Neuregulin-activated ERBB4 induces the SREBP-2 cholesterol biosynthetic pathway and increases low-density lipoprotein uptake

Abstract

We have found that ERBB4 activates sterol regulatory element binding protein-2 (SREBP-2) to enhance expression of genes essential for cholesterol metabolism including mevalonate pathway enzymes and the low-density lipoprotein receptor (LDLR). ERBB4 is unusual among receptor kinases in undergoing ligand-induced proteolytic cleavage to release a soluble intracellular domain that enters the nucleus and modifies transcription. Expression of the ERBB4 intracellular domain or activation of ERBB4 with the ligand Neuregulin 1 (NRG1) in mammary epithelial cells induced expression of SREBP target genes involved in cholesterol biosynthesis including HMGCR, HMGCS1, and LDLR beyond amounts induced through lipoprotein depletion. ERBB4 increased expression of cholesterogenic genes by enhancing abundance of the mature form of SREBP-2. NRG1 activated SREBP-2 through ERBB family kinases and PI3K, but independent from AKT or mTORC1 activity. NRG1 increased cholesterol metabolism by 1) increasing de novo biosynthesis through the mevalonate pathway, and 2) enhancing LDL binding and uptake through the LDLR. As all EGFR family receptors can appropriate ERBB4 signaling by cross-activating ERBB4, these data show that the ERBBs are linked to SREBP-regulated cholesterol metabolism, with potential impact on dyslipidemia and cancer.

Introduction

ERBB4 is essential for normal cardiac, neuronal, and mammary development and is activated by mutation in several cancers (1-4). As for other receptor kinases, ligand-induced ERBB4 Tyr phosphorylation sites recruit downstream signaling proteins. In addition, ERBB4 is unique in the ERBB family of receptor tyrosine kinases (RTKs) in that JM-a spliced isoforms undergo ligand-induced ecto-domain and intramembrane proteolysis to release a soluble intracellular fragment that enters the nucleus and modifies transcription (5-7). In this context, ERBB4 can act as both a nuclear chaperone (e.g., for STAT5) and a transcriptional co-activator (e.g., for estrogen receptor [ER], yes-associated protein [YAP], AP-2, KAP1, and TAB2/N-CoR) (6, 8-11).

ERBB4 is alternatively spliced at two sites, a juxtamembrane (JM) region and a cytoplasmic (CYT) region, to generate four receptor isoforms. The JM-a isoform contains a tumor necrosis factor-α (TNF-α) converting enzyme (TACE/ADAM17) cleavage site that is absent in JM-b that enables proteolytic cleavage of the receptor (12). Activation of TACE by ERBB4 ligands (heparin-binding EGF-like growth factor (HB-EGF), betacellulin, and neuregulins (NRG) 1-4), phorbol esters, or other agonists induces shedding of the extracellular domain (ECD) of the receptor, leaving a membrane-embedded 80 kDa isoform (m80) (13, 14). ECD shedding of JM-a (but not JM-b) isoforms enables intramembrane proteolysis at a γ-secretase cleavage site, which releases the soluble s80 form of the intracellular domain (ICD) (13). Tissue-specific alternative splicing to produce the ERBB4 JM-a isoform endows cells with the ability to signal through the ICD and couples the ERBBs to new avenues of signaling in addition to traditional RTK signaling at the membrane. Since the ERBBs promiscuously cross-activate when co-expressed, all ERBB receptors have the potential to signal through the ERBB4 JM-a ICD.

Alternative splicing of the intracellular domain produces isoforms that differ by only 16 amino acids included in CYT-1, but absent in CYT-2. This region contains an overlapping predicted PI(3’)K p85 binding site and PPxY motif (12). The small amino acid sequence difference between CYT-1 and CYT-2 potentiates divergent biological properties: in tissue culture and mouse models, CYT-1 promotes mammary differentiation and survival phenotypes, whereas CYT-2 induces proliferation (15-18). Expression of the CYT-2 isoform is enriched in the ER⁺HER2⁻ subtype of breast cancer (19).

In a transcriptional analysis of pathways activated by the full-length (FL) and ICD ERBB4 JM-a isoforms, we found that both FL and ICD ERBB4 CYT-1 and CYT-2 increase expression of several genes associated with cholesterol metabolism including 3-Hydroxy-3-Methylglutaryl (HMG)-Coenzyme A (CoA)-Reductase (HMGCR), HMG-CoA-Synthase 1 (HMGCS1), and Low-Density Lipoprotein Receptor (LDLR) (20).

Cholesterol is an essential component of cell membranes and acts as a signaling molecule, as a precursor to steroid hormones, and as a major nutritional source for neonates in milk (21, 22). Cholesterol homeostasis is coordinated by sterol regulatory element binding proteins (SREBPs), which promote expression of enzymes for de novo cholesterol production via the mevalonate pathway and expression of the LDLR to enhance extracellular lipoprotein uptake (23). When cholesterol abundance is high, SREBPs bind to SREBP cleavage-activating protein (SCAP) in the endoplasmic reticulum (ER), which forms a complex with insulin-induced gene (INSIG) proteins that sequester SREBPs in the ER and prevent cleavage. When intracellular cholesterol accumulation is low, SCAP transports SREBPs from the ER to the Golgi apparatus where they are subsequently cleaved to release amino-terminal fragments that bind to promoters and activate gene transcription.

The ERBBs have been linked to SREBP signaling. EGF activates fatty acid synthase through SREBP-1 in prostate and breast carcinoma cells and in glioblastoma (GBM) to enhance fatty acid synthesis (24-26). In GBM, activated EGFR induces cleavage of SREBP-1 following 6 hours of EGF treatment and increasing through 24 hours. EGFR activation of SREBP-1 occurs through PI3K and AKT signaling (26) and is independent of mTORC1 in GBM cell lines (25). Targeting fatty acid synthesis reduced tumor growth in a xenograft mouse model of EGFR-activated GBM (25). EGFRvIII activates a PI3K and SREBP-1-dependent tumor survival pathway by increasing LDLR in GBM through SREBP-1 but not SREBP-2 (27). Targeting LDLR with a liver X receptor agonist caused LDLR degradation and led to tumor cell death in an in vivo GBM model (27). EGFR has not been reported to activate SREBP-2, and there is little information on intersection of ERBB signaling with cholesterogenic pathways.

NRG1-activated ERBB4 regulation of cholesterogenic genes that we identified by transcriptional profiling (10, 18) could occur through an SREBP-dependent or SREBP-independent mechanism. Since SREBP-2 is the major regulator of genes promoting cholesterol synthesis, NRG1-induced increases in expression of these genes may work through enhanced SREBP-2 cleavage. Alternatively, ERBB4 might directly bind SREBP-2 and enhance nuclear translocation or transcriptional activation of SREBP-2 target genes. In order to address these possibilities, we have investigated the ability of the ERBB signaling network to activate the SREBP-2-regulated cholesterol biosynthetic pathway through ERBB4. The results indicate that NRG1 binding to ERBB4 activates SREBP-2 by inducing cleavage, and that this is accompanied by a corresponding increase in the expression of cholesterol metabolic enzymes, LDL uptake, and cholesterol biosynthesis.

