Activation of Src induces mitochondrial localisation of de2-7EGFR (EGFRvIII) in glioma cells: implications for glucose metabolism

Glioblastoma multiforme (GBM) is the most common and aggressive form of brain neoplasma affecting adults (Maher et al., 2001). The amplification of the EGFR gene is a common event in GBMs and is often accompanied by EGFR gene rearrangement (Ekstrand et al., 1992; Sugawa et al., 1990; Wong et al., 1992; Yamazaki et al., 1990); with the most common EGFR mutant found being the de2-7EGFR (or EGFRvIII) (Frederick et al., 2000). This mutant consists of an in-frame deletion spanning exons 2–7 of the coding sequence, resulting in the deletion of 267 amino acid residues from the extracellular domain and the insertion of a novel glycine residue at the junction site (Humphrey et al., 1991; Sugawa et al., 1990). As a result of this truncation, the de2-7EGFR is unable to bind any known ligand. Despite this, de2-7EGFR displays low level constitutive kinase activity that leads to the prolonged activation of downstream signalling pathways (Chakravarti et al., 2004; Li et al., 2004; Moscatello et al., 1998; Narita et al., 2002), partially due to the impaired internalisation and subsequent down-regulation of the receptor (Nishikawa et al., 1994; Schmidt et al., 2003).

Previous studies have demonstrated that the human-derived U87MG glioma cells expressing the de2-7EGFR have an in vivo growth advantage over the wild-type (wt) EGFR (Nishikawa et al., 1994). The enhanced tumorgenicity mediated by de2-7EGFR-expressing cells in part results from direct association or crosstalk between this truncated receptor and other cell-surface receptors such as the wtEGFR and Met (Huang et al., 2007; Luwor et al., 2001; Pillay et al., 2009). Prolonged activation of the PI3K–Akt pathway appears to be a central element of signalling in both GBM tumour samples (Chakravarti et al., 2004), as well as in human-derived GBM cell lines expressing the de2-7EGFR (Li et al., 2004; Moscatello et al., 1998; Narita et al., 2002). Recently, we demonstrated that the de2-7EGFR expressed in U87MG cells is constitutively phosphorylated at tyrosine 845 (Y845) by a member of the Src family kinases (SFKs) (Johns et al., 2007). Given that Y845 has been identified as the site responsible for the activation of Stat3 signalling by the wtEGFR (Mizoguchi et al., 2006), activation of this pathway might also be related to de2-7EGFR tumorgenicity.

There are two reports from the same group showing that wtEGFR can translocate to the mitochondria (Boerner et al., 2004; Demory et al., 2009). The authors hypothesised a mitochondrial localisation after showing that a phosphorylated, but not unphosphorylated, peptide containing Y845 bound the mitochondrial protein CoxII. They then showed that the wtEGFR could translocate to the mitochondria following ligand stimulation in the presence of Src, where it can phosphorylate CoxII. Mitochondrial localisation of wtEGFR appeared to be important in mediating the EGF protection of breast cancer cells from adriamycin-induced apoptosis. One concern is that this group did not show that their mitochondrial preparations were free of contaminating membranes from other organelles.

Using multiple techniques, we now demonstrate that the de2-7EGFR expressed in human-derived glioma cells is also colocalised with the mitochondria, an observation dramatically enhanced by activation of Src. Using the SFK inhibitor Dasatinib, as well as catalytically impaired Src or Y845 mutants, we demonstrated that the translocation of the de2-7EGFR to the mitochondria is dependent upon the phosphorylation of Y845 by Src. We also demonstrate in this present study, that the de2-7EGFR located at the mitochondria is fully glycosylated and constitutively active, implicating a functionally significant role for this receptor in the mitochondria.

The detection of ER-associated high-mannose forms of the de2-7EGFR on the plasma membrane (Johns et al., 2005) shows that the normal quality control mechanisms associated with glycoproteins might be overwhelmed by this mutant receptor. Therefore, using confocal microscopy techniques, we examined the localisation of the de2-7EGFR in human U87MG glioma cells. Confocal microscopy images acquired from fixed and permeabilised U87MG.Δ2-7 cells, immunostained with the de2-7EGFR-specific monoclonal antibody (mAb) 806 (Johns et al., 2002), demonstrated colocalisation with cadherins at the plasma membrane (supplementary material Fig. S1A) and the existence of a large intracellular pool of de2-7EGFR (supplementary material Fig. S1A).

We then determined whether the intracellular de2-7EGFR in U87MG glioma cells was localised within the ER or Golgi using organelle-specific antibodies. It was difficult to detect the de2-7EGFR in the ER by confocal microscopy (supplementary material Fig. S1B) and subsequent western blotting only detected small amounts of unglycosylated nascent receptor (supplementary material Fig. S2). Thus, only a small component of the intracellular de2-7EGFR is located within the ER. Localisation of the de2-7EGFR to the Golgi complex was examined using mAb 806 and the Golgi antibody ab24586, which specifically binds giantin. Images acquired by confocal microscopy clearly demonstrate the localisation of the de2-7EGFR within the Golgi complex (supplementary material Fig. S1C). These findings suggest that the de2-7EGFR is rapidly transported from the ER to the Golgi, where the process of terminal glycosylation is delayed, resulting in the accumulation of the high-mannose form of de2-7EGFR.

