The Parkinson's disease–linked proteins Fbxo7 and Parkin interact to mediate mitophagy

Journal name:
Nature Neuroscience
Volume:
16,
Pages:
1257–1265
Year published:
DOI:
doi:10.1038/nn.3489
Received
Accepted
Published online

Abstract

Compelling evidence indicates that two autosomal recessive Parkinson's disease genes, PINK1 (PARK6) and Parkin (PARK2), cooperate to mediate the autophagic clearance of damaged mitochondria (mitophagy). Mutations in the F-box domain–containing protein Fbxo7 (encoded by PARK15) also cause early-onset autosomal recessive Parkinson's disease, by an unknown mechanism. Here we show that Fbxo7 participates in mitochondrial maintenance through direct interaction with PINK1 and Parkin and acts in Parkin-mediated mitophagy. Cells with reduced Fbxo7 expression showed deficiencies in translocation of Parkin to mitochondria, ubiquitination of mitofusin 1 and mitophagy. In Drosophila, ectopic overexpression of Fbxo7 rescued loss of Parkin, supporting a functional relationship between the two proteins. Parkinson's disease–causing mutations in Fbxo7 interfered with this process, emphasizing the importance of mitochondrial dysfunction in Parkinson's disease pathogenesis.

At a glance

Figures

  1. The N-terminal Ubl domain of Fbxo7 interacts directly with Parkin.
    Figure 1: The N-terminal Ubl domain of Fbxo7 interacts directly with Parkin.

    (a) Diagram of Fbxo7 isoforms 1 and 2, showing the location of functional domains and the Parkinson's disease–associated mutations. (b) Coimmuno-precipitation (IP) of Fbxo7-HA and Flag-Parkin in whole-cell lysates from U2OS cells over-expressing both proteins. IB, immunoblot. Molecular mass markers are in kilodaltons. (c) Coimmunoprecipitation of endogenous Fbxo7 with Flag-Parkin in HEK293T cells transfected with Flag-Parkin or a control protein (EGFP). (d) Flag immunoprecipitation was repeated in U2OS cells transfected with Flag-Parkin and either full-length (1–522) T7-Fbxo7 or an N-terminal truncation lacking the Ubl domain (89–522). (e) As in d, using lysates from U2OS cells expressing Flag-Parkin and T7-Fbxo7, either full length (1–522) or a C-terminal deletion of the proline-rich region (1–398). * indicates IgG heavy chain. (f) In vitro–translated Flag-Parkin was incubated with bacterially expressed GST or GST fused to the Fbxo7 Ubl domain (1–88) immobilized on glutathione beads. Bead-bound proteins and inputs were analyzed by immunoblotting with anti-Parkin antibodies. (g) The disease-causing mutation T22M interferes with Fbxo7's interaction with Parkin. Coimmunoprecipitation was performed as in b using lysates from U2OS cells expressing Flag-Parkin and WT or T22M Fbxo7-HA. (h) Coimmunoprecipitation of Flag-Parkin and Fbxo7 in the mitochondrial and cytosolic fractions of HEK293T cells overexpressing both proteins. CxVα and GAPDH are markers for mitochondria and cytoplasm, respectively. All western blots are representative of experiments performed at least three times. Full-length blots are presented in Supplementary Figure 9.

  2. Fbxo7 participates in CCCP-induced accumulation of Parkin at the mitochondria.
    Figure 2: Fbxo7 participates in CCCP-induced accumulation of Parkin at the mitochondria.

