Comprehensive silencing of target-sharing microRNAs is a mechanism for SIRT1 overexpression in cancer

Abstract

Overexpression of SIRT1 is frequently observed in various types of cancers, suggesting its potential role in malignancies. However, the molecular basis of how SIRT1 is elevated in cancer is less understood. Here we show that cancer-related SIRT1 overexpression is due to evasion of Sirt1 mRNA from repression by a group of Sirt1-targeting microRNAs (miRNAs) that might be robustly silenced in cancer. Our comprehensive library-based screening and subsequent miRNA gene profiling revealed a housekeeping gene-like broad expression pattern and strong CpG island-association of the Sirt1-targeting miRNA genes. This suggests aberrant CpG DNA methylation as the mechanistic background for malignant SIRT1 elevation. Our work also provides an example where epigenetic mechanisms cause the group-wide regulation of miRNAs sharing a common key target.

Abbreviations

SIRT1

sirtuin (silent mating type information regulation 2 homolog) 1 (S. cerevisiae)

Sir2

Silent information regulator 2

CGIs

CpG islands

NAD

nicotinamide adenine dinucleotide

p53

protein 53

FOXO

forkhead box O

E2F1

E2F transcription factor 1

NF-κB

nuclear factor kappa-B

PGC-1α

peroxisome proliferative activated receptor, gamma, coactivator 1 alpha

LXR

liver X receptor

MyoD

myogenic differentiation 1

pri-miRNA

primary miRNA

3′UTR

3′ untranslated region

5-Aza-dC

5-aza-2’-deoxycytidine

PUMA

p53-upregulated modulator of apoptosis

PARP

poly(ADP-ribose) polymerase-1

HIC1

hypermethylated in cancer 1

OVCA1

ovarian cancer-associated gene 1 protein

OVCA2

ovarian cancer-associated gene 2 protein

SMG6

smg-6 homolog, nonsense mediated mRNA decay factor (C. elegans)

Introduction

Tumorigenesis is the accumulation of detrimental genetic alterations which promote malignant transformation of a cell.¹ The onset and progression of tumorigenesis involve genetic changes causing inactivation of variety of tumor suppressors and the overexpression of oncogenes.² Furthermore, recent work shows that epigenetic mechanisms such as DNA methylation are also causally involved in pathogenic gene expression in malignant cells.³ Methylation of CpG islands (CGIs), genomic regions with a significantly high frequency of cytosine-guanine dinucleotides, is frequently altered in tumors.^3,4 Dense methylation of CGIs present in the promoter regions of genes in malignant cells results in silencing of their expression⁵ and could be a key pathogenic step, especially for silencing genes with tumor suppressive functions.

SIRT1, the mammalian ortholog of yeast Sir2, is an NAD⁺-dependent histone deacetylase which participates in numerous cellular events including apoptosis, stress responses, metabolism, and aging.^6-9 Through its deacetylation activity, SIRT1 modulates functions of various critical molecules such as p53, FOXO, E2F1, NF-κB, PGC-1α, LXR, and MyoD, illustrating its critical and multifaceted roles in cellular physiology.¹⁰ Given its role in the regulation of these critical targets, SIRT1 has also been linked to malignancies and is suggested to be important for cancer progression.^6-9 Notably, SIRT1 is frequently overexpressed in many types of tumors, also indicating its critical role in tumorigenesis.^11-14 Despite extensive studies on SIRT1 biology, however, the causal mechanism leading aberrant elevation of SIRT1 in cancer cells is still elusive.

MicroRNAs (miRNAs) are a growing class of functional noncoding RNAs that play a key role in the regulation of gene expression, by repressing target genes at the post-transcriptional level.¹⁵ The biogenesis of functional mature miRNAs consists of a stepwise processing of a primary miRNA transcript.¹⁶ Cumulative evidences support the idea that the expression of miRNAs is largely regulated at the step of pri-miRNA transcription, and depends on the genetic context around the miRNA gene.^17,18 Therefore, identifying and characterizing the unique genetic context associated with miRNA genes is important to understand miRNA biology.

In a cell, dozens of co-expressed miRNAs constitute a molecular system affecting various genetic pathways. Each single miRNA may regulate numerous target mRNAs¹⁵ while one mRNA can be also regulated by multiple miRNAs.¹⁹ In this way, the miRNA system modulates a broad spectrum of the genes and is a key component of the gene regulatory network. Thus, for some critical target mRNAs, it is likely that there is a group of multiple miRNAs directly regulating the same target.