The specificity of mAb 806 for de2-7EGFR was confirmed by its lack of staining on U87MG parental cells that express the wtEGFR (supplementary material Fig. S3). The mitochondrial localisation of de2-7EGFR in U87MG.Δ2-7 cells was also assessed, and revealed colocalisation between mAb 806 and MitoTracker Red (MTR) (Fig. 1A). A small amount of de2-7EGFR was located with the mitochondria (Fig. 1A, overlay), as confirmed by voxel analysis of the acquired images (Fig. 1A, image zoom). Given this result, we examined whether the Src-dependent phosphorylation site of Y845 in the de2-7EGFR cells was phosphorylated. Consistent with our previous report (Johns et al., 2007), Y845 was phosphorylated and this phosphorylation was blocked by SFK inhibitors (Fig. 1B), whereas phosphorylation of a major trans-phosphorylation site, Y1173, was unaffected by either SFK inhibitor.

We further investigated this result by determining the expression level of Src in U87MG.Δ2-7 cells. Using intracellular FACS, U87MG.Δ2-7 cells showed low but reproducible levels of Src (Fig. 1C, bottom left). Because the mitochondria localisation of the wtEGFR required overexpression of Src, we generated U87MG.Δ2-7 cells expressing modest levels of activated Src or a dominant-negative form of Src (DNSrc). As expected, transfection of these constructs led to an increase in total Src (Fig. 1C, bottom), but retained similar levels of surface de2-7EGFR as determined by FACS (Fig. 1C, top).

We examined the influence of Src activation on the translocation of the de2-7EGFR to the mitochondria. U87MG.Δ2-7_DNSrc cells were devoid of mitochondrial localised de2-7EGFR (Fig. 1D, top). By contrast, de2-7EGFR was clearly colocalised with the mitochondria in cells expressing activated Src (Fig. 1D, bottom). Indeed, the general increase in intracellular de2-7EGFR was very evident (Fig. 1D, compare top left panel with bottom left panel). Multiple ‘z-section stack’ confocal microscopy images of U87MG.Δ2-7 cell lines (supplementary material Fig. S4) were analysed using ImageJ (NIH). Voxel analysis of these images, showed that significantly more of the total de2-7EGFR expressed in U87MG.Δ2-7_Src cells was localised to the mitochondria compared with U87MG.Δ2-7_DNSrc cells (24.4% versus 4.1%, respectively; P<0.0001) (Fig. 1E). This result suggests that the de2-7EGFR translocates to the mitochondria in a Src-dependent manner.

To support the confocal microscopy analysis, we purified mitochondria and analysed it for the presence of de2-7EGFR by immunoblotting. Fig. 2A shows the western blot analysis of isolated mitochondria washed multiple times with high-salt buffer. Along with mitochondrial marker Cpn60, this mitochondria fraction contained detectable amounts of lysosomal and ER proteins but was totally devoid of markers for plasma membrane or Golgi (Fig. 2A); the two other organelles contained full-length de2-7EGFR. Additional high-salt washing of the mitochondrial fraction from Fig. 2A resulted in a fraction devoid of mitochondria but enriched in endoplasmic reticulum. This ‘Endo’ fraction also contained some lysosomal membrane (Fig. 2B). Full-length de2-7EGFR was not detected in this Endo fraction (Fig. 2C), which is consistent with our observation that the ER only contains a small amount of a low molecular mass (100 kDa) nascent form of de2-7EGFR (supplementary material Fig. S2). Furthermore, we have previously demonstrated that de2-7EGFR in lysosomal compartments is degraded (Perera et al., 2005). Because the mitochondrial-enriched sample only contained full-length de2-7EGFR in a western blot (Fig. 2A, bottom panel) it must be derived from the mitochondria component of the preparation, and not the ER or lysosomal components.

Having demonstrated that Src activation is required for de2-7EGFR localisation to the mitochondria, we sought to confirm the importance of Src kinase activity in modulating mitochondrial translocation. To address this question, mitochondria were isolated from U87MG.Δ2-7_Src cells treated with either the SFK inhibitor Dasatinib (150 ng/ml) or vehicle, under serum-free conditions. This dose of Dasatinib is sufficient to inhibit the constitutive active Src expressed in these cells (Lu et al., 2009). Purified mitochondria were resolved by SDS-PAGE with equal loading confirmed by immunoblotting for the mitochondrial marker antibody, Cpn60 (Fig. 3A, bottom panel). Immunoblot analysis of the samples using mAb 806, which under denaturing conditions acts as a pan-EGFR antibody (Fig. 3A, top panel), demonstrated that both the de2-7EGFR and wtEGFR are localised within the mitochondria of U87MG.Δ2-7_Src control cells. Dasatinib completely ablated localisation of wtEGFR within mitochondria and caused a marked reduction in the level of de2-7EGFR localised within the mitochondria (Fig. 3A, top panel). Quantification of four independent experiments demonstrated a significant 33% reduction in de2-7EGFR localisation within the mitochondria in cells treated with Dasatinib relative to vehicle control (P=0.002; Fig. 3B, left panel). Because the half-life of de2-7EGFR in mitochondria is unknown, it is possible that the remaining receptor observed after 24 hours of Dasatinib treatment is part of the pre-existing pool within the mitochondria.

