2017-He-Dynamics of phosphoinositide conversio(1).pdf

4 1 0|N A T U R E|V O L 5 5 2|2 1/2 8 D E C E M B E R 2 0 1 7LETTERdoi:10.1038/nature25146Dynamics of phosphoinositide conversion in clathrin-mediated endocytic trafficKangmin He1,2,3,Robert Marsland III4,Srigokul Upadhyayula1,2,3,Eli Song2,Song Dang2,Benjamin R.Capraro1,2,Weiming Wang1,2,Wesley Skillern2,Raphael Gaudin1,2,Minghe Ma2&Tom Kirchhausen1,2,3Vesicular carriers transport proteins and lipids from one organelle to another,recognizing specific identifiers for the donor and acceptor membranes.Two important identifiers are phosphoinositides and GTP-bound GTPases,which provide well-defined but mutable labels.Phosphatidylinositol and its phosphorylated derivatives are present on the cytosolic faces of most cellular membranes1,2.Reversible phosphorylation of its headgroup produces seven distinct phosphoinositides.In endocytic traffic,phosphatidylinositol-4,5-biphosphate marks the plasma membrane,and phosphatidylinositol-3-phosphate and phosphatidylinositol-4-phosphate mark distinct endosomal compartments2,3.It is unknown what sequence of changes in lipid content confers on the vesicles their distinct identity at each intermediate step.Here we describe coincidence-detecting sensors that selectively report the phosphoinositide composition of clathrin-associated structures,and the use of these sensors to follow the dynamics of phosphoinositide conversion during endocytosis.The membrane of an assembling coated pit,in equilibrium with the surrounding plasma membrane,contains phosphatidylinositol-4,5-biphosphate and a smaller amount of phosphatidylinositol-4-phosphate.Closure of the vesicle interrupts free exchange with the plasma membrane.A substantial burst of phosphatidylinositol-4-phosphate immediately after budding coincides with a burst of phosphatidylinositol-3-phosphate,distinct from any later encounter with the phosphatidylinositol-3-phosphate pool in early endosomes;phosphatidylinositol-3,4-biphosphate and the GTPase Rab5 then appear and remain as the uncoating vesicles mature into Rab5-positive endocytic intermediates.Our observations show that a cascade of molecular conversions,made possible by the separation of a vesicle from its parent membrane,can label membrane-traffic intermediates and determine their destinations.To design the new sensors,we capitalized on the way in which auxilin and epsin associate with clathrin coats46.Auxilins(in mammalian cells,auxilin1(Aux1)and auxilin2 or GAK)require both a clathrin-binding domain and a phosphatase and tensin homologue(PTEN)-like domain for effective recruitment to newly budded clathrin-coated vesicles4,7,8(Extended Data Fig.1a,Supplementary Video 1).Binding to clathrin depends on the geometry of the clathrin lattice9,and neither domain is effective on its own at normal intracellular concentrations.Epsin also has both clathrin-binding and lipid-binding domains5,6.We proposed previously that the auxilin PTEN-like domain interacts with a specific phosphoinositide in the coat-engulfed membrane and that auxilins are effectively coincidence detectors4.We have therefore prepared a series of sensors(Extended Data Figs 14)in which a phosphoinositide-binding domain of known specificity is combined with the Aux1(Fig.1a,Extended Data Fig.1j)or epsin1(Extended Data Fig.4e)clathrin-binding domain and an enhanced GFP(EGFP)or mCherry fluorophore.We validated their properties as described in Extended Data Figs 14 and the Supplementary Discussion.In most experiments,we followed recruitment of these sensors in gene-edited SUM159 cells expressing the clathrin light chain A joined to the fluorescent marker TagRFP(CLTATagRFP)(Extended Data Fig.1b).Cells were imaged by total internal reflection