Extracellular growth factors and mitogens cooperate to drive mitochondrial biogenesis

Proliferating cells normally duplicate their organelles before they divide. Likewise, growing post-mitotic cells increase many of their organelles with increasing cell size. In both cases, organelle biogenesis must be coordinated with cell growth, suggesting that the processes are coupled. However, there are exceptions where cells can preferentially increase specific organelles without overall growth: two examples of such independent regulation are brown adipocytes exposed to low temperatures, which increase their mitochondrial content to produce heat (reviewed by Puigserver and Spiegelman, 2003) and B-lymphocytes exposed to antigen, which extend their endoplasmic reticulum to produce antibodies for secretion (Rush et al., 1991). Organelle biogenesis also occurs in non-growing cells to maintain homeostasis during normal organelle turnover (Klionsky and Emr, 2000). Yet, despite the importance of the regulation of organelle biogenesis, the signalling pathways involved are poorly characterised.

Mitochondria have several characteristics that make them attractive for studying the regulation of organelle biogenesis. The amount of mitochondria differs greatly in different cell types, moreover, the amount of mitochondria in a specific cell type can change in response to extracellular stimuli, such as nutrients, temperature or exercise, indicating that the amounts of mitochondria in a cell are tightly regulated (Hood, 2001; Nicholls et al., 1986; Nisoli et al., 2003; Scarpulla, 2008; Wu et al., 1999). Mitochondria are the primary sites of energy production in eukaryotic cells and it is thought that variations in mitochondrial amount reflect the energy requirements of different cell types in specific conditions (Wallace, 2005). However, in addition to their role as a power supply for eukaryotic cells, mitochondria are centres of biogenesis – utilising carbohydrate, fatty acids and amino acids to produce intermediates for new lipid and protein production – and so biosynthetic requirements might also be important. Mitochondria also contain their own genome (mtDNA), which encodes a small sub-set of the proteins that are essential for mitochondrial function. Therefore, mitochondrial biogenesis requires the coordinated synthesis of proteins encoded by both the mitochondrial and nuclear genomes. Although it has been shown that mtDNA replication can be regulated independently of nuclear DNA replication, little is known about how it is coordinated with cell growth and division (Moraes, 2001).

Studies, performed mainly in skeletal muscle cells and adipocytes, have established that stimuli such as exercise, cold, starvation and nitric oxide (NO) can promote mitochondrial biogenesis. A number of transcription factors (NRF1, NRF2, ERRα and PPARδ) and coactivators (PGC-1-α, PGC-1-β, PRC) have been shown to be part of the molecular machinery involved in promoting mitochondrial biogenesis in these conditions (Finck and Kelly, 2006; Handschin et al., 2007; Jager et al., 2007; Meirhaeghe et al., 2003; Nicholls et al., 1986; Nisoli et al., 2003; Puigserver and Spiegelman, 2003; Puigserver et al., 1998; Scarpulla, 2006; Scarpulla, 2008). In the present study, we have used primary Schwann cells to study how extracellular growth and mitogenic factors can stimulate mitochondrial biogenesis. Schwann cells have several advantages that make them a useful cell-type for studying biogenesis. They are primary cells that can be maintained in the absence of growth factors and mitogens without significant cell death, providing a clean background to assess the effects of extracellular signals (Conlon et al., 2001; Mathon et al., 2001). In addition, two well-characterised extracellular signal proteins have distinct effects on Schwann cell behaviour: insulin-like growth factor 1 (IGF1; hereafter referred to as IGF), on its own, promotes cell growth (cell enlargement) but not cell-cycle progression, whereas neuregulin (NRG), on its own, promotes cell-cycle progression but not cell growth, thus allowing growth and mitogenic pathways to be studied separately. However, when the two factors are added together, they synergise to drive cell-cycle progression but not to drive cell growth. So, cells treated with NRG and IGF add volume at the same rate as cells treated with IGF alone, but only the cells treated with NRG proliferate (Conlon et al., 2001; Echave et al., 2007).

We report here that NRG and IGF synergise to drive mitochondrial biogenesis but not cell growth, leading to an increase in mitochondrial density as well as mtDNA copy number. This synergistic effect is evident at the gene expression level and involves the transcription factor ERRα (estrogen-related receptor α). Interestingly, although both the ERK (extracellular signal-related kinase; also known as MAPK1) and the phosphoinositide 3-kinase (PI3K) signalling pathways are required, the effect is independent of Akt and TOR. This ability to regulate mitochondrial biogenesis independently of cell growth allows these cells to regulate the density of mitochondria according to their metabolic needs.

