A Naturally Occurring HER2 Carboxy-Terminal Fragment Promotes Mammary Tumor Growth and Metastasis

Materials.All plasmid constructs of HER2 were derived from a cDNA clone identical to the published sequence gi:183986. The different cDNA constructs were made using standard PCR, sequencing, and cloning techniques.

Antibodies were from Cell Signaling (anti-P-HER2 [no. 2249], anti-P-Erk1/2 [no. 9101], anti-Erk1/2 [no. 9102], anti-P-Jnk [no. 9251], anti-P-Akt [no. 9275 and 9271], anti-Akt [no. 9272], anti-P-Src [no. 2101], and anti-P-PLCg1 [no. 2821]), BD Biosciences (anti-ITGA2, anti-ITGA5, and anti-ITGB1), Santa Cruz Biotechnology (anti-MET and anti-PHLDA), Upstate (anti-EphA2), Trevigen (anti-GAPDH), Abcam (anti-LDH), BioGenex (anti-HER2 [CB11]), Amersham (anti-rabbit immunoglobulin G [IgG] and anti-mouse IgG, both horseradish peroxidase linked), and Invitrogen (anti-mouse IgG linked to Alexa Fluor 488).

Lapatinib was kindly provided by GlaxoSmithKline, Research Triangle Park, NJ.

Cell culture.MCF7 Tet-Off cells (BD Biosciences) were maintained at 37°C and 5% CO₂ in Dulbecco's minimal essential medium/F-12 (1:1) (Gibco) containing 10% fetal bovine serum (Gibco), 4 mM l-glutamine (PAA Laboratories), 0.2 mg/ml G418 (Gibco), and 1 μg/ml doxycycline (Sigma). The BT474 cells were cultured in the same medium but without G418 and doxycycline. Cells were transfected with the various expression plasmids by using FuGENE6 (Roche). Single stable clones with pUHD10-3h-based plasmids integrated were selected with 0.1 mg/ml hygromycin B (Invitrogen). Expression from pUHD10-3 h-encoded cDNAs of HER2 and CTFs was induced by removing doxycycline. First the cells were detached with 0.5% trypsin-EDTA (GIBCO) and washed three times by centrifugation, and the medium was changed 10 h after seeding in culture dishes. Homogeneity of the individual clones was checked by immunofluorescence confocal microscopy with an antibody against the cytoplasmic domain of HER2. Two independently selected stable clones (i.e., -A and -B in Fig. S6 in the supplemental material) were used in the experiments presented throughout this report.

P95 was induced in BT474 cells by treatment with 0.75 mM APMA (4-aminophenyl mercuric acetate) for 20 min, with or without 1 h of pretreatment with 1 mM of the inhibitor 1,10-phenanthroline.

Biochemical methods.Extracts for immunoblots were prepared in modified radioimmunoprecipitation assay (RIPA) buffer (20 mM NaH₂PO₄/NaOH, pH 7.4, 150 mM NaCl, 1% Triton X-100, 5 mM EDTA, 100 mM phenylmethylsulfonyl fluoride, 25 mM NaF, 16 μg/ml aprotinin, 10 μg/ml leupeptin, and 1.3 mM Na₃VO₄), and protein concentrations were determined with DC protein assay reagents (Bio-Rad). Samples were mixed with loading buffer (final concentrations: 62 mM Tris, pH 6.8, 12% glycerol, 2.5% sodium dodecyl sulfate [SDS]) with or without 5% beta-mercaptoethanol and incubated at 99°C for 5 min before fractionation of 15 μg of protein by SDS-polyacrylamide gel electrophoresis (PAGE). Specific signals in Western blots were quantified with the software ImageJ 1.38 (NIH).

Cells for immunofluorescence microscopy seeded on glass coverslips were washed with phosphate-buffered saline, fixed with 4% paraformaldehyde for 20 min, and permeabilized with 0.2% Triton X-100 for 10 min. For blocking and antibody binding, we used phosphate-buffered saline with 1% bovine serum albumin, 0.1% saponin, and 0.02% NaN₃, and for mounting, we used Vectashield with DAPI (4′,6′-diamidino-2-phenylindole) (Vector Laboratories).

Glycosylation was examined by incubating ON at 37°C modified RIPA cell lysate (15 μg of protein) with or without 1 μl (1 unit) of N-glycosidase F (Roche).