In the current study, we adopt a combination of a library-based high-throughput screen and miRNA gene profiling to identify an evolutionarily conserved group of Sirt1-targeting miRNAs and show how they might be coordinately regulated. Furthermore, we propose a mechanism involving group-wide miRNA gene silencing for SIRT1 overexpression in cancer cells, thereby showing the therapeutic potential for the artificial restoration of a Sirt1-targeting miRNA against chemoresistant tumor cells.

Results

To examine potential roles of miRNAs in the regulation of SIRT1, we performed library-based screening of miRNAs targeting the 3′ UTR of human and murine Sirt1 mRNA (hSirt1 and mSirt1, respectively) (Fig. S1A). Synthetic precursors of 470 human and 379 murine miRNA species from the corresponding libraries were individually transfected in the HeLa cells stably expressing a reporter (the firefly luciferase) gene carrying the 3′UTR of hSirt1 (for the human library test) or mSirt1 (for the mouse library test) at their 3′ end, and a control reporter gene (as negative control for both library tests) (Fig. S1A).

The miRNAs which downregulated the reporter signals in the precursor library assay were considered as potential regulators of Sirt1 3′UTR. Classifying all examined miRNAs by canonical “seed” sequences (either 8mer, 7mer-m8, 7mer-A1, or 6mer; Fig. 1A)²⁰ in the 3′UTR of mouse and human Sirt1 mRNAs, the miRNAs with high affinity seed-sequences, especially 8mer seeds, tend to show repressive capacity in the 3′UTR of both hSirt1 and mSirt1, supporting the overall validity of our screening to distinguish Sirt1-targeting miRNAs.

Figure 1.

Library-based high-throughput miRNA screening identified the Sirt1-targeting miRNA group. (A) Human or mouse Sirt1 3′UTR reporter signals from the transfectants for each miRNA categorized by canonical seeds (8mer, 7mer-m8, 7mer-A, 6mer, and no site) were plotted as the cumulative fraction per bin. (B) Human and mouse Sirt1 3′UTR reporter signals from the assay are presented in a scatter plot and the correlation coefficient was calculated. Data for the miRNAs conserved between human and mouse (241 species) were shown. (C) Venn diagram for the overlap of human and mouse Sirt1 3′UTR-targeting miRNAs. (D) List of the miRNAs regulating Sirt1 3′UTR activity. The seed type (Fig. 1A) of each miRNA is shown. (E) Top ten list of the human or mouse specific miRNAs regulating Sirt1 3′UTR activity. (F) Mapping of the predicted target sites for the validated miRNAs with the relative experimental signal values. The means ± SD (n = 4) are shown.

Comparing relative effects of the miRNA species in hSirt1 and mSirt1 3′UTR regulation, a positive correlation was readily observed (Fig. 1B). Since conservation of miRNA regulation among human and mouse could indicate the biological significance of miRNA-target pairings, we selected 35 miRNAs showing repressive activity in the regulation of both hSirt1 and mSirt1 3′UTR reporters (Fig. 1C). These miRNA were also assessed for the presence of the canonical seeds in the hSirt1 and mSirt1 3′UTR, resulting in selection of 13 miRNAs (Fig. 1D and E). Furthermore, miRNA targeting sites conserved in hSirt1 and mSirt1 3′UTRs at similar positions were sought and 12 miRNA species were identified as Sirt1-targeting miRNA candidates (Fig. 1F).

We then examined whether these candidate miRNAs could regulate Sirt1 mRNA in cells and found that nine miRNAs (miR-9–5p, -22–3p, -132–3p, -199a-3p, -199a-5p, -199b-5p, -200b-3p, -212–3p, -217) among the 12 candidates could actually repress endogenous Sirt1 mRNA (Fig. S1B).

Having identified the Sirt1-targeting miRNA group, the genomic structure of each miRNA gene was then examined. Interestingly, all of the miRNA genes, except Mir-217 gene, are closely associated with CGIs and are reported to relate with CGIs (Fig. 2A; Fig. S2). Consistent with the notion that CpG islands are often located near the promoters of housekeeping genes,²¹ the CGI-associated Sirt1-targeting miRNAs are also broadly expressed in various tissues (Fig. 2B). Of note, the expression of miR-217 is exclusively detected in the pancreas,²² suggesting its role in SIRT1 regulation might be pancreas specific.