Immunoblot analysis using an anti-EGFR antibody that specifically recognises phosphorylation of Y845 confirmed that the de2-7EGFR localised within the mitochondria is phosphorylated at this Src phosphorylation site (Fig. 3A, second panel). As expected, Dasatinib led to a significant reduction in the level of Y845-phosphorylated de2-7EGFR within the mitochondria (P=0.01; Fig. 3B, middle panel). Phosphorylation at a major auto-phosphorylation site of the de2-7EGFR (Y1173) was also seen in receptor isolated from mitochondria (Fig. 3A, third panel), further confirming that mitochondrial-located de2-7EGFR is active and potentially capable of signalling within this organelle. However, unlike results for phosphorylation at Y845, the amount of de2-7EGFR phosphorylated at Y1173 was not significantly lower in cells treated with Dasatinib (P>0.05) (Fig. 3B, right panel).

The influence of Dasatinib on de2-7EGFR localisation to the mitochondria was further examined by confocal microscopy. Voxel analysis of acquired confocal microscopy images demonstrated a significant reduction in the level of de2-7EGFR colocalised with the mitochondria (Fig. 3C, image zoom). Computational analysis of cells revealed varying levels of de2-7EGFR colocalised in the mitochondria, which is consistent with the heterogeneity of cell lines used (Fig. 3D, left panel). However, quantification of cells examined clearly showed a significant reduction in de2-7EGFR in the mitochondria following Dasatinib treatment (Fig. 3D, right panel; P=0.002). Indeed, the 33% was virtually identical to that obtained by immunoblotting analysis (cf. Fig. 3B, left panel).

Having demonstrated that Src activity modulates the translocation of de2-7EGFR to the mitochondria, we wanted to confirm this result in a system that is not reliant on transfected Src. Thus, U87MG.Δ2-7 cells were grown on uncoated coverslips or coverslips covered with a thin layer of matrigel and then examined by confocal microscopy. Matrigel contains a number of ECM proteins known to activate Src and other molecules through engagement with integrins (Clark and Brugge, 1995). A significant increase in mitochondrial de2-7EGFR was demonstrated following plating of U87MG.Δ2-7 cells on matrigel (Fig. 4A, bottom panel). Voxel analysis of cells revealed that virtually all cells grown on matrigel and all mitochondria in these cells contained de2-7EGFR (Fig. 4B). Quantitative analysis showed that the amount of mitochondria colocalised with de2-7EGFR following engagement with matrigel increased from 33% to 99% (P<0.0001; Fig. 4B). To test whether the ECM component of Matrigel could mediate phosphorylation of the Y845 residue, U87MG.Δ2-7 cells were plated on uncoated plastic or plastic coated with collagen, which makes up 30% of the Matrigel ECM. Collagen stimulated a robust phosphorylation of de2-7EGFR at multiple tyrosine sites including Y845, when compared with results on plastic (supplementary material Fig. S5). Interestingly, collagen stimulated Y845 phosphorylation above the level of de2-7EGFR from cells grown in serum.

All our evidence suggests that the phosphorylation of Y845 by Src is required for de2-7EGFR mitochondrial localisation. Therefore we constructed a cell line co-expressing de2-7EGFR with a Y845F modification (i.e. unable to be phosphorylated by Src) and activated Src. FACS analysis confirmed that the level of de2-7EGFR expressed on the surface of U87MG.Δ2-7_845F cells was not compromised following the transfection of cells to produce the U87MG.Δ2-7_845F.Src cell line (Fig. 5A, top panel). Intracellular FACS analysis clearly demonstrates the increase in Src expression following transfection with activated Src (Fig. 5A, bottom panel).

Analysis of U87MG.Δ2-7_845F cells by confocal microscopy clearly demonstrates the lack of de2-7EGFR colocalised with the mitochondria (Fig. 5B, top panel). More importantly, expression of activated Src in the U87MG.Δ2-7_845F cells failed to induce the mitochondrial localisation of de2-7EGFR Y845F (Fig. 5B, middle panel). As previously, the U87MG.Δ2-7_Src cells showed clear localisation of de2-7EGFR in the mitochondria (Fig. 5B, bottom panel). These results demonstrate that the activation of Y845 by Src is essential for the translocation of the receptor to the mitochondria and is not merely the result of increased Src activity.

