|Home | About | Journals | Submit | Contact Us | Français|
Normal epithelial cells require matrix attachment for survival and the ability of tumour cells to survive outside their natural extracellular matrix (ECM) niches is dependent on acquisition of anchorage independence1. While apoptosis is the most rapid mechanism for eliminating cells lacking appropriate ECM attachment2, recent reports suggest that non-apoptotic death processes prevent survival when apoptosis is inhibited in matrix-deprived cells3,4. Here we demonstrate that detachment of mammary epithelial cells from ECM causes an ATP deficiency due to loss of glucose transport. Overexpression of ErbB2 rescues the ATP deficiency by restoring glucose uptake through stabilization of EGFR and PI(3)K activation and this rescue is dependent on glucose-stimulated flux through the antioxidant-generating pentose phosphate pathway (PPP). Interestingly, we found that the ATP deficiency could be rescued by antioxidant treatment without rescue of glucose uptake. This rescue was found to be dependent on stimulation of fatty acid oxidation (FAO), which is inhibited by detachment-induced reactive oxygen species (ROS). The significance of these findings was supported by evidence of an elevation in ROS in matrix-deprived cells in the luminal space of mammary acini and that antioxidants facilitate the survival of these cells and enhance anchorage-independent colony formation. These results reveal both the importance of matrix attachment in regulating metabolic activity and an unanticipated mechanism for cell survival in altered matrix environments through antioxidant restoration of ATP generation.
Epithelial cells are dependent on interactions with specific extracellular matrix (ECM) components for survival, proliferation, and differentiation functions5. Loss of matrix attachment of cultured epithelial cells activates a caspase-mediated apoptotic program known as anoikis2. In glandular cancers, like breast cancer, tumour cells are displaced from their normal matrix niches in the early stages of tumourigenesis when they proliferate into the lumen of hollow glandular structures. Filling of the luminal space is one of the hallmarks of early tumourigenesis.
Studies of luminal filling in both three-dimensional (3D) structures of MCF-10A mammary epithelial cells and the developing mammary gland have demonstrated that apoptosis is involved in clearance of centrally localised cells that lack matrix attachment; however, inhibition of apoptosis is not sufficient for survival of matrix-deprived cells in the luminal space3,4,6,7. Interestingly, several oncogenes, including ErbB2, have been shown to rescue cells from anoikis and prevent clearance of luminal cells in 3D acinar structures8, suggesting that these oncogenes prevent luminal clearance programs in addition to anoikis.
Another striking feature of both matrix-detached MCF-10A cells and centrally located, matrix-deprived cells in MCF-10A acini is the induction of autophagy9,10. As autophagy is a catabolic process commonly upregulated under conditions of starvation11, the association of this process with matrix deprivation suggests that ECM attachment regulates metabolic activity as well as apoptosis. Here we elucidate the basis for the metabolic defects in matrix-detached cells, demonstrate that oncogenes can rescue these defects through restoration of glucose uptake and enhancement of antioxidant capacity, and unexpectedly find that antioxidants alone can rescue matrix-detached cells via restoration of ATP generation through FAO. Lastly, we demonstrate that antioxidants promote anchorage-independent survival in two in vitro models of tumourigenesis.
To investigate whether ECM regulates cellular metabolism, we examined cellular ATP levels in MCF-10A cells cultured on adherent or non-adherent plates. We detected a substantial reduction in ATP in both MCF-10A (Fig. 1a) and primary human mammary epithelial cells (HMEC, Supplementary Fig. 2) that had been detached from the ECM for 24 hours. We confirmed these results in lysates normalized for total protein (Fig. 1b) and by measuring the ATP/ADP ratio (Fig. 1c). The reduction of ATP in MCF-10A cells occurred between 12 and 24 hours after ECM detachment (Supplementary Fig. 7), was not affected by inhibition of apoptosis (Fig. 1a and Supplementary Fig. 4,5) or autophagy (Supplementary Fig. 6), and was rescued by addition of reconstituted basement membrane (Supplementary Fig. 3). Expression of ErbB2 in these cells (ErbB2-MCF-10A) prevented the reduction in ATP following matrix detachment (Fig. 1a,c), suggesting that ErbB2 circumvents the matrix requirement for ATP production.