Figure 2.

For figure legend, see page 1351. Figure 2 (See previous page). CGI-association of the Sirt1-targeting miRNA group and their epigenetic regulation by DNA methylation in cancer cells. (A) The relative expression levels of the miRNAs in the cells transfected with the vectors (shown in the left side) untreated or treated with the indicated DNA methyltransferases. Barcodes shown under the vectors indicate methylation sites for the DNA methyltransferases. CpG islands were shown with green box. Data are from the real-time PCR assays. The error bars indicate SD (n = 3). *P< 0.05, **P < 0.01, ***P < 0.001 by Student's t test. (B) The expression pattern of the Sirt1-targeting miRNAs in human tissues is shown as a heat map created based on miRNAmap data. (C) The expression pattern of Sirt1-targeting miRNAs in human normal (N) and tumor (T) tissues as demonstrated in (b). Average expression of samples was shown as a heat map. Sample numbers were described under the map. (D) Upregulation of Sirt1-targeting miRNAs and downregulation of Sirt1 mRNA in 5-aza-dC-treated cancer cell lines. The expression levels of Sirt1 mRNA and the miRNAs in the indicated cell lines treated with 5-aza-dC for 4 d were measured by real-time PCR. (E) The Dicer protein levels in NB4 cells transfected with Dicer siRNA were examined by western blot analysis (upper panel). The expression levels of Sirt1 mRNA in NB4 cells treated with Dicer siRNA and 5-aza-dC were measured by real-time PCR (lower panel). The error bars indicate SD (n = 3). **P < 0.01 by a Student t test.

The close association with CGI suggests that the Sirt1-targeting miRNAs could be regulated by CpG DNA methylation. Indeed, a methylation-sensitivity assay demonstrated that the expression of all eight CGI-associated Sirt1-targeting miRNAs were highly repressed upon in vitro DNA methylation by SssI methyltransferase (Fig. 2A; Fig. S3).

Given the housekeeping gene-like expression pattern and sensitivity to DNA methylation, we then asked whether the group of Sirt1-targeting miRNAs were suppressed in various cancer types where CGIs are tend to be highly methylated. Using a bioinformatics-based expression array analysis, we found that, in comparison to healthy normal counterpart tissues, most of the Sirt1-targeting miRNAs were downregulated (Fig. 2C).

We then hypothesized that SIRT1 overexpression in many types of cancer might be partly due to epigenetic silencing of the Sirt1-targeting miRNAs. To test this, we treated various tumor cell lines with a methyltransferase inhibitor, 5-Aza-dC (5-aza-2′-deoxycytidine), to liberate the Sirt1-targeting miRNAs from epigenetic repression by DNA methylation and examined possible correlation of the Sirt1-targeting miRNA induction and endogenous Sirt1 mRNA level. In the cell lines, 5-Aza-dC treatment suppressed Sirt1 mRNA and induced Sirt1-targeting miRNAs to a various degree (Fig. 2D; Fig. S4A). Moreover, suppression of SIRT1 by 5-aza-dC was recovered when Dicer, which is a component of miRNA machinery, was depleted with siRNA (Fig. 2E). Altogether, we propose that group-wide silencing of the Sirt1-targeting miRNAs is a mechanism for SIRT1 overexpression in cancer.

To explore the physiological consequences of SIRT1 repression by the miRNA group and SIRT1 overexpression due to its deregulation in cancer, we sought to reproduce a repressive condition for SIRT1 in cancer cell lines by reintroduction of a Sirt1-targeting miRNA species. To this end, we focused on miR-132 and miR-212, which share the same seed site and are generated from a single primary transcript.²³ The human miR-132/miR-212 gene is located in a vast CpG island on 17p13.3, one of the most frequently altered chromosomal regions in variety of human cancers (Fig. 3A),²⁴ making miR-132 and miR-212 as an ideal example for testing.

Figure 3.