Membrane and cytosolic fractionation of breast tumor tissue samples was achieved by ultracentrifugation precipitation of membranes. While frozen, the samples were cut in small pieces, mixed with separation buffer (50 mM Tris, pH 7.4, 150 mM NaCl, 250 mM sucrose, 10 mM EDTA, 1 mM phenylmethylsulfonyl fluoride, 5 mM NaF, 10 μg/ml aprotinin, 10 μg/ml leupeptin, 1 μg/ml pepstatin, and 1 mM Na₃VO₄), and homogenized with a Polytron instrument and a 22-gauge needle. Then, nonbroken material and organelles were removed by centrifugation three times, each for 10 min at 10,000 × g. The cleared lysates were centrifuged for 1 h at 100,000 × g. The resulting supernatants containing the cytosolic proteins were collected and centrifuged for an additional 1 h at 100,000 × g to remove traces of membrane proteins. The pellets from the first 1 h of centrifugation containing the membranes were washed in separation buffer containing an additional 1 M NaCl, in order to release membrane interacting proteins. Membranes were then recovered by centrifugation again for 1 h at 100,000 × g and finally suspended in modified RIPA buffer. For Western analysis, we used 50 μg of cytosolic proteins and a volume of membrane protein samples corresponding to the same percentage of the input.

Transcriptomic analysis of cell line model.For the transcriptomic analysis, we purified total RNA (Qiagen; RNeasy) from the MCF7 Tet-Off stable clones seeded 15 or 60 h earlier in the presence or absence of doxycycline. For immunoblot analyses, the same cells were seeded in parallel dishes (see Fig. S6 in the supplemental material). For the 15-h time point, we used the following clones: -A and -B (vector), H2-A and H2-B (HER2), 611-A and 611-B (611-CTF), 676-A and 676-B (676-CTF), and 687-A and 687-B (687-CTF). The following clones were used for the 60-h time point: -A (vector), H2-A (HER2), 611-A (611-CTF), 676-A (676-CTF), and 687-A (687-CTF). The integrity of the total RNA samples was validated in an Agilent BioAnalyzer Nanochip before amplification with a one-cycle target labeling protocol and by analysis on Affymetrix GeneChip expression probe arrays (Human Genome U133 Plus 2.0) at the UCTS facility in Vall d'Hebron University Hospital. For the 15-h samples, the RNA preparations were run twice, independently on arrays. For the 60-h samples, except for clones 676-A and 687-A, two independent RNA preparations of each condition were analyzed. Except for the first array run of the 15-h samples of clone H2-A, the RNA samples from clones in the presence and absence of doxycycline were analyzed completely in parallel at the facility. The data files of in total 56 arrays were analyzed in the program ArrayAssist 5.5.1 (Stratagene) with probe levels normalized by the RMA (robust multichip average) algorithm. Consistency of the data sets was verified by determining how many of the 54,675 probe sets varied more than twofold when comparing the doxycycline presence and absence data from a clone in the same array run (see Table SII, column 5, in the supplemental material). In the case of the 60-h samples of clones 676-A and 687-A, only 1 and 21 probe sets representing 0 and 17 genes, respectively, varied more than twofold. For the rest of the conditions, we took advantage of the experimental duplication to determine the number of probe sets with more than twofold differences in pairwise t tests with a P of <0.05 (see Table SII, column 7, in the supplemental material). Expression of 611-CTF for 15 and 60 h, and HER2 for 60 h, led to significant changes of 120, 690, and 150 probe sets, respectively, representing in total 624 different genes. Subsequently, all probe sets of these genes with more than twofold changes in at least one of the conditions were exported to Excel, where the average n-fold induction for each gene in each condition was calculated (see Table SI in the supplemental material).

Transcriptomic analysis of publicly available data on primary breast tumors.For expression analysis in breast tumors, we downloaded the gene array data with GEO accession numbers GSE1456 and GSE3494. The GSE1456 data set consists of 159 profiles of primary breast tumors collected at the Karolinska Hospital in Sweden from 1 January 1994 to 31 December 1996, with clinicopathological information available on all patients. The GSE3494 data set consists of 251 profiles of primary breast tumors collected in Uppsala County in Sweden from 1 January 1987 to 31 December 1989, with clinicopathological information available on 236 of the patients. The two data sets had been obtained by RNeasy Mini kit (Qiagen) extraction of total RNA followed by Affymetrix U133 A and B array analysis. The Excel files of all samples with patient information from the two data sets were opened as two independent projects with RMA normalization without baseline transformation in the program GeneSpring GX 9.0. Of the 624 genes identified as regulated by 611-CTF and HER2 in our cell line model, 599 represented by 1,416 probe sets were found in the U133 data. Without filtering for minimum detection threshold or expression variation, we exported all values of the 1,416 probe sets from the 159- and 236-profile data sets to Excel. Here the data sets were baseline and log₂ transformed and then fused to give a data set of 395 profiles. The HER2 status of the original 159-profile, but not the 236-profile, data set was available. In the 236-profile data set, we defined patients as HER2 positive if the value of at least one of the three different probe sets targeting HER2 was more than 2.5 times higher than its total average. If this definition had been applied to the 159-profile data set, five tumors would have been classified as false positive and another five as false negative with respect to the determinations made by the pathologist. In order to examine the importance of different subsets of genes regulated in our cell line analysis, we extracted all probe set values representing the chosen genes in all 395 tumor profiles and performed unsupervised hierarchical average linkage clustering with correlation-centered similarity metrics in the program Cluster (Michael Eisen, Stanford University). Clustering results were imported in the program Treeview in order to save heatmaps as .bmp files and dendrograms as .ps files. Kaplan-Meier survival analyses of clustered groups were done with the Excel add-in XLSTAT 2008.