Given the evidence that EGFR is downregulated in detached cells and that its overexpression can rescue anoikis12, we investigated the effect of ErbB2 on EGFR expression in detached cells. Indeed, ErbB2 expression caused a striking stabilization of EGFR following ECM detachment (Fig. 1d) that was critical for the rescue of ATP (Fig. 1e). This stabilization correlated with the maintenance of ERK activation and an enhancement of PI(3)K/Akt signaling (Fig. 1d); however, inhibition of PI(3)K (but not ERK) abrogated the ErbB2 rescue of ATP (Fig. 1f,g). Furthermore, expression of a constitutively active variant of PI(3)K (PIK3CA E545K) or Akt (Myr-Akt1) is sufficient to rescue detachment-induced ATP (Fig. 1h). These results indicate that ErbB2 rescues the metabolic defect in matrix detached cells by preventing the downregulation of EGFR andthus maintaining the activation of the PI(3)K pathway.
Given the critical role of PI(3)K/Akt in stimulating glucose transport13,14, we investigated whether glucose uptake is altered in matrix-detached cells. Indeed, we found a striking deficiency in glucose uptake in detached cells that was rescued (in a PI(3)K dependent fashion, Supplementary Fig. 8) by ErbB2 expression (Fig. 2a). In addition, treatment of detached cells with methyl pyruvate (MP), which provides substrates for the tricarboxylic acid (TCA) cycle, causes a substantial increase in ATP (Fig. 2b) that is blocked by the mitochondrial Complex 1 inhibitor rotenone (Supplementary Fig. 9). This suggests that mitochondria of detached cells retain the capacity to produce ATP. To determine if the rescue of glucose uptake by ErbB2 is important for its ability to elevate ATP levels in detached cells, we treated ErbB2-MCF-10A cells with 2-deoxyglucose (2DG), a glucose analogue that inhibits glycolysis. 2DG treatment completely abolished the rescue of ATP by ErbB2 (Fig. 2c), confirming the importance of glucose uptake in the rescue of ATP by ErbB2.
Following cellular uptake, glucose can be further metabolised to generate ATP by glycolysis/oxidative phosphorylation, or it can be driven down the pentose phosphate pathway (PPP) by glucose-6-phosphate dehydrogenase (G6PDH)15. Since the PPP is a major source of cellular NADPH (which provides reducing equivalents), we examined the effects of matrix detachment on ROS production. ECM detachment induced a significant increase in ROS (Fig. 2d) and decrease in reduced glutathione (Fig. 2e), both of which were reversed by ErbB2 (Fig. 2d,f). In support of the possibility that the lack of PPP flux is responsible for the increase in ROS, we found that the reduction in glucose uptake precedes the elevation in ROS levels (Supplementary Fig. 10).
Previous studies have shown that reducing ROS through the stimulation of PPP flux can promote cell survival16,17; thus we hypothesised that PPP flux may be important for ErbB2 to rescue ATP levels in detached cells. Indeed, the treatment of detached ErbB2-MCF-10A cells with the PPP inhibitors dehydroisoandrosterone (DHEA) or 6-aminonicotinamide (6-AN) abrogated the ability of ErbB2 to rescue ATP (Fig. 2g) and led to an increase in cellular ROS (Supplementary Fig. 11). Interestingly, we found that matrix-detachment caused a significant increase in G6PDH protein (Fig. 2h), and siRNA-mediated reduction of G6PDH in detached ErbB2-MCF-10A cells caused a substantial decrease in ATP levels (Fig. 2i) and increase in ROS (Fig. 2j). The evidence that loss of matrix attachment causes an increase is ROS is consistent with studies in endothelial cells18 and in the developing mammary gland4.
The induction of ROS in suspended cells led us to investigate the effects of ROS neutralization on ATP in matrix-detached cells. Treatment of detached MCF-10A cells with the antioxidants N-acetyl-L-cysteine (NAC) or Trolox (a water-soluble Vitamin E derivative) significantly elevated ATP in detached cells (Fig. 3a) independent of any changes in glucose uptake (Fig. 3b). These data suggest that ROS inhibits ATP production through a metabolic pathway that could otherwise compensate for loss of glucose uptake.