    (a) Fbxo7 relocates from the cytosolic to the mitochondrial fractions of HEK293T cells treated with CCCP (10 μM). IB, immunoblot. (b) Fbxo7 levels are increased in Flag-Parkin complexes immunoprecipitated (IP) from the mitochondrial fraction of HEK293T cells transfected with Flag-Parkin and untagged Fbxo7 following 1 or 3 h treatment with CCCP (10 μM). (c) Parkin localization at the mitochondria was assessed by immunocytochemistry in SH-SY5Y cells transfected with Flag-Parkin plus scrambled (scr) or Fbxo7 siRNA, following 1 or 3 h treatment with CCCP (10 μM). Cells were scored visually for the colocalization of Flag-Parkin with HtrA2, a mitochondrial marker. Images are displayed for cells transfected as indicated, following 0 or 3 h CCCP treatment. For corresponding images at 1 h treatment, see Supplementary Figure 2c. Nuclei (blue) were stained with DAPI. Scale bars, 10 μm. (d) Loss of Flag-Parkin translocation upon Fbxo7 silencing is rescued by WT and R378G Fbxo7, but not by T22M Fbxo7, by R498X Fbxo7 or by Fbxo7 in which the mitochondrial targeting sequence is mutated (mt-MTS). For c,d, histograms indicate the percentage of cells in which Parkin localized to the mitochondria. Data are presented as mean of three experiments ± s.e.m. *P < 0.05, **P < 0.01 compared to cells transfected with Flag-Parkin plus Fbxo7 siRNA. (e) T7-tagged WT Fbxo7 localizes to both cytosolic (C) and mitochondrial (M) fractions of transfected HEK293T cells, but MTS mutant (mt-MTS) Fbxo7 localizes only to the cytosolic fraction. CxVα and PDHE1α are mitochondrial markers; GAPDH is a cytosolic marker. All western blots are representative of experiments performed at least three times. Full-length blots are presented in Supplementary Figure 9.

  3. Expression of Fbxo7 rescues parkin mutant phenotypes.
    Figure 3: Expression of Fbxo7 rescues parkin mutant phenotypes.

    (a,b) Overexpression of Fbxo7 by da-GAL4 (da>Fbxo7) suppresses climbing (a) and flight (b) defects of parkin mutants. (c) Overexpression of Fbxo7 also suppresses dopaminergic (DA) neurodegeneration parkin mutants. (df) Top and middle panels, toluidine blue–stained sections of adult thorax. Bottom panels, transmission electron micrographs of muscle. Fbxo7 overexpression suppresses muscle degeneration and mitochondrial disruption in parkin mutants. Scale bars: 200 μm (top), 20 μm (middle), 2 μm (bottom). Images are representative of three animals per genotype. (g,h) Overexpression of Fbxo7 pathogenic mutants, mt-MTS or isoform 2 by da-GAL4 fails to rescue climbing (g) and flight (h) deficits in parkin mutants. Control genotype is park25/+; da-GAL4/+. All transgenic lines are site-directed integration except WT(4), which is random integration line 4 (see Supplementary Fig. 3 and Online Methods). Histograms indicate mean ± s.e.m. One-way ANOVA with Bonferroni correction (***P < 0.001, **P < 0.01). For climbing and flight assays, at least 50 flies were assessed; number of flies is shown for each bar.

  4. PINK1 interacts directly with the amino terminus of Fbxo7.
    Figure 4: PINK1 interacts directly with the amino terminus of Fbxo7.

    (a) Coimmunoprecipitation (IP) of PINK1-Myc and Flag-Fbxo7 in whole-cell lysates from U2OS cells overexpressing both proteins. IB, immunoblot. (b) Coimmunoprecipitation of PINK1-Myc with full-length and two N-terminally deleted Flag-Fbxo7 forms. PINK1-Myc is detected at low levels in complex with Flag-Fbxo7(89–522) but not with Flag-Fbxo7(129–522). (c) As in b, using lysates from U2OS cells expressing PINK1-Myc, and Flag-Fbxo7 containing either N- or C-terminal truncations. (d) In vitro coprecipitation experiments were performed using in vitro–translated (IVT) PINK1-Myc and either GST or GST fusions of Fbxo7(1–398) or Fbxo7(129–398) immobilized on glutathione beads. (e) Competitive binding assays using immobilized GST-Fbxo7(1–398) incubated with IVT Flag-Parkin and/or full length (top panel) or N-terminally truncated (bottom panel) PINK1-Myc. Input and bead-bound proteins were analyzed by immunoblotting as indicated. (f) As in e, immobilized GST-ΔN-PINK1 (top panel), GST-Parkin (middle panel) and immobilized GST alone (bottom panel) were incubated with combinations of IVT Flag-Parkin, T7-ΔN-PINK1 and Fbxo7-HA as indicated. Input and bead-bound proteins were analyzed by immunoblotting with the indicated antibodies. All western blots were performed a minimum of three times. Full-length blots are presented in Supplementary Figure 9.