To assess the effects of NRG and IGF on mitochondrial biogenesis, we used similar conditions to those used previously to analyse cell size (Conlon et al., 2001; Echave et al., 2007). Proliferating Schwann cells were cell-cycle arrested in S phase, by the addition of aphidicolin, to ensure that all cells were at the same stage of the cell cycle and to remove the effects of different division rates on the number of mitochondria per cell. Aphidicolin, an inhibitor of DNA polymerase α, blocks nuclear DNA replication but, because it is inactive against the mitochondrial DNA polymerase DNApol γ, it has no effect on mitochondrial DNA replication. Aphidicolin-arrested cells were transferred into defined medium (without IGF or NRG) for 24 hours prior to stimulation with IGF, NRG or both. As reported previously (Conlon et al., 2001), IGF promoted an increase in cell volume, whereas NRG addition did not and did not increase the growth caused by IGF (Fig. 1A). To measure the amount of mitochondria, we used the mitochondrial-specific dye, MitoTracker Green, which is reportedly independent of membrane potential, and quantified total mitochondrial volume using stacked images obtained by scanning confocal microscopy (Casley et al., 2002). The addition of either NRG or IGF alone did not significantly increase the amount of mitochondria per cell compared with untreated cells, whereas the combination of NRG and IGF caused a dramatic increase in the amount of mitochondria (Fig. 1B,C; supplementary material Fig. S1A). To confirm the results, we repeated the experiments with a second dye (MitoTracker CMXROS) and obtained essentially the same results (supplementary material Fig. S1B,C). These results demonstrate that the two factors, IGF and NRG, act synergistically to drive mitochondrial biogenesis.

As NRG and IGF synergise to stimulate mitochondrial biogenesis but not cell growth, treatment with both factors should increase mitochondrial density. To determine if this was the case, we used the mitochondrial dye (TMRM), together with CellTracker, to measure the percentage of the cell volume occupied by mitochondria. As shown in Fig. 1D, mitochondria constituted ∼7% of cell volume in untreated cells, and in cells treated with IGF or NRG the percentage was similar. Strikingly, in cells treated with both NRG and IGF, mitochondrial density increased greatly, with mitochondria occupying >20% of the cell volume. To confirm these relative mitochondrial densities we calculated them using a different method, by dividing the mean mitochondrial volume measured in the MitoTracker studies (Fig. 1C) by the mean cell volumes measured in a Coulter counter, and, as shown in Fig. 1E, we obtained a similar result. This increase in mitochondrial density clearly demonstrates that the increase in mitochondrial biogenesis in response to extracellular factors does not just reflect an increase in overall cell growth. Instead, specific extracellular signals can act preferentially to increase the amount of mitochondria within a cell without a parallel increase in cellular volume. Following on from this, cells proliferating in steady-state conditions, driven by NRG and IGF, should also have a comparable mitochondrial density to the aphidicolin-arrested cells. This indeed was the case, in that cells proliferating in NRG and IGF had a higher density of mitochondria than similar sized cells in the absence of these factors and a similar density to cells arrested in aphidicolin and treated with both NRG and IGF (Fig. 1E). To address whether the increase in mitochondrial density had a physiological impact on the cells, we measured oxygen uptake. These experiments showed that consistent with increased mitochondrial amount, the cells pre-treated with NRG and IGF had significantly higher levels of oxygen-dependent metabolism compared with control cells (Fig. 1F).

To analyse the effects of NRG and IGF on mtDNA replication we took two approaches: (1) using immunofluorescence we measured the induction of mtDNA replication by monitoring bromodeoxyuridine (BrdU) incorporation into mtDNA; (2) we measured the amount of mtDNA by real-time PCR. Pulse labelling of aphidicolin-arrested cells with BrdU resulted in a three- to fourfold increase in BrdU immunofluorescence per cell in response to treatment with both NRG and IGF (Fig. 1G); addition of either factor alone had no detectable effect (data not shown). Real-time PCR analysis showed that treatment with NRG and IGF for 24 hours resulted in a ∼50% increase in the number of copies of mitochondrial DNA compared with control cells (Fig. 1H; supplementary material Fig. S1D); again, addition of either factor alone had no detectable effect. These results show that NRG and IGF synergise in stimulating mitochondrial DNA replication, as they have been shown to do in stimulating nuclear DNA replication.

