the proliferation of MM134 cells (Supplemental Figure 3A), this growth suppression was not equivalent to that observed with anti-estrogens or with WNT4 siRNA. Whereas siWNT4 parallels treatment with fulvestrant and causes a G1 arrest, siFOXM1 causes a G2/M arrest (Supplemental Figure 3B-C), consistent with established roles of FOXM1 in progression through mitosis [34]. This lack of phenocopy between siWNT4/anti-estrogen and siFOXM1 suggested that decreased FOXM1 levels upon WNT4 knockdown may be indicative of knockdown-induced cell cycle arrest, rather than identifying FOXM1 as a direct WNT4 target. To confirm this, we examined whether CDKN1A/p21 knockdown, which partially restores proliferation after siWNT4 in ILC cells [15], would restore FOXM1 levels (ie. since p21 knockdown partially releases the G1 arrest, FOXM1 levels would increase despite WNT4 knockdown). In both MM134 and 44PE, siCDKN1A partially restored the FOXM1 decrease caused by siWNT4 (Supplemental Figure 3D), supporting that FOXM1 protein level is acting as a marker of cell cycle and G2/M progression. Consistent with a lack of a direct role in WNT4-driven cell cycle progression, we also did not observe any suppression of ER-driven transcription in MM134 by siFOXM1 (Supplemental Figure 3E). These data suggest that FOXM1 is not a direct target of ER and WNT4 in ILC, but is an indirect target of ER/WNT4-driven cell cycle progression.

WNT4 regulates mTOR signaling downstream of mTOR kinase activity
Our RPPA data showed estrogen treatment of ILC cells led to activation of mTOR activity (e.g. 4EBP1, p70S6K, and S6 phosphorylation), but siWNT4 specifically reduced S6 phosphorylation without altering p70S6K phosphorylation (T389). We confirmed via immunoblotting that targeting ER, but not WNT4, reduced p70S6K-pT389; this phosphorylation was mTOR-dependent, as treatment with everolimus completely blocked T389 phosphorylation ( Figure 3A). Subsequent phosphorylation of S6 was reduced by siWNT4 or by inhibiting ER, mTOR, or p70S6K (S6K1 inhibitor PF-04708671 [35]). These data suggest that WNT4 mediates S6K1 activity downstream of mTOR phosphorylation of S6K1. Consistent with this, WNT4 over-expression modestly delayed the loss of S6K1 and S6 phosphorylation caused by everolimus treatment, but ultimately downstream mTOR signaling was completely suppressed by everolimus ( Figure 3B). WNT4 over-expression in either MM134 or 44PE did not change the effect of everolimus on cell proliferation ( Figure 3C). Notably, though WNT4 knockdown strongly suppresses proliferation of both MM134 and 44PE, these models were resistant vs sensitive, respectively, to treatment with single-agent everolimus. Taken together, WNT4 is at least in part necessary for S6K1 activity (ie. phosphorylation of S6) but S6K1 remains wholly dependent on mTOR kinase activity and thus sensitive to mTOR inhibition. A necessary-but-not-sufficient role for WNT4 in this pathway is consistent with similar characterization of WNT4 in the mammary gland [17,36], and also suggests that other ER:WNT4 targets are critical mediators of proliferation and survival in ILC cells.

ER:WNT4 regulation of MCL1 is associated with metabolic dysregulation
Estrogen caused a decrease in levels of the BCL2-family protein MCL1, and this was reversed by WNT4 knockdown (Figures 1-2). ER regulation of MCL1 likely occurs post-transcriptionally, as qPCR determined that MCL1 gene expression is not regulated by estrogen in ILC cells (Supplemental Figure   4A). Translational regulation of MCL1 has been previously linked to mTOR, with activation of mTOR signaling associated with increased MCL1 levels [37]. However, our data show that estrogen activates mTOR while suppressing MCL1, and targeting mTOR signaling does not similarly affect MCL1 levels ( Figure 3D). Thus ER:WNT4 regulation of MCL1 is post-transcriptional but independent of mTOR signaling.
Given the canonical role of MCL1 as an anti-apoptotic BCL2-family protein [38], estrogen-driven reduction in MCL1 levels contrasts the important roles of ER:WNT4 in mediating cell survival [15].