Results

ERBB4 ICD enriches for SREBP-regulated cholesterol biosynthesis genes

Our earlier transcriptional analysis revealed that ERBB4 induces genes involved in cholesterol metabolism (18). Both NRG1-activated FL ERBB4 and the constitutively active ERBB4 ICD CYT-1 and CYT-2 isoforms enhance expression of genes associated with cholesterol metabolism including HMGCR, HMGCS1, DHCR7, DHCR24, LDLR, FDFT1, FDPS, IDI1, MVD, SQLE, LSS, NSDHL, SC5DL, INSIG1, ACLY, and ACSS2 (see fig. S1 for full list).

To further analyze SREBP-dependent cholesterol pathway activation we used gene set enrichment analysis (GSEA) to test for overlap of ERBB4 ICD-induced genes and genes associated with cholesterol metabolism. Engineered expression of ERBB4 ICD CYT-1 or CYT-2 in MCF10A cells, which do not express endogenous ERBB4, induced enrichment of transcripts associated with the REACTOME_CHOLESTEROL_BIOSYNTHESIS gene signature [Normalized Enrichment Score (NES) = 2.33, p = 1.46 × 10⁻⁴ (CYT-1) or NES = 2.42, p < 0.0001 (CYT-2)] and the HORTON_SREBF_TARGET gene signature [NES = 2.40, p < 0.0001 (CYT-1) or NES = 2.32, p < 0.0001 (CYT-2)] (Fig. 1). Genes showing core enrichment in each data set are listed in fig. S1, A and B.

Figure 1. ERBB4 ICD expression enriches for SREBP target genes and cholesterol biosynthesis.

(A) Gene Set Enrichment Analysis (GSEA) in MCF10A cells expressing ERBB4 ICD CYT-1. Upper plot shows enrichment for REACTOME_CHOLESTEROL_BIOSYNTHESIS gene set (Normalized Enrichment Score=2.33, p=1.47x10⁻⁴). Lower plot shows enrichment for HORTON_SREBF_TARGETS gene set (Normalized Enrichment Score=2.40, p<0.0001).

(B) Gene Set Enrichment Analysis (GSEA) in MCF10A cells expressing ERBB4 ICD CYT-2. Upper plot shows enrichment for REACTOME_CHOLESTEROL_BIOSYNTHESIS gene set (Normalized Enrichment Score=2.42, p<0.0001). Lower plot shows enrichment for HORTON_SREBF_TARGETS gene set (Normalized Enrichment Score=2.32, p<0.0001).

SREBPs (also denoted sterol regulatory element-binding transcription factors [SREBFs]) are master regulators of genes involved in cholesterol metabolism and fatty acid synthesis. The SREBP isoform SREBP-1c activates genes involved in fatty acid synthesis (including FASN [fatty acid synthase], ACLY [ATP citrate lyase], and ACC [acetyl-CoA carboxylase]) whereas SREBP-2 predominantly activates genes involved in cholesterol metabolism (including HMGCR, HMGCS1, and LDLR) (21). SREBP-1a is capable of activating both fatty acid synthesis and cholesterol metabolism (28). To test if ERBB4 broadly activates multiple SREBPs, we analyzed the ERBB4 ICD transcriptional data for increases in genes involved in fatty acid synthesis. We found no significant enrichment for the fatty acid synthesis gene sets KEGG_FATTY_ACID_METABOLISM or KEGG_BIOSYNTHESIS_OF_UNSATURATED_FA. Only a few fatty acid synthesis genes were significantly altered upon expression of ERBB4 CYT-1 or CYT-2 ICD (fig. S2, A and B). Unlike cholesterol metabolism genes, fatty acid synthesis genes were both up- and down-regulated (fig. S2, C and D).

NRG1 increases expression of SREBP-regulated genes

As the ERBB4 ICD used for these experiments is encoded by an artificial construct, we next determined if stimulation of FL ERBB4 with its ligand NRG1 also activates these cholesterol metabolism genes in the T47D mammary carcinoma cell line, which endogenously expresses ERBB4. FL ERBB4 might stimulate cholesterogenic genes either by activation of SREBPs or through SREBP-independent mechanisms, so we compared the ability of NRG1 to increase SREBP-target genes under conditions where SREBP is active or inhibited.

Use of lipoprotein-deficient serum (LPDS) reduces the bioavailability of exogenous cholesterol and therefore activates SREBPs. The increased mRNA expression of the SREBP-regulated genes HMGCR, HMGCS1, and LDLR observed in cells grown in LPDS was reduced when cells were supplemented with exogenous cholesterol in the form of low-density lipoproteins (LDLs), which inhibit the processing of SREBPs by inducing the formation of a complex containing SCAP, SREBP, and INSIG1 that sequesters SREBPs in the ER (fig. S3A; compare LPDS to LPDS replenished with LDL).

Interestingly, under LPDS conditions in which SREBPs are activated, NRG1 further increased HMGCR, HMGCS1, and LDLR mRNA an additional 2- to 3-fold (fig. S3A). This induction was more apparent with 2 hours of NRG1 treatment than 4 hours of treatment. Under conditions inhibiting SREBP activation (+LDL), LDL reduced but did not completely abolish NRG1-induction of HMGCR, HMGCS1, and LDLR mRNA (fig. S3A). NRG1-stimulated SREBP-target gene expression was slightly reduced following four hours of LDL incubation (2 hours pre-treatment and 2 hours with LDL and NRG1 co-incubation) compared with LPDS controls, and was reduced by ~50% following 6 hours of LDL treatment (2 hour LDL and 4 hour LDL plus NRG1 co-incubation) (fig. S3A).

As LDL and NRG1 regulate SREBP target genes with different kinetics, we conducted experiments during periods of maximal NRG1 gene expression (2 hours) and maximal LDL-induced inhibition of SREBP activity (6 hours). Two hours of NRG1 treatment increased HMGCR 2-fold, HMGCS1 2-fold, and LDLR 3-fold (Fig. 2A). NRG1 weakly increased HMGCR and HMCS1 expression, 1.2- and 1.3-fold, respectively, in the presence of LDL. However, in the presence of exogenous LDL, NRG1 still induced a substantial 3-fold increase in LDLR despite lower absolute amounts of LDLR expression (Fig. 2A).

Figure 2. NRG1 activates SREBP-2 cleavage and enhances expression of cholesterogenic genes.

(A) RT-PCR of HMGCR, HMGCS1, or LDLR in T47D cells incubated in LPDS for 24 hours followed by concomitant NRG1 (50 ng/ml) treatment for the final 2 hours. LDL treated cells (+LDL) were pre-treated with LDL (50 μg/ml) for 4 hours prior to the addition of NRG1 to the media. Data are means (SD) from 3 experiments.

(B) Immunoblot of SREBP-2 cleavage in T47D cells treated as in (A). Blots are representative of 3 experiments. “p”, uncleaved SREBP-2 precursor; “m” cleaved SREBP-2 mature form.

(C) Immunoblot of HMGCR and LDLR abundance in T47D cells incubated in LPDS for 24 hours followed by concomitant NRG1 (50 ng/ml) treatment for the final 0, 2, 4, or 6 hours. Negative control cells were incubated in parallel in the presence of FBS (+serum). Positive control cells were incubated in LPDS along with 1 μM simvastatin (+simv) for 24 hours. Blots are representative of 3 experiments.