To obtain a global overview of the transcriptional response to NRG and IGF, we performed microarray analysis of cells treated with each factor alone or in combination and compared the results with those in untreated cells. We then performed GO analysis of the microarray data to determine the response of mitochondria-related genes. As shown in Fig. 2A, when either NRG or IGF were added alone, there was little effect on the levels of mitochondria-related transcripts. However, when NRG and IGF were added together, there was a synergistic increase in the number of genes affected, with a highly significant enrichment of mitochondrial genes (P=0.00345; Fig. 2A and supplementary material Table S1). Thus the synergy between NRG and IGF to drive mitochondrial biogenesis and to increase mitochondrial density is reflected at the level of mitochondrial gene expression. To verify these results and to analyse the gene expression response further, we used real-time PCR to measure cytochrome c and prohibitin mRNA levels, two mitochondria-related RNAs that were upregulated in the microarrays and previously used as markers of mitochondrial biogenesis (Kelly and Scarpulla, 2004; Merkwirth and Langer, 2009). It should be noted however, that although prohibitin was located mainly in the mitochondria there was some in the nucleus (supplementary material Fig. S2A). In agreement with the microarray analysis, treatment of cells with NRG or IGF alone resulted in a small induction of these two mRNAs (cytochrome c, two- to threefold; prohibitin, 1.5- to twofold), whereas a much larger response was seen when NRG and IGF were added together [cytochrome c, five- to sixfold; prohibitin four- to 4.5-fold; Fig. 2B). Noticeably, the fold changes detected by real-time PCR were larger than those seen in the microarrays, suggesting that the PCR analysis was the more sensitive assay. This was confirmed by the analysis of Isocitrate dehydrogenase, IDH3 – a gene linked to mitochondrial biogenesis – as a change in this mRNA was not detected in the microarray assay but was detected by real-time PCR (Fig. 2B). To determine whether the increases in RNA resulted in corresponding increases in the proteins they encode, we analysed the levels of cytochrome c and prohibitin proteins by western blotting and found that the levels of both proteins increased significantly following the co-addition of NRG and IGF (Fig. 2C).

Previous studies have established that both NRG and IGF signal through the ERK and the PI3K signalling pathways (Schlessinger, 2000). Using inhibitors of these pathways we found that the NRG-induced proliferation of Schwann cells was strictly dependent on ERK signalling (Fig. 3A; and not shown). By contrast, the IGF-induced growth of these cells was independent of ERK. IGF-induced growth, however, was abolished by inhibition of the PI3K pathway (Fig. 3B). Consistent with these findings, western blotting showed that whereas both NRG and IGF activated both pathways, NRG caused a more sustained activation of the ERK pathway and entry into S phase, whereas IGF caused a more sustained activation of the PI3K pathway but did not trigger significant entry into S phase (Fig. 3C; supplementary material Fig. S2B). Thus, in this cell system, sustained ERK signalling promotes proliferation but not cell growth, whereas sustained PI3K signalling promotes cell growth. To study the role of these signalling pathways in mitochondrial biogenesis we used the levels of mRNA encoding mitochondrial proteins as a read-out of the synthetic pathways involved. As shown in Fig. 3D and supplementary material Fig. S2C, pre-treatment of the cells with inhibitors to these pathways showed that, in contrast to the effects on cell growth, the increase in cytochrome c and prohibitin mRNAs depended on both the ERK and PI3K pathways. Thus, NRG and IGF synergise via the ERK and PI3K pathways to stimulate mitochondrial biogenesis, just as they do to stimulate cell-cycle progression. To test whether these pathways might be important in regulating mitochondrial biogenesis in other cell types, we investigated the role of these pathways in a second cell type, REF52s (a rat fibroblast cell line). Similarly to the effects in Schwann cells, we found that extracellular (serum) induction of cytochrome c was also dependent on the ERK and PI3K signalling pathways (supplementary material Fig. S2D).

Akt and TOR are important downstream components of the PI3K signalling pathway and have been shown to be crucial regulators of cell growth (Bhaskar and Hay, 2007; Sarbassov et al., 2005). However, as shown in Fig. 4A, inhibition of Akt and TOR had no effect on the ability of NRG and IGF to increase cytochrome c and prohibitin mRNA levels. By contrast, and consistent with their reported role in cell growth, inhibition of Akt or TOR blocked the ability of IGF to stimulate cell growth (Fig. 4B; supplementary material Fig. S3A). To confirm the role of PI3K, we used a second PI3K inhibitor PIK4, which preferentially inhibits the class 1 family of these kinases, and this gave similar results (supplementary material Fig. S3B), as did three other inhibitors of PI3K-PI-103, ZSTK474 and wortmanin (supplementary material Fig. S3C,D). This indicates that although PI3K signalling cooperates with ERK signalling to drive mitochondrial biogenesis it is independent of Akt or TOR.

To confirm these findings, we tested whether constitutive Akt signalling could induce the expression of mitochondrial genes. We isolated pools of Schwann cells expressing a tamoxifen-inducible form of Akt (NSAktER) (Kohn et al., 1998). Activation of Akt in these cells (in the absence of extracellular factors) resulted in phosphorylation of the AktER fusion protein and the activation of S6 kinase, and was sufficient to promote an increase in cell size (Fig. 5A). It did not, however, promote an increase in cytochrome c and prohibitin mRNA levels or mtDNA copy number (Fig. 5B,C). Moreover, activation of Akt in the presence of NRG did not augment the levels of cytochrome c induced by NRG alone (Fig. 5D).