TG mice.Transgenic (TG) 611 and TG 687 mice were engineered by cloning the sequences encoding 687-CTF and 611-CTF into the multiple cloning site II downstream of the Rous sarcoma virus-enhanced mouse mammary tumor virus long terminal repeat of the pMB vector (a kind gift from Marcos Malumbres, CNIO, Madrid, Spain). Founder lines were generated by microinjecting linearized plasmid DNA into fertilized oocytes harvested from superovulated FVB mice in the Centre of Animal Biotechnology and Gene Therapy (Centre de Biotecnologia Animal i Teràpia Gènica, Universitat Autònoma de Barcelona). Founder mice were genotyped by Southern hybridization analysis. After identification of founder animals, routine colony maintenance was performed by PCR genotyping. The male and female FVB/N-Tg(MMTVneu)202J mice were obtained from the Jackson Laboratory (Bar Harbor, ME).

Whole mounts and histology.Mammary glands were mounted on glass slides, fixed overnight in 4% paraformaldehyde, and transferred to 70% ethanol. The slides were rinsed in water for 5 min and stained in a filtered solution of 0.2% carmine for 24 h. Glands were then dehydrated sequentially with decreasing concentrations of ethanol and then defatted and stored in methyl salicylate. For histological analysis, fixed glands were blocked in paraffin, sectioned, and stained with hematoxylin and eosin.

Generation of a cellular model to characterize CTFs of HER2.Alternative initiation of translation of the mRNA encoding HER2 from methionine codons 611 and 687 leads to synthesis of two CTFs that differ by only 76 amino acids (1). However, this sequence could be functionally relevant because it contains a short extracellular region, a transmembrane domain, and a NLS (Fig. 1A and B).

Proteolytic shedding of the HER2 extracellular domain occurs by alpha-secretase cleavage after alanine 645 or arginine 647 (44). This cleavage generates a CTF, P95, with five to eight extracellular amino acid residues. Many products of the alpha-secretases are subsequently cleaved by the gamma-secretase complex. A putative gamma-secretase cleavage of P95 would generate an intracellular soluble fragment starting around lysine 676 and containing the NLS (Fig. 1A and B).

To individually express the different CTFs, we constructed plasmids with the corresponding cDNAs (Fig. 1A) under the control of a promoter repressible by the tetracycline analog doxycycline and transfected them into MCF7 cells. This cell line expresses low levels of HER2 and undetectable levels of CTFs and has been widely used to study signaling pathways involved in tumor progression.

Western blot analysis of stable clones confirmed that each cDNA construct produced a characteristic set of CTFs (Fig. 1C). Detailed characterization of these HER2 isoforms by immunofluorescence microscopy (Fig. 1D; also see Fig. S1 in the supplemental material) and a variety of biochemical techniques (see Fig. S2 in the supplemental material) showed that CTFs containing the transmembrane domain, 611- and 648-CTFs, were efficiently delivered to the cell surface plasma membrane. The two 611-CTF species corresponded to an intracellular precursor and a glycosylated cell surface transmembrane form. The 676- and 687-CTFs were soluble and localized to the cytoplasm and nucleus. Consistent with the presence of a NLS, the nuclear levels of 676-CTF were higher than those of 687-CTF (see Fig. S1 and S2 in the supplemental material).

Collectively these results demonstrated that the cell lines generated constitute an appropriate model to characterize the different CTFs of HER2.

CTFs expressed in human breast tumors.To determine the type of CTFs expressed by breast tumor cells, we compared them with CTFs from the transfected cell lines. First, we analyzed BT474 cells, which are derived from a human mammary carcinoma, because they overexpress HER2 as well as several CTFs (6, 9). Analysis by Western blotting showed that one of the CTFs expressed comigrated with transfected 611-CTF (Fig. 2A; also data not shown). Another fragment was identified as P95 because it migrated as 648-CTF and could be upregulated by APMA, a mercurial compound known to activate the metalloproteases that cleave HER2 (26). Furthermore, this upregulation could be blocked by 1,10-phenanthroline, a classic metalloprotease inhibitor that prevents the shedding of HER2 (8, 25). APMA treatment induced the disappearance of the low-molecular-weight fragments that comigrated with 676- and 687-CTFs (Fig. 2A). This effect was likely due to cell permeabilization by the mercurial compound.

• Open in new tab
• Download powerpoint

FIG. 2.