(D) Immunoblot of SREBP-2 cleavage and HMGCR abundance in T47D cells incubated in LPDS for 48 hours with concomitant NRG1 (50 ng/ml) treatment for the final 0.5, 1, 2, 4, 6, 8, or 24 hours. Positive control cells were incubated in LPDS along with 1 μM simvastatin (+simv) for 24 hours. “long”, longer exposure; “p”, uncleaved SREBP-2 precursor; “m” cleaved SREBP-2 mature form. Arrowhead marks ICD.

(E) Immunoblot of HMGCR abundance and SREBP-2 cleavage in T47D cells incubated in LPDS along with PF-429242 for 24 hours followed by concomitant NRG1 (50 ng/ml) treatment for the final 6 hours. Blots are representative of 3 experiments. p = precursor form of SREBP-2 (full-length, 120kDa), m = mature form of SREPB-2 (cleaved, 60kDa). GAPDH (H), (L), or (S) = GAPDH for HMGCR, LDLR, or SREBP-2 samples. “p”, uncleaved SREBP-2 precursor; “m” cleaved SREBP-2 mature form. Blots are representative of 3 experiments. *p<0.05; **p<0.01, ***p<0.001, ****p<0.0001.

NRG1 stimulates SREBP-2 cleavage

Since NRG1 induces multiple SREBP-2 target genes, we determined whether NRG1 activates SREBP-2 cleavage. Indeed, under lipoprotein depletion, incubation with NRG1 for 2 hours induced a 2-fold increase in the amount of mature, cleaved SREBP-2 (Fig. 2B “m”). However, 4 hours of NRG1 treatment did not strongly affect SREBP-2 cleavage (fig. S3B). Addition of LDL reduced the absolute amount of cleaved SREBP-2 in the absence of NRG1. NRG1 still induced a 2-fold increase in cleaved SREBP-2 in the presence of LDL with 2 hours in NRG1, but not 4 hours (Fig. 2B and fig. S3B). Taken together, these data indicate that NRG1 activates SREBP-2 to induce SREBP-2-regulated genes and that NRG1 may partially override sterol-mediated inhibition of SREBP-2 cleavage in the presence of cholesterol in the media in the form of LDL.

NRG1 induces acute activation of SREBP-2 followed by increases in protein expression of cholesterogenic genes

To determine if NRG1-dependent increases in expression of SREBP-2 target genes are recapitulated at the protein level, we analyzed HMGCR and LDL receptor protein expression. T47D cells were incubated with NRG1 for up to 6 hours following 24 hours of lipoprotein depletion. Incubation in serum (negative control) strongly inhibited HMGCR protein expression. NRG1 induced a 2- to 3-fold increase in both HMGCR and LDLR protein abundance as early as 4 hours following NRG1 treatment (Fig. 2C). The HMGCR inhibitor simvastatin was used as a positive control for HMGCR expression since inhibition of cholesterol biosynthesis with simvastatin activates the SREBP pathway, thereby increasing HMGCR protein abundance. The increases in expression elicited by NRG1 augmented that induced by lipoprotein depletion, suggesting that ERBB4 can activate cholesterol biosynthesis genes beyond the amounts induced by lipoprotein depletion.

We next compared timing of SREBP-2 cleavage and changes in HMGCR expression in NRG1-stimulated T47D cells (Fig. 2D). NRG1 induced SREBP-2 cleavage as early as 30 min following NRG1 addition, and the mature “m” form of SREBP-2 continued to increase in abundance through 2 hours of stimulation. Abundance of mature SREBP-2 returned to baseline amounts induced by lipoprotein depletion after 4 hours of NRG1 treatment. HMGCR expression increased through 6 hours of NRG1 treatment and returned to baseline within 24 hours. ERBB4 cleavage (Fig. 2D, arrowhead) was detected 30 min following NRG1 treatment, was sustained through 1 hour and returned to baseline at 2 hours (Fig. 2D). Presence of the ERBB4 ICD temporally preceded the increase in the mature form of SREBP-2, which was followed by increased abundance of HMGCR. Therefore, transient activation of ERBB4 was sufficient to activate SREBP-2 and induce persistent expression of cholesterogenic genes.

Inhibition of SREBP cleavage reduces NRG1-induced expression of HMGCR

NRG1 enhances SREBP-2 cleavage. To determine if SREBP mediates the NRG1 induction of cholesterol metabolism genes, we used a serine protease inhibitor that is selective for SREBP site 1 protease (S1P), PF-429242, to block SREBP cleavage. Co-treatment of T47D cells with LPDS and PF-429242 for 24 hours reduced both precursor and mature SREBP-2 protein abundance (Fig. 2E). Since SREBP-2 activates its own promoter, leading to feed-forward induction, inhibition of SREBP-2 cleavage was expected to reduce total SREBP-2 protein (29). PF-429242 reduced the abundance of HMGCR protein induced by 6 hours of NRG1 stimulation in a dose-dependent fashion (Fig. 2E). NRG1 increased HMGCR relative to cells cultured in PF-429242 alone, so NRG1 can still induce SREBP-2 cleavage despite reduced expression of SREBP-2. NRG1 did not affect SREBP-2 cleavage at the 6-hour time point, which is consistent with our previous data.

We confirmed that PF-429242 treatment blocks SREBP-2 cleavage and does not inhibit ERBB4 phosphorylation or cleavage following 2 hours of NRG1 stimulation (fig. S4, A and B). In fact, PF-429242 at the highest dose enhanced accumulation of the ERBB4 cleavage product (fig. S4B), which was nearly undetectable in the absence of inhibitor after 2 hours NRG1 treatment (Fig. 2D). Overall, these data confirm that NRG1 activation of SREBP target genes is at least partially dependent on SREBPs. We cannot rule out the possibility that SREBP-1 is also involved since S1P also cleaves the SREBP-1 isoform.

Collectively, these data support the conclusion that NRG1 activates SREBP-2 target genes involved in the biosynthesis and uptake of cholesterol, and does so beyond amounts induced by lipoprotein depletion.

SREBPs mediate NRG1-induced activation of cholesterogenic genes through ERBB kinases and PI3K, independent of AKT and mTORC1

The mechanism by which SREBP activation is regulated by growth factor receptors such as the insulin receptor and the EGFR is not completely understood. In prostate carcinoma and GBM cell lines and tumors, EGFR activates SREBP-1 to increase fatty acid synthesis through a PI3K and AKT-dependent mechanism that in GBM is mTORC1-independent (25, 26). In keratinocytes, EGF and TGFα induce HMGCR through MAP kinase pathways but the involvement of SREBPs in this process was not determined (30).

To test if ERBB kinase activity is required for NRG1 activation of SREBP-2 cleavage we used a pan-ERBB kinase inhibitor, lapatinib, and an EGFR kinase inhibitor, erlotinib. In control cells treated with DMSO, NRG1 induced a 2-fold increase in SREBP-2 cleavage with concomitant upregulation of phosphorylation of ERBB4, EGFR, AKT, and p70 S6 kinase (Fig. 3A). Inhibition of ERBB kinases with lapatinib reduced phosphorylation of ERBB4 and EGFR and diminished NRG1-induced cleavage of SREBP-2. Lapatinib also lowered baseline and NRG1-induced phosphorylation of AKT and p70 S6 kinase. Treatment with the EGFR inhibitor erlotinib did not block NRG1-induced phosphorylation of EGFR, presumably through cross-phosphorylation of EGFR by ERBB4. However, erlotinib did reduce baseline EGFR phosphorylation. EGFR kinase inhibition did not greatly affect NRG1 activation of SREBP-2 cleavage. Erlotinib lowered baseline and NRG1-induced phosphorylation of p70 S6 kinase but had no effect on AKT or ERBB4 phosphorylation.