  5. Functional interaction of Fbxo7 with PINK1.
    Figure 5: Functional interaction of Fbxo7 with PINK1.

    (a) PINK1 localization at the mitochondria was assessed by immunocytochemistry in SH-SY5Y cells transfected with PINK1-HA plus scrambled (scr) or Fbxo7 siRNA following 1 or 3 h treatment with CCCP (10 μM). Cells were scored visually for the colocalization of PINK1-HA with complex Vβ subunit (CxVβ), a mitochondrial marker. Nuclei (blue) are stained with DAPI. Histograms indicate the percentage of cells in which PINK1-HA accumulated at the mitochondria. Data are presented as mean ± s.e.m.; *P < 0.05. Representative images are displayed for cells transfected as indicated, following 0 or 3 h CCCP treatment. For corresponding images at 0 h and 1 h treatment, see Supplementary Figure 5a. Scale bars, 10 μm. (b) Fbxo7 accumulation in the mitochondrial fraction following treatment with CCCP (10 μM) is impaired in SH-SY5Y cells transfected with PINK1 siRNA compared to scrambled siRNA (scr). IB, immunoblot. Full-length blots are presented in Supplementary Figure 9. (cf) Overexpression of Fbxo7 does not rescue climbing (c,e) or flight (d,f) defects in PINK1 male mutants (PINK1B9) or PINK1:parkin double mutants (PINK1B9;park25, daG4). Control genotypes are PINK1B9/+; da-GAL4/+ (c,d) and da-GAL4/+ (e,f). Histograms indicate mean ± s.e.m. One-way ANOVA with Bonferroni correction (***P < 0.001). For climbing and flight assays, at least 50 flies were assessed; number of flies is shown for each bar for cf.

  6. Fbxo7 promotes Mfn1 ubiquitination and restores Mfn levels and mitochondrial morphology in Parkin- but not PINK1-deficient cells.
    Figure 6: Fbxo7 promotes Mfn1 ubiquitination and restores Mfn levels and mitochondrial morphology in Parkin- but not PINK1-deficient cells.

    (a,b) Ubiquitination of Mfn1 following treatment with CCCP (10 μM) is reduced in the mitochondrial fraction of both SH-SY5Y cells stably expressing Fbxo7 short hairpin RNA (Fbxo7 KD) compared to an empty vector control line (a) and in patient fibroblasts with homozygous R378G mutation compared to fibroblasts from healthy controls (b). Arrowheads indicate ubiquitinated Mfn1. IB, immunoblot. (c,d) Fbxo7 expression restores elevated Mfn steady-state levels in parkin (c) but not PINK1 (d) mutant Drosophila. Histograms show mean ± s.e.m. of densitometry analysis of Mfn immunoblots above, normalized to complex Vα (CxVα). Control genotypes are park25/+; da-GAL4/+ (c) and PINK1B9/+; da-GAL4/+ (d). (e) Mitochondria in control Drosophila S2R+ cells stained with MitoTracker Red show a heterogeneous morphology, with a mixture of tubules and fragmented mitochondria. RNAi knockdown of parkin or PINK1 causes excessive fusion and elongated mitochondria compared to control double-stranded RNA (Caenorhabditis elegans gene ZK686.3). Expression of Fbxo7 restores parkin but not PINK1 knockdown phenotype to WT appearance. Scale bar shows 5 μm. (f) Quantification of mitochondrial morphology in dsRNA treated cells. Scoring system: 1, fragmented; 2, WT; 3, tubular; 4, hyper-fused (clumped). Histograms indicate mean ± s.e.m. Two-tailed Student t-tests (***P < 0.001, *P < 0.05). All western blots were performed a minimum of three times and images are representative of 100 cells scored per condition. Full-length blots are presented in Supplementary Figure 9.

