Mesenchymal–epithelial interactions involving epiregulin in tuberous sclerosis complex hamartomas

Keratinocyte Proliferation and mTOR Activation in TSC Skin Tumors.

Tissue sections were stained for nuclear antigen Ki-67 as a marker of cell proliferation and phospho-S6 as a marker of mTOR activation. The number of Ki-67-positive cells in the epidermis of TSC AFs and PFs was clearly larger than that in the same patient's normal-appearing skin (Fig. 1). Cells that were positive for Ki-67 were seen mainly in the basal layer of the epidermis, with few apparent in the dermis.

Few cells in the patient's normal-appearing skin were positive for phospho-S6. These cells were restricted almost entirely to a thin layer of cells comprising the granular layer of the epidermis; almost no cells in the dermis were positive for phospho-S6 (Fig. 2A). In contrast, the acanthotic epidermis of TSC skin tumors was strongly positive for phospho-S6 (Fig. 2C). The dermis also contained spindle- and stellate-shaped cells positive for phospho-S6 scattered between collagen bundles, with greater numbers located adjacent to the epidermis. Acanthosis in TSC skin tumors was accompanied by both increased keratinocyte proliferation and mTOR activation.

Two-Hit Cells Are Present in the Dermis, but Not the Epidermis, of TSC Skin Tumors.

The pattern of staining for phospho-S6 in TSC skin tumors appeared consistent with the possibility that keratinocytes and/or scattered fibroblast-like cells in the dermis represented two-hit neoplastic cells. To identify these two-hit cells, samples were analyzed for LOH at the TSC1 and TSC2 loci. LOH was not detected in DNA extracted from microdissected epidermis or dermis of three different AFs (data not shown). Because LOH in two-hit cells could be masked by cellular heterogeneity, we turned to interphase fluorescence in situ hybridization (I-FISH), which allows individual nuclei to be scored for allelic deletion of TSC1 or TSC2, based on the presence of only one signal per nucleus rather than the expected two signals per nucleus for each gene. Keratinocytes derived from three AFs did not show allelic deletion of TSC1 or TSC2 (Table 1); almost all nuclei showed two signals each for TSC1 and TSC2, with 0–4% showing only one signal. (In normal tissues, one signal is often observed in a small percentage of cells because of insufficient hybridization or overlapping signals.) Touch preparations of the dermal side of normal-appearing skin and cultured fibroblasts derived from normal-appearing skin from TSC patients showed similar patterns, with 0–8% showing one signal for TSC1 and 2–12% showing a single signal for TSC2. Among touch preparations or cells cultured from AFs and PFs only 0–8% showed one signal for TSC1, but 11–51% of nuclei had one signal for TSC2. These results are consistent with allelic deletion of TSC2 in a fraction of cells located in the dermis of AFs and PFs, but not in the keratinocytes associated with these tumors, which suggested that the epidermal changes in the tumor originated from effects of tumor mesenchymal cells.

Epiregulin mRNA Content of Fibroblast-Like Cells Derived from TSC Skin Tumors Is Greater Than That of the Patient's Fibroblasts from Normal-Appearing Skin.

RNA from fibroblast-like cells grown from four AFs, three PFs, and normal fibroblasts from four patients were analyzed by using Affymetrix GeneChips. In total, amounts of 22 probe sets representing 20 genes were increased, and 33 probe sets representing 28 genes were decreased by 3-fold or more (tumor vs. normal) in both AFs and PFs [supporting information (SI) Table 3]. The mRNA with the greatest mean elevation in AFs and PFs was epiregulin, which was 11.8-fold that of the patient's normal fibroblasts in AFs and 22.5-fold in PFs. Another mRNA that was highly overexpressed was MCP-1, which was 9.5-fold the control level in AFs and 3.9-fold in PFs, consistent with our earlier experiments using filter-based arrays (17). Epiregulin, an EGF family member, was investigated further because it appeared relevant to the observed epidermal changes. Amounts of epiregulin mRNA measured in samples from 17 patients using real-time PCR were 3.7- to 690-fold the paired controls in AF cells (n = 13, P = 0.001) and 4.5- to 5,660-fold in PF cells (n = 9, P = 0.008; Table 2). The addition of 100 nM rapamycin for 24 h did not significantly change epiregulin mRNA levels (P = 0.84; SI Table 4).

Release of Epiregulin Protein by Fibroblast-Like Cells from TSC Skin Tumors.