For figure legend, see page 1353.Figure 3 (See previous page). Evidence for the biological role of the miRNA groupregulation of SIRT1. (A) The genomic structure of the human mir-212/mir-132 locus on 17p13.3. (B) Methylation-specific PCR (MSP) for the mir-212/mir-132 and Hic1 gene promoters in different tumor cell lines (left panel). The methylation status of CpG sites close to the mir-212/mir-132 gene promoter in tumor cell lines was assessed by the bisulfite sequencing assay (right panel). Twelve CpG sites were identified and indicated as a single circle. Each horizontal line represents one sequence result in which open and closed circles denote unmethylated and methylated CpG sites, respectively. (C) A549 cells were transfected with miRNA precursors or hSIRT1 siRNA and exposed to etoposide for the indicated hours. Acethylation of the p53 tumor suppressor protein was then examined by western blotting. Re-blotting with anti-β-actin antibody was performed as a loading control. (D) A549 cells were transfected with the indicated miRNA precursors and exposed to etoposide. Caspase 3/7 activity was measured at the indicated times. Mean ± SD. Statistics vs miR-NC samples, *: P = 0.000001 (n = 4), **: P = 0.0000007 (n = 4) (E) A549 cells were transfected with miRNA precursors or hSIRT1 siRNA and exposed to etoposide for the indicated hours. PARP cleavage was then examined by western blotting with an anti-PARP antibody. Re-blotting with anti-β-actin antibody was performed as a loading control. (F) A549 cells were untreated or transfected with indicated miRNA precursors or hSIRT1 siRNA, treated with Etoposide, and monitored for cell viability by a cell-titer assay. Mean ± SD. Statistics vs miR-NC samples, *: P = 0.0000014 (n = 3), **: P = 0.00000008 (n = 3), ***: P = 0.00000007 (n = 3).

The CpG sites near the promoter of mir-132/mir-212 gene are densely methylated in several tumor cell lines (Fig. 3B). DNA methylation in this region is quite rare in the genomic DNA sample from a healthy donor, as revealed by methylation specific PCR and also by direct bisulfite sequencing of the sequences proximal to the promoter (Fig. 3B). In agreement with this, miR-132 expression is extremely low in tumor cell lines (Fig. S4B) and induced in several tumor cell lines by the 5-Aza-dC treatment (Fig. S5A–C).

We tested whether miR-132/miR-212 would have an effect on SIRT1-mediated control of p53 function, such as the induction of cellular apoptosis upon DNA damage. A549 cells were first transfected with the precursors of miR-132, miR-212, hSirt1-specific siRNA, or a negative control RNA, and then exposed to the genotoxic compound etoposide. A strong decrease in the level of SIRT1 protein was consistently observed in samples transfected with miR-132, miR-212 or hSirt1-specific siRNA but not in cells inoculated with an irrelevant control RNA (Fig. 3C; Fig. S5D). Following exposure to etoposide, a much higher acetylation of p53 at Lys382 in cells transfected with miR-132, miR-212, or siRNA against hSIRT1 was observed when compared with the negative control and to non-transfected samples (Fig. 3C). The cellular p53 level was identical in all samples examined. Consistent with increased acetylation of p53, a p53-regulated gene, PUMA, was more efficiently induced in cells transfected with miR-132, miR-212, or hSIRT1 siRNA than control and non-treated cells (Fig. S5E). This suggests that p53 activity, induced by DNA damage, is probably augmented due to the increase in p53 acetylation caused by miR-132/miR-212-mediated repression of SIRT1. This conclusion is also supported by a direct measurement of p53 transcriptional activity with a reporter assay (Fig. S5F). Furthermore, increased p53 activity caused by miR-132/miR-212 during the DNA-damage response was also accompanied by elevated apoptosis-related signals such as Caspase3/7 activity and PARP cleavage (Fig. 3D and E). Indeed, cell death induced by etoposide treatment was significantly accelerated by the introduction of miR-132, miR-212 or SIRT1 siRNA (Fig. 3F; Fig. S5G), probably due to increased p53 activity with its enhanced acetylation.

Taken together, these data demonstrate a potential role of the miRNA group in the repression of SIRT1 during p53-dependent apoptosis. Aberrant expression of the Sirt1-targeting miRNA group due to DNA methylation would cause abnormal deregulation of p53-dependent events and might be involved in tumorigenesis.