  7. Fbxo7 is important for mitophagy.
    Figure 7: Fbxo7 is important for mitophagy.

    (a) Treatment with CCCP (10 μM) results in an increase in LC3-II in the mitochondrial but not the cytosolic fraction of cells stably expressing the empty shRNA vector (control), and this is delayed in stable Fbxo7 knockdown (KD) SH-SY5Y cells. (b) As in a, an accumulation of LC3-II was observed in the mitochondrial fraction of healthy control fibroblasts following 1 and 5 h CCCP treatment, but this was reduced in fibroblasts from a patient carrying the R378G mutation. Western blots were performed a minimum of three times. (c) Mitochondrial accumulation of p62 following CCCP treatment is inhibited by Fbxo7 siRNA. Flag-Parkin overexpressing SH-SY5Y cells were transfected with scrambled (scr) or Fbxo7 siRNA as indicated and treated with either DMSO or CCCP (10 μM) for 6 h. Colocalization of p62 with HtrA2, a mitochondrial marker, was assessed by Pearson's correlation coefficient (Rr) on a cell-by-cell basis. Histogram shows the percentage of cells in which Rr was greater than 0.5. Data are represented as mean ± s.e.m., *P < 0.05. Scale bars, 10 μm. (d) Mitophagy was analyzed in untransfected (UT) SH-SY5Y cells or in stable Flag-Parkin overexpressing SH-SY5Y cells transfected with either scrambled (scr) or Fbxo7 siRNA. Histogram indicates the percentage of cells with no remaining mitochondria following 24 h treatment with CCCP (10 μM) for each condition. Complex Vβ subunit (CxVβ) was used as a mitochondrial marker. Data are presented as mean ± s.e.m., **P < 0.01. Scale bars, 10 μm. (e) Mitochondrial mass was measured in Flag-Parkin overexpressing SH-SY5Y cells transfected with scrambled (scr) or Fbxo7 siRNA and treated for 24 h with either dimethylsulfoxide (DMSO) vehicle or CCCP (10 μM). For representative images, see Supplementary Figure 8d. Full-length blots are presented in Supplementary Figure 9.