Figure 3. NRG1 induces SREBP-2 cleavage through ERBB kinases, independent of AKT and mTORC1.

(A) Immunoblot of SREBP-2 cleavage in T47D cells incubated in LPDS for 24 hours followed by concomitant NRG1 (50 ng/ml) treatment for the final 2 hours. Cells were pre-treated for 30 min prior to the addition of NRG1 with DMSO control or drugs targeting the following kinases: pan-ERBB (lapatinib, 1 μM), EGFR (erlotinib, 1 μM), PI3K (GDC-0941, 1 μM), or AKT (MK-2206, 1 μM). Drug efficacy was confirmed by blotting for inhibition of phosphorylation of ERBB4 (P-Tyr¹⁰⁵⁶), EGFR (P-Tyr¹⁰⁶⁸), AKT (P-Ser⁴⁷³), or p70 S6 kinase (P-Thr³⁸⁹). Blots are representative of 2 experiments.

(B) Immunoblot of SREBP-2 cleavage and phosphorylation of p70 S6 kinase (P-S6K) and ERBB4 in T47D cells incubated in LPDS for 24 hours followed by concomitant NRG1 (50 ng/ml) treatment for the final 2 hours. Cells were pre-treated for 30 min prior to the addition of NRG1 with DMSO control or rapamycin (1 or 10 nM). “p”, precursor form of SREBP-2 (full-length, 120kDa), “m”, mature form of SREPB-2 (cleaved, 60 kDa).

(C) Immunoblot of HMGCR protein abundance and phosphorylation of ERBB4 at Tyr1056 in T47D cells incubated in LPDS for 24 hours followed by NRG1 (50 ng/ml) treatment for the final 6 hours. Cells were pre-treated for 30 min prior to the addition of NRG1 with DMSO control or wortmannin (1 μM), GDC-0941 (1 μM), BYL-719 (1 μM), dual PI3K/mTOR inhibitor (1 μM), rapamycin (100 nM), AZD-8055 (1 μM), MK-2206 (1 μM), or SB-203580 (1 μM). Blots are representative of 3 experiments.

(D) Immunoblot of SREBP-2 cleavage and phosphorylation of ERBB4 at Tyr1056 in T47D cells incubated in LPDS for 24 hours followed by NRG1 (50 ng/ml) treatment for the final 2 hours. Cells were treated as in (C). Blots are representative of 3 experiments.

(E) Immunoblot of SREBP-2 cleavage, phosphorylation of ERBB4 at Tyr1056, and phosphorylation of p70 S6 kinase at Thr389 in T47D cells incubated in LPDS for 24 hours followed by NRG1 (50 ng/ml) treatment for the final 2 hours. Cells were treated as in (C) in the presence or absence of NRG1. Blots are representative of 3 experiments.

We next examined whether NRG1 activates SREBP-2 through the PI3K pathway since inhibitors of PI3K, AKT, and mTORC1 can reduce SREBP cleavage in various cellular contexts. PI3K inhibition with GDC-0941 reduced baseline and NRG1-induced phosphorylation of AKT and p70 S6 kinase (Fig. 3A). GDC-0941 raised baseline SREBP-2 cleavage, and NRG1 did not elicit any further increase in mature SREBP-2. AKT inhibition with MK-2206 strongly reduced baseline and NRG1-induced AKT phosphorylation, without affecting NRG1 activation of SREBP-2 cleavage (Fig. 3A). mTORC1 inhibition with rapamycin (Fig. 3B) caused a dose-dependent reduction in p70 S6 kinase phosphorylation, but did not alter the abundance of mature SREBP-2 induced by NRG1.

Having ruled out contributions from AKT and mTORC1, we tested a more extensive panel of PI3K and mTOR inhibitors for effects on ERBB4 phosphorylation and NRG-dependent HMGCR production (Fig. 3C) and for impact on SREBP-2 cleavage (Fig. 3D). Four PI3K inhibitors: wortmannin (PIKK superfamily), GDC-0941, BYL-719 (PI3Kα), and a dual PI3K and mTOR inhibitor reduced HMGCR in the presence of NRG1. Of these, wortmannin, BYL-719, and the dual PI3K and mTOR inhibitor also blocked NRG1 activation of SREBP-2 cleavage. Rapamycin (mTORC1), AZD-8055 (mTORC1 and mTORC2), MK-2206 (AKT inhibitor), and SB-203580 (p38 MAPK) all partially reduced HMGCR, and only AZD-8055 (mTORC1 and mTORC2) also consistently blocked NRG1 activation of SREBP-2 cleavage. The PI3K inhibitors, wortmannin and BYL-719, and the dual mTORC1 and mTORC2 inhibitor, AZD-8055, were the most effective agents at blocking both NRG1 activation of SREBP-2 and upregulation of HMGCR. Each of these drugs reduced NRG1-dependent phosphorylation of p70 S6 kinase without affecting ERBB4 phosphorylation (Fig. 3E). Overall, these results indicate that NRG1 activates SREBP-2 through a mechanism requiring ERBB4 kinase activity and PI3K, and independent of AKT or mTORC1.

ERBB4 cleavage is not required for NRG1 increases in HMGCR

ERBB4 ICD produced by cleavage of JM-a isoforms has unique intracellular signaling properties. Since both the ERBB4 ICD and the FL receptor may be capable of activating SREBP signaling, we determined whether cleavage and release of the ERBB4 ICD is required for SREBP-2 signaling. We produced MCF10A pINDUCER20 cell lines expressing doxycycline (DOX)-inducible ERBB4 JM-a CYT-2 or the TACE-insensitive spliced isoform ERBB4 JM-b CYT-2. As expected, NRG1 or PMA treatment of JM-a, but not JM-b ERBB4 induced production of the approximately 80kD JM-a cleavage product (fig. S5; compare FL and smaller forms at 5 ng/ml DOX induction). In engineered MCF10A cells without DOX induction of pINDUCER20-ERBB4, NRG1 had no effect (Fig. 4) or slightly enhanced production of HMGCR (Fig. 4), either through leaky production of ERBB4 by the uninduced pINDUCER-regulated cDNAs, or through activity of endogenous ERBB3. DOX induction of ERBB4 enhanced NRG1-dependent accumulation of HMGCR, and this occurred comparably in cells expressing either JM-a or JM-b (Fig. 4). Despite inducing ERBB4 cleavage in JM-a ERBB4 cells, PMA did not consistently increase HMGCR above baseline (Fig. 4). Hence, ERBB4 cleavage mediated exclusively through the JM-a isoform is neither necessary nor sufficient (in the absence of NRG1) for upregulation of HMGCR.

Figure 4. NRG1-induced HMGCR is similar for ERBB4 juxtamembrane domain isoforms JM-a and JM-b.

