1. Cell Biology
  2. Human Biology and Medicine
Download icon

Mitochondrial biogenesis is transcriptionally repressed in lysosomal lipid storage diseases

  1. King Faisal Yambire
  2. Lorena Fernandez-Mosquera
  3. Robert Steinfeld
  4. Christiane Mühle
  5. Elina Ikonen
  6. Ira Milosevic
  7. Nuno Raimundo  Is a corresponding author
  1. University Medical Center Goettingen, Germany
  2. International Max-Planck Research School in Neuroscience, Germany
  3. Friedrich-Alexander University Erlangen-Nürnberg (FAU), Germany
  4. University of Helsinki, Biomedicum Helsinki, Finland
Research Communication
  • Cited 1
  • Views 1,843
  • Annotations
Cite this article as: eLife 2019;8:e39598 doi: 10.7554/eLife.39598

Abstract

Perturbations in mitochondrial function and homeostasis are pervasive in lysosomal storage diseases, but the underlying mechanisms remain unknown. Here, we report a transcriptional program that represses mitochondrial biogenesis and function in lysosomal storage diseases Niemann-Pick type C (NPC) and acid sphingomyelinase deficiency (ASM), in patient cells and mouse tissues. This mechanism is mediated by the transcription factors KLF2 and ETV1, which are both induced in NPC and ASM patient cells. Mitochondrial biogenesis and function defects in these cells are rescued by the silencing of KLF2 or ETV1. Increased ETV1 expression is regulated by KLF2, while the increase of KLF2 protein levels in NPC and ASM stems from impaired signaling downstream sphingosine-1-phosphate receptor 1 (S1PR1), which normally represses KLF2. In patient cells, S1PR1 is barely detectable at the plasma membrane and thus unable to repress KLF2. This manuscript provides a mechanistic pathway for the prevalent mitochondrial defects in lysosomal storage diseases.

Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that all the issues have been addressed (see decision letter).

https://doi.org/10.7554/eLife.39598.001

Introduction

Lysosomal storage diseases are a group of severe diseases caused by mutations in genes encoding for lysosomal proteins, and are referred to as storage diseases because one common phenotype is the accumulation of undigested substrates in the lysosomes, with the consequent enlargement and loss of function of the organelle (Parenti et al., 2015). The lysosomes have far-reaching roles beyond the ‘recycling bin’ paradigm, and are key players in nutrient sensing and metabolic regulation (Ballabio, 2016; Lim and Zoncu, 2016; Settembre et al., 2013). Furthermore, lysosomes are essential for the process of macroautophagy, and thus for the selective autophagy of mitochondria, the main mechanism to degrade dysfunctional mitochondria (Pickles et al., 2018). Mitochondrial perturbations have been widely reported in several lysosomal storage diseases (Platt et al., 2012; Plotegher and Duchen, 2017), including neuronal ceroid lipofuscinosis, Gaucher and Niemann-Pick diseases (Jolly et al., 2002; Lim et al., 2015; Osellame et al., 2013; Torres et al., 2017; Woś et al., 2016). Nevertheless, it remains unclear why mitochondrial dysfunction is so prevalent in lysosomal storage diseases.

In this study, we focus on two lysosomal storage diseases, Niemann-Pick type C (NPC) and acid sphingomyelinase (ASM) deficiency. NPC is caused by mutations in the gene NPC1 or, less commonly, NPC2 (Patterson and Walkley, 2017; Schuchman and Wasserstein, 2016). NPC1 and NPC2 encode proteins involved in sphingomyelin and cholesterol efflux from the lysosome (Platt, 2014). ASM deficiency, also known as Niemann-Pick A/B, is caused by mutations in the gene SMPD1 encoding acid sphingomyelinase. ASM catalyzes the breakdown of sphingomyelin into ceramide and phosphorylcholine (Schuchman and Wasserstein, 2016). Interestingly, accumulation of cholesterol, sphingosine, sphingomyelin and glycosphingolipids in the lysosomes are observed both in Niemann-Pick and ASM deficiency cells and tissues (Leventhal et al., 2001; Vanier, 1983).

The NPC1 knock-out mouse (NPC1 KO) and a knock-in of the most common NPC1 patient mutation I1061T (Praggastis et al., 2015) are established models of Niemann-Pick type C disease (Loftus et al., 1997). Both NPC1 KO and NPC1I1061T mice recapitulate most of the neuropathological phenotypes of the disease, with the disease onset occurring earlier in the NPC1 KO mice. The ASM knock-out mouse (ASM KO) is a widely used model of ASM deficiency (Horinouchi et al., 1995).

Mitochondria are fundamental metabolic organelles in the cell, harboring key pathways for aerobic metabolism such as the citrate cycle, the key integrator metabolic pathway, as well as the respiratory chain and oxidative phosphorylation, Fe-S cluster and heme synthesis (Pagliarini and Rutter, 2013). They are also recognized as a major cellular signaling platform, with far-reaching implications on cell proliferation, stem cell maintenance, cellular immunity and cell death (Kasahara and Scorrano, 2014; Raimundo, 2014). Mitochondria are composed of about 1000 proteins, of which only 13 are encoded by mitochondrial DNA (mtDNA) (Pagliarini et al., 2008). The other ~1000 proteins are encoded by nuclear genes, and imported to the different sub-mitochondrial compartments (e.g., matrix, inner membrane, outer membrane, intermembrane space) by dedicated pathways (Wiedemann and Pfanner, 2017).

The large number of proteins that are nuclear-encoded and imported to mitochondria imply the need for regulatory steps that ensure the coordination of the process of mitochondrial biogenesis. This is often regulated at transcript level, by transcription factors that promote the expression of nuclear genes encoding for mitochondrial proteins (Scarpulla et al., 2012). One of the best characterized is the nuclear respiratory factor 1 (NRF1), which stimulates the expression of many subunits of the respiratory chain and oxidative phosphorylation, and also of genes necessary for mtDNA maintenance and expression, such as TFAM (Evans and Scarpulla, 1989; Evans and Scarpulla, 1990). Other transcription factors, such as estrongen-related receptor α (ERRα) and the oncogene myc, also act as positive regulators of mitochondrial biogenesis (Herzog et al., 2006; Li et al., 2005). Several co-activators also participate in the regulation of mitochondrial biogenesis, of which the co-activator PGC1α (peroxisome proliferator-activated receptor-gamma, co-activator 1 α) is the best characterized (Wu et al., 1999). PGC1a can interact with NRF1 or ERRα and stimulate mitochondrial biogenesis (Scarpulla et al., 2012). No transcriptional repressors of mitochondrial biogenesis have so far been described. Impaired or uncoordinated mitochondrial biogenesis often results in impaired mitochondria leading to pathological consequences (Cotney et al., 2009; Raimundo et al., 2012).

Here, we identify the transcription factors KLF2 and ETV1 as transcriptional repressors of mitochondrial biogenesis. The up-regulation of these two proteins in patient cells and mouse tissues of two lysosomal diseases, Niemann-Pick type C and ASM deficiency, underlies the mitochondrial defects observed in these syndromes. The silencing of ETV1 and, particularly, KLF2, is sufficient to return mitochondrial biogenesis and function to control levels.

Results

Expression of mitochondria-related genes is decreased in NPC1 KO tissues

Mitochondrial homeostasis and function is impaired in many lysosomal storage diseases. The two main axes of mitochondrial homeostasis are biogenesis and demise (by selective autophagy, designated mitophagy). Given that lysosomal diseases are characterized by impaired autophagy (Settembre et al., 2008), it is expectable that mitophagy is also impaired. However, it remains unknown how mitochondrial biogenesis is affected in lysosomal storage diseases.

To assess mitochondrial biogenesis at transcript level in a systematic manner, we resorted to a publicly-available transcriptome dataset of NPC1 KO mice liver and brain, the two tissues most affected in Niemann-Pick type C. The dataset included both pre-symptomatic and symptomatic animals (Alam et al., 2012). To monitor the effects of Niemann-Pick disease on transcriptional regulation of mitochondrial biogenesis, we started by establishing a comprehensive list of mitochondria-related genes. We used a published mitochondrial proteome (MitoCarta, (Pagliarini et al., 2008) see Materials and methods for details), and converted the protein names to the corresponding ENSEMBL gene name to generate the ‘mitochondria-associated gene list’. The process is illustrated in Figure 1A. We prepared a second list which included only the respiratory chain and oxidative phosphorylation subunits (‘RC/OXPHOS gene list’). As controls, we prepared ‘gene lists’ for lysosomes, peroxisomes, Golgi and endoplasmic reticulum using the same strategy. The proteomes used to build the organelle-specific gene lists are detailed in the methods section (Table 1). Next, we used transcriptome data from asymptomatic and symptomatic brain and liver of NPC1 KO and corresponding WT littermates to determine how the organelle gene lists were affected.

Table 1
Sources of organelle-specific proteomes.
https://doi.org/10.7554/eLife.39598.002
DatasetNumber of genesReference (source)
Mitochondria1049(Pagliarini et al., 2008)
Respiratory chain subunits108(Pagliarini et al., 2008)
Lysosomes435(Skon et al., 2013)
Peroxisomes254(Hollenhorst et al., 2007)
Endoplasmic reticulum297(Herzog et al., 2006)
Golgi (COP I) Vesicles86(Dugas et al., 2006)
Figure 1 with 2 supplements see all
Mitochondrial genes are down-regulated in brain and liver of symptomatic NPC1 KO mice.

(a) Schematic representation of the in silico approach. The list of mitochondria-related genes was built by converting the MitoCarta proteome inventory into a transcript list. This was then crossed with the differentially-expressed gene list of brain and liver in symptomatic and asymptomatic NPC1 KO (n = 11; brain, n = 6; liver) versus WT (n = 5; brain, n = 6; liver) mice. (b–c) Decreased expression of genes encoding ~1000 mitochondrial proteins (panel b) and ~100 respiratory chain subunits (panel c) in brain and liver of NPC1 KO mice. The plots (b–c) represent the variation in gene expression comparing the fold change between the average expression in NPC1 KO over NPC1 WT. Error bars denote standard error of the mean (s.e.m.). This variation is represented as the difference from average WT expression (e.g., a 20% increase in the mutant mice is shown as 20%, while −25% denotes a 25% decrease). Statistical analyses using t-test with Bonferroni correction, adjusted p-values ***p<0.001.

https://doi.org/10.7554/eLife.39598.003

First, we assessed the average expression of lysosomal genes in NPC1 KO brain and liver, to verify the validity of our ‘organelle gene list’ approach in this dataset. We have shown earlier that the average expression level of an organelle-gene list is a good indicator of the activity of the transcriptional program of biogenesis for that organelle (Fernández-Mosquera et al., 2017). The average expression of lysosomal genes was significantly increased in the asymptomatic NPC1 KO brain and liver (Figure 1—figure supplement 1A), and increased further with the onset of the disease in NPC1 KO brain and liver (Figure 1—figure supplement 1A), in agreement with the expected increase in the expression of lysosomal genes in lysosomal storage diseases.

Then, we measured the average expression of the ‘mitochondrial gene list’ in NPC1 KO brain and liver. Mitochondria-associated genes were up-regulated in pre-symptomatic NPC1 KO brain, and down-regulated in symptomatic brain (Figure 1B). In the liver, the average expression of mitochondria-associated genes was not significantly changed in the pre-symptomatic group, but was robustly decreased in the symptomatic NPC1 KO mice (Figure 1B). When looking only at the ‘RC/OXPHOS gene list’, the pattern was similar but the magnitude of the changes was more robust (Figure 1C). These results are not due to a small number of genes skewing the whole population, since the proportion of mitochondrial genes in the differentially expressed gene lists for NPC1 KO brain (Figure 1—figure supplement 2A–C) and liver (Figure 1—figure supplement 2D–F) increases robustly (about 5-fold) with disease onset. These results highlight a general trend towards a global down-regulation of mitochondrial genes under chronic lysosomal malfunction.

In order to determine if this effect was specific to mitochondria or also observed in other organelles, we tested how the average expression of peroxisomal-, endoplasmic reticulum- and Golgi-specific genes was affected. The expression of peroxisomal genes was not affected in NPC1 KO brain, but was down-regulated in both asymptomatic and symptomatic NPC1 KO liver (Figure 1—figure supplement 1B). The expression of endoplasmic reticulum-related and Golgi-related genes was not significantly altered (Figure 1—figure supplement 1B). These results suggest that lysosomal stress caused by absence of Npc1 in multiple tissues specifically affects the expression of mitochondrial genes, although disease onset also results in a liver-specific repression of peroxisomal genes.

Mitochondrial biogenesis and function are impaired in NPC and ASM patient cells and tissues

To verify the results from the large-scale transcriptional analysis of NPC1 KO tissues, we tested the expression of several genes encoding for mitochondrial proteins in the livers of NPC1 KO mice. The genes tested encode for subunits of the respiratory chain complex I (NDUFS3 and ND6), complex II (SDHA), complex III (CYTB) and complex IV (COX5A, COX1). ND6, CYTB and COX1 are encoded by mtDNA, while all the others are nuclear-encoded. We observed a robust and consistent decrease in the transcript levels of mitochondria-related genes in the livers of NPC1 KO mice (Figure 2A) compared to their respective WT littermates. A similar reduction on the expression of mitochondria-associated genes was also observed in NPC patient fibroblasts (Figure 2B) whose lysosomal phenotype has already been characterized (Park et al., 2003).

Impaired mitochondrial biogenesis and function in mouse and cellular models of Niemann-Pick disease.

The transcript levels of several nuclear-encoded and mitochondrial DNA (mtDNA)-encoded mitochondria-related genes were measured. (a) transcript levels of mitochondria-related genes are decreased in the liver of NPC1 knockout mice (NPC1 KO), a model of Niemann-Pick type C. The plot shows mean ± s.e.m. T-test p-values ***p<0.001, n = 9 (b) transcript levels of mitochondria-related genes are decreased in the fibroblasts of a patient with compound heterozygote NPC1 mutations (GM18398 Coriell Repository). The plot shows mean ± s.e.m. T-test p-values *p<0.05 **p<0.01 ***p<0.001, n = 3 (c) transcript levels of mitochondria-related genes are decreased in the liver of acid sphingomyelinase knockout (ASM KO) mice, a model of acid sphingomyelinase deficiency. The plot shows mean ± s.e.m. T-test p-values *p<0.05 **p<0.01, n = 8. (d) transcript levels of mitochondria-related genes are decreased in fibroblasts from a patient with acid sphingomyelinase deficiency (only 5% of ASM activity left) and in the ASM-2 patient line. The plot shows mean ± s.e.m. T-test p-values *p<0.05 **p<0.01 ***p<0.001, n = 3. Further characterization of the lysosomal defects in the fibroblasts of this patient are presented in Figure 3—figure supplement 1. (e–f) mitochondrial superoxide levels, as assessed by the fluorescence intensity of the superoxide-sensitive mitochondria-targeted dye MitoSox, measured by flow cytometry, are increased in NPC fibroblasts (panel e) and in ASM-1 and ASM-2 patient fibroblasts (panel f); histogram plots are representative of three biological replicates. Quantifications denote mean ± s.e.m..T-test p-values ***p<0.001, n = 3.

https://doi.org/10.7554/eLife.39598.006

The accumulation of cholesterol and sphingomyelin in the lysosomes is common to both NPC and acid shingomyelinase (ASM) deficiency (Pentchev et al., 1984Reagan et al., 2000; Leventhal et al., 2001; Herzog et al., 2006; Lloyd-Evans et al., 2008; Suzuki et al., 2012; Skon et al., 2013; Platt, 2014). However, while mitochondria in NPC also present increased levels of cholesterol, this does not happen in ASM deficiency (Torres et al., 2017). Since excessive mitochondrial cholesterol can impair mitochondrial function (Torres et al., 2017), we tested if ASM deficiency would also have a repressive effect on mitochondrial biogenesis. Similar to the NPC findings, we observed a decrease in the expression of mitochondria-associated genes in the ASM KO liver compared to the WT littermates (Figure 2C) as well as in two different patient fibroblasts of ASM deficiency (Figure 2D).