In a recent report, it was shown that the increase in fatty acid oxidation and TCA-cycle activity following exercise-induced cardiac hypertrophy in mice is dependent on PI3K, independent of TOR and dependent on PKCζ and PKCλ (O'Neill et al., 2007). We therefore investigated whether PKC is involved in the increase in mitochondrial gene expression stimulated by NRG+IGF. Initially, we used the general PKC inhibitor bisindolylmaleimide I (BIM), which is able to block the activity of most PKC isoforms (Davies et al., 2000; Davis et al., 1992). As shown in supplementary material Fig. S4A, it had no effect on the increased expression of mitochondria-related genes induced by NRG+IGF but it did block the phosphorylation of MARCKS (supplementary material Fig. S4B), a target of both novel and conventional PKCs (Disatnik et al., 2002; Gallant et al., 2005). To investigate the role of the atypical PKCs that are not blocked by BIM, we used an inhibitor specific for PKCζ and PKCλ, with similar results (supplementary material Fig. S4A). These results show that PKC is not required for NRG plus IGF to increase the expression of mitochondrial genes.

Oncogenic Ras signalling results in constitutive activation of the ERK and PI3K pathways, which should, based on our results, lead to an increase in mitochondrial density. To circumvent the p53-dependent cell-cycle arrest induced in primary cells by oncogenic Ras signalling (Lloyd et al., 1997), we analysed the effects of oncogenic Ras signalling in Schwann cells expressing a dominant-negative form of p53 (DNp53). We arrested DNp53 cells expressing either Ras (DNp53Ras) or the empty vector (DNp53V) in aphidicolin, before transferring them to defined medium in the absence of NRG and IGF. We isolated mRNA 24 hours later and measured the levels of cytochrome c and prohibitin mRNAs by real-time PCR and found that oncogenic Ras signalling was sufficient to increase cytochrome c and prohibitin mRNA levels (Fig. 6A). Moreover, confocal imaging showed that, in the absence of NRG and IGF, oncogenic Ras signalling was sufficient to increase the amount of mitochondria (Fig. 6B), and western blotting confirmed an increase in cyctochrome c levels. Importantly, this increase in mitochondria resulted in higher levels of oxygen-dependent metabolism (Fig. 6C). Similar to stimulation by NRG and IGF, the stimulation by oncogenic Ras was dependent on ERK and PI3K signalling but independent of TOR (Fig. 6A). In complete medium, in the presence of NRG plus IGF, DNp53V and DNp53Ras cells proliferated at a similar rate (not shown) and have a similar cell size (Fig. 6D). However, the amount of mitochondria per cell was significantly higher in the Ras-expressing cells (Fig. 6E), which corresponded with higher levels of cytochrome c mRNA and protein (Fig. 6F). Thus oncogenic Ras is sufficient to drive mitochondrial biogenesis and resulted in an even greater density of mitochondria than that stimulated by extracellular factors.

In our microarray analysis, we observed that NRG and IGF increased the mRNA encoding PGC-1-related-coactivator (PRC), a transcriptional coactivator previously linked to mitochondrial biogenesis and reported to be induced by serum (Vercauteren et al., 2006). We confirmed the NRG and IGF induction of PRC by real-time PCR analysis and showed that it was dependent on both ERK and PI3K signalling (Fig. 7A,B). To test the role of PRC in the induction of mitochondrial biogenesis, we used siRNA to reduce the levels of PRC and then measured the levels of cytochrome c mRNA following NRG and IGF treatment. As shown in Fig. 7C, knockdown of PRC had no effect on the ability of NRG and IGF to increase cytochrome c mRNA levels, indicating that the induction of PRC was not required for the response. A number of transcription factors and other transcriptional coactivators have been reported to have a role in mitochondrial biogenesis (Scarpulla, 2006). We analysed the levels of mRNAs encoding a number of these factors by real-time PCR, following NRG and IGF treatment and found that one of these, the transcription factor ERRα was synergistically increased in response to NRG plus IGF; by contrast, the levels of the mRNAs encoding PGC-1-α, PGC-1-β, TFAM and NRF1 and NRF2 remained unchanged by these factors (Fig. 7A and not shown). The NRG plus IGF-stimulated increase in ERRα mRNA levels was dependent on both ERK and PI3K signalling (Fig. 7B). Moreover, knockdown of ERRα, using two different siRNAs, significantly inhibited the increase in cytochrome c and IDH mRNAs induced by NRG and IGF, indicating a crucial role for this transcription factor in driving the response (Fig. 7D; supplementary material Fig. S4C). We could not test the effects of ERRα knockdown on basal levels of gene expression, as we were unable to detect significant knockdown of the very low levels of ERRα in these conditions, in the absence of extracellular factors. To test whether blocking this transcription factor resulted in an inhibition of mitochondrial biogenesis, we measured the amount of mitochondria in response to NRG and IGF, following knockdown of ERRα. As shown in Fig. 7E, we found that knockdown of ERRα blocked the ability of NRG and IGF to induce mitochondrial biogenesis, indicating an important role for a transcriptional response and specifically an important role for the transcription factor ERRα in this process.