Immunoblot of ERBB4 cleavage and HMGCR levels in MCF10A cells expressing pINDUCER20 encoding doxycycline (DOX)-inducible JM-a or JM-b, CYT-2 isoforms of ERBB4. Cells were incubated in Opti-MEM reduced serum media in the presence or absence of DOX (5 ng/ml) for 24 hours followed by concomitant DMSO, NRG1 (50 ng/ml) or PMA (100 ng/ml) treatment for the final 3 hours. Arrowheads mark ICD, FL marks the position of full-length ERBB4. Blots are representative of 3 experiments.

NRG1 activates SREBP-2 through ERBB4

Although EGF activates SREBP-1 cleavage in prostate carcinoma cells and GBM, the impact of EGF on SREBP-2 has not been reported. We directly compared the ability of EGF and NRG1 to induce SREBP-2 cleavage and upregulation of HMGCR in T47D cells. NRG1 increased mature SREBP-2 after 2 hours of treatment but not 24 hours (Fig. 5A). EGF did not alter amount of mature SREBP-2 compared with baseline accumulation induced in LPDS at 2 hours or 24 hours (Fig. 5A). NRG1 increased HMGCR protein expression 2-fold and EGF increased expression of HMGCR 1.5-fold. NRG1 consistently induced greater HMGCR protein abundance than EGF, and HMGCR abundance correlated with ERBB4 phosphorylation (Fig. 5B).

Figure 5. NRG1 activates SREBP-2 through ERBB4.

(A) Immunoblot of SREBP-2 cleavage and ERBB4 (P-Tyr¹⁰⁵⁶) and EGFR (P-Tyr¹⁰⁶⁸) phosphorylation in T47D cells incubated in LPDS for 24 hours followed by concomitant NRG1 (N, 50 ng/ml) or EGF (E, 50 ng/ml) treatment for the final 2 or 24 hours. Positive control cells were incubated in LPDS along with 1 μM simvastatin (+simv) for 24 hours. Blots are representative of 3 experiments. “p”, uncleaved SREBP-2 precursor; “m” cleaved SREBP-2 mature form.

(B) Immunoblot of HMGCR and ERBB4 (P-Tyr¹⁰⁵⁶) and EGFR (P-Tyr¹⁰⁶⁸) phosphorylation in T47D cells incubated in LPDS for 24 hours followed by concomitant NRG1 (N, 50 ng/ml) or EGF (E, 50 ng/ml) treatment for the final 6 hours. Negative control cells were incubated in parallel in the presence of FBS (+serum). Positive control cells were incubated in LPDS along with 1 μM simvastatin (+simv) for 24 hours. Blots are representative of 3 experiments.

(C) Immunoblot of HMGCR protein abundance and ERBB4 (P-Tyr¹⁰⁵⁶) and EGFR (P-Tyr¹⁰⁶⁸) phosphorylation in T47D cells incubated in LPDS for 24 hours followed by concomitant NRG1 (N, 50 ng/ml) or EGF (E, 50 ng/ml) treatment for the final 6 hours or 10 min. Blots are representative of 2 experiments.

(D) Immunoblot of HMGCR protein abundance and ERBB4 (P-Tyr¹⁰⁵⁶) and EGFR (P-Tyr¹⁰⁶⁸) phosphorylation in MCF10A cells expressing pINDUCER20 encoding doxycycline (DOX)-inducible ERBB4 JM-a, CYT-2 incubated in Opti-MEM reduced serum media in the presence or absence of DOX (50 ng/ml) for 24 hours followed by concomitant NRG1 (N, 50 ng/ml) or EGF (E, 50 ng/ml) treatment for the final 6 hours. Blots are representative of 3 experiments.

After 2 or 4 hours of treatment, NRG1-induced ERBB4 phosphorylation was strong, but EGF-induced EGFR phosphorylation was considerably weaker (Fig. 5A and 5B). As T47D cells express roughly similar amounts of EGFR and ERBB4, this may be explained by suboptimal activity of EGF, or by differences in timing of EGFR and ERBB4 downregulation and/or dephosphorylation. Hence, we compared NRG1- or EGF-induced phosphorylation of ERBB4 and EGFR following 10 min of stimulation and 6 hours of stimulation (Fig. 5C). In fact, EGFR and ERBB4 were similarly phosphorylated with 10 min of incubation with their respective ligands, but EGFR phosphorylation at Tyr¹⁰⁶⁸ decayed much more rapidly than did ERBB4 phosphorylation. Hence, despite comparable initial activation of ERBB4 and EGFR by their respective ligands, NRG1 more robustly activates SREBP-2 than does EGF.

MCF10A cells express endogenous EGFR and ERBB3, but not endogenous ERBB4. As another approach to determining the ERBB4 dependency of the NRG1 response, we evaluated NRG1 activation of SREBP-2 in MCF10A cells engineered to express ERBB4 under control of DOX-inducible promoter. MCF10A pINDUCER20 ERBB4 JM-a CYT-2 cells were cultured in reduced serum media in the presence or absence of DOX for 24 hours and co-cultured with NRG1 or EGF for the final 6 hours. DOX treatment induced ERBB4 CYT-2 protein expression, which was absent in untreated cells (Fig. 5D, fig. S6B, S6C). NRG1 induced ERBB4 phosphorylation and increased HMGCR abundance two-fold when ERBB4 was expressed but had no effect on HMGCR when ERBB4 was not expressed. EGF induced EGFR phosphorylation and increased HMGCR abundance 2-fold in the presence or absence of ERBB4, indicating that EGF increases HMGCR expression through EGFR in MCF10A cells.

To quantitatively compare the ability of NRG1 and EGF to activate SREBP target genes we measured the ability of each ligand to induce LDLR, HMGCR, and HMGCS1 mRNA expression in MCF10A pINDUCER20 ERBB4 JM-a CYT-2 cells in the absence or presence of DOX (fig. S6A, S6B). LDLR was induced comparably with EGF treatment regardless of whether ERBB4 expression was induced with DOX. Consistent with the HMGCR protein phenotype in Figure 4D, NRG1 induced LDLR expression only when ERBB4 was present (+DOX). There were no changes in HMGCR or HMGCS1 mRNA in the presence of ERBB4. Collectively, these data demonstrate that NRG1 activates SREBP-2 through ERBB4.

NRG1 increases LDL binding to LDLR and LDL uptake

To determine if NRG1-induced changes in cholesterol genes and SREBP-2 cleavage functionally affects cholesterol metabolism, we first examined the impact of NRG1 on binding of LDL to LDLR and on LDL uptake (Fig. 6, A and B). Incubation of T47D cells in LPDS for 24 hours increased LDL binding by 73% (±4.7%, p<0.0001) and uptake by 59% (±6.1%, p<0.001) compared with cells grown in the presence of serum (FBS), demonstrating effective lipoprotein depletion (Fig. 6A and 6B). Eight hours incubation with NRG1 increased LDL binding by 43% (±6.4%, p<0.01) and uptake by 45% (±4.7%, p<0.001) compared with LPDS treated cells. In the presence of serum, NRG1 had no effect on LDL binding but there was a trend toward enhanced LDL uptake (+34% ±16.0%, p=0.104). These data are consistent with NRG1-induced SREBP-2 cleavage and activation of SREBP-2 regulated cholesterol genes and indicate that NRG1 has a functional impact on cholesterol metabolism.

Figure 6. NRG1 increases LDL binding and uptake, and enhances biosynthesis of cholesterol from [2-¹³C]-acetate precursor.