However, reduction of MCL1 protein has been identified as critical to facilitate post-partum involution of uterine tissue [39]. Based on this, we hypothesized that estrogen-mediated repression of MCL1 may "prime" cells for apoptosis ( [40]; ie. facilitate involution), but that anti-estrogens (inducing MCL1) would make ILC cells resistant to apoptosis. To test this, we treated ILC cells with BH3 mimetics alone or in combination with fulvestrant, or combined with MCL1-specific BH3 mimetic A-1210477. Though blocking MCL1 could moderately sensitize cells to other BH3 mimetics, fulvestrant caused no shifts in sensitivity to any tested combination of BH3 mimetics (Supplemental Figure 4B). This suggests ER:WNT4 regulation of MCL1 levels is not related to control of intrinsic apoptosis. MCL1 also plays a critical role in regulating mitochondrial dynamics [41][42][43], and based on this we hypothesized that ER:WNT4 signaling may regulate mitochondrial function and metabolism in ILC cells. Consistent with this, siWNT4 causes both an increase in intracellular reactive oxygen species (ROS)( Figure 3E) and a decrease in total ATP content per cell ( Figure 3F). These effects were not observed with siESR1, which parallels our prior observations that WNT4 knockdown but not inhibition of ER induces cell death in ILC cells [15]. These data suggest that regulation of MCL1 is related to a critical role for WNT4 in mitochondrial metabolism.

WNT4 signaling is re-activated during anti-estrogen resistance in ILC cells
We previously reported that long-term estrogen deprived (LTED; mimicking AI-resistance) variants of ILC cell lines remain dependent on WNT4 [15], and hypothesized that WNT4-driven pathways identified in the parental cells would also be active in LTED models. LTED models 44:LTED/A and 134:LTED/E (derived from 44PE and MM134, respectively [15]) were analyzed by RPPA and compared to hormonedeprived parental cells to identify signaling adaptations during LTED. LTED cells were also transfected with siNT or siWNT4 as above to identify WNT4-driven signaling during LTED.
The majority of protein changes in LTED v parental comparisons were shared between models ( Figure   4A, Supplemental Figure 5A). Of note, this included upregulation of FASN, confirming the transcriptional upregulation of FASN and lipid metabolism genes in these models that we recently reported (Supplemental Figure 5B) [44]. RPPA changes specific to 44:LTED/A (increased ER and OCT4) vs 134:LTED/E (decreased ER, increased phospho-NFκB) (Supplemental Figure 5B) also confirmed model-specific features of LTED that are linked to WNT4 upregulation as we reported previously [15]. Comparing protein changes in LTED models versus those caused by WNT4 knockdown (Figure 4B-C) identified 14 protein changes that are induced during LTED and mediated by WNT4 signaling (Figure 4C-D). WNT4-driven signaling in LTED included key ER:WNT4 targets described above, eg. S6 phosphorylation and MCL1 suppression, which were confirmed by immunoblotting ( Figure 4E). These WNT4-driven signaling changes were also confirmed in our additional ILC:LTED cell lines (Supplemental Figure 6). Taken together, WNT4 regulation of downstream pathways, eg. cell cycle, mTOR signaling, and metabolism, remain active and critical to the LTED phenotype in ILC cells.

Atypical WNT4 signaling functions in diverse tumor types
To determine whether similar WNT4-driven signaling functions in tumor tissues (including other than ILC), we explored data from the Cancer Genome Atlas (TCGA). We examined tumors related to tissues that require WNT4 in development (eg. kidney, adrenal, lung, ovary, uterus [45][46][47][48]), and identified tumors that over-expressed WNT4 mRNA (Figure 5A). Comparing WNT4 high versus other tumors, we looked for differences in protein signaling in RPPA data (reverse phase protein array). Renal clear cell carcinoma (RCC; total n=515, high WNT4 n=89), lung adenocarcinoma (LuA; total n=533, high WNT4 n=57), and serous ovarian cancer (OvCa; total n=307, high WNT4 n=42) presented differences in protein signaling by RPPA in WNT4 high tumors (Supplemental Figure 7A). Of note, other tumor types including ILC were limited by smaller sample sizes with RPPA data (eg. ILC, total n=160, high WNT4 n=29). Comparing signaling changes in RCC, LuA, and OvCa (Figure 5B), OvCa was unique in having an upregulation of PI3K/mTOR pathway phospho-proteins in high WNT4 tumors (Supplemental Figure 7A-B). Shared signaling changes were limited; differentially regulated proteins in all 3 tumors types (n=10) were not similarly induced/repressed in all three tumors, with the exception of high WNT4 being associated with decreased ER levels (Supplemental Figure 7C). Changes in total β-catenin in high WNT4 tumors was seen in LuA and RCC, but not OvCa (Supplemental Figure 7D), suggesting that WNT4 in OvCa may activate non-canonical signaling that converges on mTOR as in ILC. Similar to our observations in ILC cell lines, high WNT4 in OvCa was associated with increased phosphorylation of 4EBP1, p70-S6K, and S6 ( Figure 5C). Further, high WNT4 expression in OvCa was associated with shorter overall survival (median OS, 28.5mo v 45.0mo; Figure 5D). Though RPPA data for ILC tumors were too limited to identify associations with active mTOR signaling, high WNT4 was associated with decreased PTEN protein levels, consistent with activated downstream signaling (Supplemental Figure   7E). Additionally, high WNT4 was associated with decreased phospho-Akt-S473, paralleling low phospho-Akt levels in our cell line data. These data suggest that non-canonical WNT4 signaling via the mTOR pathway may be critical in multiple tumor types, including ILC and OvCa.