After incubation of cells in medium containing [³⁵S]methionine for 24 h, samples of medium were subjected to immunoprecipitation with antibodies against epiregulin, which yielded immunoreactive proteins of ≈5 kDa by using medium from fibroblast-like cells of TSC skin tumors but not from patients' normal-appearing skin (Fig. 3 and SI Fig. 6). The addition of unlabeled recombinant epiregulin protein before immunoprecipitation eliminated this band but not other, apparently nonspecific bands. Incubation of cells with EGF did not measurably alter epiregulin release by TSC tumor cells or normal skin fibroblasts (Fig. 3). Searchlight Protein Array analysis showed that the levels of epiregulin in conditioned medium from PF cells of two patients were 12.8 and 43.6 pg/ml, respectively, whereas the protein in conditioned medium from their normal TSC fibroblasts was undetectable. Changes in levels of other EGF family members (EGF, TGF-α, heparin-binding EGF-like growth factor, and amphiregulin) and keratinocyte growth factor, if seen, were inconsistent (data not shown).

Keratinocyte Proliferation and Phosphorylation of Ribosomal Protein S6 Enhanced by Epiregulin.

Recombinant epiregulin added in vitro to normal human keratinocytes stimulated cell proliferation in a concentration-dependent manner (Fig. 4), with a maximum 1.9-fold increase in the rate of BrdU incorporation with 20 ng/ml. Treatment of keratinocytes with epiregulin increased phosphorylation of ribosomal protein S6 within 10 min, which persisted at 1 h and decreased to basal levels after 24 h (Fig. 5 A and C). The epiregulin effect was concentration-dependent and maximal at 100 ng/ml (Fig. 5 B and D). Epiregulin increased phosphorylation of p70S6K similar to ribosomal protein S6 (data not shown). The data indicate that epiregulin enhanced keratinocyte proliferation and mTOR activation.

Tumor Sample and Cell Culture.

Patients meeting clinical criteria for diagnosis of TSC (40) were enrolled in protocol 00-H-0051 approved by the National Heart, Lung, and Blood Institute/National Institutes of Health (NIH) Institutional Review Board. Samples of AFs, PFs, and normal-appearing skin were obtained and bisected, with one portion used for routine pathology and the other used for frozen sections or cell culture. Fibroblasts from normal-appearing skin or fibroblast-like cells from AFs and PFs were isolated from explants as described in ref. 41 and then cultured in DMEM supplemented with 10% FBS, 100 units/ml penicillin, and 100 μg/ml streptomycin. Keratinocytes from TSC skin samples were obtained by using standard methods (42). Keratinocytes were cultured in serum-free medium (keratinocyte-SFM; Invitrogen), supplemented with EGF and bovine pituitary extract. For studies of the effects of epiregulin on proliferation and cell signaling, keratinocytes from foreskins of unidentified normal neonates were used (provided by Jonathan Vogel, National Cancer Institute, NIH).

Immunohistochemical Staining.

Sections were incubated with mouse anti-human Ki-67 monoclonal antibody (1 μg/ml, clone MIB-1; DakoCytomation), rabbit anti-phospho-S6 (1:200; Cell Signaling Technology), or source- and subtype-matched control IgG (DakoCytomation). Sections were stained with a Vectastain Elite ABC kit and 3,3′-diaminobenzidine peroxidase substrate kit or Vectastain ABC-alkaline phosphatase with Vector red substrate (Vector Laboratories) and counterstained with Mayer's hematoxylin solution (Sigma–Aldrich).

I-FISH.

Cells obtained from touch preparations or culture of TSC normal-appearing skin and skin tumors were analyzed for allelic deletion of TSC1 and TSC2 by using I-FISH as was described in ref. 43. At least 50 nuclei were scored for each sample.

Gene Expression Array.

Early passage cells (passages 3–5) were grown until 70–80% confluent. Total RNA was extracted by using an RNeasy minicolumn (Qiagen), and gene expression analysis was performed as described in the Affymetrix expression analysis technical manual by using 10 μg of total RNA. Fragmented, biotinylated cRNA was hybridized to human genome U133A GeneChips (Affymetrix). Fluorescent probes were added, and intensities were quantified with a GeneArray scanner. Arrays that passed initial quality control requirements were analyzed with Affymetrix MAS 5.1. A total of 11 arrays from four patients were analyzed. Each PF and AF were first compared with the respective patient's control cells from behind the ear or the back. Analysis of the intensities showed that differences between the back and ear locations were small compared with differences between controls and tumor tissue. Also, a comparison of results using normalization methods based on individual controls vs. grouped controls showed little effect (with 994 of 1,180 possible probe sets in common). Probe sets that were identified as “present or marginal” by MAS 5.1 were used for subsequent analysis. Lists of probe sets were constructed for evaluation including those with threefold change and/or statistically significant with a false discovery rate <1% by using a Benjamini–Hochberg multiple testing correction.