It has previously been demonstrated that cancer cells deprived of glucose maintain ATP production through FAO19. Given that detached cells are glucose-deprived and antioxidants can promote ATP generation, we hypothesised that detachment-induced ROS could inhibit FAO. Indeed, we found that FAO was markedly reduced in detached MCF-10A cells (Fig. 3c) and Trolox treatment substantially elevated FAO (Fig. 3d). In addition, treatment of antioxidant-supplemented cells with etomoxir, an inhibitor of fatty acid transport into the mitochondria19, prevented the rescue of ATP by antioxidants in detached cells (Fig. 3e). Furthermore, methyl malate treatment (which generates NADPH through an alternative pathway) elevated ATP levels, lowered ROS levels, and caused an increase in FAO in detached MCF-10A cells (Supplementary Fig. 12a–c) suggesting that NADPH production is sufficient to elevate FAO. Interestingly, ErbB2 reduced FAO in attached and suspended cells (Fig. 3c), consistent with previous reports showing PI(3)K/Akt mediated reduction of FAO19,20. However, treatment of detached ErbB2-MCF-10A cells with etomoxir caused a dose-dependent reduction in the ATP levels (Fig. 3f), suggesting that the residual FAO significantly contributes to ATP production in these cells.
To extend our analysis to a model with more physiologic relevance, we utilised the MCF-10A 3D cell culture model in which the inner, matrix-deprived cells show evidence of metabolic impairment (based on upregulation of autophagy9,10). While it is not feasible to measure ATP directly in these structures, we used two-photon microscopy to examine the native fluorescence of NADH and NADPH [NAD(P)H] and obtain an assessment of metabolic differences between the inner and outer cells21. Since NADH is the principal electron donor in glycolytic and oxidative metabolism, the native fluorescence of NAD(P)H represents a non-invasive fluorescent reporter of the metabolic state22. Images of acinar structures on Day 8 revealed a dichotomy in NAD(P)H fluorescence intensity between the inner and outer cells of the majority of structures, whereas most Day 4 structures showed homogenous fluorescence (Fig. 4a, Supplementary Fig. 13). Furthermore, we found that ROS were detectable exclusively in the centrally localized cells of Day 7 acini (Fig. 4c). Together, there results provide evidence that there are differences in metabolic activity and ROS accumulation in the matrix-deprived centrally localized cells of acini, as were observed in monolayer and suspension cultures (Figures 1–3). In addition, Trolox treatment prevented the dichotomy in NAD(P)H fluorescence in 3D culture (Fig. 4b, Supplementary Fig. 14), suggesting that ROS significantly contributes to the metabolic dichotomy between the inner and outer cells.
To understand the implications of these findings with regards to the clearance of cells from the luminal space, we studied the effects of antioxidant treatment on the survival of cells located in the center of acini. We treated acini with either NAC or Trolox and monitored luminal filling over time. Interestingly, we found that NAC or Trolox treatment significantly reduced luminal clearance (Fig. 4d) independent of effects on caspase activation (Supplementary Fig. 15), suggesting that the elimination of ROS can preserve the viability of matrix-deprived cells in the center of acini. By extrapolation to our studies showing ATP reduction and ROS generation in matrix-deprived cells, we speculate that the viability is a consequence of rescuing these metabolic impairments. In addition, the late clearing of acini observed in Bcl-2-MCF-10A cells is suppressed by Trolox. (Fig. 4d). These results suggest that suppression of both metabolic impairments and apoptosis are required for anchorage-independent survival.
The evidence that antioxidants can rescue metabolic defects of ECM-deprived cells and promote luminal filling in MCF-10A acini raises the question whether antioxidants could promote the transforming activity of mammary epithelial cells. To examine this, we assayed anchorage-independent colony formation in soft agar in MCF-10A cells expressing oncogenes that promote hyperproliferation (human papilloma virus E7 protein) and suppress apoptosis (Bcl-2). Neither E7 nor Bcl-2 expression rescued ATP in matrix detached cells (data not shown) and the E7/Bcl2 cells exhibited a relatively weak capacity to form colonies in this assay (Fig. 4e). Trolox treatment induced a substantial increase in the number and average size of colonies, suggesting that antioxidant treatment can enhance the transforming activity of cells that harbour oncogenic insults. We observed similar results with ErbB2-MCF-10A (Supplementary Fig. 16) and BT-474 cells, an ErbB2 expressing breast cancer cell line (Supplementary Fig. 17).
In summary, our study highlights the possibility that glucose deprivation could lead to ROS production during tumourigenesis and force selection for alterations that allow escape from oxidative damage. Furthermore, we demonstrate that antioxidants rescue cells from the need for glycolysis through stimulation of FAO. Thompson and coworkers19 have previously shown that stimulation of FAO through AMPK/p53 is another mechanism for the survival of cells in the absence of glucose. Thus, our study suggests that deprivation of matrix may limit glucose accessibility during tumourigenesis, and reveals strategies whereby a tumour could escape from metabolic stress due to these conditions by stimulating anchorage-independent glucose transport and/or eliminating ROS.
In addition, these data demonstrate that antioxidants promote the survival of cells that lack attachment to the ECM and raise the question whether antioxidants may have dichotomous activities with respect to tumourigenesis -- that is, suppressing tumourigenesis by preventing oxidative damage to DNA23,24, and promoting tumourigenesis by allowing survival of cells that are metabolically impaired (e.g. in altered matrix environments). In support of this, expression of SOD2, a mitochondrial protein that reduces oxidative stress caused by respiratory chain leak, is elevated in more advanced and higher grade mammary tumours25,26. Furthermore, a recent study has revealed that enhanced PPP flux and increased antioxidant capacity correlates with metastasis of breast cancer cells to the brain27. Lastly, randomised trials have demonstrated both anti-neoplastic and neoplastic effects of antioxidants, with neoplastic effects associated with patients at higher risk due to smoking and alcohol consumption28, or patients undergoing chemo- or radiation therapy29. Our work provides a biological rationale for these findings, as antioxidant activity may promote the survival of pre-initiated tumour cells in unnatural matrix environments and thus enhance malignancy.
MCF-10A cells and their variants were all cultured as described on http://brugge.med.harvard.edu/. Other cell lines were cultured as described in the supplementary information. All assays on ECM detached cells were carried out 24 hours after plating on poly-HEMA coated plates unless otherwise noted. ATP assays were conducted using either the ATPlite assay (PerkinElmer, Waltham, MA), the ATP determination kit (Invitrogen, Carlsbad, CA), or the ATP/ADP Ratio Assay Kit (BioAssay Systems, Hayward, CA). Glucose uptake assays were performed using the Amplex Red Glucose Assay Kit (Invitrogen). ROS was measured using carboxy-H2DCF-DA in detached/attached cells and in 3D culture. To measure reduced glutathione, we used chloromethylcoumarin (CMAC, Invitrogen). FAO was measured by monitoring the release of 14CO2 after addition of 1-14C-Oleic Acid. 3D culture of mammary acini was completed according to the protocol at http://brugge.med.harvard.edu/. Native fluorescence of NAD(P)H was measured using two-photon microscopy. Soft agar assays were performed in the presence or absence of antioxidants and colony formation/size was determined using ImageJ.
Full methods and any associated references are available in the online version of the paper at www.nature.com/nature.
We thank Michael Overholtzer, Ghassan Mouneimne, Marco Mazzone, and Cheuk Leung for critical reading of the manuscript and/or experimental assistance. We thank Arnaud Mailleux, Alice Kaanta, Veronica Schafer, Alicia Zhou, Kaylene Simpson, and the members of the Brugge lab for experimental assistance, helpful comments, and/or discussion. This work was supported by a grant from the National Cancer Institute (J.S.B.) and a grant from the NIH (P.P.). Z.T.S. is the recipient of a Ruth L. Kirschstein National Research Service Award (NRSA) for Postdoctoral Fellows from the National Cancer Institute; L.S. an NCI Mentored Quantitative Research Development Award (K25); A.R.G. a National Science Graduate Research Fellowship; and H.Y.I. an NCI K08 Award.
Supplementary information is linked to the online version of the paper at www.nature.com/nature. A figure summarising the main result of this paper (Supplementary Fig. 1) is included in the supplementary information.
Author ContributionsZ.T.S. and J.S.B. were responsible for the overall study design. Z.T.S., A.R.G., H.Y.I., and S.G. conducted experiments. L.S. and Z.J. conducted the experiments measuring native fluorescence of NAD(P)H in 3D cell culture. Z.G.-H. and P.P. designed the fatty acid oxidation studies and Z.T.S. and Z.G.-H. conducted the fatty acid oxidation assays. Z.T.S. and J.S.B. drafted the manuscript and all other authors made revisions.
Reprints and permissions information is available at npg.nature.com/reprintsandpermissions.