In Schwann cells, growth and cell-cycle progression can be independently regulated by distinct extracellular factors: IGF driving cell growth, NRG stimulating proliferation, with a synergistic effect on the cell-cycle but not on growth. This division of labour allows a quiescent cell to grow without dividing and allows proliferating cells to divide at different sizes. The synergy of the pathways permits sustained, robust proliferation only in the presence of growth (Conlon and Raff, 2003; Conlon et al., 2001; Echave et al., 2007). In the present study, we show that IGF-stimulated growth is associated with sustained signalling through the PI3K pathway and is dependent on both Akt and TOR, whereas NRG-induced proliferation is associated with sustained signalling through the ERK pathway and is dependent on this pathway. Using this cell system we then investigated how these growth and proliferative pathways regulate mitochondrial biogenesis.

IGF or NRG alone do not stimulate a significant amount of mitochondrial biogenesis; however, there is substantial synergy when they act together, just as when they stimulate cell-cycle progression. In the presence of both IGF and NRG, cell growth is the same as in IGF alone (they add volume at the same rate) but mitochondrial biogenesis increases disproportionally, so the density of mitochondria increases. This resulting increase in mitochondrial density might be required to meet the great energy and biosynthetic demands of cell proliferation other than the addition of cell volume, such as DNA replication and the mechanics of cell division (DeBerardinis et al., 2008). Interestingly, this increase in mitochondrial density stimulated by the combination of IGF and NRG is seen both in proliferating cells and cells arrested in the cell cycle by aphidicolin. This would suggest that extra mitochondria are not generated in response to the demand for energy or biosynthesis by these processes but rather in preparation for these events.

Similarly, the combination of NRG and IGF increased the number of copies of mtDNA in a cell, whereas neither factor alone had a significant effect. This involved an increase in the synthesis of mtDNA, as the effect could be observed using immunofluorescence of freshly incorporated BrdU. Although much is known about the proteins that regulate the replication of mtDNA, what is not well understood is how the amount of mitochondrial DNA is measured, maintained or changed in response to extrinsic stimuli (Moraes, 2001; Scarpulla, 2008). MtDNA replication is known to occur independently of nuclear DNA replication, a finding confirmed here, as cells arrested with aphidicolin were still able to replicate their mtDNA. It is thought that mtDNA replication is a continual process, with turnover of mtDNA seen in post-mitotic cells in vitro (Kai et al., 2006) and it is thought to occur in post-mitotic tissues throughout life (Moraes, 2001). Here, we found that mtDNA levels increased in response to the same extracellular factors that drive mitochondrial biogenesis, providing a mechanism by which a cell can maintain the ratio of mtDNA:total mitochondria mass, but the downstream mechanisms remain unknown.

The ability of IGF and NRG to synergise to drive mitochondrial biogenesis was mirrored by their ability to increase the levels of mitochondria-related mRNAs. Microarray analysis indicated a synergistic and preferential induction of these genes – findings corroborated by real-time PCR. Interestingly, another class of genes that were preferentially upregulated were cell-cycle progression genes, consistent with the ability of NRG and IGF to cooperate to drive cell-cycle progression. This gene-expression response depended on both the PI3K and ERK signalling pathways. The importance of these pathways was also seen in a second cell type, the fibroblast cell line, REF52, in response to serum stimulation. Notably, in previous studies it was found that the correlation between ErbB2 and ERRα levels in breast cancer cells was also dependent on both ERK and PI3K signalling and that a NRG-induced increase in some mitochondrial transcripts in cardiocytes depended on ERK (Ariazi et al., 2007; Giraud et al., 2005). Together these results indicate an important role for the ERK and PI3K pathways in extracellular growth and mitogenic stimulation of mitochondrial biogenesis.

Interestingly, gene expression in Schwann cells in response to NRG plus IGF was independent of both Akt and TOR, which are required for the increases in cell size in response to growth factor signalling in these cells. Presumably, the separation of the pathways regulating mitochondrial biogenesis from the pathways regulating cell growth allows a cell to vary mitochondrial density in response to requirements other than cell size, which might be important in certain conditions. Other studies have found roles for TOR in regulating mitochondrial biogenesis. In mammalian cells, it has been reported that inhibition of TOR lowers mitochondrial membrane potential, oxygen consumption and ATP synthetic capacity (Schieke et al., 2006) and that TOR activity is required for the maintenance of mitochondrial gene expression (Cunningham et al., 2007). By contrast, studies in yeast, worms and mammals have suggested that inhibition of TOR can result in increased mitochondrial number and increased oxygen consumption (Bonawitz et al., 2007; Guarente, 2008). Undoubtedly, long-term inhibition of TOR has effects on mitochondria biology. TOR activity is crucial for the appropriate translation of many transcripts and it might be expected that long-term treatment with rapamycin would eventually affect any transcriptional response to extracellular growth factors and mitogens. Unfortunately, we have been unable to test this in our system, as long-term treatment with rapamycin is toxic. However, in agreement with other studies we did find that rapamycin treatment of Schwann cells resulted in a loss of mitochondrial potential and a dramatic change in mitochondrial structure (P.E. and A.C.L., unpublished observations). This effect is distinct from what we observed in untreated cells in the absence of factors; in these conditions, TOR activity was also very low (as measured by phosphorylation of downstream targets) but mitochondrial structure and potential were relatively similar to those in cells in the presence of factors (there are just less mitochondria). This might suggest either a toxic effect of rapamycin on mitochondrial physiology or indicative that rapamycin decreases TOR activity to an even greater extent than growth factor removal. Where our results are clear, is that the initial transcriptional induction of mitochondrial gene expression following NRG and IGF addition, which is required for the increase in mitochondrial biogenesis, is independent of TOR.