Conclusions and Discussion

Unlike most previous studies focusing on a single miRNA for one target, our current study implicates group-wide regulation of miRNAs sharing a common key target. Thus, future studies exploring possible miRNA regulation of critical genes should aim to identify multiple miRNAs as a group targeting the desired genes and profile the group-wide regulatory mechanisms. Our finding showing that most of the Sirt1-targeting miRNA genes are sensitive to epigenetic silencing by DNA methylation, provides a mechanistic link between SIRT1 overexpression in many types of cancer and aberrant DNA methylation as a common feature of mammalian tumors (Fig. S6). Our screening enabled to identify the miRNAs which have potential to strongly-regulate SIRT1 among several miRNAs have been shown to regulate SIRT1.²⁵ p53-dependent miRNA, miR-34a has been also shown to regulate SIRT1,²⁵ but we confirmed that not mouse but human miR-34a regulates SIRT1 3′UTR (see Supplemental Materials). miR-34 family is also methylated in many types of cancer,²⁶ indicating DNA methylation of miR-34 family might facilitate SIRT1 overexpression in human cancer. Since other genes such as known oncogenes could be regulated by similar mechanisms in tumor cells, developing experimental models for studying miRNA group-wide regulatory mechanisms would be an interesting challenge in cancer miRNA genomics.

Together with the notion that the Sirt1-targeting miRNAs show a housekeeping gene-like expression pattern (Fig. 2B), the presence of multiple Sirt1-targeting miRNAs indicates that Sirt1 mRNA is under constitutive repressive control by those miRNAs in a healthy cell. Considering the presence of conserved miRNA target sites in the 3′UTR of Sirt1 mRNA (Fig. 1F), this “default control” of Sirt1 mRNA has been established during evolution for general physiological consequences antagonistic to SIRT1 function, such as deacetylation of p53.

In this study, based on the idea that the Sirt1-targeting miRNAs are silenced in cancer cells, we could also demonstrate that reintroducing one of the Sirt1-targeting miRNAs can lead to the efficient downregulation of SIRT1 and the rescue of the expected biological actions antagonized by SIRT1 in cancer cells. Here, we could reverse the resistant phenotype of A549 lung carcinoma cells against a genotoxic drug, etoposide, through reactivation of p53 by cancelling the repressive function of SIRT1. Due to multifaceted roles of SIRT1 in cancer cells, similar therapeutic approaches could be possible. For example, the reversal of drug resistance by artificial administration of a miRNA represents a potential application of miRNAs as a molecular adjuvant for various cancer chemotherapies.

Materials and Methods

Reagents, synthetic miRNAs, and siRNA

Etoposide was purchased from ALEXIS. All-trans retinoic acid and 5-Aza-2′-deoxycytidine was obtained from Sigma. FirstChoice→ Human Total RNA Survey Panel was purchased from Applied Biosystems-Ambion. The Pre-miR™ miRNA Precursors of hsa-miR-132, hsa-miR-212, hsa-miR-1 and the negative control #1 were all supplied by Applied Biosystems-Ambion. The siRNA against SIRT1 and Dicer was synthesized by MWG biotech and RNAi Inc. in each.

Cell culture conditions and DNA constructs

Conditions of cell cultures and DNA constructs are described in the Supplemental Information.

Establishment of stable cell lines

HeLa cells were seeded in 6-well plate at 2 ×10⁵ cells. After overnight culture, pmirGLO mouse SIRT1 3′UTR, pmirGLO human SIRT1 3′UTR or pmirGLO Control Reporter vector were transfected into cells by using Lipofectamine 2000 (Invitrogen) according to the manufacture’s protocol. After 2 d in culture, cells were replated in 10 cm dishes in fresh culture medium containing 500 μg/ml G418 and incubated until colonies arose. More than 80 colonies were picked from each transfectant and separately replated into 96-well plates. The clones expressing high levels of Firefly and Renilla and exhibiting normal proliferation and shape in comparison with HeLa cells were selected as stable transfectants and used for the library-based screening experiment.

Library-based screening of miRNAs regulating Sirt1 3′UTR

Pre-miR™ miRNA Precursor Library – Human Ver.3 and Mouse Ver.3 was obtained from Applied Biosystems-Ambion. For the screening assay, HeLa cells (1 × 10⁴/well in 100 μl) were seeded in 96-well plates. After overnight culture, cells were transfected with each individual species of miRNA precursor (30 nM) from the library using Lipofectamine RNAiMAX transfection reagent (Invitrogen) and the reverse transfection method according to the manufacturer’s recommendation. Following 36 h of cultivation, cells were subjected to a dual luciferase assay (Promega).