Discussion
Clinical and laboratory data support that ILC represents a unique context for estrogen receptor signaling, which may be mediated in part by ER induced WNT4 expression. The Wnt ligand WNT4 plays a critical role in the normal development of the mammary gland, and plays a similar role in endocrine response and resistance in ILC [15]. However, given the myriad tissue-specific pathways targeted by WNT4, and dysfunctional β-catenin-dependent signaling in ILC [4,15,16], it is unclear how WNT4 mediates proliferation and survival in ILC cells. To address this, we used RPPA analyses to profile ER-driven signaling in ILC cells, identify ER-driven signaling that requires WNT4 (ie. ER:WNT4 signaling), and assess the activity of these pathways in anti-estrogen resistance. Our studies identified that in ILC cells, ER:WNT4 signaling mediates downstream activity of mTOR signaling via p70-S6K, as WNT4 knockdown suppresses S6 phosphorylation. We also identified that ER:WNT4 signaling led to a suppression of total MCL1 levels, which was not associated with differential sensitivity to pro-apoptotic dugs, but instead with metabolic dysfunction upon WNT4 knockdown. Parallel signaling mediated by WNT4 was identified in anti-estrogen-resistant models of ILC, and related signaling pathways were also linked to WNT4 overexpression in serous ovarian cancer. These data provide new insight in to a poorly understood Wnt signaling pathway, and identify potential downstream signaling pathways that may be targetable to inhibit cell proliferation and survival mediated by WNT4.
Though WNT4 itself has not been mechanistically linked directly to mTOR signaling previously, βcatenin-independent Wnt signaling was shown by Inoki et al to regulate mTOR activity via GSK3, independent of Akt phosphorylation [29]. Since we did not detect E2-induced AKT phosphorylation but GSK3 phosphorylation was strongly E2-induced in ILC cells (Figure 1), this may represent the pathway linking WNT4 to mTOR. GSK3 phosphorylation was also among the most strongly increased signaling events in high WNT4-expressing OvCa (Supplemental Figure 7B). However, GSK3 phosphorylation did not appear to be affected by WNT4 knockdown in ILC cells by RPPA (Figure 2, Supplemental   Figure 2). Further, whereas Inoki et al also showed that Wnt signaling modulated p70-S6K phosphorylation (T389), we observed that while estrogen did increase phosphorylation at this residue, WNT4 knockdown had no effect on p70-S6K-T389 phosphorylation (Figure 3). Direct inhibition of p70-S6K using PF-4708671 also does not block T389 phosphorylation [35], but WNT4 may regulate p70-S6K phosphorylation at an alternative site, or regulate p70-S6K cellular localization [49].
Three independent large-scale genomic analyses of ILC cohorts have described activation of PI3K/Akt signaling in ILC as a major feature of this tumor type, including increased activity of this pathway versus matched IDC tumors [4][5][6]. Pathway activation is linked to reduced PTEN protein levels in ILC, increased phospho-Akt, and increased phospho-p70-S6K [4]. These observations parallel our data identifying that ER drives downstream mTOR signaling in ILC cells. In contrast, we observed that ILC cells have low levels of total and phospho-Akt versus MCF-7 cells. As described above, WNT4 in ILC may represent an alternative pathway for the activation of mTOR signaling, potentially in concert with estrogen regulation of WNT4 and low PTEN, as high WNT4-expressing ILC also had decreased phospho-Akt (Supplemental Figure 7E). A recent study by Teo et al linked Akt activation to loss of Ecadherin (the hallmark feature of ILC [4]), but these studies identified this link using ER-negative ILC models (murine p53/CDH1-null lines and IPH-926) [50]. These observations suggest that even within ILC, subsets with distinct modes of PI3K/Akt/mTOR signaling exist. Supporting this, among Luminal A ILC, TCGA RPPA identified differential levels and phosphorylation of PI3K/Akt/mTOR pathway proteins across mRNA subtypes (e.g. low p70S6K and Raptor in Reactive; high phospho-PRAS40 and phospho-mTOR in immune; Proliferative lacked distinct differences in this pathway). Another important context for PI3K-pathway signaling in ILC is likely the tumor microenvironment and metastasis.