Real-Time PCR.

Tumor cells or normal fibroblasts grown from AFs, PFs, or normal skin were incubated in 1% FBS/DMEM with or without 100 nM rapamycin for 24 h before isolation of total RNA from the cells by RNeasy mini kit (Qiagen). Epiregulin mRNA levels were quantified by real-time PCR using the TaqMan universal master mix with assay-on-demand epiregulin primers (Applied Biosystems, Hs 00154995_AI), and values were expressed relative to levels in the same cells of 18S rRNA by using assay-on-demand primers Hs 99999901_S1. Real-time PCR was performed by using the GeneAmp 5700 sequence detection system (Applied Biosystems).

Immunoprecipitation and Multiplex Assay.

Epiregulin protein in conditioned medium of TSC skin tumor cells was detected by using metabolic labeling and immunoprecipitation as reported in ref. 30. Briefly, patient normal fibroblasts or skin tumor cells (1 × 10⁶ cells per 10-cm dish) were labeled in 5 ml of methionine-free DMEM containing 10% dialyzed FBS (Invitrogen) and 1 mCi of [³⁵S]methionine (Amersham Pharmacia Biotech). To some, 100 ng/ml EGF was added at the beginning of the labeling. After 24 h, conditioned media were collected, and an aliquot (1.25 ml) of each medium was used for immunoprecipitation with rabbit anti-epiregulin IgG. The immunoprecipitants were subjected to electrophoresis in 18% Tris-glycine gels (Invitrogen) and analyzed by using the FLA-5100 Bio-Imaging System (Fujifilm Medical System, Inc.). Epiregulin protein released by TSC skin tumor cells was quantified by Thermo Scientific SearchLight multiplex assay (Pierce) according to the manufacturer's procedures.

Cell Proliferation.

Keratinocytes from neonatal human foreskins were added to 96-well Costar Black plates with clear bottoms (1 × 10⁴ cells per well). The next day, wells were washed with keratinocyte-SFM without bovine pituitary extract and human recombinant EGF, and cells were incubated for 24 h followed by incubation for 24 h more in the same medium containing the indicated concentrations of recombinant epiregulin. Cells were labeled with 10 μM BrdU for 3 h, and incorporated BrdU was quantified by using a chemiluminescent ELISA kit (Roche), with chemiluminescence measured with a Fluoroskan Ascent FL (Thermo Labsystems).

Western Blot Analysis.

Neonatal human foreskin keratinocytes were plated on 60-mm dishes at 6 × 10⁵ cells in keratinocyte-SFM. The next day, medium was changed to keratinocyte-SFM without EGF and bovine pituitary extract and incubated for 24 h. Cells were treated with epiregulin as indicated and lysed in protein extraction buffer [20 mM Tris (pH 7.5), 150 mM NaCl, 1% Nonidet P-40, 20 mM NaF, 2.5 mM Na₂P₄O₇, 1 mM β-glycerophosphate, 1 mM benzamidine, 10 mM p-nitrophenyl phosphate, 1 mM phenylmethylsulfonyl fluoride]. Equivalent samples of total proteins were separated in 10% (wt/vol) polyacrylamide gels and transferred to 0.45-μm Invitrolon PVDF membranes (Invitrogen) before immunoblotting with 5 μg of anti-phospho-S6 ribosomal protein (Ser-235/236) or anti-S6 ribosomal protein antibodies (Cell Signaling), horseradish peroxidase-conjugated anti-rabbit antibodies (GE Healthcare), and a SuperSignal West Pico chemiluminescence detection kit (Pierce). Band intensity was measured by using a Kodak Capture DC 290 imaging system (Eastman Kodak).

Statistical Analysis.

Epiregulin mRNA data were evaluated by using Wilcoxon signed-rank test. Proliferation (chemiluminescence values) and S6 phosphorylation (ratio of phospho- to total S6 on a log scale) were assessed by using two-way ANOVA with the primary variable epiregulin or time adjusted for differences in baseline level among experiments, followed by Dunnett t test to compare each treatment group with control. Significance was defined as P < 0.05.