Characterization of CTFs from BT474 cells and breast cancer samples. (A) BT474 cells were treated with APMA, 1,10-phenanthroline, and/or control solvent, as indicated. Cell lysates were analyzed by Western blotting with an antibody against the cytoplasmic domain of HER2. (B) Lysates of BT474 cells treated with APMA and membrane fraction from tumor sample 108 were incubated overnight with or without N-glycosidase F, followed by Western blot analysis with an antibody against the cytoplasmic domain of HER2. (C) Tissue samples of human mammary tumors were fractionated, and equal amounts of total soluble (S) and membrane (M) fractions were analyzed by Western blotting with an antibody against the cytoplasmic domain of HER2 or, as a control, with an antibody against the cytosolic protein LDH.

611-CTF includes the N-glycosylated Asn-629, which is absent in P95 (Fig. 1A; also see Fig. S2 in the supplemental material). Thus, to further support the identifications made in Fig. 2A, we analyzed the N-glycosylation status of the CTFs from BT474 cells treated with APMA. The result of N-glycosidase F treatment was consistent with the identification of 611-CTF and P95 (Fig. 2B, left panel).

Next, we extended the analysis to human mammary tumor tissue samples selected on the basis of their high HER2 expression. Like with the BT474 cells, we identified the different CTFs from tumor samples by comparing their electrophoretic migration patterns with those of the transfected CTFs (see Fig. S3 in the supplemental material; also data not shown). Fractionation of the tumor samples showed that, as expected, the fragments identified as 611-CTF and P95 were membrane bound, while the CTFs comigrating with the 676- and 687-CTFs were largely soluble (Fig. 2C). Treatment with N-glycosidase F confirmed that the candidate 611-CTF was glycosylated, while P95 was not (Fig. 2B, right panel). These results showed that tumor samples contain both P95 and a fragment identical to that generated by alternative initiation of translation from methionine 611, as well as different soluble fragments that migrate as 676- and 687-CTFs.

In addition to malignant epithelial cells overexpressing HER2 and CTFs, the tumors contain a variety of stromal cell types that do not express, or express low levels of, HER2 and CTFs. Furthermore, the levels of total HER2 in individual epithelial tumor cells vary considerably, as seen in immunohistochemistry analyses (see, for example, reference 26). Nevertheless, the average levels of HER2 in tumors as determined by Western analysis appear less variable and similar to the expression level in our HER2 MCF7 stable transfectant (see Fig. S4 in the supplemental material). Thus, to quantitatively compare the expression of individual CTFs, we normalized their levels to the level of HER2 in the same tumor sample or cell line (Tables 1 and 2, respectively).

Transduction of signals by CTFs in cell lines.Activation of the intrinsic tyrosine kinase activity of HER2 leads to autophosphorylation and subsequent activation of signal transduction pathways. Incubation of immunoprecipitated CTFs with [γ-³²P]ATP led to radioactive labeling in all cases (see Fig. S5 in the supplemental material). Thus, all the model CTFs appear to be correctly folded and endowed with kinase activity.

We then monitored the statuses of specific components of the MAPK, Akt, Src, and PLCgamma signal transduction pathways (Fig. 3A). Expression of HER2 led to a progressive time-dependent accumulation of active components of these pathways (Fig. 3B; also see Fig. S6 in the supplemental material). Within the time frame chosen, the activation induced by HER2 did not reach a plateau. In contrast, 611-CTF induced a rapid and acute increase in the levels of active components that later decreased (P-Erk1/2, P-Akt, and P-Jnk) or reached a plateau (P-Src). In cells expressing 648-CTF, the activation was kinetically comparable to that in cells expressing 611-CTF, but the intensity was clearly lower (Fig. 3; also see Fig. S6 in the supplemental material).

• Open in new tab
• Download powerpoint

FIG. 3.

Activation of signal transduction pathways in cells expressing the different CTFs. (A) The MCF7 clones stably transfected with vector (-), HER2, or 611-, 648-, 676-, or 687-CTF, as in Fig. 1C, were washed with media with or without doxycycline (Dox), cultured for 24 h, lysed, and analyzed by Western blotting with the indicated antibodies. Lanes marked with an arrow were used for quantification. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (B) Western blot signals from the expressed constructs and the indicated phosphoproteins in the three independent experiments described for panel A (and in Fig. S6 in the supplemental material) at different time points were quantified. The level of each phosphoprotein was divided by the level of expressed HER2 or CTF in the same sample, followed by normalization to the level of phosphoprotein in cells transfected with empty vector. The averages are presented as bars ± standard deviations.

Importantly, the signaling activation was nearly identical in cells expressing two different levels of 611-CTFs (see Fig. S6, clone 611-A and 611-B in the supplemental material; also data not shown). Despite expressing levels that differed by ∼5-fold, the subsequent increases in phosphorylated signal transducers were kinetically and quantitatively similar. This result indicates that signal saturation was reached at relatively low levels of 611-CTF. The level of 611-CTF expressed in the 611-B cell line was comparable to those found in human mammary tumors (Table 1, tumor 145, and Table 2). Thus, pathophysiological levels of 611-CTF expression are signaling competent.