References

  1. Narendra, D., Tanaka, A., Suen, D.F. & Youle, R.J. Parkin is recruited selectively to impaired mitochondria and promotes their autophagy. J. Cell Biol. 183, 795803 (2008).
  2. Vives-Bauza, C. et al. PINK1-dependent recruitment of Parkin to mitochondria in mitophagy. Proc. Natl. Acad. Sci. USA 107, 378383 (2010).
  3. Narendra, D.P. et al. PINK1 is selectively stabilized on impaired mitochondria to activate Parkin. PLoS Biol. 8, e1000298 (2010).
  4. Geisler, S. et al. PINK1/Parkin-mediated mitophagy is dependent on VDAC1 and p62/SQSTM1. Nat. Cell Biol. 12, 119131 (2010).
  5. Deas, E., Wood, N.W. & Plun-Favreau, H. Mitophagy and Parkinson's disease: the PINK1-parkin link. Biochim. Biophys. Acta 1813, 623633 (2011).
  6. Jin, S.M. et al. Mitochondrial membrane potential regulates PINK1 import and proteolytic destabilization by PARL. J. Cell Biol. 191, 933942 (2010).
  7. Whitworth, A.J. et al. Rhomboid-7 and HtrA2/Omi act in a common pathway with the Parkinson's disease factors Pink1 and Parkin. Dis. Model. Mech. 1, 168174 discussion 173 (2008).
  8. Ziviani, E., Tao, R.N. & Whitworth, A.J. Drosophila parkin requires PINK1 for mitochondrial translocation and ubiquitinates mitofusin. Proc. Natl. Acad. Sci. USA 107, 50185023 (2010).
  9. Poole, A.C., Thomas, R.E., Yu, S., Vincow, E.S. & Pallanck, L. The mitochondrial fusion-promoting factor mitofusin is a substrate of the PINK1/parkin pathway. PLoS ONE 5, e10054 (2010).
  10. Gegg, M.E. et al. Mitofusin 1 and mitofusin 2 are ubiquitinated in a PINK1/parkin-dependent manner upon induction of mitophagy. Hum. Mol. Genet. 19, 48614870 (2010).
  11. Tanaka, A. Parkin-mediated selective mitochondrial autophagy, mitophagy: Parkin purges damaged organelles from the vital mitochondrial network. FEBS Lett. 584, 13861392 (2010).
  12. Rakovic, A. et al. Mutations in PINK1 and Parkin impair ubiquitination of Mitofusins in human fibroblasts. PLoS ONE 6, e16746 (2011).
  13. Sarraf, S.A. et al. Landscape of the PARKIN-dependent ubiquitylome in response to mitochondrial depolarization. Nature 496, 372376 (2013).
  14. Tanaka, A. et al. Proteasome and p97 mediate mitophagy and degradation of mitofusins induced by Parkin. J. Cell Biol. 191, 13671380 (2010).
  15. Shojaee, S. et al. Genome-wide linkage analysis of a Parkinsonian-pyramidal syndrome pedigree by 500 K SNP arrays. Am. J. Hum. Genet. 82, 13751384 (2008).
  16. Di Fonzo, A. et al. FBXO7 mutations cause autosomal recessive, early-onset parkinsonian-pyramidal syndrome. Neurology 72, 240245 (2009).
  17. Paisán-Ruiz, C. et al. Early-onset L-dopa-responsive parkinsonism with pyramidal signs due to ATP13A2, PLA2G6, FBXO7 and spatacsin mutations. Mov. Disord. 25, 17911800 (2010).
  18. Skowyra, D., Craig, K.L., Tyers, M., Elledge, S.J. & Harper, J.W. F-box proteins are receptors that recruit phosphorylated substrates to the SCF ubiquitin-ligase complex. Cell 91, 209219 (1997).
  19. Hsu, J.M., Lee, Y.C., Yu, C.T. & Huang, C.Y. Fbx7 functions in the SCF complex regulating Cdk1-cyclin B-phosphorylated hepatoma up-regulated protein (HURP) proteolysis by a proline-rich region. J. Biol. Chem. 279, 3259232602 (2004).
  20. Laman, H. et al. Transforming activity of Fbxo7 is mediated specifically through regulation of cyclin D/cdk6. EMBO J. 24, 31043116 (2005).
  21. Kuiken, H.J. et al. Identification of F-box only protein 7 as a negative regulator of NF-κB signalling. J. Cell Mol. Med. 16, 21402149 (2012).
  22. Kirk, R. et al. Structure of a conserved dimerization domain within the F-box protein Fbxo7 and the PI31 proteasome inhibitor. J. Biol. Chem. 283, 2232522335 (2008).
  23. Chang, Y.F., Cheng, C.M., Chang, L.K., Jong, Y.J. & Yuo, C.Y. The F-box protein Fbxo7 interacts with human inhibitor of apoptosis protein cIAP1 and promotes cIAP1 ubiquitination. Biochem. Biophys. Res. Commun. 342, 10221026 (2006).
  24. Greene, J.C. et al. Mitochondrial pathology and apoptotic muscle degeneration in Drosophila parkin mutants. Proc. Natl. Acad. Sci. USA 100, 40784083 (2003).
  25. Whitworth, A.J. et al. Increased glutathione S-transferase activity rescues dopaminergic neuron loss in a Drosophila model of Parkinson's disease. Proc. Natl. Acad. Sci. USA 102, 80248029 (2005).
  26. Park, J. et al. Mitochondrial dysfunction in Drosophila PINK1 mutants is complemented by parkin. Nature 441, 11571161 (2006).
  27. Clark, I.E. et al. Drosophila pink1 is required for mitochondrial function and interacts genetically with parkin. Nature 441, 11621166 (2006).
  28. Bader, M. et al. A conserved F box regulatory complex controls proteasome activity in Drosophila. Cell 145, 371382 (2011).
  29. Bader, M., Arama, E. & Steller, H. A novel F-box protein is required for caspase activation during cellular remodeling in Drosophila. Development 137, 16791688 (2010).
  30. Yang, Y. et al. Mitochondrial pathology and muscle and dopaminergic neuron degeneration caused by inactivation of Drosophila Pink1 is rescued by Parkin. Proc. Natl. Acad. Sci. USA 103, 1079310798 (2006).
  31. Exner, N. et al. Loss-of-function of human PINK1 results in mitochondrial pathology and can be rescued by parkin. J. Neurosci. 27, 1241312418 (2007).
  32. Song, S. et al. Characterization of PINK1 (PTEN-induced Putative Kinase 1) mutations associated with Parkinson disease in mammalian cells and Drosophila. J. Biol. Chem. 288, 56605672 (2013).
  33. Klionsky, D.J. et al. Guidelines for the use and interpretation of assays for monitoring autophagy in higher eukaryotes. Autophagy 4, 151175 (2008).
  34. Ding, W.X. et al. Nix is critical to two distinct phases of mitophagy, reactive oxygen species-mediated autophagy induction and Parkin-ubiquitin-p62-mediated mitochondrial priming. J. Biol. Chem. 285, 2787927890 (2010).
  35. Lee, S.J. et al. A functional role for the p62–ERK1 axis in the control of energy homeostasis and adipogenesis. EMBO Rep. 11, 226232 (2010).
  36. Klionsky, D.J. et al. Guidelines for the use and interpretation of assays for monitoring autophagy. Autophagy 8, 445544 (2012).
  37. Nelson, D.E. & Laman, H. A competitive binding mechanism between SKP1 and exportin 1 (CRM1) controls the localization of a subset of F-box proteins. J. Biol. Chem. 286, 1980419815 (2011).
  38. Kondapalli, C. et al. PINK1 is activated by mitochondrial membrane potential depolarization and stimulates Parkin E3 ligase activity by phosphorylating serine 65. Open Biol. 2, 120080 (2012).
  39. Nakai, K. & Kanehisa, M. A knowledge base for predicting protein localization sites in eukaryotic cells. Genomics 14, 897911 (1992).
  40. Claros, M.G. & Vincens, P. Computational method to predict mitochondrially imported proteins and their targeting sequences. Eur. J. Biochem. 241, 779786 (1996).
  41. Small, I., Peeters, N., Legeai, F. & Lurin, C. Predotar: a tool for rapidly screening proteomes for N-terminal targeting sequences. Proteomics 4, 15811590 (2004).
  42. Dickins, R.A. et al. Probing tumor phenotypes using stable and regulated synthetic microRNA precursors. Nat. Genet. 37, 12891295 (2005).
  43. Ardley, H.C. et al. Inhibition of proteasomal activity causes inclusion formation in neuronal and non-neuronal cells overexpressing Parkin. Mol. Biol. Cell 14, 45414556 (2003).
  44. Plun-Favreau, H. et al. The mitochondrial protease HtrA2 is regulated by Parkinson's disease–associated kinase PINK1. Nat. Cell Biol. 9, 12431252 (2007).
  45. Samali, A., Cai, J., Zhivotovsky, B., Jones, D.P. & Orrenius, S. Presence of a pre-apoptotic complex of pro-caspase-3, Hsp60 and Hsp10 in the mitochondrial fraction of Jurkat cells. EMBO J. 18, 20402048 (1999).
  46. Tain, L.S. et al. Drosophila HtrA2 is dispensable for apoptosis but acts downstream of PINK1 independently from Parkin. Cell Death Differ. 16, 11181125 (2009).
  47. Sullivan, W., Ashburner, M. & Hawley, R.S. Drosophila Protocols Ch.13, 240–241 (Cold Spring Harbor Laboratory Press, Cold Spring Harbor, New York, USA, 2000).