(A) Flow cytometry analysis of diI-LDL binding in T47D cells incubated in LPDS or in the presence of FBS for 24 hours. During the final 6 hours, cells were treated with or without NRG1 (50 ng/mL). Data are representative of 3 experiments.

(B) Flow cytometry analysis of diI-LDL uptake as in (A). Data points from each trial are shown. Filled circles, squares, and triangles represent trials 1, 2, and 3, respectively. n.s., not significant, **p<0.01, ***p<0.001, ****p<0.0001.

(C), (D) Cholesterol-trimethylsilane isotopomer spectra from T47D cells incubated without (“•“) or with NRG1 (“NRG”) for 6 hours (Fig. 6C) or 24 hours (Fig. 6D), or in the presence of simvastatin (“Simv”) to inhibit cholesterol synthesis. NRG promoted cholesterol synthesis from [2-¹³C] acetate as evidenced by the higher fractional abundance of [¹³C]-cholesterol (M+3->M+7), whereas, simvastatin inhibited cholesterol synthesis from [2-¹³C]acetate resulting in less ¹³C incorporation. Representative data from one of three experiments described in supplementary Table S1 are shown.

NRG1 enhances de novo cholesterol biosynthesis

We next evaluated the impact of NRG1 on de novo cholesterol biosynthesis. T47D cells in LPDS were cultured with NRG1 for 6 hours or 24 hours in the presence of heavy-labeled [2-¹³C]-acetate or non-labeled acetate (Fig. 6C, 6D). Relative [¹³C] incorporation into cholesterol isotopomers was evaluated using gas chromatograph-mass spectrometry (GC-MS). 24 hour incubation with NRG1 significantly increased incorporation of [¹³C] into cholesterol at peaks m+3 through m+7 (Fig. 6D). A similar trend was observed with 6 hours of NRG1 treatment, but was not statistically significant. In contrast, simvastatin treatment significantly reduced [¹³C]-acetate conversion into cholesterol, in accordance with inhibition of HMGCR by this agent (Fig. 6C, 6D).

We used Isotopomer Spectral analysis (ISA) to estimate the parameters of cholesterol synthesis (31, 32) from the isotopomer labeling patterns (Fig. 6D). For 24 hours NRG1 treatment, ISA revealed a 20% increase in the fraction of newly synthesized cholesterol from 27.8% for control cells to 33.4% for NRG1-treated cells (p= 0.0002), with no significant increase for 6 hours treatment (Supplementary Table S1). According to ISA, acetate contributed 47% to 49% of the lipogenic acetyl CoA pool and this value was not affected by NRG1 (Supplementary Table S1). Thus, the increased labeling of NRG1 treated cells is the result of increased de novo synthesis. Overall, these findings demonstrate that NRG1 stimulates the mevalonate pathway leading to increased cholesterol biosynthesis.

Discussion

The ERBBs are major regulators of cell growth, a process that requires sufficient nutrients and metabolites. However, little is known about how the ERBBs alter cell metabolism. We found that NRG1-activated ERBB4 enhances cholesterogenic gene expression through SREBP-2, a master regulator of cholesterol metabolism. Activated ERBB4 increased the abundance of both mevalonate pathway enzymes and the LDLR- the two primary routes for generating intracellular cholesterol. NRG1 significantly increased the binding and uptake of LDL as well as de novo cholesterol biosynthesis, demonstrating the functional impact of ERBB4 to SREBP-2 signaling.

NRG1 enhanced SREBP-2 cleavage and SREBP-target gene expression above amounts induced by lipoprotein depletion, indicating that NRG1 is capable of activating SREBP signaling beyond that induced physiologically by sterol depletion. Exogenous LDL, which inhibits SREBP cleavage through SCAP, reduced NRG1-induced expression of cholesterogenic genes, suggesting that active SREBP is necessary for the NRG1 effect. Furthermore, NRG1-induced LDL uptake was impaired in the presence of FBS, which contains lipoproteins. A chemical inhibitor of SREBP cleavage (PF-429242) partially blocked the ability of NRG1 to induce HMGCR expression. Collectively, these data demonstrate that ERBB4 strongly activates the expression of genes regulating cholesterol metabolism through SREBP-2.

ERBB4 might activate SREBP-2 at multiple levels. We find that both the ICD and FL ERBB4 are capable of activating SREBP-2 signaling. However, release of the ICD does not appear to be required. FL ERBB4 activates SREBP-2 through PI3K, but independent from AKT and mTORC1. The soluble ERBB4 ICD might also activate PI3K or, alternatively, it could interact with SREBP in the ER, Golgi, or nucleus.

NRG1 activates SREBP-2 through the ERBB4 kinase, since treatment with the pan-ERBB kinase inhibitor lapatinib reduced NRG1 induction of SREBP-2 cleavage and HMGCR protein expression. NRG1 binds ERBB4, and also ERBB3, which has no kinase activity. Therefore, ERBB4 kinase activity appears to be necessary for the effect of NRG1 on SREBP-2. Since ERBB4 can cross-activate the other ERBB receptors, it is possible that lapatinib blocks SREBP cleavage by inhibiting EGFR kinase activity. However, erlotinib treatment did not prevent SREBP-2 cleavage.

Active PI3K appears to be required for NRG1 activation of SREBP-2. Multiple PI3K inhibitors blocked both NRG-1 induced SREBP-2 cleavage and increases in HMGCR protein abundance. NRG1 activation of SREBP-2 does not seem to require AKT or mTORC1. Despite inhibition of AKT activity by MK-2206 and mTORC1 activity by rapamycin, NRG1 still increased SREBP-2 cleavage. This sets ERBB4 to SREBP-2 signaling apart from mechanisms of SREBP activation that are proposed to require AKT, including EGFR and insulin receptor signaling, and suggests that ERBB4 activates SREBP-2 through AKT-independent PI3K signaling arms. Since a dual mTORC1 and mTORC2 inhibitor, but not rapamycin (targeting mTORC1), blocked SREBP-2 signaling, it is possible that mTORC2 facilitates ERBB4 activation of SREBP-2.

Our data show that the ERBB4 ICD is capable of activating SREBP-2, independent of the cytoplasmic isoform (CYT-1 or CYT-2) that is expressed. So, it is likely that ERBB4 activates SREBP-2 through a domain outside of the alternatively spliced 16 amino acid region present in CYT-1. ERBB4-mediated activation of SREBP signaling is likely dependent upon paracrine or autocrine ERBB-ligand expression and the unique milieu of ERBB receptors present to form homo- or heterodimers.

The similar abilities of JM-a and JM-b isoforms to activate SREBP-2 implies that formation of the ERBB4 ICD is not required. This interpretation is not definitive without further experimentation, as γ-secretase cleavage of ERBB4 in lung alveolar cells does not require TACE activity (33). Nonetheless, it is possible that expression of soluble ICD in the original transcription profiling experiments (17) activated PI3K signaling, and subsequent SREBP signaling. We are also investigating alternative possibilities involving ERBB4 directly enhancing SREBP-2 activation,including stabilization of the SCAP complex with SREBP, completion of the SCAP complex away from INSIG1 allowing for ER to Golgi transport and SREBP cleavage, and/or promotionof nuclear import of mature SREBP-2.