Even at the latest time point examined, expression of soluble 676- or 687-CTF did not lead to any changes in the statuses of the signal transducers (Fig. 3; also see Fig. S6 in the supplemental material). Therefore, neither cytoplasmic nor nuclear CTFs were able to activate, directly or indirectly, the MAPK, Akt, PLCgamma, or Src pathway.

611-CTF regulates the expression of a group of genes that correlates with poor prognosis in breast cancer patients.Despite engaging the same pathways, the activation levels of signal transduction by HER2 and transmembrane CTFs, particularly 611-CTF, were kinetically and quantitatively different (Fig. 3B). Thus, 611-CTF might regulate the expression of a distinct group of genes. To test this, we analyzed the transcriptome of cells expressing HER2 or 611-CTF at different time points. In addition, as an unbiased way to examine the possible activities of soluble nuclear or cytoplasmic CTFs, we included in the analysis cells expressing 676- or 687-CTF.

Consistent with the kinetic analyses, we found that 611-CTF expression led to a rapid regulation of gene expression and produced a change in the transcriptome of MCF7 cells that was far more profound than that induced by HER2 (Fig. 4A; also see Fig. S7A and Table SI in the supplemental material). In contrast, virtually no transcripts were affected by the expression of 676- or 687-CTF. The technical quality of the gene arrays was validated by verifying the regulation of several transcripts with real-time quantitative PCR (see Fig. S7B in the supplemental material).

• Open in new tab
• Download powerpoint

FIG. 4.

Transcriptomic analysis with HER2 and 611-CTF. (A) Venn diagram showing the number and overlap of genes regulated after 15 or 60 h of expression of 611-CTF or HER2 in MCF7. The identities and n-fold induction of in total 624 genes in the seven different groups are listed in Table SI in the supplemental material. GAPDH, glyceraldehyde-3-phosphate dehydrogenase. (B) Cells of the MCF7 clones stably transfected with HER2 or 611-CTF were washed with media with or without doxycycline (Dox), cultured for 15, 30, 60, or 90 h, lysed, and analyzed by Western blotting with the indicated antibodies. (C) Unsupervised hierarchical clustering of 395 primary breast tumors from two independent publicly available studies, on the basis of expression levels of the 76-gene signature of 611-CTF (see Fig. S11 in the supplemental material). *, patients that died from breast cancer; **, mean survival time (years). (D) Kaplan-Meier survival plots of the two patient groups classified in panel C.

Regulation of gene transcription is not always followed by corresponding changes in protein levels. We therefore sought validation of the increased expression of some gene products by Western blot analysis. Confirming the existence of genes specifically regulated by 611-CTFs, MET, MMP1, and the cell adhesion molecule integrin alpha 2 were not upregulated by HER2 (Fig. 4B; also see Fig. S8 in the supplemental material). Conversely, integrin alpha 5 was preferentially upregulated by HER2. Finally, PHLDA1, EPHA2, and integrin beta 1 were upregulated by both HER2 and 611-CTF, although as expected, to a higher extent by the latter. Part of these results was validated in a different cell line (see Fig. S9 in the supplemental material).

Hierarchical clustering of induction levels of the 624 genes identified as targets of HER2 and 611-CTF in MCF7 cells (Fig. 4A) divided them into 357 and 267 generally up- and downregulated genes, respectively (see Fig. S10A in the supplemental material). To investigate the expression patterns of these genes in human breast cancer, we combined two publicly available gene array data sets (24, 30). The resulting data set contained 395 transcriptomic profiles of primary breast tumors with corresponding clinical information on the patients (Table 3). Unsupervised clustering of the tumor profiles on the basis of the expression levels of the 357 upregulated genes (in MCF7) resulted in two distinct groups (see Fig. S10B in the supplemental material). The smaller of the two groups contained a high proportion of patients with HER2-positive tumors and patients (HER2-positive and -negative) that died from breast cancer (see Fig. S10C in the supplemental material). This result is in agreement with the known ability of HER2 to regulate the expression of genes important for tumor progression (43).

To determine the importance of genes preferentially regulated by 611-CTF, we repeated the unsupervised clustering of the 395 profiles on the basis of the 76 genes upregulated more than threefold by 611-CTF and less than twofold by HER2 (see Table SI in the supplemental material). Again the profiles fell into two groups, of which the smaller one had a higher incidence of patients that died from breast cancer (Fig. 4C and D). The 76-gene signature contains many genes that are causally involved in metastasis, like those for MMP1, ANGPTL4, MET, and IL-11, as well as genes that contribute to various aspects of malignant development, for example, those for CD44, BCL2A1, ADAM9, PLAUR, EPHA1, and components of the EGFR pathway such as EGFR and TGFalpha (see Fig. S11 in the supplemental material). These results demonstrated that the genes regulated by 611-CTF play an essential role in breast cancer.