To assess if this down-regulation of mitochondrial biogenesis in NPC and ASM deficiency had functional consequences for respiratory chain efficiency, we measured the amounts of mitochondrial superoxide, a by-product of the mitochondrial respiratory chain known to be produced in higher amounts when mitochondria are not functioning optimally (Raimundo et al., 2012; Raimundo, 2014), which can be estimated using a superoxide-sensitive mitochondria-targeted dye, MitoSox. We observed an increase in MitoSox intensity in patient fibroblasts with NPC (Figure 2E) and ASM deficiency (Figure 2F) denoting increased superoxide levels which are indicative of poor mitochondrial performance. Altogether, these results show that the biogenesis of mitochondria is repressed in NPC- and ASM-deficient cells and tissues, and that the existing mitochondria are not functioning optimally. Furthermore, the mitochondrial impairments are likely unrelated to the levels of cholesterol in mitochondria (known to be high in NPC but normal in ASM; Torres et al., 2017), and seem rather a consequence of the lysosomal saturation in NPC and ASM deficiency.

Impaired mitochondrial respiration in NPC1 and ASM deficiency

To further characterize the impact of lysosomal disease on mitochondrial function, we focused on the ASM-deficient fibroblasts, which showed a more robust decrease of mitochondrial biogenesis than NPC and do not have the confounding factor of excessive mitochondrial cholesterol. We used cells from two patients of ASM deficiency, one of which (ASM-2) had the lysosomal phenotype already characterized (Corcelle-Termeau et al., 2016). Additionally, we also employed a line from a patient (ASM-1) with compound heterozygous loss-of-function mutations in SMPD1 (the gene encoding ASM), which has severe ASM deficiency (5% activity left). The lysosomal impairments in this line have not yet been characterized besides patient diagnosis; therefore, we first evaluated lysosomal function in these fibroblasts. One of the consequences of lysosomal dysfunction is the accumulation of autophagic substrates, such as the protein p62 (also known as Sequestosome 1, SQSTM1) as well as autophagosomes (Settembre et al., 2008). We assessed the levels of p62/SQSTM1 and LC3B-II, a marker of autophagosomal mass, by Western blot, and found both sharply increased in the ASM-1 fibroblasts, as expected (Figure 3—figure supplement 1A). We also assessed the lysosomal proteolytic capacity, by measuring the degradation of the lysosomal substrate DQ-BSA. DQ-BSA is a polymer of fluorescently-tagged bovine serum albumin, which accumulates in the lysosomes. The fluorescence is quenched in the polymeric form and detectable in the monomers. As the lysosomal proteases start cleaving DQ-BSA and releasing monomers, fluorescence starts increasing, and the rate of fluorescence increase is proportional to the activity of lysosomal proteases. We observed a strong decrease in DQ-BSA degradation rate in the ASM-1 fibroblasts (Figure 3—figure supplement 1B). These results support a strong impairment of lysosomal function in ASM-1 cells used in this study, in line with the cellular phenotype of the disease and the described phenotype of ASM-2.

We then set to characterize mitochondrial function. First, we monitored the oxygen consumption rate (OCR). This was done with a high-throughput real-time respirometer, which allows the measurement under multiple conditions, such as basal medium, inhibition of oxidative phosphorylation (when OCR is inhibited) and uncoupled respiratory chain (when OCR occurs unrestrained). We observed a robust decrease in OCR in ASM-1 fibroblasts which lasted across all conditions tested: basal medium, inhibition of the oxidative phosphorylation with oligomycin, and uncoupling of respiratory chain and oxidative phosphorylation by FCCP (Figure 3A). We determined that the ASM-1 fibroblasts have ~70% decrease in the OCR compared to the control cells in basal conditions and in maximal (uncoupled) conditions (Figure 3B). We also monitored the OCR in ASM-2 and NPC fibroblasts (Figure 3C) and observed that they also presented a robust decrease in OCR both at basal (~50% down) and maximal conditions (Figure 3D). Importantly, the expression of wild-type ASM in ASM-1 and ASM-2 fibroblasts increased OCR significantly, as did expression of wild-type NPC1 in NPC1-deficient fibroblasts (Figure 3—figure supplement 1C–D). Furthermore, introduction of wild-type ASM relieved the inhibition in the expression of mitochondrial genes, denoted by the increase in their transcript levels (Figure 3—figure supplement 1E). Expression of wild-type NPC1 in NPC1- deficient fibroblasts had a similar result (Figure 3—figure supplement 1E). These results show that the robust decrease in mitochondrial biogenesis and respiration observed in ASM- and NPC1-deficient fibroblasts are a specific consequence of the loss of ASM or NPC1 activity, respectively.

Figure 3 with 4 supplements see all
Mitochondrial function and mitochondrial mass are impaired in acid sphingomyelinase (ASM)- and NPC1-deficient patient fibroblasts.

(a,c) ASM- and NPC1-deficient fibroblasts have substantially lower O2 Consumption Rate (OCR) than controls. OCR was measured using whole cells, sequentially in basal conditions (complete medium), after oxidative phosphorylation inhibition using the ATPase inhibitor oligomycin, after uncouling the respiratory chain from oxidative phosphorylation using the uncoupler FCCP, and after inhibition of the respiratory chain using complex I inhibitor rotenone and complex III inhibitor antimycin. The measurements were made in a 96-well plate using a SeaHorse Extracellular Flux analyser. The mean ± s.e.m. of at least eight wells per cell line is plotted over time. OCR was normalized to the amount of protein in each well. (b,d) Reduced basal and maximal (uncoupled) OCR in ASM1-deficient fibroblasts quantified from the curves in (a) and in ASM2- and NPC1- deficient fibroblasts quantified from profiles in (c) and bar graphs are presented as mean ± s.e.m. T-test p-value ***p<0.001, n = 3.

https://doi.org/10.7554/eLife.39598.007

We then tested if the decrease in mitochondrial biogenesis in ASM- and NPC1-deficient fibroblasts resulted in decreased mitochondrial mass. We stained the cells with a dye that specifically targets mitochondria independently of the mitochondrial inner membrane potential, Mitotracker green, and measured the intensity of this dye in the different lines. We observed that there was increased Mitotracker green signal in the three lines with lysosomal defects (Figure 3—figure supplement 2A), suggesting increased mitochondrial mass. We then assessed the protein levels of respiratory chain proteins. Despite their transcripts were repressed, we did not observe a similar reduction in the protein levels (Figure 3—figure supplement 2B). These results suggest that despite the transcriptional repression of mitochondrial biogenesis, there is accumulation of mitochondria, likely due to the impact of impaired lysosomal function on mitophagy. This would result in the accumulation of damaged (e.g. uncoupled) mitochondria in the cytoplasm, which would normally be removed in cells with functioning lysosomes/autophagy pathway. Thus, we stained the cells with two dyes that accumulate specifically in mitochondria, Mitotracker green (independent of mitochondrial inner membrane potential) and Mitotracker red (potential-dependent mitochondrial accumulation), and analyzed the intensity of the signals by flow cytometry. We quantified the proportion of cells with low-potential mitochondria (essentially, a simultaneous decrease in the red intensity, and increase in the green intensity). This proportion is increased in all lysosomal patient cells tested (Figure 3—figure supplement 2C). Control cells treated with the mitochondrial uncoupler CCCP were used as positive control for this assay, and show a massive increase in the proportion of cells with uncoupled mitochondria. Therefore, the ASM- and NPC1-deficient cells present accumulation of dysfunctional mitochondria in the cytoplasm, despite the repression in mitochondrial biogenesis at transcript level. These results are further underscored by the decrease in mitochondrial respiratory activity and the increase in superoxide levels in the ASM- and NPC1-deficient cells, as shown above (Figure 2E–F and Figure 3).

Acid sphingomyelinase generates ceramide, which is itself a powerful signaling lipid, and can be metabolized by acid ceramidase into sphingosine and other signaling lipids. Notably, desipramine inhibits both acid sphingomyelinase and acid ceramidase (Herzog et al., 2006). We then tested if simultaneous pharmacological blockage of ASM and acid ceramidase with desipramine yields the same mitochondrial phenotypes as observed in the ASM patient cells. In the desipramine-treated cells we observed decreased mitochondrial biogenesis (lower transcript levels of genes encoding mitochondrial proteins, Figure 3—figure supplement 3A) and accumulation of damaged mitochondria, as shown by increased superoxide levels (Figure 3—figure supplement 3B) and decreased respiration (Figure 3—figure supplement 3C–D). Thus, pharmacological inhibition of both ASM and acid ceramidase yielded similar results to ASM patient cells, suggesting that acid ceramidase is not relevant for the phenotypes observed.

Since one of the known consequences of ASM deficiency is accumulation of cholesterol in the lysosomes (Lloyd-Evans et al., 2008; Yeo et al., 2014), and given that we observed similar perturbations on mitochondrial homeostasis in ASM- and NPC1-deficient patient fibroblasts, we tested if pharmacological inhibition of NPC1 would also be sufficient to impact mitochondrial biogenesis and function. We treated control cells with the NPC1 inhibitor U18666A (West et al., 2015), and observed decreased expression of mitochondria-associated genes (Figure 3—figure supplement 4A), and increased mitochondrial superoxide levels (Figure 3—figure supplement 4B). Finally, treatment with U18666A resulted in lower respiration with ~30% lower basal OCR and ~50% lower uncoupled OCR (Figure 3—figure supplement 4C–D). Thus, pharmacological inhibition of ASM or NPC1, similar to genetic defects in these proteins, is sufficient to cause decreased expression of mitochondrial genes and impaired mitochondrial respiratory chain activity.

KLF2 and ETV1 are up-regulated in NPC1 KO tissues and repress transcription of mitochondria-associated genes

Having established a clear mitochondrial phenotype in NPC1 and ASM deficiency, we set out to identify the underlying mechanism. The robust decrease in the expression of hundreds of mitochondria-related genes in NPC1 KO brain (Figure 1—figure supplement 2B–C) and NPC1 KO liver (Figure 1—figure supplement 2E–F) suggests the involvement of a coordinated transcriptional program, and therefore of transcriptional regulators such as transcription factors. To determine which transcription factors might be mediating the repression of mitochondria-associated genes, we took an unbiased bottom-up approach to determine potential transcriptional regulators. Given that the whole mitochondrial gene list has ~1000 genes, we focused on the RC/OXPHOS list, which shows the same behavior as the complete mitochondrial gene list (as shown in Figure 1) and has a more manageable size (~100 genes). Using the Genomatix Gene2Promoter tool, we obtained the genomic sequences (Mus musculus) of the promoter regions of the RC/OXPHOS genes, from −500 base pairs upstream the transcription start site, to +100 base pairs downstream. This region is sufficient to account for the regulation of gene expression by transcription factors in many promoters of mitochondrial genes (Gleyzer et al., 2005; Virbasius and Scarpulla, 1994). We then used Genomatix Matinspector tool to analyze the gene promoters for transcription factor binding sites (cis-elements), and identified those statistically enriched (illustrated in Figure 4A). The most overrepresented cis-elements in the promoters of RC/OXPHOS genes were the transcription factor families SP1, E2F, Krueppel-like factors (KLF) and ETS factors (Table 2). In parallel, as control, we carried out a similar approach for the lysosomal gene list (whose expression is increased, in contrast to the mitochondrial genes) and observed that the SP1 and E2F families were also significantly enriched in the promoters of lysosomal genes (Supplementary file 1). Given that the expression of lysosomal genes and mitochondrial genes is affected in opposite ways, we reasoned that it would be unlikely that the same transcription factors were driving two opposite processes. For this reason, we proceeded only with the KLF and ETS families, which only scored as significantly enriched in the mitochondrial promoters (Table 2).

Table 2
Transcription factors with statistically enriched cis-elements in the promoters of genes encoding for subunits of mitochondrial respiratory chain.
https://doi.org/10.7554/eLife.39598.012
Transcription factor familyp-value
(Fisher's exact test)
Transcription factors
 SP11.52E-09SP1, SP4
 E2F2.79E-08E2F1, E2F2, E2F3, E2F4
 KLF0.000265KLF2, KLF6, KLF7, KLF15
 ETS0.000796ELK1, SPI1, ETV1
Figure 4 with 5 supplements see all
Transcription factors Etv1 and Klf2 are induced in Niemann-Pick and involved in the regulation of mitochondrial biogenesis.

(a) Venn diagram illustrating the intersection between the list of transcription factors (TFs) that are significantly activated or repressed in tissues of Npc1-/- mice, and the list of TFs that are predicted to regulate the expression of mitochondrial respiratory chain genes, which yields KLF2 and ETV1 as hits. (b) Increased Klf2 and Etv1 protein levels in ASM-deficient and NPC fibroblasts, shown in a representative (out of three biologically independent experiments) western blot of whole cell extracts, with quantification (mean ± s.e.m., n = 3) of band densities in the adjacent plot. T-test p-values **p<0.01 (c) Increased nuclear localization of Klf2 and Etv1 in ASM-deficient fibroblasts. Blots are representative of biological triplicates with quantifications in (d) shown as mean ± s.e.m. T-test p-values *p<0.05, **p<0.01 and ***p<0.001 (e) Overexpression of ETV1WT (full length ETV1) and of ETV11-334, lacking the C-terminus which includes the DNA-binding domain. Representative western blot, quantification of band densities normalized to empty vector control from two independent experiments with two technical replicates each on the right panel (mean ± s.e.m.) (f) Overexpression of ETV1WT significantly down-regulates the transcript levels of most mitochondria-related genes, while ETV11-334, unable to bind DNA, causes an increase in transcript levels. The plots show mean ± s.e.m. T-test p-values **p<0.01 ***p<0.001, n = 2 with three technical replicates each.

https://doi.org/10.7554/eLife.39598.013

Next, we again resorted to the transcriptome dataset of NPC1 KO brain and liver to determine if any transcription factors in the KLF2 and ETS families were predicted to have increased or decreased activity during NPC disease progression. Using Ingenuity Pathway Analysis, we determined which transcription factors scored as significant regulators in these tissues (Supplementary file 2). The only transcription factor of the KLF family meeting the criteria was KLF2. Several ETS family transcription factors have redundant binding sites (Hollenhorst et al., 2007), so we tested the three members that scored in the Genomatix promoter analysis, SPI1, ELK1 and ETV1. SPI1 is expressed in macrophages and not expressed in fibroblasts (Feng et al., 2008; Suzuki et al., 2012), and accordingly we could not detect the expression of SPI1 in control or patient fibroblasts, either at transcript or protein (data not shown). While ELK1 was not changed at transcript level (Figure 4—figure supplement 1A), ETV1 was significantly increased in ASM deficiency patient fibroblasts (Figure 4—figure supplement 1A). The transcript levels of KLF2 were not changed in ASM deficiency (Figure 4—figure supplement 1B).

We then focused on KLF2 and ETV1 (Figure 4A). First, we tested if the levels of these proteins were affected in NPC1- or ASM-deficient fibroblasts, by Western blotting. We found that both KLF2 and ETV1 were robustly up-regulated in both ASM patient lines (Figure 4B). In the NPC1- deficient cells, KLF2 was robustly increased, but ETV1 was not significantly changed (Figure 4B). Given that many transcription factors shuttle between the nucleus and the cytoplasm, we prepared nuclear extracts to verify if KLF2 and ETV1 were enriched in the nucleus of the patient cells. We observed that there was a clear increase in nuclear KLF2 and ETV1 in ASM-1 cells (Figure 4C). Similarly, ASM-2 and NPC1-deficient cells also had increased nuclear KLF2 (Figure 4—figure supplement 1C) and ETV1 (Figure 4—figure supplement 1D). Thus, KLF2 and ETV1 are likely more active both in ASM- and NPC1-deficient cells.

We again compared the ASM-deficient fibroblasts with control fibroblasts treated with the inhibitor of both ASM and acid ceramidase. Desipramine-treated fibroblasts yielded a similar result: both KLF2 and ETV1 are up-regulated at protein level (Figure 4—figure supplement 2A, quantified in Figure 4—figure supplement 2B) but only ETV1 transcript levels are significantly changed (Figure 4—figure supplement 2C). Altogether, these results suggest that the accumulation of KLF2 in response to lysosomal lipid storage is regulated post-translationally, while ETV1 is regulated at transcript level. The nuclear localization of both transcription factors is likely another regulatory step for KLF2 and ETV1 in ASM- and NPC1-deficient cells.