In thinking about candidate proteins downstream of PI3K, other than Akt and TOR, we noted that PKCζ and λ had recently been shown to be required for changes in mitochondrial metabolism in the heart following exercise (O'Neill et al., 2007). There are important differences between this system and the effect of NRG and IGF on mitochondria biogenesis, in that in the study by O'Neill and colleagues increased mitochondrial function was not accompanied by changes in mitochondria amount or mtDNA copy number. In our system, we could find no evidence of a role for any member of the PKC family in regulating the expression of mitochondria-related genes downstream of NRG and IGF. Moreover, in results not shown, we found that inhibitors of Src and RSK also had no effect on NRG- and IGF-stimulated changes in mitochondria-related gene expression. Further studies will be required to identify the relevant intracellular signalling pathways.

It is well established that oncogenic Ras signalling constitutively activates the ERK and PI3K signalling pathways. It has also been reported that cancer is associated with aberrant mitochondrial metabolism (Gogvadze et al., 2008). We find that oncogenic Ras is sufficient to drive mitochondrial biogenesis in the absence of extracellular factors and that this requires both the ERK and PI3K pathways but not TOR. Although we have not ruled out that this is an S-phase-specific effect, importantly, this aberrant signalling is sufficient to promote an increase in mitochondrial density in proliferating cells, despite the presence of NRG and IGF, indicating that constitutive Ras signalling results in a higher density of mitochondria than that induced by physiological growth and mitogenic factors. This is likely to contribute to the aberrant mitochondrial metabolism of cancer cells.

Many of the key components of the mitochondrial transcriptional machinery have been identified (Finck and Kelly, 2006; Scarpulla, 2006). Of these, we found that only the transcription factor ERRα and the nuclear coactivator PRC were increased at the mRNA level in response to NRG and IGF. It has previously been reported that PRC mRNA is induced by serum (Vercauteren et al., 2006) and it seems probable that the signalling pathways involved are similar to those induced by NRG and IGF. However, we found that siRNA knockdown of PRC had no effect on the NRG plus IGF-induced increase in mitochondria-related genes, suggesting that PRC is not a crucial mediator of this response. The knockdown of ERRα, however, significantly inhibited the induction of mitochondrial biogenesis. This demonstrates that the transcriptional response to NRG and IGF is required for these factors to drive mitochondrial biogenesis, and that ERRα is an important mediator of this response.

The ERR family of transcription factors, of which ERRα is the best characterised, are thought to have little transcriptional activity in the absence of transcriptional coactivators (Mootha et al., 2004; Schreiber et al., 2003). Instead they have been shown to synergise with the PGC-1 family of transcriptional coactivators to drive the transcription of proteins involved in mitochondrial biogenesis as well as other metabolic processes (Carrier et al., 2004; Huss et al., 2002; Mootha et al., 2004; Schreiber et al., 2004; Schreiber et al., 2003). In particular, there is strong evidence that ERRα is a crucial mediator of PGC-1-α-stimulated mitochondrial biogenesis (Mootha et al., 2004; Schreiber et al., 2004). Consistent with our studies, expression of a constitutively activated form of ERRα has been shown to be capable of activating the transcription of many of the mitochondria-related genes that we found are induced in response to NRG and IGF and to increase mitochondrial biogenesis, as measured by a relative increase in mitochondrial DNA (Schreiber et al., 2004). Although our studies indicate a crucial role for ERRα in mediating the NRG plus IGF-induced increase in mitochondria, we have not identified the transcriptional coactivators that are involved. PRC is the only family member induced at the RNA level following NRG plus IGF treatment but siRNA analysis indicated that this induction is not required. Schwann cells express both PGC-1-α and PGC-1-β and it is possible that either or both of these are involved, especially as the PGC-1 family are subject to multiple post-translational modifications and have been reported to be regulated by NRG (Canto et al., 2007; Lin et al., 2005). Alternatively, other transcriptional coactivators might be involved. Other transcription factors might also be important: for example ERRα has been shown to cooperate with NRF1 and NRF2 (Schreiber et al., 2004) and NRG has been shown to phosphorylate and induce NRF2-mediated mitochondrial gene expression (Canto et al., 2007; Fromm and Burden, 2001). Further studies will be required to dissect what is likely to be a highly regulated process.