Mechanism of activation of 611-CTF.Previous reports have shown that deletions leading to an imbalance of cysteine residues in the extracellular domain of HER2 result in constitutive receptor activation due to the formation of disulfide-bonded dimers (37). Since 611-CTF contains an odd number of extracellular cysteine residues (Fig. 1A), we investigated whether it is activated through a similar mechanism. Analysis of cell lysates in the absence of a reducing agent unveiled the existence of species with the expected electrophoretic migration of 611-CTF dimers (Fig. 5A, upper panel). Furthermore, coexpressed FLAG- and hemagglutinin-tagged 611-CTFs efficiently coimmunoprecipitated, supporting the supposition that the high-molecular-weight complexes represent 611-CTF homodimers (see Fig. S12A in the supplemental material). The high-molecular-weight complexes were phosphorylated, indicating that they are active (Fig. 5A, bottom panel). Preincubation of cells with lapatinib, a tyrosine kinase inhibitor currently in clinical trials which targets both HER2 and EGFR, prevented the phosphorylation of 611-CTF dimers but not their generation (Fig. 5A). Thus, dimerization stabilized by disulfide bonds and subsequent transphosphorylation provided a feasible explanation for the hyperactivity of 611-CTF.

• Open in new tab
• Download powerpoint

FIG. 5.

Active 611-CTF complexes are maintained by disulfide bonds. (A) The MCF7 clones stably transfected with vector (-), HER2, or 611-, 648-, 676-, or 687-CTF, as in Fig. 1C, were washed with media without doxycycline, cultured for 24 h in the presence or absence of 1 mM lapatinib, lysed, fractionated by SDS-PAGE in the presence or absence of beta-mercaptoethanol, and analyzed by Western blotting with the indicated antibodies. Arrow, dimers. (B) Schematic showing the primary sequence of the extracellular juxtamembrane region of HER2 with the transmembrane domain (TM), the position of amino acids 611 and 648, and the five juxtamembrane cysteines marked. The schematics below the sequence show the different cysteine substitutions inserted in the cDNA of 611-CTF. (C) MCF7 cells were transiently transfected with the empty vector (-), the vector containing the cDNA of HER2, or wild-type 648- or 611-CTF or CTF-611 with one (C-S), three (3C-A), or five (5C-A) cysteines substituted. After 48 h, the cells were lysed and analyzed as in panel A. (D) The same lysates as in panel C with beta-mercaptoethanol were analyzed by Western blotting with the indicated antibodies. The specific signals of three independent experiments were quantified as in Fig. 3B. Statistical analysis of the normalized ratios using Student's t test showed statistically significant differences for P-Erk1/2 and P-Akt (S473) between cells expressing 611/5C-A and wild-type 611-CTF (*, P < 0.01).

To corroborate this conclusion, we transiently transfected MCF7 cells with 611-CTF constructs with one or several of these cysteines mutated (Fig. 5B). Mutation of one cysteine reduced the generation of the high-molecular-weight 611-CTF complexes (Fig. 5C). Mutation of three and five cysteines further reduced and almost prevented, respectively, the generation of the complexes, further supporting the fact that they are maintained by disulfide bonds. Analysis of P-Erk1/2 and P-Akt levels in cells expressing the different constructs verified that the cysteines are required for the full activity of 611-CTF (Fig. 5D). Since the transfection efficiency of MCF7 cells was low (∼20%), we used the highly transfectable HEK293T cell line to confirm that the formation of disulfide-bridged complexes is the mechanism of hyperactivation of 611-CTFs (see Fig. S12B in the supplemental material).

Generation of animal models to characterize the effect of CTF expression in vivo.Mouse models have been instrumental in showing the oncogenic potential and relevance of HER2 in tumor progression (38). To characterize its oncogenic potential, we established TG mice expressing 611-CTF under the control of the mouse mammary tumor virus long terminal repeat, which is preferentially active in the mammary gland. Although the cellular models indicated that soluble intracellular CTFs are inactive, to further explore the consequences of expression of these fragments, we also generated TG animals expressing 687-CTF. As a control, we used the classical and well-characterized model expressing wild-type HER2 (i.e., rat neu).

At an age of 7 weeks, the levels of 611-CTF expressed in the heterozygous TG 611 lines F3 and F2 were approximately equal to and one-third of, respectively, the levels of endogenous HER2, while the level in F1 was below the detection threshold (Fig. 6A). The levels of 687-CTF in the homozygous lines developed varied from approximately double to half the levels of HER2 in the TG 687 lines F2 and F1, respectively (Fig. 6A).

• Open in new tab
• Download powerpoint

FIG. 6.