Having demonstrated that de2-7EGFR colocalises with the mitochondria, we examined the outer membrane of intact mitochondria isolated from U87MG and U87MG.Δ2-7_Src cells for de2-7EGFR localisation by electron microscopy (EM). De2-7EGFR on the outer membrane surface of mitochondria was visualised using mAb 806 and a gold-labelled anti-mouse secondary antibody. The mitochondria from U87MG.Δ2-7_Src cells (Fig. 6A), but not from U87MG parental cells, were positive for surface de2-7EGFR. Of a total of 126 mitochondria from U87MG.Δ2-7_Src cells analysed by EM, 76% were positive for de2-7EGFR, compared with 0% of 82 mitochondria from U87MG cells (Chi square=77.94, d.f.=1, P<0.0001). This result confirms the presence of de2-7EGFR in the mitochondria by a third independent technique, and suggests that the de2-7EGFR is localised in the outer mitochondrial membrane, although the orientation of the receptor is not known.

We treated purified mitochondria isolated from U87MG.Δ2-7_Src cells with proteinase K (PK) (Gough et al., 2009; Vande Velde et al., 2008). The level of de2-7EGFR detected in the mitochondria was significantly reduced in samples treated with PK compared with sham-treated samples (Fig. 6B, top panel). Thus, consistent with the EM data, this demonstrates that some of the de2-7EGFR is located on the surface of mitochondria and accessible to PK. However, the presence of some PK-resistant de2-7EGFR suggests that some of the de2-7EGFR is located within the mitochondria (Fig. 6B, top panel).

Given that de2-7EGFR preferentially activates the PI3-K–Akt pathway in glioma cell lines and tumour samples (Chakravarti et al., 2004; Li et al., 2004; Moscatello et al., 1998; Narita et al., 2002), which in turn increases cellular metabolism by increased aerobic glycolysis and lipogenesis (Guo et al., 2009; Swinnen et al., 2006; Warburg, 1956b), we examined the effect of low glucose on de2-7EGFR localisation. There was an increase in de2-7EGFR colocalised with the mitochondria following a reduction of glucose in the cell medium (Fig. 7A). A detailed statistical analysis showed that the proportion of mitochondria containing de2-7EGFR was significantly higher in the 1.6 g/l cohort compared with cells in medium containing 4.5 g/l of glucose (Fig. 7B).

The translocation of de2-7EGFR to the mitochondria under low-glucose conditions suggests that this receptor has a function in protecting glucose-stressed glioma cells. Therefore, we examined the proliferation and viability of U87MG and U87MG.Δ2-7 glioma cell lines in different concentrations of glucose. Both U87MG and U87MG.Δ2-7 cells proliferated in medium containing 25 mM glucose (Fig. 8A,B). Reduction to 5 mM glucose had no effect on U87MG.Δ2-7 cells until day 4, when there was a small decrease in proliferation (Fig. 8B). By contrast, U87MG cells displayed significantly reduced proliferation in 5 mM glucose on all days (Fig. 8A). Neither cell line proliferated at concentrations of glucose less than 1 mM (Fig. 8A,B). U87MG cells were viable in 25 and 5 mM glucose but had low viability in concentrations of 1 mM and less (Fig. 8C). By contrast, U87MG.Δ2-7 cells remained viable at concentrations as low as 0.5 mM glucose, with a significant drop in viability only seen in the complete absence of glucose (Fig. 8D). Thus U87MG.Δ2-7 cells show increased proliferation and survival under glucose stress. U87MG cell lines are null for PTEN, which is a crucial endogenous inhibitor of PI3K. Restoration of PTEN in U87MG.Δ2-7 cells completely re-established the glucose-sensitivity phenotype (supplementary material Fig. S6), highlighting the central role of PI3K in this process. Overall, these results suggest that de2-7EGFR in the mitochondria has a role in regulating glucose metabolism.

As shown in Fig. 9A, transfection of de2-7EGFR in to U87MG cells increased basal mitochondrial OCR but decreased extracellular acidification rate ECAR or glycolysis rate (i.e. the U87MG.Δ2-7 cells moved to a more oxidative metabolic phenotype). Given this observation, we explored whether de2-7EGFR alters mitochondrial function. Mitochondrial function was assessed by the sequential addition of oligomycin (ATP synthase inhibitor), FCCP (uncoupler of oxidative phosphorylation) and antimycin (complex III inhibitor) to cells; thus allowing determination of the ATP-coupled respiration, leaked respiration (proton leak) and maximal respiration capacity. These compounds were titrated for their optimal concentrations in separate experiments. As shown in Fig. 9B, maximal mitochondrial respiration capacity (i.e. FCCP-stimulated OCR) was significantly higher in U87MG.Δ2-7 cells than in U87MG cells. This suggests that the electron transport chain activity is unregulated by de2-7EGFR. Hence, de2-7EGFR not only elevated basal oxygen consumption rate, but also increased mitochondrial bioenergetic function, either through an increase in the number of mitochondria or raised levels of enzymes of the electron transport complex.

When expressed in a variety of cell lines de2-7EGFR runs as two distinct bands on SDS-PAGE (Johns et al., 2005). These two bands have also been observed in patient samples (Mellinghoff et al., 2005). Previously, we demonstrated that the lower molecular mass form of the de2-7EGFR corresponds to an immature, high-mannose form of the receptor (Johns et al., 2005). Most of this receptor is found inside the cell, although a small amount was misdirected to the cell surface. The studies described here show that this form of the receptor is concentrated in the Golgi, presumably awaiting terminal glycosylation, not in the ER as might be expected. Because this high-mannose form of the de2-7EGFR is constitutively active, this pool of the receptor might contribute to its tumorigenicity. The observation that signalling molecules downstream of de2-7EGFR, such as Ras, are present and active in the Golgi supports this possibility (Chiu et al., 2002).