Our results indicate that ERRα has a major role in the induction of mitochondria biogenesis in response to extracellular growth factors and mitogens, yet ERRα mice appear relatively normal. ERRα^–/– mice are slim and resistant to high-fat induced obesity (Luo et al., 2003) and later studies found important roles for ERRα in the biogenesis of mitochondria in brown adipose tissue and adaptive thermogenesis (Villena et al., 2007) and in the adaptive bioenergetic response to stress-induced heart failure (Huss et al., 2007). However, although several studies have reported that these mice have decreased expression of mitochondrial genes including cytochrome c, the phenotype is less severe than might be predicted from our studies (Herzog et al., 2006; Luo et al., 2003). One potential explanation for the disparity of these findings is that there is strong evidence of adaptive responses in the mice, particularly for the upregulation of the family member, ERRγ (Alaynick et al., 2007; Dufour et al., 2007; Herzog et al., 2006; Huss et al., 2007; Huss et al., 2004). It will be of great interest to determine the effects of loss of all ERR family members in mice.

Previous studies have indicated an important role for ERRα in regulating mitochondrial biogenesis in non-proliferative tissues in response to energetic requirements. Here, we report an important role for this transcription factor in linking extracellular growth and mitogenic signals to mitochondrial biogenesis. Importantly, our results also demonstrate that by uncoupling the pathways that drive mitochondrial biogenesis from those increasing cell size, extracellular factor signalling can result in changes in the density of mitochondria in cells. This is likely to be an important mechanism to couple mitochondrial biogenesis to the metabolic requirements of growing and proliferating cells.

Schwann cells were purified from P7 rat sciatic nerve by immunopanning (Mathon et al., 2001), and expanded on poly-D-lysine (PDL) in DMEM supplemented with 3% FCS, 1 μM forskolin and recombinant neuregulin (NRG) (complete medium). REF52 cells (a rat fibroblast cell line) were expanded in DMEM supplemented with 10% FCS.

Cells were plated in complete medium on dishes coated with PDL and fibronectin for 24 hours before being treated with aphidicolin (1 μg/ml) in complete medium for a further 24 hours. Then cells were switched into defined medium (DMEM containing 100 μg/ml transferrin, 100 μg/ml BSA, 16.1 μg/ml putrescine and 39 ng/ml selenium) plus aphidicolin for 24 hours. The factors to be tested were added in defined medium plus aphidicolin for a further 24 hours: NRG 20 ng/ml (R&D Systems), IGF 100ng/ml (Autogen Bioclear) or both. At these concentrations, NRG and IGF were saturating for proliferation or cell growth, respectively. For inhibitor studies, performed in the absence of aphidicolin, cells were treated with either 15 μM U0126 (Promega), 15 μM LY294002 (BioMol), 10 μM triciribine, 0.1 μM rapamycin, 10 μM BIM, 1 μM pseudo-substrate inhibitor specific for PKCζ and PKCλ (Calbiochem), 0.25 μM PIK4 (528116 PI3Kα inhibitor VIII), 200 nM wortmannin, 10 μM compound C, 0.5 μM PI-103 (Calbiochem), 250 nM ZSTK474 (Alexis Biochemicals), 1 μM or 7.5 μM PD184352 (REF52 cells; Axon) or appropriate vehicle control.

To measure cell volume in a Coulter counter, cells were trypsinised, resuspended in Isoton II and >5000 cells counted from triplicate plates. The data were analysed with Coulter multisizer Accucomp colour software (Beckman-Coulter). All experiments were repeated three times with similar results.

Cells were loaded with 100 nM MitoTracker CMXROS probe (Molecular Probes) according to the manufacturer's instructions. The cells were fixed with 4% paraformaldehyde (PFA) and permeabilised with cold acetone for 5 minutes. Cells were then washed several times with PBS. Serial sections were examined with a Nikon E1000 confocal microscope. Volume determination from stacked images was performed using Volocity software. Mitochondrial dyes are sensitive to mitochondrial membrane potential. Therefore to avoid bias in the measurements due to changes in potential, the intensity of the staining was ignored, and only the number of pixels corresponding to mitochondria was quantified. Identical measurements were also performed with 100 nM MitoTracker Green (Molecular Probes). To measure the percentage of the cell occupied by mitochondria, cells were loaded with two different dyes, TMRM (Invitrogen) and CellTracker (Invitrogen), according to the manufacturer's instructions, to stain for mitochondria and the cell cytoplasm, respectively. Following 20 minutes incubation at 37°C, sections were taken with a Leica SPE confocal microscope.

To visualize mtDNA replication, BrdU (10 μM) was added to the medium of aphidicolin-arrested cells for 24 hours. The cells were then fixed and labelled with a monoclonal anti-BrdU antibody (Roche) and Hoechst 33342.