Expression of 611-CTF in mouse mammary glands results in mild abnormal alveolar development. (A) Total protein lysates from mammary glands of 7-week-old TG mice were analyzed by Western blotting with an antibody against the cytoplasmic domain of HER2 (left). The average expression, normalized to HER2 in wild-type (WT) mice, in four independent Western blot analyses is presented as bars ± standard deviations (right). (B) Carmine-stained whole mounts (left), ×25 magnifications of the same whole mounts (middle), and hematoxylin and eosin-stained sections of mammary glands (right; magnification, ×400) from 7-week-old animals.

Mammary glands of the TG animals exhibited no macroscopic abnormalities at 7 weeks of age. However, as previously reported (14), morphological examination of carmine-stained whole mounts revealed hyperplasic abnormalities in the mammary ductal trees of TG HER2 mice (Fig. 6B). Similar abnormalities, albeit less pronounced, were present in all three lines of TG 611 mice (Fig. 6B; also data not shown). In contrast, the glands of TG 687 mice were indistinguishable from those of wild-type mice.

611-CTF expression leads to the development of aggressive mammary tumors.Despite the more-pronounced hyperplasia in TG HER2 mice, the three lines of TG 611 animals developed more aggressive tumors, in terms of number of tumors per animal (Fig. 7A and B), tumor onset (Fig. 7C and Table 4), and tumor growth (Fig. 7D). No tumors or abnormalities were observed in TG 687 animals even after a follow-up of more than one year (Fig. 7C).

• Open in new tab
• Download powerpoint

FIG. 7.

TG 611-CTF mice develop mammary adenocarcinomas more aggressive than those in HER2-expressing mice. (A) Representative TG HER2 (35-week-old) and 611-CTF (25-week-old) mice. (B) Numbers of tumors that had developed in 40-week-old mice. Animals with tumors in every gland were counted as 10. The averages are presented as bars ± standard deviations. (C) Tumor-free survival curves of TG HER2 (n = 32), 611-CTF (n = 7), and 687-CTF (n = 26) mice. (D) Individual tumors from the indicated TG animals were measured at different time points. Each point represents the average ± standard deviation of ten individual tumors.

Histological analysis of the tumors showed the same typical invasive solid nodular carcinomas induced by HER2 (27) in the TG 611 mice (see Fig. S13A in the supplemental material). The only histological difference between the tumors initiated by HER2 and 611-CTF was a higher number of mitotic images in the ones from TG 611 mice (see Fig. S13A in the supplemental material; also data not shown).

As previously shown, TG HER2 mice developed lung metastasis (14). Three to six weeks after detection of breast tumors by palpation, immunohistochemical analyses of sacrificed animals showed that approximately one-fourth of TG HER2 mice (two of nine) had nodules in the lungs (data not shown). Histological analysis of these lung metastases confirmed the expression of HER2, and staining against cytokeratin 18 verified that the cells originated from the primary tumor (see Fig. S13B in the supplemental material). In comparison to that for the TG HER2 animals, the frequency of detectable metastases in mice expressing 611-CTF was more than double (five of nine) (data not shown). This shows that the tumors initiated by this CTF have a more-pronounced tendency to invade the lungs.

Overexpression of HER2 alone does not seem sufficient to generate mammary tumors in mice. The TG HER2 mice only develop tumors after genetic activation of the transgene (35). This consists of in-frame deletions, insertions, or mutations that affect the number of cysteine residues in the extracellular domain. The imbalance of cysteines results in the constitutive dimerization through the formation of intermolecular disulfide bonds (36). In agreement with this, we identified a variety of genetic alterations in the transgene from tumors developed in TG HER2 animals (see Fig. S14 in the supplemental material). All 13 independent alterations identified affected the number of cysteines, confirming that this is the preferred mechanism of oncogenic activation in mice.

In contrast, and consistent with the constitutive activity of 611-CTF, we did not find any genetic alteration in the transgene encoding 611-CTF in 23 independent tumors samples (data not shown). Collectively, these results clearly demonstrate the hyperactivity and oncogenic potential of 611-CTF in vivo and suggest that expression of this CTF could contribute to mammary tumor progression in humans.

The transmembrane domain of HER2 facilitates incorporation of CTFs into the secretory pathway.To analyze their functions, we generated stable transfectants expressing each individual CTF. This approach was straightforward for the soluble CTFs, because their expression from cDNA constructs is not expected to affect their subcellular localization. For the transmembrane CTFs, this strategy was feasible because CTFs lacking a signal peptide but containing the transmembrane domain efficiently incorporated into the secretory pathway (Fig. 1; also see Fig. S1 and S2 in the supplemental material). As a typical type I transmembrane protein, HER2 starts with an N-terminal signal peptide that inserts the nascent peptide into the lumen of the endoplasmic reticulum (ER). However, when located near the N terminus, as in 611- and 648-CTFs, the transmembrane domain was also able to recruit the necessary signal proteins to ensure insertion into the ER. Insertion in the membrane of type Ib integral proteins or single-pass type III membrane proteins, such as the neuregulins and cytochrome P450, is also directed by the transmembrane domain (3). However, to our knowledge HER2 is the first reported case in which this function has been attributed to the transmembrane domain of a type I protein that, in the full-length version, also contains a signal peptide.