Bioinformatics

For target site prediction, 1810bp and 1588bp of human (uc001jnd.1, corresponding RefSeq Accession: NM_012238) and mouse (uc007fke.1, corresponding RefSseq Accession: NM_019812) SIRT1 3′UTR sequences including poly-A sequences, respectively, were obtained from UCSC genome browser (http://genome.ucsc.edu/). Human and mouse mature miRNA sequences were downloaded from miRBase (version 17) (http://microrna.sanger.ac.uk/sequences/ftp.shtml). Target site prediction of each miRNA was conducted with a minimum 6-nt seed match at 1–7, 2–8 or 2–7 positions from the 5′ end of miRNAs. For the 12 Sirt1-targeting miRNAs identified by the library-based screening, we conducted (1) the target site mapping using target site prediction described above and (2) comparison of the predicted target sites between human and mouse to find target sites with an evolutionarily conserved structure of miRNA-mRNA pairing and relative positioning in 3′UTRs. Finally, nine miRNAs were defined as authentic miRNA species targeting SIRT1. The CpG island in the miRNA loci was predicted using Methprimer (http://www.urogene.org/methprimer/index1.html). miRNA expression data was obtained from miRNAMap (http://mirnamap.mbc.nctu.edu.tw/). Heat maps were produced with Java TreeView software (http://jtreeview.sourceforge.net/).

Western blotting and quantitative real-time RT-PCR

Western blotting was performed as described previously. A rabbit monoclonal antibody against human SIRT1 (ab32441) was obtained from Abcam. The antibodies against p53 (#9282), acetyl-p53 (at Lys382: #2525), PARP (#9542), and β-actin (#4967) were all purchased from Cell Signaling Technology. The antibody against Dicer (ab13502) was purchased from Abcam. Conditions of Real-time RT-PCR and the primer sequences are described in the Supplemental Materials.

Methylation specific PCR (MSP) and bisulfite sequencing

Genomic DNA was isolated from cells using DNeasy@ Blood and Tissue Kit (Qiagen). Bisulfite conversion of genomic DNA was performed using EpiTect@ Bisulfite Kit (Qiagen) according to manufacturer’s recommendation. For MSP analyses, bisufite DNA samples were subjected to PCR with the specific primer pairs amplifying either methylated or unmethylated target sequences (the primers are listed in Table S1). For the detection of HIC1 (both methylated [M] and unmethylated [U] status) and unmethylated mir-212/mir-132 genes, AmpliTaq Gold@ (Applied Biosystems) was used for the amplification with the conditions for HIC1 (M&U: 95°C-10’, 30 cycles of 95°C-30” > 60°C-30,” 72°C-10’) and for unmethylated mir-212/mir-132 (95°C-10’, 37 cycles of 95°C -30” > 52°C -30” > 72°C -30,” 72°C -10’). The methylated sites for the mir-212/mir-132 gene were amplified by GoTaq polymerase (Promega) with the specific reaction parameter (95°C-2’, 37 cycles of 95°C-1’ > 55°C-30” > 72°C-3′, 72°C -10’). For sequencing, bisulfite-modified DNA was subjected to the PCR with the specific primers (Table S1) using AmpliTaq Gold@ with the specific condition (95°C -10’, 37 cycles of 95°C -30” > 50°C -30” > 72°C -30,” 72°C -10’). Resultant PCR products were purified, cloned into a TA cloning vector, pGEM-T easy, and sequenced.

Caspase 3/7 and Cell viability assays

A549 cells were reverse transfected with the indicated synthetic RNAs in 96-well clear-bottom plates (1 × 10⁴ cells /well in 100μl) using Lipofectamine RNAi MAX. After 48 h of incubation, cells were treated either by etoposide (1μM) or by DMSO (vehicle). At the indicated time points, samples were subjected to the Caspase 3/7 and cell viability assays using Apo-ONE Homogeneous Caspase-3/7 Assay (Pomega) and CellTiter-Glo Luminescent Cell Viability Assay (Pomega), respectively, according to the manufacturer’s instructions.

Statistics

The experimental values were statistically examined by the two-tailed Student’s t test. The number of samples (n) and actual P values are presented in the figures or legends.

Disclosure of Potential Conflicts of Interest

No potential conflicts of interest were disclosed.

Acknowledgments

We thank P. Nielsen for fruitful discussion.

Funding

This work was supported by funds from Max-Planck Gesellschaft and DFG SFB 620, (to T.F.).

Supplemental Materials

Supplemental data for this article can be accessed on the publisher's website.