Total DNA was extracted and purified from Schwann cells with a DNeasy kit (Qiagen) and used to quantify the amount of mtDNA with respect to the genomic DNA present in the preparation, by quantitative PCR. Primers to quantify mtDNA were complementary to the D-loop region (forward: 5′-GGTTCTTACTTCAGGGCCATCA-3′, reverse: 5′-GATTAGACCCGTTACCATCGAGAT-3′). Primers to quantify genomic DNA were directed against the GAPDH gene sequence.

To measure respiration rate, intact cells were resuspended in defined medium at a density of 6×10⁶ cells per ml. Oxygen consumption was measured over 5 minutes in an equilibrated Clark-type oxygen electrode thermostatically maintained at 37°C. All data were obtained using a PowerLab system with Chart recording software.

Statistical analysis was performed using GraphPad Prism, version 4.0c for Macintosh (GraphPad). Experiments were analysed using either the Man-Whitney tests or one-way ANOVA followed by Tukey's post-hoc test.

Protein extraction and western blotting were carried out as previously described (Lloyd et al., 1997). Blots were analysed using antibodies to prohibitin, cytochrome c, p-MARCKS, p-Akt, Akt, p-S6 kinase (Cell Signaling Technology), p-AMPK (Upstate) and p-ERK and ERK (Sigma). Proteins were visualised using ECL plus (Amersham).

Cells were fixed in 4% paraformaldehyde and permeabilised with 0.5% Triton X-100 before staining with primary antibodies to ERRα (Abcam; 1:250) or prohibitin (Abcam; 1:250).

Total RNA (5 μg) was prepared using Qiagen RNeasy Plus kit. Labelling and hybridisation was performed at the CRUK Molecular Biology Core Facility at the Paterson Institute for Cancer Research, Manchester, UK. Data were analysed using the open source bioinformatics platform Bioconductor 2.1, running on R 2.6.0. Probeset expression measures were calculated using the Robust Multichip Average (RMA) default method (Affymetrix). Differential gene expression was assessed between replicate groups using an empirical Bayes' t-test as implemented in the `limma' package (Smyth, 2004). Any Probesets with a P-value ≤0.006 and absolute fold change ≥1.23 were categorised as differentially expressed. All data were deposited in the GEO database (id GSE 10775). For the GO analysis, a standard hypergeometric test from Bioconductor's `GOstats' package was used to test for over-representation of gene ontology (GO) terms associated with each gene list against a background of all genes on the microarray platform (Falcon and Gentleman, 2007).

Total RNA was obtained using the RNeasy Plus Minikit (Qiagen) and equal amounts of RNA reverse transcribed with Superscript II reverse transcriptase (Invitrogen). Quantitative PCR was performed using the Dynamo SYBR Green qPCR kit (Finnzymes) and the Opticon 2 DNA engine (MJ Research). Gene-specific primer pairs for cytochrome c (forward 5′-GGAGGCAAGCATAAGACTG-3′, reverse 5′-GTCTGCCCTTTCTCCCTTCT-3′), TFAM (forward 5′-GAAAGCACAAATCAAGAGG-3′, reverse 5′-CTGCTTTTCATCATGAGACA-3′), ERRα (forward 5′-TTGAAGATGCTGAGGCTGTG-3′, reverse 5′-CCAGCTTCACCCCATAGAAA-3′), prohibitin (forward 5′-GGAGTCGAGGTGAACTCTGC-3′, reverse 5′-TCGAGAGCAGCAGTCAAAGA-3′), PRC (forward 5′-TGGGATCCTGAAGGAATCTG-3′, reverse 5′-TCCTCCTTGAGTGAGCTGGT-3′), PGC-1 (forward 5′-ATGTGTCGCCTTCTTGCTCT-3′, reverse 5′-ATCTACTGCCTGGGGACCTT-3′), IDH3 (forward 5′-GGAACAACCACAGGAGCAAT-3′, reverse 5′-GAACTGGGACGGGTCTTGTA-3′) were designed using the Primer3 algorithm (Rozen and Skaletsky, 2000). All experiments were performed in duplicate and each experiment shown is representative of three similar experiments.

Small interfering RNA duplexes were prepared according to manufacturer's instructions (Qiagen) and specific knockdown conditions were optimised using the HiPerfect transfection reagent (Qiagen). A final concentration of 1 nM of duplex was used to achieve knockdown as assessed by quantitative RT-PCR. Target sequences for duplexes against PRC were 5′-TAACGAGTATTGAACAAATTA-3′ (duplex 1) and CAGGAGATCTGTGCTAGTTA (duplex 2). For ERRα the target sequences were 5′-AAGCTAGTGCTCAGCTCTCTA-3′ (duplex 1) and 5′-CTCTGTGATCTGTTTGATAGA-3′ (duplex 2).