Given that ETV1 and KLF2 are predicted by our promoter analysis to have binding sites in the promoters of the genes encoding for respiratory chain subunits, and that increased expression of these two transcription factors correlates with repression of respiratory chain genes, we reasoned that KLF2 and ETV1 might be mediating this repression. To explore this possibility, we took advantage of another publicly available transcriptome dataset of erythroid cells of KLF2 KO and WT mice (GSE27602) (Redmond et al., 2011). We observed an increase in the average transcript levels of the ‘mitochondria gene list’ in the KLF2 KO cells compared to the WT littermates (Figure 4—figure supplement 3). The effect is also observed, with higher magnitude, when measuring the average expression of the genes encoding for respiratory chain subunits (Figure 4—figure supplement 3). These results suggest that KLF2 is able to repress mitochondrial biogenesis in vivo. This effect is likely direct, since analysis of a KLF2 ChIP-Seq dataset (Yeo et al., 2014) reveals a large number of target genes encoding mitochondrial proteins, including several respiratory chain subunits (Figure 4—figure supplement 4), in agreement with our in silico promoter analysis. Notably, ETV1 and other transcription factors regulating mitochondrial biogenesis, such as NRF1, were also identified as KLF2 transcriptional targets in the same dataset (Figure 4—figure supplement 4).

In addition, it is noteworthy that several known ETV1 targets are mitochondrial genes, as previously shown by chromatin immunoprecipitation (Baena et al., 2013) and illustrated in Figure 4—figure supplement 5. To test if the effect of ETV1 on the expression of mitochondria-related genes is direct, we expressed full length ETV1 (ETV1FL) as well as ETV1 lacking the DNA-binding domain (ETV11-334) in control fibroblasts (Figure 4E) and evaluated the effect on the expression of mitochondria-related genes. The overexpression of ETV1FL elicited a decrease in the transcript levels of most mitochondria-associated genes (Figure 4F). However, ETV11-334 did not repress the transcript levels of these mitochondrial-related genes (Figure 4F). This result is coherent with the role of ETV1 as a repressor of mitochondrial biogenesis, and further demonstrates that this repression occurs via direct binding of ETV1 to DNA (Janknecht, 1996), thus validating our in silico promoter analysis. The unexpected increase in the transcript levels of mitochondria-related genes under overexpression of ETV11-334, unable to bind DNA, may be explained by ETV1 functioning as a homodimer (Poon, 2012). Therefore, overexpression of a mutant unable to bind DNA might titrate out the wild-type ETV1, thus effectively functioning as a dominant-negative ETV1 isoform, with the consequent activation of mitochondrial biogenesis.

Silencing of KLF2 and ETV1 in ASM- and NPC1-deficiency rescues mitochondrial biogenesis and function

To test if KLF2 and ETV1 were indeed repressing mitochondrial biogenesis in ASM-deficient cells, we knocked-down ETV1 (Figure 5A) and KLF2 (Figure 5B), independently, in ASM-deficient fibroblasts. Given that the ASM-1 and ASM-2 fibroblasts had the same mitochondrial phenotype, and showed similar patterns of KLF2 and ETV1 behavior, at this point we focused on ASM-1, which had a slightly more robust effect.

Figure 5 with 1 supplement see all
Silencing of ETV1 or KLF2 rescues mitochondrial biogenesis and function in Niemann-Pick fibroblasts.

Using siRNA-mediated silencing, we knocked-down Etv1 (a) or Klf2 (b) in ASM1-deficient fibroblasts, which brought the protein levels of mitochondrial protein TFAM, and of mitochondrial biogenesis regulator NRF1 to control levels, as shown in a representative western blot of whole cell extracts, with quantification of band densities in the adjacent plots as mean ± s.em., n = 3. Scrambled siRNA was used as control in both control and ASM-deficient cells for all experiments involving ETV1 or KLF2 silencing. T-test p-values *p<0.05, **p<0.01 ***p<0.001 (c) Silencing of ETV1 or KLF2 increases the transcript levels of mitochondrial genes, as assessed by qPCR. The data is presented in a Heatmap, in which blue denotes decrease in expression compared to the control cells (white represents no change relative to the control values) and red denotes increase. Note the mostly decreased (blue) mitochondrial genes in ASM-deficient cells and their turn to red (increased expression) when ETV1 or KLF2 are silenced (n = 3). (d–e) Silencing of either Klf2 or Etv1 partially rescues the decreased basal and maximal OCR in ASM-deficient fibroblasts as measured by real time respirometry. The plot shows the mean ± s.e.m., n = 3. T-test p-values **p<0.01 and ***p<0.001. (f) Robust silencing of KLF2 or ETV1 in NPC1-deficient cells shows accordingly, significantly decreased transcript levels of KLF2 and ETV1. Graphs represent mean ± s.e.m, n = 3 with T-test p-values ***p<0.001 (g) KLF2 or ETV1 knockdowns in NPC1-deficient cells increases the transcript levels of mitochondrial genes, which is presented as a Heatmap. Note the mostly increased (red) mitochondrial genes when KLF2 or ETV1 are silenced relative to Scrambled siRNA (white) in NPC1-deficient cells (n = 3).

https://doi.org/10.7554/eLife.39598.019

The knock-downs of ETV1 and KLF2 were both effective (Figure 5A–B). Interestingly, transcription factor nuclear respiratory factor 1 (NRF1), a known inducer of mitochondria-related gene expression, was also sharply down-regulated in ASM-deficient fibroblasts, and was rescued by the silencing of ETV1 or of KLF2. This result suggests a compound effect of repression of mitochondria-related genes by KLF2 and ETV1, combined with decreased activation of the expression of the same genes by NRF1. We have shown above (Figure 3—figure supplement 2) that the ASM-deficient cells accumulate mitochondria, and for that reason show increased levels of mitochondrial proteins. Nevertheless, some mitochondrial proteins are present at lower levels in ASM1-deficient cells, of which TFAM is a notable example. TFAM is a target of NRF1, and its protein levels are sharply decreased in ASM-1 cells, but are readily normalized by silencing of ETV1 (Figure 5A) or KLF2 (Figure 5B). Importantly, the transcript levels of genes encoding mitochondrial proteins, which are down-regulated in ASM-deficient fibroblasts, were increased by the silencing of ETV1 and even more robustly increased by KLF2 silencing (Figure 5C). This pattern also includes NRF1 and its closely related protein nuclear respiratory factor 2 (NRF2, also known as GABPA), again suggesting that these two transcription factors may be repressed by KLF2 and ETV1. Importantly, the improvement in the expression of mitochondria-associated genes by silencing KLF2 or ETV1 is not due to an improvement of the lysosomal phenotype. We measured readouts of lysosomal function such as the accumulation of autophagosomal marker LC3BII or autophagy substrate p62, by Western blot, and found that silencing of KLF2 or ETV1 had no impact on the lysosomal dysfunction in ASM-deficient cells (Figure 5—figure supplement 1). Finally, mitochondrial respiration was partly rescued in ASM-1 fibroblasts by the knock-down of ETV1 and robustly rescued by KLF2 silencing (Figure 5D), both under basal and maximal electron flow conditions (Figure 5E).

To ensure that the effect of KLF2 and ETV1 on mitochondria is not limited to ASM-deficient cells, we also tested how the silencing of KLF2 and ETV1 impacts NPC1-deficient cells. The knock-downs were robust (Figure 5F), and resulted in a strong increase in the expression of mitochondria-related genes (Figure 5G).

Altogether, these results show that KLF2 and ETV1, two transcription factors that are increased in ASM- and NPC1-deficient fibroblasts and hyperactive in NPC1 KO tissues, repress mitochondrial biogenesis and that their silencing restores mitochondrial biogenesis and function in ASM- and NPC1-deficient fibroblasts.

KLF2 regulates ETV1 in an ERK-dependent manner

The silencing of KLF2 had a more robust effect on the recovery of mitochondrial function than the silencing of ETV1. For this reason, we set to understand if these transcription factors work in parallel pathways or if they are epistatic. We observed that the silencing of KLF2 in ASM-deficient fibroblasts results in the ablation of ETV1 (Figure 6A), while ETV1 silencing has only a minor effect on KLF2 (Figure 6A). Since we have shown above that ETV1 is regulated at transcript level (Figure 4—figure supplement 1A), this result implies that KLF2 regulates (activates) the transcription of the gene encoding ETV1, in agreement with the increased transcript levels of ETV1 in ASM-deficient fibroblasts. These findings are also validated by the results of the KLF2 ChIP-Seq analysis, whose target genes include ETV1 (Figure 4—figure supplement 4). These results suggest that KLF2 and ETV1 are epistatic, with ETV1 downstream of KLF2.

ETV1 up-regulation is dependent on KLF2 and ERK.

(a) Silencing of KLF2 in ASM-deficient fibroblasts results in reduced levels of ETV1, shown by a representative western blot of whole cell extracts, with quantification of band densities (mean ± s.e.m, n = 3) in adjacent plots. One way ANOVA p-values ***p<0.001. (b) ASM-deficient fibroblasts show increased ERK and mTORC1 activities, reduced AKT activity and unchanged AMPK activity, as shown by a representative western blot of whole cell extracts with band density quantification presented in the adjacent plot as mean ± s.e.m., n = 3. T-test p-values **p<0.01 ***p<0.001 (c) ERK inhibition by treatment with U0126 (20 µM, 16 hr) in ASM-deficient fibroblasts results in reduced ETV1 levels but does not affect KLF2, as shown by a representative western blot, with band density quantification in the adjacent plot depicted as mean ± s.e.m. for biological triplicates. T-test p-values **p<0.01 ***p<0.001.

https://doi.org/10.7554/eLife.39598.021

Next, we tested known signaling modulators of KLF2 or ETV1 in ASM-deficient fibroblasts. Akt signaling down-regulates KLF2 (Skon et al., 2013), and we observed that Akt seems deactivated in ASM-deficient fibroblasts, as assessed by decreased phosphorylation of Akt Serine 473 (Figure 6B). ERK is a positive effector of ETV1 (Janknecht, 1996), and we found ERK signaling increased in ASM-deficient fibroblasts (Figure 6B). mTORC1 signaling is often involved in lysosomal stress signaling, and we found it activated in ASM-deficient fibroblasts, as assessed by the phosphorylation of p70S6 kinase (P70S6K) Threonine 389 (Figure 6B). AMPK signaling, which regulates mTORC1 as well as biogenesis of mitochondria and lysosomes, was not affected, as assessed by phosphorylation of AMPK target acetyl-CoA carboxylase (ACC) or of the activating phosphorylation of AMPK itself (Figure 6B). Inhibition of mTORC1 signaling in ASM-deficient fibroblasts by treatment with the mTORC1 inhibitor torin1 had no effect on the expression of mitochondria-related genes or mitochondrial function (data not shown).

We next tested if the increased ERK signaling was related to the increased levels of ETV1. We treated the ASM-deficient fibroblasts with the ERK inhibitor U0126, which led to the ablation of ERK signaling, as expected (Figure 6C). KLF2 was mostly unaffected by ERK inhibition (Figure 6C). However, ETV1 was returned to control levels (Figure 6C). This result suggests that KLF2 can only trigger ETV1 expression in the presence of active ERK signaling.

S1PR1 signaling dynamically regulates KLF2 and mitochondrial biogenesis and function

Next, we sought to identify the mechanism leading to KLF2 up-regulation. Since one of the consequences of lysosomal malfunction is the stalling of the autophagy pathway, we tested if KLF2 could be induced by perturbations in autophagy, such as inhibition of autophagosome formation (Atg5 silencing) or inhibition of the fusion of autophagosomes to lysosomes (syntaxin 17 silencing). However, no effect was observed in KLF2 (data not shown).

KLF2 is known to be negatively regulated by Akt signaling (Skon et al., 2013), which is repressed in ASM-deficient fibroblasts (Figure 6B). Interestingly, one of the genes induced by KLF2 is the sphingosine-1-phosphate receptor 1 (S1PR1) (Skon et al., 2013), which we find up-regulated at transcript level in ASM-deficient fibroblasts (Figure 7—figure supplement 1). S1PR1 and KLF2 are part of a signaling network in which the activity of the receptor represses its own expression by downregulating KLF2 via Akt activation (Sinclair et al., 2008; Skon et al., 2013). Interestingly, the S1PR1 receptor has been previously shown to affect mitochondrial function in T cells, but the mechanisms remained unexplored (Mendoza et al., 2017). Furthermore, the levels of sphingosine-1-phosphate (S1P) are decreased in the plasma of NPC1 patients (Fan et al., 2013), suggesting that signaling elicited by S1P may be down-regulated.

Given the connections between S1PR1, KLF2 and our findings implicating KLF2 in the regulation of mitochondrial-related gene expression, we decided to test if perturbation of the S1PR1 pathway in ASM- or NPC1-deficient cells could explain the up-regulation of KLF2 and, accordingly, the expression of mitochondria-related genes. To this end, we first sought to establish that S1PR1 can regulate mitochondrial biogenesis and function in healthy cells. We treated control fibroblasts with either a selective agonist (Sew2871) or with a selective inhibitor (W146) of S1PR1, and measured the effects on mitochondria. We observed that the activation of S1PR1 by the agonist Sew2871 results in increased transcript levels of mitochondria-related genes (Figure 7A). Reciprocally, inhibition of S1PR1 by W146 leads to decreased transcript levels of these genes (Figure 7B). Furthermore, activation of S1PR1 results in increased mitochondrial OCR under basal and uncoupled conditions (Figure 7C, quantified in 7E), while the inhibition of the receptor results in a robust inhibition of mitochondrial OCR (Figure 7D, quantified in 7F). Finally, we observed that KLF2 responds as expected to S1PR1 activity. When S1PR1 is activated, KLF2 levels decrease (Figure 7G), while inhibition of S1PR1 results in increased KLF2 abundance (Figure 7H). ETV1 shows a similar pattern, decreasing when S1PR1 is activated (Figure 7G) and increasing in response to S1PR1 inhibition (Figure 7H). Notably, the protein levels of mitochondrial proteins TFAM, cytochrome oxidase I (mtCOI),succinate dehydrogenase subunit b (SDHB) and porin (VDAC1) are all increased when KLF2 and ETV1 are down-regulated (S1PR1 activation, Figure 7G), and all decreased when KLF2 and ETV1 levels are increased (S1PR1 inhibition, Figure 7H). These results underscore that the S1PR1-KLF2-ETV1 mitochondrial biogenesis pathway can be dynamically regulated in control fibroblasts. Furthermore, these data suggest that the S1PR1 pathway may be down-regulated in ASM-deficient fibroblasts, given the increased levels of KLF2 and the decreased expression of mitochondria-related genes. Interestingly, the expression of sphingosine kinase 1 (SPHK1), which generates S1P that can be exported to the extracellular space, is down-regulated in ASM-deficient fibroblasts (Figure 7—figure supplement 1). Similarly, SPHK2, which generates S1P intracellularly, in mitochondria and endoplasmic reticulum, is also down-regulated in ASM-deficient fibroblasts (Figure 7—figure supplement 1). Altogether, these results suggest that S1P signaling via S1PR1 is profoundly down-regulated in ASM-deficient fibroblasts, and that this event is at the root of the up-regulation of KLF2 and its downstream consequences, particularly ETV1 induction and inhibition of mitochondrial biogenesis.

Figure 7 with 1 supplement see all
Dynamic regulation of S1PR1 activity impacts mitochondrial biogenesis and function.