fluorescence(TIRF)microscopy,with illumination at an angle chosen to decrease sensitivity to sample depth but to increase sensitivity with respect to spinning-disk confocal fluorescence microscopy.We used a previously developed 2D-tracking computational framework for automated detec-tion and tracking of the fluorescently tagged coated structures10.To follow the sensors on internal membranes,we used a lattice light-sheet microscope11 to visualize the full cellular volume.The phosphatidylinositol-4,5-biphosphate(PtdIns(4,5)P2)fluo-rescent sensor EGFPPH(PLC 1)-Aux1 was present in all plasma membrane coated pits of CLTATagRFP+/+cells(Fig.1b,Extended Data Fig.4a,Supplementary Video 2).Unlike intact Aux1,which appears in a burst immediately after scission of a coated vesicle from the plasma membrane(Extended Data Fig.1a,Supplementary Video 1),the sensor accumulated in clathrin-coated pits as they formed,followed by a gradual loss coinciding with disassembly of the clathrin coat(Figs 1b,2a).The sensor did not associate with any clathrin-coated structures in endosomal membranes or in the trans-Golgi network(Extended Data Fig.2a),both of which lack substantial concentrations of PtdIns(4,5)P22,3.As a control for the phosphoinositide-binding spec-ificity of the PtdIns(4,5)P2 sensor,we showed that it failed to appear in coated pits if we introduced point mutations into the PH(PLC 1)domain that are known to prevent PtdIns(4,5)P2 binding12,13(Figs 1b,2a).Moreover,EGFPPH(PLC 1)alone accumulated throughout the plasma membrane and not in fluorescence-enriched spots that colocalized with clathrin-coated pits(Figs 1b,2a,Supplementary Video 2).In the plasma membrane,PtdIns(4,5)P2 is essential for initiating and sustaining assembly of endocytic coated pits1416.Depletion of plasma membrane PtdIns(4,5)P2 by rapid,light-activated transfer of the inositol 5-phosphatase module of inositol polyphosphate 5-phosphatase OCRL from the cytosol to the plasma membrane14 prevented new AP2 adaptor complex fluorescent spots from appearing(no initiation of endocytic coated pits)and stalled those already present(no matura-tion of pits)(Extended Data Fig.2b).The PtdIns(4,5)P2 sensor was not recruited to the stalled pitsa stringent test of its lipid specificity(Extended Data Fig.2c).The phosphatidylinositol-3-phosphate(PtdIns3P)sensor EGFP2 FYVE(Hrs)-Aux1 appeared in a burst that coincided with clathrin coat disassembly(Fig.2b,Supplementary Video 3).The Aux1 clathrin-binding domain ensured specificity of the sensor for clathrin-containing structures(Fig.2b).There was no sensor recruitment for a 1Department of Cell Biology,Harvard Medical School,200 Longwood Ave,Boston,Massachusetts 02115,USA.2Program in Cellular and Molecular Medicine,Boston Childrens Hospital,200 Longwood Ave,Boston,Massachusetts 02115,USA.3Department of Pediatrics,Harvard Medical School,200 Longwood Ave,Boston,Massachusetts 02115,USA.4Physics of Living Systems Group,Massachusetts Institute of Technology,400 Technology Square,Cambridge,Massachusetts 02139,USA.Present addresses:Department of Physics,Boston University,590 Commonwealth Ave,Boston,Massachusetts 02215,USA(R.M.);National Laboratory of Biomacromolecules,CAS Center for Excellence in Biomacromolecules,Institute of Biophysics,Chinese Academy of Sciences,Beijing 100101,China(E.S.);Institute of Viral and Liver DiseaseINSERM U1110,3 rue Koeberl,Strasbourg 67000,France(R.G.).2017 Macmillan Publishers Limited,part of Springer Nature.All rights reserved.Letter reSeArCH2 1/2 8 D E C E M B E R 2 0 1 7|V O L 5 5 2|N A T U R E|4 1 1disabled variant with mutations in the PtdIns3P binding site17(Fig.2b,Extended Data Fig.4a).The PtdIns3P sensor recruitment mimicked the normal pattern of Aux1 association4,7:absent from assembling coated