Download references

Author information

  1. These authors contributed equally to this work.

    • Victoria S Burchell,
    • David E Nelson &
    • Alvaro Sanchez-Martinez

Affiliations

  1. Department of Molecular Neuroscience, University College London Institute of Neurology, London, UK.

    • Victoria S Burchell,
    • Marta Delgado-Camprubi,
    • Selina Wray,
    • Patrick A Lewis,
    • Henry Houlden,
    • Andrey Y Abramov,
    • John Hardy,
    • Nicholas W Wood &
    • Helene Plun-Favreau
  2. Department of Pathology, University of Cambridge, Cambridge, UK.

    • David E Nelson,
    • Suzanne J Randle &
    • Heike Laman
  3. Medical Research Council Centre for Developmental and Biomedical Genetics, Sheffield, UK.

    • Alvaro Sanchez-Martinez,
    • Rachael M Ivatt,
    • Joe H Pogson &
    • Alexander J Whitworth
  4. Department of Biomedical Sciences, University of Sheffield, Sheffield, UK.

    • Alvaro Sanchez-Martinez,
    • Rachael M Ivatt,
    • Joe H Pogson &
    • Alexander J Whitworth

Contributions

H.P.-F., H.L. and A.J.W. conceived the study. V.S.B., D.E.N., A.S.-M., A.J.W., H.L. and H.P.-F. designed and performed experiments. M.D.-C., R.M.I., J.H.P. and S.J.R. performed experiments. H.H., P.A.L. and S.W. obtained and cultured the patient fibroblasts. P.A.L., A.Y.A., J.H. and N.W.W. contributed to the design of the study. V.S.B., D.E.N., A.S.-M., H.P.-F., A.J.W. and H.L. wrote the manuscript.

Competing financial interests

The authors declare no competing financial interests.

Corresponding authors

Correspondence to:

Author details

Supplementary information

PDF files

  1. Supplementary Text and Figures (10,178 KB)

    Supplementary Figures 1–9

Additional data