Members of the ErbB family are generally anti-apoptotic, and signalling downstream from these receptors can modulate the activity of mitochondria-associated proteins, including the Bcl-2 family (Goel et al., 2007). Recently, it was demonstrated that the transforming activity of ErbB2 could be reduced by overexpressing DecR1 (Ursini-Siegel et al., 2007); a mitochondrial enzyme intimately involved in fatty acid β-oxidation. More surprising are the recent observations that a cleaved intracellular fragment derived from ErbB4 and the full-length wtEGFR can translocate to the mitochondria (Boerner et al., 2004; Naresh et al., 2006). The function of EGFR and ErbB4 located in the mitochondria remains unknown, although a role for EGFR in anti-apoptosis was indicated by its potential interaction with CoxII. Collectively, these studies show a growing relationship between several aspects of mitochondria function and activation of ErbB family members.

Our data definitively show that the cancer-specific de2-7EGFR can also translocate to the mitochondria. This translocation required activation of Src, but not necessarily the overexpression of Src. Unlike the published data for the wtEGFR (Boerner et al., 2004), mitochondrial localisation of de2-7EGFR was entirely dependent on the phosphorylation of Y845 because mutation of this site completely abrogated mitochondrial localisation, even in the presence of constitutively active Src. Because the mitochondrially localised de2-7EGFR was fully glycosylated, it was not derived from the high-mannose pool of intracellular receptor that resides in the Golgi. There is a precedent for these proteins moving from the plasma membrane to other cellular compartments. Both the wtEGFR and de2-7EGFR can traffic to the nucleus (Carpenter and Liao, 2009; de la Iglesia et al., 2008). In the case of the wtEGFR this involves retrograde transport of the receptor from the cell surface by endosomes to the ER. In the ER, the wtEGFR is extracted from the lipid bilayer by interacting with Sec61. Further interaction with chaperones, such as heat shock protein 70 (HSP70), promote the extrusion of wtEGFR into the cytoplasm, where it associates with importin-β, which leads to nuclear import (Carpenter and Liao, 2009). Given that mitochondrial de2-7EGFR is fully glycosylated, we propose that mitochondrial localisation of both the de2-7 and wtEGFR is mediated by a similar process, except once in the cytoplasm, it is directed to the mitochondria. Our results suggest that phosphorylation of the Y845 site in the de2-7EGFR is crucial in the decision to direct it to the mitochondria. Direct confirmation that the sec61 pathway is responsible for de2-7EGFR transport to the mitochondria is the subject of ongoing studies.

De2-7EGFR signalling in glioma cells protects them from apoptosis by increasing expression of Bcl-X_L and therefore does not require the physical presence of the receptor in the mitochondria (Nagane et al., 1998). The wtEGFR was reported to interact with CoxII in the mitochondria by binding to Y845 (Boerner et al., 2004); however, we could not detect a direct interaction of de2-7EGFR with COXII, even though it is constitutively activated at Y845 (data not shown). Recently, it has been reported that de2-7EGFR enhances lipogenesis in U87MG glioma cells (Guo et al., 2009), firmly establishing a role for this receptor in cell metabolism. The above report, and our observations that de2-7EGFR shows increased mitochondrial localisation under low-glucose conditions, indicates that it has a role in modulating cell metabolism. Indeed, the ability of de2-7EGFR to shift U87MG cells towards mitochondrial oxidative metabolism at the expense of aerobic glycolysis, strongly supports a function for this receptor in cell metabolism. Furthermore, this de2-7EGFR-mediated change in cell metabolism should make U87MG.Δ2-7 cells less dependant on glucose; an outcome that is entirely consistent with the enhanced cell survival of U87MG.Δ2-7 cells grown in low-glucose conditions. Given the convergence between metabolic and apoptotic pathways (Bonnet et al., 2007), a central role for de2-7EGFR in metabolic processes related to apoptosis is an attractive hypothesis that we are exploring.

De2-7EGFR is resistant to several EGFR-targeted therapeutics (Fukai et al., 2008; Heimberger et al., 2002; Montgomery, 2002) and other therapeutics show only modest activity (Johns et al., 2007). Our observation that there are two intracellular pools of activated de2-7EGFR (Golgi and mitochondria), provides a possible mechanism for this resistance. Clearly EGFR-specific antibodies would not effectively target these intracellular pools of de2-7EGFR. Given that some of the de2-7EGFR appears to be found in the internal mitochondrial compartments, it is also likely that some small molecules might not effectively inhibit this pool of receptor. Designing EGFR inhibitors that can penetrate the mitochondria membranes might have greater efficacy against this receptor. We have shown that the anti-tumour activity of a de2-7EGFR-specific antibody (mAb806) against U87MG.Δ2-7 glioma xenografts is significantly enhanced if Src activity is inhibited genetically (Johns et al., 2007) or pharmacologically, using dasatinib (Lu et al., 2009). This was achieved using doses of dasatinib that had no therapeutic effect as a sole agent. Our current studies showing that dasatinib prevents the translocation of de2-7EGFR provides an explanation for this increased anti-tumour activity; when there is no intracellular pool of de2-7EGFR, mAb806 can effectively block signalling at the cell surface.