pits and appearing in a burst immediately after dynamin-catalysed release of coated vesicles(Extended Data Fig.1a).The phosphatidylinositol-4-phosphate(PtdIns4P)sensor EGFPP4M(DrrA)-Aux1 accumulated at a low,steady rate along with clathrin and AP2,and then,like native Aux14,7,appeared in an acute burst just after budding(Fig.2c,Extended Data Fig.2eg,Supplementary Video 4).The burst required membrane scission,as it was absent from the stalled(that is,persistent)coated pits in cells treated with the small-molecule dynamin inhibitor dynasore-OH or depleted of dynamin2 by small interfering RNA(siRNA)(Extended Data Fig.2e,f).EGFPP4M(DrrA)18,which lacked the clathrin-binding region,labelled the plasma membrane diffusely(Extended Data Fig.4a)as well as the Golgi apparatus and late endosomes/lysosomes.We con-firmed the specificity of PtdIns4P binding by mutation of PtdIns4P binding residues19(Fig.2c).In cells subjected to acute light-activated depletion of PtdIns(4,5)P2,PtdIns4P appeared in pits stalled by loss of PtdIns(4,5)P2,but the PtdIns4P burst accompanying uncoating did not occur(Extended Data Fig.2d).PtdIns4P bursts appeared after membrane scission in the coated vesicles that had formed at the onset of light-activated depletion,when PtdIns(4,5)P2 had not yet been sufficiently depleted.We enhanced the temporal precision of measurements between the onset of the loss of signal from the PtdIns(4,5)P2 sensor and the burst of signal from the PtdIns4P sensor by co-expressing the PtdIns(4,5)P2 and PtdIns4P sensors in the same cells(Extended Data Fig.2g).PtdIns(4,5)P2 had begun to disappear at the time of onset of the PtdIns4P burst,which immediately followed conversion of a coated pit into a coated vesicle.The PtdIns(3,4)P2 sensor,EGFP2 PH(TAPP1)-Aux1,appeared in coated vesicles but not in coated pits;the association remained even after uncoating had finished(Fig.2d,Supplementary Video 5).Depletion from coated pits of the proposed PtdIns(3,4)P2 effector SNX920 did not induce capture of the PtdIns(3,4)P2 sensor(Extended Data Fig.5a),showing that protection of PtdIns(3,4)P2 by SNX9 cannot account for sensor exclusion.Acute dynamin accumula-tion leads to membrane scission,transforming the coated pit into a vesicle and releasing it from the plasma membrane21,22.In gene-edited dynamin2EGFP+/+SUM159 cells21,recruitment of the PtdIns(3,4)P2 sensor began only after the burst accumulation of dynamin2EGFP was complete(Extended Data Fig.5b),consistent with absence of the PtdIns(3,4)P2 sensor from stalled coated pits in cells treated with dynasore-OH or depleted of dynamin2(Extended Data Fig.5a).The onset of recruitment of the PtdIns(3,4)P2 sensor coincided with vesicle release from the plasma membrane,and continued during the subsequent loss of the clathrin signal(Fig.2d,Extended Data Fig.5a,Supplementary Video 5).Unlike the burst recruitment of Aux1 or of the PtdIns3P and PtdIns4P sensors,the PtdIns(3,4)P2 sensor remained asso-ciated with the vesicular carrier even after uncoating had ended(Fig.2d,Extended Data Fig.5a).This late association of the PtdIns(3,4)P2 sensor with the uncoated vesicles was possible because a few clathrin molecules remained on the uncoated vesicle(Extended Data Fig.5c,d).As with the other sensors,we confirmed the specificity of the PtdIns(3,4)P2 sensor by showing that a version with mutations at the positions of PtdIns(3,4)P2-binding residues23 failed to appear in coated structures(Fig.2d).The presence in most cells of multiple enzymes for catalysing specific phosphoinositide interconversions suggests functional redundancy in generating the phosphoinositide dynamics just outlined.As described in more detail in Fig.3,Extended Data Figs 68 and the Supplementary Discussion,we used gene editing