ERBB4 activation of SREBP-2 appears to occur though a mechanism distinct from EGFR regulation of SREBP-1 (Fig. 7). NRG1, but not EGF, activates SREBP-2 cleavage in T47D cells leading to concomitant increases in HMGCR. ERBB4 is required for NRG1-induced increases in HMGCR. Unlike EGFR activation of SREBP-1 in GBM cells, ERBB4 induction of SREBP-2 cleavage does not depend on AKT or EGFR kinase activity. NRG1 activates SREBP-2 cleavage acutely following 0.5 to 2 hours of NRG1 treatment in breast cancer cell lines whereas EGF induces SREBP-1 cleavage from 6 to 24 hours in GBM cell lines (25). These data suggest that NRG1 and ERBB4 activate SREBP-2 distinctly from EGFR but do not rule out the possibility that EGFR might activate SREBP-1a to regulate cholesterol biosynthesis genes in parallel with or in place of ERBB4 to SREBP-2 in some tissues.

Figure 7. Model of ERBB4 activation of SREBP-2/cholesterol signaling.

NRG1 activates ERBB4 and induces proteolytic cleavage and release of membrane- anchored and soluble forms of the ICD. FL and ICD ERBB4 contribute to SREBP-2 activation through PI3K signaling pathways and, hypothetically, through direct interactions. ERBB4 enhances SREBP-2 cleavage through ERBB kinase activity but independent of AKT and mTORC1 resulting in increased expression of cholesterogenic genes (including HMGCR, HMGCS1, and LDLR) and increased LDL uptake and cholesterol synthesis. EGFR regulates fatty acid synthesis and increases LDLR expression through SREBP-1 (24-26), so activated EGFR might also regulate cholesterogenesis in parallel with ERBB4.

Several ERBB4-regulated biological processes might rely on SREBP-2 regulation of cholesterol metabolism. In the mammary gland, EGFR, ERBB2, and ERBB3 are important for ductal tissue expansion at puberty and in pregnancy, whereas ERBB4 plays an essential role in tissue differentiation and the expression of milk proteins during lactation through ICD activation of STAT5 (2, 3, 16). The lactating mouse mammary gland secretes a milk lipid equivalent to its entire body weight over a 20-day lactation cycle (35). ERBB4 might help coordinate production of milk proteins with the anabolic shift that occurs at onset of lactation by increasing expression of mevalonate pathway enzymes and LDLR.

In the brain, NRG1 stimulates myelination through ERBB4, and cholesterol is a major component of myelin. SREBP-2 might mediate the effects of NRG1 and ERBB4 in the brain including: neuronal cell migration, NMDA receptor signaling, timing of astrogenesis, and differentiation of radial glia. NRG1 and ERBB4 are both schizophrenia-linked genes. ERBB4 to SREBP-2 signaling could be altered in schizophrenia where disruption of myelination is believed to contribute to pathogenesis (36).

ERBB4 is activated by mutation in several cancers including melanoma (14%, COSMIC) and lung cancer (5%, COSMIC). CD74-NRG1 fusions are amplified in a subset of human lung cancers, and ligand-activated ERBB4 is enriched in chemotherapy-resistant mouse models of lung cancer (37, 38). ERBB4 to SREBP-2 signaling may be coopted in cancerous cells.

Increased mevalonate pathway activity and cholesterol abundance are seen in several cancers, including prostate and breast cancer. In prostate cancer, cholesterol is elevated due to enhanced activity of the mevalonate pathway and activation of SREBPs (39, 40). High mRNA expression of HMGCR and other genes of the mevalonate pathway correlates with poor prognosis in primary breast cancer (41). Furthermore, upregulation of the mevalonate pathway by mutant p53 is necessary and sufficient for induction of spheroids by nonmalignant mammary epithelial cells in three-dimensional cell culture (42). Introduction of activated PI3K or KRAS into mammary cells upregulates lipid biosynthesis through mTORC1 and mTORC2, leading to activation of SREBP-1 and SREBP-2, and knockdown experiments indicate the importance of both SREBP-1 and SREBP-2 on lipogenesis and proliferation (43). Ectopic expression of HMGCR, the rate-limiting enzyme in the mevalonate pathway, promotes cell growth and cooperates with RAS to drive the transformation of primary mouse embryonic fibroblasts (41). Therefore, ERBB4 activation of SREBP-2 has the potential to contribute to cancer cell growth and survival.

ERBBs were linked to metabolic regulation as early as 1965, when Stanley Cohen found that injection of EGF into newborn rodents induces fatty liver (including some elevation in cholesteryl esters) (44), and now reinforced with the finding that the Dsk5 gain-of-function mutation in the EGFR induces fatty liver with elevated HMGCR and FAS in mice (45).

The union of the ERBB receptors, major regulators of cell growth and development, and SREBPs, master regulators of cholesterol and fatty acid homeostasis, sheds light on potential biological roles for ERBB to SREBP signaling in development and disease. ERBB4 can couple all expressed ERBB receptors to cholesterol metabolism through receptor cross-activation and is unusual in its ability to undergo cleavage and act as a transcriptional co-activator. The discovery that ERBB4 activates cholesterol metabolism through SREBP-2 highlights an underappreciated connection between the ERBBs and cell metabolism. The network of signaling interactions through which the four ERBBs modulate cholesterol and fat biosynthesis have important implications for cellular signaling and metabolism. The importance of ERBBs including ERBB4 as cancer drivers, and of SREBP pathways and cholesterol in cancer cell maintenance (43, 46, 47) means that further elucidation of this signaling axis will provide insights into cancer metabolic dysregulation.

Materials and methods

(1) Reagents

The following reagents were used: NRG1 (Sigma), EGF (Sigma), lapatinib (Selleck), erlotinib (LC Laboratories), GDC0941 (Selleck), MK2206 (Selleck), rapamycin (LC Laboratories), PF-429242 (Tocris), and simvastatin (Selleck).

(2) Cell culture

T47D human breast cancer cells [American Type Culture Collection (ATCC)] were maintained in RPMI (Life Technologies) with 10% FBS, 1% pen/strep, and insulin (5 μg/ml, Gibco). Lipoprotein-deficient serum was prepared as described previously (48). Briefly, fumed silica was added to FBS (heat-inactivated, 70 mg silica/mL FBS) and was mixed overnight at 4°C. The mixture was ultracentrifuged at 11,290 × g for 10 min at 4°C to pellet lipoproteins and the LPDS layer was transferred to a fresh tube and filter sterilized twice. LPDS media (10% LPDS, RPMI) was tested for the ability to induce HMGCR following 24 hours treatment to confirm depletion of lipoproteins. MCF10A human breast cancer cells (ATCC) were maintained in DMEM/F12 (Life Technologies) with 5% horse serum, 1% penicillin/streptomycin, insulin (10 μg/ml, Gibco), EGF (20 ng/ml, Sigma), hydrocortisone (0.5 μg/ml, Sigma), and cholera toxin (100 ng/ml, Sigma).

(3) Quantitative RT-PCR

RNA was isolated with the RNeasy Mini Plus kit using QIAshredder columns (Qiagen). cDNA was prepared using the iScript kit (Bio-Rad). Real-time PCR was performed using Universal TaqMan Master Mix (Applied Biosystems) coupled with TaqMan FAM-labeled probes and ran on a ViiA 7 RT-PCR machine (Life Technologies). Relative mRNA expression was determined using the 2^−ΔΔCt method with GAPDH as the reference gene. The following Taqman probes were used: LDLR (Hs00181192_m1), HMGCR (Hs00168352_m1, HMGCS1 (Hs00940429_m1), GAPDH (Hs02758991_g1).