(a) Transcript levels of mitochondrial-related genes increase upon activation of S1PR1 with the agonist Sew2871 (5 µM, 16 hr; DMSO as vehicle control), as measured by qPCR. Plots show mean ± s.e.m., n = 3. T-test p-values *p<0.05 and **p<0.01 (b) Transcript levels of mitochondrial-related genes decrease upon inhibition of S1PR1 with the competitive antagonist W146 (10 µM, 16 hr; methanol as vehicle control), as measured by qPCR. Plots show mean ± s.e.m., n = 3. T-test p-values **p<0.01 and ***p<0.001 (c) Increased OCR in cells treated with the S1PR1 agonist Sew2871 compared to vehicle control (DMSO), quantified in panel (e). (d) Decreased OCR in cells treated with the S1PR1 antagonist W146 compared to vehicle control (methanol), quantified in panel (f). Quantifications in e and f represent mean ± s.em., n = 3 with T-test p-values ***p<0.001. (g) Representative blots showing decreased protein levels of KLF2 and ETV1, and increased amounts of mitochondrial proteins VDAC1, TFAM, CO1 and SDHB, in cells treated with S1PR1 agonist Sew2871, assessed by western blots of whole cell extracts, using HPRT as loading control. Adjacent plot depicts the fold difference in band density relative to vehicle control (DMSO) as mean ± s.e.m., n = 2 with technical triplicates (the line on zero denotes no change relative to the controls, negative numbers show decrease in fold change, positive numbers show increased fold change). T-test p-value *p<0.05 (h) Representative blots depicting increased protein levels of KLF2 and ETV1, and decreased amounts of mitochondrial proteins VDAC1, TFAM, CO1 and SDHB, in cells treated with S1PR1 antagonist W146, assessed by western blots of whole cell extracts, using HPRT as loading control. Adjacent plot shows the difference in fold band density compared to vehicle control (methanol) and depicted as average ± s.e.m., n = 2 with technical triplicates. T-test p-value *p<0.05, **p<0.01 and ***p<0.001.

https://doi.org/10.7554/eLife.39598.022

S1PR1 is mislocalized in ASM-deficient cells and unresponsive to activators

Given the apparent down-regulation of S1PR1 signaling in ASM deficiency, we set to test if reactivation of the S1PR1 pathway in ASM-deficient fibroblasts would rescue the expression of mitochondria-related genes as well as mitochondrial function. We treated control and ASM-deficient fibroblasts with the S1PR1 agonist Sew2871, and in agreement with our data shown above (Figure 7A), we found an increase in the expression of mitochondria-related genes in control fibroblasts (Figure 8B). However, and surprisingly, the ASM-deficient fibroblasts did not respond to the treatment with the S1PR1 agonist: no change was observed in the transcript levels of mitochondria-related genes (Figure 8B). Similar results were obtained when using S1P instead of the agonist (data not shown). These results suggest that the S1PR1 receptor is absent or inaccessible to extracellular cues, implying that it may be sequestered away from the plasma membrane. The protein levels of S1PR1 are not changed in ASM-deficient fibroblasts (Figure 8C). Therefore, we tested if S1PR1 localization at the plasma membrane was affected in ASM-deficient cells. We used a PE-conjugated antibody against S1PR1 for flow cytometry, in non-permeabilized cells, and determined the amount of plasma membrane labelling in control and ASM-deficient fibroblasts. As negative control, we treated cells with FTY720, which antagonizes S1PR1 signaling by promoting its endocytosis. The treatment with FTY720 reduced the levels of S1PR1 at the plasma membrane, which were robustly decreased in ASM-deficient cells. Thus, the mislocalization of S1PR1 in ASM-deficient cells, and consequent decreased signaling, explain the increase in KLF2 signaling and its downstream consequences.

S1PR1 signaling in Niemann-Pick disease.

(a) Schematic illustration of sphingosine-1-phospate (S1P) signaling. S1P is generated from sphingosine by the kinases SPHK1 (plasma membrane) and SPHK2 (endoplasmic reticulum and mitochondria), and can be transported out of the cell. Extracellular S1P can activate several receptors (S1PR1-5). Specifically, stimulation of S1PR1 triggers Akt signaling which regulates KLF2 levels. Expression of S1PR1 is regulated by KLF2, which as shown by our data also activates ETV1. Sew2871 is an agonist of S1PR1, and W146 is an antagonist. (b) Treatment of control fibroblasts with S1PR1 agonist Sew2871 (5 µM, 16 hr) results in increased transcript levels of mitochondria-related genes in control fibroblasts, but has no effect on ASM-deficient fibroblasts. The data is presented in a Heatmap for n = 3, in which blue denotes decrease in expression compared to the control cells (white represents no change relative to the control values) and red denotes increase. Note the mostly increased (red) mitochondrial genes in control fibroblasts treated with Sew2871, while blue in ASM fibroblasts regardless of the treatment. (c) Protein levels of S1PR1 are not changed in ASM fibroblasts, as measured by western blot using whole cell extracts. Adjacent plot shows the quantification presented as mean ± s.e.m., n = 3. T-test p-value>0.05. (d) Staining of S1PR1 present at the plasma membrane, in non-permeabilized cells, measured by flow cytometry. FTY720 triggers the endocytosis if S1PR1 and was used as a negative control for surface staining. Note the barely detectable surface S1PR1 levels. Plots represent the average fraction of S1PR1 levels normalized to vehicle treated control cells and depicted as mean ± s.e.m., n = 3. T.test p-value ***p<0.001 (e) Staurosporine-treated ASM-deficient cells with KLF2 and ETV1 silencing show increased apoptotic cell population relative to control ASM-deficient cells as measured by flow cytometry with Annexin-V and Propidium iodide staining. Quantifications are depicted as mean ± s.e.m., n = 5. T-test p-value ***p<0.001. (f) Staurosporine-treated ASM-deficient fibroblasts with either silencing control, KLF2 or ETV1 knockdowns show increased protein amounts of Cleaved PARP and Cleaved Caspase three levels in cells with KLF2 or ETV1 silencing and quantifications in adjacent graphs show mean ± s.em., n = 3. T-test p-value **p<0.01 and ***p<0.001. (g) Decreased cell viability as measured by Glo Titer Assay in ASM-deficient fibroblast with either KLF2 or ETV1 silencing. Plots represent mean± s.e.m., n = 2 with six technical replicates per condition. T-test p-value, ***p<0.001.

https://doi.org/10.7554/eLife.39598.024

Upregulation of KLF2 and ETV1 has a protective effect in ASM-deficient cells

Finally, we sought to test if induction of KLF2 and ETV1 in ASM-deficient cells contributes to Niemann Pick disease pathology or if it is a protective mechanism. We again resorted to the silencing of KLF2 and ETV1 by siRNA, and measured cell death using the Annexin-V/propidium iodide flow cytometry assay. We observed that ~38% of ASM-deficient cells are apoptotic (high Annexin V signal), while this number increases to ~55% when KLF2 or ETV1 are silenced (Figure 8E). Accordingly, the amount of cleaved (active) caspase-3 and of cleaved PARP (caspase-3 target) are increased when KLF2 and ETV1 are silenced (Figure 8F). Finally, cell viability, measured by the CellTiter-Glo assay, is robustly decreased when KLF2 or ETV1 are silenced (Figure 8G). Thus, the up-regulation of KLF2 and ETV1 in ASM-deficient cells is a protective mechanism.

Discussion

This study addresses a novel mechanism by which mitochondria are impaired in lysosomal lipid storage diseases. We show here that the transcription factors KLF2 and ETV1 repress the expression of genes encoding mitochondrial proteins. Both KLF2 and ETV1 are up-regulated in patient cells from Niemann-Pick type C and acid sphingomyelinase (ASM) deficiency, and their silencing, particularly KLF2, is sufficient to return mitochondrial biogenesis and function to control levels. Decreased signaling through sphingosine-1-phosphate receptor 1 (S1PR1) activates KLF2, which induces the expression of ETV1, culminating in the down-regulation of mitochondrial biogenesis.

The transcriptional regulation of mitochondrial biogenesis is known since the identification of the transcription factor nuclear respiratory factor 1 (NRF1), which induces the expression of many respiratory chain and mtDNA maintenance genes (Scarpulla et al., 2012). Several other transcription factors have been shown to stimulate mitochondrial biogenesis, such as estrogen related receptor α (ERRα) or the oncogene myc (Scarpulla et al., 2012). The role of the co-activator PGC1α (peroxisome proliferation activated receptor gamma, co-activator 1α) has also been shown to promote NRF1- and ERRα-mediated mitochondrial biogenesis (Wu et al., 1999). However, to our knowledge, no transcription factor has previously been shown to repress mitochondrial biogenesis. Thus, the roles of KLF2 and ETV1 as repressors of mitochondrial biogenesis, shown in this manuscript, open a new paradigm on the transcriptional regulation of the mitochondrial biogenesis. Interestingly, another Krüppel-like factor, KLF4, was recently shown to promote mitochondrial biogenesis in the heart (Liao et al., 2015), implying that the repressive behavior of KLF2 is a specificity of this transcription factor and not a characteristic transversal to the whole Krüppel-like factor family.

Notably, the transcription factor NRF1, which is a known positive regulator of mitochondrial biogenesis and is down-regulated in fibroblasts with acid sphingomyelinase deficiency, is also repressed by KLF2 and ETV1. It therefore seems that KLF2, ETV1 and NRF1 may form a transcriptional regulatory network that dynamically regulates mitochondrial biogenesis, with ‘accelerator’ (NRF1) and ‘brakes’ (KLF2 and ETV1). The transcriptional network between KLF2, ETV1 and NRF1, as well as the involvement of other transcription factors such as ERRα, myc, or co-activators such as PGC1α, warrants further research.

It is particularly interesting that a transcriptional network repressing mitochondrial biogenesis appears robustly active in lysosomal diseases. The role of lysosomes in cellular function has been subject of increasing attention, both regarding its physiological roles as a signaling platform as well as the pathological consequences of lysosomal defects in lysosomal storage diseases (Settembre et al., 2008; Ballabio and Gieselmann, 2009; Perera and Zoncu, 2016; Platt et al., 2012). Numerous studies describe the impact of lysosomal defects on the function of other organelles, particularly mitochondria, in several lysosomal storage diseases (Diogo et al., 2018; Plotegher and Duchen, 2017; Raimundo et al., 2016; Torres et al., 2017). Mitochondria are usually impaired in cells and tissues with primary lysosomal defects, with decreased oxygen consumption and increased production of superoxide and other reactive oxygen species (Jolly et al., 2002; Plotegher and Duchen, 2017). However, this is often usually attributed to a decrease in autophagy (and mitophagy), with the consequent accumulation of damaged mitochondria in the cytoplasm. Our data in cellular and mouse models of Niemann-Pick-C disease and acid sphingomyelinase deficiency shows, however, that in addition to defective autophagy there is a signaling mechanism based on the induction of two transcription factors, KLF2 and ETV1, which repress mitochondrial biogenesis. This may represent a signaling circuit in which the cells with lysosomal defects repress the generation of an organelle whose degradation requires lysosomal function. It may also be a consequence of the accumulation of lipids such as sphingomyelin and cholesterol in the lysosomes in Niemann-Pick type C and acid sphingomyelinase deficiency, which is likely to result in deficiency of those lipids in other cellular locations. Thus, one conceivable cellular adaptation would be shutting down the mitochondrial respiratory chain and citrate cycle, which would allow to shunt citrate to the cytoplasm, where it can be converted by acetyl-CoA lyase to acetyl-CoA and used for lipid synthesis (Bauer et al., 2005; Wellen et al., 2009). Interestingly, a reciprocal mechanism seems to exist, since chronic mitochondrial defects result in repression of lysosomal biogenesis (Nezich et al., 2015; Woś et al., 2016; Fernández-Mosquera et al., 2017) and function (Demers-Lamarche et al., 2016; Fernandez-Mosquera et al., 2018). The interplay between mitochondria and lysosomes is a relatively novel concept that is only now being grasped (Diogo et al., 2018; Raimundo et al., 2016). The existence of cross-talk mechanisms involving transcriptional networks implies that the communication between these two organelles goes beyond metabolic cues, and involves complex cellular signaling.

The up-regulation of KLF2 in ASM-deficient cells seems to be a consequence of impaired sphingosine-1-phosphate (S1P) signaling through S1P receptor 1 (S1PR1). This receptor had previously been implicated in the regulation of mitochondrial function in T cells, but the mechanism remained unclear (Mendoza et al., 2017). We show in this study that S1PR1 is a bona fide bi-directional regulator of mitochondrial function via the effect of KLF2 and ETV1 on mitochondrial biogenesis. Indeed, both the activation and the inhibition of S1PR1 in control cells impacted mitochondrial biogenesis and function. This effect can be interpreted in diverse biological scenarios. For example, in the context of the role of S1P and S1PR1 in angiogenesis, decreased signaling could be interpreted as impaired angiogenesis, thus inefficient delivery of O2, and the cells respond by shutting down the major O2 consumption cellular component – mitochondria. Interestingly, the acid sphingomyelinase-deficient fibroblasts were non-responsive to agonists of S1PR1, which suggests that the receptor may be sequestered away from the plasma membrane in the patient cells. In support of this hypothesis, the amount of S1PR1 in the plasma membrane of the ASM-deficient cells is negligible, while the total protein levels of S1PR1 are similar to control cells. This result implies a mistargeting of S1PR1 in acid sphingomyelinase deficiency and Niemann-Pick disease, which is akin to other proteins aberrantly mislocalized away from the plasma membrane in these diseases, such as Met receptor tyrosine kinase or K-Ras (Schuchman and Wasserstein, 2016; Praggastis et al., 2015). Thus, a therapeutic strategy targeting the receptor activity would likely be insufficient.

The contribution of the signaling pathways mediating communication between mitochondria and lysosomes and their roles in pathology certainly warrants further exploration, not just in mitochondrial and lysosomal diseases but also in the context of neurodegenerative diseases that arise from defects of either of these organelles.

Materials and methods

Drugs and cellular treatments

The following drugs were used for cellular treatments: 1 μM Oligomycin (Sigma, O4876), 2 μM Carbonyl cyanide 3-fluorophenylhydrazone (FCCP) (Sigma, C2920), 1 μM Rotenone (Sigma, R8875), 1 μM Antimycin (Sigma, A8674), 40 µM Desipramine (Biotrend, BG0162), 5 µM Sew2871 (Cayman, 10006440), 20 µM U0126 (Millipore, 662005), 10 µM U18666A (Cayman, 10009085), 10 µM W146 (Sigma-Aldrich, W1020), 2 µM FTY720 (Selleckchem, S5002) and 4 µM Staurosporine (Sigma-Aldrich, 37095).

Cell culture and transient transfections

Control and Niemann-Pick patient fibroblasts were grown in DMEM high glucose medium (Gibco, 11965) supplemented with 10% fetal bovine serum and 1% Penicillin/Streptomycin at 37°C and 5% CO2, in a humidified incubator, unless otherwise stated. ASM-1 patient fibroblasts retained about 5% of the control activity of acid sphingomyelinase, and were collected and maintained according to the ethical guidelines of the UMG. Control and patient fibroblasts were transfected with siRNAs for ETV1 or KLF2 using electroporation (Amaxa kit, Lonza, V4XP-1024) or with scrambled control siRNA following manufacturer’s protocol. Additional control, human, adult primary fibroblasts were obtained ATCC (PCS-201–012). NPC1 patient cells and an additional ASM patient line (referred to as ASM-2 in the text) were obtained from Coriell Institute for Medical Research (GM18398, GM13205). The cell lines were not authenticated for cross-type contamination and were tested periodically for mycoplasma. The use of human cells for these studies was approved by the Ethical Commission of the Universitätsmedizin Göttingen.

Mouse tissues

The NPC1 mice are maintained at the University of Helsinki, and the ASM mice are maintained at the University of Erlangen. In both cases, the maintenance of these animals is approved under the directive 2010/63/EU.

XF medium

XF assay medium (Seahorse Bioscience, 100965–000) was supplemented with sodium pyruvate, glutamax and glucose following manufacturer’s recipe and the pH of medium was adjusted to 7.4.

Oxygen consumption rate measurements

OCR was measured in fibroblasts using the XF96 Extracellular Flux analyzer (Seahorse Bioscience). Briefly, cells were seeded at 2 x 104 cells per well in XF96 cell culture multi-well plates in DMEM medium and incubated for 24 hr in the growth conditions stated for all cell cultures. XF96 cartridges were incubated overnight in XF calibrant at 37°C in a non-CO2 incubator. Prior to OCR measurements, the growth medium of cells was exchanged with XF medium and incubated at 37°C in a non-CO2 incubator for 1 hr. Inhibitors were diluted to appropriate concentrations in XF medium and loaded into corresponding microwells in the XF96cartridge plate. Following equilibration of sensor cartridges, XF96 cell culture plate was loaded into the XF96 Extracellular Flux analyzer at 37°C and OCR was measured after cycles of mixing and acquiring data (basal) or inhibitor injection, mixing and data acquisition.

Western blotting

Whole cell extracts of cultured fibroblast were prepared in 1.5% n-dodesylmaltoside (Roth, CN26.2) in PBS supplemented with protease and phosphatase inhibitor cocktail (Thermoscientific, 78442) as described (Raimundo et al., 2009). Protein concentrations of whole cell extracts were determined using a Bradford assay (Bio-Rad, 500–0006). 50 μg of sample proteins per well were subjected to Sodium dodecyl sulfate -polyacrylamide gel electrophoresis (SDS-PAGE) and transferred to polyvinylidene fluoride (PVDF) membranes (Amersham, Life Technologies). After blocking in 5% Milk in TBS tween, membranes were immunoblotted with the following antibodies: Sqstm1 (Abcam, ab110252), Hprt (Abcam, ab10479), Klf2 (Abcam, ab203591), Etv1 (Abcam, ab184120), Lc3b (Cell signaling, 3868), Pan Akt (Cell signaling, 4691), Phospho Akt (Cell signaling, 4060), Total Erk1/2(Cell signaling, 4695), Phospho Erk1/2 (Cell signaling, 4376), Tfam (Abcam, ab138351), P70s6k1 (Cell signaling, 2708), Phospho P70s6k1 (9234), Acc (Cell signaling, 3676), Phospho Acc (Cell signaling, 3661), Ampkα (Cell signaling, 5832), Phospho Ampkα (Cell signaling, 2535), Nrf1 (Abcam, ab175932), Atp5a, Uqcrc2, Sdhb and mtCO1 OXPHOS cocktail (Abcam, ab110413), S1pr1 (Abcam, ab125074), Cleaved PARP (Cell signaling, 5625), Cleaved Caspase-3 (Cell signaling, 9664), Laminin A/C (Cell signaling,) and Vdac1 (Abcam, ab14734). Band densitometric quantifications were determined using ImageJ software 1.48 v. Following normalization with Hprt, all control samples of each experiment were centered at one to ease relative comparisons with experimental samples.