and partial knockdown(KD)with RNAi to test for potential roles for phosphatidylinositol 4-kinase type III(PI4KIII),phosphatidylinositol 4-phosphate 5-kinase type I(PIPKI),synaptojanin1(Synj1)and OCRL in the reactions that affect coated pits and their scission,and potential roles for the class II phosphatidylinositol 3-kinase C2(PI3K-C2),the class III phosphatidylinositol 3-kinase Vps34 and inositol polyphosphate-4-phosphatase type I(INPP4A)in generating PtdIns3P in coated vesicles.Our data indicate that PI4KIII and PIPKI generate PtdIns4P and a fraction of PtdIns(4,5)P2 in coated pits at the plasma membrane(Fig.3a,Extended Data Fig.6ad)and that both Synj1 and OCRLphosphatases that have been suggested to be part of the conversion cascade24,25are major sources of the compositional changes we detect in PtdIns(4,5)P2 and PtdIns4P(Fig.3b,c,Extended Data Fig.7h).These enzymes thus appear to have redundant functions in the cells we have used,even though their arrival times at the coated structures overlapped only partly:Synj1 recruitment began during coated pit formation and continued as uncoating proceeded(Extended Data Fig.6e,f),while OCRL recruitment began at the onset of uncoating(Extended Data Fig.7ad).We have also identified the kinase PI3K-C2 and the phosphatase INPP4A as the enzymes that generate PtdIns3P from phosphatidylinositol and PtdIns(3,4)P2 in coated vesicles(Fig.3d,Extended Data Fig.8).We have not yet identified the enzyme(s)that generate PtdIns(3,4)P2 in coated vesicles.Rab5 is an early endosome-specific small GTPase26.We made three gene-edited SUM159 cell lines:EGFPRab5a+/+,expressing EGFPRab5a at both alleles;EGFPRab5c+/+,expressing EGFPRab5c at both alleles;and EGFPRab5a+/+EGFPRab5c+/+,expressing both tagged proteins at all four alleles(Extended Data Fig.9a).In no case did we detect any EGFPRab5 molecules in coated pits or coated vesicles,even ab300 sPTEN-like domainEGFPClathrinbindingJdomainPhosphoinositide-binding domain Aux114208144208149100 sClathrinbindingCLTATagRFP+/+Ptdlns(4,5)P2 sensorsPhosphoinositide sensorEGFPPH(PLC1)EGFPPH(PLC1)-Aux1EGFPPH(PLC1)-mt-Aux1ClathrinPhosphoinositideMembraneFigure 1|Cellular localization of phosphoinositide-specific,auxilin1-based PtdIns(4,5)P2 sensors.a,Left,domain organization of mammalian Aux1 and of fluorescently tagged Aux1-based phosphoinositide sensors.Right,diagram of sensor-coat association.b,Localization of a general PtdIns(4,5)P2 sensor(EGFPPH(PLC 1),a mutated Aux1-based PtdIns(4,5)P2 sensor defective in binding PtdIns(4,5)P2(EGFP PH(PLC 1)-mt-Aux1),and a wild-type Aux1-based PtdIns(4,5)P2 sensor(EGFPPH(PLC 1)-Aux1).Top,distribution of PtdIns(4,5)P2 sensor at a single time point;middle,CLTATagRFP superposed on PtdIns(4,5)P2 sensor(green),including enlarged region(square box);bottom,corresponding kymographs from 300-s time series imaged every 2s by spinning-disk confocal microscopy.EGFP channel in the enlarged regions and kymographs shifted laterally by six pixels.Images are representative of at least three independent experiments.Scale bars,5 m.2017 Macmillan Publishers Limited,part of Springer Nature.All rights reserved.LetterreSeArCH4 1 2|N A T U R E|V O L 5 5 2|2 1/2 8 D E C E M B E R 2 0 1 7using TIRF with single-molecule sensitivity(Fig.4a,Extended Data Fig.9b,d,Supplementary Video 6).The onset of Rab5 recruitment coincided in a few instances with the very end of uncoating(Fig.4a,Extended Data Fig.9b,d),but the TIRF geometry did not allow us to follow most of the uncoated vesicles as they moved into the cell.We obtained similar results in gene-edited human SVGA cells27(Extended Data Fig.9e)and in SUM159 cells transiently expressing EGFPRab5a(Extended Data Fig.9f).As expected,we also found EGFPRab5a in small endosomal vesicl