611-CTF controls genes that predict poor prognosis in breast cancer.We have previously shown that the expression of CTFs correlates with poor prognosis and increased likelihood of developing metastasis (26, 32). However in these studies, patients were classified as CTF positive or negative without taking into consideration the type of CTFs expressed. The functional characterization presented here indicates that, while the expression of soluble CTFs is likely irrelevant, the transmembrane CTFs, especially 611-CTF, could play a causal role in the particularly poor clinical outcome of certain HER2-positive patients. In agreement with this suggestion, our transcriptomic analysis showed that the group of genes preferentially upregulated by 611-CTF predicted poor prognosis (Fig. 4; also see Fig. S11 in the supplemental material). In fact, several of the genes regulated by 611-CTF have been shown to participate in tumor spreading. Interesting examples include genes that play a causal role in metastasis, such as those for ANGPTL4 (29) and IL-11 (17). Furthermore, we validated target genes involved in metastasis, such as those for MET, EPHA2, and MMP1 (4, 11, 13), at the protein level (Fig. 4; also see Fig. S8 and S9 in the supplemental material).

Mechanism of activation of HER2 and CTFs.The invariable presence of a mutated transgene in tumors from TG HER2 mice suggests that, although it likely contributes, the mere overexpression of the receptor is not sufficient to drive malignant transformation in this animal model. Although the mutations are diverse, they all cause an imbalance in the number of cysteine residues (35) (see Fig. S14 in the supplemental material). This molecular abnormality results in constitutive activation of the receptor through the formation of covalent homodimers (36). Thus, intermolecular disulfide bonding appears to be the favored mechanism of oncogenic activation of HER2 in vivo. In agreement with this conclusion, we did not find any mutations in the transgenes of tumors induced by 611-CTF expression. Because of its intrinsic ability to form disulfide-bonded dimers (Fig. 5), this HER2 isoform does not require mutations in order to generate aggressive tumors in mice (Fig. 6 and 7).

In contrast to the situation in TG mice, HER2 is not frequently mutated in samples from human mammary tumors (34). Thus, it has been concluded that overexpression may be enough to activate HER2 in human cancers (16). Our results point to an additional possibility. Cells expressing even very low levels of 611-CTF in humans may have an advantage similar to that of cells expressing the mutant forms of HER2 in mice. This hypothesis may explain, at least in part, why HER2 mutations are rare in human tumors.

611-CTF in human tumors.It could be argued that if 611-CTF plays an important role in HER2-induced tumorigenesis, expression of this isoform, rather than mutations in the transgene, should occur in the HER2 TG mice. However, although the regulation of alternative initiation of translation in mammalian genes is poorly understood (18), it is clear that it does require cis-acting sequences that might be absent in the cDNA constructs used in the animal model. Nevertheless, a better understanding of the mechanism(s) responsible for the translation of 611-CTF is warranted in order to more precisely define its role in cancer. Furthermore, it cannot be ruled out that in the highly complex and heterogeneous tumor setting, in addition to alternative initiation of translation, other mechanisms may contribute to the generation of 611-CTF. It is not unlikely that incomplete gene duplication, aberrant mRNA splicing, and/or initiation of transcription lead to synthesis of 5′-truncated HER2 mRNA transcripts. Should these transcripts lack a region affecting the sequence between methionine codons 1 and 347 (the upstream methionine codon nearest to methionine 611), they may result in synthesis of 611-CTF.

Interestingly, several studies have reported the existence of an alternatively spliced form of HER2, known as ΔHER2, that lacks an exon encoding 16 amino acids of the extracellular juxtamembrane region (19). This deletion also causes an imbalance in the number of cysteines and the generation of dimers maintained by intermolecular disulfide bonds (37). Since ΔHER2 was detected in most breast cancer samples analyzed and accounted for 4 to 9% of total HER2 transcripts, this HER2 isoform likely contributes to tumor progression too.

Even though its activity was much lower than that of 611-CTF, 648-CTF (P95) was also capable of signaling (Fig. 3). Since this fragment of HER2 completely lacks the extracellular domain, its mechanism of activation remains unknown. In this respect, it was recently suggested that an undefined CTF of HER2 can form active complexes with HER3 (42). However, the generation of TG mice expressing the P95 isoform is needed in order to establish its pathophysiological importance.

Our finding that 611-CTF is a hyperactive form of HER2 with a clear link to metastasis-promoting genes suggests that in future studies, classification of patients according to the presence of this fragment could lead to improved clinical correlations. The development of tools to specifically and robustly detect the presence of this fragment would be instrumental in accomplishing this goal.