Subcellular fractionation

Patient and control fibroblasts were harvested at 80% confluence by scraping in ice-cold PBS. Nuclear and cytosolic fractions were isolated from the cell pellets using a nuclei/cytosol fractionation kit (BioVision, K266). Nuclear and cytosolic proteins, along with whole cell extracts, were subjected to Western blot analyses.

Measurement of lysosomal proteolytic capacity

Lysosomal proteolytic capacity was measured using the DQ Red BSA Dye (Molecular Probes, D-12051) following manufacturer’s protocol. Briefly, 100 ul of 1mg/ml dye was added to 10 ml of warm DMEM medium. Previously plated cells in a transparent 96 well-plate were loaded with 100 ul per well each of the dye containing medium and incubated at 37°C for 1 hr. Cells were then washed twice with warm PBS and the medium was replaced with 100 µL/well of warm EBSS medium. The kinetics of DQ Red BSA digestion were recorded at respective excitation and emission maxima of 590 nm and 620 nm in a multi-plate reader over a 4 hr period.

Quantitative RT-PCR

RNA extraction and purification from fibroblasts were performed using Crystal RNA mini Kit (Biolab, 31-01-404). From mouse livers, RNA was extracted using the TRI Reagent (Sigma-Aldrich, T9424). RNA concentration and quality were determined using Nanodrop (PeqLab) and cDNA was synthesized with iScript cDNA synthesis kit (Bio-Rad, 178–8991) following manufacturer’s protocol. Each 8 μl q-PCR was made of 4 μl diluted cDNA, 0.2 μl of each primer (from 25 μM stock) and 3.6 μl of iTaq Universal Sybr Green Supermix (Bio-Rad, 172–5124) and ran on the QuantStudio 6 Flex Real-Time PCR system (Applied Biosystems). Transcript levels measured by quantitative PCR (qPCR) were determined by the ΔΔCT method using HPRT and GAPDH (not shown) as reference genes. Unless otherwise indicated, qPCR experiments of at least three biologically independent experiments always included at least technical triplicates. For the determination of relative expression (fold change), all control samples were centered at one by normalizing the expression of experimental samples to those of the corresponding controls.

Flow cytometry

Measurement of mitochondrial superoxide levels using MitoSOX Red Mitochondrial superoxide indicator (Molecular Probes, M36008) was performed by flow cytometry according to the manufacturer’s instructions. For S1PR1 plasma membrane localization, 1 × 106 control and ASM deficient fibroblasts treated with or without 2 µM FTY720 were labelled in suspension with 10 µL of PE-conjugated S1PR1 antibody (R and D systems, FAB2016P) for 1 hr, washed twice in isotonic PBS supplemented with 1% BSA, resuspended in 200–400 uL of buffer and subjected to flow cytometry analyses for the surface expression of S1PR1. For apoptosis measurements, 1 × 105 cells were plated 24 hr prior to flow cytometric determinations. Cells were then treated for 1 hr with 4 µM Staurosporine, harvested and stained in suspension with Annexin V (BD Pharmingen, 556419) and Propidium iodide (Sigma-Aldrich, P4170) in the dark for 20 min and analyzed by flow cytometry. Analyses of flow cytometry results were done using FlowJo v10 (FlowJo, LLC).

Cell viability

Measurement of cell viability in patient fibroblast was carried out using the Cell Titer-Glo Luminiscent cell viability assay (Promega, G7570) following manufacturer’s protocol.

Dataset selection

In order to identify transcriptional signatures mediating interactions between organelles in Niemann-Pick pathology, we mined for microarray data involving Niemann-Pick mouse models from the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo). Criteria for dataset selection included datasets with multiple replicates from several tissues. The dataset selected was GSE39621, which includes samples of brain, liver and spleen of mice before and after 6 weeks of age, when the symptoms of the disease start manifesting. Given that the spleen may contain immune cells in addition to splenocytes, and likely to have many more of non-splenocytes in the disease case, since spleen enlargement is a hallmark of the disease, we considered that the control and Npc1-/- were not directly comparable and thus used only the data relative to brain and liver.

Organelle-specific gene lists

We obtained organelle proteomes from up-to-date and comprehensive databases for mitochondrial (and respiratory chain subunits), lysosomal, peroxisomal, endoplasmic reticulum and Golgi proteomes (Table 1). These protein IDs were converted to NCBI gene symbols, which were then used to identify the corresponding probeset names for different microarray matrices.

Microarray data analysis

We obtained mouse Npc1 wildtype, Npc1+/- and Npc1-/- in asymptomatic (less than 6 weeks old) and symptomatic (more than 6 weeks old) brain, liver and spleen from the GEO database (Alam et al., 2012). The controls for the NPC1 dataset are the wt mice in the brain but the heterozygous mice in the other tissues. We used the software GeneSpring (Agilent Technologies, Santa Clara, CA) to normalize the datasets by robust multi-array averaging (RMA) to normalize datasets (Raimundo et al., 2009). The datasets for all tissues originating from the same knock-out mouse and corresponding controls were normalized together. After normalization, we determined which transcripts had significantly different expression between Npc1-/- and controls for each individual tissue, using ANOVA. We also calculated the fold change from probe expression values between lysosomal disease and control mice for each tissue. The statistical filter was set at p-value<0.05, and the transcripts that pass the filter for each tissue represent the corresponding transcriptional signature.

To calculate the average expression of organelle-specific gene lists, we normalized each transcript to the average of the control samples, and calculated the average of the expression levels of all genes in each organelle-specific gene list. To determine if the difference observed between Npc1-/- and controls was significant, we calculated the t-test p-value (unpaired, unequal variance) for the whole gene set using Microsoft Excel. Given that the lists have hundreds of genes, we performed a Bonferroni post-hoc correction. The adjusted p-values<0,05 were considered significant.

Pathway analysis and identification of transcriptional regulators

We employed a multi-dimensional strategy aimed at the identification of signaling pathways, as described (Raimundo et al., 2012; Raimundo et al., 2009; Schroeder et al., 2013; West et al., 2015). The transcriptional lists were imported to the software Ingenuity Pathway Analysis (IPA) (http://www.ingenuity.com), which then determines which pathways and transcriptional regulators are statistically enriched, using Fisher’s exact test. The statistical threshold was set at p<0.01.

Promoter analysis

To perform promoter analysis on the respiratory chain genes, we imported the respiratory chain gene list to the software Genomatix Suite (www.genomatix.de). Then we set a pipeline within the software suite, by first defining the promoters of the respiratory chain genes and then determining which transcription factors (TF) had binding sites on them. To locate the promoters, we use the Genomatix tool Gene2Promoter, and defined the promoter region from 500 base pairs upstream (−500) the transcription start site (TSS) until 100 base pairs downstream the TSS (+100). Given that some genes may have more than one promoter due to alternative splicing, we selected only the promoters that drive the expression of the transcript leading to the protein that functions as a respiratory chain subunit. The promoter sequences were then used to determine cis-elements and identify the corresponding TF, limiting the search to those TF that had at least a binding site in at least 85% of the promoters. The software provides a statistical assessment of the enrichment of the binding sites for each TF family in the promoters under analysis. We set a threshold of p<0.05 for the Fisher's exact test p-value for each TF family enrichment. Then, we determine, for each significantly enriched family, which individual TF are included, and select as relevant TF those that have a binding site in at least 50% of the promoters under analysis.

Statistical analysis

Statistical analyses were carried out using Graph Pad Prism 6 and 7 softwares. Unless otherwise stated in the corresponding figure legends, all measures throughout this manuscript were summarized as graphs displaying mean ± s.e.m., of at least three independent biological replicates. The means of the corresponding controls are typically centered at one to ensure easier comparisons unless otherwise stated. For in vivo experiments, n represents number of mice of each genotype used in this study. For cell culture work, n refers to the number of independent experiments carried out with different stocks of each cell line. Each n included at least technical duplicates for cells and the standard errors of the means were calculated from the means of the numbers of independent biological replicates (n) with their technical replicates. Differences between group means were determined by the unpaired Welch’s t-test, assuming unequal variances between two groups and One way ANOVA for multi-group (at least three) comparisons; *p<0.05; **p<0.01; ***p<0.001; ns, non-significant p>0.05.

Accession numbers

The publicly-available transcriptome datasets used in this study are GSE39621 for Niemann Pick’s disease mouse model (Npc1-/-) (Alam et al., 2012) and GSE27602 for Klf2-/- mice (Redmond et al., 2011). The accession numbers for the ETV1 ChIP-ChIP (Baena et al., 2013) and KLF2 ChIP-Seq (Yeo et al., 2014) datasets are GSE39388 and E-MTAB-2365 respectively.

References

  1. 1
  2. 2
  3. 3
  4. 4
    Lysosomal disorders: from storage to cellular damage
    1. A Ballabio
    2. V Gieselmann
    (2009)
    Biochimica Et Biophysica Acta (BBA) - Molecular Cell Research 1793:684–696.
    https://doi.org/10.1016/j.bbamcr.2008.12.001
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
    Interaction of nuclear factors with multiple sites in the somatic cytochrome c promoter. characterization of upstream NRF-1, ATF, and intron Sp1 recognition sequences
    1. MJ Evans
    2. RC Scarpulla
    (1989)
    The Journal of Biological Chemistry 264:14361–14368.
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
    Mitochondrial respiratory chain deficiency impairs lysosomal hydrolysis via AMPK-PIKfyve-PI(3,5)P2-MCOLN1 pathway in press
    1. L Fernandez-Mosquera
    2. KF Yambire
    3. RF Couto
    4. LF Pereyra
    5. K Pabis
    6. A Ponsford
    7. M Stagi
    8. I Milosevic
    9. N Raimundo
    (2018)
    Autophagy.
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
  37. 37
  38. 38
  39. 39
    A genetic storage disorder in BALB/C mice with a metabolic block in esterification of exogenous cholesterol
    1. PG Pentchev
    2. AD Boothe
    3. HS Kruth
    4. H Weintroub
    5. J Stivers
    6. RO Brady
    (1984)
    The Journal of Biological Chemistry 259:5784–5791.
  40. 40
  41. 41
  42. 42
  43. 43
  44. 44
  45. 45
  46. 46
  47. 47
  48. 48
  49. 49
  50. 50
  51. 51
  52. 52
  53. 53
  54. 54
  55. 55
    Types A and B Niemann-Pick disease
    1. EH Schuchman
    2. MP Wasserstein
    (2016)
    Pediatric Endocrinology Reviews : PER 13:674–681.
  56. 56
  57. 57
  58. 58
  59. 59
  60. 60
  61. 61
  62. 62
  63. 63
  64. 64
  65. 65
  66. 66
  67. 67
  68. 68
  69. 69

Decision letter

  1. Agnieszka Chacinska
    Reviewing Editor; University of Warsaw, Poland
  2. Ivan Dikic
    Senior Editor; Goethe University Frankfurt, Germany
  3. Julia Sellin
    Reviewer; University of Bonn, Germany

In the interests of transparency, eLife includes the editorial decision letter, peer reviews, and accompanying author responses.

[Editorial note: This article has been through an editorial process in which the authors decide how to respond to the issues raised during peer review. The Reviewing Editor's assessment is that all the issues have been addressed.]

Thank you for submitting your article "Mitochondrial biogenesis is transcriptionally repressed in lysosomal lipid storage diseases" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by a Reviewing Editor and Ivan Dikic as the Senior Editor. The following individual involved in review of your submission has agreed to reveal her identity: Julia Sellin (Reviewer #2). The other reviewers remain anonymous.

The Reviewing Editor has highlighted the concerns that require revision and/or responses, and we have included the separate reviews below for your consideration. If you have any questions, please do not hesitate to contact us.

The reviewers find the paper interesting but they all point out several short comings of this work. Some of the limitations, highlighted by the comments of reviewer 1 and especially reviewer 2, are of technical or statistical nature and need to be fully addressed. Furthermore, the rescue and requested protein levels experiments suggested in several places are important. Finally, critical experiments to substantiate the notion that alterations in the mitochondrial biogenesis impact disease should be performed.

More precise wording based on the reviewers’ suggestions is required to improve the manuscript. Thorough attention to the reviewers’ comments will greatly improve the manuscript.

Separate reviews (please respond to each point):

Reviewer #1:

In this paper, the authors show that the accumulation of transcription factors KLF2 and ETV1 represses mitochondrial biogenesis in Niemann-Pick disease patient fibroblasts. Using publicly available transcriptome data of Niemann-Pick type C model mouse tissues and in silico promoter analysis of respiratory chain genes, they discovered a link between KLF2, ETV1 and mitochondrial genes. Interestingly they discover that in sphingomyelinase deficient patient cells sphingosine-1-phosphate receptor 1 localization at plasma membrane is altered, resulting in a link between S1PR signaling and KLF2.

This article is interesting because it shows for the first-time that KLF2 and ETV1 are transcriptional repressors for mitochondrial biogenesis and that lysosomal dysfunction affects mitochondria via cellular signaling beyond autophagy. While certain conclusions are supported by the presented data, this manuscript requires further experimentation to confirm and clarify key results.

Specific points:

In this article, authors analyze two acid sphingomyelinase deficient patient cell lines. To clarify which cell line is analyzed in each experiment, authors are requested to name and specify them in figures. In Corcelle-Termeau et al., 2016, three NPA fibroblasts were analyzed. Authors should describe which cell line they analyzed in this paper. Is it possible to show mutation sites of the in-house NPA fibroblast?

In Figure 3, authors analyzed only a single cell line from an ASM-deficient patient. It is not clear whether the reduced OCR and mtDNA level are specific features of this cell line or general. Authors should analyze mitochondrial functions in other fibroblasts including GM18398 as in Figure 3. Rescue experiments with wild type sphingomyelinase and NPC1 are also needed to clarify causal relationships.

In Figure 3A, authors measured OCR by Seahorse and showed quantifications in the bar graphs. Although control and ASM cells supposedly responded to antimycin and rotenone, the observed values for the ASM cells are considerably lower than control cells. Authors need to discuss why non-mitochondrial respiration is lower in ASM cells. In order to identify mitochondrial OCR, non-mitochondrial respiration should be subtracted from other measurements (Figure 3B, Figure 3—figure supplement 3D, Figure 3—figure supplement 4D, 5D, 7D, and 7F).

In order to check if the respiratory chain defects are observed at the protein level, the authors should clarify the abundance of respiratory chain components by SDS-PAGE and Western blot analysis.

In the final paragraph of subsection “Impaired mitochondrial respiration in NPC and ASM deficiency”, authors claimed U18666A treatment results in accumulation of lysosomal sphingomyelin. However, a previous report (Baulies et al., 2015, Scientific Reports) showed U18666A did not affect lysosomal distribution of sphingomyelin. Authors are requested to show appropriate references for the comment. Authors also describe that desipramine induces mitochondrial dysfunction. Since desipramine is an inhibitor of both acid sphingomyelinase and acid ceramidase, the results seem not to support the conclusion that the effects of mitochondrial impairment is independent of acid ceramidase activity as described in the last sentences.

Authors found the enrichment of KLF2 and ETV1 in mitochondrial promoters by in silico analysis and analyzed ETV1 ChIp targets based on the data of Baena et al., 2013. In order to clarify the results, authors should perform ChIP-qPCR and luciferase reporter assay to verify the direct interaction of the transcription factors to some representative mitochondrial promoters.

In Figure 4Bb, authors observed the levels of KLF2 and ETV1 in NPC and ASM deficient fibroblasts by Western blotting and quantified the bands. As ETV1 was not altered in NPC fibroblasts even though KLF2 was up-regulated, authors should explain how the differences between NPC and ASM deficient fibroblasts arose. Silencing experiments of KLF2s are also required to clarify whether increased KLF2 represses mitochondrial biogenesis in NPC fibroblasts as Figure 5.

Authors described that KLF2 is upregulated post-translationally, but not transcriptionally, through deactivation of Akt in ASM deficient fibroblasts (Figure 6B). Akt transcriptionally regulates Klf2 expression in activated CD8+ T cells (Skon, Nat Immunol, 2013) and cited articles (Sinclair et al., 2008; Skon et al., 20013) did not show proteasomal degradation of KLF2 through Akt. Akt signaling can elicit proteasomal degradation of FoxO1, an inducer of Klf2 transcription (Plas, D.R. and Thompson, C.B, 2003, J. Biol. Chem.). Posttranslational regulation of KLF2 by Akt in Niemann-Pick fibroblasts needs to be analyzed properly.

Authors described KLF2 induces ETV1 through ERK signaling, however ERK activation by KLF2 shown in Figure 8A was not assessed in this article. In order to clarify whether KLF2 induces ERK activation, authors should test the ERK phosphorylation by KLF2 silencing in ASM cells.

In Figure 7G and H, authors show that an S1PR1 agonist leads to KLF2 inhibition, whereas S1PR1 inhibition increases KLF2 in control fibroblasts. Is ETV1 is also altered in control fibroblasts treated with Sew2871 and W146?

In Figure 8D, authors showed S1PR1 at the plasma membrane in ASM-deficient cells was less than control cells. Since they could detect S1PR1 and the reduction of S1PR1 by FTY720 treatment in ASM-deficient cells, they should revise the text of "in patient cells, S1PR1 is undetectable at the plasma membrane" in the Abstract.

Minor points:

1) Legend of Figure 3D is missing.

2) Kirkegaard et al., 2010 analyzed Niemann-Pick disease (NPD) A and B fibroblasts but not type C. Authors should cite appropriate references for NPC fibroblasts.

3) In Baena et al., Genes Dev, 2013, they took advantage of ChIP-on-chip with LNCaP cells, but not ChIpSeq. I would ask the authors to correct the figure legend of Supplementary Figure 9. Authors also should show the accession code of the dataset.

4) As two siRNA duplex sequences of KLF2 and ETV1, respectively, are shown in Supplementary file 4, authors should clarify which siRNAs they used in this research.

5) (B) and (F) in the legends of Figure 5 should be modified as (C) and (D).

6) References of Oninla et al., 2014, Lu et al., 2015, Zhu et al., 2016, and Cho et al., 2015 are missing. References of (53) and (54) in Materials and methods should be shown properly.

7) 50μM chloroquine (Σ, C6628) in Material and methods is not used in this article.

8) Nomenclature and mixed cases should be unified.

Reviewer #2:

In the present study, Yambire et al. discover that aberrant S1P signaling via S1PR1 and dysregulation of its downstream targets ETV1 and KLF2 cause mitochondrial abnormalities in Niemann-Pick animal models and patient cells (NPC1 and ASM mutants). Analysis of organelle specific gene lists in these mutants situations revealed, besides a to-be-expected lysosomal phenotype, an overall reduction of mitochondria-relevant genes, likely impacting on mitochondrial biogenesis and function. Further analysis yields the candidate transcription factors ETV1 and KLF2, which were subsequently analyzed in detail with respect to mitochondrial gene regulation and impact on mitochondrial function. A candidate approach to search for relevant regulators of ETV1 and KLF2 revealed the involvement of S1P signaling in mitochondrial phenotype induction and progression.

Overall, the study by Yambire et al. discovers a pathway contributing to NPC pathology and provides a novel explanation for mitochondrial abnormalities observed in this disease, which is an important field-specific contribution. Furthermore, their organelle gene list approach identifies the first known negative regulator of overall mitochondrial function, making the study interesting for a broader audience as well.

While I do not doubt the overall findings of the study and their relevance for the field, I have some concerns with regard to data presentation issues and unclear information about biological and technical replicates, especially in case of real time RT PCR results.

For example, in Figure 2A-D (real time data of mitochondrial gene expression in NPC and ASM mutant tissues (mice) and cells (patients)), it is stated that "one experiment out of two" is shown – what constitutes an experiment, and why is only one of those shown? Why not combine all biological replicates into one graph? Or is "one experiment" a technical replicate? What is N=8 in Figure 2A? Number of mice? What n is underlying Figure 2B? (three plates, therefore I assume n=3?). (should be n, not capitalized N, for n=sample size, which normally should represent biological replicates). What is the difference between "independent experiment" and "independent replicates" in the second sentence of the Figure 2 legend? Similar issues arise in other figures (e.g., Figure 4D). The Materials and methods section "Statistical Analysis" is also not quite clear on biological vs. technical replicates. Figure 3—figure supplement 1A does not give any information on sample size. Additionally, the use of standard deviation and s.e.m. changes between experiments, and it is unclear to the reader what the rationale between these choices is.

The presentation of real time PCR data was calculated as ΔΔCT (mentioned in the figure legend of Figure 2, not in the Materials and methods section, which I would have preferred). However, that method normally results in -fold change values (ΔΔCT = log2(fold-change)), which should be used in the y axis title (instead of "Gene expression/ reference gene"). -fold change values also mean that the control condition (in Figure 2A, that would be NPC1+/+) is automatically 1 (as the control condition is not changed relative to the control condition, therefore the -fold change is 1). A standard deviation of this value does therefore not make sense, unless all the data was normalized to one biological replicate (which is not explained in the legend, nor the Materials and methods section). In Figure 1B and C (and Figure 1—figure supplement 1), the y axis title states "Fold change (% of control)", which is a contradiction – it can be either -fold change, or% of controls. Furthermore, the graph seems to show neither of those two possibilities, but "difference to control in% ", which would explain the negative values. In Figure 4—figure supplement 3, a log FC (I assume log -fold change?) of overall gene expression is given in the figure – which would mean that the average -fold change for mitochondrial genes is 1.25fold and for respiratory chain 1.4fold, not 10% and 15% respectively, which is stated in the legend. The normalization to KLF2+/+ would indeed result in a log -fold change of 0 for KLF2+/+, which fits to the graph. Which one is correct, the graph or the legend?

Strictly speaking, HPRT and GAPDH are "reference genes", not "control genes" (e.g., figure legend of Figure 2).

In Figure 4C (quantification of bands), the normalized band intensity is given, but it is unclear to what it was normalized (I assume to the empty vector control band intensity?).

Figure 7 G and H are unclear to me – normalization to control would imply values relative to 1, not to 0 (band intensity (experiment)/band intensity (control)=0 would mean no band in experiment). What exactly is depicted? Band intensity (experiment) – band intensity (control), i.e., not-normalized values?

While the study is conducted in some depth, I would have liked to see analysis of Porin protein amounts or citrate synthase activity as an estimate of mitochondrial abundance/mass. Tfam expression and mtDNA only give a good estimate for mtDNA copy number, which not always correlates with mitochondrial mass and biogenesis. Secondly, it would be interesting to see where the S1PR1 receptor is located in ASM mutant cells with an antibody staining, as a relatively easy experiment to confirm and expand the FACS data (Figure 8D). Lastly, in subsection “KLF2 regulates ETV1 in an ERK-dependent manner”, AMPK activity is estimated by analyzing ACC phosphorylation as an AMPK target. I wonder why the authors did not check also for AMPK phosphorylation itself, as ACC phosphorylation is not always a direct measure for AMPK activity, as far as I know.

Sometimes there is a tendency to overstate results slightly. For example, in subsection “Silencing of KLF2 and ETV1 in ASM deficiency rescues mitochondrial biogenesis and function” the authors claim that protein levels of mitochondrial genes are rescued by ETV1 and KLF2 overexpression, but only Tfam was analyzed in this respect. In subsection “KLF2 regulates ETV1 in an ERK-dependent manner” (concerning Figure 6A), it is stated that ETV1 knockdown has no effect on KFL2 expression, which I do not agree with, although the effect is much smaller than the other way around. While this does not really change overall interpretation of the results, the change is statistically significant according to the graph.

I am a bit unclear about the paragraph in subsection “KLF2 and ETV1 are up-regulated in NPC1-/- tissues and repress transcription of mitochondria-associated genes”. What was the result of the IPA (in terms of what was determined – TFs that are differentially expressed in NPC1 mutant tissues, or relevant/enriched pathways with THEIR according TFs, i.e. TFs with predicted higher activity in the analyzed tissue?). In Figure 4A, the former is described ("Active/repressed TFs in NPC1-/- bain and liver", left circle), but the text implies the latter (subsection “KLF2 and ETV1 are up-regulated in NPC1-/- tissues and repress transcription of mitochondria-associated genes” paragraph two). The Materials and methods section is also not totally clear on this (statistically enriched in terms of what? Expression levels of the TFs? Number of pathway targets differentially expressed?).

Lastly, I was wondering why NRF1 (and 2) did not show up in the initial screening for candidates as described in Figure 4A. They should recognize cis-elements in mitochondrial genes, and I would have expected them to come up in an ingenuity pathway analysis. Maybe the authors could discuss this?

Minor Comments:

• Subsection “Expression of mitochondria-related genes is decreased in NPC1-/- tissues”: Hard to find the "gene lists", as they are not referenced the first time they are mentioned in the results

• Subsection “KLF2 and ETV1 are up-regulated in NPC1-/- tissues and repress transcription of mitochondria-associated genes”, final paragraph: "Given that ETV1 and KLF2 are predicted by our promoter analysis to have binding sites in.…" – to have binding sites implies the presence of these sites in the enhancer of ETV1 and KLF2 – I suggest "to recognize binding sites…".

• Subsection “KLF2 regulates ETV1 in an ERK-dependent manner”: Figure not cited (…as we have shown above, ETV1 is regulated at transcript level.…) (found in Figure 4—figure supplement 1)

• Subsection “KLF2 regulates ETV1 in an ERK-dependent manner”: ERK is not an effector (i.e. downstream), but a regulator of ETV1?

• Figure 1A: type very small, hard to read. The same is true for Figure 8A

• Figure 5 is numbered f instead of d and b instead of c

• Figure 7: It would be helpful to indicate in the figure itself that Sew2871 is an activator, and W146 an inhibitor of S1PR1 (in the title of a and b, for example). In subsection “S1PR1 signaling dynamically regulates KLF2 and mitochondrial biogenesis and function”, Figure 7 B is not cited in main text. I would prefer if all Sew2871 data would be in the left column and all W146 in the right column for easier grasp of concept.

• Table 1 was hard to find (in the pdf, it is not in the Materials and methods section, but at the end of the pdf).

• Accession number of KLF2-/- mouse dataset is missing in Materials and methods

• Figure 6A: asterisks for significance unclear (*** and ***** defined in legend, but there is ****** in figure – or is that 2x*** for two adjacent comparisons?)

• An explanatory description of Figure 9 would be helpful

Additional data files and statistical comments:

See major comments. I suggest to provide all source data for original experiments (mainly real time RT PCRs and WB band intensity plots, including uncropped Western blots as pictures).

Reviewer #3:

Yambire and colleagues address an underlying cause of the previously documented mitochondrial dysfunction that occurs in lysosomal storage diseases. They report the intriguing findings that in Niemann Pick Disease, deactivation of sphingosine receptor reduces mitochondrial biogenesis by activating the repressive transcription factors KLF2 and ETV1. The elucidation of a network of transcription factors that presumably tunes mitochondrial biogenesis (Nrf1 is the activator, KLF2/ETV1 are repressors) is certainly of interest, especially in the context of lysosomal storage disease. The authors demonstrate that during lysosomal stress, cells harbor fewer mitochondrial proteins with reduced activity. The data are clearly presented and the manuscript nicely written. I have a couple of modest concerns.

Does de-repression of KLF2/ETV1 affect any aspect of modeled Niemann Pick Disease? The authors write the manuscript as to suggest that increased mitochondrial biogenesis will improve aspects of disease. However, it may not. The KLF2 and ETV1 knockdown cells should be well-suited for such studies.

What is the relationship between S1P and mitochondria? Presumably the repression of mitochondrial biogenesis during S1PR impairment evolved for some reason other than to cause mitochondrial dysfunction in Niemann Pick Disease. Some additional discussion related to this would be appreciated.

[Editors' note: further revisions were suggested, as described below.]

Thank you for resubmitting your work entitled "Mitochondrial biogenesis is transcriptionally repressed in lysosomal lipid storage diseases" for further consideration at eLife. Your revised article has been favorably evaluated by Ivan Dikic (Senior Editor), a Reviewing Editor, and three reviewers.

The manuscript has been improved but there are some remaining issues and concerns with respect to statistics and data presentation that need to be addressed before publication, as outlined in the reviewer comments below:

• The statistics paragraph of Materials and methods is still very short and lacks some general considerations about the statistics used. It also does not cite the Cumming, Fidler and Vaux 2007 paper referenced in the point to point reply to my previous comments.

• In fact, Cumming, Fidler and Vaux 2007 do not encourage the usage of standard deviation. Furthermore, since almost all original experiments show rather small sample sizes (n=2-5) after clarification of biological vs. technical replicates, individual data points should be depicted, and the use of st.dev., s.em. and t-test is not appropriate, as is stated in Cumming, Fidler and Vaux (Rule 4) and many other guidelines on statistics for biologists. (It is not appropriate to use (inferential) statistics on technical replicates, as they do not influence sample size and all parameters calculated from n, like st. dev., s.e.m or p-value – which means that the sample sizes are in almost all experiments quite small, sometimes only n=2 or 3 in these cases, error bars and p-values are dubious.)

• Figure 1: as I have stated before, "fold change (% of controls)" does not make sense, as fold change means that the experiment is x-fold increased relative to control, i.e. a fold change of 2 means it is doubled, a fold change of 0.5 means its only half of the control condition. This cannot be expressed as% , which is of course also a term for change, but not for FOLD change. I suggest again to write "difference to control (%)". There is no explanation about error bar usage (Rule 1 in Cumming, Fidler and Vaux). There is no number of n given (Rule 2 in Cumming, Fidler and Vaux).

• Figure 1—figure supplement 1: see figure 1.

• Figure 4—figure supplement 3 (previous Figure 8) – figure legend is still incorrect – a log FC of 0.1 is NOT 10% increase! (10 to the power of 0.1 = 1.25-fold change, which means an increase of expression of +25%, or to 125% of control). I stated this in the first review and am disappointed that it is not corrected.

• Figure 4D (prev. 4C): I would prefer if the answer with respect to normalization from the point to point reply were also stated in the figure legend (band density, normalized to empty vector control).

Furthermore, we very strongly encourage you to provide original data, for example blot quantifications and numbers behind the charts to support the understanding and replication of the findings.

https://doi.org/10.7554/eLife.39598.040

Author response

Reviewer #1:

[…] This article is interesting because it shows for the first-time that KLF2 and ETV1 are transcriptional repressors for mitochondrial biogenesis and that lysosomal dysfunction affects mitochondria via cellular signaling beyond autophagy. While certain conclusions are supported by the presented data, this manuscript requires further experimentation to confirm and clarify key results.

We have performed a number of new experiments and added several new pieces of data, detailed below.

Specific points:

In this article, authors analyze two acid sphingomyelinase deficient patient cell lines. To clarify which cell line is analyzed in each experiment, authors are requested to name and specify them in figures. In Corcelle-Termeau et al., 2016, three NPA fibroblasts were analyzed. Authors should describe which cell line they analyzed in this paper. Is it possible to show mutation sites of the in-house NPA fibroblast?

We have now added the mutation sites of the in-house NPA fibroblasts. The figures now specify which results are from ASM-1 (the in-house NPA fibroblasts) or from ASM-2 (NPA fibroblasts obtained from Coriell Institute). We have also added in the Materials and methods all the cell lines used in this study.

In Figure 3, authors analyzed only a single cell line from an ASM-deficient patient. It is not clear whether the reduced OCR and mtDNA level are specific features of this cell line or general. Authors should analyze mitochondrial functions in other fibroblasts including GM18398 as in Figure 3. Rescue experiments with wild type sphingomyelinase and NPC1 are also needed to clarify causal relationships.

We have analyzed the key mitochondrial phenotypes in all ASM and NPC lines. Mitochondrial biogenesis and superoxide levels are shown in Figure 2, the OCR is shown in Figure 3. The results were similar to the ASM-1 line that we showed in the previous version. In addition, rescue experiments with ASMwt and NPC1wt, showed increased mitochondrial gene expression, associated with increased OCR in all the patient lines – these results are now included in Figure 3—figure supplement 1C-E.

In Figure 3A, authors measured OCR by Seahorse and showed quantifications in the bar graphs. Although control and ASM cells supposedly responded to antimycin and rotenone, the observed values for the ASM cells are considerably lower than control cells. Authors need to discuss why non-mitochondrial respiration is lower in ASM cells. In order to identify mitochondrial OCR, non-mitochondrial respiration should be subtracted from other measurements (Figure 3B, Figure 3—figure supplement 3D, Figure 3—figure supplement 4D, 5D, 7D, and 7F).

We thank the referee for noting this unclear point. The mitochondrial OCR calculations have been updated with subtraction of non-mitochondrial OCR. The real-time respirometry approach with the SeaHorse Extracellular Flux Analyzer typically results in a minor (usually negligible) O2 consumption rate when the respiratory chain is blocked with antimycin and rotenone. We have optimized the concentrations of the reagents for the SeaHorse assay (oligomycin, FCCP, antimycin and rotenone) for fibroblast cells, several years ago, and use always the same, for standardization purposes. So the protocol is not optimized for every single cell line, otherwise it wouldn’t be possible to compare them. As for the reason why there is a slight difference in non-mitochondria O2 consumption between ASM-1 and controls, we have not addressed that in particular. A possible explanation is that the concentrations of antimycin and rotenone used did not inhibit 100% of the activity of the respiratory chain, and since there is higher O2 consumption in control cells than ASM-deficient, there would still be a slight difference if only 5% or 1% of the electrons are passing through inhibited complexes I and III. It could also be that enzymes that consume O2 in the cytoplasm (e.g., dioxigenases such as prolyl hydroxylases, etc.) are less abundant or active in the ASM-deficient cells, or that enzymes that produce O2 (e.g., catalase) are more active.

In order to check if the respiratory chain defects are observed at the protein level, the authors should clarify the abundance of respiratory chain components by SDS-PAGE and Western blot analysis.

We thank the referee for raising this issue. We have performed several experiments to tackle this question, and have included new data on the protein levels of mitochondrial respiratory chain subunits in all lines tested. In addition, we verified mitochondrial mass by flow cytometry. These new results are shown in the Figure 3—figure supplement 2. Results from all lines irrespective of technique used, showed that mitochondrial mass is not reduced unlike at transcript level. The likely reason been that, as defective lysosomal storage disorders, the patient lines accumulate non-dysfunctional mitochondria due to the inability to complete mitophagy. We validated this reasoning by flow cytometry where we assessed the amount of mitotracker green (potential insensitive) vs red (functional mitochondria) and found significantly increased amount of dysfunctional mitochondria (mitotracker red negative but positive for mitotracker green) in all the patient lines. Importantly, in control fibroblasts without defective autophagy, manipulating S1PR1 activity and subsequent differential regulation of KLF2 and ETV1 levels, alters mitochondrial biogenesis both at transcript and at protein levels (Figure 7G-H). We have now addressed the discrepancy between mitochondrial biogenesis (transcript levels) and mitochondrial mass (result of the balance between decreased mitochondrial biogenesis and severely impaired mitophagy).

In the final paragraph of subsection “Impaired mitochondrial respiration in NPC and ASM deficiency”, authors claimed U18666A treatment results in accumulation of lysosomal sphingomyelin. However, a previous report (Baulies et al., 2015, Scientific Reports) showed U18666A did not affect lysosomal distribution of sphingomyelin. Authors are requested to show appropriate references for the comment. Authors also describe that desipramine induces mitochondrial dysfunction. Since desipramine is an inhibitor of both acid sphingomyelinase and acid ceramidase, the results seem not to support the conclusion that the effects of mitochondrial impairment is independent of acid ceramidase activity as described in the last sentences.

We thank the referee for noting that we were not clear on this point. We intended to state that there are biochemical overlaps between ASM and NPC disorders with secondary accumulation of cholesterol and sphingomyelin respectively as suggested previously in Platt, 2014, and Schuchman and Desnick, 2017. That notwithstanding, experimental evidence by Santos et al., 2015 showed reduced sphingomyelinase activity in rat astrocytes treated with U18666A. It is also worthy of note that Baulies et al., 2017 demonstrated convincingly that sphingomyelin homeostasis was not altered following 24h of U18666A treatment in primary mouse hepatocytes. Given that we treated human fibroblasts with U18666A for 72h, it is likely that the different models and times of treatment used, contributed to the discrepancies in our findings. Nevertheless, we have modified the sentence to make it less definite.

Authors found the enrichment of KLF2 and ETV1 in mitochondrial promoters by in silico analysis and analyzed ETV1 ChIp targets based on the data of Baena et al., 2013. In order to clarify the results, authors should perform ChIP-qPCR and luciferase reporter assay to verify the direct interaction of the transcription factors to some representative mitochondrial promoters.

We thank the referee for this suggestion. We have analyzed publicly-available Klf2 ChIP-seq data (arrayexpress E-MTAB-2365; Yeo et al., 2014). We found a large number of KLF2 targets (i.e., that meet the significance cut-off for the ChIP-seq analysis) in genes that are included in the “mitochondrial gene list”. These are now shown in Figure 4—figure supplement 4.

In Figure 4B, authors observed the levels of KLF2 and ETV1 in NPC and ASM deficient fibroblasts by Western blotting and quantified the bands. As ETV1 was not altered in NPC fibroblasts even though KLF2 was up-regulated, authors should explain how the differences between NPC and ASM deficient fibroblasts arose. Silencing experiments of KLF2s are also required to clarify whether increased KLF2 represses mitochondrial biogenesis in NPC fibroblasts as Figure 5.

We thank the referee for bringing this point up. We work with other transcription factors (TF) in the lab, and often the total protein level of the TF is not necessarily indicative of activity because of shuttling between the nucleus and the cytoplasm. For that reason, we prepared nuclear and non-nuclear “cytoplasmic” extracts, to test if KLF2 and ETV1 were enriched in the nucleus. We find that there is a clear increase of KLF2 and ETV1 in the nucleus of ASM-1 cells (now shown in Figure 4C), but also in the nucleus of ASM-2 (now shown in Figure 4—figure supplement 1) and NPC cells (now shown in Figure 4—figure supplement 1D).

We also silenced KLF2 and ETV1 in NPC fibroblasts (now shown in Figure 5F), and observed the expected increase in the expression of mitochondrial genes (now shown in Figure 5G).

Authors described that KLF2 is upregulated post-translationally, but not transcriptionally, through deactivation of Akt in ASM deficient fibroblasts (Figure 6B). Akt transcriptionally regulates Klf2 expression in activated CD8+ T cells (Skon, Nat Immunol, 2013) and cited articles (Sinclair et al., 2008; Skon et al., 20013) did not show proteasomal degradation of KLF2 through Akt. Akt signaling can elicit proteasomal degradation of FoxO1, an inducer of Klf2 transcription (Plas, D.R. and. Thompson, C.B, 2003, J. Biol. Chem.). Posttranslational regulation of KLF2 by Akt in Niemann-Pick fibroblasts needs to be analyzed properly.

We thank the referee for pointing out this unclear point. We have revised the text accordingly to avoid suggesting that Akt directly targets KLF2 for proteosomal degradation. Pursuing the exact mechanism of KLF2 post-translational regulation is a very interesting question, which we are working on, but this manuscript already covers so many aspects that we feel adding another layer of detail in this particular subject would decrease the readability of the manuscript.

Authors described KLF2 induces ETV1 through ERK signaling, however ERK activation by KLF2 shown in Figure 8A was not assessed in this article. In order to clarify whether KLF2 induces ERK activation, authors should test the ERK phosphorylation by KLF2 silencing in ASM cells.

We thank the referee for this comment, which results from lack of clarity in the way we described the KLF2-ETV1 interaction. We don’t intend to suggest that KLF2 induces ERK activation, at least we do not yet have the experimental evidence to say so. Rather, we imply that the induction of ETV1 by KLF2 requires ERK activation, which when inhibited (U0126) in Figure 6B abrogates ETV1 induction by KLF2. We have modified the text accordingly.

In Figure 7G and H, authors show that an S1PR1 agonist leads to KLF2 inhibition, whereas S1PR1 inhibition increases KLF2 in control fibroblasts. Is ETV1 is also altered in control fibroblasts treated with Sew2871 and W146?

This is an excellent point. We have now included data on ETV1 and VDAC1 levels in Sew2871 and W146 treated fibroblasts. ETV1 behaves in complete agreement with KLF2, and VDAC behaves in agreement with the mitochondrial proteins (opposite to KLF2 and ETV1).

In Figure 8D, authors showed S1PR1 at the plasma membrane in ASM-deficient cells was less than control cells. Since they could detect S1PR1 and the reduction of S1PR1 by FTY720 treatment in ASM-deficient cells, they should revise the text of "in patient cells, S1PR1 is undetectable at the plasma membrane" in the Abstract.

Again, thank you for pointing out our incorrect wording. The text has been revised accordingly, we now state that S1PR1 is barely detectable in patient cells.

Minor points:

1) Legend of Figure 3D is missing.

Corrected accordingly.

2) Kirkegaard et al., 2010 analyzed Niemann-Pick disease (NPD) A and B fibroblasts but not type C. Authors should cite appropriate references for NPC fibroblasts.

Point noted and corrected.

3) In Baena et al., Genes Dev, 2013, they took advantage of ChIP-on-chip with LNCaP cells, but not ChIpSeq. I would ask the authors to correct the figure legend of Figure 4—figure supplement 4. Authors also should show the accession code of the dataset.

Reviewer’s point is noted and corrected accordingly. Accession code for datasets has been included.

4) As two siRNA duplex sequences of KLF2 and ETV1, respectively, are shown in Supplementary file 4, authors should clarify which siRNAs they used in this research.

Both siRNA duplexes were combined for efficient knockdowns.

5) (B) and (F) in the legends of Figure 5 should be modified as (C) and (D).

Legends have been updated.

6) References of Oninla et al., 2014, Lu et al., 2015, Zhu et al., 2016, and Cho et al., 2015 are missing. References of (53) and (54) in Materials and methods should be shown properly.

References now updated

7) 50μM chloroquine (Σ, C6628) in Material and methods is not used in this article.

Point well noted, and chloroquine is taken out of the article.

8) Nomenclature and mixed cases should be unified.

We have now unified the nomenclature.

Reviewer #2:

[…] While I do not doubt the overall findings of the study and their relevance for the field, I have some concerns with regard to data presentation issues and unclear information about biological and technical replicates, especially in case of real time RT PCR results.

We thank the referee for pointing out that the legends were not clear. We have now revised the legends accordingly, as detailed below.

For example, in Figure 2A-D (real time data of mitochondrial gene expression in NPC and ASM mutant tissues (mice) and cells (patients)), it is stated that "one experiment out of two" is shown – what constitutes an experiment, and why is only one of those shown? Why not combine all biological replicates into one graph? Or is "one experiment" a technical replicate? What is N=8 in Figure 2A? Number of mice? What n is underlying Figure 2B? (three plates, therefore I assume n=3?). (should be n, not capitalized N, for n=sample size, which normally should represent biological replicates). What is the difference between "independent experiment" and "independent replicates" in the second sentence of the Figure 2 legend? Similar issues arise in other figures (e.g., Figure 4D). The Materials and methods section "Statistical Analysis" is also not quite clear on biological vs. technical replicates. Figure 3—figure supplement 1A does not give any information on sample size. Additionally, the use of standard deviation and s.e.m. changes between experiments, and it is unclear to the reader what the rationale between these choices is.

We thank the referee for this comment. We agree that the legends were over-complicated and unclear. Legends have been revised to clarify this confusing description. All the N= are now shown as n= and indeed they correspond to a biological replicate (mouse, or cell plate grown independently). Independent replicates means different biological samples (e.g., eight mice are eight different biological replicates; three biological replicates for cells means three plates of cells, collected and processed independently, e.g., three protein extracts or three RNA extracts, etc..). Regarding the standard deviation and the standard error of the mean, we follow the guidelines by Cumming, Fidler and Vaux (2007), J. Cell Biol.. Specifically, for experiments in which we have a moderate number of biological replicates and technical replicates, we use the standard deviation, because it is a descriptive error bar. However, when the sample/replicate number is higher (we set an arbitrary number of at least 20), such as in the analysis of the gene lists (n is in the hundreds) or when using plate-based assays (plate reader, respirometry), we have a large number of technical replicates, and in these cases the mean of the result is much closer to the mean of the population, and thus we use inferential error bars, i.e., the standard error of the mean.

The presentation of real time PCR data was calculated as ΔΔCT (mentioned in the figure legend of Figure 2, not in the Materials and methods section, which I would have preferred). However, that method normally results in -fold change values (ΔΔCT = log2(fold-change)), which should be used in the y axis title (instead of "Gene expression/ reference gene"). -fold change values also mean that the control condition (in Figure 2A, that would be NPC1+/+) is automatically 1 (as the control condition is not changed relative to the control condition, therefore the -fold change is 1). A standard deviation of this value does therefore not make sense, unless all the data was normalized to one biological replicate (which is not explained in the legend, nor the Materials and methods section).

qCPR data analyses now moved to Materials and methods section and text revised for clarity. Indeed, when performing a qPCR we have multiple biological replicates of the control samples, we set one of them as reference sample and normalize all the other samples (including the other control replicates) to that sample – that is where the standard deviation comes from. This is now clarified in the Materials and methods.

In Figure 1B and C (and Figure 1—figure supplement 1), the y axis title states "Fold change (% of control)", which is a contradiction – it can be either -fold change, or% of controls. Furthermore, the graph seems to show neither of those two possibilities, but "difference to control in% ", which would explain the negative values. In Figure 4—figure supplement 3, a log FC (I assume log -fold change?) of overall gene expression is given in the figure – which would mean that the average -fold change for mitochondrial genes is 1.25fold and for respiratory chain 1.4fold, not 10% and 15% respectively, which is stated in the legend. The normalization to KLF2+/+ would indeed result in a log -fold change of 0 for KLF2+/+, which fits to the graph. Which one is correct, the graph or the legend?

Indeed, logFC means log of the fold change. The text has been rewritten for clarity.

Strictly speaking, HPRT and GAPDH are "reference genes", not "control genes" (e.g., figure legend of Figure 2).

HPRT and GAPDH are now referred to as reference genes in text.

In Figure 4C (quantification of bands), the normalized band intensity is given, but it is unclear to what it was normalized (I assume to the empty vector control band intensity?).

Normalized to empty vector control.

Figure 7 G and H are unclear to me – normalization to control would imply values relative to 1, not to 0 (band intensity (experiment)/band intensity (control)=0 would mean no band in experiment). What exactly is depicted? Band intensity (experiment) – band intensity (control), i.e., not-normalized values?

Point noted and revised as normalized band density (difference from vehicle treatment).

While the study is conducted in some depth, I would have liked to see analysis of Porin protein amounts or citrate synthase activity as an estimate of mitochondrial abundance/mass. Tfam expression and mtDNA only give a good estimate for mtDNA copy number, which not always correlates with mitochondrial mass and biogenesis. Secondly, it would be interesting to see where the S1PR1 receptor is located in ASM mutant cells with an antibody staining, as a relatively easy experiment to confirm and expand the FACS data (Figure 8D). Lastly, in subsection “KLF2 regulates ETV1 in an ERK-dependent manner”, AMPK activity is estimated by analyzing ACC phosphorylation as an AMPK target. I wonder why the authors did not check also for AMPK phosphorylation itself, as ACC phosphorylation is not always a direct measure for AMPK activity, as far as I know.

This point is well taken, and responded to above since reviewer 1 also had similar concerns. AMPK phosphorylation has been included in the data in Figure 6B.

Sometimes there is a tendency to overstate results slightly. For example, in subsection “Silencing of KLF2 and ETV1 in ASM deficiency rescues mitochondrial biogenesis and function” the authors claim that protein levels of mitochondrial genes are rescued by ETV1 and KLF2 overexpression, but only Tfam was analyzed in this respect.

We stated that TFAM was just an example of such rescue following ETV1 and KLF2 silencing (not overexpression). In the figures, we also show that NRF1 levels are rescued. Given the confounding effect of defective mitochondrial clearance (Figure 3—figure supplement 2) and the fact that ETV1 and KLF2 silencing does not alter defective autophagy in these models (Figure 5—figure supplement 1), assessing mitochondrial proteins directly was found to be not useful. However, in Figure 7, in the absence of the confounding effect caused by impaired autophagy, it is clearly shown that all respiratory chain subunits tested behave similarly to observed changes in transcript levels.