The observation that most cancer cells rely on aerobic glycosylation rather than oxidative phosphorylation (the Warburg effect) (Warburg, 1956a) has long demonstrated an association between the mitochondria and cancer beyond its obvious role in apoptosis. The metabolic requirements and associated role of mitochondria in cancer cells is a topic of increasing interest (Vander Heiden et al., 2009), with one recent paper showing that the pharmacological reversal of the Warburg effect induces cell death in a range of cancer cell lines (Bonnet et al., 2007). As noted above, there is growing evidence that ErbB family members have a role in cell metabolism and mitochondria function. Our work showing significant targeting of the de2-7EGFR to the mitochondria strongly implicates a more direct role for some members of this family in mitochondrial activity.

mAb806, which preferentially recognises de2-7EGFR in U87MG glioblastoma cells and the irrelevant isotype-matched IgG_2b antibody (clone 100-310), have been described in detail previously (Johns et al., 2002; Luwor et al., 2001). Both antibodies were produced in the Biological Development Facility (BDF; Ludwig Institute for Cancer Research Ltd, Melbourne). Serum from rabbits immunised with the mitochondrial protein Chaperone-60 (Cpn60), was used as a mitochondrial marker. Antibodies raised against the following antigens were used: giantin (ab24586), pan cadherin (ab6529) (Abcam, Cambridge, MA); calnexin (C-terminal 575–594); v-Src 327 (Ab-1) (Calbiochem, Gibbstown, NJ); Golgi matrix protein 130 (GM-130), lysosome-associated membrane protein 1 (LAMP-1) (BD Transduction Laboratories, Franklin Lakes, NJ); MitoTracker Red (MTR), secondary antibodies: mouse Alexa Fluor 488 and rabbit Alexa Fluor 594 IgG (Molecular Probes, Grand Island, NY); EGFR-pY845, Src-pY418 and EGFR-pY1173 (Biosource International, Cell Signaling Laboratories, Danvers, MA); HRP-conjugated secondary antibody raised against goat IgG (1:10,000), pan phospho-EGFR, pY20 (Santa Cruz Biotechnology, La Jolla, CA); gold-conjugated secondary antibody raised against mouse IgG (1:500) (Sigma, St Louis, MO). HRP-conjugated secondary antibodies raised against mouse (1:10,000) and rabbit (1:2000) IgG were purchased from Chemicon Australia (Millipore, North Ryde, NSW, Australia).

The human derived glioblastoma cell lines U87MG, U87MG.Δ2-7 and U87MG.Δ2-7_DNSrc have been described previously (Johns et al., 2007; Nagane et al., 1996; Nishikawa et al., 1994) All cell lines were routinely maintained in DMEM/F-12 (Life Technologies) supplemented with 5% FCS (CSL), 2 mM Glutamax (Sigma) and 2 mM penicillin-streptomycin (Life Technologies). In addition, the transfected U87MG.Δ2-7 cell line was maintained in, 400 μg/ml Geneticin (Life Technologies), whereas the U87MG.Δ2-7_DNSrc cell line was maintained in both Geneticin and 100 μg/ml hygromycin (Roche Diagnostics, Mannheim, Germany).

A mutant form of the de2-7EGFR containing a Y845F substitution was obtained by site-directed mutagenesis. The mutant receptor (Δ2-7_845F) was subsequently introduced into U87MG parental cells using previously described methods (Nishikawa et al., 1994). Clones were screened initially by FACS analysis to confirm de2-7EGFR expression. Cells which stained positively for de2-7EGFR were sorted by FACS for high expression and expanded. This population of cells was tested by western blotting for de2-7EGFR expression (mAb806) and loss of phosphorylated Y845 using a phosphorylation-specific antibody against the site.

A PmeI fragment containing activated Src cDNA (Y529F mutation) was subcloned into the pcDNA3.1/Hygro⁺ vector (Life Technologies) before transfection of the U87MG.Δ2-7 and U87MG.Δ2-7_845F cell lines. Transfected cells were screened by FACS to confirm that expression of de2-7EGFR was retained. Clones were then subjected to further analysis by intracellular FACS and western blot using the Src-specific antibody, v-Src 327. Finally, clones showing increased Src expression were western blotted with the phosphorylated-Src-specific antibody, Y418 (Biosource International) to ensure the Src was active. All cell lines were tested by FACS to ensure levels of de2-7EGFR and Src were equal in the different cell lines.

U87MG and U87MG.Δ2-7 cells were plated at 4×10⁴ cells per well in six-well tissue culture plates in 10% fetal bovine serum. On the next day, cells were washed in phosphate-buffered saline, and medium was changed to one containing 10% FBS, glutamine, pyruvate, and glucose at doses of 25, 5, 1, 0.5 and 0 mM. Floating cells were collected and combined with trypsinised plated cells and viable and non-viable cells counted using a hemocytometer on each day.