(4) Immunoblotting

Cells were lysed in buffers described below that were supplemented with protease inhibitor cocktail (Roche) and phosphatase inhibitors 2 and 3 (Sigma). For analysis of SREBPs (49), cells were lysed in SREBP lysis buffer (20 mM Tris-HCl pH 8.0, 120 mM KCl, 1 mM DTT, 2 mM EGTA, 0.1%Triton-X 100, 0.5% Nonidet P40). Samples were diluted in 2X Laemmli sample buffer and incubated at 100°C for 5 min. For detection of HMGCR (50), cells were lysed in lysis buffer A (10 mM Tris-HCl, 1% (w/v) SDS, 100 mM NaCl, 1 mM EDTA, 1 mM EGTA). Then, an equal volume of lysis buffer B was added (62.5 mM Tris-HCl, 15% (w/v) SDS, 8 M urea, 15% glycerol, 100 mM DTT). Samples were diluted in 6X Laemmli sample buffer (without 2-Mercaptoethanol) and incubated at 37°C for 20 min. For analysis of LDLR (51), cells were lysed in LDLR lysis buffer (50 mM Tris-HCL pH 6.8, 2.6 mM CaCl₂, 1% Triton-X 100).

Samples were loaded onto 4-12% Bis-Tris gradient gels (NuPAGE, Life Technologies) and run in MOPS buffer. Protein was transferred to PVDF membranes at constant amperage at 500 mA for 1 hour. Membranes were blocked in 5% milk/TBST (HMGCR) or 5% BSA/TBST (SREBP, LDLR) Membranes were incubated with horseradish peroxidase-conjugated secondary antibodies and developed by chemiluminescence (Pierce). The following antibodies were used for immunoblotting: SREBP-2 (BD Pharmingen, cat. no. 557037, mouse), HMGCR (from mouse hybridoma), LDLR (Cayman Chemical, cat. no. 10007665, rabbit), phosphorylated ERBB4 at Tyr¹⁰⁵⁶ (Santa Cruz, sc-33040, rabbit), phosphorylated EGFR at Tyr¹⁰⁶⁸ (Cell Signaling, cat. no. 3777, rabbit), EGFR (Santa Cruz, sc-03, rabbit), phosphorylated S6 kinase at Thr³⁸⁹ (Cell Signaling, cat. no. 9205, rabbit), p70 S6 kinase (Cell Signaling, cat. no. 9202, rabbit), phosphorylated AKT at Ser⁴⁷³ (Cell Signaling, cat. no. 4060, rabbit), AKT (Cell Signaling, cat. no. 9272, rabbit), and GAPDH (Santa Cruz, sc-25778, rabbit).

(5) LDL uptake and binding assays

T47D cells were incubated for 24 hours in 10% lipoprotein-deficient serum (LPDS) medium or grown in the presence of 10% FBS. During the final 6 hours, cells were incubated with or without NRG1 (50 ng/mL). For LDL uptake (51), diI-LDL (30 μg/mL) alone or diI-LDL along with unlabeled LDL (600 μg/mL), to account for unspecific uptake, was added to the medium for 2 hours. Cells were washed with PBS, trypsinized, and fixed in 1% paraformaldehyde (PFA)/PBS and stored at 4°C. For LDL binding, cells were treated with NRG1 for 6 hours then incubated on ice for 30 min to stop LDLR endocytosis. Then diI-LDL alone or diI-LDL along with unlabeled LDL, as above indicated, was added for 1.5 hours on ice. Cells were washed with PBS, trypsinized, and fixed in 1% PFA/PBS and stored at 4°C. LDL binding and uptake was analyzed by flow cytometry using the FACSCalibur (BD Biosciences) system. Specific diI-LDL uptake or binding was calculated using geometric mean after subtracting non-specific LDL uptake or binding. Samples were normalized to LPDS treated cells (52).

(6) Cholesterol Synthesis by Isotopomer Spectral Analysis

Fractional amounts of newly synthesized cholesterol were determined from the mass isotopomer distribution of cholesterol from cells incubated for 6 hours and 24 hours with [2-¹³C] acetate (Cambridge Isotope Laboratories) (31, 32). T47D cells were grown in LPDS media for 24 hours. Simvastatin (1 μM) was added to control wells simultaneously with LPDS. Media was aspirated and fresh LPDS was added along with NRG1 (50 ng/ml) and 2 mM acetate or [2-¹³C] acetate. Following 6 or 24 hours of incubation with NRG1 and labeled acetate, cells were scraped into PBS, pelleted, and flash frozen in liquid nitrogen. Cells were resuspended in 500 μl water, disrupted by pulse sonication and lipids extracted in 3 ml CH₃Cl:CH₃OH (2:1). The organic phase was dried under N2 gas and silylated with 150 μl N,O-bis(trimethylsilyl)trifluoroacetamide (BSTFA) at 60^oC for 30 min. Cholesterol isotopomer distribution was determined at the Yale Mouse Metabolic Phenotyping Center from GC-MS (HP6890-5975C MSD) in the EI mode monitoring masses from M-1 to M+18 (367-368). ISA was performed as previously described (31, 32).

(7) Gene set enrichment analysis

GSEA was performed on transcription profiles from MCF10A cells overexpressing ERBB4 ICD CYT-1 or CYT-2. Our identification of genes significantly altered by ERBB4 ICD (adjusted p-value < 0.05) compared with empty vector controls was previously reported (18), GEO GSE57339). The REACTOME_CHOLESTEROL_BIOSYNTHESIS and HORTON_SREBF_TARGETS (53) gene sets were manually curated from the MSigDB_v4.0 (Broad Institute). ERBB4 CYT-1 genes (n=6865) or CYT-2 genes (n=5965) were first rank ordered by fold-change in expression over vector-transfected cells. This list was evaluated with GSEA under default settings (GSEAPreranked, 10,000 permutations; (54), http://www.broad.mit.edu/gsea/).

(8) Statistical analysis

Two-tailed Student’s t-tests were performed where appropriate. Error bars represent SD. *p<0.05, **p<0.01, ***p<0.001, ****p<0.0001.

Supplementary Material

Fig. S1

Supplementary Figure S1: Gene lists for cholesterol/SREBP enrichment in the ERBB4 ICD data set.

Fig. S2

Supplementary Figure S2: Fatty acid synthesis genes in the ERBB4 ICD data set.

Fig. S3

Supplementary Figure S3: NRG1 transcriptionally activates expression of SREBP-regulated genes and induces SREBP-2 cleavage.

Fig. S4

Supplementary Figure S4: PF-429242 reduces SREBP-2 cleavage and does not inhibit ERBB4 phosphorylation.

Fig. S5

Supplementary Figure S5: Impact of NRG1 and PMA on cleavage of ERBB4 juxtamembrane isoforms JM-a and JM-b.

Fig. S6

Supplementary Figure S6: Regulation of cholesterogenic gene mRNA and HMGCR protein by EGF and NRG1 in MCF10A cells.

Table S1

Supplementary Table 1. Determination of cholesterol biosynthesis by ISA.

One-Sentence Summary: ERBB4 connects the ERBB growth regulatory network with cholesterol metabolism.