In subsection “KLF2 regulates ETV1 in an ERK-dependent manner” (concerning Figure 6A), it is stated that ETV1 knockdown has no effect on KFL2 expression, which I do not agree with, although the effect is much smaller than the other way around. While this does not really change overall interpretation of the results, the change is statistically significant according to the graph.

We revised the wording, and it is now stated as negligible effect.

I am a bit unclear about the paragraph in subsection “KLF2 and ETV1 are up-regulated in NPC1-/- tissues and repress transcription of mitochondria-associated genes”. What was the result of the IPA (in terms of what was determined – TFs that are differentially expressed in NPC1 mutant tissues, or relevant/enriched pathways with THEIR according TFs, i.e. TFs with predicted higher activity in the analyzed tissue?). In Figure 4A, the former is described ("Active/repressed TFs in NPC1-/- bain and liver", left circle), but the text implies the latter (subsection “KLF2 and ETV1 are up-regulated in NPC1-/- tissues and repress transcription of mitochondria-associated genes” paragraph two). The Materials and methods section is also not totally clear on this (statistically enriched in terms of what? Expression levels of the TFs? Number of pathway targets differentially expressed?).

The text has now been revised for clarity. The TFs identified by IPA are filtered based on statistically over-represented targets in our gene lists, using Fisher’s exact test. Depending on the known effect of the TF on each target (activation/repression), it is then calculated the Z-score, which determines if the TF is predicted to be active or repressed. The TFs themselves are not necessarily differentially expressed, since many of these proteins are not regulated at transcript level, so we didn’t use this as a filter criterion.

Lastly, I was wondering why NRF1 (and 2) did not show up in the initial screening for candidates as described in Figure 4A. They should recognize cis-elements in mitochondrial genes, and I would have expected them to come up in an ingenuity pathway analysis. Maybe the authors could discuss this?

Cis-elements for NRF1 scored in many respiratory chain genes, but not in many others, and didn’t meet the statistical cut-off (Fisher’s exact test). While it is usually assumed that NRF1 triggers the expression of all mitochondrial genes, that is a notion that was extrapolated out of a few genes encoding mitochondrial proteins, and that does not necessarily hold true for the whole mitochondrial proteome, or even for the whole respiratory chain. NRF2 (GABPA) had less cis-elements than NRF1 in the respiratory chain subunits promoters. We didn’t discuss this in the text because we took un unbiased approach, and therefore to discuss specifically NRF1 or NRF2 would leave out several other TFs that trigger some aspects of mitochondrial biogenesis. We are addressing the transcriptional regulation of mitochondrial biogenesis in a systematic manner in the context of another project.

Minor Comments:

• Subsection “Expression of mitochondria-related genes is decreased in NPC1-/- tissues”: Hard to find the "gene lists", as they are not referenced the first time they are mentioned in the results.

Gene list referenced sooner for clarity.

• Subsection “KLF2 and ETV1 are up-regulated in NPC1-/- tissues and repress transcription of mitochondria-associated genes”, final paragraph: "Given that ETV1 and KLF2 are predicted by our promoter analysis to have binding sites in.…" – to have binding sites implies the presence of these sites in the enhancer of ETV1 and KLF2 – I suggest "to recognize binding sites…".

Point noted and revised as suggested.

• Subsection “KLF2 regulates ETV1 in an ERK-dependent manner”: Figure not cited (…as we have shown above, ETV1 is regulated at transcript level.…) (found in Figure 4—figure supplement 2).

This point is well noted and Figures have been cited again and again for clarity.

• Subsection “KLF2 regulates ETV1 in an ERK-dependent manner”: ERK is not an effector (i.e. downstream), but a regulator of ETV1?

Point noted. ERK is indeed a regulator.

• Figure 1A: type very small, hard to read. The same is true for Figure 8A

Revised.

• Figure 5 is numbered F instead of D and B instead of C.

Updated.

• Figure 7: It would be helpful to indicate in the figure itself that Sew2871 is an activator, and W146 an inhibitor of S1PR1 (in the title of a and b, for example). In subsection “S1PR1 signaling dynamically regulates KLF2 and mitochondrial biogenesis and function”, Figure 7 B is not cited in main text. I would prefer if all Sew2871 data would be in the left column and all W146 in the right column for easier grasp of concept.

Figures rearranged, and 7B cited.

• Table 1 was hard to find (in the pdf, it is not in the Materials and methods section, but at the end of the pdf).

Tables have been rearranged accordingly.

• Accession number of KLF2-/- mouse dataset is missing in Materials and methods

All accession numbers for datasets used in this study have now been included.

• Figure 6A: asterisks for significance unclear (*** and ***** defined in legend, but there is ****** in figure – or is that 2x*** for two adjacent comparisons?)

Well noted! 2x*** for adjacent comparisons, revised accordingly.

• An explanatory description of Figure 9 would be helpful.

Graphic representation of publicly available Etv1 ChIP-chip data, highlighting how many of its target genes encode proteins involved in different aspects of mitochondrial function. Etv1 target genes from the ChIP-seq data were crossed with the mitochondrial ‘gene list’ described above to obtain targets of Etv1 which encode mitochondrial proteins. We note that NRF1 and PPARG, although they do not encode bona fide mitochondrial proteins, they were found to be direct targets of Etv1 and thus their inclusion.

Reviewer #3:

Yambire and colleagues address an underlying cause of the previously documented mitochondrial dysfunction that occurs in lysosomal storage diseases. They report the intriguing findings that in Niemann Pick Disease, deactivation of sphingosine receptor reduces mitochondrial biogenesis by activating the repressive transcription factors KLF2 and ETV1. The elucidation of a network of transcription factors that presumably tunes mitochondrial biogenesis (Nrf1 is the activator, KLF2/ETV1 are repressors) is certainly of interest, especially in the context of lysosomal storage disease. The authors demonstrate that during lysosomal stress, cells harbor fewer mitochondrial proteins with reduced activity. The data are clearly presented and the manuscript nicely written. I have a couple of modest concerns.

We thank the referee for the positive comments on our manuscript.

Does de-repression of KLF2/ETV1 affect any aspect of modeled Niemann Pick Disease? The authors write the manuscript as to suggest that increased mitochondrial biogenesis will improve aspects of disease. However, it may not. The KLF2 and ETV1 knockdown cells should be well-suited for such studies.

We thank the referee for bringing up this important point. We have addressed it with new experiments. Although the text may give that impression, it was not our intention to suggest that increasing mitochondrial biogenesis will alleviate aspects of NP disease. On the contrary, we aimed at clarifying the involvement of mitochondrial and lysosomal crosstalk in the pathophysiology of the disease. Interestingly, we indeed find that increasing mitochondrial biogenesis via KLF2 and ETV1 silencing in ASM and NPC1 models is not beneficial since it increases apoptosis with a subsequent reduction in cell viability (shown in Figure 8E-G). These findings suggest that the induction of KLF2 and ETV1 in ASM and NPC1 is a protective adaptive mechanism to promote survival with a compromising repression of mitochondrial biogenesis.

What is the relationship between S1P and mitochondria? Presumably the repression of mitochondrial biogenesis during S1PR impairment evolved for some reason other than to cause mitochondrial dysfunction in Niemann Pick Disease. Some additional discussion related to this would be appreciated.

The role of S1P signaling in mitochondrial biogenesis does not pertain only to NP disease. Indeed, in hepatocytes, S1P signaling was found to be necessary for mitochondrial biogenesis via PKA-PGC1a. Furthermore, S1P signaling in T cells was shown to be essential for the maintenance of mitochondrial content and function. As well as our results, these findings suggest that S1P signaling might be a previously underappreciated regulatory loop for mitochondrial biogenesis. Given that S1P signaling was shown to engage PGC1a, and we have shown for the first time that lack of its signaling engages repressors of mitochondrial biogenesis, S1P signaling might be a crucial denominator for maintaining mitochondrial content. We have now addressed this question in the discussion.

[Editors' note: further revisions were suggested, as described below.]

The manuscript has been improved but there are some remaining issues and concerns with respect to statistics and data presentation that need to be addressed before publication, as outlined in the reviewer comments below:

• The statistics paragraph of Materials and methods is still very short and lacks some general considerations about the statistics used. It also does not cite the Cumming, Fidler and Vaux 2007 paper referenced in the point to point reply to my previous comments.

We have included further details in the statistical description in the Materials and methods. We must note that we provide information on the statistical treatment for every panel, in the respective figure legend. We have also included clarification on biological replicates and their technical replicates thereof.

• In fact, Cumming, Fidler and Vaux 2007 do not encourage the usage of standard deviation. Furthermore, since almost all original experiments show rather small sample sizes (n=2-5) after clarification of biological vs. technical replicates, individual data points should be depicted, and the use of st.dev., s.em. and t-test is not appropriate, as is stated in Cumming, Fidler and Vaux (Rule 4) and many other guidelines on statistics for biologists. (It is not appropriate to use (inferential) statistics on technical replicates, as they do not influence sample size and all parameters calculated from n, like st. dev., s.e.m or p-value – which means that the sample sizes are in almost all experiments quite small, sometimes only n=2 or 3 – in these cases, error bars and p-values are dubious.)

We respectfully disagree with reviewer #2 on this instance. Cumming, Fidler and Vaux neither encourage nor discourage the use of standard deviation, although they suggest that standard errors or confidence intervals can be used instead. It is also standard practice in biology to present the standard deviation, including papers signed by referee #2 (Buelow et al., MBoC 2017 and Sellin et al., PLOS Biology 2018). We must respectfully note that in this manuscript error bars are mainly presented as s.e.m. in the graphs (as opposed to the consistent miscommunication in the initial figure legends in which we identified as SD. We have therefore accurately updated the figure legends to address this pertinent concern. In support of this, we are including an image (Author response image 1) for Figure 2A,C to illustrate the point that the error bars represent s.e.m. and not SD.

Author response image 1

However, we must point out that we disagree with the conclusion of the referee regarding the number of biological replicates. We are using patient cells. So in the most extreme interpretation, everything is a technical replicate, because it comes from n=1 patient. That notwithstanding, we define biological replicates as follows: for in vivoexperiments n refers to the number of mice for each genotype used in this study and for all cell culture experiments we state that n refers to the number of independent experiments carried out with different stocks of each cell line. We also included several technical replicates for each biological experiment. This definition is standard and prescribed by others including Cumming, Fidler and Vaux. In this case, the sample size for almost all cell culture experiments in this manuscript is at least n=3, which is commonplace for such work. Despite the perceived reduced number of samples, we employed many different cell lines from the same and related genetic defects, to sustain the validity of our biological conclusions. We show that by using pharmacological approaches the same conclusions are reached. We show that by using mouse models (n=8 or 9), the same conclusions are reached. How will the same dubious conclusions be arrived at using several models and approaches? Besides, the instances with n=2 involve the robust and highly reproducible (as evidenced by the technical replicates) real-time respirometry experiments with the Seahorse FluxAnalyzer where pharmacological inhibition of ASM and NPC were carried out. We don’t find it necessary to include more replicates in all the experiments just for the sake of statistical satisfaction without recourse to the biology.

For this particular point, we choose to take advantage of the possibility associated with the “Research Communications” in which the authors decide if and how to reply to the referees’ comments. We did not apply this to the rest of the comments in the first review and in the current one, but in this particular instance we choose not to further engage in what we consider a discussion detached from the biological conclusions of our study.

• Figure 1: as I have stated before, "fold change (% of controls)" does not make sense, as fold change means that the experiment is x-fold increased relative to control, i.e. a fold change of 2 means it is doubled, a fold change of 0.5 means its only half of the control condition. This cannot be expressed as% , which is of course also a term for change, but not for FOLD change. I suggest again to write "difference to control (%)". There is no explanation about error bar usage (Rule 1 in Cumming, Fidler and Vaux). There is no number of n given (Rule 2 in Cumming, Fidler and Vaux).

We corrected the text to “Fold change (difference from WT)”. The number of genes analyzed is alluded to in the legend, and clearly detailed in the organelle proteome table and the number of biological samples (mice) is now included in the figure legends.

• Figure 1—figure supplement 1: see Figure 1.

Since the same datasets were used for the analyses, the number of biological samples is the same as stated above and corrected in the figure legends.

• Figure 4—figure supplement 3 – figure legend is still incorrect – a log FC of 0.1 is NOT 10% increase! (10 to the power of 0.1 = 1.25-fold change, which means an increase of expression of +25%, or to 125% of control). I stated this in the first review and am disappointed that it is not corrected.

We respectfully note this comment and we have corrected it accordingly.

• Figure 4D: I would prefer if the answer with respect to normalization from the point to point reply were also stated in the figure legend (band density, normalized to empty vector control).

This point is noted and included in the Materials and methods section for experiments in which we normalized the results of experimental samples to those of the corresponding controls to ensure easy comparisons.

https://doi.org/10.7554/eLife.39598.041

Article and author information

Author details

  1. King Faisal Yambire

    1. Institute of Cellular Biochemistry, University Medical Center Goettingen, Goettingen, Germany
    2. International Max-Planck Research School in Neuroscience, Goettingen, Germany
    3. European Neuroscience Institute Goettingen, University Medical Center Goettingen, Goettingen, Germany
    Contribution
    Data curation, Formal analysis, Validation, Investigation, Methodology, Writing—original draft, Project administration, Writing—review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2417-450X
  2. Lorena Fernandez-Mosquera

    Institute of Cellular Biochemistry, University Medical Center Goettingen, Goettingen, Germany
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-4606-5056
  3. Robert Steinfeld

    Klinik für Kinder- und Jugendmedizin, University Medical Center Goettingen, Goettingen, Germany
    Contribution
    Resources, Validation
    Competing interests
    No competing interests declared
  4. Christiane Mühle

    Department of Psychiatry and Psychotherapy, Friedrich-Alexander University Erlangen-Nürnberg (FAU), Erlangen, Germany
    Contribution
    Resources, Validation
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-7517-9154
  5. Elina Ikonen

    Department of Anatomy, Faculty of Medicine, University of Helsinki, Biomedicum Helsinki, Helsinki, Finland
    Contribution
    Resources, Formal analysis, Methodology
    Competing interests
    No competing interests declared
  6. Ira Milosevic

    European Neuroscience Institute Goettingen, University Medical Center Goettingen, Goettingen, Germany
    Contribution
    Validation, Methodology
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6440-3763
  7. Nuno Raimundo

    Institute of Cellular Biochemistry, University Medical Center Goettingen, Goettingen, Germany
    Contribution
    Conceptualization, Resources, Data curation, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    nuno.raimundo@med.uni-goettingen.de
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-5988-9129

Funding

Deutsche Forschungsgemeinschaft (GRK2162/1)

  • Christiane Mühle

Academy of Finland (312491)

  • Elina Ikonen

Deutsche Forschungsgemeinschaft (SFB1190)

  • Ira Milosevic
  • Nuno Raimundo

Deutsche Forschungsgemeinschaft (Emmy-Noether Award)

  • Ira Milosevic

Schram Stiftung

  • Ira Milosevic

H2020 European Research Council (337327)

  • Nuno Raimundo

Acid Maltase Deficiency Association (Research Grant)

  • Nuno Raimundo

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This research was supported by ERC Starting Grant 337327 and AMDA Research Grant (NR); Deutsche Forschungsgemeinschaft Emmy-Noether Award and Schram Stiftung Grant (IM); Deutsche Forschungsgemeinschaft SFB1190 (NR and IM); Academy of Finland Grant 312491 (EI); Deutsche Forschungsgemeinschaft (GRK2162/1), Neurodevelopment and Vulnerability of the Central Nervous System (CM).

We thank Dr. Ralf Janknecht for the ETV1 constructs and Dr. Roberto Zoncu for the NPC1 constructs.

Ethics

Animal experimentation: All mice were handled according to the rules of law under application at the University of Helsinki (Finland) and at the University of Erlangen (Germany), according to the directive 2010/63/EU.

Senior Editor

  1. Ivan Dikic, Goethe University Frankfurt, Germany

Reviewing Editor

  1. Agnieszka Chacinska, University of Warsaw, Poland

Reviewer

  1. Julia Sellin, University of Bonn, Germany

Publication history

  1. Received: July 2, 2018
  2. Accepted: February 11, 2019
  3. Version of Record published: February 18, 2019 (version 1)

Copyright

© 2019, Yambire et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,843
    Page views
  • 286
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Cell Biology
    2. Developmental Biology
    Péter Lőrincz et al.
    Research Advance
    1. Cell Biology
    2. Plant Biology
    Andreas S Richter et al.
    Research Article