A crude ER sample was isolated from 10⁷ exponentially growing U87MG.Δ2-7 cells, using the Endoplasmic Reticulum Isolation Kit (Sigma), as per the manufacturer's instructions. The final ER pellet obtained using this kit was further purified using discontinuous sucrose gradients. Briefly, the ER pellet was resuspended in Buffer A (20 mM HEPES, pH 7.6, 220 mM mannitol, 70 mM sucrose, 1 mM EDTA and 0.5 mM PMSF), before loading the ER sample on top of a discontinuous sucrose density gradient consisting of 15%, 30% and 60% sucrose solutions. The gradient was subjected to ultracentrifugation at 24,000 r.p.m. at 4°C for 18 hours, using a Beckman Ultracentrifuge fitted with a SW28 rotor. 0.5–1.0 ml fractions were collected from the top of the centrifuge and subsequently tested against a panel of organelle marker antibodies via western blot. Fractions that tested positive against the ER marker antibody calnexin, but negative against all other organelle marker antibodies, were pooled.

6×10⁶ exponentially growing cells were harvested, washed twice in ice-cold PBS, and ‘crude’ mitochondria isolated, in accordance with previously described methods (Johnston et al., 2002; McKenzie et al., 2006). The mitochondrial pellet was stripped of contaminating organelles using a high-salt wash buffer (250 mM sucrose, 1 mM EDTA, 4 M NaCl, 10 mM Tris-HCl, pH 7.4) (Crowley and Payne, 1998; Walter and Blobel, 1983). Following ten high-salt wash steps, the mitochondria were re-isolated by centrifugation at 8000 g at 4°C for 15 minutes.

The locality of mitochondria-associated de2-7EGFR was examined biochemically as detailed elsewhere (Gough et al., 2009; Vande Velde et al., 2008), with some modifications. Briefly, samples were halved before proteinase K (PK) digestion (+PK) or a ‘sham digest’ in Buffer B (500 mM Sucrose, 10 mM HEPES, pH 7.4) (−PK) at 37°C for 15 minutes. Digestion reactions were stopped by incubating mitochondria with PMSF at 4°C for 5 minutes. Following two wash steps in Buffer B, mitochondria were re-isolated at 8000 g, 4°C for 15 minutes and resolved by SDS-PAGE in 2× sample reducing buffer [2.0 ml of Glycerol, 4% (w/v) SDS, 0.005% (w/v) Bromophenol Blue, 0.5 ml of β-mercaptoethanol, 125 mM Tris-HCl, pH 6.8, and ddH₂O].

Whole-cell lysates were prepared by lysis in RIPA buffer (50 mM Tris-HCl, pH 7.5, 150 mM NaCl, 5 mM EDTA, 200 mM Na₃VO₄, 0.5% sodium deoxycholate, 0.05% SDS, 10 mM NaF and protease inhibitor cocktail set 1 from Calbiochem), then sonicated for 15 minutes and centrifuged at 14,000 g for 30 minutes at 4°C. The supernatant (lysate) was collected and added to equal volumes of 2× sample reducing buffer. Mitochondrial pellets were resuspended in RIPA lysis buffer and sonicated for 15 minutes. Following sonication, equal volumes of 2× sample reducing buffer were added to the sample. All samples were heated at 95°C for 5 minutes, directly before electrophoresis.

Cell lysates were immunoprecipitated with relevant antibodies using methods described by us in detail previously (Gan et al., 2007; Johns et al., 2004). Samples were resolved on NuPAGE gels (either 3–8% or 4–12%) as previously described (Gan et al., 2007), followed by the electro-transfer of proteins onto PVDF membranes using the iBlot Gel Transfer System (Life Technologies) as per the manufacturer's instructions. Membranes were blotted with antibodies as detailed in the relevant figure legends, before detection of bands by chemiluminescence radiography using the Storm 804 Phosphorimager (Amersham Biosciences, Piscataway, NJ). Analysis of western blots were performed using ImageQuant TL Image Analysis Software, version 2005 (Gan et al., 2007; Hagar et al., 2003).

The extracellular matrix gel, matrigel (Sigma) was prepared in serum-free DMEM/F-12 medium (1:50), before application of a thin layer (100 μl) to wells holding sterile 18 mm glass coverslips. Following a 1 hour incubation step at room temperature, excess matrigel was removed and coverslips were subsequently washed twice with serum-free DMEM/F-12 medium. U87MG.Δ2-7 cells (2×10⁵ per well) were grown on matrigel-coated coverslips, under normal culture conditions for 24 hours.