Tasdemir et al recently reported that in ultra-low attachment conditions (ie. requiring anchorageindependence), ILC cell lines uniquely sustained PI3K-pathway activation versus IDC cell lines [51].
Further mechanistic studies are needed to link biomarkers with PI3K/Akt/mTOR signaling activity to identify personalized targets for therapy.
The putative link between ER:WNT signaling and MCL1/mitochondrial function suggests that WNT4 is critical for cellular metabolism in ILC. We observed that WNT4 knockdown leads to an accumulation of cellular ROS that parallels impaired ATP production (Figure 3), and both of these phenotypes precede the induction of cell death (~4d post-siWNT4 transfection, [15]). This suggests that WNT4-mediated control of metabolism and/or mitochondrial function are critical to ILC cell survival. Wnt signaling (both β-catenin-dependent and -independent signaling) has been previously linked to these processes [52][53][54], but roles for WNT4 in mitochondrial function have not previously described. However, one recent study demonstrated that Wnt4 overexpression could rescue a defect in mitochondrial function and dynamics caused by deletion of PTEN-inducible kinase 1 (PINK1) in Drosophila [55]. Though the mechanism of rescue is unclear, Wnt4 over-expression rescued flight defects in the PINK1 knockout flies, increasing ATP production and restoring mitochondria membrane potential in flight muscles.
Future studies will further examine whether WNT4 plays a similar role in cancer cells. Metabolic defects caused by WNT4 knockdown may also be caused by suppressing WNT4 signaling using anti-estrogens or mTOR pathway inhibitors. Targeting therapy-induced metabolic vulnerability may be a powerful combination treatment approach for WNT4-driven cancers.
The similar signaling pathways associated with WNT4 in ILC and OvCa may parallel the critical role of WNT4 in the tissues of origin, as well as related tumor biology. In addition to being required for mammary gland development, WNT4 is required for the development of the ovary and Mullerian tissues [45,56], female sex differentiation [56,57], and fertility [46,58]. Accordingly, WNT4 dysfunction is linked to a range of endocrine and gynecologic pathologies, including endometriosis [59], uterine fibroids [60], and ovarian cancer [61]. For ovarian cancer that originates from fallopian tube epithelium (FTE), transformed FTE cells must migrate to and invade the ovary to establish a tumor. WNT4 is required for cell migration and ovary invasion in murine PTEN-null models of FTE-derived ovarian cancer [62], suggesting that WNT4 is critical in early ovarian tumorigenesis. Perhaps consistent with this role of WNT4 in migration/invasion, both OvCa and ILC metastasize to the abdomen/peritoneal cavity, ILC being unique among breast cancers in this regard [63,64]. Defining WNT4 signaling in these tumors may provide new insight in to targeting metastatic OvCa and ILC.
WNT4 signaling is a key mediator of endocrine response and resistance in ILC, but in particular given the dysfunction of β-catenin-dependent signaling in ILC, WNT4 signaling is poorly defined. In this study, we used RPPA analyses to identify mTOR signaling and MCL1/mitochondrial function as two key downstream effectors of WNT4 in ILC. Parallel pathways associated with WNT4 were also identified in serous ovarian cancer, suggesting that WNT4 signaling is important in multiple tumor types. Future studies will determine how WNT4 is mechanistically linked to these targets, including identification of the WNT4 receptor in ILC. Furthering our understanding of WNT4 signaling will improve opportunities for precision treatment approaches by targeting WNT4 signaling for patients with ILC or OvCa, and support the development of strategies to overcome anti-estrogen resistance for patients with ILC.

T o t a l p S 4 7 3 p T 3 0 8 T o t a l p S 4 7 3 p T 3 0 8 T o t a l p S 4 7 3 p T 3 0 8
Akt: