|Home | About | Journals | Submit | Contact Us | Français|
A high frequency of somatic mutations has been found in breast cancers within the gene encoding the catalytic p110α subunit of PI3K, PIK3CA. Using isogenic human breast epithelial cells, we have previously demonstrated that oncogenic PIK3CA “hotspot” mutations predict for response to the toxic effects of lithium. However, other somatic genetic alterations occur within this pathway in breast cancers, and it is possible that these changes may also predict for lithium sensitivity. We overexpressed the epidermal growth factor receptor (EGFR) into the non-tumorigenic human breast epithelial cell line MCF-10A, and compared these cells to isogenic cell lines previously created via somatic cell gene targeting to model Pten loss, PIK3CA mutations, and the invariant AKT1 mutation, E17K. EGFR overexpressing clones were capable of cellular proliferation in the absence of EGF and were sensitive to lithium similar to the results previously seen with cells harboring PIK3CA mutations. In contrast, AKT1 E17K cells and PTEN−/− cells displayed resistance or partial sensitivity to lithium, respectively. Western blot analysis demonstrated that lithium sensitivity correlated with significant decreases in both PI3K and MAPK signaling that were observed only in EGFR overexpressing and mutant PIK3CA cell lines. These studies demonstrate that EGFR overexpression and PIK3CA mutations are predictors of response to lithium, whereas Pten loss and AKT1 E17K mutations do not predict for lithium sensitivity. Our findings may have important implications for the use of these genetic lesions in breast cancer patients as predictive markers of response to emerging PI3K pathway inhibitors.
The phosphatidylinositol 3-kinase (PI3K) pathway regulates many important cellular processes including angiogenesis, proliferation and apoptosis.1 The catalytic and regulatory subunits of the human PIK3CA gene were cloned over fifteen years ago2 and somatic mutations in the gene encoding for the PI3K p110α catalytic subunit, PIK3CA, have subsequently been identified in many cancers (reviewed in ref. 3). The frequency of PIK3CA mutations in human breast cancers ranges in studies from 8–40%4–7 with an average of 25%, an observation that supports the significance of PI3K in breast cancer biology and underscores its importance as a potential therapeutic target.
Additional somatic alterations are also found in key genes that lie within the PI3K pathway. For example, the phosphatase and tensin homolog (Pten) protein is a tumor suppressor that reverses the effects of PI3K by dephosphorylating the 3′ phosphate of the inositol ring in phosphatidylinositol-(3,4,5)-trisphosphate resulting in phosphatidylinositol-(4,5)-bisphosphate. Although rarely mutated in breast cancer, diminished levels of Pten expression through loss of heterozygosity and/or epigenetic silencing mechanisms are observed in up to 48% of tumors.8,9 Furthermore, aberrant Pten activity in breast cancers has been associated with metastasis and poor survival.8,10 Another critical member of the PI3K pathway is Akt. Akt family members are frequently activated in cancers via phosphorylation. Recently a single hotspot mutation, G49A:E17K, in the pleckstrin homology domain of AKT1 has been described, with the highest frequency of mutations found in human breast cancers.11 Subsequent studies have confirmed the relatively low but consistent frequency of this mutation ranging from 1.8–8%.12–15
Aberrant activation of the PI3K pathway in breast cancers also occurs through the human erbB receptor tyrosine kinase (RTK) family of transmembrane receptors which includes epidermal growth factor receptor (EGFR), HER2, HER3 and HER4. Although RTK activation leads to MAPK signaling via Ras, Raf, Mek and Erk, it is now known that RTK activation also results in signaling through the PI3K pathway via Ras/p110α binding as well as through the intermediate molecule IRS-1.16,17 EGFR overexpression has been reported in breast cancers and is associated with resistance to hormonal therapy and reduced disease free survival.18–20 In addition, approximately 40% of triple negative/basal type breast cancers are associated with overexpression of EGFR. Moreover, HER2 and concurrent EGFR expression is found in 21% of breast tumors.21 EGFR is therefore an attractive target in breast cancer, but to date clinical trials of single agent tyrosine kinase inhibitors have been disappointing.22 Successful translation for benefit will require a better understanding of the complex pathways involved with EGFR signaling leading to novel combinations of cytotoxic therapies.23
We and others have previously created physiologic in vitro models of aberrant oncogenic PI3K pathway signaling by employing somatic cell gene targeting in the human breast epithelial cell line, MCF-10A.24–27 MCF-10A cells are spontaneously immortalized cells and provide an ideal model for these experiments because they are human, mostly diploid, non-tumorigenic and are genetically stable by FISH,28 karyotype,29 copy number variation and microsatellite analyses (data not shown). The use of paired isogenic cell lines provides a unique opportunity to study the effects of oncogenic PIK3CA mutations on downstream signaling pathways with less concern about potential confounding genetic anomalies. Consequently, we determined that lithium, an FDA-approved therapy for bipolar disorder, has selective anti-neoplastic properties against human breast and colon cancer cell lines that harbor oncogenic PIK3CA mutations.25 However, other genetic alterations in the PI3K pathway such as Pten loss, AKT1 E17K mutations and EGFR overexpression occur in human breast cancers. Although previous work has suggested that mutations in genes within a common pathway are functionally equivalent and, therefore, rarely occur concurrently in human malignancies, this notion has been recently challenged.13 Indeed, our own previous studies have demonstrated dramatic phenotypic differences between PIK3CA and AKT1 mutations.15 It therefore remains an open question as to whether genetic lesions in the same pathway will be equivalent in their ability to predict for response to a given pathway inhibitor. Therefore, we sought to determine if these genetic alterations would also recapitulate equivalent lithium sensitivity by employing our isogenic cell based system. Using isogenic MCF-10A derived PTEN knock out cells (PTEN−/−),26 AKT1 E17K knock in cells (AKT1 mutant)15 and EGFR overexpressing cells, we determined that only EGFR overexpression exhibited a similar signal transduction pattern and sensitivity to lithium similar to mutant PIK3CA knock in cell lines. This work has potential implications for the development of predictive biomarkers of response to future targeted therapies.
Based upon our previous study in reference 25, oncogenic PIK3CA mutations appear to activate both MAPK and PI3K pathways in a manner akin to erbB receptor activation. This led us to hypothesize that EGFR overexpression, which is present in a significant fraction of triple negative breast cancers, may also result in lithium sensitivity. To effectively model this, we chose to overexpress a human EGFR cDNA in the non-tumorigenic breast epithelial cell line, MCF-10A. Using a retroviral infection strategy, we were able to isolate three EGFR-overexpressing clones, which were named EGFR7, EGFR8 and EGFR9. Furthermore, we also isolated a G418-resistant, non-EGFR overexpressing control clone transduced with an empty vector (EV). Overexpression of EGFR protein was confirmed by western blot using antibodies against total EGFR protein (Fig. 1).
EGFR overexpression is thought to lead to homodimerization of EGFR as well as heterodimerization with other erbB family members resulting in autophosphorylation and activation of these RTKs. Because MCF-10A cells require exogenous EGF for cellular proliferation, we tested the EGFR overexpressing clones for their ability to proliferate under EGF free conditions as well as physiologic concentrations of EGF (0.2 ng/ml) as previously described in reference 25. Figure 2 displays the proliferation of parental MCF-10A cells, our three independently-derived EGFR overexpressing clones, and an empty vector control in the absence and presence of 0.2 ng/mL EGF. All three of our EGFR overexpressing clones were capable of statistically-significant, EGF-independent proliferation, compared to control cells as measured after 72 hours by Student's t-test (p < 0.05). In the presence of EGF, the increased rate of proliferation was also statistically significant (p < 0.05). The empty vector clone was not capable of EGF-independent growth and it's growth in the presence of EGF was not statistically different from parental MCF10A cells (p > 0.05). These EGF independent growth properties were similar to what we have previously described for PIK3CA knock in mutant and PTEN knock out MCF-10A derived cell lines.25,26 Although theoretically it is possible that EGF independent growth was the result of increased EGF ligands secreted by these genetically modified cell lines, western blot analysis and co-culture using transwell assays did not support this hypothesis (Sup. Fig. 1).
Although PIK3CA mutations, Pten loss and AKT1 E17K mutations were originally thought to be functionally equivalent based upon their mutual exclusivity in human cancers,7,11 recent studies have now demonstrated differences in pathway activation among these genetic alterations as well as rare cancers that have lost Pten and concurrently harbor an activating PIK3CA mutation.13,30 In order to perform comparative analyses between PIK3CA knock in, PTEN knock out, AKT1 E17K knock in and EGFR overexpressing cells in the same MCF-10A background, we first performed Western blot analyses to determine the degree of MAPK and PI3K pathway activation by comparing relative levels of phosphorylated and total Akt and Erk in the absence of exogenous EGF and in the presence of physiologic concentrations of EGF (Fig. 3). The rationale for examining our cell lines in the absence and presence of EGF stems from our previous work demonstrating that EGF exposure at varying levels can lead to dramatic differences in biochemical signaling as well as response to drugs such as the mTOR inhibitor, rapamycin.25 It should be noted that blots were performed separately for no EGF versus 0.2 ng/ml EGF conditions, such that comparisons can only be made between cell lines within each EGF culture condition. Direct comparisons of signaling pathways between no EGF versus 0.2 ng/ml of EGF have been previously described for PIK3CA mutant, PTEN knock out and AKT1 E17K cell lines.15,25,26
Representative clones for PIK3CA exon 9 knock in, PTEN knock out and AKT1 E17K knock in were used as these cell lines have all previously been described to be indistinct from their clonal sibs.15,25,26 Because EGFR overexpressing clones have not yet been characterized, all three clones were used for these studies. Previously, we have described that AKT1 E17K knock in cells do not proliferate in the absence of EGF and concordantly, they display minimal phosphorylation of Akt and Erk.15 As seen in Figure 3, AKT1 E17K cells did not demonstrate any significant activation of the PI3K or MAPK pathways relative to parental MCF-10A cells, as shown by the minimal phosphorylation of Akt and Erk under EGF free and physiologic concentrations of EGF (0.2 ng/ml). However, in the presence of EGF, there was a slight but reproducible increase in ERK phosphorylation in AKT1 E17K cells relative to parental MCF-10A cells (Fig. 3 and right panel). The reason for this is unclear, but reaffirms the notion that signaling via EGF/EGFR can lead to unexpected and varying responses depending on the genetic alterations present within a given cell. In contrast, phosphorylated Akt was increased in the PTEN−/− cell line, but this was not as pronounced as in the three EGFR overexpressing clones or the PIK3CA knock in cell line in conditions without exogenous EGF (Fig. 3 and left panel), though was comparable to these cell lines in conditions with 0.2 ng/ml EGF (Fig. 3 and right panel).
EGFR overexpressing clones, PIK3CA knock in cells and PTEN−/− cells also demonstrated activation of the MAPK pathway as displayed by the increased levels of phosphorylated Erk relative to total Erk both in the absence and presence of 0.2 ng/mL EGF (Fig. 3). Interestingly, in the absence of EGF the PTEN−/− cell lines exhibited a pronounced increase in Erk phosphorylation compared to the EGFR overexpressing cell lines or the PIK3CA knock in cell line (Fig. 3 and left panel). However, in the presence of EGF, Erk phosphorylation in EGFR overexpressing clones was increased compared with parental and AKT1 E17K cells, but was decreased relative to both PTEN−/− cells and PIK3CA knock in cells (Fig. 3 and right panel). The cause for these differences are unknown, but these results are consistent with our previous observations in PTEN−/− cell lines showing that the presence or absence of EGF as well as duration of exposure to this growth factor can influence the level of Erk phosphorylation.26 Thus, our biochemical analyses reaffirm that the presence of an AKT1 E17K mutation alone does not confer significant oncogenic pathway signaling in human breast epithelial cells. In contrast, the presence of a PIK3CA oncogenic mutation, the loss of Pten, or overexpression of EGFR does indeed result in Akt and Erk phosphorylation in a manner similar to that found in breast cancer cells. However, the level and pattern of activation seen in these pathways is distinctly different between these three sets of cell lines, as evidenced by the varying levels of phosphorylation seen under conditions with and without exogenous EGF. This further underscores the previously unrecognized complexity of crosstalk that occurs between these important pathways.
Because the mechanism of action of lithium has not been fully elucidated, we wanted to assess whether EGFR overexpression, Pten loss or the AKT1 E17K mutation could also predict for lithium sensitivity. There were two main reasons for formally testing this hypothesis. First, because one of lithium's targets has been suggested to be GSK3β, which is modulated by Akt and Erk activation, we hypothesized that EGFR overexpression and Pten loss via gene targeting would predict for lithium sensitivity based upon our previous data that PTEN knock out leads to increased Akt and Erk activation26 and our own data presented in this study that EGFR overexpression also activates both the MAPK and PI3K pathways (Fig. 3). Second, despite the minimal pathway signaling seen with AKT1 E17K knock in cells,15 it was still formally possible that this mutation could activate other pathways that would lead to lithium sensitivity. There are in fact, examples where the same drug/compound can be extremely effective in various cancers with very different somatic alterations such as the case with BCR/ABL translocations in chronic myelogenous leukemia and certain C-KIT mutations in gastrointestinal stromal tumors both predicting for response to the small molecule inhibitor imatinib.31 In addition, Wang et al. recently demonstrated that MLL leukemias are also sensitive to lithium treatment.32
We previously described the selective anti-neoplastic properties of lithium in vitro and in vivo using human breast and colon cancer cell lines that harbor activating mutations in PIK3CA.25 A standard 10 mM concentration of lithium choloride (LiCl) was used in vitro based upon our initial tests25 and the doses previously reported in studies examining the effects of lithium in various in vitro systems and their correlation to in vivo models.32–36 It should be noted that although therapeutic lithium serum levels are 0.8–1.2 mEq/L, wide variations between serum lithium levels and intracellular concentrations of lithium have been reported.37
We hypothesized based on the biochemical results shown in Figure 3, that EGFR overexpression and PTEN loss, but not the AKT1 E17K mutation would also predict for sensitivity to lithium. We therefore performed multiple growth assays with and without EGF to compare the growth inhibitory effects of lithium in our panel of cell lines. As stated previously, the rationale for testing lithium toxicity under varying EGF conditions stems from our previous observations that differences in EGF concentration can have a profound effect on the downstream signaling cascades imparted by genetic mutations and their relative resistance and sensitivity to drugs such as rapamycin.25 In addition, parental MCF-10A, AKT1 E17K and empty vector control cells do not proliferate without EGF, but do proliferate with 0.2 ng/ml EGF and therefore can be used as isogenic counterparts when cultured with EGF. Importantly, we have previously described that parental MCF-10A cells and control cells proliferate at approximately an equal rate in 0.2 ng/ml EGF as mutant PIK3CA knock in cells in the absence of EGF, yet the former cell lines are resistant to lithium while the mutant PIK3CA clones were uniformly sensitive to lithium under these conditions. This strongly suggests that the effects of lithium are not simply due to increased cell proliferation.
Using identical conditions to our previous work, we found that treatment with LiCl significantly inhibited the growth of cells that overexpressed EGFR, similar to the response seen with the PIK3CA knock in cell line (Fig. 4). These effects were also observed at physiologic concentrations of EGF (Fig. 4A vs. B). Using a pair-wise comparison one-way ANOVA across cell lines, we found a statistically significant decrease in the proliferation of PTEN−/−, EGFR#7, EGFR#8, EGFR#9 and PIK3CAEx9 as compared with parental MCF-10A cells (p < 0.05). In contrast, the parental MCF-10A and AKT1 E17K cell lines were not significantly inhibited by LiCl when cultured in 0.2 ng/ml EGF (p > 0.05). As stated above, the effect of LiCl in parental MCF-10A and AKT1 E17K cells could not be ascertained in the absence of EGF as these cells do not proliferate under these conditions. Interestingly, the PTEN−/− cell line demonstrated intermediate sensitivity to LiCl, when compared to the response seen with PIK3CA knock in and EGFR overexpressing cell lines. Thus, although the biochemical pathways activated by mutant PIK3CA, Pten loss and EGFR overexpression appear similar, they do not uniformly predict for sensitivity to lithium treatment.
To uncover the potential reasons for the differential responses to lithium seen in our panel of cell lines, we performed Western blotting to elucidate any biochemical changes in the MAPK or PI3K pathways elicited by lithium exposure. Although we have previously demonstrated an increase in total GSK3β in PIK3CA knock in cells upon lithium treatment,25 this was not consistently observed in any of the EGFR overexpressing clones or PTEN−/− cells suggesting that increases in GSK3β are not the key mediator of lithium toxicity (data not shown). In addition, we also examined levels of phosphorylated p70S6Kinase, a marker of mTOR activation, but detectable levels were only present in PIK3CA mutant cells and EGFR overexpressing clones grown in 0.2 ng/ml EGF, with no appreciable change upon lithium exposure (Sup. Fig. 2). However, consistent differences in Akt and Erk phosphorylation were seen in lithium sensitive cell lines. For example, increased phosphorylation of Akt was seen at baseline in PTEN−/−, EGFR7, EGFR8, EGFR9 and PIK3CA knock in cells and this was significantly reduced in the presence of LiCl in the EGFR overexpressing and PIK3CA knock in cells (Fig. 5 and Sup. Fig. 3). The decrease in Akt phosphorylation was far less pronounced in the PTEN−/− cells and was not appreciable in AKT1 E17K cells consistent with the response to lithium treatment observed in the growth assays. In contrast, Erk phosphorylation was slightly decreased in PTEN−/− and PIK3CA knock in cells and moderately decreased in EGFR overexpressing clones upon lithium treatment when no EGF was added to the growth medium (Fig. 5 and left panel and Sup. Fig. 3, left panel). However, marked decreases in Erk phosphorylation were seen in EGFR overexpressing cell lines and PIK3CA knock in cells when exposed to lithium under physiologic concentrations of EGF, but interestingly Erk phosphorylation appeared to be unaffected in PTEN−/− cell lines after lithium treatment under these conditions (Fig. 5 and right panel). As expected, in all cases where lithium demonstrated some inhibition of proliferation, slight to moderate decreases in cyclin D1 protein were noted (Fig. 5 and Sup. Fig. 3).
The fact that lithium's effects were most pronounced on EGFR overexpressing and mutant PIK3CA knock in clones presented the intriguing possibility that perhaps lithium's effects were being mediated more proximally in the MAPK/PI3K pathways. To explore this hypothesis, we initially performed in vitro PI3K competitive kinase/ELISA assays in the presence and absence of lithium, but were unable to discern any consistent differences in PI3K in vitro activity (data not shown). We then performed Western blot analyses to examine EGFR tyrosine phosphorylation (1173), a marker of EGFR activation, in the presence and absence of lithium in EGFR overexpressing, mutant PIK3CA and parental MCF-10A control cells (Sup. Fig. 4). Interestingly, these results demonstrated that total EGFR levels decreased in parental MCF-10A, mutant PIK3CA, PTEN null and AKT1 E17K cells in the presence and absence of EGF upon lithium exposure. Although there appears to also be decreased EGFR phosphorylation with lithium in EGF containing conditions, this is likely due to total EGFR levels decreasing. EGFR overexpressing clones however, did not show any appreciable change in total or phosphorylated EGFR with lithium treatment regardless of whether EGF was present in the media. However, due to the fact that decreases in total EGFR did not correlate with sensitivity to lithium, it can be concluded that changes in total EGFR protein or its phosphorylation are not responsible for the growth inhibitory effects of lithium in mutant PIK3CA and EGFR overexpressing cell lines.
In sum, our results demonstrate that different genetic alterations within the same pathway are not biologically or functionally equivalent, as evidenced by the differences in biochemical pathway activation and responses to EGF stimulation. Moreover, because of the isogenic nature of our cell line panel, these data strongly suggest that genetic changes within a common pathway may not uniformly predict for sensitivity to a given pathway inhibitor.
Cell-based models of human cancers have traditionally been used for drug discovery. Though their use has led to the development of effective therapies, lack of proper control cells has hindered the ability to validate a given compound's true target or mechanism of action and has often led to disappointing results in clinical trials. For example, the small molecule inhibitors gefitinib and erlotinib were originally proposed to be effective agents against lung cancer cells overexpressing EGFR although only 10% of patients responded to these drugs. Retrospective examination of these tumors demonstrated that certain EGFR mutations were associated with responses,38–40 thus these mutations, rather than EGFR overexpression, have become better predictors of response to gefitinib and erlotinib.41 Recently, the use of isogenic non-tumorigenic human breast epithelial cell lines has been proposed as a model for drug discovery using targeted gene replacement strategies.24 Indeed, these studies demonstrated that EGFR mutations knocked into human breast epithelial cells dramatically sensitized cells to erlotinib thus recapitulating human clinical trial data. Additionally, prospective proof of principle concepts using isogenic cell lines to develop cancer therapies have been realized. A recent exciting clinical trial demonstrated the efficacy of poly (ADP-ribose) polymerase (PARP) inhibitors in BRCA mutation positive breast cancers.42 This was the direct result of a “synthetic lethal” drug screen employing isogenic cell lines to identify PARP inhibitors' selectivity for BRCA null cells.43,44 The data from these clinical trials demonstrate that better defined genetic models of cancer can ultimately result in highly effective targeted therapies and equally important, reliable predictors of response to those therapies.
Uncontrolled activity of the PI3K signaling pathway is found in many cancers. Our study further characterizes the disparate effects of various aberrations that occur within the PI3K pathway in human breast cancers using our unique library of isogenic cell lines. Loss of the tumor suppressor Pten is common in breast cancers.8,9 We verify here that such loss is associated with increased constitutive activation of the MAPK and PI3K pathways; however, the level of Akt activation is not as extensive as that caused by a PIK3CA mutation and does not confer equivalent sensitivity to the toxic effects of lithium. In contrast to PIK3CA mutations and Pten loss, the presence of an isolated AKT1 E17K mutation leads to minimal changes in the MAPK and PI3K pathways and as such, these cells appear relatively resistant to lithium compared to cells harboring an oncogenic PIK3CA mutation.
We demonstrate in this study that overexpression of EGFR leads to activation of multiple oncogenic pathways in a similar manner as somatic cell knock in of mutant PIK3CA. Thus a mutation of PIK3CA appears to mimic a proximal alteration resulting in the complex activation of multiple pathways including PI3K and MAPK signaling. Although activation of these two pathways by EGFR overexpression was expected based upon previous knowledge of RTK signaling, the signaling induced by mutant PIK3CA is confirmed to be of greater complexity as we previously reported in reference 25. We hypothesize that “rewiring” by mutant PIK3CA leads to MAPK pathway activation, perhaps by virtue of the known binding that occurs via Ras and the Ras binding domain of p110a. Our results demonstrate that targeted therapies against key components in both the MAPK and PI3K pathways may yield the most effective results in cancers that harbor PIK3CA mutations. This may be especially relevant given the current inability to isolate mutant specific PI3K inhibitors. Interestingly, various inhibitors of the PI3K and MAPK pathways such as wortmannin, LY294002, U0126 and rapamycin have shown some activity in our PIK3CA knock in and PTEN knock out cell lines,15,25,26 but depending on the culture conditions, non-selective toxicity due to off target effects was also present making interpretation of dual pathway inhibition difficult to assess.
Interestingly, loss of Pten led to only a partial sensitivity to lithium treatment despite baseline increases in Akt and Erk phosphorylation. However, in the absence of exogenous EGF, PTEN knock out cells had relatively less Akt phosphorylation and relatively more Erk phosphorylation compared to the highly lithium sensitive EGFR overexpressing and mutant PIK3CA cell lines. Intriguingly, lithium treatment of PTEN−/− cells modestly decreased Akt phosphorylation, yet there was no appreciable effect on Erk phosphorylation under physiologic EGF concentrations, a result that was distinct from EGFR overexpressing and mutant PIK3CA clones. Because of the isogenic nature of these cell lines, our results provide further evidence against the notion that pathway alterations in the PI3K pathway are functionally similar. During the preparation of this manuscript, Parsons and colleagues reported the newly discovered role of P-Rex2a in regulating Pten activity, as well as increased P-Rex2a activity in cancers with PIK3CA mutations.45 It is tempting to speculate that changes in P-Rex2a activity may account for some of the differences seen in our isogenic system between PTEN−/− and mutant PIK3CA knock in cell lines.
Lithium, an FDA approved drug, is classically thought to be an inhibitor of GSK3β, but paradoxically, decreased tumor growth has been seen in many systems using lithium and other GSK3β inhibitors.32–36 In our previous studies, we demonstrated that lithium decreases epithelial cell proliferation in cells harboring mutant PIK3CA and this was correlated with the upregulation of GSK3β, a known growth suppressor. However, in the current study, increases in GSK3β were not seen in EGFR overexpressing clones, despite the fact that these cell lines were otherwise phenotypically identical to mutant PIK3CA knock in cells (data not shown). More surprising was the finding that Akt and Erk, two molecular mediators of GSK3β phosphorylation and its subsequent inactivation,35,46,47 demonstrated significantly decreased phosphorylation in EGFR overexpressing and mutant PIK3CA knock in cell lines upon lithium treatment. We conclude that significant activation of both PI3K and MAPK pathways beyond a critical threshold may be required for lithium sensitivity. These results suggest that lithium's mechanism of action may not be at the level of GSK3β, but rather more proximal in the signaling cascades induced by these oncogenic stimuli. We therefore hypothesized that lithium may affect more upstream events such as PI3K activity or phosphorylation of the EGFR itself. However, our analyses excluded these possibilities. The fact that Pten null cells displayed only partial sensitivity to lithium and did not display decreases in Erk phosphorylation with lithium treatment suggests once again that loss of Pten is functionally distinct from mutant PIK3CA. Equally important, because our experiments were performed in isogenic cell lines, direct and conclusive comparison of different genetic alterations within the same pathway can be made. Thus, our study suggests that the somatic changes that occur in human cancers may not be used interchangeably as predictors of response to a given pathway inhibitor as evidenced by our varying results with lithium sensitivity.
Our results have significant potential for clinical relevance beyond the ability to use isogenic cell lines to validate targets of therapy and predictive markers of response. Based on the experiments presented here, there is the possibility of using this information clinically for the treatment of breast cancer patients whose tumors harbor mutant PIK3CA and/or EGFR overexpression. For example, combination therapy with lithium may be a strategy to prevent drug resistance, since previous studies have noted that EGFR expression is associated with resistance to hormonal and chemotherapies in breast cancers.48 In addition, the majority of breast tumors harboring a PIK3CA mutation are also hormone receptor positive, yet activation of the PI3K pathway is recognized as a mechanism of endocrine resistance,49 and resistance to HER2targeted agents.50,51 Thus we envision that the addition of lithium therapy to current chemo, hormone and HER2 directed therapies may augment the clinical efficacy of these agents and perhaps reduce the emergence of drug resistant tumors.
MCF-10A, empty vector control and AKT1 E17K mutant cells were maintained in DMEM:F12 supplemented with 5% horse serum, 20 ng/mL Epidermal Growth Factor (EGF), 10 µg/mL insulin, 0.5 µg/mL hydrocortisone and 0.1 µg/mL cholera toxin. All supplements were purchased from Sigma-Aldrich, St. Louis, MO. PIK3CA Exon 9 (E545K) knock in cells (PIK3CAEx9), PTEN−/− gene targeted clones of MCF-10A and EGFR overexpressing cells were grown in identical conditions except no EGF was added to the medium. EGFR overexpressing cells (clones labeled EGFR 7, EGFR 8, EGFR 9) were further supplemented with 180 µg/mL G418 (Invitrogen, Carlsbad, CA). All cell lines were grown at 37°C in 5% CO2.
A human EGFR cDNA was stably expressed in MCF-10A cells using the retroviral expression vector pFBneo, which was a kind gift from Dr. Anil K. Rustgi (University of Pennsylvania, PA). Retrovirus containing the coding sequence for EGFR was generated using Fugene6 (Roche Diagnostics, Indianapolis, IN) per the manufacturer's protocol in HEK-293T cells. Purified retrovirus was then used to infect MCF-10A cells following the manufacturer's protocol. Stable transformants were selected using 180 µg/mL G418 (Invitrogen, Carlsbad, CA). EGFR expression was confirmed by western blot using antibodies against total EGFR protein. Parental MCF-10A cells were also stably transduced in parallel with an empty retroviral expression vector pFBneo (named Empty Vector or EV) and selected in the same manner to serve as controls for all experiments.
Cells were prepared by seeding each cell line in DMEM:F12 medium without phenol red, supplemented with 1% charcoal dextran-treated fetal bovine serum (Hyclone), 10 µg/mL insulin, 0.5 µg/mL hydrocortisone, 0.1 µg/mL cholera toxin at a density of 100,000 cells per 25 cm2. Medium was changed to either EGF-free or 0.2 ng/mL EGF-containing medium in the absence and presence of 10 mM LiCl on days 1 and 4 as indicated. Cells were counted and evaluated for viability on days 1 and 6 using a Vi-CELL Cell Viability Analyzer (Beckman Coulter). All assays and growth conditions were performed in triplicate and repeated at least three times.
Lysates for cells grown in each experimental condition were prepared as previously described in reference 52. Western blotting was performed using the NuPage XCell SureLock electrophoresis system (Invitrogen, Carlsbad, CA) and PVDF membranes (Invitrogen, Carlsbad, CA). Primary antibodies were added overnight at 4°C, while secondary antibodies, conjugated with horseradish peroxidase were added for 1 hr at RT. Antibodies used in this study were anti-EGFR rabbit antibody (2232; Cell Signaling Technology), anti-phospho EGFR (Tyr 1173) rabbit anti-body (4407L; Cell Signaling Technology), anti-AKT rabbit antibody (9272; Cell Signaling Technology), anti-phospho AKT (Ser 473) rabbit antibody (9271; Cell Signaling Technology), anti-p42/p44 MAP kinase rabbit antibody (9102; Cell Signaling Technology), anti-phospho p42/p44 MAP kinase (Thr-202/Tyr-204) mouse antibody (9106; Cell Signaling Technology), anti-cyclin D1 rabbit antibody (2922; Cell Signaling Technology), anti-GSK3β rabbit antibody (9315; Cell Signaling Technology), anti-p70S6 Kinase rabbit antibody (9202; Cell Signaling Technology), anti-phospho p70S6 Kinase rabbit antibody (9205S; Cell Signaling Technology) anti-amphiregulin rabbit antibody (ab48191; Abcam), anti-TGFalpha mouse antibody (ab9578; Abcam) and anti-GAPDH mouse antibody (6C5) (ab8245; Abcam). All primary antibodies were used at 1:1,000 dilutions except the anti-GAPDH antibody, which was used at a 1:40,000 dilution. Blots were exposed to Kodak XAR film using chemiluminescence for detection (Perkin Elmer). All experiments were performed at least three times, with representative figures shown.
Cells were seeded at approximately 10% confluency in bottom and top chambers of 6 well plates with transwell inserts (Cat# 3412, Corning) using DMEM:F12 medium without phenol red, supplemented with 1% charcoal dextran-treated fetal bovine serum (Hyclone), 10 µg/mL insulin, 0.5 µg/mL hydrocortisone, 0.1 µg/mL cholera toxin, without EGF. Medium was changed on days 1 and 4 and cells were then stained on day 6 with crystal violet to visualize cell proliferation. All experiments were performed in triplicate.
Standard error of the mean (SEM) was calculated for each proliferation assay. Statistical analyses were performed using a two-tailed Student's t-test and a one-way ANOVA across cell lines, which were calculated using Microsoft Excel and ezANOVA. A p value less than 0.05 was considered statistically significant.
We thank Anil Rustgi for providing the DNA vectors.
ASCO Young Investigator Award (M.J.H., D.P.C.); DOD Breast Cancer Research Program BC087658 (M.J.H.), W81XWH-06-1-0325 (J.P. Gustin), BC083057 (M.M.); KG090199 (J.D.L.), BCTR0707684 (B.H.P.); Flight Attendant Medical Research Institute (FAMRI) (J.D.L., H.K.), the V Foundation (J.D.L.), the Maryland Cigarette Restitution Fund (J.D.L.), the Avon Foundation (J.D.L., B.H.P.), NIH CA088843 (J.D.L., B.H.P.), CA109274 (J.P. Garay, B.H.P.), GM007309 (G.M.W.) CA009071 (D.P.C., D.J.); Susan G. Komen for the Cure PDF0707944 (A.M.A.) and the Breast Cancer Research Foundation (B.H.P.).
B.H.P. has received prior research funding from GlaxoSmithKline (GSK) though none of the studies reported here were supported by GSK. B.H.P. is a consultant for GSK and is on the scientific advisory board for Horizon Discovery LTD., and is entitled to payments for these services. These arrangements are managed according to the Johns Hopkins University conflict of interest policy.