Cells grown overnight on 18 mm glass coverslips, were washed twice with PBS, fixed with 4% paraformaldehyde in PBS for 10 minutes, permeabilised with 0.1% Triton X-100 for 10 minutes at room temperature, and subsequently blocked with Image-iT™ FX signal enhancer (Molecular Probes) as per the manufacturer's instructions. The cells were then immunostained for 1 hour at room temperature with primary antibodies, washed twice with 0.1% Triton X-100 in PBS and counterstained with Alexa-Fluor-488- and/or 596-conjugated secondary antibodies (Molecular Probes, 1:2000), for 20 minutes at room temperature. Once washed, coverslips were mounted onto superfrost glass slides over a droplet of mounting media (0.23 g of 1,4-diazabicyclo[2.2.2]octane; Sigma) in 200 μl of ddH₂O and subsequently examined. The mitochondria of viable cells were stained with 500 nM MTR (Molecular Probes) in serum-free medium at 37°C for 30 minutes, before fixing, permeabilisation and subsequent immunostaining of cells.

Cells were examined and images acquired using a TCS SP2 confocal laser-scanning microscope (Leica). In all experiments, multiple z-section scans (14–16) at 0.15 μm increments spanning the entire depth of the cell were acquired under the same confocal microscope settings (i.e. sequential scans, with wavelengths set as follows: green, 500–535; red, 570–650), using a planApo 100×/1.40 oil immersion, objective lens (Leica). The only exception was where detector gain adjustments were performed to normalise saturation levels in the green channel.

Colocalisation was determined following the voxel analysis of acquired images using a colocalisation macro operated under ImageJ v1.38d software (National Institutes of Health). This plug-in considers two voxels (from the red A, mitochondria and green B, de2-7EGFR channels) as colocalised if their respective intensities are higher than the threshold of their respective channels (default setting: 50 grey levels), and their ratio of intensity is higher than the default ratio setting value of 50% (Gotliv et al., 2006). Following analysis, the program creates a new image where colocalised voxels are highlighted in white. In all cases, computational analysis began with a region of interest (ROI) drawn either around the circumference of the cell in order to determine the percentage of total de2-7EGFR (channel B) expressed by the cell colocalised with mitochondria (channel A), or drawn around the mitochondria in order to determine the percentage of mitochondria within the cell colocalised with de2-7EGFR. Calculations were performed using ImageJ (NIH) and an established formula ([ΣBi_{{i ε Coloc}}/ΣAi_{{i ε Mask}}]3100), detailed elsewhere (Abramoff et al., 2004).

Mitochondria from human glioblastoma cell lines were isolated and purified as detailed above. Once purified, the mitochondrial samples were incubated with 1 μg/ml mAb806, in 1% HSA in PBS (v/v), on ice for 2 hours. Samples were subsequently washed with ice-cold PBS to remove unbound antibody and counterstained on ice for 2 hours with a gold-conjugated secondary antibody (Sigma) at a 1:500 dilution in Buffer C (0.5 M NaCl, 0.1% BSA, 0.05% Tween20, 5% fetal bovine serum, buffered to pH 7.4). The mitochondrial samples were washed again to remove unbound gold, fixed at room temperature for 10 minutes in 2.4% paraformaldehyde in PBS, washed again and subsequently resuspended in a 1:1 solution consisting of mitochondria storage buffer (Johnston et al., 2002) and agarose. Thereafter, labelled mitochondria were processed and embedded in resin, sectioned and photographed using a JEOL JEM-1200 EX transmission electron microscope.

U87MG and U87MG.Δ2-7 cells were maintained in DMEM containing 25 mM glucose, 4 mM glutamine and 1 mM pyruvate (Invitrogen, Carlsbad, CA). Oligomycin, carbonyl cyanide p-trifluoromethoxyphenylhydrazone (FCCP) and antimycin were obtained from Sigma, and stock solutions were prepared following the manufacturer's instruction. An extracellular flux (XF) analyzer (Seahorse Bioscience, North Billerica, MA) was used to determine metabolic phenotype of U87MG and U87MG.Δ2-7 cells. XF Analyzer simultaneously monitors oxygen consumption rate (OCR) and extracellular acidification rate (ECAR), which are indicators of mitochondrial respiration and aerobic glycolysis (lactate production), respectively, of cultured cells (Wu et al., 2007). The OCR and ECAR measurements were carried out as described previously (Wu et al., 2007). Briefly, cells were plated at densities of 20,000 cell per well in XF96 cell culture plates and 40,000/well in XF24 cell culture plates, and incubated for 24 hours in 37°C incubators at 10% CO₂ and 100% humidity. Before measurements, the growth medium was replaced with 600 μl assay medium (Seahorse Bioscience), a low buffered DMEM containing no bicarbonate, and incubated for 45 minutes in a 37°C non-CO₂ incubator. The concentration of glucose, glutamine and pyruvate concentrations in the assay medium are as indicated in the figure legends. Basal oxygen consumption rate and extracellular acidification rate was determined using XF96 Extracellular Flux analyzer. Mitochondrial function of the cells was assessed using XF24 analyzer. After the measurements, cells were removed from each well with trypsin-EDAT and the number of viable cells was determined using ViCell (Beckman Coulter, CA). The number of cells in each well was used to normalise OCR and ECAR was obtained.

This grant was funded in part by the National Health & Medical Council of Australia (Project Grant 433615 and 1012020) and the James S. McDonnell Foundation (#220020173).